1 /* 2 * 2.5 block I/O model 3 * 4 * Copyright (C) 2001 Jens Axboe <[email protected]> 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 #include <asm/io.h> 30 31 #define BIO_DEBUG 32 33 #ifdef BIO_DEBUG 34 #define BIO_BUG_ON BUG_ON 35 #else 36 #define BIO_BUG_ON 37 #endif 38 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) 42 43 /* 44 * was unsigned short, but we might as well be ready for > 64kB I/O pages 45 */ 46 struct bio_vec { 47 struct page *bv_page; 48 unsigned int bv_len; 49 unsigned int bv_offset; 50 }; 51 52 struct bio_set; 53 struct bio; 54 struct bio_integrity_payload; 55 typedef void (bio_end_io_t) (struct bio *, int); 56 typedef void (bio_destructor_t) (struct bio *); 57 58 /* 59 * main unit of I/O for the block layer and lower layers (ie drivers and 60 * stacking drivers) 61 */ 62 struct bio { 63 sector_t bi_sector; /* device address in 512 byte 64 sectors */ 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, 69 * top bits priority 70 */ 71 72 unsigned short bi_vcnt; /* how many bio_vec's */ 73 unsigned short bi_idx; /* current index into bvl_vec */ 74 75 /* Number of segments in this BIO after 76 * physical address coalescing is performed. 77 */ 78 unsigned int bi_phys_segments; 79 80 unsigned int bi_size; /* residual I/O count */ 81 82 unsigned int bi_max_vecs; /* max bvl_vecs we can hold */ 83 84 unsigned int bi_comp_cpu; /* completion CPU */ 85 86 struct bio_vec *bi_io_vec; /* the actual vec list */ 87 88 bio_end_io_t *bi_end_io; 89 atomic_t bi_cnt; /* pin count */ 90 91 void *bi_private; 92 #if defined(CONFIG_BLK_DEV_INTEGRITY) 93 struct bio_integrity_payload *bi_integrity; /* data integrity */ 94 #endif 95 96 bio_destructor_t *bi_destructor; /* destructor */ 97 }; 98 99 /* 100 * bio flags 101 */ 102 #define BIO_UPTODATE 0 /* ok after I/O completion */ 103 #define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */ 104 #define BIO_EOF 2 /* out-out-bounds error */ 105 #define BIO_SEG_VALID 3 /* bi_phys_segments valid */ 106 #define BIO_CLONED 4 /* doesn't own data */ 107 #define BIO_BOUNCED 5 /* bio is a bounce bio */ 108 #define BIO_USER_MAPPED 6 /* contains user pages */ 109 #define BIO_EOPNOTSUPP 7 /* not supported */ 110 #define BIO_CPU_AFFINE 8 /* complete bio on same CPU as submitted */ 111 #define BIO_NULL_MAPPED 9 /* contains invalid user pages */ 112 #define BIO_FS_INTEGRITY 10 /* fs owns integrity data, not block layer */ 113 #define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag))) 114 115 /* 116 * top 4 bits of bio flags indicate the pool this bio came from 117 */ 118 #define BIO_POOL_BITS (4) 119 #define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS) 120 #define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET) 121 #define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET) 122 123 /* 124 * bio bi_rw flags 125 * 126 * bit 0 -- data direction 127 * If not set, bio is a read from device. If set, it's a write to device. 128 * bit 1 -- rw-ahead when set 129 * bit 2 -- barrier 130 * Insert a serialization point in the IO queue, forcing previously 131 * submitted IO to be completed before this oen is issued. 132 * bit 3 -- fail fast, don't want low level driver retries 133 * bit 4 -- synchronous I/O hint: the block layer will unplug immediately 134 * Note that this does NOT indicate that the IO itself is sync, just 135 * that the block layer will not postpone issue of this IO by plugging. 136 * bit 5 -- metadata request 137 * Used for tracing to differentiate metadata and data IO. May also 138 * get some preferential treatment in the IO scheduler 139 * bit 6 -- discard sectors 140 * Informs the lower level device that this range of sectors is no longer 141 * used by the file system and may thus be freed by the device. Used 142 * for flash based storage. 