1 /* SPDX-License-Identifier: GPL-2.0 */
8 #include <linux/highmem.h>
9 #include <linux/mempool.h>
10 #include <linux/ioprio.h>
13 /* struct bio, bio_vec and BIO_* flags are defined in blk_types.h */
14 #include <linux/blk_types.h>
19 #define BIO_BUG_ON BUG_ON
24 #define BIO_MAX_PAGES 256
26 #define bio_prio(bio) (bio)->bi_ioprio
27 #define bio_set_prio(bio, prio) ((bio)->bi_ioprio = prio)
29 #define bio_iter_iovec(bio, iter) \
30 bvec_iter_bvec((bio)->bi_io_vec, (iter))
32 #define bio_iter_page(bio, iter) \
33 bvec_iter_page((bio)->bi_io_vec, (iter))
34 #define bio_iter_len(bio, iter) \
35 bvec_iter_len((bio)->bi_io_vec, (iter))
36 #define bio_iter_offset(bio, iter) \
37 bvec_iter_offset((bio)->bi_io_vec, (iter))
39 #define bio_page(bio) bio_iter_page((bio), (bio)->bi_iter)
40 #define bio_offset(bio) bio_iter_offset((bio), (bio)->bi_iter)
41 #define bio_iovec(bio) bio_iter_iovec((bio), (bio)->bi_iter)
43 #define bio_multiple_segments(bio) \
44 ((bio)->bi_iter.bi_size != bio_iovec(bio).bv_len)
46 #define bvec_iter_sectors(iter) ((iter).bi_size >> 9)
47 #define bvec_iter_end_sector(iter) ((iter).bi_sector + bvec_iter_sectors((iter)))
49 #define bio_sectors(bio) bvec_iter_sectors((bio)->bi_iter)
50 #define bio_end_sector(bio) bvec_iter_end_sector((bio)->bi_iter)
53 * Return the data direction, READ or WRITE.
55 #define bio_data_dir(bio) \
56 (op_is_write(bio_op(bio)) ? WRITE : READ)
59 * Check whether this bio carries any data or not. A NULL bio is allowed.
61 static inline bool bio_has_data(struct bio *bio)
64 bio->bi_iter.bi_size &&
65 bio_op(bio) != REQ_OP_DISCARD &&
66 bio_op(bio) != REQ_OP_SECURE_ERASE &&
67 bio_op(bio) != REQ_OP_WRITE_ZEROES)
73 static inline bool bio_no_advance_iter(const struct bio *bio)
75 return bio_op(bio) == REQ_OP_DISCARD ||
76 bio_op(bio) == REQ_OP_SECURE_ERASE ||
77 bio_op(bio) == REQ_OP_WRITE_SAME ||
78 bio_op(bio) == REQ_OP_WRITE_ZEROES;
81 static inline bool bio_mergeable(struct bio *bio)
83 if (bio->bi_opf & REQ_NOMERGE_FLAGS)
89 static inline unsigned int bio_cur_bytes(struct bio *bio)
91 if (bio_has_data(bio))
92 return bio_iovec(bio).bv_len;
93 else /* dataless requests such as discard */
94 return bio->bi_iter.bi_size;
97 static inline void *bio_data(struct bio *bio)
99 if (bio_has_data(bio))
100 return page_address(bio_page(bio)) + bio_offset(bio);
106 * bio_full - check if the bio is full
108 * @len: length of one segment to be added
110 * Return true if @bio is full and one segment with @len bytes can't be
111 * added to the bio, otherwise return false
113 static inline bool bio_full(struct bio *bio, unsigned len)
115 if (bio->bi_vcnt >= bio->bi_max_vecs)
118 if (bio->bi_iter.bi_size > UINT_MAX - len)
124 static inline bool bio_next_segment(const struct bio *bio,
125 struct bvec_iter_all *iter)
127 if (iter->idx >= bio->bi_vcnt)
130 bvec_advance(&bio->bi_io_vec[iter->idx], iter);
135 * drivers should _never_ use the all version - the bio may have been split
136 * before it got to the driver and the driver won't own all of it
138 #define bio_for_each_segment_all(bvl, bio, iter) \
139 for (bvl = bvec_init_iter_all(&iter); bio_next_segment((bio), &iter); )
141 static inline void bio_advance_iter(const struct bio *bio,
142 struct bvec_iter *iter, unsigned int bytes)
144 iter->bi_sector += bytes >> 9;
146 if (bio_no_advance_iter(bio))
147 iter->bi_size -= bytes;
149 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
150 /* TODO: It is reasonable to complete bio with error here. */
153 #define __bio_for_each_segment(bvl, bio, iter, start) \
154 for (iter = (start); \
156 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
157 bio_advance_iter((bio), &(iter), (bvl).bv_len))
159 #define bio_for_each_segment(bvl, bio, iter) \
160 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
162 #define __bio_for_each_bvec(bvl, bio, iter, start) \
163 for (iter = (start); \
165 ((bvl = mp_bvec_iter_bvec((bio)->bi_io_vec, (iter))), 1); \
166 bio_advance_iter((bio), &(iter), (bvl).bv_len))
168 /* iterate over multi-page bvec */
169 #define bio_for_each_bvec(bvl, bio, iter) \
170 __bio_for_each_bvec(bvl, bio, iter, (bio)->bi_iter)
173 * Iterate over all multi-page bvecs. Drivers shouldn't use this version for the
174 * same reasons as bio_for_each_segment_all().
