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> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | |
Li Zefan | 5578213 | 2009-06-09 13:43:05 +0800 | [diff] [blame] | 32 | #include <trace/events/block.h> |
Ingo Molnar | 0bfc245 | 2008-11-26 11:59:56 +0100 | [diff] [blame] | 33 | |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 34 | /* |
| 35 | * Test patch to inline a certain number of bi_io_vec's inside the bio |
| 36 | * itself, to shrink a bio data allocation from two mempool calls to one |
| 37 | */ |
| 38 | #define BIO_INLINE_VECS 4 |
| 39 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | /* |
| 41 | * if you change this list, also change bvec_alloc or things will |
| 42 | * break badly! cannot be bigger than what you can fit into an |
| 43 | * unsigned short |
| 44 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 45 | #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 46 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 47 | BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), |
| 48 | }; |
| 49 | #undef BV |
| 50 | |
| 51 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 52 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by |
| 53 | * IO code that does not need private memory pools. |
| 54 | */ |
Martin K. Petersen | 51d654e | 2008-06-17 18:59:56 +0200 | [diff] [blame] | 55 | struct bio_set *fs_bio_set; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 56 | EXPORT_SYMBOL(fs_bio_set); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 57 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 58 | /* |
| 59 | * Our slab pool management |
| 60 | */ |
| 61 | struct bio_slab { |
| 62 | struct kmem_cache *slab; |
| 63 | unsigned int slab_ref; |
| 64 | unsigned int slab_size; |
| 65 | char name[8]; |
| 66 | }; |
| 67 | static DEFINE_MUTEX(bio_slab_lock); |
| 68 | static struct bio_slab *bio_slabs; |
| 69 | static unsigned int bio_slab_nr, bio_slab_max; |
| 70 | |
| 71 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) |
| 72 | { |
| 73 | unsigned int sz = sizeof(struct bio) + extra_size; |
| 74 | struct kmem_cache *slab = NULL; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 75 | struct bio_slab *bslab, *new_bio_slabs; |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 76 | unsigned int new_bio_slab_max; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 77 | unsigned int i, entry = -1; |
| 78 | |
| 79 | mutex_lock(&bio_slab_lock); |
| 80 | |
| 81 | i = 0; |
| 82 | while (i < bio_slab_nr) { |
Thiago Farina | f06f135 | 2010-01-19 14:07:09 +0100 | [diff] [blame] | 83 | bslab = &bio_slabs[i]; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 84 | |
| 85 | if (!bslab->slab && entry == -1) |
| 86 | entry = i; |
| 87 | else if (bslab->slab_size == sz) { |
| 88 | slab = bslab->slab; |
| 89 | bslab->slab_ref++; |
| 90 | break; |
| 91 | } |
| 92 | i++; |
| 93 | } |
| 94 | |
| 95 | if (slab) |
| 96 | goto out_unlock; |
| 97 | |
| 98 | if (bio_slab_nr == bio_slab_max && entry == -1) { |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 99 | new_bio_slab_max = bio_slab_max << 1; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 100 | new_bio_slabs = krealloc(bio_slabs, |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 101 | new_bio_slab_max * sizeof(struct bio_slab), |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 102 | GFP_KERNEL); |
| 103 | if (!new_bio_slabs) |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 104 | goto out_unlock; |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 105 | bio_slab_max = new_bio_slab_max; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 106 | bio_slabs = new_bio_slabs; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 107 | } |
| 108 | if (entry == -1) |
| 109 | entry = bio_slab_nr++; |
| 110 | |
| 111 | bslab = &bio_slabs[entry]; |
| 112 | |
| 113 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); |
Mikulas Patocka | 6a24148 | 2014-03-28 15:51:55 -0400 | [diff] [blame] | 114 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
| 115 | SLAB_HWCACHE_ALIGN, NULL); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 116 | if (!slab) |
| 117 | goto out_unlock; |
| 118 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 119 | bslab->slab = slab; |
| 120 | bslab->slab_ref = 1; |
| 121 | bslab->slab_size = sz; |
| 122 | out_unlock: |
| 123 | mutex_unlock(&bio_slab_lock); |
| 124 | return slab; |
| 125 | } |
| 126 | |
| 127 | static void bio_put_slab(struct bio_set *bs) |
| 128 | { |
| 129 | struct bio_slab *bslab = NULL; |
| 130 | unsigned int i; |
| 131 | |
| 132 | mutex_lock(&bio_slab_lock); |
| 133 | |
| 134 | for (i = 0; i < bio_slab_nr; i++) { |
| 135 | if (bs->bio_slab == bio_slabs[i].slab) { |
| 136 | bslab = &bio_slabs[i]; |
| 137 | break; |
| 138 | } |
| 139 | } |
| 140 | |
| 141 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) |
| 142 | goto out; |
| 143 | |
| 144 | WARN_ON(!bslab->slab_ref); |
| 145 | |
| 146 | if (--bslab->slab_ref) |
| 147 | goto out; |
| 148 | |
| 149 | kmem_cache_destroy(bslab->slab); |
| 150 | bslab->slab = NULL; |
| 151 | |
| 152 | out: |
| 153 | mutex_unlock(&bio_slab_lock); |
| 154 | } |
| 155 | |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 156 | unsigned int bvec_nr_vecs(unsigned short idx) |
| 157 | { |
| 158 | return bvec_slabs[idx].nr_vecs; |
| 159 | } |
| 160 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 161 | 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] | 162 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 163 | if (!idx) |
| 164 | return; |
| 165 | idx--; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 166 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 167 | BIO_BUG_ON(idx >= BVEC_POOL_NR); |
| 168 | |
| 169 | if (idx == BVEC_POOL_MAX) { |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 170 | mempool_free(bv, pool); |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 171 | } else { |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 172 | struct biovec_slab *bvs = bvec_slabs + idx; |
| 173 | |
| 174 | kmem_cache_free(bvs->slab, bv); |
| 175 | } |
| 176 | } |
| 177 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 178 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
| 179 | mempool_t *pool) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 180 | { |
| 181 | struct bio_vec *bvl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | |
| 183 | /* |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 184 | * see comment near bvec_array define! |
| 185 | */ |
| 186 | switch (nr) { |
| 187 | case 1: |
| 188 | *idx = 0; |
| 189 | break; |
| 190 | case 2 ... 4: |
| 191 | *idx = 1; |
| 192 | break; |
| 193 | case 5 ... 16: |
| 194 | *idx = 2; |
| 195 | break; |
| 196 | case 17 ... 64: |
| 197 | *idx = 3; |
| 198 | break; |
| 199 | case 65 ... 128: |
| 200 | *idx = 4; |
| 201 | break; |
| 202 | case 129 ... BIO_MAX_PAGES: |
| 203 | *idx = 5; |
| 204 | break; |
| 205 | default: |
| 206 | return NULL; |
| 207 | } |
| 208 | |
| 209 | /* |
| 210 | * idx now points to the pool we want to allocate from. only the |
| 211 | * 1-vec entry pool is mempool backed. |
| 212 | */ |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 213 | if (*idx == BVEC_POOL_MAX) { |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 214 | fallback: |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 215 | bvl = mempool_alloc(pool, gfp_mask); |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 216 | } else { |
| 217 | struct biovec_slab *bvs = bvec_slabs + *idx; |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 218 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 219 | |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 220 | /* |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 221 | * Make this allocation restricted and don't dump info on |
| 222 | * allocation failures, since we'll fallback to the mempool |
| 223 | * in case of failure. |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 224 | */ |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 225 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
| 226 | |
| 227 | /* |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 228 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 229 | * is set, retry with the 1-entry mempool |
| 230 | */ |
| 231 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 232 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 233 | *idx = BVEC_POOL_MAX; |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 234 | goto fallback; |
| 235 | } |
| 236 | } |
| 237 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 238 | (*idx)++; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 239 | return bvl; |
| 240 | } |
| 241 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 242 | static void __bio_free(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 243 | { |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 244 | bio_disassociate_task(bio); |
Jens Axboe | 992c5dd | 2007-07-18 13:18:08 +0200 | [diff] [blame] | 245 | |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 246 | if (bio_integrity(bio)) |
Kent Overstreet | 1e2a410f | 2012-09-06 15:34:56 -0700 | [diff] [blame] | 247 | bio_integrity_free(bio); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 248 | } |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 249 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 250 | static void bio_free(struct bio *bio) |
| 251 | { |
| 252 | struct bio_set *bs = bio->bi_pool; |
| 253 | void *p; |
| 254 | |
| 255 | __bio_free(bio); |
| 256 | |
| 257 | if (bs) { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 258 | bvec_free(bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio)); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 259 | |
| 260 | /* |
| 261 | * If we have front padding, adjust the bio pointer before freeing |
| 262 | */ |
| 263 | p = bio; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 264 | p -= bs->front_pad; |
| 265 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 266 | mempool_free(p, bs->bio_pool); |
| 267 | } else { |
| 268 | /* Bio was allocated by bio_kmalloc() */ |
| 269 | kfree(bio); |
| 270 | } |
Peter Osterlund | 3676347 | 2005-09-06 15:16:42 -0700 | [diff] [blame] | 271 | } |
| 272 | |
Arjan van de Ven | 858119e | 2006-01-14 13:20:43 -0800 | [diff] [blame] | 273 | void bio_init(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 274 | { |
Jens Axboe | 2b94de5 | 2007-07-18 13:14:03 +0200 | [diff] [blame] | 275 | memset(bio, 0, sizeof(*bio)); |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 276 | atomic_set(&bio->__bi_remaining, 1); |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 277 | atomic_set(&bio->__bi_cnt, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 278 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 279 | EXPORT_SYMBOL(bio_init); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 280 | |
| 281 | /** |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 282 | * bio_reset - reinitialize a bio |
| 283 | * @bio: bio to reset |
| 284 | * |
| 285 | * Description: |
| 286 | * After calling bio_reset(), @bio will be in the same state as a freshly |
| 287 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are |
| 288 | * preserved are the ones that are initialized by bio_alloc_bioset(). See |
| 289 | * comment in struct bio. |
| 290 | */ |
| 291 | void bio_reset(struct bio *bio) |
| 292 | { |
| 293 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); |
| 294 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 295 | __bio_free(bio); |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 296 | |
| 297 | memset(bio, 0, BIO_RESET_BYTES); |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 298 | bio->bi_flags = flags; |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 299 | atomic_set(&bio->__bi_remaining, 1); |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 300 | } |
| 301 | EXPORT_SYMBOL(bio_reset); |
| 302 | |
