projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Add 'discard' request handling
[linux-2.6-block.git] / include / linux / bio.h
CommitLineData
1da177e4
LT		 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>
22e2c507 25#include <linux/ioprio.h>
1da177e4 26
02a5e0ac
DH		 27#ifdef CONFIG_BLOCK
28
1da177e4
LT		 29/* Platforms may set this to teach the BIO layer about IOMMU hardware. */
30#include <asm/io.h>
31
32#if defined(BIO_VMERGE_MAX_SIZE) && defined(BIO_VMERGE_BOUNDARY)
33#define BIOVEC_VIRT_START_SIZE(x) (bvec_to_phys(x) & (BIO_VMERGE_BOUNDARY - 1))
34#define BIOVEC_VIRT_OVERSIZE(x) ((x) > BIO_VMERGE_MAX_SIZE)
35#else
36#define BIOVEC_VIRT_START_SIZE(x) 0
37#define BIOVEC_VIRT_OVERSIZE(x) 0
38#endif
39
40#ifndef BIO_VMERGE_BOUNDARY
41#define BIO_VMERGE_BOUNDARY 0
42#endif
43
44#define BIO_DEBUG
45
46#ifdef BIO_DEBUG
47#define BIO_BUG_ON BUG_ON
48#else
49#define BIO_BUG_ON
50#endif
51
d84a8477 52#define BIO_MAX_PAGES 256
1da177e4
LT		 53#define BIO_MAX_SIZE (BIO_MAX_PAGES << PAGE_CACHE_SHIFT)
54#define BIO_MAX_SECTORS (BIO_MAX_SIZE >> 9)
55
56/*
57 * was unsigned short, but we might as well be ready for > 64kB I/O pages
58 */
59struct bio_vec {
60 struct page *bv_page;
61 unsigned int bv_len;
62 unsigned int bv_offset;
63};
64
65struct bio_set;
66struct bio;
7ba1ba12 67struct bio_integrity_payload;
6712ecf8 68typedef void (bio_end_io_t) (struct bio *, int);
1da177e4
LT		 69typedef void (bio_destructor_t) (struct bio *);
70
71/*
72 * main unit of I/O for the block layer and lower layers (ie drivers and
73 * stacking drivers)
74 */
75struct bio {
2c2345c2
RG		 76 sector_t bi_sector; /* device address in 512 byte
77 sectors */
1da177e4
LT		 78 struct bio *bi_next; /* request queue link */
79 struct block_device *bi_bdev;
80 unsigned long bi_flags; /* status, command, etc */
81 unsigned long bi_rw; /* bottom bits READ/WRITE,
82 * top bits priority
83 */
84
85 unsigned short bi_vcnt; /* how many bio_vec's */
86 unsigned short bi_idx; /* current index into bvl_vec */
87
88 /* Number of segments in this BIO after
89 * physical address coalescing is performed.
90 */
91 unsigned short bi_phys_segments;
92
93 /* Number of segments after physical and DMA remapping
94 * hardware coalescing is performed.
