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 | ||
1da177e4 | 650 | /** |
c66a14d0 KO |
651 | * bio_add_pc_page - attempt to add page to bio |
652 | * @q: the target queue | |
653 | * @bio: destination bio | |
654 | * @page: page to add | |
655 | * @len: vec entry length | |
656 | * @offset: vec entry offset | |
1da177e4 | 657 | * |
c66a14d0 KO |
658 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
659 | * number of reasons, such as the bio being full or target block device | |
660 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
661 | * so it is always possible to add a single page to an empty bio. | |
662 | * | |
663 | * This should only be used by REQ_PC bios. | |
1da177e4 | 664 | */ |
c66a14d0 KO |
665 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
666 | *page, unsigned int len, unsigned int offset) | |
1da177e4 LT |
667 | { |
668 | int retried_segments = 0; | |
669 | struct bio_vec *bvec; | |
670 | ||
671 | /* | |
672 | * cloned bio must not modify vec list | |
673 | */ | |
674 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
675 | return 0; | |
676 | ||
c66a14d0 | 677 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
678 | return 0; |
679 | ||
80cfd548 JA |
680 | /* |
681 | * For filesystems with a blocksize smaller than the pagesize | |
682 | * we will often be called with the same page as last time and | |
683 | * a consecutive offset. Optimize this special case. | |
684 | */ | |
685 | if (bio->bi_vcnt > 0) { | |
686 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
687 | ||
688 | if (page == prev->bv_page && | |
689 | offset == prev->bv_offset + prev->bv_len) { | |
690 | prev->bv_len += len; | |
fcbf6a08 | 691 | bio->bi_iter.bi_size += len; |
80cfd548 JA |
692 | goto done; |
693 | } | |
66cb45aa JA |
694 | |
695 | /* | |
696 | * If the queue doesn't support SG gaps and adding this | |
697 | * offset would create a gap, disallow it. | |
698 | */ | |
03100aad | 699 | if (bvec_gap_to_prev(q, prev, offset)) |
66cb45aa | 700 | return 0; |
80cfd548 JA |
701 | } |
702 | ||
0aa69fd3 | 703 | if (bio_full(bio)) |
1da177e4 LT |
704 | return 0; |
705 | ||
706 | /* | |
fcbf6a08 ML |
707 | * setup the new entry, we might clear it again later if we |
708 | * cannot add the page | |
709 | */ | |
710 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
711 | bvec->bv_page = page; | |
712 | bvec->bv_len = len; | |
713 | bvec->bv_offset = offset; | |
714 | bio->bi_vcnt++; | |
715 | bio->bi_phys_segments++; | |
716 | bio->bi_iter.bi_size += len; | |
717 | ||
718 | /* | |
719 | * Perform a recount if the number of segments is greater | |
720 | * than queue_max_segments(q). | |
1da177e4 LT |
721 | */ |
722 | ||
fcbf6a08 | 723 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
1da177e4 LT |
724 | |
725 | if (retried_segments) | |
fcbf6a08 | 726 | goto failed; |
1da177e4 LT |
727 | |
728 | retried_segments = 1; | |
729 | blk_recount_segments(q, bio); | |
730 | } | |
731 | ||
1da177e4 | 732 | /* If we may be able to merge these biovecs, force a recount */ |
3dccdae5 | 733 | if (bio->bi_vcnt > 1 && biovec_phys_mergeable(q, bvec - 1, bvec)) |
b7c44ed9 | 734 | bio_clear_flag(bio, BIO_SEG_VALID); |
1da177e4 | 735 | |
80cfd548 | 736 | done: |
1da177e4 | 737 | return len; |
fcbf6a08 ML |
738 | |
739 | failed: | |
740 | bvec->bv_page = NULL; | |
741 | bvec->bv_len = 0; | |
742 | bvec->bv_offset = 0; | |
743 | bio->bi_vcnt--; | |
744 | bio->bi_iter.bi_size -= len; | |
745 | blk_recount_segments(q, bio); | |
746 | return 0; | |
1da177e4 | 747 | } |
a112a71d | 748 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 749 | |
1da177e4 | 750 | /** |
0aa69fd3 CH |
751 | * __bio_try_merge_page - try appending data to an existing bvec. |
752 | * @bio: destination bio | |
753 | * @page: page to add | |
754 | * @len: length of the data to add | |
755 | * @off: offset of the data in @page | |
07173c3e ML |
756 | * @same_page: if %true only merge if the new data is in the same physical |
757 | * page as the last segment of the bio. | |
1da177e4 | 758 | * |
0aa69fd3 CH |
759 | * Try to add the data at @page + @off to the last bvec of @bio. This is a |
760 | * a useful optimisation for file systems with a block size smaller than the | |
761 | * page size. | |
762 | * | |
763 | * Return %true on success or %false on failure. | |
1da177e4 | 764 | */ |
0aa69fd3 | 765 | bool __bio_try_merge_page(struct bio *bio, struct page *page, |
07173c3e | 766 | unsigned int len, unsigned int off, bool same_page) |
1da177e4 | 767 | { |
c66a14d0 | 768 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
0aa69fd3 | 769 | return false; |
762380ad | 770 | |
c66a14d0 | 771 | if (bio->bi_vcnt > 0) { |
0aa69fd3 | 772 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
07173c3e ML |
773 | phys_addr_t vec_end_addr = page_to_phys(bv->bv_page) + |
774 | bv->bv_offset + bv->bv_len - 1; | |
775 | phys_addr_t page_addr = page_to_phys(page); | |
776 | ||
777 | if (vec_end_addr + 1 != page_addr + off) | |
778 | return false; | |
779 | if (same_page && (vec_end_addr & PAGE_MASK) != page_addr) | |
780 | return false; | |
781 | ||
782 | bv->bv_len += len; | |
783 | bio->bi_iter.bi_size += len; | |
784 | return true; | |
c66a14d0 | 785 | } |
0aa69fd3 CH |
786 | return false; |
787 | } | |
788 | EXPORT_SYMBOL_GPL(__bio_try_merge_page); | |
c66a14d0 | 789 | |
0aa69fd3 CH |
790 | /** |
791 | * __bio_add_page - add page to a bio in a new segment | |
792 | * @bio: destination bio | |
793 | * @page: page to add | |
794 | * @len: length of the data to add | |
795 | * @off: offset of the data in @page | |
796 | * | |
797 | * Add the data at @page + @off to @bio as a new bvec. The caller must ensure | |
798 | * that @bio has space for another bvec. | |
799 | */ | |
800 | void __bio_add_page(struct bio *bio, struct page *page, | |
801 | unsigned int len, unsigned int off) | |
802 | { | |
803 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
c66a14d0 | 804 | |
0aa69fd3 CH |
805 | WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)); |
806 | WARN_ON_ONCE(bio_full(bio)); | |
807 | ||
808 | bv->bv_page = page; | |
809 | bv->bv_offset = off; | |
810 | bv->bv_len = len; | |
c66a14d0 | 811 | |
c66a14d0 | 812 | bio->bi_iter.bi_size += len; |
0aa69fd3 CH |
813 | bio->bi_vcnt++; |
814 | } | |
815 | EXPORT_SYMBOL_GPL(__bio_add_page); | |
816 | ||
817 | /** | |
818 | * bio_add_page - attempt to add page to bio | |
819 | * @bio: destination bio | |
820 | * @page: page to add | |
821 | * @len: vec entry length | |
822 | * @offset: vec entry offset | |
823 | * | |
824 | * Attempt to add a page to the bio_vec maplist. This will only fail | |
825 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. | |
826 | */ | |
827 | int bio_add_page(struct bio *bio, struct page *page, | |
828 | unsigned int len, unsigned int offset) | |
829 | { | |
07173c3e | 830 | if (!__bio_try_merge_page(bio, page, len, offset, false)) { |
0aa69fd3 CH |
831 | if (bio_full(bio)) |
832 | return 0; | |
833 | __bio_add_page(bio, page, len, offset); | |
834 | } | |
c66a14d0 | 835 | return len; |
1da177e4 | 836 | } |
a112a71d | 837 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 838 | |