Christoph Hellwig | 38f8baa | 2016-03-11 17:34:51 +0100 | [diff] [blame] | 303 | static struct bio *__bio_chain_endio(struct bio *bio) |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 304 | { |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 305 | struct bio *parent = bio->bi_private; |
| 306 | |
Christoph Hellwig | af3e3a5 | 2016-03-11 17:34:50 +0100 | [diff] [blame] | 307 | if (!parent->bi_error) |
| 308 | parent->bi_error = bio->bi_error; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 309 | bio_put(bio); |
Christoph Hellwig | 38f8baa | 2016-03-11 17:34:51 +0100 | [diff] [blame] | 310 | return parent; |
| 311 | } |
| 312 | |
| 313 | static void bio_chain_endio(struct bio *bio) |
| 314 | { |
| 315 | bio_endio(__bio_chain_endio(bio)); |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 316 | } |
| 317 | |
| 318 | /** |
| 319 | * bio_chain - chain bio completions |
Randy Dunlap | 1051a90 | 2014-04-20 16:03:31 -0700 | [diff] [blame] | 320 | * @bio: the target bio |
| 321 | * @parent: the @bio's parent bio |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 322 | * |
| 323 | * The caller won't have a bi_end_io called when @bio completes - instead, |
| 324 | * @parent's bi_end_io won't be called until both @parent and @bio have |
| 325 | * completed; the chained bio will also be freed when it completes. |
| 326 | * |
| 327 | * The caller must not set bi_private or bi_end_io in @bio. |
| 328 | */ |
| 329 | void bio_chain(struct bio *bio, struct bio *parent) |
| 330 | { |
| 331 | BUG_ON(bio->bi_private || bio->bi_end_io); |
| 332 | |
| 333 | bio->bi_private = parent; |
| 334 | bio->bi_end_io = bio_chain_endio; |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 335 | bio_inc_remaining(parent); |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 336 | } |
| 337 | EXPORT_SYMBOL(bio_chain); |
| 338 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 339 | static void bio_alloc_rescue(struct work_struct *work) |
| 340 | { |
| 341 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); |
| 342 | struct bio *bio; |
| 343 | |
| 344 | while (1) { |
| 345 | spin_lock(&bs->rescue_lock); |
| 346 | bio = bio_list_pop(&bs->rescue_list); |
| 347 | spin_unlock(&bs->rescue_lock); |
| 348 | |
| 349 | if (!bio) |
| 350 | break; |
| 351 | |
| 352 | generic_make_request(bio); |
| 353 | } |
| 354 | } |
| 355 | |
| 356 | static void punt_bios_to_rescuer(struct bio_set *bs) |
| 357 | { |
| 358 | struct bio_list punt, nopunt; |
| 359 | struct bio *bio; |
| 360 | |
| 361 | /* |
| 362 | * In order to guarantee forward progress we must punt only bios that |
| 363 | * were allocated from this bio_set; otherwise, if there was a bio on |
| 364 | * there for a stacking driver higher up in the stack, processing it |
| 365 | * could require allocating bios from this bio_set, and doing that from |
| 366 | * our own rescuer would be bad. |
| 367 | * |
| 368 | * Since bio lists are singly linked, pop them all instead of trying to |
| 369 | * remove from the middle of the list: |
| 370 | */ |
| 371 | |
| 372 | bio_list_init(&punt); |
| 373 | bio_list_init(&nopunt); |
| 374 | |
| 375 | while ((bio = bio_list_pop(current->bio_list))) |
| 376 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
| 377 | |
| 378 | *current->bio_list = nopunt; |
| 379 | |
| 380 | spin_lock(&bs->rescue_lock); |
| 381 | bio_list_merge(&bs->rescue_list, &punt); |
| 382 | spin_unlock(&bs->rescue_lock); |
| 383 | |
| 384 | queue_work(bs->rescue_workqueue, &bs->rescue_work); |
| 385 | } |
| 386 | |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 387 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 388 | * bio_alloc_bioset - allocate a bio for I/O |
| 389 | * @gfp_mask: the GFP_ mask given to the slab allocator |
| 390 | * @nr_iovecs: number of iovecs to pre-allocate |
Jaak Ristioja | db18efa | 2010-01-15 12:05:07 +0200 | [diff] [blame] | 391 | * @bs: the bio_set to allocate from. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 392 | * |
| 393 | * Description: |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 394 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
| 395 | * backed by the @bs's mempool. |
| 396 | * |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 397 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
| 398 | * always be able to allocate a bio. This is due to the mempool guarantees. |
| 399 | * To make this work, callers must never allocate more than 1 bio at a time |
| 400 | * from this pool. Callers that need to allocate more than 1 bio must always |
| 401 | * submit the previously allocated bio for IO before attempting to allocate |
| 402 | * a new one. Failure to do so can cause deadlocks under memory pressure. |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 403 | * |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 404 | * Note that when running under generic_make_request() (i.e. any block |
| 405 | * driver), bios are not submitted until after you return - see the code in |
| 406 | * generic_make_request() that converts recursion into iteration, to prevent |
| 407 | * stack overflows. |
| 408 | * |
| 409 | * This would normally mean allocating multiple bios under |
| 410 | * generic_make_request() would be susceptible to deadlocks, but we have |
| 411 | * deadlock avoidance code that resubmits any blocked bios from a rescuer |
| 412 | * thread. |
| 413 | * |
| 414 | * However, we do not guarantee forward progress for allocations from other |
| 415 | * mempools. Doing multiple allocations from the same mempool under |
| 416 | * generic_make_request() should be avoided - instead, use bio_set's front_pad |
| 417 | * for per bio allocations. |
| 418 | * |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 419 | * RETURNS: |
| 420 | * Pointer to new bio on success, NULL on failure. |
| 421 | */ |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 422 | 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] | 423 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 424 | gfp_t saved_gfp = gfp_mask; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 425 | unsigned front_pad; |
| 426 | unsigned inline_vecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 427 | struct bio_vec *bvl = NULL; |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 428 | struct bio *bio; |
| 429 | void *p; |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 430 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 431 | if (!bs) { |
| 432 | if (nr_iovecs > UIO_MAXIOV) |
| 433 | return NULL; |
| 434 | |
| 435 | p = kmalloc(sizeof(struct bio) + |
| 436 | nr_iovecs * sizeof(struct bio_vec), |
| 437 | gfp_mask); |
| 438 | front_pad = 0; |
| 439 | inline_vecs = nr_iovecs; |
| 440 | } else { |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 441 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
| 442 | if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0)) |
| 443 | return NULL; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 444 | /* |
| 445 | * generic_make_request() converts recursion to iteration; this |
| 446 | * means if we're running beneath it, any bios we allocate and |
| 447 | * submit will not be submitted (and thus freed) until after we |
| 448 | * return. |
| 449 | * |
| 450 | * This exposes us to a potential deadlock if we allocate |
| 451 | * multiple bios from the same bio_set() while running |
| 452 | * underneath generic_make_request(). If we were to allocate |
| 453 | * multiple bios (say a stacking block driver that was splitting |
| 454 | * bios), we would deadlock if we exhausted the mempool's |
| 455 | * reserve. |
| 456 | * |
| 457 | * We solve this, and guarantee forward progress, with a rescuer |
| 458 | * workqueue per bio_set. If we go to allocate and there are |
| 459 | * bios on current->bio_list, we first try the allocation |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 460 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
| 461 | * bios we would be blocking to the rescuer workqueue before |
| 462 | * we retry with the original gfp_flags. |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 463 | */ |
| 464 | |
| 465 | if (current->bio_list && !bio_list_empty(current->bio_list)) |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 466 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 467 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 468 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 469 | if (!p && gfp_mask != saved_gfp) { |
| 470 | punt_bios_to_rescuer(bs); |
| 471 | gfp_mask = saved_gfp; |
| 472 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
| 473 | } |
| 474 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 475 | front_pad = bs->front_pad; |
| 476 | inline_vecs = BIO_INLINE_VECS; |
| 477 | } |
| 478 | |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 479 | if (unlikely(!p)) |
| 480 | return NULL; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 481 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 482 | bio = p + front_pad; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 483 | bio_init(bio); |
| 484 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 485 | if (nr_iovecs > inline_vecs) { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 486 | unsigned long idx = 0; |
| 487 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 488 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 489 | if (!bvl && gfp_mask != saved_gfp) { |
| 490 | punt_bios_to_rescuer(bs); |
| 491 | gfp_mask = saved_gfp; |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 492 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 493 | } |
| 494 | |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 495 | if (unlikely(!bvl)) |
| 496 | goto err_free; |
Kent Overstreet | a38352e | 2012-05-25 13:03:11 -0700 | [diff] [blame] | 497 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 498 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 499 | } else if (nr_iovecs) { |
| 500 | bvl = bio->bi_inline_vecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 501 | } |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 502 | |
| 503 | bio->bi_pool = bs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 504 | bio->bi_max_vecs = nr_iovecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 505 | bio->bi_io_vec = bvl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 506 | return bio; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 507 | |
| 508 | err_free: |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 509 | mempool_free(p, bs->bio_pool); |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 510 | return NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 511 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 512 | EXPORT_SYMBOL(bio_alloc_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 513 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 514 | void zero_fill_bio(struct bio *bio) |
| 515 | { |
| 516 | unsigned long flags; |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 517 | struct bio_vec bv; |
| 518 | struct bvec_iter iter; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 519 | |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 520 | bio_for_each_segment(bv, bio, iter) { |
| 521 | char *data = bvec_kmap_irq(&bv, &flags); |
| 522 | memset(data, 0, bv.bv_len); |
| 523 | flush_dcache_page(bv.bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 524 | bvec_kunmap_irq(data, &flags); |
| 525 | } |
| 526 | } |
| 527 | EXPORT_SYMBOL(zero_fill_bio); |
| 528 | |
| 529 | /** |
| 530 | * bio_put - release a reference to a bio |
| 531 | * @bio: bio to release reference to |
| 532 | * |
| 533 | * Description: |
| 534 | * 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] | 535 | * 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] | 536 | **/ |
| 537 | void bio_put(struct bio *bio) |
| 538 | { |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 539 | if (!bio_flagged(bio, BIO_REFFED)) |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 540 | bio_free(bio); |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 541 | else { |
