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
blk_try_enter_queue(struct request_queue * q,bool pm)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
bio_queue_enter(struct bio * bio)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
blk_wait_io(struct completion * done)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 #define BIO_INLINE_VECS 4
98 struct bio_vec *bvec_alloc(mempool_t *pool, unsigned short *nr_vecs,
99 gfp_t gfp_mask);
100 void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned short nr_vecs);
101
102 bool bvec_try_merge_hw_page(struct request_queue *q, struct bio_vec *bv,
103 struct page *page, unsigned len, unsigned offset,
104 bool *same_page);
105
biovec_phys_mergeable(struct request_queue * q,struct bio_vec * vec1,struct bio_vec * vec2)106 static inline bool biovec_phys_mergeable(struct request_queue *q,
107 struct bio_vec *vec1, struct bio_vec *vec2)
108 {
109 unsigned long mask = queue_segment_boundary(q);
110 phys_addr_t addr1 = bvec_phys(vec1);
111 phys_addr_t addr2 = bvec_phys(vec2);
112
113 /*
114 * Merging adjacent physical pages may not work correctly under KMSAN
115 * if their metadata pages aren't adjacent. Just disable merging.
116 */
117 if (IS_ENABLED(CONFIG_KMSAN))
118 return false;
119
120 if (addr1 + vec1->bv_len != addr2)
121 return false;
122 if (xen_domain() && !xen_biovec_phys_mergeable(vec1, vec2->bv_page))
123 return false;
124 if ((addr1 | mask) != ((addr2 + vec2->bv_len - 1) | mask))
125 return false;
126 return true;
127 }
128
__bvec_gap_to_prev(const struct queue_limits * lim,struct bio_vec * bprv,unsigned int offset)129 static inline bool __bvec_gap_to_prev(const struct queue_limits *lim,
130 struct bio_vec *bprv, unsigned int offset)
131 {
132 return (offset & lim->virt_boundary_mask) ||
133 ((bprv->bv_offset + bprv->bv_len) & lim->virt_boundary_mask);
134 }
135
136 /*
137 * Check if adding a bio_vec after bprv with offset would create a gap in
138 * the SG list. Most drivers don't care about this, but some do.
139 */
bvec_gap_to_prev(const struct queue_limits * lim,struct bio_vec * bprv,unsigned int offset)140 static inline bool bvec_gap_to_prev(const struct queue_limits *lim,
141 struct bio_vec *bprv, unsigned int offset)
142 {
143 if (!lim->virt_boundary_mask)
144 return false;
145 return __bvec_gap_to_prev(lim, bprv, offset);
146 }
147
rq_mergeable(struct request * rq)148 static inline bool rq_mergeable(struct request *rq)
149 {
150 if (blk_rq_is_passthrough(rq))
151 return false;
152
153 if (req_op(rq) == REQ_OP_FLUSH)
154 return false;
155
156 if (req_op(rq) == REQ_OP_WRITE_ZEROES)
157 return false;
158
159 if (req_op(rq) == REQ_OP_ZONE_APPEND)
160 return false;
161
162 if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
163 return false;
164 if (rq->rq_flags & RQF_NOMERGE_FLAGS)
165 return false;
166
167 return true;
168 }
169
170 /*
171 * There are two different ways to handle DISCARD merges:
172 * 1) If max_discard_segments > 1, the driver treats every bio as a range and
173 * send the bios to controller together. The ranges don't need to be
174 * contiguous.
175 * 2) Otherwise, the request will be normal read/write requests. The ranges
176 * need to be contiguous.