576ed913 CH |
839 | #define PAGE_PTRS_PER_BVEC (sizeof(struct bio_vec) / sizeof(struct page *)) |
840 | ||
2cefe4db | 841 | /** |
17d51b10 | 842 | * __bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio |
2cefe4db KO |
843 | * @bio: bio to add pages to |
844 | * @iter: iov iterator describing the region to be mapped | |
845 | * | |
17d51b10 | 846 | * Pins pages from *iter and appends them to @bio's bvec array. The |
2cefe4db | 847 | * pages will have to be released using put_page() when done. |
17d51b10 MW |
848 | * For multi-segment *iter, this function only adds pages from the |
849 | * the next non-empty segment of the iov iterator. | |
2cefe4db | 850 | */ |
17d51b10 | 851 | static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) |
2cefe4db | 852 | { |
576ed913 CH |
853 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; |
854 | unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt; | |
2cefe4db KO |
855 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; |
856 | struct page **pages = (struct page **)bv; | |
576ed913 CH |
857 | ssize_t size, left; |
858 | unsigned len, i; | |
b403ea24 | 859 | size_t offset; |
576ed913 CH |
860 | |
861 | /* | |
862 | * Move page array up in the allocated memory for the bio vecs as far as | |
863 | * possible so that we can start filling biovecs from the beginning | |
864 | * without overwriting the temporary page array. | |
865 | */ | |
866 | BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2); | |
867 | pages += entries_left * (PAGE_PTRS_PER_BVEC - 1); | |
2cefe4db KO |
868 | |
869 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); | |
870 | if (unlikely(size <= 0)) | |
871 | return size ? size : -EFAULT; | |
2cefe4db | 872 | |
576ed913 CH |
873 | for (left = size, i = 0; left > 0; left -= len, i++) { |
874 | struct page *page = pages[i]; | |
2cefe4db | 875 | |
576ed913 CH |
876 | len = min_t(size_t, PAGE_SIZE - offset, left); |
877 | if (WARN_ON_ONCE(bio_add_page(bio, page, len, offset) != len)) | |
878 | return -EINVAL; | |
879 | offset = 0; | |
2cefe4db KO |
880 | } |
881 | ||
2cefe4db KO |
882 | iov_iter_advance(iter, size); |
883 | return 0; | |
884 | } | |
17d51b10 MW |
885 | |
886 | /** | |
887 | * bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio | |
888 | * @bio: bio to add pages to | |
889 | * @iter: iov iterator describing the region to be mapped | |
890 | * | |
891 | * Pins pages from *iter and appends them to @bio's bvec array. The | |
892 | * pages will have to be released using put_page() when done. | |
893 | * The function tries, but does not guarantee, to pin as many pages as | |
894 | * fit into the bio, or are requested in *iter, whatever is smaller. | |
895 | * If MM encounters an error pinning the requested pages, it stops. | |
896 | * Error is returned only if 0 pages could be pinned. | |
897 | */ | |
898 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) | |
899 | { | |
900 | unsigned short orig_vcnt = bio->bi_vcnt; | |
901 | ||
902 | do { | |
903 | int ret = __bio_iov_iter_get_pages(bio, iter); | |
904 | ||
905 | if (unlikely(ret)) | |
906 | return bio->bi_vcnt > orig_vcnt ? 0 : ret; | |
907 | ||
908 | } while (iov_iter_count(iter) && !bio_full(bio)); | |
909 | ||
910 | return 0; | |
911 | } | |
2cefe4db | 912 | |
4246a0b6 | 913 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 | 914 | { |
65e53aab | 915 | complete(bio->bi_private); |
9e882242 KO |
916 | } |
917 | ||
918 | /** | |
919 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
920 | * @bio: The &struct bio which describes the I/O |
921 | * | |
922 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
923 | * bio_endio() on failure. | |
3d289d68 JK |
924 | * |
925 | * WARNING: Unlike to how submit_bio() is usually used, this function does not | |
926 | * result in bio reference to be consumed. The caller must drop the reference | |
927 | * on his own. | |
9e882242 | 928 | */ |
4e49ea4a | 929 | int submit_bio_wait(struct bio *bio) |
9e882242 | 930 | { |
e319e1fb | 931 | DECLARE_COMPLETION_ONSTACK_MAP(done, bio->bi_disk->lockdep_map); |
9e882242 | 932 | |
65e53aab | 933 | bio->bi_private = &done; |
9e882242 | 934 | bio->bi_end_io = submit_bio_wait_endio; |
1eff9d32 | 935 | bio->bi_opf |= REQ_SYNC; |
4e49ea4a | 936 | submit_bio(bio); |
65e53aab | 937 | wait_for_completion_io(&done); |
9e882242 | 938 | |
65e53aab | 939 | return blk_status_to_errno(bio->bi_status); |
9e882242 KO |
940 | } |
941 | EXPORT_SYMBOL(submit_bio_wait); | |
942 | ||
054bdf64 KO |
943 | /** |
944 | * bio_advance - increment/complete a bio by some number of bytes | |
945 | * @bio: bio to advance | |
946 | * @bytes: number of bytes to complete | |
947 | * | |
948 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
949 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
950 | * be updated on the last bvec as well. | |
951 | * | |
952 | * @bio will then represent the remaining, uncompleted portion of the io. | |
953 | */ | |
954 | void bio_advance(struct bio *bio, unsigned bytes) | |
955 | { | |
956 | if (bio_integrity(bio)) | |
957 | bio_integrity_advance(bio, bytes); | |
958 | ||
4550dd6c | 959 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
960 | } |
961 | EXPORT_SYMBOL(bio_advance); | |
962 | ||
45db54d5 KO |
963 | void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter, |
964 | struct bio *src, struct bvec_iter *src_iter) | |
16ac3d63 | 965 | { |
1cb9dda4 | 966 | struct bio_vec src_bv, dst_bv; |
16ac3d63 | 967 | void *src_p, *dst_p; |
1cb9dda4 | 968 | unsigned bytes; |
16ac3d63 | 969 | |
45db54d5 KO |
970 | while (src_iter->bi_size && dst_iter->bi_size) { |
971 | src_bv = bio_iter_iovec(src, *src_iter); | |
972 | dst_bv = bio_iter_iovec(dst, *dst_iter); | |
1cb9dda4 KO |
973 | |
974 | bytes = min(src_bv.bv_len, dst_bv.bv_len); | |
16ac3d63 | 975 | |
1cb9dda4 KO |
976 | src_p = kmap_atomic(src_bv.bv_page); |
977 | dst_p = kmap_atomic(dst_bv.bv_page); | |
16ac3d63 | 978 | |
1cb9dda4 KO |
979 | memcpy(dst_p + dst_bv.bv_offset, |
980 | src_p + src_bv.bv_offset, | |
16ac3d63 KO |
981 | bytes); |
982 | ||
983 | kunmap_atomic(dst_p); | |
984 | kunmap_atomic(src_p); | |
985 | ||
6e6e811d KO |
986 | flush_dcache_page(dst_bv.bv_page); |
987 | ||
45db54d5 KO |
988 | bio_advance_iter(src, src_iter, bytes); |
989 | bio_advance_iter(dst, dst_iter, bytes); | |
16ac3d63 KO |
990 | } |
991 | } | |
38a72dac KO |
992 | EXPORT_SYMBOL(bio_copy_data_iter); |
993 | ||
994 | /** | |
45db54d5 KO |
995 | * bio_copy_data - copy contents of data buffers from one bio to another |
996 | * @src: source bio | |
997 | * @dst: destination bio | |
38a72dac KO |
998 | * |
999 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
1000 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
1001 | */ | |
1002 | void bio_copy_data(struct bio *dst, struct bio *src) | |
1003 | { | |
45db54d5 KO |
1004 | struct bvec_iter src_iter = src->bi_iter; |
1005 | struct bvec_iter dst_iter = dst->bi_iter; | |
1006 | ||
1007 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); | |
38a72dac | 1008 | } |
16ac3d63 KO |
1009 | EXPORT_SYMBOL(bio_copy_data); |
