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 /* 271 * We should strive for 1 << (PAGE_SHIFT + MAX_PAGECACHE_ORDER) 272 * however we constrain this to what we can validate and test. 273 */ 274 #define BLK_MAX_BLOCK_SIZE SZ_64K 275 276 /* blk_validate_limits() validates bsize, so drivers don't usually need to */ 277 static inline int blk_validate_block_size(unsigned long bsize) 278 { 279 if (bsize < 512 || bsize > BLK_MAX_BLOCK_SIZE || !is_power_of_2(bsize)) 280 return -EINVAL; 281 282 return 0; 283 } 284 285 static inline bool blk_op_is_passthrough(blk_opf_t op) 286 { 287 op &= REQ_OP_MASK; 288 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 289 } 290 291 /* flags set by the driver in queue_limits.features */ 292 typedef unsigned int __bitwise blk_features_t; 293 294 /* supports a volatile write cache */ 295 #define BLK_FEAT_WRITE_CACHE ((__force blk_features_t)(1u << 0)) 296 297 /* supports passing on the FUA bit */ 298 #define BLK_FEAT_FUA ((__force blk_features_t)(1u << 1)) 299 300 /* rotational device (hard drive or floppy) */ 301 #define BLK_FEAT_ROTATIONAL ((__force blk_features_t)(1u << 2)) 302 303 /* contributes to the random number pool */ 304 #define BLK_FEAT_ADD_RANDOM ((__force blk_features_t)(1u << 3)) 305 306 /* do disk/partitions IO accounting */ 307 #define BLK_FEAT_IO_STAT ((__force blk_features_t)(1u << 4)) 308 309 /* don't modify data until writeback is done */ 310 #define BLK_FEAT_STABLE_WRITES ((__force blk_features_t)(1u << 5)) 311 312 /* always completes in submit context */ 313 #define BLK_FEAT_SYNCHRONOUS ((__force blk_features_t)(1u << 6)) 314 315 /* supports REQ_NOWAIT */ 316 #define BLK_FEAT_NOWAIT ((__force blk_features_t)(1u << 7)) 317 318 /* supports DAX */ 319 #define BLK_FEAT_DAX ((__force blk_features_t)(1u << 8)) 320 321 /* supports I/O polling */ 322 #define BLK_FEAT_POLL ((__force blk_features_t)(1u << 9)) 323 324 /* is a zoned device */ 325 #define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10)) 326 327 /* supports PCI(e) p2p requests */ 328 #define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12)) 329 330 /* skip this queue in blk_mq_(un)quiesce_tagset */ 331 #define BLK_FEAT_SKIP_TAGSET_QUIESCE ((__force blk_features_t)(1u << 13)) 332 333 /* bounce all highmem pages */ 334 #define BLK_FEAT_BOUNCE_HIGH ((__force blk_features_t)(1u << 14)) 335 336 /* undocumented magic for bcache */ 337 #define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \ 338 ((__force blk_features_t)(1u << 15)) 339 340 /* atomic writes enabled */ 341 #define BLK_FEAT_ATOMIC_WRITES \ 342 ((__force blk_features_t)(1u << 16)) 343 344 /* 345 * Flags automatically inherited when stacking limits. 346 */ 347 #define BLK_FEAT_INHERIT_MASK \ 348 (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ 349 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH | \ 350 BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE) 351 352 /* internal flags in queue_limits.flags */ 353 typedef unsigned int __bitwise blk_flags_t; 354 355 /* do not send FLUSH/FUA commands despite advertising a write cache */ 356 #define BLK_FLAG_WRITE_CACHE_DISABLED ((__force blk_flags_t)(1u << 0)) 357 358 /* I/O topology is misaligned */ 359 #define BLK_FLAG_MISALIGNED ((__force blk_flags_t)(1u << 1)) 360 361 /* passthrough command IO accounting */ 362 #define BLK_FLAG_IOSTATS_PASSTHROUGH ((__force blk_flags_t)(1u << 2)) 363 364 struct queue_limits { 365 blk_features_t features; 366 blk_flags_t flags; 367 unsigned long seg_boundary_mask; 368 unsigned long virt_boundary_mask; 369 370 unsigned int max_hw_sectors; 371 unsigned int max_dev_sectors; 372 unsigned int chunk_sectors; 373 unsigned int max_sectors; 374 unsigned int max_user_sectors; 375 unsigned int max_segment_size; 376 unsigned int physical_block_size; 377 unsigned int logical_block_size; 378 unsigned int alignment_offset; 379 unsigned int io_min; 380 unsigned int io_opt; 381 unsigned int max_discard_sectors; 382 unsigned int max_hw_discard_sectors; 383 unsigned int max_user_discard_sectors; 384 unsigned int max_secure_erase_sectors; 385 unsigned int max_write_zeroes_sectors; 386 unsigned int max_hw_zone_append_sectors; 387 unsigned int max_zone_append_sectors; 388 unsigned int discard_granularity; 389 unsigned int discard_alignment; 390 unsigned int zone_write_granularity; 391 392 /* atomic write limits */ 393 unsigned int atomic_write_hw_max; 394 unsigned int atomic_write_max_sectors; 395 unsigned int atomic_write_hw_boundary; 396 unsigned int atomic_write_boundary_sectors; 397 unsigned int atomic_write_hw_unit_min; 398 unsigned int atomic_write_unit_min; 399 unsigned int atomic_write_hw_unit_max; 400 unsigned int atomic_write_unit_max; 401 402 unsigned short max_segments; 403 unsigned short max_integrity_segments; 404 unsigned short max_discard_segments; 405 406 unsigned int max_open_zones; 407 unsigned int max_active_zones; 408 409 /* 410 * Drivers that set dma_alignment to less than 511 must be prepared to 411 * handle individual bvec's that are not a multiple of a SECTOR_SIZE 412 * due to possible offsets. 413 */ 414 unsigned int dma_alignment; 415 unsigned int dma_pad_mask; 416 417 struct blk_integrity integrity; 418 }; 419 420 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, 421 void *data); 422 423 #define BLK_ALL_ZONES ((unsigned int)-1) 424 int blkdev_report_zones(struct block_device *bdev, sector_t sector, 425 unsigned int nr_zones, report_zones_cb cb, void *data); 426 int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, 427 sector_t sectors, sector_t nr_sectors); 428 int blk_revalidate_disk_zones(struct gendisk *disk); 429 430 /* 431 * Independent access ranges: struct blk_independent_access_range describes 432 * a range of contiguous sectors that can be accessed using device command 433 * execution resources that are independent from the resources used for 434 * other access ranges. This is typically found with single-LUN multi-actuator 435 * HDDs where each access range is served by a different set of heads. 