]>
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> | |
a27bb332 | 22 | #include <linux/uio.h> |
852c788f | 23 | #include <linux/iocontext.h> |
1da177e4 LT |
24 | #include <linux/slab.h> |
25 | #include <linux/init.h> | |
26 | #include <linux/kernel.h> | |
630d9c47 | 27 | #include <linux/export.h> |
1da177e4 LT |
28 | #include <linux/mempool.h> |
29 | #include <linux/workqueue.h> | |
852c788f | 30 | #include <linux/cgroup.h> |
1da177e4 | 31 | |
55782138 | 32 | #include <trace/events/block.h> |
9e234eea | 33 | #include "blk.h" |
0bfc2455 | 34 | |
392ddc32 JA |
35 | /* |
36 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
37 | * itself, to shrink a bio data allocation from two mempool calls to one | |
38 | */ | |
39 | #define BIO_INLINE_VECS 4 | |
40 | ||
1da177e4 LT |
41 | /* |
42 | * if you change this list, also change bvec_alloc or things will | |
43 | * break badly! cannot be bigger than what you can fit into an | |
44 | * unsigned short | |
45 | */ | |
bd5c4fac | 46 | #define BV(x, n) { .nr_vecs = x, .name = "biovec-"#n } |
ed996a52 | 47 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
bd5c4fac | 48 | BV(1, 1), BV(4, 4), BV(16, 16), BV(64, 64), BV(128, 128), BV(BIO_MAX_PAGES, max), |
1da177e4 LT |
49 | }; |
50 | #undef BV | |
51 | ||
1da177e4 LT |
52 | /* |
53 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
54 | * IO code that does not need private memory pools. | |
55 | */ | |
51d654e1 | 56 | struct bio_set *fs_bio_set; |
3f86a82a | 57 | EXPORT_SYMBOL(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; |
386bc35a | 77 | unsigned int new_bio_slab_max; |
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) { | |
386bc35a | 100 | new_bio_slab_max = bio_slab_max << 1; |
389d7b26 | 101 | new_bio_slabs = krealloc(bio_slabs, |
386bc35a | 102 | new_bio_slab_max * sizeof(struct bio_slab), |
389d7b26 AK |
103 | GFP_KERNEL); |
104 | if (!new_bio_slabs) | |
bb799ca0 | 105 | goto out_unlock; |
386bc35a | 106 | bio_slab_max = new_bio_slab_max; |
389d7b26 | 107 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
108 | } |
109 | if (entry == -1) | |
110 | entry = bio_slab_nr++; | |
111 | ||
112 | bslab = &bio_slabs[entry]; | |
113 | ||
114 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
6a241483 MP |
115 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
116 | SLAB_HWCACHE_ALIGN, NULL); | |
bb799ca0 JA |
117 | if (!slab) |
118 | goto out_unlock; | |
119 | ||
bb799ca0 JA |
120 | bslab->slab = slab; |
121 | bslab->slab_ref = 1; | |
122 | bslab->slab_size = sz; | |
123 | out_unlock: | |
124 | mutex_unlock(&bio_slab_lock); | |
125 | return slab; | |
126 | } | |
127 | ||
128 | static void bio_put_slab(struct bio_set *bs) | |
129 | { | |
130 | struct bio_slab *bslab = NULL; | |
131 | unsigned int i; | |
132 | ||
133 | mutex_lock(&bio_slab_lock); | |
134 | ||
135 | for (i = 0; i < bio_slab_nr; i++) { | |
136 | if (bs->bio_slab == bio_slabs[i].slab) { | |
137 | bslab = &bio_slabs[i]; | |
138 | break; | |
139 | } | |
140 | } | |
141 | ||
142 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
143 | goto out; | |
144 | ||
145 | WARN_ON(!bslab->slab_ref); | |
146 | ||
147 | if (--bslab->slab_ref) | |
148 | goto out; | |
149 | ||
150 | kmem_cache_destroy(bslab->slab); | |
151 | bslab->slab = NULL; | |
152 | ||
153 | out: | |
154 | mutex_unlock(&bio_slab_lock); | |
155 | } | |
156 | ||
7ba1ba12 MP |
157 | unsigned int bvec_nr_vecs(unsigned short idx) |
158 | { | |
159 | return bvec_slabs[idx].nr_vecs; | |
160 | } | |
161 | ||
9f060e22 | 162 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
bb799ca0 | 163 | { |
ed996a52 CH |
164 | if (!idx) |
165 | return; | |
166 | idx--; | |
167 | ||
168 | BIO_BUG_ON(idx >= BVEC_POOL_NR); | |
bb799ca0 | 169 | |
ed996a52 | 170 | if (idx == BVEC_POOL_MAX) { |
9f060e22 | 171 | mempool_free(bv, pool); |
ed996a52 | 172 | } else { |
bb799ca0 JA |
173 | struct biovec_slab *bvs = bvec_slabs + idx; |
174 | ||
175 | kmem_cache_free(bvs->slab, bv); | |
176 | } | |
177 | } | |
178 | ||
9f060e22 KO |
179 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
180 | mempool_t *pool) | |
1da177e4 LT |
181 | { |
182 | struct bio_vec *bvl; | |
1da177e4 | 183 | |
7ff9345f JA |
184 | /* |
185 | * see comment near bvec_array define! | |
186 | */ | |
187 | switch (nr) { | |
188 | case 1: | |
189 | *idx = 0; | |
190 | break; | |
191 | case 2 ... 4: | |
192 | *idx = 1; | |
193 | break; | |
194 | case 5 ... 16: | |
195 | *idx = 2; | |
196 | break; | |
197 | case 17 ... 64: | |
198 | *idx = 3; | |
199 | break; | |
200 | case 65 ... 128: | |
201 | *idx = 4; | |
202 | break; | |
203 | case 129 ... BIO_MAX_PAGES: | |
204 | *idx = 5; | |
205 | break; | |
206 | default: | |
207 | return NULL; | |
208 | } | |
209 | ||
210 | /* | |
211 | * idx now points to the pool we want to allocate from. only the | |
212 | * 1-vec entry pool is mempool backed. | |
213 | */ | |
ed996a52 | 214 | if (*idx == BVEC_POOL_MAX) { |
7ff9345f | 215 | fallback: |
9f060e22 | 216 | bvl = mempool_alloc(pool, gfp_mask); |
7ff9345f JA |
217 | } else { |
218 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
d0164adc | 219 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
7ff9345f | 220 | |
0a0d96b0 | 221 | /* |
7ff9345f JA |
222 | * Make this allocation restricted and don't dump info on |
223 | * allocation failures, since we'll fallback to the mempool | |
224 | * in case of failure. | |
0a0d96b0 | 225 | */ |
7ff9345f | 226 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 227 | |
0a0d96b0 | 228 | /* |
d0164adc | 229 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
7ff9345f | 230 | * is set, retry with the 1-entry mempool |
0a0d96b0 | 231 | */ |
7ff9345f | 232 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
d0164adc | 233 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
ed996a52 | 234 | *idx = BVEC_POOL_MAX; |
7ff9345f JA |
235 | goto fallback; |
236 | } | |
237 | } | |
238 | ||
ed996a52 | 239 | (*idx)++; |
1da177e4 LT |
240 | return bvl; |
241 | } | |
242 | ||
9ae3b3f5 | 243 | void bio_uninit(struct bio *bio) |
1da177e4 | 244 | { |
4254bba1 | 245 | bio_disassociate_task(bio); |
4254bba1 | 246 | } |
9ae3b3f5 | 247 | EXPORT_SYMBOL(bio_uninit); |
7ba1ba12 | 248 | |
4254bba1 KO |
249 | static void bio_free(struct bio *bio) |
250 | { | |
251 | struct bio_set *bs = bio->bi_pool; | |
252 | void *p; | |
253 | ||
9ae3b3f5 | 254 | bio_uninit(bio); |
4254bba1 KO |
255 | |
256 | if (bs) { | |
8aa6ba2f | 257 | bvec_free(&bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio)); |
4254bba1 KO |
258 | |
259 | /* | |
260 | * If we have front padding, adjust the bio pointer before freeing | |
261 | */ | |
262 | p = bio; | |
bb799ca0 JA |
263 | p -= bs->front_pad; |
264 | ||
8aa6ba2f | 265 | mempool_free(p, &bs->bio_pool); |
4254bba1 KO |
266 | } else { |
267 | /* Bio was allocated by bio_kmalloc() */ | |
268 | kfree(bio); | |
269 | } | |
3676347a PO |
270 | } |
271 | ||
9ae3b3f5 JA |
272 | /* |
273 | * Users of this function have their own bio allocation. Subsequently, | |
274 | * they must remember to pair any call to bio_init() with bio_uninit() | |
275 | * when IO has completed, or when the bio is released. | |
276 | */ | |
3a83f467 ML |
277 | void bio_init(struct bio *bio, struct bio_vec *table, |
278 | unsigned short max_vecs) | |
1da177e4 | 279 | { |
2b94de55 | 280 | memset(bio, 0, sizeof(*bio)); |
c4cf5261 | 281 | atomic_set(&bio->__bi_remaining, 1); |
dac56212 | 282 | atomic_set(&bio->__bi_cnt, 1); |
3a83f467 ML |
283 | |
284 | bio->bi_io_vec = table; | |
285 | bio->bi_max_vecs = max_vecs; | |
1da177e4 | 286 | } |
a112a71d | 287 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 288 | |
f44b48c7 KO |
289 | /** |
290 | * bio_reset - reinitialize a bio | |
291 | * @bio: bio to reset | |
292 | * | |
293 | * Description: | |
294 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
295 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
296 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
297 | * comment in struct bio. | |
298 | */ | |
299 | void bio_reset(struct bio *bio) | |
300 | { | |
301 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
302 | ||
9ae3b3f5 | 303 | bio_uninit(bio); |
f44b48c7 KO |
304 | |
305 | memset(bio, 0, BIO_RESET_BYTES); | |
4246a0b6 | 306 | bio->bi_flags = flags; |
c4cf5261 | 307 | atomic_set(&bio->__bi_remaining, 1); |
f44b48c7 KO |
308 | } |
309 | EXPORT_SYMBOL(bio_reset); | |
310 | ||
38f8baae | 311 | static struct bio *__bio_chain_endio(struct bio *bio) |
196d38bc | 312 | { |
4246a0b6 CH |
313 | struct bio *parent = bio->bi_private; |
314 | ||
4e4cbee9 CH |
315 | if (!parent->bi_status) |
316 | parent->bi_status = bio->bi_status; | |
196d38bc | 317 | bio_put(bio); |
38f8baae CH |
318 | return parent; |
319 | } | |
320 | ||
321 | static void bio_chain_endio(struct bio *bio) | |
322 | { | |
323 | bio_endio(__bio_chain_endio(bio)); | |
196d38bc KO |
324 | } |
325 | ||
326 | /** | |
327 | * bio_chain - chain bio completions | |
1051a902 RD |