1010 | ||
45db54d5 KO |
1011 | /** |
1012 | * bio_list_copy_data - copy contents of data buffers from one chain of bios to | |
1013 | * another | |
1014 | * @src: source bio list | |
1015 | * @dst: destination bio list | |
1016 | * | |
1017 | * Stops when it reaches the end of either the @src list or @dst list - that is, | |
1018 | * copies min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of | |
1019 | * bios). | |
1020 | */ | |
1021 | void bio_list_copy_data(struct bio *dst, struct bio *src) | |
1022 | { | |
1023 | struct bvec_iter src_iter = src->bi_iter; | |
1024 | struct bvec_iter dst_iter = dst->bi_iter; | |
1025 | ||
1026 | while (1) { | |
1027 | if (!src_iter.bi_size) { | |
1028 | src = src->bi_next; | |
1029 | if (!src) | |
1030 | break; | |
1031 | ||
1032 | src_iter = src->bi_iter; | |
1033 | } | |
1034 | ||
1035 | if (!dst_iter.bi_size) { | |
1036 | dst = dst->bi_next; | |
1037 | if (!dst) | |
1038 | break; | |
1039 | ||
1040 | dst_iter = dst->bi_iter; | |
1041 | } | |
1042 | ||
1043 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); | |
1044 | } | |
1045 | } | |
1046 | EXPORT_SYMBOL(bio_list_copy_data); | |
1047 | ||
1da177e4 | 1048 | struct bio_map_data { |
152e283f | 1049 | int is_our_pages; |
26e49cfc KO |
1050 | struct iov_iter iter; |
1051 | struct iovec iov[]; | |
1da177e4 LT |
1052 | }; |
1053 | ||
0e5b935d | 1054 | static struct bio_map_data *bio_alloc_map_data(struct iov_iter *data, |
76029ff3 | 1055 | gfp_t gfp_mask) |
1da177e4 | 1056 | { |
0e5b935d AV |
1057 | struct bio_map_data *bmd; |
1058 | if (data->nr_segs > UIO_MAXIOV) | |
f3f63c1c | 1059 | return NULL; |
1da177e4 | 1060 | |
0e5b935d AV |
1061 | bmd = kmalloc(sizeof(struct bio_map_data) + |
1062 | sizeof(struct iovec) * data->nr_segs, gfp_mask); | |
1063 | if (!bmd) | |
1064 | return NULL; | |
1065 | memcpy(bmd->iov, data->iov, sizeof(struct iovec) * data->nr_segs); | |
1066 | bmd->iter = *data; | |
1067 | bmd->iter.iov = bmd->iov; | |
1068 | return bmd; | |
1da177e4 LT |
1069 | } |
1070 | ||
9124d3fe DP |
1071 | /** |
1072 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1073 | * @bio: The &struct bio which describes the I/O as destination | |
1074 | * @iter: iov_iter as source | |
1075 | * | |
1076 | * Copy all pages from iov_iter to bio. | |
1077 | * Returns 0 on success, or error on failure. | |
1078 | */ | |
98a09d61 | 1079 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter *iter) |
c5dec1c3 | 1080 | { |
9124d3fe | 1081 | int i; |
c5dec1c3 | 1082 | struct bio_vec *bvec; |
6dc4f100 | 1083 | struct bvec_iter_all iter_all; |
c5dec1c3 | 1084 | |
6dc4f100 | 1085 | bio_for_each_segment_all(bvec, bio, i, iter_all) { |
9124d3fe | 1086 | ssize_t ret; |
c5dec1c3 | 1087 | |
9124d3fe DP |
1088 | ret = copy_page_from_iter(bvec->bv_page, |
1089 | bvec->bv_offset, | |
1090 | bvec->bv_len, | |
98a09d61 | 1091 | iter); |
9124d3fe | 1092 | |
98a09d61 | 1093 | if (!iov_iter_count(iter)) |
9124d3fe DP |
1094 | break; |
1095 | ||
1096 | if (ret < bvec->bv_len) | |
1097 | return -EFAULT; | |
c5dec1c3 FT |
1098 | } |
1099 | ||
9124d3fe DP |
1100 | return 0; |
1101 | } | |
1102 | ||
1103 | /** | |
1104 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1105 | * @bio: The &struct bio which describes the I/O as source | |
1106 | * @iter: iov_iter as destination | |
1107 | * | |
1108 | * Copy all pages from bio to iov_iter. | |
1109 | * Returns 0 on success, or error on failure. | |
1110 | */ | |
1111 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1112 | { | |
1113 | int i; | |
1114 | struct bio_vec *bvec; | |
6dc4f100 | 1115 | struct bvec_iter_all iter_all; |
9124d3fe | 1116 | |
6dc4f100 | 1117 | bio_for_each_segment_all(bvec, bio, i, iter_all) { |
9124d3fe DP |
1118 | ssize_t ret; |
1119 | ||
1120 | ret = copy_page_to_iter(bvec->bv_page, | |
1121 | bvec->bv_offset, | |
1122 | bvec->bv_len, | |
1123 | &iter); | |
1124 | ||
1125 | if (!iov_iter_count(&iter)) | |
1126 | break; | |
1127 | ||
1128 | if (ret < bvec->bv_len) | |
1129 | return -EFAULT; | |
1130 | } | |
1131 | ||
1132 | return 0; | |
c5dec1c3 FT |
1133 | } |
1134 | ||
491221f8 | 1135 | void bio_free_pages(struct bio *bio) |
1dfa0f68 CH |
1136 | { |
1137 | struct bio_vec *bvec; | |
1138 | int i; | |
6dc4f100 | 1139 | struct bvec_iter_all iter_all; |
1dfa0f68 | 1140 | |
6dc4f100 | 1141 | bio_for_each_segment_all(bvec, bio, i, iter_all) |
1dfa0f68 CH |
1142 | __free_page(bvec->bv_page); |
1143 | } | |
491221f8 | 1144 | EXPORT_SYMBOL(bio_free_pages); |
1dfa0f68 | 1145 | |
1da177e4 LT |
1146 | /** |
1147 | * bio_uncopy_user - finish previously mapped bio | |
1148 | * @bio: bio being terminated | |
1149 | * | |
ddad8dd0 | 1150 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1151 | * to user space in case of a read. |
1152 | */ | |
1153 | int bio_uncopy_user(struct bio *bio) | |
1154 | { | |
1155 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1156 | int ret = 0; |
1da177e4 | 1157 | |
35dc2483 RD |
1158 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1159 | /* | |
1160 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1161 | * don't copy into a random user address space, just free |
1162 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1163 | */ |
2d99b55d HR |
1164 | if (!current->mm) |
1165 | ret = -EINTR; | |
1166 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1167 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1168 | if (bmd->is_our_pages) |
1169 | bio_free_pages(bio); | |
35dc2483 | 1170 | } |
c8db4448 | 1171 | kfree(bmd); |
1da177e4 LT |
1172 | bio_put(bio); |
1173 | return ret; | |
1174 | } | |
1175 | ||
1176 | /** | |
c5dec1c3 | 1177 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1178 | * @q: destination block queue |
1179 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1180 | * @iter: iovec iterator | |
1181 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1182 | * |
1183 | * Prepares and returns a bio for indirect user io, bouncing data | |
1184 | * to/from kernel pages as necessary. Must be paired with | |
1185 | * call bio_uncopy_user() on io completion. | |
1186 | */ | |
152e283f FT |
1187 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1188 | struct rq_map_data *map_data, | |
e81cef5d | 1189 | struct iov_iter *iter, |
26e49cfc | 1190 | gfp_t gfp_mask) |
1da177e4 | 1191 | { |
1da177e4 | 1192 | struct bio_map_data *bmd; |
1da177e4 LT |
1193 | struct page *page; |
1194 | struct bio *bio; | |
d16d44eb AV |
1195 | int i = 0, ret; |
1196 | int nr_pages; | |
26e49cfc | 1197 | unsigned int len = iter->count; |
bd5cecea | 1198 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1199 | |
0e5b935d | 1200 | bmd = bio_alloc_map_data(iter, gfp_mask); |
1da177e4 LT |
1201 | if (!bmd) |
1202 | return ERR_PTR(-ENOMEM); | |
1203 | ||
26e49cfc KO |
1204 | /* |
1205 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1206 | * The caller provided iov might point to an on-stack or otherwise | |
1207 | * shortlived one. | |
1208 | */ | |
1209 | bmd->is_our_pages = map_data ? 0 : 1; | |
26e49cfc | 1210 | |
d16d44eb AV |
1211 | nr_pages = DIV_ROUND_UP(offset + len, PAGE_SIZE); |
1212 | if (nr_pages > BIO_MAX_PAGES) | |
