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>
34 #define BIO_BUG_ON BUG_ON
39 #define BIO_MAX_PAGES 256
40 #define BIO_MAX_SIZE (BIO_MAX_PAGES << PAGE_CACHE_SHIFT)
41 #define BIO_MAX_SECTORS (BIO_MAX_SIZE >> 9)
44 * was unsigned short, but we might as well be ready for > 64kB I/O pages
49 unsigned int bv_offset;
54 struct bio_integrity_payload;
55 typedef void (bio_end_io_t) (struct bio *, int);
56 typedef void (bio_destructor_t) (struct bio *);
59 * main unit of I/O for the block layer and lower layers (ie drivers and
63 sector_t bi_sector; /* device address in 512 byte
65 struct bio *bi_next; /* request queue link */
66 struct block_device *bi_bdev;
67 unsigned long bi_flags; /* status, command, etc */
68 unsigned long bi_rw; /* bottom bits READ/WRITE,
72 unsigned short bi_vcnt; /* how many bio_vec's */
73 unsigned short bi_idx; /* current index into bvl_vec */
75 /* Number of segments in this BIO after
76 * physical address coalescing is performed.
78 unsigned int bi_phys_segments;
80 unsigned int bi_size; /* residual I/O count */
83 * To keep track of the max segment size, we account for the
84 * sizes of the first and last mergeable segments in this bio.
86 unsigned int bi_seg_front_size;
87 unsigned int bi_seg_back_size;
89 unsigned int bi_max_vecs; /* max bvl_vecs we can hold */
91 unsigned int bi_comp_cpu; /* completion CPU */
93 atomic_t bi_cnt; /* pin count */
95 struct bio_vec *bi_io_vec; /* the actual vec list */
97 bio_end_io_t *bi_end_io;
100 #if defined(CONFIG_BLK_DEV_INTEGRITY)
101 struct bio_integrity_payload *bi_integrity; /* data integrity */
104 bio_destructor_t *bi_destructor; /* destructor */
107 * We can inline a number of vecs at the end of the bio, to avoid
108 * double allocations for a small number of bio_vecs. This member
109 * MUST obviously be kept at the very end of the bio.
111 struct bio_vec bi_inline_vecs[0];
117 #define BIO_UPTODATE 0 /* ok after I/O completion */
118 #define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */
119 #define BIO_EOF 2 /* out-out-bounds error */
120 #define BIO_SEG_VALID 3 /* bi_phys_segments valid */
121 #define BIO_CLONED 4 /* doesn't own data */
122 #define BIO_BOUNCED 5 /* bio is a bounce bio */
123 #define BIO_USER_MAPPED 6 /* contains user pages */
124 #define BIO_EOPNOTSUPP 7 /* not supported */
125 #define BIO_CPU_AFFINE 8 /* complete bio on same CPU as submitted */
126 #define BIO_NULL_MAPPED 9 /* contains invalid user pages */
127 #define BIO_FS_INTEGRITY 10 /* fs owns integrity data, not block layer */
128 #define BIO_QUIET 11 /* Make BIO Quiet */
129 #define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag)))
132 * top 4 bits of bio flags indicate the pool this bio came from
134 #define BIO_POOL_BITS (4)
135 #define BIO_POOL_NONE ((1UL << BIO_POOL_BITS) - 1)
136 #define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS)
137 #define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET)
138 #define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET)
143 * bit 0 -- data direction
144 * If not set, bio is a read from device. If set, it's a write to device.
145 * bit 1 -- fail fast device errors
146 * bit 2 -- fail fast transport errors
147 * bit 3 -- fail fast driver errors
148 * bit 4 -- rw-ahead when set
150 * Insert a serialization point in the IO queue, forcing previously
151 * submitted IO to be completed before this one is issued.
152 * bit 6 -- synchronous I/O hint.
153 * bit 7 -- Unplug the device immediately after submitting this bio.
154 * bit 8 -- metadata request
155 * Used for tracing to differentiate metadata and data IO. May also
156 * get some preferential treatment in the IO scheduler
157 * bit 9 -- discard sectors
158 * Informs the lower level device that this range of sectors is no longer
159 * used by the file system and may thus be freed by the device. Used
160 * for flash based storage.
161 * Don't want driver retries for any fast fail whatever the reason.
162 * bit 10 -- Tell the IO scheduler not to wait for more requests after this
163 one has been submitted, even if it is a SYNC request.
