1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Copyright (C) 2001 Jens Axboe <axboe@suse.de>
8 #include <linux/highmem.h>
9 #include <linux/mempool.h>
10 #include <linux/ioprio.h>
11 /* struct bio, bio_vec and BIO_* flags are defined in blk_types.h */
12 #include <linux/blk_types.h>
17 #define BIO_BUG_ON BUG_ON
22 #define BIO_MAX_PAGES 256
24 #define bio_prio(bio) (bio)->bi_ioprio
25 #define bio_set_prio(bio, prio) ((bio)->bi_ioprio = prio)
27 #define bio_iter_iovec(bio, iter) \
28 bvec_iter_bvec((bio)->bi_io_vec, (iter))
30 #define bio_iter_page(bio, iter) \
31 bvec_iter_page((bio)->bi_io_vec, (iter))
32 #define bio_iter_len(bio, iter) \
33 bvec_iter_len((bio)->bi_io_vec, (iter))
34 #define bio_iter_offset(bio, iter) \
35 bvec_iter_offset((bio)->bi_io_vec, (iter))
37 #define bio_page(bio) bio_iter_page((bio), (bio)->bi_iter)
38 #define bio_offset(bio) bio_iter_offset((bio), (bio)->bi_iter)
39 #define bio_iovec(bio) bio_iter_iovec((bio), (bio)->bi_iter)
41 #define bvec_iter_sectors(iter) ((iter).bi_size >> 9)
42 #define bvec_iter_end_sector(iter) ((iter).bi_sector + bvec_iter_sectors((iter)))
44 #define bio_sectors(bio) bvec_iter_sectors((bio)->bi_iter)
45 #define bio_end_sector(bio) bvec_iter_end_sector((bio)->bi_iter)
48 * Return the data direction, READ or WRITE.
50 #define bio_data_dir(bio) \
51 (op_is_write(bio_op(bio)) ? WRITE : READ)
54 * Check whether this bio carries any data or not. A NULL bio is allowed.
56 static inline bool bio_has_data(struct bio *bio)
59 bio->bi_iter.bi_size &&
60 bio_op(bio) != REQ_OP_DISCARD &&
61 bio_op(bio) != REQ_OP_SECURE_ERASE &&
62 bio_op(bio) != REQ_OP_WRITE_ZEROES)
68 static inline bool bio_no_advance_iter(const struct bio *bio)
70 return bio_op(bio) == REQ_OP_DISCARD ||
71 bio_op(bio) == REQ_OP_SECURE_ERASE ||
72 bio_op(bio) == REQ_OP_WRITE_SAME ||
73 bio_op(bio) == REQ_OP_WRITE_ZEROES;
76 static inline bool bio_mergeable(struct bio *bio)
78 if (bio->bi_opf & REQ_NOMERGE_FLAGS)
84 static inline unsigned int bio_cur_bytes(struct bio *bio)
86 if (bio_has_data(bio))
87 return bio_iovec(bio).bv_len;
88 else /* dataless requests such as discard */
89 return bio->bi_iter.bi_size;
92 static inline void *bio_data(struct bio *bio)
94 if (bio_has_data(bio))
95 return page_address(bio_page(bio)) + bio_offset(bio);
101 * bio_full - check if the bio is full
103 * @len: length of one segment to be added
105 * Return true if @bio is full and one segment with @len bytes can't be
106 * added to the bio, otherwise return false
108 static inline bool bio_full(struct bio *bio, unsigned len)
110 if (bio->bi_vcnt >= bio->bi_max_vecs)
113 if (bio->bi_iter.bi_size > UINT_MAX - len)
119 static inline bool bio_next_segment(const struct bio *bio,
120 struct bvec_iter_all *iter)
122 if (iter->idx >= bio->bi_vcnt)
125 bvec_advance(&bio->bi_io_vec[iter->idx], iter);
130 * drivers should _never_ use the all version - the bio may have been split
131 * before it got to the driver and the driver won't own all of it
133 #define bio_for_each_segment_all(bvl, bio, iter) \
134 for (bvl = bvec_init_iter_all(&iter); bio_next_segment((bio), &iter); )
136 static inline void bio_advance_iter(const struct bio *bio,
137 struct bvec_iter *iter, unsigned int bytes)
139 iter->bi_sector += bytes >> 9;
141 if (bio_no_advance_iter(bio))
142 iter->bi_size -= bytes;
144 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
