Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
Jens Axboe | 0fe2347 | 2006-09-04 15:41:16 +0200 | [diff] [blame] | 2 | * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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> |
Kent Overstreet | a27bb33 | 2013-05-07 16:19:08 -0700 | [diff] [blame] | 22 | #include <linux/uio.h> |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 23 | #include <linux/iocontext.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include <linux/slab.h> |
| 25 | #include <linux/init.h> |
| 26 | #include <linux/kernel.h> |
Paul Gortmaker | 630d9c4 | 2011-11-16 23:57:37 -0500 | [diff] [blame] | 27 | #include <linux/export.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 28 | #include <linux/mempool.h> |
| 29 | #include <linux/workqueue.h> |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 30 | #include <linux/cgroup.h> |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 31 | #include <scsi/sg.h> /* for struct sg_iovec */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 32 | |
Li Zefan | 5578213 | 2009-06-09 13:43:05 +0800 | [diff] [blame] | 33 | #include <trace/events/block.h> |
Ingo Molnar | 0bfc245 | 2008-11-26 11:59:56 +0100 | [diff] [blame] | 34 | |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 35 | /* |
| 36 | * Test patch to inline a certain number of bi_io_vec's inside the bio |
| 37 | * itself, to shrink a bio data allocation from two mempool calls to one |
| 38 | */ |
| 39 | #define BIO_INLINE_VECS 4 |
| 40 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 41 | /* |
| 42 | * if you change this list, also change bvec_alloc or things will |
| 43 | * break badly! cannot be bigger than what you can fit into an |
| 44 | * unsigned short |
| 45 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 46 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
Martin K. Petersen | df67714 | 2011-03-08 08:28:01 +0100 | [diff] [blame] | 47 | static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 48 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
| 49 | }; |
| 50 | #undef BV |
| 51 | |
| 52 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by |
| 54 | * IO code that does not need private memory pools. |
| 55 | */ |
Martin K. Petersen | 51d654e | 2008-06-17 18:59:56 +0200 | [diff] [blame] | 56 | struct bio_set *fs_bio_set; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 57 | EXPORT_SYMBOL(fs_bio_set); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 58 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 59 | /* |
| 60 | * Our slab pool management |
| 61 | */ |
| 62 | struct bio_slab { |
| 63 | struct kmem_cache *slab; |
| 64 | unsigned int slab_ref; |
| 65 | unsigned int slab_size; |
| 66 | char name[8]; |
| 67 | }; |
| 68 | static DEFINE_MUTEX(bio_slab_lock); |
| 69 | static struct bio_slab *bio_slabs; |
| 70 | static unsigned int bio_slab_nr, bio_slab_max; |
| 71 | |
| 72 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) |
| 73 | { |
| 74 | unsigned int sz = sizeof(struct bio) + extra_size; |
| 75 | struct kmem_cache *slab = NULL; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 76 | struct bio_slab *bslab, *new_bio_slabs; |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 77 | unsigned int new_bio_slab_max; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 78 | unsigned int i, entry = -1; |
| 79 | |
| 80 | mutex_lock(&bio_slab_lock); |
| 81 | |
| 82 | i = 0; |
| 83 | while (i < bio_slab_nr) { |
Thiago Farina | f06f135 | 2010-01-19 14:07:09 +0100 | [diff] [blame] | 84 | bslab = &bio_slabs[i]; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 85 | |
| 86 | if (!bslab->slab && entry == -1) |
| 87 | entry = i; |
| 88 | else if (bslab->slab_size == sz) { |
| 89 | slab = bslab->slab; |
| 90 | bslab->slab_ref++; |
| 91 | break; |
| 92 | } |
| 93 | i++; |
| 94 | } |
| 95 | |
| 96 | if (slab) |
| 97 | goto out_unlock; |
| 98 | |
| 99 | if (bio_slab_nr == bio_slab_max && entry == -1) { |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 100 | new_bio_slab_max = bio_slab_max << 1; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 101 | new_bio_slabs = krealloc(bio_slabs, |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 102 | new_bio_slab_max * sizeof(struct bio_slab), |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 103 | GFP_KERNEL); |
| 104 | if (!new_bio_slabs) |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 105 | goto out_unlock; |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 106 | bio_slab_max = new_bio_slab_max; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 107 | bio_slabs = new_bio_slabs; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 108 | } |
| 109 | if (entry == -1) |
| 110 | entry = bio_slab_nr++; |
| 111 | |
| 112 | bslab = &bio_slabs[entry]; |
| 113 | |
| 114 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); |
Mikulas Patocka | 6a24148 | 2014-03-28 15:51:55 -0400 | [diff] [blame] | 115 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
| 116 | SLAB_HWCACHE_ALIGN, NULL); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 117 | if (!slab) |
| 118 | goto out_unlock; |
| 119 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 120 | bslab->slab = slab; |
| 121 | bslab->slab_ref = 1; |
| 122 | bslab->slab_size = sz; |
| 123 | out_unlock: |
| 124 | mutex_unlock(&bio_slab_lock); |
| 125 | return slab; |
| 126 | } |
| 127 | |
| 128 | static void bio_put_slab(struct bio_set *bs) |
| 129 | { |
| 130 | struct bio_slab *bslab = NULL; |
| 131 | unsigned int i; |
| 132 | |
| 133 | mutex_lock(&bio_slab_lock); |
| 134 | |
| 135 | for (i = 0; i < bio_slab_nr; i++) { |
| 136 | if (bs->bio_slab == bio_slabs[i].slab) { |
| 137 | bslab = &bio_slabs[i]; |
| 138 | break; |
| 139 | } |
| 140 | } |
| 141 | |
| 142 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) |
| 143 | goto out; |
| 144 | |
| 145 | WARN_ON(!bslab->slab_ref); |
| 146 | |
| 147 | if (--bslab->slab_ref) |
| 148 | goto out; |
| 149 | |
| 150 | kmem_cache_destroy(bslab->slab); |
| 151 | bslab->slab = NULL; |
| 152 | |
| 153 | out: |
| 154 | mutex_unlock(&bio_slab_lock); |
| 155 | } |
| 156 | |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 157 | unsigned int bvec_nr_vecs(unsigned short idx) |
| 158 | { |
| 159 | return bvec_slabs[idx].nr_vecs; |
| 160 | } |
| 161 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 162 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx) |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 163 | { |
| 164 | BIO_BUG_ON(idx >= BIOVEC_NR_POOLS); |
| 165 | |
| 166 | if (idx == BIOVEC_MAX_IDX) |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 167 | mempool_free(bv, pool); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 168 | else { |
| 169 | struct biovec_slab *bvs = bvec_slabs + idx; |
| 170 | |
| 171 | kmem_cache_free(bvs->slab, bv); |
| 172 | } |
| 173 | } |
| 174 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 175 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
| 176 | mempool_t *pool) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 177 | { |
| 178 | struct bio_vec *bvl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 179 | |
| 180 | /* |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 181 | * see comment near bvec_array define! |
| 182 | */ |
| 183 | switch (nr) { |
| 184 | case 1: |
| 185 | *idx = 0; |
| 186 | break; |
| 187 | case 2 ... 4: |
| 188 | *idx = 1; |
| 189 | break; |
| 190 | case 5 ... 16: |
| 191 | *idx = 2; |
| 192 | break; |
| 193 | case 17 ... 64: |
| 194 | *idx = 3; |
| 195 | break; |
| 196 | case 65 ... 128: |
| 197 | *idx = 4; |
| 198 | break; |
| 199 | case 129 ... BIO_MAX_PAGES: |
| 200 | *idx = 5; |
| 201 | break; |
| 202 | default: |
| 203 | return NULL; |
| 204 | } |
| 205 | |
| 206 | /* |
| 207 | * idx now points to the pool we want to allocate from. only the |
| 208 | * 1-vec entry pool is mempool backed. |
| 209 | */ |
| 210 | if (*idx == BIOVEC_MAX_IDX) { |
| 211 | fallback: |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 212 | bvl = mempool_alloc(pool, gfp_mask); |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 213 | } else { |
| 214 | struct biovec_slab *bvs = bvec_slabs + *idx; |
| 215 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 216 | |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 217 | /* |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 218 | * Make this allocation restricted and don't dump info on |
| 219 | * allocation failures, since we'll fallback to the mempool |
| 220 | * in case of failure. |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 221 | */ |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 222 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
| 223 | |
| 224 | /* |
| 225 | * Try a slab allocation. If this fails and __GFP_WAIT |
| 226 | * is set, retry with the 1-entry mempool |
| 227 | */ |
| 228 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
| 229 | if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) { |
| 230 | *idx = BIOVEC_MAX_IDX; |
| 231 | goto fallback; |
| 232 | } |
| 233 | } |
| 234 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 235 | return bvl; |
| 236 | } |
| 237 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 238 | static void __bio_free(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 239 | { |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 240 | bio_disassociate_task(bio); |
Jens Axboe | 992c5dd | 2007-07-18 13:18:08 +0200 | [diff] [blame] | 241 | |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 242 | if (bio_integrity(bio)) |
Kent Overstreet | 1e2a410f | 2012-09-06 15:34:56 -0700 | [diff] [blame] | 243 | bio_integrity_free(bio); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 244 | } |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 245 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 246 | static void bio_free(struct bio *bio) |
| 247 | { |
| 248 | struct bio_set *bs = bio->bi_pool; |
| 249 | void *p; |
| 250 | |
| 251 | __bio_free(bio); |
| 252 | |
| 253 | if (bs) { |
Kent Overstreet | a38352e | 2012-05-25 13:03:11 -0700 | [diff] [blame] | 254 | if (bio_flagged(bio, BIO_OWNS_VEC)) |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 255 | bvec_free(bs->bvec_pool, bio->bi_io_vec, BIO_POOL_IDX(bio)); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 256 | |
| 257 | /* |
| 258 | * If we have front padding, adjust the bio pointer before freeing |
| 259 | */ |
| 260 | p = bio; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 261 | p -= bs->front_pad; |
| 262 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 263 | mempool_free(p, bs->bio_pool); |
| 264 | } else { |
| 265 | /* Bio was allocated by bio_kmalloc() */ |
| 266 | kfree(bio); |
| 267 | } |
Peter Osterlund | 3676347 | 2005-09-06 15:16:42 -0700 | [diff] [blame] | 268 | } |
| 269 | |
Arjan van de Ven | 858119e | 2006-01-14 13:20:43 -0800 | [diff] [blame] | 270 | void bio_init(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 271 | { |
Jens Axboe | 2b94de5 | 2007-07-18 13:14:03 +0200 | [diff] [blame] | 272 | memset(bio, 0, sizeof(*bio)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 273 | bio->bi_flags = 1 << BIO_UPTODATE; |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 274 | atomic_set(&bio->bi_remaining, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 275 | atomic_set(&bio->bi_cnt, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 276 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 277 | EXPORT_SYMBOL(bio_init); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 278 | |
| 279 | /** |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 280 | * bio_reset - reinitialize a bio |
| 281 | * @bio: bio to reset |
| 282 | * |
| 283 | * Description: |
| 284 | * After calling bio_reset(), @bio will be in the same state as a freshly |
| 285 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are |
| 286 | * preserved are the ones that are initialized by bio_alloc_bioset(). See |
| 287 | * comment in struct bio. |
| 288 | */ |
| 289 | void bio_reset(struct bio *bio) |
| 290 | { |
| 291 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); |
| 292 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 293 | __bio_free(bio); |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 294 | |
