1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_BLKDEV_H 3 #define _LINUX_BLKDEV_H 4 5 #include <linux/sched.h> 6 #include <linux/sched/clock.h> 7 8 #ifdef CONFIG_BLOCK 9 10 #include <linux/major.h> 11 #include <linux/genhd.h> 12 #include <linux/list.h> 13 #include <linux/llist.h> 14 #include <linux/timer.h> 15 #include <linux/workqueue.h> 16 #include <linux/pagemap.h> 17 #include <linux/backing-dev-defs.h> 18 #include <linux/wait.h> 19 #include <linux/mempool.h> 20 #include <linux/pfn.h> 21 #include <linux/bio.h> 22 #include <linux/stringify.h> 23 #include <linux/gfp.h> 24 #include <linux/bsg.h> 25 #include <linux/smp.h> 26 #include <linux/rcupdate.h> 27 #include <linux/percpu-refcount.h> 28 #include <linux/scatterlist.h> 29 #include <linux/blkzoned.h> 30 #include <linux/seqlock.h> 31 #include <linux/u64_stats_sync.h> 32 33 struct module; 34 struct scsi_ioctl_command; 35 36 struct request_queue; 37 struct elevator_queue; 38 struct blk_trace; 39 struct request; 40 struct sg_io_hdr; 41 struct bsg_job; 42 struct blkcg_gq; 43 struct blk_flush_queue; 44 struct pr_ops; 45 struct rq_wb; 46 struct blk_queue_stats; 47 struct blk_stat_callback; 48 49 #define BLKDEV_MIN_RQ 4 50 #define BLKDEV_MAX_RQ 128 /* Default maximum */ 51 52 /* Must be consistent with blk_mq_poll_stats_bkt() */ 53 #define BLK_MQ_POLL_STATS_BKTS 16 54 55 /* 56 * Maximum number of blkcg policies allowed to be registered concurrently. 57 * Defined here to simplify include dependency. 58 */ 59 #define BLKCG_MAX_POLS 3 60 61 typedef void (rq_end_io_fn)(struct request *, blk_status_t); 62 63 #define BLK_RL_SYNCFULL (1U << 0) 64 #define BLK_RL_ASYNCFULL (1U << 1) 65 66 struct request_list { 67 struct request_queue *q; /* the queue this rl belongs to */ 68 #ifdef CONFIG_BLK_CGROUP 69 struct blkcg_gq *blkg; /* blkg this request pool belongs to */ 70 #endif 71 /* 72 * count[], starved[], and wait[] are indexed by 73 * BLK_RW_SYNC/BLK_RW_ASYNC 74 */ 75 int count[2]; 76 int starved[2]; 77 mempool_t *rq_pool; 78 wait_queue_head_t wait[2]; 79 unsigned int flags; 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 /* Look at ->special_vec for the actual data payload instead of the 124 bio chain. */ 125 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18)) 126 /* The per-zone write lock is held for this request */ 127 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) 128 /* timeout is expired */ 129 #define RQF_MQ_TIMEOUT_EXPIRED ((__force req_flags_t)(1 << 20)) 130 /* already slept for hybrid poll */ 131 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 21)) 132 133 /* flags that prevent us from merging requests: */ 134 #define RQF_NOMERGE_FLAGS \ 135 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD) 136 137 /* 138 * Try to put the fields that are referenced together in the same cacheline. 139 * 140 * If you modify this structure, make sure to update blk_rq_init() and 141 * especially blk_mq_rq_ctx_init() to take care of the added fields. 142 */ 143 struct request { 144 struct request_queue *q; 145 struct blk_mq_ctx *mq_ctx; 146 147 int cpu; 148 unsigned int cmd_flags; /* op and common flags */ 149 req_flags_t rq_flags; 150 151 int internal_tag; 152 153 /* the following two fields are internal, NEVER access directly */ 154 unsigned int __data_len; /* total data len */ 155 int tag; 156 sector_t __sector; /* sector cursor */ 157 158 struct bio *bio; 159 struct bio *biotail; 160 161 struct list_head queuelist; 162 163 /* 164 * The hash is used inside the scheduler, and killed once the 165 * request reaches the dispatch list. The ipi_list is only used 166 * to queue the request for softirq completion, which is long 167 * after the request has been unhashed (and even removed from 168 * the dispatch list). 169 */ 170 union { 171 struct hlist_node hash; /* merge hash */ 172 struct list_head ipi_list; 173 }; 174 175 /* 176 * The rb_node is only used inside the io scheduler, requests 177 * are pruned when moved to the dispatch queue. So let the 178 * completion_data share space with the rb_node. 179 */ 180 union { 181 struct rb_node rb_node; /* sort/lookup */ 182 struct bio_vec special_vec; 183 void *completion_data; 184 int error_count; /* for legacy drivers, don't use */ 185 }; 186 187 /* 188 * Three pointers are available for the IO schedulers, if they need 189 * more they have to dynamically allocate it. Flush requests are 190 * never put on the IO scheduler. So let the flush fields share 191 * space with the elevator data. 192 */ 193 union { 194 struct { 195 struct io_cq *icq; 196 void *priv[2]; 197 } elv; 198 199 struct { 200 unsigned int seq; 201 struct list_head list; 202 rq_end_io_fn *saved_end_io; 203 } flush; 204 }; 205 206 struct gendisk *rq_disk; 207 struct hd_struct *part; 208 unsigned long start_time; 209 struct blk_issue_stat issue_stat; 210 /* Number of scatter-gather DMA addr+len pairs after 211 * physical address coalescing is performed. 212 */ 213 unsigned short nr_phys_segments; 214 215 #if defined(CONFIG_BLK_DEV_INTEGRITY) 216 unsigned short nr_integrity_segments; 217 #endif 218 219 unsigned short write_hint; 220 unsigned short ioprio; 221 222 unsigned int timeout; 223 224 void *special; /* opaque pointer available for LLD use */ 225 226 unsigned int extra_len; /* length of alignment and padding */ 227 228 /* 229 * On blk-mq, the lower bits of ->gstate (generation number and 230 * state) carry the MQ_RQ_* state value and the upper bits the 231 * generation number which is monotonically incremented and used to 232 * distinguish the reuse instances. 233 * 234 * ->gstate_seq allows updates to ->gstate and other fields 235 * (currently ->deadline) during request start to be read 236 * atomically from the timeout path, so that it can operate on a 237 * coherent set of information. 238 */ 239 seqcount_t gstate_seq; 240 u64 gstate; 241 242 /* 243 * ->aborted_gstate is used by the timeout to claim a specific 244 * recycle instance of this request. See blk_mq_timeout_work(). 245 */ 246 struct u64_stats_sync aborted_gstate_sync; 247 u64 aborted_gstate; 248 249 /* access through blk_rq_set_deadline, blk_rq_deadline */ 250 unsigned long __deadline; 251 252 struct list_head timeout_list; 253 254 union { 255 struct __call_single_data csd; 256 u64 fifo_time; 257 }; 258 259 /* 260 * completion callback. 261 */ 262 rq_end_io_fn *end_io; 263 void *end_io_data; 264 265 /* for bidi */ 266 struct request *next_rq; 267 268 #ifdef CONFIG_BLK_CGROUP 269 struct request_list *rl; /* rl this rq is alloced from */ 270 unsigned long long start_time_ns; 271 unsigned long long io_start_time_ns; /* when passed to hardware */ 272 #endif 273 }; 274 275 static inline bool blk_op_is_scsi(unsigned int op) 276 { 277 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT; 278 } 279 280 static inline bool blk_op_is_private(unsigned int op) 281 { 282 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 283 } 284 285 static inline bool blk_rq_is_scsi(struct request *rq) 286 { 287 return blk_op_is_scsi(req_op(rq)); 288 } 289 290 static inline bool blk_rq_is_private(struct request *rq) 291 { 292 return blk_op_is_private(req_op(rq)); 293 } 294 295 static inline bool blk_rq_is_passthrough(struct request *rq) 296 { 297 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq); 298 } 299 300 static inline bool bio_is_passthrough(struct bio *bio) 301 { 302 unsigned op = bio_op(bio); 303 304 return blk_op_is_scsi(op) || blk_op_is_private(op); 305 } 306 307 static inline unsigned short req_get_ioprio(struct request *req) 308 { 309 return req->ioprio; 310 } 311 312 #include <linux/elevator.h> 313 314 struct blk_queue_ctx; 315 316 typedef void (request_fn_proc) (struct request_queue *q); 317 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio); 318 typedef bool (poll_q_fn) (struct request_queue *q, blk_qc_t); 319 typedef int (prep_rq_fn) (struct request_queue *, struct request *); 320 typedef void (unprep_rq_fn) (struct request_queue *, struct request *); 321 322 struct bio_vec; 323 typedef void (softirq_done_fn)(struct request *); 324 typedef int (dma_drain_needed_fn)(struct request *); 325 typedef int (lld_busy_fn) (struct request_queue *q); 326 typedef int (bsg_job_fn) (struct bsg_job *); 327 typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t); 328 typedef void (exit_rq_fn)(struct request_queue *, struct request *); 329 330 enum blk_eh_timer_return { 331 BLK_EH_NOT_HANDLED, 332 BLK_EH_HANDLED, 333 BLK_EH_RESET_TIMER, 334 }; 335 336 typedef enum blk_eh_timer_return (rq_timed_out_fn)(struct request *); 337 338 enum blk_queue_state { 339 Queue_down, 340 Queue_up, 341 }; 342 343 struct blk_queue_tag { 344 struct request **tag_index; /* map of busy tags */ 345 unsigned long *tag_map; /* bit map of free/busy tags */ 346 int max_depth; /* what we will send to device */ 347 int real_max_depth; /* what the array can hold */ 348 atomic_t refcnt; /* map can be shared */ 349 int alloc_policy; /* tag allocation policy */ 350 int next_tag; /* next tag */ 351 }; 352 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ 353 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ 354 355 #define BLK_SCSI_MAX_CMDS (256) 356 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) 357 358 /* 359 * Zoned block device models (zoned limit). 