include/linux/bio.h

   1 /*
   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>
  25 #include <linux/ioprio.h>
  26
  27 /* Platforms may set this to teach the BIO layer about IOMMU hardware. */
  28 #include <asm/io.h>
  29
  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)
  33 #else
  34 #define BIOVEC_VIRT_START_SIZE(x)       0
  35 #define BIOVEC_VIRT_OVERSIZE(x)         0
  36 #endif
  37
  38 #ifndef BIO_VMERGE_BOUNDARY
  39 #define BIO_VMERGE_BOUNDARY     0
  40 #endif
  41
  42 #define BIO_DEBUG
  43
  44 #ifdef BIO_DEBUG
  45 #define BIO_BUG_ON      BUG_ON
  46 #else
  47 #define BIO_BUG_ON
  48 #endif
  49
  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)
  53
  54 /*
  55  * was unsigned short, but we might as well be ready for > 64kB I/O pages
  56  */
  57 struct bio_vec {
  58         struct page     *bv_page;
  59         unsigned int    bv_len;
  60         unsigned int    bv_offset;
  61 };
  62
  63 struct bio_set;
  64 struct bio;
  65 typedef int (bio_end_io_t) (struct bio *, unsigned int, int);
  66 typedef void (bio_destructor_t) (struct bio *);
  67
  68 /*
  69  * main unit of I/O for the block layer and lower layers (ie drivers and
  70  * stacking drivers)
  71  */
  72 struct bio {
  73         sector_t                bi_sector;
  74         struct bio              *bi_next;       /* request queue link */
  75         struct block_device     *bi_bdev;
  76         unsigned long           bi_flags;       /* status, command, etc */
  77         unsigned long           bi_rw;          /* bottom bits READ/WRITE,
  78                                                  * top bits priority
  79                                                  */
  80
  81         unsigned short          bi_vcnt;        /* how many bio_vec's */
  82         unsigned short          bi_idx;         /* current index into bvl_vec */
  83
  84         /* Number of segments in this BIO after
  85          * physical address coalescing is performed.
  86          */
  87         unsigned short          bi_phys_segments;
  88
  89         /* Number of segments after physical and DMA remapping
  90          * hardware coalescing is performed.
  91          */
  92         unsigned short          bi_hw_segments;
  93
  94         unsigned int            bi_size;        /* residual I/O count */
  95
  96         /*
  97          * To keep track of the max hw size, we account for the
  98          * sizes of the first and last virtually mergeable segments
  99          * in this bio
 100          */
 101         unsigned int            bi_hw_front_size;
 102         unsigned int            bi_hw_back_size;
 103
 104         unsigned int            bi_max_vecs;    /* max bvl_vecs we can hold */
 105
 106         struct bio_vec          *bi_io_vec;     /* the actual vec list */
 107
 108         bio_end_io_t            *bi_end_io;
 109         atomic_t                bi_cnt;         /* pin count */
 110
 111         void                    *bi_private;
 112
 113         bio_destructor_t        *bi_destructor; /* destructor */
 114         struct bio_set          *bi_set;        /* memory pools set */
 115 };
 116
 117 /*
 118  * bio flags
 119  */
 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)))
 129
 130 /*
 131  * top 4 bits of bio flags indicate the pool this bio came from
 132  */
 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)
 137
 138 /*
 139  * bio bi_rw flags
 140  *
 141  * bit 0 -- read (not set) or write (set)
 142  * bit 1 -- rw-ahead when set
 143  * bit 2 -- barrier
 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
 146  */
 147 #define BIO_RW          0
 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
 153 /*
 154  * upper 16 bits of bi_rw define the io priority of this bio
 155  */
 156 #define BIO_PRIO_SHIFT  (8 * sizeof(unsigned long) - IOPRIO_BITS)
 157 #define bio_prio(bio)   ((bio)->bi_rw >> BIO_PRIO_SHIFT)
 158 #define bio_prio_valid(bio)     ioprio_valid(bio_prio(bio))
 159
 160 #define bio_set_prio(bio, prio)         do {                    \
 161         WARN_ON(prio >= (1 << IOPRIO_BITS));                    \
