]>
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> |
08e18eab | 31 | #include <linux/blk-cgroup.h> |
1da177e4 | 32 | |
55782138 | 33 | #include <trace/events/block.h> |
9e234eea | 34 | #include "blk.h" |
67b42d0b | 35 | #include "blk-rq-qos.h" |
0bfc2455 | 36 | |
392ddc32 JA |
37 | /* |
38 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
39 | * itself, to shrink a bio data allocation from two mempool calls to one | |
40 | */ | |
41 | #define BIO_INLINE_VECS 4 | |
42 | ||
1da177e4 LT |
43 | /* |
44 | * if you change this list, also change bvec_alloc or things will | |
45 | * break badly! cannot be bigger than what you can fit into an | |
46 | * unsigned short | |
47 | */ | |
bd5c4fac | 48 | #define BV(x, n) { .nr_vecs = x, .name = "biovec-"#n } |
ed996a52 | 49 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
bd5c4fac | 50 | BV(1, 1), BV(4, 4), BV(16, 16), BV(64, 64), BV(128, 128), BV(BIO_MAX_PAGES, max), |
1da177e4 LT |
51 | }; |
52 | #undef BV | |
53 | ||
1da177e4 LT |
54 | /* |
55 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
56 | * IO code that does not need private memory pools. | |
57 | */ | |
f4f8154a | 58 | struct bio_set fs_bio_set; |
3f86a82a | 59 | EXPORT_SYMBOL(fs_bio_set); |
1da177e4 | 60 | |
bb799ca0 JA |
61 | /* |
62 | * Our slab pool management | |
63 | */ | |
64 | struct bio_slab { | |
65 | struct kmem_cache *slab; | |
66 | unsigned int slab_ref; | |
67 | unsigned int slab_size; | |
68 | char name[8]; | |
69 | }; | |
70 | static DEFINE_MUTEX(bio_slab_lock); | |
71 | static struct bio_slab *bio_slabs; | |
72 | static unsigned int bio_slab_nr, bio_slab_max; | |
73 | ||
74 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
75 | { | |
76 | unsigned int sz = sizeof(struct bio) + extra_size; | |
77 | struct kmem_cache *slab = NULL; | |
389d7b26 | 78 | struct bio_slab *bslab, *new_bio_slabs; |
386bc35a | 79 | unsigned int new_bio_slab_max; |
bb799ca0 JA |
80 | unsigned int i, entry = -1; |
81 | ||
82 | mutex_lock(&bio_slab_lock); | |
83 | ||
84 | i = 0; | |
85 | while (i < bio_slab_nr) { | |
f06f135d | 86 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
87 | |
88 | if (!bslab->slab && entry == -1) | |
89 | entry = i; | |
90 | else if (bslab->slab_size == sz) { | |
91 | slab = bslab->slab; | |
92 | bslab->slab_ref++; | |
93 | break; | |
94 | } | |
95 | i++; | |
96 | } | |
97 | ||
98 | if (slab) | |
99 | goto out_unlock; | |
100 | ||
101 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
386bc35a | 102 | new_bio_slab_max = bio_slab_max << 1; |
389d7b26 | 103 | new_bio_slabs = krealloc(bio_slabs, |
386bc35a | 104 | new_bio_slab_max * sizeof(struct bio_slab), |
389d7b26 AK |
105 | GFP_KERNEL); |
106 | if (!new_bio_slabs) | |
bb799ca0 | 107 | goto out_unlock; |
386bc35a | 108 | bio_slab_max = new_bio_slab_max; |
389d7b26 | 109 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
110 | } |
111 | if (entry == -1) | |
112 | entry = bio_slab_nr++; | |
113 | ||
114 | bslab = &bio_slabs[entry]; | |
115 | ||
116 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
6a241483 MP |
117 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
118 | SLAB_HWCACHE_ALIGN, NULL); | |
bb799ca0 JA |
119 | if (!slab) |
120 | goto out_unlock; | |
121 | ||
bb799ca0 JA |
122 | bslab->slab = slab; |
123 | bslab->slab_ref = 1; | |
124 | bslab->slab_size = sz; | |
125 | out_unlock: | |
126 | mutex_unlock(&bio_slab_lock); | |
127 | return slab; | |
128 | } | |
129 | ||
130 | static void bio_put_slab(struct bio_set *bs) | |
131 | { | |
132 | struct bio_slab *bslab = NULL; | |
133 | unsigned int i; | |
134 | ||
135 | mutex_lock(&bio_slab_lock); | |
136 | ||
137 | for (i = 0; i < bio_slab_nr; i++) { | |
138 | if (bs->bio_slab == bio_slabs[i].slab) { | |
139 | bslab = &bio_slabs[i]; | |
140 | break; | |
141 | } | |
142 | } | |
143 | ||
144 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
145 | goto out; | |
146 | ||
147 | WARN_ON(!bslab->slab_ref); | |
148 | ||
149 | if (--bslab->slab_ref) | |
150 | goto out; | |
151 | ||
152 | kmem_cache_destroy(bslab->slab); | |
153 | bslab->slab = NULL; | |
154 | ||
155 | out: | |
156 | mutex_unlock(&bio_slab_lock); | |
157 | } | |
158 | ||
7ba1ba12 MP |
159 | unsigned int bvec_nr_vecs(unsigned short idx) |
160 | { | |
d6c02a9b | 161 | return bvec_slabs[--idx].nr_vecs; |
7ba1ba12 MP |
162 | } |
163 | ||
9f060e22 | 164 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
bb799ca0 | 165 | { |
ed996a52 CH |
166 | if (!idx) |
167 | return; | |
168 | idx--; | |
169 | ||
170 | BIO_BUG_ON(idx >= BVEC_POOL_NR); | |
bb799ca0 | 171 | |
ed996a52 | 172 | if (idx == BVEC_POOL_MAX) { |
9f060e22 | 173 | mempool_free(bv, pool); |
ed996a52 | 174 | } else { |
bb799ca0 JA |
175 | struct biovec_slab *bvs = bvec_slabs + idx; |
176 | ||
177 | kmem_cache_free(bvs->slab, bv); | |
178 | } | |
179 | } | |
180 | ||
9f060e22 KO |
181 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
182 | mempool_t *pool) | |
1da177e4 LT |
183 | { |
184 | struct bio_vec *bvl; | |
1da177e4 | 185 | |
7ff9345f JA |
186 | /* |
187 | * see comment near bvec_array define! | |
188 | */ | |
189 | switch (nr) { | |
190 | case 1: | |
191 | *idx = 0; | |
192 | break; | |
193 | case 2 ... 4: | |
194 | *idx = 1; | |
195 | break; | |
196 | case 5 ... 16: | |
197 | *idx = 2; | |
198 | break; | |
199 | case 17 ... 64: | |
200 | *idx = 3; | |
201 | break; | |
202 | case 65 ... 128: | |
203 | *idx = 4; | |
204 | break; | |
205 | case 129 ... BIO_MAX_PAGES: | |
206 | *idx = 5; | |
207 | break; | |
208 | default: | |
209 | return NULL; | |
210 | } | |
211 | ||
212 | /* | |
213 | * idx now points to the pool we want to allocate from. only the | |
214 | * 1-vec entry pool is mempool backed. | |
215 | */ | |
ed996a52 | 216 | if (*idx == BVEC_POOL_MAX) { |
7ff9345f | 217 | fallback: |
9f060e22 | 218 | bvl = mempool_alloc(pool, gfp_mask); |
7ff9345f JA |
219 | } else { |
220 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
d0164adc | 221 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
7ff9345f | 222 | |
0a0d96b0 | 223 | /* |
7ff9345f JA |
224 | * Make this allocation restricted and don't dump info on |
225 | * allocation failures, since we'll fallback to the mempool | |
226 | * in case of failure. | |
0a0d96b0 | 227 | */ |
7ff9345f | 228 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 229 | |
0a0d96b0 | 230 | /* |
d0164adc | 231 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
7ff9345f | 232 | * is set, retry with the 1-entry mempool |
0a0d96b0 | 233 | */ |
7ff9345f | 234 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
d0164adc | 235 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
ed996a52 | 236 | *idx = BVEC_POOL_MAX; |
7ff9345f JA |
237 | goto fallback; |
238 | } | |
239 | } | |
240 | ||
ed996a52 | 241 | (*idx)++; |
1da177e4 LT |
242 | return bvl; |
243 | } | |
244 | ||
9ae3b3f5 | 245 | void bio_uninit(struct bio *bio) |
1da177e4 | 246 | { |
6f70fb66 | 247 | bio_disassociate_blkg(bio); |
4254bba1 | 248 | } |
9ae3b3f5 | 249 | EXPORT_SYMBOL(bio_uninit); |
7ba1ba12 | 250 | |
4254bba1 KO |
251 | static void bio_free(struct bio *bio) |
252 | { | |
253 | struct bio_set *bs = bio->bi_pool; | |
254 | void *p; | |
255 | ||
9ae3b3f5 | 256 | bio_uninit(bio); |
4254bba1 KO |
257 | |
258 | if (bs) { | |
8aa6ba2f | 259 | bvec_free(&bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio)); |
4254bba1 KO |
260 | |
261 | /* | |
262 | * If we have front padding, adjust the bio pointer before freeing | |
263 | */ | |
264 | p = bio; | |
bb799ca0 JA |
265 | p -= bs->front_pad; |
266 | ||
8aa6ba2f | 267 | mempool_free(p, &bs->bio_pool); |
4254bba1 KO |
268 | } else { |
269 | /* Bio was allocated by bio_kmalloc() */ | |
270 | kfree(bio); | |
271 | } | |
3676347a PO |
272 | } |
273 | ||
9ae3b3f5 JA |
274 | /* |
275 | * Users of this function have their own bio allocation. Subsequently, | |
276 | * they must remember to pair any call to bio_init() with bio_uninit() | |
277 | * when IO has completed, or when the bio is released. | |
278 | */ | |
3a83f467 ML |
279 | void bio_init(struct bio *bio, struct bio_vec *table, |
280 | unsigned short max_vecs) | |
1da177e4 | 281 | { |
2b94de55 | 282 | memset(bio, 0, sizeof(*bio)); |
c4cf5261 | 283 | atomic_set(&bio->__bi_remaining, 1); |
dac56212 | 284 | atomic_set(&bio->__bi_cnt, 1); |
3a83f467 ML |
285 | |
286 | bio->bi_io_vec = table; | |
287 | bio->bi_max_vecs = max_vecs; | |
1da177e4 | 288 | } |
a112a71d | 289 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 290 | |
f44b48c7 KO |
291 | /** |
292 | * bio_reset - reinitialize a bio | |
293 | * @bio: bio to reset | |
294 | * | |
295 | * Description: | |
296 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
297 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
298 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
299 | * comment in struct bio. | |
300 | */ | |
301 | void bio_reset(struct bio *bio) | |
302 | { | |
303 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
304 | ||
9ae3b3f5 | 305 | bio_uninit(bio); |
f44b48c7 KO |
306 | |
307 | memset(bio, 0, BIO_RESET_BYTES); | |
4246a0b6 | 308 | bio->bi_flags = flags; |
c4cf5261 | 309 | atomic_set(&bio->__bi_remaining, 1); |
f44b48c7 KO |
310 | } |
311 | EXPORT_SYMBOL(bio_reset); | |
312 | ||
38f8baae | 313 | static struct bio *__bio_chain_endio(struct bio *bio) |
196d38bc | 314 | { |
4246a0b6 CH |
315 | struct bio *parent = bio->bi_private; |
316 | ||
4e4cbee9 CH |
317 | if (!parent->bi_status) |
318 | parent->bi_status = bio->bi_status; | |
196d38bc | 319 | bio_put(bio); |
38f8baae CH |
320 | return parent; |
321 | } | |
322 | ||
323 | static void bio_chain_endio(struct bio *bio) | |
324 | { | |
325 | bio_endio(__bio_chain_endio(bio)); | |
196d38bc KO |
326 | } |
327 | ||
328 | /** | |
329 | * bio_chain - chain bio completions | |
1051a902 RD |
