blob: d91cfbf7ebc411b9c240e6cbcbfa1c60e6078b02 [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>
James Bottomley f1970ba2005-06-20 14:06:52 +020029#include <scsi/sg.h> /* for struct sg_iovec */
Linus Torvalds1da177e2005-04-16 15:20:36 -070030
31#define BIO_POOL_SIZE 256
32
Eric Dumazetfa3536c2006-03-26 01:37:24 -080033static kmem_cache_t *bio_slab __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
35#define BIOVEC_NR_POOLS 6
36
37/*
38 * a small number of entries is fine, not going to be performance critical.
39 * basically we just need to survive
40 */
41#define BIO_SPLIT_ENTRIES 8
Eric Dumazetfa3536c2006-03-26 01:37:24 -080042mempool_t *bio_split_pool __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
44struct biovec_slab {
45 int nr_vecs;
46 char *name;
47 kmem_cache_t *slab;
48};
49
50/*
51 * if you change this list, also change bvec_alloc or things will
52 * break badly! cannot be bigger than what you can fit into an
53 * unsigned short
54 */
55
56#define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
Christoph Lameter6c036522005-07-07 17:56:59 -070057static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
Linus Torvalds1da177e2005-04-16 15:20:36 -070058 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES),
59};
60#undef BV
61
62/*
63 * bio_set is used to allow other portions of the IO system to
64 * allocate their own private memory pools for bio and iovec structures.
65 * These memory pools in turn all allocate from the bio_slab
66 * and the bvec_slabs[].
67 */
68struct bio_set {
69 mempool_t *bio_pool;
70 mempool_t *bvec_pools[BIOVEC_NR_POOLS];
71};
72
73/*
74 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
75 * IO code that does not need private memory pools.
76 */
77static struct bio_set *fs_bio_set;
78
Al Virodd0fc662005-10-07 07:46:04 +010079static inline struct 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 -070080{
81 struct bio_vec *bvl;
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
83 /*
84 * see comment near bvec_array define!
85 */
86 switch (nr) {
87 case 1 : *idx = 0; break;
88 case 2 ... 4: *idx = 1; break;
89 case 5 ... 16: *idx = 2; break;
90 case 17 ... 64: *idx = 3; break;
91 case 65 ... 128: *idx = 4; break;
92 case 129 ... BIO_MAX_PAGES: *idx = 5; break;
93 default:
94 return NULL;
95 }
96 /*
97 * idx now points to the pool we want to allocate from
98 */
99
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 bvl = mempool_alloc(bs->bvec_pools[*idx], gfp_mask);
Andreas Mohrbf02c082006-10-11 01:22:24 -0700101 if (bvl) {
102 struct biovec_slab *bp = bvec_slabs + *idx;
103
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104 memset(bvl, 0, bp->nr_vecs * sizeof(struct bio_vec));
Andreas Mohrbf02c082006-10-11 01:22:24 -0700105 }
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{
112 const int pool_idx = BIO_POOL_IDX(bio);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
114 BIO_BUG_ON(pool_idx >= BIOVEC_NR_POOLS);
115
Peter Osterlund36763472005-09-06 15:16:42 -0700116 mempool_free(bio->bi_io_vec, bio_set->bvec_pools[pool_idx]);
117 mempool_free(bio, bio_set->bio_pool);
118}
119
120/*
121 * default destructor for a bio allocated with bio_alloc_bioset()
122 */
123static void bio_fs_destructor(struct bio *bio)
124{
125 bio_free(bio, fs_bio_set);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126}
127
Arjan van de Ven858119e2006-01-14 13:20:43 -0800128void bio_init(struct bio *bio)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129{
130 bio->bi_next = NULL;
Jens Axboe0ea60b52006-01-09 14:45:10 +0100131 bio->bi_bdev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 bio->bi_flags = 1 << BIO_UPTODATE;
133 bio->bi_rw = 0;
134 bio->bi_vcnt = 0;
135 bio->bi_idx = 0;
136 bio->bi_phys_segments = 0;
137 bio->bi_hw_segments = 0;
138 bio->bi_hw_front_size = 0;
139 bio->bi_hw_back_size = 0;
140 bio->bi_size = 0;
141 bio->bi_max_vecs = 0;
142 bio->bi_end_io = NULL;
143 atomic_set(&bio->bi_cnt, 1);
144 bio->bi_private = NULL;
145}
146
147/**
148 * bio_alloc_bioset - allocate a bio for I/O
149 * @gfp_mask: the GFP_ mask given to the slab allocator
150 * @nr_iovecs: number of iovecs to pre-allocate
Martin Waitz67be2dd2005-05-01 08:59:26 -0700151 * @bs: the bio_set to allocate from
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 *
153 * Description:
154 * bio_alloc_bioset will first try it's on mempool to satisfy the allocation.
155 * If %__GFP_WAIT is set then we will block on the internal pool waiting
156 * for a &struct bio to become free.
157 *
158 * allocate bio and iovecs from the memory pools specified by the
159 * bio_set structure.
