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