1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* 3 * Portions Copyright (C) 1992 Drew Eckhardt 4 */ 5 #ifndef _LINUX_BLKDEV_H 6 #define _LINUX_BLKDEV_H 7 8 #include <linux/types.h> 9 #include <linux/blk_types.h> 10 #include <linux/device.h> 11 #include <linux/list.h> 12 #include <linux/llist.h> 13 #include <linux/minmax.h> 14 #include <linux/timer.h> 15 #include <linux/workqueue.h> 16 #include <linux/wait.h> 17 #include <linux/bio.h> 18 #include <linux/gfp.h> 19 #include <linux/kdev_t.h> 20 #include <linux/rcupdate.h> 21 #include <linux/percpu-refcount.h> 22 #include <linux/blkzoned.h> 23 #include <linux/sched.h> 24 #include <linux/sbitmap.h> 25 #include <linux/uuid.h> 26 #include <linux/xarray.h> 27 #include <linux/file.h> 28 #include <linux/lockdep.h> 29 30 struct module; 31 struct request_queue; 32 struct elevator_queue; 33 struct blk_trace; 34 struct request; 35 struct sg_io_hdr; 36 struct blkcg_gq; 37 struct blk_flush_queue; 38 struct kiocb; 39 struct pr_ops; 40 struct rq_qos; 41 struct blk_queue_stats; 42 struct blk_stat_callback; 43 struct blk_crypto_profile; 44 45 extern const struct device_type disk_type; 46 extern const struct device_type part_type; 47 extern const struct class block_class; 48 49 /* 50 * Maximum number of blkcg policies allowed to be registered concurrently. 51 * Defined here to simplify include dependency. 52 */ 53 #define BLKCG_MAX_POLS 6 54 55 #define DISK_MAX_PARTS 256 56 #define DISK_NAME_LEN 32 57 58 #define PARTITION_META_INFO_VOLNAMELTH 64 59 /* 60 * Enough for the string representation of any kind of UUID plus NULL. 61 * EFI UUID is 36 characters. MSDOS UUID is 11 characters. 62 */ 63 #define PARTITION_META_INFO_UUIDLTH (UUID_STRING_LEN + 1) 64 65 struct partition_meta_info { 66 char uuid[PARTITION_META_INFO_UUIDLTH]; 67 u8 volname[PARTITION_META_INFO_VOLNAMELTH]; 68 }; 69 70 /** 71 * DOC: genhd capability flags 72 * 73 * ``GENHD_FL_REMOVABLE``: indicates that the block device gives access to 74 * removable media. When set, the device remains present even when media is not 75 * inserted. Shall not be set for devices which are removed entirely when the 76 * media is removed. 77 * 78 * ``GENHD_FL_HIDDEN``: the block device is hidden; it doesn't produce events, 79 * doesn't appear in sysfs, and can't be opened from userspace or using 80 * blkdev_get*. Used for the underlying components of multipath devices. 81 * 82 * ``GENHD_FL_NO_PART``: partition support is disabled. The kernel will not 83 * scan for partitions from add_disk, and users can't add partitions manually. 84 * 85 */ 86 enum { 87 GENHD_FL_REMOVABLE = 1 << 0, 88 GENHD_FL_HIDDEN = 1 << 1, 89 GENHD_FL_NO_PART = 1 << 2, 90 }; 91 92 enum { 93 DISK_EVENT_MEDIA_CHANGE = 1 << 0, /* media changed */ 94 DISK_EVENT_EJECT_REQUEST = 1 << 1, /* eject requested */ 95 }; 96 97 enum { 98 /* Poll even if events_poll_msecs is unset */ 99 DISK_EVENT_FLAG_POLL = 1 << 0, 100 /* Forward events to udev */ 101 DISK_EVENT_FLAG_UEVENT = 1 << 1, 102 /* Block event polling when open for exclusive write */ 103 DISK_EVENT_FLAG_BLOCK_ON_EXCL_WRITE = 1 << 2, 104 }; 105 106 struct disk_events; 107 struct badblocks; 108 109 enum blk_integrity_checksum { 110 BLK_INTEGRITY_CSUM_NONE = 0, 111 BLK_INTEGRITY_CSUM_IP = 1, 112 BLK_INTEGRITY_CSUM_CRC = 2, 113 BLK_INTEGRITY_CSUM_CRC64 = 3, 114 } __packed ; 115 116 struct blk_integrity { 117 unsigned char flags; 118 enum blk_integrity_checksum csum_type; 119 unsigned char tuple_size; 120 unsigned char pi_offset; 121 unsigned char interval_exp; 122 unsigned char tag_size; 123 }; 124 125 typedef unsigned int __bitwise blk_mode_t; 126 127 /* open for reading */ 128 #define BLK_OPEN_READ ((__force blk_mode_t)(1 << 0)) 129 /* open for writing */ 130 #define BLK_OPEN_WRITE ((__force blk_mode_t)(1 << 1)) 131 /* open exclusively (vs other exclusive openers */ 132 #define BLK_OPEN_EXCL ((__force blk_mode_t)(1 << 2)) 133 /* opened with O_NDELAY */ 134 #define BLK_OPEN_NDELAY ((__force blk_mode_t)(1 << 3)) 135 /* open for "writes" only for ioctls (specialy hack for floppy.c) */ 136 #define BLK_OPEN_WRITE_IOCTL ((__force blk_mode_t)(1 << 4)) 137 /* open is exclusive wrt all other BLK_OPEN_WRITE opens to the device */ 138 #define BLK_OPEN_RESTRICT_WRITES ((__force blk_mode_t)(1 << 5)) 139 /* return partition scanning errors */ 140 #define BLK_OPEN_STRICT_SCAN ((__force blk_mode_t)(1 << 6)) 141 142 struct gendisk { 143 /* 144 * major/first_minor/minors should not be set by any new driver, the 145 * block core will take care of allocating them automatically. 146 */ 147 int major; 148 int first_minor; 149 int minors; 150 151 char disk_name[DISK_NAME_LEN]; /* name of major driver */ 152 153 unsigned short events; /* supported events */ 154 unsigned short event_flags; /* flags related to event processing */ 155 156 struct xarray part_tbl; 157 struct block_device *part0; 158 159 const struct block_device_operations *fops; 160 struct request_queue *queue; 161 void *private_data; 162 163 struct bio_set bio_split; 164 165 int flags; 166 unsigned long state; 167 #define GD_NEED_PART_SCAN 0 168 #define GD_READ_ONLY 1 169 #define GD_DEAD 2 170 #define GD_NATIVE_CAPACITY 3 171 #define GD_ADDED 4 172 #define GD_SUPPRESS_PART_SCAN 5 173 #define GD_OWNS_QUEUE 6 174 175 struct mutex open_mutex; /* open/close mutex */ 176 unsigned open_partitions; /* number of open partitions */ 177 178 struct backing_dev_info *bdi; 179 struct kobject queue_kobj; /* the queue/ directory */ 180 struct kobject *slave_dir; 181 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 182 struct list_head slave_bdevs; 183 #endif 184 struct timer_rand_state *random; 185 atomic_t sync_io; /* RAID */ 186 struct disk_events *ev; 187 188 #ifdef CONFIG_BLK_DEV_ZONED 189 /* 190 * Zoned block device information. Reads of this information must be 191 * protected with blk_queue_enter() / blk_queue_exit(). Modifying this 192 * information is only allowed while no requests are being processed. 193 * See also blk_mq_freeze_queue() and blk_mq_unfreeze_queue(). 194 */ 195 unsigned int nr_zones; 196 unsigned int zone_capacity; 197 unsigned int last_zone_capacity; 198 unsigned long __rcu *conv_zones_bitmap; 199 unsigned int zone_wplugs_hash_bits; 200 spinlock_t zone_wplugs_lock; 201 struct mempool_s *zone_wplugs_pool; 202 struct hlist_head *zone_wplugs_hash; 203 struct workqueue_struct *zone_wplugs_wq; 204 #endif /* CONFIG_BLK_DEV_ZONED */ 205 206 #if IS_ENABLED(CONFIG_CDROM) 207 struct cdrom_device_info *cdi; 208 #endif 209 int node_id; 210 struct badblocks *bb; 211 struct lockdep_map lockdep_map; 212 u64 diskseq; 213 blk_mode_t open_mode; 214 215 /* 216 * Independent sector access ranges. This is always NULL for 217 * devices that do not have multiple independent access ranges. 