1/* SPDX-License-Identifier: GPL-2.0 */
2/*
3 * Portions Copyright (C) 1992 Drew Eckhardt
4 */
5#ifndef _LINUX_BLKDEV_H
6#define _LINUX_BLKDEV_H
7
8#include <linux/types.h>
9#include <linux/blk_types.h>
10#include <linux/device.h>
11#include <linux/list.h>
12#include <linux/llist.h>
13#include <linux/minmax.h>
14#include <linux/timer.h>
15#include <linux/workqueue.h>
16#include <linux/wait.h>
17#include <linux/bio.h>
18#include <linux/gfp.h>
19#include <linux/kdev_t.h>
20#include <linux/rcupdate.h>
21#include <linux/percpu-refcount.h>
22#include <linux/blkzoned.h>
23#include <linux/sched.h>
24#include <linux/sbitmap.h>
25#include <linux/uuid.h>
26#include <linux/xarray.h>
27#include <linux/file.h>
28
29struct module;
30struct request_queue;
31struct elevator_queue;
32struct blk_trace;
33struct request;
34struct sg_io_hdr;
35struct blkcg_gq;
36struct blk_flush_queue;
37struct kiocb;
38struct pr_ops;
39struct rq_qos;
40struct blk_queue_stats;
41struct blk_stat_callback;
42struct blk_crypto_profile;
43
44extern const struct device_type disk_type;
45extern const struct device_type part_type;
46extern const struct class block_class;
47
48/*
49 * Maximum number of blkcg policies allowed to be registered concurrently.
50 * Defined here to simplify include dependency.
51 */
52#define BLKCG_MAX_POLS 6
53
54#define DISK_MAX_PARTS 256
55#define DISK_NAME_LEN 32
56
57#define PARTITION_META_INFO_VOLNAMELTH 64
58/*
59 * Enough for the string representation of any kind of UUID plus NULL.
60 * EFI UUID is 36 characters. MSDOS UUID is 11 characters.
61 */
62#define PARTITION_META_INFO_UUIDLTH (UUID_STRING_LEN + 1)
63
64struct partition_meta_info {
65 char uuid[PARTITION_META_INFO_UUIDLTH];
66 u8 volname[PARTITION_META_INFO_VOLNAMELTH];
67};
68
69/**
70 * DOC: genhd capability flags
71 *
72 * ``GENHD_FL_REMOVABLE``: indicates that the block device gives access to
73 * removable media. When set, the device remains present even when media is not
74 * inserted. Shall not be set for devices which are removed entirely when the
75 * media is removed.
76 *
77 * ``GENHD_FL_HIDDEN``: the block device is hidden; it doesn't produce events,
78 * doesn't appear in sysfs, and can't be opened from userspace or using
79 * blkdev_get*. Used for the underlying components of multipath devices.
80 *
81 * ``GENHD_FL_NO_PART``: partition support is disabled. The kernel will not
82 * scan for partitions from add_disk, and users can't add partitions manually.
83 *
84 */
85enum {
86 GENHD_FL_REMOVABLE = 1 << 0,
87 GENHD_FL_HIDDEN = 1 << 1,
88 GENHD_FL_NO_PART = 1 << 2,
89};
90
91enum {
92 DISK_EVENT_MEDIA_CHANGE = 1 << 0, /* media changed */
93 DISK_EVENT_EJECT_REQUEST = 1 << 1, /* eject requested */
94};
95
96enum {
97 /* Poll even if events_poll_msecs is unset */
98 DISK_EVENT_FLAG_POLL = 1 << 0,
99 /* Forward events to udev */
100 DISK_EVENT_FLAG_UEVENT = 1 << 1,
101 /* Block event polling when open for exclusive write */
102 DISK_EVENT_FLAG_BLOCK_ON_EXCL_WRITE = 1 << 2,
103};
104
105struct disk_events;
106struct badblocks;
107
108struct blk_integrity {
109 const struct blk_integrity_profile *profile;
110 unsigned char flags;
111 unsigned char tuple_size;
112 unsigned char pi_offset;
113 unsigned char interval_exp;
114 unsigned char tag_size;
115};
116
117typedef unsigned int __bitwise blk_mode_t;
118
119/* open for reading */
120#define BLK_OPEN_READ ((__force blk_mode_t)(1 << 0))
121/* open for writing */
122#define BLK_OPEN_WRITE ((__force blk_mode_t)(1 << 1))
123/* open exclusively (vs other exclusive openers */
124#define BLK_OPEN_EXCL ((__force blk_mode_t)(1 << 2))
125/* opened with O_NDELAY */
126#define BLK_OPEN_NDELAY ((__force blk_mode_t)(1 << 3))
127/* open for "writes" only for ioctls (specialy hack for floppy.c) */
128#define BLK_OPEN_WRITE_IOCTL ((__force blk_mode_t)(1 << 4))
129/* open is exclusive wrt all other BLK_OPEN_WRITE opens to the device */
130#define BLK_OPEN_RESTRICT_WRITES ((__force blk_mode_t)(1 << 5))
131/* return partition scanning errors */
132#define BLK_OPEN_STRICT_SCAN ((__force blk_mode_t)(1 << 6))
133
134struct gendisk {
135 /*
136 * major/first_minor/minors should not be set by any new driver, the
137 * block core will take care of allocating them automatically.
138 */
139 int major;
140 int first_minor;
141 int minors;
142
143 char disk_name[DISK_NAME_LEN]; /* name of major driver */
144
145 unsigned short events; /* supported events */
146 unsigned short event_flags; /* flags related to event processing */
147
148 struct xarray part_tbl;
149 struct block_device *part0;
150
151 const struct block_device_operations *fops;
152 struct request_queue *queue;
153 void *private_data;
154
155 struct bio_set bio_split;
156
157 int flags;
158 unsigned long state;
159#define GD_NEED_PART_SCAN 0
160#define GD_READ_ONLY 1
161#define GD_DEAD 2
162#define GD_NATIVE_CAPACITY 3
163#define GD_ADDED 4
164#define GD_SUPPRESS_PART_SCAN 5
165#define GD_OWNS_QUEUE 6
166
167 struct mutex open_mutex; /* open/close mutex */
168 unsigned open_partitions; /* number of open partitions */
169
170 struct backing_dev_info *bdi;
171 struct kobject queue_kobj; /* the queue/ directory */
172 struct kobject *slave_dir;
173#ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
174 struct list_head slave_bdevs;
175#endif
176 struct timer_rand_state *random;
177 atomic_t sync_io; /* RAID */
178 struct disk_events *ev;
179
180#ifdef CONFIG_BLK_DEV_ZONED
181 /*
182 * Zoned block device information for request dispatch control.
