1 /* SPDX-License-Identifier: GPL-2.0 */ 2 #ifndef _LINUX_BLKDEV_H 3 #define _LINUX_BLKDEV_H 4 5 #include <linux/sched.h> 6 #include <linux/sched/clock.h> 7 8 #ifdef CONFIG_BLOCK 9 10 #include <linux/major.h> 11 #include <linux/genhd.h> 12 #include <linux/list.h> 13 #include <linux/llist.h> 14 #include <linux/timer.h> 15 #include <linux/workqueue.h> 16 #include <linux/pagemap.h> 17 #include <linux/backing-dev-defs.h> 18 #include <linux/wait.h> 19 #include <linux/mempool.h> 20 #include <linux/pfn.h> 21 #include <linux/bio.h> 22 #include <linux/stringify.h> 23 #include <linux/gfp.h> 24 #include <linux/bsg.h> 25 #include <linux/smp.h> 26 #include <linux/rcupdate.h> 27 #include <linux/percpu-refcount.h> 28 #include <linux/scatterlist.h> 29 #include <linux/blkzoned.h> 30 31 struct module; 32 struct scsi_ioctl_command; 33 34 struct request_queue; 35 struct elevator_queue; 36 struct blk_trace; 37 struct request; 38 struct sg_io_hdr; 39 struct bsg_job; 40 struct blkcg_gq; 41 struct blk_flush_queue; 42 struct pr_ops; 43 struct rq_qos; 44 struct blk_queue_stats; 45 struct blk_stat_callback; 46 47 #define BLKDEV_MIN_RQ 4 48 #define BLKDEV_MAX_RQ 128 /* Default maximum */ 49 50 /* Must be consistent with blk_mq_poll_stats_bkt() */ 51 #define BLK_MQ_POLL_STATS_BKTS 16 52 53 /* Doing classic polling */ 54 #define BLK_MQ_POLL_CLASSIC -1 55 56 /* 57 * Maximum number of blkcg policies allowed to be registered concurrently. 58 * Defined here to simplify include dependency. 59 */ 60 #define BLKCG_MAX_POLS 5 61 62 typedef void (rq_end_io_fn)(struct request *, blk_status_t); 63 64 /* 65 * request flags */ 66 typedef __u32 __bitwise req_flags_t; 67 68 /* elevator knows about this request */ 69 #define RQF_SORTED ((__force req_flags_t)(1 << 0)) 70 /* drive already may have started this one */ 71 #define RQF_STARTED ((__force req_flags_t)(1 << 1)) 72 /* may not be passed by ioscheduler */ 73 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3)) 74 /* request for flush sequence */ 75 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4)) 76 /* merge of different types, fail separately */ 77 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5)) 78 /* track inflight for MQ */ 79 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6)) 80 /* don't call prep for this one */ 81 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7)) 82 /* set for "ide_preempt" requests and also for requests for which the SCSI 83 "quiesce" state must be ignored. */ 84 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8)) 85 /* contains copies of user pages */ 86 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9)) 87 /* vaguely specified driver internal error. Ignored by the block layer */ 88 #define RQF_FAILED ((__force req_flags_t)(1 << 10)) 89 /* don't warn about errors */ 90 #define RQF_QUIET ((__force req_flags_t)(1 << 11)) 91 /* elevator private data attached */ 92 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12)) 93 /* account into disk and partition IO statistics */ 94 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13)) 95 /* request came from our alloc pool */ 96 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14)) 97 /* runtime pm request */ 98 #define RQF_PM ((__force req_flags_t)(1 << 15)) 99 /* on IO scheduler merge hash */ 100 #define RQF_HASHED ((__force req_flags_t)(1 << 16)) 101 /* track IO completion time */ 102 #define RQF_STATS ((__force req_flags_t)(1 << 17)) 103 /* Look at ->special_vec for the actual data payload instead of the 104 bio chain. */ 105 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18)) 106 /* The per-zone write lock is held for this request */ 107 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) 108 /* already slept for hybrid poll */ 109 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20)) 110 /* ->timeout has been called, don't expire again */ 111 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21)) 112 113 /* flags that prevent us from merging requests: */ 114 #define RQF_NOMERGE_FLAGS \ 115 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD) 116 117 /* 118 * Request state for blk-mq. 119 */ 120 enum mq_rq_state { 121 MQ_RQ_IDLE = 0, 122 MQ_RQ_IN_FLIGHT = 1, 123 MQ_RQ_COMPLETE = 2, 124 }; 125 126 /* 127 * Try to put the fields that are referenced together in the same cacheline. 128 * 129 * If you modify this structure, make sure to update blk_rq_init() and 130 * especially blk_mq_rq_ctx_init() to take care of the added fields. 131 */ 132 struct request { 133 struct request_queue *q; 134 struct blk_mq_ctx *mq_ctx; 135 struct blk_mq_hw_ctx *mq_hctx; 136 137 unsigned int cmd_flags; /* op and common flags */ 138 req_flags_t rq_flags; 139 140 int tag; 141 int internal_tag; 142 143 /* the following two fields are internal, NEVER access directly */ 144 unsigned int __data_len; /* total data len */ 145 sector_t __sector; /* sector cursor */ 146 147 struct bio *bio; 148 struct bio *biotail; 149 150 struct list_head queuelist; 151 152 /* 153 * The hash is used inside the scheduler, and killed once the 154 * request reaches the dispatch list. The ipi_list is only used 155 * to queue the request for softirq completion, which is long 156 * after the request has been unhashed (and even removed from 157 * the dispatch list). 158 */ 159 union { 160 struct hlist_node hash; /* merge hash */ 161 struct list_head ipi_list; 162 }; 163 164 /* 165 * The rb_node is only used inside the io scheduler, requests 166 * are pruned when moved to the dispatch queue. So let the 167 * completion_data share space with the rb_node. 168 */ 169 union { 170 struct rb_node rb_node; /* sort/lookup */ 171 struct bio_vec special_vec; 172 void *completion_data; 173 int error_count; /* for legacy drivers, don't use */ 174 }; 175 176 /* 177 * Three pointers are available for the IO schedulers, if they need 178 * more they have to dynamically allocate it. Flush requests are 179 * never put on the IO scheduler. So let the flush fields share 180 * space with the elevator data. 181 */ 182 union { 183 struct { 184 struct io_cq *icq; 185 void *priv[2]; 186 } elv; 187 188 struct { 189 unsigned int seq; 190 struct list_head list; 191 rq_end_io_fn *saved_end_io; 192 } flush; 193 }; 194 195 struct gendisk *rq_disk; 196 struct hd_struct *part; 197 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 198 /* Time that the first bio started allocating this request. */ 199 u64 alloc_time_ns; 200 #endif 201 /* Time that this request was allocated for this IO. */ 202 u64 start_time_ns; 203 /* Time that I/O was submitted to the device. */ 204 u64 io_start_time_ns; 205 206 #ifdef CONFIG_BLK_WBT 207 unsigned short wbt_flags; 208 #endif 209 /* 210 * rq sectors used for blk stats. It has the same value 211 * with blk_rq_sectors(rq), except that it never be zeroed 212 * by completion. 213 */ 214 unsigned short stats_sectors; 215 216 /* 217 * Number of scatter-gather DMA addr+len pairs after 218 * physical address coalescing is performed. 