| 295 | memset(bio, 0, BIO_RESET_BYTES); |
| 296 | bio->bi_flags = flags|(1 << BIO_UPTODATE); |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 297 | atomic_set(&bio->bi_remaining, 1); |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 298 | } |
| 299 | EXPORT_SYMBOL(bio_reset); |
| 300 | |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 301 | static void bio_chain_endio(struct bio *bio, int error) |
| 302 | { |
| 303 | bio_endio(bio->bi_private, error); |
| 304 | bio_put(bio); |
| 305 | } |
| 306 | |
| 307 | /** |
| 308 | * bio_chain - chain bio completions |
Randy Dunlap | 1051a90 | 2014-04-20 16:03:31 -0700 | [diff] [blame] | 309 | * @bio: the target bio |
| 310 | * @parent: the @bio's parent bio |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 311 | * |
| 312 | * The caller won't have a bi_end_io called when @bio completes - instead, |
| 313 | * @parent's bi_end_io won't be called until both @parent and @bio have |
| 314 | * completed; the chained bio will also be freed when it completes. |
| 315 | * |
| 316 | * The caller must not set bi_private or bi_end_io in @bio. |
| 317 | */ |
| 318 | void bio_chain(struct bio *bio, struct bio *parent) |
| 319 | { |
| 320 | BUG_ON(bio->bi_private || bio->bi_end_io); |
| 321 | |
| 322 | bio->bi_private = parent; |
| 323 | bio->bi_end_io = bio_chain_endio; |
| 324 | atomic_inc(&parent->bi_remaining); |
| 325 | } |
| 326 | EXPORT_SYMBOL(bio_chain); |
| 327 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 328 | static void bio_alloc_rescue(struct work_struct *work) |
| 329 | { |
| 330 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); |
| 331 | struct bio *bio; |
| 332 | |
| 333 | while (1) { |
| 334 | spin_lock(&bs->rescue_lock); |
| 335 | bio = bio_list_pop(&bs->rescue_list); |
| 336 | spin_unlock(&bs->rescue_lock); |
| 337 | |
| 338 | if (!bio) |
| 339 | break; |
| 340 | |
| 341 | generic_make_request(bio); |
| 342 | } |
| 343 | } |
| 344 | |
| 345 | static void punt_bios_to_rescuer(struct bio_set *bs) |
| 346 | { |
| 347 | struct bio_list punt, nopunt; |
| 348 | struct bio *bio; |
| 349 | |
| 350 | /* |
| 351 | * In order to guarantee forward progress we must punt only bios that |
| 352 | * were allocated from this bio_set; otherwise, if there was a bio on |
| 353 | * there for a stacking driver higher up in the stack, processing it |
| 354 | * could require allocating bios from this bio_set, and doing that from |
| 355 | * our own rescuer would be bad. |
| 356 | * |
| 357 | * Since bio lists are singly linked, pop them all instead of trying to |
| 358 | * remove from the middle of the list: |
| 359 | */ |
| 360 | |
| 361 | bio_list_init(&punt); |
| 362 | bio_list_init(&nopunt); |
| 363 | |
| 364 | while ((bio = bio_list_pop(current->bio_list))) |
| 365 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
| 366 | |
| 367 | *current->bio_list = nopunt; |
| 368 | |
| 369 | spin_lock(&bs->rescue_lock); |
| 370 | bio_list_merge(&bs->rescue_list, &punt); |
| 371 | spin_unlock(&bs->rescue_lock); |
| 372 | |
| 373 | queue_work(bs->rescue_workqueue, &bs->rescue_work); |
| 374 | } |
| 375 | |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 376 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 377 | * bio_alloc_bioset - allocate a bio for I/O |
| 378 | * @gfp_mask: the GFP_ mask given to the slab allocator |
| 379 | * @nr_iovecs: number of iovecs to pre-allocate |
Jaak Ristioja | db18efa | 2010-01-15 12:05:07 +0200 | [diff] [blame] | 380 | * @bs: the bio_set to allocate from. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 381 | * |
| 382 | * Description: |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 383 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
| 384 | * backed by the @bs's mempool. |
| 385 | * |
| 386 | * When @bs is not NULL, if %__GFP_WAIT is set then bio_alloc will always be |
| 387 | * able to allocate a bio. This is due to the mempool guarantees. To make this |
| 388 | * work, callers must never allocate more than 1 bio at a time from this pool. |
| 389 | * Callers that need to allocate more than 1 bio must always submit the |
| 390 | * previously allocated bio for IO before attempting to allocate a new one. |
| 391 | * Failure to do so can cause deadlocks under memory pressure. |
| 392 | * |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 393 | * Note that when running under generic_make_request() (i.e. any block |
| 394 | * driver), bios are not submitted until after you return - see the code in |
| 395 | * generic_make_request() that converts recursion into iteration, to prevent |
| 396 | * stack overflows. |
| 397 | * |
| 398 | * This would normally mean allocating multiple bios under |
| 399 | * generic_make_request() would be susceptible to deadlocks, but we have |
| 400 | * deadlock avoidance code that resubmits any blocked bios from a rescuer |
| 401 | * thread. |
| 402 | * |
| 403 | * However, we do not guarantee forward progress for allocations from other |
| 404 | * mempools. Doing multiple allocations from the same mempool under |
| 405 | * generic_make_request() should be avoided - instead, use bio_set's front_pad |
| 406 | * for per bio allocations. |
| 407 | * |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 408 | * RETURNS: |
| 409 | * Pointer to new bio on success, NULL on failure. |
| 410 | */ |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 411 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 412 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 413 | gfp_t saved_gfp = gfp_mask; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 414 | unsigned front_pad; |
| 415 | unsigned inline_vecs; |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 416 | unsigned long idx = BIO_POOL_NONE; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 417 | struct bio_vec *bvl = NULL; |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 418 | struct bio *bio; |
| 419 | void *p; |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 420 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 421 | if (!bs) { |
| 422 | if (nr_iovecs > UIO_MAXIOV) |
| 423 | return NULL; |
| 424 | |
| 425 | p = kmalloc(sizeof(struct bio) + |
| 426 | nr_iovecs * sizeof(struct bio_vec), |
| 427 | gfp_mask); |
| 428 | front_pad = 0; |
| 429 | inline_vecs = nr_iovecs; |
| 430 | } else { |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame^] | 431 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
| 432 | if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0)) |
| 433 | return NULL; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 434 | /* |
| 435 | * generic_make_request() converts recursion to iteration; this |
| 436 | * means if we're running beneath it, any bios we allocate and |
| 437 | * submit will not be submitted (and thus freed) until after we |
| 438 | * return. |
| 439 | * |
| 440 | * This exposes us to a potential deadlock if we allocate |
| 441 | * multiple bios from the same bio_set() while running |
| 442 | * underneath generic_make_request(). If we were to allocate |
| 443 | * multiple bios (say a stacking block driver that was splitting |
| 444 | * bios), we would deadlock if we exhausted the mempool's |
| 445 | * reserve. |
| 446 | * |
| 447 | * We solve this, and guarantee forward progress, with a rescuer |
| 448 | * workqueue per bio_set. If we go to allocate and there are |
| 449 | * bios on current->bio_list, we first try the allocation |
| 450 | * without __GFP_WAIT; if that fails, we punt those bios we |
| 451 | * would be blocking to the rescuer workqueue before we retry |
| 452 | * with the original gfp_flags. |
| 453 | */ |
| 454 | |
| 455 | if (current->bio_list && !bio_list_empty(current->bio_list)) |
| 456 | gfp_mask &= ~__GFP_WAIT; |
| 457 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 458 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 459 | if (!p && gfp_mask != saved_gfp) { |
| 460 | punt_bios_to_rescuer(bs); |
| 461 | gfp_mask = saved_gfp; |
| 462 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
| 463 | } |
| 464 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 465 | front_pad = bs->front_pad; |
| 466 | inline_vecs = BIO_INLINE_VECS; |
| 467 | } |
| 468 | |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 469 | if (unlikely(!p)) |
| 470 | return NULL; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 471 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 472 | bio = p + front_pad; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 473 | bio_init(bio); |
| 474 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 475 | if (nr_iovecs > inline_vecs) { |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 476 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 477 | if (!bvl && gfp_mask != saved_gfp) { |
| 478 | punt_bios_to_rescuer(bs); |
| 479 | gfp_mask = saved_gfp; |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 480 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 481 | } |
| 482 | |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 483 | if (unlikely(!bvl)) |
| 484 | goto err_free; |
Kent Overstreet | a38352e | 2012-05-25 13:03:11 -0700 | [diff] [blame] | 485 | |
| 486 | bio->bi_flags |= 1 << BIO_OWNS_VEC; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 487 | } else if (nr_iovecs) { |
| 488 | bvl = bio->bi_inline_vecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 489 | } |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 490 | |
| 491 | bio->bi_pool = bs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 492 | bio->bi_flags |= idx << BIO_POOL_OFFSET; |
| 493 | bio->bi_max_vecs = nr_iovecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 494 | bio->bi_io_vec = bvl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 495 | return bio; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 496 | |
| 497 | err_free: |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 498 | mempool_free(p, bs->bio_pool); |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 499 | return NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 500 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 501 | EXPORT_SYMBOL(bio_alloc_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 502 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 503 | void zero_fill_bio(struct bio *bio) |
| 504 | { |
| 505 | unsigned long flags; |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 506 | struct bio_vec bv; |
| 507 | struct bvec_iter iter; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 508 | |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 509 | bio_for_each_segment(bv, bio, iter) { |
| 510 | char *data = bvec_kmap_irq(&bv, &flags); |
| 511 | memset(data, 0, bv.bv_len); |
| 512 | flush_dcache_page(bv.bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 513 | bvec_kunmap_irq(data, &flags); |
| 514 | } |
| 515 | } |
| 516 | EXPORT_SYMBOL(zero_fill_bio); |
| 517 | |
| 518 | /** |
| 519 | * bio_put - release a reference to a bio |
| 520 | * @bio: bio to release reference to |
| 521 | * |
| 522 | * Description: |
| 523 | * Put a reference to a &struct bio, either one you have gotten with |
Alberto Bertogli | ad0bf11 | 2009-11-02 11:39:22 +0100 | [diff] [blame] | 524 | * bio_alloc, bio_get or bio_clone. The last put of a bio will free it. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 525 | **/ |
| 526 | void bio_put(struct bio *bio) |
| 527 | { |
| 528 | BIO_BUG_ON(!atomic_read(&bio->bi_cnt)); |
| 529 | |
| 530 | /* |
| 531 | * last put frees it |
| 532 | */ |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 533 | if (atomic_dec_and_test(&bio->bi_cnt)) |
| 534 | bio_free(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 535 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 536 | EXPORT_SYMBOL(bio_put); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 537 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 538 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 539 | { |
| 540 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) |
| 541 | blk_recount_segments(q, bio); |
| 542 | |
| 543 | return bio->bi_phys_segments; |
| 544 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 545 | EXPORT_SYMBOL(bio_phys_segments); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 546 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 547 | /** |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 548 | * __bio_clone_fast - clone a bio that shares the original bio's biovec |