183 * nr_zones is the total number of zones of the device. This is always
184 * 0 for regular block devices. conv_zones_bitmap is a bitmap of nr_zones
185 * bits which indicates if a zone is conventional (bit set) or
186 * sequential (bit clear). seq_zones_wlock is a bitmap of nr_zones
187 * bits which indicates if a zone is write locked, that is, if a write
188 * request targeting the zone was dispatched.
189 *
190 * Reads of this information must be protected with blk_queue_enter() /
191 * blk_queue_exit(). Modifying this information is only allowed while
192 * no requests are being processed. See also blk_mq_freeze_queue() and
193 * blk_mq_unfreeze_queue().
194 */
195 unsigned int nr_zones;
196 unsigned long *conv_zones_bitmap;
197 unsigned long *seq_zones_wlock;
198#endif /* CONFIG_BLK_DEV_ZONED */
199
200#if IS_ENABLED(CONFIG_CDROM)
201 struct cdrom_device_info *cdi;
202#endif
203 int node_id;
204 struct badblocks *bb;
205 struct lockdep_map lockdep_map;
206 u64 diskseq;
207 blk_mode_t open_mode;
208
209 /*
210 * Independent sector access ranges. This is always NULL for
211 * devices that do not have multiple independent access ranges.
212 */
213 struct blk_independent_access_ranges *ia_ranges;
214};
215
216static inline bool disk_live(struct gendisk *disk)
217{
218 return !inode_unhashed(inode: disk->part0->bd_inode);
219}
220
221/**
222 * disk_openers - returns how many openers are there for a disk
223 * @disk: disk to check
224 *
225 * This returns the number of openers for a disk. Note that this value is only
226 * stable if disk->open_mutex is held.
227 *
228 * Note: Due to a quirk in the block layer open code, each open partition is
229 * only counted once even if there are multiple openers.
230 */
231static inline unsigned int disk_openers(struct gendisk *disk)
232{
233 return atomic_read(v: &disk->part0->bd_openers);
234}
235
236/*
237 * The gendisk is refcounted by the part0 block_device, and the bd_device
238 * therein is also used for device model presentation in sysfs.
239 */
240#define dev_to_disk(device) \
241 (dev_to_bdev(device)->bd_disk)
242#define disk_to_dev(disk) \
243 (&((disk)->part0->bd_device))
244
245#if IS_REACHABLE(CONFIG_CDROM)
246#define disk_to_cdi(disk) ((disk)->cdi)
247#else
248#define disk_to_cdi(disk) NULL
249#endif
250
251static inline dev_t disk_devt(struct gendisk *disk)
252{
253 return MKDEV(disk->major, disk->first_minor);
254}
255
256static inline int blk_validate_block_size(unsigned long bsize)
257{
258 if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(n: bsize))
259 return -EINVAL;
260
261 return 0;
262}
263
264static inline bool blk_op_is_passthrough(blk_opf_t op)
265{
266 op &= REQ_OP_MASK;
267 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
268}
269
270/*
271 * BLK_BOUNCE_NONE: never bounce (default)
272 * BLK_BOUNCE_HIGH: bounce all highmem pages
273 */
274enum blk_bounce {
275 BLK_BOUNCE_NONE,
276 BLK_BOUNCE_HIGH,
277};
278
279struct queue_limits {
280 enum blk_bounce bounce;
281 unsigned long seg_boundary_mask;
282 unsigned long virt_boundary_mask;
283
284 unsigned int max_hw_sectors;
285 unsigned int max_dev_sectors;
286 unsigned int chunk_sectors;
287 unsigned int max_sectors;
288 unsigned int max_user_sectors;
289 unsigned int max_segment_size;
290 unsigned int physical_block_size;
291 unsigned int logical_block_size;
292 unsigned int alignment_offset;
293 unsigned int io_min;
294 unsigned int io_opt;
295 unsigned int max_discard_sectors;
296 unsigned int max_hw_discard_sectors;
297 unsigned int max_user_discard_sectors;
298 unsigned int max_secure_erase_sectors;
299 unsigned int max_write_zeroes_sectors;
300 unsigned int max_zone_append_sectors;
301 unsigned int discard_granularity;
302 unsigned int discard_alignment;
303 unsigned int zone_write_granularity;
304
305 unsigned short max_segments;
306 unsigned short max_integrity_segments;
307 unsigned short max_discard_segments;
308
309 unsigned char misaligned;
310 unsigned char discard_misaligned;
311 unsigned char raid_partial_stripes_expensive;
312 bool zoned;
313 unsigned int max_open_zones;
314 unsigned int max_active_zones;
315
316 /*
317 * Drivers that set dma_alignment to less than 511 must be prepared to
318 * handle individual bvec's that are not a multiple of a SECTOR_SIZE
319 * due to possible offsets.
320 */
321 unsigned int dma_alignment;
322};
323
324typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
325 void *data);
326
327void disk_set_zoned(struct gendisk *disk);
328
329#define BLK_ALL_ZONES ((unsigned int)-1)
330int blkdev_report_zones(struct block_device *bdev, sector_t sector,
331 unsigned int nr_zones, report_zones_cb cb, void *data);
332int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op,
333 sector_t sectors, sector_t nr_sectors);
334int blk_revalidate_disk_zones(struct gendisk *disk,
335 void (*update_driver_data)(struct gendisk *disk));
336
337/*
338 * Independent access ranges: struct blk_independent_access_range describes
339 * a range of contiguous sectors that can be accessed using device command
340 * execution resources that are independent from the resources used for
341 * other access ranges. This is typically found with single-LUN multi-actuator
342 * HDDs where each access range is served by a different set of heads.
343 * The set of independent ranges supported by the device is defined using
344 * struct blk_independent_access_ranges. The independent ranges must not overlap
345 * and must include all sectors within the disk capacity (no sector holes
346 * allowed).
347 * For a device with multiple ranges, requests targeting sectors in different
348 * ranges can be executed in parallel. A request can straddle an access range
349 * boundary.
350 */
351struct blk_independent_access_range {
352 struct kobject kobj;
353 sector_t sector;
354 sector_t nr_sectors;
355};
356
357struct blk_independent_access_ranges {
358 struct kobject kobj;
359 bool sysfs_registered;
360 unsigned int nr_ia_ranges;
361 struct blk_independent_access_range ia_range[];
362};
363
364struct request_queue {
365 /*
366 * The queue owner gets to use this for whatever they like.