219 */ 220 unsigned short nr_phys_segments; 221 222 #if defined(CONFIG_BLK_DEV_INTEGRITY) 223 unsigned short nr_integrity_segments; 224 #endif 225 226 unsigned short write_hint; 227 unsigned short ioprio; 228 229 unsigned int extra_len; /* length of alignment and padding */ 230 231 enum mq_rq_state state; 232 refcount_t ref; 233 234 unsigned int timeout; 235 unsigned long deadline; 236 237 union { 238 struct __call_single_data csd; 239 u64 fifo_time; 240 }; 241 242 /* 243 * completion callback. 244 */ 245 rq_end_io_fn *end_io; 246 void *end_io_data; 247 }; 248 249 static inline bool blk_op_is_scsi(unsigned int op) 250 { 251 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT; 252 } 253 254 static inline bool blk_op_is_private(unsigned int op) 255 { 256 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 257 } 258 259 static inline bool blk_rq_is_scsi(struct request *rq) 260 { 261 return blk_op_is_scsi(req_op(rq)); 262 } 263 264 static inline bool blk_rq_is_private(struct request *rq) 265 { 266 return blk_op_is_private(req_op(rq)); 267 } 268 269 static inline bool blk_rq_is_passthrough(struct request *rq) 270 { 271 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq); 272 } 273 274 static inline bool bio_is_passthrough(struct bio *bio) 275 { 276 unsigned op = bio_op(bio); 277 278 return blk_op_is_scsi(op) || blk_op_is_private(op); 279 } 280 281 static inline unsigned short req_get_ioprio(struct request *req) 282 { 283 return req->ioprio; 284 } 285 286 #include <linux/elevator.h> 287 288 struct blk_queue_ctx; 289 290 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio); 291 292 struct bio_vec; 293 typedef int (dma_drain_needed_fn)(struct request *); 294 295 enum blk_eh_timer_return { 296 BLK_EH_DONE, /* drivers has completed the command */ 297 BLK_EH_RESET_TIMER, /* reset timer and try again */ 298 }; 299 300 enum blk_queue_state { 301 Queue_down, 302 Queue_up, 303 }; 304 305 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ 306 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ 307 308 #define BLK_SCSI_MAX_CMDS (256) 309 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) 310 311 /* 312 * Zoned block device models (zoned limit). 313 */ 314 enum blk_zoned_model { 315 BLK_ZONED_NONE, /* Regular block device */ 316 BLK_ZONED_HA, /* Host-aware zoned block device */ 317 BLK_ZONED_HM, /* Host-managed zoned block device */ 318 }; 319 320 struct queue_limits { 321 unsigned long bounce_pfn; 322 unsigned long seg_boundary_mask; 323 unsigned long virt_boundary_mask; 324 325 unsigned int max_hw_sectors; 326 unsigned int max_dev_sectors; 327 unsigned int chunk_sectors; 328 unsigned int max_sectors; 329 unsigned int max_segment_size; 330 unsigned int physical_block_size; 331 unsigned int alignment_offset; 332 unsigned int io_min; 333 unsigned int io_opt; 334 unsigned int max_discard_sectors; 335 unsigned int max_hw_discard_sectors; 336 unsigned int max_write_same_sectors; 337 unsigned int max_write_zeroes_sectors; 338 unsigned int discard_granularity; 339 unsigned int discard_alignment; 340 341 unsigned short logical_block_size; 342 unsigned short max_segments; 343 unsigned short max_integrity_segments; 344 unsigned short max_discard_segments; 345 346 unsigned char misaligned; 347 unsigned char discard_misaligned; 348 unsigned char raid_partial_stripes_expensive; 349 enum blk_zoned_model zoned; 350 }; 351 352 #ifdef CONFIG_BLK_DEV_ZONED 353 354 /* 355 * Maximum number of zones to report with a single report zones command. 356 */ 357 #define BLK_ZONED_REPORT_MAX_ZONES 8192U 358 359 extern unsigned int blkdev_nr_zones(struct block_device *bdev); 360 extern int blkdev_report_zones(struct block_device *bdev, 361 sector_t sector, struct blk_zone *zones, 362 unsigned int *nr_zones); 363 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors, 364 sector_t nr_sectors, gfp_t gfp_mask); 365 extern int blk_revalidate_disk_zones(struct gendisk *disk); 366 367 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, 368 unsigned int cmd, unsigned long arg); 369 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode, 370 unsigned int cmd, unsigned long arg); 371 372 #else /* CONFIG_BLK_DEV_ZONED */ 373 374 static inline unsigned int blkdev_nr_zones(struct block_device *bdev) 375 { 376 return 0; 377 } 378 379 static inline int blk_revalidate_disk_zones(struct gendisk *disk) 380 { 381 return 0; 382 } 383 384 static inline int blkdev_report_zones_ioctl(struct block_device *bdev, 385 fmode_t mode, unsigned int cmd, 386 unsigned long arg) 387 { 388 return -ENOTTY; 389 } 390 391 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev, 392 fmode_t mode, unsigned int cmd, 393 unsigned long arg) 394 { 395 return -ENOTTY; 396 } 397 398 #endif /* CONFIG_BLK_DEV_ZONED */ 399 400 struct request_queue { 401 struct request *last_merge; 402 struct elevator_queue *elevator; 403 404 struct blk_queue_stats *stats; 405 struct rq_qos *rq_qos; 406 407 make_request_fn *make_request_fn; 408 dma_drain_needed_fn *dma_drain_needed; 409 410 const struct blk_mq_ops *mq_ops; 411 412 /* sw queues */ 413 struct blk_mq_ctx __percpu *queue_ctx; 414 unsigned int nr_queues; 415 416 unsigned int queue_depth; 417 418 /* hw dispatch queues */ 419 struct blk_mq_hw_ctx **queue_hw_ctx; 420 unsigned int nr_hw_queues; 421 422 struct backing_dev_info *backing_dev_info; 423 424 /* 425 * The queue owner gets to use this for whatever they like. 426 * ll_rw_blk doesn't touch it. 427 */ 428 void *queuedata; 429 430 /* 431 * various queue flags, see QUEUE_* below 432 */ 433 unsigned long queue_flags; 434 /* 435 * Number of contexts that have called blk_set_pm_only(). If this 436 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are 437 * processed. 438 */ 439 atomic_t pm_only; 440 441 /* 442 * ida allocated id for this queue. Used to index queues from 443 * ioctx. 444 */ 445 int id; 446 447 /* 448 * queue needs bounce pages for pages above this limit 449 */ 450 gfp_t bounce_gfp; 451 452 spinlock_t queue_lock; 453 454 /* 455 * queue kobject 456 */ 457 struct kobject kobj; 458 459 /* 460 * mq queue kobject 461 */ 462 struct kobject *mq_kobj; 463 464 #ifdef CONFIG_BLK_DEV_INTEGRITY 465 struct blk_integrity integrity; 466 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 467 468 #ifdef CONFIG_PM 469 struct device *dev; 470 int rpm_status; 471 unsigned int nr_pending; 472 #endif 473 474 /* 475 * queue settings 476 */ 477 unsigned long nr_requests; /* Max # of requests */ 478 479 unsigned int dma_drain_size; 480 void *dma_drain_buffer; 481 unsigned int dma_pad_mask; 482 unsigned int dma_alignment; 483 484 unsigned int rq_timeout; 485 int poll_nsec; 486 487 struct blk_stat_callback *poll_cb; 488 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS]; 489 490 struct timer_list timeout; 491 struct work_struct timeout_work; 492 493 struct list_head icq_list; 494 #ifdef CONFIG_BLK_CGROUP 495 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 496 struct blkcg_gq *root_blkg; 497 struct list_head blkg_list; 498 #endif 499 500 struct queue_limits limits; 501 502 unsigned int required_elevator_features; 503 504 #ifdef CONFIG_BLK_DEV_ZONED 505 /* 506 * Zoned block device information for request dispatch control. 