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 #include <linux/bug.h> 27 28 #ifdef CONFIG_BLOCK 29 30 #include <asm/io.h> 31 32 /* struct bio, bio_vec and BIO_* flags are defined in blk_types.h */ 33 #include <linux/blk_types.h> 34 35 #define BIO_DEBUG 36 37 #ifdef BIO_DEBUG 38 #define BIO_BUG_ON BUG_ON 39 #else 40 #define BIO_BUG_ON 41 #endif 42 43 #define BIO_MAX_PAGES 256 44 #define BIO_MAX_SIZE (BIO_MAX_PAGES << PAGE_CACHE_SHIFT) 45 #define BIO_MAX_SECTORS (BIO_MAX_SIZE >> 9) 46 47 /* 48 * upper 16 bits of bi_rw define the io priority of this bio 49 */ 50 #define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS) 51 #define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT) 52 #define bio_prio_valid(bio) ioprio_valid(bio_prio(bio)) 53 54 #define bio_set_prio(bio, prio) do { \ 55 WARN_ON(prio >= (1 << IOPRIO_BITS)); \ 56 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \ 57 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \ 58 } while (0) 59 60 /* 61 * various member access, note that bio_data should of course not be used 62 * on highmem page vectors 63 */ 64 #define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)])) 65 #define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx) 66 #define bio_page(bio) bio_iovec((bio))->bv_page 67 #define bio_offset(bio) bio_iovec((bio))->bv_offset 68 #define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx) 69 #define bio_sectors(bio) ((bio)->bi_size >> 9) 70 71 static inline unsigned int bio_cur_bytes(struct bio *bio) 72 { 73 if (bio->bi_vcnt) 74 return bio_iovec(bio)->bv_len; 75 else /* dataless requests such as discard */ 76 return bio->bi_size; 77 } 78 79 static inline void *bio_data(struct bio *bio) 80 { 81 if (bio->bi_vcnt) 82 return page_address(bio_page(bio)) + bio_offset(bio); 83 84 return NULL; 85 } 86 87 static inline int bio_has_allocated_vec(struct bio *bio) 88 { 89 return bio->bi_io_vec && bio->bi_io_vec != bio->bi_inline_vecs; 90 } 91 92 /* 93 * will die 94 */ 95 #define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio))) 96 #define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset) 97 98 /* 99 * queues that have highmem support enabled may still need to revert to 100 * PIO transfers occasionally and thus map high pages temporarily. For 101 * permanent PIO fall back, user is probably better off disabling highmem 102 * I/O completely on that queue (see ide-dma for example) 103 */ 104 #define __bio_kmap_atomic(bio, idx, kmtype) \ 105 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page) + \ 106 bio_iovec_idx((bio), (idx))->bv_offset) 107 108 #define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr) 109 110 /* 111 * merge helpers etc 112 */ 113 114 #define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1) 115 #define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx) 116 117 /* Default implementation of BIOVEC_PHYS_MERGEABLE */ 118 #define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 119 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2))) 120 121 /* 122 * allow arch override, for eg virtualized architectures (put in asm/io.h) 123 */ 124 #ifndef BIOVEC_PHYS_MERGEABLE 125 #define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 126 __BIOVEC_PHYS_MERGEABLE(vec1, vec2) 127 #endif 128 129 #define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \ 130 (((addr1) | (mask)) == (((addr2) - 1) | (mask))) 131 #define BIOVEC_SEG_BOUNDARY(q, b1, b2) \ 132 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, queue_segment_boundary((q))) 133 #define BIO_SEG_BOUNDARY(q, b1, b2) \ 134 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2))) 135 136 #define bio_io_error(bio) bio_endio((bio), -EIO) 137 138 /* 139 * drivers should not use the __ version unless they _really_ want to 140 * run through the entire bio and not just pending pieces 141 */ 142 #define __bio_for_each_segment(bvl, bio, i, start_idx) \ 143 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \ 144 i < (bio)->bi_vcnt; \ 145 bvl++, i++) 146 147 #define bio_for_each_segment(bvl, bio, i) \ 148 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx) 149 150 /* 151 * get a reference to a bio, so it won't disappear. the intended use is 152 * something like: 153 * 154 * bio_get(bio); 155 * submit_bio(rw, bio); 156 * if (bio->bi_flags ...) 