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 atomic_t nr_zone_wplugs; 201 spinlock_t zone_wplugs_lock; 202 struct mempool_s *zone_wplugs_pool; 203 struct hlist_head *zone_wplugs_hash; 204 struct workqueue_struct *zone_wplugs_wq; 205 #endif /* CONFIG_BLK_DEV_ZONED */ 206 207 #if IS_ENABLED(CONFIG_CDROM) 208 struct cdrom_device_info *cdi; 209 #endif 210 int node_id; 211 struct badblocks *bb; 212 struct lockdep_map lockdep_map; 213 u64 diskseq; 214 blk_mode_t open_mode; 215 216 /* 217 * Independent sector access ranges. This is always NULL for 218 * devices that do not have multiple independent access ranges. 219 */ 220 struct blk_independent_access_ranges *ia_ranges; 221 }; 222 223 /** 224 * disk_openers - returns how many openers are there for a disk 225 * @disk: disk to check 226 * 227 * This returns the number of openers for a disk. Note that this value is only 228 * stable if disk->open_mutex is held. 229 * 230 * Note: Due to a quirk in the block layer open code, each open partition is 231 * only counted once even if there are multiple openers. 232 */ 233 static inline unsigned int disk_openers(struct gendisk *disk) 234 { 235 return atomic_read(&disk->part0->bd_openers); 236 } 237 238 /** 239 * disk_has_partscan - return %true if partition scanning is enabled on a disk 240 * @disk: disk to check 241 * 242 * Returns %true if partitions scanning is enabled for @disk, or %false if 243 * partition scanning is disabled either permanently or temporarily. 244 */ 245 static inline bool disk_has_partscan(struct gendisk *disk) 246 { 247 return !(disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN)) && 248 !test_bit(GD_SUPPRESS_PART_SCAN, &disk->state); 249 } 250 251 /* 252 * The gendisk is refcounted by the part0 block_device, and the bd_device 253 * therein is also used for device model presentation in sysfs. 254 */ 255 #define dev_to_disk(device) \ 256 (dev_to_bdev(device)->bd_disk) 257 #define disk_to_dev(disk) \ 258 (&((disk)->part0->bd_device)) 259 260 #if IS_REACHABLE(CONFIG_CDROM) 261 #define disk_to_cdi(disk) ((disk)->cdi) 262 #else 263 #define disk_to_cdi(disk) NULL 264 #endif 265 266 static inline dev_t disk_devt(struct gendisk *disk) 267 { 268 return MKDEV(disk->major, disk->first_minor); 269 } 270 271 /* blk_validate_limits() validates bsize, so drivers don't usually need to */ 272 static inline int blk_validate_block_size(unsigned long bsize) 273 { 274 if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(bsize)) 275 return -EINVAL; 276 277 return 0; 278 } 279 280 static inline bool blk_op_is_passthrough(blk_opf_t op) 281 { 282 op &= REQ_OP_MASK; 283 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 284 } 285 286 /* flags set by the driver in queue_limits.features */ 287 typedef unsigned int __bitwise blk_features_t; 288 289 /* supports a volatile write cache */ 290 #define BLK_FEAT_WRITE_CACHE ((__force blk_features_t)(1u << 0)) 291 292 /* supports passing on the FUA bit */ 293 #define BLK_FEAT_FUA ((__force blk_features_t)(1u << 1)) 294 295 /* rotational device (hard drive or floppy) */ 296 #define BLK_FEAT_ROTATIONAL ((__force blk_features_t)(1u << 2)) 297 298 /* contributes to the random number pool */ 299 #define BLK_FEAT_ADD_RANDOM ((__force blk_features_t)(1u << 3)) 300 301 /* do disk/partitions IO accounting */ 302 #define BLK_FEAT_IO_STAT ((__force blk_features_t)(1u << 4)) 303 304 /* don't modify data until writeback is done */ 305 #define BLK_FEAT_STABLE_WRITES ((__force blk_features_t)(1u << 5)) 306 307 /* always completes in submit context */ 308 #define BLK_FEAT_SYNCHRONOUS ((__force blk_features_t)(1u << 6)) 309 310 /* supports REQ_NOWAIT */ 311 #define BLK_FEAT_NOWAIT ((__force blk_features_t)(1u << 7)) 312 313 /* supports DAX */ 314 #define BLK_FEAT_DAX ((__force blk_features_t)(1u << 8)) 315 316 /* supports I/O polling */ 317 #define BLK_FEAT_POLL ((__force blk_features_t)(1u << 9)) 318 319 /* is a zoned device */ 320 #define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10)) 321 322 /* supports PCI(e) p2p requests */ 323 #define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12)) 324 325 /* skip this queue in blk_mq_(un)quiesce_tagset */ 326 #define BLK_FEAT_SKIP_TAGSET_QUIESCE ((__force blk_features_t)(1u << 13)) 327 328 /* bounce all highmem pages */ 329 #define BLK_FEAT_BOUNCE_HIGH ((__force blk_features_t)(1u << 14)) 330 331 /* undocumented magic for bcache */ 332 #define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \ 333 ((__force blk_features_t)(1u << 15)) 334 335 /* atomic writes enabled */ 336 #define BLK_FEAT_ATOMIC_WRITES \ 337 ((__force blk_features_t)(1u << 16)) 338 339 /* 340 * Flags automatically inherited when stacking limits. 341 */ 342 #define BLK_FEAT_INHERIT_MASK \ 343 (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ 344 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH | \ 345 BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE) 346 347 /* internal flags in queue_limits.flags */ 348 typedef unsigned int __bitwise blk_flags_t; 349 350 /* do not send FLUSH/FUA commands despite advertising a write cache */ 351 #define BLK_FLAG_WRITE_CACHE_DISABLED ((__force blk_flags_t)(1u << 0)) 352 353 /* I/O topology is misaligned */ 354 #define BLK_FLAG_MISALIGNED ((__force blk_flags_t)(1u << 1)) 355 356 /* passthrough command IO accounting */ 357 #define BLK_FLAG_IOSTATS_PASSTHROUGH ((__force blk_flags_t)(1u << 2)) 358 359 struct queue_limits { 360 blk_features_t features; 361 blk_flags_t flags; 362 unsigned long seg_boundary_mask; 363 unsigned long virt_boundary_mask; 364 365 unsigned int max_hw_sectors; 366 unsigned int max_dev_sectors; 367 unsigned int chunk_sectors; 368 unsigned int max_sectors; 369 unsigned int max_user_sectors; 370 unsigned int max_segment_size; 371 unsigned int min_segment_size; 372 unsigned int physical_block_size; 373 unsigned int logical_block_size; 374 unsigned int alignment_offset; 375 unsigned int io_min; 376 unsigned int io_opt; 377 unsigned int max_discard_sectors; 378 unsigned int max_hw_discard_sectors; 379 unsigned int max_user_discard_sectors; 380 unsigned int max_secure_erase_sectors; 381 unsigned int max_write_zeroes_sectors; 382 unsigned int max_hw_zone_append_sectors; 383 unsigned int max_zone_append_sectors; 384 unsigned int discard_granularity; 385 unsigned int discard_alignment; 386 unsigned int zone_write_granularity; 387 388 /* atomic write limits */ 389 unsigned int atomic_write_hw_max; 390 unsigned int atomic_write_max_sectors; 391 unsigned int atomic_write_hw_boundary; 392 unsigned int atomic_write_boundary_sectors; 393 unsigned int atomic_write_hw_unit_min; 394 unsigned int atomic_write_unit_min; 395 unsigned int atomic_write_hw_unit_max; 396 unsigned int atomic_write_unit_max; 397 398 unsigned short max_segments; 399 unsigned short max_integrity_segments; 400 unsigned short max_discard_segments; 401 402 unsigned int max_open_zones; 403 unsigned int max_active_zones; 404 405 /* 406 * Drivers that set dma_alignment to less than 511 must be prepared to 407 * handle individual bvec's that are not a multiple of a SECTOR_SIZE 408 * due to possible offsets. 