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> |
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 | ||
1da177e4 LT |
40 | /* |
41 | * if you change this list, also change bvec_alloc or things will | |
42 | * break badly! cannot be bigger than what you can fit into an | |
43 | * unsigned short | |
44 | */ | |
1da177e4 | 45 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
ed996a52 | 46 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
1da177e4 LT |
47 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
48 | }; | |
49 | #undef BV | |
50 | ||
1da177e4 LT |
51 | /* |
52 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
53 | * IO code that does not need private memory pools. | |
54 | */ | |
51d654e1 | 55 | struct bio_set *fs_bio_set; |
3f86a82a | 56 | EXPORT_SYMBOL(fs_bio_set); |
1da177e4 | 57 | |
bb799ca0 JA |
58 | /* |
59 | * Our slab pool management | |
60 | */ | |
61 | struct bio_slab { | |
62 | struct kmem_cache *slab; | |
63 | unsigned int slab_ref; | |
64 | unsigned int slab_size; | |
65 | char name[8]; | |
66 | }; | |
67 | static DEFINE_MUTEX(bio_slab_lock); | |
68 | static struct bio_slab *bio_slabs; | |
69 | static unsigned int bio_slab_nr, bio_slab_max; | |
70 | ||
71 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
72 | { | |
73 | unsigned int sz = sizeof(struct bio) + extra_size; | |
74 | struct kmem_cache *slab = NULL; | |
389d7b26 | 75 | struct bio_slab *bslab, *new_bio_slabs; |
386bc35a | 76 | unsigned int new_bio_slab_max; |
bb799ca0 JA |
77 | unsigned int i, entry = -1; |
78 | ||
79 | mutex_lock(&bio_slab_lock); | |
80 | ||
81 | i = 0; | |
82 | while (i < bio_slab_nr) { | |
f06f135d | 83 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
84 | |
85 | if (!bslab->slab && entry == -1) | |
86 | entry = i; | |
87 | else if (bslab->slab_size == sz) { | |
88 | slab = bslab->slab; | |
89 | bslab->slab_ref++; | |
90 | break; | |
91 | } | |
92 | i++; | |
93 | } | |
94 | ||
95 | if (slab) | |
96 | goto out_unlock; | |
97 | ||
98 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
386bc35a | 99 | new_bio_slab_max = bio_slab_max << 1; |
389d7b26 | 100 | new_bio_slabs = krealloc(bio_slabs, |
386bc35a | 101 | new_bio_slab_max * sizeof(struct bio_slab), |
389d7b26 AK |
102 | GFP_KERNEL); |
103 | if (!new_bio_slabs) | |
bb799ca0 | 104 | goto out_unlock; |
386bc35a | 105 | bio_slab_max = new_bio_slab_max; |
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); | |
6a241483 MP |
114 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
115 | SLAB_HWCACHE_ALIGN, NULL); | |
bb799ca0 JA |
116 | if (!slab) |
117 | goto out_unlock; | |
118 | ||
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 | ||
9f060e22 | 161 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
bb799ca0 | 162 | { |
ed996a52 CH |
163 | if (!idx) |
164 | return; | |
165 | idx--; | |
166 | ||
167 | BIO_BUG_ON(idx >= BVEC_POOL_NR); | |
bb799ca0 | 168 | |
ed996a52 | 169 | if (idx == BVEC_POOL_MAX) { |
9f060e22 | 170 | mempool_free(bv, pool); |
ed996a52 | 171 | } else { |
bb799ca0 JA |
172 | struct biovec_slab *bvs = bvec_slabs + idx; |
173 | ||
174 | kmem_cache_free(bvs->slab, bv); | |
175 | } | |
176 | } | |
177 | ||
9f060e22 KO |
178 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
179 | mempool_t *pool) | |
1da177e4 LT |
180 | { |
181 | struct bio_vec *bvl; | |
1da177e4 | 182 | |
7ff9345f JA |
183 | /* |
184 | * see comment near bvec_array define! | |
185 | */ | |
186 | switch (nr) { | |
187 | case 1: | |
188 | *idx = 0; | |
189 | break; | |
190 | case 2 ... 4: | |
191 | *idx = 1; | |
192 | break; | |
193 | case 5 ... 16: | |
194 | *idx = 2; | |
195 | break; | |
196 | case 17 ... 64: | |
197 | *idx = 3; | |
198 | break; | |
199 | case 65 ... 128: | |
200 | *idx = 4; | |
201 | break; | |
202 | case 129 ... BIO_MAX_PAGES: | |
203 | *idx = 5; | |
204 | break; | |
205 | default: | |
206 | return NULL; | |
207 | } | |
208 | ||
209 | /* | |
210 | * idx now points to the pool we want to allocate from. only the | |
211 | * 1-vec entry pool is mempool backed. | |
212 | */ | |
ed996a52 | 213 | if (*idx == BVEC_POOL_MAX) { |
7ff9345f | 214 | fallback: |
9f060e22 | 215 | bvl = mempool_alloc(pool, gfp_mask); |
7ff9345f JA |
216 | } else { |
217 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
d0164adc | 218 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
7ff9345f | 219 | |
0a0d96b0 | 220 | /* |
7ff9345f JA |
221 | * Make this allocation restricted and don't dump info on |
222 | * allocation failures, since we'll fallback to the mempool | |
223 | * in case of failure. | |
0a0d96b0 | 224 | */ |
7ff9345f | 225 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 226 | |
0a0d96b0 | 227 | /* |
d0164adc | 228 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
7ff9345f | 229 | * is set, retry with the 1-entry mempool |
0a0d96b0 | 230 | */ |
7ff9345f | 231 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
d0164adc | 232 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
ed996a52 | 233 | *idx = BVEC_POOL_MAX; |
7ff9345f JA |
234 | goto fallback; |
235 | } | |
236 | } | |
237 | ||
ed996a52 | 238 | (*idx)++; |
1da177e4 LT |
239 | return bvl; |
240 | } | |
241 | ||
4254bba1 | 242 | static void __bio_free(struct bio *bio) |
1da177e4 | 243 | { |
4254bba1 | 244 | bio_disassociate_task(bio); |
1da177e4 | 245 | |
7ba1ba12 | 246 | if (bio_integrity(bio)) |
1e2a410f | 247 | bio_integrity_free(bio); |
4254bba1 | 248 | } |
7ba1ba12 | 249 | |
4254bba1 KO |
250 | static void bio_free(struct bio *bio) |
251 | { | |
252 | struct bio_set *bs = bio->bi_pool; | |
253 | void *p; | |
254 | ||
255 | __bio_free(bio); | |
256 | ||
257 | if (bs) { | |
ed996a52 | 258 | bvec_free(bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio)); |
4254bba1 KO |
259 | |
260 | /* | |
261 | * If we have front padding, adjust the bio pointer before freeing | |
262 | */ | |
263 | p = bio; | |
bb799ca0 JA |
264 | p -= bs->front_pad; |
265 | ||
4254bba1 KO |
266 | mempool_free(p, bs->bio_pool); |
267 | } else { | |
268 | /* Bio was allocated by bio_kmalloc() */ | |
269 | kfree(bio); | |
270 | } | |
3676347a PO |
271 | } |
272 | ||
3a83f467 ML |
273 | void bio_init(struct bio *bio, struct bio_vec *table, |
274 | unsigned short max_vecs) | |
1da177e4 | 275 | { |
2b94de55 | 276 | memset(bio, 0, sizeof(*bio)); |
c4cf5261 | 277 | atomic_set(&bio->__bi_remaining, 1); |
dac56212 | 278 | atomic_set(&bio->__bi_cnt, 1); |
3a83f467 ML |
279 | |
280 | bio->bi_io_vec = table; | |
281 | bio->bi_max_vecs = max_vecs; | |
1da177e4 | 282 | } |
a112a71d | 283 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 284 | |
f44b48c7 KO |
285 | /** |
286 | * bio_reset - reinitialize a bio | |
287 | * @bio: bio to reset | |
288 | * | |
289 | * Description: | |
290 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
291 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
292 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
293 | * comment in struct bio. | |
294 | */ | |
295 | void bio_reset(struct bio *bio) | |
296 | { | |
297 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
298 | ||
4254bba1 | 299 | __bio_free(bio); |
f44b48c7 KO |
300 | |
301 | memset(bio, 0, BIO_RESET_BYTES); | |
4246a0b6 | 302 | bio->bi_flags = flags; |
c4cf5261 | 303 | atomic_set(&bio->__bi_remaining, 1); |
f44b48c7 KO |
304 | } |
305 | EXPORT_SYMBOL(bio_reset); | |
306 | ||
38f8baae | 307 | static struct bio *__bio_chain_endio(struct bio *bio) |
196d38bc | 308 | { |
4246a0b6 CH |
309 | struct bio *parent = bio->bi_private; |
310 | ||
af3e3a52 CH |
311 | if (!parent->bi_error) |
312 | parent->bi_error = bio->bi_error; | |
196d38bc | 313 | bio_put(bio); |
38f8baae CH |
314 | return parent; |
315 | } | |
316 | ||
317 | static void bio_chain_endio(struct bio *bio) | |
318 | { | |
319 | bio_endio(__bio_chain_endio(bio)); | |
196d38bc KO |
320 | } |
321 | ||
322 | /** | |
323 | * bio_chain - chain bio completions | |
1051a902 RD |
324 | * @bio: the target bio |
325 | * @parent: the @bio's parent bio | |
196d38bc KO |
326 | * |
327 | * The caller won't have a bi_end_io called when @bio completes - instead, | |
328 | * @parent's bi_end_io won't be called until both @parent and @bio have | |
329 | * completed; the chained bio will also be freed when it completes. | |
330 | * | |
331 | * The caller must not set bi_private or bi_end_io in @bio. | |
332 | */ | |
333 | void bio_chain(struct bio *bio, struct bio *parent) | |
334 | { | |
335 | BUG_ON(bio->bi_private || bio->bi_end_io); | |
336 | ||
337 | bio->bi_private = parent; | |
338 | bio->bi_end_io = bio_chain_endio; | |
c4cf5261 | 339 | bio_inc_remaining(parent); |
196d38bc KO |
340 | } |
341 | EXPORT_SYMBOL(bio_chain); | |
342 | ||
df2cb6da KO |
343 | static void bio_alloc_rescue(struct work_struct *work) |
344 | { | |
345 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); | |
346 | struct bio *bio; | |
347 | ||
348 | while (1) { | |
349 | spin_lock(&bs->rescue_lock); | |
350 | bio = bio_list_pop(&bs->rescue_list); | |
351 | spin_unlock(&bs->rescue_lock); | |
352 | ||
353 | if (!bio) | |
354 | break; | |
355 | ||
356 | generic_make_request(bio); | |
357 | } | |
358 | } | |
359 | ||
360 | static void punt_bios_to_rescuer(struct bio_set *bs) | |
