1 /* SPDX-License-Identifier: GPL-2.0 */
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>
13 #include <linux/uio.h>
18 #define BIO_BUG_ON BUG_ON
23 #define BIO_MAX_VECS 256U
25 static inline unsigned int bio_max_segs(unsigned int nr_segs)
27 return min(nr_segs, BIO_MAX_VECS);
30 #define bio_prio(bio) (bio)->bi_ioprio
31 #define bio_set_prio(bio, prio) ((bio)->bi_ioprio = prio)
33 #define bio_iter_iovec(bio, iter) \
34 bvec_iter_bvec((bio)->bi_io_vec, (iter))
36 #define bio_iter_page(bio, iter) \
37 bvec_iter_page((bio)->bi_io_vec, (iter))
38 #define bio_iter_len(bio, iter) \
39 bvec_iter_len((bio)->bi_io_vec, (iter))
40 #define bio_iter_offset(bio, iter) \
41 bvec_iter_offset((bio)->bi_io_vec, (iter))
43 #define bio_page(bio) bio_iter_page((bio), (bio)->bi_iter)
44 #define bio_offset(bio) bio_iter_offset((bio), (bio)->bi_iter)
45 #define bio_iovec(bio) bio_iter_iovec((bio), (bio)->bi_iter)
47 #define bio_multiple_segments(bio) \
48 ((bio)->bi_iter.bi_size != bio_iovec(bio).bv_len)
50 #define bvec_iter_sectors(iter) ((iter).bi_size >> 9)
51 #define bvec_iter_end_sector(iter) ((iter).bi_sector + bvec_iter_sectors((iter)))
53 #define bio_sectors(bio) bvec_iter_sectors((bio)->bi_iter)
54 #define bio_end_sector(bio) bvec_iter_end_sector((bio)->bi_iter)
57 * Return the data direction, READ or WRITE.
59 #define bio_data_dir(bio) \
60 (op_is_write(bio_op(bio)) ? WRITE : READ)
63 * Check whether this bio carries any data or not. A NULL bio is allowed.
65 static inline bool bio_has_data(struct bio *bio)
68 bio->bi_iter.bi_size &&
69 bio_op(bio) != REQ_OP_DISCARD &&
70 bio_op(bio) != REQ_OP_SECURE_ERASE &&
71 bio_op(bio) != REQ_OP_WRITE_ZEROES)
77 static inline bool bio_no_advance_iter(const struct bio *bio)
79 return bio_op(bio) == REQ_OP_DISCARD ||
80 bio_op(bio) == REQ_OP_SECURE_ERASE ||
81 bio_op(bio) == REQ_OP_WRITE_SAME ||
82 bio_op(bio) == REQ_OP_WRITE_ZEROES;
85 static inline bool bio_mergeable(struct bio *bio)
87 if (bio->bi_opf & REQ_NOMERGE_FLAGS)
93 static inline unsigned int bio_cur_bytes(struct bio *bio)
95 if (bio_has_data(bio))
96 return bio_iovec(bio).bv_len;
97 else /* dataless requests such as discard */
98 return bio->bi_iter.bi_size;
101 static inline void *bio_data(struct bio *bio)
103 if (bio_has_data(bio))
104 return page_address(bio_page(bio)) + bio_offset(bio);
109 extern unsigned int bio_max_size(struct bio *bio);
112 * bio_full - check if the bio is full
114 * @len: length of one segment to be added
116 * Return true if @bio is full and one segment with @len bytes can't be
117 * added to the bio, otherwise return false
119 static inline bool bio_full(struct bio *bio, unsigned len)
121 if (bio->bi_vcnt >= bio->bi_max_vecs)
124 if (bio->bi_iter.bi_size > bio_max_size(bio) - len)
130 static inline bool bio_next_segment(const struct bio *bio,
131 struct bvec_iter_all *iter)
133 if (iter->idx >= bio->bi_vcnt)
136 bvec_advance(&bio->bi_io_vec[iter->idx], iter);
141 * drivers should _never_ use the all version - the bio may have been split
142 * before it got to the driver and the driver won't own all of it
144 #define bio_for_each_segment_all(bvl, bio, iter) \
145 for (bvl = bvec_init_iter_all(&iter); bio_next_segment((bio), &iter); )
147 static inline void bio_advance_iter(const struct bio *bio,
148 struct bvec_iter *iter, unsigned int bytes)
150 iter->bi_sector += bytes >> 9;
