Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
0fe23479 | 2 | * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk> |
1da177e4 LT |
3 | * |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of the GNU General Public License version 2 as | |
6 | * published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
11 | * GNU General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public Licens | |
14 | * along with this program; if not, write to the Free Software | |
15 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111- | |
16 | * | |
17 | */ | |
18 | #include <linux/mm.h> | |
19 | #include <linux/swap.h> | |
20 | #include <linux/bio.h> | |
21 | #include <linux/blkdev.h> | |
a27bb332 | 22 | #include <linux/uio.h> |
852c788f | 23 | #include <linux/iocontext.h> |
1da177e4 LT |
24 | #include <linux/slab.h> |
25 | #include <linux/init.h> | |
26 | #include <linux/kernel.h> | |
630d9c47 | 27 | #include <linux/export.h> |
1da177e4 LT |
28 | #include <linux/mempool.h> |
29 | #include <linux/workqueue.h> | |
852c788f | 30 | #include <linux/cgroup.h> |
1da177e4 | 31 | |
55782138 | 32 | #include <trace/events/block.h> |
0bfc2455 | 33 | |
392ddc32 JA |
34 | /* |
35 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
36 | * itself, to shrink a bio data allocation from two mempool calls to one | |
37 | */ | |
38 | #define BIO_INLINE_VECS 4 | |
39 | ||
1da177e4 LT |
40 | /* |
41 | * if you change this list, also change bvec_alloc or things will | |
42 | * break badly! cannot be bigger than what you can fit into an | |
43 | * unsigned short | |
44 | */ | |
1da177e4 | 45 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
df677140 | 46 | static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = { |
1da177e4 LT |
47 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
48 | }; | |
49 | #undef BV | |
50 | ||
1da177e4 LT |
51 | /* |
52 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
53 | * IO code that does not need private memory pools. | |
54 | */ | |
51d654e1 | 55 | struct bio_set *fs_bio_set; |
3f86a82a | 56 | EXPORT_SYMBOL(fs_bio_set); |
1da177e4 | 57 | |
bb799ca0 JA |
58 | /* |
59 | * Our slab pool management | |
60 | */ | |
61 | struct bio_slab { | |
62 | struct kmem_cache *slab; | |
63 | unsigned int slab_ref; | |
64 | unsigned int slab_size; | |
65 | char name[8]; | |
66 | }; | |
67 | static DEFINE_MUTEX(bio_slab_lock); | |
68 | static struct bio_slab *bio_slabs; | |
69 | static unsigned int bio_slab_nr, bio_slab_max; | |
70 | ||
71 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
72 | { | |
73 | unsigned int sz = sizeof(struct bio) + extra_size; | |
74 | struct kmem_cache *slab = NULL; | |
389d7b26 | 75 | struct bio_slab *bslab, *new_bio_slabs; |
386bc35a | 76 | unsigned int new_bio_slab_max; |
bb799ca0 JA |
77 | unsigned int i, entry = -1; |
78 | ||
79 | mutex_lock(&bio_slab_lock); | |
80 | ||
81 | i = 0; | |
82 | while (i < bio_slab_nr) { | |
f06f135d | 83 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
84 | |
85 | if (!bslab->slab && entry == -1) | |
86 | entry = i; | |
87 | else if (bslab->slab_size == sz) { | |
88 | slab = bslab->slab; | |
89 | bslab->slab_ref++; | |
90 | break; | |
91 | } | |
92 | i++; | |
93 | } | |
94 | ||
95 | if (slab) | |
96 | goto out_unlock; | |
97 | ||
98 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
386bc35a | 99 | new_bio_slab_max = bio_slab_max << 1; |
389d7b26 | 100 | new_bio_slabs = krealloc(bio_slabs, |
386bc35a | 101 | new_bio_slab_max * sizeof(struct bio_slab), |
389d7b26 AK |
102 | GFP_KERNEL); |
103 | if (!new_bio_slabs) | |
bb799ca0 | 104 | goto out_unlock; |
386bc35a | 105 | bio_slab_max = new_bio_slab_max; |
389d7b26 | 106 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
107 | } |
108 | if (entry == -1) | |
109 | entry = bio_slab_nr++; | |
110 | ||
111 | bslab = &bio_slabs[entry]; | |
112 | ||
113 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
6a241483 MP |
114 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
115 | SLAB_HWCACHE_ALIGN, NULL); | |
bb799ca0 JA |
116 | if (!slab) |
117 | goto out_unlock; | |
118 | ||
bb799ca0 JA |
119 | bslab->slab = slab; |
120 | bslab->slab_ref = 1; | |
121 | bslab->slab_size = sz; | |
122 | out_unlock: | |
123 | mutex_unlock(&bio_slab_lock); | |
124 | return slab; | |
125 | } | |
126 | ||
127 | static void bio_put_slab(struct bio_set *bs) | |
128 | { | |
129 | struct bio_slab *bslab = NULL; | |
130 | unsigned int i; | |
131 | ||
132 | mutex_lock(&bio_slab_lock); | |
133 | ||
134 | for (i = 0; i < bio_slab_nr; i++) { | |
135 | if (bs->bio_slab == bio_slabs[i].slab) { | |
136 | bslab = &bio_slabs[i]; | |
137 | break; | |
138 | } | |
139 | } | |
140 | ||
141 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
142 | goto out; | |
143 | ||
144 | WARN_ON(!bslab->slab_ref); | |
145 | ||
146 | if (--bslab->slab_ref) | |
147 | goto out; | |
148 | ||
149 | kmem_cache_destroy(bslab->slab); | |
150 | bslab->slab = NULL; | |
151 | ||
152 | out: | |
153 | mutex_unlock(&bio_slab_lock); | |
154 | } | |
155 | ||
7ba1ba12 MP |
156 | unsigned int bvec_nr_vecs(unsigned short idx) |
157 | { | |
158 | return bvec_slabs[idx].nr_vecs; | |
159 | } | |
160 | ||
9f060e22 | 161 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
bb799ca0 JA |
162 | { |
163 | BIO_BUG_ON(idx >= BIOVEC_NR_POOLS); | |
164 | ||
165 | if (idx == BIOVEC_MAX_IDX) | |
9f060e22 | 166 | mempool_free(bv, pool); |
bb799ca0 JA |
167 | else { |
168 | struct biovec_slab *bvs = bvec_slabs + idx; | |
169 | ||
170 | kmem_cache_free(bvs->slab, bv); | |
171 | } | |
172 | } | |
173 | ||
9f060e22 KO |
174 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
175 | mempool_t *pool) | |
1da177e4 LT |
176 | { |
177 | struct bio_vec *bvl; | |
1da177e4 | 178 | |
7ff9345f JA |
179 | /* |
180 | * see comment near bvec_array define! | |
181 | */ | |
182 | switch (nr) { | |
183 | case 1: | |
184 | *idx = 0; | |
185 | break; | |
186 | case 2 ... 4: | |
187 | *idx = 1; | |
188 | break; | |
189 | case 5 ... 16: | |
190 | *idx = 2; | |
191 | break; | |
192 | case 17 ... 64: | |
193 | *idx = 3; | |
194 | break; | |
195 | case 65 ... 128: | |
196 | *idx = 4; | |
197 | break; | |
198 | case 129 ... BIO_MAX_PAGES: | |
199 | *idx = 5; | |
200 | break; | |
201 | default: | |
202 | return NULL; | |
203 | } | |
204 | ||
205 | /* | |
206 | * idx now points to the pool we want to allocate from. only the | |
207 | * 1-vec entry pool is mempool backed. | |
208 | */ | |
209 | if (*idx == BIOVEC_MAX_IDX) { | |
210 | fallback: | |
9f060e22 | 211 | bvl = mempool_alloc(pool, gfp_mask); |
7ff9345f JA |
212 | } else { |
213 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
d0164adc | 214 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
7ff9345f | 215 | |
0a0d96b0 | 216 | /* |
7ff9345f JA |
217 | * Make this allocation restricted and don't dump info on |
218 | * allocation failures, since we'll fallback to the mempool | |
219 | * in case of failure. | |
0a0d96b0 | 220 | */ |
7ff9345f | 221 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 222 | |
0a0d96b0 | 223 | /* |
d0164adc | 224 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
7ff9345f | 225 | * is set, retry with the 1-entry mempool |
0a0d96b0 | 226 | */ |
7ff9345f | 227 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
d0164adc | 228 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
7ff9345f JA |
229 | *idx = BIOVEC_MAX_IDX; |
230 | goto fallback; | |
231 | } | |
232 | } | |
233 | ||
1da177e4 LT |
234 | return bvl; |
235 | } | |
236 | ||
4254bba1 | 237 | static void __bio_free(struct bio *bio) |
1da177e4 | 238 | { |
4254bba1 | 239 | bio_disassociate_task(bio); |
1da177e4 | 240 | |
7ba1ba12 | 241 | if (bio_integrity(bio)) |
1e2a410f | 242 | bio_integrity_free(bio); |
4254bba1 | 243 | } |
7ba1ba12 | 244 | |
4254bba1 KO |
245 | static void bio_free(struct bio *bio) |
246 | { | |
247 | struct bio_set *bs = bio->bi_pool; | |
248 | void *p; | |
249 | ||
250 | __bio_free(bio); | |
251 | ||
252 | if (bs) { | |
a38352e0 | 253 | if (bio_flagged(bio, BIO_OWNS_VEC)) |
9f060e22 | 254 | bvec_free(bs->bvec_pool, bio->bi_io_vec, BIO_POOL_IDX(bio)); |
4254bba1 KO |
255 | |
256 | /* | |
257 | * If we have front padding, adjust the bio pointer before freeing | |
258 | */ | |
259 | p = bio; | |
bb799ca0 JA |
260 | p -= bs->front_pad; |
261 | ||
4254bba1 KO |
262 | mempool_free(p, bs->bio_pool); |
263 | } else { | |
264 | /* Bio was allocated by bio_kmalloc() */ | |
265 | kfree(bio); | |
266 | } | |
3676347a PO |
267 | } |
268 | ||
858119e1 | 269 | void bio_init(struct bio *bio) |
1da177e4 | 270 | { |
2b94de55 | 271 | memset(bio, 0, sizeof(*bio)); |
c4cf5261 | 272 | atomic_set(&bio->__bi_remaining, 1); |
dac56212 | 273 | atomic_set(&bio->__bi_cnt, 1); |
1da177e4 | 274 | } |
a112a71d | 275 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 276 | |
f44b48c7 KO |
277 | /** |
278 | * bio_reset - reinitialize a bio | |
279 | * @bio: bio to reset | |
280 | * | |
281 | * Description: | |
282 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
283 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
