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 | */ |
Mikulas Patocka | 6bae912 | 2018-03-21 12:49:29 -0400 | [diff] [blame] | 45 | #define BV(x, n) { .nr_vecs = x, .name = "biovec-"#n } |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 46 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
Mikulas Patocka | 6bae912 | 2018-03-21 12:49:29 -0400 | [diff] [blame] | 47 | BV(1, 1), BV(4, 4), BV(16, 16), BV(64, 64), BV(128, 128), BV(BIO_MAX_PAGES, max), |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | { |
Greg Edwards | bbab86e | 2018-08-08 13:27:53 -0600 | [diff] [blame] | 158 | return bvec_slabs[--idx].nr_vecs; |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 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 | |
NeilBrown | 5959cde | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 375 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 376 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
NeilBrown | 5959cde | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 377 | current->bio_list[0] = nopunt; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 378 | |
NeilBrown | 5959cde | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 379 | bio_list_init(&nopunt); |
| 380 | while ((bio = bio_list_pop(¤t->bio_list[1]))) |
| 381 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
| 382 | current->bio_list[1] = nopunt; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 383 | |
| 384 | spin_lock(&bs->rescue_lock); |
| 385 | bio_list_merge(&bs->rescue_list, &punt); |
| 386 | spin_unlock(&bs->rescue_lock); |
| 387 | |
| 388 | queue_work(bs->rescue_workqueue, &bs->rescue_work); |
| 389 | } |
| 390 | |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 391 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 392 | * bio_alloc_bioset - allocate a bio for I/O |
| 393 | * @gfp_mask: the GFP_ mask given to the slab allocator |
| 394 | * @nr_iovecs: number of iovecs to pre-allocate |
Jaak Ristioja | db18efa | 2010-01-15 12:05:07 +0200 | [diff] [blame] | 395 | * @bs: the bio_set to allocate from. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 396 | * |
| 397 | * Description: |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 398 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
| 399 | * backed by the @bs's mempool. |
| 400 | * |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 401 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
| 402 | * always be able to allocate a bio. This is due to the mempool guarantees. |
| 403 | * To make this work, callers must never allocate more than 1 bio at a time |
| 404 | * from this pool. Callers that need to allocate more than 1 bio must always |
| 405 | * submit the previously allocated bio for IO before attempting to allocate |
| 406 | * 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] | 407 | * |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 408 | * Note that when running under generic_make_request() (i.e. any block |
| 409 | * driver), bios are not submitted until after you return - see the code in |
| 410 | * generic_make_request() that converts recursion into iteration, to prevent |
| 411 | * stack overflows. |
| 412 | * |
| 413 | * This would normally mean allocating multiple bios under |
| 414 | * generic_make_request() would be susceptible to deadlocks, but we have |
| 415 | * deadlock avoidance code that resubmits any blocked bios from a rescuer |
| 416 | * thread. |
| 417 | * |
| 418 | * However, we do not guarantee forward progress for allocations from other |
| 419 | * mempools. Doing multiple allocations from the same mempool under |
| 420 | * generic_make_request() should be avoided - instead, use bio_set's front_pad |
| 421 | * for per bio allocations. |
| 422 | * |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 423 | * RETURNS: |
| 424 | * Pointer to new bio on success, NULL on failure. |
| 425 | */ |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 426 | 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] | 427 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 428 | gfp_t saved_gfp = gfp_mask; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 429 | unsigned front_pad; |
| 430 | unsigned inline_vecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 431 | struct bio_vec *bvl = NULL; |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 432 | struct bio *bio; |
| 433 | void *p; |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 434 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 435 | if (!bs) { |
