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