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
17 static inline unsigned int bio_max_segs(unsigned int nr_segs)
19 return min(nr_segs, BIO_MAX_VECS);
22 #define bio_prio(bio) (bio)->bi_ioprio
23 #define bio_set_prio(bio, prio) ((bio)->bi_ioprio = prio)
25 #define bio_iter_iovec(bio, iter) \
26 bvec_iter_bvec((bio)->bi_io_vec, (iter))
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))
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)
39 #define bvec_iter_sectors(iter) ((iter).bi_size >> 9)
40 #define bvec_iter_end_sector(iter) ((iter).bi_sector + bvec_iter_sectors((iter)))
42 #define bio_sectors(bio) bvec_iter_sectors((bio)->bi_iter)
43 #define bio_end_sector(bio) bvec_iter_end_sector((bio)->bi_iter)
46 * Return the data direction, READ or WRITE.
48 #define bio_data_dir(bio) \
49 (op_is_write(bio_op(bio)) ? WRITE : READ)
52 * Check whether this bio carries any data or not. A NULL bio is allowed.
54 static inline bool bio_has_data(struct bio *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)
66 static inline bool bio_no_advance_iter(const struct bio *bio)
68 return bio_op(bio) == REQ_OP_DISCARD ||
69 bio_op(bio) == REQ_OP_SECURE_ERASE ||
70 bio_op(bio) == REQ_OP_WRITE_ZEROES;
73 static inline void *bio_data(struct bio *bio)
75 if (bio_has_data(bio))
76 return page_address(bio_page(bio)) + bio_offset(bio);
81 static inline bool bio_next_segment(const struct bio *bio,
82 struct bvec_iter_all *iter)
84 if (iter->idx >= bio->bi_vcnt)
87 bvec_advance(&bio->bi_io_vec[iter->idx], iter);
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
95 #define bio_for_each_segment_all(bvl, bio, iter) \
96 for (bvl = bvec_init_iter_all(&iter); bio_next_segment((bio), &iter); )
98 static inline void bio_advance_iter(const struct bio *bio,
99 struct bvec_iter *iter, unsigned int bytes)
101 iter->bi_sector += bytes >> 9;
103 if (bio_no_advance_iter(bio))
104 iter->bi_size -= bytes;
106 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
107 /* TODO: It is reasonable to complete bio with error here. */
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,
115 iter->bi_sector += bytes >> 9;
117 if (bio_no_advance_iter(bio))
118 iter->bi_size -= bytes;
120 bvec_iter_advance_single(bio->bi_io_vec, iter, bytes);
123 void __bio_advance(struct bio *, unsigned bytes);
126 * bio_advance - increment/complete a bio by some number of bytes
127 * @bio: bio to advance
128 * @nbytes: number of bytes to complete
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.
134 * @bio will then represent the remaining, uncompleted portion of the io.
136 static inline void bio_advance(struct bio *bio, unsigned int nbytes)
138 if (nbytes == bio->bi_iter.bi_size) {
139 bio->bi_iter.bi_size = 0;
142 __bio_advance(bio, nbytes);
145 #define __bio_for_each_segment(bvl, bio, iter, start) \
146 for (iter = (start); \
148 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
149 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
151 #define bio_for_each_segment(bvl, bio, iter) \
152 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
154 #define __bio_for_each_bvec(bvl, bio, iter, start) \
155 for (iter = (start); \
157 ((bvl = mp_bvec_iter_bvec((bio)->bi_io_vec, (iter))), 1); \
158 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
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)
165 * Iterate over all multi-page bvecs. Drivers shouldn't use this version for the
166 * same reasons as bio_for_each_segment_all().
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++)
172 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
174 static inline unsigned bio_segments(struct bio *bio)
178 struct bvec_iter iter;
181 * We special case discard/write same/write zeroes, because they
182 * interpret bi_size differently:
185 switch (bio_op(bio)) {
187 case REQ_OP_SECURE_ERASE:
188 case REQ_OP_WRITE_ZEROES:
194 bio_for_each_segment(bv, bio, iter)
201 * get a reference to a bio, so it won't disappear. the intended use is
205 * submit_bio(rw, bio);
206 * if (bio->bi_flags ...)
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 ...)
