]>
Commit | Line | Data |
---|---|---|
8c16567d | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
0fe23479 | 3 | * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk> |
1da177e4 LT |
4 | */ |
5 | #include <linux/mm.h> | |
6 | #include <linux/swap.h> | |
7 | #include <linux/bio.h> | |
8 | #include <linux/blkdev.h> | |
a27bb332 | 9 | #include <linux/uio.h> |
852c788f | 10 | #include <linux/iocontext.h> |
1da177e4 LT |
11 | #include <linux/slab.h> |
12 | #include <linux/init.h> | |
13 | #include <linux/kernel.h> | |
630d9c47 | 14 | #include <linux/export.h> |
1da177e4 LT |
15 | #include <linux/mempool.h> |
16 | #include <linux/workqueue.h> | |
852c788f | 17 | #include <linux/cgroup.h> |
08e18eab | 18 | #include <linux/blk-cgroup.h> |
1da177e4 | 19 | |
55782138 | 20 | #include <trace/events/block.h> |
9e234eea | 21 | #include "blk.h" |
67b42d0b | 22 | #include "blk-rq-qos.h" |
0bfc2455 | 23 | |
392ddc32 JA |
24 | /* |
25 | * Test patch to inline a certain number of bi_io_vec's inside the bio | |
26 | * itself, to shrink a bio data allocation from two mempool calls to one | |
27 | */ | |
28 | #define BIO_INLINE_VECS 4 | |
29 | ||
1da177e4 LT |
30 | /* |
31 | * if you change this list, also change bvec_alloc or things will | |
32 | * break badly! cannot be bigger than what you can fit into an | |
33 | * unsigned short | |
34 | */ | |
bd5c4fac | 35 | #define BV(x, n) { .nr_vecs = x, .name = "biovec-"#n } |
ed996a52 | 36 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
bd5c4fac | 37 | BV(1, 1), BV(4, 4), BV(16, 16), BV(64, 64), BV(128, 128), BV(BIO_MAX_PAGES, max), |
1da177e4 LT |
38 | }; |
39 | #undef BV | |
40 | ||
1da177e4 LT |
41 | /* |
42 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by | |
43 | * IO code that does not need private memory pools. | |
44 | */ | |
f4f8154a | 45 | struct bio_set fs_bio_set; |
3f86a82a | 46 | EXPORT_SYMBOL(fs_bio_set); |
1da177e4 | 47 | |
bb799ca0 JA |
48 | /* |
49 | * Our slab pool management | |
50 | */ | |
51 | struct bio_slab { | |
52 | struct kmem_cache *slab; | |
53 | unsigned int slab_ref; | |
54 | unsigned int slab_size; | |
55 | char name[8]; | |
56 | }; | |
57 | static DEFINE_MUTEX(bio_slab_lock); | |
58 | static struct bio_slab *bio_slabs; | |
59 | static unsigned int bio_slab_nr, bio_slab_max; | |
60 | ||
61 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) | |
62 | { | |
63 | unsigned int sz = sizeof(struct bio) + extra_size; | |
64 | struct kmem_cache *slab = NULL; | |
389d7b26 | 65 | struct bio_slab *bslab, *new_bio_slabs; |
386bc35a | 66 | unsigned int new_bio_slab_max; |
bb799ca0 JA |
67 | unsigned int i, entry = -1; |
68 | ||
69 | mutex_lock(&bio_slab_lock); | |
70 | ||
71 | i = 0; | |
72 | while (i < bio_slab_nr) { | |
f06f135d | 73 | bslab = &bio_slabs[i]; |
bb799ca0 JA |
74 | |
75 | if (!bslab->slab && entry == -1) | |
76 | entry = i; | |
77 | else if (bslab->slab_size == sz) { | |
78 | slab = bslab->slab; | |
79 | bslab->slab_ref++; | |
80 | break; | |
81 | } | |
82 | i++; | |
83 | } | |
84 | ||
85 | if (slab) | |
86 | goto out_unlock; | |
87 | ||
88 | if (bio_slab_nr == bio_slab_max && entry == -1) { | |
386bc35a | 89 | new_bio_slab_max = bio_slab_max << 1; |
389d7b26 | 90 | new_bio_slabs = krealloc(bio_slabs, |
386bc35a | 91 | new_bio_slab_max * sizeof(struct bio_slab), |
389d7b26 AK |
92 | GFP_KERNEL); |
93 | if (!new_bio_slabs) | |
bb799ca0 | 94 | goto out_unlock; |
386bc35a | 95 | bio_slab_max = new_bio_slab_max; |
389d7b26 | 96 | bio_slabs = new_bio_slabs; |
bb799ca0 JA |
97 | } |
98 | if (entry == -1) | |
99 | entry = bio_slab_nr++; | |
100 | ||
101 | bslab = &bio_slabs[entry]; | |
102 | ||
103 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); | |
6a241483 MP |
104 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
105 | SLAB_HWCACHE_ALIGN, NULL); | |
bb799ca0 JA |
106 | if (!slab) |
107 | goto out_unlock; | |
108 | ||
bb799ca0 JA |
109 | bslab->slab = slab; |
110 | bslab->slab_ref = 1; | |
111 | bslab->slab_size = sz; | |
112 | out_unlock: | |
113 | mutex_unlock(&bio_slab_lock); | |
114 | return slab; | |
115 | } | |
116 | ||
117 | static void bio_put_slab(struct bio_set *bs) | |
118 | { | |
119 | struct bio_slab *bslab = NULL; | |
120 | unsigned int i; | |
121 | ||
122 | mutex_lock(&bio_slab_lock); | |
123 | ||
124 | for (i = 0; i < bio_slab_nr; i++) { | |
125 | if (bs->bio_slab == bio_slabs[i].slab) { | |
126 | bslab = &bio_slabs[i]; | |
127 | break; | |
128 | } | |
129 | } | |
130 | ||
131 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) | |
132 | goto out; | |
133 | ||
134 | WARN_ON(!bslab->slab_ref); | |
135 | ||
136 | if (--bslab->slab_ref) | |
137 | goto out; | |
138 | ||
139 | kmem_cache_destroy(bslab->slab); | |
140 | bslab->slab = NULL; | |
141 | ||
142 | out: | |
143 | mutex_unlock(&bio_slab_lock); | |
144 | } | |
145 | ||
7ba1ba12 MP |
146 | unsigned int bvec_nr_vecs(unsigned short idx) |
147 | { | |
d6c02a9b | 148 | return bvec_slabs[--idx].nr_vecs; |
7ba1ba12 MP |
149 | } |
150 | ||
9f060e22 | 151 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
bb799ca0 | 152 | { |
ed996a52 CH |
153 | if (!idx) |
154 | return; | |
155 | idx--; | |
156 | ||
157 | BIO_BUG_ON(idx >= BVEC_POOL_NR); | |
bb799ca0 | 158 | |
ed996a52 | 159 | if (idx == BVEC_POOL_MAX) { |
9f060e22 | 160 | mempool_free(bv, pool); |
ed996a52 | 161 | } else { |
bb799ca0 JA |
162 | struct biovec_slab *bvs = bvec_slabs + idx; |
163 | ||
164 | kmem_cache_free(bvs->slab, bv); | |
165 | } | |
166 | } | |
167 | ||
9f060e22 KO |
168 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
169 | mempool_t *pool) | |
1da177e4 LT |
170 | { |
171 | struct bio_vec *bvl; | |
1da177e4 | 172 | |
7ff9345f JA |
173 | /* |
174 | * see comment near bvec_array define! | |
175 | */ | |
176 | switch (nr) { | |
177 | case 1: | |
178 | *idx = 0; | |
179 | break; | |
180 | case 2 ... 4: | |
181 | *idx = 1; | |
182 | break; | |
183 | case 5 ... 16: | |
184 | *idx = 2; | |
185 | break; | |
186 | case 17 ... 64: | |
187 | *idx = 3; | |
188 | break; | |
189 | case 65 ... 128: | |
190 | *idx = 4; | |
191 | break; | |
192 | case 129 ... BIO_MAX_PAGES: | |
193 | *idx = 5; | |
194 | break; | |
195 | default: | |
196 | return NULL; | |
197 | } | |
198 | ||
199 | /* | |
200 | * idx now points to the pool we want to allocate from. only the | |
201 | * 1-vec entry pool is mempool backed. | |
202 | */ | |
ed996a52 | 203 | if (*idx == BVEC_POOL_MAX) { |
7ff9345f | 204 | fallback: |
9f060e22 | 205 | bvl = mempool_alloc(pool, gfp_mask); |
7ff9345f JA |
206 | } else { |
207 | struct biovec_slab *bvs = bvec_slabs + *idx; | |
d0164adc | 208 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
7ff9345f | 209 | |
0a0d96b0 | 210 | /* |
7ff9345f JA |
211 | * Make this allocation restricted and don't dump info on |
212 | * allocation failures, since we'll fallback to the mempool | |
213 | * in case of failure. | |
0a0d96b0 | 214 | */ |
7ff9345f | 215 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
1da177e4 | 216 | |
0a0d96b0 | 217 | /* |
d0164adc | 218 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
7ff9345f | 219 | * is set, retry with the 1-entry mempool |
0a0d96b0 | 220 | */ |
7ff9345f | 221 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
d0164adc | 222 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
ed996a52 | 223 | *idx = BVEC_POOL_MAX; |
7ff9345f JA |
224 | goto fallback; |
225 | } | |
226 | } | |
227 | ||
ed996a52 | 228 | (*idx)++; |
1da177e4 LT |
229 | return bvl; |
230 | } | |
231 | ||
9ae3b3f5 | 232 | void bio_uninit(struct bio *bio) |
1da177e4 | 233 | { |
6f70fb66 | 234 | bio_disassociate_blkg(bio); |
4254bba1 | 235 | } |
9ae3b3f5 | 236 | EXPORT_SYMBOL(bio_uninit); |
7ba1ba12 | 237 | |
4254bba1 KO |
238 | static void bio_free(struct bio *bio) |
239 | { | |
240 | struct bio_set *bs = bio->bi_pool; | |
241 | void *p; | |
242 | ||
9ae3b3f5 | 243 | bio_uninit(bio); |
4254bba1 KO |
244 | |
245 | if (bs) { | |
8aa6ba2f | 246 | bvec_free(&bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio)); |
4254bba1 KO |
247 | |
248 | /* | |
249 | * If we have front padding, adjust the bio pointer before freeing | |
250 | */ | |
251 | p = bio; | |
bb799ca0 JA |
252 | p -= bs->front_pad; |
253 | ||
8aa6ba2f | 254 | mempool_free(p, &bs->bio_pool); |
4254bba1 KO |
255 | } else { |
256 | /* Bio was allocated by bio_kmalloc() */ | |
257 | kfree(bio); | |
258 | } | |
3676347a PO |
259 | } |
260 | ||
9ae3b3f5 JA |
261 | /* |
262 | * Users of this function have their own bio allocation. Subsequently, | |
263 | * they must remember to pair any call to bio_init() with bio_uninit() | |
264 | * when IO has completed, or when the bio is released. | |
265 | */ | |
3a83f467 ML |
266 | void bio_init(struct bio *bio, struct bio_vec *table, |
267 | unsigned short max_vecs) | |
1da177e4 | 268 | { |
2b94de55 | 269 | memset(bio, 0, sizeof(*bio)); |
c4cf5261 | 270 | atomic_set(&bio->__bi_remaining, 1); |
dac56212 | 271 | atomic_set(&bio->__bi_cnt, 1); |
3a83f467 ML |
272 | |
273 | bio->bi_io_vec = table; | |
274 | bio->bi_max_vecs = max_vecs; | |
1da177e4 | 275 | } |
a112a71d | 276 | EXPORT_SYMBOL(bio_init); |
1da177e4 | 277 | |
f44b48c7 KO |
278 | /** |
279 | * bio_reset - reinitialize a bio | |
280 | * @bio: bio to reset | |
281 | * | |
282 | * Description: | |
283 | * After calling bio_reset(), @bio will be in the same state as a freshly | |
284 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are | |
285 | * preserved are the ones that are initialized by bio_alloc_bioset(). See | |
286 | * comment in struct bio. | |
287 | */ | |
288 | void bio_reset(struct bio *bio) | |
289 | { | |
290 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); | |
291 | ||
9ae3b3f5 | 292 | bio_uninit(bio); |
f44b48c7 KO |
293 | |
294 | memset(bio, 0, BIO_RESET_BYTES); | |
4246a0b6 | 295 | bio->bi_flags = flags; |
c4cf5261 | 296 | atomic_set(&bio->__bi_remaining, 1); |
f44b48c7 KO |
297 | } |
298 | EXPORT_SYMBOL(bio_reset); | |
299 | ||
38f8baae | 300 | static struct bio *__bio_chain_endio(struct bio *bio) |
196d38bc | 301 | { |
4246a0b6 CH |
302 | struct bio *parent = bio->bi_private; |
303 | ||
4e4cbee9 CH |
304 | if (!parent->bi_status) |
305 | parent->bi_status = bio->bi_status; | |
196d38bc | 306 | bio_put(bio); |
38f8baae CH |
307 | return parent; |
308 | } | |
309 | ||
310 | static void bio_chain_endio(struct bio *bio) | |
311 | { | |
312 | bio_endio(__bio_chain_endio(bio)); | |
196d38bc KO |
313 | } |
314 | ||
315 | /** | |
316 | * bio_chain - chain bio completions | |
1051a902 RD |
317 | * @bio: the target bio |
