]>
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 | /** | |
441 | * bio_clone - clone a bio | |
442 | * @bio: bio to clone | |
443 | * @gfp_mask: allocation priority | |
444 | * | |
445 | * Like __bio_clone, only also allocates the returned bio | |
446 | */ | |
dd0fc66f | 447 | struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask) |
1da177e4 | 448 | { |
395c72a7 | 449 | struct bio *b = bio_alloc(gfp_mask, bio->bi_max_vecs); |
1da177e4 | 450 | |
7ba1ba12 MP |
451 | if (!b) |
452 | return NULL; | |
453 | ||
7ba1ba12 MP |
454 | __bio_clone(b, bio); |
455 | ||
456 | if (bio_integrity(bio)) { | |
457 | int ret; | |
458 | ||
1e2a410f | 459 | ret = bio_integrity_clone(b, bio, gfp_mask); |
7ba1ba12 | 460 | |
059ea331 LZ |
461 | if (ret < 0) { |
462 | bio_put(b); | |
7ba1ba12 | 463 | return NULL; |
059ea331 | 464 | } |
3676347a | 465 | } |
1da177e4 LT |
466 | |
467 | return b; | |
468 | } | |
a112a71d | 469 | EXPORT_SYMBOL(bio_clone); |
1da177e4 LT |
470 | |
471 | /** | |
472 | * bio_get_nr_vecs - return approx number of vecs | |
473 | * @bdev: I/O target | |
474 | * | |
475 | * Return the approximate number of pages we can send to this target. | |
476 | * There's no guarantee that you will be able to fit this number of pages | |
477 | * into a bio, it does not account for dynamic restrictions that vary | |
478 | * on offset. | |
479 | */ | |
480 | int bio_get_nr_vecs(struct block_device *bdev) | |
481 | { | |
165125e1 | 482 | struct request_queue *q = bdev_get_queue(bdev); |
f908ee94 BS |
483 | int nr_pages; |
484 | ||
485 | nr_pages = min_t(unsigned, | |
5abebfdd KO |
486 | queue_max_segments(q), |
487 | queue_max_sectors(q) / (PAGE_SIZE >> 9) + 1); | |
f908ee94 BS |
488 | |
489 | return min_t(unsigned, nr_pages, BIO_MAX_PAGES); | |
490 | ||
1da177e4 | 491 | } |
a112a71d | 492 | EXPORT_SYMBOL(bio_get_nr_vecs); |
1da177e4 | 493 | |
165125e1 | 494 | static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page |
defd94b7 MC |
495 | *page, unsigned int len, unsigned int offset, |
496 | unsigned short max_sectors) | |
1da177e4 LT |
497 | { |
498 | int retried_segments = 0; | |
499 | struct bio_vec *bvec; | |
500 | ||
501 | /* | |
502 | * cloned bio must not modify vec list | |
503 | */ | |
504 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
505 | return 0; | |
506 | ||
80cfd548 | 507 | if (((bio->bi_size + len) >> 9) > max_sectors) |
1da177e4 LT |
508 | return 0; |
509 | ||
80cfd548 JA |
510 | /* |
511 | * For filesystems with a blocksize smaller than the pagesize | |
512 | * we will often be called with the same page as last time and | |
513 | * a consecutive offset. Optimize this special case. | |
514 | */ | |
515 | if (bio->bi_vcnt > 0) { | |
516 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
517 | ||
518 | if (page == prev->bv_page && | |
519 | offset == prev->bv_offset + prev->bv_len) { | |
1d616585 | 520 | unsigned int prev_bv_len = prev->bv_len; |
80cfd548 | 521 | prev->bv_len += len; |
cc371e66 AK |
522 | |
523 | if (q->merge_bvec_fn) { | |
524 | struct bvec_merge_data bvm = { | |
1d616585 DM |
525 | /* prev_bvec is already charged in |
526 | bi_size, discharge it in order to | |
527 | simulate merging updated prev_bvec | |
528 | as new bvec. */ | |
cc371e66 AK |
529 | .bi_bdev = bio->bi_bdev, |
530 | .bi_sector = bio->bi_sector, | |
1d616585 | 531 | .bi_size = bio->bi_size - prev_bv_len, |
cc371e66 AK |
532 | .bi_rw = bio->bi_rw, |
533 | }; | |
534 | ||
8bf8c376 | 535 | if (q->merge_bvec_fn(q, &bvm, prev) < prev->bv_len) { |
cc371e66 AK |
536 | prev->bv_len -= len; |
537 | return 0; | |
538 | } | |
80cfd548 JA |
539 | } |
540 | ||
541 | goto done; | |
542 | } | |
543 | } | |
544 | ||
545 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
546 | return 0; |
547 | ||
548 | /* | |
549 | * we might lose a segment or two here, but rather that than | |
550 | * make this too complex. | |
551 | */ | |
552 | ||
8a78362c | 553 | while (bio->bi_phys_segments >= queue_max_segments(q)) { |
1da177e4 LT |
554 | |
555 | if (retried_segments) | |
556 | return 0; | |
557 | ||
558 | retried_segments = 1; | |
559 | blk_recount_segments(q, bio); | |
560 | } | |
561 | ||
562 | /* | |
563 | * setup the new entry, we might clear it again later if we | |
564 | * cannot add the page | |
565 | */ | |
566 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
567 | bvec->bv_page = page; | |
568 | bvec->bv_len = len; | |
569 | bvec->bv_offset = offset; | |
570 | ||
571 | /* | |
572 | * if queue has other restrictions (eg varying max sector size | |
573 | * depending on offset), it can specify a merge_bvec_fn in the | |
574 | * queue to get further control | |
575 | */ | |
576 | if (q->merge_bvec_fn) { | |
cc371e66 AK |
577 | struct bvec_merge_data bvm = { |
578 | .bi_bdev = bio->bi_bdev, | |
579 | .bi_sector = bio->bi_sector, | |
580 | .bi_size = bio->bi_size, | |
581 | .bi_rw = bio->bi_rw, | |
582 | }; | |
583 | ||
1da177e4 LT |
584 | /* |
585 | * merge_bvec_fn() returns number of bytes it can accept | |
586 | * at this offset | |
587 | */ | |
8bf8c376 | 588 | if (q->merge_bvec_fn(q, &bvm, bvec) < bvec->bv_len) { |
1da177e4 LT |
589 | bvec->bv_page = NULL; |
590 | bvec->bv_len = 0; | |
591 | bvec->bv_offset = 0; | |
592 | return 0; | |
593 | } | |
594 | } | |
595 | ||
596 | /* If we may be able to merge these biovecs, force a recount */ | |
b8b3e16c | 597 | if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
1da177e4 LT |
598 | bio->bi_flags &= ~(1 << BIO_SEG_VALID); |
599 | ||
600 | bio->bi_vcnt++; | |
601 | bio->bi_phys_segments++; | |
80cfd548 | 602 | done: |
1da177e4 LT |
603 | bio->bi_size += len; |
604 | return len; | |
605 | } | |