177 */
blk_discard_mergable(struct request * req)178 static inline bool blk_discard_mergable(struct request *req)
179 {
180 if (req_op(req) == REQ_OP_DISCARD &&
181 queue_max_discard_segments(req->q) > 1)
182 return true;
183 return false;
184 }
185
blk_rq_get_max_segments(struct request * rq)186 static inline unsigned int blk_rq_get_max_segments(struct request *rq)
187 {
188 if (req_op(rq) == REQ_OP_DISCARD)
189 return queue_max_discard_segments(rq->q);
190 return queue_max_segments(rq->q);
191 }
192
blk_queue_get_max_sectors(struct request * rq)193 static inline unsigned int blk_queue_get_max_sectors(struct request *rq)
194 {
195 struct request_queue *q = rq->q;
196 enum req_op op = req_op(rq);
197
198 if (unlikely(op == REQ_OP_DISCARD || op == REQ_OP_SECURE_ERASE))
199 return min(q->limits.max_discard_sectors,
200 UINT_MAX >> SECTOR_SHIFT);
201
202 if (unlikely(op == REQ_OP_WRITE_ZEROES))
203 return q->limits.max_write_zeroes_sectors;
204
205 if (rq->cmd_flags & REQ_ATOMIC)
206 return q->limits.atomic_write_max_sectors;
207
208 return q->limits.max_sectors;
209 }
210
211 #ifdef CONFIG_BLK_DEV_INTEGRITY
212 void blk_flush_integrity(void);
213 void bio_integrity_free(struct bio *bio);
214
215 /*
216 * Integrity payloads can either be owned by the submitter, in which case
217 * bio_uninit will free them, or owned and generated by the block layer,
218 * in which case we'll verify them here (for reads) and free them before
219 * the bio is handed back to the submitted.
220 */
221 bool __bio_integrity_endio(struct bio *bio);
bio_integrity_endio(struct bio * bio)222 static inline bool bio_integrity_endio(struct bio *bio)
223 {
224 struct bio_integrity_payload *bip = bio_integrity(bio);
225
226 if (bip && (bip->bip_flags & BIP_BLOCK_INTEGRITY))
227 return __bio_integrity_endio(bio);
228 return true;
229 }
230
231 bool blk_integrity_merge_rq(struct request_queue *, struct request *,
232 struct request *);
233 bool blk_integrity_merge_bio(struct request_queue *, struct request *,
234 struct bio *);
235
integrity_req_gap_back_merge(struct request * req,struct bio * next)236 static inline bool integrity_req_gap_back_merge(struct request *req,
237 struct bio *next)
238 {
239 struct bio_integrity_payload *bip = bio_integrity(req->bio);
240 struct bio_integrity_payload *bip_next = bio_integrity(next);
241
242 return bvec_gap_to_prev(&req->q->limits,
243 &bip->bip_vec[bip->bip_vcnt - 1],
244 bip_next->bip_vec[0].bv_offset);
245 }
246
integrity_req_gap_front_merge(struct request * req,struct bio * bio)247 static inline bool integrity_req_gap_front_merge(struct request *req,
248 struct bio *bio)
249 {
250 struct bio_integrity_payload *bip = bio_integrity(bio);
251 struct bio_integrity_payload *bip_next = bio_integrity(req->bio);
252
253 return bvec_gap_to_prev(&req->q->limits,
254 &bip->bip_vec[bip->bip_vcnt - 1],
255 bip_next->bip_vec[0].bv_offset);
256 }
257
258 extern const struct attribute_group blk_integrity_attr_group;
259 #else /* CONFIG_BLK_DEV_INTEGRITY */
blk_integrity_merge_rq(struct request_queue * rq,struct request * r1,struct request * r2)260 static inline bool blk_integrity_merge_rq(struct request_queue *rq,
261 struct request *r1, struct request *r2)
262 {
263 return true;
264 }
blk_integrity_merge_bio(struct request_queue * rq,struct request * r,struct bio * b)265 static inline bool blk_integrity_merge_bio(struct request_queue *rq,
266 struct request *r, struct bio *b)
267 {
268 return true;
269 }
integrity_req_gap_back_merge(struct request * req,struct bio * next)270 static inline bool integrity_req_gap_back_merge(struct request *req,
271 struct bio *next)
272 {
273 return false;
274 }
integrity_req_gap_front_merge(struct request * req,struct bio * bio)275 static inline bool integrity_req_gap_front_merge(struct request *req,
276 struct bio *bio)
277 {
278 return false;
279 }
280
blk_flush_integrity(void)281 static inline void blk_flush_integrity(void)
282 {
283 }
bio_integrity_endio(struct bio * bio)284 static inline bool bio_integrity_endio(struct bio *bio)
285 {
286 return true;
287 }
bio_integrity_free(struct bio * bio)288 static inline void bio_integrity_free(struct bio *bio)
289 {
290 }
291 #endif /* CONFIG_BLK_DEV_INTEGRITY */