| 542 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); |
| 543 | |
| 544 | /* |
| 545 | * last put frees it |
| 546 | */ |
| 547 | if (atomic_dec_and_test(&bio->__bi_cnt)) |
| 548 | bio_free(bio); |
| 549 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 550 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 551 | EXPORT_SYMBOL(bio_put); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 552 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 553 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 554 | { |
| 555 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) |
| 556 | blk_recount_segments(q, bio); |
| 557 | |
| 558 | return bio->bi_phys_segments; |
| 559 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 560 | EXPORT_SYMBOL(bio_phys_segments); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 561 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 562 | /** |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 563 | * __bio_clone_fast - clone a bio that shares the original bio's biovec |
| 564 | * @bio: destination bio |
| 565 | * @bio_src: bio to clone |
| 566 | * |
| 567 | * Clone a &bio. Caller will own the returned bio, but not |
| 568 | * the actual data it points to. Reference count of returned |
| 569 | * bio will be one. |
| 570 | * |
| 571 | * Caller must ensure that @bio_src is not freed before @bio. |
| 572 | */ |
| 573 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) |
| 574 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 575 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 576 | |
| 577 | /* |
| 578 | * most users will be overriding ->bi_bdev with a new target, |
| 579 | * so we don't set nor calculate new physical/hw segment counts here |
| 580 | */ |
| 581 | bio->bi_bdev = bio_src->bi_bdev; |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 582 | bio_set_flag(bio, BIO_CLONED); |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 583 | bio->bi_opf = bio_src->bi_opf; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 584 | bio->bi_iter = bio_src->bi_iter; |
| 585 | bio->bi_io_vec = bio_src->bi_io_vec; |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 586 | |
| 587 | bio_clone_blkcg_association(bio, bio_src); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 588 | } |
| 589 | EXPORT_SYMBOL(__bio_clone_fast); |
| 590 | |
| 591 | /** |
| 592 | * bio_clone_fast - clone a bio that shares the original bio's biovec |
| 593 | * @bio: bio to clone |
| 594 | * @gfp_mask: allocation priority |
| 595 | * @bs: bio_set to allocate from |
| 596 | * |
| 597 | * Like __bio_clone_fast, only also allocates the returned bio |
| 598 | */ |
| 599 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) |
| 600 | { |
| 601 | struct bio *b; |
| 602 | |
| 603 | b = bio_alloc_bioset(gfp_mask, 0, bs); |
| 604 | if (!b) |
| 605 | return NULL; |
| 606 | |
| 607 | __bio_clone_fast(b, bio); |
| 608 | |
| 609 | if (bio_integrity(bio)) { |
| 610 | int ret; |
| 611 | |
| 612 | ret = bio_integrity_clone(b, bio, gfp_mask); |
| 613 | |
| 614 | if (ret < 0) { |
| 615 | bio_put(b); |
| 616 | return NULL; |
| 617 | } |
| 618 | } |
| 619 | |
| 620 | return b; |
| 621 | } |
| 622 | EXPORT_SYMBOL(bio_clone_fast); |
| 623 | |
| 624 | /** |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 625 | * bio_clone_bioset - clone a bio |
| 626 | * @bio_src: bio to clone |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 627 | * @gfp_mask: allocation priority |
Kent Overstreet | bf800ef | 2012-09-06 15:35:02 -0700 | [diff] [blame] | 628 | * @bs: bio_set to allocate from |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 629 | * |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 630 | * Clone bio. Caller will own the returned bio, but not the actual data it |
| 631 | * points to. Reference count of returned bio will be one. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 632 | */ |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 633 | 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] | 634 | struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 635 | { |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 636 | struct bvec_iter iter; |
| 637 | struct bio_vec bv; |
| 638 | struct bio *bio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 639 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 640 | /* |
| 641 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from |
| 642 | * bio_src->bi_io_vec to bio->bi_io_vec. |
| 643 | * |
| 644 | * We can't do that anymore, because: |
| 645 | * |
| 646 | * - The point of cloning the biovec is to produce a bio with a biovec |
| 647 | * the caller can modify: bi_idx and bi_bvec_done should be 0. |
| 648 | * |
| 649 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if |
| 650 | * we tried to clone the whole thing bio_alloc_bioset() would fail. |
| 651 | * But the clone should succeed as long as the number of biovecs we |
| 652 | * actually need to allocate is fewer than BIO_MAX_PAGES. |
| 653 | * |
| 654 | * - Lastly, bi_vcnt should not be looked at or relied upon by code |
| 655 | * that does not own the bio - reason being drivers don't use it for |
| 656 | * iterating over the biovec anymore, so expecting it to be kept up |
| 657 | * to date (i.e. for clones that share the parent biovec) is just |
| 658 | * asking for trouble and would force extra work on |
| 659 | * __bio_clone_fast() anyways. |
| 660 | */ |
| 661 | |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 662 | bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs); |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 663 | if (!bio) |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 664 | return NULL; |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 665 | bio->bi_bdev = bio_src->bi_bdev; |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 666 | bio->bi_opf = bio_src->bi_opf; |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 667 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; |
| 668 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 669 | |
Adrian Hunter | 7afafc8 | 2016-08-16 10:59:35 +0300 | [diff] [blame] | 670 | switch (bio_op(bio)) { |
| 671 | case REQ_OP_DISCARD: |
| 672 | case REQ_OP_SECURE_ERASE: |
| 673 | break; |
| 674 | case REQ_OP_WRITE_SAME: |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 675 | bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0]; |
Adrian Hunter | 7afafc8 | 2016-08-16 10:59:35 +0300 | [diff] [blame] | 676 | break; |
| 677 | default: |
| 678 | bio_for_each_segment(bv, bio_src, iter) |
| 679 | bio->bi_io_vec[bio->bi_vcnt++] = bv; |
| 680 | break; |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 681 | } |
| 682 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 683 | if (bio_integrity(bio_src)) { |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 684 | int ret; |
| 685 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 686 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
Li Zefan | 059ea33 | 2009-03-09 10:42:45 +0100 | [diff] [blame] | 687 | if (ret < 0) { |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 688 | bio_put(bio); |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 689 | return NULL; |
Li Zefan | 059ea33 | 2009-03-09 10:42:45 +0100 | [diff] [blame] | 690 | } |
Peter Osterlund | 3676347 | 2005-09-06 15:16:42 -0700 | [diff] [blame] | 691 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 692 | |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 693 | bio_clone_blkcg_association(bio, bio_src); |
| 694 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 695 | return bio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 696 | } |
Kent Overstreet | bf800ef | 2012-09-06 15:35:02 -0700 | [diff] [blame] | 697 | EXPORT_SYMBOL(bio_clone_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 698 | |
| 699 | /** |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 700 | * bio_add_pc_page - attempt to add page to bio |
| 701 | * @q: the target queue |
| 702 | * @bio: destination bio |
| 703 | * @page: page to add |
| 704 | * @len: vec entry length |
| 705 | * @offset: vec entry offset |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 706 | * |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 707 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
| 708 | * number of reasons, such as the bio being full or target block device |
| 709 | * limitations. The target block device must allow bio's up to PAGE_SIZE, |
| 710 | * so it is always possible to add a single page to an empty bio. |
| 711 | * |
| 712 | * This should only be used by REQ_PC bios. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 713 | */ |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 714 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
| 715 | *page, unsigned int len, unsigned int offset) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 716 | { |
| 717 | int retried_segments = 0; |
| 718 | struct bio_vec *bvec; |
| 719 | |
| 720 | /* |
| 721 | * cloned bio must not modify vec list |
| 722 | */ |
| 723 | if (unlikely(bio_flagged(bio, BIO_CLONED))) |
| 724 | return 0; |
| 725 | |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 726 | if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 727 | return 0; |
| 728 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 729 | /* |
| 730 | * For filesystems with a blocksize smaller than the pagesize |
| 731 | * we will often be called with the same page as last time and |
| 732 | * a consecutive offset. Optimize this special case. |
| 733 | */ |
| 734 | if (bio->bi_vcnt > 0) { |
| 735 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 736 | |
| 737 | if (page == prev->bv_page && |
| 738 | offset == prev->bv_offset + prev->bv_len) { |
| 739 | prev->bv_len += len; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 740 | bio->bi_iter.bi_size += len; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 741 | goto done; |
| 742 | } |
Jens Axboe | 66cb45a | 2014-06-24 16:22:24 -0600 | [diff] [blame] | 743 | |
| 744 | /* |
| 745 | * If the queue doesn't support SG gaps and adding this |
| 746 | * offset would create a gap, disallow it. |
| 747 | */ |
Keith Busch | 03100aa | 2015-08-19 14:24:05 -0700 | [diff] [blame] | 748 | if (bvec_gap_to_prev(q, prev, offset)) |
Jens Axboe | 66cb45a | 2014-06-24 16:22:24 -0600 | [diff] [blame] | 749 | return 0; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 750 | } |
| 751 | |
| 752 | if (bio->bi_vcnt >= bio->bi_max_vecs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 753 | return 0; |
| 754 | |
| 755 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 756 | * setup the new entry, we might clear it again later if we |
| 757 | * cannot add the page |
| 758 | */ |
| 759 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; |
| 760 | bvec->bv_page = page; |
| 761 | bvec->bv_len = len; |
| 762 | bvec->bv_offset = offset; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 763 | bio->bi_vcnt++; |
| 764 | bio->bi_phys_segments++; |
| 765 | bio->bi_iter.bi_size += len; |
| 766 | |
| 767 | /* |
| 768 | * Perform a recount if the number of segments is greater |
| 769 | * than queue_max_segments(q). |
| 770 | */ |
| 771 | |
| 772 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
| 773 | |
| 774 | if (retried_segments) |
| 775 | goto failed; |
| 776 | |
| 777 | retried_segments = 1; |
| 778 | blk_recount_segments(q, bio); |
| 779 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 780 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 781 | /* If we may be able to merge these biovecs, force a recount */ |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 782 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 783 | bio_clear_flag(bio, BIO_SEG_VALID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 784 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 785 | done: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 786 | return len; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 787 | |
| 788 | failed: |
| 789 | bvec->bv_page = NULL; |