361 | { | |
362 | struct bio_list punt, nopunt; | |
363 | struct bio *bio; | |
364 | ||
365 | /* | |
366 | * In order to guarantee forward progress we must punt only bios that | |
367 | * were allocated from this bio_set; otherwise, if there was a bio on | |
368 | * there for a stacking driver higher up in the stack, processing it | |
369 | * could require allocating bios from this bio_set, and doing that from | |
370 | * our own rescuer would be bad. | |
371 | * | |
372 | * Since bio lists are singly linked, pop them all instead of trying to | |
373 | * remove from the middle of the list: | |
374 | */ | |
375 | ||
376 | bio_list_init(&punt); | |
377 | bio_list_init(&nopunt); | |
378 | ||
f5fe1b51 | 379 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
df2cb6da | 380 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
f5fe1b51 | 381 | current->bio_list[0] = nopunt; |
df2cb6da | 382 | |
f5fe1b51 N |
383 | bio_list_init(&nopunt); |
384 | while ((bio = bio_list_pop(¤t->bio_list[1]))) | |
385 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); | |
386 | current->bio_list[1] = nopunt; | |
df2cb6da KO |
387 | |
388 | spin_lock(&bs->rescue_lock); | |
389 | bio_list_merge(&bs->rescue_list, &punt); | |
390 | spin_unlock(&bs->rescue_lock); | |
391 | ||
392 | queue_work(bs->rescue_workqueue, &bs->rescue_work); | |
393 | } | |
394 | ||
1da177e4 LT |
395 | /** |
396 | * bio_alloc_bioset - allocate a bio for I/O | |
397 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
398 | * @nr_iovecs: number of iovecs to pre-allocate | |
db18efac | 399 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
400 | * |
401 | * Description: | |
3f86a82a KO |
402 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
403 | * backed by the @bs's mempool. | |
404 | * | |
d0164adc MG |
405 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
406 | * always be able to allocate a bio. This is due to the mempool guarantees. | |
407 | * To make this work, callers must never allocate more than 1 bio at a time | |
408 | * from this pool. Callers that need to allocate more than 1 bio must always | |
409 | * submit the previously allocated bio for IO before attempting to allocate | |
410 | * a new one. Failure to do so can cause deadlocks under memory pressure. | |
3f86a82a | 411 | * |
df2cb6da KO |
412 | * Note that when running under generic_make_request() (i.e. any block |
413 | * driver), bios are not submitted until after you return - see the code in | |
414 | * generic_make_request() that converts recursion into iteration, to prevent | |
415 | * stack overflows. | |
416 | * | |
417 | * This would normally mean allocating multiple bios under | |
418 | * generic_make_request() would be susceptible to deadlocks, but we have | |
419 | * deadlock avoidance code that resubmits any blocked bios from a rescuer | |
420 | * thread. | |
421 | * | |
422 | * However, we do not guarantee forward progress for allocations from other | |
423 | * mempools. Doing multiple allocations from the same mempool under | |
424 | * generic_make_request() should be avoided - instead, use bio_set's front_pad | |
425 | * for per bio allocations. | |
426 | * | |
3f86a82a KO |
427 | * RETURNS: |
428 | * Pointer to new bio on success, NULL on failure. | |
429 | */ | |
dd0fc66f | 430 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) |
1da177e4 | 431 | { |
df2cb6da | 432 | gfp_t saved_gfp = gfp_mask; |
3f86a82a KO |
433 | unsigned front_pad; |
434 | unsigned inline_vecs; | |
34053979 | 435 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
436 | struct bio *bio; |
437 | void *p; | |
438 | ||
3f86a82a KO |
439 | if (!bs) { |
440 | if (nr_iovecs > UIO_MAXIOV) | |
441 | return NULL; | |
442 | ||
443 | p = kmalloc(sizeof(struct bio) + | |
444 | nr_iovecs * sizeof(struct bio_vec), | |
445 | gfp_mask); | |
446 | front_pad = 0; | |
447 | inline_vecs = nr_iovecs; | |
448 | } else { | |
d8f429e1 JN |
449 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
450 | if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0)) | |
451 | return NULL; | |
df2cb6da KO |
452 | /* |
453 | * generic_make_request() converts recursion to iteration; this | |
454 | * means if we're running beneath it, any bios we allocate and | |
455 | * submit will not be submitted (and thus freed) until after we | |
456 | * return. | |
457 | * | |
458 | * This exposes us to a potential deadlock if we allocate | |
459 | * multiple bios from the same bio_set() while running | |
460 | * underneath generic_make_request(). If we were to allocate | |
461 | * multiple bios (say a stacking block driver that was splitting | |
462 | * bios), we would deadlock if we exhausted the mempool's | |
463 | * reserve. | |
464 | * | |
465 | * We solve this, and guarantee forward progress, with a rescuer | |
466 | * workqueue per bio_set. If we go to allocate and there are | |
467 | * bios on current->bio_list, we first try the allocation | |
d0164adc MG |
468 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
469 | * bios we would be blocking to the rescuer workqueue before | |
470 | * we retry with the original gfp_flags. | |
df2cb6da KO |
471 | */ |
472 | ||
f5fe1b51 N |
473 | if (current->bio_list && |
474 | (!bio_list_empty(¤t->bio_list[0]) || | |
475 | !bio_list_empty(¤t->bio_list[1]))) | |
d0164adc | 476 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
df2cb6da | 477 | |
3f86a82a | 478 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
df2cb6da KO |
479 | if (!p && gfp_mask != saved_gfp) { |
480 | punt_bios_to_rescuer(bs); | |
481 | gfp_mask = saved_gfp; | |
482 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
483 | } | |
484 | ||
3f86a82a KO |
485 | front_pad = bs->front_pad; |
486 | inline_vecs = BIO_INLINE_VECS; | |
487 | } | |
488 | ||
451a9ebf TH |
489 | if (unlikely(!p)) |
490 | return NULL; | |
1da177e4 | 491 | |
3f86a82a | 492 | bio = p + front_pad; |
3a83f467 | 493 | bio_init(bio, NULL, 0); |
34053979 | 494 | |
3f86a82a | 495 | if (nr_iovecs > inline_vecs) { |
ed996a52 CH |
496 | unsigned long idx = 0; |
497 | ||
9f060e22 | 498 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
499 | if (!bvl && gfp_mask != saved_gfp) { |
500 | punt_bios_to_rescuer(bs); | |
501 | gfp_mask = saved_gfp; | |
9f060e22 | 502 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
503 | } |
504 | ||
34053979 IM |
505 | if (unlikely(!bvl)) |
506 | goto err_free; | |
a38352e0 | 507 | |
ed996a52 | 508 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
3f86a82a KO |
509 | } else if (nr_iovecs) { |
510 | bvl = bio->bi_inline_vecs; | |
1da177e4 | 511 | } |
3f86a82a KO |
512 | |
513 | bio->bi_pool = bs; | |
34053979 | 514 | bio->bi_max_vecs = nr_iovecs; |
34053979 | 515 | bio->bi_io_vec = bvl; |
1da177e4 | 516 | return bio; |
34053979 IM |
517 | |
518 | err_free: | |
451a9ebf | 519 | mempool_free(p, bs->bio_pool); |
34053979 | 520 | return NULL; |
1da177e4 | 521 | } |
a112a71d | 522 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 523 | |
1da177e4 LT |
524 | void zero_fill_bio(struct bio *bio) |
525 | { | |
526 | unsigned long flags; | |
7988613b KO |
527 | struct bio_vec bv; |
528 | struct bvec_iter iter; | |
1da177e4 | 529 | |
7988613b KO |
530 | bio_for_each_segment(bv, bio, iter) { |
531 | char *data = bvec_kmap_irq(&bv, &flags); | |
532 | memset(data, 0, bv.bv_len); | |
533 | flush_dcache_page(bv.bv_page); | |
1da177e4 LT |
534 | bvec_kunmap_irq(data, &flags); |
535 | } | |
536 | } | |
537 | EXPORT_SYMBOL(zero_fill_bio); | |
538 | ||
539 | /** | |
540 | * bio_put - release a reference to a bio | |
541 | * @bio: bio to release reference to | |
542 | * | |
543 | * Description: | |
544 | * Put a reference to a &struct bio, either one you have gotten with | |
ad0bf110 | 545 | * bio_alloc, bio_get or bio_clone. The last put of a bio will free it. |
1da177e4 LT |
546 | **/ |
547 | void bio_put(struct bio *bio) | |
548 | { | |
dac56212 | 549 | if (!bio_flagged(bio, BIO_REFFED)) |
4254bba1 | 550 | bio_free(bio); |
dac56212 JA |
551 | else { |
552 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); | |
553 | ||
554 | /* | |
555 | * last put frees it | |
556 | */ | |
557 | if (atomic_dec_and_test(&bio->__bi_cnt)) | |
558 | bio_free(bio); | |
559 | } | |
1da177e4 | 560 | } |
a112a71d | 561 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 562 | |
165125e1 | 563 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
564 | { |
565 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
566 | blk_recount_segments(q, bio); | |
567 | ||
568 | return bio->bi_phys_segments; | |
569 | } | |
a112a71d | 570 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 571 | |
59d276fe KO |
572 | /** |
573 | * __bio_clone_fast - clone a bio that shares the original bio's biovec | |
574 | * @bio: destination bio | |
575 | * @bio_src: bio to clone | |
576 | * | |
577 | * Clone a &bio. Caller will own the returned bio, but not | |
578 | * the actual data it points to. Reference count of returned | |
579 | * bio will be one. | |
580 | * | |
581 | * Caller must ensure that @bio_src is not freed before @bio. | |
582 | */ | |
583 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) | |
584 | { | |
