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
29 struct module;
30 struct request_queue;
31 struct elevator_queue;
32 struct blk_trace;
33 struct request;
34 struct sg_io_hdr;
35 struct blkcg_gq;
36 struct blk_flush_queue;
37 struct kiocb;
38 struct pr_ops;
39 struct rq_qos;
40 struct blk_queue_stats;
41 struct blk_stat_callback;
42 struct blk_crypto_profile;
43
44 extern const struct device_type disk_type;
45 extern const struct device_type part_type;
46 extern 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
64 struct 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 */
85 enum {
86 GENHD_FL_REMOVABLE = 1 << 0,
87 GENHD_FL_HIDDEN = 1 << 1,
88 GENHD_FL_NO_PART = 1 << 2,
89 };
90
91 enum {
92 DISK_EVENT_MEDIA_CHANGE = 1 << 0, /* media changed */
93 DISK_EVENT_EJECT_REQUEST = 1 << 1, /* eject requested */
94 };
95
96 enum {
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
105 struct disk_events;
106 struct badblocks;
107
108 enum blk_integrity_checksum {
109 BLK_INTEGRITY_CSUM_NONE = 0,
110 BLK_INTEGRITY_CSUM_IP = 1,
111 BLK_INTEGRITY_CSUM_CRC = 2,
112 BLK_INTEGRITY_CSUM_CRC64 = 3,
113 } __packed ;
114
115 struct blk_integrity {
116 unsigned char flags;
117 enum blk_integrity_checksum csum_type;
118 unsigned char tuple_size;
119 unsigned char pi_offset;
120 unsigned char interval_exp;
121 unsigned char tag_size;
122 };
123
124 typedef unsigned int __bitwise blk_mode_t;
125
126 /* open for reading */
127 #define BLK_OPEN_READ ((__force blk_mode_t)(1 << 0))
128 /* open for writing */
129 #define BLK_OPEN_WRITE ((__force blk_mode_t)(1 << 1))
130 /* open exclusively (vs other exclusive openers */
131 #define BLK_OPEN_EXCL ((__force blk_mode_t)(1 << 2))
132 /* opened with O_NDELAY */
133 #define BLK_OPEN_NDELAY ((__force blk_mode_t)(1 << 3))
134 /* open for "writes" only for ioctls (specialy hack for floppy.c) */
135 #define BLK_OPEN_WRITE_IOCTL ((__force blk_mode_t)(1 << 4))
136 /* open is exclusive wrt all other BLK_OPEN_WRITE opens to the device */
137 #define BLK_OPEN_RESTRICT_WRITES ((__force blk_mode_t)(1 << 5))
138 /* return partition scanning errors */
139 #define BLK_OPEN_STRICT_SCAN ((__force blk_mode_t)(1 << 6))
140
141 struct gendisk {
142 /*
143 * major/first_minor/minors should not be set by any new driver, the
144 * block core will take care of allocating them automatically.
145 */
146 int major;
147 int first_minor;
148 int minors;
149
150 char disk_name[DISK_NAME_LEN]; /* name of major driver */
151
152 unsigned short events; /* supported events */
153 unsigned short event_flags; /* flags related to event processing */
154
155 struct xarray part_tbl;
156 struct block_device *part0;
157
158 const struct block_device_operations *fops;
159 struct request_queue *queue;
160 void *private_data;
161
162 struct bio_set bio_split;
163
164 int flags;
165 unsigned long state;
166 #define GD_NEED_PART_SCAN 0
167 #define GD_READ_ONLY 1
168 #define GD_DEAD 2
169 #define GD_NATIVE_CAPACITY 3
170 #define GD_ADDED 4
171 #define GD_SUPPRESS_PART_SCAN 5
172 #define GD_OWNS_QUEUE 6
173
174 struct mutex open_mutex; /* open/close mutex */
175 unsigned open_partitions; /* number of open partitions */
176
177 struct backing_dev_info *bdi;
178 struct kobject queue_kobj; /* the queue/ directory */
179 struct kobject *slave_dir;
180 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
181 struct list_head slave_bdevs;
182 #endif
183 struct timer_rand_state *random;
184 atomic_t sync_io; /* RAID */
185 struct disk_events *ev;
186
187 #ifdef CONFIG_BLK_DEV_ZONED
188 /*
189 * Zoned block device information. Reads of this information must be
190 * protected with blk_queue_enter() / blk_queue_exit(). Modifying this
191 * information is only allowed while no requests are being processed.
192 * See also blk_mq_freeze_queue() and blk_mq_unfreeze_queue().
193 */
194 unsigned int nr_zones;
195 unsigned int zone_capacity;
196 unsigned int last_zone_capacity;
197 unsigned long *conv_zones_bitmap;
198 unsigned int zone_wplugs_hash_bits;
199 spinlock_t zone_wplugs_lock;
200 struct mempool_s *zone_wplugs_pool;
201 struct hlist_head *zone_wplugs_hash;
202 struct list_head zone_wplugs_err_list;
203 struct work_struct zone_wplugs_work;
204 struct workqueue_struct *zone_wplugs_wq;
205 #endif /* CONFIG_BLK_DEV_ZONED */
206
207 #if IS_ENABLED(CONFIG_CDROM)
208 struct cdrom_device_info *cdi;
209 #endif
210 int node_id;
211 struct badblocks *bb;
212 struct lockdep_map lockdep_map;
213 u64 diskseq;
214 blk_mode_t open_mode;
215
216 /*
217 * Independent sector access ranges. This is always NULL for
218 * devices that do not have multiple independent access ranges.
219 */
220 struct blk_independent_access_ranges *ia_ranges;
221 };
222
223 /**
224 * disk_openers - returns how many openers are there for a disk
225 * @disk: disk to check
226 *
227 * This returns the number of openers for a disk. Note that this value is only
228 * stable if disk->open_mutex is held.
229 *
230 * Note: Due to a quirk in the block layer open code, each open partition is
231 * only counted once even if there are multiple openers.
232 */
233 static inline unsigned int disk_openers(struct gendisk *disk)
234 {
235 return atomic_read(&disk->part0->bd_openers);
236 }
237
238 /**
239 * disk_has_partscan - return %true if partition scanning is enabled on a disk
240 * @disk: disk to check
241 *
242 * Returns %true if partitions scanning is enabled for @disk, or %false if
243 * partition scanning is disabled either permanently or temporarily.
244 */
245 static inline bool disk_has_partscan(struct gendisk *disk)
246 {
247 return !(disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN)) &&
248 !test_bit(GD_SUPPRESS_PART_SCAN, &disk->state);
249 }
250
251 /*
252 * The gendisk is refcounted by the part0 block_device, and the bd_device
253 * therein is also used for device model presentation in sysfs.