284 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
285 | * comment in struct bio. | |
286 | */ | |
287 | void bio_reset(struct bio *bio) | |
288 | { | |
289 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
290 | ||
4254bba1 | 291 | __bio_free(bio); |
f44b48c7 KO |
292 | |
293 | memset(bio, 0, BIO_RESET_BYTES); | |
4246a0b6 | 294 | bio->bi_flags = flags; |
c4cf5261 | 295 | atomic_set(&bio->__bi_remaining, 1); |
f44b48c7 KO |
296 | } |
297 | EXPORT_SYMBOL(bio_reset); | |
298 | ||
38f8baae | 299 | static struct bio *__bio_chain_endio(struct bio *bio) |
196d38bc | 300 | { |
4246a0b6 CH |
301 | struct bio *parent = bio->bi_private; |
302 | ||
af3e3a52 CH |
303 | if (!parent->bi_error) |
304 | parent->bi_error = bio->bi_error; | |
196d38bc | 305 | bio_put(bio); |
38f8baae CH |
306 | return parent; |
307 | } | |
308 | ||
309 | static void bio_chain_endio(struct bio *bio) | |
310 | { | |
311 | bio_endio(__bio_chain_endio(bio)); | |
196d38bc KO |
312 | } |
313 | ||
314 | /** | |
315 | * bio_chain - chain bio completions | |
1051a902 RD |
316 | * @bio: the target bio |
317 | * @parent: the @bio's parent bio | |
196d38bc KO |
318 | * |
319 | * The caller won't have a bi_end_io called when @bio completes - instead, | |
320 | * @parent's bi_end_io won't be called until both @parent and @bio have | |
321 | * completed; the chained bio will also be freed when it completes. | |
322 | * | |
323 | * The caller must not set bi_private or bi_end_io in @bio. | |
324 | */ | |
325 | void bio_chain(struct bio *bio, struct bio *parent) | |
326 | { | |
327 | BUG_ON(bio->bi_private || bio->bi_end_io); | |
328 | ||
329 | bio->bi_private = parent; | |
330 | bio->bi_end_io = bio_chain_endio; | |
c4cf5261 | 331 | bio_inc_remaining(parent); |
196d38bc KO |
332 | } |
333 | EXPORT_SYMBOL(bio_chain); | |
334 | ||
df2cb6da KO |
335 | static void bio_alloc_rescue(struct work_struct *work) |
336 | { | |
337 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); | |
338 | struct bio *bio; | |
339 | ||
340 | while (1) { | |
341 | spin_lock(&bs->rescue_lock); | |
342 | bio = bio_list_pop(&bs->rescue_list); | |
343 | spin_unlock(&bs->rescue_lock); | |
344 | ||
345 | if (!bio) | |
346 | break; | |
347 | ||
348 | generic_make_request(bio); | |
349 | } | |
350 | } | |
351 | ||
352 | static void punt_bios_to_rescuer(struct bio_set *bs) | |
353 | { | |
354 | struct bio_list punt, nopunt; | |
355 | struct bio *bio; | |
356 | ||
357 | /* | |
358 | * In order to guarantee forward progress we must punt only bios that | |
359 | * were allocated from this bio_set; otherwise, if there was a bio on | |
360 | * there for a stacking driver higher up in the stack, processing it | |
361 | * could require allocating bios from this bio_set, and doing that from | |
362 | * our own rescuer would be bad. | |
363 | * | |
364 | * Since bio lists are singly linked, pop them all instead of trying to | |
365 | * remove from the middle of the list: | |
366 | */ | |
367 | ||
368 | bio_list_init(&punt); | |
369 | bio_list_init(&nopunt); | |
370 | ||
371 | while ((bio = bio_list_pop(current->bio_list))) | |
372 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); | |
373 | ||
374 | *current->bio_list = nopunt; | |
375 | ||
376 | spin_lock(&bs->rescue_lock); | |
377 | bio_list_merge(&bs->rescue_list, &punt); | |
378 | spin_unlock(&bs->rescue_lock); | |
379 | ||
380 | queue_work(bs->rescue_workqueue, &bs->rescue_work); | |
381 | } | |
382 | ||
1da177e4 LT |
383 | /** |
384 | * bio_alloc_bioset - allocate a bio for I/O | |
385 | * @gfp_mask: the GFP_ mask given to the slab allocator | |
386 | * @nr_iovecs: number of iovecs to pre-allocate | |
db18efac | 387 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
388 | * |
389 | * Description: | |
3f86a82a KO |
390 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
391 | * backed by the @bs's mempool. | |
392 | * | |
d0164adc MG |
393 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
394 | * always be able to allocate a bio. This is due to the mempool guarantees. | |
395 | * To make this work, callers must never allocate more than 1 bio at a time | |
396 | * from this pool. Callers that need to allocate more than 1 bio must always | |
397 | * submit the previously allocated bio for IO before attempting to allocate | |
398 | * a new one. Failure to do so can cause deadlocks under memory pressure. | |
3f86a82a | 399 | * |
df2cb6da KO |
400 | * Note that when running under generic_make_request() (i.e. any block |
401 | * driver), bios are not submitted until after you return - see the code in | |
402 | * generic_make_request() that converts recursion into iteration, to prevent | |
403 | * stack overflows. | |
404 | * | |
405 | * This would normally mean allocating multiple bios under | |
406 | * generic_make_request() would be susceptible to deadlocks, but we have | |
407 | * deadlock avoidance code that resubmits any blocked bios from a rescuer | |
408 | * thread. | |
409 | * | |
410 | * However, we do not guarantee forward progress for allocations from other | |
411 | * mempools. Doing multiple allocations from the same mempool under | |
412 | * generic_make_request() should be avoided - instead, use bio_set's front_pad | |
413 | * for per bio allocations. | |
414 | * | |
3f86a82a KO |
415 | * RETURNS: |
416 | * Pointer to new bio on success, NULL on failure. | |
417 | */ | |
dd0fc66f | 418 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) |
1da177e4 | 419 | { |
df2cb6da | 420 | gfp_t saved_gfp = gfp_mask; |
3f86a82a KO |
421 | unsigned front_pad; |
422 | unsigned inline_vecs; | |
451a9ebf | 423 | unsigned long idx = BIO_POOL_NONE; |
34053979 | 424 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
425 | struct bio *bio; |
426 | void *p; | |
427 | ||
3f86a82a KO |
428 | if (!bs) { |
429 | if (nr_iovecs > UIO_MAXIOV) | |
430 | return NULL; | |
431 | ||
432 | p = kmalloc(sizeof(struct bio) + | |
433 | nr_iovecs * sizeof(struct bio_vec), | |
434 | gfp_mask); | |
435 | front_pad = 0; | |
436 | inline_vecs = nr_iovecs; | |
437 | } else { | |
d8f429e1 JN |
438 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
439 | if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0)) | |
440 | return NULL; | |
df2cb6da KO |
441 | /* |
442 | * generic_make_request() converts recursion to iteration; this | |
443 | * means if we're running beneath it, any bios we allocate and | |
444 | * submit will not be submitted (and thus freed) until after we | |
445 | * return. | |
446 | * | |
447 | * This exposes us to a potential deadlock if we allocate | |
448 | * multiple bios from the same bio_set() while running | |
449 | * underneath generic_make_request(). If we were to allocate | |
450 | * multiple bios (say a stacking block driver that was splitting | |
451 | * bios), we would deadlock if we exhausted the mempool's | |
452 | * reserve. | |
453 | * | |
454 | * We solve this, and guarantee forward progress, with a rescuer | |
455 | * workqueue per bio_set. If we go to allocate and there are | |
456 | * bios on current->bio_list, we first try the allocation | |
d0164adc MG |
457 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
458 | * bios we would be blocking to the rescuer workqueue before | |
459 | * we retry with the original gfp_flags. | |
df2cb6da KO |
460 | */ |
461 | ||
462 | if (current->bio_list && !bio_list_empty(current->bio_list)) | |
d0164adc | 463 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
df2cb6da | 464 | |
3f86a82a | 465 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
df2cb6da KO |
466 | if (!p && gfp_mask != saved_gfp) { |
467 | punt_bios_to_rescuer(bs); | |
468 | gfp_mask = saved_gfp; | |
469 | p = mempool_alloc(bs->bio_pool, gfp_mask); | |
470 | } | |
471 | ||
3f86a82a KO |
472 | front_pad = bs->front_pad; |
473 | inline_vecs = BIO_INLINE_VECS; | |
474 | } | |
475 | ||
451a9ebf TH |
476 | if (unlikely(!p)) |
477 | return NULL; | |
1da177e4 | 478 | |
3f86a82a | 479 | bio = p + front_pad; |
34053979 IM |
480 | bio_init(bio); |
481 | ||
3f86a82a | 482 | if (nr_iovecs > inline_vecs) { |
9f060e22 | 483 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
484 | if (!bvl && gfp_mask != saved_gfp) { |
485 | punt_bios_to_rescuer(bs); | |
486 | gfp_mask = saved_gfp; | |
9f060e22 | 487 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
df2cb6da KO |
488 | } |
489 | ||
34053979 IM |
490 | if (unlikely(!bvl)) |
491 | goto err_free; | |
a38352e0 | 492 | |
b7c44ed9 | 493 | bio_set_flag(bio, BIO_OWNS_VEC); |
3f86a82a KO |
494 | } else if (nr_iovecs) { |
495 | bvl = bio->bi_inline_vecs; | |
1da177e4 | 496 | } |
3f86a82a KO |
497 | |
498 | bio->bi_pool = bs; | |
34053979 IM |
499 | bio->bi_flags |= idx << BIO_POOL_OFFSET; |
500 | bio->bi_max_vecs = nr_iovecs; | |
34053979 | 501 | bio->bi_io_vec = bvl; |
1da177e4 | 502 | return bio; |
34053979 IM |
503 | |
504 | err_free: | |
451a9ebf | 505 | mempool_free(p, bs->bio_pool); |
34053979 | 506 | return NULL; |
1da177e4 | 507 | } |
a112a71d | 508 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 509 | |
