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 /* 83 * To keep track of the max segment size, we account for the 84 * sizes of the first and last mergeable segments in this bio. 85 */ 86 unsigned int bi_seg_front_size; 87 unsigned int bi_seg_back_size; 88 89 unsigned int bi_max_vecs; /* max bvl_vecs we can hold */ 90 91 unsigned int bi_comp_cpu; /* completion CPU */ 92 93 struct bio_vec *bi_io_vec; /* the actual vec list */ 94 95 bio_end_io_t *bi_end_io; 96 atomic_t bi_cnt; /* pin count */ 97 98 void *bi_private; 99 #if defined(CONFIG_BLK_DEV_INTEGRITY) 100 struct bio_integrity_payload *bi_integrity; /* data integrity */ 101 #endif 102 103 bio_destructor_t *bi_destructor; /* destructor */ 104 }; 105 106 /* 107 * bio flags 108 */ 109 #define BIO_UPTODATE 0 /* ok after I/O completion */ 110 #define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */ 111 #define BIO_EOF 2 /* out-out-bounds error */ 112 #define BIO_SEG_VALID 3 /* bi_phys_segments valid */ 113 #define BIO_CLONED 4 /* doesn't own data */ 114 #define BIO_BOUNCED 5 /* bio is a bounce bio */ 115 #define BIO_USER_MAPPED 6 /* contains user pages */ 116 #define BIO_EOPNOTSUPP 7 /* not supported */ 117 #define BIO_CPU_AFFINE 8 /* complete bio on same CPU as submitted */ 118 #define BIO_NULL_MAPPED 9 /* contains invalid user pages */ 119 #define BIO_FS_INTEGRITY 10 /* fs owns integrity data, not block layer */ 120 #define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag))) 121 122 /* 123 * top 4 bits of bio flags indicate the pool this bio came from 124 */ 125 #define BIO_POOL_BITS (4) 126 #define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS) 127 #define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET) 128 #define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET) 129 130 /* 131 * bio bi_rw flags 132 * 133 * bit 0 -- data direction 134 * If not set, bio is a read from device. If set, it's a write to device. 135 * bit 1 -- rw-ahead when set 136 * bit 2 -- barrier 137 * Insert a serialization point in the IO queue, forcing previously 138 * submitted IO to be completed before this oen is issued. 139 * bit 3 -- synchronous I/O hint: the block layer will unplug immediately 140 * Note that this does NOT indicate that the IO itself is sync, just 141 * that the block layer will not postpone issue of this IO by plugging. 142 * bit 4 -- metadata request 143 * Used for tracing to differentiate metadata and data IO. May also 144 * get some preferential treatment in the IO scheduler 145 * bit 5 -- discard sectors 146 * Informs the lower level device that this range of sectors is no longer 147 * used by the file system and may thus be freed by the device. Used 148 * for flash based storage. 149 * bit 6 -- fail fast device errors 150 * bit 7 -- fail fast transport errors 151 * bit 8 -- fail fast driver errors 152 * Don't want driver retries for any fast fail whatever the reason. 153 */ 154 #define BIO_RW 0 /* Must match RW in req flags (blkdev.h) */ 155 #define BIO_RW_AHEAD 1 /* Must match FAILFAST in req flags */ 156 #define BIO_RW_BARRIER 2 157 #define BIO_RW_SYNC 3 158 #define BIO_RW_META 4 159 #define BIO_RW_DISCARD 5 160 #define BIO_RW_FAILFAST_DEV 6 161 #define BIO_RW_FAILFAST_TRANSPORT 7 162 #define BIO_RW_FAILFAST_DRIVER 8 163 164 /* 165 * upper 16 bits of bi_rw define the io priority of this bio 166 */ 167 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS) 168 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT) 169 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio)) 170 171 #define bio_set_prio(bio, prio) do { \ 172 WARN_ON(prio >= (1 << IOPRIO_BITS)); \ 173 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \ 174 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \ 175 } while (0) 176 177 /* 178 * various member access, note that bio_data should of course not be used 179 * on highmem page vectors 180 */ 181 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)])) 182 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx) 183 #define bio_page(bio) bio_iovec((bio))->bv_page 184 #define bio_offset(bio) bio_iovec((bio))->bv_offset 