409 */ 410 unsigned int dma_alignment; 411 unsigned int dma_pad_mask; 412 413 struct blk_integrity integrity; 414 }; 415 416 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, 417 void *data); 418 419 #define BLK_ALL_ZONES ((unsigned int)-1) 420 int blkdev_report_zones(struct block_device *bdev, sector_t sector, 421 unsigned int nr_zones, report_zones_cb cb, void *data); 422 int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, 423 sector_t sectors, sector_t nr_sectors); 424 int blk_revalidate_disk_zones(struct gendisk *disk); 425 426 /* 427 * Independent access ranges: struct blk_independent_access_range describes 428 * a range of contiguous sectors that can be accessed using device command 429 * execution resources that are independent from the resources used for 430 * other access ranges. This is typically found with single-LUN multi-actuator 431 * HDDs where each access range is served by a different set of heads. 432 * The set of independent ranges supported by the device is defined using 433 * struct blk_independent_access_ranges. The independent ranges must not overlap 434 * and must include all sectors within the disk capacity (no sector holes 435 * allowed). 436 * For a device with multiple ranges, requests targeting sectors in different 437 * ranges can be executed in parallel. A request can straddle an access range 438 * boundary. 439 */ 440 struct blk_independent_access_range { 441 struct kobject kobj; 442 sector_t sector; 443 sector_t nr_sectors; 444 }; 445 446 struct blk_independent_access_ranges { 447 struct kobject kobj; 448 bool sysfs_registered; 449 unsigned int nr_ia_ranges; 450 struct blk_independent_access_range ia_range[]; 451 }; 452 453 struct request_queue { 454 /* 455 * The queue owner gets to use this for whatever they like. 456 * ll_rw_blk doesn't touch it. 457 */ 458 void *queuedata; 459 460 struct elevator_queue *elevator; 461 462 const struct blk_mq_ops *mq_ops; 463 464 /* sw queues */ 465 struct blk_mq_ctx __percpu *queue_ctx; 466 467 /* 468 * various queue flags, see QUEUE_* below 469 */ 470 unsigned long queue_flags; 471 472 unsigned int rq_timeout; 473 474 unsigned int queue_depth; 475 476 refcount_t refs; 477 478 /* hw dispatch queues */ 479 unsigned int nr_hw_queues; 480 struct xarray hctx_table; 481 482 struct percpu_ref q_usage_counter; 483 struct lock_class_key io_lock_cls_key; 484 struct lockdep_map io_lockdep_map; 485 486 struct lock_class_key q_lock_cls_key; 487 struct lockdep_map q_lockdep_map; 488 489 struct request *last_merge; 490 491 spinlock_t queue_lock; 492 493 int quiesce_depth; 494 495 struct gendisk *disk; 496 497 /* 498 * mq queue kobject 499 */ 500 struct kobject *mq_kobj; 501 502 struct queue_limits limits; 503 504 #ifdef CONFIG_PM 505 struct device *dev; 506 enum rpm_status rpm_status; 507 #endif 508 509 /* 510 * Number of contexts that have called blk_set_pm_only(). If this 511 * counter is above zero then only RQF_PM requests are processed. 512 */ 513 atomic_t pm_only; 514 515 struct blk_queue_stats *stats; 516 struct rq_qos *rq_qos; 517 struct mutex rq_qos_mutex; 518 519 /* 520 * ida allocated id for this queue. Used to index queues from 521 * ioctx. 522 */ 523 int id; 524 525 /* 526 * queue settings 527 */ 528 unsigned long nr_requests; /* Max # of requests */ 529 530 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 531 struct blk_crypto_profile *crypto_profile; 532 struct kobject *crypto_kobject; 533 #endif 534 535 struct timer_list timeout; 536 struct work_struct timeout_work; 537 538 atomic_t nr_active_requests_shared_tags; 539 540 struct blk_mq_tags *sched_shared_tags; 541 542 struct list_head icq_list; 543 #ifdef CONFIG_BLK_CGROUP 544 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 545 struct blkcg_gq *root_blkg; 546 struct list_head blkg_list; 547 struct mutex blkcg_mutex; 548 #endif 549 550 int node; 551 552 spinlock_t requeue_lock; 553 struct list_head requeue_list; 554 struct delayed_work requeue_work; 555 556 #ifdef CONFIG_BLK_DEV_IO_TRACE 557 struct blk_trace __rcu *blk_trace; 558 #endif 559 /* 560 * for flush operations 561 */ 562 struct blk_flush_queue *fq; 563 struct list_head flush_list; 564 565 struct mutex sysfs_lock; 566 struct mutex limits_lock; 567 568 /* 569 * for reusing dead hctx instance in case of updating 570 * nr_hw_queues 571 */ 572 struct list_head unused_hctx_list; 573 spinlock_t unused_hctx_lock; 574 575 int mq_freeze_depth; 576 577 #ifdef CONFIG_BLK_DEV_THROTTLING 578 /* Throttle data */ 579 struct throtl_data *td; 580 #endif 581 struct rcu_head rcu_head; 582 #ifdef CONFIG_LOCKDEP 583 struct task_struct *mq_freeze_owner; 584 int mq_freeze_owner_depth; 585 /* 586 * Records disk & queue state in current context, used in unfreeze 587 * queue 588 */ 589 bool mq_freeze_disk_dead; 590 bool mq_freeze_queue_dying; 591 #endif 592 wait_queue_head_t mq_freeze_wq; 593 /* 594 * Protect concurrent access to q_usage_counter by 595 * percpu_ref_kill() and percpu_ref_reinit(). 596 */ 597 struct mutex mq_freeze_lock; 598 599 struct blk_mq_tag_set *tag_set; 600 struct list_head tag_set_list; 601 602 struct dentry *debugfs_dir; 603 struct dentry *sched_debugfs_dir; 604 struct dentry *rqos_debugfs_dir; 605 /* 606 * Serializes all debugfs metadata operations using the above dentries. 