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> | |
f1970baf | 28 | #include <scsi/sg.h> /* for struct sg_iovec */ |
1da177e4 | 29 | |
55782138 | 30 | #include <trace/events/block.h> |
0bfc2455 | 31 | |
392ddc32 JA |
32 | /* |
33 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
34 | * itself, to shrink a bio data allocation from two mempool calls to one | |
35 | */ | |
36 | #define BIO_INLINE_VECS 4 | |
37 | ||
6feef531 | 38 | static mempool_t *bio_split_pool __read_mostly; |
1da177e4 | 39 | |
1da177e4 LT |
40 | /* |
41 | * if you change this list, also change bvec_alloc or things will | |
42 | * break badly! cannot be bigger than what you can fit into an | |
43 | * unsigned short | |
44 | */ | |
1da177e4 | 45 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
bb799ca0 | 46 | struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = { |
1da177e4 LT |
47 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
48 | }; | |
49 | #undef BV | |
50 | ||
1da177e4 LT |
51 | /* |
52 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
53 | * IO code that does not need private memory pools. | |
54 | */ | |
51d654e1 | 55 | struct bio_set *fs_bio_set; |
1da177e4 | 56 | |
bb799ca0 JA |
57 | /* |
58 | * Our slab pool management | |
59 | */ | |
60 | struct bio_slab { | |
61 | struct kmem_cache *slab; | |
62 | unsigned int slab_ref; | |
63 | unsigned int slab_size; | |
64 | char name[8]; | |
65 | }; | |
66 | static DEFINE_MUTEX(bio_slab_lock); | |
67 | static struct bio_slab *bio_slabs; | |
68 | static unsigned int bio_slab_nr, bio_slab_max; | |
69 | ||
70 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
71 | { | |
72 | unsigned int sz = sizeof(struct bio) + extra_size; | |
73 | struct kmem_cache *slab = NULL; | |
74 | struct bio_slab *bslab; | |
75 | unsigned int i, entry = -1; | |
76 | ||
77 | mutex_lock(&bio_slab_lock); | |
78 | ||
79 | i = 0; | |
80 | while (i < bio_slab_nr) { | |
f06f135d | 81 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
82 | |
83 | if (!bslab->slab && entry == -1) | |
84 | entry = i; | |
85 | else if (bslab->slab_size == sz) { | |
86 | slab = bslab->slab; | |
87 | bslab->slab_ref++; | |
88 | break; | |
89 | } | |
90 | i++; | |
91 | } | |
92 | ||
93 | if (slab) | |
94 | goto out_unlock; | |
95 | ||
96 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
97 | bio_slab_max <<= 1; | |
98 | bio_slabs = krealloc(bio_slabs, | |
99 | bio_slab_max * sizeof(struct bio_slab), | |
100 | GFP_KERNEL); | |
101 | if (!bio_slabs) | |
102 | goto out_unlock; | |
103 | } | |
104 | if (entry == -1) | |
105 | entry = bio_slab_nr++; | |
106 | ||
107 | bslab = &bio_slabs[entry]; | |
108 | ||
109 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
110 | slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL); | |
111 | if (!slab) | |
112 | goto out_unlock; | |
113 | ||
114 | printk("bio: create slab <%s> at %d\n", bslab->name, entry); | |
115 | bslab->slab = slab; | |
116 | bslab->slab_ref = 1; | |
117 | bslab->slab_size = sz; | |
118 | out_unlock: | |
119 | mutex_unlock(&bio_slab_lock); | |
120 | return slab; | |
121 | } | |
122 | ||
123 | static void bio_put_slab(struct bio_set *bs) | |
124 | { | |
125 | struct bio_slab *bslab = NULL; | |
126 | unsigned int i; | |
127 | ||
128 | mutex_lock(&bio_slab_lock); | |
129 | ||
130 | for (i = 0; i < bio_slab_nr; i++) { | |
131 | if (bs->bio_slab == bio_slabs[i].slab) { | |
132 | bslab = &bio_slabs[i]; | |
133 | break; | |
134 | } | |
135 | } | |
136 | ||
137 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
138 | goto out; | |
139 | ||
140 | WARN_ON(!bslab->slab_ref); | |
141 | ||
142 | if (--bslab->slab_ref) | |
143 | goto out; | |
144 | ||
145 | kmem_cache_destroy(bslab->slab); | |
146 | bslab->slab = NULL; | |
147 | ||
148 | out: | |
149 | mutex_unlock(&bio_slab_lock); | |
150 | } | |
151 | ||
7ba1ba12 MP |
152 | unsigned int bvec_nr_vecs(unsigned short idx) |
153 | { | |
154 | return bvec_slabs[idx].nr_vecs; | |
155 | } | |
156 | ||
bb799ca0 JA |
157 | void bvec_free_bs(struct bio_set *bs, struct bio_vec *bv, unsigned int idx) |
158 | { | |
159 | BIO_BUG_ON(idx >= BIOVEC_NR_POOLS); | |
160 | ||
161 | if (idx == BIOVEC_MAX_IDX) | |
162 | mempool_free(bv, bs->bvec_pool); | |
163 | else { | |
164 | struct biovec_slab *bvs = bvec_slabs + idx; | |
165 | ||
166 | kmem_cache_free(bvs->slab, bv); | |
167 | } | |
168 | } | |
169 | ||
7ff9345f JA |
170 | struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, |
171 | struct bio_set *bs) | |
1da177e4 LT |
172 | { |
173 | struct bio_vec *bvl; | |
1da177e4 | 174 | |
7ff9345f JA |
175 | /* |
176 | * see comment near bvec_array define! | |
177 | */ | |
178 | switch (nr) { | |
179 | case 1: | |
180 | *idx = 0; | |
181 | break; | |
182 | case 2 ... 4: | |
183 | *idx = 1; | |
184 | break; | |
185 | case 5 ... 16: | |
186 | *idx = 2; | |
187 | break; | |
188 | case 17 ... 64: | |
189 | *idx = 3; | |
190 | break; | |
191 | case 65 ... 128: | |
192 | *idx = 4; | |
193 | break; | |
194 | case 129 ... BIO_MAX_PAGES: | |
195 | *idx = 5; | |
196 | break; | |
197 | default: | |
198 | return NULL; | |
199 | } | |
200 | ||
201 | /* | |
202 | * idx now points to the pool we want to allocate from. only the | |
203 | * 1-vec entry pool is mempool backed. | |
204 | */ | |
205 | if (*idx == BIOVEC_MAX_IDX) { | |
206 | fallback: | |
207 | bvl = mempool_alloc(bs->bvec_pool, gfp_mask); | |
208 | } else { | |
209 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
210 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO); | |
211 | ||
0a0d96b0 | 212 | /* |
7ff9345f JA |
213 | * Make this allocation restricted and don't dump info on |
214 | * allocation failures, since we'll fallback to the mempool | |
215 | * in case of failure. | |
0a0d96b0 | 216 | */ |
7ff9345f | 217 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 218 | |
0a0d96b0 | 219 | /* |
7ff9345f JA |
220 | * Try a slab allocation. If this fails and __GFP_WAIT |
221 | * is set, retry with the 1-entry mempool | |
0a0d96b0 | 222 | */ |
7ff9345f JA |
223 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
224 | if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) { | |
225 | *idx = BIOVEC_MAX_IDX; | |
226 | goto fallback; | |
227 | } | |
228 | } | |
229 | ||
1da177e4 LT |
230 | return bvl; |
231 | } | |
232 | ||
7ff9345f | 233 | void bio_free(struct bio *bio, struct bio_set *bs) |
1da177e4 | 234 | { |
bb799ca0 | 235 | void *p; |
1da177e4 | 236 | |
392ddc32 | 237 | if (bio_has_allocated_vec(bio)) |
bb799ca0 | 238 | bvec_free_bs(bs, bio->bi_io_vec, BIO_POOL_IDX(bio)); |
1da177e4 | 239 | |
7ba1ba12 | 240 | if (bio_integrity(bio)) |
7878cba9 | 241 | bio_integrity_free(bio, bs); |
7ba1ba12 | 242 | |
bb799ca0 JA |
243 | /* |
244 | * If we have front padding, adjust the bio pointer before freeing | |
245 | */ | |
246 | p = bio; | |
247 | if (bs->front_pad) | |
248 | p -= bs->front_pad; | |
249 | ||
250 | mempool_free(p, bs->bio_pool); | |
3676347a | 251 | } |
a112a71d | 252 | EXPORT_SYMBOL(bio_free); |
3676347a | 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 | 260 | } |
a112a71d | 261 | EXPORT_SYMBOL(bio_init); |
1da177e4 LT |
262 | |
263 | /** | |
264 | * bio_alloc_bioset - allocate a bio for I/O | |
265 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
