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 fi->folio = page_folio(bvec->bv_page);
290 fi->offset = bvec->bv_offset +
291 PAGE_SIZE * (bvec->bv_page - &fi->folio->page);
292 fi->_seg_count = bvec->bv_len;
293 fi->length = min(folio_size(fi->folio) - fi->offset, fi->_seg_count);
294 fi->_next = folio_next(fi->folio);
298 static inline void bio_next_folio(struct folio_iter *fi, struct bio *bio)
300 fi->_seg_count -= fi->length;
301 if (fi->_seg_count) {
302 fi->folio = fi->_next;
304 fi->length = min(folio_size(fi->folio), fi->_seg_count);
305 fi->_next = folio_next(fi->folio);
306 } else if (fi->_i + 1 < bio->bi_vcnt) {
307 bio_first_folio(fi, bio, fi->_i + 1);
314 * bio_for_each_folio_all - Iterate over each folio in a bio.
315 * @fi: struct folio_iter which is updated for each folio.
316 * @bio: struct bio to iterate over.
318 #define bio_for_each_folio_all(fi, bio) \
319 for (bio_first_folio(&fi, bio, 0); fi.folio; bio_next_folio(&fi, bio))
322 BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */
323 BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */
324 BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */
325 BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */
326 BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */
330 * bio integrity payload
332 struct bio_integrity_payload {
333 struct bio *bip_bio; /* parent bio */
335 struct bvec_iter bip_iter;
337 unsigned short bip_vcnt; /* # of integrity bio_vecs */
338 unsigned short bip_max_vcnt; /* integrity bio_vec slots */
339 unsigned short bip_flags; /* control flags */
341 struct bvec_iter bio_iter; /* for rewinding parent bio */
343 struct work_struct bip_work; /* I/O completion */
345 struct bio_vec *bip_vec;
346 struct bio_vec bip_inline_vecs[];/* embedded bvec array */
349 #if defined(CONFIG_BLK_DEV_INTEGRITY)
351 static inline struct bio_integrity_payload *bio_integrity(struct bio *bio)
353 if (bio->bi_opf & REQ_INTEGRITY)
354 return bio->bi_integrity;
359 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
361 struct bio_integrity_payload *bip = bio_integrity(bio);
364 return bip->bip_flags & flag;
369 static inline sector_t bip_get_seed(struct bio_integrity_payload *bip)
371 return bip->bip_iter.bi_sector;
374 static inline void bip_set_seed(struct bio_integrity_payload *bip,
377 bip->bip_iter.bi_sector = seed;
380 #endif /* CONFIG_BLK_DEV_INTEGRITY */
382 void bio_trim(struct bio *bio, sector_t offset, sector_t size);
383 extern struct bio *bio_split(struct bio *bio, int sectors,
384 gfp_t gfp, struct bio_set *bs);
385 struct bio *bio_split_rw(struct bio *bio, const struct queue_limits *lim,
386 unsigned *segs, struct bio_set *bs, unsigned max_bytes);
389 * bio_next_split - get next @sectors from a bio, splitting if necessary
391 * @sectors: number of sectors to split from the front of @bio
393 * @bs: bio set to allocate from
395 * Return: a bio representing the next @sectors of @bio - if the bio is smaller
396 * than @sectors, returns the original bio unchanged.
398 static inline struct bio *bio_next_split(struct bio *bio, int sectors,
399 gfp_t gfp, struct bio_set *bs)
401 if (sectors >= bio_sectors(bio))
404 return bio_split(bio, sectors, gfp, bs);
408 BIOSET_NEED_BVECS = BIT(0),
409 BIOSET_NEED_RESCUER = BIT(1),
410 BIOSET_PERCPU_CACHE = BIT(2),
412 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
413 extern void bioset_exit(struct bio_set *);
414 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
416 struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs,
417 blk_opf_t opf, gfp_t gfp_mask,
419 struct bio *bio_kmalloc(unsigned short nr_vecs, gfp_t gfp_mask);
420 extern void bio_put(struct bio *);
422 struct bio *bio_alloc_clone(struct block_device *bdev, struct bio *bio_src,
423 gfp_t gfp, struct bio_set *bs);
424 int bio_init_clone(struct block_device *bdev, struct bio *bio,
425 struct bio *bio_src, gfp_t gfp);
427 extern struct bio_set fs_bio_set;
429 static inline struct bio *bio_alloc(struct block_device *bdev,
430 unsigned short nr_vecs, blk_opf_t opf, gfp_t gfp_mask)
432 return bio_alloc_bioset(bdev, nr_vecs, opf, gfp_mask, &fs_bio_set);
435 void submit_bio(struct bio *bio);
437 extern void bio_endio(struct bio *);
439 static inline void bio_io_error(struct bio *bio)
441 bio->bi_status = BLK_STS_IOERR;
445 static inline void bio_wouldblock_error(struct bio *bio)
447 bio_set_flag(bio, BIO_QUIET);
448 bio->bi_status = BLK_STS_AGAIN;
453 * Calculate number of bvec segments that should be allocated to fit data
454 * pointed by @iter. If @iter is backed by bvec it's going to be reused
455 * instead of allocating a new one.