328 | * @bio: the target bio |
329 | * @parent: the @bio's parent bio | |
196d38bc KO |
330 | * |
331 | * The caller won't have a bi_end_io called when @bio completes - instead, | |
332 | * @parent's bi_end_io won't be called until both @parent and @bio have | |
333 | * completed; the chained bio will also be freed when it completes. | |
334 | * | |
335 | * The caller must not set bi_private or bi_end_io in @bio. | |
336 | */ | |
337 | void bio_chain(struct bio *bio, struct bio *parent) | |
338 | { | |
339 | BUG_ON(bio->bi_private || bio->bi_end_io); | |
340 | ||
341 | bio->bi_private = parent; | |
342 | bio->bi_end_io = bio_chain_endio; | |
c4cf5261 | 343 | bio_inc_remaining(parent); |
196d38bc KO |
344 | } |
345 | EXPORT_SYMBOL(bio_chain); | |
346 | ||
df2cb6da KO |
347 | static void bio_alloc_rescue(struct work_struct *work) |
348 | { | |
349 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); | |
350 | struct bio *bio; | |
351 | ||
352 | while (1) { | |
353 | spin_lock(&bs->rescue_lock); | |
354 | bio = bio_list_pop(&bs->rescue_list); | |
355 | spin_unlock(&bs->rescue_lock); | |
356 | ||
357 | if (!bio) | |
358 | break; | |
359 | ||
360 | generic_make_request(bio); | |
361 | } | |
362 | } | |
363 | ||
364 | static void punt_bios_to_rescuer(struct bio_set *bs) | |
365 | { | |
366 | struct bio_list punt, nopunt; | |
367 | struct bio *bio; | |
368 | ||
47e0fb46 N |
369 | if (WARN_ON_ONCE(!bs->rescue_workqueue)) |
370 | return; | |
df2cb6da KO |
371 | /* |
372 | * In order to guarantee forward progress we must punt only bios that | |
373 | * were allocated from this bio_set; otherwise, if there was a bio on | |
374 | * there for a stacking driver higher up in the stack, processing it | |
375 | * could require allocating bios from this bio_set, and doing that from | |
376 | * our own rescuer would be bad. | |
377 | * | |
378 | * Since bio lists are singly linked, pop them all instead of trying to | |
379 | * remove from the middle of the list: | |
380 | */ | |
381 | ||
382 | bio_list_init(&punt); | |
383 | bio_list_init(&nopunt); | |
384 | ||
f5fe1b51 | 385 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
df2cb6da | 386 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
f5fe1b51 | 387 | current->bio_list[0] = nopunt; |
df2cb6da | 388 | |
f5fe1b51 N |
389 | bio_list_init(&nopunt); |
390 | while ((bio = bio_list_pop(¤t->bio_list[1]))) | |
391 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); | |
392 | current->bio_list[1] = nopunt; | |
df2cb6da KO |
393 | |
394 | spin_lock(&bs->rescue_lock); | |
395 | bio_list_merge(&bs->rescue_list, &punt); | |
396 | spin_unlock(&bs->rescue_lock); | |
397 | ||
398 | queue_work(bs->rescue_workqueue, &bs->rescue_work); | |
399 | } | |
400 | ||
1da177e4 LT |
401 | /** |
402 | * bio_alloc_bioset - allocate a bio for I/O | |
519c8e9f | 403 | * @gfp_mask: the GFP_* mask given to the slab allocator |
1da177e4 | 404 | * @nr_iovecs: number of iovecs to pre-allocate |
db18efac | 405 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
406 | * |
407 | * Description: | |
3f86a82a KO |
408 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
409 | * backed by the @bs's mempool. | |
410 | * | |
d0164adc MG |
411 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
412 | * always be able to allocate a bio. This is due to the mempool guarantees. | |
413 | * To make this work, callers must never allocate more than 1 bio at a time | |
414 | * from this pool. Callers that need to allocate more than 1 bio must always | |
415 | * submit the previously allocated bio for IO before attempting to allocate | |
416 | * a new one. Failure to do so can cause deadlocks under memory pressure. | |
3f86a82a | 417 | * |
df2cb6da KO |
418 | * Note that when running under generic_make_request() (i.e. any block |
419 | * driver), bios are not submitted until after you return - see the code in | |
420 | * generic_make_request() that converts recursion into iteration, to prevent | |
421 | * stack overflows. | |
422 | * | |
423 | * This would normally mean allocating multiple bios under | |
424 | * generic_make_request() would be susceptible to deadlocks, but we have | |
425 | * deadlock avoidance code that resubmits any blocked bios from a rescuer | |
426 | * thread. | |
427 | * | |
428 | * However, we do not guarantee forward progress for allocations from other | |
429 | * mempools. Doing multiple allocations from the same mempool under | |
430 | * generic_make_request() should be avoided - instead, use bio_set's front_pad | |
431 | * for per bio allocations. | |
432 | * | |
3f86a82a KO |
433 | * RETURNS: |
434 | * Pointer to new bio on success, NULL on failure. | |
435 | */ | |
7a88fa19 DC |
436 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, unsigned int nr_iovecs, |
437 | struct bio_set *bs) | |
1da177e4 | 438 | { |
df2cb6da | 439 | gfp_t saved_gfp = gfp_mask; |
3f86a82a KO |
440 | unsigned front_pad; |
441 | unsigned inline_vecs; | |
34053979 | 442 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
443 | struct bio *bio; |
444 | void *p; | |
445 | ||
3f86a82a KO |
446 | if (!bs) { |
447 | if (nr_iovecs > UIO_MAXIOV) | |
448 | return NULL; | |
449 | ||
450 | p = kmalloc(sizeof(struct bio) + | |
451 | nr_iovecs * sizeof(struct bio_vec), | |
452 | gfp_mask); | |
453 | front_pad = 0; | |
454 | inline_vecs = nr_iovecs; | |
455 | } else { | |
d8f429e1 | 456 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
8aa6ba2f KO |
457 | if (WARN_ON_ONCE(!mempool_initialized(&bs->bvec_pool) && |
458 | nr_iovecs > 0)) | |
d8f429e1 | 459 | return NULL; |
df2cb6da KO |
460 | /* |
461 | * generic_make_request() converts recursion to iteration; this | |
462 | * means if we're running beneath it, any bios we allocate and | |
463 | * submit will not be submitted (and thus freed) until after we | |
464 | * return. | |
465 | * | |
466 | * This exposes us to a potential deadlock if we allocate | |
467 | * multiple bios from the same bio_set() while running | |
468 | * underneath generic_make_request(). If we were to allocate | |
469 | * multiple bios (say a stacking block driver that was splitting | |
470 | * bios), we would deadlock if we exhausted the mempool's | |
471 | * reserve. | |
472 | * | |
473 | * We solve this, and guarantee forward progress, with a rescuer | |
474 | * workqueue per bio_set. If we go to allocate and there are | |
475 | * bios on current->bio_list, we first try the allocation | |
d0164adc MG |
476 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
477 | * bios we would be blocking to the rescuer workqueue before | |
478 | * we retry with the original gfp_flags. | |
df2cb6da KO |
479 | */ |
480 | ||
f5fe1b51 N |
481 | if (current->bio_list && |
482 | (!bio_list_empty(¤t->bio_list[0]) || | |
47e0fb46 N |
483 | !bio_list_empty(¤t->bio_list[1])) && |
484 | bs->rescue_workqueue) | |
d0164adc | 485 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
df2cb6da | 486 | |
8aa6ba2f | 487 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
df2cb6da KO |
488 | if (!p && gfp_mask != saved_gfp) { |
489 | punt_bios_to_rescuer(bs); | |
490 | gfp_mask = saved_gfp; | |
8aa6ba2f | 491 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
df2cb6da KO |
492 | } |
493 | ||
3f86a82a KO |
494 | front_pad = bs->front_pad; |
495 | inline_vecs = BIO_INLINE_VECS; | |
496 | } | |
497 | ||
451a9ebf TH |
498 | if (unlikely(!p)) |
499 | return NULL; | |
1da177e4 | 500 | |
3f86a82a | 501 | bio = p + front_pad; |
3a83f467 | 502 | bio_init(bio, NULL, 0); |
34053979 | 503 | |
3f86a82a | 504 | if (nr_iovecs > inline_vecs) { |
ed996a52 CH |
505 | unsigned long idx = 0; |
506 | ||
8aa6ba2f | 507 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, &bs->bvec_pool); |
df2cb6da KO |
508 | if (!bvl && gfp_mask != saved_gfp) { |
509 | punt_bios_to_rescuer(bs); | |
510 | gfp_mask = saved_gfp; | |
8aa6ba2f | 511 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, &bs->bvec_pool); |
df2cb6da KO |
512 | } |
513 | ||
34053979 IM |
514 | if (unlikely(!bvl)) |
515 | goto err_free; | |
a38352e0 | 516 | |
ed996a52 | 517 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
3f86a82a KO |
518 | } else if (nr_iovecs) { |
519 | bvl = bio->bi_inline_vecs; | |
1da177e4 | 520 | } |
3f86a82a KO |
521 | |
522 | bio->bi_pool = bs; | |
34053979 | 523 | bio->bi_max_vecs = nr_iovecs; |
34053979 | 524 | bio->bi_io_vec = bvl; |
1da177e4 | 525 | return bio; |
34053979 IM |
526 | |
527 | err_free: | |
8aa6ba2f | 528 | mempool_free(p, &bs->bio_pool); |
34053979 | 529 | return NULL; |
1da177e4 | 530 | } |
a112a71d | 531 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 532 | |
1da177e4 LT |
533 | void zero_fill_bio(struct bio *bio) |
534 | { | |
535 | unsigned long flags; | |
7988613b KO |
536 | struct bio_vec bv; |
537 | struct bvec_iter iter; | |
1da177e4 | 538 | |
7988613b KO |
539 | bio_for_each_segment(bv, bio, iter) { |
540 | char *data = bvec_kmap_irq(&bv, &flags); | |
541 | memset(data, 0, bv.bv_len); | |
542 | flush_dcache_page(bv.bv_page); | |
