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-defs.h> 16 #include <linux/wait.h> 17 #include <linux/mempool.h> 18 #include <linux/pfn.h> 19 #include <linux/bio.h> 20 #include <linux/stringify.h> 21 #include <linux/gfp.h> 22 #include <linux/bsg.h> 23 #include <linux/smp.h> 24 #include <linux/rcupdate.h> 25 #include <linux/percpu-refcount.h> 26 #include <linux/scatterlist.h> 27 #include <linux/blkzoned.h> 28 29 struct module; 30 struct scsi_ioctl_command; 31 32 struct request_queue; 33 struct elevator_queue; 34 struct blk_trace; 35 struct request; 36 struct sg_io_hdr; 37 struct bsg_job; 38 struct blkcg_gq; 39 struct blk_flush_queue; 40 struct pr_ops; 41 struct rq_wb; 42 43 #define BLKDEV_MIN_RQ 4 44 #define BLKDEV_MAX_RQ 128 /* Default maximum */ 45 46 /* 47 * Maximum number of blkcg policies allowed to be registered concurrently. 48 * Defined here to simplify include dependency. 49 */ 50 #define BLKCG_MAX_POLS 2 51 52 typedef void (rq_end_io_fn)(struct request *, int); 53 54 #define BLK_RL_SYNCFULL (1U << 0) 55 #define BLK_RL_ASYNCFULL (1U << 1) 56 57 struct request_list { 58 struct request_queue *q; /* the queue this rl belongs to */ 59 #ifdef CONFIG_BLK_CGROUP 60 struct blkcg_gq *blkg; /* blkg this request pool belongs to */ 61 #endif 62 /* 63 * count[], starved[], and wait[] are indexed by 64 * BLK_RW_SYNC/BLK_RW_ASYNC 65 */ 66 int count[2]; 67 int starved[2]; 68 mempool_t *rq_pool; 69 wait_queue_head_t wait[2]; 70 unsigned int flags; 71 }; 72 73 /* 74 * request command types 75 */ 76 enum rq_cmd_type_bits { 77 REQ_TYPE_FS = 1, /* fs request */ 78 REQ_TYPE_BLOCK_PC, /* scsi command */ 79 REQ_TYPE_DRV_PRIV, /* driver defined types from here */ 80 }; 81 82 /* 83 * request flags */ 84 typedef __u32 __bitwise req_flags_t; 85 86 /* elevator knows about this request */ 87 #define RQF_SORTED ((__force req_flags_t)(1 << 0)) 88 /* drive already may have started this one */ 89 #define RQF_STARTED ((__force req_flags_t)(1 << 1)) 90 /* uses tagged queueing */ 91 #define RQF_QUEUED ((__force req_flags_t)(1 << 2)) 92 /* may not be passed by ioscheduler */ 93 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3)) 94 /* request for flush sequence */ 95 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4)) 96 /* merge of different types, fail separately */ 97 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5)) 98 /* track inflight for MQ */ 99 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6)) 100 /* don't call prep for this one */ 101 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7)) 102 /* set for "ide_preempt" requests and also for requests for which the SCSI 103 "quiesce" state must be ignored. */ 104 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8)) 105 /* contains copies of user pages */ 106 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9)) 107 /* vaguely specified driver internal error. Ignored by the block layer */ 108 #define RQF_FAILED ((__force req_flags_t)(1 << 10)) 109 /* don't warn about errors */ 110 #define RQF_QUIET ((__force req_flags_t)(1 << 11)) 111 /* elevator private data attached */ 112 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12)) 113 /* account I/O stat */ 114 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13)) 115 /* request came from our alloc pool */ 116 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14)) 117 /* runtime pm request */ 118 #define RQF_PM ((__force req_flags_t)(1 << 15)) 119 /* on IO scheduler merge hash */ 120 #define RQF_HASHED ((__force req_flags_t)(1 << 16)) 121 /* IO stats tracking on */ 122 #define RQF_STATS ((__force req_flags_t)(1 << 17)) 123 124 /* flags that prevent us from merging requests: */ 125 #define RQF_NOMERGE_FLAGS \ 126 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ) 127 128 #define BLK_MAX_CDB 16 129 130 /* 131 * Try to put the fields that are referenced together in the same cacheline. 132 * 133 * If you modify this structure, make sure to update blk_rq_init() and 134 * especially blk_mq_rq_ctx_init() to take care of the added fields. 135 */ 136 struct request { 137 struct list_head queuelist; 138 union { 139 struct call_single_data csd; 140 u64 fifo_time; 141 }; 142 143 struct request_queue *q; 144 struct blk_mq_ctx *mq_ctx; 145 146 int cpu; 147 unsigned cmd_type; 148 unsigned int cmd_flags; /* op and common flags */ 149 req_flags_t rq_flags; 150 unsigned long atomic_flags; 151 152 /* the following two fields are internal, NEVER access directly */ 153 unsigned int __data_len; /* total data len */ 154 sector_t __sector; /* sector cursor */ 155 156 struct bio *bio; 157 struct bio *biotail; 158 159 /* 160 * The hash is used inside the scheduler, and killed once the 161 * request reaches the dispatch list. The ipi_list is only used 162 * to queue the request for softirq completion, which is long 163 * after the request has been unhashed (and even removed from 164 * the dispatch list). 165 */ 166 union { 167 struct hlist_node hash; /* merge hash */ 168 struct list_head ipi_list; 169 }; 170 171 /* 172 * The rb_node is only used inside the io scheduler, requests 173 * are pruned when moved to the dispatch queue. So let the 174 * completion_data share space with the rb_node. 175 */ 176 union { 177 struct rb_node rb_node; /* sort/lookup */ 178 void *completion_data; 179 }; 180 181 /* 182 * Three pointers are available for the IO schedulers, if they need 183 * more they have to dynamically allocate it. Flush requests are 184 * never put on the IO scheduler. So let the flush fields share 185 * space with the elevator data. 186 */ 187 union { 188 struct { 189 struct io_cq *icq; 190 void *priv[2]; 191 } elv; 192 193 struct { 194 unsigned int seq; 195 struct list_head list; 196 rq_end_io_fn *saved_end_io; 197 } flush; 198 }; 199 200 struct gendisk *rq_disk; 201 struct hd_struct *part; 202 unsigned long start_time; 203 struct blk_issue_stat issue_stat; 204 #ifdef CONFIG_BLK_CGROUP 205 struct request_list *rl; /* rl this rq is alloced from */ 206 unsigned long long start_time_ns; 207 unsigned long long io_start_time_ns; /* when passed to hardware */ 208 #endif 209 /* Number of scatter-gather DMA addr+len pairs after 210 * physical address coalescing is performed. 211 */ 212 unsigned short nr_phys_segments; 213 #if defined(CONFIG_BLK_DEV_INTEGRITY) 214 unsigned short nr_integrity_segments; 215 #endif 216 217 unsigned short ioprio; 218 219 void *special; /* opaque pointer available for LLD use */ 220 221 int tag; 222 int errors; 223 224 /* 225 * when request is used as a packet command carrier 226 */ 227 unsigned char __cmd[BLK_MAX_CDB]; 228 unsigned char *cmd; 229 unsigned short cmd_len; 230 231 unsigned int extra_len; /* length of alignment and padding */ 232 unsigned int sense_len; 233 unsigned int resid_len; /* residual count */ 234 void *sense; 235 236 unsigned long deadline; 237 struct list_head timeout_list; 238 unsigned int timeout; 239 int retries; 240 241 /* 242 * completion callback. 