318 | * @parent: the @bio's parent bio | |
196d38bc KO |
319 | * |
320 | * The caller won't have a bi_end_io called when @bio completes - instead, | |
321 | * @parent's bi_end_io won't be called until both @parent and @bio have | |
322 | * completed; the chained bio will also be freed when it completes. | |
323 | * | |
324 | * The caller must not set bi_private or bi_end_io in @bio. | |
325 | */ | |
326 | void bio_chain(struct bio *bio, struct bio *parent) | |
327 | { | |
328 | BUG_ON(bio->bi_private || bio->bi_end_io); | |
329 | ||
330 | bio->bi_private = parent; | |
331 | bio->bi_end_io = bio_chain_endio; | |
c4cf5261 | 332 | bio_inc_remaining(parent); |
196d38bc KO |
333 | } |
334 | EXPORT_SYMBOL(bio_chain); | |
335 | ||
df2cb6da KO |
336 | static void bio_alloc_rescue(struct work_struct *work) |
337 | { | |
338 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); | |
339 | struct bio *bio; | |
340 | ||
341 | while (1) { | |
342 | spin_lock(&bs->rescue_lock); | |
343 | bio = bio_list_pop(&bs->rescue_list); | |
344 | spin_unlock(&bs->rescue_lock); | |
345 | ||
346 | if (!bio) | |
347 | break; | |
348 | ||
349 | generic_make_request(bio); | |
350 | } | |
351 | } | |
352 | ||
353 | static void punt_bios_to_rescuer(struct bio_set *bs) | |
354 | { | |
355 | struct bio_list punt, nopunt; | |
356 | struct bio *bio; | |
357 | ||
47e0fb46 N |
358 | if (WARN_ON_ONCE(!bs->rescue_workqueue)) |
359 | return; | |
df2cb6da KO |
360 | /* |
361 | * In order to guarantee forward progress we must punt only bios that | |
362 | * were allocated from this bio_set; otherwise, if there was a bio on | |
363 | * there for a stacking driver higher up in the stack, processing it | |
364 | * could require allocating bios from this bio_set, and doing that from | |
365 | * our own rescuer would be bad. | |
366 | * | |
367 | * Since bio lists are singly linked, pop them all instead of trying to | |
368 | * remove from the middle of the list: | |
369 | */ | |
370 | ||
371 | bio_list_init(&punt); | |
372 | bio_list_init(&nopunt); | |
373 | ||
f5fe1b51 | 374 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
df2cb6da | 375 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
f5fe1b51 | 376 | current->bio_list[0] = nopunt; |
df2cb6da | 377 | |
f5fe1b51 N |
378 | bio_list_init(&nopunt); |
379 | while ((bio = bio_list_pop(¤t->bio_list[1]))) | |
380 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); | |
381 | current->bio_list[1] = nopunt; | |
df2cb6da KO |
382 | |
383 | spin_lock(&bs->rescue_lock); | |
384 | bio_list_merge(&bs->rescue_list, &punt); | |
385 | spin_unlock(&bs->rescue_lock); | |
386 | ||
387 | queue_work(bs->rescue_workqueue, &bs->rescue_work); | |
388 | } | |
389 | ||
1da177e4 LT |
390 | /** |
391 | * bio_alloc_bioset - allocate a bio for I/O | |
519c8e9f | 392 | * @gfp_mask: the GFP_* mask given to the slab allocator |
1da177e4 | 393 | * @nr_iovecs: number of iovecs to pre-allocate |
db18efac | 394 | * @bs: the bio_set to allocate from. |
1da177e4 LT |
395 | * |
396 | * Description: | |
3f86a82a KO |
397 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
398 | * backed by the @bs's mempool. | |
399 | * | |
d0164adc MG |
400 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
401 | * always be able to allocate a bio. This is due to the mempool guarantees. | |
402 | * To make this work, callers must never allocate more than 1 bio at a time | |
403 | * from this pool. Callers that need to allocate more than 1 bio must always | |
404 | * submit the previously allocated bio for IO before attempting to allocate | |
405 | * a new one. Failure to do so can cause deadlocks under memory pressure. | |
3f86a82a | 406 | * |
df2cb6da KO |
407 | * Note that when running under generic_make_request() (i.e. any block |
408 | * driver), bios are not submitted until after you return - see the code in | |
409 | * generic_make_request() that converts recursion into iteration, to prevent | |
410 | * stack overflows. | |
411 | * | |
412 | * This would normally mean allocating multiple bios under | |
413 | * generic_make_request() would be susceptible to deadlocks, but we have | |
414 | * deadlock avoidance code that resubmits any blocked bios from a rescuer | |
415 | * thread. | |
416 | * | |
417 | * However, we do not guarantee forward progress for allocations from other | |
418 | * mempools. Doing multiple allocations from the same mempool under | |
419 | * generic_make_request() should be avoided - instead, use bio_set's front_pad | |
420 | * for per bio allocations. | |
421 | * | |
3f86a82a KO |
422 | * RETURNS: |
423 | * Pointer to new bio on success, NULL on failure. | |
424 | */ | |
7a88fa19 DC |
425 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, unsigned int nr_iovecs, |
426 | struct bio_set *bs) | |
1da177e4 | 427 | { |
df2cb6da | 428 | gfp_t saved_gfp = gfp_mask; |
3f86a82a KO |
429 | unsigned front_pad; |
430 | unsigned inline_vecs; | |
34053979 | 431 | struct bio_vec *bvl = NULL; |
451a9ebf TH |
432 | struct bio *bio; |
433 | void *p; | |
434 | ||
3f86a82a KO |
435 | if (!bs) { |
436 | if (nr_iovecs > UIO_MAXIOV) | |
437 | return NULL; | |
438 | ||
439 | p = kmalloc(sizeof(struct bio) + | |
440 | nr_iovecs * sizeof(struct bio_vec), | |
441 | gfp_mask); | |
442 | front_pad = 0; | |
443 | inline_vecs = nr_iovecs; | |
444 | } else { | |
d8f429e1 | 445 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
8aa6ba2f KO |
446 | if (WARN_ON_ONCE(!mempool_initialized(&bs->bvec_pool) && |
447 | nr_iovecs > 0)) | |
d8f429e1 | 448 | return NULL; |
df2cb6da KO |
449 | /* |
450 | * generic_make_request() converts recursion to iteration; this | |
451 | * means if we're running beneath it, any bios we allocate and | |
452 | * submit will not be submitted (and thus freed) until after we | |
453 | * return. | |
454 | * | |
455 | * This exposes us to a potential deadlock if we allocate | |
456 | * multiple bios from the same bio_set() while running | |
457 | * underneath generic_make_request(). If we were to allocate | |
458 | * multiple bios (say a stacking block driver that was splitting | |
459 | * bios), we would deadlock if we exhausted the mempool's | |
460 | * reserve. | |
461 | * | |
462 | * We solve this, and guarantee forward progress, with a rescuer | |
463 | * workqueue per bio_set. If we go to allocate and there are | |
464 | * bios on current->bio_list, we first try the allocation | |
d0164adc MG |
465 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
466 | * bios we would be blocking to the rescuer workqueue before | |
467 | * we retry with the original gfp_flags. | |
df2cb6da KO |
468 | */ |
469 | ||
f5fe1b51 N |
470 | if (current->bio_list && |
471 | (!bio_list_empty(¤t->bio_list[0]) || | |
47e0fb46 N |
472 | !bio_list_empty(¤t->bio_list[1])) && |
473 | bs->rescue_workqueue) | |
d0164adc | 474 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
df2cb6da | 475 | |
8aa6ba2f | 476 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
df2cb6da KO |
477 | if (!p && gfp_mask != saved_gfp) { |
478 | punt_bios_to_rescuer(bs); | |
479 | gfp_mask = saved_gfp; | |
8aa6ba2f | 480 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
df2cb6da KO |
481 | } |
482 | ||
3f86a82a KO |
483 | front_pad = bs->front_pad; |
484 | inline_vecs = BIO_INLINE_VECS; | |
485 | } | |
486 | ||
451a9ebf TH |
487 | if (unlikely(!p)) |
488 | return NULL; | |
1da177e4 | 489 | |
3f86a82a | 490 | bio = p + front_pad; |
3a83f467 | 491 | bio_init(bio, NULL, 0); |
34053979 | 492 | |
3f86a82a | 493 | if (nr_iovecs > inline_vecs) { |
ed996a52 CH |
494 | unsigned long idx = 0; |
495 | ||
8aa6ba2f | 496 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, &bs->bvec_pool); |
df2cb6da KO |
497 | if (!bvl && gfp_mask != saved_gfp) { |
498 | punt_bios_to_rescuer(bs); | |
499 | gfp_mask = saved_gfp; | |
8aa6ba2f | 500 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, &bs->bvec_pool); |
df2cb6da KO |
501 | } |
502 | ||
34053979 IM |
503 | if (unlikely(!bvl)) |
504 | goto err_free; | |
a38352e0 | 505 | |
ed996a52 | 506 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
3f86a82a KO |
507 | } else if (nr_iovecs) { |
508 | bvl = bio->bi_inline_vecs; | |
1da177e4 | 509 | } |
3f86a82a KO |
510 | |
511 | bio->bi_pool = bs; | |
34053979 | 512 | bio->bi_max_vecs = nr_iovecs; |
34053979 | 513 | bio->bi_io_vec = bvl; |
1da177e4 | 514 | return bio; |
34053979 IM |
515 | |
516 | err_free: | |
8aa6ba2f | 517 | mempool_free(p, &bs->bio_pool); |
34053979 | 518 | return NULL; |
1da177e4 | 519 | } |
a112a71d | 520 | EXPORT_SYMBOL(bio_alloc_bioset); |
1da177e4 | 521 | |
38a72dac | 522 | void zero_fill_bio_iter(struct bio *bio, struct bvec_iter start) |
1da177e4 LT |
523 | { |
524 | unsigned long flags; | |
7988613b KO |
525 | struct bio_vec bv; |
526 | struct bvec_iter iter; | |
1da177e4 | 527 | |
38a72dac | 528 | __bio_for_each_segment(bv, bio, iter, start) { |
7988613b KO |
529 | char *data = bvec_kmap_irq(&bv, &flags); |
530 | memset(data, 0, bv.bv_len); | |
531 | flush_dcache_page(bv.bv_page); | |
1da177e4 LT |
532 | bvec_kunmap_irq(data, &flags); |
533 | } | |
534 | } | |
38a72dac | 535 | EXPORT_SYMBOL(zero_fill_bio_iter); |
1da177e4 LT |
536 | |
537 | /** | |
538 | * bio_put - release a reference to a bio | |
539 | * @bio: bio to release reference to | |
540 | * | |
541 | * Description: | |
542 | * Put a reference to a &struct bio, either one you have gotten with | |
9b10f6a9 | 543 | * bio_alloc, bio_get or bio_clone_*. The last put of a bio will free it. |
1da177e4 LT |
544 | **/ |
545 | void bio_put(struct bio *bio) | |
546 | { | |
dac56212 | 547 | if (!bio_flagged(bio, BIO_REFFED)) |
4254bba1 | 548 | bio_free(bio); |
dac56212 JA |
549 | else { |
550 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); | |
551 | ||
552 | /* | |
553 | * last put frees it | |
554 | */ | |
555 | if (atomic_dec_and_test(&bio->__bi_cnt)) | |
556 | bio_free(bio); | |
557 | } | |
1da177e4 | 558 | } |
a112a71d | 559 | EXPORT_SYMBOL(bio_put); |
1da177e4 | 560 | |
6c210aa5 | 561 | int bio_phys_segments(struct request_queue *q, struct bio *bio) |
1da177e4 LT |
562 | { |
563 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) | |
564 | blk_recount_segments(q, bio); | |
565 | ||
566 | return bio->bi_phys_segments; | |
567 | } | |
568 | ||
59d276fe KO |
569 | /** |
570 | * __bio_clone_fast - clone a bio that shares the original bio's biovec | |
571 | * @bio: destination bio | |
572 | * @bio_src: bio to clone | |
573 | * | |
574 | * Clone a &bio. Caller will own the returned bio, but not | |
