1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2001 Jens Axboe <axboe@suse.de>
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
14 #define BIO_MAX_INLINE_VECS UIO_MAXIOV
18 static inline unsigned int bio_max_segs(unsigned int nr_segs)
20 return min(nr_segs, BIO_MAX_VECS);
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);
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 folio *bio_first_folio_all(struct bio *bio)
256 return page_folio(bio_first_page_all(bio));
259 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
261 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
262 return &bio->bi_io_vec[bio->bi_vcnt - 1];
266 * struct folio_iter - State for iterating all folios in a bio.
267 * @folio: The current folio we're iterating. NULL after the last folio.
268 * @offset: The byte offset within the current folio.
269 * @length: The number of bytes in this iteration (will not cross folio
276 /* private: for use by the iterator */
282 static inline void bio_first_folio(struct folio_iter *fi, struct bio *bio,
285 struct bio_vec *bvec = bio_first_bvec_all(bio) + i;
287 if (unlikely(i >= bio->bi_vcnt)) {
292 fi->folio = page_folio(bvec->bv_page);
293 fi->offset = bvec->bv_offset +
294 PAGE_SIZE * folio_page_idx(fi->folio, bvec->bv_page);
295 fi->_seg_count = bvec->bv_len;
296 fi->length = min(folio_size(fi->folio) - fi->offset, fi->_seg_count);
297 fi->_next = folio_next(fi->folio);
301 static inline void bio_next_folio(struct folio_iter *fi, struct bio *bio)
303 fi->_seg_count -= fi->length;
304 if (fi->_seg_count) {
305 fi->folio = fi->_next;
307 fi->length = min(folio_size(fi->folio), fi->_seg_count);
308 fi->_next = folio_next(fi->folio);
310 bio_first_folio(fi, bio, fi->_i + 1);
315 * bio_for_each_folio_all - Iterate over each folio in a bio.
316 * @fi: struct folio_iter which is updated for each folio.
317 * @bio: struct bio to iterate over.
319 #define bio_for_each_folio_all(fi, bio) \
320 for (bio_first_folio(&fi, bio, 0); fi.folio; bio_next_folio(&fi, bio))
322 void bio_trim(struct bio *bio, sector_t offset, sector_t size);
323 extern struct bio *bio_split(struct bio *bio, int sectors,
324 gfp_t gfp, struct bio_set *bs);
325 int bio_split_rw_at(struct bio *bio, const struct queue_limits *lim,
326 unsigned *segs, unsigned max_bytes);
329 * bio_next_split - get next @sectors from a bio, splitting if necessary
331 * @sectors: number of sectors to split from the front of @bio
333 * @bs: bio set to allocate from
335 * Return: a bio representing the next @sectors of @bio - if the bio is smaller
336 * than @sectors, returns the original bio unchanged.
338 static inline struct bio *bio_next_split(struct bio *bio, int sectors,
339 gfp_t gfp, struct bio_set *bs)
341 if (sectors >= bio_sectors(bio))
344 return bio_split(bio, sectors, gfp, bs);
348 BIOSET_NEED_BVECS = BIT(0),
349 BIOSET_NEED_RESCUER = BIT(1),
350 BIOSET_PERCPU_CACHE = BIT(2),
352 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
353 extern void bioset_exit(struct bio_set *);
354 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
356 struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs,
357 blk_opf_t opf, gfp_t gfp_mask,
359 struct bio *bio_kmalloc(unsigned short nr_vecs, gfp_t gfp_mask);
360 extern void bio_put(struct bio *);
362 struct bio *bio_alloc_clone(struct block_device *bdev, struct bio *bio_src,
363 gfp_t gfp, struct bio_set *bs);
364 int bio_init_clone(struct block_device *bdev, struct bio *bio,
365 struct bio *bio_src, gfp_t gfp);
367 extern struct bio_set fs_bio_set;
369 static inline struct bio *bio_alloc(struct block_device *bdev,
370 unsigned short nr_vecs, blk_opf_t opf, gfp_t gfp_mask)
372 return bio_alloc_bioset(bdev, nr_vecs, opf, gfp_mask, &fs_bio_set);
375 void submit_bio(struct bio *bio);
377 extern void bio_endio(struct bio *);
379 static inline void bio_io_error(struct bio *bio)
381 bio->bi_status = BLK_STS_IOERR;
385 static inline void bio_wouldblock_error(struct bio *bio)
387 bio_set_flag(bio, BIO_QUIET);
388 bio->bi_status = BLK_STS_AGAIN;
393 * Calculate number of bvec segments that should be allocated to fit data
394 * pointed by @iter. If @iter is backed by bvec it's going to be reused
395 * instead of allocating a new one.