243 */ 244 rq_end_io_fn *end_io; 245 void *end_io_data; 246 247 /* for bidi */ 248 struct request *next_rq; 249 }; 250 251 static inline unsigned short req_get_ioprio(struct request *req) 252 { 253 return req->ioprio; 254 } 255 256 #include <linux/elevator.h> 257 258 struct blk_queue_ctx; 259 260 typedef void (request_fn_proc) (struct request_queue *q); 261 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio); 262 typedef int (prep_rq_fn) (struct request_queue *, struct request *); 263 typedef void (unprep_rq_fn) (struct request_queue *, struct request *); 264 265 struct bio_vec; 266 typedef void (softirq_done_fn)(struct request *); 267 typedef int (dma_drain_needed_fn)(struct request *); 268 typedef int (lld_busy_fn) (struct request_queue *q); 269 typedef int (bsg_job_fn) (struct bsg_job *); 270 271 enum blk_eh_timer_return { 272 BLK_EH_NOT_HANDLED, 273 BLK_EH_HANDLED, 274 BLK_EH_RESET_TIMER, 275 }; 276 277 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *); 278 279 enum blk_queue_state { 280 Queue_down, 281 Queue_up, 282 }; 283 284 struct blk_queue_tag { 285 struct request **tag_index; /* map of busy tags */ 286 unsigned long *tag_map; /* bit map of free/busy tags */ 287 int busy; /* current depth */ 288 int max_depth; /* what we will send to device */ 289 int real_max_depth; /* what the array can hold */ 290 atomic_t refcnt; /* map can be shared */ 291 int alloc_policy; /* tag allocation policy */ 292 int next_tag; /* next tag */ 293 }; 294 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ 295 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ 296 297 #define BLK_SCSI_MAX_CMDS (256) 298 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) 299 300 /* 301 * Zoned block device models (zoned limit). 302 */ 303 enum blk_zoned_model { 304 BLK_ZONED_NONE, /* Regular block device */ 305 BLK_ZONED_HA, /* Host-aware zoned block device */ 306 BLK_ZONED_HM, /* Host-managed zoned block device */ 307 }; 308 309 struct queue_limits { 310 unsigned long bounce_pfn; 311 unsigned long seg_boundary_mask; 312 unsigned long virt_boundary_mask; 313 314 unsigned int max_hw_sectors; 315 unsigned int max_dev_sectors; 316 unsigned int chunk_sectors; 317 unsigned int max_sectors; 318 unsigned int max_segment_size; 319 unsigned int physical_block_size; 320 unsigned int alignment_offset; 321 unsigned int io_min; 322 unsigned int io_opt; 323 unsigned int max_discard_sectors; 324 unsigned int max_hw_discard_sectors; 325 unsigned int max_write_same_sectors; 326 unsigned int max_write_zeroes_sectors; 327 unsigned int discard_granularity; 328 unsigned int discard_alignment; 329 330 unsigned short logical_block_size; 331 unsigned short max_segments; 332 unsigned short max_integrity_segments; 333 334 unsigned char misaligned; 335 unsigned char discard_misaligned; 336 unsigned char cluster; 337 unsigned char discard_zeroes_data; 338 unsigned char raid_partial_stripes_expensive; 339 enum blk_zoned_model zoned; 340 }; 341 342 #ifdef CONFIG_BLK_DEV_ZONED 343 344 struct blk_zone_report_hdr { 345 unsigned int nr_zones; 346 u8 padding[60]; 347 }; 348 349 extern int blkdev_report_zones(struct block_device *bdev, 350 sector_t sector, struct blk_zone *zones, 351 unsigned int *nr_zones, gfp_t gfp_mask); 352 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors, 353 sector_t nr_sectors, gfp_t gfp_mask); 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 int blkdev_report_zones_ioctl(struct block_device *bdev, 363 fmode_t mode, unsigned int cmd, 364 unsigned long arg) 365 { 366 return -ENOTTY; 367 } 368 369 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev, 370 fmode_t mode, unsigned int cmd, 371 unsigned long arg) 372 { 373 return -ENOTTY; 374 } 375 376 #endif /* CONFIG_BLK_DEV_ZONED */ 377 378 struct request_queue { 379 /* 380 * Together with queue_head for cacheline sharing 381 */ 382 struct list_head queue_head; 383 struct request *last_merge; 384 struct elevator_queue *elevator; 385 int nr_rqs[2]; /* # allocated [a]sync rqs */ 386 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */ 387 388 struct rq_wb *rq_wb; 389 390 /* 391 * If blkcg is not used, @q->root_rl serves all requests. If blkcg 392 * is used, root blkg allocates from @q->root_rl and all other 393 * blkgs from their own blkg->rl. Which one to use should be 394 * determined using bio_request_list(). 395 */ 396 struct request_list root_rl; 397 398 request_fn_proc *request_fn; 399 make_request_fn *make_request_fn; 400 prep_rq_fn *prep_rq_fn; 401 unprep_rq_fn *unprep_rq_fn; 402 softirq_done_fn *softirq_done_fn; 403 rq_timed_out_fn *rq_timed_out_fn; 404 dma_drain_needed_fn *dma_drain_needed; 405 lld_busy_fn *lld_busy_fn; 406 407 struct blk_mq_ops *mq_ops; 408 409 unsigned int *mq_map; 410 411 /* sw queues */ 412 struct blk_mq_ctx __percpu *queue_ctx; 413 unsigned int nr_queues; 414 415 unsigned int queue_depth; 416 417 /* hw dispatch queues */ 418 struct blk_mq_hw_ctx **queue_hw_ctx; 419 unsigned int nr_hw_queues; 420 421 /* 422 * Dispatch queue sorting 423 */ 424 sector_t end_sector; 425 struct request *boundary_rq; 426 427 /* 428 * Delayed queue handling 429 */ 430 struct delayed_work delay_work; 431 432 struct backing_dev_info backing_dev_info; 433 434 /* 435 * The queue owner gets to use this for whatever they like. 436 * ll_rw_blk doesn't touch it. 437 */ 438 void *queuedata; 439 440 /* 441 * various queue flags, see QUEUE_* below 442 */ 443 unsigned long queue_flags; 444 445 /* 446 * ida allocated id for this queue. Used to index queues from 447 * ioctx. 448 */ 449 int id; 450 451 /* 452 * queue needs bounce pages for pages above this limit 453 */ 454 gfp_t bounce_gfp; 455 456 /* 457 * protects queue structures from reentrancy. ->__queue_lock should 458 * _never_ be used directly, it is queue private. always use 459 * ->queue_lock. 460 */ 461 spinlock_t __queue_lock; 462 spinlock_t *queue_lock; 463 464 /* 465 * queue kobject 466 */ 467 struct kobject kobj; 468 469 /* 470 * mq queue kobject 471 */ 472 struct kobject mq_kobj; 473 474 #ifdef CONFIG_BLK_DEV_INTEGRITY 475 struct blk_integrity integrity; 476 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 477 478 #ifdef CONFIG_PM 479 struct device *dev; 480 int rpm_status; 481 unsigned int nr_pending; 482 #endif 483 484 /* 485 * queue settings 486 */ 487 unsigned long nr_requests; /* Max # of requests */ 488 unsigned int nr_congestion_on; 489 unsigned int nr_congestion_off; 490 unsigned int nr_batching; 491 492 unsigned int dma_drain_size; 493 void *dma_drain_buffer; 494 unsigned int dma_pad_mask; 495 unsigned int dma_alignment; 496 497 struct blk_queue_tag *queue_tags; 498 struct list_head tag_busy_list; 499 500 unsigned int nr_sorted; 501 unsigned int in_flight[2]; 502 503 struct blk_rq_stat rq_stats[2]; 504 505 /* 506 * Number of active block driver functions for which blk_drain_queue() 507 * must wait. Must be incremented around functions that unlock the 508 * queue_lock internally, e.g. scsi_request_fn(). 