4 * Copyright (C) 2001 Jens Axboe <axboe@suse.de>
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public Licens
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
23 #include <linux/highmem.h>
24 #include <linux/mempool.h>
25 #include <linux/ioprio.h>
26 #include <linux/bug.h>
32 /* struct bio, bio_vec and BIO_* flags are defined in blk_types.h */
33 #include <linux/blk_types.h>
38 #define BIO_BUG_ON BUG_ON
43 #define BIO_MAX_PAGES 256
44 #define BIO_MAX_SIZE (BIO_MAX_PAGES << PAGE_CACHE_SHIFT)
45 #define BIO_MAX_SECTORS (BIO_MAX_SIZE >> 9)
48 * upper 16 bits of bi_rw define the io priority of this bio
50 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
51 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
52 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
54 #define bio_set_prio(bio, prio) do { \
55 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
56 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
57 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
61 * various member access, note that bio_data should of course not be used
62 * on highmem page vectors
64 #define __bvec_iter_bvec(bvec, iter) (&(bvec)[(iter).bi_idx])
66 #define bvec_iter_page(bvec, iter) \
67 (__bvec_iter_bvec((bvec), (iter))->bv_page)
69 #define bvec_iter_len(bvec, iter) \
71 __bvec_iter_bvec((bvec), (iter))->bv_len - (iter).bi_bvec_done)
73 #define bvec_iter_offset(bvec, iter) \
74 (__bvec_iter_bvec((bvec), (iter))->bv_offset + (iter).bi_bvec_done)
76 #define bvec_iter_bvec(bvec, iter) \
78 .bv_page = bvec_iter_page((bvec), (iter)), \
79 .bv_len = bvec_iter_len((bvec), (iter)), \
80 .bv_offset = bvec_iter_offset((bvec), (iter)), \
83 #define bio_iter_iovec(bio, iter) \
84 bvec_iter_bvec((bio)->bi_io_vec, (iter))
86 #define bio_iter_page(bio, iter) \
87 bvec_iter_page((bio)->bi_io_vec, (iter))
88 #define bio_iter_len(bio, iter) \
89 bvec_iter_len((bio)->bi_io_vec, (iter))
90 #define bio_iter_offset(bio, iter) \
91 bvec_iter_offset((bio)->bi_io_vec, (iter))
93 #define bio_page(bio) bio_iter_page((bio), (bio)->bi_iter)
94 #define bio_offset(bio) bio_iter_offset((bio), (bio)->bi_iter)
95 #define bio_iovec(bio) bio_iter_iovec((bio), (bio)->bi_iter)
97 #define bio_multiple_segments(bio) \
98 ((bio)->bi_iter.bi_size != bio_iovec(bio).bv_len)
99 #define bio_sectors(bio) ((bio)->bi_iter.bi_size >> 9)
100 #define bio_end_sector(bio) ((bio)->bi_iter.bi_sector + bio_sectors((bio)))
103 * Check whether this bio carries any data or not. A NULL bio is allowed.
105 static inline bool bio_has_data(struct bio *bio)
108 bio->bi_iter.bi_size &&
109 !(bio->bi_rw & REQ_DISCARD))
115 static inline bool bio_is_rw(struct bio *bio)
117 if (!bio_has_data(bio))
120 if (bio->bi_rw & BIO_NO_ADVANCE_ITER_MASK)
126 static inline bool bio_mergeable(struct bio *bio)
128 if (bio->bi_rw & REQ_NOMERGE_FLAGS)
134 static inline unsigned int bio_cur_bytes(struct bio *bio)
136 if (bio_has_data(bio))
137 return bio_iovec(bio).bv_len;
138 else /* dataless requests such as discard */
139 return bio->bi_iter.bi_size;
142 static inline void *bio_data(struct bio *bio)
144 if (bio_has_data(bio))
145 return page_address(bio_page(bio)) + bio_offset(bio);
153 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
154 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
157 * queues that have highmem support enabled may still need to revert to
158 * PIO transfers occasionally and thus map high pages temporarily. For
159 * permanent PIO fall back, user is probably better off disabling highmem
160 * I/O completely on that queue (see ide-dma for example)
162 #define __bio_kmap_atomic(bio, iter) \
163 (kmap_atomic(bio_iter_iovec((bio), (iter)).bv_page) + \
164 bio_iter_iovec((bio), (iter)).bv_offset)
166 #define __bio_kunmap_atomic(addr) kunmap_atomic(addr)
172 /* Default implementation of BIOVEC_PHYS_MERGEABLE */
173 #define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
174 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
177 * allow arch override, for eg virtualized architectures (put in asm/io.h)
179 #ifndef BIOVEC_PHYS_MERGEABLE
180 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
181 __BIOVEC_PHYS_MERGEABLE(vec1, vec2)
184 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
185 (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
186 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
187 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, queue_segment_boundary((q)))
190 * Check if adding a bio_vec after bprv with offset would create a gap in
191 * the SG list. Most drivers don't care about this, but some do.
