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 30 struct module; 31 struct request_queue; 32 struct elevator_queue; 33 struct blk_trace; 34 struct request; 35 struct sg_io_hdr; 36 struct blkcg_gq; 37 struct blk_flush_queue; 38 struct kiocb; 39 struct pr_ops; 40 struct rq_qos; 41 struct blk_queue_stats; 42 struct blk_stat_callback; 43 struct blk_crypto_profile; 44 45 extern const struct device_type disk_type; 46 extern const struct device_type part_type; 47 extern const struct class block_class; 48 49 /* 50 * Maximum number of blkcg policies allowed to be registered concurrently. 51 * Defined here to simplify include dependency. 52 */ 53 #define BLKCG_MAX_POLS 6 54 55 #define DISK_MAX_PARTS 256 56 #define DISK_NAME_LEN 32 57 58 #define PARTITION_META_INFO_VOLNAMELTH 64 59 /* 60 * Enough for the string representation of any kind of UUID plus NULL. 61 * EFI UUID is 36 characters. MSDOS UUID is 11 characters. 62 */ 63 #define PARTITION_META_INFO_UUIDLTH (UUID_STRING_LEN + 1) 64 65 struct partition_meta_info { 66 char uuid[PARTITION_META_INFO_UUIDLTH]; 67 u8 volname[PARTITION_META_INFO_VOLNAMELTH]; 68 }; 69 70 /** 71 * DOC: genhd capability flags 72 * 73 * ``GENHD_FL_REMOVABLE``: indicates that the block device gives access to 74 * removable media. When set, the device remains present even when media is not 75 * inserted. Shall not be set for devices which are removed entirely when the 76 * media is removed. 77 * 78 * ``GENHD_FL_HIDDEN``: the block device is hidden; it doesn't produce events, 79 * doesn't appear in sysfs, and can't be opened from userspace or using 80 * blkdev_get*. Used for the underlying components of multipath devices. 81 * 82 * ``GENHD_FL_NO_PART``: partition support is disabled. The kernel will not 83 * scan for partitions from add_disk, and users can't add partitions manually. 84 * 85 */ 86 enum { 87 GENHD_FL_REMOVABLE = 1 << 0, 88 GENHD_FL_HIDDEN = 1 << 1, 89 GENHD_FL_NO_PART = 1 << 2, 90 }; 91 92 enum { 93 DISK_EVENT_MEDIA_CHANGE = 1 << 0, /* media changed */ 94 DISK_EVENT_EJECT_REQUEST = 1 << 1, /* eject requested */ 95 }; 96 97 enum { 98 /* Poll even if events_poll_msecs is unset */ 99 DISK_EVENT_FLAG_POLL = 1 << 0, 100 /* Forward events to udev */ 101 DISK_EVENT_FLAG_UEVENT = 1 << 1, 102 /* Block event polling when open for exclusive write */ 103 DISK_EVENT_FLAG_BLOCK_ON_EXCL_WRITE = 1 << 2, 104 }; 105 106 struct disk_events; 107 struct badblocks; 108 109 enum blk_integrity_checksum { 110 BLK_INTEGRITY_CSUM_NONE = 0, 111 BLK_INTEGRITY_CSUM_IP = 1, 112 BLK_INTEGRITY_CSUM_CRC = 2, 113 BLK_INTEGRITY_CSUM_CRC64 = 3, 114 } __packed ; 115 116 struct blk_integrity { 117 unsigned char flags; 118 enum blk_integrity_checksum csum_type; 119 unsigned char tuple_size; 120 unsigned char pi_offset; 121 unsigned char interval_exp; 122 unsigned char tag_size; 123 }; 124 125 typedef unsigned int __bitwise blk_mode_t; 126 127 /* open for reading */ 128 #define BLK_OPEN_READ ((__force blk_mode_t)(1 << 0)) 129 /* open for writing */ 130 #define BLK_OPEN_WRITE ((__force blk_mode_t)(1 << 1)) 131 /* open exclusively (vs other exclusive openers */ 132 #define BLK_OPEN_EXCL ((__force blk_mode_t)(1 << 2)) 133 /* opened with O_NDELAY */ 134 #define BLK_OPEN_NDELAY ((__force blk_mode_t)(1 << 3)) 135 /* open for "writes" only for ioctls (specialy hack for floppy.c) */ 136 #define BLK_OPEN_WRITE_IOCTL ((__force blk_mode_t)(1 << 4)) 137 /* open is exclusive wrt all other BLK_OPEN_WRITE opens to the device */ 138 #define BLK_OPEN_RESTRICT_WRITES ((__force blk_mode_t)(1 << 5)) 139 /* return partition scanning errors */ 140 #define BLK_OPEN_STRICT_SCAN ((__force blk_mode_t)(1 << 6)) 141 142 struct gendisk { 143 /* 144 * major/first_minor/minors should not be set by any new driver, the 145 * block core will take care of allocating them automatically. 146 */ 147 int major; 148 int first_minor; 149 int minors; 150 151 char disk_name[DISK_NAME_LEN]; /* name of major driver */ 152 153 unsigned short events; /* supported events */ 154 unsigned short event_flags; /* flags related to event processing */ 155 156 struct xarray part_tbl; 157 struct block_device *part0; 158 159 const struct block_device_operations *fops; 160 struct request_queue *queue; 161 void *private_data; 162 163 struct bio_set bio_split; 164 165 int flags; 166 unsigned long state; 167 #define GD_NEED_PART_SCAN 0 168 #define GD_READ_ONLY 1 169 #define GD_DEAD 2 170 #define GD_NATIVE_CAPACITY 3 171 #define GD_ADDED 4 172 #define GD_SUPPRESS_PART_SCAN 5 173 #define GD_OWNS_QUEUE 6 174 175 struct mutex open_mutex; /* open/close mutex */ 176 unsigned open_partitions; /* number of open partitions */ 177 178 struct backing_dev_info *bdi; 179 struct kobject queue_kobj; /* the queue/ directory */ 180 struct kobject *slave_dir; 181 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 182 struct list_head slave_bdevs; 183 #endif 184 struct timer_rand_state *random; 185 atomic_t sync_io; /* RAID */ 186 struct disk_events *ev; 187 188 #ifdef CONFIG_BLK_DEV_ZONED 189 /* 190 * Zoned block device information. Reads of this information must be 191 * protected with blk_queue_enter() / blk_queue_exit(). Modifying this 192 * information is only allowed while no requests are being processed. 193 * See also blk_mq_freeze_queue() and blk_mq_unfreeze_queue(). 194 */ 195 unsigned int nr_zones; 196 unsigned int zone_capacity; 197 unsigned int last_zone_capacity; 198 unsigned long __rcu *conv_zones_bitmap; 199 unsigned int zone_wplugs_hash_bits; 200 spinlock_t zone_wplugs_lock; 201 struct mempool_s *zone_wplugs_pool; 202 struct hlist_head *zone_wplugs_hash; 203 struct workqueue_struct *zone_wplugs_wq; 204 #endif /* CONFIG_BLK_DEV_ZONED */ 205 206 #if IS_ENABLED(CONFIG_CDROM) 207 struct cdrom_device_info *cdi; 208 #endif 209 int node_id; 210 struct badblocks *bb; 211 struct lockdep_map lockdep_map; 212 u64 diskseq; 213 blk_mode_t open_mode; 214 215 /* 216 * Independent sector access ranges. This is always NULL for 217 * devices that do not have multiple independent access ranges. 218 */ 219 struct blk_independent_access_ranges *ia_ranges; 220 }; 221 222 /** 223 * disk_openers - returns how many openers are there for a disk 224 * @disk: disk to check 225 * 226 * This returns the number of openers for a disk. Note that this value is only 227 * stable if disk->open_mutex is held. 228 * 229 * Note: Due to a quirk in the block layer open code, each open partition is 230 * only counted once even if there are multiple openers. 231 */ 232 static inline unsigned int disk_openers(struct gendisk *disk) 233 { 234 return atomic_read(&disk->part0->bd_openers); 235 } 236 237 /** 238 * disk_has_partscan - return %true if partition scanning is enabled on a disk 239 * @disk: disk to check 240 * 241 * Returns %true if partitions scanning is enabled for @disk, or %false if 242 * partition scanning is disabled either permanently or temporarily. 