509 */ 510 unsigned int request_fn_active; 511 512 unsigned int rq_timeout; 513 int poll_nsec; 514 struct timer_list timeout; 515 struct work_struct timeout_work; 516 struct list_head timeout_list; 517 518 struct list_head icq_list; 519 #ifdef CONFIG_BLK_CGROUP 520 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 521 struct blkcg_gq *root_blkg; 522 struct list_head blkg_list; 523 #endif 524 525 struct queue_limits limits; 526 527 /* 528 * sg stuff 529 */ 530 unsigned int sg_timeout; 531 unsigned int sg_reserved_size; 532 int node; 533 #ifdef CONFIG_BLK_DEV_IO_TRACE 534 struct blk_trace *blk_trace; 535 #endif 536 /* 537 * for flush operations 538 */ 539 struct blk_flush_queue *fq; 540 541 struct list_head requeue_list; 542 spinlock_t requeue_lock; 543 struct delayed_work requeue_work; 544 545 struct mutex sysfs_lock; 546 547 int bypass_depth; 548 atomic_t mq_freeze_depth; 549 550 #if defined(CONFIG_BLK_DEV_BSG) 551 bsg_job_fn *bsg_job_fn; 552 int bsg_job_size; 553 struct bsg_class_device bsg_dev; 554 #endif 555 556 #ifdef CONFIG_BLK_DEV_THROTTLING 557 /* Throttle data */ 558 struct throtl_data *td; 559 #endif 560 struct rcu_head rcu_head; 561 wait_queue_head_t mq_freeze_wq; 562 struct percpu_ref q_usage_counter; 563 struct list_head all_q_node; 564 565 struct blk_mq_tag_set *tag_set; 566 struct list_head tag_set_list; 567 struct bio_set *bio_split; 568 569 bool mq_sysfs_init_done; 570 }; 571 572 #define QUEUE_FLAG_QUEUED 1 /* uses generic tag queueing */ 573 #define QUEUE_FLAG_STOPPED 2 /* queue is stopped */ 574 #define QUEUE_FLAG_SYNCFULL 3 /* read queue has been filled */ 575 #define QUEUE_FLAG_ASYNCFULL 4 /* write queue has been filled */ 576 #define QUEUE_FLAG_DYING 5 /* queue being torn down */ 577 #define QUEUE_FLAG_BYPASS 6 /* act as dumb FIFO queue */ 578 #define QUEUE_FLAG_BIDI 7 /* queue supports bidi requests */ 579 #define QUEUE_FLAG_NOMERGES 8 /* disable merge attempts */ 580 #define QUEUE_FLAG_SAME_COMP 9 /* complete on same CPU-group */ 581 #define QUEUE_FLAG_FAIL_IO 10 /* fake timeout */ 582 #define QUEUE_FLAG_STACKABLE 11 /* supports request stacking */ 583 #define QUEUE_FLAG_NONROT 12 /* non-rotational device (SSD) */ 584 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ 585 #define QUEUE_FLAG_IO_STAT 13 /* do IO stats */ 586 #define QUEUE_FLAG_DISCARD 14 /* supports DISCARD */ 587 #define QUEUE_FLAG_NOXMERGES 15 /* No extended merges */ 588 #define QUEUE_FLAG_ADD_RANDOM 16 /* Contributes to random pool */ 589 #define QUEUE_FLAG_SECERASE 17 /* supports secure erase */ 590 #define QUEUE_FLAG_SAME_FORCE 18 /* force complete on same CPU */ 591 #define QUEUE_FLAG_DEAD 19 /* queue tear-down finished */ 592 #define QUEUE_FLAG_INIT_DONE 20 /* queue is initialized */ 593 #define QUEUE_FLAG_NO_SG_MERGE 21 /* don't attempt to merge SG segments*/ 594 #define QUEUE_FLAG_POLL 22 /* IO polling enabled if set */ 595 #define QUEUE_FLAG_WC 23 /* Write back caching */ 596 #define QUEUE_FLAG_FUA 24 /* device supports FUA writes */ 597 #define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */ 598 #define QUEUE_FLAG_DAX 26 /* device supports DAX */ 599 #define QUEUE_FLAG_STATS 27 /* track rq completion times */ 600 601 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 602 (1 << QUEUE_FLAG_STACKABLE) | \ 603 (1 << QUEUE_FLAG_SAME_COMP) | \ 604 (1 << QUEUE_FLAG_ADD_RANDOM)) 605 606 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 607 (1 << QUEUE_FLAG_STACKABLE) | \ 608 (1 << QUEUE_FLAG_SAME_COMP) | \ 609 (1 << QUEUE_FLAG_POLL)) 610 611 static inline void queue_lockdep_assert_held(struct request_queue *q) 612 { 613 if (q->queue_lock) 614 lockdep_assert_held(q->queue_lock); 615 } 616 617 static inline void queue_flag_set_unlocked(unsigned int flag, 618 struct request_queue *q) 619 { 620 __set_bit(flag, &q->queue_flags); 621 } 622 623 static inline int queue_flag_test_and_clear(unsigned int flag, 624 struct request_queue *q) 625 { 626 queue_lockdep_assert_held(q); 627 628 if (test_bit(flag, &q->queue_flags)) { 629 __clear_bit(flag, &q->queue_flags); 630 return 1; 631 } 632 633 return 0; 634 } 635 636 static inline int queue_flag_test_and_set(unsigned int flag, 637 struct request_queue *q) 638 { 639 queue_lockdep_assert_held(q); 640 641 if (!test_bit(flag, &q->queue_flags)) { 642 __set_bit(flag, &q->queue_flags); 643 return 0; 644 } 645 646 return 1; 647 } 648 649 static inline void queue_flag_set(unsigned int flag, struct request_queue *q) 650 { 651 queue_lockdep_assert_held(q); 652 __set_bit(flag, &q->queue_flags); 653 } 654 655 static inline void queue_flag_clear_unlocked(unsigned int flag, 656 struct request_queue *q) 657 { 658 __clear_bit(flag, &q->queue_flags); 659 } 660 661 static inline int queue_in_flight(struct request_queue *q) 662 { 663 return q->in_flight[0] + q->in_flight[1]; 664 } 665 666 static inline void queue_flag_clear(unsigned int flag, struct request_queue *q) 667 { 668 queue_lockdep_assert_held(q); 669 __clear_bit(flag, &q->queue_flags); 670 } 671 672 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags) 673 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 674 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 675 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) 676 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags) 677 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 678 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 679 #define blk_queue_noxmerges(q) \ 680 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 681 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) 682 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) 683 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) 684 #define blk_queue_stackable(q) \ 685 test_bit(QUEUE_FLAG_STACKABLE, &(q)->queue_flags) 686 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) 687 #define blk_queue_secure_erase(q) \ 688 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags)) 689 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) 690 691 #define blk_noretry_request(rq) \ 692 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 693 REQ_FAILFAST_DRIVER)) 694 695 #define blk_account_rq(rq) \ 696 (((rq)->rq_flags & RQF_STARTED) && \ 697 ((rq)->cmd_type == REQ_TYPE_FS)) 698 699 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1) 700 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL) 701 /* rq->queuelist of dequeued request must be list_empty() */ 702 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist)) 703 704 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 705 706 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ) 707 708 /* 709 * Driver can handle struct request, if it either has an old style 710 * request_fn defined, or is blk-mq based. 711 */ 712 static inline bool queue_is_rq_based(struct request_queue *q) 713 { 714 return q->request_fn || q->mq_ops; 715 } 716 717 static inline unsigned int blk_queue_cluster(struct request_queue *q) 718 { 719 return q->limits.cluster; 720 } 721 722 static inline enum blk_zoned_model 723 blk_queue_zoned_model(struct request_queue *q) 724 { 725 return q->limits.zoned; 726 } 727 728 static inline bool blk_queue_is_zoned(struct request_queue *q) 729 { 730 switch (blk_queue_zoned_model(q)) { 731 case BLK_ZONED_HA: 732 case BLK_ZONED_HM: 733 return true; 734 default: 735 return false; 736 } 737 } 738 739 static inline unsigned int blk_queue_zone_size(struct request_queue *q) 740 { 741 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; 742 } 743 744 static inline bool rq_is_sync(struct request *rq) 745 { 746 return op_is_sync(rq->cmd_flags); 747 } 748 749 static inline bool blk_rl_full(struct request_list *rl, bool sync) 750 { 751 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 752 753 return rl->flags & flag; 754 } 755 756 static inline void blk_set_rl_full(struct request_list *rl, bool sync) 757 { 758 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 759 760 rl->flags |= flag; 761 } 762 763 static inline void blk_clear_rl_full(struct request_list *rl, bool sync) 764 { 765 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 766 767 rl->flags &= ~flag; 768 } 769 770 static inline bool rq_mergeable(struct request *rq) 771 { 772 if (rq->cmd_type != REQ_TYPE_FS) 773 return false; 774 775 if (req_op(rq) == REQ_OP_FLUSH) 776 return false; 777 778 if (req_op(rq) == REQ_OP_WRITE_ZEROES) 779 return false; 780 781 if (rq->cmd_flags & REQ_NOMERGE_FLAGS) 782 return false; 783 if (rq->rq_flags & RQF_NOMERGE_FLAGS) 