| 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
| 2 | #ifndef BLK_INTERNAL_H |
| 3 | #define BLK_INTERNAL_H |
| 4 | |
| 5 | #include <linux/bio-integrity.h> |
| 6 | #include <linux/blk-crypto.h> |
| 7 | #include <linux/lockdep.h> |
| 8 | #include <linux/memblock.h> /* for max_pfn/max_low_pfn */ |
| 9 | #include <linux/sched/sysctl.h> |
| 10 | #include <linux/timekeeping.h> |
| 11 | #include <xen/xen.h> |
| 12 | #include "blk-crypto-internal.h" |
| 13 | |
| 14 | struct elevator_type; |
| 15 | |
| 16 | #define BLK_DEV_MAX_SECTORS (LLONG_MAX >> 9) |
| 17 | #define BLK_MIN_SEGMENT_SIZE 4096 |
| 18 | |
| 19 | /* Max future timer expiry for timeouts */ |
| 20 | #define BLK_MAX_TIMEOUT (5 * HZ) |
| 21 | |
| 22 | extern struct dentry *blk_debugfs_root; |
| 23 | |
| 24 | struct blk_flush_queue { |
| 25 | spinlock_t mq_flush_lock; |
| 26 | unsigned int flush_pending_idx:1; |
| 27 | unsigned int flush_running_idx:1; |
| 28 | blk_status_t rq_status; |
| 29 | unsigned long flush_pending_since; |
| 30 | struct list_head flush_queue[2]; |
| 31 | unsigned long flush_data_in_flight; |
| 32 | struct request *flush_rq; |
| 33 | }; |
| 34 | |
| 35 | bool is_flush_rq(struct request *req); |
| 36 | |
| 37 | struct blk_flush_queue *blk_alloc_flush_queue(int node, int cmd_size, |
| 38 | gfp_t flags); |
| 39 | void blk_free_flush_queue(struct blk_flush_queue *q); |
| 40 | |
| 41 | bool __blk_mq_unfreeze_queue(struct request_queue *q, bool force_atomic); |
| 42 | bool blk_queue_start_drain(struct request_queue *q); |
| 43 | bool __blk_freeze_queue_start(struct request_queue *q, |
| 44 | struct task_struct *owner); |
| 45 | int __bio_queue_enter(struct request_queue *q, struct bio *bio); |
| 46 | void submit_bio_noacct_nocheck(struct bio *bio); |
| 47 | void bio_await_chain(struct bio *bio); |
| 48 | |
| 49 | static inline bool blk_try_enter_queue(struct request_queue *q, bool pm) |
| 50 | { |
| 51 | rcu_read_lock(); |
| 52 | if (!percpu_ref_tryget_live_rcu(&q->q_usage_counter)) |
| 53 | goto fail; |
| 54 | |
| 55 | /* |
| 56 | * The code that increments the pm_only counter must ensure that the |
| 57 | * counter is globally visible before the queue is unfrozen. |
| 58 | */ |
| 59 | if (blk_queue_pm_only(q) && |
| 60 | (!pm || queue_rpm_status(q) == RPM_SUSPENDED)) |
| 61 | goto fail_put; |
| 62 | |
| 63 | rcu_read_unlock(); |
| 64 | return true; |
| 65 | |
| 66 | fail_put: |
| 67 | blk_queue_exit(q); |
| 68 | fail: |
| 69 | rcu_read_unlock(); |
| 70 | return false; |
| 71 | } |
| 72 | |
| 73 | static inline int bio_queue_enter(struct bio *bio) |
| 74 | { |
| 75 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
| 76 | |
| 77 | if (blk_try_enter_queue(q, false)) { |
| 78 | rwsem_acquire_read(&q->io_lockdep_map, 0, 0, _RET_IP_); |
| 79 | rwsem_release(&q->io_lockdep_map, _RET_IP_); |
| 80 | return 0; |
| 81 | } |
| 82 | return __bio_queue_enter(q, bio); |
| 83 | } |
| 84 | |
| 85 | static inline void blk_wait_io(struct completion *done) |
| 86 | { |
| 87 | /* Prevent hang_check timer from firing at us during very long I/O */ |
| 88 | unsigned long timeout = sysctl_hung_task_timeout_secs * HZ / 2; |
| 89 | |
| 90 | if (timeout) |
| 91 | while (!wait_for_completion_io_timeout(done, timeout)) |
| 92 | ; |
| 93 | else |
| 94 | wait_for_completion_io(done); |
| 95 | } |
| 96 | |
| 97 | struct block_device *blkdev_get_no_open(dev_t dev, bool autoload); |
| 98 | void blkdev_put_no_open(struct block_device *bdev); |
| 99 | |
| 100 | #define BIO_INLINE_VECS 4 |
| 101 | struct bio_vec *bvec_alloc(mempool_t *pool, unsigned short *nr_vecs, |
| 102 | gfp_t gfp_mask); |
| 103 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned short nr_vecs); |
| 104 | |