330 | * @bio: the target bio |
331 | * @parent: the @bio's parent bio | |
196d38bc KO |
332 | * |
333 | * The caller won't have a bi_end_io called when @bio completes - instead, | |
334 | * @parent's bi_end_io won't be called until both @parent and @bio have | |
335 | * completed; the chained bio will also be freed when it completes. | |
336 | * | |
337 | * The caller must not set bi_private or bi_end_io in @bio. | |
338 | */ | |
339 | void bio_chain(struct bio *bio, struct bio *parent) | |
340 | { | |
341 | BUG_ON(bio->bi_private || bio->bi_end_io); | |
342 | ||
343 | bio->bi_private = parent; | |
344 | bio->bi_end_io = bio_chain_endio; | |
c4cf5261 | 345 | bio_inc_remaining(parent); |
196d38bc KO |
346 | } |
347 | EXPORT_SYMBOL(bio_chain); | |
348 | ||
df2cb6da KO |
349 | static void bio_alloc_rescue(struct work_struct *work) |
350 | { | |
351 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); | |
352 | struct bio *bio; | |
353 | ||
354 | while (1) { | |
355 | spin_lock(&bs->rescue_lock); | |
356 | bio = bio_list_pop(&bs->rescue_list); | |
357 | spin_unlock(&bs->rescue_lock); | |
358 | ||
359 | if (!bio) | |
360 | break; | |
361 | ||
362 | generic_make_request(bio); | |
363 | } | |
364 | } | |
365 | ||
366 | static void punt_bios_to_rescuer(struct bio_set *bs) | |
367 | { | |
368 | struct bio_list punt, nopunt; | |
369 | struct bio *bio; | |
370 | ||
47e0fb46 N |
371 | if (WARN_ON_ONCE(!bs->rescue_workqueue)) |
372 | return; | |
df2cb6da KO |
373 | /* |
374 | * In order to guarantee forward progress we must punt only bios that | |
375 | * were allocated from this bio_set; otherwise, if there was a bio on | |
376 | * there for a stacking driver higher up in the stack, processing it | |
377 | * could require allocating bios from this bio_set, and doing that from | |
378 | * our own rescuer would be bad. | |
379 | * | |
380 | * Since bio lists are singly linked, pop them all instead of trying to | |
381 | * remove from the middle of the list: | |
382 | */ | |
383 | ||
384 | bio_list_init(&punt); | |
385 | bio_list_init(&nopunt); | |
386 | ||
f5fe1b51 | 387 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
df2cb6da | 388 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
f5fe1b51 | 389 | current->bio_list[0] = nopunt; |
df2cb6da | 390 | |
f5fe1b51 N |
391 | bio_list_init(&nopunt); |
392 | while ((bio = bio_list_pop(¤t->bio_list[1]))) | |
393 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); | |
394 | current->bio_list[1] = nopunt; | |
df2cb6da KO |
395 | |
396 | spin_lock(&bs->rescue_lock); | |
397 | bio_list_merge(&bs->rescue_list, &punt); | |
398 | spin_unlock(&bs->rescue_lock); | |
399 | ||
400 | queue_work(bs->rescue_workqueue, &bs->rescue_work); | |
401 | } | |
402 | ||
1da177e4 LT |
403 | /** |
404 | * bio_alloc_bioset - allocate a bio for I/O | |
519c8e9f | 405 | * @gfp_mask: the GFP_* mask given to the slab allocator |
1da177e4 | 406 | * @nr_iovecs: number of iovecs to pre-allocate |
db18efac | 407 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
408 | * |
409 | * Description: | |
3f86a82a KO |
410 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
411 | * backed by the @bs's mempool. | |
412 | * | |
d0164adc MG |
413 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
414 | * always be able to allocate a bio. This is due to the mempool guarantees. | |
415 | * To make this work, callers must never allocate more than 1 bio at a time | |
416 | * from this pool. Callers that need to allocate more than 1 bio must always | |
417 | * submit the previously allocated bio for IO before attempting to allocate | |
418 | * a new one. Failure to do so can cause deadlocks under memory pressure. | |
3f86a82a | 419 | * |
df2cb6da KO |
420 | * Note that when running under generic_make_request() (i.e. any block |
421 | * driver), bios are not submitted until after you return - see the code in | |
422 | * generic_make_request() that converts recursion into iteration, to prevent | |
423 | * stack overflows. | |
424 | * | |
425 | * This would normally mean allocating multiple bios under | |
426 | * generic_make_request() would be susceptible to deadlocks, but we have | |
427 | * deadlock avoidance code that resubmits any blocked bios from a rescuer | |
428 | * thread. | |
429 | * | |
430 | * However, we do not guarantee forward progress for allocations from other | |
431 | * mempools. Doing multiple allocations from the same mempool under | |
432 | * generic_make_request() should be avoided - instead, use bio_set's front_pad | |
433 | * for per bio allocations. | |
434 | * | |
3f86a82a KO |
435 | * RETURNS: |
436 | * Pointer to new bio on success, NULL on failure. | |
437 | */ | |
7a88fa19 DC |
438 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, unsigned int nr_iovecs, |
439 | struct bio_set *bs) | |
1da177e4 | 440 | { |
df2cb6da | 441 | gfp_t saved_gfp = gfp_mask; |
3f86a82a KO |
442 | unsigned front_pad; |
443 | unsigned inline_vecs; | |
34053979 | 444 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
445 | struct bio *bio; |
446 | void *p; | |
447 | ||
3f86a82a KO |
448 | if (!bs) { |
449 | if (nr_iovecs > UIO_MAXIOV) | |
450 | return NULL; | |
451 | ||
452 | p = kmalloc(sizeof(struct bio) + | |
453 | nr_iovecs * sizeof(struct bio_vec), | |
454 | gfp_mask); | |
455 | front_pad = 0; | |
456 | inline_vecs = nr_iovecs; | |
457 | } else { | |
d8f429e1 | 458 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
8aa6ba2f KO |
459 | if (WARN_ON_ONCE(!mempool_initialized(&bs->bvec_pool) && |
460 | nr_iovecs > 0)) | |
d8f429e1 | 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 | |
8aa6ba2f | 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; | |
8aa6ba2f | 493 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
df2cb6da KO |
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 | ||
8aa6ba2f | 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; | |
8aa6ba2f | 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: | |
8aa6ba2f | 530 | mempool_free(p, &bs->bio_pool); |
34053979 | 531 | return NULL; |
1da177e4 | 532 | } |
a112a71d | 533 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 534 | |
38a72dac | 535 | void zero_fill_bio_iter(struct bio *bio, struct bvec_iter start) |
1da177e4 LT |
536 | { |
537 | unsigned long flags; | |
7988613b KO |
538 | struct bio_vec bv; |
539 | struct bvec_iter iter; | |
1da177e4 | 540 | |
38a72dac | 541 | __bio_for_each_segment(bv, bio, iter, start) { |
7988613b KO |
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 | } | |
38a72dac | 548 | EXPORT_SYMBOL(zero_fill_bio_iter); |
1da177e4 LT |
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 | |
6c210aa5 | 574 | 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 | } | |
581 | ||
59d276fe KO |
582 | /** |
583 | * __bio_clone_fast - clone a bio that shares the original bio's biovec | |
584 | * @bio: destination bio | |
585 | * @bio_src: bio to clone | |
586 | * | |
587 | * Clone a &bio. Caller will own the returned bio, but not | |
588 | * the actual data it points to. Reference count of returned | |
589 | * bio will be one. | |
590 | * | |
591 | * Caller must ensure that @bio_src is not freed before @bio. | |
592 | */ | |
593 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) | |
594 | { | |
ed996a52 | 595 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
59d276fe KO |
596 | |
597 | /* | |
74d46992 | 598 | * most users will be overriding ->bi_disk with a new target, |
59d276fe KO |
599 | * so we don't set nor calculate new physical/hw segment counts here |
600 | */ | |
74d46992 | 601 | bio->bi_disk = bio_src->bi_disk; |
62530ed8 | 602 | bio->bi_partno = bio_src->bi_partno; |
b7c44ed9 | 603 | bio_set_flag(bio, BIO_CLONED); |
111be883 SL |
604 | if (bio_flagged(bio_src, BIO_THROTTLED)) |
605 | bio_set_flag(bio, BIO_THROTTLED); | |
1eff9d32 | 606 | bio->bi_opf = bio_src->bi_opf; |
ca474b73 | 607 | bio->bi_ioprio = bio_src->bi_ioprio; |
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; | |
20bd723e | 611 | |
db6638d7 | 612 | bio_clone_blkg_association(bio, bio_src); |
e439bedf | 613 | blkcg_bio_issue_init(bio); |
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 | ||
5919482e ML |
650 | static inline bool page_is_mergeable(const struct bio_vec *bv, |
651 | struct page *page, unsigned int len, unsigned int off, | |
652 | bool same_page) | |
653 | { | |
654 | phys_addr_t vec_end_addr = page_to_phys(bv->bv_page) + | |
655 | bv->bv_offset + bv->bv_len - 1; | |
656 | phys_addr_t page_addr = page_to_phys(page); | |
657 | ||
658 | if (vec_end_addr + 1 != page_addr + off) | |
659 | return false; | |
660 | if (xen_domain() && !xen_biovec_phys_mergeable(bv, page)) | |
661 | return false; | |
52d52d1c CH |
662 | |
663 | if ((vec_end_addr & PAGE_MASK) != page_addr) { | |
664 | if (same_page) | |
665 | return false; | |
666 | if (pfn_to_page(PFN_DOWN(vec_end_addr)) + 1 != page) | |
667 | return false; | |
668 | } | |
5919482e | 669 | |
551879a4 ML |
670 | WARN_ON_ONCE(same_page && (len + off) > PAGE_SIZE); |
671 | ||
5919482e ML |
672 | return true; |
673 | } | |
674 | ||
489fbbcb ML |
675 | /* |
676 | * Check if the @page can be added to the current segment(@bv), and make | |
677 | * sure to call it only if page_is_mergeable(@bv, @page) is true | |
678 | */ | |
679 | static bool can_add_page_to_seg(struct request_queue *q, | |
680 | struct bio_vec *bv, struct page *page, unsigned len, | |
681 | unsigned offset) | |
682 | { | |
683 | unsigned long mask = queue_segment_boundary(q); | |
684 | phys_addr_t addr1 = page_to_phys(bv->bv_page) + bv->bv_offset; | |
685 | phys_addr_t addr2 = page_to_phys(page) + offset + len - 1; | |
686 | ||
687 | if ((addr1 | mask) != (addr2 | mask)) | |
688 | return false; | |
689 | ||
690 | if (bv->bv_len + len > queue_max_segment_size(q)) | |
