blkdev.h

   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#include <linux/lockdep.h>
  29
  30struct module;
  31struct request_queue;
  32struct elevator_queue;
  33struct blk_trace;
  34struct request;
  35struct sg_io_hdr;
  36struct blkcg_gq;
  37struct blk_flush_queue;
  38struct kiocb;
  39struct pr_ops;
  40struct rq_qos;
  41struct blk_report_zones_args;
  42struct blk_queue_stats;
  43struct blk_stat_callback;
  44struct blk_crypto_profile;
  45
  46extern const struct device_type disk_type;
  47extern const struct device_type part_type;
  48extern const struct class block_class;
  49
  50/*
  51 * Maximum number of blkcg policies allowed to be registered concurrently.
  52 * Defined here to simplify include dependency.
  53 */
  54#define BLKCG_MAX_POLS		6
  55
  56#define DISK_MAX_PARTS			256
  57#define DISK_NAME_LEN			32
  58
  59#define PARTITION_META_INFO_VOLNAMELTH	64
  60/*
  61 * Enough for the string representation of any kind of UUID plus NULL.
  62 * EFI UUID is 36 characters. MSDOS UUID is 11 characters.
  63 */
  64#define PARTITION_META_INFO_UUIDLTH	(UUID_STRING_LEN + 1)
  65
  66struct partition_meta_info {
  67	char uuid[PARTITION_META_INFO_UUIDLTH];
  68	u8 volname[PARTITION_META_INFO_VOLNAMELTH];
  69};
  70
  71/**
  72 * DOC: genhd capability flags
  73 *
  74 * ``GENHD_FL_REMOVABLE``: indicates that the block device gives access to
  75 * removable media.  When set, the device remains present even when media is not
  76 * inserted.  Shall not be set for devices which are removed entirely when the
  77 * media is removed.
  78 *
  79 * ``GENHD_FL_HIDDEN``: the block device is hidden; it doesn't produce events,
  80 * doesn't appear in sysfs, and can't be opened from userspace or using
  81 * blkdev_get*. Used for the underlying components of multipath devices.
  82 *
  83 * ``GENHD_FL_NO_PART``: partition support is disabled.  The kernel will not
  84 * scan for partitions from add_disk, and users can't add partitions manually.
  85 *
  86 */
  87enum {
  88	GENHD_FL_REMOVABLE			= 1 << 0,
  89	GENHD_FL_HIDDEN				= 1 << 1,
  90	GENHD_FL_NO_PART			= 1 << 2,
  91};
  92
  93enum {
  94	DISK_EVENT_MEDIA_CHANGE			= 1 << 0, /* media changed */
  95	DISK_EVENT_EJECT_REQUEST		= 1 << 1, /* eject requested */
  96};
  97
  98enum {
  99	/* Poll even if events_poll_msecs is unset */
 100	DISK_EVENT_FLAG_POLL			= 1 << 0,
 101	/* Forward events to udev */
 102	DISK_EVENT_FLAG_UEVENT			= 1 << 1,
 103	/* Block event polling when open for exclusive write */
 104	DISK_EVENT_FLAG_BLOCK_ON_EXCL_WRITE	= 1 << 2,
 105};
 106
 107struct disk_events;
 108struct badblocks;
 109
 110enum blk_integrity_checksum {
 111	BLK_INTEGRITY_CSUM_NONE		= 0,
 112	BLK_INTEGRITY_CSUM_IP		= 1,
 113	BLK_INTEGRITY_CSUM_CRC		= 2,
 114	BLK_INTEGRITY_CSUM_CRC64	= 3,
 115} __packed ;
 116
 117struct blk_integrity {
 118	unsigned char				flags;
 119	enum blk_integrity_checksum		csum_type;
 120	unsigned char				metadata_size;
 121	unsigned char				pi_offset;
 122	unsigned char				interval_exp;
 123	unsigned char				tag_size;
 124	unsigned char				pi_tuple_size;
 125};
 126
 127typedef unsigned int __bitwise blk_mode_t;
 128
 129/* open for reading */
 130#define BLK_OPEN_READ		((__force blk_mode_t)(1 << 0))
 131/* open for writing */
 132#define BLK_OPEN_WRITE		((__force blk_mode_t)(1 << 1))
 133/* open exclusively (vs other exclusive openers */
 134#define BLK_OPEN_EXCL		((__force blk_mode_t)(1 << 2))
 135/* opened with O_NDELAY */
 136#define BLK_OPEN_NDELAY		((__force blk_mode_t)(1 << 3))
 137/* open for "writes" only for ioctls (specialy hack for floppy.c) */
 138#define BLK_OPEN_WRITE_IOCTL	((__force blk_mode_t)(1 << 4))
 139/* open is exclusive wrt all other BLK_OPEN_WRITE opens to the device */
 140#define BLK_OPEN_RESTRICT_WRITES	((__force blk_mode_t)(1 << 5))
 141/* return partition scanning errors */
 142#define BLK_OPEN_STRICT_SCAN	((__force blk_mode_t)(1 << 6))
 143
 144struct gendisk {
 145	/*
 146	 * major/first_minor/minors should not be set by any new driver, the
 147	 * block core will take care of allocating them automatically.
 148	 */
 149	int major;
 150	int first_minor;
 151	int minors;
 152
 153	char disk_name[DISK_NAME_LEN];	/* name of major driver */
 154
 155	unsigned short events;		/* supported events */
 156	unsigned short event_flags;	/* flags related to event processing */
 157
 158	struct xarray part_tbl;
 159	struct block_device *part0;
 160
 161	const struct block_device_operations *fops;
 162	struct request_queue *queue;
 163	void *private_data;
 164
 165	struct bio_set bio_split;
 166
 167	int flags;
 168	unsigned long state;
 169#define GD_NEED_PART_SCAN		0
 170#define GD_READ_ONLY			1
 171#define GD_DEAD				2
 172#define GD_NATIVE_CAPACITY		3
 173#define GD_ADDED			4
 174#define GD_SUPPRESS_PART_SCAN		5
 175#define GD_OWNS_QUEUE			6
 176#define GD_ZONE_APPEND_USED		7
 177
 178	struct mutex open_mutex;	/* open/close mutex */
 179	unsigned open_partitions;	/* number of open partitions */
 180
 181	struct backing_dev_info	*bdi;
 182	struct kobject queue_kobj;	/* the queue/ directory */
 183	struct kobject *slave_dir;
 184#ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
 185	struct list_head slave_bdevs;
 186#endif
 187	struct timer_rand_state *random;
 188	struct disk_events *ev;
 189
 190#ifdef CONFIG_BLK_DEV_ZONED
 191	/*
 192	 * Zoned block device information. Reads of this information must be
 193	 * protected with blk_queue_enter() / blk_queue_exit(). Modifying this
 194	 * information is only allowed while no requests are being processed.
 195	 * See also blk_mq_freeze_queue() and blk_mq_unfreeze_queue().
 196	 */
 197	unsigned int		nr_zones;
 198	unsigned int		zone_capacity;
 199	unsigned int		last_zone_capacity;
 200	u8 __rcu		*zones_cond;
 201	unsigned int		zone_wplugs_hash_bits;
 202	atomic_t		nr_zone_wplugs;
 203	spinlock_t		zone_wplugs_lock;
 204	struct mempool		*zone_wplugs_pool;
 205	struct hlist_head	*zone_wplugs_hash;
 206	struct workqueue_struct *zone_wplugs_wq;
 207#endif /* CONFIG_BLK_DEV_ZONED */
 208
 209#if IS_ENABLED(CONFIG_CDROM)
 210	struct cdrom_device_info *cdi;
 211#endif
 212	int node_id;
 213	struct badblocks *bb;
 214	struct lockdep_map lockdep_map;
 215	u64 diskseq;
 216	blk_mode_t open_mode;
 217
 218	/*
 219	 * Independent sector access ranges. This is always NULL for
 220	 * devices that do not have multiple independent access ranges.
 221	 */
 222	struct blk_independent_access_ranges *ia_ranges;
 223
 224	struct mutex rqos_state_mutex;	/* rqos state change mutex */
 225};
 226
 227/**
 228 * disk_openers - returns how many openers are there for a disk
 229 * @disk: disk to check
 230 *
 231 * This returns the number of openers for a disk.  Note that this value is only
 232 * stable if disk->open_mutex is held.
 233 *
 234 * Note: Due to a quirk in the block layer open code, each open partition is
 235 * only counted once even if there are multiple openers.