360 */ 361 enum blk_zoned_model { 362 BLK_ZONED_NONE, /* Regular block device */ 363 BLK_ZONED_HA, /* Host-aware zoned block device */ 364 BLK_ZONED_HM, /* Host-managed zoned block device */ 365 }; 366 367 struct queue_limits { 368 unsigned long bounce_pfn; 369 unsigned long seg_boundary_mask; 370 unsigned long virt_boundary_mask; 371 372 unsigned int max_hw_sectors; 373 unsigned int max_dev_sectors; 374 unsigned int chunk_sectors; 375 unsigned int max_sectors; 376 unsigned int max_segment_size; 377 unsigned int physical_block_size; 378 unsigned int alignment_offset; 379 unsigned int io_min; 380 unsigned int io_opt; 381 unsigned int max_discard_sectors; 382 unsigned int max_hw_discard_sectors; 383 unsigned int max_write_same_sectors; 384 unsigned int max_write_zeroes_sectors; 385 unsigned int discard_granularity; 386 unsigned int discard_alignment; 387 388 unsigned short logical_block_size; 389 unsigned short max_segments; 390 unsigned short max_integrity_segments; 391 unsigned short max_discard_segments; 392 393 unsigned char misaligned; 394 unsigned char discard_misaligned; 395 unsigned char cluster; 396 unsigned char raid_partial_stripes_expensive; 397 enum blk_zoned_model zoned; 398 }; 399 400 #ifdef CONFIG_BLK_DEV_ZONED 401 402 struct blk_zone_report_hdr { 403 unsigned int nr_zones; 404 u8 padding[60]; 405 }; 406 407 extern int blkdev_report_zones(struct block_device *bdev, 408 sector_t sector, struct blk_zone *zones, 409 unsigned int *nr_zones, gfp_t gfp_mask); 410 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors, 411 sector_t nr_sectors, gfp_t gfp_mask); 412 413 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, 414 unsigned int cmd, unsigned long arg); 415 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode, 416 unsigned int cmd, unsigned long arg); 417 418 #else /* CONFIG_BLK_DEV_ZONED */ 419 420 static inline int blkdev_report_zones_ioctl(struct block_device *bdev, 421 fmode_t mode, unsigned int cmd, 422 unsigned long arg) 423 { 424 return -ENOTTY; 425 } 426 427 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev, 428 fmode_t mode, unsigned int cmd, 429 unsigned long arg) 430 { 431 return -ENOTTY; 432 } 433 434 #endif /* CONFIG_BLK_DEV_ZONED */ 435 436 struct request_queue { 437 /* 438 * Together with queue_head for cacheline sharing 439 */ 440 struct list_head queue_head; 441 struct request *last_merge; 442 struct elevator_queue *elevator; 443 int nr_rqs[2]; /* # allocated [a]sync rqs */ 444 int nr_rqs_elvpriv; /* # allocated rqs w/ elvpriv */ 445 446 atomic_t shared_hctx_restart; 447 448 struct blk_queue_stats *stats; 449 struct rq_wb *rq_wb; 450 451 /* 452 * If blkcg is not used, @q->root_rl serves all requests. If blkcg 453 * is used, root blkg allocates from @q->root_rl and all other 454 * blkgs from their own blkg->rl. Which one to use should be 455 * determined using bio_request_list(). 456 */ 457 struct request_list root_rl; 458 459 request_fn_proc *request_fn; 460 make_request_fn *make_request_fn; 461 poll_q_fn *poll_fn; 462 prep_rq_fn *prep_rq_fn; 463 unprep_rq_fn *unprep_rq_fn; 464 softirq_done_fn *softirq_done_fn; 465 rq_timed_out_fn *rq_timed_out_fn; 466 dma_drain_needed_fn *dma_drain_needed; 467 lld_busy_fn *lld_busy_fn; 468 /* Called just after a request is allocated */ 469 init_rq_fn *init_rq_fn; 470 /* Called just before a request is freed */ 471 exit_rq_fn *exit_rq_fn; 472 /* Called from inside blk_get_request() */ 473 void (*initialize_rq_fn)(struct request *rq); 474 475 const struct blk_mq_ops *mq_ops; 476 477 unsigned int *mq_map; 478 479 /* sw queues */ 480 struct blk_mq_ctx __percpu *queue_ctx; 481 unsigned int nr_queues; 482 483 unsigned int queue_depth; 484 485 /* hw dispatch queues */ 486 struct blk_mq_hw_ctx **queue_hw_ctx; 487 unsigned int nr_hw_queues; 488 489 /* 490 * Dispatch queue sorting 491 */ 492 sector_t end_sector; 493 struct request *boundary_rq; 494 495 /* 496 * Delayed queue handling 497 */ 498 struct delayed_work delay_work; 499 500 struct backing_dev_info *backing_dev_info; 501 502 /* 503 * The queue owner gets to use this for whatever they like. 504 * ll_rw_blk doesn't touch it. 505 */ 506 void *queuedata; 507 508 /* 509 * various queue flags, see QUEUE_* below 510 */ 511 unsigned long queue_flags; 512 513 /* 514 * ida allocated id for this queue. Used to index queues from 515 * ioctx. 516 */ 517 int id; 518 519 /* 520 * queue needs bounce pages for pages above this limit 521 */ 522 gfp_t bounce_gfp; 523 524 /* 525 * protects queue structures from reentrancy. ->__queue_lock should 526 * _never_ be used directly, it is queue private. always use 527 * ->queue_lock. 528 */ 529 spinlock_t __queue_lock; 530 spinlock_t *queue_lock; 531 532 /* 533 * queue kobject 534 */ 535 struct kobject kobj; 536 537 /* 538 * mq queue kobject 539 */ 540 struct kobject mq_kobj; 541 542 #ifdef CONFIG_BLK_DEV_INTEGRITY 543 struct blk_integrity integrity; 544 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 545 546 #ifdef CONFIG_PM 547 struct device *dev; 548 int rpm_status; 549 unsigned int nr_pending; 550 #endif 551 552 /* 553 * queue settings 554 */ 555 unsigned long nr_requests; /* Max # of requests */ 556 unsigned int nr_congestion_on; 557 unsigned int nr_congestion_off; 558 unsigned int nr_batching; 559 560 unsigned int dma_drain_size; 561 void *dma_drain_buffer; 562 unsigned int dma_pad_mask; 563 unsigned int dma_alignment; 564 565 struct blk_queue_tag *queue_tags; 566 struct list_head tag_busy_list; 567 568 unsigned int nr_sorted; 569 unsigned int in_flight[2]; 570 571 /* 572 * Number of active block driver functions for which blk_drain_queue() 573 * must wait. Must be incremented around functions that unlock the 574 * queue_lock internally, e.g. scsi_request_fn(). 575 */ 576 unsigned int request_fn_active; 577 578 unsigned int rq_timeout; 579 int poll_nsec; 580 581 struct blk_stat_callback *poll_cb; 582 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS]; 583 584 struct timer_list timeout; 585 struct work_struct timeout_work; 586 struct list_head timeout_list; 587 588 struct list_head icq_list; 589 #ifdef CONFIG_BLK_CGROUP 590 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 591 struct blkcg_gq *root_blkg; 592 struct list_head blkg_list; 593 #endif 594 595 struct queue_limits limits; 596 597 /* 598 * Zoned block device information for request dispatch control. 599 * nr_zones is the total number of zones of the device. This is always 600 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones 601 * bits which indicates if a zone is conventional (bit clear) or 602 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones 603 * bits which indicates if a zone is write locked, that is, if a write 604 * request targeting the zone was dispatched. All three fields are 605 * initialized by the low level device driver (e.g. scsi/sd.c). 606 * Stacking drivers (device mappers) may or may not initialize 607 * these fields. 608 * 609 * Reads of this information must be protected with blk_queue_enter() / 610 * blk_queue_exit(). Modifying this information is only allowed while 611 * no requests are being processed. See also blk_mq_freeze_queue() and 612 * blk_mq_unfreeze_queue(). 