367 * ll_rw_blk doesn't touch it.
368 */
369 void *queuedata;
370
371 struct elevator_queue *elevator;
372
373 const struct blk_mq_ops *mq_ops;
374
375 /* sw queues */
376 struct blk_mq_ctx __percpu *queue_ctx;
377
378 /*
379 * various queue flags, see QUEUE_* below
380 */
381 unsigned long queue_flags;
382
383 unsigned int rq_timeout;
384
385 unsigned int queue_depth;
386
387 refcount_t refs;
388
389 /* hw dispatch queues */
390 unsigned int nr_hw_queues;
391 struct xarray hctx_table;
392
393 struct percpu_ref q_usage_counter;
394
395 struct request *last_merge;
396
397 spinlock_t queue_lock;
398
399 int quiesce_depth;
400
401 struct gendisk *disk;
402
403 /*
404 * mq queue kobject
405 */
406 struct kobject *mq_kobj;
407
408 struct queue_limits limits;
409
410#ifdef CONFIG_BLK_DEV_INTEGRITY
411 struct blk_integrity integrity;
412#endif /* CONFIG_BLK_DEV_INTEGRITY */
413
414#ifdef CONFIG_PM
415 struct device *dev;
416 enum rpm_status rpm_status;
417#endif
418
419 /*
420 * Number of contexts that have called blk_set_pm_only(). If this
421 * counter is above zero then only RQF_PM requests are processed.
422 */
423 atomic_t pm_only;
424
425 struct blk_queue_stats *stats;
426 struct rq_qos *rq_qos;
427 struct mutex rq_qos_mutex;
428
429 /*
430 * ida allocated id for this queue. Used to index queues from
431 * ioctx.
432 */
433 int id;
434
435 unsigned int dma_pad_mask;
436
437 /*
438 * queue settings
439 */
440 unsigned long nr_requests; /* Max # of requests */
441
442#ifdef CONFIG_BLK_INLINE_ENCRYPTION
443 struct blk_crypto_profile *crypto_profile;
444 struct kobject *crypto_kobject;
445#endif
446
447 struct timer_list timeout;
448 struct work_struct timeout_work;
449
450 atomic_t nr_active_requests_shared_tags;
451
452 unsigned int required_elevator_features;
453
454 struct blk_mq_tags *sched_shared_tags;
455
456 struct list_head icq_list;
457#ifdef CONFIG_BLK_CGROUP
458 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
459 struct blkcg_gq *root_blkg;
460 struct list_head blkg_list;
461 struct mutex blkcg_mutex;
462#endif
463
464 int node;
465
466 spinlock_t requeue_lock;
467 struct list_head requeue_list;
468 struct delayed_work requeue_work;
469
470#ifdef CONFIG_BLK_DEV_IO_TRACE
471 struct blk_trace __rcu *blk_trace;
472#endif
473 /*
474 * for flush operations
475 */
476 struct blk_flush_queue *fq;
477 struct list_head flush_list;
478
479 struct mutex sysfs_lock;
480 struct mutex sysfs_dir_lock;
481 struct mutex limits_lock;
482
483 /*
484 * for reusing dead hctx instance in case of updating
485 * nr_hw_queues
486 */
487 struct list_head unused_hctx_list;
488 spinlock_t unused_hctx_lock;
489
490 int mq_freeze_depth;
491
492#ifdef CONFIG_BLK_DEV_THROTTLING
493 /* Throttle data */
494 struct throtl_data *td;
495#endif
496 struct rcu_head rcu_head;
497 wait_queue_head_t mq_freeze_wq;
498 /*
499 * Protect concurrent access to q_usage_counter by
500 * percpu_ref_kill() and percpu_ref_reinit().
501 */
502 struct mutex mq_freeze_lock;
503
504 struct blk_mq_tag_set *tag_set;
505 struct list_head tag_set_list;
506
507 struct dentry *debugfs_dir;
508 struct dentry *sched_debugfs_dir;
509 struct dentry *rqos_debugfs_dir;
510 /*
511 * Serializes all debugfs metadata operations using the above dentries.
512 */
513 struct mutex debugfs_mutex;
514
515 bool mq_sysfs_init_done;
516};
517
518/* Keep blk_queue_flag_name[] in sync with the definitions below */
519#define QUEUE_FLAG_STOPPED 0 /* queue is stopped */
520#define QUEUE_FLAG_DYING 1 /* queue being torn down */
521#define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */
522#define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */
523#define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */
524#define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */
525#define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
526#define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */
527#define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */
528#define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */
529#define QUEUE_FLAG_SYNCHRONOUS 11 /* always completes in submit context */
530#define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */
531#define QUEUE_FLAG_HW_WC 13 /* Write back caching supported */
532#define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */
533#define QUEUE_FLAG_STABLE_WRITES 15 /* don't modify blks until WB is done */
534#define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */
535#define QUEUE_FLAG_WC 17 /* Write back caching */
536#define QUEUE_FLAG_FUA 18 /* device supports FUA writes */
537#define QUEUE_FLAG_DAX 19 /* device supports DAX */
538#define QUEUE_FLAG_STATS 20 /* track IO start and completion times */
539#define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */
540#define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */
541#define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */
542#define QUEUE_FLAG_ZONE_RESETALL 26 /* supports Zone Reset All */
543#define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */
544#define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */
545#define QUEUE_FLAG_NOWAIT 29 /* device supports NOWAIT */
546#define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */
547#define QUEUE_FLAG_SKIP_TAGSET_QUIESCE 31 /* quiesce_tagset skip the queue*/
548
549#define QUEUE_FLAG_MQ_DEFAULT ((1UL << QUEUE_FLAG_IO_STAT) | \
550 (1UL << QUEUE_FLAG_SAME_COMP) | \
551 (1UL << QUEUE_FLAG_NOWAIT))
552
553void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
554void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
555bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q);
556
557#define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags)
558#define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
559#define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
560#define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
561#define blk_queue_noxmerges(q) \
562 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
563#define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags)
564#define blk_queue_stable_writes(q) \
565 test_bit(QUEUE_FLAG_STABLE_WRITES, &(q)->queue_flags)
566#define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags)
567#define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags)
568#define blk_queue_zone_resetall(q) \
569 test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags)
