Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
0fe23479 | 2 | * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk> |
1da177e4 LT |
3 | * |
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> | |
2056a782 | 28 | #include <linux/blktrace_api.h> |
f1970baf | 29 | #include <scsi/sg.h> /* for struct sg_iovec */ |
1da177e4 | 30 | |
55782138 | 31 | #include <trace/events/block.h> |
0bfc2455 | 32 | |
392ddc32 JA |
33 | /* |
34 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
35 | * itself, to shrink a bio data allocation from two mempool calls to one | |
36 | */ | |
37 | #define BIO_INLINE_VECS 4 | |
38 | ||
6feef531 | 39 | static mempool_t *bio_split_pool __read_mostly; |
1da177e4 | 40 | |
1da177e4 LT |
41 | /* |
42 | * if you change this list, also change bvec_alloc or things will | |
43 | * break badly! cannot be bigger than what you can fit into an | |
44 | * unsigned short | |
45 | */ | |
1da177e4 | 46 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
bb799ca0 | 47 | struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = { |
1da177e4 LT |
48 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
49 | }; | |
50 | #undef BV | |
51 | ||
1da177e4 LT |
52 | /* |
53 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
54 | * IO code that does not need private memory pools. | |
55 | */ | |
51d654e1 | 56 | struct bio_set *fs_bio_set; |
1da177e4 | 57 | |
bb799ca0 JA |
58 | /* |
59 | * Our slab pool management | |
60 | */ | |
61 | struct bio_slab { | |
62 | struct kmem_cache *slab; | |
63 | unsigned int slab_ref; | |
64 | unsigned int slab_size; | |
65 | char name[8]; | |
66 | }; | |
67 | static DEFINE_MUTEX(bio_slab_lock); | |
68 | static struct bio_slab *bio_slabs; | |
69 | static unsigned int bio_slab_nr, bio_slab_max; | |
70 | ||
71 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
72 | { | |
73 | unsigned int sz = sizeof(struct bio) + extra_size; | |
74 | struct kmem_cache *slab = NULL; | |
75 | struct bio_slab *bslab; | |
76 | unsigned int i, entry = -1; | |
77 | ||
78 | mutex_lock(&bio_slab_lock); | |
79 | ||
80 | i = 0; | |
81 | while (i < bio_slab_nr) { | |
82 | struct bio_slab *bslab = &bio_slabs[i]; | |
83 | ||
84 | if (!bslab->slab && entry == -1) | |
85 | entry = i; | |
86 | else if (bslab->slab_size == sz) { | |
87 | slab = bslab->slab; | |
88 | bslab->slab_ref++; | |
89 | break; | |
90 | } | |
91 | i++; | |
92 | } | |
93 | ||
94 | if (slab) | |
95 | goto out_unlock; | |
96 | ||
97 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
98 | bio_slab_max <<= 1; | |
99 | bio_slabs = krealloc(bio_slabs, | |
100 | bio_slab_max * sizeof(struct bio_slab), | |
101 | GFP_KERNEL); | |
102 | if (!bio_slabs) | |
103 | goto out_unlock; | |
104 | } | |
105 | if (entry == -1) | |
106 | entry = bio_slab_nr++; | |
107 | ||
108 | bslab = &bio_slabs[entry]; | |
109 | ||
110 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
111 | slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL); | |
112 | if (!slab) | |
113 | goto out_unlock; | |
114 | ||
115 | printk("bio: create slab <%s> at %d\n", bslab->name, entry); | |
116 | bslab->slab = slab; | |
117 | bslab->slab_ref = 1; | |
118 | bslab->slab_size = sz; | |
119 | out_unlock: | |
120 | mutex_unlock(&bio_slab_lock); | |
121 | return slab; | |
122 | } | |
123 | ||
124 | static void bio_put_slab(struct bio_set *bs) | |
125 | { | |
126 | struct bio_slab *bslab = NULL; | |
127 | unsigned int i; | |
128 | ||
129 | mutex_lock(&bio_slab_lock); | |
130 | ||
131 | for (i = 0; i < bio_slab_nr; i++) { | |
132 | if (bs->bio_slab == bio_slabs[i].slab) { | |
133 | bslab = &bio_slabs[i]; | |
134 | break; | |
135 | } | |
136 | } | |
137 | ||
138 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
139 | goto out; | |
140 | ||
141 | WARN_ON(!bslab->slab_ref); | |
142 | ||
143 | if (--bslab->slab_ref) | |
144 | goto out; | |
145 | ||
146 | kmem_cache_destroy(bslab->slab); | |
147 | bslab->slab = NULL; | |
148 | ||
149 | out: | |
150 | mutex_unlock(&bio_slab_lock); | |
151 | } | |
152 | ||
7ba1ba12 MP |
153 | unsigned int bvec_nr_vecs(unsigned short idx) |
154 | { | |
155 | return bvec_slabs[idx].nr_vecs; | |
156 | } | |
157 | ||
bb799ca0 JA |
158 | void bvec_free_bs(struct bio_set *bs, struct bio_vec *bv, unsigned int idx) |
159 | { | |
160 | BIO_BUG_ON(idx >= BIOVEC_NR_POOLS); | |
161 | ||
162 | if (idx == BIOVEC_MAX_IDX) | |
163 | mempool_free(bv, bs->bvec_pool); | |
164 | else { | |
165 | struct biovec_slab *bvs = bvec_slabs + idx; | |
166 | ||
167 | kmem_cache_free(bvs->slab, bv); | |
168 | } | |
169 | } | |
170 | ||
7ff9345f JA |
171 | struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, |
172 | struct bio_set *bs) | |
1da177e4 LT |
173 | { |
174 | struct bio_vec *bvl; | |
1da177e4 | 175 | |
7ff9345f JA |
176 | /* |
177 | * see comment near bvec_array define! | |
178 | */ | |
179 | switch (nr) { | |
180 | case 1: | |
181 | *idx = 0; | |
182 | break; | |
183 | case 2 ... 4: | |
184 | *idx = 1; | |
185 | break; | |
186 | case 5 ... 16: | |
187 | *idx = 2; | |
188 | break; | |
189 | case 17 ... 64: | |
190 | *idx = 3; | |
191 | break; | |
192 | case 65 ... 128: | |
193 | *idx = 4; | |
194 | break; | |
195 | case 129 ... BIO_MAX_PAGES: | |
196 | *idx = 5; | |
197 | break; | |
198 | default: | |
199 | return NULL; | |
200 | } | |
201 | ||
202 | /* | |
203 | * idx now points to the pool we want to allocate from. only the | |
204 | * 1-vec entry pool is mempool backed. | |
205 | */ | |
206 | if (*idx == BIOVEC_MAX_IDX) { | |
207 | fallback: | |
208 | bvl = mempool_alloc(bs->bvec_pool, gfp_mask); | |
209 | } else { | |
210 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
211 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO); | |
212 | ||
0a0d96b0 | 213 | /* |
7ff9345f JA |
214 | * Make this allocation restricted and don't dump info on |
215 | * allocation failures, since we'll fallback to the mempool | |
216 | * in case of failure. | |
0a0d96b0 | 217 | */ |
7ff9345f | 218 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 219 | |
0a0d96b0 | 220 | /* |
7ff9345f JA |
221 | * Try a slab allocation. If this fails and __GFP_WAIT |
222 | * is set, retry with the 1-entry mempool | |
0a0d96b0 | 223 | */ |
7ff9345f JA |
224 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
225 | if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) { | |
226 | *idx = BIOVEC_MAX_IDX; | |
227 | goto fallback; | |
228 | } | |
229 | } | |
230 | ||
1da177e4 LT |
231 | return bvl; |
232 | } | |
233 | ||
7ff9345f | 234 | void bio_free(struct bio *bio, struct bio_set *bs) |
1da177e4 | 235 | { |
bb799ca0 | 236 | void *p; |
1da177e4 | 237 | |
392ddc32 | 238 | if (bio_has_allocated_vec(bio)) |
bb799ca0 | 239 | bvec_free_bs(bs, bio->bi_io_vec, BIO_POOL_IDX(bio)); |
1da177e4 | 240 | |
7ba1ba12 | 241 | if (bio_integrity(bio)) |
6d2a78e7 | 242 | bio_integrity_free(bio); |
7ba1ba12 | 243 | |
bb799ca0 JA |
244 | /* |
245 | * If we have front padding, adjust the bio pointer before freeing | |
246 | */ | |
247 | p = bio; | |
248 | if (bs->front_pad) | |
249 | p -= bs->front_pad; | |
250 | ||
251 | mempool_free(p, bs->bio_pool); | |