613 */ 614 unsigned int nr_zones; 615 unsigned long *seq_zones_bitmap; 616 unsigned long *seq_zones_wlock; 617 618 /* 619 * sg stuff 620 */ 621 unsigned int sg_timeout; 622 unsigned int sg_reserved_size; 623 int node; 624 #ifdef CONFIG_BLK_DEV_IO_TRACE 625 struct blk_trace *blk_trace; 626 struct mutex blk_trace_mutex; 627 #endif 628 /* 629 * for flush operations 630 */ 631 struct blk_flush_queue *fq; 632 633 struct list_head requeue_list; 634 spinlock_t requeue_lock; 635 struct delayed_work requeue_work; 636 637 struct mutex sysfs_lock; 638 639 int bypass_depth; 640 atomic_t mq_freeze_depth; 641 642 #if defined(CONFIG_BLK_DEV_BSG) 643 bsg_job_fn *bsg_job_fn; 644 struct bsg_class_device bsg_dev; 645 #endif 646 647 #ifdef CONFIG_BLK_DEV_THROTTLING 648 /* Throttle data */ 649 struct throtl_data *td; 650 #endif 651 struct rcu_head rcu_head; 652 wait_queue_head_t mq_freeze_wq; 653 struct percpu_ref q_usage_counter; 654 struct list_head all_q_node; 655 656 struct blk_mq_tag_set *tag_set; 657 struct list_head tag_set_list; 658 struct bio_set *bio_split; 659 660 #ifdef CONFIG_BLK_DEBUG_FS 661 struct dentry *debugfs_dir; 662 struct dentry *sched_debugfs_dir; 663 #endif 664 665 bool mq_sysfs_init_done; 666 667 size_t cmd_size; 668 void *rq_alloc_data; 669 670 struct work_struct release_work; 671 672 #define BLK_MAX_WRITE_HINTS 5 673 u64 write_hints[BLK_MAX_WRITE_HINTS]; 674 }; 675 676 #define QUEUE_FLAG_QUEUED 0 /* uses generic tag queueing */ 677 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */ 678 #define QUEUE_FLAG_DYING 2 /* queue being torn down */ 679 #define QUEUE_FLAG_BYPASS 3 /* act as dumb FIFO queue */ 680 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */ 681 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */ 682 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */ 683 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */ 684 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */ 685 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ 686 #define QUEUE_FLAG_IO_STAT 10 /* do IO stats */ 687 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */ 688 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */ 689 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */ 690 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */ 691 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */ 692 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */ 693 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */ 694 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/ 695 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */ 696 #define QUEUE_FLAG_WC 20 /* Write back caching */ 697 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */ 698 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */ 699 #define QUEUE_FLAG_DAX 23 /* device supports DAX */ 700 #define QUEUE_FLAG_STATS 24 /* track rq completion times */ 701 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */ 702 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */ 703 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */ 704 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */ 705 #define QUEUE_FLAG_PREEMPT_ONLY 29 /* only process REQ_PREEMPT requests */ 706 707 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 708 (1 << QUEUE_FLAG_SAME_COMP) | \ 709 (1 << QUEUE_FLAG_ADD_RANDOM)) 710 711 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 712 (1 << QUEUE_FLAG_SAME_COMP) | \ 713 (1 << QUEUE_FLAG_POLL)) 714 715 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 716 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 717 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); 718 bool blk_queue_flag_test_and_clear(unsigned int flag, struct request_queue *q); 719 720 #define blk_queue_tagged(q) test_bit(QUEUE_FLAG_QUEUED, &(q)->queue_flags) 721 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 722 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 723 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) 724 #define blk_queue_bypass(q) test_bit(QUEUE_FLAG_BYPASS, &(q)->queue_flags) 725 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 726 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 727 #define blk_queue_noxmerges(q) \ 728 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 729 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) 730 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) 731 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) 732 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) 733 #define blk_queue_secure_erase(q) \ 734 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags)) 735 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) 736 #define blk_queue_scsi_passthrough(q) \ 737 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags) 738 739 #define blk_noretry_request(rq) \ 740 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 741 REQ_FAILFAST_DRIVER)) 742 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 743 #define blk_queue_preempt_only(q) \ 744 test_bit(QUEUE_FLAG_PREEMPT_ONLY, &(q)->queue_flags) 745 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags) 746 747 extern int blk_set_preempt_only(struct request_queue *q); 748 extern void blk_clear_preempt_only(struct request_queue *q); 749 750 static inline int queue_in_flight(struct request_queue *q) 751 { 752 return q->in_flight[0] + q->in_flight[1]; 753 } 754 755 static inline bool blk_account_rq(struct request *rq) 756 { 757 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq); 758 } 759 760 #define blk_rq_cpu_valid(rq) ((rq)->cpu != -1) 761 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL) 762 /* rq->queuelist of dequeued request must be list_empty() */ 763 #define blk_queued_rq(rq) (!list_empty(&(rq)->queuelist)) 764 765 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 766 767 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ) 768 769 /* 770 * Driver can handle struct request, if it either has an old style 771 * request_fn defined, or is blk-mq based. 772 */ 773 static inline bool queue_is_rq_based(struct request_queue *q) 774 { 775 return q->request_fn || q->mq_ops; 776 } 777 778 static inline unsigned int blk_queue_cluster(struct request_queue *q) 779 { 780 return q->limits.cluster; 781 } 782 783 static inline enum blk_zoned_model 784 blk_queue_zoned_model(struct request_queue *q) 785 { 786 return q->limits.zoned; 787 } 788 789 static inline bool blk_queue_is_zoned(struct request_queue *q) 790 { 791 switch (blk_queue_zoned_model(q)) { 792 case BLK_ZONED_HA: 793 case BLK_ZONED_HM: 794 return true; 795 default: 796 return false; 797 } 798 } 799 800 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q) 801 { 802 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; 803 } 804 805 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 806 { 807 return q->nr_zones; 808 } 809 810 static inline unsigned int blk_queue_zone_no(struct request_queue *q, 811 sector_t sector) 812 { 813 if (!blk_queue_is_zoned(q)) 814 return 0; 815 return sector >> ilog2(q->limits.chunk_sectors); 816 } 817 818 static inline bool blk_queue_zone_is_seq(struct request_queue *q, 819 sector_t sector) 820 { 821 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap) 822 return false; 823 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap); 824 } 825 826 static inline bool rq_is_sync(struct request *rq) 827 { 828 return op_is_sync(rq->cmd_flags); 829 } 830 831 static inline bool blk_rl_full(struct request_list *rl, bool sync) 832 { 833 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 834 835 return rl->flags & flag; 836 } 837 838 static inline void blk_set_rl_full(struct request_list *rl, bool sync) 839 { 840 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 841 842 rl->flags |= flag; 843 } 844 845 static inline void blk_clear_rl_full(struct request_list *rl, bool sync) 846 { 847 unsigned int flag = sync ? BLK_RL_SYNCFULL : BLK_RL_ASYNCFULL; 848 849 rl->flags &= ~flag; 850 } 851 852 static inline bool rq_mergeable(struct request *rq) 853 { 854 if (blk_rq_is_passthrough(rq)) 855 return false; 856 857 if (req_op(rq) == REQ_OP_FLUSH) 858 return false; 859 860 if (req_op(rq) == REQ_OP_WRITE_ZEROES) 861 return false; 862 863 if (rq->cmd_flags & REQ_NOMERGE_FLAGS) 864 return false; 865 if (rq->rq_flags & RQF_NOMERGE_FLAGS) 866 return false; 867 868 return true; 869 } 870 871 