1#ifndef _LINUX_BLKDEV_H 2#define _LINUX_BLKDEV_H 3 4#include <linux/sched.h> 5 6#ifdef CONFIG_BLOCK 7 8#include <linux/major.h> 9#include <linux/genhd.h> 10#include <linux/list.h> 11#include <linux/llist.h> 12#include <linux/timer.h> 13#include <linux/workqueue.h> 14#include <linux/pagemap.h> 15#include <linux/backing-dev.h> 16#include <linux/wait.h> 17#include <linux/mempool.h> 18#include <linux/bio.h> 19#include <linux/stringify.h> 20#include <linux/gfp.h> 21#include <linux/bsg.h> 22#include <linux/smp.h> 23#include <linux/rcupdate.h> 24#include <linux/percpu-refcount.h> 25 26#include <asm/scatterlist.h> 27 28struct module; 29struct scsi_ioctl_command; 30 31struct request_queue; 32struct elevator_queue; 33struct request_pm_state; 34struct blk_trace; 35struct request; 36struct sg_io_hdr; 37struct bsg_job; 38struct blkcg_gq; 39struct blk_flush_queue; 40 41#define BLKDEV_MIN_RQ 4 42#define BLKDEV_MAX_RQ 128 /* Default maximum */ 43 44/* 45 * Maximum number of blkcg policies allowed to be registered concurrently. 46 * Defined here to simplify include dependency. 47 */ 48#define BLKCG_MAX_POLS 2 49 50struct request; 51typedef void (rq_end_io_fn)(struct request *, int); 52 53#define BLK_RL_SYNCFULL (1U << 0) 54#define BLK_RL_ASYNCFULL (1U << 1) 55 56struct request_list { 57 struct request_queue *q; /* the queue this rl belongs to */ 58#ifdef CONFIG_BLK_CGROUP 59 struct blkcg_gq *blkg; /* blkg this request pool belongs to */ 60#endif 61 /* 62 * count[], starved[], and wait[] are indexed by 63 * BLK_RW_SYNC/BLK_RW_ASYNC 64 */ 65 int count[2]; 66 int starved[2]; 67 mempool_t *rq_pool; 68 wait_queue_head_t wait[2]; 69 unsigned int flags; 70}; 71 72/* 73 * request command types 74 */ 75enum rq_cmd_type_bits { 76 REQ_TYPE_FS = 1, /* fs request */ 77 REQ_TYPE_BLOCK_PC, /* scsi command */ 78 REQ_TYPE_SENSE, /* sense request */ 79 REQ_TYPE_PM_SUSPEND, /* suspend request */ 80 REQ_TYPE_PM_RESUME, /* resume request */ 81 REQ_TYPE_PM_SHUTDOWN, /* shutdown request */ 82 REQ_TYPE_SPECIAL, /* driver defined type */ 83 /* 84 * for ATA/ATAPI devices. this really doesn't belong here, ide should 85 * use REQ_TYPE_SPECIAL and use rq->cmd[0] with the range of driver 86 * private REQ_LB opcodes to differentiate what type of request this is 87 */ 88 REQ_TYPE_ATA_TASKFILE, 89 REQ_TYPE_ATA_PC, 90}; 91 92#define BLK_MAX_CDB 16 93 94/* 95 * Try to put the fields that are referenced together in the same cacheline. 96 * 97 * If you modify this structure, make sure to update blk_rq_init() and 98 * especially blk_mq_rq_ctx_init() to take care of the added fields. 99 */ 100struct request { 101 struct list_head queuelist; 102 union { 103 struct call_single_data csd; 104 unsigned long fifo_time; 105 }; 106 107 struct request_queue *q; 108 struct blk_mq_ctx *mq_ctx; 109 110 u64 cmd_flags; 111 enum rq_cmd_type_bits cmd_type; 112 unsigned long atomic_flags; 113 114 int cpu; 115 116 /* the following two fields are internal, NEVER access directly */ 117 unsigned int __data_len; /* total data len */ 118 sector_t __sector; /* sector cursor */ 119 120 struct bio *bio; 121 struct bio *biotail; 122 123 /* 124 * The hash is used inside the scheduler, and killed once the 125 * request reaches the dispatch list. The ipi_list is only used 126 * to queue the request for softirq completion, which is long 127 * after the request has been unhashed (and even removed from 128 * the dispatch list). 129 */ 130 union { 131 struct hlist_node hash; /* merge hash */ 132 struct list_head ipi_list; 133 }; 134 135 /* 136 * The rb_node is only used inside the io scheduler, requests 137 * are pruned when moved to the dispatch queue. So let the 138 * completion_data share space with the rb_node. 139 */ 140 union { 141 struct rb_node rb_node; /* sort/lookup */ 142 void *completion_data; 143 }; 144 145 /* 146 * Three pointers are available for the IO schedulers, if they need 147 * more they have to dynamically allocate it. Flush requests are 148 * never put on the IO scheduler. So let the flush fields share 149 * space with the elevator data. 150 */ 151 union { 152 struct { 153 struct io_cq *icq; 154 void *priv[2]; 155 } elv; 156 157 struct { 158 unsigned int seq; 159 struct list_head list; 160 rq_end_io_fn *saved_end_io; 161 } flush; 162 }; 163 164 struct gendisk *rq_disk; 165 struct hd_struct *part; 166 unsigned long start_time; 167#ifdef CONFIG_BLK_CGROUP 168 struct request_list *rl; /* rl this rq is alloced from */ 169 unsigned long long start_time_ns; 170 unsigned long long io_start_time_ns; /* when passed to hardware */ 171#endif 172 /* Number of scatter-gather DMA addr+len pairs after 173 * physical address coalescing is performed. 174 */ 175 unsigned short nr_phys_segments; 176#if defined(CONFIG_BLK_DEV_INTEGRITY) 177 unsigned short nr_integrity_segments; 178#endif 179 180 unsigned short ioprio; 181 182 void *special; /* opaque pointer available for LLD use */ 183 184 int tag; 185 int errors; 186 187 /* 188 * when request is used as a packet command carrier 189 */ 190 unsigned char __cmd[BLK_MAX_CDB]; 191 unsigned char *cmd; 192 unsigned short cmd_len; 193 194 unsigned int extra_len; /* length of alignment and padding */ 195 unsigned int sense_len; 196 unsigned int resid_len; /* residual count */ 197 void *sense; 198 199 unsigned long deadline; 200 struct list_head timeout_list; 201 unsigned int timeout; 202 int retries; 203 204 /* 205 * completion callback. 206 */ 207 rq_end_io_fn *end_io; 208 void *end_io_data; 209 210 /* for bidi */ 211 struct request *next_rq; 212}; 213 214static inline unsigned short req_get_ioprio(struct request *req) 215{ 216 return req->ioprio; 217} 218 219/* 220 * State information carried for REQ_TYPE_PM_SUSPEND and REQ_TYPE_PM_RESUME 221 * requests. Some step values could eventually be made generic. 