ed996a52 | 585 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
59d276fe KO |
586 | |
587 | /* | |
588 | * most users will be overriding ->bi_bdev with a new target, | |
589 | * so we don't set nor calculate new physical/hw segment counts here | |
590 | */ | |
591 | bio->bi_bdev = bio_src->bi_bdev; | |
b7c44ed9 | 592 | bio_set_flag(bio, BIO_CLONED); |
1eff9d32 | 593 | bio->bi_opf = bio_src->bi_opf; |
59d276fe KO |
594 | bio->bi_iter = bio_src->bi_iter; |
595 | bio->bi_io_vec = bio_src->bi_io_vec; | |
20bd723e PV |
596 | |
597 | bio_clone_blkcg_association(bio, bio_src); | |
59d276fe KO |
598 | } |
599 | EXPORT_SYMBOL(__bio_clone_fast); | |
600 | ||
601 | /** | |
602 | * bio_clone_fast - clone a bio that shares the original bio's biovec | |
603 | * @bio: bio to clone | |
604 | * @gfp_mask: allocation priority | |
605 | * @bs: bio_set to allocate from | |
606 | * | |
607 | * Like __bio_clone_fast, only also allocates the returned bio | |
608 | */ | |
609 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) | |
610 | { | |
611 | struct bio *b; | |
612 | ||
613 | b = bio_alloc_bioset(gfp_mask, 0, bs); | |
614 | if (!b) | |
615 | return NULL; | |
616 | ||
617 | __bio_clone_fast(b, bio); | |
618 | ||
619 | if (bio_integrity(bio)) { | |
620 | int ret; | |
621 | ||
622 | ret = bio_integrity_clone(b, bio, gfp_mask); | |
623 | ||
624 | if (ret < 0) { | |
625 | bio_put(b); | |
626 | return NULL; | |
627 | } | |
628 | } | |
629 | ||
630 | return b; | |
631 | } | |
632 | EXPORT_SYMBOL(bio_clone_fast); | |
633 | ||
c18a1e09 ML |
634 | static struct bio *__bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask, |
635 | struct bio_set *bs, int offset, | |
636 | int size) | |
1da177e4 | 637 | { |
bdb53207 KO |
638 | struct bvec_iter iter; |
639 | struct bio_vec bv; | |
640 | struct bio *bio; | |
c18a1e09 ML |
641 | struct bvec_iter iter_src = bio_src->bi_iter; |
642 | ||
643 | /* for supporting partial clone */ | |
644 | if (offset || size != bio_src->bi_iter.bi_size) { | |
645 | bio_advance_iter(bio_src, &iter_src, offset); | |
646 | iter_src.bi_size = size; | |
647 | } | |
1da177e4 | 648 | |
bdb53207 KO |
649 | /* |
650 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from | |
651 | * bio_src->bi_io_vec to bio->bi_io_vec. | |
652 | * | |
653 | * We can't do that anymore, because: | |
654 | * | |
655 | * - The point of cloning the biovec is to produce a bio with a biovec | |
656 | * the caller can modify: bi_idx and bi_bvec_done should be 0. | |
657 | * | |
658 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if | |
659 | * we tried to clone the whole thing bio_alloc_bioset() would fail. | |
660 | * But the clone should succeed as long as the number of biovecs we | |
661 | * actually need to allocate is fewer than BIO_MAX_PAGES. | |
662 | * | |
663 | * - Lastly, bi_vcnt should not be looked at or relied upon by code | |
664 | * that does not own the bio - reason being drivers don't use it for | |
665 | * iterating over the biovec anymore, so expecting it to be kept up | |
666 | * to date (i.e. for clones that share the parent biovec) is just | |
667 | * asking for trouble and would force extra work on | |
668 | * __bio_clone_fast() anyways. | |
669 | */ | |
670 | ||
c18a1e09 ML |
671 | bio = bio_alloc_bioset(gfp_mask, __bio_segments(bio_src, |
672 | &iter_src), bs); | |
bdb53207 | 673 | if (!bio) |
7ba1ba12 | 674 | return NULL; |
bdb53207 | 675 | bio->bi_bdev = bio_src->bi_bdev; |
1eff9d32 | 676 | bio->bi_opf = bio_src->bi_opf; |
bdb53207 KO |
677 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; |
678 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; | |
7ba1ba12 | 679 | |
7afafc8a AH |
680 | switch (bio_op(bio)) { |
681 | case REQ_OP_DISCARD: | |
682 | case REQ_OP_SECURE_ERASE: | |
a6f0788e | 683 | case REQ_OP_WRITE_ZEROES: |
7afafc8a AH |
684 | break; |
685 | case REQ_OP_WRITE_SAME: | |
8423ae3d | 686 | bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0]; |
7afafc8a AH |
687 | break; |
688 | default: | |
c18a1e09 | 689 | __bio_for_each_segment(bv, bio_src, iter, iter_src) |
7afafc8a AH |
690 | bio->bi_io_vec[bio->bi_vcnt++] = bv; |
691 | break; | |
8423ae3d KO |
692 | } |
693 | ||
bdb53207 KO |
694 | if (bio_integrity(bio_src)) { |
695 | int ret; | |
7ba1ba12 | 696 | |
bdb53207 | 697 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
059ea331 | 698 | if (ret < 0) { |
bdb53207 | 699 | bio_put(bio); |
7ba1ba12 | 700 | return NULL; |
059ea331 | 701 | } |
3676347a | 702 | } |
1da177e4 | 703 | |
20bd723e PV |
704 | bio_clone_blkcg_association(bio, bio_src); |
705 | ||
bdb53207 | 706 | return bio; |
1da177e4 | 707 | } |
c18a1e09 ML |
708 | |
709 | /** | |
710 | * bio_clone_bioset - clone a bio | |
711 | * @bio_src: bio to clone | |
712 | * @gfp_mask: allocation priority | |
713 | * @bs: bio_set to allocate from | |
714 | * | |
715 | * Clone bio. Caller will own the returned bio, but not the actual data it | |
716 | * points to. Reference count of returned bio will be one. | |
717 | */ | |
718 | struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask, | |
719 | struct bio_set *bs) | |
720 | { | |
721 | return __bio_clone_bioset(bio_src, gfp_mask, bs, 0, | |
722 | bio_src->bi_iter.bi_size); | |
723 | } | |
bf800ef1 | 724 | EXPORT_SYMBOL(bio_clone_bioset); |
1da177e4 | 725 | |
c18a1e09 ML |
726 | /** |
727 | * bio_clone_bioset_partial - clone a partial bio | |
728 | * @bio_src: bio to clone | |
729 | * @gfp_mask: allocation priority | |
730 | * @bs: bio_set to allocate from | |
731 | * @offset: cloned starting from the offset | |
732 | * @size: size for the cloned bio | |
733 | * | |
734 | * Clone bio. Caller will own the returned bio, but not the actual data it | |
735 | * points to. Reference count of returned bio will be one. | |
736 | */ | |
737 | struct bio *bio_clone_bioset_partial(struct bio *bio_src, gfp_t gfp_mask, | |
738 | struct bio_set *bs, int offset, | |
739 | int size) | |
740 | { | |
741 | return __bio_clone_bioset(bio_src, gfp_mask, bs, offset, size); | |
742 | } | |
743 | EXPORT_SYMBOL(bio_clone_bioset_partial); | |
744 | ||
1da177e4 | 745 | /** |
c66a14d0 KO |
746 | * bio_add_pc_page - attempt to add page to bio |
747 | * @q: the target queue | |
748 | * @bio: destination bio | |
749 | * @page: page to add | |
750 | * @len: vec entry length | |
751 | * @offset: vec entry offset | |
1da177e4 | 752 | * |
c66a14d0 KO |
753 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
754 | * number of reasons, such as the bio being full or target block device | |
755 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
756 | * so it is always possible to add a single page to an empty bio. | |
757 | * | |
758 | * This should only be used by REQ_PC bios. | |
1da177e4 | 759 | */ |
c66a14d0 KO |
760 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
761 | *page, unsigned int len, unsigned int offset) | |
1da177e4 LT |
762 | { |
763 | int retried_segments = 0; | |
764 | struct bio_vec *bvec; | |
765 | ||
766 | /* | |
767 | * cloned bio must not modify vec list | |
768 | */ | |
769 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
770 | return 0; | |
771 | ||
c66a14d0 | 772 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
773 | return 0; |
774 | ||
80cfd548 JA |
775 | /* |
776 | * For filesystems with a blocksize smaller than the pagesize | |
777 | * we will often be called with the same page as last time and | |
778 | * a consecutive offset. Optimize this special case. | |
779 | */ | |
780 | if (bio->bi_vcnt > 0) { | |
781 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
782 | ||
783 | if (page == prev->bv_page && | |
784 | offset == prev->bv_offset + prev->bv_len) { | |
785 | prev->bv_len += len; | |
fcbf6a08 | 786 | bio->bi_iter.bi_size += len; |
80cfd548 JA |
787 | goto done; |
788 | } | |
66cb45aa JA |
789 | |
790 | /* | |
791 | * If the queue doesn't support SG gaps and adding this | |
792 | * offset would create a gap, disallow it. | |
793 | */ | |
03100aad | 794 | if (bvec_gap_to_prev(q, prev, offset)) |
66cb45aa | 795 | return 0; |
80cfd548 JA |
796 | } |
797 | ||
798 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
799 | return 0; |
800 | ||
801 | /* | |
fcbf6a08 ML |
802 | * setup the new entry, we might clear it again later if we |
803 | * cannot add the page | |
804 | */ | |
805 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
806 | bvec->bv_page = page; | |
807 | bvec->bv_len = len; | |
808 | bvec->bv_offset = offset; | |
809 | bio->bi_vcnt++; | |
810 | bio->bi_phys_segments++; | |
811 | bio->bi_iter.bi_size += len; | |
812 | ||
813 | /* | |
814 | * Perform a recount if the number of segments is greater | |
815 | * than queue_max_segments(q). | |
1da177e4 LT |
816 | */ |
817 | ||
fcbf6a08 | 818 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
1da177e4 LT |
819 | |
820 | if (retried_segments) | |
fcbf6a08 | 821 | goto failed; |
1da177e4 LT |
822 | |
823 | retried_segments = 1; | |
824 | blk_recount_segments(q, bio); | |
825 | } | |
826 | ||
1da177e4 | 827 | /* If we may be able to merge these biovecs, force a recount */ |
fcbf6a08 | 828 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
b7c44ed9 | 829 | bio_clear_flag(bio, BIO_SEG_VALID); |
1da177e4 | 830 | |
80cfd548 | 831 | done: |