143 */ 144 #define BIO_RW 0 /* Must match RW in req flags (blkdev.h) */ 145 #define BIO_RW_AHEAD 1 /* Must match FAILFAST in req flags */ 146 #define BIO_RW_BARRIER 2 147 #define BIO_RW_FAILFAST 3 148 #define BIO_RW_SYNC 4 149 #define BIO_RW_META 5 150 #define BIO_RW_DISCARD 6 151 152 /* 153 * upper 16 bits of bi_rw define the io priority of this bio 154 */ 155 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS) 156 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT) 157 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio)) 158 159 #define bio_set_prio(bio, prio) do { \ 160 WARN_ON(prio >= (1 << IOPRIO_BITS)); \ 161 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \ 162 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \ 163 } while (0) 164 165 /* 166 * various member access, note that bio_data should of course not be used 167 * on highmem page vectors 168 */ 169 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)])) 170 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx) 171 #define bio_page(bio) bio_iovec((bio))->bv_page 172 #define bio_offset(bio) bio_iovec((bio))->bv_offset 173 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx) 174 #define bio_sectors(bio) ((bio)->bi_size >> 9) 175 #define bio_barrier(bio) ((bio)->bi_rw & (1 << BIO_RW_BARRIER)) 176 #define bio_sync(bio) ((bio)->bi_rw & (1 << BIO_RW_SYNC)) 177 #define bio_failfast(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST)) 178 #define bio_rw_ahead(bio) ((bio)->bi_rw & (1 << BIO_RW_AHEAD)) 179 #define bio_rw_meta(bio) ((bio)->bi_rw & (1 << BIO_RW_META)) 180 #define bio_discard(bio) ((bio)->bi_rw & (1 << BIO_RW_DISCARD)) 181 #define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio) && !bio_discard(bio)) 182 183 static inline unsigned int bio_cur_sectors(struct bio *bio) 184 { 185 if (bio->bi_vcnt) 186 return bio_iovec(bio)->bv_len >> 9; 187 else /* dataless requests such as discard */ 188 return bio->bi_size >> 9; 189 } 190 191 static inline void *bio_data(struct bio *bio) 192 { 193 if (bio->bi_vcnt) 194 return page_address(bio_page(bio)) + bio_offset(bio); 195 196 return NULL; 197 } 198 199 /* 200 * will die 201 */ 202 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio))) 203 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset) 204 205 /* 206 * queues that have highmem support enabled may still need to revert to 207 * PIO transfers occasionally and thus map high pages temporarily. For 208 * permanent PIO fall back, user is probably better off disabling highmem 209 * I/O completely on that queue (see ide-dma for example) 210 */ 211 #define __bio_kmap_atomic(bio, idx, kmtype) \ 212 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \ 213 bio_iovec_idx((bio), (idx))->bv_offset) 214 215 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype) 216 217 /* 218 * merge helpers etc 219 */ 220 221 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1) 222 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx) 223 224 /* 225 * allow arch override, for eg virtualized architectures (put in asm/io.h) 226 */ 227 #ifndef BIOVEC_PHYS_MERGEABLE 228 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 229 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2))) 230 #endif 231 232 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \ 233 (((addr1) | (mask)) == (((addr2) - 1) | (mask))) 234 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \ 235 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask) 236 #define BIO_SEG_BOUNDARY(q, b1, b2) \ 237 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2))) 238 239 #define bio_io_error(bio) bio_endio((bio), -EIO) 240 241 /* 242 * drivers should not use the __ version unless they _really_ want to 243 * run through the entire bio and not just pending pieces 244 */ 245 #define __bio_for_each_segment(bvl, bio, i, start_idx) \ 246 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \ 247 i < (bio)->bi_vcnt; \ 248 bvl++, i++) 249 250 #define bio_for_each_segment(bvl, bio, i) \ 251 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx) 252 253 /* 254 * get a reference to a bio, so it won't disappear. the intended use is 255 * something like: 256 * 257 * bio_get(bio); 258 * submit_bio(rw, bio); 259 * if (bio->bi_flags ...) 260 * do_something 261 * bio_put(bio); 262 * 263 * without the bio_get(), it could potentially complete I/O before submit_bio 264 * returns. and then bio would be freed memory when if (bio->bi_flags ...) 