1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3 * Copyright (C) 2001 Jens Axboe <axboe@suse.de>
4 */
5 #ifndef __LINUX_BIO_H
6 #define __LINUX_BIO_H
7
8 #include <linux/mempool.h>
9 /* struct bio, bio_vec and BIO_* flags are defined in blk_types.h */
10 #include <linux/blk_types.h>
11 #include <linux/uio.h>
12
13 #define BIO_MAX_VECS 256U
14
15 struct queue_limits;
16
17 static inline unsigned int bio_max_segs(unsigned int nr_segs)
18 {
19 return min(nr_segs, BIO_MAX_VECS);
20 }
21
22 #define bio_prio(bio) (bio)->bi_ioprio
23 #define bio_set_prio(bio, prio) ((bio)->bi_ioprio = prio)
24
25 #define bio_iter_iovec(bio, iter) \
26 bvec_iter_bvec((bio)->bi_io_vec, (iter))
27
28 #define bio_iter_page(bio, iter) \
29 bvec_iter_page((bio)->bi_io_vec, (iter))
30 #define bio_iter_len(bio, iter) \
31 bvec_iter_len((bio)->bi_io_vec, (iter))
32 #define bio_iter_offset(bio, iter) \
33 bvec_iter_offset((bio)->bi_io_vec, (iter))
34
35 #define bio_page(bio) bio_iter_page((bio), (bio)->bi_iter)
36 #define bio_offset(bio) bio_iter_offset((bio), (bio)->bi_iter)
37 #define bio_iovec(bio) bio_iter_iovec((bio), (bio)->bi_iter)
38
39 #define bvec_iter_sectors(iter) ((iter).bi_size >> 9)
40 #define bvec_iter_end_sector(iter) ((iter).bi_sector + bvec_iter_sectors((iter)))
41
42 #define bio_sectors(bio) bvec_iter_sectors((bio)->bi_iter)
43 #define bio_end_sector(bio) bvec_iter_end_sector((bio)->bi_iter)
44
45 /*
46 * Return the data direction, READ or WRITE.
47 */
48 #define bio_data_dir(bio) \
49 (op_is_write(bio_op(bio)) ? WRITE : READ)
50
51 /*
52 * Check whether this bio carries any data or not. A NULL bio is allowed.
53 */
54 static inline bool bio_has_data(struct bio *bio)
55 {
56 if (bio &&
57 bio->bi_iter.bi_size &&
58 bio_op(bio) != REQ_OP_DISCARD &&
59 bio_op(bio) != REQ_OP_SECURE_ERASE &&
60 bio_op(bio) != REQ_OP_WRITE_ZEROES)
61 return true;
62
63 return false;
64 }
65
66 static inline bool bio_no_advance_iter(const struct bio *bio)
67 {
68 return bio_op(bio) == REQ_OP_DISCARD ||
69 bio_op(bio) == REQ_OP_SECURE_ERASE ||
70 bio_op(bio) == REQ_OP_WRITE_ZEROES;
71 }
72
73 static inline void *bio_data(struct bio *bio)
74 {
75 if (bio_has_data(bio))
76 return page_address(bio_page(bio)) + bio_offset(bio);
77
78 return NULL;
79 }
80
81 static inline bool bio_next_segment(const struct bio *bio,
82 struct bvec_iter_all *iter)
83 {
84 if (iter->idx >= bio->bi_vcnt)
85 return false;
86
87 bvec_advance(&bio->bi_io_vec[iter->idx], iter);
88 return true;
89 }
90
91 /*
92 * drivers should _never_ use the all version - the bio may have been split
93 * before it got to the driver and the driver won't own all of it
94 */
95 #define bio_for_each_segment_all(bvl, bio, iter) \
96 for (bvl = bvec_init_iter_all(&iter); bio_next_segment((bio), &iter); )
97
98 static inline void bio_advance_iter(const struct bio *bio,
99 struct bvec_iter *iter, unsigned int bytes)
100 {
101 iter->bi_sector += bytes >> 9;
102
103 if (bio_no_advance_iter(bio))
104 iter->bi_size -= bytes;
105 else
106 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
107 /* TODO: It is reasonable to complete bio with error here. */
108 }
109
110 /* @bytes should be less or equal to bvec[i->bi_idx].bv_len */
111 static inline void bio_advance_iter_single(const struct bio *bio,
112 struct bvec_iter *iter,
113 unsigned int bytes)
114 {
115 iter->bi_sector += bytes >> 9;
116
117 if (bio_no_advance_iter(bio))
118 iter->bi_size -= bytes;
119 else
120 bvec_iter_advance_single(bio->bi_io_vec, iter, bytes);
121 }
122
123 void __bio_advance(struct bio *, unsigned bytes);
124
125 /**
126 * bio_advance - increment/complete a bio by some number of bytes
127 * @bio: bio to advance
128 * @nbytes: number of bytes to complete
129 *
130 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
131 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
132 * be updated on the last bvec as well.
