]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
0fe23479 | 2 | * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk> |
1da177e4 LT |
3 | * |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of the GNU General Public License version 2 as | |
6 | * published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
11 | * GNU General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public Licens | |
14 | * along with this program; if not, write to the Free Software | |
15 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111- | |
16 | * | |
17 | */ | |
18 | #include <linux/mm.h> | |
19 | #include <linux/swap.h> | |
20 | #include <linux/bio.h> | |
21 | #include <linux/blkdev.h> | |
a27bb332 | 22 | #include <linux/uio.h> |
852c788f | 23 | #include <linux/iocontext.h> |
1da177e4 LT |
24 | #include <linux/slab.h> |
25 | #include <linux/init.h> | |
26 | #include <linux/kernel.h> | |
630d9c47 | 27 | #include <linux/export.h> |
1da177e4 LT |
28 | #include <linux/mempool.h> |
29 | #include <linux/workqueue.h> | |
852c788f | 30 | #include <linux/cgroup.h> |
1da177e4 | 31 | |
55782138 | 32 | #include <trace/events/block.h> |
9e234eea | 33 | #include "blk.h" |
0bfc2455 | 34 | |
392ddc32 JA |
35 | /* |
36 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
37 | * itself, to shrink a bio data allocation from two mempool calls to one | |
38 | */ | |
39 | #define BIO_INLINE_VECS 4 | |
40 | ||
1da177e4 LT |
41 | /* |
42 | * if you change this list, also change bvec_alloc or things will | |
43 | * break badly! cannot be bigger than what you can fit into an | |
44 | * unsigned short | |
45 | */ | |
1da177e4 | 46 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
ed996a52 | 47 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
1da177e4 LT |
48 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
49 | }; | |
50 | #undef BV | |
51 | ||
1da177e4 LT |
52 | /* |
53 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
54 | * IO code that does not need private memory pools. | |
55 | */ | |
51d654e1 | 56 | struct bio_set *fs_bio_set; |
3f86a82a | 57 | EXPORT_SYMBOL(fs_bio_set); |
1da177e4 | 58 | |
bb799ca0 JA |
59 | /* |
60 | * Our slab pool management | |
61 | */ | |
62 | struct bio_slab { | |
63 | struct kmem_cache *slab; | |
64 | unsigned int slab_ref; | |
65 | unsigned int slab_size; | |
66 | char name[8]; | |
67 | }; | |
68 | static DEFINE_MUTEX(bio_slab_lock); | |
69 | static struct bio_slab *bio_slabs; | |
70 | static unsigned int bio_slab_nr, bio_slab_max; | |
71 | ||
72 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
73 | { | |
74 | unsigned int sz = sizeof(struct bio) + extra_size; | |
75 | struct kmem_cache *slab = NULL; | |
389d7b26 | 76 | struct bio_slab *bslab, *new_bio_slabs; |
386bc35a | 77 | unsigned int new_bio_slab_max; |
bb799ca0 JA |
78 | unsigned int i, entry = -1; |
79 | ||
80 | mutex_lock(&bio_slab_lock); | |
81 | ||
82 | i = 0; | |
83 | while (i < bio_slab_nr) { | |
f06f135d | 84 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
85 | |
86 | if (!bslab->slab && entry == -1) | |
87 | entry = i; | |
88 | else if (bslab->slab_size == sz) { | |
89 | slab = bslab->slab; | |
90 | bslab->slab_ref++; | |
91 | break; | |
92 | } | |
93 | i++; | |
94 | } | |
95 | ||
96 | if (slab) | |
97 | goto out_unlock; | |
98 | ||
99 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
386bc35a | 100 | new_bio_slab_max = bio_slab_max << 1; |
389d7b26 | 101 | new_bio_slabs = krealloc(bio_slabs, |
386bc35a | 102 | new_bio_slab_max * sizeof(struct bio_slab), |
389d7b26 AK |
103 | GFP_KERNEL); |
104 | if (!new_bio_slabs) | |
bb799ca0 | 105 | goto out_unlock; |
386bc35a | 106 | bio_slab_max = new_bio_slab_max; |
389d7b26 | 107 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
108 | } |
109 | if (entry == -1) | |
110 | entry = bio_slab_nr++; | |
111 | ||
112 | bslab = &bio_slabs[entry]; | |
113 | ||
114 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
6a241483 MP |
115 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
116 | SLAB_HWCACHE_ALIGN, NULL); | |
bb799ca0 JA |
117 | if (!slab) |
118 | goto out_unlock; | |
119 | ||
bb799ca0 JA |
120 | bslab->slab = slab; |
121 | bslab->slab_ref = 1; | |
122 | bslab->slab_size = sz; | |
123 | out_unlock: | |
124 | mutex_unlock(&bio_slab_lock); | |
125 | return slab; | |
126 | } | |
127 | ||
128 | static void bio_put_slab(struct bio_set *bs) | |
129 | { | |
130 | struct bio_slab *bslab = NULL; | |
131 | unsigned int i; | |
132 | ||
133 | mutex_lock(&bio_slab_lock); | |
134 | ||
135 | for (i = 0; i < bio_slab_nr; i++) { | |
136 | if (bs->bio_slab == bio_slabs[i].slab) { | |
137 | bslab = &bio_slabs[i]; | |
138 | break; | |
139 | } | |
140 | } | |
141 | ||
142 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
143 | goto out; | |
144 | ||
145 | WARN_ON(!bslab->slab_ref); | |
146 | ||
147 | if (--bslab->slab_ref) | |
148 | goto out; | |
149 | ||
150 | kmem_cache_destroy(bslab->slab); | |
151 | bslab->slab = NULL; | |
152 | ||
153 | out: | |
154 | mutex_unlock(&bio_slab_lock); | |
155 | } | |
156 | ||
7ba1ba12 MP |
157 | unsigned int bvec_nr_vecs(unsigned short idx) |
158 | { | |
159 | return bvec_slabs[idx].nr_vecs; | |
160 | } | |
161 | ||
9f060e22 | 162 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
bb799ca0 | 163 | { |
ed996a52 CH |
164 | if (!idx) |
165 | return; | |
166 | idx--; | |
167 | ||
168 | BIO_BUG_ON(idx >= BVEC_POOL_NR); | |
bb799ca0 | 169 | |
ed996a52 | 170 | if (idx == BVEC_POOL_MAX) { |
9f060e22 | 171 | mempool_free(bv, pool); |
ed996a52 | 172 | } else { |
bb799ca0 JA |
173 | struct biovec_slab *bvs = bvec_slabs + idx; |
174 | ||
175 | kmem_cache_free(bvs->slab, bv); | |
176 | } | |
177 | } | |
178 | ||
9f060e22 KO |
179 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
180 | mempool_t *pool) | |
1da177e4 LT |
181 | { |
182 | struct bio_vec *bvl; | |
1da177e4 | 183 | |
7ff9345f JA |
184 | /* |
185 | * see comment near bvec_array define! | |
186 | */ | |
187 | switch (nr) { | |
188 | case 1: | |
189 | *idx = 0; | |
190 | break; | |
191 | case 2 ... 4: | |
192 | *idx = 1; | |
193 | break; | |
194 | case 5 ... 16: | |
195 | *idx = 2; | |
196 | break; | |
197 | case 17 ... 64: | |
198 | *idx = 3; | |
199 | break; | |
200 | case 65 ... 128: | |
201 | *idx = 4; | |
202 | break; | |
203 | case 129 ... BIO_MAX_PAGES: | |
204 | *idx = 5; | |
205 | break; | |
206 | default: | |
207 | return NULL; | |
208 | } | |
209 | ||
210 | /* | |
211 | * idx now points to the pool we want to allocate from. only the | |
212 | * 1-vec entry pool is mempool backed. | |
213 | */ | |
ed996a52 | 214 | if (*idx == BVEC_POOL_MAX) { |
7ff9345f | 215 | fallback: |
9f060e22 | 216 | bvl = mempool_alloc(pool, gfp_mask); |
7ff9345f JA |
217 | } else { |
218 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
d0164adc | 219 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
7ff9345f | 220 | |
0a0d96b0 | 221 | /* |
7ff9345f JA |
222 | * Make this allocation restricted and don't dump info on |
223 | * allocation failures, since we'll fallback to the mempool | |
224 | * in case of failure. | |
0a0d96b0 | 225 | */ |
7ff9345f | 226 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 227 | |
0a0d96b0 | 228 | /* |
d0164adc | 229 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
7ff9345f | 230 | * is set, retry with the 1-entry mempool |
0a0d96b0 | 231 | */ |
7ff9345f | 232 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
d0164adc | 233 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
ed996a52 | 234 | *idx = BVEC_POOL_MAX; |
7ff9345f JA |
235 | goto fallback; |
236 | } | |
237 | } | |
238 | ||
ed996a52 | 239 | (*idx)++; |
1da177e4 LT |
240 | return bvl; |
241 | } | |
242 | ||
9ae3b3f5 | 243 | void bio_uninit(struct bio *bio) |
1da177e4 | 244 | { |
4254bba1 | 245 | bio_disassociate_task(bio); |
1da177e4 | 246 | |
7ba1ba12 | 247 | if (bio_integrity(bio)) |
1e2a410f | 248 | bio_integrity_free(bio); |
4254bba1 | 249 | } |
9ae3b3f5 | 250 | EXPORT_SYMBOL(bio_uninit); |
7ba1ba12 | 251 | |
4254bba1 KO |
252 | static void bio_free(struct bio *bio) |
253 | { | |
254 | struct bio_set *bs = bio->bi_pool; | |
255 | void *p; | |
256 | ||
9ae3b3f5 | 257 | bio_uninit(bio); |
4254bba1 KO |
258 | |
259 | if (bs) { | |
ed996a52 | 260 | bvec_free(bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio)); |
4254bba1 KO |
261 | |
262 | /* | |
263 | * If we have front padding, adjust the bio pointer before freeing | |
264 | */ | |
265 | p = bio; | |
bb799ca0 JA |
266 | p -= bs->front_pad; |
267 | ||
4254bba1 KO |
268 | mempool_free(p, bs->bio_pool); |
269 | } else { | |
270 | /* Bio was allocated by bio_kmalloc() */ | |
271 | kfree(bio); | |
272 | } | |
3676347a PO |
273 | } |
274 | ||
9ae3b3f5 JA |
275 | /* |
276 | * Users of this function have their own bio allocation. Subsequently, | |
277 | * they must remember to pair any call to bio_init() with bio_uninit() | |
278 | * when IO has completed, or when the bio is released. | |
279 | */ | |
3a83f467 ML |
280 | void bio_init(struct bio *bio, struct bio_vec *table, |
281 | unsigned short max_vecs) | |
1da177e4 | 282 | { |
2b94de55 | 283 | memset(bio, 0, sizeof(*bio)); |
c4cf5261 | 284 | atomic_set(&bio->__bi_remaining, 1); |
dac56212 | 285 | atomic_set(&bio->__bi_cnt, 1); |
3a83f467 ML |
286 | |
287 | bio->bi_io_vec = table; | |
288 | bio->bi_max_vecs = max_vecs; | |
1da177e4 | 289 | } |
a112a71d | 290 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 291 | |
f44b48c7 KO |
292 | /** |