218 */ 219 struct blk_independent_access_ranges *ia_ranges; 220 }; 221 222 /** 223 * disk_openers - returns how many openers are there for a disk 224 * @disk: disk to check 225 * 226 * This returns the number of openers for a disk. Note that this value is only 227 * stable if disk->open_mutex is held. 228 * 229 * Note: Due to a quirk in the block layer open code, each open partition is 230 * only counted once even if there are multiple openers. 231 */ 232 static inline unsigned int disk_openers(struct gendisk *disk) 233 { 234 return atomic_read(&disk->part0->bd_openers); 235 } 236 237 /** 238 * disk_has_partscan - return %true if partition scanning is enabled on a disk 239 * @disk: disk to check 240 * 241 * Returns %true if partitions scanning is enabled for @disk, or %false if 242 * partition scanning is disabled either permanently or temporarily. 243 */ 244 static inline bool disk_has_partscan(struct gendisk *disk) 245 { 246 return !(disk->flags & (GENHD_FL_NO_PART | GENHD_FL_HIDDEN)) && 247 !test_bit(GD_SUPPRESS_PART_SCAN, &disk->state); 248 } 249 250 /* 251 * The gendisk is refcounted by the part0 block_device, and the bd_device 252 * therein is also used for device model presentation in sysfs. 253 */ 254 #define dev_to_disk(device) \ 255 (dev_to_bdev(device)->bd_disk) 256 #define disk_to_dev(disk) \ 257 (&((disk)->part0->bd_device)) 258 259 #if IS_REACHABLE(CONFIG_CDROM) 260 #define disk_to_cdi(disk) ((disk)->cdi) 261 #else 262 #define disk_to_cdi(disk) NULL 263 #endif 264 265 static inline dev_t disk_devt(struct gendisk *disk) 266 { 267 return MKDEV(disk->major, disk->first_minor); 268 } 269 270 /* blk_validate_limits() validates bsize, so drivers don't usually need to */ 271 static inline int blk_validate_block_size(unsigned long bsize) 272 { 273 if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(bsize)) 274 return -EINVAL; 275 276 return 0; 277 } 278 279 static inline bool blk_op_is_passthrough(blk_opf_t op) 280 { 281 op &= REQ_OP_MASK; 282 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 283 } 284 285 /* flags set by the driver in queue_limits.features */ 286 typedef unsigned int __bitwise blk_features_t; 287 288 /* supports a volatile write cache */ 289 #define BLK_FEAT_WRITE_CACHE ((__force blk_features_t)(1u << 0)) 290 291 /* supports passing on the FUA bit */ 292 #define BLK_FEAT_FUA ((__force blk_features_t)(1u << 1)) 293 294 /* rotational device (hard drive or floppy) */ 295 #define BLK_FEAT_ROTATIONAL ((__force blk_features_t)(1u << 2)) 296 297 /* contributes to the random number pool */ 298 #define BLK_FEAT_ADD_RANDOM ((__force blk_features_t)(1u << 3)) 299 300 /* do disk/partitions IO accounting */ 301 #define BLK_FEAT_IO_STAT ((__force blk_features_t)(1u << 4)) 302 303 /* don't modify data until writeback is done */ 304 #define BLK_FEAT_STABLE_WRITES ((__force blk_features_t)(1u << 5)) 305 306 /* always completes in submit context */ 307 #define BLK_FEAT_SYNCHRONOUS ((__force blk_features_t)(1u << 6)) 308 309 /* supports REQ_NOWAIT */ 310 #define BLK_FEAT_NOWAIT ((__force blk_features_t)(1u << 7)) 311 312 /* supports DAX */ 313 #define BLK_FEAT_DAX ((__force blk_features_t)(1u << 8)) 314 315 /* supports I/O polling */ 316 #define BLK_FEAT_POLL ((__force blk_features_t)(1u << 9)) 317 318 /* is a zoned device */ 319 #define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10)) 320 321 /* supports PCI(e) p2p requests */ 322 #define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12)) 323 324 /* skip this queue in blk_mq_(un)quiesce_tagset */ 325 #define BLK_FEAT_SKIP_TAGSET_QUIESCE ((__force blk_features_t)(1u << 13)) 326 327 /* bounce all highmem pages */ 328 #define BLK_FEAT_BOUNCE_HIGH ((__force blk_features_t)(1u << 14)) 329 330 /* undocumented magic for bcache */ 331 #define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \ 332 ((__force blk_features_t)(1u << 15)) 333 334 /* atomic writes enabled */ 335 #define BLK_FEAT_ATOMIC_WRITES \ 336 ((__force blk_features_t)(1u << 16)) 337 338 /* 339 * Flags automatically inherited when stacking limits. 340 */ 341 #define BLK_FEAT_INHERIT_MASK \ 342 (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ 343 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH | \ 344 BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE) 345 346 /* internal flags in queue_limits.flags */ 347 typedef unsigned int __bitwise blk_flags_t; 348 349 /* do not send FLUSH/FUA commands despite advertising a write cache */ 350 #define BLK_FLAG_WRITE_CACHE_DISABLED ((__force blk_flags_t)(1u << 0)) 351 352 /* I/O topology is misaligned */ 353 #define BLK_FLAG_MISALIGNED ((__force blk_flags_t)(1u << 1)) 354 355 /* passthrough command IO accounting */ 356 #define BLK_FLAG_IOSTATS_PASSTHROUGH ((__force blk_flags_t)(1u << 2)) 357 358 struct queue_limits { 359 blk_features_t features; 360 blk_flags_t flags; 361 unsigned long seg_boundary_mask; 362 unsigned long virt_boundary_mask; 363 364 unsigned int max_hw_sectors; 365 unsigned int max_dev_sectors; 366 unsigned int chunk_sectors; 367 unsigned int max_sectors; 368 unsigned int max_user_sectors; 369 unsigned int max_segment_size; 370 unsigned int physical_block_size; 371 unsigned int logical_block_size; 372 unsigned int alignment_offset; 373 unsigned int io_min; 374 unsigned int io_opt; 375 unsigned int max_discard_sectors; 376 unsigned int max_hw_discard_sectors; 377 unsigned int max_user_discard_sectors; 378 unsigned int max_secure_erase_sectors; 379 unsigned int max_write_zeroes_sectors; 380 unsigned int max_hw_zone_append_sectors; 381 unsigned int max_zone_append_sectors; 382 unsigned int discard_granularity; 383 unsigned int discard_alignment; 384 unsigned int zone_write_granularity; 385 386 /* atomic write limits */ 387 unsigned int atomic_write_hw_max; 388 unsigned int atomic_write_max_sectors; 389 unsigned int atomic_write_hw_boundary; 390 unsigned int atomic_write_boundary_sectors; 391 unsigned int atomic_write_hw_unit_min; 392 unsigned int atomic_write_unit_min; 393 unsigned int atomic_write_hw_unit_max; 394 unsigned int atomic_write_unit_max; 395 396 unsigned short max_segments; 397 unsigned short max_integrity_segments; 398 unsigned short max_discard_segments; 399 400 unsigned int max_open_zones; 401 unsigned int max_active_zones; 402 403 /* 404 * Drivers that set dma_alignment to less than 511 must be prepared to 405 * handle individual bvec's that are not a multiple of a SECTOR_SIZE 406 * due to possible offsets. 407 */ 408 unsigned int dma_alignment; 409 unsigned int dma_pad_mask; 410 411 struct blk_integrity integrity; 412 }; 413 414 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, 415 void *data); 416 417 #define BLK_ALL_ZONES ((unsigned int)-1) 418 int blkdev_report_zones(struct block_device *bdev, sector_t sector, 419 unsigned int nr_zones, report_zones_cb cb, void *data); 420 int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, 421 sector_t sectors, sector_t nr_sectors); 422 int blk_revalidate_disk_zones(struct gendisk *disk); 423 424 /* 425 * Independent access ranges: struct blk_independent_access_range describes 426 * a range of contiguous sectors that can be accessed using device command 427 * execution resources that are independent from the resources used for 428 * other access ranges. This is typically found with single-LUN multi-actuator 429 * HDDs where each access range is served by a different set of heads. 