254 */
255 #define dev_to_disk(device) \
256 (dev_to_bdev(device)->bd_disk)
257 #define disk_to_dev(disk) \
258 (&((disk)->part0->bd_device))
259
260 #if IS_REACHABLE(CONFIG_CDROM)
261 #define disk_to_cdi(disk) ((disk)->cdi)
262 #else
263 #define disk_to_cdi(disk) NULL
264 #endif
265
266 static inline dev_t disk_devt(struct gendisk *disk)
267 {
268 return MKDEV(disk->major, disk->first_minor);
269 }
270
271 /* blk_validate_limits() validates bsize, so drivers don't usually need to */
272 static inline int blk_validate_block_size(unsigned long bsize)
273 {
274 if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(bsize))
275 return -EINVAL;
276
277 return 0;
278 }
279
280 static inline bool blk_op_is_passthrough(blk_opf_t op)
281 {
282 op &= REQ_OP_MASK;
283 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
284 }
285
286 /* flags set by the driver in queue_limits.features */
287 typedef unsigned int __bitwise blk_features_t;
288
289 /* supports a volatile write cache */
290 #define BLK_FEAT_WRITE_CACHE ((__force blk_features_t)(1u << 0))
291
292 /* supports passing on the FUA bit */
293 #define BLK_FEAT_FUA ((__force blk_features_t)(1u << 1))
294
295 /* rotational device (hard drive or floppy) */
296 #define BLK_FEAT_ROTATIONAL ((__force blk_features_t)(1u << 2))
297
298 /* contributes to the random number pool */
299 #define BLK_FEAT_ADD_RANDOM ((__force blk_features_t)(1u << 3))
300
301 /* do disk/partitions IO accounting */
302 #define BLK_FEAT_IO_STAT ((__force blk_features_t)(1u << 4))
303
304 /* don't modify data until writeback is done */
305 #define BLK_FEAT_STABLE_WRITES ((__force blk_features_t)(1u << 5))
306
307 /* always completes in submit context */
308 #define BLK_FEAT_SYNCHRONOUS ((__force blk_features_t)(1u << 6))
309
310 /* supports REQ_NOWAIT */
311 #define BLK_FEAT_NOWAIT ((__force blk_features_t)(1u << 7))
312
313 /* supports DAX */
314 #define BLK_FEAT_DAX ((__force blk_features_t)(1u << 8))
315
316 /* supports I/O polling */
317 #define BLK_FEAT_POLL ((__force blk_features_t)(1u << 9))
318
319 /* is a zoned device */
320 #define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10))
321
322 /* supports PCI(e) p2p requests */
323 #define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12))
324
325 /* skip this queue in blk_mq_(un)quiesce_tagset */
326 #define BLK_FEAT_SKIP_TAGSET_QUIESCE ((__force blk_features_t)(1u << 13))
327
328 /* bounce all highmem pages */
329 #define BLK_FEAT_BOUNCE_HIGH ((__force blk_features_t)(1u << 14))
330
331 /* undocumented magic for bcache */
332 #define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \
333 ((__force blk_features_t)(1u << 15))
334
335 /*
336 * Flags automatically inherited when stacking limits.
337 */
338 #define BLK_FEAT_INHERIT_MASK \
339 (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \
340 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH | \
341 BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE)
342
343 /* internal flags in queue_limits.flags */
344 typedef unsigned int __bitwise blk_flags_t;
345
346 /* do not send FLUSH/FUA commands despite advertising a write cache */
347 #define BLK_FLAG_WRITE_CACHE_DISABLED ((__force blk_flags_t)(1u << 0))
348
349 /* I/O topology is misaligned */
350 #define BLK_FLAG_MISALIGNED ((__force blk_flags_t)(1u << 1))
351
352 struct queue_limits {
353 blk_features_t features;
354 blk_flags_t flags;
355 unsigned long seg_boundary_mask;
356 unsigned long virt_boundary_mask;
357
358 unsigned int max_hw_sectors;
359 unsigned int max_dev_sectors;
360 unsigned int chunk_sectors;
361 unsigned int max_sectors;
362 unsigned int max_user_sectors;
363 unsigned int max_segment_size;
364 unsigned int physical_block_size;
365 unsigned int logical_block_size;
366 unsigned int alignment_offset;
367 unsigned int io_min;
368 unsigned int io_opt;
369 unsigned int max_discard_sectors;
370 unsigned int max_hw_discard_sectors;
371 unsigned int max_user_discard_sectors;
372 unsigned int max_secure_erase_sectors;
373 unsigned int max_write_zeroes_sectors;
374 unsigned int max_zone_append_sectors;
375 unsigned int discard_granularity;
376 unsigned int discard_alignment;
377 unsigned int zone_write_granularity;
378
379 /* atomic write limits */
380 unsigned int atomic_write_hw_max;
381 unsigned int atomic_write_max_sectors;
382 unsigned int atomic_write_hw_boundary;
383 unsigned int atomic_write_boundary_sectors;
384 unsigned int atomic_write_hw_unit_min;
385 unsigned int atomic_write_unit_min;
386 unsigned int atomic_write_hw_unit_max;
387 unsigned int atomic_write_unit_max;
388
389 unsigned short max_segments;
390 unsigned short max_integrity_segments;
391 unsigned short max_discard_segments;
392
393 unsigned int max_open_zones;
394 unsigned int max_active_zones;
395
396 /*
397 * Drivers that set dma_alignment to less than 511 must be prepared to
398 * handle individual bvec's that are not a multiple of a SECTOR_SIZE
399 * due to possible offsets.
400 */
401 unsigned int dma_alignment;
402 unsigned int dma_pad_mask;
403
404 struct blk_integrity integrity;
405 };
406
407 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
408 void *data);
409
410 #define BLK_ALL_ZONES ((unsigned int)-1)
411 int blkdev_report_zones(struct block_device *bdev, sector_t sector,
412 unsigned int nr_zones, report_zones_cb cb, void *data);
413 int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op,
414 sector_t sectors, sector_t nr_sectors);
415 int blk_revalidate_disk_zones(struct gendisk *disk);
416
417 /*
418 * Independent access ranges: struct blk_independent_access_range describes
419 * a range of contiguous sectors that can be accessed using device command
420 * execution resources that are independent from the resources used for
421 * other access ranges. This is typically found with single-LUN multi-actuator
422 * HDDs where each access range is served by a different set of heads.
423 * The set of independent ranges supported by the device is defined using
424 * struct blk_independent_access_ranges. The independent ranges must not overlap
425 * and must include all sectors within the disk capacity (no sector holes
426 * allowed).