1213 | nr_pages = BIO_MAX_PAGES; | |
26e49cfc | 1214 | |
1da177e4 | 1215 | ret = -ENOMEM; |
a9e9dc24 | 1216 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1217 | if (!bio) |
1218 | goto out_bmd; | |
1219 | ||
1da177e4 | 1220 | ret = 0; |
56c451f4 FT |
1221 | |
1222 | if (map_data) { | |
e623ddb4 | 1223 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1224 | i = map_data->offset / PAGE_SIZE; |
1225 | } | |
1da177e4 | 1226 | while (len) { |
e623ddb4 | 1227 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1228 | |
56c451f4 FT |
1229 | bytes -= offset; |
1230 | ||
1da177e4 LT |
1231 | if (bytes > len) |
1232 | bytes = len; | |
1233 | ||
152e283f | 1234 | if (map_data) { |
e623ddb4 | 1235 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1236 | ret = -ENOMEM; |
1237 | break; | |
1238 | } | |
e623ddb4 FT |
1239 | |
1240 | page = map_data->pages[i / nr_pages]; | |
1241 | page += (i % nr_pages); | |
1242 | ||
1243 | i++; | |
1244 | } else { | |
152e283f | 1245 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1246 | if (!page) { |
1247 | ret = -ENOMEM; | |
1248 | break; | |
1249 | } | |
1da177e4 LT |
1250 | } |
1251 | ||
56c451f4 | 1252 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 1253 | break; |
1da177e4 LT |
1254 | |
1255 | len -= bytes; | |
56c451f4 | 1256 | offset = 0; |
1da177e4 LT |
1257 | } |
1258 | ||
1259 | if (ret) | |
1260 | goto cleanup; | |
1261 | ||
2884d0be AV |
1262 | if (map_data) |
1263 | map_data->offset += bio->bi_iter.bi_size; | |
1264 | ||
1da177e4 LT |
1265 | /* |
1266 | * success | |
1267 | */ | |
00e23707 | 1268 | if ((iov_iter_rw(iter) == WRITE && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1269 | (map_data && map_data->from_user)) { |
98a09d61 | 1270 | ret = bio_copy_from_iter(bio, iter); |
c5dec1c3 FT |
1271 | if (ret) |
1272 | goto cleanup; | |
98a09d61 | 1273 | } else { |
f55adad6 KB |
1274 | if (bmd->is_our_pages) |
1275 | zero_fill_bio(bio); | |
98a09d61 | 1276 | iov_iter_advance(iter, bio->bi_iter.bi_size); |
1da177e4 LT |
1277 | } |
1278 | ||
26e49cfc | 1279 | bio->bi_private = bmd; |
2884d0be AV |
1280 | if (map_data && map_data->null_mapped) |
1281 | bio_set_flag(bio, BIO_NULL_MAPPED); | |
1da177e4 LT |
1282 | return bio; |
1283 | cleanup: | |
152e283f | 1284 | if (!map_data) |
1dfa0f68 | 1285 | bio_free_pages(bio); |
1da177e4 LT |
1286 | bio_put(bio); |
1287 | out_bmd: | |
c8db4448 | 1288 | kfree(bmd); |
1da177e4 LT |
1289 | return ERR_PTR(ret); |
1290 | } | |
1291 | ||
37f19e57 CH |
1292 | /** |
1293 | * bio_map_user_iov - map user iovec into bio | |
1294 | * @q: the struct request_queue for the bio | |
1295 | * @iter: iovec iterator | |
1296 | * @gfp_mask: memory allocation flags | |
1297 | * | |
1298 | * Map the user space address into a bio suitable for io to a block | |
1299 | * device. Returns an error pointer in case of error. | |
1300 | */ | |
1301 | struct bio *bio_map_user_iov(struct request_queue *q, | |
e81cef5d | 1302 | struct iov_iter *iter, |
37f19e57 | 1303 | gfp_t gfp_mask) |
1da177e4 | 1304 | { |
26e49cfc | 1305 | int j; |
1da177e4 | 1306 | struct bio *bio; |
076098e5 | 1307 | int ret; |
2b04e8f6 | 1308 | struct bio_vec *bvec; |
6dc4f100 | 1309 | struct bvec_iter_all iter_all; |
1da177e4 | 1310 | |
b282cc76 | 1311 | if (!iov_iter_count(iter)) |
1da177e4 LT |
1312 | return ERR_PTR(-EINVAL); |
1313 | ||
b282cc76 | 1314 | bio = bio_kmalloc(gfp_mask, iov_iter_npages(iter, BIO_MAX_PAGES)); |
1da177e4 LT |
1315 | if (!bio) |
1316 | return ERR_PTR(-ENOMEM); | |
1317 | ||
0a0f1513 | 1318 | while (iov_iter_count(iter)) { |
629e42bc | 1319 | struct page **pages; |
076098e5 AV |
1320 | ssize_t bytes; |
1321 | size_t offs, added = 0; | |
1322 | int npages; | |
1da177e4 | 1323 | |
0a0f1513 | 1324 | bytes = iov_iter_get_pages_alloc(iter, &pages, LONG_MAX, &offs); |
076098e5 AV |
1325 | if (unlikely(bytes <= 0)) { |
1326 | ret = bytes ? bytes : -EFAULT; | |
f1970baf | 1327 | goto out_unmap; |
99172157 | 1328 | } |
f1970baf | 1329 | |
076098e5 | 1330 | npages = DIV_ROUND_UP(offs + bytes, PAGE_SIZE); |
f1970baf | 1331 | |
98f0bc99 AV |
1332 | if (unlikely(offs & queue_dma_alignment(q))) { |
1333 | ret = -EINVAL; | |
1334 | j = 0; | |
1335 | } else { | |
1336 | for (j = 0; j < npages; j++) { | |
1337 | struct page *page = pages[j]; | |
1338 | unsigned int n = PAGE_SIZE - offs; | |
1339 | unsigned short prev_bi_vcnt = bio->bi_vcnt; | |
f1970baf | 1340 | |
98f0bc99 AV |
1341 | if (n > bytes) |
1342 | n = bytes; | |
95d78c28 | 1343 | |
98f0bc99 AV |
1344 | if (!bio_add_pc_page(q, bio, page, n, offs)) |
1345 | break; | |
1da177e4 | 1346 | |
98f0bc99 AV |
1347 | /* |
1348 | * check if vector was merged with previous | |
1349 | * drop page reference if needed | |
1350 | */ | |
1351 | if (bio->bi_vcnt == prev_bi_vcnt) | |
1352 | put_page(page); | |
1353 | ||
1354 | added += n; | |
1355 | bytes -= n; | |
1356 | offs = 0; | |
1357 | } | |
0a0f1513 | 1358 | iov_iter_advance(iter, added); |
f1970baf | 1359 | } |
1da177e4 | 1360 | /* |
f1970baf | 1361 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1362 | */ |
629e42bc | 1363 | while (j < npages) |
09cbfeaf | 1364 | put_page(pages[j++]); |
629e42bc | 1365 | kvfree(pages); |
e2e115d1 AV |
1366 | /* couldn't stuff something into bio? */ |
1367 | if (bytes) | |
1368 | break; | |
1da177e4 LT |
1369 | } |
1370 | ||
b7c44ed9 | 1371 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1372 | |
1373 | /* | |
5fad1b64 | 1374 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
37f19e57 CH |
1375 | * it would normally disappear when its bi_end_io is run. |
1376 | * however, we need it for the unmap, so grab an extra | |
1377 | * reference to it | |
1378 | */ | |
1379 | bio_get(bio); | |
1da177e4 | 1380 | return bio; |
f1970baf JB |
1381 | |
1382 | out_unmap: | |
6dc4f100 | 1383 | bio_for_each_segment_all(bvec, bio, j, iter_all) { |
2b04e8f6 | 1384 | put_page(bvec->bv_page); |
f1970baf | 1385 | } |
1da177e4 LT |
1386 | bio_put(bio); |
1387 | return ERR_PTR(ret); | |
1388 | } | |
1389 | ||
1da177e4 LT |
1390 | static void __bio_unmap_user(struct bio *bio) |
1391 | { | |
1392 | struct bio_vec *bvec; | |
1393 | int i; | |
6dc4f100 | 1394 | struct bvec_iter_all iter_all; |
1da177e4 LT |
1395 | |
1396 | /* | |
1397 | * make sure we dirty pages we wrote to | |
1398 | */ | |
6dc4f100 | 1399 | bio_for_each_segment_all(bvec, bio, i, iter_all) { |
1da177e4 LT |
1400 | if (bio_data_dir(bio) == READ) |
1401 | set_page_dirty_lock(bvec->bv_page); | |
1402 | ||
09cbfeaf | 1403 | put_page(bvec->bv_page); |
1da177e4 LT |
1404 | } |
1405 | ||
1406 | bio_put(bio); | |
1407 | } | |
1408 | ||
1409 | /** | |
1410 | * bio_unmap_user - unmap a bio | |
1411 | * @bio: the bio being unmapped | |
1412 | * | |
5fad1b64 BVA |
1413 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
1414 | * process context. | |
1da177e4 LT |
1415 | * |
1416 | * bio_unmap_user() may sleep. | |
1417 | */ | |
1418 | void bio_unmap_user(struct bio *bio) | |
1419 | { | |
1420 | __bio_unmap_user(bio); | |
1421 | bio_put(bio); | |
1422 | } | |
1423 | ||
4246a0b6 | 1424 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1425 | { |
