1 /* SPDX-License-Identifier: GPL-2.0 */
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>
13 #define BIO_MAX_VECS 256U
15 static inline unsigned int bio_max_segs(unsigned int nr_segs)
17 return min(nr_segs, BIO_MAX_VECS);
20 #define bio_prio(bio) (bio)->bi_ioprio
21 #define bio_set_prio(bio, prio) ((bio)->bi_ioprio = prio)
23 #define bio_iter_iovec(bio, iter) \
24 bvec_iter_bvec((bio)->bi_io_vec, (iter))
26 #define bio_iter_page(bio, iter) \
27 bvec_iter_page((bio)->bi_io_vec, (iter))
28 #define bio_iter_len(bio, iter) \
29 bvec_iter_len((bio)->bi_io_vec, (iter))
30 #define bio_iter_offset(bio, iter) \
31 bvec_iter_offset((bio)->bi_io_vec, (iter))
33 #define bio_page(bio) bio_iter_page((bio), (bio)->bi_iter)
34 #define bio_offset(bio) bio_iter_offset((bio), (bio)->bi_iter)
35 #define bio_iovec(bio) bio_iter_iovec((bio), (bio)->bi_iter)
37 #define bvec_iter_sectors(iter) ((iter).bi_size >> 9)
38 #define bvec_iter_end_sector(iter) ((iter).bi_sector + bvec_iter_sectors((iter)))
40 #define bio_sectors(bio) bvec_iter_sectors((bio)->bi_iter)
41 #define bio_end_sector(bio) bvec_iter_end_sector((bio)->bi_iter)
44 * Return the data direction, READ or WRITE.
46 #define bio_data_dir(bio) \
47 (op_is_write(bio_op(bio)) ? WRITE : READ)
50 * Check whether this bio carries any data or not. A NULL bio is allowed.
52 static inline bool bio_has_data(struct bio *bio)
55 bio->bi_iter.bi_size &&
56 bio_op(bio) != REQ_OP_DISCARD &&
57 bio_op(bio) != REQ_OP_SECURE_ERASE &&
58 bio_op(bio) != REQ_OP_WRITE_ZEROES)
64 static inline bool bio_no_advance_iter(const struct bio *bio)
66 return bio_op(bio) == REQ_OP_DISCARD ||
67 bio_op(bio) == REQ_OP_SECURE_ERASE ||
68 bio_op(bio) == REQ_OP_WRITE_SAME ||
69 bio_op(bio) == REQ_OP_WRITE_ZEROES;
72 static inline void *bio_data(struct bio *bio)
74 if (bio_has_data(bio))
75 return page_address(bio_page(bio)) + bio_offset(bio);
80 static inline bool bio_next_segment(const struct bio *bio,
81 struct bvec_iter_all *iter)
83 if (iter->idx >= bio->bi_vcnt)
86 bvec_advance(&bio->bi_io_vec[iter->idx], iter);
91 * drivers should _never_ use the all version - the bio may have been split
92 * before it got to the driver and the driver won't own all of it
94 #define bio_for_each_segment_all(bvl, bio, iter) \
95 for (bvl = bvec_init_iter_all(&iter); bio_next_segment((bio), &iter); )
97 static inline void bio_advance_iter(const struct bio *bio,
98 struct bvec_iter *iter, unsigned int bytes)
100 iter->bi_sector += bytes >> 9;
102 if (bio_no_advance_iter(bio))
103 iter->bi_size -= bytes;
105 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
106 /* TODO: It is reasonable to complete bio with error here. */
109 /* @bytes should be less or equal to bvec[i->bi_idx].bv_len */
110 static inline void bio_advance_iter_single(const struct bio *bio,
111 struct bvec_iter *iter,
114 iter->bi_sector += bytes >> 9;
116 if (bio_no_advance_iter(bio))
117 iter->bi_size -= bytes;
119 bvec_iter_advance_single(bio->bi_io_vec, iter, bytes);
122 void __bio_advance(struct bio *, unsigned bytes);
125 * bio_advance - increment/complete a bio by some number of bytes
126 * @bio: bio to advance
127 * @nbytes: number of bytes to complete
129 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
130 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
131 * be updated on the last bvec as well.
