at v2.6.29 18 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#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 */ 46struct bio_vec { 47 struct page *bv_page; 48 unsigned int bv_len; 49 unsigned int bv_offset; 50}; 51 52struct bio_set; 53struct bio; 54struct bio_integrity_payload; 55typedef void (bio_end_io_t) (struct bio *, int); 56typedef 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 */ 62struct 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 atomic_t bi_cnt; /* pin count */ 94 95 struct bio_vec *bi_io_vec; /* the actual vec list */ 96 97 bio_end_io_t *bi_end_io; 98 99 void *bi_private; 100#if defined(CONFIG_BLK_DEV_INTEGRITY) 101 struct bio_integrity_payload *bi_integrity; /* data integrity */ 102#endif 103 104 bio_destructor_t *bi_destructor; /* destructor */ 105 106 /* 107 * We can inline a number of vecs at the end of the bio, to avoid 108 * double allocations for a small number of bio_vecs. This member 109 * MUST obviously be kept at the very end of the bio. 110 */ 111 struct bio_vec bi_inline_vecs[0]; 112}; 113 114/* 115 * bio flags 116 */ 117#define BIO_UPTODATE 0 /* ok after I/O completion */ 118#define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */ 119#define BIO_EOF 2 /* out-out-bounds error */ 120#define BIO_SEG_VALID 3 /* bi_phys_segments valid */ 121#define BIO_CLONED 4 /* doesn't own data */ 122#define BIO_BOUNCED 5 /* bio is a bounce bio */ 123#define BIO_USER_MAPPED 6 /* contains user pages */ 124#define BIO_EOPNOTSUPP 7 /* not supported */ 125#define BIO_CPU_AFFINE 8 /* complete bio on same CPU as submitted */ 126#define BIO_NULL_MAPPED 9 /* contains invalid user pages */ 127#define BIO_FS_INTEGRITY 10 /* fs owns integrity data, not block layer */ 128#define BIO_QUIET 11 /* Make BIO Quiet */ 129#define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag))) 130 131/* 132 * top 4 bits of bio flags indicate the pool this bio came from 133 */ 134#define BIO_POOL_BITS (4) 135#define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS) 136#define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET) 137#define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET) 138 139/* 140 * bio bi_rw flags 141 * 142 * bit 0 -- data direction 143 * If not set, bio is a read from device. If set, it's a write to device. 144 * bit 1 -- rw-ahead when set 145 * bit 2 -- barrier 146 * Insert a serialization point in the IO queue, forcing previously 147 * submitted IO to be completed before this one is issued. 148 * bit 3 -- synchronous I/O hint: the block layer will unplug immediately 149 * Note that this does NOT indicate that the IO itself is sync, just 150 * that the block layer will not postpone issue of this IO by plugging. 151 * bit 4 -- metadata request 152 * Used for tracing to differentiate metadata and data IO. May also 153 * get some preferential treatment in the IO scheduler 154 * bit 5 -- discard sectors 155 * Informs the lower level device that this range of sectors is no longer 156 * used by the file system and may thus be freed by the device. Used 157 * for flash based storage. 158 * bit 6 -- fail fast device errors 159 * bit 7 -- fail fast transport errors 160 * bit 8 -- fail fast driver errors 161 * Don't want driver retries for any fast fail whatever the reason. 162 */ 163#define BIO_RW 0 /* Must match RW in req flags (blkdev.h) */ 164#define BIO_RW_AHEAD 1 /* Must match FAILFAST in req flags */ 165#define BIO_RW_BARRIER 2 166#define BIO_RW_SYNCIO 3 167#define BIO_RW_UNPLUG 4 168#define BIO_RW_META 5 169#define BIO_RW_DISCARD 6 170#define BIO_RW_FAILFAST_DEV 7 171#define BIO_RW_FAILFAST_TRANSPORT 8 172#define BIO_RW_FAILFAST_DRIVER 9 173 174#define bio_rw_flagged(bio, flag) ((bio)->bi_rw & (1 << (flag))) 175 176/* 177 * Old defines, these should eventually be replaced by direct usage of 178 * bio_rw_flagged() 179 */ 180#define bio_barrier(bio) bio_rw_flagged(bio, BIO_RW_BARRIER) 181#define bio_sync(bio) bio_rw_flagged(bio, BIO_RW_SYNCIO) 182#define bio_unplug(bio) bio_rw_flagged(bio, BIO_RW_UNPLUG) 183#define bio_failfast_dev(bio) bio_rw_flagged(bio, BIO_RW_FAILFAST_DEV) 184#define bio_failfast_transport(bio) \ 185 bio_rw_flagged(bio, BIO_RW_FAILFAST_TRANSPORT) 186#define bio_failfast_driver(bio) \ 187 bio_rw_flagged(bio, BIO_RW_FAILFAST_DRIVER) 188#define bio_rw_ahead(bio) bio_rw_flagged(bio, BIO_RW_AHEAD) 189#define