427 * For a device with multiple ranges, requests targeting sectors in different
428 * ranges can be executed in parallel. A request can straddle an access range
429 * boundary.
430 */
431 struct blk_independent_access_range {
432 struct kobject kobj;
433 sector_t sector;
434 sector_t nr_sectors;
435 };
436
437 struct blk_independent_access_ranges {
438 struct kobject kobj;
439 bool sysfs_registered;
440 unsigned int nr_ia_ranges;
441 struct blk_independent_access_range ia_range[];
442 };
443
444 struct request_queue {
445 /*
446 * The queue owner gets to use this for whatever they like.
447 * ll_rw_blk doesn't touch it.
448 */
449 void *queuedata;
450
451 struct elevator_queue *elevator;
452
453 const struct blk_mq_ops *mq_ops;
454
455 /* sw queues */
456 struct blk_mq_ctx __percpu *queue_ctx;
457
458 /*
459 * various queue flags, see QUEUE_* below
460 */
461 unsigned long queue_flags;
462
463 unsigned int rq_timeout;
464
465 unsigned int queue_depth;
466
467 refcount_t refs;
468
469 /* hw dispatch queues */
470 unsigned int nr_hw_queues;
471 struct xarray hctx_table;
472
473 struct percpu_ref q_usage_counter;
474
475 struct request *last_merge;
476
477 spinlock_t queue_lock;
478
479 int quiesce_depth;
480
481 struct gendisk *disk;
482
483 /*
484 * mq queue kobject
485 */
486 struct kobject *mq_kobj;
487
488 struct queue_limits limits;
489
490 #ifdef CONFIG_PM
491 struct device *dev;
492 enum rpm_status rpm_status;
493 #endif
494
495 /*
496 * Number of contexts that have called blk_set_pm_only(). If this
497 * counter is above zero then only RQF_PM requests are processed.
498 */
499 atomic_t pm_only;
500
501 struct blk_queue_stats *stats;
502 struct rq_qos *rq_qos;
503 struct mutex rq_qos_mutex;
504
505 /*
506 * ida allocated id for this queue. Used to index queues from
507 * ioctx.
508 */
509 int id;
510
511 /*
512 * queue settings
513 */
514 unsigned long nr_requests; /* Max # of requests */
515
516 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
517 struct blk_crypto_profile *crypto_profile;
518 struct kobject *crypto_kobject;
519 #endif
520
521 struct timer_list timeout;
522 struct work_struct timeout_work;
523
524 atomic_t nr_active_requests_shared_tags;
525
526 struct blk_mq_tags *sched_shared_tags;
527
528 struct list_head icq_list;
529 #ifdef CONFIG_BLK_CGROUP
530 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS);
531 struct blkcg_gq *root_blkg;
532 struct list_head blkg_list;
533 struct mutex blkcg_mutex;
534 #endif
535
536 int node;
537
538 spinlock_t requeue_lock;
539 struct list_head requeue_list;
540 struct delayed_work requeue_work;
541
542 #ifdef CONFIG_BLK_DEV_IO_TRACE
543 struct blk_trace __rcu *blk_trace;
544 #endif
545 /*
546 * for flush operations
547 */
548 struct blk_flush_queue *fq;
549 struct list_head flush_list;
550
551 struct mutex sysfs_lock;
552 struct mutex sysfs_dir_lock;
553 struct mutex limits_lock;
554
555 /*
556 * for reusing dead hctx instance in case of updating
557 * nr_hw_queues
558 */
559 struct list_head unused_hctx_list;
560 spinlock_t unused_hctx_lock;
561
562 int mq_freeze_depth;
563
564 #ifdef CONFIG_BLK_DEV_THROTTLING
565 /* Throttle data */
566 struct throtl_data *td;
567 #endif
568 struct rcu_head rcu_head;
569 wait_queue_head_t mq_freeze_wq;
570 /*
571 * Protect concurrent access to q_usage_counter by
572 * percpu_ref_kill() and percpu_ref_reinit().
573 */
574 struct mutex mq_freeze_lock;
575
576 struct blk_mq_tag_set *tag_set;
577 struct list_head tag_set_list;
578
579 struct dentry *debugfs_dir;
580 struct dentry *sched_debugfs_dir;
581 struct dentry *rqos_debugfs_dir;
582 /*
583 * Serializes all debugfs metadata operations using the above dentries.
584 */
585 struct mutex debugfs_mutex;
586
587 bool mq_sysfs_init_done;
588 };
589
590 /* Keep blk_queue_flag_name[] in sync with the definitions below */
591 enum {
592 QUEUE_FLAG_DYING, /* queue being torn down */
593 QUEUE_FLAG_NOMERGES, /* disable merge attempts */
594 QUEUE_FLAG_SAME_COMP, /* complete on same CPU-group */
595 QUEUE_FLAG_FAIL_IO, /* fake timeout */
596 QUEUE_FLAG_NOXMERGES, /* No extended merges */
597 QUEUE_FLAG_SAME_FORCE, /* force complete on same CPU */
598 QUEUE_FLAG_INIT_DONE, /* queue is initialized */
599 QUEUE_FLAG_STATS, /* track IO start and completion times */
600 QUEUE_FLAG_REGISTERED, /* queue has been registered to a disk */
601 QUEUE_FLAG_QUIESCED, /* queue has been quiesced */
602 QUEUE_FLAG_RQ_ALLOC_TIME, /* record rq->alloc_time_ns */
603 QUEUE_FLAG_HCTX_ACTIVE, /* at least one blk-mq hctx is active */
604 QUEUE_FLAG_SQ_SCHED, /* single queue style io dispatch */
605 QUEUE_FLAG_MAX
606 };
607
608 #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP)
609
610 void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
611 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
612
613 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
614 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
615 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
616 #define blk_queue_noxmerges(q) \
617 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
618 #define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL))
619 #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT)
620 #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX)
621 #define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA)
622 #ifdef CONFIG_BLK_RQ_ALLOC_TIME
623 #define blk_queue_rq_alloc_time(q) \
624 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
625 #else
626 #define blk_queue_rq_alloc_time(q) false
627 #endif
628
629 #define blk_noretry_request(rq) \
630 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
631 REQ_FAILFAST_DRIVER))
632 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
633 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only)
634 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
635 #define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags)
636 #define blk_queue_skip_tagset_quiesce(q) \
637 ((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE)
638
639 extern void blk_set_pm_only(struct request_queue *q);
640 extern void blk_clear_pm_only(struct request_queue *q);
641
642 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist)
643
644 #define dma_map_bvec(dev, bv, dir, attrs) \
645 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
646 (dir), (attrs))
647
648 static inline bool queue_is_mq(struct request_queue *q)
649 {
650 return q->mq_ops;
651 }
652
653 #ifdef CONFIG_PM
654 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
655 {
656 return q->rpm_status;
657 }
658 #else
659 static inline enum rpm_status queue_rpm_status(struct request_queue *q)
660 {
661 return RPM_ACTIVE;
662 }
663 #endif
664
665 static inline bool blk_queue_is_zoned(struct request_queue *q)
666 {
667 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) &&
668 (q->limits.features & BLK_FEAT_ZONED);
669 }
670
671 #ifdef CONFIG_BLK_DEV_ZONED
672 static inline unsigned int disk_nr_zones(struct gendisk *disk)
673 {
674 return disk->nr_zones;
675 }
676 bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs);