575 | * the actual data it points to. Reference count of returned | |
576 | * bio will be one. | |
577 | * | |
578 | * Caller must ensure that @bio_src is not freed before @bio. | |
579 | */ | |
580 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) | |
581 | { | |
ed996a52 | 582 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
59d276fe KO |
583 | |
584 | /* | |
74d46992 | 585 | * most users will be overriding ->bi_disk with a new target, |
59d276fe KO |
586 | * so we don't set nor calculate new physical/hw segment counts here |
587 | */ | |
74d46992 | 588 | bio->bi_disk = bio_src->bi_disk; |
62530ed8 | 589 | bio->bi_partno = bio_src->bi_partno; |
b7c44ed9 | 590 | bio_set_flag(bio, BIO_CLONED); |
111be883 SL |
591 | if (bio_flagged(bio_src, BIO_THROTTLED)) |
592 | bio_set_flag(bio, BIO_THROTTLED); | |
1eff9d32 | 593 | bio->bi_opf = bio_src->bi_opf; |
ca474b73 | 594 | bio->bi_ioprio = bio_src->bi_ioprio; |
cb6934f8 | 595 | bio->bi_write_hint = bio_src->bi_write_hint; |
59d276fe KO |
596 | bio->bi_iter = bio_src->bi_iter; |
597 | bio->bi_io_vec = bio_src->bi_io_vec; | |
20bd723e | 598 | |
db6638d7 | 599 | bio_clone_blkg_association(bio, bio_src); |
e439bedf | 600 | blkcg_bio_issue_init(bio); |
59d276fe KO |
601 | } |
602 | EXPORT_SYMBOL(__bio_clone_fast); | |
603 | ||
604 | /** | |
605 | * bio_clone_fast - clone a bio that shares the original bio's biovec | |
606 | * @bio: bio to clone | |
607 | * @gfp_mask: allocation priority | |
608 | * @bs: bio_set to allocate from | |
609 | * | |
610 | * Like __bio_clone_fast, only also allocates the returned bio | |
611 | */ | |
612 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) | |
613 | { | |
614 | struct bio *b; | |
615 | ||
616 | b = bio_alloc_bioset(gfp_mask, 0, bs); | |
617 | if (!b) | |
618 | return NULL; | |
619 | ||
620 | __bio_clone_fast(b, bio); | |
621 | ||
622 | if (bio_integrity(bio)) { | |
623 | int ret; | |
624 | ||
625 | ret = bio_integrity_clone(b, bio, gfp_mask); | |
626 | ||
627 | if (ret < 0) { | |
628 | bio_put(b); | |
629 | return NULL; | |
630 | } | |
631 | } | |
632 | ||
633 | return b; | |
634 | } | |
635 | EXPORT_SYMBOL(bio_clone_fast); | |
636 | ||
5919482e ML |
637 | static inline bool page_is_mergeable(const struct bio_vec *bv, |
638 | struct page *page, unsigned int len, unsigned int off, | |
ff896738 | 639 | bool *same_page) |
5919482e ML |
640 | { |
641 | phys_addr_t vec_end_addr = page_to_phys(bv->bv_page) + | |
642 | bv->bv_offset + bv->bv_len - 1; | |
643 | phys_addr_t page_addr = page_to_phys(page); | |
644 | ||
645 | if (vec_end_addr + 1 != page_addr + off) | |
646 | return false; | |
647 | if (xen_domain() && !xen_biovec_phys_mergeable(bv, page)) | |
648 | return false; | |
52d52d1c | 649 | |
ff896738 CH |
650 | *same_page = ((vec_end_addr & PAGE_MASK) == page_addr); |
651 | if (!*same_page && pfn_to_page(PFN_DOWN(vec_end_addr)) + 1 != page) | |
652 | return false; | |
5919482e ML |
653 | return true; |
654 | } | |
655 | ||
489fbbcb ML |
656 | /* |
657 | * Check if the @page can be added to the current segment(@bv), and make | |
658 | * sure to call it only if page_is_mergeable(@bv, @page) is true | |
659 | */ | |
660 | static bool can_add_page_to_seg(struct request_queue *q, | |
661 | struct bio_vec *bv, struct page *page, unsigned len, | |
662 | unsigned offset) | |
663 | { | |
664 | unsigned long mask = queue_segment_boundary(q); | |
665 | phys_addr_t addr1 = page_to_phys(bv->bv_page) + bv->bv_offset; | |
666 | phys_addr_t addr2 = page_to_phys(page) + offset + len - 1; | |
667 | ||
668 | if ((addr1 | mask) != (addr2 | mask)) | |
669 | return false; | |
670 | ||
671 | if (bv->bv_len + len > queue_max_segment_size(q)) | |
672 | return false; | |
673 | ||
674 | return true; | |
675 | } | |
676 | ||
1da177e4 | 677 | /** |
19047087 | 678 | * __bio_add_pc_page - attempt to add page to passthrough bio |
c66a14d0 KO |
679 | * @q: the target queue |
680 | * @bio: destination bio | |
681 | * @page: page to add | |
682 | * @len: vec entry length | |
683 | * @offset: vec entry offset | |
19047087 | 684 | * @put_same_page: put the page if it is same with last added page |
1da177e4 | 685 | * |
c66a14d0 KO |
686 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
687 | * number of reasons, such as the bio being full or target block device | |
688 | * limitations. The target block device must allow bio's up to PAGE_SIZE, | |
689 | * so it is always possible to add a single page to an empty bio. | |
690 | * | |
5a8ce240 | 691 | * This should only be used by passthrough bios. |
1da177e4 | 692 | */ |
4713839d | 693 | static int __bio_add_pc_page(struct request_queue *q, struct bio *bio, |
19047087 ML |
694 | struct page *page, unsigned int len, unsigned int offset, |
695 | bool put_same_page) | |
1da177e4 | 696 | { |
1da177e4 | 697 | struct bio_vec *bvec; |
ff896738 | 698 | bool same_page = false; |
1da177e4 LT |
699 | |
700 | /* | |
701 | * cloned bio must not modify vec list | |
702 | */ | |
703 | if (unlikely(bio_flagged(bio, BIO_CLONED))) | |
704 | return 0; | |
705 | ||
c66a14d0 | 706 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
1da177e4 LT |
707 | return 0; |
708 | ||
80cfd548 | 709 | if (bio->bi_vcnt > 0) { |
5a8ce240 | 710 | bvec = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
80cfd548 | 711 | |
5a8ce240 ML |
712 | if (page == bvec->bv_page && |
713 | offset == bvec->bv_offset + bvec->bv_len) { | |
19047087 ML |
714 | if (put_same_page) |
715 | put_page(page); | |
5a8ce240 | 716 | bvec->bv_len += len; |
80cfd548 JA |
717 | goto done; |
718 | } | |
66cb45aa JA |
719 | |
720 | /* | |
721 | * If the queue doesn't support SG gaps and adding this | |
722 | * offset would create a gap, disallow it. | |
723 | */ | |
5a8ce240 | 724 | if (bvec_gap_to_prev(q, bvec, offset)) |
66cb45aa | 725 | return 0; |
489fbbcb | 726 | |
ff896738 | 727 | if (page_is_mergeable(bvec, page, len, offset, &same_page) && |
dcdca753 CH |
728 | can_add_page_to_seg(q, bvec, page, len, offset)) { |
729 | bvec->bv_len += len; | |
730 | goto done; | |
731 | } | |
80cfd548 JA |
732 | } |
733 | ||
0aa69fd3 | 734 | if (bio_full(bio)) |
1da177e4 LT |
735 | return 0; |
736 | ||
489fbbcb ML |
737 | if (bio->bi_phys_segments >= queue_max_segments(q)) |
738 | return 0; | |
739 | ||
fcbf6a08 ML |
740 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; |
741 | bvec->bv_page = page; | |
742 | bvec->bv_len = len; | |
743 | bvec->bv_offset = offset; | |
744 | bio->bi_vcnt++; | |
80cfd548 | 745 | done: |
dcdca753 | 746 | bio->bi_iter.bi_size += len; |
489fbbcb ML |
747 | bio->bi_phys_segments = bio->bi_vcnt; |
748 | bio_set_flag(bio, BIO_SEG_VALID); | |
1da177e4 LT |
749 | return len; |
750 | } | |
19047087 ML |
751 | |
752 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, | |
753 | struct page *page, unsigned int len, unsigned int offset) | |
754 | { | |
755 | return __bio_add_pc_page(q, bio, page, len, offset, false); | |
756 | } | |
a112a71d | 757 | EXPORT_SYMBOL(bio_add_pc_page); |
6e68af66 | 758 | |
1da177e4 | 759 | /** |
0aa69fd3 CH |
760 | * __bio_try_merge_page - try appending data to an existing bvec. |
761 | * @bio: destination bio | |
551879a4 | 762 | * @page: start page to add |
0aa69fd3 | 763 | * @len: length of the data to add |
551879a4 | 764 | * @off: offset of the data relative to @page |
ff896738 | 765 | * @same_page: return if the segment has been merged inside the same page |
1da177e4 | 766 | * |
0aa69fd3 CH |
767 | * Try to add the data at @page + @off to the last bvec of @bio. This is a |
768 | * a useful optimisation for file systems with a block size smaller than the | |
769 | * page size. | |
770 | * | |
551879a4 ML |
771 | * Warn if (@len, @off) crosses pages in case that @same_page is true. |
772 | * | |
0aa69fd3 | 773 | * Return %true on success or %false on failure. |
1da177e4 | 774 | */ |
0aa69fd3 | 775 | bool __bio_try_merge_page(struct bio *bio, struct page *page, |
ff896738 | 776 | unsigned int len, unsigned int off, bool *same_page) |
1da177e4 | 777 | { |
c66a14d0 | 778 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
0aa69fd3 | 779 | return false; |
762380ad | 780 | |
c66a14d0 | 781 | if (bio->bi_vcnt > 0) { |
0aa69fd3 | 782 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
5919482e ML |
783 | |
784 | if (page_is_mergeable(bv, page, len, off, same_page)) { | |
785 | bv->bv_len += len; | |
786 | bio->bi_iter.bi_size += len; | |
787 | return true; | |
788 | } | |
c66a14d0 | 789 | } |
0aa69fd3 CH |
790 | return false; |
791 | } | |
792 | EXPORT_SYMBOL_GPL(__bio_try_merge_page); | |
c66a14d0 | 793 | |
0aa69fd3 | 794 | /** |
551879a4 | 795 | * __bio_add_page - add page(s) to a bio in a new segment |
0aa69fd3 | 796 | * @bio: destination bio |
551879a4 ML |
797 | * @page: start page to add |
798 | * @len: length of the data to add, may cross pages | |
799 | * @off: offset of the data relative to @page, may cross pages | |
0aa69fd3 CH |
800 | * |
801 | * Add the data at @page + @off to @bio as a new bvec. The caller must ensure | |
802 | * that @bio has space for another bvec. | |
803 | */ | |
804 | void __bio_add_page(struct bio *bio, struct page *page, | |
805 | unsigned int len, unsigned int off) | |
806 | { | |
807 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt]; | |
c66a14d0 | 808 | |
0aa69fd3 CH |
809 | WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)); |
810 | WARN_ON_ONCE(bio_full(bio)); | |
811 | ||
812 | bv->bv_page = page; | |
813 | bv->bv_offset = off; | |
814 | bv->bv_len = len; | |
c66a14d0 | 815 | |
c66a14d0 | 816 | bio->bi_iter.bi_size += len; |
0aa69fd3 CH |
817 | bio->bi_vcnt++; |
818 | } | |
819 | EXPORT_SYMBOL_GPL(__bio_add_page); | |
820 | ||
821 | /** | |
551879a4 | 822 | * bio_add_page - attempt to add page(s) to bio |
0aa69fd3 | 823 | * @bio: destination bio |
551879a4 ML |
824 | * @page: start page to add |
825 | * @len: vec entry length, may cross pages | |
826 | * @offset: vec entry offset relative to @page, may cross pages | |
0aa69fd3 | 827 | * |
551879a4 | 828 | * Attempt to add page(s) to the bio_vec maplist. This will only fail |
0aa69fd3 CH |
829 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. |
830 | */ | |
831 | int bio_add_page(struct bio *bio, struct page *page, | |
832 | unsigned int len, unsigned int offset) | |
833 | { | |
ff896738 CH |
834 | bool same_page = false; |
835 | ||
836 | if (!__bio_try_merge_page(bio, page, len, offset, &same_page)) { | |
0aa69fd3 CH |
837 | if (bio_full(bio)) |
838 | return 0; | |
839 | __bio_add_page(bio, page, len, offset); | |
