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>
27 /* Platforms may set this to teach the BIO layer about IOMMU hardware. */
30 #if defined(BIO_VMERGE_MAX_SIZE) && defined(BIO_VMERGE_BOUNDARY)
31 #define BIOVEC_VIRT_START_SIZE(x) (bvec_to_phys(x) & (BIO_VMERGE_BOUNDARY - 1))
32 #define BIOVEC_VIRT_OVERSIZE(x) ((x) > BIO_VMERGE_MAX_SIZE)
34 #define BIOVEC_VIRT_START_SIZE(x) 0
35 #define BIOVEC_VIRT_OVERSIZE(x) 0
38 #ifndef BIO_VMERGE_BOUNDARY
39 #define BIO_VMERGE_BOUNDARY 0
45 #define BIO_BUG_ON BUG_ON
50 #define BIO_MAX_PAGES 256
51 #define BIO_MAX_SIZE (BIO_MAX_PAGES << PAGE_CACHE_SHIFT)
52 #define BIO_MAX_SECTORS (BIO_MAX_SIZE >> 9)
55 * was unsigned short, but we might as well be ready for > 64kB I/O pages
60 unsigned int bv_offset;
65 typedef int (bio_end_io_t) (struct bio *, unsigned int, int);
66 typedef void (bio_destructor_t) (struct bio *);
69 * main unit of I/O for the block layer and lower layers (ie drivers and
73 sector_t bi_sector; /* device address in 512 byte
75 struct bio *bi_next; /* request queue link */
76 struct block_device *bi_bdev;
77 unsigned long bi_flags; /* status, command, etc */
78 unsigned long bi_rw; /* bottom bits READ/WRITE,
82 unsigned short bi_vcnt; /* how many bio_vec's */
83 unsigned short bi_idx; /* current index into bvl_vec */
85 /* Number of segments in this BIO after
86 * physical address coalescing is performed.
88 unsigned short bi_phys_segments;
90 /* Number of segments after physical and DMA remapping
91 * hardware coalescing is performed.
93 unsigned short bi_hw_segments;
95 unsigned int bi_size; /* residual I/O count */
98 * To keep track of the max hw size, we account for the
99 * sizes of the first and last virtually mergeable segments
102 unsigned int bi_hw_front_size;
103 unsigned int bi_hw_back_size;
105 unsigned int bi_max_vecs; /* max bvl_vecs we can hold */
107 struct bio_vec *bi_io_vec; /* the actual vec list */
109 bio_end_io_t *bi_end_io;
110 atomic_t bi_cnt; /* pin count */
114 bio_destructor_t *bi_destructor; /* destructor */
120 #define BIO_UPTODATE 0 /* ok after I/O completion */
121 #define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */
122 #define BIO_EOF 2 /* out-out-bounds error */
123 #define BIO_SEG_VALID 3 /* nr_hw_seg valid */
124 #define BIO_CLONED 4 /* doesn't own data */
125 #define BIO_BOUNCED 5 /* bio is a bounce bio */
126 #define BIO_USER_MAPPED 6 /* contains user pages */
127 #define BIO_EOPNOTSUPP 7 /* not supported */
128 #define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag)))
131 * top 4 bits of bio flags indicate the pool this bio came from
133 #define BIO_POOL_BITS (4)
134 #define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS)
135 #define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET)
136 #define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET)
141 * bit 0 -- read (not set) or write (set)
142 * bit 1 -- rw-ahead when set
144 * bit 3 -- fail fast, don't want low level driver retries
145 * bit 4 -- synchronous I/O hint: the block layer will unplug immediately
148 #define BIO_RW_AHEAD 1
149 #define BIO_RW_BARRIER 2
150 #define BIO_RW_FAILFAST 3
151 #define BIO_RW_SYNC 4
152 #define BIO_RW_META 5
155 * upper 16 bits of bi_rw define the io priority of this bio
157 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
158 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
159 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
161 #define bio_set_prio(bio, prio) do { \
162 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
163 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
164 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
168 * various member access, note that bio_data should of course not be used
169 * on highmem page vectors
171 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
172 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx)
173 #define bio_page(bio) bio_iovec((bio))->bv_page
174 #define bio_offset(bio) bio_iovec((bio))->bv_offset
175 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx)
176 #define bio_sectors(bio) ((bio)->bi_size >> 9)
177 #define bio_cur_sectors(bio) (bio_iovec(bio)->bv_len >> 9)
178 #define bio_data(bio) (page_address(bio_page((bio))) + bio_offset((bio)))
179 #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))
183 #define bio_rw_meta(bio) ((bio)->bi_rw & (1 << BIO_RW_META))
188 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
189 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
192 * queues that have highmem support enabled may still need to revert to
