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> | |
852c788f | 22 | #include <linux/iocontext.h> |
1da177e4 LT |
23 | #include <linux/slab.h> |
24 | #include <linux/init.h> | |
25 | #include <linux/kernel.h> | |
630d9c47 | 26 | #include <linux/export.h> |
1da177e4 LT |
27 | #include <linux/mempool.h> |
28 | #include <linux/workqueue.h> | |
852c788f | 29 | #include <linux/cgroup.h> |
f1970baf | 30 | #include <scsi/sg.h> /* for struct sg_iovec */ |
1da177e4 | 31 | |
55782138 | 32 | #include <trace/events/block.h> |
0bfc2455 | 33 | |
392ddc32 JA |
34 | /* |
35 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
36 | * itself, to shrink a bio data allocation from two mempool calls to one | |
37 | */ | |
38 | #define BIO_INLINE_VECS 4 | |
39 | ||
6feef531 | 40 | static mempool_t *bio_split_pool __read_mostly; |
1da177e4 | 41 | |
1da177e4 LT |
42 | /* |
43 | * if you change this list, also change bvec_alloc or things will | |
44 | * break badly! cannot be bigger than what you can fit into an | |
45 | * unsigned short | |
46 | */ | |
1da177e4 | 47 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
df677140 | 48 | static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = { |
1da177e4 LT |
49 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
50 | }; | |
51 | #undef BV | |
52 | ||
1da177e4 LT |
53 | /* |
54 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
55 | * IO code that does not need private memory pools. | |
56 | */ | |
51d654e1 | 57 | struct bio_set *fs_bio_set; |
3f86a82a | 58 | EXPORT_SYMBOL(fs_bio_set); |
1da177e4 | 59 | |
bb799ca0 JA |
60 | /* |
61 | * Our slab pool management | |
62 | */ | |
63 | struct bio_slab { | |
64 | struct kmem_cache *slab; | |
65 | unsigned int slab_ref; | |
66 | unsigned int slab_size; | |
67 | char name[8]; | |
68 | }; | |
69 | static DEFINE_MUTEX(bio_slab_lock); | |
70 | static struct bio_slab *bio_slabs; | |
71 | static unsigned int bio_slab_nr, bio_slab_max; | |
72 | ||
73 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
74 | { | |
75 | unsigned int sz = sizeof(struct bio) + extra_size; | |
76 | struct kmem_cache *slab = NULL; | |
389d7b26 | 77 | struct bio_slab *bslab, *new_bio_slabs; |
bb799ca0 JA |
78 | unsigned int i, entry = -1; |
79 | ||
80 | mutex_lock(&bio_slab_lock); | |
81 | ||
82 | i = 0; | |
83 | while (i < bio_slab_nr) { | |
f06f135d | 84 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
85 | |
86 | if (!bslab->slab && entry == -1) | |
87 | entry = i; | |
88 | else if (bslab->slab_size == sz) { | |
89 | slab = bslab->slab; | |
90 | bslab->slab_ref++; | |
91 | break; | |
92 | } | |
93 | i++; | |
94 | } | |
95 | ||
96 | if (slab) | |
97 | goto out_unlock; | |
98 | ||
99 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
100 | bio_slab_max <<= 1; | |
389d7b26 AK |
101 | new_bio_slabs = krealloc(bio_slabs, |
102 | bio_slab_max * sizeof(struct bio_slab), | |
103 | GFP_KERNEL); | |
104 | if (!new_bio_slabs) | |
bb799ca0 | 105 | goto out_unlock; |
389d7b26 | 106 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
107 | } |
108 | if (entry == -1) | |
109 | entry = bio_slab_nr++; | |
110 | ||
111 | bslab = &bio_slabs[entry]; | |
112 | ||
113 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
114 | slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL); | |
115 | if (!slab) | |
116 | goto out_unlock; | |
117 | ||
80cdc6da | 118 | printk(KERN_INFO "bio: create slab <%s> at %d\n", bslab->name, entry); |
bb799ca0 JA |
119 | bslab->slab = slab; |
120 | bslab->slab_ref = 1; | |
121 | bslab->slab_size = sz; | |
122 | out_unlock: | |
123 | mutex_unlock(&bio_slab_lock); | |
124 | return slab; | |
125 | } | |
126 | ||
127 | static void bio_put_slab(struct bio_set *bs) | |
128 | { | |
129 | struct bio_slab *bslab = NULL; | |
130 | unsigned int i; | |
131 | ||
132 | mutex_lock(&bio_slab_lock); | |
133 | ||
134 | for (i = 0; i < bio_slab_nr; i++) { | |
135 | if (bs->bio_slab == bio_slabs[i].slab) { | |
136 | bslab = &bio_slabs[i]; | |
137 | break; | |
138 | } | |
139 | } | |
140 | ||
141 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
142 | goto out; | |
143 | ||
144 | WARN_ON(!bslab->slab_ref); | |
145 | ||
146 | if (--bslab->slab_ref) | |
147 | goto out; | |
148 | ||
149 | kmem_cache_destroy(bslab->slab); | |
150 | bslab->slab = NULL; | |
151 | ||
152 | out: | |
153 | mutex_unlock(&bio_slab_lock); | |
154 | } | |
155 | ||
7ba1ba12 MP |
156 | unsigned int bvec_nr_vecs(unsigned short idx) |
157 | { | |
158 | return bvec_slabs[idx].nr_vecs; | |
159 | } | |
160 | ||
bb799ca0 JA |
161 | void bvec_free_bs(struct bio_set *bs, struct bio_vec *bv, unsigned int idx) |
162 | { | |
163 | BIO_BUG_ON(idx >= BIOVEC_NR_POOLS); | |
164 | ||
165 | if (idx == BIOVEC_MAX_IDX) | |
166 | mempool_free(bv, bs->bvec_pool); | |
167 | else { | |
168 | struct biovec_slab *bvs = bvec_slabs + idx; | |
169 | ||
170 | kmem_cache_free(bvs->slab, bv); | |
171 | } | |
172 | } | |
173 | ||
7ff9345f JA |
174 | struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, |
175 | struct bio_set *bs) | |
1da177e4 LT |
176 | { |
177 | struct bio_vec *bvl; | |
1da177e4 | 178 | |
7ff9345f JA |
179 | /* |
180 | * see comment near bvec_array define! | |
181 | */ | |
182 | switch (nr) { | |
183 | case 1: | |
184 | *idx = 0; | |
185 | break; | |
186 | case 2 ... 4: | |
187 | *idx = 1; | |
188 | break; | |
189 | case 5 ... 16: | |
190 | *idx = 2; | |
191 | break; | |
192 | case 17 ... 64: | |
193 | *idx = 3; | |
194 | break; | |
195 | case 65 ... 128: | |
196 | *idx = 4; | |
197 | break; | |
198 | case 129 ... BIO_MAX_PAGES: | |
199 | *idx = 5; | |
200 | break; | |
201 | default: | |
202 | return NULL; | |
203 | } | |
204 | ||
205 | /* | |
206 | * idx now points to the pool we want to allocate from. only the | |
207 | * 1-vec entry pool is mempool backed. | |
208 | */ | |
209 | if (*idx == BIOVEC_MAX_IDX) { | |
210 | fallback: | |
211 | bvl = mempool_alloc(bs->bvec_pool, gfp_mask); | |
212 | } else { | |
213 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
214 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO); | |
215 | ||
0a0d96b0 | 216 | /* |
7ff9345f JA |
217 | * Make this allocation restricted and don't dump info on |
218 | * allocation failures, since we'll fallback to the mempool | |
219 | * in case of failure. | |
0a0d96b0 | 220 | */ |
7ff9345f | 221 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 222 | |
0a0d96b0 | 223 | /* |
7ff9345f JA |
224 | * Try a slab allocation. If this fails and __GFP_WAIT |
225 | * is set, retry with the 1-entry mempool | |
0a0d96b0 | 226 | */ |
7ff9345f JA |
227 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
228 | if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) { | |
229 | *idx = BIOVEC_MAX_IDX; | |
230 | goto fallback; | |
