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 | */ | |
92e3d3f6 | 46 | #define BV(x, n) { .nr_vecs = x, .name = "biovec-"#n } |
ed996a52 | 47 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
92e3d3f6 | 48 | BV(1, 1), BV(4, 4), BV(16, 16), BV(64, 64), BV(128, 128), BV(BIO_MAX_PAGES, max), |
1da177e4 LT |
49 | }; |
50 | #undef BV | |
51 | ||
1da177e4 LT |
52 | /* |
53 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
54 | * IO code that does not need private memory pools. | |
55 | */ | |
51d654e1 | 56 | struct bio_set *fs_bio_set; |
3f86a82a | 57 | EXPORT_SYMBOL(fs_bio_set); |
1da177e4 | 58 | |
bb799ca0 JA |
59 | /* |
60 | * Our slab pool management | |
61 | */ | |
62 | struct bio_slab { | |
63 | struct kmem_cache *slab; | |
64 | unsigned int slab_ref; | |
65 | unsigned int slab_size; | |
66 | char name[8]; | |
67 | }; | |
68 | static DEFINE_MUTEX(bio_slab_lock); | |
69 | static struct bio_slab *bio_slabs; | |
70 | static unsigned int bio_slab_nr, bio_slab_max; | |
71 | ||
72 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
73 | { | |
74 | unsigned int sz = sizeof(struct bio) + extra_size; | |
75 | struct kmem_cache *slab = NULL; | |
389d7b26 | 76 | struct bio_slab *bslab, *new_bio_slabs; |
386bc35a | 77 | unsigned int new_bio_slab_max; |
bb799ca0 JA |
78 | unsigned int i, entry = -1; |
79 | ||
80 | mutex_lock(&bio_slab_lock); | |
81 | ||
82 | i = 0; | |
83 | while (i < bio_slab_nr) { | |
f06f135d | 84 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
85 | |
86 | if (!bslab->slab && entry == -1) | |
87 | entry = i; | |
88 | else if (bslab->slab_size == sz) { | |
89 | slab = bslab->slab; | |
90 | bslab->slab_ref++; | |
91 | break; | |
92 | } | |
93 | i++; | |
94 | } | |
95 | ||
96 | if (slab) | |
97 | goto out_unlock; | |
98 | ||
99 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
386bc35a | 100 | new_bio_slab_max = bio_slab_max << 1; |
389d7b26 | 101 | new_bio_slabs = krealloc(bio_slabs, |
386bc35a | 102 | new_bio_slab_max * sizeof(struct bio_slab), |
389d7b26 AK |
103 | GFP_KERNEL); |
104 | if (!new_bio_slabs) | |
bb799ca0 | 105 | goto out_unlock; |
386bc35a | 106 | bio_slab_max = new_bio_slab_max; |
389d7b26 | 107 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
108 | } |
109 | if (entry == -1) | |
110 | entry = bio_slab_nr++; | |
111 | ||
112 | bslab = &bio_slabs[entry]; | |
113 | ||
114 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
6a241483 MP |
115 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
116 | SLAB_HWCACHE_ALIGN, NULL); | |
bb799ca0 JA |
117 | if (!slab) |
118 | goto out_unlock; | |
119 | ||
bb799ca0 JA |
120 | bslab->slab = slab; |
121 | bslab->slab_ref = 1; | |
122 | bslab->slab_size = sz; | |
123 | out_unlock: | |
124 | mutex_unlock(&bio_slab_lock); | |
125 | return slab; | |
126 | } | |
127 | ||
128 | static void bio_put_slab(struct bio_set *bs) | |
129 | { | |
130 | struct bio_slab *bslab = NULL; | |
131 | unsigned int i; | |
132 | ||
133 | mutex_lock(&bio_slab_lock); | |
134 | ||
135 | for (i = 0; i < bio_slab_nr; i++) { | |
136 | if (bs->bio_slab == bio_slabs[i].slab) { | |
137 | bslab = &bio_slabs[i]; | |
138 | break; | |
139 | } | |
140 | } | |
141 | ||
142 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
143 | goto out; | |
144 | ||
145 | WARN_ON(!bslab->slab_ref); | |
146 | ||
147 | if (--bslab->slab_ref) | |
148 | goto out; | |
149 | ||
150 | kmem_cache_destroy(bslab->slab); | |
151 | bslab->slab = NULL; | |
152 | ||
153 | out: | |
154 | mutex_unlock(&bio_slab_lock); | |
155 | } | |
156 | ||
7ba1ba12 MP |
157 | unsigned int bvec_nr_vecs(unsigned short idx) |
158 | { | |
159 | return bvec_slabs[idx].nr_vecs; | |
160 | } | |
161 | ||
9f060e22 | 162 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
bb799ca0 | 163 | { |
ed996a52 CH |
164 | if (!idx) |
165 | return; | |
166 | idx--; | |
167 | ||
168 | BIO_BUG_ON(idx >= BVEC_POOL_NR); | |
bb799ca0 | 169 | |
ed996a52 | 170 | if (idx == BVEC_POOL_MAX) { |
9f060e22 | 171 | mempool_free(bv, pool); |
ed996a52 | 172 | } else { |
bb799ca0 JA |
173 | struct biovec_slab *bvs = bvec_slabs + idx; |
174 | ||
175 | kmem_cache_free(bvs->slab, bv); | |
176 | } | |
177 | } | |
178 | ||
9f060e22 KO |
179 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
180 | mempool_t *pool) | |
1da177e4 LT |
181 | { |
182 | struct bio_vec *bvl; | |
1da177e4 | 183 | |
7ff9345f JA |
184 | /* |
185 | * see comment near bvec_array define! | |
186 | */ | |
187 | switch (nr) { | |
188 | case 1: | |
189 | *idx = 0; | |
190 | break; | |
191 | case 2 ... 4: | |
192 | *idx = 1; | |
193 | break; | |
194 | case 5 ... 16: | |
195 | *idx = 2; | |
196 | break; | |
197 | case 17 ... 64: | |
198 | *idx = 3; | |
199 | break; | |
200 | case 65 ... 128: | |
201 | *idx = 4; | |
202 | break; | |
203 | case 129 ... BIO_MAX_PAGES: | |
204 | *idx = 5; | |
205 | break; | |
206 | default: | |
207 | return NULL; | |
208 | } | |
209 | ||
210 | /* | |
211 | * idx now points to the pool we want to allocate from. only the | |
212 | * 1-vec entry pool is mempool backed. | |
213 | */ | |
ed996a52 | 214 | if (*idx == BVEC_POOL_MAX) { |
7ff9345f | 215 | fallback: |
9f060e22 | 216 | bvl = mempool_alloc(pool, gfp_mask); |
7ff9345f JA |
217 | } else { |
218 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
d0164adc | 219 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
7ff9345f | 220 | |
0a0d96b0 | 221 | /* |
7ff9345f JA |
222 | * Make this allocation restricted and don't dump info on |
223 | * allocation failures, since we'll fallback to the mempool | |
224 | * in case of failure. | |
0a0d96b0 | 225 | */ |
7ff9345f | 226 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 227 | |
0a0d96b0 | 228 | /* |
d0164adc | 229 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
7ff9345f | 230 | * is set, retry with the 1-entry mempool |
0a0d96b0 | 231 | */ |
7ff9345f | 232 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
d0164adc | 233 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
ed996a52 | 234 | *idx = BVEC_POOL_MAX; |
7ff9345f JA |
235 | goto fallback; |
236 | } | |
237 | } | |
238 | ||
ed996a52 | 239 | (*idx)++; |
1da177e4 LT |
240 | return bvl; |
241 | } | |
242 | ||
9ae3b3f5 | 243 | void bio_uninit(struct bio *bio) |
1da177e4 | 244 | { |
4254bba1 | 245 | bio_disassociate_task(bio); |
4254bba1 | 246 | } |
9ae3b3f5 | 247 | EXPORT_SYMBOL(bio_uninit); |
7ba1ba12 | 248 | |
4254bba1 KO |
249 | static void bio_free(struct bio *bio) |
250 | { | |
251 | struct bio_set *bs = bio->bi_pool; | |
252 | void *p; | |
253 | ||
9ae3b3f5 | 254 | bio_uninit(bio); |
4254bba1 KO |
255 | |
256 | if (bs) { | |
ed996a52 | 257 | bvec_free(bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio)); |
4254bba1 KO |
258 | |
259 | /* | |
260 | * If we have front padding, adjust the bio pointer before freeing | |
261 | */ | |
262 | p = bio; | |
bb799ca0 JA |
263 | p -= bs->front_pad; |
264 | ||
4254bba1 KO |
265 | mempool_free(p, bs->bio_pool); |
266 | } else { | |
267 | /* Bio was allocated by bio_kmalloc() */ | |
268 | kfree(bio); | |
269 | } | |
3676347a PO |
270 | } |
271 | ||
9ae3b3f5 JA |
272 | /* |
273 | * Users of this function have their own bio allocation. Subsequently, | |
274 | * they must remember to pair any call to bio_init() with bio_uninit() | |
275 | * when IO has completed, or when the bio is released. | |
276 | */ | |
3a83f467 ML |
277 | void bio_init(struct bio *bio, struct bio_vec *table, |
278 | unsigned short max_vecs) | |
1da177e4 | 279 | { |
2b94de55 | 280 | memset(bio, 0, sizeof(*bio)); |
c4cf5261 | 281 | atomic_set(&bio->__bi_remaining, 1); |
dac56212 | 282 | atomic_set(&bio->__bi_cnt, 1); |
3a83f467 ML |
283 | |
284 | bio->bi_io_vec = table; | |
285 | bio->bi_max_vecs = max_vecs; | |
1da177e4 | 286 | } |
a112a71d | 287 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 288 | |
f44b48c7 KO |
289 | /** |
290 | * bio_reset - reinitialize a bio | |
291 | * @bio: bio to reset | |
292 | * | |
293 | * Description: | |
294 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
295 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
296 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
297 | * comment in struct bio. | |
298 | */ | |
299 | void bio_reset(struct bio *bio) | |
300 | { | |
301 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
302 | ||
9ae3b3f5 | 303 | bio_uninit(bio); |
f44b48c7 KO |
304 | |
305 | memset(bio, 0, BIO_RESET_BYTES); | |
4246a0b6 | 306 | bio->bi_flags = flags; |
c4cf5261 | 307 | atomic_set(&bio->__bi_remaining, 1); |
f44b48c7 KO |
308 | } |
309 | EXPORT_SYMBOL(bio_reset); | |
310 | ||
38f8baae | 311 | static struct bio *__bio_chain_endio(struct bio *bio) |
196d38bc | 312 | { |
4246a0b6 CH |
313 | struct bio *parent = bio->bi_private; |
314 | ||
4e4cbee9 CH |
315 | if (!parent->bi_status) |
316 | parent->bi_status = bio->bi_status; | |
196d38bc | 317 | bio_put(bio); |
38f8baae CH |
318 | return parent; |
319 | } | |
320 | ||
321 | static void bio_chain_endio(struct bio *bio) | |
322 | { | |
323 | bio_endio(__bio_chain_endio(bio)); | |
196d38bc KO |
324 | } |
325 | ||
326 | /** | |
327 | * bio_chain - chain bio completions | |
1051a902 RD |
328 | * @bio: the target bio |
329 | * @parent: the @bio's parent bio | |
196d38bc KO |
330 | * |
331 | * The caller won't have a bi_end_io called when @bio completes - instead, | |
332 | * @parent's bi_end_io won't be called until both @parent and @bio have | |
333 | * completed; the chained bio will also be freed when it completes. | |
334 | * | |