507 * nr_zones is the total number of zones of the device. This is always 508 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones 509 * bits which indicates if a zone is conventional (bit clear) or 510 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones 511 * bits which indicates if a zone is write locked, that is, if a write 512 * request targeting the zone was dispatched. All three fields are 513 * initialized by the low level device driver (e.g. scsi/sd.c). 514 * Stacking drivers (device mappers) may or may not initialize 515 * these fields. 516 * 517 * Reads of this information must be protected with blk_queue_enter() / 518 * blk_queue_exit(). Modifying this information is only allowed while 519 * no requests are being processed. See also blk_mq_freeze_queue() and 520 * blk_mq_unfreeze_queue(). 521 */ 522 unsigned int nr_zones; 523 unsigned long *seq_zones_bitmap; 524 unsigned long *seq_zones_wlock; 525 #endif /* CONFIG_BLK_DEV_ZONED */ 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 struct mutex blk_trace_mutex; 536 #endif 537 /* 538 * for flush operations 539 */ 540 struct blk_flush_queue *fq; 541 542 struct list_head requeue_list; 543 spinlock_t requeue_lock; 544 struct delayed_work requeue_work; 545 546 struct mutex sysfs_lock; 547 struct mutex sysfs_dir_lock; 548 549 /* 550 * for reusing dead hctx instance in case of updating 551 * nr_hw_queues 552 */ 553 struct list_head unused_hctx_list; 554 spinlock_t unused_hctx_lock; 555 556 int mq_freeze_depth; 557 558 #if defined(CONFIG_BLK_DEV_BSG) 559 struct bsg_class_device bsg_dev; 560 #endif 561 562 #ifdef CONFIG_BLK_DEV_THROTTLING 563 /* Throttle data */ 564 struct throtl_data *td; 565 #endif 566 struct rcu_head rcu_head; 567 wait_queue_head_t mq_freeze_wq; 568 /* 569 * Protect concurrent access to q_usage_counter by 570 * percpu_ref_kill() and percpu_ref_reinit(). 571 */ 572 struct mutex mq_freeze_lock; 573 struct percpu_ref q_usage_counter; 574 575 struct blk_mq_tag_set *tag_set; 576 struct list_head tag_set_list; 577 struct bio_set bio_split; 578 579 #ifdef CONFIG_BLK_DEBUG_FS 580 struct dentry *debugfs_dir; 581 struct dentry *sched_debugfs_dir; 582 struct dentry *rqos_debugfs_dir; 583 #endif 584 585 bool mq_sysfs_init_done; 586 587 size_t cmd_size; 588 589 struct work_struct release_work; 590 591 #define BLK_MAX_WRITE_HINTS 5 592 u64 write_hints[BLK_MAX_WRITE_HINTS]; 593 }; 594 595 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */ 596 #define QUEUE_FLAG_DYING 1 /* queue being torn down */ 597 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */ 598 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ 599 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ 600 #define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */ 601 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ 602 #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */ 603 #define QUEUE_FLAG_DISCARD 8 /* supports DISCARD */ 604 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ 605 #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */ 606 #define QUEUE_FLAG_SECERASE 11 /* supports secure erase */ 607 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ 608 #define QUEUE_FLAG_DEAD 13 /* queue tear-down finished */ 609 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ 610 #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ 611 #define QUEUE_FLAG_WC 17 /* Write back caching */ 612 #define QUEUE_FLAG_FUA 18 /* device supports FUA writes */ 613 #define QUEUE_FLAG_DAX 19 /* device supports DAX */ 614 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ 615 #define QUEUE_FLAG_POLL_STATS 21 /* collecting stats for hybrid polling */ 616 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ 617 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23 /* queue supports SCSI commands */ 618 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ 619 #define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */ 620 #define QUEUE_FLAG_ZONE_RESETALL 26 /* supports Zone Reset All */ 621 #define QUEUE_FLAG_RQ_ALLOC_TIME 27 /* record rq->alloc_time_ns */ 622 623 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 624 (1 << QUEUE_FLAG_SAME_COMP)) 625 626 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 627 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 628 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); 629 630 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 631 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 632 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) 633 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 634 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 635 #define blk_queue_noxmerges(q) \ 636 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 637 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) 638 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) 639 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) 640 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) 641 #define blk_queue_zone_resetall(q) \ 642 test_bit(QUEUE_FLAG_ZONE_RESETALL, &(q)->queue_flags) 643 #define blk_queue_secure_erase(q) \ 644 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags)) 645 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) 646 #define blk_queue_scsi_passthrough(q) \ 647 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags) 648 #define blk_queue_pci_p2pdma(q) \ 649 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags) 650 #ifdef CONFIG_BLK_RQ_ALLOC_TIME 651 #define blk_queue_rq_alloc_time(q) \ 652 test_bit(QUEUE_FLAG_RQ_ALLOC_TIME, &(q)->queue_flags) 653 #else 654 #define blk_queue_rq_alloc_time(q) false 655 #endif 656 657 #define blk_noretry_request(rq) \ 658 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 659 REQ_FAILFAST_DRIVER)) 660 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 661 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 662 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags) 663 #define blk_queue_registered(q) test_bit(QUEUE_FLAG_REGISTERED, &(q)->queue_flags) 664 665 extern void blk_set_pm_only(struct request_queue *q); 666 extern void blk_clear_pm_only(struct request_queue *q); 667 668 static inline bool blk_account_rq(struct request *rq) 669 { 670 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq); 671 } 672 673 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 674 675 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ) 676 677 #define rq_dma_dir(rq) \ 678 (op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE) 679 680 #define dma_map_bvec(dev, bv, dir, attrs) \ 681 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 682 (dir), (attrs)) 683 684 static inline bool queue_is_mq(struct request_queue *q) 685 { 686 return q->mq_ops; 687 } 688 689 static inline enum blk_zoned_model 690 blk_queue_zoned_model(struct request_queue *q) 691 { 692 return q->limits.zoned; 693 } 694 695 static inline bool blk_queue_is_zoned(struct request_queue *q) 696 { 697 switch (blk_queue_zoned_model(q)) { 698 case BLK_ZONED_HA: 699 case BLK_ZONED_HM: 700 return true; 701 default: 702 return false; 703 } 704 } 705 706 static inline sector_t blk_queue_zone_sectors(struct request_queue *q) 707 { 708 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; 709 } 710 711 #ifdef CONFIG_BLK_DEV_ZONED 712 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 713 { 714 return blk_queue_is_zoned(q) ? q->nr_zones : 0; 715 } 716 717 static inline unsigned int blk_queue_zone_no(struct request_queue *q, 718 sector_t sector) 719 { 720 if (!blk_queue_is_zoned(q)) 721 return 0; 722 return sector >> ilog2(q->limits.chunk_sectors); 723 } 724 725 static inline bool blk_queue_zone_is_seq(struct request_queue *q, 726 sector_t sector) 727 { 728 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap) 729 return false; 730 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap); 731 } 732 #else /* CONFIG_BLK_DEV_ZONED */ 733 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 734 { 735 return 0; 736 } 737 #endif /* CONFIG_BLK_DEV_ZONED */ 738 739 static inline bool rq_is_sync(struct request *rq) 740 { 741 return op_is_sync(rq->cmd_flags); 742 } 743 744 static inline bool rq_mergeable(struct request *rq) 745 { 746 if (blk_rq_is_passthrough(rq)) 747 return false; 748 749 if (req_op(rq) == REQ_OP_FLUSH) 750 return false; 751 752 if (req_op(rq) == REQ_OP_WRITE_ZEROES) 753 return false; 754 755 if (rq->cmd_flags & REQ_NOMERGE_FLAGS) 756 return false; 757 if (rq->rq_flags & RQF_NOMERGE_FLAGS) 758 return false; 759 760 return true; 761 } 762 763 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) 764 { 765 if (bio_page(a) == bio_page(b) && 766 bio_offset(a) == bio_offset(b)) 767 return true; 768 769 return false; 770 } 771 772 static inline unsigned int blk_queue_depth(struct request_queue *q) 773 { 774 if (q->queue_depth) 775 return q->queue_depth; 776 777 return q->nr_requests; 778 } 779 780 extern unsigned long blk_max_low_pfn, blk_max_pfn; 781 782 /* 783 * standard bounce addresses: 784 * 785 * BLK_BOUNCE_HIGH : bounce all highmem pages 786 * BLK_BOUNCE_ANY : don't bounce anything 787 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary 788 */ 789 790 #if BITS_PER_LONG == 32 791 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) 792 #else 793 #define BLK_BOUNCE_HIGH -1ULL 794 #endif 795 #define BLK_BOUNCE_ANY (-1ULL) 796 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) 797 798 /* 799 * default timeout for SG_IO if none specified 800 */ 801 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 802 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 803 804 struct rq_map_data { 805 struct page **pages; 806 int page_order; 807 int nr_entries; 808 unsigned long offset; 809 int null_mapped; 810 int from_user; 811 }; 812 813 struct req_iterator { 814 struct bvec_iter iter; 815 struct bio *bio; 816 }; 817 818 /* This should not be used directly - use rq_for_each_segment */ 819 #define for_each_bio(_bio) \ 820 for (; _bio; _bio = _bio->bi_next) 821 #define __rq_for_each_bio(_bio, rq) \ 822 if ((rq->bio)) \ 823 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) 824 825 #define rq_for_each_segment(bvl, _rq, _iter) \ 826 __rq_for_each_bio(_iter.bio, _rq) \ 827 bio_for_each_segment(bvl, _iter.bio, _iter.iter) 828 829 #define rq_for_each_bvec(bvl, _rq, _iter) \ 830 __rq_for_each_bio(_iter.bio, _rq) \ 831 bio_for_each_bvec(bvl, _iter.bio, _iter.iter) 832 833 #define rq_iter_last(bvec, _iter) \ 834 (_iter.bio->bi_next == NULL && \ 835 bio_iter_last(bvec, _iter.iter)) 836 837 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 838 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 839 #endif 840 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 841 extern void rq_flush_dcache_pages(struct request *rq); 842 #else 843 static inline void rq_flush_dcache_pages(struct request *rq) 844 { 845 } 846 #endif 847 848 extern int blk_register_queue(struct gendisk *disk); 849 extern void blk_unregister_queue(struct gendisk *disk); 850 extern blk_qc_t generic_make_request(struct bio *bio); 851 extern blk_qc_t direct_make_request(struct bio *bio); 852 extern void blk_rq_init(struct request_queue *q, struct request *rq); 853 extern void blk_put_request(struct request *); 854 extern struct request *blk_get_request(struct request_queue *, unsigned int op, 855 blk_mq_req_flags_t flags); 856 extern int blk_lld_busy(struct request_queue *q); 857 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, 858 struct bio_set *bs, gfp_t gfp_mask, 859 int (*bio_ctr)(struct bio *, struct bio *, void *), 860 void *data); 861 extern void blk_rq_unprep_clone(struct request *rq); 862 extern blk_status_t blk_insert_cloned_request(struct request_queue *q, 863 struct request *rq); 864 extern int blk_rq_append_bio(struct request *rq, struct bio **bio); 865 extern void blk_queue_split(struct request_queue *, struct bio **); 866 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); 867 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, 868 unsigned int, void __user *); 869 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, 870 unsigned int, void __user *); 871 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, 872 struct scsi_ioctl_command __user *); 873 874 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 875 extern void blk_queue_exit(struct request_queue *q); 876 extern void blk_sync_queue(struct request_queue *q); 877 extern int blk_rq_map_user(struct request_queue *, struct request *, 878 struct rq_map_data *, void __user *, unsigned long, 879 gfp_t); 880 extern int blk_rq_unmap_user(struct bio *); 881 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); 882 extern int blk_rq_map_user_iov(struct request_queue *, struct request *, 883 struct rq_map_data *, const struct iov_iter *, 884 gfp_t); 885 extern void blk_execute_rq(struct request_queue *, struct gendisk *, 886 struct request *, int); 887 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, 888 struct request *, int, rq_end_io_fn *); 889 890 /* Helper to convert REQ_OP_XXX to its string format XXX */ 891 extern const char *blk_op_str(unsigned int op); 892 893 int blk_status_to_errno(blk_status_t status); 894 blk_status_t errno_to_blk_status(int errno); 895 896 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin); 897 898 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 899 { 900 return bdev->bd_disk->queue; /* this is never NULL */ 901 } 902 903 /* 904 * The basic unit of block I/O is a sector. It is used in a number of contexts 905 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9 906 * bytes. Variables of type sector_t represent an offset or size that is a 907 * multiple of 512 bytes. Hence these two constants. 