436 * The set of independent ranges supported by the device is defined using 437 * struct blk_independent_access_ranges. The independent ranges must not overlap 438 * and must include all sectors within the disk capacity (no sector holes 439 * allowed). 440 * For a device with multiple ranges, requests targeting sectors in different 441 * ranges can be executed in parallel. A request can straddle an access range 442 * boundary. 443 */ 444 struct blk_independent_access_range { 445 struct kobject kobj; 446 sector_t sector; 447 sector_t nr_sectors; 448 }; 449 450 struct blk_independent_access_ranges { 451 struct kobject kobj; 452 bool sysfs_registered; 453 unsigned int nr_ia_ranges; 454 struct blk_independent_access_range ia_range[]; 455 }; 456 457 struct request_queue { 458 /* 459 * The queue owner gets to use this for whatever they like. 460 * ll_rw_blk doesn't touch it. 461 */ 462 void *queuedata; 463 464 struct elevator_queue *elevator; 465 466 const struct blk_mq_ops *mq_ops; 467 468 /* sw queues */ 469 struct blk_mq_ctx __percpu *queue_ctx; 470 471 /* 472 * various queue flags, see QUEUE_* below 473 */ 474 unsigned long queue_flags; 475 476 unsigned int rq_timeout; 477 478 unsigned int queue_depth; 479 480 refcount_t refs; 481 482 /* hw dispatch queues */ 483 unsigned int nr_hw_queues; 484 struct xarray hctx_table; 485 486 struct percpu_ref q_usage_counter; 487 struct lock_class_key io_lock_cls_key; 488 struct lockdep_map io_lockdep_map; 489 490 struct lock_class_key q_lock_cls_key; 491 struct lockdep_map q_lockdep_map; 492 493 struct request *last_merge; 494 495 spinlock_t queue_lock; 496 497 int quiesce_depth; 498 499 struct gendisk *disk; 500 501 /* 502 * mq queue kobject 503 */ 504 struct kobject *mq_kobj; 505 506 struct queue_limits limits; 507 508 #ifdef CONFIG_PM 509 struct device *dev; 510 enum rpm_status rpm_status; 511 #endif 512 513 /* 514 * Number of contexts that have called blk_set_pm_only(). If this 515 * counter is above zero then only RQF_PM requests are processed. 516 */ 517 atomic_t pm_only; 518 519 struct blk_queue_stats *stats; 520 struct rq_qos *rq_qos; 521 struct mutex rq_qos_mutex; 522 523 /* 524 * ida allocated id for this queue. Used to index queues from 525 * ioctx. 526 */ 527 int id; 528 529 /* 530 * queue settings 531 */ 532 unsigned long nr_requests; /* Max # of requests */ 533 534 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 535 struct blk_crypto_profile *crypto_profile; 536 struct kobject *crypto_kobject; 537 #endif 538 539 struct timer_list timeout; 540 struct work_struct timeout_work; 541 542 atomic_t nr_active_requests_shared_tags; 543 544 struct blk_mq_tags *sched_shared_tags; 545 546 struct list_head icq_list; 547 #ifdef CONFIG_BLK_CGROUP 548 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 549 struct blkcg_gq *root_blkg; 550 struct list_head blkg_list; 551 struct mutex blkcg_mutex; 552 #endif 553 554 int node; 555 556 spinlock_t requeue_lock; 557 struct list_head requeue_list; 558 struct delayed_work requeue_work; 559 560 #ifdef CONFIG_BLK_DEV_IO_TRACE 561 struct blk_trace __rcu *blk_trace; 562 #endif 563 /* 564 * for flush operations 565 */ 566 struct blk_flush_queue *fq; 567 struct list_head flush_list; 568 569 struct mutex sysfs_lock; 570 struct mutex limits_lock; 571 572 /* 573 * for reusing dead hctx instance in case of updating 574 * nr_hw_queues 575 */ 576 struct list_head unused_hctx_list; 577 spinlock_t unused_hctx_lock; 578 579 int mq_freeze_depth; 580 581 #ifdef CONFIG_BLK_DEV_THROTTLING 582 /* Throttle data */ 583 struct throtl_data *td; 584 #endif 585 struct rcu_head rcu_head; 586 #ifdef CONFIG_LOCKDEP 587 struct task_struct *mq_freeze_owner; 588 int mq_freeze_owner_depth; 589 /* 590 * Records disk & queue state in current context, used in unfreeze 591 * queue 592 */ 593 bool mq_freeze_disk_dead; 594 bool mq_freeze_queue_dying; 595 #endif 596 wait_queue_head_t mq_freeze_wq; 597 /* 598 * Protect concurrent access to q_usage_counter by 599 * percpu_ref_kill() and percpu_ref_reinit(). 600 */ 601 struct mutex mq_freeze_lock; 602 603 struct blk_mq_tag_set *tag_set; 604 struct list_head tag_set_list; 605 606 struct dentry *debugfs_dir; 607 struct dentry *sched_debugfs_dir; 608 struct dentry *rqos_debugfs_dir; 609 /* 610 * Serializes all debugfs metadata operations using the above dentries. 611 */ 612 struct mutex debugfs_mutex; 613 }; 614 615 /* Keep blk_queue_flag_name[] in sync with the definitions below */ 616 enum { 617 QUEUE_FLAG_DYING, /* queue being torn down */ 618 QUEUE_FLAG_NOMERGES, /* disable merge attempts */ 619 QUEUE_FLAG_SAME_COMP, /* complete on same CPU-group */ 620 QUEUE_FLAG_FAIL_IO, /* fake timeout */ 621 QUEUE_FLAG_NOXMERGES, /* No extended merges */ 622 QUEUE_FLAG_SAME_FORCE, /* force complete on same CPU */ 623 QUEUE_FLAG_INIT_DONE, /* queue is initialized */ 624 QUEUE_FLAG_STATS, /* track IO start and completion times */ 625 QUEUE_FLAG_REGISTERED, /* queue has been registered to a disk */ 626 QUEUE_FLAG_QUIESCED, /* queue has been quiesced */ 627 QUEUE_FLAG_RQ_ALLOC_TIME, /* record rq->alloc_time_ns */ 628 QUEUE_FLAG_HCTX_ACTIVE, /* at least one blk-mq hctx is active */ 629 QUEUE_FLAG_SQ_SCHED, /* single queue style io dispatch */ 630 QUEUE_FLAG_MAX 631 }; 632 633 #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) 634 635 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 636 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 637 638 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 639 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 640 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 641 #define blk_queue_noxmerges(q) \ 642 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 643 #define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL)) 