blob: 4da441337d6effd31990afc5616d7772010d0f47 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * 2.5 block I/O model
3 *
4 * Copyright (C) 2001 Jens Axboe <axboe@suse.de>
5 *
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.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
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-
19 */
20#ifndef __LINUX_BIO_H
21#define __LINUX_BIO_H
22
23#include <linux/highmem.h>
24#include <linux/mempool.h>
Jens Axboe22e2c502005-06-27 10:55:12 +020025#include <linux/ioprio.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026
David Howells02a5e0a2007-08-11 22:34:32 +020027#ifdef CONFIG_BLOCK
28
Linus Torvalds1da177e2005-04-16 15:20:36 -070029/* Platforms may set this to teach the BIO layer about IOMMU hardware. */
30#include <asm/io.h>
31
32#if defined(BIO_VMERGE_MAX_SIZE) && defined(BIO_VMERGE_BOUNDARY)
33#define BIOVEC_VIRT_START_SIZE(x) (bvec_to_phys(x) & (BIO_VMERGE_BOUNDARY - 1))
34#define BIOVEC_VIRT_OVERSIZE(x) ((x) > BIO_VMERGE_MAX_SIZE)
35#else
36#define BIOVEC_VIRT_START_SIZE(x) 0
37#define BIOVEC_VIRT_OVERSIZE(x) 0
38#endif
39
40#ifndef BIO_VMERGE_BOUNDARY
41#define BIO_VMERGE_BOUNDARY 0
42#endif
43
44#define BIO_DEBUG
45
46#ifdef BIO_DEBUG
47#define BIO_BUG_ON BUG_ON
48#else
49#define BIO_BUG_ON
50#endif
51
Alexey Dobriyand84a8472006-06-25 05:49:32 -070052#define BIO_MAX_PAGES 256
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#define BIO_MAX_SIZE (BIO_MAX_PAGES << PAGE_CACHE_SHIFT)
54#define BIO_MAX_SECTORS (BIO_MAX_SIZE >> 9)
55
56/*
57 * was unsigned short, but we might as well be ready for > 64kB I/O pages
58 */
59struct bio_vec {
60 struct page *bv_page;
61 unsigned int bv_len;
62 unsigned int bv_offset;
63};
64
65struct bio_set;
66struct bio;
NeilBrown6712ecf2007-09-27 12:47:43 +020067typedef void (bio_end_io_t) (struct bio *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068typedef void (bio_destructor_t) (struct bio *);
69
70/*
71 * main unit of I/O for the block layer and lower layers (ie drivers and
72 * stacking drivers)
73 */
74struct bio {
Roger Gammans2c2345c2006-10-04 13:37:45 +020075 sector_t bi_sector; /* device address in 512 byte
76 sectors */
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 struct bio *bi_next; /* request queue link */
78 struct block_device *bi_bdev;
79 unsigned long bi_flags; /* status, command, etc */
80 unsigned long bi_rw; /* bottom bits READ/WRITE,
81 * top bits priority
82 */
83
84 unsigned short bi_vcnt; /* how many bio_vec's */
85 unsigned short bi_idx; /* current index into bvl_vec */
86
87 /* Number of segments in this BIO after
88 * physical address coalescing is performed.
89 */
90 unsigned short bi_phys_segments;
91
92 /* Number of segments after physical and DMA remapping
93 * hardware coalescing is performed.