1da177e4 | 832 | return len; |
fcbf6a08 ML |
833 | |
834 | failed: | |
835 | bvec->bv_page = NULL; | |
836 | bvec->bv_len = 0; | |
837 | bvec->bv_offset = 0; | |
838 | bio->bi_vcnt--; | |
839 | bio->bi_iter.bi_size -= len; | |
840 | blk_recount_segments(q, bio); | |
841 | return 0; | |
1da177e4 | 842 | } |
a112a71d | 843 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 844 | |
1da177e4 LT |
845 | /** |
846 | * bio_add_page - attempt to add page to bio | |
847 | * @bio: destination bio | |
848 | * @page: page to add | |
849 | * @len: vec entry length | |
850 | * @offset: vec entry offset | |
851 | * | |
c66a14d0 KO |
852 | * Attempt to add a page to the bio_vec maplist. This will only fail |
853 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. | |
1da177e4 | 854 | */ |
c66a14d0 KO |
855 | int bio_add_page(struct bio *bio, struct page *page, |
856 | unsigned int len, unsigned int offset) | |
1da177e4 | 857 | { |
c66a14d0 KO |
858 | struct bio_vec *bv; |
859 | ||
860 | /* | |
861 | * cloned bio must not modify vec list | |
862 | */ | |
863 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) | |
864 | return 0; | |
762380ad | 865 | |
c66a14d0 KO |
866 | /* |
867 | * For filesystems with a blocksize smaller than the pagesize | |
868 | * we will often be called with the same page as last time and | |
869 | * a consecutive offset. Optimize this special case. | |
870 | */ | |
871 | if (bio->bi_vcnt > 0) { | |
872 | bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
58a4915a | 873 | |
c66a14d0 KO |
874 | if (page == bv->bv_page && |
875 | offset == bv->bv_offset + bv->bv_len) { | |
876 | bv->bv_len += len; | |
877 | goto done; | |
878 | } | |
879 | } | |
880 | ||
881 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
882 | return 0; | |
883 | ||
884 | bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
885 | bv->bv_page = page; | |
886 | bv->bv_len = len; | |
887 | bv->bv_offset = offset; | |
888 | ||
889 | bio->bi_vcnt++; | |
890 | done: | |
891 | bio->bi_iter.bi_size += len; | |
892 | return len; | |
1da177e4 | 893 | } |
a112a71d | 894 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 895 | |
2cefe4db KO |
896 | /** |
897 | * bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio | |
898 | * @bio: bio to add pages to | |
899 | * @iter: iov iterator describing the region to be mapped | |
900 | * | |
901 | * Pins as many pages from *iter and appends them to @bio's bvec array. The | |
902 | * pages will have to be released using put_page() when done. | |
903 | */ | |
904 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) | |
905 | { | |
906 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; | |
907 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; | |
908 | struct page **pages = (struct page **)bv; | |
909 | size_t offset, diff; | |
910 | ssize_t size; | |
911 | ||
912 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); | |
913 | if (unlikely(size <= 0)) | |
914 | return size ? size : -EFAULT; | |
915 | nr_pages = (size + offset + PAGE_SIZE - 1) / PAGE_SIZE; | |
916 | ||
917 | /* | |
918 | * Deep magic below: We need to walk the pinned pages backwards | |
919 | * because we are abusing the space allocated for the bio_vecs | |
920 | * for the page array. Because the bio_vecs are larger than the | |
921 | * page pointers by definition this will always work. But it also | |
922 | * means we can't use bio_add_page, so any changes to it's semantics | |
923 | * need to be reflected here as well. | |
924 | */ | |
925 | bio->bi_iter.bi_size += size; | |
926 | bio->bi_vcnt += nr_pages; | |
927 | ||
928 | diff = (nr_pages * PAGE_SIZE - offset) - size; | |
929 | while (nr_pages--) { | |
930 | bv[nr_pages].bv_page = pages[nr_pages]; | |
931 | bv[nr_pages].bv_len = PAGE_SIZE; | |
932 | bv[nr_pages].bv_offset = 0; | |
933 | } | |
934 | ||
935 | bv[0].bv_offset += offset; | |
936 | bv[0].bv_len -= offset; | |
937 | if (diff) | |
938 | bv[bio->bi_vcnt - 1].bv_len -= diff; | |
939 | ||
940 | iov_iter_advance(iter, size); | |
941 | return 0; | |
942 | } | |
943 | EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages); | |
944 | ||
9e882242 KO |
945 | struct submit_bio_ret { |
946 | struct completion event; | |
947 | int error; | |
948 | }; | |
949 | ||
4246a0b6 | 950 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 KO |
951 | { |
952 | struct submit_bio_ret *ret = bio->bi_private; | |
953 | ||
4246a0b6 | 954 | ret->error = bio->bi_error; |
9e882242 KO |
955 | complete(&ret->event); |
956 | } | |
957 | ||
958 | /** | |
959 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
960 | * @bio: The &struct bio which describes the I/O |
961 | * | |
962 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
963 | * bio_endio() on failure. | |
964 | */ | |
4e49ea4a | 965 | int submit_bio_wait(struct bio *bio) |
9e882242 KO |
966 | { |
967 | struct submit_bio_ret ret; | |
968 | ||
9e882242 KO |
969 | init_completion(&ret.event); |
970 | bio->bi_private = &ret; | |
971 | bio->bi_end_io = submit_bio_wait_endio; | |
1eff9d32 | 972 | bio->bi_opf |= REQ_SYNC; |
4e49ea4a | 973 | submit_bio(bio); |
d57d6115 | 974 | wait_for_completion_io(&ret.event); |
9e882242 KO |
975 | |
976 | return ret.error; | |
977 | } | |
978 | EXPORT_SYMBOL(submit_bio_wait); | |
979 | ||
054bdf64 KO |
980 | /** |
981 | * bio_advance - increment/complete a bio by some number of bytes | |
982 | * @bio: bio to advance | |
983 | * @bytes: number of bytes to complete | |
984 | * | |
985 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
986 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
987 | * be updated on the last bvec as well. | |
988 | * | |
989 | * @bio will then represent the remaining, uncompleted portion of the io. | |
990 | */ | |
991 | void bio_advance(struct bio *bio, unsigned bytes) | |
992 | { | |
993 | if (bio_integrity(bio)) | |
994 | bio_integrity_advance(bio, bytes); | |
995 | ||
4550dd6c | 996 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
997 | } |
998 | EXPORT_SYMBOL(bio_advance); | |
999 | ||
a0787606 KO |
1000 | /** |
1001 | * bio_alloc_pages - allocates a single page for each bvec in a bio | |
1002 | * @bio: bio to allocate pages for | |
1003 | * @gfp_mask: flags for allocation | |
1004 | * | |
1005 | * Allocates pages up to @bio->bi_vcnt. | |
1006 | * | |
1007 | * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are | |
1008 | * freed. | |
1009 | */ | |
1010 | int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask) | |
1011 | { | |
1012 | int i; | |
1013 | struct bio_vec *bv; | |
1014 | ||
1015 | bio_for_each_segment_all(bv, bio, i) { | |
1016 | bv->bv_page = alloc_page(gfp_mask); | |
1017 | if (!bv->bv_page) { | |
1018 | while (--bv >= bio->bi_io_vec) | |
1019 | __free_page(bv->bv_page); | |
1020 | return -ENOMEM; | |
1021 | } | |
1022 | } | |
1023 | ||
1024 | return 0; | |
1025 | } | |
1026 | EXPORT_SYMBOL(bio_alloc_pages); | |
1027 | ||
16ac3d63 KO |
1028 | /** |
1029 | * bio_copy_data - copy contents of data buffers from one chain of bios to | |
1030 | * another | |
1031 | * @src: source bio list | |
1032 | * @dst: destination bio list | |
1033 | * | |
1034 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats | |
1035 | * @src and @dst as linked lists of bios. | |
1036 | * | |
1037 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
1038 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
1039 | */ | |
1040 | void bio_copy_data(struct bio *dst, struct bio *src) | |
1041 | { | |
1cb9dda4 KO |
1042 | struct bvec_iter src_iter, dst_iter; |
1043 | struct bio_vec src_bv, dst_bv; | |
16ac3d63 | 1044 | void *src_p, *dst_p; |
1cb9dda4 | 1045 | unsigned bytes; |
16ac3d63 | 1046 | |
1cb9dda4 KO |
1047 | src_iter = src->bi_iter; |
1048 | dst_iter = dst->bi_iter; | |
16ac3d63 KO |
1049 | |
1050 | while (1) { | |
1cb9dda4 KO |
1051 | if (!src_iter.bi_size) { |
1052 | src = src->bi_next; | |
1053 | if (!src) | |
1054 | break; | |
16ac3d63 | 1055 | |
1cb9dda4 | 1056 | src_iter = src->bi_iter; |
16ac3d63 KO |
1057 | } |
1058 | ||
1cb9dda4 KO |
1059 | if (!dst_iter.bi_size) { |
1060 | dst = dst->bi_next; | |
1061 | if (!dst) | |
1062 | break; | |
16ac3d63 | 1063 | |
1cb9dda4 | 1064 | dst_iter = dst->bi_iter; |
16ac3d63 KO |
1065 | } |
1066 | ||
1cb9dda4 KO |
1067 | src_bv = bio_iter_iovec(src, src_iter); |
1068 | dst_bv = bio_iter_iovec(dst, dst_iter); | |
1069 | ||
1070 | bytes = min(src_bv.bv_len, dst_bv.bv_len); | |
16ac3d63 | 1071 | |
1cb9dda4 KO |
1072 | src_p = kmap_atomic(src_bv.bv_page); |
1073 | dst_p = kmap_atomic(dst_bv.bv_page); | |
16ac3d63 | 1074 | |
1cb9dda4 KO |
1075 | memcpy(dst_p + dst_bv.bv_offset, |
1076 | src_p + src_bv.bv_offset, | |
16ac3d63 KO |
1077 | bytes); |
1078 | ||
1079 | kunmap_atomic(dst_p); | |
1080 | kunmap_atomic(src_p); | |
1081 | ||
1cb9dda4 KO |
1082 | bio_advance_iter(src, &src_iter, bytes); |
1083 | bio_advance_iter(dst, &dst_iter, bytes); | |
16ac3d63 KO |
1084 | } |
1085 | } | |
1086 | EXPORT_SYMBOL(bio_copy_data); | |
1087 | ||
1da177e4 | 1088 | struct bio_map_data { |
152e283f | 1089 | int is_our_pages; |
26e49cfc KO |
1090 | struct iov_iter iter; |
1091 | struct iovec iov[]; | |
1da177e4 LT |
1092 | }; |
1093 | ||
