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 enum { 287 /* supports a volatile write cache */ 288 BLK_FEAT_WRITE_CACHE = (1u << 0), 289 290 /* supports passing on the FUA bit */ 291 BLK_FEAT_FUA = (1u << 1), 292 293 /* rotational device (hard drive or floppy) */ 294 BLK_FEAT_ROTATIONAL = (1u << 2), 295 296 /* contributes to the random number pool */ 297 BLK_FEAT_ADD_RANDOM = (1u << 3), 298 299 /* do disk/partitions IO accounting */ 300 BLK_FEAT_IO_STAT = (1u << 4), 301 302 /* don't modify data until writeback is done */ 303 BLK_FEAT_STABLE_WRITES = (1u << 5), 304 305 /* always completes in submit context */ 306 BLK_FEAT_SYNCHRONOUS = (1u << 6), 307 308 /* supports REQ_NOWAIT */ 309 BLK_FEAT_NOWAIT = (1u << 7), 310 311 /* supports DAX */ 312 BLK_FEAT_DAX = (1u << 8), 313 314 /* supports I/O polling */ 315 BLK_FEAT_POLL = (1u << 9), 316 317 /* is a zoned device */ 318 BLK_FEAT_ZONED = (1u << 10), 319 320 /* supports Zone Reset All */ 321 BLK_FEAT_ZONE_RESETALL = (1u << 11), 322 323 /* supports PCI(e) p2p requests */ 324 BLK_FEAT_PCI_P2PDMA = (1u << 12), 325 326 /* skip this queue in blk_mq_(un)quiesce_tagset */ 327 BLK_FEAT_SKIP_TAGSET_QUIESCE = (1u << 13), 328 329 /* bounce all highmem pages */ 330 BLK_FEAT_BOUNCE_HIGH = (1u << 14), 331 }; 332 333 /* 334 * Flags automatically inherited when stacking limits. 335 */ 336 #define BLK_FEAT_INHERIT_MASK \ 337 (BLK_FEAT_WRITE_CACHE | BLK_FEAT_FUA | BLK_FEAT_ROTATIONAL | \ 338 BLK_FEAT_STABLE_WRITES | BLK_FEAT_ZONED | BLK_FEAT_BOUNCE_HIGH) 339 340 /* internal flags in queue_limits.flags */ 341 enum { 342 /* do not send FLUSH or FUA command despite advertised write cache */ 343 BLK_FLAGS_WRITE_CACHE_DISABLED = (1u << 31), 344 }; 345 346 /* 347 * BLK_BOUNCE_NONE: never bounce (default) 348 * BLK_BOUNCE_HIGH: bounce all highmem pages 349 */ 350 enum blk_bounce { 351 BLK_BOUNCE_NONE, 352 BLK_BOUNCE_HIGH, 353 }; 354 355 struct queue_limits { 356 unsigned int features; 357 unsigned int flags; 358 unsigned long seg_boundary_mask; 359 unsigned long virt_boundary_mask; 360 361 unsigned int max_hw_sectors; 362 unsigned int max_dev_sectors; 363 unsigned int chunk_sectors; 364 unsigned int max_sectors; 365 unsigned int max_user_sectors; 366 unsigned int max_segment_size; 367 unsigned int physical_block_size; 368 unsigned int logical_block_size; 369 unsigned int alignment_offset; 370 unsigned int io_min; 371 unsigned int io_opt; 372 unsigned int max_discard_sectors; 373 unsigned int max_hw_discard_sectors; 374 unsigned int max_user_discard_sectors; 375 unsigned int max_secure_erase_sectors; 376 unsigned int max_write_zeroes_sectors; 377 unsigned int max_zone_append_sectors; 378 unsigned int discard_granularity; 379 unsigned int discard_alignment; 380 unsigned int zone_write_granularity; 381 382 unsigned short max_segments; 383 unsigned short max_integrity_segments; 384 unsigned short max_discard_segments; 385 386 unsigned char misaligned; 387 unsigned char discard_misaligned; 388 unsigned char raid_partial_stripes_expensive; 389 unsigned int max_open_zones; 390 unsigned int max_active_zones; 391 392 /* 393 * Drivers that set dma_alignment to less than 511 must be prepared to 394 * handle individual bvec's that are not a multiple of a SECTOR_SIZE 395 * due to possible offsets. 396 */ 397 unsigned int dma_alignment; 398 399 struct blk_integrity integrity; 400 }; 401 402 typedef int (*report_zones_cb)(struct blk_zone *zone, unsigned int idx, 403 void *data); 404 405 #define BLK_ALL_ZONES ((unsigned int)-1) 406 int blkdev_report_zones(struct block_device *bdev, sector_t sector, 407 unsigned int nr_zones, report_zones_cb cb, void *data); 408 int blkdev_zone_mgmt(struct block_device *bdev, enum req_op op, 409 sector_t sectors, sector_t nr_sectors); 410 int blk_revalidate_disk_zones(struct gendisk *disk); 411 412 /* 413 * Independent access ranges: struct blk_independent_access_range describes 414 * a range of contiguous sectors that can be accessed using device command 415 * execution resources that are independent from the resources used for 416 * other access ranges. This is typically found with single-LUN multi-actuator 417 * HDDs where each access range is served by a different set of heads. 418 * The set of independent ranges supported by the device is defined using 419 * struct blk_independent_access_ranges. The independent ranges must not overlap 420 * and must include all sectors within the disk capacity (no sector holes 421 * allowed). 422 * For a device with multiple ranges, requests targeting sectors in different 423 * ranges can be executed in parallel. A request can straddle an access range 424 * boundary. 