 236 */
 237static inline unsigned int disk_openers(struct gendisk *disk)
 238{
 239	return atomic_read(&disk->part0->bd_openers);
 240}
 241
 242/**
 243 * disk_has_partscan - return %true if partition scanning is enabled on a disk
 244 * @disk: disk to check
 245 *
 246 * Returns %true if partitions scanning is enabled for @disk, or %false if
 247 * partition scanning is disabled either permanently or temporarily.
 248 */
 249static inline bool disk_has_partscan(struct gendisk *disk)
 250{
 251	return !(disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN)) &&
 252		!test_bit(GD_SUPPRESS_PART_SCAN, &disk->state);
 253}
 254
 255/*
 256 * The gendisk is refcounted by the part0 block_device, and the bd_device
 257 * therein is also used for device model presentation in sysfs.
 258 */
 259#define dev_to_disk(device) \
 260	(dev_to_bdev(device)->bd_disk)
 261#define disk_to_dev(disk) \
 262	(&((disk)->part0->bd_device))
 263
 264#if IS_REACHABLE(CONFIG_CDROM)
 265#define disk_to_cdi(disk)	((disk)->cdi)
 266#else
 267#define disk_to_cdi(disk)	NULL
 268#endif
 269
 270static inline dev_t disk_devt(struct gendisk *disk)
 271{
 272	return MKDEV(disk->major, disk->first_minor);
 273}
 274
 275#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 276/*
 277 * We should strive for 1 << (PAGE_SHIFT + MAX_PAGECACHE_ORDER)
 278 * however we constrain this to what we can validate and test.
 279 */
 280#define BLK_MAX_BLOCK_SIZE      SZ_64K
 281#else
 282#define BLK_MAX_BLOCK_SIZE      PAGE_SIZE
 283#endif
 284
 285
 286/* blk_validate_limits() validates bsize, so drivers don't usually need to */
 287static inline int blk_validate_block_size(unsigned long bsize)
 288{
 289	if (bsize < 512 || bsize > BLK_MAX_BLOCK_SIZE || !is_power_of_2(bsize))
 290		return -EINVAL;
 291
 292	return 0;
 293}
 294
 295static inline bool blk_op_is_passthrough(blk_opf_t op)
 296{
 297	op &= REQ_OP_MASK;
 298	return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT;
 299}
 300
 301/* flags set by the driver in queue_limits.features */
 302typedef unsigned int __bitwise blk_features_t;
 303
 304/* supports a volatile write cache */
 305#define BLK_FEAT_WRITE_CACHE		((__force blk_features_t)(1u << 0))
 306
 307/* supports passing on the FUA bit */
 308#define BLK_FEAT_FUA			((__force blk_features_t)(1u << 1))
 309
 310/* rotational device (hard drive or floppy) */
 311#define BLK_FEAT_ROTATIONAL		((__force blk_features_t)(1u << 2))
 312
 313/* contributes to the random number pool */
 314#define BLK_FEAT_ADD_RANDOM		((__force blk_features_t)(1u << 3))
 315
 316/* do disk/partitions IO accounting */
 317#define BLK_FEAT_IO_STAT		((__force blk_features_t)(1u << 4))
 318
 319/* don't modify data until writeback is done */
 320#define BLK_FEAT_STABLE_WRITES		((__force blk_features_t)(1u << 5))
 321
 322/* always completes in submit context */
 323#define BLK_FEAT_SYNCHRONOUS		((__force blk_features_t)(1u << 6))
 324
 325/* supports REQ_NOWAIT */
 326#define BLK_FEAT_NOWAIT			((__force blk_features_t)(1u << 7))
 327
 328/* supports DAX */
 329#define BLK_FEAT_DAX			((__force blk_features_t)(1u << 8))
 330
 331/* supports I/O polling */
 332#define BLK_FEAT_POLL			((__force blk_features_t)(1u << 9))
 333
 334/* is a zoned device */
 335#define BLK_FEAT_ZONED			((__force blk_features_t)(1u << 10))
 336
 337/* supports PCI(e) p2p requests */
 338#define BLK_FEAT_PCI_P2PDMA		((__force blk_features_t)(1u << 12))
 339
 340/* skip this queue in blk_mq_(un)quiesce_tagset */
 341#define BLK_FEAT_SKIP_TAGSET_QUIESCE	((__force blk_features_t)(1u << 13))
 342
 343/* undocumented magic for bcache */
 344#define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \
 345	((__force blk_features_t)(1u << 15))
 346
 347/* atomic writes enabled */
 348#define BLK_FEAT_ATOMIC_WRITES \
 349	((__force blk_features_t)(1u << 16))
 350
 351/*
 352 * Flags automatically inherited when stacking limits.
 353 */
 354#define BLK_FEAT_INHERIT_MASK \
 355	(BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \
 356	 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | \
 357	 BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE)
 358
 359/* internal flags in queue_limits.flags */
 360typedef unsigned int __bitwise blk_flags_t;
 361
 362/* do not send FLUSH/FUA commands despite advertising a write cache */
 363#define BLK_FLAG_WRITE_CACHE_DISABLED	((__force blk_flags_t)(1u << 0))
 364
 365/* I/O topology is misaligned */
 366#define BLK_FLAG_MISALIGNED		((__force blk_flags_t)(1u << 1))
 367
 368/* passthrough command IO accounting */
 369#define BLK_FLAG_IOSTATS_PASSTHROUGH	((__force blk_flags_t)(1u << 2))
 370
 371struct queue_limits {
 372	blk_features_t		features;
 373	blk_flags_t		flags;
 374	unsigned long		seg_boundary_mask;
 375	unsigned long		virt_boundary_mask;
 376
 377	unsigned int		max_hw_sectors;
 378	unsigned int		max_dev_sectors;
 379	unsigned int		chunk_sectors;
 380	unsigned int		max_sectors;
 381	unsigned int		max_user_sectors;
 382	unsigned int		max_segment_size;
 383	unsigned int		max_fast_segment_size;
 384	unsigned int		physical_block_size;
 385	unsigned int		logical_block_size;
 386	unsigned int		alignment_offset;
 387	unsigned int		io_min;
 388	unsigned int		io_opt;
 389	unsigned int		max_discard_sectors;
 390	unsigned int		max_hw_discard_sectors;
 391	unsigned int		max_user_discard_sectors;
 392	unsigned int		max_secure_erase_sectors;
 393	unsigned int		max_write_zeroes_sectors;
 394	unsigned int		max_wzeroes_unmap_sectors;
 395	unsigned int		max_hw_wzeroes_unmap_sectors;
 396	unsigned int		max_user_wzeroes_unmap_sectors;
 397	unsigned int		max_hw_zone_append_sectors;
 398	unsigned int		max_zone_append_sectors;
 399	unsigned int		discard_granularity;
 400	unsigned int		discard_alignment;
 401	unsigned int		zone_write_granularity;
 402
 403	/* atomic write limits */
 404	unsigned int		atomic_write_hw_max;
 405	unsigned int		atomic_write_max_sectors;
 406	unsigned int		atomic_write_hw_boundary;
 407	unsigned int		atomic_write_boundary_sectors;
 408	unsigned int		atomic_write_hw_unit_min;
 409	unsigned int		atomic_write_unit_min;
 410	unsigned int		atomic_write_hw_unit_max;
 411	unsigned int		atomic_write_unit_max;
 412
 413	unsigned short		max_segments;
 414	unsigned short		max_integrity_segments;
 415	unsigned short		max_discard_segments;
 416
 417	unsigned short		max_write_streams;
 418	unsigned int		write_stream_granularity;
 419
 420	unsigned int		max_open_zones;
 421	unsigned int		max_active_zones;
 422
 423	/*
 424	 * Drivers that set dma_alignment to less than 511 must be prepared to
 425	 * handle individual bvec's that are not a multiple of a SECTOR_SIZE
 426	 * due to possible offsets.
 427	 */
 428	unsigned int		dma_alignment;
 429	unsigned int		dma_pad_mask;
 430
 431	struct blk_integrity	integrity;
 432};
 433
 434typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx,
 435			       void *data);
 436
 437int disk_report_zone(struct gendisk *disk, struct blk_zone *zone,
 438		     unsigned int idx, struct blk_report_zones_args *args);
 439
 440int blkdev_get_zone_info(struct block_device *bdev, sector_t sector,
 441			 struct blk_zone *zone);
 442
 443#define BLK_ALL_ZONES  ((unsigned int)-1)
 444int blkdev_report_zones(struct block_device *bdev, sector_t sector,
 445		unsigned int nr_zones, report_zones_cb cb, void *data);
 446int blkdev_report_zones_cached(struct block_device *bdev, sector_t sector,
 447		unsigned int nr_zones, report_zones_cb cb, void *data);
 448int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op,
 449		sector_t sectors, sector_t nr_sectors);
 450int blk_revalidate_disk_zones(struct gendisk *disk);
 451
 452/*
 453 * Independent access ranges: struct blk_independent_access_range describes
 454 * a range of contiguous sectors that can be accessed using device command
 455 * execution resources that are independent from the resources used for
 456 * other access ranges. This is typically found with single-LUN multi-actuator
 457 * HDDs where each access range is served by a different set of heads.