7410b3c6 | 1094 | static struct bio_map_data *bio_alloc_map_data(unsigned int iov_count, |
76029ff3 | 1095 | gfp_t gfp_mask) |
1da177e4 | 1096 | { |
f3f63c1c JA |
1097 | if (iov_count > UIO_MAXIOV) |
1098 | return NULL; | |
1da177e4 | 1099 | |
c8db4448 | 1100 | return kmalloc(sizeof(struct bio_map_data) + |
26e49cfc | 1101 | sizeof(struct iovec) * iov_count, gfp_mask); |
1da177e4 LT |
1102 | } |
1103 | ||
9124d3fe DP |
1104 | /** |
1105 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1106 | * @bio: The &struct bio which describes the I/O as destination | |
1107 | * @iter: iov_iter as source | |
1108 | * | |
1109 | * Copy all pages from iov_iter to bio. | |
1110 | * Returns 0 on success, or error on failure. | |
1111 | */ | |
1112 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter iter) | |
c5dec1c3 | 1113 | { |
9124d3fe | 1114 | int i; |
c5dec1c3 | 1115 | struct bio_vec *bvec; |
c5dec1c3 | 1116 | |
d74c6d51 | 1117 | bio_for_each_segment_all(bvec, bio, i) { |
9124d3fe | 1118 | ssize_t ret; |
c5dec1c3 | 1119 | |
9124d3fe DP |
1120 | ret = copy_page_from_iter(bvec->bv_page, |
1121 | bvec->bv_offset, | |
1122 | bvec->bv_len, | |
1123 | &iter); | |
1124 | ||
1125 | if (!iov_iter_count(&iter)) | |
1126 | break; | |
1127 | ||
1128 | if (ret < bvec->bv_len) | |
1129 | return -EFAULT; | |
c5dec1c3 FT |
1130 | } |
1131 | ||
9124d3fe DP |
1132 | return 0; |
1133 | } | |
1134 | ||
1135 | /** | |
1136 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1137 | * @bio: The &struct bio which describes the I/O as source | |
1138 | * @iter: iov_iter as destination | |
1139 | * | |
1140 | * Copy all pages from bio to iov_iter. | |
1141 | * Returns 0 on success, or error on failure. | |
1142 | */ | |
1143 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1144 | { | |
1145 | int i; | |
1146 | struct bio_vec *bvec; | |
1147 | ||
1148 | bio_for_each_segment_all(bvec, bio, i) { | |
1149 | ssize_t ret; | |
1150 | ||
1151 | ret = copy_page_to_iter(bvec->bv_page, | |
1152 | bvec->bv_offset, | |
1153 | bvec->bv_len, | |
1154 | &iter); | |
1155 | ||
1156 | if (!iov_iter_count(&iter)) | |
1157 | break; | |
1158 | ||
1159 | if (ret < bvec->bv_len) | |
1160 | return -EFAULT; | |
1161 | } | |
1162 | ||
1163 | return 0; | |
c5dec1c3 FT |
1164 | } |
1165 | ||
491221f8 | 1166 | void bio_free_pages(struct bio *bio) |
1dfa0f68 CH |
1167 | { |
1168 | struct bio_vec *bvec; | |
1169 | int i; | |
1170 | ||
1171 | bio_for_each_segment_all(bvec, bio, i) | |
1172 | __free_page(bvec->bv_page); | |
1173 | } | |
491221f8 | 1174 | EXPORT_SYMBOL(bio_free_pages); |
1dfa0f68 | 1175 | |
1da177e4 LT |
1176 | /** |
1177 | * bio_uncopy_user - finish previously mapped bio | |
1178 | * @bio: bio being terminated | |
1179 | * | |
ddad8dd0 | 1180 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1181 | * to user space in case of a read. |
1182 | */ | |
1183 | int bio_uncopy_user(struct bio *bio) | |
1184 | { | |
1185 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1186 | int ret = 0; |
1da177e4 | 1187 | |
35dc2483 RD |
1188 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1189 | /* | |
1190 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1191 | * don't copy into a random user address space, just free |
1192 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1193 | */ |
2d99b55d HR |
1194 | if (!current->mm) |
1195 | ret = -EINTR; | |
1196 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1197 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1198 | if (bmd->is_our_pages) |
1199 | bio_free_pages(bio); | |
35dc2483 | 1200 | } |
c8db4448 | 1201 | kfree(bmd); |
1da177e4 LT |
1202 | bio_put(bio); |
1203 | return ret; | |
1204 | } | |
1205 | ||
1206 | /** | |
c5dec1c3 | 1207 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1208 | * @q: destination block queue |
1209 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1210 | * @iter: iovec iterator | |
1211 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1212 | * |
1213 | * Prepares and returns a bio for indirect user io, bouncing data | |
1214 | * to/from kernel pages as necessary. Must be paired with | |
1215 | * call bio_uncopy_user() on io completion. | |
1216 | */ | |
152e283f FT |
1217 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1218 | struct rq_map_data *map_data, | |
26e49cfc KO |
1219 | const struct iov_iter *iter, |
1220 | gfp_t gfp_mask) | |
1da177e4 | 1221 | { |
1da177e4 | 1222 | struct bio_map_data *bmd; |
1da177e4 LT |
1223 | struct page *page; |
1224 | struct bio *bio; | |
1225 | int i, ret; | |
c5dec1c3 | 1226 | int nr_pages = 0; |
26e49cfc | 1227 | unsigned int len = iter->count; |
bd5cecea | 1228 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1229 | |
26e49cfc | 1230 | for (i = 0; i < iter->nr_segs; i++) { |
c5dec1c3 FT |
1231 | unsigned long uaddr; |
1232 | unsigned long end; | |
1233 | unsigned long start; | |
1234 | ||
26e49cfc KO |
1235 | uaddr = (unsigned long) iter->iov[i].iov_base; |
1236 | end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1) | |
1237 | >> PAGE_SHIFT; | |
c5dec1c3 FT |
1238 | start = uaddr >> PAGE_SHIFT; |
1239 | ||
cb4644ca JA |
1240 | /* |
1241 | * Overflow, abort | |
1242 | */ | |
1243 | if (end < start) | |
1244 | return ERR_PTR(-EINVAL); | |
1245 | ||
c5dec1c3 | 1246 | nr_pages += end - start; |
c5dec1c3 FT |
1247 | } |
1248 | ||
69838727 FT |
1249 | if (offset) |
1250 | nr_pages++; | |
1251 | ||
26e49cfc | 1252 | bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask); |
1da177e4 LT |
1253 | if (!bmd) |
1254 | return ERR_PTR(-ENOMEM); | |
1255 | ||
26e49cfc KO |
1256 | /* |
1257 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1258 | * The caller provided iov might point to an on-stack or otherwise | |
1259 | * shortlived one. | |
1260 | */ | |
1261 | bmd->is_our_pages = map_data ? 0 : 1; | |
1262 | memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs); | |
1263 | iov_iter_init(&bmd->iter, iter->type, bmd->iov, | |
1264 | iter->nr_segs, iter->count); | |
1265 | ||
1da177e4 | 1266 | ret = -ENOMEM; |
a9e9dc24 | 1267 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1268 | if (!bio) |
1269 | goto out_bmd; | |
1270 | ||
1da177e4 | 1271 | ret = 0; |
56c451f4 FT |
1272 | |
1273 | if (map_data) { | |
e623ddb4 | 1274 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1275 | i = map_data->offset / PAGE_SIZE; |
1276 | } | |
1da177e4 | 1277 | while (len) { |
e623ddb4 | 1278 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1279 | |
56c451f4 FT |
1280 | bytes -= offset; |
1281 | ||
1da177e4 LT |
1282 | if (bytes > len) |
1283 | bytes = len; | |
1284 | ||
152e283f | 1285 | if (map_data) { |
e623ddb4 | 1286 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1287 | ret = -ENOMEM; |
1288 | break; | |
1289 | } | |
e623ddb4 FT |
1290 | |
1291 | page = map_data->pages[i / nr_pages]; | |
1292 | page += (i % nr_pages); | |
1293 | ||
1294 | i++; | |
1295 | } else { | |
152e283f | 1296 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1297 | if (!page) { |
1298 | ret = -ENOMEM; | |
1299 | break; | |
1300 | } | |
1da177e4 LT |
1301 | } |
1302 | ||
56c451f4 | 1303 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 1304 | break; |
1da177e4 LT |
1305 | |
1306 | len -= bytes; | |
56c451f4 | 1307 | offset = 0; |
1da177e4 LT |
1308 | } |
1309 | ||
1310 | if (ret) | |
1311 | goto cleanup; | |
1312 | ||
1313 | /* | |
1314 | * success | |
1315 | */ | |
26e49cfc | 1316 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1317 | (map_data && map_data->from_user)) { |
9124d3fe | 1318 | ret = bio_copy_from_iter(bio, *iter); |
c5dec1c3 FT |
1319 | if (ret) |
1320 | goto cleanup; | |
1da177e4 LT |
1321 | } |
1322 | ||
26e49cfc | 1323 | bio->bi_private = bmd; |
1da177e4 LT |
1324 | return bio; |
1325 | cleanup: | |
152e283f | 1326 | if (!map_data) |
1dfa0f68 | 1327 | bio_free_pages(bio); |
1da177e4 LT |
1328 | bio_put(bio); |
1329 | out_bmd: | |
c8db4448 | 1330 | kfree(bmd); |
1da177e4 LT |
1331 | return ERR_PTR(ret); |
1332 | } | |
1333 | ||
37f19e57 CH |
1334 | /** |
1335 | * bio_map_user_iov - map user iovec into bio | |
1336 | * @q: the struct request_queue for the bio | |
1337 | * @iter: iovec iterator | |
1338 | * @gfp_mask: memory allocation flags | |
1339 | * | |
1340 | * Map the user space address into a bio suitable for io to a block | |
1341 | * device. Returns an error pointer in case of error. | |
1342 | */ | |
1343 | struct bio *bio_map_user_iov(struct request_queue *q, | |
1344 | const struct iov_iter *iter, | |
1345 | gfp_t gfp_mask) | |
1da177e4 | 1346 | { |
26e49cfc | 1347 | int j; |
f1970baf | 1348 | int nr_pages = 0; |
1da177e4 LT |
1349 | struct page **pages; |
1350 | struct bio *bio; | |
f1970baf JB |
1351 | int cur_page = 0; |
1352 | int ret, offset; | |
26e49cfc KO |
1353 | struct iov_iter i; |
1354 | struct iovec iov; | |
1da177e4 | 1355 | |
26e49cfc KO |
1356 | iov_for_each(iov, i, *iter) { |
1357 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1358 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1359 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1360 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1361 | ||
cb4644ca JA |
1362 | /* |
1363 | * Overflow, abort | |
1364 | */ | |
1365 | if (end < start) | |