607 */ 608 struct mutex debugfs_mutex; 609 }; 610 611 /* Keep blk_queue_flag_name[] in sync with the definitions below */ 612 enum { 613 QUEUE_FLAG_DYING, /* queue being torn down */ 614 QUEUE_FLAG_NOMERGES, /* disable merge attempts */ 615 QUEUE_FLAG_SAME_COMP, /* complete on same CPU-group */ 616 QUEUE_FLAG_FAIL_IO, /* fake timeout */ 617 QUEUE_FLAG_NOXMERGES, /* No extended merges */ 618 QUEUE_FLAG_SAME_FORCE, /* force complete on same CPU */ 619 QUEUE_FLAG_INIT_DONE, /* queue is initialized */ 620 QUEUE_FLAG_STATS, /* track IO start and completion times */ 621 QUEUE_FLAG_REGISTERED, /* queue has been registered to a disk */ 622 QUEUE_FLAG_QUIESCED, /* queue has been quiesced */ 623 QUEUE_FLAG_RQ_ALLOC_TIME, /* record rq->alloc_time_ns */ 624 QUEUE_FLAG_HCTX_ACTIVE, /* at least one blk-mq hctx is active */ 625 QUEUE_FLAG_SQ_SCHED, /* single queue style io dispatch */ 626 QUEUE_FLAG_MAX 627 }; 628 629 #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) 630 631 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 632 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 633 634 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 635 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 636 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 637 #define blk_queue_noxmerges(q) \ 638 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 639 #define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL)) 640 #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) 641 #define blk_queue_passthrough_stat(q) \ 642 ((q)->limits.flags & BLK_FLAG_IOSTATS_PASSTHROUGH) 643 #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) 644 #define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA) 645 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 646 #define blk_queue_rq_alloc_time(q) \ 647 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) 648 #else 649 #define blk_queue_rq_alloc_time(q) false 650 #endif 651 652 #define blk_noretry_request(rq) \ 653 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 654 REQ_FAILFAST_DRIVER)) 655 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 656 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 657 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) 658 #define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags) 659 #define blk_queue_skip_tagset_quiesce(q) \ 660 ((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE) 661 662 extern void blk_set_pm_only(struct request_queue *q); 663 extern void blk_clear_pm_only(struct request_queue *q); 664 665 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 666 667 #define dma_map_bvec(dev, bv, dir, attrs) \ 668 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 669 (dir), (attrs)) 670 671 static inline bool queue_is_mq(struct request_queue *q) 672 { 673 return q->mq_ops; 674 } 675 676 #ifdef CONFIG_PM 677 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 678 { 679 return q->rpm_status; 680 } 681 #else 682 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 683 { 684 return RPM_ACTIVE; 685 } 686 #endif 687 688 static inline bool blk_queue_is_zoned(struct request_queue *q) 689 { 690 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && 691 (q->limits.features & BLK_FEAT_ZONED); 692 } 693 694 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) 695 { 696 if (!blk_queue_is_zoned(disk->queue)) 697 return 0; 698 return sector >> ilog2(disk->queue->limits.chunk_sectors); 699 } 700 701 static inline unsigned int bdev_max_open_zones(struct block_device *bdev) 702 { 703 return bdev->bd_disk->queue->limits.max_open_zones; 704 } 705 706 static inline unsigned int bdev_max_active_zones(struct block_device *bdev) 707 { 708 return bdev->bd_disk->queue->limits.max_active_zones; 709 } 710 711 static inline unsigned int blk_queue_depth(struct request_queue *q) 712 { 713 if (q->queue_depth) 714 return q->queue_depth; 715 716 return q->nr_requests; 717 } 718 719 /* 720 * default timeout for SG_IO if none specified 721 */ 722 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 723 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 724 725 /* This should not be used directly - use rq_for_each_segment */ 726 #define for_each_bio(_bio) \ 727 for (; _bio; _bio = _bio->bi_next) 728 729 int __must_check add_disk_fwnode(struct device *parent, struct gendisk *disk, 730 const struct attribute_group **groups, 731 struct fwnode_handle *fwnode); 732 int __must_check device_add_disk(struct device *parent, struct gendisk *disk, 733 const struct attribute_group **groups); 734 static inline int __must_check add_disk(struct gendisk *disk) 735 { 736 return device_add_disk(NULL, disk, NULL); 737 } 738 void del_gendisk(struct gendisk *gp); 739 void invalidate_disk(struct gendisk *disk); 740 void set_disk_ro(struct gendisk *disk, bool read_only); 741 void disk_uevent(struct gendisk *disk, enum kobject_action action); 742 743 static inline u8 bdev_partno(const struct block_device *bdev) 744 { 745 return atomic_read(&bdev->__bd_flags) & BD_PARTNO; 746 } 747 748 static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag) 749 { 750 return atomic_read(&bdev->__bd_flags) & flag; 751 } 752 753 static inline void bdev_set_flag(struct block_device *bdev, unsigned flag) 754 { 755 atomic_or(flag, &bdev->__bd_flags); 756 } 757 758 static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag) 759 { 760 atomic_andnot(flag, &bdev->__bd_flags); 761 } 762 763 static inline bool get_disk_ro(struct gendisk *disk) 764 { 765 return bdev_test_flag(disk->part0, BD_READ_ONLY) || 766 test_bit(GD_READ_ONLY, &disk->state); 767 } 768 769 static inline bool bdev_read_only(struct block_device *bdev) 770 { 771 return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk); 772 } 773 774 bool set_capacity_and_notify(struct gendisk *disk, sector_t size); 775 void disk_force_media_change(struct gendisk *disk); 776 void bdev_mark_dead(struct block_device *bdev, bool surprise); 777 778 void add_disk_randomness(struct gendisk *disk) __latent_entropy; 779 void rand_initialize_disk(struct gendisk *disk); 780 781 static inline sector_t get_start_sect(struct block_device *bdev) 782 { 783 return bdev->bd_start_sect; 784 } 785 786 static inline sector_t bdev_nr_sectors(struct block_device *bdev) 787 { 788 return