133 *
134 * @bio will then represent the remaining, uncompleted portion of the io.
135 */
136 static inline void bio_advance(struct bio *bio, unsigned int nbytes)
137 {
138 if (nbytes == bio->bi_iter.bi_size) {
139 bio->bi_iter.bi_size = 0;
140 return;
141 }
142 __bio_advance(bio, nbytes);
143 }
144
145 #define __bio_for_each_segment(bvl, bio, iter, start) \
146 for (iter = (start); \
147 (iter).bi_size && \
148 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
149 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
150
151 #define bio_for_each_segment(bvl, bio, iter) \
152 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
153
154 #define __bio_for_each_bvec(bvl, bio, iter, start) \
155 for (iter = (start); \
156 (iter).bi_size && \
157 ((bvl = mp_bvec_iter_bvec((bio)->bi_io_vec, (iter))), 1); \
158 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
159
160 /* iterate over multi-page bvec */
161 #define bio_for_each_bvec(bvl, bio, iter) \
162 __bio_for_each_bvec(bvl, bio, iter, (bio)->bi_iter)
163
164 /*
165 * Iterate over all multi-page bvecs. Drivers shouldn't use this version for the
166 * same reasons as bio_for_each_segment_all().
167 */
168 #define bio_for_each_bvec_all(bvl, bio, i) \
169 for (i = 0, bvl = bio_first_bvec_all(bio); \
170 i < (bio)->bi_vcnt; i++, bvl++)
171
172 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
173
174 static inline unsigned bio_segments(struct bio *bio)
175 {
176 unsigned segs = 0;
177 struct bio_vec bv;
178 struct bvec_iter iter;
179
180 /*
181 * We special case discard/write same/write zeroes, because they
182 * interpret bi_size differently:
183 */
184
185 switch (bio_op(bio)) {
186 case REQ_OP_DISCARD:
187 case REQ_OP_SECURE_ERASE:
188 case REQ_OP_WRITE_ZEROES:
189 return 0;
190 default:
191 break;
192 }
193
194 bio_for_each_segment(bv, bio, iter)
195 segs++;
196
197 return segs;
198 }
199
200 /*
201 * get a reference to a bio, so it won't disappear. the intended use is
202 * something like:
203 *
204 * bio_get(bio);
205 * submit_bio(rw, bio);
206 * if (bio->bi_flags ...)
