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_OFFSET (BITS_PER_LONG - BIO_POOL_BITS)
136 #define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET)
137 #define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET)
142 * bit 0 -- data direction
143 * If not set, bio is a read from device. If set, it's a write to device.
144 * bit 1 -- rw-ahead when set
146 * Insert a serialization point in the IO queue, forcing previously
147 * submitted IO to be completed before this one is issued.
148 * bit 3 -- synchronous I/O hint.
149 * bit 4 -- Unplug the device immediately after submitting this bio.
150 * bit 5 -- metadata request
151 * Used for tracing to differentiate metadata and data IO. May also
152 * get some preferential treatment in the IO scheduler
153 * bit 6 -- discard sectors
154 * Informs the lower level device that this range of sectors is no longer
155 * used by the file system and may thus be freed by the device. Used
156 * for flash based storage.
157 * bit 7 -- fail fast device errors
158 * bit 8 -- fail fast transport errors
159 * bit 9 -- fail fast driver errors
160 * Don't want driver retries for any fast fail whatever the reason.
161 * bit 10 -- Tell the IO scheduler not to wait for more requests after this
162 one has been submitted, even if it is a SYNC request.
164 #define BIO_RW 0 /* Must match RW in req flags (blkdev.h) */
165 #define BIO_RW_AHEAD 1 /* Must match FAILFAST in req flags */
166 #define BIO_RW_BARRIER 2
167 #define BIO_RW_SYNCIO 3
168 #define BIO_RW_UNPLUG 4
169 #define BIO_RW_META 5
170 #define BIO_RW_DISCARD 6
171 #define BIO_RW_FAILFAST_DEV 7
172 #define BIO_RW_FAILFAST_TRANSPORT 8
173 #define BIO_RW_FAILFAST_DRIVER 9
174 #define BIO_RW_NOIDLE 10
176 #define bio_rw_flagged(bio, flag) ((bio)->bi_rw & (1 << (flag)))
179 * Old defines, these should eventually be replaced by direct usage of
182 #define bio_barrier(bio) bio_rw_flagged(bio, BIO_RW_BARRIER)
183 #define bio_sync(bio) bio_rw_flagged(bio, BIO_RW_SYNCIO)
184 #define bio_unplug(bio) bio_rw_flagged(bio, BIO_RW_UNPLUG)
185 #define bio_failfast_dev(bio) bio_rw_flagged(bio, BIO_RW_FAILFAST_DEV)
186 #define bio_failfast_transport(bio) \
187 bio_rw_flagged(bio, BIO_RW_FAILFAST_TRANSPORT)
188 #define bio_failfast_driver(bio) \
189 bio_rw_flagged(bio, BIO_RW_FAILFAST_DRIVER)
190 #define bio_rw_ahead(bio) bio_rw_flagged(bio, BIO_RW_AHEAD)
191 #define bio_rw_meta(bio) bio_rw_flagged(bio, BIO_RW_META)
192 #define bio_discard(bio) bio_rw_flagged(bio, BIO_RW_DISCARD)
193 #define bio_noidle(bio) bio_rw_flagged(bio, BIO_RW_NOIDLE)
196 * upper 16 bits of bi_rw define the io priority of this bio
198 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
199 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
200 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
202 #define bio_set_prio(bio, prio) do { \
203 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
204 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
205 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
209 * various member access, note that bio_data should of course not be used
210 * on highmem page vectors
212 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
213 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx)
214 #define bio_page(bio) bio_iovec((bio))->bv_page
215 #define bio_offset(bio) bio_iovec((bio))->bv_offset
216 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx)
217 #define bio_sectors(bio) ((bio)->bi_size >> 9)
218 #define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio) && !bio_discard(bio))
220 static inline unsigned int bio_cur_sectors(struct bio *bio)
223 return bio_iovec(bio)->bv_len >> 9;
224 else /* dataless requests such as discard */
225 return bio->bi_size >> 9;
228 static inline void *bio_data(struct bio *bio)
231 return page_address(bio_page(bio)) + bio_offset(bio);
236 static inline int bio_has_allocated_vec(struct bio *bio)
238 return bio->bi_io_vec && bio->bi_io_vec != bio->bi_inline_vecs;
244 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
245 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
248 * queues that have highmem support enabled may still need to revert to
249 * PIO transfers occasionally and thus map high pages temporarily. For
250 * permanent PIO fall back, user is probably better off disabling highmem
251 * I/O completely on that queue (see ide-dma for example)
253 #define __bio_kmap_atomic(bio, idx, kmtype) \
254 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \
255 bio_iovec_idx((bio), (idx))->bv_offset)
257 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
263 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
264 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx)
266 /* Default implementation of BIOVEC_PHYS_MERGEABLE */
267 #define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
268 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