bdev->bd_nr_sectors; 789 } 790 791 static inline loff_t bdev_nr_bytes(struct block_device *bdev) 792 { 793 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT; 794 } 795 796 static inline sector_t get_capacity(struct gendisk *disk) 797 { 798 return bdev_nr_sectors(disk->part0); 799 } 800 801 static inline u64 sb_bdev_nr_blocks(struct super_block *sb) 802 { 803 return bdev_nr_sectors(sb->s_bdev) >> 804 (sb->s_blocksize_bits - SECTOR_SHIFT); 805 } 806 807 #ifdef CONFIG_BLK_DEV_ZONED 808 static inline unsigned int disk_nr_zones(struct gendisk *disk) 809 { 810 return disk->nr_zones; 811 } 812 bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); 813 814 /** 815 * disk_zone_capacity - returns the zone capacity of zone containing @sector 816 * @disk: disk to work with 817 * @sector: sector number within the querying zone 818 * 819 * Returns the zone capacity of a zone containing @sector. @sector can be any 820 * sector in the zone. 821 */ 822 static inline unsigned int disk_zone_capacity(struct gendisk *disk, 823 sector_t sector) 824 { 825 sector_t zone_sectors = disk->queue->limits.chunk_sectors; 826 827 if (sector + zone_sectors >= get_capacity(disk)) 828 return disk->last_zone_capacity; 829 return disk->zone_capacity; 830 } 831 static inline unsigned int bdev_zone_capacity(struct block_device *bdev, 832 sector_t pos) 833 { 834 return disk_zone_capacity(bdev->bd_disk, pos); 835 } 836 #else /* CONFIG_BLK_DEV_ZONED */ 837 static inline unsigned int disk_nr_zones(struct gendisk *disk) 838 { 839 return 0; 840 } 841 static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) 842 { 843 return false; 844 } 845 #endif /* CONFIG_BLK_DEV_ZONED */ 846 847 static inline unsigned int bdev_nr_zones(struct block_device *bdev) 848 { 849 return disk_nr_zones(bdev->bd_disk); 850 } 851 852 int bdev_disk_changed(struct gendisk *disk, bool invalidate); 853 854 void put_disk(struct gendisk *disk); 855 struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node, 856 struct lock_class_key *lkclass); 857 858 /** 859 * blk_alloc_disk - allocate a gendisk structure 860 * @lim: queue limits to be used for this disk. 861 * @node_id: numa node to allocate on 862 * 863 * Allocate and pre-initialize a gendisk structure for use with BIO based 864 * drivers. 865 * 866 * Returns an ERR_PTR on error, else the allocated disk. 867 * 868 * Context: can sleep 869 */ 870 #define blk_alloc_disk(lim, node_id) \ 871 ({ \ 872 static struct lock_class_key __key; \ 873 \ 874 __blk_alloc_disk(lim, node_id, &__key); \ 875 }) 876 877 int __register_blkdev(unsigned int major, const char *name, 878 void (*probe)(dev_t devt)); 879 #define register_blkdev(major, name) \ 880 __register_blkdev(major, name, NULL) 881 void unregister_blkdev(unsigned int major, const char *name); 882 883 bool disk_check_media_change(struct gendisk *disk); 884 void set_capacity(struct gendisk *disk, sector_t size); 885 886 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 887 int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk); 888 void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk); 889 #else 890 static inline int bd_link_disk_holder(struct block_device *bdev, 891 struct gendisk *disk) 892 { 893 return 0; 894 } 895 static inline void bd_unlink_disk_holder(struct block_device *bdev, 896 struct gendisk *disk) 897 { 898 } 899 #endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */ 900 901 dev_t part_devt(struct gendisk *disk, u8 partno); 902 void inc_diskseq(struct gendisk *disk); 903 void blk_request_module(dev_t devt); 904 905 extern int blk_register_queue(struct gendisk *disk); 906 extern void blk_unregister_queue(struct gendisk *disk); 907 void submit_bio_noacct(struct bio *bio); 908 struct bio *bio_split_to_limits(struct bio *bio); 909 910 extern int blk_lld_busy(struct request_queue *q); 911 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 912 extern void blk_queue_exit(struct request_queue *q); 913 extern void blk_sync_queue(struct request_queue *q); 914 915 /* Helper to convert REQ_OP_XXX to its string format XXX */ 916 extern const char *blk_op_str(enum req_op op); 917 918 int blk_status_to_errno(blk_status_t status); 919 blk_status_t errno_to_blk_status(int errno); 920 const char *blk_status_to_str(blk_status_t status); 921 922 /* only poll the hardware once, don't continue until a completion was found */ 923 #define BLK_POLL_ONESHOT (1 << 0) 924 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags); 925 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, 926 unsigned int flags); 927 928 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 929 { 930 return bdev->bd_queue; /* this is never NULL */ 931 } 932 933 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */ 934 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond); 935 936 static inline unsigned int bio_zone_no(struct bio *bio) 937 { 938 return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector); 939 } 940 941 static inline bool bio_straddles_zones(struct bio *bio) 942 { 943 return bio_sectors(bio) && 944 bio_zone_no(bio) != 945 disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1); 946 } 947 948 /* 949 * Return how much within the boundary is left to be used for I/O at a given 950 * offset. 951 */ 952 static inline unsigned int blk_boundary_sectors_left(sector_t offset, 953 unsigned int boundary_sectors) 954 { 955 if (unlikely(!is_power_of_2(boundary_sectors))) 956 return boundary_sectors - sector_div(offset, boundary_sectors); 957 return boundary_sectors - (offset & (boundary_sectors - 1)); 958 } 959 960 /** 961 * queue_limits_start_update - start an atomic update of queue limits 962 * @q: queue to update 963 * 964 * This functions starts an atomic update of the queue limits. It takes a lock 965 * to prevent other updates and returns a snapshot of the current limits that 966 * the caller can modify. The caller must call queue_limits_commit_update() 967 * to finish the update. 968 * 969 * Context: process context. 