266 | * @nr_iovecs: number of iovecs to pre-allocate | |
0a0d96b0 | 267 | * @bs: the bio_set to allocate from. If %NULL, just use kmalloc |
1da177e4 LT |
268 | * |
269 | * Description: | |
0a0d96b0 | 270 | * bio_alloc_bioset will first try its own mempool to satisfy the allocation. |
1da177e4 | 271 | * If %__GFP_WAIT is set then we will block on the internal pool waiting |
0a0d96b0 JA |
272 | * for a &struct bio to become free. If a %NULL @bs is passed in, we will |
273 | * fall back to just using @kmalloc to allocate the required memory. | |
1da177e4 | 274 | * |
af901ca1 | 275 | * Note that the caller must set ->bi_destructor on successful return |
bb799ca0 JA |
276 | * of a bio, to do the appropriate freeing of the bio once the reference |
277 | * count drops to zero. | |
1da177e4 | 278 | **/ |
dd0fc66f | 279 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) |
1da177e4 | 280 | { |
451a9ebf | 281 | unsigned long idx = BIO_POOL_NONE; |
34053979 | 282 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
283 | struct bio *bio; |
284 | void *p; | |
285 | ||
286 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
287 | if (unlikely(!p)) | |
288 | return NULL; | |
289 | bio = p + bs->front_pad; | |
1da177e4 | 290 | |
34053979 IM |
291 | bio_init(bio); |
292 | ||
293 | if (unlikely(!nr_iovecs)) | |
294 | goto out_set; | |
295 | ||
296 | if (nr_iovecs <= BIO_INLINE_VECS) { | |
297 | bvl = bio->bi_inline_vecs; | |
298 | nr_iovecs = BIO_INLINE_VECS; | |
299 | } else { | |
300 | bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs); | |
301 | if (unlikely(!bvl)) | |
302 | goto err_free; | |
303 | ||
304 | nr_iovecs = bvec_nr_vecs(idx); | |
1da177e4 | 305 | } |
451a9ebf | 306 | out_set: |
34053979 IM |
307 | bio->bi_flags |= idx << BIO_POOL_OFFSET; |
308 | bio->bi_max_vecs = nr_iovecs; | |
34053979 | 309 | bio->bi_io_vec = bvl; |
1da177e4 | 310 | return bio; |
34053979 IM |
311 | |
312 | err_free: | |
451a9ebf | 313 | mempool_free(p, bs->bio_pool); |
34053979 | 314 | return NULL; |
1da177e4 | 315 | } |
a112a71d | 316 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 317 | |
451a9ebf TH |
318 | static void bio_fs_destructor(struct bio *bio) |
319 | { | |
320 | bio_free(bio, fs_bio_set); | |
321 | } | |
322 | ||
323 | /** | |
324 | * bio_alloc - allocate a new bio, memory pool backed | |
325 | * @gfp_mask: allocation mask to use | |
326 | * @nr_iovecs: number of iovecs | |
327 | * | |
5f04eeb8 AB |
328 | * bio_alloc will allocate a bio and associated bio_vec array that can hold |
329 | * at least @nr_iovecs entries. Allocations will be done from the | |
330 | * fs_bio_set. Also see @bio_alloc_bioset and @bio_kmalloc. | |
331 | * | |
332 | * If %__GFP_WAIT is set, then bio_alloc will always be able to allocate | |
333 | * a bio. This is due to the mempool guarantees. To make this work, callers | |
334 | * must never allocate more than 1 bio at a time from this pool. Callers | |
335 | * that need to allocate more than 1 bio must always submit the previously | |
336 | * allocated bio for IO before attempting to allocate a new one. Failure to | |
337 | * do so can cause livelocks under memory pressure. | |
451a9ebf TH |
338 | * |
339 | * RETURNS: | |
340 | * Pointer to new bio on success, NULL on failure. | |
341 | */ | |
342 | struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs) | |
343 | { | |
344 | struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set); | |
345 | ||
346 | if (bio) | |
347 | bio->bi_destructor = bio_fs_destructor; | |
348 | ||
349 | return bio; | |
350 | } | |
a112a71d | 351 | EXPORT_SYMBOL(bio_alloc); |
451a9ebf TH |
352 | |
353 | static void bio_kmalloc_destructor(struct bio *bio) | |
354 | { | |
355 | if (bio_integrity(bio)) | |
7878cba9 | 356 | bio_integrity_free(bio, fs_bio_set); |
451a9ebf TH |
357 | kfree(bio); |
358 | } | |
359 | ||
86c824b9 | 360 | /** |
5f04eeb8 | 361 | * bio_kmalloc - allocate a bio for I/O using kmalloc() |
86c824b9 JA |
362 | * @gfp_mask: the GFP_ mask given to the slab allocator |
363 | * @nr_iovecs: number of iovecs to pre-allocate | |
364 | * | |
365 | * Description: | |
5f04eeb8 AB |
366 | * Allocate a new bio with @nr_iovecs bvecs. If @gfp_mask contains |
367 | * %__GFP_WAIT, the allocation is guaranteed to succeed. | |
86c824b9 JA |
368 | * |
369 | **/ | |
0a0d96b0 JA |
370 | struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs) |
371 | { | |
451a9ebf | 372 | struct bio *bio; |
0a0d96b0 | 373 | |
451a9ebf TH |
374 | bio = kmalloc(sizeof(struct bio) + nr_iovecs * sizeof(struct bio_vec), |
375 | gfp_mask); | |
376 | if (unlikely(!bio)) | |
377 | return NULL; | |
378 | ||
379 | bio_init(bio); | |
380 | bio->bi_flags |= BIO_POOL_NONE << BIO_POOL_OFFSET; | |
381 | bio->bi_max_vecs = nr_iovecs; | |
382 | bio->bi_io_vec = bio->bi_inline_vecs; | |
383 | bio->bi_destructor = bio_kmalloc_destructor; | |
0a0d96b0 JA |
384 | |
385 | return bio; | |
386 | } | |
a112a71d | 387 | EXPORT_SYMBOL(bio_kmalloc); |
0a0d96b0 | 388 | |
1da177e4 LT |
389 | void zero_fill_bio(struct bio *bio) |
390 | { | |
391 | unsigned long flags; | |
392 | struct bio_vec *bv; | |
393 | int i; | |
394 | ||
395 | bio_for_each_segment(bv, bio, i) { | |
396 | char *data = bvec_kmap_irq(bv, &flags); | |
397 | memset(data, 0, bv->bv_len); | |
398 | flush_dcache_page(bv->bv_page); | |
399 | bvec_kunmap_irq(data, &flags); | |
400 | } | |
401 | } | |
402 | EXPORT_SYMBOL(zero_fill_bio); | |
403 | ||
404 | /** | |
405 | * bio_put - release a reference to a bio | |
406 | * @bio: bio to release reference to | |
407 | * | |
408 | * Description: | |
409 | * Put a reference to a &struct bio, either one you have gotten with | |
ad0bf110 | 410 | * bio_alloc, bio_get or bio_clone. The last put of a bio will free it. |
1da177e4 LT |
411 | **/ |
412 | void bio_put(struct bio *bio) | |
413 | { | |
414 | BIO_BUG_ON(!atomic_read(&bio->bi_cnt)); | |
415 | ||
416 | /* | |
417 | * last put frees it | |
418 | */ | |
419 | if (atomic_dec_and_test(&bio->bi_cnt)) { | |
420 | bio->bi_next = NULL; | |
421 | bio->bi_destructor(bio); | |
422 | } | |
423 | } | |
a112a71d | 424 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 425 | |
165125e1 | 426 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
427 | { |
428 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
429 | blk_recount_segments(q, bio); | |
430 | ||
431 | return bio->bi_phys_segments; | |
432 | } | |
a112a71d | 433 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 434 | |
1da177e4 LT |
435 | /** |
436 | * __bio_clone - clone a bio | |
437 | * @bio: destination bio | |
438 | * @bio_src: bio to clone | |
439 | * | |
440 | * Clone a &bio. Caller will own the returned bio, but not | |
441 | * the actual data it points to. Reference count of returned | |
442 | * bio will be one. | |
443 | */ | |
858119e1 | 444 | void __bio_clone(struct bio *bio, struct bio *bio_src) |
1da177e4 | 445 | { |
e525e153 AM |
446 | memcpy(bio->bi_io_vec, bio_src->bi_io_vec, |
447 | bio_src->bi_max_vecs * sizeof(struct bio_vec)); | |
1da177e4 | 448 | |
5d84070e JA |
449 | /* |
450 | * most users will be overriding ->bi_bdev with a new target, | |
451 | * so we don't set nor calculate new physical/hw segment counts here | |