207 * do_something
208 * bio_put(bio);
209 *
210 * without the bio_get(), it could potentially complete I/O before submit_bio
211 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
212 * runs
213 */
214 static inline void bio_get(struct bio *bio)
215 {
216 bio->bi_flags |= (1 << BIO_REFFED);
217 smp_mb__before_atomic();
218 atomic_inc(&bio->__bi_cnt);
219 }
220
221 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
222 {
223 if (count != 1) {
224 bio->bi_flags |= (1 << BIO_REFFED);
225 smp_mb();
226 }
227 atomic_set(&bio->__bi_cnt, count);
228 }
229
230 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
231 {
232 return bio->bi_flags & (1U << bit);
233 }
234
235 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
236 {
237 bio->bi_flags |= (1U << bit);
238 }
239
240 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
241 {
242 bio->bi_flags &= ~(1U << bit);
243 }
244
245 static inline struct bio_vec *bio_first_bvec_all(struct bio *bio)
246 {
247 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
248 return bio->bi_io_vec;
249 }
250
251 static inline struct page *bio_first_page_all(struct bio *bio)
252 {
253 return bio_first_bvec_all(bio)->bv_page;
254 }
255
256 static inline struct folio *bio_first_folio_all(struct bio *bio)
257 {
258 return page_folio(bio_first_page_all(bio));
259 }
260
261 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
262 {
263 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
264 return &bio->bi_io_vec[bio->bi_vcnt - 1];
265 }
266
267 /**
268 * struct folio_iter - State for iterating all folios in a bio.
269 * @folio: The current folio we're iterating. NULL after the last folio.
270 * @offset: The byte offset within the current folio.
271 * @length: The number of bytes in this iteration (will not cross folio
272 * boundary).
273 */
274 struct folio_iter {
275 struct folio *folio;
276 size_t offset;
277 size_t length;
278 /* private: for use by the iterator */
279 struct folio *_next;
280 size_t _seg_count;
281 int _i;
282 };
283
284 static inline void bio_first_folio(struct folio_iter *fi, struct bio *bio,
285 int i)
286 {
287 struct bio_vec *bvec = bio_first_bvec_all(bio) + i;
288
289 if (unlikely(i >= bio->bi_vcnt)) {
290 fi->folio = NULL;
291 return;
292 }
293
294 fi->folio = page_folio(bvec->bv_page);
295 fi->offset = bvec->bv_offset +
296 PAGE_SIZE * (bvec->bv_page - &fi->folio->page);
297 fi->_seg_count = bvec->bv_len;
298 fi->length = min(folio_size(fi->folio) - fi->offset, fi->_seg_count);
299 fi->_next = folio_next(fi->folio);
300 fi->_i = i;
301 }
302
303 static inline void bio_next_folio(struct folio_iter *fi, struct bio *bio)
304 {
305 fi->_seg_count -= fi->length;
306 if (fi->_seg_count) {
307 fi->folio = fi->_next;
308 fi->offset = 0;
309 fi->length = min(folio_size(fi->folio), fi->_seg_count);
310 fi->_next = folio_next(fi->folio);
311 } else {
312 bio_first_folio(fi, bio, fi->_i + 1);
313 }
314 }
315
316 /**
317 * bio_for_each_folio_all - Iterate over each folio in a bio.
318 * @fi: struct folio_iter which is updated for each folio.
319 * @bio: struct bio to iterate over.
320 */
321 #define bio_for_each_folio_all(fi, bio) \
322 for (bio_first_folio(&fi, bio, 0); fi.folio; bio_next_folio(&fi, bio))
323
324 void bio_trim(struct bio *bio, sector_t offset, sector_t size);
325 extern struct bio *bio_split(struct bio *bio, int sectors,
326 gfp_t gfp, struct bio_set *bs);
327 struct bio *bio_split_rw(struct bio *bio, const struct queue_limits *lim,
328 unsigned *segs, struct bio_set *bs, unsigned max_bytes);
329
330 /**
331 * bio_next_split - get next @sectors from a bio, splitting if necessary
332 * @bio: bio to split
333 * @sectors: number of sectors to split from the front of @bio
334 * @gfp: gfp mask
335 * @bs: bio set to allocate from
336 *
337 * Return: a bio representing the next @sectors of @bio - if the bio is smaller
338 * than @sectors, returns the original bio unchanged.