160 **/
Al Virodd0fc662005-10-07 07:46:04 +0100161struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162{
163 struct bio *bio = mempool_alloc(bs->bio_pool, gfp_mask);
164
165 if (likely(bio)) {
166 struct bio_vec *bvl = NULL;
167
168 bio_init(bio);
169 if (likely(nr_iovecs)) {
Andreas Mohrbf02c082006-10-11 01:22:24 -0700170 unsigned long idx = 0; /* shut up gcc */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
172 bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs);
173 if (unlikely(!bvl)) {
174 mempool_free(bio, bs->bio_pool);
175 bio = NULL;
176 goto out;
177 }
178 bio->bi_flags |= idx << BIO_POOL_OFFSET;
179 bio->bi_max_vecs = bvec_slabs[idx].nr_vecs;
180 }
181 bio->bi_io_vec = bvl;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 }
183out:
184 return bio;
185}
186
Al Virodd0fc662005-10-07 07:46:04 +0100187struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188{
Peter Osterlund36763472005-09-06 15:16:42 -0700189 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
190
191 if (bio)
192 bio->bi_destructor = bio_fs_destructor;
193
194 return bio;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195}
196
197void zero_fill_bio(struct bio *bio)
198{
199 unsigned long flags;
200 struct bio_vec *bv;
201 int i;
202
203 bio_for_each_segment(bv, bio, i) {
204 char *data = bvec_kmap_irq(bv, &flags);
205 memset(data, 0, bv->bv_len);
206 flush_dcache_page(bv->bv_page);
207 bvec_kunmap_irq(data, &flags);
208 }
209}
210EXPORT_SYMBOL(zero_fill_bio);
211
212/**
213 * bio_put - release a reference to a bio
214 * @bio: bio to release reference to
215 *
216 * Description:
217 * Put a reference to a &struct bio, either one you have gotten with
218 * bio_alloc or bio_get. The last put of a bio will free it.
219 **/
220void bio_put(struct bio *bio)
221{
222 BIO_BUG_ON(!atomic_read(&bio->bi_cnt));
223
224 /*
225 * last put frees it
226 */
227 if (atomic_dec_and_test(&bio->bi_cnt)) {
228 bio->bi_next = NULL;
229 bio->bi_destructor(bio);
230 }
231}
232
233inline int bio_phys_segments(request_queue_t *q, struct bio *bio)
234{
235 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
236 blk_recount_segments(q, bio);
237
238 return bio->bi_phys_segments;
239}
240
241inline int bio_hw_segments(request_queue_t *q, struct bio *bio)
242{
243 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
244 blk_recount_segments(q, bio);
245
246 return bio->bi_hw_segments;
247}
248
249/**
250 * __bio_clone - clone a bio
251 * @bio: destination bio
252 * @bio_src: bio to clone
253 *
254 * Clone a &bio. Caller will own the returned bio, but not
255 * the actual data it points to. Reference count of returned
256 * bio will be one.
257 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800258void __bio_clone(struct bio *bio, struct bio *bio_src)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259{
260 request_queue_t *q = bdev_get_queue(bio_src->bi_bdev);
261
Andrew Mortone525e152005-08-07 09:42:12 -0700262 memcpy(bio->bi_io_vec, bio_src->bi_io_vec,
263 bio_src->bi_max_vecs * sizeof(struct bio_vec));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
265 bio->bi_sector = bio_src->bi_sector;
266 bio->bi_bdev = bio_src->bi_bdev;
267 bio->bi_flags |= 1 << BIO_CLONED;
268 bio->bi_rw = bio_src->bi_rw;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 bio->bi_vcnt = bio_src->bi_vcnt;
270 bio->bi_size = bio_src->bi_size;
Andrew Mortona5453be2005-07-28 01:07:18 -0700271 bio->bi_idx = bio_src->bi_idx;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 bio_phys_segments(q, bio);
273 bio_hw_segments(q, bio);
274}
275
276/**
277 * bio_clone - clone a bio
278 * @bio: bio to clone
279 * @gfp_mask: allocation priority
280 *
281 * Like __bio_clone, only also allocates the returned bio
282 */
Al Virodd0fc662005-10-07 07:46:04 +0100283struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284{
285 struct bio *b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, fs_bio_set);
286
Peter Osterlund36763472005-09-06 15:16:42 -0700287 if (b) {
288 b->bi_destructor = bio_fs_destructor;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 __bio_clone(b, bio);
Peter Osterlund36763472005-09-06 15:16:42 -0700290 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291
292 return b;
293}
294
295/**
296 * bio_get_nr_vecs - return approx number of vecs
297 * @bdev: I/O target
298 *
299 * Return the approximate number of pages we can send to this target.
300 * There's no guarantee that you will be able to fit this number of pages
301 * into a bio, it does not account for dynamic restrictions that vary
302 * on offset.
303 */
304int bio_get_nr_vecs(struct block_device *bdev)
305{
306 request_queue_t *q = bdev_get_queue(bdev);
307 int nr_pages;
308
309 nr_pages = ((q->max_sectors << 9) + PAGE_SIZE - 1) >> PAGE_SHIFT;
310 if (nr_pages > q->max_phys_segments)
311 nr_pages = q->max_phys_segments;
312 if (nr_pages > q->max_hw_segments)
313 nr_pages = q->max_hw_segments;
314
315 return nr_pages;
316}
317
318static int __bio_add_page(request_queue_t *q, struct bio *bio, struct page
Mike Christiedefd94b2005-12-05 02:37:06 -0600319 *page, unsigned int len, unsigned int offset,
320 unsigned short max_sectors)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321{
322 int retried_segments = 0;
323 struct bio_vec *bvec;
324
325 /*
326 * cloned bio must not modify vec list
327 */
328 if (unlikely(bio_flagged(bio, BIO_CLONED)))
329 return 0;
330
Jens Axboe80cfd542006-01-06 09:43:28 +0100331 if (((bio->bi_size + len) >> 9) > max_sectors)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 return 0;
333
Jens Axboe80cfd542006-01-06 09:43:28 +0100334 /*
335 * For filesystems with a blocksize smaller than the pagesize
336 * we will often be called with the same page as last time and
337 * a consecutive offset. Optimize this special case.