95 */
96 unsigned short bi_hw_segments;
97
98 unsigned int bi_size; /* residual I/O count */
99
100 /*
101 * To keep track of the max hw size, we account for the
102 * sizes of the first and last virtually mergeable segments
103 * in this bio
104 */
105 unsigned int bi_hw_front_size;
106 unsigned int bi_hw_back_size;
107
108 unsigned int bi_max_vecs; /* max bvl_vecs we can hold */
109
110 struct bio_vec *bi_io_vec; /* the actual vec list */
111
112 bio_end_io_t *bi_end_io;
113 atomic_t bi_cnt; /* pin count */
114
115 void *bi_private;
7ba1ba12
MP		 116#if defined(CONFIG_BLK_DEV_INTEGRITY)
117 struct bio_integrity_payload *bi_integrity; /* data integrity */
118#endif
1da177e4
LT		 119
120 bio_destructor_t *bi_destructor; /* destructor */
1da177e4
LT		 121};
122
123/*
124 * bio flags
125 */
126#define BIO_UPTODATE 0 /* ok after I/O completion */
127#define BIO_RW_BLOCK 1 /* RW_AHEAD set, and read/write would block */
128#define BIO_EOF 2 /* out-out-bounds error */
129#define BIO_SEG_VALID 3 /* nr_hw_seg valid */
130#define BIO_CLONED 4 /* doesn't own data */
131#define BIO_BOUNCED 5 /* bio is a bounce bio */
132#define BIO_USER_MAPPED 6 /* contains user pages */
133#define BIO_EOPNOTSUPP 7 /* not supported */
134#define bio_flagged(bio, flag) ((bio)->bi_flags & (1 << (flag)))
135
136/*
137 * top 4 bits of bio flags indicate the pool this bio came from
138 */
139#define BIO_POOL_BITS (4)
140#define BIO_POOL_OFFSET (BITS_PER_LONG - BIO_POOL_BITS)
141#define BIO_POOL_MASK (1UL << BIO_POOL_OFFSET)
142#define BIO_POOL_IDX(bio) ((bio)->bi_flags >> BIO_POOL_OFFSET)
143
144/*
145 * bio bi_rw flags
146 *
147 * bit 0 -- read (not set) or write (set)
148 * bit 1 -- rw-ahead when set
149 * bit 2 -- barrier
150 * bit 3 -- fail fast, don't want low level driver retries
151 * bit 4 -- synchronous I/O hint: the block layer will unplug immediately
fb2dce86
DW		 152 * bit 5 -- metadata request
153 * bit 6 -- discard sectors
1da177e4 154 */
d628eaef
DW		 155#define BIO_RW 0 /* Must match RW in req flags (blkdev.h) */
156#define BIO_RW_AHEAD 1 /* Must match FAILFAST in req flags */
1da177e4
LT		 157#define BIO_RW_BARRIER 2
158#define BIO_RW_FAILFAST 3
159#define BIO_RW_SYNC 4
5404bc7a 160#define BIO_RW_META 5
fb2dce86 161#define BIO_RW_DISCARD 6
1da177e4 162
22e2c507
JA		 163/*
164 * upper 16 bits of bi_rw define the io priority of this bio
165 */
166#define BIO_PRIO_SHIFT (8 * sizeof(unsigned long) - IOPRIO_BITS)
167#define bio_prio(bio) ((bio)->bi_rw >> BIO_PRIO_SHIFT)
168#define bio_prio_valid(bio) ioprio_valid(bio_prio(bio))
169
170#define bio_set_prio(bio, prio) do { \
171 WARN_ON(prio >= (1 << IOPRIO_BITS)); \
172 (bio)->bi_rw &= ((1UL << BIO_PRIO_SHIFT) - 1); \
173 (bio)->bi_rw |= ((unsigned long) (prio) << BIO_PRIO_SHIFT); \
174} while (0)
175
1da177e4
LT		 176/*
177 * various member access, note that bio_data should of course not be used
178 * on highmem page vectors
179 */
180#define bio_iovec_idx(bio, idx) (&((bio)->bi_io_vec[(idx)]))
181#define bio_iovec(bio) bio_iovec_idx((bio), (bio)->bi_idx)
182#define bio_page(bio) bio_iovec((bio))->bv_page
183#define bio_offset(bio) bio_iovec((bio))->bv_offset
184#define bio_segments(bio) ((bio)->bi_vcnt - (bio)->bi_idx)
185#define bio_sectors(bio) ((bio)->bi_size >> 9)
1da177e4