1da177e4 LT |
510 | void zero_fill_bio(struct bio *bio) |
511 | { | |
512 | unsigned long flags; | |
7988613b KO |
513 | struct bio_vec bv; |
514 | struct bvec_iter iter; | |
1da177e4 | 515 | |
7988613b KO |
516 | bio_for_each_segment(bv, bio, iter) { |
517 | char *data = bvec_kmap_irq(&bv, &flags); | |
518 | memset(data, 0, bv.bv_len); | |
519 | flush_dcache_page(bv.bv_page); | |
1da177e4 LT |
520 | bvec_kunmap_irq(data, &flags); |
521 | } | |
522 | } | |
523 | EXPORT_SYMBOL(zero_fill_bio); | |
524 | ||
525 | /** | |
526 | * bio_put - release a reference to a bio | |
527 | * @bio: bio to release reference to | |
528 | * | |
529 | * Description: | |
530 | * Put a reference to a &struct bio, either one you have gotten with | |
ad0bf110 | 531 | * bio_alloc, bio_get or bio_clone. The last put of a bio will free it. |
1da177e4 LT |
532 | **/ |
533 | void bio_put(struct bio *bio) | |
534 | { | |
dac56212 | 535 | if (!bio_flagged(bio, BIO_REFFED)) |
4254bba1 | 536 | bio_free(bio); |
dac56212 JA |
537 | else { |
538 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); | |
539 | ||
540 | /* | |
541 | * last put frees it | |
542 | */ | |
543 | if (atomic_dec_and_test(&bio->__bi_cnt)) | |
544 | bio_free(bio); | |
545 | } | |
1da177e4 | 546 | } |
a112a71d | 547 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 548 | |
165125e1 | 549 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
550 | { |
551 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
552 | blk_recount_segments(q, bio); | |
553 | ||
554 | return bio->bi_phys_segments; | |
555 | } | |
a112a71d | 556 | EXPORT_SYMBOL(bio_phys_segments); |
1da177e4 | 557 | |
59d276fe KO |
558 | /** |
559 | * __bio_clone_fast - clone a bio that shares the original bio's biovec | |
560 | * @bio: destination bio | |
561 | * @bio_src: bio to clone | |
562 | * | |
563 | * Clone a &bio. Caller will own the returned bio, but not | |
564 | * the actual data it points to. Reference count of returned | |
565 | * bio will be one. | |
566 | * | |
567 | * Caller must ensure that @bio_src is not freed before @bio. | |
568 | */ | |
569 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) | |
570 | { | |
571 | BUG_ON(bio->bi_pool && BIO_POOL_IDX(bio) != BIO_POOL_NONE); | |
572 | ||
573 | /* | |
574 | * most users will be overriding ->bi_bdev with a new target, | |
575 | * so we don't set nor calculate new physical/hw segment counts here | |
576 | */ | |
577 | bio->bi_bdev = bio_src->bi_bdev; | |
b7c44ed9 | 578 | bio_set_flag(bio, BIO_CLONED); |
59d276fe KO |
579 | bio->bi_rw = bio_src->bi_rw; |
580 | bio->bi_iter = bio_src->bi_iter; | |
581 | bio->bi_io_vec = bio_src->bi_io_vec; | |
582 | } | |
583 | EXPORT_SYMBOL(__bio_clone_fast); | |
584 | ||
585 | /** | |
586 | * bio_clone_fast - clone a bio that shares the original bio's biovec | |
587 | * @bio: bio to clone | |
588 | * @gfp_mask: allocation priority | |
589 | * @bs: bio_set to allocate from | |
590 | * | |
591 | * Like __bio_clone_fast, only also allocates the returned bio | |
592 | */ | |
593 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) | |
594 | { | |
595 | struct bio *b; | |
596 | ||
597 | b = bio_alloc_bioset(gfp_mask, 0, bs); | |
598 | if (!b) | |
599 | return NULL; | |
600 | ||
601 | __bio_clone_fast(b, bio); | |
602 | ||
603 | if (bio_integrity(bio)) { | |
604 | int ret; | |
605 | ||
606 | ret = bio_integrity_clone(b, bio, gfp_mask); | |
607 | ||
608 | if (ret < 0) { | |
609 | bio_put(b); | |
610 | return NULL; | |
611 | } | |
612 | } | |
613 | ||
614 | return b; | |
615 | } | |
616 | EXPORT_SYMBOL(bio_clone_fast); | |
617 | ||
1da177e4 | 618 | /** |
bdb53207 KO |
619 | * bio_clone_bioset - clone a bio |
620 | * @bio_src: bio to clone | |
1da177e4 | 621 | * @gfp_mask: allocation priority |
bf800ef1 | 622 | * @bs: bio_set to allocate from |
1da177e4 | 623 | * |
bdb53207 KO |
624 | * Clone bio. Caller will own the returned bio, but not the actual data it |
625 | * points to. Reference count of returned bio will be one. | |
1da177e4 | 626 | */ |
bdb53207 | 627 | struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask, |
bf800ef1 | 628 | struct bio_set *bs) |
1da177e4 | 629 | { |
bdb53207 KO |
630 | struct bvec_iter iter; |
631 | struct bio_vec bv; | |
632 | struct bio *bio; | |
1da177e4 | 633 | |
bdb53207 KO |
634 | /* |
635 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from | |
636 | * bio_src->bi_io_vec to bio->bi_io_vec. | |
637 | * | |
638 | * We can't do that anymore, because: | |
639 | * | |
640 | * - The point of cloning the biovec is to produce a bio with a biovec | |
641 | * the caller can modify: bi_idx and bi_bvec_done should be 0. | |
642 | * | |
643 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if | |
644 | * we tried to clone the whole thing bio_alloc_bioset() would fail. | |
645 | * But the clone should succeed as long as the number of biovecs we | |
646 | * actually need to allocate is fewer than BIO_MAX_PAGES. | |
647 | * | |
648 | * - Lastly, bi_vcnt should not be looked at or relied upon by code | |
649 | * that does not own the bio - reason being drivers don't use it for | |
650 | * iterating over the biovec anymore, so expecting it to be kept up | |
651 | * to date (i.e. for clones that share the parent biovec) is just | |
652 | * asking for trouble and would force extra work on | |
653 | * __bio_clone_fast() anyways. | |
654 | */ | |
655 | ||
8423ae3d | 656 | bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs); |
bdb53207 | 657 | if (!bio) |
7ba1ba12 | 658 | return NULL; |
bdb53207 KO |
659 | bio->bi_bdev = bio_src->bi_bdev; |
660 | bio->bi_rw = bio_src->bi_rw; | |
661 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; | |
662 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; | |
7ba1ba12 | 663 | |
95fe6c1a | 664 | if (bio_op(bio) == REQ_OP_DISCARD) |
8423ae3d KO |
665 | goto integrity_clone; |
666 | ||
95fe6c1a | 667 | if (bio_op(bio) == REQ_OP_WRITE_SAME) { |
8423ae3d KO |
668 | bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0]; |
669 | goto integrity_clone; | |
670 | } | |
671 | ||
bdb53207 KO |
672 | bio_for_each_segment(bv, bio_src, iter) |
673 | bio->bi_io_vec[bio->bi_vcnt++] = bv; | |
7ba1ba12 | 674 | |
8423ae3d | 675 | integrity_clone: |
bdb53207 KO |
676 | if (bio_integrity(bio_src)) { |
677 | int ret; | |
7ba1ba12 | 678 | |
bdb53207 | 679 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
059ea331 | 680 | if (ret < 0) { |
bdb53207 | 681 | bio_put(bio); |
7ba1ba12 | 682 | return NULL; |
059ea331 | 683 | } |
3676347a | 684 | } |
1da177e4 | 685 | |
bdb53207 | 686 | return bio; |
1da177e4 | 687 | } |
bf800ef1 | 688 | EXPORT_SYMBOL(bio_clone_bioset); |
1da177e4 LT |
689 | |
690 | /** | |
c66a14d0 KO |
691 | * bio_add_pc_page - attempt to add page to bio |
692 | * @q: the target queue | |
693 | * @bio: destination bio | |
694 | * @page: page to add | |
695 | * @len: vec entry length | |
696 | * @offset: vec entry offset | |
1da177e4 | 697 | * |
c66a14d0 KO |
698 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
699 | * number of reasons, such as the bio being full or target block device | |
700 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
701 | * so it is always possible to add a single page to an empty bio. | |
702 | * | |
703 | * This should only be used by REQ_PC bios. | |
1da177e4 | 704 | */ |
c66a14d0 KO |
705 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
706 | *page, unsigned int len, unsigned int offset) | |
1da177e4 LT |
707 | { |
708 | int retried_segments = 0; | |
709 | struct bio_vec *bvec; | |
710 | ||
711 | /* | |
712 | * cloned bio must not modify vec list | |
713 | */ | |
714 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
715 | return 0; | |
716 | ||
c66a14d0 | 717 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
718 | return 0; |
719 | ||
80cfd548 JA |
720 | /* |
721 | * For filesystems with a blocksize smaller than the pagesize | |
722 | * we will often be called with the same page as last time and | |
723 | * a consecutive offset. Optimize this special case. | |
724 | */ | |
725 | if (bio->bi_vcnt > 0) { | |
726 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
727 | ||
728 | if (page == prev->bv_page && | |
729 | offset == prev->bv_offset + prev->bv_len) { | |
730 | prev->bv_len += len; | |
fcbf6a08 | 731 | bio->bi_iter.bi_size += len; |
80cfd548 JA |
732 | goto done; |
733 | } | |
66cb45aa JA |
734 | |
735 | /* | |
736 | * If the queue doesn't support SG gaps and adding this | |
737 | * offset would create a gap, disallow it. | |
738 | */ | |
03100aad | 739 | if (bvec_gap_to_prev(q, prev, offset)) |
66cb45aa | 740 | return 0; |
80cfd548 JA |
741 | } |
742 | ||
743 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
1da177e4 LT |
744 | return 0; |
745 | ||
746 | /* | |
fcbf6a08 ML |
747 | * setup the new entry, we might clear it again later if we |
748 | * cannot add the page | |
749 | */ | |
750 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; | |
751 | bvec->bv_page = page; | |
752 | bvec->bv_len = len; | |