452 | */ | |
1da177e4 LT |
453 | bio->bi_sector = bio_src->bi_sector; |
454 | bio->bi_bdev = bio_src->bi_bdev; | |
455 | bio->bi_flags |= 1 << BIO_CLONED; | |
456 | bio->bi_rw = bio_src->bi_rw; | |
1da177e4 LT |
457 | bio->bi_vcnt = bio_src->bi_vcnt; |
458 | bio->bi_size = bio_src->bi_size; | |
a5453be4 | 459 | bio->bi_idx = bio_src->bi_idx; |
1da177e4 | 460 | } |
a112a71d | 461 | EXPORT_SYMBOL(__bio_clone); |
1da177e4 LT |
462 | |
463 | /** | |
464 | * bio_clone - clone a bio | |
465 | * @bio: bio to clone | |
466 | * @gfp_mask: allocation priority | |
467 | * | |
468 | * Like __bio_clone, only also allocates the returned bio | |
469 | */ | |
dd0fc66f | 470 | struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask) |
1da177e4 LT |
471 | { |
472 | struct bio *b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, fs_bio_set); | |
473 | ||
7ba1ba12 MP |
474 | if (!b) |
475 | return NULL; | |
476 | ||
477 | b->bi_destructor = bio_fs_destructor; | |
478 | __bio_clone(b, bio); | |
479 | ||
480 | if (bio_integrity(bio)) { | |
481 | int ret; | |
482 | ||
7878cba9 | 483 | ret = bio_integrity_clone(b, bio, gfp_mask, fs_bio_set); |
7ba1ba12 | 484 | |
059ea331 LZ |
485 | if (ret < 0) { |
486 | bio_put(b); | |
7ba1ba12 | 487 | return NULL; |
059ea331 | 488 | } |
3676347a | 489 | } |
1da177e4 LT |
490 | |
491 | return b; | |
492 | } | |
a112a71d | 493 | EXPORT_SYMBOL(bio_clone); |
1da177e4 LT |
494 | |
495 | /** | |
496 | * bio_get_nr_vecs - return approx number of vecs | |
497 | * @bdev: I/O target | |
498 | * | |
499 | * Return the approximate number of pages we can send to this target. | |
500 | * There's no guarantee that you will be able to fit this number of pages | |
501 | * into a bio, it does not account for dynamic restrictions that vary | |
502 | * on offset. | |
503 | */ | |
504 | int bio_get_nr_vecs(struct block_device *bdev) | |
505 | { | |
165125e1 | 506 | struct request_queue *q = bdev_get_queue(bdev); |
1da177e4 LT |
507 | int nr_pages; |
508 | ||
ae03bf63 MP |
509 | nr_pages = ((queue_max_sectors(q) << 9) + PAGE_SIZE - 1) >> PAGE_SHIFT; |
510 | if (nr_pages > queue_max_phys_segments(q)) | |
511 | nr_pages = queue_max_phys_segments(q); | |
512 | if (nr_pages > queue_max_hw_segments(q)) | |
513 | nr_pages = queue_max_hw_segments(q); | |
1da177e4 LT |
514 | |
515 | return nr_pages; | |
516 | } | |
a112a71d | 517 | EXPORT_SYMBOL(bio_get_nr_vecs); |
1da177e4 | 518 | |
165125e1 | 519 | static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page |
defd94b7 MC |
520 | *page, unsigned int len, unsigned int offset, |
521 | unsigned short max_sectors) | |
1da177e4 LT |
522 | { |
523 | int retried_segments = 0; | |
524 | struct bio_vec *bvec; | |
525 | ||
526 | /* | |
527 | * cloned bio must not modify vec list | |
528 | */ | |
529 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
530 | return 0; | |
531 | ||
80cfd548 | 532 | if (((bio->bi_size + len) >> 9) > max_sectors) |
1da177e4 LT |
533 | return 0; |
534 | ||
80cfd548 JA |
535 | /* |
536 | * For filesystems with a blocksize smaller than the pagesize | |
537 | * we will often be called with the same page as last time and | |
538 | * a consecutive offset. Optimize this special case. | |
539 | */ | |
540 | if (bio->bi_vcnt > 0) { | |
541 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
542 | ||
543 | if (page == prev->bv_page && | |
544 | offset == prev->bv_offset + prev->bv_len) { | |
1d616585 | 545 | unsigned int prev_bv_len = prev->bv_len; |
80cfd548 | 546 | prev->bv_len += len; |
cc371e66 AK |
547 | |
548 | if (q->merge_bvec_fn) { | |
549 | struct bvec_merge_data bvm = { | |
1d616585 DM |
550 | /* prev_bvec is already charged in |
551 | bi_size, discharge it in order to | |
552 | simulate merging updated prev_bvec | |
553 | as new bvec. */ | |
cc371e66 AK |
554 | .bi_bdev = bio->bi_bdev, |
555 | .bi_sector = bio->bi_sector, | |
1d616585 | 556 | .bi_size = bio->bi_size - prev_bv_len, |
cc371e66 AK |
557 | .bi_rw = bio->bi_rw, |
558 | }; | |
559 | ||
560 | if (q->merge_bvec_fn(q, &bvm, prev) < len) { | |
561 | prev->bv_len -= len; | |
562 | return 0; | |
563 | } | |
80cfd548 JA |
564 | } |
565 | ||
566 | goto done; | |
567 | } | |
568 | } | |
569 | ||
570 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
571 | return 0; |
572 | ||
573 | /* | |
574 | * we might lose a segment or two here, but rather that than | |
575 | * make this too complex. | |
576 | */ | |
577 | ||
ae03bf63 MP |
578 | while (bio->bi_phys_segments >= queue_max_phys_segments(q) |
579 | || bio->bi_phys_segments >= queue_max_hw_segments(q)) { | |
1da177e4 LT |
580 | |
581 | if (retried_segments) | |
582 | return 0; | |
583 | ||
584 | retried_segments = 1; | |
585 | blk_recount_segments(q, bio); | |
586 | } | |
587 | ||
588 | /* | |
589 | * setup the new entry, we might clear it again later if we | |
590 | * cannot add the page | |
591 | */ | |
592 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
593 | bvec->bv_page = page; | |
594 | bvec->bv_len = len; | |
595 | bvec->bv_offset = offset; | |
596 | ||
597 | /* | |
598 | * if queue has other restrictions (eg varying max sector size | |
599 | * depending on offset), it can specify a merge_bvec_fn in the | |
600 | * queue to get further control | |
601 | */ | |
602 | if (q->merge_bvec_fn) { | |
cc371e66 AK |
603 | struct bvec_merge_data bvm = { |
604 | .bi_bdev = bio->bi_bdev, | |
605 | .bi_sector = bio->bi_sector, | |
606 | .bi_size = bio->bi_size, | |
607 | .bi_rw = bio->bi_rw, | |
608 | }; | |
609 | ||
1da177e4 LT |
610 | /* |
611 | * merge_bvec_fn() returns number of bytes it can accept | |
612 | * at this offset | |
613 | */ | |
cc371e66 | 614 | if (q->merge_bvec_fn(q, &bvm, bvec) < len) { |
1da177e4 LT |
615 | bvec->bv_page = NULL; |
616 | bvec->bv_len = 0; | |
617 | bvec->bv_offset = 0; | |
618 | return 0; | |
619 | } | |
620 | } | |
621 | ||
622 | /* If we may be able to merge these biovecs, force a recount */ | |
b8b3e16c | 623 | if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
1da177e4 LT |
624 | bio->bi_flags &= ~(1 << BIO_SEG_VALID); |
625 | ||
626 | bio->bi_vcnt++; | |
627 | bio->bi_phys_segments++; | |
80cfd548 | 628 | done: |
1da177e4 LT |
629 | bio->bi_size += len; |
630 | return len; | |
631 | } | |
632 | ||
6e68af66 MC |
633 | /** |
634 | * bio_add_pc_page - attempt to add page to bio | |
fddfdeaf | 635 | * @q: the target queue |
6e68af66 MC |
636 | * @bio: destination bio |
637 | * @page: page to add | |
638 | * @len: vec entry length | |
639 | * @offset: vec entry offset | |
640 | * | |
641 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
642 | * number of reasons, such as the bio being full or target block | |
643 | * device limitations. The target block device must allow bio's | |
644 | * smaller than PAGE_SIZE, so it is always possible to add a single | |
645 | * page to an empty bio. This should only be used by REQ_PC bios. | |
646 | */ | |
165125e1 | 647 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page, |
6e68af66 MC |
648 | unsigned int len, unsigned int offset) |
649 | { | |
ae03bf63 MP |
650 | return __bio_add_page(q, bio, page, len, offset, |
651 | queue_max_hw_sectors(q)); | |
6e68af66 | 652 | } |
a112a71d | 653 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 654 | |
