blob: df99c882b807549f25c30b13416a3ae8ca43d7e0 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Jens Axboe0fe23472006-09-04 15:41:16 +02002 * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk>
Linus Torvalds1da177e2005-04-16 15:20:36 -07003 *
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public Licens
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
16 *
17 */
18#include <linux/mm.h>
19#include <linux/swap.h>
20#include <linux/bio.h>
21#include <linux/blkdev.h>
22#include <linux/slab.h>
23#include <linux/init.h>
24#include <linux/kernel.h>
25#include <linux/module.h>
26#include <linux/mempool.h>
27#include <linux/workqueue.h>
Jens Axboe2056a782006-03-23 20:00:26 +010028#include <linux/blktrace_api.h>
Arnaldo Carvalho de Melo5f3ea372008-10-30 08:34:33 +010029#include <trace/block.h>
James Bottomley f1970ba2005-06-20 14:06:52 +020030#include <scsi/sg.h> /* for struct sg_iovec */
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
Ingo Molnar0bfc2452008-11-26 11:59:56 +010032DEFINE_TRACE(block_split);
33
Christoph Lametere18b8902006-12-06 20:33:20 -080034static struct kmem_cache *bio_slab __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Denis ChengRq6feef532008-10-09 08:57:05 +020036static mempool_t *bio_split_pool __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Linus Torvalds1da177e2005-04-16 15:20:36 -070038/*
39 * if you change this list, also change bvec_alloc or things will
40 * break badly! cannot be bigger than what you can fit into an
41 * unsigned short
42 */
43
44#define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
Christoph Lameter6c036522005-07-07 17:56:59 -070045static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES),
47};
48#undef BV
49
50/*
Linus Torvalds1da177e2005-04-16 15:20:36 -070051 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
52 * IO code that does not need private memory pools.
53 */
Martin K. Petersen51d654e2008-06-17 18:59:56 +020054struct bio_set *fs_bio_set;
Linus Torvalds1da177e2005-04-16 15:20:36 -070055
Martin K. Petersen7ba1ba12008-06-30 20:04:41 +020056unsigned int bvec_nr_vecs(unsigned short idx)
57{
58 return bvec_slabs[idx].nr_vecs;
59}
60
Martin K. Petersen51d654e2008-06-17 18:59:56 +020061struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, struct bio_set *bs)
Linus Torvalds1da177e2005-04-16 15:20:36 -070062{
63 struct bio_vec *bvl;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
65 /*
Jens Axboe0a0d96b2008-09-11 13:17:37 +020066 * If 'bs' is given, lookup the pool and do the mempool alloc.
67 * If not, this is a bio_kmalloc() allocation and just do a
68 * kzalloc() for the exact number of vecs right away.
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 */
Jens Axboe0a0d96b2008-09-11 13:17:37 +020070 if (bs) {
71 /*
72 * see comment near bvec_array define!
73 */
74 switch (nr) {
75 case 1:
76 *idx = 0;
77 break;
78 case 2 ... 4:
79 *idx = 1;
80 break;
81 case 5 ... 16:
82 *idx = 2;
83 break;
84 case 17 ... 64:
85 *idx = 3;
86 break;
87 case 65 ... 128:
88 *idx = 4;
89 break;
90 case 129 ... BIO_MAX_PAGES:
91 *idx = 5;
92 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 default:
94 return NULL;
Jens Axboe0a0d96b2008-09-11 13:17:37 +020095 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
Jens Axboe0a0d96b2008-09-11 13:17:37 +020097 /*
98 * idx now points to the pool we want to allocate from
99 */
100 bvl = mempool_alloc(bs->bvec_pools[*idx], gfp_mask);
101 if (bvl)
102 memset(bvl, 0,
103 bvec_nr_vecs(*idx) * sizeof(struct bio_vec));
104 } else
105 bvl = kzalloc(nr * sizeof(struct bio_vec), gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
107 return bvl;
108}
109
Peter Osterlund36763472005-09-06 15:16:42 -0700110void bio_free(struct bio *bio, struct bio_set *bio_set)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111{
Jens Axboe992c5dd2007-07-18 13:18:08 +0200112 if (bio->bi_io_vec) {
113 const int pool_idx = BIO_POOL_IDX(bio);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
Jens Axboe992c5dd2007-07-18 13:18:08 +0200115 BIO_BUG_ON(pool_idx >= BIOVEC_NR_POOLS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Jens Axboe992c5dd2007-07-18 13:18:08 +0200117 mempool_free(bio->bi_io_vec, bio_set->bvec_pools[pool_idx]);
118 }
119
Martin K. Petersen7ba1ba12008-06-30 20:04:41 +0200120 if (bio_integrity(bio))
121 bio_integrity_free(bio, bio_set);
122
Peter Osterlund36763472005-09-06 15:16:42 -0700123 mempool_free(bio, bio_set->bio_pool);
124}
125
126/*
127 * default destructor for a bio allocated with bio_alloc_bioset()
128 */
129static void bio_fs_destructor(struct bio *bio)
130{
131 bio_free(bio, fs_bio_set);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132}
133
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200134static void bio_kmalloc_destructor(struct bio *bio)
135{
136 kfree(bio->bi_io_vec);
137 kfree(bio);
138}
139
Arjan van de Ven858119e2006-01-14 13:20:43 -0800140void bio_init(struct bio *bio)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141{
Jens Axboe2b94de52007-07-18 13:14:03 +0200142 memset(bio, 0, sizeof(*bio));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 bio->bi_flags = 1 << BIO_UPTODATE;
Jens Axboec7c22e42008-09-13 20:26:01 +0200144 bio->bi_comp_cpu = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 atomic_set(&bio->bi_cnt, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146}
147
148/**
149 * bio_alloc_bioset - allocate a bio for I/O
150 * @gfp_mask: the GFP_ mask given to the slab allocator
151 * @nr_iovecs: number of iovecs to pre-allocate
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200152 * @bs: the bio_set to allocate from. If %NULL, just use kmalloc
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 *
154 * Description:
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200155 * bio_alloc_bioset will first try its own mempool to satisfy the allocation.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 * If %__GFP_WAIT is set then we will block on the internal pool waiting
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200157 * for a &struct bio to become free. If a %NULL @bs is passed in, we will
158 * fall back to just using @kmalloc to allocate the required memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 *
160 * allocate bio and iovecs from the memory pools specified by the
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200161 * bio_set structure, or @kmalloc if none given.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 **/
Al Virodd0fc662005-10-07 07:46:04 +0100163struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164{
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200165 struct bio *bio;
166
167 if (bs)
168 bio = mempool_alloc(bs->bio_pool, gfp_mask);
169 else
170 bio = kmalloc(sizeof(*bio), gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
172 if (likely(bio)) {
173 struct bio_vec *bvl = NULL;
174
175 bio_init(bio);
176 if (likely(nr_iovecs)) {
Jens Axboeeeae1d42008-05-07 13:26:27 +0200177 unsigned long uninitialized_var(idx);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178
179 bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs);
180 if (unlikely(!bvl)) {
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200181 if (bs)
182 mempool_free(bio, bs->bio_pool);
183 else
184 kfree(bio);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 bio = NULL;
186 goto out;
187 }
188 bio->bi_flags |= idx << BIO_POOL_OFFSET;
Denis ChengRq1ac0ae02008-08-04 11:56:30 +0200189 bio->bi_max_vecs = bvec_nr_vecs(idx);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 }
191 bio->bi_io_vec = bvl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 }
193out:
194 return bio;
195}
196
Al Virodd0fc662005-10-07 07:46:04 +0100197struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198{
Peter Osterlund36763472005-09-06 15:16:42 -0700199 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
200
201 if (bio)
202 bio->bi_destructor = bio_fs_destructor;
203
204 return bio;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205}
206
Jens Axboe0a0d96b2008-09-11 13:17:37 +0200207/*
208 * Like bio_alloc(), but doesn't use a mempool backing. This means that
209 * it CAN fail, but while bio_alloc() can only be used for allocations
210 * that have a short (finite) life span, bio_kmalloc() should be used
211 * for more permanent bio allocations (like allocating some bio's for
212 * initalization or setup purposes).