243 */ 244 static inline bool disk_has_partscan(struct gendisk *disk) 245 { 246 return !(disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN)) && 247 !test_bit(GD_SUPPRESS_PART_SCAN, &disk->state); 248 } 249 250 /* 251 * The gendisk is refcounted by the part0 block_device, and the bd_device 252 * therein is also used for device model presentation in sysfs. 253 */ 254 #define dev_to_disk(device) \ 255 (dev_to_bdev(device)->bd_disk) 256 #define disk_to_dev(disk) \ 257 (&((disk)->part0->bd_device)) 258 259 #if IS_REACHABLE(CONFIG_CDROM) 260 #define disk_to_cdi(disk) ((disk)->cdi) 261 #else 262 #define disk_to_cdi(disk) NULL 263 #endif 264 265 static inline dev_t disk_devt(struct gendisk *disk) 266 { 267 return MKDEV(disk->major, disk->first_minor); 268 } 269 270 /* blk_validate_limits() validates bsize, so drivers don't usually need to */ 271 static inline int blk_validate_block_size(unsigned long bsize) 272 { 273 if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(bsize)) 274 return -EINVAL; 275 276 return 0; 277 } 278 279 static inline bool blk_op_is_passthrough(blk_opf_t op) 280 { 281 op &= REQ_OP_MASK; 282 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 283 } 284 285 /* flags set by the driver in queue_limits.features */ 286 typedef unsigned int __bitwise blk_features_t; 287 288 /* supports a volatile write cache */ 289 #define BLK_FEAT_WRITE_CACHE ((__force blk_features_t)(1u << 0)) 290 291 /* supports passing on the FUA bit */ 292 #define BLK_FEAT_FUA ((__force blk_features_t)(1u << 1)) 293 294 /* rotational device (hard drive or floppy) */ 295 #define BLK_FEAT_ROTATIONAL ((__force blk_features_t)(1u << 2)) 296 297 /* contributes to the random number pool */ 298 #define BLK_FEAT_ADD_RANDOM ((__force blk_features_t)(1u << 3)) 299 300 /* do disk/partitions IO accounting */ 301 #define BLK_FEAT_IO_STAT ((__force blk_features_t)(1u << 4)) 302 303 /* don't modify data until writeback is done */ 304 #define BLK_FEAT_STABLE_WRITES ((__force blk_features_t)(1u << 5)) 305 306 /* always completes in submit context */ 307 #define BLK_FEAT_SYNCHRONOUS ((__force blk_features_t)(1u << 6)) 308 309 /* supports REQ_NOWAIT */ 310 #define BLK_FEAT_NOWAIT ((__force blk_features_t)(1u << 7)) 311 312 /* supports DAX */ 313 #define BLK_FEAT_DAX ((__force blk_features_t)(1u << 8)) 314 315 /* supports I/O polling */ 316 #define BLK_FEAT_POLL ((__force blk_features_t)(1u << 9)) 317 318 /* is a zoned device */ 319 #define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10)) 320 321 /* supports PCI(e) p2p requests */ 322 #define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12)) 323 324 /* skip this queue in blk_mq_(un)quiesce_tagset */ 325 #define BLK_FEAT_SKIP_TAGSET_QUIESCE ((__force blk_features_t)(1u << 13)) 326 327 /* bounce all highmem pages */ 328 #define BLK_FEAT_BOUNCE_HIGH ((__force blk_features_t)(1u << 14)) 329 330 /* undocumented magic for bcache */ 331 #define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \ 332 ((__force blk_features_t)(1u << 15)) 333 334 /* atomic writes enabled */ 335 #define BLK_FEAT_ATOMIC_WRITES \ 336 ((__force blk_features_t)(1u << 16)) 337 338 /* 339 * Flags automatically inherited when stacking limits. 340 */ 341 #define BLK_FEAT_INHERIT_MASK \ 342 (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ 343 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH | \ 344 BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE) 345 346 /* internal flags in queue_limits.flags */ 347 typedef unsigned int __bitwise blk_flags_t; 348 349 /* do not send FLUSH/FUA commands despite advertising a write cache */ 350 #define BLK_FLAG_WRITE_CACHE_DISABLED ((__force blk_flags_t)(1u << 0)) 351 352 /* I/O topology is misaligned */ 353 #define BLK_FLAG_MISALIGNED ((__force blk_flags_t)(1u << 1)) 354 355 /* passthrough command IO accounting */ 356 #define BLK_FLAG_IOSTATS_PASSTHROUGH ((__force blk_flags_t)(1u << 2)) 357 358 struct queue_limits { 359 blk_features_t features; 360 blk_flags_t flags; 361 unsigned long seg_boundary_mask; 362 unsigned long virt_boundary_mask; 363 364 unsigned int max_hw_sectors; 365 unsigned int max_dev_sectors; 366 unsigned int chunk_sectors; 367 unsigned int max_sectors; 368 unsigned int max_user_sectors; 369 unsigned int max_segment_size; 370 unsigned int physical_block_size; 371 unsigned int logical_block_size; 372 unsigned int alignment_offset; 373 unsigned int io_min; 374 unsigned int io_opt; 375 unsigned int max_discard_sectors; 376 unsigned int max_hw_discard_sectors; 377 unsigned int max_user_discard_sectors; 378 unsigned int max_secure_erase_sectors; 379 unsigned int max_write_zeroes_sectors; 380 unsigned int max_hw_zone_append_sectors; 381 unsigned int max_zone_append_sectors; 382 unsigned int discard_granularity; 383 unsigned int discard_alignment; 384 unsigned int zone_write_granularity; 385 386 /* atomic write limits */ 387 unsigned int atomic_write_hw_max; 388 unsigned int atomic_write_max_sectors; 389 unsigned int atomic_write_hw_boundary; 390 unsigned int atomic_write_boundary_sectors; 391 unsigned int atomic_write_hw_unit_min; 392 unsigned int atomic_write_unit_min; 393 unsigned int atomic_write_hw_unit_max; 394 unsigned int atomic_write_unit_max; 395 396 unsigned short max_segments; 397 unsigned short max_integrity_segments; 398 unsigned short max_discard_segments; 399 400 unsigned int max_open_zones; 401 unsigned int max_active_zones; 402 403 /* 404 * Drivers that set dma_alignment to less than 511 must be prepared to 405 * handle individual bvec's that are not a multiple of a SECTOR_SIZE 406 * due to possible offsets. 407 */ 408 unsigned int dma_alignment; 409 unsigned int dma_pad_mask; 410 411 struct blk_integrity integrity; 412 }; 413 414 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, 415 void *data); 416 417 #define BLK_ALL_ZONES ((unsigned int)-1) 418 int blkdev_report_zones(struct block_device *bdev, sector_t sector, 419 unsigned int nr_zones, report_zones_cb cb, void *data); 420 int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, 421 sector_t sectors, sector_t nr_sectors); 422 int blk_revalidate_disk_zones(struct gendisk *disk); 423 424 /* 425 * Independent access ranges: struct blk_independent_access_range describes 426 * a range of contiguous sectors that can be accessed using device command 427 * execution resources that are independent from the resources used for 428 * other access ranges. This is typically found with single-LUN multi-actuator 429 * HDDs where each access range is served by a different set of heads. 430 * The set of independent ranges supported by the device is defined using 431 * struct blk_independent_access_ranges. The independent ranges must not overlap 432 * and must include all sectors within the disk capacity (no sector holes 433 * allowed). 434 * For a device with multiple ranges, requests targeting sectors in different 435 * ranges can be executed in parallel. A request can straddle an access range 436 * boundary. 