| 105 | bool bvec_try_merge_hw_page(struct request_queue *q, struct bio_vec *bv, |
| 106 | struct page *page, unsigned len, unsigned offset); |
| 107 | |
| 108 | static inline bool biovec_phys_mergeable(struct request_queue *q, |
| 109 | struct bio_vec *vec1, struct bio_vec *vec2) |
| 110 | { |
| 111 | unsigned long mask = queue_segment_boundary(q); |
| 112 | phys_addr_t addr1 = bvec_phys(vec1); |
| 113 | phys_addr_t addr2 = bvec_phys(vec2); |
| 114 | |
| 115 | /* |
| 116 | * Merging adjacent physical pages may not work correctly under KMSAN |
| 117 | * if their metadata pages aren't adjacent. Just disable merging. |
| 118 | */ |
| 119 | if (IS_ENABLED(CONFIG_KMSAN)) |
| 120 | return false; |
| 121 | |
| 122 | if (addr1 + vec1->bv_len != addr2) |
| 123 | return false; |
| 124 | if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page)) |
| 125 | return false; |
| 126 | if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask)) |
| 127 | return false; |
| 128 | return true; |
| 129 | } |
| 130 | |
| 131 | static inline bool __bvec_gap_to_prev(const struct queue_limits *lim, |
| 132 | struct bio_vec *bprv, unsigned int offset) |
| 133 | { |
| 134 | return (offset & lim->virt_boundary_mask) || |
| 135 | ((bprv->bv_offset + bprv->bv_len) & lim->virt_boundary_mask); |
| 136 | } |
| 137 | |
| 138 | /* |
| 139 | * Check if adding a bio_vec after bprv with offset would create a gap in |
| 140 | * the SG list. Most drivers don't care about this, but some do. |
| 141 | */ |
| 142 | static inline bool bvec_gap_to_prev(const struct queue_limits *lim, |
| 143 | struct bio_vec *bprv, unsigned int offset) |
| 144 | { |
| 145 | if (!lim->virt_boundary_mask) |
| 146 | return false; |
| 147 | return __bvec_gap_to_prev(lim, bprv, offset); |
| 148 | } |
| 149 | |
| 150 | static inline bool rq_mergeable(struct request *rq) |
| 151 | { |
| 152 | if (blk_rq_is_passthrough(rq)) |
| 153 | return false; |
| 154 | |
| 155 | if (req_op(rq) == REQ_OP_FLUSH) |
| 156 | return false; |
| 157 | |
| 158 | if (req_op(rq) == REQ_OP_WRITE_ZEROES) |
| 159 | return false; |
| 160 | |
| 161 | if (req_op(rq) == REQ_OP_ZONE_APPEND) |
| 162 | return false; |
| 163 | |
| 164 | if (rq->cmd_flags & REQ_NOMERGE_FLAGS) |
| 165 | return false; |
| 166 | if (rq->rq_flags & RQF_NOMERGE_FLAGS) |
| 167 | return false; |
| 168 | |
| 169 | return true; |
| 170 | } |
| 171 | |
| 172 | /* |
| 173 | * There are two different ways to handle DISCARD merges: |
| 174 | * 1) If max_discard_segments > 1, the driver treats every bio as a range and |
| 175 | * send the bios to controller together. The ranges don't need to be |
| 176 | * contiguous. |
| 177 | * 2) Otherwise, the request will be normal read/write requests. The ranges |
| 178 | * need to be contiguous. |
| 179 | */ |
| 180 | static inline bool blk_discard_mergable(struct request *req) |
| 181 | { |
| 182 | if (req_op(req) == REQ_OP_DISCARD && |
| 183 | queue_max_discard_segments(req->q) > 1) |
| 184 | return true; |
| 185 | return false; |
| 186 | } |
| 187 | |
| 188 | static inline unsigned int blk_rq_get_max_segments(struct request *rq) |
| 189 | { |
| 190 | if (req_op(rq) == REQ_OP_DISCARD) |
| 191 | return queue_max_discard_segments(rq->q); |
| 192 | return queue_max_segments(rq->q); |
| 193 | } |
| 194 | |
| 195 | static inline unsigned int blk_queue_get_max_sectors(struct request *rq) |
| 196 | { |
| 197 | struct request_queue *q = rq->q; |
| 198 | enum req_op op = req_op(rq); |
| 199 | |
| 200 | if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE)) |
| 201 | return min(q->limits.max_discard_sectors, |
| 202 | UINT_MAX >> SECTOR_SHIFT); |
| 203 | |
| 204 | if (unlikely(op == REQ_OP_WRITE_ZEROES)) |
