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 */
82 unsigned int bi_max_vecs; /* max bvl_vecs we can hold */
84 unsigned int bi_comp_cpu; /* completion CPU */
86 struct bio_vec *bi_io_vec; /* the actual vec list */
88 bio_end_io_t *bi_end_io;
89 atomic_t bi_cnt; /* pin count */
92 #if defined(CONFIG_BLK_DEV_INTEGRITY)
93 struct bio_integrity_payload *bi_integrity; /* data integrity */
96 bio_destructor_t *bi_destructor; /* destructor */
102 #define BIO_UPTODATE 0 /* ok after I/O completion */
103 #define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */
104 #define BIO_EOF 2 /* out-out-bounds error */
105 #define BIO_SEG_VALID 3 /* bi_phys_segments valid */
106 #define BIO_CLONED 4 /* doesn't own data */
107 #define BIO_BOUNCED 5 /* bio is a bounce bio */
108 #define BIO_USER_MAPPED 6 /* contains user pages */
109 #define BIO_EOPNOTSUPP 7 /* not supported */
110 #define BIO_CPU_AFFINE 8 /* complete bio on same CPU as submitted */
111 #define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag)))
114 * top 4 bits of bio flags indicate the pool this bio came from
116 #define BIO_POOL_BITS (4)
117 #define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS)
118 #define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET)
119 #define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET)
124 * bit 0 -- read (not set) or write (set)
125 * bit 1 -- rw-ahead when set
127 * bit 3 -- fail fast, don't want low level driver retries
128 * bit 4 -- synchronous I/O hint: the block layer will unplug immediately
129 * bit 5 -- metadata request
130 * bit 6 -- discard sectors
132 #define BIO_RW 0 /* Must match RW in req flags (blkdev.h) */
133 #define BIO_RW_AHEAD 1 /* Must match FAILFAST in req flags */
134 #define BIO_RW_BARRIER 2
135 #define BIO_RW_FAILFAST 3
136 #define BIO_RW_SYNC 4
137 #define BIO_RW_META 5
138 #define BIO_RW_DISCARD 6
141 * upper 16 bits of bi_rw define the io priority of this bio
143 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
144 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
145 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
147 #define bio_set_prio(bio, prio) do { \
148 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
149 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
150 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
154 * various member access, note that bio_data should of course not be used
155 * on highmem page vectors
157 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
158 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx)
159 #define bio_page(bio) bio_iovec((bio))->bv_page
160 #define bio_offset(bio) bio_iovec((bio))->bv_offset
161 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx)
162 #define bio_sectors(bio) ((bio)->bi_size >> 9)
163 #define bio_barrier(bio) ((bio)->bi_rw & (1 << BIO_RW_BARRIER))
164 #define bio_sync(bio) ((bio)->bi_rw & (1 << BIO_RW_SYNC))
165 #define bio_failfast(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST))
166 #define bio_rw_ahead(bio) ((bio)->bi_rw & (1 << BIO_RW_AHEAD))
167 #define bio_rw_meta(bio) ((bio)->bi_rw & (1 << BIO_RW_META))
168 #define bio_discard(bio) ((bio)->bi_rw & (1 << BIO_RW_DISCARD))
169 #define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio) && !bio_discard(bio))
171 static inline unsigned int bio_cur_sectors(struct bio *bio)
174 return bio_iovec(bio)->bv_len >> 9;
175 else /* dataless requests such as discard */
176 return bio->bi_size >> 9;
179 static inline void *bio_data(struct bio *bio)
182 return page_address(bio_page(bio)) + bio_offset(bio);
190 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
191 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
194 * queues that have highmem support enabled may still need to revert to
195 * PIO transfers occasionally and thus map high pages temporarily. For
196 * permanent PIO fall back, user is probably better off disabling highmem
197 * I/O completely on that queue (see ide-dma for example)
199 #define __bio_kmap_atomic(bio, idx, kmtype) \
200 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \
201 bio_iovec_idx((bio), (idx))->bv_offset)
203 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
209 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
210 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx)
213 * allow arch override, for eg virtualized architectures (put in asm/io.h)
215 #ifndef BIOVEC_PHYS_MERGEABLE
216 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
217 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
220 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
221 (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
222 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
223 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask)
224 #define BIO_SEG_BOUNDARY(q, b1, b2) \
225 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
227 #define bio_io_error(bio) bio_endio((bio), -EIO)
230 * drivers should not use the __ version unless they _really_ want to
231 * run through the entire bio and not just pending pieces
233 #define __bio_for_each_segment(bvl, bio, i, start_idx) \
234 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \
235 i < (bio)->bi_vcnt; \
238 #define bio_for_each_segment(bvl, bio, i) \
239 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
242 * get a reference to a bio, so it won't disappear. the intended use is
246 * submit_bio(rw, bio);
247 * if (bio->bi_flags ...)