293 | * bio_reset - reinitialize a bio | |
294 | * @bio: bio to reset | |
295 | * | |
296 | * Description: | |
297 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
298 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
299 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
300 | * comment in struct bio. | |
301 | */ | |
302 | void bio_reset(struct bio *bio) | |
303 | { | |
304 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
305 | ||
9ae3b3f5 | 306 | bio_uninit(bio); |
f44b48c7 KO |
307 | |
308 | memset(bio, 0, BIO_RESET_BYTES); | |
4246a0b6 | 309 | bio->bi_flags = flags; |
c4cf5261 | 310 | atomic_set(&bio->__bi_remaining, 1); |
f44b48c7 KO |
311 | } |
312 | EXPORT_SYMBOL(bio_reset); | |
313 | ||
38f8baae | 314 | static struct bio *__bio_chain_endio(struct bio *bio) |
196d38bc | 315 | { |
4246a0b6 CH |
316 | struct bio *parent = bio->bi_private; |
317 | ||
4e4cbee9 CH |
318 | if (!parent->bi_status) |
319 | parent->bi_status = bio->bi_status; | |
196d38bc | 320 | bio_put(bio); |
38f8baae CH |
321 | return parent; |
322 | } | |
323 | ||
324 | static void bio_chain_endio(struct bio *bio) | |
325 | { | |
326 | bio_endio(__bio_chain_endio(bio)); | |
196d38bc KO |
327 | } |
328 | ||
329 | /** | |
330 | * bio_chain - chain bio completions | |
1051a902 RD |
331 | * @bio: the target bio |
332 | * @parent: the @bio's parent bio | |
196d38bc KO |
333 | * |
334 | * The caller won't have a bi_end_io called when @bio completes - instead, | |
335 | * @parent's bi_end_io won't be called until both @parent and @bio have | |
336 | * completed; the chained bio will also be freed when it completes. | |
337 | * | |
338 | * The caller must not set bi_private or bi_end_io in @bio. | |
339 | */ | |
340 | void bio_chain(struct bio *bio, struct bio *parent) | |
341 | { | |
342 | BUG_ON(bio->bi_private || bio->bi_end_io); | |
343 | ||
344 | bio->bi_private = parent; | |
345 | bio->bi_end_io = bio_chain_endio; | |
c4cf5261 | 346 | bio_inc_remaining(parent); |
196d38bc KO |
347 | } |
348 | EXPORT_SYMBOL(bio_chain); | |
349 | ||
df2cb6da KO |
350 | static void bio_alloc_rescue(struct work_struct *work) |
351 | { | |
352 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); | |
353 | struct bio *bio; | |
354 | ||
355 | while (1) { | |
356 | spin_lock(&bs->rescue_lock); | |
357 | bio = bio_list_pop(&bs->rescue_list); | |
358 | spin_unlock(&bs->rescue_lock); | |
359 | ||
360 | if (!bio) | |
361 | break; | |
362 | ||
363 | generic_make_request(bio); | |
364 | } | |
365 | } | |
366 | ||
367 | static void punt_bios_to_rescuer(struct bio_set *bs) | |
368 | { | |
369 | struct bio_list punt, nopunt; | |
370 | struct bio *bio; | |
371 | ||
47e0fb46 N |
372 | if (WARN_ON_ONCE(!bs->rescue_workqueue)) |
373 | return; | |
df2cb6da KO |
374 | /* |
375 | * In order to guarantee forward progress we must punt only bios that | |
376 | * were allocated from this bio_set; otherwise, if there was a bio on | |
377 | * there for a stacking driver higher up in the stack, processing it | |
378 | * could require allocating bios from this bio_set, and doing that from | |
379 | * our own rescuer would be bad. | |
380 | * | |
381 | * Since bio lists are singly linked, pop them all instead of trying to | |
382 | * remove from the middle of the list: | |
383 | */ | |
384 | ||
385 | bio_list_init(&punt); | |
386 | bio_list_init(&nopunt); | |
387 | ||
f5fe1b51 | 388 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
df2cb6da | 389 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
f5fe1b51 | 390 | current->bio_list[0] = nopunt; |
df2cb6da | 391 | |
f5fe1b51 N |
392 | bio_list_init(&nopunt); |
393 | while ((bio = bio_list_pop(¤t->bio_list[1]))) | |
394 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); | |
395 | current->bio_list[1] = nopunt; | |
df2cb6da KO |
396 | |
397 | spin_lock(&bs->rescue_lock); | |
398 | bio_list_merge(&bs->rescue_list, &punt); | |
399 | spin_unlock(&bs->rescue_lock); | |
400 | ||
401 | queue_work(bs->rescue_workqueue, &bs->rescue_work); | |
402 | } | |
403 | ||
1da177e4 LT |
404 | /** |
405 | * bio_alloc_bioset - allocate a bio for I/O | |
406 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
407 | * @nr_iovecs: number of iovecs to pre-allocate | |
db18efac | 408 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
409 | * |
410 | * Description: | |
3f86a82a KO |
411 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
412 | * backed by the @bs's mempool. | |
413 | * | |
d0164adc MG |
414 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
415 | * always be able to allocate a bio. This is due to the mempool guarantees. | |
416 | * To make this work, callers must never allocate more than 1 bio at a time | |
417 | * from this pool. Callers that need to allocate more than 1 bio must always | |
418 | * submit the previously allocated bio for IO before attempting to allocate | |
419 | * a new one. Failure to do so can cause deadlocks under memory pressure. | |
3f86a82a | 420 | * |
df2cb6da KO |
421 | * Note that when running under generic_make_request() (i.e. any block |
422 | * driver), bios are not submitted until after you return - see the code in | |
423 | * generic_make_request() that converts recursion into iteration, to prevent | |
424 | * stack overflows. | |
425 | * | |
426 | * This would normally mean allocating multiple bios under | |
427 | * generic_make_request() would be susceptible to deadlocks, but we have | |
428 | * deadlock avoidance code that resubmits any blocked bios from a rescuer | |
429 | * thread. | |
430 | * | |
431 | * However, we do not guarantee forward progress for allocations from other | |
432 | * mempools. Doing multiple allocations from the same mempool under | |
433 | * generic_make_request() should be avoided - instead, use bio_set's front_pad | |
434 | * for per bio allocations. | |
435 | * | |
3f86a82a KO |
436 | * RETURNS: |
437 | * Pointer to new bio on success, NULL on failure. | |
438 | */ | |
7a88fa19 DC |
439 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, unsigned int nr_iovecs, |
440 | struct bio_set *bs) | |
1da177e4 | 441 | { |
df2cb6da | 442 | gfp_t saved_gfp = gfp_mask; |
3f86a82a KO |
443 | unsigned front_pad; |
444 | unsigned inline_vecs; | |
34053979 | 445 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
446 | struct bio *bio; |
447 | void *p; | |
448 | ||
3f86a82a KO |
449 | if (!bs) { |
450 | if (nr_iovecs > UIO_MAXIOV) | |
451 | return NULL; | |
452 | ||
453 | p = kmalloc(sizeof(struct bio) + | |
454 | nr_iovecs * sizeof(struct bio_vec), | |
455 | gfp_mask); | |
456 | front_pad = 0; | |
457 | inline_vecs = nr_iovecs; | |
458 | } else { | |
d8f429e1 JN |
459 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
460 | if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0)) | |
461 | return NULL; | |
df2cb6da KO |
462 | /* |
463 | * generic_make_request() converts recursion to iteration; this | |
464 | * means if we're running beneath it, any bios we allocate and | |
465 | * submit will not be submitted (and thus freed) until after we | |
466 | * return. | |
467 | * | |
468 | * This exposes us to a potential deadlock if we allocate | |
469 | * multiple bios from the same bio_set() while running | |
470 | * underneath generic_make_request(). If we were to allocate | |
471 | * multiple bios (say a stacking block driver that was splitting | |
472 | * bios), we would deadlock if we exhausted the mempool's | |
473 | * reserve. | |
474 | * | |
475 | * We solve this, and guarantee forward progress, with a rescuer | |
476 | * workqueue per bio_set. If we go to allocate and there are | |
477 | * bios on current->bio_list, we first try the allocation | |
d0164adc MG |
478 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
479 | * bios we would be blocking to the rescuer workqueue before | |
480 | * we retry with the original gfp_flags. | |
df2cb6da KO |
481 | */ |
482 | ||
f5fe1b51 N |
483 | if (current->bio_list && |
484 | (!bio_list_empty(¤t->bio_list[0]) || | |
47e0fb46 N |
485 | !bio_list_empty(¤t->bio_list[1])) && |
486 | bs->rescue_workqueue) | |
d0164adc | 487 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
df2cb6da | 488 | |
3f86a82a | 489 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
df2cb6da KO |
490 | if (!p && gfp_mask != saved_gfp) { |
491 | punt_bios_to_rescuer(bs); | |
492 | gfp_mask = saved_gfp; | |
493 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
494 | } | |
495 | ||
3f86a82a KO |
496 | front_pad = bs->front_pad; |
497 | inline_vecs = BIO_INLINE_VECS; | |
498 | } | |
499 | ||
451a9ebf TH |
500 | if (unlikely(!p)) |
501 | return NULL; | |
1da177e4 | 502 | |
3f86a82a | 503 | bio = p + front_pad; |
3a83f467 | 504 | bio_init(bio, NULL, 0); |
34053979 | 505 | |
3f86a82a | 506 | if (nr_iovecs > inline_vecs) { |
ed996a52 CH |
507 | unsigned long idx = 0; |
508 | ||
9f060e22 | 509 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
510 | if (!bvl && gfp_mask != saved_gfp) { |
511 | punt_bios_to_rescuer(bs); | |
512 | gfp_mask = saved_gfp; | |
9f060e22 | 513 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
514 | } |
515 | ||
34053979 IM |
516 | if (unlikely(!bvl)) |
517 | goto err_free; | |
a38352e0 | 518 | |
ed996a52 | 519 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
3f86a82a KO |
520 | } else if (nr_iovecs) { |
521 | bvl = bio->bi_inline_vecs; | |
1da177e4 | 522 | } |
3f86a82a KO |
523 | |
524 | bio->bi_pool = bs; | |
34053979 | 525 | bio->bi_max_vecs = nr_iovecs; |
34053979 | 526 | bio->bi_io_vec = bvl; |