| 549 | * @bio: destination bio |
| 550 | * @bio_src: bio to clone |
| 551 | * |
| 552 | * Clone a &bio. Caller will own the returned bio, but not |
| 553 | * the actual data it points to. Reference count of returned |
| 554 | * bio will be one. |
| 555 | * |
| 556 | * Caller must ensure that @bio_src is not freed before @bio. |
| 557 | */ |
| 558 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) |
| 559 | { |
| 560 | BUG_ON(bio->bi_pool && BIO_POOL_IDX(bio) != BIO_POOL_NONE); |
| 561 | |
| 562 | /* |
| 563 | * most users will be overriding ->bi_bdev with a new target, |
| 564 | * so we don't set nor calculate new physical/hw segment counts here |
| 565 | */ |
| 566 | bio->bi_bdev = bio_src->bi_bdev; |
| 567 | bio->bi_flags |= 1 << BIO_CLONED; |
| 568 | bio->bi_rw = bio_src->bi_rw; |
| 569 | bio->bi_iter = bio_src->bi_iter; |
| 570 | bio->bi_io_vec = bio_src->bi_io_vec; |
| 571 | } |
| 572 | EXPORT_SYMBOL(__bio_clone_fast); |
| 573 | |
| 574 | /** |
| 575 | * bio_clone_fast - clone a bio that shares the original bio's biovec |
| 576 | * @bio: bio to clone |
| 577 | * @gfp_mask: allocation priority |
| 578 | * @bs: bio_set to allocate from |
| 579 | * |
| 580 | * Like __bio_clone_fast, only also allocates the returned bio |
| 581 | */ |
| 582 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) |
| 583 | { |
| 584 | struct bio *b; |
| 585 | |
| 586 | b = bio_alloc_bioset(gfp_mask, 0, bs); |
| 587 | if (!b) |
| 588 | return NULL; |
| 589 | |
| 590 | __bio_clone_fast(b, bio); |
| 591 | |
| 592 | if (bio_integrity(bio)) { |
| 593 | int ret; |
| 594 | |
| 595 | ret = bio_integrity_clone(b, bio, gfp_mask); |
| 596 | |
| 597 | if (ret < 0) { |
| 598 | bio_put(b); |
| 599 | return NULL; |
| 600 | } |
| 601 | } |
| 602 | |
| 603 | return b; |
| 604 | } |
| 605 | EXPORT_SYMBOL(bio_clone_fast); |
| 606 | |
| 607 | /** |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 608 | * bio_clone_bioset - clone a bio |
| 609 | * @bio_src: bio to clone |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 610 | * @gfp_mask: allocation priority |
Kent Overstreet | bf800ef | 2012-09-06 15:35:02 -0700 | [diff] [blame] | 611 | * @bs: bio_set to allocate from |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 612 | * |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 613 | * Clone bio. Caller will own the returned bio, but not the actual data it |
| 614 | * points to. Reference count of returned bio will be one. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 615 | */ |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 616 | struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask, |
Kent Overstreet | bf800ef | 2012-09-06 15:35:02 -0700 | [diff] [blame] | 617 | struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 618 | { |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 619 | struct bvec_iter iter; |
| 620 | struct bio_vec bv; |
| 621 | struct bio *bio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 622 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 623 | /* |
| 624 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from |
| 625 | * bio_src->bi_io_vec to bio->bi_io_vec. |
| 626 | * |
| 627 | * We can't do that anymore, because: |
| 628 | * |
| 629 | * - The point of cloning the biovec is to produce a bio with a biovec |
| 630 | * the caller can modify: bi_idx and bi_bvec_done should be 0. |
| 631 | * |
| 632 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if |
| 633 | * we tried to clone the whole thing bio_alloc_bioset() would fail. |
| 634 | * But the clone should succeed as long as the number of biovecs we |
| 635 | * actually need to allocate is fewer than BIO_MAX_PAGES. |
| 636 | * |
| 637 | * - Lastly, bi_vcnt should not be looked at or relied upon by code |
| 638 | * that does not own the bio - reason being drivers don't use it for |
| 639 | * iterating over the biovec anymore, so expecting it to be kept up |
| 640 | * to date (i.e. for clones that share the parent biovec) is just |
| 641 | * asking for trouble and would force extra work on |
| 642 | * __bio_clone_fast() anyways. |
| 643 | */ |
| 644 | |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 645 | bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs); |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 646 | if (!bio) |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 647 | return NULL; |
| 648 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 649 | bio->bi_bdev = bio_src->bi_bdev; |
| 650 | bio->bi_rw = bio_src->bi_rw; |
| 651 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; |
| 652 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 653 | |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 654 | if (bio->bi_rw & REQ_DISCARD) |
| 655 | goto integrity_clone; |
| 656 | |
| 657 | if (bio->bi_rw & REQ_WRITE_SAME) { |
| 658 | bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0]; |
| 659 | goto integrity_clone; |
| 660 | } |
| 661 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 662 | bio_for_each_segment(bv, bio_src, iter) |
| 663 | bio->bi_io_vec[bio->bi_vcnt++] = bv; |
| 664 | |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 665 | integrity_clone: |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 666 | if (bio_integrity(bio_src)) { |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 667 | int ret; |
| 668 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 669 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
Li Zefan | 059ea33 | 2009-03-09 10:42:45 +0100 | [diff] [blame] | 670 | if (ret < 0) { |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 671 | bio_put(bio); |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 672 | return NULL; |
Li Zefan | 059ea33 | 2009-03-09 10:42:45 +0100 | [diff] [blame] | 673 | } |
Peter Osterlund | 3676347 | 2005-09-06 15:16:42 -0700 | [diff] [blame] | 674 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 675 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 676 | return bio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 677 | } |
Kent Overstreet | bf800ef | 2012-09-06 15:35:02 -0700 | [diff] [blame] | 678 | EXPORT_SYMBOL(bio_clone_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 679 | |
| 680 | /** |
| 681 | * bio_get_nr_vecs - return approx number of vecs |
| 682 | * @bdev: I/O target |
| 683 | * |
| 684 | * Return the approximate number of pages we can send to this target. |
| 685 | * There's no guarantee that you will be able to fit this number of pages |
| 686 | * into a bio, it does not account for dynamic restrictions that vary |
| 687 | * on offset. |
| 688 | */ |
| 689 | int bio_get_nr_vecs(struct block_device *bdev) |
| 690 | { |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 691 | struct request_queue *q = bdev_get_queue(bdev); |
Bernd Schubert | f908ee9 | 2012-05-11 16:36:44 +0200 | [diff] [blame] | 692 | int nr_pages; |
| 693 | |
| 694 | nr_pages = min_t(unsigned, |
Kent Overstreet | 5abebfd | 2012-02-08 22:07:18 +0100 | [diff] [blame] | 695 | queue_max_segments(q), |
| 696 | queue_max_sectors(q) / (PAGE_SIZE >> 9) + 1); |
Bernd Schubert | f908ee9 | 2012-05-11 16:36:44 +0200 | [diff] [blame] | 697 | |
| 698 | return min_t(unsigned, nr_pages, BIO_MAX_PAGES); |
| 699 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 700 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 701 | EXPORT_SYMBOL(bio_get_nr_vecs); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 702 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 703 | static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 704 | *page, unsigned int len, unsigned int offset, |
Akinobu Mita | 34f2fd8 | 2013-11-18 22:11:42 +0900 | [diff] [blame] | 705 | unsigned int max_sectors) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 706 | { |
| 707 | int retried_segments = 0; |
| 708 | struct bio_vec *bvec; |
| 709 | |
| 710 | /* |
| 711 | * cloned bio must not modify vec list |
| 712 | */ |
| 713 | if (unlikely(bio_flagged(bio, BIO_CLONED))) |
| 714 | return 0; |
| 715 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 716 | if (((bio->bi_iter.bi_size + len) >> 9) > max_sectors) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 717 | return 0; |
| 718 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 719 | /* |
| 720 | * For filesystems with a blocksize smaller than the pagesize |
| 721 | * we will often be called with the same page as last time and |
| 722 | * a consecutive offset. Optimize this special case. |
| 723 | */ |
| 724 | if (bio->bi_vcnt > 0) { |
| 725 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 726 | |
| 727 | if (page == prev->bv_page && |
| 728 | offset == prev->bv_offset + prev->bv_len) { |
Dmitry Monakhov | 1d61658 | 2010-01-27 22:44:36 +0300 | [diff] [blame] | 729 | unsigned int prev_bv_len = prev->bv_len; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 730 | prev->bv_len += len; |
Alasdair G Kergon | cc371e6 | 2008-07-03 09:53:43 +0200 | [diff] [blame] | 731 | |
| 732 | if (q->merge_bvec_fn) { |
| 733 | struct bvec_merge_data bvm = { |
Dmitry Monakhov | 1d61658 | 2010-01-27 22:44:36 +0300 | [diff] [blame] | 734 | /* prev_bvec is already charged in |
| 735 | bi_size, discharge it in order to |
| 736 | simulate merging updated prev_bvec |
| 737 | as new bvec. */ |
Alasdair G Kergon | cc371e6 | 2008-07-03 09:53:43 +0200 | [diff] [blame] | 738 | .bi_bdev = bio->bi_bdev, |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 739 | .bi_sector = bio->bi_iter.bi_sector, |
| 740 | .bi_size = bio->bi_iter.bi_size - |
| 741 | prev_bv_len, |
Alasdair G Kergon | cc371e6 | 2008-07-03 09:53:43 +0200 | [diff] [blame] | 742 | .bi_rw = bio->bi_rw, |
| 743 | }; |
| 744 | |
Dmitry Monakhov | 8bf8c37 | 2010-03-03 06:28:06 +0300 | [diff] [blame] | 745 | if (q->merge_bvec_fn(q, &bvm, prev) < prev->bv_len) { |
Alasdair G Kergon | cc371e6 | 2008-07-03 09:53:43 +0200 | [diff] [blame] | 746 | prev->bv_len -= len; |
| 747 | return 0; |
| 748 | } |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 749 | } |
| 750 | |
| 751 | goto done; |
| 752 | } |
Jens Axboe | 66cb45a | 2014-06-24 16:22:24 -0600 | [diff] [blame] | 753 | |
| 754 | /* |
| 755 | * If the queue doesn't support SG gaps and adding this |
| 756 | * offset would create a gap, disallow it. |
| 757 | */ |
| 758 | if (q->queue_flags & (1 << QUEUE_FLAG_SG_GAPS) && |
| 759 | bvec_gap_to_prev(prev, offset)) |
| 760 | return 0; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 761 | } |
| 762 | |
| 763 | if (bio->bi_vcnt >= bio->bi_max_vecs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 764 | return 0; |
| 765 | |
| 766 | /* |
| 767 | * we might lose a segment or two here, but rather that than |
| 768 | * make this too complex. |
| 769 | */ |
| 770 | |
Martin K. Petersen | 8a78362 | 2010-02-26 00:20:39 -0500 | [diff] [blame] | 771 | while (bio->bi_phys_segments >= queue_max_segments(q)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 772 | |
| 773 | if (retried_segments) |
| 774 | return 0; |
| 775 | |
| 776 | retried_segments = 1; |
| 777 | blk_recount_segments(q, bio); |
| 778 | } |
| 779 | |
| 780 | /* |
| 781 | * setup the new entry, we might clear it again later if we |
| 782 | * cannot add the page |
| 783 | */ |
| 784 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; |
| 785 | bvec->bv_page = page; |
| 786 | bvec->bv_len = len; |
| 787 | bvec->bv_offset = offset; |
| 788 | |
| 789 | /* |
| 790 | * if queue has other restrictions (eg varying max sector size |
| 791 | * depending on offset), it can specify a merge_bvec_fn in the |
| 792 | * queue to get further control |
| 793 | */ |
| 794 | if (q->merge_bvec_fn) { |
Alasdair G Kergon | cc371e6 | 2008-07-03 09:53:43 +0200 | [diff] [blame] | 795 | struct bvec_merge_data bvm = { |
| 796 | .bi_bdev = bio->bi_bdev, |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 797 | .bi_sector = bio->bi_iter.bi_sector, |
| 798 | .bi_size = bio->bi_iter.bi_size, |
Alasdair G Kergon | cc371e6 | 2008-07-03 09:53:43 +0200 | [diff] [blame] | 799 | .bi_rw = bio->bi_rw, |
| 800 | }; |
| 801 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 802 | /* |