425 */ 426 struct blk_independent_access_range { 427 struct kobject kobj; 428 sector_t sector; 429 sector_t nr_sectors; 430 }; 431 432 struct blk_independent_access_ranges { 433 struct kobject kobj; 434 bool sysfs_registered; 435 unsigned int nr_ia_ranges; 436 struct blk_independent_access_range ia_range[]; 437 }; 438 439 struct request_queue { 440 /* 441 * The queue owner gets to use this for whatever they like. 442 * ll_rw_blk doesn't touch it. 443 */ 444 void *queuedata; 445 446 struct elevator_queue *elevator; 447 448 const struct blk_mq_ops *mq_ops; 449 450 /* sw queues */ 451 struct blk_mq_ctx __percpu *queue_ctx; 452 453 /* 454 * various queue flags, see QUEUE_* below 455 */ 456 unsigned long queue_flags; 457 458 unsigned int rq_timeout; 459 460 unsigned int queue_depth; 461 462 refcount_t refs; 463 464 /* hw dispatch queues */ 465 unsigned int nr_hw_queues; 466 struct xarray hctx_table; 467 468 struct percpu_ref q_usage_counter; 469 470 struct request *last_merge; 471 472 spinlock_t queue_lock; 473 474 int quiesce_depth; 475 476 struct gendisk *disk; 477 478 /* 479 * mq queue kobject 480 */ 481 struct kobject *mq_kobj; 482 483 struct queue_limits limits; 484 485 #ifdef CONFIG_PM 486 struct device *dev; 487 enum rpm_status rpm_status; 488 #endif 489 490 /* 491 * Number of contexts that have called blk_set_pm_only(). If this 492 * counter is above zero then only RQF_PM requests are processed. 493 */ 494 atomic_t pm_only; 495 496 struct blk_queue_stats *stats; 497 struct rq_qos *rq_qos; 498 struct mutex rq_qos_mutex; 499 500 /* 501 * ida allocated id for this queue. Used to index queues from 502 * ioctx. 503 */ 504 int id; 505 506 unsigned int dma_pad_mask; 507 508 /* 509 * queue settings 510 */ 511 unsigned long nr_requests; /* Max # of requests */ 512 513 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 514 struct blk_crypto_profile *crypto_profile; 515 struct kobject *crypto_kobject; 516 #endif 517 518 struct timer_list timeout; 519 struct work_struct timeout_work; 520 521 atomic_t nr_active_requests_shared_tags; 522 523 struct blk_mq_tags *sched_shared_tags; 524 525 struct list_head icq_list; 526 #ifdef CONFIG_BLK_CGROUP 527 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 528 struct blkcg_gq *root_blkg; 529 struct list_head blkg_list; 530 struct mutex blkcg_mutex; 531 #endif 532 533 int node; 534 535 spinlock_t requeue_lock; 536 struct list_head requeue_list; 537 struct delayed_work requeue_work; 538 539 #ifdef CONFIG_BLK_DEV_IO_TRACE 540 struct blk_trace __rcu *blk_trace; 541 #endif 542 /* 543 * for flush operations 544 */ 545 struct blk_flush_queue *fq; 546 struct list_head flush_list; 547 548 struct mutex sysfs_lock; 549 struct mutex sysfs_dir_lock; 550 struct mutex limits_lock; 551 552 /* 553 * for reusing dead hctx instance in case of updating 554 * nr_hw_queues 555 */ 556 struct list_head unused_hctx_list; 557 spinlock_t unused_hctx_lock; 558 559 int mq_freeze_depth; 560 561 #ifdef CONFIG_BLK_DEV_THROTTLING 562 /* Throttle data */ 563 struct throtl_data *td; 564 #endif 565 struct rcu_head rcu_head; 566 wait_queue_head_t mq_freeze_wq; 567 /* 568 * Protect concurrent access to q_usage_counter by 569 * percpu_ref_kill() and percpu_ref_reinit(). 570 */ 571 struct mutex mq_freeze_lock; 572 573 struct blk_mq_tag_set *tag_set; 574 struct list_head tag_set_list; 575 576 struct dentry *debugfs_dir; 577 struct dentry *sched_debugfs_dir; 578 struct dentry *rqos_debugfs_dir; 579 /* 580 * Serializes all debugfs metadata operations using the above dentries. 581 */ 582 struct mutex debugfs_mutex; 583 584 bool mq_sysfs_init_done; 585 }; 586 587 /* Keep blk_queue_flag_name[] in sync with the definitions below */ 588 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */ 589 #define QUEUE_FLAG_DYING 1 /* queue being torn down */ 590 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */ 591 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ 592 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ 593 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ 594 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ 595 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ 596 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ 597 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ 598 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ 599 #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ 600 #define QUEUE_FLAG_HCTX_ACTIVE 28 /* at least one blk-mq hctx is active */ 601 #define QUEUE_FLAG_SQ_SCHED 30 /* single queue style io dispatch */ 602 603 #define QUEUE_FLAG_MQ_DEFAULT (1UL << QUEUE_FLAG_SAME_COMP) 604 605 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 606 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 607 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); 608 609 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 610 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 611 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 612 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 613 #define blk_queue_noxmerges(q) \ 614 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 615 #define blk_queue_nonrot(q) ((q)->limits.features & BLK_FEAT_ROTATIONAL) 616 #define blk_queue_io_stat(q) ((q)->limits.features & BLK_FEAT_IO_STAT) 617 #define blk_queue_zone_resetall(q) \ 618 ((q)->limits.features & BLK_FEAT_ZONE_RESETALL) 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 unsigned int bdev_nr_zones(struct block_device *bdev); 672 673 static inline unsigned int disk_nr_zones(struct gendisk *disk) 674 { 675 return blk_queue_is_zoned(disk->queue) ? disk->nr_zones : 0; 676 } 677 678 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) 679 { 680 if (!blk_queue_is_zoned(disk->queue)) 681 return 0; 682 return sector >> ilog2(disk->queue->limits.chunk_sectors); 683 } 684 685 static inline unsigned int bdev_max_open_zones(struct block_device *bdev) 686 { 687 return bdev->bd_disk->queue->limits.max_open_zones; 688 } 689 690 static inline unsigned int bdev_max_active_zones(struct block_device *bdev) 691 { 692 return bdev->bd_disk->queue->limits.max_active_zones; 693 } 694 695 bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs); 696 #else /* CONFIG_BLK_DEV_ZONED */ 697 static inline unsigned int bdev_nr_zones(struct block_device *bdev) 698 { 699 return 0; 700 } 701 702 static inline unsigned int disk_nr_zones(struct gendisk *disk) 703 { 704 return 0; 705 } 706 static inline unsigned int disk_zone_no(struct gendisk *disk, sector_t sector) 707 { 708 return 0; 709 } 710 static inline unsigned int bdev_max_open_zones(struct block_device *bdev) 711 { 712 return 0; 713 } 714 715 static inline unsigned int bdev_max_active_zones(struct block_device *bdev) 716 { 717 return 0; 718 } 719 static inline bool blk_zone_plug_bio(struct bio *bio, unsigned int nr_segs) 720 { 721 return false; 722 } 723 #endif /* CONFIG_BLK_DEV_ZONED */ 724 725 static inline unsigned int blk_queue_depth(struct request_queue *q) 726 { 727 if (q->queue_depth) 728 return q->queue_depth; 729 730 return q->nr_requests; 731 } 732 733 /* 734 * default timeout for SG_IO if none specified 735 */ 736 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 737 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 738 739 /* This should not be used directly - use rq_for_each_segment */ 740 #define for_each_bio(_bio) \ 741 for (; _bio; _bio = _bio->bi_next) 742 743 int __must_check device_add_disk(struct device *parent, struct gendisk *disk, 744 const struct attribute_group **groups); 745 static inline int __must_check add_disk(struct gendisk *disk) 746 { 747 return device_add_disk(NULL, disk, NULL); 748 } 749 void del_gendisk(struct gendisk *gp); 750 void invalidate_disk(struct gendisk *disk); 751 void set_disk_ro(struct gendisk *disk, bool read_only); 752 void disk_uevent(struct gendisk *disk, enum kobject_action action); 753 754 static inline u8 bdev_partno(const struct block_device *bdev) 755 { 756 return atomic_read(&bdev->__bd_flags) & BD_PARTNO; 757 } 758 759 static inline bool bdev_test_flag(const struct block_device *bdev, unsigned flag) 760 { 761 return atomic_read(&bdev->__bd_flags) & flag; 762 } 763 764 static inline void bdev_set_flag(struct block_device *bdev, unsigned flag) 765 { 766 atomic_or(flag, &bdev->__bd_flags); 767 } 768 769 static inline void bdev_clear_flag(struct block_device *bdev, unsigned flag) 770 { 771 atomic_andnot(flag, &bdev->__bd_flags); 772 } 773 774 static inline int get_disk_ro(struct gendisk *disk) 775 { 776 return bdev_test_flag(disk->part0, BD_READ_ONLY) || 777 test_bit(GD_READ_ONLY, &disk->state); 778 } 779 780 static inline int bdev_read_only(struct block_device *bdev) 781 { 782 return bdev_test_flag(bdev, BD_READ_ONLY) || get_disk_ro(bdev->bd_disk); 783 } 784 785 bool set_capacity_and_notify(struct gendisk *disk, sector_t size); 786 void disk_force_media_change(struct gendisk *disk); 787 void bdev_mark_dead(struct block_device *bdev, bool surprise); 788 789 void add_disk_randomness(struct gendisk *disk) __latent_entropy; 790 void rand_initialize_disk(struct gendisk *disk); 791 792 static inline sector_t get_start_sect(struct block_device *bdev) 793 { 794 return bdev->bd_start_sect; 795 } 796 797 static inline sector_t bdev_nr_sectors(struct block_device *bdev) 798 { 799 return bdev->bd_nr_sectors; 800 } 801 802 static inline loff_t bdev_nr_bytes(struct block_device *bdev) 803 { 804 return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT; 805 } 806 807 static inline sector_t get_capacity(struct gendisk *disk) 808 { 809 return bdev_nr_sectors(disk->part0); 810 } 811 812 static inline u64 sb_bdev_nr_blocks(struct super_block *sb) 813 { 814 return bdev_nr_sectors(sb->s_bdev) >> 815 (sb->s_blocksize_bits - SECTOR_SHIFT); 816 } 817 818 int bdev_disk_changed(struct gendisk *disk, bool invalidate); 819 820 void put_disk(struct gendisk *disk); 821 struct gendisk *__blk_alloc_disk(struct queue_limits *lim, int node, 822 struct lock_class_key *lkclass); 823 824 /** 825 * blk_alloc_disk - allocate a gendisk structure 826 * @lim: queue limits to be used for this disk. 827 * @node_id: numa node to allocate on 828 * 829 * Allocate and pre-initialize a gendisk structure for use with BIO based 830 * drivers. 