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; |
1da177e4 | 58 | |
bb799ca0 JA |
59 | /* |
60 | * Our slab pool management | |
61 | */ | |
62 | struct bio_slab { | |
63 | struct kmem_cache *slab; | |
64 | unsigned int slab_ref; | |
65 | unsigned int slab_size; | |
66 | char name[8]; | |
67 | }; | |
68 | static DEFINE_MUTEX(bio_slab_lock); | |
69 | static struct bio_slab *bio_slabs; | |
70 | static unsigned int bio_slab_nr, bio_slab_max; | |
71 | ||
72 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
73 | { | |
74 | unsigned int sz = sizeof(struct bio) + extra_size; | |
75 | struct kmem_cache *slab = NULL; | |
389d7b26 | 76 | struct bio_slab *bslab, *new_bio_slabs; |
bb799ca0 JA |
77 | unsigned int i, entry = -1; |
78 | ||
79 | mutex_lock(&bio_slab_lock); | |
80 | ||
81 | i = 0; | |
82 | while (i < bio_slab_nr) { | |
f06f135d | 83 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
84 | |
85 | if (!bslab->slab && entry == -1) | |
86 | entry = i; | |
87 | else if (bslab->slab_size == sz) { | |
88 | slab = bslab->slab; | |
89 | bslab->slab_ref++; | |
90 | break; | |
91 | } | |
92 | i++; | |
93 | } | |
94 | ||
95 | if (slab) | |
96 | goto out_unlock; | |
97 | ||
98 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
99 | bio_slab_max <<= 1; | |
389d7b26 AK |
100 | new_bio_slabs = krealloc(bio_slabs, |
101 | bio_slab_max * sizeof(struct bio_slab), | |
102 | GFP_KERNEL); | |
103 | if (!new_bio_slabs) | |
bb799ca0 | 104 | goto out_unlock; |
389d7b26 | 105 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
106 | } |
107 | if (entry == -1) | |
108 | entry = bio_slab_nr++; | |
109 | ||
110 | bslab = &bio_slabs[entry]; | |
111 | ||
112 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
113 | slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL); | |
114 | if (!slab) | |
115 | goto out_unlock; | |
116 | ||
80cdc6da | 117 | printk(KERN_INFO "bio: create slab <%s> at %d\n", bslab->name, entry); |
bb799ca0 JA |
118 | bslab->slab = slab; |
119 | bslab->slab_ref = 1; | |
120 | bslab->slab_size = sz; | |
121 | out_unlock: | |
122 | mutex_unlock(&bio_slab_lock); | |
123 | return slab; | |
124 | } | |
125 | ||
126 | static void bio_put_slab(struct bio_set *bs) | |
127 | { | |
128 | struct bio_slab *bslab = NULL; | |
129 | unsigned int i; | |
130 | ||
131 | mutex_lock(&bio_slab_lock); | |
132 | ||
133 | for (i = 0; i < bio_slab_nr; i++) { | |
134 | if (bs->bio_slab == bio_slabs[i].slab) { | |
135 | bslab = &bio_slabs[i]; | |
136 | break; | |
137 | } | |
138 | } | |
139 | ||
140 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
141 | goto out; | |
142 | ||
143 | WARN_ON(!bslab->slab_ref); | |
144 | ||
145 | if (--bslab->slab_ref) | |
146 | goto out; | |
147 | ||
148 | kmem_cache_destroy(bslab->slab); | |
149 | bslab->slab = NULL; | |
150 | ||
151 | out: | |
152 | mutex_unlock(&bio_slab_lock); | |
153 | } | |
154 | ||
7ba1ba12 MP |
155 | unsigned int bvec_nr_vecs(unsigned short idx) |
156 | { | |
157 | return bvec_slabs[idx].nr_vecs; | |
158 | } | |
159 | ||
bb799ca0 JA |
160 | void bvec_free_bs(struct bio_set *bs, struct bio_vec *bv, unsigned int idx) |
161 | { | |
162 | BIO_BUG_ON(idx >= BIOVEC_NR_POOLS); | |
163 | ||
164 | if (idx == BIOVEC_MAX_IDX) | |
165 | mempool_free(bv, bs->bvec_pool); | |
166 | else { | |
167 | struct biovec_slab *bvs = bvec_slabs + idx; | |
168 | ||
169 | kmem_cache_free(bvs->slab, bv); | |
170 | } | |
171 | } | |
172 | ||
7ff9345f JA |
173 | struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, |
174 | struct bio_set *bs) | |
1da177e4 LT |
175 | { |
176 | struct bio_vec *bvl; | |
1da177e4 | 177 | |
7ff9345f JA |
178 | /* |
179 | * see comment near bvec_array define! | |
180 | */ | |
181 | switch (nr) { | |
182 | case 1: | |
183 | *idx = 0; | |
184 | break; | |
185 | case 2 ... 4: | |
186 | *idx = 1; | |
187 | break; | |
188 | case 5 ... 16: | |
189 | *idx = 2; | |
190 | break; | |
191 | case 17 ... 64: | |
192 | *idx = 3; | |
193 | break; | |
194 | case 65 ... 128: | |
195 | *idx = 4; | |
196 | break; | |
197 | case 129 ... BIO_MAX_PAGES: | |
198 | *idx = 5; | |
199 | break; | |
200 | default: | |
201 | return NULL; | |
202 | } | |
203 | ||
204 | /* | |
205 | * idx now points to the pool we want to allocate from. only the | |
206 | * 1-vec entry pool is mempool backed. | |
207 | */ | |
208 | if (*idx == BIOVEC_MAX_IDX) { | |
209 | fallback: | |
210 | bvl = mempool_alloc(bs->bvec_pool, gfp_mask); | |
211 | } else { | |
212 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
213 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO); | |
214 | ||
0a0d96b0 | 215 | /* |
7ff9345f JA |
216 | * Make this allocation restricted and don't dump info on |
217 | * allocation failures, since we'll fallback to the mempool | |
218 | * in case of failure. | |
0a0d96b0 | 219 | */ |
7ff9345f | 220 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 221 | |
0a0d96b0 | 222 | /* |
7ff9345f JA |
223 | * Try a slab allocation. If this fails and __GFP_WAIT |
224 | * is set, retry with the 1-entry mempool | |
0a0d96b0 | 225 | */ |
7ff9345f JA |
226 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
227 | if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) { | |
228 | *idx = BIOVEC_MAX_IDX; | |
229 | goto fallback; | |
230 | } | |
231 | } | |
232 | ||
1da177e4 LT |
233 | return bvl; |
234 | } | |
235 | ||
4254bba1 | 236 | static void __bio_free(struct bio *bio) |
1da177e4 | 237 | { |
4254bba1 | 238 | bio_disassociate_task(bio); |
1da177e4 | 239 | |
7ba1ba12 | 240 | if (bio_integrity(bio)) |
1e2a410f | 241 | bio_integrity_free(bio); |
4254bba1 | 242 | } |
7ba1ba12 | 243 | |
4254bba1 KO |
244 | static void bio_free(struct bio *bio) |
245 | { | |
246 | struct bio_set *bs = bio->bi_pool; | |
247 | void *p; | |
248 | ||
249 | __bio_free(bio); | |
250 | ||
251 | if (bs) { | |
252 | if (bio_has_allocated_vec(bio)) | |
253 | bvec_free_bs(bs, bio->bi_io_vec, BIO_POOL_IDX(bio)); | |
254 | ||
255 | /* | |
256 | * If we have front padding, adjust the bio pointer before freeing | |
257 | */ | |
258 | p = bio; | |
bb799ca0 JA |
259 | p -= bs->front_pad; |
260 | ||
4254bba1 KO |
261 | mempool_free(p, bs->bio_pool); |
262 | } else { | |
263 | /* Bio was allocated by bio_kmalloc() */ | |
264 | kfree(bio); | |
265 | } | |
3676347a PO |
266 | } |
267 | ||
858119e1 | 268 | void bio_init(struct bio *bio) |
1da177e4 | 269 | { |
2b94de55 | 270 | memset(bio, 0, sizeof(*bio)); |
1da177e4 | 271 | bio->bi_flags = 1 << BIO_UPTODATE; |
1da177e4 | 272 | atomic_set(&bio->bi_cnt, 1); |
1da177e4 | 273 | } |
a112a71d | 274 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 275 | |
f44b48c7 KO |
276 | /** |
277 | * bio_reset - reinitialize a bio | |
278 | * @bio: bio to reset | |
279 | * | |
280 | * Description: | |
281 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
282 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
283 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
284 | * comment in struct bio. | |
285 | */ | |
286 | void bio_reset(struct bio *bio) | |
287 | { | |
288 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
289 | ||
4254bba1 | 290 | __bio_free(bio); |
f44b48c7 KO |
291 | |
292 | memset(bio, 0, BIO_RESET_BYTES); | |
293 | bio->bi_flags = flags|(1 << BIO_UPTODATE); | |
294 | } | |
295 | EXPORT_SYMBOL(bio_reset); | |
296 | ||
1da177e4 LT |
297 | /** |
298 | * bio_alloc_bioset - allocate a bio for I/O | |
299 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