185 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx) 186 #define bio_sectors(bio) ((bio)->bi_size >> 9) 187 #define bio_barrier(bio) ((bio)->bi_rw & (1 << BIO_RW_BARRIER)) 188 #define bio_sync(bio) ((bio)->bi_rw & (1 << BIO_RW_SYNC)) 189 #define bio_failfast_dev(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST_DEV)) 190 #define bio_failfast_transport(bio) \ 191 ((bio)->bi_rw & (1 << BIO_RW_FAILFAST_TRANSPORT)) 192 #define bio_failfast_driver(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST_DRIVER)) 193 #define bio_rw_ahead(bio) ((bio)->bi_rw & (1 << BIO_RW_AHEAD)) 194 #define bio_rw_meta(bio) ((bio)->bi_rw & (1 << BIO_RW_META)) 195 #define bio_discard(bio) ((bio)->bi_rw & (1 << BIO_RW_DISCARD)) 196 #define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio) && !bio_discard(bio)) 197 198 static inline unsigned int bio_cur_sectors(struct bio *bio) 199 { 200 if (bio->bi_vcnt) 201 return bio_iovec(bio)->bv_len >> 9; 202 else /* dataless requests such as discard */ 203 return bio->bi_size >> 9; 204 } 205 206 static inline void *bio_data(struct bio *bio) 207 { 208 if (bio->bi_vcnt) 209 return page_address(bio_page(bio)) + bio_offset(bio); 210 211 return NULL; 212 } 213 214 /* 215 * will die 216 */ 217 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio))) 218 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset) 219 220 /* 221 * queues that have highmem support enabled may still need to revert to 222 * PIO transfers occasionally and thus map high pages temporarily. For 223 * permanent PIO fall back, user is probably better off disabling highmem 224 * I/O completely on that queue (see ide-dma for example) 225 */ 226 #define __bio_kmap_atomic(bio, idx, kmtype) \ 227 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \ 228 bio_iovec_idx((bio), (idx))->bv_offset) 229 230 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype) 231 232 /* 233 * merge helpers etc 234 */ 235 236 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1) 237 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx) 238 239 /* Default implementation of BIOVEC_PHYS_MERGEABLE */ 240 #define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 241 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2))) 242 243 /* 244 * allow arch override, for eg virtualized architectures (put in asm/io.h) 245 */ 246 #ifndef BIOVEC_PHYS_MERGEABLE 247 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 248 __BIOVEC_PHYS_MERGEABLE(vec1, vec2) 249 #endif 250 251 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \ 252 (((addr1) | (mask)) == (((addr2) - 1) | (mask))) 253 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \ 254 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask) 255 #define BIO_SEG_BOUNDARY(q, b1, b2) \ 256 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2))) 257 258 #define bio_io_error(bio) bio_endio((bio), -EIO) 259 260 /* 261 * drivers should not use the __ version unless they _really_ want to 262 * run through the entire bio and not just pending pieces 263 */ 264 #define __bio_for_each_segment(bvl, bio, i, start_idx) \ 265 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \ 266 i < (bio)->bi_vcnt; \ 267 bvl++, i++) 268 269 #define bio_for_each_segment(bvl, bio, i) \ 270 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx) 271 272 /* 273 * get a reference to a bio, so it won't disappear. the intended use is 274 * something like: 275 * 276 * bio_get(bio); 277 * submit_bio(rw, bio); 278 * if (bio->bi_flags ...) 279 * do_something 280 * bio_put(bio); 281 * 282 * without the bio_get(), it could potentially complete I/O before submit_bio 283 * returns. and then bio would be freed memory when if (bio->bi_flags ...) 