908 */ 909 #ifndef SECTOR_SHIFT 910 #define SECTOR_SHIFT 9 911 #endif 912 #ifndef SECTOR_SIZE 913 #define SECTOR_SIZE (1 << SECTOR_SHIFT) 914 #endif 915 916 /* 917 * blk_rq_pos() : the current sector 918 * blk_rq_bytes() : bytes left in the entire request 919 * blk_rq_cur_bytes() : bytes left in the current segment 920 * blk_rq_err_bytes() : bytes left till the next error boundary 921 * blk_rq_sectors() : sectors left in the entire request 922 * blk_rq_cur_sectors() : sectors left in the current segment 923 * blk_rq_stats_sectors() : sectors of the entire request used for stats 924 */ 925 static inline sector_t blk_rq_pos(const struct request *rq) 926 { 927 return rq->__sector; 928 } 929 930 static inline unsigned int blk_rq_bytes(const struct request *rq) 931 { 932 return rq->__data_len; 933 } 934 935 static inline int blk_rq_cur_bytes(const struct request *rq) 936 { 937 return rq->bio ? bio_cur_bytes(rq->bio) : 0; 938 } 939 940 extern unsigned int blk_rq_err_bytes(const struct request *rq); 941 942 static inline unsigned int blk_rq_sectors(const struct request *rq) 943 { 944 return blk_rq_bytes(rq) >> SECTOR_SHIFT; 945 } 946 947 static inline unsigned int blk_rq_cur_sectors(const struct request *rq) 948 { 949 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT; 950 } 951 952 static inline unsigned int blk_rq_stats_sectors(const struct request *rq) 953 { 954 return rq->stats_sectors; 955 } 956 957 #ifdef CONFIG_BLK_DEV_ZONED 958 static inline unsigned int blk_rq_zone_no(struct request *rq) 959 { 960 return blk_queue_zone_no(rq->q, blk_rq_pos(rq)); 961 } 962 963 static inline unsigned int blk_rq_zone_is_seq(struct request *rq) 964 { 965 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq)); 966 } 967 #endif /* CONFIG_BLK_DEV_ZONED */ 968 969 /* 970 * Some commands like WRITE SAME have a payload or data transfer size which 971 * is different from the size of the request. Any driver that supports such 972 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to 973 * calculate the data transfer size. 974 */ 975 static inline unsigned int blk_rq_payload_bytes(struct request *rq) 976 { 977 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 978 return rq->special_vec.bv_len; 979 return blk_rq_bytes(rq); 980 } 981 982 /* 983 * Return the first full biovec in the request. The caller needs to check that 984 * there are any bvecs before calling this helper. 985 */ 986 static inline struct bio_vec req_bvec(struct request *rq) 987 { 988 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 989 return rq->special_vec; 990 return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter); 991 } 992 993 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, 994 int op) 995 { 996 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) 997 return min(q->limits.max_discard_sectors, 998 UINT_MAX >> SECTOR_SHIFT); 999 1000 if (unlikely(op == REQ_OP_WRITE_SAME)) 1001 return q->limits.max_write_same_sectors; 1002 1003 if (unlikely(op == REQ_OP_WRITE_ZEROES)) 1004 return q->limits.max_write_zeroes_sectors; 1005 1006 return q->limits.max_sectors; 1007 } 1008 1009 /* 1010 * Return maximum size of a request at given offset. Only valid for 1011 * file system requests. 1012 */ 1013 static inline unsigned int blk_max_size_offset(struct request_queue *q, 1014 sector_t offset) 1015 { 1016 if (!q->limits.chunk_sectors) 1017 return q->limits.max_sectors; 1018 1019 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors - 1020 (offset & (q->limits.chunk_sectors - 1)))); 1021 } 1022 1023 static inline unsigned int blk_rq_get_max_sectors(struct request *rq, 1024 sector_t offset) 1025 { 1026 struct request_queue *q = rq->q; 1027 1028 if (blk_rq_is_passthrough(rq)) 1029 return q->limits.max_hw_sectors; 1030 1031 if (!q->limits.chunk_sectors || 1032 req_op(rq) == REQ_OP_DISCARD || 1033 req_op(rq) == REQ_OP_SECURE_ERASE) 1034 return blk_queue_get_max_sectors(q, req_op(rq)); 1035 1036 return min(blk_max_size_offset(q, offset), 1037 blk_queue_get_max_sectors(q, req_op(rq))); 1038 } 1039 1040 static inline unsigned int blk_rq_count_bios(struct request *rq) 1041 { 1042 unsigned int nr_bios = 0; 1043 struct bio *bio; 1044 1045 __rq_for_each_bio(bio, rq) 1046 nr_bios++; 1047 1048 return nr_bios; 1049 } 1050 1051 void blk_steal_bios(struct bio_list *list, struct request *rq); 1052 1053 /* 1054 * Request completion related functions. 1055 * 1056 * blk_update_request() completes given number of bytes and updates 1057 * the request without completing it. 1058 */ 1059 extern bool blk_update_request(struct request *rq, blk_status_t error, 1060 unsigned int nr_bytes); 1061 1062 extern void __blk_complete_request(struct request *); 1063 extern void blk_abort_request(struct request *); 1064 1065 /* 1066 * Access functions for manipulating queue properties 1067 */ 1068 extern void blk_cleanup_queue(struct request_queue *); 1069 extern void blk_queue_make_request(struct request_queue *, make_request_fn *); 1070 extern void blk_queue_bounce_limit(struct request_queue *, u64); 1071 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); 1072 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); 1073 extern void blk_queue_max_segments(struct request_queue *, unsigned short); 1074 extern void blk_queue_max_discard_segments(struct request_queue *, 1075 unsigned short); 1076 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); 1077 extern void blk_queue_max_discard_sectors(struct request_queue *q, 1078 unsigned int max_discard_sectors); 1079 extern void blk_queue_max_write_same_sectors(struct request_queue *q, 1080 unsigned int max_write_same_sectors); 1081 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, 1082 unsigned int max_write_same_sectors); 1083 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); 1084 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); 1085 extern void blk_queue_alignment_offset(struct request_queue *q, 1086 unsigned int alignment); 1087 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 1088 extern void blk_queue_io_min(struct request_queue *q, unsigned int min); 1089 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 1090 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); 1091 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1092 extern void blk_set_default_limits(struct queue_limits *lim); 1093 extern void blk_set_stacking_limits(struct queue_limits *lim); 1094 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1095 sector_t offset); 1096 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, 1097 sector_t offset); 1098 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, 1099 sector_t offset); 1100 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); 1101 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 1102 extern int blk_queue_dma_drain(struct request_queue *q, 1103 dma_drain_needed_fn *dma_drain_needed, 1104 void *buf, unsigned int size); 1105 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); 1106 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long); 1107 extern void blk_queue_dma_alignment(struct request_queue *, int); 1108 extern void blk_queue_update_dma_alignment(struct request_queue *, int); 1109 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1110 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); 1111 extern void blk_queue_required_elevator_features(struct request_queue *q, 1112 unsigned int features); 1113 1114 /* 1115 * Number of physical segments as sent to the device. 