133 * @bio will then represent the remaining, uncompleted portion of the io.
135 static inline void bio_advance(struct bio *bio, unsigned int nbytes)
137 if (nbytes == bio->bi_iter.bi_size) {
138 bio->bi_iter.bi_size = 0;
141 __bio_advance(bio, nbytes);
144 #define __bio_for_each_segment(bvl, bio, iter, start) \
145 for (iter = (start); \
147 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
148 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
150 #define bio_for_each_segment(bvl, bio, iter) \
151 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
153 #define __bio_for_each_bvec(bvl, bio, iter, start) \
154 for (iter = (start); \
156 ((bvl = mp_bvec_iter_bvec((bio)->bi_io_vec, (iter))), 1); \
157 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
159 /* iterate over multi-page bvec */
160 #define bio_for_each_bvec(bvl, bio, iter) \
161 __bio_for_each_bvec(bvl, bio, iter, (bio)->bi_iter)
164 * Iterate over all multi-page bvecs. Drivers shouldn't use this version for the
165 * same reasons as bio_for_each_segment_all().
167 #define bio_for_each_bvec_all(bvl, bio, i) \
168 for (i = 0, bvl = bio_first_bvec_all(bio); \
169 i < (bio)->bi_vcnt; i++, bvl++)
171 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
173 static inline unsigned bio_segments(struct bio *bio)
177 struct bvec_iter iter;
180 * We special case discard/write same/write zeroes, because they
181 * interpret bi_size differently:
184 switch (bio_op(bio)) {
186 case REQ_OP_SECURE_ERASE:
187 case REQ_OP_WRITE_ZEROES:
189 case REQ_OP_WRITE_SAME:
195 bio_for_each_segment(bv, bio, iter)
202 * get a reference to a bio, so it won't disappear. the intended use is
206 * submit_bio(rw, bio);
207 * if (bio->bi_flags ...)
211 * without the bio_get(), it could potentially complete I/O before submit_bio
212 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
215 static inline void bio_get(struct bio *bio)
217 bio->bi_flags |= (1 << BIO_REFFED);
218 smp_mb__before_atomic();
219 atomic_inc(&bio->__bi_cnt);
222 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
225 bio->bi_flags |= (1 << BIO_REFFED);
228 atomic_set(&bio->__bi_cnt, count);
231 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
233 return (bio->bi_flags & (1U << bit)) != 0;
236 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
238 bio->bi_flags |= (1U << bit);
241 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
243 bio->bi_flags &= ~(1U << bit);
246 static inline struct bio_vec *bio_first_bvec_all(struct bio *bio)
248 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
249 return bio->bi_io_vec;
252 static inline struct page *bio_first_page_all(struct bio *bio)
254 return bio_first_bvec_all(bio)->bv_page;
257 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
259 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
260 return &bio->bi_io_vec[bio->bi_vcnt - 1];
264 * struct folio_iter - State for iterating all folios in a bio.
265 * @folio: The current folio we're iterating. NULL after the last folio.
266 * @offset: The byte offset within the current folio.
267 * @length: The number of bytes in this iteration (will not cross folio
274 /* private: for use by the iterator */
279 static inline void bio_first_folio(struct folio_iter *fi, struct bio *bio,
282 struct bio_vec *bvec = bio_first_bvec_all(bio) + i;
284 fi->folio = page_folio(bvec->bv_page);
285 fi->offset = bvec->bv_offset +
286 PAGE_SIZE * (bvec->bv_page - &fi->folio->page);
287 fi->_seg_count = bvec->bv_len;
288 fi->length = min(folio_size(fi->folio) - fi->offset, fi->_seg_count);
292 static inline void bio_next_folio(struct folio_iter *fi, struct bio *bio)
294 fi->_seg_count -= fi->length;
295 if (fi->_seg_count) {
296 fi->folio = folio_next(fi->folio);
298 fi->length = min(folio_size(fi->folio), fi->_seg_count);
299 } else if (fi->_i + 1 < bio->bi_vcnt) {
300 bio_first_folio(fi, bio, fi->_i + 1);
307 * bio_for_each_folio_all - Iterate over each folio in a bio.