840 | } | |
c66a14d0 | 841 | return len; |
1da177e4 | 842 | } |
a112a71d | 843 | EXPORT_SYMBOL(bio_add_page); |
1da177e4 | 844 | |
7321ecbf CH |
845 | static void bio_get_pages(struct bio *bio) |
846 | { | |
847 | struct bvec_iter_all iter_all; | |
848 | struct bio_vec *bvec; | |
7321ecbf | 849 | |
2b070cfe | 850 | bio_for_each_segment_all(bvec, bio, iter_all) |
7321ecbf CH |
851 | get_page(bvec->bv_page); |
852 | } | |
853 | ||
854 | static void bio_release_pages(struct bio *bio) | |
855 | { | |
856 | struct bvec_iter_all iter_all; | |
857 | struct bio_vec *bvec; | |
7321ecbf | 858 | |
2b070cfe | 859 | bio_for_each_segment_all(bvec, bio, iter_all) |
7321ecbf CH |
860 | put_page(bvec->bv_page); |
861 | } | |
862 | ||
6d0c48ae JA |
863 | static int __bio_iov_bvec_add_pages(struct bio *bio, struct iov_iter *iter) |
864 | { | |
865 | const struct bio_vec *bv = iter->bvec; | |
866 | unsigned int len; | |
867 | size_t size; | |
868 | ||
869 | if (WARN_ON_ONCE(iter->iov_offset > bv->bv_len)) | |
870 | return -EINVAL; | |
871 | ||
872 | len = min_t(size_t, bv->bv_len - iter->iov_offset, iter->count); | |
873 | size = bio_add_page(bio, bv->bv_page, len, | |
874 | bv->bv_offset + iter->iov_offset); | |
a10584c3 CH |
875 | if (unlikely(size != len)) |
876 | return -EINVAL; | |
a10584c3 CH |
877 | iov_iter_advance(iter, size); |
878 | return 0; | |
6d0c48ae JA |
879 | } |
880 | ||
576ed913 CH |
881 | #define PAGE_PTRS_PER_BVEC (sizeof(struct bio_vec) / sizeof(struct page *)) |
882 | ||
2cefe4db | 883 | /** |
17d51b10 | 884 | * __bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio |
2cefe4db KO |
885 | * @bio: bio to add pages to |
886 | * @iter: iov iterator describing the region to be mapped | |
887 | * | |
17d51b10 | 888 | * Pins pages from *iter and appends them to @bio's bvec array. The |
2cefe4db | 889 | * pages will have to be released using put_page() when done. |
17d51b10 MW |
890 | * For multi-segment *iter, this function only adds pages from the |
891 | * the next non-empty segment of the iov iterator. | |
2cefe4db | 892 | */ |
17d51b10 | 893 | static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) |
2cefe4db | 894 | { |
576ed913 CH |
895 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; |
896 | unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt; | |
2cefe4db KO |
897 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; |
898 | struct page **pages = (struct page **)bv; | |
576ed913 CH |
899 | ssize_t size, left; |
900 | unsigned len, i; | |
b403ea24 | 901 | size_t offset; |
576ed913 CH |
902 | |
903 | /* | |
904 | * Move page array up in the allocated memory for the bio vecs as far as | |
905 | * possible so that we can start filling biovecs from the beginning | |
906 | * without overwriting the temporary page array. | |
907 | */ | |
908 | BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2); | |
909 | pages += entries_left * (PAGE_PTRS_PER_BVEC - 1); | |
2cefe4db KO |
910 | |
911 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); | |
912 | if (unlikely(size <= 0)) | |
913 | return size ? size : -EFAULT; | |
2cefe4db | 914 | |
576ed913 CH |
915 | for (left = size, i = 0; left > 0; left -= len, i++) { |
916 | struct page *page = pages[i]; | |
2cefe4db | 917 | |
576ed913 CH |
918 | len = min_t(size_t, PAGE_SIZE - offset, left); |
919 | if (WARN_ON_ONCE(bio_add_page(bio, page, len, offset) != len)) | |
920 | return -EINVAL; | |
921 | offset = 0; | |
2cefe4db KO |
922 | } |
923 | ||
2cefe4db KO |
924 | iov_iter_advance(iter, size); |
925 | return 0; | |
926 | } | |
17d51b10 MW |
927 | |
928 | /** | |
6d0c48ae | 929 | * bio_iov_iter_get_pages - add user or kernel pages to a bio |
17d51b10 | 930 | * @bio: bio to add pages to |
6d0c48ae JA |
931 | * @iter: iov iterator describing the region to be added |
932 | * | |
933 | * This takes either an iterator pointing to user memory, or one pointing to | |
934 | * kernel pages (BVEC iterator). If we're adding user pages, we pin them and | |
935 | * map them into the kernel. On IO completion, the caller should put those | |
399254aa JA |
936 | * pages. If we're adding kernel pages, and the caller told us it's safe to |
937 | * do so, we just have to add the pages to the bio directly. We don't grab an | |
938 | * extra reference to those pages (the user should already have that), and we | |
939 | * don't put the page on IO completion. The caller needs to check if the bio is | |
940 | * flagged BIO_NO_PAGE_REF on IO completion. If it isn't, then pages should be | |
941 | * released. | |
17d51b10 | 942 | * |
17d51b10 | 943 | * The function tries, but does not guarantee, to pin as many pages as |
6d0c48ae JA |
944 | * fit into the bio, or are requested in *iter, whatever is smaller. If |
945 | * MM encounters an error pinning the requested pages, it stops. Error | |
946 | * is returned only if 0 pages could be pinned. | |
17d51b10 MW |
947 | */ |
948 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) | |
949 | { | |
6d0c48ae | 950 | const bool is_bvec = iov_iter_is_bvec(iter); |
14eacf12 CH |
951 | int ret; |
952 | ||
953 | if (WARN_ON_ONCE(bio->bi_vcnt)) | |
954 | return -EINVAL; | |
17d51b10 MW |
955 | |
956 | do { | |
6d0c48ae JA |
957 | if (is_bvec) |
958 | ret = __bio_iov_bvec_add_pages(bio, iter); | |
959 | else | |
960 | ret = __bio_iov_iter_get_pages(bio, iter); | |
14eacf12 | 961 | } while (!ret && iov_iter_count(iter) && !bio_full(bio)); |
17d51b10 | 962 | |
7321ecbf CH |
963 | if (iov_iter_bvec_no_ref(iter)) |
964 | bio_set_flag(bio, BIO_NO_PAGE_REF); | |
0257c0ed | 965 | else if (is_bvec) |
7321ecbf CH |
966 | bio_get_pages(bio); |
967 | ||
14eacf12 | 968 | return bio->bi_vcnt ? 0 : ret; |
17d51b10 | 969 | } |
2cefe4db | 970 | |
4246a0b6 | 971 | static void submit_bio_wait_endio(struct bio *bio) |
9e882242 | 972 | { |
65e53aab | 973 | complete(bio->bi_private); |
9e882242 KO |
974 | } |
975 | ||
976 | /** | |
977 | * submit_bio_wait - submit a bio, and wait until it completes | |
9e882242 KO |
978 | * @bio: The &struct bio which describes the I/O |
979 | * | |
980 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from | |
981 | * bio_endio() on failure. | |
3d289d68 JK |
982 | * |
983 | * WARNING: Unlike to how submit_bio() is usually used, this function does not | |
984 | * result in bio reference to be consumed. The caller must drop the reference | |
985 | * on his own. | |
9e882242 | 986 | */ |
4e49ea4a | 987 | int submit_bio_wait(struct bio *bio) |
9e882242 | 988 | { |
e319e1fb | 989 | DECLARE_COMPLETION_ONSTACK_MAP(done, bio->bi_disk->lockdep_map); |
9e882242 | 990 | |
65e53aab | 991 | bio->bi_private = &done; |
9e882242 | 992 | bio->bi_end_io = submit_bio_wait_endio; |
1eff9d32 | 993 | bio->bi_opf |= REQ_SYNC; |
4e49ea4a | 994 | submit_bio(bio); |
65e53aab | 995 | wait_for_completion_io(&done); |
9e882242 | 996 | |
65e53aab | 997 | return blk_status_to_errno(bio->bi_status); |
9e882242 KO |
998 | } |
999 | EXPORT_SYMBOL(submit_bio_wait); | |
1000 | ||
054bdf64 KO |
1001 | /** |
1002 | * bio_advance - increment/complete a bio by some number of bytes | |
1003 | * @bio: bio to advance | |
1004 | * @bytes: number of bytes to complete | |
1005 | * | |
1006 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to | |
1007 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will | |
1008 | * be updated on the last bvec as well. | |
1009 | * | |
1010 | * @bio will then represent the remaining, uncompleted portion of the io. | |
1011 | */ | |
1012 | void bio_advance(struct bio *bio, unsigned bytes) | |
1013 | { | |
1014 | if (bio_integrity(bio)) | |
1015 | bio_integrity_advance(bio, bytes); | |
1016 | ||
4550dd6c | 1017 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
054bdf64 KO |
1018 | } |
1019 | EXPORT_SYMBOL(bio_advance); | |
1020 | ||
45db54d5 KO |
1021 | void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter, |
1022 | struct bio *src, struct bvec_iter *src_iter) | |
16ac3d63 | 1023 | { |
1cb9dda4 | 1024 | struct bio_vec src_bv, dst_bv; |
16ac3d63 | 1025 | void *src_p, *dst_p; |
1cb9dda4 | 1026 | unsigned bytes; |
16ac3d63 | 1027 | |
45db54d5 KO |
1028 | while (src_iter->bi_size && dst_iter->bi_size) { |
1029 | src_bv = bio_iter_iovec(src, *src_iter); | |
1030 | dst_bv = bio_iter_iovec(dst, *dst_iter); | |
1cb9dda4 KO |
1031 | |
1032 | bytes = min(src_bv.bv_len, dst_bv.bv_len); | |
16ac3d63 | 1033 | |
1cb9dda4 KO |
1034 | src_p = kmap_atomic(src_bv.bv_page); |
1035 | dst_p = kmap_atomic(dst_bv.bv_page); | |
16ac3d63 | 1036 | |
1cb9dda4 KO |
1037 | memcpy(dst_p + dst_bv.bv_offset, |
1038 | src_p + src_bv.bv_offset, | |
16ac3d63 KO |
1039 | bytes); |
1040 | ||
1041 | kunmap_atomic(dst_p); | |
1042 | kunmap_atomic(src_p); | |
1043 | ||
6e6e811d KO |
1044 | flush_dcache_page(dst_bv.bv_page); |
1045 | ||
45db54d5 KO |
1046 | bio_advance_iter(src, src_iter, bytes); |
1047 | bio_advance_iter(dst, dst_iter, bytes); | |
16ac3d63 KO |
1048 | } |
1049 | } | |
38a72dac KO |
1050 | EXPORT_SYMBOL(bio_copy_data_iter); |
1051 | ||
1052 | /** | |
45db54d5 KO |
1053 | * bio_copy_data - copy contents of data buffers from one bio to another |
1054 | * @src: source bio | |
1055 | * @dst: destination bio | |
38a72dac KO |
1056 | * |
1057 | * Stops when it reaches the end of either @src or @dst - that is, copies | |
1058 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). | |
1059 | */ | |
1060 | void bio_copy_data(struct bio *dst, struct bio *src) | |
1061 | { | |
45db54d5 KO |
1062 | struct bvec_iter src_iter = src->bi_iter; |
1063 | struct bvec_iter dst_iter = dst->bi_iter; | |
1064 | ||
1065 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); | |
38a72dac | 1066 | } |
16ac3d63 KO |
1067 | EXPORT_SYMBOL(bio_copy_data); |
1068 | ||
45db54d5 KO |
1069 | /** |
1070 | * bio_list_copy_data - copy contents of data buffers from one chain of bios to | |
1071 | * another | |
1072 | * @src: source bio list | |
1073 | * @dst: destination bio list | |
1074 | * | |
1075 | * Stops when it reaches the end of either the @src list or @dst list - that is, | |
1076 | * copies min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of | |
1077 | * bios). | |
1078 | */ | |
1079 | void bio_list_copy_data(struct bio *dst, struct bio *src) | |
1080 | { | |
1081 | struct bvec_iter src_iter = src->bi_iter; | |
1082 | struct bvec_iter dst_iter = dst->bi_iter; | |
1083 | ||
1084 | while (1) { | |
1085 | if (!src_iter.bi_size) { | |
1086 | src = src->bi_next; | |
1087 | if (!src) | |
1088 | break; | |
1089 | ||
1090 | src_iter = src->bi_iter; | |
1091 | } | |
