1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_BLKDEV_H 3 #define _LINUX_BLKDEV_H 4 5 #include <linux/sched.h> 6 #include <linux/sched/clock.h> 7 8 #ifdef CONFIG_BLOCK 9 10 #include <linux/major.h> 11 #include <linux/genhd.h> 12 #include <linux/list.h> 13 #include <linux/llist.h> 14 #include <linux/timer.h> 15 #include <linux/workqueue.h> 16 #include <linux/pagemap.h> 17 #include <linux/backing-dev-defs.h> 18 #include <linux/wait.h> 19 #include <linux/mempool.h> 20 #include <linux/pfn.h> 21 #include <linux/bio.h> 22 #include <linux/stringify.h> 23 #include <linux/gfp.h> 24 #include <linux/bsg.h> 25 #include <linux/smp.h> 26 #include <linux/rcupdate.h> 27 #include <linux/percpu-refcount.h> 28 #include <linux/scatterlist.h> 29 #include <linux/blkzoned.h> 30 31 struct module; 32 struct scsi_ioctl_command; 33 34 struct request_queue; 35 struct elevator_queue; 36 struct blk_trace; 37 struct request; 38 struct sg_io_hdr; 39 struct bsg_job; 40 struct blkcg_gq; 41 struct blk_flush_queue; 42 struct pr_ops; 43 struct rq_qos; 44 struct blk_queue_stats; 45 struct blk_stat_callback; 46 47 #define BLKDEV_MIN_RQ 4 48 #define BLKDEV_MAX_RQ 128 /* Default maximum */ 49 50 /* Must be consistent with blk_mq_poll_stats_bkt() */ 51 #define BLK_MQ_POLL_STATS_BKTS 16 52 53 /* Doing classic polling */ 54 #define BLK_MQ_POLL_CLASSIC -1 55 56 /* 57 * Maximum number of blkcg policies allowed to be registered concurrently. 58 * Defined here to simplify include dependency. 59 */ 60 #define BLKCG_MAX_POLS 5 61 62 typedef void (rq_end_io_fn)(struct request *, blk_status_t); 63 64 /* 65 * request flags */ 66 typedef __u32 __bitwise req_flags_t; 67 68 /* elevator knows about this request */ 69 #define RQF_SORTED ((__force req_flags_t)(1 << 0)) 70 /* drive already may have started this one */ 71 #define RQF_STARTED ((__force req_flags_t)(1 << 1)) 72 /* may not be passed by ioscheduler */ 73 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3)) 74 /* request for flush sequence */ 75 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4)) 76 /* merge of different types, fail separately */ 77 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5)) 78 /* track inflight for MQ */ 79 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6)) 80 /* don't call prep for this one */ 81 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7)) 82 /* set for "ide_preempt" requests and also for requests for which the SCSI 83 "quiesce" state must be ignored. */ 84 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8)) 85 /* contains copies of user pages */ 86 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9)) 87 /* vaguely specified driver internal error. Ignored by the block layer */ 88 #define RQF_FAILED ((__force req_flags_t)(1 << 10)) 89 /* don't warn about errors */ 90 #define RQF_QUIET ((__force req_flags_t)(1 << 11)) 91 /* elevator private data attached */ 92 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12)) 93 /* account into disk and partition IO statistics */ 94 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13)) 95 /* request came from our alloc pool */ 96 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14)) 97 /* runtime pm request */ 98 #define RQF_PM ((__force req_flags_t)(1 << 15)) 99 /* on IO scheduler merge hash */ 100 #define RQF_HASHED ((__force req_flags_t)(1 << 16)) 101 /* track IO completion time */ 102 #define RQF_STATS ((__force req_flags_t)(1 << 17)) 103 /* Look at ->special_vec for the actual data payload instead of the 104 bio chain. */ 105 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18)) 106 /* The per-zone write lock is held for this request */ 107 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) 108 /* already slept for hybrid poll */ 109 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20)) 110 /* ->timeout has been called, don't expire again */ 111 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21)) 112 113 /* flags that prevent us from merging requests: */ 114 #define RQF_NOMERGE_FLAGS \ 115 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD) 116 117 /* 118 * Request state for blk-mq. 119 */ 120 enum mq_rq_state { 121 MQ_RQ_IDLE = 0, 122 MQ_RQ_IN_FLIGHT = 1, 123 MQ_RQ_COMPLETE = 2, 124 }; 125 126 /* 127 * Try to put the fields that are referenced together in the same cacheline. 128 * 129 * If you modify this structure, make sure to update blk_rq_init() and 130 * especially blk_mq_rq_ctx_init() to take care of the added fields. 131 */ 132 struct request { 133 struct request_queue *q; 134 struct blk_mq_ctx *mq_ctx; 135 struct blk_mq_hw_ctx *mq_hctx; 136 137 unsigned int cmd_flags; /* op and common flags */ 138 req_flags_t rq_flags; 139 140 int tag; 141 int internal_tag; 142 143 /* the following two fields are internal, NEVER access directly */ 144 unsigned int __data_len; /* total data len */ 145 sector_t __sector; /* sector cursor */ 146 147 struct bio *bio; 148 struct bio *biotail; 149 150 struct list_head queuelist; 151 152 /* 153 * The hash is used inside the scheduler, and killed once the 154 * request reaches the dispatch list. The ipi_list is only used 155 * to queue the request for softirq completion, which is long 156 * after the request has been unhashed (and even removed from 157 * the dispatch list). 158 */ 159 union { 160 struct hlist_node hash; /* merge hash */ 161 struct list_head ipi_list; 162 }; 163 164 /* 165 * The rb_node is only used inside the io scheduler, requests 166 * are pruned when moved to the dispatch queue. So let the 167 * completion_data share space with the rb_node. 168 */ 169 union { 170 struct rb_node rb_node; /* sort/lookup */ 171 struct bio_vec special_vec; 172 void *completion_data; 173 int error_count; /* for legacy drivers, don't use */ 174 }; 175 176 /* 177 * Three pointers are available for the IO schedulers, if they need 178 * more they have to dynamically allocate it. Flush requests are 179 * never put on the IO scheduler. So let the flush fields share 180 * space with the elevator data. 181 */ 182 union { 183 struct { 184 struct io_cq *icq; 185 void *priv[2]; 186 } elv; 187 188 struct { 189 unsigned int seq; 190 struct list_head list; 191 rq_end_io_fn *saved_end_io; 192 } flush; 193 }; 194 195 struct gendisk *rq_disk; 196 struct hd_struct *part; 197 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 198 /* Time that the first bio started allocating this request. */ 199 u64 alloc_time_ns; 200 #endif 201 /* Time that this request was allocated for this IO. */ 202 u64 start_time_ns; 203 /* Time that I/O was submitted to the device. */ 204 u64 io_start_time_ns; 205 206 #ifdef CONFIG_BLK_WBT 207 unsigned short wbt_flags; 208 #endif 209 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW 210 unsigned short throtl_size; 211 #endif 212 213 /* 214 * Number of scatter-gather DMA addr+len pairs after 215 * physical address coalescing is performed. 