784 return false; 785 786 return true; 787 } 788 789 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) 790 { 791 if (bio_data(a) == bio_data(b)) 792 return true; 793 794 return false; 795 } 796 797 static inline unsigned int blk_queue_depth(struct request_queue *q) 798 { 799 if (q->queue_depth) 800 return q->queue_depth; 801 802 return q->nr_requests; 803 } 804 805 /* 806 * q->prep_rq_fn return values 807 */ 808 enum { 809 BLKPREP_OK, /* serve it */ 810 BLKPREP_KILL, /* fatal error, kill, return -EIO */ 811 BLKPREP_DEFER, /* leave on queue */ 812 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */ 813 }; 814 815 extern unsigned long blk_max_low_pfn, blk_max_pfn; 816 817 /* 818 * standard bounce addresses: 819 * 820 * BLK_BOUNCE_HIGH : bounce all highmem pages 821 * BLK_BOUNCE_ANY : don't bounce anything 822 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary 823 */ 824 825 #if BITS_PER_LONG == 32 826 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) 827 #else 828 #define BLK_BOUNCE_HIGH -1ULL 829 #endif 830 #define BLK_BOUNCE_ANY (-1ULL) 831 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) 832 833 /* 834 * default timeout for SG_IO if none specified 835 */ 836 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 837 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 838 839 #ifdef CONFIG_BOUNCE 840 extern int init_emergency_isa_pool(void); 841 extern void blk_queue_bounce(struct request_queue *q, struct bio **bio); 842 #else 843 static inline int init_emergency_isa_pool(void) 844 { 845 return 0; 846 } 847 static inline void blk_queue_bounce(struct request_queue *q, struct bio **bio) 848 { 849 } 850 #endif /* CONFIG_MMU */ 851 852 struct rq_map_data { 853 struct page **pages; 854 int page_order; 855 int nr_entries; 856 unsigned long offset; 857 int null_mapped; 858 int from_user; 859 }; 860 861 struct req_iterator { 862 struct bvec_iter iter; 863 struct bio *bio; 864 }; 865 866 /* This should not be used directly - use rq_for_each_segment */ 867 #define for_each_bio(_bio) \ 868 for (; _bio; _bio = _bio->bi_next) 869 #define __rq_for_each_bio(_bio, rq) \ 870 if ((rq->bio)) \ 871 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) 872 873 #define rq_for_each_segment(bvl, _rq, _iter) \ 874 __rq_for_each_bio(_iter.bio, _rq) \ 875 bio_for_each_segment(bvl, _iter.bio, _iter.iter) 876 877 #define rq_iter_last(bvec, _iter) \ 878 (_iter.bio->bi_next == NULL && \ 879 bio_iter_last(bvec, _iter.iter)) 880 881 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 882 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 883 #endif 884 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 885 extern void rq_flush_dcache_pages(struct request *rq); 886 #else 887 static inline void rq_flush_dcache_pages(struct request *rq) 888 { 889 } 890 #endif 891 892 #ifdef CONFIG_PRINTK 893 #define vfs_msg(sb, level, fmt, ...) \ 894 __vfs_msg(sb, level, fmt, ##__VA_ARGS__) 895 #else 896 #define vfs_msg(sb, level, fmt, ...) \ 897 do { \ 898 no_printk(fmt, ##__VA_ARGS__); \ 899 __vfs_msg(sb, "", " "); \ 900 } while (0) 901 #endif 902 903 extern int blk_register_queue(struct gendisk *disk); 904 extern void blk_unregister_queue(struct gendisk *disk); 905 extern blk_qc_t generic_make_request(struct bio *bio); 906 extern void blk_rq_init(struct request_queue *q, struct request *rq); 907 extern void blk_put_request(struct request *); 908 extern void __blk_put_request(struct request_queue *, struct request *); 909 extern struct request *blk_get_request(struct request_queue *, int, gfp_t); 910 extern void blk_rq_set_block_pc(struct request *); 911 extern void blk_requeue_request(struct request_queue *, struct request *); 912 extern void blk_add_request_payload(struct request *rq, struct page *page, 913 int offset, unsigned int len); 914 extern int blk_lld_busy(struct request_queue *q); 915 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, 916 struct bio_set *bs, gfp_t gfp_mask, 917 int (*bio_ctr)(struct bio *, struct bio *, void *), 918 void *data); 919 extern void blk_rq_unprep_clone(struct request *rq); 920 extern int blk_insert_cloned_request(struct request_queue *q, 921 struct request *rq); 922 extern int blk_rq_append_bio(struct request *rq, struct bio *bio); 923 extern void blk_delay_queue(struct request_queue *, unsigned long); 924 extern void blk_queue_split(struct request_queue *, struct bio **, 925 struct bio_set *); 926 extern void blk_recount_segments(struct request_queue *, struct bio *); 927 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); 928 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, 929 unsigned int, void __user *); 930 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, 931 unsigned int, void __user *); 932 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, 933 struct scsi_ioctl_command __user *); 934 935 extern int blk_queue_enter(struct request_queue *q, bool nowait); 936 extern void blk_queue_exit(struct request_queue *q); 937 extern void blk_start_queue(struct request_queue *q); 938 extern void blk_start_queue_async(struct request_queue *q); 939 extern void blk_stop_queue(struct request_queue *q); 940 extern void blk_sync_queue(struct request_queue *q); 941 extern void __blk_stop_queue(struct request_queue *q); 942 extern void __blk_run_queue(struct request_queue *q); 943 extern void __blk_run_queue_uncond(struct request_queue *q); 944 extern void blk_run_queue(struct request_queue *); 945 extern void blk_run_queue_async(struct request_queue *q); 946 extern void blk_mq_quiesce_queue(struct request_queue *q); 947 extern int blk_rq_map_user(struct request_queue *, struct request *, 948 struct rq_map_data *, void __user *, unsigned long, 949 gfp_t); 950 extern int blk_rq_unmap_user(struct bio *); 951 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); 952 extern int blk_rq_map_user_iov(struct request_queue *, struct request *, 953 struct rq_map_data *, const struct iov_iter *, 954 gfp_t); 955 extern int blk_execute_rq(struct request_queue *, struct gendisk *, 956 struct request *, int); 957 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, 958 struct request *, int, rq_end_io_fn *); 959 960 bool blk_mq_poll(struct request_queue *q, blk_qc_t cookie); 961 962 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 963 { 964 return bdev->bd_disk->queue; /* this is never NULL */ 965 } 966 967 /* 968 * blk_rq_pos() : the current sector 969 * blk_rq_bytes() : bytes left in the entire request 970 * blk_rq_cur_bytes() : bytes left in the current segment 971 * blk_rq_err_bytes() : bytes left till the next error boundary 972 * blk_rq_sectors() : sectors left in the entire request 973 * blk_rq_cur_sectors() : sectors left in the current segment 974 */ 975 static inline sector_t blk_rq_pos(const struct request *rq) 976 { 977 return rq->__sector; 978 } 979 980 static inline unsigned int blk_rq_bytes(const struct request *rq) 981 { 982 return rq->__data_len; 983 } 984 985 static inline int blk_rq_cur_bytes(const struct request *rq) 986 { 987 return rq->bio ? bio_cur_bytes(rq->bio) : 0; 988 } 989 990 extern unsigned int blk_rq_err_bytes(const struct request *rq); 991 992 static inline unsigned int blk_rq_sectors(const struct request *rq) 993 { 994 return blk_rq_bytes(rq) >> 9; 995 } 996 997 static inline unsigned int blk_rq_cur_sectors(const struct request *rq) 998 { 999 return blk_rq_cur_bytes(rq) >> 9; 1000 } 1001 1002 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, 1003 int op) 1004 { 1005 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) 1006 return min(q->limits.max_discard_sectors, UINT_MAX >> 9); 1007 1008 if (unlikely(op == REQ_OP_WRITE_SAME)) 1009 return q->limits.max_write_same_sectors; 1010 1011 if (unlikely(op == REQ_OP_WRITE_ZEROES)) 1012 return q->limits.max_write_zeroes_sectors; 1013 1014 return q->limits.max_sectors; 1015 } 1016 1017 /* 1018 * Return maximum size of a request at given offset. Only valid for 1019 * file system requests. 