292
293 unsigned long blk_rq_timeout(unsigned long timeout);
294 void blk_add_timer(struct request *req);
295
296 enum bio_merge_status {
297 BIO_MERGE_OK,
298 BIO_MERGE_NONE,
299 BIO_MERGE_FAILED,
300 };
301
302 enum bio_merge_status bio_attempt_back_merge(struct request *req,
303 struct bio *bio, unsigned int nr_segs);
304 bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
305 unsigned int nr_segs);
306 bool blk_bio_list_merge(struct request_queue *q, struct list_head *list,
307 struct bio *bio, unsigned int nr_segs);
308
309 /*
310 * Plug flush limits
311 */
312 #define BLK_MAX_REQUEST_COUNT 32
313 #define BLK_PLUG_FLUSH_SIZE (128 * 1024)
314
315 /*
316 * Internal elevator interface
317 */
318 #define ELV_ON_HASH(rq) ((rq)->rq_flags & RQF_HASHED)
319
320 bool blk_insert_flush(struct request *rq);
321
322 int elevator_switch(struct request_queue *q, struct elevator_type *new_e);
323 void elevator_disable(struct request_queue *q);
324 void elevator_exit(struct request_queue *q);
325 int elv_register_queue(struct request_queue *q, bool uevent);
326 void elv_unregister_queue(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 */
bio_may_need_split(struct bio * bio,const struct queue_limits * lim)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 */
__bio_split_to_limits(struct bio * bio,const struct queue_limits * lim,unsigned int * nr_segs)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 int ll_back_merge_fn(struct request *req, struct bio *bio,
408 unsigned int nr_segs);
409 bool blk_attempt_req_merge(struct request_queue *q, struct request *rq,
410 struct request *next);
411 unsigned int blk_recalc_rq_segments(struct request *rq);
412 bool blk_rq_merge_ok(struct request *rq, struct bio *bio);
413 enum elv_merge blk_try_merge(struct request *rq, struct bio *bio);
414
415 int blk_set_default_limits(struct queue_limits *lim);
416 void blk_apply_bdi_limits(struct backing_dev_info *bdi,
417 struct queue_limits *lim);
418 int blk_dev_init(void);
419
420 void update_io_ticks(struct block_device *part, unsigned long now, bool end);
421 unsigned int part_in_flight(struct block_device *part);
422
req_set_nomerge(struct request_queue * q,struct request * req)423 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
424 {
425 req->cmd_flags |= REQ_NOMERGE;
426 if (req == q->last_merge)
427 q->last_merge = NULL;
428 }
429
430 /*
431 * Internal io_context interface
432 */
433 struct io_cq *ioc_find_get_icq(struct request_queue *q);
434 struct io_cq *ioc_lookup_icq(struct request_queue *q);
435 #ifdef CONFIG_BLK_ICQ
436 void ioc_clear_queue(struct request_queue *q);
437 #else
ioc_clear_queue(struct request_queue * q)438 static inline void ioc_clear_queue(struct request_queue *q)
439 {
440 }
441 #endif /* CONFIG_BLK_ICQ */
442
443 struct bio *__blk_queue_bounce(struct bio *bio, struct request_queue *q);
444
blk_queue_may_bounce(struct request_queue * q)445 static inline bool blk_queue_may_bounce(struct request_queue *q)
446 {
447 return IS_ENABLED(CONFIG_BOUNCE) &&
448 (q->limits.features & BLK_FEAT_BOUNCE_HIGH) &&
449 max_low_pfn >= max_pfn;
450 }
451
blk_queue_bounce(struct bio * bio,struct request_queue * q)452 static inline struct bio *blk_queue_bounce(struct bio *bio,
453 struct request_queue *q)
454 {
455 if (unlikely(blk_queue_may_bounce(q) && bio_has_data(bio)))
456 return __blk_queue_bounce(bio, q);
457 return bio;
458 }
459
460 #ifdef CONFIG_BLK_DEV_ZONED
461 void disk_init_zone_resources(struct gendisk *disk);
462 void disk_free_zone_resources(struct gendisk *disk);
bio_zone_write_plugging(struct bio * bio)463 static inline bool bio_zone_write_plugging(struct bio *bio)
464 {
465 return bio_flagged(bio, BIO_ZONE_WRITE_PLUGGING);
466 }
467 void blk_zone_write_plug_bio_merged(struct bio *bio);
468 void blk_zone_write_plug_init_request(struct request *rq);
blk_zone_update_request_bio(struct request * rq,struct bio * bio)469 static inline void blk_zone_update_request_bio(struct request *rq,
470 struct bio *bio)
471 {
472 /*
473 * For zone append requests, the request sector indicates the location
474 * at which the BIO data was written. Return this value to the BIO
475 * issuer through the BIO iter sector.