214 static inline void bio_get(struct bio *bio)
216 bio->bi_flags |= (1 << BIO_REFFED);
217 smp_mb__before_atomic();
218 atomic_inc(&bio->__bi_cnt);
221 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
224 bio->bi_flags |= (1 << BIO_REFFED);
227 atomic_set(&bio->__bi_cnt, count);
230 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
232 return bio->bi_flags & (1U << bit);
235 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
237 bio->bi_flags |= (1U << bit);
240 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
242 bio->bi_flags &= ~(1U << bit);
245 static inline struct bio_vec *bio_first_bvec_all(struct bio *bio)
247 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
248 return bio->bi_io_vec;
251 static inline struct page *bio_first_page_all(struct bio *bio)
253 return bio_first_bvec_all(bio)->bv_page;
256 static inline struct folio *bio_first_folio_all(struct bio *bio)
258 return page_folio(bio_first_page_all(bio));
261 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
263 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
264 return &bio->bi_io_vec[bio->bi_vcnt - 1];
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
278 /* private: for use by the iterator */
284 static inline void bio_first_folio(struct folio_iter *fi, struct bio *bio,
287 struct bio_vec *bvec = bio_first_bvec_all(bio) + i;
289 if (unlikely(i >= bio->bi_vcnt)) {
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);
303 static inline void bio_next_folio(struct folio_iter *fi, struct bio *bio)
305 fi->_seg_count -= fi->length;
306 if (fi->_seg_count) {
307 fi->folio = fi->_next;
309 fi->length = min(folio_size(fi->folio), fi->_seg_count);
310 fi->_next = folio_next(fi->folio);
312 bio_first_folio(fi, bio, fi->_i + 1);
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.
321 #define bio_for_each_folio_all(fi, bio) \
322 for (bio_first_folio(&fi, bio, 0); fi.folio; bio_next_folio(&fi, bio))
325 BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */
326 BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */
327 BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */
328 BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */
329 BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */
330 BIP_INTEGRITY_USER = 1 << 5, /* Integrity payload is user address */
331 BIP_COPY_USER = 1 << 6, /* Kernel bounce buffer in use */
335 * bio integrity payload
337 struct bio_integrity_payload {
338 struct bio *bip_bio; /* parent bio */
340 struct bvec_iter bip_iter;
342 unsigned short bip_vcnt; /* # of integrity bio_vecs */
343 unsigned short bip_max_vcnt; /* integrity bio_vec slots */
344 unsigned short bip_flags; /* control flags */
346 struct bvec_iter bio_iter; /* for rewinding parent bio */
348 struct work_struct bip_work; /* I/O completion */
350 struct bio_vec *bip_vec;
351 struct bio_vec bip_inline_vecs[];/* embedded bvec array */
354 #if defined(CONFIG_BLK_DEV_INTEGRITY)
356 static inline struct bio_integrity_payload *bio_integrity(struct bio *bio)
358 if (bio->bi_opf & REQ_INTEGRITY)
359 return bio->bi_integrity;
364 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
366 struct bio_integrity_payload *bip = bio_integrity(bio);
369 return bip->bip_flags & flag;
374 static inline sector_t bip_get_seed(struct bio_integrity_payload *bip)
376 return bip->bip_iter.bi_sector;
379 static inline void bip_set_seed(struct bio_integrity_payload *bip,
382 bip->bip_iter.bi_sector = seed;
385 #endif /* CONFIG_BLK_DEV_INTEGRITY */
387 void bio_trim(struct bio *bio, sector_t offset, sector_t size);
388 extern struct bio *bio_split(struct bio *bio, int sectors,
389 gfp_t gfp, struct bio_set *bs);
390 struct bio *bio_split_rw(struct bio *bio, const struct queue_limits *lim,
391 unsigned *segs, struct bio_set *bs, unsigned max_bytes);
394 * bio_next_split - get next @sectors from a bio, splitting if necessary
396 * @sectors: number of sectors to split from the front of @bio
398 * @bs: bio set to allocate from
400 * Return: a bio representing the next @sectors of @bio - if the bio is smaller
401 * than @sectors, returns the original bio unchanged.
403 static inline struct bio *bio_next_split(struct bio *bio, int sectors,
404 gfp_t gfp, struct bio_set *bs)
406 if (sectors >= bio_sectors(bio))
409 return bio_split(bio, sectors, gfp, bs);
413 BIOSET_NEED_BVECS = BIT(0),
414 BIOSET_NEED_RESCUER = BIT(1),
415 BIOSET_PERCPU_CACHE = BIT(2),
417 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
418 extern void bioset_exit(struct bio_set *);
419 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
421 struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs,
422 blk_opf_t opf, gfp_t gfp_mask,
424 struct bio *bio_kmalloc(unsigned short nr_vecs, gfp_t gfp_mask);
425 extern void bio_put(struct bio *);
427 struct bio *bio_alloc_clone(struct block_device *bdev, struct bio *bio_src,
428 gfp_t gfp, struct bio_set *bs);
429 int bio_init_clone(struct block_device *bdev, struct bio *bio,
430 struct bio *bio_src, gfp_t gfp);
432 extern struct bio_set fs_bio_set;
434 static inline struct bio *bio_alloc(struct block_device *bdev,
435 unsigned short nr_vecs, blk_opf_t opf, gfp_t gfp_mask)
437 return bio_alloc_bioset(bdev, nr_vecs, opf, gfp_mask, &fs_bio_set);
440 void submit_bio(struct bio *bio);
442 extern void bio_endio(struct bio *);
444 static inline void bio_io_error(struct bio *bio)
446 bio->bi_status = BLK_STS_IOERR;
450 static inline void bio_wouldblock_error(struct bio *bio)
452 bio_set_flag(bio, BIO_QUIET);
453 bio->bi_status = BLK_STS_AGAIN;
458 * Calculate number of bvec segments that should be allocated to fit data
459 * pointed by @iter. If @iter is backed by bvec it's going to be reused
460 * instead of allocating a new one.