 458 * The set of independent ranges supported by the device is defined using
 459 * struct blk_independent_access_ranges. The independent ranges must not overlap
 460 * and must include all sectors within the disk capacity (no sector holes
 461 * allowed).
 462 * For a device with multiple ranges, requests targeting sectors in different
 463 * ranges can be executed in parallel. A request can straddle an access range
 464 * boundary.
 465 */
 466struct blk_independent_access_range {
 467	struct kobject		kobj;
 468	sector_t		sector;
 469	sector_t		nr_sectors;
 470};
 471
 472struct blk_independent_access_ranges {
 473	struct kobject				kobj;
 474	bool					sysfs_registered;
 475	unsigned int				nr_ia_ranges;
 476	struct blk_independent_access_range	ia_range[];
 477};
 478
 479struct request_queue {
 480	/*
 481	 * The queue owner gets to use this for whatever they like.
 482	 * ll_rw_blk doesn't touch it.
 483	 */
 484	void			*queuedata;
 485
 486	struct elevator_queue	*elevator;
 487
 488	const struct blk_mq_ops	*mq_ops;
 489
 490	/* sw queues */
 491	struct blk_mq_ctx __percpu	*queue_ctx;
 492
 493	/*
 494	 * various queue flags, see QUEUE_* below
 495	 */
 496	unsigned long		queue_flags;
 497
 498	unsigned int __data_racy rq_timeout;
 499
 500	unsigned int		queue_depth;
 501
 502	refcount_t		refs;
 503
 504	/* hw dispatch queues */
 505	unsigned int		nr_hw_queues;
 506	struct blk_mq_hw_ctx * __rcu *queue_hw_ctx;
 507
 508	struct percpu_ref	q_usage_counter;
 509	struct lock_class_key	io_lock_cls_key;
 510	struct lockdep_map	io_lockdep_map;
 511
 512	struct lock_class_key	q_lock_cls_key;
 513	struct lockdep_map	q_lockdep_map;
 514
 515	struct request		*last_merge;
 516
 517	spinlock_t		queue_lock;
 518
 519	int			quiesce_depth;
 520
 521	struct gendisk		*disk;
 522
 523	/*
 524	 * mq queue kobject
 525	 */
 526	struct kobject *mq_kobj;
 527
 528	struct queue_limits	limits;
 529
 530#ifdef CONFIG_PM
 531	struct device		*dev;
 532	enum rpm_status		rpm_status;
 533#endif
 534
 535	/*
 536	 * Number of contexts that have called blk_set_pm_only(). If this
 537	 * counter is above zero then only RQF_PM requests are processed.
 538	 */
 539	atomic_t		pm_only;
 540
 541	struct blk_queue_stats	*stats;
 542	struct rq_qos		*rq_qos;
 543	struct mutex		rq_qos_mutex;
 544
 545	/*
 546	 * ida allocated id for this queue.  Used to index queues from
 547	 * ioctx.
 548	 */
 549	int			id;
 550
 551	/*
 552	 * queue settings
 553	 */
 554	unsigned long		nr_requests;	/* Max # of requests */
 555
 556#ifdef CONFIG_BLK_INLINE_ENCRYPTION
 557	struct blk_crypto_profile *crypto_profile;
 558	struct kobject *crypto_kobject;
 559#endif
 560
 561	struct timer_list	timeout;
 562	struct work_struct	timeout_work;
 563
 564	atomic_t		nr_active_requests_shared_tags;
 565
 566	struct blk_mq_tags	*sched_shared_tags;
 567
 568	struct list_head	icq_list;
 569#ifdef CONFIG_BLK_CGROUP
 570	DECLARE_BITMAP		(blkcg_pols, BLKCG_MAX_POLS);
 571	struct blkcg_gq		*root_blkg;
 572	struct list_head	blkg_list;
 573	struct mutex		blkcg_mutex;
 574#endif
 575
 576	int			node;
 577
 578	spinlock_t		requeue_lock;
 579	struct list_head	requeue_list;
 580	struct delayed_work	requeue_work;
 581
 582#ifdef CONFIG_BLK_DEV_IO_TRACE
 583	struct blk_trace __rcu	*blk_trace;
 584#endif
 585	/*
 586	 * for flush operations
 587	 */
 588	struct blk_flush_queue	*fq;
 589	struct list_head	flush_list;
 590
 591	/*
 592	 * Protects against I/O scheduler switching, particularly when updating
 593	 * q->elevator. Since the elevator update code path may also modify q->
 594	 * nr_requests and wbt latency, this lock also protects the sysfs attrs
 595	 * nr_requests and wbt_lat_usec. Additionally the nr_hw_queues update
 596	 * may modify hctx tags, reserved-tags and cpumask, so this lock also
 597	 * helps protect the hctx sysfs/debugfs attrs. To ensure proper locking
 598	 * order during an elevator or nr_hw_queue update, first freeze the
 599	 * queue, then acquire ->elevator_lock.
 600	 */
 601	struct mutex		elevator_lock;
 602
 603	struct mutex		sysfs_lock;
 604	/*
 605	 * Protects queue limits and also sysfs attribute read_ahead_kb.
 606	 */
 607	struct mutex		limits_lock;
 608
 609	/*
 610	 * for reusing dead hctx instance in case of updating
 611	 * nr_hw_queues
 612	 */
 613	struct list_head	unused_hctx_list;
 614	spinlock_t		unused_hctx_lock;
 615
 616	int			mq_freeze_depth;
 617
 618#ifdef CONFIG_BLK_DEV_THROTTLING
 619	/* Throttle data */
 620	struct throtl_data *td;
 621#endif
 622	struct rcu_head		rcu_head;
 623#ifdef CONFIG_LOCKDEP
 624	struct task_struct	*mq_freeze_owner;
 625	int			mq_freeze_owner_depth;
 626	/*
 627	 * Records disk & queue state in current context, used in unfreeze
 628	 * queue
 629	 */
 630	bool			mq_freeze_disk_dead;
 631	bool			mq_freeze_queue_dying;
 632#endif
 633	wait_queue_head_t	mq_freeze_wq;
 634	/*
 635	 * Protect concurrent access to q_usage_counter by
 636	 * percpu_ref_kill() and percpu_ref_reinit().
 637	 */
 638	struct mutex		mq_freeze_lock;
 639
 640	struct blk_mq_tag_set	*tag_set;
 641	struct list_head	tag_set_list;
 642
 643	struct dentry		*debugfs_dir;
 644	struct dentry		*sched_debugfs_dir;
 645	struct dentry		*rqos_debugfs_dir;
 646	/*
 647	 * Serializes all debugfs metadata operations using the above dentries.