LT		 186#define bio_barrier(bio) ((bio)->bi_rw & (1 << BIO_RW_BARRIER))
187#define bio_sync(bio) ((bio)->bi_rw & (1 << BIO_RW_SYNC))
188#define bio_failfast(bio) ((bio)->bi_rw & (1 << BIO_RW_FAILFAST))
189#define bio_rw_ahead(bio) ((bio)->bi_rw & (1 << BIO_RW_AHEAD))
5404bc7a 190#define bio_rw_meta(bio) ((bio)->bi_rw & (1 << BIO_RW_META))
a9c701e5 191#define bio_empty_barrier(bio) (bio_barrier(bio) && !bio_has_data(bio))
fb2dce86 192#define bio_discard(bio) ((bio)->bi_rw & (1 << BIO_RW_DISCARD))
bf2de6f5
JA		 193
194static inline unsigned int bio_cur_sectors(struct bio *bio)
195{
196 if (bio->bi_vcnt)
197 return bio_iovec(bio)->bv_len >> 9;
fb2dce86
DW		 198 else /* dataless requests such as discard */
199 return bio->bi_size >> 9;
bf2de6f5
JA		 200}
201
202static inline void *bio_data(struct bio *bio)
203{
204 if (bio->bi_vcnt)
205 return page_address(bio_page(bio)) + bio_offset(bio);
206
207 return NULL;
208}
1da177e4
LT		 209
210/*
211 * will die
212 */
213#define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long) bio_offset((bio)))
214#define bvec_to_phys(bv) (page_to_phys((bv)->bv_page) + (unsigned long) (bv)->bv_offset)
215
216/*
217 * queues that have highmem support enabled may still need to revert to
218 * PIO transfers occasionally and thus map high pages temporarily. For
219 * permanent PIO fall back, user is probably better off disabling highmem
220 * I/O completely on that queue (see ide-dma for example)
221 */
222#define __bio_kmap_atomic(bio, idx, kmtype) \
223 (kmap_atomic(bio_iovec_idx((bio), (idx))->bv_page, kmtype) + \
224 bio_iovec_idx((bio), (idx))->bv_offset)
225
226#define __bio_kunmap_atomic(addr, kmtype) kunmap_atomic(addr, kmtype)
227
228/*
229 * merge helpers etc
230 */
231
232#define __BVEC_END(bio) bio_iovec_idx((bio), (bio)->bi_vcnt - 1)
233#define __BVEC_START(bio) bio_iovec_idx((bio), (bio)->bi_idx)
234
235/*
236 * allow arch override, for eg virtualized architectures (put in asm/io.h)
237 */
238#ifndef BIOVEC_PHYS_MERGEABLE
239#define BIOVEC_PHYS_MERGEABLE(vec1, vec2) \
240 ((bvec_to_phys((vec1)) + (vec1)->bv_len) == bvec_to_phys((vec2)))
241#endif
242
243#define BIOVEC_VIRT_MERGEABLE(vec1, vec2) \
244 ((((bvec_to_phys((vec1)) + (vec1)->bv_len) | bvec_to_phys((vec2))) & (BIO_VMERGE_BOUNDARY - 1)) == 0)
245#define __BIO_SEG_BOUNDARY(addr1, addr2, mask) \
246 (((addr1) | (mask)) == (((addr2) - 1) | (mask)))
247#define BIOVEC_SEG_BOUNDARY(q, b1, b2) \
248 __BIO_SEG_BOUNDARY(bvec_to_phys((b1)), bvec_to_phys((b2)) + (b2)->bv_len, (q)->seg_boundary_mask)
249#define BIO_SEG_BOUNDARY(q, b1, b2) \
250 BIOVEC_SEG_BOUNDARY((q), __BVEC_END((b1)), __BVEC_START((b2)))
251
6712ecf8 252#define bio_io_error(bio) bio_endio((bio), -EIO)
1da177e4
LT		 253
254/*
255 * drivers should not use the __ version unless they _really_ want to
256 * run through the entire bio and not just pending pieces
257 */
258#define __bio_for_each_segment(bvl, bio, i, start_idx) \
259 for (bvl = bio_iovec_idx((bio), (start_idx)), i = (start_idx); \
260 i < (bio)->bi_vcnt; \
261 bvl++, i++)
262
263#define bio_for_each_segment(bvl, bio, i) \
264 __bio_for_each_segment(bvl, bio, i, (bio)->bi_idx)
265
266/*
267 * get a reference to a bio, so it won't disappear. the intended use is
268 * something like:
269 *
270 * bio_get(bio);
271 * submit_bio(rw, bio);
272 * if (bio->bi_flags ...)