213 */
214struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs)
215{
216 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, NULL);
217
218 if (bio)
219 bio->bi_destructor = bio_kmalloc_destructor;
220
221 return bio;
222}
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224void zero_fill_bio(struct bio *bio)
225{
226 unsigned long flags;
227 struct bio_vec *bv;
228 int i;
229
230 bio_for_each_segment(bv, bio, i) {
231 char *data = bvec_kmap_irq(bv, &flags);
232 memset(data, 0, bv->bv_len);
233 flush_dcache_page(bv->bv_page);
234 bvec_kunmap_irq(data, &flags);
235 }
236}
237EXPORT_SYMBOL(zero_fill_bio);
238
239/**
240 * bio_put - release a reference to a bio
241 * @bio: bio to release reference to
242 *
243 * Description:
244 * Put a reference to a &struct bio, either one you have gotten with
245 * bio_alloc or bio_get. The last put of a bio will free it.
246 **/
247void bio_put(struct bio *bio)
248{
249 BIO_BUG_ON(!atomic_read(&bio->bi_cnt));
250
251 /*
252 * last put frees it
253 */
254 if (atomic_dec_and_test(&bio->bi_cnt)) {
255 bio->bi_next = NULL;
256 bio->bi_destructor(bio);
257 }
258}
259
Jens Axboe165125e2007-07-24 09:28:11 +0200260inline int bio_phys_segments(struct request_queue *q, struct bio *bio)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261{
262 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
263 blk_recount_segments(q, bio);
264
265 return bio->bi_phys_segments;
266}
267
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268/**
269 * __bio_clone - clone a bio
270 * @bio: destination bio
271 * @bio_src: bio to clone
272 *
273 * Clone a &bio. Caller will own the returned bio, but not
274 * the actual data it points to. Reference count of returned
275 * bio will be one.
276 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800277void __bio_clone(struct bio *bio, struct bio *bio_src)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
Andrew Mortone525e152005-08-07 09:42:12 -0700279 memcpy(bio->bi_io_vec, bio_src->bi_io_vec,
280 bio_src->bi_max_vecs * sizeof(struct bio_vec));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281
Jens Axboe5d840702008-01-25 12:44:44 +0100282 /*
283 * most users will be overriding ->bi_bdev with a new target,
284 * so we don't set nor calculate new physical/hw segment counts here
285 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 bio->bi_sector = bio_src->bi_sector;
287 bio->bi_bdev = bio_src->bi_bdev;
288 bio->bi_flags |= 1 << BIO_CLONED;
289 bio->bi_rw = bio_src->bi_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 bio->bi_vcnt = bio_src->bi_vcnt;
291 bio->bi_size = bio_src->bi_size;
Andrew Mortona5453be2005-07-28 01:07:18 -0700292 bio->bi_idx = bio_src->bi_idx;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293}
294
295/**
296 * bio_clone - clone a bio
297 * @bio: bio to clone
298 * @gfp_mask: allocation priority
299 *
300 * Like __bio_clone, only also allocates the returned bio
301 */
Al Virodd0fc662005-10-07 07:46:04 +0100302struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303{
304 struct bio *b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, fs_bio_set);
305
Martin K. Petersen7ba1ba12008-06-30 20:04:41 +0200306 if (!b)
307 return NULL;
308
309 b->bi_destructor = bio_fs_destructor;
310 __bio_clone(b, bio);
311
312 if (bio_integrity(bio)) {
313 int ret;
314
315 ret = bio_integrity_clone(b, bio, fs_bio_set);
316
317 if (ret < 0)
318 return NULL;
Peter Osterlund36763472005-09-06 15:16:42 -0700319 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
321 return b;
322}
323
324/**
325 * bio_get_nr_vecs - return approx number of vecs
326 * @bdev: I/O target
327 *
328 * Return the approximate number of pages we can send to this target.
329 * There's no guarantee that you will be able to fit this number of pages
330 * into a bio, it does not account for dynamic restrictions that vary
331 * on offset.
332 */
333int bio_get_nr_vecs(struct block_device *bdev)
334{
Jens Axboe165125e2007-07-24 09:28:11 +0200335 struct request_queue *q = bdev_get_queue(bdev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 int nr_pages;
337
338 nr_pages = ((q->max_sectors << 9) + PAGE_SIZE - 1) >> PAGE_SHIFT;
339 if (nr_pages > q->max_phys_segments)
340 nr_pages = q->max_phys_segments;
341 if (nr_pages > q->max_hw_segments)
342 nr_pages = q->max_hw_segments;
343
344 return nr_pages;
345}
346
Jens Axboe165125e2007-07-24 09:28:11 +0200347static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page
Mike Christiedefd94b2005-12-05 02:37:06 -0600348 *page, unsigned int len, unsigned int offset,
349 unsigned short max_sectors)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350{
351 int retried_segments = 0;
352 struct bio_vec *bvec;
353
354 /*
355 * cloned bio must not modify vec list
356 */
357 if (unlikely(bio_flagged(bio, BIO_CLONED)))
358 return 0;
359
Jens Axboe80cfd542006-01-06 09:43:28 +0100360 if (((bio->bi_size + len) >> 9) > max_sectors)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 return 0;
362
Jens Axboe80cfd542006-01-06 09:43:28 +0100363 /*
364 * For filesystems with a blocksize smaller than the pagesize
365 * we will often be called with the same page as last time and
366 * a consecutive offset. Optimize this special case.