1116 * 1117 * Normally this is the number of discontiguous data segments sent by the 1118 * submitter. But for data-less command like discard we might have no 1119 * actual data segments submitted, but the driver might have to add it's 1120 * own special payload. In that case we still return 1 here so that this 1121 * special payload will be mapped. 1122 */ 1123 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq) 1124 { 1125 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 1126 return 1; 1127 return rq->nr_phys_segments; 1128 } 1129 1130 /* 1131 * Number of discard segments (or ranges) the driver needs to fill in. 1132 * Each discard bio merged into a request is counted as one segment. 1133 */ 1134 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq) 1135 { 1136 return max_t(unsigned short, rq->nr_phys_segments, 1); 1137 } 1138 1139 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); 1140 extern void blk_dump_rq_flags(struct request *, char *); 1141 extern long nr_blockdev_pages(void); 1142 1143 bool __must_check blk_get_queue(struct request_queue *); 1144 struct request_queue *blk_alloc_queue(gfp_t); 1145 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id); 1146 extern void blk_put_queue(struct request_queue *); 1147 extern void blk_set_queue_dying(struct request_queue *); 1148 1149 /* 1150 * blk_plug permits building a queue of related requests by holding the I/O 1151 * fragments for a short period. This allows merging of sequential requests 1152 * into single larger request. As the requests are moved from a per-task list to 1153 * the device's request_queue in a batch, this results in improved scalability 1154 * as the lock contention for request_queue lock is reduced. 1155 * 1156 * It is ok not to disable preemption when adding the request to the plug list 1157 * or when attempting a merge, because blk_schedule_flush_list() will only flush 1158 * the plug list when the task sleeps by itself. For details, please see 1159 * schedule() where blk_schedule_flush_plug() is called. 1160 */ 1161 struct blk_plug { 1162 struct list_head mq_list; /* blk-mq requests */ 1163 struct list_head cb_list; /* md requires an unplug callback */ 1164 unsigned short rq_count; 1165 bool multiple_queues; 1166 }; 1167 #define BLK_MAX_REQUEST_COUNT 16 1168 #define BLK_PLUG_FLUSH_SIZE (128 * 1024) 1169 1170 struct blk_plug_cb; 1171 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1172 struct blk_plug_cb { 1173 struct list_head list; 1174 blk_plug_cb_fn callback; 1175 void *data; 1176 }; 1177 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1178 void *data, int size); 1179 extern void blk_start_plug(struct blk_plug *); 1180 extern void blk_finish_plug(struct blk_plug *); 1181 extern void blk_flush_plug_list(struct blk_plug *, bool); 1182 1183 static inline void blk_flush_plug(struct task_struct *tsk) 1184 { 1185 struct blk_plug *plug = tsk->plug; 1186 1187 if (plug) 1188 blk_flush_plug_list(plug, false); 1189 } 1190 1191 static inline void blk_schedule_flush_plug(struct task_struct *tsk) 1192 { 1193 struct blk_plug *plug = tsk->plug; 1194 1195 if (plug) 1196 blk_flush_plug_list(plug, true); 1197 } 1198 1199 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1200 { 1201 struct blk_plug *plug = tsk->plug; 1202 1203 return plug && 1204 (!list_empty(&plug->mq_list) || 1205 !list_empty(&plug->cb_list)); 1206 } 1207 1208 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); 1209 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, 1210 sector_t nr_sects, gfp_t gfp_mask, struct page *page); 1211 1212 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */ 1213 1214 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1215 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); 1216 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1217 sector_t nr_sects, gfp_t gfp_mask, int flags, 1218 struct bio **biop); 1219 1220 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1221 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1222 1223 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1224 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1225 unsigned flags); 1226 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1227 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1228 1229 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1230 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1231 { 1232 return blkdev_issue_discard(sb->s_bdev, 1233 block << (sb->s_blocksize_bits - 1234 SECTOR_SHIFT), 1235 nr_blocks << (sb->s_blocksize_bits - 1236 SECTOR_SHIFT), 1237 gfp_mask, flags); 1238 } 1239 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1240 sector_t nr_blocks, gfp_t gfp_mask) 1241 { 1242 return blkdev_issue_zeroout(sb->s_bdev, 1243 block << (sb->s_blocksize_bits - 1244 SECTOR_SHIFT), 1245 nr_blocks << (sb->s_blocksize_bits - 1246 SECTOR_SHIFT), 1247 gfp_mask, 0); 1248 } 1249 1250 extern int blk_verify_command(unsigned char *cmd, fmode_t mode); 1251 1252 enum blk_default_limits { 1253 BLK_MAX_SEGMENTS = 128, 1254 BLK_SAFE_MAX_SECTORS = 255, 1255 BLK_DEF_MAX_SECTORS = 2560, 1256 BLK_MAX_SEGMENT_SIZE = 65536, 1257 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1258 }; 1259 1260 static inline unsigned long queue_segment_boundary(const struct request_queue *q) 1261 { 1262 return q->limits.seg_boundary_mask; 1263 } 1264 1265 static inline unsigned long queue_virt_boundary(const struct request_queue *q) 1266 { 1267 return q->limits.virt_boundary_mask; 1268 } 1269 1270 static inline unsigned int queue_max_sectors(const struct request_queue *q) 1271 { 1272 return q->limits.max_sectors; 1273 } 1274 1275 static inline unsigned int queue_max_hw_sectors(const struct request_queue *q) 1276 { 1277 return q->limits.max_hw_sectors; 1278 } 1279 1280 static inline unsigned short queue_max_segments(const struct request_queue *q) 1281 { 1282 return q->limits.max_segments; 1283 } 1284 1285 static inline unsigned short queue_max_discard_segments(const struct request_queue *q) 1286 { 1287 return q->limits.max_discard_segments; 1288 } 1289 1290 static inline unsigned int queue_max_segment_size(const struct request_queue *q) 1291 { 1292 return q->limits.max_segment_size; 1293 } 1294 1295 static inline unsigned short queue_logical_block_size(const struct request_queue *q) 1296 { 1297 int retval = 512; 1298 1299 if (q && q->limits.logical_block_size) 1300 retval = q->limits.logical_block_size; 1301 1302 return retval; 1303 } 1304 1305 static inline unsigned short bdev_logical_block_size(struct block_device *bdev) 1306 { 1307 return queue_logical_block_size(bdev_get_queue(bdev)); 1308 } 1309 1310 static inline unsigned int queue_physical_block_size(const struct request_queue *q) 1311 { 1312 return q->limits.physical_block_size; 1313 } 1314 1315 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1316 { 1317 return queue_physical_block_size(bdev_get_queue(bdev)); 1318 } 1319 1320 static inline unsigned int queue_io_min(const