1da177e4 LT |
655 | /** |
656 | * bio_add_page - attempt to add page to bio | |
657 | * @bio: destination bio | |
658 | * @page: page to add | |
659 | * @len: vec entry length | |
660 | * @offset: vec entry offset | |
661 | * | |
662 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
663 | * number of reasons, such as the bio being full or target block | |
664 | * device limitations. The target block device must allow bio's | |
665 | * smaller than PAGE_SIZE, so it is always possible to add a single | |
666 | * page to an empty bio. | |
667 | */ | |
668 | int bio_add_page(struct bio *bio, struct page *page, unsigned int len, | |
669 | unsigned int offset) | |
670 | { | |
defd94b7 | 671 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
ae03bf63 | 672 | return __bio_add_page(q, bio, page, len, offset, queue_max_sectors(q)); |
1da177e4 | 673 | } |
a112a71d | 674 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 LT |
675 | |
676 | struct bio_map_data { | |
677 | struct bio_vec *iovecs; | |
c5dec1c3 | 678 | struct sg_iovec *sgvecs; |
152e283f FT |
679 | int nr_sgvecs; |
680 | int is_our_pages; | |
1da177e4 LT |
681 | }; |
682 | ||
c5dec1c3 | 683 | static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio, |
152e283f FT |
684 | struct sg_iovec *iov, int iov_count, |
685 | int is_our_pages) | |
1da177e4 LT |
686 | { |
687 | memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt); | |
c5dec1c3 FT |
688 | memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count); |
689 | bmd->nr_sgvecs = iov_count; | |
152e283f | 690 | bmd->is_our_pages = is_our_pages; |
1da177e4 LT |
691 | bio->bi_private = bmd; |
692 | } | |
693 | ||
694 | static void bio_free_map_data(struct bio_map_data *bmd) | |
695 | { | |
696 | kfree(bmd->iovecs); | |
c5dec1c3 | 697 | kfree(bmd->sgvecs); |
1da177e4 LT |
698 | kfree(bmd); |
699 | } | |
700 | ||
76029ff3 FT |
701 | static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count, |
702 | gfp_t gfp_mask) | |
1da177e4 | 703 | { |
76029ff3 | 704 | struct bio_map_data *bmd = kmalloc(sizeof(*bmd), gfp_mask); |
1da177e4 LT |
705 | |
706 | if (!bmd) | |
707 | return NULL; | |
708 | ||
76029ff3 | 709 | bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask); |
c5dec1c3 FT |
710 | if (!bmd->iovecs) { |
711 | kfree(bmd); | |
712 | return NULL; | |
713 | } | |
714 | ||
76029ff3 | 715 | bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask); |
c5dec1c3 | 716 | if (bmd->sgvecs) |
1da177e4 LT |
717 | return bmd; |
718 | ||
c5dec1c3 | 719 | kfree(bmd->iovecs); |
1da177e4 LT |
720 | kfree(bmd); |
721 | return NULL; | |
722 | } | |
723 | ||
aefcc28a | 724 | static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs, |
ecb554a8 FT |
725 | struct sg_iovec *iov, int iov_count, |
726 | int to_user, int from_user, int do_free_page) | |
c5dec1c3 FT |
727 | { |
728 | int ret = 0, i; | |
729 | struct bio_vec *bvec; | |
730 | int iov_idx = 0; | |
731 | unsigned int iov_off = 0; | |
c5dec1c3 FT |
732 | |
733 | __bio_for_each_segment(bvec, bio, i, 0) { | |
734 | char *bv_addr = page_address(bvec->bv_page); | |
aefcc28a | 735 | unsigned int bv_len = iovecs[i].bv_len; |
c5dec1c3 FT |
736 | |
737 | while (bv_len && iov_idx < iov_count) { | |
738 | unsigned int bytes; | |
0e0c6212 | 739 | char __user *iov_addr; |
c5dec1c3 FT |
740 | |
741 | bytes = min_t(unsigned int, | |
742 | iov[iov_idx].iov_len - iov_off, bv_len); | |
743 | iov_addr = iov[iov_idx].iov_base + iov_off; | |
744 | ||
745 | if (!ret) { | |
ecb554a8 | 746 | if (to_user) |
c5dec1c3 FT |
747 | ret = copy_to_user(iov_addr, bv_addr, |
748 | bytes); | |
749 | ||
ecb554a8 FT |
750 | if (from_user) |
751 | ret = copy_from_user(bv_addr, iov_addr, | |
752 | bytes); | |
753 | ||
c5dec1c3 FT |
754 | if (ret) |
755 | ret = -EFAULT; | |
756 | } | |
757 | ||
758 | bv_len -= bytes; | |
759 | bv_addr += bytes; | |
760 | iov_addr += bytes; | |
761 | iov_off += bytes; | |
762 | ||
763 | if (iov[iov_idx].iov_len == iov_off) { | |
764 | iov_idx++; | |
765 | iov_off = 0; | |
766 | } | |
767 | } | |
768 | ||
152e283f | 769 | if (do_free_page) |
c5dec1c3 FT |
770 | __free_page(bvec->bv_page); |
771 | } | |
772 | ||
773 | return ret; | |
774 | } | |
775 | ||
1da177e4 LT |
776 | /** |
777 | * bio_uncopy_user - finish previously mapped bio | |
778 | * @bio: bio being terminated | |
779 | * | |
780 | * Free pages allocated from bio_copy_user() and write back data | |
781 | * to user space in case of a read. | |
782 | */ | |
783 | int bio_uncopy_user(struct bio *bio) | |
784 | { | |
785 | struct bio_map_data *bmd = bio->bi_private; | |
81882766 | 786 | int ret = 0; |
1da177e4 | 787 | |
81882766 FT |
788 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) |
789 | ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs, | |
ecb554a8 FT |
790 | bmd->nr_sgvecs, bio_data_dir(bio) == READ, |
791 | 0, bmd->is_our_pages); | |
1da177e4 LT |
792 | bio_free_map_data(bmd); |
793 | bio_put(bio); | |
794 | return ret; | |
795 | } | |
a112a71d | 796 | EXPORT_SYMBOL(bio_uncopy_user); |
1da177e4 LT |
797 | |
798 | /** | |
c5dec1c3 | 799 | * bio_copy_user_iov - copy user data to bio |
1da177e4 | 800 | * @q: destination block queue |
152e283f | 801 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
802 | * @iov: the iovec. |
803 | * @iov_count: number of elements in the iovec | |
1da177e4 | 804 | * @write_to_vm: bool indicating writing to pages or not |
a3bce90e | 805 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
806 | * |
807 | * Prepares and returns a bio for indirect user io, bouncing data | |
808 | * to/from kernel pages as necessary. Must be paired with | |
809 | * call bio_uncopy_user() on io completion. | |
810 | */ | |
152e283f FT |
811 | struct bio *bio_copy_user_iov(struct request_queue *q, |
812 | struct rq_map_data *map_data, | |
813 | struct sg_iovec *iov, int iov_count, | |
814 | int write_to_vm, gfp_t gfp_mask) | |
1da177e4 | 815 | { |
1da177e4 LT |
816 | struct bio_map_data *bmd; |
817 | struct bio_vec *bvec; | |
818 | struct page *page; | |
819 | struct bio *bio; | |
820 | int i, ret; | |
c5dec1c3 FT |
821 | int nr_pages = 0; |
822 | unsigned int len = 0; | |
56c451f4 | 823 | unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0; |
1da177e4 | 824 | |
c5dec1c3 FT |
825 | for (i = 0; i < iov_count; i++) { |
826 | unsigned long uaddr; | |
827 | unsigned long end; | |
828 | unsigned long start; | |
829 | ||
830 | uaddr = (unsigned long)iov[i].iov_base; | |
831 | end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
832 | start = uaddr >> PAGE_SHIFT; | |
833 | ||
834 | nr_pages += end - start; | |
835 | len += iov[i].iov_len; | |
836 | } | |
837 | ||
69838727 FT |
838 | if (offset) |
839 | nr_pages++; | |
840 | ||
a3bce90e | 841 | bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask); |
1da177e4 LT |
842 | if (!bmd) |
843 | return ERR_PTR(-ENOMEM); | |
844 | ||
1da177e4 | 845 | ret = -ENOMEM; |
a9e9dc24 | 846 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
847 | if (!bio) |
848 | goto out_bmd; | |
849 | ||
850 | bio->bi_rw |= (!write_to_vm << BIO_RW); | |
851 | ||
852 | ret = 0; | |
56c451f4 FT |
853 | |
854 | if (map_data) { | |
e623ddb4 | 855 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