462 static inline int bio_iov_vecs_to_alloc(struct iov_iter *iter, int max_segs)
464 if (iov_iter_is_bvec(iter))
466 return iov_iter_npages(iter, max_segs);
469 struct request_queue;
471 extern int submit_bio_wait(struct bio *bio);
472 void bio_init(struct bio *bio, struct block_device *bdev, struct bio_vec *table,
473 unsigned short max_vecs, blk_opf_t opf);
474 extern void bio_uninit(struct bio *);
475 void bio_reset(struct bio *bio, struct block_device *bdev, blk_opf_t opf);
476 void bio_chain(struct bio *, struct bio *);
478 int __must_check bio_add_page(struct bio *bio, struct page *page, unsigned len,
480 bool __must_check bio_add_folio(struct bio *bio, struct folio *folio,
481 size_t len, size_t off);
482 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
483 unsigned int, unsigned int);
484 int bio_add_zone_append_page(struct bio *bio, struct page *page,
485 unsigned int len, unsigned int offset);
486 void __bio_add_page(struct bio *bio, struct page *page,
487 unsigned int len, unsigned int off);
488 void bio_add_folio_nofail(struct bio *bio, struct folio *folio, size_t len,
490 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
491 void bio_iov_bvec_set(struct bio *bio, struct iov_iter *iter);
492 void __bio_release_pages(struct bio *bio, bool mark_dirty);
493 extern void bio_set_pages_dirty(struct bio *bio);
494 extern void bio_check_pages_dirty(struct bio *bio);
496 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
497 struct bio *src, struct bvec_iter *src_iter);
498 extern void bio_copy_data(struct bio *dst, struct bio *src);
499 extern void bio_free_pages(struct bio *bio);
500 void guard_bio_eod(struct bio *bio);
501 void zero_fill_bio_iter(struct bio *bio, struct bvec_iter iter);
503 static inline void zero_fill_bio(struct bio *bio)
505 zero_fill_bio_iter(bio, bio->bi_iter);
508 static inline void bio_release_pages(struct bio *bio, bool mark_dirty)
510 if (bio_flagged(bio, BIO_PAGE_PINNED))
511 __bio_release_pages(bio, mark_dirty);
514 #define bio_dev(bio) \
515 disk_devt((bio)->bi_bdev->bd_disk)
517 #ifdef CONFIG_BLK_CGROUP
518 void bio_associate_blkg(struct bio *bio);
519 void bio_associate_blkg_from_css(struct bio *bio,
520 struct cgroup_subsys_state *css);
521 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
522 void blkcg_punt_bio_submit(struct bio *bio);
523 #else /* CONFIG_BLK_CGROUP */
524 static inline void bio_associate_blkg(struct bio *bio) { }
525 static inline void bio_associate_blkg_from_css(struct bio *bio,
526 struct cgroup_subsys_state *css)
528 static inline void bio_clone_blkg_association(struct bio *dst,
530 static inline void blkcg_punt_bio_submit(struct bio *bio)
534 #endif /* CONFIG_BLK_CGROUP */
536 static inline void bio_set_dev(struct bio *bio, struct block_device *bdev)
538 bio_clear_flag(bio, BIO_REMAPPED);
539 if (bio->bi_bdev != bdev)
540 bio_clear_flag(bio, BIO_BPS_THROTTLED);
542 bio_associate_blkg(bio);
546 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
548 * A bio_list anchors a singly-linked list of bios chained through the bi_next
549 * member of the bio. The bio_list also caches the last list member to allow
550 * fast access to the tail.
