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 /* Time that I/O was submitted to the kernel. */ 198 u64 start_time_ns; 199 /* Time that I/O was submitted to the device. */ 200 u64 io_start_time_ns; 201 202 #ifdef CONFIG_BLK_WBT 203 unsigned short wbt_flags; 204 #endif 205 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW 206 unsigned short throtl_size; 207 #endif 208 209 /* 210 * Number of scatter-gather DMA addr+len pairs after 211 * physical address coalescing is performed. 212 */ 213 unsigned short nr_phys_segments; 214 215 #if defined(CONFIG_BLK_DEV_INTEGRITY) 216 unsigned short nr_integrity_segments; 217 #endif 218 219 unsigned short write_hint; 220 unsigned short ioprio; 221 222 unsigned int extra_len; /* length of alignment and padding */ 223 224 enum mq_rq_state state; 225 refcount_t ref; 226 227 unsigned int timeout; 228 unsigned long deadline; 229 230 union { 231 struct __call_single_data csd; 232 u64 fifo_time; 233 }; 234 235 /* 236 * completion callback. 237 */ 238 rq_end_io_fn *end_io; 239 void *end_io_data; 240 }; 241 242 static inline bool blk_op_is_scsi(unsigned int op) 243 { 244 return op == REQ_OP_SCSI_IN || op == REQ_OP_SCSI_OUT; 245 } 246 247 static inline bool blk_op_is_private(unsigned int op) 248 { 249 return op == REQ_OP_DRV_IN || op == REQ_OP_DRV_OUT; 250 } 251 252 static inline bool blk_rq_is_scsi(struct request *rq) 253 { 254 return blk_op_is_scsi(req_op(rq)); 255 } 256 257 static inline bool blk_rq_is_private(struct request *rq) 258 { 259 return blk_op_is_private(req_op(rq)); 260 } 261 262 static inline bool blk_rq_is_passthrough(struct request *rq) 263 { 264 return blk_rq_is_scsi(rq) || blk_rq_is_private(rq); 265 } 266 267 static inline bool bio_is_passthrough(struct bio *bio) 268 { 269 unsigned op = bio_op(bio); 270 271 return blk_op_is_scsi(op) || blk_op_is_private(op); 272 } 273 274 static inline unsigned short req_get_ioprio(struct request *req) 275 { 276 return req->ioprio; 277 } 278 279 #include <linux/elevator.h> 280 281 struct blk_queue_ctx; 282 283 typedef blk_qc_t (make_request_fn) (struct request_queue *q, struct bio *bio); 284 285 struct bio_vec; 286 typedef int (dma_drain_needed_fn)(struct request *); 287 288 enum blk_eh_timer_return { 289 BLK_EH_DONE, /* drivers has completed the command */ 290 BLK_EH_RESET_TIMER, /* reset timer and try again */ 291 }; 292 293 enum blk_queue_state { 294 Queue_down, 295 Queue_up, 296 }; 297 298 #define BLK_TAG_ALLOC_FIFO 0 /* allocate starting from 0 */ 299 #define BLK_TAG_ALLOC_RR 1 /* allocate starting from last allocated tag */ 300 301 #define BLK_SCSI_MAX_CMDS (256) 302 #define BLK_SCSI_CMD_PER_LONG (BLK_SCSI_MAX_CMDS / (sizeof(long) * 8)) 303 304 /* 305 * Zoned block device models (zoned limit). 306 */ 307 enum blk_zoned_model { 308 BLK_ZONED_NONE, /* Regular block device */ 309 BLK_ZONED_HA, /* Host-aware zoned block device */ 310 BLK_ZONED_HM, /* Host-managed zoned block device */ 311 }; 312 313 struct queue_limits { 314 unsigned long bounce_pfn; 315 unsigned long seg_boundary_mask; 316 unsigned long virt_boundary_mask; 317 318 unsigned int max_hw_sectors; 319 unsigned int max_dev_sectors; 320 unsigned int chunk_sectors; 321 unsigned int max_sectors; 322 unsigned int max_segment_size; 323 unsigned int physical_block_size; 324 unsigned int alignment_offset; 325 unsigned int io_min; 326 unsigned int io_opt; 327 unsigned int max_discard_sectors; 328 unsigned int max_hw_discard_sectors; 329 unsigned int max_write_same_sectors; 330 unsigned int max_write_zeroes_sectors; 331 unsigned int discard_granularity; 332 unsigned int discard_alignment; 333 334 unsigned short logical_block_size; 335 unsigned short max_segments; 336 unsigned short max_integrity_segments; 337 unsigned short max_discard_segments; 338 339 unsigned char misaligned; 340 unsigned char discard_misaligned; 341 unsigned char raid_partial_stripes_expensive; 342 enum blk_zoned_model zoned; 343 }; 344 345 #ifdef CONFIG_BLK_DEV_ZONED 346 347 extern unsigned int blkdev_nr_zones(struct block_device *bdev); 348 extern int blkdev_report_zones(struct block_device *bdev, 349 sector_t sector, struct blk_zone *zones, 350 unsigned int *nr_zones, gfp_t gfp_mask); 351 extern int blkdev_reset_zones(struct block_device *bdev, sector_t sectors, 352 sector_t nr_sectors, gfp_t gfp_mask); 353 extern int blk_revalidate_disk_zones(struct gendisk *disk); 354 355 extern int blkdev_report_zones_ioctl(struct block_device *bdev, fmode_t mode, 356 unsigned int cmd, unsigned long arg); 357 extern int blkdev_reset_zones_ioctl(struct block_device *bdev, fmode_t mode, 358 unsigned int cmd, unsigned long arg); 359 360 #else /* CONFIG_BLK_DEV_ZONED */ 361 362 static inline unsigned int blkdev_nr_zones(struct block_device *bdev) 363 { 364 return 0; 365 } 366 367 static inline int blk_revalidate_disk_zones(struct gendisk *disk) 368 { 369 return 0; 370 } 371 372 static inline int blkdev_report_zones_ioctl(struct block_device *bdev, 373 fmode_t mode, unsigned int cmd, 374 unsigned long arg) 375 { 376 return -ENOTTY; 377 } 378 379 static inline int blkdev_reset_zones_ioctl(struct block_device *bdev, 380 fmode_t mode, unsigned int cmd, 381 unsigned long arg) 382 { 383 return -ENOTTY; 384 } 385 386 #endif /* CONFIG_BLK_DEV_ZONED */ 387 388 struct request_queue { 389 /* 390 * Together with queue_head for cacheline sharing 391 */ 392 struct list_head queue_head; 393 struct request *last_merge; 394 struct elevator_queue *elevator; 395 396 struct blk_queue_stats *stats; 397 struct rq_qos *rq_qos; 398 399 make_request_fn *make_request_fn; 400 dma_drain_needed_fn *dma_drain_needed; 401 402 const struct blk_mq_ops *mq_ops; 403 404 /* sw queues */ 405 struct blk_mq_ctx __percpu *queue_ctx; 406 unsigned int nr_queues; 407 408 unsigned int queue_depth; 409 410 /* hw dispatch queues */ 411 struct blk_mq_hw_ctx **queue_hw_ctx; 412 unsigned int nr_hw_queues; 413 414 struct backing_dev_info *backing_dev_info; 415 416 /* 417 * The queue owner gets to use this for whatever they like. 418 * ll_rw_blk doesn't touch it. 419 */ 420 void *queuedata; 421 422 /* 423 * various queue flags, see QUEUE_* below 424 */ 425 unsigned long queue_flags; 426 /* 427 * Number of contexts that have called blk_set_pm_only(). If this 428 * counter is above zero then only RQF_PM and RQF_PREEMPT requests are 429 * processed. 