| 436 | if (nr_iovecs > UIO_MAXIOV) |
| 437 | return NULL; |
| 438 | |
| 439 | p = kmalloc(sizeof(struct bio) + |
| 440 | nr_iovecs * sizeof(struct bio_vec), |
| 441 | gfp_mask); |
| 442 | front_pad = 0; |
| 443 | inline_vecs = nr_iovecs; |
| 444 | } else { |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 445 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
| 446 | if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0)) |
| 447 | return NULL; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 448 | /* |
| 449 | * generic_make_request() converts recursion to iteration; this |
| 450 | * means if we're running beneath it, any bios we allocate and |
| 451 | * submit will not be submitted (and thus freed) until after we |
| 452 | * return. |
| 453 | * |
| 454 | * This exposes us to a potential deadlock if we allocate |
| 455 | * multiple bios from the same bio_set() while running |
| 456 | * underneath generic_make_request(). If we were to allocate |
| 457 | * multiple bios (say a stacking block driver that was splitting |
| 458 | * bios), we would deadlock if we exhausted the mempool's |
| 459 | * reserve. |
| 460 | * |
| 461 | * We solve this, and guarantee forward progress, with a rescuer |
| 462 | * workqueue per bio_set. If we go to allocate and there are |
| 463 | * bios on current->bio_list, we first try the allocation |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 464 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
| 465 | * bios we would be blocking to the rescuer workqueue before |
| 466 | * we retry with the original gfp_flags. |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 467 | */ |
| 468 | |
NeilBrown | 5959cde | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 469 | if (current->bio_list && |
| 470 | (!bio_list_empty(¤t->bio_list[0]) || |
| 471 | !bio_list_empty(¤t->bio_list[1]))) |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 472 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 473 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 474 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 475 | if (!p && gfp_mask != saved_gfp) { |
| 476 | punt_bios_to_rescuer(bs); |
| 477 | gfp_mask = saved_gfp; |
| 478 | p = mempool_alloc(bs->bio_pool, gfp_mask); |
| 479 | } |
| 480 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 481 | front_pad = bs->front_pad; |
| 482 | inline_vecs = BIO_INLINE_VECS; |
| 483 | } |
| 484 | |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 485 | if (unlikely(!p)) |
| 486 | return NULL; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 487 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 488 | bio = p + front_pad; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 489 | bio_init(bio); |
| 490 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 491 | if (nr_iovecs > inline_vecs) { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 492 | unsigned long idx = 0; |
| 493 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 494 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 495 | if (!bvl && gfp_mask != saved_gfp) { |
| 496 | punt_bios_to_rescuer(bs); |
| 497 | gfp_mask = saved_gfp; |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 498 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 499 | } |
| 500 | |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 501 | if (unlikely(!bvl)) |
| 502 | goto err_free; |
Kent Overstreet | a38352e | 2012-05-25 13:03:11 -0700 | [diff] [blame] | 503 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 504 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 505 | } else if (nr_iovecs) { |
| 506 | bvl = bio->bi_inline_vecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 507 | } |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 508 | |
| 509 | bio->bi_pool = bs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 510 | bio->bi_max_vecs = nr_iovecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 511 | bio->bi_io_vec = bvl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 512 | return bio; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 513 | |
| 514 | err_free: |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 515 | mempool_free(p, bs->bio_pool); |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 516 | return NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 517 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 518 | EXPORT_SYMBOL(bio_alloc_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 519 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 520 | void zero_fill_bio(struct bio *bio) |