430 * The set of independent ranges supported by the device is defined using 431 * struct blk_independent_access_ranges. The independent ranges must not overlap 432 * and must include all sectors within the disk capacity (no sector holes 433 * allowed). 434 * For a device with multiple ranges, requests targeting sectors in different 435 * ranges can be executed in parallel. A request can straddle an access range 436 * boundary. 437 */ 438 struct blk_independent_access_range { 439 struct kobject kobj; 440 sector_t sector; 441 sector_t nr_sectors; 442 }; 443 444 struct blk_independent_access_ranges { 445 struct kobject kobj; 446 bool sysfs_registered; 447 unsigned int nr_ia_ranges; 448 struct blk_independent_access_range ia_range[]; 449 }; 450 451 struct request_queue { 452 /* 453 * The queue owner gets to use this for whatever they like. 454 * ll_rw_blk doesn't touch it. 455 */ 456 void *queuedata; 457 458 struct elevator_queue *elevator; 459 460 const struct blk_mq_ops *mq_ops; 461 462 /* sw queues */ 463 struct blk_mq_ctx __percpu *queue_ctx; 464 465 /* 466 * various queue flags, see QUEUE_* below 467 */ 468 unsigned long queue_flags; 469 470 unsigned int rq_timeout; 471 472 unsigned int queue_depth; 473 474 refcount_t refs; 475 476 /* hw dispatch queues */ 477 unsigned int nr_hw_queues; 478 struct xarray hctx_table; 479 480 struct percpu_ref q_usage_counter; 481 struct lock_class_key io_lock_cls_key; 482 struct lockdep_map io_lockdep_map; 483 484 struct lock_class_key q_lock_cls_key; 485 struct lockdep_map q_lockdep_map; 486 487 struct request *last_merge; 488 489 spinlock_t queue_lock; 490 491 int quiesce_depth; 492 493 struct gendisk *disk; 494 495 /* 496 * mq queue kobject 497 */ 498 struct kobject *mq_kobj; 499 500 struct queue_limits limits; 501 502 #ifdef CONFIG_PM 503 struct device *dev; 504 enum rpm_status rpm_status; 505 #endif 506 507 /* 508 * Number of contexts that have called blk_set_pm_only(). If this 509 * counter is above zero then only RQF_PM requests are processed. 510 */ 511 atomic_t pm_only; 512 513 struct blk_queue_stats *stats; 514 struct rq_qos *rq_qos; 515 struct mutex rq_qos_mutex; 516 517 /* 518 * ida allocated id for this queue. Used to index queues from 519 * ioctx. 520 */ 521 int id; 522 523 /* 524 * queue settings 525 */ 526 unsigned long nr_requests; /* Max # of requests */ 527 528 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 529 struct blk_crypto_profile *crypto_profile; 530 struct kobject *crypto_kobject; 531 #endif 532 533 struct timer_list timeout; 534 struct work_struct timeout_work; 535 536 atomic_t nr_active_requests_shared_tags; 537 538 struct blk_mq_tags *sched_shared_tags; 539 540 struct list_head icq_list; 541 #ifdef CONFIG_BLK_CGROUP 542 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 543 struct blkcg_gq *root_blkg; 544 struct list_head blkg_list; 545 struct mutex blkcg_mutex; 546 #endif 547 548 int node; 549 550 spinlock_t requeue_lock; 551 struct list_head requeue_list; 552 struct delayed_work requeue_work; 553 554 #ifdef CONFIG_BLK_DEV_IO_TRACE 555 struct blk_trace __rcu *blk_trace; 556 #endif 557 /* 558 * for flush operations 559 */ 560 struct blk_flush_queue *fq; 561 struct list_head flush_list; 562 563 struct mutex sysfs_lock; 564 struct mutex sysfs_dir_lock; 565 struct mutex limits_lock; 566 567 /* 568 * for reusing dead hctx instance in case of updating 569 * nr_hw_queues 570 */ 571 struct list_head unused_hctx_list; 572 spinlock_t unused_hctx_lock; 573 574 int mq_freeze_depth; 575 576 #ifdef CONFIG_BLK_DEV_THROTTLING 577 /* Throttle data */ 578 struct throtl_data *td; 579 #endif 580 struct rcu_head rcu_head; 581 #ifdef CONFIG_LOCKDEP 582 struct task_struct *mq_freeze_owner; 583 int mq_freeze_owner_depth; 584 /* 585 * Records disk & queue state in current context, used in unfreeze 586 * queue 587 */ 588 bool mq_freeze_disk_dead; 589 bool mq_freeze_queue_dying; 590 #endif 591 wait_queue_head_t mq_freeze_wq; 592 /* 593 * Protect concurrent access to q_usage_counter by 594 * percpu_ref_kill() and percpu_ref_reinit(). 595 */ 596 struct mutex mq_freeze_lock; 597 598 struct blk_mq_tag_set *tag_set; 599 struct list_head tag_set_list; 600 601 struct dentry *debugfs_dir; 602 struct dentry *sched_debugfs_dir; 603 struct dentry *rqos_debugfs_dir; 604 /* 605 * Serializes all debugfs metadata operations using the above dentries. 606 */ 607 struct mutex debugfs_mutex; 608 609 bool mq_sysfs_init_done; 610 }; 611 612 /* Keep blk_queue_flag_name[] in sync with the definitions below */ 613 enum { 614 QUEUE_FLAG_DYING, /* queue being torn down */ 615 QUEUE_FLAG_NOMERGES, /* disable merge attempts */ 616 QUEUE_FLAG_SAME_COMP, /* complete on same CPU-group */ 617 QUEUE_FLAG_FAIL_IO, /* fake timeout */ 618 QUEUE_FLAG_NOXMERGES, /* No extended merges */ 619 QUEUE_FLAG_SAME_FORCE, /* force complete on same CPU */ 620 QUEUE_FLAG_INIT_DONE, /* queue is initialized */ 621 QUEUE_FLAG_STATS, /* track IO start and completion times */ 622 QUEUE_FLAG_REGISTERED, /* queue has been registered to a disk */ 623 QUEUE_FLAG_QUIESCED, /* queue has been quiesced */ 624 QUEUE_FLAG_RQ_ALLOC_TIME, /* record rq->alloc_time_ns */ 625 QUEUE_FLAG_HCTX_ACTIVE, /* at least one blk-mq hctx is active */ 626 QUEUE_FLAG_SQ_SCHED, /* single queue style io dispatch */ 627 QUEUE_FLAG_MAX 628 }; 629 630 #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) 631 632 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 633 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 634 635 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 636 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 637 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 638 #define blk_queue_noxmerges(q) \ 639 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 640 #define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL)) 641 #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) 642 #define blk_queue_passthrough_stat(q) \ 643 ((q)->limits.flags & BLK_FLAG_IOSTATS_PASSTHROUGH) 644 #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) 645 #define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA) 646 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 647 #define blk_queue_rq_alloc_time(q) \ 648 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) 649 #else 650 #define blk_queue_rq_alloc_time(q) false 651 #endif 652 653 #define blk_noretry_request(rq) \ 654 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 655 REQ_FAILFAST_DRIVER)) 656 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 657 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 