1020 */ 1021 static inline unsigned int blk_max_size_offset(struct request_queue *q, 1022 sector_t offset) 1023 { 1024 if (!q->limits.chunk_sectors) 1025 return q->limits.max_sectors; 1026 1027 return q->limits.chunk_sectors - 1028 (offset & (q->limits.chunk_sectors - 1)); 1029 } 1030 1031 static inline unsigned int blk_rq_get_max_sectors(struct request *rq, 1032 sector_t offset) 1033 { 1034 struct request_queue *q = rq->q; 1035 1036 if (unlikely(rq->cmd_type != REQ_TYPE_FS)) 1037 return q->limits.max_hw_sectors; 1038 1039 if (!q->limits.chunk_sectors || 1040 req_op(rq) == REQ_OP_DISCARD || 1041 req_op(rq) == REQ_OP_SECURE_ERASE) 1042 return blk_queue_get_max_sectors(q, req_op(rq)); 1043 1044 return min(blk_max_size_offset(q, offset), 1045 blk_queue_get_max_sectors(q, req_op(rq))); 1046 } 1047 1048 static inline unsigned int blk_rq_count_bios(struct request *rq) 1049 { 1050 unsigned int nr_bios = 0; 1051 struct bio *bio; 1052 1053 __rq_for_each_bio(bio, rq) 1054 nr_bios++; 1055 1056 return nr_bios; 1057 } 1058 1059 /* 1060 * Request issue related functions. 1061 */ 1062 extern struct request *blk_peek_request(struct request_queue *q); 1063 extern void blk_start_request(struct request *rq); 1064 extern struct request *blk_fetch_request(struct request_queue *q); 1065 1066 /* 1067 * Request completion related functions. 1068 * 1069 * blk_update_request() completes given number of bytes and updates 1070 * the request without completing it. 1071 * 1072 * blk_end_request() and friends. __blk_end_request() must be called 1073 * with the request queue spinlock acquired. 1074 * 1075 * Several drivers define their own end_request and call 1076 * blk_end_request() for parts of the original function. 1077 * This prevents code duplication in drivers. 1078 */ 1079 extern bool blk_update_request(struct request *rq, int error, 1080 unsigned int nr_bytes); 1081 extern void blk_finish_request(struct request *rq, int error); 1082 extern bool blk_end_request(struct request *rq, int error, 1083 unsigned int nr_bytes); 1084 extern void blk_end_request_all(struct request *rq, int error); 1085 extern bool blk_end_request_cur(struct request *rq, int error); 1086 extern bool blk_end_request_err(struct request *rq, int error); 1087 extern bool __blk_end_request(struct request *rq, int error, 1088 unsigned int nr_bytes); 1089 extern void __blk_end_request_all(struct request *rq, int error); 1090 extern bool __blk_end_request_cur(struct request *rq, int error); 1091 extern bool __blk_end_request_err(struct request *rq, int error); 1092 1093 extern void blk_complete_request(struct request *); 1094 extern void __blk_complete_request(struct request *); 1095 extern void blk_abort_request(struct request *); 1096 extern void blk_unprep_request(struct request *); 1097 1098 /* 1099 * Access functions for manipulating queue properties 1100 */ 1101 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn, 1102 spinlock_t *lock, int node_id); 1103 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *); 1104 extern struct request_queue *blk_init_allocated_queue(struct request_queue *, 1105 request_fn_proc *, spinlock_t *); 1106 extern void blk_cleanup_queue(struct request_queue *); 1107 extern void blk_queue_make_request(struct request_queue *, make_request_fn *); 1108 extern void blk_queue_bounce_limit(struct request_queue *, u64); 1109 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); 1110 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); 1111 extern void blk_queue_max_segments(struct request_queue *, unsigned short); 1112 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); 1113 extern void blk_queue_max_discard_sectors(struct request_queue *q, 1114 unsigned int max_discard_sectors); 1115 extern void blk_queue_max_write_same_sectors(struct request_queue *q, 1116 unsigned int max_write_same_sectors); 1117 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, 1118 unsigned int max_write_same_sectors); 1119 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); 1120 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); 1121 extern void blk_queue_alignment_offset(struct request_queue *q, 1122 unsigned int alignment); 1123 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 1124 extern void blk_queue_io_min(struct request_queue *q, unsigned int min); 1125 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 1126 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); 1127 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1128 extern void blk_set_default_limits(struct queue_limits *lim); 1129 extern void blk_set_stacking_limits(struct queue_limits *lim); 1130 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1131 sector_t offset); 1132 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, 1133 sector_t offset); 1134 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, 1135 sector_t offset); 1136 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); 1137 extern void blk_queue_dma_pad(struct request_queue *, unsigned int); 1138 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 1139 extern int blk_queue_dma_drain(struct request_queue *q, 1140 dma_drain_needed_fn *dma_drain_needed, 1141 void *buf, unsigned int size); 1142 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn); 1143 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); 1144 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long); 1145 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn); 1146 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn); 1147 extern void blk_queue_dma_alignment(struct request_queue *, int); 1148 extern void blk_queue_update_dma_alignment(struct request_queue *, int); 1149 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *); 1150 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *); 1151 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1152 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable); 1153 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); 1154 extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev); 1155 1156 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); 1157 extern void blk_dump_rq_flags(struct request *, char *); 1158 extern long nr_blockdev_pages(void); 1159 1160 bool __must_check blk_get_queue(struct request_queue *); 1161 struct request_queue *blk_alloc_queue(gfp_t); 1162 struct request_queue *blk_alloc_queue_node(gfp_t, int); 1163 extern void blk_put_queue(struct request_queue *); 1164 extern void blk_set_queue_dying(struct request_queue *); 1165 1166 /* 1167 * block layer runtime pm functions 1168 */ 1169 #ifdef CONFIG_PM 1170 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev); 1171 extern int blk_pre_runtime_suspend(struct request_queue *q); 1172 extern void blk_post_runtime_suspend(struct request_queue *q, int err); 1173 extern void blk_pre_runtime_resume(struct request_queue *q); 1174 extern void blk_post_runtime_resume(struct request_queue *q, int err); 1175 extern void blk_set_runtime_active(struct request_queue *q); 1176 #else 1177 static inline void blk_pm_runtime_init(struct request_queue *q, 1178 struct device *dev) {} 1179 static inline int blk_pre_runtime_suspend(struct request_queue *q) 1180 { 1181 return -ENOSYS; 1182 } 1183 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {} 1184 static inline void blk_pre_runtime_resume(struct request_queue *q) {} 1185 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {} 1186 static inline void blk_set_runtime_active(struct request_queue *q) {} 1187 #endif 1188 1189 /* 1190 * blk_plug permits building a queue of related requests by holding the I/O 1191 * fragments for a short period. This allows merging of sequential requests 1192 * into single larger request. As the requests are moved from a per-task list to 1193 * the device's request_queue in a batch, this results in improved scalability 1194 * as the lock contention for request_queue lock is reduced. 