753 | bvec->bv_offset = offset; | |
754 | bio->bi_vcnt++; | |
755 | bio->bi_phys_segments++; | |
756 | bio->bi_iter.bi_size += len; | |
757 | ||
758 | /* | |
759 | * Perform a recount if the number of segments is greater | |
760 | * than queue_max_segments(q). | |
1da177e4 LT |
761 | */ |
762 | ||
fcbf6a08 | 763 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
1da177e4 LT |
764 | |
765 | if (retried_segments) | |
fcbf6a08 | 766 | goto failed; |
1da177e4 LT |
767 | |
768 | retried_segments = 1; | |
769 | blk_recount_segments(q, bio); | |
770 | } | |
771 | ||
1da177e4 | 772 | /* If we may be able to merge these biovecs, force a recount */ |
fcbf6a08 | 773 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
b7c44ed9 | 774 | bio_clear_flag(bio, BIO_SEG_VALID); |
1da177e4 | 775 | |
80cfd548 | 776 | done: |
1da177e4 | 777 | return len; |
fcbf6a08 ML |
778 | |
779 | failed: | |
780 | bvec->bv_page = NULL; | |
781 | bvec->bv_len = 0; | |
782 | bvec->bv_offset = 0; | |
783 | bio->bi_vcnt--; | |
784 | bio->bi_iter.bi_size -= len; | |
785 | blk_recount_segments(q, bio); | |
786 | return 0; | |
1da177e4 | 787 | } |
a112a71d | 788 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 789 | |
1da177e4 LT |
790 | /** |
791 | * bio_add_page - attempt to add page to bio | |
792 | * @bio: destination bio | |
793 | * @page: page to add | |
794 | * @len: vec entry length | |
795 | * @offset: vec entry offset | |
796 | * | |
c66a14d0 KO |
797 | * Attempt to add a page to the bio_vec maplist. This will only fail |
798 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. | |
1da177e4 | 799 | */ |
c66a14d0 KO |
800 | int bio_add_page(struct bio *bio, struct page *page, |
801 | unsigned int len, unsigned int offset) | |
1da177e4 | 802 | { |
c66a14d0 KO |
803 | struct bio_vec *bv; |
804 | ||
805 | /* | |
806 | * cloned bio must not modify vec list | |
807 | */ | |
808 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) | |
809 | return 0; | |
762380ad | 810 | |
c66a14d0 KO |
811 | /* |
812 | * For filesystems with a blocksize smaller than the pagesize | |
813 | * we will often be called with the same page as last time and | |
814 | * a consecutive offset. Optimize this special case. | |
815 | */ | |
816 | if (bio->bi_vcnt > 0) { | |
817 | bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; | |
58a4915a | 818 | |
c66a14d0 KO |
819 | if (page == bv->bv_page && |
820 | offset == bv->bv_offset + bv->bv_len) { | |
821 | bv->bv_len += len; | |
822 | goto done; | |
823 | } | |
824 | } | |
825 | ||
826 | if (bio->bi_vcnt >= bio->bi_max_vecs) | |
827 | return 0; | |
828 | ||
829 | bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
830 | bv->bv_page = page; | |
831 | bv->bv_len = len; | |
832 | bv->bv_offset = offset; | |
833 | ||
834 | bio->bi_vcnt++; | |
835 | done: | |
836 | bio->bi_iter.bi_size += len; | |
837 | return len; | |
1da177e4 | 838 | } |
a112a71d | 839 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 840 | |
9e882242 KO |
841 | struct submit_bio_ret { |
842 | struct completion event; | |
843 | int error; | |
844 | }; | |
845 | ||
4246a0b6 | 846 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 KO |
847 | { |
848 | struct submit_bio_ret *ret = bio->bi_private; | |
849 | ||
4246a0b6 | 850 | ret->error = bio->bi_error; |
9e882242 KO |
851 | complete(&ret->event); |
852 | } | |
853 | ||
854 | /** | |
855 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
856 | * @bio: The &struct bio which describes the I/O |
857 | * | |
858 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
859 | * bio_endio() on failure. | |
860 | */ | |
4e49ea4a | 861 | int submit_bio_wait(struct bio *bio) |
9e882242 KO |
862 | { |
863 | struct submit_bio_ret ret; | |
864 | ||
9e882242 KO |
865 | init_completion(&ret.event); |
866 | bio->bi_private = &ret; | |
867 | bio->bi_end_io = submit_bio_wait_endio; | |
4e49ea4a MC |
868 | bio->bi_rw |= REQ_SYNC; |
869 | submit_bio(bio); | |
d57d6115 | 870 | wait_for_completion_io(&ret.event); |
9e882242 KO |
871 | |
872 | return ret.error; | |
873 | } | |
874 | EXPORT_SYMBOL(submit_bio_wait); | |
875 | ||
054bdf64 KO |
876 | /** |
877 | * bio_advance - increment/complete a bio by some number of bytes | |
878 | * @bio: bio to advance | |
879 | * @bytes: number of bytes to complete | |
880 | * | |
881 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
882 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
883 | * be updated on the last bvec as well. | |
884 | * | |
885 | * @bio will then represent the remaining, uncompleted portion of the io. | |
886 | */ | |
887 | void bio_advance(struct bio *bio, unsigned bytes) | |
888 | { | |
889 | if (bio_integrity(bio)) | |
890 | bio_integrity_advance(bio, bytes); | |
891 | ||
4550dd6c | 892 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
893 | } |
894 | EXPORT_SYMBOL(bio_advance); | |
895 | ||
a0787606 KO |
896 | /** |
897 | * bio_alloc_pages - allocates a single page for each bvec in a bio | |
898 | * @bio: bio to allocate pages for | |
899 | * @gfp_mask: flags for allocation | |
900 | * | |
901 | * Allocates pages up to @bio->bi_vcnt. | |
902 | * | |
903 | * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are | |
904 | * freed. | |
905 | */ | |
906 | int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask) | |
907 | { | |
908 | int i; | |
909 | struct bio_vec *bv; | |
910 | ||
911 | bio_for_each_segment_all(bv, bio, i) { | |
912 | bv->bv_page = alloc_page(gfp_mask); | |
913 | if (!bv->bv_page) { | |
914 | while (--bv >= bio->bi_io_vec) | |
915 | __free_page(bv->bv_page); | |
916 | return -ENOMEM; | |
917 | } | |
918 | } | |
919 | ||
920 | return 0; | |
921 | } | |
922 | EXPORT_SYMBOL(bio_alloc_pages); | |
923 | ||
16ac3d63 KO |
924 | /** |
925 | * bio_copy_data - copy contents of data buffers from one chain of bios to | |
926 | * another | |
927 | * @src: source bio list | |
928 | * @dst: destination bio list | |
929 | * | |
930 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats | |
931 | * @src and @dst as linked lists of bios. | |
932 | * | |
933 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
934 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
935 | */ | |
936 | void bio_copy_data(struct bio *dst, struct bio *src) | |
937 | { | |
1cb9dda4 KO |
938 | struct bvec_iter src_iter, dst_iter; |
939 | struct bio_vec src_bv, dst_bv; | |
16ac3d63 | 940 | void *src_p, *dst_p; |
1cb9dda4 | 941 | unsigned bytes; |
16ac3d63 | 942 | |
1cb9dda4 KO |
943 | src_iter = src->bi_iter; |
944 | dst_iter = dst->bi_iter; | |
16ac3d63 KO |
945 | |
946 | while (1) { | |
1cb9dda4 KO |
947 | if (!src_iter.bi_size) { |
948 | src = src->bi_next; | |
949 | if (!src) | |
950 | break; | |
16ac3d63 | 951 | |
1cb9dda4 | 952 | src_iter = src->bi_iter; |
16ac3d63 KO |
953 | } |
954 | ||
1cb9dda4 KO |
955 | if (!dst_iter.bi_size) { |
956 | dst = dst->bi_next; | |
957 | if (!dst) | |
958 | break; | |
16ac3d63 | 959 | |
1cb9dda4 | 960 | dst_iter = dst->bi_iter; |
16ac3d63 KO |
961 | } |
962 | ||
1cb9dda4 KO |
963 | src_bv = bio_iter_iovec(src, src_iter); |
964 | dst_bv = bio_iter_iovec(dst, dst_iter); | |
965 | ||
966 | bytes = min(src_bv.bv_len, dst_bv.bv_len); | |
16ac3d63 | 967 | |
1cb9dda4 KO |
968 | src_p = kmap_atomic(src_bv.bv_page); |
969 | dst_p = kmap_atomic(dst_bv.bv_page); | |
16ac3d63 | 970 | |
1cb9dda4 KO |
971 | memcpy(dst_p + dst_bv.bv_offset, |
972 | src_p + src_bv.bv_offset, | |
16ac3d63 KO |
973 | bytes); |
974 | ||
975 | kunmap_atomic(dst_p); | |
976 | kunmap_atomic(src_p); | |
977 | ||
1cb9dda4 KO |
978 | bio_advance_iter(src, &src_iter, bytes); |
979 | bio_advance_iter(dst, &dst_iter, bytes); | |
16ac3d63 KO |
980 | } |
981 | } | |
982 | EXPORT_SYMBOL(bio_copy_data); | |
983 | ||
1da177e4 | 984 | struct bio_map_data { |
152e283f | 985 | int is_our_pages; |
26e49cfc KO |
986 | struct iov_iter iter; |
987 | struct iovec iov[]; | |
1da177e4 LT |
988 | }; |
989 | ||
7410b3c6 | 990 | static struct bio_map_data *bio_alloc_map_data(unsigned int iov_count, |
76029ff3 | 991 | gfp_t gfp_mask) |
1da177e4 | 992 | { |
f3f63c1c JA |
993 | if (iov_count > UIO_MAXIOV) |
994 | return NULL; | |
1da177e4 | 995 | |
c8db4448 | 996 | return kmalloc(sizeof(struct bio_map_data) + |
26e49cfc | 997 | sizeof(struct iovec) * iov_count, gfp_mask); |
1da177e4 LT |
998 | } |
999 | ||
9124d3fe DP |
1000 | /** |
1001 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1002 | * @bio: The &struct bio which describes the I/O as destination | |
1003 | * @iter: iov_iter as source | |
1004 | * | |
1005 | * Copy all pages from iov_iter to bio. | |
1006 | * Returns 0 on success, or error on failure. | |
1007 | */ | |
1008 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter iter) | |
c5dec1c3 | 1009 | { |
9124d3fe | 1010 | int i; |
c5dec1c3 | 1011 | struct bio_vec *bvec; |
c5dec1c3 | 1012 | |
d74c6d51 | 1013 | bio_for_each_segment_all(bvec, bio, i) { |
9124d3fe | 1014 | ssize_t ret; |
c5dec1c3 | 1015 | |
9124d3fe DP |
1016 | ret = copy_page_from_iter(bvec->bv_page, |
1017 | bvec->bv_offset, | |
1018 | bvec->bv_len, | |
1019 | &iter); | |