157 * do_something 158 * bio_put(bio); 159 * 160 * without the bio_get(), it could potentially complete I/O before submit_bio 161 * returns. and then bio would be freed memory when if (bio->bi_flags ...) 162 * runs 163 */ 164 #define bio_get(bio) atomic_inc(&(bio)->bi_cnt) 165 166 #if defined(CONFIG_BLK_DEV_INTEGRITY) 167 /* 168 * bio integrity payload 169 */ 170 struct bio_integrity_payload { 171 struct bio *bip_bio; /* parent bio */ 172 173 sector_t bip_sector; /* virtual start sector */ 174 175 void *bip_buf; /* generated integrity data */ 176 bio_end_io_t *bip_end_io; /* saved I/O completion fn */ 177 178 unsigned int bip_size; 179 180 unsigned short bip_slab; /* slab the bip came from */ 181 unsigned short bip_vcnt; /* # of integrity bio_vecs */ 182 unsigned short bip_idx; /* current bip_vec index */ 183 184 struct work_struct bip_work; /* I/O completion */ 185 struct bio_vec bip_vec[0]; /* embedded bvec array */ 186 }; 187 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 188 189 /* 190 * A bio_pair is used when we need to split a bio. 191 * This can only happen for a bio that refers to just one 192 * page of data, and in the unusual situation when the 193 * page crosses a chunk/device boundary 194 * 195 * The address of the master bio is stored in bio1.bi_private 196 * The address of the pool the pair was allocated from is stored 197 * in bio2.bi_private 198 */ 199 struct bio_pair { 200 struct bio bio1, bio2; 201 struct bio_vec bv1, bv2; 202 #if defined(CONFIG_BLK_DEV_INTEGRITY) 203 struct bio_integrity_payload bip1, bip2; 204 struct bio_vec iv1, iv2; 205 #endif 206 atomic_t cnt; 207 int error; 208 }; 209 extern struct bio_pair *bio_split(struct bio *bi, int first_sectors); 210 extern void bio_pair_release(struct bio_pair *dbio); 211 212 extern struct bio_set *bioset_create(unsigned int, unsigned int); 213 extern void bioset_free(struct bio_set *); 214 215 extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *); 216 extern void bio_put(struct bio *); 217 218 extern void __bio_clone(struct bio *, struct bio *); 219 extern struct bio *bio_clone_bioset(struct bio *, gfp_t, struct bio_set *bs); 220 221 extern struct bio_set *fs_bio_set; 222 223 static inline struct bio *bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs) 224 { 225 return bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set); 226 } 227 228 static inline struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask) 229 { 230 return bio_clone_bioset(bio, gfp_mask, fs_bio_set); 231 } 232 233 static inline struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned int nr_iovecs) 234 { 235 return bio_alloc_bioset(gfp_mask, nr_iovecs, NULL); 236 } 237 238 static inline struct bio *bio_clone_kmalloc(struct bio *bio, gfp_t gfp_mask) 239 { 240 return bio_clone_bioset(bio, gfp_mask, NULL); 241 242 } 243 244 extern void bio_endio(struct bio *, int); 245 struct request_queue; 246 extern int bio_phys_segments(struct request_queue *, struct bio *); 247 248 extern void bio_init(struct bio *); 249 extern void bio_reset(struct bio *); 250 251 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int); 252 extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *, 253 unsigned int, unsigned int); 254 extern int bio_get_nr_vecs(struct block_device *); 255 extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int); 256 extern struct bio *bio_map_user(struct request_queue *, struct block_device *, 257 unsigned long, unsigned int, int, gfp_t); 258 struct sg_iovec; 259 struct rq_map_data; 260 extern struct bio *bio_map_user_iov(struct request_queue *, 261 struct block_device *, 262 struct sg_iovec *, int, int, gfp_t); 263 extern void bio_unmap_user(struct bio *); 264 extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int, 265 gfp_t); 266 extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int, 267 gfp_t, int); 268 extern void bio_set_pages_dirty(struct bio *bio); 269 extern void bio_check_pages_dirty(struct bio *bio); 270 271 #ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 272 # error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 273 #endif 274 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 275 extern void bio_flush_dcache_pages(struct bio *bi); 276 #else 277 static inline void bio_flush_dcache_pages(struct bio *bi) 278 { 279 } 280 #endif 281 282 extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *, 283 unsigned long, unsigned int, int, gfp_t); 284 