300 | * @nr_iovecs: number of iovecs to pre-allocate | |
db18efac | 301 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
302 | * |
303 | * Description: | |
db18efac | 304 | * bio_alloc_bioset will try its own mempool to satisfy the allocation. |
1da177e4 | 305 | * If %__GFP_WAIT is set then we will block on the internal pool waiting |
db18efac | 306 | * for a &struct bio to become free. |
1da177e4 | 307 | **/ |
dd0fc66f | 308 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) |
1da177e4 | 309 | { |
451a9ebf | 310 | unsigned long idx = BIO_POOL_NONE; |
34053979 | 311 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
312 | struct bio *bio; |
313 | void *p; | |
314 | ||
315 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
316 | if (unlikely(!p)) | |
317 | return NULL; | |
318 | bio = p + bs->front_pad; | |
1da177e4 | 319 | |
34053979 | 320 | bio_init(bio); |
395c72a7 | 321 | bio->bi_pool = bs; |
34053979 IM |
322 | |
323 | if (unlikely(!nr_iovecs)) | |
324 | goto out_set; | |
325 | ||
326 | if (nr_iovecs <= BIO_INLINE_VECS) { | |
327 | bvl = bio->bi_inline_vecs; | |
328 | nr_iovecs = BIO_INLINE_VECS; | |
329 | } else { | |
330 | bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs); | |
331 | if (unlikely(!bvl)) | |
332 | goto err_free; | |
333 | ||
334 | nr_iovecs = bvec_nr_vecs(idx); | |
1da177e4 | 335 | } |
451a9ebf | 336 | out_set: |
34053979 IM |
337 | bio->bi_flags |= idx << BIO_POOL_OFFSET; |
338 | bio->bi_max_vecs = nr_iovecs; | |
34053979 | 339 | bio->bi_io_vec = bvl; |
1da177e4 | 340 | return bio; |
34053979 IM |
341 | |
342 | err_free: | |
451a9ebf | 343 | mempool_free(p, bs->bio_pool); |
34053979 | 344 | return NULL; |
1da177e4 | 345 | } |
a112a71d | 346 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 347 | |
451a9ebf TH |
348 | /** |
349 | * bio_alloc - allocate a new bio, memory pool backed | |
350 | * @gfp_mask: allocation mask to use | |
351 | * @nr_iovecs: number of iovecs | |
352 | * | |
5f04eeb8 AB |
353 | * bio_alloc will allocate a bio and associated bio_vec array that can hold |
354 | * at least @nr_iovecs entries. Allocations will be done from the | |
355 | * fs_bio_set. Also see @bio_alloc_bioset and @bio_kmalloc. | |
356 | * | |
357 | * If %__GFP_WAIT is set, then bio_alloc will always be able to allocate | |
358 | * a bio. This is due to the mempool guarantees. To make this work, callers | |
359 | * must never allocate more than 1 bio at a time from this pool. Callers | |
360 | * that need to allocate more than 1 bio must always submit the previously | |
361 | * allocated bio for IO before attempting to allocate a new one. Failure to | |
362 | * do so can cause livelocks under memory pressure. | |
451a9ebf TH |
363 | * |
364 | * RETURNS: | |
365 | * Pointer to new bio on success, NULL on failure. | |
366 | */ | |
121f0994 | 367 | struct bio *bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs) |
451a9ebf | 368 | { |
395c72a7 | 369 | return bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set); |
451a9ebf | 370 | } |
a112a71d | 371 | EXPORT_SYMBOL(bio_alloc); |
451a9ebf | 372 | |
86c824b9 | 373 | /** |
5f04eeb8 | 374 | * bio_kmalloc - allocate a bio for I/O using kmalloc() |
86c824b9 JA |
375 | * @gfp_mask: the GFP_ mask given to the slab allocator |
376 | * @nr_iovecs: number of iovecs to pre-allocate | |
377 | * | |
378 | * Description: | |
5f04eeb8 AB |
379 | * Allocate a new bio with @nr_iovecs bvecs. If @gfp_mask contains |
380 | * %__GFP_WAIT, the allocation is guaranteed to succeed. | |
86c824b9 JA |
381 | * |
382 | **/ | |
121f0994 | 383 | struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned int nr_iovecs) |
0a0d96b0 | 384 | { |
451a9ebf | 385 | struct bio *bio; |
0a0d96b0 | 386 | |
f3f63c1c JA |
387 | if (nr_iovecs > UIO_MAXIOV) |
388 | return NULL; | |
389 | ||
451a9ebf TH |
390 | bio = kmalloc(sizeof(struct bio) + nr_iovecs * sizeof(struct bio_vec), |
391 | gfp_mask); | |
392 | if (unlikely(!bio)) | |
393 | return NULL; | |
394 | ||
395 | bio_init(bio); | |
396 | bio->bi_flags |= BIO_POOL_NONE << BIO_POOL_OFFSET; | |
397 | bio->bi_max_vecs = nr_iovecs; | |
398 | bio->bi_io_vec = bio->bi_inline_vecs; | |
0a0d96b0 JA |
399 | |
400 | return bio; | |
401 | } | |
a112a71d | 402 | EXPORT_SYMBOL(bio_kmalloc); |
0a0d96b0 | 403 | |
1da177e4 LT |
404 | void zero_fill_bio(struct bio *bio) |
405 | { | |
406 | unsigned long flags; | |
407 | struct bio_vec *bv; | |
408 | int i; | |
409 | ||
410 | bio_for_each_segment(bv, bio, i) { | |
411 | char *data = bvec_kmap_irq(bv, &flags); | |
412 | memset(data, 0, bv->bv_len); | |
413 | flush_dcache_page(bv->bv_page); | |
414 | bvec_kunmap_irq(data, &flags); | |
415 | } | |
416 | } | |
417 | EXPORT_SYMBOL(zero_fill_bio); | |
418 | ||
419 | /** | |
420 | * bio_put - release a reference to a bio | |
421 | * @bio: bio to release reference to | |
422 | * | |
423 | * Description: | |
424 | * Put a reference to a &struct bio, either one you have gotten with | |
ad0bf110 | 425 | * bio_alloc, bio_get or bio_clone. The last put of a bio will free it. |
1da177e4 LT |
426 | **/ |
427 | void bio_put(struct bio *bio) | |
428 | { | |
429 | BIO_BUG_ON(!atomic_read(&bio->bi_cnt)); | |
430 | ||
431 | /* | |
432 | * last put frees it | |
433 | */ | |
4254bba1 KO |
434 | if (atomic_dec_and_test(&bio->bi_cnt)) |
435 | bio_free(bio); | |
1da177e4 | 436 | } |
a112a71d | 437 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 438 | |
165125e1 | 439 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
440 | { |
441 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
442 | blk_recount_segments(q, bio); | |
443 | ||
444 | return bio->bi_phys_segments; | |
445 | } | |
a112a71d | 446 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 447 | |
1da177e4 LT |
448 | /** |
449 | * __bio_clone - clone a bio | |
450 | * @bio: destination bio | |
451 | * @bio_src: bio to clone | |
452 | * | |
453 | * Clone a &bio. Caller will own the returned bio, but not | |
454 | * the actual data it points to. Reference count of returned | |
455 | * bio will be one. | |
456 | */ | |
858119e1 | 457 | void __bio_clone(struct bio *bio, struct bio *bio_src) |
1da177e4 | 458 | { |
e525e153 AM |
459 | memcpy(bio->bi_io_vec, bio_src->bi_io_vec, |
460 | bio_src->bi_max_vecs * sizeof(struct bio_vec)); | |
1da177e4 | 461 | |
5d84070e JA |
462 | /* |
463 | * most users will be overriding ->bi_bdev with a new target, | |
464 | * so we don't set nor calculate new physical/hw segment counts here | |
465 | */ | |
1da177e4 LT |
466 | bio->bi_sector = bio_src->bi_sector; |
467 | bio->bi_bdev = bio_src->bi_bdev; | |
468 | bio->bi_flags |= 1 << BIO_CLONED; | |
469 | bio->bi_rw = bio_src->bi_rw; | |
1da177e4 LT |
470 | bio->bi_vcnt = bio_src->bi_vcnt; |
471 | bio->bi_size = bio_src->bi_size; | |
a5453be4 | 472 | bio->bi_idx = bio_src->bi_idx; |
1da177e4 | 473 | } |
a112a71d | 474 | EXPORT_SYMBOL(__bio_clone); |
1da177e4 LT |
475 | |
476 | /** | |
477 | * bio_clone - clone a bio | |
478 | * @bio: bio to clone | |
479 | * @gfp_mask: allocation priority | |
480 | * | |
481 | * Like __bio_clone, only also allocates the returned bio | |
482 | */ | |
dd0fc66f | 483 | struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask) |
1da177e4 | 484 | { |
395c72a7 | 485 | struct bio *b = bio_alloc(gfp_mask, bio->bi_max_vecs); |
1da177e4 | 486 | |
7ba1ba12 MP |
487 | if (!b) |
488 | return NULL; | |
489 | ||
7ba1ba12 MP |
490 | __bio_clone(b, bio); |
491 | ||
