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