658 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) 659 #define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags) 660 #define blk_queue_skip_tagset_quiesce(q) \ 661 ((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE) 662 663 extern void blk_set_pm_only(struct request_queue *q); 664 extern void blk_clear_pm_only(struct request_queue *q); 665 666 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 667 668 #define dma_map_bvec(dev, bv, dir, attrs) \ 669 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 670 (dir), (attrs)) 671 672 static inline bool queue_is_mq(struct request_queue *q) 673 { 674 return q->mq_ops; 675 } 676 677 #ifdef CONFIG_PM 678 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 679 { 680 return q->rpm_status; 681 } 682 #else 683 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 684 { 685 return RPM_ACTIVE; 686 } 687 #endif 688 689 static inline bool blk_queue_is_zoned(struct request_queue *q) 690 { 691 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && 692 (q->limits.features & BLK_FEAT_ZONED); 693 } 694 695 #ifdef CONFIG_BLK_DEV_ZONED 696 static inline unsigned int disk_nr_zones(struct gendisk *disk) 697 { 698 return disk->nr_zones; 699 } 700 bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); 701 #else /* CONFIG_BLK_DEV_ZONED */ 702 static inline unsigned int disk_nr_zones(struct gendisk *disk) 703 { 704 return 0; 705 } 706 static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) 707 { 708 return false; 709 } 710 #endif /* CONFIG_BLK_DEV_ZONED */ 711 712 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) 713 { 714 if (!blk_queue_is_zoned(disk->queue)) 715 return 0; 716 return sector >> ilog2(disk->queue->limits.chunk_sectors); 717 } 718 719 static inline unsigned int bdev_nr_zones(struct block_device *bdev) 720 { 721 return disk_nr_zones(bdev->bd_disk); 722 } 723 724 static inline unsigned int bdev_max_open_zones(struct block_device *bdev) 725 { 726 return bdev->bd_disk->queue->limits.max_open_zones; 727 } 728 729 static inline unsigned int bdev_max_active_zones(struct block_device *bdev) 730 { 731 return bdev->bd_disk->queue->limits.max_active_zones; 732 } 733 734 static inline unsigned int blk_queue_depth(struct request_queue *q) 735 { 736 if (q->queue_depth) 737 return q->queue_depth; 738 739 return q->nr_requests; 740 } 741 742 /* 743 * default timeout for SG_IO if none specified 744 */ 745 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 746 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 747 748 /* This should not be used directly - use rq_for_each_segment */ 749 #define for_each_bio(_bio) \ 750 for (; _bio; _bio = _bio->bi_next) 751 752 int __must_check add_disk_fwnode(struct device *parent, struct gendisk *disk, 753 const struct attribute_group **groups, 754 struct fwnode_handle *fwnode); 755 int __must_check device_add_disk(struct device *parent, struct gendisk *disk, 756 const struct attribute_group **groups); 757 static inline int __must_check add_disk(struct gendisk *disk) 758 { 759 return device_add_disk(NULL, disk, NULL); 760 } 761 void del_gendisk(struct gendisk *gp); 762 void invalidate_disk(struct gendisk *disk); 763 void set_disk_ro(struct gendisk *disk, bool read_only); 764 void disk_uevent(struct gendisk *disk, enum kobject_action action); 765 766 static inline u8 bdev_partno(const struct block_device *bdev) 767 { 768 return atomic_read(&bdev->__bd_flags) & BD_PARTNO; 769 } 770 771 static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag) 772 { 773 return atomic_read(&bdev->__bd_flags) & flag; 774 } 775 776 static inline void bdev_set_flag(struct block_device *bdev, unsigned flag) 777 { 778 atomic_or(flag, &bdev->__bd_flags); 779 } 780 781 static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag) 782 { 783 atomic_andnot(flag, &bdev->__bd_flags); 784 } 785 786 static inline bool get_disk_ro(struct gendisk *disk) 787 { 788 return bdev_test_flag(disk->part0, BD_READ_ONLY) || 789 test_bit(GD_READ_ONLY, &disk->state); 790 } 791 792 static inline bool bdev_read_only(struct block_device *bdev) 793 { 794 return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk); 795 } 796 797 bool set_capacity_and_notify(struct gendisk *disk, sector_t size); 798 void disk_force_media_change(struct gendisk *disk); 799 void bdev_mark_dead(struct block_device *bdev, bool surprise); 800 801 void add_disk_randomness(struct gendisk *disk) __latent_entropy; 802 void rand_initialize_disk(struct gendisk *disk); 803 804 static inline sector_t get_start_sect(struct block_device *bdev) 805 { 806 return bdev->bd_start_sect; 807 } 808 809 static inline sector_t bdev_nr_sectors(struct block_device *bdev) 810 { 811 return bdev->bd_nr_sectors; 812 } 813 814 static inline loff_t bdev_nr_bytes(struct block_device *bdev) 815 { 816 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT; 817 } 818 819 static inline sector_t get_capacity(struct gendisk *disk) 820 { 821 return bdev_nr_sectors(disk->part0); 822 } 823 824 static inline u64 sb_bdev_nr_blocks(struct super_block *sb) 825 { 826 return bdev_nr_sectors(sb->s_bdev) >> 827 (sb->s_blocksize_bits - SECTOR_SHIFT); 828 } 829 830 int bdev_disk_changed(struct gendisk *disk, bool invalidate); 831 832 void put_disk(struct gendisk *disk); 833 struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node, 834 struct lock_class_key *lkclass); 835 836 /** 837 * blk_alloc_disk - allocate a gendisk structure 838 * @lim: queue limits to be used for this disk. 839 * @node_id: numa node to allocate on 840 * 841 * Allocate and pre-initialize a gendisk structure for use with BIO based 842 * drivers. 843 * 844 * Returns an ERR_PTR on error, else the allocated disk. 845 * 846 * Context: can sleep 847 */ 848 #define blk_alloc_disk(lim, node_id) \ 849 ({ \ 850 static struct lock_class_key __key; \ 851 \ 852 __blk_alloc_disk(lim, node_id, &__key); \ 853 }) 854 855 int __register_blkdev(unsigned int major, const char *name, 856 void (*probe)(dev_t devt)); 857 #define register_blkdev(major, name) \ 858 __register_blkdev(major, name, NULL) 859 void unregister_blkdev(unsigned int major, const char *name); 860 861 bool disk_check_media_change(struct gendisk *disk); 862 void set_capacity(struct gendisk *disk, sector_t size); 863 864 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 865 int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk); 866 void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk); 867 #else 868 static inline int bd_link_disk_holder(struct block_device *bdev, 869 struct gendisk *disk) 870 { 871 return 0; 872 } 873 static inline void bd_unlink_disk_holder(struct block_device *bdev, 874 struct gendisk *disk) 875 { 876 } 877 #endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */ 878 879 dev_t part_devt(struct gendisk *disk, u8 partno); 880 void inc_diskseq(struct gendisk *disk); 881 void blk_request_module(dev_t devt); 882 883 extern int