1092 | ||
1093 | if (!dst_iter.bi_size) { | |
1094 | dst = dst->bi_next; | |
1095 | if (!dst) | |
1096 | break; | |
1097 | ||
1098 | dst_iter = dst->bi_iter; | |
1099 | } | |
1100 | ||
1101 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); | |
1102 | } | |
1103 | } | |
1104 | EXPORT_SYMBOL(bio_list_copy_data); | |
1105 | ||
1da177e4 | 1106 | struct bio_map_data { |
152e283f | 1107 | int is_our_pages; |
26e49cfc KO |
1108 | struct iov_iter iter; |
1109 | struct iovec iov[]; | |
1da177e4 LT |
1110 | }; |
1111 | ||
0e5b935d | 1112 | static struct bio_map_data *bio_alloc_map_data(struct iov_iter *data, |
76029ff3 | 1113 | gfp_t gfp_mask) |
1da177e4 | 1114 | { |
0e5b935d AV |
1115 | struct bio_map_data *bmd; |
1116 | if (data->nr_segs > UIO_MAXIOV) | |
f3f63c1c | 1117 | return NULL; |
1da177e4 | 1118 | |
0e5b935d AV |
1119 | bmd = kmalloc(sizeof(struct bio_map_data) + |
1120 | sizeof(struct iovec) * data->nr_segs, gfp_mask); | |
1121 | if (!bmd) | |
1122 | return NULL; | |
1123 | memcpy(bmd->iov, data->iov, sizeof(struct iovec) * data->nr_segs); | |
1124 | bmd->iter = *data; | |
1125 | bmd->iter.iov = bmd->iov; | |
1126 | return bmd; | |
1da177e4 LT |
1127 | } |
1128 | ||
9124d3fe DP |
1129 | /** |
1130 | * bio_copy_from_iter - copy all pages from iov_iter to bio | |
1131 | * @bio: The &struct bio which describes the I/O as destination | |
1132 | * @iter: iov_iter as source | |
1133 | * | |
1134 | * Copy all pages from iov_iter to bio. | |
1135 | * Returns 0 on success, or error on failure. | |
1136 | */ | |
98a09d61 | 1137 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter *iter) |
c5dec1c3 | 1138 | { |
c5dec1c3 | 1139 | struct bio_vec *bvec; |
6dc4f100 | 1140 | struct bvec_iter_all iter_all; |
c5dec1c3 | 1141 | |
2b070cfe | 1142 | bio_for_each_segment_all(bvec, bio, iter_all) { |
9124d3fe | 1143 | ssize_t ret; |
c5dec1c3 | 1144 | |
9124d3fe DP |
1145 | ret = copy_page_from_iter(bvec->bv_page, |
1146 | bvec->bv_offset, | |
1147 | bvec->bv_len, | |
98a09d61 | 1148 | iter); |
9124d3fe | 1149 | |
98a09d61 | 1150 | if (!iov_iter_count(iter)) |
9124d3fe DP |
1151 | break; |
1152 | ||
1153 | if (ret < bvec->bv_len) | |
1154 | return -EFAULT; | |
c5dec1c3 FT |
1155 | } |
1156 | ||
9124d3fe DP |
1157 | return 0; |
1158 | } | |
1159 | ||
1160 | /** | |
1161 | * bio_copy_to_iter - copy all pages from bio to iov_iter | |
1162 | * @bio: The &struct bio which describes the I/O as source | |
1163 | * @iter: iov_iter as destination | |
1164 | * | |
1165 | * Copy all pages from bio to iov_iter. | |
1166 | * Returns 0 on success, or error on failure. | |
1167 | */ | |
1168 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) | |
1169 | { | |
9124d3fe | 1170 | struct bio_vec *bvec; |
6dc4f100 | 1171 | struct bvec_iter_all iter_all; |
9124d3fe | 1172 | |
2b070cfe | 1173 | bio_for_each_segment_all(bvec, bio, iter_all) { |
9124d3fe DP |
1174 | ssize_t ret; |
1175 | ||
1176 | ret = copy_page_to_iter(bvec->bv_page, | |
1177 | bvec->bv_offset, | |
1178 | bvec->bv_len, | |
1179 | &iter); | |
1180 | ||
1181 | if (!iov_iter_count(&iter)) | |
1182 | break; | |
1183 | ||
1184 | if (ret < bvec->bv_len) | |
1185 | return -EFAULT; | |
1186 | } | |
1187 | ||
1188 | return 0; | |
c5dec1c3 FT |
1189 | } |
1190 | ||
491221f8 | 1191 | void bio_free_pages(struct bio *bio) |
1dfa0f68 CH |
1192 | { |
1193 | struct bio_vec *bvec; | |
6dc4f100 | 1194 | struct bvec_iter_all iter_all; |
1dfa0f68 | 1195 | |
2b070cfe | 1196 | bio_for_each_segment_all(bvec, bio, iter_all) |
1dfa0f68 CH |
1197 | __free_page(bvec->bv_page); |
1198 | } | |
491221f8 | 1199 | EXPORT_SYMBOL(bio_free_pages); |
1dfa0f68 | 1200 | |
1da177e4 LT |
1201 | /** |
1202 | * bio_uncopy_user - finish previously mapped bio | |
1203 | * @bio: bio being terminated | |
1204 | * | |
ddad8dd0 | 1205 | * Free pages allocated from bio_copy_user_iov() and write back data |
1da177e4 LT |
1206 | * to user space in case of a read. |
1207 | */ | |
1208 | int bio_uncopy_user(struct bio *bio) | |
1209 | { | |
1210 | struct bio_map_data *bmd = bio->bi_private; | |
1dfa0f68 | 1211 | int ret = 0; |
1da177e4 | 1212 | |
35dc2483 RD |
1213 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
1214 | /* | |
1215 | * if we're in a workqueue, the request is orphaned, so | |
2d99b55d HR |
1216 | * don't copy into a random user address space, just free |
1217 | * and return -EINTR so user space doesn't expect any data. | |
35dc2483 | 1218 | */ |
2d99b55d HR |
1219 | if (!current->mm) |
1220 | ret = -EINTR; | |
1221 | else if (bio_data_dir(bio) == READ) | |
9124d3fe | 1222 | ret = bio_copy_to_iter(bio, bmd->iter); |
1dfa0f68 CH |
1223 | if (bmd->is_our_pages) |
1224 | bio_free_pages(bio); | |
35dc2483 | 1225 | } |
c8db4448 | 1226 | kfree(bmd); |
1da177e4 LT |
1227 | bio_put(bio); |
1228 | return ret; | |
1229 | } | |
1230 | ||
1231 | /** | |
c5dec1c3 | 1232 | * bio_copy_user_iov - copy user data to bio |
26e49cfc KO |
1233 | * @q: destination block queue |
1234 | * @map_data: pointer to the rq_map_data holding pages (if necessary) | |
1235 | * @iter: iovec iterator | |
1236 | * @gfp_mask: memory allocation flags | |
1da177e4 LT |
1237 | * |
1238 | * Prepares and returns a bio for indirect user io, bouncing data | |
1239 | * to/from kernel pages as necessary. Must be paired with | |
1240 | * call bio_uncopy_user() on io completion. | |
1241 | */ | |
152e283f FT |
1242 | struct bio *bio_copy_user_iov(struct request_queue *q, |
1243 | struct rq_map_data *map_data, | |
e81cef5d | 1244 | struct iov_iter *iter, |
26e49cfc | 1245 | gfp_t gfp_mask) |
1da177e4 | 1246 | { |
1da177e4 | 1247 | struct bio_map_data *bmd; |
1da177e4 LT |
1248 | struct page *page; |
1249 | struct bio *bio; | |
d16d44eb AV |
1250 | int i = 0, ret; |
1251 | int nr_pages; | |
26e49cfc | 1252 | unsigned int len = iter->count; |
bd5cecea | 1253 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
1da177e4 | 1254 | |
0e5b935d | 1255 | bmd = bio_alloc_map_data(iter, gfp_mask); |
1da177e4 LT |
1256 | if (!bmd) |
1257 | return ERR_PTR(-ENOMEM); | |
1258 | ||
26e49cfc KO |
1259 | /* |
1260 | * We need to do a deep copy of the iov_iter including the iovecs. | |
1261 | * The caller provided iov might point to an on-stack or otherwise | |
1262 | * shortlived one. | |
1263 | */ | |
1264 | bmd->is_our_pages = map_data ? 0 : 1; | |
26e49cfc | 1265 | |
d16d44eb AV |
1266 | nr_pages = DIV_ROUND_UP(offset + len, PAGE_SIZE); |
1267 | if (nr_pages > BIO_MAX_PAGES) | |
1268 | nr_pages = BIO_MAX_PAGES; | |
26e49cfc | 1269 | |
1da177e4 | 1270 | ret = -ENOMEM; |
a9e9dc24 | 1271 | bio = bio_kmalloc(gfp_mask, nr_pages); |
1da177e4 LT |
1272 | if (!bio) |
1273 | goto out_bmd; | |
1274 | ||
1da177e4 | 1275 | ret = 0; |
56c451f4 FT |
1276 | |
1277 | if (map_data) { | |
e623ddb4 | 1278 | nr_pages = 1 << map_data->page_order; |
56c451f4 FT |
1279 | i = map_data->offset / PAGE_SIZE; |
1280 | } | |
1da177e4 | 1281 | while (len) { |
e623ddb4 | 1282 | unsigned int bytes = PAGE_SIZE; |
1da177e4 | 1283 | |
56c451f4 FT |
1284 | bytes -= offset; |
1285 | ||
1da177e4 LT |
1286 | if (bytes > len) |
1287 | bytes = len; | |
1288 | ||
152e283f | 1289 | if (map_data) { |
e623ddb4 | 1290 | if (i == map_data->nr_entries * nr_pages) { |
152e283f FT |
1291 | ret = -ENOMEM; |
1292 | break; | |
1293 | } | |
e623ddb4 FT |
1294 | |
1295 | page = map_data->pages[i / nr_pages]; | |
1296 | page += (i % nr_pages); | |
1297 | ||
1298 | i++; | |
1299 | } else { | |
152e283f | 1300 | page = alloc_page(q->bounce_gfp | gfp_mask); |
e623ddb4 FT |
1301 | if (!page) { |
1302 | ret = -ENOMEM; | |
1303 | break; | |
1304 | } | |
1da177e4 LT |
1305 | } |
1306 | ||
a3761c3c JG |
1307 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) { |
1308 | if (!map_data) | |
1309 | __free_page(page); | |
1da177e4 | 1310 | break; |
a3761c3c | 1311 | } |
1da177e4 LT |
1312 | |
1313 | len -= bytes; | |
56c451f4 | 1314 | offset = 0; |
1da177e4 LT |
1315 | } |
1316 | ||
1317 | if (ret) | |
1318 | goto cleanup; | |
1319 | ||
2884d0be AV |
1320 | if (map_data) |
1321 | map_data->offset += bio->bi_iter.bi_size; | |
1322 | ||
1da177e4 LT |
1323 | /* |
1324 | * success | |
1325 | */ | |
00e23707 | 1326 | if ((iov_iter_rw(iter) == WRITE && (!map_data || !map_data->null_mapped)) || |
ecb554a8 | 1327 | (map_data && map_data->from_user)) { |
98a09d61 | 1328 | ret = bio_copy_from_iter(bio, iter); |
c5dec1c3 FT |
1329 | if (ret) |
1330 | goto cleanup; | |
98a09d61 | 1331 | } else { |
f55adad6 KB |
1332 | if (bmd->is_our_pages) |
1333 | zero_fill_bio(bio); | |
98a09d61 | 1334 | iov_iter_advance(iter, bio->bi_iter.bi_size); |
1da177e4 LT |
1335 | } |
1336 | ||
26e49cfc | 1337 | bio->bi_private = bmd; |
2884d0be AV |
1338 | if (map_data && map_data->null_mapped) |
1339 | bio_set_flag(bio, BIO_NULL_MAPPED); | |
1da177e4 LT |
1340 | return bio; |
1341 | cleanup: | |
152e283f | 1342 | if (!map_data) |
1dfa0f68 | 1343 | bio_free_pages(bio); |
1da177e4 LT |
1344 | bio_put(bio); |
1345 | out_bmd: | |
c8db4448 | 1346 | kfree(bmd); |
1da177e4 LT |
1347 | return ERR_PTR(ret); |
1348 | } | |
1349 | ||
37f19e57 CH |
1350 | /** |
1351 | * bio_map_user_iov - map user iovec into bio | |
1352 | * @q: the struct request_queue for the bio | |
1353 | * @iter: iovec iterator | |
1354 | * @gfp_mask: memory allocation flags | |
1355 | * | |
1356 | * Map the user space address into a bio suitable for io to a block | |
1357 | * device. Returns an error pointer in case of error. | |
1358 | */ | |
1359 | struct bio *bio_map_user_iov(struct request_queue *q, | |
e81cef5d | 1360 | struct iov_iter *iter, |
37f19e57 | 1361 | gfp_t gfp_mask) |
1da177e4 | 1362 | { |
26e49cfc | 1363 | int j; |
1da177e4 | 1364 | struct bio *bio; |
076098e5 | 1365 | int ret; |
2b04e8f6 | 1366 | struct bio_vec *bvec; |
6dc4f100 | 1367 | struct bvec_iter_all iter_all; |
1da177e4 | 1368 | |
b282cc76 | 1369 | if (!iov_iter_count(iter)) |
1da177e4 LT |
1370 | return ERR_PTR(-EINVAL); |
1371 | ||
b282cc76 | 1372 | bio = bio_kmalloc(gfp_mask, iov_iter_npages(iter, BIO_MAX_PAGES)); |
1da177e4 LT |
1373 | if (!bio) |
1374 | return ERR_PTR(-ENOMEM); | |
1375 | ||
0a0f1513 | 1376 | while (iov_iter_count(iter)) { |
629e42bc | 1377 | struct page **pages; |
076098e5 AV |
1378 | ssize_t bytes; |
1379 | size_t offs, added = 0; | |
1380 | int npages; | |
1da177e4 | 1381 | |
0a0f1513 | 1382 | bytes = iov_iter_get_pages_alloc(iter, &pages, LONG_MAX, &offs); |
076098e5 AV |
1383 | if (unlikely(bytes <= 0)) { |
1384 | ret = bytes ? bytes : -EFAULT; | |
f1970baf | 1385 | goto out_unmap; |
99172157 | 1386 | } |