970 */ 971 static inline struct queue_limits 972 queue_limits_start_update(struct request_queue *q) 973 { 974 mutex_lock(&q->limits_lock); 975 return q->limits; 976 } 977 int queue_limits_commit_update_frozen(struct request_queue *q, 978 struct queue_limits *lim); 979 int queue_limits_commit_update(struct request_queue *q, 980 struct queue_limits *lim); 981 int queue_limits_set(struct request_queue *q, struct queue_limits *lim); 982 int blk_validate_limits(struct queue_limits *lim); 983 984 /** 985 * queue_limits_cancel_update - cancel an atomic update of queue limits 986 * @q: queue to update 987 * 988 * This functions cancels an atomic update of the queue limits started by 989 * queue_limits_start_update() and should be used when an error occurs after 990 * starting update. 991 */ 992 static inline void queue_limits_cancel_update(struct request_queue *q) 993 { 994 mutex_unlock(&q->limits_lock); 995 } 996 997 /* 998 * These helpers are for drivers that have sloppy feature negotiation and might 999 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O 1000 * completion handler when the device returned an indicator that the respective 1001 * feature is not actually supported. They are racy and the driver needs to 1002 * cope with that. Try to avoid this scheme if you can. 1003 */ 1004 static inline void blk_queue_disable_discard(struct request_queue *q) 1005 { 1006 q->limits.max_discard_sectors = 0; 1007 } 1008 1009 static inline void blk_queue_disable_secure_erase(struct request_queue *q) 1010 { 1011 q->limits.max_secure_erase_sectors = 0; 1012 } 1013 1014 static inline void blk_queue_disable_write_zeroes(struct request_queue *q) 1015 { 1016 q->limits.max_write_zeroes_sectors = 0; 1017 } 1018 1019 /* 1020 * Access functions for manipulating queue properties 1021 */ 1022 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1023 extern void blk_set_stacking_limits(struct queue_limits *lim); 1024 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1025 sector_t offset); 1026 void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, 1027 sector_t offset, const char *pfx); 1028 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1029 1030 struct blk_independent_access_ranges * 1031 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges); 1032 void disk_set_independent_access_ranges(struct gendisk *disk, 1033 struct blk_independent_access_ranges *iars); 1034 1035 bool __must_check blk_get_queue(struct request_queue *); 1036 extern void blk_put_queue(struct request_queue *); 1037 1038 void blk_mark_disk_dead(struct gendisk *disk); 1039 1040 struct rq_list { 1041 struct request *head; 1042 struct request *tail; 1043 }; 1044 1045 #ifdef CONFIG_BLOCK 1046 /* 1047 * blk_plug permits building a queue of related requests by holding the I/O 1048 * fragments for a short period. This allows merging of sequential requests 1049 * into single larger request. As the requests are moved from a per-task list to 1050 * the device's request_queue in a batch, this results in improved scalability 1051 * as the lock contention for request_queue lock is reduced. 1052 * 1053 * It is ok not to disable preemption when adding the request to the plug list 1054 * or when attempting a merge. For details, please see schedule() where 1055 * blk_flush_plug() is called. 1056 */ 1057 struct blk_plug { 1058 struct rq_list mq_list; /* blk-mq requests */ 1059 1060 /* if ios_left is > 1, we can batch tag/rq allocations */ 1061 struct rq_list cached_rqs; 1062 u64 cur_ktime; 1063 unsigned short nr_ios; 1064 1065 unsigned short rq_count; 1066 1067 bool multiple_queues; 1068 bool has_elevator; 1069 1070 struct list_head cb_list; /* md requires an unplug callback */ 1071 }; 1072 1073 struct blk_plug_cb; 1074 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1075 struct blk_plug_cb { 1076 struct list_head list; 1077 blk_plug_cb_fn callback; 1078 void *data; 1079 }; 1080 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1081 void *data, int size); 1082 extern void blk_start_plug(struct blk_plug *); 1083 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short); 1084 extern void blk_finish_plug(struct blk_plug *); 1085 1086 void __blk_flush_plug(struct blk_plug *plug, bool from_schedule); 1087 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1088 { 1089 if (plug) 1090 __blk_flush_plug(plug, async); 1091 } 1092 1093 /* 1094 * tsk == current here 1095 */ 1096 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1097 { 1098 struct blk_plug *plug = tsk->plug; 1099 1100 if (plug) 1101 plug->cur_ktime = 0; 1102 current->flags &= ~PF_BLOCK_TS; 1103 } 1104 1105 int blkdev_issue_flush(struct block_device *bdev); 1106 long nr_blockdev_pages(void); 1107 #else /* CONFIG_BLOCK */ 1108 struct blk_plug { 1109 }; 1110 1111 static inline void blk_start_plug_nr_ios(struct blk_plug *plug, 1112 unsigned short nr_ios) 1113 { 1114 } 1115 1116 static inline void blk_start_plug(struct blk_plug *plug) 1117 { 1118 } 1119 1120 static inline void blk_finish_plug(struct blk_plug *plug) 1121 { 1122 } 1123 1124 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1125 { 1126 } 1127 1128 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1129 { 1130 } 1131 1132 static inline int blkdev_issue_flush(struct block_device *bdev) 1133 { 1134 return 0; 1135 } 1136 1137 static inline long nr_blockdev_pages(void) 1138 { 1139 return 0; 1140 } 1141 #endif /* CONFIG_BLOCK */ 1142 1143 extern void blk_io_schedule(void); 1144 1145 int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1146 sector_t nr_sects, gfp_t gfp_mask); 1147 int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1148 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop); 1149 int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, 1150 sector_t nr_sects, gfp_t gfp); 1151 1152 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1153 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1154 #define BLKDEV_ZERO_KILLABLE (1 << 2) /* interruptible by fatal signals */ 1155 1156 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1157 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1158 unsigned flags); 1159 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1160 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1161 1162 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1163 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1164 { 1165 return blkdev_issue_discard(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); 1171 } 1172 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1173 sector_t nr_blocks, gfp_t gfp_mask) 1174 { 1175 return blkdev_issue_zeroout(sb->s_bdev, 1176 block << (sb->s_blocksize_bits - 1177 SECTOR_SHIFT), 1178 nr_blocks << (sb->s_blocksize_bits - 1179 SECTOR_SHIFT), 1180 gfp_mask, 0); 1181 } 1182 1183 static inline bool bdev_is_partition(struct block_device *bdev) 1184 { 1185 return bdev_partno(bdev) != 0; 1186 } 1187 1188 enum blk_default_limits { 1189 BLK_MAX_SEGMENTS = 128, 1190 BLK_SAFE_MAX_SECTORS = 255, 1191 BLK_MAX_SEGMENT_SIZE = 65536, 1192 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1193 }; 1194 1195 /* 1196 * Default upper limit for the software max_sectors limit used for 1197 * regular file system I/O. This can be increased through sysfs. 