644 #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) 645 #define blk_queue_passthrough_stat(q) \ 646 ((q)->limits.flags & BLK_FLAG_IOSTATS_PASSTHROUGH) 647 #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) 648 #define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA) 649 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 650 #define blk_queue_rq_alloc_time(q) \ 651 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) 652 #else 653 #define blk_queue_rq_alloc_time(q) false 654 #endif 655 656 #define blk_noretry_request(rq) \ 657 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 658 REQ_FAILFAST_DRIVER)) 659 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 660 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 661 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) 662 #define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags) 663 #define blk_queue_skip_tagset_quiesce(q) \ 664 ((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE) 665 666 extern void blk_set_pm_only(struct request_queue *q); 667 extern void blk_clear_pm_only(struct request_queue *q); 668 669 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 670 671 #define dma_map_bvec(dev, bv, dir, attrs) \ 672 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 673 (dir), (attrs)) 674 675 static inline bool queue_is_mq(struct request_queue *q) 676 { 677 return q->mq_ops; 678 } 679 680 #ifdef CONFIG_PM 681 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 682 { 683 return q->rpm_status; 684 } 685 #else 686 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 687 { 688 return RPM_ACTIVE; 689 } 690 #endif 691 692 static inline bool blk_queue_is_zoned(struct request_queue *q) 693 { 694 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && 695 (q->limits.features & BLK_FEAT_ZONED); 696 } 697 698 #ifdef CONFIG_BLK_DEV_ZONED 699 static inline unsigned int disk_nr_zones(struct gendisk *disk) 700 { 701 return disk->nr_zones; 702 } 703 bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); 704 #else /* CONFIG_BLK_DEV_ZONED */ 705 static inline unsigned int disk_nr_zones(struct gendisk *disk) 706 { 707 return 0; 708 } 709 static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) 710 { 711 return false; 712 } 713 #endif /* CONFIG_BLK_DEV_ZONED */ 714 715 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) 716 { 717 if (!blk_queue_is_zoned(disk->queue)) 718 return 0; 719 return sector >> ilog2(disk->queue->limits.chunk_sectors); 720 } 721 722 static inline unsigned int bdev_nr_zones(struct block_device *bdev) 723 { 724 return disk_nr_zones(bdev->bd_disk); 725 } 726 727 static inline unsigned int bdev_max_open_zones(struct block_device *bdev) 728 { 729 return bdev->bd_disk->queue->limits.max_open_zones; 730 } 731 732 static inline unsigned int bdev_max_active_zones(struct block_device *bdev) 733 { 734 return bdev->bd_disk->queue->limits.max_active_zones; 735 } 736 737 static inline unsigned int blk_queue_depth(struct request_queue *q) 738 { 739 if (q->queue_depth) 740 return q->queue_depth; 741 742 return q->nr_requests; 743 } 744 745 /* 746 * default timeout for SG_IO if none specified 747 */ 748 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 749 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 750 751 /* This should not be used directly - use rq_for_each_segment */ 752 #define for_each_bio(_bio) \ 753 for (; _bio; _bio = _bio->bi_next) 754 755 int __must_check add_disk_fwnode(struct device *parent, struct gendisk *disk, 756 const struct attribute_group **groups, 757 struct fwnode_handle *fwnode); 758 int __must_check device_add_disk(struct device *parent, struct gendisk *disk, 759 const struct attribute_group **groups); 760 static inline int __must_check add_disk(struct gendisk *disk) 761 { 762 return device_add_disk(NULL, disk, NULL); 763 } 764 void del_gendisk(struct gendisk *gp); 765 void invalidate_disk(struct gendisk *disk); 766 void set_disk_ro(struct gendisk *disk, bool read_only); 767 void disk_uevent(struct gendisk *disk, enum kobject_action action); 768 769 static inline u8 bdev_partno(const struct block_device *bdev) 770 { 771 return atomic_read(&bdev->__bd_flags) & BD_PARTNO; 772 } 773 774 static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag) 775 { 776 return atomic_read(&bdev->__bd_flags) & flag; 777 } 778 779 static inline void bdev_set_flag(struct block_device *bdev, unsigned flag) 780 { 781 atomic_or(flag, &bdev->__bd_flags); 782 } 783 784 static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag) 785 { 786 atomic_andnot(flag, &bdev->__bd_flags); 787 } 788 789 static inline bool get_disk_ro(struct gendisk *disk) 790 { 791 return bdev_test_flag(disk->part0, BD_READ_ONLY) || 792 test_bit(GD_READ_ONLY, &disk->state); 793 } 794 795 static inline bool bdev_read_only(struct block_device *bdev) 796 { 797 return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk); 798 } 799 800 bool set_capacity_and_notify(struct gendisk *disk, sector_t size); 801 void disk_force_media_change(struct gendisk *disk); 802 void bdev_mark_dead(struct block_device *bdev, bool surprise); 803 804 void add_disk_randomness(struct gendisk *disk) __latent_entropy; 805 void rand_initialize_disk(struct gendisk *disk); 806 807 static inline sector_t get_start_sect(struct block_device *bdev) 808 { 809 return bdev->bd_start_sect; 810 } 811 812 static inline sector_t bdev_nr_sectors(struct block_device *bdev) 813 { 814 return bdev->bd_nr_sectors; 815 } 816 817 static inline loff_t bdev_nr_bytes(struct block_device *bdev) 818 { 819 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT; 820 } 821 822 static inline sector_t get_capacity(struct gendisk *disk) 823 { 824 return bdev_nr_sectors(disk->part0); 825 } 826 827 static inline u64 sb_bdev_nr_blocks(struct super_block *sb) 828 { 829 return bdev_nr_sectors(sb->s_bdev) >> 830 (sb->s_blocksize_bits - SECTOR_SHIFT); 831 } 832 833 int bdev_disk_changed(struct gendisk *disk, bool invalidate); 834 835 void put_disk(struct gendisk *disk); 836 struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node, 837 struct lock_class_key *lkclass); 838 839 /** 840 * blk_alloc_disk - allocate a gendisk structure 841 * @lim: queue limits to be used for this disk. 842 * @node_id: numa node to allocate on 843 * 844 * Allocate and pre-initialize a gendisk structure for use with BIO based 845 * drivers. 