at v2.6.39 16 kB view raw
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#include <asm/io.h> 30 31/* struct bio, bio_vec and BIO_* flags are defined in blk_types.h */ 32#include <linux/blk_types.h> 33 34#define BIO_DEBUG 35 36#ifdef BIO_DEBUG 37#define BIO_BUG_ON BUG_ON 38#else 39#define BIO_BUG_ON 40#endif 41 42#define BIO_MAX_PAGES 256 43#define BIO_MAX_SIZE (BIO_MAX_PAGES << PAGE_CACHE_SHIFT) 44#define BIO_MAX_SECTORS (BIO_MAX_SIZE >> 9) 45 46/* 47 * upper 16 bits of bi_rw define the io priority of this bio 48 */ 49#define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS) 50#define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT) 51#define bio_prio_valid(bio) ioprio_valid(bio_prio(bio)) 52 53#define bio_set_prio(bio, prio) do { \ 54 WARN_ON(prio >= (1 << IOPRIO_BITS)); \ 55 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \ 56 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \ 57} while (0) 58 59/* 60 * various member access, note that bio_data should of course not be used 61 * on highmem page vectors 62 */ 63#define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)])) 64#define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx) 65#define bio_page(bio) bio_iovec((bio))->bv_page 66#define bio_offset(bio) bio_iovec((bio))->bv_offset 67#define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx) 68#define bio_sectors(bio) ((bio)->bi_size >> 9) 69 70static inline unsigned int bio_cur_bytes(struct bio *bio) 71{ 72 if (bio->bi_vcnt) 73 return bio_iovec(bio)->bv_len; 74 else /* dataless requests such as discard */ 75 return bio->bi_size; 76} 77 78static inline void *bio_data(struct bio *bio) 79{ 80 if (bio->bi_vcnt) 81 return page_address(bio_page(bio)) + bio_offset(bio); 82 83 return NULL; 84} 85 86static inline int bio_has_allocated_vec(struct bio *bio) 87{ 88 return bio->bi_io_vec && bio->bi_io_vec != bio->bi_inline_vecs; 89} 90 91/* 92 * will die 93 */ 94#define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio))) 95#define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset) 96 97/* 98 * queues that have highmem support enabled may still need to revert to 99 * PIO transfers occasionally and thus map high pages temporarily. For 100 * permanent PIO fall back, user is probably better off disabling highmem 101 * I/O completely on that queue (see ide-dma for example) 102 */ 103#define __bio_kmap_atomic(bio, idx, kmtype) \ 104 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \ 105 bio_iovec_idx((bio), (idx))->bv_offset) 106 107#define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype) 108 109/* 110 * merge helpers etc 111 */ 112 113#define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1) 114#define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx) 115 116/* Default implementation of BIOVEC_PHYS_MERGEABLE */ 117#define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 118 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2))) 119 120/* 121 * allow arch override, for eg virtualized architectures (put in asm/io.h) 122 */ 123#ifndef BIOVEC_PHYS_MERGEABLE 124#define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 125 __BIOVEC_PHYS_MERGEABLE(vec1, vec2) 126#endif 127 128#define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \ 129 (((addr1) | (mask)) == (((addr2) - 1) | (mask))) 130#define BIOVEC_SEG_BOUNDARY(q, b1, b2) \ 131 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, queue_segment_boundary((q))) 132#define BIO_SEG_BOUNDARY(q, b1, b2) \ 133 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2))) 134 135#define bio_io_error(bio) bio_endio((bio), -EIO) 136 137/* 138 * drivers should not use the __ version unless they _really_ want to 139 * run through the entire bio and not just pending pieces 140 */ 141#define __bio_for_each_segment(bvl, bio, i, start_idx) \ 142 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \ 143 i < (bio)->bi_vcnt; \ 144 bvl++, i++) 145 146#define bio_for_each_segment(bvl, bio, i) \ 147 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx) 148 149/* 150 * get a reference to a bio, so it won't disappear. the intended use is 151 * something like: 152 * 153 * bio_get(bio); 154 * submit_bio(rw, bio); 155 * if (bio->bi_flags ...) 