f1970baf | 1387 | |
076098e5 | 1388 | npages = DIV_ROUND_UP(offs + bytes, PAGE_SIZE); |
f1970baf | 1389 | |
98f0bc99 AV |
1390 | if (unlikely(offs & queue_dma_alignment(q))) { |
1391 | ret = -EINVAL; | |
1392 | j = 0; | |
1393 | } else { | |
1394 | for (j = 0; j < npages; j++) { | |
1395 | struct page *page = pages[j]; | |
1396 | unsigned int n = PAGE_SIZE - offs; | |
f1970baf | 1397 | |
98f0bc99 AV |
1398 | if (n > bytes) |
1399 | n = bytes; | |
95d78c28 | 1400 | |
19047087 ML |
1401 | if (!__bio_add_pc_page(q, bio, page, n, offs, |
1402 | true)) | |
98f0bc99 | 1403 | break; |
1da177e4 | 1404 | |
98f0bc99 AV |
1405 | added += n; |
1406 | bytes -= n; | |
1407 | offs = 0; | |
1408 | } | |
0a0f1513 | 1409 | iov_iter_advance(iter, added); |
f1970baf | 1410 | } |
1da177e4 | 1411 | /* |
f1970baf | 1412 | * release the pages we didn't map into the bio, if any |
1da177e4 | 1413 | */ |
629e42bc | 1414 | while (j < npages) |
09cbfeaf | 1415 | put_page(pages[j++]); |
629e42bc | 1416 | kvfree(pages); |
e2e115d1 AV |
1417 | /* couldn't stuff something into bio? */ |
1418 | if (bytes) | |
1419 | break; | |
1da177e4 LT |
1420 | } |
1421 | ||
b7c44ed9 | 1422 | bio_set_flag(bio, BIO_USER_MAPPED); |
37f19e57 CH |
1423 | |
1424 | /* | |
5fad1b64 | 1425 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
37f19e57 CH |
1426 | * it would normally disappear when its bi_end_io is run. |
1427 | * however, we need it for the unmap, so grab an extra | |
1428 | * reference to it | |
1429 | */ | |
1430 | bio_get(bio); | |
1da177e4 | 1431 | return bio; |
f1970baf JB |
1432 | |
1433 | out_unmap: | |
2b070cfe | 1434 | bio_for_each_segment_all(bvec, bio, iter_all) { |
2b04e8f6 | 1435 | put_page(bvec->bv_page); |
f1970baf | 1436 | } |
1da177e4 LT |
1437 | bio_put(bio); |
1438 | return ERR_PTR(ret); | |
1439 | } | |
1440 | ||
1da177e4 LT |
1441 | static void __bio_unmap_user(struct bio *bio) |
1442 | { | |
1443 | struct bio_vec *bvec; | |
6dc4f100 | 1444 | struct bvec_iter_all iter_all; |
1da177e4 LT |
1445 | |
1446 | /* | |
1447 | * make sure we dirty pages we wrote to | |
1448 | */ | |
2b070cfe | 1449 | bio_for_each_segment_all(bvec, bio, iter_all) { |
1da177e4 LT |
1450 | if (bio_data_dir(bio) == READ) |
1451 | set_page_dirty_lock(bvec->bv_page); | |
1452 | ||
09cbfeaf | 1453 | put_page(bvec->bv_page); |
1da177e4 LT |
1454 | } |
1455 | ||
1456 | bio_put(bio); | |
1457 | } | |
1458 | ||
1459 | /** | |
1460 | * bio_unmap_user - unmap a bio | |
1461 | * @bio: the bio being unmapped | |
1462 | * | |
5fad1b64 BVA |
1463 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
1464 | * process context. | |
1da177e4 LT |
1465 | * |
1466 | * bio_unmap_user() may sleep. | |
1467 | */ | |
1468 | void bio_unmap_user(struct bio *bio) | |
1469 | { | |
1470 | __bio_unmap_user(bio); | |
1471 | bio_put(bio); | |
1472 | } | |
1473 | ||
4246a0b6 | 1474 | static void bio_map_kern_endio(struct bio *bio) |
b823825e | 1475 | { |
b823825e | 1476 | bio_put(bio); |
b823825e JA |
1477 | } |
1478 | ||
75c72b83 CH |
1479 | /** |
1480 | * bio_map_kern - map kernel address into bio | |
1481 | * @q: the struct request_queue for the bio | |
1482 | * @data: pointer to buffer to map | |
1483 | * @len: length in bytes | |
1484 | * @gfp_mask: allocation flags for bio allocation | |
1485 | * | |
1486 | * Map the kernel address into a bio suitable for io to a block | |
1487 | * device. Returns an error pointer in case of error. | |
1488 | */ | |
1489 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, | |
1490 | gfp_t gfp_mask) | |
df46b9a4 MC |
1491 | { |
1492 | unsigned long kaddr = (unsigned long)data; | |
1493 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1494 | unsigned long start = kaddr >> PAGE_SHIFT; | |
1495 | const int nr_pages = end - start; | |
1496 | int offset, i; | |
1497 | struct bio *bio; | |
1498 | ||
a9e9dc24 | 1499 | bio = bio_kmalloc(gfp_mask, nr_pages); |
df46b9a4 MC |
1500 | if (!bio) |
1501 | return ERR_PTR(-ENOMEM); | |
1502 | ||
1503 | offset = offset_in_page(kaddr); | |
1504 | for (i = 0; i < nr_pages; i++) { | |
1505 | unsigned int bytes = PAGE_SIZE - offset; | |
1506 | ||
1507 | if (len <= 0) | |
1508 | break; | |
1509 | ||
1510 | if (bytes > len) | |
1511 | bytes = len; | |
1512 | ||
defd94b7 | 1513 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
75c72b83 CH |
1514 | offset) < bytes) { |
1515 | /* we don't support partial mappings */ | |
1516 | bio_put(bio); | |
1517 | return ERR_PTR(-EINVAL); | |
1518 | } | |
df46b9a4 MC |
1519 | |
1520 | data += bytes; | |
1521 | len -= bytes; | |
1522 | offset = 0; | |
1523 | } | |
1524 | ||
b823825e | 1525 | bio->bi_end_io = bio_map_kern_endio; |
df46b9a4 MC |
1526 | return bio; |
1527 | } | |
a112a71d | 1528 | EXPORT_SYMBOL(bio_map_kern); |
df46b9a4 | 1529 | |
4246a0b6 | 1530 | static void bio_copy_kern_endio(struct bio *bio) |
68154e90 | 1531 | { |
1dfa0f68 CH |
1532 | bio_free_pages(bio); |
1533 | bio_put(bio); | |
1534 | } | |
1535 | ||
4246a0b6 | 1536 | static void bio_copy_kern_endio_read(struct bio *bio) |
1dfa0f68 | 1537 | { |
42d2683a | 1538 | char *p = bio->bi_private; |
1dfa0f68 | 1539 | struct bio_vec *bvec; |
6dc4f100 | 1540 | struct bvec_iter_all iter_all; |
68154e90 | 1541 | |
2b070cfe | 1542 | bio_for_each_segment_all(bvec, bio, iter_all) { |
1dfa0f68 | 1543 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
c8db4448 | 1544 | p += bvec->bv_len; |
68154e90 FT |
1545 | } |
1546 | ||
4246a0b6 | 1547 | bio_copy_kern_endio(bio); |
68154e90 FT |
1548 | } |
1549 | ||
1550 | /** | |
1551 | * bio_copy_kern - copy kernel address into bio | |
1552 | * @q: the struct request_queue for the bio | |
1553 | * @data: pointer to buffer to copy | |
1554 | * @len: length in bytes | |
1555 | * @gfp_mask: allocation flags for bio and page allocation | |
ffee0259 | 1556 | * @reading: data direction is READ |
68154e90 FT |
1557 | * |
1558 | * copy the kernel address into a bio suitable for io to a block | |
1559 | * device. Returns an error pointer in case of error. | |
1560 | */ | |
1561 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, | |
1562 | gfp_t gfp_mask, int reading) | |
1563 | { | |
42d2683a CH |
1564 | unsigned long kaddr = (unsigned long)data; |
1565 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
1566 | unsigned long start = kaddr >> PAGE_SHIFT; | |
42d2683a CH |
1567 | struct bio *bio; |
1568 | void *p = data; | |
1dfa0f68 | 1569 | int nr_pages = 0; |
68154e90 | 1570 | |
42d2683a CH |
1571 | /* |
1572 | * Overflow, abort | |
1573 | */ | |
1574 | if (end < start) | |
1575 | return ERR_PTR(-EINVAL); | |
68154e90 | 1576 | |
42d2683a CH |
1577 | nr_pages = end - start; |
1578 | bio = bio_kmalloc(gfp_mask, nr_pages); | |
1579 | if (!bio) | |
1580 | return ERR_PTR(-ENOMEM); | |
68154e90 | 1581 | |
42d2683a CH |
1582 | while (len) { |
1583 | struct page *page; | |
1584 | unsigned int bytes = PAGE_SIZE; | |
68154e90 | 1585 | |
42d2683a CH |
1586 | if (bytes > len) |
1587 | bytes = len; | |
1588 | ||
1589 | page = alloc_page(q->bounce_gfp | gfp_mask); | |
1590 | if (!page) | |
1591 | goto cleanup; | |
1592 | ||
1593 | if (!reading) | |
1594 | memcpy(page_address(page), p, bytes); | |
1595 | ||
1596 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) | |
1597 | break; | |
1598 | ||
1599 | len -= bytes; | |
1600 | p += bytes; | |
68154e90 FT |
1601 | } |
1602 | ||
1dfa0f68 CH |
1603 | if (reading) { |
1604 | bio->bi_end_io = bio_copy_kern_endio_read; | |
1605 | bio->bi_private = data; | |
1606 | } else { | |
1607 | bio->bi_end_io = bio_copy_kern_endio; | |
1dfa0f68 | 1608 | } |
76029ff3 | 1609 | |
68154e90 | 1610 | return bio; |
42d2683a CH |
1611 | |
1612 | cleanup: | |
1dfa0f68 | 1613 | bio_free_pages(bio); |
42d2683a CH |
1614 | bio_put(bio); |
1615 | return ERR_PTR(-ENOMEM); | |
68154e90 FT |
1616 | } |
1617 | ||
1da177e4 LT |
1618 | /* |
1619 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions | |
1620 | * for performing direct-IO in BIOs. | |
1621 | * | |
1622 | * The problem is that we cannot run set_page_dirty() from interrupt context | |
1623 | * because the required locks are not interrupt-safe. So what we can do is to | |
1624 | * mark the pages dirty _before_ performing IO. And in interrupt context, | |
1625 | * check that the pages are still dirty. If so, fine. If not, redirty them | |
1626 | * in process context. | |
1627 | * | |
1628 | * We special-case compound pages here: normally this means reads into hugetlb | |
1629 | * pages. The logic in here doesn't really work right for compound pages | |
1630 | * because the VM does not uniformly chase down the head page in all cases. | |
1631 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't | |
1632 | * handle them at all. So we skip compound pages here at an early stage. | |
1633 | * | |
1634 | * Note that this code is very hard to test under normal circumstances because | |
1635 | * direct-io pins the pages with get_user_pages(). This makes | |
1636 | * is_page_cache_freeable return false, and the VM will not clean the pages. | |
0d5c3eba | 1637 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
1da177e4 LT |
1638 | * pagecache. |
1639 | * | |
1640 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the | |
1641 | * deferred bio dirtying paths. | |
1642 | */ | |
1643 | ||
1644 | /* | |
1645 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. | |
1646 | */ | |
1647 | void bio_set_pages_dirty(struct bio *bio) | |
1648 | { | |
cb34e057 | 1649 | struct bio_vec *bvec; |
6dc4f100 | 1650 | struct bvec_iter_all iter_all; |
1da177e4 | 1651 | |
2b070cfe | 1652 | bio_for_each_segment_all(bvec, bio, iter_all) { |
3bb50983 CH |
1653 | if (!PageCompound(bvec->bv_page)) |
1654 | set_page_dirty_lock(bvec->bv_page); | |
1da177e4 LT |
1655 | } |
1656 | } | |
1657 | ||
1da177e4 LT |
1658 | /* |
1659 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. | |
1660 | * If they are, then fine. If, however, some pages are clean then they must | |
1661 | * have been written out during the direct-IO read. So we take another ref on | |
24d5493f | 1662 | * the BIO and re-dirty the pages in process context. |
1da177e4 LT |
1663 | * |
1664 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from | |
ea1754a0 KS |
1665 | * here on. It will run one put_page() against each page and will run one |
1666 | * bio_put() against the BIO. | |
1da177e4 LT |
1667 | */ |
1668 | ||
65f27f38 | 1669 | static void bio_dirty_fn(struct work_struct *work); |
1da177e4 | 1670 | |
65f27f38 | 1671 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
1da177e4 LT |