335 | * The caller must not set bi_private or bi_end_io in @bio. | |
336 | */ | |
337 | void bio_chain(struct bio *bio, struct bio *parent) | |
338 | { | |
339 | BUG_ON(bio->bi_private || bio->bi_end_io); | |
340 | ||
341 | bio->bi_private = parent; | |
342 | bio->bi_end_io = bio_chain_endio; | |
c4cf5261 | 343 | bio_inc_remaining(parent); |
196d38bc KO |
344 | } |
345 | EXPORT_SYMBOL(bio_chain); | |
346 | ||
df2cb6da KO |
347 | static void bio_alloc_rescue(struct work_struct *work) |
348 | { | |
349 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); | |
350 | struct bio *bio; | |
351 | ||
352 | while (1) { | |
353 | spin_lock(&bs->rescue_lock); | |
354 | bio = bio_list_pop(&bs->rescue_list); | |
355 | spin_unlock(&bs->rescue_lock); | |
356 | ||
357 | if (!bio) | |
358 | break; | |
359 | ||
360 | generic_make_request(bio); | |
361 | } | |
362 | } | |
363 | ||
364 | static void punt_bios_to_rescuer(struct bio_set *bs) | |
365 | { | |
366 | struct bio_list punt, nopunt; | |
367 | struct bio *bio; | |
368 | ||
47e0fb46 N |
369 | if (WARN_ON_ONCE(!bs->rescue_workqueue)) |
370 | return; | |
df2cb6da KO |
371 | /* |
372 | * In order to guarantee forward progress we must punt only bios that | |
373 | * were allocated from this bio_set; otherwise, if there was a bio on | |
374 | * there for a stacking driver higher up in the stack, processing it | |
375 | * could require allocating bios from this bio_set, and doing that from | |
376 | * our own rescuer would be bad. | |
377 | * | |
378 | * Since bio lists are singly linked, pop them all instead of trying to | |
379 | * remove from the middle of the list: | |
380 | */ | |
381 | ||
382 | bio_list_init(&punt); | |
383 | bio_list_init(&nopunt); | |
384 | ||
f5fe1b51 | 385 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
df2cb6da | 386 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
f5fe1b51 | 387 | current->bio_list[0] = nopunt; |
df2cb6da | 388 | |
f5fe1b51 N |
389 | bio_list_init(&nopunt); |
390 | while ((bio = bio_list_pop(¤t->bio_list[1]))) | |
391 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); | |
392 | current->bio_list[1] = nopunt; | |
df2cb6da KO |
393 | |
394 | spin_lock(&bs->rescue_lock); | |
395 | bio_list_merge(&bs->rescue_list, &punt); | |
396 | spin_unlock(&bs->rescue_lock); | |
397 | ||
398 | queue_work(bs->rescue_workqueue, &bs->rescue_work); | |
399 | } | |
400 | ||
1da177e4 LT |
401 | /** |
402 | * bio_alloc_bioset - allocate a bio for I/O | |
403 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
404 | * @nr_iovecs: number of iovecs to pre-allocate | |
db18efac | 405 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
406 | * |
407 | * Description: | |
3f86a82a KO |
408 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
409 | * backed by the @bs's mempool. | |
410 | * | |
d0164adc MG |
411 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
412 | * always be able to allocate a bio. This is due to the mempool guarantees. | |
413 | * To make this work, callers must never allocate more than 1 bio at a time | |
414 | * from this pool. Callers that need to allocate more than 1 bio must always | |
415 | * submit the previously allocated bio for IO before attempting to allocate | |
416 | * a new one. Failure to do so can cause deadlocks under memory pressure. | |
3f86a82a | 417 | * |
df2cb6da KO |
418 | * Note that when running under generic_make_request() (i.e. any block |
419 | * driver), bios are not submitted until after you return - see the code in | |
420 | * generic_make_request() that converts recursion into iteration, to prevent | |
421 | * stack overflows. | |
422 | * | |
423 | * This would normally mean allocating multiple bios under | |
424 | * generic_make_request() would be susceptible to deadlocks, but we have | |
425 | * deadlock avoidance code that resubmits any blocked bios from a rescuer | |
426 | * thread. | |
427 | * | |
428 | * However, we do not guarantee forward progress for allocations from other | |
429 | * mempools. Doing multiple allocations from the same mempool under | |
430 | * generic_make_request() should be avoided - instead, use bio_set's front_pad | |
431 | * for per bio allocations. | |
432 | * | |
3f86a82a KO |
433 | * RETURNS: |
434 | * Pointer to new bio on success, NULL on failure. | |
435 | */ | |
7a88fa19 DC |
436 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, unsigned int nr_iovecs, |
437 | struct bio_set *bs) | |
1da177e4 | 438 | { |
df2cb6da | 439 | gfp_t saved_gfp = gfp_mask; |
3f86a82a KO |
440 | unsigned front_pad; |
441 | unsigned inline_vecs; | |
34053979 | 442 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
443 | struct bio *bio; |
444 | void *p; | |
445 | ||
3f86a82a KO |
446 | if (!bs) { |
447 | if (nr_iovecs > UIO_MAXIOV) | |
448 | return NULL; | |
449 | ||
450 | p = kmalloc(sizeof(struct bio) + | |
451 | nr_iovecs * sizeof(struct bio_vec), | |
452 | gfp_mask); | |
453 | front_pad = 0; | |
454 | inline_vecs = nr_iovecs; | |
455 | } else { | |
d8f429e1 JN |
456 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
457 | if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0)) | |
458 | return NULL; | |
df2cb6da KO |
459 | /* |
460 | * generic_make_request() converts recursion to iteration; this | |
461 | * means if we're running beneath it, any bios we allocate and | |
462 | * submit will not be submitted (and thus freed) until after we | |
463 | * return. | |
464 | * | |
465 | * This exposes us to a potential deadlock if we allocate | |
466 | * multiple bios from the same bio_set() while running | |
467 | * underneath generic_make_request(). If we were to allocate | |
468 | * multiple bios (say a stacking block driver that was splitting | |
469 | * bios), we would deadlock if we exhausted the mempool's | |
470 | * reserve. | |
471 | * | |
472 | * We solve this, and guarantee forward progress, with a rescuer | |
473 | * workqueue per bio_set. If we go to allocate and there are | |
474 | * bios on current->bio_list, we first try the allocation | |
d0164adc MG |
475 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
476 | * bios we would be blocking to the rescuer workqueue before | |
477 | * we retry with the original gfp_flags. | |
df2cb6da KO |
478 | */ |
479 | ||
f5fe1b51 N |
480 | if (current->bio_list && |
481 | (!bio_list_empty(¤t->bio_list[0]) || | |
47e0fb46 N |
482 | !bio_list_empty(¤t->bio_list[1])) && |
483 | bs->rescue_workqueue) | |
d0164adc | 484 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
df2cb6da | 485 | |
3f86a82a | 486 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
df2cb6da KO |
487 | if (!p && gfp_mask != saved_gfp) { |
488 | punt_bios_to_rescuer(bs); | |
489 | gfp_mask = saved_gfp; | |
490 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
491 | } | |
492 | ||
3f86a82a KO |
493 | front_pad = bs->front_pad; |
494 | inline_vecs = BIO_INLINE_VECS; | |
495 | } | |
496 | ||
451a9ebf TH |
497 | if (unlikely(!p)) |
498 | return NULL; | |
1da177e4 | 499 | |
3f86a82a | 500 | bio = p + front_pad; |
3a83f467 | 501 | bio_init(bio, NULL, 0); |
34053979 | 502 | |
3f86a82a | 503 | if (nr_iovecs > inline_vecs) { |
ed996a52 CH |
504 | unsigned long idx = 0; |
505 | ||
9f060e22 | 506 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
507 | if (!bvl && gfp_mask != saved_gfp) { |
508 | punt_bios_to_rescuer(bs); | |
509 | gfp_mask = saved_gfp; | |
9f060e22 | 510 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
511 | } |
512 | ||
34053979 IM |
513 | if (unlikely(!bvl)) |
514 | goto err_free; | |
a38352e0 | 515 | |
ed996a52 | 516 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
3f86a82a KO |
517 | } else if (nr_iovecs) { |
518 | bvl = bio->bi_inline_vecs; | |
1da177e4 | 519 | } |
3f86a82a KO |
520 | |
521 | bio->bi_pool = bs; | |
34053979 | 522 | bio->bi_max_vecs = nr_iovecs; |
34053979 | 523 | bio->bi_io_vec = bvl; |
1da177e4 | 524 | return bio; |
34053979 IM |
525 | |
526 | err_free: | |
451a9ebf | 527 | mempool_free(p, bs->bio_pool); |
34053979 | 528 | return NULL; |
1da177e4 | 529 | } |
a112a71d | 530 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 531 | |
1da177e4 LT |
532 | void zero_fill_bio(struct bio *bio) |
533 | { | |
534 | unsigned long flags; | |
7988613b KO |
535 | struct bio_vec bv; |
536 | struct bvec_iter iter; | |
1da177e4 | 537 | |
7988613b KO |
538 | bio_for_each_segment(bv, bio, iter) { |
539 | char *data = bvec_kmap_irq(&bv, &flags); | |
540 | memset(data, 0, bv.bv_len); | |
541 | flush_dcache_page(bv.bv_page); | |
1da177e4 LT |
542 | bvec_kunmap_irq(data, &flags); |
543 | } | |
544 | } | |
545 | EXPORT_SYMBOL(zero_fill_bio); | |
546 | ||
547 | /** | |
548 | * bio_put - release a reference to a bio | |
549 | * @bio: bio to release reference to | |
550 | * | |
551 | * Description: | |
552 | * Put a reference to a &struct bio, either one you have gotten with | |
9b10f6a9 | 553 | * bio_alloc, bio_get or bio_clone_*. The last put of a bio will free it. |
1da177e4 LT |
554 | **/ |
555 | void bio_put(struct bio *bio) | |
556 | { | |
dac56212 | 557 | if (!bio_flagged(bio, BIO_REFFED)) |
4254bba1 | 558 | bio_free(bio); |
dac56212 JA |
559 | else { |
560 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); | |
561 | ||
562 | /* | |
563 | * last put frees it | |
564 | */ | |
565 | if (atomic_dec_and_test(&bio->__bi_cnt)) | |
566 | bio_free(bio); | |
567 | } | |
1da177e4 | 568 | } |
a112a71d | 569 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 570 | |
165125e1 | 571 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
572 | { |