216 */ 217 unsigned short nr_phys_segments; 218 219 #if defined(CONFIG_BLK_DEV_INTEGRITY) 220 unsigned short nr_integrity_segments; 221 #endif 222 223 unsigned short write_hint; 224 unsigned short ioprio; 225 226 unsigned int extra_len; /* length of alignment and padding */ 227 228 enum mq_rq_state state; 229 refcount_t ref; 230 231 unsigned int timeout; 232 unsigned long deadline; 233 234 union { 235 struct __call_single_data csd; 236 u64 fifo_time; 237 }; 238 239 /* 240 * completion callback. 241 */ 242 rq_end_io_fn *end_io; 243 void *end_io_data; 244 }; 245 246 static inline bool blk_op_is_scsi(unsigned int op) 247 { 248 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT; 249 } 250 251 static inline bool blk_op_is_private(unsigned int op) 252 { 253 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 254 } 255 256 static inline bool blk_rq_is_scsi(struct request *rq) 257 { 258 return blk_op_is_scsi(req_op(rq)); 259 } 260 261 static inline bool blk_rq_is_private(struct request *rq) 262 { 263 return blk_op_is_private(req_op(rq)); 264 } 265 266 static inline bool blk_rq_is_passthrough(struct request *rq) 267 { 268 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq); 269 } 270 271 static inline bool bio_is_passthrough(struct bio *bio) 272 { 273 unsigned op = bio_op(bio); 274 275 return blk_op_is_scsi(op) || blk_op_is_private(op); 276 } 277 278 static inline unsigned short req_get_ioprio(struct request *req) 279 { 280 return req->ioprio; 281 } 282 283 #include <linux/elevator.h> 284 285 struct blk_queue_ctx; 286 287 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio); 288 289 struct bio_vec; 290 typedef int (dma_drain_needed_fn)(struct request *); 291 292 enum blk_eh_timer_return { 293 BLK_EH_DONE, /* drivers has completed the command */ 294 BLK_EH_RESET_TIMER, /* reset timer and try again */ 295 }; 296 297 enum blk_queue_state { 298 Queue_down, 299 Queue_up, 300 }; 301 302 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ 303 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ 304 305 #define BLK_SCSI_MAX_CMDS (256) 306 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) 307 308 /* 309 * Zoned block device models (zoned limit). 310 */ 311 enum blk_zoned_model { 312 BLK_ZONED_NONE, /* Regular block device */ 313 BLK_ZONED_HA, /* Host-aware zoned block device */ 314 BLK_ZONED_HM, /* Host-managed zoned block device */ 315 }; 316 317 struct queue_limits { 318 unsigned long bounce_pfn; 319 unsigned long seg_boundary_mask; 320 unsigned long virt_boundary_mask; 321 322 unsigned int max_hw_sectors; 323 unsigned int max_dev_sectors; 324 unsigned int chunk_sectors; 325 unsigned int max_sectors; 326 unsigned int max_segment_size; 327 unsigned int physical_block_size; 328 unsigned int alignment_offset; 329 unsigned int io_min; 330 unsigned int io_opt; 331 unsigned int max_discard_sectors; 332 unsigned int max_hw_discard_sectors; 333 unsigned int max_write_same_sectors; 334 unsigned int max_write_zeroes_sectors; 335 unsigned int discard_granularity; 336 unsigned int discard_alignment; 337 338 unsigned short logical_block_size; 339 unsigned short max_segments; 340 unsigned short max_integrity_segments; 341 unsigned short max_discard_segments; 342 343 unsigned char misaligned; 344 unsigned char discard_misaligned; 345 unsigned char raid_partial_stripes_expensive; 346 enum blk_zoned_model zoned; 347 }; 348 349 #ifdef CONFIG_BLK_DEV_ZONED 350 351 /* 352 * Maximum number of zones to report with a single report zones command. 353 */ 354 #define BLK_ZONED_REPORT_MAX_ZONES 8192U 355 356 extern unsigned int blkdev_nr_zones(struct block_device *bdev); 357 extern int blkdev_report_zones(struct block_device *bdev, 358 sector_t sector, struct blk_zone *zones, 359 unsigned int *nr_zones); 360 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors, 361 sector_t nr_sectors, gfp_t gfp_mask); 362 extern int blk_revalidate_disk_zones(struct gendisk *disk); 363 364 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, 365 unsigned int cmd, unsigned long arg); 366 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode, 367 unsigned int cmd, unsigned long arg); 368 369 #else /* CONFIG_BLK_DEV_ZONED */ 370 371 static inline unsigned int blkdev_nr_zones(struct block_device *bdev) 372 { 373 return 0; 374 } 375 376 static inline int blk_revalidate_disk_zones(struct gendisk *disk) 377 { 378 return 0; 379 } 380 381 static inline int blkdev_report_zones_ioctl(struct block_device *bdev, 382 fmode_t mode, unsigned int cmd, 383 unsigned long arg) 384 { 385 return -ENOTTY; 386 } 387 388 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev, 389 fmode_t mode, unsigned int cmd, 390 unsigned long arg) 391 { 392 return -ENOTTY; 393 } 394 395 #endif /* CONFIG_BLK_DEV_ZONED */ 396 397 struct request_queue { 398 struct request *last_merge; 399 struct elevator_queue *elevator; 400 401 struct blk_queue_stats *stats; 402 struct rq_qos *rq_qos; 403 404 make_request_fn *make_request_fn; 405 dma_drain_needed_fn *dma_drain_needed; 406 407 const struct blk_mq_ops *mq_ops; 408 409 /* sw queues */ 410 struct blk_mq_ctx __percpu *queue_ctx; 411 unsigned int nr_queues; 412 413 unsigned int queue_depth; 414 415 /* hw dispatch queues */ 416 struct blk_mq_hw_ctx **queue_hw_ctx; 417 unsigned int nr_hw_queues; 418 419 struct backing_dev_info *backing_dev_info; 420 421 /* 422 * The queue owner gets to use this for whatever they like. 423 * ll_rw_blk doesn't touch it. 424 */ 425 void *queuedata; 426 427 /* 428 * various queue flags, see QUEUE_* below 429 */ 430 unsigned long queue_flags; 431 /* 432 * Number of contexts that have called blk_set_pm_only(). If this 433 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are 434 * processed. 435 */ 436 atomic_t pm_only; 437 438 /* 439 * ida allocated id for this queue. Used to index queues from 440 * ioctx. 441 */ 442 int id; 443 444 /* 445 * queue needs bounce pages for pages above this limit 446 */ 447 gfp_t bounce_gfp; 448 449 spinlock_t queue_lock; 450 451 /* 452 * queue kobject 453 */ 454 struct kobject kobj; 455 456 /* 457 * mq queue kobject 458 */ 459 struct kobject *mq_kobj; 460 461 #ifdef CONFIG_BLK_DEV_INTEGRITY 462 struct blk_integrity integrity; 463 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 464 465 #ifdef CONFIG_PM 466 struct device *dev; 467 int rpm_status; 468 unsigned int nr_pending; 469 #endif 470 471 /* 472 * queue settings 473 */ 474 unsigned long nr_requests; /* Max # of requests */ 475 476 unsigned int dma_drain_size; 477 void *dma_drain_buffer; 478 unsigned int dma_pad_mask; 479 unsigned int dma_alignment; 480 481 unsigned int rq_timeout; 482 int poll_nsec; 483 484 struct blk_stat_callback *poll_cb; 485 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS]; 486 487 struct timer_list timeout; 488 struct work_struct timeout_work; 489 490 struct list_head icq_list; 491 #ifdef CONFIG_BLK_CGROUP 492 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 493 struct blkcg_gq *root_blkg; 494 struct list_head blkg_list; 495 #endif 496 497 struct queue_limits limits; 498 499 unsigned int required_elevator_features; 500 501 #ifdef CONFIG_BLK_DEV_ZONED 502 /* 503 * Zoned block device information for request dispatch control. 