1198 * 1199 * Not to be confused with the max_hw_sector limit that is entirely 1200 * controlled by the driver, usually based on hardware limits. 1201 */ 1202 #define BLK_DEF_MAX_SECTORS_CAP 2560u 1203 1204 static inline struct queue_limits *bdev_limits(struct block_device *bdev) 1205 { 1206 return &bdev_get_queue(bdev)->limits; 1207 } 1208 1209 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1210 { 1211 return q->limits.seg_boundary_mask; 1212 } 1213 1214 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1215 { 1216 return q->limits.virt_boundary_mask; 1217 } 1218 1219 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1220 { 1221 return q->limits.max_sectors; 1222 } 1223 1224 static inline unsigned int queue_max_bytes(struct request_queue *q) 1225 { 1226 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9; 1227 } 1228 1229 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1230 { 1231 return q->limits.max_hw_sectors; 1232 } 1233 1234 static inline unsigned short queue_max_segments(const struct request_queue *q) 1235 { 1236 return q->limits.max_segments; 1237 } 1238 1239 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1240 { 1241 return q->limits.max_discard_segments; 1242 } 1243 1244 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1245 { 1246 return q->limits.max_segment_size; 1247 } 1248 1249 static inline bool queue_emulates_zone_append(struct request_queue *q) 1250 { 1251 return blk_queue_is_zoned(q) && !q->limits.max_hw_zone_append_sectors; 1252 } 1253 1254 static inline bool bdev_emulates_zone_append(struct block_device *bdev) 1255 { 1256 return queue_emulates_zone_append(bdev_get_queue(bdev)); 1257 } 1258 1259 static inline unsigned int 1260 bdev_max_zone_append_sectors(struct block_device *bdev) 1261 { 1262 return bdev_limits(bdev)->max_zone_append_sectors; 1263 } 1264 1265 static inline unsigned int bdev_max_segments(struct block_device *bdev) 1266 { 1267 return queue_max_segments(bdev_get_queue(bdev)); 1268 } 1269 1270 static inline unsigned queue_logical_block_size(const struct request_queue *q) 1271 { 1272 return q->limits.logical_block_size; 1273 } 1274 1275 static inline unsigned int bdev_logical_block_size(struct block_device *bdev) 1276 { 1277 return queue_logical_block_size(bdev_get_queue(bdev)); 1278 } 1279 1280 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1281 { 1282 return q->limits.physical_block_size; 1283 } 1284 1285 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1286 { 1287 return queue_physical_block_size(bdev_get_queue(bdev)); 1288 } 1289 1290 static inline unsigned int queue_io_min(const struct request_queue *q) 1291 { 1292 return q->limits.io_min; 1293 } 1294 1295 static inline unsigned int bdev_io_min(struct block_device *bdev) 1296 { 1297 return queue_io_min(bdev_get_queue(bdev)); 1298 } 1299 1300 static inline unsigned int queue_io_opt(const struct request_queue *q) 1301 { 1302 return q->limits.io_opt; 1303 } 1304 1305 static inline unsigned int bdev_io_opt(struct block_device *bdev) 1306 { 1307 return queue_io_opt(bdev_get_queue(bdev)); 1308 } 1309 1310 static inline unsigned int 1311 queue_zone_write_granularity(const struct request_queue *q) 1312 { 1313 return q->limits.zone_write_granularity; 1314 } 1315 1316 static inline unsigned int 1317 bdev_zone_write_granularity(struct block_device *bdev) 1318 { 1319 return queue_zone_write_granularity(bdev_get_queue(bdev)); 1320 } 1321 1322 int bdev_alignment_offset(struct block_device *bdev); 1323 unsigned int bdev_discard_alignment(struct block_device *bdev); 1324 1325 static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev) 1326 { 1327 return bdev_limits(bdev)->max_discard_sectors; 1328 } 1329 1330 static inline unsigned int bdev_discard_granularity(struct block_device *bdev) 1331 { 1332 return bdev_limits(bdev)->discard_granularity; 1333 } 1334 1335 static inline unsigned int 1336 bdev_max_secure_erase_sectors(struct block_device *bdev) 1337 { 1338 return bdev_limits(bdev)->max_secure_erase_sectors; 1339 } 1340 1341 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1342 { 1343 return bdev_limits(bdev)->max_write_zeroes_sectors; 1344 } 1345 1346 static inline bool bdev_nonrot(struct block_device *bdev) 1347 { 1348 return blk_queue_nonrot(bdev_get_queue(bdev)); 1349 } 1350 1351 static inline bool bdev_synchronous(struct block_device *bdev) 1352 { 1353 return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS; 1354 } 1355 1356 static inline bool bdev_stable_writes(struct block_device *bdev) 1357 { 1358 struct request_queue *q = bdev_get_queue(bdev); 1359 1360 if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && 1361 q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) 1362 return true; 1363 return q->limits.features & BLK_FEAT_STABLE_WRITES; 1364 } 1365 1366 static inline bool blk_queue_write_cache(struct request_queue *q) 1367 { 1368 return (q->limits.features & BLK_FEAT_WRITE_CACHE) && 1369 !