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 /* 54 * Maximum number of blkcg policies allowed to be registered concurrently. 55 * Defined here to simplify include dependency. 56 */ 57 #define BLKCG_MAX_POLS 5 58 59 typedef void (rq_end_io_fn)(struct request *, blk_status_t); 60 61 /* 62 * request flags */ 63 typedef __u32 __bitwise req_flags_t; 64 65 /* elevator knows about this request */ 66 #define RQF_SORTED ((__force req_flags_t)(1 << 0)) 67 /* drive already may have started this one */ 68 #define RQF_STARTED ((__force req_flags_t)(1 << 1)) 69 /* may not be passed by ioscheduler */ 70 #define RQF_SOFTBARRIER ((__force req_flags_t)(1 << 3)) 71 /* request for flush sequence */ 72 #define RQF_FLUSH_SEQ ((__force req_flags_t)(1 << 4)) 73 /* merge of different types, fail separately */ 74 #define RQF_MIXED_MERGE ((__force req_flags_t)(1 << 5)) 75 /* track inflight for MQ */ 76 #define RQF_MQ_INFLIGHT ((__force req_flags_t)(1 << 6)) 77 /* don't call prep for this one */ 78 #define RQF_DONTPREP ((__force req_flags_t)(1 << 7)) 79 /* set for "ide_preempt" requests and also for requests for which the SCSI 80 "quiesce" state must be ignored. */ 81 #define RQF_PREEMPT ((__force req_flags_t)(1 << 8)) 82 /* contains copies of user pages */ 83 #define RQF_COPY_USER ((__force req_flags_t)(1 << 9)) 84 /* vaguely specified driver internal error. Ignored by the block layer */ 85 #define RQF_FAILED ((__force req_flags_t)(1 << 10)) 86 /* don't warn about errors */ 87 #define RQF_QUIET ((__force req_flags_t)(1 << 11)) 88 /* elevator private data attached */ 89 #define RQF_ELVPRIV ((__force req_flags_t)(1 << 12)) 90 /* account into disk and partition IO statistics */ 91 #define RQF_IO_STAT ((__force req_flags_t)(1 << 13)) 92 /* request came from our alloc pool */ 93 #define RQF_ALLOCED ((__force req_flags_t)(1 << 14)) 94 /* runtime pm request */ 95 #define RQF_PM ((__force req_flags_t)(1 << 15)) 96 /* on IO scheduler merge hash */ 97 #define RQF_HASHED ((__force req_flags_t)(1 << 16)) 98 /* track IO completion time */ 99 #define RQF_STATS ((__force req_flags_t)(1 << 17)) 100 /* Look at ->special_vec for the actual data payload instead of the 101 bio chain. */ 102 #define RQF_SPECIAL_PAYLOAD ((__force req_flags_t)(1 << 18)) 103 /* The per-zone write lock is held for this request */ 104 #define RQF_ZONE_WRITE_LOCKED ((__force req_flags_t)(1 << 19)) 105 /* already slept for hybrid poll */ 106 #define RQF_MQ_POLL_SLEPT ((__force req_flags_t)(1 << 20)) 107 /* ->timeout has been called, don't expire again */ 108 #define RQF_TIMED_OUT ((__force req_flags_t)(1 << 21)) 109 110 /* flags that prevent us from merging requests: */ 111 #define RQF_NOMERGE_FLAGS \ 112 (RQF_STARTED | RQF_SOFTBARRIER | RQF_FLUSH_SEQ | RQF_SPECIAL_PAYLOAD) 113 114 /* 115 * Request state for blk-mq. 116 */ 117 enum mq_rq_state { 118 MQ_RQ_IDLE = 0, 119 MQ_RQ_IN_FLIGHT = 1, 120 MQ_RQ_COMPLETE = 2, 121 }; 122 123 /* 124 * Try to put the fields that are referenced together in the same cacheline. 125 * 126 * If you modify this structure, make sure to update blk_rq_init() and 127 * especially blk_mq_rq_ctx_init() to take care of the added fields. 128 */ 129 struct request { 130 struct request_queue *q; 131 struct blk_mq_ctx *mq_ctx; 132 struct blk_mq_hw_ctx *mq_hctx; 133 134 unsigned int cmd_flags; /* op and common flags */ 135 req_flags_t rq_flags; 136 137 int internal_tag; 138 139 /* the following two fields are internal, NEVER access directly */ 140 unsigned int __data_len; /* total data len */ 141 int tag; 142 sector_t __sector; /* sector cursor */ 143 144 struct bio *bio; 145 struct bio *biotail; 146 147 struct list_head queuelist; 148 149 /* 150 * The hash is used inside the scheduler, and killed once the 151 * request reaches the dispatch list. The ipi_list is only used 152 * to queue the request for softirq completion, which is long 153 * after the request has been unhashed (and even removed from 154 * the dispatch list). 155 */ 156 union { 157 struct hlist_node hash; /* merge hash */ 158 struct list_head ipi_list; 159 }; 160 161 /* 162 * The rb_node is only used inside the io scheduler, requests 163 * are pruned when moved to the dispatch queue. So let the 164 * completion_data share space with the rb_node. 165 */ 166 union { 167 struct rb_node rb_node; /* sort/lookup */ 168 struct bio_vec special_vec; 169 void *completion_data; 170 int error_count; /* for legacy drivers, don't use */ 171 }; 172 173 /* 174 * Three pointers are available for the IO schedulers, if they need 175 * more they have to dynamically allocate it. Flush requests are 176 * never put on the IO scheduler. So let the flush fields share 177 * space with the elevator data. 178 */ 179 union { 180 struct { 181 struct io_cq *icq; 182 void *priv[2]; 183 } elv; 184 185 struct { 186 unsigned int seq; 187 struct list_head list; 188 rq_end_io_fn *saved_end_io; 189 } flush; 190 }; 191 192 struct gendisk *rq_disk; 193 struct hd_struct *part; 194 /* Time that I/O was submitted to the kernel. */ 195 u64 start_time_ns; 196 /* Time that I/O was submitted to the device. */ 197 u64 io_start_time_ns; 198 199 #ifdef CONFIG_BLK_WBT 200 unsigned short wbt_flags; 201 #endif 202 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW 203 unsigned short throtl_size; 204 #endif 205 206 /* 207 * Number of scatter-gather DMA addr+len pairs after 208 * physical address coalescing is performed. 209 */ 210 unsigned short nr_phys_segments; 211 212 #if defined(CONFIG_BLK_DEV_INTEGRITY) 213 unsigned short nr_integrity_segments; 214 #endif 215 216 unsigned short write_hint; 217 unsigned short ioprio; 218 219 void *special; /* opaque pointer available for LLD use */ 220 221 unsigned int extra_len; /* length of alignment and padding */ 222 223 enum mq_rq_state state; 224 refcount_t ref; 225 226 unsigned int timeout; 227 unsigned long deadline; 228 229 union { 230 struct __call_single_data csd; 231 u64 fifo_time; 232 }; 233 234 /* 235 * completion callback. 