3676347a PO |
252 | } |
253 | ||
858119e1 | 254 | void bio_init(struct bio *bio) |
1da177e4 | 255 | { |
2b94de55 | 256 | memset(bio, 0, sizeof(*bio)); |
1da177e4 | 257 | bio->bi_flags = 1 << BIO_UPTODATE; |
c7c22e4d | 258 | bio->bi_comp_cpu = -1; |
1da177e4 | 259 | atomic_set(&bio->bi_cnt, 1); |
1da177e4 LT |
260 | } |
261 | ||
262 | /** | |
263 | * bio_alloc_bioset - allocate a bio for I/O | |
264 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
265 | * @nr_iovecs: number of iovecs to pre-allocate | |
0a0d96b0 | 266 | * @bs: the bio_set to allocate from. If %NULL, just use kmalloc |
1da177e4 LT |
267 | * |
268 | * Description: | |
0a0d96b0 | 269 | * bio_alloc_bioset will first try its own mempool to satisfy the allocation. |
1da177e4 | 270 | * If %__GFP_WAIT is set then we will block on the internal pool waiting |
0a0d96b0 JA |
271 | * for a &struct bio to become free. If a %NULL @bs is passed in, we will |
272 | * fall back to just using @kmalloc to allocate the required memory. | |
1da177e4 | 273 | * |
bb799ca0 JA |
274 | * Note that the caller must set ->bi_destructor on succesful return |
275 | * of a bio, to do the appropriate freeing of the bio once the reference | |
276 | * count drops to zero. | |
1da177e4 | 277 | **/ |
dd0fc66f | 278 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) |
1da177e4 | 279 | { |
451a9ebf | 280 | unsigned long idx = BIO_POOL_NONE; |
34053979 | 281 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
282 | struct bio *bio; |
283 | void *p; | |
284 | ||
285 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
286 | if (unlikely(!p)) | |
287 | return NULL; | |
288 | bio = p + bs->front_pad; | |
1da177e4 | 289 | |
34053979 IM |
290 | bio_init(bio); |
291 | ||
292 | if (unlikely(!nr_iovecs)) | |
293 | goto out_set; | |
294 | ||
295 | if (nr_iovecs <= BIO_INLINE_VECS) { | |
296 | bvl = bio->bi_inline_vecs; | |
297 | nr_iovecs = BIO_INLINE_VECS; | |
298 | } else { | |
299 | bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs); | |
300 | if (unlikely(!bvl)) | |
301 | goto err_free; | |
302 | ||
303 | nr_iovecs = bvec_nr_vecs(idx); | |
1da177e4 | 304 | } |
451a9ebf | 305 | out_set: |
34053979 IM |
306 | bio->bi_flags |= idx << BIO_POOL_OFFSET; |
307 | bio->bi_max_vecs = nr_iovecs; | |
34053979 | 308 | bio->bi_io_vec = bvl; |
1da177e4 | 309 | return bio; |
34053979 IM |
310 | |
311 | err_free: | |
451a9ebf | 312 | mempool_free(p, bs->bio_pool); |
34053979 | 313 | return NULL; |
1da177e4 LT |
314 | } |
315 | ||
451a9ebf TH |
316 | static void bio_fs_destructor(struct bio *bio) |
317 | { | |
318 | bio_free(bio, fs_bio_set); | |
319 | } | |
320 | ||
321 | /** | |
322 | * bio_alloc - allocate a new bio, memory pool backed | |
323 | * @gfp_mask: allocation mask to use | |
324 | * @nr_iovecs: number of iovecs | |
325 | * | |
326 | * Allocate a new bio with @nr_iovecs bvecs. If @gfp_mask | |
327 | * contains __GFP_WAIT, the allocation is guaranteed to succeed. | |
328 | * | |
329 | * RETURNS: | |
330 | * Pointer to new bio on success, NULL on failure. | |
331 | */ | |
332 | struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs) | |
333 | { | |
334 | struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set); | |
335 | ||
336 | if (bio) | |
337 | bio->bi_destructor = bio_fs_destructor; | |
338 | ||
339 | return bio; | |
340 | } | |
341 | ||
342 | static void bio_kmalloc_destructor(struct bio *bio) | |
343 | { | |
344 | if (bio_integrity(bio)) | |
345 | bio_integrity_free(bio); | |
346 | kfree(bio); | |
347 | } | |
348 | ||
86c824b9 JA |
349 | /** |
350 | * bio_alloc - allocate a bio for I/O | |
351 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
352 | * @nr_iovecs: number of iovecs to pre-allocate | |
353 | * | |
354 | * Description: | |
355 | * bio_alloc will allocate a bio and associated bio_vec array that can hold | |
356 | * at least @nr_iovecs entries. Allocations will be done from the | |
357 | * fs_bio_set. Also see @bio_alloc_bioset. | |
358 | * | |
359 | * If %__GFP_WAIT is set, then bio_alloc will always be able to allocate | |
360 | * a bio. This is due to the mempool guarantees. To make this work, callers | |
361 | * must never allocate more than 1 bio at the time from this pool. Callers | |
362 | * that need to allocate more than 1 bio must always submit the previously | |
363 | * allocate bio for IO before attempting to allocate a new one. Failure to | |
364 | * do so can cause livelocks under memory pressure. | |
365 | * | |
366 | **/ | |
0a0d96b0 JA |
367 | struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs) |
368 | { | |
451a9ebf | 369 | struct bio *bio; |
0a0d96b0 | 370 | |
451a9ebf TH |
371 | bio = kmalloc(sizeof(struct bio) + nr_iovecs * sizeof(struct bio_vec), |
372 | gfp_mask); | |
373 | if (unlikely(!bio)) | |
374 | return NULL; | |
375 | ||
376 | bio_init(bio); | |
377 | bio->bi_flags |= BIO_POOL_NONE << BIO_POOL_OFFSET; | |
378 | bio->bi_max_vecs = nr_iovecs; | |
379 | bio->bi_io_vec = bio->bi_inline_vecs; | |
380 | bio->bi_destructor = bio_kmalloc_destructor; | |
0a0d96b0 JA |
381 | |
382 | return bio; | |
383 | } | |
384 | ||
1da177e4 LT |
385 | void zero_fill_bio(struct bio *bio) |
386 | { | |
387 | unsigned long flags; | |
388 | struct bio_vec *bv; | |
389 | int i; | |
390 | ||
391 | bio_for_each_segment(bv, bio, i) { | |
392 | char *data = bvec_kmap_irq(bv, &flags); | |
393 | memset(data, 0, bv->bv_len); | |
394 | flush_dcache_page(bv->bv_page); | |
395 | bvec_kunmap_irq(data, &flags); | |
396 | } | |
397 | } | |
398 | EXPORT_SYMBOL(zero_fill_bio); | |
399 | ||
400 | /** | |
401 | * bio_put - release a reference to a bio | |
402 | * @bio: bio to release reference to | |
403 | * | |
404 | * Description: | |
405 | * Put a reference to a &struct bio, either one you have gotten with | |
406 | * bio_alloc or bio_get. The last put of a bio will free it. | |
407 | **/ | |
408 | void bio_put(struct bio *bio) | |
409 | { | |
410 | BIO_BUG_ON(!atomic_read(&bio->bi_cnt)); | |
411 | ||
412 | /* | |
413 | * last put frees it | |
414 | */ | |
415 | if (atomic_dec_and_test(&bio->bi_cnt)) { | |
416 | bio->bi_next = NULL; | |
417 | bio->bi_destructor(bio); | |
418 | } | |
419 | } | |
420 | ||
165125e1 | 421 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
422 | { |
423 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
424 | blk_recount_segments(q, bio); | |
425 | ||
426 | return bio->bi_phys_segments; | |
427 | } | |
428 | ||
1da177e4 LT |
429 | /** |
430 | * __bio_clone - clone a bio | |
431 | * @bio: destination bio | |
432 | * @bio_src: bio to clone | |
433 | * | |
434 | * Clone a &bio. Caller will own the returned bio, but not | |
435 | * the actual data it points to. Reference count of returned | |
436 | * bio will be one. | |
437 | */ | |
858119e1 | 438 | void __bio_clone(struct bio *bio, struct bio *bio_src) |
1da177e4 | 439 | { |
e525e153 AM |
440 | memcpy(bio->bi_io_vec, bio_src->bi_io_vec, |
441 | bio_src->bi_max_vecs * sizeof(struct bio_vec)); | |
1da177e4 | 442 | |
5d84070e JA |
443 | /* |
444 | * most users will be overriding ->bi_bdev with a new target, | |