231 | } | |
232 | } | |
233 | ||
1da177e4 LT |
234 | return bvl; |
235 | } | |
236 | ||
4254bba1 | 237 | static void __bio_free(struct bio *bio) |
1da177e4 | 238 | { |
4254bba1 | 239 | bio_disassociate_task(bio); |
1da177e4 | 240 | |
7ba1ba12 | 241 | if (bio_integrity(bio)) |
1e2a410f | 242 | bio_integrity_free(bio); |
4254bba1 | 243 | } |
7ba1ba12 | 244 | |
4254bba1 KO |
245 | static void bio_free(struct bio *bio) |
246 | { | |
247 | struct bio_set *bs = bio->bi_pool; | |
248 | void *p; | |
249 | ||
250 | __bio_free(bio); | |
251 | ||
252 | if (bs) { | |
253 | if (bio_has_allocated_vec(bio)) | |
254 | bvec_free_bs(bs, bio->bi_io_vec, BIO_POOL_IDX(bio)); | |
255 | ||
256 | /* | |
257 | * If we have front padding, adjust the bio pointer before freeing | |
258 | */ | |
259 | p = bio; | |
bb799ca0 JA |
260 | p -= bs->front_pad; |
261 | ||
4254bba1 KO |
262 | mempool_free(p, bs->bio_pool); |
263 | } else { | |
264 | /* Bio was allocated by bio_kmalloc() */ | |
265 | kfree(bio); | |
266 | } | |
3676347a PO |
267 | } |
268 | ||
858119e1 | 269 | void bio_init(struct bio *bio) |
1da177e4 | 270 | { |
2b94de55 | 271 | memset(bio, 0, sizeof(*bio)); |
1da177e4 | 272 | bio->bi_flags = 1 << BIO_UPTODATE; |
1da177e4 | 273 | atomic_set(&bio->bi_cnt, 1); |
1da177e4 | 274 | } |
a112a71d | 275 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 276 | |
f44b48c7 KO |
277 | /** |
278 | * bio_reset - reinitialize a bio | |
279 | * @bio: bio to reset | |
280 | * | |
281 | * Description: | |
282 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
283 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
284 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
285 | * comment in struct bio. | |
286 | */ | |
287 | void bio_reset(struct bio *bio) | |
288 | { | |
289 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
290 | ||
4254bba1 | 291 | __bio_free(bio); |
f44b48c7 KO |
292 | |
293 | memset(bio, 0, BIO_RESET_BYTES); | |
294 | bio->bi_flags = flags|(1 << BIO_UPTODATE); | |
295 | } | |
296 | EXPORT_SYMBOL(bio_reset); | |
297 | ||
1da177e4 LT |
298 | /** |
299 | * bio_alloc_bioset - allocate a bio for I/O | |
300 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
301 | * @nr_iovecs: number of iovecs to pre-allocate | |
db18efac | 302 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
303 | * |
304 | * Description: | |
3f86a82a KO |
305 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
306 | * backed by the @bs's mempool. | |
307 | * | |
308 | * When @bs is not NULL, if %__GFP_WAIT is set then bio_alloc will always be | |
309 | * able to allocate a bio. This is due to the mempool guarantees. To make this | |
310 | * work, callers must never allocate more than 1 bio at a time from this pool. | |
311 | * Callers that need to allocate more than 1 bio must always submit the | |
312 | * previously allocated bio for IO before attempting to allocate a new one. | |
313 | * Failure to do so can cause deadlocks under memory pressure. | |
314 | * | |
315 | * RETURNS: | |
316 | * Pointer to new bio on success, NULL on failure. | |
317 | */ | |
dd0fc66f | 318 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) |
1da177e4 | 319 | { |
3f86a82a KO |
320 | unsigned front_pad; |
321 | unsigned inline_vecs; | |
451a9ebf | 322 | unsigned long idx = BIO_POOL_NONE; |
34053979 | 323 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
324 | struct bio *bio; |
325 | void *p; | |
326 | ||
3f86a82a KO |
327 | if (!bs) { |
328 | if (nr_iovecs > UIO_MAXIOV) | |
329 | return NULL; | |
330 | ||
331 | p = kmalloc(sizeof(struct bio) + | |
332 | nr_iovecs * sizeof(struct bio_vec), | |
333 | gfp_mask); | |
334 | front_pad = 0; | |
335 | inline_vecs = nr_iovecs; | |
336 | } else { | |
337 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
338 | front_pad = bs->front_pad; | |
339 | inline_vecs = BIO_INLINE_VECS; | |
340 | } | |
341 | ||
451a9ebf TH |
342 | if (unlikely(!p)) |
343 | return NULL; | |
1da177e4 | 344 | |
3f86a82a | 345 | bio = p + front_pad; |
34053979 IM |
346 | bio_init(bio); |
347 | ||
3f86a82a | 348 | if (nr_iovecs > inline_vecs) { |
34053979 IM |
349 | bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs); |
350 | if (unlikely(!bvl)) | |
351 | goto err_free; | |
3f86a82a KO |
352 | } else if (nr_iovecs) { |
353 | bvl = bio->bi_inline_vecs; | |
1da177e4 | 354 | } |
3f86a82a KO |
355 | |
356 | bio->bi_pool = bs; | |
34053979 IM |
357 | bio->bi_flags |= idx << BIO_POOL_OFFSET; |
358 | bio->bi_max_vecs = nr_iovecs; | |
34053979 | 359 | bio->bi_io_vec = bvl; |
1da177e4 | 360 | return bio; |
34053979 IM |
361 | |
362 | err_free: | |
451a9ebf | 363 | mempool_free(p, bs->bio_pool); |
34053979 | 364 | return NULL; |
1da177e4 | 365 | } |
a112a71d | 366 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 367 | |
1da177e4 LT |
368 | void zero_fill_bio(struct bio *bio) |
369 | { | |
370 | unsigned long flags; | |
371 | struct bio_vec *bv; | |
372 | int i; | |
373 | ||
374 | bio_for_each_segment(bv, bio, i) { | |
375 | char *data = bvec_kmap_irq(bv, &flags); | |
376 | memset(data, 0, bv->bv_len); | |
377 | flush_dcache_page(bv->bv_page); | |
378 | bvec_kunmap_irq(data, &flags); | |
379 | } | |
380 | } | |
381 | EXPORT_SYMBOL(zero_fill_bio); | |
382 | ||
383 | /** | |
384 | * bio_put - release a reference to a bio | |
385 | * @bio: bio to release reference to | |
386 | * | |
387 | * Description: | |
388 | * Put a reference to a &struct bio, either one you have gotten with | |
ad0bf110 | 389 | * bio_alloc, bio_get or bio_clone. The last put of a bio will free it. |
1da177e4 LT |
390 | **/ |
391 | void bio_put(struct bio *bio) | |
392 | { | |
393 | BIO_BUG_ON(!atomic_read(&bio->bi_cnt)); | |
394 | ||
395 | /* | |
396 | * last put frees it | |
397 | */ | |
4254bba1 KO |
398 | if (atomic_dec_and_test(&bio->bi_cnt)) |
399 | bio_free(bio); | |
1da177e4 | 400 | } |
a112a71d | 401 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 402 | |
165125e1 | 403 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
404 | { |
405 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
406 | blk_recount_segments(q, bio); | |
407 | ||
408 | return bio->bi_phys_segments; | |
409 | } | |
a112a71d | 410 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 411 | |
1da177e4 LT |
412 | /** |
413 | * __bio_clone - clone a bio | |
414 | * @bio: destination bio | |
415 | * @bio_src: bio to clone | |
416 | * | |
417 | * Clone a &bio. Caller will own the returned bio, but not | |
418 | * the actual data it points to. Reference count of returned | |
419 | * bio will be one. | |
420 | */ | |
858119e1 | 421 | void __bio_clone(struct bio *bio, struct bio *bio_src) |
1da177e4 | 422 | { |
e525e153 AM |
423 | memcpy(bio->bi_io_vec, bio_src->bi_io_vec, |
424 | bio_src->bi_max_vecs * sizeof(struct bio_vec)); | |