677 #else /* CONFIG_BLK_DEV_ZONED */
678 static inline unsigned int disk_nr_zones(struct gendisk *disk)
679 {
680 return 0;
681 }
682 static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs)
683 {
684 return false;
685 }
686 #endif /* CONFIG_BLK_DEV_ZONED */
687
688 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector)
689 {
690 if (!blk_queue_is_zoned(disk->queue))
691 return 0;
692 return sector >> ilog2(disk->queue->limits.chunk_sectors);
693 }
694
695 static inline unsigned int bdev_nr_zones(struct block_device *bdev)
696 {
697 return disk_nr_zones(bdev->bd_disk);
698 }
699
700 static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
701 {
702 return bdev->bd_disk->queue->limits.max_open_zones;
703 }
704
705 static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
706 {
707 return bdev->bd_disk->queue->limits.max_active_zones;
708 }
709
710 static inline unsigned int blk_queue_depth(struct request_queue *q)
711 {
712 if (q->queue_depth)
713 return q->queue_depth;
714
715 return q->nr_requests;
716 }
717
718 /*
719 * default timeout for SG_IO if none specified
720 */
721 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ)
722 #define BLK_MIN_SG_TIMEOUT (7 * HZ)
723
724 /* This should not be used directly - use rq_for_each_segment */
725 #define for_each_bio(_bio) \
726 for (; _bio; _bio = _bio->bi_next)
727
728 int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
729 const struct attribute_group **groups);
730 static inline int __must_check add_disk(struct gendisk *disk)
731 {
732 return device_add_disk(NULL, disk, NULL);
733 }
734 void del_gendisk(struct gendisk *gp);
735 void invalidate_disk(struct gendisk *disk);
736 void set_disk_ro(struct gendisk *disk, bool read_only);
737 void disk_uevent(struct gendisk *disk, enum kobject_action action);
738
739 static inline u8 bdev_partno(const struct block_device *bdev)
740 {
741 return atomic_read(&bdev->__bd_flags) & BD_PARTNO;
742 }
743
744 static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag)
745 {
746 return atomic_read(&bdev->__bd_flags) & flag;
747 }
748
749 static inline void bdev_set_flag(struct block_device *bdev, unsigned flag)
750 {
751 atomic_or(flag, &bdev->__bd_flags);
752 }
753
754 static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag)
755 {
756 atomic_andnot(flag, &bdev->__bd_flags);
757 }
758
759 static inline int get_disk_ro(struct gendisk *disk)
760 {
761 return bdev_test_flag(disk->part0, BD_READ_ONLY) ||
762 test_bit(GD_READ_ONLY, &disk->state);
763 }
764
765 static inline int bdev_read_only(struct block_device *bdev)
766 {
767 return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk);
768 }
769
770 bool set_capacity_and_notify(struct gendisk *disk, sector_t size);
771 void disk_force_media_change(struct gendisk *disk);
772 void bdev_mark_dead(struct block_device *bdev, bool surprise);
773
774 void add_disk_randomness(struct gendisk *disk) __latent_entropy;
775 void rand_initialize_disk(struct gendisk *disk);
776
777 static inline sector_t get_start_sect(struct block_device *bdev)
778 {
779 return bdev->bd_start_sect;
780 }
781
782 static inline sector_t bdev_nr_sectors(struct block_device *bdev)
783 {
784 return bdev->bd_nr_sectors;
785 }
786
787 static inline loff_t bdev_nr_bytes(struct block_device *bdev)
788 {
789 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT;
790 }
791
792 static inline sector_t get_capacity(struct gendisk *disk)
793 {
794 return bdev_nr_sectors(disk->part0);
795 }
796
797 static inline u64 sb_bdev_nr_blocks(struct super_block *sb)
798 {
799 return bdev_nr_sectors(sb->s_bdev) >>
800 (sb->s_blocksize_bits - SECTOR_SHIFT);
801 }
802
803 int bdev_disk_changed(struct gendisk *disk, bool invalidate);
804
805 void put_disk(struct gendisk *disk);
806 struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node,
807 struct lock_class_key *lkclass);
808
809 /**
810 * blk_alloc_disk - allocate a gendisk structure
811 * @lim: queue limits to be used for this disk.
812 * @node_id: numa node to allocate on
813 *
814 * Allocate and pre-initialize a gendisk structure for use with BIO based
815 * drivers.
816 *
817 * Returns an ERR_PTR on error, else the allocated disk.
818 *
819 * Context: can sleep
820 */
821 #define blk_alloc_disk(lim, node_id) \
822 ({ \
823 static struct lock_class_key __key; \
824 \
825 __blk_alloc_disk(lim, node_id, &__key); \
826 })
827
828 int __register_blkdev(unsigned int major, const char *name,
829 void (*probe)(dev_t devt));
830 #define register_blkdev(major, name) \
831 __register_blkdev(major, name, NULL)
832 void unregister_blkdev(unsigned int major, const char *name);
833
834 bool disk_check_media_change(struct gendisk *disk);
835 void set_capacity(struct gendisk *disk, sector_t size);
836
837 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
838 int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk);
839 void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk);
840 #else
841 static inline int bd_link_disk_holder(struct block_device *bdev,
842 struct gendisk *disk)
843 {
844 return 0;
845 }
846 static inline void bd_unlink_disk_holder(struct block_device *bdev,
847 struct gendisk *disk)
848 {
849 }
850 #endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */
851
852 dev_t part_devt(struct gendisk *disk, u8 partno);
853 void inc_diskseq(struct gendisk *disk);
854 void blk_request_module(dev_t devt);
855
856 extern int blk_register_queue(struct gendisk *disk);
857 extern void blk_unregister_queue(struct gendisk *disk);
858 void submit_bio_noacct(struct bio *bio);
859 struct bio *bio_split_to_limits(struct bio *bio);
860
861 extern int blk_lld_busy(struct request_queue *q);
862 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
863 extern void blk_queue_exit(struct request_queue *q);
864 extern void blk_sync_queue(struct request_queue *q);
865
866 /* Helper to convert REQ_OP_XXX to its string format XXX */
867 extern const char *blk_op_str(enum req_op op);
868
869 int blk_status_to_errno(blk_status_t status);
870 blk_status_t errno_to_blk_status(int errno);
871 const char *blk_status_to_str(blk_status_t status);
872
873 /* only poll the hardware once, don't continue until a completion was found */
874 #define BLK_POLL_ONESHOT (1 << 0)
875 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags);
876 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob,
877 unsigned int flags);
878
879 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
880 {
881 return bdev->bd_queue; /* this is never NULL */
882 }
883
884 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
885 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
886
887 static inline unsigned int bio_zone_no(struct bio *bio)
888 {
889 return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector);
890 }
891
892 static inline bool bio_straddles_zones(struct bio *bio)
893 {
894 return bio_sectors(bio) &&
895 bio_zone_no(bio) !=
896 disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1);
897 }
898
899 /*
900 * Return how much within the boundary is left to be used for I/O at a given
901 * offset.