573 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
574 | blk_recount_segments(q, bio); | |
575 | ||
576 | return bio->bi_phys_segments; | |
577 | } | |
a112a71d | 578 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 579 | |
59d276fe KO |
580 | /** |
581 | * __bio_clone_fast - clone a bio that shares the original bio's biovec | |
582 | * @bio: destination bio | |
583 | * @bio_src: bio to clone | |
584 | * | |
585 | * Clone a &bio. Caller will own the returned bio, but not | |
586 | * the actual data it points to. Reference count of returned | |
587 | * bio will be one. | |
588 | * | |
589 | * Caller must ensure that @bio_src is not freed before @bio. | |
590 | */ | |
591 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) | |
592 | { | |
ed996a52 | 593 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
59d276fe KO |
594 | |
595 | /* | |
74d46992 | 596 | * most users will be overriding ->bi_disk with a new target, |
59d276fe KO |
597 | * so we don't set nor calculate new physical/hw segment counts here |
598 | */ | |
74d46992 | 599 | bio->bi_disk = bio_src->bi_disk; |
5d62da3a | 600 | bio->bi_partno = bio_src->bi_partno; |
b7c44ed9 | 601 | bio_set_flag(bio, BIO_CLONED); |
3ef1c33f SL |
602 | if (bio_flagged(bio_src, BIO_THROTTLED)) |
603 | bio_set_flag(bio, BIO_THROTTLED); | |
1eff9d32 | 604 | bio->bi_opf = bio_src->bi_opf; |
49e3bbe3 BK |
605 | #ifdef CONFIG_CRYPTO_DISKCIPHER_DUN |
606 | bio->bi_iter.bi_dun = bio_src->bi_iter.bi_dun; | |
607 | #endif | |
cb6934f8 | 608 | bio->bi_write_hint = bio_src->bi_write_hint; |
59d276fe KO |
609 | bio->bi_iter = bio_src->bi_iter; |
610 | bio->bi_io_vec = bio_src->bi_io_vec; | |
03c98711 | 611 | bio->bi_aux_private = bio_src->bi_aux_private; |
20bd723e PV |
612 | |
613 | bio_clone_blkcg_association(bio, bio_src); | |
59d276fe KO |
614 | } |
615 | EXPORT_SYMBOL(__bio_clone_fast); | |
616 | ||
617 | /** | |
618 | * bio_clone_fast - clone a bio that shares the original bio's biovec | |
619 | * @bio: bio to clone | |
620 | * @gfp_mask: allocation priority | |
621 | * @bs: bio_set to allocate from | |
622 | * | |
623 | * Like __bio_clone_fast, only also allocates the returned bio | |
624 | */ | |
625 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) | |
626 | { | |
627 | struct bio *b; | |
628 | ||
629 | b = bio_alloc_bioset(gfp_mask, 0, bs); | |
630 | if (!b) | |
631 | return NULL; | |
632 | ||
633 | __bio_clone_fast(b, bio); | |
634 | ||
635 | if (bio_integrity(bio)) { | |
636 | int ret; | |
637 | ||
638 | ret = bio_integrity_clone(b, bio, gfp_mask); | |
639 | ||
640 | if (ret < 0) { | |
641 | bio_put(b); | |
642 | return NULL; | |
643 | } | |
644 | } | |
645 | ||
646 | return b; | |
647 | } | |
648 | EXPORT_SYMBOL(bio_clone_fast); | |
649 | ||
f4595875 SL |
650 | /** |
651 | * bio_clone_bioset - clone a bio | |
652 | * @bio_src: bio to clone | |
653 | * @gfp_mask: allocation priority | |
654 | * @bs: bio_set to allocate from | |
655 | * | |
656 | * Clone bio. Caller will own the returned bio, but not the actual data it | |
657 | * points to. Reference count of returned bio will be one. | |
658 | */ | |
659 | struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask, | |
660 | struct bio_set *bs) | |
1da177e4 | 661 | { |
bdb53207 KO |
662 | struct bvec_iter iter; |
663 | struct bio_vec bv; | |
664 | struct bio *bio; | |
1da177e4 | 665 | |
bdb53207 KO |
666 | /* |
667 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from | |
668 | * bio_src->bi_io_vec to bio->bi_io_vec. | |
669 | * | |
670 | * We can't do that anymore, because: | |
671 | * | |
672 | * - The point of cloning the biovec is to produce a bio with a biovec | |
673 | * the caller can modify: bi_idx and bi_bvec_done should be 0. | |
674 | * | |
675 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if | |
676 | * we tried to clone the whole thing bio_alloc_bioset() would fail. | |
677 | * But the clone should succeed as long as the number of biovecs we | |
678 | * actually need to allocate is fewer than BIO_MAX_PAGES. | |
679 | * | |
680 | * - Lastly, bi_vcnt should not be looked at or relied upon by code | |
681 | * that does not own the bio - reason being drivers don't use it for | |
682 | * iterating over the biovec anymore, so expecting it to be kept up | |
683 | * to date (i.e. for clones that share the parent biovec) is just | |
684 | * asking for trouble and would force extra work on | |
685 | * __bio_clone_fast() anyways. | |
686 | */ | |
687 | ||
f4595875 | 688 | bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs); |
bdb53207 | 689 | if (!bio) |
7ba1ba12 | 690 | return NULL; |
74d46992 | 691 | bio->bi_disk = bio_src->bi_disk; |
1eff9d32 | 692 | bio->bi_opf = bio_src->bi_opf; |
cb6934f8 | 693 | bio->bi_write_hint = bio_src->bi_write_hint; |
bdb53207 KO |
694 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; |
695 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; | |
03c98711 | 696 | bio->bi_aux_private = bio_src->bi_aux_private; |
7ba1ba12 | 697 | |
7afafc8a AH |
698 | switch (bio_op(bio)) { |
699 | case REQ_OP_DISCARD: | |
700 | case REQ_OP_SECURE_ERASE: | |
a6f0788e | 701 | case REQ_OP_WRITE_ZEROES: |
7afafc8a AH |
702 | break; |
703 | case REQ_OP_WRITE_SAME: | |
8423ae3d | 704 | bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0]; |
7afafc8a AH |
705 | break; |
706 | default: | |
f4595875 | 707 | bio_for_each_segment(bv, bio_src, iter) |
7afafc8a AH |
708 | bio->bi_io_vec[bio->bi_vcnt++] = bv; |
709 | break; | |
8423ae3d KO |
710 | } |
711 | ||
bdb53207 KO |
712 | if (bio_integrity(bio_src)) { |
713 | int ret; | |
7ba1ba12 | 714 | |
bdb53207 | 715 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
059ea331 | 716 | if (ret < 0) { |
bdb53207 | 717 | bio_put(bio); |
7ba1ba12 | 718 | return NULL; |
059ea331 | 719 | } |
3676347a | 720 | } |
1da177e4 | 721 | |
49e3bbe3 BK |
722 | #ifdef CONFIG_CRYPTO_DISKCIPHER_DUN |
723 | bio->bi_iter.bi_dun = bio_src->bi_iter.bi_dun; | |
724 | #endif | |
20bd723e PV |
725 | bio_clone_blkcg_association(bio, bio_src); |
726 | ||
bdb53207 | 727 | return bio; |
1da177e4 | 728 | } |
bf800ef1 | 729 | EXPORT_SYMBOL(bio_clone_bioset); |
1da177e4 LT |
730 | |
731 | /** | |
c66a14d0 KO |
732 | * bio_add_pc_page - attempt to add page to bio |
733 | * @q: the target queue | |
734 | * @bio: destination bio | |
735 | * @page: page to add | |
736 | * @len: vec entry length | |
737 | * @offset: vec entry offset | |
1da177e4 | 738 | * |
c66a14d0 KO |
739 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
740 | * number of reasons, such as the bio being full or target block device | |
741 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
742 | * so it is always possible to add a single page to an empty bio. | |
743 | * | |
744 | * This should only be used by REQ_PC bios. | |
1da177e4 | 745 | */ |
c66a14d0 KO |
746 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
747 | *page, unsigned int len, unsigned int offset) | |
1da177e4 LT |
748 | { |
749 | int retried_segments = 0; | |
750 | struct bio_vec *bvec; | |
751 | ||
752 | /* | |
753 | * cloned bio must not modify vec list | |
754 | */ | |
755 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
756 | return 0; | |
757 | ||
c66a14d0 | 758 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
759 | return 0; |
760 | ||
80cfd548 JA |
761 | /* |
762 | * For filesystems with a blocksize smaller than the pagesize | |
763 | * we will often be called with the same page as last time and | |
764 | * a consecutive offset. Optimize this special case. | |
765 | */ | |
766 | if (bio->bi_vcnt > 0) { | |
767 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
768 | ||
769 | if (page == prev->bv_page && | |
770 | offset == prev->bv_offset + prev->bv_len) { | |
771 | prev->bv_len += len; | |
fcbf6a08 | 772 | bio->bi_iter.bi_size += len; |
80cfd548 JA |
773 | goto done; |
774 | } | |
66cb45aa JA |
775 | |
776 | /* | |
777 | * If the queue doesn't support SG gaps and adding this | |
778 | * offset would create a gap, disallow it. | |
779 | */ | |
03100aad | 780 | if (bvec_gap_to_prev(q, prev, offset)) |
66cb45aa | 781 | return 0; |
80cfd548 JA |
782 | } |
783 | ||
784 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
785 | return 0; |
786 | ||
787 | /* | |
fcbf6a08 ML |
788 | * setup the new entry, we might clear it again later if we |
789 | * cannot add the page | |
790 | */ | |
791 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
792 | bvec->bv_page = page; | |
793 | bvec->bv_len = len; | |
794 | bvec->bv_offset = offset; | |
795 | bio->bi_vcnt++; | |
796 | bio->bi_phys_segments++; | |
797 | bio->bi_iter.bi_size += len; | |
798 | ||
799 | /* | |
800 | * Perform a recount if the number of segments is greater | |
801 | * than queue_max_segments(q). | |
1da177e4 LT |
802 | */ |
803 | ||
fcbf6a08 | 804 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
1da177e4 LT |
805 | |
806 | if (retried_segments) | |
fcbf6a08 | 807 | goto failed; |
1da177e4 LT |
808 | |
809 | retried_segments = 1; | |
810 | blk_recount_segments(q, bio); | |
811 | } | |
812 | ||
1da177e4 | 813 | /* If we may be able to merge these biovecs, force a recount */ |
fcbf6a08 | 814 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
b7c44ed9 | 815 | bio_clear_flag(bio, BIO_SEG_VALID); |
1da177e4 | 816 | |
80cfd548 | 817 | done: |
1da177e4 | 818 | return len; |
fcbf6a08 ML |
819 | |
820 | failed: | |
821 | bvec->bv_page = NULL; | |
822 | bvec->bv_len = 0; | |
823 | bvec->bv_offset = 0; | |
824 | bio->bi_vcnt--; | |