1da177e4 LT |
543 | bvec_kunmap_irq(data, &flags); |
544 | } | |
545 | } | |
546 | EXPORT_SYMBOL(zero_fill_bio); | |
547 | ||
548 | /** | |
549 | * bio_put - release a reference to a bio | |
550 | * @bio: bio to release reference to | |
551 | * | |
552 | * Description: | |
553 | * Put a reference to a &struct bio, either one you have gotten with | |
9b10f6a9 | 554 | * bio_alloc, bio_get or bio_clone_*. The last put of a bio will free it. |
1da177e4 LT |
555 | **/ |
556 | void bio_put(struct bio *bio) | |
557 | { | |
dac56212 | 558 | if (!bio_flagged(bio, BIO_REFFED)) |
4254bba1 | 559 | bio_free(bio); |
dac56212 JA |
560 | else { |
561 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); | |
562 | ||
563 | /* | |
564 | * last put frees it | |
565 | */ | |
566 | if (atomic_dec_and_test(&bio->__bi_cnt)) | |
567 | bio_free(bio); | |
568 | } | |
1da177e4 | 569 | } |
a112a71d | 570 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 571 | |
165125e1 | 572 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
573 | { |
574 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
575 | blk_recount_segments(q, bio); | |
576 | ||
577 | return bio->bi_phys_segments; | |
578 | } | |
a112a71d | 579 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 580 | |
59d276fe KO |
581 | /** |
582 | * __bio_clone_fast - clone a bio that shares the original bio's biovec | |
583 | * @bio: destination bio | |
584 | * @bio_src: bio to clone | |
585 | * | |
586 | * Clone a &bio. Caller will own the returned bio, but not | |
587 | * the actual data it points to. Reference count of returned | |
588 | * bio will be one. | |
589 | * | |
590 | * Caller must ensure that @bio_src is not freed before @bio. | |
591 | */ | |
592 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) | |
593 | { | |
ed996a52 | 594 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
59d276fe KO |
595 | |
596 | /* | |
74d46992 | 597 | * most users will be overriding ->bi_disk with a new target, |
59d276fe KO |
598 | * so we don't set nor calculate new physical/hw segment counts here |
599 | */ | |
74d46992 | 600 | bio->bi_disk = bio_src->bi_disk; |
62530ed8 | 601 | bio->bi_partno = bio_src->bi_partno; |
b7c44ed9 | 602 | bio_set_flag(bio, BIO_CLONED); |
111be883 SL |
603 | if (bio_flagged(bio_src, BIO_THROTTLED)) |
604 | bio_set_flag(bio, BIO_THROTTLED); | |
1eff9d32 | 605 | bio->bi_opf = bio_src->bi_opf; |
cb6934f8 | 606 | bio->bi_write_hint = bio_src->bi_write_hint; |
59d276fe KO |
607 | bio->bi_iter = bio_src->bi_iter; |
608 | bio->bi_io_vec = bio_src->bi_io_vec; | |
20bd723e PV |
609 | |
610 | bio_clone_blkcg_association(bio, bio_src); | |
59d276fe KO |
611 | } |
612 | EXPORT_SYMBOL(__bio_clone_fast); | |
613 | ||
614 | /** | |
615 | * bio_clone_fast - clone a bio that shares the original bio's biovec | |
616 | * @bio: bio to clone | |
617 | * @gfp_mask: allocation priority | |
618 | * @bs: bio_set to allocate from | |
619 | * | |
620 | * Like __bio_clone_fast, only also allocates the returned bio | |
621 | */ | |
622 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) | |
623 | { | |
624 | struct bio *b; | |
625 | ||
626 | b = bio_alloc_bioset(gfp_mask, 0, bs); | |
627 | if (!b) | |
628 | return NULL; | |
629 | ||
630 | __bio_clone_fast(b, bio); | |
631 | ||
632 | if (bio_integrity(bio)) { | |
633 | int ret; | |
634 | ||
635 | ret = bio_integrity_clone(b, bio, gfp_mask); | |
636 | ||
637 | if (ret < 0) { | |
638 | bio_put(b); | |
639 | return NULL; | |
640 | } | |
641 | } | |
642 | ||
643 | return b; | |
644 | } | |
645 | EXPORT_SYMBOL(bio_clone_fast); | |
646 | ||
f4595875 SL |
647 | /** |
648 | * bio_clone_bioset - clone a bio | |
649 | * @bio_src: bio to clone | |
650 | * @gfp_mask: allocation priority | |
651 | * @bs: bio_set to allocate from | |
652 | * | |
653 | * Clone bio. Caller will own the returned bio, but not the actual data it | |
654 | * points to. Reference count of returned bio will be one. | |
655 | */ | |
656 | struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask, | |
657 | struct bio_set *bs) | |
1da177e4 | 658 | { |
bdb53207 KO |
659 | struct bvec_iter iter; |
660 | struct bio_vec bv; | |
661 | struct bio *bio; | |
1da177e4 | 662 | |
bdb53207 KO |
663 | /* |
664 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from | |
665 | * bio_src->bi_io_vec to bio->bi_io_vec. | |
666 | * | |
667 | * We can't do that anymore, because: | |
668 | * | |
669 | * - The point of cloning the biovec is to produce a bio with a biovec | |
670 | * the caller can modify: bi_idx and bi_bvec_done should be 0. | |
671 | * | |
672 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if | |
673 | * we tried to clone the whole thing bio_alloc_bioset() would fail. | |
674 | * But the clone should succeed as long as the number of biovecs we | |
675 | * actually need to allocate is fewer than BIO_MAX_PAGES. | |
676 | * | |
677 | * - Lastly, bi_vcnt should not be looked at or relied upon by code | |
678 | * that does not own the bio - reason being drivers don't use it for | |
679 | * iterating over the biovec anymore, so expecting it to be kept up | |
680 | * to date (i.e. for clones that share the parent biovec) is just | |
681 | * asking for trouble and would force extra work on | |
682 | * __bio_clone_fast() anyways. | |
683 | */ | |
684 | ||
f4595875 | 685 | bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs); |
bdb53207 | 686 | if (!bio) |
7ba1ba12 | 687 | return NULL; |
74d46992 | 688 | bio->bi_disk = bio_src->bi_disk; |
1eff9d32 | 689 | bio->bi_opf = bio_src->bi_opf; |
cb6934f8 | 690 | bio->bi_write_hint = bio_src->bi_write_hint; |
bdb53207 KO |
691 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; |
692 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; | |
7ba1ba12 | 693 | |
7afafc8a AH |
694 | switch (bio_op(bio)) { |
695 | case REQ_OP_DISCARD: | |
696 | case REQ_OP_SECURE_ERASE: | |
a6f0788e | 697 | case REQ_OP_WRITE_ZEROES: |
7afafc8a AH |
698 | break; |
699 | case REQ_OP_WRITE_SAME: | |
8423ae3d | 700 | bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0]; |
7afafc8a AH |
701 | break; |
702 | default: | |
f4595875 | 703 | bio_for_each_segment(bv, bio_src, iter) |
7afafc8a AH |
704 | bio->bi_io_vec[bio->bi_vcnt++] = bv; |
705 | break; | |
8423ae3d KO |
706 | } |
707 | ||
bdb53207 KO |
708 | if (bio_integrity(bio_src)) { |
709 | int ret; | |
7ba1ba12 | 710 | |
bdb53207 | 711 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
059ea331 | 712 | if (ret < 0) { |
bdb53207 | 713 | bio_put(bio); |
7ba1ba12 | 714 | return NULL; |
059ea331 | 715 | } |
3676347a | 716 | } |
1da177e4 | 717 | |
20bd723e PV |
718 | bio_clone_blkcg_association(bio, bio_src); |
719 | ||
bdb53207 | 720 | return bio; |
1da177e4 | 721 | } |
bf800ef1 | 722 | EXPORT_SYMBOL(bio_clone_bioset); |
1da177e4 LT |
723 | |
724 | /** | |
c66a14d0 KO |
725 | * bio_add_pc_page - attempt to add page to bio |
726 | * @q: the target queue | |
727 | * @bio: destination bio | |
728 | * @page: page to add | |
729 | * @len: vec entry length | |
730 | * @offset: vec entry offset | |
1da177e4 | 731 | * |
c66a14d0 KO |
732 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
733 | * number of reasons, such as the bio being full or target block device | |
734 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
735 | * so it is always possible to add a single page to an empty bio. | |
736 | * | |
737 | * This should only be used by REQ_PC bios. | |
1da177e4 | 738 | */ |
c66a14d0 KO |
739 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
740 | *page, unsigned int len, unsigned int offset) | |
1da177e4 LT |
741 | { |
742 | int retried_segments = 0; | |
743 | struct bio_vec *bvec; | |
744 | ||
745 | /* | |
746 | * cloned bio must not modify vec list | |
747 | */ | |
748 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
749 | return 0; | |
750 | ||
c66a14d0 | 751 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
752 | return 0; |
753 | ||
80cfd548 JA |
754 | /* |
755 | * For filesystems with a blocksize smaller than the pagesize | |
756 | * we will often be called with the same page as last time and | |
757 | * a consecutive offset. Optimize this special case. | |
758 | */ | |
759 | if (bio->bi_vcnt > 0) { | |
760 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
761 | ||
762 | if (page == prev->bv_page && | |
763 | offset == prev->bv_offset + prev->bv_len) { | |
764 | prev->bv_len += len; | |
fcbf6a08 | 765 | bio->bi_iter.bi_size += len; |
80cfd548 JA |
766 | goto done; |
767 | } | |
66cb45aa JA |
768 | |
769 | /* | |
770 | * If the queue doesn't support SG gaps and adding this | |
771 | * offset would create a gap, disallow it. | |
772 | */ | |
03100aad | 773 | if (bvec_gap_to_prev(q, prev, offset)) |
66cb45aa | 774 | return 0; |
80cfd548 JA |
775 | } |
776 | ||
777 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
778 | return 0; |
779 | ||
780 | /* | |
fcbf6a08 ML |
781 | * setup the new entry, we might clear it again later if we |
782 | * cannot add the page | |
