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 /* blk_validate_limits() validates bsize, so drivers don't usually need to */ 272 static inline int blk_validate_block_size(unsigned long bsize) 273 { 274 if (bsize < 512 || bsize > PAGE_SIZE || !is_power_of_2(bsize)) 275 return -EINVAL; 276 277 return 0; 278 } 279 280 static inline bool blk_op_is_passthrough(blk_opf_t op) 281 { 282 op &= REQ_OP_MASK; 283 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 284 } 285 286 /* flags set by the driver in queue_limits.features */ 287 typedef unsigned int __bitwise blk_features_t; 288 289 /* supports a volatile write cache */ 290 #define BLK_FEAT_WRITE_CACHE ((__force blk_features_t)(1u << 0)) 291 292 /* supports passing on the FUA bit */ 293 #define BLK_FEAT_FUA ((__force blk_features_t)(1u << 1)) 294 295 /* rotational device (hard drive or floppy) */ 296 #define BLK_FEAT_ROTATIONAL ((__force blk_features_t)(1u << 2)) 297 298 /* contributes to the random number pool */ 299 #define BLK_FEAT_ADD_RANDOM ((__force blk_features_t)(1u << 3)) 300 301 /* do disk/partitions IO accounting */ 302 #define BLK_FEAT_IO_STAT ((__force blk_features_t)(1u << 4)) 303 304 /* don't modify data until writeback is done */ 305 #define BLK_FEAT_STABLE_WRITES ((__force blk_features_t)(1u << 5)) 306 307 /* always completes in submit context */ 308 #define BLK_FEAT_SYNCHRONOUS ((__force blk_features_t)(1u << 6)) 309 310 /* supports REQ_NOWAIT */ 311 #define BLK_FEAT_NOWAIT ((__force blk_features_t)(1u << 7)) 312 313 /* supports DAX */ 314 #define BLK_FEAT_DAX ((__force blk_features_t)(1u << 8)) 315 316 /* supports I/O polling */ 317 #define BLK_FEAT_POLL ((__force blk_features_t)(1u << 9)) 318 319 /* is a zoned device */ 320 #define BLK_FEAT_ZONED ((__force blk_features_t)(1u << 10)) 321 322 /* supports PCI(e) p2p requests */ 323 #define BLK_FEAT_PCI_P2PDMA ((__force blk_features_t)(1u << 12)) 324 325 /* skip this queue in blk_mq_(un)quiesce_tagset */ 326 #define BLK_FEAT_SKIP_TAGSET_QUIESCE ((__force blk_features_t)(1u << 13)) 327 328 /* bounce all highmem pages */ 329 #define BLK_FEAT_BOUNCE_HIGH ((__force blk_features_t)(1u << 14)) 330 331 /* undocumented magic for bcache */ 332 #define BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE \ 333 ((__force blk_features_t)(1u << 15)) 334 335 /* 336 * Flags automatically inherited when stacking limits. 337 */ 338 #define BLK_FEAT_INHERIT_MASK \ 339 (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ 340 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH | \ 341 BLK_FEAT_RAID_PARTIAL_STRIPES_EXPENSIVE) 342 343 /* internal flags in queue_limits.flags */ 344 typedef unsigned int __bitwise blk_flags_t; 345 346 /* do not send FLUSH/FUA commands despite advertising a write cache */ 347 #define BLK_FLAG_WRITE_CACHE_DISABLED ((__force blk_flags_t)(1u << 0)) 348 349 /* I/O topology is misaligned */ 350 #define BLK_FLAG_MISALIGNED ((__force blk_flags_t)(1u << 1)) 351 352 struct queue_limits { 353 blk_features_t features; 354 blk_flags_t flags; 355 unsigned long seg_boundary_mask; 356 unsigned long virt_boundary_mask; 357 358 unsigned int max_hw_sectors; 359 unsigned int max_dev_sectors; 360 unsigned int chunk_sectors; 361 unsigned int max_sectors; 362 unsigned int max_user_sectors; 363 unsigned int max_segment_size; 364 unsigned int physical_block_size; 365 unsigned int logical_block_size; 366 unsigned int alignment_offset; 367 unsigned int io_min; 368 unsigned int io_opt; 369 unsigned int max_discard_sectors; 370 unsigned int max_hw_discard_sectors; 371 unsigned int max_user_discard_sectors; 372 unsigned int max_secure_erase_sectors; 373 unsigned int max_write_zeroes_sectors; 374 unsigned int max_zone_append_sectors; 375 unsigned int discard_granularity; 376 unsigned int discard_alignment; 377 unsigned int zone_write_granularity; 378 379 /* atomic write limits */ 380 unsigned int atomic_write_hw_max; 381 unsigned int atomic_write_max_sectors; 382 unsigned int atomic_write_hw_boundary; 383 unsigned int atomic_write_boundary_sectors; 384 unsigned int atomic_write_hw_unit_min; 385 unsigned int atomic_write_unit_min; 386 unsigned int atomic_write_hw_unit_max; 387 unsigned int atomic_write_unit_max; 388 389 unsigned short max_segments; 390 unsigned short max_integrity_segments; 391 unsigned short max_discard_segments; 392 393 unsigned int max_open_zones; 394 unsigned int max_active_zones; 395 396 /* 397 * Drivers that set dma_alignment to less than 511 must be prepared to 398 * handle individual bvec's that are not a multiple of a SECTOR_SIZE 399 * due to possible offsets. 400 */ 401 unsigned int dma_alignment; 402 unsigned int dma_pad_mask; 403 404 struct blk_integrity integrity; 405 }; 406 407 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, 408 void *data); 409 410 #define BLK_ALL_ZONES ((unsigned int)-1) 411 int blkdev_report_zones(struct block_device *bdev, sector_t sector, 412 unsigned int nr_zones, report_zones_cb cb, void *data); 413 int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, 414 sector_t sectors, sector_t nr_sectors); 415 int blk_revalidate_disk_zones(struct gendisk *disk); 416 417 /* 418 * Independent access ranges: struct blk_independent_access_range describes 419 * a range of contiguous sectors that can be accessed using device command 420 * execution resources that are independent from the resources used for 421 * other access ranges. This is typically found with single-LUN multi-actuator 422 * HDDs where each access range is served by a different set of heads. 