222 */ 223struct request_pm_state 224{ 225 /* PM state machine step value, currently driver specific */ 226 int pm_step; 227 /* requested PM state value (S1, S2, S3, S4, ...) */ 228 u32 pm_state; 229 void* data; /* for driver use */ 230}; 231 232#include <linux/elevator.h> 233 234struct blk_queue_ctx; 235 236typedef void (request_fn_proc) (struct request_queue *q); 237typedef void (make_request_fn) (struct request_queue *q, struct bio *bio); 238typedef int (prep_rq_fn) (struct request_queue *, struct request *); 239typedef void (unprep_rq_fn) (struct request_queue *, struct request *); 240 241struct bio_vec; 242struct bvec_merge_data { 243 struct block_device *bi_bdev; 244 sector_t bi_sector; 245 unsigned bi_size; 246 unsigned long bi_rw; 247}; 248typedef int (merge_bvec_fn) (struct request_queue *, struct bvec_merge_data *, 249 struct bio_vec *); 250typedef void (softirq_done_fn)(struct request *); 251typedef int (dma_drain_needed_fn)(struct request *); 252typedef int (lld_busy_fn) (struct request_queue *q); 253typedef int (bsg_job_fn) (struct bsg_job *); 254 255enum blk_eh_timer_return { 256 BLK_EH_NOT_HANDLED, 257 BLK_EH_HANDLED, 258 BLK_EH_RESET_TIMER, 259}; 260 261typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *); 262 263enum blk_queue_state { 264 Queue_down, 265 Queue_up, 266}; 267 268struct blk_queue_tag { 269 struct request **tag_index; /* map of busy tags */ 270 unsigned long *tag_map; /* bit map of free/busy tags */ 271 int busy; /* current depth */ 272 int max_depth; /* what we will send to device */ 273 int real_max_depth; /* what the array can hold */ 274 atomic_t refcnt; /* map can be shared */ 275 int alloc_policy; /* tag allocation policy */ 276 int next_tag; /* next tag */ 277}; 278#define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ 279#define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ 280 281#define BLK_SCSI_MAX_CMDS (256) 282#define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) 283 284struct queue_limits { 285 unsigned long bounce_pfn; 286 unsigned long seg_boundary_mask; 287 288 unsigned int max_hw_sectors; 289 unsigned int chunk_sectors; 290 unsigned int max_sectors; 291 unsigned int max_segment_size; 292 unsigned int physical_block_size; 293 unsigned int alignment_offset; 294 unsigned int io_min; 295 unsigned int io_opt; 296 unsigned int max_discard_sectors; 297 unsigned int max_write_same_sectors; 298 unsigned int discard_granularity; 299 unsigned int discard_alignment; 300 301 unsigned short logical_block_size; 302 unsigned short max_segments; 303 unsigned short max_integrity_segments; 304 305 unsigned char misaligned; 306 unsigned char discard_misaligned; 307 unsigned char cluster; 308 unsigned char discard_zeroes_data; 309 unsigned char raid_partial_stripes_expensive; 310}; 311 312struct request_queue { 313 /* 314 * Together with queue_head for cacheline sharing 315 */ 316 struct list_head queue_head; 317 struct request *last_merge; 318 struct elevator_queue *elevator; 319 int nr_rqs[2]; /* # allocated [a]sync rqs */ 320 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */ 321 322 /* 323 * If blkcg is not used, @q->root_rl serves all requests. If blkcg 324 * is used, root blkg allocates from @q->root_rl and all other 325 * blkgs from their own blkg->rl. Which one to use should be 326 * determined using bio_request_list(). 327 */ 328 struct request_list root_rl; 329 330 request_fn_proc *request_fn; 331 make_request_fn *make_request_fn; 332 prep_rq_fn *prep_rq_fn; 333 unprep_rq_fn *unprep_rq_fn; 334 merge_bvec_fn *merge_bvec_fn; 335 softirq_done_fn *softirq_done_fn; 336 rq_timed_out_fn *rq_timed_out_fn; 337 dma_drain_needed_fn *dma_drain_needed; 338 lld_busy_fn *lld_busy_fn; 339 340 struct blk_mq_ops *mq_ops; 341 342 unsigned int *mq_map; 343 344 /* sw queues */ 345 struct blk_mq_ctx __percpu *queue_ctx; 346 unsigned int nr_queues; 347 348 /* hw dispatch queues */ 349 struct blk_mq_hw_ctx **queue_hw_ctx; 350 unsigned int nr_hw_queues; 351 352 /* 353 * Dispatch queue sorting 354 */ 355 sector_t end_sector; 356 struct request *boundary_rq; 357 358 /* 359 * Delayed queue handling 360 */ 361 struct delayed_work delay_work; 362 363 struct backing_dev_info backing_dev_info; 364 365 /* 366 * The queue owner gets to use this for whatever they like. 367 * ll_rw_blk doesn't touch it. 368 */ 369 void *queuedata; 370 371 /* 372 * various queue flags, see QUEUE_* below 373 */ 374 unsigned long queue_flags; 375 376 /* 377 * ida allocated id for this queue. Used to index queues from 378 * ioctx. 379 */ 380 int id; 381 382 /* 383 * queue needs bounce pages for pages above this limit 384 */ 385 gfp_t bounce_gfp; 386 387 /* 388 * protects queue structures from reentrancy. ->__queue_lock should 389 * _never_ be used directly, it is queue private. always use 390 * ->queue_lock. 391 */ 392 spinlock_t __queue_lock; 393 spinlock_t *queue_lock; 394 395 /* 396 * queue kobject 397 */ 398 struct kobject kobj; 399 400 /* 401 * mq queue kobject 402 */ 403 struct kobject mq_kobj; 404 405#ifdef CONFIG_PM 406 struct device *dev; 407 int rpm_status; 408 unsigned int nr_pending; 409#endif 410 411 /* 412 * queue settings 413 */ 414 unsigned long nr_requests; /* Max # of requests */ 415 unsigned int nr_congestion_on; 416 unsigned int nr_congestion_off; 417 unsigned int nr_batching; 418 419 unsigned int dma_drain_size; 420 void *dma_drain_buffer; 421 unsigned int dma_pad_mask; 422 unsigned int dma_alignment; 423 424 struct blk_queue_tag *queue_tags; 425 struct list_head tag_busy_list; 426 427 unsigned int nr_sorted; 428 unsigned int in_flight[2]; 429 /* 430 * Number of active block driver functions for which blk_drain_queue() 431 * must wait. Must be incremented around functions that unlock the 432 * queue_lock internally, e.g. scsi_request_fn(). 433 */ 434 unsigned int request_fn_active; 435 436 unsigned int rq_timeout; 437 struct timer_list timeout; 438 struct list_head timeout_list; 439 440 struct list_head icq_list; 441#ifdef CONFIG_BLK_CGROUP 442 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 443 struct blkcg_gq *root_blkg; 444 struct list_head blkg_list; 445#endif 446 447 struct queue_limits limits; 448 449 /* 450 * sg stuff 451 */ 452 unsigned int sg_timeout; 453 unsigned int sg_reserved_size; 454 int node; 455#ifdef CONFIG_BLK_DEV_IO_TRACE 456 struct blk_trace *blk_trace; 457#endif 458 /* 459 * for flush operations 460 */ 461 unsigned int flush_flags; 462 unsigned int flush_not_queueable:1; 463 struct blk_flush_queue *fq; 464 465 struct list_head requeue_list; 466 spinlock_t requeue_lock; 467 struct work_struct requeue_work; 468 469 struct mutex sysfs_lock; 470 471 int bypass_depth; 472 int mq_freeze_depth; 473 474#if defined(CONFIG_BLK_DEV_BSG) 475 bsg_job_fn *bsg_job_fn; 476 int bsg_job_size; 477 struct bsg_class_device bsg_dev; 478#endif 479 480#ifdef CONFIG_BLK_DEV_THROTTLING 481 /* Throttle data */ 482 struct throtl_data *td; 483#endif 484 struct rcu_head rcu_head; 485 wait_queue_head_t mq_freeze_wq; 486 struct percpu_ref mq_usage_counter; 487 struct list_head all_q_node; 488 489 struct blk_mq_tag_set *tag_set; 490 struct list_head tag_set_list; 491}; 492 