176 #define bio_for_each_bvec_all(bvl, bio, i) \
177 for (i = 0, bvl = bio_first_bvec_all(bio); \
178 i < (bio)->bi_vcnt; i++, bvl++) \
180 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
182 static inline unsigned bio_segments(struct bio *bio)
186 struct bvec_iter iter;
189 * We special case discard/write same/write zeroes, because they
190 * interpret bi_size differently:
193 switch (bio_op(bio)) {
195 case REQ_OP_SECURE_ERASE:
196 case REQ_OP_WRITE_ZEROES:
198 case REQ_OP_WRITE_SAME:
204 bio_for_each_segment(bv, bio, iter)
211 * get a reference to a bio, so it won't disappear. the intended use is
215 * submit_bio(rw, bio);
216 * if (bio->bi_flags ...)
220 * without the bio_get(), it could potentially complete I/O before submit_bio
221 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
224 static inline void bio_get(struct bio *bio)
226 bio->bi_flags |= (1 << BIO_REFFED);
227 smp_mb__before_atomic();
228 atomic_inc(&bio->__bi_cnt);
231 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
234 bio->bi_flags |= (1 << BIO_REFFED);
237 atomic_set(&bio->__bi_cnt, count);
240 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
242 return (bio->bi_flags & (1U << bit)) != 0;
245 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
247 bio->bi_flags |= (1U << bit);
250 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
252 bio->bi_flags &= ~(1U << bit);
255 static inline void bio_get_first_bvec(struct bio *bio, struct bio_vec *bv)
257 *bv = bio_iovec(bio);
260 static inline void bio_get_last_bvec(struct bio *bio, struct bio_vec *bv)
262 struct bvec_iter iter = bio->bi_iter;
265 if (unlikely(!bio_multiple_segments(bio))) {
266 *bv = bio_iovec(bio);
270 bio_advance_iter(bio, &iter, iter.bi_size);
272 if (!iter.bi_bvec_done)
273 idx = iter.bi_idx - 1;
274 else /* in the middle of bvec */
277 *bv = bio->bi_io_vec[idx];
280 * iter.bi_bvec_done records actual length of the last bvec
281 * if this bio ends in the middle of one io vector
283 if (iter.bi_bvec_done)
284 bv->bv_len = iter.bi_bvec_done;
287 static inline struct bio_vec *bio_first_bvec_all(struct bio *bio)
289 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
290 return bio->bi_io_vec;
293 static inline struct page *bio_first_page_all(struct bio *bio)
295 return bio_first_bvec_all(bio)->bv_page;
298 static inline struct bio_vec *bio_last_bvec_all(struct bio *bio)
300 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
301 return &bio->bi_io_vec[bio->bi_vcnt - 1];
305 BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */
306 BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */
307 BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */
308 BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */
309 BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */
313 * bio integrity payload
315 struct bio_integrity_payload {
316 struct bio *bip_bio; /* parent bio */
318 struct bvec_iter bip_iter;
320 unsigned short bip_slab; /* slab the bip came from */
321 unsigned short bip_vcnt; /* # of integrity bio_vecs */
322 unsigned short bip_max_vcnt; /* integrity bio_vec slots */
323 unsigned short bip_flags; /* control flags */
325 struct bvec_iter bio_iter; /* for rewinding parent bio */
327 struct work_struct bip_work; /* I/O completion */
329 struct bio_vec *bip_vec;
330 struct bio_vec bip_inline_vecs[];/* embedded bvec array */
333 #if defined(CONFIG_BLK_DEV_INTEGRITY)
335 static inline struct bio_integrity_payload *bio_integrity(struct bio *bio)
337 if (bio->bi_opf & REQ_INTEGRITY)
338 return bio->bi_integrity;
343 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
345 struct bio_integrity_payload *bip = bio_integrity(bio);
348 return bip->bip_flags & flag;
353 static inline sector_t bip_get_seed(struct bio_integrity_payload *bip)
355 return bip->bip_iter.bi_sector;
358 static inline void bip_set_seed(struct bio_integrity_payload *bip,
361 bip->bip_iter.bi_sector = seed;
364 #endif /* CONFIG_BLK_DEV_INTEGRITY */
366 extern void bio_trim(struct bio *bio, int offset, int size);
367 extern struct bio *bio_split(struct bio *bio, int sectors,
368 gfp_t gfp, struct bio_set *bs);
371 * bio_next_split - get next @sectors from a bio, splitting if necessary
373 * @sectors: number of sectors to split from the front of @bio
375 * @bs: bio set to allocate from
377 * Returns a bio representing the next @sectors of @bio - if the bio is smaller
378 * than @sectors, returns the original bio unchanged.