492 | if (bio_integrity(bio)) { | |
493 | int ret; | |
494 | ||
1e2a410f | 495 | ret = bio_integrity_clone(b, bio, gfp_mask); |
7ba1ba12 | 496 | |
059ea331 LZ |
497 | if (ret < 0) { |
498 | bio_put(b); | |
7ba1ba12 | 499 | return NULL; |
059ea331 | 500 | } |
3676347a | 501 | } |
1da177e4 LT |
502 | |
503 | return b; | |
504 | } | |
a112a71d | 505 | EXPORT_SYMBOL(bio_clone); |
1da177e4 LT |
506 | |
507 | /** | |
508 | * bio_get_nr_vecs - return approx number of vecs | |
509 | * @bdev: I/O target | |
510 | * | |
511 | * Return the approximate number of pages we can send to this target. | |
512 | * There's no guarantee that you will be able to fit this number of pages | |
513 | * into a bio, it does not account for dynamic restrictions that vary | |
514 | * on offset. | |
515 | */ | |
516 | int bio_get_nr_vecs(struct block_device *bdev) | |
517 | { | |
165125e1 | 518 | struct request_queue *q = bdev_get_queue(bdev); |
f908ee94 BS |
519 | int nr_pages; |
520 | ||
521 | nr_pages = min_t(unsigned, | |
5abebfdd KO |
522 | queue_max_segments(q), |
523 | queue_max_sectors(q) / (PAGE_SIZE >> 9) + 1); | |
f908ee94 BS |
524 | |
525 | return min_t(unsigned, nr_pages, BIO_MAX_PAGES); | |
526 | ||
1da177e4 | 527 | } |
a112a71d | 528 | EXPORT_SYMBOL(bio_get_nr_vecs); |
1da177e4 | 529 | |
165125e1 | 530 | static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page |
defd94b7 MC |
531 | *page, unsigned int len, unsigned int offset, |
532 | unsigned short max_sectors) | |
1da177e4 LT |
533 | { |
534 | int retried_segments = 0; | |
535 | struct bio_vec *bvec; | |
536 | ||
537 | /* | |
538 | * cloned bio must not modify vec list | |
539 | */ | |
540 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
541 | return 0; | |
542 | ||
80cfd548 | 543 | if (((bio->bi_size + len) >> 9) > max_sectors) |
1da177e4 LT |
544 | return 0; |
545 | ||
80cfd548 JA |
546 | /* |
547 | * For filesystems with a blocksize smaller than the pagesize | |
548 | * we will often be called with the same page as last time and | |
549 | * a consecutive offset. Optimize this special case. | |
550 | */ | |
551 | if (bio->bi_vcnt > 0) { | |
552 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
553 | ||
554 | if (page == prev->bv_page && | |
555 | offset == prev->bv_offset + prev->bv_len) { | |
1d616585 | 556 | unsigned int prev_bv_len = prev->bv_len; |
80cfd548 | 557 | prev->bv_len += len; |
cc371e66 AK |
558 | |
559 | if (q->merge_bvec_fn) { | |
560 | struct bvec_merge_data bvm = { | |
1d616585 DM |
561 | /* prev_bvec is already charged in |
562 | bi_size, discharge it in order to | |
563 | simulate merging updated prev_bvec | |
564 | as new bvec. */ | |
cc371e66 AK |
565 | .bi_bdev = bio->bi_bdev, |
566 | .bi_sector = bio->bi_sector, | |
1d616585 | 567 | .bi_size = bio->bi_size - prev_bv_len, |
cc371e66 AK |
568 | .bi_rw = bio->bi_rw, |
569 | }; | |
570 | ||
8bf8c376 | 571 | if (q->merge_bvec_fn(q, &bvm, prev) < prev->bv_len) { |
cc371e66 AK |
572 | prev->bv_len -= len; |
573 | return 0; | |
574 | } | |
80cfd548 JA |
575 | } |
576 | ||
577 | goto done; | |
578 | } | |
579 | } | |
580 | ||
581 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
582 | return 0; |
583 | ||
584 | /* | |
585 | * we might lose a segment or two here, but rather that than | |
586 | * make this too complex. | |
587 | */ | |
588 | ||
8a78362c | 589 | while (bio->bi_phys_segments >= queue_max_segments(q)) { |
1da177e4 LT |
590 | |
591 | if (retried_segments) | |
592 | return 0; | |
593 | ||
594 | retried_segments = 1; | |
595 | blk_recount_segments(q, bio); | |
596 | } | |
597 | ||
598 | /* | |
599 | * setup the new entry, we might clear it again later if we | |
600 | * cannot add the page | |
601 | */ | |
602 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
603 | bvec->bv_page = page; | |
604 | bvec->bv_len = len; | |
605 | bvec->bv_offset = offset; | |
606 | ||
607 | /* | |
608 | * if queue has other restrictions (eg varying max sector size | |
609 | * depending on offset), it can specify a merge_bvec_fn in the | |
610 | * queue to get further control | |
611 | */ | |
612 | if (q->merge_bvec_fn) { | |
cc371e66 AK |
613 | struct bvec_merge_data bvm = { |
614 | .bi_bdev = bio->bi_bdev, | |
615 | .bi_sector = bio->bi_sector, | |
616 | .bi_size = bio->bi_size, | |
617 | .bi_rw = bio->bi_rw, | |
618 | }; | |
619 | ||
1da177e4 LT |
620 | /* |
621 | * merge_bvec_fn() returns number of bytes it can accept | |
622 | * at this offset | |
623 | */ | |
8bf8c376 | 624 | if (q->merge_bvec_fn(q, &bvm, bvec) < bvec->bv_len) { |
1da177e4 LT |
625 | bvec->bv_page = NULL; |
626 | bvec->bv_len = 0; | |
627 | bvec->bv_offset = 0; | |
628 | return 0; | |
629 | } | |
630 | } | |
631 | ||
632 | /* If we may be able to merge these biovecs, force a recount */ | |
b8b3e16c | 633 | if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
1da177e4 LT |
634 | bio->bi_flags &= ~(1 << BIO_SEG_VALID); |
635 | ||
636 | bio->bi_vcnt++; | |
637 | bio->bi_phys_segments++; | |
80cfd548 | 638 | done: |
1da177e4 LT |
639 | bio->bi_size += len; |
640 | return len; | |
641 | } | |
642 | ||
6e68af66 MC |
643 | /** |
644 | * bio_add_pc_page - attempt to add page to bio | |
fddfdeaf | 645 | * @q: the target queue |
6e68af66 MC |
646 | * @bio: destination bio |
647 | * @page: page to add | |
648 | * @len: vec entry length | |
649 | * @offset: vec entry offset | |
650 | * | |
651 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
c6428084 AG |
652 | * number of reasons, such as the bio being full or target block device |
653 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
654 | * so it is always possible to add a single page to an empty bio. | |
655 | * | |
656 | * This should only be used by REQ_PC bios. | |
6e68af66 | 657 | */ |
165125e1 | 658 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page, |
6e68af66 MC |
659 | unsigned int len, unsigned int offset) |
660 | { | |
ae03bf63 MP |
661 | return __bio_add_page(q, bio, page, len, offset, |
662 | queue_max_hw_sectors(q)); | |
6e68af66 | 663 | } |
a112a71d | 664 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 665 | |
1da177e4 LT |
666 | /** |
667 | * bio_add_page - attempt to add page to bio | |
668 | * @bio: destination bio | |
669 | * @page: page to add | |
670 | * @len: vec entry length | |
671 | * @offset: vec entry offset | |
672 | * | |
673 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
c6428084 AG |
674 | * number of reasons, such as the bio being full or target block device |
675 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
676 | * so it is always possible to add a single page to an empty bio. | |
1da177e4 LT |
677 | */ |
678 | int bio_add_page(struct bio *bio, struct page *page, unsigned int len, | |
679 | unsigned int offset) | |
680 | { | |
defd94b7 | 681 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
ae03bf63 | 682 | return __bio_add_page(q, bio, page, len, offset, queue_max_sectors(q)); |
1da177e4 | 683 | } |
a112a71d | 684 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 LT |
685 | |
686 | struct bio_map_data { | |
687 | struct bio_vec *iovecs; | |
c5dec1c3 | 688 | struct sg_iovec *sgvecs; |
152e283f FT |
689 | int nr_sgvecs; |
690 | int is_our_pages; | |
1da177e4 LT |
691 | }; |
692 | ||
c5dec1c3 | 693 | static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio, |
152e283f FT |
694 | struct sg_iovec *iov, int iov_count, |
695 | int is_our_pages) | |
1da177e4 LT |
696 | { |
