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_ZEROES;
71 static inline void *bio_data(struct bio *bio)
73 if (bio_has_data(bio))
74 return page_address(bio_page(bio)) + bio_offset(bio);
79 static inline bool bio_next_segment(const struct bio *bio,
80 struct bvec_iter_all *iter)
82 if (iter->idx >= bio->bi_vcnt)
85 bvec_advance(&bio->bi_io_vec[iter->idx], iter);
90 * drivers should _never_ use the all version - the bio may have been split
91 * before it got to the driver and the driver won't own all of it
93 #define bio_for_each_segment_all(bvl, bio, iter) \
94 for (bvl = bvec_init_iter_all(&iter); bio_next_segment((bio), &iter); )
96 static inline void bio_advance_iter(const struct bio *bio,
97 struct bvec_iter *iter, unsigned int bytes)
99 iter->bi_sector += bytes >> 9;
101 if (bio_no_advance_iter(bio))
102 iter->bi_size -= bytes;
104 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
105 /* TODO: It is reasonable to complete bio with error here. */
108 /* @bytes should be less or equal to bvec[i->bi_idx].bv_len */
109 static inline void bio_advance_iter_single(const struct bio *bio,
110 struct bvec_iter *iter,
113 iter->bi_sector += bytes >> 9;
115 if (bio_no_advance_iter(bio))
116 iter->bi_size -= bytes;
118 bvec_iter_advance_single(bio->bi_io_vec, iter, bytes);
121 void __bio_advance(struct bio *, unsigned bytes);
124 * bio_advance - increment/complete a bio by some number of bytes
125 * @bio: bio to advance
126 * @nbytes: number of bytes to complete
128 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
129 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
130 * be updated on the last bvec as well.
132 * @bio will then represent the remaining, uncompleted portion of the io.
134 static inline void bio_advance(struct bio *bio, unsigned int nbytes)
136 if (nbytes == bio->bi_iter.bi_size) {
137 bio->bi_iter.bi_size = 0;
140 __bio_advance(bio, nbytes);
143 #define __bio_for_each_segment(bvl, bio, iter, start) \
144 for (iter = (start); \
146 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
147 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
149 #define bio_for_each_segment(bvl, bio, iter) \
150 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
152 #define __bio_for_each_bvec(bvl, bio, iter, start) \
153 for (iter = (start); \
155 ((bvl = mp_bvec_iter_bvec((bio)->bi_io_vec, (iter))), 1); \
156 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
158 /* iterate over multi-page bvec */
159 #define bio_for_each_bvec(bvl, bio, iter) \
160 __bio_for_each_bvec(bvl, bio, iter, (bio)->bi_iter)
163 * Iterate over all multi-page bvecs. Drivers shouldn't use this version for the
164 * same reasons as bio_for_each_segment_all().
166 #define bio_for_each_bvec_all(bvl, bio, i) \
167 for (i = 0, bvl = bio_first_bvec_all(bio); \
168 i < (bio)->bi_vcnt; i++, bvl++)
170 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
172 static inline unsigned bio_segments(struct bio *bio)
176 struct bvec_iter iter;
179 * We special case discard/write same/write zeroes, because they
180 * interpret bi_size differently:
183 switch (bio_op(bio)) {
185 case REQ_OP_SECURE_ERASE:
186 case REQ_OP_WRITE_ZEROES:
192 bio_for_each_segment(bv, bio, iter)
199 * get a reference to a bio, so it won't disappear. the intended use is
203 * submit_bio(rw, bio);
204 * if (bio->bi_flags ...)
208 * without the bio_get(), it could potentially complete I/O before submit_bio
209 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
212 static inline void bio_get(struct bio *bio)
214 bio->bi_flags |= (1 << BIO_REFFED);
215 smp_mb__before_atomic();
216 atomic_inc(&bio->__bi_cnt);
219 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
222 bio->bi_flags |= (1 << BIO_REFFED);
225 atomic_set(&bio->__bi_cnt, count);
228 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
230 return (bio->bi_flags & (1U << bit)) != 0;
233 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
235 bio->bi_flags |= (1U << bit);
238 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
240 bio->bi_flags &= ~(1U << bit);
243 static inline struct bio_vec *bio_first_bvec_all(struct bio *bio)
245 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
246 return bio->bi_io_vec;
249 static inline struct page *bio_first_page_all(struct bio *bio)
251 return bio_first_bvec_all(bio)->bv_page;
254 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
256 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
257 return &bio->bi_io_vec[bio->bi_vcnt - 1];
261 * struct folio_iter - State for iterating all folios in a bio.
262 * @folio: The current folio we're iterating. NULL after the last folio.
263 * @offset: The byte offset within the current folio.