1da177e4 | 425 | |
5d84070e JA |
426 | /* |
427 | * most users will be overriding ->bi_bdev with a new target, | |
428 | * so we don't set nor calculate new physical/hw segment counts here | |
429 | */ | |
1da177e4 LT |
430 | bio->bi_sector = bio_src->bi_sector; |
431 | bio->bi_bdev = bio_src->bi_bdev; | |
432 | bio->bi_flags |= 1 << BIO_CLONED; | |
433 | bio->bi_rw = bio_src->bi_rw; | |
1da177e4 LT |
434 | bio->bi_vcnt = bio_src->bi_vcnt; |
435 | bio->bi_size = bio_src->bi_size; | |
a5453be4 | 436 | bio->bi_idx = bio_src->bi_idx; |
1da177e4 | 437 | } |
a112a71d | 438 | EXPORT_SYMBOL(__bio_clone); |
1da177e4 LT |
439 | |
440 | /** | |
bf800ef1 | 441 | * bio_clone_bioset - clone a bio |
1da177e4 LT |
442 | * @bio: bio to clone |
443 | * @gfp_mask: allocation priority | |
bf800ef1 | 444 | * @bs: bio_set to allocate from |
1da177e4 LT |
445 | * |
446 | * Like __bio_clone, only also allocates the returned bio | |
447 | */ | |
bf800ef1 KO |
448 | struct bio *bio_clone_bioset(struct bio *bio, gfp_t gfp_mask, |
449 | struct bio_set *bs) | |
1da177e4 | 450 | { |
bf800ef1 | 451 | struct bio *b; |
1da177e4 | 452 | |
bf800ef1 | 453 | b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, bs); |
7ba1ba12 MP |
454 | if (!b) |
455 | return NULL; | |
456 | ||
7ba1ba12 MP |
457 | __bio_clone(b, bio); |
458 | ||
459 | if (bio_integrity(bio)) { | |
460 | int ret; | |
461 | ||
1e2a410f | 462 | ret = bio_integrity_clone(b, bio, gfp_mask); |
7ba1ba12 | 463 | |
059ea331 LZ |
464 | if (ret < 0) { |
465 | bio_put(b); | |
7ba1ba12 | 466 | return NULL; |
059ea331 | 467 | } |
3676347a | 468 | } |
1da177e4 LT |
469 | |
470 | return b; | |
471 | } | |
bf800ef1 | 472 | EXPORT_SYMBOL(bio_clone_bioset); |
1da177e4 LT |
473 | |
474 | /** | |
475 | * bio_get_nr_vecs - return approx number of vecs | |
476 | * @bdev: I/O target | |
477 | * | |
478 | * Return the approximate number of pages we can send to this target. | |
479 | * There's no guarantee that you will be able to fit this number of pages | |
480 | * into a bio, it does not account for dynamic restrictions that vary | |
481 | * on offset. | |
482 | */ | |
483 | int bio_get_nr_vecs(struct block_device *bdev) | |
484 | { | |
165125e1 | 485 | struct request_queue *q = bdev_get_queue(bdev); |
f908ee94 BS |
486 | int nr_pages; |
487 | ||
488 | nr_pages = min_t(unsigned, | |
5abebfdd KO |
489 | queue_max_segments(q), |
490 | queue_max_sectors(q) / (PAGE_SIZE >> 9) + 1); | |
f908ee94 BS |
491 | |
492 | return min_t(unsigned, nr_pages, BIO_MAX_PAGES); | |
493 | ||
1da177e4 | 494 | } |
a112a71d | 495 | EXPORT_SYMBOL(bio_get_nr_vecs); |
1da177e4 | 496 | |
165125e1 | 497 | static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page |
defd94b7 MC |
498 | *page, unsigned int len, unsigned int offset, |
499 | unsigned short max_sectors) | |
1da177e4 LT |
500 | { |
501 | int retried_segments = 0; | |
502 | struct bio_vec *bvec; | |
503 | ||
504 | /* | |
505 | * cloned bio must not modify vec list | |
506 | */ | |
507 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
508 | return 0; | |
509 | ||
80cfd548 | 510 | if (((bio->bi_size + len) >> 9) > max_sectors) |
1da177e4 LT |
511 | return 0; |
512 | ||
80cfd548 JA |
513 | /* |
514 | * For filesystems with a blocksize smaller than the pagesize | |
515 | * we will often be called with the same page as last time and | |
516 | * a consecutive offset. Optimize this special case. | |
517 | */ | |
518 | if (bio->bi_vcnt > 0) { | |
519 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
520 | ||
521 | if (page == prev->bv_page && | |
522 | offset == prev->bv_offset + prev->bv_len) { | |
1d616585 | 523 | unsigned int prev_bv_len = prev->bv_len; |
80cfd548 | 524 | prev->bv_len += len; |
cc371e66 AK |
525 | |
526 | if (q->merge_bvec_fn) { | |
527 | struct bvec_merge_data bvm = { | |
1d616585 DM |
528 | /* prev_bvec is already charged in |
529 | bi_size, discharge it in order to | |
530 | simulate merging updated prev_bvec | |
531 | as new bvec. */ | |
cc371e66 AK |
532 | .bi_bdev = bio->bi_bdev, |
533 | .bi_sector = bio->bi_sector, | |
1d616585 | 534 | .bi_size = bio->bi_size - prev_bv_len, |
cc371e66 AK |
535 | .bi_rw = bio->bi_rw, |
536 | }; | |
537 | ||
8bf8c376 | 538 | if (q->merge_bvec_fn(q, &bvm, prev) < prev->bv_len) { |
cc371e66 AK |
539 | prev->bv_len -= len; |
540 | return 0; | |
541 | } | |
80cfd548 JA |
542 | } |
543 | ||
544 | goto done; | |
545 | } | |
546 | } | |
547 | ||
548 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
549 | return 0; |
550 | ||
551 | /* | |
552 | * we might lose a segment or two here, but rather that than | |
553 | * make this too complex. | |
554 | */ | |
555 | ||
8a78362c | 556 | while (bio->bi_phys_segments >= queue_max_segments(q)) { |
1da177e4 LT |
557 | |
558 | if (retried_segments) | |
559 | return 0; | |
560 | ||
561 | retried_segments = 1; | |
562 | blk_recount_segments(q, bio); | |
563 | } | |
564 | ||
565 | /* | |
566 | * setup the new entry, we might clear it again later if we | |
567 | * cannot add the page | |
568 | */ | |
569 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
570 | bvec->bv_page = page; | |
571 | bvec->bv_len = len; | |
572 | bvec->bv_offset = offset; | |
573 | ||
574 | /* | |
575 | * if queue has other restrictions (eg varying max sector size | |
576 | * depending on offset), it can specify a merge_bvec_fn in the | |
577 | * queue to get further control | |
578 | */ | |
579 | if (q->merge_bvec_fn) { | |
cc371e66 AK |
580 | struct bvec_merge_data bvm = { |
581 | .bi_bdev = bio->bi_bdev, | |
582 | .bi_sector = bio->bi_sector, | |
583 | .bi_size = bio->bi_size, | |
584 | .bi_rw = bio->bi_rw, | |
585 | }; | |
586 | ||
1da177e4 LT |
587 | /* |
588 | * merge_bvec_fn() returns number of bytes it can accept | |
589 | * at this offset | |
590 | */ | |
8bf8c376 | 591 | if (q->merge_bvec_fn(q, &bvm, bvec) < bvec->bv_len) { |
1da177e4 LT |
592 | bvec->bv_page = NULL; |
593 | bvec->bv_len = 0; | |
594 | bvec->bv_offset = 0; | |
595 | return 0; | |
596 | } | |
597 | } | |
598 | ||
599 | /* If we may be able to merge these biovecs, force a recount */ | |
b8b3e16c | 600 | if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
1da177e4 LT |
601 | bio->bi_flags &= ~(1 << BIO_SEG_VALID); |
602 | ||
603 | bio->bi_vcnt++; | |
604 | bio->bi_phys_segments++; | |
80cfd548 | 605 | done: |
1da177e4 LT |
606 | bio->bi_size += len; |
607 | return len; | |
608 | } | |
609 | ||
6e68af66 MC |
610 | /** |
611 | * bio_add_pc_page - attempt to add page to bio | |
fddfdeaf | 612 | * @q: the target queue |
6e68af66 MC |
613 | * @bio: destination bio |
614 | * @page: page to add | |
615 | * @len: vec entry length | |
616 | * @offset: vec entry offset | |
617 | * | |
618 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