570#define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
571#define blk_queue_pci_p2pdma(q) \
572 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags)
573#ifdef CONFIG_BLK_RQ_ALLOC_TIME
574#define blk_queue_rq_alloc_time(q) \
575 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
576#else
577#define blk_queue_rq_alloc_time(q) false
578#endif
579
580#define blk_noretry_request(rq) \
581 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
582 REQ_FAILFAST_DRIVER))
583#define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
584#define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
585#define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
586#define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags)
587#define blk_queue_skip_tagset_quiesce(q) \
588 test_bit(QUEUE_FLAG_SKIP_TAGSET_QUIESCE, &(q)->queue_flags)
589
590extern void blk_set_pm_only(struct request_queue *q);
591extern void blk_clear_pm_only(struct request_queue *q);
592
593#define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
594
595#define dma_map_bvec(dev, bv, dir, attrs) \
596 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
597 (dir), (attrs))
598
599static inline bool queue_is_mq(struct request_queue *q)
600{
601 return q->mq_ops;
602}
603
604#ifdef CONFIG_PM
605static inline enum rpm_status queue_rpm_status(struct request_queue *q)
606{
607 return q->rpm_status;
608}
609#else
610static inline enum rpm_status queue_rpm_status(struct request_queue *q)
611{
612 return RPM_ACTIVE;
613}
614#endif
615
616static inline bool blk_queue_is_zoned(struct request_queue *q)
617{
618 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && q->limits.zoned;
619}
620
621#ifdef CONFIG_BLK_DEV_ZONED
622unsigned int bdev_nr_zones(struct block_device *bdev);
623
624static inline unsigned int disk_nr_zones(struct gendisk *disk)
625{
626 return blk_queue_is_zoned(q: disk->queue) ? disk->nr_zones : 0;
627}
628
629static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector)
630{
631 if (!blk_queue_is_zoned(q: disk->queue))
632 return 0;
633 return sector >> ilog2(disk->queue->limits.chunk_sectors);
634}
635
636static inline bool disk_zone_is_seq(struct gendisk *disk, sector_t sector)
637{
638 if (!blk_queue_is_zoned(q: disk->queue))
639 return false;
640 if (!disk->conv_zones_bitmap)
641 return true;
642 return !test_bit(disk_zone_no(disk, sector), disk->conv_zones_bitmap);
643}
644
645static inline void disk_set_max_open_zones(struct gendisk *disk,
646 unsigned int max_open_zones)
647{
648 disk->queue->limits.max_open_zones = max_open_zones;
649}
650
651static inline void disk_set_max_active_zones(struct gendisk *disk,
652 unsigned int max_active_zones)
653{
654 disk->queue->limits.max_active_zones = max_active_zones;
655}
656
657static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
658{
659 return bdev->bd_disk->queue->limits.max_open_zones;
660}
661
662static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
663{
664 return bdev->bd_disk->queue->limits.max_active_zones;
665}
666
667#else /* CONFIG_BLK_DEV_ZONED */
668static inline unsigned int bdev_nr_zones(struct block_device *bdev)
669{
670 return 0;
671}
672
673static inline unsigned int disk_nr_zones(struct gendisk *disk)
674{
675 return 0;
676}
677static inline bool disk_zone_is_seq(struct gendisk *disk, sector_t sector)
678{
679 return false;
680}
681static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector)
682{
683 return 0;
684}
685static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
686{
687 return 0;
688}
689
690static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
691{
692 return 0;
693}
694#endif /* CONFIG_BLK_DEV_ZONED */
695
696static inline unsigned int blk_queue_depth(struct request_queue *q)
697{
698 if (q->queue_depth)
699 return q->queue_depth;
700
701 return q->nr_requests;
702}
703
704/*
705 * default timeout for SG_IO if none specified
706 */
707#define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
708#define BLK_MIN_SG_TIMEOUT (7 * HZ)
709
710/* This should not be used directly - use rq_for_each_segment */
711#define for_each_bio(_bio) \
712 for (; _bio; _bio = _bio->bi_next)
713
714int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
715 const struct attribute_group **groups);
716static inline int __must_check add_disk(struct gendisk *disk)
717{
718 return device_add_disk(NULL, disk, NULL);
719}
720void del_gendisk(struct gendisk *gp);
721void invalidate_disk(struct gendisk *disk);
722void set_disk_ro(struct gendisk *disk, bool read_only);
723void disk_uevent(struct gendisk *disk, enum kobject_action action);
724
725static inline int get_disk_ro(struct gendisk *disk)
726{
727 return disk->part0->bd_read_only ||
728 test_bit(GD_READ_ONLY, &disk->state);
729}
730
731static inline int bdev_read_only(struct block_device *bdev)
732{
733 return bdev->bd_read_only || get_disk_ro(disk: bdev->bd_disk);
734}
735
736bool set_capacity_and_notify(struct gendisk *disk, sector_t size);
737void disk_force_media_change(struct gendisk *disk);
738void bdev_mark_dead(struct block_device *bdev, bool surprise);
739
740void add_disk_randomness(struct gendisk *disk) __latent_entropy;
741void rand_initialize_disk(struct gendisk *disk);
742
743static inline sector_t get_start_sect(struct block_device *bdev)
744{
745 return bdev->bd_start_sect;
746}
747
748static inline sector_t bdev_nr_sectors(struct block_device *bdev)
749{
750 return bdev->bd_nr_sectors;
751}
752
753static inline loff_t bdev_nr_bytes(struct block_device *bdev)
754{
755 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT;
756}
757
758static inline sector_t get_capacity(struct gendisk *disk)
759{
760 return bdev_nr_sectors(bdev: disk->part0);
761}
762
763static inline u64 sb_bdev_nr_blocks(struct super_block *sb)
764{
765 return bdev_nr_sectors(bdev: sb->s_bdev) >>
766 (sb->s_blocksize_bits - SECTOR_SHIFT);
767}
768
769int bdev_disk_changed(struct gendisk *disk, bool invalidate);
770
771void put_disk(struct gendisk *disk);
772struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node,
773 struct lock_class_key *lkclass);
774
775/**
776 * blk_alloc_disk - allocate a gendisk structure
777 * @lim: queue limits to be used for this disk.
778 * @node_id: numa node to allocate on
779 *
780 * Allocate and pre-initialize a gendisk structure for use with BIO based
781 * drivers.
782 *
783 * Returns an ERR_PTR on error, else the allocated disk.