606 | ||
6e68af66 MC |
607 | /** |
608 | * bio_add_pc_page - attempt to add page to bio | |
fddfdeaf | 609 | * @q: the target queue |
6e68af66 MC |
610 | * @bio: destination bio |
611 | * @page: page to add | |
612 | * @len: vec entry length | |
613 | * @offset: vec entry offset | |
614 | * | |
615 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
c6428084 AG |
616 | * number of reasons, such as the bio being full or target block device |
617 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
618 | * so it is always possible to add a single page to an empty bio. | |
619 | * | |
620 | * This should only be used by REQ_PC bios. | |
6e68af66 | 621 | */ |
165125e1 | 622 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page, |
6e68af66 MC |
623 | unsigned int len, unsigned int offset) |
624 | { | |
ae03bf63 MP |
625 | return __bio_add_page(q, bio, page, len, offset, |
626 | queue_max_hw_sectors(q)); | |
6e68af66 | 627 | } |
a112a71d | 628 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 629 | |
1da177e4 LT |
630 | /** |
631 | * bio_add_page - attempt to add page to bio | |
632 | * @bio: destination bio | |
633 | * @page: page to add | |
634 | * @len: vec entry length | |
635 | * @offset: vec entry offset | |
636 | * | |
637 | * Attempt to add a page to the bio_vec maplist. This can fail for a | |
c6428084 AG |
638 | * number of reasons, such as the bio being full or target block device |
639 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
640 | * so it is always possible to add a single page to an empty bio. | |
1da177e4 LT |
641 | */ |
642 | int bio_add_page(struct bio *bio, struct page *page, unsigned int len, | |
643 | unsigned int offset) | |
644 | { | |
defd94b7 | 645 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
ae03bf63 | 646 | return __bio_add_page(q, bio, page, len, offset, queue_max_sectors(q)); |
1da177e4 | 647 | } |
a112a71d | 648 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 LT |
649 | |
650 | struct bio_map_data { | |
651 | struct bio_vec *iovecs; | |
c5dec1c3 | 652 | struct sg_iovec *sgvecs; |
152e283f FT |
653 | int nr_sgvecs; |
654 | int is_our_pages; | |
1da177e4 LT |
655 | }; |
656 | ||
c5dec1c3 | 657 | static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio, |
152e283f FT |
658 | struct sg_iovec *iov, int iov_count, |
659 | int is_our_pages) | |
1da177e4 LT |
660 | { |
661 | memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt); | |
c5dec1c3 FT |
662 | memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count); |
663 | bmd->nr_sgvecs = iov_count; | |
152e283f | 664 | bmd->is_our_pages = is_our_pages; |
1da177e4 LT |
665 | bio->bi_private = bmd; |
666 | } | |
667 | ||
668 | static void bio_free_map_data(struct bio_map_data *bmd) | |
669 | { | |
670 | kfree(bmd->iovecs); | |
c5dec1c3 | 671 | kfree(bmd->sgvecs); |
1da177e4 LT |
672 | kfree(bmd); |
673 | } | |
674 | ||
121f0994 DC |
675 | static struct bio_map_data *bio_alloc_map_data(int nr_segs, |
676 | unsigned int iov_count, | |
76029ff3 | 677 | gfp_t gfp_mask) |
1da177e4 | 678 | { |
f3f63c1c JA |
679 | struct bio_map_data *bmd; |
680 | ||
681 | if (iov_count > UIO_MAXIOV) | |
682 | return NULL; | |
1da177e4 | 683 | |
f3f63c1c | 684 | bmd = kmalloc(sizeof(*bmd), gfp_mask); |
1da177e4 LT |
685 | if (!bmd) |
686 | return NULL; | |
687 | ||
76029ff3 | 688 | bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask); |
c5dec1c3 FT |
689 | if (!bmd->iovecs) { |
690 | kfree(bmd); | |
691 | return NULL; | |
692 | } | |
693 | ||
76029ff3 | 694 | bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask); |
c5dec1c3 | 695 | if (bmd->sgvecs) |
1da177e4 LT |
696 | return bmd; |
697 | ||
c5dec1c3 | 698 | kfree(bmd->iovecs); |
1da177e4 LT |
699 | kfree(bmd); |
700 | return NULL; | |
701 | } | |
702 | ||
aefcc28a | 703 | static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs, |
ecb554a8 FT |
704 | struct sg_iovec *iov, int iov_count, |
705 | int to_user, int from_user, int do_free_page) | |
c5dec1c3 FT |
706 | { |
707 | int ret = 0, i; | |
708 | struct bio_vec *bvec; | |
709 | int iov_idx = 0; | |
710 | unsigned int iov_off = 0; | |
c5dec1c3 FT |
711 | |
712 | __bio_for_each_segment(bvec, bio, i, 0) { | |
713 | char *bv_addr = page_address(bvec->bv_page); | |
aefcc28a | 714 | unsigned int bv_len = iovecs[i].bv_len; |
c5dec1c3 FT |
715 | |
716 | while (bv_len && iov_idx < iov_count) { | |
717 | unsigned int bytes; | |
0e0c6212 | 718 | char __user *iov_addr; |
c5dec1c3 FT |
719 | |
720 | bytes = min_t(unsigned int, | |
721 | iov[iov_idx].iov_len - iov_off, bv_len); | |
722 | iov_addr = iov[iov_idx].iov_base + iov_off; | |
723 | ||
724 | if (!ret) { | |
ecb554a8 | 725 | if (to_user) |
c5dec1c3 FT |
726 | ret = copy_to_user(iov_addr, bv_addr, |
727 | bytes); | |
728 | ||
ecb554a8 FT |
729 | if (from_user) |
730 | ret = copy_from_user(bv_addr, iov_addr, | |
731 | bytes); | |
732 | ||
c5dec1c3 FT |
733 | if (ret) |
734 | ret = -EFAULT; | |
735 | } | |
736 | ||
737 | bv_len -= bytes; | |
738 | bv_addr += bytes; | |
739 | iov_addr += bytes; | |
740 | iov_off += bytes; | |
741 | ||
742 | if (iov[iov_idx].iov_len == iov_off) { | |
743 | iov_idx++; | |
744 | iov_off = 0; | |
745 | } | |
746 | } | |
747 | ||
152e283f | 748 | if (do_free_page) |
c5dec1c3 FT |
749 | __free_page(bvec->bv_page); |
750 | } | |
751 | ||
752 | return ret; | |
753 | } | |
754 | ||
1da177e4 LT |
755 | /** |
756 | * bio_uncopy_user - finish previously mapped bio | |
757 | * @bio: bio being terminated | |
758 | * | |
759 | * Free pages allocated from bio_copy_user() and write back data | |
760 | * to user space in case of a read. | |
761 | */ | |
762 | int bio_uncopy_user(struct bio *bio) | |