145 /* TODO: It is reasonable to complete bio with error here. */
148 #define __bio_for_each_segment(bvl, bio, iter, start) \
149 for (iter = (start); \
151 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
152 bio_advance_iter((bio), &(iter), (bvl).bv_len))
154 #define bio_for_each_segment(bvl, bio, iter) \
155 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
157 #define __bio_for_each_bvec(bvl, bio, iter, start) \
158 for (iter = (start); \
160 ((bvl = mp_bvec_iter_bvec((bio)->bi_io_vec, (iter))), 1); \
161 bio_advance_iter((bio), &(iter), (bvl).bv_len))
163 /* iterate over multi-page bvec */
164 #define bio_for_each_bvec(bvl, bio, iter) \
165 __bio_for_each_bvec(bvl, bio, iter, (bio)->bi_iter)
168 * Iterate over all multi-page bvecs. Drivers shouldn't use this version for the
169 * same reasons as bio_for_each_segment_all().
171 #define bio_for_each_bvec_all(bvl, bio, i) \
172 for (i = 0, bvl = bio_first_bvec_all(bio); \
173 i < (bio)->bi_vcnt; i++, bvl++) \
175 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
177 static inline unsigned bio_segments(struct bio *bio)
181 struct bvec_iter iter;
184 * We special case discard/write same/write zeroes, because they
185 * interpret bi_size differently:
188 switch (bio_op(bio)) {
190 case REQ_OP_SECURE_ERASE:
191 case REQ_OP_WRITE_ZEROES:
193 case REQ_OP_WRITE_SAME:
199 bio_for_each_segment(bv, bio, iter)
206 * get a reference to a bio, so it won't disappear. the intended use is
210 * submit_bio(rw, bio);
211 * if (bio->bi_flags ...)
215 * without the bio_get(), it could potentially complete I/O before submit_bio
216 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
219 static inline void bio_get(struct bio *bio)
221 bio->bi_flags |= (1 << BIO_REFFED);
222 smp_mb__before_atomic();
223 atomic_inc(&bio->__bi_cnt);
226 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
229 bio->bi_flags |= (1 << BIO_REFFED);
232 atomic_set(&bio->__bi_cnt, count);
235 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
237 return (bio->bi_flags & (1U << bit)) != 0;
240 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
242 bio->bi_flags |= (1U << bit);
245 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
247 bio->bi_flags &= ~(1U << bit);
250 static inline void bio_get_first_bvec(struct bio *bio, struct bio_vec *bv)
252 *bv = mp_bvec_iter_bvec(bio->bi_io_vec, bio->bi_iter);
255 static inline void bio_get_last_bvec(struct bio *bio, struct bio_vec *bv)
257 struct bvec_iter iter = bio->bi_iter;
260 bio_get_first_bvec(bio, bv);
261 if (bv->bv_len == bio->bi_iter.bi_size)
262 return; /* this bio only has a single bvec */
264 bio_advance_iter(bio, &iter, iter.bi_size);
266 if (!iter.bi_bvec_done)
267 idx = iter.bi_idx - 1;
268 else /* in the middle of bvec */
271 *bv = bio->bi_io_vec[idx];
274 * iter.bi_bvec_done records actual length of the last bvec
275 * if this bio ends in the middle of one io vector
277 if (iter.bi_bvec_done)
278 bv->bv_len = iter.bi_bvec_done;
281 static inline struct bio_vec *bio_first_bvec_all(struct bio *bio)
283 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
284 return bio->bi_io_vec;
287 static inline struct page *bio_first_page_all(struct bio *bio)
289 return bio_first_bvec_all(bio)->bv_page;
292 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
294 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
295 return &bio->bi_io_vec[bio->bi_vcnt - 1];
299 BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */
300 BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */
301 BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */
302 BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */
303 BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */
307 * bio integrity payload
309 struct bio_integrity_payload {
310 struct bio *bip_bio; /* parent bio */
312 struct bvec_iter bip_iter;
314 unsigned short bip_slab; /* slab the bip came from */
315 unsigned short bip_vcnt; /* # of integrity bio_vecs */
316 unsigned short bip_max_vcnt; /* integrity bio_vec slots */
317 unsigned short bip_flags; /* control flags */
319 struct bvec_iter bio_iter; /* for rewinding parent bio */
321 struct work_struct bip_work; /* I/O completion */
323 struct bio_vec *bip_vec;
324 struct bio_vec bip_inline_vecs[];/* embedded bvec array */
327 #if defined(CONFIG_BLK_DEV_INTEGRITY)
329 static inline struct bio_integrity_payload *bio_integrity(struct bio *bio)
331 if (bio->bi_opf & REQ_INTEGRITY)
332 return bio->bi_integrity;
337 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
339 struct bio_integrity_payload *bip = bio_integrity(bio);
342 return bip->bip_flags & flag;
347 static inline sector_t bip_get_seed(struct bio_integrity_payload *bip)
349 return bip->bip_iter.bi_sector;
352 static inline void bip_set_seed(struct bio_integrity_payload *bip,
355 bip->bip_iter.bi_sector = seed;
358 #endif /* CONFIG_BLK_DEV_INTEGRITY */
360 extern void bio_trim(struct bio *bio, int offset, int size);
361 extern struct bio *bio_split(struct bio *bio, int sectors,
362 gfp_t gfp, struct bio_set *bs);
365 * bio_next_split - get next @sectors from a bio, splitting if necessary
367 * @sectors: number of sectors to split from the front of @bio
369 * @bs: bio set to allocate from
371 * Returns a bio representing the next @sectors of @bio - if the bio is smaller
372 * than @sectors, returns the original bio unchanged.
374 static inline struct bio *bio_next_split(struct bio *bio, int sectors,
375 gfp_t gfp, struct bio_set *bs)
377 if (sectors >= bio_sectors(bio))
380 return bio_split(bio, sectors, gfp, bs);
384 BIOSET_NEED_BVECS = BIT(0),
385 BIOSET_NEED_RESCUER = BIT(1),
387 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
388 extern void bioset_exit(struct bio_set *);
389 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
390 extern int bioset_init_from_src(struct bio_set *bs, struct bio_set *src);
392 extern struct bio *bio_alloc_bioset(gfp_t, unsigned int, struct bio_set *);
393 extern void bio_put(struct bio *);
395 extern void __bio_clone_fast(struct bio *, struct bio *);
396 extern struct bio *bio_clone_fast(struct bio *, gfp_t, struct bio_set *);
398 extern struct bio_set fs_bio_set;
400 static inline struct bio *bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
402 return bio_alloc_bioset(gfp_mask, nr_iovecs, &fs_bio_set);
405 static inline struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned int nr_iovecs)
407 return bio_alloc_bioset(gfp_mask, nr_iovecs, NULL);
410 extern blk_qc_t submit_bio(struct bio *);
412 extern void bio_endio(struct bio *);
414 static inline void bio_io_error(struct bio *bio)
416 bio->bi_status = BLK_STS_IOERR;
420 static inline void bio_wouldblock_error(struct bio *bio)
422 bio_set_flag(bio, BIO_QUIET);
423 bio->bi_status = BLK_STS_AGAIN;
427 struct request_queue;
429 extern int submit_bio_wait(struct bio *bio);