783 | */ | |
784 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
785 | bvec->bv_page = page; | |
786 | bvec->bv_len = len; | |
787 | bvec->bv_offset = offset; | |
788 | bio->bi_vcnt++; | |
789 | bio->bi_phys_segments++; | |
790 | bio->bi_iter.bi_size += len; | |
791 | ||
792 | /* | |
793 | * Perform a recount if the number of segments is greater | |
794 | * than queue_max_segments(q). | |
1da177e4 LT |
795 | */ |
796 | ||
fcbf6a08 | 797 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
1da177e4 LT |
798 | |
799 | if (retried_segments) | |
fcbf6a08 | 800 | goto failed; |
1da177e4 LT |
801 | |
802 | retried_segments = 1; | |
803 | blk_recount_segments(q, bio); | |
804 | } | |
805 | ||
1da177e4 | 806 | /* If we may be able to merge these biovecs, force a recount */ |
fcbf6a08 | 807 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
b7c44ed9 | 808 | bio_clear_flag(bio, BIO_SEG_VALID); |
1da177e4 | 809 | |
80cfd548 | 810 | done: |
1da177e4 | 811 | return len; |
fcbf6a08 ML |
812 | |
813 | failed: | |
814 | bvec->bv_page = NULL; | |
815 | bvec->bv_len = 0; | |
816 | bvec->bv_offset = 0; | |
817 | bio->bi_vcnt--; | |
818 | bio->bi_iter.bi_size -= len; | |
819 | blk_recount_segments(q, bio); | |
820 | return 0; | |
1da177e4 | 821 | } |
a112a71d | 822 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 823 | |
1da177e4 LT |
824 | /** |
825 | * bio_add_page - attempt to add page to bio | |
826 | * @bio: destination bio | |
827 | * @page: page to add | |
828 | * @len: vec entry length | |
829 | * @offset: vec entry offset | |
830 | * | |
c66a14d0 KO |
831 | * Attempt to add a page to the bio_vec maplist. This will only fail |
832 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. | |
1da177e4 | 833 | */ |
c66a14d0 KO |
834 | int bio_add_page(struct bio *bio, struct page *page, |
835 | unsigned int len, unsigned int offset) | |
1da177e4 | 836 | { |
c66a14d0 KO |
837 | struct bio_vec *bv; |
838 | ||
839 | /* | |
840 | * cloned bio must not modify vec list | |
841 | */ | |
842 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) | |
843 | return 0; | |
762380ad | 844 | |
c66a14d0 KO |
845 | /* |
846 | * For filesystems with a blocksize smaller than the pagesize | |
847 | * we will often be called with the same page as last time and | |
848 | * a consecutive offset. Optimize this special case. | |
849 | */ | |
850 | if (bio->bi_vcnt > 0) { | |
851 | bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
58a4915a | 852 | |
c66a14d0 KO |
853 | if (page == bv->bv_page && |
854 | offset == bv->bv_offset + bv->bv_len) { | |
855 | bv->bv_len += len; | |
856 | goto done; | |
857 | } | |
858 | } | |
859 | ||
860 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
861 | return 0; | |
862 | ||
863 | bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
864 | bv->bv_page = page; | |
865 | bv->bv_len = len; | |
866 | bv->bv_offset = offset; | |
867 | ||
868 | bio->bi_vcnt++; | |
869 | done: | |
870 | bio->bi_iter.bi_size += len; | |
871 | return len; | |
1da177e4 | 872 | } |
a112a71d | 873 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 874 | |
2cefe4db KO |
875 | /** |
876 | * bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio | |
877 | * @bio: bio to add pages to | |
878 | * @iter: iov iterator describing the region to be mapped | |
879 | * | |
880 | * Pins as many pages from *iter and appends them to @bio's bvec array. The | |
881 | * pages will have to be released using put_page() when done. | |
882 | */ | |
883 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) | |
884 | { | |
885 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; | |
886 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; | |
887 | struct page **pages = (struct page **)bv; | |
888 | size_t offset, diff; | |
889 | ssize_t size; | |
890 | ||
891 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); | |
892 | if (unlikely(size <= 0)) | |
893 | return size ? size : -EFAULT; | |
894 | nr_pages = (size + offset + PAGE_SIZE - 1) / PAGE_SIZE; | |
895 | ||
896 | /* | |
897 | * Deep magic below: We need to walk the pinned pages backwards | |
898 | * because we are abusing the space allocated for the bio_vecs | |
899 | * for the page array. Because the bio_vecs are larger than the | |
900 | * page pointers by definition this will always work. But it also | |
901 | * means we can't use bio_add_page, so any changes to it's semantics | |
902 | * need to be reflected here as well. | |
903 | */ | |
904 | bio->bi_iter.bi_size += size; | |
905 | bio->bi_vcnt += nr_pages; | |
906 | ||
907 | diff = (nr_pages * PAGE_SIZE - offset) - size; | |
908 | while (nr_pages--) { | |
909 | bv[nr_pages].bv_page = pages[nr_pages]; | |
910 | bv[nr_pages].bv_len = PAGE_SIZE; | |
911 | bv[nr_pages].bv_offset = 0; | |
912 | } | |
913 | ||
914 | bv[0].bv_offset += offset; | |
915 | bv[0].bv_len -= offset; | |
916 | if (diff) | |
917 | bv[bio->bi_vcnt - 1].bv_len -= diff; | |
918 | ||
919 | iov_iter_advance(iter, size); | |
920 | return 0; | |
921 | } | |
922 | EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages); | |
923 | ||
4246a0b6 | 924 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 | 925 | { |
65e53aab | 926 | complete(bio->bi_private); |
9e882242 KO |
927 | } |
928 | ||
929 | /** | |
930 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
931 | * @bio: The &struct bio which describes the I/O |
932 | * | |
933 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
934 | * bio_endio() on failure. | |
3d289d68 JK |
935 | * |
936 | * WARNING: Unlike to how submit_bio() is usually used, this function does not | |
937 | * result in bio reference to be consumed. The caller must drop the reference | |
938 | * on his own. | |
9e882242 | 939 | */ |
4e49ea4a | 940 | int submit_bio_wait(struct bio *bio) |
9e882242 | 941 | { |
e319e1fb | 942 | DECLARE_COMPLETION_ONSTACK_MAP(done, bio->bi_disk->lockdep_map); |
9e882242 | 943 | |
65e53aab | 944 | bio->bi_private = &done; |
9e882242 | 945 | bio->bi_end_io = submit_bio_wait_endio; |
1eff9d32 | 946 | bio->bi_opf |= REQ_SYNC; |
4e49ea4a | 947 | submit_bio(bio); |
65e53aab | 948 | wait_for_completion_io(&done); |
9e882242 | 949 | |
65e53aab | 950 | return blk_status_to_errno(bio->bi_status); |
9e882242 KO |
951 | } |
952 | EXPORT_SYMBOL(submit_bio_wait); | |
953 | ||
054bdf64 KO |
954 | /** |
955 | * bio_advance - increment/complete a bio by some number of bytes | |
956 | * @bio: bio to advance | |
957 | * @bytes: number of bytes to complete | |
958 | * | |
959 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
960 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
961 | * be updated on the last bvec as well. | |
962 | * | |
963 | * @bio will then represent the remaining, uncompleted portion of the io. | |
964 | */ | |
965 | void bio_advance(struct bio *bio, unsigned bytes) | |
966 | { | |
967 | if (bio_integrity(bio)) | |
968 | bio_integrity_advance(bio, bytes); | |
969 | ||
4550dd6c | 970 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
971 | } |
972 | EXPORT_SYMBOL(bio_advance); | |
973 | ||
16ac3d63 KO |
974 | /** |
975 | * bio_copy_data - copy contents of data buffers from one chain of bios to | |
976 | * another | |
977 | * @src: source bio list | |
978 | * @dst: destination bio list | |
979 | * | |
980 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats | |
981 | * @src and @dst as linked lists of bios. | |
982 | * | |
983 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
984 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
985 | */ | |
986 | void bio_copy_data(struct bio *dst, struct bio *src) | |
987 | { | |
1cb9dda4 KO |
988 | struct bvec_iter src_iter, dst_iter; |
989 | struct bio_vec src_bv, dst_bv; | |
16ac3d63 | 990 | void *src_p, *dst_p; |
1cb9dda4 | 991 | unsigned bytes; |
16ac3d63 | 992 | |
1cb9dda4 KO |
993 | src_iter = src->bi_iter; |
994 | dst_iter = dst->bi_iter; | |
16ac3d63 KO |
995 | |
996 | while (1) { | |
1cb9dda4 KO |
997 | if (!src_iter.bi_size) { |
998 | src = src->bi_next; | |
999 | if (!src) | |
1000 | break; | |
16ac3d63 | 1001 | |
1cb9dda4 | 1002 | src_iter = src->bi_iter; |
16ac3d63 KO |
1003 | } |
1004 | ||
1cb9dda4 KO |
1005 | if (!dst_iter.bi_size) { |
1006 | dst = dst->bi_next; | |
1007 | if (!dst) | |
1008 | break; | |
16ac3d63 | 1009 | |
1cb9dda4 | 1010 | dst_iter = dst->bi_iter; |
16ac3d63 KO |
1011 | } |
1012 | ||
1cb9dda4 KO |
1013 | src_bv = bio_iter_iovec(src, src_iter); |
1014 | dst_bv = bio_iter_iovec(dst, dst_iter); | |
1015 | ||
1016 | bytes = min(src_bv.bv_len, dst_bv.bv_len); | |
16ac3d63 | 1017 | |
1cb9dda4 KO |
1018 | src_p = kmap_atomic(src_bv.bv_page); |
1019 | dst_p = kmap_atomic(dst_bv.bv_page); | |
16ac3d63 | 1020 | |
1cb9dda4 KO |
1021 | memcpy(dst_p + dst_bv.bv_offset, |
1022 | src_p + src_bv.bv_offset, | |
16ac3d63 KO |
1023 | bytes); |
1024 | ||
1025 | kunmap_atomic(dst_p); | |
1026 | kunmap_atomic(src_p); | |
1027 | ||
1cb9dda4 KO |
1028 | bio_advance_iter(src, &src_iter, bytes); |