| 803 | * merge_bvec_fn() returns number of bytes it can accept |
| 804 | * at this offset |
| 805 | */ |
Dmitry Monakhov | 8bf8c37 | 2010-03-03 06:28:06 +0300 | [diff] [blame] | 806 | if (q->merge_bvec_fn(q, &bvm, bvec) < bvec->bv_len) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 807 | bvec->bv_page = NULL; |
| 808 | bvec->bv_len = 0; |
| 809 | bvec->bv_offset = 0; |
| 810 | return 0; |
| 811 | } |
| 812 | } |
| 813 | |
| 814 | /* If we may be able to merge these biovecs, force a recount */ |
Mikulas Patocka | b8b3e16 | 2008-08-15 10:15:19 +0200 | [diff] [blame] | 815 | if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 816 | bio->bi_flags &= ~(1 << BIO_SEG_VALID); |
| 817 | |
| 818 | bio->bi_vcnt++; |
| 819 | bio->bi_phys_segments++; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 820 | done: |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 821 | bio->bi_iter.bi_size += len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 822 | return len; |
| 823 | } |
| 824 | |
| 825 | /** |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 826 | * bio_add_pc_page - attempt to add page to bio |
Jens Axboe | fddfdea | 2006-01-31 15:24:34 +0100 | [diff] [blame] | 827 | * @q: the target queue |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 828 | * @bio: destination bio |
| 829 | * @page: page to add |
| 830 | * @len: vec entry length |
| 831 | * @offset: vec entry offset |
| 832 | * |
| 833 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
Andreas Gruenbacher | c642808 | 2011-05-27 14:52:09 +0200 | [diff] [blame] | 834 | * number of reasons, such as the bio being full or target block device |
| 835 | * limitations. The target block device must allow bio's up to PAGE_SIZE, |
| 836 | * so it is always possible to add a single page to an empty bio. |
| 837 | * |
| 838 | * This should only be used by REQ_PC bios. |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 839 | */ |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 840 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page, |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 841 | unsigned int len, unsigned int offset) |
| 842 | { |
Martin K. Petersen | ae03bf6 | 2009-05-22 17:17:50 -0400 | [diff] [blame] | 843 | return __bio_add_page(q, bio, page, len, offset, |
| 844 | queue_max_hw_sectors(q)); |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 845 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 846 | EXPORT_SYMBOL(bio_add_pc_page); |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 847 | |
| 848 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 849 | * bio_add_page - attempt to add page to bio |
| 850 | * @bio: destination bio |
| 851 | * @page: page to add |
| 852 | * @len: vec entry length |
| 853 | * @offset: vec entry offset |
| 854 | * |
| 855 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
Andreas Gruenbacher | c642808 | 2011-05-27 14:52:09 +0200 | [diff] [blame] | 856 | * number of reasons, such as the bio being full or target block device |
| 857 | * limitations. The target block device must allow bio's up to PAGE_SIZE, |
| 858 | * so it is always possible to add a single page to an empty bio. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 859 | */ |
| 860 | int bio_add_page(struct bio *bio, struct page *page, unsigned int len, |
| 861 | unsigned int offset) |
| 862 | { |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 863 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
Jens Axboe | 58a4915 | 2014-06-10 12:53:56 -0600 | [diff] [blame] | 864 | unsigned int max_sectors; |
Jens Axboe | 762380a | 2014-06-05 13:38:39 -0600 | [diff] [blame] | 865 | |
Jens Axboe | 58a4915 | 2014-06-10 12:53:56 -0600 | [diff] [blame] | 866 | max_sectors = blk_max_size_offset(q, bio->bi_iter.bi_sector); |
| 867 | if ((max_sectors < (len >> 9)) && !bio->bi_iter.bi_size) |
| 868 | max_sectors = len >> 9; |
| 869 | |
| 870 | return __bio_add_page(q, bio, page, len, offset, max_sectors); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 871 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 872 | EXPORT_SYMBOL(bio_add_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 873 | |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 874 | struct submit_bio_ret { |
| 875 | struct completion event; |
| 876 | int error; |
| 877 | }; |
| 878 | |
| 879 | static void submit_bio_wait_endio(struct bio *bio, int error) |
| 880 | { |
| 881 | struct submit_bio_ret *ret = bio->bi_private; |
| 882 | |
| 883 | ret->error = error; |
| 884 | complete(&ret->event); |
| 885 | } |
| 886 | |
| 887 | /** |
| 888 | * submit_bio_wait - submit a bio, and wait until it completes |
| 889 | * @rw: whether to %READ or %WRITE, or maybe to %READA (read ahead) |
| 890 | * @bio: The &struct bio which describes the I/O |
| 891 | * |
| 892 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from |
| 893 | * bio_endio() on failure. |
| 894 | */ |
| 895 | int submit_bio_wait(int rw, struct bio *bio) |
| 896 | { |
| 897 | struct submit_bio_ret ret; |
| 898 | |
| 899 | rw |= REQ_SYNC; |
| 900 | init_completion(&ret.event); |
| 901 | bio->bi_private = &ret; |
| 902 | bio->bi_end_io = submit_bio_wait_endio; |
| 903 | submit_bio(rw, bio); |
| 904 | wait_for_completion(&ret.event); |
| 905 | |
| 906 | return ret.error; |
| 907 | } |
| 908 | EXPORT_SYMBOL(submit_bio_wait); |
| 909 | |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 910 | /** |
| 911 | * bio_advance - increment/complete a bio by some number of bytes |
| 912 | * @bio: bio to advance |
| 913 | * @bytes: number of bytes to complete |
| 914 | * |
| 915 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to |
| 916 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will |
| 917 | * be updated on the last bvec as well. |
| 918 | * |
| 919 | * @bio will then represent the remaining, uncompleted portion of the io. |
| 920 | */ |
| 921 | void bio_advance(struct bio *bio, unsigned bytes) |
| 922 | { |
| 923 | if (bio_integrity(bio)) |
| 924 | bio_integrity_advance(bio, bytes); |
| 925 | |
Kent Overstreet | 4550dd6 | 2013-08-07 14:26:21 -0700 | [diff] [blame] | 926 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 927 | } |
| 928 | EXPORT_SYMBOL(bio_advance); |
| 929 | |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 930 | /** |
Kent Overstreet | a078760 | 2012-09-10 14:03:28 -0700 | [diff] [blame] | 931 | * bio_alloc_pages - allocates a single page for each bvec in a bio |
| 932 | * @bio: bio to allocate pages for |
| 933 | * @gfp_mask: flags for allocation |
| 934 | * |
| 935 | * Allocates pages up to @bio->bi_vcnt. |
| 936 | * |
| 937 | * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are |
| 938 | * freed. |
| 939 | */ |
| 940 | int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask) |
| 941 | { |
| 942 | int i; |
| 943 | struct bio_vec *bv; |
| 944 | |
| 945 | bio_for_each_segment_all(bv, bio, i) { |
| 946 | bv->bv_page = alloc_page(gfp_mask); |
| 947 | if (!bv->bv_page) { |
| 948 | while (--bv >= bio->bi_io_vec) |
| 949 | __free_page(bv->bv_page); |
| 950 | return -ENOMEM; |
| 951 | } |
| 952 | } |
| 953 | |
| 954 | return 0; |
| 955 | } |
| 956 | EXPORT_SYMBOL(bio_alloc_pages); |
| 957 | |
| 958 | /** |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 959 | * bio_copy_data - copy contents of data buffers from one chain of bios to |
| 960 | * another |
| 961 | * @src: source bio list |
| 962 | * @dst: destination bio list |
| 963 | * |
| 964 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats |
| 965 | * @src and @dst as linked lists of bios. |
| 966 | * |
| 967 | * Stops when it reaches the end of either @src or @dst - that is, copies |
| 968 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). |
| 969 | */ |
| 970 | void bio_copy_data(struct bio *dst, struct bio *src) |
| 971 | { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 972 | struct bvec_iter src_iter, dst_iter; |
| 973 | struct bio_vec src_bv, dst_bv; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 974 | void *src_p, *dst_p; |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 975 | unsigned bytes; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 976 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 977 | src_iter = src->bi_iter; |
| 978 | dst_iter = dst->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 979 | |
| 980 | while (1) { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 981 | if (!src_iter.bi_size) { |
| 982 | src = src->bi_next; |
| 983 | if (!src) |
| 984 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 985 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 986 | src_iter = src->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 987 | } |
| 988 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 989 | if (!dst_iter.bi_size) { |
| 990 | dst = dst->bi_next; |
| 991 | if (!dst) |
| 992 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 993 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 994 | dst_iter = dst->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 995 | } |
| 996 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 997 | src_bv = bio_iter_iovec(src, src_iter); |
| 998 | dst_bv = bio_iter_iovec(dst, dst_iter); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 999 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1000 | bytes = min(src_bv.bv_len, dst_bv.bv_len); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1001 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1002 | src_p = kmap_atomic(src_bv.bv_page); |
| 1003 | dst_p = kmap_atomic(dst_bv.bv_page); |
| 1004 | |
| 1005 | memcpy(dst_p + dst_bv.bv_offset, |
| 1006 | src_p + src_bv.bv_offset, |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1007 | bytes); |
| 1008 | |
| 1009 | kunmap_atomic(dst_p); |
| 1010 | kunmap_atomic(src_p); |
| 1011 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1012 | bio_advance_iter(src, &src_iter, bytes); |
| 1013 | bio_advance_iter(dst, &dst_iter, bytes); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1014 | } |
| 1015 | } |
| 1016 | EXPORT_SYMBOL(bio_copy_data); |
| 1017 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1018 | struct bio_map_data { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1019 | int nr_sgvecs; |
| 1020 | int is_our_pages; |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1021 | struct sg_iovec sgvecs[]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1022 | }; |
| 1023 | |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1024 | static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio, |
Al Viro | 86d564c | 2014-02-08 20:42:52 -0500 | [diff] [blame] | 1025 | const struct sg_iovec *iov, int iov_count, |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1026 | int is_our_pages) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1027 | { |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1028 | memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count); |
| 1029 | bmd->nr_sgvecs = iov_count; |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1030 | bmd->is_our_pages = is_our_pages; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1031 | bio->bi_private = bmd; |
| 1032 | } |
| 1033 | |
Fabian Frederick | 7410b3c | 2014-04-22 15:09:07 -0600 | [diff] [blame] | 1034 | static struct bio_map_data *bio_alloc_map_data(unsigned int iov_count, |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1035 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1036 | { |
Jens Axboe | f3f63c1 | 2010-10-29 11:46:56 -0600 | [diff] [blame] | 1037 | if (iov_count > UIO_MAXIOV) |
| 1038 | return NULL; |
| 1039 | |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1040 | return kmalloc(sizeof(struct bio_map_data) + |
| 1041 | sizeof(struct sg_iovec) * iov_count, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1042 | } |
| 1043 | |
Al Viro | 86d564c | 2014-02-08 20:42:52 -0500 | [diff] [blame] | 1044 | static int __bio_copy_iov(struct bio *bio, const struct sg_iovec *iov, int iov_count, |
FUJITA Tomonori | ecb554a | 2009-07-09 14:46:53 +0200 | [diff] [blame] | 1045 | int to_user, int from_user, int do_free_page) |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1046 | { |
| 1047 | int ret = 0, i; |
| 1048 | struct bio_vec *bvec; |
| 1049 | int iov_idx = 0; |