1195 * 1196 * It is ok not to disable preemption when adding the request to the plug list 1197 * or when attempting a merge, because blk_schedule_flush_list() will only flush 1198 * the plug list when the task sleeps by itself. For details, please see 1199 * schedule() where blk_schedule_flush_plug() is called. 1200 */ 1201 struct blk_plug { 1202 struct list_head list; /* requests */ 1203 struct list_head mq_list; /* blk-mq requests */ 1204 struct list_head cb_list; /* md requires an unplug callback */ 1205 }; 1206 #define BLK_MAX_REQUEST_COUNT 16 1207 #define BLK_PLUG_FLUSH_SIZE (128 * 1024) 1208 1209 struct blk_plug_cb; 1210 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1211 struct blk_plug_cb { 1212 struct list_head list; 1213 blk_plug_cb_fn callback; 1214 void *data; 1215 }; 1216 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1217 void *data, int size); 1218 extern void blk_start_plug(struct blk_plug *); 1219 extern void blk_finish_plug(struct blk_plug *); 1220 extern void blk_flush_plug_list(struct blk_plug *, bool); 1221 1222 static inline void blk_flush_plug(struct task_struct *tsk) 1223 { 1224 struct blk_plug *plug = tsk->plug; 1225 1226 if (plug) 1227 blk_flush_plug_list(plug, false); 1228 } 1229 1230 static inline void blk_schedule_flush_plug(struct task_struct *tsk) 1231 { 1232 struct blk_plug *plug = tsk->plug; 1233 1234 if (plug) 1235 blk_flush_plug_list(plug, true); 1236 } 1237 1238 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1239 { 1240 struct blk_plug *plug = tsk->plug; 1241 1242 return plug && 1243 (!list_empty(&plug->list) || 1244 !list_empty(&plug->mq_list) || 1245 !list_empty(&plug->cb_list)); 1246 } 1247 1248 /* 1249 * tag stuff 1250 */ 1251 extern int blk_queue_start_tag(struct request_queue *, struct request *); 1252 extern struct request *blk_queue_find_tag(struct request_queue *, int); 1253 extern void blk_queue_end_tag(struct request_queue *, struct request *); 1254 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int); 1255 extern void blk_queue_free_tags(struct request_queue *); 1256 extern int blk_queue_resize_tags(struct request_queue *, int); 1257 extern void blk_queue_invalidate_tags(struct request_queue *); 1258 extern struct blk_queue_tag *blk_init_tags(int, int); 1259 extern void blk_free_tags(struct blk_queue_tag *); 1260 1261 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt, 1262 int tag) 1263 { 1264 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth)) 1265 return NULL; 1266 return bqt->tag_index[tag]; 1267 } 1268 1269 1270 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */ 1271 #define BLKDEV_DISCARD_ZERO (1 << 1) /* must reliably zero data */ 1272 1273 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); 1274 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1275 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); 1276 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1277 sector_t nr_sects, gfp_t gfp_mask, int flags, 1278 struct bio **biop); 1279 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, 1280 sector_t nr_sects, gfp_t gfp_mask, struct page *page); 1281 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1282 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1283 bool discard); 1284 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1285 sector_t nr_sects, gfp_t gfp_mask, bool discard); 1286 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1287 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1288 { 1289 return blkdev_issue_discard(sb->s_bdev, block << (sb->s_blocksize_bits - 9), 1290 nr_blocks << (sb->s_blocksize_bits - 9), 1291 gfp_mask, flags); 1292 } 1293 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1294 sector_t nr_blocks, gfp_t gfp_mask) 1295 { 1296 return blkdev_issue_zeroout(sb->s_bdev, 1297 block << (sb->s_blocksize_bits - 9), 1298 nr_blocks << (sb->s_blocksize_bits - 9), 1299 gfp_mask, true); 1300 } 1301 1302 extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm); 1303 1304 enum blk_default_limits { 1305 BLK_MAX_SEGMENTS = 128, 1306 BLK_SAFE_MAX_SECTORS = 255, 1307 BLK_DEF_MAX_SECTORS = 2560, 1308 BLK_MAX_SEGMENT_SIZE = 65536, 1309 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1310 }; 1311 1312 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist) 1313 1314 static inline unsigned long queue_bounce_pfn(struct request_queue *q) 1315 { 1316 return q->limits.bounce_pfn; 1317 } 1318 1319 static inline unsigned long queue_segment_boundary(struct request_queue *q) 1320 { 1321 return q->limits.seg_boundary_mask; 1322 } 1323 1324 static inline unsigned long queue_virt_boundary(struct request_queue *q) 1325 { 1326 return q->limits.virt_boundary_mask; 1327 } 1328 1329 static inline unsigned int queue_max_sectors(struct request_queue *q) 1330 { 1331 return q->limits.max_sectors; 1332 } 1333 1334 static inline unsigned int queue_max_hw_sectors(struct request_queue *q) 1335 { 1336 return q->limits.max_hw_sectors; 1337 } 1338 1339 static inline unsigned short queue_max_segments(struct request_queue *q) 1340 { 1341 return q->limits.max_segments; 1342 } 1343 1344 static inline unsigned int queue_max_segment_size(struct request_queue *q) 1345 { 1346 return q->limits.max_segment_size; 1347 } 1348 1349 static inline unsigned short queue_logical_block_size(struct request_queue *q) 1350 { 1351 int retval = 512; 1352 1353 if (q && q->limits.logical_block_size) 1354 retval = q->limits.logical_block_size; 1355 1356 return retval; 1357 } 1358 1359 static inline unsigned short bdev_logical_block_size(struct block_device *bdev) 1360 { 1361 return queue_logical_block_size(bdev_get_queue(bdev)); 1362 } 1363 1364 static inline unsigned int queue_physical_block_size(struct request_queue *q) 1365 { 1366 return q->limits.physical_block_size; 1367 } 1368 1369 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1370 { 1371 return queue_physical_block_size(bdev_get_queue(bdev)); 1372 } 1373 1374 static inline unsigned int queue_io_min(struct request_queue *q) 1375 { 1376 return q->limits.io_min; 1377 } 1378 1379 static inline int bdev_io_min(struct block_device *bdev) 1380 { 1381 return queue_io_min(bdev_get_queue(bdev)); 1382 } 1383 1384 static inline unsigned int queue_io_opt(struct request_queue *q) 1385 { 1386 return q->limits.io_opt; 1387 } 1388 1389 static inline int bdev_io_opt(struct block_device *bdev) 1390 { 1391 return queue_io_opt(bdev_get_queue(bdev)); 1392 } 1393 1394 static inline int queue_alignment_offset(struct request_queue *q) 1395 { 1396 if (q->limits.misaligned) 1397 return -1; 1398 1399 return q->limits.alignment_offset; 1400 } 1401 1402 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) 1403 { 1404 unsigned int granularity = max(lim->physical_block_size, lim->io_min); 1405 unsigned int alignment = sector_div(sector, granularity >> 9) << 9; 1406 1407 return (granularity + lim->alignment_offset - alignment) % granularity; 1408 } 1409 1410 static inline int bdev_alignment_offset(struct block_device *bdev) 1411 { 1412 struct request_queue *q = bdev_get_queue(bdev); 1413 1414 if (q->limits.misaligned) 1415 return -1; 1416 1417 if (bdev != bdev->bd_contains) 1418 return bdev->bd_part->alignment_offset; 1419 1420 return q->limits.alignment_offset; 1421 } 1422 1423 static inline int queue_discard_alignment(struct request_queue *q) 1424 { 1425 if (q->limits.discard_misaligned) 1426 return -1; 1427 1428 return q->limits.discard_alignment; 1429 } 1430 1431 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) 1432 { 1433 unsigned int alignment, granularity, offset; 1434 1435 if (!lim->max_discard_sectors) 1436 return 0; 1437 1438 /* Why are these in bytes, not sectors? */ 1439 alignment = lim->discard_alignment >> 9; 1440 granularity = lim->discard_granularity >> 9; 1441 if (!granularity) 1442 return 0; 1443 1444 /* Offset of the partition start in 'granularity' sectors */ 1445 offset = sector_div(sector, granularity); 1446 1447 /* And why do we do this modulus *again* in blkdev_issue_discard()? */ 1448 offset = (granularity + alignment - offset) % granularity; 1449 1450 /* Turn it back into bytes, gaah */ 1451 return offset << 9; 1452 } 1453 1454 static inline int bdev_discard_alignment(struct block_device *bdev) 1455 { 1456 struct request_queue *q = bdev_get_queue(bdev); 1457 1458 if (bdev != bdev->bd_contains) 1459 return bdev->bd_part->discard_alignment; 1460 1461 return q->limits.discard_alignment; 1462 } 1463 1464 static inline unsigned int queue_discard_zeroes_data(struct request_queue *q) 1465 { 1466 if (q->limits.max_discard_sectors && q->limits.discard_zeroes_data == 1) 1467 return 1; 1468 1469 return 0; 1470 } 1471 1472 static inline unsigned int bdev_discard_zeroes_data(struct block_device *bdev) 1473 { 1474 return queue_discard_zeroes_data(bdev_get_queue(bdev)); 1475 } 1476 1477 static inline unsigned int bdev_write_same(struct block_device *bdev) 1478 { 1479 struct request_queue *q = bdev_get_queue(bdev); 1480 1481 if (q) 1482 return q->limits.max_write_same_sectors; 1483 1484 return 0; 1485 } 1486 1487 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1488 { 1489 struct request_queue *q = bdev_get_queue(bdev); 1490 1491 if (q) 1492 return q->limits.max_write_zeroes_sectors; 1493 1494 return 0; 1495 } 1496 1497 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev) 1498 { 1499 struct request_queue *q = bdev_get_queue(bdev); 1500 1501 if (q) 1502 return blk_queue_zoned_model(q); 1503 1504 return BLK_ZONED_NONE; 1505 } 1506 1507 static inline bool bdev_is_zoned(struct block_device *bdev) 1508 { 1509 struct request_queue *q = bdev_get_queue(bdev); 1510 1511 if (q) 1512 return blk_queue_is_zoned(q); 1513 1514 return false; 1515 } 1516 1517 static inline unsigned int bdev_zone_size(struct block_device *bdev) 1518 { 1519 struct request_queue *q = bdev_get_queue(bdev); 1520 1521 if (q) 1522 return blk_queue_zone_size(q); 1523 1524 return 0; 1525 } 1526 1527 static inline int queue_dma_alignment(struct request_queue *q) 1528 { 1529 return q ? q->dma_alignment : 511; 1530 } 1531 1532 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1533 unsigned int len) 1534 { 1535 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1536 return !(addr & alignment) && !(len & alignment); 1537 } 1538 1539 /* assumes size > 256 */ 1540 static inline unsigned int blksize_bits(unsigned int size) 1541 { 1542 unsigned int bits = 8; 1543 do { 1544 bits++; 1545 size >>= 1; 1546 } while (size > 256); 1547 return bits; 1548 } 1549 1550 static inline unsigned int block_size(struct block_device *bdev) 1551 { 1552 return bdev->bd_block_size; 1553 } 1554 1555 static inline bool queue_flush_queueable(struct request_queue *q) 1556 { 1557 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); 1558 } 1559 1560 typedef struct {struct page *v;} Sector; 1561 1562 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); 1563 1564 static inline void put_dev_sector(Sector p) 1565 { 1566 put_page(p.v); 1567 } 1568 1569 static inline bool __bvec_gap_to_prev(struct request_queue *q, 1570 struct bio_vec *bprv, unsigned int offset) 1571 { 1572 return offset || 1573 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q)); 1574 } 1575 1576 /* 1577 * Check if adding a bio_vec after bprv with offset would create a gap in 1578 * the SG list. Most drivers don't care about this, but some do. 1579 */ 1580 static inline bool bvec_gap_to_prev(struct request_queue *q, 1581 struct bio_vec *bprv, unsigned int offset) 1582 { 1583 if (!queue_virt_boundary(q)) 1584 return false; 1585 return __bvec_gap_to_prev(q, bprv, offset); 1586 } 1587 1588 static inline bool bio_will_gap(struct request_queue *q, struct bio *prev, 1589 struct bio *next) 1590 { 1591 if (bio_has_data(prev) && queue_virt_boundary(q)) { 1592 struct bio_vec pb, nb; 1593 1594 bio_get_last_bvec(prev, &pb); 1595 bio_get_first_bvec(next, &nb); 1596 1597 return __bvec_gap_to_prev(q, &pb, nb.bv_offset); 1598 } 1599 1600 return false; 1601 } 1602 1603 static inline bool req_gap_back_merge(struct request *req, struct bio *bio) 1604 { 1605 return bio_will_gap(req->q, req->biotail, bio); 1606 } 1607 1608 static inline bool req_gap_front_merge(struct request *req, struct bio *bio) 1609 { 1610 return bio_will_gap(req->q, bio, req->bio); 1611 } 1612 1613 int kblockd_schedule_work(struct work_struct *work); 1614 int kblockd_schedule_work_on(int cpu, struct work_struct *work); 1615 int kblockd_schedule_delayed_work(struct delayed_work *dwork, unsigned long delay); 1616 int kblockd_schedule_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1617 1618 #ifdef CONFIG_BLK_CGROUP 1619 /* 1620 * This should not be using sched_clock(). A real patch is in progress 1621 * to fix this up, until that is in place we need to disable preemption 1622 * around sched_clock() in this function and set_io_start_time_ns(). 1623 */ 1624 static inline void set_start_time_ns(struct request *req) 1625 { 1626 preempt_disable(); 1627 req->start_time_ns = sched_clock(); 1628 preempt_enable(); 1629 } 1630 1631 static inline void set_io_start_time_ns(struct request *req) 1632 { 1633 preempt_disable(); 1634 req->io_start_time_ns = sched_clock(); 1635 preempt_enable(); 1636 } 1637 1638 static inline uint64_t rq_start_time_ns(struct request *req) 1639 { 1640 return req->start_time_ns; 1641 } 1642 1643 static inline uint64_t rq_io_start_time_ns(struct request *req) 1644 { 1645 return req->io_start_time_ns; 1646 } 1647 #else 1648 static inline void set_start_time_ns(struct request *req) {} 1649 static inline void set_io_start_time_ns(struct request *req) {} 1650 static inline uint64_t rq_start_time_ns(struct request *req) 1651 { 1652 return 0; 1653 } 1654 static inline uint64_t rq_io_start_time_ns(struct request *req) 1655 { 1656 return 0; 1657 } 1658 #endif 1659 1660 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1661 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1662 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1663 