168 BIO_RW_FAILFAST_TRANSPORT,
169 BIO_RW_FAILFAST_DRIVER,
170 /* above flags must match REQ_* */
180 #define bio_rw_flagged(bio, flag) ((bio)->bi_rw & (1 << (flag)))
183 * Old defines, these should eventually be replaced by direct usage of
186 #define bio_barrier(bio) bio_rw_flagged(bio, BIO_RW_BARRIER)
187 #define bio_sync(bio) bio_rw_flagged(bio, BIO_RW_SYNCIO)
188 #define bio_unplug(bio) bio_rw_flagged(bio, BIO_RW_UNPLUG)
189 #define bio_failfast_dev(bio) bio_rw_flagged(bio, BIO_RW_FAILFAST_DEV)
190 #define bio_failfast_transport(bio) \
191 bio_rw_flagged(bio, BIO_RW_FAILFAST_TRANSPORT)
192 #define bio_failfast_driver(bio) \
193 bio_rw_flagged(bio, BIO_RW_FAILFAST_DRIVER)
194 #define bio_rw_ahead(bio) bio_rw_flagged(bio, BIO_RW_AHEAD)
195 #define bio_rw_meta(bio) bio_rw_flagged(bio, BIO_RW_META)
196 #define bio_discard(bio) bio_rw_flagged(bio, BIO_RW_DISCARD)
197 #define bio_noidle(bio) bio_rw_flagged(bio, BIO_RW_NOIDLE)
200 * upper 16 bits of bi_rw define the io priority of this bio
202 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
203 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
204 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
206 #define bio_set_prio(bio, prio) do { \
207 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
208 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
209 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
213 * various member access, note that bio_data should of course not be used
214 * on highmem page vectors
216 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
217 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx)
218 #define bio_page(bio) bio_iovec((bio))->bv_page
219 #define bio_offset(bio) bio_iovec((bio))->bv_offset
220 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx)
221 #define bio_sectors(bio) ((bio)->bi_size >> 9)
222 #define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio) && !bio_discard(bio))
224 static inline unsigned int bio_cur_bytes(struct bio *bio)
227 return bio_iovec(bio)->bv_len;
228 else /* dataless requests such as discard */
232 static inline void *bio_data(struct bio *bio)
235 return page_address(bio_page(bio)) + bio_offset(bio);
240 static inline int bio_has_allocated_vec(struct bio *bio)
242 return bio->bi_io_vec && bio->bi_io_vec != bio->bi_inline_vecs;
248 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
249 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
252 * queues that have highmem support enabled may still need to revert to
253 * PIO transfers occasionally and thus map high pages temporarily. For
254 * permanent PIO fall back, user is probably better off disabling highmem
255 * I/O completely on that queue (see ide-dma for example)
257 #define __bio_kmap_atomic(bio, idx, kmtype) \
258 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \
259 bio_iovec_idx((bio), (idx))->bv_offset)
261 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
267 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
268 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx)
270 /* Default implementation of BIOVEC_PHYS_MERGEABLE */
271 #define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
272 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
275 * allow arch override, for eg virtualized architectures (put in asm/io.h)
277 #ifndef BIOVEC_PHYS_MERGEABLE
278 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
279 __BIOVEC_PHYS_MERGEABLE(vec1, vec2)
282 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
283 (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
284 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
285 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, queue_segment_boundary((q)))
286 #define BIO_SEG_BOUNDARY(q, b1, b2) \
287 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
289 #define bio_io_error(bio) bio_endio((bio), -EIO)
292 * drivers should not use the __ version unless they _really_ want to
293 * run through the entire bio and not just pending pieces
295 #define __bio_for_each_segment(bvl, bio, i, start_idx) \
296 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \
297 i < (bio)->bi_vcnt; \
300 #define bio_for_each_segment(bvl, bio, i) \
301 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
304 * get a reference to a bio, so it won't disappear. the intended use is
308 * submit_bio(rw, bio);
309 * if (bio->bi_flags ...)