251 * without the bio_get(), it could potentially complete I/O before submit_bio
252 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
255 #define bio_get(bio) atomic_inc(&(bio)->bi_cnt)
257 #if defined(CONFIG_BLK_DEV_INTEGRITY)
259 * bio integrity payload
261 struct bio_integrity_payload {
262 struct bio *bip_bio; /* parent bio */
263 struct bio_vec *bip_vec; /* integrity data vector */
265 sector_t bip_sector; /* virtual start sector */
267 void *bip_buf; /* generated integrity data */
268 bio_end_io_t *bip_end_io; /* saved I/O completion fn */
270 int bip_error; /* saved I/O error */
271 unsigned int bip_size;
273 unsigned short bip_pool; /* pool the ivec came from */
274 unsigned short bip_vcnt; /* # of integrity bio_vecs */
275 unsigned short bip_idx; /* current bip_vec index */
277 struct work_struct bip_work; /* I/O completion */
279 #endif /* CONFIG_BLK_DEV_INTEGRITY */
282 * A bio_pair is used when we need to split a bio.
283 * This can only happen for a bio that refers to just one
284 * page of data, and in the unusual situation when the
285 * page crosses a chunk/device boundary
287 * The address of the master bio is stored in bio1.bi_private
288 * The address of the pool the pair was allocated from is stored
292 struct bio bio1, bio2;
293 struct bio_vec bv1, bv2;
294 #if defined(CONFIG_BLK_DEV_INTEGRITY)
295 struct bio_integrity_payload bip1, bip2;
296 struct bio_vec iv1, iv2;
301 extern struct bio_pair *bio_split(struct bio *bi, mempool_t *pool,
303 extern mempool_t *bio_split_pool;
304 extern void bio_pair_release(struct bio_pair *dbio);
306 extern struct bio_set *bioset_create(int, int);
307 extern void bioset_free(struct bio_set *);
309 extern struct bio *bio_alloc(gfp_t, int);
310 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
311 extern void bio_put(struct bio *);
312 extern void bio_free(struct bio *, struct bio_set *);
314 extern void bio_endio(struct bio *, int);
315 struct request_queue;
316 extern int bio_phys_segments(struct request_queue *, struct bio *);
318 extern void __bio_clone(struct bio *, struct bio *);
319 extern struct bio *bio_clone(struct bio *, gfp_t);
321 extern void bio_init(struct bio *);
323 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
324 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
325 unsigned int, unsigned int);
326 extern int bio_get_nr_vecs(struct block_device *);
327 extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
328 unsigned long, unsigned int, int);
330 extern struct bio *bio_map_user_iov(struct request_queue *,
331 struct block_device *,
332 struct sg_iovec *, int, int);
333 extern void bio_unmap_user(struct bio *);
334 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
336 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
338 extern void bio_set_pages_dirty(struct bio *bio);
339 extern void bio_check_pages_dirty(struct bio *bio);
340 extern struct bio *bio_copy_user(struct request_queue *, unsigned long, unsigned int, int);
341 extern struct bio *bio_copy_user_iov(struct request_queue *, struct sg_iovec *,
343 extern int bio_uncopy_user(struct bio *);
344 void zero_fill_bio(struct bio *bio);
345 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *);
346 extern unsigned int bvec_nr_vecs(unsigned short idx);
349 * Allow queuer to specify a completion CPU for this bio
351 static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu)
353 bio->bi_comp_cpu = cpu;
357 * bio_set is used to allow other portions of the IO system to
358 * allocate their own private memory pools for bio and iovec structures.