271 * allow arch override, for eg virtualized architectures (put in asm/io.h)
273 #ifndef BIOVEC_PHYS_MERGEABLE
274 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
275 __BIOVEC_PHYS_MERGEABLE(vec1, vec2)
278 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
279 (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
280 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
281 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask)
282 #define BIO_SEG_BOUNDARY(q, b1, b2) \
283 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
285 #define bio_io_error(bio) bio_endio((bio), -EIO)
288 * drivers should not use the __ version unless they _really_ want to
289 * run through the entire bio and not just pending pieces
291 #define __bio_for_each_segment(bvl, bio, i, start_idx) \
292 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \
293 i < (bio)->bi_vcnt; \
296 #define bio_for_each_segment(bvl, bio, i) \
297 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
300 * get a reference to a bio, so it won't disappear. the intended use is
304 * submit_bio(rw, bio);
305 * if (bio->bi_flags ...)
309 * without the bio_get(), it could potentially complete I/O before submit_bio
310 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
313 #define bio_get(bio) atomic_inc(&(bio)->bi_cnt)
315 #if defined(CONFIG_BLK_DEV_INTEGRITY)
317 * bio integrity payload
319 struct bio_integrity_payload {
320 struct bio *bip_bio; /* parent bio */
321 struct bio_vec *bip_vec; /* integrity data vector */
323 sector_t bip_sector; /* virtual start sector */
325 void *bip_buf; /* generated integrity data */
326 bio_end_io_t *bip_end_io; /* saved I/O completion fn */
328 unsigned int bip_size;
330 unsigned short bip_pool; /* pool the ivec came from */
331 unsigned short bip_vcnt; /* # of integrity bio_vecs */
332 unsigned short bip_idx; /* current bip_vec index */
334 struct work_struct bip_work; /* I/O completion */
336 #endif /* CONFIG_BLK_DEV_INTEGRITY */
339 * A bio_pair is used when we need to split a bio.
340 * This can only happen for a bio that refers to just one
341 * page of data, and in the unusual situation when the
342 * page crosses a chunk/device boundary
344 * The address of the master bio is stored in bio1.bi_private
345 * The address of the pool the pair was allocated from is stored
349 struct bio bio1, bio2;
350 struct bio_vec bv1, bv2;
351 #if defined(CONFIG_BLK_DEV_INTEGRITY)
352 struct bio_integrity_payload bip1, bip2;
353 struct bio_vec iv1, iv2;
358 extern struct bio_pair *bio_split(struct bio *bi, int first_sectors);
359 extern void bio_pair_release(struct bio_pair *dbio);
361 extern struct bio_set *bioset_create(unsigned int, unsigned int);
362 extern void bioset_free(struct bio_set *);
364 extern struct bio *bio_alloc(gfp_t, int);
365 extern struct bio *bio_kmalloc(gfp_t, int);
366 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
367 extern void bio_put(struct bio *);
368 extern void bio_free(struct bio *, struct bio_set *);
370 extern void bio_endio(struct bio *, int);
371 struct request_queue;
372 extern int bio_phys_segments(struct request_queue *, struct bio *);
374 extern void __bio_clone(struct bio *, struct bio *);
375 extern struct bio *bio_clone(struct bio *, gfp_t);
377 extern void bio_init(struct bio *);
379 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
380 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
381 unsigned int, unsigned int);
382 extern int bio_get_nr_vecs(struct block_device *);
383 extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int);
384 extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
385 unsigned long, unsigned int, int, gfp_t);
388 extern struct bio *bio_map_user_iov(struct request_queue *,
389 struct block_device *,
390 struct sg_iovec *, int, int, gfp_t);
391 extern void bio_unmap_user(struct bio *);
392 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
394 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
396 extern void bio_set_pages_dirty(struct bio *bio);
397 extern void bio_check_pages_dirty(struct bio *bio);
398 extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *,
399 unsigned long, unsigned int, int, gfp_t);
400 extern struct bio *bio_copy_user_iov(struct request_queue *,
401 struct rq_map_data *, struct sg_iovec *,
403 extern int bio_uncopy_user(struct bio *);
404 void zero_fill_bio(struct bio *bio);
405 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *);
406 extern void bvec_free_bs(struct bio_set *, struct bio_vec *, unsigned int);
407 extern unsigned int bvec_nr_vecs(unsigned short idx);
410 * Allow queuer to specify a completion CPU for this bio
412 static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu)
414 bio->bi_comp_cpu = cpu;
418 * bio_set is used to allow other portions of the IO system to
419 * allocate their own private memory pools for bio and iovec structures.