504 * nr_zones is the total number of zones of the device. This is always 505 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones 506 * bits which indicates if a zone is conventional (bit clear) or 507 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones 508 * bits which indicates if a zone is write locked, that is, if a write 509 * request targeting the zone was dispatched. All three fields are 510 * initialized by the low level device driver (e.g. scsi/sd.c). 511 * Stacking drivers (device mappers) may or may not initialize 512 * these fields. 513 * 514 * Reads of this information must be protected with blk_queue_enter() / 515 * blk_queue_exit(). Modifying this information is only allowed while 516 * no requests are being processed. See also blk_mq_freeze_queue() and 517 * blk_mq_unfreeze_queue(). 518 */ 519 unsigned int nr_zones; 520 unsigned long *seq_zones_bitmap; 521 unsigned long *seq_zones_wlock; 522 #endif /* CONFIG_BLK_DEV_ZONED */ 523 524 /* 525 * sg stuff 526 */ 527 unsigned int sg_timeout; 528 unsigned int sg_reserved_size; 529 int node; 530 #ifdef CONFIG_BLK_DEV_IO_TRACE 531 struct blk_trace *blk_trace; 532 struct mutex blk_trace_mutex; 533 #endif 534 /* 535 * for flush operations 536 */ 537 struct blk_flush_queue *fq; 538 539 struct list_head requeue_list; 540 spinlock_t requeue_lock; 541 struct delayed_work requeue_work; 542 543 struct mutex sysfs_lock; 544 struct mutex sysfs_dir_lock; 545 546 /* 547 * for reusing dead hctx instance in case of updating 548 * nr_hw_queues 549 */ 550 struct list_head unused_hctx_list; 551 spinlock_t unused_hctx_lock; 552 553 int mq_freeze_depth; 554 555 #if defined(CONFIG_BLK_DEV_BSG) 556 struct bsg_class_device bsg_dev; 557 #endif 558 559 #ifdef CONFIG_BLK_DEV_THROTTLING 560 /* Throttle data */ 561 struct throtl_data *td; 562 #endif 563 struct rcu_head rcu_head; 564 wait_queue_head_t mq_freeze_wq; 565 /* 566 * Protect concurrent access to q_usage_counter by 567 * percpu_ref_kill() and percpu_ref_reinit(). 568 */ 569 struct mutex mq_freeze_lock; 570 struct percpu_ref q_usage_counter; 571 572 struct blk_mq_tag_set *tag_set; 573 struct list_head tag_set_list; 574 struct bio_set bio_split; 575 576 #ifdef CONFIG_BLK_DEBUG_FS 577 struct dentry *debugfs_dir; 578 struct dentry *sched_debugfs_dir; 579 struct dentry *rqos_debugfs_dir; 580 #endif 581 582 bool mq_sysfs_init_done; 583 584 size_t cmd_size; 585 586 struct work_struct release_work; 587 588 #define BLK_MAX_WRITE_HINTS 5 589 u64 write_hints[BLK_MAX_WRITE_HINTS]; 590 }; 591 592 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */ 593 #define QUEUE_FLAG_DYING 1 /* queue being torn down */ 594 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */ 595 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ 596 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ 597 #define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */ 598 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ 599 #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */ 600 #define QUEUE_FLAG_DISCARD 8 /* supports DISCARD */ 601 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ 602 #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */ 603 #define QUEUE_FLAG_SECERASE 11 /* supports secure erase */ 604 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ 605 #define QUEUE_FLAG_DEAD 13 /* queue tear-down finished */ 606 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ 607 #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ 608 #define QUEUE_FLAG_WC 17 /* Write back caching */ 609 #define QUEUE_FLAG_FUA 18 /* device supports FUA writes */ 610 #define QUEUE_FLAG_DAX 19 /* device supports DAX */ 611 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ 612 #define QUEUE_FLAG_POLL_STATS 21 /* collecting stats for hybrid polling */ 613 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ 614 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23 /* queue supports SCSI commands */ 615 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ 616 #define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */ 617 #define QUEUE_FLAG_ZONE_RESETALL 26 /* supports Zone Reset All */ 618 #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ 619 620 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 621 (1 << QUEUE_FLAG_SAME_COMP)) 622 623 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 624 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 625 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); 626 627 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 628 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 629 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) 630 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 631 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 632 #define blk_queue_noxmerges(q) \ 633 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 634 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) 635 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) 636 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) 637 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) 638 #define blk_queue_zone_resetall(q) \ 639 test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags) 640 #define blk_queue_secure_erase(q) \ 641 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags)) 642 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) 643 #define blk_queue_scsi_passthrough(q) \ 644 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags) 645 #define blk_queue_pci_p2pdma(q) \ 646 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags) 647 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 648 #define blk_queue_rq_alloc_time(q) \ 649 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) 650 #else 651 #define blk_queue_rq_alloc_time(q) false 652 #endif 653 654 #define blk_noretry_request(rq) \ 655 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 656 REQ_FAILFAST_DRIVER)) 657 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 658 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 659 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags) 660 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) 661 662 extern void blk_set_pm_only(struct request_queue *q); 663 extern void blk_clear_pm_only(struct request_queue *q); 664 665 static inline bool blk_account_rq(struct request *rq) 666 { 667 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq); 668 } 669 670 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 671 672 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ) 673 674 #define rq_dma_dir(rq) \ 675 (op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE) 676 677 #define dma_map_bvec(dev, bv, dir, attrs) \ 678 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 679 (dir), (attrs)) 680 681 static inline bool queue_is_mq(struct request_queue *q) 682 { 683 return q->mq_ops; 684 } 685 686 static inline enum blk_zoned_model 687 