193 static inline bool bvec_gap_to_prev(struct bio_vec *bprv, unsigned int offset)
195 return offset || ((bprv->bv_offset + bprv->bv_len) & (PAGE_SIZE - 1));
199 * drivers should _never_ use the all version - the bio may have been split
200 * before it got to the driver and the driver won't own all of it
202 #define bio_for_each_segment_all(bvl, bio, i) \
203 for (i = 0, bvl = (bio)->bi_io_vec; i < (bio)->bi_vcnt; i++, bvl++)
205 static inline void bvec_iter_advance(struct bio_vec *bv, struct bvec_iter *iter,
208 WARN_ONCE(bytes > iter->bi_size,
209 "Attempted to advance past end of bvec iter\n");
212 unsigned len = min(bytes, bvec_iter_len(bv, *iter));
215 iter->bi_size -= len;
216 iter->bi_bvec_done += len;
218 if (iter->bi_bvec_done == __bvec_iter_bvec(bv, *iter)->bv_len) {
219 iter->bi_bvec_done = 0;
225 #define for_each_bvec(bvl, bio_vec, iter, start) \
226 for (iter = (start); \
228 ((bvl = bvec_iter_bvec((bio_vec), (iter))), 1); \
229 bvec_iter_advance((bio_vec), &(iter), (bvl).bv_len))
232 static inline void bio_advance_iter(struct bio *bio, struct bvec_iter *iter,
235 iter->bi_sector += bytes >> 9;
237 if (bio->bi_rw & BIO_NO_ADVANCE_ITER_MASK)
238 iter->bi_size -= bytes;
240 bvec_iter_advance(bio->bi_io_vec, iter, bytes);
243 #define __bio_for_each_segment(bvl, bio, iter, start) \
244 for (iter = (start); \
246 ((bvl = bio_iter_iovec((bio), (iter))), 1); \
247 bio_advance_iter((bio), &(iter), (bvl).bv_len))
249 #define bio_for_each_segment(bvl, bio, iter) \
250 __bio_for_each_segment(bvl, bio, iter, (bio)->bi_iter)
252 #define bio_iter_last(bvec, iter) ((iter).bi_size == (bvec).bv_len)
254 static inline unsigned bio_segments(struct bio *bio)
258 struct bvec_iter iter;
261 * We special case discard/write same, because they interpret bi_size
265 if (bio->bi_rw & REQ_DISCARD)
268 if (bio->bi_rw & REQ_WRITE_SAME)
271 bio_for_each_segment(bv, bio, iter)
278 * get a reference to a bio, so it won't disappear. the intended use is
282 * submit_bio(rw, bio);
283 * if (bio->bi_flags ...)
287 * without the bio_get(), it could potentially complete I/O before submit_bio
288 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
291 static inline void bio_get(struct bio *bio)
293 bio->bi_flags |= (1 << BIO_REFFED);
294 smp_mb__before_atomic();
295 atomic_inc(&bio->__bi_cnt);
298 static inline void bio_cnt_set(struct bio *bio, unsigned int count)
301 bio->bi_flags |= (1 << BIO_REFFED);
302 smp_mb__before_atomic();
304 atomic_set(&bio->__bi_cnt, count);
307 static inline bool bio_flagged(struct bio *bio, unsigned int bit)
309 return (bio->bi_flags & (1UL << bit)) != 0;
312 static inline void bio_set_flag(struct bio *bio, unsigned int bit)
314 bio->bi_flags |= (1UL << bit);
317 static inline void bio_clear_flag(struct bio *bio, unsigned int bit)
319 bio->bi_flags &= ~(1UL << bit);
323 BIP_BLOCK_INTEGRITY = 1 << 0, /* block layer owns integrity data */
324 BIP_MAPPED_INTEGRITY = 1 << 1, /* ref tag has been remapped */
325 BIP_CTRL_NOCHECK = 1 << 2, /* disable HBA integrity checking */
326 BIP_DISK_NOCHECK = 1 << 3, /* disable disk integrity checking */
327 BIP_IP_CHECKSUM = 1 << 4, /* IP checksum */
330 #if defined(CONFIG_BLK_DEV_INTEGRITY)
332 static inline struct bio_integrity_payload *bio_integrity(struct bio *bio)
334 if (bio->bi_rw & REQ_INTEGRITY)
335 return bio->bi_integrity;
341 * bio integrity payload
343 struct bio_integrity_payload {
344 struct bio *bip_bio; /* parent bio */
346 struct bvec_iter bip_iter;
348 bio_end_io_t *bip_end_io; /* saved I/O completion fn */
350 unsigned short bip_slab; /* slab the bip came from */
351 unsigned short bip_vcnt; /* # of integrity bio_vecs */
352 unsigned short bip_max_vcnt; /* integrity bio_vec slots */
353 unsigned short bip_flags; /* control flags */
355 struct work_struct bip_work; /* I/O completion */
357 struct bio_vec *bip_vec;
358 struct bio_vec bip_inline_vecs[0];/* embedded bvec array */
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);
407 extern struct bio_set *bioset_create(unsigned int, unsigned int);
408 extern struct bio_set *bioset_create_nobvec(unsigned int, unsigned int);
409 extern void bioset_free(struct bio_set *);
410 extern mempool_t *biovec_create_pool(int pool_entries);
412 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
413 extern void bio_put(struct bio *);
415 extern void __bio_clone_fast(struct bio *, struct bio *);
416 extern struct bio *bio_clone_fast(struct bio *, gfp_t, struct bio_set *);
417 extern struct bio *bio_clone_bioset(struct bio *, gfp_t, struct bio_set *bs);
419 extern struct bio_set *fs_bio_set;