236 */ 237 rq_end_io_fn *end_io; 238 void *end_io_data; 239 240 /* for bidi */ 241 struct request *next_rq; 242 }; 243 244 static inline bool blk_op_is_scsi(unsigned int op) 245 { 246 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT; 247 } 248 249 static inline bool blk_op_is_private(unsigned int op) 250 { 251 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 252 } 253 254 static inline bool blk_rq_is_scsi(struct request *rq) 255 { 256 return blk_op_is_scsi(req_op(rq)); 257 } 258 259 static inline bool blk_rq_is_private(struct request *rq) 260 { 261 return blk_op_is_private(req_op(rq)); 262 } 263 264 static inline bool blk_rq_is_passthrough(struct request *rq) 265 { 266 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq); 267 } 268 269 static inline bool bio_is_passthrough(struct bio *bio) 270 { 271 unsigned op = bio_op(bio); 272 273 return blk_op_is_scsi(op) || blk_op_is_private(op); 274 } 275 276 static inline unsigned short req_get_ioprio(struct request *req) 277 { 278 return req->ioprio; 279 } 280 281 #include <linux/elevator.h> 282 283 struct blk_queue_ctx; 284 285 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio); 286 287 struct bio_vec; 288 typedef int (dma_drain_needed_fn)(struct request *); 289 290 enum blk_eh_timer_return { 291 BLK_EH_DONE, /* drivers has completed the command */ 292 BLK_EH_RESET_TIMER, /* reset timer and try again */ 293 }; 294 295 enum blk_queue_state { 296 Queue_down, 297 Queue_up, 298 }; 299 300 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ 301 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ 302 303 #define BLK_SCSI_MAX_CMDS (256) 304 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) 305 306 /* 307 * Zoned block device models (zoned limit). 308 */ 309 enum blk_zoned_model { 310 BLK_ZONED_NONE, /* Regular block device */ 311 BLK_ZONED_HA, /* Host-aware zoned block device */ 312 BLK_ZONED_HM, /* Host-managed zoned block device */ 313 }; 314 315 struct queue_limits { 316 unsigned long bounce_pfn; 317 unsigned long seg_boundary_mask; 318 unsigned long virt_boundary_mask; 319 320 unsigned int max_hw_sectors; 321 unsigned int max_dev_sectors; 322 unsigned int chunk_sectors; 323 unsigned int max_sectors; 324 unsigned int max_segment_size; 325 unsigned int physical_block_size; 326 unsigned int alignment_offset; 327 unsigned int io_min; 328 unsigned int io_opt; 329 unsigned int max_discard_sectors; 330 unsigned int max_hw_discard_sectors; 331 unsigned int max_write_same_sectors; 332 unsigned int max_write_zeroes_sectors; 333 unsigned int discard_granularity; 334 unsigned int discard_alignment; 335 336 unsigned short logical_block_size; 337 unsigned short max_segments; 338 unsigned short max_integrity_segments; 339 unsigned short max_discard_segments; 340 341 unsigned char misaligned; 342 unsigned char discard_misaligned; 343 unsigned char raid_partial_stripes_expensive; 344 enum blk_zoned_model zoned; 345 }; 346 347 #ifdef CONFIG_BLK_DEV_ZONED 348 349 extern unsigned int blkdev_nr_zones(struct block_device *bdev); 350 extern int blkdev_report_zones(struct block_device *bdev, 351 sector_t sector, struct blk_zone *zones, 352 unsigned int *nr_zones, gfp_t gfp_mask); 353 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors, 354 sector_t nr_sectors, gfp_t gfp_mask); 355 extern int blk_revalidate_disk_zones(struct gendisk *disk); 356 357 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, 358 unsigned int cmd, unsigned long arg); 359 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode, 360 unsigned int cmd, unsigned long arg); 361 362 #else /* CONFIG_BLK_DEV_ZONED */ 363 364 static inline unsigned int blkdev_nr_zones(struct block_device *bdev) 365 { 366 return 0; 367 } 368 369 static inline int blk_revalidate_disk_zones(struct gendisk *disk) 370 { 371 return 0; 372 } 373 374 static inline int blkdev_report_zones_ioctl(struct block_device *bdev, 375 fmode_t mode, unsigned int cmd, 376 unsigned long arg) 377 { 378 return -ENOTTY; 379 } 380 381 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev, 382 fmode_t mode, unsigned int cmd, 383 unsigned long arg) 384 { 385 return -ENOTTY; 386 } 387 388 #endif /* CONFIG_BLK_DEV_ZONED */ 389 390 struct request_queue { 391 /* 392 * Together with queue_head for cacheline sharing 393 */ 394 struct list_head queue_head; 395 struct request *last_merge; 396 struct elevator_queue *elevator; 397 398 struct blk_queue_stats *stats; 399 struct rq_qos *rq_qos; 400 401 make_request_fn *make_request_fn; 402 dma_drain_needed_fn *dma_drain_needed; 403 404 const struct blk_mq_ops *mq_ops; 405 406 /* sw queues */ 407 struct blk_mq_ctx __percpu *queue_ctx; 408 unsigned int nr_queues; 409 410 unsigned int queue_depth; 411 412 /* hw dispatch queues */ 413 struct blk_mq_hw_ctx **queue_hw_ctx; 414 unsigned int nr_hw_queues; 415 416 struct backing_dev_info *backing_dev_info; 417 418 /* 419 * The queue owner gets to use this for whatever they like. 420 * ll_rw_blk doesn't touch it. 421 */ 422 void *queuedata; 423 424 /* 425 * various queue flags, see QUEUE_* below 426 */ 427 unsigned long queue_flags; 428 /* 429 * Number of contexts that have called blk_set_pm_only(). If this 430 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are 431 * processed. 432 */ 433 atomic_t pm_only; 434 435 /* 436 * ida allocated id for this queue. Used to index queues from 437 * ioctx. 438 */ 439 int id; 440 441 /* 442 * queue needs bounce pages for pages above this limit 443 */ 444 gfp_t bounce_gfp; 445 446 spinlock_t queue_lock; 447 448 /* 449 * queue kobject 450 */ 451 struct kobject kobj; 452 453 /* 454 * mq queue kobject 455 */ 456 struct kobject *mq_kobj; 457 458 #ifdef CONFIG_BLK_DEV_INTEGRITY 459 struct blk_integrity integrity; 460 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 461 462 #ifdef CONFIG_PM 463 struct device *dev; 464 int rpm_status; 465 unsigned int nr_pending; 466 #endif 467 468 /* 469 * queue settings 470 */ 471 unsigned long nr_requests; /* Max # of requests */ 472 473 unsigned int dma_drain_size; 474 void *dma_drain_buffer; 475 unsigned int dma_pad_mask; 476 unsigned int dma_alignment; 477 478 unsigned int rq_timeout; 479 int poll_nsec; 480 481 struct blk_stat_callback *poll_cb; 482 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS]; 483 484 struct timer_list timeout; 485 struct work_struct timeout_work; 486 487 struct list_head icq_list; 488 #ifdef CONFIG_BLK_CGROUP 489 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 490 struct blkcg_gq *root_blkg; 491 struct list_head blkg_list; 492 #endif 493 494 struct queue_limits limits; 495 496 #ifdef CONFIG_BLK_DEV_ZONED 497 /* 498 * Zoned block device information for request dispatch control. 499 * nr_zones is the total number of zones of the device. This is always 500 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones 501 * bits which indicates if a zone is conventional (bit clear) or 502 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones 503 * bits which indicates if a zone is write locked, that is, if a write 504 * request targeting the zone was dispatched. All three fields are 505 * initialized by the low level device driver (e.g. scsi/sd.c). 506 * Stacking drivers (device mappers) may or may not initialize 507 * these fields. 508 * 509 * Reads of this information must be protected with blk_queue_enter() / 510 * blk_queue_exit(). Modifying this information is only allowed while 511 * no requests are being processed. See also blk_mq_freeze_queue() and 512 * blk_mq_unfreeze_queue(). 