445 | * so we don't set nor calculate new physical/hw segment counts here | |
446 | */ | |
1da177e4 LT |
447 | bio->bi_sector = bio_src->bi_sector; |
448 | bio->bi_bdev = bio_src->bi_bdev; | |
449 | bio->bi_flags |= 1 << BIO_CLONED; | |
450 | bio->bi_rw = bio_src->bi_rw; | |
1da177e4 LT |
451 | bio->bi_vcnt = bio_src->bi_vcnt; |
452 | bio->bi_size = bio_src->bi_size; | |
a5453be4 | 453 | bio->bi_idx = bio_src->bi_idx; |
1da177e4 LT |
454 | } |
455 | ||
456 | /** | |
457 | * bio_clone - clone a bio | |
458 | * @bio: bio to clone | |
459 | * @gfp_mask: allocation priority | |
460 | * | |
461 | * Like __bio_clone, only also allocates the returned bio | |
462 | */ | |
dd0fc66f | 463 | struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask) |
1da177e4 LT |
464 | { |
465 | struct bio *b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, fs_bio_set); | |
466 | ||
7ba1ba12 MP |
467 | if (!b) |
468 | return NULL; | |
469 | ||
470 | b->bi_destructor = bio_fs_destructor; | |
471 | __bio_clone(b, bio); | |
472 | ||
473 | if (bio_integrity(bio)) { | |
474 | int ret; | |
475 | ||
6d2a78e7 | 476 | ret = bio_integrity_clone(b, bio, gfp_mask); |
7ba1ba12 | 477 | |
059ea331 LZ |
478 | if (ret < 0) { |
479 | bio_put(b); | |
7ba1ba12 | 480 | return NULL; |
059ea331 | 481 | } |
3676347a | 482 | } |
1da177e4 LT |
483 | |
484 | return b; | |
485 | } | |
486 | ||
487 | /** | |
488 | * bio_get_nr_vecs - return approx number of vecs | |
489 | * @bdev: I/O target | |
490 | * | |
491 | * Return the approximate number of pages we can send to this target. | |
492 | * There's no guarantee that you will be able to fit this number of pages | |
493 | * into a bio, it does not account for dynamic restrictions that vary | |
494 | * on offset. | |
495 | */ | |
496 | int bio_get_nr_vecs(struct block_device *bdev) | |
497 | { | |
165125e1 | 498 | struct request_queue *q = bdev_get_queue(bdev); |
1da177e4 LT |
499 | int nr_pages; |
500 | ||
ae03bf63 MP |
501 | nr_pages = ((queue_max_sectors(q) << 9) + PAGE_SIZE - 1) >> PAGE_SHIFT; |
502 | if (nr_pages > queue_max_phys_segments(q)) | |
503 | nr_pages = queue_max_phys_segments(q); | |
504 | if (nr_pages > queue_max_hw_segments(q)) | |
505 | nr_pages = queue_max_hw_segments(q); | |
1da177e4 LT |
506 | |
507 | return nr_pages; | |
508 | } | |
509 | ||
165125e1 | 510 | static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page |
defd94b7 MC |
511 | *page, unsigned int len, unsigned int offset, |
512 | unsigned short max_sectors) | |
1da177e4 LT |
513 | { |
514 | int retried_segments = 0; | |
515 | struct bio_vec *bvec; | |
516 | ||
517 | /* | |
518 | * cloned bio must not modify vec list | |
519 | */ | |
520 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
521 | return 0; | |
522 | ||
80cfd548 | 523 | if (((bio->bi_size + len) >> 9) > max_sectors) |
1da177e4 LT |
524 | return 0; |
525 | ||
80cfd548 JA |
526 | /* |
527 | * For filesystems with a blocksize smaller than the pagesize | |
528 | * we will often be called with the same page as last time and | |
529 | * a consecutive offset. Optimize this special case. | |
530 | */ | |
531 | if (bio->bi_vcnt > 0) { | |
532 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
533 | ||
534 | if (page == prev->bv_page && | |
535 | offset == prev->bv_offset + prev->bv_len) { | |
536 | prev->bv_len += len; | |
cc371e66 AK |
537 | |
538 | if (q->merge_bvec_fn) { | |
539 | struct bvec_merge_data bvm = { | |
540 | .bi_bdev = bio->bi_bdev, | |
541 | .bi_sector = bio->bi_sector, | |
542 | .bi_size = bio->bi_size, | |
543 | .bi_rw = bio->bi_rw, | |
544 | }; | |
545 | ||
546 | if (q->merge_bvec_fn(q, &bvm, prev) < len) { | |
547 | prev->bv_len -= len; | |
548 | return 0; | |
549 | } | |
80cfd548 JA |
550 | } |
551 | ||
552 | goto done; | |
553 | } | |
554 | } | |
555 | ||
556 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
557 | return 0; |
558 | ||
559 | /* | |
560 | * we might lose a segment or two here, but rather that than | |
561 | * make this too complex. | |
562 | */ | |
563 | ||
ae03bf63 MP |
564 | while (bio->bi_phys_segments >= queue_max_phys_segments(q) |
565 | || bio->bi_phys_segments >= queue_max_hw_segments(q)) { | |
1da177e4 LT |
566 | |
567 | if (retried_segments) | |
568 | return 0; | |
569 | ||
570 | retried_segments = 1; | |
571 | blk_recount_segments(q, bio); | |
572 | } | |
573 | ||
574 | /* | |
575 | * setup the new entry, we might clear it again later if we | |
576 | * cannot add the page | |
577 | */ | |
578 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
579 | bvec->bv_page = page; | |
580 | bvec->bv_len = len; | |
581 | bvec->bv_offset = offset; | |
582 | ||
583 | /* | |
584 | * if queue has other restrictions (eg varying max sector size | |
585 | * depending on offset), it can specify a merge_bvec_fn in the | |
586 | * queue to get further control | |
587 | */ | |
588 | if (q->merge_bvec_fn) { | |
cc371e66 AK |
589 | struct bvec_merge_data bvm = { |
590 | .bi_bdev = bio->bi_bdev, | |
591 | .bi_sector = bio->bi_sector, | |
592 | .bi_size = bio->bi_size, | |
593 | .bi_rw = bio->bi_rw, | |
594 | }; | |
595 | ||
1da177e4 LT |
596 | /* |
597 | * merge_bvec_fn() returns number of bytes it can accept | |
598 | * at this offset | |
599 | */ | |
cc371e66 | 600 | if (q->merge_bvec_fn(q, &bvm, bvec) < len) { |
1da177e4 LT |
601 | bvec->bv_page = NULL; |
602 | bvec->bv_len = 0; | |
603 | bvec->bv_offset = 0; | |
604 | return 0; | |
605 | } | |
606 | } | |
607 | ||
608 | /* If we may be able to merge these biovecs, force a recount */ | |
b8b3e16c | 609 | if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
1da177e4 LT |
610 | bio->bi_flags &= ~(1 << BIO_SEG_VALID); |
611 | ||
612 | bio->bi_vcnt++; | |
613 | bio->bi_phys_segments++; | |
80cfd548 | 614 | done: |
1da177e4 LT |
615 | bio->bi_size += len; |
616 | return len; | |
617 | } | |
618 | ||
6e68af66 MC |
619 | /** |
620 | * bio_add_pc_page - attempt to add page to bio | |
fddfdeaf | 621 | * @q: the target queue |
6e68af66 MC |
622 | * @bio: destination bio |
623 | * @page: page to add | |
624 | * @len: vec entry length | |
625 | * @offset: vec entry offset | |
626 | * | |
627 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
628 | * number of reasons, such as the bio being full or target block | |
629 | * device limitations. The target block device must allow bio's | |
630 | * smaller than PAGE_SIZE, so it is always possible to add a single | |
631 | * page to an empty bio. This should only be used by REQ_PC bios. | |
632 | */ | |
165125e1 | 633 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page, |
6e68af66 MC |
634 | unsigned int len, unsigned int offset) |
635 | { | |
ae03bf63 MP |
636 | return __bio_add_page(q, bio, page, len, offset, |
637 | queue_max_hw_sectors(q)); | |
6e68af66 MC |
638 | } |
639 | ||
1da177e4 LT |
640 | /** |
641 | * bio_add_page - attempt to add page to bio | |
642 | * @bio: destination bio | |
643 | * @page: page to add | |
644 | * @len: vec entry length | |
645 | * @offset: vec entry offset | |
646 | * | |
647 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