902 */
903 static inline unsigned int blk_boundary_sectors_left(sector_t offset,
904 unsigned int boundary_sectors)
905 {
906 if (unlikely(!is_power_of_2(boundary_sectors)))
907 return boundary_sectors - sector_div(offset, boundary_sectors);
908 return boundary_sectors - (offset & (boundary_sectors - 1));
909 }
910
911 /**
912 * queue_limits_start_update - start an atomic update of queue limits
913 * @q: queue to update
914 *
915 * This functions starts an atomic update of the queue limits. It takes a lock
916 * to prevent other updates and returns a snapshot of the current limits that
917 * the caller can modify. The caller must call queue_limits_commit_update()
918 * to finish the update.
919 *
920 * Context: process context. The caller must have frozen the queue or ensured
921 * that there is outstanding I/O by other means.
922 */
923 static inline struct queue_limits
924 queue_limits_start_update(struct request_queue *q)
925 {
926 mutex_lock(&q->limits_lock);
927 return q->limits;
928 }
929 int queue_limits_commit_update(struct request_queue *q,
930 struct queue_limits *lim);
931 int queue_limits_set(struct request_queue *q, struct queue_limits *lim);
932
933 /**
934 * queue_limits_cancel_update - cancel an atomic update of queue limits
935 * @q: queue to update
936 *
937 * This functions cancels an atomic update of the queue limits started by
938 * queue_limits_start_update() and should be used when an error occurs after
939 * starting update.
940 */
941 static inline void queue_limits_cancel_update(struct request_queue *q)
942 {
943 mutex_unlock(&q->limits_lock);
944 }
945
946 /*
947 * These helpers are for drivers that have sloppy feature negotiation and might
948 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O
949 * completion handler when the device returned an indicator that the respective
950 * feature is not actually supported. They are racy and the driver needs to
951 * cope with that. Try to avoid this scheme if you can.
952 */
953 static inline void blk_queue_disable_discard(struct request_queue *q)
954 {
955 q->limits.max_discard_sectors = 0;
956 }
957
958 static inline void blk_queue_disable_secure_erase(struct request_queue *q)
959 {
960 q->limits.max_secure_erase_sectors = 0;
961 }
962
963 static inline void blk_queue_disable_write_zeroes(struct request_queue *q)
964 {
965 q->limits.max_write_zeroes_sectors = 0;
966 }
967
968 /*
969 * Access functions for manipulating queue properties
970 */
971 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min);
972 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt);
973 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
974 extern void blk_set_stacking_limits(struct queue_limits *lim);
975 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
976 sector_t offset);
977 void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev,
978 sector_t offset, const char *pfx);
979 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
980
981 struct blk_independent_access_ranges *
982 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges);
983 void disk_set_independent_access_ranges(struct gendisk *disk,
984 struct blk_independent_access_ranges *iars);
985
986 bool __must_check blk_get_queue(struct request_queue *);
987 extern void blk_put_queue(struct request_queue *);
988
989 void blk_mark_disk_dead(struct gendisk *disk);
990
991 #ifdef CONFIG_BLOCK
992 /*
993 * blk_plug permits building a queue of related requests by holding the I/O
994 * fragments for a short period. This allows merging of sequential requests
995 * into single larger request. As the requests are moved from a per-task list to
996 * the device's request_queue in a batch, this results in improved scalability
997 * as the lock contention for request_queue lock is reduced.
998 *
999 * It is ok not to disable preemption when adding the request to the plug list
1000 * or when attempting a merge. For details, please see schedule() where
1001 * blk_flush_plug() is called.
1002 */
1003 struct blk_plug {
1004 struct request *mq_list; /* blk-mq requests */
1005
1006 /* if ios_left is > 1, we can batch tag/rq allocations */
1007 struct request *cached_rq;
1008 u64 cur_ktime;
1009 unsigned short nr_ios;
1010
1011 unsigned short rq_count;
1012
1013 bool multiple_queues;
1014 bool has_elevator;
1015
1016 struct list_head cb_list; /* md requires an unplug callback */
1017 };
1018
1019 struct blk_plug_cb;
1020 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1021 struct blk_plug_cb {
1022 struct list_head list;
1023 blk_plug_cb_fn callback;
1024 void *data;
1025 };
1026 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1027 void *data, int size);
1028 extern void blk_start_plug(struct blk_plug *);
1029 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short);
1030 extern void blk_finish_plug(struct blk_plug *);
1031
1032 void __blk_flush_plug(struct blk_plug *plug, bool from_schedule);
1033 static inline void blk_flush_plug(struct blk_plug *plug, bool async)
1034 {
1035 if (plug)
1036 __blk_flush_plug(plug, async);
1037 }
1038
1039 /*
1040 * tsk == current here
1041 */
1042 static inline void blk_plug_invalidate_ts(struct task_struct *tsk)
1043 {
1044 struct blk_plug *plug = tsk->plug;
1045
1046 if (plug)
1047 plug->cur_ktime = 0;
1048 current->flags &= ~PF_BLOCK_TS;
1049 }
1050
1051 int blkdev_issue_flush(struct block_device *bdev);
1052 long nr_blockdev_pages(void);
1053 #else /* CONFIG_BLOCK */
1054 struct blk_plug {
1055 };
1056
1057 static inline void blk_start_plug_nr_ios(struct blk_plug *plug,
1058 unsigned short nr_ios)
1059 {
1060 }
1061
1062 static inline void blk_start_plug(struct blk_plug *plug)
1063 {
1064 }
1065
1066 static inline void blk_finish_plug(struct blk_plug *plug)
1067 {
1068 }
1069
1070 static inline void blk_flush_plug(struct blk_plug *plug, bool async)
1071 {
1072 }
1073
1074 static inline void blk_plug_invalidate_ts(struct task_struct *tsk)
1075 {
1076 }
1077
1078 static inline int blkdev_issue_flush(struct block_device *bdev)
1079 {
1080 return 0;
1081 }
1082
1083 static inline long nr_blockdev_pages(void)
1084 {
1085 return 0;
1086 }
1087 #endif /* CONFIG_BLOCK */
1088
1089 extern void blk_io_schedule(void);
1090
1091 int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1092 sector_t nr_sects, gfp_t gfp_mask);