284 * runs 285 */ 286 #define bio_get(bio) atomic_inc(&(bio)->bi_cnt) 287 288 #if defined(CONFIG_BLK_DEV_INTEGRITY) 289 /* 290 * bio integrity payload 291 */ 292 struct bio_integrity_payload { 293 struct bio *bip_bio; /* parent bio */ 294 struct bio_vec *bip_vec; /* integrity data vector */ 295 296 sector_t bip_sector; /* virtual start sector */ 297 298 void *bip_buf; /* generated integrity data */ 299 bio_end_io_t *bip_end_io; /* saved I/O completion fn */ 300 301 int bip_error; /* saved I/O error */ 302 unsigned int bip_size; 303 304 unsigned short bip_pool; /* pool the ivec came from */ 305 unsigned short bip_vcnt; /* # of integrity bio_vecs */ 306 unsigned short bip_idx; /* current bip_vec index */ 307 308 struct work_struct bip_work; /* I/O completion */ 309 }; 310 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 311 312 /* 313 * A bio_pair is used when we need to split a bio. 314 * This can only happen for a bio that refers to just one 315 * page of data, and in the unusual situation when the 316 * page crosses a chunk/device boundary 317 * 318 * The address of the master bio is stored in bio1.bi_private 319 * The address of the pool the pair was allocated from is stored 320 * in bio2.bi_private 321 */ 322 struct bio_pair { 323 struct bio bio1, bio2; 324 struct bio_vec bv1, bv2; 325 #if defined(CONFIG_BLK_DEV_INTEGRITY) 326 struct bio_integrity_payload bip1, bip2; 327 struct bio_vec iv1, iv2; 328 #endif 329 atomic_t cnt; 330 int error; 331 }; 332 extern struct bio_pair *bio_split(struct bio *bi, int first_sectors); 333 extern void bio_pair_release(struct bio_pair *dbio); 334 335 extern struct bio_set *bioset_create(int, int); 336 extern void bioset_free(struct bio_set *); 337 338 extern struct bio *bio_alloc(gfp_t, int); 339 extern struct bio *bio_kmalloc(gfp_t, int); 340 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *); 341 extern void bio_put(struct bio *); 342 extern void bio_free(struct bio *, struct bio_set *); 343 344 extern void bio_endio(struct bio *, int); 345 struct request_queue; 346 extern int bio_phys_segments(struct request_queue *, struct bio *); 347 348 extern void __bio_clone(struct bio *, struct bio *); 349 extern struct bio *bio_clone(struct bio *, gfp_t); 350 351 extern void bio_init(struct bio *); 352 353 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int); 354 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *, 355 unsigned int, unsigned int); 356 extern int bio_get_nr_vecs(struct block_device *); 357 extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int); 358 extern struct bio *bio_map_user(struct request_queue *, struct block_device *, 359 unsigned long, unsigned int, int, gfp_t); 360 struct sg_iovec; 361 struct rq_map_data; 362 extern struct bio *bio_map_user_iov(struct request_queue *, 363 struct block_device *, 364 struct sg_iovec *, int, int, gfp_t); 365 extern void bio_unmap_user(struct bio *); 366 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int, 367 gfp_t); 368 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int, 369 gfp_t, int); 370 extern void bio_set_pages_dirty(struct bio *bio); 371 extern void bio_check_pages_dirty(struct bio *bio); 372 extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *, 373 unsigned long, unsigned int, int, gfp_t); 374 extern struct bio *bio_copy_user_iov(struct request_queue *, 375 struct rq_map_data *, struct sg_iovec *, 376 int, int, gfp_t); 377 extern int bio_uncopy_user(struct bio *); 378 void zero_fill_bio(struct bio *bio); 379 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *); 380 extern unsigned int bvec_nr_vecs(unsigned short idx); 381 382 /* 383 * Allow queuer to specify a completion CPU for this bio 384 */ 385 static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu) 386 { 387 bio->bi_comp_cpu = cpu; 388 } 389 390 /* 391 * bio_set is used to allow other portions of the IO system to 392 * allocate their own private memory pools for bio and iovec structures. 393 * These memory pools in turn all allocate from the bio_slab 394 * and the bvec_slabs[]. 395 */ 396 #define BIO_POOL_SIZE 2 397 #define BIOVEC_NR_POOLS 6 398 399 struct bio_set { 400 mempool_t *bio_pool; 401 #if defined(CONFIG_BLK_DEV_INTEGRITY) 402 mempool_t *bio_integrity_pool; 403 #endif 404 mempool_t *bvec_pools[BIOVEC_NR_POOLS]; 405 }; 406 407 struct biovec_slab { 408 int nr_vecs; 409 char *name; 410 struct kmem_cache *slab; 411 }; 412 413 extern struct bio_set *fs_bio_set; 414 415 /* 416 * a small number of entries is fine, not going to be performance critical. 