825 | bio->bi_iter.bi_size -= len; | |
826 | blk_recount_segments(q, bio); | |
827 | return 0; | |
1da177e4 | 828 | } |
a112a71d | 829 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 830 | |
1da177e4 LT |
831 | /** |
832 | * bio_add_page - attempt to add page to bio | |
833 | * @bio: destination bio | |
834 | * @page: page to add | |
835 | * @len: vec entry length | |
836 | * @offset: vec entry offset | |
837 | * | |
c66a14d0 KO |
838 | * Attempt to add a page to the bio_vec maplist. This will only fail |
839 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. | |
1da177e4 | 840 | */ |
c66a14d0 KO |
841 | int bio_add_page(struct bio *bio, struct page *page, |
842 | unsigned int len, unsigned int offset) | |
1da177e4 | 843 | { |
c66a14d0 KO |
844 | struct bio_vec *bv; |
845 | ||
846 | /* | |
847 | * cloned bio must not modify vec list | |
848 | */ | |
849 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) | |
850 | return 0; | |
762380ad | 851 | |
c66a14d0 KO |
852 | /* |
853 | * For filesystems with a blocksize smaller than the pagesize | |
854 | * we will often be called with the same page as last time and | |
855 | * a consecutive offset. Optimize this special case. | |
856 | */ | |
857 | if (bio->bi_vcnt > 0) { | |
858 | bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
58a4915a | 859 | |
c66a14d0 KO |
860 | if (page == bv->bv_page && |
861 | offset == bv->bv_offset + bv->bv_len) { | |
862 | bv->bv_len += len; | |
863 | goto done; | |
864 | } | |
865 | } | |
866 | ||
867 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
868 | return 0; | |
869 | ||
870 | bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
871 | bv->bv_page = page; | |
872 | bv->bv_len = len; | |
873 | bv->bv_offset = offset; | |
874 | ||
875 | bio->bi_vcnt++; | |
876 | done: | |
877 | bio->bi_iter.bi_size += len; | |
878 | return len; | |
1da177e4 | 879 | } |
a112a71d | 880 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 881 | |
2cefe4db KO |
882 | /** |
883 | * bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio | |
884 | * @bio: bio to add pages to | |
885 | * @iter: iov iterator describing the region to be mapped | |
886 | * | |
887 | * Pins as many pages from *iter and appends them to @bio's bvec array. The | |
888 | * pages will have to be released using put_page() when done. | |
889 | */ | |
890 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) | |
891 | { | |
892 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; | |
893 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; | |
894 | struct page **pages = (struct page **)bv; | |
895 | size_t offset, diff; | |
896 | ssize_t size; | |
897 | ||
898 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); | |
899 | if (unlikely(size <= 0)) | |
900 | return size ? size : -EFAULT; | |
901 | nr_pages = (size + offset + PAGE_SIZE - 1) / PAGE_SIZE; | |
902 | ||
903 | /* | |
904 | * Deep magic below: We need to walk the pinned pages backwards | |
905 | * because we are abusing the space allocated for the bio_vecs | |
906 | * for the page array. Because the bio_vecs are larger than the | |
907 | * page pointers by definition this will always work. But it also | |
908 | * means we can't use bio_add_page, so any changes to it's semantics | |
909 | * need to be reflected here as well. | |
910 | */ | |
911 | bio->bi_iter.bi_size += size; | |
912 | bio->bi_vcnt += nr_pages; | |
913 | ||
914 | diff = (nr_pages * PAGE_SIZE - offset) - size; | |
915 | while (nr_pages--) { | |
916 | bv[nr_pages].bv_page = pages[nr_pages]; | |
917 | bv[nr_pages].bv_len = PAGE_SIZE; | |
918 | bv[nr_pages].bv_offset = 0; | |
919 | } | |
920 | ||
921 | bv[0].bv_offset += offset; | |
922 | bv[0].bv_len -= offset; | |
923 | if (diff) | |
924 | bv[bio->bi_vcnt - 1].bv_len -= diff; | |
925 | ||
926 | iov_iter_advance(iter, size); | |
927 | return 0; | |
928 | } | |
929 | EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages); | |
930 | ||
9e882242 KO |
931 | struct submit_bio_ret { |
932 | struct completion event; | |
933 | int error; | |
934 | }; | |
935 | ||
4246a0b6 | 936 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 KO |
937 | { |
938 | struct submit_bio_ret *ret = bio->bi_private; | |
939 | ||
4e4cbee9 | 940 | ret->error = blk_status_to_errno(bio->bi_status); |
9e882242 KO |
941 | complete(&ret->event); |
942 | } | |
943 | ||
944 | /** | |
945 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
946 | * @bio: The &struct bio which describes the I/O |
947 | * | |
948 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
949 | * bio_endio() on failure. | |
3d289d68 JK |
950 | * |
951 | * WARNING: Unlike to how submit_bio() is usually used, this function does not | |
952 | * result in bio reference to be consumed. The caller must drop the reference | |
953 | * on his own. | |
9e882242 | 954 | */ |
4e49ea4a | 955 | int submit_bio_wait(struct bio *bio) |
9e882242 KO |
956 | { |
957 | struct submit_bio_ret ret; | |
958 | ||
9e882242 KO |
959 | init_completion(&ret.event); |
960 | bio->bi_private = &ret; | |
961 | bio->bi_end_io = submit_bio_wait_endio; | |
1eff9d32 | 962 | bio->bi_opf |= REQ_SYNC; |
4e49ea4a | 963 | submit_bio(bio); |
d57d6115 | 964 | wait_for_completion_io(&ret.event); |
9e882242 KO |
965 | |
966 | return ret.error; | |
967 | } | |
968 | EXPORT_SYMBOL(submit_bio_wait); | |
969 | ||
054bdf64 KO |
970 | /** |
971 | * bio_advance - increment/complete a bio by some number of bytes | |
972 | * @bio: bio to advance | |
973 | * @bytes: number of bytes to complete | |
974 | * | |
975 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
976 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
977 | * be updated on the last bvec as well. | |
978 | * | |
979 | * @bio will then represent the remaining, uncompleted portion of the io. | |
980 | */ | |
981 | void bio_advance(struct bio *bio, unsigned bytes) | |
982 | { | |
983 | if (bio_integrity(bio)) | |
984 | bio_integrity_advance(bio, bytes); | |
985 | ||
4550dd6c | 986 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
987 | } |
988 | EXPORT_SYMBOL(bio_advance); | |
989 | ||
a0787606 KO |
990 | /** |
991 | * bio_alloc_pages - allocates a single page for each bvec in a bio | |
992 | * @bio: bio to allocate pages for | |
993 | * @gfp_mask: flags for allocation | |
994 | * | |
995 | * Allocates pages up to @bio->bi_vcnt. | |
996 | * | |
997 | * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are | |
998 | * freed. | |
999 | */ | |
1000 | int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask) | |
1001 | { | |
1002 | int i; | |
1003 | struct bio_vec *bv; | |
1004 | ||
1005 | bio_for_each_segment_all(bv, bio, i) { | |
1006 | bv->bv_page = alloc_page(gfp_mask); | |
1007 | if (!bv->bv_page) { | |
1008 | while (--bv >= bio->bi_io_vec) | |
1009 | __free_page(bv->bv_page); | |
1010 | return -ENOMEM; | |
1011 | } | |
1012 | } | |
1013 | ||
1014 | return 0; | |
1015 | } | |
1016 | EXPORT_SYMBOL(bio_alloc_pages); | |
1017 | ||
16ac3d63 KO |
1018 | /** |
1019 | * bio_copy_data - copy contents of data buffers from one chain of bios to | |
1020 | * another | |
1021 | * @src: source bio list | |
1022 | * @dst: destination bio list | |
1023 | * | |
1024 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats | |
1025 | * @src and @dst as linked lists of bios. | |
1026 | * | |
1027 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
1028 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
1029 | */ | |
1030 | void bio_copy_data(struct bio *dst, struct bio *src) | |
1031 | { | |
1cb9dda4 KO |
1032 | struct bvec_iter src_iter, dst_iter; |
1033 | struct bio_vec src_bv, dst_bv; | |
16ac3d63 | 1034 | void *src_p, *dst_p; |
1cb9dda4 | 1035 | unsigned bytes; |
16ac3d63 | 1036 | |
1cb9dda4 KO |
1037 | src_iter = src->bi_iter; |
1038 | dst_iter = dst->bi_iter; | |
16ac3d63 KO |
1039 | |
1040 | while (1) { | |
1cb9dda4 KO |
1041 | if (!src_iter.bi_size) { |
1042 | src = src->bi_next; | |
1043 | if (!src) | |
1044 | break; | |
16ac3d63 | 1045 | |
1cb9dda4 | 1046 | src_iter = src->bi_iter; |
16ac3d63 KO |
1047 | } |
1048 | ||
1cb9dda4 KO |
1049 | if (!dst_iter.bi_size) { |
1050 | dst = dst->bi_next; | |
1051 | if (!dst) | |
1052 | break; | |
16ac3d63 | 1053 | |
1cb9dda4 | 1054 | dst_iter = dst->bi_iter; |
16ac3d63 KO |
1055 | } |
1056 | ||
1cb9dda4 KO |
1057 | src_bv = bio_iter_iovec(src, src_iter); |
1058 | dst_bv = bio_iter_iovec(dst, dst_iter); | |
1059 | ||
1060 | bytes = min(src_bv.bv_len, dst_bv.bv_len); | |
16ac3d63 | 1061 | |
1cb9dda4 KO |
1062 | src_p = kmap_atomic(src_bv.bv_page); |
1063 | dst_p = kmap_atomic(dst_bv.bv_page); | |
16ac3d63 | 1064 | |
1cb9dda4 KO |
1065 | memcpy(dst_p + dst_bv.bv_offset, |
1066 | src_p + src_bv.bv_offset, | |
16ac3d63 KO |
1067 | bytes); |
1068 | ||
1069 | kunmap_atomic(dst_p); | |
1070 | kunmap_atomic(src_p); | |
1071 | ||
1cb9dda4 KO |
1072 | bio_advance_iter(src, &src_iter, bytes); |
1073 | bio_advance_iter(dst, &dst_iter, bytes); | |
16ac3d63 KO |
1074 | } |
1075 | } | |
1076 | EXPORT_SYMBOL(bio_copy_data); | |
1077 | ||
1da177e4 | 1078 | struct bio_map_data { |
152e283f | 1079 | int is_our_pages; |
26e49cfc KO |
1080 | struct iov_iter iter; |
1081 | struct iovec iov[]; | |
1da177e4 LT |
1082 | }; |
1083 | ||
7410b3c6 | 1084 | static struct bio_map_data *bio_alloc_map_data(unsigned int iov_count, |
76029ff3 | 1085 | gfp_t gfp_mask) |
1da177e4 | 1086 | { |
f3f63c1c JA |
1087 | if (iov_count > UIO_MAXIOV) |
1088 | return NULL; | |
1da177e4 | 1089 | |
c8db4448 | 1090 | return kmalloc(sizeof(struct bio_map_data) + |