152 if (bio_no_advance_iter(bio))
153 iter->bi_size -= bytes;
155 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
156 /* TODO: It is reasonable to complete bio with error here. */
159 /* @bytes should be less or equal to bvec[i->bi_idx].bv_len */
160 static inline void bio_advance_iter_single(const struct bio *bio,
161 struct bvec_iter *iter,
164 iter->bi_sector += bytes >> 9;
166 if (bio_no_advance_iter(bio))
167 iter->bi_size -= bytes;
169 bvec_iter_advance_single(bio->bi_io_vec, iter, bytes);
172 #define __bio_for_each_segment(bvl, bio, iter, start) \
173 for (iter = (start); \
175 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
176 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
178 #define bio_for_each_segment(bvl, bio, iter) \
179 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
181 #define __bio_for_each_bvec(bvl, bio, iter, start) \
182 for (iter = (start); \
184 ((bvl = mp_bvec_iter_bvec((bio)->bi_io_vec, (iter))), 1); \
185 bio_advance_iter_single((bio), &(iter), (bvl).bv_len))
187 /* iterate over multi-page bvec */
188 #define bio_for_each_bvec(bvl, bio, iter) \
189 __bio_for_each_bvec(bvl, bio, iter, (bio)->bi_iter)
192 * Iterate over all multi-page bvecs. Drivers shouldn't use this version for the
193 * same reasons as bio_for_each_segment_all().
195 #define bio_for_each_bvec_all(bvl, bio, i) \
196 for (i = 0, bvl = bio_first_bvec_all(bio); \
197 i < (bio)->bi_vcnt; i++, bvl++) \
199 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
201 static inline unsigned bio_segments(struct bio *bio)
205 struct bvec_iter iter;
208 * We special case discard/write same/write zeroes, because they
209 * interpret bi_size differently:
212 switch (bio_op(bio)) {
214 case REQ_OP_SECURE_ERASE:
215 case REQ_OP_WRITE_ZEROES:
217 case REQ_OP_WRITE_SAME:
223 bio_for_each_segment(bv, bio, iter)
230 * get a reference to a bio, so it won't disappear. the intended use is
234 * submit_bio(rw, bio);
235 * if (bio->bi_flags ...)
239 * without the bio_get(), it could potentially complete I/O before submit_bio
240 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
243 static inline void bio_get(struct bio *bio)
245 bio->bi_flags |= (1 << BIO_REFFED);
246 smp_mb__before_atomic();
247 atomic_inc(&bio->__bi_cnt);
250 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
253 bio->bi_flags |= (1 << BIO_REFFED);
256 atomic_set(&bio->__bi_cnt, count);
259 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
261 return (bio->bi_flags & (1U << bit)) != 0;
264 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
266 bio->bi_flags |= (1U << bit);
269 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
271 bio->bi_flags &= ~(1U << bit);
274 static inline void bio_get_first_bvec(struct bio *bio, struct bio_vec *bv)
276 *bv = bio_iovec(bio);
279 static inline void bio_get_last_bvec(struct bio *bio, struct bio_vec *bv)
281 struct bvec_iter iter = bio->bi_iter;
284 if (unlikely(!bio_multiple_segments(bio))) {
285 *bv = bio_iovec(bio);
289 bio_advance_iter(bio, &iter, iter.bi_size);
291 if (!iter.bi_bvec_done)
292 idx = iter.bi_idx - 1;
293 else /* in the middle of bvec */
296 *bv = bio->bi_io_vec[idx];
299 * iter.bi_bvec_done records actual length of the last bvec
300 * if this bio ends in the middle of one io vector
302 if (iter.bi_bvec_done)
303 bv->bv_len = iter.bi_bvec_done;