1366 | return ERR_PTR(-EINVAL); | |
1367 | ||
f1970baf JB |
1368 | nr_pages += end - start; |
1369 | /* | |
a441b0d0 | 1370 | * buffer must be aligned to at least logical block size for now |
f1970baf | 1371 | */ |
ad2d7225 | 1372 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
1373 | return ERR_PTR(-EINVAL); |
1374 | } | |
1375 | ||
1376 | if (!nr_pages) | |
1da177e4 LT |
1377 | return ERR_PTR(-EINVAL); |
1378 | ||
a9e9dc24 | 1379 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1380 | if (!bio) |
1381 | return ERR_PTR(-ENOMEM); | |
1382 | ||
1383 | ret = -ENOMEM; | |
a3bce90e | 1384 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
1385 | if (!pages) |
1386 | goto out; | |
1387 | ||
26e49cfc KO |
1388 | iov_for_each(iov, i, *iter) { |
1389 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1390 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1391 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1392 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1393 | const int local_nr_pages = end - start; | |
1394 | const int page_limit = cur_page + local_nr_pages; | |
cb4644ca | 1395 | |
f5dd33c4 | 1396 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
26e49cfc KO |
1397 | (iter->type & WRITE) != WRITE, |
1398 | &pages[cur_page]); | |
99172157 JA |
1399 | if (ret < local_nr_pages) { |
1400 | ret = -EFAULT; | |
f1970baf | 1401 | goto out_unmap; |
99172157 | 1402 | } |
f1970baf | 1403 | |
bd5cecea | 1404 | offset = offset_in_page(uaddr); |
f1970baf JB |
1405 | for (j = cur_page; j < page_limit; j++) { |
1406 | unsigned int bytes = PAGE_SIZE - offset; | |
1407 | ||
1408 | if (len <= 0) | |
1409 | break; | |
1410 | ||
1411 | if (bytes > len) | |
1412 | bytes = len; | |
1413 | ||
1414 | /* | |
1415 | * sorry... | |
1416 | */ | |
defd94b7 MC |
1417 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1418 | bytes) | |
f1970baf JB |
1419 | break; |
1420 | ||
1421 | len -= bytes; | |
1422 | offset = 0; | |
1423 | } | |
1da177e4 | 1424 | |
f1970baf | 1425 | cur_page = j; |
1da177e4 | 1426 | /* |
f1970baf | 1427 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1428 | */ |
f1970baf | 1429 | while (j < page_limit) |
09cbfeaf | 1430 | put_page(pages[j++]); |
1da177e4 LT |
1431 | } |
1432 | ||
1da177e4 LT |
1433 | kfree(pages); |
1434 | ||
b7c44ed9 | 1435 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1436 | |
1437 | /* | |
5fad1b64 | 1438 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
37f19e57 CH |
1439 | * it would normally disappear when its bi_end_io is run. |
1440 | * however, we need it for the unmap, so grab an extra | |
1441 | * reference to it | |
1442 | */ | |
1443 | bio_get(bio); | |
1da177e4 | 1444 | return bio; |
f1970baf JB |
1445 | |
1446 | out_unmap: | |
26e49cfc KO |
1447 | for (j = 0; j < nr_pages; j++) { |
1448 | if (!pages[j]) | |
f1970baf | 1449 | break; |
09cbfeaf | 1450 | put_page(pages[j]); |
f1970baf JB |
1451 | } |
1452 | out: | |
1da177e4 LT |
1453 | kfree(pages); |
1454 | bio_put(bio); | |
1455 | return ERR_PTR(ret); | |
1456 | } | |
1457 | ||
1da177e4 LT |
1458 | static void __bio_unmap_user(struct bio *bio) |
1459 | { | |
1460 | struct bio_vec *bvec; | |
1461 | int i; | |
1462 | ||
1463 | /* | |
1464 | * make sure we dirty pages we wrote to | |
1465 | */ | |
d74c6d51 | 1466 | bio_for_each_segment_all(bvec, bio, i) { |
1da177e4 LT |
1467 | if (bio_data_dir(bio) == READ) |
1468 | set_page_dirty_lock(bvec->bv_page); | |
1469 | ||
09cbfeaf | 1470 | put_page(bvec->bv_page); |
1da177e4 LT |
1471 | } |
1472 | ||
1473 | bio_put(bio); | |
1474 | } | |
1475 | ||
1476 | /** | |
1477 | * bio_unmap_user - unmap a bio | |
1478 | * @bio: the bio being unmapped | |
1479 | * | |
5fad1b64 BVA |
1480 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
1481 | * process context. | |
1da177e4 LT |
1482 | * |
1483 | * bio_unmap_user() may sleep. | |
1484 | */ | |
1485 | void bio_unmap_user(struct bio *bio) | |
1486 | { | |
1487 | __bio_unmap_user(bio); | |
1488 | bio_put(bio); | |
1489 | } | |
1490 | ||
4246a0b6 | 1491 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1492 | { |
b823825e | 1493 | bio_put(bio); |
b823825e JA |
1494 | } |
1495 | ||
75c72b83 CH |
1496 | /** |
1497 | * bio_map_kern - map kernel address into bio | |
1498 | * @q: the struct request_queue for the bio | |
1499 | * @data: pointer to buffer to map | |
1500 | * @len: length in bytes | |
1501 | * @gfp_mask: allocation flags for bio allocation | |
1502 | * | |
1503 | * Map the kernel address into a bio suitable for io to a block | |
1504 | * device. Returns an error pointer in case of error. | |
1505 | */ | |
1506 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1507 | gfp_t gfp_mask) | |
df46b9a4 MC |
1508 | { |
1509 | unsigned long kaddr = (unsigned long)data; | |
1510 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1511 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1512 | const int nr_pages = end - start; | |
1513 | int offset, i; | |
1514 | struct bio *bio; | |
1515 | ||
a9e9dc24 | 1516 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1517 | if (!bio) |
1518 | return ERR_PTR(-ENOMEM); | |
1519 | ||
1520 | offset = offset_in_page(kaddr); | |
1521 | for (i = 0; i < nr_pages; i++) { | |
1522 | unsigned int bytes = PAGE_SIZE - offset; | |
1523 | ||
1524 | if (len <= 0) | |
1525 | break; | |
1526 | ||
1527 | if (bytes > len) | |
1528 | bytes = len; | |
1529 | ||
defd94b7 | 1530 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1531 | offset) < bytes) { |
1532 | /* we don't support partial mappings */ | |
1533 | bio_put(bio); | |
1534 | return ERR_PTR(-EINVAL); | |
1535 | } | |
df46b9a4 MC |
1536 | |
1537 | data += bytes; | |
1538 | len -= bytes; | |
1539 | offset = 0; | |
1540 | } | |
1541 | ||
b823825e | 1542 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1543 | return bio; |
1544 | } | |
a112a71d | 1545 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1546 | |
4246a0b6 | 1547 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1548 | { |
1dfa0f68 CH |
1549 | bio_free_pages(bio); |
1550 | bio_put(bio); | |
1551 | } | |
1552 | ||
4246a0b6 | 1553 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1554 | { |
42d2683a | 1555 | char *p = bio->bi_private; |
1dfa0f68 | 1556 | struct bio_vec *bvec; |
68154e90 FT |
1557 | int i; |
1558 | ||
d74c6d51 | 1559 | bio_for_each_segment_all(bvec, bio, i) { |
1dfa0f68 | 1560 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1561 | p += bvec->bv_len; |
68154e90 FT |
1562 | } |
1563 | ||
4246a0b6 | 1564 | bio_copy_kern_endio(bio); |
68154e90 FT |
1565 | } |
1566 | ||
1567 | /** | |
1568 | * bio_copy_kern - copy kernel address into bio | |
1569 | * @q: the struct request_queue for the bio | |
1570 | * @data: pointer to buffer to copy | |
1571 | * @len: length in bytes | |
1572 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1573 | * @reading: data direction is READ |
68154e90 FT |
1574 | * |
1575 | * copy the kernel address into a bio suitable for io to a block | |
1576 | * device. Returns an error pointer in case of error. | |
1577 | */ | |
1578 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1579 | gfp_t gfp_mask, int reading) | |
1580 | { | |
42d2683a CH |
1581 | unsigned long kaddr = (unsigned long)data; |
1582 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1583 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1584 | struct bio *bio; |
1585 | void *p = data; | |
1dfa0f68 | 1586 | int nr_pages = 0; |
68154e90 | 1587 | |
42d2683a CH |
1588 | /* |
1589 | * Overflow, abort | |
1590 | */ | |
1591 | if (end < start) | |
1592 | return ERR_PTR(-EINVAL); | |
68154e90 | 1593 | |
42d2683a CH |
1594 | nr_pages = end - start; |
1595 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1596 | if (!bio) | |
1597 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1598 | |
42d2683a CH |
1599 | while (len) { |
1600 | struct page *page; | |
1601 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1602 | |
42d2683a CH |
1603 | if (bytes > len) |
1604 | bytes = len; | |
1605 | ||
1606 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1607 | if (!page) | |
1608 | goto cleanup; | |
1609 | ||
1610 | if (!reading) | |
1611 | memcpy(page_address(page), p, bytes); | |
1612 | ||
1613 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1614 | break; | |
1615 | ||
1616 | len -= bytes; | |
1617 | p += bytes; | |
68154e90 FT |
1618 | } |
1619 | ||
1dfa0f68 CH |
1620 | if (reading) { |
1621 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1622 | bio->bi_private = data; | |
1623 | } else { | |
1624 | bio->bi_end_io = bio_copy_kern_endio; | |
1dfa0f68 | 1625 | } |
76029ff3 | 1626 | |
68154e90 | 1627 | return bio; |
42d2683a CH |
1628 | |
1629 | cleanup: | |
1dfa0f68 | 1630 | bio_free_pages(bio); |
42d2683a CH |
1631 | bio_put(bio); |
1632 | return ERR_PTR(-ENOMEM); | |
68154e90 FT |
1633 | } |
1634 | ||
1da177e4 LT |
1635 | /* |
1636 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1637 | * for performing direct-IO in BIOs. | |
1638 | * | |
1639 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1640 | * because the required locks are not interrupt-safe. So what we can do is to | |