| 205 | return q->limits.max_write_zeroes_sectors; |
| 206 | |
| 207 | if (rq->cmd_flags & REQ_ATOMIC) |
| 208 | return q->limits.atomic_write_max_sectors; |
| 209 | |
| 210 | return q->limits.max_sectors; |
| 211 | } |
| 212 | |
| 213 | #ifdef CONFIG_BLK_DEV_INTEGRITY |
| 214 | void blk_flush_integrity(void); |
| 215 | void bio_integrity_free(struct bio *bio); |
| 216 | |
| 217 | /* |
| 218 | * Integrity payloads can either be owned by the submitter, in which case |
| 219 | * bio_uninit will free them, or owned and generated by the block layer, |
| 220 | * in which case we'll verify them here (for reads) and free them before |
| 221 | * the bio is handed back to the submitted. |
| 222 | */ |
| 223 | bool __bio_integrity_endio(struct bio *bio); |
| 224 | static inline bool bio_integrity_endio(struct bio *bio) |
| 225 | { |
| 226 | struct bio_integrity_payload *bip = bio_integrity(bio); |
| 227 | |
| 228 | if (bip && (bip->bip_flags & BIP_BLOCK_INTEGRITY)) |
| 229 | return __bio_integrity_endio(bio); |
| 230 | return true; |
| 231 | } |
| 232 | |
| 233 | bool blk_integrity_merge_rq(struct request_queue *, struct request *, |
| 234 | struct request *); |
| 235 | bool blk_integrity_merge_bio(struct request_queue *, struct request *, |
| 236 | struct bio *); |
| 237 | |
| 238 | static inline bool integrity_req_gap_back_merge(struct request *req, |
| 239 | struct bio *next) |
| 240 | { |
| 241 | struct bio_integrity_payload *bip = bio_integrity(req->bio); |
| 242 | struct bio_integrity_payload *bip_next = bio_integrity(next); |
| 243 | |
| 244 | return bvec_gap_to_prev(&req->q->limits, |
| 245 | &bip->bip_vec[bip->bip_vcnt - 1], |
| 246 | bip_next->bip_vec[0].bv_offset); |
| 247 | } |
| 248 | |
| 249 | static inline bool integrity_req_gap_front_merge(struct request *req, |
| 250 | struct bio *bio) |
| 251 | { |
| 252 | struct bio_integrity_payload *bip = bio_integrity(bio); |
| 253 | struct bio_integrity_payload *bip_next = bio_integrity(req->bio); |
| 254 | |
| 255 | return bvec_gap_to_prev(&req->q->limits, |
| 256 | &bip->bip_vec[bip->bip_vcnt - 1], |
| 257 | bip_next->bip_vec[0].bv_offset); |
| 258 | } |
| 259 | |
| 260 | extern const struct attribute_group blk_integrity_attr_group; |
| 261 | #else /* CONFIG_BLK_DEV_INTEGRITY */ |
| 262 | static inline bool blk_integrity_merge_rq(struct request_queue *rq, |
| 263 | struct request *r1, struct request *r2) |
| 264 | { |
| 265 | return true; |
| 266 | } |
| 267 | static inline bool blk_integrity_merge_bio(struct request_queue *rq, |
| 268 | struct request *r, struct bio *b) |
| 269 | { |
| 270 | return true; |
| 271 | } |
| 272 | static inline bool integrity_req_gap_back_merge(struct request *req, |
| 273 | struct bio *next) |
| 274 | { |
| 275 | return false; |
| 276 | } |
| 277 | static inline bool integrity_req_gap_front_merge(struct request *req, |
| 278 | struct bio *bio) |
| 279 | { |
| 280 | return false; |
| 281 | } |
| 282 | |
| 283 | static inline void blk_flush_integrity(void) |
| 284 | { |
| 285 | } |
| 286 | static inline bool bio_integrity_endio(struct bio *bio) |
| 287 | { |
| 288 | return true; |
| 289 | } |
| 290 | static inline void bio_integrity_free(struct bio *bio) |
| 291 | { |
| 292 | } |
| 293 | #endif /* CONFIG_BLK_DEV_INTEGRITY */ |
| 294 | |
| 295 | unsigned long blk_rq_timeout(unsigned long timeout); |
| 296 | void blk_add_timer(struct request *req); |
| 297 | |
| 298 | enum bio_merge_status { |
| 299 | BIO_MERGE_OK, |
| 300 | BIO_MERGE_NONE, |
| 301 | BIO_MERGE_FAILED, |
| 302 | }; |
| 303 | |
| 304 | enum bio_merge_status bio_attempt_back_merge(struct request *req, |
| 305 | struct bio *bio, unsigned int nr_segs); |
| 306 | bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio, |