697 | memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt); | |
c5dec1c3 FT |
698 | memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count); |
699 | bmd->nr_sgvecs = iov_count; | |
152e283f | 700 | bmd->is_our_pages = is_our_pages; |
1da177e4 LT |
701 | bio->bi_private = bmd; |
702 | } | |
703 | ||
704 | static void bio_free_map_data(struct bio_map_data *bmd) | |
705 | { | |
706 | kfree(bmd->iovecs); | |
c5dec1c3 | 707 | kfree(bmd->sgvecs); |
1da177e4 LT |
708 | kfree(bmd); |
709 | } | |
710 | ||
121f0994 DC |
711 | static struct bio_map_data *bio_alloc_map_data(int nr_segs, |
712 | unsigned int iov_count, | |
76029ff3 | 713 | gfp_t gfp_mask) |
1da177e4 | 714 | { |
f3f63c1c JA |
715 | struct bio_map_data *bmd; |
716 | ||
717 | if (iov_count > UIO_MAXIOV) | |
718 | return NULL; | |
1da177e4 | 719 | |
f3f63c1c | 720 | bmd = kmalloc(sizeof(*bmd), gfp_mask); |
1da177e4 LT |
721 | if (!bmd) |
722 | return NULL; | |
723 | ||
76029ff3 | 724 | bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask); |
c5dec1c3 FT |
725 | if (!bmd->iovecs) { |
726 | kfree(bmd); | |
727 | return NULL; | |
728 | } | |
729 | ||
76029ff3 | 730 | bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask); |
c5dec1c3 | 731 | if (bmd->sgvecs) |
1da177e4 LT |
732 | return bmd; |
733 | ||
c5dec1c3 | 734 | kfree(bmd->iovecs); |
1da177e4 LT |
735 | kfree(bmd); |
736 | return NULL; | |
737 | } | |
738 | ||
aefcc28a | 739 | static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs, |
ecb554a8 FT |
740 | struct sg_iovec *iov, int iov_count, |
741 | int to_user, int from_user, int do_free_page) | |
c5dec1c3 FT |
742 | { |
743 | int ret = 0, i; | |
744 | struct bio_vec *bvec; | |
745 | int iov_idx = 0; | |
746 | unsigned int iov_off = 0; | |
c5dec1c3 FT |
747 | |
748 | __bio_for_each_segment(bvec, bio, i, 0) { | |
749 | char *bv_addr = page_address(bvec->bv_page); | |
aefcc28a | 750 | unsigned int bv_len = iovecs[i].bv_len; |
c5dec1c3 FT |
751 | |
752 | while (bv_len && iov_idx < iov_count) { | |
753 | unsigned int bytes; | |
0e0c6212 | 754 | char __user *iov_addr; |
c5dec1c3 FT |
755 | |
756 | bytes = min_t(unsigned int, | |
757 | iov[iov_idx].iov_len - iov_off, bv_len); | |
758 | iov_addr = iov[iov_idx].iov_base + iov_off; | |
759 | ||
760 | if (!ret) { | |
ecb554a8 | 761 | if (to_user) |
c5dec1c3 FT |
762 | ret = copy_to_user(iov_addr, bv_addr, |
763 | bytes); | |
764 | ||
ecb554a8 FT |
765 | if (from_user) |
766 | ret = copy_from_user(bv_addr, iov_addr, | |
767 | bytes); | |
768 | ||
c5dec1c3 FT |
769 | if (ret) |
770 | ret = -EFAULT; | |
771 | } | |
772 | ||
773 | bv_len -= bytes; | |
774 | bv_addr += bytes; | |
775 | iov_addr += bytes; | |
776 | iov_off += bytes; | |
777 | ||
778 | if (iov[iov_idx].iov_len == iov_off) { | |
779 | iov_idx++; | |
780 | iov_off = 0; | |
781 | } | |
782 | } | |
783 | ||
152e283f | 784 | if (do_free_page) |
c5dec1c3 FT |
785 | __free_page(bvec->bv_page); |
786 | } | |
787 | ||
788 | return ret; | |
789 | } | |
790 | ||
1da177e4 LT |
791 | /** |
792 | * bio_uncopy_user - finish previously mapped bio | |
793 | * @bio: bio being terminated | |
794 | * | |
795 | * Free pages allocated from bio_copy_user() and write back data | |
796 | * to user space in case of a read. | |
797 | */ | |
798 | int bio_uncopy_user(struct bio *bio) | |
799 | { | |
800 | struct bio_map_data *bmd = bio->bi_private; | |
81882766 | 801 | int ret = 0; |
1da177e4 | 802 | |
81882766 FT |
803 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) |
804 | ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs, | |
ecb554a8 FT |
805 | bmd->nr_sgvecs, bio_data_dir(bio) == READ, |
806 | 0, bmd->is_our_pages); | |
1da177e4 LT |
807 | bio_free_map_data(bmd); |
808 | bio_put(bio); | |
809 | return ret; | |
810 | } | |
a112a71d | 811 | EXPORT_SYMBOL(bio_uncopy_user); |
1da177e4 LT |
812 | |
813 | /** | |
c5dec1c3 | 814 | * bio_copy_user_iov - copy user data to bio |
1da177e4 | 815 | * @q: destination block queue |
152e283f | 816 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
817 | * @iov: the iovec. |
818 | * @iov_count: number of elements in the iovec | |
1da177e4 | 819 | * @write_to_vm: bool indicating writing to pages or not |
a3bce90e | 820 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
821 | * |
822 | * Prepares and returns a bio for indirect user io, bouncing data | |
823 | * to/from kernel pages as necessary. Must be paired with | |
824 | * call bio_uncopy_user() on io completion. | |
825 | */ | |
152e283f FT |
826 | struct bio *bio_copy_user_iov(struct request_queue *q, |
827 | struct rq_map_data *map_data, | |
828 | struct sg_iovec *iov, int iov_count, | |
829 | int write_to_vm, gfp_t gfp_mask) | |
1da177e4 | 830 | { |
1da177e4 LT |
831 | struct bio_map_data *bmd; |
832 | struct bio_vec *bvec; | |
833 | struct page *page; | |
834 | struct bio *bio; | |
835 | int i, ret; | |
c5dec1c3 FT |
836 | int nr_pages = 0; |
837 | unsigned int len = 0; | |
56c451f4 | 838 | unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0; |
1da177e4 | 839 | |
c5dec1c3 FT |
840 | for (i = 0; i < iov_count; i++) { |
841 | unsigned long uaddr; | |
842 | unsigned long end; | |
843 | unsigned long start; | |
844 | ||
845 | uaddr = (unsigned long)iov[i].iov_base; | |
846 | end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
847 | start = uaddr >> PAGE_SHIFT; | |
848 | ||
cb4644ca JA |
849 | /* |
850 | * Overflow, abort | |
851 | */ | |
852 | if (end < start) | |
853 | return ERR_PTR(-EINVAL); | |
854 | ||
c5dec1c3 FT |
855 | nr_pages += end - start; |
856 | len += iov[i].iov_len; | |
857 | } | |
858 | ||
69838727 FT |
859 | if (offset) |
860 | nr_pages++; | |
861 | ||
a3bce90e | 862 | bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask); |
1da177e4 LT |
863 | if (!bmd) |
864 | return ERR_PTR(-ENOMEM); | |
865 | ||
1da177e4 | 866 | ret = -ENOMEM; |
a9e9dc24 | 867 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
868 | if (!bio) |
869 | goto out_bmd; | |
870 | ||
7b6d91da CH |
871 | if (!write_to_vm) |
872 | bio->bi_rw |= REQ_WRITE; | |
1da177e4 LT |
873 | |
874 | ret = 0; | |
56c451f4 FT |
875 | |
876 | if (map_data) { | |
e623ddb4 | 877 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
878 | i = map_data->offset / PAGE_SIZE; |
879 | } | |
1da177e4 | 880 | while (len) { |
e623ddb4 | 881 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 882 | |
56c451f4 FT |
883 | bytes -= offset; |
884 | ||
1da177e4 LT |
885 | if (bytes > len) |
886 | bytes = len; | |
887 | ||
152e283f | 888 | if (map_data) { |
e623ddb4 | 889 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
890 | ret = -ENOMEM; |
891 | break; | |
892 | } | |
e623ddb4 FT |
893 | |
894 | page = map_data->pages[i / nr_pages]; | |
895 | page += (i % nr_pages); | |
896 | ||
897 | i++; | |
898 | } else { | |
152e283f | 899 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
900 | if (!page) { |
901 | ret = -ENOMEM; | |
902 | break; | |
903 | } | |
1da177e4 LT |
904 | } |
905 | ||
56c451f4 | 906 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 907 | break; |
1da177e4 LT |
908 | |
909 | len -= bytes; | |
56c451f4 | 910 | offset = 0; |
1da177e4 LT |
911 | } |
912 | ||
913 | if (ret) | |
914 | goto cleanup; | |
915 | ||
916 | /* | |
917 | * success | |
918 | */ | |
ecb554a8 FT |
919 | if ((!write_to_vm && (!map_data || !map_data->null_mapped)) || |
920 | (map_data && map_data->from_user)) { | |
921 | ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 1, 0); | |
c5dec1c3 FT |