367 */
368 if (bio->bi_vcnt > 0) {
369 struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1];
370
371 if (page == prev->bv_page &&
372 offset == prev->bv_offset + prev->bv_len) {
373 prev->bv_len += len;
Alasdair G Kergoncc371e62008-07-03 09:53:43 +0200374
375 if (q->merge_bvec_fn) {
376 struct bvec_merge_data bvm = {
377 .bi_bdev = bio->bi_bdev,
378 .bi_sector = bio->bi_sector,
379 .bi_size = bio->bi_size,
380 .bi_rw = bio->bi_rw,
381 };
382
383 if (q->merge_bvec_fn(q, &bvm, prev) < len) {
384 prev->bv_len -= len;
385 return 0;
386 }
Jens Axboe80cfd542006-01-06 09:43:28 +0100387 }
388
389 goto done;
390 }
391 }
392
393 if (bio->bi_vcnt >= bio->bi_max_vecs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 return 0;
395
396 /*
397 * we might lose a segment or two here, but rather that than
398 * make this too complex.
399 */
400
401 while (bio->bi_phys_segments >= q->max_phys_segments
Mikulas Patocka5df97b92008-08-15 10:20:02 +0200402 || bio->bi_phys_segments >= q->max_hw_segments) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403
404 if (retried_segments)
405 return 0;
406
407 retried_segments = 1;
408 blk_recount_segments(q, bio);
409 }
410
411 /*
412 * setup the new entry, we might clear it again later if we
413 * cannot add the page
414 */
415 bvec = &bio->bi_io_vec[bio->bi_vcnt];
416 bvec->bv_page = page;
417 bvec->bv_len = len;
418 bvec->bv_offset = offset;
419
420 /*
421 * if queue has other restrictions (eg varying max sector size
422 * depending on offset), it can specify a merge_bvec_fn in the
423 * queue to get further control
424 */
425 if (q->merge_bvec_fn) {
Alasdair G Kergoncc371e62008-07-03 09:53:43 +0200426 struct bvec_merge_data bvm = {
427 .bi_bdev = bio->bi_bdev,
428 .bi_sector = bio->bi_sector,
429 .bi_size = bio->bi_size,
430 .bi_rw = bio->bi_rw,
431 };
432
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 /*
434 * merge_bvec_fn() returns number of bytes it can accept
435 * at this offset
436 */
Alasdair G Kergoncc371e62008-07-03 09:53:43 +0200437 if (q->merge_bvec_fn(q, &bvm, bvec) < len) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 bvec->bv_page = NULL;
439 bvec->bv_len = 0;
440 bvec->bv_offset = 0;
441 return 0;
442 }
443 }
444
445 /* If we may be able to merge these biovecs, force a recount */
Mikulas Patockab8b3e162008-08-15 10:15:19 +0200446 if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 bio->bi_flags &= ~(1 << BIO_SEG_VALID);
448
449 bio->bi_vcnt++;
450 bio->bi_phys_segments++;
Jens Axboe80cfd542006-01-06 09:43:28 +0100451 done:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 bio->bi_size += len;
453 return len;
454}
455
456/**
Mike Christie6e68af62005-11-11 05:30:27 -0600457 * bio_add_pc_page - attempt to add page to bio
Jens Axboefddfdea2006-01-31 15:24:34 +0100458 * @q: the target queue
Mike Christie6e68af62005-11-11 05:30:27 -0600459 * @bio: destination bio
460 * @page: page to add
461 * @len: vec entry length
462 * @offset: vec entry offset
463 *
464 * Attempt to add a page to the bio_vec maplist. This can fail for a
465 * number of reasons, such as the bio being full or target block
466 * device limitations. The target block device must allow bio's
467 * smaller than PAGE_SIZE, so it is always possible to add a single
468 * page to an empty bio. This should only be used by REQ_PC bios.
469 */
Jens Axboe165125e2007-07-24 09:28:11 +0200470int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page,
Mike Christie6e68af62005-11-11 05:30:27 -0600471 unsigned int len, unsigned int offset)
472{
Mike Christiedefd94b2005-12-05 02:37:06 -0600473 return __bio_add_page(q, bio, page, len, offset, q->max_hw_sectors);
Mike Christie6e68af62005-11-11 05:30:27 -0600474}
475
476/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 * bio_add_page - attempt to add page to bio
478 * @bio: destination bio
479 * @page: page to add
480 * @len: vec entry length
481 * @offset: vec entry offset
482 *
483 * Attempt to add a page to the bio_vec maplist. This can fail for a
484 * number of reasons, such as the bio being full or target block
485 * device limitations. The target block device must allow bio's
486 * smaller than PAGE_SIZE, so it is always possible to add a single
487 * page to an empty bio.
488 */
489int bio_add_page(struct bio *bio, struct page *page, unsigned int len,
490 unsigned int offset)
491{
Mike Christiedefd94b2005-12-05 02:37:06 -0600492 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
493 return __bio_add_page(q, bio, page, len, offset, q->max_sectors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494}
495
496struct bio_map_data {
497 struct bio_vec *iovecs;
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200498 struct sg_iovec *sgvecs;
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900499 int nr_sgvecs;
500 int is_our_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501};
502
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200503static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio,
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900504 struct sg_iovec *iov, int iov_count,
505 int is_our_pages)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506{
507 memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt);
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200508 memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count);
509 bmd->nr_sgvecs = iov_count;
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900510 bmd->is_our_pages = is_our_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 bio->bi_private = bmd;
512}
513
514static void bio_free_map_data(struct bio_map_data *bmd)
515{
516 kfree(bmd->iovecs);
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200517 kfree(bmd->sgvecs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 kfree(bmd);
519}
520
FUJITA Tomonori76029ff2008-08-25 20:36:08 +0200521static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count,
522 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523{
FUJITA Tomonori76029ff2008-08-25 20:36:08 +0200524 struct bio_map_data *bmd = kmalloc(sizeof(*bmd), gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525
526 if (!bmd)
527 return NULL;
528
FUJITA Tomonori76029ff2008-08-25 20:36:08 +0200529 bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask);
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200530 if (!bmd->iovecs) {
531 kfree(bmd);
532 return NULL;
533 }
534
FUJITA Tomonori76029ff2008-08-25 20:36:08 +0200535 bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask);
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200536 if (bmd->sgvecs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 return bmd;
538
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200539 kfree(bmd->iovecs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 kfree(bmd);
541 return NULL;
542}
543
FUJITA Tomonoriaefcc282008-08-25 20:36:08 +0200544static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs,
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900545 struct sg_iovec *iov, int iov_count, int uncopy,
546 int do_free_page)
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200547{
548 int ret = 0, i;
549 struct bio_vec *bvec;
550 int iov_idx = 0;
551 unsigned int iov_off = 0;
552 int read = bio_data_dir(bio) == READ;
553
554 __bio_for_each_segment(bvec, bio, i, 0) {
555 char *bv_addr = page_address(bvec->bv_page);
FUJITA Tomonoriaefcc282008-08-25 20:36:08 +0200556 unsigned int bv_len = iovecs[i].bv_len;
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200557
558 while (bv_len && iov_idx < iov_count) {
559 unsigned int bytes;
560 char *iov_addr;
561
562 bytes = min_t(unsigned int,
563 iov[iov_idx].iov_len - iov_off, bv_len);
564 iov_addr = iov[iov_idx].iov_base + iov_off;
565
566 if (!ret) {
567 if (!read && !uncopy)
568 ret = copy_from_user(bv_addr, iov_addr,
569 bytes);
570 if (read && uncopy)
571 ret = copy_to_user(iov_addr, bv_addr,
572 bytes);
573
574 if (ret)
575 ret = -EFAULT;
576 }
577
578 bv_len -= bytes;
579 bv_addr += bytes;
580 iov_addr += bytes;
581 iov_off += bytes;
582
583 if (iov[iov_idx].iov_len == iov_off) {
584 iov_idx++;
585 iov_off = 0;
586 }
587 }
588
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900589 if (do_free_page)
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200590 __free_page(bvec->bv_page);
591 }
592
593 return ret;
594}
595
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596/**
597 * bio_uncopy_user - finish previously mapped bio
598 * @bio: bio being terminated
599 *
600 * Free pages allocated from bio_copy_user() and write back data
601 * to user space in case of a read.