397 static inline int bio_iov_vecs_to_alloc(struct iov_iter *iter, int max_segs)
399 if (iov_iter_is_bvec(iter))
401 return iov_iter_npages(iter, max_segs);
404 struct request_queue;
406 void bio_init(struct bio *bio, struct block_device *bdev, struct bio_vec *table,
407 unsigned short max_vecs, blk_opf_t opf);
408 extern void bio_uninit(struct bio *);
409 void bio_reset(struct bio *bio, struct block_device *bdev, blk_opf_t opf);
410 void bio_chain(struct bio *, struct bio *);
412 int __must_check bio_add_page(struct bio *bio, struct page *page, unsigned len,
414 bool __must_check bio_add_folio(struct bio *bio, struct folio *folio,
415 size_t len, size_t off);
416 void __bio_add_page(struct bio *bio, struct page *page,
417 unsigned int len, unsigned int off);
418 void bio_add_folio_nofail(struct bio *bio, struct folio *folio, size_t len,
420 void bio_add_virt_nofail(struct bio *bio, void *vaddr, unsigned len);
423 * bio_add_max_vecs - number of bio_vecs needed to add data to a bio
424 * @kaddr: kernel virtual address to add
425 * @len: length in bytes to add
427 * Calculate how many bio_vecs need to be allocated to add the kernel virtual
428 * address range in [@kaddr:@len] in the worse case.
430 static inline unsigned int bio_add_max_vecs(void *kaddr, unsigned int len)
432 if (is_vmalloc_addr(kaddr))
433 return DIV_ROUND_UP(offset_in_page(kaddr) + len, PAGE_SIZE);
437 unsigned int bio_add_vmalloc_chunk(struct bio *bio, void *vaddr, unsigned len);
438 bool bio_add_vmalloc(struct bio *bio, void *vaddr, unsigned int len);
440 int submit_bio_wait(struct bio *bio);
441 int bdev_rw_virt(struct block_device *bdev, sector_t sector, void *data,
442 size_t len, enum req_op op);
444 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
445 void bio_iov_bvec_set(struct bio *bio, const struct iov_iter *iter);
446 void __bio_release_pages(struct bio *bio, bool mark_dirty);
447 extern void bio_set_pages_dirty(struct bio *bio);
448 extern void bio_check_pages_dirty(struct bio *bio);
450 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
451 struct bio *src, struct bvec_iter *src_iter);
452 extern void bio_copy_data(struct bio *dst, struct bio *src);
453 extern void bio_free_pages(struct bio *bio);
454 void guard_bio_eod(struct bio *bio);
455 void zero_fill_bio_iter(struct bio *bio, struct bvec_iter iter);
457 static inline void zero_fill_bio(struct bio *bio)
459 zero_fill_bio_iter(bio, bio->bi_iter);
462 static inline void bio_release_pages(struct bio *bio, bool mark_dirty)
464 if (bio_flagged(bio, BIO_PAGE_PINNED))
465 __bio_release_pages(bio, mark_dirty);
468 #define bio_dev(bio) \
469 disk_devt((bio)->bi_bdev->bd_disk)
471 #ifdef CONFIG_BLK_CGROUP
472 void bio_associate_blkg(struct bio *bio);
473 void bio_associate_blkg_from_css(struct bio *bio,
474 struct cgroup_subsys_state *css);
475 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
476 void blkcg_punt_bio_submit(struct bio *bio);
477 #else /* CONFIG_BLK_CGROUP */
478 static inline void bio_associate_blkg(struct bio *bio) { }
479 static inline void bio_associate_blkg_from_css(struct bio *bio,
480 struct cgroup_subsys_state *css)
482 static inline void bio_clone_blkg_association(struct bio *dst,
484 static inline void blkcg_punt_bio_submit(struct bio *bio)
488 #endif /* CONFIG_BLK_CGROUP */
490 static inline void bio_set_dev(struct bio *bio, struct block_device *bdev)
492 bio_clear_flag(bio, BIO_REMAPPED);
493 if (bio->bi_bdev != bdev)
494 bio_clear_flag(bio, BIO_BPS_THROTTLED);
496 bio_associate_blkg(bio);
500 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
502 * A bio_list anchors a singly-linked list of bios chained through the bi_next
503 * member of the bio. The bio_list also caches the last list member to allow
504 * fast access to the tail.