1029 | bio_advance_iter(dst, &dst_iter, bytes); | |
16ac3d63 KO |
1030 | } |
1031 | } | |
1032 | EXPORT_SYMBOL(bio_copy_data); | |
1033 | ||
1da177e4 | 1034 | struct bio_map_data { |
152e283f | 1035 | int is_our_pages; |
26e49cfc KO |
1036 | struct iov_iter iter; |
1037 | struct iovec iov[]; | |
1da177e4 LT |
1038 | }; |
1039 | ||
0e5b935d | 1040 | static struct bio_map_data *bio_alloc_map_data(struct iov_iter *data, |
76029ff3 | 1041 | gfp_t gfp_mask) |
1da177e4 | 1042 | { |
0e5b935d AV |
1043 | struct bio_map_data *bmd; |
1044 | if (data->nr_segs > UIO_MAXIOV) | |
f3f63c1c | 1045 | return NULL; |
1da177e4 | 1046 | |
0e5b935d AV |
1047 | bmd = kmalloc(sizeof(struct bio_map_data) + |
1048 | sizeof(struct iovec) * data->nr_segs, gfp_mask); | |
1049 | if (!bmd) | |
1050 | return NULL; | |
1051 | memcpy(bmd->iov, data->iov, sizeof(struct iovec) * data->nr_segs); | |
1052 | bmd->iter = *data; | |
1053 | bmd->iter.iov = bmd->iov; | |
1054 | return bmd; | |
1da177e4 LT |
1055 | } |
1056 | ||
9124d3fe DP |
1057 | /** |
1058 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1059 | * @bio: The &struct bio which describes the I/O as destination | |
1060 | * @iter: iov_iter as source | |
1061 | * | |
1062 | * Copy all pages from iov_iter to bio. | |
1063 | * Returns 0 on success, or error on failure. | |
1064 | */ | |
98a09d61 | 1065 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter *iter) |
c5dec1c3 | 1066 | { |
9124d3fe | 1067 | int i; |
c5dec1c3 | 1068 | struct bio_vec *bvec; |
c5dec1c3 | 1069 | |
d74c6d51 | 1070 | bio_for_each_segment_all(bvec, bio, i) { |
9124d3fe | 1071 | ssize_t ret; |
c5dec1c3 | 1072 | |
9124d3fe DP |
1073 | ret = copy_page_from_iter(bvec->bv_page, |
1074 | bvec->bv_offset, | |
1075 | bvec->bv_len, | |
98a09d61 | 1076 | iter); |
9124d3fe | 1077 | |
98a09d61 | 1078 | if (!iov_iter_count(iter)) |
9124d3fe DP |
1079 | break; |
1080 | ||
1081 | if (ret < bvec->bv_len) | |
1082 | return -EFAULT; | |
c5dec1c3 FT |
1083 | } |
1084 | ||
9124d3fe DP |
1085 | return 0; |
1086 | } | |
1087 | ||
1088 | /** | |
1089 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1090 | * @bio: The &struct bio which describes the I/O as source | |
1091 | * @iter: iov_iter as destination | |
1092 | * | |
1093 | * Copy all pages from bio to iov_iter. | |
1094 | * Returns 0 on success, or error on failure. | |
1095 | */ | |
1096 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1097 | { | |
1098 | int i; | |
1099 | struct bio_vec *bvec; | |
1100 | ||
1101 | bio_for_each_segment_all(bvec, bio, i) { | |
1102 | ssize_t ret; | |
1103 | ||
1104 | ret = copy_page_to_iter(bvec->bv_page, | |
1105 | bvec->bv_offset, | |
1106 | bvec->bv_len, | |
1107 | &iter); | |
1108 | ||
1109 | if (!iov_iter_count(&iter)) | |
1110 | break; | |
1111 | ||
1112 | if (ret < bvec->bv_len) | |
1113 | return -EFAULT; | |
1114 | } | |
1115 | ||
1116 | return 0; | |
c5dec1c3 FT |
1117 | } |
1118 | ||
491221f8 | 1119 | void bio_free_pages(struct bio *bio) |
1dfa0f68 CH |
1120 | { |
1121 | struct bio_vec *bvec; | |
1122 | int i; | |
1123 | ||
1124 | bio_for_each_segment_all(bvec, bio, i) | |
1125 | __free_page(bvec->bv_page); | |
1126 | } | |
491221f8 | 1127 | EXPORT_SYMBOL(bio_free_pages); |
1dfa0f68 | 1128 | |
1da177e4 LT |
1129 | /** |
1130 | * bio_uncopy_user - finish previously mapped bio | |
1131 | * @bio: bio being terminated | |
1132 | * | |
ddad8dd0 | 1133 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1134 | * to user space in case of a read. |
1135 | */ | |
1136 | int bio_uncopy_user(struct bio *bio) | |
1137 | { | |
1138 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1139 | int ret = 0; |
1da177e4 | 1140 | |
35dc2483 RD |
1141 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1142 | /* | |
1143 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1144 | * don't copy into a random user address space, just free |
1145 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1146 | */ |
2d99b55d HR |
1147 | if (!current->mm) |
1148 | ret = -EINTR; | |
1149 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1150 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1151 | if (bmd->is_our_pages) |
1152 | bio_free_pages(bio); | |
35dc2483 | 1153 | } |
c8db4448 | 1154 | kfree(bmd); |
1da177e4 LT |
1155 | bio_put(bio); |
1156 | return ret; | |
1157 | } | |
1158 | ||
1159 | /** | |
c5dec1c3 | 1160 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1161 | * @q: destination block queue |
1162 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1163 | * @iter: iovec iterator | |
1164 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1165 | * |
1166 | * Prepares and returns a bio for indirect user io, bouncing data | |
1167 | * to/from kernel pages as necessary. Must be paired with | |
1168 | * call bio_uncopy_user() on io completion. | |
1169 | */ | |
152e283f FT |
1170 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1171 | struct rq_map_data *map_data, | |
e81cef5d | 1172 | struct iov_iter *iter, |
26e49cfc | 1173 | gfp_t gfp_mask) |
1da177e4 | 1174 | { |
1da177e4 | 1175 | struct bio_map_data *bmd; |
1da177e4 LT |
1176 | struct page *page; |
1177 | struct bio *bio; | |
d16d44eb AV |
1178 | int i = 0, ret; |
1179 | int nr_pages; | |
26e49cfc | 1180 | unsigned int len = iter->count; |
bd5cecea | 1181 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1182 | |
0e5b935d | 1183 | bmd = bio_alloc_map_data(iter, gfp_mask); |
1da177e4 LT |
1184 | if (!bmd) |
1185 | return ERR_PTR(-ENOMEM); | |
1186 | ||
26e49cfc KO |
1187 | /* |
1188 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1189 | * The caller provided iov might point to an on-stack or otherwise | |
1190 | * shortlived one. | |
1191 | */ | |
1192 | bmd->is_our_pages = map_data ? 0 : 1; | |
26e49cfc | 1193 | |
d16d44eb AV |
1194 | nr_pages = DIV_ROUND_UP(offset + len, PAGE_SIZE); |
1195 | if (nr_pages > BIO_MAX_PAGES) | |
1196 | nr_pages = BIO_MAX_PAGES; | |
26e49cfc | 1197 | |
1da177e4 | 1198 | ret = -ENOMEM; |
a9e9dc24 | 1199 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1200 | if (!bio) |
1201 | goto out_bmd; | |
1202 | ||
1da177e4 | 1203 | ret = 0; |
56c451f4 FT |
1204 | |
1205 | if (map_data) { | |
e623ddb4 | 1206 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1207 | i = map_data->offset / PAGE_SIZE; |
1208 | } | |
1da177e4 | 1209 | while (len) { |
e623ddb4 | 1210 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1211 | |
56c451f4 FT |
1212 | bytes -= offset; |
1213 | ||
1da177e4 LT |
1214 | if (bytes > len) |
1215 | bytes = len; | |
1216 | ||
152e283f | 1217 | if (map_data) { |
e623ddb4 | 1218 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1219 | ret = -ENOMEM; |
1220 | break; | |
1221 | } | |
e623ddb4 FT |
1222 | |
1223 | page = map_data->pages[i / nr_pages]; | |
1224 | page += (i % nr_pages); | |
1225 | ||
1226 | i++; | |
1227 | } else { | |
152e283f | 1228 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1229 | if (!page) { |
1230 | ret = -ENOMEM; | |
1231 | break; | |
1232 | } | |
1da177e4 LT |
1233 | } |
1234 | ||
56c451f4 | 1235 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 1236 | break; |
1da177e4 LT |
1237 | |
1238 | len -= bytes; | |
56c451f4 | 1239 | offset = 0; |
1da177e4 LT |
1240 | } |
1241 | ||
1242 | if (ret) | |
1243 | goto cleanup; | |
1244 | ||
2884d0be AV |
1245 | if (map_data) |
1246 | map_data->offset += bio->bi_iter.bi_size; | |
1247 | ||
1da177e4 LT |
1248 | /* |
1249 | * success | |
1250 | */ | |
26e49cfc | 1251 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1252 | (map_data && map_data->from_user)) { |
98a09d61 | 1253 | ret = bio_copy_from_iter(bio, iter); |
c5dec1c3 FT |
1254 | if (ret) |
1255 | goto cleanup; | |
98a09d61 AV |
1256 | } else { |
1257 | iov_iter_advance(iter, bio->bi_iter.bi_size); | |
1da177e4 LT |
1258 | } |
1259 | ||
26e49cfc | 1260 | bio->bi_private = bmd; |
2884d0be AV |
1261 | if (map_data && map_data->null_mapped) |
1262 | bio_set_flag(bio, BIO_NULL_MAPPED); | |
1da177e4 LT |
1263 | return bio; |
1264 | cleanup: | |
152e283f | 1265 | if (!map_data) |
1dfa0f68 | 1266 | bio_free_pages(bio); |
1da177e4 LT |
1267 | bio_put(bio); |
1268 | out_bmd: | |
c8db4448 | 1269 | kfree(bmd); |
1da177e4 LT |
1270 | return ERR_PTR(ret); |
1271 | } | |
1272 | ||
37f19e57 CH |
1273 | /** |
1274 | * bio_map_user_iov - map user iovec into bio | |
1275 | * @q: the struct request_queue for the bio | |
1276 | * @iter: iovec iterator | |
1277 | * @gfp_mask: memory allocation flags | |
1278 | * | |
1279 | * Map the user space address into a bio suitable for io to a block | |
1280 | * device. Returns an error pointer in case of error. | |
1281 | */ | |
1282 | struct bio *bio_map_user_iov(struct request_queue *q, | |
e81cef5d | 1283 | struct iov_iter *iter, |
37f19e57 | 1284 | gfp_t gfp_mask) |
1da177e4 | 1285 | { |
26e49cfc | 1286 | int j; |
1da177e4 | 1287 | struct bio *bio; |
076098e5 | 1288 | int ret; |
2b04e8f6 | 1289 | struct bio_vec *bvec; |
1da177e4 | 1290 | |
b282cc76 | 1291 | if (!iov_iter_count(iter)) |