1da177e4 | 527 | return bio; |
34053979 IM |
528 | |
529 | err_free: | |
451a9ebf | 530 | mempool_free(p, bs->bio_pool); |
34053979 | 531 | return NULL; |
1da177e4 | 532 | } |
a112a71d | 533 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 534 | |
1da177e4 LT |
535 | void zero_fill_bio(struct bio *bio) |
536 | { | |
537 | unsigned long flags; | |
7988613b KO |
538 | struct bio_vec bv; |
539 | struct bvec_iter iter; | |
1da177e4 | 540 | |
7988613b KO |
541 | bio_for_each_segment(bv, bio, iter) { |
542 | char *data = bvec_kmap_irq(&bv, &flags); | |
543 | memset(data, 0, bv.bv_len); | |
544 | flush_dcache_page(bv.bv_page); | |
1da177e4 LT |
545 | bvec_kunmap_irq(data, &flags); |
546 | } | |
547 | } | |
548 | EXPORT_SYMBOL(zero_fill_bio); | |
549 | ||
550 | /** | |
551 | * bio_put - release a reference to a bio | |
552 | * @bio: bio to release reference to | |
553 | * | |
554 | * Description: | |
555 | * Put a reference to a &struct bio, either one you have gotten with | |
9b10f6a9 | 556 | * bio_alloc, bio_get or bio_clone_*. The last put of a bio will free it. |
1da177e4 LT |
557 | **/ |
558 | void bio_put(struct bio *bio) | |
559 | { | |
dac56212 | 560 | if (!bio_flagged(bio, BIO_REFFED)) |
4254bba1 | 561 | bio_free(bio); |
dac56212 JA |
562 | else { |
563 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); | |
564 | ||
565 | /* | |
566 | * last put frees it | |
567 | */ | |
568 | if (atomic_dec_and_test(&bio->__bi_cnt)) | |
569 | bio_free(bio); | |
570 | } | |
1da177e4 | 571 | } |
a112a71d | 572 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 573 | |
165125e1 | 574 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
575 | { |
576 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
577 | blk_recount_segments(q, bio); | |
578 | ||
579 | return bio->bi_phys_segments; | |
580 | } | |
a112a71d | 581 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 582 | |
59d276fe KO |
583 | /** |
584 | * __bio_clone_fast - clone a bio that shares the original bio's biovec | |
585 | * @bio: destination bio | |
586 | * @bio_src: bio to clone | |
587 | * | |
588 | * Clone a &bio. Caller will own the returned bio, but not | |
589 | * the actual data it points to. Reference count of returned | |
590 | * bio will be one. | |
591 | * | |
592 | * Caller must ensure that @bio_src is not freed before @bio. | |
593 | */ | |
594 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) | |
595 | { | |
ed996a52 | 596 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
59d276fe KO |
597 | |
598 | /* | |
599 | * most users will be overriding ->bi_bdev with a new target, | |
600 | * so we don't set nor calculate new physical/hw segment counts here | |
601 | */ | |
602 | bio->bi_bdev = bio_src->bi_bdev; | |
b7c44ed9 | 603 | bio_set_flag(bio, BIO_CLONED); |
1eff9d32 | 604 | bio->bi_opf = bio_src->bi_opf; |
cb6934f8 | 605 | bio->bi_write_hint = bio_src->bi_write_hint; |
59d276fe KO |
606 | bio->bi_iter = bio_src->bi_iter; |
607 | bio->bi_io_vec = bio_src->bi_io_vec; | |
20bd723e PV |
608 | |
609 | bio_clone_blkcg_association(bio, bio_src); | |
59d276fe KO |
610 | } |
611 | EXPORT_SYMBOL(__bio_clone_fast); | |
612 | ||
613 | /** | |
614 | * bio_clone_fast - clone a bio that shares the original bio's biovec | |
615 | * @bio: bio to clone | |
616 | * @gfp_mask: allocation priority | |
617 | * @bs: bio_set to allocate from | |
618 | * | |
619 | * Like __bio_clone_fast, only also allocates the returned bio | |
620 | */ | |
621 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) | |
622 | { | |
623 | struct bio *b; | |
624 | ||
625 | b = bio_alloc_bioset(gfp_mask, 0, bs); | |
626 | if (!b) | |
627 | return NULL; | |
628 | ||
629 | __bio_clone_fast(b, bio); | |
630 | ||
631 | if (bio_integrity(bio)) { | |
632 | int ret; | |
633 | ||
634 | ret = bio_integrity_clone(b, bio, gfp_mask); | |
635 | ||
636 | if (ret < 0) { | |
637 | bio_put(b); | |
638 | return NULL; | |
639 | } | |
640 | } | |
641 | ||
642 | return b; | |
643 | } | |
644 | EXPORT_SYMBOL(bio_clone_fast); | |
645 | ||
f4595875 SL |
646 | /** |
647 | * bio_clone_bioset - clone a bio | |
648 | * @bio_src: bio to clone | |
649 | * @gfp_mask: allocation priority | |
650 | * @bs: bio_set to allocate from | |
651 | * | |
652 | * Clone bio. Caller will own the returned bio, but not the actual data it | |
653 | * points to. Reference count of returned bio will be one. | |
654 | */ | |
655 | struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask, | |
656 | struct bio_set *bs) | |
1da177e4 | 657 | { |
bdb53207 KO |
658 | struct bvec_iter iter; |
659 | struct bio_vec bv; | |
660 | struct bio *bio; | |
1da177e4 | 661 | |
bdb53207 KO |
662 | /* |
663 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from | |
664 | * bio_src->bi_io_vec to bio->bi_io_vec. | |
665 | * | |
666 | * We can't do that anymore, because: | |
667 | * | |
668 | * - The point of cloning the biovec is to produce a bio with a biovec | |
669 | * the caller can modify: bi_idx and bi_bvec_done should be 0. | |
670 | * | |
671 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if | |
672 | * we tried to clone the whole thing bio_alloc_bioset() would fail. | |
673 | * But the clone should succeed as long as the number of biovecs we | |
674 | * actually need to allocate is fewer than BIO_MAX_PAGES. | |
675 | * | |
676 | * - Lastly, bi_vcnt should not be looked at or relied upon by code | |
677 | * that does not own the bio - reason being drivers don't use it for | |
678 | * iterating over the biovec anymore, so expecting it to be kept up | |
679 | * to date (i.e. for clones that share the parent biovec) is just | |
680 | * asking for trouble and would force extra work on | |
681 | * __bio_clone_fast() anyways. | |
682 | */ | |
683 | ||
f4595875 | 684 | bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs); |
bdb53207 | 685 | if (!bio) |
7ba1ba12 | 686 | return NULL; |
bdb53207 | 687 | bio->bi_bdev = bio_src->bi_bdev; |
1eff9d32 | 688 | bio->bi_opf = bio_src->bi_opf; |
cb6934f8 | 689 | bio->bi_write_hint = bio_src->bi_write_hint; |
bdb53207 KO |
690 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; |
691 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; | |
7ba1ba12 | 692 | |
7afafc8a AH |
693 | switch (bio_op(bio)) { |
694 | case REQ_OP_DISCARD: | |
695 | case REQ_OP_SECURE_ERASE: | |
a6f0788e | 696 | case REQ_OP_WRITE_ZEROES: |
7afafc8a AH |
697 | break; |
698 | case REQ_OP_WRITE_SAME: | |
8423ae3d | 699 | bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0]; |
7afafc8a AH |
700 | break; |
701 | default: | |
f4595875 | 702 | bio_for_each_segment(bv, bio_src, iter) |
7afafc8a AH |
703 | bio->bi_io_vec[bio->bi_vcnt++] = bv; |
704 | break; | |
8423ae3d KO |
705 | } |
706 | ||
bdb53207 KO |
707 | if (bio_integrity(bio_src)) { |
708 | int ret; | |
7ba1ba12 | 709 | |
bdb53207 | 710 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
059ea331 | 711 | if (ret < 0) { |
bdb53207 | 712 | bio_put(bio); |
7ba1ba12 | 713 | return NULL; |
059ea331 | 714 | } |
3676347a | 715 | } |
1da177e4 | 716 | |
20bd723e PV |
717 | bio_clone_blkcg_association(bio, bio_src); |
718 | ||
bdb53207 | 719 | return bio; |
1da177e4 | 720 | } |
bf800ef1 | 721 | EXPORT_SYMBOL(bio_clone_bioset); |
1da177e4 LT |
722 | |
723 | /** | |
c66a14d0 KO |
724 | * bio_add_pc_page - attempt to add page to bio |
725 | * @q: the target queue | |
726 | * @bio: destination bio | |
727 | * @page: page to add | |
728 | * @len: vec entry length | |
729 | * @offset: vec entry offset | |
1da177e4 | 730 | * |
c66a14d0 KO |
731 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
732 | * number of reasons, such as the bio being full or target block device | |
733 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
734 | * so it is always possible to add a single page to an empty bio. | |
735 | * | |
736 | * This should only be used by REQ_PC bios. | |
1da177e4 | 737 | */ |
c66a14d0 KO |
738 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
739 | *page, unsigned int len, unsigned int offset) | |
1da177e4 LT |
740 | { |
741 | int retried_segments = 0; | |
742 | struct bio_vec *bvec; | |
743 | ||
744 | /* | |
745 | * cloned bio must not modify vec list | |
746 | */ | |
747 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
748 | return 0; | |
749 | ||
c66a14d0 | 750 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
751 | return 0; |
752 | ||
80cfd548 JA |
753 | /* |
754 | * For filesystems with a blocksize smaller than the pagesize | |
755 | * we will often be called with the same page as last time and | |
756 | * a consecutive offset. Optimize this special case. | |
757 | */ | |
758 | if (bio->bi_vcnt > 0) { | |
759 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
760 | ||
761 | if (page == prev->bv_page && | |
762 | offset == prev->bv_offset + prev->bv_len) { | |
763 | prev->bv_len += len; | |
fcbf6a08 | 764 | bio->bi_iter.bi_size += len; |
80cfd548 JA |
765 | goto done; |
766 | } | |
66cb45aa JA |
767 | |
768 | /* | |
769 | * If the queue doesn't support SG gaps and adding this | |
770 | * offset would create a gap, disallow it. | |
771 | */ | |
03100aad | 772 | if (bvec_gap_to_prev(q, prev, offset)) |
66cb45aa | 773 | return 0; |
80cfd548 JA |
774 | } |
775 | ||
776 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
777 | return 0; |
778 | ||
779 | /* | |
fcbf6a08 ML |
780 | * setup the new entry, we might clear it again later if we |
781 | * cannot add the page | |
782 | */ | |
783 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