94 */
95 unsigned short bi_hw_segments;
96
97 unsigned int bi_size; /* residual I/O count */
98
99 /*
100 * To keep track of the max hw size, we account for the
101 * sizes of the first and last virtually mergeable segments
102 * in this bio
103 */
104 unsigned int bi_hw_front_size;
105 unsigned int bi_hw_back_size;
106
107 unsigned int bi_max_vecs; /* max bvl_vecs we can hold */
108
109 struct bio_vec *bi_io_vec; /* the actual vec list */
110
111 bio_end_io_t *bi_end_io;
112 atomic_t bi_cnt; /* pin count */
113
114 void *bi_private;
115
116 bio_destructor_t *bi_destructor; /* destructor */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117};
118
119/*
120 * bio flags
121 */
122#define BIO_UPTODATE 0 /* ok after I/O completion */
123#define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */
124#define BIO_EOF 2 /* out-out-bounds error */
125#define BIO_SEG_VALID 3 /* nr_hw_seg valid */
126#define BIO_CLONED 4 /* doesn't own data */
127#define BIO_BOUNCED 5 /* bio is a bounce bio */
128#define BIO_USER_MAPPED 6 /* contains user pages */
129#define BIO_EOPNOTSUPP 7 /* not supported */
130#define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag)))
131
132/*
133 * top 4 bits of bio flags indicate the pool this bio came from
134 */
135#define BIO_POOL_BITS (4)
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)
139
140/*
141 * bio bi_rw flags
142 *
143 * bit 0 -- read (not set) or write (set)
144 * bit 1 -- rw-ahead when set
145 * bit 2 -- barrier
146 * bit 3 -- fail fast, don't want low level driver retries
147 * bit 4 -- synchronous I/O hint: the block layer will unplug immediately
148 */
149#define BIO_RW 0
150#define BIO_RW_AHEAD 1
151#define BIO_RW_BARRIER 2
152#define BIO_RW_FAILFAST 3
153#define BIO_RW_SYNC 4
Jens Axboe5404bc72006-08-10 09:01:02 +0200154#define BIO_RW_META 5
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
156/*
Jens Axboe22e2c502005-06-27 10:55:12 +0200157 * upper 16 bits of bi_rw define the io priority of this bio
158 */
159#define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
160#define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
161#define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
162
163#define bio_set_prio(bio, prio) do { \
164 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
165 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
166 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
167} while (0)
168
169/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 * various member access, note that bio_data should of course not be used
171 * on highmem page vectors
172 */
173#define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
174#define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx)
175#define bio_page(bio) bio_iovec((bio))->bv_page
176#define bio_offset(bio) bio_iovec((bio))->bv_offset
177#define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx)
178#define bio_sectors(bio) ((bio)->bi_size >> 9)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179#define bio_barrier(bio) ((bio)->bi_rw & (1 << BIO_RW_BARRIER))
180#define bio_sync(bio) ((bio)->bi_rw & (1 << BIO_RW_SYNC))
181#define bio_failfast(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST))
182#define bio_rw_ahead(bio) ((bio)->bi_rw & (1 << BIO_RW_AHEAD))
Jens Axboe5404bc72006-08-10 09:01:02 +0200183#define bio_rw_meta(bio) ((bio)->bi_rw & (1 << BIO_RW_META))
Jens Axboebf2de6f2007-09-27 13:01:25 +0200184#define bio_empty_barrier(bio) (bio_barrier(bio) && !(bio)->bi_size)
185
186static inline unsigned int bio_cur_sectors(struct bio *bio)
187{
188 if (bio->bi_vcnt)
189 return bio_iovec(bio)->bv_len >> 9;
190
191 return 0;
192}
193
194static inline void *bio_data(struct bio *bio)
195{
196 if (bio->bi_vcnt)
197 return page_address(bio_page(bio)) + bio_offset(bio);
198
199 return NULL;
200}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201
202/*
203 * will die
204 */
205#define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
206#define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
207
208/*
209 * queues that have highmem support enabled may still need to revert to
210 * PIO transfers occasionally and thus map high pages temporarily. For
211 * permanent PIO fall back, user is probably better off disabling highmem
212 * I/O completely on that queue (see ide-dma for example)
213 */
214#define __bio_kmap_atomic(bio, idx, kmtype) \
215 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \
216 bio_iovec_idx((bio), (idx))->bv_offset)
217
218#define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
219
220/*
221 * merge helpers etc
222 */
223
224#define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
225#define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx)
226
227/*
228 * allow arch override, for eg virtualized architectures (put in asm/io.h)
229 */
230#ifndef BIOVEC_PHYS_MERGEABLE
231#define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
232 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
233#endif
234
235#define BIOVEC_VIRT_MERGEABLE(vec1, vec2) \
236 ((((bvec_to_phys((vec1)) + (vec1)->bv_len) | bvec_to_phys((vec2))) & (BIO_VMERGE_BOUNDARY - 1)) == 0)
237#define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
238 (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
239#define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
240 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask)
241#define BIO_SEG_BOUNDARY(q, b1, b2) \
242 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
243
NeilBrown6712ecf2007-09-27 12:47:43 +0200244#define bio_io_error(bio) bio_endio((bio), -EIO)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245
246/*
247 * drivers should not use the __ version unless they _really_ want to
248 * run through the entire bio and not just pending pieces
249 */
250#define __bio_for_each_segment(bvl, bio, i, start_idx) \
251 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \
252 i < (bio)->bi_vcnt; \
253 bvl++, i++)
254
255#define bio_for_each_segment(bvl, bio, i) \
256 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
257
258/*
259 * get a reference to a bio, so it won't disappear. the intended use is
260 * something like:
261 *
262 * bio_get(bio);
263 * submit_bio(rw, bio);
264 * if (bio->bi_flags ...)