1020 | ||
1021 | if (!iov_iter_count(&iter)) | |
1022 | break; | |
1023 | ||
1024 | if (ret < bvec->bv_len) | |
1025 | return -EFAULT; | |
c5dec1c3 FT |
1026 | } |
1027 | ||
9124d3fe DP |
1028 | return 0; |
1029 | } | |
1030 | ||
1031 | /** | |
1032 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1033 | * @bio: The &struct bio which describes the I/O as source | |
1034 | * @iter: iov_iter as destination | |
1035 | * | |
1036 | * Copy all pages from bio to iov_iter. | |
1037 | * Returns 0 on success, or error on failure. | |
1038 | */ | |
1039 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1040 | { | |
1041 | int i; | |
1042 | struct bio_vec *bvec; | |
1043 | ||
1044 | bio_for_each_segment_all(bvec, bio, i) { | |
1045 | ssize_t ret; | |
1046 | ||
1047 | ret = copy_page_to_iter(bvec->bv_page, | |
1048 | bvec->bv_offset, | |
1049 | bvec->bv_len, | |
1050 | &iter); | |
1051 | ||
1052 | if (!iov_iter_count(&iter)) | |
1053 | break; | |
1054 | ||
1055 | if (ret < bvec->bv_len) | |
1056 | return -EFAULT; | |
1057 | } | |
1058 | ||
1059 | return 0; | |
c5dec1c3 FT |
1060 | } |
1061 | ||
1dfa0f68 CH |
1062 | static void bio_free_pages(struct bio *bio) |
1063 | { | |
1064 | struct bio_vec *bvec; | |
1065 | int i; | |
1066 | ||
1067 | bio_for_each_segment_all(bvec, bio, i) | |
1068 | __free_page(bvec->bv_page); | |
1069 | } | |
1070 | ||
1da177e4 LT |
1071 | /** |
1072 | * bio_uncopy_user - finish previously mapped bio | |
1073 | * @bio: bio being terminated | |
1074 | * | |
ddad8dd0 | 1075 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1076 | * to user space in case of a read. |
1077 | */ | |
1078 | int bio_uncopy_user(struct bio *bio) | |
1079 | { | |
1080 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1081 | int ret = 0; |
1da177e4 | 1082 | |
35dc2483 RD |
1083 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1084 | /* | |
1085 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1086 | * don't copy into a random user address space, just free |
1087 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1088 | */ |
2d99b55d HR |
1089 | if (!current->mm) |
1090 | ret = -EINTR; | |
1091 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1092 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1093 | if (bmd->is_our_pages) |
1094 | bio_free_pages(bio); | |
35dc2483 | 1095 | } |
c8db4448 | 1096 | kfree(bmd); |
1da177e4 LT |
1097 | bio_put(bio); |
1098 | return ret; | |
1099 | } | |
a112a71d | 1100 | EXPORT_SYMBOL(bio_uncopy_user); |
1da177e4 LT |
1101 | |
1102 | /** | |
c5dec1c3 | 1103 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1104 | * @q: destination block queue |
1105 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1106 | * @iter: iovec iterator | |
1107 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1108 | * |
1109 | * Prepares and returns a bio for indirect user io, bouncing data | |
1110 | * to/from kernel pages as necessary. Must be paired with | |
1111 | * call bio_uncopy_user() on io completion. | |
1112 | */ | |
152e283f FT |
1113 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1114 | struct rq_map_data *map_data, | |
26e49cfc KO |
1115 | const struct iov_iter *iter, |
1116 | gfp_t gfp_mask) | |
1da177e4 | 1117 | { |
1da177e4 | 1118 | struct bio_map_data *bmd; |
1da177e4 LT |
1119 | struct page *page; |
1120 | struct bio *bio; | |
1121 | int i, ret; | |
c5dec1c3 | 1122 | int nr_pages = 0; |
26e49cfc | 1123 | unsigned int len = iter->count; |
bd5cecea | 1124 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1125 | |
26e49cfc | 1126 | for (i = 0; i < iter->nr_segs; i++) { |
c5dec1c3 FT |
1127 | unsigned long uaddr; |
1128 | unsigned long end; | |
1129 | unsigned long start; | |
1130 | ||
26e49cfc KO |
1131 | uaddr = (unsigned long) iter->iov[i].iov_base; |
1132 | end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1) | |
1133 | >> PAGE_SHIFT; | |
c5dec1c3 FT |
1134 | start = uaddr >> PAGE_SHIFT; |
1135 | ||
cb4644ca JA |
1136 | /* |
1137 | * Overflow, abort | |
1138 | */ | |
1139 | if (end < start) | |
1140 | return ERR_PTR(-EINVAL); | |
1141 | ||
c5dec1c3 | 1142 | nr_pages += end - start; |
c5dec1c3 FT |
1143 | } |
1144 | ||
69838727 FT |
1145 | if (offset) |
1146 | nr_pages++; | |
1147 | ||
26e49cfc | 1148 | bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask); |
1da177e4 LT |
1149 | if (!bmd) |
1150 | return ERR_PTR(-ENOMEM); | |
1151 | ||
26e49cfc KO |
1152 | /* |
1153 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1154 | * The caller provided iov might point to an on-stack or otherwise | |
1155 | * shortlived one. | |
1156 | */ | |
1157 | bmd->is_our_pages = map_data ? 0 : 1; | |
1158 | memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs); | |
1159 | iov_iter_init(&bmd->iter, iter->type, bmd->iov, | |
1160 | iter->nr_segs, iter->count); | |
1161 | ||
1da177e4 | 1162 | ret = -ENOMEM; |
a9e9dc24 | 1163 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1164 | if (!bio) |
1165 | goto out_bmd; | |
1166 | ||
26e49cfc | 1167 | if (iter->type & WRITE) |
95fe6c1a | 1168 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
1da177e4 LT |
1169 | |
1170 | ret = 0; | |
56c451f4 FT |
1171 | |
1172 | if (map_data) { | |
e623ddb4 | 1173 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1174 | i = map_data->offset / PAGE_SIZE; |
1175 | } | |
1da177e4 | 1176 | while (len) { |
e623ddb4 | 1177 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1178 | |
56c451f4 FT |
1179 | bytes -= offset; |
1180 | ||
1da177e4 LT |
1181 | if (bytes > len) |
1182 | bytes = len; | |
1183 | ||
152e283f | 1184 | if (map_data) { |
e623ddb4 | 1185 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1186 | ret = -ENOMEM; |
1187 | break; | |
1188 | } | |
e623ddb4 FT |
1189 | |
1190 | page = map_data->pages[i / nr_pages]; | |
1191 | page += (i % nr_pages); | |
1192 | ||
1193 | i++; | |
1194 | } else { | |
152e283f | 1195 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1196 | if (!page) { |
1197 | ret = -ENOMEM; | |
1198 | break; | |
1199 | } | |
1da177e4 LT |
1200 | } |
1201 | ||
56c451f4 | 1202 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
1da177e4 | 1203 | break; |
1da177e4 LT |
1204 | |
1205 | len -= bytes; | |
56c451f4 | 1206 | offset = 0; |
1da177e4 LT |
1207 | } |
1208 | ||
1209 | if (ret) | |
1210 | goto cleanup; | |
1211 | ||
1212 | /* | |
1213 | * success | |
1214 | */ | |
26e49cfc | 1215 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1216 | (map_data && map_data->from_user)) { |
9124d3fe | 1217 | ret = bio_copy_from_iter(bio, *iter); |
c5dec1c3 FT |
1218 | if (ret) |
1219 | goto cleanup; | |
1da177e4 LT |
1220 | } |
1221 | ||
26e49cfc | 1222 | bio->bi_private = bmd; |
1da177e4 LT |
1223 | return bio; |
1224 | cleanup: | |
152e283f | 1225 | if (!map_data) |
1dfa0f68 | 1226 | bio_free_pages(bio); |
1da177e4 LT |
1227 | bio_put(bio); |
1228 | out_bmd: | |
c8db4448 | 1229 | kfree(bmd); |
1da177e4 LT |
1230 | return ERR_PTR(ret); |
1231 | } | |
1232 | ||
37f19e57 CH |
1233 | /** |
1234 | * bio_map_user_iov - map user iovec into bio | |
1235 | * @q: the struct request_queue for the bio | |
1236 | * @iter: iovec iterator | |
1237 | * @gfp_mask: memory allocation flags | |
1238 | * | |
1239 | * Map the user space address into a bio suitable for io to a block | |
1240 | * device. Returns an error pointer in case of error. | |
1241 | */ | |
1242 | struct bio *bio_map_user_iov(struct request_queue *q, | |
1243 | const struct iov_iter *iter, | |
1244 | gfp_t gfp_mask) | |
1da177e4 | 1245 | { |
26e49cfc | 1246 | int j; |
f1970baf | 1247 | int nr_pages = 0; |
1da177e4 LT |
1248 | struct page **pages; |
1249 | struct bio *bio; | |
f1970baf JB |
1250 | int cur_page = 0; |
1251 | int ret, offset; | |
26e49cfc KO |
1252 | struct iov_iter i; |
1253 | struct iovec iov; | |
1da177e4 | 1254 | |
26e49cfc KO |
1255 | iov_for_each(iov, i, *iter) { |
1256 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1257 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1258 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1259 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1260 | ||
cb4644ca JA |
1261 | /* |
1262 | * Overflow, abort | |
1263 | */ | |
1264 | if (end < start) | |
1265 | return ERR_PTR(-EINVAL); | |
1266 | ||
f1970baf JB |
1267 | nr_pages += end - start; |
1268 | /* | |
ad2d7225 | 1269 | * buffer must be aligned to at least hardsector size for now |
f1970baf | 1270 | */ |
ad2d7225 | 1271 | if (uaddr & queue_dma_alignment(q)) |
f1970baf JB |
1272 | return ERR_PTR(-EINVAL); |
1273 | } | |
1274 | ||
1275 | if (!nr_pages) | |
1da177e4 LT |
1276 | return ERR_PTR(-EINVAL); |
1277 | ||
a9e9dc24 | 1278 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1279 | if (!bio) |
1280 | return ERR_PTR(-ENOMEM); | |
1281 | ||
1282 | ret = -ENOMEM; | |
a3bce90e | 1283 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
1da177e4 LT |
1284 | if (!pages) |
1285 | goto out; | |
1286 | ||
26e49cfc KO |
1287 | iov_for_each(iov, i, *iter) { |
1288 | unsigned long uaddr = (unsigned long) iov.iov_base; | |