430 */ 431 atomic_t pm_only; 432 433 /* 434 * ida allocated id for this queue. Used to index queues from 435 * ioctx. 436 */ 437 int id; 438 439 /* 440 * queue needs bounce pages for pages above this limit 441 */ 442 gfp_t bounce_gfp; 443 444 spinlock_t queue_lock; 445 446 /* 447 * queue kobject 448 */ 449 struct kobject kobj; 450 451 /* 452 * mq queue kobject 453 */ 454 struct kobject *mq_kobj; 455 456 #ifdef CONFIG_BLK_DEV_INTEGRITY 457 struct blk_integrity integrity; 458 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 459 460 #ifdef CONFIG_PM 461 struct device *dev; 462 int rpm_status; 463 unsigned int nr_pending; 464 #endif 465 466 /* 467 * queue settings 468 */ 469 unsigned long nr_requests; /* Max # of requests */ 470 471 unsigned int dma_drain_size; 472 void *dma_drain_buffer; 473 unsigned int dma_pad_mask; 474 unsigned int dma_alignment; 475 476 unsigned int rq_timeout; 477 int poll_nsec; 478 479 struct blk_stat_callback *poll_cb; 480 struct blk_rq_stat poll_stat[BLK_MQ_POLL_STATS_BKTS]; 481 482 struct timer_list timeout; 483 struct work_struct timeout_work; 484 485 struct list_head icq_list; 486 #ifdef CONFIG_BLK_CGROUP 487 DECLARE_BITMAP (blkcg_pols, BLKCG_MAX_POLS); 488 struct blkcg_gq *root_blkg; 489 struct list_head blkg_list; 490 #endif 491 492 struct queue_limits limits; 493 494 #ifdef CONFIG_BLK_DEV_ZONED 495 /* 496 * Zoned block device information for request dispatch control. 497 * nr_zones is the total number of zones of the device. This is always 498 * 0 for regular block devices. seq_zones_bitmap is a bitmap of nr_zones 499 * bits which indicates if a zone is conventional (bit clear) or 500 * sequential (bit set). seq_zones_wlock is a bitmap of nr_zones 501 * bits which indicates if a zone is write locked, that is, if a write 502 * request targeting the zone was dispatched. All three fields are 503 * initialized by the low level device driver (e.g. scsi/sd.c). 504 * Stacking drivers (device mappers) may or may not initialize 505 * these fields. 506 * 507 * Reads of this information must be protected with blk_queue_enter() / 508 * blk_queue_exit(). Modifying this information is only allowed while 509 * no requests are being processed. See also blk_mq_freeze_queue() and 510 * blk_mq_unfreeze_queue(). 511 */ 512 unsigned int nr_zones; 513 unsigned long *seq_zones_bitmap; 514 unsigned long *seq_zones_wlock; 515 #endif /* CONFIG_BLK_DEV_ZONED */ 516 517 /* 518 * sg stuff 519 */ 520 unsigned int sg_timeout; 521 unsigned int sg_reserved_size; 522 int node; 523 #ifdef CONFIG_BLK_DEV_IO_TRACE 524 struct blk_trace *blk_trace; 525 struct mutex blk_trace_mutex; 526 #endif 527 /* 528 * for flush operations 529 */ 530 struct blk_flush_queue *fq; 531 532 struct list_head requeue_list; 533 spinlock_t requeue_lock; 534 struct delayed_work requeue_work; 535 536 struct mutex sysfs_lock; 537 538 /* 539 * for reusing dead hctx instance in case of updating 540 * nr_hw_queues 541 */ 542 struct list_head unused_hctx_list; 543 spinlock_t unused_hctx_lock; 544 545 int mq_freeze_depth; 546 547 #if defined(CONFIG_BLK_DEV_BSG) 548 struct bsg_class_device bsg_dev; 549 #endif 550 551 #ifdef CONFIG_BLK_DEV_THROTTLING 552 /* Throttle data */ 553 struct throtl_data *td; 554 #endif 555 struct rcu_head rcu_head; 556 wait_queue_head_t mq_freeze_wq; 557 /* 558 * Protect concurrent access to q_usage_counter by 559 * percpu_ref_kill() and percpu_ref_reinit(). 560 */ 561 struct mutex mq_freeze_lock; 562 struct percpu_ref q_usage_counter; 563 564 struct blk_mq_tag_set *tag_set; 565 struct list_head tag_set_list; 566 struct bio_set bio_split; 567 568 #ifdef CONFIG_BLK_DEBUG_FS 569 struct dentry *debugfs_dir; 570 struct dentry *sched_debugfs_dir; 571 struct dentry *rqos_debugfs_dir; 572 #endif 573 574 bool mq_sysfs_init_done; 575 576 size_t cmd_size; 577 578 struct work_struct release_work; 579 580 #define BLK_MAX_WRITE_HINTS 5 581 u64 write_hints[BLK_MAX_WRITE_HINTS]; 582 }; 583 584 #define QUEUE_FLAG_STOPPED 0 /* queue is stopped */ 585 #define QUEUE_FLAG_DYING 1 /* queue being torn down */ 586 #define QUEUE_FLAG_NOMERGES 3 /* disable merge attempts */ 587 #define QUEUE_FLAG_SAME_COMP 4 /* complete on same CPU-group */ 588 #define QUEUE_FLAG_FAIL_IO 5 /* fake timeout */ 589 #define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */ 590 #define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */ 591 #define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */ 592 #define QUEUE_FLAG_DISCARD 8 /* supports DISCARD */ 593 #define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */ 594 #define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */ 595 #define QUEUE_FLAG_SECERASE 11 /* supports secure erase */ 596 #define QUEUE_FLAG_SAME_FORCE 12 /* force complete on same CPU */ 597 #define QUEUE_FLAG_DEAD 13 /* queue tear-down finished */ 598 #define QUEUE_FLAG_INIT_DONE 14 /* queue is initialized */ 599 #define QUEUE_FLAG_POLL 16 /* IO polling enabled if set */ 600 #define QUEUE_FLAG_WC 17 /* Write back caching */ 601 #define QUEUE_FLAG_FUA 18 /* device supports FUA writes */ 602 #define QUEUE_FLAG_DAX 19 /* device supports DAX */ 603 #define QUEUE_FLAG_STATS 20 /* track IO start and completion times */ 604 #define QUEUE_FLAG_POLL_STATS 21 /* collecting stats for hybrid polling */ 605 #define QUEUE_FLAG_REGISTERED 22 /* queue has been registered to a disk */ 606 #define QUEUE_FLAG_SCSI_PASSTHROUGH 23 /* queue supports SCSI commands */ 607 #define QUEUE_FLAG_QUIESCED 24 /* queue has been quiesced */ 608 #define QUEUE_FLAG_PCI_P2PDMA 25 /* device supports PCI p2p requests */ 609 610 #define QUEUE_FLAG_MQ_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \ 611 (1 << QUEUE_FLAG_SAME_COMP)) 612 613 void blk_queue_flag_set(unsigned int flag, struct request_queue *q); 614 void blk_queue_flag_clear(unsigned int flag, struct request_queue *q); 615 bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q); 616 617 #define blk_queue_stopped(q) test_bit(QUEUE_FLAG_STOPPED, &(q)->queue_flags) 618 #define blk_queue_dying(q) test_bit(QUEUE_FLAG_DYING, &(q)->queue_flags) 619 #define blk_queue_dead(q) test_bit(QUEUE_FLAG_DEAD, &(q)->queue_flags) 620 #define blk_queue_init_done(q) test_bit(QUEUE_FLAG_INIT_DONE, &(q)->queue_flags) 621 #define blk_queue_nomerges(q) test_bit(QUEUE_FLAG_NOMERGES, &(q)->queue_flags) 622 #define blk_queue_noxmerges(q) \ 623 test_bit(QUEUE_FLAG_NOXMERGES, &(q)->queue_flags) 624 #define