 162         (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1);          \
 163         (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT);     \
 164 } while (0)
 165
 166 /*
 167  * various member access, note that bio_data should of course not be used
 168  * on highmem page vectors
 169  */
 170 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
 171 #define bio_iovec(bio)          bio_iovec_idx((bio), (bio)->bi_idx)
 172 #define bio_page(bio)           bio_iovec((bio))->bv_page
 173 #define bio_offset(bio)         bio_iovec((bio))->bv_offset
 174 #define bio_segments(bio)       ((bio)->bi_vcnt - (bio)->bi_idx)
 175 #define bio_sectors(bio)        ((bio)->bi_size >> 9)
 176 #define bio_cur_sectors(bio)    (bio_iovec(bio)->bv_len >> 9)
 177 #define bio_data(bio)           (page_address(bio_page((bio))) + bio_offset((bio)))
 178 #define bio_barrier(bio)        ((bio)->bi_rw & (1 << BIO_RW_BARRIER))
 179 #define bio_sync(bio)           ((bio)->bi_rw & (1 << BIO_RW_SYNC))
 180 #define bio_failfast(bio)       ((bio)->bi_rw & (1 << BIO_RW_FAILFAST))
 181 #define bio_rw_ahead(bio)       ((bio)->bi_rw & (1 << BIO_RW_AHEAD))
 182
 183 /*
 184  * will die
 185  */
 186 #define bio_to_phys(bio)        (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
 187 #define bvec_to_phys(bv)        (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
 188
 189 /*
 190  * queues that have highmem support enabled may still need to revert to
 191  * PIO transfers occasionally and thus map high pages temporarily. For
 192  * permanent PIO fall back, user is probably better off disabling highmem
 193  * I/O completely on that queue (see ide-dma for example)
 194  */
 195 #define __bio_kmap_atomic(bio, idx, kmtype)                             \
 196         (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) +    \
 197                 bio_iovec_idx((bio), (idx))->bv_offset)
 198
 199 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
 200
 201 /*
 202  * merge helpers etc
 203  */
 204
 205 #define __BVEC_END(bio)         bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
 206 #define __BVEC_START(bio)       bio_iovec_idx((bio), (bio)->bi_idx)
 207
 208 /*
 209  * allow arch override, for eg virtualized architectures (put in asm/io.h)
 210  */
 211 #ifndef BIOVEC_PHYS_MERGEABLE
 212 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2)       \
 213         ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
 214 #endif
 215
 216 #define BIOVEC_VIRT_MERGEABLE(vec1, vec2)       \
 217         ((((bvec_to_phys((vec1)) + (vec1)->bv_len) | bvec_to_phys((vec2))) & (BIO_VMERGE_BOUNDARY - 1)) == 0)
 218 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
 219         (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
 220 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
 221         __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask)
 222 #define BIO_SEG_BOUNDARY(q, b1, b2) \
 223         BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
 224
 225 #define bio_io_error(bio, bytes) bio_endio((bio), (bytes), -EIO)
 226
 227 /*
 228  * drivers should not use the __ version unless they _really_ want to
 229  * run through the entire bio and not just pending pieces
 230  */
 231 #define __bio_for_each_segment(bvl, bio, i, start_idx)                  \
 232         for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx);  \
 233              i < (bio)->bi_vcnt;                                        \
 234              bvl++, i++)
 235
 236 #define bio_for_each_segment(bvl, bio, i)                               \
 237         __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
 238
 239 /*
 240  * get a reference to a bio, so it won't disappear. the intended use is
 241  * something like:
 242  *
 243  * bio_get(bio);
 244  * submit_bio(rw, bio);
 245  * if (bio->bi_flags ...)
 246  *      do_something
 247  * bio_put(bio);
 248  *
 249  * without the bio_get(), it could potentially complete I/O before submit_bio
 250  * returns. and then bio would be freed memory when if (bio->bi_flags ...)