513 */ 514 unsigned int nr_zones; 515 unsigned long *seq_zones_bitmap; 516 unsigned long *seq_zones_wlock; 517 #endif /* CONFIG_BLK_DEV_ZONED */ 518 519 /* 520 * sg stuff 521 */ 522 unsigned int sg_timeout; 523 unsigned int sg_reserved_size; 524 int node; 525 #ifdef CONFIG_BLK_DEV_IO_TRACE 526 struct blk_trace *blk_trace; 527 struct mutex blk_trace_mutex; 528 #endif 529 /* 530 * for flush operations 531 */ 532 struct blk_flush_queue *fq; 533 534 struct list_head requeue_list; 535 spinlock_t requeue_lock; 536 struct delayed_work requeue_work; 537 538 struct mutex sysfs_lock; 539 540 atomic_t mq_freeze_depth; 541 542 #if defined(CONFIG_BLK_DEV_BSG) 543 struct bsg_class_device bsg_dev; 544 #endif 545 546 #ifdef CONFIG_BLK_DEV_THROTTLING 547 /* Throttle data */ 548 struct throtl_data *td; 549 #endif 550 struct rcu_head rcu_head; 551 wait_queue_head_t mq_freeze_wq; 552 struct percpu_ref q_usage_counter; 553 struct list_head all_q_node; 554 555 struct blk_mq_tag_set *tag_set; 556 struct list_head tag_set_list; 557 struct bio_set bio_split; 558 559 #ifdef CONFIG_BLK_DEBUG_FS 560 struct dentry *debugfs_dir; 561 struct dentry *sched_debugfs_dir; 562 struct dentry *rqos_debugfs_dir; 563 #endif 564 565 bool mq_sysfs_init_done; 566 567 size_t cmd_size; 568 569 struct work_struct release_work; 570 571 #define BLK_MAX_WRITE_HINTS 5 572 u64 write_hints[BLK_MAX_WRITE_HINTS]; 573 }; 574 575 #define QUEUE_FLAG_STOPPED 1 /* queue is stopped */ 576 #define QUEUE_FLAG_DYING 2 /* queue being torn down */ 577 #define QUEUE_FLAG_BIDI 4 /* queue supports bidi requests */ 578 #define QUEUE_FLAG_NOMERGES 5 /* disable merge attempts */ 579 #define QUEUE_FLAG_SAME_COMP 6 /* complete on same CPU-group */ 580 #define QUEUE_FLAG_FAIL_IO 7 /* fake timeout */ 581 #define QUEUE_FLAG_NONROT 9 /* non-rotational device (SSD) */ 582 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ 583 #define QUEUE_FLAG_IO_STAT 10 /* do disk/partitions IO accounting */ 584 #define QUEUE_FLAG_DISCARD 11 /* supports DISCARD */ 585 #define QUEUE_FLAG_NOXMERGES 12 /* No extended merges */ 586 #define QUEUE_FLAG_ADD_RANDOM 13 /* Contributes to random pool */ 587 #define QUEUE_FLAG_SECERASE 14 /* supports secure erase */ 588 #define QUEUE_FLAG_SAME_FORCE 15 /* force complete on same CPU */ 589 #define QUEUE_FLAG_DEAD 16 /* queue tear-down finished */ 590 #define QUEUE_FLAG_INIT_DONE 17 /* queue is initialized */ 591 #define QUEUE_FLAG_NO_SG_MERGE 18 /* don't attempt to merge SG segments*/ 592 #define QUEUE_FLAG_POLL 19 /* IO polling enabled if set */ 593 #define QUEUE_FLAG_WC 20 /* Write back caching */ 594 #define QUEUE_FLAG_FUA 21 /* device supports FUA writes */ 595 #define QUEUE_FLAG_FLUSH_NQ 22 /* flush not queueuable */ 596 #define QUEUE_FLAG_DAX 23 /* device supports DAX */ 597 #define QUEUE_FLAG_STATS 24 /* track IO start and completion times */ 598 #define QUEUE_FLAG_POLL_STATS 25 /* collecting stats for hybrid polling */ 599 #define QUEUE_FLAG_REGISTERED 26 /* queue has been registered to a disk */ 600 #define QUEUE_FLAG_SCSI_PASSTHROUGH 27 /* queue supports SCSI commands */ 601 #define QUEUE_FLAG_QUIESCED 28 /* queue has been quiesced */ 602 #define QUEUE_FLAG_PCI_P2PDMA 29 /* device supports PCI p2p requests */ 603 604 #define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 605 (1 << QUEUE_FLAG_SAME_COMP) | \ 606 (1 << QUEUE_FLAG_ADD_RANDOM)) 607 608 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 609 (1 << QUEUE_FLAG_SAME_COMP)) 610 611 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 612 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 613 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); 614 615 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 616 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 617 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) 618 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 619 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 620 #define blk_queue_noxmerges(q) \ 621 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 622 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) 623 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) 624 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) 625 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) 626 #define blk_queue_secure_erase(q) \ 627 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags)) 628 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) 629 #define blk_queue_scsi_passthrough(q) \ 630 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags) 631 #define blk_queue_pci_p2pdma(q) \ 632 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags) 633 634 #define blk_noretry_request(rq) \ 635 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 636 REQ_FAILFAST_DRIVER)) 637 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 638 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 639 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags) 640 641 extern void blk_set_pm_only(struct request_queue *q); 642 extern void blk_clear_pm_only(struct request_queue *q); 643 644 static inline bool blk_account_rq(struct request *rq) 645 { 646 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq); 647 } 648 649 #define blk_bidi_rq(rq) ((rq)->next_rq != NULL) 650 651 #define list_entry_rq(ptr) list_entry((ptr), struct request, queuelist) 652 653 #define rq_data_dir(rq) (op_is_write(req_op(rq)) ? WRITE : READ) 654 655 static inline bool queue_is_mq(struct request_queue *q) 656 { 657 return q->mq_ops; 658 } 659 660 static inline enum blk_zoned_model 661 blk_queue_zoned_model(struct request_queue *q) 662 { 663 return q->limits.zoned; 664 } 665 666 static inline bool blk_queue_is_zoned(struct request_queue *q) 667 { 668 switch (blk_queue_zoned_model(q)) { 669 case BLK_ZONED_HA: 670 case BLK_ZONED_HM: 671 return true; 672 default: 673 return false; 674 } 675 } 676 677 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q) 678 { 679 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; 680 } 681 682 #ifdef CONFIG_BLK_DEV_ZONED 683 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 684 { 685 return blk_queue_is_zoned(q) ? q->nr_zones : 0; 686 } 687 688 static inline unsigned int blk_queue_zone_no(struct request_queue *q, 689 sector_t sector) 690 { 691 if (!blk_queue_is_zoned(q)) 692 return 0; 693 return sector >> ilog2(q->limits.chunk_sectors); 694 } 695 696 static inline bool blk_queue_zone_is_seq(struct request_queue *q, 697 sector_t sector) 698 { 699 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap) 700 return false; 701 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap); 702 } 703 #else /* CONFIG_BLK_DEV_ZONED */ 704 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 705 { 706 return 0; 707 } 708 #endif /* CONFIG_BLK_DEV_ZONED */ 709 