c6428084 AG |
619 | * number of reasons, such as the bio being full or target block device |
620 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
621 | * so it is always possible to add a single page to an empty bio. | |
622 | * | |
623 | * This should only be used by REQ_PC bios. | |
6e68af66 | 624 | */ |
165125e1 | 625 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page, |
6e68af66 MC |
626 | unsigned int len, unsigned int offset) |
627 | { | |
ae03bf63 MP |
628 | return __bio_add_page(q, bio, page, len, offset, |
629 | queue_max_hw_sectors(q)); | |
6e68af66 | 630 | } |
a112a71d | 631 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 632 | |
1da177e4 LT |
633 | /** |
634 | * bio_add_page - attempt to add page to bio | |
635 | * @bio: destination bio | |
636 | * @page: page to add | |
637 | * @len: vec entry length | |
638 | * @offset: vec entry offset | |
639 | * | |
640 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
c6428084 AG |
641 | * number of reasons, such as the bio being full or target block device |
642 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
643 | * so it is always possible to add a single page to an empty bio. | |
1da177e4 LT |
644 | */ |
645 | int bio_add_page(struct bio *bio, struct page *page, unsigned int len, | |
646 | unsigned int offset) | |
647 | { | |
defd94b7 | 648 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
ae03bf63 | 649 | return __bio_add_page(q, bio, page, len, offset, queue_max_sectors(q)); |
1da177e4 | 650 | } |
a112a71d | 651 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 LT |
652 | |
653 | struct bio_map_data { | |
654 | struct bio_vec *iovecs; | |
c5dec1c3 | 655 | struct sg_iovec *sgvecs; |
152e283f FT |
656 | int nr_sgvecs; |
657 | int is_our_pages; | |
1da177e4 LT |
658 | }; |
659 | ||
c5dec1c3 | 660 | static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio, |
152e283f FT |
661 | struct sg_iovec *iov, int iov_count, |
662 | int is_our_pages) | |
1da177e4 LT |
663 | { |
664 | memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt); | |
c5dec1c3 FT |
665 | memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count); |
666 | bmd->nr_sgvecs = iov_count; | |
152e283f | 667 | bmd->is_our_pages = is_our_pages; |
1da177e4 LT |
668 | bio->bi_private = bmd; |
669 | } | |
670 | ||
671 | static void bio_free_map_data(struct bio_map_data *bmd) | |
672 | { | |
673 | kfree(bmd->iovecs); | |
c5dec1c3 | 674 | kfree(bmd->sgvecs); |
1da177e4 LT |
675 | kfree(bmd); |
676 | } | |
677 | ||
121f0994 DC |
678 | static struct bio_map_data *bio_alloc_map_data(int nr_segs, |
679 | unsigned int iov_count, | |
76029ff3 | 680 | gfp_t gfp_mask) |
1da177e4 | 681 | { |
f3f63c1c JA |
682 | struct bio_map_data *bmd; |
683 | ||
684 | if (iov_count > UIO_MAXIOV) | |
685 | return NULL; | |
1da177e4 | 686 | |
f3f63c1c | 687 | bmd = kmalloc(sizeof(*bmd), gfp_mask); |
1da177e4 LT |
688 | if (!bmd) |
689 | return NULL; | |
690 | ||
76029ff3 | 691 | bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask); |
c5dec1c3 FT |
692 | if (!bmd->iovecs) { |
693 | kfree(bmd); | |
694 | return NULL; | |
695 | } | |
696 | ||
76029ff3 | 697 | bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask); |
c5dec1c3 | 698 | if (bmd->sgvecs) |
1da177e4 LT |
699 | return bmd; |
700 | ||
c5dec1c3 | 701 | kfree(bmd->iovecs); |
1da177e4 LT |
702 | kfree(bmd); |
703 | return NULL; | |
704 | } | |
705 | ||
aefcc28a | 706 | static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs, |
ecb554a8 FT |
707 | struct sg_iovec *iov, int iov_count, |
708 | int to_user, int from_user, int do_free_page) | |
c5dec1c3 FT |
709 | { |
710 | int ret = 0, i; | |
711 | struct bio_vec *bvec; | |
712 | int iov_idx = 0; | |
713 | unsigned int iov_off = 0; | |
c5dec1c3 FT |
714 | |
715 | __bio_for_each_segment(bvec, bio, i, 0) { | |
716 | char *bv_addr = page_address(bvec->bv_page); | |
aefcc28a | 717 | unsigned int bv_len = iovecs[i].bv_len; |
c5dec1c3 FT |
718 | |
719 | while (bv_len && iov_idx < iov_count) { | |
720 | unsigned int bytes; | |
0e0c6212 | 721 | char __user *iov_addr; |
c5dec1c3 FT |
722 | |
723 | bytes = min_t(unsigned int, | |
724 | iov[iov_idx].iov_len - iov_off, bv_len); | |
725 | iov_addr = iov[iov_idx].iov_base + iov_off; | |
726 | ||
727 | if (!ret) { | |
ecb554a8 | 728 | if (to_user) |
c5dec1c3 FT |
729 | ret = copy_to_user(iov_addr, bv_addr, |
730 | bytes); | |
731 | ||
ecb554a8 FT |
732 | if (from_user) |
733 | ret = copy_from_user(bv_addr, iov_addr, | |
734 | bytes); | |
735 | ||
c5dec1c3 FT |
736 | if (ret) |
737 | ret = -EFAULT; | |
738 | } | |
739 | ||
740 | bv_len -= bytes; | |
741 | bv_addr += bytes; | |
742 | iov_addr += bytes; | |
743 | iov_off += bytes; | |
744 | ||
745 | if (iov[iov_idx].iov_len == iov_off) { | |
746 | iov_idx++; | |
747 | iov_off = 0; | |
748 | } | |
749 | } | |
750 | ||
152e283f | 751 | if (do_free_page) |
c5dec1c3 FT |
752 | __free_page(bvec->bv_page); |
753 | } | |
754 | ||
755 | return ret; | |
756 | } | |
757 | ||
1da177e4 LT |
758 | /** |
759 | * bio_uncopy_user - finish previously mapped bio | |
760 | * @bio: bio being terminated | |
761 | * | |
762 | * Free pages allocated from bio_copy_user() and write back data | |
763 | * to user space in case of a read. | |
764 | */ | |
765 | int bio_uncopy_user(struct bio *bio) | |
766 | { | |
767 | struct bio_map_data *bmd = bio->bi_private; | |
81882766 | 768 | int ret = 0; |
1da177e4 | 769 | |
81882766 FT |
770 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) |
771 | ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs, | |
ecb554a8 FT |
772 | bmd->nr_sgvecs, bio_data_dir(bio) == READ, |
773 | 0, bmd->is_our_pages); | |
1da177e4 LT |
774 | bio_free_map_data(bmd); |
775 | bio_put(bio); | |
776 | return ret; | |
777 | } | |
a112a71d | 778 | EXPORT_SYMBOL(bio_uncopy_user); |
1da177e4 LT |
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 | ||
cb4644ca JA |
816 | /* |
817 | * Overflow, abort | |
818 | */ | |
819 | if (end < start) | |
820 | return ERR_PTR(-EINVAL); | |
821 | ||
c5dec1c3 FT |
822 | nr_pages += end - start; |
823 | len += iov[i].iov_len; | |
824 | } | |
825 | ||
69838727 FT |
826 | if (offset) |
827 | nr_pages++; | |
828 | ||
a3bce90e | 829 | bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask); |
1da177e4 LT |
830 | if (!bmd) |
831 | return ERR_PTR(-ENOMEM); | |
832 | ||
1da177e4 | 833 | ret = -ENOMEM; |
a9e9dc24 | 834 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
835 | if (!bio) |
836 | goto out_bmd; | |
837 | ||
7b6d91da CH |
838 | if (!write_to_vm) |
839 | bio->bi_rw |= REQ_WRITE; | |
1da177e4 LT |
840 | |
841 | ret = 0; | |
56c451f4 FT |
842 | |
843 | if (map_data) { | |
e623ddb4 | 844 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