831 * 832 * Returns an ERR_PTR on error, else the allocated disk. 833 * 834 * Context: can sleep 835 */ 836 #define blk_alloc_disk(lim, node_id) \ 837 ({ \ 838 static struct lock_class_key __key; \ 839 \ 840 __blk_alloc_disk(lim, node_id, &__key); \ 841 }) 842 843 int __register_blkdev(unsigned int major, const char *name, 844 void (*probe)(dev_t devt)); 845 #define register_blkdev(major, name) \ 846 __register_blkdev(major, name, NULL) 847 void unregister_blkdev(unsigned int major, const char *name); 848 849 bool disk_check_media_change(struct gendisk *disk); 850 void set_capacity(struct gendisk *disk, sector_t size); 851 852 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 853 int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk); 854 void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk); 855 #else 856 static inline int bd_link_disk_holder(struct block_device *bdev, 857 struct gendisk *disk) 858 { 859 return 0; 860 } 861 static inline void bd_unlink_disk_holder(struct block_device *bdev, 862 struct gendisk *disk) 863 { 864 } 865 #endif /* CONFIG_BLOCK_HOLDER_DEPRECATED */ 866 867 dev_t part_devt(struct gendisk *disk, u8 partno); 868 void inc_diskseq(struct gendisk *disk); 869 void blk_request_module(dev_t devt); 870 871 extern int blk_register_queue(struct gendisk *disk); 872 extern void blk_unregister_queue(struct gendisk *disk); 873 void submit_bio_noacct(struct bio *bio); 874 struct bio *bio_split_to_limits(struct bio *bio); 875 876 extern int blk_lld_busy(struct request_queue *q); 877 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 878 extern void blk_queue_exit(struct request_queue *q); 879 extern void blk_sync_queue(struct request_queue *q); 880 881 /* Helper to convert REQ_OP_XXX to its string format XXX */ 882 extern const char *blk_op_str(enum req_op op); 883 884 int blk_status_to_errno(blk_status_t status); 885 blk_status_t errno_to_blk_status(int errno); 886 const char *blk_status_to_str(blk_status_t status); 887 888 /* only poll the hardware once, don't continue until a completion was found */ 889 #define BLK_POLL_ONESHOT (1 << 0) 890 int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags); 891 int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, 892 unsigned int flags); 893 894 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 895 { 896 return bdev->bd_queue; /* this is never NULL */ 897 } 898 899 /* Helper to convert BLK_ZONE_ZONE_XXX to its string format XXX */ 900 const char *blk_zone_cond_str(enum blk_zone_cond zone_cond); 901 902 static inline unsigned int bio_zone_no(struct bio *bio) 903 { 904 return disk_zone_no(bio->bi_bdev->bd_disk, bio->bi_iter.bi_sector); 905 } 906 907 static inline bool bio_straddles_zones(struct bio *bio) 908 { 909 return bio_sectors(bio) && 910 bio_zone_no(bio) != 911 disk_zone_no(bio->bi_bdev->bd_disk, bio_end_sector(bio) - 1); 912 } 913 914 /* 915 * Return how much of the chunk is left to be used for I/O at a given offset. 916 */ 917 static inline unsigned int blk_chunk_sectors_left(sector_t offset, 918 unsigned int chunk_sectors) 919 { 920 if (unlikely(!is_power_of_2(chunk_sectors))) 921 return chunk_sectors - sector_div(offset, chunk_sectors); 922 return chunk_sectors - (offset & (chunk_sectors - 1)); 923 } 924 925 /** 926 * queue_limits_start_update - start an atomic update of queue limits 927 * @q: queue to update 928 * 929 * This functions starts an atomic update of the queue limits. It takes a lock 930 * to prevent other updates and returns a snapshot of the current limits that 931 * the caller can modify. The caller must call queue_limits_commit_update() 932 * to finish the update. 933 * 934 * Context: process context. The caller must have frozen the queue or ensured 935 * that there is outstanding I/O by other means. 936 */ 937 static inline struct queue_limits 938 queue_limits_start_update(struct request_queue *q) 939 { 940 mutex_lock(&q->limits_lock); 941 return q->limits; 942 } 943 int queue_limits_commit_update(struct request_queue *q, 944 struct queue_limits *lim); 945 int queue_limits_set(struct request_queue *q, struct queue_limits *lim); 946 947 /** 948 * queue_limits_cancel_update - cancel an atomic update of queue limits 949 * @q: queue to update 950 * 951 * This functions cancels an atomic update of the queue limits started by 952 * queue_limits_start_update() and should be used when an error occurs after 953 * starting update. 954 */ 955 static inline void queue_limits_cancel_update(struct request_queue *q) 956 { 957 mutex_unlock(&q->limits_lock); 958 } 959 960 /* 961 * These helpers are for drivers that have sloppy feature negotiation and might 962 * have to disable DISCARD, WRITE_ZEROES or SECURE_DISCARD from the I/O 963 * completion handler when the device returned an indicator that the respective 964 * feature is not actually supported. They are racy and the driver needs to 965 * cope with that. Try to avoid this scheme if you can. 966 */ 967 static inline void blk_queue_disable_discard(struct request_queue *q) 968 { 969 q->limits.max_discard_sectors = 0; 970 } 971 972 static inline void blk_queue_disable_secure_erase(struct request_queue *q) 973 { 974 q->limits.max_secure_erase_sectors = 0; 975 } 976 977 static inline void blk_queue_disable_write_zeroes(struct request_queue *q) 978 { 979 q->limits.max_write_zeroes_sectors = 0; 980 } 981 982 /* 983 * Access functions for manipulating queue properties 984 */ 985 void disk_update_readahead(struct gendisk *disk); 986 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 987 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 988 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 989 extern void blk_set_stacking_limits(struct queue_limits *lim); 990 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 991 sector_t offset); 992 void queue_limits_stack_bdev(struct queue_limits *t, struct block_device *bdev, 993 sector_t offset, const char *pfx); 994 