265 * do_something
266 * bio_put(bio);
267 *
268 * without the bio_get(), it could potentially complete I/O before submit_bio
269 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
270 * runs
271 */
272#define bio_get(bio) atomic_inc(&(bio)->bi_cnt)
273
274
275/*
276 * A bio_pair is used when we need to split a bio.
277 * This can only happen for a bio that refers to just one
278 * page of data, and in the unusual situation when the
279 * page crosses a chunk/device boundary
280 *
281 * The address of the master bio is stored in bio1.bi_private
282 * The address of the pool the pair was allocated from is stored
283 * in bio2.bi_private
284 */
285struct bio_pair {
286 struct bio bio1, bio2;
287 struct bio_vec bv1, bv2;
288 atomic_t cnt;
289 int error;
290};
291extern struct bio_pair *bio_split(struct bio *bi, mempool_t *pool,
292 int first_sectors);
293extern mempool_t *bio_split_pool;
294extern void bio_pair_release(struct bio_pair *dbio);
295
Jens Axboe59725112007-04-02 10:06:42 +0200296extern struct bio_set *bioset_create(int, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297extern void bioset_free(struct bio_set *);
298
Al Virodd0fc662005-10-07 07:46:04 +0100299extern struct bio *bio_alloc(gfp_t, int);
300extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301extern void bio_put(struct bio *);
Peter Osterlund36763472005-09-06 15:16:42 -0700302extern void bio_free(struct bio *, struct bio_set *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
NeilBrown6712ecf2007-09-27 12:47:43 +0200304extern void bio_endio(struct bio *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305struct request_queue;
306extern int bio_phys_segments(struct request_queue *, struct bio *);
307extern int bio_hw_segments(struct request_queue *, struct bio *);
308
309extern void __bio_clone(struct bio *, struct bio *);
Al Virodd0fc662005-10-07 07:46:04 +0100310extern struct bio *bio_clone(struct bio *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311
312extern void bio_init(struct bio *);
313
314extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
Mike Christie6e68af62005-11-11 05:30:27 -0600315extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
316 unsigned int, unsigned int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317extern int bio_get_nr_vecs(struct block_device *);
318extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
319 unsigned long, unsigned int, int);
James Bottomley f1970ba2005-06-20 14:06:52 +0200320struct sg_iovec;
321extern struct bio *bio_map_user_iov(struct request_queue *,
322 struct block_device *,
323 struct sg_iovec *, int, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324extern void bio_unmap_user(struct bio *);
Mike Christie df46b9a2005-06-20 14:04:44 +0200325extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
Al Viro27496a82005-10-21 03:20:48 -0400326 gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327extern void bio_set_pages_dirty(struct bio *bio);
328extern void bio_check_pages_dirty(struct bio *bio);
Chen, Kenneth We61c9012006-12-13 00:34:36 -0800329extern void bio_release_pages(struct bio *bio);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330extern struct bio *bio_copy_user(struct request_queue *, unsigned long, unsigned int, int);
331extern int bio_uncopy_user(struct bio *);
332void zero_fill_bio(struct bio *bio);
333
334#ifdef CONFIG_HIGHMEM
335/*
336 * remember to add offset! and never ever reenable interrupts between a
337 * bvec_kmap_irq and bvec_kunmap_irq!!
338 *
339 * This function MUST be inlined - it plays with the CPU interrupt flags.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 */
Adrian Bunkc2d08da2005-09-10 00:27:18 -0700341static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342{
343 unsigned long addr;
344
345 /*
346 * might not be a highmem page, but the preempt/irq count
347 * balancing is a lot nicer this way
348 */
349 local_irq_save(*flags);
350 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
351
352 BUG_ON(addr & ~PAGE_MASK);
353
354 return (char *) addr + bvec->bv_offset;
355}
356
Adrian Bunkc2d08da2005-09-10 00:27:18 -0700357static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358{
359 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
360
361 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
362 local_irq_restore(*flags);
363}
364
365#else
366#define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset)
367#define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0)
368#endif
369
Adrian Bunkc2d08da2005-09-10 00:27:18 -0700370static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 unsigned long *flags)
372{
373 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
374}
375#define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
376
377#define bio_kmap_irq(bio, flags) \
378 __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
379#define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
380
David Howells02a5e0a2007-08-11 22:34:32 +0200381#endif /* CONFIG_BLOCK */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382#endif /* __LINUX_BIO_H */