| 790 | bvec->bv_len = 0; |
| 791 | bvec->bv_offset = 0; |
| 792 | bio->bi_vcnt--; |
| 793 | bio->bi_iter.bi_size -= len; |
| 794 | blk_recount_segments(q, bio); |
| 795 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 796 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 797 | EXPORT_SYMBOL(bio_add_pc_page); |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 798 | |
| 799 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 800 | * bio_add_page - attempt to add page to bio |
| 801 | * @bio: destination bio |
| 802 | * @page: page to add |
| 803 | * @len: vec entry length |
| 804 | * @offset: vec entry offset |
| 805 | * |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 806 | * Attempt to add a page to the bio_vec maplist. This will only fail |
| 807 | * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 808 | */ |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 809 | int bio_add_page(struct bio *bio, struct page *page, |
| 810 | unsigned int len, unsigned int offset) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 811 | { |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 812 | struct bio_vec *bv; |
Jens Axboe | 762380a | 2014-06-05 13:38:39 -0600 | [diff] [blame] | 813 | |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 814 | /* |
| 815 | * cloned bio must not modify vec list |
| 816 | */ |
| 817 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
| 818 | return 0; |
Jens Axboe | 58a4915 | 2014-06-10 12:53:56 -0600 | [diff] [blame] | 819 | |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 820 | /* |
| 821 | * For filesystems with a blocksize smaller than the pagesize |
| 822 | * we will often be called with the same page as last time and |
| 823 | * a consecutive offset. Optimize this special case. |
| 824 | */ |
| 825 | if (bio->bi_vcnt > 0) { |
| 826 | bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 827 | |
| 828 | if (page == bv->bv_page && |
| 829 | offset == bv->bv_offset + bv->bv_len) { |
| 830 | bv->bv_len += len; |
| 831 | goto done; |
| 832 | } |
| 833 | } |
| 834 | |
| 835 | if (bio->bi_vcnt >= bio->bi_max_vecs) |
| 836 | return 0; |
| 837 | |
| 838 | bv = &bio->bi_io_vec[bio->bi_vcnt]; |
| 839 | bv->bv_page = page; |
| 840 | bv->bv_len = len; |
| 841 | bv->bv_offset = offset; |
| 842 | |
| 843 | bio->bi_vcnt++; |
| 844 | done: |
| 845 | bio->bi_iter.bi_size += len; |
| 846 | return len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 847 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 848 | EXPORT_SYMBOL(bio_add_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 849 | |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 850 | struct submit_bio_ret { |
| 851 | struct completion event; |
| 852 | int error; |
| 853 | }; |
| 854 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 855 | static void submit_bio_wait_endio(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 856 | { |
| 857 | struct submit_bio_ret *ret = bio->bi_private; |
| 858 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 859 | ret->error = bio->bi_error; |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 860 | complete(&ret->event); |
| 861 | } |
| 862 | |
| 863 | /** |
| 864 | * submit_bio_wait - submit a bio, and wait until it completes |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 865 | * @bio: The &struct bio which describes the I/O |
| 866 | * |
| 867 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from |
| 868 | * bio_endio() on failure. |
| 869 | */ |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 870 | int submit_bio_wait(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 871 | { |
| 872 | struct submit_bio_ret ret; |
| 873 | |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 874 | init_completion(&ret.event); |
| 875 | bio->bi_private = &ret; |
| 876 | bio->bi_end_io = submit_bio_wait_endio; |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 877 | bio->bi_opf |= REQ_SYNC; |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 878 | submit_bio(bio); |
Stephane Gasparini | d57d611 | 2016-02-09 17:07:38 +0100 | [diff] [blame] | 879 | wait_for_completion_io(&ret.event); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 880 | |
| 881 | return ret.error; |
| 882 | } |
| 883 | EXPORT_SYMBOL(submit_bio_wait); |
| 884 | |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 885 | /** |
| 886 | * bio_advance - increment/complete a bio by some number of bytes |
| 887 | * @bio: bio to advance |
| 888 | * @bytes: number of bytes to complete |
| 889 | * |
| 890 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to |
| 891 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will |
| 892 | * be updated on the last bvec as well. |
| 893 | * |
| 894 | * @bio will then represent the remaining, uncompleted portion of the io. |
| 895 | */ |
| 896 | void bio_advance(struct bio *bio, unsigned bytes) |
| 897 | { |
| 898 | if (bio_integrity(bio)) |
| 899 | bio_integrity_advance(bio, bytes); |
| 900 | |
Kent Overstreet | 4550dd6 | 2013-08-07 14:26:21 -0700 | [diff] [blame] | 901 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 902 | } |
| 903 | EXPORT_SYMBOL(bio_advance); |
| 904 | |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 905 | /** |
Kent Overstreet | a078760 | 2012-09-10 14:03:28 -0700 | [diff] [blame] | 906 | * bio_alloc_pages - allocates a single page for each bvec in a bio |
| 907 | * @bio: bio to allocate pages for |
| 908 | * @gfp_mask: flags for allocation |
| 909 | * |
| 910 | * Allocates pages up to @bio->bi_vcnt. |
| 911 | * |
| 912 | * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are |
| 913 | * freed. |
| 914 | */ |
| 915 | int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask) |
| 916 | { |
| 917 | int i; |
| 918 | struct bio_vec *bv; |
| 919 | |
| 920 | bio_for_each_segment_all(bv, bio, i) { |
| 921 | bv->bv_page = alloc_page(gfp_mask); |
| 922 | if (!bv->bv_page) { |
| 923 | while (--bv >= bio->bi_io_vec) |
| 924 | __free_page(bv->bv_page); |
| 925 | return -ENOMEM; |
| 926 | } |
| 927 | } |
| 928 | |
| 929 | return 0; |
| 930 | } |
| 931 | EXPORT_SYMBOL(bio_alloc_pages); |
| 932 | |
| 933 | /** |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 934 | * bio_copy_data - copy contents of data buffers from one chain of bios to |
| 935 | * another |
| 936 | * @src: source bio list |
| 937 | * @dst: destination bio list |
| 938 | * |
| 939 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats |
| 940 | * @src and @dst as linked lists of bios. |
| 941 | * |
| 942 | * Stops when it reaches the end of either @src or @dst - that is, copies |
| 943 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). |
| 944 | */ |
| 945 | void bio_copy_data(struct bio *dst, struct bio *src) |
| 946 | { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 947 | struct bvec_iter src_iter, dst_iter; |
| 948 | struct bio_vec src_bv, dst_bv; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 949 | void *src_p, *dst_p; |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 950 | unsigned bytes; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 951 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 952 | src_iter = src->bi_iter; |
| 953 | dst_iter = dst->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 954 | |
| 955 | while (1) { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 956 | if (!src_iter.bi_size) { |
| 957 | src = src->bi_next; |
| 958 | if (!src) |
| 959 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 960 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 961 | src_iter = src->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 962 | } |
| 963 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 964 | if (!dst_iter.bi_size) { |
| 965 | dst = dst->bi_next; |
| 966 | if (!dst) |
| 967 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 968 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 969 | dst_iter = dst->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 970 | } |
| 971 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 972 | src_bv = bio_iter_iovec(src, src_iter); |
| 973 | dst_bv = bio_iter_iovec(dst, dst_iter); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 974 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 975 | bytes = min(src_bv.bv_len, dst_bv.bv_len); |
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_p = kmap_atomic(src_bv.bv_page); |
| 978 | dst_p = kmap_atomic(dst_bv.bv_page); |
| 979 | |
| 980 | memcpy(dst_p + dst_bv.bv_offset, |
| 981 | src_p + src_bv.bv_offset, |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 982 | bytes); |
| 983 | |
| 984 | kunmap_atomic(dst_p); |
| 985 | kunmap_atomic(src_p); |
| 986 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 987 | bio_advance_iter(src, &src_iter, bytes); |
| 988 | bio_advance_iter(dst, &dst_iter, bytes); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 989 | } |
| 990 | } |
| 991 | EXPORT_SYMBOL(bio_copy_data); |
| 992 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 993 | struct bio_map_data { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 994 | int is_our_pages; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 995 | struct iov_iter iter; |
| 996 | struct iovec iov[]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 997 | }; |
| 998 | |
Fabian Frederick | 7410b3c | 2014-04-22 15:09:07 -0600 | [diff] [blame] | 999 | 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] | 1000 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1001 | { |
Jens Axboe | f3f63c1 | 2010-10-29 11:46:56 -0600 | [diff] [blame] | 1002 | if (iov_count > UIO_MAXIOV) |
| 1003 | return NULL; |
| 1004 | |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1005 | return kmalloc(sizeof(struct bio_map_data) + |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1006 | sizeof(struct iovec) * iov_count, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1007 | } |
| 1008 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1009 | /** |
| 1010 | * bio_copy_from_iter - copy all pages from iov_iter to bio |
| 1011 | * @bio: The &struct bio which describes the I/O as destination |
| 1012 | * @iter: iov_iter as source |
| 1013 | * |
| 1014 | * Copy all pages from iov_iter to bio. |
| 1015 | * Returns 0 on success, or error on failure. |
| 1016 | */ |
| 1017 | static int bio_copy_from_iter(struct bio *bio, struct iov_iter iter) |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1018 | { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1019 | int i; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1020 | struct bio_vec *bvec; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1021 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1022 | bio_for_each_segment_all(bvec, bio, i) { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1023 | ssize_t ret; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1024 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1025 | ret = copy_page_from_iter(bvec->bv_page, |
| 1026 | bvec->bv_offset, |
| 1027 | bvec->bv_len, |
| 1028 | &iter); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1029 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1030 | if (!iov_iter_count(&iter)) |
| 1031 | break; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1032 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1033 | if (ret < bvec->bv_len) |
| 1034 | return -EFAULT; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1035 | } |
| 1036 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1037 | return 0; |
| 1038 | } |
| 1039 | |
| 1040 | /** |
| 1041 | * bio_copy_to_iter - copy all pages from bio to iov_iter |
| 1042 | * @bio: The &struct bio which describes the I/O as source |
| 1043 | * @iter: iov_iter as destination |
| 1044 | * |
| 1045 | * Copy all pages from bio to iov_iter. |
| 1046 | * Returns 0 on success, or error on failure. |
| 1047 | */ |
| 1048 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) |
| 1049 | { |
| 1050 | int i; |
| 1051 | struct bio_vec *bvec; |
| 1052 | |
| 1053 | bio_for_each_segment_all(bvec, bio, i) { |