557 static inline int bio_list_empty(const struct bio_list *bl)
559 return bl->head == NULL;
562 static inline void bio_list_init(struct bio_list *bl)
564 bl->head = bl->tail = NULL;
567 #define BIO_EMPTY_LIST { NULL, NULL }
569 #define bio_list_for_each(bio, bl) \
570 for (bio = (bl)->head; bio; bio = bio->bi_next)
572 static inline unsigned bio_list_size(const struct bio_list *bl)
577 bio_list_for_each(bio, bl)
583 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
588 bl->tail->bi_next = bio;
595 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
597 bio->bi_next = bl->head;
605 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
611 bl->tail->bi_next = bl2->head;
613 bl->head = bl2->head;
615 bl->tail = bl2->tail;
618 static inline void bio_list_merge_head(struct bio_list *bl,
619 struct bio_list *bl2)
625 bl2->tail->bi_next = bl->head;
627 bl->tail = bl2->tail;
629 bl->head = bl2->head;
632 static inline struct bio *bio_list_peek(struct bio_list *bl)
637 static inline struct bio *bio_list_pop(struct bio_list *bl)
639 struct bio *bio = bl->head;
642 bl->head = bl->head->bi_next;
652 static inline struct bio *bio_list_get(struct bio_list *bl)
654 struct bio *bio = bl->head;
656 bl->head = bl->tail = NULL;
662 * Increment chain count for the bio. Make sure the CHAIN flag update
663 * is visible before the raised count.
665 static inline void bio_inc_remaining(struct bio *bio)
667 bio_set_flag(bio, BIO_CHAIN);
668 smp_mb__before_atomic();
669 atomic_inc(&bio->__bi_remaining);
673 * bio_set is used to allow other portions of the IO system to
674 * allocate their own private memory pools for bio and iovec structures.
675 * These memory pools in turn all allocate from the bio_slab
676 * and the bvec_slabs[].
678 #define BIO_POOL_SIZE 2
681 struct kmem_cache *bio_slab;
682 unsigned int front_pad;
685 * per-cpu bio alloc cache
687 struct bio_alloc_cache __percpu *cache;
691 #if defined(CONFIG_BLK_DEV_INTEGRITY)
692 mempool_t bio_integrity_pool;
693 mempool_t bvec_integrity_pool;
696 unsigned int back_pad;
698 * Deadlock avoidance for stacking block drivers: see comments in
699 * bio_alloc_bioset() for details
701 spinlock_t rescue_lock;
702 struct bio_list rescue_list;
703 struct work_struct rescue_work;
704 struct workqueue_struct *rescue_workqueue;
707 * Hot un-plug notifier for the per-cpu cache, if used
709 struct hlist_node cpuhp_dead;
712 static inline bool bioset_initialized(struct bio_set *bs)
714 return bs->bio_slab != NULL;
717 #if defined(CONFIG_BLK_DEV_INTEGRITY)
719 #define bip_for_each_vec(bvl, bip, iter) \
720 for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter)
722 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \
724 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter)
726 int bio_integrity_map_user(struct bio *bio, void __user *ubuf, ssize_t len, u32 seed);
727 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
728 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
729 extern bool bio_integrity_prep(struct bio *);
730 extern void bio_integrity_advance(struct bio *, unsigned int);
731 extern void bio_integrity_trim(struct bio *);
732 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
733 extern int bioset_integrity_create(struct bio_set *, int);
734 extern void bioset_integrity_free(struct bio_set *);
735 extern void bio_integrity_init(void);
737 #else /* CONFIG_BLK_DEV_INTEGRITY */
739 static inline void *bio_integrity(struct bio *bio)
744 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size)
749 static inline void bioset_integrity_free (struct bio_set *bs)
754 static inline bool bio_integrity_prep(struct bio *bio)
759 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
765 static inline void bio_integrity_advance(struct bio *bio,
766 unsigned int bytes_done)
771 static inline void bio_integrity_trim(struct bio *bio)
776 static inline void bio_integrity_init(void)
781 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
786 static inline void *bio_integrity_alloc(struct bio * bio, gfp_t gfp,
789 return ERR_PTR(-EINVAL);
792 static inline int bio_integrity_add_page(struct bio *bio, struct page *page,
793 unsigned int len, unsigned int offset)
798 static inline int bio_integrity_map_user(struct bio *bio, void __user *ubuf,
799 ssize_t len, u32 seed)
804 #endif /* CONFIG_BLK_DEV_INTEGRITY */
807 * Mark a bio as polled. Note that for async polled IO, the caller must
808 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
809 * We cannot block waiting for requests on polled IO, as those completions
810 * must be found by the caller. This is different than IRQ driven IO, where
811 * it's safe to wait for IO to complete.
813 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
815 bio->bi_opf |= REQ_POLLED;
816 if (kiocb->ki_flags & IOCB_NOWAIT)
817 bio->bi_opf |= REQ_NOWAIT;
820 static inline void bio_clear_polled(struct bio *bio)
822 bio->bi_opf &= ~REQ_POLLED;
825 struct bio *blk_next_bio(struct bio *bio, struct block_device *bdev,
826 unsigned int nr_pages, blk_opf_t opf, gfp_t gfp);
828 #endif /* __LINUX_BIO_H */