423 * The set of independent ranges supported by the device is defined using 424 * struct blk_independent_access_ranges. The independent ranges must not overlap 425 * and must include all sectors within the disk capacity (no sector holes 426 * allowed). 427 * For a device with multiple ranges, requests targeting sectors in different 428 * ranges can be executed in parallel. A request can straddle an access range 429 * boundary. 430 */ 431 struct blk_independent_access_range { 432 struct kobject kobj; 433 sector_t sector; 434 sector_t nr_sectors; 435 }; 436 437 struct blk_independent_access_ranges { 438 struct kobject kobj; 439 bool sysfs_registered; 440 unsigned int nr_ia_ranges; 441 struct blk_independent_access_range ia_range[]; 442 }; 443 444 struct request_queue { 445 /* 446 * The queue owner gets to use this for whatever they like. 447 * ll_rw_blk doesn't touch it. 448 */ 449 void *queuedata; 450 451 struct elevator_queue *elevator; 452 453 const struct blk_mq_ops *mq_ops; 454 455 /* sw queues */ 456 struct blk_mq_ctx __percpu *queue_ctx; 457 458 /* 459 * various queue flags, see QUEUE_* below 460 */ 461 unsigned long queue_flags; 462 463 unsigned int rq_timeout; 464 465 unsigned int queue_depth; 466 467 refcount_t refs; 468 469 /* hw dispatch queues */ 470 unsigned int nr_hw_queues; 471 struct xarray hctx_table; 472 473 struct percpu_ref q_usage_counter; 474 475 struct request *last_merge; 476 477 spinlock_t queue_lock; 478 479 int quiesce_depth; 480 481 struct gendisk *disk; 482 483 /* 484 * mq queue kobject 485 */ 486 struct kobject *mq_kobj; 487 488 struct queue_limits limits; 489 490 #ifdef CONFIG_PM 491 struct device *dev; 492 enum rpm_status rpm_status; 493 #endif 494 495 /* 496 * Number of contexts that have called blk_set_pm_only(). If this 497 * counter is above zero then only RQF_PM requests are processed. 498 */ 499 atomic_t pm_only; 500 501 struct blk_queue_stats *stats; 502 struct rq_qos *rq_qos; 503 struct mutex rq_qos_mutex; 504 505 /* 506 * ida allocated id for this queue. Used to index queues from 507 * ioctx. 508 */ 509 int id; 510 511 /* 512 * queue settings 513 */ 514 unsigned long nr_requests; /* Max # of requests */ 515 516 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 517 struct blk_crypto_profile *crypto_profile; 518 struct kobject *crypto_kobject; 519 #endif 520 521 struct timer_list timeout; 522 struct work_struct timeout_work; 523 524 atomic_t nr_active_requests_shared_tags; 525 526 struct blk_mq_tags *sched_shared_tags; 527 528 struct list_head icq_list; 529 #ifdef CONFIG_BLK_CGROUP 530 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 531 struct blkcg_gq *root_blkg; 532 struct list_head blkg_list; 533 struct mutex blkcg_mutex; 534 #endif 535 536 int node; 537 538 spinlock_t requeue_lock; 539 struct list_head requeue_list; 540 struct delayed_work requeue_work; 541 542 #ifdef CONFIG_BLK_DEV_IO_TRACE 543 struct blk_trace __rcu *blk_trace; 544 #endif 545 /* 546 * for flush operations 547 */ 548 struct blk_flush_queue *fq; 549 struct list_head flush_list; 550 551 struct mutex sysfs_lock; 552 struct mutex sysfs_dir_lock; 553 struct mutex limits_lock; 554 555 /* 556 * for reusing dead hctx instance in case of updating 557 * nr_hw_queues 558 */ 559 struct list_head unused_hctx_list; 560 spinlock_t unused_hctx_lock; 561 562 int mq_freeze_depth; 563 564 #ifdef CONFIG_BLK_DEV_THROTTLING 565 /* Throttle data */ 566 struct throtl_data *td; 567 #endif 568 struct rcu_head rcu_head; 569 wait_queue_head_t mq_freeze_wq; 570 /* 571 * Protect concurrent access to q_usage_counter by 572 * percpu_ref_kill() and percpu_ref_reinit(). 573 */ 574 struct mutex mq_freeze_lock; 575 576 struct blk_mq_tag_set *tag_set; 577 struct list_head tag_set_list; 578 579 struct dentry *debugfs_dir; 580 struct dentry *sched_debugfs_dir; 581 struct dentry *rqos_debugfs_dir; 582 /* 583 * Serializes all debugfs metadata operations using the above dentries. 584 */ 585 struct mutex debugfs_mutex; 586 587 bool mq_sysfs_init_done; 588 }; 589 590 /* Keep blk_queue_flag_name[] in sync with the definitions below */ 591 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */ 592 #define QUEUE_FLAG_DYING 1 /* queue being torn down */ 593 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */ 594 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ 595 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ 596 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ 597 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ 598 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ 599 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ 600 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ 601 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ 602 #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ 603 #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */ 604 #define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */ 605 606 #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) 607 608 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 609 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 610 611 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 612 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 613 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 614 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 615 #define blk_queue_noxmerges(q) \ 616 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 617 #define blk_queue_nonrot(q) (!