extern struct bio *bio_copy_user_iov(struct request_queue *, 285 struct rq_map_data *, struct sg_iovec *, 286 int, int, gfp_t); 287 extern int bio_uncopy_user(struct bio *); 288 void zero_fill_bio(struct bio *bio); 289 extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *); 290 extern void bvec_free_bs(struct bio_set *, struct bio_vec *, unsigned int); 291 extern unsigned int bvec_nr_vecs(unsigned short idx); 292 293 #ifdef CONFIG_BLK_CGROUP 294 int bio_associate_current(struct bio *bio); 295 void bio_disassociate_task(struct bio *bio); 296 #else /* CONFIG_BLK_CGROUP */ 297 static inline int bio_associate_current(struct bio *bio) { return -ENOENT; } 298 static inline void bio_disassociate_task(struct bio *bio) { } 299 #endif /* CONFIG_BLK_CGROUP */ 300 301 /* 302 * bio_set is used to allow other portions of the IO system to 303 * allocate their own private memory pools for bio and iovec structures. 304 * These memory pools in turn all allocate from the bio_slab 305 * and the bvec_slabs[]. 306 */ 307 #define BIO_POOL_SIZE 2 308 #define BIOVEC_NR_POOLS 6 309 #define BIOVEC_MAX_IDX (BIOVEC_NR_POOLS - 1) 310 311 struct bio_set { 312 struct kmem_cache *bio_slab; 313 unsigned int front_pad; 314 315 mempool_t *bio_pool; 316 #if defined(CONFIG_BLK_DEV_INTEGRITY) 317 mempool_t *bio_integrity_pool; 318 #endif 319 mempool_t *bvec_pool; 320 }; 321 322 struct biovec_slab { 323 int nr_vecs; 324 char *name; 325 struct kmem_cache *slab; 326 }; 327 328 /* 329 * a small number of entries is fine, not going to be performance critical. 330 * basically we just need to survive 331 */ 332 #define BIO_SPLIT_ENTRIES 2 333 334 #ifdef CONFIG_HIGHMEM 335 /* 336 * remember never ever reenable interrupts between a bvec_kmap_irq and 337 * bvec_kunmap_irq! 338 */ 339 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags) 340 { 341 unsigned long addr; 342 343 /* 344 * might not be a highmem page, but the preempt/irq count 345 * balancing is a lot nicer this way 346 */ 347 local_irq_save(*flags); 348 addr = (unsigned long) kmap_atomic(bvec->bv_page); 349 350 BUG_ON(addr & ~PAGE_MASK); 351 352 return (char *) addr + bvec->bv_offset; 353 } 354 355 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags) 356 { 357 unsigned long ptr = (unsigned long) buffer & PAGE_MASK; 358 359 kunmap_atomic((void *) ptr); 360 local_irq_restore(*flags); 361 } 362 363 #else 364 static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags) 365 { 366 return page_address(bvec->bv_page) + bvec->bv_offset; 367 } 368 369 static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags) 370 { 371 *flags = 0; 372 } 373 #endif 374 375 static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx, 376 unsigned long *flags) 377 { 378 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags); 379 } 380 #define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags) 381 382 #define bio_kmap_irq(bio, flags) \ 383 __bio_kmap_irq((bio), (bio)->bi_idx, (flags)) 384 #define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags) 385 386 /* 387 * Check whether this bio carries any data or not. A NULL bio is allowed. 388 */ 389 static inline bool bio_has_data(struct bio *bio) 390 { 391 if (bio && bio->bi_vcnt) 392 return true; 393 394 return false; 395 } 396 397 static inline bool bio_is_rw(struct bio *bio) 398 { 399 if (!bio_has_data(bio)) 400 return false; 401 402 if (bio->bi_rw & REQ_WRITE_SAME) 403 return false; 404 405 return true; 406 } 407 408 static inline bool bio_mergeable(struct bio *bio) 409 { 410 if (bio->bi_rw & REQ_NOMERGE_FLAGS) 411 return false; 412 413 return true; 414 } 415 416 /* 417 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop. 418 * 419 * A bio_list anchors a singly-linked list of bios chained through the bi_next 420 * member of the bio. The bio_list also caches the last list member to allow 421 * fast access to the tail. 422 */ 423 struct bio_list { 424 struct bio *head; 425 struct bio *tail; 426 }; 427 428 static inline int bio_list_empty(const struct bio_list *bl) 429 { 430 return bl->head == NULL; 431 } 432 433 static inline void bio_list_init(struct bio_list *bl) 434 { 435 bl->head = bl->tail = NULL; 436 } 437 438 #define bio_list_for_each(bio, bl) \ 439 for (bio = (bl)->head; bio; bio = bio->bi_next) 440 441 static inline unsigned bio_list_size(const struct bio_list *bl) 442 { 443 unsigned sz = 0; 444 struct bio *bio; 