417 * basically we just need to survive 418 */ 419 #define BIO_SPLIT_ENTRIES 2 420 421 #ifdef CONFIG_HIGHMEM 422 /* 423 * remember to add offset! and never ever reenable interrupts between a 424 * bvec_kmap_irq and bvec_kunmap_irq!! 425 * 426 * This function MUST be inlined - it plays with the CPU interrupt flags. 427 */ 428 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags) 429 { 430 unsigned long addr; 431 432 /* 433 * might not be a highmem page, but the preempt/irq count 434 * balancing is a lot nicer this way 435 */ 436 local_irq_save(*flags); 437 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ); 438 439 BUG_ON(addr & ~PAGE_MASK); 440 441 return (char *) addr + bvec->bv_offset; 442 } 443 444 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags) 445 { 446 unsigned long ptr = (unsigned long) buffer & PAGE_MASK; 447 448 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ); 449 local_irq_restore(*flags); 450 } 451 452 #else 453 #define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset) 454 #define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0) 455 #endif 456 457 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx, 458 unsigned long *flags) 459 { 460 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags); 461 } 462 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags) 463 464 #define bio_kmap_irq(bio, flags) \ 465 __bio_kmap_irq((bio), (bio)->bi_idx, (flags)) 466 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags) 467 468 /* 469 * Check whether this bio carries any data or not. A NULL bio is allowed. 470 */ 471 static inline int bio_has_data(struct bio *bio) 472 { 473 return bio && bio->bi_io_vec != NULL; 474 } 475 476 #if defined(CONFIG_BLK_DEV_INTEGRITY) 477 478 #define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)])) 479 #define bip_vec(bip) bip_vec_idx(bip, 0) 480 481 #define __bip_for_each_vec(bvl, bip, i, start_idx) \ 482 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \ 483 i < (bip)->bip_vcnt; \ 484 bvl++, i++) 485 486 #define bip_for_each_vec(bvl, bip, i) \ 487 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx) 488 489 #define bio_integrity(bio) (bio->bi_integrity != NULL) 490 491 extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *); 492 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int); 493 extern void bio_integrity_free(struct bio *, struct bio_set *); 494 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int); 495 extern int bio_integrity_enabled(struct bio *bio); 496 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int); 497 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int); 498 extern int bio_integrity_prep(struct bio *); 499 extern void bio_integrity_endio(struct bio *, int); 500 extern void bio_integrity_advance(struct bio *, unsigned int); 501 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int); 502 extern void bio_integrity_split(struct bio *, struct bio_pair *, int); 503 extern int bio_integrity_clone(struct bio *, struct bio *, struct bio_set *); 504 extern int bioset_integrity_create(struct bio_set *, int); 505 extern void bioset_integrity_free(struct bio_set *); 506 extern void bio_integrity_init_slab(void); 507 508 #else /* CONFIG_BLK_DEV_INTEGRITY */ 509 510 #define bio_integrity(a) (0) 511 #define bioset_integrity_create(a, b) (0) 512 #define bio_integrity_prep(a) (0) 513 #define bio_integrity_enabled(a) (0) 514 #define bio_integrity_clone(a, b, c) (0) 515 #define bioset_integrity_free(a) do { } while (0) 516 #define bio_integrity_free(a, b) do { } while (0) 517 #define bio_integrity_endio(a, b) do { } while (0) 518 #define bio_integrity_advance(a, b) do { } while (0) 519 #define bio_integrity_trim(a, b, c) do { } while (0) 520 #define bio_integrity_split(a, b, c) do { } while (0) 521 #define bio_integrity_set_tag(a, b, c) do { } while (0) 522 #define bio_integrity_get_tag(a, b, c) do { } while (0) 523 #define bio_integrity_init_slab(a) do { } while (0) 524 525 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 526 527 #endif /* CONFIG_BLOCK */ 528 #endif /* __LINUX_BIO_H */ 529