648 | * number of reasons, such as the bio being full or target block | |
649 | * device limitations. The target block device must allow bio's | |
650 | * smaller than PAGE_SIZE, so it is always possible to add a single | |
651 | * page to an empty bio. | |
652 | */ | |
653 | int bio_add_page(struct bio *bio, struct page *page, unsigned int len, | |
654 | unsigned int offset) | |
655 | { | |
defd94b7 | 656 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
ae03bf63 | 657 | return __bio_add_page(q, bio, page, len, offset, queue_max_sectors(q)); |
1da177e4 LT |
658 | } |
659 | ||
660 | struct bio_map_data { | |
661 | struct bio_vec *iovecs; | |
c5dec1c3 | 662 | struct sg_iovec *sgvecs; |
152e283f FT |
663 | int nr_sgvecs; |
664 | int is_our_pages; | |
1da177e4 LT |
665 | }; |
666 | ||
c5dec1c3 | 667 | static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio, |
152e283f FT |
668 | struct sg_iovec *iov, int iov_count, |
669 | int is_our_pages) | |
1da177e4 LT |
670 | { |
671 | memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt); | |
c5dec1c3 FT |
672 | memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count); |
673 | bmd->nr_sgvecs = iov_count; | |
152e283f | 674 | bmd->is_our_pages = is_our_pages; |
1da177e4 LT |
675 | bio->bi_private = bmd; |
676 | } | |
677 | ||
678 | static void bio_free_map_data(struct bio_map_data *bmd) | |
679 | { | |
680 | kfree(bmd->iovecs); | |
c5dec1c3 | 681 | kfree(bmd->sgvecs); |
1da177e4 LT |
682 | kfree(bmd); |
683 | } | |
684 | ||
76029ff3 FT |
685 | static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count, |
686 | gfp_t gfp_mask) | |
1da177e4 | 687 | { |
76029ff3 | 688 | struct bio_map_data *bmd = kmalloc(sizeof(*bmd), gfp_mask); |
1da177e4 LT |
689 | |
690 | if (!bmd) | |
691 | return NULL; | |
692 | ||
76029ff3 | 693 | bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask); |
c5dec1c3 FT |
694 | if (!bmd->iovecs) { |
695 | kfree(bmd); | |
696 | return NULL; | |
697 | } | |
698 | ||
76029ff3 | 699 | bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask); |
c5dec1c3 | 700 | if (bmd->sgvecs) |
1da177e4 LT |
701 | return bmd; |
702 | ||
c5dec1c3 | 703 | kfree(bmd->iovecs); |
1da177e4 LT |
704 | kfree(bmd); |
705 | return NULL; | |
706 | } | |
707 | ||
aefcc28a | 708 | static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs, |
152e283f FT |
709 | struct sg_iovec *iov, int iov_count, int uncopy, |
710 | int do_free_page) | |
c5dec1c3 FT |
711 | { |
712 | int ret = 0, i; | |
713 | struct bio_vec *bvec; | |
714 | int iov_idx = 0; | |
715 | unsigned int iov_off = 0; | |
716 | int read = bio_data_dir(bio) == READ; | |
717 | ||
718 | __bio_for_each_segment(bvec, bio, i, 0) { | |
719 | char *bv_addr = page_address(bvec->bv_page); | |
aefcc28a | 720 | unsigned int bv_len = iovecs[i].bv_len; |
c5dec1c3 FT |
721 | |
722 | while (bv_len && iov_idx < iov_count) { | |
723 | unsigned int bytes; | |
0e0c6212 | 724 | char __user *iov_addr; |
c5dec1c3 FT |
725 | |
726 | bytes = min_t(unsigned int, | |
727 | iov[iov_idx].iov_len - iov_off, bv_len); | |
728 | iov_addr = iov[iov_idx].iov_base + iov_off; | |
729 | ||
730 | if (!ret) { | |
731 | if (!read && !uncopy) | |
732 | ret = copy_from_user(bv_addr, iov_addr, | |
733 | bytes); | |
734 | if (read && uncopy) | |
735 | ret = copy_to_user(iov_addr, bv_addr, | |
736 | bytes); | |
737 | ||
738 | if (ret) | |
739 | ret = -EFAULT; | |
740 | } | |
741 | ||
742 | bv_len -= bytes; | |
743 | bv_addr += bytes; | |
744 | iov_addr += bytes; | |
745 | iov_off += bytes; | |
746 | ||
747 | if (iov[iov_idx].iov_len == iov_off) { | |
748 | iov_idx++; | |
749 | iov_off = 0; | |
750 | } | |
751 | } | |
752 | ||
152e283f | 753 | if (do_free_page) |
c5dec1c3 FT |
754 | __free_page(bvec->bv_page); |
755 | } | |
756 | ||
757 | return ret; | |
758 | } | |
759 | ||
1da177e4 LT |
760 | /** |
761 | * bio_uncopy_user - finish previously mapped bio | |
762 | * @bio: bio being terminated | |
763 | * | |
764 | * Free pages allocated from bio_copy_user() and write back data | |
765 | * to user space in case of a read. | |
766 | */ | |
767 | int bio_uncopy_user(struct bio *bio) | |
768 | { | |
769 | struct bio_map_data *bmd = bio->bi_private; | |
81882766 | 770 | int ret = 0; |
1da177e4 | 771 | |
81882766 FT |
772 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) |
773 | ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs, | |
774 | bmd->nr_sgvecs, 1, bmd->is_our_pages); | |
1da177e4 LT |
775 | bio_free_map_data(bmd); |
776 | bio_put(bio); | |
777 | return ret; | |
778 | } | |
779 | ||
780 | /** | |
c5dec1c3 | 781 | * bio_copy_user_iov - copy user data to bio |
1da177e4 | 782 | * @q: destination block queue |
152e283f | 783 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
784 | * @iov: the iovec. |
785 | * @iov_count: number of elements in the iovec | |
1da177e4 | 786 | * @write_to_vm: bool indicating writing to pages or not |
a3bce90e | 787 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
788 | * |
789 | * Prepares and returns a bio for indirect user io, bouncing data | |
790 | * to/from kernel pages as necessary. Must be paired with | |
791 | * call bio_uncopy_user() on io completion. | |
792 | */ | |
152e283f FT |
793 | struct bio *bio_copy_user_iov(struct request_queue *q, |
794 | struct rq_map_data *map_data, | |
795 | struct sg_iovec *iov, int iov_count, | |
796 | int write_to_vm, gfp_t gfp_mask) | |
1da177e4 | 797 | { |
1da177e4 LT |
798 | struct bio_map_data *bmd; |
799 | struct bio_vec *bvec; | |
800 | struct page *page; | |
801 | struct bio *bio; | |
802 | int i, ret; | |
c5dec1c3 FT |
803 | int nr_pages = 0; |
804 | unsigned int len = 0; | |
56c451f4 | 805 | unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0; |
1da177e4 | 806 | |
c5dec1c3 FT |
807 | for (i = 0; i < iov_count; i++) { |
808 | unsigned long uaddr; | |
809 | unsigned long end; | |
810 | unsigned long start; | |
811 | ||
812 | uaddr = (unsigned long)iov[i].iov_base; | |
813 | end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
814 | start = uaddr >> PAGE_SHIFT; | |
815 | ||
816 | nr_pages += end - start; | |
817 | len += iov[i].iov_len; | |
818 | } | |
819 | ||
69838727 FT |
820 | if (offset) |
821 | nr_pages++; | |
822 | ||
a3bce90e | 823 | bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask); |
1da177e4 LT |
824 | if (!bmd) |
825 | return ERR_PTR(-ENOMEM); | |
826 | ||
1da177e4 | 827 | ret = -ENOMEM; |
a9e9dc24 | 828 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
829 | if (!bio) |
830 | goto out_bmd; | |
831 | ||
832 | bio->bi_rw |= (!write_to_vm << BIO_RW); | |
833 | ||
834 | ret = 0; | |
56c451f4 FT |
835 | |
836 | if (map_data) { | |
e623ddb4 | 837 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
838 | i = map_data->offset / PAGE_SIZE; |
839 | } | |
1da177e4 | 840 | while (len) { |
e623ddb4 | 841 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 842 | |
56c451f4 FT |
843 | bytes -= offset; |
844 | ||
1da177e4 LT |
845 | if (bytes > len) |
846 | bytes = len; | |
847 | ||
152e283f | 848 | if (map_data) { |
e623ddb4 | 849 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