691 | return false; | |
692 | ||
693 | return true; | |
694 | } | |
695 | ||
1da177e4 | 696 | /** |
19047087 | 697 | * __bio_add_pc_page - attempt to add page to passthrough bio |
c66a14d0 KO |
698 | * @q: the target queue |
699 | * @bio: destination bio | |
700 | * @page: page to add | |
701 | * @len: vec entry length | |
702 | * @offset: vec entry offset | |
19047087 | 703 | * @put_same_page: put the page if it is same with last added page |
1da177e4 | 704 | * |
c66a14d0 KO |
705 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
706 | * number of reasons, such as the bio being full or target block device | |
707 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
708 | * so it is always possible to add a single page to an empty bio. | |
709 | * | |
5a8ce240 | 710 | * This should only be used by passthrough bios. |
1da177e4 | 711 | */ |
19047087 ML |
712 | int __bio_add_pc_page(struct request_queue *q, struct bio *bio, |
713 | struct page *page, unsigned int len, unsigned int offset, | |
714 | bool put_same_page) | |
1da177e4 | 715 | { |
1da177e4 LT |
716 | struct bio_vec *bvec; |
717 | ||
718 | /* | |
719 | * cloned bio must not modify vec list | |
720 | */ | |
721 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
722 | return 0; | |
723 | ||
c66a14d0 | 724 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
725 | return 0; |
726 | ||
80cfd548 JA |
727 | /* |
728 | * For filesystems with a blocksize smaller than the pagesize | |
729 | * we will often be called with the same page as last time and | |
730 | * a consecutive offset. Optimize this special case. | |
731 | */ | |
732 | if (bio->bi_vcnt > 0) { | |
5a8ce240 | 733 | bvec = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
80cfd548 | 734 | |
5a8ce240 ML |
735 | if (page == bvec->bv_page && |
736 | offset == bvec->bv_offset + bvec->bv_len) { | |
19047087 ML |
737 | if (put_same_page) |
738 | put_page(page); | |
489fbbcb | 739 | bvec_merge: |
5a8ce240 | 740 | bvec->bv_len += len; |
fcbf6a08 | 741 | bio->bi_iter.bi_size += len; |
80cfd548 JA |
742 | goto done; |
743 | } | |
66cb45aa JA |
744 | |
745 | /* | |
746 | * If the queue doesn't support SG gaps and adding this | |
747 | * offset would create a gap, disallow it. | |
748 | */ | |
5a8ce240 | 749 | if (bvec_gap_to_prev(q, bvec, offset)) |
66cb45aa | 750 | return 0; |
489fbbcb ML |
751 | |
752 | if (page_is_mergeable(bvec, page, len, offset, false) && | |
753 | can_add_page_to_seg(q, bvec, page, len, offset)) | |
754 | goto bvec_merge; | |
80cfd548 JA |
755 | } |
756 | ||
0aa69fd3 | 757 | if (bio_full(bio)) |
1da177e4 LT |
758 | return 0; |
759 | ||
489fbbcb ML |
760 | if (bio->bi_phys_segments >= queue_max_segments(q)) |
761 | return 0; | |
762 | ||
1da177e4 | 763 | /* |
fcbf6a08 ML |
764 | * setup the new entry, we might clear it again later if we |
765 | * cannot add the page | |
766 | */ | |
767 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
768 | bvec->bv_page = page; | |
769 | bvec->bv_len = len; | |
770 | bvec->bv_offset = offset; | |
771 | bio->bi_vcnt++; | |
fcbf6a08 ML |
772 | bio->bi_iter.bi_size += len; |
773 | ||
80cfd548 | 774 | done: |
489fbbcb ML |
775 | bio->bi_phys_segments = bio->bi_vcnt; |
776 | bio_set_flag(bio, BIO_SEG_VALID); | |
1da177e4 LT |
777 | return len; |
778 | } | |
19047087 ML |
779 | EXPORT_SYMBOL(__bio_add_pc_page); |
780 | ||
781 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, | |
782 | struct page *page, unsigned int len, unsigned int offset) | |
783 | { | |
784 | return __bio_add_pc_page(q, bio, page, len, offset, false); | |
785 | } | |
a112a71d | 786 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 787 | |
1da177e4 | 788 | /** |
0aa69fd3 CH |
789 | * __bio_try_merge_page - try appending data to an existing bvec. |
790 | * @bio: destination bio | |
551879a4 | 791 | * @page: start page to add |
0aa69fd3 | 792 | * @len: length of the data to add |
551879a4 | 793 | * @off: offset of the data relative to @page |
07173c3e ML |
794 | * @same_page: if %true only merge if the new data is in the same physical |
795 | * page as the last segment of the bio. | |
1da177e4 | 796 | * |
0aa69fd3 CH |
797 | * Try to add the data at @page + @off to the last bvec of @bio. This is a |
798 | * a useful optimisation for file systems with a block size smaller than the | |
799 | * page size. | |
800 | * | |
551879a4 ML |
801 | * Warn if (@len, @off) crosses pages in case that @same_page is true. |
802 | * | |
0aa69fd3 | 803 | * Return %true on success or %false on failure. |
1da177e4 | 804 | */ |
0aa69fd3 | 805 | bool __bio_try_merge_page(struct bio *bio, struct page *page, |
07173c3e | 806 | unsigned int len, unsigned int off, bool same_page) |
1da177e4 | 807 | { |
c66a14d0 | 808 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
0aa69fd3 | 809 | return false; |
762380ad | 810 | |
c66a14d0 | 811 | if (bio->bi_vcnt > 0) { |
0aa69fd3 | 812 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
5919482e ML |
813 | |
814 | if (page_is_mergeable(bv, page, len, off, same_page)) { | |
815 | bv->bv_len += len; | |
816 | bio->bi_iter.bi_size += len; | |
817 | return true; | |
818 | } | |
c66a14d0 | 819 | } |
0aa69fd3 CH |
820 | return false; |
821 | } | |
822 | EXPORT_SYMBOL_GPL(__bio_try_merge_page); | |
c66a14d0 | 823 | |
0aa69fd3 | 824 | /** |
551879a4 | 825 | * __bio_add_page - add page(s) to a bio in a new segment |
0aa69fd3 | 826 | * @bio: destination bio |
551879a4 ML |
827 | * @page: start page to add |
828 | * @len: length of the data to add, may cross pages | |
829 | * @off: offset of the data relative to @page, may cross pages | |
0aa69fd3 CH |
830 | * |
831 | * Add the data at @page + @off to @bio as a new bvec. The caller must ensure | |
832 | * that @bio has space for another bvec. | |
833 | */ | |
834 | void __bio_add_page(struct bio *bio, struct page *page, | |
835 | unsigned int len, unsigned int off) | |
836 | { | |
837 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
c66a14d0 | 838 | |
0aa69fd3 CH |
839 | WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)); |
840 | WARN_ON_ONCE(bio_full(bio)); | |
841 | ||
842 | bv->bv_page = page; | |
843 | bv->bv_offset = off; | |
844 | bv->bv_len = len; | |
c66a14d0 | 845 | |
c66a14d0 | 846 | bio->bi_iter.bi_size += len; |
0aa69fd3 CH |
847 | bio->bi_vcnt++; |
848 | } | |
849 | EXPORT_SYMBOL_GPL(__bio_add_page); | |
850 | ||
851 | /** | |
551879a4 | 852 | * bio_add_page - attempt to add page(s) to bio |
0aa69fd3 | 853 | * @bio: destination bio |
551879a4 ML |
854 | * @page: start page to add |
855 | * @len: vec entry length, may cross pages | |
856 | * @offset: vec entry offset relative to @page, may cross pages | |
0aa69fd3 | 857 | * |
551879a4 | 858 | * Attempt to add page(s) to the bio_vec maplist. This will only fail |
0aa69fd3 CH |
859 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. |
860 | */ | |
861 | int bio_add_page(struct bio *bio, struct page *page, | |
862 | unsigned int len, unsigned int offset) | |
863 | { | |
07173c3e | 864 | if (!__bio_try_merge_page(bio, page, len, offset, false)) { |
0aa69fd3 CH |
865 | if (bio_full(bio)) |
866 | return 0; | |
867 | __bio_add_page(bio, page, len, offset); | |
868 | } | |
c66a14d0 | 869 | return len; |
1da177e4 | 870 | } |
a112a71d | 871 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 872 | |
7321ecbf CH |
873 | static void bio_get_pages(struct bio *bio) |
874 | { | |
875 | struct bvec_iter_all iter_all; | |
876 | struct bio_vec *bvec; | |
7321ecbf | 877 | |
2b070cfe | 878 | bio_for_each_segment_all(bvec, bio, iter_all) |
7321ecbf CH |
879 | get_page(bvec->bv_page); |
880 | } | |
881 | ||
882 | static void bio_release_pages(struct bio *bio) | |
883 | { | |
884 | struct bvec_iter_all iter_all; | |
885 | struct bio_vec *bvec; | |
7321ecbf | 886 | |
2b070cfe | 887 | bio_for_each_segment_all(bvec, bio, iter_all) |
7321ecbf CH |
888 | put_page(bvec->bv_page); |
889 | } | |
890 | ||
6d0c48ae JA |
891 | static int __bio_iov_bvec_add_pages(struct bio *bio, struct iov_iter *iter) |
892 | { | |
893 | const struct bio_vec *bv = iter->bvec; | |
894 | unsigned int len; | |
895 | size_t size; | |
896 | ||
897 | if (WARN_ON_ONCE(iter->iov_offset > bv->bv_len)) | |
898 | return -EINVAL; | |
899 | ||
900 | len = min_t(size_t, bv->bv_len - iter->iov_offset, iter->count); | |
901 | size = bio_add_page(bio, bv->bv_page, len, | |
902 | bv->bv_offset + iter->iov_offset); | |
a10584c3 CH |
903 | if (unlikely(size != len)) |
904 | return -EINVAL; | |
a10584c3 CH |
905 | iov_iter_advance(iter, size); |
906 | return 0; | |
6d0c48ae JA |
907 | } |
908 | ||
576ed913 CH |
909 | #define PAGE_PTRS_PER_BVEC (sizeof(struct bio_vec) / sizeof(struct page *)) |
910 | ||
2cefe4db | 911 | /** |
17d51b10 | 912 | * __bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio |
2cefe4db KO |
913 | * @bio: bio to add pages to |
914 | * @iter: iov iterator describing the region to be mapped | |
915 | * | |
17d51b10 | 916 | * Pins pages from *iter and appends them to @bio's bvec array. The |
2cefe4db | 917 | * pages will have to be released using put_page() when done. |
17d51b10 MW |
918 | * For multi-segment *iter, this function only adds pages from the |
919 | * the next non-empty segment of the iov iterator. | |
2cefe4db | 920 | */ |
17d51b10 | 921 | static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) |
2cefe4db | 922 | { |
576ed913 CH |
923 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; |
924 | unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt; | |
2cefe4db KO |
925 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; |
926 | struct page **pages = (struct page **)bv; | |
576ed913 CH |
927 | ssize_t size, left; |
928 | unsigned len, i; | |
b403ea24 | 929 | size_t offset; |
576ed913 CH |
930 | |
931 | /* | |