846 * 847 * Returns an ERR_PTR on error, else the allocated disk. 848 * 849 * Context: can sleep 850 */ 851 #define blk_alloc_disk(lim, node_id) \ 852 ({ \ 853 static struct lock_class_key __key; \ 854 \ 855 __blk_alloc_disk(lim, node_id, &__key); \ 856 }) 857 858 int __register_blkdev(unsigned int major, const char *name, 859 void (*probe)(dev_t devt)); 860 #define register_blkdev(major, name) \ 861 __register_blkdev(major, name, NULL) 862 void unregister_blkdev(unsigned int major, const char *name); 863 864 bool disk_check_media_change(struct gendisk *disk); 865 void set_capacity(struct gendisk *disk, sector_t size); 866 867 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 868 int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk); 869 void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk); 870 #else 871 static inline int bd_link_disk_holder(struct block_device *bdev, 872 struct gendisk *disk) 873 { 874 return 0; 875 } 876 static inline void bd_unlink_disk_holder(struct block_device *bdev, 877 struct gendisk *disk) 878 { 879 } 880 #endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */ 881 882 dev_t part_devt(struct gendisk *disk, u8 partno); 883 void inc_diskseq(struct gendisk *disk); 884 void blk_request_module(dev_t devt); 885 886 extern int blk_register_queue(struct gendisk *disk); 887 extern void blk_unregister_queue(struct gendisk *disk); 888 void submit_bio_noacct(struct bio *bio); 889 struct bio *bio_split_to_limits(struct bio *bio); 890 891 extern int blk_lld_busy(struct request_queue *q); 892 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 893 extern void blk_queue_exit(struct request_queue *q); 894 extern void blk_sync_queue(struct request_queue *q); 895 896 /* Helper to convert REQ_OP_XXX to its string format XXX */ 897 extern const char *blk_op_str(enum req_op op); 898 899 int blk_status_to_errno(blk_status_t status); 900 blk_status_t errno_to_blk_status(int errno); 901 const char *blk_status_to_str(blk_status_t status); 902 903 /* only poll the hardware once, don't continue until a completion was found */ 904 #define BLK_POLL_ONESHOT (1 << 0) 905 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags); 906 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, 907 unsigned int flags); 908 909 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 910 { 911 return bdev->bd_queue; /* this is never NULL */ 912 } 913 914 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */ 915 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond); 916 917 static inline unsigned int bio_zone_no(struct bio *bio) 918 { 919 return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector); 920 } 921 922 static inline bool bio_straddles_zones(struct bio *bio) 923 { 924 return bio_sectors(bio) && 925 bio_zone_no(bio) != 926 disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1); 927 } 928 929 /* 930 * Return how much within the boundary is left to be used for I/O at a given 931 * offset. 932 */ 933 static inline unsigned int blk_boundary_sectors_left(sector_t offset, 934 unsigned int boundary_sectors) 935 { 936 if (unlikely(!is_power_of_2(boundary_sectors))) 937 return boundary_sectors - sector_div(offset, boundary_sectors); 938 return boundary_sectors - (offset & (boundary_sectors - 1)); 939 } 940 941 /** 942 * queue_limits_start_update - start an atomic update of queue limits 943 * @q: queue to update 944 * 945 * This functions starts an atomic update of the queue limits. It takes a lock 946 * to prevent other updates and returns a snapshot of the current limits that 947 * the caller can modify. The caller must call queue_limits_commit_update() 948 * to finish the update. 949 * 950 * Context: process context. 951 */ 952 static inline struct queue_limits 953 queue_limits_start_update(struct request_queue *q) 954 { 955 mutex_lock(&q->limits_lock); 956 return q->limits; 957 } 958 int queue_limits_commit_update_frozen(struct request_queue *q, 959 struct queue_limits *lim); 960 int queue_limits_commit_update(struct request_queue *q, 961 struct queue_limits *lim); 962 int queue_limits_set(struct request_queue *q, struct queue_limits *lim); 963 int blk_validate_limits(struct queue_limits *lim); 964 965 /** 966 * queue_limits_cancel_update - cancel an atomic update of queue limits 967 * @q: queue to update 968 * 969 * This functions cancels an atomic update of the queue limits started by 970 * queue_limits_start_update() and should be used when an error occurs after 971 * starting update. 972 */ 973 static inline void queue_limits_cancel_update(struct request_queue *q) 974 { 975 mutex_unlock(&q->limits_lock); 976 } 977 978 /* 979 * These helpers are for drivers that have sloppy feature negotiation and might 980 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O 981 * completion handler when the device returned an indicator that the respective 982 * feature is not actually supported. They are racy and the driver needs to 983 * cope with that. Try to avoid this scheme if you can. 984 */ 985 static inline void blk_queue_disable_discard(struct request_queue *q) 986 { 987 q->limits.max_discard_sectors = 0; 988 } 989 990 static inline void blk_queue_disable_secure_erase(struct request_queue *q) 991 { 992 q->limits.max_secure_erase_sectors = 0; 993 } 994 995 static inline void blk_queue_disable_write_zeroes(struct request_queue *q) 996 { 997 q->limits.max_write_zeroes_sectors = 0; 998 } 999 1000 /* 1001 * Access functions for manipulating queue properties 1002 */ 1003 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1004 extern void blk_set_stacking_limits(struct queue_limits *lim); 1005 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1006 sector_t offset); 1007 void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, 1008 sector_t offset, const char *pfx); 1009 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1010 1011 struct blk_independent_access_ranges * 1012 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges); 1013 void disk_set_independent_access_ranges(struct gendisk *disk, 1014 struct blk_independent_access_ranges *iars); 1015 1016 bool __must_check blk_get_queue(struct request_queue *); 1017 extern void blk_put_queue(struct request_queue *); 1018 1019 void blk_mark_disk_dead(struct gendisk *disk); 1020 1021 struct rq_list { 1022 struct request *head; 1023 struct request *tail; 1024 }; 1025 1026 #ifdef CONFIG_BLOCK 1027 /* 1028 * blk_plug permits building a queue of related requests by holding the I/O 1029 * fragments for a short period. This allows merging of sequential requests 1030 * into single larger request. As the requests are moved from a per-task list to 1031 * the device's request_queue in a batch, this results in improved scalability 1032 * as the lock contention for request_queue lock is reduced. 1033 * 1034 * It is ok not to disable preemption when adding the request to the plug list 1035 * or when attempting a merge. For details, please see schedule() where 1036 * blk_flush_plug() is called. 1037 */ 1038 struct blk_plug { 1039 struct rq_list mq_list; /* blk-mq requests */ 1040 1041 /* if ios_left is > 1, we can batch tag/rq allocations */ 1042 struct rq_list cached_rqs; 1043 u64 cur_ktime; 1044 unsigned short nr_ios; 1045 1046 unsigned short rq_count; 1047 1048 bool multiple_queues; 1049 bool has_elevator; 1050 1051 struct list_head cb_list; /* md requires an unplug callback */ 1052 }; 1053 1054 struct blk_plug_cb; 1055 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1056 struct blk_plug_cb { 1057 struct list_head list; 1058 blk_plug_cb_fn callback; 1059 void *data; 1060 }; 1061 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1062 void *data, int size); 1063 extern void blk_start_plug(struct blk_plug *); 1064 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short); 1065 extern void blk_finish_plug(struct blk_plug *); 1066 1067 void __blk_flush_plug(struct blk_plug *plug, bool from_schedule); 1068 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1069 { 1070 if (plug) 1071 __blk_flush_plug(plug, async); 1072 } 1073 1074 /* 1075 * tsk == current here 1076 */ 1077 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1078 { 1079 struct blk_plug *plug = tsk->plug; 1080 1081 if (plug) 1082 plug->cur_ktime = 0; 1083 current->flags &= ~PF_BLOCK_TS; 1084 } 1085 1086 int blkdev_issue_flush(struct block_device *bdev); 1087 long nr_blockdev_pages(void); 1088 #else /* CONFIG_BLOCK */ 1089 struct blk_plug { 1090 }; 1091 1092 static inline void blk_start_plug_nr_ios(struct blk_plug *plug, 1093 unsigned short nr_ios) 1094 { 1095 } 1096 1097 static inline void blk_start_plug(struct blk_plug *plug) 1098 { 1099 } 1100 1101 static inline void blk_finish_plug(struct blk_plug *plug) 1102 { 1103 } 1104 1105 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1106 { 1107 } 1108 1109 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1110 { 1111 } 1112 1113 static inline int blkdev_issue_flush(struct block_device *bdev) 1114 { 1115 return 0; 1116 } 1117 1118 static inline long nr_blockdev_pages(void) 1119 { 1120 return 0; 1121 } 1122 #endif /* CONFIG_BLOCK */ 1123 1124 extern void blk_io_schedule(void); 1125 1126 int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1127 sector_t nr_sects, gfp_t gfp_mask); 1128 int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1129 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop); 1130 int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, 1131 sector_t nr_sects, gfp_t gfp); 1132 1133 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1134 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1135 #define BLKDEV_ZERO_KILLABLE (1 << 2) /* interruptible by fatal signals */ 1136 1137 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1138 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1139 unsigned flags); 1140 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1141 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1142 1143 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1144 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1145 { 1146 return blkdev_issue_discard(sb->s_bdev, 1147 block << (sb->s_blocksize_bits - 1148 SECTOR_SHIFT), 1149 nr_blocks << (sb->s_blocksize_bits - 1150 SECTOR_SHIFT), 1151 gfp_mask); 1152 } 1153 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1154 sector_t nr_blocks, gfp_t gfp_mask) 1155 { 1156 return blkdev_issue_zeroout(sb->s_bdev, 1157 block << (sb->s_blocksize_bits - 1158 SECTOR_SHIFT), 1159 nr_blocks << (sb->s_blocksize_bits - 1160 SECTOR_SHIFT), 1161 gfp_mask, 0); 1162 } 1163 1164 static inline bool bdev_is_partition(struct block_device *bdev) 1165 { 1166 return bdev_partno(bdev) != 0; 1167 } 1168 1169 enum blk_default_limits { 1170 BLK_MAX_SEGMENTS = 128, 1171 BLK_SAFE_MAX_SECTORS = 255, 1172 BLK_MAX_SEGMENT_SIZE = 65536, 1173 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1174 }; 1175 1176 /* 1177 * Default upper limit for the software max_sectors limit used for 1178 * regular file system I/O. This can be increased through sysfs. 1179 * 1180 * Not to be confused with the max_hw_sector limit that is entirely 1181 * controlled by the driver, usually based on hardware limits. 