264 * @length: The number of bytes in this iteration (will not cross folio
271 /* private: for use by the iterator */
277 static inline void bio_first_folio(struct folio_iter *fi, struct bio *bio,
280 struct bio_vec *bvec = bio_first_bvec_all(bio) + i;
282 fi->folio = page_folio(bvec->bv_page);
283 fi->offset = bvec->bv_offset +
284 PAGE_SIZE * (bvec->bv_page - &fi->folio->page);
285 fi->_seg_count = bvec->bv_len;
286 fi->length = min(folio_size(fi->folio) - fi->offset, fi->_seg_count);
287 fi->_next = folio_next(fi->folio);
291 static inline void bio_next_folio(struct folio_iter *fi, struct bio *bio)
293 fi->_seg_count -= fi->length;
294 if (fi->_seg_count) {
295 fi->folio = fi->_next;
297 fi->length = min(folio_size(fi->folio), fi->_seg_count);
298 fi->_next = folio_next(fi->folio);
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);
407 struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs,
408 blk_opf_t opf, gfp_t gfp_mask,
410 struct bio *bio_kmalloc(unsigned short nr_vecs, gfp_t gfp_mask);
411 extern void bio_put(struct bio *);
413 struct bio *bio_alloc_clone(struct block_device *bdev, struct bio *bio_src,
414 gfp_t gfp, struct bio_set *bs);
415 int bio_init_clone(struct block_device *bdev, struct bio *bio,
416 struct bio *bio_src, gfp_t gfp);
418 extern struct bio_set fs_bio_set;
420 static inline struct bio *bio_alloc(struct block_device *bdev,
421 unsigned short nr_vecs, blk_opf_t opf, gfp_t gfp_mask)
423 return bio_alloc_bioset(bdev, nr_vecs, opf, gfp_mask, &fs_bio_set);
426 void submit_bio(struct bio *bio);
428 extern void bio_endio(struct bio *);
430 static inline void bio_io_error(struct bio *bio)
432 bio->bi_status = BLK_STS_IOERR;
436 static inline void bio_wouldblock_error(struct bio *bio)
438 bio_set_flag(bio, BIO_QUIET);
439 bio->bi_status = BLK_STS_AGAIN;
444 * Calculate number of bvec segments that should be allocated to fit data
445 * pointed by @iter. If @iter is backed by bvec it's going to be reused
446 * instead of allocating a new one.
448 static inline int bio_iov_vecs_to_alloc(struct iov_iter *iter, int max_segs)
450 if (iov_iter_is_bvec(iter))
452 return iov_iter_npages(iter, max_segs);
455 struct request_queue;
457 extern int submit_bio_wait(struct bio *bio);
458 void bio_init(struct bio *bio, struct block_device *bdev, struct bio_vec *table,
459 unsigned short max_vecs, blk_opf_t opf);
460 extern void bio_uninit(struct bio *);
461 void bio_reset(struct bio *bio, struct block_device *bdev, blk_opf_t opf);
462 void bio_chain(struct bio *, struct bio *);
464 int bio_add_page(struct bio *, struct page *, unsigned len, unsigned off);
465 bool bio_add_folio(struct bio *, struct folio *, size_t len, size_t off);
466 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
467 unsigned int, unsigned int);
468 int bio_add_zone_append_page(struct bio *bio, struct page *page,
469 unsigned int len, unsigned int offset);
470 void __bio_add_page(struct bio *bio, struct page *page,
471 unsigned int len, unsigned int off);
472 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
473 void bio_iov_bvec_set(struct bio *bio, struct iov_iter *iter);
474 void __bio_release_pages(struct bio *bio, bool mark_dirty);
475 extern void bio_set_pages_dirty(struct bio *bio);
476 extern void bio_check_pages_dirty(struct bio *bio);
478 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
479 struct bio *src, struct bvec_iter *src_iter);
480 extern void bio_copy_data(struct bio *dst, struct bio *src);
481 extern void bio_free_pages(struct bio *bio);
482 void guard_bio_eod(struct bio *bio);
483 void zero_fill_bio(struct bio *bio);
485 static inline void bio_release_pages(struct bio *bio, bool mark_dirty)
487 if (!bio_flagged(bio, BIO_NO_PAGE_REF))
488 __bio_release_pages(bio, mark_dirty);
491 #define bio_dev(bio) \
492 disk_devt((bio)->bi_bdev->bd_disk)
494 #ifdef CONFIG_BLK_CGROUP
495 void bio_associate_blkg(struct bio *bio);
496 void bio_associate_blkg_from_css(struct bio *bio,
497 struct cgroup_subsys_state *css);
498 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
499 #else /* CONFIG_BLK_CGROUP */
500 static inline void bio_associate_blkg(struct bio *bio) { }
501 static inline void bio_associate_blkg_from_css(struct bio *bio,
502 struct cgroup_subsys_state *css)
504 static inline void bio_clone_blkg_association(struct bio *dst,
506 #endif /* CONFIG_BLK_CGROUP */
508 static inline void bio_set_dev(struct bio *bio, struct block_device *bdev)
510 bio_clear_flag(bio, BIO_REMAPPED);
511 if (bio->bi_bdev != bdev)
512 bio_clear_flag(bio, BIO_BPS_THROTTLED);
514 bio_associate_blkg(bio);
518 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
520 * A bio_list anchors a singly-linked list of bios chained through the bi_next
521 * member of the bio. The bio_list also caches the last list member to allow
522 * fast access to the tail.