493#define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */ 494#define QUEUE_FLAG_STOPPED 2 /* queue is stopped */ 495#define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */ 496#define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */ 497#define QUEUE_FLAG_DYING 5 /* queue being torn down */ 498#define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */ 499#define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */ 500#define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */ 501#define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */ 502#define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */ 503#define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */ 504#define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */ 505#define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ 506#define QUEUE_FLAG_IO_STAT 13 /* do IO stats */ 507#define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */ 508#define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */ 509#define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */ 510#define QUEUE_FLAG_SECDISCARD 17 /* supports SECDISCARD */ 511#define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */ 512#define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */ 513#define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */ 514#define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/ 515#define QUEUE_FLAG_SG_GAPS 22 /* queue doesn't support SG gaps */ 516 517#define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 518 (1 << QUEUE_FLAG_STACKABLE) | \ 519 (1 << QUEUE_FLAG_SAME_COMP) | \ 520 (1 << QUEUE_FLAG_ADD_RANDOM)) 521 522#define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 523 (1 << QUEUE_FLAG_STACKABLE) | \ 524 (1 << QUEUE_FLAG_SAME_COMP)) 525 526static inline void queue_lockdep_assert_held(struct request_queue *q) 527{ 528 if (q->queue_lock) 529 lockdep_assert_held(q->queue_lock); 530} 531 532static inline void queue_flag_set_unlocked(unsigned int flag, 533 struct request_queue *q) 534{ 535 __set_bit(flag, &q->queue_flags); 536} 537 538static inline int queue_flag_test_and_clear(unsigned int flag, 539 struct request_queue *q) 540{ 541 queue_lockdep_assert_held(q); 542 543 if (test_bit(flag, &q->queue_flags)) { 544 __clear_bit(flag, &q->queue_flags); 545 return 1; 546 } 547 548 return 0; 549} 550 551static inline int queue_flag_test_and_set(unsigned int flag, 552 struct request_queue *q) 553{ 554 queue_lockdep_assert_held(q); 555 556 if (!test_bit(flag, &q->queue_flags)) { 557 __set_bit(flag, &q->queue_flags); 558 return 0; 559 } 560 561 return 1; 562} 563 564static inline void queue_flag_set(unsigned int flag, struct request_queue *q) 565{ 566 queue_lockdep_assert_held(q); 567 __set_bit(flag, &q->queue_flags); 568} 569 570static inline void queue_flag_clear_unlocked(unsigned int flag, 571 struct request_queue *q) 572{ 573 __clear_bit(flag, &q->queue_flags); 574} 575 576static inline int queue_in_flight(struct request_queue *q) 577{ 578 return q->in_flight[0] + q->in_flight[1]; 579} 580 581static inline void queue_flag_clear(unsigned int flag, struct request_queue *q) 582{ 583 queue_lockdep_assert_held(q); 584 __clear_bit(flag, &q->queue_flags); 585} 586 587#define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags) 588#define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 589#define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 590#define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) 591#define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags) 592#define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 593#define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 594#define blk_queue_noxmerges(q) \ 595 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 596#define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) 597#define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) 598#define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) 599#define blk_queue_stackable(q) \ 600 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags) 601#define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) 602#define blk_queue_secdiscard(q) (blk_queue_discard(q) && \ 603 test_bit(QUEUE_FLAG_SECDISCARD, &(q)->queue_flags)) 604 605#define blk_noretry_request(rq) \ 606 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 607 REQ_FAILFAST_DRIVER)) 608 609#define blk_account_rq(rq) \ 610 (((rq)->cmd_flags & REQ_STARTED) && \ 611 ((rq)->cmd_type == REQ_TYPE_FS)) 612 613#define blk_pm_request(rq) \ 614 ((rq)->cmd_type == REQ_TYPE_PM_SUSPEND || \ 615 (rq)->cmd_type == REQ_TYPE_PM_RESUME) 616 617#define blk_rq_cpu_valid(rq) ((rq)->cpu != -1) 618#define blk_bidi_rq(rq) ((rq)->next_rq != NULL) 619/* rq->queuelist of dequeued request must be list_empty() */ 620#define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist)) 621 622#define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 623 624#define rq_data_dir(rq) (((rq)->cmd_flags & 1) != 0) 625 626/* 627 * Driver can handle struct request, if it either has an old style 628 * request_fn defined, or is blk-mq based. 629 */ 630static inline bool queue_is_rq_based(struct request_queue *q) 631{ 632 return q->request_fn || q->mq_ops; 633} 634 635static inline unsigned int blk_queue_cluster(struct request_queue *q) 636{ 637 return q->limits.cluster; 638} 639 640/* 641 * We regard a request as sync, if either a read or a sync write 642 */ 643static inline bool rw_is_sync(unsigned int rw_flags) 644{ 645 return !