339 */
340 static inline struct bio *bio_next_split(struct bio *bio, int sectors,
341 gfp_t gfp, struct bio_set *bs)
342 {
343 if (sectors >= bio_sectors(bio))
344 return bio;
345
346 return bio_split(bio, sectors, gfp, bs);
347 }
348
349 enum {
350 BIOSET_NEED_BVECS = BIT(0),
351 BIOSET_NEED_RESCUER = BIT(1),
352 BIOSET_PERCPU_CACHE = BIT(2),
353 };
354 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
355 extern void bioset_exit(struct bio_set *);
356 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
357
358 struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs,
359 blk_opf_t opf, gfp_t gfp_mask,
360 struct bio_set *bs);
361 struct bio *bio_kmalloc(unsigned short nr_vecs, gfp_t gfp_mask);
362 extern void bio_put(struct bio *);
363
364 struct bio *bio_alloc_clone(struct block_device *bdev, struct bio *bio_src,
365 gfp_t gfp, struct bio_set *bs);
366 int bio_init_clone(struct block_device *bdev, struct bio *bio,
367 struct bio *bio_src, gfp_t gfp);
368
369 extern struct bio_set fs_bio_set;
370
371 static inline struct bio *bio_alloc(struct block_device *bdev,
372 unsigned short nr_vecs, blk_opf_t opf, gfp_t gfp_mask)
373 {
374 return bio_alloc_bioset(bdev, nr_vecs, opf, gfp_mask, &fs_bio_set);
375 }
376
377 void submit_bio(struct bio *bio);
378
379 extern void bio_endio(struct bio *);
380
381 static inline void bio_io_error(struct bio *bio)
382 {
383 bio->bi_status = BLK_STS_IOERR;
384 bio_endio(bio);
385 }
386
387 static inline void bio_wouldblock_error(struct bio *bio)
388 {
389 bio_set_flag(bio, BIO_QUIET);
390 bio->bi_status = BLK_STS_AGAIN;
391 bio_endio(bio);
392 }
393
394 /*
395 * Calculate number of bvec segments that should be allocated to fit data
396 * pointed by @iter. If @iter is backed by bvec it's going to be reused
397 * instead of allocating a new one.
398 */
399 static inline int bio_iov_vecs_to_alloc(struct iov_iter *iter, int max_segs)
400 {
401 if (iov_iter_is_bvec(iter))
402 return 0;
403 return iov_iter_npages(iter, max_segs);
404 }
405
406 struct request_queue;
407
408 extern int submit_bio_wait(struct bio *bio);
409 void bio_init(struct bio *bio, struct block_device *bdev, struct bio_vec *table,
410 unsigned short max_vecs, blk_opf_t opf);
411 extern void bio_uninit(struct bio *);
412 void bio_reset(struct bio *bio, struct block_device *bdev, blk_opf_t opf);
413 void bio_chain(struct bio *, struct bio *);
414
415 int __must_check bio_add_page(struct bio *bio, struct page *page, unsigned len,
416 unsigned off);
417 bool __must_check bio_add_folio(struct bio *bio, struct folio *folio,
418 size_t len, size_t off);
419 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
420 unsigned int, unsigned int);
421 int bio_add_zone_append_page(struct bio *bio, struct page *page,
422 unsigned int len, unsigned int offset);
423 void __bio_add_page(struct bio *bio, struct page *page,
424 unsigned int len, unsigned int off);
425 void bio_add_folio_nofail(struct bio *bio, struct folio *folio, size_t len,
426 size_t off);
427 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
428 void bio_iov_bvec_set(struct bio *bio, struct iov_iter *iter);
429 void __bio_release_pages(struct bio *bio, bool mark_dirty);
430 extern void bio_set_pages_dirty(struct bio *bio);
431 extern void bio_check_pages_dirty(struct bio *bio);
432
433 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
434 struct bio *src, struct bvec_iter *src_iter);
435 extern void bio_copy_data(struct bio *dst, struct bio *src);
436 extern void bio_free_pages(struct bio *bio);
437 void guard_bio_eod(struct bio *bio);
438 void zero_fill_bio_iter(struct bio *bio, struct bvec_iter iter);
439
440 static inline void zero_fill_bio(struct bio *bio)
441 {
442 zero_fill_bio_iter(bio, bio->bi_iter);
443 }
444