bio_rw_meta(bio) bio_rw_flagged(bio, BIO_RW_META) 190#define bio_discard(bio) bio_rw_flagged(bio, BIO_RW_DISCARD) 191 192/* 193 * upper 16 bits of bi_rw define the io priority of this bio 194 */ 195#define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS) 196#define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT) 197#define bio_prio_valid(bio) ioprio_valid(bio_prio(bio)) 198 199#define bio_set_prio(bio, prio) do { \ 200 WARN_ON(prio >= (1 << IOPRIO_BITS)); \ 201 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \ 202 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \ 203} while (0) 204 205/* 206 * various member access, note that bio_data should of course not be used 207 * on highmem page vectors 208 */ 209#define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)])) 210#define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx) 211#define bio_page(bio) bio_iovec((bio))->bv_page 212#define bio_offset(bio) bio_iovec((bio))->bv_offset 213#define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx) 214#define bio_sectors(bio) ((bio)->bi_size >> 9) 215#define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio) && !bio_discard(bio)) 216 217static inline unsigned int bio_cur_sectors(struct bio *bio) 218{ 219 if (bio->bi_vcnt) 220 return bio_iovec(bio)->bv_len >> 9; 221 else /* dataless requests such as discard */ 222 return bio->bi_size >> 9; 223} 224 225static inline void *bio_data(struct bio *bio) 226{ 227 if (bio->bi_vcnt) 228 return page_address(bio_page(bio)) + bio_offset(bio); 229 230 return NULL; 231} 232 233static inline int bio_has_allocated_vec(struct bio *bio) 234{ 235 return bio->bi_io_vec && bio->bi_io_vec != bio->bi_inline_vecs; 236} 237 238/* 239 * will die 240 */ 241#define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio))) 242#define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset) 243 244/* 245 * queues that have highmem support enabled may still need to revert to 246 * PIO transfers occasionally and thus map high pages temporarily. For 247 * permanent PIO fall back, user is probably better off disabling highmem 248 * I/O completely on that queue (see ide-dma for example) 249 */ 250#define __bio_kmap_atomic(bio, idx, kmtype) \ 251 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \ 252 bio_iovec_idx((bio), (idx))->bv_offset) 253 254#define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype) 255 256/* 257 * merge helpers etc 258 */ 259 260#define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1) 261#define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx) 262 263/* Default implementation of BIOVEC_PHYS_MERGEABLE */ 264#define __BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 265 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2))) 266 267/* 268 * allow arch override, for eg virtualized architectures (put in asm/io.h) 269 */ 270#ifndef BIOVEC_PHYS_MERGEABLE 271#define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \ 272 __BIOVEC_PHYS_MERGEABLE(vec1, vec2) 273#endif 274 275#define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \ 276 (((addr1) | (mask)) == (((addr2) - 1) | (mask))) 277#define BIOVEC_SEG_BOUNDARY(q, b1, b2) \ 278 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask) 279#define BIO_SEG_BOUNDARY(q, b1, b2) \ 280 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2))) 281 282#define bio_io_error(bio) bio_endio((bio), -EIO) 283 284/* 285 * drivers should not use the __ version unless they _really_ want to 286 * run through the entire bio and not just pending pieces 287 */ 288#define __bio_for_each_segment(bvl, bio, i, start_idx) \ 289 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \ 290 i < (bio)->bi_vcnt; \ 291 bvl++, i++) 292 293#define bio_for_each_segment(bvl, bio, i) \ 294 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx) 295 296/* 297 * get a reference to a bio, so it won't disappear. the intended use is 298 * something like: 299 * 300 * bio_get(bio); 301 * submit_bio(rw, bio); 302 * if (bio->bi_flags ...) 303 * do_something 304 * bio_put(bio); 305 * 306 * without the bio_get(), it could potentially complete I/O before submit_bio 307 * returns. and then bio would be freed memory when if (bio->bi_flags ...) 