763 | { | |
764 | struct bio_map_data *bmd = bio->bi_private; | |
81882766 | 765 | int ret = 0; |
1da177e4 | 766 | |
81882766 FT |
767 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) |
768 | ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs, | |
ecb554a8 FT |
769 | bmd->nr_sgvecs, bio_data_dir(bio) == READ, |
770 | 0, bmd->is_our_pages); | |
1da177e4 LT |
771 | bio_free_map_data(bmd); |
772 | bio_put(bio); | |
773 | return ret; | |
774 | } | |
a112a71d | 775 | EXPORT_SYMBOL(bio_uncopy_user); |
1da177e4 LT |
776 | |
777 | /** | |
c5dec1c3 | 778 | * bio_copy_user_iov - copy user data to bio |
1da177e4 | 779 | * @q: destination block queue |
152e283f | 780 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
781 | * @iov: the iovec. |
782 | * @iov_count: number of elements in the iovec | |
1da177e4 | 783 | * @write_to_vm: bool indicating writing to pages or not |
a3bce90e | 784 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
785 | * |
786 | * Prepares and returns a bio for indirect user io, bouncing data | |
787 | * to/from kernel pages as necessary. Must be paired with | |
788 | * call bio_uncopy_user() on io completion. | |
789 | */ | |
152e283f FT |
790 | struct bio *bio_copy_user_iov(struct request_queue *q, |
791 | struct rq_map_data *map_data, | |
792 | struct sg_iovec *iov, int iov_count, | |
793 | int write_to_vm, gfp_t gfp_mask) | |
1da177e4 | 794 | { |
1da177e4 LT |
795 | struct bio_map_data *bmd; |
796 | struct bio_vec *bvec; | |
797 | struct page *page; | |
798 | struct bio *bio; | |
799 | int i, ret; | |
c5dec1c3 FT |
800 | int nr_pages = 0; |
801 | unsigned int len = 0; | |
56c451f4 | 802 | unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0; |
1da177e4 | 803 | |
c5dec1c3 FT |
804 | for (i = 0; i < iov_count; i++) { |
805 | unsigned long uaddr; | |
806 | unsigned long end; | |
807 | unsigned long start; | |
808 | ||
809 | uaddr = (unsigned long)iov[i].iov_base; | |
810 | end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
811 | start = uaddr >> PAGE_SHIFT; | |
812 | ||
cb4644ca JA |
813 | /* |
814 | * Overflow, abort | |
815 | */ | |
816 | if (end < start) | |
817 | return ERR_PTR(-EINVAL); | |
818 | ||
c5dec1c3 FT |
819 | nr_pages += end - start; |
820 | len += iov[i].iov_len; | |
821 | } | |
822 | ||
69838727 FT |
823 | if (offset) |
824 | nr_pages++; | |
825 | ||
a3bce90e | 826 | bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask); |
1da177e4 LT |
827 | if (!bmd) |
828 | return ERR_PTR(-ENOMEM); | |
829 | ||
1da177e4 | 830 | ret = -ENOMEM; |
a9e9dc24 | 831 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
832 | if (!bio) |
833 | goto out_bmd; | |
834 | ||
7b6d91da CH |
835 | if (!write_to_vm) |
836 | bio->bi_rw |= REQ_WRITE; | |
1da177e4 LT |
837 | |
838 | ret = 0; | |
56c451f4 FT |
839 | |
840 | if (map_data) { | |
e623ddb4 | 841 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
842 | i = map_data->offset / PAGE_SIZE; |
843 | } | |
1da177e4 | 844 | while (len) { |
e623ddb4 | 845 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 846 | |
56c451f4 FT |
847 | bytes -= offset; |
848 | ||
1da177e4 LT |
849 | if (bytes > len) |
850 | bytes = len; | |
851 | ||
152e283f | 852 | if (map_data) { |
e623ddb4 | 853 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
854 | ret = -ENOMEM; |
855 | break; | |
856 | } | |
e623ddb4 FT |
857 | |
858 | page = map_data->pages[i / nr_pages]; | |
859 | page += (i % nr_pages); | |
860 | ||
861 | i++; | |
862 | } else { | |
152e283f | 863 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
864 | if (!page) { |
865 | ret = -ENOMEM; | |
866 | break; | |
867 | } | |
1da177e4 LT |
868 | } |
869 | ||
56c451f4 | 870 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 871 | break; |
1da177e4 LT |
872 | |
873 | len -= bytes; | |
56c451f4 | 874 | offset = 0; |
1da177e4 LT |
875 | } |
876 | ||
877 | if (ret) | |
878 | goto cleanup; | |
879 | ||
880 | /* | |
881 | * success | |
882 | */ | |
ecb554a8 FT |
883 | if ((!write_to_vm && (!map_data || !map_data->null_mapped)) || |
884 | (map_data && map_data->from_user)) { | |
885 | ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 1, 0); | |
c5dec1c3 FT |
886 | if (ret) |
887 | goto cleanup; | |
1da177e4 LT |
888 | } |
889 | ||
152e283f | 890 | bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1); |
1da177e4 LT |
891 | return bio; |
892 | cleanup: | |
152e283f FT |
893 | if (!map_data) |
894 | bio_for_each_segment(bvec, bio, i) | |
895 | __free_page(bvec->bv_page); | |
1da177e4 LT |
896 | |
897 | bio_put(bio); | |
898 | out_bmd: | |
899 | bio_free_map_data(bmd); | |
900 | return ERR_PTR(ret); | |
901 | } | |
902 | ||
c5dec1c3 FT |
903 | /** |
904 | * bio_copy_user - copy user data to bio | |
905 | * @q: destination block queue | |
152e283f | 906 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
c5dec1c3 FT |
907 | * @uaddr: start of user address |
908 | * @len: length in bytes | |
909 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 910 | * @gfp_mask: memory allocation flags |
c5dec1c3 FT |
911 | * |
912 | * Prepares and returns a bio for indirect user io, bouncing data | |
913 | * to/from kernel pages as necessary. Must be paired with | |
914 | * call bio_uncopy_user() on io completion. | |
915 | */ | |
152e283f FT |
916 | struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data, |
917 | unsigned long uaddr, unsigned int len, | |
918 | int write_to_vm, gfp_t gfp_mask) | |
c5dec1c3 FT |
919 | { |
920 | struct sg_iovec iov; | |
921 | ||
922 | iov.iov_base = (void __user *)uaddr; | |
923 | iov.iov_len = len; | |
924 | ||
152e283f | 925 | return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask); |