((q)->limits.features & BLK_FEAT_ROTATIONAL)) 618 #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) 619 #define blk_queue_dax(q) ((q)->limits.features & BLK_FEAT_DAX) 620 #define blk_queue_pci_p2pdma(q) ((q)->limits.features & BLK_FEAT_PCI_P2PDMA) 621 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 622 #define blk_queue_rq_alloc_time(q) \ 623 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) 624 #else 625 #define blk_queue_rq_alloc_time(q) false 626 #endif 627 628 #define blk_noretry_request(rq) \ 629 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 630 REQ_FAILFAST_DRIVER)) 631 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 632 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 633 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) 634 #define blk_queue_sq_sched(q) test_bit(QUEUE_FLAG_SQ_SCHED, &(q)->queue_flags) 635 #define blk_queue_skip_tagset_quiesce(q) \ 636 ((q)->limits.features & BLK_FEAT_SKIP_TAGSET_QUIESCE) 637 638 extern void blk_set_pm_only(struct request_queue *q); 639 extern void blk_clear_pm_only(struct request_queue *q); 640 641 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 642 643 #define dma_map_bvec(dev, bv, dir, attrs) \ 644 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 645 (dir), (attrs)) 646 647 static inline bool queue_is_mq(struct request_queue *q) 648 { 649 return q->mq_ops; 650 } 651 652 #ifdef CONFIG_PM 653 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 654 { 655 return q->rpm_status; 656 } 657 #else 658 static inline enum rpm_status queue_rpm_status(struct request_queue *q) 659 { 660 return RPM_ACTIVE; 661 } 662 #endif 663 664 static inline bool blk_queue_is_zoned(struct request_queue *q) 665 { 666 return IS_ENABLED(CONFIG_BLK_DEV_ZONED) && 667 (q->limits.features & BLK_FEAT_ZONED); 668 } 669 670 #ifdef CONFIG_BLK_DEV_ZONED 671 static inline unsigned int disk_nr_zones(struct gendisk *disk) 672 { 673 return disk->nr_zones; 674 } 675 bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); 676 #else /* CONFIG_BLK_DEV_ZONED */ 677 static inline unsigned int disk_nr_zones(struct gendisk *disk) 678 { 679 return 0; 680 } 681 static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) 682 { 683 return false; 684 } 685 #endif /* CONFIG_BLK_DEV_ZONED */ 686 687 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) 688 { 689 if (!blk_queue_is_zoned(disk->queue)) 690 return 0; 691 return sector >> ilog2(disk->queue->limits.chunk_sectors); 692 } 693 694 static inline unsigned int bdev_nr_zones(struct block_device *bdev) 695 { 696 return disk_nr_zones(bdev->bd_disk); 697 } 698 699 static inline unsigned int bdev_max_open_zones(struct block_device *bdev) 700 { 701 return bdev->bd_disk->queue->limits.max_open_zones; 702 } 703 704 static inline unsigned int bdev_max_active_zones(struct block_device *bdev) 705 { 706 return bdev->bd_disk->queue->limits.max_active_zones; 707 } 708 709 static inline unsigned int blk_queue_depth(struct request_queue *q) 710 { 711 if (q->queue_depth) 712 return q->queue_depth; 713 714 return q->nr_requests; 715 } 716 717 /* 718 * default timeout for SG_IO if none specified 719 */ 720 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 721 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 722 723 /* This should not be used directly - use rq_for_each_segment */ 724 #define for_each_bio(_bio) \ 725 for (; _bio; _bio = _bio->bi_next) 726 727 int __must_check device_add_disk(struct device *parent, struct gendisk *disk, 728 const struct attribute_group **groups); 729 static inline int __must_check add_disk(struct gendisk *disk) 730 { 731 return device_add_disk(NULL, disk, NULL); 732 } 733 void del_gendisk(struct gendisk *gp); 734 void invalidate_disk(struct gendisk *disk); 735 void set_disk_ro(struct gendisk *disk, bool read_only); 736 void disk_uevent(struct gendisk *disk, enum kobject_action action); 737 738 static inline u8 bdev_partno(const struct block_device *bdev) 739 { 740 return atomic_read(&bdev->__bd_flags) & BD_PARTNO; 741 } 742 743 static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag) 744 { 745 return atomic_read(&bdev->__bd_flags) & flag; 746 } 747 748 static inline void bdev_set_flag(struct block_device *bdev, unsigned flag) 749 { 750 atomic_or(flag, &bdev->__bd_flags); 751 } 752 753 static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag) 754 { 755 atomic_andnot(flag, &bdev->__bd_flags); 756 } 757 758 static inline int get_disk_ro(struct gendisk *disk) 759 { 760 return bdev_test_flag(disk->part0, BD_READ_ONLY) || 761 test_bit(GD_READ_ONLY, &disk->state); 762 } 763 764 static inline int bdev_read_only(struct block_device *bdev) 765 { 766 return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk); 767 } 768 769 bool set_capacity_and_notify(struct gendisk *disk, sector_t size); 770 void disk_force_media_change(struct gendisk *disk); 771 void bdev_mark_dead(struct block_device *bdev, bool surprise); 772 773 void add_disk_randomness(struct gendisk *disk) __latent_entropy; 774 void rand_initialize_disk(struct gendisk *disk); 775 776 static inline sector_t get_start_sect(struct block_device *bdev) 777 { 778 return bdev->bd_start_sect; 779 } 780 781 static inline sector_t bdev_nr_sectors(struct block_device *bdev) 782 { 783 return bdev->bd_nr_sectors; 