1093 int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1094 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop);
1095 int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector,
1096 sector_t nr_sects, gfp_t gfp);
1097
1098 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */
1099 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */
1100 #define BLKDEV_ZERO_KILLABLE (1 << 2) /* interruptible by fatal signals */
1101
1102 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1103 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1104 unsigned flags);
1105 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1106 sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1107
1108 static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1109 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1110 {
1111 return blkdev_issue_discard(sb->s_bdev,
1112 block << (sb->s_blocksize_bits -
1113 SECTOR_SHIFT),
1114 nr_blocks << (sb->s_blocksize_bits -
1115 SECTOR_SHIFT),
1116 gfp_mask);
1117 }
1118 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1119 sector_t nr_blocks, gfp_t gfp_mask)
1120 {
1121 return blkdev_issue_zeroout(sb->s_bdev,
1122 block << (sb->s_blocksize_bits -
1123 SECTOR_SHIFT),
1124 nr_blocks << (sb->s_blocksize_bits -
1125 SECTOR_SHIFT),
1126 gfp_mask, 0);
1127 }
1128
1129 static inline bool bdev_is_partition(struct block_device *bdev)
1130 {
1131 return bdev_partno(bdev) != 0;
1132 }
1133
1134 enum blk_default_limits {
1135 BLK_MAX_SEGMENTS = 128,
1136 BLK_SAFE_MAX_SECTORS = 255,
1137 BLK_MAX_SEGMENT_SIZE = 65536,
1138 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL,
1139 };
1140
1141 /*
1142 * Default upper limit for the software max_sectors limit used for
1143 * regular file system I/O. This can be increased through sysfs.
1144 *
1145 * Not to be confused with the max_hw_sector limit that is entirely
1146 * controlled by the driver, usually based on hardware limits.
1147 */
1148 #define BLK_DEF_MAX_SECTORS_CAP 2560u
1149
1150 static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1151 {
1152 return q->limits.seg_boundary_mask;
1153 }
1154
1155 static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1156 {
1157 return q->limits.virt_boundary_mask;
1158 }
1159
1160 static inline unsigned int queue_max_sectors(const struct request_queue *q)
1161 {
1162 return q->limits.max_sectors;
1163 }
1164
1165 static inline unsigned int queue_max_bytes(struct request_queue *q)
1166 {
1167 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9;
1168 }
1169
1170 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1171 {
1172 return q->limits.max_hw_sectors;
1173 }
1174
1175 static inline unsigned short queue_max_segments(const struct request_queue *q)
1176 {
1177 return q->limits.max_segments;
1178 }
1179
1180 static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1181 {
1182 return q->limits.max_discard_segments;
1183 }
1184
1185 static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1186 {
1187 return q->limits.max_segment_size;
1188 }
1189
1190 static inline unsigned int queue_limits_max_zone_append_sectors(struct queue_limits *l)
1191 {
1192 unsigned int max_sectors = min(l->chunk_sectors, l->max_hw_sectors);
1193
1194 return min_not_zero(l->max_zone_append_sectors, max_sectors);
1195 }
1196
1197 static inline unsigned int queue_max_zone_append_sectors(struct request_queue *q)
1198 {
1199 if (!blk_queue_is_zoned(q))
1200 return 0;
1201
1202 return queue_limits_max_zone_append_sectors(&q->limits);
1203 }
1204
1205 static inline bool queue_emulates_zone_append(struct request_queue *q)
1206 {
1207 return blk_queue_is_zoned(q) && !q->limits.max_zone_append_sectors;
1208 }
1209
1210 static inline bool bdev_emulates_zone_append(struct block_device *bdev)
1211 {
1212 return queue_emulates_zone_append(bdev_get_queue(bdev));
1213 }
1214
1215 static inline unsigned int
1216 bdev_max_zone_append_sectors(struct block_device *bdev)
1217 {
1218 return queue_max_zone_append_sectors(bdev_get_queue(bdev));
1219 }
1220
1221 static inline unsigned int bdev_max_segments(struct block_device *bdev)
1222 {
1223 return queue_max_segments(bdev_get_queue(bdev));
1224 }
1225
1226 static inline unsigned queue_logical_block_size(const struct request_queue *q)
1227 {
1228 return q->limits.logical_block_size;
1229 }
1230
1231 static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1232 {
1233 return queue_logical_block_size(bdev_get_queue(bdev));
1234 }
1235
1236 static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1237 {
1238 return q->limits.physical_block_size;
1239 }
1240
1241 static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1242 {
1243 return queue_physical_block_size(bdev_get_queue(bdev));
1244 }
1245
1246 static inline unsigned int queue_io_min(const struct request_queue *q)
1247 {
1248 return q->limits.io_min;
1249 }
1250
1251 static inline int bdev_io_min(struct block_device *bdev)
1252 {
1253 return queue_io_min(bdev_get_queue(bdev));
1254 }
1255
1256 static inline unsigned int queue_io_opt(const struct request_queue *q)
1257 {
1258 return q->limits.io_opt;
1259 }
1260
1261 static inline int bdev_io_opt(struct block_device *bdev)
1262 {
1263 return queue_io_opt(bdev_get_queue(bdev));
1264 }
1265
1266 static inline unsigned int
1267 queue_zone_write_granularity(const struct request_queue *q)
1268 {
1269 return q->limits.zone_write_granularity;
1270 }
1271
1272 static inline unsigned int
1273 bdev_zone_write_granularity(struct block_device *bdev)
1274 {
1275 return queue_zone_write_granularity(bdev_get_queue(bdev));
1276 }
1277
1278 int bdev_alignment_offset(struct block_device *bdev);
1279 unsigned int bdev_discard_alignment(struct block_device *bdev);
1280
1281 static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev)
1282 {
1283 return bdev_get_queue(bdev)->limits.max_discard_sectors;
1284 }
1285
1286 static inline unsigned int bdev_discard_granularity(struct block_device *bdev)
1287 {
1288 return bdev_get_queue(bdev)->limits.discard_granularity;
1289 }
1290
1291 static inline unsigned int
1292 bdev_max_secure_erase_sectors(struct block_device *bdev)
1293 {
1294 return bdev_get_queue(bdev)->limits.max_secure_erase_sectors;
1295 }
1296
1297 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1298 {
1299 return bdev_get_queue(bdev)->limits.max_write_zeroes_sectors;
1300 }
1301
1302 static inline bool bdev_nonrot(struct block_device *bdev)