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) 872 { 873 if (bio_page(a) == bio_page(b) && 874 bio_offset(a) == bio_offset(b)) 875 return true; 876 877 return false; 878 } 879 880 static inline unsigned int blk_queue_depth(struct request_queue *q) 881 { 882 if (q->queue_depth) 883 return q->queue_depth; 884 885 return q->nr_requests; 886 } 887 888 /* 889 * q->prep_rq_fn return values 890 */ 891 enum { 892 BLKPREP_OK, /* serve it */ 893 BLKPREP_KILL, /* fatal error, kill, return -EIO */ 894 BLKPREP_DEFER, /* leave on queue */ 895 BLKPREP_INVALID, /* invalid command, kill, return -EREMOTEIO */ 896 }; 897 898 extern unsigned long blk_max_low_pfn, blk_max_pfn; 899 900 /* 901 * standard bounce addresses: 902 * 903 * BLK_BOUNCE_HIGH : bounce all highmem pages 904 * BLK_BOUNCE_ANY : don't bounce anything 905 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary 906 */ 907 908 #if BITS_PER_LONG == 32 909 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) 910 #else 911 #define BLK_BOUNCE_HIGH -1ULL 912 #endif 913 #define BLK_BOUNCE_ANY (-1ULL) 914 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) 915 916 /* 917 * default timeout for SG_IO if none specified 918 */ 919 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 920 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 921 922 struct rq_map_data { 923 struct page **pages; 924 int page_order; 925 int nr_entries; 926 unsigned long offset; 927 int null_mapped; 928 int from_user; 929 }; 930 931 struct req_iterator { 932 struct bvec_iter iter; 933 struct bio *bio; 934 }; 935 936 /* This should not be used directly - use rq_for_each_segment */ 937 #define for_each_bio(_bio) \ 938 for (; _bio; _bio = _bio->bi_next) 939 #define __rq_for_each_bio(_bio, rq) \ 940 if ((rq->bio)) \ 941 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) 942 943 #define rq_for_each_segment(bvl, _rq, _iter) \ 944 __rq_for_each_bio(_iter.bio, _rq) \ 945 bio_for_each_segment(bvl, _iter.bio, _iter.iter) 946 947 #define rq_iter_last(bvec, _iter) \ 948 (_iter.bio->bi_next == NULL && \ 949 bio_iter_last(bvec, _iter.iter)) 950 951 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 952 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 953 #endif 954 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 955 extern void rq_flush_dcache_pages(struct request *rq); 956 #else 957 static inline void rq_flush_dcache_pages(struct request *rq) 958 { 959 } 960 #endif 961 962 extern int blk_register_queue(struct gendisk *disk); 963 extern void blk_unregister_queue(struct gendisk *disk); 964 extern blk_qc_t generic_make_request(struct bio *bio); 965 extern blk_qc_t direct_make_request(struct bio *bio); 966 extern void blk_rq_init(struct request_queue *q, struct request *rq); 967 extern void blk_init_request_from_bio(struct request *req, struct bio *bio); 968 extern void blk_put_request(struct request *); 969 extern void __blk_put_request(struct request_queue *, struct request *); 970 extern struct request *blk_get_request_flags(struct request_queue *, 971 unsigned int op, 972 blk_mq_req_flags_t flags); 973 extern struct request *blk_get_request(struct request_queue *, unsigned int op, 974 gfp_t gfp_mask); 975 extern void blk_requeue_request(struct request_queue *, struct request *); 976 extern int blk_lld_busy(struct request_queue *q); 977 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, 978 struct bio_set *bs, gfp_t gfp_mask, 979 int (*bio_ctr)(struct bio *, struct bio *, void *), 980 void *data); 981 extern void blk_rq_unprep_clone(struct request *rq); 982 extern blk_status_t blk_insert_cloned_request(struct request_queue *q, 983 struct request *rq); 984 extern int blk_rq_append_bio(struct request *rq, struct bio **bio); 985 extern void blk_delay_queue(struct request_queue *, unsigned long); 986 extern void blk_queue_split(struct request_queue *, struct bio **); 987 extern void blk_recount_segments(struct request_queue *, struct bio *); 988 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); 989 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, 990 unsigned int, void __user *); 991 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, 992 unsigned int, void __user *); 993 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, 994 struct scsi_ioctl_command __user *); 995 996 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 997 extern void blk_queue_exit(struct request_queue *q); 998 extern void blk_start_queue(struct request_queue *q); 999 extern void blk_start_queue_async(struct request_queue *q); 1000 extern void blk_stop_queue(struct request_queue *q); 1001 extern void blk_sync_queue(struct request_queue *q); 1002 extern void __blk_stop_queue(struct request_queue *q); 1003 extern void __blk_run_queue(struct request_queue *q); 1004 extern void __blk_run_queue_uncond(struct request_queue *q); 1005 extern void blk_run_queue(struct request_queue *); 1006 extern void blk_run_queue_async(struct request_queue *q); 1007 extern int blk_rq_map_user(struct request_queue *, struct request *, 1008 struct rq_map_data *, void __user *, unsigned long, 1009 gfp_t); 1010 extern int blk_rq_unmap_user(struct bio *); 1011 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); 1012 extern int blk_rq_map_user_iov(struct request_queue *, struct request *, 1013 struct rq_map_data *, const struct iov_iter *, 1014 gfp_t); 1015 extern void blk_execute_rq(struct request_queue *, struct gendisk *, 1016 struct request *, int); 1017 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, 1018 struct request *, int, rq_end_io_fn *); 1019 1020 int blk_status_to_errno(blk_status_t status); 1021 blk_status_t errno_to_blk_status(int errno); 1022 1023 bool blk_poll(struct request_queue *q, blk_qc_t cookie); 1024 1025 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 1026 { 1027 return bdev->bd_disk->queue; /* this is never NULL */ 1028 } 1029 1030 /* 1031 * The basic unit of block I/O is a sector. It is used in a number of contexts 1032 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9 1033 * bytes. Variables of type sector_t represent an offset or size that is a 1034 * multiple of 512 bytes. Hence these two constants. 1035 */ 1036 #ifndef SECTOR_SHIFT 1037 #define SECTOR_SHIFT 9 1038 #endif 1039 #ifndef SECTOR_SIZE 1040 #define SECTOR_SIZE (1 << SECTOR_SHIFT) 1041 #endif 1042 1043 /* 1044 * blk_rq_pos() : the current sector 1045 * blk_rq_bytes() : bytes left in the entire request 1046 * blk_rq_cur_bytes() : bytes left in the current segment 1047 * blk_rq_err_bytes() : bytes left till the next error boundary 1048 * blk_rq_sectors() : sectors left in the entire request 1049 * blk_rq_cur_sectors() : sectors left in the current segment 1050 */ 1051 static inline sector_t blk_rq_pos(const struct request *rq) 1052 { 1053 return rq->__sector; 1054 } 1055 1056 static inline unsigned int blk_rq_bytes(const struct request *rq) 1057 { 1058 return rq->__data_len; 1059 } 1060 1061 static inline int blk_rq_cur_bytes(const struct request *rq) 1062 { 1063 return rq->bio ? bio_cur_bytes(rq->bio) : 0; 1064 } 1065 1066 extern unsigned int blk_rq_err_bytes(const struct request *rq); 1067 1068 static inline unsigned int blk_rq_sectors(const struct request *rq) 1069 { 1070 return blk_rq_bytes(rq) >> SECTOR_SHIFT; 1071 } 1072 1073 static inline unsigned int blk_rq_cur_sectors(const struct request *rq) 1074 { 1075 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT; 1076 } 1077 1078 static inline unsigned int blk_rq_zone_no(struct request *rq) 1079 { 1080 return blk_queue_zone_no(rq->q, blk_rq_pos(rq)); 1081 } 1082 1083 static inline unsigned int blk_rq_zone_is_seq(struct request *rq) 1084 { 1085 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq)); 1086 } 1087 1088 /* 1089 * Some commands like WRITE SAME have a payload or data transfer size which 1090 * is different from the size of the request. Any driver that supports such 1091 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to 1092 * calculate the data transfer size. 1093 */ 1094 static inline unsigned int blk_rq_payload_bytes(struct request *rq) 1095 { 1096 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 1097 return rq->special_vec.bv_len; 1098 return blk_rq_bytes(rq); 1099 } 1100 1101 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, 1102 int op) 1103 { 1104 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) 1105 return min(q->limits.max_discard_sectors, 1106 UINT_MAX >> SECTOR_SHIFT); 1107 1108 if (unlikely(op == REQ_OP_WRITE_SAME)) 1109 return q->limits.max_write_same_sectors; 1110 1111 if (unlikely(op == REQ_OP_WRITE_ZEROES)) 1112 return q->limits.max_write_zeroes_sectors; 1113 1114 return q->limits.max_sectors; 1115 } 1116 1117 /* 1118 * Return maximum size of a request at given offset. Only valid for 1119 * file system requests. 