26e49cfc | 1091 | sizeof(struct iovec) * iov_count, gfp_mask); |
1da177e4 LT |
1092 | } |
1093 | ||
9124d3fe DP |
1094 | /** |
1095 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1096 | * @bio: The &struct bio which describes the I/O as destination | |
1097 | * @iter: iov_iter as source | |
1098 | * | |
1099 | * Copy all pages from iov_iter to bio. | |
1100 | * Returns 0 on success, or error on failure. | |
1101 | */ | |
1102 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter iter) | |
c5dec1c3 | 1103 | { |
9124d3fe | 1104 | int i; |
c5dec1c3 | 1105 | struct bio_vec *bvec; |
c5dec1c3 | 1106 | |
d74c6d51 | 1107 | bio_for_each_segment_all(bvec, bio, i) { |
9124d3fe | 1108 | ssize_t ret; |
c5dec1c3 | 1109 | |
9124d3fe DP |
1110 | ret = copy_page_from_iter(bvec->bv_page, |
1111 | bvec->bv_offset, | |
1112 | bvec->bv_len, | |
1113 | &iter); | |
1114 | ||
1115 | if (!iov_iter_count(&iter)) | |
1116 | break; | |
1117 | ||
1118 | if (ret < bvec->bv_len) | |
1119 | return -EFAULT; | |
c5dec1c3 FT |
1120 | } |
1121 | ||
9124d3fe DP |
1122 | return 0; |
1123 | } | |
1124 | ||
1125 | /** | |
1126 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1127 | * @bio: The &struct bio which describes the I/O as source | |
1128 | * @iter: iov_iter as destination | |
1129 | * | |
1130 | * Copy all pages from bio to iov_iter. | |
1131 | * Returns 0 on success, or error on failure. | |
1132 | */ | |
1133 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1134 | { | |
1135 | int i; | |
1136 | struct bio_vec *bvec; | |
1137 | ||
1138 | bio_for_each_segment_all(bvec, bio, i) { | |
1139 | ssize_t ret; | |
1140 | ||
1141 | ret = copy_page_to_iter(bvec->bv_page, | |
1142 | bvec->bv_offset, | |
1143 | bvec->bv_len, | |
1144 | &iter); | |
1145 | ||
1146 | if (!iov_iter_count(&iter)) | |
1147 | break; | |
1148 | ||
1149 | if (ret < bvec->bv_len) | |
1150 | return -EFAULT; | |
1151 | } | |
1152 | ||
1153 | return 0; | |
c5dec1c3 FT |
1154 | } |
1155 | ||
491221f8 | 1156 | void bio_free_pages(struct bio *bio) |
1dfa0f68 CH |
1157 | { |
1158 | struct bio_vec *bvec; | |
1159 | int i; | |
1160 | ||
1161 | bio_for_each_segment_all(bvec, bio, i) | |
1162 | __free_page(bvec->bv_page); | |
1163 | } | |
491221f8 | 1164 | EXPORT_SYMBOL(bio_free_pages); |
1dfa0f68 | 1165 | |
1da177e4 LT |
1166 | /** |
1167 | * bio_uncopy_user - finish previously mapped bio | |
1168 | * @bio: bio being terminated | |
1169 | * | |
ddad8dd0 | 1170 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1171 | * to user space in case of a read. |
1172 | */ | |
1173 | int bio_uncopy_user(struct bio *bio) | |
1174 | { | |
1175 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1176 | int ret = 0; |
1da177e4 | 1177 | |
35dc2483 RD |
1178 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1179 | /* | |
1180 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1181 | * don't copy into a random user address space, just free |
1182 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1183 | */ |
2d99b55d HR |
1184 | if (!current->mm) |
1185 | ret = -EINTR; | |
1186 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1187 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1188 | if (bmd->is_our_pages) |
1189 | bio_free_pages(bio); | |
35dc2483 | 1190 | } |
c8db4448 | 1191 | kfree(bmd); |
1da177e4 LT |
1192 | bio_put(bio); |
1193 | return ret; | |
1194 | } | |
1195 | ||
1196 | /** | |
c5dec1c3 | 1197 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1198 | * @q: destination block queue |
1199 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1200 | * @iter: iovec iterator | |
1201 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1202 | * |
1203 | * Prepares and returns a bio for indirect user io, bouncing data | |
1204 | * to/from kernel pages as necessary. Must be paired with | |
1205 | * call bio_uncopy_user() on io completion. | |
1206 | */ | |
152e283f FT |
1207 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1208 | struct rq_map_data *map_data, | |
26e49cfc KO |
1209 | const struct iov_iter *iter, |
1210 | gfp_t gfp_mask) | |
1da177e4 | 1211 | { |
1da177e4 | 1212 | struct bio_map_data *bmd; |
1da177e4 LT |
1213 | struct page *page; |
1214 | struct bio *bio; | |
1215 | int i, ret; | |
c5dec1c3 | 1216 | int nr_pages = 0; |
26e49cfc | 1217 | unsigned int len = iter->count; |
bd5cecea | 1218 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1219 | |
26e49cfc | 1220 | for (i = 0; i < iter->nr_segs; i++) { |
c5dec1c3 FT |
1221 | unsigned long uaddr; |
1222 | unsigned long end; | |
1223 | unsigned long start; | |
1224 | ||
26e49cfc KO |
1225 | uaddr = (unsigned long) iter->iov[i].iov_base; |
1226 | end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1) | |
1227 | >> PAGE_SHIFT; | |
c5dec1c3 FT |
1228 | start = uaddr >> PAGE_SHIFT; |
1229 | ||
cb4644ca JA |
1230 | /* |
1231 | * Overflow, abort | |
1232 | */ | |
1233 | if (end < start) | |
1234 | return ERR_PTR(-EINVAL); | |
1235 | ||
c5dec1c3 | 1236 | nr_pages += end - start; |
c5dec1c3 FT |
1237 | } |
1238 | ||
69838727 FT |
1239 | if (offset) |
1240 | nr_pages++; | |
1241 | ||
26e49cfc | 1242 | bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask); |
1da177e4 LT |
1243 | if (!bmd) |
1244 | return ERR_PTR(-ENOMEM); | |
1245 | ||
26e49cfc KO |
1246 | /* |
1247 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1248 | * The caller provided iov might point to an on-stack or otherwise | |
1249 | * shortlived one. | |
1250 | */ | |
1251 | bmd->is_our_pages = map_data ? 0 : 1; | |
1252 | memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs); | |
1cfd0ddd AV |
1253 | bmd->iter = *iter; |
1254 | bmd->iter.iov = bmd->iov; | |
26e49cfc | 1255 | |
1da177e4 | 1256 | ret = -ENOMEM; |
a9e9dc24 | 1257 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1258 | if (!bio) |
1259 | goto out_bmd; | |
1260 | ||
1da177e4 | 1261 | ret = 0; |
56c451f4 FT |
1262 | |
1263 | if (map_data) { | |
e623ddb4 | 1264 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1265 | i = map_data->offset / PAGE_SIZE; |
1266 | } | |
1da177e4 | 1267 | while (len) { |
e623ddb4 | 1268 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1269 | |
56c451f4 FT |
1270 | bytes -= offset; |
1271 | ||
1da177e4 LT |
1272 | if (bytes > len) |
1273 | bytes = len; | |
1274 | ||
152e283f | 1275 | if (map_data) { |
e623ddb4 | 1276 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1277 | ret = -ENOMEM; |
1278 | break; | |
1279 | } | |
e623ddb4 FT |
1280 | |
1281 | page = map_data->pages[i / nr_pages]; | |
1282 | page += (i % nr_pages); | |
1283 | ||
1284 | i++; | |
1285 | } else { | |
152e283f | 1286 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1287 | if (!page) { |
1288 | ret = -ENOMEM; | |
1289 | break; | |
1290 | } | |
1da177e4 LT |
1291 | } |
1292 | ||
56c451f4 | 1293 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 1294 | break; |
1da177e4 LT |
1295 | |
1296 | len -= bytes; | |
56c451f4 | 1297 | offset = 0; |
1da177e4 LT |
1298 | } |
1299 | ||
1300 | if (ret) | |
1301 | goto cleanup; | |
1302 | ||
1303 | /* | |
1304 | * success | |
1305 | */ | |
26e49cfc | 1306 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1307 | (map_data && map_data->from_user)) { |
9124d3fe | 1308 | ret = bio_copy_from_iter(bio, *iter); |
c5dec1c3 FT |
1309 | if (ret) |
1310 | goto cleanup; | |
1da177e4 LT |
1311 | } |
1312 | ||
26e49cfc | 1313 | bio->bi_private = bmd; |
1da177e4 LT |
1314 | return bio; |
1315 | cleanup: | |
152e283f | 1316 | if (!map_data) |
1dfa0f68 | 1317 | bio_free_pages(bio); |
1da177e4 LT |
1318 | bio_put(bio); |
1319 | out_bmd: | |
c8db4448 | 1320 | kfree(bmd); |
1da177e4 LT |
1321 | return ERR_PTR(ret); |
1322 | } | |
1323 | ||
37f19e57 CH |
1324 | /** |
1325 | * bio_map_user_iov - map user iovec into bio | |
1326 | * @q: the struct request_queue for the bio | |
1327 | * @iter: iovec iterator | |
1328 | * @gfp_mask: memory allocation flags | |
1329 | * | |
1330 | * Map the user space address into a bio suitable for io to a block | |
1331 | * device. Returns an error pointer in case of error. | |
1332 | */ | |
1333 | struct bio *bio_map_user_iov(struct request_queue *q, | |
1334 | const struct iov_iter *iter, | |
1335 | gfp_t gfp_mask) | |
1da177e4 | 1336 | { |
26e49cfc | 1337 | int j; |
f1970baf | 1338 | int nr_pages = 0; |
1da177e4 LT |
1339 | struct page **pages; |
1340 | struct bio *bio; | |
f1970baf JB |
1341 | int cur_page = 0; |
1342 | int ret, offset; | |
26e49cfc KO |
1343 | struct iov_iter i; |
1344 | struct iovec iov; | |
2b04e8f6 | 1345 | struct bio_vec *bvec; |
1da177e4 | 1346 | |
26e49cfc KO |
1347 | iov_for_each(iov, i, *iter) { |
1348 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1349 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1350 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1351 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1352 | ||
cb4644ca JA |
1353 | /* |
1354 | * Overflow, abort | |
1355 | */ | |
1356 | if (end < start) | |
1357 | return ERR_PTR(-EINVAL); | |
1358 | ||
f1970baf JB |
1359 | nr_pages += end - start; |
1360 | /* | |
a441b0d0 | 1361 | * buffer must be aligned to at least logical block size for now |
f1970baf | 1362 | */ |
ad2d7225 | 1363 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
1364 | return ERR_PTR(-EINVAL); |
1365 | } | |
1366 | ||
1367 | if (!nr_pages) | |
1da177e4 LT |
1368 | return ERR_PTR(-EINVAL); |
1369 | ||
a9e9dc24 | 1370 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1371 | if (!bio) |
1372 | return ERR_PTR(-ENOMEM); | |