784 } 785 786 static inline loff_t bdev_nr_bytes(struct block_device *bdev) 787 { 788 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT; 789 } 790 791 static inline sector_t get_capacity(struct gendisk *disk) 792 { 793 return bdev_nr_sectors(disk->part0); 794 } 795 796 static inline u64 sb_bdev_nr_blocks(struct super_block *sb) 797 { 798 return bdev_nr_sectors(sb->s_bdev) >> 799 (sb->s_blocksize_bits - SECTOR_SHIFT); 800 } 801 802 int bdev_disk_changed(struct gendisk *disk, bool invalidate); 803 804 void put_disk(struct gendisk *disk); 805 struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node, 806 struct lock_class_key *lkclass); 807 808 /** 809 * blk_alloc_disk - allocate a gendisk structure 810 * @lim: queue limits to be used for this disk. 811 * @node_id: numa node to allocate on 812 * 813 * Allocate and pre-initialize a gendisk structure for use with BIO based 814 * drivers. 815 * 816 * Returns an ERR_PTR on error, else the allocated disk. 817 * 818 * Context: can sleep 819 */ 820 #define blk_alloc_disk(lim, node_id) \ 821 ({ \ 822 static struct lock_class_key __key; \ 823 \ 824 __blk_alloc_disk(lim, node_id, &__key); \ 825 }) 826 827 int __register_blkdev(unsigned int major, const char *name, 828 void (*probe)(dev_t devt)); 829 #define register_blkdev(major, name) \ 830 __register_blkdev(major, name, NULL) 831 void unregister_blkdev(unsigned int major, const char *name); 832 833 bool disk_check_media_change(struct gendisk *disk); 834 void set_capacity(struct gendisk *disk, sector_t size); 835 836 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 837 int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk); 838 void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk); 839 #else 840 static inline int bd_link_disk_holder(struct block_device *bdev, 841 struct gendisk *disk) 842 { 843 return 0; 844 } 845 static inline void bd_unlink_disk_holder(struct block_device *bdev, 846 struct gendisk *disk) 847 { 848 } 849 #endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */ 850 851 dev_t part_devt(struct gendisk *disk, u8 partno); 852 void inc_diskseq(struct gendisk *disk); 853 void blk_request_module(dev_t devt); 854 855 extern int blk_register_queue(struct gendisk *disk); 856 extern void blk_unregister_queue(struct gendisk *disk); 857 void submit_bio_noacct(struct bio *bio); 858 struct bio *bio_split_to_limits(struct bio *bio); 859 860 extern int blk_lld_busy(struct request_queue *q); 861 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 862 extern void blk_queue_exit(struct request_queue *q); 863 extern void blk_sync_queue(struct request_queue *q); 864 865 /* Helper to convert REQ_OP_XXX to its string format XXX */ 866 extern const char *blk_op_str(enum req_op op); 867 868 int blk_status_to_errno(blk_status_t status); 869 blk_status_t errno_to_blk_status(int errno); 870 const char *blk_status_to_str(blk_status_t status); 871 872 /* only poll the hardware once, don't continue until a completion was found */ 873 #define BLK_POLL_ONESHOT (1 << 0) 874 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags); 875 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, 876 unsigned int flags); 877 878 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 879 { 880 return bdev->bd_queue; /* this is never NULL */ 881 } 882 883 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */ 884 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond); 885 886 static inline unsigned int bio_zone_no(struct bio *bio) 887 { 888 return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector); 889 } 890 891 static inline bool bio_straddles_zones(struct bio *bio) 892 { 893 return bio_sectors(bio) && 894 bio_zone_no(bio) != 895 disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1); 896 } 897 898 /* 899 * Return how much within the boundary is left to be used for I/O at a given 900 * offset. 901 */ 902 static inline unsigned int blk_boundary_sectors_left(sector_t offset, 903 unsigned int boundary_sectors) 904 { 905 if (unlikely(!is_power_of_2(boundary_sectors))) 906 return boundary_sectors - sector_div(offset, boundary_sectors); 907 return boundary_sectors - (offset & (boundary_sectors - 1)); 908 } 909 910 /** 911 * queue_limits_start_update - start an atomic update of queue limits 912 * @q: queue to update 913 * 914 * This functions starts an atomic update of the queue limits. It takes a lock 915 * to prevent other updates and returns a snapshot of the current limits that 916 * the caller can modify. The caller must call queue_limits_commit_update() 917 * to finish the update. 918 * 919 * Context: process context. The caller must have frozen the queue or ensured 920 * that there is outstanding I/O by other means. 921 */ 922 static inline struct queue_limits 923 queue_limits_start_update(struct request_queue *q) 924 { 925 mutex_lock(&q->limits_lock); 926 return q->limits; 927 } 928 int queue_limits_commit_update(struct request_queue *q, 929 struct queue_limits *lim); 930 int queue_limits_set(struct request_queue *q, struct queue_limits *lim); 931 932 /** 933 * queue_limits_cancel_update - cancel an atomic update of queue limits 934 * @q: queue to update 935 * 936 * This functions cancels an atomic update of the queue limits started by 937 * queue_limits_start_update() and should be used when an error occurs after 938 * starting update. 