845 | i = map_data->offset / PAGE_SIZE; |
846 | } | |
1da177e4 | 847 | while (len) { |
e623ddb4 | 848 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 849 | |
56c451f4 FT |
850 | bytes -= offset; |
851 | ||
1da177e4 LT |
852 | if (bytes > len) |
853 | bytes = len; | |
854 | ||
152e283f | 855 | if (map_data) { |
e623ddb4 | 856 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
857 | ret = -ENOMEM; |
858 | break; | |
859 | } | |
e623ddb4 FT |
860 | |
861 | page = map_data->pages[i / nr_pages]; | |
862 | page += (i % nr_pages); | |
863 | ||
864 | i++; | |
865 | } else { | |
152e283f | 866 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
867 | if (!page) { |
868 | ret = -ENOMEM; | |
869 | break; | |
870 | } | |
1da177e4 LT |
871 | } |
872 | ||
56c451f4 | 873 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 874 | break; |
1da177e4 LT |
875 | |
876 | len -= bytes; | |
56c451f4 | 877 | offset = 0; |
1da177e4 LT |
878 | } |
879 | ||
880 | if (ret) | |
881 | goto cleanup; | |
882 | ||
883 | /* | |
884 | * success | |
885 | */ | |
ecb554a8 FT |
886 | if ((!write_to_vm && (!map_data || !map_data->null_mapped)) || |
887 | (map_data && map_data->from_user)) { | |
888 | ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 1, 0); | |
c5dec1c3 FT |
889 | if (ret) |
890 | goto cleanup; | |
1da177e4 LT |
891 | } |
892 | ||
152e283f | 893 | bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1); |
1da177e4 LT |
894 | return bio; |
895 | cleanup: | |
152e283f FT |
896 | if (!map_data) |
897 | bio_for_each_segment(bvec, bio, i) | |
898 | __free_page(bvec->bv_page); | |
1da177e4 LT |
899 | |
900 | bio_put(bio); | |
901 | out_bmd: | |
902 | bio_free_map_data(bmd); | |
903 | return ERR_PTR(ret); | |
904 | } | |
905 | ||
c5dec1c3 FT |
906 | /** |
907 | * bio_copy_user - copy user data to bio | |
908 | * @q: destination block queue | |
152e283f | 909 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
910 | * @uaddr: start of user address |
911 | * @len: length in bytes | |
912 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 913 | * @gfp_mask: memory allocation flags |
c5dec1c3 FT |
914 | * |
915 | * Prepares and returns a bio for indirect user io, bouncing data | |
916 | * to/from kernel pages as necessary. Must be paired with | |
917 | * call bio_uncopy_user() on io completion. | |
918 | */ | |
152e283f FT |
919 | struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data, |
920 | unsigned long uaddr, unsigned int len, | |
921 | int write_to_vm, gfp_t gfp_mask) | |
c5dec1c3 FT |
922 | { |
923 | struct sg_iovec iov; | |
924 | ||
925 | iov.iov_base = (void __user *)uaddr; | |
926 | iov.iov_len = len; | |
927 | ||
152e283f | 928 | return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask); |
c5dec1c3 | 929 | } |
a112a71d | 930 | EXPORT_SYMBOL(bio_copy_user); |
c5dec1c3 | 931 | |
165125e1 | 932 | static struct bio *__bio_map_user_iov(struct request_queue *q, |
f1970baf JB |
933 | struct block_device *bdev, |
934 | struct sg_iovec *iov, int iov_count, | |
a3bce90e | 935 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 | 936 | { |
f1970baf JB |
937 | int i, j; |
938 | int nr_pages = 0; | |
1da177e4 LT |
939 | struct page **pages; |
940 | struct bio *bio; | |
f1970baf JB |
941 | int cur_page = 0; |
942 | int ret, offset; | |
1da177e4 | 943 | |
f1970baf JB |
944 | for (i = 0; i < iov_count; i++) { |
945 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
946 | unsigned long len = iov[i].iov_len; | |
947 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
948 | unsigned long start = uaddr >> PAGE_SHIFT; | |
949 | ||
cb4644ca JA |
950 | /* |
951 | * Overflow, abort | |
952 | */ | |
953 | if (end < start) | |
954 | return ERR_PTR(-EINVAL); | |
955 | ||
f1970baf JB |
956 | nr_pages += end - start; |
957 | /* | |
ad2d7225 | 958 | * buffer must be aligned to at least hardsector size for now |
f1970baf | 959 | */ |
ad2d7225 | 960 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
961 | return ERR_PTR(-EINVAL); |
962 | } | |
963 | ||
964 | if (!nr_pages) | |
1da177e4 LT |
965 | return ERR_PTR(-EINVAL); |
966 | ||
a9e9dc24 | 967 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
968 | if (!bio) |
969 | return ERR_PTR(-ENOMEM); | |
970 | ||
971 | ret = -ENOMEM; | |
a3bce90e | 972 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
973 | if (!pages) |
974 | goto out; | |
975 | ||
f1970baf JB |
976 | for (i = 0; i < iov_count; i++) { |
977 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
978 | unsigned long len = iov[i].iov_len; | |
979 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
980 | unsigned long start = uaddr >> PAGE_SHIFT; | |
981 | const int local_nr_pages = end - start; | |
982 | const int page_limit = cur_page + local_nr_pages; | |
cb4644ca | 983 | |
f5dd33c4 NP |
984 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
985 | write_to_vm, &pages[cur_page]); | |
99172157 JA |
986 | if (ret < local_nr_pages) { |
987 | ret = -EFAULT; | |
f1970baf | 988 | goto out_unmap; |
99172157 | 989 | } |
f1970baf JB |
990 | |
991 | offset = uaddr & ~PAGE_MASK; | |
992 | for (j = cur_page; j < page_limit; j++) { | |
993 | unsigned int bytes = PAGE_SIZE - offset; | |
994 | ||
995 | if (len <= 0) | |
996 | break; | |
997 | ||
998 | if (bytes > len) | |
999 | bytes = len; | |
1000 | ||
1001 | /* | |
1002 | * sorry... | |
1003 | */ | |
defd94b7 MC |
1004 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1005 | bytes) | |
f1970baf JB |
1006 | break; |
1007 | ||
1008 | len -= bytes; | |
1009 | offset = 0; | |
1010 | } | |
1da177e4 | 1011 | |
f1970baf | 1012 | cur_page = j; |
1da177e4 | 1013 | /* |
f1970baf | 1014 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1015 | */ |
f1970baf JB |
1016 | while (j < page_limit) |
1017 | page_cache_release(pages[j++]); | |
1da177e4 LT |
1018 | } |
1019 | ||
1da177e4 LT |
1020 | kfree(pages); |
1021 | ||
1022 | /* | |
1023 | * set data direction, and check if mapped pages need bouncing | |
1024 | */ | |
1025 | if (!write_to_vm) | |
7b6d91da | 1026 | bio->bi_rw |= REQ_WRITE; |
1da177e4 | 1027 | |
f1970baf | 1028 | bio->bi_bdev = bdev; |
1da177e4 LT |
1029 | bio->bi_flags |= (1 << BIO_USER_MAPPED); |
1030 | return bio; | |
f1970baf JB |
1031 | |
1032 | out_unmap: | |
1033 | for (i = 0; i < nr_pages; i++) { | |
1034 | if(!pages[i]) | |
1035 | break; | |
1036 | page_cache_release(pages[i]); | |
1037 | } | |
1038 | out: | |
1da177e4 LT |
1039 | kfree(pages); |
1040 | bio_put(bio); | |
1041 | return ERR_PTR(ret); | |
1042 | } | |
1043 | ||
1044 | /** | |
1045 | * bio_map_user - map user address into bio | |
165125e1 | 1046 | * @q: the struct request_queue for the bio |
1da177e4 LT |
1047 | * @bdev: destination block device |
1048 | * @uaddr: start of user address | |