922 | if (ret) |
923 | goto cleanup; | |
1da177e4 LT |
924 | } |
925 | ||
152e283f | 926 | bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1); |
1da177e4 LT |
927 | return bio; |
928 | cleanup: | |
152e283f FT |
929 | if (!map_data) |
930 | bio_for_each_segment(bvec, bio, i) | |
931 | __free_page(bvec->bv_page); | |
1da177e4 LT |
932 | |
933 | bio_put(bio); | |
934 | out_bmd: | |
935 | bio_free_map_data(bmd); | |
936 | return ERR_PTR(ret); | |
937 | } | |
938 | ||
c5dec1c3 FT |
939 | /** |
940 | * bio_copy_user - copy user data to bio | |
941 | * @q: destination block queue | |
152e283f | 942 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
943 | * @uaddr: start of user address |
944 | * @len: length in bytes | |
945 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 946 | * @gfp_mask: memory allocation flags |
c5dec1c3 FT |
947 | * |
948 | * Prepares and returns a bio for indirect user io, bouncing data | |
949 | * to/from kernel pages as necessary. Must be paired with | |
950 | * call bio_uncopy_user() on io completion. | |
951 | */ | |
152e283f FT |
952 | struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data, |
953 | unsigned long uaddr, unsigned int len, | |
954 | int write_to_vm, gfp_t gfp_mask) | |
c5dec1c3 FT |
955 | { |
956 | struct sg_iovec iov; | |
957 | ||
958 | iov.iov_base = (void __user *)uaddr; | |
959 | iov.iov_len = len; | |
960 | ||
152e283f | 961 | return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask); |
c5dec1c3 | 962 | } |
a112a71d | 963 | EXPORT_SYMBOL(bio_copy_user); |
c5dec1c3 | 964 | |
165125e1 | 965 | static struct bio *__bio_map_user_iov(struct request_queue *q, |
f1970baf JB |
966 | struct block_device *bdev, |
967 | struct sg_iovec *iov, int iov_count, | |
a3bce90e | 968 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 | 969 | { |
f1970baf JB |
970 | int i, j; |
971 | int nr_pages = 0; | |
1da177e4 LT |
972 | struct page **pages; |
973 | struct bio *bio; | |
f1970baf JB |
974 | int cur_page = 0; |
975 | int ret, offset; | |
1da177e4 | 976 | |
f1970baf JB |
977 | for (i = 0; i < iov_count; i++) { |
978 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
979 | unsigned long len = iov[i].iov_len; | |
980 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
981 | unsigned long start = uaddr >> PAGE_SHIFT; | |
982 | ||
cb4644ca JA |
983 | /* |
984 | * Overflow, abort | |
985 | */ | |
986 | if (end < start) | |
987 | return ERR_PTR(-EINVAL); | |
988 | ||
f1970baf JB |
989 | nr_pages += end - start; |
990 | /* | |
ad2d7225 | 991 | * buffer must be aligned to at least hardsector size for now |
f1970baf | 992 | */ |
ad2d7225 | 993 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
994 | return ERR_PTR(-EINVAL); |
995 | } | |
996 | ||
997 | if (!nr_pages) | |
1da177e4 LT |
998 | return ERR_PTR(-EINVAL); |
999 | ||
a9e9dc24 | 1000 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1001 | if (!bio) |
1002 | return ERR_PTR(-ENOMEM); | |
1003 | ||
1004 | ret = -ENOMEM; | |
a3bce90e | 1005 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
1006 | if (!pages) |
1007 | goto out; | |
1008 | ||
f1970baf JB |
1009 | for (i = 0; i < iov_count; i++) { |
1010 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
1011 | unsigned long len = iov[i].iov_len; | |
1012 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1013 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1014 | const int local_nr_pages = end - start; | |
1015 | const int page_limit = cur_page + local_nr_pages; | |
cb4644ca | 1016 | |
f5dd33c4 NP |
1017 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
1018 | write_to_vm, &pages[cur_page]); | |
99172157 JA |
1019 | if (ret < local_nr_pages) { |
1020 | ret = -EFAULT; | |
f1970baf | 1021 | goto out_unmap; |
99172157 | 1022 | } |
f1970baf JB |
1023 | |
1024 | offset = uaddr & ~PAGE_MASK; | |
1025 | for (j = cur_page; j < page_limit; j++) { | |
1026 | unsigned int bytes = PAGE_SIZE - offset; | |
1027 | ||
1028 | if (len <= 0) | |
1029 | break; | |
1030 | ||
1031 | if (bytes > len) | |
1032 | bytes = len; | |
1033 | ||
1034 | /* | |
1035 | * sorry... | |
1036 | */ | |
defd94b7 MC |
1037 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1038 | bytes) | |
f1970baf JB |
1039 | break; |
1040 | ||
1041 | len -= bytes; | |
1042 | offset = 0; | |
1043 | } | |
1da177e4 | 1044 | |
f1970baf | 1045 | cur_page = j; |
1da177e4 | 1046 | /* |
f1970baf | 1047 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1048 | */ |
f1970baf JB |
1049 | while (j < page_limit) |
1050 | page_cache_release(pages[j++]); | |
1da177e4 LT |
1051 | } |
1052 | ||
1da177e4 LT |
1053 | kfree(pages); |
1054 | ||
1055 | /* | |
1056 | * set data direction, and check if mapped pages need bouncing | |
1057 | */ | |
1058 | if (!write_to_vm) | |
7b6d91da | 1059 | bio->bi_rw |= REQ_WRITE; |
1da177e4 | 1060 | |
f1970baf | 1061 | bio->bi_bdev = bdev; |
1da177e4 LT |
1062 | bio->bi_flags |= (1 << BIO_USER_MAPPED); |
1063 | return bio; | |
f1970baf JB |
1064 | |
1065 | out_unmap: | |
1066 | for (i = 0; i < nr_pages; i++) { | |
1067 | if(!pages[i]) | |
1068 | break; | |
1069 | page_cache_release(pages[i]); | |
1070 | } | |
1071 | out: | |
1da177e4 LT |
1072 | kfree(pages); |
1073 | bio_put(bio); | |
1074 | return ERR_PTR(ret); | |
1075 | } | |
1076 | ||
1077 | /** | |
1078 | * bio_map_user - map user address into bio | |
165125e1 | 1079 | * @q: the struct request_queue for the bio |
1da177e4 LT |
1080 | * @bdev: destination block device |
1081 | * @uaddr: start of user address | |
1082 | * @len: length in bytes | |
1083 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1084 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
1085 | * |
1086 | * Map the user space address into a bio suitable for io to a block | |
1087 | * device. Returns an error pointer in case of error. | |
1088 | */ | |
165125e1 | 1089 | struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev, |
a3bce90e FT |
1090 | unsigned long uaddr, unsigned int len, int write_to_vm, |
1091 | gfp_t gfp_mask) | |
f1970baf JB |
1092 | { |
1093 | struct sg_iovec iov; | |
1094 | ||
3f70353e | 1095 | iov.iov_base = (void __user *)uaddr; |
f1970baf JB |
1096 | iov.iov_len = len; |
1097 | ||
a3bce90e | 1098 | return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask); |
f1970baf | 1099 | } |
a112a71d | 1100 | EXPORT_SYMBOL(bio_map_user); |
f1970baf JB |
1101 | |
1102 | /** | |
1103 | * bio_map_user_iov - map user sg_iovec table into bio | |
165125e1 | 1104 | * @q: the struct request_queue for the bio |
f1970baf JB |
1105 | * @bdev: destination block device |
1106 | * @iov: the iovec. | |
1107 | * @iov_count: number of elements in the iovec | |
1108 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1109 | * @gfp_mask: memory allocation flags |
f1970baf JB |
1110 | * |
1111 | * Map the user space address into a bio suitable for io to a block | |
1112 | * device. Returns an error pointer in case of error. | |
1113 | */ | |
165125e1 | 1114 | struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev, |
f1970baf | 1115 | struct sg_iovec *iov, int iov_count, |
a3bce90e | 1116 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 LT |
1117 | { |
1118 | struct bio *bio; | |
1119 | ||
a3bce90e FT |
1120 | bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm, |
1121 | gfp_mask); | |
1da177e4 LT |
1122 | if (IS_ERR(bio)) |
1123 | return bio; | |
1124 | ||
1125 | /* | |
1126 | * subtle -- if __bio_map_user() ended up bouncing a bio, | |