blk_queue_nonrot(q) test_bit(QUEUE_FLAG_NONROT, &(q)->queue_flags) 625 #define blk_queue_io_stat(q) test_bit(QUEUE_FLAG_IO_STAT, &(q)->queue_flags) 626 #define blk_queue_add_random(q) test_bit(QUEUE_FLAG_ADD_RANDOM, &(q)->queue_flags) 627 #define blk_queue_discard(q) test_bit(QUEUE_FLAG_DISCARD, &(q)->queue_flags) 628 #define blk_queue_secure_erase(q) \ 629 (test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags)) 630 #define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags) 631 #define blk_queue_scsi_passthrough(q) \ 632 test_bit(QUEUE_FLAG_SCSI_PASSTHROUGH, &(q)->queue_flags) 633 #define blk_queue_pci_p2pdma(q) \ 634 test_bit(QUEUE_FLAG_PCI_P2PDMA, &(q)->queue_flags) 635 636 #define blk_noretry_request(rq) \ 637 ((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \ 638 REQ_FAILFAST_DRIVER)) 639 #define blk_queue_quiesced(q) test_bit(QUEUE_FLAG_QUIESCED, &(q)->queue_flags) 640 #define blk_queue_pm_only(q) atomic_read(&(q)->pm_only) 641 #define blk_queue_fua(q) test_bit(QUEUE_FLAG_FUA, &(q)->queue_flags) 642 643 extern void blk_set_pm_only(struct request_queue *q); 644 extern void blk_clear_pm_only(struct request_queue *q); 645 646 static inline bool blk_account_rq(struct request *rq) 647 { 648 return (rq->rq_flags & RQF_STARTED) && !blk_rq_is_passthrough(rq); 649 } 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 #define rq_dma_dir(rq) \ 656 (op_is_write(req_op(rq)) ? DMA_TO_DEVICE : DMA_FROM_DEVICE) 657 658 #define dma_map_bvec(dev, bv, dir, attrs) \ 659 dma_map_page_attrs(dev, (bv)->bv_page, (bv)->bv_offset, (bv)->bv_len, \ 660 (dir), (attrs)) 661 662 static inline bool queue_is_mq(struct request_queue *q) 663 { 664 return q->mq_ops; 665 } 666 667 static inline enum blk_zoned_model 668 blk_queue_zoned_model(struct request_queue *q) 669 { 670 return q->limits.zoned; 671 } 672 673 static inline bool blk_queue_is_zoned(struct request_queue *q) 674 { 675 switch (blk_queue_zoned_model(q)) { 676 case BLK_ZONED_HA: 677 case BLK_ZONED_HM: 678 return true; 679 default: 680 return false; 681 } 682 } 683 684 static inline unsigned int blk_queue_zone_sectors(struct request_queue *q) 685 { 686 return blk_queue_is_zoned(q) ? q->limits.chunk_sectors : 0; 687 } 688 689 #ifdef CONFIG_BLK_DEV_ZONED 690 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 691 { 692 return blk_queue_is_zoned(q) ? q->nr_zones : 0; 693 } 694 695 static inline unsigned int blk_queue_zone_no(struct request_queue *q, 696 sector_t sector) 697 { 698 if (!blk_queue_is_zoned(q)) 699 return 0; 700 return sector >> ilog2(q->limits.chunk_sectors); 701 } 702 703 static inline bool blk_queue_zone_is_seq(struct request_queue *q, 704 sector_t sector) 705 { 706 if (!blk_queue_is_zoned(q) || !q->seq_zones_bitmap) 707 return false; 708 return test_bit(blk_queue_zone_no(q, sector), q->seq_zones_bitmap); 709 } 710 #else /* CONFIG_BLK_DEV_ZONED */ 711 static inline unsigned int blk_queue_nr_zones(struct request_queue *q) 712 { 713 return 0; 714 } 715 #endif /* CONFIG_BLK_DEV_ZONED */ 716 717 static inline bool rq_is_sync(struct request *rq) 718 { 719 return op_is_sync(rq->cmd_flags); 720 } 721 722 static inline bool rq_mergeable(struct request *rq) 723 { 724 if (blk_rq_is_passthrough(rq)) 725 return false; 726 727 if (req_op(rq) == REQ_OP_FLUSH) 728 return false; 729 730 if (req_op(rq) == REQ_OP_WRITE_ZEROES) 731 return false; 732 733 if (rq->cmd_flags & REQ_NOMERGE_FLAGS) 734 return false; 735 if (rq->rq_flags & RQF_NOMERGE_FLAGS) 736 return false; 737 738 return true; 739 } 740 741 static inline bool blk_write_same_mergeable(struct bio *a, struct bio *b) 742 { 743 if (bio_page(a) == bio_page(b) && 744 bio_offset(a) == bio_offset(b)) 745 return true; 746 747 return false; 748 } 749 750 static inline unsigned int blk_queue_depth(struct request_queue *q) 751 { 752 if (q->queue_depth) 753 return q->queue_depth; 754 755 return q->nr_requests; 756 } 757 758 extern unsigned long blk_max_low_pfn, blk_max_pfn; 759 760 /* 761 * standard bounce addresses: 762 * 763 * BLK_BOUNCE_HIGH : bounce all highmem pages 764 * BLK_BOUNCE_ANY : don't bounce anything 765 * BLK_BOUNCE_ISA : bounce pages above ISA DMA boundary 766 */ 767 768 #if BITS_PER_LONG == 32 769 #define BLK_BOUNCE_HIGH ((u64)blk_max_low_pfn << PAGE_SHIFT) 770 #else 771 #define BLK_BOUNCE_HIGH -1ULL 772 #endif 773 #define BLK_BOUNCE_ANY (-1ULL) 774 #define BLK_BOUNCE_ISA (DMA_BIT_MASK(24)) 775 776 /* 777 * default timeout for SG_IO if none specified 778 */ 779 #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) 780 #define BLK_MIN_SG_TIMEOUT (7 * HZ) 781 782 struct rq_map_data { 783 struct page **pages; 784 int page_order; 785 int nr_entries; 786 unsigned long offset; 787 int null_mapped; 788 int from_user; 789 }; 790 791 struct req_iterator { 792 struct bvec_iter iter; 793 struct bio *bio; 794 }; 795 796 /* This should not be used directly - use rq_for_each_segment */ 797 #define for_each_bio(_bio) \ 798 for (; _bio; _bio = _bio->bi_next) 799 #define __rq_for_each_bio(_bio, rq) \ 800 if ((rq->bio)) \ 801 for (_bio = (rq)->bio; _bio; _bio = _bio->bi_next) 802 803 #define rq_for_each_segment(bvl, _rq, _iter) \ 804 __rq_for_each_bio(_iter.bio, _rq) \ 805 bio_for_each_segment(bvl, _iter.bio, _iter.iter) 806 807 #define rq_for_each_bvec(bvl, _rq, _iter) \ 808 __rq_for_each_bio(_iter.bio, _rq) \ 809 bio_for_each_bvec(bvl, _iter.bio, _iter.iter) 810 811 #define rq_iter_last(bvec, _iter) \ 812 (_iter.bio->bi_next == NULL && \ 813 bio_iter_last(bvec, _iter.iter)) 814 815 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 816 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 817 #endif 818 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 819 extern void rq_flush_dcache_pages(struct request *rq); 820 #else 821 static inline void rq_flush_dcache_pages(struct request *rq) 822 { 823 } 824 #endif 825 826 extern int blk_register_queue(struct gendisk *disk); 827 extern void blk_unregister_queue(struct gendisk *disk); 828 extern blk_qc_t generic_make_request(struct bio *bio); 829 extern blk_qc_t direct_make_request(struct bio *bio); 830 extern void blk_rq_init(struct request_queue *q, struct request *rq); 831 extern void blk_put_request(struct request *); 832 extern struct request *blk_get_request(struct request_queue *, unsigned int op, 833 blk_mq_req_flags_t flags); 834 extern int blk_lld_busy(struct request_queue *q); 835 