784 *
785 * Context: can sleep
786 */
787#define blk_alloc_disk(lim, node_id) \
788({ \
789 static struct lock_class_key __key; \
790 \
791 __blk_alloc_disk(lim, node_id, &__key); \
792})
793
794int __register_blkdev(unsigned int major, const char *name,
795 void (*probe)(dev_t devt));
796#define register_blkdev(major, name) \
797 __register_blkdev(major, name, NULL)
798void unregister_blkdev(unsigned int major, const char *name);
799
800bool disk_check_media_change(struct gendisk *disk);
801void set_capacity(struct gendisk *disk, sector_t size);
802
803#ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
804int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk);
805void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk);
806#else
807static inline int bd_link_disk_holder(struct block_device *bdev,
808 struct gendisk *disk)
809{
810 return 0;
811}
812static inline void bd_unlink_disk_holder(struct block_device *bdev,
813 struct gendisk *disk)
814{
815}
816#endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */
817
818dev_t part_devt(struct gendisk *disk, u8 partno);
819void inc_diskseq(struct gendisk *disk);
820void blk_request_module(dev_t devt);
821
822extern int blk_register_queue(struct gendisk *disk);
823extern void blk_unregister_queue(struct gendisk *disk);
824void submit_bio_noacct(struct bio *bio);
825struct bio *bio_split_to_limits(struct bio *bio);
826
827extern int blk_lld_busy(struct request_queue *q);
828extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
829extern void blk_queue_exit(struct request_queue *q);
830extern void blk_sync_queue(struct request_queue *q);
831
832/* Helper to convert REQ_OP_XXX to its string format XXX */
833extern const char *blk_op_str(enum req_op op);
834
835int blk_status_to_errno(blk_status_t status);
836blk_status_t errno_to_blk_status(int errno);
837const char *blk_status_to_str(blk_status_t status);
838
839/* only poll the hardware once, don't continue until a completion was found */
840#define BLK_POLL_ONESHOT (1 << 0)
841int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags);
842int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob,
843 unsigned int flags);
844
845static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
846{
847 return bdev->bd_queue; /* this is never NULL */
848}
849
850/* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
851const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
852
853static inline unsigned int bio_zone_no(struct bio *bio)
854{
855 return disk_zone_no(disk: bio->bi_bdev->bd_disk, sector: bio->bi_iter.bi_sector);
856}
857
858static inline unsigned int bio_zone_is_seq(struct bio *bio)
859{
860 return disk_zone_is_seq(disk: bio->bi_bdev->bd_disk, sector: bio->bi_iter.bi_sector);
861}
862
863/*
864 * Return how much of the chunk is left to be used for I/O at a given offset.
865 */
866static inline unsigned int blk_chunk_sectors_left(sector_t offset,
867 unsigned int chunk_sectors)
868{
869 if (unlikely(!is_power_of_2(chunk_sectors)))
870 return chunk_sectors - sector_div(offset, chunk_sectors);
871 return chunk_sectors - (offset & (chunk_sectors - 1));
872}
873
874/**
875 * queue_limits_start_update - start an atomic update of queue limits
876 * @q: queue to update
877 *
878 * This functions starts an atomic update of the queue limits. It takes a lock
879 * to prevent other updates and returns a snapshot of the current limits that
880 * the caller can modify. The caller must call queue_limits_commit_update()
881 * to finish the update.
882 *
883 * Context: process context. The caller must have frozen the queue or ensured
884 * that there is outstanding I/O by other means.
885 */
886static inline struct queue_limits
887queue_limits_start_update(struct request_queue *q)
888 __acquires(q->limits_lock)
889{
890 mutex_lock(&q->limits_lock);
891 return q->limits;
892}
893int queue_limits_commit_update(struct request_queue *q,
894 struct queue_limits *lim);
895int queue_limits_set(struct request_queue *q, struct queue_limits *lim);
896
897/*
898 * Access functions for manipulating queue properties
899 */
900void blk_queue_bounce_limit(struct request_queue *q, enum blk_bounce limit);
901extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int);
902extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int);
903extern void blk_queue_max_segments(struct request_queue *, unsigned short);
904extern void blk_queue_max_discard_segments(struct request_queue *,
905 unsigned short);
906void blk_queue_max_secure_erase_sectors(struct request_queue *q,
907 unsigned int max_sectors);
908extern void blk_queue_max_segment_size(struct request_queue *, unsigned int);
909extern void blk_queue_max_discard_sectors(struct request_queue *q,
910 unsigned int max_discard_sectors);
911extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q,
912 unsigned int max_write_same_sectors);
913extern void blk_queue_logical_block_size(struct request_queue *, unsigned int);
914extern void blk_queue_max_zone_append_sectors(struct request_queue *q,
915 unsigned int max_zone_append_sectors);
916extern void blk_queue_physical_block_size(struct request_queue *, unsigned int);
917void blk_queue_zone_write_granularity(struct request_queue *q,
918 unsigned int size);
919extern void blk_queue_alignment_offset(struct request_queue *q,
920 unsigned int alignment);
921void disk_update_readahead(struct gendisk *disk);
922extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
923extern void blk_queue_io_min(struct request_queue *q, unsigned int min);
924extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
925extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt);
926extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
927extern void blk_set_stacking_limits(struct queue_limits *lim);
928extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
929 sector_t offset);
930void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev,
931 sector_t offset, const char *pfx);
932extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int);
933extern void blk_queue_segment_boundary(struct request_queue *, unsigned long);
934extern void blk_queue_virt_boundary(struct request_queue *, unsigned long);
935extern void blk_queue_dma_alignment(struct request_queue *, int);
936extern void blk_queue_update_dma_alignment(struct request_queue *, int);
937extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
938extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua);
939
940struct blk_independent_access_ranges *
941disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges);
942void disk_set_independent_access_ranges(struct gendisk *disk,
943 struct blk_independent_access_ranges *iars);
944
945/*
946 * Elevator features for blk_queue_required_elevator_features:
947 */
948/* Supports zoned block devices sequential write constraint */
949#define ELEVATOR_F_ZBD_SEQ_WRITE (1U << 0)
950
951extern void blk_queue_required_elevator_features(struct request_queue *q,
952 unsigned int features);
953extern bool blk_queue_can_use_dma_map_merging(struct request_queue *q,
954 struct device *dev);
955
956bool __must_check blk_get_queue(struct request_queue *);
957extern void blk_put_queue(struct request_queue *);
958
959void blk_mark_disk_dead(struct gendisk *disk);
960
961#ifdef CONFIG_BLOCK
962/*
963 * blk_plug permits building a queue of related requests by holding the I/O
964 * fragments for a short period. This allows merging of sequential requests
965 * into single larger request. As the requests are moved from a per-task list to
966 * the device's request_queue in a batch, this results in improved scalability
967 * as the lock contention for request_queue lock is reduced.