193 * PIO transfers occasionally and thus map high pages temporarily. For
194 * permanent PIO fall back, user is probably better off disabling highmem
195 * I/O completely on that queue (see ide-dma for example)
197 #define __bio_kmap_atomic(bio, idx, kmtype) \
198 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \
199 bio_iovec_idx((bio), (idx))->bv_offset)
201 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
207 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
208 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx)
211 * allow arch override, for eg virtualized architectures (put in asm/io.h)
213 #ifndef BIOVEC_PHYS_MERGEABLE
214 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
215 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
218 #define BIOVEC_VIRT_MERGEABLE(vec1, vec2) \
219 ((((bvec_to_phys((vec1)) + (vec1)->bv_len) | bvec_to_phys((vec2))) & (BIO_VMERGE_BOUNDARY - 1)) == 0)
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, bytes) bio_endio((bio), (bytes), -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)
259 * A bio_pair is used when we need to split a bio.
260 * This can only happen for a bio that refers to just one
261 * page of data, and in the unusual situation when the
262 * page crosses a chunk/device boundary
264 * The address of the master bio is stored in bio1.bi_private
265 * The address of the pool the pair was allocated from is stored
269 struct bio bio1, bio2;
270 struct bio_vec bv1, bv2;
274 extern struct bio_pair *bio_split(struct bio *bi, mempool_t *pool,
276 extern mempool_t *bio_split_pool;
277 extern void bio_pair_release(struct bio_pair *dbio);
279 extern struct bio_set *bioset_create(int, int, int);
280 extern void bioset_free(struct bio_set *);
282 extern struct bio *bio_alloc(gfp_t, int);
283 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
284 extern void bio_put(struct bio *);
285 extern void bio_free(struct bio *, struct bio_set *);
287 extern void bio_endio(struct bio *, unsigned int, int);
288 struct request_queue;
289 extern int bio_phys_segments(struct request_queue *, struct bio *);
290 extern int bio_hw_segments(struct request_queue *, struct bio *);
292 extern void __bio_clone(struct bio *, struct bio *);
293 extern struct bio *bio_clone(struct bio *, gfp_t);
295 extern void bio_init(struct bio *);
297 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
298 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
299 unsigned int, unsigned int);
300 extern int bio_get_nr_vecs(struct block_device *);
301 extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
302 unsigned long, unsigned int, int);
304 extern struct bio *bio_map_user_iov(struct request_queue *,
305 struct block_device *,
306 struct sg_iovec *, int, int);
307 extern void bio_unmap_user(struct bio *);
308 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
310 extern void bio_set_pages_dirty(struct bio *bio);
311 extern void bio_check_pages_dirty(struct bio *bio);
312 extern void bio_release_pages(struct bio *bio);
313 extern struct bio *bio_copy_user(struct request_queue *, unsigned long, unsigned int, int);
314 extern int bio_uncopy_user(struct bio *);
315 void zero_fill_bio(struct bio *bio);
317 #ifdef CONFIG_HIGHMEM
319 * remember to add offset! and never ever reenable interrupts between a
320 * bvec_kmap_irq and bvec_kunmap_irq!!
322 * This function MUST be inlined - it plays with the CPU interrupt flags.
324 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
329 * might not be a highmem page, but the preempt/irq count
330 * balancing is a lot nicer this way
332 local_irq_save(*flags);
333 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
335 BUG_ON(addr & ~PAGE_MASK);
337 return (char *) addr + bvec->bv_offset;
340 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
342 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
344 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
345 local_irq_restore(*flags);
349 #define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset)
350 #define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0)
353 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
354 unsigned long *flags)
356 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
358 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
360 #define bio_kmap_irq(bio, flags) \
361 __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
362 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
364 #endif /* __LINUX_BIO_H */