1120 */ 1121 static inline unsigned int blk_max_size_offset(struct request_queue *q, 1122 sector_t offset) 1123 { 1124 if (!q->limits.chunk_sectors) 1125 return q->limits.max_sectors; 1126 1127 return q->limits.chunk_sectors - 1128 (offset & (q->limits.chunk_sectors - 1)); 1129 } 1130 1131 static inline unsigned int blk_rq_get_max_sectors(struct request *rq, 1132 sector_t offset) 1133 { 1134 struct request_queue *q = rq->q; 1135 1136 if (blk_rq_is_passthrough(rq)) 1137 return q->limits.max_hw_sectors; 1138 1139 if (!q->limits.chunk_sectors || 1140 req_op(rq) == REQ_OP_DISCARD || 1141 req_op(rq) == REQ_OP_SECURE_ERASE) 1142 return blk_queue_get_max_sectors(q, req_op(rq)); 1143 1144 return min(blk_max_size_offset(q, offset), 1145 blk_queue_get_max_sectors(q, req_op(rq))); 1146 } 1147 1148 static inline unsigned int blk_rq_count_bios(struct request *rq) 1149 { 1150 unsigned int nr_bios = 0; 1151 struct bio *bio; 1152 1153 __rq_for_each_bio(bio, rq) 1154 nr_bios++; 1155 1156 return nr_bios; 1157 } 1158 1159 /* 1160 * Request issue related functions. 1161 */ 1162 extern struct request *blk_peek_request(struct request_queue *q); 1163 extern void blk_start_request(struct request *rq); 1164 extern struct request *blk_fetch_request(struct request_queue *q); 1165 1166 void blk_steal_bios(struct bio_list *list, struct request *rq); 1167 1168 /* 1169 * Request completion related functions. 1170 * 1171 * blk_update_request() completes given number of bytes and updates 1172 * the request without completing it. 1173 * 1174 * blk_end_request() and friends. __blk_end_request() must be called 1175 * with the request queue spinlock acquired. 1176 * 1177 * Several drivers define their own end_request and call 1178 * blk_end_request() for parts of the original function. 1179 * This prevents code duplication in drivers. 1180 */ 1181 extern bool blk_update_request(struct request *rq, blk_status_t error, 1182 unsigned int nr_bytes); 1183 extern void blk_finish_request(struct request *rq, blk_status_t error); 1184 extern bool blk_end_request(struct request *rq, blk_status_t error, 1185 unsigned int nr_bytes); 1186 extern void blk_end_request_all(struct request *rq, blk_status_t error); 1187 extern bool __blk_end_request(struct request *rq, blk_status_t error, 1188 unsigned int nr_bytes); 1189 extern void __blk_end_request_all(struct request *rq, blk_status_t error); 1190 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error); 1191 1192 extern void blk_complete_request(struct request *); 1193 extern void __blk_complete_request(struct request *); 1194 extern void blk_abort_request(struct request *); 1195 extern void blk_unprep_request(struct request *); 1196 1197 /* 1198 * Access functions for manipulating queue properties 1199 */ 1200 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn, 1201 spinlock_t *lock, int node_id); 1202 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *); 1203 extern int blk_init_allocated_queue(struct request_queue *); 1204 extern void blk_cleanup_queue(struct request_queue *); 1205 extern void blk_queue_make_request(struct request_queue *, make_request_fn *); 1206 extern void blk_queue_bounce_limit(struct request_queue *, u64); 1207 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); 1208 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); 1209 extern void blk_queue_max_segments(struct request_queue *, unsigned short); 1210 extern void blk_queue_max_discard_segments(struct request_queue *, 1211 unsigned short); 1212 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); 1213 extern void blk_queue_max_discard_sectors(struct request_queue *q, 1214 unsigned int max_discard_sectors); 1215 extern void blk_queue_max_write_same_sectors(struct request_queue *q, 1216 unsigned int max_write_same_sectors); 1217 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, 1218 unsigned int max_write_same_sectors); 1219 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); 1220 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); 1221 extern void blk_queue_alignment_offset(struct request_queue *q, 1222 unsigned int alignment); 1223 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 1224 extern void blk_queue_io_min(struct request_queue *q, unsigned int min); 1225 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 1226 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); 1227 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1228 extern void blk_set_default_limits(struct queue_limits *lim); 1229 extern void blk_set_stacking_limits(struct queue_limits *lim); 1230 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1231 sector_t offset); 1232 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, 1233 sector_t offset); 1234 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, 1235 sector_t offset); 1236 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); 1237 extern void blk_queue_dma_pad(struct request_queue *, unsigned int); 1238 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 1239 extern int blk_queue_dma_drain(struct request_queue *q, 1240 dma_drain_needed_fn *dma_drain_needed, 1241 void *buf, unsigned int size); 1242 extern void blk_queue_lld_busy(struct request_queue *q, lld_busy_fn *fn); 1243 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); 1244 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long); 1245 extern void blk_queue_prep_rq(struct request_queue *, prep_rq_fn *pfn); 1246 extern void blk_queue_unprep_rq(struct request_queue *, unprep_rq_fn *ufn); 1247 extern void blk_queue_dma_alignment(struct request_queue *, int); 1248 extern void blk_queue_update_dma_alignment(struct request_queue *, int); 1249 extern void blk_queue_softirq_done(struct request_queue *, softirq_done_fn *); 1250 extern void blk_queue_rq_timed_out(struct request_queue *, rq_timed_out_fn *); 1251 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1252 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable); 1253 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); 1254 1255 /* 1256 * Number of physical segments as sent to the device. 1257 * 1258 * Normally this is the number of discontiguous data segments sent by the 1259 * submitter. But for data-less command like discard we might have no 1260 * actual data segments submitted, but the driver might have to add it's 1261 * own special payload. In that case we still return 1 here so that this 1262 * special payload will be mapped. 1263 */ 1264 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq) 1265 { 1266 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 1267 return 1; 1268 return rq->nr_phys_segments; 1269 } 1270 1271 /* 1272 * Number of discard segments (or ranges) the driver needs to fill in. 1273 * Each discard bio merged into a request is counted as one segment. 