602 */
603int bio_uncopy_user(struct bio *bio)
604{
605 struct bio_map_data *bmd = bio->bi_private;
FUJITA Tomonori81882762008-09-02 16:20:19 +0900606 int ret = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
FUJITA Tomonori81882762008-09-02 16:20:19 +0900608 if (!bio_flagged(bio, BIO_NULL_MAPPED))
609 ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs,
610 bmd->nr_sgvecs, 1, bmd->is_our_pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 bio_free_map_data(bmd);
612 bio_put(bio);
613 return ret;
614}
615
616/**
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200617 * bio_copy_user_iov - copy user data to bio
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 * @q: destination block queue
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900619 * @map_data: pointer to the rq_map_data holding pages (if necessary)
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200620 * @iov: the iovec.
621 * @iov_count: number of elements in the iovec
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 * @write_to_vm: bool indicating writing to pages or not
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900623 * @gfp_mask: memory allocation flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 *
625 * Prepares and returns a bio for indirect user io, bouncing data
626 * to/from kernel pages as necessary. Must be paired with
627 * call bio_uncopy_user() on io completion.
628 */
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900629struct bio *bio_copy_user_iov(struct request_queue *q,
630 struct rq_map_data *map_data,
631 struct sg_iovec *iov, int iov_count,
632 int write_to_vm, gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 struct bio_map_data *bmd;
635 struct bio_vec *bvec;
636 struct page *page;
637 struct bio *bio;
638 int i, ret;
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200639 int nr_pages = 0;
640 unsigned int len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200642 for (i = 0; i < iov_count; i++) {
643 unsigned long uaddr;
644 unsigned long end;
645 unsigned long start;
646
647 uaddr = (unsigned long)iov[i].iov_base;
648 end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT;
649 start = uaddr >> PAGE_SHIFT;
650
651 nr_pages += end - start;
652 len += iov[i].iov_len;
653 }
654
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900655 bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 if (!bmd)
657 return ERR_PTR(-ENOMEM);
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 ret = -ENOMEM;
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900660 bio = bio_alloc(gfp_mask, nr_pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 if (!bio)
662 goto out_bmd;
663
664 bio->bi_rw |= (!write_to_vm << BIO_RW);
665
666 ret = 0;
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900667 i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 while (len) {
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900669 unsigned int bytes;
670
671 if (map_data)
672 bytes = 1U << (PAGE_SHIFT + map_data->page_order);
673 else
674 bytes = PAGE_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
676 if (bytes > len)
677 bytes = len;
678
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900679 if (map_data) {
680 if (i == map_data->nr_entries) {
681 ret = -ENOMEM;
682 break;
683 }
684 page = map_data->pages[i++];
685 } else
686 page = alloc_page(q->bounce_gfp | gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 if (!page) {
688 ret = -ENOMEM;
689 break;
690 }
691
Mike Christie0e75f902006-12-01 10:40:55 +0100692 if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
695 len -= bytes;
696 }
697
698 if (ret)
699 goto cleanup;
700
701 /*
702 * success
703 */
704 if (!write_to_vm) {
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900705 ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 0);
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200706 if (ret)
707 goto cleanup;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 }
709
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900710 bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 return bio;
712cleanup:
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900713 if (!map_data)
714 bio_for_each_segment(bvec, bio, i)
715 __free_page(bvec->bv_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716
717 bio_put(bio);
718out_bmd:
719 bio_free_map_data(bmd);
720 return ERR_PTR(ret);
721}
722
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200723/**
724 * bio_copy_user - copy user data to bio
725 * @q: destination block queue
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900726 * @map_data: pointer to the rq_map_data holding pages (if necessary)
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200727 * @uaddr: start of user address
728 * @len: length in bytes
729 * @write_to_vm: bool indicating writing to pages or not
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900730 * @gfp_mask: memory allocation flags
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200731 *
732 * Prepares and returns a bio for indirect user io, bouncing data
733 * to/from kernel pages as necessary. Must be paired with
734 * call bio_uncopy_user() on io completion.
735 */
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900736struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data,
737 unsigned long uaddr, unsigned int len,
738 int write_to_vm, gfp_t gfp_mask)
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200739{
740 struct sg_iovec iov;
741
742 iov.iov_base = (void __user *)uaddr;
743 iov.iov_len = len;
744
FUJITA Tomonori152e2832008-08-28 16:17:06 +0900745 return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask);
FUJITA Tomonoric5dec1c2008-04-11 12:56:49 +0200746}
747
Jens Axboe165125e2007-07-24 09:28:11 +0200748static struct bio *__bio_map_user_iov(struct request_queue *q,
James Bottomley f1970ba2005-06-20 14:06:52 +0200749 struct block_device *bdev,
750 struct sg_iovec *iov, int iov_count,
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900751 int write_to_vm, gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752{
James Bottomley f1970ba2005-06-20 14:06:52 +0200753 int i, j;
754 int nr_pages = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 struct page **pages;
756 struct bio *bio;
James Bottomley f1970ba2005-06-20 14:06:52 +0200757 int cur_page = 0;
758 int ret, offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
James Bottomley f1970ba2005-06-20 14:06:52 +0200760 for (i = 0; i < iov_count; i++) {
761 unsigned long uaddr = (unsigned long)iov[i].iov_base;
762 unsigned long len = iov[i].iov_len;
763 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
764 unsigned long start = uaddr >> PAGE_SHIFT;
765
766 nr_pages += end - start;
767 /*
Mike Christiead2d7222006-12-01 10:40:20 +0100768 * buffer must be aligned to at least hardsector size for now
James Bottomley f1970ba2005-06-20 14:06:52 +0200769 */
Mike Christiead2d7222006-12-01 10:40:20 +0100770 if (uaddr & queue_dma_alignment(q))
James Bottomley f1970ba2005-06-20 14:06:52 +0200771 return ERR_PTR(-EINVAL);
772 }
773
774 if (!nr_pages)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 return ERR_PTR(-EINVAL);
776
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900777 bio = bio_alloc(gfp_mask, nr_pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 if (!bio)
779 return ERR_PTR(-ENOMEM);
780
781 ret = -ENOMEM;
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900782 pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 if (!pages)
784 goto out;
785
James Bottomley f1970ba2005-06-20 14:06:52 +0200786 for (i = 0; i < iov_count; i++) {
787 unsigned long uaddr = (unsigned long)iov[i].iov_base;
788 unsigned long len = iov[i].iov_len;
789 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
790 unsigned long start = uaddr >> PAGE_SHIFT;
791 const int local_nr_pages = end - start;
792 const int page_limit = cur_page + local_nr_pages;
793
Nick Pigginf5dd33c2008-07-25 19:45:25 -0700794 ret = get_user_pages_fast(uaddr, local_nr_pages,
795 write_to_vm, &pages[cur_page]);
Jens Axboe99172152006-06-16 13:02:29 +0200796 if (ret < local_nr_pages) {
797 ret = -EFAULT;
James Bottomley f1970ba2005-06-20 14:06:52 +0200798 goto out_unmap;
Jens Axboe99172152006-06-16 13:02:29 +0200799 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
James Bottomley f1970ba2005-06-20 14:06:52 +0200801 offset = uaddr & ~PAGE_MASK;
802 for (j = cur_page; j < page_limit; j++) {
803 unsigned int bytes = PAGE_SIZE - offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804
James Bottomley f1970ba2005-06-20 14:06:52 +0200805 if (len <= 0)
806 break;
807
808 if (bytes > len)
809 bytes = len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810
James Bottomley f1970ba2005-06-20 14:06:52 +0200811 /*
812 * sorry...