306 static inline struct bio_vec *bio_first_bvec_all(struct bio *bio)
308 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
309 return bio->bi_io_vec;
312 static inline struct page *bio_first_page_all(struct bio *bio)
314 return bio_first_bvec_all(bio)->bv_page;
317 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
319 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
320 return &bio->bi_io_vec[bio->bi_vcnt - 1];
324 BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */
325 BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */
326 BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */
327 BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */
328 BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */
332 * bio integrity payload
334 struct bio_integrity_payload {
335 struct bio *bip_bio; /* parent bio */
337 struct bvec_iter bip_iter;
339 unsigned short bip_vcnt; /* # of integrity bio_vecs */
340 unsigned short bip_max_vcnt; /* integrity bio_vec slots */
341 unsigned short bip_flags; /* control flags */
343 struct bvec_iter bio_iter; /* for rewinding parent bio */
345 struct work_struct bip_work; /* I/O completion */
347 struct bio_vec *bip_vec;
348 struct bio_vec bip_inline_vecs[];/* embedded bvec array */
351 #if defined(CONFIG_BLK_DEV_INTEGRITY)
353 static inline struct bio_integrity_payload *bio_integrity(struct bio *bio)
355 if (bio->bi_opf & REQ_INTEGRITY)
356 return bio->bi_integrity;
361 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
363 struct bio_integrity_payload *bip = bio_integrity(bio);
366 return bip->bip_flags & flag;
371 static inline sector_t bip_get_seed(struct bio_integrity_payload *bip)
373 return bip->bip_iter.bi_sector;
376 static inline void bip_set_seed(struct bio_integrity_payload *bip,
379 bip->bip_iter.bi_sector = seed;
382 #endif /* CONFIG_BLK_DEV_INTEGRITY */
384 extern void bio_trim(struct bio *bio, int offset, int size);
385 extern struct bio *bio_split(struct bio *bio, int sectors,
386 gfp_t gfp, struct bio_set *bs);
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 * Returns 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),
411 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
412 extern void bioset_exit(struct bio_set *);
413 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
414 extern int bioset_init_from_src(struct bio_set *bs, struct bio_set *src);
416 struct bio *bio_alloc_bioset(gfp_t gfp, unsigned short nr_iovecs,
418 struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned short nr_iovecs);
419 extern void bio_put(struct bio *);
421 extern void __bio_clone_fast(struct bio *, struct bio *);
422 extern struct bio *bio_clone_fast(struct bio *, gfp_t, struct bio_set *);
424 extern struct bio_set fs_bio_set;
426 static inline struct bio *bio_alloc(gfp_t gfp_mask, unsigned short nr_iovecs)
428 return bio_alloc_bioset(gfp_mask, nr_iovecs, &fs_bio_set);
431 extern blk_qc_t submit_bio(struct bio *);
433 extern void bio_endio(struct bio *);
435 static inline void bio_io_error(struct bio *bio)
437 bio->bi_status = BLK_STS_IOERR;
441 static inline void bio_wouldblock_error(struct bio *bio)
443 bio_set_flag(bio, BIO_QUIET);
444 bio->bi_status = BLK_STS_AGAIN;
449 * Calculate number of bvec segments that should be allocated to fit data
450 * pointed by @iter. If @iter is backed by bvec it's going to be reused
451 * instead of allocating a new one.