b823825e | 1426 | bio_put(bio); |
b823825e JA |
1427 | } |
1428 | ||
75c72b83 CH |
1429 | /** |
1430 | * bio_map_kern - map kernel address into bio | |
1431 | * @q: the struct request_queue for the bio | |
1432 | * @data: pointer to buffer to map | |
1433 | * @len: length in bytes | |
1434 | * @gfp_mask: allocation flags for bio allocation | |
1435 | * | |
1436 | * Map the kernel address into a bio suitable for io to a block | |
1437 | * device. Returns an error pointer in case of error. | |
1438 | */ | |
1439 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1440 | gfp_t gfp_mask) | |
df46b9a4 MC |
1441 | { |
1442 | unsigned long kaddr = (unsigned long)data; | |
1443 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1444 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1445 | const int nr_pages = end - start; | |
1446 | int offset, i; | |
1447 | struct bio *bio; | |
1448 | ||
a9e9dc24 | 1449 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1450 | if (!bio) |
1451 | return ERR_PTR(-ENOMEM); | |
1452 | ||
1453 | offset = offset_in_page(kaddr); | |
1454 | for (i = 0; i < nr_pages; i++) { | |
1455 | unsigned int bytes = PAGE_SIZE - offset; | |
1456 | ||
1457 | if (len <= 0) | |
1458 | break; | |
1459 | ||
1460 | if (bytes > len) | |
1461 | bytes = len; | |
1462 | ||
defd94b7 | 1463 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1464 | offset) < bytes) { |
1465 | /* we don't support partial mappings */ | |
1466 | bio_put(bio); | |
1467 | return ERR_PTR(-EINVAL); | |
1468 | } | |
df46b9a4 MC |
1469 | |
1470 | data += bytes; | |
1471 | len -= bytes; | |
1472 | offset = 0; | |
1473 | } | |
1474 | ||
b823825e | 1475 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1476 | return bio; |
1477 | } | |
a112a71d | 1478 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1479 | |
4246a0b6 | 1480 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1481 | { |
1dfa0f68 CH |
1482 | bio_free_pages(bio); |
1483 | bio_put(bio); | |
1484 | } | |
1485 | ||
4246a0b6 | 1486 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1487 | { |
42d2683a | 1488 | char *p = bio->bi_private; |
1dfa0f68 | 1489 | struct bio_vec *bvec; |
68154e90 | 1490 | int i; |
6dc4f100 | 1491 | struct bvec_iter_all iter_all; |
68154e90 | 1492 | |
6dc4f100 | 1493 | bio_for_each_segment_all(bvec, bio, i, iter_all) { |
1dfa0f68 | 1494 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1495 | p += bvec->bv_len; |
68154e90 FT |
1496 | } |
1497 | ||
4246a0b6 | 1498 | bio_copy_kern_endio(bio); |
68154e90 FT |
1499 | } |
1500 | ||
1501 | /** | |
1502 | * bio_copy_kern - copy kernel address into bio | |
1503 | * @q: the struct request_queue for the bio | |
1504 | * @data: pointer to buffer to copy | |
1505 | * @len: length in bytes | |
1506 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1507 | * @reading: data direction is READ |
68154e90 FT |
1508 | * |
1509 | * copy the kernel address into a bio suitable for io to a block | |
1510 | * device. Returns an error pointer in case of error. | |
1511 | */ | |
1512 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1513 | gfp_t gfp_mask, int reading) | |
1514 | { | |
42d2683a CH |
1515 | unsigned long kaddr = (unsigned long)data; |
1516 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1517 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1518 | struct bio *bio; |
1519 | void *p = data; | |
1dfa0f68 | 1520 | int nr_pages = 0; |
68154e90 | 1521 | |
42d2683a CH |
1522 | /* |
1523 | * Overflow, abort | |
1524 | */ | |
1525 | if (end < start) | |
1526 | return ERR_PTR(-EINVAL); | |
68154e90 | 1527 | |
42d2683a CH |
1528 | nr_pages = end - start; |
1529 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1530 | if (!bio) | |
1531 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1532 | |
42d2683a CH |
1533 | while (len) { |
1534 | struct page *page; | |
1535 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1536 | |
42d2683a CH |
1537 | if (bytes > len) |
1538 | bytes = len; | |
1539 | ||
1540 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1541 | if (!page) | |
1542 | goto cleanup; | |
1543 | ||
1544 | if (!reading) | |
1545 | memcpy(page_address(page), p, bytes); | |
1546 | ||
1547 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1548 | break; | |
1549 | ||
1550 | len -= bytes; | |
1551 | p += bytes; | |
68154e90 FT |
1552 | } |
1553 | ||
1dfa0f68 CH |
1554 | if (reading) { |
1555 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1556 | bio->bi_private = data; | |
1557 | } else { | |
1558 | bio->bi_end_io = bio_copy_kern_endio; | |
1dfa0f68 | 1559 | } |
76029ff3 | 1560 | |
68154e90 | 1561 | return bio; |
42d2683a CH |
1562 | |
1563 | cleanup: | |
1dfa0f68 | 1564 | bio_free_pages(bio); |
42d2683a CH |
1565 | bio_put(bio); |
1566 | return ERR_PTR(-ENOMEM); | |
68154e90 FT |
1567 | } |
1568 | ||
1da177e4 LT |
1569 | /* |
1570 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1571 | * for performing direct-IO in BIOs. | |
1572 | * | |
1573 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1574 | * because the required locks are not interrupt-safe. So what we can do is to | |
1575 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1576 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1577 | * in process context. | |
1578 | * | |
1579 | * We special-case compound pages here: normally this means reads into hugetlb | |
1580 | * pages. The logic in here doesn't really work right for compound pages | |
1581 | * because the VM does not uniformly chase down the head page in all cases. | |
1582 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1583 | * handle them at all. So we skip compound pages here at an early stage. | |
1584 | * | |
1585 | * Note that this code is very hard to test under normal circumstances because | |
1586 | * direct-io pins the pages with get_user_pages(). This makes | |
1587 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1588 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1589 | * pagecache. |
1590 | * | |
1591 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1592 | * deferred bio dirtying paths. | |
1593 | */ | |
1594 | ||
1595 | /* | |
1596 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1597 | */ | |
1598 | void bio_set_pages_dirty(struct bio *bio) | |
1599 | { | |
cb34e057 | 1600 | struct bio_vec *bvec; |
1da177e4 | 1601 | int i; |
6dc4f100 | 1602 | struct bvec_iter_all iter_all; |
1da177e4 | 1603 | |
6dc4f100 | 1604 | bio_for_each_segment_all(bvec, bio, i, iter_all) { |
3bb50983 CH |
1605 | if (!PageCompound(bvec->bv_page)) |
1606 | set_page_dirty_lock(bvec->bv_page); | |
1da177e4 LT |
1607 | } |
1608 | } | |
1609 | ||
86b6c7a7 | 1610 | static void bio_release_pages(struct bio *bio) |
1da177e4 | 1611 | { |
cb34e057 | 1612 | struct bio_vec *bvec; |
1da177e4 | 1613 | int i; |
6dc4f100 | 1614 | struct bvec_iter_all iter_all; |
1da177e4 | 1615 | |
6dc4f100 | 1616 | bio_for_each_segment_all(bvec, bio, i, iter_all) |
24d5493f | 1617 | put_page(bvec->bv_page); |
1da177e4 LT |
1618 | } |
1619 | ||
1620 | /* | |
1621 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1622 | * If they are, then fine. If, however, some pages are clean then they must | |