blk_queue_zoned_model(struct request_queue *q) 688 { 689 return q->limits.zoned; 690 } 691 692 static inline bool blk_queue_is_zoned(struct request_queue *q) 693 { 694 switch (blk_queue_zoned_model(q)) { 695 case BLK_ZONED_HA: 696 case BLK_ZONED_HM: 697 return true; 698 default: 699 return false; 700 } 701 } 702 703 static inline sector_t blk_queue_zone_sectors(struct request_queue *q) 704 { 705 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; 706 } 707 708 #ifdef CONFIG_BLK_DEV_ZONED 709 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 710 { 711 return blk_queue_is_zoned(q) ? q->nr_zones : 0; 712 } 713 714 static inline unsigned int blk_queue_zone_no(struct request_queue *q, 715 sector_t sector) 716 { 717 if (!blk_queue_is_zoned(q)) 718 return 0; 719 return sector >> ilog2(q->limits.chunk_sectors); 720 } 721 722 static inline bool blk_queue_zone_is_seq(struct request_queue *q, 723 sector_t sector) 724 { 725 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap) 726 return false; 727 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap); 728 } 729 #else /* CONFIG_BLK_DEV_ZONED */ 730 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 731 { 732 return 0; 733 } 734 #endif /* CONFIG_BLK_DEV_ZONED */ 735 736 static inline bool rq_is_sync(struct request *rq) 737 { 738 return op_is_sync(rq->cmd_flags); 739 } 740 741 static inline bool rq_mergeable(struct request *rq) 742 { 743 if (blk_rq_is_passthrough(rq)) 744 return false; 745 746 if (req_op(rq) == REQ_OP_FLUSH) 747 return false; 748 749 if (req_op(rq) == REQ_OP_WRITE_ZEROES) 750 return false; 751 752 if (rq->cmd_flags & REQ_NOMERGE_FLAGS) 753 return false; 754 if (rq->rq_flags & RQF_NOMERGE_FLAGS) 755 return false; 756 757 return true; 758 } 759 760 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) 761 { 762 if (bio_page(a) == bio_page(b) && 763 bio_offset(a) == bio_offset(b)) 764 return true; 765 766 return false; 767 } 768 769 static inline unsigned int blk_queue_depth(struct request_queue *q) 770 { 771 if (q->queue_depth) 772 return q->queue_depth; 773 774 return q->nr_requests; 775 } 776 777 extern unsigned long blk_max_low_pfn, blk_max_pfn; 778 779 /* 780 * standard bounce addresses: 781 * 782 * BLK_BOUNCE_HIGH : bounce all highmem pages 783 * BLK_BOUNCE_ANY : don't bounce anything 784 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary 785 */ 786 787 #if BITS_PER_LONG == 32 788 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) 789 #else 790 #define BLK_BOUNCE_HIGH -1ULL 791 #endif 792 #define BLK_BOUNCE_ANY (-1ULL) 793 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) 794 795 /* 796 * default timeout for SG_IO if none specified 797 */ 798 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 799 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 800 801 struct rq_map_data { 802 struct page **pages; 803 int page_order; 804 int nr_entries; 805 unsigned long offset; 806 int null_mapped; 807 int from_user; 808 }; 809 810 struct req_iterator { 811 struct bvec_iter iter; 812 struct bio *bio; 813 }; 814 815 /* This should not be used directly - use rq_for_each_segment */ 816 #define for_each_bio(_bio) \ 817 for (; _bio; _bio = _bio->bi_next) 818 #define __rq_for_each_bio(_bio, rq) \ 819 if ((rq->bio)) \ 820 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) 821 822 #define rq_for_each_segment(bvl, _rq, _iter) \ 823 __rq_for_each_bio(_iter.bio, _rq) \ 824 bio_for_each_segment(bvl, _iter.bio, _iter.iter) 825 826 #define rq_for_each_bvec(bvl, _rq, _iter) \ 827 __rq_for_each_bio(_iter.bio, _rq) \ 828 bio_for_each_bvec(bvl, _iter.bio, _iter.iter) 829 830 #define rq_iter_last(bvec, _iter) \ 831 (_iter.bio->bi_next == NULL && \ 832 bio_iter_last(bvec, _iter.iter)) 833 834 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 835 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 836 #endif 837 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 838 extern void rq_flush_dcache_pages(struct request *rq); 839 #else 840 static inline void rq_flush_dcache_pages(struct request *rq) 841 { 842 } 843 #endif 844 845 extern int blk_register_queue(struct gendisk *disk); 846 extern void blk_unregister_queue(struct gendisk *disk); 847 extern blk_qc_t generic_make_request(struct bio *bio); 848 extern blk_qc_t direct_make_request(struct bio *bio); 849 extern void blk_rq_init(struct request_queue *q, struct request *rq); 850 extern void blk_put_request(struct request *); 851 extern struct request *blk_get_request(struct request_queue *, unsigned int op, 852 blk_mq_req_flags_t flags); 853 extern int blk_lld_busy(struct request_queue *q); 854 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, 855 struct bio_set *bs, gfp_t gfp_mask, 856 int (*bio_ctr)(struct bio *, struct bio *, void *), 857 void *data); 858 extern void blk_rq_unprep_clone(struct request *rq); 859 extern blk_status_t blk_insert_cloned_request(struct request_queue *q, 860 struct request *rq); 861 extern int blk_rq_append_bio(struct request *rq, struct bio **bio); 862 extern void blk_queue_split(struct request_queue *, struct bio **); 863 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); 864 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, 865 unsigned int, void __user *); 866 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, 867 unsigned int, void __user *); 868 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, 869 struct scsi_ioctl_command __user *); 870 871 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 872 extern void blk_queue_exit(struct request_queue *q); 873 extern void blk_sync_queue(struct request_queue *q); 874 extern int blk_rq_map_user(struct request_queue *, struct request *, 875 struct rq_map_data *, void __user *, unsigned long, 876 gfp_t); 877 extern int blk_rq_unmap_user(struct bio *); 878 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); 879 extern int blk_rq_map_user_iov(struct request_queue *, struct request *, 880 struct rq_map_data *, const struct iov_iter *, 881 gfp_t); 882 extern void blk_execute_rq(struct request_queue *, struct gendisk *, 883 struct request *, int); 884 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, 885 struct request *, int, rq_end_io_fn *); 886 887 /* Helper to convert REQ_OP_XXX to its string format XXX */ 888 extern const char *blk_op_str(unsigned int op); 889 890 int blk_status_to_errno(blk_status_t status); 891 blk_status_t errno_to_blk_status(int errno); 892 893 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin); 894 895 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 896 { 897 return bdev->bd_disk->queue; /* this is never NULL */ 898 } 899 900 /* 901 * The basic unit of block I/O is a sector. It is used in a number of contexts 902 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9 903 * bytes. Variables of type sector_t represent an offset or size that is a 904 * multiple of 512 bytes. Hence these two constants. 