437 */ 438 struct blk_independent_access_range { 439 struct kobject kobj; 440 sector_t sector; 441 sector_t nr_sectors; 442 }; 443 444 struct blk_independent_access_ranges { 445 struct kobject kobj; 446 bool sysfs_registered; 447 unsigned int nr_ia_ranges; 448 struct blk_independent_access_range ia_range[]; 449 }; 450 451 struct request_queue { 452 /* 453 * The queue owner gets to use this for whatever they like. 454 * ll_rw_blk doesn't touch it. 455 */ 456 void *queuedata; 457 458 struct elevator_queue *elevator; 459 460 const struct blk_mq_ops *mq_ops; 461 462 /* sw queues */ 463 struct blk_mq_ctx __percpu *queue_ctx; 464 465 /* 466 * various queue flags, see QUEUE_* below 467 */ 468 unsigned long queue_flags; 469 470 unsigned int rq_timeout; 471 472 unsigned int queue_depth; 473 474 refcount_t refs; 475 476 /* hw dispatch queues */ 477 unsigned int nr_hw_queues; 478 struct xarray hctx_table; 479 480 struct percpu_ref q_usage_counter; 481 struct lock_class_key io_lock_cls_key; 482 struct lockdep_map io_lockdep_map; 483 484 struct lock_class_key q_lock_cls_key; 485 struct lockdep_map q_lockdep_map; 486 487 struct request *last_merge; 488 489 spinlock_t queue_lock; 490 491 int quiesce_depth; 492 493 struct gendisk *disk; 494 495 /* 496 * mq queue kobject 497 */ 498 struct kobject *mq_kobj; 499 500 struct queue_limits limits; 501 502 #ifdef CONFIG_PM 503 struct device *dev; 504 enum rpm_status rpm_status; 505 #endif 506 507 /* 508 * Number of contexts that have called blk_set_pm_only(). If this 509 * counter is above zero then only RQF_PM requests are processed. 510 */ 511 atomic_t pm_only; 512 513 struct blk_queue_stats *stats; 514 struct rq_qos *rq_qos; 515 struct mutex rq_qos_mutex; 516 517 /* 518 * ida allocated id for this queue. Used to index queues from 519 * ioctx. 520 */ 521 int id; 522 523 /* 524 * queue settings 525 */ 526 unsigned long nr_requests; /* Max # of requests */ 527 528 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 529 struct blk_crypto_profile *crypto_profile; 530 struct kobject *crypto_kobject; 531 #endif 532 533 struct timer_list timeout; 534 struct work_struct timeout_work; 535 536 atomic_t nr_active_requests_shared_tags; 537 538 struct blk_mq_tags *sched_shared_tags; 539 540 struct list_head icq_list; 541 #ifdef CONFIG_BLK_CGROUP 542 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 543 struct blkcg_gq *root_blkg; 544 struct list_head blkg_list; 545 struct mutex blkcg_mutex; 546 #endif 547 548 int node; 549 550 spinlock_t requeue_lock; 551 struct list_head requeue_list; 552 struct delayed_work requeue_work; 553 554 #ifdef CONFIG_BLK_DEV_IO_TRACE 555 struct blk_trace __rcu *blk_trace; 556 #endif 557 /* 558 * for flush operations 559 */ 560 struct blk_flush_queue *fq; 561 struct list_head flush_list; 562 563 /* 564 * Protects against I/O scheduler switching, particularly when updating 565 * q->elevator. Since the elevator update code path may also modify q-> 566 * nr_requests and wbt latency, this lock also protects the sysfs attrs 567 * nr_requests and wbt_lat_usec. Additionally the nr_hw_queues update 568 * may modify hctx tags, reserved-tags and cpumask, so this lock also 569 * helps protect the hctx sysfs/debugfs attrs. To ensure proper locking 570 * order during an elevator or nr_hw_queue update, first freeze the 571 * queue, then acquire ->elevator_lock. 572 */ 573 struct mutex elevator_lock; 574 575 struct mutex sysfs_lock; 576 /* 577 * Protects queue limits and also sysfs attribute read_ahead_kb. 578 */ 579 struct mutex limits_lock; 580 581 /* 582 * for reusing dead hctx instance in case of updating 583 * nr_hw_queues 584 */ 585 struct list_head unused_hctx_list; 586 spinlock_t unused_hctx_lock; 587 588 int mq_freeze_depth; 589 590 #ifdef CONFIG_BLK_DEV_THROTTLING 591 /* Throttle data */ 592 struct throtl_data *td; 593 #endif 594 struct rcu_head rcu_head; 595 #ifdef CONFIG_LOCKDEP 596 struct task_struct *mq_freeze_owner; 597 int mq_freeze_owner_depth; 598 /* 599 * Records disk & queue state in current context, used in unfreeze 600 * queue 601 */ 602 bool mq_freeze_disk_dead; 603 bool mq_freeze_queue_dying; 604 #endif 605 wait_queue_head_t mq_freeze_wq; 606 /* 607 * Protect concurrent access to q_usage_counter by 608 * percpu_ref_kill() and percpu_ref_reinit(). 609 */ 610 struct mutex mq_freeze_lock; 611 612 struct blk_mq_tag_set *tag_set; 613 struct list_head tag_set_list; 614 615 struct dentry *debugfs_dir; 616 struct dentry *sched_debugfs_dir; 617 struct dentry *rqos_debugfs_dir; 618 /* 619 * Serializes all debugfs metadata operations using the above dentries. 620 */ 621 struct mutex debugfs_mutex; 622 }; 623 624 /* Keep blk_queue_flag_name[] in sync with the definitions below */ 625 enum { 626 QUEUE_FLAG_DYING, /* queue being torn down */ 627 QUEUE_FLAG_NOMERGES, /* disable merge attempts */ 628 QUEUE_FLAG_SAME_COMP, /* complete on same CPU-group */ 629 QUEUE_FLAG_FAIL_IO, /* fake timeout */ 630 QUEUE_FLAG_NOXMERGES, /* No extended merges */ 631 QUEUE_FLAG_SAME_FORCE, /* force complete on same CPU */ 632 QUEUE_FLAG_INIT_DONE, /* queue is initialized */ 633 QUEUE_FLAG_STATS, /* track IO start and completion times */ 634 QUEUE_FLAG_REGISTERED, /* queue has been registered to a disk */ 635 QUEUE_FLAG_QUIESCED, /* queue has been quiesced */ 636 QUEUE_FLAG_RQ_ALLOC_TIME, /* record rq->alloc_time_ns */ 637 QUEUE_FLAG_HCTX_ACTIVE, /* at least one blk-mq hctx is active */ 638 QUEUE_FLAG_SQ_SCHED, /* single queue style io dispatch */ 639 QUEUE_FLAG_MAX 640 }; 641 642 #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) 643 644 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 645 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 646 647 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 648 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 649 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 650 #define blk_queue_noxmerges(q) \ 651 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 652 #define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL)) 653 #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) 654 #define blk_queue_passthrough_stat(q) \ 655 ((q)->limits.flags & BLK_FLAG_IOSTATS_PASSTHROUGH) 656 #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) 657 #define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA) 658 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 659 #define blk_queue_rq_alloc_time(q) \ 660 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) 661 #else 662 #define blk_queue_rq_alloc_time(q) false 663 #endif 664 665 #define blk_noretry_request(rq) \ 666 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 667 REQ_FAILFAST_DRIVER)) 668 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 669 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 670 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) 