1289 | unsigned long len = iov.iov_len; | |
f1970baf JB |
1290 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1291 | unsigned long start = uaddr >> PAGE_SHIFT; | |
1292 | const int local_nr_pages = end - start; | |
1293 | const int page_limit = cur_page + local_nr_pages; | |
cb4644ca | 1294 | |
f5dd33c4 | 1295 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
26e49cfc KO |
1296 | (iter->type & WRITE) != WRITE, |
1297 | &pages[cur_page]); | |
99172157 JA |
1298 | if (ret < local_nr_pages) { |
1299 | ret = -EFAULT; | |
f1970baf | 1300 | goto out_unmap; |
99172157 | 1301 | } |
f1970baf | 1302 | |
bd5cecea | 1303 | offset = offset_in_page(uaddr); |
f1970baf JB |
1304 | for (j = cur_page; j < page_limit; j++) { |
1305 | unsigned int bytes = PAGE_SIZE - offset; | |
1306 | ||
1307 | if (len <= 0) | |
1308 | break; | |
1309 | ||
1310 | if (bytes > len) | |
1311 | bytes = len; | |
1312 | ||
1313 | /* | |
1314 | * sorry... | |
1315 | */ | |
defd94b7 MC |
1316 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
1317 | bytes) | |
f1970baf JB |
1318 | break; |
1319 | ||
1320 | len -= bytes; | |
1321 | offset = 0; | |
1322 | } | |
1da177e4 | 1323 | |
f1970baf | 1324 | cur_page = j; |
1da177e4 | 1325 | /* |
f1970baf | 1326 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1327 | */ |
f1970baf | 1328 | while (j < page_limit) |
09cbfeaf | 1329 | put_page(pages[j++]); |
1da177e4 LT |
1330 | } |
1331 | ||
1da177e4 LT |
1332 | kfree(pages); |
1333 | ||
1334 | /* | |
1335 | * set data direction, and check if mapped pages need bouncing | |
1336 | */ | |
26e49cfc | 1337 | if (iter->type & WRITE) |
95fe6c1a | 1338 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
1da177e4 | 1339 | |
b7c44ed9 | 1340 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1341 | |
1342 | /* | |
1343 | * subtle -- if __bio_map_user() ended up bouncing a bio, | |
1344 | * it would normally disappear when its bi_end_io is run. | |
1345 | * however, we need it for the unmap, so grab an extra | |
1346 | * reference to it | |
1347 | */ | |
1348 | bio_get(bio); | |
1da177e4 | 1349 | return bio; |
f1970baf JB |
1350 | |
1351 | out_unmap: | |
26e49cfc KO |
1352 | for (j = 0; j < nr_pages; j++) { |
1353 | if (!pages[j]) | |
f1970baf | 1354 | break; |
09cbfeaf | 1355 | put_page(pages[j]); |
f1970baf JB |
1356 | } |
1357 | out: | |
1da177e4 LT |
1358 | kfree(pages); |
1359 | bio_put(bio); | |
1360 | return ERR_PTR(ret); | |
1361 | } | |
1362 | ||
1da177e4 LT |
1363 | static void __bio_unmap_user(struct bio *bio) |
1364 | { | |
1365 | struct bio_vec *bvec; | |
1366 | int i; | |
1367 | ||
1368 | /* | |
1369 | * make sure we dirty pages we wrote to | |
1370 | */ | |
d74c6d51 | 1371 | bio_for_each_segment_all(bvec, bio, i) { |
1da177e4 LT |
1372 | if (bio_data_dir(bio) == READ) |
1373 | set_page_dirty_lock(bvec->bv_page); | |
1374 | ||
09cbfeaf | 1375 | put_page(bvec->bv_page); |
1da177e4 LT |
1376 | } |
1377 | ||
1378 | bio_put(bio); | |
1379 | } | |
1380 | ||
1381 | /** | |
1382 | * bio_unmap_user - unmap a bio | |
1383 | * @bio: the bio being unmapped | |
1384 | * | |
1385 | * Unmap a bio previously mapped by bio_map_user(). Must be called with | |
1386 | * a process context. | |
1387 | * | |
1388 | * bio_unmap_user() may sleep. | |
1389 | */ | |
1390 | void bio_unmap_user(struct bio *bio) | |
1391 | { | |
1392 | __bio_unmap_user(bio); | |
1393 | bio_put(bio); | |
1394 | } | |
a112a71d | 1395 | EXPORT_SYMBOL(bio_unmap_user); |
1da177e4 | 1396 | |
4246a0b6 | 1397 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1398 | { |
b823825e | 1399 | bio_put(bio); |
b823825e JA |
1400 | } |
1401 | ||
75c72b83 CH |
1402 | /** |
1403 | * bio_map_kern - map kernel address into bio | |
1404 | * @q: the struct request_queue for the bio | |
1405 | * @data: pointer to buffer to map | |
1406 | * @len: length in bytes | |
1407 | * @gfp_mask: allocation flags for bio allocation | |
1408 | * | |
1409 | * Map the kernel address into a bio suitable for io to a block | |
1410 | * device. Returns an error pointer in case of error. | |
1411 | */ | |
1412 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1413 | gfp_t gfp_mask) | |
df46b9a4 MC |
1414 | { |
1415 | unsigned long kaddr = (unsigned long)data; | |
1416 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1417 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1418 | const int nr_pages = end - start; | |
1419 | int offset, i; | |
1420 | struct bio *bio; | |
1421 | ||
a9e9dc24 | 1422 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1423 | if (!bio) |
1424 | return ERR_PTR(-ENOMEM); | |
1425 | ||
1426 | offset = offset_in_page(kaddr); | |
1427 | for (i = 0; i < nr_pages; i++) { | |
1428 | unsigned int bytes = PAGE_SIZE - offset; | |
1429 | ||
1430 | if (len <= 0) | |
1431 | break; | |
1432 | ||
1433 | if (bytes > len) | |
1434 | bytes = len; | |
1435 | ||
defd94b7 | 1436 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1437 | offset) < bytes) { |
1438 | /* we don't support partial mappings */ | |
1439 | bio_put(bio); | |
1440 | return ERR_PTR(-EINVAL); | |
1441 | } | |
df46b9a4 MC |
1442 | |
1443 | data += bytes; | |
1444 | len -= bytes; | |
1445 | offset = 0; | |
1446 | } | |
1447 | ||
b823825e | 1448 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1449 | return bio; |
1450 | } | |
a112a71d | 1451 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1452 | |
4246a0b6 | 1453 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1454 | { |
1dfa0f68 CH |
1455 | bio_free_pages(bio); |
1456 | bio_put(bio); | |
1457 | } | |
1458 | ||
4246a0b6 | 1459 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1460 | { |
42d2683a | 1461 | char *p = bio->bi_private; |
1dfa0f68 | 1462 | struct bio_vec *bvec; |
68154e90 FT |
1463 | int i; |
1464 | ||
d74c6d51 | 1465 | bio_for_each_segment_all(bvec, bio, i) { |
1dfa0f68 | 1466 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1467 | p += bvec->bv_len; |
68154e90 FT |
1468 | } |
1469 | ||
4246a0b6 | 1470 | bio_copy_kern_endio(bio); |
68154e90 FT |
1471 | } |
1472 | ||
1473 | /** | |
1474 | * bio_copy_kern - copy kernel address into bio | |
1475 | * @q: the struct request_queue for the bio | |
1476 | * @data: pointer to buffer to copy | |
1477 | * @len: length in bytes | |
1478 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1479 | * @reading: data direction is READ |
68154e90 FT |
1480 | * |
1481 | * copy the kernel address into a bio suitable for io to a block | |
1482 | * device. Returns an error pointer in case of error. | |
1483 | */ | |
1484 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1485 | gfp_t gfp_mask, int reading) | |
1486 | { | |
42d2683a CH |
1487 | unsigned long kaddr = (unsigned long)data; |
1488 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1489 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1490 | struct bio *bio; |
1491 | void *p = data; | |
1dfa0f68 | 1492 | int nr_pages = 0; |
68154e90 | 1493 | |
42d2683a CH |
1494 | /* |
1495 | * Overflow, abort | |
1496 | */ | |
1497 | if (end < start) | |
1498 | return ERR_PTR(-EINVAL); | |
68154e90 | 1499 | |
42d2683a CH |
1500 | nr_pages = end - start; |
1501 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1502 | if (!bio) | |
1503 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1504 | |
42d2683a CH |
1505 | while (len) { |
1506 | struct page *page; | |
1507 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1508 | |
42d2683a CH |
1509 | if (bytes > len) |
1510 | bytes = len; | |
1511 | ||
1512 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1513 | if (!page) | |
1514 | goto cleanup; | |
1515 | ||
1516 | if (!reading) | |
1517 | memcpy(page_address(page), p, bytes); | |
1518 | ||
1519 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1520 | break; | |
1521 | ||
1522 | len -= bytes; | |
1523 | p += bytes; | |
68154e90 FT |
1524 | } |
1525 | ||
1dfa0f68 CH |
1526 | if (reading) { |
1527 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1528 | bio->bi_private = data; | |
1529 | } else { | |
1530 | bio->bi_end_io = bio_copy_kern_endio; | |
95fe6c1a | 1531 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
1dfa0f68 | 1532 | } |
76029ff3 | 1533 | |
68154e90 | 1534 | return bio; |
42d2683a CH |
1535 | |
1536 | cleanup: | |
1dfa0f68 | 1537 | bio_free_pages(bio); |
42d2683a CH |
1538 | bio_put(bio); |
1539 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1540 | } |
a112a71d | 1541 | EXPORT_SYMBOL(bio_copy_kern); |
68154e90 | 1542 | |
1da177e4 LT |
1543 | /* |
1544 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1545 | * for performing direct-IO in BIOs. | |
1546 | * | |
1547 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1548 | * because the required locks are not interrupt-safe. So what we can do is to | |
1549 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1550 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1551 | * in process context. | |
1552 | * | |
1553 | * We special-case compound pages here: normally this means reads into hugetlb | |
1554 | * pages. The logic in here doesn't really work right for compound pages | |