905 */ 906 #ifndef SECTOR_SHIFT 907 #define SECTOR_SHIFT 9 908 #endif 909 #ifndef SECTOR_SIZE 910 #define SECTOR_SIZE (1 << SECTOR_SHIFT) 911 #endif 912 913 /* 914 * blk_rq_pos() : the current sector 915 * blk_rq_bytes() : bytes left in the entire request 916 * blk_rq_cur_bytes() : bytes left in the current segment 917 * blk_rq_err_bytes() : bytes left till the next error boundary 918 * blk_rq_sectors() : sectors left in the entire request 919 * blk_rq_cur_sectors() : sectors left in the current segment 920 */ 921 static inline sector_t blk_rq_pos(const struct request *rq) 922 { 923 return rq->__sector; 924 } 925 926 static inline unsigned int blk_rq_bytes(const struct request *rq) 927 { 928 return rq->__data_len; 929 } 930 931 static inline int blk_rq_cur_bytes(const struct request *rq) 932 { 933 return rq->bio ? bio_cur_bytes(rq->bio) : 0; 934 } 935 936 extern unsigned int blk_rq_err_bytes(const struct request *rq); 937 938 static inline unsigned int blk_rq_sectors(const struct request *rq) 939 { 940 return blk_rq_bytes(rq) >> SECTOR_SHIFT; 941 } 942 943 static inline unsigned int blk_rq_cur_sectors(const struct request *rq) 944 { 945 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT; 946 } 947 948 #ifdef CONFIG_BLK_DEV_ZONED 949 static inline unsigned int blk_rq_zone_no(struct request *rq) 950 { 951 return blk_queue_zone_no(rq->q, blk_rq_pos(rq)); 952 } 953 954 static inline unsigned int blk_rq_zone_is_seq(struct request *rq) 955 { 956 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq)); 957 } 958 #endif /* CONFIG_BLK_DEV_ZONED */ 959 960 /* 961 * Some commands like WRITE SAME have a payload or data transfer size which 962 * is different from the size of the request. Any driver that supports such 963 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to 964 * calculate the data transfer size. 965 */ 966 static inline unsigned int blk_rq_payload_bytes(struct request *rq) 967 { 968 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 969 return rq->special_vec.bv_len; 970 return blk_rq_bytes(rq); 971 } 972 973 /* 974 * Return the first full biovec in the request. The caller needs to check that 975 * there are any bvecs before calling this helper. 976 */ 977 static inline struct bio_vec req_bvec(struct request *rq) 978 { 979 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 980 return rq->special_vec; 981 return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter); 982 } 983 984 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, 985 int op) 986 { 987 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) 988 return min(q->limits.max_discard_sectors, 989 UINT_MAX >> SECTOR_SHIFT); 990 991 if (unlikely(op == REQ_OP_WRITE_SAME)) 992 return q->limits.max_write_same_sectors; 993 994 if (unlikely(op == REQ_OP_WRITE_ZEROES)) 995 return q->limits.max_write_zeroes_sectors; 996 997 return q->limits.max_sectors; 998 } 999 1000 /* 1001 * Return maximum size of a request at given offset. Only valid for 1002 * file system requests. 1003 */ 1004 static inline unsigned int blk_max_size_offset(struct request_queue *q, 1005 sector_t offset) 1006 { 1007 if (!q->limits.chunk_sectors) 1008 return q->limits.max_sectors; 1009 1010 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors - 1011 (offset & (q->limits.chunk_sectors - 1)))); 1012 } 1013 1014 static inline unsigned int blk_rq_get_max_sectors(struct request *rq, 1015 sector_t offset) 1016 { 1017 struct request_queue *q = rq->q; 1018 1019 if (blk_rq_is_passthrough(rq)) 1020 return q->limits.max_hw_sectors; 1021 1022 if (!q->limits.chunk_sectors || 1023 req_op(rq) == REQ_OP_DISCARD || 1024 req_op(rq) == REQ_OP_SECURE_ERASE) 1025 return blk_queue_get_max_sectors(q, req_op(rq)); 1026 1027 return min(blk_max_size_offset(q, offset), 1028 blk_queue_get_max_sectors(q, req_op(rq))); 1029 } 1030 1031 static inline unsigned int blk_rq_count_bios(struct request *rq) 1032 { 1033 unsigned int nr_bios = 0; 1034 struct bio *bio; 1035 1036 __rq_for_each_bio(bio, rq) 1037 nr_bios++; 1038 1039 return nr_bios; 1040 } 1041 1042 void blk_steal_bios(struct bio_list *list, struct request *rq); 1043 1044 /* 1045 * Request completion related functions. 1046 * 1047 * blk_update_request() completes given number of bytes and updates 1048 * the request without completing it. 1049 */ 1050 extern bool blk_update_request(struct request *rq, blk_status_t error, 1051 unsigned int nr_bytes); 1052 1053 extern void __blk_complete_request(struct request *); 1054 extern void blk_abort_request(struct request *); 1055 1056 /* 1057 * Access functions for manipulating queue properties 1058 */ 1059 extern void blk_cleanup_queue(struct request_queue *); 1060 extern void blk_queue_make_request(struct request_queue *, make_request_fn *); 1061 extern void blk_queue_bounce_limit(struct request_queue *, u64); 1062 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); 1063 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); 1064 extern void blk_queue_max_segments(struct request_queue *, unsigned short); 1065 extern void blk_queue_max_discard_segments(struct request_queue *, 1066 unsigned short); 1067 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); 1068 extern void blk_queue_max_discard_sectors(struct request_queue *q, 1069 unsigned int max_discard_sectors); 1070 extern void blk_queue_max_write_same_sectors(struct request_queue *q, 1071 unsigned int max_write_same_sectors); 1072 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, 1073 unsigned int max_write_same_sectors); 1074 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); 1075 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); 1076 extern void blk_queue_alignment_offset(struct request_queue *q, 1077 unsigned int alignment); 1078 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 1079 extern void blk_queue_io_min(struct request_queue *q, unsigned int min); 1080 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 1081 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); 1082 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1083 extern void blk_set_default_limits(struct queue_limits *lim); 1084 extern void blk_set_stacking_limits(struct queue_limits *lim); 1085 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1086 sector_t offset); 1087 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, 1088 sector_t offset); 1089 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, 1090 sector_t offset); 1091 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); 1092 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 1093 extern int blk_queue_dma_drain(struct request_queue *q, 1094 dma_drain_needed_fn *dma_drain_needed, 1095 void *buf, unsigned int size); 1096 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); 1097 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long); 1098 extern void blk_queue_dma_alignment(struct request_queue *, int); 1099 extern void blk_queue_update_dma_alignment(struct request_queue *, int); 1100 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1101 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); 1102 extern void blk_queue_required_elevator_features(struct request_queue *q, 1103 unsigned int features); 1104 1105 /* 1106 * Number of physical segments as sent to the device. 