671 #define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags) 672 #define blk_queue_skip_tagset_quiesce(q) \ 673 ((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE) 674 675 extern void blk_set_pm_only(struct request_queue *q); 676 extern void blk_clear_pm_only(struct request_queue *q); 677 678 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 679 680 #define dma_map_bvec(dev, bv, dir, attrs) \ 681 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 682 (dir), (attrs)) 683 684 static inline bool queue_is_mq(struct request_queue *q) 685 { 686 return q->mq_ops; 687 } 688 689 #ifdef CONFIG_PM 690 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 691 { 692 return q->rpm_status; 693 } 694 #else 695 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 696 { 697 return RPM_ACTIVE; 698 } 699 #endif 700 701 static inline bool blk_queue_is_zoned(struct request_queue *q) 702 { 703 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && 704 (q->limits.features & BLK_FEAT_ZONED); 705 } 706 707 #ifdef CONFIG_BLK_DEV_ZONED 708 static inline unsigned int disk_nr_zones(struct gendisk *disk) 709 { 710 return disk->nr_zones; 711 } 712 bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); 713 #else /* CONFIG_BLK_DEV_ZONED */ 714 static inline unsigned int disk_nr_zones(struct gendisk *disk) 715 { 716 return 0; 717 } 718 static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) 719 { 720 return false; 721 } 722 #endif /* CONFIG_BLK_DEV_ZONED */ 723 724 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) 725 { 726 if (!blk_queue_is_zoned(disk->queue)) 727 return 0; 728 return sector >> ilog2(disk->queue->limits.chunk_sectors); 729 } 730 731 static inline unsigned int bdev_nr_zones(struct block_device *bdev) 732 { 733 return disk_nr_zones(bdev->bd_disk); 734 } 735 736 static inline unsigned int bdev_max_open_zones(struct block_device *bdev) 737 { 738 return bdev->bd_disk->queue->limits.max_open_zones; 739 } 740 741 static inline unsigned int bdev_max_active_zones(struct block_device *bdev) 742 { 743 return bdev->bd_disk->queue->limits.max_active_zones; 744 } 745 746 static inline unsigned int blk_queue_depth(struct request_queue *q) 747 { 748 if (q->queue_depth) 749 return q->queue_depth; 750 751 return q->nr_requests; 752 } 753 754 /* 755 * default timeout for SG_IO if none specified 756 */ 757 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 758 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 759 760 /* This should not be used directly - use rq_for_each_segment */ 761 #define for_each_bio(_bio) \ 762 for (; _bio; _bio = _bio->bi_next) 763 764 int __must_check add_disk_fwnode(struct device *parent, struct gendisk *disk, 765 const struct attribute_group **groups, 766 struct fwnode_handle *fwnode); 767 int __must_check device_add_disk(struct device *parent, struct gendisk *disk, 768 const struct attribute_group **groups); 769 static inline int __must_check add_disk(struct gendisk *disk) 770 { 771 return device_add_disk(NULL, disk, NULL); 772 } 773 void del_gendisk(struct gendisk *gp); 774 void invalidate_disk(struct gendisk *disk); 775 void set_disk_ro(struct gendisk *disk, bool read_only); 776 void disk_uevent(struct gendisk *disk, enum kobject_action action); 777 778 static inline u8 bdev_partno(const struct block_device *bdev) 779 { 780 return atomic_read(&bdev->__bd_flags) & BD_PARTNO; 781 } 782 783 static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag) 784 { 785 return atomic_read(&bdev->__bd_flags) & flag; 786 } 787 788 static inline void bdev_set_flag(struct block_device *bdev, unsigned flag) 789 { 790 atomic_or(flag, &bdev->__bd_flags); 791 } 792 793 static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag) 794 { 795 atomic_andnot(flag, &bdev->__bd_flags); 796 } 797 798 static inline bool get_disk_ro(struct gendisk *disk) 799 { 800 return bdev_test_flag(disk->part0, BD_READ_ONLY) || 801 test_bit(GD_READ_ONLY, &disk->state); 802 } 803 804 static inline bool bdev_read_only(struct block_device *bdev) 805 { 806 return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk); 807 } 808 809 bool set_capacity_and_notify(struct gendisk *disk, sector_t size); 810 void disk_force_media_change(struct gendisk *disk); 811 void bdev_mark_dead(struct block_device *bdev, bool surprise); 812 813 void add_disk_randomness(struct gendisk *disk) __latent_entropy; 814 void rand_initialize_disk(struct gendisk *disk); 815 816 static inline sector_t get_start_sect(struct block_device *bdev) 817 { 818 return bdev->bd_start_sect; 819 } 820 821 static inline sector_t bdev_nr_sectors(struct block_device *bdev) 822 { 823 return bdev->bd_nr_sectors; 824 } 825 826 static inline loff_t bdev_nr_bytes(struct block_device *bdev) 827 { 828 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT; 829 } 830 831 static inline sector_t get_capacity(struct gendisk *disk) 832 { 833 return bdev_nr_sectors(disk->part0); 834 } 835 836 static inline u64 sb_bdev_nr_blocks(struct super_block *sb) 837 { 838 return bdev_nr_sectors(sb->s_bdev) >> 839 (sb->s_blocksize_bits - SECTOR_SHIFT); 840 } 841 842 int bdev_disk_changed(struct gendisk *disk, bool invalidate); 843 844 void put_disk(struct gendisk *disk); 845 struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node, 846 struct lock_class_key *lkclass); 847 848 /** 849 * blk_alloc_disk - allocate a gendisk structure 850 * @lim: queue limits to be used for this disk. 851 * @node_id: numa node to allocate on 852 * 853 * Allocate and pre-initialize a gendisk structure for use with BIO based 854 * drivers. 855 * 856 * Returns an ERR_PTR on error, else the allocated disk. 857 * 858 * Context: can sleep 859 */ 860 #define blk_alloc_disk(lim, node_id) \ 861 ({ \ 862 static struct lock_class_key __key; \ 863 \ 864 __blk_alloc_disk(lim, node_id, &__key); \ 865 }) 866 867 int __register_blkdev(unsigned int major, const char *name, 868 void (*probe)(dev_t devt)); 869 #define register_blkdev(major, name) \ 870 __register_blkdev(major, name, NULL) 871 void unregister_blkdev(unsigned int major, const char *name); 872 873 bool disk_check_media_change(struct gendisk *disk); 874 void set_capacity(struct gendisk *disk, sector_t size); 875 876 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 877 int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk); 878 void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk); 879 #else 880 static inline int bd_link_disk_holder(struct block_device *bdev, 881 struct gendisk *disk) 882 { 883 return 0; 884 } 885 static inline void bd_unlink_disk_holder(struct block_device *bdev, 886 struct gendisk *disk) 887 { 888 } 889 #endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */ 890 891 dev_t part_devt(struct gendisk *disk, u8 partno); 892 void inc_diskseq(struct gendisk *disk); 893 void blk_request_module(dev_t devt); 894 895 extern int blk_register_queue(struct gendisk *disk); 896 extern