| 521 | { |
| 522 | unsigned long flags; |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 523 | struct bio_vec bv; |
| 524 | struct bvec_iter iter; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 525 | |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 526 | bio_for_each_segment(bv, bio, iter) { |
| 527 | char *data = bvec_kmap_irq(&bv, &flags); |
| 528 | memset(data, 0, bv.bv_len); |
| 529 | flush_dcache_page(bv.bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 530 | bvec_kunmap_irq(data, &flags); |
| 531 | } |
| 532 | } |
| 533 | EXPORT_SYMBOL(zero_fill_bio); |
| 534 | |
| 535 | /** |
| 536 | * bio_put - release a reference to a bio |
| 537 | * @bio: bio to release reference to |
| 538 | * |
| 539 | * Description: |
| 540 | * 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] | 541 | * 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] | 542 | **/ |
| 543 | void bio_put(struct bio *bio) |
| 544 | { |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 545 | if (!bio_flagged(bio, BIO_REFFED)) |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 546 | bio_free(bio); |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 547 | else { |
| 548 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); |
| 549 | |
| 550 | /* |
| 551 | * last put frees it |
| 552 | */ |
| 553 | if (atomic_dec_and_test(&bio->__bi_cnt)) |
| 554 | bio_free(bio); |
| 555 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 556 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 557 | EXPORT_SYMBOL(bio_put); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 558 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 559 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 560 | { |
| 561 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) |
| 562 | blk_recount_segments(q, bio); |
| 563 | |
| 564 | return bio->bi_phys_segments; |
| 565 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 566 | EXPORT_SYMBOL(bio_phys_segments); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 567 | |
Neeraj Soni | 36c6512 | 2018-04-18 21:04:46 +0530 | [diff] [blame] | 568 | static inline void bio_clone_crypt_key(struct bio *dst, const struct bio *src) |
| 569 | { |
| 570 | #ifdef CONFIG_PFK |
Neeraj Soni | 36c6512 | 2018-04-18 21:04:46 +0530 | [diff] [blame] | 571 | dst->bi_iter.bi_dun = src->bi_iter.bi_dun; |
Jaegeuk Kim | 3bd012f | 2018-04-20 19:26:09 -0700 | [diff] [blame] | 572 | #ifdef CONFIG_DM_DEFAULT_KEY |
| 573 | dst->bi_crypt_key = src->bi_crypt_key; |
| 574 | dst->bi_crypt_skip = src->bi_crypt_skip; |
| 575 | #endif |
Neeraj Soni | 36c6512 | 2018-04-18 21:04:46 +0530 | [diff] [blame] | 576 | dst->bi_dio_inode = src->bi_dio_inode; |
Jaegeuk Kim | 5190c81 | 2018-04-22 16:05:02 -0700 | [diff] [blame] | 577 | #endif |
Neeraj Soni | 36c6512 | 2018-04-18 21:04:46 +0530 | [diff] [blame] | 578 | } |
| 579 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 580 | /** |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 581 | * __bio_clone_fast - clone a bio that shares the original bio's biovec |
| 582 | * @bio: destination bio |
| 583 | * @bio_src: bio to clone |
| 584 | * |
| 585 | * Clone a &bio. Caller will own the returned bio, but not |
| 586 | * the actual data it points to. Reference count of returned |
| 587 | * bio will be one. |
| 588 | * |
| 589 | * Caller must ensure that @bio_src is not freed before @bio. |
| 590 | */ |
| 591 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) |
| 592 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 593 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 594 | |
| 595 | /* |
| 596 | * most users will be overriding ->bi_bdev with a new target, |
| 597 | * so we don't set nor calculate new physical/hw segment counts here |
| 598 | */ |
| 599 | bio->bi_bdev = bio_src->bi_bdev; |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 600 | bio_set_flag(bio, BIO_CLONED); |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 601 | bio->bi_opf = bio_src->bi_opf; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 602 | bio->bi_iter = bio_src->bi_iter; |
| 603 | bio->bi_io_vec = bio_src->bi_io_vec; |
Jaegeuk Kim | 5190c81 | 2018-04-22 16:05:02 -0700 | [diff] [blame] | 604 | #ifdef CONFIG_PFK |
Neeraj Soni | c692cb9 | 2018-04-18 17:20:22 +0530 | [diff] [blame] | 605 | bio->bi_dio_inode = bio_src->bi_dio_inode; |
Jaegeuk Kim | 5190c81 | 2018-04-22 16:05:02 -0700 | [diff] [blame] | 606 | #endif |