1049 | * @len: length in bytes | |
1050 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1051 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
1052 | * |
1053 | * Map the user space address into a bio suitable for io to a block | |
1054 | * device. Returns an error pointer in case of error. | |
1055 | */ | |
165125e1 | 1056 | struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev, |
a3bce90e FT |
1057 | unsigned long uaddr, unsigned int len, int write_to_vm, |
1058 | gfp_t gfp_mask) | |
f1970baf JB |
1059 | { |
1060 | struct sg_iovec iov; | |
1061 | ||
3f70353e | 1062 | iov.iov_base = (void __user *)uaddr; |
f1970baf JB |
1063 | iov.iov_len = len; |
1064 | ||
a3bce90e | 1065 | return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask); |
f1970baf | 1066 | } |
a112a71d | 1067 | EXPORT_SYMBOL(bio_map_user); |
f1970baf JB |
1068 | |
1069 | /** | |
1070 | * bio_map_user_iov - map user sg_iovec table into bio | |
165125e1 | 1071 | * @q: the struct request_queue for the bio |
f1970baf JB |
1072 | * @bdev: destination block device |
1073 | * @iov: the iovec. | |
1074 | * @iov_count: number of elements in the iovec | |
1075 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1076 | * @gfp_mask: memory allocation flags |
f1970baf JB |
1077 | * |
1078 | * Map the user space address into a bio suitable for io to a block | |
1079 | * device. Returns an error pointer in case of error. | |
1080 | */ | |
165125e1 | 1081 | struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev, |
f1970baf | 1082 | struct sg_iovec *iov, int iov_count, |
a3bce90e | 1083 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 LT |
1084 | { |
1085 | struct bio *bio; | |
1086 | ||
a3bce90e FT |
1087 | bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm, |
1088 | gfp_mask); | |
1da177e4 LT |
1089 | if (IS_ERR(bio)) |
1090 | return bio; | |
1091 | ||
1092 | /* | |
1093 | * subtle -- if __bio_map_user() ended up bouncing a bio, | |
1094 | * it would normally disappear when its bi_end_io is run. | |
1095 | * however, we need it for the unmap, so grab an extra | |
1096 | * reference to it | |
1097 | */ | |
1098 | bio_get(bio); | |
1099 | ||
0e75f906 | 1100 | return bio; |
1da177e4 LT |
1101 | } |
1102 | ||
1103 | static void __bio_unmap_user(struct bio *bio) | |
1104 | { | |
1105 | struct bio_vec *bvec; | |
1106 | int i; | |
1107 | ||
1108 | /* | |
1109 | * make sure we dirty pages we wrote to | |
1110 | */ | |
1111 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1112 | if (bio_data_dir(bio) == READ) | |
1113 | set_page_dirty_lock(bvec->bv_page); | |
1114 | ||
1115 | page_cache_release(bvec->bv_page); | |
1116 | } | |
1117 | ||
1118 | bio_put(bio); | |
1119 | } | |
1120 | ||
1121 | /** | |
1122 | * bio_unmap_user - unmap a bio | |
1123 | * @bio: the bio being unmapped | |
1124 | * | |
1125 | * Unmap a bio previously mapped by bio_map_user(). Must be called with | |
1126 | * a process context. | |
1127 | * | |
1128 | * bio_unmap_user() may sleep. | |
1129 | */ | |
1130 | void bio_unmap_user(struct bio *bio) | |
1131 | { | |
1132 | __bio_unmap_user(bio); | |
1133 | bio_put(bio); | |
1134 | } | |
a112a71d | 1135 | EXPORT_SYMBOL(bio_unmap_user); |
1da177e4 | 1136 | |
6712ecf8 | 1137 | static void bio_map_kern_endio(struct bio *bio, int err) |
b823825e | 1138 | { |
b823825e | 1139 | bio_put(bio); |
b823825e JA |
1140 | } |
1141 | ||
165125e1 | 1142 | static struct bio *__bio_map_kern(struct request_queue *q, void *data, |
27496a8c | 1143 | unsigned int len, gfp_t gfp_mask) |
df46b9a4 MC |
1144 | { |
1145 | unsigned long kaddr = (unsigned long)data; | |
1146 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1147 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1148 | const int nr_pages = end - start; | |
1149 | int offset, i; | |
1150 | struct bio *bio; | |
1151 | ||
a9e9dc24 | 1152 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1153 | if (!bio) |
1154 | return ERR_PTR(-ENOMEM); | |
1155 | ||
1156 | offset = offset_in_page(kaddr); | |
1157 | for (i = 0; i < nr_pages; i++) { | |
1158 | unsigned int bytes = PAGE_SIZE - offset; | |
1159 | ||
1160 | if (len <= 0) | |
1161 | break; | |
1162 | ||
1163 | if (bytes > len) | |
1164 | bytes = len; | |
1165 | ||
defd94b7 MC |
1166 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
1167 | offset) < bytes) | |
df46b9a4 MC |
1168 | break; |
1169 | ||
1170 | data += bytes; | |
1171 | len -= bytes; | |
1172 | offset = 0; | |
1173 | } | |
1174 | ||
b823825e | 1175 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1176 | return bio; |
1177 | } | |
1178 | ||
1179 | /** | |
1180 | * bio_map_kern - map kernel address into bio | |
165125e1 | 1181 | * @q: the struct request_queue for the bio |
df46b9a4 MC |
1182 | * @data: pointer to buffer to map |
1183 | * @len: length in bytes | |
1184 | * @gfp_mask: allocation flags for bio allocation | |
1185 | * | |
1186 | * Map the kernel address into a bio suitable for io to a block | |
1187 | * device. Returns an error pointer in case of error. | |
1188 | */ | |
165125e1 | 1189 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
27496a8c | 1190 | gfp_t gfp_mask) |
df46b9a4 MC |
1191 | { |
1192 | struct bio *bio; | |
1193 | ||
1194 | bio = __bio_map_kern(q, data, len, gfp_mask); | |
1195 | if (IS_ERR(bio)) | |
1196 | return bio; | |
1197 | ||
1198 | if (bio->bi_size == len) | |
1199 | return bio; | |
1200 | ||
1201 | /* | |
1202 | * Don't support partial mappings. | |
1203 | */ | |
1204 | bio_put(bio); | |
1205 | return ERR_PTR(-EINVAL); | |
1206 | } | |
a112a71d | 1207 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1208 | |
68154e90 FT |
1209 | static void bio_copy_kern_endio(struct bio *bio, int err) |
1210 | { | |
1211 | struct bio_vec *bvec; | |
1212 | const int read = bio_data_dir(bio) == READ; | |
76029ff3 | 1213 | struct bio_map_data *bmd = bio->bi_private; |
68154e90 | 1214 | int i; |
76029ff3 | 1215 | char *p = bmd->sgvecs[0].iov_base; |
68154e90 FT |
1216 | |
1217 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1218 | char *addr = page_address(bvec->bv_page); | |
76029ff3 | 1219 | int len = bmd->iovecs[i].bv_len; |
68154e90 | 1220 | |
4fc981ef | 1221 | if (read) |
76029ff3 | 1222 | memcpy(p, addr, len); |
68154e90 FT |
1223 | |
1224 | __free_page(bvec->bv_page); | |
76029ff3 | 1225 | p += len; |
68154e90 FT |
1226 | } |
1227 | ||
76029ff3 | 1228 | bio_free_map_data(bmd); |
68154e90 FT |
1229 | bio_put(bio); |
1230 | } | |
1231 | ||
1232 | /** | |
1233 | * bio_copy_kern - copy kernel address into bio | |
1234 | * @q: the struct request_queue for the bio | |
1235 | * @data: pointer to buffer to copy | |
1236 | * @len: length in bytes | |
1237 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1238 | * @reading: data direction is READ |