850 | ret = -ENOMEM; |
851 | break; | |
852 | } | |
e623ddb4 FT |
853 | |
854 | page = map_data->pages[i / nr_pages]; | |
855 | page += (i % nr_pages); | |
856 | ||
857 | i++; | |
858 | } else { | |
152e283f | 859 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
860 | if (!page) { |
861 | ret = -ENOMEM; | |
862 | break; | |
863 | } | |
1da177e4 LT |
864 | } |
865 | ||
56c451f4 | 866 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 867 | break; |
1da177e4 LT |
868 | |
869 | len -= bytes; | |
56c451f4 | 870 | offset = 0; |
1da177e4 LT |
871 | } |
872 | ||
873 | if (ret) | |
874 | goto cleanup; | |
875 | ||
876 | /* | |
877 | * success | |
878 | */ | |
97ae77a1 | 879 | if (!write_to_vm && (!map_data || !map_data->null_mapped)) { |
152e283f | 880 | ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 0); |
c5dec1c3 FT |
881 | if (ret) |
882 | goto cleanup; | |
1da177e4 LT |
883 | } |
884 | ||
152e283f | 885 | bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1); |
1da177e4 LT |
886 | return bio; |
887 | cleanup: | |
152e283f FT |
888 | if (!map_data) |
889 | bio_for_each_segment(bvec, bio, i) | |
890 | __free_page(bvec->bv_page); | |
1da177e4 LT |
891 | |
892 | bio_put(bio); | |
893 | out_bmd: | |
894 | bio_free_map_data(bmd); | |
895 | return ERR_PTR(ret); | |
896 | } | |
897 | ||
c5dec1c3 FT |
898 | /** |
899 | * bio_copy_user - copy user data to bio | |
900 | * @q: destination block queue | |
152e283f | 901 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
902 | * @uaddr: start of user address |
903 | * @len: length in bytes | |
904 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 905 | * @gfp_mask: memory allocation flags |
c5dec1c3 FT |
906 | * |
907 | * Prepares and returns a bio for indirect user io, bouncing data | |
908 | * to/from kernel pages as necessary. Must be paired with | |
909 | * call bio_uncopy_user() on io completion. | |
910 | */ | |
152e283f FT |
911 | struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data, |
912 | unsigned long uaddr, unsigned int len, | |
913 | int write_to_vm, gfp_t gfp_mask) | |
c5dec1c3 FT |
914 | { |
915 | struct sg_iovec iov; | |
916 | ||
917 | iov.iov_base = (void __user *)uaddr; | |
918 | iov.iov_len = len; | |
919 | ||
152e283f | 920 | return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask); |
c5dec1c3 FT |
921 | } |
922 | ||
165125e1 | 923 | static struct bio *__bio_map_user_iov(struct request_queue *q, |
f1970baf JB |
924 | struct block_device *bdev, |
925 | struct sg_iovec *iov, int iov_count, | |
a3bce90e | 926 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 | 927 | { |
f1970baf JB |
928 | int i, j; |
929 | int nr_pages = 0; | |
1da177e4 LT |
930 | struct page **pages; |
931 | struct bio *bio; | |
f1970baf JB |
932 | int cur_page = 0; |
933 | int ret, offset; | |
1da177e4 | 934 | |
f1970baf JB |
935 | for (i = 0; i < iov_count; i++) { |
936 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
937 | unsigned long len = iov[i].iov_len; | |
938 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
939 | unsigned long start = uaddr >> PAGE_SHIFT; | |
940 | ||
941 | nr_pages += end - start; | |
942 | /* | |
ad2d7225 | 943 | * buffer must be aligned to at least hardsector size for now |
f1970baf | 944 | */ |
ad2d7225 | 945 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
946 | return ERR_PTR(-EINVAL); |
947 | } | |
948 | ||
949 | if (!nr_pages) | |
1da177e4 LT |
950 | return ERR_PTR(-EINVAL); |
951 | ||
a9e9dc24 | 952 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
953 | if (!bio) |
954 | return ERR_PTR(-ENOMEM); | |
955 | ||
956 | ret = -ENOMEM; | |
a3bce90e | 957 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
958 | if (!pages) |
959 | goto out; | |
960 | ||
f1970baf JB |
961 | for (i = 0; i < iov_count; i++) { |
962 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
963 | unsigned long len = iov[i].iov_len; | |
964 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
965 | unsigned long start = uaddr >> PAGE_SHIFT; | |
966 | const int local_nr_pages = end - start; | |
967 | const int page_limit = cur_page + local_nr_pages; | |
968 | ||
f5dd33c4 NP |
969 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
970 | write_to_vm, &pages[cur_page]); | |
99172157 JA |
971 | if (ret < local_nr_pages) { |
972 | ret = -EFAULT; | |
f1970baf | 973 | goto out_unmap; |
99172157 | 974 | } |
f1970baf JB |
975 | |
976 | offset = uaddr & ~PAGE_MASK; | |
977 | for (j = cur_page; j < page_limit; j++) { | |
978 | unsigned int bytes = PAGE_SIZE - offset; | |
979 | ||
980 | if (len <= 0) | |
981 | break; | |
982 | ||
983 | if (bytes > len) | |
984 | bytes = len; | |
985 | ||
986 | /* | |
987 | * sorry... | |
988 | */ | |
defd94b7 MC |
989 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
990 | bytes) | |
f1970baf JB |
991 | break; |
992 | ||
993 | len -= bytes; | |
994 | offset = 0; | |
995 | } | |
1da177e4 | 996 | |
f1970baf | 997 | cur_page = j; |
1da177e4 | 998 | /* |
f1970baf | 999 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1000 | */ |
f1970baf JB |
1001 | while (j < page_limit) |
1002 | page_cache_release(pages[j++]); | |
1da177e4 LT |
1003 | } |
1004 | ||
1da177e4 LT |
1005 | kfree(pages); |
1006 | ||
1007 | /* | |
1008 | * set data direction, and check if mapped pages need bouncing | |
1009 | */ | |
1010 | if (!write_to_vm) | |
1011 | bio->bi_rw |= (1 << BIO_RW); | |
1012 | ||
f1970baf | 1013 | bio->bi_bdev = bdev; |
1da177e4 LT |
1014 | bio->bi_flags |= (1 << BIO_USER_MAPPED); |
1015 | return bio; | |
f1970baf JB |
1016 | |
1017 | out_unmap: | |
1018 | for (i = 0; i < nr_pages; i++) { | |
1019 | if(!pages[i]) | |
1020 | break; | |
1021 | page_cache_release(pages[i]); | |
1022 | } | |
1023 | out: | |
1da177e4 LT |
1024 | kfree(pages); |
1025 | bio_put(bio); | |
1026 | return ERR_PTR(ret); | |
1027 | } | |
1028 | ||
1029 | /** | |
1030 | * bio_map_user - map user address into bio | |
165125e1 | 1031 | * @q: the struct request_queue for the bio |
1da177e4 LT |
1032 | * @bdev: destination block device |
1033 | * @uaddr: start of user address | |
1034 | * @len: length in bytes | |
1035 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1036 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
1037 | * |
1038 | * Map the user space address into a bio suitable for io to a block | |
1039 | * device. Returns an error pointer in case of error. | |
1040 | */ | |
165125e1 | 1041 | struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev, |
a3bce90e FT |
1042 | unsigned long uaddr, unsigned int len, int write_to_vm, |
1043 | gfp_t gfp_mask) | |
f1970baf JB |
1044 | { |
1045 | struct sg_iovec iov; | |
1046 | ||
3f70353e | 1047 | iov.iov_base = (void __user *)uaddr; |
f1970baf JB |
1048 | iov.iov_len = len; |
1049 | ||
a3bce90e | 1050 | return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask); |
f1970baf JB |
1051 | } |
1052 | ||
1053 | /** | |
1054 | * bio_map_user_iov - map user sg_iovec table into bio | |
165125e1 | 1055 | * @q: the struct request_queue for the bio |
f1970baf JB |
1056 | * @bdev: destination block device |