| 307 | unsigned int nr_segs); |
| 308 | bool blk_bio_list_merge(struct request_queue *q, struct list_head *list, |
| 309 | struct bio *bio, unsigned int nr_segs); |
| 310 | |
| 311 | /* |
| 312 | * Plug flush limits |
| 313 | */ |
| 314 | #define BLK_MAX_REQUEST_COUNT 32 |
| 315 | #define BLK_PLUG_FLUSH_SIZE (128 * 1024) |
| 316 | |
| 317 | /* |
| 318 | * Internal elevator interface |
| 319 | */ |
| 320 | #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED) |
| 321 | |
| 322 | bool blk_insert_flush(struct request *rq); |
| 323 | |
| 324 | void elv_update_nr_hw_queues(struct request_queue *q); |
| 325 | void elevator_set_default(struct request_queue *q); |
| 326 | void elevator_set_none(struct request_queue *q); |
| 327 | |
| 328 | ssize_t part_size_show(struct device *dev, struct device_attribute *attr, |
| 329 | char *buf); |
| 330 | ssize_t part_stat_show(struct device *dev, struct device_attribute *attr, |
| 331 | char *buf); |
| 332 | ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, |
| 333 | char *buf); |
| 334 | ssize_t part_fail_show(struct device *dev, struct device_attribute *attr, |
| 335 | char *buf); |
| 336 | ssize_t part_fail_store(struct device *dev, struct device_attribute *attr, |
| 337 | const char *buf, size_t count); |
| 338 | ssize_t part_timeout_show(struct device *, struct device_attribute *, char *); |
| 339 | ssize_t part_timeout_store(struct device *, struct device_attribute *, |
| 340 | const char *, size_t); |
| 341 | |
| 342 | struct bio *bio_split_discard(struct bio *bio, const struct queue_limits *lim, |
| 343 | unsigned *nsegs); |
| 344 | struct bio *bio_split_write_zeroes(struct bio *bio, |
| 345 | const struct queue_limits *lim, unsigned *nsegs); |
| 346 | struct bio *bio_split_rw(struct bio *bio, const struct queue_limits *lim, |
| 347 | unsigned *nr_segs); |
| 348 | struct bio *bio_split_zone_append(struct bio *bio, |
| 349 | const struct queue_limits *lim, unsigned *nr_segs); |
| 350 | |
| 351 | /* |
| 352 | * All drivers must accept single-segments bios that are smaller than PAGE_SIZE. |
| 353 | * |
| 354 | * This is a quick and dirty check that relies on the fact that bi_io_vec[0] is |
| 355 | * always valid if a bio has data. The check might lead to occasional false |
| 356 | * positives when bios are cloned, but compared to the performance impact of |
| 357 | * cloned bios themselves the loop below doesn't matter anyway. |
| 358 | */ |
| 359 | static inline bool bio_may_need_split(struct bio *bio, |
| 360 | const struct queue_limits *lim) |
| 361 | { |
| 362 | if (lim->chunk_sectors) |
| 363 | return true; |
| 364 | if (bio->bi_vcnt != 1) |
| 365 | return true; |
| 366 | return bio->bi_io_vec->bv_len + bio->bi_io_vec->bv_offset > |
| 367 | lim->min_segment_size; |
| 368 | } |
| 369 | |
| 370 | /** |
| 371 | * __bio_split_to_limits - split a bio to fit the queue limits |
| 372 | * @bio: bio to be split |
| 373 | * @lim: queue limits to split based on |
| 374 | * @nr_segs: returns the number of segments in the returned bio |
| 375 | * |
| 376 | * Check if @bio needs splitting based on the queue limits, and if so split off |
| 377 | * a bio fitting the limits from the beginning of @bio and return it. @bio is |
| 378 | * shortened to the remainder and re-submitted. |
| 379 | * |
| 380 | * The split bio is allocated from @q->bio_split, which is provided by the |
| 381 | * block layer. |
| 382 | */ |
| 383 | static inline struct bio *__bio_split_to_limits(struct bio *bio, |
| 384 | const struct queue_limits *lim, unsigned int *nr_segs) |
| 385 | { |
| 386 | switch (bio_op(bio)) { |
| 387 | case REQ_OP_READ: |
| 388 | case REQ_OP_WRITE: |