856 | i = map_data->offset / PAGE_SIZE; |
857 | } | |
1da177e4 | 858 | while (len) { |
e623ddb4 | 859 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 860 | |
56c451f4 FT |
861 | bytes -= offset; |
862 | ||
1da177e4 LT |
863 | if (bytes > len) |
864 | bytes = len; | |
865 | ||
152e283f | 866 | if (map_data) { |
e623ddb4 | 867 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
868 | ret = -ENOMEM; |
869 | break; | |
870 | } | |
e623ddb4 FT |
871 | |
872 | page = map_data->pages[i / nr_pages]; | |
873 | page += (i % nr_pages); | |
874 | ||
875 | i++; | |
876 | } else { | |
152e283f | 877 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
878 | if (!page) { |
879 | ret = -ENOMEM; | |
880 | break; | |
881 | } | |
1da177e4 LT |
882 | } |
883 | ||
56c451f4 | 884 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 885 | break; |
1da177e4 LT |
886 | |
887 | len -= bytes; | |
56c451f4 | 888 | offset = 0; |
1da177e4 LT |
889 | } |
890 | ||
891 | if (ret) | |
892 | goto cleanup; | |
893 | ||
894 | /* | |
895 | * success | |
896 | */ | |
ecb554a8 FT |
897 | if ((!write_to_vm && (!map_data || !map_data->null_mapped)) || |
898 | (map_data && map_data->from_user)) { | |
899 | ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 1, 0); | |
c5dec1c3 FT |
900 | if (ret) |
901 | goto cleanup; | |
1da177e4 LT |
902 | } |
903 | ||
152e283f | 904 | bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1); |
1da177e4 LT |
905 | return bio; |
906 | cleanup: | |
152e283f FT |
907 | if (!map_data) |
908 | bio_for_each_segment(bvec, bio, i) | |
909 | __free_page(bvec->bv_page); | |
1da177e4 LT |
910 | |
911 | bio_put(bio); | |
912 | out_bmd: | |
913 | bio_free_map_data(bmd); | |
914 | return ERR_PTR(ret); | |
915 | } | |
916 | ||
c5dec1c3 FT |
917 | /** |
918 | * bio_copy_user - copy user data to bio | |
919 | * @q: destination block queue | |
152e283f | 920 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
921 | * @uaddr: start of user address |
922 | * @len: length in bytes | |
923 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 924 | * @gfp_mask: memory allocation flags |
c5dec1c3 FT |
925 | * |
926 | * Prepares and returns a bio for indirect user io, bouncing data | |
927 | * to/from kernel pages as necessary. Must be paired with | |
928 | * call bio_uncopy_user() on io completion. | |
929 | */ | |
152e283f FT |
930 | struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data, |
931 | unsigned long uaddr, unsigned int len, | |
932 | int write_to_vm, gfp_t gfp_mask) | |
c5dec1c3 FT |
933 | { |
934 | struct sg_iovec iov; | |
935 | ||
936 | iov.iov_base = (void __user *)uaddr; | |
937 | iov.iov_len = len; | |
938 | ||
152e283f | 939 | return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask); |
c5dec1c3 | 940 | } |
a112a71d | 941 | EXPORT_SYMBOL(bio_copy_user); |
c5dec1c3 | 942 | |
165125e1 | 943 | static struct bio *__bio_map_user_iov(struct request_queue *q, |
f1970baf JB |
944 | struct block_device *bdev, |
945 | struct sg_iovec *iov, int iov_count, | |
a3bce90e | 946 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 | 947 | { |
f1970baf JB |
948 | int i, j; |
949 | int nr_pages = 0; | |
1da177e4 LT |
950 | struct page **pages; |
951 | struct bio *bio; | |
f1970baf JB |
952 | int cur_page = 0; |
953 | int ret, offset; | |
1da177e4 | 954 | |
f1970baf JB |
955 | for (i = 0; i < iov_count; i++) { |
956 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
957 | unsigned long len = iov[i].iov_len; | |
958 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
959 | unsigned long start = uaddr >> PAGE_SHIFT; | |
960 | ||
961 | nr_pages += end - start; | |
962 | /* | |
ad2d7225 | 963 | * buffer must be aligned to at least hardsector size for now |
f1970baf | 964 | */ |
ad2d7225 | 965 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
966 | return ERR_PTR(-EINVAL); |
967 | } | |
968 | ||
969 | if (!nr_pages) | |
1da177e4 LT |
970 | return ERR_PTR(-EINVAL); |
971 | ||
a9e9dc24 | 972 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
973 | if (!bio) |
974 | return ERR_PTR(-ENOMEM); | |
975 | ||
976 | ret = -ENOMEM; | |
a3bce90e | 977 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
978 | if (!pages) |
979 | goto out; | |
980 | ||
f1970baf JB |
981 | for (i = 0; i < iov_count; i++) { |
982 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
983 | unsigned long len = iov[i].iov_len; | |
984 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
985 | unsigned long start = uaddr >> PAGE_SHIFT; | |
986 | const int local_nr_pages = end - start; | |
987 | const int page_limit = cur_page + local_nr_pages; | |
988 | ||
f5dd33c4 NP |
989 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
990 | write_to_vm, &pages[cur_page]); | |
99172157 JA |
991 | if (ret < local_nr_pages) { |
992 | ret = -EFAULT; | |
f1970baf | 993 | goto out_unmap; |
99172157 | 994 | } |
f1970baf JB |
995 | |
996 | offset = uaddr & ~PAGE_MASK; | |
997 | for (j = cur_page; j < page_limit; j++) { | |
998 | unsigned int bytes = PAGE_SIZE - offset; | |
999 | ||
1000 | if (len <= 0) | |
1001 | break; | |
1002 | ||
1003 | if (bytes > len) | |
1004 | bytes = len; | |
1005 | ||
1006 | /* | |
1007 | * sorry... | |
1008 | */ | |
defd94b7 MC |
1009 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1010 | bytes) | |
f1970baf JB |
1011 | break; |
1012 | ||
1013 | len -= bytes; | |
1014 | offset = 0; | |
1015 | } | |
1da177e4 | 1016 | |
f1970baf | 1017 | cur_page = j; |
1da177e4 | 1018 | /* |
f1970baf | 1019 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1020 | */ |
f1970baf JB |
1021 | while (j < page_limit) |
1022 | page_cache_release(pages[j++]); | |
1da177e4 LT |
1023 | } |
1024 | ||
1da177e4 LT |
1025 | kfree(pages); |
1026 | ||
1027 | /* | |
1028 | * set data direction, and check if mapped pages need bouncing | |
1029 | */ | |
1030 | if (!write_to_vm) | |
1031 | bio->bi_rw |= (1 << BIO_RW); | |
1032 | ||
f1970baf | 1033 | bio->bi_bdev = bdev; |
1da177e4 LT |
1034 | bio->bi_flags |= (1 << BIO_USER_MAPPED); |
1035 | return bio; | |
f1970baf JB |
1036 | |
1037 | out_unmap: | |
1038 | for (i = 0; i < nr_pages; i++) { | |
1039 | if(!pages[i]) | |
1040 | break; | |
1041 | page_cache_release(pages[i]); | |
1042 | } | |
1043 | out: | |
1da177e4 LT |
1044 | kfree(pages); |
1045 | bio_put(bio); | |
1046 | return ERR_PTR(ret); | |
1047 | } | |
1048 | ||
1049 | /** | |
1050 | * bio_map_user - map user address into bio | |
165125e1 | 1051 | * @q: the struct request_queue for the bio |
1da177e4 LT |
1052 | * @bdev: destination block device |
1053 | * @uaddr: start of user address | |
1054 | * @len: length in bytes | |
1055 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1056 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
1057 | * |
1058 | * Map the user space address into a bio suitable for io to a block | |
1059 | * device. Returns an error pointer in case of error. | |
1060 | */ | |
165125e1 | 1061 | struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev, |
a3bce90e FT |
1062 | unsigned long uaddr, unsigned int len, int write_to_vm, |
1063 | gfp_t gfp_mask) | |
f1970baf JB |
1064 | { |
1065 | struct sg_iovec iov; | |
1066 | ||
3f70353e | 1067 | iov.iov_base = (void __user *)uaddr; |
f1970baf JB |