1057 | * @iov: the iovec. | |
1058 | * @iov_count: number of elements in the iovec | |
1059 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1060 | * @gfp_mask: memory allocation flags |
f1970baf JB |
1061 | * |
1062 | * Map the user space address into a bio suitable for io to a block | |
1063 | * device. Returns an error pointer in case of error. | |
1064 | */ | |
165125e1 | 1065 | struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev, |
f1970baf | 1066 | struct sg_iovec *iov, int iov_count, |
a3bce90e | 1067 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 LT |
1068 | { |
1069 | struct bio *bio; | |
1070 | ||
a3bce90e FT |
1071 | bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm, |
1072 | gfp_mask); | |
1da177e4 LT |
1073 | if (IS_ERR(bio)) |
1074 | return bio; | |
1075 | ||
1076 | /* | |
1077 | * subtle -- if __bio_map_user() ended up bouncing a bio, | |
1078 | * it would normally disappear when its bi_end_io is run. | |
1079 | * however, we need it for the unmap, so grab an extra | |
1080 | * reference to it | |
1081 | */ | |
1082 | bio_get(bio); | |
1083 | ||
0e75f906 | 1084 | return bio; |
1da177e4 LT |
1085 | } |
1086 | ||
1087 | static void __bio_unmap_user(struct bio *bio) | |
1088 | { | |
1089 | struct bio_vec *bvec; | |
1090 | int i; | |
1091 | ||
1092 | /* | |
1093 | * make sure we dirty pages we wrote to | |
1094 | */ | |
1095 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1096 | if (bio_data_dir(bio) == READ) | |
1097 | set_page_dirty_lock(bvec->bv_page); | |
1098 | ||
1099 | page_cache_release(bvec->bv_page); | |
1100 | } | |
1101 | ||
1102 | bio_put(bio); | |
1103 | } | |
1104 | ||
1105 | /** | |
1106 | * bio_unmap_user - unmap a bio | |
1107 | * @bio: the bio being unmapped | |
1108 | * | |
1109 | * Unmap a bio previously mapped by bio_map_user(). Must be called with | |
1110 | * a process context. | |
1111 | * | |
1112 | * bio_unmap_user() may sleep. | |
1113 | */ | |
1114 | void bio_unmap_user(struct bio *bio) | |
1115 | { | |
1116 | __bio_unmap_user(bio); | |
1117 | bio_put(bio); | |
1118 | } | |
1119 | ||
6712ecf8 | 1120 | static void bio_map_kern_endio(struct bio *bio, int err) |
b823825e | 1121 | { |
b823825e | 1122 | bio_put(bio); |
b823825e JA |
1123 | } |
1124 | ||
1125 | ||
165125e1 | 1126 | static struct bio *__bio_map_kern(struct request_queue *q, void *data, |
27496a8c | 1127 | unsigned int len, gfp_t gfp_mask) |
df46b9a4 MC |
1128 | { |
1129 | unsigned long kaddr = (unsigned long)data; | |
1130 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1131 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1132 | const int nr_pages = end - start; | |
1133 | int offset, i; | |
1134 | struct bio *bio; | |
1135 | ||
a9e9dc24 | 1136 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1137 | if (!bio) |
1138 | return ERR_PTR(-ENOMEM); | |
1139 | ||
1140 | offset = offset_in_page(kaddr); | |
1141 | for (i = 0; i < nr_pages; i++) { | |
1142 | unsigned int bytes = PAGE_SIZE - offset; | |
1143 | ||
1144 | if (len <= 0) | |
1145 | break; | |
1146 | ||
1147 | if (bytes > len) | |
1148 | bytes = len; | |
1149 | ||
defd94b7 MC |
1150 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
1151 | offset) < bytes) | |
df46b9a4 MC |
1152 | break; |
1153 | ||
1154 | data += bytes; | |
1155 | len -= bytes; | |
1156 | offset = 0; | |
1157 | } | |
1158 | ||
b823825e | 1159 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1160 | return bio; |
1161 | } | |
1162 | ||
1163 | /** | |
1164 | * bio_map_kern - map kernel address into bio | |
165125e1 | 1165 | * @q: the struct request_queue for the bio |
df46b9a4 MC |
1166 | * @data: pointer to buffer to map |
1167 | * @len: length in bytes | |
1168 | * @gfp_mask: allocation flags for bio allocation | |
1169 | * | |
1170 | * Map the kernel address into a bio suitable for io to a block | |
1171 | * device. Returns an error pointer in case of error. | |
1172 | */ | |
165125e1 | 1173 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
27496a8c | 1174 | gfp_t gfp_mask) |
df46b9a4 MC |
1175 | { |
1176 | struct bio *bio; | |
1177 | ||
1178 | bio = __bio_map_kern(q, data, len, gfp_mask); | |
1179 | if (IS_ERR(bio)) | |
1180 | return bio; | |
1181 | ||
1182 | if (bio->bi_size == len) | |
1183 | return bio; | |
1184 | ||
1185 | /* | |
1186 | * Don't support partial mappings. | |
1187 | */ | |
1188 | bio_put(bio); | |
1189 | return ERR_PTR(-EINVAL); | |
1190 | } | |
1191 | ||
68154e90 FT |
1192 | static void bio_copy_kern_endio(struct bio *bio, int err) |
1193 | { | |
1194 | struct bio_vec *bvec; | |
1195 | const int read = bio_data_dir(bio) == READ; | |
76029ff3 | 1196 | struct bio_map_data *bmd = bio->bi_private; |
68154e90 | 1197 | int i; |
76029ff3 | 1198 | char *p = bmd->sgvecs[0].iov_base; |
68154e90 FT |
1199 | |
1200 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1201 | char *addr = page_address(bvec->bv_page); | |
76029ff3 | 1202 | int len = bmd->iovecs[i].bv_len; |
68154e90 | 1203 | |
4fc981ef | 1204 | if (read) |
76029ff3 | 1205 | memcpy(p, addr, len); |
68154e90 FT |
1206 | |
1207 | __free_page(bvec->bv_page); | |
76029ff3 | 1208 | p += len; |
68154e90 FT |
1209 | } |
1210 | ||
76029ff3 | 1211 | bio_free_map_data(bmd); |
68154e90 FT |
1212 | bio_put(bio); |
1213 | } | |
1214 | ||
1215 | /** | |
1216 | * bio_copy_kern - copy kernel address into bio | |
1217 | * @q: the struct request_queue for the bio | |
1218 | * @data: pointer to buffer to copy | |
1219 | * @len: length in bytes | |
1220 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1221 | * @reading: data direction is READ |
68154e90 FT |
1222 | * |
1223 | * copy the kernel address into a bio suitable for io to a block | |
1224 | * device. Returns an error pointer in case of error. | |
1225 | */ | |
1226 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1227 | gfp_t gfp_mask, int reading) | |
1228 | { | |
68154e90 FT |
1229 | struct bio *bio; |
1230 | struct bio_vec *bvec; | |
4d8ab62e | 1231 | int i; |
68154e90 | 1232 | |
4d8ab62e FT |
1233 | bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask); |
1234 | if (IS_ERR(bio)) | |
1235 | return bio; | |
68154e90 FT |
1236 | |
1237 | if (!reading) { | |
1238 | void *p = data; | |
1239 | ||
1240 | bio_for_each_segment(bvec, bio, i) { | |
1241 | char *addr = page_address(bvec->bv_page); | |
1242 | ||
1243 | memcpy(addr, p, bvec->bv_len); | |
1244 | p += bvec->bv_len; | |
1245 | } | |
1246 | } | |
1247 | ||
68154e90 | 1248 | bio->bi_end_io = bio_copy_kern_endio; |
76029ff3 | 1249 | |
68154e90 | 1250 | return bio; |
68154e90 FT |
1251 | } |
1252 | ||
1da177e4 LT |
1253 | /* |
1254 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1255 | * for performing direct-IO in BIOs. | |
1256 | * | |
1257 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1258 | * because the required locks are not interrupt-safe. So what we can do is to | |
1259 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1260 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1261 | * in process context. | |
1262 | * | |