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 995 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 996 997 struct blk_independent_access_ranges * 998 disk_alloc_independent_access_ranges(struct gendisk *disk, int nr_ia_ranges); 999 void disk_set_independent_access_ranges(struct gendisk *disk, 1000 struct blk_independent_access_ranges *iars); 1001 1002 bool __must_check blk_get_queue(struct request_queue *); 1003 extern void blk_put_queue(struct request_queue *); 1004 1005 void blk_mark_disk_dead(struct gendisk *disk); 1006 1007 #ifdef CONFIG_BLOCK 1008 /* 1009 * blk_plug permits building a queue of related requests by holding the I/O 1010 * fragments for a short period. This allows merging of sequential requests 1011 * into single larger request. As the requests are moved from a per-task list to 1012 * the device's request_queue in a batch, this results in improved scalability 1013 * as the lock contention for request_queue lock is reduced. 1014 * 1015 * It is ok not to disable preemption when adding the request to the plug list 1016 * or when attempting a merge. For details, please see schedule() where 1017 * blk_flush_plug() is called. 1018 */ 1019 struct blk_plug { 1020 struct request *mq_list; /* blk-mq requests */ 1021 1022 /* if ios_left is > 1, we can batch tag/rq allocations */ 1023 struct request *cached_rq; 1024 u64 cur_ktime; 1025 unsigned short nr_ios; 1026 1027 unsigned short rq_count; 1028 1029 bool multiple_queues; 1030 bool has_elevator; 1031 1032 struct list_head cb_list; /* md requires an unplug callback */ 1033 }; 1034 1035 struct blk_plug_cb; 1036 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1037 struct blk_plug_cb { 1038 struct list_head list; 1039 blk_plug_cb_fn callback; 1040 void *data; 1041 }; 1042 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1043 void *data, int size); 1044 extern void blk_start_plug(struct blk_plug *); 1045 extern void blk_start_plug_nr_ios(struct blk_plug *, unsigned short); 1046 extern void blk_finish_plug(struct blk_plug *); 1047 1048 void __blk_flush_plug(struct blk_plug *plug, bool from_schedule); 1049 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1050 { 1051 if (plug) 1052 __blk_flush_plug(plug, async); 1053 } 1054 1055 /* 1056 * tsk == current here 1057 */ 1058 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1059 { 1060 struct blk_plug *plug = tsk->plug; 1061 1062 if (plug) 1063 plug->cur_ktime = 0; 1064 current->flags &= ~PF_BLOCK_TS; 1065 } 1066 1067 int blkdev_issue_flush(struct block_device *bdev); 1068 long nr_blockdev_pages(void); 1069 #else /* CONFIG_BLOCK */ 1070 struct blk_plug { 1071 }; 1072 1073 static inline void blk_start_plug_nr_ios(struct blk_plug *plug, 1074 unsigned short nr_ios) 1075 { 1076 } 1077 1078 static inline void blk_start_plug(struct blk_plug *plug) 1079 { 1080 } 1081 1082 static inline void blk_finish_plug(struct blk_plug *plug) 1083 { 1084 } 1085 1086 static inline void blk_flush_plug(struct blk_plug *plug, bool async) 1087 { 1088 } 1089 1090 static inline void blk_plug_invalidate_ts(struct task_struct *tsk) 1091 { 1092 } 1093 1094 static inline int blkdev_issue_flush(struct block_device *bdev) 1095 { 1096 return 0; 1097 } 1098 1099 static inline long nr_blockdev_pages(void) 1100 { 1101 return 0; 1102 } 1103 #endif /* CONFIG_BLOCK */ 1104 1105 extern void blk_io_schedule(void); 1106 1107 int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1108 sector_t nr_sects, gfp_t gfp_mask); 1109 int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1110 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop); 1111 int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector, 1112 sector_t nr_sects, gfp_t gfp); 1113 1114 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1115 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1116 1117 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1118 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1119 unsigned flags); 1120 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1121 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1122 1123 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1124 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1125 { 1126 return blkdev_issue_discard(sb->s_bdev, 1127 block << (sb->s_blocksize_bits - 1128 SECTOR_SHIFT), 1129 nr_blocks << (sb->s_blocksize_bits - 1130 SECTOR_SHIFT), 1131 gfp_mask); 1132 } 1133 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1134 sector_t nr_blocks, gfp_t gfp_mask) 1135 { 1136 return blkdev_issue_zeroout(sb->s_bdev, 1137 block << (sb->s_blocksize_bits - 1138 SECTOR_SHIFT), 1139 nr_blocks << (sb->s_blocksize_bits - 1140 SECTOR_SHIFT), 1141 gfp_mask, 0); 1142 } 1143 1144 static inline bool bdev_is_partition(struct block_device *bdev) 1145 { 1146 return bdev_partno(bdev) != 0; 1147 } 1148 1149 enum blk_default_limits { 1150 BLK_MAX_SEGMENTS = 128, 1151 BLK_SAFE_MAX_SECTORS = 255, 1152 BLK_MAX_SEGMENT_SIZE = 65536, 1153 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1154 }; 1155 1156 /* 1157 * Default upper limit for the software max_sectors limit used for 1158 * regular file system I/O. This can be increased through sysfs. 