1127 | * it would normally disappear when its bi_end_io is run. | |
1128 | * however, we need it for the unmap, so grab an extra | |
1129 | * reference to it | |
1130 | */ | |
1131 | bio_get(bio); | |
1132 | ||
0e75f906 | 1133 | return bio; |
1da177e4 LT |
1134 | } |
1135 | ||
1136 | static void __bio_unmap_user(struct bio *bio) | |
1137 | { | |
1138 | struct bio_vec *bvec; | |
1139 | int i; | |
1140 | ||
1141 | /* | |
1142 | * make sure we dirty pages we wrote to | |
1143 | */ | |
1144 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1145 | if (bio_data_dir(bio) == READ) | |
1146 | set_page_dirty_lock(bvec->bv_page); | |
1147 | ||
1148 | page_cache_release(bvec->bv_page); | |
1149 | } | |
1150 | ||
1151 | bio_put(bio); | |
1152 | } | |
1153 | ||
1154 | /** | |
1155 | * bio_unmap_user - unmap a bio | |
1156 | * @bio: the bio being unmapped | |
1157 | * | |
1158 | * Unmap a bio previously mapped by bio_map_user(). Must be called with | |
1159 | * a process context. | |
1160 | * | |
1161 | * bio_unmap_user() may sleep. | |
1162 | */ | |
1163 | void bio_unmap_user(struct bio *bio) | |
1164 | { | |
1165 | __bio_unmap_user(bio); | |
1166 | bio_put(bio); | |
1167 | } | |
a112a71d | 1168 | EXPORT_SYMBOL(bio_unmap_user); |
1da177e4 | 1169 | |
6712ecf8 | 1170 | static void bio_map_kern_endio(struct bio *bio, int err) |
b823825e | 1171 | { |
b823825e | 1172 | bio_put(bio); |
b823825e JA |
1173 | } |
1174 | ||
165125e1 | 1175 | static struct bio *__bio_map_kern(struct request_queue *q, void *data, |
27496a8c | 1176 | unsigned int len, gfp_t gfp_mask) |
df46b9a4 MC |
1177 | { |
1178 | unsigned long kaddr = (unsigned long)data; | |
1179 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1180 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1181 | const int nr_pages = end - start; | |
1182 | int offset, i; | |
1183 | struct bio *bio; | |
1184 | ||
a9e9dc24 | 1185 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1186 | if (!bio) |
1187 | return ERR_PTR(-ENOMEM); | |
1188 | ||
1189 | offset = offset_in_page(kaddr); | |
1190 | for (i = 0; i < nr_pages; i++) { | |
1191 | unsigned int bytes = PAGE_SIZE - offset; | |
1192 | ||
1193 | if (len <= 0) | |
1194 | break; | |
1195 | ||
1196 | if (bytes > len) | |
1197 | bytes = len; | |
1198 | ||
defd94b7 MC |
1199 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
1200 | offset) < bytes) | |
df46b9a4 MC |
1201 | break; |
1202 | ||
1203 | data += bytes; | |
1204 | len -= bytes; | |
1205 | offset = 0; | |
1206 | } | |
1207 | ||
b823825e | 1208 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1209 | return bio; |
1210 | } | |
1211 | ||
1212 | /** | |
1213 | * bio_map_kern - map kernel address into bio | |
165125e1 | 1214 | * @q: the struct request_queue for the bio |
df46b9a4 MC |
1215 | * @data: pointer to buffer to map |
1216 | * @len: length in bytes | |
1217 | * @gfp_mask: allocation flags for bio allocation | |
1218 | * | |
1219 | * Map the kernel address into a bio suitable for io to a block | |
1220 | * device. Returns an error pointer in case of error. | |
1221 | */ | |
165125e1 | 1222 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
27496a8c | 1223 | gfp_t gfp_mask) |
df46b9a4 MC |
1224 | { |
1225 | struct bio *bio; | |
1226 | ||
1227 | bio = __bio_map_kern(q, data, len, gfp_mask); | |
1228 | if (IS_ERR(bio)) | |
1229 | return bio; | |
1230 | ||
1231 | if (bio->bi_size == len) | |
1232 | return bio; | |
1233 | ||
1234 | /* | |
1235 | * Don't support partial mappings. | |
1236 | */ | |
1237 | bio_put(bio); | |
1238 | return ERR_PTR(-EINVAL); | |
1239 | } | |
a112a71d | 1240 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1241 | |
68154e90 FT |
1242 | static void bio_copy_kern_endio(struct bio *bio, int err) |
1243 | { | |
1244 | struct bio_vec *bvec; | |
1245 | const int read = bio_data_dir(bio) == READ; | |
76029ff3 | 1246 | struct bio_map_data *bmd = bio->bi_private; |
68154e90 | 1247 | int i; |
76029ff3 | 1248 | char *p = bmd->sgvecs[0].iov_base; |
68154e90 FT |
1249 | |
1250 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1251 | char *addr = page_address(bvec->bv_page); | |
76029ff3 | 1252 | int len = bmd->iovecs[i].bv_len; |
68154e90 | 1253 | |
4fc981ef | 1254 | if (read) |
76029ff3 | 1255 | memcpy(p, addr, len); |
68154e90 FT |
1256 | |
1257 | __free_page(bvec->bv_page); | |
76029ff3 | 1258 | p += len; |
68154e90 FT |
1259 | } |
1260 | ||
76029ff3 | 1261 | bio_free_map_data(bmd); |
68154e90 FT |
1262 | bio_put(bio); |
1263 | } | |
1264 | ||
1265 | /** | |
1266 | * bio_copy_kern - copy kernel address into bio | |
1267 | * @q: the struct request_queue for the bio | |
1268 | * @data: pointer to buffer to copy | |
1269 | * @len: length in bytes | |
1270 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1271 | * @reading: data direction is READ |
68154e90 FT |
1272 | * |
1273 | * copy the kernel address into a bio suitable for io to a block | |
1274 | * device. Returns an error pointer in case of error. | |
1275 | */ | |
1276 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1277 | gfp_t gfp_mask, int reading) | |
1278 | { | |
68154e90 FT |
1279 | struct bio *bio; |
1280 | struct bio_vec *bvec; | |
4d8ab62e | 1281 | int i; |
68154e90 | 1282 | |
4d8ab62e FT |
1283 | bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask); |
1284 | if (IS_ERR(bio)) | |
1285 | return bio; | |
68154e90 FT |
1286 | |
1287 | if (!reading) { | |
1288 | void *p = data; | |
1289 | ||
1290 | bio_for_each_segment(bvec, bio, i) { | |
1291 | char *addr = page_address(bvec->bv_page); | |
1292 | ||
1293 | memcpy(addr, p, bvec->bv_len); | |
1294 | p += bvec->bv_len; | |
1295 | } | |
1296 | } | |
1297 | ||
68154e90 | 1298 | bio->bi_end_io = bio_copy_kern_endio; |
76029ff3 | 1299 | |
68154e90 | 1300 | return bio; |
68154e90 | 1301 | } |
a112a71d | 1302 | EXPORT_SYMBOL(bio_copy_kern); |
68154e90 | 1303 | |
1da177e4 LT |
1304 | /* |
1305 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1306 | * for performing direct-IO in BIOs. | |
1307 | * | |
1308 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1309 | * because the required locks are not interrupt-safe. So what we can do is to | |
1310 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1311 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1312 | * in process context. | |
1313 | * | |
1314 | * We special-case compound pages here: normally this means reads into hugetlb | |
1315 | * pages. The logic in here doesn't really work right for compound pages | |
1316 | * because the VM does not uniformly chase down the head page in all cases. | |
1317 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1318 | * handle them at all. So we skip compound pages here at an early stage. | |
1319 | * | |
1320 | * Note that this code is very hard to test under normal circumstances because | |
1321 | * direct-io pins the pages with get_user_pages(). This makes | |
1322 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1323 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1324 | * pagecache. |
1325 | * | |
1326 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1327 | * deferred bio dirtying paths. | |
1328 | */ | |
1329 | ||
1330 | /* | |
1331 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1332 | */ | |
1333 | void bio_set_pages_dirty(struct bio *bio) | |
1334 | { | |
1335 | struct bio_vec *bvec = bio->bi_io_vec; | |
1336 | int i; | |
1337 | ||