1263 | * We special-case compound pages here: normally this means reads into hugetlb | |
1264 | * pages. The logic in here doesn't really work right for compound pages | |
1265 | * because the VM does not uniformly chase down the head page in all cases. | |
1266 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1267 | * handle them at all. So we skip compound pages here at an early stage. | |
1268 | * | |
1269 | * Note that this code is very hard to test under normal circumstances because | |
1270 | * direct-io pins the pages with get_user_pages(). This makes | |
1271 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
1272 | * But other code (eg, pdflush) could clean the pages if they are mapped | |
1273 | * pagecache. | |
1274 | * | |
1275 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1276 | * deferred bio dirtying paths. | |
1277 | */ | |
1278 | ||
1279 | /* | |
1280 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1281 | */ | |
1282 | void bio_set_pages_dirty(struct bio *bio) | |
1283 | { | |
1284 | struct bio_vec *bvec = bio->bi_io_vec; | |
1285 | int i; | |
1286 | ||
1287 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1288 | struct page *page = bvec[i].bv_page; | |
1289 | ||
1290 | if (page && !PageCompound(page)) | |
1291 | set_page_dirty_lock(page); | |
1292 | } | |
1293 | } | |
1294 | ||
86b6c7a7 | 1295 | static void bio_release_pages(struct bio *bio) |
1da177e4 LT |
1296 | { |
1297 | struct bio_vec *bvec = bio->bi_io_vec; | |
1298 | int i; | |
1299 | ||
1300 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1301 | struct page *page = bvec[i].bv_page; | |
1302 | ||
1303 | if (page) | |
1304 | put_page(page); | |
1305 | } | |
1306 | } | |
1307 | ||
1308 | /* | |
1309 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1310 | * If they are, then fine. If, however, some pages are clean then they must | |
1311 | * have been written out during the direct-IO read. So we take another ref on | |
1312 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1313 | * | |
1314 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
1315 | * here on. It will run one page_cache_release() against each page and will | |
1316 | * run one bio_put() against the BIO. | |
1317 | */ | |
1318 | ||
65f27f38 | 1319 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1320 | |
65f27f38 | 1321 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1322 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1323 | static struct bio *bio_dirty_list; | |
1324 | ||
1325 | /* | |
1326 | * This runs in process context | |
1327 | */ | |
65f27f38 | 1328 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1329 | { |
1330 | unsigned long flags; | |
1331 | struct bio *bio; | |
1332 | ||
1333 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1334 | bio = bio_dirty_list; | |
1335 | bio_dirty_list = NULL; | |
1336 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1337 | ||
1338 | while (bio) { | |
1339 | struct bio *next = bio->bi_private; | |
1340 | ||
1341 | bio_set_pages_dirty(bio); | |
1342 | bio_release_pages(bio); | |
1343 | bio_put(bio); | |
1344 | bio = next; | |
1345 | } | |
1346 | } | |
1347 | ||
1348 | void bio_check_pages_dirty(struct bio *bio) | |
1349 | { | |
1350 | struct bio_vec *bvec = bio->bi_io_vec; | |
1351 | int nr_clean_pages = 0; | |
1352 | int i; | |
1353 | ||
1354 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1355 | struct page *page = bvec[i].bv_page; | |
1356 | ||
1357 | if (PageDirty(page) || PageCompound(page)) { | |
1358 | page_cache_release(page); | |
1359 | bvec[i].bv_page = NULL; | |
1360 | } else { | |
1361 | nr_clean_pages++; | |
1362 | } | |
1363 | } | |
1364 | ||
1365 | if (nr_clean_pages) { | |
1366 | unsigned long flags; | |
1367 | ||
1368 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1369 | bio->bi_private = bio_dirty_list; | |
1370 | bio_dirty_list = bio; | |
1371 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1372 | schedule_work(&bio_dirty_work); | |
1373 | } else { | |
1374 | bio_put(bio); | |
1375 | } | |
1376 | } | |
1377 | ||
1378 | /** | |
1379 | * bio_endio - end I/O on a bio | |
1380 | * @bio: bio | |
1da177e4 LT |
1381 | * @error: error, if any |
1382 | * | |
1383 | * Description: | |
6712ecf8 | 1384 | * bio_endio() will end I/O on the whole bio. bio_endio() is the |
5bb23a68 N |
1385 | * preferred way to end I/O on a bio, it takes care of clearing |
1386 | * BIO_UPTODATE on error. @error is 0 on success, and and one of the | |
1387 | * established -Exxxx (-EIO, for instance) error values in case | |
1388 | * something went wrong. Noone should call bi_end_io() directly on a | |
1389 | * bio unless they own it and thus know that it has an end_io | |
1390 | * function. | |
1da177e4 | 1391 | **/ |
6712ecf8 | 1392 | void bio_endio(struct bio *bio, int error) |
1da177e4 LT |
1393 | { |
1394 | if (error) | |
1395 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
9cc54d40 N |
1396 | else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) |
1397 | error = -EIO; | |
1da177e4 | 1398 | |
5bb23a68 | 1399 | if (bio->bi_end_io) |
6712ecf8 | 1400 | bio->bi_end_io(bio, error); |
1da177e4 LT |
1401 | } |
1402 | ||
1403 | void bio_pair_release(struct bio_pair *bp) | |
1404 | { | |
1405 | if (atomic_dec_and_test(&bp->cnt)) { | |
1406 | struct bio *master = bp->bio1.bi_private; | |
1407 | ||
6712ecf8 | 1408 | bio_endio(master, bp->error); |
1da177e4 LT |
1409 | mempool_free(bp, bp->bio2.bi_private); |
1410 | } | |
1411 | } | |
1412 | ||
6712ecf8 | 1413 | static void bio_pair_end_1(struct bio *bi, int err) |
1da177e4 LT |
1414 | { |
1415 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio1); | |
1416 | ||
1417 | if (err) | |
1418 | bp->error = err; | |
1419 | ||
1da177e4 | 1420 | bio_pair_release(bp); |
1da177e4 LT |
1421 | } |
1422 | ||
6712ecf8 | 1423 | static void bio_pair_end_2(struct bio *bi, int err) |
1da177e4 LT |
1424 | { |
1425 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio2); | |
1426 | ||
1427 | if (err) | |
1428 | bp->error = err; | |
1429 | ||
1da177e4 | 1430 | bio_pair_release(bp); |
1da177e4 LT |
1431 | } |
1432 | ||
1433 | /* | |
c7eee1b8 | 1434 | * split a bio - only worry about a bio with a single page in its iovec |
1da177e4 | 1435 | */ |
6feef531 | 1436 | struct bio_pair *bio_split(struct bio *bi, int first_sectors) |
1da177e4 | 1437 | { |
6feef531 | 1438 | struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO); |
1da177e4 LT |
1439 | |
1440 | if (!bp) | |
1441 | return bp; | |
1442 | ||
5f3ea37c | 1443 | trace_block_split(bdev_get_queue(bi->bi_bdev), bi, |
2056a782 JA |
1444 | bi->bi_sector + first_sectors); |
1445 | ||
1da177e4 LT |
1446 | BUG_ON(bi->bi_vcnt != 1); |
1447 | BUG_ON(bi->bi_idx != 0); | |
1448 | atomic_set(&bp->cnt, 3); | |
1449 | bp->error = 0; | |
1450 | bp->bio1 = *bi; | |
1451 | bp->bio2 = *bi; | |
1452 | bp->bio2.bi_sector += first_sectors; | |
1453 | bp->bio2.bi_size -= first_sectors << 9; | |
1454 | bp->bio1.bi_size = first_sectors << 9; | |
1455 | ||
1456 | bp->bv1 = bi->bi_io_vec[0]; | |
1457 | bp->bv2 = bi->bi_io_vec[0]; | |
1458 | bp->bv2.bv_offset += first_sectors << 9; | |