338 */
339 if (bio->bi_vcnt > 0) {
340 struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1];
341
342 if (page == prev->bv_page &&
343 offset == prev->bv_offset + prev->bv_len) {
344 prev->bv_len += len;
345 if (q->merge_bvec_fn &&
346 q->merge_bvec_fn(q, bio, prev) < len) {
347 prev->bv_len -= len;
348 return 0;
349 }
350
351 goto done;
352 }
353 }
354
355 if (bio->bi_vcnt >= bio->bi_max_vecs)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 return 0;
357
358 /*
359 * we might lose a segment or two here, but rather that than
360 * make this too complex.
361 */
362
363 while (bio->bi_phys_segments >= q->max_phys_segments
364 || bio->bi_hw_segments >= q->max_hw_segments
365 || BIOVEC_VIRT_OVERSIZE(bio->bi_size)) {
366
367 if (retried_segments)
368 return 0;
369
370 retried_segments = 1;
371 blk_recount_segments(q, bio);
372 }
373
374 /*
375 * setup the new entry, we might clear it again later if we
376 * cannot add the page
377 */
378 bvec = &bio->bi_io_vec[bio->bi_vcnt];
379 bvec->bv_page = page;
380 bvec->bv_len = len;
381 bvec->bv_offset = offset;
382
383 /*
384 * if queue has other restrictions (eg varying max sector size
385 * depending on offset), it can specify a merge_bvec_fn in the
386 * queue to get further control
387 */
388 if (q->merge_bvec_fn) {
389 /*
390 * merge_bvec_fn() returns number of bytes it can accept
391 * at this offset
392 */
393 if (q->merge_bvec_fn(q, bio, bvec) < len) {
394 bvec->bv_page = NULL;
395 bvec->bv_len = 0;
396 bvec->bv_offset = 0;
397 return 0;
398 }
399 }
400
401 /* If we may be able to merge these biovecs, force a recount */
402 if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec) ||
403 BIOVEC_VIRT_MERGEABLE(bvec-1, bvec)))
404 bio->bi_flags &= ~(1 << BIO_SEG_VALID);
405
406 bio->bi_vcnt++;
407 bio->bi_phys_segments++;
408 bio->bi_hw_segments++;
Jens Axboe80cfd542006-01-06 09:43:28 +0100409 done:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 bio->bi_size += len;
411 return len;
412}
413
414/**
Mike Christie6e68af62005-11-11 05:30:27 -0600415 * bio_add_pc_page - attempt to add page to bio
Jens Axboefddfdea2006-01-31 15:24:34 +0100416 * @q: the target queue
Mike Christie6e68af62005-11-11 05:30:27 -0600417 * @bio: destination bio
418 * @page: page to add
419 * @len: vec entry length
420 * @offset: vec entry offset
421 *
422 * Attempt to add a page to the bio_vec maplist. This can fail for a
423 * number of reasons, such as the bio being full or target block
424 * device limitations. The target block device must allow bio's
425 * smaller than PAGE_SIZE, so it is always possible to add a single
426 * page to an empty bio. This should only be used by REQ_PC bios.
427 */
428int bio_add_pc_page(request_queue_t *q, struct bio *bio, struct page *page,
429 unsigned int len, unsigned int offset)
430{
Mike Christiedefd94b2005-12-05 02:37:06 -0600431 return __bio_add_page(q, bio, page, len, offset, q->max_hw_sectors);
Mike Christie6e68af62005-11-11 05:30:27 -0600432}
433
434/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 * bio_add_page - attempt to add page to bio
436 * @bio: destination bio
437 * @page: page to add
438 * @len: vec entry length
439 * @offset: vec entry offset
440 *
441 * Attempt to add a page to the bio_vec maplist. This can fail for a
442 * number of reasons, such as the bio being full or target block
443 * device limitations. The target block device must allow bio's
444 * smaller than PAGE_SIZE, so it is always possible to add a single
445 * page to an empty bio.
446 */
447int bio_add_page(struct bio *bio, struct page *page, unsigned int len,
448 unsigned int offset)
449{
Mike Christiedefd94b2005-12-05 02:37:06 -0600450 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
451 return __bio_add_page(q, bio, page, len, offset, q->max_sectors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452}
453
454struct bio_map_data {
455 struct bio_vec *iovecs;
456 void __user *userptr;
457};
458
459static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio)
460{
461 memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt);
462 bio->bi_private = bmd;
463}
464
465static void bio_free_map_data(struct bio_map_data *bmd)
466{
467 kfree(bmd->iovecs);
468 kfree(bmd);
469}
470
471static struct bio_map_data *bio_alloc_map_data(int nr_segs)
472{
473 struct bio_map_data *bmd = kmalloc(sizeof(*bmd), GFP_KERNEL);
474
475 if (!bmd)
476 return NULL;
477
478 bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, GFP_KERNEL);
479 if (bmd->iovecs)
480 return bmd;
481
482 kfree(bmd);
483 return NULL;
484}
485
486/**
487 * bio_uncopy_user - finish previously mapped bio
488 * @bio: bio being terminated
489 *
490 * Free pages allocated from bio_copy_user() and write back data
491 * to user space in case of a read.