430 extern void bio_advance(struct bio *, unsigned);
432 extern void bio_init(struct bio *bio, struct bio_vec *table,
433 unsigned short max_vecs);
434 extern void bio_uninit(struct bio *);
435 extern void bio_reset(struct bio *);
436 void bio_chain(struct bio *, struct bio *);
438 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
439 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
440 unsigned int, unsigned int);
441 bool __bio_try_merge_page(struct bio *bio, struct page *page,
442 unsigned int len, unsigned int off, bool *same_page);
443 void __bio_add_page(struct bio *bio, struct page *page,
444 unsigned int len, unsigned int off);
445 int bio_iov_iter_get_pages(struct bio *bio, 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_list_copy_data(struct bio *dst, struct bio *src);
454 extern void bio_free_pages(struct bio *bio);
455 void zero_fill_bio_iter(struct bio *bio, struct bvec_iter iter);
456 void bio_truncate(struct bio *bio, unsigned new_size);
457 void guard_bio_eod(struct bio *bio);
459 static inline void zero_fill_bio(struct bio *bio)
461 zero_fill_bio_iter(bio, bio->bi_iter);
464 extern struct bio_vec *bvec_alloc(gfp_t, int, unsigned long *, mempool_t *);
465 extern void bvec_free(mempool_t *, struct bio_vec *, unsigned int);
466 extern unsigned int bvec_nr_vecs(unsigned short idx);
467 extern const char *bio_devname(struct bio *bio, char *buffer);
469 #define bio_set_dev(bio, bdev) \
471 if ((bio)->bi_disk != (bdev)->bd_disk) \
472 bio_clear_flag(bio, BIO_THROTTLED);\
473 (bio)->bi_disk = (bdev)->bd_disk; \
474 (bio)->bi_partno = (bdev)->bd_partno; \
475 bio_associate_blkg(bio); \
478 #define bio_copy_dev(dst, src) \
480 (dst)->bi_disk = (src)->bi_disk; \
481 (dst)->bi_partno = (src)->bi_partno; \
482 bio_clone_blkg_association(dst, src); \
485 #define bio_dev(bio) \
486 disk_devt((bio)->bi_disk)
488 #ifdef CONFIG_BLK_CGROUP
489 void bio_associate_blkg(struct bio *bio);
490 void bio_associate_blkg_from_css(struct bio *bio,
491 struct cgroup_subsys_state *css);
492 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
493 #else /* CONFIG_BLK_CGROUP */
494 static inline void bio_associate_blkg(struct bio *bio) { }
495 static inline void bio_associate_blkg_from_css(struct bio *bio,
496 struct cgroup_subsys_state *css)
498 static inline void bio_clone_blkg_association(struct bio *dst,
500 #endif /* CONFIG_BLK_CGROUP */
502 #ifdef CONFIG_HIGHMEM
504 * remember never ever reenable interrupts between a bvec_kmap_irq and
507 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
512 * might not be a highmem page, but the preempt/irq count
513 * balancing is a lot nicer this way
515 local_irq_save(*flags);
516 addr = (unsigned long) kmap_atomic(bvec->bv_page);
518 BUG_ON(addr & ~PAGE_MASK);
520 return (char *) addr + bvec->bv_offset;
523 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
525 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
527 kunmap_atomic((void *) ptr);
528 local_irq_restore(*flags);
532 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
534 return page_address(bvec->bv_page) + bvec->bv_offset;
537 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
544 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
546 * A bio_list anchors a singly-linked list of bios chained through the bi_next
547 * member of the bio. The bio_list also caches the last list member to allow
548 * fast access to the tail.