359 * These memory pools in turn all allocate from the bio_slab
360 * and the bvec_slabs[].
362 #define BIO_POOL_SIZE 2
363 #define BIOVEC_NR_POOLS 6
367 #if defined(CONFIG_BLK_DEV_INTEGRITY)
368 mempool_t *bio_integrity_pool;
370 mempool_t *bvec_pools[BIOVEC_NR_POOLS];
376 struct kmem_cache *slab;
379 extern struct bio_set *fs_bio_set;
382 * a small number of entries is fine, not going to be performance critical.
383 * basically we just need to survive
385 #define BIO_SPLIT_ENTRIES 2
387 #ifdef CONFIG_HIGHMEM
389 * remember to add offset! and never ever reenable interrupts between a
390 * bvec_kmap_irq and bvec_kunmap_irq!!
392 * This function MUST be inlined - it plays with the CPU interrupt flags.
394 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
399 * might not be a highmem page, but the preempt/irq count
400 * balancing is a lot nicer this way
402 local_irq_save(*flags);
403 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
405 BUG_ON(addr & ~PAGE_MASK);
407 return (char *) addr + bvec->bv_offset;
410 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
412 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
414 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
415 local_irq_restore(*flags);
419 #define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset)
420 #define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0)
423 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
424 unsigned long *flags)
426 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
428 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
430 #define bio_kmap_irq(bio, flags) \
431 __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
432 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
435 * Check whether this bio carries any data or not. A NULL bio is allowed.
437 static inline int bio_has_data(struct bio *bio)
439 return bio && bio->bi_io_vec != NULL;
442 #if defined(CONFIG_BLK_DEV_INTEGRITY)
444 #define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)]))
445 #define bip_vec(bip) bip_vec_idx(bip, 0)
447 #define __bip_for_each_vec(bvl, bip, i, start_idx) \
448 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \
449 i < (bip)->bip_vcnt; \
452 #define bip_for_each_vec(bvl, bip, i) \
453 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx)
455 static inline int bio_integrity(struct bio *bio)
457 #if defined(CONFIG_BLK_DEV_INTEGRITY)
458 return bio->bi_integrity != NULL;
464 extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *);
465 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
466 extern void bio_integrity_free(struct bio *, struct bio_set *);
467 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
468 extern int bio_integrity_enabled(struct bio *bio);
469 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int);
470 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int);
471 extern int bio_integrity_prep(struct bio *);
472 extern void bio_integrity_endio(struct bio *, int);
473 extern void bio_integrity_advance(struct bio *, unsigned int);
474 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int);
475 extern void bio_integrity_split(struct bio *, struct bio_pair *, int);
476 extern int bio_integrity_clone(struct bio *, struct bio *, struct bio_set *);
477 extern int bioset_integrity_create(struct bio_set *, int);
478 extern void bioset_integrity_free(struct bio_set *);
479 extern void bio_integrity_init_slab(void);
481 #else /* CONFIG_BLK_DEV_INTEGRITY */
483 #define bio_integrity(a) (0)
484 #define bioset_integrity_create(a, b) (0)
485 #define bio_integrity_prep(a) (0)
486 #define bio_integrity_enabled(a) (0)
487 #define bio_integrity_clone(a, b, c) (0)
488 #define bioset_integrity_free(a) do { } while (0)
489 #define bio_integrity_free(a, b) do { } while (0)
490 #define bio_integrity_endio(a, b) do { } while (0)
491 #define bio_integrity_advance(a, b) do { } while (0)
492 #define bio_integrity_trim(a, b, c) do { } while (0)
493 #define bio_integrity_split(a, b, c) do { } while (0)
494 #define bio_integrity_set_tag(a, b, c) do { } while (0)
495 #define bio_integrity_get_tag(a, b, c) do { } while (0)
496 #define bio_integrity_init_slab(a) do { } while (0)
498 #endif /* CONFIG_BLK_DEV_INTEGRITY */
500 #endif /* CONFIG_BLOCK */
501 #endif /* __LINUX_BIO_H */