1274 */ 1275 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq) 1276 { 1277 return max_t(unsigned short, rq->nr_phys_segments, 1); 1278 } 1279 1280 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); 1281 extern void blk_dump_rq_flags(struct request *, char *); 1282 extern long nr_blockdev_pages(void); 1283 1284 bool __must_check blk_get_queue(struct request_queue *); 1285 struct request_queue *blk_alloc_queue(gfp_t); 1286 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id, 1287 spinlock_t *lock); 1288 extern void blk_put_queue(struct request_queue *); 1289 extern void blk_set_queue_dying(struct request_queue *); 1290 1291 /* 1292 * block layer runtime pm functions 1293 */ 1294 #ifdef CONFIG_PM 1295 extern void blk_pm_runtime_init(struct request_queue *q, struct device *dev); 1296 extern int blk_pre_runtime_suspend(struct request_queue *q); 1297 extern void blk_post_runtime_suspend(struct request_queue *q, int err); 1298 extern void blk_pre_runtime_resume(struct request_queue *q); 1299 extern void blk_post_runtime_resume(struct request_queue *q, int err); 1300 extern void blk_set_runtime_active(struct request_queue *q); 1301 #else 1302 static inline void blk_pm_runtime_init(struct request_queue *q, 1303 struct device *dev) {} 1304 static inline int blk_pre_runtime_suspend(struct request_queue *q) 1305 { 1306 return -ENOSYS; 1307 } 1308 static inline void blk_post_runtime_suspend(struct request_queue *q, int err) {} 1309 static inline void blk_pre_runtime_resume(struct request_queue *q) {} 1310 static inline void blk_post_runtime_resume(struct request_queue *q, int err) {} 1311 static inline void blk_set_runtime_active(struct request_queue *q) {} 1312 #endif 1313 1314 /* 1315 * blk_plug permits building a queue of related requests by holding the I/O 1316 * fragments for a short period. This allows merging of sequential requests 1317 * into single larger request. As the requests are moved from a per-task list to 1318 * the device's request_queue in a batch, this results in improved scalability 1319 * as the lock contention for request_queue lock is reduced. 1320 * 1321 * It is ok not to disable preemption when adding the request to the plug list 1322 * or when attempting a merge, because blk_schedule_flush_list() will only flush 1323 * the plug list when the task sleeps by itself. For details, please see 1324 * schedule() where blk_schedule_flush_plug() is called. 1325 */ 1326 struct blk_plug { 1327 struct list_head list; /* requests */ 1328 struct list_head mq_list; /* blk-mq requests */ 1329 struct list_head cb_list; /* md requires an unplug callback */ 1330 }; 1331 #define BLK_MAX_REQUEST_COUNT 16 1332 #define BLK_PLUG_FLUSH_SIZE (128 * 1024) 1333 1334 struct blk_plug_cb; 1335 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1336 struct blk_plug_cb { 1337 struct list_head list; 1338 blk_plug_cb_fn callback; 1339 void *data; 1340 }; 1341 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1342 void *data, int size); 1343 extern void blk_start_plug(struct blk_plug *); 1344 extern void blk_finish_plug(struct blk_plug *); 1345 extern void blk_flush_plug_list(struct blk_plug *, bool); 1346 1347 static inline void blk_flush_plug(struct task_struct *tsk) 1348 { 1349 struct blk_plug *plug = tsk->plug; 1350 1351 if (plug) 1352 blk_flush_plug_list(plug, false); 1353 } 1354 1355 static inline void blk_schedule_flush_plug(struct task_struct *tsk) 1356 { 1357 struct blk_plug *plug = tsk->plug; 1358 1359 if (plug) 1360 blk_flush_plug_list(plug, true); 1361 } 1362 1363 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1364 { 1365 struct blk_plug *plug = tsk->plug; 1366 1367 return plug && 1368 (!list_empty(&plug->list) || 1369 !list_empty(&plug->mq_list) || 1370 !list_empty(&plug->cb_list)); 1371 } 1372 1373 /* 1374 * tag stuff 1375 */ 1376 extern int blk_queue_start_tag(struct request_queue *, struct request *); 1377 extern struct request *blk_queue_find_tag(struct request_queue *, int); 1378 extern void blk_queue_end_tag(struct request_queue *, struct request *); 1379 extern int blk_queue_init_tags(struct request_queue *, int, struct blk_queue_tag *, int); 1380 extern void blk_queue_free_tags(struct request_queue *); 1381 extern int blk_queue_resize_tags(struct request_queue *, int); 1382 extern void blk_queue_invalidate_tags(struct request_queue *); 1383 extern struct blk_queue_tag *blk_init_tags(int, int); 1384 extern void blk_free_tags(struct blk_queue_tag *); 1385 1386 static inline struct request *blk_map_queue_find_tag(struct blk_queue_tag *bqt, 1387 int tag) 1388 { 1389 if (unlikely(bqt == NULL || tag >= bqt->real_max_depth)) 1390 return NULL; 1391 return bqt->tag_index[tag]; 1392 } 1393 1394 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); 1395 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, 1396 sector_t nr_sects, gfp_t gfp_mask, struct page *page); 1397 1398 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */ 1399 1400 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1401 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); 1402 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1403 sector_t nr_sects, gfp_t gfp_mask, int flags, 1404 struct bio **biop); 1405 1406 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1407 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1408 1409 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1410 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1411 unsigned flags); 1412 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1413 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1414 1415 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1416 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1417 { 1418 return blkdev_issue_discard(sb->s_bdev, 1419 block << (sb->s_blocksize_bits - 1420 SECTOR_SHIFT), 1421 nr_blocks << (sb->s_blocksize_bits - 1422 SECTOR_SHIFT), 1423 gfp_mask, flags); 1424 } 1425 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1426 sector_t nr_blocks, gfp_t gfp_mask) 1427 { 1428 return blkdev_issue_zeroout(sb->s_bdev, 1429 block << (sb->s_blocksize_bits - 1430 SECTOR_SHIFT), 1431 nr_blocks << (sb->s_blocksize_bits - 1432 SECTOR_SHIFT), 1433 gfp_mask, 0); 1434 } 1435 1436 extern int blk_verify_command(unsigned char *cmd, fmode_t mode); 1437 1438 enum blk_default_limits { 1439 BLK_MAX_SEGMENTS = 128, 1440 BLK_SAFE_MAX_SECTORS = 255, 1441 BLK_DEF_MAX_SECTORS = 2560, 1442 BLK_MAX_SEGMENT_SIZE = 65536, 1443 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1444 }; 1445 1446 #define blkdev_entry_to_request(entry) list_entry((entry), struct request, queuelist) 1447 1448 static inline unsigned long queue_segment_boundary(struct request_queue *q) 1449 { 1450 return q->limits.seg_boundary_mask; 1451 } 1452 1453 static inline unsigned long queue_virt_boundary(struct request_queue *q) 1454 { 1455 return q->limits.virt_boundary_mask; 1456 } 1457 1458 static inline unsigned int queue_max_sectors(struct request_queue *q) 1459 { 1460 return q->limits.max_sectors; 1461 } 1462 1463 static inline unsigned int queue_max_hw_sectors(struct request_queue *q) 1464 { 1465 return q->limits.max_hw_sectors; 1466 } 1467 1468 static inline unsigned short queue_max_segments(struct request_queue *q) 1469 { 1470 return q->limits.max_segments; 1471 } 1472 1473 static inline unsigned short queue_max_discard_segments(struct request_queue *q) 1474 { 1475 return q->limits.max_discard_segments; 1476 } 1477 1478 static inline unsigned int queue_max_segment_size(struct request_queue *q) 1479 { 1480 return q->limits.max_segment_size; 1481 } 1482 1483 static inline unsigned short queue_logical_block_size(struct request_queue *q) 1484 { 1485 int retval = 512; 1486 1487 if (q && q->limits.logical_block_size) 1488 retval = q->limits.logical_block_size; 1489 1490 return retval; 1491 } 1492 1493 static inline unsigned short bdev_logical_block_size(struct block_device *bdev) 1494 { 1495 return queue_logical_block_size(bdev_get_queue(bdev)); 1496 } 1497 1498 static inline unsigned int queue_physical_block_size(struct request_queue *q) 1499 { 1500 return q->limits.physical_block_size; 1501 } 1502 1503 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1504 { 1505 return queue_physical_block_size(bdev_get_queue(bdev)); 1506 } 1507 1508 static inline unsigned int queue_io_min(struct request_queue *q) 1509 { 1510 return q->limits.io_min; 1511 } 1512 1513 static inline int bdev_io_min(struct block_device *bdev) 1514 { 1515 return queue_io_min(bdev_get_queue(bdev)); 1516 } 1517 1518 static inline unsigned int queue_io_opt(struct request_queue *q) 1519 { 1520 return q->limits.io_opt; 1521 } 1522 1523 static inline int bdev_io_opt(struct block_device *bdev) 1524 { 1525 return queue_io_opt(bdev_get_queue(bdev)); 1526 } 1527 1528 static inline int queue_alignment_offset(struct request_queue *q) 1529 { 1530 if (q->limits.misaligned) 1531 return -1; 1532 1533 return q->limits.alignment_offset; 1534 } 1535 1536 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) 1537 { 1538 unsigned int granularity = max(lim->physical_block_size, lim->io_min); 1539 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) 1540 << SECTOR_SHIFT; 1541 1542 return (granularity + lim->alignment_offset - alignment) % granularity; 1543 } 1544 1545 static inline int bdev_alignment_offset(struct block_device *bdev) 1546 { 1547 struct request_queue *q = bdev_get_queue(bdev); 1548 1549 if (q->limits.misaligned) 1550 return -1; 1551 1552 if (bdev != bdev->bd_contains) 1553 return bdev->bd_part->alignment_offset; 1554 1555 return q->limits.alignment_offset; 1556 } 1557 1558 static inline int queue_discard_alignment(struct request_queue *q) 1559 { 1560 if (q->limits.discard_misaligned) 1561 return -1; 1562 1563 return q->limits.discard_alignment; 1564 } 1565 1566 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) 1567 { 1568 unsigned int alignment, granularity, offset; 1569 1570 if (!lim->max_discard_sectors) 1571 return 0; 1572 1573 /* Why are these in bytes, not sectors? */ 1574 alignment = lim->discard_alignment >> SECTOR_SHIFT; 1575 granularity = lim->discard_granularity >> SECTOR_SHIFT; 1576 if (!granularity) 1577 return 0; 1578 1579 /* Offset of the partition start in 'granularity' sectors */ 1580 offset = sector_div(sector, granularity); 1581 1582 /* And why do we do this modulus *again* in blkdev_issue_discard()? */ 1583 offset = (granularity + alignment - offset) % granularity; 1584 1585 /* Turn it back into bytes, gaah */ 1586 return offset << SECTOR_SHIFT; 1587 } 1588 1589 static inline int bdev_discard_alignment(struct block_device *bdev) 1590 { 1591 struct request_queue *q = bdev_get_queue(bdev); 1592 1593 if (bdev != bdev->bd_contains) 1594 return bdev->bd_part->discard_alignment; 1595 1596 return q->limits.discard_alignment; 1597 } 1598 1599 static inline unsigned int bdev_write_same(struct block_device *bdev) 1600 { 1601 struct request_queue *q = bdev_get_queue(bdev); 1602 1603 if (q) 1604 return q->limits.max_write_same_sectors; 1605 1606 return 0; 1607 } 1608 1609 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1610 { 1611 struct request_queue *q = bdev_get_queue(bdev); 1612 1613 if (q) 1614 return q->limits.max_write_zeroes_sectors; 1615 1616 return 0; 1617 } 1618 1619 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev) 1620 { 1621 struct request_queue *q = bdev_get_queue(bdev); 1622 1623 if (q) 1624 return blk_queue_zoned_model(q); 1625 1626 return BLK_ZONED_NONE; 1627 } 1628 1629 static inline bool bdev_is_zoned(struct block_device *bdev) 1630 { 1631 struct request_queue *q = bdev_get_queue(bdev); 1632 1633 if (q) 1634 return blk_queue_is_zoned(q); 1635 1636 return false; 1637 } 1638 1639 static inline unsigned int bdev_zone_sectors(struct block_device *bdev) 1640 { 1641 struct request_queue *q = bdev_get_queue(bdev); 1642 1643 if (q) 1644 return blk_queue_zone_sectors(q); 1645 return 0; 1646 } 1647 1648 static inline unsigned int bdev_nr_zones(struct block_device *bdev) 1649 { 1650 struct request_queue *q = bdev_get_queue(bdev); 1651 1652 if (q) 1653 return blk_queue_nr_zones(q); 1654 return 0; 1655 } 1656 1657 static inline int queue_dma_alignment(struct request_queue *q) 1658 { 1659 return q ? q->dma_alignment : 511; 1660 } 1661 1662 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1663 unsigned int len) 1664 { 1665 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1666 return !(addr & alignment) && !(len & alignment); 1667 } 1668 1669 /* assumes size > 256 */ 1670 static inline unsigned int blksize_bits(unsigned int size) 1671 { 1672 unsigned int bits = 8; 1673 do { 1674 bits++; 1675 size >>= 1; 1676 } while (size > 256); 1677 return bits; 1678 } 1679 1680 static inline unsigned int block_size(struct block_device *bdev) 1681 { 1682 return bdev->bd_block_size; 1683 } 1684 1685 static inline bool queue_flush_queueable(struct request_queue *q) 1686 { 1687 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); 1688 } 1689 1690 typedef struct {struct page *v;} Sector; 1691 1692 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); 1693 1694 static inline void put_dev_sector(Sector p) 1695 { 1696 put_page(p.v); 1697 } 1698 1699 static inline bool __bvec_gap_to_prev(struct request_queue *q, 1700 struct bio_vec *bprv, unsigned int offset) 1701 { 1702 return offset || 1703 ((bprv->bv_offset + bprv->bv_len) & queue_virt_boundary(q)); 1704 } 1705 1706 /* 1707 * Check if adding a bio_vec after bprv with offset would create a gap in 1708 * the SG list. Most drivers don't care about this, but some do. 1709 */ 1710 static inline bool bvec_gap_to_prev(struct request_queue *q, 1711 struct bio_vec *bprv, unsigned int offset) 1712 { 1713 if (!queue_virt_boundary(q)) 1714 return false; 1715 return __bvec_gap_to_prev(q, bprv, offset); 1716 } 1717 1718 /* 1719 * Check if the two bvecs from two bios can be merged to one segment. 1720 * If yes, no need to check gap between the two bios since the 1st bio 1721 * and the 1st bvec in the 2nd bio can be handled in one segment. 1722 */ 1723 static inline bool bios_segs_mergeable(struct request_queue *q, 1724 struct bio *prev, struct bio_vec *prev_last_bv, 1725 struct bio_vec *next_first_bv) 1726 { 1727 if (!BIOVEC_PHYS_MERGEABLE(prev_last_bv, next_first_bv)) 1728 return false; 1729 if (!BIOVEC_SEG_BOUNDARY(q, prev_last_bv, next_first_bv)) 1730 return false; 1731 if (prev->bi_seg_back_size + next_first_bv->bv_len > 1732 queue_max_segment_size(q)) 1733 return false; 1734 return true; 1735 } 1736 1737 static inline bool bio_will_gap(struct request_queue *q, 1738 struct request *prev_rq, 1739 struct bio *prev, 1740 struct bio *next) 1741 { 1742 if (bio_has_data(prev) && queue_virt_boundary(q)) { 1743 struct bio_vec pb, nb; 1744 1745 /* 1746 * don't merge if the 1st bio starts with non-zero 1747 * offset, otherwise it is quite difficult to respect 1748 * sg gap limit. We work hard to merge a huge number of small 1749 * single bios in case of mkfs. 1750 */ 1751 if (prev_rq) 1752 bio_get_first_bvec(prev_rq->bio, &pb); 1753 else 1754 bio_get_first_bvec(prev, &pb); 1755 if (pb.bv_offset) 1756 return true; 1757 1758 /* 1759 * We don't need to worry about the situation that the 1760 * merged segment ends in unaligned virt boundary: 1761 * 1762 * - if 'pb' ends aligned, the merged segment ends aligned 1763 * - if 'pb' ends unaligned, the next bio must include 1764 * one single bvec of 'nb', otherwise the 'nb' can't 1765 * merge with 'pb' 1766 */ 1767 bio_get_last_bvec(prev, &pb); 1768 bio_get_first_bvec(next, &nb); 1769 1770 if (!bios_segs_mergeable(q, prev, &pb, &nb)) 1771 return __bvec_gap_to_prev(q, &pb, nb.bv_offset); 1772 } 1773 1774 return false; 1775 } 1776 1777 static inline bool req_gap_back_merge(struct request *req, struct bio *bio) 1778 { 1779 return bio_will_gap(req->q, req, req->biotail, bio); 1780 } 1781 1782 static inline bool req_gap_front_merge(struct request *req, struct bio *bio) 1783 { 1784 return bio_will_gap(req->q, NULL, bio, req->bio); 1785 } 1786 1787 int kblockd_schedule_work(struct work_struct *work); 1788 int kblockd_schedule_work_on(int cpu, struct work_struct *work); 1789 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1790 1791 #ifdef CONFIG_BLK_CGROUP 1792 /* 1793 * This should not be using sched_clock(). A real patch is in progress 1794 * to fix this up, until that is in place we need to disable preemption 1795 * around sched_clock() in this function and set_io_start_time_ns(). 