308 * runs 309 */ 310#define bio_get(bio) atomic_inc(&(bio)->bi_cnt) 311 312#if defined(CONFIG_BLK_DEV_INTEGRITY) 313/* 314 * bio integrity payload 315 */ 316struct bio_integrity_payload { 317 struct bio *bip_bio; /* parent bio */ 318 struct bio_vec *bip_vec; /* integrity data vector */ 319 320 sector_t bip_sector; /* virtual start sector */ 321 322 void *bip_buf; /* generated integrity data */ 323 bio_end_io_t *bip_end_io; /* saved I/O completion fn */ 324 325 unsigned int bip_size; 326 327 unsigned short bip_pool; /* pool the ivec came from */ 328 unsigned short bip_vcnt; /* # of integrity bio_vecs */ 329 unsigned short bip_idx; /* current bip_vec index */ 330 331 struct work_struct bip_work; /* I/O completion */ 332}; 333#endif /* CONFIG_BLK_DEV_INTEGRITY */ 334 335/* 336 * A bio_pair is used when we need to split a bio. 337 * This can only happen for a bio that refers to just one 338 * page of data, and in the unusual situation when the 339 * page crosses a chunk/device boundary 340 * 341 * The address of the master bio is stored in bio1.bi_private 342 * The address of the pool the pair was allocated from is stored 343 * in bio2.bi_private 344 */ 345struct bio_pair { 346 struct bio bio1, bio2; 347 struct bio_vec bv1, bv2; 348#if defined(CONFIG_BLK_DEV_INTEGRITY) 349 struct bio_integrity_payload bip1, bip2; 350 struct bio_vec iv1, iv2; 351#endif 352 atomic_t cnt; 353 int error; 354}; 355extern struct bio_pair *bio_split(struct bio *bi, int first_sectors); 356extern void bio_pair_release(struct bio_pair *dbio); 357 358extern struct bio_set *bioset_create(unsigned int, unsigned int); 359extern void bioset_free(struct bio_set *); 360 361extern struct bio *bio_alloc(gfp_t, int); 362extern struct bio *bio_kmalloc(gfp_t, int); 363extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *); 364extern void bio_put(struct bio *); 365extern void bio_free(struct bio *, struct bio_set *); 366 367extern void bio_endio(struct bio *, int); 368struct request_queue; 369extern int bio_phys_segments(struct request_queue *, struct bio *); 370 371extern void __bio_clone(struct bio *, struct bio *); 372extern struct bio *bio_clone(struct bio *, gfp_t); 373 374extern void bio_init(struct bio *); 375 376extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int); 377extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *, 378 unsigned int, unsigned int); 379extern int bio_get_nr_vecs(struct block_device *); 380extern sector_t bio_sector_offset(struct bio *, unsigned short, unsigned int); 381extern struct bio *bio_map_user(struct request_queue *, struct block_device *, 382 unsigned long, unsigned int, int, gfp_t); 383struct sg_iovec; 384struct rq_map_data; 385extern struct bio *bio_map_user_iov(struct request_queue *, 386 struct block_device *, 387 struct sg_iovec *, int, int, gfp_t); 388extern void bio_unmap_user(struct bio *); 389extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int, 390 gfp_t); 391extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int, 392 gfp_t, int); 393extern void bio_set_pages_dirty(struct bio *bio); 394extern void bio_check_pages_dirty(struct bio *bio); 395extern struct bio *bio_copy_user(struct request_queue *, struct rq_map_data *, 396 unsigned long, unsigned int, int, gfp_t); 397extern struct bio *bio_copy_user_iov(struct request_queue *, 398 struct rq_map_data *, struct sg_iovec *, 399 int, int, gfp_t); 400extern int bio_uncopy_user(struct bio *); 401void zero_fill_bio(struct bio *bio); 402extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *); 403extern void bvec_free_bs(struct bio_set *, struct bio_vec *, unsigned int); 404extern unsigned int bvec_nr_vecs(unsigned short idx); 405 406/* 407 * Allow queuer to specify a completion CPU for this bio 408 */ 409static inline void bio_set_completion_cpu(struct bio *bio, unsigned int cpu) 410{ 411 bio->bi_comp_cpu = cpu; 412} 413 414/* 415 * bio_set is used to allow other portions of the IO system to 416 * allocate their own private memory pools for bio and iovec structures. 417 * These memory pools in turn all allocate from the bio_slab 418 * and the bvec_slabs[]. 419 */ 420#define BIO_POOL_SIZE 2 421#define BIOVEC_NR_POOLS 6 422#define BIOVEC_MAX_IDX (BIOVEC_NR_POOLS - 1) 423 424struct bio_set { 425 struct kmem_cache *bio_slab; 426 unsigned int front_pad; 427 428 mempool_t *bio_pool; 429#if defined(CONFIG_BLK_DEV_INTEGRITY) 430 mempool_t *bio_integrity_pool; 431#endif 432 mempool_t *bvec_pool; 433}; 434 435struct biovec_slab { 436 int nr_vecs; 437 char *name; 438 struct kmem_cache *slab; 439}; 440 441extern struct bio_set *fs_bio_set; 442extern struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly; 443 444/* 445 * a small number of entries is fine, not going to be performance critical. 