457 static inline int bio_iov_vecs_to_alloc(struct iov_iter *iter, int max_segs)
459 if (iov_iter_is_bvec(iter))
461 return iov_iter_npages(iter, max_segs);
464 struct request_queue;
466 extern int submit_bio_wait(struct bio *bio);
467 void bio_init(struct bio *bio, struct block_device *bdev, struct bio_vec *table,
468 unsigned short max_vecs, blk_opf_t opf);
469 extern void bio_uninit(struct bio *);
470 void bio_reset(struct bio *bio, struct block_device *bdev, blk_opf_t opf);
471 void bio_chain(struct bio *, struct bio *);
473 int __must_check bio_add_page(struct bio *bio, struct page *page, unsigned len,
475 bool __must_check bio_add_folio(struct bio *bio, struct folio *folio,
476 size_t len, size_t off);
477 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
478 unsigned int, unsigned int);
479 int bio_add_zone_append_page(struct bio *bio, struct page *page,
480 unsigned int len, unsigned int offset);
481 void __bio_add_page(struct bio *bio, struct page *page,
482 unsigned int len, unsigned int off);
483 void bio_add_folio_nofail(struct bio *bio, struct folio *folio, size_t len,
485 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
486 void bio_iov_bvec_set(struct bio *bio, struct iov_iter *iter);
487 void __bio_release_pages(struct bio *bio, bool mark_dirty);
488 extern void bio_set_pages_dirty(struct bio *bio);
489 extern void bio_check_pages_dirty(struct bio *bio);
491 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
492 struct bio *src, struct bvec_iter *src_iter);
493 extern void bio_copy_data(struct bio *dst, struct bio *src);
494 extern void bio_free_pages(struct bio *bio);
495 void guard_bio_eod(struct bio *bio);
496 void zero_fill_bio_iter(struct bio *bio, struct bvec_iter iter);
498 static inline void zero_fill_bio(struct bio *bio)
500 zero_fill_bio_iter(bio, bio->bi_iter);
503 static inline void bio_release_pages(struct bio *bio, bool mark_dirty)
505 if (bio_flagged(bio, BIO_PAGE_PINNED))
506 __bio_release_pages(bio, mark_dirty);
509 #define bio_dev(bio) \
510 disk_devt((bio)->bi_bdev->bd_disk)
512 #ifdef CONFIG_BLK_CGROUP
513 void bio_associate_blkg(struct bio *bio);
514 void bio_associate_blkg_from_css(struct bio *bio,
515 struct cgroup_subsys_state *css);
516 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
517 void blkcg_punt_bio_submit(struct bio *bio);
518 #else /* CONFIG_BLK_CGROUP */
519 static inline void bio_associate_blkg(struct bio *bio) { }
520 static inline void bio_associate_blkg_from_css(struct bio *bio,
521 struct cgroup_subsys_state *css)
523 static inline void bio_clone_blkg_association(struct bio *dst,
525 static inline void blkcg_punt_bio_submit(struct bio *bio)
529 #endif /* CONFIG_BLK_CGROUP */
531 static inline void bio_set_dev(struct bio *bio, struct block_device *bdev)
533 bio_clear_flag(bio, BIO_REMAPPED);
534 if (bio->bi_bdev != bdev)
535 bio_clear_flag(bio, BIO_BPS_THROTTLED);
537 bio_associate_blkg(bio);
541 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
543 * A bio_list anchors a singly-linked list of bios chained through the bi_next
544 * member of the bio. The bio_list also caches the last list member to allow
545 * fast access to the tail.