void blk_unregister_queue(struct gendisk *disk); 897 void submit_bio_noacct(struct bio *bio); 898 struct bio *bio_split_to_limits(struct bio *bio); 899 900 extern int blk_lld_busy(struct request_queue *q); 901 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 902 extern void blk_queue_exit(struct request_queue *q); 903 extern void blk_sync_queue(struct request_queue *q); 904 905 /* Helper to convert REQ_OP_XXX to its string format XXX */ 906 extern const char *blk_op_str(enum req_op op); 907 908 int blk_status_to_errno(blk_status_t status); 909 blk_status_t errno_to_blk_status(int errno); 910 const char *blk_status_to_str(blk_status_t status); 911 912 /* only poll the hardware once, don't continue until a completion was found */ 913 #define BLK_POLL_ONESHOT (1 << 0) 914 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags); 915 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, 916 unsigned int flags); 917 918 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 919 { 920 return bdev->bd_queue; /* this is never NULL */ 921 } 922 923 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */ 924 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond); 925 926 static inline unsigned int bio_zone_no(struct bio *bio) 927 { 928 return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector); 929 } 930 931 static inline bool bio_straddles_zones(struct bio *bio) 932 { 933 return bio_sectors(bio) && 934 bio_zone_no(bio) != 935 disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1); 936 } 937 938 /* 939 * Return how much within the boundary is left to be used for I/O at a given 940 * offset. 941 */ 942 static inline unsigned int blk_boundary_sectors_left(sector_t offset, 943 unsigned int boundary_sectors) 944 { 945 if (unlikely(!is_power_of_2(boundary_sectors))) 946 return boundary_sectors - sector_div(offset, boundary_sectors); 947 return boundary_sectors - (offset & (boundary_sectors - 1)); 948 } 949 950 /** 951 * queue_limits_start_update - start an atomic update of queue limits 952 * @q: queue to update 953 * 954 * This functions starts an atomic update of the queue limits. It takes a lock 955 * to prevent other updates and returns a snapshot of the current limits that 956 * the caller can modify. The caller must call queue_limits_commit_update() 957 * to finish the update. 958 * 959 * Context: process context. 960 */ 961 static inline struct queue_limits 962 queue_limits_start_update(struct request_queue *q) 963 { 964 mutex_lock(&q->limits_lock); 965 return q->limits; 966 } 967 int queue_limits_commit_update_frozen(struct request_queue *q, 968 struct queue_limits *lim); 969 int queue_limits_commit_update(struct request_queue *q, 970 struct queue_limits *lim); 971 int queue_limits_set(struct request_queue *q, struct queue_limits *lim); 972 int blk_validate_limits(struct queue_limits *lim); 973 974 /** 975 * queue_limits_cancel_update - cancel an atomic update of queue limits 976 * @q: queue to update 977 * 978 * This functions cancels an atomic update of the queue limits started by 979 * queue_limits_start_update() and should be used when an error occurs after 980 * starting update. 981 */ 982 static inline void queue_limits_cancel_update(struct request_queue *q) 983 { 984 mutex_unlock(&q->limits_lock); 985 } 986 987 /* 988 * These helpers are for drivers that have sloppy feature negotiation and might 989 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O 990 * completion handler when the device returned an indicator that the respective 991 * feature is not actually supported. They are racy and the driver needs to 992 * cope with that. Try to avoid this scheme if you can. 993 */ 994 static inline void blk_queue_disable_discard(struct request_queue *q) 995 { 996 q->limits.max_discard_sectors = 0; 997 } 998 999 static inline void blk_queue_disable_secure_erase(struct request_queue *q) 1000 { 1001 q->limits.max_secure_erase_sectors = 0; 1002 } 1003 1004 static inline void blk_queue_disable_write_zeroes(struct request_queue *q) 1005 { 1006 q->limits.max_write_zeroes_sectors = 0; 1007 } 1008 1009 /* 1010 * Access functions for manipulating queue properties 1011 */ 1012 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1013 extern void blk_set_stacking_limits(struct queue_limits *lim); 1014 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1015 sector_t offset); 1016 void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, 1017 sector_t offset, const char *pfx); 1018 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1019 1020 struct blk_independent_access_ranges * 1021 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges); 1022 void disk_set_independent_access_ranges(struct gendisk *disk, 1023 struct blk_independent_access_ranges *iars); 1024 1025 bool __must_check blk_get_queue(struct request_queue *); 1026 extern void blk_put_queue(struct request_queue *); 1027 1028 void blk_mark_disk_dead(struct gendisk *disk); 1029 1030 struct rq_list { 1031 struct request *head; 1032 struct request *tail; 1033 }; 1034 1035 #ifdef CONFIG_BLOCK 1036 /* 1037 * blk_plug permits building a queue of related requests by holding the I/O 1038 * fragments for a short period. This allows merging of sequential requests 1039 * into single larger request. As the requests are moved from a per-task list to 1040 * the device's request_queue in a batch, this results in improved scalability 1041 * as the lock contention for request_queue lock is reduced. 1042 * 1043 * It is ok not to disable preemption when adding the request to the plug list 1044 * or when attempting a merge. For details, please see schedule() where 1045 * blk_flush_plug() is called. 1046 */ 1047 struct blk_plug { 1048 struct rq_list mq_list; /* blk-mq requests */ 1049 1050 /* if ios_left is > 1, we can batch tag/rq allocations */ 1051 struct rq_list cached_rqs; 1052 u64 cur_ktime; 1053 unsigned short nr_ios; 1054 1055 unsigned short rq_count; 1056 1057 bool multiple_queues; 1058 bool has_elevator; 1059 1060 struct list_head cb_list; /* md requires an unplug callback */ 1061 }; 1062 1063 struct blk_plug_cb; 1064 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1065 struct blk_plug_cb { 1066 struct list_head list; 1067 blk_plug_cb_fn callback; 1068 void *data; 1069 }; 1070 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1071 void *data, int size); 1072 extern void blk_start_plug(struct blk_plug *); 1073 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short); 1074 extern void blk_finish_plug(struct blk_plug *); 1075 1076 void __blk_flush_plug(struct blk_plug *plug, bool from_schedule); 1077 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1078 { 1079 if (plug) 1080 __blk_flush_plug(plug, async); 1081 } 1082 1083 /* 1084 * tsk == current here 1085 */ 1086 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1087 { 1088 struct blk_plug *plug = tsk->plug; 1089 1090 if (plug) 1091 