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 oen is issued.
148 * bit 3 -- synchronous I/O hint: the block layer will unplug immediately
149 * Note that this does NOT indicate that the IO itself is sync, just
150 * that the block layer will not postpone issue of this IO by plugging.
151 * bit 4 -- metadata request
152 * Used for tracing to differentiate metadata and data IO. May also
153 * get some preferential treatment in the IO scheduler
154 * bit 5 -- discard sectors
155 * Informs the lower level device that this range of sectors is no longer
156 * used by the file system and may thus be freed by the device. Used
157 * for flash based storage.
158 * bit 6 -- fail fast device errors
159 * bit 7 -- fail fast transport errors
160 * bit 8 -- fail fast driver errors
161 * Don't want driver retries for any fast fail whatever the reason.
163 #define BIO_RW 0 /* Must match RW in req flags (blkdev.h) */
164 #define BIO_RW_AHEAD 1 /* Must match FAILFAST in req flags */
165 #define BIO_RW_BARRIER 2
166 #define BIO_RW_SYNC 3
167 #define BIO_RW_META 4
168 #define BIO_RW_DISCARD 5
169 #define BIO_RW_FAILFAST_DEV 6
170 #define BIO_RW_FAILFAST_TRANSPORT 7
171 #define BIO_RW_FAILFAST_DRIVER 8
174 * upper 16 bits of bi_rw define the io priority of this bio
176 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
177 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
178 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
180 #define bio_set_prio(bio, prio) do { \
181 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
182 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
183 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
187 * various member access, note that bio_data should of course not be used
188 * on highmem page vectors
190 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
191 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx)
192 #define bio_page(bio) bio_iovec((bio))->bv_page
193 #define bio_offset(bio) bio_iovec((bio))->bv_offset
194 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx)
195 #define bio_sectors(bio) ((bio)->bi_size >> 9)
196 #define bio_barrier(bio) ((bio)->bi_rw & (1 << BIO_RW_BARRIER))
197 #define bio_sync(bio) ((bio)->bi_rw & (1 << BIO_RW_SYNC))
198 #define bio_failfast_dev(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST_DEV))
199 #define bio_failfast_transport(bio) \
200 ((bio)->bi_rw & (1 << BIO_RW_FAILFAST_TRANSPORT))
201 #define bio_failfast_driver(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST_DRIVER))
202 #define bio_rw_ahead(bio) ((bio)->bi_rw & (1 << BIO_RW_AHEAD))
203 #define bio_rw_meta(bio) ((bio)->bi_rw & (1 << BIO_RW_META))
204 #define bio_discard(bio) ((bio)->bi_rw & (1 << BIO_RW_DISCARD))
205 #define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio) && !bio_discard(bio))
207 static inline unsigned int bio_cur_sectors(struct bio *bio)
210 return bio_iovec(bio)->bv_len >> 9;
211 else /* dataless requests such as discard */
212 return bio->bi_size >> 9;
215 static inline void *bio_data(struct bio *bio)
218 return page_address(bio_page(bio)) + bio_offset(bio);
223 static inline int bio_has_allocated_vec(struct bio *bio)
225 return bio->bi_io_vec && bio->bi_io_vec != bio->bi_inline_vecs;
231 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
232 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
235 * queues that have highmem support enabled may still need to revert to
236 * PIO transfers occasionally and thus map high pages temporarily. For
237 * permanent PIO fall back, user is probably better off disabling highmem
238 * I/O completely on that queue (see ide-dma for example)
240 #define __bio_kmap_atomic(bio, idx, kmtype) \
241 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \
242 bio_iovec_idx((bio), (idx))->bv_offset)
244 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
250 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
251 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx)
253 /* Default implementation of BIOVEC_PHYS_MERGEABLE */
254 #define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
255 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
258 * allow arch override, for eg virtualized architectures (put in asm/io.h)
260 #ifndef BIOVEC_PHYS_MERGEABLE
261 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
262 __BIOVEC_PHYS_MERGEABLE(vec1, vec2)
265 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
266 (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
267 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
268 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask)
269 #define BIO_SEG_BOUNDARY(q, b1, b2) \
270 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
272 #define bio_io_error(bio) bio_endio((bio), -EIO)
275 * drivers should not use the __ version unless they _really_ want to
276 * run through the entire bio and not just pending pieces
278 #define __bio_for_each_segment(bvl, bio, i, start_idx) \
279 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \
280 i < (bio)->bi_vcnt; \
283 #define bio_for_each_segment(bvl, bio, i) \
284 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
287 * get a reference to a bio, so it won't disappear. the intended use is
291 * submit_bio(rw, bio);
292 * if (bio->bi_flags ...)