1068 | iov.iov_len = len; |
1069 | ||
a3bce90e | 1070 | return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask); |
f1970baf | 1071 | } |
a112a71d | 1072 | EXPORT_SYMBOL(bio_map_user); |
f1970baf JB |
1073 | |
1074 | /** | |
1075 | * bio_map_user_iov - map user sg_iovec table into bio | |
165125e1 | 1076 | * @q: the struct request_queue for the bio |
f1970baf JB |
1077 | * @bdev: destination block device |
1078 | * @iov: the iovec. | |
1079 | * @iov_count: number of elements in the iovec | |
1080 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1081 | * @gfp_mask: memory allocation flags |
f1970baf JB |
1082 | * |
1083 | * Map the user space address into a bio suitable for io to a block | |
1084 | * device. Returns an error pointer in case of error. | |
1085 | */ | |
165125e1 | 1086 | struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev, |
f1970baf | 1087 | struct sg_iovec *iov, int iov_count, |
a3bce90e | 1088 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 LT |
1089 | { |
1090 | struct bio *bio; | |
1091 | ||
a3bce90e FT |
1092 | bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm, |
1093 | gfp_mask); | |
1da177e4 LT |
1094 | if (IS_ERR(bio)) |
1095 | return bio; | |
1096 | ||
1097 | /* | |
1098 | * subtle -- if __bio_map_user() ended up bouncing a bio, | |
1099 | * it would normally disappear when its bi_end_io is run. | |
1100 | * however, we need it for the unmap, so grab an extra | |
1101 | * reference to it | |
1102 | */ | |
1103 | bio_get(bio); | |
1104 | ||
0e75f906 | 1105 | return bio; |
1da177e4 LT |
1106 | } |
1107 | ||
1108 | static void __bio_unmap_user(struct bio *bio) | |
1109 | { | |
1110 | struct bio_vec *bvec; | |
1111 | int i; | |
1112 | ||
1113 | /* | |
1114 | * make sure we dirty pages we wrote to | |
1115 | */ | |
1116 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1117 | if (bio_data_dir(bio) == READ) | |
1118 | set_page_dirty_lock(bvec->bv_page); | |
1119 | ||
1120 | page_cache_release(bvec->bv_page); | |
1121 | } | |
1122 | ||
1123 | bio_put(bio); | |
1124 | } | |
1125 | ||
1126 | /** | |
1127 | * bio_unmap_user - unmap a bio | |
1128 | * @bio: the bio being unmapped | |
1129 | * | |
1130 | * Unmap a bio previously mapped by bio_map_user(). Must be called with | |
1131 | * a process context. | |
1132 | * | |
1133 | * bio_unmap_user() may sleep. | |
1134 | */ | |
1135 | void bio_unmap_user(struct bio *bio) | |
1136 | { | |
1137 | __bio_unmap_user(bio); | |
1138 | bio_put(bio); | |
1139 | } | |
a112a71d | 1140 | EXPORT_SYMBOL(bio_unmap_user); |
1da177e4 | 1141 | |
6712ecf8 | 1142 | static void bio_map_kern_endio(struct bio *bio, int err) |
b823825e | 1143 | { |
b823825e | 1144 | bio_put(bio); |
b823825e JA |
1145 | } |
1146 | ||
165125e1 | 1147 | static struct bio *__bio_map_kern(struct request_queue *q, void *data, |
27496a8c | 1148 | unsigned int len, gfp_t gfp_mask) |
df46b9a4 MC |
1149 | { |
1150 | unsigned long kaddr = (unsigned long)data; | |
1151 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1152 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1153 | const int nr_pages = end - start; | |
1154 | int offset, i; | |
1155 | struct bio *bio; | |
1156 | ||
a9e9dc24 | 1157 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1158 | if (!bio) |
1159 | return ERR_PTR(-ENOMEM); | |
1160 | ||
1161 | offset = offset_in_page(kaddr); | |
1162 | for (i = 0; i < nr_pages; i++) { | |
1163 | unsigned int bytes = PAGE_SIZE - offset; | |
1164 | ||
1165 | if (len <= 0) | |
1166 | break; | |
1167 | ||
1168 | if (bytes > len) | |
1169 | bytes = len; | |
1170 | ||
defd94b7 MC |
1171 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
1172 | offset) < bytes) | |
df46b9a4 MC |
1173 | break; |
1174 | ||
1175 | data += bytes; | |
1176 | len -= bytes; | |
1177 | offset = 0; | |
1178 | } | |
1179 | ||
b823825e | 1180 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1181 | return bio; |
1182 | } | |
1183 | ||
1184 | /** | |
1185 | * bio_map_kern - map kernel address into bio | |
165125e1 | 1186 | * @q: the struct request_queue for the bio |
df46b9a4 MC |
1187 | * @data: pointer to buffer to map |
1188 | * @len: length in bytes | |
1189 | * @gfp_mask: allocation flags for bio allocation | |
1190 | * | |
1191 | * Map the kernel address into a bio suitable for io to a block | |
1192 | * device. Returns an error pointer in case of error. | |
1193 | */ | |
165125e1 | 1194 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
27496a8c | 1195 | gfp_t gfp_mask) |
df46b9a4 MC |
1196 | { |
1197 | struct bio *bio; | |
1198 | ||
1199 | bio = __bio_map_kern(q, data, len, gfp_mask); | |
1200 | if (IS_ERR(bio)) | |
1201 | return bio; | |
1202 | ||
1203 | if (bio->bi_size == len) | |
1204 | return bio; | |
1205 | ||
1206 | /* | |
1207 | * Don't support partial mappings. | |
1208 | */ | |
1209 | bio_put(bio); | |
1210 | return ERR_PTR(-EINVAL); | |
1211 | } | |
a112a71d | 1212 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1213 | |
68154e90 FT |
1214 | static void bio_copy_kern_endio(struct bio *bio, int err) |
1215 | { | |
1216 | struct bio_vec *bvec; | |
1217 | const int read = bio_data_dir(bio) == READ; | |
76029ff3 | 1218 | struct bio_map_data *bmd = bio->bi_private; |
68154e90 | 1219 | int i; |
76029ff3 | 1220 | char *p = bmd->sgvecs[0].iov_base; |
68154e90 FT |
1221 | |
1222 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1223 | char *addr = page_address(bvec->bv_page); | |
76029ff3 | 1224 | int len = bmd->iovecs[i].bv_len; |
68154e90 | 1225 | |
4fc981ef | 1226 | if (read) |
76029ff3 | 1227 | memcpy(p, addr, len); |
68154e90 FT |
1228 | |
1229 | __free_page(bvec->bv_page); | |
76029ff3 | 1230 | p += len; |
68154e90 FT |
1231 | } |
1232 | ||
76029ff3 | 1233 | bio_free_map_data(bmd); |
68154e90 FT |
1234 | bio_put(bio); |
1235 | } | |
1236 | ||
1237 | /** | |
1238 | * bio_copy_kern - copy kernel address into bio | |
1239 | * @q: the struct request_queue for the bio | |
1240 | * @data: pointer to buffer to copy | |
1241 | * @len: length in bytes | |
1242 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1243 | * @reading: data direction is READ |
68154e90 FT |
1244 | * |
1245 | * copy the kernel address into a bio suitable for io to a block | |
1246 | * device. Returns an error pointer in case of error. | |
1247 | */ | |
1248 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1249 | gfp_t gfp_mask, int reading) | |
1250 | { | |
68154e90 FT |
1251 | struct bio *bio; |
1252 | struct bio_vec *bvec; | |
4d8ab62e | 1253 | int i; |
68154e90 | 1254 | |
4d8ab62e FT |
1255 | bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask); |
1256 | if (IS_ERR(bio)) | |
1257 | return bio; | |
68154e90 FT |
1258 | |
1259 | if (!reading) { | |
1260 | void *p = data; | |
1261 | ||
1262 | bio_for_each_segment(bvec, bio, i) { | |
1263 | char *addr = page_address(bvec->bv_page); | |
1264 | ||
1265 | memcpy(addr, p, bvec->bv_len); | |
1266 | p += bvec->bv_len; | |
1267 | } | |
1268 | } | |
1269 | ||
68154e90 | 1270 | bio->bi_end_io = bio_copy_kern_endio; |
76029ff3 | 1271 | |
68154e90 | 1272 | return bio; |
68154e90 | 1273 | } |
a112a71d | 1274 | EXPORT_SYMBOL(bio_copy_kern); |
68154e90 | 1275 | |
1da177e4 LT |
1276 | /* |