445 446 bio_list_for_each(bio, bl) 447 sz++; 448 449 return sz; 450 } 451 452 static inline void bio_list_add(struct bio_list *bl, struct bio *bio) 453 { 454 bio->bi_next = NULL; 455 456 if (bl->tail) 457 bl->tail->bi_next = bio; 458 else 459 bl->head = bio; 460 461 bl->tail = bio; 462 } 463 464 static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio) 465 { 466 bio->bi_next = bl->head; 467 468 bl->head = bio; 469 470 if (!bl->tail) 471 bl->tail = bio; 472 } 473 474 static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2) 475 { 476 if (!bl2->head) 477 return; 478 479 if (bl->tail) 480 bl->tail->bi_next = bl2->head; 481 else 482 bl->head = bl2->head; 483 484 bl->tail = bl2->tail; 485 } 486 487 static inline void bio_list_merge_head(struct bio_list *bl, 488 struct bio_list *bl2) 489 { 490 if (!bl2->head) 491 return; 492 493 if (bl->head) 494 bl2->tail->bi_next = bl->head; 495 else 496 bl->tail = bl2->tail; 497 498 bl->head = bl2->head; 499 } 500 501 static inline struct bio *bio_list_peek(struct bio_list *bl) 502 { 503 return bl->head; 504 } 505 506 static inline struct bio *bio_list_pop(struct bio_list *bl) 507 { 508 struct bio *bio = bl->head; 509 510 if (bio) { 511 bl->head = bl->head->bi_next; 512 if (!bl->head) 513 bl->tail = NULL; 514 515 bio->bi_next = NULL; 516 } 517 518 return bio; 519 } 520 521 static inline struct bio *bio_list_get(struct bio_list *bl) 522 { 523 struct bio *bio = bl->head; 524 525 bl->head = bl->tail = NULL; 526 527 return bio; 528 } 529 530 #if defined(CONFIG_BLK_DEV_INTEGRITY) 531 532 #define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)])) 533 #define bip_vec(bip) bip_vec_idx(bip, 0) 534 535 #define __bip_for_each_vec(bvl, bip, i, start_idx) \ 536 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \ 537 i < (bip)->bip_vcnt; \ 538 bvl++, i++) 539 540 #define bip_for_each_vec(bvl, bip, i) \ 541 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx) 542 543 #define bio_for_each_integrity_vec(_bvl, _bio, _iter) \ 544 for_each_bio(_bio) \ 545 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter) 546 547 #define bio_integrity(bio) (bio->bi_integrity != NULL) 548 549 extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int); 550 extern void bio_integrity_free(struct bio *); 551 extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int); 552 extern int bio_integrity_enabled(struct bio *bio); 553 extern int bio_integrity_set_tag(struct bio *, void *, unsigned int); 554 extern int bio_integrity_get_tag(struct bio *, void *, unsigned int); 555 extern int bio_integrity_prep(struct bio *); 556 extern void bio_integrity_endio(struct bio *, int); 557 extern void bio_integrity_advance(struct bio *, unsigned int); 558 extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int); 559 extern void bio_integrity_split(struct bio *, struct bio_pair *, int); 560 extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t); 561 extern int bioset_integrity_create(struct bio_set *, int); 562 extern void bioset_integrity_free(struct bio_set *); 563 extern void bio_integrity_init(void); 564 565 #else /* CONFIG_BLK_DEV_INTEGRITY */ 566 567 static inline int bio_integrity(struct bio *bio) 568 { 569 return 0; 570 } 571 572 static inline int bio_integrity_enabled(struct bio *bio) 573 { 574 return 0; 575 } 576 577 static inline int bioset_integrity_create(struct bio_set *bs, int pool_size) 578 { 579 return 0; 580 } 581 582 static inline void bioset_integrity_free (struct bio_set *bs) 583 { 584 return; 585 } 586 587 static inline int bio_integrity_prep(struct bio *bio) 588 { 589 return 0; 590 } 591 592 static inline void bio_integrity_free(struct bio *bio) 593 { 594 return; 595 } 596 597 static inline int bio_integrity_clone(struct bio *bio, struct bio *bio_src, 598 gfp_t gfp_mask) 599 { 600 return 0; 601 } 602 603 static inline void bio_integrity_split(struct bio *bio, struct bio_pair *bp, 604 int sectors) 605 { 606 return; 607 } 608 609 static inline void bio_integrity_advance(struct bio *bio, 610 unsigned int bytes_done) 611 { 612 return; 613 } 614 615 static inline void bio_integrity_trim(struct bio *bio, unsigned int offset, 616 unsigned int sectors) 617 { 618 return; 619 } 620 621 static inline void bio_integrity_init(void) 622 { 623 return; 624 } 625 626 #endif /* CONFIG_BLK_DEV_INTEGRITY */ 627 628 #endif /* CONFIG_BLOCK */ 629 #endif /* __LINUX_BIO_H */ 630