1da177e4 LT |
1292 | return ERR_PTR(-EINVAL); |
1293 | ||
b282cc76 | 1294 | bio = bio_kmalloc(gfp_mask, iov_iter_npages(iter, BIO_MAX_PAGES)); |
1da177e4 LT |
1295 | if (!bio) |
1296 | return ERR_PTR(-ENOMEM); | |
1297 | ||
0a0f1513 | 1298 | while (iov_iter_count(iter)) { |
629e42bc | 1299 | struct page **pages; |
076098e5 AV |
1300 | ssize_t bytes; |
1301 | size_t offs, added = 0; | |
1302 | int npages; | |
1da177e4 | 1303 | |
0a0f1513 | 1304 | bytes = iov_iter_get_pages_alloc(iter, &pages, LONG_MAX, &offs); |
076098e5 AV |
1305 | if (unlikely(bytes <= 0)) { |
1306 | ret = bytes ? bytes : -EFAULT; | |
f1970baf | 1307 | goto out_unmap; |
99172157 | 1308 | } |
f1970baf | 1309 | |
076098e5 | 1310 | npages = DIV_ROUND_UP(offs + bytes, PAGE_SIZE); |
f1970baf | 1311 | |
98f0bc99 AV |
1312 | if (unlikely(offs & queue_dma_alignment(q))) { |
1313 | ret = -EINVAL; | |
1314 | j = 0; | |
1315 | } else { | |
1316 | for (j = 0; j < npages; j++) { | |
1317 | struct page *page = pages[j]; | |
1318 | unsigned int n = PAGE_SIZE - offs; | |
1319 | unsigned short prev_bi_vcnt = bio->bi_vcnt; | |
f1970baf | 1320 | |
98f0bc99 AV |
1321 | if (n > bytes) |
1322 | n = bytes; | |
95d78c28 | 1323 | |
98f0bc99 AV |
1324 | if (!bio_add_pc_page(q, bio, page, n, offs)) |
1325 | break; | |
1da177e4 | 1326 | |
98f0bc99 AV |
1327 | /* |
1328 | * check if vector was merged with previous | |
1329 | * drop page reference if needed | |
1330 | */ | |
1331 | if (bio->bi_vcnt == prev_bi_vcnt) | |
1332 | put_page(page); | |
1333 | ||
1334 | added += n; | |
1335 | bytes -= n; | |
1336 | offs = 0; | |
1337 | } | |
0a0f1513 | 1338 | iov_iter_advance(iter, added); |
f1970baf | 1339 | } |
1da177e4 | 1340 | /* |
f1970baf | 1341 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1342 | */ |
629e42bc | 1343 | while (j < npages) |
09cbfeaf | 1344 | put_page(pages[j++]); |
629e42bc | 1345 | kvfree(pages); |
e2e115d1 AV |
1346 | /* couldn't stuff something into bio? */ |
1347 | if (bytes) | |
1348 | break; | |
1da177e4 LT |
1349 | } |
1350 | ||
b7c44ed9 | 1351 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1352 | |
1353 | /* | |
5fad1b64 | 1354 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
37f19e57 CH |
1355 | * it would normally disappear when its bi_end_io is run. |
1356 | * however, we need it for the unmap, so grab an extra | |
1357 | * reference to it | |
1358 | */ | |
1359 | bio_get(bio); | |
1da177e4 | 1360 | return bio; |
f1970baf JB |
1361 | |
1362 | out_unmap: | |
2b04e8f6 AV |
1363 | bio_for_each_segment_all(bvec, bio, j) { |
1364 | put_page(bvec->bv_page); | |
f1970baf | 1365 | } |
1da177e4 LT |
1366 | bio_put(bio); |
1367 | return ERR_PTR(ret); | |
1368 | } | |
1369 | ||
1da177e4 LT |
1370 | static void __bio_unmap_user(struct bio *bio) |
1371 | { | |
1372 | struct bio_vec *bvec; | |
1373 | int i; | |
1374 | ||
1375 | /* | |
1376 | * make sure we dirty pages we wrote to | |
1377 | */ | |
d74c6d51 | 1378 | bio_for_each_segment_all(bvec, bio, i) { |
1da177e4 LT |
1379 | if (bio_data_dir(bio) == READ) |
1380 | set_page_dirty_lock(bvec->bv_page); | |
1381 | ||
09cbfeaf | 1382 | put_page(bvec->bv_page); |
1da177e4 LT |
1383 | } |
1384 | ||
1385 | bio_put(bio); | |
1386 | } | |
1387 | ||
1388 | /** | |
1389 | * bio_unmap_user - unmap a bio | |
1390 | * @bio: the bio being unmapped | |
1391 | * | |
5fad1b64 BVA |
1392 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
1393 | * process context. | |
1da177e4 LT |
1394 | * |
1395 | * bio_unmap_user() may sleep. | |
1396 | */ | |
1397 | void bio_unmap_user(struct bio *bio) | |
1398 | { | |
1399 | __bio_unmap_user(bio); | |
1400 | bio_put(bio); | |
1401 | } | |
1402 | ||
4246a0b6 | 1403 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1404 | { |
b823825e | 1405 | bio_put(bio); |
b823825e JA |
1406 | } |
1407 | ||
75c72b83 CH |
1408 | /** |
1409 | * bio_map_kern - map kernel address into bio | |
1410 | * @q: the struct request_queue for the bio | |
1411 | * @data: pointer to buffer to map | |
1412 | * @len: length in bytes | |
1413 | * @gfp_mask: allocation flags for bio allocation | |
1414 | * | |
1415 | * Map the kernel address into a bio suitable for io to a block | |
1416 | * device. Returns an error pointer in case of error. | |
1417 | */ | |
1418 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1419 | gfp_t gfp_mask) | |
df46b9a4 MC |
1420 | { |
1421 | unsigned long kaddr = (unsigned long)data; | |
1422 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1423 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1424 | const int nr_pages = end - start; | |
1425 | int offset, i; | |
1426 | struct bio *bio; | |
1427 | ||
a9e9dc24 | 1428 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1429 | if (!bio) |
1430 | return ERR_PTR(-ENOMEM); | |
1431 | ||
1432 | offset = offset_in_page(kaddr); | |
1433 | for (i = 0; i < nr_pages; i++) { | |
1434 | unsigned int bytes = PAGE_SIZE - offset; | |
1435 | ||
1436 | if (len <= 0) | |
1437 | break; | |
1438 | ||
1439 | if (bytes > len) | |
1440 | bytes = len; | |
1441 | ||
defd94b7 | 1442 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1443 | offset) < bytes) { |
1444 | /* we don't support partial mappings */ | |
1445 | bio_put(bio); | |
1446 | return ERR_PTR(-EINVAL); | |
1447 | } | |
df46b9a4 MC |
1448 | |
1449 | data += bytes; | |
1450 | len -= bytes; | |
1451 | offset = 0; | |
1452 | } | |
1453 | ||
b823825e | 1454 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1455 | return bio; |
1456 | } | |
a112a71d | 1457 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1458 | |
4246a0b6 | 1459 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1460 | { |
1dfa0f68 CH |
1461 | bio_free_pages(bio); |
1462 | bio_put(bio); | |
1463 | } | |
1464 | ||
4246a0b6 | 1465 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1466 | { |
42d2683a | 1467 | char *p = bio->bi_private; |
1dfa0f68 | 1468 | struct bio_vec *bvec; |
68154e90 FT |
1469 | int i; |
1470 | ||
d74c6d51 | 1471 | bio_for_each_segment_all(bvec, bio, i) { |
1dfa0f68 | 1472 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1473 | p += bvec->bv_len; |
68154e90 FT |
1474 | } |
1475 | ||
4246a0b6 | 1476 | bio_copy_kern_endio(bio); |
68154e90 FT |
1477 | } |
1478 | ||
1479 | /** | |
1480 | * bio_copy_kern - copy kernel address into bio | |
1481 | * @q: the struct request_queue for the bio | |
1482 | * @data: pointer to buffer to copy | |
1483 | * @len: length in bytes | |
1484 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1485 | * @reading: data direction is READ |
68154e90 FT |
1486 | * |
1487 | * copy the kernel address into a bio suitable for io to a block | |
1488 | * device. Returns an error pointer in case of error. | |
1489 | */ | |
1490 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1491 | gfp_t gfp_mask, int reading) | |
1492 | { | |
42d2683a CH |
1493 | unsigned long kaddr = (unsigned long)data; |
1494 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1495 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1496 | struct bio *bio; |
1497 | void *p = data; | |
1dfa0f68 | 1498 | int nr_pages = 0; |
68154e90 | 1499 | |
42d2683a CH |
1500 | /* |
1501 | * Overflow, abort | |
1502 | */ | |
1503 | if (end < start) | |
1504 | return ERR_PTR(-EINVAL); | |
68154e90 | 1505 | |
42d2683a CH |
1506 | nr_pages = end - start; |
1507 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1508 | if (!bio) | |
1509 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1510 | |
42d2683a CH |
1511 | while (len) { |
1512 | struct page *page; | |
1513 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1514 | |
42d2683a CH |
1515 | if (bytes > len) |
1516 | bytes = len; | |
1517 | ||
1518 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1519 | if (!page) | |
1520 | goto cleanup; | |
1521 | ||
1522 | if (!reading) | |
1523 | memcpy(page_address(page), p, bytes); | |
1524 | ||
1525 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1526 | break; | |
1527 | ||
1528 | len -= bytes; | |
1529 | p += bytes; | |
68154e90 FT |
1530 | } |
1531 | ||
1dfa0f68 CH |
1532 | if (reading) { |
1533 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1534 | bio->bi_private = data; | |
1535 | } else { | |
1536 | bio->bi_end_io = bio_copy_kern_endio; | |
1dfa0f68 | 1537 | } |
76029ff3 | 1538 | |
68154e90 | 1539 | return bio; |
42d2683a CH |
1540 | |
1541 | cleanup: | |
1dfa0f68 | 1542 | bio_free_pages(bio); |
42d2683a CH |
1543 | bio_put(bio); |
1544 | return ERR_PTR(-ENOMEM); | |
68154e90 FT |
1545 | } |
1546 | ||
1da177e4 LT |
1547 | /* |
1548 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1549 | * for performing direct-IO in BIOs. | |
1550 | * | |
1551 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1552 | * because the required locks are not interrupt-safe. So what we can do is to | |
1553 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1554 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1555 | * in process context. | |
1556 | * | |
