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 #ifdef CONFIG_BLOCK
  28
  29 /* 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
  52 #define BIO_MAX_PAGES           256
  53 #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  */
  59 struct bio_vec {
  60         struct page     *bv_page;
  61         unsigned int    bv_len;
  62         unsigned int    bv_offset;
  63 };
  64
  65 struct bio_set;
  66 struct bio;
  67 struct bio_integrity_payload;
  68 typedef void (bio_end_io_t) (struct bio *, int);
  69 typedef void (bio_destructor_t) (struct bio *);
  70
  71 /*
  72  * main unit of I/O for the block layer and lower layers (ie drivers and
  73  * stacking drivers)
  74  */
  75 struct bio {
  76         sector_t                bi_sector;      /* device address in 512 byte
  77                                                    sectors */
  78         struct bio              *bi_next;       /* request queue link */
  79         struct block_device     *bi_bdev;
  80         unsigned long           bi_flags;       /* status, command, etc */
  81         unsigned long           bi_rw;          /* bottom bits READ/WRITE,
  82                                                  * top bits priority
  83                                                  */
  84
  85         unsigned short          bi_vcnt;        /* how many bio_vec's */
  86         unsigned short          bi_idx;         /* current index into bvl_vec */
  87
  88         /* Number of segments in this BIO after
  89          * physical address coalescing is performed.
  90          */
  91         unsigned short          bi_phys_segments;
  92
  93         /* Number of segments after physical and DMA remapping
  94          * hardware coalescing is performed.
  95          */
  96         unsigned short          bi_hw_segments;
  97
  98         unsigned int            bi_size;        /* residual I/O count */
  99
 100         /*
 101          * To keep track of the max hw size, we account for the
 102          * sizes of the first and last virtually mergeable segments
 103          * in this bio
 104          */
 105         unsigned int            bi_hw_front_size;
 106         unsigned int            bi_hw_back_size;
 107
 108         unsigned int            bi_max_vecs;    /* max bvl_vecs we can hold */
 109
 110         struct bio_vec          *bi_io_vec;     /* the actual vec list */
 111
 112         bio_end_io_t            *bi_end_io;
 113         atomic_t                bi_cnt;         /* pin count */
 114
 115         void                    *bi_private;
 116 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 117         struct bio_integrity_payload *bi_integrity;  /* data integrity */
 118 #endif
 119
 120         bio_destructor_t        *bi_destructor; /* destructor */
 121 };
 122
 123 /*
 124  * bio flags
 125  */
 126 #define BIO_UPTODATE    0       /* ok after I/O completion */
 127 #define BIO_RW_BLOCK    1       /* RW_AHEAD set, and read/write would block */
 128 #define BIO_EOF         2       /* out-out-bounds error */
 129 #define BIO_SEG_VALID   3       /* nr_hw_seg valid */
 130 #define BIO_CLONED      4       /* doesn't own data */
 131 #define BIO_BOUNCED     5       /* bio is a bounce bio */
 132 #define BIO_USER_MAPPED 6       /* contains user pages */
 133 #define BIO_EOPNOTSUPP  7       /* not supported */
 134 #define bio_flagged(bio, flag)  ((bio)->bi_flags & (1 << (flag)))
 135
 136 /*
 137  * top 4 bits of bio flags indicate the pool this bio came from
 138  */
 139 #define BIO_POOL_BITS           (4)
 140 #define BIO_POOL_OFFSET         (BITS_PER_LONG - BIO_POOL_BITS)
 141 #define BIO_POOL_MASK           (1UL << BIO_POOL_OFFSET)
 142 #define BIO_POOL_IDX(bio)       ((bio)->bi_flags >> BIO_POOL_OFFSET)
 143
 144 /*
 145  * bio bi_rw flags
 146  *
 147  * bit 0 -- read (not set) or write (set)
 148  * bit 1 -- rw-ahead when set
 149  * bit 2 -- barrier
 150  * bit 3 -- fail fast, don't want low level driver retries
 151  * bit 4 -- synchronous I/O hint: the block layer will unplug immediately
 152  */
 153 #define BIO_RW          0
 154 #define BIO_RW_AHEAD    1
 155 #define BIO_RW_BARRIER  2
 156 #define BIO_RW_FAILFAST 3
 157 #define BIO_RW_SYNC     4
 158 #define BIO_RW_META     5
 159
 160 /*
 161  * upper 16 bits of bi_rw define the io priority of this bio
 162  */
 163 #define BIO_PRIO_SHIFT  (8 * sizeof(unsigned long) - IOPRIO_BITS)
 164 #define bio_prio(bio)   ((bio)->bi_rw >> BIO_PRIO_SHIFT)
 165 #define bio_prio_valid(bio)     ioprio_valid(bio_prio(bio))
 166