492 */
493int bio_uncopy_user(struct bio *bio)
494{
495 struct bio_map_data *bmd = bio->bi_private;
496 const int read = bio_data_dir(bio) == READ;
497 struct bio_vec *bvec;
498 int i, ret = 0;
499
500 __bio_for_each_segment(bvec, bio, i, 0) {
501 char *addr = page_address(bvec->bv_page);
502 unsigned int len = bmd->iovecs[i].bv_len;
503
504 if (read && !ret && copy_to_user(bmd->userptr, addr, len))
505 ret = -EFAULT;
506
507 __free_page(bvec->bv_page);
508 bmd->userptr += len;
509 }
510 bio_free_map_data(bmd);
511 bio_put(bio);
512 return ret;
513}
514
515/**
516 * bio_copy_user - copy user data to bio
517 * @q: destination block queue
518 * @uaddr: start of user address
519 * @len: length in bytes
520 * @write_to_vm: bool indicating writing to pages or not
521 *
522 * Prepares and returns a bio for indirect user io, bouncing data
523 * to/from kernel pages as necessary. Must be paired with
524 * call bio_uncopy_user() on io completion.
525 */
526struct bio *bio_copy_user(request_queue_t *q, unsigned long uaddr,
527 unsigned int len, int write_to_vm)
528{
529 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
530 unsigned long start = uaddr >> PAGE_SHIFT;
531 struct bio_map_data *bmd;
532 struct bio_vec *bvec;
533 struct page *page;
534 struct bio *bio;
535 int i, ret;
536
537 bmd = bio_alloc_map_data(end - start);
538 if (!bmd)
539 return ERR_PTR(-ENOMEM);
540
541 bmd->userptr = (void __user *) uaddr;
542
543 ret = -ENOMEM;
544 bio = bio_alloc(GFP_KERNEL, end - start);
545 if (!bio)
546 goto out_bmd;
547
548 bio->bi_rw |= (!write_to_vm << BIO_RW);
549
550 ret = 0;
551 while (len) {
552 unsigned int bytes = PAGE_SIZE;
553
554 if (bytes > len)
555 bytes = len;
556
557 page = alloc_page(q->bounce_gfp | GFP_KERNEL);
558 if (!page) {
559 ret = -ENOMEM;
560 break;
561 }
562
Mike Christiedefd94b2005-12-05 02:37:06 -0600563 if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 ret = -EINVAL;
565 break;
566 }
567
568 len -= bytes;
569 }
570
571 if (ret)
572 goto cleanup;
573
574 /*
575 * success
576 */
577 if (!write_to_vm) {
578 char __user *p = (char __user *) uaddr;
579
580 /*
581 * for a write, copy in data to kernel pages
582 */
583 ret = -EFAULT;
584 bio_for_each_segment(bvec, bio, i) {
585 char *addr = page_address(bvec->bv_page);
586
587 if (copy_from_user(addr, p, bvec->bv_len))
588 goto cleanup;
589 p += bvec->bv_len;
590 }
591 }
592
593 bio_set_map_data(bmd, bio);
594 return bio;
595cleanup:
596 bio_for_each_segment(bvec, bio, i)
597 __free_page(bvec->bv_page);
598
599 bio_put(bio);
600out_bmd:
601 bio_free_map_data(bmd);
602 return ERR_PTR(ret);
603}
604
James Bottomley f1970ba2005-06-20 14:06:52 +0200605static struct bio *__bio_map_user_iov(request_queue_t *q,
606 struct block_device *bdev,
607 struct sg_iovec *iov, int iov_count,
608 int write_to_vm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609{
James Bottomley f1970ba2005-06-20 14:06:52 +0200610 int i, j;
611 int nr_pages = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 struct page **pages;
613 struct bio *bio;
James Bottomley f1970ba2005-06-20 14:06:52 +0200614 int cur_page = 0;
615 int ret, offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
James Bottomley f1970ba2005-06-20 14:06:52 +0200617 for (i = 0; i < iov_count; i++) {
618 unsigned long uaddr = (unsigned long)iov[i].iov_base;
619 unsigned long len = iov[i].iov_len;
620 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
621 unsigned long start = uaddr >> PAGE_SHIFT;
622
623 nr_pages += end - start;
624 /*
Mike Christiead2d7222006-12-01 10:40:20 +0100625 * buffer must be aligned to at least hardsector size for now
James Bottomley f1970ba2005-06-20 14:06:52 +0200626 */
Mike Christiead2d7222006-12-01 10:40:20 +0100627 if (uaddr & queue_dma_alignment(q))
James Bottomley f1970ba2005-06-20 14:06:52 +0200628 return ERR_PTR(-EINVAL);
629 }
630
631 if (!nr_pages)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 return ERR_PTR(-EINVAL);
633
634 bio = bio_alloc(GFP_KERNEL, nr_pages);
635 if (!bio)
636 return ERR_PTR(-ENOMEM);
637
638 ret = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -0800639 pages = kcalloc(nr_pages, sizeof(struct page *), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 if (!pages)
641 goto out;
642
James Bottomley f1970ba2005-06-20 14:06:52 +0200643 for (i = 0; i < iov_count; i++) {
644 unsigned long uaddr = (unsigned long)iov[i].iov_base;
645 unsigned long len = iov[i].iov_len;
646 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
647 unsigned long start = uaddr >> PAGE_SHIFT;
648 const int local_nr_pages = end - start;
649 const int page_limit = cur_page + local_nr_pages;
650
651 down_read(&current->mm->mmap_sem);
652 ret = get_user_pages(current, current->mm, uaddr,
653 local_nr_pages,
654 write_to_vm, 0, &pages[cur_page], NULL);
655 up_read(&current->mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Jens Axboe99172152006-06-16 13:02:29 +0200657 if (ret < local_nr_pages) {
658 ret = -EFAULT;
James Bottomley f1970ba2005-06-20 14:06:52 +0200659 goto out_unmap;
Jens Axboe99172152006-06-16 13:02:29 +0200660 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
James Bottomley f1970ba2005-06-20 14:06:52 +0200662 offset = uaddr & ~PAGE_MASK;
663 for (j = cur_page; j < page_limit; j++) {
664 unsigned int bytes = PAGE_SIZE - offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
James Bottomley f1970ba2005-06-20 14:06:52 +0200666 if (len <= 0)
667 break;
668
669 if (bytes > len)
670 bytes = len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
James Bottomley f1970ba2005-06-20 14:06:52 +0200672 /*
673 * sorry...