blk_register_queue(struct gendisk *disk); 884 extern void blk_unregister_queue(struct gendisk *disk); 885 void submit_bio_noacct(struct bio *bio); 886 struct bio *bio_split_to_limits(struct bio *bio); 887 888 extern int blk_lld_busy(struct request_queue *q); 889 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 890 extern void blk_queue_exit(struct request_queue *q); 891 extern void blk_sync_queue(struct request_queue *q); 892 893 /* Helper to convert REQ_OP_XXX to its string format XXX */ 894 extern const char *blk_op_str(enum req_op op); 895 896 int blk_status_to_errno(blk_status_t status); 897 blk_status_t errno_to_blk_status(int errno); 898 const char *blk_status_to_str(blk_status_t status); 899 900 /* only poll the hardware once, don't continue until a completion was found */ 901 #define BLK_POLL_ONESHOT (1 << 0) 902 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags); 903 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, 904 unsigned int flags); 905 906 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 907 { 908 return bdev->bd_queue; /* this is never NULL */ 909 } 910 911 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */ 912 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond); 913 914 static inline unsigned int bio_zone_no(struct bio *bio) 915 { 916 return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector); 917 } 918 919 static inline bool bio_straddles_zones(struct bio *bio) 920 { 921 return bio_sectors(bio) && 922 bio_zone_no(bio) != 923 disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1); 924 } 925 926 /* 927 * Return how much within the boundary is left to be used for I/O at a given 928 * offset. 929 */ 930 static inline unsigned int blk_boundary_sectors_left(sector_t offset, 931 unsigned int boundary_sectors) 932 { 933 if (unlikely(!is_power_of_2(boundary_sectors))) 934 return boundary_sectors - sector_div(offset, boundary_sectors); 935 return boundary_sectors - (offset & (boundary_sectors - 1)); 936 } 937 938 /** 939 * queue_limits_start_update - start an atomic update of queue limits 940 * @q: queue to update 941 * 942 * This functions starts an atomic update of the queue limits. It takes a lock 943 * to prevent other updates and returns a snapshot of the current limits that 944 * the caller can modify. The caller must call queue_limits_commit_update() 945 * to finish the update. 946 * 947 * Context: process context. 948 */ 949 static inline struct queue_limits 950 queue_limits_start_update(struct request_queue *q) 951 { 952 mutex_lock(&q->limits_lock); 953 return q->limits; 954 } 955 int queue_limits_commit_update_frozen(struct request_queue *q, 956 struct queue_limits *lim); 957 int queue_limits_commit_update(struct request_queue *q, 958 struct queue_limits *lim); 959 int queue_limits_set(struct request_queue *q, struct queue_limits *lim); 960 int blk_validate_limits(struct queue_limits *lim); 961 962 /** 963 * queue_limits_cancel_update - cancel an atomic update of queue limits 964 * @q: queue to update 965 * 966 * This functions cancels an atomic update of the queue limits started by 967 * queue_limits_start_update() and should be used when an error occurs after 968 * starting update. 969 */ 970 static inline void queue_limits_cancel_update(struct request_queue *q) 971 { 972 mutex_unlock(&q->limits_lock); 973 } 974 975 /* 976 * These helpers are for drivers that have sloppy feature negotiation and might 977 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O 978 * completion handler when the device returned an indicator that the respective 979 * feature is not actually supported. They are racy and the driver needs to 980 * cope with that. Try to avoid this scheme if you can. 981 */ 982 static inline void blk_queue_disable_discard(struct request_queue *q) 983 { 984 q->limits.max_discard_sectors = 0; 985 } 986 987 static inline void blk_queue_disable_secure_erase(struct request_queue *q) 988 { 989 q->limits.max_secure_erase_sectors = 0; 990 } 991 992 static inline void blk_queue_disable_write_zeroes(struct request_queue *q) 993 { 994 q->limits.max_write_zeroes_sectors = 0; 995 } 996 997 /* 998 * Access functions for manipulating queue properties 999 */ 1000 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1001 extern void blk_set_stacking_limits(struct queue_limits *lim); 1002 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1003 sector_t offset); 1004 void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, 1005 sector_t offset, const char *pfx); 1006 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1007 1008 struct blk_independent_access_ranges * 1009 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges); 1010 void disk_set_independent_access_ranges(struct gendisk *disk, 1011 struct blk_independent_access_ranges *iars); 1012 1013 bool __must_check blk_get_queue(struct request_queue *); 1014 extern void blk_put_queue(struct request_queue *); 1015 1016 void blk_mark_disk_dead(struct gendisk *disk); 1017 1018 struct rq_list { 1019 struct request *head; 1020 struct request *tail; 1021 }; 1022 1023 #ifdef CONFIG_BLOCK 1024 /* 1025 * blk_plug permits building a queue of related requests by holding the I/O 1026 * fragments for a short period. This allows merging of sequential requests 1027 * into single larger request. As the requests are moved from a per-task list to 1028 * the device's request_queue in a batch, this results in improved scalability 1029 * as the lock contention for request_queue lock is reduced. 1030 * 1031 * It is ok not to disable preemption when adding the request to the plug list 1032 * or when attempting a merge. For details, please see schedule() where 1033 * blk_flush_plug() is called. 1034 */ 1035 struct blk_plug { 1036 struct rq_list mq_list; /* blk-mq requests */ 1037 1038 /* if ios_left is > 1, we can batch tag/rq allocations */ 1039 struct rq_list cached_rqs; 1040 u64 cur_ktime; 1041 unsigned short nr_ios; 1042 1043 unsigned short rq_count; 1044 1045 bool multiple_queues; 1046 bool has_elevator; 1047 1048 struct list_head cb_list; /* md requires an unplug callback */ 1049 }; 1050 1051 struct blk_plug_cb; 1052 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1053 struct blk_plug_cb { 1054 struct list_head list; 1055 blk_plug_cb_fn callback; 1056 void *data; 1057 }; 1058 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1059 void *data, int size); 1060 extern void blk_start_plug(struct blk_plug *); 1061 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short); 1062 extern void blk_finish_plug(struct blk_plug *); 1063 1064 void __blk_flush_plug(struct blk_plug *plug, bool from_schedule); 1065 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1066 { 1067 if (plug) 1068 __blk_flush_plug(plug, async); 1069 } 1070 1071 /* 1072 * tsk == current here 1073 */ 1074 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1075 { 1076 struct blk_plug *plug = tsk->plug; 1077 