struct request_queue *q) 1321 { 1322 return q->limits.io_min; 1323 } 1324 1325 static inline int bdev_io_min(struct block_device *bdev) 1326 { 1327 return queue_io_min(bdev_get_queue(bdev)); 1328 } 1329 1330 static inline unsigned int queue_io_opt(const struct request_queue *q) 1331 { 1332 return q->limits.io_opt; 1333 } 1334 1335 static inline int bdev_io_opt(struct block_device *bdev) 1336 { 1337 return queue_io_opt(bdev_get_queue(bdev)); 1338 } 1339 1340 static inline int queue_alignment_offset(const struct request_queue *q) 1341 { 1342 if (q->limits.misaligned) 1343 return -1; 1344 1345 return q->limits.alignment_offset; 1346 } 1347 1348 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) 1349 { 1350 unsigned int granularity = max(lim->physical_block_size, lim->io_min); 1351 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) 1352 << SECTOR_SHIFT; 1353 1354 return (granularity + lim->alignment_offset - alignment) % granularity; 1355 } 1356 1357 static inline int bdev_alignment_offset(struct block_device *bdev) 1358 { 1359 struct request_queue *q = bdev_get_queue(bdev); 1360 1361 if (q->limits.misaligned) 1362 return -1; 1363 1364 if (bdev != bdev->bd_contains) 1365 return bdev->bd_part->alignment_offset; 1366 1367 return q->limits.alignment_offset; 1368 } 1369 1370 static inline int queue_discard_alignment(const struct request_queue *q) 1371 { 1372 if (q->limits.discard_misaligned) 1373 return -1; 1374 1375 return q->limits.discard_alignment; 1376 } 1377 1378 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) 1379 { 1380 unsigned int alignment, granularity, offset; 1381 1382 if (!lim->max_discard_sectors) 1383 return 0; 1384 1385 /* Why are these in bytes, not sectors? */ 1386 alignment = lim->discard_alignment >> SECTOR_SHIFT; 1387 granularity = lim->discard_granularity >> SECTOR_SHIFT; 1388 if (!granularity) 1389 return 0; 1390 1391 /* Offset of the partition start in 'granularity' sectors */ 1392 offset = sector_div(sector, granularity); 1393 1394 /* And why do we do this modulus *again* in blkdev_issue_discard()? */ 1395 offset = (granularity + alignment - offset) % granularity; 1396 1397 /* Turn it back into bytes, gaah */ 1398 return offset << SECTOR_SHIFT; 1399 } 1400 1401 static inline int bdev_discard_alignment(struct block_device *bdev) 1402 { 1403 struct request_queue *q = bdev_get_queue(bdev); 1404 1405 if (bdev != bdev->bd_contains) 1406 return bdev->bd_part->discard_alignment; 1407 1408 return q->limits.discard_alignment; 1409 } 1410 1411 static inline unsigned int bdev_write_same(struct block_device *bdev) 1412 { 1413 struct request_queue *q = bdev_get_queue(bdev); 1414 1415 if (q) 1416 return q->limits.max_write_same_sectors; 1417 1418 return 0; 1419 } 1420 1421 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1422 { 1423 struct request_queue *q = bdev_get_queue(bdev); 1424 1425 if (q) 1426 return q->limits.max_write_zeroes_sectors; 1427 1428 return 0; 1429 } 1430 1431 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev) 1432 { 1433 struct request_queue *q = bdev_get_queue(bdev); 1434 1435 if (q) 1436 return blk_queue_zoned_model(q); 1437 1438 return BLK_ZONED_NONE; 1439 } 1440 1441 static inline bool bdev_is_zoned(struct block_device *bdev) 1442 { 1443 struct request_queue *q = bdev_get_queue(bdev); 1444 1445 if (q) 1446 return blk_queue_is_zoned(q); 1447 1448 return false; 1449 } 1450 1451 static inline sector_t bdev_zone_sectors(struct block_device *bdev) 1452 { 1453 struct request_queue *q = bdev_get_queue(bdev); 1454 1455 if (q) 1456 return blk_queue_zone_sectors(q); 1457 return 0; 1458 } 1459 1460 static inline int queue_dma_alignment(const struct request_queue *q) 1461 { 1462 return q ? q->dma_alignment : 511; 1463 } 1464 1465 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1466 unsigned int len) 1467 { 1468 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1469 return !(addr & alignment) && !(len & alignment); 1470 } 1471 1472 /* assumes size > 256 */ 1473 static inline unsigned int blksize_bits(unsigned int size) 1474 { 1475 unsigned int bits = 8; 1476 do { 1477 bits++; 1478 size >>= 1; 1479 } while (size > 256); 1480 return bits; 1481 } 1482 1483 static inline unsigned int block_size(struct block_device *bdev) 1484 { 1485 return bdev->bd_block_size; 1486 } 1487 1488 typedef struct {struct page *v;} Sector; 1489 1490 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); 1491 1492 static inline void put_dev_sector(Sector p) 1493 { 1494 put_page(p.v); 1495 } 1496 1497 int kblockd_schedule_work(struct work_struct *work); 1498 int kblockd_schedule_work_on(int cpu, struct work_struct *work); 1499 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1500 1501 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1502 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1503 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1504 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1505 1506 #if defined(CONFIG_BLK_DEV_INTEGRITY) 1507 1508 enum blk_integrity_flags { 1509 BLK_INTEGRITY_VERIFY = 1 << 0, 1510 BLK_INTEGRITY_GENERATE = 1 << 1, 1511 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, 1512 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, 1513 }; 1514 1515 struct blk_integrity_iter { 1516 void *prot_buf; 1517 void *data_buf; 1518 sector_t seed; 1519 unsigned int data_size; 1520 unsigned short interval; 1521 const char *disk_name; 1522 }; 1523 1524 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *); 1525 1526 struct blk_integrity_profile { 1527 integrity_processing_fn *generate_fn; 1528 integrity_processing_fn *verify_fn; 1529 const char *name; 1530 }; 1531 1532 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *); 1533 extern void blk_integrity_unregister(struct gendisk *); 1534 extern int blk_integrity_compare(struct gendisk *, struct gendisk *); 1535 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, 1536 struct scatterlist *); 1537 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); 1538 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, 1539 struct request *); 1540 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, 1541 struct bio *); 1542 1543 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1544 { 1545 struct blk_integrity *bi = &disk->queue->integrity; 1546 1547 if (!bi->profile) 1548 return NULL; 1549 1550 return bi; 1551 } 1552 1553 static inline 1554 struct blk_integrity *bdev_get_integrity(struct block_device *bdev) 1555 { 1556 return blk_get_integrity(bdev->bd_disk); 1557 } 1558 1559 static inline bool blk_integrity_rq(struct request *rq) 1560 { 1561 return rq->cmd_flags & REQ_INTEGRITY; 1562 } 1563 1564 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1565 unsigned int segs) 1566 { 1567 q->limits.max_integrity_segments = segs; 1568 } 1569 1570 static inline unsigned short 1571 queue_max_integrity_segments(const struct request_queue *q) 1572 { 1573 return q->limits.max_integrity_segments; 1574 } 1575 1576 /** 1577 * bio_integrity_intervals - Return number of integrity intervals for a bio 1578 * @bi: blk_integrity profile for device 1579 * @sectors: Size of the bio in 512-byte sectors 1580 * 1581 * Description: The block layer calculates everything in 512 byte 1582 * sectors but integrity metadata is done in terms of the data integrity 1583 * interval size of the storage device. Convert the block layer sectors 1584 * to the appropriate number of integrity intervals. 