813 */
Mike Christiedefd94b2005-12-05 02:37:06 -0600814 if (bio_add_pc_page(q, bio, pages[j], bytes, offset) <
815 bytes)
James Bottomley f1970ba2005-06-20 14:06:52 +0200816 break;
817
818 len -= bytes;
819 offset = 0;
820 }
821
822 cur_page = j;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 /*
James Bottomley f1970ba2005-06-20 14:06:52 +0200824 * release the pages we didn't map into the bio, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 */
James Bottomley f1970ba2005-06-20 14:06:52 +0200826 while (j < page_limit)
827 page_cache_release(pages[j++]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 }
829
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 kfree(pages);
831
832 /*
833 * set data direction, and check if mapped pages need bouncing
834 */
835 if (!write_to_vm)
836 bio->bi_rw |= (1 << BIO_RW);
837
James Bottomley f1970ba2005-06-20 14:06:52 +0200838 bio->bi_bdev = bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 bio->bi_flags |= (1 << BIO_USER_MAPPED);
840 return bio;
James Bottomley f1970ba2005-06-20 14:06:52 +0200841
842 out_unmap:
843 for (i = 0; i < nr_pages; i++) {
844 if(!pages[i])
845 break;
846 page_cache_release(pages[i]);
847 }
848 out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 kfree(pages);
850 bio_put(bio);
851 return ERR_PTR(ret);
852}
853
854/**
855 * bio_map_user - map user address into bio
Jens Axboe165125e2007-07-24 09:28:11 +0200856 * @q: the struct request_queue for the bio
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 * @bdev: destination block device
858 * @uaddr: start of user address
859 * @len: length in bytes
860 * @write_to_vm: bool indicating writing to pages or not
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900861 * @gfp_mask: memory allocation flags
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 *
863 * Map the user space address into a bio suitable for io to a block
864 * device. Returns an error pointer in case of error.
865 */
Jens Axboe165125e2007-07-24 09:28:11 +0200866struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev,
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900867 unsigned long uaddr, unsigned int len, int write_to_vm,
868 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869{
James Bottomley f1970ba2005-06-20 14:06:52 +0200870 struct sg_iovec iov;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871
viro@ZenIV.linux.org.uk3f703532005-09-09 16:53:56 +0100872 iov.iov_base = (void __user *)uaddr;
James Bottomley f1970ba2005-06-20 14:06:52 +0200873 iov.iov_len = len;
874
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900875 return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask);
James Bottomley f1970ba2005-06-20 14:06:52 +0200876}
877
878/**
879 * bio_map_user_iov - map user sg_iovec table into bio
Jens Axboe165125e2007-07-24 09:28:11 +0200880 * @q: the struct request_queue for the bio
James Bottomley f1970ba2005-06-20 14:06:52 +0200881 * @bdev: destination block device
882 * @iov: the iovec.
883 * @iov_count: number of elements in the iovec
884 * @write_to_vm: bool indicating writing to pages or not
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900885 * @gfp_mask: memory allocation flags
James Bottomley f1970ba2005-06-20 14:06:52 +0200886 *
887 * Map the user space address into a bio suitable for io to a block
888 * device. Returns an error pointer in case of error.
889 */
Jens Axboe165125e2007-07-24 09:28:11 +0200890struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev,
James Bottomley f1970ba2005-06-20 14:06:52 +0200891 struct sg_iovec *iov, int iov_count,
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900892 int write_to_vm, gfp_t gfp_mask)
James Bottomley f1970ba2005-06-20 14:06:52 +0200893{
894 struct bio *bio;
James Bottomley f1970ba2005-06-20 14:06:52 +0200895
FUJITA Tomonoria3bce902008-08-28 16:17:05 +0900896 bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm,
897 gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 if (IS_ERR(bio))
899 return bio;
900
901 /*
902 * subtle -- if __bio_map_user() ended up bouncing a bio,
903 * it would normally disappear when its bi_end_io is run.
904 * however, we need it for the unmap, so grab an extra
905 * reference to it
906 */
907 bio_get(bio);
908
Mike Christie0e75f902006-12-01 10:40:55 +0100909 return bio;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910}
911
912static void __bio_unmap_user(struct bio *bio)
913{
914 struct bio_vec *bvec;
915 int i;
916
917 /*
918 * make sure we dirty pages we wrote to
919 */
920 __bio_for_each_segment(bvec, bio, i, 0) {
921 if (bio_data_dir(bio) == READ)
922 set_page_dirty_lock(bvec->bv_page);
923
924 page_cache_release(bvec->bv_page);
925 }
926
927 bio_put(bio);
928}
929
930/**
931 * bio_unmap_user - unmap a bio
932 * @bio: the bio being unmapped
933 *
934 * Unmap a bio previously mapped by bio_map_user(). Must be called with
935 * a process context.
936 *
937 * bio_unmap_user() may sleep.