1182 */ 1183 #define BLK_DEF_MAX_SECTORS_CAP 2560u 1184 1185 static inline struct queue_limits *bdev_limits(struct block_device *bdev) 1186 { 1187 return &bdev_get_queue(bdev)->limits; 1188 } 1189 1190 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1191 { 1192 return q->limits.seg_boundary_mask; 1193 } 1194 1195 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1196 { 1197 return q->limits.virt_boundary_mask; 1198 } 1199 1200 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1201 { 1202 return q->limits.max_sectors; 1203 } 1204 1205 static inline unsigned int queue_max_bytes(struct request_queue *q) 1206 { 1207 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9; 1208 } 1209 1210 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1211 { 1212 return q->limits.max_hw_sectors; 1213 } 1214 1215 static inline unsigned short queue_max_segments(const struct request_queue *q) 1216 { 1217 return q->limits.max_segments; 1218 } 1219 1220 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1221 { 1222 return q->limits.max_discard_segments; 1223 } 1224 1225 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1226 { 1227 return q->limits.max_segment_size; 1228 } 1229 1230 static inline bool queue_emulates_zone_append(struct request_queue *q) 1231 { 1232 return blk_queue_is_zoned(q) && !q->limits.max_hw_zone_append_sectors; 1233 } 1234 1235 static inline bool bdev_emulates_zone_append(struct block_device *bdev) 1236 { 1237 return queue_emulates_zone_append(bdev_get_queue(bdev)); 1238 } 1239 1240 static inline unsigned int 1241 bdev_max_zone_append_sectors(struct block_device *bdev) 1242 { 1243 return bdev_limits(bdev)->max_zone_append_sectors; 1244 } 1245 1246 static inline unsigned int bdev_max_segments(struct block_device *bdev) 1247 { 1248 return queue_max_segments(bdev_get_queue(bdev)); 1249 } 1250 1251 static inline unsigned queue_logical_block_size(const struct request_queue *q) 1252 { 1253 return q->limits.logical_block_size; 1254 } 1255 1256 static inline unsigned int bdev_logical_block_size(struct block_device *bdev) 1257 { 1258 return queue_logical_block_size(bdev_get_queue(bdev)); 1259 } 1260 1261 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1262 { 1263 return q->limits.physical_block_size; 1264 } 1265 1266 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1267 { 1268 return queue_physical_block_size(bdev_get_queue(bdev)); 1269 } 1270 1271 static inline unsigned int queue_io_min(const struct request_queue *q) 1272 { 1273 return q->limits.io_min; 1274 } 1275 1276 static inline unsigned int bdev_io_min(struct block_device *bdev) 1277 { 1278 return queue_io_min(bdev_get_queue(bdev)); 1279 } 1280 1281 static inline unsigned int queue_io_opt(const struct request_queue *q) 1282 { 1283 return q->limits.io_opt; 1284 } 1285 1286 static inline unsigned int bdev_io_opt(struct block_device *bdev) 1287 { 1288 return queue_io_opt(bdev_get_queue(bdev)); 1289 } 1290 1291 static inline unsigned int 1292 queue_zone_write_granularity(const struct request_queue *q) 1293 { 1294 return q->limits.zone_write_granularity; 1295 } 1296 1297 static inline unsigned int 1298 bdev_zone_write_granularity(struct block_device *bdev) 1299 { 1300 return queue_zone_write_granularity(bdev_get_queue(bdev)); 1301 } 1302 1303 int bdev_alignment_offset(struct block_device *bdev); 1304 unsigned int bdev_discard_alignment(struct block_device *bdev); 1305 1306 static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev) 1307 { 1308 return bdev_limits(bdev)->max_discard_sectors; 1309 } 1310 1311 static inline unsigned int bdev_discard_granularity(struct block_device *bdev) 1312 { 1313 return bdev_limits(bdev)->discard_granularity; 1314 } 1315 1316 static inline unsigned int 1317 bdev_max_secure_erase_sectors(struct block_device *bdev) 1318 { 1319 return bdev_limits(bdev)->max_secure_erase_sectors; 1320 } 1321 1322 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1323 { 1324 return bdev_limits(bdev)->max_write_zeroes_sectors; 1325 } 1326 1327 static inline bool bdev_nonrot(struct block_device *bdev) 1328 { 1329 return blk_queue_nonrot(bdev_get_queue(bdev)); 1330 } 1331 1332 static inline bool bdev_synchronous(struct block_device *bdev) 1333 { 1334 return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS; 1335 } 1336 1337 static inline bool bdev_stable_writes(struct block_device *bdev) 1338 { 1339 struct request_queue *q = bdev_get_queue(bdev); 1340 1341 if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && 1342 q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) 1343 return true; 1344 return q->limits.features & BLK_FEAT_STABLE_WRITES; 1345 } 1346 1347 static inline bool blk_queue_write_cache(struct request_queue *q) 1348 { 1349 return (q->limits.features & BLK_FEAT_WRITE_CACHE) && 1350 !(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED); 1351 } 1352 1353 static inline bool bdev_write_cache(struct block_device *bdev) 1354 { 1355 return blk_queue_write_cache(bdev_get_queue(bdev)); 1356 } 1357 1358 static inline bool bdev_fua(struct block_device *bdev) 1359 { 1360 return bdev_limits(bdev)->features & BLK_FEAT_FUA; 1361 } 1362 1363 static inline bool bdev_nowait(struct block_device *bdev) 1364 { 1365 return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT; 1366 } 1367 1368 static inline bool bdev_is_zoned(struct block_device *bdev) 1369 { 1370 return blk_queue_is_zoned(bdev_get_queue(bdev)); 1371 } 1372 1373 static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec) 1374 { 1375 return disk_zone_no(bdev->bd_disk, sec); 1376 } 1377 1378 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1379 { 1380 struct request_queue *q = bdev_get_queue(bdev); 1381 1382 if (!blk_queue_is_zoned(q)) 1383 return 0; 1384 return q->limits.chunk_sectors; 1385 } 1386 1387 static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev, 1388 sector_t sector) 1389 { 1390 return sector & (bdev_zone_sectors(bdev) - 1); 1391 } 1392 1393 static inline sector_t bio_offset_from_zone_start(struct