674 */
Mike Christiedefd94b2005-12-05 02:37:06 -0600675 if (bio_add_pc_page(q, bio, pages[j], bytes, offset) <
676 bytes)
James Bottomley f1970ba2005-06-20 14:06:52 +0200677 break;
678
679 len -= bytes;
680 offset = 0;
681 }
682
683 cur_page = j;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 /*
James Bottomley f1970ba2005-06-20 14:06:52 +0200685 * release the pages we didn't map into the bio, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 */
James Bottomley f1970ba2005-06-20 14:06:52 +0200687 while (j < page_limit)
688 page_cache_release(pages[j++]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 }
690
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 kfree(pages);
692
693 /*
694 * set data direction, and check if mapped pages need bouncing
695 */
696 if (!write_to_vm)
697 bio->bi_rw |= (1 << BIO_RW);
698
James Bottomley f1970ba2005-06-20 14:06:52 +0200699 bio->bi_bdev = bdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 bio->bi_flags |= (1 << BIO_USER_MAPPED);
701 return bio;
James Bottomley f1970ba2005-06-20 14:06:52 +0200702
703 out_unmap:
704 for (i = 0; i < nr_pages; i++) {
705 if(!pages[i])
706 break;
707 page_cache_release(pages[i]);
708 }
709 out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 kfree(pages);
711 bio_put(bio);
712 return ERR_PTR(ret);
713}
714
715/**
716 * bio_map_user - map user address into bio
Martin Waitz67be2dd2005-05-01 08:59:26 -0700717 * @q: the request_queue_t for the bio
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 * @bdev: destination block device
719 * @uaddr: start of user address
720 * @len: length in bytes
721 * @write_to_vm: bool indicating writing to pages or not
722 *
723 * Map the user space address into a bio suitable for io to a block
724 * device. Returns an error pointer in case of error.
725 */
726struct bio *bio_map_user(request_queue_t *q, struct block_device *bdev,
727 unsigned long uaddr, unsigned int len, int write_to_vm)
728{
James Bottomley f1970ba2005-06-20 14:06:52 +0200729 struct sg_iovec iov;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730
viro@ZenIV.linux.org.uk3f703532005-09-09 16:53:56 +0100731 iov.iov_base = (void __user *)uaddr;
James Bottomley f1970ba2005-06-20 14:06:52 +0200732 iov.iov_len = len;
733
734 return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm);
735}
736
737/**
738 * bio_map_user_iov - map user sg_iovec table into bio
739 * @q: the request_queue_t for the bio
740 * @bdev: destination block device
741 * @iov: the iovec.
742 * @iov_count: number of elements in the iovec
743 * @write_to_vm: bool indicating writing to pages or not
744 *
745 * Map the user space address into a bio suitable for io to a block
746 * device. Returns an error pointer in case of error.
747 */
748struct bio *bio_map_user_iov(request_queue_t *q, struct block_device *bdev,
749 struct sg_iovec *iov, int iov_count,
750 int write_to_vm)
751{
752 struct bio *bio;
753 int len = 0, i;
754
755 bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756
757 if (IS_ERR(bio))
758 return bio;
759
760 /*
761 * subtle -- if __bio_map_user() ended up bouncing a bio,
762 * it would normally disappear when its bi_end_io is run.
763 * however, we need it for the unmap, so grab an extra
764 * reference to it
765 */
766 bio_get(bio);
767
James Bottomley f1970ba2005-06-20 14:06:52 +0200768 for (i = 0; i < iov_count; i++)
769 len += iov[i].iov_len;
770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 if (bio->bi_size == len)
772 return bio;
773
774 /*
775 * don't support partial mappings
776 */
777 bio_endio(bio, bio->bi_size, 0);
778 bio_unmap_user(bio);
779 return ERR_PTR(-EINVAL);
780}
781
782static void __bio_unmap_user(struct bio *bio)
783{
784 struct bio_vec *bvec;
785 int i;
786
787 /*
788 * make sure we dirty pages we wrote to
789 */
790 __bio_for_each_segment(bvec, bio, i, 0) {
791 if (bio_data_dir(bio) == READ)
792 set_page_dirty_lock(bvec->bv_page);
793
794 page_cache_release(bvec->bv_page);
795 }
796
797 bio_put(bio);
798}
799
800/**
801 * bio_unmap_user - unmap a bio
802 * @bio: the bio being unmapped
803 *
804 * Unmap a bio previously mapped by bio_map_user(). Must be called with
805 * a process context.
806 *
807 * bio_unmap_user() may sleep.