476 * For plugged zone writes, which include emulated zone append, we need
477 * the original BIO sector so that blk_zone_write_plug_bio_endio() can
478 * lookup the zone write plug.
479 */
480 if (req_op(rq) == REQ_OP_ZONE_APPEND || bio_zone_write_plugging(bio))
481 bio->bi_iter.bi_sector = rq->__sector;
482 }
483 void blk_zone_write_plug_bio_endio(struct bio *bio);
blk_zone_bio_endio(struct bio * bio)484 static inline void blk_zone_bio_endio(struct bio *bio)
485 {
486 /*
487 * For write BIOs to zoned devices, signal the completion of the BIO so
488 * that the next write BIO can be submitted by zone write plugging.
489 */
490 if (bio_zone_write_plugging(bio))
491 blk_zone_write_plug_bio_endio(bio);
492 }
493
494 void blk_zone_write_plug_finish_request(struct request *rq);
blk_zone_finish_request(struct request * rq)495 static inline void blk_zone_finish_request(struct request *rq)
496 {
497 if (rq->rq_flags & RQF_ZONE_WRITE_PLUGGING)
498 blk_zone_write_plug_finish_request(rq);
499 }
500 int blkdev_report_zones_ioctl(struct block_device *bdev, unsigned int cmd,
501 unsigned long arg);
502 int blkdev_zone_mgmt_ioctl(struct block_device *bdev, blk_mode_t mode,
503 unsigned int cmd, unsigned long arg);
504 #else /* CONFIG_BLK_DEV_ZONED */
disk_init_zone_resources(struct gendisk * disk)505 static inline void disk_init_zone_resources(struct gendisk *disk)
506 {
507 }
disk_free_zone_resources(struct gendisk * disk)508 static inline void disk_free_zone_resources(struct gendisk *disk)
509 {
510 }
bio_zone_write_plugging(struct bio * bio)511 static inline bool bio_zone_write_plugging(struct bio *bio)
512 {
513 return false;
514 }
blk_zone_write_plug_bio_merged(struct bio * bio)515 static inline void blk_zone_write_plug_bio_merged(struct bio *bio)
516 {
517 }
blk_zone_write_plug_init_request(struct request * rq)518 static inline void blk_zone_write_plug_init_request(struct request *rq)
519 {
520 }
blk_zone_update_request_bio(struct request * rq,struct bio * bio)521 static inline void blk_zone_update_request_bio(struct request *rq,
522 struct bio *bio)
523 {
524 }
blk_zone_bio_endio(struct bio * bio)525 static inline void blk_zone_bio_endio(struct bio *bio)
526 {
527 }
blk_zone_finish_request(struct request * rq)528 static inline void blk_zone_finish_request(struct request *rq)
529 {
530 }
blkdev_report_zones_ioctl(struct block_device * bdev,unsigned int cmd,unsigned long arg)531 static inline int blkdev_report_zones_ioctl(struct block_device *bdev,
532 unsigned int cmd, unsigned long arg)
533 {
534 return -ENOTTY;
535 }
blkdev_zone_mgmt_ioctl(struct block_device * bdev,blk_mode_t mode,unsigned int cmd,unsigned long arg)536 static inline int blkdev_zone_mgmt_ioctl(struct block_device *bdev,
537 blk_mode_t mode, unsigned int cmd, unsigned long arg)
538 {
539 return -ENOTTY;
540 }
541 #endif /* CONFIG_BLK_DEV_ZONED */
542
543 struct block_device *bdev_alloc(struct gendisk *disk, u8 partno);
544 void bdev_add(struct block_device *bdev, dev_t dev);
545 void bdev_unhash(struct block_device *bdev);
546 void bdev_drop(struct block_device *bdev);
547
548 int blk_alloc_ext_minor(void);
549 void blk_free_ext_minor(unsigned int minor);
550 #define ADDPART_FLAG_NONE 0
551 #define ADDPART_FLAG_RAID 1
552 #define ADDPART_FLAG_WHOLEDISK 2
553 #define ADDPART_FLAG_READONLY 4
554 int bdev_add_partition(struct gendisk *disk, int partno, sector_t start,
555 sector_t length);
556 int bdev_del_partition(struct gendisk *disk, int partno);
557 int bdev_resize_partition(struct gendisk *disk, int partno, sector_t start,
558 sector_t length);
559 void drop_partition(struct block_device *part);
560
561 void bdev_set_nr_sectors(struct block_device *bdev, sector_t sectors);
562
563 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id,
564 struct lock_class_key *lkclass);
565
566 /*
567 * Clean up a page appropriately, where the page may be pinned, may have a
568 * ref taken on it or neither.