 251  * runs
 252  */
 253 #define bio_get(bio)    atomic_inc(&(bio)->bi_cnt)
 254
 255
 256 /*
 257  * A bio_pair is used when we need to split a bio.
 258  * This can only happen for a bio that refers to just one
 259  * page of data, and in the unusual situation when the
 260  * page crosses a chunk/device boundary
 261  *
 262  * The address of the master bio is stored in bio1.bi_private
 263  * The address of the pool the pair was allocated from is stored
 264  *   in bio2.bi_private
 265  */
 266 struct bio_pair {
 267         struct bio      bio1, bio2;
 268         struct bio_vec  bv1, bv2;
 269         atomic_t        cnt;
 270         int             error;
 271 };
 272 extern struct bio_pair *bio_split(struct bio *bi, mempool_t *pool,
 273                                   int first_sectors);
 274 extern mempool_t *bio_split_pool;
 275 extern void bio_pair_release(struct bio_pair *dbio);
 276
 277 extern struct bio_set *bioset_create(int, int, int);
 278 extern void bioset_free(struct bio_set *);
 279
 280 extern struct bio *bio_alloc(unsigned int __nocast, int);
 281 extern struct bio *bio_alloc_bioset(unsigned int __nocast, int, struct bio_set *);
 282 extern void bio_put(struct bio *);
 283
 284 extern void bio_endio(struct bio *, unsigned int, int);
 285 struct request_queue;
 286 extern int bio_phys_segments(struct request_queue *, struct bio *);
 287 extern int bio_hw_segments(struct request_queue *, struct bio *);
 288
 289 extern void __bio_clone(struct bio *, struct bio *);
 290 extern struct bio *bio_clone(struct bio *, unsigned int __nocast);
 291
 292 extern void bio_init(struct bio *);
 293
 294 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
 295 extern int bio_get_nr_vecs(struct block_device *);
 296 extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
 297                                 unsigned long, unsigned int, int);
 298 extern void bio_unmap_user(struct bio *);
 299 extern void bio_set_pages_dirty(struct bio *bio);
 300 extern void bio_check_pages_dirty(struct bio *bio);
 301 extern struct bio *bio_copy_user(struct request_queue *, unsigned long, unsigned int, int);
 302 extern int bio_uncopy_user(struct bio *);
 303 void zero_fill_bio(struct bio *bio);
 304
 305 #ifdef CONFIG_HIGHMEM
 306 /*
 307  * remember to add offset! and never ever reenable interrupts between a
 308  * bvec_kmap_irq and bvec_kunmap_irq!!
 309  *
 310  * This function MUST be inlined - it plays with the CPU interrupt flags.
 311  * Hence the `extern inline'.
 312  */
 313 extern inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
 314 {
 315         unsigned long addr;
 316
 317         /*
 318          * might not be a highmem page, but the preempt/irq count
 319          * balancing is a lot nicer this way
 320          */
 321         local_irq_save(*flags);
 322         addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
 323
 324         BUG_ON(addr & ~PAGE_MASK);
 325
 326         return (char *) addr + bvec->bv_offset;
 327 }
 328
 329 extern inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
 330 {
 331         unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
 332
 333         kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
 334         local_irq_restore(*flags);
 335 }
 336
 337 #else
 338 #define bvec_kmap_irq(bvec, flags)      (page_address((bvec)->bv_page) + (bvec)->bv_offset)
 339 #define bvec_kunmap_irq(buf, flags)     do { *(flags) = 0; } while (0)
 340 #endif
 341
 342 extern inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
 343                                    unsigned long *flags)
 344 {
 345         return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
 346 }
 347 #define __bio_kunmap_irq(buf, flags)    bvec_kunmap_irq(buf, flags)
 348
 349 #define bio_kmap_irq(bio, flags) \
 350         __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
 351 #define bio_kunmap_irq(buf,flags)       __bio_kunmap_irq(buf, flags)
 352
 353 #endif /* __LINUX_BIO_H */