156 * do_something 157 * bio_put(bio); 158 * 159 * without the bio_get(), it could potentially complete I/O before submit_bio 160 * returns. and then bio would be freed memory when if (bio->bi_flags ...) 161 * runs 162 */ 163#define bio_get(bio) atomic_inc(&(bio)->bi_cnt) 164 165#if defined(CONFIG_BLK_DEV_INTEGRITY) 166/* 167 * bio integrity payload 168 */ 169struct bio_integrity_payload { 170 struct bio *bip_bio; /* parent bio */ 171 172 sector_t bip_sector; /* virtual start sector */ 173 174 void *bip_buf; /* generated integrity data */ 175 bio_end_io_t *bip_end_io; /* saved I/O completion fn */ 176 177 unsigned int bip_size; 178 179 unsigned short bip_slab; /* slab the bip came from */ 180 unsigned short bip_vcnt; /* # of integrity bio_vecs */ 181 unsigned short bip_idx; /* current bip_vec index */ 182 183 struct work_struct bip_work; /* I/O completion */ 184 struct bio_vec bip_vec[0]; /* embedded bvec array */ 185}; 186#endif /* CONFIG_BLK_DEV_INTEGRITY */ 187 188/* 189 * A bio_pair is used when we need to split a bio. 190 * This can only happen for a bio that refers to just one 191 * page of data, and in the unusual situation when the 192 * page crosses a chunk/device boundary 193 * 194 * The address of the master bio is stored in bio1.bi_private 195 * The address of the pool the pair was allocated from is stored 196 * in bio2.bi_private 197 */ 198struct bio_pair { 199 struct bio bio1, bio2; 200 struct bio_vec bv1, bv2; 201#if defined(CONFIG_BLK_DEV_INTEGRITY) 202 struct bio_integrity_payload bip1, bip2; 203 struct bio_vec iv1, iv2; 204#endif 205 atomic_t cnt; 206 int error; 207}; 208extern struct bio_pair *bio_split(struct bio *bi, int first_sectors); 209extern void bio_pair_release(struct bio_pair *dbio); 210 211extern struct bio_set *bioset_create(unsigned int, unsigned int); 212extern void bioset_free(struct bio_set *); 213 214extern struct bio *bio_alloc(gfp_t, int); 215extern struct bio *bio_kmalloc(gfp_t, int); 216extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *); 217extern void bio_put(struct bio *); 218extern void bio_free(struct bio *, struct bio_set *); 219 220extern void bio_endio(struct bio *, int); 221struct request_queue; 222extern int bio_phys_segments(struct request_queue *, struct bio *); 223 224extern void __bio_clone(struct bio *, struct bio *); 225extern struct bio *bio_clone(struct bio *, gfp_t); 226 227extern void bio_init(struct bio *); 228 229extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int); 230extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *, 231 unsigned int, unsigned int); 232extern int bio_get_nr_vecs(struct block_device *); 233extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int); 234extern struct bio *bio_map_user(struct request_queue *, struct block_device *, 235 unsigned long, unsigned int, int, gfp_t); 236struct sg_iovec; 237struct rq_map_data; 238extern struct bio *bio_map_user_iov(struct request_queue *, 239 struct block_device *, 240 struct sg_iovec *, int, int, gfp_t); 241extern void bio_unmap_user(struct bio *); 242extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int, 243 gfp_t); 244extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int, 245 gfp_t, int); 246extern void bio_set_pages_dirty(struct bio *bio); 247extern void bio_check_pages_dirty(struct bio *bio); 248 249#ifndef ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 250# error "You should define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE for your platform" 251#endif 252#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 253extern void bio_flush_dcache_pages(struct bio *bi); 254#else 255static inline void bio_flush_dcache_pages(struct bio *bi) 256{ 257} 258#endif 259 260extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *, 