1557 | * We special-case compound pages here: normally this means reads into hugetlb | |
1558 | * pages. The logic in here doesn't really work right for compound pages | |
1559 | * because the VM does not uniformly chase down the head page in all cases. | |
1560 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1561 | * handle them at all. So we skip compound pages here at an early stage. | |
1562 | * | |
1563 | * Note that this code is very hard to test under normal circumstances because | |
1564 | * direct-io pins the pages with get_user_pages(). This makes | |
1565 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1566 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1567 | * pagecache. |
1568 | * | |
1569 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1570 | * deferred bio dirtying paths. | |
1571 | */ | |
1572 | ||
1573 | /* | |
1574 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1575 | */ | |
1576 | void bio_set_pages_dirty(struct bio *bio) | |
1577 | { | |
cb34e057 | 1578 | struct bio_vec *bvec; |
1da177e4 LT |
1579 | int i; |
1580 | ||
cb34e057 KO |
1581 | bio_for_each_segment_all(bvec, bio, i) { |
1582 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1583 | |
1584 | if (page && !PageCompound(page)) | |
1585 | set_page_dirty_lock(page); | |
1586 | } | |
1587 | } | |
1588 | ||
86b6c7a7 | 1589 | static void bio_release_pages(struct bio *bio) |
1da177e4 | 1590 | { |
cb34e057 | 1591 | struct bio_vec *bvec; |
1da177e4 LT |
1592 | int i; |
1593 | ||
cb34e057 KO |
1594 | bio_for_each_segment_all(bvec, bio, i) { |
1595 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1596 | |
1597 | if (page) | |
1598 | put_page(page); | |
1599 | } | |
1600 | } | |
1601 | ||
1602 | /* | |
1603 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1604 | * If they are, then fine. If, however, some pages are clean then they must | |
1605 | * have been written out during the direct-IO read. So we take another ref on | |
1606 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1607 | * | |
1608 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1609 | * here on. It will run one put_page() against each page and will run one |
1610 | * bio_put() against the BIO. | |
1da177e4 LT |
1611 | */ |
1612 | ||
65f27f38 | 1613 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1614 | |
65f27f38 | 1615 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1616 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1617 | static struct bio *bio_dirty_list; | |
1618 | ||
1619 | /* | |
1620 | * This runs in process context | |
1621 | */ | |
65f27f38 | 1622 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1623 | { |
1624 | unsigned long flags; | |
1625 | struct bio *bio; | |
1626 | ||
1627 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1628 | bio = bio_dirty_list; | |
1629 | bio_dirty_list = NULL; | |
1630 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1631 | ||
1632 | while (bio) { | |
1633 | struct bio *next = bio->bi_private; | |
1634 | ||
1635 | bio_set_pages_dirty(bio); | |
1636 | bio_release_pages(bio); | |
1637 | bio_put(bio); | |
1638 | bio = next; | |
1639 | } | |
1640 | } | |
1641 | ||
1642 | void bio_check_pages_dirty(struct bio *bio) | |
1643 | { | |
cb34e057 | 1644 | struct bio_vec *bvec; |
1da177e4 LT |
1645 | int nr_clean_pages = 0; |
1646 | int i; | |
1647 | ||
cb34e057 KO |
1648 | bio_for_each_segment_all(bvec, bio, i) { |
1649 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1650 | |
1651 | if (PageDirty(page) || PageCompound(page)) { | |
09cbfeaf | 1652 | put_page(page); |
cb34e057 | 1653 | bvec->bv_page = NULL; |
1da177e4 LT |
1654 | } else { |
1655 | nr_clean_pages++; | |
1656 | } | |
1657 | } | |
1658 | ||
1659 | if (nr_clean_pages) { | |
1660 | unsigned long flags; | |
1661 | ||
1662 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1663 | bio->bi_private = bio_dirty_list; | |
1664 | bio_dirty_list = bio; | |
1665 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1666 | schedule_work(&bio_dirty_work); | |
1667 | } else { | |
1668 | bio_put(bio); | |
1669 | } | |
1670 | } | |
1671 | ||
d62e26b3 JA |
1672 | void generic_start_io_acct(struct request_queue *q, int rw, |
1673 | unsigned long sectors, struct hd_struct *part) | |
394ffa50 GZ |
1674 | { |
1675 | int cpu = part_stat_lock(); | |
1676 | ||
d62e26b3 | 1677 | part_round_stats(q, cpu, part); |
394ffa50 GZ |
1678 | part_stat_inc(cpu, part, ios[rw]); |
1679 | part_stat_add(cpu, part, sectors[rw], sectors); | |
d62e26b3 | 1680 | part_inc_in_flight(q, part, rw); |
394ffa50 GZ |
1681 | |
1682 | part_stat_unlock(); | |
1683 | } | |
1684 | EXPORT_SYMBOL(generic_start_io_acct); | |
1685 | ||
d62e26b3 JA |
1686 | void generic_end_io_acct(struct request_queue *q, int rw, |
1687 | struct hd_struct *part, unsigned long start_time) | |
394ffa50 GZ |
1688 | { |
1689 | unsigned long duration = jiffies - start_time; | |
1690 | int cpu = part_stat_lock(); | |
1691 | ||
1692 | part_stat_add(cpu, part, ticks[rw], duration); | |
d62e26b3 JA |
1693 | part_round_stats(q, cpu, part); |
1694 | part_dec_in_flight(q, part, rw); | |
394ffa50 GZ |
1695 | |
1696 | part_stat_unlock(); | |
1697 | } | |
1698 | EXPORT_SYMBOL(generic_end_io_acct); | |
1699 | ||
2d4dc890 IL |
1700 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1701 | void bio_flush_dcache_pages(struct bio *bi) | |
1702 | { | |
7988613b KO |
1703 | struct bio_vec bvec; |
1704 | struct bvec_iter iter; | |
2d4dc890 | 1705 | |
7988613b KO |
1706 | bio_for_each_segment(bvec, bi, iter) |
1707 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1708 | } |
1709 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1710 | #endif | |
1711 | ||
c4cf5261 JA |
1712 | static inline bool bio_remaining_done(struct bio *bio) |
1713 | { | |
1714 | /* | |
1715 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1716 | * we always end io on the first invocation. | |
1717 | */ | |
1718 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1719 | return true; | |
1720 | ||
1721 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1722 | ||
326e1dbb | 1723 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1724 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1725 | return true; |
326e1dbb | 1726 | } |
c4cf5261 JA |
1727 | |
1728 | return false; | |
1729 | } | |
1730 | ||
1da177e4 LT |
1731 | /** |
1732 | * bio_endio - end I/O on a bio | |
1733 | * @bio: bio | |
1da177e4 LT |
1734 | * |
1735 | * Description: | |
4246a0b6 CH |
1736 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1737 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1738 | * bio unless they own it and thus know that it has an end_io function. | |
fbbaf700 N |
1739 | * |
1740 | * bio_endio() can be called several times on a bio that has been chained | |
1741 | * using bio_chain(). The ->bi_end_io() function will only be called the | |
1742 | * last time. At this point the BLK_TA_COMPLETE tracing event will be | |
1743 | * generated if BIO_TRACE_COMPLETION is set. | |
1da177e4 | 1744 | **/ |
4246a0b6 | 1745 | void bio_endio(struct bio *bio) |
1da177e4 | 1746 | { |
ba8c6967 | 1747 | again: |
2b885517 | 1748 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1749 | return; |
7c20f116 CH |
1750 | if (!bio_integrity_endio(bio)) |
1751 | return; | |
1da177e4 | 1752 | |
ba8c6967 CH |
1753 | /* |
1754 | * Need to have a real endio function for chained bios, otherwise | |
1755 | * various corner cases will break (like stacking block devices that | |
1756 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1757 | * recursion and blowing the stack. Tail call optimization would | |
1758 | * handle this, but compiling with frame pointers also disables | |
1759 | * gcc's sibling call optimization. | |
1760 | */ | |
1761 | if (bio->bi_end_io == bio_chain_endio) { | |
1762 | bio = __bio_chain_endio(bio); | |
1763 | goto again; | |
196d38bc | 1764 | } |
ba8c6967 | 1765 | |
74d46992 CH |
1766 | if (bio->bi_disk && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
1767 | trace_block_bio_complete(bio->bi_disk->queue, bio, | |
a462b950 | 1768 | blk_status_to_errno(bio->bi_status)); |
fbbaf700 N |
1769 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
1770 | } | |
1771 | ||
9e234eea | 1772 | blk_throtl_bio_endio(bio); |
b222dd2f SL |
1773 | /* release cgroup info */ |
1774 | bio_uninit(bio); | |
ba8c6967 CH |
1775 | if (bio->bi_end_io) |
1776 | bio->bi_end_io(bio); | |
1da177e4 | 1777 | } |
a112a71d | 1778 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1779 | |
20d0189b KO |
1780 | /** |
1781 | * bio_split - split a bio | |
1782 | * @bio: bio to split | |
1783 | * @sectors: number of sectors to split from the front of @bio | |
1784 | * @gfp: gfp mask | |
1785 | * @bs: bio set to allocate from | |
1786 | * | |
1787 | * Allocates and returns a new bio which represents @sectors from the start of | |
1788 | * @bio, and updates @bio to represent the remaining sectors. | |
1789 | * | |
f3f5da62 MP |
1790 | * Unless this is a discard request the newly allocated bio will point |