| 389 | if (bio_may_need_split(bio, lim)) |
| 390 | return bio_split_rw(bio, lim, nr_segs); |
| 391 | *nr_segs = 1; |
| 392 | return bio; |
| 393 | case REQ_OP_ZONE_APPEND: |
| 394 | return bio_split_zone_append(bio, lim, nr_segs); |
| 395 | case REQ_OP_DISCARD: |
| 396 | case REQ_OP_SECURE_ERASE: |
| 397 | return bio_split_discard(bio, lim, nr_segs); |
| 398 | case REQ_OP_WRITE_ZEROES: |
| 399 | return bio_split_write_zeroes(bio, lim, nr_segs); |
| 400 | default: |
| 401 | /* other operations can't be split */ |
| 402 | *nr_segs = 0; |
| 403 | return bio; |
| 404 | } |
| 405 | } |
| 406 | |
| 407 | /** |
| 408 | * get_max_segment_size() - maximum number of bytes to add as a single segment |
| 409 | * @lim: Request queue limits. |
| 410 | * @paddr: address of the range to add |
| 411 | * @len: maximum length available to add at @paddr |
| 412 | * |
| 413 | * Returns the maximum number of bytes of the range starting at @paddr that can |
| 414 | * be added to a single segment. |
| 415 | */ |
| 416 | static inline unsigned get_max_segment_size(const struct queue_limits *lim, |
| 417 | phys_addr_t paddr, unsigned int len) |
| 418 | { |
| 419 | /* |
| 420 | * Prevent an overflow if mask = ULONG_MAX and offset = 0 by adding 1 |
| 421 | * after having calculated the minimum. |
| 422 | */ |
| 423 | return min_t(unsigned long, len, |
| 424 | min(lim->seg_boundary_mask - (lim->seg_boundary_mask & paddr), |
| 425 | (unsigned long)lim->max_segment_size - 1) + 1); |
| 426 | } |
| 427 | |
| 428 | int ll_back_merge_fn(struct request *req, struct bio *bio, |
| 429 | unsigned int nr_segs); |
| 430 | bool blk_attempt_req_merge(struct request_queue *q, struct request *rq, |
| 431 | struct request *next); |
| 432 | unsigned int blk_recalc_rq_segments(struct request *rq); |
| 433 | bool blk_rq_merge_ok(struct request *rq, struct bio *bio); |
| 434 | enum elv_merge blk_try_merge(struct request *rq, struct bio *bio); |
| 435 | |
| 436 | int blk_set_default_limits(struct queue_limits *lim); |
| 437 | void blk_apply_bdi_limits(struct backing_dev_info *bdi, |
| 438 | struct queue_limits *lim); |
| 439 | int blk_dev_init(void); |
| 440 | |
| 441 | void update_io_ticks(struct block_device *part, unsigned long now, bool end); |
| 442 | |
| 443 | static inline void req_set_nomerge(struct request_queue *q, struct request *req) |
| 444 | { |
| 445 | req->cmd_flags |= REQ_NOMERGE; |
| 446 | if (req == q->last_merge) |
| 447 | q->last_merge = NULL; |
| 448 | } |
| 449 | |
| 450 | /* |
| 451 | * Internal io_context interface |
| 452 | */ |
| 453 | struct io_cq *ioc_find_get_icq(struct request_queue *q); |
| 454 | struct io_cq *ioc_lookup_icq(struct request_queue *q); |
| 455 | #ifdef CONFIG_BLK_ICQ |
| 456 | void ioc_clear_queue(struct request_queue *q); |
| 457 | #else |
| 458 | static inline void ioc_clear_queue(struct request_queue *q) |
| 459 | { |
| 460 | } |
| 461 | #endif /* CONFIG_BLK_ICQ */ |
| 462 | |
| 463 | #ifdef CONFIG_BLK_DEV_ZONED |
| 464 | void disk_init_zone_resources(struct gendisk *disk); |
| 465 | void disk_free_zone_resources(struct gendisk *disk); |
| 466 | static inline bool bio_zone_write_plugging(struct bio *bio) |
| 467 | { |
| 468 | return bio_flagged(bio, BIO_ZONE_WRITE_PLUGGING); |
| 469 | } |
| 470 | void blk_zone_write_plug_bio_merged(struct bio *bio); |
| 471 | void blk_zone_write_plug_init_request(struct request *rq); |
| 472 | static inline void blk_zone_update_request_bio(struct request *rq, |
| 473 | struct bio *bio) |
| 474 | { |
| 475 | /* |
| 476 | * For zone append requests, the request sector indicates the location |
| 477 | * at which the BIO data was written. Return this value to the BIO |
| 478 | * issuer through the BIO iter sector. |