1277 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1278 | * for performing direct-IO in BIOs. | |
1279 | * | |
1280 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1281 | * because the required locks are not interrupt-safe. So what we can do is to | |
1282 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1283 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1284 | * in process context. | |
1285 | * | |
1286 | * We special-case compound pages here: normally this means reads into hugetlb | |
1287 | * pages. The logic in here doesn't really work right for compound pages | |
1288 | * because the VM does not uniformly chase down the head page in all cases. | |
1289 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1290 | * handle them at all. So we skip compound pages here at an early stage. | |
1291 | * | |
1292 | * Note that this code is very hard to test under normal circumstances because | |
1293 | * direct-io pins the pages with get_user_pages(). This makes | |
1294 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
1295 | * But other code (eg, pdflush) could clean the pages if they are mapped | |
1296 | * pagecache. | |
1297 | * | |
1298 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1299 | * deferred bio dirtying paths. | |
1300 | */ | |
1301 | ||
1302 | /* | |
1303 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1304 | */ | |
1305 | void bio_set_pages_dirty(struct bio *bio) | |
1306 | { | |
1307 | struct bio_vec *bvec = bio->bi_io_vec; | |
1308 | int i; | |
1309 | ||
1310 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1311 | struct page *page = bvec[i].bv_page; | |
1312 | ||
1313 | if (page && !PageCompound(page)) | |
1314 | set_page_dirty_lock(page); | |
1315 | } | |
1316 | } | |
1317 | ||
86b6c7a7 | 1318 | static void bio_release_pages(struct bio *bio) |
1da177e4 LT |
1319 | { |
1320 | struct bio_vec *bvec = bio->bi_io_vec; | |
1321 | int i; | |
1322 | ||
1323 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1324 | struct page *page = bvec[i].bv_page; | |
1325 | ||
1326 | if (page) | |
1327 | put_page(page); | |
1328 | } | |
1329 | } | |
1330 | ||
1331 | /* | |
1332 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1333 | * If they are, then fine. If, however, some pages are clean then they must | |
1334 | * have been written out during the direct-IO read. So we take another ref on | |
1335 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1336 | * | |
1337 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
1338 | * here on. It will run one page_cache_release() against each page and will | |
1339 | * run one bio_put() against the BIO. | |
1340 | */ | |
1341 | ||
65f27f38 | 1342 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1343 | |
65f27f38 | 1344 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1345 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1346 | static struct bio *bio_dirty_list; | |
1347 | ||
1348 | /* | |
1349 | * This runs in process context | |
1350 | */ | |
65f27f38 | 1351 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1352 | { |
1353 | unsigned long flags; | |
1354 | struct bio *bio; | |
1355 | ||
1356 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1357 | bio = bio_dirty_list; | |
1358 | bio_dirty_list = NULL; | |
1359 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1360 | ||
1361 | while (bio) { | |
1362 | struct bio *next = bio->bi_private; | |
1363 | ||
1364 | bio_set_pages_dirty(bio); | |
1365 | bio_release_pages(bio); | |
1366 | bio_put(bio); | |
1367 | bio = next; | |
1368 | } | |
1369 | } | |
1370 | ||
1371 | void bio_check_pages_dirty(struct bio *bio) | |
1372 | { | |
1373 | struct bio_vec *bvec = bio->bi_io_vec; | |
1374 | int nr_clean_pages = 0; | |
1375 | int i; | |
1376 | ||
1377 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1378 | struct page *page = bvec[i].bv_page; | |
1379 | ||
1380 | if (PageDirty(page) || PageCompound(page)) { | |
1381 | page_cache_release(page); | |
1382 | bvec[i].bv_page = NULL; | |
1383 | } else { | |
1384 | nr_clean_pages++; | |
1385 | } | |
1386 | } | |
1387 | ||
1388 | if (nr_clean_pages) { | |
1389 | unsigned long flags; | |
1390 | ||
1391 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1392 | bio->bi_private = bio_dirty_list; | |
1393 | bio_dirty_list = bio; | |
1394 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1395 | schedule_work(&bio_dirty_work); | |
1396 | } else { | |
1397 | bio_put(bio); | |
1398 | } | |
1399 | } | |
1400 | ||
2d4dc890 IL |
1401 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1402 | void bio_flush_dcache_pages(struct bio *bi) | |
1403 | { | |
1404 | int i; | |
1405 | struct bio_vec *bvec; | |
1406 | ||
1407 | bio_for_each_segment(bvec, bi, i) | |
1408 | flush_dcache_page(bvec->bv_page); | |
1409 | } | |
1410 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1411 | #endif | |
1412 | ||
1da177e4 LT |
1413 | /** |
1414 | * bio_endio - end I/O on a bio | |
1415 | * @bio: bio | |
1da177e4 LT |
1416 | * @error: error, if any |
1417 | * | |
1418 | * Description: | |
6712ecf8 | 1419 | * bio_endio() will end I/O on the whole bio. bio_endio() is the |
5bb23a68 N |
1420 | * preferred way to end I/O on a bio, it takes care of clearing |
1421 | * BIO_UPTODATE on error. @error is 0 on success, and and one of the | |
1422 | * established -Exxxx (-EIO, for instance) error values in case | |
1423 | * something went wrong. Noone should call bi_end_io() directly on a | |
1424 | * bio unless they own it and thus know that it has an end_io | |
1425 | * function. | |
1da177e4 | 1426 | **/ |
6712ecf8 | 1427 | void bio_endio(struct bio *bio, int error) |
1da177e4 LT |
1428 | { |
1429 | if (error) | |
1430 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
9cc54d40 N |
1431 | else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) |
1432 | error = -EIO; | |
1da177e4 | 1433 | |
5bb23a68 | 1434 | if (bio->bi_end_io) |
6712ecf8 | 1435 | bio->bi_end_io(bio, error); |
1da177e4 | 1436 | } |
a112a71d | 1437 | EXPORT_SYMBOL(bio_endio); |
1da177e4 LT |
1438 | |
1439 | void bio_pair_release(struct bio_pair *bp) | |
1440 | { | |
1441 | if (atomic_dec_and_test(&bp->cnt)) { | |
1442 | struct bio *master = bp->bio1.bi_private; | |
1443 | ||
6712ecf8 | 1444 | bio_endio(master, bp->error); |
1da177e4 LT |
1445 | mempool_free(bp, bp->bio2.bi_private); |
1446 | } | |
1447 | } | |
a112a71d | 1448 | EXPORT_SYMBOL(bio_pair_release); |
1da177e4 | 1449 | |
6712ecf8 | 1450 | static void bio_pair_end_1(struct bio *bi, int err) |
1da177e4 LT |
1451 | { |
1452 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio1); | |
1453 | ||
1454 | if (err) | |
1455 | bp->error = err; | |
1456 | ||
1da177e4 | 1457 | bio_pair_release(bp); |
1da177e4 LT |
1458 | } |
1459 | ||
6712ecf8 | 1460 | static void bio_pair_end_2(struct bio *bi, int err) |
1da177e4 LT |
1461 | { |
1462 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio2); | |
1463 | ||
1464 | if (err) | |
1465 | bp->error = err; | |
1466 | ||
1da177e4 | 1467 | bio_pair_release(bp); |
1da177e4 LT |
1468 | } |
1469 | ||
1470 | /* | |
c7eee1b8 | 1471 | * split a bio - only worry about a bio with a single page in its iovec |
1da177e4 | 1472 | */ |
6feef531 | 1473 | struct bio_pair *bio_split(struct bio *bi, int first_sectors) |