1373 | ||
1374 | ret = -ENOMEM; | |
a3bce90e | 1375 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
1376 | if (!pages) |
1377 | goto out; | |
1378 | ||
26e49cfc KO |
1379 | iov_for_each(iov, i, *iter) { |
1380 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1381 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1382 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1383 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1384 | const int local_nr_pages = end - start; | |
1385 | const int page_limit = cur_page + local_nr_pages; | |
cb4644ca | 1386 | |
f5dd33c4 | 1387 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
26e49cfc KO |
1388 | (iter->type & WRITE) != WRITE, |
1389 | &pages[cur_page]); | |
2b04e8f6 AV |
1390 | if (unlikely(ret < local_nr_pages)) { |
1391 | for (j = cur_page; j < page_limit; j++) { | |
1392 | if (!pages[j]) | |
1393 | break; | |
1394 | put_page(pages[j]); | |
1395 | } | |
99172157 | 1396 | ret = -EFAULT; |
f1970baf | 1397 | goto out_unmap; |
99172157 | 1398 | } |
f1970baf | 1399 | |
bd5cecea | 1400 | offset = offset_in_page(uaddr); |
f1970baf JB |
1401 | for (j = cur_page; j < page_limit; j++) { |
1402 | unsigned int bytes = PAGE_SIZE - offset; | |
95d78c28 | 1403 | unsigned short prev_bi_vcnt = bio->bi_vcnt; |
f1970baf JB |
1404 | |
1405 | if (len <= 0) | |
1406 | break; | |
1407 | ||
1408 | if (bytes > len) | |
1409 | bytes = len; | |
1410 | ||
1411 | /* | |
1412 | * sorry... | |
1413 | */ | |
defd94b7 MC |
1414 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1415 | bytes) | |
f1970baf JB |
1416 | break; |
1417 | ||
95d78c28 VM |
1418 | /* |
1419 | * check if vector was merged with previous | |
1420 | * drop page reference if needed | |
1421 | */ | |
1422 | if (bio->bi_vcnt == prev_bi_vcnt) | |
1423 | put_page(pages[j]); | |
1424 | ||
f1970baf JB |
1425 | len -= bytes; |
1426 | offset = 0; | |
1427 | } | |
1da177e4 | 1428 | |
f1970baf | 1429 | cur_page = j; |
1da177e4 | 1430 | /* |
f1970baf | 1431 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1432 | */ |
f1970baf | 1433 | while (j < page_limit) |
09cbfeaf | 1434 | put_page(pages[j++]); |
1da177e4 LT |
1435 | } |
1436 | ||
1da177e4 LT |
1437 | kfree(pages); |
1438 | ||
b7c44ed9 | 1439 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1440 | |
1441 | /* | |
5fad1b64 | 1442 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
37f19e57 CH |
1443 | * it would normally disappear when its bi_end_io is run. |
1444 | * however, we need it for the unmap, so grab an extra | |
1445 | * reference to it | |
1446 | */ | |
1447 | bio_get(bio); | |
1da177e4 | 1448 | return bio; |
f1970baf JB |
1449 | |
1450 | out_unmap: | |
2b04e8f6 AV |
1451 | bio_for_each_segment_all(bvec, bio, j) { |
1452 | put_page(bvec->bv_page); | |
f1970baf JB |
1453 | } |
1454 | out: | |
1da177e4 LT |
1455 | kfree(pages); |
1456 | bio_put(bio); | |
1457 | return ERR_PTR(ret); | |
1458 | } | |
1459 | ||
1da177e4 LT |
1460 | static void __bio_unmap_user(struct bio *bio) |
1461 | { | |
1462 | struct bio_vec *bvec; | |
1463 | int i; | |
1464 | ||
1465 | /* | |
1466 | * make sure we dirty pages we wrote to | |
1467 | */ | |
d74c6d51 | 1468 | bio_for_each_segment_all(bvec, bio, i) { |
1da177e4 LT |
1469 | if (bio_data_dir(bio) == READ) |
1470 | set_page_dirty_lock(bvec->bv_page); | |
1471 | ||
09cbfeaf | 1472 | put_page(bvec->bv_page); |
1da177e4 LT |
1473 | } |
1474 | ||
1475 | bio_put(bio); | |
1476 | } | |
1477 | ||
1478 | /** | |
1479 | * bio_unmap_user - unmap a bio | |
1480 | * @bio: the bio being unmapped | |
1481 | * | |
5fad1b64 BVA |
1482 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
1483 | * process context. | |
1da177e4 LT |
1484 | * |
1485 | * bio_unmap_user() may sleep. | |
1486 | */ | |
1487 | void bio_unmap_user(struct bio *bio) | |
1488 | { | |
1489 | __bio_unmap_user(bio); | |
1490 | bio_put(bio); | |
1491 | } | |
1492 | ||
4246a0b6 | 1493 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1494 | { |
b823825e | 1495 | bio_put(bio); |
b823825e JA |
1496 | } |
1497 | ||
75c72b83 CH |
1498 | /** |
1499 | * bio_map_kern - map kernel address into bio | |
1500 | * @q: the struct request_queue for the bio | |
1501 | * @data: pointer to buffer to map | |
1502 | * @len: length in bytes | |
1503 | * @gfp_mask: allocation flags for bio allocation | |
1504 | * | |
1505 | * Map the kernel address into a bio suitable for io to a block | |
1506 | * device. Returns an error pointer in case of error. | |
1507 | */ | |
1508 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1509 | gfp_t gfp_mask) | |
df46b9a4 MC |
1510 | { |
1511 | unsigned long kaddr = (unsigned long)data; | |
1512 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1513 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1514 | const int nr_pages = end - start; | |
1515 | int offset, i; | |
1516 | struct bio *bio; | |
1517 | ||
a9e9dc24 | 1518 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1519 | if (!bio) |
1520 | return ERR_PTR(-ENOMEM); | |
1521 | ||
1522 | offset = offset_in_page(kaddr); | |
1523 | for (i = 0; i < nr_pages; i++) { | |
1524 | unsigned int bytes = PAGE_SIZE - offset; | |
1525 | ||
1526 | if (len <= 0) | |
1527 | break; | |
1528 | ||
1529 | if (bytes > len) | |
1530 | bytes = len; | |
1531 | ||
defd94b7 | 1532 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1533 | offset) < bytes) { |
1534 | /* we don't support partial mappings */ | |
1535 | bio_put(bio); | |
1536 | return ERR_PTR(-EINVAL); | |
1537 | } | |
df46b9a4 MC |
1538 | |
1539 | data += bytes; | |
1540 | len -= bytes; | |
1541 | offset = 0; | |
1542 | } | |
1543 | ||
b823825e | 1544 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1545 | return bio; |
1546 | } | |
a112a71d | 1547 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1548 | |
4246a0b6 | 1549 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1550 | { |
1dfa0f68 CH |
1551 | bio_free_pages(bio); |
1552 | bio_put(bio); | |
1553 | } | |
1554 | ||
4246a0b6 | 1555 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1556 | { |
42d2683a | 1557 | char *p = bio->bi_private; |
1dfa0f68 | 1558 | struct bio_vec *bvec; |
68154e90 FT |
1559 | int i; |
1560 | ||
d74c6d51 | 1561 | bio_for_each_segment_all(bvec, bio, i) { |
1dfa0f68 | 1562 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1563 | p += bvec->bv_len; |
68154e90 FT |
1564 | } |
1565 | ||
4246a0b6 | 1566 | bio_copy_kern_endio(bio); |
68154e90 FT |
1567 | } |
1568 | ||
1569 | /** | |
1570 | * bio_copy_kern - copy kernel address into bio | |
1571 | * @q: the struct request_queue for the bio | |
1572 | * @data: pointer to buffer to copy | |
1573 | * @len: length in bytes | |
1574 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1575 | * @reading: data direction is READ |
68154e90 FT |
1576 | * |
1577 | * copy the kernel address into a bio suitable for io to a block | |
1578 | * device. Returns an error pointer in case of error. | |
1579 | */ | |
1580 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1581 | gfp_t gfp_mask, int reading) | |
1582 | { | |
42d2683a CH |
1583 | unsigned long kaddr = (unsigned long)data; |
1584 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1585 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1586 | struct bio *bio; |
1587 | void *p = data; | |
1dfa0f68 | 1588 | int nr_pages = 0; |
68154e90 | 1589 | |
42d2683a CH |
1590 | /* |
1591 | * Overflow, abort | |
1592 | */ | |
1593 | if (end < start) | |
1594 | return ERR_PTR(-EINVAL); | |
68154e90 | 1595 | |
42d2683a CH |
1596 | nr_pages = end - start; |
1597 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1598 | if (!bio) | |
1599 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1600 | |
42d2683a CH |
1601 | while (len) { |
1602 | struct page *page; | |
1603 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1604 | |
42d2683a CH |
1605 | if (bytes > len) |
1606 | bytes = len; | |
1607 | ||
1608 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1609 | if (!page) | |
1610 | goto cleanup; | |
1611 | ||
1612 | if (!reading) | |
1613 | memcpy(page_address(page), p, bytes); | |
1614 | ||
1615 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1616 | break; | |
1617 | ||
1618 | len -= bytes; | |
1619 | p += bytes; | |
68154e90 FT |
1620 | } |
1621 | ||
1dfa0f68 CH |
1622 | if (reading) { |
1623 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1624 | bio->bi_private = data; | |
1625 | } else { | |
1626 | bio->bi_end_io = bio_copy_kern_endio; | |
1dfa0f68 | 1627 | } |
76029ff3 | 1628 | |
68154e90 | 1629 | return bio; |
42d2683a CH |
1630 | |
1631 | cleanup: | |
1dfa0f68 | 1632 | bio_free_pages(bio); |
42d2683a CH |
1633 | bio_put(bio); |
1634 | return ERR_PTR(-ENOMEM); | |
68154e90 FT |
1635 | } |
1636 | ||
1da177e4 LT |
1637 | /* |
1638 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1639 | * for performing direct-IO in BIOs. | |
1640 | * | |
1641 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1642 | * because the required locks are not interrupt-safe. So what we can do is to | |
1643 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1644 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1645 | * in process context. | |
1646 | * | |
1647 | * We special-case compound pages here: normally this means reads into hugetlb | |
1648 | * pages. The logic in here doesn't really work right for compound pages | |
1649 | * because the VM does not uniformly chase down the head page in all cases. | |