529 static inline int bio_list_empty(const struct bio_list *bl)
531 return bl->head == NULL;
534 static inline void bio_list_init(struct bio_list *bl)
536 bl->head = bl->tail = NULL;
539 #define BIO_EMPTY_LIST { NULL, NULL }
541 #define bio_list_for_each(bio, bl) \
542 for (bio = (bl)->head; bio; bio = bio->bi_next)
544 static inline unsigned bio_list_size(const struct bio_list *bl)
549 bio_list_for_each(bio, bl)
555 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
560 bl->tail->bi_next = bio;
567 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
569 bio->bi_next = bl->head;
577 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
583 bl->tail->bi_next = bl2->head;
585 bl->head = bl2->head;
587 bl->tail = bl2->tail;
590 static inline void bio_list_merge_head(struct bio_list *bl,
591 struct bio_list *bl2)
597 bl2->tail->bi_next = bl->head;
599 bl->tail = bl2->tail;
601 bl->head = bl2->head;
604 static inline struct bio *bio_list_peek(struct bio_list *bl)
609 static inline struct bio *bio_list_pop(struct bio_list *bl)
611 struct bio *bio = bl->head;
614 bl->head = bl->head->bi_next;
624 static inline struct bio *bio_list_get(struct bio_list *bl)
626 struct bio *bio = bl->head;
628 bl->head = bl->tail = NULL;
634 * Increment chain count for the bio. Make sure the CHAIN flag update
635 * is visible before the raised count.
637 static inline void bio_inc_remaining(struct bio *bio)
639 bio_set_flag(bio, BIO_CHAIN);
640 smp_mb__before_atomic();
641 atomic_inc(&bio->__bi_remaining);
645 * bio_set is used to allow other portions of the IO system to
646 * allocate their own private memory pools for bio and iovec structures.
647 * These memory pools in turn all allocate from the bio_slab
648 * and the bvec_slabs[].
650 #define BIO_POOL_SIZE 2
653 struct kmem_cache *bio_slab;
654 unsigned int front_pad;
657 * per-cpu bio alloc cache
659 struct bio_alloc_cache __percpu *cache;
663 #if defined(CONFIG_BLK_DEV_INTEGRITY)
664 mempool_t bio_integrity_pool;
665 mempool_t bvec_integrity_pool;
668 unsigned int back_pad;
670 * Deadlock avoidance for stacking block drivers: see comments in
671 * bio_alloc_bioset() for details
673 spinlock_t rescue_lock;
674 struct bio_list rescue_list;
675 struct work_struct rescue_work;
676 struct workqueue_struct *rescue_workqueue;
679 * Hot un-plug notifier for the per-cpu cache, if used
681 struct hlist_node cpuhp_dead;
684 static inline bool bioset_initialized(struct bio_set *bs)
686 return bs->bio_slab != NULL;
689 #if defined(CONFIG_BLK_DEV_INTEGRITY)
691 #define bip_for_each_vec(bvl, bip, iter) \
692 for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter)
694 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \
696 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter)
698 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
699 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
700 extern bool bio_integrity_prep(struct bio *);
701 extern void bio_integrity_advance(struct bio *, unsigned int);
702 extern void bio_integrity_trim(struct bio *);
703 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
704 extern int bioset_integrity_create(struct bio_set *, int);
705 extern void bioset_integrity_free(struct bio_set *);
706 extern void bio_integrity_init(void);
708 #else /* CONFIG_BLK_DEV_INTEGRITY */
710 static inline void *bio_integrity(struct bio *bio)
715 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size)
720 static inline void bioset_integrity_free (struct bio_set *bs)
725 static inline bool bio_integrity_prep(struct bio *bio)
730 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
736 static inline void bio_integrity_advance(struct bio *bio,
737 unsigned int bytes_done)
742 static inline void bio_integrity_trim(struct bio *bio)
747 static inline void bio_integrity_init(void)
752 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
757 static inline void *bio_integrity_alloc(struct bio * bio, gfp_t gfp,
760 return ERR_PTR(-EINVAL);
763 static inline int bio_integrity_add_page(struct bio *bio, struct page *page,
764 unsigned int len, unsigned int offset)
769 #endif /* CONFIG_BLK_DEV_INTEGRITY */
772 * Mark a bio as polled. Note that for async polled IO, the caller must
773 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
774 * We cannot block waiting for requests on polled IO, as those completions
775 * must be found by the caller. This is different than IRQ driven IO, where
776 * it's safe to wait for IO to complete.
778 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
780 bio->bi_opf |= REQ_POLLED;
781 if (!is_sync_kiocb(kiocb))
782 bio->bi_opf |= REQ_NOWAIT;
785 static inline void bio_clear_polled(struct bio *bio)
787 /* can't support alloc cache if we turn off polling */
788 bio->bi_opf &= ~(REQ_POLLED | REQ_ALLOC_CACHE);
791 struct bio *blk_next_bio(struct bio *bio, struct block_device *bdev,
792 unsigned int nr_pages, blk_opf_t opf, gfp_t gfp);
794 #endif /* __LINUX_BIO_H */