273 * do_something
274 * bio_put(bio);
275 *
276 * without the bio_get(), it could potentially complete I/O before submit_bio
277 * returns. and then bio would be freed memory when if (bio->bi_flags ...)
278 * runs
279 */
280#define bio_get(bio) atomic_inc(&(bio)->bi_cnt)
281
7ba1ba12
MP		 282#if defined(CONFIG_BLK_DEV_INTEGRITY)
283/*
284 * bio integrity payload
285 */
286struct bio_integrity_payload {
287 struct bio *bip_bio; /* parent bio */
288 struct bio_vec *bip_vec; /* integrity data vector */
289
290 sector_t bip_sector; /* virtual start sector */
291
292 void *bip_buf; /* generated integrity data */
293 bio_end_io_t *bip_end_io; /* saved I/O completion fn */
294
295 int bip_error; /* saved I/O error */
296 unsigned int bip_size;
297
298 unsigned short bip_pool; /* pool the ivec came from */
299 unsigned short bip_vcnt; /* # of integrity bio_vecs */
300 unsigned short bip_idx; /* current bip_vec index */
301
302 struct work_struct bip_work; /* I/O completion */
303};
304#endif /* CONFIG_BLK_DEV_INTEGRITY */
1da177e4
LT		 305
306/*
307 * A bio_pair is used when we need to split a bio.
308 * This can only happen for a bio that refers to just one
309 * page of data, and in the unusual situation when the
310 * page crosses a chunk/device boundary
311 *
312 * The address of the master bio is stored in bio1.bi_private
313 * The address of the pool the pair was allocated from is stored
314 * in bio2.bi_private
315 */
316struct bio_pair {
7ba1ba12
MP		 317 struct bio bio1, bio2;
318 struct bio_vec bv1, bv2;
319#if defined(CONFIG_BLK_DEV_INTEGRITY)
320 struct bio_integrity_payload bip1, bip2;
321 struct bio_vec iv1, iv2;
322#endif
323 atomic_t cnt;
324 int error;
1da177e4
LT		 325};
326extern struct bio_pair *bio_split(struct bio *bi, mempool_t *pool,
327 int first_sectors);
328extern mempool_t *bio_split_pool;
329extern void bio_pair_release(struct bio_pair *dbio);
330
5972511b 331extern struct bio_set *bioset_create(int, int);
1da177e4
LT		 332extern void bioset_free(struct bio_set *);
333
dd0fc66f
AV		 334extern struct bio *bio_alloc(gfp_t, int);
335extern struct bio *bio_alloc_bioset(gfp_t, int, struct bio_set *);
1da177e4 336extern void bio_put(struct bio *);
3676347a 337extern void bio_free(struct bio *, struct bio_set *);
1da177e4 338
6712ecf8 339extern void bio_endio(struct bio *, int);
1da177e4
LT		 340struct request_queue;
341extern int bio_phys_segments(struct request_queue *, struct bio *);
342extern int bio_hw_segments(struct request_queue *, struct bio *);
343
344extern void __bio_clone(struct bio *, struct bio *);
dd0fc66f 345extern struct bio *bio_clone(struct bio *, gfp_t);
1da177e4
LT		 346
347extern void bio_init(struct bio *);
348
349extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);
6e68af66
MC		 350extern int bio_add_pc_page(struct request_queue *, struct bio *, struct page *,
351 unsigned int, unsigned int);
1da177e4
LT		 352extern int bio_get_nr_vecs(struct block_device *);
353extern struct bio *bio_map_user(struct request_queue *, struct block_device *,
354 unsigned long, unsigned int, int);
f1970baf
JB		 355struct sg_iovec;
356extern struct bio *bio_map_user_iov(struct request_queue *,
357 struct block_device *,
358 struct sg_iovec *, int, int);
1da177e4 359extern void bio_unmap_user(struct bio *);