 167 #define bio_set_prio(bio, prio)         do {                    \
 168         WARN_ON(prio >= (1 << IOPRIO_BITS));                    \
 169         (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1);          \
 170         (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT);     \
 171 } while (0)
 172
 173 /*
 174  * various member access, note that bio_data should of course not be used
 175  * on highmem page vectors
 176  */
 177 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
 178 #define bio_iovec(bio)          bio_iovec_idx((bio), (bio)->bi_idx)
 179 #define bio_page(bio)           bio_iovec((bio))->bv_page
 180 #define bio_offset(bio)         bio_iovec((bio))->bv_offset
 181 #define bio_segments(bio)       ((bio)->bi_vcnt - (bio)->bi_idx)
 182 #define bio_sectors(bio)        ((bio)->bi_size >> 9)
 183 #define bio_barrier(bio)        ((bio)->bi_rw & (1 << BIO_RW_BARRIER))
 184 #define bio_sync(bio)           ((bio)->bi_rw & (1 << BIO_RW_SYNC))
 185 #define bio_failfast(bio)       ((bio)->bi_rw & (1 << BIO_RW_FAILFAST))
 186 #define bio_rw_ahead(bio)       ((bio)->bi_rw & (1 << BIO_RW_AHEAD))
 187 #define bio_rw_meta(bio)        ((bio)->bi_rw & (1 << BIO_RW_META))
 188 #define bio_empty_barrier(bio)  (bio_barrier(bio) && !(bio)->bi_size)
 189
 190 static inline unsigned int bio_cur_sectors(struct bio *bio)
 191 {
 192         if (bio->bi_vcnt)
 193                 return bio_iovec(bio)->bv_len >> 9;
 194
 195         return 0;
 196 }
 197
 198 static inline void *bio_data(struct bio *bio)
 199 {
 200         if (bio->bi_vcnt)
 201                 return page_address(bio_page(bio)) + bio_offset(bio);
 202
 203         return NULL;
 204 }
 205
 206 /*
 207  * will die
 208  */
 209 #define bio_to_phys(bio)        (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
 210 #define bvec_to_phys(bv)        (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
 211
 212 /*
 213  * queues that have highmem support enabled may still need to revert to
 214  * PIO transfers occasionally and thus map high pages temporarily. For
 215  * permanent PIO fall back, user is probably better off disabling highmem
 216  * I/O completely on that queue (see ide-dma for example)
 217  */
 218 #define __bio_kmap_atomic(bio, idx, kmtype)                             \
 219         (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) +    \
 220                 bio_iovec_idx((bio), (idx))->bv_offset)
 221
 222 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
 223
 224 /*
 225  * merge helpers etc
 226  */
 227
 228 #define __BVEC_END(bio)         bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
 229 #define __BVEC_START(bio)       bio_iovec_idx((bio), (bio)->bi_idx)
 230
 231 /*
 232  * allow arch override, for eg virtualized architectures (put in asm/io.h)
 233  */
 234 #ifndef BIOVEC_PHYS_MERGEABLE
 235 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2)       \
 236         ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
 237 #endif
 238
 239 #define BIOVEC_VIRT_MERGEABLE(vec1, vec2)       \
 240         ((((bvec_to_phys((vec1)) + (vec1)->bv_len) | bvec_to_phys((vec2))) & (BIO_VMERGE_BOUNDARY - 1)) == 0)
 241 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
 242         (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
 243 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
 244         __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask)
 245 #define BIO_SEG_BOUNDARY(q, b1, b2) \
 246         BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
 247
 248 #define bio_io_error(bio) bio_endio((bio), -EIO)
 249
 250 /*
 251  * drivers should not use the __ version unless they _really_ want to
 252  * run through the entire bio and not just pending pieces
 253  */
 254 #define __bio_for_each_segment(bvl, bio, i, start_idx)                  \
 255         for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx);  \
 256              i < (bio)->bi_vcnt;                                        \
 257              bvl++, i++)
 258
 259 #define bio_for_each_segment(bvl, bio, i)                               \
 260         __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
 261
 262 /*
 263  * get a reference to a bio, so it won't disappear. the intended use is
 264  * something like:
 265  *
 266  * bio_get(bio);
 267  * submit_bio(rw, bio);
 268  * if (bio->bi_flags ...)
 269  *      do_something
 270  * bio_put(bio);
 271  *
 272  * without the bio_get(), it could potentially complete I/O before submit_bio
 273  * returns. and then bio would be freed memory when if (bio->bi_flags ...)