265 * runs 266 */ 267 #define bio_get(bio) atomic_inc(&(bio)->bi_cnt) 268 269 #if defined(CONFIG_BLK_DEV_INTEGRITY) 270 /* 271 * bio integrity payload 272 */ 273 struct bio_integrity_payload { 274 struct bio *bip_bio; /* parent bio */ 275 struct bio_vec *bip_vec; /* integrity data vector */ 276 277 sector_t bip_sector; /* virtual start sector */ 278 279 void *bip_buf; /* generated integrity data */ 280 bio_end_io_t *bip_end_io; /* saved I/O completion fn */ 281 282 int bip_error; /* saved I/O error */ 283 unsigned int bip_size; 284 285 unsigned short bip_pool; /* pool the ivec came from */ 286 unsigned short bip_vcnt; /* # of integrity bio_vecs */ 287 unsigned short bip_idx; /* current bip_vec index */ 288 289 struct work_struct bip_work; /* I/O completion */ 290 }; 291 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 292 293 /* 294 * A bio_pair is used when we need to split a bio. 295 * This can only happen for a bio that refers to just one 296 * page of data, and in the unusual situation when the 297 * page crosses a chunk/device boundary 298 * 299 * The address of the master bio is stored in bio1.bi_private 300 * The address of the pool the pair was allocated from is stored 301 * in bio2.bi_private 302 */ 303 struct bio_pair { 304 struct bio bio1, bio2; 305 struct bio_vec bv1, bv2; 306 #if defined(CONFIG_BLK_DEV_INTEGRITY) 307 struct bio_integrity_payload bip1, bip2; 308 struct bio_vec iv1, iv2; 309 #endif 310 atomic_t cnt; 311 int error; 312 }; 313 extern struct bio_pair *bio_split(struct bio *bi, int first_sectors); 314 extern void bio_pair_release(struct bio_pair *dbio); 315 316 extern struct bio_set *bioset_create(int, int); 317 extern void bioset_free(struct bio_set *); 318 319 extern struct bio *bio_alloc(gfp_t, int); 320 extern struct bio *bio_kmalloc(gfp_t, int); 321 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *); 322 extern void bio_put(struct bio *); 323 extern void bio_free(struct bio *, struct bio_set *); 324 325 extern void bio_endio(struct bio *, int); 326 struct request_queue; 327 extern int bio_phys_segments(struct request_queue *, struct bio *); 328 329 extern void __bio_clone(struct bio *, struct bio *); 330 extern struct bio *bio_clone(struct bio *, gfp_t); 331 332 extern void bio_init(struct bio *); 333 334 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int); 335 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *, 336 unsigned int, unsigned int); 337 extern int bio_get_nr_vecs(struct block_device *); 338 extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int); 339 extern struct bio *bio_map_user(struct request_queue *, struct block_device *, 340 unsigned long, unsigned int, int, gfp_t); 341 struct sg_iovec; 342 struct rq_map_data; 343 extern struct bio *bio_map_user_iov(struct request_queue *, 344 struct block_device *, 345 struct sg_iovec *, int, int, gfp_t); 346 extern void bio_unmap_user(struct bio *); 347 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int, 348 gfp_t); 349 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int, 350 gfp_t, int); 351 extern void bio_set_pages_dirty(struct bio *bio); 352 extern void bio_check_pages_dirty(struct bio *bio); 353 extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *, 354 unsigned long, unsigned int, int, gfp_t); 355 extern struct bio *bio_copy_user_iov(struct request_queue *, 356 struct rq_map_data *, struct sg_iovec *, 357 int, int, gfp_t); 358 extern int bio_uncopy_user(struct bio *); 359 void zero_fill_bio(struct bio *bio); 360 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *); 361 extern unsigned int bvec_nr_vecs(unsigned short idx); 362 363 /* 364 * Allow queuer to specify a completion CPU for this bio 365 */ 366 static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu) 367 { 368 bio->bi_comp_cpu = cpu; 369 } 370 371 /* 372 * bio_set is used to allow other portions of the IO system to 373 * allocate their own private memory pools for bio and iovec structures. 374 * These memory pools in turn all allocate from the bio_slab 375 * and the bvec_slabs[]. 376 */ 377 #define BIO_POOL_SIZE 2 378 #define BIOVEC_NR_POOLS 6 379 380 struct bio_set { 381 mempool_t *bio_pool; 382 #if defined(CONFIG_BLK_DEV_INTEGRITY) 383 mempool_t *bio_integrity_pool; 384 #endif 385 mempool_t *bvec_pools[BIOVEC_NR_POOLS]; 386 }; 387 388 struct biovec_slab { 389 int nr_vecs; 390 char *name; 391 struct kmem_cache *slab; 392 }; 393 394 extern struct bio_set *fs_bio_set; 395 396 /* 397 * a small number of entries is fine, not going to be performance critical. 