 648	 */
 649	struct mutex		debugfs_mutex;
 650};
 651
 652/* Keep blk_queue_flag_name[] in sync with the definitions below */
 653enum {
 654	QUEUE_FLAG_DYING,		/* queue being torn down */
 655	QUEUE_FLAG_NOMERGES,		/* disable merge attempts */
 656	QUEUE_FLAG_SAME_COMP,		/* complete on same CPU-group */
 657	QUEUE_FLAG_FAIL_IO,		/* fake timeout */
 658	QUEUE_FLAG_NOXMERGES,		/* No extended merges */
 659	QUEUE_FLAG_SAME_FORCE,		/* force complete on same CPU */
 660	QUEUE_FLAG_INIT_DONE,		/* queue is initialized */
 661	QUEUE_FLAG_STATS,		/* track IO start and completion times */
 662	QUEUE_FLAG_REGISTERED,		/* queue has been registered to a disk */
 663	QUEUE_FLAG_QUIESCED,		/* queue has been quiesced */
 664	QUEUE_FLAG_RQ_ALLOC_TIME,	/* record rq->alloc_time_ns */
 665	QUEUE_FLAG_HCTX_ACTIVE,		/* at least one blk-mq hctx is active */
 666	QUEUE_FLAG_SQ_SCHED,		/* single queue style io dispatch */
 667	QUEUE_FLAG_DISABLE_WBT_DEF,	/* for sched to disable/enable wbt */
 668	QUEUE_FLAG_NO_ELV_SWITCH,	/* can't switch elevator any more */
 669	QUEUE_FLAG_QOS_ENABLED,		/* qos is enabled */
 670	QUEUE_FLAG_BIO_ISSUE_TIME,	/* record bio->issue_time_ns */
 671	QUEUE_FLAG_MAX
 672};
 673
 674#define QUEUE_FLAG_MQ_DEFAULT	(1UL << QUEUE_FLAG_SAME_COMP)
 675
 676void blk_queue_flag_set(unsigned int flag, struct request_queue *q);
 677void blk_queue_flag_clear(unsigned int flag, struct request_queue *q);
 678
 679#define blk_queue_dying(q)	test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags)
 680#define blk_queue_init_done(q)	test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags)
 681#define blk_queue_nomerges(q)	test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags)
 682#define blk_queue_noxmerges(q)	\
 683	test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags)
 684#define blk_queue_nonrot(q)	(!((q)->limits.features & BLK_FEAT_ROTATIONAL))
 685#define blk_queue_io_stat(q)	((q)->limits.features & BLK_FEAT_IO_STAT)
 686#define blk_queue_passthrough_stat(q)	\
 687	((q)->limits.flags & BLK_FLAG_IOSTATS_PASSTHROUGH)
 688#define blk_queue_dax(q)	((q)->limits.features & BLK_FEAT_DAX)
 689#define blk_queue_pci_p2pdma(q)	((q)->limits.features & BLK_FEAT_PCI_P2PDMA)
 690#ifdef CONFIG_BLK_RQ_ALLOC_TIME
 691#define blk_queue_rq_alloc_time(q)	\
 692	test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags)
 693#else
 694#define blk_queue_rq_alloc_time(q)	false
 695#endif
 696
 697#define blk_noretry_request(rq) \
 698	((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
 699			     REQ_FAILFAST_DRIVER))
 700#define blk_queue_quiesced(q)	test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags)
 701#define blk_queue_pm_only(q)	atomic_read(&(q)->pm_only)
 702#define blk_queue_registered(q)	test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags)
 703#define blk_queue_sq_sched(q)	test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags)
 704#define blk_queue_skip_tagset_quiesce(q) \
 705	((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE)
 706#define blk_queue_disable_wbt(q)	\
 707	test_bit(QUEUE_FLAG_DISABLE_WBT_DEF, &(q)->queue_flags)
 708#define blk_queue_no_elv_switch(q)	\
 709	test_bit(QUEUE_FLAG_NO_ELV_SWITCH, &(q)->queue_flags)
 710
 711extern void blk_set_pm_only(struct request_queue *q);
 712extern void blk_clear_pm_only(struct request_queue *q);
 713
 714#define list_entry_rq(ptr)	list_entry((ptr), struct request, queuelist)
 715
 716#define dma_map_bvec(dev, bv, dir, attrs) \
 717	dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \
 718	(dir), (attrs))
 719
 720static inline bool queue_is_mq(struct request_queue *q)
 721{
 722	return q->mq_ops;
 723}
 724
 725#ifdef CONFIG_PM
 726static inline enum rpm_status queue_rpm_status(struct request_queue *q)
 727{
 728	return q->rpm_status;
 729}
 730#else
 731static inline enum rpm_status queue_rpm_status(struct request_queue *q)
 732{
 733	return RPM_ACTIVE;
 734}
 735#endif
 736
 737static inline bool blk_queue_is_zoned(struct request_queue *q)
 738{
 739	return IS_ENABLED(CONFIG_BLK_DEV_ZONED) &&
 740		(q->limits.features & BLK_FEAT_ZONED);
 741}
 742
 743static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector)
 744{
 745	if (!blk_queue_is_zoned(disk->queue))
 746		return 0;
 747	return sector >> ilog2(disk->queue->limits.chunk_sectors);
 748}
 749
 750static inline unsigned int bdev_max_open_zones(struct block_device *bdev)
 751{
 752	return bdev->bd_disk->queue->limits.max_open_zones;
 753}
 754
 755static inline unsigned int bdev_max_active_zones(struct block_device *bdev)
 756{
 757	return bdev->bd_disk->queue->limits.max_active_zones;
 758}
 759
 760static inline unsigned int blk_queue_depth(struct request_queue *q)
 761{
 762	if (q->queue_depth)
 763		return q->queue_depth;
 764
 765	return q->nr_requests;
 766}
 767
 768/*
 769 * default timeout for SG_IO if none specified
 770 */
 771#define BLK_DEFAULT_SG_TIMEOUT	(60 * HZ)
 772#define BLK_MIN_SG_TIMEOUT	(7 * HZ)
 773
 774/* This should not be used directly - use rq_for_each_segment */
 775#define for_each_bio(_bio)		\
 776	for (; _bio; _bio = _bio->bi_next)
 777
 778int __must_check add_disk_fwnode(struct device *parent, struct gendisk *disk,
 779				 const struct attribute_group **groups,
 780				 struct fwnode_handle *fwnode);
 781int __must_check device_add_disk(struct device *parent, struct gendisk *disk,
 782				 const struct attribute_group **groups);
 783static inline int __must_check add_disk(struct gendisk *disk)
 784{
 785	return device_add_disk(NULL, disk, NULL);
 786}
 787void del_gendisk(struct gendisk *gp);
 788void invalidate_disk(struct gendisk *disk);
 789void set_disk_ro(struct gendisk *disk, bool read_only);
 790void disk_uevent(struct gendisk *disk, enum kobject_action action);
 791
 792static inline u8 bdev_partno(const struct block_device *bdev)
 793{
 794	return atomic_read(&bdev->__bd_flags) & BD_PARTNO;
 795}
 796
 797static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag)
 798{
 799	return atomic_read(&bdev->__bd_flags) & flag;
 800}
 801
 802static inline void bdev_set_flag(struct block_device *bdev, unsigned flag)
 803{
 804	atomic_or(flag, &bdev->__bd_flags);
 805}
 806
 807static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag)
 808{
 809	atomic_andnot(flag, &bdev->__bd_flags);
 810}
 811
 812static inline bool get_disk_ro(struct gendisk *disk)
 813{
 814	return bdev_test_flag(disk->part0, BD_READ_ONLY) ||
 815		test_bit(GD_READ_ONLY, &disk->state);
 816}
 817
 818static inline bool bdev_read_only(struct block_device *bdev)
 819{
 820	return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk);
 821}
 822
 823bool set_capacity_and_notify(struct gendisk *disk, sector_t size);
 824void disk_force_media_change(struct gendisk *disk);
 825void bdev_mark_dead(struct block_device *bdev, bool surprise);
 826
 827void add_disk_randomness(struct gendisk *disk) __latent_entropy;
 828void rand_initialize_disk(struct gendisk *disk);
 829
 830static inline sector_t get_start_sect(struct block_device *bdev)
 831{
 832	return bdev->bd_start_sect;
 833}
 834
 835static inline sector_t bdev_nr_sectors(struct block_device *bdev)
 836{
 837	return bdev->bd_nr_sectors;
 838}
 839
 840static inline loff_t bdev_nr_bytes(struct block_device *bdev)
 841{
 842	return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT;
 843}
 844
 845static inline sector_t get_capacity(struct gendisk *disk)
 846{
 847	return bdev_nr_sectors(disk->part0);
 848}
 849
 850static inline u64 sb_bdev_nr_blocks(struct super_block *sb)
 851{
 852	return bdev_nr_sectors(sb->s_bdev) >>
 853		(sb->s_blocksize_bits - SECTOR_SHIFT);
 854}
 855
 856#ifdef CONFIG_BLK_DEV_ZONED
 857static inline unsigned int disk_nr_zones(struct gendisk *disk)
 858{
 859	return disk->nr_zones;
 860}
 861
 862/**
 863 * bio_needs_zone_write_plugging - Check if a BIO needs to be handled with zone
 864 *				   write plugging
 865 * @bio: The BIO being submitted
 866 *
 867 * Return true whenever @bio execution needs to be handled through zone
 868 * write plugging (using blk_zone_plug_bio()). Return false otherwise.
 869 */
 870static inline bool bio_needs_zone_write_plugging(struct bio *bio)
 871{
 872	enum req_op op = bio_op(bio);
 873
 874	/*
 875	 * Only zoned block devices have a zone write plug hash table. But not
 876	 * all of them have one (e.g. DM devices may not need one).