784 | bvec->bv_page = page; | |
785 | bvec->bv_len = len; | |
786 | bvec->bv_offset = offset; | |
787 | bio->bi_vcnt++; | |
788 | bio->bi_phys_segments++; | |
789 | bio->bi_iter.bi_size += len; | |
790 | ||
791 | /* | |
792 | * Perform a recount if the number of segments is greater | |
793 | * than queue_max_segments(q). | |
1da177e4 LT |
794 | */ |
795 | ||
fcbf6a08 | 796 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
1da177e4 LT |
797 | |
798 | if (retried_segments) | |
fcbf6a08 | 799 | goto failed; |
1da177e4 LT |
800 | |
801 | retried_segments = 1; | |
802 | blk_recount_segments(q, bio); | |
803 | } | |
804 | ||
1da177e4 | 805 | /* If we may be able to merge these biovecs, force a recount */ |
fcbf6a08 | 806 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
b7c44ed9 | 807 | bio_clear_flag(bio, BIO_SEG_VALID); |
1da177e4 | 808 | |
80cfd548 | 809 | done: |
1da177e4 | 810 | return len; |
fcbf6a08 ML |
811 | |
812 | failed: | |
813 | bvec->bv_page = NULL; | |
814 | bvec->bv_len = 0; | |
815 | bvec->bv_offset = 0; | |
816 | bio->bi_vcnt--; | |
817 | bio->bi_iter.bi_size -= len; | |
818 | blk_recount_segments(q, bio); | |
819 | return 0; | |
1da177e4 | 820 | } |
a112a71d | 821 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 822 | |
1da177e4 LT |
823 | /** |
824 | * bio_add_page - attempt to add page to bio | |
825 | * @bio: destination bio | |
826 | * @page: page to add | |
827 | * @len: vec entry length | |
828 | * @offset: vec entry offset | |
829 | * | |
c66a14d0 KO |
830 | * Attempt to add a page to the bio_vec maplist. This will only fail |
831 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. | |
1da177e4 | 832 | */ |
c66a14d0 KO |
833 | int bio_add_page(struct bio *bio, struct page *page, |
834 | unsigned int len, unsigned int offset) | |
1da177e4 | 835 | { |
c66a14d0 KO |
836 | struct bio_vec *bv; |
837 | ||
838 | /* | |
839 | * cloned bio must not modify vec list | |
840 | */ | |
841 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) | |
842 | return 0; | |
762380ad | 843 | |
c66a14d0 KO |
844 | /* |
845 | * For filesystems with a blocksize smaller than the pagesize | |
846 | * we will often be called with the same page as last time and | |
847 | * a consecutive offset. Optimize this special case. | |
848 | */ | |
849 | if (bio->bi_vcnt > 0) { | |
850 | bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
58a4915a | 851 | |
c66a14d0 KO |
852 | if (page == bv->bv_page && |
853 | offset == bv->bv_offset + bv->bv_len) { | |
854 | bv->bv_len += len; | |
855 | goto done; | |
856 | } | |
857 | } | |
858 | ||
859 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
860 | return 0; | |
861 | ||
862 | bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
863 | bv->bv_page = page; | |
864 | bv->bv_len = len; | |
865 | bv->bv_offset = offset; | |
866 | ||
867 | bio->bi_vcnt++; | |
868 | done: | |
869 | bio->bi_iter.bi_size += len; | |
870 | return len; | |
1da177e4 | 871 | } |
a112a71d | 872 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 873 | |
2cefe4db KO |
874 | /** |
875 | * bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio | |
876 | * @bio: bio to add pages to | |
877 | * @iter: iov iterator describing the region to be mapped | |
878 | * | |
879 | * Pins as many pages from *iter and appends them to @bio's bvec array. The | |
880 | * pages will have to be released using put_page() when done. | |
881 | */ | |
882 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) | |
883 | { | |
884 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; | |
885 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; | |
886 | struct page **pages = (struct page **)bv; | |
887 | size_t offset, diff; | |
888 | ssize_t size; | |
889 | ||
890 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); | |
891 | if (unlikely(size <= 0)) | |
892 | return size ? size : -EFAULT; | |
893 | nr_pages = (size + offset + PAGE_SIZE - 1) / PAGE_SIZE; | |
894 | ||
895 | /* | |
896 | * Deep magic below: We need to walk the pinned pages backwards | |
897 | * because we are abusing the space allocated for the bio_vecs | |
898 | * for the page array. Because the bio_vecs are larger than the | |
899 | * page pointers by definition this will always work. But it also | |
900 | * means we can't use bio_add_page, so any changes to it's semantics | |
901 | * need to be reflected here as well. | |
902 | */ | |
903 | bio->bi_iter.bi_size += size; | |
904 | bio->bi_vcnt += nr_pages; | |
905 | ||
906 | diff = (nr_pages * PAGE_SIZE - offset) - size; | |
907 | while (nr_pages--) { | |
908 | bv[nr_pages].bv_page = pages[nr_pages]; | |
909 | bv[nr_pages].bv_len = PAGE_SIZE; | |
910 | bv[nr_pages].bv_offset = 0; | |
911 | } | |
912 | ||
913 | bv[0].bv_offset += offset; | |
914 | bv[0].bv_len -= offset; | |
915 | if (diff) | |
916 | bv[bio->bi_vcnt - 1].bv_len -= diff; | |
917 | ||
918 | iov_iter_advance(iter, size); | |
919 | return 0; | |
920 | } | |
921 | EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages); | |
922 | ||
9e882242 KO |
923 | struct submit_bio_ret { |
924 | struct completion event; | |
925 | int error; | |
926 | }; | |
927 | ||
4246a0b6 | 928 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 KO |
929 | { |
930 | struct submit_bio_ret *ret = bio->bi_private; | |
931 | ||
4e4cbee9 | 932 | ret->error = blk_status_to_errno(bio->bi_status); |
9e882242 KO |
933 | complete(&ret->event); |
934 | } | |
935 | ||
936 | /** | |
937 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
938 | * @bio: The &struct bio which describes the I/O |
939 | * | |
940 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
941 | * bio_endio() on failure. | |
942 | */ | |
4e49ea4a | 943 | int submit_bio_wait(struct bio *bio) |
9e882242 KO |
944 | { |
945 | struct submit_bio_ret ret; | |
946 | ||
9e882242 KO |
947 | init_completion(&ret.event); |
948 | bio->bi_private = &ret; | |
949 | bio->bi_end_io = submit_bio_wait_endio; | |
1eff9d32 | 950 | bio->bi_opf |= REQ_SYNC; |
4e49ea4a | 951 | submit_bio(bio); |
d57d6115 | 952 | wait_for_completion_io(&ret.event); |
9e882242 KO |
953 | |
954 | return ret.error; | |
955 | } | |
956 | EXPORT_SYMBOL(submit_bio_wait); | |
957 | ||
054bdf64 KO |
958 | /** |
959 | * bio_advance - increment/complete a bio by some number of bytes | |
960 | * @bio: bio to advance | |
961 | * @bytes: number of bytes to complete | |
962 | * | |
963 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
964 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
965 | * be updated on the last bvec as well. | |
966 | * | |
967 | * @bio will then represent the remaining, uncompleted portion of the io. | |
968 | */ | |
969 | void bio_advance(struct bio *bio, unsigned bytes) | |
970 | { | |
971 | if (bio_integrity(bio)) | |
972 | bio_integrity_advance(bio, bytes); | |
973 | ||
4550dd6c | 974 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
975 | } |
976 | EXPORT_SYMBOL(bio_advance); | |
977 | ||
a0787606 KO |
978 | /** |
979 | * bio_alloc_pages - allocates a single page for each bvec in a bio | |
980 | * @bio: bio to allocate pages for | |
981 | * @gfp_mask: flags for allocation | |
982 | * | |
983 | * Allocates pages up to @bio->bi_vcnt. | |
984 | * | |
985 | * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are | |
986 | * freed. | |
987 | */ | |
988 | int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask) | |
989 | { | |
990 | int i; | |
991 | struct bio_vec *bv; | |
992 | ||
993 | bio_for_each_segment_all(bv, bio, i) { | |
994 | bv->bv_page = alloc_page(gfp_mask); | |
995 | if (!bv->bv_page) { | |
996 | while (--bv >= bio->bi_io_vec) | |
997 | __free_page(bv->bv_page); | |
998 | return -ENOMEM; | |
999 | } | |
1000 | } | |
1001 | ||
1002 | return 0; | |
1003 | } | |
1004 | EXPORT_SYMBOL(bio_alloc_pages); | |
1005 | ||
16ac3d63 KO |
1006 | /** |
1007 | * bio_copy_data - copy contents of data buffers from one chain of bios to | |
1008 | * another | |
1009 | * @src: source bio list | |
1010 | * @dst: destination bio list | |
1011 | * | |
1012 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats | |
1013 | * @src and @dst as linked lists of bios. | |
1014 | * | |
1015 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
1016 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
1017 | */ | |
1018 | void bio_copy_data(struct bio *dst, struct bio *src) | |
1019 | { | |
1cb9dda4 KO |
1020 | struct bvec_iter src_iter, dst_iter; |
1021 | struct bio_vec src_bv, dst_bv; | |
16ac3d63 | 1022 | void *src_p, *dst_p; |
1cb9dda4 | 1023 | unsigned bytes; |
16ac3d63 | 1024 | |
1cb9dda4 KO |
1025 | src_iter = src->bi_iter; |
1026 | dst_iter = dst->bi_iter; | |
16ac3d63 KO |
1027 | |
1028 | while (1) { | |
1cb9dda4 KO |
1029 | if (!src_iter.bi_size) { |
1030 | src = src->bi_next; | |
1031 | if (!src) | |
1032 | break; | |
16ac3d63 | 1033 | |
1cb9dda4 | 1034 | src_iter = src->bi_iter; |
16ac3d63 KO |
1035 | } |
1036 | ||
1cb9dda4 KO |
1037 | if (!dst_iter.bi_size) { |
1038 | dst = dst->bi_next; | |
1039 | if (!dst) | |
1040 | break; | |
16ac3d63 | 1041 | |
1cb9dda4 | 1042 | dst_iter = dst->bi_iter; |
16ac3d63 KO |
1043 | } |
1044 | ||
1cb9dda4 KO |
1045 | src_bv = bio_iter_iovec(src, src_iter); |
1046 | dst_bv = bio_iter_iovec(dst, dst_iter); | |
1047 | ||
1048 | bytes = min(src_bv.bv_len, dst_bv.bv_len); | |
16ac3d63 | 1049 | |
1cb9dda4 KO |
1050 | src_p = kmap_atomic(src_bv.bv_page); |
1051 | dst_p = kmap_atomic(dst_bv.bv_page); | |
16ac3d63 | 1052 | |
1cb9dda4 KO |