939 */ 940 static inline void queue_limits_cancel_update(struct request_queue *q) 941 { 942 mutex_unlock(&q->limits_lock); 943 } 944 945 /* 946 * These helpers are for drivers that have sloppy feature negotiation and might 947 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O 948 * completion handler when the device returned an indicator that the respective 949 * feature is not actually supported. They are racy and the driver needs to 950 * cope with that. Try to avoid this scheme if you can. 951 */ 952 static inline void blk_queue_disable_discard(struct request_queue *q) 953 { 954 q->limits.max_discard_sectors = 0; 955 } 956 957 static inline void blk_queue_disable_secure_erase(struct request_queue *q) 958 { 959 q->limits.max_secure_erase_sectors = 0; 960 } 961 962 static inline void blk_queue_disable_write_zeroes(struct request_queue *q) 963 { 964 q->limits.max_write_zeroes_sectors = 0; 965 } 966 967 /* 968 * Access functions for manipulating queue properties 969 */ 970 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 971 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 972 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 973 extern void blk_set_stacking_limits(struct queue_limits *lim); 974 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 975 sector_t offset); 976 void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, 977 sector_t offset, const char *pfx); 978 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 979 980 struct blk_independent_access_ranges * 981 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges); 982 void disk_set_independent_access_ranges(struct gendisk *disk, 983 struct blk_independent_access_ranges *iars); 984 985 bool __must_check blk_get_queue(struct request_queue *); 986 extern void blk_put_queue(struct request_queue *); 987 988 void blk_mark_disk_dead(struct gendisk *disk); 989 990 #ifdef CONFIG_BLOCK 991 /* 992 * blk_plug permits building a queue of related requests by holding the I/O 993 * fragments for a short period. This allows merging of sequential requests 994 * into single larger request. As the requests are moved from a per-task list to 995 * the device's request_queue in a batch, this results in improved scalability 996 * as the lock contention for request_queue lock is reduced. 997 * 998 * It is ok not to disable preemption when adding the request to the plug list 999 * or when attempting a merge. For details, please see schedule() where 1000 * blk_flush_plug() is called. 1001 */ 1002 struct blk_plug { 1003 struct request *mq_list; /* blk-mq requests */ 1004 1005 /* if ios_left is > 1, we can batch tag/rq allocations */ 1006 struct request *cached_rq; 1007 u64 cur_ktime; 1008 unsigned short nr_ios; 1009 1010 unsigned short rq_count; 1011 1012 bool multiple_queues; 1013 bool has_elevator; 1014 1015 struct list_head cb_list; /* md requires an unplug callback */ 1016 }; 1017 1018 struct blk_plug_cb; 1019 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1020 struct blk_plug_cb { 1021 struct list_head list; 1022 blk_plug_cb_fn callback; 1023 void *data; 1024 }; 1025 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1026 void *data, int size); 1027 extern void blk_start_plug(struct blk_plug *); 1028 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short); 1029 extern void blk_finish_plug(struct blk_plug *); 1030 1031 void __blk_flush_plug(struct blk_plug *plug, bool from_schedule); 1032 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1033 { 1034 if (plug) 1035 __blk_flush_plug(plug, async); 1036 } 1037 1038 /* 1039 * tsk == current here 1040 */ 1041 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1042 { 1043 struct blk_plug *plug = tsk->plug; 1044 1045 if (plug) 1046 plug->cur_ktime = 0; 1047 current->flags &= ~PF_BLOCK_TS; 1048 } 1049 1050 int blkdev_issue_flush(struct block_device *bdev); 1051 long nr_blockdev_pages(void); 1052 #else /* CONFIG_BLOCK */ 1053 struct blk_plug { 1054 }; 1055 1056 static inline void blk_start_plug_nr_ios(struct blk_plug *plug, 1057 unsigned short nr_ios) 1058 { 1059 } 1060 1061 static inline void blk_start_plug(struct blk_plug *plug) 1062 { 1063 } 1064 1065 static inline void blk_finish_plug(struct blk_plug *plug) 1066 { 1067 } 1068 1069 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1070 { 1071 } 1072 1073 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1074 { 1075 } 1076 1077 static inline int blkdev_issue_flush(struct block_device *bdev) 1078 { 1079 return 0; 1080 } 1081 1082 static inline long nr_blockdev_pages(void) 1083 { 1084 return 0; 1085 } 1086 #endif /* CONFIG_BLOCK */ 1087 1088 extern void blk_io_schedule(void); 1089 1090 int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1091 sector_t nr_sects, gfp_t gfp_mask); 1092 int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1093 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop); 1094 int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, 1095 sector_t nr_sects, gfp_t gfp); 1096 1097 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1098 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1099 #define BLKDEV_ZERO_KILLABLE (1 << 2) /* interruptible by fatal signals */ 1100 1101 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1102 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1103 unsigned flags); 1104 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1105 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1106 1107 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1108 