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src, 836 struct bio_set *bs, gfp_t gfp_mask, 837 int (*bio_ctr)(struct bio *, struct bio *, void *), 838 void *data); 839 extern void blk_rq_unprep_clone(struct request *rq); 840 extern blk_status_t blk_insert_cloned_request(struct request_queue *q, 841 struct request *rq); 842 extern int blk_rq_append_bio(struct request *rq, struct bio **bio); 843 extern void blk_queue_split(struct request_queue *, struct bio **); 844 extern int scsi_verify_blk_ioctl(struct block_device *, unsigned int); 845 extern int scsi_cmd_blk_ioctl(struct block_device *, fmode_t, 846 unsigned int, void __user *); 847 extern int scsi_cmd_ioctl(struct request_queue *, struct gendisk *, fmode_t, 848 unsigned int, void __user *); 849 extern int sg_scsi_ioctl(struct request_queue *, struct gendisk *, fmode_t, 850 struct scsi_ioctl_command __user *); 851 852 extern int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags); 853 extern void blk_queue_exit(struct request_queue *q); 854 extern void blk_sync_queue(struct request_queue *q); 855 extern int blk_rq_map_user(struct request_queue *, struct request *, 856 struct rq_map_data *, void __user *, unsigned long, 857 gfp_t); 858 extern int blk_rq_unmap_user(struct bio *); 859 extern int blk_rq_map_kern(struct request_queue *, struct request *, void *, unsigned int, gfp_t); 860 extern int blk_rq_map_user_iov(struct request_queue *, struct request *, 861 struct rq_map_data *, const struct iov_iter *, 862 gfp_t); 863 extern void blk_execute_rq(struct request_queue *, struct gendisk *, 864 struct request *, int); 865 extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, 866 struct request *, int, rq_end_io_fn *); 867 868 /* Helper to convert REQ_OP_XXX to its string format XXX */ 869 extern const char *blk_op_str(unsigned int op); 870 871 int blk_status_to_errno(blk_status_t status); 872 blk_status_t errno_to_blk_status(int errno); 873 874 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin); 875 876 static inline struct request_queue *bdev_get_queue(struct block_device *bdev) 877 { 878 return bdev->bd_disk->queue; /* this is never NULL */ 879 } 880 881 /* 882 * The basic unit of block I/O is a sector. It is used in a number of contexts 883 * in Linux (blk, bio, genhd). The size of one sector is 512 = 2**9 884 * bytes. Variables of type sector_t represent an offset or size that is a 885 * multiple of 512 bytes. Hence these two constants. 886 */ 887 #ifndef SECTOR_SHIFT 888 #define SECTOR_SHIFT 9 889 #endif 890 #ifndef SECTOR_SIZE 891 #define SECTOR_SIZE (1 << SECTOR_SHIFT) 892 #endif 893 894 /* 895 * blk_rq_pos() : the current sector 896 * blk_rq_bytes() : bytes left in the entire request 897 * blk_rq_cur_bytes() : bytes left in the current segment 898 * blk_rq_err_bytes() : bytes left till the next error boundary 899 * blk_rq_sectors() : sectors left in the entire request 900 * blk_rq_cur_sectors() : sectors left in the current segment 901 */ 902 static inline sector_t blk_rq_pos(const struct request *rq) 903 { 904 return rq->__sector; 905 } 906 907 static inline unsigned int blk_rq_bytes(const struct request *rq) 908 { 909 return rq->__data_len; 910 } 911 912 static inline int blk_rq_cur_bytes(const struct request *rq) 913 { 914 return rq->bio ? bio_cur_bytes(rq->bio) : 0; 915 } 916 917 extern unsigned int blk_rq_err_bytes(const struct request *rq); 918 919 static inline unsigned int blk_rq_sectors(const struct request *rq) 920 { 921 return blk_rq_bytes(rq) >> SECTOR_SHIFT; 922 } 923 924 static inline unsigned int blk_rq_cur_sectors(const struct request *rq) 925 { 926 return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT; 927 } 928 929 #ifdef CONFIG_BLK_DEV_ZONED 930 static inline unsigned int blk_rq_zone_no(struct request *rq) 931 { 932 return blk_queue_zone_no(rq->q, blk_rq_pos(rq)); 933 } 934 935 static inline unsigned int blk_rq_zone_is_seq(struct request *rq) 936 { 937 return blk_queue_zone_is_seq(rq->q, blk_rq_pos(rq)); 938 } 939 #endif /* CONFIG_BLK_DEV_ZONED */ 940 941 /* 942 * Some commands like WRITE SAME have a payload or data transfer size which 943 * is different from the size of the request. Any driver that supports such 944 * commands using the RQF_SPECIAL_PAYLOAD flag needs to use this helper to 945 * calculate the data transfer size. 946 */ 947 static inline unsigned int blk_rq_payload_bytes(struct request *rq) 948 { 949 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 950 return rq->special_vec.bv_len; 951 return blk_rq_bytes(rq); 952 } 953 954 /* 955 * Return the first full biovec in the request. The caller needs to check that 956 * there are any bvecs before calling this helper. 957 */ 958 static inline struct bio_vec req_bvec(struct request *rq) 959 { 960 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 961 return rq->special_vec; 962 return mp_bvec_iter_bvec(rq->bio->bi_io_vec, rq->bio->bi_iter); 963 } 964 965 static inline unsigned int blk_queue_get_max_sectors(struct request_queue *q, 966 int op) 967 { 968 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) 969 return min(q->limits.max_discard_sectors, 970 UINT_MAX >> SECTOR_SHIFT); 971 972 if (unlikely(op == REQ_OP_WRITE_SAME)) 973 return q->limits.max_write_same_sectors; 974 975 if (unlikely(op == REQ_OP_WRITE_ZEROES)) 976 return q->limits.max_write_zeroes_sectors; 977 978 return q->limits.max_sectors; 979 } 980 981 /* 982 * Return maximum size of a request at given offset. Only valid for 983 * file system requests. 984 */ 985 static inline unsigned int blk_max_size_offset(struct request_queue *q, 986 sector_t offset) 987 { 988 if (!q->limits.chunk_sectors) 989 return q->limits.max_sectors; 990 991 return min(q->limits.max_sectors, (unsigned int)(q->limits.chunk_sectors - 992 (offset & (q->limits.chunk_sectors - 1)))); 993 } 994 995 static inline unsigned int blk_rq_get_max_sectors(struct request *rq, 996 sector_t offset) 997 { 998 struct request_queue *q = rq->q; 999 1000 if (blk_rq_is_passthrough(rq)) 1001 return q->limits.max_hw_sectors; 1002 1003 if (!q->limits.chunk_sectors || 1004 req_op(rq) == REQ_OP_DISCARD || 1005 req_op(rq) == REQ_OP_SECURE_ERASE) 1006 return blk_queue_get_max_sectors(q, req_op(rq)); 1007 1008 return min(blk_max_size_offset(q, offset), 1009 blk_queue_get_max_sectors(q, req_op(rq))); 1010 } 1011 1012 static inline unsigned int blk_rq_count_bios(struct request *rq) 1013 { 1014 unsigned int nr_bios = 0; 1015 struct bio *bio; 1016 1017 __rq_for_each_bio(bio, rq) 1018 nr_bios++; 1019 1020 return nr_bios; 1021 } 1022 1023 void blk_steal_bios(struct bio_list *list, struct request *rq); 1024 1025 /* 1026 * Request completion related functions. 