1053 | memcpy(dst_p + dst_bv.bv_offset, |
1054 | src_p + src_bv.bv_offset, | |
16ac3d63 KO |
1055 | bytes); |
1056 | ||
1057 | kunmap_atomic(dst_p); | |
1058 | kunmap_atomic(src_p); | |
1059 | ||
1cb9dda4 KO |
1060 | bio_advance_iter(src, &src_iter, bytes); |
1061 | bio_advance_iter(dst, &dst_iter, bytes); | |
16ac3d63 KO |
1062 | } |
1063 | } | |
1064 | EXPORT_SYMBOL(bio_copy_data); | |
1065 | ||
1da177e4 | 1066 | struct bio_map_data { |
152e283f | 1067 | int is_our_pages; |
26e49cfc KO |
1068 | struct iov_iter iter; |
1069 | struct iovec iov[]; | |
1da177e4 LT |
1070 | }; |
1071 | ||
7410b3c6 | 1072 | static struct bio_map_data *bio_alloc_map_data(unsigned int iov_count, |
76029ff3 | 1073 | gfp_t gfp_mask) |
1da177e4 | 1074 | { |
f3f63c1c JA |
1075 | if (iov_count > UIO_MAXIOV) |
1076 | return NULL; | |
1da177e4 | 1077 | |
c8db4448 | 1078 | return kmalloc(sizeof(struct bio_map_data) + |
26e49cfc | 1079 | sizeof(struct iovec) * iov_count, gfp_mask); |
1da177e4 LT |
1080 | } |
1081 | ||
9124d3fe DP |
1082 | /** |
1083 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1084 | * @bio: The &struct bio which describes the I/O as destination | |
1085 | * @iter: iov_iter as source | |
1086 | * | |
1087 | * Copy all pages from iov_iter to bio. | |
1088 | * Returns 0 on success, or error on failure. | |
1089 | */ | |
1090 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter iter) | |
c5dec1c3 | 1091 | { |
9124d3fe | 1092 | int i; |
c5dec1c3 | 1093 | struct bio_vec *bvec; |
c5dec1c3 | 1094 | |
d74c6d51 | 1095 | bio_for_each_segment_all(bvec, bio, i) { |
9124d3fe | 1096 | ssize_t ret; |
c5dec1c3 | 1097 | |
9124d3fe DP |
1098 | ret = copy_page_from_iter(bvec->bv_page, |
1099 | bvec->bv_offset, | |
1100 | bvec->bv_len, | |
1101 | &iter); | |
1102 | ||
1103 | if (!iov_iter_count(&iter)) | |
1104 | break; | |
1105 | ||
1106 | if (ret < bvec->bv_len) | |
1107 | return -EFAULT; | |
c5dec1c3 FT |
1108 | } |
1109 | ||
9124d3fe DP |
1110 | return 0; |
1111 | } | |
1112 | ||
1113 | /** | |
1114 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1115 | * @bio: The &struct bio which describes the I/O as source | |
1116 | * @iter: iov_iter as destination | |
1117 | * | |
1118 | * Copy all pages from bio to iov_iter. | |
1119 | * Returns 0 on success, or error on failure. | |
1120 | */ | |
1121 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1122 | { | |
1123 | int i; | |
1124 | struct bio_vec *bvec; | |
1125 | ||
1126 | bio_for_each_segment_all(bvec, bio, i) { | |
1127 | ssize_t ret; | |
1128 | ||
1129 | ret = copy_page_to_iter(bvec->bv_page, | |
1130 | bvec->bv_offset, | |
1131 | bvec->bv_len, | |
1132 | &iter); | |
1133 | ||
1134 | if (!iov_iter_count(&iter)) | |
1135 | break; | |
1136 | ||
1137 | if (ret < bvec->bv_len) | |
1138 | return -EFAULT; | |
1139 | } | |
1140 | ||
1141 | return 0; | |
c5dec1c3 FT |
1142 | } |
1143 | ||
491221f8 | 1144 | void bio_free_pages(struct bio *bio) |
1dfa0f68 CH |
1145 | { |
1146 | struct bio_vec *bvec; | |
1147 | int i; | |
1148 | ||
1149 | bio_for_each_segment_all(bvec, bio, i) | |
1150 | __free_page(bvec->bv_page); | |
1151 | } | |
491221f8 | 1152 | EXPORT_SYMBOL(bio_free_pages); |
1dfa0f68 | 1153 | |
1da177e4 LT |
1154 | /** |
1155 | * bio_uncopy_user - finish previously mapped bio | |
1156 | * @bio: bio being terminated | |
1157 | * | |
ddad8dd0 | 1158 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1159 | * to user space in case of a read. |
1160 | */ | |
1161 | int bio_uncopy_user(struct bio *bio) | |
1162 | { | |
1163 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1164 | int ret = 0; |
1da177e4 | 1165 | |
35dc2483 RD |
1166 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1167 | /* | |
1168 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1169 | * don't copy into a random user address space, just free |
1170 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1171 | */ |
2d99b55d HR |
1172 | if (!current->mm) |
1173 | ret = -EINTR; | |
1174 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1175 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1176 | if (bmd->is_our_pages) |
1177 | bio_free_pages(bio); | |
35dc2483 | 1178 | } |
c8db4448 | 1179 | kfree(bmd); |
1da177e4 LT |
1180 | bio_put(bio); |
1181 | return ret; | |
1182 | } | |
1183 | ||
1184 | /** | |
c5dec1c3 | 1185 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1186 | * @q: destination block queue |
1187 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1188 | * @iter: iovec iterator | |
1189 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1190 | * |
1191 | * Prepares and returns a bio for indirect user io, bouncing data | |
1192 | * to/from kernel pages as necessary. Must be paired with | |
1193 | * call bio_uncopy_user() on io completion. | |
1194 | */ | |
152e283f FT |
1195 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1196 | struct rq_map_data *map_data, | |
26e49cfc KO |
1197 | const struct iov_iter *iter, |
1198 | gfp_t gfp_mask) | |
1da177e4 | 1199 | { |
1da177e4 | 1200 | struct bio_map_data *bmd; |
1da177e4 LT |
1201 | struct page *page; |
1202 | struct bio *bio; | |
1203 | int i, ret; | |
c5dec1c3 | 1204 | int nr_pages = 0; |
26e49cfc | 1205 | unsigned int len = iter->count; |
bd5cecea | 1206 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1207 | |
26e49cfc | 1208 | for (i = 0; i < iter->nr_segs; i++) { |
c5dec1c3 FT |
1209 | unsigned long uaddr; |
1210 | unsigned long end; | |
1211 | unsigned long start; | |
1212 | ||
26e49cfc KO |
1213 | uaddr = (unsigned long) iter->iov[i].iov_base; |
1214 | end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1) | |
1215 | >> PAGE_SHIFT; | |
c5dec1c3 FT |
1216 | start = uaddr >> PAGE_SHIFT; |
1217 | ||
cb4644ca JA |
1218 | /* |
1219 | * Overflow, abort | |
1220 | */ | |
1221 | if (end < start) | |
1222 | return ERR_PTR(-EINVAL); | |
1223 | ||
c5dec1c3 | 1224 | nr_pages += end - start; |
c5dec1c3 FT |
1225 | } |
1226 | ||
69838727 FT |
1227 | if (offset) |
1228 | nr_pages++; | |
1229 | ||
26e49cfc | 1230 | bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask); |
1da177e4 LT |
1231 | if (!bmd) |
1232 | return ERR_PTR(-ENOMEM); | |
1233 | ||
26e49cfc KO |
1234 | /* |
1235 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1236 | * The caller provided iov might point to an on-stack or otherwise | |
1237 | * shortlived one. | |
1238 | */ | |
1239 | bmd->is_our_pages = map_data ? 0 : 1; | |
1240 | memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs); | |
1241 | iov_iter_init(&bmd->iter, iter->type, bmd->iov, | |
1242 | iter->nr_segs, iter->count); | |
1243 | ||
1da177e4 | 1244 | ret = -ENOMEM; |
a9e9dc24 | 1245 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1246 | if (!bio) |
1247 | goto out_bmd; | |
1248 | ||
1da177e4 | 1249 | ret = 0; |
56c451f4 FT |
1250 | |
1251 | if (map_data) { | |
e623ddb4 | 1252 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1253 | i = map_data->offset / PAGE_SIZE; |
1254 | } | |
1da177e4 | 1255 | while (len) { |
e623ddb4 | 1256 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1257 | |
56c451f4 FT |
1258 | bytes -= offset; |
1259 | ||
1da177e4 LT |
1260 | if (bytes > len) |
1261 | bytes = len; | |
1262 | ||
152e283f | 1263 | if (map_data) { |
e623ddb4 | 1264 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1265 | ret = -ENOMEM; |
1266 | break; | |
1267 | } | |
e623ddb4 FT |
1268 | |
1269 | page = map_data->pages[i / nr_pages]; | |
1270 | page += (i % nr_pages); | |
1271 | ||
1272 | i++; | |
1273 | } else { | |
152e283f | 1274 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1275 | if (!page) { |
1276 | ret = -ENOMEM; | |
1277 | break; | |
1278 | } | |
1da177e4 LT |
1279 | } |
1280 | ||
56c451f4 | 1281 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 1282 | break; |
1da177e4 LT |
1283 | |
1284 | len -= bytes; | |
56c451f4 | 1285 | offset = 0; |
1da177e4 LT |
1286 | } |
1287 | ||
1288 | if (ret) | |
1289 | goto cleanup; | |
1290 | ||
1291 | /* | |
1292 | * success | |
1293 | */ | |
26e49cfc | 1294 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1295 | (map_data && map_data->from_user)) { |
9124d3fe | 1296 | ret = bio_copy_from_iter(bio, *iter); |
c5dec1c3 FT |
1297 | if (ret) |
1298 | goto cleanup; | |
1da177e4 LT |
1299 | } |
1300 | ||
26e49cfc | 1301 | bio->bi_private = bmd; |
1da177e4 LT |
1302 | return bio; |
1303 | cleanup: | |
152e283f | 1304 | if (!map_data) |
1dfa0f68 | 1305 | bio_free_pages(bio); |
1da177e4 LT |
1306 | bio_put(bio); |
1307 | out_bmd: | |
c8db4448 | 1308 | kfree(bmd); |
1da177e4 LT |
1309 | return ERR_PTR(ret); |
1310 | } | |
1311 | ||
37f19e57 CH |
1312 | /** |
1313 | * bio_map_user_iov - map user iovec into bio | |
1314 | * @q: the struct request_queue for the bio | |
1315 | * @iter: iovec iterator | |
1316 | * @gfp_mask: memory allocation flags | |
1317 | * | |
1318 | * Map the user space address into a bio suitable for io to a block | |
1319 | * device. Returns an error pointer in case of error. | |
1320 | */ | |
1321 | struct bio *bio_map_user_iov(struct request_queue *q, | |
1322 | const struct iov_iter *iter, | |
1323 | gfp_t gfp_mask) | |
1da177e4 | 1324 | { |
26e49cfc | 1325 | int j; |
f1970baf | 1326 | int nr_pages = 0; |
1da177e4 LT |
1327 | struct page **pages; |
1328 | struct bio *bio; | |
f1970baf JB |
1329 | int cur_page = 0; |
1330 | int ret, offset; | |
26e49cfc KO |
1331 | struct iov_iter i; |
1332 | struct iovec iov; | |
1da177e4 | 1333 | |
26e49cfc KO |
1334 | iov_for_each(iov, i, *iter) { |