308 * @fi: struct folio_iter which is updated for each folio.
309 * @bio: struct bio to iterate over.
311 #define bio_for_each_folio_all(fi, bio) \
312 for (bio_first_folio(&fi, bio, 0); fi.folio; bio_next_folio(&fi, bio))
315 BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */
316 BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */
317 BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */
318 BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */
319 BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */
323 * bio integrity payload
325 struct bio_integrity_payload {
326 struct bio *bip_bio; /* parent bio */
328 struct bvec_iter bip_iter;
330 unsigned short bip_vcnt; /* # of integrity bio_vecs */
331 unsigned short bip_max_vcnt; /* integrity bio_vec slots */
332 unsigned short bip_flags; /* control flags */
334 struct bvec_iter bio_iter; /* for rewinding parent bio */
336 struct work_struct bip_work; /* I/O completion */
338 struct bio_vec *bip_vec;
339 struct bio_vec bip_inline_vecs[];/* embedded bvec array */
342 #if defined(CONFIG_BLK_DEV_INTEGRITY)
344 static inline struct bio_integrity_payload *bio_integrity(struct bio *bio)
346 if (bio->bi_opf & REQ_INTEGRITY)
347 return bio->bi_integrity;
352 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
354 struct bio_integrity_payload *bip = bio_integrity(bio);
357 return bip->bip_flags & flag;
362 static inline sector_t bip_get_seed(struct bio_integrity_payload *bip)
364 return bip->bip_iter.bi_sector;
367 static inline void bip_set_seed(struct bio_integrity_payload *bip,
370 bip->bip_iter.bi_sector = seed;
373 #endif /* CONFIG_BLK_DEV_INTEGRITY */
375 void bio_trim(struct bio *bio, sector_t offset, sector_t size);
376 extern struct bio *bio_split(struct bio *bio, int sectors,
377 gfp_t gfp, struct bio_set *bs);
380 * bio_next_split - get next @sectors from a bio, splitting if necessary
382 * @sectors: number of sectors to split from the front of @bio
384 * @bs: bio set to allocate from
386 * Return: a bio representing the next @sectors of @bio - if the bio is smaller
387 * than @sectors, returns the original bio unchanged.
389 static inline struct bio *bio_next_split(struct bio *bio, int sectors,
390 gfp_t gfp, struct bio_set *bs)
392 if (sectors >= bio_sectors(bio))
395 return bio_split(bio, sectors, gfp, bs);
399 BIOSET_NEED_BVECS = BIT(0),
400 BIOSET_NEED_RESCUER = BIT(1),
401 BIOSET_PERCPU_CACHE = BIT(2),
403 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
404 extern void bioset_exit(struct bio_set *);
405 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
406 extern int bioset_init_from_src(struct bio_set *bs, struct bio_set *src);
408 struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs,
409 unsigned int opf, gfp_t gfp_mask,
411 struct bio *bio_alloc_kiocb(struct kiocb *kiocb, struct block_device *bdev,
412 unsigned short nr_vecs, unsigned int opf, struct bio_set *bs);
413 struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned short nr_iovecs);
414 extern void bio_put(struct bio *);
416 struct bio *bio_alloc_clone(struct block_device *bdev, struct bio *bio_src,
417 gfp_t gfp, struct bio_set *bs);
418 int bio_init_clone(struct block_device *bdev, struct bio *bio,
419 struct bio *bio_src, gfp_t gfp);
421 extern struct bio_set fs_bio_set;
423 static inline struct bio *bio_alloc(struct block_device *bdev,
424 unsigned short nr_vecs, unsigned int opf, gfp_t gfp_mask)
426 return bio_alloc_bioset(bdev, nr_vecs, opf, gfp_mask, &fs_bio_set);
429 void submit_bio(struct bio *bio);
431 extern void bio_endio(struct bio *);
433 static inline void bio_io_error(struct bio *bio)
435 bio->bi_status = BLK_STS_IOERR;
439 static inline void bio_wouldblock_error(struct bio *bio)
441 bio_set_flag(bio, BIO_QUIET);
442 bio->bi_status = BLK_STS_AGAIN;
447 * Calculate number of bvec segments that should be allocated to fit data
448 * pointed by @iter. If @iter is backed by bvec it's going to be reused
449 * instead of allocating a new one.