968 *
969 * It is ok not to disable preemption when adding the request to the plug list
970 * or when attempting a merge. For details, please see schedule() where
971 * blk_flush_plug() is called.
972 */
973struct blk_plug {
974 struct request *mq_list; /* blk-mq requests */
975
976 /* if ios_left is > 1, we can batch tag/rq allocations */
977 struct request *cached_rq;
978 u64 cur_ktime;
979 unsigned short nr_ios;
980
981 unsigned short rq_count;
982
983 bool multiple_queues;
984 bool has_elevator;
985
986 struct list_head cb_list; /* md requires an unplug callback */
987};
988
989struct blk_plug_cb;
990typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
991struct blk_plug_cb {
992 struct list_head list;
993 blk_plug_cb_fn callback;
994 void *data;
995};
996extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
997 void *data, int size);
998extern void blk_start_plug(struct blk_plug *);
999extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short);
1000extern void blk_finish_plug(struct blk_plug *);
1001
1002void __blk_flush_plug(struct blk_plug *plug, bool from_schedule);
1003static inline void blk_flush_plug(struct blk_plug *plug, bool async)
1004{
1005 if (plug)
1006 __blk_flush_plug(plug, from_schedule: async);
1007}
1008
1009/*
1010 * tsk == current here
1011 */
1012static inline void blk_plug_invalidate_ts(struct task_struct *tsk)
1013{
1014 struct blk_plug *plug = tsk->plug;
1015
1016 if (plug)
1017 plug->cur_ktime = 0;
1018 current->flags &= ~PF_BLOCK_TS;
1019}
1020
1021int blkdev_issue_flush(struct block_device *bdev);
1022long nr_blockdev_pages(void);
1023#else /* CONFIG_BLOCK */
1024struct blk_plug {
1025};
1026
1027static inline void blk_start_plug_nr_ios(struct blk_plug *plug,
1028 unsigned short nr_ios)
1029{
1030}
1031
1032static inline void blk_start_plug(struct blk_plug *plug)
1033{
1034}
1035
1036static inline void blk_finish_plug(struct blk_plug *plug)
1037{
1038}
1039
1040static inline void blk_flush_plug(struct blk_plug *plug, bool async)
1041{
1042}
1043
1044static inline void blk_plug_invalidate_ts(struct task_struct *tsk)
1045{
1046}
1047
1048static inline int blkdev_issue_flush(struct block_device *bdev)
1049{
1050 return 0;
1051}
1052
1053static inline long nr_blockdev_pages(void)
1054{
1055 return 0;
1056}
1057#endif /* CONFIG_BLOCK */
1058
1059extern void blk_io_schedule(void);
1060
1061int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1062 sector_t nr_sects, gfp_t gfp_mask);
1063int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1064 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop);
1065int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector,
1066 sector_t nr_sects, gfp_t gfp);
1067
1068#define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1069#define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1070
1071extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1072 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1073 unsigned flags);
1074extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1075 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1076
1077static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1078 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1079{
1080 return blkdev_issue_discard(bdev: sb->s_bdev,
1081 sector: block << (sb->s_blocksize_bits -
1082 SECTOR_SHIFT),
1083 nr_sects: nr_blocks << (sb->s_blocksize_bits -
1084 SECTOR_SHIFT),
1085 gfp_mask);
1086}
1087static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1088 sector_t nr_blocks, gfp_t gfp_mask)
1089{
1090 return blkdev_issue_zeroout(bdev: sb->s_bdev,
1091 sector: block << (sb->s_blocksize_bits -
1092 SECTOR_SHIFT),
1093 nr_sects: nr_blocks << (sb->s_blocksize_bits -
1094 SECTOR_SHIFT),
1095 gfp_mask, flags: 0);
1096}
1097
1098static inline bool bdev_is_partition(struct block_device *bdev)
1099{
1100 return bdev->bd_partno;
1101}
1102
1103enum blk_default_limits {
1104 BLK_MAX_SEGMENTS = 128,
1105 BLK_SAFE_MAX_SECTORS = 255,
1106 BLK_MAX_SEGMENT_SIZE = 65536,
1107 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1108};
1109
1110/*
1111 * Default upper limit for the software max_sectors limit used for
1112 * regular file system I/O. This can be increased through sysfs.
1113 *
1114 * Not to be confused with the max_hw_sector limit that is entirely
1115 * controlled by the driver, usually based on hardware limits.
1116 */
1117#define BLK_DEF_MAX_SECTORS_CAP 2560u
1118
1119static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1120{
1121 return q->limits.seg_boundary_mask;
1122}
1123
1124static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1125{
1126 return q->limits.virt_boundary_mask;
1127}
1128
1129static inline unsigned int queue_max_sectors(const struct request_queue *q)
1130{
1131 return q->limits.max_sectors;
1132}
1133
1134static inline unsigned int queue_max_bytes(struct request_queue *q)
1135{
1136 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9;
1137}
1138
1139static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1140{
1141 return q->limits.max_hw_sectors;
1142}
1143
1144static inline unsigned short queue_max_segments(const struct request_queue *q)
1145{
1146 return q->limits.max_segments;
1147}
1148
1149static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1150{
1151 return q->limits.max_discard_segments;
1152}
1153
1154static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1155{
1156 return q->limits.max_segment_size;
1157}
1158
1159static inline unsigned int queue_max_zone_append_sectors(const struct request_queue *q)
1160{
1161
1162 const struct queue_limits *l = &q->limits;
1163
1164 return min(l->max_zone_append_sectors, l->max_sectors);
1165}
1166
1167static inline unsigned int
1168bdev_max_zone_append_sectors(struct block_device *bdev)
1169{
1170 return queue_max_zone_append_sectors(q: bdev_get_queue(bdev));
1171}
1172
1173static inline unsigned int bdev_max_segments(struct block_device *bdev)
1174{
1175 return queue_max_segments(q: bdev_get_queue(bdev));
1176}
1177
1178static inline unsigned queue_logical_block_size(const struct request_queue *q)
1179{
1180 int retval = 512;
1181
1182 if (q && q->limits.logical_block_size)
1183 retval = q->limits.logical_block_size;
1184
1185 return retval;
1186}
1187
1188static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1189{
1190 return queue_logical_block_size(q: bdev_get_queue(bdev));
1191}
1192
1193static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1194{
1195 return q->limits.physical_block_size;
1196}
1197
1198static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1199{
1200 return queue_physical_block_size(q: bdev_get_queue(bdev));
1201}
1202
1203static inline unsigned int queue_io_min(const struct request_queue *q)
1204{
1205 return q->limits.io_min;
1206}
1207
1208static inline int bdev_io_min(struct block_device *bdev)
1209{
1210 return queue_io_min(q: bdev_get_queue(bdev));
1211}
1212