1338 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1339 | struct page *page = bvec[i].bv_page; | |
1340 | ||
1341 | if (page && !PageCompound(page)) | |
1342 | set_page_dirty_lock(page); | |
1343 | } | |
1344 | } | |
1345 | ||
86b6c7a7 | 1346 | static void bio_release_pages(struct bio *bio) |
1da177e4 LT |
1347 | { |
1348 | struct bio_vec *bvec = bio->bi_io_vec; | |
1349 | int i; | |
1350 | ||
1351 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1352 | struct page *page = bvec[i].bv_page; | |
1353 | ||
1354 | if (page) | |
1355 | put_page(page); | |
1356 | } | |
1357 | } | |
1358 | ||
1359 | /* | |
1360 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1361 | * If they are, then fine. If, however, some pages are clean then they must | |
1362 | * have been written out during the direct-IO read. So we take another ref on | |
1363 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1364 | * | |
1365 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
1366 | * here on. It will run one page_cache_release() against each page and will | |
1367 | * run one bio_put() against the BIO. | |
1368 | */ | |
1369 | ||
65f27f38 | 1370 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1371 | |
65f27f38 | 1372 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1373 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1374 | static struct bio *bio_dirty_list; | |
1375 | ||
1376 | /* | |
1377 | * This runs in process context | |
1378 | */ | |
65f27f38 | 1379 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1380 | { |
1381 | unsigned long flags; | |
1382 | struct bio *bio; | |
1383 | ||
1384 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1385 | bio = bio_dirty_list; | |
1386 | bio_dirty_list = NULL; | |
1387 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1388 | ||
1389 | while (bio) { | |
1390 | struct bio *next = bio->bi_private; | |
1391 | ||
1392 | bio_set_pages_dirty(bio); | |
1393 | bio_release_pages(bio); | |
1394 | bio_put(bio); | |
1395 | bio = next; | |
1396 | } | |
1397 | } | |
1398 | ||
1399 | void bio_check_pages_dirty(struct bio *bio) | |
1400 | { | |
1401 | struct bio_vec *bvec = bio->bi_io_vec; | |
1402 | int nr_clean_pages = 0; | |
1403 | int i; | |
1404 | ||
1405 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1406 | struct page *page = bvec[i].bv_page; | |
1407 | ||
1408 | if (PageDirty(page) || PageCompound(page)) { | |
1409 | page_cache_release(page); | |
1410 | bvec[i].bv_page = NULL; | |
1411 | } else { | |
1412 | nr_clean_pages++; | |
1413 | } | |
1414 | } | |
1415 | ||
1416 | if (nr_clean_pages) { | |
1417 | unsigned long flags; | |
1418 | ||
1419 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1420 | bio->bi_private = bio_dirty_list; | |
1421 | bio_dirty_list = bio; | |
1422 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1423 | schedule_work(&bio_dirty_work); | |
1424 | } else { | |
1425 | bio_put(bio); | |
1426 | } | |
1427 | } | |
1428 | ||
2d4dc890 IL |
1429 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1430 | void bio_flush_dcache_pages(struct bio *bi) | |
1431 | { | |
1432 | int i; | |
1433 | struct bio_vec *bvec; | |
1434 | ||
1435 | bio_for_each_segment(bvec, bi, i) | |
1436 | flush_dcache_page(bvec->bv_page); | |
1437 | } | |
1438 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1439 | #endif | |
1440 | ||
1da177e4 LT |
1441 | /** |
1442 | * bio_endio - end I/O on a bio | |
1443 | * @bio: bio | |
1da177e4 LT |
1444 | * @error: error, if any |
1445 | * | |
1446 | * Description: | |
6712ecf8 | 1447 | * bio_endio() will end I/O on the whole bio. bio_endio() is the |
5bb23a68 N |
1448 | * preferred way to end I/O on a bio, it takes care of clearing |
1449 | * BIO_UPTODATE on error. @error is 0 on success, and and one of the | |
1450 | * established -Exxxx (-EIO, for instance) error values in case | |
25985edc | 1451 | * something went wrong. No one should call bi_end_io() directly on a |
5bb23a68 N |
1452 | * bio unless they own it and thus know that it has an end_io |
1453 | * function. | |
1da177e4 | 1454 | **/ |
6712ecf8 | 1455 | void bio_endio(struct bio *bio, int error) |
1da177e4 LT |
1456 | { |
1457 | if (error) | |
1458 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
9cc54d40 N |
1459 | else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) |
1460 | error = -EIO; | |
1da177e4 | 1461 | |
5bb23a68 | 1462 | if (bio->bi_end_io) |
6712ecf8 | 1463 | bio->bi_end_io(bio, error); |
1da177e4 | 1464 | } |
a112a71d | 1465 | EXPORT_SYMBOL(bio_endio); |
1da177e4 LT |
1466 | |
1467 | void bio_pair_release(struct bio_pair *bp) | |
1468 | { | |
1469 | if (atomic_dec_and_test(&bp->cnt)) { | |
1470 | struct bio *master = bp->bio1.bi_private; | |
1471 | ||
6712ecf8 | 1472 | bio_endio(master, bp->error); |
1da177e4 LT |
1473 | mempool_free(bp, bp->bio2.bi_private); |
1474 | } | |
1475 | } | |
a112a71d | 1476 | EXPORT_SYMBOL(bio_pair_release); |
1da177e4 | 1477 | |
6712ecf8 | 1478 | static void bio_pair_end_1(struct bio *bi, int err) |
1da177e4 LT |
1479 | { |
1480 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio1); | |
1481 | ||
1482 | if (err) | |
1483 | bp->error = err; | |
1484 | ||
1da177e4 | 1485 | bio_pair_release(bp); |
1da177e4 LT |
1486 | } |
1487 | ||
6712ecf8 | 1488 | static void bio_pair_end_2(struct bio *bi, int err) |
1da177e4 LT |
1489 | { |
1490 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio2); | |
1491 | ||
1492 | if (err) | |
1493 | bp->error = err; | |
1494 | ||
1da177e4 | 1495 | bio_pair_release(bp); |
1da177e4 LT |
1496 | } |
1497 | ||
1498 | /* | |
c7eee1b8 | 1499 | * split a bio - only worry about a bio with a single page in its iovec |
1da177e4 | 1500 | */ |
6feef531 | 1501 | struct bio_pair *bio_split(struct bio *bi, int first_sectors) |
1da177e4 | 1502 | { |
6feef531 | 1503 | struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO); |
1da177e4 LT |
1504 | |
1505 | if (!bp) | |
1506 | return bp; | |
1507 | ||
5f3ea37c | 1508 | trace_block_split(bdev_get_queue(bi->bi_bdev), bi, |
2056a782 JA |
1509 | bi->bi_sector + first_sectors); |
1510 | ||
1da177e4 LT |
1511 | BUG_ON(bi->bi_vcnt != 1); |
1512 | BUG_ON(bi->bi_idx != 0); | |
1513 | atomic_set(&bp->cnt, 3); | |
1514 | bp->error = 0; | |
1515 | bp->bio1 = *bi; | |
1516 | bp->bio2 = *bi; | |
1517 | bp->bio2.bi_sector += first_sectors; | |
1518 | bp->bio2.bi_size -= first_sectors << 9; | |
1519 | bp->bio1.bi_size = first_sectors << 9; | |
1520 | ||
1521 | bp->bv1 = bi->bi_io_vec[0]; | |
1522 | bp->bv2 = bi->bi_io_vec[0]; | |
1523 | bp->bv2.bv_offset += first_sectors << 9; | |
1524 | bp->bv2.bv_len -= first_sectors << 9; | |
1525 | bp->bv1.bv_len = first_sectors << 9; | |
1526 | ||
1527 | bp->bio1.bi_io_vec = &bp->bv1; | |
1528 | bp->bio2.bi_io_vec = &bp->bv2; | |
1529 | ||
a2eb0c10 N |
1530 | bp->bio1.bi_max_vecs = 1; |
1531 | bp->bio2.bi_max_vecs = 1; | |
1532 | ||
1da177e4 LT |
1533 | bp->bio1.bi_end_io = bio_pair_end_1; |
1534 | bp->bio2.bi_end_io = bio_pair_end_2; | |
1535 | ||
1536 | bp->bio1.bi_private = bi; | |
6feef531 | 1537 | bp->bio2.bi_private = bio_split_pool; |
1da177e4 | 1538 | |
7ba1ba12 MP |
1539 | if (bio_integrity(bi)) |
1540 | bio_integrity_split(bi, bp, first_sectors); | |
1541 | ||
1da177e4 LT |
1542 | return bp; |
1543 | } | |
a112a71d | 1544 | EXPORT_SYMBOL(bio_split); |
1da177e4 | 1545 | |
ad3316bf MP |
1546 | /** |
1547 | * bio_sector_offset - Find hardware sector offset in bio | |
1548 | * @bio: bio to inspect | |
1549 | * @index: bio_vec index | |