c5dec1c3 | 926 | } |
a112a71d | 927 | EXPORT_SYMBOL(bio_copy_user); |
c5dec1c3 | 928 | |
165125e1 | 929 | static struct bio *__bio_map_user_iov(struct request_queue *q, |
f1970baf JB |
930 | struct block_device *bdev, |
931 | struct sg_iovec *iov, int iov_count, | |
a3bce90e | 932 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 | 933 | { |
f1970baf JB |
934 | int i, j; |
935 | int nr_pages = 0; | |
1da177e4 LT |
936 | struct page **pages; |
937 | struct bio *bio; | |
f1970baf JB |
938 | int cur_page = 0; |
939 | int ret, offset; | |
1da177e4 | 940 | |
f1970baf JB |
941 | for (i = 0; i < iov_count; i++) { |
942 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
943 | unsigned long len = iov[i].iov_len; | |
944 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
945 | unsigned long start = uaddr >> PAGE_SHIFT; | |
946 | ||
cb4644ca JA |
947 | /* |
948 | * Overflow, abort | |
949 | */ | |
950 | if (end < start) | |
951 | return ERR_PTR(-EINVAL); | |
952 | ||
f1970baf JB |
953 | nr_pages += end - start; |
954 | /* | |
ad2d7225 | 955 | * buffer must be aligned to at least hardsector size for now |
f1970baf | 956 | */ |
ad2d7225 | 957 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
958 | return ERR_PTR(-EINVAL); |
959 | } | |
960 | ||
961 | if (!nr_pages) | |
1da177e4 LT |
962 | return ERR_PTR(-EINVAL); |
963 | ||
a9e9dc24 | 964 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
965 | if (!bio) |
966 | return ERR_PTR(-ENOMEM); | |
967 | ||
968 | ret = -ENOMEM; | |
a3bce90e | 969 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
970 | if (!pages) |
971 | goto out; | |
972 | ||
f1970baf JB |
973 | for (i = 0; i < iov_count; i++) { |
974 | unsigned long uaddr = (unsigned long)iov[i].iov_base; | |
975 | unsigned long len = iov[i].iov_len; | |
976 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
977 | unsigned long start = uaddr >> PAGE_SHIFT; | |
978 | const int local_nr_pages = end - start; | |
979 | const int page_limit = cur_page + local_nr_pages; | |
cb4644ca | 980 | |
f5dd33c4 NP |
981 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
982 | write_to_vm, &pages[cur_page]); | |
99172157 JA |
983 | if (ret < local_nr_pages) { |
984 | ret = -EFAULT; | |
f1970baf | 985 | goto out_unmap; |
99172157 | 986 | } |
f1970baf JB |
987 | |
988 | offset = uaddr & ~PAGE_MASK; | |
989 | for (j = cur_page; j < page_limit; j++) { | |
990 | unsigned int bytes = PAGE_SIZE - offset; | |
991 | ||
992 | if (len <= 0) | |
993 | break; | |
994 | ||
995 | if (bytes > len) | |
996 | bytes = len; | |
997 | ||
998 | /* | |
999 | * sorry... | |
1000 | */ | |
defd94b7 MC |
1001 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1002 | bytes) | |
f1970baf JB |
1003 | break; |
1004 | ||
1005 | len -= bytes; | |
1006 | offset = 0; | |
1007 | } | |
1da177e4 | 1008 | |
f1970baf | 1009 | cur_page = j; |
1da177e4 | 1010 | /* |
f1970baf | 1011 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1012 | */ |
f1970baf JB |
1013 | while (j < page_limit) |
1014 | page_cache_release(pages[j++]); | |
1da177e4 LT |
1015 | } |
1016 | ||
1da177e4 LT |
1017 | kfree(pages); |
1018 | ||
1019 | /* | |
1020 | * set data direction, and check if mapped pages need bouncing | |
1021 | */ | |
1022 | if (!write_to_vm) | |
7b6d91da | 1023 | bio->bi_rw |= REQ_WRITE; |
1da177e4 | 1024 | |
f1970baf | 1025 | bio->bi_bdev = bdev; |
1da177e4 LT |
1026 | bio->bi_flags |= (1 << BIO_USER_MAPPED); |
1027 | return bio; | |
f1970baf JB |
1028 | |
1029 | out_unmap: | |
1030 | for (i = 0; i < nr_pages; i++) { | |
1031 | if(!pages[i]) | |
1032 | break; | |
1033 | page_cache_release(pages[i]); | |
1034 | } | |
1035 | out: | |
1da177e4 LT |
1036 | kfree(pages); |
1037 | bio_put(bio); | |
1038 | return ERR_PTR(ret); | |
1039 | } | |
1040 | ||
1041 | /** | |
1042 | * bio_map_user - map user address into bio | |
165125e1 | 1043 | * @q: the struct request_queue for the bio |
1da177e4 LT |
1044 | * @bdev: destination block device |
1045 | * @uaddr: start of user address | |
1046 | * @len: length in bytes | |
1047 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1048 | * @gfp_mask: memory allocation flags |
1da177e4 LT |
1049 | * |
1050 | * Map the user space address into a bio suitable for io to a block | |
1051 | * device. Returns an error pointer in case of error. | |
1052 | */ | |
165125e1 | 1053 | struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev, |
a3bce90e FT |
1054 | unsigned long uaddr, unsigned int len, int write_to_vm, |
1055 | gfp_t gfp_mask) | |
f1970baf JB |
1056 | { |
1057 | struct sg_iovec iov; | |
1058 | ||
3f70353e | 1059 | iov.iov_base = (void __user *)uaddr; |
f1970baf JB |
1060 | iov.iov_len = len; |
1061 | ||
a3bce90e | 1062 | return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask); |
f1970baf | 1063 | } |
a112a71d | 1064 | EXPORT_SYMBOL(bio_map_user); |
f1970baf JB |
1065 | |
1066 | /** | |
1067 | * bio_map_user_iov - map user sg_iovec table into bio | |
165125e1 | 1068 | * @q: the struct request_queue for the bio |
f1970baf JB |
1069 | * @bdev: destination block device |
1070 | * @iov: the iovec. | |
1071 | * @iov_count: number of elements in the iovec | |
1072 | * @write_to_vm: bool indicating writing to pages or not | |
a3bce90e | 1073 | * @gfp_mask: memory allocation flags |
f1970baf JB |
1074 | * |
1075 | * Map the user space address into a bio suitable for io to a block | |
1076 | * device. Returns an error pointer in case of error. | |
1077 | */ | |
165125e1 | 1078 | struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev, |
f1970baf | 1079 | struct sg_iovec *iov, int iov_count, |
a3bce90e | 1080 | int write_to_vm, gfp_t gfp_mask) |