1650 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1651 | * handle them at all. So we skip compound pages here at an early stage. | |
1652 | * | |
1653 | * Note that this code is very hard to test under normal circumstances because | |
1654 | * direct-io pins the pages with get_user_pages(). This makes | |
1655 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1656 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1657 | * pagecache. |
1658 | * | |
1659 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1660 | * deferred bio dirtying paths. | |
1661 | */ | |
1662 | ||
1663 | /* | |
1664 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1665 | */ | |
1666 | void bio_set_pages_dirty(struct bio *bio) | |
1667 | { | |
cb34e057 | 1668 | struct bio_vec *bvec; |
1da177e4 LT |
1669 | int i; |
1670 | ||
cb34e057 KO |
1671 | bio_for_each_segment_all(bvec, bio, i) { |
1672 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1673 | |
1674 | if (page && !PageCompound(page)) | |
1675 | set_page_dirty_lock(page); | |
1676 | } | |
1677 | } | |
1678 | ||
86b6c7a7 | 1679 | static void bio_release_pages(struct bio *bio) |
1da177e4 | 1680 | { |
cb34e057 | 1681 | struct bio_vec *bvec; |
1da177e4 LT |
1682 | int i; |
1683 | ||
cb34e057 KO |
1684 | bio_for_each_segment_all(bvec, bio, i) { |
1685 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1686 | |
1687 | if (page) | |
1688 | put_page(page); | |
1689 | } | |
1690 | } | |
1691 | ||
1692 | /* | |
1693 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1694 | * If they are, then fine. If, however, some pages are clean then they must | |
1695 | * have been written out during the direct-IO read. So we take another ref on | |
1696 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1697 | * | |
1698 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1699 | * here on. It will run one put_page() against each page and will run one |
1700 | * bio_put() against the BIO. | |
1da177e4 LT |
1701 | */ |
1702 | ||
65f27f38 | 1703 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1704 | |
65f27f38 | 1705 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1706 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1707 | static struct bio *bio_dirty_list; | |
1708 | ||
1709 | /* | |
1710 | * This runs in process context | |
1711 | */ | |
65f27f38 | 1712 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1713 | { |
1714 | unsigned long flags; | |
1715 | struct bio *bio; | |
1716 | ||
1717 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1718 | bio = bio_dirty_list; | |
1719 | bio_dirty_list = NULL; | |
1720 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1721 | ||
1722 | while (bio) { | |
1723 | struct bio *next = bio->bi_private; | |
1724 | ||
1725 | bio_set_pages_dirty(bio); | |
1726 | bio_release_pages(bio); | |
1727 | bio_put(bio); | |
1728 | bio = next; | |
1729 | } | |
1730 | } | |
1731 | ||
1732 | void bio_check_pages_dirty(struct bio *bio) | |
1733 | { | |
cb34e057 | 1734 | struct bio_vec *bvec; |
1da177e4 LT |
1735 | int nr_clean_pages = 0; |
1736 | int i; | |
1737 | ||
cb34e057 KO |
1738 | bio_for_each_segment_all(bvec, bio, i) { |
1739 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1740 | |
1741 | if (PageDirty(page) || PageCompound(page)) { | |
09cbfeaf | 1742 | put_page(page); |
cb34e057 | 1743 | bvec->bv_page = NULL; |
1da177e4 LT |
1744 | } else { |
1745 | nr_clean_pages++; | |
1746 | } | |
1747 | } | |
1748 | ||
1749 | if (nr_clean_pages) { | |
1750 | unsigned long flags; | |
1751 | ||
1752 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1753 | bio->bi_private = bio_dirty_list; | |
1754 | bio_dirty_list = bio; | |
1755 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1756 | schedule_work(&bio_dirty_work); | |
1757 | } else { | |
1758 | bio_put(bio); | |
1759 | } | |
1760 | } | |
1761 | ||
d62e26b3 JA |
1762 | void generic_start_io_acct(struct request_queue *q, int rw, |
1763 | unsigned long sectors, struct hd_struct *part) | |
394ffa50 GZ |
1764 | { |
1765 | int cpu = part_stat_lock(); | |
1766 | ||
d62e26b3 | 1767 | part_round_stats(q, cpu, part); |
394ffa50 GZ |
1768 | part_stat_inc(cpu, part, ios[rw]); |
1769 | part_stat_add(cpu, part, sectors[rw], sectors); | |
d62e26b3 | 1770 | part_inc_in_flight(q, part, rw); |
394ffa50 GZ |
1771 | |
1772 | part_stat_unlock(); | |
1773 | } | |
1774 | EXPORT_SYMBOL(generic_start_io_acct); | |
1775 | ||
d62e26b3 JA |
1776 | void generic_end_io_acct(struct request_queue *q, int rw, |
1777 | struct hd_struct *part, unsigned long start_time) | |
394ffa50 GZ |
1778 | { |
1779 | unsigned long duration = jiffies - start_time; | |
1780 | int cpu = part_stat_lock(); | |
1781 | ||
1782 | part_stat_add(cpu, part, ticks[rw], duration); | |
d62e26b3 JA |
1783 | part_round_stats(q, cpu, part); |
1784 | part_dec_in_flight(q, part, rw); | |
394ffa50 GZ |
1785 | |
1786 | part_stat_unlock(); | |
1787 | } | |
1788 | EXPORT_SYMBOL(generic_end_io_acct); | |
1789 | ||
2d4dc890 IL |
1790 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1791 | void bio_flush_dcache_pages(struct bio *bi) | |
1792 | { | |
7988613b KO |
1793 | struct bio_vec bvec; |
1794 | struct bvec_iter iter; | |
2d4dc890 | 1795 | |
7988613b KO |
1796 | bio_for_each_segment(bvec, bi, iter) |
1797 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1798 | } |
1799 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1800 | #endif | |
1801 | ||
c4cf5261 JA |
1802 | static inline bool bio_remaining_done(struct bio *bio) |
1803 | { | |
1804 | /* | |
1805 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1806 | * we always end io on the first invocation. | |
1807 | */ | |
1808 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1809 | return true; | |
1810 | ||
1811 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1812 | ||
326e1dbb | 1813 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1814 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1815 | return true; |
326e1dbb | 1816 | } |
c4cf5261 JA |
1817 | |
1818 | return false; | |
1819 | } | |
1820 | ||
1da177e4 LT |
1821 | /** |
1822 | * bio_endio - end I/O on a bio | |
1823 | * @bio: bio | |
1da177e4 LT |
1824 | * |
1825 | * Description: | |
4246a0b6 CH |
1826 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1827 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1828 | * bio unless they own it and thus know that it has an end_io function. | |
fbbaf700 N |
1829 | * |
1830 | * bio_endio() can be called several times on a bio that has been chained | |
1831 | * using bio_chain(). The ->bi_end_io() function will only be called the | |
1832 | * last time. At this point the BLK_TA_COMPLETE tracing event will be | |
1833 | * generated if BIO_TRACE_COMPLETION is set. | |
1da177e4 | 1834 | **/ |
4246a0b6 | 1835 | void bio_endio(struct bio *bio) |
1da177e4 | 1836 | { |
ba8c6967 | 1837 | again: |
2b885517 | 1838 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1839 | return; |
7c20f116 CH |
1840 | if (!bio_integrity_endio(bio)) |
1841 | return; | |
1da177e4 | 1842 | |
ba8c6967 CH |
1843 | /* |
1844 | * Need to have a real endio function for chained bios, otherwise | |
1845 | * various corner cases will break (like stacking block devices that | |
1846 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1847 | * recursion and blowing the stack. Tail call optimization would | |
1848 | * handle this, but compiling with frame pointers also disables | |
1849 | * gcc's sibling call optimization. | |
1850 | */ | |
1851 | if (bio->bi_end_io == bio_chain_endio) { | |
1852 | bio = __bio_chain_endio(bio); | |
1853 | goto again; | |
196d38bc | 1854 | } |
ba8c6967 | 1855 | |
74d46992 CH |
1856 | if (bio->bi_disk && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
1857 | trace_block_bio_complete(bio->bi_disk->queue, bio, | |
a462b950 | 1858 | blk_status_to_errno(bio->bi_status)); |
fbbaf700 N |
1859 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
1860 | } | |
1861 | ||
9e234eea | 1862 | blk_throtl_bio_endio(bio); |
b222dd2f SL |
1863 | /* release cgroup info */ |
1864 | bio_uninit(bio); | |
ba8c6967 CH |
1865 | if (bio->bi_end_io) |
1866 | bio->bi_end_io(bio); | |
1da177e4 | 1867 | } |
a112a71d | 1868 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1869 | |
20d0189b KO |
1870 | /** |
1871 | * bio_split - split a bio | |
1872 | * @bio: bio to split | |
1873 | * @sectors: number of sectors to split from the front of @bio | |
1874 | * @gfp: gfp mask | |
1875 | * @bs: bio set to allocate from | |
1876 | * | |
1877 | * Allocates and returns a new bio which represents @sectors from the start of | |
1878 | * @bio, and updates @bio to represent the remaining sectors. | |
1879 | * | |
f3f5da62 MP |
1880 | * Unless this is a discard request the newly allocated bio will point |
1881 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1882 | * @bio is not freed before the split. | |
20d0189b KO |
1883 | */ |
1884 | struct bio *bio_split(struct bio *bio, int sectors, | |
1885 | gfp_t gfp, struct bio_set *bs) | |
1886 | { | |
1887 | struct bio *split = NULL; | |
1888 | ||
1889 | BUG_ON(sectors <= 0); | |
1890 | BUG_ON(sectors >= bio_sectors(bio)); | |
1891 | ||
f9d03f96 | 1892 | split = bio_clone_fast(bio, gfp, bs); |
20d0189b KO |
1893 | if (!split) |
1894 | return NULL; | |
1895 | ||
1896 | split->bi_iter.bi_size = sectors << 9; | |
1897 | ||
1898 | if (bio_integrity(split)) | |
fbd08e76 | 1899 | bio_integrity_trim(split); |
20d0189b KO |
1900 | |
1901 | bio_advance(bio, split->bi_iter.bi_size); | |
1902 | ||