| 1054 | ssize_t ret; |
| 1055 | |
| 1056 | ret = copy_page_to_iter(bvec->bv_page, |
| 1057 | bvec->bv_offset, |
| 1058 | bvec->bv_len, |
| 1059 | &iter); |
| 1060 | |
| 1061 | if (!iov_iter_count(&iter)) |
| 1062 | break; |
| 1063 | |
| 1064 | if (ret < bvec->bv_len) |
| 1065 | return -EFAULT; |
| 1066 | } |
| 1067 | |
| 1068 | return 0; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1069 | } |
| 1070 | |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1071 | static void bio_free_pages(struct bio *bio) |
| 1072 | { |
| 1073 | struct bio_vec *bvec; |
| 1074 | int i; |
| 1075 | |
| 1076 | bio_for_each_segment_all(bvec, bio, i) |
| 1077 | __free_page(bvec->bv_page); |
| 1078 | } |
| 1079 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1080 | /** |
| 1081 | * bio_uncopy_user - finish previously mapped bio |
| 1082 | * @bio: bio being terminated |
| 1083 | * |
Christoph Hellwig | ddad8dd | 2015-01-18 16:16:29 +0100 | [diff] [blame] | 1084 | * Free pages allocated from bio_copy_user_iov() and write back data |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1085 | * to user space in case of a read. |
| 1086 | */ |
| 1087 | int bio_uncopy_user(struct bio *bio) |
| 1088 | { |
| 1089 | struct bio_map_data *bmd = bio->bi_private; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1090 | int ret = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1091 | |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1092 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
| 1093 | /* |
| 1094 | * if we're in a workqueue, the request is orphaned, so |
Hannes Reinecke | 2d99b55 | 2016-02-12 09:39:15 +0100 | [diff] [blame] | 1095 | * don't copy into a random user address space, just free |
| 1096 | * and return -EINTR so user space doesn't expect any data. |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1097 | */ |
Hannes Reinecke | 2d99b55 | 2016-02-12 09:39:15 +0100 | [diff] [blame] | 1098 | if (!current->mm) |
| 1099 | ret = -EINTR; |
| 1100 | else if (bio_data_dir(bio) == READ) |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1101 | ret = bio_copy_to_iter(bio, bmd->iter); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1102 | if (bmd->is_our_pages) |
| 1103 | bio_free_pages(bio); |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1104 | } |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1105 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1106 | bio_put(bio); |
| 1107 | return ret; |
| 1108 | } |
| 1109 | |
| 1110 | /** |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1111 | * bio_copy_user_iov - copy user data to bio |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1112 | * @q: destination block queue |
| 1113 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
| 1114 | * @iter: iovec iterator |
| 1115 | * @gfp_mask: memory allocation flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1116 | * |
| 1117 | * Prepares and returns a bio for indirect user io, bouncing data |
| 1118 | * to/from kernel pages as necessary. Must be paired with |
| 1119 | * call bio_uncopy_user() on io completion. |
| 1120 | */ |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1121 | struct bio *bio_copy_user_iov(struct request_queue *q, |
| 1122 | struct rq_map_data *map_data, |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1123 | const struct iov_iter *iter, |
| 1124 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1125 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1126 | struct bio_map_data *bmd; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1127 | struct page *page; |
| 1128 | struct bio *bio; |
| 1129 | int i, ret; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1130 | int nr_pages = 0; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1131 | unsigned int len = iter->count; |
Geliang Tang | bd5cece | 2015-11-21 17:27:31 +0800 | [diff] [blame] | 1132 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1133 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1134 | for (i = 0; i < iter->nr_segs; i++) { |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1135 | unsigned long uaddr; |
| 1136 | unsigned long end; |
| 1137 | unsigned long start; |
| 1138 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1139 | uaddr = (unsigned long) iter->iov[i].iov_base; |
| 1140 | end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1) |
| 1141 | >> PAGE_SHIFT; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1142 | start = uaddr >> PAGE_SHIFT; |
| 1143 | |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1144 | /* |
| 1145 | * Overflow, abort |
| 1146 | */ |
| 1147 | if (end < start) |
| 1148 | return ERR_PTR(-EINVAL); |
| 1149 | |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1150 | nr_pages += end - start; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1151 | } |
| 1152 | |
FUJITA Tomonori | 6983872 | 2009-04-28 20:24:29 +0200 | [diff] [blame] | 1153 | if (offset) |
| 1154 | nr_pages++; |
| 1155 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1156 | bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1157 | if (!bmd) |
| 1158 | return ERR_PTR(-ENOMEM); |
| 1159 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1160 | /* |
| 1161 | * We need to do a deep copy of the iov_iter including the iovecs. |
| 1162 | * The caller provided iov might point to an on-stack or otherwise |
| 1163 | * shortlived one. |
| 1164 | */ |
| 1165 | bmd->is_our_pages = map_data ? 0 : 1; |
| 1166 | memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs); |
| 1167 | iov_iter_init(&bmd->iter, iter->type, bmd->iov, |
| 1168 | iter->nr_segs, iter->count); |
| 1169 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1170 | ret = -ENOMEM; |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1171 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1172 | if (!bio) |
| 1173 | goto out_bmd; |
| 1174 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1175 | if (iter->type & WRITE) |
Mike Christie | 95fe6c1 | 2016-06-05 14:31:48 -0500 | [diff] [blame] | 1176 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1177 | |
| 1178 | ret = 0; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1179 | |
| 1180 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1181 | nr_pages = 1 << map_data->page_order; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1182 | i = map_data->offset / PAGE_SIZE; |
| 1183 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1184 | while (len) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1185 | unsigned int bytes = PAGE_SIZE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1186 | |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1187 | bytes -= offset; |
| 1188 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1189 | if (bytes > len) |
| 1190 | bytes = len; |
| 1191 | |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1192 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1193 | if (i == map_data->nr_entries * nr_pages) { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1194 | ret = -ENOMEM; |
| 1195 | break; |
| 1196 | } |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1197 | |
| 1198 | page = map_data->pages[i / nr_pages]; |
| 1199 | page += (i % nr_pages); |
| 1200 | |
| 1201 | i++; |
| 1202 | } else { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1203 | page = alloc_page(q->bounce_gfp | gfp_mask); |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1204 | if (!page) { |
| 1205 | ret = -ENOMEM; |
| 1206 | break; |
| 1207 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1208 | } |
| 1209 | |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1210 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1211 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1212 | |
| 1213 | len -= bytes; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1214 | offset = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1215 | } |
| 1216 | |
| 1217 | if (ret) |
| 1218 | goto cleanup; |
| 1219 | |
| 1220 | /* |
| 1221 | * success |
| 1222 | */ |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1223 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
FUJITA Tomonori | ecb554a | 2009-07-09 14:46:53 +0200 | [diff] [blame] | 1224 | (map_data && map_data->from_user)) { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1225 | ret = bio_copy_from_iter(bio, *iter); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1226 | if (ret) |
| 1227 | goto cleanup; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1228 | } |
| 1229 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1230 | bio->bi_private = bmd; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1231 | return bio; |
| 1232 | cleanup: |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1233 | if (!map_data) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1234 | bio_free_pages(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1235 | bio_put(bio); |
| 1236 | out_bmd: |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1237 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1238 | return ERR_PTR(ret); |
| 1239 | } |
| 1240 | |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1241 | /** |
| 1242 | * bio_map_user_iov - map user iovec into bio |
| 1243 | * @q: the struct request_queue for the bio |
| 1244 | * @iter: iovec iterator |
| 1245 | * @gfp_mask: memory allocation flags |
| 1246 | * |
| 1247 | * Map the user space address into a bio suitable for io to a block |
| 1248 | * device. Returns an error pointer in case of error. |
| 1249 | */ |
| 1250 | struct bio *bio_map_user_iov(struct request_queue *q, |
| 1251 | const struct iov_iter *iter, |
| 1252 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1253 | { |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1254 | int j; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1255 | int nr_pages = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1256 | struct page **pages; |
| 1257 | struct bio *bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1258 | int cur_page = 0; |
| 1259 | int ret, offset; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1260 | struct iov_iter i; |
| 1261 | struct iovec iov; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1262 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1263 | iov_for_each(iov, i, *iter) { |
| 1264 | unsigned long uaddr = (unsigned long) iov.iov_base; |
| 1265 | unsigned long len = iov.iov_len; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1266 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1267 | unsigned long start = uaddr >> PAGE_SHIFT; |
| 1268 | |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1269 | /* |
| 1270 | * Overflow, abort |
| 1271 | */ |
| 1272 | if (end < start) |
| 1273 | return ERR_PTR(-EINVAL); |
| 1274 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1275 | nr_pages += end - start; |
| 1276 | /* |
Mike Christie | ad2d722 | 2006-12-01 10:40:20 +0100 | [diff] [blame] | 1277 | * buffer must be aligned to at least hardsector size for now |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1278 | */ |
Mike Christie | ad2d722 | 2006-12-01 10:40:20 +0100 | [diff] [blame] | 1279 | if (uaddr & queue_dma_alignment(q)) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1280 | return ERR_PTR(-EINVAL); |
| 1281 | } |
| 1282 | |
| 1283 | if (!nr_pages) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1284 | return ERR_PTR(-EINVAL); |
| 1285 | |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1286 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1287 | if (!bio) |
| 1288 | return ERR_PTR(-ENOMEM); |
| 1289 | |
| 1290 | ret = -ENOMEM; |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1291 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1292 | if (!pages) |
| 1293 | goto out; |
| 1294 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1295 | iov_for_each(iov, i, *iter) { |
| 1296 | unsigned long uaddr = (unsigned long) iov.iov_base; |
| 1297 | unsigned long len = iov.iov_len; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1298 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1299 | unsigned long start = uaddr >> PAGE_SHIFT; |