808 */
809void bio_unmap_user(struct bio *bio)
810{
811 __bio_unmap_user(bio);
812 bio_put(bio);
813}
814
Jens Axboeb8238252005-06-20 14:05:27 +0200815static int bio_map_kern_endio(struct bio *bio, unsigned int bytes_done, int err)
816{
817 if (bio->bi_size)
818 return 1;
819
820 bio_put(bio);
821 return 0;
822}
823
824
Mike Christie df46b9a2005-06-20 14:04:44 +0200825static struct bio *__bio_map_kern(request_queue_t *q, void *data,
Al Viro27496a82005-10-21 03:20:48 -0400826 unsigned int len, gfp_t gfp_mask)
Mike Christie df46b9a2005-06-20 14:04:44 +0200827{
828 unsigned long kaddr = (unsigned long)data;
829 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
830 unsigned long start = kaddr >> PAGE_SHIFT;
831 const int nr_pages = end - start;
832 int offset, i;
833 struct bio *bio;
834
835 bio = bio_alloc(gfp_mask, nr_pages);
836 if (!bio)
837 return ERR_PTR(-ENOMEM);
838
839 offset = offset_in_page(kaddr);
840 for (i = 0; i < nr_pages; i++) {
841 unsigned int bytes = PAGE_SIZE - offset;
842
843 if (len <= 0)
844 break;
845
846 if (bytes > len)
847 bytes = len;
848
Mike Christiedefd94b2005-12-05 02:37:06 -0600849 if (bio_add_pc_page(q, bio, virt_to_page(data), bytes,
850 offset) < bytes)
Mike Christie df46b9a2005-06-20 14:04:44 +0200851 break;
852
853 data += bytes;
854 len -= bytes;
855 offset = 0;
856 }
857
Jens Axboeb8238252005-06-20 14:05:27 +0200858 bio->bi_end_io = bio_map_kern_endio;
Mike Christie df46b9a2005-06-20 14:04:44 +0200859 return bio;
860}
861
862/**
863 * bio_map_kern - map kernel address into bio
864 * @q: the request_queue_t for the bio
865 * @data: pointer to buffer to map
866 * @len: length in bytes
867 * @gfp_mask: allocation flags for bio allocation
868 *
869 * Map the kernel address into a bio suitable for io to a block
870 * device. Returns an error pointer in case of error.
871 */
872struct bio *bio_map_kern(request_queue_t *q, void *data, unsigned int len,
Al Viro27496a82005-10-21 03:20:48 -0400873 gfp_t gfp_mask)
Mike Christie df46b9a2005-06-20 14:04:44 +0200874{
875 struct bio *bio;
876
877 bio = __bio_map_kern(q, data, len, gfp_mask);
878 if (IS_ERR(bio))
879 return bio;
880
881 if (bio->bi_size == len)
882 return bio;
883
884 /*
885 * Don't support partial mappings.
886 */
887 bio_put(bio);
888 return ERR_PTR(-EINVAL);
889}
890
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891/*
892 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
893 * for performing direct-IO in BIOs.
894 *
895 * The problem is that we cannot run set_page_dirty() from interrupt context
896 * because the required locks are not interrupt-safe. So what we can do is to
897 * mark the pages dirty _before_ performing IO. And in interrupt context,
898 * check that the pages are still dirty. If so, fine. If not, redirty them
899 * in process context.
900 *
901 * We special-case compound pages here: normally this means reads into hugetlb
902 * pages. The logic in here doesn't really work right for compound pages
903 * because the VM does not uniformly chase down the head page in all cases.
904 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
905 * handle them at all. So we skip compound pages here at an early stage.
906 *
907 * Note that this code is very hard to test under normal circumstances because
908 * direct-io pins the pages with get_user_pages(). This makes
909 * is_page_cache_freeable return false, and the VM will not clean the pages.
910 * But other code (eg, pdflush) could clean the pages if they are mapped
911 * pagecache.
912 *
913 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
914 * deferred bio dirtying paths.
915 */
916
917/*
918 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
919 */
920void bio_set_pages_dirty(struct bio *bio)
921{
922 struct bio_vec *bvec = bio->bi_io_vec;
923 int i;
924
925 for (i = 0; i < bio->bi_vcnt; i++) {
926 struct page *page = bvec[i].bv_page;
927
928 if (page && !PageCompound(page))
929 set_page_dirty_lock(page);
930 }
931}
932
933static void bio_release_pages(struct bio *bio)
934{
935 struct bio_vec *bvec = bio->bi_io_vec;
936 int i;
937
938 for (i = 0; i < bio->bi_vcnt; i++) {
939 struct page *page = bvec[i].bv_page;
940
941 if (page)
942 put_page(page);
943 }
944}
945
946/*
947 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
948 * If they are, then fine. If, however, some pages are clean then they must
949 * have been written out during the direct-IO read. So we take another ref on
950 * the BIO and the offending pages and re-dirty the pages in process context.
951 *
952 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
953 * here on. It will run one page_cache_release() against each page and will
954 * run one bio_put() against the BIO.