bio *bio) 1394 { 1395 return bdev_offset_from_zone_start(bio->bi_bdev, 1396 bio->bi_iter.bi_sector); 1397 } 1398 1399 static inline bool bdev_is_zone_start(struct block_device *bdev, 1400 sector_t sector) 1401 { 1402 return bdev_offset_from_zone_start(bdev, sector) == 0; 1403 } 1404 1405 /** 1406 * bdev_zone_is_seq - check if a sector belongs to a sequential write zone 1407 * @bdev: block device to check 1408 * @sector: sector number 1409 * 1410 * Check if @sector on @bdev is contained in a sequential write required zone. 1411 */ 1412 static inline bool bdev_zone_is_seq(struct block_device *bdev, sector_t sector) 1413 { 1414 bool is_seq = false; 1415 1416 #if IS_ENABLED(CONFIG_BLK_DEV_ZONED) 1417 if (bdev_is_zoned(bdev)) { 1418 struct gendisk *disk = bdev->bd_disk; 1419 unsigned long *bitmap; 1420 1421 rcu_read_lock(); 1422 bitmap = rcu_dereference(disk->conv_zones_bitmap); 1423 is_seq = !bitmap || 1424 !test_bit(disk_zone_no(disk, sector), bitmap); 1425 rcu_read_unlock(); 1426 } 1427 #endif 1428 1429 return is_seq; 1430 } 1431 1432 int blk_zone_issue_zeroout(struct block_device *bdev, sector_t sector, 1433 sector_t nr_sects, gfp_t gfp_mask); 1434 1435 static inline unsigned int queue_dma_alignment(const struct request_queue *q) 1436 { 1437 return q->limits.dma_alignment; 1438 } 1439 1440 static inline unsigned int 1441 queue_atomic_write_unit_max_bytes(const struct request_queue *q) 1442 { 1443 return q->limits.atomic_write_unit_max; 1444 } 1445 1446 static inline unsigned int 1447 queue_atomic_write_unit_min_bytes(const struct request_queue *q) 1448 { 1449 return q->limits.atomic_write_unit_min; 1450 } 1451 1452 static inline unsigned int 1453 queue_atomic_write_boundary_bytes(const struct request_queue *q) 1454 { 1455 return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT; 1456 } 1457 1458 static inline unsigned int 1459 queue_atomic_write_max_bytes(const struct request_queue *q) 1460 { 1461 return q->limits.atomic_write_max_sectors << SECTOR_SHIFT; 1462 } 1463 1464 static inline unsigned int bdev_dma_alignment(struct block_device *bdev) 1465 { 1466 return queue_dma_alignment(bdev_get_queue(bdev)); 1467 } 1468 1469 static inline bool bdev_iter_is_aligned(struct block_device *bdev, 1470 struct iov_iter *iter) 1471 { 1472 return iov_iter_is_aligned(iter, bdev_dma_alignment(bdev), 1473 bdev_logical_block_size(bdev) - 1); 1474 } 1475 1476 static inline unsigned int 1477 blk_lim_dma_alignment_and_pad(struct queue_limits *lim) 1478 { 1479 return lim->dma_alignment | lim->dma_pad_mask; 1480 } 1481 1482 static inline bool blk_rq_aligned(struct request_queue *q, unsigned long addr, 1483 unsigned int len) 1484 { 1485 unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits); 1486 1487 return !(addr & alignment) && !(len & alignment); 1488 } 1489 1490 /* assumes size > 256 */ 1491 static inline unsigned int blksize_bits(unsigned int size) 1492 { 1493 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT; 1494 } 1495 1496 int kblockd_schedule_work(struct work_struct *work); 1497 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1498 1499 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1500 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1501 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1502 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1503 1504 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 1505 1506 bool blk_crypto_register(struct blk_crypto_profile *profile, 1507 struct request_queue *q); 1508 1509 #else /* CONFIG_BLK_INLINE_ENCRYPTION */ 1510 1511 static inline bool blk_crypto_register(struct blk_crypto_profile *profile, 1512 struct request_queue *q) 1513 { 1514 return true; 1515 } 1516 1517 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */ 1518 1519 enum blk_unique_id { 1520 /* these match the Designator Types specified in SPC */ 1521 BLK_UID_T10 = 1, 1522 BLK_UID_EUI64 = 2, 1523 BLK_UID_NAA = 3, 1524 }; 1525 1526 struct block_device_operations { 1527 void (*submit_bio)(struct bio *bio); 1528 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob, 1529 unsigned int flags); 1530 int (*open)(struct gendisk *disk, blk_mode_t mode); 1531 void (*release)(struct gendisk *disk); 1532 int (*ioctl)(struct block_device *bdev, blk_mode_t mode, 1533 unsigned cmd, unsigned long arg); 1534 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode, 1535 unsigned cmd, unsigned long arg); 1536 unsigned int (*check_events) (struct gendisk *disk, 1537 unsigned int clearing); 1538 void (*unlock_native_capacity) (struct gendisk *); 1539 int (*getgeo)(struct block_device *, struct hd_geometry *); 1540 int (*set_read_only)(struct block_device *bdev, bool ro); 1541 void (*free_disk)(struct gendisk *disk); 1542 /* this callback is with swap_lock and sometimes page table lock held */ 1543 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1544 int (*report_zones)(struct gendisk *, sector_t sector, 1545 unsigned int nr_zones, report_zones_cb cb, void *data); 1546 char *(*devnode)(struct gendisk *disk, umode_t *mode); 1547 /* returns the length of the identifier or a negative errno: */ 1548 int (*get_unique_id)(struct gendisk *disk, u8 id[16], 1549 enum blk_unique_id id_type); 1550 struct module *owner; 1551 const struct pr_ops *pr_ops; 1552 1553 /* 1554 * Special callback for probing GPT entry at a given sector. 1555 * Needed by Android devices, used by GPT scanner and MMC blk 1556 * driver. 1557 */ 1558 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector); 1559 }; 1560 1561 #ifdef CONFIG_COMPAT 1562 extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t, 1563 unsigned int, unsigned long); 1564 #else 1565 #define blkdev_compat_ptr_ioctl NULL 1566 #endif 1567 1568 static inline void blk_wake_io_task(struct task_struct *waiter) 1569 { 1570 /* 1571 * If we're polling, the task itself is doing the completions. For 1572 * that case, we don't need to signal a wakeup, it's enough to just 1573 * mark us as RUNNING. 