1159 * 1160 * Not to be confused with the max_hw_sector limit that is entirely 1161 * controlled by the driver, usually based on hardware limits. 1162 */ 1163 #define BLK_DEF_MAX_SECTORS_CAP 2560u 1164 1165 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1166 { 1167 return q->limits.seg_boundary_mask; 1168 } 1169 1170 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1171 { 1172 return q->limits.virt_boundary_mask; 1173 } 1174 1175 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1176 { 1177 return q->limits.max_sectors; 1178 } 1179 1180 static inline unsigned int queue_max_bytes(struct request_queue *q) 1181 { 1182 return min_t(unsigned int, queue_max_sectors(q), INT_MAX >> 9) << 9; 1183 } 1184 1185 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1186 { 1187 return q->limits.max_hw_sectors; 1188 } 1189 1190 static inline unsigned short queue_max_segments(const struct request_queue *q) 1191 { 1192 return q->limits.max_segments; 1193 } 1194 1195 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1196 { 1197 return q->limits.max_discard_segments; 1198 } 1199 1200 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1201 { 1202 return q->limits.max_segment_size; 1203 } 1204 1205 static inline unsigned int queue_limits_max_zone_append_sectors(struct queue_limits *l) 1206 { 1207 unsigned int max_sectors = min(l->chunk_sectors, l->max_hw_sectors); 1208 1209 return min_not_zero(l->max_zone_append_sectors, max_sectors); 1210 } 1211 1212 static inline unsigned int queue_max_zone_append_sectors(struct request_queue *q) 1213 { 1214 if (!blk_queue_is_zoned(q)) 1215 return 0; 1216 1217 return queue_limits_max_zone_append_sectors(&q->limits); 1218 } 1219 1220 static inline bool queue_emulates_zone_append(struct request_queue *q) 1221 { 1222 return blk_queue_is_zoned(q) && !q->limits.max_zone_append_sectors; 1223 } 1224 1225 static inline bool bdev_emulates_zone_append(struct block_device *bdev) 1226 { 1227 return queue_emulates_zone_append(bdev_get_queue(bdev)); 1228 } 1229 1230 static inline unsigned int 1231 bdev_max_zone_append_sectors(struct block_device *bdev) 1232 { 1233 return queue_max_zone_append_sectors(bdev_get_queue(bdev)); 1234 } 1235 1236 static inline unsigned int bdev_max_segments(struct block_device *bdev) 1237 { 1238 return queue_max_segments(bdev_get_queue(bdev)); 1239 } 1240 1241 static inline unsigned queue_logical_block_size(const struct request_queue *q) 1242 { 1243 int retval = 512; 1244 1245 if (q && q->limits.logical_block_size) 1246 retval = q->limits.logical_block_size; 1247 1248 return retval; 1249 } 1250 1251 static inline unsigned int bdev_logical_block_size(struct block_device *bdev) 1252 { 1253 return queue_logical_block_size(bdev_get_queue(bdev)); 1254 } 1255 1256 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1257 { 1258 return q->limits.physical_block_size; 1259 } 1260 1261 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1262 { 1263 return queue_physical_block_size(bdev_get_queue(bdev)); 1264 } 1265 1266 static inline unsigned int queue_io_min(const struct request_queue *q) 1267 { 1268 return q->limits.io_min; 1269 } 1270 1271 static inline int bdev_io_min(struct block_device *bdev) 1272 { 1273 return queue_io_min(bdev_get_queue(bdev)); 1274 } 1275 1276 static inline unsigned int queue_io_opt(const struct request_queue *q) 1277 { 1278 return q->limits.io_opt; 1279 } 1280 1281 static inline int bdev_io_opt(struct block_device *bdev) 1282 { 1283 return queue_io_opt(bdev_get_queue(bdev)); 1284 } 1285 1286 static inline unsigned int 1287 queue_zone_write_granularity(const struct request_queue *q) 1288 { 1289 return q->limits.zone_write_granularity; 1290 } 1291 1292 static inline unsigned int 1293 bdev_zone_write_granularity(struct block_device *bdev) 1294 { 1295 return queue_zone_write_granularity(bdev_get_queue(bdev)); 1296 } 1297 1298 int bdev_alignment_offset(struct block_device *bdev); 1299 unsigned int bdev_discard_alignment(struct block_device *bdev); 1300 1301 static inline unsigned int bdev_max_discard_sectors(struct block_device *bdev) 1302 { 1303 return bdev_get_queue(bdev)->limits.max_discard_sectors; 1304 } 1305 1306 static inline unsigned int bdev_discard_granularity(struct block_device *bdev) 1307 { 1308 return bdev_get_queue(bdev)->limits.discard_granularity; 1309 } 1310 1311 static inline unsigned int 1312 bdev_max_secure_erase_sectors(struct block_device *bdev) 1313 { 1314 return bdev_get_queue(bdev)->limits.max_secure_erase_sectors; 1315 } 1316 1317 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1318 { 1319 struct request_queue *q = bdev_get_queue(bdev); 1320 1321 if (q) 1322 return q->limits.max_write_zeroes_sectors; 1323 1324 return 0; 1325 } 1326 1327 static inline bool bdev_nonrot(struct block_device *bdev) 1328 { 1329 return blk_queue_nonrot(bdev_get_queue(bdev)); 1330 } 1331 1332 static inline bool bdev_synchronous(struct block_device *bdev) 1333 { 1334 return bdev->bd_disk->queue->limits.features & BLK_FEAT_SYNCHRONOUS; 1335 } 1336 1337 static inline bool bdev_stable_writes(struct block_device *bdev) 1338 { 1339 struct request_queue *q = bdev_get_queue(bdev); 1340 1341 if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY) && 1342 q->limits.integrity.csum_type != BLK_INTEGRITY_CSUM_NONE) 1343 return true; 1344 return q->limits.features & BLK_FEAT_STABLE_WRITES; 1345 } 1346 1347 static inline bool blk_queue_write_cache(struct request_queue *q) 1348 { 1349 return (q->limits.features & BLK_FEAT_WRITE_CACHE) && 1350 !