1335 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1336 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1337 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1338 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1339 | ||
cb4644ca JA |
1340 | /* |
1341 | * Overflow, abort | |
1342 | */ | |
1343 | if (end < start) | |
1344 | return ERR_PTR(-EINVAL); | |
1345 | ||
f1970baf JB |
1346 | nr_pages += end - start; |
1347 | /* | |
a441b0d0 | 1348 | * buffer must be aligned to at least logical block size for now |
f1970baf | 1349 | */ |
ad2d7225 | 1350 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
1351 | return ERR_PTR(-EINVAL); |
1352 | } | |
1353 | ||
1354 | if (!nr_pages) | |
1da177e4 LT |
1355 | return ERR_PTR(-EINVAL); |
1356 | ||
a9e9dc24 | 1357 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1358 | if (!bio) |
1359 | return ERR_PTR(-ENOMEM); | |
1360 | ||
1361 | ret = -ENOMEM; | |
a3bce90e | 1362 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
1363 | if (!pages) |
1364 | goto out; | |
1365 | ||
26e49cfc KO |
1366 | iov_for_each(iov, i, *iter) { |
1367 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1368 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1369 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1370 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1371 | const int local_nr_pages = end - start; | |
1372 | const int page_limit = cur_page + local_nr_pages; | |
cb4644ca | 1373 | |
f5dd33c4 | 1374 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
26e49cfc KO |
1375 | (iter->type & WRITE) != WRITE, |
1376 | &pages[cur_page]); | |
99172157 JA |
1377 | if (ret < local_nr_pages) { |
1378 | ret = -EFAULT; | |
f1970baf | 1379 | goto out_unmap; |
99172157 | 1380 | } |
f1970baf | 1381 | |
bd5cecea | 1382 | offset = offset_in_page(uaddr); |
f1970baf JB |
1383 | for (j = cur_page; j < page_limit; j++) { |
1384 | unsigned int bytes = PAGE_SIZE - offset; | |
1385 | ||
1386 | if (len <= 0) | |
1387 | break; | |
1388 | ||
1389 | if (bytes > len) | |
1390 | bytes = len; | |
1391 | ||
1392 | /* | |
1393 | * sorry... | |
1394 | */ | |
defd94b7 MC |
1395 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1396 | bytes) | |
f1970baf JB |
1397 | break; |
1398 | ||
1399 | len -= bytes; | |
1400 | offset = 0; | |
1401 | } | |
1da177e4 | 1402 | |
f1970baf | 1403 | cur_page = j; |
1da177e4 | 1404 | /* |
f1970baf | 1405 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1406 | */ |
f1970baf | 1407 | while (j < page_limit) |
09cbfeaf | 1408 | put_page(pages[j++]); |
1da177e4 LT |
1409 | } |
1410 | ||
1da177e4 LT |
1411 | kfree(pages); |
1412 | ||
b7c44ed9 | 1413 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1414 | |
1415 | /* | |
5fad1b64 | 1416 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
37f19e57 CH |
1417 | * it would normally disappear when its bi_end_io is run. |
1418 | * however, we need it for the unmap, so grab an extra | |
1419 | * reference to it | |
1420 | */ | |
1421 | bio_get(bio); | |
1da177e4 | 1422 | return bio; |
f1970baf JB |
1423 | |
1424 | out_unmap: | |
26e49cfc KO |
1425 | for (j = 0; j < nr_pages; j++) { |
1426 | if (!pages[j]) | |
f1970baf | 1427 | break; |
09cbfeaf | 1428 | put_page(pages[j]); |
f1970baf JB |
1429 | } |
1430 | out: | |
1da177e4 LT |
1431 | kfree(pages); |
1432 | bio_put(bio); | |
1433 | return ERR_PTR(ret); | |
1434 | } | |
1435 | ||
1da177e4 LT |
1436 | static void __bio_unmap_user(struct bio *bio) |
1437 | { | |
1438 | struct bio_vec *bvec; | |
1439 | int i; | |
1440 | ||
1441 | /* | |
1442 | * make sure we dirty pages we wrote to | |
1443 | */ | |
d74c6d51 | 1444 | bio_for_each_segment_all(bvec, bio, i) { |
1da177e4 LT |
1445 | if (bio_data_dir(bio) == READ) |
1446 | set_page_dirty_lock(bvec->bv_page); | |
1447 | ||
09cbfeaf | 1448 | put_page(bvec->bv_page); |
1da177e4 LT |
1449 | } |
1450 | ||
1451 | bio_put(bio); | |
1452 | } | |
1453 | ||
1454 | /** | |
1455 | * bio_unmap_user - unmap a bio | |
1456 | * @bio: the bio being unmapped | |
1457 | * | |
5fad1b64 BVA |
1458 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
1459 | * process context. | |
1da177e4 LT |
1460 | * |
1461 | * bio_unmap_user() may sleep. | |
1462 | */ | |
1463 | void bio_unmap_user(struct bio *bio) | |
1464 | { | |
1465 | __bio_unmap_user(bio); | |
1466 | bio_put(bio); | |
1467 | } | |
1468 | ||
4246a0b6 | 1469 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1470 | { |
b823825e | 1471 | bio_put(bio); |
b823825e JA |
1472 | } |
1473 | ||
75c72b83 CH |
1474 | /** |
1475 | * bio_map_kern - map kernel address into bio | |
1476 | * @q: the struct request_queue for the bio | |
1477 | * @data: pointer to buffer to map | |
1478 | * @len: length in bytes | |
1479 | * @gfp_mask: allocation flags for bio allocation | |
1480 | * | |
1481 | * Map the kernel address into a bio suitable for io to a block | |
1482 | * device. Returns an error pointer in case of error. | |
1483 | */ | |
1484 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1485 | gfp_t gfp_mask) | |
df46b9a4 MC |
1486 | { |
1487 | unsigned long kaddr = (unsigned long)data; | |
1488 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1489 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1490 | const int nr_pages = end - start; | |
1491 | int offset, i; | |
1492 | struct bio *bio; | |
1493 | ||
a9e9dc24 | 1494 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1495 | if (!bio) |
1496 | return ERR_PTR(-ENOMEM); | |
1497 | ||
1498 | offset = offset_in_page(kaddr); | |
1499 | for (i = 0; i < nr_pages; i++) { | |
1500 | unsigned int bytes = PAGE_SIZE - offset; | |
1501 | ||
1502 | if (len <= 0) | |
1503 | break; | |
1504 | ||
1505 | if (bytes > len) | |
1506 | bytes = len; | |
1507 | ||
defd94b7 | 1508 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1509 | offset) < bytes) { |
1510 | /* we don't support partial mappings */ | |
1511 | bio_put(bio); | |
1512 | return ERR_PTR(-EINVAL); | |
1513 | } | |
df46b9a4 MC |
1514 | |
1515 | data += bytes; | |
1516 | len -= bytes; | |
1517 | offset = 0; | |
1518 | } | |
1519 | ||
b823825e | 1520 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1521 | return bio; |
1522 | } | |
a112a71d | 1523 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1524 | |
4246a0b6 | 1525 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1526 | { |
1dfa0f68 CH |
1527 | bio_free_pages(bio); |
1528 | bio_put(bio); | |
1529 | } | |
1530 | ||
4246a0b6 | 1531 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1532 | { |
42d2683a | 1533 | char *p = bio->bi_private; |
1dfa0f68 | 1534 | struct bio_vec *bvec; |
68154e90 FT |
1535 | int i; |
1536 | ||
d74c6d51 | 1537 | bio_for_each_segment_all(bvec, bio, i) { |
1dfa0f68 | 1538 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1539 | p += bvec->bv_len; |
68154e90 FT |
1540 | } |
1541 | ||
4246a0b6 | 1542 | bio_copy_kern_endio(bio); |
68154e90 FT |
1543 | } |
1544 | ||
1545 | /** | |
1546 | * bio_copy_kern - copy kernel address into bio | |
1547 | * @q: the struct request_queue for the bio | |
1548 | * @data: pointer to buffer to copy | |
1549 | * @len: length in bytes | |
1550 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1551 | * @reading: data direction is READ |
68154e90 FT |
1552 | * |
1553 | * copy the kernel address into a bio suitable for io to a block | |
1554 | * device. Returns an error pointer in case of error. | |
1555 | */ | |
1556 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1557 | gfp_t gfp_mask, int reading) | |
1558 | { | |
42d2683a CH |
1559 | unsigned long kaddr = (unsigned long)data; |
1560 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1561 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1562 | struct bio *bio; |
1563 | void *p = data; | |
1dfa0f68 | 1564 | int nr_pages = 0; |
68154e90 | 1565 | |
42d2683a CH |
1566 | /* |
1567 | * Overflow, abort | |
1568 | */ | |
1569 | if (end < start) | |
1570 | return ERR_PTR(-EINVAL); | |
68154e90 | 1571 | |
42d2683a CH |
1572 | nr_pages = end - start; |
1573 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1574 | if (!bio) | |
1575 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1576 | |
42d2683a CH |
1577 | while (len) { |
1578 | struct page *page; | |
1579 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1580 | |
42d2683a CH |
1581 | if (bytes > len) |
1582 | bytes = len; | |
1583 | ||
1584 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1585 | if (!page) | |
1586 | goto cleanup; | |
1587 | ||
1588 | if (!reading) | |
1589 | memcpy(page_address(page), p, bytes); | |
1590 | ||
1591 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1592 | break; | |
1593 | ||
1594 | len -= bytes; | |
1595 | p += bytes; | |
68154e90 FT |
1596 | } |
1597 | ||
1dfa0f68 CH |
1598 | if (reading) { |
1599 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1600 | bio->bi_private = data; | |
1601 | } else { | |
1602 | bio->bi_end_io = bio_copy_kern_endio; | |
1dfa0f68 | 1603 | } |
76029ff3 | 1604 | |
68154e90 | 1605 | return bio; |
42d2683a CH |
1606 | |
1607 | cleanup: | |
1dfa0f68 | 1608 | bio_free_pages(bio); |
42d2683a CH |
1609 | bio_put(bio); |
1610 | return ERR_PTR(-ENOMEM); | |
68154e90 FT |
1611 | } |
1612 | ||
1da177e4 LT |
1613 | /* |
1614 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1615 | * for performing direct-IO in BIOs. | |
1616 | * | |
1617 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1618 | * because the required locks are not interrupt-safe. So what we can do is to | |