| 1300 | const int local_nr_pages = end - start; |
| 1301 | const int page_limit = cur_page + local_nr_pages; |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1302 | |
Nick Piggin | f5dd33c | 2008-07-25 19:45:25 -0700 | [diff] [blame] | 1303 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1304 | (iter->type & WRITE) != WRITE, |
| 1305 | &pages[cur_page]); |
Jens Axboe | 9917215 | 2006-06-16 13:02:29 +0200 | [diff] [blame] | 1306 | if (ret < local_nr_pages) { |
| 1307 | ret = -EFAULT; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1308 | goto out_unmap; |
Jens Axboe | 9917215 | 2006-06-16 13:02:29 +0200 | [diff] [blame] | 1309 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1310 | |
Geliang Tang | bd5cece | 2015-11-21 17:27:31 +0800 | [diff] [blame] | 1311 | offset = offset_in_page(uaddr); |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1312 | for (j = cur_page; j < page_limit; j++) { |
| 1313 | unsigned int bytes = PAGE_SIZE - offset; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1314 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1315 | if (len <= 0) |
| 1316 | break; |
| 1317 | |
| 1318 | if (bytes > len) |
| 1319 | bytes = len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1320 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1321 | /* |
| 1322 | * sorry... |
| 1323 | */ |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 1324 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
| 1325 | bytes) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1326 | break; |
| 1327 | |
| 1328 | len -= bytes; |
| 1329 | offset = 0; |
| 1330 | } |
| 1331 | |
| 1332 | cur_page = j; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1333 | /* |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1334 | * release the pages we didn't map into the bio, if any |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1335 | */ |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1336 | while (j < page_limit) |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1337 | put_page(pages[j++]); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1338 | } |
| 1339 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1340 | kfree(pages); |
| 1341 | |
| 1342 | /* |
| 1343 | * set data direction, and check if mapped pages need bouncing |
| 1344 | */ |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1345 | if (iter->type & WRITE) |
Mike Christie | 95fe6c1 | 2016-06-05 14:31:48 -0500 | [diff] [blame] | 1346 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1347 | |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1348 | bio_set_flag(bio, BIO_USER_MAPPED); |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1349 | |
| 1350 | /* |
| 1351 | * subtle -- if __bio_map_user() ended up bouncing a bio, |
| 1352 | * it would normally disappear when its bi_end_io is run. |
| 1353 | * however, we need it for the unmap, so grab an extra |
| 1354 | * reference to it |
| 1355 | */ |
| 1356 | bio_get(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1357 | return bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1358 | |
| 1359 | out_unmap: |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1360 | for (j = 0; j < nr_pages; j++) { |
| 1361 | if (!pages[j]) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1362 | break; |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1363 | put_page(pages[j]); |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1364 | } |
| 1365 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1366 | kfree(pages); |
| 1367 | bio_put(bio); |
| 1368 | return ERR_PTR(ret); |
| 1369 | } |
| 1370 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1371 | static void __bio_unmap_user(struct bio *bio) |
| 1372 | { |
| 1373 | struct bio_vec *bvec; |
| 1374 | int i; |
| 1375 | |
| 1376 | /* |
| 1377 | * make sure we dirty pages we wrote to |
| 1378 | */ |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1379 | bio_for_each_segment_all(bvec, bio, i) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1380 | if (bio_data_dir(bio) == READ) |
| 1381 | set_page_dirty_lock(bvec->bv_page); |
| 1382 | |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1383 | put_page(bvec->bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1384 | } |
| 1385 | |
| 1386 | bio_put(bio); |
| 1387 | } |
| 1388 | |
| 1389 | /** |
| 1390 | * bio_unmap_user - unmap a bio |
| 1391 | * @bio: the bio being unmapped |
| 1392 | * |
| 1393 | * Unmap a bio previously mapped by bio_map_user(). Must be called with |
| 1394 | * a process context. |
| 1395 | * |
| 1396 | * bio_unmap_user() may sleep. |
| 1397 | */ |
| 1398 | void bio_unmap_user(struct bio *bio) |
| 1399 | { |
| 1400 | __bio_unmap_user(bio); |
| 1401 | bio_put(bio); |
| 1402 | } |
| 1403 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1404 | static void bio_map_kern_endio(struct bio *bio) |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1405 | { |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1406 | bio_put(bio); |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1407 | } |
| 1408 | |
Christoph Hellwig | 75c72b8 | 2015-01-18 16:16:32 +0100 | [diff] [blame] | 1409 | /** |
| 1410 | * bio_map_kern - map kernel address into bio |
| 1411 | * @q: the struct request_queue for the bio |
| 1412 | * @data: pointer to buffer to map |
| 1413 | * @len: length in bytes |
| 1414 | * @gfp_mask: allocation flags for bio allocation |
| 1415 | * |
| 1416 | * Map the kernel address into a bio suitable for io to a block |
| 1417 | * device. Returns an error pointer in case of error. |
| 1418 | */ |
| 1419 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
| 1420 | gfp_t gfp_mask) |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1421 | { |
| 1422 | unsigned long kaddr = (unsigned long)data; |
| 1423 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1424 | unsigned long start = kaddr >> PAGE_SHIFT; |
| 1425 | const int nr_pages = end - start; |
| 1426 | int offset, i; |
| 1427 | struct bio *bio; |
| 1428 | |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1429 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1430 | if (!bio) |
| 1431 | return ERR_PTR(-ENOMEM); |
| 1432 | |
| 1433 | offset = offset_in_page(kaddr); |
| 1434 | for (i = 0; i < nr_pages; i++) { |
| 1435 | unsigned int bytes = PAGE_SIZE - offset; |
| 1436 | |
| 1437 | if (len <= 0) |
| 1438 | break; |
| 1439 | |
| 1440 | if (bytes > len) |
| 1441 | bytes = len; |
| 1442 | |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 1443 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
Christoph Hellwig | 75c72b8 | 2015-01-18 16:16:32 +0100 | [diff] [blame] | 1444 | offset) < bytes) { |
| 1445 | /* we don't support partial mappings */ |
| 1446 | bio_put(bio); |
| 1447 | return ERR_PTR(-EINVAL); |
| 1448 | } |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1449 | |
| 1450 | data += bytes; |
| 1451 | len -= bytes; |
| 1452 | offset = 0; |
| 1453 | } |
| 1454 | |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1455 | bio->bi_end_io = bio_map_kern_endio; |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1456 | return bio; |
| 1457 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1458 | EXPORT_SYMBOL(bio_map_kern); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1459 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1460 | static void bio_copy_kern_endio(struct bio *bio) |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1461 | { |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1462 | bio_free_pages(bio); |
| 1463 | bio_put(bio); |
| 1464 | } |
| 1465 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1466 | static void bio_copy_kern_endio_read(struct bio *bio) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1467 | { |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1468 | char *p = bio->bi_private; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1469 | struct bio_vec *bvec; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1470 | int i; |
| 1471 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1472 | bio_for_each_segment_all(bvec, bio, i) { |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1473 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1474 | p += bvec->bv_len; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1475 | } |
| 1476 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1477 | bio_copy_kern_endio(bio); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1478 | } |
| 1479 | |
| 1480 | /** |
| 1481 | * bio_copy_kern - copy kernel address into bio |
| 1482 | * @q: the struct request_queue for the bio |
| 1483 | * @data: pointer to buffer to copy |
| 1484 | * @len: length in bytes |
| 1485 | * @gfp_mask: allocation flags for bio and page allocation |
Randy Dunlap | ffee025 | 2008-04-30 09:08:54 +0200 | [diff] [blame] | 1486 | * @reading: data direction is READ |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1487 | * |
| 1488 | * copy the kernel address into a bio suitable for io to a block |
| 1489 | * device. Returns an error pointer in case of error. |
| 1490 | */ |
| 1491 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, |
| 1492 | gfp_t gfp_mask, int reading) |
| 1493 | { |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1494 | unsigned long kaddr = (unsigned long)data; |
| 1495 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1496 | unsigned long start = kaddr >> PAGE_SHIFT; |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1497 | struct bio *bio; |
| 1498 | void *p = data; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1499 | int nr_pages = 0; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1500 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1501 | /* |
| 1502 | * Overflow, abort |
| 1503 | */ |
| 1504 | if (end < start) |
| 1505 | return ERR_PTR(-EINVAL); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1506 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1507 | nr_pages = end - start; |
| 1508 | bio = bio_kmalloc(gfp_mask, nr_pages); |
| 1509 | if (!bio) |
| 1510 | return ERR_PTR(-ENOMEM); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1511 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1512 | while (len) { |
| 1513 | struct page *page; |
| 1514 | unsigned int bytes = PAGE_SIZE; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1515 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1516 | if (bytes > len) |
| 1517 | bytes = len; |
| 1518 | |
| 1519 | page = alloc_page(q->bounce_gfp | gfp_mask); |
| 1520 | if (!page) |
| 1521 | goto cleanup; |
| 1522 | |
| 1523 | if (!reading) |
| 1524 | memcpy(page_address(page), p, bytes); |
| 1525 | |
| 1526 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) |
| 1527 | break; |
| 1528 | |
| 1529 | len -= bytes; |
| 1530 | p += bytes; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1531 | } |
| 1532 | |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1533 | if (reading) { |
| 1534 | bio->bi_end_io = bio_copy_kern_endio_read; |
| 1535 | bio->bi_private = data; |
| 1536 | } else { |
| 1537 | bio->bi_end_io = bio_copy_kern_endio; |
Mike Christie | 95fe6c1 | 2016-06-05 14:31:48 -0500 | [diff] [blame] | 1538 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1539 | } |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1540 | |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1541 | return bio; |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1542 | |
| 1543 | cleanup: |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1544 | bio_free_pages(bio); |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1545 | bio_put(bio); |
| 1546 | return ERR_PTR(-ENOMEM); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1547 | } |
| 1548 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1549 | /* |
| 1550 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions |
| 1551 | * for performing direct-IO in BIOs. |
| 1552 | * |
| 1553 | * The problem is that we cannot run set_page_dirty() from interrupt context |