plug->cur_ktime = 0; 1092 current->flags &= ~PF_BLOCK_TS; 1093 } 1094 1095 int blkdev_issue_flush(struct block_device *bdev); 1096 long nr_blockdev_pages(void); 1097 #else /* CONFIG_BLOCK */ 1098 struct blk_plug { 1099 }; 1100 1101 static inline void blk_start_plug_nr_ios(struct blk_plug *plug, 1102 unsigned short nr_ios) 1103 { 1104 } 1105 1106 static inline void blk_start_plug(struct blk_plug *plug) 1107 { 1108 } 1109 1110 static inline void blk_finish_plug(struct blk_plug *plug) 1111 { 1112 } 1113 1114 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1115 { 1116 } 1117 1118 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1119 { 1120 } 1121 1122 static inline int blkdev_issue_flush(struct block_device *bdev) 1123 { 1124 return 0; 1125 } 1126 1127 static inline long nr_blockdev_pages(void) 1128 { 1129 return 0; 1130 } 1131 #endif /* CONFIG_BLOCK */ 1132 1133 extern void blk_io_schedule(void); 1134 1135 int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1136 sector_t nr_sects, gfp_t gfp_mask); 1137 int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1138 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop); 1139 int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, 1140 sector_t nr_sects, gfp_t gfp); 1141 1142 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1143 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1144 #define BLKDEV_ZERO_KILLABLE (1 << 2) /* interruptible by fatal signals */ 1145 1146 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1147 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1148 unsigned flags); 1149 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1150 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1151 1152 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1153 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1154 { 1155 return blkdev_issue_discard(sb->s_bdev, 1156 block << (sb->s_blocksize_bits - 1157 SECTOR_SHIFT), 1158 nr_blocks << (sb->s_blocksize_bits - 1159 SECTOR_SHIFT), 1160 gfp_mask); 1161 } 1162 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1163 sector_t nr_blocks, gfp_t gfp_mask) 1164 { 1165 return blkdev_issue_zeroout(sb->s_bdev, 1166 block << (sb->s_blocksize_bits - 1167 SECTOR_SHIFT), 1168 nr_blocks << (sb->s_blocksize_bits - 1169 SECTOR_SHIFT), 1170 gfp_mask, 0); 1171 } 1172 1173 static inline bool bdev_is_partition(struct block_device *bdev) 1174 { 1175 return bdev_partno(bdev) != 0; 1176 } 1177 1178 enum blk_default_limits { 1179 BLK_MAX_SEGMENTS = 128, 1180 BLK_SAFE_MAX_SECTORS = 255, 1181 BLK_MAX_SEGMENT_SIZE = 65536, 1182 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1183 }; 1184 1185 /* 1186 * Default upper limit for the software max_sectors limit used for 1187 * regular file system I/O. This can be increased through sysfs. 1188 * 1189 * Not to be confused with the max_hw_sector limit that is entirely 1190 * controlled by the driver, usually based on hardware limits. 1191 */ 1192 #define BLK_DEF_MAX_SECTORS_CAP 2560u 1193 1194 static inline struct queue_limits *bdev_limits(struct block_device *bdev) 1195 { 1196 return &bdev_get_queue(bdev)->limits; 1197 } 1198 1199 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1200 { 1201 return q->limits.seg_boundary_mask; 1202 } 1203 1204 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1205 { 1206 return q->limits.virt_boundary_mask; 1207 } 1208 1209 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1210 { 1211 return q->limits.max_sectors; 1212 } 1213 1214 static inline unsigned int queue_max_bytes(struct request_queue *q) 1215 { 1216 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9; 1217 } 1218 1219 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1220 { 1221 return q->limits.max_hw_sectors; 1222 } 1223 1224 static inline unsigned short queue_max_segments(const struct request_queue *q) 1225 { 1226 return q->limits.max_segments; 1227 } 1228 1229 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1230 { 1231 return q->limits.max_discard_segments; 1232 } 1233 1234 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1235 { 1236 return q->limits.max_segment_size; 1237 } 1238 1239 static inline bool queue_emulates_zone_append(struct request_queue *q) 1240 { 1241 return blk_queue_is_zoned(q) && !q->limits.max_hw_zone_append_sectors; 1242 } 1243 1244 static inline bool bdev_emulates_zone_append(struct block_device *bdev) 1245 { 1246 return queue_emulates_zone_append(bdev_get_queue(bdev)); 1247 } 1248 1249 static inline unsigned int 1250 bdev_max_zone_append_sectors(struct block_device *bdev) 1251 { 1252 return bdev_limits(bdev)->max_zone_append_sectors; 1253 } 1254 1255 static inline unsigned int bdev_max_segments(struct block_device *bdev) 1256 { 1257 return queue_max_segments(bdev_get_queue(bdev)); 1258 } 1259 1260 static inline unsigned queue_logical_block_size(const struct request_queue *q) 1261 { 1262 return q->limits.logical_block_size; 1263 } 1264 1265 static inline unsigned int bdev_logical_block_size(struct block_device *bdev) 1266 { 1267 return queue_logical_block_size(bdev_get_queue(bdev)); 1268 } 1269 1270 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1271 { 1272 return q->limits.physical_block_size; 1273 } 1274 1275 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1276 { 1277 return queue_physical_block_size(bdev_get_queue(bdev)); 1278 } 1279 1280 static inline unsigned int queue_io_min(const struct request_queue *q) 1281 { 1282 return q->limits.io_min; 1283 } 1284 1285 static inline unsigned int bdev_io_min(struct block_device *bdev) 1286 { 1287 return queue_io_min(bdev_get_queue(bdev)); 1288 } 1289 1290 static inline unsigned int queue_io_opt(const struct request_queue *q) 1291 { 1292 return q->limits.io_opt; 1293 } 1294 1295 static inline unsigned int bdev_io_opt(struct block_device *bdev) 1296 { 1297 return queue_io_opt(bdev_get_queue(bdev)); 1298 } 1299 1300 static inline unsigned int 1301 queue_zone_write_granularity(const struct request_queue *q) 1302 { 1303 return q->limits.zone_write_granularity; 1304 } 1305 1306 static inline unsigned int 1307 bdev_zone_write_granularity(struct block_device *bdev) 1308 { 1309 return queue_zone_write_granularity(bdev_get_queue(bdev)); 1310 } 1311 1312 int bdev_alignment_offset(struct block_device *bdev); 1313 unsigned int bdev_discard_alignment(struct block_device *bdev); 1314 1315 static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev) 1316 { 1317 return bdev_limits(bdev)->max_discard_sectors; 1318 } 1319 1320 static inline unsigned int bdev_discard_granularity(struct block_device *bdev) 1321 { 1322 return bdev_limits(bdev)->discard_granularity; 1323 } 1324 1325 static inline unsigned int 1326 bdev_max_secure_erase_sectors(struct block_device *bdev) 1327 { 1328 return bdev_limits(bdev)->max_secure_erase_sectors; 1329 } 1330 1331 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1332 { 1333 return bdev_limits(bdev)->max_write_zeroes_sectors; 1334 } 1335 1336 static inline bool bdev_nonrot(struct block_device *bdev) 1337 { 1338 return blk_queue_nonrot(bdev_get_queue(bdev)); 1339 } 1340 1341 static inline bool bdev_synchronous(struct block_device *bdev) 1342 { 1343 return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS; 1344 } 1345 1346 static inline bool bdev_stable_writes(struct block_device *bdev) 1347 { 1348 struct request_queue *q = bdev_get_queue(bdev); 1349 1350 if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && 1351 q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) 1352 return true; 1353 return q->limits.features & BLK_FEAT_STABLE_WRITES; 1354 } 1355 1356 static inline bool blk_queue_write_cache(struct request_queue *q) 1357 { 1358 return (q->limits.features & BLK_FEAT_WRITE_CACHE) && 1359 !