(rw_flags & REQ_WRITE) || (rw_flags & REQ_SYNC); 646} 647 648static inline bool rq_is_sync(struct request *rq) 649{ 650 return rw_is_sync(rq->cmd_flags); 651} 652 653static inline bool blk_rl_full(struct request_list *rl, bool sync) 654{ 655 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 656 657 return rl->flags & flag; 658} 659 660static inline void blk_set_rl_full(struct request_list *rl, bool sync) 661{ 662 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 663 664 rl->flags |= flag; 665} 666 667static inline void blk_clear_rl_full(struct request_list *rl, bool sync) 668{ 669 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 670 671 rl->flags &= ~flag; 672} 673 674static inline bool rq_mergeable(struct request *rq) 675{ 676 if (rq->cmd_type != REQ_TYPE_FS) 677 return false; 678 679 if (rq->cmd_flags & REQ_NOMERGE_FLAGS) 680 return false; 681 682 return true; 683} 684 685static inline bool blk_check_merge_flags(unsigned int flags1, 686 unsigned int flags2) 687{ 688 if ((flags1 & REQ_DISCARD) != (flags2 & REQ_DISCARD)) 689 return false; 690 691 if ((flags1 & REQ_SECURE) != (flags2 & REQ_SECURE)) 692 return false; 693 694 if ((flags1 & REQ_WRITE_SAME) != (flags2 & REQ_WRITE_SAME)) 695 return false; 696 697 return true; 698} 699 700static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) 701{ 702 if (bio_data(a) == bio_data(b)) 703 return true; 704 705 return false; 706} 707 708/* 709 * q->prep_rq_fn return values 710 */ 711#define BLKPREP_OK 0 /* serve it */ 712#define BLKPREP_KILL 1 /* fatal error, kill */ 713#define BLKPREP_DEFER 2 /* leave on queue */ 714 715extern unsigned long blk_max_low_pfn, blk_max_pfn; 716 717/* 718 * standard bounce addresses: 719 * 720 * BLK_BOUNCE_HIGH : bounce all highmem pages 721 * BLK_BOUNCE_ANY : don't bounce anything 722 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary 723 */ 724 725#if BITS_PER_LONG == 32 726#define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) 727#else 728#define BLK_BOUNCE_HIGH -1ULL 729#endif 730#define BLK_BOUNCE_ANY (-1ULL) 731#define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) 732 733/* 734 * default timeout for SG_IO if none specified 735 */ 736#define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 737#define BLK_MIN_SG_TIMEOUT (7 * HZ) 738 739#ifdef CONFIG_BOUNCE 740extern int init_emergency_isa_pool(void); 741extern void blk_queue_bounce(struct request_queue *q, struct bio **bio); 742#else 743static inline int init_emergency_isa_pool(void) 744{ 745 return 0; 746} 747static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio) 748{ 749} 750#endif /* CONFIG_MMU */ 751 752struct rq_map_data { 753 struct page **pages; 754 int page_order; 755 int nr_entries; 756 unsigned long offset; 757 int null_mapped; 758 int from_user; 759}; 760 761struct req_iterator { 762 struct bvec_iter iter; 763 struct bio *bio; 764}; 765 766/* This should not be used directly - use rq_for_each_segment */ 767#define for_each_bio(_bio) \ 768 for (; _bio; _bio = _bio->bi_next) 769#define __rq_for_each_bio(_bio, rq) \ 770 if ((rq->bio)) \ 771 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) 772 773#define rq_for_each_segment(bvl, _rq, _iter) \ 774 __rq_for_each_bio(_iter.bio, _rq) \ 775 bio_for_each_segment(bvl, _iter.bio, _iter.iter) 776 777#define rq_iter_last(bvec, _iter) \ 778 (_iter.bio->bi_next == NULL && \ 779 bio_iter_last(bvec, _iter.iter)) 780 781#ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 782# error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 783#endif 784#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 785extern void rq_flush_dcache_pages(struct request *rq); 786#else 787static inline void rq_flush_dcache_pages(struct request *rq) 788{ 789} 790#endif 791 792extern int blk_register_queue(struct gendisk *disk); 793extern void blk_unregister_queue(struct gendisk *disk); 794extern void generic_make_request(struct bio *bio); 795extern void blk_rq_init(struct request_queue *q, struct request *rq); 796extern void blk_put_request(struct request *); 797extern void __blk_put_request(struct request_queue *, struct request *); 798extern struct request *blk_get_request(struct request_queue *, int, gfp_t); 799extern struct request *blk_make_request(struct request_queue *, struct bio *, 800 gfp_t); 801extern void blk_rq_set_block_pc(struct request *); 802extern void blk_requeue_request(struct request_queue *, struct request *); 803extern void blk_add_request_payload(struct request *rq, struct page *page, 804 unsigned int len); 805extern int blk_rq_check_limits(struct request_queue *q, struct request *rq); 806extern int blk_lld_busy(struct request_queue *q); 807extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, 808 struct bio_set *bs, gfp_t gfp_mask, 809 int (*bio_ctr)(struct bio *, struct bio *, void *), 810 void *data); 811extern void blk_rq_unprep_clone(struct request *rq); 812extern int blk_insert_cloned_request(struct request_queue *q, 813 struct request *rq); 814extern void blk_delay_queue(struct request_queue *, unsigned long); 815extern void blk_recount_segments(struct request_queue *, struct bio *); 816extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); 817extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, 818 unsigned int, void __user *); 819extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, 820 unsigned int, void __user *); 821extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, 822 struct scsi_ioctl_command __user *); 823 824/* 825 * A queue has just exitted congestion. Note this in the global counter of 826 * congested queues, and wake up anyone who was waiting for requests to be 827 * put back. 828 */ 829static inline void blk_clear_queue_congested(struct request_queue *q, int sync) 830{ 831 clear_bdi_congested(&q->backing_dev_info, sync); 832} 833 834/* 835 * A queue has just entered congestion. Flag that in the queue's VM-visible 836 * state flags and increment the global gounter of congested queues. 837 */ 838static inline void blk_set_queue_congested(struct request_queue *q, int sync) 839{ 840 set_bdi_congested(&q->backing_dev_info, sync); 841} 842 843extern void blk_start_queue(struct request_queue *q); 844extern void blk_stop_queue(struct request_queue *q); 845extern void blk_sync_queue(struct request_queue *q); 846extern void __blk_stop_queue(struct request_queue *q); 847extern void __blk_run_queue(struct request_queue *q); 848extern void blk_run_queue(struct request_queue *); 849extern void blk_run_queue_async(struct request_queue *q); 850extern int blk_rq_map_user(struct request_queue *, struct request *, 851 struct rq_map_data *, void __user *, unsigned long, 852 gfp_t); 853extern int blk_rq_unmap_user(struct bio *); 854extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); 855extern int blk_rq_map_user_iov(struct request_queue *, struct request *, 856 struct rq_map_data *, const struct iov_iter *, 857 gfp_t); 858extern int blk_execute_rq(struct request_queue *, struct gendisk *, 859 struct request *, int); 860extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, 861 struct request *, int, rq_end_io_fn *); 862 863static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 864{ 865 return bdev->bd_disk->queue; /* this is never NULL */ 866} 867 868/* 869 * blk_rq_pos() : the current sector 870 * blk_rq_bytes() : bytes left in the entire request 871 * blk_rq_cur_bytes() : bytes left in the current segment 872 * blk_rq_err_bytes() : bytes left till the next error boundary 873 * blk_rq_sectors() : sectors left in the entire request 874 * blk_rq_cur_sectors() : sectors left in the current segment 875 */ 876static inline sector_t blk_rq_pos(const struct request *rq) 877{ 878 return rq->__sector; 879} 880 881static inline unsigned int blk_rq_bytes(const struct request *rq) 882{ 883 return rq->__data_len; 884} 885 886static inline int blk_rq_cur_bytes(const struct request *rq) 887{ 888 return rq->bio ? bio_cur_bytes(rq->bio) : 0; 889} 890 891extern unsigned int blk_rq_err_bytes(const struct request *rq); 892 893static inline unsigned int blk_rq_sectors(const struct request *rq) 894{ 895 return blk_rq_bytes(rq) >> 9; 896} 897 898static inline unsigned int blk_rq_cur_sectors(const struct request *rq) 899{ 900 return blk_rq_cur_bytes(rq) >> 9; 901} 902 903static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, 904 unsigned int cmd_flags) 905{ 906 if (unlikely(cmd_flags & REQ_DISCARD)) 907 return min(q->limits.max_discard_sectors, UINT_MAX >> 9); 908 909 if (unlikely(cmd_flags & REQ_WRITE_SAME)) 910 return q->limits.max_write_same_sectors; 911 912 return q->limits.max_sectors; 913} 914 915/* 916 * Return maximum size of a request at given offset. Only valid for 917 * file system requests. 918 */ 919static inline unsigned int blk_max_size_offset(struct request_queue *q, 920 sector_t offset) 921{ 922 if (!q->limits.chunk_sectors) 923 return q->limits.max_sectors; 924 925 return q->limits.chunk_sectors - 926 (offset & (q->limits.chunk_sectors - 1)); 927} 928 929static inline unsigned int blk_rq_get_max_sectors(struct request *rq) 930{ 931 struct request_queue *q = rq->q; 932 933 if (unlikely(rq->cmd_type == REQ_TYPE_BLOCK_PC)) 934 return q->limits.max_hw_sectors; 935 936 if (!q->limits.chunk_sectors) 937 return blk_queue_get_max_sectors(q, rq->cmd_flags); 938 939 return min(blk_max_size_offset(q, blk_rq_pos(rq)), 940 blk_queue_get_max_sectors(q, rq->cmd_flags)); 941} 942 943static inline unsigned int blk_rq_count_bios(struct request *rq) 944{ 945 unsigned int nr_bios = 0; 946 struct bio *bio; 947 948 __rq_for_each_bio(bio, rq) 949 nr_bios++; 950 951 return nr_bios; 952} 953 954/* 955 * Request issue related functions. 956 */ 957extern struct request *blk_peek_request(struct request_queue *q); 958extern void blk_start_request(struct request *rq); 959extern struct request *blk_fetch_request(struct request_queue *q); 960 961/* 962 * Request completion related functions. 963 * 964 * blk_update_request() completes given number of bytes and updates 965 * the request without completing it. 966 * 967 * blk_end_request() and friends. __blk_end_request() must be called 968 * with the request queue spinlock acquired. 969 * 970 * Several drivers define their own end_request and call 971 * blk_end_request() for parts of the original function. 972 * This prevents code duplication in drivers. 973 */ 974extern bool blk_update_request(struct request *rq, int error, 975 unsigned int nr_bytes); 976extern void blk_finish_request(struct request *rq, int error); 977extern bool blk_end_request(struct request *rq, int error, 978 unsigned int nr_bytes); 979extern void blk_end_request_all(struct request *rq, int error); 980extern bool blk_end_request_cur(struct request *rq, int error); 981extern bool blk_end_request_err(struct request *rq, int error); 982extern bool __blk_end_request(struct request *rq, int error, 983 unsigned int nr_bytes); 984extern void __blk_end_request_all(struct request *rq, int error); 985extern bool __blk_end_request_cur(struct request *rq, int error); 986extern bool __blk_end_request_err(struct request *rq, int error); 987 988extern void blk_complete_request(struct request *); 989extern void __blk_complete_request(struct request *); 990extern void blk_abort_request(struct request *); 991extern void blk_unprep_request(struct request *); 992 993/* 994 * Access functions for manipulating queue properties 995 */ 996extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn, 997 spinlock_t *lock, int node_id); 998extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *); 999extern struct request_queue *blk_init_allocated_queue(struct request_queue *, 1000 request_fn_proc *, spinlock_t *); 1001extern void blk_cleanup_queue(struct request_queue *); 1002extern void blk_queue_make_request(struct request_queue *, make_request_fn *); 1003extern void blk_queue_bounce_limit(struct request_queue *, u64); 1004extern void blk_limits_max_hw_sectors(struct queue_limits *, unsigned int); 1005extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); 1006extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); 1007extern void blk_queue_max_segments(struct request_queue *, unsigned short); 1008extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); 1009extern void blk_queue_max_discard_sectors(struct request_queue *q, 1010 unsigned int max_discard_sectors); 1011extern void blk_queue_max_write_same_sectors(struct request_queue *q, 1012 unsigned int max_write_same_sectors); 1013extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); 1014extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); 1015extern void blk_queue_alignment_offset(struct request_queue *q, 1016 unsigned int alignment); 1017extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 1018extern void blk_queue_io_min(struct request_queue *q, unsigned int min); 1019extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 1020extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); 1021extern void blk_set_default_limits(struct queue_limits *lim); 1022extern void blk_set_stacking_limits(struct queue_limits *lim); 1023extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1024 sector_t offset); 1025extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, 1026 sector_t offset); 1027extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, 1028 sector_t offset); 1029extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); 1030extern void blk_queue_dma_pad(struct request_queue *, unsigned int); 1031extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 1032extern int blk_queue_dma_drain(struct request_queue *q, 1033 dma_drain_needed_fn *dma_drain_needed, 1034 void *buf, unsigned int size); 1035extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn); 1036extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); 1037extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn); 1038extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn); 