932 | * Move page array up in the allocated memory for the bio vecs as far as | |
933 | * possible so that we can start filling biovecs from the beginning | |
934 | * without overwriting the temporary page array. | |
935 | */ | |
936 | BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2); | |
937 | pages += entries_left * (PAGE_PTRS_PER_BVEC - 1); | |
2cefe4db KO |
938 | |
939 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); | |
940 | if (unlikely(size <= 0)) | |
941 | return size ? size : -EFAULT; | |
2cefe4db | 942 | |
576ed913 CH |
943 | for (left = size, i = 0; left > 0; left -= len, i++) { |
944 | struct page *page = pages[i]; | |
2cefe4db | 945 | |
576ed913 CH |
946 | len = min_t(size_t, PAGE_SIZE - offset, left); |
947 | if (WARN_ON_ONCE(bio_add_page(bio, page, len, offset) != len)) | |
948 | return -EINVAL; | |
949 | offset = 0; | |
2cefe4db KO |
950 | } |
951 | ||
2cefe4db KO |
952 | iov_iter_advance(iter, size); |
953 | return 0; | |
954 | } | |
17d51b10 MW |
955 | |
956 | /** | |
6d0c48ae | 957 | * bio_iov_iter_get_pages - add user or kernel pages to a bio |
17d51b10 | 958 | * @bio: bio to add pages to |
6d0c48ae JA |
959 | * @iter: iov iterator describing the region to be added |
960 | * | |
961 | * This takes either an iterator pointing to user memory, or one pointing to | |
962 | * kernel pages (BVEC iterator). If we're adding user pages, we pin them and | |
963 | * map them into the kernel. On IO completion, the caller should put those | |
399254aa JA |
964 | * pages. If we're adding kernel pages, and the caller told us it's safe to |
965 | * do so, we just have to add the pages to the bio directly. We don't grab an | |
966 | * extra reference to those pages (the user should already have that), and we | |
967 | * don't put the page on IO completion. The caller needs to check if the bio is | |
968 | * flagged BIO_NO_PAGE_REF on IO completion. If it isn't, then pages should be | |
969 | * released. | |
17d51b10 | 970 | * |
17d51b10 | 971 | * The function tries, but does not guarantee, to pin as many pages as |
6d0c48ae JA |
972 | * fit into the bio, or are requested in *iter, whatever is smaller. If |
973 | * MM encounters an error pinning the requested pages, it stops. Error | |
974 | * is returned only if 0 pages could be pinned. | |
17d51b10 MW |
975 | */ |
976 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) | |
977 | { | |
6d0c48ae | 978 | const bool is_bvec = iov_iter_is_bvec(iter); |
14eacf12 CH |
979 | int ret; |
980 | ||
981 | if (WARN_ON_ONCE(bio->bi_vcnt)) | |
982 | return -EINVAL; | |
17d51b10 MW |
983 | |
984 | do { | |
6d0c48ae JA |
985 | if (is_bvec) |
986 | ret = __bio_iov_bvec_add_pages(bio, iter); | |
987 | else | |
988 | ret = __bio_iov_iter_get_pages(bio, iter); | |
14eacf12 | 989 | } while (!ret && iov_iter_count(iter) && !bio_full(bio)); |
17d51b10 | 990 | |
7321ecbf CH |
991 | if (iov_iter_bvec_no_ref(iter)) |
992 | bio_set_flag(bio, BIO_NO_PAGE_REF); | |
0257c0ed | 993 | else if (is_bvec) |
7321ecbf CH |
994 | bio_get_pages(bio); |
995 | ||
14eacf12 | 996 | return bio->bi_vcnt ? 0 : ret; |
17d51b10 | 997 | } |
2cefe4db | 998 | |
4246a0b6 | 999 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 | 1000 | { |
65e53aab | 1001 | complete(bio->bi_private); |
9e882242 KO |
1002 | } |
1003 | ||
1004 | /** | |
1005 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
1006 | * @bio: The &struct bio which describes the I/O |
1007 | * | |
1008 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
1009 | * bio_endio() on failure. | |
3d289d68 JK |
1010 | * |
1011 | * WARNING: Unlike to how submit_bio() is usually used, this function does not | |
1012 | * result in bio reference to be consumed. The caller must drop the reference | |
1013 | * on his own. | |
9e882242 | 1014 | */ |
4e49ea4a | 1015 | int submit_bio_wait(struct bio *bio) |
9e882242 | 1016 | { |
e319e1fb | 1017 | DECLARE_COMPLETION_ONSTACK_MAP(done, bio->bi_disk->lockdep_map); |
9e882242 | 1018 | |
65e53aab | 1019 | bio->bi_private = &done; |
9e882242 | 1020 | bio->bi_end_io = submit_bio_wait_endio; |
1eff9d32 | 1021 | bio->bi_opf |= REQ_SYNC; |
4e49ea4a | 1022 | submit_bio(bio); |
65e53aab | 1023 | wait_for_completion_io(&done); |
9e882242 | 1024 | |
65e53aab | 1025 | return blk_status_to_errno(bio->bi_status); |
9e882242 KO |
1026 | } |
1027 | EXPORT_SYMBOL(submit_bio_wait); | |
1028 | ||
054bdf64 KO |
1029 | /** |
1030 | * bio_advance - increment/complete a bio by some number of bytes | |
1031 | * @bio: bio to advance | |
1032 | * @bytes: number of bytes to complete | |
1033 | * | |
1034 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
1035 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
1036 | * be updated on the last bvec as well. | |
1037 | * | |
1038 | * @bio will then represent the remaining, uncompleted portion of the io. | |
1039 | */ | |
1040 | void bio_advance(struct bio *bio, unsigned bytes) | |
1041 | { | |
1042 | if (bio_integrity(bio)) | |
1043 | bio_integrity_advance(bio, bytes); | |
1044 | ||
4550dd6c | 1045 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
1046 | } |
1047 | EXPORT_SYMBOL(bio_advance); | |
1048 | ||
45db54d5 KO |
1049 | void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter, |
1050 | struct bio *src, struct bvec_iter *src_iter) | |
16ac3d63 | 1051 | { |
1cb9dda4 | 1052 | struct bio_vec src_bv, dst_bv; |
16ac3d63 | 1053 | void *src_p, *dst_p; |
1cb9dda4 | 1054 | unsigned bytes; |
16ac3d63 | 1055 | |
45db54d5 KO |
1056 | while (src_iter->bi_size && dst_iter->bi_size) { |
1057 | src_bv = bio_iter_iovec(src, *src_iter); | |
1058 | dst_bv = bio_iter_iovec(dst, *dst_iter); | |
1cb9dda4 KO |
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 | ||
6e6e811d KO |
1072 | flush_dcache_page(dst_bv.bv_page); |
1073 | ||
45db54d5 KO |
1074 | bio_advance_iter(src, src_iter, bytes); |
1075 | bio_advance_iter(dst, dst_iter, bytes); | |
16ac3d63 KO |
1076 | } |
1077 | } | |
38a72dac KO |
1078 | EXPORT_SYMBOL(bio_copy_data_iter); |
1079 | ||
1080 | /** | |
45db54d5 KO |
1081 | * bio_copy_data - copy contents of data buffers from one bio to another |
1082 | * @src: source bio | |
1083 | * @dst: destination bio | |
38a72dac KO |
1084 | * |
1085 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
1086 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
1087 | */ | |
1088 | void bio_copy_data(struct bio *dst, struct bio *src) | |
1089 | { | |
45db54d5 KO |
1090 | struct bvec_iter src_iter = src->bi_iter; |
1091 | struct bvec_iter dst_iter = dst->bi_iter; | |
1092 | ||
1093 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); | |
38a72dac | 1094 | } |
16ac3d63 KO |
1095 | EXPORT_SYMBOL(bio_copy_data); |
1096 | ||
45db54d5 KO |
1097 | /** |
1098 | * bio_list_copy_data - copy contents of data buffers from one chain of bios to | |
1099 | * another | |
1100 | * @src: source bio list | |
1101 | * @dst: destination bio list | |
1102 | * | |
1103 | * Stops when it reaches the end of either the @src list or @dst list - that is, | |
1104 | * copies min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of | |
1105 | * bios). | |
1106 | */ | |
1107 | void bio_list_copy_data(struct bio *dst, struct bio *src) | |
1108 | { | |
1109 | struct bvec_iter src_iter = src->bi_iter; | |
1110 | struct bvec_iter dst_iter = dst->bi_iter; | |
1111 | ||
1112 | while (1) { | |
1113 | if (!src_iter.bi_size) { | |
1114 | src = src->bi_next; | |
1115 | if (!src) | |
1116 | break; | |
1117 | ||
1118 | src_iter = src->bi_iter; | |
1119 | } | |
1120 | ||
1121 | if (!dst_iter.bi_size) { | |
1122 | dst = dst->bi_next; | |
1123 | if (!dst) | |
1124 | break; | |
1125 | ||
1126 | dst_iter = dst->bi_iter; | |
1127 | } | |
1128 | ||
1129 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); | |
1130 | } | |
1131 | } | |
1132 | EXPORT_SYMBOL(bio_list_copy_data); | |
1133 | ||
1da177e4 | 1134 | struct bio_map_data { |
152e283f | 1135 | int is_our_pages; |
26e49cfc KO |
1136 | struct iov_iter iter; |
1137 | struct iovec iov[]; | |
1da177e4 LT |
1138 | }; |
1139 | ||
0e5b935d | 1140 | static struct bio_map_data *bio_alloc_map_data(struct iov_iter *data, |
76029ff3 | 1141 | gfp_t gfp_mask) |
1da177e4 | 1142 | { |
0e5b935d AV |
1143 | struct bio_map_data *bmd; |
1144 | if (data->nr_segs > UIO_MAXIOV) | |
f3f63c1c | 1145 | return NULL; |
1da177e4 | 1146 | |
0e5b935d AV |
1147 | bmd = kmalloc(sizeof(struct bio_map_data) + |
1148 | sizeof(struct iovec) * data->nr_segs, gfp_mask); | |
1149 | if (!bmd) | |
1150 | return NULL; | |
1151 | memcpy(bmd->iov, data->iov, sizeof(struct iovec) * data->nr_segs); | |
1152 | bmd->iter = *data; | |
1153 | bmd->iter.iov = bmd->iov; | |
1154 | return bmd; | |
1da177e4 LT |
1155 | } |
1156 | ||
9124d3fe DP |
1157 | /** |
1158 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1159 | * @bio: The &struct bio which describes the I/O as destination | |
1160 | * @iter: iov_iter as source | |
1161 | * | |
1162 | * Copy all pages from iov_iter to bio. | |
1163 | * Returns 0 on success, or error on failure. | |
1164 | */ | |
98a09d61 | 1165 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter *iter) |
c5dec1c3 | 1166 | { |
c5dec1c3 | 1167 | struct bio_vec *bvec; |
6dc4f100 | 1168 | struct bvec_iter_all iter_all; |
c5dec1c3 | 1169 | |
2b070cfe | 1170 | bio_for_each_segment_all(bvec, bio, iter_all) { |
9124d3fe | 1171 | ssize_t ret; |
c5dec1c3 | 1172 | |
9124d3fe DP |
1173 | ret = copy_page_from_iter(bvec->bv_page, |
1174 | bvec->bv_offset, | |
1175 | bvec->bv_len, | |
98a09d61 | 1176 | iter); |
9124d3fe | 1177 | |
98a09d61 | 1178 | if (!iov_iter_count(iter)) |
9124d3fe DP |
1179 | break; |
1180 | ||
1181 | if (ret < bvec->bv_len) | |
1182 | return -EFAULT; | |
c5dec1c3 FT |
1183 | } |
1184 | ||
9124d3fe DP |
1185 | return 0; |
1186 | } | |
1187 | ||
1188 | /** | |