1550 | * @offset: offset in bv_page | |
1551 | * | |
1552 | * Return the number of hardware sectors between beginning of bio | |
1553 | * and an end point indicated by a bio_vec index and an offset | |
1554 | * within that vector's page. | |
1555 | */ | |
1556 | sector_t bio_sector_offset(struct bio *bio, unsigned short index, | |
1557 | unsigned int offset) | |
1558 | { | |
e1defc4f | 1559 | unsigned int sector_sz; |
ad3316bf MP |
1560 | struct bio_vec *bv; |
1561 | sector_t sectors; | |
1562 | int i; | |
1563 | ||
e1defc4f | 1564 | sector_sz = queue_logical_block_size(bio->bi_bdev->bd_disk->queue); |
ad3316bf MP |
1565 | sectors = 0; |
1566 | ||
1567 | if (index >= bio->bi_idx) | |
1568 | index = bio->bi_vcnt - 1; | |
1569 | ||
1570 | __bio_for_each_segment(bv, bio, i, 0) { | |
1571 | if (i == index) { | |
1572 | if (offset > bv->bv_offset) | |
1573 | sectors += (offset - bv->bv_offset) / sector_sz; | |
1574 | break; | |
1575 | } | |
1576 | ||
1577 | sectors += bv->bv_len / sector_sz; | |
1578 | } | |
1579 | ||
1580 | return sectors; | |
1581 | } | |
1582 | EXPORT_SYMBOL(bio_sector_offset); | |
1da177e4 LT |
1583 | |
1584 | /* | |
1585 | * create memory pools for biovec's in a bio_set. | |
1586 | * use the global biovec slabs created for general use. | |
1587 | */ | |
5972511b | 1588 | static int biovec_create_pools(struct bio_set *bs, int pool_entries) |
1da177e4 | 1589 | { |
7ff9345f | 1590 | struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX; |
1da177e4 | 1591 | |
7ff9345f JA |
1592 | bs->bvec_pool = mempool_create_slab_pool(pool_entries, bp->slab); |
1593 | if (!bs->bvec_pool) | |
1594 | return -ENOMEM; | |
1da177e4 | 1595 | |
1da177e4 LT |
1596 | return 0; |
1597 | } | |
1598 | ||
1599 | static void biovec_free_pools(struct bio_set *bs) | |
1600 | { | |
7ff9345f | 1601 | mempool_destroy(bs->bvec_pool); |
1da177e4 LT |
1602 | } |
1603 | ||
1604 | void bioset_free(struct bio_set *bs) | |
1605 | { | |
1606 | if (bs->bio_pool) | |
1607 | mempool_destroy(bs->bio_pool); | |
1608 | ||
7878cba9 | 1609 | bioset_integrity_free(bs); |
1da177e4 | 1610 | biovec_free_pools(bs); |
bb799ca0 | 1611 | bio_put_slab(bs); |
1da177e4 LT |
1612 | |
1613 | kfree(bs); | |
1614 | } | |
a112a71d | 1615 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1616 | |
bb799ca0 JA |
1617 | /** |
1618 | * bioset_create - Create a bio_set | |
1619 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1620 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1621 | * | |
1622 | * Description: | |
1623 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1624 | * to ask for a number of bytes to be allocated in front of the bio. | |
1625 | * Front pad allocation is useful for embedding the bio inside | |
1626 | * another structure, to avoid allocating extra data to go with the bio. | |
1627 | * Note that the bio must be embedded at the END of that structure always, | |
1628 | * or things will break badly. | |
1629 | */ | |
1630 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) | |
1da177e4 | 1631 | { |
392ddc32 | 1632 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1633 | struct bio_set *bs; |
1da177e4 | 1634 | |
1b434498 | 1635 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1636 | if (!bs) |
1637 | return NULL; | |
1638 | ||
bb799ca0 | 1639 | bs->front_pad = front_pad; |
1b434498 | 1640 | |
392ddc32 | 1641 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1642 | if (!bs->bio_slab) { |
1643 | kfree(bs); | |
1644 | return NULL; | |
1645 | } | |
1646 | ||
1647 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
1648 | if (!bs->bio_pool) |
1649 | goto bad; | |
1650 | ||
bb799ca0 | 1651 | if (!biovec_create_pools(bs, pool_size)) |
1da177e4 LT |
1652 | return bs; |
1653 | ||
1654 | bad: | |
1655 | bioset_free(bs); | |
1656 | return NULL; | |
1657 | } | |
a112a71d | 1658 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 1659 | |
852c788f TH |
1660 | #ifdef CONFIG_BLK_CGROUP |
1661 | /** | |
1662 | * bio_associate_current - associate a bio with %current | |
1663 | * @bio: target bio | |
1664 | * | |
1665 | * Associate @bio with %current if it hasn't been associated yet. Block | |
1666 | * layer will treat @bio as if it were issued by %current no matter which | |
1667 | * task actually issues it. | |
1668 | * | |
1669 | * This function takes an extra reference of @task's io_context and blkcg | |
1670 | * which will be put when @bio is released. The caller must own @bio, | |
1671 | * ensure %current->io_context exists, and is responsible for synchronizing | |
1672 | * calls to this function. | |
1673 | */ | |
1674 | int bio_associate_current(struct bio *bio) | |
1675 | { | |
1676 | struct io_context *ioc; | |
1677 | struct cgroup_subsys_state *css; | |
1678 | ||
1679 | if (bio->bi_ioc) | |
1680 | return -EBUSY; | |
1681 | ||
1682 | ioc = current->io_context; | |
1683 | if (!ioc) | |
1684 | return -ENOENT; | |
1685 | ||
1686 | /* acquire active ref on @ioc and associate */ | |
1687 | get_io_context_active(ioc); | |
1688 | bio->bi_ioc = ioc; | |
1689 | ||
1690 | /* associate blkcg if exists */ | |
1691 | rcu_read_lock(); | |
1692 | css = task_subsys_state(current, blkio_subsys_id); | |
1693 | if (css && css_tryget(css)) | |
1694 | bio->bi_css = css; | |
1695 | rcu_read_unlock(); | |
1696 | ||
1697 | return 0; | |
1698 | } | |
1699 | ||
1700 | /** | |
1701 | * bio_disassociate_task - undo bio_associate_current() | |
1702 | * @bio: target bio | |
1703 | */ | |
1704 | void bio_disassociate_task(struct bio *bio) | |
1705 | { | |
1706 | if (bio->bi_ioc) { | |
1707 | put_io_context(bio->bi_ioc); | |
1708 | bio->bi_ioc = NULL; | |
1709 | } | |
1710 | if (bio->bi_css) { | |
1711 | css_put(bio->bi_css); | |
1712 | bio->bi_css = NULL; | |
1713 | } | |
1714 | } | |
1715 | ||
1716 | #endif /* CONFIG_BLK_CGROUP */ | |
1717 | ||
1da177e4 LT |
1718 | static void __init biovec_init_slabs(void) |
1719 | { | |
1720 | int i; | |
1721 | ||
1722 | for (i = 0; i < BIOVEC_NR_POOLS; i++) { | |
1723 | int size; | |
1724 | struct biovec_slab *bvs = bvec_slabs + i; | |
1725 | ||
a7fcd37c JA |
1726 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
1727 | bvs->slab = NULL; | |
1728 | continue; | |
1729 | } | |
a7fcd37c | 1730 | |
1da177e4 LT |
1731 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
1732 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 1733 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1734 | } |
1735 | } | |
1736 | ||
1737 | static int __init init_bio(void) | |
1738 | { | |
bb799ca0 JA |
1739 | bio_slab_max = 2; |
1740 | bio_slab_nr = 0; | |
1741 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
1742 | if (!bio_slabs) | |
1743 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 1744 | |
7878cba9 | 1745 | bio_integrity_init(); |
1da177e4 LT |
1746 | biovec_init_slabs(); |
1747 | ||
bb799ca0 | 1748 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
1da177e4 LT |
1749 | if (!fs_bio_set) |
1750 | panic("bio: can't allocate bios\n"); | |
1751 | ||
a91a2785 MP |
1752 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
1753 | panic("bio: can't create integrity pool\n"); | |
1754 | ||
0eaae62a MD |
1755 | bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES, |
1756 | sizeof(struct bio_pair)); | |
1da177e4 LT |
1757 | if (!bio_split_pool) |
1758 | panic("bio: can't create split pool\n"); | |
1759 | ||
1760 | return 0; | |
1761 | } | |
1da177e4 | 1762 | subsys_initcall(init_bio); |