296 * without the bio_get(), it could potentially complete I/O before submit_bio
297 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
300 #define bio_get(bio) atomic_inc(&(bio)->bi_cnt)
302 #if defined(CONFIG_BLK_DEV_INTEGRITY)
304 * bio integrity payload
306 struct bio_integrity_payload {
307 struct bio *bip_bio; /* parent bio */
308 struct bio_vec *bip_vec; /* integrity data vector */
310 sector_t bip_sector; /* virtual start sector */
312 void *bip_buf; /* generated integrity data */
313 bio_end_io_t *bip_end_io; /* saved I/O completion fn */
315 int bip_error; /* saved I/O error */
316 unsigned int bip_size;
318 unsigned short bip_pool; /* pool the ivec came from */
319 unsigned short bip_vcnt; /* # of integrity bio_vecs */
320 unsigned short bip_idx; /* current bip_vec index */
322 struct work_struct bip_work; /* I/O completion */
324 #endif /* CONFIG_BLK_DEV_INTEGRITY */
327 * A bio_pair is used when we need to split a bio.
328 * This can only happen for a bio that refers to just one
329 * page of data, and in the unusual situation when the
330 * page crosses a chunk/device boundary
332 * The address of the master bio is stored in bio1.bi_private
333 * The address of the pool the pair was allocated from is stored
337 struct bio bio1, bio2;
338 struct bio_vec bv1, bv2;
339 #if defined(CONFIG_BLK_DEV_INTEGRITY)
340 struct bio_integrity_payload bip1, bip2;
341 struct bio_vec iv1, iv2;
346 extern struct bio_pair *bio_split(struct bio *bi, int first_sectors);
347 extern void bio_pair_release(struct bio_pair *dbio);
349 extern struct bio_set *bioset_create(unsigned int, unsigned int);
350 extern void bioset_free(struct bio_set *);
352 extern struct bio *bio_alloc(gfp_t, int);
353 extern struct bio *bio_kmalloc(gfp_t, int);
354 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
355 extern void bio_put(struct bio *);
356 extern void bio_free(struct bio *, struct bio_set *);
358 extern void bio_endio(struct bio *, int);
359 struct request_queue;
360 extern int bio_phys_segments(struct request_queue *, struct bio *);
362 extern void __bio_clone(struct bio *, struct bio *);
363 extern struct bio *bio_clone(struct bio *, gfp_t);
365 extern void bio_init(struct bio *);
367 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
368 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
369 unsigned int, unsigned int);
370 extern int bio_get_nr_vecs(struct block_device *);
371 extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int);
372 extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
373 unsigned long, unsigned int, int, gfp_t);
376 extern struct bio *bio_map_user_iov(struct request_queue *,
377 struct block_device *,
378 struct sg_iovec *, int, int, gfp_t);
379 extern void bio_unmap_user(struct bio *);
380 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
382 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
384 extern void bio_set_pages_dirty(struct bio *bio);
385 extern void bio_check_pages_dirty(struct bio *bio);
386 extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *,
387 unsigned long, unsigned int, int, gfp_t);
388 extern struct bio *bio_copy_user_iov(struct request_queue *,
389 struct rq_map_data *, struct sg_iovec *,
391 extern int bio_uncopy_user(struct bio *);
392 void zero_fill_bio(struct bio *bio);
393 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *);
394 extern void bvec_free_bs(struct bio_set *, struct bio_vec *, unsigned int);
395 extern unsigned int bvec_nr_vecs(unsigned short idx);
398 * Allow queuer to specify a completion CPU for this bio
400 static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu)
402 bio->bi_comp_cpu = cpu;
406 * bio_set is used to allow other portions of the IO system to
407 * allocate their own private memory pools for bio and iovec structures.