68154e90 FT |
1239 | * |
1240 | * copy the kernel address into a bio suitable for io to a block | |
1241 | * device. Returns an error pointer in case of error. | |
1242 | */ | |
1243 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1244 | gfp_t gfp_mask, int reading) | |
1245 | { | |
68154e90 FT |
1246 | struct bio *bio; |
1247 | struct bio_vec *bvec; | |
4d8ab62e | 1248 | int i; |
68154e90 | 1249 | |
4d8ab62e FT |
1250 | bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask); |
1251 | if (IS_ERR(bio)) | |
1252 | return bio; | |
68154e90 FT |
1253 | |
1254 | if (!reading) { | |
1255 | void *p = data; | |
1256 | ||
1257 | bio_for_each_segment(bvec, bio, i) { | |
1258 | char *addr = page_address(bvec->bv_page); | |
1259 | ||
1260 | memcpy(addr, p, bvec->bv_len); | |
1261 | p += bvec->bv_len; | |
1262 | } | |
1263 | } | |
1264 | ||
68154e90 | 1265 | bio->bi_end_io = bio_copy_kern_endio; |
76029ff3 | 1266 | |
68154e90 | 1267 | return bio; |
68154e90 | 1268 | } |
a112a71d | 1269 | EXPORT_SYMBOL(bio_copy_kern); |
68154e90 | 1270 | |
1da177e4 LT |
1271 | /* |
1272 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1273 | * for performing direct-IO in BIOs. | |
1274 | * | |
1275 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1276 | * because the required locks are not interrupt-safe. So what we can do is to | |
1277 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1278 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1279 | * in process context. | |
1280 | * | |
1281 | * We special-case compound pages here: normally this means reads into hugetlb | |
1282 | * pages. The logic in here doesn't really work right for compound pages | |
1283 | * because the VM does not uniformly chase down the head page in all cases. | |
1284 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1285 | * handle them at all. So we skip compound pages here at an early stage. | |
1286 | * | |
1287 | * Note that this code is very hard to test under normal circumstances because | |
1288 | * direct-io pins the pages with get_user_pages(). This makes | |
1289 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1290 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1291 | * pagecache. |
1292 | * | |
1293 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1294 | * deferred bio dirtying paths. | |
1295 | */ | |
1296 | ||
1297 | /* | |
1298 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1299 | */ | |
1300 | void bio_set_pages_dirty(struct bio *bio) | |
1301 | { | |
1302 | struct bio_vec *bvec = bio->bi_io_vec; | |
1303 | int i; | |
1304 | ||
1305 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1306 | struct page *page = bvec[i].bv_page; | |
1307 | ||
1308 | if (page && !PageCompound(page)) | |
1309 | set_page_dirty_lock(page); | |
1310 | } | |
1311 | } | |
1312 | ||
86b6c7a7 | 1313 | static void bio_release_pages(struct bio *bio) |
1da177e4 LT |
1314 | { |
1315 | struct bio_vec *bvec = bio->bi_io_vec; | |
1316 | int i; | |
1317 | ||
1318 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1319 | struct page *page = bvec[i].bv_page; | |
1320 | ||
1321 | if (page) | |
1322 | put_page(page); | |
1323 | } | |
1324 | } | |
1325 | ||
1326 | /* | |
1327 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1328 | * If they are, then fine. If, however, some pages are clean then they must | |
1329 | * have been written out during the direct-IO read. So we take another ref on | |
1330 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1331 | * | |
1332 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
1333 | * here on. It will run one page_cache_release() against each page and will | |
1334 | * run one bio_put() against the BIO. | |
1335 | */ | |
1336 | ||
65f27f38 | 1337 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1338 | |
65f27f38 | 1339 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1340 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1341 | static struct bio *bio_dirty_list; | |
1342 | ||
1343 | /* | |
1344 | * This runs in process context | |
1345 | */ | |
65f27f38 | 1346 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1347 | { |
1348 | unsigned long flags; | |
1349 | struct bio *bio; | |
1350 | ||
1351 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1352 | bio = bio_dirty_list; | |
1353 | bio_dirty_list = NULL; | |
1354 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1355 | ||
1356 | while (bio) { | |
1357 | struct bio *next = bio->bi_private; | |
1358 | ||
1359 | bio_set_pages_dirty(bio); | |
1360 | bio_release_pages(bio); | |
1361 | bio_put(bio); | |
1362 | bio = next; | |
1363 | } | |
1364 | } | |
1365 | ||
1366 | void bio_check_pages_dirty(struct bio *bio) | |
1367 | { | |
1368 | struct bio_vec *bvec = bio->bi_io_vec; | |
1369 | int nr_clean_pages = 0; | |
1370 | int i; | |
1371 | ||
1372 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1373 | struct page *page = bvec[i].bv_page; | |
1374 | ||
1375 | if (PageDirty(page) || PageCompound(page)) { | |
1376 | page_cache_release(page); | |
1377 | bvec[i].bv_page = NULL; | |
1378 | } else { | |
1379 | nr_clean_pages++; | |
1380 | } | |
1381 | } | |
1382 | ||
1383 | if (nr_clean_pages) { | |
1384 | unsigned long flags; | |
1385 | ||
1386 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1387 | bio->bi_private = bio_dirty_list; | |
1388 | bio_dirty_list = bio; | |
1389 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1390 | schedule_work(&bio_dirty_work); | |
1391 | } else { | |
1392 | bio_put(bio); | |
1393 | } | |
1394 | } | |
1395 | ||
2d4dc890 IL |
1396 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1397 | void bio_flush_dcache_pages(struct bio *bi) | |
1398 | { | |
1399 | int i; | |
1400 | struct bio_vec *bvec; | |
1401 | ||
1402 | bio_for_each_segment(bvec, bi, i) | |
1403 | flush_dcache_page(bvec->bv_page); | |
1404 | } | |
1405 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1406 | #endif | |
1407 | ||
1da177e4 LT |
1408 | /** |
1409 | * bio_endio - end I/O on a bio | |
1410 | * @bio: bio | |
1da177e4 LT |
1411 | * @error: error, if any |
1412 | * | |
1413 | * Description: | |
6712ecf8 | 1414 | * bio_endio() will end I/O on the whole bio. bio_endio() is the |
5bb23a68 N |
1415 | * preferred way to end I/O on a bio, it takes care of clearing |
1416 | * BIO_UPTODATE on error. @error is 0 on success, and and one of the | |
1417 | * established -Exxxx (-EIO, for instance) error values in case | |
25985edc | 1418 | * something went wrong. No one should call bi_end_io() directly on a |
5bb23a68 N |
1419 | * bio unless they own it and thus know that it has an end_io |
1420 | * function. | |
1da177e4 | 1421 | **/ |
6712ecf8 | 1422 | void bio_endio(struct bio *bio, int error) |
1da177e4 LT |
1423 | { |