710 static inline bool rq_is_sync(struct request *rq) 711 { 712 return op_is_sync(rq->cmd_flags); 713 } 714 715 static inline bool rq_mergeable(struct request *rq) 716 { 717 if (blk_rq_is_passthrough(rq)) 718 return false; 719 720 if (req_op(rq) == REQ_OP_FLUSH) 721 return false; 722 723 if (req_op(rq) == REQ_OP_WRITE_ZEROES) 724 return false; 725 726 if (rq->cmd_flags & REQ_NOMERGE_FLAGS) 727 return false; 728 if (rq->rq_flags & RQF_NOMERGE_FLAGS) 729 return false; 730 731 return true; 732 } 733 734 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) 735 { 736 if (bio_page(a) == bio_page(b) && 737 bio_offset(a) == bio_offset(b)) 738 return true; 739 740 return false; 741 } 742 743 static inline unsigned int blk_queue_depth(struct request_queue *q) 744 { 745 if (q->queue_depth) 746 return q->queue_depth; 747 748 return q->nr_requests; 749 } 750 751 extern unsigned long blk_max_low_pfn, blk_max_pfn; 752 753 /* 754 * standard bounce addresses: 755 * 756 * BLK_BOUNCE_HIGH : bounce all highmem pages 757 * BLK_BOUNCE_ANY : don't bounce anything 758 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary 759 */ 760 761 #if BITS_PER_LONG == 32 762 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) 763 #else 764 #define BLK_BOUNCE_HIGH -1ULL 765 #endif 766 #define BLK_BOUNCE_ANY (-1ULL) 767 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) 768 769 /* 770 * default timeout for SG_IO if none specified 771 */ 772 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 773 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 774 775 struct rq_map_data { 776 struct page **pages; 777 int page_order; 778 int nr_entries; 779 unsigned long offset; 780 int null_mapped; 781 int from_user; 782 }; 783 784 struct req_iterator { 785 struct bvec_iter iter; 786 struct bio *bio; 787 }; 788 789 /* This should not be used directly - use rq_for_each_segment */ 790 #define for_each_bio(_bio) \ 791 for (; _bio; _bio = _bio->bi_next) 792 #define __rq_for_each_bio(_bio, rq) \ 793 if ((rq->bio)) \ 794 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) 795 796 #define rq_for_each_segment(bvl, _rq, _iter) \ 797 __rq_for_each_bio(_iter.bio, _rq) \ 798 bio_for_each_segment(bvl, _iter.bio, _iter.iter) 799 800 #define rq_iter_last(bvec, _iter) \ 801 (_iter.bio->bi_next == NULL && \ 802 bio_iter_last(bvec, _iter.iter)) 803 804 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 805 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 806 #endif 807 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 808 extern void rq_flush_dcache_pages(struct request *rq); 809 #else 810 static inline void rq_flush_dcache_pages(struct request *rq) 811 { 812 } 813 #endif 814 815 extern int blk_register_queue(struct gendisk *disk); 816 extern void blk_unregister_queue(struct gendisk *disk); 817 extern blk_qc_t generic_make_request(struct bio *bio); 818 extern blk_qc_t direct_make_request(struct bio *bio); 819 extern void blk_rq_init(struct request_queue *q, struct request *rq); 820 extern void blk_init_request_from_bio(struct request *req, struct bio *bio); 821 extern void blk_put_request(struct request *); 822 extern struct request *blk_get_request(struct request_queue *, unsigned int op, 823 blk_mq_req_flags_t flags); 824 extern int blk_lld_busy(struct request_queue *q); 825 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, 826 struct bio_set *bs, gfp_t gfp_mask, 827 int (*bio_ctr)(struct bio *, struct bio *, void *), 828 void *data); 829 extern void blk_rq_unprep_clone(struct request *rq); 830 extern blk_status_t blk_insert_cloned_request(struct request_queue *q, 831 struct request *rq); 832 extern int blk_rq_append_bio(struct request *rq, struct bio **bio); 833 extern void blk_queue_split(struct request_queue *, struct bio **); 834 extern void blk_recount_segments(struct request_queue *, struct bio *); 835 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); 836 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, 837 unsigned int, void __user *); 838 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, 839 unsigned int, void __user *); 840 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, 841 struct scsi_ioctl_command __user *); 842 843 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 844 extern void blk_queue_exit(struct request_queue *q); 845 extern void blk_sync_queue(struct request_queue *q); 846 extern int blk_rq_map_user(struct request_queue *, struct request *, 847 struct rq_map_data *, void __user *, unsigned long, 848 gfp_t); 849 extern int blk_rq_unmap_user(struct bio *); 850 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); 851 extern int blk_rq_map_user_iov(struct request_queue *, struct request *, 852 struct rq_map_data *, const struct iov_iter *, 853 gfp_t); 854 extern void blk_execute_rq(struct request_queue *, struct gendisk *, 855 struct request *, int); 856 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, 857 struct request *, int, rq_end_io_fn *); 858 859 int blk_status_to_errno(blk_status_t status); 860 blk_status_t errno_to_blk_status(int errno); 861 862 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin); 863 864 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 865 { 866 return bdev->bd_disk->queue; /* this is never NULL */ 867 } 868 869 /* 870 * The basic unit of block I/O is a sector. It is used in a number of contexts 871 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9 872 * bytes. Variables of type sector_t represent an offset or size that is a 873 * multiple of 512 bytes. Hence these two constants. 874 */ 875 #ifndef SECTOR_SHIFT 876 #define SECTOR_SHIFT 9 877 #endif 878 #ifndef SECTOR_SIZE 879 #define SECTOR_SIZE (1 << SECTOR_SHIFT) 880 #endif 881 882 /* 883 * blk_rq_pos() : the current sector 884 * blk_rq_bytes() : bytes left in the entire request 885 * blk_rq_cur_bytes() : bytes left in the current segment 886 * blk_rq_err_bytes() : bytes left till the next error boundary 887 * blk_rq_sectors() : sectors left in the entire request 888 * blk_rq_cur_sectors() : sectors left in the current segment 889 */ 890 static inline sector_t blk_rq_pos(const struct request *rq) 891 { 892 return rq->__sector; 893 } 894 895 static inline unsigned int blk_rq_bytes(const struct request *rq) 896 { 897 return rq->__data_len; 898 } 899 900 static inline int blk_rq_cur_bytes(const struct request *rq) 901 { 902 return rq->bio ? bio_cur_bytes(rq->bio) : 0; 903 } 904 905 extern unsigned int blk_rq_err_bytes(const struct request *rq); 906 907 static inline unsigned int blk_rq_sectors(const struct request *rq) 908 { 909 return blk_rq_bytes(rq) >> SECTOR_SHIFT; 910 } 911 912 static inline unsigned int blk_rq_cur_sectors(const struct request *rq) 913 { 914 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT; 