313 * without the bio_get(), it could potentially complete I/O before submit_bio
314 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
317 #define bio_get(bio) atomic_inc(&(bio)->bi_cnt)
319 #if defined(CONFIG_BLK_DEV_INTEGRITY)
321 * bio integrity payload
323 struct bio_integrity_payload {
324 struct bio *bip_bio; /* parent bio */
326 sector_t bip_sector; /* virtual start sector */
328 void *bip_buf; /* generated integrity data */
329 bio_end_io_t *bip_end_io; /* saved I/O completion fn */
331 unsigned int bip_size;
333 unsigned short bip_slab; /* slab the bip came from */
334 unsigned short bip_vcnt; /* # of integrity bio_vecs */
335 unsigned short bip_idx; /* current bip_vec index */
337 struct work_struct bip_work; /* I/O completion */
338 struct bio_vec bip_vec[0]; /* embedded bvec array */
340 #endif /* CONFIG_BLK_DEV_INTEGRITY */
343 * A bio_pair is used when we need to split a bio.
344 * This can only happen for a bio that refers to just one
345 * page of data, and in the unusual situation when the
346 * page crosses a chunk/device boundary
348 * The address of the master bio is stored in bio1.bi_private
349 * The address of the pool the pair was allocated from is stored
353 struct bio bio1, bio2;
354 struct bio_vec bv1, bv2;
355 #if defined(CONFIG_BLK_DEV_INTEGRITY)
356 struct bio_integrity_payload bip1, bip2;
357 struct bio_vec iv1, iv2;
362 extern struct bio_pair *bio_split(struct bio *bi, int first_sectors);
363 extern void bio_pair_release(struct bio_pair *dbio);
365 extern struct bio_set *bioset_create(unsigned int, unsigned int);
366 extern void bioset_free(struct bio_set *);
368 extern struct bio *bio_alloc(gfp_t, int);
369 extern struct bio *bio_kmalloc(gfp_t, int);
370 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
371 extern void bio_put(struct bio *);
372 extern void bio_free(struct bio *, struct bio_set *);
374 extern void bio_endio(struct bio *, int);
375 struct request_queue;
376 extern int bio_phys_segments(struct request_queue *, struct bio *);
378 extern void __bio_clone(struct bio *, struct bio *);
379 extern struct bio *bio_clone(struct bio *, gfp_t);
381 extern void bio_init(struct bio *);
383 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
384 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
385 unsigned int, unsigned int);
386 extern int bio_get_nr_vecs(struct block_device *);
387 extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int);
388 extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
389 unsigned long, unsigned int, int, gfp_t);
392 extern struct bio *bio_map_user_iov(struct request_queue *,
393 struct block_device *,
394 struct sg_iovec *, int, int, gfp_t);
395 extern void bio_unmap_user(struct bio *);
396 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
398 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
400 extern void bio_set_pages_dirty(struct bio *bio);
401 extern void bio_check_pages_dirty(struct bio *bio);
402 extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *,
403 unsigned long, unsigned int, int, gfp_t);
404 extern struct bio *bio_copy_user_iov(struct request_queue *,
405 struct rq_map_data *, struct sg_iovec *,
407 extern int bio_uncopy_user(struct bio *);
408 void zero_fill_bio(struct bio *bio);
409 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *);
410 extern void bvec_free_bs(struct bio_set *, struct bio_vec *, unsigned int);
411 extern unsigned int bvec_nr_vecs(unsigned short idx);
414 * Allow queuer to specify a completion CPU for this bio
416 static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu)
418 bio->bi_comp_cpu = cpu;
422 * bio_set is used to allow other portions of the IO system to
423 * allocate their own private memory pools for bio and iovec structures.
424 * These memory pools in turn all allocate from the bio_slab
425 * and the bvec_slabs[].
427 #define BIO_POOL_SIZE 2
428 #define BIOVEC_NR_POOLS 6
429 #define BIOVEC_MAX_IDX (BIOVEC_NR_POOLS - 1)
432 struct kmem_cache *bio_slab;
433 unsigned int front_pad;
436 #if defined(CONFIG_BLK_DEV_INTEGRITY)
437 mempool_t *bio_integrity_pool;
439 mempool_t *bvec_pool;
445 struct kmem_cache *slab;
448 extern struct bio_set *fs_bio_set;
449 extern struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly;
452 * a small number of entries is fine, not going to be performance critical.
453 * basically we just need to survive
455 #define BIO_SPLIT_ENTRIES 2
457 #ifdef CONFIG_HIGHMEM
459 * remember never ever reenable interrupts between a bvec_kmap_irq and
462 * This function MUST be inlined - it plays with the CPU interrupt flags.