(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED); 1370 } 1371 1372 static inline bool bdev_write_cache(struct block_device *bdev) 1373 { 1374 return blk_queue_write_cache(bdev_get_queue(bdev)); 1375 } 1376 1377 static inline bool bdev_fua(struct block_device *bdev) 1378 { 1379 return bdev_limits(bdev)->features & BLK_FEAT_FUA; 1380 } 1381 1382 static inline bool bdev_nowait(struct block_device *bdev) 1383 { 1384 return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT; 1385 } 1386 1387 static inline bool bdev_is_zoned(struct block_device *bdev) 1388 { 1389 return blk_queue_is_zoned(bdev_get_queue(bdev)); 1390 } 1391 1392 static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec) 1393 { 1394 return disk_zone_no(bdev->bd_disk, sec); 1395 } 1396 1397 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1398 { 1399 struct request_queue *q = bdev_get_queue(bdev); 1400 1401 if (!blk_queue_is_zoned(q)) 1402 return 0; 1403 return q->limits.chunk_sectors; 1404 } 1405 1406 static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev, 1407 sector_t sector) 1408 { 1409 return sector & (bdev_zone_sectors(bdev) - 1); 1410 } 1411 1412 static inline sector_t bio_offset_from_zone_start(struct bio *bio) 1413 { 1414 return bdev_offset_from_zone_start(bio->bi_bdev, 1415 bio->bi_iter.bi_sector); 1416 } 1417 1418 static inline bool bdev_is_zone_start(struct block_device *bdev, 1419 sector_t sector) 1420 { 1421 return bdev_offset_from_zone_start(bdev, sector) == 0; 1422 } 1423 1424 /** 1425 * bdev_zone_is_seq - check if a sector belongs to a sequential write zone 1426 * @bdev: block device to check 1427 * @sector: sector number 1428 * 1429 * Check if @sector on @bdev is contained in a sequential write required zone. 1430 */ 1431 static inline bool bdev_zone_is_seq(struct block_device *bdev, sector_t sector) 1432 { 1433 bool is_seq = false; 1434 1435 #if IS_ENABLED(CONFIG_BLK_DEV_ZONED) 1436 if (bdev_is_zoned(bdev)) { 1437 struct gendisk *disk = bdev->bd_disk; 1438 unsigned long *bitmap; 1439 1440 rcu_read_lock(); 1441 bitmap = rcu_dereference(disk->conv_zones_bitmap); 1442 is_seq = !bitmap || 1443 !test_bit(disk_zone_no(disk, sector), bitmap); 1444 rcu_read_unlock(); 1445 } 1446 #endif 1447 1448 return is_seq; 1449 } 1450 1451 int blk_zone_issue_zeroout(struct block_device *bdev, sector_t sector, 1452 sector_t nr_sects, gfp_t gfp_mask); 1453 1454 static inline unsigned int queue_dma_alignment(const struct request_queue *q) 1455 { 1456 return q->limits.dma_alignment; 1457 } 1458 1459 static inline unsigned int 1460 queue_atomic_write_unit_max_bytes(const struct request_queue *q) 1461 { 1462 return q->limits.atomic_write_unit_max; 1463 } 1464 1465 static inline unsigned int 1466 queue_atomic_write_unit_min_bytes(const struct request_queue *q) 1467 { 1468 return q->limits.atomic_write_unit_min; 1469 } 1470 1471 static inline unsigned int 1472 queue_atomic_write_boundary_bytes(const struct request_queue *q) 1473 { 1474 return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT; 1475 } 1476 1477 static inline unsigned int 1478 queue_atomic_write_max_bytes(const struct request_queue *q) 1479 { 1480 return q->limits.atomic_write_max_sectors << SECTOR_SHIFT; 1481 } 1482 1483 static inline unsigned int bdev_dma_alignment(struct block_device *bdev) 1484 { 1485 return queue_dma_alignment(bdev_get_queue(bdev)); 1486 } 1487 1488 static inline bool bdev_iter_is_aligned(struct block_device *bdev, 1489 struct iov_iter *iter) 1490 { 1491 return iov_iter_is_aligned(iter, bdev_dma_alignment(bdev), 1492 bdev_logical_block_size(bdev) - 1); 1493 } 1494 1495 static inline unsigned int 1496 blk_lim_dma_alignment_and_pad(struct queue_limits *lim) 1497 { 1498 return lim->dma_alignment | lim->dma_pad_mask; 1499 } 1500 1501 static inline bool blk_rq_aligned(struct request_queue *q, unsigned long addr, 1502 unsigned int len) 1503 { 1504 unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits); 1505 1506 return !(addr & alignment) && !(len & alignment); 1507 } 1508 1509 /* assumes size > 256 */ 1510 static inline unsigned int blksize_bits(unsigned int size) 1511 { 1512 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT; 1513 } 1514 1515 int kblockd_schedule_work(struct work_struct *work); 1516 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1517 1518 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1519 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1520 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1521 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1522 1523 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 1524 1525 bool blk_crypto_register(struct blk_crypto_profile *profile, 1526 struct request_queue *q); 1527 1528 #else /* CONFIG_BLK_INLINE_ENCRYPTION */ 1529 1530 static inline bool blk_crypto_register(struct blk_crypto_profile *profile, 1531 struct request_queue *q) 1532 { 1533 return true; 1534 } 1535 1536 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */ 1537 1538 enum blk_unique_id { 1539 /* these match the Designator Types specified in SPC */ 1540 BLK_UID_T10 = 1, 1541 BLK_UID_EUI64 = 2, 1542 BLK_UID_NAA = 3, 1543 }; 1544 1545 struct block_device_operations { 1546 void (*submit_bio)(struct bio *bio); 1547 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob, 1548 unsigned int flags); 1549 int (*open)(struct gendisk *disk, blk_mode_t mode); 1550 void (*release)(struct gendisk *disk); 1551 int (*ioctl)(struct block_device *bdev, blk_mode_t mode, 1552 unsigned cmd, unsigned long arg); 1553 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode, 1554 unsigned cmd, unsigned long arg); 1555 unsigned int (*check_events) (struct gendisk *disk, 1556 unsigned int clearing); 1557 void (*unlock_native_capacity) (struct gendisk *); 1558 int (*getgeo)(struct block_device *, struct hd_geometry *); 1559 int (*set_read_only)(struct block_device *bdev, bool ro); 1560 void (*free_disk)(struct gendisk *disk); 1561 /* this callback is with swap_lock and sometimes page table lock held */ 1562 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1563 int (*report_zones)(struct gendisk *, sector_t sector, 1564 unsigned int nr_zones, report_zones_cb cb, void *data); 1565 char *(*devnode)(struct gendisk *disk, umode_t *mode); 1566 /* returns the length of the identifier or a negative errno: */ 1567 int (*get_unique_id)(struct gendisk *disk, u8 id[16], 1568 enum blk_unique_id id_type); 1569 struct module *owner; 1570 const struct pr_ops *pr_ops; 1571 1572 /* 1573 * Special callback for probing GPT entry at a given sector. 1574 * Needed by Android devices, used by GPT scanner and MMC blk 1575 * driver. 