1078 if (plug) 1079 plug->cur_ktime = 0; 1080 current->flags &= ~PF_BLOCK_TS; 1081 } 1082 1083 int blkdev_issue_flush(struct block_device *bdev); 1084 long nr_blockdev_pages(void); 1085 #else /* CONFIG_BLOCK */ 1086 struct blk_plug { 1087 }; 1088 1089 static inline void blk_start_plug_nr_ios(struct blk_plug *plug, 1090 unsigned short nr_ios) 1091 { 1092 } 1093 1094 static inline void blk_start_plug(struct blk_plug *plug) 1095 { 1096 } 1097 1098 static inline void blk_finish_plug(struct blk_plug *plug) 1099 { 1100 } 1101 1102 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1103 { 1104 } 1105 1106 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1107 { 1108 } 1109 1110 static inline int blkdev_issue_flush(struct block_device *bdev) 1111 { 1112 return 0; 1113 } 1114 1115 static inline long nr_blockdev_pages(void) 1116 { 1117 return 0; 1118 } 1119 #endif /* CONFIG_BLOCK */ 1120 1121 extern void blk_io_schedule(void); 1122 1123 int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1124 sector_t nr_sects, gfp_t gfp_mask); 1125 int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1126 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop); 1127 int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, 1128 sector_t nr_sects, gfp_t gfp); 1129 1130 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1131 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1132 #define BLKDEV_ZERO_KILLABLE (1 << 2) /* interruptible by fatal signals */ 1133 1134 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1135 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1136 unsigned flags); 1137 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1138 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1139 1140 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1141 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1142 { 1143 return blkdev_issue_discard(sb->s_bdev, 1144 block << (sb->s_blocksize_bits - 1145 SECTOR_SHIFT), 1146 nr_blocks << (sb->s_blocksize_bits - 1147 SECTOR_SHIFT), 1148 gfp_mask); 1149 } 1150 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1151 sector_t nr_blocks, gfp_t gfp_mask) 1152 { 1153 return blkdev_issue_zeroout(sb->s_bdev, 1154 block << (sb->s_blocksize_bits - 1155 SECTOR_SHIFT), 1156 nr_blocks << (sb->s_blocksize_bits - 1157 SECTOR_SHIFT), 1158 gfp_mask, 0); 1159 } 1160 1161 static inline bool bdev_is_partition(struct block_device *bdev) 1162 { 1163 return bdev_partno(bdev) != 0; 1164 } 1165 1166 enum blk_default_limits { 1167 BLK_MAX_SEGMENTS = 128, 1168 BLK_SAFE_MAX_SECTORS = 255, 1169 BLK_MAX_SEGMENT_SIZE = 65536, 1170 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1171 }; 1172 1173 /* 1174 * Default upper limit for the software max_sectors limit used for 1175 * regular file system I/O. This can be increased through sysfs. 1176 * 1177 * Not to be confused with the max_hw_sector limit that is entirely 1178 * controlled by the driver, usually based on hardware limits. 1179 */ 1180 #define BLK_DEF_MAX_SECTORS_CAP 2560u 1181 1182 static inline struct queue_limits *bdev_limits(struct block_device *bdev) 1183 { 1184 return &bdev_get_queue(bdev)->limits; 1185 } 1186 1187 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1188 { 1189 return q->limits.seg_boundary_mask; 1190 } 1191 1192 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1193 { 1194 return q->limits.virt_boundary_mask; 1195 } 1196 1197 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1198 { 1199 return q->limits.max_sectors; 1200 } 1201 1202 static inline unsigned int queue_max_bytes(struct request_queue *q) 1203 { 1204 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9; 1205 } 1206 1207 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1208 { 1209 return q->limits.max_hw_sectors; 1210 } 1211 1212 static inline unsigned short queue_max_segments(const struct request_queue *q) 1213 { 1214 return q->limits.max_segments; 1215 } 1216 1217 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1218 { 1219 return q->limits.max_discard_segments; 1220 } 1221 1222 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1223 { 1224 return q->limits.max_segment_size; 1225 } 1226 1227 static inline bool queue_emulates_zone_append(struct request_queue *q) 1228 { 1229 return blk_queue_is_zoned(q) && !q->limits.max_hw_zone_append_sectors; 1230 } 1231 1232 static inline bool bdev_emulates_zone_append(struct block_device *bdev) 1233 { 1234 return queue_emulates_zone_append(bdev_get_queue(bdev)); 1235 } 1236 1237 static inline unsigned int 1238 bdev_max_zone_append_sectors(struct block_device *bdev) 1239 { 1240 return bdev_limits(bdev)->max_zone_append_sectors; 1241 } 1242 1243 static inline unsigned int bdev_max_segments(struct block_device *bdev) 1244 { 1245 return queue_max_segments(bdev_get_queue(bdev)); 1246 } 1247 1248 static inline unsigned queue_logical_block_size(const struct request_queue *q) 1249 { 1250 return q->limits.logical_block_size; 1251 } 1252 1253 static inline unsigned int bdev_logical_block_size(struct block_device *bdev) 1254 { 1255 return queue_logical_block_size(bdev_get_queue(bdev)); 1256 } 1257 1258 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1259 { 1260 return q->limits.physical_block_size; 1261 } 1262 1263 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1264 { 1265 return queue_physical_block_size(bdev_get_queue(bdev)); 1266 } 1267 1268 static inline unsigned int queue_io_min(const struct request_queue *q) 1269 { 1270 return q->limits.io_min; 1271 } 1272 1273 static inline unsigned int bdev_io_min(struct block_device *bdev) 1274 { 1275 return queue_io_min(bdev_get_queue(bdev)); 1276 } 1277 1278 static inline unsigned int queue_io_opt(const struct request_queue *q) 1279 { 1280 return q->limits.io_opt; 1281 } 1282 1283 static inline unsigned int bdev_io_opt(struct block_device *bdev) 1284 { 1285 return queue_io_opt(bdev_get_queue(bdev)); 1286 } 1287 1288 static inline unsigned int 1289 queue_zone_write_granularity(const struct request_queue *q) 1290 { 1291 return q->limits.zone_write_granularity; 1292 } 1293 1294 static inline unsigned int 1295 bdev_zone_write_granularity(struct block_device *bdev) 1296 { 1297 return queue_zone_write_granularity(bdev_get_queue(bdev)); 1298 } 1299 1300 int bdev_alignment_offset(struct block_device *bdev); 1301 unsigned int bdev_discard_alignment(struct block_device *bdev); 1302 1303 static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev) 1304 { 1305 return bdev_limits(bdev)->max_discard_sectors; 1306 } 1307 1308 static inline unsigned int bdev_discard_granularity(struct block_device *bdev) 1309 { 1310 return bdev_limits(bdev)->discard_granularity; 1311 } 1312 1313 static inline unsigned int 1314 bdev_max_secure_erase_sectors(struct block_device *bdev) 1315 { 1316 return bdev_limits(bdev)->max_secure_erase_sectors; 1317 } 1318 1319 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1320 { 1321 return bdev_limits(bdev)->max_write_zeroes_sectors; 1322 } 1323 1324 static inline bool bdev_nonrot(struct block_device *bdev) 1325 { 1326 return blk_queue_nonrot(bdev_get_queue(bdev)); 1327 } 1328 1329 static inline bool bdev_synchronous(struct block_device *bdev) 1330 { 1331 return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS; 1332 } 1333 1334 static inline bool bdev_stable_writes(struct block_device *bdev) 1335 { 1336 struct request_queue *q = bdev_get_queue(bdev); 1337 1338 if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && 1339 q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) 1340 return true; 1341 return q->limits.features & BLK_FEAT_STABLE_WRITES; 1342 } 1343 1344 static inline bool blk_queue_write_cache(struct request_queue *q) 1345 { 1346 return (q->limits.features & BLK_FEAT_WRITE_CACHE) && 1347 !