1189 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1190 | * @bio: The &struct bio which describes the I/O as source | |
1191 | * @iter: iov_iter as destination | |
1192 | * | |
1193 | * Copy all pages from bio to iov_iter. | |
1194 | * Returns 0 on success, or error on failure. | |
1195 | */ | |
1196 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1197 | { | |
9124d3fe | 1198 | struct bio_vec *bvec; |
6dc4f100 | 1199 | struct bvec_iter_all iter_all; |
9124d3fe | 1200 | |
2b070cfe | 1201 | bio_for_each_segment_all(bvec, bio, iter_all) { |
9124d3fe DP |
1202 | ssize_t ret; |
1203 | ||
1204 | ret = copy_page_to_iter(bvec->bv_page, | |
1205 | bvec->bv_offset, | |
1206 | bvec->bv_len, | |
1207 | &iter); | |
1208 | ||
1209 | if (!iov_iter_count(&iter)) | |
1210 | break; | |
1211 | ||
1212 | if (ret < bvec->bv_len) | |
1213 | return -EFAULT; | |
1214 | } | |
1215 | ||
1216 | return 0; | |
c5dec1c3 FT |
1217 | } |
1218 | ||
491221f8 | 1219 | void bio_free_pages(struct bio *bio) |
1dfa0f68 CH |
1220 | { |
1221 | struct bio_vec *bvec; | |
6dc4f100 | 1222 | struct bvec_iter_all iter_all; |
1dfa0f68 | 1223 | |
2b070cfe | 1224 | bio_for_each_segment_all(bvec, bio, iter_all) |
1dfa0f68 CH |
1225 | __free_page(bvec->bv_page); |
1226 | } | |
491221f8 | 1227 | EXPORT_SYMBOL(bio_free_pages); |
1dfa0f68 | 1228 | |
1da177e4 LT |
1229 | /** |
1230 | * bio_uncopy_user - finish previously mapped bio | |
1231 | * @bio: bio being terminated | |
1232 | * | |
ddad8dd0 | 1233 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1234 | * to user space in case of a read. |
1235 | */ | |
1236 | int bio_uncopy_user(struct bio *bio) | |
1237 | { | |
1238 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1239 | int ret = 0; |
1da177e4 | 1240 | |
35dc2483 RD |
1241 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1242 | /* | |
1243 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1244 | * don't copy into a random user address space, just free |
1245 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1246 | */ |
2d99b55d HR |
1247 | if (!current->mm) |
1248 | ret = -EINTR; | |
1249 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1250 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1251 | if (bmd->is_our_pages) |
1252 | bio_free_pages(bio); | |
35dc2483 | 1253 | } |
c8db4448 | 1254 | kfree(bmd); |
1da177e4 LT |
1255 | bio_put(bio); |
1256 | return ret; | |
1257 | } | |
1258 | ||
1259 | /** | |
c5dec1c3 | 1260 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1261 | * @q: destination block queue |
1262 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1263 | * @iter: iovec iterator | |
1264 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1265 | * |
1266 | * Prepares and returns a bio for indirect user io, bouncing data | |
1267 | * to/from kernel pages as necessary. Must be paired with | |
1268 | * call bio_uncopy_user() on io completion. | |
1269 | */ | |
152e283f FT |
1270 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1271 | struct rq_map_data *map_data, | |
e81cef5d | 1272 | struct iov_iter *iter, |
26e49cfc | 1273 | gfp_t gfp_mask) |
1da177e4 | 1274 | { |
1da177e4 | 1275 | struct bio_map_data *bmd; |
1da177e4 LT |
1276 | struct page *page; |
1277 | struct bio *bio; | |
d16d44eb AV |
1278 | int i = 0, ret; |
1279 | int nr_pages; | |
26e49cfc | 1280 | unsigned int len = iter->count; |
bd5cecea | 1281 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1282 | |
0e5b935d | 1283 | bmd = bio_alloc_map_data(iter, gfp_mask); |
1da177e4 LT |
1284 | if (!bmd) |
1285 | return ERR_PTR(-ENOMEM); | |
1286 | ||
26e49cfc KO |
1287 | /* |
1288 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1289 | * The caller provided iov might point to an on-stack or otherwise | |
1290 | * shortlived one. | |
1291 | */ | |
1292 | bmd->is_our_pages = map_data ? 0 : 1; | |
26e49cfc | 1293 | |
d16d44eb AV |
1294 | nr_pages = DIV_ROUND_UP(offset + len, PAGE_SIZE); |
1295 | if (nr_pages > BIO_MAX_PAGES) | |
1296 | nr_pages = BIO_MAX_PAGES; | |
26e49cfc | 1297 | |
1da177e4 | 1298 | ret = -ENOMEM; |
a9e9dc24 | 1299 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1300 | if (!bio) |
1301 | goto out_bmd; | |
1302 | ||
1da177e4 | 1303 | ret = 0; |
56c451f4 FT |
1304 | |
1305 | if (map_data) { | |
e623ddb4 | 1306 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1307 | i = map_data->offset / PAGE_SIZE; |
1308 | } | |
1da177e4 | 1309 | while (len) { |
e623ddb4 | 1310 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1311 | |
56c451f4 FT |
1312 | bytes -= offset; |
1313 | ||
1da177e4 LT |
1314 | if (bytes > len) |
1315 | bytes = len; | |
1316 | ||
152e283f | 1317 | if (map_data) { |
e623ddb4 | 1318 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1319 | ret = -ENOMEM; |
1320 | break; | |
1321 | } | |
e623ddb4 FT |
1322 | |
1323 | page = map_data->pages[i / nr_pages]; | |
1324 | page += (i % nr_pages); | |
1325 | ||
1326 | i++; | |
1327 | } else { | |
152e283f | 1328 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1329 | if (!page) { |
1330 | ret = -ENOMEM; | |
1331 | break; | |
1332 | } | |
1da177e4 LT |
1333 | } |
1334 | ||
a3761c3c JG |
1335 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) { |
1336 | if (!map_data) | |
1337 | __free_page(page); | |
1da177e4 | 1338 | break; |
a3761c3c | 1339 | } |
1da177e4 LT |
1340 | |
1341 | len -= bytes; | |
56c451f4 | 1342 | offset = 0; |
1da177e4 LT |
1343 | } |
1344 | ||
1345 | if (ret) | |
1346 | goto cleanup; | |
1347 | ||
2884d0be AV |
1348 | if (map_data) |
1349 | map_data->offset += bio->bi_iter.bi_size; | |
1350 | ||
1da177e4 LT |
1351 | /* |
1352 | * success | |
1353 | */ | |
00e23707 | 1354 | if ((iov_iter_rw(iter) == WRITE && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1355 | (map_data && map_data->from_user)) { |
98a09d61 | 1356 | ret = bio_copy_from_iter(bio, iter); |
c5dec1c3 FT |
1357 | if (ret) |
1358 | goto cleanup; | |
98a09d61 | 1359 | } else { |
f55adad6 KB |
1360 | if (bmd->is_our_pages) |
1361 | zero_fill_bio(bio); | |
98a09d61 | 1362 | iov_iter_advance(iter, bio->bi_iter.bi_size); |
1da177e4 LT |
1363 | } |
1364 | ||
26e49cfc | 1365 | bio->bi_private = bmd; |
2884d0be AV |
1366 | if (map_data && map_data->null_mapped) |
1367 | bio_set_flag(bio, BIO_NULL_MAPPED); | |
1da177e4 LT |
1368 | return bio; |
1369 | cleanup: | |
152e283f | 1370 | if (!map_data) |
1dfa0f68 | 1371 | bio_free_pages(bio); |
1da177e4 LT |
1372 | bio_put(bio); |
1373 | out_bmd: | |
c8db4448 | 1374 | kfree(bmd); |
1da177e4 LT |
1375 | return ERR_PTR(ret); |
1376 | } | |
1377 | ||
37f19e57 CH |
1378 | /** |
1379 | * bio_map_user_iov - map user iovec into bio | |
1380 | * @q: the struct request_queue for the bio | |
1381 | * @iter: iovec iterator | |
1382 | * @gfp_mask: memory allocation flags | |
1383 | * | |
1384 | * Map the user space address into a bio suitable for io to a block | |
1385 | * device. Returns an error pointer in case of error. | |
1386 | */ | |
1387 | struct bio *bio_map_user_iov(struct request_queue *q, | |
e81cef5d | 1388 | struct iov_iter *iter, |
37f19e57 | 1389 | gfp_t gfp_mask) |
1da177e4 | 1390 | { |
26e49cfc | 1391 | int j; |
1da177e4 | 1392 | struct bio *bio; |
076098e5 | 1393 | int ret; |
2b04e8f6 | 1394 | struct bio_vec *bvec; |
6dc4f100 | 1395 | struct bvec_iter_all iter_all; |
1da177e4 | 1396 | |
b282cc76 | 1397 | if (!iov_iter_count(iter)) |
1da177e4 LT |
1398 | return ERR_PTR(-EINVAL); |
1399 | ||
b282cc76 | 1400 | bio = bio_kmalloc(gfp_mask, iov_iter_npages(iter, BIO_MAX_PAGES)); |
1da177e4 LT |
1401 | if (!bio) |
1402 | return ERR_PTR(-ENOMEM); | |
1403 | ||
0a0f1513 | 1404 | while (iov_iter_count(iter)) { |
629e42bc | 1405 | struct page **pages; |
076098e5 AV |
1406 | ssize_t bytes; |
1407 | size_t offs, added = 0; | |
1408 | int npages; | |
1da177e4 | 1409 | |
0a0f1513 | 1410 | bytes = iov_iter_get_pages_alloc(iter, &pages, LONG_MAX, &offs); |
076098e5 AV |
1411 | if (unlikely(bytes <= 0)) { |
1412 | ret = bytes ? bytes : -EFAULT; | |
f1970baf | 1413 | goto out_unmap; |
99172157 | 1414 | } |
f1970baf | 1415 | |
076098e5 | 1416 | npages = DIV_ROUND_UP(offs + bytes, PAGE_SIZE); |
f1970baf | 1417 | |
98f0bc99 AV |
1418 | if (unlikely(offs & queue_dma_alignment(q))) { |
1419 | ret = -EINVAL; | |
1420 | j = 0; | |
1421 | } else { | |
1422 | for (j = 0; j < npages; j++) { | |
1423 | struct page *page = pages[j]; | |
1424 | unsigned int n = PAGE_SIZE - offs; | |
f1970baf | 1425 | |
98f0bc99 AV |
1426 | if (n > bytes) |
1427 | n = bytes; | |
95d78c28 | 1428 | |
19047087 ML |
1429 | if (!__bio_add_pc_page(q, bio, page, n, offs, |
1430 | true)) | |
98f0bc99 | 1431 | break; |
1da177e4 | 1432 | |
98f0bc99 AV |
1433 | added += n; |
1434 | bytes -= n; | |
1435 | offs = 0; | |
1436 | } | |
0a0f1513 | 1437 | iov_iter_advance(iter, added); |
f1970baf | 1438 | } |
1da177e4 | 1439 | /* |
f1970baf | 1440 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1441 | */ |
629e42bc | 1442 | while (j < npages) |
09cbfeaf | 1443 | put_page(pages[j++]); |
629e42bc | 1444 | kvfree(pages); |
e2e115d1 AV |
1445 | /* couldn't stuff something into bio? */ |
1446 | if (bytes) | |
1447 | break; | |
1da177e4 LT |
1448 | } |
1449 | ||
b7c44ed9 | 1450 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1451 | |
1452 | /* | |
5fad1b64 | 1453 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
37f19e57 CH |
1454 | * it would normally disappear when its bi_end_io is run. |
1455 | * however, we need it for the unmap, so grab an extra | |
1456 | * reference to it | |
1457 | */ | |
1458 | bio_get(bio); | |
1da177e4 | 1459 | return bio; |
f1970baf JB |
1460 | |
1461 | out_unmap: | |
2b070cfe | 1462 | bio_for_each_segment_all(bvec, bio, iter_all) { |
2b04e8f6 | 1463 | put_page(bvec->bv_page); |
f1970baf | 1464 | } |
1da177e4 LT |
1465 | bio_put(bio); |
1466 | return ERR_PTR(ret); | |