| 1050 | unsigned int iov_off = 0; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1051 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1052 | bio_for_each_segment_all(bvec, bio, i) { |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1053 | char *bv_addr = page_address(bvec->bv_page); |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1054 | unsigned int bv_len = bvec->bv_len; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1055 | |
| 1056 | while (bv_len && iov_idx < iov_count) { |
| 1057 | unsigned int bytes; |
Michal Simek | 0e0c621 | 2009-06-10 12:57:07 -0700 | [diff] [blame] | 1058 | char __user *iov_addr; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1059 | |
| 1060 | bytes = min_t(unsigned int, |
| 1061 | iov[iov_idx].iov_len - iov_off, bv_len); |
| 1062 | iov_addr = iov[iov_idx].iov_base + iov_off; |
| 1063 | |
| 1064 | if (!ret) { |
FUJITA Tomonori | ecb554a | 2009-07-09 14:46:53 +0200 | [diff] [blame] | 1065 | if (to_user) |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1066 | ret = copy_to_user(iov_addr, bv_addr, |
| 1067 | bytes); |
| 1068 | |
FUJITA Tomonori | ecb554a | 2009-07-09 14:46:53 +0200 | [diff] [blame] | 1069 | if (from_user) |
| 1070 | ret = copy_from_user(bv_addr, iov_addr, |
| 1071 | bytes); |
| 1072 | |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1073 | if (ret) |
| 1074 | ret = -EFAULT; |
| 1075 | } |
| 1076 | |
| 1077 | bv_len -= bytes; |
| 1078 | bv_addr += bytes; |
| 1079 | iov_addr += bytes; |
| 1080 | iov_off += bytes; |
| 1081 | |
| 1082 | if (iov[iov_idx].iov_len == iov_off) { |
| 1083 | iov_idx++; |
| 1084 | iov_off = 0; |
| 1085 | } |
| 1086 | } |
| 1087 | |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1088 | if (do_free_page) |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1089 | __free_page(bvec->bv_page); |
| 1090 | } |
| 1091 | |
| 1092 | return ret; |
| 1093 | } |
| 1094 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1095 | /** |
| 1096 | * bio_uncopy_user - finish previously mapped bio |
| 1097 | * @bio: bio being terminated |
| 1098 | * |
| 1099 | * Free pages allocated from bio_copy_user() and write back data |
| 1100 | * to user space in case of a read. |
| 1101 | */ |
| 1102 | int bio_uncopy_user(struct bio *bio) |
| 1103 | { |
| 1104 | struct bio_map_data *bmd = bio->bi_private; |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1105 | struct bio_vec *bvec; |
| 1106 | int ret = 0, i; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1107 | |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1108 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
| 1109 | /* |
| 1110 | * if we're in a workqueue, the request is orphaned, so |
| 1111 | * don't copy into a random user address space, just free. |
| 1112 | */ |
| 1113 | if (current->mm) |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1114 | ret = __bio_copy_iov(bio, bmd->sgvecs, bmd->nr_sgvecs, |
| 1115 | bio_data_dir(bio) == READ, |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1116 | 0, bmd->is_our_pages); |
| 1117 | else if (bmd->is_our_pages) |
| 1118 | bio_for_each_segment_all(bvec, bio, i) |
| 1119 | __free_page(bvec->bv_page); |
| 1120 | } |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1121 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1122 | bio_put(bio); |
| 1123 | return ret; |
| 1124 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1125 | EXPORT_SYMBOL(bio_uncopy_user); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1126 | |
| 1127 | /** |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1128 | * bio_copy_user_iov - copy user data to bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1129 | * @q: destination block queue |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1130 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1131 | * @iov: the iovec. |
| 1132 | * @iov_count: number of elements in the iovec |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1133 | * @write_to_vm: bool indicating writing to pages or not |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1134 | * @gfp_mask: memory allocation flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1135 | * |
| 1136 | * Prepares and returns a bio for indirect user io, bouncing data |
| 1137 | * to/from kernel pages as necessary. Must be paired with |
| 1138 | * call bio_uncopy_user() on io completion. |
| 1139 | */ |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1140 | struct bio *bio_copy_user_iov(struct request_queue *q, |
| 1141 | struct rq_map_data *map_data, |
Al Viro | 86d564c | 2014-02-08 20:42:52 -0500 | [diff] [blame] | 1142 | const struct sg_iovec *iov, int iov_count, |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1143 | int write_to_vm, gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1144 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1145 | struct bio_map_data *bmd; |
| 1146 | struct bio_vec *bvec; |
| 1147 | struct page *page; |
| 1148 | struct bio *bio; |
| 1149 | int i, ret; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1150 | int nr_pages = 0; |
| 1151 | unsigned int len = 0; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1152 | unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1153 | |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1154 | for (i = 0; i < iov_count; i++) { |
| 1155 | unsigned long uaddr; |
| 1156 | unsigned long end; |
| 1157 | unsigned long start; |
| 1158 | |
| 1159 | uaddr = (unsigned long)iov[i].iov_base; |
| 1160 | end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1161 | start = uaddr >> PAGE_SHIFT; |
| 1162 | |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1163 | /* |
| 1164 | * Overflow, abort |
| 1165 | */ |
| 1166 | if (end < start) |
| 1167 | return ERR_PTR(-EINVAL); |
| 1168 | |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1169 | nr_pages += end - start; |
| 1170 | len += iov[i].iov_len; |
| 1171 | } |
| 1172 | |
FUJITA Tomonori | 6983872 | 2009-04-28 20:24:29 +0200 | [diff] [blame] | 1173 | if (offset) |
| 1174 | nr_pages++; |
| 1175 | |
Fabian Frederick | 7410b3c | 2014-04-22 15:09:07 -0600 | [diff] [blame] | 1176 | bmd = bio_alloc_map_data(iov_count, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1177 | if (!bmd) |
| 1178 | return ERR_PTR(-ENOMEM); |
| 1179 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1180 | ret = -ENOMEM; |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1181 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1182 | if (!bio) |
| 1183 | goto out_bmd; |
| 1184 | |
Christoph Hellwig | 7b6d91d | 2010-08-07 18:20:39 +0200 | [diff] [blame] | 1185 | if (!write_to_vm) |
| 1186 | bio->bi_rw |= REQ_WRITE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1187 | |
| 1188 | ret = 0; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1189 | |
| 1190 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1191 | nr_pages = 1 << map_data->page_order; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1192 | i = map_data->offset / PAGE_SIZE; |
| 1193 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1194 | while (len) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1195 | unsigned int bytes = PAGE_SIZE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1196 | |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1197 | bytes -= offset; |
| 1198 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1199 | if (bytes > len) |
| 1200 | bytes = len; |
| 1201 | |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1202 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1203 | if (i == map_data->nr_entries * nr_pages) { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1204 | ret = -ENOMEM; |
| 1205 | break; |
| 1206 | } |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1207 | |
| 1208 | page = map_data->pages[i / nr_pages]; |
| 1209 | page += (i % nr_pages); |
| 1210 | |
| 1211 | i++; |
| 1212 | } else { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1213 | page = alloc_page(q->bounce_gfp | gfp_mask); |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1214 | if (!page) { |
| 1215 | ret = -ENOMEM; |
| 1216 | break; |
| 1217 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1218 | } |
| 1219 | |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1220 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1221 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1222 | |
| 1223 | len -= bytes; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1224 | offset = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1225 | } |
| 1226 | |
| 1227 | if (ret) |
| 1228 | goto cleanup; |
| 1229 | |
| 1230 | /* |
| 1231 | * success |
| 1232 | */ |
FUJITA Tomonori | ecb554a | 2009-07-09 14:46:53 +0200 | [diff] [blame] | 1233 | if ((!write_to_vm && (!map_data || !map_data->null_mapped)) || |
| 1234 | (map_data && map_data->from_user)) { |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1235 | ret = __bio_copy_iov(bio, iov, iov_count, 0, 1, 0); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1236 | if (ret) |
| 1237 | goto cleanup; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1238 | } |
| 1239 | |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1240 | bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1241 | return bio; |
| 1242 | cleanup: |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1243 | if (!map_data) |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1244 | bio_for_each_segment_all(bvec, bio, i) |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1245 | __free_page(bvec->bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1246 | |
| 1247 | bio_put(bio); |
| 1248 | out_bmd: |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1249 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1250 | return ERR_PTR(ret); |
| 1251 | } |
| 1252 | |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1253 | /** |
| 1254 | * bio_copy_user - copy user data to bio |
| 1255 | * @q: destination block queue |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1256 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1257 | * @uaddr: start of user address |
| 1258 | * @len: length in bytes |
| 1259 | * @write_to_vm: bool indicating writing to pages or not |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1260 | * @gfp_mask: memory allocation flags |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1261 | * |
| 1262 | * Prepares and returns a bio for indirect user io, bouncing data |
| 1263 | * to/from kernel pages as necessary. Must be paired with |
| 1264 | * call bio_uncopy_user() on io completion. |
| 1265 | */ |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1266 | struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data, |
| 1267 | unsigned long uaddr, unsigned int len, |
| 1268 | int write_to_vm, gfp_t gfp_mask) |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1269 | { |
| 1270 | struct sg_iovec iov; |
| 1271 | |
| 1272 | iov.iov_base = (void __user *)uaddr; |
| 1273 | iov.iov_len = len; |
| 1274 | |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1275 | return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1276 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1277 | EXPORT_SYMBOL(bio_copy_user); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1278 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1279 | static struct bio *__bio_map_user_iov(struct request_queue *q, |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1280 | struct block_device *bdev, |
Al Viro | 86d564c | 2014-02-08 20:42:52 -0500 | [diff] [blame] | 1281 | const struct sg_iovec *iov, int iov_count, |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1282 | int write_to_vm, gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1283 | { |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1284 | int i, j; |
| 1285 | int nr_pages = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1286 | struct page **pages; |
| 1287 | struct bio *bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1288 | int cur_page = 0; |
| 1289 | int ret, offset; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1290 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1291 | for (i = 0; i < iov_count; i++) { |
| 1292 | unsigned long uaddr = (unsigned long)iov[i].iov_base; |
| 1293 | unsigned long len = iov[i].iov_len; |