552 static inline int bio_list_empty(const struct bio_list *bl)
554 return bl->head == NULL;
557 static inline void bio_list_init(struct bio_list *bl)
559 bl->head = bl->tail = NULL;
562 #define BIO_EMPTY_LIST { NULL, NULL }
564 #define bio_list_for_each(bio, bl) \
565 for (bio = (bl)->head; bio; bio = bio->bi_next)
567 static inline unsigned bio_list_size(const struct bio_list *bl)
572 bio_list_for_each(bio, bl)
578 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
583 bl->tail->bi_next = bio;
590 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
592 bio->bi_next = bl->head;
600 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
606 bl->tail->bi_next = bl2->head;
608 bl->head = bl2->head;
610 bl->tail = bl2->tail;
613 static inline void bio_list_merge_head(struct bio_list *bl,
614 struct bio_list *bl2)
620 bl2->tail->bi_next = bl->head;
622 bl->tail = bl2->tail;
624 bl->head = bl2->head;
627 static inline struct bio *bio_list_peek(struct bio_list *bl)
632 static inline struct bio *bio_list_pop(struct bio_list *bl)
634 struct bio *bio = bl->head;
637 bl->head = bl->head->bi_next;
647 static inline struct bio *bio_list_get(struct bio_list *bl)
649 struct bio *bio = bl->head;
651 bl->head = bl->tail = NULL;
657 * Increment chain count for the bio. Make sure the CHAIN flag update
658 * is visible before the raised count.
660 static inline void bio_inc_remaining(struct bio *bio)
662 bio_set_flag(bio, BIO_CHAIN);
663 smp_mb__before_atomic();
664 atomic_inc(&bio->__bi_remaining);
668 * bio_set is used to allow other portions of the IO system to
669 * allocate their own private memory pools for bio and iovec structures.
670 * These memory pools in turn all allocate from the bio_slab
671 * and the bvec_slabs[].
673 #define BIO_POOL_SIZE 2
676 struct kmem_cache *bio_slab;
677 unsigned int front_pad;
680 * per-cpu bio alloc cache
682 struct bio_alloc_cache __percpu *cache;
686 #if defined(CONFIG_BLK_DEV_INTEGRITY)
687 mempool_t bio_integrity_pool;
688 mempool_t bvec_integrity_pool;
691 unsigned int back_pad;
693 * Deadlock avoidance for stacking block drivers: see comments in
694 * bio_alloc_bioset() for details
696 spinlock_t rescue_lock;
697 struct bio_list rescue_list;
698 struct work_struct rescue_work;
699 struct workqueue_struct *rescue_workqueue;
702 * Hot un-plug notifier for the per-cpu cache, if used
704 struct hlist_node cpuhp_dead;
707 static inline bool bioset_initialized(struct bio_set *bs)
709 return bs->bio_slab != NULL;
712 #if defined(CONFIG_BLK_DEV_INTEGRITY)
714 #define bip_for_each_vec(bvl, bip, iter) \
715 for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter)
717 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \
719 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter)
721 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
722 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
723 extern bool bio_integrity_prep(struct bio *);
724 extern void bio_integrity_advance(struct bio *, unsigned int);
725 extern void bio_integrity_trim(struct bio *);
726 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
727 extern int bioset_integrity_create(struct bio_set *, int);
728 extern void bioset_integrity_free(struct bio_set *);
729 extern void bio_integrity_init(void);
731 #else /* CONFIG_BLK_DEV_INTEGRITY */
733 static inline void *bio_integrity(struct bio *bio)
738 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size)
743 static inline void bioset_integrity_free (struct bio_set *bs)
748 static inline bool bio_integrity_prep(struct bio *bio)
753 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
759 static inline void bio_integrity_advance(struct bio *bio,
760 unsigned int bytes_done)
765 static inline void bio_integrity_trim(struct bio *bio)
770 static inline void bio_integrity_init(void)
775 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
780 static inline void *bio_integrity_alloc(struct bio * bio, gfp_t gfp,
783 return ERR_PTR(-EINVAL);
786 static inline int bio_integrity_add_page(struct bio *bio, struct page *page,
787 unsigned int len, unsigned int offset)
792 #endif /* CONFIG_BLK_DEV_INTEGRITY */
795 * Mark a bio as polled. Note that for async polled IO, the caller must
796 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
797 * We cannot block waiting for requests on polled IO, as those completions
798 * must be found by the caller. This is different than IRQ driven IO, where
799 * it's safe to wait for IO to complete.
801 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
803 bio->bi_opf |= REQ_POLLED;
804 if (kiocb->ki_flags & IOCB_NOWAIT)
805 bio->bi_opf |= REQ_NOWAIT;
808 static inline void bio_clear_polled(struct bio *bio)
810 bio->bi_opf &= ~REQ_POLLED;
813 struct bio *blk_next_bio(struct bio *bio, struct block_device *bdev,
814 unsigned int nr_pages, blk_opf_t opf, gfp_t gfp);
816 #endif /* __LINUX_BIO_H */