 877	 */
 878	if (!bio->bi_bdev->bd_disk->zone_wplugs_hash)
 879		return false;
 880
 881	/* Only write operations need zone write plugging. */
 882	if (!op_is_write(op))
 883		return false;
 884
 885	/* Ignore empty flush */
 886	if (op_is_flush(bio->bi_opf) && !bio_sectors(bio))
 887		return false;
 888
 889	/* Ignore BIOs that already have been handled by zone write plugging. */
 890	if (bio_flagged(bio, BIO_ZONE_WRITE_PLUGGING))
 891		return false;
 892
 893	/*
 894	 * All zone write operations must be handled through zone write plugging
 895	 * using blk_zone_plug_bio().
 896	 */
 897	switch (op) {
 898	case REQ_OP_ZONE_APPEND:
 899	case REQ_OP_WRITE:
 900	case REQ_OP_WRITE_ZEROES:
 901	case REQ_OP_ZONE_FINISH:
 902	case REQ_OP_ZONE_RESET:
 903	case REQ_OP_ZONE_RESET_ALL:
 904		return true;
 905	default:
 906		return false;
 907	}
 908}
 909
 910bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs);
 911
 912/**
 913 * disk_zone_capacity - returns the zone capacity of zone containing @sector
 914 * @disk:	disk to work with
 915 * @sector:	sector number within the querying zone
 916 *
 917 * Returns the zone capacity of a zone containing @sector. @sector can be any
 918 * sector in the zone.
 919 */
 920static inline unsigned int disk_zone_capacity(struct gendisk *disk,
 921					      sector_t sector)
 922{
 923	sector_t zone_sectors = disk->queue->limits.chunk_sectors;
 924
 925	if (sector + zone_sectors >= get_capacity(disk))
 926		return disk->last_zone_capacity;
 927	return disk->zone_capacity;
 928}
 929static inline unsigned int bdev_zone_capacity(struct block_device *bdev,
 930					      sector_t pos)
 931{
 932	return disk_zone_capacity(bdev->bd_disk, pos);
 933}
 934
 935bool bdev_zone_is_seq(struct block_device *bdev, sector_t sector);
 936
 937#else /* CONFIG_BLK_DEV_ZONED */
 938static inline unsigned int disk_nr_zones(struct gendisk *disk)
 939{
 940	return 0;
 941}
 942
 943static inline bool bdev_zone_is_seq(struct block_device *bdev, sector_t sector)
 944{
 945	return false;
 946}
 947
 948static inline bool bio_needs_zone_write_plugging(struct bio *bio)
 949{
 950	return false;
 951}
 952
 953static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs)
 954{
 955	return false;
 956}
 957#endif /* CONFIG_BLK_DEV_ZONED */
 958
 959static inline unsigned int bdev_nr_zones(struct block_device *bdev)
 960{
 961	return disk_nr_zones(bdev->bd_disk);
 962}
 963
 964int bdev_disk_changed(struct gendisk *disk, bool invalidate);
 965
 966void put_disk(struct gendisk *disk);
 967struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node,
 968		struct lock_class_key *lkclass);
 969
 970/**
 971 * blk_alloc_disk - allocate a gendisk structure
 972 * @lim: queue limits to be used for this disk.
 973 * @node_id: numa node to allocate on
 974 *
 975 * Allocate and pre-initialize a gendisk structure for use with BIO based
 976 * drivers.
 977 *
 978 * Returns an ERR_PTR on error, else the allocated disk.
 979 *
 980 * Context: can sleep
 981 */
 982#define blk_alloc_disk(lim, node_id)					\
 983({									\
 984	static struct lock_class_key __key;				\
 985									\
 986	__blk_alloc_disk(lim, node_id, &__key);				\
 987})
 988
 989int __register_blkdev(unsigned int major, const char *name,
 990		void (*probe)(dev_t devt));
 991#define register_blkdev(major, name) \
 992	__register_blkdev(major, name, NULL)
 993void unregister_blkdev(unsigned int major, const char *name);
 994
 995bool disk_check_media_change(struct gendisk *disk);
 996void set_capacity(struct gendisk *disk, sector_t size);
 997
 998#ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
 999int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk);
1000void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk);
1001#else
1002static inline int bd_link_disk_holder(struct block_device *bdev,
1003				      struct gendisk *disk)
1004{
1005	return 0;
1006}
1007static inline void bd_unlink_disk_holder(struct block_device *bdev,
1008					 struct gendisk *disk)
1009{
1010}
1011#endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */
1012
1013dev_t part_devt(struct gendisk *disk, u8 partno);
1014void inc_diskseq(struct gendisk *disk);
1015void blk_request_module(dev_t devt);
1016
1017extern int blk_register_queue(struct gendisk *disk);
1018extern void blk_unregister_queue(struct gendisk *disk);
1019void submit_bio_noacct(struct bio *bio);
1020struct bio *bio_split_to_limits(struct bio *bio);
1021struct bio *bio_submit_split_bioset(struct bio *bio, unsigned int split_sectors,
1022				    struct bio_set *bs);
1023
1024extern int blk_lld_busy(struct request_queue *q);
1025extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags);
1026extern void blk_queue_exit(struct request_queue *q);
1027extern void blk_sync_queue(struct request_queue *q);
1028
1029/* Helper to convert REQ_OP_XXX to its string format XXX */
1030extern const char *blk_op_str(enum req_op op);
1031
1032int blk_status_to_errno(blk_status_t status);
1033blk_status_t errno_to_blk_status(int errno);
1034const char *blk_status_to_str(blk_status_t status);
1035
1036/* only poll the hardware once, don't continue until a completion was found */
1037#define BLK_POLL_ONESHOT		(1 << 0)
1038int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags);
1039int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob,
1040			unsigned int flags);
1041
1042static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
1043{
1044	return bdev->bd_queue;	/* this is never NULL */
1045}
1046
1047/* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */
1048const char *blk_zone_cond_str(enum blk_zone_cond zone_cond);
1049
1050static inline unsigned int bio_zone_no(struct bio *bio)
1051{
1052	return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector);
1053}
1054
1055static inline bool bio_straddles_zones(struct bio *bio)
1056{
1057	return bio_sectors(bio) &&
1058		bio_zone_no(bio) !=
1059		disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1);
1060}
1061
1062/*
1063 * Return how much within the boundary is left to be used for I/O at a given
1064 * offset.
1065 */
1066static inline unsigned int blk_boundary_sectors_left(sector_t offset,
1067		unsigned int boundary_sectors)
1068{
1069	if (unlikely(!is_power_of_2(boundary_sectors)))
1070		return boundary_sectors - sector_div(offset, boundary_sectors);
1071	return boundary_sectors - (offset & (boundary_sectors - 1));
1072}
1073
1074/**
1075 * queue_limits_start_update - start an atomic update of queue limits
1076 * @q:		queue to update
1077 *
1078 * This functions starts an atomic update of the queue limits.  It takes a lock
1079 * to prevent other updates and returns a snapshot of the current limits that
1080 * the caller can modify.  The caller must call queue_limits_commit_update()
1081 * to finish the update.
1082 *
1083 * Context: process context.
1084 */
1085static inline struct queue_limits
1086queue_limits_start_update(struct request_queue *q)
1087{
1088	mutex_lock(&q->limits_lock);
1089	return q->limits;
1090}
1091int queue_limits_commit_update_frozen(struct request_queue *q,
1092		struct queue_limits *lim);
1093int queue_limits_commit_update(struct request_queue *q,
1094		struct queue_limits *lim);
1095int queue_limits_set(struct request_queue *q, struct queue_limits *lim);
1096int blk_validate_limits(struct queue_limits *lim);
1097
1098/**
1099 * queue_limits_cancel_update - cancel an atomic update of queue limits
1100 * @q:		queue to update
1101 *
1102 * This functions cancels an atomic update of the queue limits started by
1103 * queue_limits_start_update() and should be used when an error occurs after
1104 * starting update.
1105 */
1106static inline void queue_limits_cancel_update(struct request_queue *q)
1107{
1108	mutex_unlock(&q->limits_lock);
1109}
1110
1111/*
1112 * These helpers are for drivers that have sloppy feature negotiation and might
1113 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O
1114 * completion handler when the device returned an indicator that the respective
1115 * feature is not actually supported.  They are racy and the driver needs to
1116 * cope with that.  Try to avoid this scheme if you can.