fbbaf700 | 1903 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
c6c6e38a | 1904 | bio_set_flag(split, BIO_TRACE_COMPLETION); |
fbbaf700 | 1905 | |
20d0189b KO |
1906 | return split; |
1907 | } | |
1908 | EXPORT_SYMBOL(bio_split); | |
1909 | ||
6678d83f KO |
1910 | /** |
1911 | * bio_trim - trim a bio | |
1912 | * @bio: bio to trim | |
1913 | * @offset: number of sectors to trim from the front of @bio | |
1914 | * @size: size we want to trim @bio to, in sectors | |
1915 | */ | |
1916 | void bio_trim(struct bio *bio, int offset, int size) | |
1917 | { | |
1918 | /* 'bio' is a cloned bio which we need to trim to match | |
1919 | * the given offset and size. | |
6678d83f | 1920 | */ |
6678d83f KO |
1921 | |
1922 | size <<= 9; | |
4f024f37 | 1923 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1924 | return; |
1925 | ||
b7c44ed9 | 1926 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1927 | |
1928 | bio_advance(bio, offset << 9); | |
1929 | ||
4f024f37 | 1930 | bio->bi_iter.bi_size = size; |
376a78ab DM |
1931 | |
1932 | if (bio_integrity(bio)) | |
fbd08e76 | 1933 | bio_integrity_trim(bio); |
376a78ab | 1934 | |
6678d83f KO |
1935 | } |
1936 | EXPORT_SYMBOL_GPL(bio_trim); | |
1937 | ||
1da177e4 LT |
1938 | /* |
1939 | * create memory pools for biovec's in a bio_set. | |
1940 | * use the global biovec slabs created for general use. | |
1941 | */ | |
a6c39cb4 | 1942 | mempool_t *biovec_create_pool(int pool_entries) |
1da177e4 | 1943 | { |
ed996a52 | 1944 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
1da177e4 | 1945 | |
9f060e22 | 1946 | return mempool_create_slab_pool(pool_entries, bp->slab); |
1da177e4 LT |
1947 | } |
1948 | ||
1949 | void bioset_free(struct bio_set *bs) | |
1950 | { | |
df2cb6da KO |
1951 | if (bs->rescue_workqueue) |
1952 | destroy_workqueue(bs->rescue_workqueue); | |
1953 | ||
1da177e4 LT |
1954 | if (bs->bio_pool) |
1955 | mempool_destroy(bs->bio_pool); | |
1956 | ||
9f060e22 KO |
1957 | if (bs->bvec_pool) |
1958 | mempool_destroy(bs->bvec_pool); | |
1959 | ||
7878cba9 | 1960 | bioset_integrity_free(bs); |
bb799ca0 | 1961 | bio_put_slab(bs); |
1da177e4 LT |
1962 | |
1963 | kfree(bs); | |
1964 | } | |
a112a71d | 1965 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1966 | |
011067b0 N |
1967 | /** |
1968 | * bioset_create - Create a bio_set | |
1969 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1970 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
47e0fb46 N |
1971 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS |
1972 | * and %BIOSET_NEED_RESCUER | |
011067b0 N |
1973 | * |
1974 | * Description: | |
1975 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1976 | * to ask for a number of bytes to be allocated in front of the bio. | |
1977 | * Front pad allocation is useful for embedding the bio inside | |
1978 | * another structure, to avoid allocating extra data to go with the bio. | |
1979 | * Note that the bio must be embedded at the END of that structure always, | |
1980 | * or things will break badly. | |
1981 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated | |
1982 | * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast(). | |
47e0fb46 N |
1983 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to |
1984 | * dispatch queued requests when the mempool runs out of space. | |
011067b0 N |
1985 | * |
1986 | */ | |
1987 | struct bio_set *bioset_create(unsigned int pool_size, | |
1988 | unsigned int front_pad, | |
1989 | int flags) | |
1da177e4 | 1990 | { |
392ddc32 | 1991 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1992 | struct bio_set *bs; |
1da177e4 | 1993 | |
1b434498 | 1994 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1995 | if (!bs) |
1996 | return NULL; | |
1997 | ||
bb799ca0 | 1998 | bs->front_pad = front_pad; |
1b434498 | 1999 | |
df2cb6da KO |
2000 | spin_lock_init(&bs->rescue_lock); |
2001 | bio_list_init(&bs->rescue_list); | |
2002 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
2003 | ||
392ddc32 | 2004 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
2005 | if (!bs->bio_slab) { |
2006 | kfree(bs); | |
2007 | return NULL; | |
2008 | } | |
2009 | ||
2010 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
2011 | if (!bs->bio_pool) |
2012 | goto bad; | |
2013 | ||
011067b0 | 2014 | if (flags & BIOSET_NEED_BVECS) { |
d8f429e1 JN |
2015 | bs->bvec_pool = biovec_create_pool(pool_size); |
2016 | if (!bs->bvec_pool) | |
2017 | goto bad; | |
2018 | } | |
df2cb6da | 2019 | |
47e0fb46 N |
2020 | if (!(flags & BIOSET_NEED_RESCUER)) |
2021 | return bs; | |
2022 | ||
df2cb6da KO |
2023 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); |
2024 | if (!bs->rescue_workqueue) | |
2025 | goto bad; | |
1da177e4 | 2026 | |
df2cb6da | 2027 | return bs; |
1da177e4 LT |
2028 | bad: |
2029 | bioset_free(bs); | |
2030 | return NULL; | |
2031 | } | |
a112a71d | 2032 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 2033 | |
852c788f | 2034 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 TH |
2035 | |
2036 | /** | |
2037 | * bio_associate_blkcg - associate a bio with the specified blkcg | |
2038 | * @bio: target bio | |
2039 | * @blkcg_css: css of the blkcg to associate | |
2040 | * | |
2041 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will | |
2042 | * treat @bio as if it were issued by a task which belongs to the blkcg. | |
2043 | * | |
2044 | * This function takes an extra reference of @blkcg_css which will be put | |
2045 | * when @bio is released. The caller must own @bio and is responsible for | |
2046 | * synchronizing calls to this function. | |
2047 | */ | |
2048 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) | |
2049 | { | |
2050 | if (unlikely(bio->bi_css)) | |
2051 | return -EBUSY; | |
2052 | css_get(blkcg_css); | |
2053 | bio->bi_css = blkcg_css; | |
2054 | return 0; | |
2055 | } | |
5aa2a96b | 2056 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
1d933cf0 | 2057 | |
852c788f TH |
2058 | /** |
2059 | * bio_associate_current - associate a bio with %current | |
2060 | * @bio: target bio | |
2061 | * | |
2062 | * Associate @bio with %current if it hasn't been associated yet. Block | |
2063 | * layer will treat @bio as if it were issued by %current no matter which | |
2064 | * task actually issues it. | |
2065 | * | |
2066 | * This function takes an extra reference of @task's io_context and blkcg | |
2067 | * which will be put when @bio is released. The caller must own @bio, | |
2068 | * ensure %current->io_context exists, and is responsible for synchronizing | |
2069 | * calls to this function. | |
2070 | */ | |
2071 | int bio_associate_current(struct bio *bio) | |
2072 | { | |
2073 | struct io_context *ioc; | |
852c788f | 2074 | |
1d933cf0 | 2075 | if (bio->bi_css) |
852c788f TH |
2076 | return -EBUSY; |
2077 | ||
2078 | ioc = current->io_context; | |
2079 | if (!ioc) | |
2080 | return -ENOENT; | |
2081 | ||
852c788f TH |
2082 | get_io_context_active(ioc); |
2083 | bio->bi_ioc = ioc; | |
c165b3e3 | 2084 | bio->bi_css = task_get_css(current, io_cgrp_id); |
852c788f TH |
2085 | return 0; |
2086 | } | |
5aa2a96b | 2087 | EXPORT_SYMBOL_GPL(bio_associate_current); |
852c788f TH |
2088 | |
2089 | /** | |
2090 | * bio_disassociate_task - undo bio_associate_current() | |
2091 | * @bio: target bio | |
2092 | */ | |
2093 | void bio_disassociate_task(struct bio *bio) | |
2094 | { | |
2095 | if (bio->bi_ioc) { | |
2096 | put_io_context(bio->bi_ioc); | |
2097 | bio->bi_ioc = NULL; | |
2098 | } | |
2099 | if (bio->bi_css) { | |
2100 | css_put(bio->bi_css); | |
2101 | bio->bi_css = NULL; | |
2102 | } | |
2103 | } | |
2104 | ||
20bd723e PV |
2105 | /** |
2106 | * bio_clone_blkcg_association - clone blkcg association from src to dst bio | |
2107 | * @dst: destination bio | |
2108 | * @src: source bio | |
2109 | */ | |
2110 | void bio_clone_blkcg_association(struct bio *dst, struct bio *src) | |
2111 | { | |
2112 | if (src->bi_css) | |
2113 | WARN_ON(bio_associate_blkcg(dst, src->bi_css)); | |
2114 | } | |
8a8e6f84 | 2115 | EXPORT_SYMBOL_GPL(bio_clone_blkcg_association); |
852c788f TH |
2116 | #endif /* CONFIG_BLK_CGROUP */ |
2117 | ||
1da177e4 LT |
2118 | static void __init biovec_init_slabs(void) |
2119 | { | |
2120 | int i; | |
2121 | ||
ed996a52 | 2122 | for (i = 0; i < BVEC_POOL_NR; i++) { |
1da177e4 LT |
2123 | int size; |
2124 | struct biovec_slab *bvs = bvec_slabs + i; | |
2125 | ||
a7fcd37c JA |
2126 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2127 | bvs->slab = NULL; | |
2128 | continue; | |
2129 | } | |
a7fcd37c | 2130 | |
1da177e4 LT |
2131 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2132 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2133 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2134 | } |
2135 | } | |
2136 | ||
2137 | static int __init init_bio(void) | |
2138 | { | |
bb799ca0 JA |
2139 | bio_slab_max = 2; |
2140 | bio_slab_nr = 0; | |
2141 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
2142 | if (!bio_slabs) | |
2143 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2144 | |
7878cba9 | 2145 | bio_integrity_init(); |
1da177e4 LT |
2146 | biovec_init_slabs(); |
2147 | ||
011067b0 | 2148 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS); |
1da177e4 LT |
2149 | if (!fs_bio_set) |
2150 | panic("bio: can't allocate bios\n"); | |
2151 | ||
a91a2785 MP |
2152 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
2153 | panic("bio: can't create integrity pool\n"); | |
2154 | ||
1da177e4 LT |
2155 | return 0; |
2156 | } | |
1da177e4 | 2157 | subsys_initcall(init_bio); |