915 } 916 917 #ifdef CONFIG_BLK_DEV_ZONED 918 static inline unsigned int blk_rq_zone_no(struct request *rq) 919 { 920 return blk_queue_zone_no(rq->q, blk_rq_pos(rq)); 921 } 922 923 static inline unsigned int blk_rq_zone_is_seq(struct request *rq) 924 { 925 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq)); 926 } 927 #endif /* CONFIG_BLK_DEV_ZONED */ 928 929 /* 930 * Some commands like WRITE SAME have a payload or data transfer size which 931 * is different from the size of the request. Any driver that supports such 932 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to 933 * calculate the data transfer size. 934 */ 935 static inline unsigned int blk_rq_payload_bytes(struct request *rq) 936 { 937 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 938 return rq->special_vec.bv_len; 939 return blk_rq_bytes(rq); 940 } 941 942 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, 943 int op) 944 { 945 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) 946 return min(q->limits.max_discard_sectors, 947 UINT_MAX >> SECTOR_SHIFT); 948 949 if (unlikely(op == REQ_OP_WRITE_SAME)) 950 return q->limits.max_write_same_sectors; 951 952 if (unlikely(op == REQ_OP_WRITE_ZEROES)) 953 return q->limits.max_write_zeroes_sectors; 954 955 return q->limits.max_sectors; 956 } 957 958 /* 959 * Return maximum size of a request at given offset. Only valid for 960 * file system requests. 961 */ 962 static inline unsigned int blk_max_size_offset(struct request_queue *q, 963 sector_t offset) 964 { 965 if (!q->limits.chunk_sectors) 966 return q->limits.max_sectors; 967 968 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors - 969 (offset & (q->limits.chunk_sectors - 1)))); 970 } 971 972 static inline unsigned int blk_rq_get_max_sectors(struct request *rq, 973 sector_t offset) 974 { 975 struct request_queue *q = rq->q; 976 977 if (blk_rq_is_passthrough(rq)) 978 return q->limits.max_hw_sectors; 979 980 if (!q->limits.chunk_sectors || 981 req_op(rq) == REQ_OP_DISCARD || 982 req_op(rq) == REQ_OP_SECURE_ERASE) 983 return blk_queue_get_max_sectors(q, req_op(rq)); 984 985 return min(blk_max_size_offset(q, offset), 986 blk_queue_get_max_sectors(q, req_op(rq))); 987 } 988 989 static inline unsigned int blk_rq_count_bios(struct request *rq) 990 { 991 unsigned int nr_bios = 0; 992 struct bio *bio; 993 994 __rq_for_each_bio(bio, rq) 995 nr_bios++; 996 997 return nr_bios; 998 } 999 1000 void blk_steal_bios(struct bio_list *list, struct request *rq); 1001 1002 /* 1003 * Request completion related functions. 1004 * 1005 * blk_update_request() completes given number of bytes and updates 1006 * the request without completing it. 1007 * 1008 * blk_end_request() and friends. __blk_end_request() must be called 1009 * with the request queue spinlock acquired. 1010 * 1011 * Several drivers define their own end_request and call 1012 * blk_end_request() for parts of the original function. 1013 * This prevents code duplication in drivers. 1014 */ 1015 extern bool blk_update_request(struct request *rq, blk_status_t error, 1016 unsigned int nr_bytes); 1017 extern void blk_end_request_all(struct request *rq, blk_status_t error); 1018 extern bool __blk_end_request(struct request *rq, blk_status_t error, 1019 unsigned int nr_bytes); 1020 extern void __blk_end_request_all(struct request *rq, blk_status_t error); 1021 extern bool __blk_end_request_cur(struct request *rq, blk_status_t error); 1022 1023 extern void __blk_complete_request(struct request *); 1024 extern void blk_abort_request(struct request *); 1025 1026 /* 1027 * Access functions for manipulating queue properties 1028 */ 1029 extern void blk_cleanup_queue(struct request_queue *); 1030 extern void blk_queue_make_request(struct request_queue *, make_request_fn *); 1031 extern void blk_queue_bounce_limit(struct request_queue *, u64); 1032 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); 1033 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); 1034 extern void blk_queue_max_segments(struct request_queue *, unsigned short); 1035 extern void blk_queue_max_discard_segments(struct request_queue *, 1036 unsigned short); 1037 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); 1038 extern void blk_queue_max_discard_sectors(struct request_queue *q, 1039 unsigned int max_discard_sectors); 1040 extern void blk_queue_max_write_same_sectors(struct request_queue *q, 1041 unsigned int max_write_same_sectors); 1042 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, 1043 unsigned int max_write_same_sectors); 1044 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); 1045 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); 1046 extern void blk_queue_alignment_offset(struct request_queue *q, 1047 unsigned int alignment); 1048 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 1049 extern void blk_queue_io_min(struct request_queue *q, unsigned int min); 1050 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 1051 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); 1052 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1053 extern void blk_set_default_limits(struct queue_limits *lim); 1054 extern void blk_set_stacking_limits(struct queue_limits *lim); 1055 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1056 sector_t offset); 1057 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, 1058 sector_t offset); 1059 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, 1060 sector_t offset); 1061 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); 1062 extern void blk_queue_dma_pad(struct request_queue *, unsigned int); 1063 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 1064 extern int blk_queue_dma_drain(struct request_queue *q, 1065 dma_drain_needed_fn *dma_drain_needed, 1066 void *buf, unsigned int size); 1067 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); 1068 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long); 1069 extern void blk_queue_dma_alignment(struct request_queue *, int); 1070 extern void blk_queue_update_dma_alignment(struct request_queue *, int); 1071 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1072 extern void blk_queue_flush_queueable(struct request_queue *q, bool queueable); 1073 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); 1074 1075 /* 1076 * Number of physical segments as sent to the device. 1077 * 1078 * Normally this is the number of discontiguous data segments sent by the 1079 * submitter. But for data-less command like discard we might have no 1080 * actual data segments submitted, but the driver might have to add it's 1081 * own special payload. In that case we still return 1 here so that this 1082 * special payload will be mapped. 