1039extern void blk_queue_merge_bvec(struct request_queue *, merge_bvec_fn *); 1040extern void blk_queue_dma_alignment(struct request_queue *, int); 1041extern void blk_queue_update_dma_alignment(struct request_queue *, int); 1042extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *); 1043extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *); 1044extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1045extern void blk_queue_flush(struct request_queue *q, unsigned int flush); 1046extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable); 1047extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev); 1048 1049extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); 1050extern void blk_dump_rq_flags(struct request *, char *); 1051extern long nr_blockdev_pages(void); 1052 1053bool __must_check blk_get_queue(struct request_queue *); 1054struct request_queue *blk_alloc_queue(gfp_t); 1055struct request_queue *blk_alloc_queue_node(gfp_t, int); 1056extern void blk_put_queue(struct request_queue *); 1057 1058/* 1059 * block layer runtime pm functions 1060 */ 1061#ifdef CONFIG_PM 1062extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev); 1063extern int blk_pre_runtime_suspend(struct request_queue *q); 1064extern void blk_post_runtime_suspend(struct request_queue *q, int err); 1065extern void blk_pre_runtime_resume(struct request_queue *q); 1066extern void blk_post_runtime_resume(struct request_queue *q, int err); 1067#else 1068static inline void blk_pm_runtime_init(struct request_queue *q, 1069 struct device *dev) {} 1070static inline int blk_pre_runtime_suspend(struct request_queue *q) 1071{ 1072 return -ENOSYS; 1073} 1074static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {} 1075static inline void blk_pre_runtime_resume(struct request_queue *q) {} 1076static inline void blk_post_runtime_resume(struct request_queue *q, int err) {} 1077#endif 1078 1079/* 1080 * blk_plug permits building a queue of related requests by holding the I/O 1081 * fragments for a short period. This allows merging of sequential requests 1082 * into single larger request. As the requests are moved from a per-task list to 1083 * the device's request_queue in a batch, this results in improved scalability 1084 * as the lock contention for request_queue lock is reduced. 1085 * 1086 * It is ok not to disable preemption when adding the request to the plug list 1087 * or when attempting a merge, because blk_schedule_flush_list() will only flush 1088 * the plug list when the task sleeps by itself. For details, please see 1089 * schedule() where blk_schedule_flush_plug() is called. 1090 */ 1091struct blk_plug { 1092 struct list_head list; /* requests */ 1093 struct list_head mq_list; /* blk-mq requests */ 1094 struct list_head cb_list; /* md requires an unplug callback */ 1095}; 1096#define BLK_MAX_REQUEST_COUNT 16 1097 1098struct blk_plug_cb; 1099typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1100struct blk_plug_cb { 1101 struct list_head list; 1102 blk_plug_cb_fn callback; 1103 void *data; 1104}; 1105extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1106 void *data, int size); 1107extern void blk_start_plug(struct blk_plug *); 1108extern void blk_finish_plug(struct blk_plug *); 1109extern void blk_flush_plug_list(struct blk_plug *, bool); 1110 1111static inline void blk_flush_plug(struct task_struct *tsk) 1112{ 1113 struct blk_plug *plug = tsk->plug; 1114 1115 if (plug) 1116 blk_flush_plug_list(plug, false); 1117} 1118 1119static inline void blk_schedule_flush_plug(struct task_struct *tsk) 1120{ 1121 struct blk_plug *plug = tsk->plug; 1122 1123 if (plug) 1124 blk_flush_plug_list(plug, true); 1125} 1126 1127static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1128{ 1129 struct blk_plug *plug = tsk->plug; 1130 1131 return plug && 1132 (!list_empty(&plug->list) || 1133 !list_empty(&plug->mq_list) || 1134 !list_empty(&plug->cb_list)); 1135} 1136 1137/* 1138 * tag stuff 1139 */ 1140extern int blk_queue_start_tag(struct request_queue *, struct request *); 1141extern struct request *blk_queue_find_tag(struct request_queue *, int); 1142extern void blk_queue_end_tag(struct request_queue *, struct request *); 1143extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int); 1144extern void blk_queue_free_tags(struct request_queue *); 1145extern int blk_queue_resize_tags(struct request_queue *, int); 1146extern void blk_queue_invalidate_tags(struct request_queue *); 1147extern struct blk_queue_tag *blk_init_tags(int, int); 1148extern void blk_free_tags(struct blk_queue_tag *); 1149 1150static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt, 1151 int tag) 1152{ 1153 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth)) 1154 return NULL; 1155 return bqt->tag_index[tag]; 1156} 1157 1158#define BLKDEV_DISCARD_SECURE 0x01 /* secure discard */ 1159 1160extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); 1161extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1162 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); 1163extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, 1164 sector_t nr_sects, gfp_t gfp_mask, struct page *page); 1165extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1166 sector_t nr_sects, gfp_t gfp_mask, bool discard); 1167static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1168 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1169{ 1170 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9), 1171 nr_blocks << (sb->s_blocksize_bits - 9), 1172 gfp_mask, flags); 1173} 1174static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1175 sector_t nr_blocks, gfp_t gfp_mask) 1176{ 1177 return blkdev_issue_zeroout(sb->s_bdev, 1178 block << (sb->s_blocksize_bits - 9), 1179 nr_blocks << (sb->s_blocksize_bits - 9), 1180 gfp_mask, true); 1181} 1182 1183extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm); 1184 1185enum blk_default_limits { 1186 BLK_MAX_SEGMENTS = 128, 1187 BLK_SAFE_MAX_SECTORS = 255, 1188 BLK_MAX_SEGMENT_SIZE = 65536, 1189 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1190}; 1191 1192#define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist) 1193 1194static inline unsigned long queue_bounce_pfn(struct request_queue *q) 1195{ 1196 return q->limits.bounce_pfn; 1197} 1198 1199static inline unsigned long queue_segment_boundary(struct request_queue *q) 1200{ 1201 return q->limits.seg_boundary_mask; 1202} 1203 1204static inline unsigned int