445 static inline void bio_release_pages(struct bio *bio, bool mark_dirty)
446 {
447 if (bio_flagged(bio, BIO_PAGE_PINNED))
448 __bio_release_pages(bio, mark_dirty);
449 }
450
451 #define bio_dev(bio) \
452 disk_devt((bio)->bi_bdev->bd_disk)
453
454 #ifdef CONFIG_BLK_CGROUP
455 void bio_associate_blkg(struct bio *bio);
456 void bio_associate_blkg_from_css(struct bio *bio,
457 struct cgroup_subsys_state *css);
458 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
459 void blkcg_punt_bio_submit(struct bio *bio);
460 #else /* CONFIG_BLK_CGROUP */
461 static inline void bio_associate_blkg(struct bio *bio) { }
462 static inline void bio_associate_blkg_from_css(struct bio *bio,
463 struct cgroup_subsys_state *css)
464 { }
465 static inline void bio_clone_blkg_association(struct bio *dst,
466 struct bio *src) { }
467 static inline void blkcg_punt_bio_submit(struct bio *bio)
468 {
469 submit_bio(bio);
470 }
471 #endif /* CONFIG_BLK_CGROUP */
472
473 static inline void bio_set_dev(struct bio *bio, struct block_device *bdev)
474 {
475 bio_clear_flag(bio, BIO_REMAPPED);
476 if (bio->bi_bdev != bdev)
477 bio_clear_flag(bio, BIO_BPS_THROTTLED);
478 bio->bi_bdev = bdev;
479 bio_associate_blkg(bio);
480 }
481
482 /*
483 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
484 *
485 * A bio_list anchors a singly-linked list of bios chained through the bi_next
486 * member of the bio. The bio_list also caches the last list member to allow
487 * fast access to the tail.
488 */
489 struct bio_list {
490 struct bio *head;
491 struct bio *tail;
492 };
493
494 static inline int bio_list_empty(const struct bio_list *bl)
495 {
496 return bl->head == NULL;
497 }
498
499 static inline void bio_list_init(struct bio_list *bl)
500 {
501 bl->head = bl->tail = NULL;
502 }
503
504 #define BIO_EMPTY_LIST { NULL, NULL }
505
506 #define bio_list_for_each(bio, bl) \
507 for (bio = (bl)->head; bio; bio = bio->bi_next)
508
509 static inline unsigned bio_list_size(const struct bio_list *bl)
510 {
511 unsigned sz = 0;
512 struct bio *bio;
513
514 bio_list_for_each(bio, bl)
515 sz++;
516
517 return sz;
518 }
519
520 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
521 {
522 bio->bi_next = NULL;
523
524 if (bl->tail)
525 bl->tail->bi_next = bio;
526 else
527 bl->head = bio;
528
529 bl->tail = bio;
530 }
531
532 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
533 {
534 bio->bi_next = bl->head;
535
536 bl->head = bio;
537
538 if (!bl->tail)
539 bl->tail = bio;
540 }
541
542 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
543 {
544 if (!bl2->head)
545 return;
546
547 if (bl->tail)
548 bl->tail->bi_next = bl2->head;
549 else
550 bl->head = bl2->head;
551
552 bl->tail = bl2->tail;
553 }
554
555 static inline void bio_list_merge_init(struct bio_list *bl,
556 struct bio_list *bl2)
557 {
558 bio_list_merge(bl, bl2);
559 bio_list_init(bl2);
560 }
561
562 static inline void bio_list_merge_head(struct bio_list *bl,
563 struct bio_list *bl2)
564 {
565 if (!bl2->head)
566 return;
567
568 if (bl->head)
569 bl2->tail->bi_next = bl->head;
570 else
571 bl->tail = bl2->tail;
572
573 bl->head = bl2->head;
574 }
575
576 static inline struct bio *bio_list_peek(struct bio_list *bl)
577 {
578 return bl->head;
579 }
580
581 static inline struct bio *bio_list_pop(struct bio_list *bl)
582 {
583 struct bio *bio = bl->head;
584
585 if (bio) {
586 bl->head = bl->head->bi_next;
587 if (!bl->head)
588 bl->tail = NULL;
589
590 bio->bi_next = NULL;
591 }
592
593 return bio;
594 }
595
596 static inline struct bio *bio_list_get(struct bio_list *bl)
597 {
598 struct bio *bio = bl->head;
599
600 bl->head = bl->tail = NULL;
601
602 return bio;
603 }
604
605 /*
606 * Increment chain count for the bio. Make sure the CHAIN flag update
607 * is visible before the raised count.