1641 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1642 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1643 | * in process context. | |
1644 | * | |
1645 | * We special-case compound pages here: normally this means reads into hugetlb | |
1646 | * pages. The logic in here doesn't really work right for compound pages | |
1647 | * because the VM does not uniformly chase down the head page in all cases. | |
1648 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1649 | * handle them at all. So we skip compound pages here at an early stage. | |
1650 | * | |
1651 | * Note that this code is very hard to test under normal circumstances because | |
1652 | * direct-io pins the pages with get_user_pages(). This makes | |
1653 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1654 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1655 | * pagecache. |
1656 | * | |
1657 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1658 | * deferred bio dirtying paths. | |
1659 | */ | |
1660 | ||
1661 | /* | |
1662 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1663 | */ | |
1664 | void bio_set_pages_dirty(struct bio *bio) | |
1665 | { | |
cb34e057 | 1666 | struct bio_vec *bvec; |
1da177e4 LT |
1667 | int i; |
1668 | ||
cb34e057 KO |
1669 | bio_for_each_segment_all(bvec, bio, i) { |
1670 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1671 | |
1672 | if (page && !PageCompound(page)) | |
1673 | set_page_dirty_lock(page); | |
1674 | } | |
1675 | } | |
1676 | ||
86b6c7a7 | 1677 | static void bio_release_pages(struct bio *bio) |
1da177e4 | 1678 | { |
cb34e057 | 1679 | struct bio_vec *bvec; |
1da177e4 LT |
1680 | int i; |
1681 | ||
cb34e057 KO |
1682 | bio_for_each_segment_all(bvec, bio, i) { |
1683 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1684 | |
1685 | if (page) | |
1686 | put_page(page); | |
1687 | } | |
1688 | } | |
1689 | ||
1690 | /* | |
1691 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1692 | * If they are, then fine. If, however, some pages are clean then they must | |
1693 | * have been written out during the direct-IO read. So we take another ref on | |
1694 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1695 | * | |
1696 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1697 | * here on. It will run one put_page() against each page and will run one |
1698 | * bio_put() against the BIO. | |
1da177e4 LT |
1699 | */ |
1700 | ||
65f27f38 | 1701 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1702 | |
65f27f38 | 1703 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1704 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1705 | static struct bio *bio_dirty_list; | |
1706 | ||
1707 | /* | |
1708 | * This runs in process context | |
1709 | */ | |
65f27f38 | 1710 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1711 | { |
1712 | unsigned long flags; | |
1713 | struct bio *bio; | |
1714 | ||
1715 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1716 | bio = bio_dirty_list; | |
1717 | bio_dirty_list = NULL; | |
1718 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1719 | ||
1720 | while (bio) { | |
1721 | struct bio *next = bio->bi_private; | |
1722 | ||
1723 | bio_set_pages_dirty(bio); | |
1724 | bio_release_pages(bio); | |
1725 | bio_put(bio); | |
1726 | bio = next; | |
1727 | } | |
1728 | } | |
1729 | ||
1730 | void bio_check_pages_dirty(struct bio *bio) | |
1731 | { | |
cb34e057 | 1732 | struct bio_vec *bvec; |
1da177e4 LT |
1733 | int nr_clean_pages = 0; |
1734 | int i; | |
1735 | ||
cb34e057 KO |
1736 | bio_for_each_segment_all(bvec, bio, i) { |
1737 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1738 | |
1739 | if (PageDirty(page) || PageCompound(page)) { | |
09cbfeaf | 1740 | put_page(page); |
cb34e057 | 1741 | bvec->bv_page = NULL; |
1da177e4 LT |
1742 | } else { |
1743 | nr_clean_pages++; | |
1744 | } | |
1745 | } | |
1746 | ||
1747 | if (nr_clean_pages) { | |
1748 | unsigned long flags; | |
1749 | ||
1750 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1751 | bio->bi_private = bio_dirty_list; | |
1752 | bio_dirty_list = bio; | |
1753 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1754 | schedule_work(&bio_dirty_work); | |
1755 | } else { | |
1756 | bio_put(bio); | |
1757 | } | |
1758 | } | |
1759 | ||
394ffa50 GZ |
1760 | void generic_start_io_acct(int rw, unsigned long sectors, |
1761 | struct hd_struct *part) | |
1762 | { | |
1763 | int cpu = part_stat_lock(); | |
1764 | ||
1765 | part_round_stats(cpu, part); | |
1766 | part_stat_inc(cpu, part, ios[rw]); | |
1767 | part_stat_add(cpu, part, sectors[rw], sectors); | |
1768 | part_inc_in_flight(part, rw); | |
1769 | ||
1770 | part_stat_unlock(); | |
1771 | } | |
1772 | EXPORT_SYMBOL(generic_start_io_acct); | |
1773 | ||
1774 | void generic_end_io_acct(int rw, struct hd_struct *part, | |
1775 | unsigned long start_time) | |
1776 | { | |
1777 | unsigned long duration = jiffies - start_time; | |
1778 | int cpu = part_stat_lock(); | |
1779 | ||
1780 | part_stat_add(cpu, part, ticks[rw], duration); | |
1781 | part_round_stats(cpu, part); | |
1782 | part_dec_in_flight(part, rw); | |
1783 | ||
1784 | part_stat_unlock(); | |
1785 | } | |
1786 | EXPORT_SYMBOL(generic_end_io_acct); | |
1787 | ||
2d4dc890 IL |
1788 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1789 | void bio_flush_dcache_pages(struct bio *bi) | |
1790 | { | |
7988613b KO |
1791 | struct bio_vec bvec; |
1792 | struct bvec_iter iter; | |
2d4dc890 | 1793 | |
7988613b KO |
1794 | bio_for_each_segment(bvec, bi, iter) |
1795 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1796 | } |
1797 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1798 | #endif | |
1799 | ||
c4cf5261 JA |
1800 | static inline bool bio_remaining_done(struct bio *bio) |
1801 | { | |
1802 | /* | |
1803 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1804 | * we always end io on the first invocation. | |
1805 | */ | |
1806 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1807 | return true; | |
1808 | ||
1809 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1810 | ||
326e1dbb | 1811 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1812 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1813 | return true; |
326e1dbb | 1814 | } |
c4cf5261 JA |
1815 | |
1816 | return false; | |
1817 | } | |
1818 | ||
1da177e4 LT |
1819 | /** |
1820 | * bio_endio - end I/O on a bio | |
1821 | * @bio: bio | |
1da177e4 LT |
1822 | * |
1823 | * Description: | |
4246a0b6 CH |
1824 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1825 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1826 | * bio unless they own it and thus know that it has an end_io function. | |
1da177e4 | 1827 | **/ |
4246a0b6 | 1828 | void bio_endio(struct bio *bio) |
1da177e4 | 1829 | { |
ba8c6967 | 1830 | again: |
2b885517 | 1831 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1832 | return; |
1da177e4 | 1833 | |
ba8c6967 CH |
1834 | /* |
1835 | * Need to have a real endio function for chained bios, otherwise | |
1836 | * various corner cases will break (like stacking block devices that | |
1837 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1838 | * recursion and blowing the stack. Tail call optimization would | |
1839 | * handle this, but compiling with frame pointers also disables | |
1840 | * gcc's sibling call optimization. | |
1841 | */ | |
1842 | if (bio->bi_end_io == bio_chain_endio) { | |
1843 | bio = __bio_chain_endio(bio); | |
1844 | goto again; | |
196d38bc | 1845 | } |
ba8c6967 CH |
1846 | |
1847 | if (bio->bi_end_io) | |
1848 | bio->bi_end_io(bio); | |
1da177e4 | 1849 | } |
a112a71d | 1850 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1851 | |
20d0189b KO |
1852 | /** |
1853 | * bio_split - split a bio | |
1854 | * @bio: bio to split | |
1855 | * @sectors: number of sectors to split from the front of @bio | |
1856 | * @gfp: gfp mask | |
1857 | * @bs: bio set to allocate from | |
1858 | * | |
1859 | * Allocates and returns a new bio which represents @sectors from the start of | |
1860 | * @bio, and updates @bio to represent the remaining sectors. | |
1861 | * | |
f3f5da62 MP |
1862 | * Unless this is a discard request the newly allocated bio will point |
1863 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1864 | * @bio is not freed before the split. | |
20d0189b KO |
1865 | */ |
1866 | struct bio *bio_split(struct bio *bio, int sectors, | |
1867 | gfp_t gfp, struct bio_set *bs) | |
1868 | { | |
1869 | struct bio *split = NULL; | |
1870 | ||
1871 | BUG_ON(sectors <= 0); | |
1872 | BUG_ON(sectors >= bio_sectors(bio)); | |
1873 | ||
f9d03f96 | 1874 | split = bio_clone_fast(bio, gfp, bs); |
20d0189b KO |
1875 | if (!split) |
1876 | return NULL; | |
1877 | ||
1878 | split->bi_iter.bi_size = sectors << 9; | |
1879 | ||
1880 | if (bio_integrity(split)) | |
1881 | bio_integrity_trim(split, 0, sectors); | |
1882 | ||
1883 | bio_advance(bio, split->bi_iter.bi_size); | |
1884 | ||
1885 | return split; | |
1886 | } | |
1887 | EXPORT_SYMBOL(bio_split); | |
1888 | ||
6678d83f KO |
1889 | /** |
1890 | * bio_trim - trim a bio | |
1891 | * @bio: bio to trim | |
1892 | * @offset: number of sectors to trim from the front of @bio | |
1893 | * @size: size we want to trim @bio to, in sectors | |