1576 */ 1577 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector); 1578 }; 1579 1580 #ifdef CONFIG_COMPAT 1581 extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t, 1582 unsigned int, unsigned long); 1583 #else 1584 #define blkdev_compat_ptr_ioctl NULL 1585 #endif 1586 1587 static inline void blk_wake_io_task(struct task_struct *waiter) 1588 { 1589 /* 1590 * If we're polling, the task itself is doing the completions. For 1591 * that case, we don't need to signal a wakeup, it's enough to just 1592 * mark us as RUNNING. 1593 */ 1594 if (waiter == current) 1595 __set_current_state(TASK_RUNNING); 1596 else 1597 wake_up_process(waiter); 1598 } 1599 1600 unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, 1601 unsigned long start_time); 1602 void bdev_end_io_acct(struct block_device *bdev, enum req_op op, 1603 unsigned int sectors, unsigned long start_time); 1604 1605 unsigned long bio_start_io_acct(struct bio *bio); 1606 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, 1607 struct block_device *orig_bdev); 1608 1609 /** 1610 * bio_end_io_acct - end I/O accounting for bio based drivers 1611 * @bio: bio to end account for 1612 * @start_time: start time returned by bio_start_io_acct() 1613 */ 1614 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time) 1615 { 1616 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev); 1617 } 1618 1619 int set_blocksize(struct file *file, int size); 1620 1621 int lookup_bdev(const char *pathname, dev_t *dev); 1622 1623 void blkdev_show(struct seq_file *seqf, off_t offset); 1624 1625 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */ 1626 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */ 1627 #ifdef CONFIG_BLOCK 1628 #define BLKDEV_MAJOR_MAX 512 1629 #else 1630 #define BLKDEV_MAJOR_MAX 0 1631 #endif 1632 1633 struct blk_holder_ops { 1634 void (*mark_dead)(struct block_device *bdev, bool surprise); 1635 1636 /* 1637 * Sync the file system mounted on the block device. 1638 */ 1639 void (*sync)(struct block_device *bdev); 1640 1641 /* 1642 * Freeze the file system mounted on the block device. 1643 */ 1644 int (*freeze)(struct block_device *bdev); 1645 1646 /* 1647 * Thaw the file system mounted on the block device. 1648 */ 1649 int (*thaw)(struct block_device *bdev); 1650 }; 1651 1652 /* 1653 * For filesystems using @fs_holder_ops, the @holder argument passed to 1654 * helpers used to open and claim block devices via 1655 * bd_prepare_to_claim() must point to a superblock. 1656 */ 1657 extern const struct blk_holder_ops fs_holder_ops; 1658 1659 /* 1660 * Return the correct open flags for blkdev_get_by_* for super block flags 1661 * as stored in sb->s_flags. 1662 */ 1663 #define sb_open_mode(flags) \ 1664 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \ 1665 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE)) 1666 1667 struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder, 1668 const struct blk_holder_ops *hops); 1669 struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode, 1670 void *holder, const struct blk_holder_ops *hops); 1671 int bd_prepare_to_claim(struct block_device *bdev, void *holder, 1672 const struct blk_holder_ops *hops); 1673 void bd_abort_claiming(struct block_device *bdev, void *holder); 1674 1675 /* just for blk-cgroup, don't use elsewhere */ 1676 struct block_device *blkdev_get_no_open(dev_t dev); 1677 void blkdev_put_no_open(struct block_device *bdev); 1678 1679 struct block_device *I_BDEV(struct inode *inode); 1680 struct block_device *file_bdev(struct file *bdev_file); 1681 bool disk_live(struct gendisk *disk); 1682 unsigned int block_size(struct block_device *bdev); 1683 1684 #ifdef CONFIG_BLOCK 1685 void invalidate_bdev(struct block_device *bdev); 1686 int sync_blockdev(struct block_device *bdev); 1687 int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend); 1688 int sync_blockdev_nowait(struct block_device *bdev); 1689 void sync_bdevs(bool wait); 1690 void bdev_statx(struct path *, struct kstat *, u32); 1691 void printk_all_partitions(void); 1692 int __init early_lookup_bdev(const char *pathname, dev_t *dev); 1693 #else 1694 static inline void invalidate_bdev(struct block_device *bdev) 1695 { 1696 } 1697 static inline int sync_blockdev(struct block_device *bdev) 1698 { 1699 return 0; 1700 } 1701 static inline int sync_blockdev_nowait(struct block_device *bdev) 1702 { 1703 return 0; 1704 } 1705 static inline void sync_bdevs(bool wait) 1706 { 1707 } 1708 static inline void bdev_statx(struct path *path, struct kstat *stat, 1709 u32 request_mask) 1710 { 1711 } 1712 static inline void printk_all_partitions(void) 1713 { 1714 } 1715 static inline int early_lookup_bdev(const char *pathname, dev_t *dev) 1716 { 1717 return -EINVAL; 1718 } 1719 #endif /* CONFIG_BLOCK */ 1720 1721 int bdev_freeze(struct block_device *bdev); 1722 int bdev_thaw(struct block_device *bdev); 1723 void bdev_fput(struct file *bdev_file); 1724 1725 struct io_comp_batch { 1726 struct rq_list req_list; 1727 bool need_ts; 1728 void (*complete)(struct io_comp_batch *); 1729 }; 1730 1731 static inline bool blk_atomic_write_start_sect_aligned(sector_t sector, 1732 struct queue_limits *limits) 1733 { 1734 unsigned int alignment = max(limits->atomic_write_hw_unit_min, 1735 limits->atomic_write_hw_boundary); 1736 1737 return IS_ALIGNED(sector, alignment >> SECTOR_SHIFT); 1738 } 1739 1740 static inline bool bdev_can_atomic_write(struct block_device *bdev) 1741 { 1742 struct request_queue *bd_queue = bdev->bd_queue; 1743 struct queue_limits *limits = &bd_queue->limits; 1744 1745 if (!limits->atomic_write_unit_min) 1746 return false; 1747 1748 if (bdev_is_partition(bdev)) 1749 return blk_atomic_write_start_sect_aligned(bdev->bd_start_sect, 1750 limits); 1751 1752 return true; 1753 } 1754 1755 static inline unsigned int 1756 bdev_atomic_write_unit_min_bytes(struct block_device *bdev) 1757 { 1758 if (!bdev_can_atomic_write(bdev)) 1759 return 0; 1760 return queue_atomic_write_unit_min_bytes(bdev_get_queue(bdev)); 1761 } 1762 1763 static inline unsigned int 1764 bdev_atomic_write_unit_max_bytes(struct block_device *bdev) 1765 { 1766 if (!bdev_can_atomic_write(bdev)) 1767 return 0; 1768 return queue_atomic_write_unit_max_bytes(bdev_get_queue(bdev)); 1769 } 1770 1771 #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { } 1772 1773 #endif /* _LINUX_BLKDEV_H */ 1774