1619 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1620 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1621 | * in process context. | |
1622 | * | |
1623 | * We special-case compound pages here: normally this means reads into hugetlb | |
1624 | * pages. The logic in here doesn't really work right for compound pages | |
1625 | * because the VM does not uniformly chase down the head page in all cases. | |
1626 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1627 | * handle them at all. So we skip compound pages here at an early stage. | |
1628 | * | |
1629 | * Note that this code is very hard to test under normal circumstances because | |
1630 | * direct-io pins the pages with get_user_pages(). This makes | |
1631 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1632 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1633 | * pagecache. |
1634 | * | |
1635 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1636 | * deferred bio dirtying paths. | |
1637 | */ | |
1638 | ||
1639 | /* | |
1640 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1641 | */ | |
1642 | void bio_set_pages_dirty(struct bio *bio) | |
1643 | { | |
cb34e057 | 1644 | struct bio_vec *bvec; |
1da177e4 LT |
1645 | int i; |
1646 | ||
cb34e057 KO |
1647 | bio_for_each_segment_all(bvec, bio, i) { |
1648 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1649 | |
1650 | if (page && !PageCompound(page)) | |
1651 | set_page_dirty_lock(page); | |
1652 | } | |
1653 | } | |
1654 | ||
86b6c7a7 | 1655 | static void bio_release_pages(struct bio *bio) |
1da177e4 | 1656 | { |
cb34e057 | 1657 | struct bio_vec *bvec; |
1da177e4 LT |
1658 | int i; |
1659 | ||
cb34e057 KO |
1660 | bio_for_each_segment_all(bvec, bio, i) { |
1661 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1662 | |
1663 | if (page) | |
1664 | put_page(page); | |
1665 | } | |
1666 | } | |
1667 | ||
1668 | /* | |
1669 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1670 | * If they are, then fine. If, however, some pages are clean then they must | |
1671 | * have been written out during the direct-IO read. So we take another ref on | |
1672 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1673 | * | |
1674 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1675 | * here on. It will run one put_page() against each page and will run one |
1676 | * bio_put() against the BIO. | |
1da177e4 LT |
1677 | */ |
1678 | ||
65f27f38 | 1679 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1680 | |
65f27f38 | 1681 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1682 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1683 | static struct bio *bio_dirty_list; | |
1684 | ||
1685 | /* | |
1686 | * This runs in process context | |
1687 | */ | |
65f27f38 | 1688 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1689 | { |
1690 | unsigned long flags; | |
1691 | struct bio *bio; | |
1692 | ||
1693 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1694 | bio = bio_dirty_list; | |
1695 | bio_dirty_list = NULL; | |
1696 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1697 | ||
1698 | while (bio) { | |
1699 | struct bio *next = bio->bi_private; | |
1700 | ||
1701 | bio_set_pages_dirty(bio); | |
1702 | bio_release_pages(bio); | |
1703 | bio_put(bio); | |
1704 | bio = next; | |
1705 | } | |
1706 | } | |
1707 | ||
1708 | void bio_check_pages_dirty(struct bio *bio) | |
1709 | { | |
cb34e057 | 1710 | struct bio_vec *bvec; |
1da177e4 LT |
1711 | int nr_clean_pages = 0; |
1712 | int i; | |
1713 | ||
cb34e057 KO |
1714 | bio_for_each_segment_all(bvec, bio, i) { |
1715 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1716 | |
1717 | if (PageDirty(page) || PageCompound(page)) { | |
09cbfeaf | 1718 | put_page(page); |
cb34e057 | 1719 | bvec->bv_page = NULL; |
1da177e4 LT |
1720 | } else { |
1721 | nr_clean_pages++; | |
1722 | } | |
1723 | } | |
1724 | ||
1725 | if (nr_clean_pages) { | |
1726 | unsigned long flags; | |
1727 | ||
1728 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1729 | bio->bi_private = bio_dirty_list; | |
1730 | bio_dirty_list = bio; | |
1731 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1732 | schedule_work(&bio_dirty_work); | |
1733 | } else { | |
1734 | bio_put(bio); | |
1735 | } | |
1736 | } | |
1737 | ||
394ffa50 GZ |
1738 | void generic_start_io_acct(int rw, unsigned long sectors, |
1739 | struct hd_struct *part) | |
1740 | { | |
1741 | int cpu = part_stat_lock(); | |
1742 | ||
1743 | part_round_stats(cpu, part); | |
1744 | part_stat_inc(cpu, part, ios[rw]); | |
1745 | part_stat_add(cpu, part, sectors[rw], sectors); | |
1746 | part_inc_in_flight(part, rw); | |
1747 | ||
1748 | part_stat_unlock(); | |
1749 | } | |
1750 | EXPORT_SYMBOL(generic_start_io_acct); | |
1751 | ||
1752 | void generic_end_io_acct(int rw, struct hd_struct *part, | |
1753 | unsigned long start_time) | |
1754 | { | |
1755 | unsigned long duration = jiffies - start_time; | |
1756 | int cpu = part_stat_lock(); | |
1757 | ||
1758 | part_stat_add(cpu, part, ticks[rw], duration); | |
1759 | part_round_stats(cpu, part); | |
1760 | part_dec_in_flight(part, rw); | |
1761 | ||
1762 | part_stat_unlock(); | |
1763 | } | |
1764 | EXPORT_SYMBOL(generic_end_io_acct); | |
1765 | ||
2d4dc890 IL |
1766 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1767 | void bio_flush_dcache_pages(struct bio *bi) | |
1768 | { | |
7988613b KO |
1769 | struct bio_vec bvec; |
1770 | struct bvec_iter iter; | |
2d4dc890 | 1771 | |
7988613b KO |
1772 | bio_for_each_segment(bvec, bi, iter) |
1773 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1774 | } |
1775 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1776 | #endif | |
1777 | ||
c4cf5261 JA |
1778 | static inline bool bio_remaining_done(struct bio *bio) |
1779 | { | |
1780 | /* | |
1781 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1782 | * we always end io on the first invocation. | |
1783 | */ | |
1784 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1785 | return true; | |
1786 | ||
1787 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1788 | ||
326e1dbb | 1789 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1790 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1791 | return true; |
326e1dbb | 1792 | } |
c4cf5261 JA |
1793 | |
1794 | return false; | |
1795 | } | |
1796 | ||
1da177e4 LT |
1797 | /** |
1798 | * bio_endio - end I/O on a bio | |
1799 | * @bio: bio | |
1da177e4 LT |
1800 | * |
1801 | * Description: | |
4246a0b6 CH |
1802 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1803 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1804 | * bio unless they own it and thus know that it has an end_io function. | |
fbbaf700 N |
1805 | * |
1806 | * bio_endio() can be called several times on a bio that has been chained | |
1807 | * using bio_chain(). The ->bi_end_io() function will only be called the | |
1808 | * last time. At this point the BLK_TA_COMPLETE tracing event will be | |
1809 | * generated if BIO_TRACE_COMPLETION is set. | |
1da177e4 | 1810 | **/ |
4246a0b6 | 1811 | void bio_endio(struct bio *bio) |
1da177e4 | 1812 | { |
ba8c6967 | 1813 | again: |
2b885517 | 1814 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1815 | return; |
1da177e4 | 1816 | |
ba8c6967 CH |
1817 | /* |
1818 | * Need to have a real endio function for chained bios, otherwise | |
1819 | * various corner cases will break (like stacking block devices that | |
1820 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1821 | * recursion and blowing the stack. Tail call optimization would | |
1822 | * handle this, but compiling with frame pointers also disables | |
1823 | * gcc's sibling call optimization. | |
1824 | */ | |
1825 | if (bio->bi_end_io == bio_chain_endio) { | |
1826 | bio = __bio_chain_endio(bio); | |
1827 | goto again; | |
196d38bc | 1828 | } |
ba8c6967 | 1829 | |
fbbaf700 | 1830 | if (bio->bi_bdev && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
a462b950 BVA |
1831 | trace_block_bio_complete(bdev_get_queue(bio->bi_bdev), bio, |
1832 | blk_status_to_errno(bio->bi_status)); | |
fbbaf700 N |
1833 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
1834 | } | |
1835 | ||
9e234eea | 1836 | blk_throtl_bio_endio(bio); |
ba8c6967 CH |
1837 | if (bio->bi_end_io) |
1838 | bio->bi_end_io(bio); | |
1da177e4 | 1839 | } |
a112a71d | 1840 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1841 | |
20d0189b KO |
1842 | /** |
1843 | * bio_split - split a bio | |
1844 | * @bio: bio to split | |
1845 | * @sectors: number of sectors to split from the front of @bio | |
1846 | * @gfp: gfp mask | |
1847 | * @bs: bio set to allocate from | |
1848 | * | |
1849 | * Allocates and returns a new bio which represents @sectors from the start of | |
1850 | * @bio, and updates @bio to represent the remaining sectors. | |
1851 | * | |
f3f5da62 MP |
1852 | * Unless this is a discard request the newly allocated bio will point |
1853 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1854 | * @bio is not freed before the split. | |
20d0189b KO |
1855 | */ |
1856 | struct bio *bio_split(struct bio *bio, int sectors, | |
1857 | gfp_t gfp, struct bio_set *bs) | |
1858 | { | |
1859 | struct bio *split = NULL; | |
1860 | ||
1861 | BUG_ON(sectors <= 0); | |
1862 | BUG_ON(sectors >= bio_sectors(bio)); | |
1863 | ||
f9d03f96 | 1864 | split = bio_clone_fast(bio, gfp, bs); |
20d0189b KO |
1865 | if (!split) |
1866 | return NULL; | |
1867 | ||
1868 | split->bi_iter.bi_size = sectors << 9; | |
1869 | ||