1555 | * because the VM does not uniformly chase down the head page in all cases. | |
1556 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1557 | * handle them at all. So we skip compound pages here at an early stage. | |
1558 | * | |
1559 | * Note that this code is very hard to test under normal circumstances because | |
1560 | * direct-io pins the pages with get_user_pages(). This makes | |
1561 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1562 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1563 | * pagecache. |
1564 | * | |
1565 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1566 | * deferred bio dirtying paths. | |
1567 | */ | |
1568 | ||
1569 | /* | |
1570 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1571 | */ | |
1572 | void bio_set_pages_dirty(struct bio *bio) | |
1573 | { | |
cb34e057 | 1574 | struct bio_vec *bvec; |
1da177e4 LT |
1575 | int i; |
1576 | ||
cb34e057 KO |
1577 | bio_for_each_segment_all(bvec, bio, i) { |
1578 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1579 | |
1580 | if (page && !PageCompound(page)) | |
1581 | set_page_dirty_lock(page); | |
1582 | } | |
1583 | } | |
1584 | ||
86b6c7a7 | 1585 | static void bio_release_pages(struct bio *bio) |
1da177e4 | 1586 | { |
cb34e057 | 1587 | struct bio_vec *bvec; |
1da177e4 LT |
1588 | int i; |
1589 | ||
cb34e057 KO |
1590 | bio_for_each_segment_all(bvec, bio, i) { |
1591 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1592 | |
1593 | if (page) | |
1594 | put_page(page); | |
1595 | } | |
1596 | } | |
1597 | ||
1598 | /* | |
1599 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1600 | * If they are, then fine. If, however, some pages are clean then they must | |
1601 | * have been written out during the direct-IO read. So we take another ref on | |
1602 | * the BIO and the offending pages and re-dirty the pages in process context. | |
1603 | * | |
1604 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1605 | * here on. It will run one put_page() against each page and will run one |
1606 | * bio_put() against the BIO. | |
1da177e4 LT |
1607 | */ |
1608 | ||
65f27f38 | 1609 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1610 | |
65f27f38 | 1611 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1612 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1613 | static struct bio *bio_dirty_list; | |
1614 | ||
1615 | /* | |
1616 | * This runs in process context | |
1617 | */ | |
65f27f38 | 1618 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 LT |
1619 | { |
1620 | unsigned long flags; | |
1621 | struct bio *bio; | |
1622 | ||
1623 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1624 | bio = bio_dirty_list; | |
1625 | bio_dirty_list = NULL; | |
1626 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1627 | ||
1628 | while (bio) { | |
1629 | struct bio *next = bio->bi_private; | |
1630 | ||
1631 | bio_set_pages_dirty(bio); | |
1632 | bio_release_pages(bio); | |
1633 | bio_put(bio); | |
1634 | bio = next; | |
1635 | } | |
1636 | } | |
1637 | ||
1638 | void bio_check_pages_dirty(struct bio *bio) | |
1639 | { | |
cb34e057 | 1640 | struct bio_vec *bvec; |
1da177e4 LT |
1641 | int nr_clean_pages = 0; |
1642 | int i; | |
1643 | ||
cb34e057 KO |
1644 | bio_for_each_segment_all(bvec, bio, i) { |
1645 | struct page *page = bvec->bv_page; | |
1da177e4 LT |
1646 | |
1647 | if (PageDirty(page) || PageCompound(page)) { | |
09cbfeaf | 1648 | put_page(page); |
cb34e057 | 1649 | bvec->bv_page = NULL; |
1da177e4 LT |
1650 | } else { |
1651 | nr_clean_pages++; | |
1652 | } | |
1653 | } | |
1654 | ||
1655 | if (nr_clean_pages) { | |
1656 | unsigned long flags; | |
1657 | ||
1658 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1659 | bio->bi_private = bio_dirty_list; | |
1660 | bio_dirty_list = bio; | |
1661 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1662 | schedule_work(&bio_dirty_work); | |
1663 | } else { | |
1664 | bio_put(bio); | |
1665 | } | |
1666 | } | |
1667 | ||
394ffa50 GZ |
1668 | void generic_start_io_acct(int rw, unsigned long sectors, |
1669 | struct hd_struct *part) | |
1670 | { | |
1671 | int cpu = part_stat_lock(); | |
1672 | ||
1673 | part_round_stats(cpu, part); | |
1674 | part_stat_inc(cpu, part, ios[rw]); | |
1675 | part_stat_add(cpu, part, sectors[rw], sectors); | |
1676 | part_inc_in_flight(part, rw); | |
1677 | ||
1678 | part_stat_unlock(); | |
1679 | } | |
1680 | EXPORT_SYMBOL(generic_start_io_acct); | |
1681 | ||
1682 | void generic_end_io_acct(int rw, struct hd_struct *part, | |
1683 | unsigned long start_time) | |
1684 | { | |
1685 | unsigned long duration = jiffies - start_time; | |
1686 | int cpu = part_stat_lock(); | |
1687 | ||
1688 | part_stat_add(cpu, part, ticks[rw], duration); | |
1689 | part_round_stats(cpu, part); | |
1690 | part_dec_in_flight(part, rw); | |
1691 | ||
1692 | part_stat_unlock(); | |
1693 | } | |
1694 | EXPORT_SYMBOL(generic_end_io_acct); | |
1695 | ||
2d4dc890 IL |
1696 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1697 | void bio_flush_dcache_pages(struct bio *bi) | |
1698 | { | |
7988613b KO |
1699 | struct bio_vec bvec; |
1700 | struct bvec_iter iter; | |
2d4dc890 | 1701 | |
7988613b KO |
1702 | bio_for_each_segment(bvec, bi, iter) |
1703 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1704 | } |
1705 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1706 | #endif | |
1707 | ||
c4cf5261 JA |
1708 | static inline bool bio_remaining_done(struct bio *bio) |
1709 | { | |
1710 | /* | |
1711 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1712 | * we always end io on the first invocation. | |
1713 | */ | |
1714 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1715 | return true; | |
1716 | ||
1717 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1718 | ||
326e1dbb | 1719 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1720 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1721 | return true; |
326e1dbb | 1722 | } |
c4cf5261 JA |
1723 | |
1724 | return false; | |
1725 | } | |
1726 | ||
1da177e4 LT |
1727 | /** |
1728 | * bio_endio - end I/O on a bio | |
1729 | * @bio: bio | |
1da177e4 LT |
1730 | * |
1731 | * Description: | |
4246a0b6 CH |
1732 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1733 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1734 | * bio unless they own it and thus know that it has an end_io function. | |
1da177e4 | 1735 | **/ |
4246a0b6 | 1736 | void bio_endio(struct bio *bio) |
1da177e4 | 1737 | { |
ba8c6967 | 1738 | again: |
2b885517 | 1739 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1740 | return; |
1da177e4 | 1741 | |
ba8c6967 CH |
1742 | /* |
1743 | * Need to have a real endio function for chained bios, otherwise | |
1744 | * various corner cases will break (like stacking block devices that | |
1745 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1746 | * recursion and blowing the stack. Tail call optimization would | |
1747 | * handle this, but compiling with frame pointers also disables | |
1748 | * gcc's sibling call optimization. | |
1749 | */ | |
1750 | if (bio->bi_end_io == bio_chain_endio) { | |
1751 | bio = __bio_chain_endio(bio); | |
1752 | goto again; | |
196d38bc | 1753 | } |
ba8c6967 CH |
1754 | |
1755 | if (bio->bi_end_io) | |
1756 | bio->bi_end_io(bio); | |
1da177e4 | 1757 | } |
a112a71d | 1758 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1759 | |
20d0189b KO |
1760 | /** |
1761 | * bio_split - split a bio | |
1762 | * @bio: bio to split | |
1763 | * @sectors: number of sectors to split from the front of @bio | |
1764 | * @gfp: gfp mask | |
1765 | * @bs: bio set to allocate from | |
1766 | * | |
1767 | * Allocates and returns a new bio which represents @sectors from the start of | |
1768 | * @bio, and updates @bio to represent the remaining sectors. | |
1769 | * | |
f3f5da62 MP |
1770 | * Unless this is a discard request the newly allocated bio will point |
1771 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1772 | * @bio is not freed before the split. | |
20d0189b KO |
1773 | */ |
1774 | struct bio *bio_split(struct bio *bio, int sectors, | |
1775 | gfp_t gfp, struct bio_set *bs) | |
1776 | { | |
1777 | struct bio *split = NULL; | |
1778 | ||
1779 | BUG_ON(sectors <= 0); | |
1780 | BUG_ON(sectors >= bio_sectors(bio)); | |
1781 | ||
f3f5da62 MP |
1782 | /* |
1783 | * Discards need a mutable bio_vec to accommodate the payload | |
1784 | * required by the DSM TRIM and UNMAP commands. | |
1785 | */ | |
95fe6c1a | 1786 | if (bio_op(bio) == REQ_OP_DISCARD) |
f3f5da62 MP |
1787 | split = bio_clone_bioset(bio, gfp, bs); |
1788 | else | |
1789 | split = bio_clone_fast(bio, gfp, bs); | |
1790 | ||
20d0189b KO |
1791 | if (!split) |
1792 | return NULL; | |
1793 | ||
1794 | split->bi_iter.bi_size = sectors << 9; | |
1795 | ||
1796 | if (bio_integrity(split)) | |
1797 | bio_integrity_trim(split, 0, sectors); | |
1798 | ||
1799 | bio_advance(bio, split->bi_iter.bi_size); | |
1800 | ||
1801 | return split; | |
1802 | } | |
1803 | EXPORT_SYMBOL(bio_split); | |
1804 | ||
6678d83f KO |
1805 | /** |