 274  * runs
 275  */
 276 #define bio_get(bio)    atomic_inc(&(bio)->bi_cnt)
 277
 278 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 279 /*
 280  * bio integrity payload
 281  */
 282 struct bio_integrity_payload {
 283         struct bio              *bip_bio;       /* parent bio */
 284         struct bio_vec          *bip_vec;       /* integrity data vector */
 285
 286         sector_t                bip_sector;     /* virtual start sector */
 287
 288         void                    *bip_buf;       /* generated integrity data */
 289         bio_end_io_t            *bip_end_io;    /* saved I/O completion fn */
 290
 291         int                     bip_error;      /* saved I/O error */
 292         unsigned int            bip_size;
 293
 294         unsigned short          bip_pool;       /* pool the ivec came from */
 295         unsigned short          bip_vcnt;       /* # of integrity bio_vecs */
 296         unsigned short          bip_idx;        /* current bip_vec index */
 297
 298         struct work_struct      bip_work;       /* I/O completion */
 299 };
 300 #endif /* CONFIG_BLK_DEV_INTEGRITY */
 301
 302 /*
 303  * A bio_pair is used when we need to split a bio.
 304  * This can only happen for a bio that refers to just one
 305  * page of data, and in the unusual situation when the
 306  * page crosses a chunk/device boundary
 307  *
 308  * The address of the master bio is stored in bio1.bi_private
 309  * The address of the pool the pair was allocated from is stored
 310  *   in bio2.bi_private
 311  */
 312 struct bio_pair {
 313         struct bio                      bio1, bio2;
 314         struct bio_vec                  bv1, bv2;
 315 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 316         struct bio_integrity_payload    bip1, bip2;
 317         struct bio_vec                  iv1, iv2;
 318 #endif
 319         atomic_t                        cnt;
 320         int                             error;
 321 };
 322 extern struct bio_pair *bio_split(struct bio *bi, mempool_t *pool,
 323                                   int first_sectors);
 324 extern mempool_t *bio_split_pool;
 325 extern void bio_pair_release(struct bio_pair *dbio);
 326
 327 extern struct bio_set *bioset_create(int, int);
 328 extern void bioset_free(struct bio_set *);
 329
 330 extern struct bio *bio_alloc(gfp_t, int);
 331 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
 332 extern void bio_put(struct bio *);
 333 extern void bio_free(struct bio *, struct bio_set *);
 334
 335 extern void bio_endio(struct bio *, int);
 336 struct request_queue;
 337 extern int bio_phys_segments(struct request_queue *, struct bio *);
 338 extern int bio_hw_segments(struct request_queue *, struct bio *);
 339
 340 extern void __bio_clone(struct bio *, struct bio *);
 341 extern struct bio *bio_clone(struct bio *, gfp_t);
 342
 343 extern void bio_init(struct bio *);
 344
 345 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
 346 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
 347                            unsigned int, unsigned int);
 348 extern int bio_get_nr_vecs(struct block_device *);
 349 extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
 350                                 unsigned long, unsigned int, int);
 351 struct sg_iovec;
 352 extern struct bio *bio_map_user_iov(struct request_queue *,
 353                                     struct block_device *,
 354                                     struct sg_iovec *, int, int);
 355 extern void bio_unmap_user(struct bio *);
 356 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
 357                                 gfp_t);
 358 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
 359                                  gfp_t, int);
 360 extern void bio_set_pages_dirty(struct bio *bio);
 361 extern void bio_check_pages_dirty(struct bio *bio);
 362 extern struct bio *bio_copy_user(struct request_queue *, unsigned long, unsigned int, int);
 363 extern struct bio *bio_copy_user_iov(struct request_queue *, struct sg_iovec *,
 364                                      int, int);
 365 extern int bio_uncopy_user(struct bio *);
 366 void zero_fill_bio(struct bio *bio);
 367 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *);
 368 extern unsigned int bvec_nr_vecs(unsigned short idx);
 369
 370 /*
 371  * bio_set is used to allow other portions of the IO system to
 372  * allocate their own private memory pools for bio and iovec structures.
 373  * These memory pools in turn all allocate from the bio_slab
 374  * and the bvec_slabs[].
 375  */
 376 #define BIO_POOL_SIZE 2
 377 #define BIOVEC_NR_POOLS 6
 378
 379 struct bio_set {
 380         mempool_t *bio_pool;
 381 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 382         mempool_t *bio_integrity_pool;
 383 #endif
 384         mempool_t *bvec_pools[BIOVEC_NR_POOLS];
 385 };
 386
 387 struct biovec_slab {
 388         int nr_vecs;
 389         char *name;
 390         struct kmem_cache *slab;
 391 };
 392
 393 extern struct bio_set *fs_bio_set;
 394
 395 /*
 396  * a small number of entries is fine, not going to be performance critical.