1da177e4 LT |
1081 | { |
1082 | struct bio *bio; | |
1083 | ||
a3bce90e FT |
1084 | bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm, |
1085 | gfp_mask); | |
1da177e4 LT |
1086 | if (IS_ERR(bio)) |
1087 | return bio; | |
1088 | ||
1089 | /* | |
1090 | * subtle -- if __bio_map_user() ended up bouncing a bio, | |
1091 | * it would normally disappear when its bi_end_io is run. | |
1092 | * however, we need it for the unmap, so grab an extra | |
1093 | * reference to it | |
1094 | */ | |
1095 | bio_get(bio); | |
1096 | ||
0e75f906 | 1097 | return bio; |
1da177e4 LT |
1098 | } |
1099 | ||
1100 | static void __bio_unmap_user(struct bio *bio) | |
1101 | { | |
1102 | struct bio_vec *bvec; | |
1103 | int i; | |
1104 | ||
1105 | /* | |
1106 | * make sure we dirty pages we wrote to | |
1107 | */ | |
1108 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1109 | if (bio_data_dir(bio) == READ) | |
1110 | set_page_dirty_lock(bvec->bv_page); | |
1111 | ||
1112 | page_cache_release(bvec->bv_page); | |
1113 | } | |
1114 | ||
1115 | bio_put(bio); | |
1116 | } | |
1117 | ||
1118 | /** | |
1119 | * bio_unmap_user - unmap a bio | |
1120 | * @bio: the bio being unmapped | |
1121 | * | |
1122 | * Unmap a bio previously mapped by bio_map_user(). Must be called with | |
1123 | * a process context. | |
1124 | * | |
1125 | * bio_unmap_user() may sleep. | |
1126 | */ | |
1127 | void bio_unmap_user(struct bio *bio) | |
1128 | { | |
1129 | __bio_unmap_user(bio); | |
1130 | bio_put(bio); | |
1131 | } | |
a112a71d | 1132 | EXPORT_SYMBOL(bio_unmap_user); |
1da177e4 | 1133 | |
6712ecf8 | 1134 | static void bio_map_kern_endio(struct bio *bio, int err) |
b823825e | 1135 | { |
b823825e | 1136 | bio_put(bio); |
b823825e JA |
1137 | } |
1138 | ||
165125e1 | 1139 | static struct bio *__bio_map_kern(struct request_queue *q, void *data, |
27496a8c | 1140 | unsigned int len, gfp_t gfp_mask) |
df46b9a4 MC |
1141 | { |
1142 | unsigned long kaddr = (unsigned long)data; | |
1143 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1144 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1145 | const int nr_pages = end - start; | |
1146 | int offset, i; | |
1147 | struct bio *bio; | |
1148 | ||
a9e9dc24 | 1149 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1150 | if (!bio) |
1151 | return ERR_PTR(-ENOMEM); | |
1152 | ||
1153 | offset = offset_in_page(kaddr); | |
1154 | for (i = 0; i < nr_pages; i++) { | |
1155 | unsigned int bytes = PAGE_SIZE - offset; | |
1156 | ||
1157 | if (len <= 0) | |
1158 | break; | |
1159 | ||
1160 | if (bytes > len) | |
1161 | bytes = len; | |
1162 | ||
defd94b7 MC |
1163 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
1164 | offset) < bytes) | |
df46b9a4 MC |
1165 | break; |
1166 | ||
1167 | data += bytes; | |
1168 | len -= bytes; | |
1169 | offset = 0; | |
1170 | } | |
1171 | ||
b823825e | 1172 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1173 | return bio; |
1174 | } | |
1175 | ||
1176 | /** | |
1177 | * bio_map_kern - map kernel address into bio | |
165125e1 | 1178 | * @q: the struct request_queue for the bio |
df46b9a4 MC |
1179 | * @data: pointer to buffer to map |
1180 | * @len: length in bytes | |
1181 | * @gfp_mask: allocation flags for bio allocation | |
1182 | * | |
1183 | * Map the kernel address into a bio suitable for io to a block | |
1184 | * device. Returns an error pointer in case of error. | |
1185 | */ | |
165125e1 | 1186 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
27496a8c | 1187 | gfp_t gfp_mask) |
df46b9a4 MC |
1188 | { |
1189 | struct bio *bio; | |
1190 | ||
1191 | bio = __bio_map_kern(q, data, len, gfp_mask); | |
1192 | if (IS_ERR(bio)) | |
1193 | return bio; | |
1194 | ||
1195 | if (bio->bi_size == len) | |
1196 | return bio; | |
1197 | ||
1198 | /* | |
1199 | * Don't support partial mappings. | |
1200 | */ | |
1201 | bio_put(bio); | |
1202 | return ERR_PTR(-EINVAL); | |
1203 | } | |
a112a71d | 1204 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1205 | |
68154e90 FT |
1206 | static void bio_copy_kern_endio(struct bio *bio, int err) |
1207 | { | |
1208 | struct bio_vec *bvec; | |
1209 | const int read = bio_data_dir(bio) == READ; | |
76029ff3 | 1210 | struct bio_map_data *bmd = bio->bi_private; |
68154e90 | 1211 | int i; |
76029ff3 | 1212 | char *p = bmd->sgvecs[0].iov_base; |
68154e90 FT |
1213 | |
1214 | __bio_for_each_segment(bvec, bio, i, 0) { | |
1215 | char *addr = page_address(bvec->bv_page); | |
76029ff3 | 1216 | int len = bmd->iovecs[i].bv_len; |
68154e90 | 1217 | |
4fc981ef | 1218 | if (read) |
76029ff3 | 1219 | memcpy(p, addr, len); |
68154e90 FT |
1220 | |
1221 | __free_page(bvec->bv_page); | |
76029ff3 | 1222 | p += len; |
68154e90 FT |
1223 | } |
1224 | ||
76029ff3 | 1225 | bio_free_map_data(bmd); |
68154e90 FT |
1226 | bio_put(bio); |
1227 | } | |
1228 | ||
1229 | /** | |
1230 | * bio_copy_kern - copy kernel address into bio | |
1231 | * @q: the struct request_queue for the bio | |
1232 | * @data: pointer to buffer to copy | |
1233 | * @len: length in bytes | |
1234 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1235 | * @reading: data direction is READ |
68154e90 FT |
1236 | * |
1237 | * copy the kernel address into a bio suitable for io to a block | |
1238 | * device. Returns an error pointer in case of error. | |
1239 | */ | |
1240 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1241 | gfp_t gfp_mask, int reading) | |
1242 | { | |
68154e90 FT |
1243 | struct bio *bio; |
1244 | struct bio_vec *bvec; | |
4d8ab62e | 1245 | int i; |
68154e90 | 1246 | |
4d8ab62e FT |
1247 | bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask); |
1248 | if (IS_ERR(bio)) | |