1083 */ 1084 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq) 1085 { 1086 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 1087 return 1; 1088 return rq->nr_phys_segments; 1089 } 1090 1091 /* 1092 * Number of discard segments (or ranges) the driver needs to fill in. 1093 * Each discard bio merged into a request is counted as one segment. 1094 */ 1095 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq) 1096 { 1097 return max_t(unsigned short, rq->nr_phys_segments, 1); 1098 } 1099 1100 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); 1101 extern void blk_dump_rq_flags(struct request *, char *); 1102 extern long nr_blockdev_pages(void); 1103 1104 bool __must_check blk_get_queue(struct request_queue *); 1105 struct request_queue *blk_alloc_queue(gfp_t); 1106 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id); 1107 extern void blk_put_queue(struct request_queue *); 1108 extern void blk_set_queue_dying(struct request_queue *); 1109 1110 /* 1111 * blk_plug permits building a queue of related requests by holding the I/O 1112 * fragments for a short period. This allows merging of sequential requests 1113 * into single larger request. As the requests are moved from a per-task list to 1114 * the device's request_queue in a batch, this results in improved scalability 1115 * as the lock contention for request_queue lock is reduced. 1116 * 1117 * It is ok not to disable preemption when adding the request to the plug list 1118 * or when attempting a merge, because blk_schedule_flush_list() will only flush 1119 * the plug list when the task sleeps by itself. For details, please see 1120 * schedule() where blk_schedule_flush_plug() is called. 1121 */ 1122 struct blk_plug { 1123 struct list_head mq_list; /* blk-mq requests */ 1124 struct list_head cb_list; /* md requires an unplug callback */ 1125 unsigned short rq_count; 1126 bool multiple_queues; 1127 }; 1128 #define BLK_MAX_REQUEST_COUNT 16 1129 #define BLK_PLUG_FLUSH_SIZE (128 * 1024) 1130 1131 struct blk_plug_cb; 1132 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1133 struct blk_plug_cb { 1134 struct list_head list; 1135 blk_plug_cb_fn callback; 1136 void *data; 1137 }; 1138 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1139 void *data, int size); 1140 extern void blk_start_plug(struct blk_plug *); 1141 extern void blk_finish_plug(struct blk_plug *); 1142 extern void blk_flush_plug_list(struct blk_plug *, bool); 1143 1144 static inline void blk_flush_plug(struct task_struct *tsk) 1145 { 1146 struct blk_plug *plug = tsk->plug; 1147 1148 if (plug) 1149 blk_flush_plug_list(plug, false); 1150 } 1151 1152 static inline void blk_schedule_flush_plug(struct task_struct *tsk) 1153 { 1154 struct blk_plug *plug = tsk->plug; 1155 1156 if (plug) 1157 blk_flush_plug_list(plug, true); 1158 } 1159 1160 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1161 { 1162 struct blk_plug *plug = tsk->plug; 1163 1164 return plug && 1165 (!list_empty(&plug->mq_list) || 1166 !list_empty(&plug->cb_list)); 1167 } 1168 1169 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); 1170 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, 1171 sector_t nr_sects, gfp_t gfp_mask, struct page *page); 1172 1173 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */ 1174 1175 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1176 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); 1177 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1178 sector_t nr_sects, gfp_t gfp_mask, int flags, 1179 struct bio **biop); 1180 1181 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1182 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1183 1184 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1185 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1186 unsigned flags); 1187 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1188 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1189 1190 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1191 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1192 { 1193 return blkdev_issue_discard(sb->s_bdev, 1194 block << (sb->s_blocksize_bits - 1195 SECTOR_SHIFT), 1196 nr_blocks << (sb->s_blocksize_bits - 1197 SECTOR_SHIFT), 1198 gfp_mask, flags); 1199 } 1200 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1201 sector_t nr_blocks, gfp_t gfp_mask) 1202 { 1203 return blkdev_issue_zeroout(sb->s_bdev, 1204 block << (sb->s_blocksize_bits - 1205 SECTOR_SHIFT), 1206 nr_blocks << (sb->s_blocksize_bits - 1207 SECTOR_SHIFT), 1208 gfp_mask, 0); 1209 } 1210 1211 extern int blk_verify_command(unsigned char *cmd, fmode_t mode); 1212 1213 enum blk_default_limits { 1214 BLK_MAX_SEGMENTS = 128, 1215 BLK_SAFE_MAX_SECTORS = 255, 1216 BLK_DEF_MAX_SECTORS = 2560, 1217 BLK_MAX_SEGMENT_SIZE = 65536, 1218 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1219 }; 1220 1221 static inline unsigned long queue_segment_boundary(struct request_queue *q) 1222 { 1223 return q->limits.seg_boundary_mask; 1224 } 1225 1226 static inline unsigned long queue_virt_boundary(struct request_queue *q) 1227 { 1228 return q->limits.virt_boundary_mask; 1229 } 1230 1231 static inline unsigned int queue_max_sectors(struct request_queue *q) 1232 { 1233 return q->limits.max_sectors; 1234 } 1235 1236 static inline unsigned int queue_max_hw_sectors(struct request_queue *q) 1237 { 1238 return q->limits.max_hw_sectors; 1239 } 1240 1241 static inline unsigned short queue_max_segments(struct request_queue *q) 1242 { 1243 return q->limits.max_segments; 1244 } 1245 1246 static inline unsigned short queue_max_discard_segments(struct request_queue *q) 1247 { 1248 return q->limits.max_discard_segments; 1249 } 1250 1251 static inline unsigned int queue_max_segment_size(struct request_queue *q) 1252 { 1253 return q->limits.max_segment_size; 1254 } 1255 1256 static inline unsigned short queue_logical_block_size(struct request_queue *q) 1257 { 1258 int retval = 512; 1259 1260 if (q && q->limits.logical_block_size) 1261 retval = q->limits.logical_block_size; 1262 1263 return retval; 1264 } 1265 1266 static inline unsigned short bdev_logical_block_size(struct block_device *bdev) 1267 { 1268 return queue_logical_block_size(bdev_get_queue(bdev)); 1269 } 1270 1271 static inline unsigned int queue_physical_block_size(struct request_queue *q) 1272 { 1273 return q->limits.physical_block_size; 1274 } 1275 1276 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1277 { 1278 return queue_physical_block_size(bdev_get_queue(bdev)); 1279 } 1280 1281 static inline unsigned int queue_io_min(struct request_queue *q) 1282 { 1283 return q->limits.io_min; 1284 } 1285 1286 static inline int bdev_io_min(struct block_device *bdev) 1287 { 1288 return queue_io_min(bdev_get_queue(bdev)); 1289 } 1290 1291 static inline unsigned int queue_io_opt(struct