(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED); 1348 } 1349 1350 static inline bool bdev_write_cache(struct block_device *bdev) 1351 { 1352 return blk_queue_write_cache(bdev_get_queue(bdev)); 1353 } 1354 1355 static inline bool bdev_fua(struct block_device *bdev) 1356 { 1357 return bdev_limits(bdev)->features & BLK_FEAT_FUA; 1358 } 1359 1360 static inline bool bdev_nowait(struct block_device *bdev) 1361 { 1362 return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT; 1363 } 1364 1365 static inline bool bdev_is_zoned(struct block_device *bdev) 1366 { 1367 return blk_queue_is_zoned(bdev_get_queue(bdev)); 1368 } 1369 1370 static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec) 1371 { 1372 return disk_zone_no(bdev->bd_disk, sec); 1373 } 1374 1375 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1376 { 1377 struct request_queue *q = bdev_get_queue(bdev); 1378 1379 if (!blk_queue_is_zoned(q)) 1380 return 0; 1381 return q->limits.chunk_sectors; 1382 } 1383 1384 static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev, 1385 sector_t sector) 1386 { 1387 return sector & (bdev_zone_sectors(bdev) - 1); 1388 } 1389 1390 static inline sector_t bio_offset_from_zone_start(struct bio *bio) 1391 { 1392 return bdev_offset_from_zone_start(bio->bi_bdev, 1393 bio->bi_iter.bi_sector); 1394 } 1395 1396 static inline bool bdev_is_zone_start(struct block_device *bdev, 1397 sector_t sector) 1398 { 1399 return bdev_offset_from_zone_start(bdev, sector) == 0; 1400 } 1401 1402 /** 1403 * bdev_zone_is_seq - check if a sector belongs to a sequential write zone 1404 * @bdev: block device to check 1405 * @sector: sector number 1406 * 1407 * Check if @sector on @bdev is contained in a sequential write required zone. 1408 */ 1409 static inline bool bdev_zone_is_seq(struct block_device *bdev, sector_t sector) 1410 { 1411 bool is_seq = false; 1412 1413 #if IS_ENABLED(CONFIG_BLK_DEV_ZONED) 1414 if (bdev_is_zoned(bdev)) { 1415 struct gendisk *disk = bdev->bd_disk; 1416 unsigned long *bitmap; 1417 1418 rcu_read_lock(); 1419 bitmap = rcu_dereference(disk->conv_zones_bitmap); 1420 is_seq = !bitmap || 1421 !test_bit(disk_zone_no(disk, sector), bitmap); 1422 rcu_read_unlock(); 1423 } 1424 #endif 1425 1426 return is_seq; 1427 } 1428 1429 int blk_zone_issue_zeroout(struct block_device *bdev, sector_t sector, 1430 sector_t nr_sects, gfp_t gfp_mask); 1431 1432 static inline unsigned int queue_dma_alignment(const struct request_queue *q) 1433 { 1434 return q->limits.dma_alignment; 1435 } 1436 1437 static inline unsigned int 1438 queue_atomic_write_unit_max_bytes(const struct request_queue *q) 1439 { 1440 return q->limits.atomic_write_unit_max; 1441 } 1442 1443 static inline unsigned int 1444 queue_atomic_write_unit_min_bytes(const struct request_queue *q) 1445 { 1446 return q->limits.atomic_write_unit_min; 1447 } 1448 1449 static inline unsigned int 1450 queue_atomic_write_boundary_bytes(const struct request_queue *q) 1451 { 1452 return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT; 1453 } 1454 1455 static inline unsigned int 1456 queue_atomic_write_max_bytes(const struct request_queue *q) 1457 { 1458 return q->limits.atomic_write_max_sectors << SECTOR_SHIFT; 1459 } 1460 1461 static inline unsigned int bdev_dma_alignment(struct block_device *bdev) 1462 { 1463 return queue_dma_alignment(bdev_get_queue(bdev)); 1464 } 1465 1466 static inline bool bdev_iter_is_aligned(struct block_device *bdev, 1467 struct iov_iter *iter) 1468 { 1469 return iov_iter_is_aligned(iter, bdev_dma_alignment(bdev), 1470 bdev_logical_block_size(bdev) - 1); 1471 } 1472 1473 static inline unsigned int 1474 blk_lim_dma_alignment_and_pad(struct queue_limits *lim) 1475 { 1476 return lim->dma_alignment | lim->dma_pad_mask; 1477 } 1478 1479 static inline bool blk_rq_aligned(struct request_queue *q, unsigned long addr, 1480 unsigned int len) 1481 { 1482 unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits); 1483 1484 return !(addr & alignment) && !(len & alignment); 1485 } 1486 1487 /* assumes size > 256 */ 1488 static inline unsigned int blksize_bits(unsigned int size) 1489 { 1490 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT; 1491 } 1492 1493 int kblockd_schedule_work(struct work_struct *work); 1494 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1495 1496 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1497 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1498 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1499 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1500 1501 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 1502 1503 bool blk_crypto_register(struct blk_crypto_profile *profile, 1504 struct request_queue *q); 1505 1506 #else /* CONFIG_BLK_INLINE_ENCRYPTION */ 1507 1508 static inline bool blk_crypto_register(struct blk_crypto_profile *profile, 1509 struct request_queue *q) 1510 { 1511 return true; 1512 } 1513 1514 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */ 1515 1516 enum blk_unique_id { 1517 /* these match the Designator Types specified in SPC */ 1518 BLK_UID_T10 = 1, 1519 BLK_UID_EUI64 = 2, 1520 BLK_UID_NAA = 3, 1521 }; 1522 1523 struct block_device_operations { 1524 void (*submit_bio)(struct bio *bio); 1525 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob, 1526 unsigned int flags); 1527 int (*open)(struct gendisk *disk, blk_mode_t mode); 1528 void (*release)(struct gendisk *disk); 1529 int (*ioctl)(struct block_device *bdev, blk_mode_t mode, 1530 unsigned cmd, unsigned long arg); 1531 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode, 1532 unsigned cmd, unsigned long arg); 1533 unsigned int (*check_events) (struct gendisk *disk, 1534 unsigned int clearing); 1535 void (*unlock_native_capacity) (struct gendisk *); 1536 int (*getgeo)(struct block_device *, struct hd_geometry *); 1537 int (*set_read_only)(struct block_device *bdev, bool ro); 1538 void (*free_disk)(struct gendisk *disk); 1539 /* this callback is with swap_lock and sometimes page table lock held */ 1540 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1541 int (*report_zones)(struct gendisk *, sector_t sector, 1542 unsigned int nr_zones, report_zones_cb cb, void *data); 1543 char *(*devnode)(struct gendisk *disk, umode_t *mode); 1544 /* returns the length of the identifier or a negative errno: */ 1545 int (*get_unique_id)(struct gendisk *disk, u8 id[16], 1546 enum blk_unique_id id_type); 1547 struct module *owner; 1548 const struct pr_ops *pr_ops; 1549 1550 /* 1551 * Special callback for probing GPT entry at a given sector. 