1806 | * bio_trim - trim a bio | |
1807 | * @bio: bio to trim | |
1808 | * @offset: number of sectors to trim from the front of @bio | |
1809 | * @size: size we want to trim @bio to, in sectors | |
1810 | */ | |
1811 | void bio_trim(struct bio *bio, int offset, int size) | |
1812 | { | |
1813 | /* 'bio' is a cloned bio which we need to trim to match | |
1814 | * the given offset and size. | |
6678d83f | 1815 | */ |
6678d83f KO |
1816 | |
1817 | size <<= 9; | |
4f024f37 | 1818 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1819 | return; |
1820 | ||
b7c44ed9 | 1821 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1822 | |
1823 | bio_advance(bio, offset << 9); | |
1824 | ||
4f024f37 | 1825 | bio->bi_iter.bi_size = size; |
6678d83f KO |
1826 | } |
1827 | EXPORT_SYMBOL_GPL(bio_trim); | |
1828 | ||
1da177e4 LT |
1829 | /* |
1830 | * create memory pools for biovec's in a bio_set. | |
1831 | * use the global biovec slabs created for general use. | |
1832 | */ | |
a6c39cb4 | 1833 | mempool_t *biovec_create_pool(int pool_entries) |
1da177e4 | 1834 | { |
7ff9345f | 1835 | struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX; |
1da177e4 | 1836 | |
9f060e22 | 1837 | return mempool_create_slab_pool(pool_entries, bp->slab); |
1da177e4 LT |
1838 | } |
1839 | ||
1840 | void bioset_free(struct bio_set *bs) | |
1841 | { | |
df2cb6da KO |
1842 | if (bs->rescue_workqueue) |
1843 | destroy_workqueue(bs->rescue_workqueue); | |
1844 | ||
1da177e4 LT |
1845 | if (bs->bio_pool) |
1846 | mempool_destroy(bs->bio_pool); | |
1847 | ||
9f060e22 KO |
1848 | if (bs->bvec_pool) |
1849 | mempool_destroy(bs->bvec_pool); | |
1850 | ||
7878cba9 | 1851 | bioset_integrity_free(bs); |
bb799ca0 | 1852 | bio_put_slab(bs); |
1da177e4 LT |
1853 | |
1854 | kfree(bs); | |
1855 | } | |
a112a71d | 1856 | EXPORT_SYMBOL(bioset_free); |
1da177e4 | 1857 | |
d8f429e1 JN |
1858 | static struct bio_set *__bioset_create(unsigned int pool_size, |
1859 | unsigned int front_pad, | |
1860 | bool create_bvec_pool) | |
1da177e4 | 1861 | { |
392ddc32 | 1862 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
1b434498 | 1863 | struct bio_set *bs; |
1da177e4 | 1864 | |
1b434498 | 1865 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
1da177e4 LT |
1866 | if (!bs) |
1867 | return NULL; | |
1868 | ||
bb799ca0 | 1869 | bs->front_pad = front_pad; |
1b434498 | 1870 | |
df2cb6da KO |
1871 | spin_lock_init(&bs->rescue_lock); |
1872 | bio_list_init(&bs->rescue_list); | |
1873 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
1874 | ||
392ddc32 | 1875 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
bb799ca0 JA |
1876 | if (!bs->bio_slab) { |
1877 | kfree(bs); | |
1878 | return NULL; | |
1879 | } | |
1880 | ||
1881 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); | |
1da177e4 LT |
1882 | if (!bs->bio_pool) |
1883 | goto bad; | |
1884 | ||
d8f429e1 JN |
1885 | if (create_bvec_pool) { |
1886 | bs->bvec_pool = biovec_create_pool(pool_size); | |
1887 | if (!bs->bvec_pool) | |
1888 | goto bad; | |
1889 | } | |
df2cb6da KO |
1890 | |
1891 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); | |
1892 | if (!bs->rescue_workqueue) | |
1893 | goto bad; | |
1da177e4 | 1894 | |
df2cb6da | 1895 | return bs; |
1da177e4 LT |
1896 | bad: |
1897 | bioset_free(bs); | |
1898 | return NULL; | |
1899 | } | |
d8f429e1 JN |
1900 | |
1901 | /** | |
1902 | * bioset_create - Create a bio_set | |
1903 | * @pool_size: Number of bio and bio_vecs to cache in the mempool | |
1904 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1905 | * | |
1906 | * Description: | |
1907 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1908 | * to ask for a number of bytes to be allocated in front of the bio. | |
1909 | * Front pad allocation is useful for embedding the bio inside | |
1910 | * another structure, to avoid allocating extra data to go with the bio. | |
1911 | * Note that the bio must be embedded at the END of that structure always, | |
1912 | * or things will break badly. | |
1913 | */ | |
1914 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) | |
1915 | { | |
1916 | return __bioset_create(pool_size, front_pad, true); | |
1917 | } | |
a112a71d | 1918 | EXPORT_SYMBOL(bioset_create); |
1da177e4 | 1919 | |
d8f429e1 JN |
1920 | /** |
1921 | * bioset_create_nobvec - Create a bio_set without bio_vec mempool | |
1922 | * @pool_size: Number of bio to cache in the mempool | |
1923 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1924 | * | |
1925 | * Description: | |
1926 | * Same functionality as bioset_create() except that mempool is not | |
1927 | * created for bio_vecs. Saving some memory for bio_clone_fast() users. | |
1928 | */ | |
1929 | struct bio_set *bioset_create_nobvec(unsigned int pool_size, unsigned int front_pad) | |
1930 | { | |
1931 | return __bioset_create(pool_size, front_pad, false); | |
1932 | } | |
1933 | EXPORT_SYMBOL(bioset_create_nobvec); | |
1934 | ||
852c788f | 1935 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 TH |
1936 | |
1937 | /** | |
1938 | * bio_associate_blkcg - associate a bio with the specified blkcg | |
1939 | * @bio: target bio | |
1940 | * @blkcg_css: css of the blkcg to associate | |
1941 | * | |
1942 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will | |
1943 | * treat @bio as if it were issued by a task which belongs to the blkcg. | |
1944 | * | |
1945 | * This function takes an extra reference of @blkcg_css which will be put | |
1946 | * when @bio is released. The caller must own @bio and is responsible for | |
1947 | * synchronizing calls to this function. | |
1948 | */ | |
1949 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) | |
1950 | { | |
1951 | if (unlikely(bio->bi_css)) | |
1952 | return -EBUSY; | |
1953 | css_get(blkcg_css); | |
1954 | bio->bi_css = blkcg_css; | |
1955 | return 0; | |
1956 | } | |
5aa2a96b | 1957 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
1d933cf0 | 1958 | |
852c788f TH |
1959 | /** |
1960 | * bio_associate_current - associate a bio with %current | |
1961 | * @bio: target bio | |
1962 | * | |
1963 | * Associate @bio with %current if it hasn't been associated yet. Block | |
1964 | * layer will treat @bio as if it were issued by %current no matter which | |
1965 | * task actually issues it. | |
1966 | * | |
1967 | * This function takes an extra reference of @task's io_context and blkcg | |
1968 | * which will be put when @bio is released. The caller must own @bio, | |
1969 | * ensure %current->io_context exists, and is responsible for synchronizing | |
1970 | * calls to this function. | |
1971 | */ | |
1972 | int bio_associate_current(struct bio *bio) | |
1973 | { | |
1974 | struct io_context *ioc; | |
852c788f | 1975 | |
1d933cf0 | 1976 | if (bio->bi_css) |
852c788f TH |
1977 | return -EBUSY; |
1978 | ||
1979 | ioc = current->io_context; | |
1980 | if (!ioc) | |
1981 | return -ENOENT; | |
1982 | ||
852c788f TH |
1983 | get_io_context_active(ioc); |
1984 | bio->bi_ioc = ioc; | |
c165b3e3 | 1985 | bio->bi_css = task_get_css(current, io_cgrp_id); |
852c788f TH |
1986 | return 0; |
1987 | } | |
5aa2a96b | 1988 | EXPORT_SYMBOL_GPL(bio_associate_current); |
852c788f TH |
1989 | |
1990 | /** | |
1991 | * bio_disassociate_task - undo bio_associate_current() | |
1992 | * @bio: target bio | |
1993 | */ | |
1994 | void bio_disassociate_task(struct bio *bio) | |
1995 | { | |
1996 | if (bio->bi_ioc) { | |
1997 | put_io_context(bio->bi_ioc); | |
1998 | bio->bi_ioc = NULL; | |
1999 | } | |
2000 | if (bio->bi_css) { | |
2001 | css_put(bio->bi_css); | |
2002 | bio->bi_css = NULL; | |
2003 | } | |
2004 | } | |
2005 | ||
2006 | #endif /* CONFIG_BLK_CGROUP */ | |
2007 | ||
1da177e4 LT |
2008 | static void __init biovec_init_slabs(void) |
2009 | { | |
2010 | int i; | |
2011 | ||
2012 | for (i = 0; i < BIOVEC_NR_POOLS; i++) { | |
2013 | int size; | |
2014 | struct biovec_slab *bvs = bvec_slabs + i; | |
2015 | ||
a7fcd37c JA |
2016 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2017 | bvs->slab = NULL; | |
2018 | continue; | |
2019 | } | |
a7fcd37c | 2020 | |
1da177e4 LT |
2021 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2022 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2023 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2024 | } |
2025 | } | |
2026 | ||
2027 | static int __init init_bio(void) | |
2028 | { | |
bb799ca0 JA |
2029 | bio_slab_max = 2; |
2030 | bio_slab_nr = 0; | |
2031 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); | |
2032 | if (!bio_slabs) | |
2033 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2034 | |
7878cba9 | 2035 | bio_integrity_init(); |
1da177e4 LT |
2036 | biovec_init_slabs(); |
2037 | ||
bb799ca0 | 2038 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
1da177e4 LT |
2039 | if (!fs_bio_set) |
2040 | panic("bio: can't allocate bios\n"); | |
2041 | ||
a91a2785 MP |
2042 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
2043 | panic("bio: can't create integrity pool\n"); | |
2044 | ||
1da177e4 LT |
2045 | return 0; |
2046 | } | |
1da177e4 | 2047 | subsys_initcall(init_bio); |