1117 */
1118static inline void blk_queue_disable_discard(struct request_queue *q)
1119{
1120	q->limits.max_discard_sectors = 0;
1121}
1122
1123static inline void blk_queue_disable_secure_erase(struct request_queue *q)
1124{
1125	q->limits.max_secure_erase_sectors = 0;
1126}
1127
1128static inline void blk_queue_disable_write_zeroes(struct request_queue *q)
1129{
1130	q->limits.max_write_zeroes_sectors = 0;
1131	q->limits.max_wzeroes_unmap_sectors = 0;
1132}
1133
1134/*
1135 * Access functions for manipulating queue properties
1136 */
1137extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth);
1138extern void blk_set_stacking_limits(struct queue_limits *lim);
1139extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
1140			    sector_t offset);
1141void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev,
1142		sector_t offset, const char *pfx);
1143extern void blk_queue_rq_timeout(struct request_queue *, unsigned int);
1144
1145struct blk_independent_access_ranges *
1146disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges);
1147void disk_set_independent_access_ranges(struct gendisk *disk,
1148				struct blk_independent_access_ranges *iars);
1149
1150bool __must_check blk_get_queue(struct request_queue *);
1151extern void blk_put_queue(struct request_queue *);
1152
1153void blk_mark_disk_dead(struct gendisk *disk);
1154
1155struct rq_list {
1156	struct request *head;
1157	struct request *tail;
1158};
1159
1160#ifdef CONFIG_BLOCK
1161/*
1162 * blk_plug permits building a queue of related requests by holding the I/O
1163 * fragments for a short period. This allows merging of sequential requests
1164 * into single larger request. As the requests are moved from a per-task list to
1165 * the device's request_queue in a batch, this results in improved scalability
1166 * as the lock contention for request_queue lock is reduced.
1167 *
1168 * It is ok not to disable preemption when adding the request to the plug list
1169 * or when attempting a merge. For details, please see schedule() where
1170 * blk_flush_plug() is called.
1171 */
1172struct blk_plug {
1173	struct rq_list mq_list; /* blk-mq requests */
1174
1175	/* if ios_left is > 1, we can batch tag/rq allocations */
1176	struct rq_list cached_rqs;
1177	u64 cur_ktime;
1178	unsigned short nr_ios;
1179
1180	unsigned short rq_count;
1181
1182	bool multiple_queues;
1183	bool has_elevator;
1184
1185	struct list_head cb_list; /* md requires an unplug callback */
1186};
1187
1188struct blk_plug_cb;
1189typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool);
1190struct blk_plug_cb {
1191	struct list_head list;
1192	blk_plug_cb_fn callback;
1193	void *data;
1194};
1195extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug,
1196					     void *data, int size);
1197extern void blk_start_plug(struct blk_plug *);
1198extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short);
1199extern void blk_finish_plug(struct blk_plug *);
1200
1201void __blk_flush_plug(struct blk_plug *plug, bool from_schedule);
1202static inline void blk_flush_plug(struct blk_plug *plug, bool async)
1203{
1204	if (plug)
1205		__blk_flush_plug(plug, async);
1206}
1207
1208/*
1209 * tsk == current here
1210 */
1211static inline void blk_plug_invalidate_ts(struct task_struct *tsk)
1212{
1213	struct blk_plug *plug = tsk->plug;
1214
1215	if (plug)
1216		plug->cur_ktime = 0;
1217	current->flags &= ~PF_BLOCK_TS;
1218}
1219
1220int blkdev_issue_flush(struct block_device *bdev);
1221long nr_blockdev_pages(void);
1222#else /* CONFIG_BLOCK */
1223struct blk_plug {
1224};
1225
1226static inline void blk_start_plug_nr_ios(struct blk_plug *plug,
1227					 unsigned short nr_ios)
1228{
1229}
1230
1231static inline void blk_start_plug(struct blk_plug *plug)
1232{
1233}
1234
1235static inline void blk_finish_plug(struct blk_plug *plug)
1236{
1237}
1238
1239static inline void blk_flush_plug(struct blk_plug *plug, bool async)
1240{
1241}
1242
1243static inline void blk_plug_invalidate_ts(struct task_struct *tsk)
1244{
1245}
1246
1247static inline int blkdev_issue_flush(struct block_device *bdev)
1248{
1249	return 0;
1250}
1251
1252static inline long nr_blockdev_pages(void)
1253{
1254	return 0;
1255}
1256#endif /* CONFIG_BLOCK */
1257
1258extern void blk_io_schedule(void);
1259
1260int blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1261		sector_t nr_sects, gfp_t gfp_mask);
1262int __blkdev_issue_discard(struct block_device *bdev, sector_t sector,
1263		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop);
1264int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector,
1265		sector_t nr_sects, gfp_t gfp);
1266
1267#define BLKDEV_ZERO_NOUNMAP	(1 << 0)  /* do not free blocks */
1268#define BLKDEV_ZERO_NOFALLBACK	(1 << 1)  /* don't write explicit zeroes */
1269#define BLKDEV_ZERO_KILLABLE	(1 << 2)  /* interruptible by fatal signals */
1270
1271extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1272		sector_t nr_sects, gfp_t gfp_mask, struct bio **biop,
1273		unsigned flags);
1274extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
1275		sector_t nr_sects, gfp_t gfp_mask, unsigned flags);
1276
1277static inline int sb_issue_discard(struct super_block *sb, sector_t block,
1278		sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags)
1279{
1280	return blkdev_issue_discard(sb->s_bdev,
1281				    block << (sb->s_blocksize_bits -
1282					      SECTOR_SHIFT),
1283				    nr_blocks << (sb->s_blocksize_bits -
1284						  SECTOR_SHIFT),
1285				    gfp_mask);
1286}
1287static inline int sb_issue_zeroout(struct super_block *sb, sector_t block,
1288		sector_t nr_blocks, gfp_t gfp_mask)
1289{
1290	return blkdev_issue_zeroout(sb->s_bdev,
1291				    block << (sb->s_blocksize_bits -
1292					      SECTOR_SHIFT),
1293				    nr_blocks << (sb->s_blocksize_bits -
1294						  SECTOR_SHIFT),
1295				    gfp_mask, 0);
1296}
1297
1298static inline bool bdev_is_partition(struct block_device *bdev)
1299{
1300	return bdev_partno(bdev) != 0;
1301}
1302
1303enum blk_default_limits {
1304	BLK_MAX_SEGMENTS	= 128,
1305	BLK_SAFE_MAX_SECTORS	= 255,
1306	BLK_MAX_SEGMENT_SIZE	= 65536,
1307	BLK_SEG_BOUNDARY_MASK	= 0xFFFFFFFFUL,
1308};
1309
1310static inline struct queue_limits *bdev_limits(struct block_device *bdev)
1311{
1312	return &bdev_get_queue(bdev)->limits;
1313}
1314
1315static inline unsigned long queue_segment_boundary(const struct request_queue *q)
1316{
1317	return q->limits.seg_boundary_mask;
1318}
1319
1320static inline unsigned long queue_virt_boundary(const struct request_queue *q)
1321{
1322	return q->limits.virt_boundary_mask;
1323}
1324
1325static inline unsigned int queue_max_sectors(const struct request_queue *q)
1326{
1327	return q->limits.max_sectors;
1328}
1329
1330static inline unsigned int queue_max_bytes(struct request_queue *q)
1331{
1332	return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9;
1333}
1334
1335static inline unsigned int queue_max_hw_sectors(const struct request_queue *q)
1336{
1337	return q->limits.max_hw_sectors;
1338}
1339
1340static inline unsigned short queue_max_segments(const struct request_queue *q)
1341{
1342	return q->limits.max_segments;
1343}
1344
1345static inline unsigned short queue_max_discard_segments(const struct request_queue *q)
1346{
1347	return q->limits.max_discard_segments;
1348}
1349
1350static inline unsigned int queue_max_segment_size(const struct request_queue *q)
1351{
1352	return q->limits.max_segment_size;
1353}
1354
1355static inline bool queue_emulates_zone_append(struct request_queue *q)
1356{
1357	return blk_queue_is_zoned(q) && !q->limits.max_hw_zone_append_sectors;
1358}
1359
1360static inline bool bdev_emulates_zone_append(struct block_device *bdev)
1361{
1362	return queue_emulates_zone_append(bdev_get_queue(bdev));
1363}
1364
1365static inline unsigned int
1366bdev_max_zone_append_sectors(struct block_device *bdev)
1367{
1368	return bdev_limits(bdev)->max_zone_append_sectors;
1369}
1370
1371static inline unsigned int bdev_max_segments(struct block_device *bdev)
1372{
1373	return queue_max_segments(bdev_get_queue(bdev));
1374}
1375
1376static inline unsigned short bdev_max_write_streams(struct block_device *bdev)
1377{
1378	if (bdev_is_partition(bdev))
1379		return 0;
1380	return bdev_limits(bdev)->max_write_streams;
1381}
1382
1383static inline unsigned queue_logical_block_size(const struct request_queue *q)
1384{
1385	return q->limits.logical_block_size;
1386}
1387
1388static inline unsigned int bdev_logical_block_size(struct block_device *bdev)
1389{
1390	return queue_logical_block_size(bdev_get_queue(bdev));
1391}
1392
1393static inline unsigned int queue_physical_block_size(const struct request_queue *q)
1394{