1027 * 1028 * blk_update_request() completes given number of bytes and updates 1029 * the request without completing it. 1030 */ 1031 extern bool blk_update_request(struct request *rq, blk_status_t error, 1032 unsigned int nr_bytes); 1033 1034 extern void __blk_complete_request(struct request *); 1035 extern void blk_abort_request(struct request *); 1036 1037 /* 1038 * Access functions for manipulating queue properties 1039 */ 1040 extern void blk_cleanup_queue(struct request_queue *); 1041 extern void blk_queue_make_request(struct request_queue *, make_request_fn *); 1042 extern void blk_queue_bounce_limit(struct request_queue *, u64); 1043 extern void blk_queue_max_hw_sectors(struct request_queue *, unsigned int); 1044 extern void blk_queue_chunk_sectors(struct request_queue *, unsigned int); 1045 extern void blk_queue_max_segments(struct request_queue *, unsigned short); 1046 extern void blk_queue_max_discard_segments(struct request_queue *, 1047 unsigned short); 1048 extern void blk_queue_max_segment_size(struct request_queue *, unsigned int); 1049 extern void blk_queue_max_discard_sectors(struct request_queue *q, 1050 unsigned int max_discard_sectors); 1051 extern void blk_queue_max_write_same_sectors(struct request_queue *q, 1052 unsigned int max_write_same_sectors); 1053 extern void blk_queue_max_write_zeroes_sectors(struct request_queue *q, 1054 unsigned int max_write_same_sectors); 1055 extern void blk_queue_logical_block_size(struct request_queue *, unsigned short); 1056 extern void blk_queue_physical_block_size(struct request_queue *, unsigned int); 1057 extern void blk_queue_alignment_offset(struct request_queue *q, 1058 unsigned int alignment); 1059 extern void blk_limits_io_min(struct queue_limits *limits, unsigned int min); 1060 extern void blk_queue_io_min(struct request_queue *q, unsigned int min); 1061 extern void blk_limits_io_opt(struct queue_limits *limits, unsigned int opt); 1062 extern void blk_queue_io_opt(struct request_queue *q, unsigned int opt); 1063 extern void blk_set_queue_depth(struct request_queue *q, unsigned int depth); 1064 extern void blk_set_default_limits(struct queue_limits *lim); 1065 extern void blk_set_stacking_limits(struct queue_limits *lim); 1066 extern int blk_stack_limits(struct queue_limits *t, struct queue_limits *b, 1067 sector_t offset); 1068 extern int bdev_stack_limits(struct queue_limits *t, struct block_device *bdev, 1069 sector_t offset); 1070 extern void disk_stack_limits(struct gendisk *disk, struct block_device *bdev, 1071 sector_t offset); 1072 extern void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b); 1073 extern void blk_queue_update_dma_pad(struct request_queue *, unsigned int); 1074 extern int blk_queue_dma_drain(struct request_queue *q, 1075 dma_drain_needed_fn *dma_drain_needed, 1076 void *buf, unsigned int size); 1077 extern void blk_queue_segment_boundary(struct request_queue *, unsigned long); 1078 extern void blk_queue_virt_boundary(struct request_queue *, unsigned long); 1079 extern void blk_queue_dma_alignment(struct request_queue *, int); 1080 extern void blk_queue_update_dma_alignment(struct request_queue *, int); 1081 extern void blk_queue_rq_timeout(struct request_queue *, unsigned int); 1082 extern void blk_queue_write_cache(struct request_queue *q, bool enabled, bool fua); 1083 1084 /* 1085 * Number of physical segments as sent to the device. 1086 * 1087 * Normally this is the number of discontiguous data segments sent by the 1088 * submitter. But for data-less command like discard we might have no 1089 * actual data segments submitted, but the driver might have to add it's 1090 * own special payload. In that case we still return 1 here so that this 1091 * special payload will be mapped. 1092 */ 1093 static inline unsigned short blk_rq_nr_phys_segments(struct request *rq) 1094 { 1095 if (rq->rq_flags & RQF_SPECIAL_PAYLOAD) 1096 return 1; 1097 return rq->nr_phys_segments; 1098 } 1099 1100 /* 1101 * Number of discard segments (or ranges) the driver needs to fill in. 1102 * Each discard bio merged into a request is counted as one segment. 1103 */ 1104 static inline unsigned short blk_rq_nr_discard_segments(struct request *rq) 1105 { 1106 return max_t(unsigned short, rq->nr_phys_segments, 1); 1107 } 1108 1109 extern int blk_rq_map_sg(struct request_queue *, struct request *, struct scatterlist *); 1110 extern void blk_dump_rq_flags(struct request *, char *); 1111 extern long nr_blockdev_pages(void); 1112 1113 bool __must_check blk_get_queue(struct request_queue *); 1114 struct request_queue *blk_alloc_queue(gfp_t); 1115 struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id); 1116 extern void blk_put_queue(struct request_queue *); 1117 extern void blk_set_queue_dying(struct request_queue *); 1118 1119 /* 1120 * blk_plug permits building a queue of related requests by holding the I/O 1121 * fragments for a short period. This allows merging of sequential requests 1122 * into single larger request. As the requests are moved from a per-task list to 1123 * the device's request_queue in a batch, this results in improved scalability 1124 * as the lock contention for request_queue lock is reduced. 1125 * 1126 * It is ok not to disable preemption when adding the request to the plug list 1127 * or when attempting a merge, because blk_schedule_flush_list() will only flush 1128 * the plug list when the task sleeps by itself. For details, please see 1129 * schedule() where blk_schedule_flush_plug() is called. 