1107 * 1108 * Normally this is the number of discontiguous data segments sent by the 1109 * submitter. But for data-less command like discard we might have no 1110 * actual data segments submitted, but the driver might have to add it's 1111 * own special payload. In that case we still return 1 here so that this 1112 * special payload will be mapped. 1113 */ 1114 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq) 1115 { 1116 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 1117 return 1; 1118 return rq->nr_phys_segments; 1119 } 1120 1121 /* 1122 * Number of discard segments (or ranges) the driver needs to fill in. 1123 * Each discard bio merged into a request is counted as one segment. 1124 */ 1125 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq) 1126 { 1127 return max_t(unsigned short, rq->nr_phys_segments, 1); 1128 } 1129 1130 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); 1131 extern void blk_dump_rq_flags(struct request *, char *); 1132 extern long nr_blockdev_pages(void); 1133 1134 bool __must_check blk_get_queue(struct request_queue *); 1135 struct request_queue *blk_alloc_queue(gfp_t); 1136 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id); 1137 extern void blk_put_queue(struct request_queue *); 1138 extern void blk_set_queue_dying(struct request_queue *); 1139 1140 /* 1141 * blk_plug permits building a queue of related requests by holding the I/O 1142 * fragments for a short period. This allows merging of sequential requests 1143 * into single larger request. As the requests are moved from a per-task list to 1144 * the device's request_queue in a batch, this results in improved scalability 1145 * as the lock contention for request_queue lock is reduced. 1146 * 1147 * It is ok not to disable preemption when adding the request to the plug list 1148 * or when attempting a merge, because blk_schedule_flush_list() will only flush 1149 * the plug list when the task sleeps by itself. For details, please see 1150 * schedule() where blk_schedule_flush_plug() is called. 1151 */ 1152 struct blk_plug { 1153 struct list_head mq_list; /* blk-mq requests */ 1154 struct list_head cb_list; /* md requires an unplug callback */ 1155 unsigned short rq_count; 1156 bool multiple_queues; 1157 }; 1158 #define BLK_MAX_REQUEST_COUNT 16 1159 #define BLK_PLUG_FLUSH_SIZE (128 * 1024) 1160 1161 struct blk_plug_cb; 1162 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1163 struct blk_plug_cb { 1164 struct list_head list; 1165 blk_plug_cb_fn callback; 1166 void *data; 1167 }; 1168 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1169 void *data, int size); 1170 extern void blk_start_plug(struct blk_plug *); 1171 extern void blk_finish_plug(struct blk_plug *); 1172 extern void blk_flush_plug_list(struct blk_plug *, bool); 1173 1174 static inline void blk_flush_plug(struct task_struct *tsk) 1175 { 1176 struct blk_plug *plug = tsk->plug; 1177 1178 if (plug) 1179 blk_flush_plug_list(plug, false); 1180 } 1181 1182 static inline void blk_schedule_flush_plug(struct task_struct *tsk) 1183 { 1184 struct blk_plug *plug = tsk->plug; 1185 1186 if (plug) 1187 blk_flush_plug_list(plug, true); 1188 } 1189 1190 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1191 { 1192 struct blk_plug *plug = tsk->plug; 1193 1194 return plug && 1195 (!list_empty(&plug->mq_list) || 1196 !list_empty(&plug->cb_list)); 1197 } 1198 1199 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); 1200 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, 1201 sector_t nr_sects, gfp_t gfp_mask, struct page *page); 1202 1203 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */ 1204 1205 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1206 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); 1207 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1208 sector_t nr_sects, gfp_t gfp_mask, int flags, 1209 struct bio **biop); 1210 1211 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1212 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1213 1214 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1215 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1216 unsigned flags); 1217 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1218 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1219 1220 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1221 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1222 { 1223 return blkdev_issue_discard(sb->s_bdev, 1224 block << (sb->s_blocksize_bits - 1225 SECTOR_SHIFT), 1226 nr_blocks << (sb->s_blocksize_bits - 1227 SECTOR_SHIFT), 1228 gfp_mask, flags); 1229 } 1230 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1231 sector_t nr_blocks, gfp_t gfp_mask) 1232 { 1233 return blkdev_issue_zeroout(sb->s_bdev, 1234 block << (sb->s_blocksize_bits - 1235 SECTOR_SHIFT), 1236 nr_blocks << (sb->s_blocksize_bits - 1237 SECTOR_SHIFT), 1238 gfp_mask, 0); 1239 } 1240 1241 extern int blk_verify_command(unsigned char *cmd, fmode_t mode); 1242 1243 enum blk_default_limits { 1244 BLK_MAX_SEGMENTS = 128, 1245 BLK_SAFE_MAX_SECTORS = 255, 1246 BLK_DEF_MAX_SECTORS = 2560, 1247 BLK_MAX_SEGMENT_SIZE = 65536, 1248 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1249 }; 1250 1251 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1252 { 1253 return q->limits.seg_boundary_mask; 1254 } 1255 1256 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1257 { 1258 return q->limits.virt_boundary_mask; 1259 } 1260 1261 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1262 { 1263 return q->limits.max_sectors; 1264 } 1265 1266 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1267 { 1268 return q->limits.max_hw_sectors; 1269 } 1270 1271 static inline unsigned short queue_max_segments(const struct request_queue *q) 1272 { 1273 return q->limits.max_segments; 1274 } 1275 1276 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1277 { 1278 return q->limits.max_discard_segments; 1279 } 1280 1281 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1282 { 1283 return q->limits.max_segment_size; 1284 } 1285 1286 static inline unsigned short queue_logical_block_size(const struct request_queue *q) 1287 { 1288 int retval = 512; 1289 1290 if (q && q->limits.logical_block_size) 1291 retval = q->limits.logical_block_size; 1292 1293 return retval; 1294 } 1295 1296 static inline unsigned short bdev_logical_block_size(struct block_device *bdev) 1297 { 1298 return queue_logical_block_size(bdev_get_queue(bdev)); 1299 } 1300 1301 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1302 { 1303 return q->limits.physical_block_size; 1304 } 1305 1306 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1307 { 1308 return queue_physical_block_size(bdev_get_queue(bdev)); 1309 } 1310 1311 static inline unsigned int queue_io_min(const struct request_queue *q) 1312 { 1313 return q->limits.io_min; 1314 } 1315 1316 static