261 unsigned long, unsigned int, int, gfp_t); 262extern struct bio *bio_copy_user_iov(struct request_queue *, 263 struct rq_map_data *, struct sg_iovec *, 264 int, int, gfp_t); 265extern int bio_uncopy_user(struct bio *); 266void zero_fill_bio(struct bio *bio); 267extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *); 268extern void bvec_free_bs(struct bio_set *, struct bio_vec *, unsigned int); 269extern unsigned int bvec_nr_vecs(unsigned short idx); 270 271/* 272 * Allow queuer to specify a completion CPU for this bio 273 */ 274static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu) 275{ 276 bio->bi_comp_cpu = cpu; 277} 278 279/* 280 * bio_set is used to allow other portions of the IO system to 281 * allocate their own private memory pools for bio and iovec structures. 282 * These memory pools in turn all allocate from the bio_slab 283 * and the bvec_slabs[]. 284 */ 285#define BIO_POOL_SIZE 2 286#define BIOVEC_NR_POOLS 6 287#define BIOVEC_MAX_IDX (BIOVEC_NR_POOLS - 1) 288 289struct bio_set { 290 struct kmem_cache *bio_slab; 291 unsigned int front_pad; 292 293 mempool_t *bio_pool; 294#if defined(CONFIG_BLK_DEV_INTEGRITY) 295 mempool_t *bio_integrity_pool; 296#endif 297 mempool_t *bvec_pool; 298}; 299 300struct biovec_slab { 301 int nr_vecs; 302 char *name; 303 struct kmem_cache *slab; 304}; 305 306extern struct bio_set *fs_bio_set; 307 308/* 309 * a small number of entries is fine, not going to be performance critical. 310 * basically we just need to survive 311 */ 312#define BIO_SPLIT_ENTRIES 2 313 314#ifdef CONFIG_HIGHMEM 315/* 316 * remember never ever reenable interrupts between a bvec_kmap_irq and 317 * bvec_kunmap_irq! 318 */ 319static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags) 320{ 321 unsigned long addr; 322 323 /* 324 * might not be a highmem page, but the preempt/irq count 325 * balancing is a lot nicer this way 326 */ 327 local_irq_save(*flags); 328 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ); 329 330 BUG_ON(addr & ~PAGE_MASK); 331 332 return (char *) addr + bvec->bv_offset; 333} 334 335static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags) 336{ 337 unsigned long ptr = (unsigned long) buffer & PAGE_MASK; 338 339 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ); 340 local_irq_restore(*flags); 341} 342 343#else 344static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags) 345{ 346 return page_address(bvec->bv_page) + bvec->bv_offset; 347} 348 349static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags) 350{ 351 *flags = 0; 352} 353#endif 354 355static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx, 356 unsigned long *flags) 357{ 358 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags); 359} 360#define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags) 361 362#define bio_kmap_irq(bio, flags) \ 363 __bio_kmap_irq((bio), (bio)->bi_idx, (flags)) 364#define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags) 365 366/* 367 * Check whether this bio carries any data or not. A NULL bio is allowed. 368 */ 369static inline int bio_has_data(struct bio *bio) 370{ 371 return bio && bio->bi_io_vec != NULL; 372} 373 374/* 375 * BIO list management for use by remapping drivers (e.g. DM or MD) and loop. 376 * 377 * A bio_list anchors a singly-linked list of bios chained through the bi_next 378 * member of the bio. The bio_list also caches the last list member to allow 379 * fast access to the tail. 380 */ 381struct bio_list { 382 struct bio *head; 383 struct bio *tail; 384}; 385 386static inline int bio_list_empty(const struct bio_list *bl) 387{ 388 return bl->head == NULL; 389} 390 391static inline void bio_list_init(struct bio_list *bl) 392{ 393 bl->head = bl->tail = NULL; 394} 395 396#define bio_list_for_each(bio, bl) \ 397 for (bio = (bl)->head; bio; bio = bio->bi_next) 398 399static inline unsigned bio_list_size(const struct bio_list *bl) 400{ 401 unsigned sz = 0; 402 struct bio *bio; 403 404 