1130 */ 1131 struct blk_plug { 1132 struct list_head mq_list; /* blk-mq requests */ 1133 struct list_head cb_list; /* md requires an unplug callback */ 1134 unsigned short rq_count; 1135 bool multiple_queues; 1136 }; 1137 #define BLK_MAX_REQUEST_COUNT 16 1138 #define BLK_PLUG_FLUSH_SIZE (128 * 1024) 1139 1140 struct blk_plug_cb; 1141 typedef void (*blk_plug_cb_fn)(struct blk_plug_cb *, bool); 1142 struct blk_plug_cb { 1143 struct list_head list; 1144 blk_plug_cb_fn callback; 1145 void *data; 1146 }; 1147 extern struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, 1148 void *data, int size); 1149 extern void blk_start_plug(struct blk_plug *); 1150 extern void blk_finish_plug(struct blk_plug *); 1151 extern void blk_flush_plug_list(struct blk_plug *, bool); 1152 1153 static inline void blk_flush_plug(struct task_struct *tsk) 1154 { 1155 struct blk_plug *plug = tsk->plug; 1156 1157 if (plug) 1158 blk_flush_plug_list(plug, false); 1159 } 1160 1161 static inline void blk_schedule_flush_plug(struct task_struct *tsk) 1162 { 1163 struct blk_plug *plug = tsk->plug; 1164 1165 if (plug) 1166 blk_flush_plug_list(plug, true); 1167 } 1168 1169 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1170 { 1171 struct blk_plug *plug = tsk->plug; 1172 1173 return plug && 1174 (!list_empty(&plug->mq_list) || 1175 !list_empty(&plug->cb_list)); 1176 } 1177 1178 extern int blkdev_issue_flush(struct block_device *, gfp_t, sector_t *); 1179 extern int blkdev_issue_write_same(struct block_device *bdev, sector_t sector, 1180 sector_t nr_sects, gfp_t gfp_mask, struct page *page); 1181 1182 #define BLKDEV_DISCARD_SECURE (1 << 0) /* issue a secure erase */ 1183 1184 extern int blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1185 sector_t nr_sects, gfp_t gfp_mask, unsigned long flags); 1186 extern int __blkdev_issue_discard(struct block_device *bdev, sector_t sector, 1187 sector_t nr_sects, gfp_t gfp_mask, int flags, 1188 struct bio **biop); 1189 1190 #define BLKDEV_ZERO_NOUNMAP (1 << 0) /* do not free blocks */ 1191 #define BLKDEV_ZERO_NOFALLBACK (1 << 1) /* don't write explicit zeroes */ 1192 1193 extern int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1194 sector_t nr_sects, gfp_t gfp_mask, struct bio **biop, 1195 unsigned flags); 1196 extern int blkdev_issue_zeroout(struct block_device *bdev, sector_t sector, 1197 sector_t nr_sects, gfp_t gfp_mask, unsigned flags); 1198 1199 static inline int sb_issue_discard(struct super_block *sb, sector_t block, 1200 sector_t nr_blocks, gfp_t gfp_mask, unsigned long flags) 1201 { 1202 return blkdev_issue_discard(sb->s_bdev, 1203 block << (sb->s_blocksize_bits - 1204 SECTOR_SHIFT), 1205 nr_blocks << (sb->s_blocksize_bits - 1206 SECTOR_SHIFT), 1207 gfp_mask, flags); 1208 } 1209 static inline int sb_issue_zeroout(struct super_block *sb, sector_t block, 1210 sector_t nr_blocks, gfp_t gfp_mask) 1211 { 1212 return blkdev_issue_zeroout(sb->s_bdev, 1213 block << (sb->s_blocksize_bits - 1214 SECTOR_SHIFT), 1215 nr_blocks << (sb->s_blocksize_bits - 1216 SECTOR_SHIFT), 1217 gfp_mask, 0); 1218 } 1219 1220 extern int blk_verify_command(unsigned char *cmd, fmode_t mode); 1221 1222 enum blk_default_limits { 1223 BLK_MAX_SEGMENTS = 128, 1224 BLK_SAFE_MAX_SECTORS = 255, 1225 BLK_DEF_MAX_SECTORS = 2560, 1226 BLK_MAX_SEGMENT_SIZE = 65536, 1227 BLK_SEG_BOUNDARY_MASK = 0xFFFFFFFFUL, 1228 }; 1229 1230 static inline unsigned long queue_segment_boundary(struct request_queue *q) 1231 { 1232 return q->limits.seg_boundary_mask; 1233 } 1234 1235 static inline unsigned long queue_virt_boundary(struct request_queue *q) 1236 { 1237 return q->limits.virt_boundary_mask; 1238 } 1239 1240 static inline unsigned int queue_max_sectors(struct request_queue *q) 1241 { 1242 return q->limits.max_sectors; 1243 } 1244 1245 static inline unsigned int queue_max_hw_sectors(struct request_queue *q) 1246 { 1247 return q->limits.max_hw_sectors; 1248 } 1249 1250 static inline unsigned short queue_max_segments(struct request_queue *q) 1251 { 1252 return q->limits.max_segments; 1253 } 1254 1255 static inline unsigned short queue_max_discard_segments(struct request_queue *q) 1256 { 1257 return q->limits.max_discard_segments; 1258 } 1259 1260 static inline unsigned int queue_max_segment_size(struct request_queue *q) 1261 { 1262 return q->limits.max_segment_size; 1263 } 1264 1265 static inline unsigned short queue_logical_block_size(struct request_queue *q) 1266 { 1267 int retval = 512; 1268 1269 if (q && q->limits.logical_block_size) 1270 retval = q->limits.logical_block_size; 1271 1272 return retval; 1273 } 1274 1275 static inline unsigned short bdev_logical_block_size(struct block_device *bdev) 1276 { 1277 return queue_logical_block_size(bdev_get_queue(bdev)); 1278 } 1279 1280 static inline unsigned int queue_physical_block_size(struct request_queue *q) 1281 { 1282 return q->limits.physical_block_size; 1283 } 1284 1285 static inline unsigned int bdev_physical_block_size(struct block_device *bdev) 1286 { 1287 return queue_physical_block_size(bdev_get_queue(bdev)); 1288 } 1289 1290 static inline unsigned int queue_io_min(struct request_queue *q) 1291 { 1292 return q->limits.io_min; 1293 } 1294 1295 static inline int bdev_io_min(struct block_device *bdev) 1296 { 1297 return queue_io_min(bdev_get_queue(bdev)); 1298 } 1299 1300 static inline unsigned int queue_io_opt(struct request_queue *q) 1301 { 1302 return q->limits.io_opt; 1303 } 1304 1305 static inline int bdev_io_opt(struct block_device *bdev) 1306 { 1307 return queue_io_opt(bdev_get_queue(bdev)); 1308 } 1309 1310 static inline int queue_alignment_offset(struct request_queue *q) 1311 { 1312 if (q->limits.misaligned) 1313 return -1; 1314 1315 return q->limits.alignment_offset; 1316 } 1317 1318 static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t sector) 1319 { 1320 unsigned int granularity = max(lim->physical_block_size, lim->io_min); 1321 unsigned int alignment = sector_div(sector, granularity >> SECTOR_SHIFT) 1322 << SECTOR_SHIFT; 1323 1324 return (granularity + lim->alignment_offset - alignment) % granularity; 1325 } 1326 1327 static inline int bdev_alignment_offset(struct block_device *bdev) 1328 { 1329 struct request_queue *q = bdev_get_queue(bdev); 1330 1331 if (q->limits.misaligned) 1332 return -1; 1333 1334 if (bdev != bdev->bd_contains) 1335 return bdev->bd_part->alignment_offset; 1336 1337 return q->limits.alignment_offset; 1338 } 1339 1340 static inline int queue_discard_alignment(struct request_queue *q) 1341 { 1342 if (q->limits.discard_misaligned) 1343 return -1; 1344 1345 return q->limits.discard_alignment; 1346 } 1347 1348 static inline int queue_limit_discard_alignment(struct queue_limits *lim, sector_t sector) 1349 { 1350 unsigned int alignment, granularity, offset; 1351 1352 if (!lim->max_discard_sectors) 1353 return 0; 1354 1355 /* Why are these in bytes, not sectors? */ 1356 alignment = lim->discard_alignment >> SECTOR_SHIFT; 1357 granularity = lim->discard_granularity >> SECTOR_SHIFT; 1358 if (!granularity) 1359 return 0; 1360 1361 /* Offset of the partition start in 'granularity' sectors */ 1362 offset = sector_div(sector, granularity); 1363 1364 /* And why do we do this modulus *again* in blkdev_issue_discard()? */ 1365 offset = (granularity + alignment - offset) % granularity; 1366 1367 /* Turn it back into bytes, gaah */ 1368 return offset << SECTOR_SHIFT; 1369 } 1370 1371 static inline int bdev_discard_alignment(struct block_device *bdev) 1372 { 1373 struct request_queue *q = bdev_get_queue(bdev); 1374 1375 if (bdev != bdev->bd_contains) 1376 return bdev->bd_part->discard_alignment; 1377 1378 return q->limits.discard_alignment; 1379 } 1380 1381 static inline unsigned int bdev_write_same(struct block_device *bdev) 1382 { 1383 struct request_queue *q = bdev_get_queue(bdev); 1384 1385 if (q) 1386 return q->limits.max_write_same_sectors; 1387 1388 return 0; 1389 } 1390 1391 static inline unsigned int bdev_write_zeroes_sectors(struct block_device *bdev) 1392 { 1393 struct request_queue *q = bdev_get_queue(bdev); 1394 1395 if (q) 1396 return q->limits.max_write_zeroes_sectors; 1397 1398 return 0; 1399 } 1400 1401 static inline enum blk_zoned_model bdev_zoned_model(struct block_device *bdev) 1402 { 1403 struct request_queue *q = bdev_get_queue(bdev); 1404 1405 if (q) 1406 return blk_queue_zoned_model(q); 1407 1408 return BLK_ZONED_NONE; 1409 } 1410 1411 static inline bool bdev_is_zoned(struct block_device *bdev) 1412 { 1413 struct request_queue *q = bdev_get_queue(bdev); 1414 1415 if (q) 1416 return blk_queue_is_zoned(q); 1417 1418 return false; 1419 } 1420 1421 static inline unsigned int bdev_zone_sectors(struct block_device *bdev) 1422 { 1423 struct request_queue *q = bdev_get_queue(bdev); 1424 1425 if (q) 1426 return blk_queue_zone_sectors(q); 1427 return 0; 1428 } 1429 1430 static inline int queue_dma_alignment(struct request_queue *q) 1431 { 1432 return q ? q->dma_alignment : 511; 1433 } 1434 1435 static inline int blk_rq_aligned(struct request_queue *q, unsigned long addr, 1436 unsigned int len) 1437 { 1438 unsigned int alignment = queue_dma_alignment(q) | q->dma_pad_mask; 1439 return !(addr & alignment) && !(len & alignment); 1440 } 1441 1442 /* assumes size > 256 */ 1443 static inline unsigned int blksize_bits(unsigned int size) 1444 { 1445 unsigned int bits = 8; 1446 do { 1447 bits++; 1448 size >>= 1; 1449 } while (size > 256); 1450 return bits; 1451 } 1452 1453 static inline unsigned int block_size(struct block_device *bdev) 1454 { 1455 return bdev->bd_block_size; 1456 } 1457 1458 typedef struct {struct page *v;} Sector; 1459 1460 unsigned char *read_dev_sector(struct block_device *, sector_t, Sector *); 1461 1462 static inline void put_dev_sector(Sector p) 1463 { 1464 put_page(p.v); 1465 } 1466 1467 int kblockd_schedule_work(struct work_struct *work); 1468 int kblockd_schedule_work_on(int cpu, struct work_struct *work); 1469 int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, unsigned long delay); 1470 1471 #define MODULE_ALIAS_BLOCKDEV(major,minor) \ 1472 MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor)) 1473 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \ 1474 MODULE_ALIAS("block-major-" __stringify(major) "-*") 1475 1476 #if defined(CONFIG_BLK_DEV_INTEGRITY) 1477 1478 enum blk_integrity_flags { 1479 BLK_INTEGRITY_VERIFY = 1 << 0, 1480 BLK_INTEGRITY_GENERATE = 1 << 1, 1481 BLK_INTEGRITY_DEVICE_CAPABLE = 1 << 2, 1482 BLK_INTEGRITY_IP_CHECKSUM = 1 << 3, 1483 }; 1484 1485 struct blk_integrity_iter { 1486 void *prot_buf; 1487 void *data_buf; 1488 sector_t seed; 1489 unsigned int data_size; 1490 unsigned short interval; 1491 const char *disk_name; 1492 }; 1493 1494 typedef blk_status_t (integrity_processing_fn) (struct blk_integrity_iter *); 1495 1496 struct blk_integrity_profile { 1497 integrity_processing_fn *generate_fn; 1498 integrity_processing_fn *verify_fn; 1499 const char *name; 1500 }; 1501 1502 extern void blk_integrity_register(struct gendisk *, struct blk_integrity *); 1503 extern void blk_integrity_unregister(struct gendisk *); 1504 extern int blk_integrity_compare(struct gendisk *, struct gendisk *); 1505 extern int blk_rq_map_integrity_sg(struct request_queue *, struct bio *, 1506 struct scatterlist *); 1507 extern int blk_rq_count_integrity_sg(struct request_queue *, struct bio *); 1508 extern bool blk_integrity_merge_rq(struct request_queue *, struct request *, 1509 struct request *); 1510 extern bool blk_integrity_merge_bio(struct request_queue *, struct request *, 1511 struct bio *); 1512 1513 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1514 { 1515 struct blk_integrity *bi = &disk->queue->integrity; 1516 1517 if (!bi->profile) 1518 return NULL; 1519 1520 return bi; 1521 } 1522 1523 static inline 1524 struct blk_integrity *bdev_get_integrity(struct block_device *bdev) 1525 { 1526 return blk_get_integrity(bdev->bd_disk); 1527 } 1528 1529 static inline bool blk_integrity_rq(struct request *rq) 1530 { 1531 return rq->cmd_flags & REQ_INTEGRITY; 1532 } 1533 1534 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1535 unsigned int segs) 1536 { 1537 q->limits.max_integrity_segments = segs; 1538 } 1539 1540 static inline unsigned short 1541 queue_max_integrity_segments(struct request_queue *q) 1542 { 1543 return q->limits.max_integrity_segments; 1544 } 1545 1546 /** 1547 * bio_integrity_intervals - Return number of integrity intervals for a bio 1548 * @bi: blk_integrity profile for device 1549 * @sectors: Size of the bio in 512-byte sectors 1550 * 1551 * Description: The block layer calculates everything in 512 byte 1552 * sectors but integrity metadata is done in terms of the data integrity 1553 * interval size of the storage device. Convert the block layer sectors 1554 * to the appropriate number of integrity intervals. 1555 */ 1556 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1557 unsigned int sectors) 1558 { 1559 return sectors >> (bi->interval_exp - 9); 1560 } 1561 1562 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1563 unsigned int sectors) 1564 { 1565 return bio_integrity_intervals(bi, sectors) * bi->tuple_size; 1566 } 1567 1568 /* 1569 * Return the first bvec that contains integrity data. Only drivers that are 1570 * limited to a single integrity segment should use this helper. 