1467 | } | |
1468 | ||
1da177e4 LT |
1469 | static void __bio_unmap_user(struct bio *bio) |
1470 | { | |
1471 | struct bio_vec *bvec; | |
6dc4f100 | 1472 | struct bvec_iter_all iter_all; |
1da177e4 LT |
1473 | |
1474 | /* | |
1475 | * make sure we dirty pages we wrote to | |
1476 | */ | |
2b070cfe | 1477 | bio_for_each_segment_all(bvec, bio, iter_all) { |
1da177e4 LT |
1478 | if (bio_data_dir(bio) == READ) |
1479 | set_page_dirty_lock(bvec->bv_page); | |
1480 | ||
09cbfeaf | 1481 | put_page(bvec->bv_page); |
1da177e4 LT |
1482 | } |
1483 | ||
1484 | bio_put(bio); | |
1485 | } | |
1486 | ||
1487 | /** | |
1488 | * bio_unmap_user - unmap a bio | |
1489 | * @bio: the bio being unmapped | |
1490 | * | |
5fad1b64 BVA |
1491 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
1492 | * process context. | |
1da177e4 LT |
1493 | * |
1494 | * bio_unmap_user() may sleep. | |
1495 | */ | |
1496 | void bio_unmap_user(struct bio *bio) | |
1497 | { | |
1498 | __bio_unmap_user(bio); | |
1499 | bio_put(bio); | |
1500 | } | |
1501 | ||
4246a0b6 | 1502 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1503 | { |
b823825e | 1504 | bio_put(bio); |
b823825e JA |
1505 | } |
1506 | ||
75c72b83 CH |
1507 | /** |
1508 | * bio_map_kern - map kernel address into bio | |
1509 | * @q: the struct request_queue for the bio | |
1510 | * @data: pointer to buffer to map | |
1511 | * @len: length in bytes | |
1512 | * @gfp_mask: allocation flags for bio allocation | |
1513 | * | |
1514 | * Map the kernel address into a bio suitable for io to a block | |
1515 | * device. Returns an error pointer in case of error. | |
1516 | */ | |
1517 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1518 | gfp_t gfp_mask) | |
df46b9a4 MC |
1519 | { |
1520 | unsigned long kaddr = (unsigned long)data; | |
1521 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1522 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1523 | const int nr_pages = end - start; | |
1524 | int offset, i; | |
1525 | struct bio *bio; | |
1526 | ||
a9e9dc24 | 1527 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1528 | if (!bio) |
1529 | return ERR_PTR(-ENOMEM); | |
1530 | ||
1531 | offset = offset_in_page(kaddr); | |
1532 | for (i = 0; i < nr_pages; i++) { | |
1533 | unsigned int bytes = PAGE_SIZE - offset; | |
1534 | ||
1535 | if (len <= 0) | |
1536 | break; | |
1537 | ||
1538 | if (bytes > len) | |
1539 | bytes = len; | |
1540 | ||
defd94b7 | 1541 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1542 | offset) < bytes) { |
1543 | /* we don't support partial mappings */ | |
1544 | bio_put(bio); | |
1545 | return ERR_PTR(-EINVAL); | |
1546 | } | |
df46b9a4 MC |
1547 | |
1548 | data += bytes; | |
1549 | len -= bytes; | |
1550 | offset = 0; | |
1551 | } | |
1552 | ||
b823825e | 1553 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1554 | return bio; |
1555 | } | |
a112a71d | 1556 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1557 | |
4246a0b6 | 1558 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1559 | { |
1dfa0f68 CH |
1560 | bio_free_pages(bio); |
1561 | bio_put(bio); | |
1562 | } | |
1563 | ||
4246a0b6 | 1564 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1565 | { |
42d2683a | 1566 | char *p = bio->bi_private; |
1dfa0f68 | 1567 | struct bio_vec *bvec; |
6dc4f100 | 1568 | struct bvec_iter_all iter_all; |
68154e90 | 1569 | |
2b070cfe | 1570 | bio_for_each_segment_all(bvec, bio, iter_all) { |
1dfa0f68 | 1571 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1572 | p += bvec->bv_len; |
68154e90 FT |
1573 | } |
1574 | ||
4246a0b6 | 1575 | bio_copy_kern_endio(bio); |
68154e90 FT |
1576 | } |
1577 | ||
1578 | /** | |
1579 | * bio_copy_kern - copy kernel address into bio | |
1580 | * @q: the struct request_queue for the bio | |
1581 | * @data: pointer to buffer to copy | |
1582 | * @len: length in bytes | |
1583 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1584 | * @reading: data direction is READ |
68154e90 FT |
1585 | * |
1586 | * copy the kernel address into a bio suitable for io to a block | |
1587 | * device. Returns an error pointer in case of error. | |
1588 | */ | |
1589 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1590 | gfp_t gfp_mask, int reading) | |
1591 | { | |
42d2683a CH |
1592 | unsigned long kaddr = (unsigned long)data; |
1593 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1594 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1595 | struct bio *bio; |
1596 | void *p = data; | |
1dfa0f68 | 1597 | int nr_pages = 0; |
68154e90 | 1598 | |
42d2683a CH |
1599 | /* |
1600 | * Overflow, abort | |
1601 | */ | |
1602 | if (end < start) | |
1603 | return ERR_PTR(-EINVAL); | |
68154e90 | 1604 | |
42d2683a CH |
1605 | nr_pages = end - start; |
1606 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1607 | if (!bio) | |
1608 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1609 | |
42d2683a CH |
1610 | while (len) { |
1611 | struct page *page; | |
1612 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1613 | |
42d2683a CH |
1614 | if (bytes > len) |
1615 | bytes = len; | |
1616 | ||
1617 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1618 | if (!page) | |
1619 | goto cleanup; | |
1620 | ||
1621 | if (!reading) | |
1622 | memcpy(page_address(page), p, bytes); | |
1623 | ||
1624 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1625 | break; | |
1626 | ||
1627 | len -= bytes; | |
1628 | p += bytes; | |
68154e90 FT |
1629 | } |
1630 | ||
1dfa0f68 CH |
1631 | if (reading) { |
1632 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1633 | bio->bi_private = data; | |
1634 | } else { | |
1635 | bio->bi_end_io = bio_copy_kern_endio; | |
1dfa0f68 | 1636 | } |
76029ff3 | 1637 | |
68154e90 | 1638 | return bio; |
42d2683a CH |
1639 | |
1640 | cleanup: | |
1dfa0f68 | 1641 | bio_free_pages(bio); |
42d2683a CH |
1642 | bio_put(bio); |
1643 | return ERR_PTR(-ENOMEM); | |
68154e90 FT |
1644 | } |
1645 | ||
1da177e4 LT |
1646 | /* |
1647 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1648 | * for performing direct-IO in BIOs. | |
1649 | * | |
1650 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1651 | * because the required locks are not interrupt-safe. So what we can do is to | |
1652 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1653 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1654 | * in process context. | |
1655 | * | |
1656 | * We special-case compound pages here: normally this means reads into hugetlb | |
1657 | * pages. The logic in here doesn't really work right for compound pages | |
1658 | * because the VM does not uniformly chase down the head page in all cases. | |
1659 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1660 | * handle them at all. So we skip compound pages here at an early stage. | |
1661 | * | |
1662 | * Note that this code is very hard to test under normal circumstances because | |
1663 | * direct-io pins the pages with get_user_pages(). This makes | |
1664 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1665 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1666 | * pagecache. |
1667 | * | |
1668 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1669 | * deferred bio dirtying paths. | |
1670 | */ | |
1671 | ||
1672 | /* | |
1673 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1674 | */ | |
1675 | void bio_set_pages_dirty(struct bio *bio) | |
1676 | { | |
cb34e057 | 1677 | struct bio_vec *bvec; |
6dc4f100 | 1678 | struct bvec_iter_all iter_all; |
1da177e4 | 1679 | |
2b070cfe | 1680 | bio_for_each_segment_all(bvec, bio, iter_all) { |
3bb50983 CH |
1681 | if (!PageCompound(bvec->bv_page)) |
1682 | set_page_dirty_lock(bvec->bv_page); | |
1da177e4 LT |
1683 | } |
1684 | } | |
1685 | ||
1da177e4 LT |
1686 | /* |
1687 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1688 | * If they are, then fine. If, however, some pages are clean then they must | |
1689 | * have been written out during the direct-IO read. So we take another ref on | |
24d5493f | 1690 | * the BIO and re-dirty the pages in process context. |
1da177e4 LT |
1691 | * |
1692 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1693 | * here on. It will run one put_page() against each page and will run one |
1694 | * bio_put() against the BIO. | |
1da177e4 LT |
1695 | */ |
1696 | ||
65f27f38 | 1697 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1698 | |
65f27f38 | 1699 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1700 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1701 | static struct bio *bio_dirty_list; | |
1702 | ||
1703 | /* | |
1704 | * This runs in process context | |
1705 | */ | |
65f27f38 | 1706 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 | 1707 | { |
24d5493f | 1708 | struct bio *bio, *next; |
1da177e4 | 1709 | |
24d5493f CH |
1710 | spin_lock_irq(&bio_dirty_lock); |
1711 | next = bio_dirty_list; | |
1da177e4 | 1712 | bio_dirty_list = NULL; |
24d5493f | 1713 | spin_unlock_irq(&bio_dirty_lock); |
1da177e4 | 1714 | |
24d5493f CH |
1715 | while ((bio = next) != NULL) { |
1716 | next = bio->bi_private; | |
1da177e4 LT |
1717 | |
1718 | bio_set_pages_dirty(bio); | |
399254aa JA |
1719 | if (!bio_flagged(bio, BIO_NO_PAGE_REF)) |
1720 | bio_release_pages(bio); | |
1da177e4 | 1721 | bio_put(bio); |
1da177e4 LT |
1722 | } |
1723 | } | |
1724 | ||
1725 | void bio_check_pages_dirty(struct bio *bio) | |
1726 | { | |
cb34e057 | 1727 | struct bio_vec *bvec; |
24d5493f | 1728 | unsigned long flags; |
6dc4f100 | 1729 | struct bvec_iter_all iter_all; |
1da177e4 | 1730 | |
2b070cfe | 1731 | bio_for_each_segment_all(bvec, bio, iter_all) { |
24d5493f CH |
1732 | if (!PageDirty(bvec->bv_page) && !PageCompound(bvec->bv_page)) |
1733 | goto defer; | |
1da177e4 LT |
1734 | } |
1735 | ||
399254aa JA |
1736 | if (!bio_flagged(bio, BIO_NO_PAGE_REF)) |