955 */
956
957static void bio_dirty_fn(void *data);
958
959static DECLARE_WORK(bio_dirty_work, bio_dirty_fn, NULL);
960static DEFINE_SPINLOCK(bio_dirty_lock);
961static struct bio *bio_dirty_list;
962
963/*
964 * This runs in process context
965 */
966static void bio_dirty_fn(void *data)
967{
968 unsigned long flags;
969 struct bio *bio;
970
971 spin_lock_irqsave(&bio_dirty_lock, flags);
972 bio = bio_dirty_list;
973 bio_dirty_list = NULL;
974 spin_unlock_irqrestore(&bio_dirty_lock, flags);
975
976 while (bio) {
977 struct bio *next = bio->bi_private;
978
979 bio_set_pages_dirty(bio);
980 bio_release_pages(bio);
981 bio_put(bio);
982 bio = next;
983 }
984}
985
986void bio_check_pages_dirty(struct bio *bio)
987{
988 struct bio_vec *bvec = bio->bi_io_vec;
989 int nr_clean_pages = 0;
990 int i;
991
992 for (i = 0; i < bio->bi_vcnt; i++) {
993 struct page *page = bvec[i].bv_page;
994
995 if (PageDirty(page) || PageCompound(page)) {
996 page_cache_release(page);
997 bvec[i].bv_page = NULL;
998 } else {
999 nr_clean_pages++;
1000 }
1001 }
1002
1003 if (nr_clean_pages) {
1004 unsigned long flags;
1005
1006 spin_lock_irqsave(&bio_dirty_lock, flags);
1007 bio->bi_private = bio_dirty_list;
1008 bio_dirty_list = bio;
1009 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1010 schedule_work(&bio_dirty_work);
1011 } else {
1012 bio_put(bio);
1013 }
1014}
1015
1016/**
1017 * bio_endio - end I/O on a bio
1018 * @bio: bio
1019 * @bytes_done: number of bytes completed
1020 * @error: error, if any
1021 *
1022 * Description:
1023 * bio_endio() will end I/O on @bytes_done number of bytes. This may be
1024 * just a partial part of the bio, or it may be the whole bio. bio_endio()
1025 * is the preferred way to end I/O on a bio, it takes care of decrementing
1026 * bi_size and clearing BIO_UPTODATE on error. @error is 0 on success, and
1027 * and one of the established -Exxxx (-EIO, for instance) error values in
1028 * case something went wrong. Noone should call bi_end_io() directly on
1029 * a bio unless they own it and thus know that it has an end_io function.
1030 **/
1031void bio_endio(struct bio *bio, unsigned int bytes_done, int error)
1032{
1033 if (error)
1034 clear_bit(BIO_UPTODATE, &bio->bi_flags);
1035
1036 if (unlikely(bytes_done > bio->bi_size)) {
1037 printk("%s: want %u bytes done, only %u left\n", __FUNCTION__,
1038 bytes_done, bio->bi_size);
1039 bytes_done = bio->bi_size;
1040 }
1041
1042 bio->bi_size -= bytes_done;
1043 bio->bi_sector += (bytes_done >> 9);
1044
1045 if (bio->bi_end_io)
1046 bio->bi_end_io(bio, bytes_done, error);
1047}
1048
1049void bio_pair_release(struct bio_pair *bp)
1050{
1051 if (atomic_dec_and_test(&bp->cnt)) {
1052 struct bio *master = bp->bio1.bi_private;
1053
1054 bio_endio(master, master->bi_size, bp->error);
1055 mempool_free(bp, bp->bio2.bi_private);
1056 }
1057}
1058
1059static int bio_pair_end_1(struct bio * bi, unsigned int done, int err)
1060{
1061 struct bio_pair *bp = container_of(bi, struct bio_pair, bio1);
1062
1063 if (err)
1064 bp->error = err;
1065
1066 if (bi->bi_size)
1067 return 1;
1068
1069 bio_pair_release(bp);
1070 return 0;
1071}
1072
1073static int bio_pair_end_2(struct bio * bi, unsigned int done, int err)
1074{
1075 struct bio_pair *bp = container_of(bi, struct bio_pair, bio2);
1076
1077 if (err)
1078 bp->error = err;
1079
1080 if (bi->bi_size)
1081 return 1;
1082
1083 bio_pair_release(bp);
1084 return 0;
1085}
1086
1087/*
1088 * split a bio - only worry about a bio with a single page
1089 * in it's iovec
1090 */
1091struct bio_pair *bio_split(struct bio *bi, mempool_t *pool, int first_sectors)
1092{
1093 struct bio_pair *bp = mempool_alloc(pool, GFP_NOIO);
1094
1095 if (!bp)
1096 return bp;
1097
Jens Axboe2056a782006-03-23 20:00:26 +01001098 blk_add_trace_pdu_int(bdev_get_queue(bi->bi_bdev), BLK_TA_SPLIT, bi,
1099 bi->bi_sector + first_sectors);
1100
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 BUG_ON(bi->bi_vcnt != 1);
1102 BUG_ON(bi->bi_idx != 0);
1103 atomic_set(&bp->cnt, 3);
1104 bp->error = 0;
1105 bp->bio1 = *bi;
1106 bp->bio2 = *bi;
1107 bp->bio2.bi_sector += first_sectors;
1108 bp->bio2.bi_size -= first_sectors << 9;
1109 bp->bio1.bi_size = first_sectors << 9;
1110
1111 bp->bv1 = bi->bi_io_vec[0];
1112 bp->bv2 = bi->bi_io_vec[0];
1113 bp->bv2.bv_offset += first_sectors << 9;
1114 bp->bv2.bv_len -= first_sectors << 9;
1115 bp->bv1.bv_len = first_sectors << 9;
1116
1117 bp->bio1.bi_io_vec = &bp->bv1;
1118 bp->bio2.bi_io_vec = &bp->bv2;
1119
NeilBrowna2eb0c12006-05-22 22:35:27 -07001120 bp->bio1.bi_max_vecs = 1;
1121 bp->bio2.bi_max_vecs = 1;
1122
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 bp->bio1.bi_end_io = bio_pair_end_1;
1124 bp->bio2.bi_end_io = bio_pair_end_2;
1125
1126 bp->bio1.bi_private = bi;
1127 bp->bio2.bi_private = pool;
1128
1129 return bp;
1130}
1131
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132
1133/*
1134 * create memory pools for biovec's in a bio_set.