453 static inline int bio_iov_vecs_to_alloc(struct iov_iter *iter, int max_segs)
455 if (iov_iter_is_bvec(iter))
457 return iov_iter_npages(iter, max_segs);
460 struct request_queue;
462 extern int submit_bio_wait(struct bio *bio);
463 extern void bio_advance(struct bio *, unsigned);
465 extern void bio_init(struct bio *bio, struct bio_vec *table,
466 unsigned short max_vecs);
467 extern void bio_uninit(struct bio *);
468 extern void bio_reset(struct bio *);
469 void bio_chain(struct bio *, struct bio *);
471 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
472 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
473 unsigned int, unsigned int);
474 int bio_add_zone_append_page(struct bio *bio, struct page *page,
475 unsigned int len, unsigned int offset);
476 bool __bio_try_merge_page(struct bio *bio, struct page *page,
477 unsigned int len, unsigned int off, bool *same_page);
478 void __bio_add_page(struct bio *bio, struct page *page,
479 unsigned int len, unsigned int off);
480 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
481 void bio_release_pages(struct bio *bio, bool mark_dirty);
482 extern void bio_set_pages_dirty(struct bio *bio);
483 extern void bio_check_pages_dirty(struct bio *bio);
485 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
486 struct bio *src, struct bvec_iter *src_iter);
487 extern void bio_copy_data(struct bio *dst, struct bio *src);
488 extern void bio_free_pages(struct bio *bio);
489 void bio_truncate(struct bio *bio, unsigned new_size);
490 void guard_bio_eod(struct bio *bio);
491 void zero_fill_bio(struct bio *bio);
493 extern const char *bio_devname(struct bio *bio, char *buffer);
495 #define bio_set_dev(bio, bdev) \
497 bio_clear_flag(bio, BIO_REMAPPED); \
498 if ((bio)->bi_bdev != (bdev)) \
499 bio_clear_flag(bio, BIO_THROTTLED); \
500 (bio)->bi_bdev = (bdev); \
501 bio_associate_blkg(bio); \
504 #define bio_copy_dev(dst, src) \
506 bio_clear_flag(dst, BIO_REMAPPED); \
507 (dst)->bi_bdev = (src)->bi_bdev; \
508 bio_clone_blkg_association(dst, src); \
511 #define bio_dev(bio) \
512 disk_devt((bio)->bi_bdev->bd_disk)
514 #ifdef CONFIG_BLK_CGROUP
515 void bio_associate_blkg(struct bio *bio);
516 void bio_associate_blkg_from_css(struct bio *bio,
517 struct cgroup_subsys_state *css);
518 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
519 #else /* CONFIG_BLK_CGROUP */
520 static inline void bio_associate_blkg(struct bio *bio) { }
521 static inline void bio_associate_blkg_from_css(struct bio *bio,
522 struct cgroup_subsys_state *css)
524 static inline void bio_clone_blkg_association(struct bio *dst,
526 #endif /* CONFIG_BLK_CGROUP */
528 #ifdef CONFIG_HIGHMEM
530 * remember never ever reenable interrupts between a bvec_kmap_irq and
533 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
538 * might not be a highmem page, but the preempt/irq count
539 * balancing is a lot nicer this way
541 local_irq_save(*flags);
542 addr = (unsigned long) kmap_atomic(bvec->bv_page);
544 BUG_ON(addr & ~PAGE_MASK);
546 return (char *) addr + bvec->bv_offset;
549 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
551 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
553 kunmap_atomic((void *) ptr);
554 local_irq_restore(*flags);
558 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
560 return page_address(bvec->bv_page) + bvec->bv_offset;
563 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
570 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
572 * A bio_list anchors a singly-linked list of bios chained through the bi_next
573 * member of the bio. The bio_list also caches the last list member to allow
574 * fast access to the tail.