inline int bdev_io_min(struct block_device *bdev) 1317 { 1318 return queue_io_min(bdev_get_queue(bdev)); 1319 } 1320 1321 static inline unsigned int queue_io_opt(const struct request_queue *q) 1322 { 1323 return q->limits.io_opt; 1324 } 1325 1326 static inline int bdev_io_opt(struct block_device *bdev) 1327 { 1328 return queue_io_opt(bdev_get_queue(bdev)); 1329 } 1330 1331 static inline int queue_alignment_offset(const struct request_queue *q) 1332 { 1333 if (q->limits.misaligned) 1334 return -1; 1335 1336 return q->limits.alignment_offset; 1337 } 1338 1339 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) 1340 { 1341 unsigned int granularity = max(lim->physical_block_size, lim->io_min); 1342 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) 1343 << SECTOR_SHIFT; 1344 1345 return (granularity + lim->alignment_offset - alignment) % granularity; 1346 } 1347 1348 static inline int bdev_alignment_offset(struct block_device *bdev) 1349 { 1350 struct request_queue *q = bdev_get_queue(bdev); 1351 1352 if (q->limits.misaligned) 1353 return -1; 1354 1355 if (bdev != bdev->bd_contains) 1356 return bdev->bd_part->alignment_offset; 1357 1358 return q->limits.alignment_offset; 1359 } 1360 1361 static inline int queue_discard_alignment(const struct request_queue *q) 1362 { 1363 if (q->limits.discard_misaligned) 1364 return -1; 1365 1366 return q->limits.discard_alignment; 1367 } 1368 1369 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) 1370 { 1371 unsigned int alignment, granularity, offset; 1372 1373 if (!lim->max_discard_sectors) 1374 return 0; 1375 1376 /* Why are these in bytes, not sectors? */ 1377 alignment = lim->discard_alignment >> SECTOR_SHIFT; 1378 granularity = lim->discard_granularity >> SECTOR_SHIFT; 1379 if (!granularity) 1380 return 0; 1381 1382 /* Offset of the partition start in 'granularity' sectors */ 1383 offset = sector_div(sector, granularity); 1384 1385 /* And why do we do this modulus *again* in blkdev_issue_discard()? */ 1386 offset = (granularity + alignment - offset) % granularity; 1387 1388 /* Turn it back into bytes, gaah */ 1389 return offset << SECTOR_SHIFT; 1390 } 1391 1392 static inline int bdev_discard_alignment(struct block_device *bdev) 1393 { 1394 struct request_queue *q = bdev_get_queue(bdev); 1395 1396 if (bdev != bdev->bd_contains) 1397 return bdev->bd_part->discard_alignment; 1398 1399 return q->limits.discard_alignment; 1400 } 1401 1402 static inline unsigned int bdev_write_same(struct block_device *bdev) 1403 { 1404 struct request_queue *q = bdev_get_queue(bdev); 1405 1406 if (q) 1407 return q->limits.max_write_same_sectors; 1408 1409 return 0; 1410 } 1411 1412 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1413 { 1414 struct request_queue *q = bdev_get_queue(bdev); 1415 1416 if (q) 1417 return q->limits.max_write_zeroes_sectors; 1418 1419 return 0; 1420 } 1421 1422 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev) 1423 { 1424 struct request_queue *q = bdev_get_queue(bdev); 1425 1426 if (q) 1427 return blk_queue_zoned_model(q); 1428 1429 return BLK_ZONED_NONE; 1430 } 1431 1432 static inline bool bdev_is_zoned(struct block_device *bdev) 1433 { 1434 struct request_queue *q = bdev_get_queue(bdev); 1435 1436 if (q) 1437 return blk_queue_is_zoned(q); 1438 1439 return false; 1440 } 1441 1442 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1443 { 1444 struct request_queue *q = bdev_get_queue(bdev); 1445 1446 if (q) 1447 return blk_queue_zone_sectors(q); 1448 return 0; 1449 } 1450 1451 static inline int queue_dma_alignment(const struct request_queue *q) 1452 { 1453 return q ? q->dma_alignment : 511; 1454 } 1455 1456 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1457 unsigned int len) 1458 { 1459 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1460 return !(addr & alignment) && !(len & alignment); 1461 } 1462 1463 /* assumes size > 256 */ 1464 static inline unsigned int blksize_bits(unsigned int size) 1465 { 1466 unsigned int bits = 8; 1467 do { 1468 bits++; 1469 size >>= 1; 1470 } while (size > 256); 1471 return bits; 1472 } 1473 1474 static inline unsigned int block_size(struct block_device *bdev) 1475 { 1476 return bdev->bd_block_size; 1477 } 1478 1479 typedef struct {struct page *v;} Sector; 1480 1481 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); 1482 1483 static inline void put_dev_sector(Sector p) 1484 { 1485 put_page(p.v); 1486 } 1487 1488 int kblockd_schedule_work(struct work_struct *work); 1489 int kblockd_schedule_work_on(int cpu, struct work_struct *work); 1490 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1491 1492 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1493 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1494 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1495 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1496 1497 #if defined(CONFIG_BLK_DEV_INTEGRITY) 1498 1499 enum blk_integrity_flags { 1500 BLK_INTEGRITY_VERIFY = 1 << 0, 1501 BLK_INTEGRITY_GENERATE = 1 << 1, 1502 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, 1503 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, 1504 }; 1505 1506 struct blk_integrity_iter { 1507 void *prot_buf; 1508 void *data_buf; 1509 sector_t seed; 1510 unsigned int data_size; 1511 unsigned short interval; 1512 const char *disk_name; 1513 }; 1514 1515 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *); 1516 1517 struct blk_integrity_profile { 1518 integrity_processing_fn *generate_fn; 1519 integrity_processing_fn *verify_fn; 1520 const char *name; 1521 }; 1522 1523 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *); 1524 extern void blk_integrity_unregister(struct gendisk *); 1525 extern int blk_integrity_compare(struct gendisk *, struct gendisk *); 1526 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, 1527 struct scatterlist *); 1528 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); 1529 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, 1530 struct request *); 1531 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, 1532 struct bio *); 1533 1534 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1535 { 1536 struct blk_integrity *bi = &disk->queue->integrity; 1537 1538 if (!bi->profile) 1539 return NULL; 1540 1541 return bi; 1542 } 1543 1544 static inline 1545 struct blk_integrity *bdev_get_integrity(struct block_device *bdev) 1546 { 1547 return blk_get_integrity(bdev->bd_disk); 1548 } 1549 1550 static inline bool blk_integrity_rq(struct request *rq) 1551 { 1552 return rq->cmd_flags & REQ_INTEGRITY; 1553 } 1554 1555 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1556 unsigned int segs) 1557 { 1558 q->limits.max_integrity_segments = segs; 1559 } 1560 1561 static inline unsigned short 1562 queue_max_integrity_segments(const struct request_queue *q) 1563 { 1564 return q->limits.max_integrity_segments; 1565 } 1566 1567 /** 1568 * bio_integrity_intervals - Return number of integrity intervals for a bio 1569 * @bi: blk_integrity profile for device 1570 * @sectors: Size of the bio in 512-byte sectors 1571 * 1572 * Description: The block layer calculates everything in 512 byte 1573 * sectors but integrity metadata is done in terms of the data integrity 1574 * interval size of the storage device. Convert the block layer sectors 1575 * to the appropriate number of integrity intervals. 