(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED); 1360 } 1361 1362 static inline bool bdev_write_cache(struct block_device *bdev) 1363 { 1364 return blk_queue_write_cache(bdev_get_queue(bdev)); 1365 } 1366 1367 static inline bool bdev_fua(struct block_device *bdev) 1368 { 1369 return bdev_limits(bdev)->features & BLK_FEAT_FUA; 1370 } 1371 1372 static inline bool bdev_nowait(struct block_device *bdev) 1373 { 1374 return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT; 1375 } 1376 1377 static inline bool bdev_is_zoned(struct block_device *bdev) 1378 { 1379 return blk_queue_is_zoned(bdev_get_queue(bdev)); 1380 } 1381 1382 static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec) 1383 { 1384 return disk_zone_no(bdev->bd_disk, sec); 1385 } 1386 1387 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1388 { 1389 struct request_queue *q = bdev_get_queue(bdev); 1390 1391 if (!blk_queue_is_zoned(q)) 1392 return 0; 1393 return q->limits.chunk_sectors; 1394 } 1395 1396 static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev, 1397 sector_t sector) 1398 { 1399 return sector & (bdev_zone_sectors(bdev) - 1); 1400 } 1401 1402 static inline sector_t bio_offset_from_zone_start(struct bio *bio) 1403 { 1404 return bdev_offset_from_zone_start(bio->bi_bdev, 1405 bio->bi_iter.bi_sector); 1406 } 1407 1408 static inline bool bdev_is_zone_start(struct block_device *bdev, 1409 sector_t sector) 1410 { 1411 return bdev_offset_from_zone_start(bdev, sector) == 0; 1412 } 1413 1414 /** 1415 * bdev_zone_is_seq - check if a sector belongs to a sequential write zone 1416 * @bdev: block device to check 1417 * @sector: sector number 1418 * 1419 * Check if @sector on @bdev is contained in a sequential write required zone. 1420 */ 1421 static inline bool bdev_zone_is_seq(struct block_device *bdev, sector_t sector) 1422 { 1423 bool is_seq = false; 1424 1425 #if IS_ENABLED(CONFIG_BLK_DEV_ZONED) 1426 if (bdev_is_zoned(bdev)) { 1427 struct gendisk *disk = bdev->bd_disk; 1428 unsigned long *bitmap; 1429 1430 rcu_read_lock(); 1431 bitmap = rcu_dereference(disk->conv_zones_bitmap); 1432 is_seq = !bitmap || 1433 !test_bit(disk_zone_no(disk, sector), bitmap); 1434 rcu_read_unlock(); 1435 } 1436 #endif 1437 1438 return is_seq; 1439 } 1440 1441 int blk_zone_issue_zeroout(struct block_device *bdev, sector_t sector, 1442 sector_t nr_sects, gfp_t gfp_mask); 1443 1444 static inline unsigned int queue_dma_alignment(const struct request_queue *q) 1445 { 1446 return q->limits.dma_alignment; 1447 } 1448 1449 static inline unsigned int 1450 queue_atomic_write_unit_max_bytes(const struct request_queue *q) 1451 { 1452 return q->limits.atomic_write_unit_max; 1453 } 1454 1455 static inline unsigned int 1456 queue_atomic_write_unit_min_bytes(const struct request_queue *q) 1457 { 1458 return q->limits.atomic_write_unit_min; 1459 } 1460 1461 static inline unsigned int 1462 queue_atomic_write_boundary_bytes(const struct request_queue *q) 1463 { 1464 return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT; 1465 } 1466 1467 static inline unsigned int 1468 queue_atomic_write_max_bytes(const struct request_queue *q) 1469 { 1470 return q->limits.atomic_write_max_sectors << SECTOR_SHIFT; 1471 } 1472 1473 static inline unsigned int bdev_dma_alignment(struct block_device *bdev) 1474 { 1475 return queue_dma_alignment(bdev_get_queue(bdev)); 1476 } 1477 1478 static inline bool bdev_iter_is_aligned(struct block_device *bdev, 1479 struct iov_iter *iter) 1480 { 1481 return iov_iter_is_aligned(iter, bdev_dma_alignment(bdev), 1482 bdev_logical_block_size(bdev) - 1); 1483 } 1484 1485 static inline unsigned int 1486 blk_lim_dma_alignment_and_pad(struct queue_limits *lim) 1487 { 1488 return lim->dma_alignment | lim->dma_pad_mask; 1489 } 1490 1491 static inline bool blk_rq_aligned(struct request_queue *q, unsigned long addr, 1492 unsigned int len) 1493 { 1494 unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits); 1495 1496 return !(addr & alignment) && !(len & alignment); 1497 } 1498 1499 /* assumes size > 256 */ 1500 static inline unsigned int blksize_bits(unsigned int size) 1501 { 1502 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT; 1503 } 1504 1505 int kblockd_schedule_work(struct work_struct *work); 1506 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1507 1508 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1509 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1510 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1511 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1512 1513 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 1514 1515 bool blk_crypto_register(struct blk_crypto_profile *profile, 1516 struct request_queue *q); 1517 1518 #else /* CONFIG_BLK_INLINE_ENCRYPTION */ 1519 1520 static inline bool blk_crypto_register(struct blk_crypto_profile *profile, 1521 struct request_queue *q) 1522 { 1523 return true; 1524 } 1525 1526 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */ 1527 1528 enum blk_unique_id { 1529 /* these match the Designator Types specified in SPC */ 1530 BLK_UID_T10 = 1, 1531 BLK_UID_EUI64 = 2, 1532 BLK_UID_NAA = 3, 1533 }; 1534 1535 struct block_device_operations { 1536 void (*submit_bio)(struct bio *bio); 1537 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob, 1538 unsigned int flags); 1539 int (*open)(struct gendisk *disk, blk_mode_t mode); 1540 void (*release)(struct gendisk *disk); 1541 int (*ioctl)(struct block_device *bdev, blk_mode_t mode, 1542 unsigned cmd, unsigned long arg); 1543 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode, 1544 unsigned cmd, unsigned long arg); 1545 unsigned int (*check_events) (struct gendisk *disk, 1546 unsigned int clearing); 1547 void (*unlock_native_capacity) (struct gendisk *); 1548 int (*getgeo)(struct block_device *, struct hd_geometry *); 1549 int (*set_read_only)(struct block_device *bdev, bool ro); 1550 void (*free_disk)(struct gendisk *disk); 1551 /* this callback is with swap_lock and sometimes page table lock held */ 1552 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1553 int (*report_zones)(struct gendisk *, sector_t sector, 1554 unsigned int nr_zones, report_zones_cb cb, void *data); 1555 char *(*devnode)(struct gendisk *disk, umode_t *mode); 1556 /* returns the length of the identifier or a negative errno: */ 1557 int (*get_unique_id)(struct gendisk *disk, u8 id[16], 1558 enum blk_unique_id id_type); 1559 struct module *owner; 1560 const struct pr_ops *pr_ops; 1561 1562 /* 1563 * Special callback for probing GPT entry at a given sector. 