1737 | bio_release_pages(bio); | |
24d5493f CH |
1738 | bio_put(bio); |
1739 | return; | |
1740 | defer: | |
1741 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1742 | bio->bi_private = bio_dirty_list; | |
1743 | bio_dirty_list = bio; | |
1744 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1745 | schedule_work(&bio_dirty_work); | |
1da177e4 LT |
1746 | } |
1747 | ||
5b18b5a7 MP |
1748 | void update_io_ticks(struct hd_struct *part, unsigned long now) |
1749 | { | |
1750 | unsigned long stamp; | |
1751 | again: | |
1752 | stamp = READ_ONCE(part->stamp); | |
1753 | if (unlikely(stamp != now)) { | |
1754 | if (likely(cmpxchg(&part->stamp, stamp, now) == stamp)) { | |
1755 | __part_stat_add(part, io_ticks, 1); | |
1756 | } | |
1757 | } | |
1758 | if (part->partno) { | |
1759 | part = &part_to_disk(part)->part0; | |
1760 | goto again; | |
1761 | } | |
1762 | } | |
1da177e4 | 1763 | |
ddcf35d3 | 1764 | void generic_start_io_acct(struct request_queue *q, int op, |
d62e26b3 | 1765 | unsigned long sectors, struct hd_struct *part) |
394ffa50 | 1766 | { |
ddcf35d3 | 1767 | const int sgrp = op_stat_group(op); |
394ffa50 | 1768 | |
112f158f MS |
1769 | part_stat_lock(); |
1770 | ||
5b18b5a7 | 1771 | update_io_ticks(part, jiffies); |
112f158f MS |
1772 | part_stat_inc(part, ios[sgrp]); |
1773 | part_stat_add(part, sectors[sgrp], sectors); | |
ddcf35d3 | 1774 | part_inc_in_flight(q, part, op_is_write(op)); |
394ffa50 GZ |
1775 | |
1776 | part_stat_unlock(); | |
1777 | } | |
1778 | EXPORT_SYMBOL(generic_start_io_acct); | |
1779 | ||
ddcf35d3 | 1780 | void generic_end_io_acct(struct request_queue *q, int req_op, |
d62e26b3 | 1781 | struct hd_struct *part, unsigned long start_time) |
394ffa50 | 1782 | { |
5b18b5a7 MP |
1783 | unsigned long now = jiffies; |
1784 | unsigned long duration = now - start_time; | |
ddcf35d3 | 1785 | const int sgrp = op_stat_group(req_op); |
394ffa50 | 1786 | |
112f158f MS |
1787 | part_stat_lock(); |
1788 | ||
5b18b5a7 | 1789 | update_io_ticks(part, now); |
112f158f | 1790 | part_stat_add(part, nsecs[sgrp], jiffies_to_nsecs(duration)); |
5b18b5a7 | 1791 | part_stat_add(part, time_in_queue, duration); |
ddcf35d3 | 1792 | part_dec_in_flight(q, part, op_is_write(req_op)); |
394ffa50 GZ |
1793 | |
1794 | part_stat_unlock(); | |
1795 | } | |
1796 | EXPORT_SYMBOL(generic_end_io_acct); | |
1797 | ||
2d4dc890 IL |
1798 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1799 | void bio_flush_dcache_pages(struct bio *bi) | |
1800 | { | |
7988613b KO |
1801 | struct bio_vec bvec; |
1802 | struct bvec_iter iter; | |
2d4dc890 | 1803 | |
7988613b KO |
1804 | bio_for_each_segment(bvec, bi, iter) |
1805 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1806 | } |
1807 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1808 | #endif | |
1809 | ||
c4cf5261 JA |
1810 | static inline bool bio_remaining_done(struct bio *bio) |
1811 | { | |
1812 | /* | |
1813 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1814 | * we always end io on the first invocation. | |
1815 | */ | |
1816 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1817 | return true; | |
1818 | ||
1819 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1820 | ||
326e1dbb | 1821 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1822 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1823 | return true; |
326e1dbb | 1824 | } |
c4cf5261 JA |
1825 | |
1826 | return false; | |
1827 | } | |
1828 | ||
1da177e4 LT |
1829 | /** |
1830 | * bio_endio - end I/O on a bio | |
1831 | * @bio: bio | |
1da177e4 LT |
1832 | * |
1833 | * Description: | |
4246a0b6 CH |
1834 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1835 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1836 | * bio unless they own it and thus know that it has an end_io function. | |
fbbaf700 N |
1837 | * |
1838 | * bio_endio() can be called several times on a bio that has been chained | |
1839 | * using bio_chain(). The ->bi_end_io() function will only be called the | |
1840 | * last time. At this point the BLK_TA_COMPLETE tracing event will be | |
1841 | * generated if BIO_TRACE_COMPLETION is set. | |
1da177e4 | 1842 | **/ |
4246a0b6 | 1843 | void bio_endio(struct bio *bio) |
1da177e4 | 1844 | { |
ba8c6967 | 1845 | again: |
2b885517 | 1846 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1847 | return; |
7c20f116 CH |
1848 | if (!bio_integrity_endio(bio)) |
1849 | return; | |
1da177e4 | 1850 | |
67b42d0b JB |
1851 | if (bio->bi_disk) |
1852 | rq_qos_done_bio(bio->bi_disk->queue, bio); | |
1853 | ||
ba8c6967 CH |
1854 | /* |
1855 | * Need to have a real endio function for chained bios, otherwise | |
1856 | * various corner cases will break (like stacking block devices that | |
1857 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1858 | * recursion and blowing the stack. Tail call optimization would | |
1859 | * handle this, but compiling with frame pointers also disables | |
1860 | * gcc's sibling call optimization. | |
1861 | */ | |
1862 | if (bio->bi_end_io == bio_chain_endio) { | |
1863 | bio = __bio_chain_endio(bio); | |
1864 | goto again; | |
196d38bc | 1865 | } |
ba8c6967 | 1866 | |
74d46992 CH |
1867 | if (bio->bi_disk && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
1868 | trace_block_bio_complete(bio->bi_disk->queue, bio, | |
a462b950 | 1869 | blk_status_to_errno(bio->bi_status)); |
fbbaf700 N |
1870 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
1871 | } | |
1872 | ||
9e234eea | 1873 | blk_throtl_bio_endio(bio); |
b222dd2f SL |
1874 | /* release cgroup info */ |
1875 | bio_uninit(bio); | |
ba8c6967 CH |
1876 | if (bio->bi_end_io) |
1877 | bio->bi_end_io(bio); | |
1da177e4 | 1878 | } |
a112a71d | 1879 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1880 | |
20d0189b KO |
1881 | /** |
1882 | * bio_split - split a bio | |
1883 | * @bio: bio to split | |
1884 | * @sectors: number of sectors to split from the front of @bio | |
1885 | * @gfp: gfp mask | |
1886 | * @bs: bio set to allocate from | |
1887 | * | |
1888 | * Allocates and returns a new bio which represents @sectors from the start of | |
1889 | * @bio, and updates @bio to represent the remaining sectors. | |
1890 | * | |
f3f5da62 MP |
1891 | * Unless this is a discard request the newly allocated bio will point |
1892 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1893 | * @bio is not freed before the split. | |
20d0189b KO |
1894 | */ |
1895 | struct bio *bio_split(struct bio *bio, int sectors, | |
1896 | gfp_t gfp, struct bio_set *bs) | |
1897 | { | |
f341a4d3 | 1898 | struct bio *split; |
20d0189b KO |
1899 | |
1900 | BUG_ON(sectors <= 0); | |
1901 | BUG_ON(sectors >= bio_sectors(bio)); | |
1902 | ||
f9d03f96 | 1903 | split = bio_clone_fast(bio, gfp, bs); |
20d0189b KO |
1904 | if (!split) |
1905 | return NULL; | |
1906 | ||
1907 | split->bi_iter.bi_size = sectors << 9; | |
1908 | ||
1909 | if (bio_integrity(split)) | |
fbd08e76 | 1910 | bio_integrity_trim(split); |
20d0189b KO |
1911 | |
1912 | bio_advance(bio, split->bi_iter.bi_size); | |
1913 | ||
fbbaf700 | 1914 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
20d59023 | 1915 | bio_set_flag(split, BIO_TRACE_COMPLETION); |
fbbaf700 | 1916 | |
20d0189b KO |
1917 | return split; |
1918 | } | |
1919 | EXPORT_SYMBOL(bio_split); | |
1920 | ||
6678d83f KO |
1921 | /** |
1922 | * bio_trim - trim a bio | |
1923 | * @bio: bio to trim | |
1924 | * @offset: number of sectors to trim from the front of @bio | |
1925 | * @size: size we want to trim @bio to, in sectors | |
1926 | */ | |
1927 | void bio_trim(struct bio *bio, int offset, int size) | |
1928 | { | |
1929 | /* 'bio' is a cloned bio which we need to trim to match | |
1930 | * the given offset and size. | |
6678d83f | 1931 | */ |
6678d83f KO |
1932 | |
1933 | size <<= 9; | |
4f024f37 | 1934 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1935 | return; |
1936 | ||
b7c44ed9 | 1937 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1938 | |
1939 | bio_advance(bio, offset << 9); | |
1940 | ||
4f024f37 | 1941 | bio->bi_iter.bi_size = size; |
376a78ab DM |
1942 | |
1943 | if (bio_integrity(bio)) | |
fbd08e76 | 1944 | bio_integrity_trim(bio); |
376a78ab | 1945 | |
6678d83f KO |
1946 | } |
1947 | EXPORT_SYMBOL_GPL(bio_trim); | |
1948 | ||
1da177e4 LT |
1949 | /* |
1950 | * create memory pools for biovec's in a bio_set. | |
1951 | * use the global biovec slabs created for general use. | |
1952 | */ | |
8aa6ba2f | 1953 | int biovec_init_pool(mempool_t *pool, int pool_entries) |
1da177e4 | 1954 | { |
ed996a52 | 1955 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
1da177e4 | 1956 | |
8aa6ba2f | 1957 | return mempool_init_slab_pool(pool, pool_entries, bp->slab); |
1da177e4 LT |
1958 | } |
1959 | ||
917a38c7 KO |
1960 | /* |
1961 | * bioset_exit - exit a bioset initialized with bioset_init() | |
1962 | * | |
1963 | * May be called on a zeroed but uninitialized bioset (i.e. allocated with | |
1964 | * kzalloc()). | |
1965 | */ | |
1966 | void bioset_exit(struct bio_set *bs) | |
1da177e4 | 1967 | { |
df2cb6da KO |
1968 | if (bs->rescue_workqueue) |
1969 | destroy_workqueue(bs->rescue_workqueue); | |
917a38c7 | 1970 | bs->rescue_workqueue = NULL; |
df2cb6da | 1971 | |
8aa6ba2f KO |
1972 | mempool_exit(&bs->bio_pool); |
1973 | mempool_exit(&bs->bvec_pool); | |
9f060e22 | 1974 | |
7878cba9 | 1975 | bioset_integrity_free(bs); |
917a38c7 KO |
1976 | if (bs->bio_slab) |
1977 | bio_put_slab(bs); | |
1978 | bs->bio_slab = NULL; | |
1979 | } | |
1980 | EXPORT_SYMBOL(bioset_exit); | |
1da177e4 | 1981 | |
917a38c7 KO |
1982 | /** |
1983 | * bioset_init - Initialize a bio_set | |
dad08527 | 1984 | * @bs: pool to initialize |
917a38c7 KO |
1985 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
1986 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1987 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS | |
1988 | * and %BIOSET_NEED_RESCUER | |
1989 | * | |
dad08527 KO |
1990 | * Description: |
1991 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1992 | * to ask for a number of bytes to be allocated in front of the bio. | |
1993 | * Front pad allocation is useful for embedding the bio inside | |
1994 | * another structure, to avoid allocating extra data to go with the bio. | |
1995 | * Note that the bio must be embedded at the END of that structure always, | |
1996 | * or things will break badly. | |
1997 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated | |
1998 | * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast(). | |
1999 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to | |
2000 | * dispatch queued requests when the mempool runs out of space. | |
2001 | * | |
917a38c7 KO |
2002 | */ |
2003 | int bioset_init(struct bio_set *bs, | |
2004 | unsigned int pool_size, | |
2005 | unsigned int front_pad, | |
2006 | int flags) | |
2007 | { | |
2008 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); | |
2009 | ||
2010 | bs->front_pad = front_pad; | |
2011 | ||
2012 | spin_lock_init(&bs->rescue_lock); | |
2013 | bio_list_init(&bs->rescue_list); | |
2014 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
2015 | ||
2016 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); | |
2017 | if (!bs->bio_slab) | |
2018 | return -ENOMEM; | |
2019 | ||
2020 | if (mempool_init_slab_pool(&bs->bio_pool, pool_size, bs->bio_slab)) | |
2021 | goto bad; | |
2022 | ||
2023 | if ((flags & BIOSET_NEED_BVECS) && | |
2024 | biovec_init_pool(&bs->bvec_pool, pool_size)) | |
2025 | goto bad; | |
2026 | ||
2027 | if (!(flags & BIOSET_NEED_RESCUER)) | |
2028 | return 0; | |
2029 | ||
2030 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); | |
2031 | if (!bs->rescue_workqueue) | |
2032 | goto bad; | |
2033 | ||
2034 | return 0; | |
2035 | bad: | |
2036 | bioset_exit(bs); | |
2037 | return -ENOMEM; | |
2038 | } | |
2039 | EXPORT_SYMBOL(bioset_init); | |
2040 | ||
28e89fd9 JA |
2041 | /* |
2042 | * Initialize and setup a new bio_set, based on the settings from | |
2043 | * another bio_set. | |
2044 | */ | |
2045 | int bioset_init_from_src(struct bio_set *bs, struct bio_set *src) | |
2046 | { | |
2047 | int flags; | |
2048 | ||
2049 | flags = 0; | |
2050 | if (src->bvec_pool.min_nr) | |
2051 | flags |= BIOSET_NEED_BVECS; | |
2052 | if (src->rescue_workqueue) | |
2053 | flags |= BIOSET_NEED_RESCUER; | |
2054 | ||
2055 | return bioset_init(bs, src->bio_pool.min_nr, src->front_pad, flags); | |
2056 | } | |
2057 | EXPORT_SYMBOL(bioset_init_from_src); | |
2058 | ||
852c788f | 2059 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 | 2060 | |
74b7c02a | 2061 | /** |
2268c0fe | 2062 | * bio_disassociate_blkg - puts back the blkg reference if associated |
74b7c02a | 2063 | * @bio: target bio |
74b7c02a | 2064 | * |
2268c0fe | 2065 | * Helper to disassociate the blkg from @bio if a blkg is associated. |
74b7c02a | 2066 | */ |
2268c0fe | 2067 | void bio_disassociate_blkg(struct bio *bio) |
74b7c02a | 2068 | { |
2268c0fe DZ |
2069 | if (bio->bi_blkg) { |
2070 | blkg_put(bio->bi_blkg); | |
2071 | bio->bi_blkg = NULL; | |
2072 | } | |
74b7c02a | 2073 | } |
892ad71f | 2074 | EXPORT_SYMBOL_GPL(bio_disassociate_blkg); |
74b7c02a | 2075 | |
a7b39b4e | 2076 | /** |
2268c0fe | 2077 | * __bio_associate_blkg - associate a bio with the a blkg |
a7b39b4e | 2078 | * @bio: target bio |
b5f2954d | 2079 | * @blkg: the blkg to associate |
b5f2954d | 2080 | * |
beea9da0 DZ |
2081 | * This tries to associate @bio with the specified @blkg. Association failure |
2082 | * is handled by walking up the blkg tree. Therefore, the blkg associated can | |
2083 | * be anything between @blkg and the root_blkg. This situation only happens | |
2084 | * when a cgroup is dying and then the remaining bios will spill to the closest | |
2085 | * alive blkg. | |
a7b39b4e | 2086 | * |
beea9da0 DZ |
2087 | * A reference will be taken on the @blkg and will be released when @bio is |
2088 | * freed. | |
a7b39b4e | 2089 | */ |
2268c0fe | 2090 | static void __bio_associate_blkg(struct bio *bio, struct blkcg_gq *blkg) |
a7b39b4e | 2091 | { |
2268c0fe DZ |
2092 | bio_disassociate_blkg(bio); |
2093 | ||
7754f669 | 2094 | bio->bi_blkg = blkg_tryget_closest(blkg); |
a7b39b4e DZF |
2095 | } |
2096 | ||
d459d853 | 2097 | /** |
fd42df30 | 2098 | * bio_associate_blkg_from_css - associate a bio with a specified css |
d459d853 | 2099 | * @bio: target bio |
fd42df30 | 2100 | * @css: target css |
d459d853 | 2101 | * |
fd42df30 | 2102 | * Associate @bio with the blkg found by combining the css's blkg and the |
fc5a828b DZ |
2103 | * request_queue of the @bio. This falls back to the queue's root_blkg if |
2104 | * the association fails with the css. | |
d459d853 | 2105 | */ |
fd42df30 DZ |
2106 | void bio_associate_blkg_from_css(struct bio *bio, |
2107 | struct cgroup_subsys_state *css) | |
d459d853 | 2108 | { |
fc5a828b DZ |
2109 | struct request_queue *q = bio->bi_disk->queue; |
2110 | struct blkcg_gq *blkg; | |
2111 | ||
2112 | rcu_read_lock(); | |
2113 | ||
2114 | if (!css || !css->parent) | |
2115 | blkg = q->root_blkg; | |
2116 | else | |
2117 | blkg = blkg_lookup_create(css_to_blkcg(css), q); | |
2118 | ||
2119 | __bio_associate_blkg(bio, blkg); | |
2120 | ||
2121 | rcu_read_unlock(); | |
d459d853 | 2122 | } |
fd42df30 | 2123 | EXPORT_SYMBOL_GPL(bio_associate_blkg_from_css); |
d459d853 | 2124 | |
6a7f6d86 | 2125 | #ifdef CONFIG_MEMCG |
852c788f | 2126 | /** |
6a7f6d86 | 2127 | * bio_associate_blkg_from_page - associate a bio with the page's blkg |
852c788f | 2128 | * @bio: target bio |
6a7f6d86 DZ |
2129 | * @page: the page to lookup the blkcg from |
2130 | * | |
2131 | * Associate @bio with the blkg from @page's owning memcg and the respective | |
fc5a828b DZ |
2132 | * request_queue. If cgroup_e_css returns %NULL, fall back to the queue's |
2133 | * root_blkg. | |
852c788f | 2134 | */ |
6a7f6d86 | 2135 | void bio_associate_blkg_from_page(struct bio *bio, struct page *page) |
852c788f | 2136 | { |
6a7f6d86 DZ |
2137 | struct cgroup_subsys_state *css; |
2138 | ||
6a7f6d86 DZ |
2139 | if (!page->mem_cgroup) |
2140 | return; | |
2141 | ||
fc5a828b DZ |
2142 | rcu_read_lock(); |
2143 | ||
2144 | css = cgroup_e_css(page->mem_cgroup->css.cgroup, &io_cgrp_subsys); | |
2145 | bio_associate_blkg_from_css(bio, css); | |
2146 | ||
2147 | rcu_read_unlock(); | |
6a7f6d86 DZ |
2148 | } |
2149 | #endif /* CONFIG_MEMCG */ | |
2150 | ||
2268c0fe DZ |
2151 | /** |
2152 | * bio_associate_blkg - associate a bio with a blkg | |
2153 | * @bio: target bio | |
2154 | * | |
2155 | * Associate @bio with the blkg found from the bio's css and request_queue. | |
2156 | * If one is not found, bio_lookup_blkg() creates the blkg. If a blkg is | |
2157 | * already associated, the css is reused and association redone as the | |
2158 | * request_queue may have changed. | |
2159 | */ | |
2160 | void bio_associate_blkg(struct bio *bio) | |
2161 | { | |
fc5a828b | 2162 | struct cgroup_subsys_state *css; |
2268c0fe DZ |
2163 | |
2164 | rcu_read_lock(); | |
2165 | ||
db6638d7 | 2166 | if (bio->bi_blkg) |
fc5a828b | 2167 | css = &bio_blkcg(bio)->css; |
db6638d7 | 2168 | else |
fc5a828b | 2169 | css = blkcg_css(); |
2268c0fe | 2170 | |
fc5a828b | 2171 | bio_associate_blkg_from_css(bio, css); |
2268c0fe DZ |
2172 | |
2173 | rcu_read_unlock(); | |
852c788f | 2174 | } |
5cdf2e3f | 2175 | EXPORT_SYMBOL_GPL(bio_associate_blkg); |
852c788f | 2176 | |
20bd723e | 2177 | /** |
db6638d7 | 2178 | * bio_clone_blkg_association - clone blkg association from src to dst bio |
20bd723e PV |
2179 | * @dst: destination bio |
2180 | * @src: source bio | |
2181 | */ | |
db6638d7 | 2182 | void bio_clone_blkg_association(struct bio *dst, struct bio *src) |
20bd723e | 2183 | { |
6ab21879 DZ |
2184 | rcu_read_lock(); |
2185 | ||
fc5a828b | 2186 | if (src->bi_blkg) |
2268c0fe | 2187 | __bio_associate_blkg(dst, src->bi_blkg); |
6ab21879 DZ |
2188 | |
2189 | rcu_read_unlock(); | |
20bd723e | 2190 | } |
db6638d7 | 2191 | EXPORT_SYMBOL_GPL(bio_clone_blkg_association); |
852c788f TH |
2192 | #endif /* CONFIG_BLK_CGROUP */ |
2193 | ||
1da177e4 LT |
2194 | static void __init biovec_init_slabs(void) |
2195 | { | |
2196 | int i; | |
2197 | ||
ed996a52 | 2198 | for (i = 0; i < BVEC_POOL_NR; i++) { |
1da177e4 LT |
2199 | int size; |
2200 | struct biovec_slab *bvs = bvec_slabs + i; | |
2201 | ||
a7fcd37c JA |
2202 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2203 | bvs->slab = NULL; | |
2204 | continue; | |
2205 | } | |
a7fcd37c | 2206 | |
1da177e4 LT |
2207 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2208 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2209 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2210 | } |
2211 | } | |
2212 | ||
2213 | static int __init init_bio(void) | |
2214 | { | |
bb799ca0 JA |
2215 | bio_slab_max = 2; |
2216 | bio_slab_nr = 0; | |
6396bb22 KC |
2217 | bio_slabs = kcalloc(bio_slab_max, sizeof(struct bio_slab), |
2218 | GFP_KERNEL); | |
2b24e6f6 JT |
2219 | |
2220 | BUILD_BUG_ON(BIO_FLAG_LAST > BVEC_POOL_OFFSET); | |
2221 | ||
bb799ca0 JA |
2222 | if (!bio_slabs) |
2223 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2224 | |
7878cba9 | 2225 | bio_integrity_init(); |
1da177e4 LT |
2226 | biovec_init_slabs(); |
2227 | ||
f4f8154a | 2228 | if (bioset_init(&fs_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS)) |
1da177e4 LT |
2229 | panic("bio: can't allocate bios\n"); |
2230 | ||
f4f8154a | 2231 | if (bioset_integrity_create(&fs_bio_set, BIO_POOL_SIZE)) |
a91a2785 MP |
2232 | panic("bio: can't create integrity pool\n"); |
2233 | ||
1da177e4 LT |
2234 | return 0; |
2235 | } | |
1da177e4 | 2236 | subsys_initcall(init_bio); |