511 static inline int bio_list_empty(const struct bio_list *bl)
513 return bl->head == NULL;
516 static inline void bio_list_init(struct bio_list *bl)
518 bl->head = bl->tail = NULL;
521 #define BIO_EMPTY_LIST { NULL, NULL }
523 #define bio_list_for_each(bio, bl) \
524 for (bio = (bl)->head; bio; bio = bio->bi_next)
526 static inline unsigned bio_list_size(const struct bio_list *bl)
531 bio_list_for_each(bio, bl)
537 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
542 bl->tail->bi_next = bio;
549 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
551 bio->bi_next = bl->head;
559 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
565 bl->tail->bi_next = bl2->head;
567 bl->head = bl2->head;
569 bl->tail = bl2->tail;
572 static inline void bio_list_merge_init(struct bio_list *bl,
573 struct bio_list *bl2)
575 bio_list_merge(bl, bl2);
579 static inline void bio_list_merge_head(struct bio_list *bl,
580 struct bio_list *bl2)
586 bl2->tail->bi_next = bl->head;
588 bl->tail = bl2->tail;
590 bl->head = bl2->head;
593 static inline struct bio *bio_list_peek(struct bio_list *bl)
598 static inline struct bio *bio_list_pop(struct bio_list *bl)
600 struct bio *bio = bl->head;
603 bl->head = bl->head->bi_next;
613 static inline struct bio *bio_list_get(struct bio_list *bl)
615 struct bio *bio = bl->head;
617 bl->head = bl->tail = NULL;
623 * Increment chain count for the bio. Make sure the CHAIN flag update
624 * is visible before the raised count.
626 static inline void bio_inc_remaining(struct bio *bio)
628 bio_set_flag(bio, BIO_CHAIN);
629 smp_mb__before_atomic();
630 atomic_inc(&bio->__bi_remaining);
634 * bio_set is used to allow other portions of the IO system to
635 * allocate their own private memory pools for bio and iovec structures.
636 * These memory pools in turn all allocate from the bio_slab
637 * and the bvec_slabs[].
639 #define BIO_POOL_SIZE 2
642 struct kmem_cache *bio_slab;
643 unsigned int front_pad;
646 * per-cpu bio alloc cache
648 struct bio_alloc_cache __percpu *cache;
653 unsigned int back_pad;
655 * Deadlock avoidance for stacking block drivers: see comments in
656 * bio_alloc_bioset() for details
658 spinlock_t rescue_lock;
659 struct bio_list rescue_list;
660 struct work_struct rescue_work;
661 struct workqueue_struct *rescue_workqueue;
664 * Hot un-plug notifier for the per-cpu cache, if used
666 struct hlist_node cpuhp_dead;
669 static inline bool bioset_initialized(struct bio_set *bs)
671 return bs->bio_slab != NULL;
675 * Mark a bio as polled. Note that for async polled IO, the caller must
676 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
677 * We cannot block waiting for requests on polled IO, as those completions
678 * must be found by the caller. This is different than IRQ driven IO, where
679 * it's safe to wait for IO to complete.
681 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
683 bio->bi_opf |= REQ_POLLED;
684 if (kiocb->ki_flags & IOCB_NOWAIT)
685 bio->bi_opf |= REQ_NOWAIT;
688 static inline void bio_clear_polled(struct bio *bio)
690 bio->bi_opf &= ~REQ_POLLED;
694 * bio_is_zone_append - is this a zone append bio?
697 * Check if @bio is a zone append operation. Core block layer code and end_io
698 * handlers must use this instead of an open coded REQ_OP_ZONE_APPEND check
699 * because the block layer can rewrite REQ_OP_ZONE_APPEND to REQ_OP_WRITE if
700 * it is not natively supported.
702 static inline bool bio_is_zone_append(struct bio *bio)
704 if (!IS_ENABLED(CONFIG_BLK_DEV_ZONED))
706 return bio_op(bio) == REQ_OP_ZONE_APPEND ||
707 bio_flagged(bio, BIO_EMULATES_ZONE_APPEND);
710 struct bio *blk_next_bio(struct bio *bio, struct block_device *bdev,
711 unsigned int nr_pages, blk_opf_t opf, gfp_t gfp);
712 struct bio *bio_chain_and_submit(struct bio *prev, struct bio *new);
714 struct bio *blk_alloc_discard_bio(struct block_device *bdev,
715 sector_t *sector, sector_t *nr_sects, gfp_t gfp_mask);
717 #endif /* __LINUX_BIO_H */