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1109 { 1110 return blkdev_issue_discard(sb->s_bdev, 1111 block << (sb->s_blocksize_bits - 1112 SECTOR_SHIFT), 1113 nr_blocks << (sb->s_blocksize_bits - 1114 SECTOR_SHIFT), 1115 gfp_mask); 1116 } 1117 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1118 sector_t nr_blocks, gfp_t gfp_mask) 1119 { 1120 return blkdev_issue_zeroout(sb->s_bdev, 1121 block << (sb->s_blocksize_bits - 1122 SECTOR_SHIFT), 1123 nr_blocks << (sb->s_blocksize_bits - 1124 SECTOR_SHIFT), 1125 gfp_mask, 0); 1126 } 1127 1128 static inline bool bdev_is_partition(struct block_device *bdev) 1129 { 1130 return bdev_partno(bdev) != 0; 1131 } 1132 1133 enum blk_default_limits { 1134 BLK_MAX_SEGMENTS = 128, 1135 BLK_SAFE_MAX_SECTORS = 255, 1136 BLK_MAX_SEGMENT_SIZE = 65536, 1137 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1138 }; 1139 1140 /* 1141 * Default upper limit for the software max_sectors limit used for 1142 * regular file system I/O. This can be increased through sysfs. 1143 * 1144 * Not to be confused with the max_hw_sector limit that is entirely 1145 * controlled by the driver, usually based on hardware limits. 1146 */ 1147 #define BLK_DEF_MAX_SECTORS_CAP 2560u 1148 1149 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1150 { 1151 return q->limits.seg_boundary_mask; 1152 } 1153 1154 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1155 { 1156 return q->limits.virt_boundary_mask; 1157 } 1158 1159 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1160 { 1161 return q->limits.max_sectors; 1162 } 1163 1164 static inline unsigned int queue_max_bytes(struct request_queue *q) 1165 { 1166 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9; 1167 } 1168 1169 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1170 { 1171 return q->limits.max_hw_sectors; 1172 } 1173 1174 static inline unsigned short queue_max_segments(const struct request_queue *q) 1175 { 1176 return q->limits.max_segments; 1177 } 1178 1179 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1180 { 1181 return q->limits.max_discard_segments; 1182 } 1183 1184 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1185 { 1186 return q->limits.max_segment_size; 1187 } 1188 1189 static inline unsigned int queue_limits_max_zone_append_sectors(struct queue_limits *l) 1190 { 1191 unsigned int max_sectors = min(l->chunk_sectors, l->max_hw_sectors); 1192 1193 return min_not_zero(l->max_zone_append_sectors, max_sectors); 1194 } 1195 1196 static inline unsigned int queue_max_zone_append_sectors(struct request_queue *q) 1197 { 1198 if (!blk_queue_is_zoned(q)) 1199 return 0; 1200 1201 return queue_limits_max_zone_append_sectors(&q->limits); 1202 } 1203 1204 static inline bool queue_emulates_zone_append(struct request_queue *q) 1205 { 1206 return blk_queue_is_zoned(q) && !q->limits.max_zone_append_sectors; 1207 } 1208 1209 static inline bool bdev_emulates_zone_append(struct block_device *bdev) 1210 { 1211 return queue_emulates_zone_append(bdev_get_queue(bdev)); 1212 } 1213 1214 static inline unsigned int 1215 bdev_max_zone_append_sectors(struct block_device *bdev) 1216 { 1217 return queue_max_zone_append_sectors(bdev_get_queue(bdev)); 1218 } 1219 1220 static inline unsigned int bdev_max_segments(struct block_device *bdev) 1221 { 1222 return queue_max_segments(bdev_get_queue(bdev)); 1223 } 1224 1225 static inline unsigned queue_logical_block_size(const struct request_queue *q) 1226 { 1227 return q->limits.logical_block_size; 1228 } 1229 1230 static inline unsigned int bdev_logical_block_size(struct block_device *bdev) 1231 { 1232 return queue_logical_block_size(bdev_get_queue(bdev)); 1233 } 1234 1235 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1236 { 1237 return q->limits.physical_block_size; 1238 } 1239 1240 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1241 { 1242 return queue_physical_block_size(bdev_get_queue(bdev)); 1243 } 1244 1245 static inline unsigned int queue_io_min(const struct request_queue *q) 1246 { 1247 return q->limits.io_min; 1248 } 1249 1250 static inline int bdev_io_min(struct block_device *bdev) 1251 { 1252 return queue_io_min(bdev_get_queue(bdev)); 1253 } 1254 1255 static inline unsigned int queue_io_opt(const struct request_queue *q) 1256 { 1257 return q->limits.io_opt; 1258 } 1259 1260 static inline int bdev_io_opt(struct block_device *bdev) 1261 { 1262 return queue_io_opt(bdev_get_queue(bdev)); 1263 } 1264 1265 static inline unsigned int 1266 queue_zone_write_granularity(const struct request_queue *q) 1267 { 1268 return q->limits.zone_write_granularity; 1269 } 1270 1271 static inline unsigned int 1272 bdev_zone_write_granularity(struct block_device *bdev) 1273 { 1274 return queue_zone_write_granularity(bdev_get_queue(bdev)); 1275 } 1276 1277 int bdev_alignment_offset(struct block_device *bdev); 1278 unsigned int bdev_discard_alignment(struct block_device *bdev); 1279 1280 static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev) 1281 { 1282 return bdev_get_queue(bdev)->limits.max_discard_sectors; 1283 } 1284 1285 static inline unsigned int bdev_discard_granularity(struct block_device *bdev) 1286 { 1287 return bdev_get_queue(bdev)->limits.discard_granularity; 1288 } 1289 1290 static inline unsigned int 1291 bdev_max_secure_erase_sectors(struct block_device *bdev) 1292 { 1293 return bdev_get_queue(bdev)->limits.max_secure_erase_sectors; 1294 } 1295 1296 