1585 */ 1586 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1587 unsigned int sectors) 1588 { 1589 return sectors >> (bi->interval_exp - 9); 1590 } 1591 1592 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1593 unsigned int sectors) 1594 { 1595 return bio_integrity_intervals(bi, sectors) * bi->tuple_size; 1596 } 1597 1598 /* 1599 * Return the first bvec that contains integrity data. Only drivers that are 1600 * limited to a single integrity segment should use this helper. 1601 */ 1602 static inline struct bio_vec *rq_integrity_vec(struct request *rq) 1603 { 1604 if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1)) 1605 return NULL; 1606 return rq->bio->bi_integrity->bip_vec; 1607 } 1608 1609 #else /* CONFIG_BLK_DEV_INTEGRITY */ 1610 1611 struct bio; 1612 struct block_device; 1613 struct gendisk; 1614 struct blk_integrity; 1615 1616 static inline int blk_integrity_rq(struct request *rq) 1617 { 1618 return 0; 1619 } 1620 static inline int blk_rq_count_integrity_sg(struct request_queue *q, 1621 struct bio *b) 1622 { 1623 return 0; 1624 } 1625 static inline int blk_rq_map_integrity_sg(struct request_queue *q, 1626 struct bio *b, 1627 struct scatterlist *s) 1628 { 1629 return 0; 1630 } 1631 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) 1632 { 1633 return NULL; 1634 } 1635 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1636 { 1637 return NULL; 1638 } 1639 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) 1640 { 1641 return 0; 1642 } 1643 static inline void blk_integrity_register(struct gendisk *d, 1644 struct blk_integrity *b) 1645 { 1646 } 1647 static inline void blk_integrity_unregister(struct gendisk *d) 1648 { 1649 } 1650 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1651 unsigned int segs) 1652 { 1653 } 1654 static inline unsigned short queue_max_integrity_segments(const struct request_queue *q) 1655 { 1656 return 0; 1657 } 1658 static inline bool blk_integrity_merge_rq(struct request_queue *rq, 1659 struct request *r1, 1660 struct request *r2) 1661 { 1662 return true; 1663 } 1664 static inline bool blk_integrity_merge_bio(struct request_queue *rq, 1665 struct request *r, 1666 struct bio *b) 1667 { 1668 return true; 1669 } 1670 1671 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1672 unsigned int sectors) 1673 { 1674 return 0; 1675 } 1676 1677 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1678 unsigned int sectors) 1679 { 1680 return 0; 1681 } 1682 1683 static inline struct bio_vec *rq_integrity_vec(struct request *rq) 1684 { 1685 return NULL; 1686 } 1687 1688 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 1689 1690 struct block_device_operations { 1691 int (*open) (struct block_device *, fmode_t); 1692 void (*release) (struct gendisk *, fmode_t); 1693 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int); 1694 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1695 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1696 unsigned int (*check_events) (struct gendisk *disk, 1697 unsigned int clearing); 1698 /* ->media_changed() is DEPRECATED, use ->check_events() instead */ 1699 int (*media_changed) (struct gendisk *); 1700 void (*unlock_native_capacity) (struct gendisk *); 1701 int (*revalidate_disk) (struct gendisk *); 1702 int (*getgeo)(struct block_device *, struct hd_geometry *); 1703 /* this callback is with swap_lock and sometimes page table lock held */ 1704 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1705 int (*report_zones)(struct gendisk *, sector_t sector, 1706 struct blk_zone *zones, unsigned int *nr_zones); 1707 struct module *owner; 1708 const struct pr_ops *pr_ops; 1709 }; 1710 1711 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, 1712 unsigned long); 1713 extern int bdev_read_page(struct block_device *, sector_t, struct page *); 1714 extern int bdev_write_page(struct block_device *, sector_t, struct page *, 1715 struct writeback_control *); 1716 1717 #ifdef CONFIG_BLK_DEV_ZONED 1718 bool blk_req_needs_zone_write_lock(struct request *rq); 1719 void __blk_req_zone_write_lock(struct request *rq); 1720 void __blk_req_zone_write_unlock(struct request *rq); 1721 1722 static inline void blk_req_zone_write_lock(struct request *rq) 1723 { 1724 if (blk_req_needs_zone_write_lock(rq)) 1725 __blk_req_zone_write_lock(rq); 1726 } 1727 1728 static inline void blk_req_zone_write_unlock(struct request *rq) 1729 { 1730 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED) 1731 __blk_req_zone_write_unlock(rq); 1732 } 1733 1734 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1735 { 1736 return rq->q->seq_zones_wlock && 1737 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock); 1738 } 1739 1740 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1741 { 1742 if (!blk_req_needs_zone_write_lock(rq)) 1743 return true; 1744 return !blk_req_zone_is_write_locked(rq); 1745 } 1746 #else 1747 static inline bool blk_req_needs_zone_write_lock(struct request *rq) 1748 { 1749 return false; 1750 } 1751 1752 static inline void blk_req_zone_write_lock(struct request *rq) 1753 { 1754 } 1755 1756 static inline void blk_req_zone_write_unlock(struct request *rq) 1757 { 1758 } 1759 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1760 { 1761 return false; 1762 } 1763 1764 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1765 { 1766 return true; 1767 } 1768 #endif /* CONFIG_BLK_DEV_ZONED */ 1769 1770 #else /* CONFIG_BLOCK */ 1771 1772 struct block_device; 1773 1774 /* 1775 * stubs for when the block layer is configured out 1776 */ 1777 #define buffer_heads_over_limit 0 1778 1779 static inline long nr_blockdev_pages(void) 1780 { 1781 return 0; 1782 } 1783 1784 struct blk_plug { 1785 }; 1786 1787 static inline void blk_start_plug(struct blk_plug *plug) 1788 { 1789 } 1790 1791 static inline void blk_finish_plug(struct blk_plug *plug) 1792 { 1793 } 1794 1795 static inline void blk_flush_plug(struct task_struct *task) 1796 { 1797 } 1798 1799 static inline void blk_schedule_flush_plug(struct task_struct *task) 1800 { 1801 } 1802 1803 1804 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1805 { 1806 return false; 1807 } 1808 1809 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, 1810 sector_t *error_sector) 1811 { 1812 return 0; 1813 } 1814 1815 #endif /* CONFIG_BLOCK */ 1816 1817 static inline void blk_wake_io_task(struct task_struct *waiter) 1818 { 1819 /* 1820 * If we're polling, the task itself is doing the completions. For 1821 * that case, we don't need to signal a wakeup, it's enough to just 1822 * mark us as RUNNING. 1823 */ 1824 if (waiter == current) 1825 __set_current_state(TASK_RUNNING); 1826 else 1827 wake_up_process(waiter); 1828 } 1829 1830 #endif 1831