1894 | */ | |
1895 | void bio_trim(struct bio *bio, int offset, int size) | |
1896 | { | |
1897 | /* 'bio' is a cloned bio which we need to trim to match | |
1898 | * the given offset and size. | |
6678d83f | 1899 | */ |
6678d83f KO |
1900 | |
1901 | size <<= 9; | |
4f024f37 | 1902 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1903 | return; |
1904 | ||
b7c44ed9 | 1905 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1906 | |
1907 | bio_advance(bio, offset << 9); | |
1908 | ||
4f024f37 | 1909 | bio->bi_iter.bi_size = size; |
6678d83f KO |
1910 | } |
1911 | EXPORT_SYMBOL_GPL(bio_trim); | |
1912 | ||
1da177e4 LT |
1913 | /* |
1914 | * create memory pools for biovec's in a bio_set. | |
1915 | * use the global biovec slabs created for general use. | |
1916 | */ | |
a6c39cb4 | 1917 | mempool_t *biovec_create_pool(int pool_entries) |
1da177e4 | 1918 | { |
ed996a52 | 1919 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
1da177e4 | 1920 | |
9f060e22 | 1921 | return mempool_create_slab_pool(pool_entries, bp->slab); |
1da177e4 LT |
1922 | } |
1923 | ||
1924 | void bioset_free(struct bio_set *bs) | |
1925 | { | |
df2cb6da KO |
1926 | if (bs->rescue_workqueue) |
1927 | destroy_workqueue(bs->rescue_workqueue); | |
1928 | ||
1da177e4 LT |
1929 | if (bs->bio_pool) |
1930 | mempool_destroy(bs->bio_pool); | |
1931 | ||
9f060e22 KO |
1932 | if (bs->bvec_pool) |
1933 | mempool_destroy(bs->bvec_pool); | |
1934 | ||
7878cba9 | 1935 | bioset_integrity_free(bs); |
bb799ca0 | 1936 | bio_put_slab(bs); |
1da177e4 LT |
1937 | |
1938 | kfree(bs); | |
1939 | } | |
a112a71d | 1940 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1941 | |
d8f429e1 JN |
1942 | static struct bio_set *__bioset_create(unsigned int pool_size, |
1943 | unsigned int front_pad, | |
1944 | bool create_bvec_pool) | |
1da177e4 | 1945 | { |
392ddc32 | 1946 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1947 | struct bio_set *bs; |
1da177e4 | 1948 | |
1b434498 | 1949 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1950 | if (!bs) |
1951 | return NULL; | |
1952 | ||
bb799ca0 | 1953 | bs->front_pad = front_pad; |
1b434498 | 1954 | |
df2cb6da KO |
1955 | spin_lock_init(&bs->rescue_lock); |
1956 | bio_list_init(&bs->rescue_list); | |
1957 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
1958 | ||
392ddc32 | 1959 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1960 | if (!bs->bio_slab) { |
1961 | kfree(bs); | |
1962 | return NULL; | |
1963 | } | |
1964 | ||
1965 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
1966 | if (!bs->bio_pool) |
1967 | goto bad; | |
1968 | ||
d8f429e1 JN |
1969 | if (create_bvec_pool) { |
1970 | bs->bvec_pool = biovec_create_pool(pool_size); | |
1971 | if (!bs->bvec_pool) | |
1972 | goto bad; | |
1973 | } | |
df2cb6da KO |
1974 | |
1975 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); | |
1976 | if (!bs->rescue_workqueue) | |
1977 | goto bad; | |
1da177e4 | 1978 | |
df2cb6da | 1979 | return bs; |
1da177e4 LT |
1980 | bad: |
1981 | bioset_free(bs); | |
1982 | return NULL; | |
1983 | } | |
d8f429e1 JN |
1984 | |
1985 | /** | |
1986 | * bioset_create - Create a bio_set | |
1987 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1988 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1989 | * | |
1990 | * Description: | |
1991 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1992 | * to ask for a number of bytes to be allocated in front of the bio. | |
1993 | * Front pad allocation is useful for embedding the bio inside | |
1994 | * another structure, to avoid allocating extra data to go with the bio. | |
1995 | * Note that the bio must be embedded at the END of that structure always, | |
1996 | * or things will break badly. | |
1997 | */ | |
1998 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) | |
1999 | { | |
2000 | return __bioset_create(pool_size, front_pad, true); | |
2001 | } | |
a112a71d | 2002 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 2003 | |
d8f429e1 JN |
2004 | /** |
2005 | * bioset_create_nobvec - Create a bio_set without bio_vec mempool | |
2006 | * @pool_size: Number of bio to cache in the mempool | |
2007 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
2008 | * | |
2009 | * Description: | |
2010 | * Same functionality as bioset_create() except that mempool is not | |
2011 | * created for bio_vecs. Saving some memory for bio_clone_fast() users. | |
2012 | */ | |
2013 | struct bio_set *bioset_create_nobvec(unsigned int pool_size, unsigned int front_pad) | |
2014 | { | |
2015 | return __bioset_create(pool_size, front_pad, false); | |
2016 | } | |
2017 | EXPORT_SYMBOL(bioset_create_nobvec); | |
2018 | ||
852c788f | 2019 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 TH |
2020 | |
2021 | /** | |
2022 | * bio_associate_blkcg - associate a bio with the specified blkcg | |
2023 | * @bio: target bio | |
2024 | * @blkcg_css: css of the blkcg to associate | |
2025 | * | |
2026 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will | |
2027 | * treat @bio as if it were issued by a task which belongs to the blkcg. | |
2028 | * | |
2029 | * This function takes an extra reference of @blkcg_css which will be put | |
2030 | * when @bio is released. The caller must own @bio and is responsible for | |
2031 | * synchronizing calls to this function. | |
2032 | */ | |
2033 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) | |
2034 | { | |
2035 | if (unlikely(bio->bi_css)) | |
2036 | return -EBUSY; | |
2037 | css_get(blkcg_css); | |
2038 | bio->bi_css = blkcg_css; | |
2039 | return 0; | |
2040 | } | |
5aa2a96b | 2041 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
1d933cf0 | 2042 | |
852c788f TH |
2043 | /** |
2044 | * bio_associate_current - associate a bio with %current | |
2045 | * @bio: target bio | |
2046 | * | |
2047 | * Associate @bio with %current if it hasn't been associated yet. Block | |
2048 | * layer will treat @bio as if it were issued by %current no matter which | |
2049 | * task actually issues it. | |
2050 | * | |
2051 | * This function takes an extra reference of @task's io_context and blkcg | |
2052 | * which will be put when @bio is released. The caller must own @bio, | |
2053 | * ensure %current->io_context exists, and is responsible for synchronizing | |
2054 | * calls to this function. | |
2055 | */ | |
2056 | int bio_associate_current(struct bio *bio) | |
2057 | { | |
2058 | struct io_context *ioc; | |
852c788f | 2059 | |
1d933cf0 | 2060 | if (bio->bi_css) |
852c788f TH |
2061 | return -EBUSY; |
2062 | ||
2063 | ioc = current->io_context; | |
2064 | if (!ioc) | |
2065 | return -ENOENT; | |
2066 | ||
852c788f TH |
2067 | get_io_context_active(ioc); |
2068 | bio->bi_ioc = ioc; | |
c165b3e3 | 2069 | bio->bi_css = task_get_css(current, io_cgrp_id); |
852c788f TH |
2070 | return 0; |
2071 | } | |
5aa2a96b | 2072 | EXPORT_SYMBOL_GPL(bio_associate_current); |
852c788f TH |
2073 | |
2074 | /** | |
2075 | * bio_disassociate_task - undo bio_associate_current() | |
2076 | * @bio: target bio | |
2077 | */ | |
2078 | void bio_disassociate_task(struct bio *bio) | |
2079 | { | |
2080 | if (bio->bi_ioc) { | |
2081 | put_io_context(bio->bi_ioc); | |
2082 | bio->bi_ioc = NULL; | |
2083 | } | |
2084 | if (bio->bi_css) { | |
2085 | css_put(bio->bi_css); | |
2086 | bio->bi_css = NULL; | |
2087 | } | |
2088 | } | |
2089 | ||
20bd723e PV |
2090 | /** |
2091 | * bio_clone_blkcg_association - clone blkcg association from src to dst bio | |
2092 | * @dst: destination bio | |
2093 | * @src: source bio | |
2094 | */ | |
2095 | void bio_clone_blkcg_association(struct bio *dst, struct bio *src) | |
2096 | { | |
2097 | if (src->bi_css) | |
2098 | WARN_ON(bio_associate_blkcg(dst, src->bi_css)); | |
2099 | } | |
2100 | ||
852c788f TH |
2101 | #endif /* CONFIG_BLK_CGROUP */ |
2102 | ||
1da177e4 LT |
2103 | static void __init biovec_init_slabs(void) |
2104 | { | |
2105 | int i; | |
2106 | ||
ed996a52 | 2107 | for (i = 0; i < BVEC_POOL_NR; i++) { |
1da177e4 LT |
2108 | int size; |
2109 | struct biovec_slab *bvs = bvec_slabs + i; | |
2110 | ||
a7fcd37c JA |
2111 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2112 | bvs->slab = NULL; | |
2113 | continue; | |
2114 | } | |
a7fcd37c | 2115 | |
1da177e4 LT |
2116 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2117 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2118 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2119 | } |
2120 | } | |
2121 | ||
2122 | static int __init init_bio(void) | |
2123 | { | |
bb799ca0 JA |
2124 | bio_slab_max = 2; |
2125 | bio_slab_nr = 0; | |
2126 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
2127 | if (!bio_slabs) | |
2128 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2129 | |
7878cba9 | 2130 | bio_integrity_init(); |
1da177e4 LT |
2131 | biovec_init_slabs(); |
2132 | ||
bb799ca0 | 2133 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
1da177e4 LT |
2134 | if (!fs_bio_set) |
2135 | panic("bio: can't allocate bios\n"); | |
2136 | ||
a91a2785 MP |
2137 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
2138 | panic("bio: can't create integrity pool\n"); | |
2139 | ||
1da177e4 LT |
2140 | return 0; |
2141 | } | |
1da177e4 | 2142 | subsys_initcall(init_bio); |