1564 * Needed by Android devices, used by GPT scanner and MMC blk 1565 * driver. 1566 */ 1567 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector); 1568 }; 1569 1570 #ifdef CONFIG_COMPAT 1571 extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t, 1572 unsigned int, unsigned long); 1573 #else 1574 #define blkdev_compat_ptr_ioctl NULL 1575 #endif 1576 1577 static inline void blk_wake_io_task(struct task_struct *waiter) 1578 { 1579 /* 1580 * If we're polling, the task itself is doing the completions. For 1581 * that case, we don't need to signal a wakeup, it's enough to just 1582 * mark us as RUNNING. 1583 */ 1584 if (waiter == current) 1585 __set_current_state(TASK_RUNNING); 1586 else 1587 wake_up_process(waiter); 1588 } 1589 1590 unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, 1591 unsigned long start_time); 1592 void bdev_end_io_acct(struct block_device *bdev, enum req_op op, 1593 unsigned int sectors, unsigned long start_time); 1594 1595 unsigned long bio_start_io_acct(struct bio *bio); 1596 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, 1597 struct block_device *orig_bdev); 1598 1599 /** 1600 * bio_end_io_acct - end I/O accounting for bio based drivers 1601 * @bio: bio to end account for 1602 * @start_time: start time returned by bio_start_io_acct() 1603 */ 1604 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time) 1605 { 1606 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev); 1607 } 1608 1609 int set_blocksize(struct file *file, int size); 1610 1611 int lookup_bdev(const char *pathname, dev_t *dev); 1612 1613 void blkdev_show(struct seq_file *seqf, off_t offset); 1614 1615 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */ 1616 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */ 1617 #ifdef CONFIG_BLOCK 1618 #define BLKDEV_MAJOR_MAX 512 1619 #else 1620 #define BLKDEV_MAJOR_MAX 0 1621 #endif 1622 1623 struct blk_holder_ops { 1624 void (*mark_dead)(struct block_device *bdev, bool surprise); 1625 1626 /* 1627 * Sync the file system mounted on the block device. 1628 */ 1629 void (*sync)(struct block_device *bdev); 1630 1631 /* 1632 * Freeze the file system mounted on the block device. 1633 */ 1634 int (*freeze)(struct block_device *bdev); 1635 1636 /* 1637 * Thaw the file system mounted on the block device. 1638 */ 1639 int (*thaw)(struct block_device *bdev); 1640 }; 1641 1642 /* 1643 * For filesystems using @fs_holder_ops, the @holder argument passed to 1644 * helpers used to open and claim block devices via 1645 * bd_prepare_to_claim() must point to a superblock. 1646 */ 1647 extern const struct blk_holder_ops fs_holder_ops; 1648 1649 /* 1650 * Return the correct open flags for blkdev_get_by_* for super block flags 1651 * as stored in sb->s_flags. 1652 */ 1653 #define sb_open_mode(flags) \ 1654 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \ 1655 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE)) 1656 1657 struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder, 1658 const struct blk_holder_ops *hops); 1659 struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode, 1660 void *holder, const struct blk_holder_ops *hops); 1661 int bd_prepare_to_claim(struct block_device *bdev, void *holder, 1662 const struct blk_holder_ops *hops); 1663 void bd_abort_claiming(struct block_device *bdev, void *holder); 1664 1665 /* just for blk-cgroup, don't use elsewhere */ 1666 struct block_device *blkdev_get_no_open(dev_t dev); 1667 void blkdev_put_no_open(struct block_device *bdev); 1668 1669 struct block_device *I_BDEV(struct inode *inode); 1670 struct block_device *file_bdev(struct file *bdev_file); 1671 bool disk_live(struct gendisk *disk); 1672 unsigned int block_size(struct block_device *bdev); 1673 1674 #ifdef CONFIG_BLOCK 1675 void invalidate_bdev(struct block_device *bdev); 1676 int sync_blockdev(struct block_device *bdev); 1677 int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend); 1678 int sync_blockdev_nowait(struct block_device *bdev); 1679 void sync_bdevs(bool wait); 1680 void bdev_statx(struct path *, struct kstat *, u32); 1681 void printk_all_partitions(void); 1682 int __init early_lookup_bdev(const char *pathname, dev_t *dev); 1683 #else 1684 static inline void invalidate_bdev(struct block_device *bdev) 1685 { 1686 } 1687 static inline int sync_blockdev(struct block_device *bdev) 1688 { 1689 return 0; 1690 } 1691 static inline int sync_blockdev_nowait(struct block_device *bdev) 1692 { 1693 return 0; 1694 } 1695 static inline void sync_bdevs(bool wait) 1696 { 1697 } 1698 static inline void bdev_statx(struct path *path, struct kstat *stat, 1699 u32 request_mask) 1700 { 1701 } 1702 static inline void printk_all_partitions(void) 1703 { 1704 } 1705 static inline int early_lookup_bdev(const char *pathname, dev_t *dev) 1706 { 1707 return -EINVAL; 1708 } 1709 #endif /* CONFIG_BLOCK */ 1710 1711 int bdev_freeze(struct block_device *bdev); 1712 int bdev_thaw(struct block_device *bdev); 1713 void bdev_fput(struct file *bdev_file); 1714 1715 struct io_comp_batch { 1716 struct rq_list req_list; 1717 bool need_ts; 1718 void (*complete)(struct io_comp_batch *); 1719 }; 1720 1721 static inline bool blk_atomic_write_start_sect_aligned(sector_t sector, 1722 struct queue_limits *limits) 1723 { 1724 unsigned int alignment = max(limits->atomic_write_hw_unit_min, 1725 limits->atomic_write_hw_boundary); 1726 1727 return IS_ALIGNED(sector, alignment >> SECTOR_SHIFT); 1728 } 1729 1730 static inline bool bdev_can_atomic_write(struct block_device *bdev) 1731 { 1732 struct request_queue *bd_queue = bdev->bd_queue; 1733 struct queue_limits *limits = &bd_queue->limits; 1734 1735 if (!limits->atomic_write_unit_min) 1736 return false; 1737 1738 if (bdev_is_partition(bdev)) 1739 return blk_atomic_write_start_sect_aligned(bdev->bd_start_sect, 1740 limits); 1741 1742 return true; 1743 } 1744 1745 static inline unsigned int 1746 bdev_atomic_write_unit_min_bytes(struct block_device *bdev) 1747 { 1748 if (!bdev_can_atomic_write(bdev)) 1749 return 0; 1750 return queue_atomic_write_unit_min_bytes(bdev_get_queue(bdev)); 1751 } 1752 1753 static inline unsigned int 1754 bdev_atomic_write_unit_max_bytes(struct block_device *bdev) 1755 { 1756 if (!bdev_can_atomic_write(bdev)) 1757 return 0; 1758 return queue_atomic_write_unit_max_bytes(bdev_get_queue(bdev)); 1759 } 1760 1761 #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { } 1762 1763 #endif /* _LINUX_BLKDEV_H */ 1764