 397  * basically we just need to survive
 398  */
 399 #define BIO_SPLIT_ENTRIES 2
 400
 401 #ifdef CONFIG_HIGHMEM
 402 /*
 403  * remember to add offset! and never ever reenable interrupts between a
 404  * bvec_kmap_irq and bvec_kunmap_irq!!
 405  *
 406  * This function MUST be inlined - it plays with the CPU interrupt flags.
 407  */
 408 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
 409 {
 410         unsigned long addr;
 411
 412         /*
 413          * might not be a highmem page, but the preempt/irq count
 414          * balancing is a lot nicer this way
 415          */
 416         local_irq_save(*flags);
 417         addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
 418
 419         BUG_ON(addr & ~PAGE_MASK);
 420
 421         return (char *) addr + bvec->bv_offset;
 422 }
 423
 424 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
 425 {
 426         unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
 427
 428         kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
 429         local_irq_restore(*flags);
 430 }
 431
 432 #else
 433 #define bvec_kmap_irq(bvec, flags)      (page_address((bvec)->bv_page) + (bvec)->bv_offset)
 434 #define bvec_kunmap_irq(buf, flags)     do { *(flags) = 0; } while (0)
 435 #endif
 436
 437 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
 438                                    unsigned long *flags)
 439 {
 440         return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
 441 }
 442 #define __bio_kunmap_irq(buf, flags)    bvec_kunmap_irq(buf, flags)
 443
 444 #define bio_kmap_irq(bio, flags) \
 445         __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
 446 #define bio_kunmap_irq(buf,flags)       __bio_kunmap_irq(buf, flags)
 447
 448 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 449
 450 #define bip_vec_idx(bip, idx)   (&(bip->bip_vec[(idx)]))
 451 #define bip_vec(bip)            bip_vec_idx(bip, 0)
 452
 453 #define __bip_for_each_vec(bvl, bip, i, start_idx)                      \
 454         for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx);    \
 455              i < (bip)->bip_vcnt;                                       \
 456              bvl++, i++)
 457
 458 #define bip_for_each_vec(bvl, bip, i)                                   \
 459         __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx)
 460
 461 static inline int bio_integrity(struct bio *bio)
 462 {
 463 #if defined(CONFIG_BLK_DEV_INTEGRITY)
 464         return bio->bi_integrity != NULL;
 465 #else
 466         return 0;
 467 #endif
 468 }
 469
 470 extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *);
 471 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
 472 extern void bio_integrity_free(struct bio *, struct bio_set *);
 473 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
 474 extern int bio_integrity_enabled(struct bio *bio);
 475 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int);
 476 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int);
 477 extern int bio_integrity_prep(struct bio *);
 478 extern void bio_integrity_endio(struct bio *, int);
 479 extern void bio_integrity_advance(struct bio *, unsigned int);
 480 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int);
 481 extern void bio_integrity_split(struct bio *, struct bio_pair *, int);
 482 extern int bio_integrity_clone(struct bio *, struct bio *, struct bio_set *);
 483 extern int bioset_integrity_create(struct bio_set *, int);
 484 extern void bioset_integrity_free(struct bio_set *);
 485 extern void bio_integrity_init_slab(void);
 486
 487 #else /* CONFIG_BLK_DEV_INTEGRITY */
 488
 489 #define bio_integrity(a)                (0)
 490 #define bioset_integrity_create(a, b)   (0)
 491 #define bio_integrity_prep(a)           (0)
 492 #define bio_integrity_enabled(a)        (0)
 493 #define bio_integrity_clone(a, b, c)    (0)
 494 #define bioset_integrity_free(a)        do { } while (0)
 495 #define bio_integrity_free(a, b)        do { } while (0)
 496 #define bio_integrity_endio(a, b)       do { } while (0)
 497 #define bio_integrity_advance(a, b)     do { } while (0)
 498 #define bio_integrity_trim(a, b, c)     do { } while (0)
 499 #define bio_integrity_split(a, b, c)    do { } while (0)
 500 #define bio_integrity_set_tag(a, b, c)  do { } while (0)
 501 #define bio_integrity_get_tag(a, b, c)  do { } while (0)
 502 #define bio_integrity_init_slab(a)      do { } while (0)
 503
 504 #endif /* CONFIG_BLK_DEV_INTEGRITY */
 505
 506 #endif /* CONFIG_BLOCK */
 507 #endif /* __LINUX_BIO_H */