1574 */ 1575 if (waiter == current) 1576 __set_current_state(TASK_RUNNING); 1577 else 1578 wake_up_process(waiter); 1579 } 1580 1581 unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, 1582 unsigned long start_time); 1583 void bdev_end_io_acct(struct block_device *bdev, enum req_op op, 1584 unsigned int sectors, unsigned long start_time); 1585 1586 unsigned long bio_start_io_acct(struct bio *bio); 1587 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, 1588 struct block_device *orig_bdev); 1589 1590 /** 1591 * bio_end_io_acct - end I/O accounting for bio based drivers 1592 * @bio: bio to end account for 1593 * @start_time: start time returned by bio_start_io_acct() 1594 */ 1595 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time) 1596 { 1597 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev); 1598 } 1599 1600 int set_blocksize(struct file *file, int size); 1601 1602 int lookup_bdev(const char *pathname, dev_t *dev); 1603 1604 void blkdev_show(struct seq_file *seqf, off_t offset); 1605 1606 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */ 1607 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */ 1608 #ifdef CONFIG_BLOCK 1609 #define BLKDEV_MAJOR_MAX 512 1610 #else 1611 #define BLKDEV_MAJOR_MAX 0 1612 #endif 1613 1614 struct blk_holder_ops { 1615 void (*mark_dead)(struct block_device *bdev, bool surprise); 1616 1617 /* 1618 * Sync the file system mounted on the block device. 1619 */ 1620 void (*sync)(struct block_device *bdev); 1621 1622 /* 1623 * Freeze the file system mounted on the block device. 1624 */ 1625 int (*freeze)(struct block_device *bdev); 1626 1627 /* 1628 * Thaw the file system mounted on the block device. 1629 */ 1630 int (*thaw)(struct block_device *bdev); 1631 }; 1632 1633 /* 1634 * For filesystems using @fs_holder_ops, the @holder argument passed to 1635 * helpers used to open and claim block devices via 1636 * bd_prepare_to_claim() must point to a superblock. 1637 */ 1638 extern const struct blk_holder_ops fs_holder_ops; 1639 1640 /* 1641 * Return the correct open flags for blkdev_get_by_* for super block flags 1642 * as stored in sb->s_flags. 1643 */ 1644 #define sb_open_mode(flags) \ 1645 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \ 1646 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE)) 1647 1648 struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder, 1649 const struct blk_holder_ops *hops); 1650 struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode, 1651 void *holder, const struct blk_holder_ops *hops); 1652 int bd_prepare_to_claim(struct block_device *bdev, void *holder, 1653 const struct blk_holder_ops *hops); 1654 void bd_abort_claiming(struct block_device *bdev, void *holder); 1655 1656 /* just for blk-cgroup, don't use elsewhere */ 1657 struct block_device *blkdev_get_no_open(dev_t dev); 1658 void blkdev_put_no_open(struct block_device *bdev); 1659 1660 struct block_device *I_BDEV(struct inode *inode); 1661 struct block_device *file_bdev(struct file *bdev_file); 1662 bool disk_live(struct gendisk *disk); 1663 unsigned int block_size(struct block_device *bdev); 1664 1665 #ifdef CONFIG_BLOCK 1666 void invalidate_bdev(struct block_device *bdev); 1667 int sync_blockdev(struct block_device *bdev); 1668 int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend); 1669 int sync_blockdev_nowait(struct block_device *bdev); 1670 void sync_bdevs(bool wait); 1671 void bdev_statx(struct path *, struct kstat *, u32); 1672 void printk_all_partitions(void); 1673 int __init early_lookup_bdev(const char *pathname, dev_t *dev); 1674 #else 1675 static inline void invalidate_bdev(struct block_device *bdev) 1676 { 1677 } 1678 static inline int sync_blockdev(struct block_device *bdev) 1679 { 1680 return 0; 1681 } 1682 static inline int sync_blockdev_nowait(struct block_device *bdev) 1683 { 1684 return 0; 1685 } 1686 static inline void sync_bdevs(bool wait) 1687 { 1688 } 1689 static inline void bdev_statx(struct path *path, struct kstat *stat, 1690 u32 request_mask) 1691 { 1692 } 1693 static inline void printk_all_partitions(void) 1694 { 1695 } 1696 static inline int early_lookup_bdev(const char *pathname, dev_t *dev) 1697 { 1698 return -EINVAL; 1699 } 1700 #endif /* CONFIG_BLOCK */ 1701 1702 int bdev_freeze(struct block_device *bdev); 1703 int bdev_thaw(struct block_device *bdev); 1704 void bdev_fput(struct file *bdev_file); 1705 1706 struct io_comp_batch { 1707 struct rq_list req_list; 1708 bool need_ts; 1709 void (*complete)(struct io_comp_batch *); 1710 }; 1711 1712 static inline bool blk_atomic_write_start_sect_aligned(sector_t sector, 1713 struct queue_limits *limits) 1714 { 1715 unsigned int alignment = max(limits->atomic_write_hw_unit_min, 1716 limits->atomic_write_hw_boundary); 1717 1718 return IS_ALIGNED(sector, alignment >> SECTOR_SHIFT); 1719 } 1720 1721 static inline bool bdev_can_atomic_write(struct block_device *bdev) 1722 { 1723 struct request_queue *bd_queue = bdev->bd_queue; 1724 struct queue_limits *limits = &bd_queue->limits; 1725 1726 if (!limits->atomic_write_unit_min) 1727 return false; 1728 1729 if (bdev_is_partition(bdev)) 1730 return blk_atomic_write_start_sect_aligned(bdev->bd_start_sect, 1731 limits); 1732 1733 return true; 1734 } 1735 1736 static inline unsigned int 1737 bdev_atomic_write_unit_min_bytes(struct block_device *bdev) 1738 { 1739 if (!bdev_can_atomic_write(bdev)) 1740 return 0; 1741 return queue_atomic_write_unit_min_bytes(bdev_get_queue(bdev)); 1742 } 1743 1744 static inline unsigned int 1745 bdev_atomic_write_unit_max_bytes(struct block_device *bdev) 1746 { 1747 if (!bdev_can_atomic_write(bdev)) 1748 return 0; 1749 return queue_atomic_write_unit_max_bytes(bdev_get_queue(bdev)); 1750 } 1751 1752 #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { } 1753 1754 #endif /* _LINUX_BLKDEV_H */ 1755