| 479 | * For plugged zone writes, which include emulated zone append, we need |
| 480 | * the original BIO sector so that blk_zone_write_plug_bio_endio() can |
| 481 | * lookup the zone write plug. |
| 482 | */ |
| 483 | if (req_op(rq) == REQ_OP_ZONE_APPEND || |
| 484 | bio_flagged(bio, BIO_EMULATES_ZONE_APPEND)) |
| 485 | bio->bi_iter.bi_sector = rq->__sector; |
| 486 | } |
| 487 | void blk_zone_write_plug_bio_endio(struct bio *bio); |
| 488 | static inline void blk_zone_bio_endio(struct bio *bio) |
| 489 | { |
| 490 | /* |
| 491 | * For write BIOs to zoned devices, signal the completion of the BIO so |
| 492 | * that the next write BIO can be submitted by zone write plugging. |
| 493 | */ |
| 494 | if (bio_zone_write_plugging(bio)) |
| 495 | blk_zone_write_plug_bio_endio(bio); |
| 496 | } |
| 497 | |
| 498 | void blk_zone_write_plug_finish_request(struct request *rq); |
| 499 | static inline void blk_zone_finish_request(struct request *rq) |
| 500 | { |
| 501 | if (rq->rq_flags & RQF_ZONE_WRITE_PLUGGING) |
| 502 | blk_zone_write_plug_finish_request(rq); |
| 503 | } |
| 504 | int blkdev_report_zones_ioctl(struct block_device *bdev, unsigned int cmd, |
| 505 | unsigned long arg); |
| 506 | int blkdev_zone_mgmt_ioctl(struct block_device *bdev, blk_mode_t mode, |
| 507 | unsigned int cmd, unsigned long arg); |
| 508 | #else /* CONFIG_BLK_DEV_ZONED */ |
| 509 | static inline void disk_init_zone_resources(struct gendisk *disk) |
| 510 | { |
| 511 | } |
| 512 | static inline void disk_free_zone_resources(struct gendisk *disk) |
| 513 | { |
| 514 | } |
| 515 | static inline bool bio_zone_write_plugging(struct bio *bio) |
| 516 | { |
| 517 | return false; |
| 518 | } |
| 519 | static inline void blk_zone_write_plug_bio_merged(struct bio *bio) |
| 520 | { |
| 521 | } |
| 522 | static inline void blk_zone_write_plug_init_request(struct request *rq) |
| 523 | { |
| 524 | } |
| 525 | static inline void blk_zone_update_request_bio(struct request *rq, |
| 526 | struct bio *bio) |
| 527 | { |
| 528 | } |
| 529 | static inline void blk_zone_bio_endio(struct bio *bio) |
| 530 | { |
| 531 | } |
| 532 | static inline void blk_zone_finish_request(struct request *rq) |
| 533 | { |
| 534 | } |
| 535 | static inline int blkdev_report_zones_ioctl(struct block_device *bdev, |
| 536 | unsigned int cmd, unsigned long arg) |
| 537 | { |
| 538 | return -ENOTTY; |
| 539 | } |
| 540 | static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev, |
| 541 | blk_mode_t mode, unsigned int cmd, unsigned long arg) |
| 542 | { |
| 543 | return -ENOTTY; |
| 544 | } |
| 545 | #endif /* CONFIG_BLK_DEV_ZONED */ |
| 546 | |
| 547 | struct block_device *bdev_alloc(struct gendisk *disk, u8 partno); |
| 548 | void bdev_add(struct block_device *bdev, dev_t dev); |
| 549 | void bdev_unhash(struct block_device *bdev); |
| 550 | void bdev_drop(struct block_device *bdev); |
| 551 | |
| 552 | int blk_alloc_ext_minor(void); |
| 553 | void blk_free_ext_minor(unsigned int minor); |
| 554 | #define ADDPART_FLAG_NONE 0 |
| 555 | #define ADDPART_FLAG_RAID 1 |
| 556 | #define ADDPART_FLAG_WHOLEDISK 2 |
| 557 | #define ADDPART_FLAG_READONLY 4 |
| 558 | int bdev_add_partition(struct gendisk *disk, int partno, sector_t start, |
| 559 | sector_t length); |
| 560 | int bdev_del_partition(struct gendisk *disk, int partno); |
| 561 | int bdev_resize_partition(struct gendisk *disk, int partno, sector_t start, |
| 562 | sector_t length); |
| 563 | void drop_partition(struct block_device *part); |
| 564 | |
| 565 | void bdev_set_nr_sectors(struct block_device *bdev, sector_t sectors); |
| 566 | |
| 567 | struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id, |