1672 | static DEFINE_SPINLOCK(bio_dirty_lock); |
1673 | static struct bio *bio_dirty_list; | |
1674 | ||
1675 | /* | |
1676 | * This runs in process context | |
1677 | */ | |
65f27f38 | 1678 | static void bio_dirty_fn(struct work_struct *work) |
1da177e4 | 1679 | { |
24d5493f | 1680 | struct bio *bio, *next; |
1da177e4 | 1681 | |
24d5493f CH |
1682 | spin_lock_irq(&bio_dirty_lock); |
1683 | next = bio_dirty_list; | |
1da177e4 | 1684 | bio_dirty_list = NULL; |
24d5493f | 1685 | spin_unlock_irq(&bio_dirty_lock); |
1da177e4 | 1686 | |
24d5493f CH |
1687 | while ((bio = next) != NULL) { |
1688 | next = bio->bi_private; | |
1da177e4 LT |
1689 | |
1690 | bio_set_pages_dirty(bio); | |
399254aa JA |
1691 | if (!bio_flagged(bio, BIO_NO_PAGE_REF)) |
1692 | bio_release_pages(bio); | |
1da177e4 | 1693 | bio_put(bio); |
1da177e4 LT |
1694 | } |
1695 | } | |
1696 | ||
1697 | void bio_check_pages_dirty(struct bio *bio) | |
1698 | { | |
cb34e057 | 1699 | struct bio_vec *bvec; |
24d5493f | 1700 | unsigned long flags; |
6dc4f100 | 1701 | struct bvec_iter_all iter_all; |
1da177e4 | 1702 | |
2b070cfe | 1703 | bio_for_each_segment_all(bvec, bio, iter_all) { |
24d5493f CH |
1704 | if (!PageDirty(bvec->bv_page) && !PageCompound(bvec->bv_page)) |
1705 | goto defer; | |
1da177e4 LT |
1706 | } |
1707 | ||
399254aa JA |
1708 | if (!bio_flagged(bio, BIO_NO_PAGE_REF)) |
1709 | bio_release_pages(bio); | |
24d5493f CH |
1710 | bio_put(bio); |
1711 | return; | |
1712 | defer: | |
1713 | spin_lock_irqsave(&bio_dirty_lock, flags); | |
1714 | bio->bi_private = bio_dirty_list; | |
1715 | bio_dirty_list = bio; | |
1716 | spin_unlock_irqrestore(&bio_dirty_lock, flags); | |
1717 | schedule_work(&bio_dirty_work); | |
1da177e4 LT |
1718 | } |
1719 | ||
5b18b5a7 MP |
1720 | void update_io_ticks(struct hd_struct *part, unsigned long now) |
1721 | { | |
1722 | unsigned long stamp; | |
1723 | again: | |
1724 | stamp = READ_ONCE(part->stamp); | |
1725 | if (unlikely(stamp != now)) { | |
1726 | if (likely(cmpxchg(&part->stamp, stamp, now) == stamp)) { | |
1727 | __part_stat_add(part, io_ticks, 1); | |
1728 | } | |
1729 | } | |
1730 | if (part->partno) { | |
1731 | part = &part_to_disk(part)->part0; | |
1732 | goto again; | |
1733 | } | |
1734 | } | |
1da177e4 | 1735 | |
ddcf35d3 | 1736 | void generic_start_io_acct(struct request_queue *q, int op, |
d62e26b3 | 1737 | unsigned long sectors, struct hd_struct *part) |
394ffa50 | 1738 | { |
ddcf35d3 | 1739 | const int sgrp = op_stat_group(op); |
394ffa50 | 1740 | |
112f158f MS |
1741 | part_stat_lock(); |
1742 | ||
5b18b5a7 | 1743 | update_io_ticks(part, jiffies); |
112f158f MS |
1744 | part_stat_inc(part, ios[sgrp]); |
1745 | part_stat_add(part, sectors[sgrp], sectors); | |
ddcf35d3 | 1746 | part_inc_in_flight(q, part, op_is_write(op)); |
394ffa50 GZ |
1747 | |
1748 | part_stat_unlock(); | |
1749 | } | |
1750 | EXPORT_SYMBOL(generic_start_io_acct); | |
1751 | ||
ddcf35d3 | 1752 | void generic_end_io_acct(struct request_queue *q, int req_op, |
d62e26b3 | 1753 | struct hd_struct *part, unsigned long start_time) |
394ffa50 | 1754 | { |
5b18b5a7 MP |
1755 | unsigned long now = jiffies; |
1756 | unsigned long duration = now - start_time; | |
ddcf35d3 | 1757 | const int sgrp = op_stat_group(req_op); |
394ffa50 | 1758 | |
112f158f MS |
1759 | part_stat_lock(); |
1760 | ||
5b18b5a7 | 1761 | update_io_ticks(part, now); |
112f158f | 1762 | part_stat_add(part, nsecs[sgrp], jiffies_to_nsecs(duration)); |
5b18b5a7 | 1763 | part_stat_add(part, time_in_queue, duration); |
ddcf35d3 | 1764 | part_dec_in_flight(q, part, op_is_write(req_op)); |
394ffa50 GZ |
1765 | |
1766 | part_stat_unlock(); | |
1767 | } | |
1768 | EXPORT_SYMBOL(generic_end_io_acct); | |
1769 | ||
2d4dc890 IL |
1770 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1771 | void bio_flush_dcache_pages(struct bio *bi) | |
1772 | { | |
7988613b KO |
1773 | struct bio_vec bvec; |
1774 | struct bvec_iter iter; | |
2d4dc890 | 1775 | |
7988613b KO |
1776 | bio_for_each_segment(bvec, bi, iter) |
1777 | flush_dcache_page(bvec.bv_page); | |
2d4dc890 IL |
1778 | } |
1779 | EXPORT_SYMBOL(bio_flush_dcache_pages); | |
1780 | #endif | |
1781 | ||
c4cf5261 JA |
1782 | static inline bool bio_remaining_done(struct bio *bio) |
1783 | { | |
1784 | /* | |
1785 | * If we're not chaining, then ->__bi_remaining is always 1 and | |
1786 | * we always end io on the first invocation. | |
1787 | */ | |
1788 | if (!bio_flagged(bio, BIO_CHAIN)) | |
1789 | return true; | |
1790 | ||
1791 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); | |
1792 | ||
326e1dbb | 1793 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
b7c44ed9 | 1794 | bio_clear_flag(bio, BIO_CHAIN); |
c4cf5261 | 1795 | return true; |
326e1dbb | 1796 | } |
c4cf5261 JA |
1797 | |
1798 | return false; | |
1799 | } | |
1800 | ||
1da177e4 LT |
1801 | /** |
1802 | * bio_endio - end I/O on a bio | |
1803 | * @bio: bio | |
1da177e4 LT |
1804 | * |
1805 | * Description: | |
4246a0b6 CH |
1806 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
1807 | * way to end I/O on a bio. No one should call bi_end_io() directly on a | |
1808 | * bio unless they own it and thus know that it has an end_io function. | |
fbbaf700 N |
1809 | * |
1810 | * bio_endio() can be called several times on a bio that has been chained | |
1811 | * using bio_chain(). The ->bi_end_io() function will only be called the | |
1812 | * last time. At this point the BLK_TA_COMPLETE tracing event will be | |
1813 | * generated if BIO_TRACE_COMPLETION is set. | |
1da177e4 | 1814 | **/ |
4246a0b6 | 1815 | void bio_endio(struct bio *bio) |
1da177e4 | 1816 | { |
ba8c6967 | 1817 | again: |
2b885517 | 1818 | if (!bio_remaining_done(bio)) |
ba8c6967 | 1819 | return; |
7c20f116 CH |
1820 | if (!bio_integrity_endio(bio)) |
1821 | return; | |
1da177e4 | 1822 | |
67b42d0b JB |
1823 | if (bio->bi_disk) |
1824 | rq_qos_done_bio(bio->bi_disk->queue, bio); | |
1825 | ||
ba8c6967 CH |
1826 | /* |
1827 | * Need to have a real endio function for chained bios, otherwise | |
1828 | * various corner cases will break (like stacking block devices that | |
1829 | * save/restore bi_end_io) - however, we want to avoid unbounded | |
1830 | * recursion and blowing the stack. Tail call optimization would | |
1831 | * handle this, but compiling with frame pointers also disables | |
1832 | * gcc's sibling call optimization. | |
1833 | */ | |
1834 | if (bio->bi_end_io == bio_chain_endio) { | |
1835 | bio = __bio_chain_endio(bio); | |
1836 | goto again; | |
196d38bc | 1837 | } |
ba8c6967 | 1838 | |
74d46992 CH |
1839 | if (bio->bi_disk && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
1840 | trace_block_bio_complete(bio->bi_disk->queue, bio, | |
a462b950 | 1841 | blk_status_to_errno(bio->bi_status)); |
fbbaf700 N |
1842 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
1843 | } | |
1844 | ||
9e234eea | 1845 | blk_throtl_bio_endio(bio); |
b222dd2f SL |
1846 | /* release cgroup info */ |
1847 | bio_uninit(bio); | |
ba8c6967 CH |
1848 | if (bio->bi_end_io) |
1849 | bio->bi_end_io(bio); | |
1da177e4 | 1850 | } |
a112a71d | 1851 | EXPORT_SYMBOL(bio_endio); |
1da177e4 | 1852 | |
20d0189b KO |
1853 | /** |
1854 | * bio_split - split a bio | |
1855 | * @bio: bio to split | |
1856 | * @sectors: number of sectors to split from the front of @bio | |
1857 | * @gfp: gfp mask | |
1858 | * @bs: bio set to allocate from | |
1859 | * | |
1860 | * Allocates and returns a new bio which represents @sectors from the start of | |
1861 | * @bio, and updates @bio to represent the remaining sectors. | |
1862 | * | |
f3f5da62 MP |
1863 | * Unless this is a discard request the newly allocated bio will point |
1864 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that | |
1865 | * @bio is not freed before the split. | |
20d0189b KO |
1866 | */ |
1867 | struct bio *bio_split(struct bio *bio, int sectors, | |
1868 | gfp_t gfp, struct bio_set *bs) | |
1869 | { | |
f341a4d3 | 1870 | struct bio *split; |
20d0189b KO |
1871 | |
1872 | BUG_ON(sectors <= 0); | |
1873 | BUG_ON(sectors >= bio_sectors(bio)); | |
1874 | ||
f9d03f96 | 1875 | split = bio_clone_fast(bio, gfp, bs); |
20d0189b KO |
1876 | if (!split) |
1877 | return NULL; | |
1878 | ||
1879 | split->bi_iter.bi_size = sectors << 9; | |
1880 | ||
1881 | if (bio_integrity(split)) | |
fbd08e76 | 1882 | bio_integrity_trim(split); |
20d0189b KO |
1883 | |
1884 | bio_advance(bio, split->bi_iter.bi_size); | |
1885 | ||
fbbaf700 | 1886 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
20d59023 | 1887 | bio_set_flag(split, BIO_TRACE_COMPLETION); |
fbbaf700 | 1888 | |
20d0189b KO |
1889 | return split; |
1890 | } | |
1891 | EXPORT_SYMBOL(bio_split); | |
1892 | ||
6678d83f KO |
1893 | /** |
1894 | * bio_trim - trim a bio | |
1895 | * @bio: bio to trim | |
1896 | * @offset: number of sectors to trim from the front of @bio | |
1897 | * @size: size we want to trim @bio to, in sectors | |
1898 | */ | |
1899 | void bio_trim(struct bio *bio, int offset, int size) | |
1900 | { | |
1901 | /* 'bio' is a cloned bio which we need to trim to match | |
1902 | * the given offset and size. | |
6678d83f | 1903 | */ |
6678d83f KO |
1904 | |
1905 | size <<= 9; | |
4f024f37 | 1906 | if (offset == 0 && size == bio->bi_iter.bi_size) |
6678d83f KO |
1907 | return; |
1908 | ||
b7c44ed9 | 1909 | bio_clear_flag(bio, BIO_SEG_VALID); |
6678d83f KO |
1910 | |
1911 | bio_advance(bio, offset << 9); | |
1912 | ||
4f024f37 | 1913 | bio->bi_iter.bi_size = size; |
376a78ab DM |
1914 | |
1915 | if (bio_integrity(bio)) | |
fbd08e76 | 1916 | bio_integrity_trim(bio); |
376a78ab | 1917 | |
6678d83f KO |
1918 | } |
1919 | EXPORT_SYMBOL_GPL(bio_trim); | |
1920 | ||
1da177e4 LT |
1921 | /* |
1922 | * create memory pools for biovec's in a bio_set. | |
1923 | * use the global biovec slabs created for general use. | |
1924 | */ | |
8aa6ba2f | 1925 | int biovec_init_pool(mempool_t *pool, int pool_entries) |
1da177e4 | 1926 | { |
ed996a52 | 1927 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
1da177e4 | 1928 | |
8aa6ba2f | 1929 | return mempool_init_slab_pool(pool, pool_entries, bp->slab); |
1da177e4 LT |
1930 | } |
1931 | ||
917a38c7 KO |
1932 | /* |
1933 | * bioset_exit - exit a bioset initialized with bioset_init() | |
1934 | * | |
1935 | * May be called on a zeroed but uninitialized bioset (i.e. allocated with | |
1936 | * kzalloc()). | |
1937 | */ | |
1938 | void bioset_exit(struct bio_set *bs) | |
1da177e4 | 1939 | { |
df2cb6da KO |
1940 | if (bs->rescue_workqueue) |
1941 | destroy_workqueue(bs->rescue_workqueue); | |