1213static inline unsigned int queue_io_opt(const struct request_queue *q)
1214{
1215 return q->limits.io_opt;
1216}
1217
1218static inline int bdev_io_opt(struct block_device *bdev)
1219{
1220 return queue_io_opt(q: bdev_get_queue(bdev));
1221}
1222
1223static inline unsigned int
1224queue_zone_write_granularity(const struct request_queue *q)
1225{
1226 return q->limits.zone_write_granularity;
1227}
1228
1229static inline unsigned int
1230bdev_zone_write_granularity(struct block_device *bdev)
1231{
1232 return queue_zone_write_granularity(q: bdev_get_queue(bdev));
1233}
1234
1235int bdev_alignment_offset(struct block_device *bdev);
1236unsigned int bdev_discard_alignment(struct block_device *bdev);
1237
1238static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev)
1239{
1240 return bdev_get_queue(bdev)->limits.max_discard_sectors;
1241}
1242
1243static inline unsigned int bdev_discard_granularity(struct block_device *bdev)
1244{
1245 return bdev_get_queue(bdev)->limits.discard_granularity;
1246}
1247
1248static inline unsigned int
1249bdev_max_secure_erase_sectors(struct block_device *bdev)
1250{
1251 return bdev_get_queue(bdev)->limits.max_secure_erase_sectors;
1252}
1253
1254static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1255{
1256 struct request_queue *q = bdev_get_queue(bdev);
1257
1258 if (q)
1259 return q->limits.max_write_zeroes_sectors;
1260
1261 return 0;
1262}
1263
1264static inline bool bdev_nonrot(struct block_device *bdev)
1265{
1266 return blk_queue_nonrot(bdev_get_queue(bdev));
1267}
1268
1269static inline bool bdev_synchronous(struct block_device *bdev)
1270{
1271 return test_bit(QUEUE_FLAG_SYNCHRONOUS,
1272 &bdev_get_queue(bdev)->queue_flags);
1273}
1274
1275static inline bool bdev_stable_writes(struct block_device *bdev)
1276{
1277 return test_bit(QUEUE_FLAG_STABLE_WRITES,
1278 &bdev_get_queue(bdev)->queue_flags);
1279}
1280
1281static inline bool bdev_write_cache(struct block_device *bdev)
1282{
1283 return test_bit(QUEUE_FLAG_WC, &bdev_get_queue(bdev)->queue_flags);
1284}
1285
1286static inline bool bdev_fua(struct block_device *bdev)
1287{
1288 return test_bit(QUEUE_FLAG_FUA, &bdev_get_queue(bdev)->queue_flags);
1289}
1290
1291static inline bool bdev_nowait(struct block_device *bdev)
1292{
1293 return test_bit(QUEUE_FLAG_NOWAIT, &bdev_get_queue(bdev)->queue_flags);
1294}
1295
1296static inline bool bdev_is_zoned(struct block_device *bdev)
1297{
1298 return blk_queue_is_zoned(q: bdev_get_queue(bdev));
1299}
1300
1301static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec)
1302{
1303 return disk_zone_no(disk: bdev->bd_disk, sector: sec);
1304}
1305
1306/* Whether write serialization is required for @op on zoned devices. */
1307static inline bool op_needs_zoned_write_locking(enum req_op op)
1308{
1309 return op == REQ_OP_WRITE || op == REQ_OP_WRITE_ZEROES;
1310}
1311
1312static inline bool bdev_op_is_zoned_write(struct block_device *bdev,
1313 enum req_op op)
1314{
1315 return bdev_is_zoned(bdev) && op_needs_zoned_write_locking(op);
1316}
1317
1318static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1319{
1320 struct request_queue *q = bdev_get_queue(bdev);
1321
1322 if (!blk_queue_is_zoned(q))
1323 return 0;
1324 return q->limits.chunk_sectors;
1325}
1326
1327static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev,
1328 sector_t sector)
1329{
1330 return sector & (bdev_zone_sectors(bdev) - 1);
1331}
1332
1333static inline bool bdev_is_zone_start(struct block_device *bdev,
1334 sector_t sector)
1335{
1336 return bdev_offset_from_zone_start(bdev, sector) == 0;
1337}
1338
1339static inline int queue_dma_alignment(const struct request_queue *q)
1340{
1341 return q ? q->limits.dma_alignment : 511;
1342}
1343
1344static inline unsigned int bdev_dma_alignment(struct block_device *bdev)
1345{
1346 return queue_dma_alignment(q: bdev_get_queue(bdev));
1347}
1348
1349static inline bool bdev_iter_is_aligned(struct block_device *bdev,
1350 struct iov_iter *iter)
1351{
1352 return iov_iter_is_aligned(i: iter, addr_mask: bdev_dma_alignment(bdev),
1353 len_mask: bdev_logical_block_size(bdev) - 1);
1354}
1355
1356static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1357 unsigned int len)
1358{
1359 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask;
1360 return !(addr & alignment) && !(len & alignment);
1361}
1362
1363/* assumes size > 256 */
1364static inline unsigned int blksize_bits(unsigned int size)
1365{
1366 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT;
1367}
1368
1369static inline unsigned int block_size(struct block_device *bdev)
1370{
1371 return 1 << bdev->bd_inode->i_blkbits;
1372}
1373
1374int kblockd_schedule_work(struct work_struct *work);
1375int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1376
1377#define MODULE_ALIAS_BLOCKDEV(major,minor) \
1378 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1379#define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1380 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1381
1382#ifdef CONFIG_BLK_INLINE_ENCRYPTION
1383
1384bool blk_crypto_register(struct blk_crypto_profile *profile,
1385 struct request_queue *q);
1386
1387#else /* CONFIG_BLK_INLINE_ENCRYPTION */
1388
1389static inline bool blk_crypto_register(struct blk_crypto_profile *profile,
1390 struct request_queue *q)
1391{
1392 return true;
1393}
1394
1395#endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1396
1397enum blk_unique_id {
1398 /* these match the Designator Types specified in SPC */
1399 BLK_UID_T10 = 1,
1400 BLK_UID_EUI64 = 2,
1401 BLK_UID_NAA = 3,
1402};
1403
1404struct block_device_operations {
1405 void (*submit_bio)(struct bio *bio);
1406 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob,
1407 unsigned int flags);
1408 int (*open)(struct gendisk *disk, blk_mode_t mode);
1409 void (*release)(struct gendisk *disk);
1410 int (*ioctl)(struct block_device *bdev, blk_mode_t mode,
1411 unsigned cmd, unsigned long arg);
1412 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode,
1413 unsigned cmd, unsigned long arg);
1414 unsigned int (*check_events) (struct gendisk *disk,
1415 unsigned int clearing);
1416 void (*unlock_native_capacity) (struct gendisk *);
1417 int (*getgeo)(struct block_device *, struct hd_geometry *);
1418 int (*set_read_only)(struct block_device *bdev, bool ro);
1419 void (*free_disk)(struct gendisk *disk);
1420 /* this callback is with swap_lock and sometimes page table lock held */
1421 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1422 int (*report_zones)(struct gendisk *, sector_t sector,
1423 unsigned int nr_zones, report_zones_cb cb, void *data);
1424 char *(*devnode)(struct gendisk *disk, umode_t *mode);
1425 /* returns the length of the identifier or a negative errno: */
1426 int (*get_unique_id)(struct gendisk *disk, u8 id[16],
1427 enum blk_unique_id id_type);
1428 struct module *owner;
1429 const struct pr_ops *pr_ops;
1430
1431 /*
1432 * Special callback for probing GPT entry at a given sector.