451 static inline int bio_iov_vecs_to_alloc(struct iov_iter *iter, int max_segs)
453 if (iov_iter_is_bvec(iter))
455 return iov_iter_npages(iter, max_segs);
458 struct request_queue;
460 extern int submit_bio_wait(struct bio *bio);
461 void bio_init(struct bio *bio, struct block_device *bdev, struct bio_vec *table,
462 unsigned short max_vecs, unsigned int opf);
463 extern void bio_uninit(struct bio *);
464 void bio_reset(struct bio *bio, struct block_device *bdev, unsigned int opf);
465 void bio_chain(struct bio *, struct bio *);
467 int bio_add_page(struct bio *, struct page *, unsigned len, unsigned off);
468 bool bio_add_folio(struct bio *, struct folio *, size_t len, size_t off);
469 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
470 unsigned int, unsigned int);
471 int bio_add_zone_append_page(struct bio *bio, struct page *page,
472 unsigned int len, unsigned int offset);
473 void __bio_add_page(struct bio *bio, struct page *page,
474 unsigned int len, unsigned int off);
475 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
476 void bio_iov_bvec_set(struct bio *bio, struct iov_iter *iter);
477 void __bio_release_pages(struct bio *bio, bool mark_dirty);
478 extern void bio_set_pages_dirty(struct bio *bio);
479 extern void bio_check_pages_dirty(struct bio *bio);
481 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
482 struct bio *src, struct bvec_iter *src_iter);
483 extern void bio_copy_data(struct bio *dst, struct bio *src);
484 extern void bio_free_pages(struct bio *bio);
485 void guard_bio_eod(struct bio *bio);
486 void zero_fill_bio(struct bio *bio);
488 static inline void bio_release_pages(struct bio *bio, bool mark_dirty)
490 if (!bio_flagged(bio, BIO_NO_PAGE_REF))
491 __bio_release_pages(bio, mark_dirty);
494 extern const char *bio_devname(struct bio *bio, char *buffer);
496 #define bio_dev(bio) \
497 disk_devt((bio)->bi_bdev->bd_disk)
499 #ifdef CONFIG_BLK_CGROUP
500 void bio_associate_blkg(struct bio *bio);
501 void bio_associate_blkg_from_css(struct bio *bio,
502 struct cgroup_subsys_state *css);
503 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
504 #else /* CONFIG_BLK_CGROUP */
505 static inline void bio_associate_blkg(struct bio *bio) { }
506 static inline void bio_associate_blkg_from_css(struct bio *bio,
507 struct cgroup_subsys_state *css)
509 static inline void bio_clone_blkg_association(struct bio *dst,
511 #endif /* CONFIG_BLK_CGROUP */
513 static inline void bio_set_dev(struct bio *bio, struct block_device *bdev)
515 bio_clear_flag(bio, BIO_REMAPPED);
516 if (bio->bi_bdev != bdev)
517 bio_clear_flag(bio, BIO_THROTTLED);
519 bio_associate_blkg(bio);
523 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
525 * A bio_list anchors a singly-linked list of bios chained through the bi_next
526 * member of the bio. The bio_list also caches the last list member to allow
527 * fast access to the tail.
534 static inline int bio_list_empty(const struct bio_list *bl)
536 return bl->head == NULL;
539 static inline void bio_list_init(struct bio_list *bl)
541 bl->head = bl->tail = NULL;
544 #define BIO_EMPTY_LIST { NULL, NULL }
546 #define bio_list_for_each(bio, bl) \
547 for (bio = (bl)->head; bio; bio = bio->bi_next)
549 static inline unsigned bio_list_size(const struct bio_list *bl)
554 bio_list_for_each(bio, bl)
560 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
565 bl->tail->bi_next = bio;
572 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
574 bio->bi_next = bl->head;
582 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
588 bl->tail->bi_next = bl2->head;
590 bl->head = bl2->head;
592 bl->tail = bl2->tail;
595 static inline void bio_list_merge_head(struct bio_list *bl,
596 struct bio_list *bl2)
602 bl2->tail->bi_next = bl->head;
604 bl->tail = bl2->tail;
606 bl->head = bl2->head;
609 static inline struct bio *bio_list_peek(struct bio_list *bl)
614 static inline struct bio *bio_list_pop(struct bio_list *bl)
616 struct bio *bio = bl->head;
619 bl->head = bl->head->bi_next;
629 static inline struct bio *bio_list_get(struct bio_list *bl)
631 struct bio *bio = bl->head;
633 bl->head = bl->tail = NULL;
639 * Increment chain count for the bio. Make sure the CHAIN flag update
640 * is visible before the raised count.