1552 * Needed by Android devices, used by GPT scanner and MMC blk 1553 * driver. 1554 */ 1555 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector); 1556 }; 1557 1558 #ifdef CONFIG_COMPAT 1559 extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t, 1560 unsigned int, unsigned long); 1561 #else 1562 #define blkdev_compat_ptr_ioctl NULL 1563 #endif 1564 1565 static inline void blk_wake_io_task(struct task_struct *waiter) 1566 { 1567 /* 1568 * If we're polling, the task itself is doing the completions. For 1569 * that case, we don't need to signal a wakeup, it's enough to just 1570 * mark us as RUNNING. 1571 */ 1572 if (waiter == current) 1573 __set_current_state(TASK_RUNNING); 1574 else 1575 wake_up_process(waiter); 1576 } 1577 1578 unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, 1579 unsigned long start_time); 1580 void bdev_end_io_acct(struct block_device *bdev, enum req_op op, 1581 unsigned int sectors, unsigned long start_time); 1582 1583 unsigned long bio_start_io_acct(struct bio *bio); 1584 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, 1585 struct block_device *orig_bdev); 1586 1587 /** 1588 * bio_end_io_acct - end I/O accounting for bio based drivers 1589 * @bio: bio to end account for 1590 * @start_time: start time returned by bio_start_io_acct() 1591 */ 1592 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time) 1593 { 1594 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev); 1595 } 1596 1597 int set_blocksize(struct file *file, int size); 1598 1599 int lookup_bdev(const char *pathname, dev_t *dev); 1600 1601 void blkdev_show(struct seq_file *seqf, off_t offset); 1602 1603 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */ 1604 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */ 1605 #ifdef CONFIG_BLOCK 1606 #define BLKDEV_MAJOR_MAX 512 1607 #else 1608 #define BLKDEV_MAJOR_MAX 0 1609 #endif 1610 1611 struct blk_holder_ops { 1612 void (*mark_dead)(struct block_device *bdev, bool surprise); 1613 1614 /* 1615 * Sync the file system mounted on the block device. 1616 */ 1617 void (*sync)(struct block_device *bdev); 1618 1619 /* 1620 * Freeze the file system mounted on the block device. 1621 */ 1622 int (*freeze)(struct block_device *bdev); 1623 1624 /* 1625 * Thaw the file system mounted on the block device. 1626 */ 1627 int (*thaw)(struct block_device *bdev); 1628 }; 1629 1630 /* 1631 * For filesystems using @fs_holder_ops, the @holder argument passed to 1632 * helpers used to open and claim block devices via 1633 * bd_prepare_to_claim() must point to a superblock. 1634 */ 1635 extern const struct blk_holder_ops fs_holder_ops; 1636 1637 /* 1638 * Return the correct open flags for blkdev_get_by_* for super block flags 1639 * as stored in sb->s_flags. 1640 */ 1641 #define sb_open_mode(flags) \ 1642 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \ 1643 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE)) 1644 1645 struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder, 1646 const struct blk_holder_ops *hops); 1647 struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode, 1648 void *holder, const struct blk_holder_ops *hops); 1649 int bd_prepare_to_claim(struct block_device *bdev, void *holder, 1650 const struct blk_holder_ops *hops); 1651 void bd_abort_claiming(struct block_device *bdev, void *holder); 1652 1653 /* just for blk-cgroup, don't use elsewhere */ 1654 struct block_device *blkdev_get_no_open(dev_t dev); 1655 void blkdev_put_no_open(struct block_device *bdev); 1656 1657 struct block_device *I_BDEV(struct inode *inode); 1658 struct block_device *file_bdev(struct file *bdev_file); 1659 bool disk_live(struct gendisk *disk); 1660 unsigned int block_size(struct block_device *bdev); 1661 1662 #ifdef CONFIG_BLOCK 1663 void invalidate_bdev(struct block_device *bdev); 1664 int sync_blockdev(struct block_device *bdev); 1665 int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend); 1666 int sync_blockdev_nowait(struct block_device *bdev); 1667 void sync_bdevs(bool wait); 1668 void bdev_statx(struct path *, struct kstat *, u32); 1669 void printk_all_partitions(void); 1670 int __init early_lookup_bdev(const char *pathname, dev_t *dev); 1671 #else 1672 static inline void invalidate_bdev(struct block_device *bdev) 1673 { 1674 } 1675 static inline int sync_blockdev(struct block_device *bdev) 1676 { 1677 return 0; 1678 } 1679 static inline int sync_blockdev_nowait(struct block_device *bdev) 1680 { 1681 return 0; 1682 } 1683 static inline void sync_bdevs(bool wait) 1684 { 1685 } 1686 static inline void bdev_statx(struct path *path, struct kstat *stat, 1687 u32 request_mask) 1688 { 1689 } 1690 static inline void printk_all_partitions(void) 1691 { 1692 } 1693 static inline int early_lookup_bdev(const char *pathname, dev_t *dev) 1694 { 1695 return -EINVAL; 1696 } 1697 #endif /* CONFIG_BLOCK */ 1698 1699 int bdev_freeze(struct block_device *bdev); 1700 int bdev_thaw(struct block_device *bdev); 1701 void bdev_fput(struct file *bdev_file); 1702 1703 struct io_comp_batch { 1704 struct rq_list req_list; 1705 bool need_ts; 1706 void (*complete)(struct io_comp_batch *); 1707 }; 1708 1709 static inline bool blk_atomic_write_start_sect_aligned(sector_t sector, 1710 struct queue_limits *limits) 1711 { 1712 unsigned int alignment = max(limits->atomic_write_hw_unit_min, 1713 limits->atomic_write_hw_boundary); 1714 1715 return IS_ALIGNED(sector, alignment >> SECTOR_SHIFT); 1716 } 1717 1718 static inline bool bdev_can_atomic_write(struct block_device *bdev) 1719 { 1720 struct request_queue *bd_queue = bdev->bd_queue; 1721 struct queue_limits *limits = &bd_queue->limits; 1722 1723 if (!limits->atomic_write_unit_min) 1724 return false; 1725 1726 if (bdev_is_partition(bdev)) 1727 return blk_atomic_write_start_sect_aligned(bdev->bd_start_sect, 1728 limits); 1729 1730 return true; 1731 } 1732 1733 static inline unsigned int 1734 bdev_atomic_write_unit_min_bytes(struct block_device *bdev) 1735 { 1736 if (!bdev_can_atomic_write(bdev)) 1737 return 0; 1738 return queue_atomic_write_unit_min_bytes(bdev_get_queue(bdev)); 1739 } 1740 1741 static inline unsigned int 1742 bdev_atomic_write_unit_max_bytes(struct block_device *bdev) 1743 { 1744 if (!bdev_can_atomic_write(bdev)) 1745 return 0; 1746 return queue_atomic_write_unit_max_bytes(bdev_get_queue(bdev)); 1747 } 1748 1749 #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { } 1750 1751 #endif /* _LINUX_BLKDEV_H */ 1752