569 */
bio_release_page(struct bio * bio,struct page * page)570 static inline void bio_release_page(struct bio *bio, struct page *page)
571 {
572 if (bio_flagged(bio, BIO_PAGE_PINNED))
573 unpin_user_page(page);
574 }
575
576 struct request_queue *blk_alloc_queue(struct queue_limits *lim, int node_id);
577
578 int disk_scan_partitions(struct gendisk *disk, blk_mode_t mode);
579
580 int disk_alloc_events(struct gendisk *disk);
581 void disk_add_events(struct gendisk *disk);
582 void disk_del_events(struct gendisk *disk);
583 void disk_release_events(struct gendisk *disk);
584 void disk_block_events(struct gendisk *disk);
585 void disk_unblock_events(struct gendisk *disk);
586 void disk_flush_events(struct gendisk *disk, unsigned int mask);
587 extern struct device_attribute dev_attr_events;
588 extern struct device_attribute dev_attr_events_async;
589 extern struct device_attribute dev_attr_events_poll_msecs;
590
591 extern struct attribute_group blk_trace_attr_group;
592
593 blk_mode_t file_to_blk_mode(struct file *file);
594 int truncate_bdev_range(struct block_device *bdev, blk_mode_t mode,
595 loff_t lstart, loff_t lend);
596 long blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
597 int blkdev_uring_cmd(struct io_uring_cmd *cmd, unsigned int issue_flags);
598 long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg);
599
600 extern const struct address_space_operations def_blk_aops;
601
602 int disk_register_independent_access_ranges(struct gendisk *disk);
603 void disk_unregister_independent_access_ranges(struct gendisk *disk);
604
605 #ifdef CONFIG_FAIL_MAKE_REQUEST
606 bool should_fail_request(struct block_device *part, unsigned int bytes);
607 #else /* CONFIG_FAIL_MAKE_REQUEST */
should_fail_request(struct block_device * part,unsigned int bytes)608 static inline bool should_fail_request(struct block_device *part,
609 unsigned int bytes)
610 {
611 return false;
612 }
613 #endif /* CONFIG_FAIL_MAKE_REQUEST */
614
615 /*
616 * Optimized request reference counting. Ideally we'd make timeouts be more
617 * clever, as that's the only reason we need references at all... But until
618 * this happens, this is faster than using refcount_t. Also see:
619 *
620 * abc54d634334 ("io_uring: switch to atomic_t for io_kiocb reference count")
621 */
622 #define req_ref_zero_or_close_to_overflow(req) \
623 ((unsigned int) atomic_read(&(req->ref)) + 127u <= 127u)
624
req_ref_inc_not_zero(struct request * req)625 static inline bool req_ref_inc_not_zero(struct request *req)
626 {
627 return atomic_inc_not_zero(&req->ref);
628 }
629
req_ref_put_and_test(struct request * req)630 static inline bool req_ref_put_and_test(struct request *req)
631 {
632 WARN_ON_ONCE(req_ref_zero_or_close_to_overflow(req));
633 return atomic_dec_and_test(&req->ref);
634 }
635
req_ref_set(struct request * req,int value)636 static inline void req_ref_set(struct request *req, int value)
637 {
638 atomic_set(&req->ref, value);
639 }
640
req_ref_read(struct request * req)641 static inline int req_ref_read(struct request *req)
642 {
643 return atomic_read(&req->ref);
644 }
645
blk_time_get_ns(void)646 static inline u64 blk_time_get_ns(void)
647 {
648 struct blk_plug *plug = current->plug;
649
650 if (!plug || !in_task())
651 return ktime_get_ns();
652
653 /*
654 * 0 could very well be a valid time, but rather than flag "this is
655 * a valid timestamp" separately, just accept that we'll do an extra
656 * ktime_get_ns() if we just happen to get 0 as the current time.