608 */
609 static inline void bio_inc_remaining(struct bio *bio)
610 {
611 bio_set_flag(bio, BIO_CHAIN);
612 smp_mb__before_atomic();
613 atomic_inc(&bio->__bi_remaining);
614 }
615
616 /*
617 * bio_set is used to allow other portions of the IO system to
618 * allocate their own private memory pools for bio and iovec structures.
619 * These memory pools in turn all allocate from the bio_slab
620 * and the bvec_slabs[].
621 */
622 #define BIO_POOL_SIZE 2
623
624 struct bio_set {
625 struct kmem_cache *bio_slab;
626 unsigned int front_pad;
627
628 /*
629 * per-cpu bio alloc cache
630 */
631 struct bio_alloc_cache __percpu *cache;
632
633 mempool_t bio_pool;
634 mempool_t bvec_pool;
635 #if defined(CONFIG_BLK_DEV_INTEGRITY)
636 mempool_t bio_integrity_pool;
637 mempool_t bvec_integrity_pool;
638 #endif
639
640 unsigned int back_pad;
641 /*
642 * Deadlock avoidance for stacking block drivers: see comments in
643 * bio_alloc_bioset() for details
644 */
645 spinlock_t rescue_lock;
646 struct bio_list rescue_list;
647 struct work_struct rescue_work;
648 struct workqueue_struct *rescue_workqueue;
649
650 /*
651 * Hot un-plug notifier for the per-cpu cache, if used
652 */
653 struct hlist_node cpuhp_dead;
654 };
655
656 static inline bool bioset_initialized(struct bio_set *bs)
657 {
658 return bs->bio_slab != NULL;
659 }
660
661 /*
662 * Mark a bio as polled. Note that for async polled IO, the caller must
663 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
664 * We cannot block waiting for requests on polled IO, as those completions
665 * must be found by the caller. This is different than IRQ driven IO, where
666 * it's safe to wait for IO to complete.
667 */
668 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
669 {
670 bio->bi_opf |= REQ_POLLED;
671 if (kiocb->ki_flags & IOCB_NOWAIT)
672 bio->bi_opf |= REQ_NOWAIT;
673 }
674
675 static inline void bio_clear_polled(struct bio *bio)
676 {
677 bio->bi_opf &= ~REQ_POLLED;
678 }
679
680 struct bio *blk_next_bio(struct bio *bio, struct block_device *bdev,
681 unsigned int nr_pages, blk_opf_t opf, gfp_t gfp);
682 struct bio *bio_chain_and_submit(struct bio *prev, struct bio *new);
683
684 struct bio *blk_alloc_discard_bio(struct block_device *bdev,
685 sector_t *sector, sector_t *nr_sects, gfp_t gfp_mask);
686
687 #endif /* __LINUX_BIO_H */
688
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.