1870 | if (bio_integrity(split)) | |
1871 | bio_integrity_trim(split, 0, sectors); | |
1872 | ||
1873 | bio_advance(bio, split->bi_iter.bi_size); | |
1874 | ||
fbbaf700 N |
1875 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
1876 | bio_set_flag(bio, BIO_TRACE_COMPLETION); | |
1877 | ||
20d0189b KO |
1878 | return split; |
1879 | } | |
1880 | EXPORT_SYMBOL(bio_split); | |
1881 | ||
6678d83f KO |
1882 | /** |
1883 | * bio_trim - trim a bio | |
1884 | * @bio: bio to trim | |
1885 | * @offset: number of sectors to trim from the front of @bio | |
1886 | * @size: size we want to trim @bio to, in sectors | |
1887 | */ | |
1888 | void bio_trim(struct bio *bio, int offset, int size) | |
1889 | { | |
1890 | /* 'bio' is a cloned bio which we need to trim to match | |
1891 | * the given offset and size. | |
6678d83f | 1892 | */ |
6678d83f KO |
1893 | |
1894 | size <<= 9; | |
4f024f37 | 1895 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1896 | return; |
1897 | ||
b7c44ed9 | 1898 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1899 | |
1900 | bio_advance(bio, offset << 9); | |
1901 | ||
4f024f37 | 1902 | bio->bi_iter.bi_size = size; |
6678d83f KO |
1903 | } |
1904 | EXPORT_SYMBOL_GPL(bio_trim); | |
1905 | ||
1da177e4 LT |
1906 | /* |
1907 | * create memory pools for biovec's in a bio_set. | |
1908 | * use the global biovec slabs created for general use. | |
1909 | */ | |
a6c39cb4 | 1910 | mempool_t *biovec_create_pool(int pool_entries) |
1da177e4 | 1911 | { |
ed996a52 | 1912 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
1da177e4 | 1913 | |
9f060e22 | 1914 | return mempool_create_slab_pool(pool_entries, bp->slab); |
1da177e4 LT |
1915 | } |
1916 | ||
1917 | void bioset_free(struct bio_set *bs) | |
1918 | { | |
df2cb6da KO |
1919 | if (bs->rescue_workqueue) |
1920 | destroy_workqueue(bs->rescue_workqueue); | |
1921 | ||
1da177e4 LT |
1922 | if (bs->bio_pool) |
1923 | mempool_destroy(bs->bio_pool); | |
1924 | ||
9f060e22 KO |
1925 | if (bs->bvec_pool) |
1926 | mempool_destroy(bs->bvec_pool); | |
1927 | ||
7878cba9 | 1928 | bioset_integrity_free(bs); |
bb799ca0 | 1929 | bio_put_slab(bs); |
1da177e4 LT |
1930 | |
1931 | kfree(bs); | |
1932 | } | |
a112a71d | 1933 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1934 | |
011067b0 N |
1935 | /** |
1936 | * bioset_create - Create a bio_set | |
1937 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1938 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
47e0fb46 N |
1939 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS |
1940 | * and %BIOSET_NEED_RESCUER | |
011067b0 N |
1941 | * |
1942 | * Description: | |
1943 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1944 | * to ask for a number of bytes to be allocated in front of the bio. | |
1945 | * Front pad allocation is useful for embedding the bio inside | |
1946 | * another structure, to avoid allocating extra data to go with the bio. | |
1947 | * Note that the bio must be embedded at the END of that structure always, | |
1948 | * or things will break badly. | |
1949 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated | |
1950 | * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast(). | |
47e0fb46 N |
1951 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to |
1952 | * dispatch queued requests when the mempool runs out of space. | |
011067b0 N |
1953 | * |
1954 | */ | |
1955 | struct bio_set *bioset_create(unsigned int pool_size, | |
1956 | unsigned int front_pad, | |
1957 | int flags) | |
1da177e4 | 1958 | { |
392ddc32 | 1959 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1960 | struct bio_set *bs; |
1da177e4 | 1961 | |
1b434498 | 1962 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1963 | if (!bs) |
1964 | return NULL; | |
1965 | ||
bb799ca0 | 1966 | bs->front_pad = front_pad; |
1b434498 | 1967 | |
df2cb6da KO |
1968 | spin_lock_init(&bs->rescue_lock); |
1969 | bio_list_init(&bs->rescue_list); | |
1970 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
1971 | ||
392ddc32 | 1972 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1973 | if (!bs->bio_slab) { |
1974 | kfree(bs); | |
1975 | return NULL; | |
1976 | } | |
1977 | ||
1978 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
1979 | if (!bs->bio_pool) |
1980 | goto bad; | |
1981 | ||
011067b0 | 1982 | if (flags & BIOSET_NEED_BVECS) { |
d8f429e1 JN |
1983 | bs->bvec_pool = biovec_create_pool(pool_size); |
1984 | if (!bs->bvec_pool) | |
1985 | goto bad; | |
1986 | } | |
df2cb6da | 1987 | |
47e0fb46 N |
1988 | if (!(flags & BIOSET_NEED_RESCUER)) |
1989 | return bs; | |
1990 | ||
df2cb6da KO |
1991 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); |
1992 | if (!bs->rescue_workqueue) | |
1993 | goto bad; | |
1da177e4 | 1994 | |
df2cb6da | 1995 | return bs; |
1da177e4 LT |
1996 | bad: |
1997 | bioset_free(bs); | |
1998 | return NULL; | |
1999 | } | |
a112a71d | 2000 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 2001 | |
852c788f | 2002 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 TH |
2003 | |
2004 | /** | |
2005 | * bio_associate_blkcg - associate a bio with the specified blkcg | |
2006 | * @bio: target bio | |
2007 | * @blkcg_css: css of the blkcg to associate | |
2008 | * | |
2009 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will | |
2010 | * treat @bio as if it were issued by a task which belongs to the blkcg. | |
2011 | * | |
2012 | * This function takes an extra reference of @blkcg_css which will be put | |
2013 | * when @bio is released. The caller must own @bio and is responsible for | |
2014 | * synchronizing calls to this function. | |
2015 | */ | |
2016 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) | |
2017 | { | |
2018 | if (unlikely(bio->bi_css)) | |
2019 | return -EBUSY; | |
2020 | css_get(blkcg_css); | |
2021 | bio->bi_css = blkcg_css; | |
2022 | return 0; | |
2023 | } | |
5aa2a96b | 2024 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
1d933cf0 | 2025 | |
852c788f TH |
2026 | /** |
2027 | * bio_associate_current - associate a bio with %current | |
2028 | * @bio: target bio | |
2029 | * | |
2030 | * Associate @bio with %current if it hasn't been associated yet. Block | |
2031 | * layer will treat @bio as if it were issued by %current no matter which | |
2032 | * task actually issues it. | |
2033 | * | |
2034 | * This function takes an extra reference of @task's io_context and blkcg | |
2035 | * which will be put when @bio is released. The caller must own @bio, | |
2036 | * ensure %current->io_context exists, and is responsible for synchronizing | |
2037 | * calls to this function. | |
2038 | */ | |
2039 | int bio_associate_current(struct bio *bio) | |
2040 | { | |
2041 | struct io_context *ioc; | |
852c788f | 2042 | |
1d933cf0 | 2043 | if (bio->bi_css) |
852c788f TH |
2044 | return -EBUSY; |
2045 | ||
2046 | ioc = current->io_context; | |
2047 | if (!ioc) | |
2048 | return -ENOENT; | |
2049 | ||
852c788f TH |
2050 | get_io_context_active(ioc); |
2051 | bio->bi_ioc = ioc; | |
c165b3e3 | 2052 | bio->bi_css = task_get_css(current, io_cgrp_id); |
852c788f TH |
2053 | return 0; |
2054 | } | |
5aa2a96b | 2055 | EXPORT_SYMBOL_GPL(bio_associate_current); |
852c788f TH |
2056 | |
2057 | /** | |
2058 | * bio_disassociate_task - undo bio_associate_current() | |
2059 | * @bio: target bio | |
2060 | */ | |
2061 | void bio_disassociate_task(struct bio *bio) | |
2062 | { | |
2063 | if (bio->bi_ioc) { | |
2064 | put_io_context(bio->bi_ioc); | |
2065 | bio->bi_ioc = NULL; | |
2066 | } | |
2067 | if (bio->bi_css) { | |
2068 | css_put(bio->bi_css); | |
2069 | bio->bi_css = NULL; | |
2070 | } | |
2071 | } | |
2072 | ||
20bd723e PV |
2073 | /** |
2074 | * bio_clone_blkcg_association - clone blkcg association from src to dst bio | |
2075 | * @dst: destination bio | |
2076 | * @src: source bio | |
2077 | */ | |
2078 | void bio_clone_blkcg_association(struct bio *dst, struct bio *src) | |
2079 | { | |
2080 | if (src->bi_css) | |
2081 | WARN_ON(bio_associate_blkcg(dst, src->bi_css)); | |
2082 | } | |
2083 | ||
852c788f TH |
2084 | #endif /* CONFIG_BLK_CGROUP */ |
2085 | ||
1da177e4 LT |
2086 | static void __init biovec_init_slabs(void) |
2087 | { | |
2088 | int i; | |
2089 | ||
ed996a52 | 2090 | for (i = 0; i < BVEC_POOL_NR; i++) { |
1da177e4 LT |
2091 | int size; |
2092 | struct biovec_slab *bvs = bvec_slabs + i; | |
2093 | ||
a7fcd37c JA |
2094 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2095 | bvs->slab = NULL; | |
2096 | continue; | |
2097 | } | |
a7fcd37c | 2098 | |
1da177e4 LT |
2099 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2100 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2101 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2102 | } |
2103 | } | |
2104 | ||
2105 | static int __init init_bio(void) | |
2106 | { | |
bb799ca0 JA |
2107 | bio_slab_max = 2; |
2108 | bio_slab_nr = 0; | |
2109 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
2110 | if (!bio_slabs) | |
2111 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2112 | |
7878cba9 | 2113 | bio_integrity_init(); |
1da177e4 LT |
2114 | biovec_init_slabs(); |
2115 | ||
011067b0 | 2116 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS); |
1da177e4 LT |
2117 | if (!fs_bio_set) |
2118 | panic("bio: can't allocate bios\n"); | |
2119 | ||
a91a2785 MP |
2120 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
2121 | panic("bio: can't create integrity pool\n"); | |
2122 | ||
1da177e4 LT |
2123 | return 0; |
2124 | } | |
1da177e4 | 2125 | subsys_initcall(init_bio); |