421 static inline struct bio *bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
423 return bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
426 static inline struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask)
428 return bio_clone_bioset(bio, gfp_mask, fs_bio_set);
431 static inline struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned int nr_iovecs)
433 return bio_alloc_bioset(gfp_mask, nr_iovecs, NULL);
436 static inline struct bio *bio_clone_kmalloc(struct bio *bio, gfp_t gfp_mask)
438 return bio_clone_bioset(bio, gfp_mask, NULL);
442 extern void bio_endio(struct bio *);
444 static inline void bio_io_error(struct bio *bio)
446 bio->bi_error = -EIO;
450 struct request_queue;
451 extern int bio_phys_segments(struct request_queue *, struct bio *);
453 extern int submit_bio_wait(int rw, struct bio *bio);
454 extern void bio_advance(struct bio *, unsigned);
456 extern void bio_init(struct bio *);
457 extern void bio_reset(struct bio *);
458 void bio_chain(struct bio *, struct bio *);
460 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
461 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
462 unsigned int, unsigned int);
463 extern int bio_get_nr_vecs(struct block_device *);
465 extern struct bio *bio_map_user_iov(struct request_queue *,
466 const struct iov_iter *, gfp_t);
467 extern void bio_unmap_user(struct bio *);
468 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
470 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
472 extern void bio_set_pages_dirty(struct bio *bio);
473 extern void bio_check_pages_dirty(struct bio *bio);
475 void generic_start_io_acct(int rw, unsigned long sectors,
476 struct hd_struct *part);
477 void generic_end_io_acct(int rw, struct hd_struct *part,
478 unsigned long start_time);
480 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
481 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform"
483 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
484 extern void bio_flush_dcache_pages(struct bio *bi);
486 static inline void bio_flush_dcache_pages(struct bio *bi)
491 extern void bio_copy_data(struct bio *dst, struct bio *src);
492 extern int bio_alloc_pages(struct bio *bio, gfp_t gfp);
494 extern struct bio *bio_copy_user_iov(struct request_queue *,
495 struct rq_map_data *,
496 const struct iov_iter *,
498 extern int bio_uncopy_user(struct bio *);
499 void zero_fill_bio(struct bio *bio);
500 extern struct bio_vec *bvec_alloc(gfp_t, int, unsigned long *, mempool_t *);
501 extern void bvec_free(mempool_t *, struct bio_vec *, unsigned int);
502 extern unsigned int bvec_nr_vecs(unsigned short idx);
504 #ifdef CONFIG_BLK_CGROUP
505 int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css);
506 int bio_associate_current(struct bio *bio);
507 void bio_disassociate_task(struct bio *bio);
508 #else /* CONFIG_BLK_CGROUP */
509 static inline int bio_associate_blkcg(struct bio *bio,
510 struct cgroup_subsys_state *blkcg_css) { return 0; }
511 static inline int bio_associate_current(struct bio *bio) { return -ENOENT; }
512 static inline void bio_disassociate_task(struct bio *bio) { }
513 #endif /* CONFIG_BLK_CGROUP */
515 #ifdef CONFIG_HIGHMEM
517 * remember never ever reenable interrupts between a bvec_kmap_irq and
520 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
525 * might not be a highmem page, but the preempt/irq count
526 * balancing is a lot nicer this way
528 local_irq_save(*flags);
529 addr = (unsigned long) kmap_atomic(bvec->bv_page);
531 BUG_ON(addr & ~PAGE_MASK);
533 return (char *) addr + bvec->bv_offset;
536 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
538 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
540 kunmap_atomic((void *) ptr);
541 local_irq_restore(*flags);
545 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
547 return page_address(bvec->bv_page) + bvec->bv_offset;
550 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
556 static inline char *__bio_kmap_irq(struct bio *bio, struct bvec_iter iter,
557 unsigned long *flags)
559 return bvec_kmap_irq(&bio_iter_iovec(bio, iter), flags);
561 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
563 #define bio_kmap_irq(bio, flags) \
564 __bio_kmap_irq((bio), (bio)->bi_iter, (flags))
565 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
568 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
570 * A bio_list anchors a singly-linked list of bios chained through the bi_next
571 * member of the bio. The bio_list also caches the last list member to allow
572 * fast access to the tail.