464 static __always_inline char *bvec_kmap_irq(struct bio_vec *bvec,
465 unsigned long *flags)
470 * might not be a highmem page, but the preempt/irq count
471 * balancing is a lot nicer this way
473 local_irq_save(*flags);
474 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
476 BUG_ON(addr & ~PAGE_MASK);
478 return (char *) addr + bvec->bv_offset;
481 static __always_inline void bvec_kunmap_irq(char *buffer,
482 unsigned long *flags)
484 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
486 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
487 local_irq_restore(*flags);
491 #define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset)
492 #define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0)
495 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
496 unsigned long *flags)
498 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
500 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
502 #define bio_kmap_irq(bio, flags) \
503 __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
504 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
507 * Check whether this bio carries any data or not. A NULL bio is allowed.
509 static inline int bio_has_data(struct bio *bio)
511 return bio && bio->bi_io_vec != NULL;
515 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop.
517 * A bio_list anchors a singly-linked list of bios chained through the bi_next
518 * member of the bio. The bio_list also caches the last list member to allow
519 * fast access to the tail.
526 static inline int bio_list_empty(const struct bio_list *bl)
528 return bl->head == NULL;
531 static inline void bio_list_init(struct bio_list *bl)
533 bl->head = bl->tail = NULL;
536 #define bio_list_for_each(bio, bl) \
537 for (bio = (bl)->head; bio; bio = bio->bi_next)
539 static inline unsigned bio_list_size(const struct bio_list *bl)
544 bio_list_for_each(bio, bl)
550 static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
555 bl->tail->bi_next = bio;
562 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio)
564 bio->bi_next = bl->head;
572 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
578 bl->tail->bi_next = bl2->head;
580 bl->head = bl2->head;
582 bl->tail = bl2->tail;
585 static inline void bio_list_merge_head(struct bio_list *bl,
586 struct bio_list *bl2)
592 bl2->tail->bi_next = bl->head;
594 bl->tail = bl2->tail;
596 bl->head = bl2->head;
599 static inline struct bio *bio_list_peek(struct bio_list *bl)
604 static inline struct bio *bio_list_pop(struct bio_list *bl)
606 struct bio *bio = bl->head;
609 bl->head = bl->head->bi_next;
619 static inline struct bio *bio_list_get(struct bio_list *bl)
621 struct bio *bio = bl->head;
623 bl->head = bl->tail = NULL;
628 #if defined(CONFIG_BLK_DEV_INTEGRITY)
630 #define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)]))
631 #define bip_vec(bip) bip_vec_idx(bip, 0)
633 #define __bip_for_each_vec(bvl, bip, i, start_idx) \
634 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \
635 i < (bip)->bip_vcnt; \
638 #define bip_for_each_vec(bvl, bip, i) \
639 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx)
641 #define bio_integrity(bio) (bio->bi_integrity != NULL)
643 extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *);
644 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
645 extern void bio_integrity_free(struct bio *, struct bio_set *);
646 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
647 extern int bio_integrity_enabled(struct bio *bio);
648 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int);
649 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int);
650 extern int bio_integrity_prep(struct bio *);
651 extern void bio_integrity_endio(struct bio *, int);
652 extern void bio_integrity_advance(struct bio *, unsigned int);
653 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int);
654 extern void bio_integrity_split(struct bio *, struct bio_pair *, int);
655 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t, struct bio_set *);
656 extern int bioset_integrity_create(struct bio_set *, int);
657 extern void bioset_integrity_free(struct bio_set *);
658 extern void bio_integrity_init(void);
660 #else /* CONFIG_BLK_DEV_INTEGRITY */
662 #define bio_integrity(a) (0)
663 #define bioset_integrity_create(a, b) (0)
664 #define bio_integrity_prep(a) (0)
665 #define bio_integrity_enabled(a) (0)
666 #define bio_integrity_clone(a, b, c, d) (0)
667 #define bioset_integrity_free(a) do { } while (0)
668 #define bio_integrity_free(a, b) do { } while (0)
669 #define bio_integrity_endio(a, b) do { } while (0)
670 #define bio_integrity_advance(a, b) do { } while (0)
671 #define bio_integrity_trim(a, b, c) do { } while (0)
672 #define bio_integrity_split(a, b, c) do { } while (0)
673 #define bio_integrity_set_tag(a, b, c) do { } while (0)
674 #define bio_integrity_get_tag(a, b, c) do { } while (0)
675 #define bio_integrity_init(a) do { } while (0)
677 #endif /* CONFIG_BLK_DEV_INTEGRITY */
679 #endif /* CONFIG_BLOCK */
680 #endif /* __LINUX_BIO_H */