Neeraj Soni | 36c6512 | 2018-04-18 21:04:46 +0530 | [diff] [blame] | 607 | bio_clone_crypt_key(bio, bio_src); |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 608 | bio_clone_blkcg_association(bio, bio_src); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 609 | } |
| 610 | EXPORT_SYMBOL(__bio_clone_fast); |
| 611 | |
| 612 | /** |
| 613 | * bio_clone_fast - clone a bio that shares the original bio's biovec |
| 614 | * @bio: bio to clone |
| 615 | * @gfp_mask: allocation priority |
| 616 | * @bs: bio_set to allocate from |
| 617 | * |
| 618 | * Like __bio_clone_fast, only also allocates the returned bio |
| 619 | */ |
| 620 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) |
| 621 | { |
| 622 | struct bio *b; |
| 623 | |
| 624 | b = bio_alloc_bioset(gfp_mask, 0, bs); |
| 625 | if (!b) |
| 626 | return NULL; |
| 627 | |
| 628 | __bio_clone_fast(b, bio); |
| 629 | |
| 630 | if (bio_integrity(bio)) { |
| 631 | int ret; |
| 632 | |
| 633 | ret = bio_integrity_clone(b, bio, gfp_mask); |
| 634 | |
| 635 | if (ret < 0) { |
| 636 | bio_put(b); |
| 637 | return NULL; |
| 638 | } |
| 639 | } |
| 640 | |
| 641 | return b; |
| 642 | } |
| 643 | EXPORT_SYMBOL(bio_clone_fast); |
| 644 | |
| 645 | /** |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 646 | * bio_clone_bioset - clone a bio |
| 647 | * @bio_src: bio to clone |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 648 | * @gfp_mask: allocation priority |
Kent Overstreet | bf800ef | 2012-09-06 15:35:02 -0700 | [diff] [blame] | 649 | * @bs: bio_set to allocate from |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 650 | * |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 651 | * Clone bio. Caller will own the returned bio, but not the actual data it |
| 652 | * points to. Reference count of returned bio will be one. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 653 | */ |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 654 | 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] | 655 | struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 656 | { |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 657 | struct bvec_iter iter; |
| 658 | struct bio_vec bv; |
| 659 | struct bio *bio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 660 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 661 | /* |
| 662 | * Pre immutable biovecs, __bio_clone() used to just do a memcpy from |
| 663 | * bio_src->bi_io_vec to bio->bi_io_vec. |
| 664 | * |
| 665 | * We can't do that anymore, because: |
| 666 | * |
| 667 | * - The point of cloning the biovec is to produce a bio with a biovec |
| 668 | * the caller can modify: bi_idx and bi_bvec_done should be 0. |
| 669 | * |
| 670 | * - The original bio could've had more than BIO_MAX_PAGES biovecs; if |
| 671 | * we tried to clone the whole thing bio_alloc_bioset() would fail. |
| 672 | * But the clone should succeed as long as the number of biovecs we |
| 673 | * actually need to allocate is fewer than BIO_MAX_PAGES. |
| 674 | * |
| 675 | * - Lastly, bi_vcnt should not be looked at or relied upon by code |
| 676 | * that does not own the bio - reason being drivers don't use it for |
| 677 | * iterating over the biovec anymore, so expecting it to be kept up |
| 678 | * to date (i.e. for clones that share the parent biovec) is just |
| 679 | * asking for trouble and would force extra work on |
| 680 | * __bio_clone_fast() anyways. |
| 681 | */ |
| 682 | |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 683 | bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs); |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 684 | if (!bio) |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 685 | return NULL; |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 686 | bio->bi_bdev = bio_src->bi_bdev; |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 687 | bio->bi_opf = bio_src->bi_opf; |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 688 | bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector; |
| 689 | bio->bi_iter.bi_size = bio_src->bi_iter.bi_size; |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 690 | |
Adrian Hunter | 7afafc8 | 2016-08-16 10:59:35 +0300 | [diff] [blame] | 691 | switch (bio_op(bio)) { |
| 692 | case REQ_OP_DISCARD: |
| 693 | case REQ_OP_SECURE_ERASE: |
| 694 | break; |
| 695 | case REQ_OP_WRITE_SAME: |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 696 | 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] | 697 | break; |
| 698 | default: |