(q->limits.flags & BLK_FLAGS_WRITE_CACHE_DISABLED); 1351 } 1352 1353 static inline bool bdev_write_cache(struct block_device *bdev) 1354 { 1355 return blk_queue_write_cache(bdev_get_queue(bdev)); 1356 } 1357 1358 static inline bool bdev_fua(struct block_device *bdev) 1359 { 1360 return bdev_get_queue(bdev)->limits.features & BLK_FEAT_FUA; 1361 } 1362 1363 static inline bool bdev_nowait(struct block_device *bdev) 1364 { 1365 return bdev->bd_disk->queue->limits.features & BLK_FEAT_NOWAIT; 1366 } 1367 1368 static inline bool bdev_is_zoned(struct block_device *bdev) 1369 { 1370 return blk_queue_is_zoned(bdev_get_queue(bdev)); 1371 } 1372 1373 static inline unsigned int bdev_zone_no(struct block_device *bdev, sector_t sec) 1374 { 1375 return disk_zone_no(bdev->bd_disk, sec); 1376 } 1377 1378 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1379 { 1380 struct request_queue *q = bdev_get_queue(bdev); 1381 1382 if (!blk_queue_is_zoned(q)) 1383 return 0; 1384 return q->limits.chunk_sectors; 1385 } 1386 1387 static inline sector_t bdev_offset_from_zone_start(struct block_device *bdev, 1388 sector_t sector) 1389 { 1390 return sector & (bdev_zone_sectors(bdev) - 1); 1391 } 1392 1393 static inline sector_t bio_offset_from_zone_start(struct bio *bio) 1394 { 1395 return bdev_offset_from_zone_start(bio->bi_bdev, 1396 bio->bi_iter.bi_sector); 1397 } 1398 1399 static inline bool bdev_is_zone_start(struct block_device *bdev, 1400 sector_t sector) 1401 { 1402 return bdev_offset_from_zone_start(bdev, sector) == 0; 1403 } 1404 1405 static inline int queue_dma_alignment(const struct request_queue *q) 1406 { 1407 return q ? q->limits.dma_alignment : 511; 1408 } 1409 1410 static inline unsigned int bdev_dma_alignment(struct block_device *bdev) 1411 { 1412 return queue_dma_alignment(bdev_get_queue(bdev)); 1413 } 1414 1415 static inline bool bdev_iter_is_aligned(struct block_device *bdev, 1416 struct iov_iter *iter) 1417 { 1418 return iov_iter_is_aligned(iter, bdev_dma_alignment(bdev), 1419 bdev_logical_block_size(bdev) - 1); 1420 } 1421 1422 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1423 unsigned int len) 1424 { 1425 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1426 return !(addr & alignment) && !(len & alignment); 1427 } 1428 1429 /* assumes size > 256 */ 1430 static inline unsigned int blksize_bits(unsigned int size) 1431 { 1432 return order_base_2(size >> SECTOR_SHIFT) + SECTOR_SHIFT; 1433 } 1434 1435 int kblockd_schedule_work(struct work_struct *work); 1436 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1437 1438 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1439 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1440 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1441 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1442 1443 #ifdef CONFIG_BLK_INLINE_ENCRYPTION 1444 1445 bool blk_crypto_register(struct blk_crypto_profile *profile, 1446 struct request_queue *q); 1447 1448 #else /* CONFIG_BLK_INLINE_ENCRYPTION */ 1449 1450 static inline bool blk_crypto_register(struct blk_crypto_profile *profile, 1451 struct request_queue *q) 1452 { 1453 return true; 1454 } 1455 1456 #endif /* CONFIG_BLK_INLINE_ENCRYPTION */ 1457 1458 enum blk_unique_id { 1459 /* these match the Designator Types specified in SPC */ 1460 BLK_UID_T10 = 1, 1461 BLK_UID_EUI64 = 2, 1462 BLK_UID_NAA = 3, 1463 }; 1464 1465 struct block_device_operations { 1466 void (*submit_bio)(struct bio *bio); 1467 int (*poll_bio)(struct bio *bio, struct io_comp_batch *iob, 1468 unsigned int flags); 1469 int (*open)(struct gendisk *disk, blk_mode_t mode); 1470 void (*release)(struct gendisk *disk); 1471 int (*ioctl)(struct block_device *bdev, blk_mode_t mode, 1472 unsigned cmd, unsigned long arg); 1473 int (*compat_ioctl)(struct block_device *bdev, blk_mode_t mode, 1474 unsigned cmd, unsigned long arg); 1475 unsigned int (*check_events) (struct gendisk *disk, 1476 unsigned int clearing); 1477 void (*unlock_native_capacity) (struct gendisk *); 1478 int (*getgeo)(struct block_device *, struct hd_geometry *); 1479 int (*set_read_only)(struct block_device *bdev, bool ro); 1480 void (*free_disk)(struct gendisk *disk); 1481 /* this callback is with swap_lock and sometimes page table lock held */ 1482 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1483 int (*report_zones)(struct gendisk *, sector_t sector, 1484 unsigned int nr_zones, report_zones_cb cb, void *data); 1485 char *(*devnode)(struct gendisk *disk, umode_t *mode); 1486 /* returns the length of the identifier or a negative errno: */ 1487 int (*get_unique_id)(struct gendisk *disk, u8 id[16], 1488 enum blk_unique_id id_type); 1489 struct module *owner; 1490 const struct pr_ops *pr_ops; 1491 1492 /* 1493 * Special callback for probing GPT entry at a given sector. 1494 * Needed by Android devices, used by GPT scanner and MMC blk 1495 * driver. 