queue_max_sectors(struct request_queue *q) 1205{ 1206 return q->limits.max_sectors; 1207} 1208 1209static inline unsigned int queue_max_hw_sectors(struct request_queue *q) 1210{ 1211 return q->limits.max_hw_sectors; 1212} 1213 1214static inline unsigned short queue_max_segments(struct request_queue *q) 1215{ 1216 return q->limits.max_segments; 1217} 1218 1219static inline unsigned int queue_max_segment_size(struct request_queue *q) 1220{ 1221 return q->limits.max_segment_size; 1222} 1223 1224static inline unsigned short queue_logical_block_size(struct request_queue *q) 1225{ 1226 int retval = 512; 1227 1228 if (q && q->limits.logical_block_size) 1229 retval = q->limits.logical_block_size; 1230 1231 return retval; 1232} 1233 1234static inline unsigned short bdev_logical_block_size(struct block_device *bdev) 1235{ 1236 return queue_logical_block_size(bdev_get_queue(bdev)); 1237} 1238 1239static inline unsigned int queue_physical_block_size(struct request_queue *q) 1240{ 1241 return q->limits.physical_block_size; 1242} 1243 1244static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1245{ 1246 return queue_physical_block_size(bdev_get_queue(bdev)); 1247} 1248 1249static inline unsigned int queue_io_min(struct request_queue *q) 1250{ 1251 return q->limits.io_min; 1252} 1253 1254static inline int bdev_io_min(struct block_device *bdev) 1255{ 1256 return queue_io_min(bdev_get_queue(bdev)); 1257} 1258 1259static inline unsigned int queue_io_opt(struct request_queue *q) 1260{ 1261 return q->limits.io_opt; 1262} 1263 1264static inline int bdev_io_opt(struct block_device *bdev) 1265{ 1266 return queue_io_opt(bdev_get_queue(bdev)); 1267} 1268 1269static inline int queue_alignment_offset(struct request_queue *q) 1270{ 1271 if (q->limits.misaligned) 1272 return -1; 1273 1274 return q->limits.alignment_offset; 1275} 1276 1277static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) 1278{ 1279 unsigned int granularity = max(lim->physical_block_size, lim->io_min); 1280 unsigned int alignment = sector_div(sector, granularity >> 9) << 9; 1281 1282 return (granularity + lim->alignment_offset - alignment) % granularity; 1283} 1284 1285static inline int bdev_alignment_offset(struct block_device *bdev) 1286{ 1287 struct request_queue *q = bdev_get_queue(bdev); 1288 1289 if (q->limits.misaligned) 1290 return -1; 1291 1292 if (bdev != bdev->bd_contains) 1293 return bdev->bd_part->alignment_offset; 1294 1295 return q->limits.alignment_offset; 1296} 1297 1298static inline int queue_discard_alignment(struct request_queue *q) 1299{ 1300 if (q->limits.discard_misaligned) 1301 return -1; 1302 1303 return q->limits.discard_alignment; 1304} 1305 1306static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) 1307{ 1308 unsigned int alignment, granularity, offset; 1309 1310 if (!lim->max_discard_sectors) 1311 return 0; 1312 1313 /* Why are these in bytes, not sectors? */ 1314 alignment = lim->discard_alignment >> 9; 1315 granularity = lim->discard_granularity >> 9; 1316 if (!granularity) 1317 return 0; 1318 1319 /* Offset of the partition start in 'granularity' sectors */ 1320 offset = sector_div(sector, granularity); 1321 1322 /* And why do we do this modulus *again* in blkdev_issue_discard()? */ 1323 offset = (granularity + alignment - offset) % granularity; 1324 1325 /* Turn it back into bytes, gaah */ 1326 return offset << 9; 1327} 1328 1329static inline int bdev_discard_alignment(struct block_device *bdev) 1330{ 1331 struct request_queue *q = bdev_get_queue(bdev); 1332 1333 if (bdev != bdev->bd_contains) 1334 return bdev->bd_part->discard_alignment; 1335 1336 return q->limits.discard_alignment; 1337} 1338 1339static inline unsigned int queue_discard_zeroes_data(struct request_queue *q) 1340{ 1341 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1) 1342 return 1; 1343 1344 return 0; 1345} 1346 1347static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev) 1348{ 1349 return queue_discard_zeroes_data(bdev_get_queue(bdev)); 1350} 1351 1352static inline unsigned int bdev_write_same(struct block_device *bdev) 1353{ 1354 struct request_queue *q = bdev_get_queue(bdev); 1355 1356 if (q) 1357 return q->limits.max_write_same_sectors; 1358 1359 return 0; 1360} 1361 1362static inline int queue_dma_alignment(struct request_queue *q) 1363{ 1364 return q ? q->dma_alignment : 511; 1365} 1366 1367static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1368 unsigned int len) 1369{ 1370 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1371 return !(addr & alignment) && !(len & alignment); 1372} 1373 1374/* assumes size > 256 */ 1375static inline unsigned int blksize_bits(unsigned int size) 1376{ 1377 unsigned int bits = 8; 1378 do { 1379 bits++; 1380 size >>= 1; 1381 } while (size > 256); 1382 return bits; 1383} 1384 1385static inline unsigned int block_size(struct block_device *bdev) 1386{ 1387 return bdev->bd_block_size; 1388} 1389 1390static inline bool queue_flush_queueable(struct request_queue *q) 1391{ 1392 return !q->flush_not_queueable; 1393} 1394 1395typedef struct {struct page *v;} Sector; 1396 1397unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); 1398 1399static inline void put_dev_sector(Sector p) 1400{ 1401 page_cache_release(p.v); 1402} 1403 1404struct work_struct; 1405int kblockd_schedule_work(struct work_struct *work); 1406int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay); 1407int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1408 1409#ifdef CONFIG_BLK_CGROUP 1410/* 1411 * This should not be using sched_clock(). A real patch is in progress 1412 * to fix this up, until that is in place we need to disable preemption 1413 * around sched_clock() in this function and set_io_start_time_ns(). 