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