1395	return q->limits.physical_block_size;
1396}
1397
1398static inline unsigned int bdev_physical_block_size(struct block_device *bdev)
1399{
1400	return queue_physical_block_size(bdev_get_queue(bdev));
1401}
1402
1403static inline unsigned int queue_io_min(const struct request_queue *q)
1404{
1405	return q->limits.io_min;
1406}
1407
1408static inline unsigned int bdev_io_min(struct block_device *bdev)
1409{
1410	return queue_io_min(bdev_get_queue(bdev));
1411}
1412
1413static inline unsigned int queue_io_opt(const struct request_queue *q)
1414{
1415	return q->limits.io_opt;
1416}
1417
1418static inline unsigned int bdev_io_opt(struct block_device *bdev)
1419{
1420	return queue_io_opt(bdev_get_queue(bdev));
1421}
1422
1423static inline unsigned int
1424queue_zone_write_granularity(const struct request_queue *q)
1425{
1426	return q->limits.zone_write_granularity;
1427}
1428
1429static inline unsigned int
1430bdev_zone_write_granularity(struct block_device *bdev)
1431{
1432	return queue_zone_write_granularity(bdev_get_queue(bdev));
1433}
1434
1435int bdev_alignment_offset(struct block_device *bdev);
1436unsigned int bdev_discard_alignment(struct block_device *bdev);
1437
1438static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev)
1439{
1440	return bdev_limits(bdev)->max_discard_sectors;
1441}
1442
1443static inline unsigned int bdev_discard_granularity(struct block_device *bdev)
1444{
1445	return bdev_limits(bdev)->discard_granularity;
1446}
1447
1448static inline unsigned int
1449bdev_max_secure_erase_sectors(struct block_device *bdev)
1450{
1451	return bdev_limits(bdev)->max_secure_erase_sectors;
1452}
1453
1454static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev)
1455{
1456	return bdev_limits(bdev)->max_write_zeroes_sectors;
1457}
1458
1459static inline unsigned int
1460bdev_write_zeroes_unmap_sectors(struct block_device *bdev)
1461{
1462	return bdev_limits(bdev)->max_wzeroes_unmap_sectors;
1463}
1464
1465static inline bool bdev_nonrot(struct block_device *bdev)
1466{
1467	return blk_queue_nonrot(bdev_get_queue(bdev));
1468}
1469
1470static inline bool bdev_synchronous(struct block_device *bdev)
1471{
1472	return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS;
1473}
1474
1475static inline bool bdev_stable_writes(struct block_device *bdev)
1476{
1477	struct request_queue *q = bdev_get_queue(bdev);
1478
1479	if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) &&
1480	    q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE)
1481		return true;
1482	return q->limits.features & BLK_FEAT_STABLE_WRITES;
1483}
1484
1485static inline bool blk_queue_write_cache(struct request_queue *q)
1486{
1487	return (q->limits.features & BLK_FEAT_WRITE_CACHE) &&
1488		!(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED);
1489}
1490
1491static inline bool bdev_write_cache(struct block_device *bdev)
1492{
1493	return blk_queue_write_cache(bdev_get_queue(bdev));
1494}
1495
1496static inline bool bdev_fua(struct block_device *bdev)
1497{
1498	return bdev_limits(bdev)->features & BLK_FEAT_FUA;
1499}
1500
1501static inline bool bdev_nowait(struct block_device *bdev)
1502{
1503	return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT;
1504}
1505
1506static inline bool bdev_is_zoned(struct block_device *bdev)
1507{
1508	return blk_queue_is_zoned(bdev_get_queue(bdev));
1509}
1510
1511static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec)
1512{
1513	return disk_zone_no(bdev->bd_disk, sec);
1514}
1515
1516static inline sector_t bdev_zone_sectors(struct block_device *bdev)
1517{
1518	struct request_queue *q = bdev_get_queue(bdev);
1519
1520	if (!blk_queue_is_zoned(q))
1521		return 0;
1522	return q->limits.chunk_sectors;
1523}
1524
1525static inline sector_t bdev_zone_start(struct block_device *bdev,
1526				       sector_t sector)
1527{
1528	return sector & ~(bdev_zone_sectors(bdev) - 1);
1529}
1530
1531static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev,
1532						   sector_t sector)
1533{
1534	return sector & (bdev_zone_sectors(bdev) - 1);
1535}
1536
1537static inline sector_t bio_offset_from_zone_start(struct bio *bio)
1538{
1539	return bdev_offset_from_zone_start(bio->bi_bdev,
1540					   bio->bi_iter.bi_sector);
1541}
1542
1543static inline bool bdev_is_zone_start(struct block_device *bdev,
1544				      sector_t sector)
1545{
1546	return bdev_offset_from_zone_start(bdev, sector) == 0;
1547}
1548
1549/* Check whether @sector is a multiple of the zone size. */
1550static inline bool bdev_is_zone_aligned(struct block_device *bdev,
1551					sector_t sector)
1552{
1553	return bdev_is_zone_start(bdev, sector);
1554}
1555
1556int blk_zone_issue_zeroout(struct block_device *bdev, sector_t sector,
1557			   sector_t nr_sects, gfp_t gfp_mask);
1558
1559static inline unsigned int queue_dma_alignment(const struct request_queue *q)
1560{
1561	return q->limits.dma_alignment;
1562}
1563
1564static inline unsigned int
1565queue_atomic_write_unit_max_bytes(const struct request_queue *q)
1566{
1567	return q->limits.atomic_write_unit_max;
1568}
1569
1570static inline unsigned int
1571queue_atomic_write_unit_min_bytes(const struct request_queue *q)
1572{
1573	return q->limits.atomic_write_unit_min;
1574}
1575
1576static inline unsigned int
1577queue_atomic_write_boundary_bytes(const struct request_queue *q)
1578{
1579	return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT;
1580}
1581
1582static inline unsigned int
1583queue_atomic_write_max_bytes(const struct request_queue *q)
1584{
1585	return q->limits.atomic_write_max_sectors << SECTOR_SHIFT;
1586}
1587
1588static inline unsigned int bdev_dma_alignment(struct block_device *bdev)
1589{
1590	return queue_dma_alignment(bdev_get_queue(bdev));
1591}
1592
1593static inline unsigned int
1594blk_lim_dma_alignment_and_pad(struct queue_limits *lim)
1595{
1596	return lim->dma_alignment | lim->dma_pad_mask;
1597}
1598
1599static inline bool blk_rq_aligned(struct request_queue *q, unsigned long addr,
1600				 unsigned int len)
1601{
1602	unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits);
1603
1604	return !(addr & alignment) && !(len & alignment);
1605}
1606
1607/* assumes size > 256 */
1608static inline unsigned int blksize_bits(unsigned int size)
1609{
1610	return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT;
1611}
1612
1613int kblockd_schedule_work(struct work_struct *work);
1614int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay);
1615
1616#define MODULE_ALIAS_BLOCKDEV(major,minor) \
1617	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
1618#define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
1619	MODULE_ALIAS("block-major-" __stringify(major) "-*")
1620
1621#ifdef CONFIG_BLK_INLINE_ENCRYPTION
1622
1623bool blk_crypto_register(struct blk_crypto_profile *profile,
1624			 struct request_queue *q);
1625
1626#else /* CONFIG_BLK_INLINE_ENCRYPTION */
1627
1628static inline bool blk_crypto_register(struct blk_crypto_profile *profile,
1629				       struct request_queue *q)
1630{
1631	return true;
1632}
1633
1634#endif /* CONFIG_BLK_INLINE_ENCRYPTION */
1635
1636enum blk_unique_id {
1637	/* these match the Designator Types specified in SPC */
1638	BLK_UID_T10	= 1,
1639	BLK_UID_EUI64	= 2,
1640	BLK_UID_NAA	= 3,
1641};
1642
1643struct block_device_operations {
1644	void (*submit_bio)(struct bio *bio);
1645	int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob,
1646			unsigned int flags);
1647	int (*open)(struct gendisk *disk, blk_mode_t mode);
1648	void (*release)(struct gendisk *disk);
1649	int (*ioctl)(struct block_device *bdev, blk_mode_t mode,
1650			unsigned cmd, unsigned long arg);
1651	int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode,
1652			unsigned cmd, unsigned long arg);
1653	unsigned int (*check_events) (struct gendisk *disk,
1654				      unsigned int clearing);
1655	void (*unlock_native_capacity) (struct gendisk *);
1656	int (*getgeo)(struct gendisk *, struct hd_geometry *);
1657	int (*set_read_only)(struct block_device *bdev, bool ro);
1658	void (*free_disk)(struct gendisk *disk);
1659	/* this callback is with swap_lock and sometimes page table lock held */
1660	void (*swap_slot_free_notify) (struct block_device *, unsigned long);
1661	int (*report_zones)(struct gendisk *, sector_t sector,
1662			    unsigned int nr_zones,
1663			    struct blk_report_zones_args *args);
1664	char *(*devnode)(struct gendisk *disk, umode_t *mode);
1665	/* returns the length of the identifier or a negative errno: */
1666	int (*get_unique_id)(struct gendisk *disk, u8 id[16],
1667			enum blk_unique_id id_type);
1668	struct module *owner;
1669	const struct pr_ops *pr_ops;
1670
1671	/*
1672	 * Special callback for probing GPT entry at a given sector.