1249 | return bio; | |
68154e90 FT |
1250 | |
1251 | if (!reading) { | |
1252 | void *p = data; | |
1253 | ||
1254 | bio_for_each_segment(bvec, bio, i) { | |
1255 | char *addr = page_address(bvec->bv_page); | |
1256 | ||
1257 | memcpy(addr, p, bvec->bv_len); | |
1258 | p += bvec->bv_len; | |
1259 | } | |
1260 | } | |
1261 | ||
68154e90 | 1262 | bio->bi_end_io = bio_copy_kern_endio; |
76029ff3 | 1263 | |
68154e90 | 1264 | return bio; |
68154e90 | 1265 | } |
a112a71d | 1266 | EXPORT_SYMBOL(bio_copy_kern); |
68154e90 | 1267 | |
1da177e4 LT |
1268 | /* |
1269 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1270 | * for performing direct-IO in BIOs. | |
1271 | * | |
1272 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1273 | * because the required locks are not interrupt-safe. So what we can do is to | |
1274 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1275 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1276 | * in process context. | |
1277 | * | |
1278 | * We special-case compound pages here: normally this means reads into hugetlb | |
1279 | * pages. The logic in here doesn't really work right for compound pages | |
1280 | * because the VM does not uniformly chase down the head page in all cases. | |
1281 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1282 | * handle them at all. So we skip compound pages here at an early stage. | |
1283 | * | |
1284 | * Note that this code is very hard to test under normal circumstances because | |
1285 | * direct-io pins the pages with get_user_pages(). This makes | |
1286 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1287 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1288 | * pagecache. |
1289 | * | |
1290 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1291 | * deferred bio dirtying paths. | |
1292 | */ | |
1293 | ||
1294 | /* | |
1295 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1296 | */ | |
1297 | void bio_set_pages_dirty(struct bio *bio) | |
1298 | { | |
1299 | struct bio_vec *bvec = bio->bi_io_vec; | |
1300 | int i; | |
1301 | ||
1302 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1303 | struct page *page = bvec[i].bv_page; | |
1304 | ||
1305 | if (page && !PageCompound(page)) | |
1306 | set_page_dirty_lock(page); | |
1307 | } | |
1308 | } | |
1309 | ||
86b6c7a7 | 1310 | static void bio_release_pages(struct bio *bio) |
1da177e4 LT |
1311 | { |
1312 | struct bio_vec *bvec = bio->bi_io_vec; | |
1313 | int i; | |
1314 | ||
1315 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1316 | struct page *page = bvec[i].bv_page; | |
1317 | ||
1318 | if (page) | |
1319 | put_page(page); | |
1320 | } | |
1321 | } | |
1322 | ||
1323 | /* | |
1324 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1325 | * If they are, then fine. If, however, some pages are clean then they must | |
1326 | * have been written out during the direct-IO read. So we take another ref on | |
1327 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1328 | * | |
1329 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
1330 | * here on. It will run one page_cache_release() against each page and will | |
1331 | * run one bio_put() against the BIO. | |
1332 | */ | |
1333 | ||
65f27f38 | 1334 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1335 | |
65f27f38 | 1336 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1337 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1338 | static struct bio *bio_dirty_list; | |
1339 | ||
1340 | /* | |
1341 | * This runs in process context | |
1342 | */ | |
65f27f38 | 1343 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1344 | { |
1345 | unsigned long flags; | |
1346 | struct bio *bio; | |
1347 | ||
1348 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1349 | bio = bio_dirty_list; | |
1350 | bio_dirty_list = NULL; | |
1351 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1352 | ||
1353 | while (bio) { | |
1354 | struct bio *next = bio->bi_private; | |
1355 | ||
1356 | bio_set_pages_dirty(bio); | |
1357 | bio_release_pages(bio); | |
1358 | bio_put(bio); | |
1359 | bio = next; | |
1360 | } | |
1361 | } | |
1362 | ||
1363 | void bio_check_pages_dirty(struct bio *bio) | |
1364 | { | |
1365 | struct bio_vec *bvec = bio->bi_io_vec; | |
1366 | int nr_clean_pages = 0; | |
1367 | int i; | |
1368 | ||
1369 | for (i = 0; i < bio->bi_vcnt; i++) { | |
1370 | struct page *page = bvec[i].bv_page; | |
1371 | ||
1372 | if (PageDirty(page) || PageCompound(page)) { | |
1373 | page_cache_release(page); | |
1374 | bvec[i].bv_page = NULL; | |
1375 | } else { | |
1376 | nr_clean_pages++; | |
1377 | } | |
1378 | } | |
1379 | ||
1380 | if (nr_clean_pages) { | |
1381 | unsigned long flags; | |
1382 | ||
1383 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1384 | bio->bi_private = bio_dirty_list; | |
1385 | bio_dirty_list = bio; | |
1386 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1387 | schedule_work(&bio_dirty_work); | |
1388 | } else { | |
1389 | bio_put(bio); | |
1390 | } | |
1391 | } | |
1392 | ||
2d4dc890 IL |
1393 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1394 | void bio_flush_dcache_pages(struct bio *bi) | |
1395 | { | |
1396 | int i; | |
1397 | struct bio_vec *bvec; | |
1398 | ||
1399 | bio_for_each_segment(bvec, bi, i) | |
1400 | flush_dcache_page(bvec->bv_page); | |
1401 | } | |
1402 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1403 | #endif | |
1404 | ||
1da177e4 LT |
1405 | /** |
1406 | * bio_endio - end I/O on a bio | |
1407 | * @bio: bio | |
1da177e4 LT |
1408 | * @error: error, if any |
1409 | * | |
1410 | * Description: | |
6712ecf8 | 1411 | * bio_endio() will end I/O on the whole bio. bio_endio() is the |