408 * These memory pools in turn all allocate from the bio_slab
409 * and the bvec_slabs[].
411 #define BIO_POOL_SIZE 2
412 #define BIOVEC_NR_POOLS 6
413 #define BIOVEC_MAX_IDX (BIOVEC_NR_POOLS - 1)
416 struct kmem_cache *bio_slab;
417 unsigned int front_pad;
420 #if defined(CONFIG_BLK_DEV_INTEGRITY)
421 mempool_t *bio_integrity_pool;
423 mempool_t *bvec_pool;
429 struct kmem_cache *slab;
432 extern struct bio_set *fs_bio_set;
433 extern struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly;
436 * a small number of entries is fine, not going to be performance critical.
437 * basically we just need to survive
439 #define BIO_SPLIT_ENTRIES 2
441 #ifdef CONFIG_HIGHMEM
443 * remember to add offset! and never ever reenable interrupts between a
444 * bvec_kmap_irq and bvec_kunmap_irq!!
446 * This function MUST be inlined - it plays with the CPU interrupt flags.
448 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
453 * might not be a highmem page, but the preempt/irq count
454 * balancing is a lot nicer this way
456 local_irq_save(*flags);
457 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
459 BUG_ON(addr & ~PAGE_MASK);
461 return (char *) addr + bvec->bv_offset;
464 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
466 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
468 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
469 local_irq_restore(*flags);
473 #define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset)
474 #define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0)
477 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
478 unsigned long *flags)
480 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
482 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
484 #define bio_kmap_irq(bio, flags) \
485 __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
486 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
489 * Check whether this bio carries any data or not. A NULL bio is allowed.
491 static inline int bio_has_data(struct bio *bio)
493 return bio && bio->bi_io_vec != NULL;
496 #if defined(CONFIG_BLK_DEV_INTEGRITY)
498 #define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)]))
499 #define bip_vec(bip) bip_vec_idx(bip, 0)
501 #define __bip_for_each_vec(bvl, bip, i, start_idx) \
502 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \
503 i < (bip)->bip_vcnt; \
506 #define bip_for_each_vec(bvl, bip, i) \
507 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx)
509 #define bio_integrity(bio) (bio->bi_integrity != NULL)
511 extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *);
512 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
513 extern void bio_integrity_free(struct bio *, struct bio_set *);
514 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
515 extern int bio_integrity_enabled(struct bio *bio);
516 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int);
517 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int);
518 extern int bio_integrity_prep(struct bio *);
519 extern void bio_integrity_endio(struct bio *, int);
520 extern void bio_integrity_advance(struct bio *, unsigned int);
521 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int);
522 extern void bio_integrity_split(struct bio *, struct bio_pair *, int);
523 extern int bio_integrity_clone(struct bio *, struct bio *, struct bio_set *);
524 extern int bioset_integrity_create(struct bio_set *, int);
525 extern void bioset_integrity_free(struct bio_set *);
526 extern void bio_integrity_init_slab(void);
528 #else /* CONFIG_BLK_DEV_INTEGRITY */
530 #define bio_integrity(a) (0)
531 #define bioset_integrity_create(a, b) (0)
532 #define bio_integrity_prep(a) (0)
533 #define bio_integrity_enabled(a) (0)
534 #define bio_integrity_clone(a, b, c) (0)
535 #define bioset_integrity_free(a) do { } while (0)
536 #define bio_integrity_free(a, b) do { } while (0)
537 #define bio_integrity_endio(a, b) do { } while (0)
538 #define bio_integrity_advance(a, b) do { } while (0)
539 #define bio_integrity_trim(a, b, c) do { } while (0)
540 #define bio_integrity_split(a, b, c) do { } while (0)
541 #define bio_integrity_set_tag(a, b, c) do { } while (0)
542 #define bio_integrity_get_tag(a, b, c) do { } while (0)
543 #define bio_integrity_init_slab(a) do { } while (0)
545 #endif /* CONFIG_BLK_DEV_INTEGRITY */
547 #endif /* CONFIG_BLOCK */
548 #endif /* __LINUX_BIO_H */