| 1554 | * because the required locks are not interrupt-safe. So what we can do is to |
| 1555 | * mark the pages dirty _before_ performing IO. And in interrupt context, |
| 1556 | * check that the pages are still dirty. If so, fine. If not, redirty them |
| 1557 | * in process context. |
| 1558 | * |
| 1559 | * We special-case compound pages here: normally this means reads into hugetlb |
| 1560 | * pages. The logic in here doesn't really work right for compound pages |
| 1561 | * because the VM does not uniformly chase down the head page in all cases. |
| 1562 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't |
| 1563 | * handle them at all. So we skip compound pages here at an early stage. |
| 1564 | * |
| 1565 | * Note that this code is very hard to test under normal circumstances because |
| 1566 | * direct-io pins the pages with get_user_pages(). This makes |
| 1567 | * 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] | 1568 | * 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] | 1569 | * pagecache. |
| 1570 | * |
| 1571 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the |
| 1572 | * deferred bio dirtying paths. |
| 1573 | */ |
| 1574 | |
| 1575 | /* |
| 1576 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. |
| 1577 | */ |
| 1578 | void bio_set_pages_dirty(struct bio *bio) |
| 1579 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1580 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1581 | int i; |
| 1582 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1583 | bio_for_each_segment_all(bvec, bio, i) { |
| 1584 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1585 | |
| 1586 | if (page && !PageCompound(page)) |
| 1587 | set_page_dirty_lock(page); |
| 1588 | } |
| 1589 | } |
| 1590 | |
Adrian Bunk | 86b6c7a | 2008-02-18 13:48:32 +0100 | [diff] [blame] | 1591 | static void bio_release_pages(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1592 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1593 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1594 | int i; |
| 1595 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1596 | bio_for_each_segment_all(bvec, bio, i) { |
| 1597 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1598 | |
| 1599 | if (page) |
| 1600 | put_page(page); |
| 1601 | } |
| 1602 | } |
| 1603 | |
| 1604 | /* |
| 1605 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. |
| 1606 | * If they are, then fine. If, however, some pages are clean then they must |
| 1607 | * have been written out during the direct-IO read. So we take another ref on |
| 1608 | * the BIO and the offending pages and re-dirty the pages in process context. |
| 1609 | * |
| 1610 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from |
Kirill A. Shutemov | ea1754a | 2016-04-01 15:29:48 +0300 | [diff] [blame] | 1611 | * here on. It will run one put_page() against each page and will run one |
| 1612 | * bio_put() against the BIO. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1613 | */ |
| 1614 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1615 | static void bio_dirty_fn(struct work_struct *work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1616 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1617 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1618 | static DEFINE_SPINLOCK(bio_dirty_lock); |
| 1619 | static struct bio *bio_dirty_list; |
| 1620 | |
| 1621 | /* |
| 1622 | * This runs in process context |
| 1623 | */ |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1624 | static void bio_dirty_fn(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1625 | { |
| 1626 | unsigned long flags; |
| 1627 | struct bio *bio; |
| 1628 | |
| 1629 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1630 | bio = bio_dirty_list; |
| 1631 | bio_dirty_list = NULL; |
| 1632 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1633 | |
| 1634 | while (bio) { |
| 1635 | struct bio *next = bio->bi_private; |
| 1636 | |
| 1637 | bio_set_pages_dirty(bio); |
| 1638 | bio_release_pages(bio); |
| 1639 | bio_put(bio); |
| 1640 | bio = next; |
| 1641 | } |
| 1642 | } |
| 1643 | |
| 1644 | void bio_check_pages_dirty(struct bio *bio) |
| 1645 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1646 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1647 | int nr_clean_pages = 0; |
| 1648 | int i; |
| 1649 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1650 | bio_for_each_segment_all(bvec, bio, i) { |
| 1651 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1652 | |
| 1653 | if (PageDirty(page) || PageCompound(page)) { |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1654 | put_page(page); |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1655 | bvec->bv_page = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1656 | } else { |
| 1657 | nr_clean_pages++; |
| 1658 | } |
| 1659 | } |
| 1660 | |
| 1661 | if (nr_clean_pages) { |
| 1662 | unsigned long flags; |
| 1663 | |
| 1664 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1665 | bio->bi_private = bio_dirty_list; |
| 1666 | bio_dirty_list = bio; |
| 1667 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1668 | schedule_work(&bio_dirty_work); |
| 1669 | } else { |
| 1670 | bio_put(bio); |
| 1671 | } |
| 1672 | } |
| 1673 | |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1674 | void generic_start_io_acct(int rw, unsigned long sectors, |
| 1675 | struct hd_struct *part) |
| 1676 | { |
| 1677 | int cpu = part_stat_lock(); |
| 1678 | |
| 1679 | part_round_stats(cpu, part); |
| 1680 | part_stat_inc(cpu, part, ios[rw]); |
| 1681 | part_stat_add(cpu, part, sectors[rw], sectors); |
| 1682 | part_inc_in_flight(part, rw); |
| 1683 | |
| 1684 | part_stat_unlock(); |
| 1685 | } |
| 1686 | EXPORT_SYMBOL(generic_start_io_acct); |
| 1687 | |
| 1688 | void generic_end_io_acct(int rw, struct hd_struct *part, |
| 1689 | unsigned long start_time) |
| 1690 | { |
| 1691 | unsigned long duration = jiffies - start_time; |
| 1692 | int cpu = part_stat_lock(); |
| 1693 | |
| 1694 | part_stat_add(cpu, part, ticks[rw], duration); |
| 1695 | part_round_stats(cpu, part); |
| 1696 | part_dec_in_flight(part, rw); |
| 1697 | |
| 1698 | part_stat_unlock(); |
| 1699 | } |
| 1700 | EXPORT_SYMBOL(generic_end_io_acct); |
| 1701 | |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1702 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
| 1703 | void bio_flush_dcache_pages(struct bio *bi) |
| 1704 | { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1705 | struct bio_vec bvec; |
| 1706 | struct bvec_iter iter; |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1707 | |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1708 | bio_for_each_segment(bvec, bi, iter) |
| 1709 | flush_dcache_page(bvec.bv_page); |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1710 | } |
| 1711 | EXPORT_SYMBOL(bio_flush_dcache_pages); |
| 1712 | #endif |
| 1713 | |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1714 | static inline bool bio_remaining_done(struct bio *bio) |
| 1715 | { |
| 1716 | /* |
| 1717 | * If we're not chaining, then ->__bi_remaining is always 1 and |
| 1718 | * we always end io on the first invocation. |
| 1719 | */ |
| 1720 | if (!bio_flagged(bio, BIO_CHAIN)) |
| 1721 | return true; |
| 1722 | |
| 1723 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); |
| 1724 | |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1725 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1726 | bio_clear_flag(bio, BIO_CHAIN); |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1727 | return true; |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1728 | } |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1729 | |
| 1730 | return false; |
| 1731 | } |
| 1732 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1733 | /** |
| 1734 | * bio_endio - end I/O on a bio |
| 1735 | * @bio: bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1736 | * |
| 1737 | * Description: |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1738 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
| 1739 | * way to end I/O on a bio. No one should call bi_end_io() directly on a |
| 1740 | * bio unless they own it and thus know that it has an end_io function. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1741 | **/ |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1742 | void bio_endio(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1743 | { |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1744 | again: |
Christoph Hellwig | 2b88551 | 2016-03-11 17:34:53 +0100 | [diff] [blame] | 1745 | if (!bio_remaining_done(bio)) |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1746 | return; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1747 | |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1748 | /* |
| 1749 | * Need to have a real endio function for chained bios, otherwise |
| 1750 | * various corner cases will break (like stacking block devices that |
| 1751 | * save/restore bi_end_io) - however, we want to avoid unbounded |
| 1752 | * recursion and blowing the stack. Tail call optimization would |
| 1753 | * handle this, but compiling with frame pointers also disables |
| 1754 | * gcc's sibling call optimization. |
| 1755 | */ |
| 1756 | if (bio->bi_end_io == bio_chain_endio) { |
| 1757 | bio = __bio_chain_endio(bio); |
| 1758 | goto again; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1759 | } |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1760 | |
| 1761 | if (bio->bi_end_io) |
| 1762 | bio->bi_end_io(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1763 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1764 | EXPORT_SYMBOL(bio_endio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1765 | |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1766 | /** |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1767 | * bio_split - split a bio |
| 1768 | * @bio: bio to split |
| 1769 | * @sectors: number of sectors to split from the front of @bio |
| 1770 | * @gfp: gfp mask |
| 1771 | * @bs: bio set to allocate from |
| 1772 | * |
| 1773 | * Allocates and returns a new bio which represents @sectors from the start of |
| 1774 | * @bio, and updates @bio to represent the remaining sectors. |
| 1775 | * |
Martin K. Petersen | f3f5da6 | 2015-07-22 07:57:12 -0400 | [diff] [blame] | 1776 | * Unless this is a discard request the newly allocated bio will point |
| 1777 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that |
| 1778 | * @bio is not freed before the split. |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1779 | */ |
| 1780 | struct bio *bio_split(struct bio *bio, int sectors, |
| 1781 | gfp_t gfp, struct bio_set *bs) |
| 1782 | { |
| 1783 | struct bio *split = NULL; |
| 1784 | |
| 1785 | BUG_ON(sectors <= 0); |
| 1786 | BUG_ON(sectors >= bio_sectors(bio)); |
| 1787 | |
Martin K. Petersen | f3f5da6 | 2015-07-22 07:57:12 -0400 | [diff] [blame] | 1788 | /* |
| 1789 | * Discards need a mutable bio_vec to accommodate the payload |
| 1790 | * required by the DSM TRIM and UNMAP commands. |
| 1791 | */ |
Adrian Hunter | 7afafc8 | 2016-08-16 10:59:35 +0300 | [diff] [blame] | 1792 | if (bio_op(bio) == REQ_OP_DISCARD || bio_op(bio) == REQ_OP_SECURE_ERASE) |
Martin K. Petersen | f3f5da6 | 2015-07-22 07:57:12 -0400 | [diff] [blame] | 1793 | split = bio_clone_bioset(bio, gfp, bs); |
| 1794 | else |
| 1795 | split = bio_clone_fast(bio, gfp, bs); |
| 1796 | |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1797 | if (!split) |
| 1798 | return NULL; |
| 1799 | |
| 1800 | split->bi_iter.bi_size = sectors << 9; |
| 1801 | |
| 1802 | if (bio_integrity(split)) |
| 1803 | bio_integrity_trim(split, 0, sectors); |
| 1804 | |
| 1805 | bio_advance(bio, split->bi_iter.bi_size); |
| 1806 | |
| 1807 | return split; |
| 1808 | } |
| 1809 | EXPORT_SYMBOL(bio_split); |
| 1810 | |
Martin K. Petersen | ad3316b | 2008-10-01 22:42:53 -0400 | [diff] [blame] | 1811 | /** |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1812 | * bio_trim - trim a bio |