938 */
939void bio_unmap_user(struct bio *bio)
940{
941 __bio_unmap_user(bio);
942 bio_put(bio);
943}
944
NeilBrown6712ecf2007-09-27 12:47:43 +0200945static void bio_map_kern_endio(struct bio *bio, int err)
Jens Axboeb8238252005-06-20 14:05:27 +0200946{
Jens Axboeb8238252005-06-20 14:05:27 +0200947 bio_put(bio);
Jens Axboeb8238252005-06-20 14:05:27 +0200948}
949
950
Jens Axboe165125e2007-07-24 09:28:11 +0200951static struct bio *__bio_map_kern(struct request_queue *q, void *data,
Al Viro27496a82005-10-21 03:20:48 -0400952 unsigned int len, gfp_t gfp_mask)
Mike Christie df46b9a2005-06-20 14:04:44 +0200953{
954 unsigned long kaddr = (unsigned long)data;
955 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
956 unsigned long start = kaddr >> PAGE_SHIFT;
957 const int nr_pages = end - start;
958 int offset, i;
959 struct bio *bio;
960
961 bio = bio_alloc(gfp_mask, nr_pages);
962 if (!bio)
963 return ERR_PTR(-ENOMEM);
964
965 offset = offset_in_page(kaddr);
966 for (i = 0; i < nr_pages; i++) {
967 unsigned int bytes = PAGE_SIZE - offset;
968
969 if (len <= 0)
970 break;
971
972 if (bytes > len)
973 bytes = len;
974
Mike Christiedefd94b2005-12-05 02:37:06 -0600975 if (bio_add_pc_page(q, bio, virt_to_page(data), bytes,
976 offset) < bytes)
Mike Christie df46b9a2005-06-20 14:04:44 +0200977 break;
978
979 data += bytes;
980 len -= bytes;
981 offset = 0;
982 }
983
Jens Axboeb8238252005-06-20 14:05:27 +0200984 bio->bi_end_io = bio_map_kern_endio;
Mike Christie df46b9a2005-06-20 14:04:44 +0200985 return bio;
986}
987
988/**
989 * bio_map_kern - map kernel address into bio
Jens Axboe165125e2007-07-24 09:28:11 +0200990 * @q: the struct request_queue for the bio
Mike Christie df46b9a2005-06-20 14:04:44 +0200991 * @data: pointer to buffer to map
992 * @len: length in bytes
993 * @gfp_mask: allocation flags for bio allocation
994 *
995 * Map the kernel address into a bio suitable for io to a block
996 * device. Returns an error pointer in case of error.
997 */
Jens Axboe165125e2007-07-24 09:28:11 +0200998struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len,
Al Viro27496a82005-10-21 03:20:48 -0400999 gfp_t gfp_mask)
Mike Christie df46b9a2005-06-20 14:04:44 +02001000{
1001 struct bio *bio;
1002
1003 bio = __bio_map_kern(q, data, len, gfp_mask);
1004 if (IS_ERR(bio))
1005 return bio;
1006
1007 if (bio->bi_size == len)
1008 return bio;
1009
1010 /*
1011 * Don't support partial mappings.
1012 */
1013 bio_put(bio);
1014 return ERR_PTR(-EINVAL);
1015}
1016
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001017static void bio_copy_kern_endio(struct bio *bio, int err)
1018{
1019 struct bio_vec *bvec;
1020 const int read = bio_data_dir(bio) == READ;
FUJITA Tomonori76029ff2008-08-25 20:36:08 +02001021 struct bio_map_data *bmd = bio->bi_private;
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001022 int i;
FUJITA Tomonori76029ff2008-08-25 20:36:08 +02001023 char *p = bmd->sgvecs[0].iov_base;
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001024
1025 __bio_for_each_segment(bvec, bio, i, 0) {
1026 char *addr = page_address(bvec->bv_page);
FUJITA Tomonori76029ff2008-08-25 20:36:08 +02001027 int len = bmd->iovecs[i].bv_len;
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001028
1029 if (read && !err)
FUJITA Tomonori76029ff2008-08-25 20:36:08 +02001030 memcpy(p, addr, len);
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001031
1032 __free_page(bvec->bv_page);
FUJITA Tomonori76029ff2008-08-25 20:36:08 +02001033 p += len;
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001034 }
1035
FUJITA Tomonori76029ff2008-08-25 20:36:08 +02001036 bio_free_map_data(bmd);
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001037 bio_put(bio);
1038}
1039
1040/**
1041 * bio_copy_kern - copy kernel address into bio
1042 * @q: the struct request_queue for the bio
1043 * @data: pointer to buffer to copy
1044 * @len: length in bytes
1045 * @gfp_mask: allocation flags for bio and page allocation
Randy Dunlapffee0252008-04-30 09:08:54 +02001046 * @reading: data direction is READ
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001047 *
1048 * copy the kernel address into a bio suitable for io to a block
1049 * device. Returns an error pointer in case of error.
1050 */
1051struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len,
1052 gfp_t gfp_mask, int reading)
1053{
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001054 struct bio *bio;
1055 struct bio_vec *bvec;
FUJITA Tomonori4d8ab622008-08-28 15:05:57 +09001056 int i;
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001057
FUJITA Tomonori4d8ab622008-08-28 15:05:57 +09001058 bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask);
1059 if (IS_ERR(bio))
1060 return bio;
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001061
1062 if (!reading) {
1063 void *p = data;
1064
1065 bio_for_each_segment(bvec, bio, i) {
1066 char *addr = page_address(bvec->bv_page);
1067
1068 memcpy(addr, p, bvec->bv_len);
1069 p += bvec->bv_len;
1070 }
1071 }
1072
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001073 bio->bi_end_io = bio_copy_kern_endio;
FUJITA Tomonori76029ff2008-08-25 20:36:08 +02001074
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001075 return bio;
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001076}
1077
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078/*
1079 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1080 * for performing direct-IO in BIOs.
1081 *
1082 * The problem is that we cannot run set_page_dirty() from interrupt context
1083 * because the required locks are not interrupt-safe. So what we can do is to
1084 * mark the pages dirty _before_ performing IO. And in interrupt context,
1085 * check that the pages are still dirty. If so, fine. If not, redirty them
1086 * in process context.
1087 *
1088 * We special-case compound pages here: normally this means reads into hugetlb
1089 * pages. The logic in here doesn't really work right for compound pages
1090 * because the VM does not uniformly chase down the head page in all cases.
1091 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1092 * handle them at all. So we skip compound pages here at an early stage.
1093 *
1094 * Note that this code is very hard to test under normal circumstances because
1095 * direct-io pins the pages with get_user_pages(). This makes
1096 * is_page_cache_freeable return false, and the VM will not clean the pages.
1097 * But other code (eg, pdflush) could clean the pages if they are mapped
1098 * pagecache.
1099 *
1100 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1101 * deferred bio dirtying paths.