request_queue *q) 1292 { 1293 return q->limits.io_opt; 1294 } 1295 1296 static inline int bdev_io_opt(struct block_device *bdev) 1297 { 1298 return queue_io_opt(bdev_get_queue(bdev)); 1299 } 1300 1301 static inline int queue_alignment_offset(struct request_queue *q) 1302 { 1303 if (q->limits.misaligned) 1304 return -1; 1305 1306 return q->limits.alignment_offset; 1307 } 1308 1309 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) 1310 { 1311 unsigned int granularity = max(lim->physical_block_size, lim->io_min); 1312 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) 1313 << SECTOR_SHIFT; 1314 1315 return (granularity + lim->alignment_offset - alignment) % granularity; 1316 } 1317 1318 static inline int bdev_alignment_offset(struct block_device *bdev) 1319 { 1320 struct request_queue *q = bdev_get_queue(bdev); 1321 1322 if (q->limits.misaligned) 1323 return -1; 1324 1325 if (bdev != bdev->bd_contains) 1326 return bdev->bd_part->alignment_offset; 1327 1328 return q->limits.alignment_offset; 1329 } 1330 1331 static inline int queue_discard_alignment(struct request_queue *q) 1332 { 1333 if (q->limits.discard_misaligned) 1334 return -1; 1335 1336 return q->limits.discard_alignment; 1337 } 1338 1339 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) 1340 { 1341 unsigned int alignment, granularity, offset; 1342 1343 if (!lim->max_discard_sectors) 1344 return 0; 1345 1346 /* Why are these in bytes, not sectors? */ 1347 alignment = lim->discard_alignment >> SECTOR_SHIFT; 1348 granularity = lim->discard_granularity >> SECTOR_SHIFT; 1349 if (!granularity) 1350 return 0; 1351 1352 /* Offset of the partition start in 'granularity' sectors */ 1353 offset = sector_div(sector, granularity); 1354 1355 /* And why do we do this modulus *again* in blkdev_issue_discard()? */ 1356 offset = (granularity + alignment - offset) % granularity; 1357 1358 /* Turn it back into bytes, gaah */ 1359 return offset << SECTOR_SHIFT; 1360 } 1361 1362 static inline int bdev_discard_alignment(struct block_device *bdev) 1363 { 1364 struct request_queue *q = bdev_get_queue(bdev); 1365 1366 if (bdev != bdev->bd_contains) 1367 return bdev->bd_part->discard_alignment; 1368 1369 return q->limits.discard_alignment; 1370 } 1371 1372 static inline unsigned int bdev_write_same(struct block_device *bdev) 1373 { 1374 struct request_queue *q = bdev_get_queue(bdev); 1375 1376 if (q) 1377 return q->limits.max_write_same_sectors; 1378 1379 return 0; 1380 } 1381 1382 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1383 { 1384 struct request_queue *q = bdev_get_queue(bdev); 1385 1386 if (q) 1387 return q->limits.max_write_zeroes_sectors; 1388 1389 return 0; 1390 } 1391 1392 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev) 1393 { 1394 struct request_queue *q = bdev_get_queue(bdev); 1395 1396 if (q) 1397 return blk_queue_zoned_model(q); 1398 1399 return BLK_ZONED_NONE; 1400 } 1401 1402 static inline bool bdev_is_zoned(struct block_device *bdev) 1403 { 1404 struct request_queue *q = bdev_get_queue(bdev); 1405 1406 if (q) 1407 return blk_queue_is_zoned(q); 1408 1409 return false; 1410 } 1411 1412 static inline unsigned int bdev_zone_sectors(struct block_device *bdev) 1413 { 1414 struct request_queue *q = bdev_get_queue(bdev); 1415 1416 if (q) 1417 return blk_queue_zone_sectors(q); 1418 return 0; 1419 } 1420 1421 static inline int queue_dma_alignment(struct request_queue *q) 1422 { 1423 return q ? q->dma_alignment : 511; 1424 } 1425 1426 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1427 unsigned int len) 1428 { 1429 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1430 return !(addr & alignment) && !(len & alignment); 1431 } 1432 1433 /* assumes size > 256 */ 1434 static inline unsigned int blksize_bits(unsigned int size) 1435 { 1436 unsigned int bits = 8; 1437 do { 1438 bits++; 1439 size >>= 1; 1440 } while (size > 256); 1441 return bits; 1442 } 1443 1444 static inline unsigned int block_size(struct block_device *bdev) 1445 { 1446 return bdev->bd_block_size; 1447 } 1448 1449 static inline bool queue_flush_queueable(struct request_queue *q) 1450 { 1451 return !test_bit(QUEUE_FLAG_FLUSH_NQ, &q->queue_flags); 1452 } 1453 1454 typedef struct {struct page *v;} Sector; 1455 1456 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); 1457 1458 static inline void put_dev_sector(Sector p) 1459 { 1460 put_page(p.v); 1461 } 1462 1463 int kblockd_schedule_work(struct work_struct *work); 1464 int kblockd_schedule_work_on(int cpu, struct work_struct *work); 1465 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1466 1467 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1468 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1469 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1470 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1471 1472 #if defined(CONFIG_BLK_DEV_INTEGRITY) 1473 1474 enum blk_integrity_flags { 1475 BLK_INTEGRITY_VERIFY = 1 << 0, 1476 BLK_INTEGRITY_GENERATE = 1 << 1, 1477 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, 1478 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, 1479 }; 1480 1481 struct blk_integrity_iter { 1482 void *prot_buf; 1483 void *data_buf; 1484 sector_t seed; 1485 unsigned int data_size; 1486 unsigned short interval; 1487 const char *disk_name; 1488 }; 1489 1490 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *); 1491 1492 struct blk_integrity_profile { 1493 integrity_processing_fn *generate_fn; 1494 integrity_processing_fn *verify_fn; 1495 const char *name; 1496 }; 1497 1498 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *); 1499 extern void blk_integrity_unregister(struct gendisk *); 1500 extern int blk_integrity_compare(struct gendisk *, struct gendisk *); 1501 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, 1502 struct scatterlist *); 1503 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); 1504 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, 1505 struct request *); 1506 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, 1507 struct bio *); 1508 1509 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1510 { 1511 struct blk_integrity *bi = &disk->queue->integrity; 1512 1513 if (!bi->profile) 1514 return NULL; 1515 1516 return bi; 1517 } 1518 1519 static inline 1520 struct blk_integrity *bdev_get_integrity(struct block_device *bdev) 1521 { 1522 return blk_get_integrity(bdev->bd_disk); 1523 } 1524 1525 static inline bool blk_integrity_rq(struct request *rq) 1526 { 1527 return rq->cmd_flags & REQ_INTEGRITY; 1528 } 1529 1530 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1531 unsigned int segs) 1532 { 1533 q->limits.max_integrity_segments = segs; 1534 } 1535 1536 static inline unsigned short 1537 queue_max_integrity_segments(struct request_queue *q) 1538 { 1539 return q->limits.max_integrity_segments; 1540 } 1541 1542 /** 1543 * bio_integrity_intervals - Return number of integrity intervals for a bio 1544 * @bi: blk_integrity profile for device 1545 * @sectors: Size of the bio in 512-byte sectors 1546 * 1547 * Description: The block layer calculates everything in 512 byte 1548 * sectors but integrity metadata is done in terms of the data integrity 1549 * interval size of the storage device. Convert the block layer sectors 1550 * to the appropriate number of integrity intervals. 