| 1294 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1295 | unsigned long start = uaddr >> PAGE_SHIFT; |
| 1296 | |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1297 | /* |
| 1298 | * Overflow, abort |
| 1299 | */ |
| 1300 | if (end < start) |
| 1301 | return ERR_PTR(-EINVAL); |
| 1302 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1303 | nr_pages += end - start; |
| 1304 | /* |
Mike Christie | ad2d722 | 2006-12-01 10:40:20 +0100 | [diff] [blame] | 1305 | * buffer must be aligned to at least hardsector size for now |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1306 | */ |
Mike Christie | ad2d722 | 2006-12-01 10:40:20 +0100 | [diff] [blame] | 1307 | if (uaddr & queue_dma_alignment(q)) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1308 | return ERR_PTR(-EINVAL); |
| 1309 | } |
| 1310 | |
| 1311 | if (!nr_pages) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1312 | return ERR_PTR(-EINVAL); |
| 1313 | |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1314 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1315 | if (!bio) |
| 1316 | return ERR_PTR(-ENOMEM); |
| 1317 | |
| 1318 | ret = -ENOMEM; |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1319 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1320 | if (!pages) |
| 1321 | goto out; |
| 1322 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1323 | for (i = 0; i < iov_count; i++) { |
| 1324 | unsigned long uaddr = (unsigned long)iov[i].iov_base; |
| 1325 | unsigned long len = iov[i].iov_len; |
| 1326 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1327 | unsigned long start = uaddr >> PAGE_SHIFT; |
| 1328 | const int local_nr_pages = end - start; |
| 1329 | const int page_limit = cur_page + local_nr_pages; |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1330 | |
Nick Piggin | f5dd33c | 2008-07-25 19:45:25 -0700 | [diff] [blame] | 1331 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
| 1332 | write_to_vm, &pages[cur_page]); |
Jens Axboe | 9917215 | 2006-06-16 13:02:29 +0200 | [diff] [blame] | 1333 | if (ret < local_nr_pages) { |
| 1334 | ret = -EFAULT; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1335 | goto out_unmap; |
Jens Axboe | 9917215 | 2006-06-16 13:02:29 +0200 | [diff] [blame] | 1336 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1337 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1338 | offset = uaddr & ~PAGE_MASK; |
| 1339 | for (j = cur_page; j < page_limit; j++) { |
| 1340 | unsigned int bytes = PAGE_SIZE - offset; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1341 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1342 | if (len <= 0) |
| 1343 | break; |
| 1344 | |
| 1345 | if (bytes > len) |
| 1346 | bytes = len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1347 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1348 | /* |
| 1349 | * sorry... |
| 1350 | */ |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 1351 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
| 1352 | bytes) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1353 | break; |
| 1354 | |
| 1355 | len -= bytes; |
| 1356 | offset = 0; |
| 1357 | } |
| 1358 | |
| 1359 | cur_page = j; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1360 | /* |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1361 | * release the pages we didn't map into the bio, if any |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1362 | */ |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1363 | while (j < page_limit) |
| 1364 | page_cache_release(pages[j++]); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1365 | } |
| 1366 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1367 | kfree(pages); |
| 1368 | |
| 1369 | /* |
| 1370 | * set data direction, and check if mapped pages need bouncing |
| 1371 | */ |
| 1372 | if (!write_to_vm) |
Christoph Hellwig | 7b6d91d | 2010-08-07 18:20:39 +0200 | [diff] [blame] | 1373 | bio->bi_rw |= REQ_WRITE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1374 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1375 | bio->bi_bdev = bdev; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1376 | bio->bi_flags |= (1 << BIO_USER_MAPPED); |
| 1377 | return bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1378 | |
| 1379 | out_unmap: |
| 1380 | for (i = 0; i < nr_pages; i++) { |
| 1381 | if(!pages[i]) |
| 1382 | break; |
| 1383 | page_cache_release(pages[i]); |
| 1384 | } |
| 1385 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1386 | kfree(pages); |
| 1387 | bio_put(bio); |
| 1388 | return ERR_PTR(ret); |
| 1389 | } |
| 1390 | |
| 1391 | /** |
| 1392 | * bio_map_user - map user address into bio |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1393 | * @q: the struct request_queue for the bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1394 | * @bdev: destination block device |
| 1395 | * @uaddr: start of user address |
| 1396 | * @len: length in bytes |
| 1397 | * @write_to_vm: bool indicating writing to pages or not |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1398 | * @gfp_mask: memory allocation flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1399 | * |
| 1400 | * Map the user space address into a bio suitable for io to a block |
| 1401 | * device. Returns an error pointer in case of error. |
| 1402 | */ |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1403 | struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev, |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1404 | unsigned long uaddr, unsigned int len, int write_to_vm, |
| 1405 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1406 | { |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1407 | struct sg_iovec iov; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1408 | |
viro@ZenIV.linux.org.uk | 3f70353 | 2005-09-09 16:53:56 +0100 | [diff] [blame] | 1409 | iov.iov_base = (void __user *)uaddr; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1410 | iov.iov_len = len; |
| 1411 | |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1412 | return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask); |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1413 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1414 | EXPORT_SYMBOL(bio_map_user); |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1415 | |
| 1416 | /** |
| 1417 | * bio_map_user_iov - map user sg_iovec table into bio |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1418 | * @q: the struct request_queue for the bio |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1419 | * @bdev: destination block device |
| 1420 | * @iov: the iovec. |
| 1421 | * @iov_count: number of elements in the iovec |
| 1422 | * @write_to_vm: bool indicating writing to pages or not |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1423 | * @gfp_mask: memory allocation flags |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1424 | * |
| 1425 | * Map the user space address into a bio suitable for io to a block |
| 1426 | * device. Returns an error pointer in case of error. |
| 1427 | */ |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1428 | struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev, |
Al Viro | 86d564c | 2014-02-08 20:42:52 -0500 | [diff] [blame] | 1429 | const struct sg_iovec *iov, int iov_count, |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1430 | int write_to_vm, gfp_t gfp_mask) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1431 | { |
| 1432 | struct bio *bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1433 | |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1434 | bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm, |
| 1435 | gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1436 | if (IS_ERR(bio)) |
| 1437 | return bio; |
| 1438 | |
| 1439 | /* |
| 1440 | * subtle -- if __bio_map_user() ended up bouncing a bio, |
| 1441 | * it would normally disappear when its bi_end_io is run. |
| 1442 | * however, we need it for the unmap, so grab an extra |
| 1443 | * reference to it |
| 1444 | */ |
| 1445 | bio_get(bio); |
| 1446 | |
Mike Christie | 0e75f90 | 2006-12-01 10:40:55 +0100 | [diff] [blame] | 1447 | return bio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1448 | } |
| 1449 | |
| 1450 | static void __bio_unmap_user(struct bio *bio) |
| 1451 | { |
| 1452 | struct bio_vec *bvec; |
| 1453 | int i; |
| 1454 | |
| 1455 | /* |
| 1456 | * make sure we dirty pages we wrote to |
| 1457 | */ |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1458 | bio_for_each_segment_all(bvec, bio, i) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1459 | if (bio_data_dir(bio) == READ) |
| 1460 | set_page_dirty_lock(bvec->bv_page); |
| 1461 | |
| 1462 | page_cache_release(bvec->bv_page); |
| 1463 | } |
| 1464 | |
| 1465 | bio_put(bio); |
| 1466 | } |
| 1467 | |
| 1468 | /** |
| 1469 | * bio_unmap_user - unmap a bio |
| 1470 | * @bio: the bio being unmapped |
| 1471 | * |
| 1472 | * Unmap a bio previously mapped by bio_map_user(). Must be called with |
| 1473 | * a process context. |
| 1474 | * |
| 1475 | * bio_unmap_user() may sleep. |
| 1476 | */ |
| 1477 | void bio_unmap_user(struct bio *bio) |
| 1478 | { |
| 1479 | __bio_unmap_user(bio); |
| 1480 | bio_put(bio); |
| 1481 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1482 | EXPORT_SYMBOL(bio_unmap_user); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1483 | |
NeilBrown | 6712ecf | 2007-09-27 12:47:43 +0200 | [diff] [blame] | 1484 | static void bio_map_kern_endio(struct bio *bio, int err) |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1485 | { |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1486 | bio_put(bio); |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1487 | } |
| 1488 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1489 | static struct bio *__bio_map_kern(struct request_queue *q, void *data, |
Al Viro | 27496a8 | 2005-10-21 03:20:48 -0400 | [diff] [blame] | 1490 | unsigned int len, gfp_t gfp_mask) |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 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 | |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1499 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 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 | |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 1513 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
| 1514 | offset) < bytes) |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1515 | break; |
| 1516 | |
| 1517 | data += bytes; |
| 1518 | len -= bytes; |
| 1519 | offset = 0; |
| 1520 | } |
| 1521 | |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1522 | bio->bi_end_io = bio_map_kern_endio; |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1523 | return bio; |
| 1524 | } |
| 1525 | |
| 1526 | /** |
| 1527 | * bio_map_kern - map kernel address into bio |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1528 | * @q: the struct request_queue for the bio |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1529 | * @data: pointer to buffer to map |
| 1530 | * @len: length in bytes |
| 1531 | * @gfp_mask: allocation flags for bio allocation |
| 1532 | * |
| 1533 | * Map the kernel address into a bio suitable for io to a block |
| 1534 | * device. Returns an error pointer in case of error. |
| 1535 | */ |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 1536 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
Al Viro | 27496a8 | 2005-10-21 03:20:48 -0400 | [diff] [blame] | 1537 | gfp_t gfp_mask) |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1538 | { |
| 1539 | struct bio *bio; |
| 1540 | |
| 1541 | bio = __bio_map_kern(q, data, len, gfp_mask); |
| 1542 | if (IS_ERR(bio)) |
| 1543 | return bio; |
| 1544 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1545 | if (bio->bi_iter.bi_size == len) |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1546 | return bio; |
| 1547 | |
| 1548 | /* |
| 1549 | * Don't support partial mappings. |
| 1550 | */ |
| 1551 | bio_put(bio); |
| 1552 | return ERR_PTR(-EINVAL); |
| 1553 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1554 | EXPORT_SYMBOL(bio_map_kern); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1555 | |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1556 | static void bio_copy_kern_endio(struct bio *bio, int err) |
| 1557 | { |