581 static inline int bio_list_empty(const struct bio_list *bl)
583 return bl->head == NULL;
586 static inline void bio_list_init(struct bio_list *bl)
588 bl->head = bl->tail = NULL;
591 #define BIO_EMPTY_LIST { NULL, NULL }
593 #define bio_list_for_each(bio, bl) \
594 for (bio = (bl)->head; bio; bio = bio->bi_next)
596 static inline unsigned bio_list_size(const struct bio_list *bl)
601 bio_list_for_each(bio, bl)
607 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
612 bl->tail->bi_next = bio;
619 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
621 bio->bi_next = bl->head;
629 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
635 bl->tail->bi_next = bl2->head;
637 bl->head = bl2->head;
639 bl->tail = bl2->tail;
642 static inline void bio_list_merge_head(struct bio_list *bl,
643 struct bio_list *bl2)
649 bl2->tail->bi_next = bl->head;
651 bl->tail = bl2->tail;
653 bl->head = bl2->head;
656 static inline struct bio *bio_list_peek(struct bio_list *bl)
661 static inline struct bio *bio_list_pop(struct bio_list *bl)
663 struct bio *bio = bl->head;
666 bl->head = bl->head->bi_next;
676 static inline struct bio *bio_list_get(struct bio_list *bl)
678 struct bio *bio = bl->head;
680 bl->head = bl->tail = NULL;
686 * Increment chain count for the bio. Make sure the CHAIN flag update
687 * is visible before the raised count.
689 static inline void bio_inc_remaining(struct bio *bio)
691 bio_set_flag(bio, BIO_CHAIN);
692 smp_mb__before_atomic();
693 atomic_inc(&bio->__bi_remaining);
697 * bio_set is used to allow other portions of the IO system to
698 * allocate their own private memory pools for bio and iovec structures.
699 * These memory pools in turn all allocate from the bio_slab
700 * and the bvec_slabs[].
702 #define BIO_POOL_SIZE 2
705 struct kmem_cache *bio_slab;
706 unsigned int front_pad;
710 #if defined(CONFIG_BLK_DEV_INTEGRITY)
711 mempool_t bio_integrity_pool;
712 mempool_t bvec_integrity_pool;
715 unsigned int back_pad;
717 * Deadlock avoidance for stacking block drivers: see comments in
718 * bio_alloc_bioset() for details
720 spinlock_t rescue_lock;
721 struct bio_list rescue_list;
722 struct work_struct rescue_work;
723 struct workqueue_struct *rescue_workqueue;
726 static inline bool bioset_initialized(struct bio_set *bs)
728 return bs->bio_slab != NULL;
731 #if defined(CONFIG_BLK_DEV_INTEGRITY)
733 #define bip_for_each_vec(bvl, bip, iter) \
734 for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter)
736 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \
738 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter)
740 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
741 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
742 extern bool bio_integrity_prep(struct bio *);
743 extern void bio_integrity_advance(struct bio *, unsigned int);
744 extern void bio_integrity_trim(struct bio *);
745 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
746 extern int bioset_integrity_create(struct bio_set *, int);
747 extern void bioset_integrity_free(struct bio_set *);
748 extern void bio_integrity_init(void);
750 #else /* CONFIG_BLK_DEV_INTEGRITY */
752 static inline void *bio_integrity(struct bio *bio)
757 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size)
762 static inline void bioset_integrity_free (struct bio_set *bs)
767 static inline bool bio_integrity_prep(struct bio *bio)
772 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
778 static inline void bio_integrity_advance(struct bio *bio,
779 unsigned int bytes_done)
784 static inline void bio_integrity_trim(struct bio *bio)
789 static inline void bio_integrity_init(void)
794 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
799 static inline void *bio_integrity_alloc(struct bio * bio, gfp_t gfp,
802 return ERR_PTR(-EINVAL);
805 static inline int bio_integrity_add_page(struct bio *bio, struct page *page,
806 unsigned int len, unsigned int offset)
811 #endif /* CONFIG_BLK_DEV_INTEGRITY */
814 * Mark a bio as polled. Note that for async polled IO, the caller must
815 * expect -EWOULDBLOCK if we cannot allocate a request (or other resources).
816 * We cannot block waiting for requests on polled IO, as those completions
817 * must be found by the caller. This is different than IRQ driven IO, where
818 * it's safe to wait for IO to complete.
820 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
822 bio->bi_opf |= REQ_HIPRI;
823 if (!is_sync_kiocb(kiocb))
824 bio->bi_opf |= REQ_NOWAIT;
827 #endif /* __LINUX_BIO_H */