1796 */ 1797 static inline void set_start_time_ns(struct request *req) 1798 { 1799 preempt_disable(); 1800 req->start_time_ns = sched_clock(); 1801 preempt_enable(); 1802 } 1803 1804 static inline void set_io_start_time_ns(struct request *req) 1805 { 1806 preempt_disable(); 1807 req->io_start_time_ns = sched_clock(); 1808 preempt_enable(); 1809 } 1810 1811 static inline uint64_t rq_start_time_ns(struct request *req) 1812 { 1813 return req->start_time_ns; 1814 } 1815 1816 static inline uint64_t rq_io_start_time_ns(struct request *req) 1817 { 1818 return req->io_start_time_ns; 1819 } 1820 #else 1821 static inline void set_start_time_ns(struct request *req) {} 1822 static inline void set_io_start_time_ns(struct request *req) {} 1823 static inline uint64_t rq_start_time_ns(struct request *req) 1824 { 1825 return 0; 1826 } 1827 static inline uint64_t rq_io_start_time_ns(struct request *req) 1828 { 1829 return 0; 1830 } 1831 #endif 1832 1833 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1834 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1835 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1836 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1837 1838 #if defined(CONFIG_BLK_DEV_INTEGRITY) 1839 1840 enum blk_integrity_flags { 1841 BLK_INTEGRITY_VERIFY = 1 << 0, 1842 BLK_INTEGRITY_GENERATE = 1 << 1, 1843 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, 1844 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, 1845 }; 1846 1847 struct blk_integrity_iter { 1848 void *prot_buf; 1849 void *data_buf; 1850 sector_t seed; 1851 unsigned int data_size; 1852 unsigned short interval; 1853 const char *disk_name; 1854 }; 1855 1856 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *); 1857 1858 struct blk_integrity_profile { 1859 integrity_processing_fn *generate_fn; 1860 integrity_processing_fn *verify_fn; 1861 const char *name; 1862 }; 1863 1864 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *); 1865 extern void blk_integrity_unregister(struct gendisk *); 1866 extern int blk_integrity_compare(struct gendisk *, struct gendisk *); 1867 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, 1868 struct scatterlist *); 1869 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); 1870 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, 1871 struct request *); 1872 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, 1873 struct bio *); 1874 1875 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1876 { 1877 struct blk_integrity *bi = &disk->queue->integrity; 1878 1879 if (!bi->profile) 1880 return NULL; 1881 1882 return bi; 1883 } 1884 1885 static inline 1886 struct blk_integrity *bdev_get_integrity(struct block_device *bdev) 1887 { 1888 return blk_get_integrity(bdev->bd_disk); 1889 } 1890 1891 static inline bool blk_integrity_rq(struct request *rq) 1892 { 1893 return rq->cmd_flags & REQ_INTEGRITY; 1894 } 1895 1896 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1897 unsigned int segs) 1898 { 1899 q->limits.max_integrity_segments = segs; 1900 } 1901 1902 static inline unsigned short 1903 queue_max_integrity_segments(struct request_queue *q) 1904 { 1905 return q->limits.max_integrity_segments; 1906 } 1907 1908 static inline bool integrity_req_gap_back_merge(struct request *req, 1909 struct bio *next) 1910 { 1911 struct bio_integrity_payload *bip = bio_integrity(req->bio); 1912 struct bio_integrity_payload *bip_next = bio_integrity(next); 1913 1914 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1], 1915 bip_next->bip_vec[0].bv_offset); 1916 } 1917 1918 static inline bool integrity_req_gap_front_merge(struct request *req, 1919 struct bio *bio) 1920 { 1921 struct bio_integrity_payload *bip = bio_integrity(bio); 1922 struct bio_integrity_payload *bip_next = bio_integrity(req->bio); 1923 1924 return bvec_gap_to_prev(req->q, &bip->bip_vec[bip->bip_vcnt - 1], 1925 bip_next->bip_vec[0].bv_offset); 1926 } 1927 1928 #else /* CONFIG_BLK_DEV_INTEGRITY */ 1929 1930 struct bio; 1931 struct block_device; 1932 struct gendisk; 1933 struct blk_integrity; 1934 1935 static inline int blk_integrity_rq(struct request *rq) 1936 { 1937 return 0; 1938 } 1939 static inline int blk_rq_count_integrity_sg(struct request_queue *q, 1940 struct bio *b) 1941 { 1942 return 0; 1943 } 1944 static inline int blk_rq_map_integrity_sg(struct request_queue *q, 1945 struct bio *b, 1946 struct scatterlist *s) 1947 { 1948 return 0; 1949 } 1950 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) 1951 { 1952 return NULL; 1953 } 1954 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1955 { 1956 return NULL; 1957 } 1958 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) 1959 { 1960 return 0; 1961 } 1962 static inline void blk_integrity_register(struct gendisk *d, 1963 struct blk_integrity *b) 1964 { 1965 } 1966 static inline void blk_integrity_unregister(struct gendisk *d) 1967 { 1968 } 1969 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1970 unsigned int segs) 1971 { 1972 } 1973 static inline unsigned short queue_max_integrity_segments(struct request_queue *q) 1974 { 1975 return 0; 1976 } 1977 static inline bool blk_integrity_merge_rq(struct request_queue *rq, 1978 struct request *r1, 1979 struct request *r2) 1980 { 1981 return true; 1982 } 1983 static inline bool blk_integrity_merge_bio(struct request_queue *rq, 1984 struct request *r, 1985 struct bio *b) 1986 { 1987 return true; 1988 } 1989 1990 static inline bool integrity_req_gap_back_merge(struct request *req, 1991 struct bio *next) 1992 { 1993 return false; 1994 } 1995 static inline bool integrity_req_gap_front_merge(struct request *req, 1996 struct bio *bio) 1997 { 1998 return false; 1999 } 2000 2001 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 2002 2003 struct block_device_operations { 2004 int (*open) (struct block_device *, fmode_t); 2005 void (*release) (struct gendisk *, fmode_t); 2006 int (*rw_page)(struct block_device *, sector_t, struct page *, bool); 2007 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 2008 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 2009 unsigned int (*check_events) (struct gendisk *disk, 2010 unsigned int clearing); 2011 /* ->media_changed() is DEPRECATED, use ->check_events() instead */ 2012 int (*media_changed) (struct gendisk *); 2013 void (*unlock_native_capacity) (struct gendisk *); 2014 int (*revalidate_disk) (struct gendisk *); 2015 int (*getgeo)(struct block_device *, struct hd_geometry *); 2016 /* this callback is with swap_lock and sometimes page table lock held */ 2017 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 2018 struct module *owner; 2019 const struct pr_ops *pr_ops; 2020 }; 2021 2022 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, 2023 unsigned long); 2024 extern int bdev_read_page(struct block_device *, sector_t, struct page *); 2025 extern int bdev_write_page(struct block_device *, sector_t, struct page *, 2026 struct writeback_control *); 2027 2028 #ifdef CONFIG_BLK_DEV_ZONED 2029 bool blk_req_needs_zone_write_lock(struct request *rq); 2030 void __blk_req_zone_write_lock(struct request *rq); 2031 void __blk_req_zone_write_unlock(struct request *rq); 2032 2033 static inline void blk_req_zone_write_lock(struct request *rq) 2034 { 2035 if (blk_req_needs_zone_write_lock(rq)) 2036 __blk_req_zone_write_lock(rq); 2037 } 2038 2039 static inline void blk_req_zone_write_unlock(struct request *rq) 2040 { 2041 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED) 2042 __blk_req_zone_write_unlock(rq); 2043 } 2044 2045 static inline bool blk_req_zone_is_write_locked(struct request *rq) 2046 { 2047 return rq->q->seq_zones_wlock && 2048 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock); 2049 } 2050 2051 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 2052 { 2053 if (!blk_req_needs_zone_write_lock(rq)) 2054 return true; 2055 return !blk_req_zone_is_write_locked(rq); 2056 } 2057 #else 2058 static inline bool blk_req_needs_zone_write_lock(struct request *rq) 2059 { 2060 return false; 2061 } 2062 2063 static inline void blk_req_zone_write_lock(struct request *rq) 2064 { 2065 } 2066 2067 static inline void blk_req_zone_write_unlock(struct request *rq) 2068 { 2069 } 2070 static inline bool blk_req_zone_is_write_locked(struct request *rq) 2071 { 2072 return false; 2073 } 2074 2075 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 2076 { 2077 return true; 2078 } 2079 #endif /* CONFIG_BLK_DEV_ZONED */ 2080 2081 #else /* CONFIG_BLOCK */ 2082 2083 struct block_device; 2084 2085 /* 2086 * stubs for when the block layer is configured out 2087 */ 2088 #define buffer_heads_over_limit 0 2089 2090 static inline long nr_blockdev_pages(void) 2091 { 2092 return 0; 2093 } 2094 2095 struct blk_plug { 2096 }; 2097 2098 static inline void blk_start_plug(struct blk_plug *plug) 2099 { 2100 } 2101 2102 static inline void blk_finish_plug(struct blk_plug *plug) 2103 { 2104 } 2105 2106 static inline void blk_flush_plug(struct task_struct *task) 2107 { 2108 } 2109 2110 static inline void blk_schedule_flush_plug(struct task_struct *task) 2111 { 2112 } 2113 2114 2115 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 2116 { 2117 return false; 2118 } 2119 2120 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, 2121 sector_t *error_sector) 2122 { 2123 return 0; 2124 } 2125 2126 #endif /* CONFIG_BLOCK */ 2127 2128 #endif 2129