| 1813 | * @bio: bio to trim |
| 1814 | * @offset: number of sectors to trim from the front of @bio |
| 1815 | * @size: size we want to trim @bio to, in sectors |
| 1816 | */ |
| 1817 | void bio_trim(struct bio *bio, int offset, int size) |
| 1818 | { |
| 1819 | /* 'bio' is a cloned bio which we need to trim to match |
| 1820 | * the given offset and size. |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1821 | */ |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1822 | |
| 1823 | size <<= 9; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1824 | if (offset == 0 && size == bio->bi_iter.bi_size) |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1825 | return; |
| 1826 | |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1827 | bio_clear_flag(bio, BIO_SEG_VALID); |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1828 | |
| 1829 | bio_advance(bio, offset << 9); |
| 1830 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1831 | bio->bi_iter.bi_size = size; |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1832 | } |
| 1833 | EXPORT_SYMBOL_GPL(bio_trim); |
| 1834 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1835 | /* |
| 1836 | * create memory pools for biovec's in a bio_set. |
| 1837 | * use the global biovec slabs created for general use. |
| 1838 | */ |
Fabian Frederick | a6c39cb4f | 2014-04-22 15:09:05 -0600 | [diff] [blame] | 1839 | mempool_t *biovec_create_pool(int pool_entries) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1840 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 1841 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1842 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1843 | return mempool_create_slab_pool(pool_entries, bp->slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1844 | } |
| 1845 | |
| 1846 | void bioset_free(struct bio_set *bs) |
| 1847 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1848 | if (bs->rescue_workqueue) |
| 1849 | destroy_workqueue(bs->rescue_workqueue); |
| 1850 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1851 | if (bs->bio_pool) |
| 1852 | mempool_destroy(bs->bio_pool); |
| 1853 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1854 | if (bs->bvec_pool) |
| 1855 | mempool_destroy(bs->bvec_pool); |
| 1856 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 1857 | bioset_integrity_free(bs); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1858 | bio_put_slab(bs); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1859 | |
| 1860 | kfree(bs); |
| 1861 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1862 | EXPORT_SYMBOL(bioset_free); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1863 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1864 | static struct bio_set *__bioset_create(unsigned int pool_size, |
| 1865 | unsigned int front_pad, |
| 1866 | bool create_bvec_pool) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1867 | { |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 1868 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1869 | struct bio_set *bs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1870 | |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1871 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1872 | if (!bs) |
| 1873 | return NULL; |
| 1874 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1875 | bs->front_pad = front_pad; |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1876 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1877 | spin_lock_init(&bs->rescue_lock); |
| 1878 | bio_list_init(&bs->rescue_list); |
| 1879 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); |
| 1880 | |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 1881 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1882 | if (!bs->bio_slab) { |
| 1883 | kfree(bs); |
| 1884 | return NULL; |
| 1885 | } |
| 1886 | |
| 1887 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1888 | if (!bs->bio_pool) |
| 1889 | goto bad; |
| 1890 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1891 | if (create_bvec_pool) { |
| 1892 | bs->bvec_pool = biovec_create_pool(pool_size); |
| 1893 | if (!bs->bvec_pool) |
| 1894 | goto bad; |
| 1895 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1896 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1897 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); |
| 1898 | if (!bs->rescue_workqueue) |
| 1899 | goto bad; |
| 1900 | |
| 1901 | return bs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1902 | bad: |
| 1903 | bioset_free(bs); |
| 1904 | return NULL; |
| 1905 | } |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1906 | |
| 1907 | /** |
| 1908 | * bioset_create - Create a bio_set |
| 1909 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
| 1910 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1911 | * |
| 1912 | * Description: |
| 1913 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller |
| 1914 | * to ask for a number of bytes to be allocated in front of the bio. |
| 1915 | * Front pad allocation is useful for embedding the bio inside |
| 1916 | * another structure, to avoid allocating extra data to go with the bio. |
| 1917 | * Note that the bio must be embedded at the END of that structure always, |
| 1918 | * or things will break badly. |
| 1919 | */ |
| 1920 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) |
| 1921 | { |
| 1922 | return __bioset_create(pool_size, front_pad, true); |
| 1923 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1924 | EXPORT_SYMBOL(bioset_create); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1925 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1926 | /** |
| 1927 | * bioset_create_nobvec - Create a bio_set without bio_vec mempool |
| 1928 | * @pool_size: Number of bio to cache in the mempool |
| 1929 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1930 | * |
| 1931 | * Description: |
| 1932 | * Same functionality as bioset_create() except that mempool is not |
| 1933 | * created for bio_vecs. Saving some memory for bio_clone_fast() users. |
| 1934 | */ |
| 1935 | struct bio_set *bioset_create_nobvec(unsigned int pool_size, unsigned int front_pad) |
| 1936 | { |
| 1937 | return __bioset_create(pool_size, front_pad, false); |
| 1938 | } |
| 1939 | EXPORT_SYMBOL(bioset_create_nobvec); |
| 1940 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1941 | #ifdef CONFIG_BLK_CGROUP |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 1942 | |
| 1943 | /** |
| 1944 | * bio_associate_blkcg - associate a bio with the specified blkcg |
| 1945 | * @bio: target bio |
| 1946 | * @blkcg_css: css of the blkcg to associate |
| 1947 | * |
| 1948 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will |
| 1949 | * treat @bio as if it were issued by a task which belongs to the blkcg. |
| 1950 | * |
| 1951 | * This function takes an extra reference of @blkcg_css which will be put |
| 1952 | * when @bio is released. The caller must own @bio and is responsible for |
| 1953 | * synchronizing calls to this function. |
| 1954 | */ |
| 1955 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) |
| 1956 | { |
| 1957 | if (unlikely(bio->bi_css)) |
| 1958 | return -EBUSY; |
| 1959 | css_get(blkcg_css); |
| 1960 | bio->bi_css = blkcg_css; |
| 1961 | return 0; |
| 1962 | } |
Tejun Heo | 5aa2a96 | 2015-07-23 14:27:09 -0400 | [diff] [blame] | 1963 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 1964 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1965 | /** |
| 1966 | * bio_associate_current - associate a bio with %current |
| 1967 | * @bio: target bio |
| 1968 | * |
| 1969 | * Associate @bio with %current if it hasn't been associated yet. Block |
| 1970 | * layer will treat @bio as if it were issued by %current no matter which |
| 1971 | * task actually issues it. |
| 1972 | * |
| 1973 | * This function takes an extra reference of @task's io_context and blkcg |
| 1974 | * which will be put when @bio is released. The caller must own @bio, |
| 1975 | * ensure %current->io_context exists, and is responsible for synchronizing |
| 1976 | * calls to this function. |
| 1977 | */ |
| 1978 | int bio_associate_current(struct bio *bio) |
| 1979 | { |
| 1980 | struct io_context *ioc; |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1981 | |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 1982 | if (bio->bi_css) |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1983 | return -EBUSY; |
| 1984 | |
| 1985 | ioc = current->io_context; |
| 1986 | if (!ioc) |
| 1987 | return -ENOENT; |
| 1988 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1989 | get_io_context_active(ioc); |
| 1990 | bio->bi_ioc = ioc; |
Tejun Heo | c165b3e | 2015-08-18 14:55:29 -0700 | [diff] [blame] | 1991 | bio->bi_css = task_get_css(current, io_cgrp_id); |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1992 | return 0; |
| 1993 | } |
Tejun Heo | 5aa2a96 | 2015-07-23 14:27:09 -0400 | [diff] [blame] | 1994 | EXPORT_SYMBOL_GPL(bio_associate_current); |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1995 | |
| 1996 | /** |
| 1997 | * bio_disassociate_task - undo bio_associate_current() |
| 1998 | * @bio: target bio |
| 1999 | */ |
| 2000 | void bio_disassociate_task(struct bio *bio) |
| 2001 | { |
| 2002 | if (bio->bi_ioc) { |
| 2003 | put_io_context(bio->bi_ioc); |
| 2004 | bio->bi_ioc = NULL; |
| 2005 | } |
| 2006 | if (bio->bi_css) { |
| 2007 | css_put(bio->bi_css); |
| 2008 | bio->bi_css = NULL; |
| 2009 | } |
| 2010 | } |
| 2011 | |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 2012 | /** |
| 2013 | * bio_clone_blkcg_association - clone blkcg association from src to dst bio |
| 2014 | * @dst: destination bio |
| 2015 | * @src: source bio |
| 2016 | */ |
| 2017 | void bio_clone_blkcg_association(struct bio *dst, struct bio *src) |
| 2018 | { |
| 2019 | if (src->bi_css) |
| 2020 | WARN_ON(bio_associate_blkcg(dst, src->bi_css)); |
| 2021 | } |
| 2022 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2023 | #endif /* CONFIG_BLK_CGROUP */ |
| 2024 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2025 | static void __init biovec_init_slabs(void) |
| 2026 | { |
| 2027 | int i; |
| 2028 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 2029 | for (i = 0; i < BVEC_POOL_NR; i++) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2030 | int size; |
| 2031 | struct biovec_slab *bvs = bvec_slabs + i; |
| 2032 | |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2033 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
| 2034 | bvs->slab = NULL; |
| 2035 | continue; |
| 2036 | } |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2037 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2038 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
| 2039 | bvs->slab = kmem_cache_create(bvs->name, size, 0, |
Paul Mundt | 20c2df8 | 2007-07-20 10:11:58 +0900 | [diff] [blame] | 2040 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2041 | } |
| 2042 | } |
| 2043 | |
| 2044 | static int __init init_bio(void) |
| 2045 | { |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2046 | bio_slab_max = 2; |
| 2047 | bio_slab_nr = 0; |
| 2048 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); |
| 2049 | if (!bio_slabs) |
| 2050 | panic("bio: can't allocate bios\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2051 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 2052 | bio_integrity_init(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2053 | biovec_init_slabs(); |
| 2054 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2055 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2056 | if (!fs_bio_set) |
| 2057 | panic("bio: can't allocate bios\n"); |
| 2058 | |
Martin K. Petersen | a91a278 | 2011-03-17 11:11:05 +0100 | [diff] [blame] | 2059 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
| 2060 | panic("bio: can't create integrity pool\n"); |
| 2061 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2062 | return 0; |
| 2063 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2064 | subsys_initcall(init_bio); |