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1297 { 1298 struct request_queue *q = bdev_get_queue(bdev); 1299 1300 if (q) 1301 return q->limits.max_write_zeroes_sectors; 1302 1303 return 0; 1304 } 1305 1306 static inline bool bdev_nonrot(struct block_device *bdev) 1307 { 1308 return blk_queue_nonrot(bdev_get_queue(bdev)); 1309 } 1310 1311 static inline bool bdev_synchronous(struct block_device *bdev) 1312 { 1313 return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS; 1314 } 1315 1316 static inline bool bdev_stable_writes(struct block_device *bdev) 1317 { 1318 struct request_queue *q = bdev_get_queue(bdev); 1319 1320 if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && 1321 q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) 1322 return true; 1323 return q->limits.features & BLK_FEAT_STABLE_WRITES; 1324 } 1325 1326 static inline bool blk_queue_write_cache(struct request_queue *q) 1327 { 1328 return (q->limits.features & BLK_FEAT_WRITE_CACHE) && 1329 !(q->limits.flags & BLK_FLAG_WRITE_CACHE_DISABLED); 1330 } 1331 1332 static inline bool bdev_write_cache(struct block_device *bdev) 1333 { 1334 return blk_queue_write_cache(bdev_get_queue(bdev)); 1335 } 1336 1337 static inline bool bdev_fua(struct block_device *bdev) 1338 { 1339 return bdev_get_queue(bdev)->limits.features & BLK_FEAT_FUA; 1340 } 1341 1342 static inline bool bdev_nowait(struct block_device *bdev) 1343 { 1344 return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT; 1345 } 1346 1347 static inline bool bdev_is_zoned(struct block_device *bdev) 1348 { 1349 return blk_queue_is_zoned(bdev_get_queue(bdev)); 1350 } 1351 1352 static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec) 1353 { 1354 return disk_zone_no(bdev->bd_disk, sec); 1355 } 1356 1357 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1358 { 1359 struct request_queue *q = bdev_get_queue(bdev); 1360 1361 if (!blk_queue_is_zoned(q)) 1362 return 0; 1363 return q->limits.chunk_sectors; 1364 } 1365 1366 static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev, 1367 sector_t sector) 1368 { 1369 return sector & (bdev_zone_sectors(bdev) - 1); 1370 } 1371 1372 static inline sector_t bio_offset_from_zone_start(struct bio *bio) 1373 { 1374 return bdev_offset_from_zone_start(bio->bi_bdev, 1375 bio->bi_iter.bi_sector); 1376 } 1377 1378 static inline bool bdev_is_zone_start(struct block_device *bdev, 1379 sector_t sector) 1380 { 1381 return bdev_offset_from_zone_start(bdev, sector) == 0; 1382 } 1383 1384 static inline int queue_dma_alignment(const struct request_queue *q) 1385 { 1386 return q->limits.dma_alignment; 1387 } 1388 1389 static inline unsigned int 1390 queue_atomic_write_unit_max_bytes(const struct request_queue *q) 1391 { 1392 return q->limits.atomic_write_unit_max; 1393 } 1394 1395 static inline unsigned int 1396 queue_atomic_write_unit_min_bytes(const struct request_queue *q) 1397 { 1398 return q->limits.atomic_write_unit_min; 1399 } 1400 1401 static inline unsigned int 1402 queue_atomic_write_boundary_bytes(const struct request_queue *q) 1403 { 1404 return q->limits.atomic_write_boundary_sectors << SECTOR_SHIFT; 1405 } 1406 1407 static inline unsigned int 1408 queue_atomic_write_max_bytes(const struct request_queue *q) 1409 { 1410 return q->limits.atomic_write_max_sectors << SECTOR_SHIFT; 1411 } 1412 1413 static inline unsigned int bdev_dma_alignment(struct block_device *bdev) 1414 { 1415 return queue_dma_alignment(bdev_get_queue(bdev)); 1416 } 1417 1418 static inline bool bdev_iter_is_aligned(struct block_device *bdev, 1419 struct iov_iter *iter) 1420 { 1421 return iov_iter_is_aligned(iter, bdev_dma_alignment(bdev), 1422 bdev_logical_block_size(bdev) - 1); 1423 } 1424 1425 static inline int blk_lim_dma_alignment_and_pad(struct queue_limits *lim) 1426 { 1427 return lim->dma_alignment | lim->dma_pad_mask; 1428 } 1429 1430 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1431 unsigned int len) 1432 { 1433 unsigned int alignment = blk_lim_dma_alignment_and_pad(&q->limits); 1434 1435 return !(addr & alignment) && !(len & alignment); 1436 } 1437 1438 /* assumes size > 256 */ 1439 static inline unsigned int blksize_bits(unsigned int size) 1440 { 1441 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT; 1442 } 1443 1444 int kblockd_schedule_work(struct work_struct *work); 1445 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1446 1447 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1448 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1449 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1450 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1451 1452 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 1453 1454 bool blk_crypto_register(struct blk_crypto_profile *profile, 1455 struct request_queue *q); 1456 1457 #else /* CONFIG_BLK_INLINE_ENCRYPTION */ 1458 1459 static inline bool blk_crypto_register(struct blk_crypto_profile *profile, 1460 struct request_queue *q) 1461 { 1462 return true; 1463 } 1464 1465 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */ 1466 1467 enum blk_unique_id { 1468 /* these match the Designator Types specified in SPC */ 1469 BLK_UID_T10 = 1, 1470 BLK_UID_EUI64 = 2, 1471 BLK_UID_NAA = 3, 1472 }; 1473 1474 struct block_device_operations { 1475 void (*submit_bio)(struct bio *bio); 1476 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob, 1477 unsigned int flags); 1478 int (*open)(struct gendisk *disk, blk_mode_t mode); 1479 