| 1558 | struct bio_vec *bvec; |
| 1559 | const int read = bio_data_dir(bio) == READ; |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1560 | struct bio_map_data *bmd = bio->bi_private; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1561 | int i; |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1562 | char *p = bmd->sgvecs[0].iov_base; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1563 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1564 | bio_for_each_segment_all(bvec, bio, i) { |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1565 | char *addr = page_address(bvec->bv_page); |
| 1566 | |
Tejun Heo | 4fc981e | 2009-05-19 18:33:06 +0900 | [diff] [blame] | 1567 | if (read) |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1568 | memcpy(p, addr, bvec->bv_len); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1569 | |
| 1570 | __free_page(bvec->bv_page); |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1571 | p += bvec->bv_len; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1572 | } |
| 1573 | |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1574 | kfree(bmd); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1575 | bio_put(bio); |
| 1576 | } |
| 1577 | |
| 1578 | /** |
| 1579 | * bio_copy_kern - copy kernel address into bio |
| 1580 | * @q: the struct request_queue for the bio |
| 1581 | * @data: pointer to buffer to copy |
| 1582 | * @len: length in bytes |
| 1583 | * @gfp_mask: allocation flags for bio and page allocation |
Randy Dunlap | ffee025 | 2008-04-30 09:08:54 +0200 | [diff] [blame] | 1584 | * @reading: data direction is READ |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1585 | * |
| 1586 | * copy the kernel address into a bio suitable for io to a block |
| 1587 | * device. Returns an error pointer in case of error. |
| 1588 | */ |
| 1589 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, |
| 1590 | gfp_t gfp_mask, int reading) |
| 1591 | { |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1592 | struct bio *bio; |
| 1593 | struct bio_vec *bvec; |
FUJITA Tomonori | 4d8ab62 | 2008-08-28 15:05:57 +0900 | [diff] [blame] | 1594 | int i; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1595 | |
FUJITA Tomonori | 4d8ab62 | 2008-08-28 15:05:57 +0900 | [diff] [blame] | 1596 | bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask); |
| 1597 | if (IS_ERR(bio)) |
| 1598 | return bio; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1599 | |
| 1600 | if (!reading) { |
| 1601 | void *p = data; |
| 1602 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1603 | bio_for_each_segment_all(bvec, bio, i) { |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1604 | char *addr = page_address(bvec->bv_page); |
| 1605 | |
| 1606 | memcpy(addr, p, bvec->bv_len); |
| 1607 | p += bvec->bv_len; |
| 1608 | } |
| 1609 | } |
| 1610 | |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1611 | bio->bi_end_io = bio_copy_kern_endio; |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1612 | |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1613 | return bio; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1614 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1615 | EXPORT_SYMBOL(bio_copy_kern); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1616 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1617 | /* |
| 1618 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions |
| 1619 | * for performing direct-IO in BIOs. |
| 1620 | * |
| 1621 | * The problem is that we cannot run set_page_dirty() from interrupt context |
| 1622 | * because the required locks are not interrupt-safe. So what we can do is to |
| 1623 | * mark the pages dirty _before_ performing IO. And in interrupt context, |
| 1624 | * check that the pages are still dirty. If so, fine. If not, redirty them |
| 1625 | * in process context. |
| 1626 | * |
| 1627 | * We special-case compound pages here: normally this means reads into hugetlb |
| 1628 | * pages. The logic in here doesn't really work right for compound pages |
| 1629 | * because the VM does not uniformly chase down the head page in all cases. |
| 1630 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't |
| 1631 | * handle them at all. So we skip compound pages here at an early stage. |
| 1632 | * |
| 1633 | * Note that this code is very hard to test under normal circumstances because |
| 1634 | * direct-io pins the pages with get_user_pages(). This makes |
| 1635 | * is_page_cache_freeable return false, and the VM will not clean the pages. |
Artem Bityutskiy | 0d5c3eb | 2012-07-25 18:12:08 +0300 | [diff] [blame] | 1636 | * But other code (eg, flusher threads) could clean the pages if they are mapped |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1637 | * pagecache. |
| 1638 | * |
| 1639 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the |
| 1640 | * deferred bio dirtying paths. |
| 1641 | */ |
| 1642 | |
| 1643 | /* |
| 1644 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. |
| 1645 | */ |
| 1646 | void bio_set_pages_dirty(struct bio *bio) |
| 1647 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1648 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1649 | int i; |
| 1650 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1651 | bio_for_each_segment_all(bvec, bio, i) { |
| 1652 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1653 | |
| 1654 | if (page && !PageCompound(page)) |
| 1655 | set_page_dirty_lock(page); |
| 1656 | } |
| 1657 | } |
| 1658 | |
Adrian Bunk | 86b6c7a | 2008-02-18 13:48:32 +0100 | [diff] [blame] | 1659 | static void bio_release_pages(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1660 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1661 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1662 | int i; |
| 1663 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1664 | bio_for_each_segment_all(bvec, bio, i) { |
| 1665 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1666 | |
| 1667 | if (page) |
| 1668 | put_page(page); |
| 1669 | } |
| 1670 | } |
| 1671 | |
| 1672 | /* |
| 1673 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. |
| 1674 | * If they are, then fine. If, however, some pages are clean then they must |
| 1675 | * have been written out during the direct-IO read. So we take another ref on |
| 1676 | * the BIO and the offending pages and re-dirty the pages in process context. |
| 1677 | * |
| 1678 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from |
| 1679 | * here on. It will run one page_cache_release() against each page and will |
| 1680 | * run one bio_put() against the BIO. |
| 1681 | */ |
| 1682 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1683 | static void bio_dirty_fn(struct work_struct *work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1684 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1685 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1686 | static DEFINE_SPINLOCK(bio_dirty_lock); |
| 1687 | static struct bio *bio_dirty_list; |
| 1688 | |
| 1689 | /* |
| 1690 | * This runs in process context |
| 1691 | */ |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1692 | static void bio_dirty_fn(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1693 | { |
| 1694 | unsigned long flags; |
| 1695 | struct bio *bio; |
| 1696 | |
| 1697 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1698 | bio = bio_dirty_list; |
| 1699 | bio_dirty_list = NULL; |
| 1700 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1701 | |
| 1702 | while (bio) { |
| 1703 | struct bio *next = bio->bi_private; |
| 1704 | |
| 1705 | bio_set_pages_dirty(bio); |
| 1706 | bio_release_pages(bio); |
| 1707 | bio_put(bio); |
| 1708 | bio = next; |
| 1709 | } |
| 1710 | } |
| 1711 | |
| 1712 | void bio_check_pages_dirty(struct bio *bio) |
| 1713 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1714 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1715 | int nr_clean_pages = 0; |
| 1716 | int i; |
| 1717 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1718 | bio_for_each_segment_all(bvec, bio, i) { |
| 1719 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1720 | |
| 1721 | if (PageDirty(page) || PageCompound(page)) { |
| 1722 | page_cache_release(page); |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1723 | bvec->bv_page = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1724 | } else { |
| 1725 | nr_clean_pages++; |
| 1726 | } |
| 1727 | } |
| 1728 | |
| 1729 | if (nr_clean_pages) { |
| 1730 | unsigned long flags; |
| 1731 | |
| 1732 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1733 | bio->bi_private = bio_dirty_list; |
| 1734 | bio_dirty_list = bio; |
| 1735 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1736 | schedule_work(&bio_dirty_work); |
| 1737 | } else { |
| 1738 | bio_put(bio); |
| 1739 | } |
| 1740 | } |
| 1741 | |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1742 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
| 1743 | void bio_flush_dcache_pages(struct bio *bi) |
| 1744 | { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1745 | struct bio_vec bvec; |
| 1746 | struct bvec_iter iter; |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1747 | |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1748 | bio_for_each_segment(bvec, bi, iter) |
| 1749 | flush_dcache_page(bvec.bv_page); |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1750 | } |
| 1751 | EXPORT_SYMBOL(bio_flush_dcache_pages); |
| 1752 | #endif |
| 1753 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1754 | /** |
| 1755 | * bio_endio - end I/O on a bio |
| 1756 | * @bio: bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1757 | * @error: error, if any |
| 1758 | * |
| 1759 | * Description: |
NeilBrown | 6712ecf | 2007-09-27 12:47:43 +0200 | [diff] [blame] | 1760 | * bio_endio() will end I/O on the whole bio. bio_endio() is the |
NeilBrown | 5bb23a6 | 2007-09-27 12:46:13 +0200 | [diff] [blame] | 1761 | * preferred way to end I/O on a bio, it takes care of clearing |
| 1762 | * BIO_UPTODATE on error. @error is 0 on success, and and one of the |
| 1763 | * established -Exxxx (-EIO, for instance) error values in case |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 1764 | * something went wrong. No one should call bi_end_io() directly on a |
NeilBrown | 5bb23a6 | 2007-09-27 12:46:13 +0200 | [diff] [blame] | 1765 | * bio unless they own it and thus know that it has an end_io |
| 1766 | * function. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1767 | **/ |
NeilBrown | 6712ecf | 2007-09-27 12:47:43 +0200 | [diff] [blame] | 1768 | void bio_endio(struct bio *bio, int error) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1769 | { |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1770 | while (bio) { |
| 1771 | BUG_ON(atomic_read(&bio->bi_remaining) <= 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1772 | |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1773 | if (error) |
| 1774 | clear_bit(BIO_UPTODATE, &bio->bi_flags); |
| 1775 | else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) |
| 1776 | error = -EIO; |
| 1777 | |
| 1778 | if (!atomic_dec_and_test(&bio->bi_remaining)) |
| 1779 | return; |
| 1780 | |
| 1781 | /* |
| 1782 | * Need to have a real endio function for chained bios, |
| 1783 | * otherwise various corner cases will break (like stacking |
| 1784 | * block devices that save/restore bi_end_io) - however, we want |
| 1785 | * to avoid unbounded recursion and blowing the stack. Tail call |
| 1786 | * optimization would handle this, but compiling with frame |
| 1787 | * pointers also disables gcc's sibling call optimization. |
| 1788 | */ |
| 1789 | if (bio->bi_end_io == bio_chain_endio) { |
| 1790 | struct bio *parent = bio->bi_private; |
| 1791 | bio_put(bio); |
| 1792 | bio = parent; |
| 1793 | } else { |
| 1794 | if (bio->bi_end_io) |
| 1795 | bio->bi_end_io(bio, error); |
| 1796 | bio = NULL; |
| 1797 | } |
| 1798 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1799 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1800 | EXPORT_SYMBOL(bio_endio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1801 | |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1802 | /** |
| 1803 | * bio_endio_nodec - end I/O on a bio, without decrementing bi_remaining |