380 static inline struct bio *bio_next_split(struct bio *bio, int sectors,
381 gfp_t gfp, struct bio_set *bs)
383 if (sectors >= bio_sectors(bio))
386 return bio_split(bio, sectors, gfp, bs);
390 BIOSET_NEED_BVECS = BIT(0),
391 BIOSET_NEED_RESCUER = BIT(1),
393 extern int bioset_init(struct bio_set *, unsigned int, unsigned int, int flags);
394 extern void bioset_exit(struct bio_set *);
395 extern int biovec_init_pool(mempool_t *pool, int pool_entries);
396 extern int bioset_init_from_src(struct bio_set *bs, struct bio_set *src);
398 extern struct bio *bio_alloc_bioset(gfp_t, unsigned int, struct bio_set *);
399 extern void bio_put(struct bio *);
401 extern void __bio_clone_fast(struct bio *, struct bio *);
402 extern struct bio *bio_clone_fast(struct bio *, gfp_t, struct bio_set *);
404 extern struct bio_set fs_bio_set;
406 static inline struct bio *bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
408 return bio_alloc_bioset(gfp_mask, nr_iovecs, &fs_bio_set);
411 static inline struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned int nr_iovecs)
413 return bio_alloc_bioset(gfp_mask, nr_iovecs, NULL);
416 extern blk_qc_t submit_bio(struct bio *);
418 extern void bio_endio(struct bio *);
420 static inline void bio_io_error(struct bio *bio)
422 bio->bi_status = BLK_STS_IOERR;
426 static inline void bio_wouldblock_error(struct bio *bio)
428 bio_set_flag(bio, BIO_QUIET);
429 bio->bi_status = BLK_STS_AGAIN;
433 struct request_queue;
435 extern int submit_bio_wait(struct bio *bio);
436 extern void bio_advance(struct bio *, unsigned);
438 extern void bio_init(struct bio *bio, struct bio_vec *table,
439 unsigned short max_vecs);
440 extern void bio_uninit(struct bio *);
441 extern void bio_reset(struct bio *);
442 void bio_chain(struct bio *, struct bio *);
444 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
445 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
446 unsigned int, unsigned int);
447 bool __bio_try_merge_page(struct bio *bio, struct page *page,
448 unsigned int len, unsigned int off, bool *same_page);
449 void __bio_add_page(struct bio *bio, struct page *page,
450 unsigned int len, unsigned int off);
451 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter);
452 void bio_release_pages(struct bio *bio, bool mark_dirty);
453 extern void bio_set_pages_dirty(struct bio *bio);
454 extern void bio_check_pages_dirty(struct bio *bio);
456 extern void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
457 struct bio *src, struct bvec_iter *src_iter);
458 extern void bio_copy_data(struct bio *dst, struct bio *src);
459 extern void bio_list_copy_data(struct bio *dst, struct bio *src);
460 extern void bio_free_pages(struct bio *bio);
461 void zero_fill_bio_iter(struct bio *bio, struct bvec_iter iter);
462 void bio_truncate(struct bio *bio, unsigned new_size);
463 void guard_bio_eod(struct bio *bio);
465 static inline void zero_fill_bio(struct bio *bio)
467 zero_fill_bio_iter(bio, bio->bi_iter);
470 extern struct bio_vec *bvec_alloc(gfp_t, int, unsigned long *, mempool_t *);
471 extern void bvec_free(mempool_t *, struct bio_vec *, unsigned int);
472 extern unsigned int bvec_nr_vecs(unsigned short idx);
473 extern const char *bio_devname(struct bio *bio, char *buffer);
475 #define bio_set_dev(bio, bdev) \
477 if ((bio)->bi_disk != (bdev)->bd_disk) \
478 bio_clear_flag(bio, BIO_THROTTLED);\
479 (bio)->bi_disk = (bdev)->bd_disk; \
480 (bio)->bi_partno = (bdev)->bd_partno; \
481 bio_associate_blkg(bio); \
484 #define bio_copy_dev(dst, src) \
486 (dst)->bi_disk = (src)->bi_disk; \
487 (dst)->bi_partno = (src)->bi_partno; \
488 bio_clone_blkg_association(dst, src); \
491 #define bio_dev(bio) \
492 disk_devt((bio)->bi_disk)
494 #if defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP)
495 void bio_associate_blkg_from_page(struct bio *bio, struct page *page);
497 static inline void bio_associate_blkg_from_page(struct bio *bio,
498 struct page *page) { }
501 #ifdef CONFIG_BLK_CGROUP
502 void bio_disassociate_blkg(struct bio *bio);
503 void bio_associate_blkg(struct bio *bio);
504 void bio_associate_blkg_from_css(struct bio *bio,