1496 */ 1497 int (*alternative_gpt_sector)(struct gendisk *disk, sector_t *sector); 1498 }; 1499 1500 #ifdef CONFIG_COMPAT 1501 extern int blkdev_compat_ptr_ioctl(struct block_device *, blk_mode_t, 1502 unsigned int, unsigned long); 1503 #else 1504 #define blkdev_compat_ptr_ioctl NULL 1505 #endif 1506 1507 static inline void blk_wake_io_task(struct task_struct *waiter) 1508 { 1509 /* 1510 * If we're polling, the task itself is doing the completions. For 1511 * that case, we don't need to signal a wakeup, it's enough to just 1512 * mark us as RUNNING. 1513 */ 1514 if (waiter == current) 1515 __set_current_state(TASK_RUNNING); 1516 else 1517 wake_up_process(waiter); 1518 } 1519 1520 unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, 1521 unsigned long start_time); 1522 void bdev_end_io_acct(struct block_device *bdev, enum req_op op, 1523 unsigned int sectors, unsigned long start_time); 1524 1525 unsigned long bio_start_io_acct(struct bio *bio); 1526 void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, 1527 struct block_device *orig_bdev); 1528 1529 /** 1530 * bio_end_io_acct - end I/O accounting for bio based drivers 1531 * @bio: bio to end account for 1532 * @start_time: start time returned by bio_start_io_acct() 1533 */ 1534 static inline void bio_end_io_acct(struct bio *bio, unsigned long start_time) 1535 { 1536 return bio_end_io_acct_remapped(bio, start_time, bio->bi_bdev); 1537 } 1538 1539 int bdev_read_only(struct block_device *bdev); 1540 int set_blocksize(struct file *file, int size); 1541 1542 int lookup_bdev(const char *pathname, dev_t *dev); 1543 1544 void blkdev_show(struct seq_file *seqf, off_t offset); 1545 1546 #define BDEVNAME_SIZE 32 /* Largest string for a blockdev identifier */ 1547 #define BDEVT_SIZE 10 /* Largest string for MAJ:MIN for blkdev */ 1548 #ifdef CONFIG_BLOCK 1549 #define BLKDEV_MAJOR_MAX 512 1550 #else 1551 #define BLKDEV_MAJOR_MAX 0 1552 #endif 1553 1554 struct blk_holder_ops { 1555 void (*mark_dead)(struct block_device *bdev, bool surprise); 1556 1557 /* 1558 * Sync the file system mounted on the block device. 1559 */ 1560 void (*sync)(struct block_device *bdev); 1561 1562 /* 1563 * Freeze the file system mounted on the block device. 1564 */ 1565 int (*freeze)(struct block_device *bdev); 1566 1567 /* 1568 * Thaw the file system mounted on the block device. 1569 */ 1570 int (*thaw)(struct block_device *bdev); 1571 }; 1572 1573 /* 1574 * For filesystems using @fs_holder_ops, the @holder argument passed to 1575 * helpers used to open and claim block devices via 1576 * bd_prepare_to_claim() must point to a superblock. 1577 */ 1578 extern const struct blk_holder_ops fs_holder_ops; 1579 1580 /* 1581 * Return the correct open flags for blkdev_get_by_* for super block flags 1582 * as stored in sb->s_flags. 1583 */ 1584 #define sb_open_mode(flags) \ 1585 (BLK_OPEN_READ | BLK_OPEN_RESTRICT_WRITES | \ 1586 (((flags) & SB_RDONLY) ? 0 : BLK_OPEN_WRITE)) 1587 1588 struct file *bdev_file_open_by_dev(dev_t dev, blk_mode_t mode, void *holder, 1589 const struct blk_holder_ops *hops); 1590 struct file *bdev_file_open_by_path(const char *path, blk_mode_t mode, 1591 void *holder, const struct blk_holder_ops *hops); 1592 int bd_prepare_to_claim(struct block_device *bdev, void *holder, 1593 const struct blk_holder_ops *hops); 1594 void bd_abort_claiming(struct block_device *bdev, void *holder); 1595 1596 /* just for blk-cgroup, don't use elsewhere */ 1597 struct block_device *blkdev_get_no_open(dev_t dev); 1598 void blkdev_put_no_open(struct block_device *bdev); 1599 1600 struct block_device *I_BDEV(struct inode *inode); 1601 struct block_device *file_bdev(struct file *bdev_file); 1602 bool disk_live(struct gendisk *disk); 1603 unsigned int block_size(struct block_device *bdev); 1604 1605 #ifdef CONFIG_BLOCK 1606 void invalidate_bdev(struct block_device *bdev); 1607 int sync_blockdev(struct block_device *bdev); 1608 int sync_blockdev_range(struct block_device *bdev, loff_t lstart, loff_t lend); 1609 int sync_blockdev_nowait(struct block_device *bdev); 1610 void sync_bdevs(bool wait); 1611 void bdev_statx_dioalign(struct inode *inode, struct kstat *stat); 1612 void printk_all_partitions(void); 1613 int __init early_lookup_bdev(const char *pathname, dev_t *dev); 1614 #else 1615 static inline void invalidate_bdev(struct block_device *bdev) 1616 { 1617 } 1618 static inline int sync_blockdev(struct block_device *bdev) 1619 { 1620 return 0; 1621 } 1622 static inline int sync_blockdev_nowait(struct block_device *bdev) 1623 { 1624 return 0; 1625 } 1626 static inline void sync_bdevs(bool wait) 1627 { 1628 } 1629 static inline void bdev_statx_dioalign(struct inode *inode, struct kstat *stat) 1630 { 1631 } 1632 static inline void printk_all_partitions(void) 1633 { 1634 } 1635 static inline int early_lookup_bdev(const char *pathname, dev_t *dev) 1636 { 1637 return -EINVAL; 1638 } 1639 #endif /* CONFIG_BLOCK */ 1640 1641 int bdev_freeze(struct block_device *bdev); 1642 int bdev_thaw(struct block_device *bdev); 1643 void bdev_fput(struct file *bdev_file); 1644 1645 struct io_comp_batch { 1646 struct request *req_list; 1647 bool need_ts; 1648 void (*complete)(struct io_comp_batch *); 1649 }; 1650 1651 #define DEFINE_IO_COMP_BATCH(name) struct io_comp_batch name = { } 1652 1653 #endif /* _LINUX_BLKDEV_H */ 1654