555 static inline int bio_list_empty(const struct bio_list *bl)
557 return bl->head == NULL;
560 static inline void bio_list_init(struct bio_list *bl)
562 bl->head = bl->tail = NULL;
565 #define BIO_EMPTY_LIST { NULL, NULL }
567 #define bio_list_for_each(bio, bl) \
568 for (bio = (bl)->head; bio; bio = bio->bi_next)
570 static inline unsigned bio_list_size(const struct bio_list *bl)
575 bio_list_for_each(bio, bl)
581 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
586 bl->tail->bi_next = bio;
593 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
595 bio->bi_next = bl->head;
603 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
609 bl->tail->bi_next = bl2->head;
611 bl->head = bl2->head;
613 bl->tail = bl2->tail;
616 static inline void bio_list_merge_head(struct bio_list *bl,
617 struct bio_list *bl2)
623 bl2->tail->bi_next = bl->head;
625 bl->tail = bl2->tail;
627 bl->head = bl2->head;
630 static inline struct bio *bio_list_peek(struct bio_list *bl)
635 static inline struct bio *bio_list_pop(struct bio_list *bl)
637 struct bio *bio = bl->head;
640 bl->head = bl->head->bi_next;
650 static inline struct bio *bio_list_get(struct bio_list *bl)
652 struct bio *bio = bl->head;
654 bl->head = bl->tail = NULL;
660 * Increment chain count for the bio. Make sure the CHAIN flag update
661 * is visible before the raised count.
663 static inline void bio_inc_remaining(struct bio *bio)
665 bio_set_flag(bio, BIO_CHAIN);
666 smp_mb__before_atomic();
667 atomic_inc(&bio->__bi_remaining);
671 * bio_set is used to allow other portions of the IO system to
672 * allocate their own private memory pools for bio and iovec structures.
673 * These memory pools in turn all allocate from the bio_slab
674 * and the bvec_slabs[].
676 #define BIO_POOL_SIZE 2
679 struct kmem_cache *bio_slab;
680 unsigned int front_pad;
684 #if defined(CONFIG_BLK_DEV_INTEGRITY)
685 mempool_t bio_integrity_pool;
686 mempool_t bvec_integrity_pool;
690 * Deadlock avoidance for stacking block drivers: see comments in
691 * bio_alloc_bioset() for details
693 spinlock_t rescue_lock;
694 struct bio_list rescue_list;
695 struct work_struct rescue_work;
696 struct workqueue_struct *rescue_workqueue;
702 struct kmem_cache *slab;
705 static inline bool bioset_initialized(struct bio_set *bs)
707 return bs->bio_slab != NULL;
711 * a small number of entries is fine, not going to be performance critical.
712 * basically we just need to survive
714 #define BIO_SPLIT_ENTRIES 2
716 #if defined(CONFIG_BLK_DEV_INTEGRITY)
718 #define bip_for_each_vec(bvl, bip, iter) \
719 for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter)
721 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \
723 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter)
725 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
726 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
727 extern bool bio_integrity_prep(struct bio *);
728 extern void bio_integrity_advance(struct bio *, unsigned int);
729 extern void bio_integrity_trim(struct bio *);
730 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
731 extern int bioset_integrity_create(struct bio_set *, int);
732 extern void bioset_integrity_free(struct bio_set *);
733 extern void bio_integrity_init(void);
735 #else /* CONFIG_BLK_DEV_INTEGRITY */
737 static inline void *bio_integrity(struct bio *bio)
742 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size)
747 static inline void bioset_integrity_free (struct bio_set *bs)
752 static inline bool bio_integrity_prep(struct bio *bio)
757 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
763 static inline void bio_integrity_advance(struct bio *bio,
764 unsigned int bytes_done)
769 static inline void bio_integrity_trim(struct bio *bio)
774 static inline void bio_integrity_init(void)
779 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
784 static inline void *bio_integrity_alloc(struct bio * bio, gfp_t gfp,
787 return ERR_PTR(-EINVAL);
790 static inline int bio_integrity_add_page(struct bio *bio, struct page *page,
791 unsigned int len, unsigned int offset)
796 #endif /* CONFIG_BLK_DEV_INTEGRITY */
799 * Mark a bio as polled. Note that for async polled IO, the caller must
800 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
801 * We cannot block waiting for requests on polled IO, as those completions
802 * must be found by the caller. This is different than IRQ driven IO, where
803 * it's safe to wait for IO to complete.
805 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
807 bio->bi_opf |= REQ_HIPRI;
808 if (!is_sync_kiocb(kiocb))
809 bio->bi_opf |= REQ_NOWAIT;
812 #endif /* __LINUX_BIO_H */