| 568 | struct lock_class_key *lkclass); |
| 569 | |
| 570 | /* |
| 571 | * Clean up a page appropriately, where the page may be pinned, may have a |
| 572 | * ref taken on it or neither. |
| 573 | */ |
| 574 | static inline void bio_release_page(struct bio *bio, struct page *page) |
| 575 | { |
| 576 | if (bio_flagged(bio, BIO_PAGE_PINNED)) |
| 577 | unpin_user_page(page); |
| 578 | } |
| 579 | |
| 580 | struct request_queue *blk_alloc_queue(struct queue_limits *lim, int node_id); |
| 581 | |
| 582 | int disk_scan_partitions(struct gendisk *disk, blk_mode_t mode); |
| 583 | |
| 584 | int disk_alloc_events(struct gendisk *disk); |
| 585 | void disk_add_events(struct gendisk *disk); |
| 586 | void disk_del_events(struct gendisk *disk); |
| 587 | void disk_release_events(struct gendisk *disk); |
| 588 | void disk_block_events(struct gendisk *disk); |
| 589 | void disk_unblock_events(struct gendisk *disk); |
| 590 | void disk_flush_events(struct gendisk *disk, unsigned int mask); |
| 591 | extern struct device_attribute dev_attr_events; |
| 592 | extern struct device_attribute dev_attr_events_async; |
| 593 | extern struct device_attribute dev_attr_events_poll_msecs; |
| 594 | |
| 595 | extern struct attribute_group blk_trace_attr_group; |
| 596 | |
| 597 | blk_mode_t file_to_blk_mode(struct file *file); |
| 598 | int truncate_bdev_range(struct block_device *bdev, blk_mode_t mode, |
| 599 | loff_t lstart, loff_t lend); |
| 600 | long blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg); |
| 601 | int blkdev_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags); |
| 602 | long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg); |
| 603 | |
| 604 | extern const struct address_space_operations def_blk_aops; |
| 605 | |
| 606 | int disk_register_independent_access_ranges(struct gendisk *disk); |
| 607 | void disk_unregister_independent_access_ranges(struct gendisk *disk); |
| 608 | |
| 609 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
| 610 | bool should_fail_request(struct block_device *part, unsigned int bytes); |
| 611 | #else /* CONFIG_FAIL_MAKE_REQUEST */ |
| 612 | static inline bool should_fail_request(struct block_device *part, |
| 613 | unsigned int bytes) |
| 614 | { |
| 615 | return false; |
| 616 | } |
| 617 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
| 618 | |
| 619 | /* |
| 620 | * Optimized request reference counting. Ideally we'd make timeouts be more |
| 621 | * clever, as that's the only reason we need references at all... But until |
| 622 | * this happens, this is faster than using refcount_t. Also see: |
| 623 | * |
| 624 | * abc54d634334 ("io_uring: switch to atomic_t for io_kiocb reference count") |
| 625 | */ |
| 626 | #define req_ref_zero_or_close_to_overflow(req) \ |
| 627 | ((unsigned int) atomic_read(&(req->ref)) + 127u <= 127u) |
| 628 | |
| 629 | static inline bool req_ref_inc_not_zero(struct request *req) |
| 630 | { |
| 631 | return atomic_inc_not_zero(&req->ref); |
| 632 | } |
| 633 | |
| 634 | static inline bool req_ref_put_and_test(struct request *req) |
| 635 | { |
| 636 | WARN_ON_ONCE(req_ref_zero_or_close_to_overflow(req)); |
| 637 | return atomic_dec_and_test(&req->ref); |
| 638 | } |
| 639 | |
| 640 | static inline void req_ref_set(struct request *req, int value) |
| 641 | { |
| 642 | atomic_set(&req->ref, value); |
| 643 | } |
| 644 | |
| 645 | static inline int req_ref_read(struct request *req) |
| 646 | { |
| 647 | return atomic_read(&req->ref); |
| 648 | } |
| 649 | |
| 650 | static inline u64 blk_time_get_ns(void) |
| 651 | { |
| 652 | struct blk_plug *plug = current->plug; |
| 653 | |
| 654 | if (!plug || !in_task()) |
| 655 | return ktime_get_ns(); |
| 656 | |
| 657 | /* |