1459 | bp->bv2.bv_len -= first_sectors << 9; | |
1460 | bp->bv1.bv_len = first_sectors << 9; | |
1461 | ||
1462 | bp->bio1.bi_io_vec = &bp->bv1; | |
1463 | bp->bio2.bi_io_vec = &bp->bv2; | |
1464 | ||
a2eb0c10 N |
1465 | bp->bio1.bi_max_vecs = 1; |
1466 | bp->bio2.bi_max_vecs = 1; | |
1467 | ||
1da177e4 LT |
1468 | bp->bio1.bi_end_io = bio_pair_end_1; |
1469 | bp->bio2.bi_end_io = bio_pair_end_2; | |
1470 | ||
1471 | bp->bio1.bi_private = bi; | |
6feef531 | 1472 | bp->bio2.bi_private = bio_split_pool; |
1da177e4 | 1473 | |
7ba1ba12 MP |
1474 | if (bio_integrity(bi)) |
1475 | bio_integrity_split(bi, bp, first_sectors); | |
1476 | ||
1da177e4 LT |
1477 | return bp; |
1478 | } | |
1479 | ||
ad3316bf MP |
1480 | /** |
1481 | * bio_sector_offset - Find hardware sector offset in bio | |
1482 | * @bio: bio to inspect | |
1483 | * @index: bio_vec index | |
1484 | * @offset: offset in bv_page | |
1485 | * | |
1486 | * Return the number of hardware sectors between beginning of bio | |
1487 | * and an end point indicated by a bio_vec index and an offset | |
1488 | * within that vector's page. | |
1489 | */ | |
1490 | sector_t bio_sector_offset(struct bio *bio, unsigned short index, | |
1491 | unsigned int offset) | |
1492 | { | |
e1defc4f | 1493 | unsigned int sector_sz; |
ad3316bf MP |
1494 | struct bio_vec *bv; |
1495 | sector_t sectors; | |
1496 | int i; | |
1497 | ||
e1defc4f | 1498 | sector_sz = queue_logical_block_size(bio->bi_bdev->bd_disk->queue); |
ad3316bf MP |
1499 | sectors = 0; |
1500 | ||
1501 | if (index >= bio->bi_idx) | |
1502 | index = bio->bi_vcnt - 1; | |
1503 | ||
1504 | __bio_for_each_segment(bv, bio, i, 0) { | |
1505 | if (i == index) { | |
1506 | if (offset > bv->bv_offset) | |
1507 | sectors += (offset - bv->bv_offset) / sector_sz; | |
1508 | break; | |
1509 | } | |
1510 | ||
1511 | sectors += bv->bv_len / sector_sz; | |
1512 | } | |
1513 | ||
1514 | return sectors; | |
1515 | } | |
1516 | EXPORT_SYMBOL(bio_sector_offset); | |
1da177e4 LT |
1517 | |
1518 | /* | |
1519 | * create memory pools for biovec's in a bio_set. | |
1520 | * use the global biovec slabs created for general use. | |
1521 | */ | |
5972511b | 1522 | static int biovec_create_pools(struct bio_set *bs, int pool_entries) |
1da177e4 | 1523 | { |
7ff9345f | 1524 | struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX; |
1da177e4 | 1525 | |
7ff9345f JA |
1526 | bs->bvec_pool = mempool_create_slab_pool(pool_entries, bp->slab); |
1527 | if (!bs->bvec_pool) | |
1528 | return -ENOMEM; | |
1da177e4 | 1529 | |
1da177e4 LT |
1530 | return 0; |
1531 | } | |
1532 | ||
1533 | static void biovec_free_pools(struct bio_set *bs) | |
1534 | { | |
7ff9345f | 1535 | mempool_destroy(bs->bvec_pool); |
1da177e4 LT |
1536 | } |
1537 | ||
1538 | void bioset_free(struct bio_set *bs) | |
1539 | { | |
1540 | if (bs->bio_pool) | |
1541 | mempool_destroy(bs->bio_pool); | |
1542 | ||
1543 | biovec_free_pools(bs); | |
bb799ca0 | 1544 | bio_put_slab(bs); |
1da177e4 LT |
1545 | |
1546 | kfree(bs); | |
1547 | } | |
1548 | ||
bb799ca0 JA |
1549 | /** |
1550 | * bioset_create - Create a bio_set | |
1551 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1552 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1553 | * | |
1554 | * Description: | |
1555 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1556 | * to ask for a number of bytes to be allocated in front of the bio. | |
1557 | * Front pad allocation is useful for embedding the bio inside | |
1558 | * another structure, to avoid allocating extra data to go with the bio. | |
1559 | * Note that the bio must be embedded at the END of that structure always, | |
1560 | * or things will break badly. | |
1561 | */ | |
1562 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) | |
1da177e4 | 1563 | { |
392ddc32 | 1564 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1565 | struct bio_set *bs; |
1da177e4 | 1566 | |
1b434498 | 1567 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1568 | if (!bs) |
1569 | return NULL; | |
1570 | ||
bb799ca0 | 1571 | bs->front_pad = front_pad; |
1b434498 | 1572 | |
392ddc32 | 1573 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1574 | if (!bs->bio_slab) { |
1575 | kfree(bs); | |
1576 | return NULL; | |
1577 | } | |
1578 | ||
1579 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
1580 | if (!bs->bio_pool) |
1581 | goto bad; | |
1582 | ||
bb799ca0 | 1583 | if (!biovec_create_pools(bs, pool_size)) |
1da177e4 LT |
1584 | return bs; |
1585 | ||
1586 | bad: | |
1587 | bioset_free(bs); | |
1588 | return NULL; | |
1589 | } | |
1590 | ||
1591 | static void __init biovec_init_slabs(void) | |
1592 | { | |
1593 | int i; | |
1594 | ||
1595 | for (i = 0; i < BIOVEC_NR_POOLS; i++) { | |
1596 | int size; | |
1597 | struct biovec_slab *bvs = bvec_slabs + i; | |
1598 | ||
a7fcd37c JA |
1599 | #ifndef CONFIG_BLK_DEV_INTEGRITY |
1600 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { | |
1601 | bvs->slab = NULL; | |
1602 | continue; | |
1603 | } | |
1604 | #endif | |
1605 | ||
1da177e4 LT |
1606 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
1607 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 1608 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1609 | } |
1610 | } | |
1611 | ||
1612 | static int __init init_bio(void) | |
1613 | { | |
bb799ca0 JA |
1614 | bio_slab_max = 2; |
1615 | bio_slab_nr = 0; | |
1616 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
1617 | if (!bio_slabs) | |
1618 | panic("bio: can't allocate bios\n"); | |
1da177e4 LT |
1619 | |
1620 | biovec_init_slabs(); | |
1621 | ||
bb799ca0 | 1622 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
1da177e4 LT |
1623 | if (!fs_bio_set) |
1624 | panic("bio: can't allocate bios\n"); | |
1625 | ||
0eaae62a MD |
1626 | bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES, |
1627 | sizeof(struct bio_pair)); | |
1da177e4 LT |
1628 | if (!bio_split_pool) |
1629 | panic("bio: can't create split pool\n"); | |
1630 | ||
1631 | return 0; | |
1632 | } | |
1633 | ||
1634 | subsys_initcall(init_bio); | |
1635 | ||
1636 | EXPORT_SYMBOL(bio_alloc); | |
0a0d96b0 | 1637 | EXPORT_SYMBOL(bio_kmalloc); |
1da177e4 | 1638 | EXPORT_SYMBOL(bio_put); |
3676347a | 1639 | EXPORT_SYMBOL(bio_free); |
1da177e4 LT |
1640 | EXPORT_SYMBOL(bio_endio); |
1641 | EXPORT_SYMBOL(bio_init); | |
1642 | EXPORT_SYMBOL(__bio_clone); | |
1643 | EXPORT_SYMBOL(bio_clone); | |
1644 | EXPORT_SYMBOL(bio_phys_segments); | |
1da177e4 | 1645 | EXPORT_SYMBOL(bio_add_page); |
6e68af66 | 1646 | EXPORT_SYMBOL(bio_add_pc_page); |
1da177e4 | 1647 | EXPORT_SYMBOL(bio_get_nr_vecs); |
40044ce0 JA |
1648 | EXPORT_SYMBOL(bio_map_user); |
1649 | EXPORT_SYMBOL(bio_unmap_user); | |
df46b9a4 | 1650 | EXPORT_SYMBOL(bio_map_kern); |
68154e90 | 1651 | EXPORT_SYMBOL(bio_copy_kern); |
1da177e4 LT |
1652 | EXPORT_SYMBOL(bio_pair_release); |
1653 | EXPORT_SYMBOL(bio_split); | |
1da177e4 LT |
1654 | EXPORT_SYMBOL(bio_copy_user); |
1655 | EXPORT_SYMBOL(bio_uncopy_user); | |
1656 | EXPORT_SYMBOL(bioset_create); | |
1657 | EXPORT_SYMBOL(bioset_free); | |
1658 | EXPORT_SYMBOL(bio_alloc_bioset); |