1791 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1792 | * @bio is not freed before the split. | |
20d0189b KO |
1793 | */ |
1794 | struct bio *bio_split(struct bio *bio, int sectors, | |
1795 | gfp_t gfp, struct bio_set *bs) | |
1796 | { | |
f341a4d3 | 1797 | struct bio *split; |
20d0189b KO |
1798 | |
1799 | BUG_ON(sectors <= 0); | |
1800 | BUG_ON(sectors >= bio_sectors(bio)); | |
1801 | ||
f9d03f96 | 1802 | split = bio_clone_fast(bio, gfp, bs); |
20d0189b KO |
1803 | if (!split) |
1804 | return NULL; | |
1805 | ||
1806 | split->bi_iter.bi_size = sectors << 9; | |
1807 | ||
1808 | if (bio_integrity(split)) | |
fbd08e76 | 1809 | bio_integrity_trim(split); |
20d0189b KO |
1810 | |
1811 | bio_advance(bio, split->bi_iter.bi_size); | |
1812 | ||
fbbaf700 | 1813 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
20d59023 | 1814 | bio_set_flag(split, BIO_TRACE_COMPLETION); |
fbbaf700 | 1815 | |
20d0189b KO |
1816 | return split; |
1817 | } | |
1818 | EXPORT_SYMBOL(bio_split); | |
1819 | ||
6678d83f KO |
1820 | /** |
1821 | * bio_trim - trim a bio | |
1822 | * @bio: bio to trim | |
1823 | * @offset: number of sectors to trim from the front of @bio | |
1824 | * @size: size we want to trim @bio to, in sectors | |
1825 | */ | |
1826 | void bio_trim(struct bio *bio, int offset, int size) | |
1827 | { | |
1828 | /* 'bio' is a cloned bio which we need to trim to match | |
1829 | * the given offset and size. | |
6678d83f | 1830 | */ |
6678d83f KO |
1831 | |
1832 | size <<= 9; | |
4f024f37 | 1833 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1834 | return; |
1835 | ||
b7c44ed9 | 1836 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1837 | |
1838 | bio_advance(bio, offset << 9); | |
1839 | ||
4f024f37 | 1840 | bio->bi_iter.bi_size = size; |
376a78ab DM |
1841 | |
1842 | if (bio_integrity(bio)) | |
fbd08e76 | 1843 | bio_integrity_trim(bio); |
376a78ab | 1844 | |
6678d83f KO |
1845 | } |
1846 | EXPORT_SYMBOL_GPL(bio_trim); | |
1847 | ||
1da177e4 LT |
1848 | /* |
1849 | * create memory pools for biovec's in a bio_set. | |
1850 | * use the global biovec slabs created for general use. | |
1851 | */ | |
8aa6ba2f | 1852 | int biovec_init_pool(mempool_t *pool, int pool_entries) |
1da177e4 | 1853 | { |
ed996a52 | 1854 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
1da177e4 | 1855 | |
8aa6ba2f | 1856 | return mempool_init_slab_pool(pool, pool_entries, bp->slab); |
1da177e4 LT |
1857 | } |
1858 | ||
1859 | void bioset_free(struct bio_set *bs) | |
1860 | { | |
df2cb6da KO |
1861 | if (bs->rescue_workqueue) |
1862 | destroy_workqueue(bs->rescue_workqueue); | |
1863 | ||
8aa6ba2f KO |
1864 | mempool_exit(&bs->bio_pool); |
1865 | mempool_exit(&bs->bvec_pool); | |
9f060e22 | 1866 | |
7878cba9 | 1867 | bioset_integrity_free(bs); |
bb799ca0 | 1868 | bio_put_slab(bs); |
1da177e4 LT |
1869 | |
1870 | kfree(bs); | |
1871 | } | |
a112a71d | 1872 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1873 | |
011067b0 N |
1874 | /** |
1875 | * bioset_create - Create a bio_set | |
1876 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1877 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
47e0fb46 N |
1878 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS |
1879 | * and %BIOSET_NEED_RESCUER | |
011067b0 N |
1880 | * |
1881 | * Description: | |
1882 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1883 | * to ask for a number of bytes to be allocated in front of the bio. | |
1884 | * Front pad allocation is useful for embedding the bio inside | |
1885 | * another structure, to avoid allocating extra data to go with the bio. | |
1886 | * Note that the bio must be embedded at the END of that structure always, | |
1887 | * or things will break badly. | |
1888 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated | |
1889 | * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast(). | |
47e0fb46 N |
1890 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to |
1891 | * dispatch queued requests when the mempool runs out of space. | |
011067b0 N |
1892 | * |
1893 | */ | |
1894 | struct bio_set *bioset_create(unsigned int pool_size, | |
1895 | unsigned int front_pad, | |
1896 | int flags) | |
1da177e4 | 1897 | { |
392ddc32 | 1898 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1899 | struct bio_set *bs; |
1da177e4 | 1900 | |
1b434498 | 1901 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1902 | if (!bs) |
1903 | return NULL; | |
1904 | ||
bb799ca0 | 1905 | bs->front_pad = front_pad; |
1b434498 | 1906 | |
df2cb6da KO |
1907 | spin_lock_init(&bs->rescue_lock); |
1908 | bio_list_init(&bs->rescue_list); | |
1909 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
1910 | ||
392ddc32 | 1911 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1912 | if (!bs->bio_slab) { |
1913 | kfree(bs); | |
1914 | return NULL; | |
1915 | } | |
1916 | ||
8aa6ba2f | 1917 | if (mempool_init_slab_pool(&bs->bio_pool, pool_size, bs->bio_slab)) |
1da177e4 LT |
1918 | goto bad; |
1919 | ||
8aa6ba2f KO |
1920 | if ((flags & BIOSET_NEED_BVECS) && |
1921 | biovec_init_pool(&bs->bvec_pool, pool_size)) | |
1922 | goto bad; | |
df2cb6da | 1923 | |
47e0fb46 N |
1924 | if (!(flags & BIOSET_NEED_RESCUER)) |
1925 | return bs; | |
1926 | ||
df2cb6da KO |
1927 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); |
1928 | if (!bs->rescue_workqueue) | |
1929 | goto bad; | |
1da177e4 | 1930 | |
df2cb6da | 1931 | return bs; |
1da177e4 LT |
1932 | bad: |
1933 | bioset_free(bs); | |
1934 | return NULL; | |
1935 | } | |
a112a71d | 1936 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 1937 | |
852c788f | 1938 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 TH |
1939 | |
1940 | /** | |
1941 | * bio_associate_blkcg - associate a bio with the specified blkcg | |
1942 | * @bio: target bio | |
1943 | * @blkcg_css: css of the blkcg to associate | |
1944 | * | |
1945 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will | |
1946 | * treat @bio as if it were issued by a task which belongs to the blkcg. | |
1947 | * | |
1948 | * This function takes an extra reference of @blkcg_css which will be put | |
1949 | * when @bio is released. The caller must own @bio and is responsible for | |
1950 | * synchronizing calls to this function. | |
1951 | */ | |
1952 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) | |
1953 | { | |
1954 | if (unlikely(bio->bi_css)) | |
1955 | return -EBUSY; | |
1956 | css_get(blkcg_css); | |
1957 | bio->bi_css = blkcg_css; | |
1958 | return 0; | |
1959 | } | |
5aa2a96b | 1960 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
1d933cf0 | 1961 | |
852c788f TH |
1962 | /** |
1963 | * bio_disassociate_task - undo bio_associate_current() | |
1964 | * @bio: target bio | |
1965 | */ | |
1966 | void bio_disassociate_task(struct bio *bio) | |
1967 | { | |
1968 | if (bio->bi_ioc) { | |
1969 | put_io_context(bio->bi_ioc); | |
1970 | bio->bi_ioc = NULL; | |
1971 | } | |
1972 | if (bio->bi_css) { | |
1973 | css_put(bio->bi_css); | |
1974 | bio->bi_css = NULL; | |
1975 | } | |
1976 | } | |
1977 | ||
20bd723e PV |
1978 | /** |
1979 | * bio_clone_blkcg_association - clone blkcg association from src to dst bio | |
1980 | * @dst: destination bio | |
1981 | * @src: source bio | |
1982 | */ | |
1983 | void bio_clone_blkcg_association(struct bio *dst, struct bio *src) | |
1984 | { | |
1985 | if (src->bi_css) | |
1986 | WARN_ON(bio_associate_blkcg(dst, src->bi_css)); | |
1987 | } | |
8a8e6f84 | 1988 | EXPORT_SYMBOL_GPL(bio_clone_blkcg_association); |
852c788f TH |
1989 | #endif /* CONFIG_BLK_CGROUP */ |
1990 | ||
1da177e4 LT |
1991 | static void __init biovec_init_slabs(void) |
1992 | { | |
1993 | int i; | |
1994 | ||
ed996a52 | 1995 | for (i = 0; i < BVEC_POOL_NR; i++) { |
1da177e4 LT |
1996 | int size; |
1997 | struct biovec_slab *bvs = bvec_slabs + i; | |
1998 | ||
a7fcd37c JA |
1999 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2000 | bvs->slab = NULL; | |
2001 | continue; | |
2002 | } | |
a7fcd37c | 2003 | |
1da177e4 LT |
2004 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2005 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2006 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2007 | } |
2008 | } | |
2009 | ||
2010 | static int __init init_bio(void) | |
2011 | { | |
bb799ca0 JA |
2012 | bio_slab_max = 2; |
2013 | bio_slab_nr = 0; | |
2014 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
2015 | if (!bio_slabs) | |
2016 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2017 | |
7878cba9 | 2018 | bio_integrity_init(); |
1da177e4 LT |
2019 | biovec_init_slabs(); |
2020 | ||
011067b0 | 2021 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS); |
1da177e4 LT |
2022 | if (!fs_bio_set) |
2023 | panic("bio: can't allocate bios\n"); | |
2024 | ||
a91a2785 MP |
2025 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
2026 | panic("bio: can't create integrity pool\n"); | |
2027 | ||
1da177e4 LT |
2028 | return 0; |
2029 | } | |
1da177e4 | 2030 | subsys_initcall(init_bio); |