1673	 * Needed by Android devices, used by GPT scanner and MMC blk
1674	 * driver.
1675	 */
1676	int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector);
1677};
1678
1679#ifdef CONFIG_COMPAT
1680extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t,
1681				      unsigned int, unsigned long);
1682#else
1683#define blkdev_compat_ptr_ioctl NULL
1684#endif
1685
1686static inline void blk_wake_io_task(struct task_struct *waiter)
1687{
1688	/*
1689	 * If we're polling, the task itself is doing the completions. For
1690	 * that case, we don't need to signal a wakeup, it's enough to just
1691	 * mark us as RUNNING.
1692	 */
1693	if (waiter == current)
1694		__set_current_state(TASK_RUNNING);
1695	else
1696		wake_up_process(waiter);
1697}
1698
1699unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op,
1700				 unsigned long start_time);
1701void bdev_end_io_acct(struct block_device *bdev, enum req_op op,
1702		      unsigned int sectors, unsigned long start_time);
1703
1704unsigned long bio_start_io_acct(struct bio *bio);
1705void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1706		struct block_device *orig_bdev);
1707
1708/**
1709 * bio_end_io_acct - end I/O accounting for bio based drivers
1710 * @bio:	bio to end account for
1711 * @start_time:	start time returned by bio_start_io_acct()
1712 */
1713static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time)
1714{
1715	return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev);
1716}
1717
1718int bdev_validate_blocksize(struct block_device *bdev, int block_size);
1719int set_blocksize(struct file *file, int size);
1720
1721int lookup_bdev(const char *pathname, dev_t *dev);
1722
1723void blkdev_show(struct seq_file *seqf, off_t offset);
1724
1725#define BDEVNAME_SIZE	32	/* Largest string for a blockdev identifier */
1726#define BDEVT_SIZE	10	/* Largest string for MAJ:MIN for blkdev */
1727#ifdef CONFIG_BLOCK
1728#define BLKDEV_MAJOR_MAX	512
1729#else
1730#define BLKDEV_MAJOR_MAX	0
1731#endif
1732
1733struct blk_holder_ops {
1734	void (*mark_dead)(struct block_device *bdev, bool surprise);
1735
1736	/*
1737	 * Sync the file system mounted on the block device.
1738	 */
1739	void (*sync)(struct block_device *bdev);
1740
1741	/*
1742	 * Freeze the file system mounted on the block device.
1743	 */
1744	int (*freeze)(struct block_device *bdev);
1745
1746	/*
1747	 * Thaw the file system mounted on the block device.
1748	 */
1749	int (*thaw)(struct block_device *bdev);
1750};
1751
1752/*
1753 * For filesystems using @fs_holder_ops, the @holder argument passed to
1754 * helpers used to open and claim block devices via
1755 * bd_prepare_to_claim() must point to a superblock.
1756 */
1757extern const struct blk_holder_ops fs_holder_ops;
1758
1759/*
1760 * Return the correct open flags for blkdev_get_by_* for super block flags
1761 * as stored in sb->s_flags.
1762 */
1763#define sb_open_mode(flags) \
1764	(BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \
1765	 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE))
1766
1767struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder,
1768		const struct blk_holder_ops *hops);
1769struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode,
1770		void *holder, const struct blk_holder_ops *hops);
1771int bd_prepare_to_claim(struct block_device *bdev, void *holder,
1772		const struct blk_holder_ops *hops);
1773void bd_abort_claiming(struct block_device *bdev, void *holder);
1774
1775struct block_device *I_BDEV(struct inode *inode);
1776struct block_device *file_bdev(struct file *bdev_file);
1777bool disk_live(struct gendisk *disk);
1778unsigned int block_size(struct block_device *bdev);
1779
1780#ifdef CONFIG_BLOCK
1781void invalidate_bdev(struct block_device *bdev);
1782int sync_blockdev(struct block_device *bdev);
1783int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend);
1784int sync_blockdev_nowait(struct block_device *bdev);
1785void sync_bdevs(bool wait);
1786void bdev_statx(const struct path *path, struct kstat *stat, u32 request_mask);
1787void printk_all_partitions(void);
1788int __init early_lookup_bdev(const char *pathname, dev_t *dev);
1789#else
1790static inline void invalidate_bdev(struct block_device *bdev)
1791{
1792}
1793static inline int sync_blockdev(struct block_device *bdev)
1794{
1795	return 0;
1796}
1797static inline int sync_blockdev_nowait(struct block_device *bdev)
1798{
1799	return 0;
1800}
1801static inline void sync_bdevs(bool wait)
1802{
1803}
1804static inline void bdev_statx(const struct path *path, struct kstat *stat,
1805		u32 request_mask)
1806{
1807}
1808static inline void printk_all_partitions(void)
1809{
1810}
1811static inline int early_lookup_bdev(const char *pathname, dev_t *dev)
1812{
1813	return -EINVAL;
1814}
1815#endif /* CONFIG_BLOCK */
1816
1817int bdev_freeze(struct block_device *bdev);
1818int bdev_thaw(struct block_device *bdev);
1819void bdev_fput(struct file *bdev_file);
1820
1821struct io_comp_batch {
1822	struct rq_list req_list;
1823	bool need_ts;
1824	void (*complete)(struct io_comp_batch *);
1825};
1826
1827static inline bool blk_atomic_write_start_sect_aligned(sector_t sector,
1828						struct queue_limits *limits)
1829{
1830	unsigned int alignment = max(limits->atomic_write_hw_unit_min,
1831				limits->atomic_write_hw_boundary);
1832
1833	return IS_ALIGNED(sector, alignment >> SECTOR_SHIFT);
1834}
1835
1836static inline bool bdev_can_atomic_write(struct block_device *bdev)
1837{
1838	struct request_queue *bd_queue = bdev->bd_queue;
1839	struct queue_limits *limits = &bd_queue->limits;
1840
1841	if (!limits->atomic_write_unit_min)
1842		return false;
1843
1844	if (bdev_is_partition(bdev))
1845		return blk_atomic_write_start_sect_aligned(bdev->bd_start_sect,
1846							limits);
1847
1848	return true;
1849}
1850
1851static inline unsigned int
1852bdev_atomic_write_unit_min_bytes(struct block_device *bdev)
1853{
1854	if (!bdev_can_atomic_write(bdev))
1855		return 0;
1856	return queue_atomic_write_unit_min_bytes(bdev_get_queue(bdev));
1857}
1858
1859static inline unsigned int
1860bdev_atomic_write_unit_max_bytes(struct block_device *bdev)
1861{
1862	if (!bdev_can_atomic_write(bdev))
1863		return 0;
1864	return queue_atomic_write_unit_max_bytes(bdev_get_queue(bdev));
1865}
1866
1867static inline int bio_split_rw_at(struct bio *bio,
1868		const struct queue_limits *lim,
1869		unsigned *segs, unsigned max_bytes)
1870{
1871	return bio_split_io_at(bio, lim, segs, max_bytes, lim->dma_alignment);
1872}
1873
1874#define DEFINE_IO_COMP_BATCH(name)	struct io_comp_batch name = { }
1875
1876#endif /* _LINUX_BLKDEV_H */