579 static inline int bio_list_empty(const struct bio_list *bl)
581 return bl->head == NULL;
584 static inline void bio_list_init(struct bio_list *bl)
586 bl->head = bl->tail = NULL;
589 #define BIO_EMPTY_LIST { NULL, NULL }
591 #define bio_list_for_each(bio, bl) \
592 for (bio = (bl)->head; bio; bio = bio->bi_next)
594 static inline unsigned bio_list_size(const struct bio_list *bl)
599 bio_list_for_each(bio, bl)
605 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
610 bl->tail->bi_next = bio;
617 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
619 bio->bi_next = bl->head;
627 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
633 bl->tail->bi_next = bl2->head;
635 bl->head = bl2->head;
637 bl->tail = bl2->tail;
640 static inline void bio_list_merge_head(struct bio_list *bl,
641 struct bio_list *bl2)
647 bl2->tail->bi_next = bl->head;
649 bl->tail = bl2->tail;
651 bl->head = bl2->head;
654 static inline struct bio *bio_list_peek(struct bio_list *bl)
659 static inline struct bio *bio_list_pop(struct bio_list *bl)
661 struct bio *bio = bl->head;
664 bl->head = bl->head->bi_next;
674 static inline struct bio *bio_list_get(struct bio_list *bl)
676 struct bio *bio = bl->head;
678 bl->head = bl->tail = NULL;
684 * bio_set is used to allow other portions of the IO system to
685 * allocate their own private memory pools for bio and iovec structures.
686 * These memory pools in turn all allocate from the bio_slab
687 * and the bvec_slabs[].
689 #define BIO_POOL_SIZE 2
690 #define BIOVEC_NR_POOLS 6
691 #define BIOVEC_MAX_IDX (BIOVEC_NR_POOLS - 1)
694 struct kmem_cache *bio_slab;
695 unsigned int front_pad;
698 mempool_t *bvec_pool;
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;
721 * a small number of entries is fine, not going to be performance critical.
722 * basically we just need to survive
724 #define BIO_SPLIT_ENTRIES 2
726 #if defined(CONFIG_BLK_DEV_INTEGRITY)
728 #define bip_for_each_vec(bvl, bip, iter) \
729 for_each_bvec(bvl, (bip)->bip_vec, iter, (bip)->bip_iter)
731 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \
733 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter)
735 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
736 extern void bio_integrity_free(struct bio *);
737 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
738 extern bool bio_integrity_enabled(struct bio *bio);
739 extern int bio_integrity_prep(struct bio *);
740 extern void bio_integrity_endio(struct bio *);
741 extern void bio_integrity_advance(struct bio *, unsigned int);
742 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int);
743 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
744 extern int bioset_integrity_create(struct bio_set *, int);
745 extern void bioset_integrity_free(struct bio_set *);
746 extern void bio_integrity_init(void);
748 #else /* CONFIG_BLK_DEV_INTEGRITY */
750 static inline void *bio_integrity(struct bio *bio)
755 static inline bool bio_integrity_enabled(struct bio *bio)
760 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size)
765 static inline void bioset_integrity_free (struct bio_set *bs)
770 static inline int bio_integrity_prep(struct bio *bio)
775 static inline void bio_integrity_free(struct bio *bio)
780 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src,
786 static inline void bio_integrity_advance(struct bio *bio,
787 unsigned int bytes_done)
792 static inline void bio_integrity_trim(struct bio *bio, unsigned int offset,
793 unsigned int sectors)
798 static inline void bio_integrity_init(void)
803 static inline bool bio_integrity_flagged(struct bio *bio, enum bip_flags flag)
808 #endif /* CONFIG_BLK_DEV_INTEGRITY */
810 #endif /* CONFIG_BLOCK */
811 #endif /* __LINUX_BIO_H */