1303 {
1304 return blk_queue_nonrot(bdev_get_queue(bdev));
1305 }
1306
1307 static inline bool bdev_synchronous(struct block_device *bdev)
1308 {
1309 return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS;
1310 }
1311
1312 static inline bool bdev_stable_writes(struct block_device *bdev)
1313 {
1314 struct request_queue *q = bdev_get_queue(bdev);
1315
1316 if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) &&
1317 q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE)
1318 return true;
1319 return q->limits.features & BLK_FEAT_STABLE_WRITES;
1320 }
1321
1322 static inline bool blk_queue_write_cache(struct request_queue *q)
1323 {
1324 return (q->limits.features & BLK_FEAT_WRITE_CACHE) &&
1325 !(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED);
1326 }
1327
1328 static inline bool bdev_write_cache(struct block_device *bdev)
1329 {
1330 return blk_queue_write_cache(bdev_get_queue(bdev));
1331 }
1332
1333 static inline bool bdev_fua(struct block_device *bdev)
1334 {
1335 return bdev_get_queue(bdev)->limits.features & BLK_FEAT_FUA;
1336 }
1337
1338 static inline bool bdev_nowait(struct block_device *bdev)
1339 {
1340 return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT;
1341 }
1342
1343 static inline bool bdev_is_zoned(struct block_device *bdev)
1344 {
1345 return blk_queue_is_zoned(bdev_get_queue(bdev));
1346 }
1347
1348 static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec)
1349 {
1350 return disk_zone_no(bdev->bd_disk, sec);
1351 }
1352
1353 static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1354 {
1355 struct request_queue *q = bdev_get_queue(bdev);
1356
1357 if (!blk_queue_is_zoned(q))
1358 return 0;
1359 return q->limits.chunk_sectors;
1360 }
1361
1362 static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev,
1363 sector_t sector)
1364 {
1365 return sector & (bdev_zone_sectors(bdev) - 1);
1366 }
1367
1368 static inline sector_t bio_offset_from_zone_start(struct bio *bio)
1369 {
1370 return bdev_offset_from_zone_start(bio->bi_bdev,
1371 bio->bi_iter.bi_sector);
1372 }
1373
1374 static inline bool bdev_is_zone_start(struct block_device *bdev,
1375 sector_t sector)
1376 {
1377 return bdev_offset_from_zone_start(bdev, sector) == 0;
1378 }
1379
1380 static inline int queue_dma_alignment(const struct request_queue *q)
1381 {
1382 return q->limits.dma_alignment;
1383 }
1384
1385 static inline unsigned int
1386 queue_atomic_write_unit_max_bytes(const struct request_queue *q)
1387 {
1388 return q->limits.atomic_write_unit_max;
1389 }
1390
1391 static inline unsigned int
1392 queue_atomic_write_unit_min_bytes(const struct request_queue *q)
1393 {
1394 return q->limits.atomic_write_unit_min;
1395 }
1396
1397 static inline unsigned int
1398 queue_atomic_write_boundary_bytes(const struct request_queue *q)
1399 {
1400 return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT;
1401 }
1402
1403 static inline unsigned int
1404 queue_atomic_write_max_bytes(const struct request_queue *q)
1405 {
1406 return q->limits.atomic_write_max_sectors << SECTOR_SHIFT;
1407 }
1408
1409 static inline unsigned int bdev_dma_alignment(struct block_device *bdev)
1410 {
1411 return queue_dma_alignment(bdev_get_queue(bdev));
1412 }
1413
1414 static inline bool bdev_iter_is_aligned(struct block_device *bdev,
1415 struct iov_iter *iter)
1416 {
1417 return iov_iter_is_aligned(iter, bdev_dma_alignment(bdev),
1418 bdev_logical_block_size(bdev) - 1);
1419 }
1420
1421 static inline int blk_lim_dma_alignment_and_pad(struct queue_limits *lim)
1422 {
1423 return lim->dma_alignment | lim->dma_pad_mask;
1424 }
1425
1426 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr,
1427 unsigned int len)
1428 {
1429 unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits);
1430
1431 return !(addr & alignment) && !(len & alignment);
1432 }
1433
1434 /* assumes size > 256 */
1435 static inline unsigned int blksize_bits(unsigned int size)
1436 {
1437 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT;
1438 }
1439
1440 int kblockd_schedule_work(struct work_struct *work);
1441 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1442
1443 #define MODULE_ALIAS_BLOCKDEV(major,minor) \
1444 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1445 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1446 MODULE_ALIAS("block-major-" __stringify(major) "-*")
1447
1448 #ifdef CONFIG_BLK_INLINE_ENCRYPTION
1449
1450 bool blk_crypto_register(struct blk_crypto_profile *profile,
1451 struct request_queue *q);
1452
1453 #else /* CONFIG_BLK_INLINE_ENCRYPTION */
1454
1455 static inline bool blk_crypto_register(struct blk_crypto_profile *profile,
1456 struct request_queue *q)
1457 {
1458 return true;
1459 }
1460
1461 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1462
1463 enum blk_unique_id {
1464 /* these match the Designator Types specified in SPC */
1465 BLK_UID_T10 = 1,
1466 BLK_UID_EUI64 = 2,
1467 BLK_UID_NAA = 3,
1468 };
1469
1470 struct block_device_operations {
1471 void (*submit_bio)(struct bio *bio);
1472 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob,
1473 unsigned int flags);
1474 int (*open)(struct gendisk *disk, blk_mode_t mode);
1475 void (*release)(struct gendisk *disk);
1476 int (*ioctl)(struct block_device *bdev, blk_mode_t mode,
1477 unsigned cmd, unsigned long arg);
1478 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode,
1479 unsigned cmd, unsigned long arg);
1480 unsigned int (*check_events) (struct gendisk *disk,
1481 unsigned int clearing);
1482 void (*unlock_native_capacity) (struct gendisk *);
1483 int (*getgeo)(struct block_device *, struct hd_geometry *);
1484 int (*set_read_only)(struct block_device *bdev, bool ro);
1485 void (*free_disk)(struct gendisk *disk);
1486 /* this callback is with swap_lock and sometimes page table lock held */
1487 void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1488 int (*report_zones)(struct gendisk *, sector_t sector,
1489 unsigned int nr_zones, report_zones_cb cb, void *data);
1490 char *(*devnode)(struct gendisk *disk, umode_t *mode);
1491 /* returns the length of the identifier or a negative errno: */
1492 int (*get_unique_id)(struct gendisk *disk, u8 id[16],
1493 enum blk_unique_id id_type);
1494 struct module *owner;
1495 const struct pr_ops *pr_ops;
1496
1497 /*
1498 * Special callback for probing GPT entry at a given sector.