| 1804 | * @bio: bio |
| 1805 | * @error: error, if any |
| 1806 | * |
| 1807 | * For code that has saved and restored bi_end_io; thing hard before using this |
| 1808 | * function, probably you should've cloned the entire bio. |
| 1809 | **/ |
| 1810 | void bio_endio_nodec(struct bio *bio, int error) |
| 1811 | { |
| 1812 | atomic_inc(&bio->bi_remaining); |
| 1813 | bio_endio(bio, error); |
| 1814 | } |
| 1815 | EXPORT_SYMBOL(bio_endio_nodec); |
| 1816 | |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1817 | /** |
| 1818 | * bio_split - split a bio |
| 1819 | * @bio: bio to split |
| 1820 | * @sectors: number of sectors to split from the front of @bio |
| 1821 | * @gfp: gfp mask |
| 1822 | * @bs: bio set to allocate from |
| 1823 | * |
| 1824 | * Allocates and returns a new bio which represents @sectors from the start of |
| 1825 | * @bio, and updates @bio to represent the remaining sectors. |
| 1826 | * |
| 1827 | * The newly allocated bio will point to @bio's bi_io_vec; it is the caller's |
| 1828 | * responsibility to ensure that @bio is not freed before the split. |
| 1829 | */ |
| 1830 | struct bio *bio_split(struct bio *bio, int sectors, |
| 1831 | gfp_t gfp, struct bio_set *bs) |
| 1832 | { |
| 1833 | struct bio *split = NULL; |
| 1834 | |
| 1835 | BUG_ON(sectors <= 0); |
| 1836 | BUG_ON(sectors >= bio_sectors(bio)); |
| 1837 | |
| 1838 | split = bio_clone_fast(bio, gfp, bs); |
| 1839 | if (!split) |
| 1840 | return NULL; |
| 1841 | |
| 1842 | split->bi_iter.bi_size = sectors << 9; |
| 1843 | |
| 1844 | if (bio_integrity(split)) |
| 1845 | bio_integrity_trim(split, 0, sectors); |
| 1846 | |
| 1847 | bio_advance(bio, split->bi_iter.bi_size); |
| 1848 | |
| 1849 | return split; |
| 1850 | } |
| 1851 | EXPORT_SYMBOL(bio_split); |
| 1852 | |
Martin K. Petersen | ad3316b | 2008-10-01 22:42:53 -0400 | [diff] [blame] | 1853 | /** |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1854 | * bio_trim - trim a bio |
| 1855 | * @bio: bio to trim |
| 1856 | * @offset: number of sectors to trim from the front of @bio |
| 1857 | * @size: size we want to trim @bio to, in sectors |
| 1858 | */ |
| 1859 | void bio_trim(struct bio *bio, int offset, int size) |
| 1860 | { |
| 1861 | /* 'bio' is a cloned bio which we need to trim to match |
| 1862 | * the given offset and size. |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1863 | */ |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1864 | |
| 1865 | size <<= 9; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1866 | if (offset == 0 && size == bio->bi_iter.bi_size) |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1867 | return; |
| 1868 | |
| 1869 | clear_bit(BIO_SEG_VALID, &bio->bi_flags); |
| 1870 | |
| 1871 | bio_advance(bio, offset << 9); |
| 1872 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1873 | bio->bi_iter.bi_size = size; |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1874 | } |
| 1875 | EXPORT_SYMBOL_GPL(bio_trim); |
| 1876 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1877 | /* |
| 1878 | * create memory pools for biovec's in a bio_set. |
| 1879 | * use the global biovec slabs created for general use. |
| 1880 | */ |
Fabian Frederick | a6c39cb4f | 2014-04-22 15:09:05 -0600 | [diff] [blame] | 1881 | mempool_t *biovec_create_pool(int pool_entries) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1882 | { |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 1883 | struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1884 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1885 | return mempool_create_slab_pool(pool_entries, bp->slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1886 | } |
| 1887 | |
| 1888 | void bioset_free(struct bio_set *bs) |
| 1889 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1890 | if (bs->rescue_workqueue) |
| 1891 | destroy_workqueue(bs->rescue_workqueue); |
| 1892 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1893 | if (bs->bio_pool) |
| 1894 | mempool_destroy(bs->bio_pool); |
| 1895 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1896 | if (bs->bvec_pool) |
| 1897 | mempool_destroy(bs->bvec_pool); |
| 1898 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 1899 | bioset_integrity_free(bs); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1900 | bio_put_slab(bs); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1901 | |
| 1902 | kfree(bs); |
| 1903 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1904 | EXPORT_SYMBOL(bioset_free); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1905 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame^] | 1906 | static struct bio_set *__bioset_create(unsigned int pool_size, |
| 1907 | unsigned int front_pad, |
| 1908 | bool create_bvec_pool) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1909 | { |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 1910 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1911 | struct bio_set *bs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1912 | |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1913 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1914 | if (!bs) |
| 1915 | return NULL; |
| 1916 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1917 | bs->front_pad = front_pad; |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1918 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1919 | spin_lock_init(&bs->rescue_lock); |
| 1920 | bio_list_init(&bs->rescue_list); |
| 1921 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); |
| 1922 | |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 1923 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1924 | if (!bs->bio_slab) { |
| 1925 | kfree(bs); |
| 1926 | return NULL; |
| 1927 | } |
| 1928 | |
| 1929 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1930 | if (!bs->bio_pool) |
| 1931 | goto bad; |
| 1932 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame^] | 1933 | if (create_bvec_pool) { |
| 1934 | bs->bvec_pool = biovec_create_pool(pool_size); |
| 1935 | if (!bs->bvec_pool) |
| 1936 | goto bad; |
| 1937 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1938 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1939 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); |
| 1940 | if (!bs->rescue_workqueue) |
| 1941 | goto bad; |
| 1942 | |
| 1943 | return bs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1944 | bad: |
| 1945 | bioset_free(bs); |
| 1946 | return NULL; |
| 1947 | } |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame^] | 1948 | |
| 1949 | /** |
| 1950 | * bioset_create - Create a bio_set |
| 1951 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
| 1952 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1953 | * |
| 1954 | * Description: |
| 1955 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller |
| 1956 | * to ask for a number of bytes to be allocated in front of the bio. |
| 1957 | * Front pad allocation is useful for embedding the bio inside |
| 1958 | * another structure, to avoid allocating extra data to go with the bio. |
| 1959 | * Note that the bio must be embedded at the END of that structure always, |
| 1960 | * or things will break badly. |
| 1961 | */ |
| 1962 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) |
| 1963 | { |
| 1964 | return __bioset_create(pool_size, front_pad, true); |
| 1965 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1966 | EXPORT_SYMBOL(bioset_create); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1967 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame^] | 1968 | /** |
| 1969 | * bioset_create_nobvec - Create a bio_set without bio_vec mempool |
| 1970 | * @pool_size: Number of bio to cache in the mempool |
| 1971 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1972 | * |
| 1973 | * Description: |
| 1974 | * Same functionality as bioset_create() except that mempool is not |
| 1975 | * created for bio_vecs. Saving some memory for bio_clone_fast() users. |
| 1976 | */ |
| 1977 | struct bio_set *bioset_create_nobvec(unsigned int pool_size, unsigned int front_pad) |
| 1978 | { |
| 1979 | return __bioset_create(pool_size, front_pad, false); |
| 1980 | } |
| 1981 | EXPORT_SYMBOL(bioset_create_nobvec); |
| 1982 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1983 | #ifdef CONFIG_BLK_CGROUP |
| 1984 | /** |
| 1985 | * bio_associate_current - associate a bio with %current |
| 1986 | * @bio: target bio |
| 1987 | * |
| 1988 | * Associate @bio with %current if it hasn't been associated yet. Block |
| 1989 | * layer will treat @bio as if it were issued by %current no matter which |
| 1990 | * task actually issues it. |
| 1991 | * |
| 1992 | * This function takes an extra reference of @task's io_context and blkcg |
| 1993 | * which will be put when @bio is released. The caller must own @bio, |
| 1994 | * ensure %current->io_context exists, and is responsible for synchronizing |
| 1995 | * calls to this function. |
| 1996 | */ |
| 1997 | int bio_associate_current(struct bio *bio) |
| 1998 | { |
| 1999 | struct io_context *ioc; |
| 2000 | struct cgroup_subsys_state *css; |
| 2001 | |
| 2002 | if (bio->bi_ioc) |
| 2003 | return -EBUSY; |
| 2004 | |
| 2005 | ioc = current->io_context; |
| 2006 | if (!ioc) |
| 2007 | return -ENOENT; |
| 2008 | |
| 2009 | /* acquire active ref on @ioc and associate */ |
| 2010 | get_io_context_active(ioc); |
| 2011 | bio->bi_ioc = ioc; |
| 2012 | |
| 2013 | /* associate blkcg if exists */ |
| 2014 | rcu_read_lock(); |
Tejun Heo | 073219e | 2014-02-08 10:36:58 -0500 | [diff] [blame] | 2015 | css = task_css(current, blkio_cgrp_id); |
Tejun Heo | ec903c0 | 2014-05-13 12:11:01 -0400 | [diff] [blame] | 2016 | if (css && css_tryget_online(css)) |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2017 | bio->bi_css = css; |
| 2018 | rcu_read_unlock(); |
| 2019 | |
| 2020 | return 0; |
| 2021 | } |
| 2022 | |
| 2023 | /** |
| 2024 | * bio_disassociate_task - undo bio_associate_current() |
| 2025 | * @bio: target bio |
| 2026 | */ |
| 2027 | void bio_disassociate_task(struct bio *bio) |
| 2028 | { |
| 2029 | if (bio->bi_ioc) { |
| 2030 | put_io_context(bio->bi_ioc); |
| 2031 | bio->bi_ioc = NULL; |
| 2032 | } |
| 2033 | if (bio->bi_css) { |
| 2034 | css_put(bio->bi_css); |
| 2035 | bio->bi_css = NULL; |
| 2036 | } |
| 2037 | } |
| 2038 | |
| 2039 | #endif /* CONFIG_BLK_CGROUP */ |
| 2040 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2041 | static void __init biovec_init_slabs(void) |
| 2042 | { |
| 2043 | int i; |
| 2044 | |
| 2045 | for (i = 0; i < BIOVEC_NR_POOLS; i++) { |
| 2046 | int size; |
| 2047 | struct biovec_slab *bvs = bvec_slabs + i; |
| 2048 | |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2049 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
| 2050 | bvs->slab = NULL; |
| 2051 | continue; |
| 2052 | } |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2053 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2054 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
| 2055 | bvs->slab = kmem_cache_create(bvs->name, size, 0, |
Paul Mundt | 20c2df8 | 2007-07-20 10:11:58 +0900 | [diff] [blame] | 2056 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2057 | } |
| 2058 | } |
| 2059 | |
| 2060 | static int __init init_bio(void) |
| 2061 | { |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2062 | bio_slab_max = 2; |
| 2063 | bio_slab_nr = 0; |
| 2064 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); |
| 2065 | if (!bio_slabs) |
| 2066 | panic("bio: can't allocate bios\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2067 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 2068 | bio_integrity_init(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2069 | biovec_init_slabs(); |
| 2070 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2071 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2072 | if (!fs_bio_set) |
| 2073 | panic("bio: can't allocate bios\n"); |
| 2074 | |
Martin K. Petersen | a91a278 | 2011-03-17 11:11:05 +0100 | [diff] [blame] | 2075 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
| 2076 | panic("bio: can't create integrity pool\n"); |
| 2077 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2078 | return 0; |
| 2079 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2080 | subsys_initcall(init_bio); |