df46b9a4 360extern struct bio *bio_map_kern(struct request_queue *, void *, unsigned int,
27496a8c 361 gfp_t);
68154e90
FT		 362extern struct bio *bio_copy_kern(struct request_queue *, void *, unsigned int,
363 gfp_t, int);
1da177e4
LT		 364extern void bio_set_pages_dirty(struct bio *bio);
365extern void bio_check_pages_dirty(struct bio *bio);
366extern struct bio *bio_copy_user(struct request_queue *, unsigned long, unsigned int, int);
c5dec1c3
FT		 367extern struct bio *bio_copy_user_iov(struct request_queue *, struct sg_iovec *,
368 int, int);
1da177e4
LT		 369extern int bio_uncopy_user(struct bio *);
370void zero_fill_bio(struct bio *bio);
51d654e1 371extern struct bio_vec *bvec_alloc_bs(gfp_t, int, unsigned long *, struct bio_set *);
7ba1ba12 372extern unsigned int bvec_nr_vecs(unsigned short idx);
51d654e1
MP		 373
374/*
375 * bio_set is used to allow other portions of the IO system to
376 * allocate their own private memory pools for bio and iovec structures.
377 * These memory pools in turn all allocate from the bio_slab
378 * and the bvec_slabs[].
379 */
380#define BIO_POOL_SIZE 2
381#define BIOVEC_NR_POOLS 6
382
383struct bio_set {
384 mempool_t *bio_pool;
7ba1ba12
MP		 385#if defined(CONFIG_BLK_DEV_INTEGRITY)
386 mempool_t *bio_integrity_pool;
387#endif
51d654e1
MP		 388 mempool_t *bvec_pools[BIOVEC_NR_POOLS];
389};
390
391struct biovec_slab {
392 int nr_vecs;
393 char *name;
394 struct kmem_cache *slab;
395};
396
397extern struct bio_set *fs_bio_set;
398
399/*
400 * a small number of entries is fine, not going to be performance critical.
401 * basically we just need to survive
402 */
403#define BIO_SPLIT_ENTRIES 2
1da177e4
LT		 404
405#ifdef CONFIG_HIGHMEM
406/*
407 * remember to add offset! and never ever reenable interrupts between a
408 * bvec_kmap_irq and bvec_kunmap_irq!!
409 *
410 * This function MUST be inlined - it plays with the CPU interrupt flags.
1da177e4 411 */
c2d08dad 412static inline char *bvec_kmap_irq(struct bio_vec *bvec, unsigned long *flags)
1da177e4
LT		 413{
414 unsigned long addr;
415
416 /*
417 * might not be a highmem page, but the preempt/irq count
418 * balancing is a lot nicer this way
419 */
420 local_irq_save(*flags);
421 addr = (unsigned long) kmap_atomic(bvec->bv_page, KM_BIO_SRC_IRQ);
422
423 BUG_ON(addr & ~PAGE_MASK);
424
425 return (char *) addr + bvec->bv_offset;
426}
427
c2d08dad 428static inline void bvec_kunmap_irq(char *buffer, unsigned long *flags)
1da177e4
LT		 429{
430 unsigned long ptr = (unsigned long) buffer & PAGE_MASK;
431
432 kunmap_atomic((void *) ptr, KM_BIO_SRC_IRQ);
433 local_irq_restore(*flags);
434}
435
436#else
437#define bvec_kmap_irq(bvec, flags) (page_address((bvec)->bv_page) + (bvec)->bv_offset)
438#define bvec_kunmap_irq(buf, flags) do { *(flags) = 0; } while (0)
439#endif
440
c2d08dad 441static inline char *__bio_kmap_irq(struct bio *bio, unsigned short idx,
1da177e4
LT		 442 unsigned long *flags)
443{
444 return bvec_kmap_irq(bio_iovec_idx(bio, idx), flags);
445}
446#define __bio_kunmap_irq(buf, flags) bvec_kunmap_irq(buf, flags)
447
448#define bio_kmap_irq(bio, flags) \
449 __bio_kmap_irq((bio), (bio)->bi_idx, (flags))
450#define bio_kunmap_irq(buf,flags) __bio_kunmap_irq(buf, flags)
451
7a67f63b
JA		 452/*
453 * Check whether this bio carries any data or not. A NULL bio is allowed.