bio_list_for_each(bio, bl) 405 sz++; 406 407 return sz; 408} 409 410static inline void bio_list_add(struct bio_list *bl, struct bio *bio) 411{ 412 bio->bi_next = NULL; 413 414 if (bl->tail) 415 bl->tail->bi_next = bio; 416 else 417 bl->head = bio; 418 419 bl->tail = bio; 420} 421 422static inline void bio_list_add_head(struct bio_list *bl, struct bio *bio) 423{ 424 bio->bi_next = bl->head; 425 426 bl->head = bio; 427 428 if (!bl->tail) 429 bl->tail = bio; 430} 431 432static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2) 433{ 434 if (!bl2->head) 435 return; 436 437 if (bl->tail) 438 bl->tail->bi_next = bl2->head; 439 else 440 bl->head = bl2->head; 441 442 bl->tail = bl2->tail; 443} 444 445static inline void bio_list_merge_head(struct bio_list *bl, 446 struct bio_list *bl2) 447{ 448 if (!bl2->head) 449 return; 450 451 if (bl->head) 452 bl2->tail->bi_next = bl->head; 453 else 454 bl->tail = bl2->tail; 455 456 bl->head = bl2->head; 457} 458 459static inline struct bio *bio_list_peek(struct bio_list *bl) 460{ 461 return bl->head; 462} 463 464static inline struct bio *bio_list_pop(struct bio_list *bl) 465{ 466 struct bio *bio = bl->head; 467 468 if (bio) { 469 bl->head = bl->head->bi_next; 470 if (!bl->head) 471 bl->tail = NULL; 472 473 bio->bi_next = NULL; 474 } 475 476 return bio; 477} 478 479static inline struct bio *bio_list_get(struct bio_list *bl) 480{ 481 struct bio *bio = bl->head; 482 483 bl->head = bl->tail = NULL; 484 485 return bio; 486} 487 488#if defined(CONFIG_BLK_DEV_INTEGRITY) 489 490#define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)])) 491#define bip_vec(bip) bip_vec_idx(bip, 0) 492 493#define __bip_for_each_vec(bvl, bip, i, start_idx) \ 494 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \ 495 i < (bip)->bip_vcnt; \ 496 bvl++, i++) 497 498#define bip_for_each_vec(bvl, bip, i) \ 499 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx) 500 501#define bio_for_each_integrity_vec(_bvl, _bio, _iter) \ 502 for_each_bio(_bio) \ 503 bip_for_each_vec(_bvl, _bio->bi_integrity, _iter) 504 505#define bio_integrity(bio) (bio->bi_integrity != NULL) 506 507extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *); 508extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int); 509extern void bio_integrity_free(struct bio *, struct bio_set *); 510extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int); 511extern int bio_integrity_enabled(struct bio *bio); 512extern int bio_integrity_set_tag(struct bio *, void *, unsigned int); 513extern int bio_integrity_get_tag(struct bio *, void *, unsigned int); 514extern int bio_integrity_prep(struct bio *); 515extern void bio_integrity_endio(struct bio *, int); 516extern void bio_integrity_advance(struct bio *, unsigned int); 517extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int); 518extern void bio_integrity_split(struct bio *, struct bio_pair *, int); 519extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t, struct bio_set *); 520extern int bioset_integrity_create(struct bio_set *, int); 521extern void bioset_integrity_free(struct bio_set *); 522extern void bio_integrity_init(void); 523 524#else /* CONFIG_BLK_DEV_INTEGRITY */ 525 526#define bio_integrity(a) (0) 527#define bioset_integrity_create(a, b) (0) 528#define bio_integrity_prep(a) (0) 529#define bio_integrity_enabled(a) (0) 530#define bio_integrity_clone(a, b, c, d) (0) 531#define bioset_integrity_free(a) do { } while (0) 532#define bio_integrity_free(a, b) do { } while (0) 533#define bio_integrity_endio(a, b) do { } while (0) 534#define bio_integrity_advance(a, b) do { } while (0) 535#define bio_integrity_trim(a, b, c) do { } while (0) 536#define bio_integrity_split(a, b, c) do { } while (0) 537#define bio_integrity_set_tag(a, b, c) do { } while (0) 538#define bio_integrity_get_tag(a, b, c) do { } while (0) 539#define bio_integrity_init(a) do { } while (0) 540 541#endif /* CONFIG_BLK_DEV_INTEGRITY */ 542 543#endif /* CONFIG_BLOCK */ 544#endif /* __LINUX_BIO_H */