1576 */ 1577 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1578 unsigned int sectors) 1579 { 1580 return sectors >> (bi->interval_exp - 9); 1581 } 1582 1583 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1584 unsigned int sectors) 1585 { 1586 return bio_integrity_intervals(bi, sectors) * bi->tuple_size; 1587 } 1588 1589 /* 1590 * Return the first bvec that contains integrity data. Only drivers that are 1591 * limited to a single integrity segment should use this helper. 1592 */ 1593 static inline struct bio_vec *rq_integrity_vec(struct request *rq) 1594 { 1595 if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1)) 1596 return NULL; 1597 return rq->bio->bi_integrity->bip_vec; 1598 } 1599 1600 #else /* CONFIG_BLK_DEV_INTEGRITY */ 1601 1602 struct bio; 1603 struct block_device; 1604 struct gendisk; 1605 struct blk_integrity; 1606 1607 static inline int blk_integrity_rq(struct request *rq) 1608 { 1609 return 0; 1610 } 1611 static inline int blk_rq_count_integrity_sg(struct request_queue *q, 1612 struct bio *b) 1613 { 1614 return 0; 1615 } 1616 static inline int blk_rq_map_integrity_sg(struct request_queue *q, 1617 struct bio *b, 1618 struct scatterlist *s) 1619 { 1620 return 0; 1621 } 1622 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) 1623 { 1624 return NULL; 1625 } 1626 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1627 { 1628 return NULL; 1629 } 1630 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) 1631 { 1632 return 0; 1633 } 1634 static inline void blk_integrity_register(struct gendisk *d, 1635 struct blk_integrity *b) 1636 { 1637 } 1638 static inline void blk_integrity_unregister(struct gendisk *d) 1639 { 1640 } 1641 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1642 unsigned int segs) 1643 { 1644 } 1645 static inline unsigned short queue_max_integrity_segments(const struct request_queue *q) 1646 { 1647 return 0; 1648 } 1649 static inline bool blk_integrity_merge_rq(struct request_queue *rq, 1650 struct request *r1, 1651 struct request *r2) 1652 { 1653 return true; 1654 } 1655 static inline bool blk_integrity_merge_bio(struct request_queue *rq, 1656 struct request *r, 1657 struct bio *b) 1658 { 1659 return true; 1660 } 1661 1662 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1663 unsigned int sectors) 1664 { 1665 return 0; 1666 } 1667 1668 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1669 unsigned int sectors) 1670 { 1671 return 0; 1672 } 1673 1674 static inline struct bio_vec *rq_integrity_vec(struct request *rq) 1675 { 1676 return NULL; 1677 } 1678 1679 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 1680 1681 struct block_device_operations { 1682 int (*open) (struct block_device *, fmode_t); 1683 void (*release) (struct gendisk *, fmode_t); 1684 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int); 1685 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1686 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1687 unsigned int (*check_events) (struct gendisk *disk, 1688 unsigned int clearing); 1689 /* ->media_changed() is DEPRECATED, use ->check_events() instead */ 1690 int (*media_changed) (struct gendisk *); 1691 void (*unlock_native_capacity) (struct gendisk *); 1692 int (*revalidate_disk) (struct gendisk *); 1693 int (*getgeo)(struct block_device *, struct hd_geometry *); 1694 /* this callback is with swap_lock and sometimes page table lock held */ 1695 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1696 int (*report_zones)(struct gendisk *, sector_t sector, 1697 struct blk_zone *zones, unsigned int *nr_zones); 1698 struct module *owner; 1699 const struct pr_ops *pr_ops; 1700 }; 1701 1702 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, 1703 unsigned long); 1704 extern int bdev_read_page(struct block_device *, sector_t, struct page *); 1705 extern int bdev_write_page(struct block_device *, sector_t, struct page *, 1706 struct writeback_control *); 1707 1708 #ifdef CONFIG_BLK_DEV_ZONED 1709 bool blk_req_needs_zone_write_lock(struct request *rq); 1710 void __blk_req_zone_write_lock(struct request *rq); 1711 void __blk_req_zone_write_unlock(struct request *rq); 1712 1713 static inline void blk_req_zone_write_lock(struct request *rq) 1714 { 1715 if (blk_req_needs_zone_write_lock(rq)) 1716 __blk_req_zone_write_lock(rq); 1717 } 1718 1719 static inline void blk_req_zone_write_unlock(struct request *rq) 1720 { 1721 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED) 1722 __blk_req_zone_write_unlock(rq); 1723 } 1724 1725 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1726 { 1727 return rq->q->seq_zones_wlock && 1728 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock); 1729 } 1730 1731 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1732 { 1733 if (!blk_req_needs_zone_write_lock(rq)) 1734 return true; 1735 return !blk_req_zone_is_write_locked(rq); 1736 } 1737 #else 1738 static inline bool blk_req_needs_zone_write_lock(struct request *rq) 1739 { 1740 return false; 1741 } 1742 1743 static inline void blk_req_zone_write_lock(struct request *rq) 1744 { 1745 } 1746 1747 static inline void blk_req_zone_write_unlock(struct request *rq) 1748 { 1749 } 1750 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1751 { 1752 return false; 1753 } 1754 1755 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1756 { 1757 return true; 1758 } 1759 #endif /* CONFIG_BLK_DEV_ZONED */ 1760 1761 #else /* CONFIG_BLOCK */ 1762 1763 struct block_device; 1764 1765 /* 1766 * stubs for when the block layer is configured out 1767 */ 1768 #define buffer_heads_over_limit 0 1769 1770 static inline long nr_blockdev_pages(void) 1771 { 1772 return 0; 1773 } 1774 1775 struct blk_plug { 1776 }; 1777 1778 static inline void blk_start_plug(struct blk_plug *plug) 1779 { 1780 } 1781 1782 static inline void blk_finish_plug(struct blk_plug *plug) 1783 { 1784 } 1785 1786 static inline void blk_flush_plug(struct task_struct *task) 1787 { 1788 } 1789 1790 static inline void blk_schedule_flush_plug(struct task_struct *task) 1791 { 1792 } 1793 1794 1795 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1796 { 1797 return false; 1798 } 1799 1800 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, 1801 sector_t *error_sector) 1802 { 1803 return 0; 1804 } 1805 1806 #endif /* CONFIG_BLOCK */ 1807 1808 static inline void blk_wake_io_task(struct task_struct *waiter) 1809 { 1810 /* 1811 * If we're polling, the task itself is doing the completions. For 1812 * that case, we don't need to signal a wakeup, it's enough to just 1813 * mark us as RUNNING. 1814 */ 1815 if (waiter == current) 1816 __set_current_state(TASK_RUNNING); 1817 else 1818 wake_up_process(waiter); 1819 } 1820 1821 #endif 1822