1499 * Needed by Android devices, used by GPT scanner and MMC blk
1500 * driver.
1501 */
1502 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector);
1503 };
1504
1505 #ifdef CONFIG_COMPAT
1506 extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t,
1507 unsigned int, unsigned long);
1508 #else
1509 #define blkdev_compat_ptr_ioctl NULL
1510 #endif
1511
1512 static inline void blk_wake_io_task(struct task_struct *waiter)
1513 {
1514 /*
1515 * If we're polling, the task itself is doing the completions. For
1516 * that case, we don't need to signal a wakeup, it's enough to just
1517 * mark us as RUNNING.
1518 */
1519 if (waiter == current)
1520 __set_current_state(TASK_RUNNING);
1521 else
1522 wake_up_process(waiter);
1523 }
1524
1525 unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op,
1526 unsigned long start_time);
1527 void bdev_end_io_acct(struct block_device *bdev, enum req_op op,
1528 unsigned int sectors, unsigned long start_time);
1529
1530 unsigned long bio_start_io_acct(struct bio *bio);
1531 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1532 struct block_device *orig_bdev);
1533
1534 /**
1535 * bio_end_io_acct - end I/O accounting for bio based drivers
1536 * @bio: bio to end account for
1537 * @start_time: start time returned by bio_start_io_acct()
1538 */
1539 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1540 {
1541 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1542 }
1543
1544 int bdev_read_only(struct block_device *bdev);
1545 int set_blocksize(struct file *file, int size);
1546
1547 int lookup_bdev(const char *pathname, dev_t *dev);
1548
1549 void blkdev_show(struct seq_file *seqf, off_t offset);
1550
1551 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */
1552 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */
1553 #ifdef CONFIG_BLOCK
1554 #define BLKDEV_MAJOR_MAX 512
1555 #else
1556 #define BLKDEV_MAJOR_MAX 0
1557 #endif
1558
1559 struct blk_holder_ops {
1560 void (*mark_dead)(struct block_device *bdev, bool surprise);
1561
1562 /*
1563 * Sync the file system mounted on the block device.
1564 */
1565 void (*sync)(struct block_device *bdev);
1566
1567 /*
1568 * Freeze the file system mounted on the block device.
1569 */
1570 int (*freeze)(struct block_device *bdev);
1571
1572 /*
1573 * Thaw the file system mounted on the block device.
1574 */
1575 int (*thaw)(struct block_device *bdev);
1576 };
1577
1578 /*
1579 * For filesystems using @fs_holder_ops, the @holder argument passed to
1580 * helpers used to open and claim block devices via
1581 * bd_prepare_to_claim() must point to a superblock.
1582 */
1583 extern const struct blk_holder_ops fs_holder_ops;
1584
1585 /*
1586 * Return the correct open flags for blkdev_get_by_* for super block flags
1587 * as stored in sb->s_flags.
1588 */
1589 #define sb_open_mode(flags) \
1590 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \
1591 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE))
1592
1593 struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder,
1594 const struct blk_holder_ops *hops);
1595 struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode,
1596 void *holder, const struct blk_holder_ops *hops);
1597 int bd_prepare_to_claim(struct block_device *bdev, void *holder,
1598 const struct blk_holder_ops *hops);
1599 void bd_abort_claiming(struct block_device *bdev, void *holder);
1600
1601 /* just for blk-cgroup, don't use elsewhere */
1602 struct block_device *blkdev_get_no_open(dev_t dev);
1603 void blkdev_put_no_open(struct block_device *bdev);
1604
1605 struct block_device *I_BDEV(struct inode *inode);
1606 struct block_device *file_bdev(struct file *bdev_file);
1607 bool disk_live(struct gendisk *disk);
1608 unsigned int block_size(struct block_device *bdev);
1609
1610 #ifdef CONFIG_BLOCK
1611 void invalidate_bdev(struct block_device *bdev);
1612 int sync_blockdev(struct block_device *bdev);
1613 int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend);
1614 int sync_blockdev_nowait(struct block_device *bdev);
1615 void sync_bdevs(bool wait);
1616 void bdev_statx(struct path *, struct kstat *, u32);
1617 void printk_all_partitions(void);
1618 int __init early_lookup_bdev(const char *pathname, dev_t *dev);
1619 #else
1620 static inline void invalidate_bdev(struct block_device *bdev)
1621 {
1622 }
1623 static inline int sync_blockdev(struct block_device *bdev)
1624 {
1625 return 0;
1626 }
1627 static inline int sync_blockdev_nowait(struct block_device *bdev)
1628 {
1629 return 0;
1630 }
1631 static inline void sync_bdevs(bool wait)
1632 {
1633 }
1634 static inline void bdev_statx(struct path *path, struct kstat *stat,
1635 u32 request_mask)
1636 {
1637 }
1638 static inline void printk_all_partitions(void)
1639 {
1640 }
1641 static inline int early_lookup_bdev(const char *pathname, dev_t *dev)
1642 {
1643 return -EINVAL;
1644 }
1645 #endif /* CONFIG_BLOCK */
1646
1647 int bdev_freeze(struct block_device *bdev);
1648 int bdev_thaw(struct block_device *bdev);
1649 void bdev_fput(struct file *bdev_file);
1650
1651 struct io_comp_batch {
1652 struct request *req_list;
1653 bool need_ts;
1654 void (*complete)(struct io_comp_batch *);
1655 };
1656
1657 static inline bool bdev_can_atomic_write(struct block_device *bdev)
1658 {
1659 struct request_queue *bd_queue = bdev->bd_queue;
1660 struct queue_limits *limits = &bd_queue->limits;
1661
1662 if (!limits->atomic_write_unit_min)
1663 return false;
1664
1665 if (bdev_is_partition(bdev)) {
1666 sector_t bd_start_sect = bdev->bd_start_sect;
1667 unsigned int alignment =
1668 max(limits->atomic_write_unit_min,
1669 limits->atomic_write_hw_boundary);
1670
1671 if (!IS_ALIGNED(bd_start_sect, alignment >> SECTOR_SHIFT))
1672 return false;
1673 }
1674
1675 return true;
1676 }
1677
1678 #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { }
1679
1680 #endif /* _LINUX_BLKDEV_H */
1681
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.