917a38c7 | 1942 | bs->rescue_workqueue = NULL; |
df2cb6da | 1943 | |
8aa6ba2f KO |
1944 | mempool_exit(&bs->bio_pool); |
1945 | mempool_exit(&bs->bvec_pool); | |
9f060e22 | 1946 | |
7878cba9 | 1947 | bioset_integrity_free(bs); |
917a38c7 KO |
1948 | if (bs->bio_slab) |
1949 | bio_put_slab(bs); | |
1950 | bs->bio_slab = NULL; | |
1951 | } | |
1952 | EXPORT_SYMBOL(bioset_exit); | |
1da177e4 | 1953 | |
917a38c7 KO |
1954 | /** |
1955 | * bioset_init - Initialize a bio_set | |
dad08527 | 1956 | * @bs: pool to initialize |
917a38c7 KO |
1957 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
1958 | * @front_pad: Number of bytes to allocate in front of the returned bio | |
1959 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS | |
1960 | * and %BIOSET_NEED_RESCUER | |
1961 | * | |
dad08527 KO |
1962 | * Description: |
1963 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller | |
1964 | * to ask for a number of bytes to be allocated in front of the bio. | |
1965 | * Front pad allocation is useful for embedding the bio inside | |
1966 | * another structure, to avoid allocating extra data to go with the bio. | |
1967 | * Note that the bio must be embedded at the END of that structure always, | |
1968 | * or things will break badly. | |
1969 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated | |
1970 | * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast(). | |
1971 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to | |
1972 | * dispatch queued requests when the mempool runs out of space. | |
1973 | * | |
917a38c7 KO |
1974 | */ |
1975 | int bioset_init(struct bio_set *bs, | |
1976 | unsigned int pool_size, | |
1977 | unsigned int front_pad, | |
1978 | int flags) | |
1979 | { | |
1980 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); | |
1981 | ||
1982 | bs->front_pad = front_pad; | |
1983 | ||
1984 | spin_lock_init(&bs->rescue_lock); | |
1985 | bio_list_init(&bs->rescue_list); | |
1986 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); | |
1987 | ||
1988 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); | |
1989 | if (!bs->bio_slab) | |
1990 | return -ENOMEM; | |
1991 | ||
1992 | if (mempool_init_slab_pool(&bs->bio_pool, pool_size, bs->bio_slab)) | |
1993 | goto bad; | |
1994 | ||
1995 | if ((flags & BIOSET_NEED_BVECS) && | |
1996 | biovec_init_pool(&bs->bvec_pool, pool_size)) | |
1997 | goto bad; | |
1998 | ||
1999 | if (!(flags & BIOSET_NEED_RESCUER)) | |
2000 | return 0; | |
2001 | ||
2002 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); | |
2003 | if (!bs->rescue_workqueue) | |
2004 | goto bad; | |
2005 | ||
2006 | return 0; | |
2007 | bad: | |
2008 | bioset_exit(bs); | |
2009 | return -ENOMEM; | |
2010 | } | |
2011 | EXPORT_SYMBOL(bioset_init); | |
2012 | ||
28e89fd9 JA |
2013 | /* |
2014 | * Initialize and setup a new bio_set, based on the settings from | |
2015 | * another bio_set. | |
2016 | */ | |
2017 | int bioset_init_from_src(struct bio_set *bs, struct bio_set *src) | |
2018 | { | |
2019 | int flags; | |
2020 | ||
2021 | flags = 0; | |
2022 | if (src->bvec_pool.min_nr) | |
2023 | flags |= BIOSET_NEED_BVECS; | |
2024 | if (src->rescue_workqueue) | |
2025 | flags |= BIOSET_NEED_RESCUER; | |
2026 | ||
2027 | return bioset_init(bs, src->bio_pool.min_nr, src->front_pad, flags); | |
2028 | } | |
2029 | EXPORT_SYMBOL(bioset_init_from_src); | |
2030 | ||
852c788f | 2031 | #ifdef CONFIG_BLK_CGROUP |
1d933cf0 | 2032 | |
74b7c02a | 2033 | /** |
2268c0fe | 2034 | * bio_disassociate_blkg - puts back the blkg reference if associated |
74b7c02a | 2035 | * @bio: target bio |
74b7c02a | 2036 | * |
2268c0fe | 2037 | * Helper to disassociate the blkg from @bio if a blkg is associated. |
74b7c02a | 2038 | */ |
2268c0fe | 2039 | void bio_disassociate_blkg(struct bio *bio) |
74b7c02a | 2040 | { |
2268c0fe DZ |
2041 | if (bio->bi_blkg) { |
2042 | blkg_put(bio->bi_blkg); | |
2043 | bio->bi_blkg = NULL; | |
2044 | } | |
74b7c02a | 2045 | } |
892ad71f | 2046 | EXPORT_SYMBOL_GPL(bio_disassociate_blkg); |
74b7c02a | 2047 | |
a7b39b4e | 2048 | /** |
2268c0fe | 2049 | * __bio_associate_blkg - associate a bio with the a blkg |
a7b39b4e | 2050 | * @bio: target bio |
b5f2954d | 2051 | * @blkg: the blkg to associate |
b5f2954d | 2052 | * |
beea9da0 DZ |
2053 | * This tries to associate @bio with the specified @blkg. Association failure |
2054 | * is handled by walking up the blkg tree. Therefore, the blkg associated can | |
2055 | * be anything between @blkg and the root_blkg. This situation only happens | |
2056 | * when a cgroup is dying and then the remaining bios will spill to the closest | |
2057 | * alive blkg. | |
a7b39b4e | 2058 | * |
beea9da0 DZ |
2059 | * A reference will be taken on the @blkg and will be released when @bio is |
2060 | * freed. | |
a7b39b4e | 2061 | */ |
2268c0fe | 2062 | static void __bio_associate_blkg(struct bio *bio, struct blkcg_gq *blkg) |
a7b39b4e | 2063 | { |
2268c0fe DZ |
2064 | bio_disassociate_blkg(bio); |
2065 | ||
7754f669 | 2066 | bio->bi_blkg = blkg_tryget_closest(blkg); |
a7b39b4e DZF |
2067 | } |
2068 | ||
d459d853 | 2069 | /** |
fd42df30 | 2070 | * bio_associate_blkg_from_css - associate a bio with a specified css |
d459d853 | 2071 | * @bio: target bio |
fd42df30 | 2072 | * @css: target css |
d459d853 | 2073 | * |
fd42df30 | 2074 | * Associate @bio with the blkg found by combining the css's blkg and the |
fc5a828b DZ |
2075 | * request_queue of the @bio. This falls back to the queue's root_blkg if |
2076 | * the association fails with the css. | |
d459d853 | 2077 | */ |
fd42df30 DZ |
2078 | void bio_associate_blkg_from_css(struct bio *bio, |
2079 | struct cgroup_subsys_state *css) | |
d459d853 | 2080 | { |
fc5a828b DZ |
2081 | struct request_queue *q = bio->bi_disk->queue; |
2082 | struct blkcg_gq *blkg; | |
2083 | ||
2084 | rcu_read_lock(); | |
2085 | ||
2086 | if (!css || !css->parent) | |
2087 | blkg = q->root_blkg; | |
2088 | else | |
2089 | blkg = blkg_lookup_create(css_to_blkcg(css), q); | |
2090 | ||
2091 | __bio_associate_blkg(bio, blkg); | |
2092 | ||
2093 | rcu_read_unlock(); | |
d459d853 | 2094 | } |
fd42df30 | 2095 | EXPORT_SYMBOL_GPL(bio_associate_blkg_from_css); |
d459d853 | 2096 | |
6a7f6d86 | 2097 | #ifdef CONFIG_MEMCG |
852c788f | 2098 | /** |
6a7f6d86 | 2099 | * bio_associate_blkg_from_page - associate a bio with the page's blkg |
852c788f | 2100 | * @bio: target bio |
6a7f6d86 DZ |
2101 | * @page: the page to lookup the blkcg from |
2102 | * | |
2103 | * Associate @bio with the blkg from @page's owning memcg and the respective | |
fc5a828b DZ |
2104 | * request_queue. If cgroup_e_css returns %NULL, fall back to the queue's |
2105 | * root_blkg. | |
852c788f | 2106 | */ |
6a7f6d86 | 2107 | void bio_associate_blkg_from_page(struct bio *bio, struct page *page) |
852c788f | 2108 | { |
6a7f6d86 DZ |
2109 | struct cgroup_subsys_state *css; |
2110 | ||
6a7f6d86 DZ |
2111 | if (!page->mem_cgroup) |
2112 | return; | |
2113 | ||
fc5a828b DZ |
2114 | rcu_read_lock(); |
2115 | ||
2116 | css = cgroup_e_css(page->mem_cgroup->css.cgroup, &io_cgrp_subsys); | |
2117 | bio_associate_blkg_from_css(bio, css); | |
2118 | ||
2119 | rcu_read_unlock(); | |
6a7f6d86 DZ |
2120 | } |
2121 | #endif /* CONFIG_MEMCG */ | |
2122 | ||
2268c0fe DZ |
2123 | /** |
2124 | * bio_associate_blkg - associate a bio with a blkg | |
2125 | * @bio: target bio | |
2126 | * | |
2127 | * Associate @bio with the blkg found from the bio's css and request_queue. | |
2128 | * If one is not found, bio_lookup_blkg() creates the blkg. If a blkg is | |
2129 | * already associated, the css is reused and association redone as the | |
2130 | * request_queue may have changed. | |
2131 | */ | |
2132 | void bio_associate_blkg(struct bio *bio) | |
2133 | { | |
fc5a828b | 2134 | struct cgroup_subsys_state *css; |
2268c0fe DZ |
2135 | |
2136 | rcu_read_lock(); | |
2137 | ||
db6638d7 | 2138 | if (bio->bi_blkg) |
fc5a828b | 2139 | css = &bio_blkcg(bio)->css; |
db6638d7 | 2140 | else |
fc5a828b | 2141 | css = blkcg_css(); |
2268c0fe | 2142 | |
fc5a828b | 2143 | bio_associate_blkg_from_css(bio, css); |
2268c0fe DZ |
2144 | |
2145 | rcu_read_unlock(); | |
852c788f | 2146 | } |
5cdf2e3f | 2147 | EXPORT_SYMBOL_GPL(bio_associate_blkg); |
852c788f | 2148 | |
20bd723e | 2149 | /** |
db6638d7 | 2150 | * bio_clone_blkg_association - clone blkg association from src to dst bio |
20bd723e PV |
2151 | * @dst: destination bio |
2152 | * @src: source bio | |
2153 | */ | |
db6638d7 | 2154 | void bio_clone_blkg_association(struct bio *dst, struct bio *src) |
20bd723e | 2155 | { |
6ab21879 DZ |
2156 | rcu_read_lock(); |
2157 | ||
fc5a828b | 2158 | if (src->bi_blkg) |
2268c0fe | 2159 | __bio_associate_blkg(dst, src->bi_blkg); |
6ab21879 DZ |
2160 | |
2161 | rcu_read_unlock(); | |
20bd723e | 2162 | } |
db6638d7 | 2163 | EXPORT_SYMBOL_GPL(bio_clone_blkg_association); |
852c788f TH |
2164 | #endif /* CONFIG_BLK_CGROUP */ |
2165 | ||
1da177e4 LT |
2166 | static void __init biovec_init_slabs(void) |
2167 | { | |
2168 | int i; | |
2169 | ||
ed996a52 | 2170 | for (i = 0; i < BVEC_POOL_NR; i++) { |
1da177e4 LT |
2171 | int size; |
2172 | struct biovec_slab *bvs = bvec_slabs + i; | |
2173 | ||
a7fcd37c JA |
2174 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
2175 | bvs->slab = NULL; | |
2176 | continue; | |
2177 | } | |
a7fcd37c | 2178 | |
1da177e4 LT |
2179 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
2180 | bvs->slab = kmem_cache_create(bvs->name, size, 0, | |
20c2df83 | 2181 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
1da177e4 LT |
2182 | } |
2183 | } | |
2184 | ||
2185 | static int __init init_bio(void) | |
2186 | { | |
bb799ca0 JA |
2187 | bio_slab_max = 2; |
2188 | bio_slab_nr = 0; | |
6396bb22 KC |
2189 | bio_slabs = kcalloc(bio_slab_max, sizeof(struct bio_slab), |
2190 | GFP_KERNEL); | |
2b24e6f6 JT |
2191 | |
2192 | BUILD_BUG_ON(BIO_FLAG_LAST > BVEC_POOL_OFFSET); | |
2193 | ||
bb799ca0 JA |
2194 | if (!bio_slabs) |
2195 | panic("bio: can't allocate bios\n"); | |
1da177e4 | 2196 | |
7878cba9 | 2197 | bio_integrity_init(); |
1da177e4 LT |
2198 | biovec_init_slabs(); |
2199 | ||
f4f8154a | 2200 | if (bioset_init(&fs_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS)) |
1da177e4 LT |
2201 | panic("bio: can't allocate bios\n"); |
2202 | ||
f4f8154a | 2203 | if (bioset_integrity_create(&fs_bio_set, BIO_POOL_SIZE)) |
a91a2785 MP |
2204 | panic("bio: can't create integrity pool\n"); |
2205 | ||
1da177e4 LT |
2206 | return 0; |
2207 | } | |
1da177e4 | 2208 | subsys_initcall(init_bio); |