5bb23a68 N |
1412 | * preferred way to end I/O on a bio, it takes care of clearing |
1413 | * BIO_UPTODATE on error. @error is 0 on success, and and one of the | |
1414 | * established -Exxxx (-EIO, for instance) error values in case | |
25985edc | 1415 | * something went wrong. No one should call bi_end_io() directly on a |
5bb23a68 N |
1416 | * bio unless they own it and thus know that it has an end_io |
1417 | * function. | |
1da177e4 | 1418 | **/ |
6712ecf8 | 1419 | void bio_endio(struct bio *bio, int error) |
1da177e4 LT |
1420 | { |
1421 | if (error) | |
1422 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
9cc54d40 N |
1423 | else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) |
1424 | error = -EIO; | |
1da177e4 | 1425 | |
5bb23a68 | 1426 | if (bio->bi_end_io) |
6712ecf8 | 1427 | bio->bi_end_io(bio, error); |
1da177e4 | 1428 | } |
a112a71d | 1429 | EXPORT_SYMBOL(bio_endio); |
1da177e4 LT |
1430 | |
1431 | void bio_pair_release(struct bio_pair *bp) | |
1432 | { | |
1433 | if (atomic_dec_and_test(&bp->cnt)) { | |
1434 | struct bio *master = bp->bio1.bi_private; | |
1435 | ||
6712ecf8 | 1436 | bio_endio(master, bp->error); |
1da177e4 LT |
1437 | mempool_free(bp, bp->bio2.bi_private); |
1438 | } | |
1439 | } | |
a112a71d | 1440 | EXPORT_SYMBOL(bio_pair_release); |
1da177e4 | 1441 | |
6712ecf8 | 1442 | static void bio_pair_end_1(struct bio *bi, int err) |
1da177e4 LT |
1443 | { |
1444 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio1); | |
1445 | ||
1446 | if (err) | |
1447 | bp->error = err; | |
1448 | ||
1da177e4 | 1449 | bio_pair_release(bp); |
1da177e4 LT |
1450 | } |
1451 | ||
6712ecf8 | 1452 | static void bio_pair_end_2(struct bio *bi, int err) |
1da177e4 LT |
1453 | { |
1454 | struct bio_pair *bp = container_of(bi, struct bio_pair, bio2); | |
1455 | ||
1456 | if (err) | |
1457 | bp->error = err; | |
1458 | ||
1da177e4 | 1459 | bio_pair_release(bp); |
1da177e4 LT |
1460 | } |
1461 | ||
1462 | /* | |
c7eee1b8 | 1463 | * split a bio - only worry about a bio with a single page in its iovec |
1da177e4 | 1464 | */ |
6feef531 | 1465 | struct bio_pair *bio_split(struct bio *bi, int first_sectors) |
1da177e4 | 1466 | { |
6feef531 | 1467 | struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO); |
1da177e4 LT |
1468 | |
1469 | if (!bp) | |
1470 | return bp; | |
1471 | ||
5f3ea37c | 1472 | trace_block_split(bdev_get_queue(bi->bi_bdev), bi, |
2056a782 JA |
1473 | bi->bi_sector + first_sectors); |
1474 | ||
1da177e4 LT |
1475 | BUG_ON(bi->bi_vcnt != 1); |
1476 | BUG_ON(bi->bi_idx != 0); | |
1477 | atomic_set(&bp->cnt, 3); | |
1478 | bp->error = 0; | |
1479 | bp->bio1 = *bi; | |
1480 | bp->bio2 = *bi; | |
1481 | bp->bio2.bi_sector += first_sectors; | |
1482 | bp->bio2.bi_size -= first_sectors << 9; | |
1483 | bp->bio1.bi_size = first_sectors << 9; | |
1484 | ||
1485 | bp->bv1 = bi->bi_io_vec[0]; | |
1486 | bp->bv2 = bi->bi_io_vec[0]; | |
1487 | bp->bv2.bv_offset += first_sectors << 9; | |
1488 | bp->bv2.bv_len -= first_sectors << 9; | |
1489 | bp->bv1.bv_len = first_sectors << 9; | |
1490 | ||
1491 | bp->bio1.bi_io_vec = &bp->bv1; | |
1492 | bp->bio2.bi_io_vec = &bp->bv2; | |
1493 | ||
a2eb0c10 N |
1494 | bp->bio1.bi_max_vecs = 1; |
1495 | bp->bio2.bi_max_vecs = 1; | |
1496 | ||
1da177e4 LT |
1497 | bp->bio1.bi_end_io = bio_pair_end_1; |
1498 | bp->bio2.bi_end_io = bio_pair_end_2; | |
1499 | ||
1500 | bp->bio1.bi_private = bi; | |
6feef531 | 1501 | bp->bio2.bi_private = bio_split_pool; |
1da177e4 | 1502 | |
7ba1ba12 MP |
1503 | if (bio_integrity(bi)) |
1504 | bio_integrity_split(bi, bp, first_sectors); | |
1505 | ||
1da177e4 LT |
1506 | return bp; |
1507 | } | |
a112a71d | 1508 | EXPORT_SYMBOL(bio_split); |
1da177e4 | 1509 | |
ad3316bf MP |
1510 | /** |
1511 | * bio_sector_offset - Find hardware sector offset in bio | |
1512 | * @bio: bio to inspect | |
1513 | * @index: bio_vec index | |
1514 | * @offset: offset in bv_page | |
1515 | * | |
1516 | * Return the number of hardware sectors between beginning of bio | |
1517 | * and an end point indicated by a bio_vec index and an offset | |
1518 | * within that vector's page. | |
1519 | */ | |
1520 | sector_t bio_sector_offset(struct bio *bio, unsigned short index, | |
1521 | unsigned int offset) | |
1522 | { | |
e1defc4f | 1523 | unsigned int sector_sz; |
ad3316bf MP |
1524 | struct bio_vec *bv; |
1525 | sector_t sectors; | |
1526 | int i; | |
1527 | ||
e1defc4f | 1528 | sector_sz = queue_logical_block_size(bio->bi_bdev->bd_disk->queue); |
ad3316bf MP |
1529 | sectors = 0; |
1530 | ||
1531 | if (index >= bio->bi_idx) | |
1532 | index = bio->bi_vcnt - 1; | |
1533 | ||
1534 | __bio_for_each_segment(bv, bio, i, 0) { | |
1535 | if (i == index) { | |
1536 | if (offset > bv->bv_offset) | |
1537 | sectors += (offset - bv->bv_offset) / sector_sz; | |
1538 | break; | |
1539 | } | |
1540 | ||
1541 | sectors += bv->bv_len / sector_sz; | |
1542 | } | |
1543 | ||
1544 | return sectors; | |
1545 | } | |
1546 | EXPORT_SYMBOL(bio_sector_offset); | |
1da177e4 LT |
1547 | |
1548 | /* | |
1549 | * create memory pools for biovec's in a bio_set. | |
1550 | * use the global biovec slabs created for general use. | |
1551 | */ | |
5972511b | 1552 | static int biovec_create_pools(struct bio_set *bs, int pool_entries) |
1da177e4 | 1553 | { |
7ff9345f | 1554 | struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX; |
1da177e4 | 1555 | |
7ff9345f JA |
1556 | bs->bvec_pool = mempool_create_slab_pool(pool_entries, bp->slab); |
1557 | if (!bs->bvec_pool) | |
1558 | return -ENOMEM; | |
1da177e4 | 1559 | |
1da177e4 LT |
1560 | return 0; |
1561 | } | |
1562 | ||
1563 | static void biovec_free_pools(struct bio_set *bs) | |