1433 * Needed by Android devices, used by GPT scanner and MMC blk
1434 * driver.
1435 */
1436 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector);
1437};
1438
1439#ifdef CONFIG_COMPAT
1440extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t,
1441 unsigned int, unsigned long);
1442#else
1443#define blkdev_compat_ptr_ioctl NULL
1444#endif
1445
1446static inline void blk_wake_io_task(struct task_struct *waiter)
1447{
1448 /*
1449 * If we're polling, the task itself is doing the completions. For
1450 * that case, we don't need to signal a wakeup, it's enough to just
1451 * mark us as RUNNING.
1452 */
1453 if (waiter == current)
1454 __set_current_state(TASK_RUNNING);
1455 else
1456 wake_up_process(tsk: waiter);
1457}
1458
1459unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op,
1460 unsigned long start_time);
1461void bdev_end_io_acct(struct block_device *bdev, enum req_op op,
1462 unsigned int sectors, unsigned long start_time);
1463
1464unsigned long bio_start_io_acct(struct bio *bio);
1465void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1466 struct block_device *orig_bdev);
1467
1468/**
1469 * bio_end_io_acct - end I/O accounting for bio based drivers
1470 * @bio: bio to end account for
1471 * @start_time: start time returned by bio_start_io_acct()
1472 */
1473static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1474{
1475 return bio_end_io_acct_remapped(bio, start_time, orig_bdev: bio->bi_bdev);
1476}
1477
1478int bdev_read_only(struct block_device *bdev);
1479int set_blocksize(struct block_device *bdev, int size);
1480
1481int lookup_bdev(const char *pathname, dev_t *dev);
1482
1483void blkdev_show(struct seq_file *seqf, off_t offset);
1484
1485#define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */
1486#define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */
1487#ifdef CONFIG_BLOCK
1488#define BLKDEV_MAJOR_MAX 512
1489#else
1490#define BLKDEV_MAJOR_MAX 0
1491#endif
1492
1493struct blk_holder_ops {
1494 void (*mark_dead)(struct block_device *bdev, bool surprise);
1495
1496 /*
1497 * Sync the file system mounted on the block device.
1498 */
1499 void (*sync)(struct block_device *bdev);
1500
1501 /*
1502 * Freeze the file system mounted on the block device.
1503 */
1504 int (*freeze)(struct block_device *bdev);
1505
1506 /*
1507 * Thaw the file system mounted on the block device.
1508 */
1509 int (*thaw)(struct block_device *bdev);
1510};
1511
1512/*
1513 * For filesystems using @fs_holder_ops, the @holder argument passed to
1514 * helpers used to open and claim block devices via
1515 * bd_prepare_to_claim() must point to a superblock.
1516 */
1517extern const struct blk_holder_ops fs_holder_ops;
1518
1519/*
1520 * Return the correct open flags for blkdev_get_by_* for super block flags
1521 * as stored in sb->s_flags.
1522 */
1523#define sb_open_mode(flags) \
1524 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \
1525 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE))
1526
1527struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder,
1528 const struct blk_holder_ops *hops);
1529struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode,
1530 void *holder, const struct blk_holder_ops *hops);
1531int bd_prepare_to_claim(struct block_device *bdev, void *holder,
1532 const struct blk_holder_ops *hops);
1533void bd_abort_claiming(struct block_device *bdev, void *holder);
1534
1535/* just for blk-cgroup, don't use elsewhere */
1536struct block_device *blkdev_get_no_open(dev_t dev);
1537void blkdev_put_no_open(struct block_device *bdev);
1538
1539struct block_device *I_BDEV(struct inode *inode);
1540struct block_device *file_bdev(struct file *bdev_file);
1541
1542#ifdef CONFIG_BLOCK
1543void invalidate_bdev(struct block_device *bdev);
1544int sync_blockdev(struct block_device *bdev);
1545int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend);
1546int sync_blockdev_nowait(struct block_device *bdev);
1547void sync_bdevs(bool wait);
1548void bdev_statx_dioalign(struct inode *inode, struct kstat *stat);
1549void printk_all_partitions(void);
1550int __init early_lookup_bdev(const char *pathname, dev_t *dev);
1551#else
1552static inline void invalidate_bdev(struct block_device *bdev)
1553{
1554}
1555static inline int sync_blockdev(struct block_device *bdev)
1556{
1557 return 0;
1558}
1559static inline int sync_blockdev_nowait(struct block_device *bdev)
1560{
1561 return 0;
1562}
1563static inline void sync_bdevs(bool wait)
1564{
1565}
1566static inline void bdev_statx_dioalign(struct inode *inode, struct kstat *stat)
1567{
1568}
1569static inline void printk_all_partitions(void)
1570{
1571}
1572static inline int early_lookup_bdev(const char *pathname, dev_t *dev)
1573{
1574 return -EINVAL;
1575}
1576#endif /* CONFIG_BLOCK */
1577
1578int bdev_freeze(struct block_device *bdev);
1579int bdev_thaw(struct block_device *bdev);
1580void bdev_fput(struct file *bdev_file);
1581
1582struct io_comp_batch {
1583 struct request *req_list;
1584 bool need_ts;
1585 void (*complete)(struct io_comp_batch *);
1586};
1587
1588#define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { }
1589
1590#endif /* _LINUX_BLKDEV_H */
1591

source code of linux/include/linux/blkdev.h