1551 */ 1552 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1553 unsigned int sectors) 1554 { 1555 return sectors >> (bi->interval_exp - 9); 1556 } 1557 1558 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1559 unsigned int sectors) 1560 { 1561 return bio_integrity_intervals(bi, sectors) * bi->tuple_size; 1562 } 1563 1564 #else /* CONFIG_BLK_DEV_INTEGRITY */ 1565 1566 struct bio; 1567 struct block_device; 1568 struct gendisk; 1569 struct blk_integrity; 1570 1571 static inline int blk_integrity_rq(struct request *rq) 1572 { 1573 return 0; 1574 } 1575 static inline int blk_rq_count_integrity_sg(struct request_queue *q, 1576 struct bio *b) 1577 { 1578 return 0; 1579 } 1580 static inline int blk_rq_map_integrity_sg(struct request_queue *q, 1581 struct bio *b, 1582 struct scatterlist *s) 1583 { 1584 return 0; 1585 } 1586 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) 1587 { 1588 return NULL; 1589 } 1590 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1591 { 1592 return NULL; 1593 } 1594 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) 1595 { 1596 return 0; 1597 } 1598 static inline void blk_integrity_register(struct gendisk *d, 1599 struct blk_integrity *b) 1600 { 1601 } 1602 static inline void blk_integrity_unregister(struct gendisk *d) 1603 { 1604 } 1605 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1606 unsigned int segs) 1607 { 1608 } 1609 static inline unsigned short queue_max_integrity_segments(struct request_queue *q) 1610 { 1611 return 0; 1612 } 1613 static inline bool blk_integrity_merge_rq(struct request_queue *rq, 1614 struct request *r1, 1615 struct request *r2) 1616 { 1617 return true; 1618 } 1619 static inline bool blk_integrity_merge_bio(struct request_queue *rq, 1620 struct request *r, 1621 struct bio *b) 1622 { 1623 return true; 1624 } 1625 1626 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1627 unsigned int sectors) 1628 { 1629 return 0; 1630 } 1631 1632 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1633 unsigned int sectors) 1634 { 1635 return 0; 1636 } 1637 1638 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 1639 1640 struct block_device_operations { 1641 int (*open) (struct block_device *, fmode_t); 1642 void (*release) (struct gendisk *, fmode_t); 1643 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int); 1644 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1645 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1646 unsigned int (*check_events) (struct gendisk *disk, 1647 unsigned int clearing); 1648 /* ->media_changed() is DEPRECATED, use ->check_events() instead */ 1649 int (*media_changed) (struct gendisk *); 1650 void (*unlock_native_capacity) (struct gendisk *); 1651 int (*revalidate_disk) (struct gendisk *); 1652 int (*getgeo)(struct block_device *, struct hd_geometry *); 1653 /* this callback is with swap_lock and sometimes page table lock held */ 1654 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1655 int (*report_zones)(struct gendisk *, sector_t sector, 1656 struct blk_zone *zones, unsigned int *nr_zones, 1657 gfp_t gfp_mask); 1658 struct module *owner; 1659 const struct pr_ops *pr_ops; 1660 }; 1661 1662 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, 1663 unsigned long); 1664 extern int bdev_read_page(struct block_device *, sector_t, struct page *); 1665 extern int bdev_write_page(struct block_device *, sector_t, struct page *, 1666 struct writeback_control *); 1667 1668 #ifdef CONFIG_BLK_DEV_ZONED 1669 bool blk_req_needs_zone_write_lock(struct request *rq); 1670 void __blk_req_zone_write_lock(struct request *rq); 1671 void __blk_req_zone_write_unlock(struct request *rq); 1672 1673 static inline void blk_req_zone_write_lock(struct request *rq) 1674 { 1675 if (blk_req_needs_zone_write_lock(rq)) 1676 __blk_req_zone_write_lock(rq); 1677 } 1678 1679 static inline void blk_req_zone_write_unlock(struct request *rq) 1680 { 1681 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED) 1682 __blk_req_zone_write_unlock(rq); 1683 } 1684 1685 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1686 { 1687 return rq->q->seq_zones_wlock && 1688 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock); 1689 } 1690 1691 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1692 { 1693 if (!blk_req_needs_zone_write_lock(rq)) 1694 return true; 1695 return !blk_req_zone_is_write_locked(rq); 1696 } 1697 #else 1698 static inline bool blk_req_needs_zone_write_lock(struct request *rq) 1699 { 1700 return false; 1701 } 1702 1703 static inline void blk_req_zone_write_lock(struct request *rq) 1704 { 1705 } 1706 1707 static inline void blk_req_zone_write_unlock(struct request *rq) 1708 { 1709 } 1710 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1711 { 1712 return false; 1713 } 1714 1715 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1716 { 1717 return true; 1718 } 1719 #endif /* CONFIG_BLK_DEV_ZONED */ 1720 1721 #else /* CONFIG_BLOCK */ 1722 1723 struct block_device; 1724 1725 /* 1726 * stubs for when the block layer is configured out 1727 */ 1728 #define buffer_heads_over_limit 0 1729 1730 static inline long nr_blockdev_pages(void) 1731 { 1732 return 0; 1733 } 1734 1735 struct blk_plug { 1736 }; 1737 1738 static inline void blk_start_plug(struct blk_plug *plug) 1739 { 1740 } 1741 1742 static inline void blk_finish_plug(struct blk_plug *plug) 1743 { 1744 } 1745 1746 static inline void blk_flush_plug(struct task_struct *task) 1747 { 1748 } 1749 1750 static inline void blk_schedule_flush_plug(struct task_struct *task) 1751 { 1752 } 1753 1754 1755 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1756 { 1757 return false; 1758 } 1759 1760 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, 1761 sector_t *error_sector) 1762 { 1763 return 0; 1764 } 1765 1766 #endif /* CONFIG_BLOCK */ 1767 1768 static inline void blk_wake_io_task(struct task_struct *waiter) 1769 { 1770 /* 1771 * If we're polling, the task itself is doing the completions. For 1772 * that case, we don't need to signal a wakeup, it's enough to just 1773 * mark us as RUNNING. 1774 */ 1775 if (waiter == current) 1776 __set_current_state(TASK_RUNNING); 1777 else 1778 wake_up_process(waiter); 1779 } 1780 1781 #endif 1782