1424 | if (error) | |
1425 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
9cc54d40 N |
1426 | else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) |
1427 | error = -EIO; | |
1da177e4 | 1428 | |
5bb23a68 | 1429 | if (bio->bi_end_io) |
6712ecf8 | 1430 | bio->bi_end_io(bio, error); |
1da177e4 | 1431 | } |
a112a71d | 1432 | EXPORT_SYMBOL(bio_endio); |
1da177e4 LT |
1433 | |
1434 | void bio_pair_release(struct bio_pair *bp) | |
1435 | { | |
1436 | if (atomic_dec_and_test(&bp->cnt)) { | |
1437 | struct bio *master = bp->bio1.bi_private; | |
1438 | ||
6712ecf8 | 1439 | bio_endio(master, bp->error); |
1da177e4 LT |
1440 | mempool_free(bp, bp->bio2.bi_private); |
1441 | } | |
1442 | } | |
a112a71d | 1443 | EXPORT_SYMBOL(bio_pair_release); |
1da177e4 | 1444 | |
6712ecf8 | 1445 | static void bio_pair_end_1(struct bio *bi, int err) |
1da177e4 LT |
1446 | { |
1447 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio1); | |
1448 | ||
1449 | if (err) | |
1450 | bp->error = err; | |
1451 | ||
1da177e4 | 1452 | bio_pair_release(bp); |
1da177e4 LT |
1453 | } |
1454 | ||
6712ecf8 | 1455 | static void bio_pair_end_2(struct bio *bi, int err) |
1da177e4 LT |
1456 | { |
1457 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio2); | |
1458 | ||
1459 | if (err) | |
1460 | bp->error = err; | |
1461 | ||
1da177e4 | 1462 | bio_pair_release(bp); |
1da177e4 LT |
1463 | } |
1464 | ||
1465 | /* | |
c7eee1b8 | 1466 | * split a bio - only worry about a bio with a single page in its iovec |
1da177e4 | 1467 | */ |
6feef531 | 1468 | struct bio_pair *bio_split(struct bio *bi, int first_sectors) |
1da177e4 | 1469 | { |
6feef531 | 1470 | struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO); |
1da177e4 LT |
1471 | |
1472 | if (!bp) | |
1473 | return bp; | |
1474 | ||
5f3ea37c | 1475 | trace_block_split(bdev_get_queue(bi->bi_bdev), bi, |
2056a782 JA |
1476 | bi->bi_sector + first_sectors); |
1477 | ||
02f3939e | 1478 | BUG_ON(bi->bi_vcnt != 1 && bi->bi_vcnt != 0); |
1da177e4 LT |
1479 | BUG_ON(bi->bi_idx != 0); |
1480 | atomic_set(&bp->cnt, 3); | |
1481 | bp->error = 0; | |
1482 | bp->bio1 = *bi; | |
1483 | bp->bio2 = *bi; | |
1484 | bp->bio2.bi_sector += first_sectors; | |
1485 | bp->bio2.bi_size -= first_sectors << 9; | |
1486 | bp->bio1.bi_size = first_sectors << 9; | |
1487 | ||
02f3939e SL |
1488 | if (bi->bi_vcnt != 0) { |
1489 | bp->bv1 = bi->bi_io_vec[0]; | |
1490 | bp->bv2 = bi->bi_io_vec[0]; | |
4363ac7c | 1491 | |
02f3939e SL |
1492 | if (bio_is_rw(bi)) { |
1493 | bp->bv2.bv_offset += first_sectors << 9; | |
1494 | bp->bv2.bv_len -= first_sectors << 9; | |
1495 | bp->bv1.bv_len = first_sectors << 9; | |
1496 | } | |
1da177e4 | 1497 | |
02f3939e SL |
1498 | bp->bio1.bi_io_vec = &bp->bv1; |
1499 | bp->bio2.bi_io_vec = &bp->bv2; | |
1da177e4 | 1500 | |
02f3939e SL |
1501 | bp->bio1.bi_max_vecs = 1; |
1502 | bp->bio2.bi_max_vecs = 1; | |
1503 | } | |
a2eb0c10 | 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 | ||
bb799ca0 | 1623 | if (!biovec_create_pools(bs, pool_size)) |
1da177e4 LT |
1624 | return bs; |
1625 | ||
1626 | bad: | |
1627 | bioset_free(bs); | |
1628 | return NULL; | |
1629 | } | |
a112a71d | 1630 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 1631 | |
852c788f TH |
1632 | #ifdef CONFIG_BLK_CGROUP |
1633 | /** | |
1634 | * bio_associate_current - associate a bio with %current | |
1635 | * @bio: target bio | |
1636 | * | |
1637 | * Associate @bio with %current if it hasn't been associated yet. Block | |
1638 | * layer will treat @bio as if it were issued by %current no matter which | |
1639 | * task actually issues it. | |
1640 | * | |
1641 | * This function takes an extra reference of @task's io_context and blkcg | |
1642 | * which will be put when @bio is released. The caller must own @bio, | |
1643 | * ensure %current->io_context exists, and is responsible for synchronizing | |
1644 | * calls to this function. | |
1645 | */ | |
1646 | int bio_associate_current(struct bio *bio) | |
1647 | { | |
1648 | struct io_context *ioc; | |
1649 | struct cgroup_subsys_state *css; | |
1650 | ||
1651 | if (bio->bi_ioc) | |
1652 | return -EBUSY; | |
1653 | ||
1654 | ioc = current->io_context; | |
1655 | if (!ioc) | |
1656 | return -ENOENT; | |
1657 | ||
1658 | /* acquire active ref on @ioc and associate */ | |
1659 | get_io_context_active(ioc); | |
1660 | bio->bi_ioc = ioc; | |
1661 | ||
1662 | /* associate blkcg if exists */ | |
1663 | rcu_read_lock(); | |
1664 | css = task_subsys_state(current, blkio_subsys_id); | |
1665 | if (css && css_tryget(css)) | |
1666 | bio->bi_css = css; | |
1667 | rcu_read_unlock(); | |
1668 | ||
1669 | return 0; | |
1670 | } | |
1671 | ||
1672 | /** | |
1673 | * bio_disassociate_task - undo bio_associate_current() | |
1674 | * @bio: target bio | |
1675 | */ | |
1676 | void bio_disassociate_task(struct bio *bio) | |
1677 | { | |
1678 | if (bio->bi_ioc) { | |
1679 | put_io_context(bio->bi_ioc); | |
1680 | bio->bi_ioc = NULL; | |
1681 | } | |
1682 | if (bio->bi_css) { | |
1683 | css_put(bio->bi_css); | |
1684 | bio->bi_css = NULL; | |
1685 | } | |
1686 | } | |
1687 | ||
1688 | #endif /* CONFIG_BLK_CGROUP */ | |
1689 | ||
1da177e4 LT |
1690 | static void __init biovec_init_slabs(void) |
1691 | { | |
1692 | int i; | |
1693 | ||
1694 | for (i = 0; i < BIOVEC_NR_POOLS; i++) { | |
1695 | int size; | |
1696 | struct biovec_slab *bvs = bvec_slabs + i; | |
1697 | ||
a7fcd37c JA |
1698 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
1699 | bvs->slab = NULL; | |
1700 | continue; | |
1701 | } | |
a7fcd37c | 1702 | |
1da177e4 LT |
1703 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
1704 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 1705 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1706 | } |
1707 | } | |
1708 | ||
1709 | static int __init init_bio(void) | |
1710 | { | |
bb799ca0 JA |
1711 | bio_slab_max = 2; |
1712 | bio_slab_nr = 0; | |
1713 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
1714 | if (!bio_slabs) | |
1715 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 1716 | |
7878cba9 | 1717 | bio_integrity_init(); |
1da177e4 LT |
1718 | biovec_init_slabs(); |
1719 | ||
bb799ca0 | 1720 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
1da177e4 LT |
1721 | if (!fs_bio_set) |
1722 | panic("bio: can't allocate bios\n"); | |
1723 | ||
a91a2785 MP |
1724 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
1725 | panic("bio: can't create integrity pool\n"); | |
1726 | ||
0eaae62a MD |
1727 | bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES, |
1728 | sizeof(struct bio_pair)); | |
1da177e4 LT |
1729 | if (!bio_split_pool) |
1730 | panic("bio: can't create split pool\n"); | |
1731 | ||
1732 | return 0; | |
1733 | } | |
1da177e4 | 1734 | subsys_initcall(init_bio); |