1623 | * have been written out during the direct-IO read. So we take another ref on | |
24d5493f | 1624 | * the BIO and re-dirty the pages in process context. |
1da177e4 LT |
1625 | * |
1626 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1627 | * here on. It will run one put_page() against each page and will run one |
1628 | * bio_put() against the BIO. | |
1da177e4 LT |
1629 | */ |
1630 | ||
65f27f38 | 1631 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1632 | |
65f27f38 | 1633 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1634 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1635 | static struct bio *bio_dirty_list; | |
1636 | ||
1637 | /* | |
1638 | * This runs in process context | |
1639 | */ | |
65f27f38 | 1640 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 | 1641 | { |
24d5493f | 1642 | struct bio *bio, *next; |
1da177e4 | 1643 | |
24d5493f CH |
1644 | spin_lock_irq(&bio_dirty_lock); |
1645 | next = bio_dirty_list; | |
1da177e4 | 1646 | bio_dirty_list = NULL; |
24d5493f | 1647 | spin_unlock_irq(&bio_dirty_lock); |
1da177e4 | 1648 | |
24d5493f CH |
1649 | while ((bio = next) != NULL) { |
1650 | next = bio->bi_private; | |
1da177e4 LT |
1651 | |
1652 | bio_set_pages_dirty(bio); | |
1653 | bio_release_pages(bio); | |
1654 | bio_put(bio); | |
1da177e4 LT |
1655 | } |
1656 | } | |
1657 | ||
1658 | void bio_check_pages_dirty(struct bio *bio) | |
1659 | { | |
cb34e057 | 1660 | struct bio_vec *bvec; |
24d5493f | 1661 | unsigned long flags; |
1da177e4 | 1662 | int i; |
6dc4f100 | 1663 | struct bvec_iter_all iter_all; |
1da177e4 | 1664 | |
6dc4f100 | 1665 | bio_for_each_segment_all(bvec, bio, i, iter_all) { |
24d5493f CH |
1666 | if (!PageDirty(bvec->bv_page) && !PageCompound(bvec->bv_page)) |
1667 | goto defer; | |
1da177e4 LT |
1668 | } |
1669 | ||
24d5493f CH |
1670 | bio_release_pages(bio); |
1671 | bio_put(bio); | |
1672 | return; | |
1673 | defer: | |
1674 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1675 | bio->bi_private = bio_dirty_list; | |
1676 | bio_dirty_list = bio; | |
1677 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1678 | schedule_work(&bio_dirty_work); | |
1da177e4 LT |
1679 | } |
1680 | ||
5b18b5a7 MP |
1681 | void update_io_ticks(struct hd_struct *part, unsigned long now) |
1682 | { | |
1683 | unsigned long stamp; | |
1684 | again: | |
1685 | stamp = READ_ONCE(part->stamp); | |
1686 | if (unlikely(stamp != now)) { | |
1687 | if (likely(cmpxchg(&part->stamp, stamp, now) == stamp)) { | |
1688 | __part_stat_add(part, io_ticks, 1); | |
1689 | } | |
1690 | } | |
1691 | if (part->partno) { | |
1692 | part = &part_to_disk(part)->part0; | |
1693 | goto again; | |
1694 | } | |
1695 | } | |
1da177e4 | 1696 | |
ddcf35d3 | 1697 | void generic_start_io_acct(struct request_queue *q, int op, |
d62e26b3 | 1698 | unsigned long sectors, struct hd_struct *part) |
394ffa50 | 1699 | { |
ddcf35d3 | 1700 | const int sgrp = op_stat_group(op); |
394ffa50 | 1701 | |
112f158f MS |
1702 | part_stat_lock(); |
1703 | ||
5b18b5a7 | 1704 | update_io_ticks(part, jiffies); |
112f158f MS |
1705 | part_stat_inc(part, ios[sgrp]); |
1706 | part_stat_add(part, sectors[sgrp], sectors); | |
ddcf35d3 | 1707 | part_inc_in_flight(q, part, op_is_write(op)); |
394ffa50 GZ |
1708 | |
1709 | part_stat_unlock(); | |
1710 | } | |
1711 | EXPORT_SYMBOL(generic_start_io_acct); | |
1712 | ||
ddcf35d3 | 1713 | void generic_end_io_acct(struct request_queue *q, int req_op, |
d62e26b3 | 1714 | struct hd_struct *part, unsigned long start_time) |
394ffa50 | 1715 | { |
5b18b5a7 MP |
1716 | unsigned long now = jiffies; |
1717 | unsigned long duration = now - start_time; | |
ddcf35d3 | 1718 | const int sgrp = op_stat_group(req_op); |
394ffa50 | 1719 | |
112f158f MS |
1720 | part_stat_lock(); |
1721 | ||
5b18b5a7 | 1722 | update_io_ticks(part, now); |
112f158f | 1723 | part_stat_add(part, nsecs[sgrp], jiffies_to_nsecs(duration)); |
5b18b5a7 | 1724 | part_stat_add(part, time_in_queue, duration); |
ddcf35d3 | 1725 | part_dec_in_flight(q, part, op_is_write(req_op)); |
394ffa50 GZ |
1726 | |
1727 | part_stat_unlock(); | |
1728 | } | |
1729 | EXPORT_SYMBOL(generic_end_io_acct); | |
1730 | ||
2d4dc890 IL |
1731 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1732 | void bio_flush_dcache_pages(struct bio *bi) | |
1733 | { | |
7988613b KO |
1734 | struct bio_vec bvec; |
1735 | struct bvec_iter iter; | |
2d4dc890 | 1736 | |
7988613b KO |
1737 | bio_for_each_segment(bvec, bi, iter) |
1738 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1739 | } |
1740 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1741 | #endif | |
1742 | ||
c4cf5261 JA |
1743 | static inline bool bio_remaining_done(struct bio *bio) |
1744 | { | |
1745 | /* | |
1746 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1747 | * we always end io on the first invocation. | |
1748 | */ | |
1749 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1750 | return true; | |
1751 | ||
1752 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1753 | ||
326e1dbb | 1754 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1755 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1756 | return true; |
326e1dbb | 1757 | } |
c4cf5261 JA |
1758 | |
1759 | return false; | |
1760 | } | |
1761 | ||
1da177e4 LT |
1762 | /** |
1763 | * bio_endio - end I/O on a bio | |
1764 | * @bio: bio | |
1da177e4 LT |
1765 | * |
1766 | * Description: | |
4246a0b6 CH |
1767 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1768 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1769 | * bio unless they own it and thus know that it has an end_io function. | |
fbbaf700 N |
1770 | * |
1771 | * bio_endio() can be called several times on a bio that has been chained | |
1772 | * using bio_chain(). The ->bi_end_io() function will only be called the | |
1773 | * last time. At this point the BLK_TA_COMPLETE tracing event will be | |
1774 | * generated if BIO_TRACE_COMPLETION is set. | |
1da177e4 | 1775 | **/ |
4246a0b6 | 1776 | void bio_endio(struct bio *bio) |
1da177e4 | 1777 | { |
ba8c6967 | 1778 | again: |
2b885517 | 1779 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1780 | return; |
7c20f116 CH |
1781 | if (!bio_integrity_endio(bio)) |
1782 | return; | |
1da177e4 | 1783 | |
67b42d0b JB |
1784 | if (bio->bi_disk) |
1785 | rq_qos_done_bio(bio->bi_disk->queue, bio); | |
1786 | ||
ba8c6967 CH |
1787 | /* |
1788 | * Need to have a real endio function for chained bios, otherwise | |
1789 | * various corner cases will break (like stacking block devices that | |
1790 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1791 | * recursion and blowing the stack. Tail call optimization would | |
1792 | * handle this, but compiling with frame pointers also disables | |
1793 | * gcc's sibling call optimization. | |
1794 | */ | |
1795 | if (bio->bi_end_io == bio_chain_endio) { | |
1796 | bio = __bio_chain_endio(bio); | |
1797 | goto again; | |
196d38bc | 1798 | } |
ba8c6967 | 1799 | |
74d46992 CH |
1800 | if (bio->bi_disk && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