1564 | { | |
7ff9345f | 1565 | mempool_destroy(bs->bvec_pool); |
1da177e4 LT |
1566 | } |
1567 | ||
1568 | void bioset_free(struct bio_set *bs) | |
1569 | { | |
1570 | if (bs->bio_pool) | |
1571 | mempool_destroy(bs->bio_pool); | |
1572 | ||
7878cba9 | 1573 | bioset_integrity_free(bs); |
1da177e4 | 1574 | biovec_free_pools(bs); |
bb799ca0 | 1575 | bio_put_slab(bs); |
1da177e4 LT |
1576 | |
1577 | kfree(bs); | |
1578 | } | |
a112a71d | 1579 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1580 | |
bb799ca0 JA |
1581 | /** |
1582 | * bioset_create - Create a bio_set | |
1583 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1584 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1585 | * | |
1586 | * Description: | |
1587 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1588 | * to ask for a number of bytes to be allocated in front of the bio. | |
1589 | * Front pad allocation is useful for embedding the bio inside | |
1590 | * another structure, to avoid allocating extra data to go with the bio. | |
1591 | * Note that the bio must be embedded at the END of that structure always, | |
1592 | * or things will break badly. | |
1593 | */ | |
1594 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) | |
1da177e4 | 1595 | { |
392ddc32 | 1596 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1597 | struct bio_set *bs; |
1da177e4 | 1598 | |
1b434498 | 1599 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1600 | if (!bs) |
1601 | return NULL; | |
1602 | ||
bb799ca0 | 1603 | bs->front_pad = front_pad; |
1b434498 | 1604 | |
392ddc32 | 1605 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1606 | if (!bs->bio_slab) { |
1607 | kfree(bs); | |
1608 | return NULL; | |
1609 | } | |
1610 | ||
1611 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
1612 | if (!bs->bio_pool) |
1613 | goto bad; | |
1614 | ||
bb799ca0 | 1615 | if (!biovec_create_pools(bs, pool_size)) |
1da177e4 LT |
1616 | return bs; |
1617 | ||
1618 | bad: | |
1619 | bioset_free(bs); | |
1620 | return NULL; | |
1621 | } | |
a112a71d | 1622 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 1623 | |
852c788f TH |
1624 | #ifdef CONFIG_BLK_CGROUP |
1625 | /** | |
1626 | * bio_associate_current - associate a bio with %current | |
1627 | * @bio: target bio | |
1628 | * | |
1629 | * Associate @bio with %current if it hasn't been associated yet. Block | |
1630 | * layer will treat @bio as if it were issued by %current no matter which | |
1631 | * task actually issues it. | |
1632 | * | |
1633 | * This function takes an extra reference of @task's io_context and blkcg | |
1634 | * which will be put when @bio is released. The caller must own @bio, | |
1635 | * ensure %current->io_context exists, and is responsible for synchronizing | |
1636 | * calls to this function. | |
1637 | */ | |
1638 | int bio_associate_current(struct bio *bio) | |
1639 | { | |
1640 | struct io_context *ioc; | |
1641 | struct cgroup_subsys_state *css; | |
1642 | ||
1643 | if (bio->bi_ioc) | |
1644 | return -EBUSY; | |
1645 | ||
1646 | ioc = current->io_context; | |
1647 | if (!ioc) | |
1648 | return -ENOENT; | |
1649 | ||
1650 | /* acquire active ref on @ioc and associate */ | |
1651 | get_io_context_active(ioc); | |
1652 | bio->bi_ioc = ioc; | |
1653 | ||
1654 | /* associate blkcg if exists */ | |
1655 | rcu_read_lock(); | |
1656 | css = task_subsys_state(current, blkio_subsys_id); | |
1657 | if (css && css_tryget(css)) | |
1658 | bio->bi_css = css; | |
1659 | rcu_read_unlock(); | |
1660 | ||
1661 | return 0; | |
1662 | } | |
1663 | ||
1664 | /** | |
1665 | * bio_disassociate_task - undo bio_associate_current() | |
1666 | * @bio: target bio | |
1667 | */ | |
1668 | void bio_disassociate_task(struct bio *bio) | |
1669 | { | |
1670 | if (bio->bi_ioc) { | |
1671 | put_io_context(bio->bi_ioc); | |
1672 | bio->bi_ioc = NULL; | |
1673 | } | |
1674 | if (bio->bi_css) { | |
1675 | css_put(bio->bi_css); | |
1676 | bio->bi_css = NULL; | |
1677 | } | |
1678 | } | |
1679 | ||
1680 | #endif /* CONFIG_BLK_CGROUP */ | |
1681 | ||
1da177e4 LT |
1682 | static void __init biovec_init_slabs(void) |
1683 | { | |
1684 | int i; | |
1685 | ||
1686 | for (i = 0; i < BIOVEC_NR_POOLS; i++) { | |
1687 | int size; | |
1688 | struct biovec_slab *bvs = bvec_slabs + i; | |
1689 | ||
a7fcd37c JA |
1690 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
1691 | bvs->slab = NULL; | |
1692 | continue; | |
1693 | } | |
a7fcd37c | 1694 | |
1da177e4 LT |
1695 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
1696 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 1697 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
1698 | } |
1699 | } | |
1700 | ||
1701 | static int __init init_bio(void) | |
1702 | { | |
bb799ca0 JA |
1703 | bio_slab_max = 2; |
1704 | bio_slab_nr = 0; | |
1705 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
1706 | if (!bio_slabs) | |
1707 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 1708 | |
7878cba9 | 1709 | bio_integrity_init(); |
1da177e4 LT |
1710 | biovec_init_slabs(); |
1711 | ||
bb799ca0 | 1712 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
1da177e4 LT |
1713 | if (!fs_bio_set) |
1714 | panic("bio: can't allocate bios\n"); | |
1715 | ||
a91a2785 MP |
1716 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
1717 | panic("bio: can't create integrity pool\n"); | |
1718 | ||
0eaae62a MD |
1719 | bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES, |
1720 | sizeof(struct bio_pair)); | |
1da177e4 LT |
1721 | if (!bio_split_pool) |
1722 | panic("bio: can't create split pool\n"); | |
1723 | ||
1724 | return 0; | |
1725 | } | |
1da177e4 | 1726 | subsys_initcall(init_bio); |