398 * basically we just need to survive 399 */ 400 #define BIO_SPLIT_ENTRIES 2 401 402 #ifdef CONFIG_HIGHMEM 403 /* 404 * remember to add offset! and never ever reenable interrupts between a 405 * bvec_kmap_irq and bvec_kunmap_irq!! 406 * 407 * This function MUST be inlined - it plays with the CPU interrupt flags. 408 */ 409 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags) 410 { 411 unsigned long addr; 412 413 /* 414 * might not be a highmem page, but the preempt/irq count 415 * balancing is a lot nicer this way 416 */ 417 local_irq_save(*flags); 418 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ); 419 420 BUG_ON(addr & ~PAGE_MASK); 421 422 return (char *) addr + bvec->bv_offset; 423 } 424 425 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags) 426 { 427 unsigned long ptr = (unsigned long) buffer & PAGE_MASK; 428 429 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ); 430 local_irq_restore(*flags); 431 } 432 433 #else 434 #define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset) 435 #define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0) 436 #endif 437 438 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx, 439 unsigned long *flags) 440 { 441 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags); 442 } 443 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags) 444 445 #define bio_kmap_irq(bio, flags) \ 446 __bio_kmap_irq((bio), (bio)->bi_idx, (flags)) 447 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags) 448 449 /* 450 * Check whether this bio carries any data or not. A NULL bio is allowed. 451 */ 452 static inline int bio_has_data(struct bio *bio) 453 { 454 return bio && bio->bi_io_vec != NULL; 455 } 456 457 #if defined(CONFIG_BLK_DEV_INTEGRITY) 458 459 #define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)])) 460 #define bip_vec(bip) bip_vec_idx(bip, 0) 461 462 #define __bip_for_each_vec(bvl, bip, i, start_idx) \ 463 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \ 464 i < (bip)->bip_vcnt; \ 465 bvl++, i++) 466 467 #define bip_for_each_vec(bvl, bip, i) \ 468 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx) 469 470 #define bio_integrity(bio) (bio->bi_integrity != NULL) 471 472 extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *); 473 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int); 474 extern void bio_integrity_free(struct bio *, struct bio_set *); 475 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int); 476 extern int bio_integrity_enabled(struct bio *bio); 477 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int); 478 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int); 479 extern int bio_integrity_prep(struct bio *); 480 extern void bio_integrity_endio(struct bio *, int); 481 extern void bio_integrity_advance(struct bio *, unsigned int); 482 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int); 483 extern void bio_integrity_split(struct bio *, struct bio_pair *, int); 484 extern int bio_integrity_clone(struct bio *, struct bio *, struct bio_set *); 485 extern int bioset_integrity_create(struct bio_set *, int); 486 extern void bioset_integrity_free(struct bio_set *); 487 extern void bio_integrity_init_slab(void); 488 489 #else /* CONFIG_BLK_DEV_INTEGRITY */ 490 491 #define bio_integrity(a) (0) 492 #define bioset_integrity_create(a, b) (0) 493 #define bio_integrity_prep(a) (0) 494 #define bio_integrity_enabled(a) (0) 495 #define bio_integrity_clone(a, b, c) (0) 496 #define bioset_integrity_free(a) do { } while (0) 497 #define bio_integrity_free(a, b) do { } while (0) 498 #define bio_integrity_endio(a, b) do { } while (0) 499 #define bio_integrity_advance(a, b) do { } while (0) 500 #define bio_integrity_trim(a, b, c) do { } while (0) 501 #define bio_integrity_split(a, b, c) do { } while (0) 502 #define bio_integrity_set_tag(a, b, c) do { } while (0) 503 #define bio_integrity_get_tag(a, b, c) do { } while (0) 504 #define bio_integrity_init_slab(a) do { } while (0) 505 506 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 507 508 #endif /* CONFIG_BLOCK */ 509 #endif /* __LINUX_BIO_H */ 510