1135 * use the global biovec slabs created for general use.
1136 */
1137static int biovec_create_pools(struct bio_set *bs, int pool_entries, int scale)
1138{
1139 int i;
1140
1141 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1142 struct biovec_slab *bp = bvec_slabs + i;
1143 mempool_t **bvp = bs->bvec_pools + i;
1144
Milan Broz50be3452006-09-04 15:37:57 +02001145 if (pool_entries > 1 && i >= scale)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 pool_entries >>= 1;
1147
Matthew Dobson93d23412006-03-26 01:37:50 -08001148 *bvp = mempool_create_slab_pool(pool_entries, bp->slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 if (!*bvp)
1150 return -ENOMEM;
1151 }
1152 return 0;
1153}
1154
1155static void biovec_free_pools(struct bio_set *bs)
1156{
1157 int i;
1158
1159 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1160 mempool_t *bvp = bs->bvec_pools[i];
1161
1162 if (bvp)
1163 mempool_destroy(bvp);
1164 }
1165
1166}
1167
1168void bioset_free(struct bio_set *bs)
1169{
1170 if (bs->bio_pool)
1171 mempool_destroy(bs->bio_pool);
1172
1173 biovec_free_pools(bs);
1174
1175 kfree(bs);
1176}
1177
1178struct bio_set *bioset_create(int bio_pool_size, int bvec_pool_size, int scale)
1179{
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001180 struct bio_set *bs = kzalloc(sizeof(*bs), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
1182 if (!bs)
1183 return NULL;
1184
Matthew Dobson93d23412006-03-26 01:37:50 -08001185 bs->bio_pool = mempool_create_slab_pool(bio_pool_size, bio_slab);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 if (!bs->bio_pool)
1187 goto bad;
1188
1189 if (!biovec_create_pools(bs, bvec_pool_size, scale))
1190 return bs;
1191
1192bad:
1193 bioset_free(bs);
1194 return NULL;
1195}
1196
1197static void __init biovec_init_slabs(void)
1198{
1199 int i;
1200
1201 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1202 int size;
1203 struct biovec_slab *bvs = bvec_slabs + i;
1204
1205 size = bvs->nr_vecs * sizeof(struct bio_vec);
1206 bvs->slab = kmem_cache_create(bvs->name, size, 0,
1207 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1208 }
1209}
1210
1211static int __init init_bio(void)
1212{
1213 int megabytes, bvec_pool_entries;
1214 int scale = BIOVEC_NR_POOLS;
1215
1216 bio_slab = kmem_cache_create("bio", sizeof(struct bio), 0,
1217 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1218
1219 biovec_init_slabs();
1220
1221 megabytes = nr_free_pages() >> (20 - PAGE_SHIFT);
1222
1223 /*
1224 * find out where to start scaling
1225 */
1226 if (megabytes <= 16)
1227 scale = 0;
1228 else if (megabytes <= 32)
1229 scale = 1;
1230 else if (megabytes <= 64)
1231 scale = 2;
1232 else if (megabytes <= 96)
1233 scale = 3;
1234 else if (megabytes <= 128)
1235 scale = 4;
1236
1237 /*
Benjamin LaHaiseb0e6e962006-03-23 03:01:08 -08001238 * Limit number of entries reserved -- mempools are only used when
1239 * the system is completely unable to allocate memory, so we only
1240 * need enough to make progress.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 */
Benjamin LaHaiseb0e6e962006-03-23 03:01:08 -08001242 bvec_pool_entries = 1 + scale;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243
1244 fs_bio_set = bioset_create(BIO_POOL_SIZE, bvec_pool_entries, scale);
1245 if (!fs_bio_set)
1246 panic("bio: can't allocate bios\n");
1247
Matthew Dobson0eaae62a2006-03-26 01:37:47 -08001248 bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES,
1249 sizeof(struct bio_pair));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 if (!bio_split_pool)
1251 panic("bio: can't create split pool\n");
1252
1253 return 0;
1254}
1255
1256subsys_initcall(init_bio);
1257
1258EXPORT_SYMBOL(bio_alloc);
1259EXPORT_SYMBOL(bio_put);
Peter Osterlund36763472005-09-06 15:16:42 -07001260EXPORT_SYMBOL(bio_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261EXPORT_SYMBOL(bio_endio);
1262EXPORT_SYMBOL(bio_init);
1263EXPORT_SYMBOL(__bio_clone);
1264EXPORT_SYMBOL(bio_clone);
1265EXPORT_SYMBOL(bio_phys_segments);
1266EXPORT_SYMBOL(bio_hw_segments);
1267EXPORT_SYMBOL(bio_add_page);
Mike Christie6e68af62005-11-11 05:30:27 -06001268EXPORT_SYMBOL(bio_add_pc_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269EXPORT_SYMBOL(bio_get_nr_vecs);
1270EXPORT_SYMBOL(bio_map_user);
1271EXPORT_SYMBOL(bio_unmap_user);
Mike Christie df46b9a2005-06-20 14:04:44 +02001272EXPORT_SYMBOL(bio_map_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273EXPORT_SYMBOL(bio_pair_release);
1274EXPORT_SYMBOL(bio_split);
1275EXPORT_SYMBOL(bio_split_pool);
1276EXPORT_SYMBOL(bio_copy_user);
1277EXPORT_SYMBOL(bio_uncopy_user);
1278EXPORT_SYMBOL(bioset_create);
1279EXPORT_SYMBOL(bioset_free);
1280EXPORT_SYMBOL(bio_alloc_bioset);