454 */
455static inline int bio_has_data(struct bio *bio)
456{
457 return bio && bio->bi_io_vec != NULL;
458}
459
7ba1ba12
MP		 460#if defined(CONFIG_BLK_DEV_INTEGRITY)
461
462#define bip_vec_idx(bip, idx) (&(bip->bip_vec[(idx)]))
463#define bip_vec(bip) bip_vec_idx(bip, 0)
464
465#define __bip_for_each_vec(bvl, bip, i, start_idx) \
466 for (bvl = bip_vec_idx((bip), (start_idx)), i = (start_idx); \
467 i < (bip)->bip_vcnt; \
468 bvl++, i++)
469
470#define bip_for_each_vec(bvl, bip, i) \
471 __bip_for_each_vec(bvl, bip, i, (bip)->bip_idx)
472
6e2401ad
JA		 473static inline int bio_integrity(struct bio *bio)
474{
475#if defined(CONFIG_BLK_DEV_INTEGRITY)
476 return bio->bi_integrity != NULL;
477#else
478 return 0;
479#endif
480}
7ba1ba12
MP		 481
482extern struct bio_integrity_payload *bio_integrity_alloc_bioset(struct bio *, gfp_t, unsigned int, struct bio_set *);
483extern struct bio_integrity_payload *bio_integrity_alloc(struct bio *, gfp_t, unsigned int);
484extern void bio_integrity_free(struct bio *, struct bio_set *);
485extern int bio_integrity_add_page(struct bio *, struct page *, unsigned int, unsigned int);
486extern int bio_integrity_enabled(struct bio *bio);
487extern int bio_integrity_set_tag(struct bio *, void *, unsigned int);
488extern int bio_integrity_get_tag(struct bio *, void *, unsigned int);
489extern int bio_integrity_prep(struct bio *);
490extern void bio_integrity_endio(struct bio *, int);
491extern void bio_integrity_advance(struct bio *, unsigned int);
492extern void bio_integrity_trim(struct bio *, unsigned int, unsigned int);
493extern void bio_integrity_split(struct bio *, struct bio_pair *, int);
494extern int bio_integrity_clone(struct bio *, struct bio *, struct bio_set *);
495extern int bioset_integrity_create(struct bio_set *, int);
496extern void bioset_integrity_free(struct bio_set *);
497extern void bio_integrity_init_slab(void);
498
499#else /* CONFIG_BLK_DEV_INTEGRITY */
500
501#define bio_integrity(a) (0)
502#define bioset_integrity_create(a, b) (0)
503#define bio_integrity_prep(a) (0)
504#define bio_integrity_enabled(a) (0)
505#define bio_integrity_clone(a, b, c) (0)
506#define bioset_integrity_free(a) do { } while (0)
507#define bio_integrity_free(a, b) do { } while (0)
508#define bio_integrity_endio(a, b) do { } while (0)
509#define bio_integrity_advance(a, b) do { } while (0)
510#define bio_integrity_trim(a, b, c) do { } while (0)
511#define bio_integrity_split(a, b, c) do { } while (0)
512#define bio_integrity_set_tag(a, b, c) do { } while (0)
513#define bio_integrity_get_tag(a, b, c) do { } while (0)
514#define bio_integrity_init_slab(a) do { } while (0)
515
516#endif /* CONFIG_BLK_DEV_INTEGRITY */
517
02a5e0ac 518#endif /* CONFIG_BLOCK */
1da177e4 519#endif /* __LINUX_BIO_H */
Linux 6.x block layer and io_uring tree(s)