| 658 | * 0 could very well be a valid time, but rather than flag "this is |
| 659 | * a valid timestamp" separately, just accept that we'll do an extra |
| 660 | * ktime_get_ns() if we just happen to get 0 as the current time. |
| 661 | */ |
| 662 | if (!plug->cur_ktime) { |
| 663 | plug->cur_ktime = ktime_get_ns(); |
| 664 | current->flags |= PF_BLOCK_TS; |
| 665 | } |
| 666 | return plug->cur_ktime; |
| 667 | } |
| 668 | |
| 669 | static inline ktime_t blk_time_get(void) |
| 670 | { |
| 671 | return ns_to_ktime(blk_time_get_ns()); |
| 672 | } |
| 673 | |
| 674 | /* |
| 675 | * From most significant bit: |
| 676 | * 1 bit: reserved for other usage, see below |
| 677 | * 12 bits: original size of bio |
| 678 | * 51 bits: issue time of bio |
| 679 | */ |
| 680 | #define BIO_ISSUE_RES_BITS 1 |
| 681 | #define BIO_ISSUE_SIZE_BITS 12 |
| 682 | #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS) |
| 683 | #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS) |
| 684 | #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1) |
| 685 | #define BIO_ISSUE_SIZE_MASK \ |
| 686 | (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT) |
| 687 | #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1)) |
| 688 | |
| 689 | /* Reserved bit for blk-throtl */ |
| 690 | #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63) |
| 691 | |
| 692 | static inline u64 __bio_issue_time(u64 time) |
| 693 | { |
| 694 | return time & BIO_ISSUE_TIME_MASK; |
| 695 | } |
| 696 | |
| 697 | static inline u64 bio_issue_time(struct bio_issue *issue) |
| 698 | { |
| 699 | return __bio_issue_time(issue->value); |
| 700 | } |
| 701 | |
| 702 | static inline sector_t bio_issue_size(struct bio_issue *issue) |
| 703 | { |
| 704 | return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT); |
| 705 | } |
| 706 | |
| 707 | static inline void bio_issue_init(struct bio_issue *issue, |
| 708 | sector_t size) |
| 709 | { |
| 710 | size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1; |
| 711 | issue->value = ((issue->value & BIO_ISSUE_RES_MASK) | |
| 712 | (blk_time_get_ns() & BIO_ISSUE_TIME_MASK) | |
| 713 | ((u64)size << BIO_ISSUE_SIZE_SHIFT)); |
| 714 | } |
| 715 | |
| 716 | void bdev_release(struct file *bdev_file); |
| 717 | int bdev_open(struct block_device *bdev, blk_mode_t mode, void *holder, |
| 718 | const struct blk_holder_ops *hops, struct file *bdev_file); |
| 719 | int bdev_permission(dev_t dev, blk_mode_t mode, void *holder); |
| 720 | |
| 721 | void blk_integrity_generate(struct bio *bio); |
| 722 | void blk_integrity_verify_iter(struct bio *bio, struct bvec_iter *saved_iter); |
| 723 | void blk_integrity_prepare(struct request *rq); |
| 724 | void blk_integrity_complete(struct request *rq, unsigned int nr_bytes); |
| 725 | |
| 726 | #ifdef CONFIG_LOCKDEP |
| 727 | static inline void blk_freeze_acquire_lock(struct request_queue *q) |
| 728 | { |
| 729 | if (!q->mq_freeze_disk_dead) |
| 730 | rwsem_acquire(&q->io_lockdep_map, 0, 1, _RET_IP_); |
| 731 | if (!q->mq_freeze_queue_dying) |
| 732 | rwsem_acquire(&q->q_lockdep_map, 0, 1, _RET_IP_); |
| 733 | } |
| 734 | |
| 735 | static inline void blk_unfreeze_release_lock(struct request_queue *q) |
| 736 | { |
| 737 | if (!q->mq_freeze_queue_dying) |
| 738 | rwsem_release(&q->q_lockdep_map, _RET_IP_); |
| 739 | if (!q->mq_freeze_disk_dead) |
| 740 | rwsem_release(&q->io_lockdep_map, _RET_IP_); |
| 741 | } |
| 742 | #else |
| 743 | static inline void blk_freeze_acquire_lock(struct request_queue *q) |
| 744 | { |
| 745 | } |
| 746 | static inline void blk_unfreeze_release_lock(struct request_queue *q) |
| 747 | { |
| 748 | } |
| 749 | #endif |
| 750 | |
| 751 | #endif /* BLK_INTERNAL_H */ |