1da177e4 | 1474 | { |
6feef531 | 1475 | struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO); |
1da177e4 LT |
1476 | |
1477 | if (!bp) | |
1478 | return bp; | |
1479 | ||
5f3ea37c | 1480 | trace_block_split(bdev_get_queue(bi->bi_bdev), bi, |
2056a782 JA |
1481 | bi->bi_sector + first_sectors); |
1482 | ||
1da177e4 LT |
1483 | BUG_ON(bi->bi_vcnt != 1); |
1484 | BUG_ON(bi->bi_idx != 0); | |
1485 | atomic_set(&bp->cnt, 3); | |
1486 | bp->error = 0; | |
1487 | bp->bio1 = *bi; | |
1488 | bp->bio2 = *bi; | |
1489 | bp->bio2.bi_sector += first_sectors; | |
1490 | bp->bio2.bi_size -= first_sectors << 9; | |
1491 | bp->bio1.bi_size = first_sectors << 9; | |
1492 | ||
1493 | bp->bv1 = bi->bi_io_vec[0]; | |
1494 | bp->bv2 = bi->bi_io_vec[0]; | |
1495 | bp->bv2.bv_offset += first_sectors << 9; | |
1496 | bp->bv2.bv_len -= first_sectors << 9; | |
1497 | bp->bv1.bv_len = first_sectors << 9; | |
1498 | ||
1499 | bp->bio1.bi_io_vec = &bp->bv1; | |
1500 | bp->bio2.bi_io_vec = &bp->bv2; | |
1501 | ||
a2eb0c10 N |
1502 | bp->bio1.bi_max_vecs = 1; |
1503 | bp->bio2.bi_max_vecs = 1; | |
1504 | ||
1da177e4 LT |
1505 | bp->bio1.bi_end_io = bio_pair_end_1; |
1506 | bp->bio2.bi_end_io = bio_pair_end_2; | |
1507 | ||
1508 | bp->bio1.bi_private = bi; | |
6feef531 | 1509 | bp->bio2.bi_private = bio_split_pool; |
1da177e4 | 1510 | |
7ba1ba12 MP |
1511 | if (bio_integrity(bi)) |
1512 | bio_integrity_split(bi, bp, first_sectors); | |
1513 | ||
1da177e4 LT |
1514 | return bp; |
1515 | } | |
a112a71d | 1516 | EXPORT_SYMBOL(bio_split); |
1da177e4 | 1517 | |
ad3316bf MP |
1518 | /** |
1519 | * bio_sector_offset - Find hardware sector offset in bio | |
1520 | * @bio: bio to inspect | |
1521 | * @index: bio_vec index | |
1522 | * @offset: offset in bv_page | |
1523 | * | |
1524 | * Return the number of hardware sectors between beginning of bio | |
1525 | * and an end point indicated by a bio_vec index and an offset | |
1526 | * within that vector's page. | |
1527 | */ | |
1528 | sector_t bio_sector_offset(struct bio *bio, unsigned short index, | |
1529 | unsigned int offset) | |
1530 | { | |
e1defc4f | 1531 | unsigned int sector_sz; |
ad3316bf MP |
1532 | struct bio_vec *bv; |
1533 | sector_t sectors; | |
1534 | int i; | |
1535 | ||
e1defc4f | 1536 | sector_sz = queue_logical_block_size(bio->bi_bdev->bd_disk->queue); |
ad3316bf MP |
1537 | sectors = 0; |
1538 | ||
1539 | if (index >= bio->bi_idx) | |
1540 | index = bio->bi_vcnt - 1; | |
1541 | ||
1542 | __bio_for_each_segment(bv, bio, i, 0) { | |
1543 | if (i == index) { | |
1544 | if (offset > bv->bv_offset) | |
1545 | sectors += (offset - bv->bv_offset) / sector_sz; | |
1546 | break; | |
1547 | } | |
1548 | ||
1549 | sectors += bv->bv_len / sector_sz; | |
1550 | } | |
1551 | ||
1552 | return sectors; | |
1553 | } | |
1554 | EXPORT_SYMBOL(bio_sector_offset); | |
1da177e4 LT |
1555 | |
1556 | /* | |
1557 | * create memory pools for biovec's in a bio_set. | |
1558 | * use the global biovec slabs created for general use. | |
1559 | */ | |
5972511b | 1560 | static int biovec_create_pools(struct bio_set *bs, int pool_entries) |
1da177e4 | 1561 | { |
7ff9345f | 1562 | struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX; |
1da177e4 | 1563 | |
7ff9345f JA |
1564 | bs->bvec_pool = mempool_create_slab_pool(pool_entries, bp->slab); |
1565 | if (!bs->bvec_pool) | |
1566 | return -ENOMEM; | |
1da177e4 | 1567 | |
1da177e4 LT |
1568 | return 0; |
1569 | } | |
1570 | ||
1571 | static void biovec_free_pools(struct bio_set *bs) | |
1572 | { | |
7ff9345f | 1573 | mempool_destroy(bs->bvec_pool); |
1da177e4 LT |
1574 | } |
1575 | ||
1576 | void bioset_free(struct bio_set *bs) | |
1577 | { | |
1578 | if (bs->bio_pool) | |
1579 | mempool_destroy(bs->bio_pool); | |
1580 | ||
7878cba9 | 1581 | bioset_integrity_free(bs); |
1da177e4 | 1582 | biovec_free_pools(bs); |
bb799ca0 | 1583 | bio_put_slab(bs); |
1da177e4 LT |
1584 | |
1585 | kfree(bs); | |
1586 | } | |
a112a71d | 1587 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1588 | |
bb799ca0 JA |
1589 | /** |
1590 | * bioset_create - Create a bio_set | |
1591 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1592 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1593 | * | |
1594 | * Description: | |
1595 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1596 | * to ask for a number of bytes to be allocated in front of the bio. | |
1597 | * Front pad allocation is useful for embedding the bio inside | |
1598 | * another structure, to avoid allocating extra data to go with the bio. | |
1599 | * Note that the bio must be embedded at the END of that structure always, | |
1600 | * or things will break badly. | |
1601 | */ | |
1602 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) | |
1da177e4 | 1603 | { |
392ddc32 | 1604 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1605 | struct bio_set *bs; |
1da177e4 | 1606 | |
1b434498 | 1607 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1608 | if (!bs) |
1609 | return NULL; | |
1610 | ||
bb799ca0 | 1611 | bs->front_pad = front_pad; |
1b434498 | 1612 | |
392ddc32 | 1613 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1614 | if (!bs->bio_slab) { |
1615 | kfree(bs); | |
1616 | return NULL; | |
1617 | } | |
1618 | ||
1619 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
1620 | if (!bs->bio_pool) |
1621 | goto bad; | |
1622 | ||
7878cba9 MP |
1623 | if (bioset_integrity_create(bs, pool_size)) |
1624 | goto bad; | |
1625 | ||
bb799ca0 | 1626 | if (!biovec_create_pools(bs, pool_size)) |
1da177e4 LT |
1627 | return bs; |
1628 | ||
1629 | bad: | |
1630 | bioset_free(bs); | |
1631 | return NULL; | |
1632 | } | |
a112a71d | 1633 | EXPORT_SYMBOL(bioset_create); |
1da177e4 LT |
1634 | |
1635 | static void __init biovec_init_slabs(void) | |
1636 | { | |
1637 | int i; | |
1638 | ||
1639 | for (i = 0; i < BIOVEC_NR_POOLS; i++) { | |
1640 | int size; | |
1641 | struct biovec_slab *bvs = bvec_slabs + i; | |
1642 | ||
a7fcd37c JA |
1643 | #ifndef CONFIG_BLK_DEV_INTEGRITY |
1644 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { | |
1645 | bvs->slab = NULL; | |
1646 | continue; | |
1647 | } | |
1648 | #endif | |
1649 | ||
1da177e4 LT |
1650 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
1651 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 1652 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1653 | } |
1654 | } | |
1655 | ||
1656 | static int __init init_bio(void) | |
1657 | { | |
bb799ca0 JA |
1658 | bio_slab_max = 2; |
1659 | bio_slab_nr = 0; | |
1660 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
1661 | if (!bio_slabs) | |
1662 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 1663 | |
7878cba9 | 1664 | bio_integrity_init(); |
1da177e4 LT |
1665 | biovec_init_slabs(); |
1666 | ||
bb799ca0 | 1667 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
1da177e4 LT |
1668 | if (!fs_bio_set) |
1669 | panic("bio: can't allocate bios\n"); | |
1670 | ||
0eaae62a MD |
1671 | bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES, |
1672 | sizeof(struct bio_pair)); | |
1da177e4 LT |
1673 | if (!bio_split_pool) |
1674 | panic("bio: can't create split pool\n"); | |
1675 | ||
1676 | return 0; | |
1677 | } | |
1da177e4 | 1678 | subsys_initcall(init_bio); |