446 * basically we just need to survive 447 */ 448#define BIO_SPLIT_ENTRIES 2 449 450#ifdef CONFIG_HIGHMEM 451/* 452 * remember never ever reenable interrupts between a bvec_kmap_irq and 453 * bvec_kunmap_irq! 454 * 455 * This function MUST be inlined - it plays with the CPU interrupt flags. 456 */ 457static __always_inline char *bvec_kmap_irq(struct bio_vec *bvec, 458 unsigned long *flags) 459{ 460 unsigned long addr; 461 462 /* 463 * might not be a highmem page, but the preempt/irq count 464 * balancing is a lot nicer this way 465 */ 466 local_irq_save(*flags); 467 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ); 468 469 BUG_ON(addr & ~PAGE_MASK); 470 471 return (char *) addr + bvec->bv_offset; 472} 473 474static __always_inline void bvec_kunmap_irq(char *buffer, 475 unsigned long *flags) 476{ 477 unsigned long ptr = (unsigned long) buffer & PAGE_MASK; 478 479 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ); 480 local_irq_restore(*flags); 481} 482 483#else 484#define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset) 485#define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0) 486#endif 487 488static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx, 489 unsigned long *flags) 490{ 491 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags); 492} 493#define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags) 494 495#define bio_kmap_irq(bio, flags) \ 496 __bio_kmap_irq((bio), (bio)->bi_idx, (flags)) 497#define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags) 498 499/* 500 * Check whether this bio carries any data or not. A NULL bio is allowed. 501 */ 502static inline int bio_has_data(struct bio *bio) 503{ 504 return bio && bio->bi_io_vec != NULL; 505} 506 507#if defined(CONFIG_BLK_DEV_INTEGRITY) 508 509#define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)])) 510#define bip_vec(bip) bip_vec_idx(bip, 0) 511 512#define __bip_for_each_vec(bvl, bip, i, start_idx) \ 513 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \ 514 i < (bip)->bip_vcnt; \ 515 bvl++, i++) 516 517#define bip_for_each_vec(bvl, bip, i) \ 518 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx) 519 520#define bio_integrity(bio) (bio->bi_integrity != NULL) 521 522extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *); 523extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int); 524extern void bio_integrity_free(struct bio *, struct bio_set *); 525extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int); 526extern int bio_integrity_enabled(struct bio *bio); 527extern int bio_integrity_set_tag(struct bio *, void *, unsigned int); 528extern int bio_integrity_get_tag(struct bio *, void *, unsigned int); 529extern int bio_integrity_prep(struct bio *); 530extern void bio_integrity_endio(struct bio *, int); 531extern void bio_integrity_advance(struct bio *, unsigned int); 532extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int); 533extern void bio_integrity_split(struct bio *, struct bio_pair *, int); 534extern int bio_integrity_clone(struct bio *, struct bio *, gfp_t, struct bio_set *); 535extern int bioset_integrity_create(struct bio_set *, int); 536extern void bioset_integrity_free(struct bio_set *); 537extern void bio_integrity_init_slab(void); 538 539#else /* CONFIG_BLK_DEV_INTEGRITY */ 540 541#define bio_integrity(a) (0) 542#define bioset_integrity_create(a, b) (0) 543#define bio_integrity_prep(a) (0) 544#define bio_integrity_enabled(a) (0) 545#define bio_integrity_clone(a, b, c,d ) (0) 546#define bioset_integrity_free(a) do { } while (0) 547#define bio_integrity_free(a, b) do { } while (0) 548#define bio_integrity_endio(a, b) do { } while (0) 549#define bio_integrity_advance(a, b) do { } while (0) 550#define bio_integrity_trim(a, b, c) do { } while (0) 551#define bio_integrity_split(a, b, c) do { } while (0) 552#define bio_integrity_set_tag(a, b, c) do { } while (0) 553#define bio_integrity_get_tag(a, b, c) do { } while (0) 554#define bio_integrity_init_slab(a) do { } while (0) 555 556#endif /* CONFIG_BLK_DEV_INTEGRITY */ 557 558#endif /* CONFIG_BLOCK */ 559#endif /* __LINUX_BIO_H */