1801 | trace_block_bio_complete(bio->bi_disk->queue, bio, | |
a462b950 | 1802 | blk_status_to_errno(bio->bi_status)); |
fbbaf700 N |
1803 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
1804 | } | |
1805 | ||
9e234eea | 1806 | blk_throtl_bio_endio(bio); |
b222dd2f SL |
1807 | /* release cgroup info */ |
1808 | bio_uninit(bio); | |
ba8c6967 CH |
1809 | if (bio->bi_end_io) |
1810 | bio->bi_end_io(bio); | |
1da177e4 | 1811 | } |
a112a71d | 1812 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1813 | |
20d0189b KO |
1814 | /** |
1815 | * bio_split - split a bio | |
1816 | * @bio: bio to split | |
1817 | * @sectors: number of sectors to split from the front of @bio | |
1818 | * @gfp: gfp mask | |
1819 | * @bs: bio set to allocate from | |
1820 | * | |
1821 | * Allocates and returns a new bio which represents @sectors from the start of | |
1822 | * @bio, and updates @bio to represent the remaining sectors. | |
1823 | * | |
f3f5da62 MP |
1824 | * Unless this is a discard request the newly allocated bio will point |
1825 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1826 | * @bio is not freed before the split. | |
20d0189b KO |
1827 | */ |
1828 | struct bio *bio_split(struct bio *bio, int sectors, | |
1829 | gfp_t gfp, struct bio_set *bs) | |
1830 | { | |
f341a4d3 | 1831 | struct bio *split; |
20d0189b KO |
1832 | |
1833 | BUG_ON(sectors <= 0); | |
1834 | BUG_ON(sectors >= bio_sectors(bio)); | |
1835 | ||
f9d03f96 | 1836 | split = bio_clone_fast(bio, gfp, bs); |
20d0189b KO |
1837 | if (!split) |
1838 | return NULL; | |
1839 | ||
1840 | split->bi_iter.bi_size = sectors << 9; | |
1841 | ||
1842 | if (bio_integrity(split)) | |
fbd08e76 | 1843 | bio_integrity_trim(split); |
20d0189b KO |
1844 | |
1845 | bio_advance(bio, split->bi_iter.bi_size); | |
1846 | ||
fbbaf700 | 1847 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
20d59023 | 1848 | bio_set_flag(split, BIO_TRACE_COMPLETION); |
fbbaf700 | 1849 | |
20d0189b KO |
1850 | return split; |
1851 | } | |
1852 | EXPORT_SYMBOL(bio_split); | |
1853 | ||
6678d83f KO |
1854 | /** |
1855 | * bio_trim - trim a bio | |
1856 | * @bio: bio to trim | |
1857 | * @offset: number of sectors to trim from the front of @bio | |
1858 | * @size: size we want to trim @bio to, in sectors | |
1859 | */ | |
1860 | void bio_trim(struct bio *bio, int offset, int size) | |
1861 | { | |
1862 | /* 'bio' is a cloned bio which we need to trim to match | |
1863 | * the given offset and size. | |
6678d83f | 1864 | */ |
6678d83f KO |
1865 | |
1866 | size <<= 9; | |
4f024f37 | 1867 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1868 | return; |
1869 | ||
b7c44ed9 | 1870 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1871 | |
1872 | bio_advance(bio, offset << 9); | |
1873 | ||
4f024f37 | 1874 | bio->bi_iter.bi_size = size; |
376a78ab DM |
1875 | |
1876 | if (bio_integrity(bio)) | |
fbd08e76 | 1877 | bio_integrity_trim(bio); |
376a78ab | 1878 | |
6678d83f KO |
1879 | } |
1880 | EXPORT_SYMBOL_GPL(bio_trim); | |
1881 | ||
1da177e4 LT |
1882 | /* |
1883 | * create memory pools for biovec's in a bio_set. | |
1884 | * use the global biovec slabs created for general use. | |
1885 | */ | |
8aa6ba2f | 1886 | int biovec_init_pool(mempool_t *pool, int pool_entries) |
1da177e4 | 1887 | { |
ed996a52 | 1888 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
1da177e4 | 1889 | |
8aa6ba2f | 1890 | return mempool_init_slab_pool(pool, pool_entries, bp->slab); |
1da177e4 LT |
1891 | } |
1892 | ||
917a38c7 KO |
1893 | /* |
1894 | * bioset_exit - exit a bioset initialized with bioset_init() | |
1895 | * | |
1896 | * May be called on a zeroed but uninitialized bioset (i.e. allocated with | |
1897 | * kzalloc()). | |
1898 | */ | |
1899 | void bioset_exit(struct bio_set *bs) | |
1da177e4 | 1900 | { |
df2cb6da KO |
1901 | if (bs->rescue_workqueue) |
1902 | destroy_workqueue(bs->rescue_workqueue); | |
917a38c7 | 1903 | bs->rescue_workqueue = NULL; |
df2cb6da | 1904 | |
8aa6ba2f KO |
1905 | mempool_exit(&bs->bio_pool); |
1906 | mempool_exit(&bs->bvec_pool); | |
9f060e22 | 1907 | |
7878cba9 | 1908 | bioset_integrity_free(bs); |
917a38c7 KO |
1909 | if (bs->bio_slab) |
1910 | bio_put_slab(bs); | |
1911 | bs->bio_slab = NULL; | |
1912 | } | |
1913 | EXPORT_SYMBOL(bioset_exit); | |
1da177e4 | 1914 | |
917a38c7 KO |
1915 | /** |
1916 | * bioset_init - Initialize a bio_set | |
dad08527 | 1917 | * @bs: pool to initialize |
917a38c7 KO |
1918 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
1919 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1920 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS | |
1921 | * and %BIOSET_NEED_RESCUER | |
1922 | * | |
dad08527 KO |
1923 | * Description: |
1924 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1925 | * to ask for a number of bytes to be allocated in front of the bio. | |
1926 | * Front pad allocation is useful for embedding the bio inside | |
1927 | * another structure, to avoid allocating extra data to go with the bio. | |
1928 | * Note that the bio must be embedded at the END of that structure always, | |
1929 | * or things will break badly. | |
1930 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated | |
1931 | * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast(). | |
1932 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to | |
1933 | * dispatch queued requests when the mempool runs out of space. | |
1934 | * | |
917a38c7 KO |
1935 | */ |
1936 | int bioset_init(struct bio_set *bs, | |
1937 | unsigned int pool_size, | |
1938 | unsigned int front_pad, | |
1939 | int flags) | |
1940 | { | |
1941 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); | |
1942 | ||
1943 | bs->front_pad = front_pad; | |
1944 | ||
1945 | spin_lock_init(&bs->rescue_lock); | |
1946 | bio_list_init(&bs->rescue_list); | |
1947 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
1948 | ||
1949 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); | |
1950 | if (!bs->bio_slab) | |
1951 | return -ENOMEM; | |
1952 | ||
1953 | if (mempool_init_slab_pool(&bs->bio_pool, pool_size, bs->bio_slab)) | |
1954 | goto bad; | |
1955 | ||
1956 | if ((flags & BIOSET_NEED_BVECS) && | |
1957 | biovec_init_pool(&bs->bvec_pool, pool_size)) | |
1958 | goto bad; | |
1959 | ||
1960 | if (!(flags & BIOSET_NEED_RESCUER)) | |
1961 | return 0; | |
1962 | ||
1963 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); | |
1964 | if (!bs->rescue_workqueue) | |
1965 | goto bad; | |
1966 | ||
1967 | return 0; | |
1968 | bad: | |
1969 | bioset_exit(bs); | |
1970 | return -ENOMEM; | |
1971 | } | |
1972 | EXPORT_SYMBOL(bioset_init); | |
1973 | ||
28e89fd9 JA |
1974 | /* |
1975 | * Initialize and setup a new bio_set, based on the settings from | |
1976 | * another bio_set. | |
1977 | */ | |
1978 | int bioset_init_from_src(struct bio_set *bs, struct bio_set *src) | |
1979 | { | |
1980 | int flags; | |
1981 | ||
1982 | flags = 0; | |
1983 | if (src->bvec_pool.min_nr) | |
1984 | flags |= BIOSET_NEED_BVECS; | |
1985 | if (src->rescue_workqueue) | |
1986 | flags |= BIOSET_NEED_RESCUER; | |
1987 | ||
1988 | return bioset_init(bs, src->bio_pool.min_nr, src->front_pad, flags); | |