void (*release)(struct gendisk *disk); 1480 int (*ioctl)(struct block_device *bdev, blk_mode_t mode, 1481 unsigned cmd, unsigned long arg); 1482 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode, 1483 unsigned cmd, unsigned long arg); 1484 unsigned int (*check_events) (struct gendisk *disk, 1485 unsigned int clearing); 1486 void (*unlock_native_capacity) (struct gendisk *); 1487 int (*getgeo)(struct block_device *, struct hd_geometry *); 1488 int (*set_read_only)(struct block_device *bdev, bool ro); 1489 void (*free_disk)(struct gendisk *disk); 1490 /* this callback is with swap_lock and sometimes page table lock held */ 1491 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1492 int (*report_zones)(struct gendisk *, sector_t sector, 1493 unsigned int nr_zones, report_zones_cb cb, void *data); 1494 char *(*devnode)(struct gendisk *disk, umode_t *mode); 1495 /* returns the length of the identifier or a negative errno: */ 1496 int (*get_unique_id)(struct gendisk *disk, u8 id[16], 1497 enum blk_unique_id id_type); 1498 struct module *owner; 1499 const struct pr_ops *pr_ops; 1500 1501 /* 1502 * Special callback for probing GPT entry at a given sector. 1503 * Needed by Android devices, used by GPT scanner and MMC blk 1504 * driver. 1505 */ 1506 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector); 1507 }; 1508 1509 #ifdef CONFIG_COMPAT 1510 extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t, 1511 unsigned int, unsigned long); 1512 #else 1513 #define blkdev_compat_ptr_ioctl NULL 1514 #endif 1515 1516 static inline void blk_wake_io_task(struct task_struct *waiter) 1517 { 1518 /* 1519 * If we're polling, the task itself is doing the completions. For 1520 * that case, we don't need to signal a wakeup, it's enough to just 1521 * mark us as RUNNING. 1522 */ 1523 if (waiter == current) 1524 __set_current_state(TASK_RUNNING); 1525 else 1526 wake_up_process(waiter); 1527 } 1528 1529 unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, 1530 unsigned long start_time); 1531 void bdev_end_io_acct(struct block_device *bdev, enum req_op op, 1532 unsigned int sectors, unsigned long start_time); 1533 1534 unsigned long bio_start_io_acct(struct bio *bio); 1535 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, 1536 struct block_device *orig_bdev); 1537 1538 /** 1539 * bio_end_io_acct - end I/O accounting for bio based drivers 1540 * @bio: bio to end account for 1541 * @start_time: start time returned by bio_start_io_acct() 1542 */ 1543 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time) 1544 { 1545 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev); 1546 } 1547 1548 int bdev_read_only(struct block_device *bdev); 1549 int set_blocksize(struct file *file, int size); 1550 1551 int lookup_bdev(const char *pathname, dev_t *dev); 1552 1553 void blkdev_show(struct seq_file *seqf, off_t offset); 1554 1555 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */ 1556 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */ 1557 #ifdef CONFIG_BLOCK 1558 #define BLKDEV_MAJOR_MAX 512 1559 #else 1560 #define BLKDEV_MAJOR_MAX 0 1561 #endif 1562 1563 struct blk_holder_ops { 1564 void (*mark_dead)(struct block_device *bdev, bool surprise); 1565 1566 /* 1567 * Sync the file system mounted on the block device. 1568 */ 1569 void (*sync)(struct block_device *bdev); 1570 1571 /* 1572 * Freeze the file system mounted on the block device. 1573 */ 1574 int (*freeze)(struct block_device *bdev); 1575 1576 /* 1577 * Thaw the file system mounted on the block device. 1578 */ 1579 int (*thaw)(struct block_device *bdev); 1580 }; 1581 1582 /* 1583 * For filesystems using @fs_holder_ops, the @holder argument passed to 1584 * helpers used to open and claim block devices via 1585 * bd_prepare_to_claim() must point to a superblock. 1586 */ 1587 extern const struct blk_holder_ops fs_holder_ops; 1588 1589 /* 1590 * Return the correct open flags for blkdev_get_by_* for super block flags 1591 * as stored in sb->s_flags. 1592 */ 1593 #define sb_open_mode(flags) \ 1594 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \ 1595 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE)) 1596 1597 struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder, 1598 const struct blk_holder_ops *hops); 1599 struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode, 1600 void *holder, const struct blk_holder_ops *hops); 1601 int bd_prepare_to_claim(struct block_device *bdev, void *holder, 1602 const struct blk_holder_ops *hops); 1603 void bd_abort_claiming(struct block_device *bdev, void *holder); 1604 1605 /* just for blk-cgroup, don't use elsewhere */ 1606 struct block_device *blkdev_get_no_open(dev_t dev); 1607 void blkdev_put_no_open(struct block_device *bdev); 1608 1609 struct block_device *I_BDEV(struct inode *inode); 1610 struct block_device *file_bdev(struct file *bdev_file); 1611 bool disk_live(struct gendisk *disk); 1612 unsigned int block_size(struct block_device *bdev); 1613 1614 #ifdef CONFIG_BLOCK 1615 void invalidate_bdev(struct block_device *bdev); 1616 int sync_blockdev(struct block_device *bdev); 1617 int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend); 1618 int sync_blockdev_nowait(struct block_device *bdev); 1619 void sync_bdevs(bool wait); 1620 void bdev_statx(struct path *, struct kstat *, u32); 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 path *path, 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