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1664 1665 #if defined(CONFIG_BLK_DEV_INTEGRITY) 1666 1667 enum blk_integrity_flags { 1668 BLK_INTEGRITY_VERIFY = 1 << 0, 1669 BLK_INTEGRITY_GENERATE = 1 << 1, 1670 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, 1671 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, 1672 }; 1673 1674 struct blk_integrity_iter { 1675 void *prot_buf; 1676 void *data_buf; 1677 sector_t seed; 1678 unsigned int data_size; 1679 unsigned short interval; 1680 const char *disk_name; 1681 }; 1682 1683 typedef int (integrity_processing_fn) (struct blk_integrity_iter *); 1684 1685 struct blk_integrity_profile { 1686 integrity_processing_fn *generate_fn; 1687 integrity_processing_fn *verify_fn; 1688 const char *name; 1689 }; 1690 1691 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *); 1692 extern void blk_integrity_unregister(struct gendisk *); 1693 extern int blk_integrity_compare(struct gendisk *, struct gendisk *); 1694 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, 1695 struct scatterlist *); 1696 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); 1697 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, 1698 struct request *); 1699 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, 1700 struct bio *); 1701 1702 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1703 { 1704 struct blk_integrity *bi = &disk->queue->integrity; 1705 1706 if (!bi->profile) 1707 return NULL; 1708 1709 return bi; 1710 } 1711 1712 static inline 1713 struct blk_integrity *bdev_get_integrity(struct block_device *bdev) 1714 { 1715 return blk_get_integrity(bdev->bd_disk); 1716 } 1717 1718 static inline bool blk_integrity_rq(struct request *rq) 1719 { 1720 return rq->cmd_flags & REQ_INTEGRITY; 1721 } 1722 1723 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1724 unsigned int segs) 1725 { 1726 q->limits.max_integrity_segments = segs; 1727 } 1728 1729 static inline unsigned short 1730 queue_max_integrity_segments(struct request_queue *q) 1731 { 1732 return q->limits.max_integrity_segments; 1733 } 1734 1735 static inline bool integrity_req_gap_back_merge(struct request *req, 1736 struct bio *next) 1737 { 1738 struct bio_integrity_payload *bip = bio_integrity(req->bio); 1739 struct bio_integrity_payload *bip_next = bio_integrity(next); 1740 1741 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1], 1742 bip_next->bip_vec[0].bv_offset); 1743 } 1744 1745 static inline bool integrity_req_gap_front_merge(struct request *req, 1746 struct bio *bio) 1747 { 1748 struct bio_integrity_payload *bip = bio_integrity(bio); 1749 struct bio_integrity_payload *bip_next = bio_integrity(req->bio); 1750 1751 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1], 1752 bip_next->bip_vec[0].bv_offset); 1753 } 1754 1755 #else /* CONFIG_BLK_DEV_INTEGRITY */ 1756 1757 struct bio; 1758 struct block_device; 1759 struct gendisk; 1760 struct blk_integrity; 1761 1762 static inline int blk_integrity_rq(struct request *rq) 1763 { 1764 return 0; 1765 } 1766 static inline int blk_rq_count_integrity_sg(struct request_queue *q, 1767 struct bio *b) 1768 { 1769 return 0; 1770 } 1771 static inline int blk_rq_map_integrity_sg(struct request_queue *q, 1772 struct bio *b, 1773 struct scatterlist *s) 1774 { 1775 return 0; 1776 } 1777 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) 1778 { 1779 return NULL; 1780 } 1781 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1782 { 1783 return NULL; 1784 } 1785 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) 1786 { 1787 return 0; 1788 } 1789 static inline void blk_integrity_register(struct gendisk *d, 1790 struct blk_integrity *b) 1791 { 1792 } 1793 static inline void blk_integrity_unregister(struct gendisk *d) 1794 { 1795 } 1796 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1797 unsigned int segs) 1798 { 1799 } 1800 static inline unsigned short queue_max_integrity_segments(struct request_queue *q) 1801 { 1802 return 0; 1803 } 1804 static inline bool blk_integrity_merge_rq(struct request_queue *rq, 1805 struct request *r1, 1806 struct request *r2) 1807 { 1808 return true; 1809 } 1810 static inline bool blk_integrity_merge_bio(struct request_queue *rq, 1811 struct request *r, 1812 struct bio *b) 1813 { 1814 return true; 1815 } 1816 1817 static inline bool integrity_req_gap_back_merge(struct request *req, 1818 struct bio *next) 1819 { 1820 return false; 1821 } 1822 static inline bool integrity_req_gap_front_merge(struct request *req, 1823 struct bio *bio) 1824 { 1825 return false; 1826 } 1827 1828 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 1829 1830 /** 1831 * struct blk_dax_ctl - control and output parameters for ->direct_access 1832 * @sector: (input) offset relative to a block_device 1833 * @addr: (output) kernel virtual address for @sector populated by driver 1834 * @pfn: (output) page frame number for @addr populated by driver 1835 * @size: (input) number of bytes requested 1836 */ 1837 struct blk_dax_ctl { 1838 sector_t sector; 1839 void *addr; 1840 long size; 1841 pfn_t pfn; 1842 }; 1843 1844 struct block_device_operations { 1845 int (*open) (struct block_device *, fmode_t); 1846 void (*release) (struct gendisk *, fmode_t); 1847 int (*rw_page)(struct block_device *, sector_t, struct page *, bool); 1848 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1849 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1850 long (*direct_access)(struct block_device *, sector_t, void **, pfn_t *, 1851 long); 1852 unsigned int (*check_events) (struct gendisk *disk, 1853 unsigned int clearing); 1854 /* ->media_changed() is DEPRECATED, use ->check_events() instead */ 1855 int (*media_changed) (struct gendisk *); 1856 void (*unlock_native_capacity) (struct gendisk *); 1857 int (*revalidate_disk) (struct gendisk *); 1858 int (*getgeo)(struct block_device *, struct hd_geometry *); 1859 /* this callback is with swap_lock and sometimes page table lock held */ 1860 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1861 struct module *owner; 1862 const struct pr_ops *pr_ops; 1863 }; 1864 1865 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, 1866 unsigned long); 1867 extern int bdev_read_page(struct block_device *, sector_t, struct page *); 1868 extern int bdev_write_page(struct block_device *, sector_t, struct page *, 1869 struct writeback_control *); 1870 extern long bdev_direct_access(struct block_device *, struct blk_dax_ctl *); 1871 extern int bdev_dax_supported(struct super_block *, int); 1872 extern bool bdev_dax_capable(struct block_device *); 1873 #else /* CONFIG_BLOCK */ 1874 1875 struct block_device; 1876 1877 /* 1878 * stubs for when the block layer is configured out 1879 */ 1880 #define buffer_heads_over_limit 0 1881 1882 static inline long nr_blockdev_pages(void) 1883 { 1884 return 0; 1885 } 1886 1887 struct blk_plug { 1888 }; 1889 1890 static inline void blk_start_plug(struct blk_plug *plug) 1891 { 1892 } 1893 1894 static inline void blk_finish_plug(struct blk_plug *plug) 1895 { 1896 } 1897 1898 static inline void blk_flush_plug(struct task_struct *task) 1899 { 1900 } 1901 1902 static inline void blk_schedule_flush_plug(struct task_struct *task) 1903 { 1904 } 1905 1906 1907 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1908 { 1909 return false; 1910 } 1911 1912 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, 1913 sector_t *error_sector) 1914 { 1915 return 0; 1916 } 1917 1918 #endif /* CONFIG_BLOCK */ 1919 1920 #endif 1921