420 * These memory pools in turn all allocate from the bio_slab
421 * and the bvec_slabs[].
423 #define BIO_POOL_SIZE 2
424 #define BIOVEC_NR_POOLS 6
425 #define BIOVEC_MAX_IDX (BIOVEC_NR_POOLS - 1)
428 struct kmem_cache *bio_slab;
429 unsigned int front_pad;
432 mempool_t *bvec_pool;
438 struct kmem_cache *slab;
441 extern struct bio_set *fs_bio_set;
442 extern struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly;
445 * a small number of entries is fine, not going to be performance critical.
446 * basically we just need to survive
448 #define BIO_SPLIT_ENTRIES 2
450 #ifdef CONFIG_HIGHMEM
452 * remember never ever reenable interrupts between a bvec_kmap_irq and
455 * This function MUST be inlined - it plays with the CPU interrupt flags.
457 static __always_inline char *bvec_kmap_irq(struct bio_vec *bvec,
458 unsigned long *flags)
463 * might not be a highmem page, but the preempt/irq count
464 * balancing is a lot nicer this way
466 local_irq_save(*flags);
467 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
469 BUG_ON(addr & ~PAGE_MASK);
471 return (char *) addr + bvec->bv_offset;
474 static __always_inline void bvec_kunmap_irq(char *buffer,
475 unsigned long *flags)
477 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
479 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
480 local_irq_restore(*flags);
484 #define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset)
485 #define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0)
488 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
489 unsigned long *flags)
491 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
493 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
495 #define bio_kmap_irq(bio, flags) \
496 __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
497 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
500 * Check whether this bio carries any data or not. A NULL bio is allowed.
502 static inline int bio_has_data(struct bio *bio)
504 return bio && bio->bi_io_vec != NULL;
507 #if defined(CONFIG_BLK_DEV_INTEGRITY)
509 #define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)]))
510 #define bip_vec(bip) bip_vec_idx(bip, 0)
512 #define __bip_for_each_vec(bvl, bip, i, start_idx) \
513 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \
514 i < (bip)->bip_vcnt; \
517 #define bip_for_each_vec(bvl, bip, i) \
518 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx)
520 #define bio_integrity(bio) (bio->bi_integrity != NULL)
522 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
523 extern void bio_integrity_free(struct bio *);
524 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
525 extern int bio_integrity_enabled(struct bio *bio);
526 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int);
527 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int);
528 extern int bio_integrity_prep(struct bio *);
529 extern void bio_integrity_endio(struct bio *, int);
530 extern void bio_integrity_advance(struct bio *, unsigned int);
531 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int);
532 extern void bio_integrity_split(struct bio *, struct bio_pair *, int);
533 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t);
535 #else /* CONFIG_BLK_DEV_INTEGRITY */
537 #define bio_integrity(a) (0)
538 #define bio_integrity_prep(a) (0)
539 #define bio_integrity_enabled(a) (0)
540 #define bio_integrity_clone(a, b, c) (0)
541 #define bio_integrity_free(a) do { } while (0)
542 #define bio_integrity_endio(a, b) do { } while (0)
543 #define bio_integrity_advance(a, b) do { } while (0)
544 #define bio_integrity_trim(a, b, c) do { } while (0)
545 #define bio_integrity_split(a, b, c) do { } while (0)
546 #define bio_integrity_set_tag(a, b, c) do { } while (0)
547 #define bio_integrity_get_tag(a, b, c) do { } while (0)
549 #endif /* CONFIG_BLK_DEV_INTEGRITY */
551 #endif /* CONFIG_BLOCK */
552 #endif /* __LINUX_BIO_H */