642 static inline void bio_inc_remaining(struct bio *bio)
644 bio_set_flag(bio, BIO_CHAIN);
645 smp_mb__before_atomic();
646 atomic_inc(&bio->__bi_remaining);
650 * bio_set is used to allow other portions of the IO system to
651 * allocate their own private memory pools for bio and iovec structures.
652 * These memory pools in turn all allocate from the bio_slab
653 * and the bvec_slabs[].
655 #define BIO_POOL_SIZE 2
658 struct kmem_cache *bio_slab;
659 unsigned int front_pad;
662 * per-cpu bio alloc cache
664 struct bio_alloc_cache __percpu *cache;
668 #if defined(CONFIG_BLK_DEV_INTEGRITY)
669 mempool_t bio_integrity_pool;
670 mempool_t bvec_integrity_pool;
673 unsigned int back_pad;
675 * Deadlock avoidance for stacking block drivers: see comments in
676 * bio_alloc_bioset() for details
678 spinlock_t rescue_lock;
679 struct bio_list rescue_list;
680 struct work_struct rescue_work;
681 struct workqueue_struct *rescue_workqueue;
684 * Hot un-plug notifier for the per-cpu cache, if used
686 struct hlist_node cpuhp_dead;
689 static inline bool bioset_initialized(struct bio_set *bs)
691 return bs->bio_slab != NULL;
694 #if defined(CONFIG_BLK_DEV_INTEGRITY)
696 #define bip_for_each_vec(bvl, bip, iter) \
697 for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter)
699 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \
701 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter)
703 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
704 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
705 extern bool bio_integrity_prep(struct bio *);
706 extern void bio_integrity_advance(struct bio *, unsigned int);
707 extern void bio_integrity_trim(struct bio *);
708 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
709 extern int bioset_integrity_create(struct bio_set *, int);
710 extern void bioset_integrity_free(struct bio_set *);
711 extern void bio_integrity_init(void);
713 #else /* CONFIG_BLK_DEV_INTEGRITY */
715 static inline void *bio_integrity(struct bio *bio)
720 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size)
725 static inline void bioset_integrity_free (struct bio_set *bs)
730 static inline bool bio_integrity_prep(struct bio *bio)
735 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
741 static inline void bio_integrity_advance(struct bio *bio,
742 unsigned int bytes_done)
747 static inline void bio_integrity_trim(struct bio *bio)
752 static inline void bio_integrity_init(void)
757 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
762 static inline void *bio_integrity_alloc(struct bio * bio, gfp_t gfp,
765 return ERR_PTR(-EINVAL);
768 static inline int bio_integrity_add_page(struct bio *bio, struct page *page,
769 unsigned int len, unsigned int offset)
774 #endif /* CONFIG_BLK_DEV_INTEGRITY */
777 * Mark a bio as polled. Note that for async polled IO, the caller must
778 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
779 * We cannot block waiting for requests on polled IO, as those completions
780 * must be found by the caller. This is different than IRQ driven IO, where
781 * it's safe to wait for IO to complete.
783 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
785 bio->bi_opf |= REQ_POLLED;
786 if (!is_sync_kiocb(kiocb))
787 bio->bi_opf |= REQ_NOWAIT;
790 struct bio *blk_next_bio(struct bio *bio, struct block_device *bdev,
791 unsigned int nr_pages, unsigned int opf, gfp_t gfp);
793 #endif /* __LINUX_BIO_H */