657 */
658 if (!plug->cur_ktime) {
659 plug->cur_ktime = ktime_get_ns();
660 current->flags |= PF_BLOCK_TS;
661 }
662 return plug->cur_ktime;
663 }
664
blk_time_get(void)665 static inline ktime_t blk_time_get(void)
666 {
667 return ns_to_ktime(blk_time_get_ns());
668 }
669
670 /*
671 * From most significant bit:
672 * 1 bit: reserved for other usage, see below
673 * 12 bits: original size of bio
674 * 51 bits: issue time of bio
675 */
676 #define BIO_ISSUE_RES_BITS 1
677 #define BIO_ISSUE_SIZE_BITS 12
678 #define BIO_ISSUE_RES_SHIFT (64 - BIO_ISSUE_RES_BITS)
679 #define BIO_ISSUE_SIZE_SHIFT (BIO_ISSUE_RES_SHIFT - BIO_ISSUE_SIZE_BITS)
680 #define BIO_ISSUE_TIME_MASK ((1ULL << BIO_ISSUE_SIZE_SHIFT) - 1)
681 #define BIO_ISSUE_SIZE_MASK \
682 (((1ULL << BIO_ISSUE_SIZE_BITS) - 1) << BIO_ISSUE_SIZE_SHIFT)
683 #define BIO_ISSUE_RES_MASK (~((1ULL << BIO_ISSUE_RES_SHIFT) - 1))
684
685 /* Reserved bit for blk-throtl */
686 #define BIO_ISSUE_THROTL_SKIP_LATENCY (1ULL << 63)
687
__bio_issue_time(u64 time)688 static inline u64 __bio_issue_time(u64 time)
689 {
690 return time & BIO_ISSUE_TIME_MASK;
691 }
692
bio_issue_time(struct bio_issue * issue)693 static inline u64 bio_issue_time(struct bio_issue *issue)
694 {
695 return __bio_issue_time(issue->value);
696 }
697
bio_issue_size(struct bio_issue * issue)698 static inline sector_t bio_issue_size(struct bio_issue *issue)
699 {
700 return ((issue->value & BIO_ISSUE_SIZE_MASK) >> BIO_ISSUE_SIZE_SHIFT);
701 }
702
bio_issue_init(struct bio_issue * issue,sector_t size)703 static inline void bio_issue_init(struct bio_issue *issue,
704 sector_t size)
705 {
706 size &= (1ULL << BIO_ISSUE_SIZE_BITS) - 1;
707 issue->value = ((issue->value & BIO_ISSUE_RES_MASK) |
708 (blk_time_get_ns() & BIO_ISSUE_TIME_MASK) |
709 ((u64)size << BIO_ISSUE_SIZE_SHIFT));
710 }
711
712 void bdev_release(struct file *bdev_file);
713 int bdev_open(struct block_device *bdev, blk_mode_t mode, void *holder,
714 const struct blk_holder_ops *hops, struct file *bdev_file);
715 int bdev_permission(dev_t dev, blk_mode_t mode, void *holder);
716
717 void blk_integrity_generate(struct bio *bio);
718 void blk_integrity_verify(struct bio *bio);
719 void blk_integrity_prepare(struct request *rq);
720 void blk_integrity_complete(struct request *rq, unsigned int nr_bytes);
721
722 #ifdef CONFIG_LOCKDEP
blk_freeze_acquire_lock(struct request_queue * q)723 static inline void blk_freeze_acquire_lock(struct request_queue *q)
724 {
725 if (!q->mq_freeze_disk_dead)
726 rwsem_acquire(&q->io_lockdep_map, 0, 1, _RET_IP_);
727 if (!q->mq_freeze_queue_dying)
728 rwsem_acquire(&q->q_lockdep_map, 0, 1, _RET_IP_);
729 }
730
blk_unfreeze_release_lock(struct request_queue * q)731 static inline void blk_unfreeze_release_lock(struct request_queue *q)
732 {
733 if (!q->mq_freeze_queue_dying)
734 rwsem_release(&q->q_lockdep_map, _RET_IP_);
735 if (!q->mq_freeze_disk_dead)
736 rwsem_release(&q->io_lockdep_map, _RET_IP_);
737 }
738 #else
blk_freeze_acquire_lock(struct request_queue * q)739 static inline void blk_freeze_acquire_lock(struct request_queue *q)
740 {
741 }
blk_unfreeze_release_lock(struct request_queue * q)742 static inline void blk_unfreeze_release_lock(struct request_queue *q)
743 {
744 }
745 #endif
746
747 #endif /* BLK_INTERNAL_H */
748