1414 */ 1415static inline void set_start_time_ns(struct request *req) 1416{ 1417 preempt_disable(); 1418 req->start_time_ns = sched_clock(); 1419 preempt_enable(); 1420} 1421 1422static inline void set_io_start_time_ns(struct request *req) 1423{ 1424 preempt_disable(); 1425 req->io_start_time_ns = sched_clock(); 1426 preempt_enable(); 1427} 1428 1429static inline uint64_t rq_start_time_ns(struct request *req) 1430{ 1431 return req->start_time_ns; 1432} 1433 1434static inline uint64_t rq_io_start_time_ns(struct request *req) 1435{ 1436 return req->io_start_time_ns; 1437} 1438#else 1439static inline void set_start_time_ns(struct request *req) {} 1440static inline void set_io_start_time_ns(struct request *req) {} 1441static inline uint64_t rq_start_time_ns(struct request *req) 1442{ 1443 return 0; 1444} 1445static inline uint64_t rq_io_start_time_ns(struct request *req) 1446{ 1447 return 0; 1448} 1449#endif 1450 1451#define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1452 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1453#define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1454 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1455 1456#if defined(CONFIG_BLK_DEV_INTEGRITY) 1457 1458enum blk_integrity_flags { 1459 BLK_INTEGRITY_VERIFY = 1 << 0, 1460 BLK_INTEGRITY_GENERATE = 1 << 1, 1461 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, 1462 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, 1463}; 1464 1465struct blk_integrity_iter { 1466 void *prot_buf; 1467 void *data_buf; 1468 sector_t seed; 1469 unsigned int data_size; 1470 unsigned short interval; 1471 const char *disk_name; 1472}; 1473 1474typedef int (integrity_processing_fn) (struct blk_integrity_iter *); 1475 1476struct blk_integrity { 1477 integrity_processing_fn *generate_fn; 1478 integrity_processing_fn *verify_fn; 1479 1480 unsigned short flags; 1481 unsigned short tuple_size; 1482 unsigned short interval; 1483 unsigned short tag_size; 1484 1485 const char *name; 1486 1487 struct kobject kobj; 1488}; 1489 1490extern bool blk_integrity_is_initialized(struct gendisk *); 1491extern int blk_integrity_register(struct gendisk *, struct blk_integrity *); 1492extern void blk_integrity_unregister(struct gendisk *); 1493extern int blk_integrity_compare(struct gendisk *, struct gendisk *); 1494extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, 1495 struct scatterlist *); 1496extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); 1497extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, 1498 struct request *); 1499extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, 1500 struct bio *); 1501 1502static inline 1503struct blk_integrity *bdev_get_integrity(struct block_device *bdev) 1504{ 1505 return bdev->bd_disk->integrity; 1506} 1507 1508static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1509{ 1510 return disk->integrity; 1511} 1512 1513static inline bool blk_integrity_rq(struct request *rq) 1514{ 1515 return rq->cmd_flags & REQ_INTEGRITY; 1516} 1517 1518static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1519 unsigned int segs) 1520{ 1521 q->limits.max_integrity_segments = segs; 1522} 1523 1524static inline unsigned short 1525queue_max_integrity_segments(struct request_queue *q) 1526{ 1527 return q->limits.max_integrity_segments; 1528} 1529 1530#else /* CONFIG_BLK_DEV_INTEGRITY */ 1531 1532struct bio; 1533struct block_device; 1534struct gendisk; 1535struct blk_integrity; 1536 1537static inline int blk_integrity_rq(struct request *rq) 1538{ 1539 return 0; 1540} 1541static inline int blk_rq_count_integrity_sg(struct request_queue *q, 1542 struct bio *b) 1543{ 1544 return 0; 1545} 1546static inline int blk_rq_map_integrity_sg(struct request_queue *q, 1547 struct bio *b, 1548 struct scatterlist *s) 1549{ 1550 return 0; 1551} 1552static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) 1553{ 1554 return NULL; 1555} 1556static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1557{ 1558 return NULL; 1559} 1560static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) 1561{ 1562 return 0; 1563} 1564static inline int blk_integrity_register(struct gendisk *d, 1565 struct blk_integrity *b) 1566{ 1567 return 0; 1568} 1569static inline void blk_integrity_unregister(struct gendisk *d) 1570{ 1571} 1572static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1573 unsigned int segs) 1574{ 1575} 1576static inline unsigned short queue_max_integrity_segments(struct request_queue *q) 1577{ 1578 return 0; 1579} 1580static inline bool blk_integrity_merge_rq(struct request_queue *rq, 1581 struct request *r1, 1582 struct request *r2) 1583{ 1584 return true; 1585} 1586static inline bool blk_integrity_merge_bio(struct request_queue *rq, 1587 struct request *r, 1588 struct bio *b) 1589{ 1590 return true; 1591} 1592static inline bool blk_integrity_is_initialized(struct gendisk *g) 1593{ 1594 return 0; 1595} 1596 1597#endif /* CONFIG_BLK_DEV_INTEGRITY */ 1598 1599struct block_device_operations { 1600 int (*open) (struct block_device *, fmode_t); 1601 void (*release) (struct gendisk *, fmode_t); 1602 int (*rw_page)(struct block_device *, sector_t, struct page *, int rw); 1603 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1604 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1605 long (*direct_access)(struct block_device *, sector_t, 1606 void **, unsigned long *pfn, long size); 1607 unsigned int (*check_events) (struct gendisk *disk, 1608 unsigned int clearing); 1609 /* ->media_changed() is DEPRECATED, use ->check_events() instead */ 1610 int (*media_changed) (struct gendisk *); 1611 void (*unlock_native_capacity) (struct gendisk *); 1612 int (*revalidate_disk) (struct gendisk *); 1613 int (*getgeo)(struct block_device *, struct hd_geometry *); 1614 /* this callback is with swap_lock and sometimes page table lock held */ 1615 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1616 struct module *owner; 1617}; 1618 1619extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, 1620 unsigned long); 1621extern int bdev_read_page(struct block_device *, sector_t, struct page *); 1622extern int bdev_write_page(struct block_device *, sector_t, struct page *, 1623 struct writeback_control *); 1624extern long bdev_direct_access(struct block_device *, sector_t, void **addr, 1625 unsigned long *pfn, long size); 1626#else /* CONFIG_BLOCK */ 1627 1628struct block_device; 1629 1630/* 1631 * stubs for when the block layer is configured out 1632 */ 1633#define buffer_heads_over_limit 0 1634 1635static inline long nr_blockdev_pages(void) 1636{ 1637 return 0; 1638} 1639 1640struct blk_plug { 1641}; 1642 1643static inline void blk_start_plug(struct blk_plug *plug) 1644{ 1645} 1646 1647static inline void blk_finish_plug(struct blk_plug *plug) 1648{ 1649} 1650 1651static inline void blk_flush_plug(struct task_struct *task) 1652{ 1653} 1654 1655static inline void blk_schedule_flush_plug(struct task_struct *task) 1656{ 1657} 1658 1659 1660static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1661{ 1662 return false; 1663} 1664 1665static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, 1666 sector_t *error_sector) 1667{ 1668 return 0; 1669} 1670 1671#endif /* CONFIG_BLOCK */ 1672 1673#endif 1674