1989 | } | |
1990 | EXPORT_SYMBOL(bioset_init_from_src); | |
1991 | ||
852c788f | 1992 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 | 1993 | |
74b7c02a | 1994 | /** |
2268c0fe | 1995 | * bio_disassociate_blkg - puts back the blkg reference if associated |
74b7c02a | 1996 | * @bio: target bio |
74b7c02a | 1997 | * |
2268c0fe | 1998 | * Helper to disassociate the blkg from @bio if a blkg is associated. |
74b7c02a | 1999 | */ |
2268c0fe | 2000 | void bio_disassociate_blkg(struct bio *bio) |
74b7c02a | 2001 | { |
2268c0fe DZ |
2002 | if (bio->bi_blkg) { |
2003 | blkg_put(bio->bi_blkg); | |
2004 | bio->bi_blkg = NULL; | |
2005 | } | |
74b7c02a | 2006 | } |
892ad71f | 2007 | EXPORT_SYMBOL_GPL(bio_disassociate_blkg); |
74b7c02a | 2008 | |
a7b39b4e | 2009 | /** |
2268c0fe | 2010 | * __bio_associate_blkg - associate a bio with the a blkg |
a7b39b4e | 2011 | * @bio: target bio |
b5f2954d | 2012 | * @blkg: the blkg to associate |
b5f2954d | 2013 | * |
beea9da0 DZ |
2014 | * This tries to associate @bio with the specified @blkg. Association failure |
2015 | * is handled by walking up the blkg tree. Therefore, the blkg associated can | |
2016 | * be anything between @blkg and the root_blkg. This situation only happens | |
2017 | * when a cgroup is dying and then the remaining bios will spill to the closest | |
2018 | * alive blkg. | |
a7b39b4e | 2019 | * |
beea9da0 DZ |
2020 | * A reference will be taken on the @blkg and will be released when @bio is |
2021 | * freed. | |
a7b39b4e | 2022 | */ |
2268c0fe | 2023 | static void __bio_associate_blkg(struct bio *bio, struct blkcg_gq *blkg) |
a7b39b4e | 2024 | { |
2268c0fe DZ |
2025 | bio_disassociate_blkg(bio); |
2026 | ||
7754f669 | 2027 | bio->bi_blkg = blkg_tryget_closest(blkg); |
a7b39b4e DZF |
2028 | } |
2029 | ||
d459d853 | 2030 | /** |
fd42df30 | 2031 | * bio_associate_blkg_from_css - associate a bio with a specified css |
d459d853 | 2032 | * @bio: target bio |
fd42df30 | 2033 | * @css: target css |
d459d853 | 2034 | * |
fd42df30 | 2035 | * Associate @bio with the blkg found by combining the css's blkg and the |
fc5a828b DZ |
2036 | * request_queue of the @bio. This falls back to the queue's root_blkg if |
2037 | * the association fails with the css. | |
d459d853 | 2038 | */ |
fd42df30 DZ |
2039 | void bio_associate_blkg_from_css(struct bio *bio, |
2040 | struct cgroup_subsys_state *css) | |
d459d853 | 2041 | { |
fc5a828b DZ |
2042 | struct request_queue *q = bio->bi_disk->queue; |
2043 | struct blkcg_gq *blkg; | |
2044 | ||
2045 | rcu_read_lock(); | |
2046 | ||
2047 | if (!css || !css->parent) | |
2048 | blkg = q->root_blkg; | |
2049 | else | |
2050 | blkg = blkg_lookup_create(css_to_blkcg(css), q); | |
2051 | ||
2052 | __bio_associate_blkg(bio, blkg); | |
2053 | ||
2054 | rcu_read_unlock(); | |
d459d853 | 2055 | } |
fd42df30 | 2056 | EXPORT_SYMBOL_GPL(bio_associate_blkg_from_css); |
d459d853 | 2057 | |
6a7f6d86 | 2058 | #ifdef CONFIG_MEMCG |
852c788f | 2059 | /** |
6a7f6d86 | 2060 | * bio_associate_blkg_from_page - associate a bio with the page's blkg |
852c788f | 2061 | * @bio: target bio |
6a7f6d86 DZ |
2062 | * @page: the page to lookup the blkcg from |
2063 | * | |
2064 | * Associate @bio with the blkg from @page's owning memcg and the respective | |
fc5a828b DZ |
2065 | * request_queue. If cgroup_e_css returns %NULL, fall back to the queue's |
2066 | * root_blkg. | |
852c788f | 2067 | */ |
6a7f6d86 | 2068 | void bio_associate_blkg_from_page(struct bio *bio, struct page *page) |
852c788f | 2069 | { |
6a7f6d86 DZ |
2070 | struct cgroup_subsys_state *css; |
2071 | ||
6a7f6d86 DZ |
2072 | if (!page->mem_cgroup) |
2073 | return; | |
2074 | ||
fc5a828b DZ |
2075 | rcu_read_lock(); |
2076 | ||
2077 | css = cgroup_e_css(page->mem_cgroup->css.cgroup, &io_cgrp_subsys); | |
2078 | bio_associate_blkg_from_css(bio, css); | |
2079 | ||
2080 | rcu_read_unlock(); | |
6a7f6d86 DZ |
2081 | } |
2082 | #endif /* CONFIG_MEMCG */ | |
2083 | ||
2268c0fe DZ |
2084 | /** |
2085 | * bio_associate_blkg - associate a bio with a blkg | |
2086 | * @bio: target bio | |
2087 | * | |
2088 | * Associate @bio with the blkg found from the bio's css and request_queue. | |
2089 | * If one is not found, bio_lookup_blkg() creates the blkg. If a blkg is | |
2090 | * already associated, the css is reused and association redone as the | |
2091 | * request_queue may have changed. | |
2092 | */ | |
2093 | void bio_associate_blkg(struct bio *bio) | |
2094 | { | |
fc5a828b | 2095 | struct cgroup_subsys_state *css; |
2268c0fe DZ |
2096 | |
2097 | rcu_read_lock(); | |
2098 | ||
db6638d7 | 2099 | if (bio->bi_blkg) |
fc5a828b | 2100 | css = &bio_blkcg(bio)->css; |
db6638d7 | 2101 | else |
fc5a828b | 2102 | css = blkcg_css(); |
2268c0fe | 2103 | |
fc5a828b | 2104 | bio_associate_blkg_from_css(bio, css); |
2268c0fe DZ |
2105 | |
2106 | rcu_read_unlock(); | |
852c788f | 2107 | } |
5cdf2e3f | 2108 | EXPORT_SYMBOL_GPL(bio_associate_blkg); |
852c788f | 2109 | |
20bd723e | 2110 | /** |
db6638d7 | 2111 | * bio_clone_blkg_association - clone blkg association from src to dst bio |
20bd723e PV |
2112 | * @dst: destination bio |
2113 | * @src: source bio | |
2114 | */ | |
db6638d7 | 2115 | void bio_clone_blkg_association(struct bio *dst, struct bio *src) |
20bd723e | 2116 | { |
6ab21879 DZ |
2117 | rcu_read_lock(); |
2118 | ||
fc5a828b | 2119 | if (src->bi_blkg) |
2268c0fe | 2120 | __bio_associate_blkg(dst, src->bi_blkg); |
6ab21879 DZ |
2121 | |
2122 | rcu_read_unlock(); | |
20bd723e | 2123 | } |
db6638d7 | 2124 | EXPORT_SYMBOL_GPL(bio_clone_blkg_association); |
852c788f TH |
2125 | #endif /* CONFIG_BLK_CGROUP */ |
2126 | ||
1da177e4 LT |
2127 | static void __init biovec_init_slabs(void) |
2128 | { | |
2129 | int i; | |
2130 | ||
ed996a52 | 2131 | for (i = 0; i < BVEC_POOL_NR; i++) { |
1da177e4 LT |
2132 | int size; |
2133 | struct biovec_slab *bvs = bvec_slabs + i; | |
2134 | ||
a7fcd37c JA |
2135 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2136 | bvs->slab = NULL; | |
2137 | continue; | |
2138 | } | |
a7fcd37c | 2139 | |
1da177e4 LT |
2140 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2141 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2142 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2143 | } |
2144 | } | |
2145 | ||
2146 | static int __init init_bio(void) | |
2147 | { | |
bb799ca0 JA |
2148 | bio_slab_max = 2; |
2149 | bio_slab_nr = 0; | |
6396bb22 KC |
2150 | bio_slabs = kcalloc(bio_slab_max, sizeof(struct bio_slab), |
2151 | GFP_KERNEL); | |
bb799ca0 JA |
2152 | if (!bio_slabs) |
2153 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2154 | |
7878cba9 | 2155 | bio_integrity_init(); |
1da177e4 LT |
2156 | biovec_init_slabs(); |
2157 | ||
f4f8154a | 2158 | if (bioset_init(&fs_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS)) |
1da177e4 LT |
2159 | panic("bio: can't allocate bios\n"); |
2160 | ||
f4f8154a | 2161 | if (bioset_integrity_create(&fs_bio_set, BIO_POOL_SIZE)) |
a91a2785 MP |
2162 | panic("bio: can't create integrity pool\n"); |
2163 | ||
1da177e4 LT |
2164 | return 0; |
2165 | } | |
1da177e4 | 2166 | subsys_initcall(init_bio); |