1102 */
1103
1104/*
1105 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1106 */
1107void bio_set_pages_dirty(struct bio *bio)
1108{
1109 struct bio_vec *bvec = bio->bi_io_vec;
1110 int i;
1111
1112 for (i = 0; i < bio->bi_vcnt; i++) {
1113 struct page *page = bvec[i].bv_page;
1114
1115 if (page && !PageCompound(page))
1116 set_page_dirty_lock(page);
1117 }
1118}
1119
Adrian Bunk86b6c7a2008-02-18 13:48:32 +01001120static void bio_release_pages(struct bio *bio)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121{
1122 struct bio_vec *bvec = bio->bi_io_vec;
1123 int i;
1124
1125 for (i = 0; i < bio->bi_vcnt; i++) {
1126 struct page *page = bvec[i].bv_page;
1127
1128 if (page)
1129 put_page(page);
1130 }
1131}
1132
1133/*
1134 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1135 * If they are, then fine. If, however, some pages are clean then they must
1136 * have been written out during the direct-IO read. So we take another ref on
1137 * the BIO and the offending pages and re-dirty the pages in process context.
1138 *
1139 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1140 * here on. It will run one page_cache_release() against each page and will
1141 * run one bio_put() against the BIO.
1142 */
1143
David Howells65f27f32006-11-22 14:55:48 +00001144static void bio_dirty_fn(struct work_struct *work);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
David Howells65f27f32006-11-22 14:55:48 +00001146static DECLARE_WORK(bio_dirty_work, bio_dirty_fn);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147static DEFINE_SPINLOCK(bio_dirty_lock);
1148static struct bio *bio_dirty_list;
1149
1150/*
1151 * This runs in process context
1152 */
David Howells65f27f32006-11-22 14:55:48 +00001153static void bio_dirty_fn(struct work_struct *work)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154{
1155 unsigned long flags;
1156 struct bio *bio;
1157
1158 spin_lock_irqsave(&bio_dirty_lock, flags);
1159 bio = bio_dirty_list;
1160 bio_dirty_list = NULL;
1161 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1162
1163 while (bio) {
1164 struct bio *next = bio->bi_private;
1165
1166 bio_set_pages_dirty(bio);
1167 bio_release_pages(bio);
1168 bio_put(bio);
1169 bio = next;
1170 }
1171}
1172
1173void bio_check_pages_dirty(struct bio *bio)
1174{
1175 struct bio_vec *bvec = bio->bi_io_vec;
1176 int nr_clean_pages = 0;
1177 int i;
1178
1179 for (i = 0; i < bio->bi_vcnt; i++) {
1180 struct page *page = bvec[i].bv_page;
1181
1182 if (PageDirty(page) || PageCompound(page)) {
1183 page_cache_release(page);
1184 bvec[i].bv_page = NULL;
1185 } else {
1186 nr_clean_pages++;
1187 }
1188 }
1189
1190 if (nr_clean_pages) {
1191 unsigned long flags;
1192
1193 spin_lock_irqsave(&bio_dirty_lock, flags);
1194 bio->bi_private = bio_dirty_list;
1195 bio_dirty_list = bio;
1196 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1197 schedule_work(&bio_dirty_work);
1198 } else {
1199 bio_put(bio);
1200 }
1201}
1202
1203/**
1204 * bio_endio - end I/O on a bio
1205 * @bio: bio
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 * @error: error, if any
1207 *
1208 * Description:
NeilBrown6712ecf2007-09-27 12:47:43 +02001209 * bio_endio() will end I/O on the whole bio. bio_endio() is the
NeilBrown5bb23a62007-09-27 12:46:13 +02001210 * preferred way to end I/O on a bio, it takes care of clearing
1211 * BIO_UPTODATE on error. @error is 0 on success, and and one of the
1212 * established -Exxxx (-EIO, for instance) error values in case
1213 * something went wrong. Noone should call bi_end_io() directly on a
1214 * bio unless they own it and thus know that it has an end_io
1215 * function.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 **/
NeilBrown6712ecf2007-09-27 12:47:43 +02001217void bio_endio(struct bio *bio, int error)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218{
1219 if (error)
1220 clear_bit(BIO_UPTODATE, &bio->bi_flags);
NeilBrown9cc54d42007-09-27 12:46:12 +02001221 else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
1222 error = -EIO;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223
NeilBrown5bb23a62007-09-27 12:46:13 +02001224 if (bio->bi_end_io)
NeilBrown6712ecf2007-09-27 12:47:43 +02001225 bio->bi_end_io(bio, error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226}
1227
1228void bio_pair_release(struct bio_pair *bp)
1229{
1230 if (atomic_dec_and_test(&bp->cnt)) {
1231 struct bio *master = bp->bio1.bi_private;
1232
NeilBrown6712ecf2007-09-27 12:47:43 +02001233 bio_endio(master, bp->error);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 mempool_free(bp, bp->bio2.bi_private);
1235 }
1236}
1237
NeilBrown6712ecf2007-09-27 12:47:43 +02001238static void bio_pair_end_1(struct bio *bi, int err)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239{
1240 struct bio_pair *bp = container_of(bi, struct bio_pair, bio1);
1241
1242 if (err)
1243 bp->error = err;
1244
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 bio_pair_release(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246}
1247
NeilBrown6712ecf2007-09-27 12:47:43 +02001248static void bio_pair_end_2(struct bio *bi, int err)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249{
1250 struct bio_pair *bp = container_of(bi, struct bio_pair, bio2);
1251
1252 if (err)
1253 bp->error = err;
1254
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 bio_pair_release(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256}
1257
1258/*
1259 * split a bio - only worry about a bio with a single page
1260 * in it's iovec
1261 */
Denis ChengRq6feef532008-10-09 08:57:05 +02001262struct bio_pair *bio_split(struct bio *bi, int first_sectors)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263{
Denis ChengRq6feef532008-10-09 08:57:05 +02001264 struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265
1266 if (!bp)
1267 return bp;
1268
Arnaldo Carvalho de Melo5f3ea372008-10-30 08:34:33 +01001269 trace_block_split(bdev_get_queue(bi->bi_bdev), bi,
Jens Axboe2056a782006-03-23 20:00:26 +01001270 bi->bi_sector + first_sectors);
1271
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 BUG_ON(bi->bi_vcnt != 1);
1273 BUG_ON(bi->bi_idx != 0);
1274 atomic_set(&bp->cnt, 3);
1275 bp->error = 0;
1276 bp->bio1 = *bi;
1277 bp->bio2 = *bi;
1278 bp->bio2.bi_sector += first_sectors;
1279 bp->bio2.bi_size -= first_sectors << 9;
1280 bp->bio1.bi_size = first_sectors << 9;
1281
1282 bp->bv1 = bi->bi_io_vec[0];
1283 bp->bv2 = bi->bi_io_vec[0];
1284 bp->bv2.bv_offset += first_sectors << 9;
1285 bp->bv2.bv_len -= first_sectors << 9;
1286 bp->bv1.bv_len = first_sectors << 9;
1287
1288 bp->bio1.bi_io_vec = &bp->bv1;
1289 bp->bio2.bi_io_vec = &bp->bv2;
1290
NeilBrowna2eb0c12006-05-22 22:35:27 -07001291 bp->bio1.bi_max_vecs = 1;
1292 bp->bio2.bi_max_vecs = 1;
1293
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 bp->bio1.bi_end_io = bio_pair_end_1;
1295 bp->bio2.bi_end_io = bio_pair_end_2;
1296
1297 bp->bio1.bi_private = bi;
Denis ChengRq6feef532008-10-09 08:57:05 +02001298 bp->bio2.bi_private = bio_split_pool;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
Martin K. Petersen7ba1ba12008-06-30 20:04:41 +02001300 if (bio_integrity(bi))
1301 bio_integrity_split(bi, bp, first_sectors);
1302
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 return bp;
1304}
1305
Martin K. Petersenad3316b2008-10-01 22:42:53 -04001306/**
1307 * bio_sector_offset - Find hardware sector offset in bio
1308 * @bio: bio to inspect
1309 * @index: bio_vec index
1310 * @offset: offset in bv_page
1311 *
1312 * Return the number of hardware sectors between beginning of bio
1313 * and an end point indicated by a bio_vec index and an offset
1314 * within that vector's page.