505 struct cgroup_subsys_state *css);
506 void bio_clone_blkg_association(struct bio *dst, struct bio *src);
507 #else /* CONFIG_BLK_CGROUP */
508 static inline void bio_disassociate_blkg(struct bio *bio) { }
509 static inline void bio_associate_blkg(struct bio *bio) { }
510 static inline void bio_associate_blkg_from_css(struct bio *bio,
511 struct cgroup_subsys_state *css)
513 static inline void bio_clone_blkg_association(struct bio *dst,
515 #endif /* CONFIG_BLK_CGROUP */
517 #ifdef CONFIG_HIGHMEM
519 * remember never ever reenable interrupts between a bvec_kmap_irq and
522 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
527 * might not be a highmem page, but the preempt/irq count
528 * balancing is a lot nicer this way
530 local_irq_save(*flags);
531 addr = (unsigned long) kmap_atomic(bvec->bv_page);
533 BUG_ON(addr & ~PAGE_MASK);
535 return (char *) addr + bvec->bv_offset;
538 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
540 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
542 kunmap_atomic((void *) ptr);
543 local_irq_restore(*flags);
547 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
549 return page_address(bvec->bv_page) + bvec->bv_offset;
552 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
559 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
561 * A bio_list anchors a singly-linked list of bios chained through the bi_next
562 * member of the bio. The bio_list also caches the last list member to allow
563 * fast access to the tail.
570 static inline int bio_list_empty(const struct bio_list *bl)
572 return bl->head == NULL;
575 static inline void bio_list_init(struct bio_list *bl)
577 bl->head = bl->tail = NULL;
580 #define BIO_EMPTY_LIST { NULL, NULL }
582 #define bio_list_for_each(bio, bl) \
583 for (bio = (bl)->head; bio; bio = bio->bi_next)
585 static inline unsigned bio_list_size(const struct bio_list *bl)
590 bio_list_for_each(bio, bl)
596 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
601 bl->tail->bi_next = bio;
608 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
610 bio->bi_next = bl->head;
618 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
624 bl->tail->bi_next = bl2->head;
626 bl->head = bl2->head;
628 bl->tail = bl2->tail;
631 static inline void bio_list_merge_head(struct bio_list *bl,
632 struct bio_list *bl2)
638 bl2->tail->bi_next = bl->head;
640 bl->tail = bl2->tail;
642 bl->head = bl2->head;
645 static inline struct bio *bio_list_peek(struct bio_list *bl)
650 static inline struct bio *bio_list_pop(struct bio_list *bl)
652 struct bio *bio = bl->head;
655 bl->head = bl->head->bi_next;
665 static inline struct bio *bio_list_get(struct bio_list *bl)
667 struct bio *bio = bl->head;
669 bl->head = bl->tail = NULL;
675 * Increment chain count for the bio. Make sure the CHAIN flag update
676 * is visible before the raised count.
678 static inline void bio_inc_remaining(struct bio *bio)
680 bio_set_flag(bio, BIO_CHAIN);
681 smp_mb__before_atomic();
682 atomic_inc(&bio->__bi_remaining);
686 * bio_set is used to allow other portions of the IO system to
687 * allocate their own private memory pools for bio and iovec structures.
688 * These memory pools in turn all allocate from the bio_slab
689 * and the bvec_slabs[].
691 #define BIO_POOL_SIZE 2
694 struct kmem_cache *bio_slab;
695 unsigned int front_pad;
699 #if defined(CONFIG_BLK_DEV_INTEGRITY)
700 mempool_t bio_integrity_pool;
701 mempool_t bvec_integrity_pool;
705 * Deadlock avoidance for stacking block drivers: see comments in
706 * bio_alloc_bioset() for details
708 spinlock_t rescue_lock;
709 struct bio_list rescue_list;
710 struct work_struct rescue_work;
711 struct workqueue_struct *rescue_workqueue;
717 struct kmem_cache *slab;
720 static inline bool bioset_initialized(struct bio_set *bs)
722 return bs->bio_slab != NULL;
726 * a small number of entries is fine, not going to be performance critical.
727 * basically we just need to survive
729 #define BIO_SPLIT_ENTRIES 2
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 /* CONFIG_BLOCK */
828 #endif /* __LINUX_BIO_H */