1571 */ 1572 static inline struct bio_vec *rq_integrity_vec(struct request *rq) 1573 { 1574 if (WARN_ON_ONCE(queue_max_integrity_segments(rq->q) > 1)) 1575 return NULL; 1576 return rq->bio->bi_integrity->bip_vec; 1577 } 1578 1579 #else /* CONFIG_BLK_DEV_INTEGRITY */ 1580 1581 struct bio; 1582 struct block_device; 1583 struct gendisk; 1584 struct blk_integrity; 1585 1586 static inline int blk_integrity_rq(struct request *rq) 1587 { 1588 return 0; 1589 } 1590 static inline int blk_rq_count_integrity_sg(struct request_queue *q, 1591 struct bio *b) 1592 { 1593 return 0; 1594 } 1595 static inline int blk_rq_map_integrity_sg(struct request_queue *q, 1596 struct bio *b, 1597 struct scatterlist *s) 1598 { 1599 return 0; 1600 } 1601 static inline struct blk_integrity *bdev_get_integrity(struct block_device *b) 1602 { 1603 return NULL; 1604 } 1605 static inline struct blk_integrity *blk_get_integrity(struct gendisk *disk) 1606 { 1607 return NULL; 1608 } 1609 static inline int blk_integrity_compare(struct gendisk *a, struct gendisk *b) 1610 { 1611 return 0; 1612 } 1613 static inline void blk_integrity_register(struct gendisk *d, 1614 struct blk_integrity *b) 1615 { 1616 } 1617 static inline void blk_integrity_unregister(struct gendisk *d) 1618 { 1619 } 1620 static inline void blk_queue_max_integrity_segments(struct request_queue *q, 1621 unsigned int segs) 1622 { 1623 } 1624 static inline unsigned short queue_max_integrity_segments(struct request_queue *q) 1625 { 1626 return 0; 1627 } 1628 static inline bool blk_integrity_merge_rq(struct request_queue *rq, 1629 struct request *r1, 1630 struct request *r2) 1631 { 1632 return true; 1633 } 1634 static inline bool blk_integrity_merge_bio(struct request_queue *rq, 1635 struct request *r, 1636 struct bio *b) 1637 { 1638 return true; 1639 } 1640 1641 static inline unsigned int bio_integrity_intervals(struct blk_integrity *bi, 1642 unsigned int sectors) 1643 { 1644 return 0; 1645 } 1646 1647 static inline unsigned int bio_integrity_bytes(struct blk_integrity *bi, 1648 unsigned int sectors) 1649 { 1650 return 0; 1651 } 1652 1653 static inline struct bio_vec *rq_integrity_vec(struct request *rq) 1654 { 1655 return NULL; 1656 } 1657 1658 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 1659 1660 struct block_device_operations { 1661 int (*open) (struct block_device *, fmode_t); 1662 void (*release) (struct gendisk *, fmode_t); 1663 int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int); 1664 int (*ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1665 int (*compat_ioctl) (struct block_device *, fmode_t, unsigned, unsigned long); 1666 unsigned int (*check_events) (struct gendisk *disk, 1667 unsigned int clearing); 1668 /* ->media_changed() is DEPRECATED, use ->check_events() instead */ 1669 int (*media_changed) (struct gendisk *); 1670 void (*unlock_native_capacity) (struct gendisk *); 1671 int (*revalidate_disk) (struct gendisk *); 1672 int (*getgeo)(struct block_device *, struct hd_geometry *); 1673 /* this callback is with swap_lock and sometimes page table lock held */ 1674 void (*swap_slot_free_notify) (struct block_device *, unsigned long); 1675 int (*report_zones)(struct gendisk *, sector_t sector, 1676 struct blk_zone *zones, unsigned int *nr_zones, 1677 gfp_t gfp_mask); 1678 struct module *owner; 1679 const struct pr_ops *pr_ops; 1680 }; 1681 1682 extern int __blkdev_driver_ioctl(struct block_device *, fmode_t, unsigned int, 1683 unsigned long); 1684 extern int bdev_read_page(struct block_device *, sector_t, struct page *); 1685 extern int bdev_write_page(struct block_device *, sector_t, struct page *, 1686 struct writeback_control *); 1687 1688 #ifdef CONFIG_BLK_DEV_ZONED 1689 bool blk_req_needs_zone_write_lock(struct request *rq); 1690 void __blk_req_zone_write_lock(struct request *rq); 1691 void __blk_req_zone_write_unlock(struct request *rq); 1692 1693 static inline void blk_req_zone_write_lock(struct request *rq) 1694 { 1695 if (blk_req_needs_zone_write_lock(rq)) 1696 __blk_req_zone_write_lock(rq); 1697 } 1698 1699 static inline void blk_req_zone_write_unlock(struct request *rq) 1700 { 1701 if (rq->rq_flags & RQF_ZONE_WRITE_LOCKED) 1702 __blk_req_zone_write_unlock(rq); 1703 } 1704 1705 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1706 { 1707 return rq->q->seq_zones_wlock && 1708 test_bit(blk_rq_zone_no(rq), rq->q->seq_zones_wlock); 1709 } 1710 1711 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1712 { 1713 if (!blk_req_needs_zone_write_lock(rq)) 1714 return true; 1715 return !blk_req_zone_is_write_locked(rq); 1716 } 1717 #else 1718 static inline bool blk_req_needs_zone_write_lock(struct request *rq) 1719 { 1720 return false; 1721 } 1722 1723 static inline void blk_req_zone_write_lock(struct request *rq) 1724 { 1725 } 1726 1727 static inline void blk_req_zone_write_unlock(struct request *rq) 1728 { 1729 } 1730 static inline bool blk_req_zone_is_write_locked(struct request *rq) 1731 { 1732 return false; 1733 } 1734 1735 static inline bool blk_req_can_dispatch_to_zone(struct request *rq) 1736 { 1737 return true; 1738 } 1739 #endif /* CONFIG_BLK_DEV_ZONED */ 1740 1741 #else /* CONFIG_BLOCK */ 1742 1743 struct block_device; 1744 1745 /* 1746 * stubs for when the block layer is configured out 1747 */ 1748 #define buffer_heads_over_limit 0 1749 1750 static inline long nr_blockdev_pages(void) 1751 { 1752 return 0; 1753 } 1754 1755 struct blk_plug { 1756 }; 1757 1758 static inline void blk_start_plug(struct blk_plug *plug) 1759 { 1760 } 1761 1762 static inline void blk_finish_plug(struct blk_plug *plug) 1763 { 1764 } 1765 1766 static inline void blk_flush_plug(struct task_struct *task) 1767 { 1768 } 1769 1770 static inline void blk_schedule_flush_plug(struct task_struct *task) 1771 { 1772 } 1773 1774 1775 static inline bool blk_needs_flush_plug(struct task_struct *tsk) 1776 { 1777 return false; 1778 } 1779 1780 static inline int blkdev_issue_flush(struct block_device *bdev, gfp_t gfp_mask, 1781 sector_t *error_sector) 1782 { 1783 return 0; 1784 } 1785 1786 #endif /* CONFIG_BLOCK */ 1787 1788 static inline void blk_wake_io_task(struct task_struct *waiter) 1789 { 1790 /* 1791 * If we're polling, the task itself is doing the completions. For 1792 * that case, we don't need to signal a wakeup, it's enough to just 1793 * mark us as RUNNING. 1794 */ 1795 if (waiter == current) 1796 __set_current_state(TASK_RUNNING); 1797 else 1798 wake_up_process(waiter); 1799 } 1800 1801 #endif 1802