1315 */
1316sector_t bio_sector_offset(struct bio *bio, unsigned short index,
1317 unsigned int offset)
1318{
1319 unsigned int sector_sz = queue_hardsect_size(bio->bi_bdev->bd_disk->queue);
1320 struct bio_vec *bv;
1321 sector_t sectors;
1322 int i;
1323
1324 sectors = 0;
1325
1326 if (index >= bio->bi_idx)
1327 index = bio->bi_vcnt - 1;
1328
1329 __bio_for_each_segment(bv, bio, i, 0) {
1330 if (i == index) {
1331 if (offset > bv->bv_offset)
1332 sectors += (offset - bv->bv_offset) / sector_sz;
1333 break;
1334 }
1335
1336 sectors += bv->bv_len / sector_sz;
1337 }
1338
1339 return sectors;
1340}
1341EXPORT_SYMBOL(bio_sector_offset);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
1343/*
1344 * create memory pools for biovec's in a bio_set.
1345 * use the global biovec slabs created for general use.
1346 */
Jens Axboe59725112007-04-02 10:06:42 +02001347static int biovec_create_pools(struct bio_set *bs, int pool_entries)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348{
1349 int i;
1350
1351 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1352 struct biovec_slab *bp = bvec_slabs + i;
1353 mempool_t **bvp = bs->bvec_pools + i;
1354
Matthew Dobson93d23412006-03-26 01:37:50 -08001355 *bvp = mempool_create_slab_pool(pool_entries, bp->slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 if (!*bvp)
1357 return -ENOMEM;
1358 }
1359 return 0;
1360}
1361
1362static void biovec_free_pools(struct bio_set *bs)
1363{
1364 int i;
1365
1366 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1367 mempool_t *bvp = bs->bvec_pools[i];
1368
1369 if (bvp)
1370 mempool_destroy(bvp);
1371 }
1372
1373}
1374
1375void bioset_free(struct bio_set *bs)
1376{
1377 if (bs->bio_pool)
1378 mempool_destroy(bs->bio_pool);
1379
Martin K. Petersen7ba1ba12008-06-30 20:04:41 +02001380 bioset_integrity_free(bs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 biovec_free_pools(bs);
1382
1383 kfree(bs);
1384}
1385
Jens Axboe59725112007-04-02 10:06:42 +02001386struct bio_set *bioset_create(int bio_pool_size, int bvec_pool_size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001388 struct bio_set *bs = kzalloc(sizeof(*bs), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389
1390 if (!bs)
1391 return NULL;
1392
Matthew Dobson93d23412006-03-26 01:37:50 -08001393 bs->bio_pool = mempool_create_slab_pool(bio_pool_size, bio_slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 if (!bs->bio_pool)
1395 goto bad;
1396
Martin K. Petersen7ba1ba12008-06-30 20:04:41 +02001397 if (bioset_integrity_create(bs, bio_pool_size))
1398 goto bad;
1399
Jens Axboe59725112007-04-02 10:06:42 +02001400 if (!biovec_create_pools(bs, bvec_pool_size))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 return bs;
1402
1403bad:
1404 bioset_free(bs);
1405 return NULL;
1406}
1407
1408static void __init biovec_init_slabs(void)
1409{
1410 int i;
1411
1412 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1413 int size;
1414 struct biovec_slab *bvs = bvec_slabs + i;
1415
1416 size = bvs->nr_vecs * sizeof(struct bio_vec);
1417 bvs->slab = kmem_cache_create(bvs->name, size, 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001418 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 }
1420}
1421
1422static int __init init_bio(void)
1423{
Christoph Lameter0a31bd52007-05-06 14:49:57 -07001424 bio_slab = KMEM_CACHE(bio, SLAB_HWCACHE_ALIGN|SLAB_PANIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425
Martin K. Petersen7ba1ba12008-06-30 20:04:41 +02001426 bio_integrity_init_slab();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 biovec_init_slabs();
1428
Jens Axboe59725112007-04-02 10:06:42 +02001429 fs_bio_set = bioset_create(BIO_POOL_SIZE, 2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 if (!fs_bio_set)
1431 panic("bio: can't allocate bios\n");
1432
Matthew Dobson0eaae62a2006-03-26 01:37:47 -08001433 bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES,
1434 sizeof(struct bio_pair));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 if (!bio_split_pool)
1436 panic("bio: can't create split pool\n");
1437
1438 return 0;
1439}
1440
1441subsys_initcall(init_bio);
1442
1443EXPORT_SYMBOL(bio_alloc);
Jens Axboe0a0d96b2008-09-11 13:17:37 +02001444EXPORT_SYMBOL(bio_kmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445EXPORT_SYMBOL(bio_put);
Peter Osterlund36763472005-09-06 15:16:42 -07001446EXPORT_SYMBOL(bio_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447EXPORT_SYMBOL(bio_endio);
1448EXPORT_SYMBOL(bio_init);
1449EXPORT_SYMBOL(__bio_clone);
1450EXPORT_SYMBOL(bio_clone);
1451EXPORT_SYMBOL(bio_phys_segments);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452EXPORT_SYMBOL(bio_add_page);
Mike Christie6e68af62005-11-11 05:30:27 -06001453EXPORT_SYMBOL(bio_add_pc_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454EXPORT_SYMBOL(bio_get_nr_vecs);
Jens Axboe40044ce2008-03-17 21:14:40 +01001455EXPORT_SYMBOL(bio_map_user);
1456EXPORT_SYMBOL(bio_unmap_user);
Mike Christie df46b9a2005-06-20 14:04:44 +02001457EXPORT_SYMBOL(bio_map_kern);
FUJITA Tomonori68154e92008-04-25 12:47:50 +02001458EXPORT_SYMBOL(bio_copy_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459EXPORT_SYMBOL(bio_pair_release);
1460EXPORT_SYMBOL(bio_split);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461EXPORT_SYMBOL(bio_copy_user);
1462EXPORT_SYMBOL(bio_uncopy_user);
1463EXPORT_SYMBOL(bioset_create);
1464EXPORT_SYMBOL(bioset_free);
1465EXPORT_SYMBOL(bio_alloc_bioset);