| 699 | bio_for_each_segment(bv, bio_src, iter) |
| 700 | bio->bi_io_vec[bio->bi_vcnt++] = bv; |
| 701 | break; |
Kent Overstreet | 8423ae3 | 2014-02-10 17:45:50 -0800 | [diff] [blame] | 702 | } |
| 703 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 704 | if (bio_integrity(bio_src)) { |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 705 | int ret; |
| 706 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 707 | ret = bio_integrity_clone(bio, bio_src, gfp_mask); |
Li Zefan | 059ea33 | 2009-03-09 10:42:45 +0100 | [diff] [blame] | 708 | if (ret < 0) { |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 709 | bio_put(bio); |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 710 | return NULL; |
Li Zefan | 059ea33 | 2009-03-09 10:42:45 +0100 | [diff] [blame] | 711 | } |
Peter Osterlund | 3676347 | 2005-09-06 15:16:42 -0700 | [diff] [blame] | 712 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 713 | |
Neeraj Soni | 69e981b | 2018-08-16 15:23:25 +0530 | [diff] [blame] | 714 | bio_clone_crypt_key(bio, bio_src); |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 715 | bio_clone_blkcg_association(bio, bio_src); |
| 716 | |
Kent Overstreet | bdb5320 | 2013-11-23 17:26:46 -0800 | [diff] [blame] | 717 | return bio; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 718 | } |
Kent Overstreet | bf800ef | 2012-09-06 15:35:02 -0700 | [diff] [blame] | 719 | EXPORT_SYMBOL(bio_clone_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 720 | |
| 721 | /** |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 722 | * bio_add_pc_page - attempt to add page to bio |
| 723 | * @q: the target queue |
| 724 | * @bio: destination bio |
| 725 | * @page: page to add |
| 726 | * @len: vec entry length |
| 727 | * @offset: vec entry offset |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 728 | * |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 729 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
| 730 | * number of reasons, such as the bio being full or target block device |
| 731 | * limitations. The target block device must allow bio's up to PAGE_SIZE, |
| 732 | * so it is always possible to add a single page to an empty bio. |
| 733 | * |
| 734 | * This should only be used by REQ_PC bios. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 735 | */ |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 736 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
| 737 | *page, unsigned int len, unsigned int offset) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 738 | { |
| 739 | int retried_segments = 0; |
| 740 | struct bio_vec *bvec; |
| 741 | |
| 742 | /* |
| 743 | * cloned bio must not modify vec list |
| 744 | */ |
| 745 | if (unlikely(bio_flagged(bio, BIO_CLONED))) |
| 746 | return 0; |
| 747 | |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 748 | 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] | 749 | return 0; |
| 750 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 751 | /* |
| 752 | * For filesystems with a blocksize smaller than the pagesize |
| 753 | * we will often be called with the same page as last time and |
| 754 | * a consecutive offset. Optimize this special case. |
| 755 | */ |
| 756 | if (bio->bi_vcnt > 0) { |
| 757 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 758 | |
| 759 | if (page == prev->bv_page && |
| 760 | offset == prev->bv_offset + prev->bv_len) { |
| 761 | prev->bv_len += len; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 762 | bio->bi_iter.bi_size += len; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 763 | goto done; |
| 764 | } |
Jens Axboe | 66cb45a | 2014-06-24 16:22:24 -0600 | [diff] [blame] | 765 | |
| 766 | /* |
| 767 | * If the queue doesn't support SG gaps and adding this |
| 768 | * offset would create a gap, disallow it. |
| 769 | */ |
Keith Busch | 03100aa | 2015-08-19 14:24:05 -0700 | [diff] [blame] | 770 | if (bvec_gap_to_prev(q, prev, offset)) |
Jens Axboe | 66cb45a | 2014-06-24 16:22:24 -0600 | [diff] [blame] | 771 | return 0; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 772 | } |
| 773 | |
| 774 | if (bio->bi_vcnt >= bio->bi_max_vecs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 775 | return 0; |
| 776 | |
| 777 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 778 | * setup the new entry, we might clear it again later if we |
| 779 | * cannot add the page |
| 780 | */ |
| 781 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; |
| 782 | bvec->bv_page = page; |
| 783 | bvec->bv_len = len; |
| 784 | bvec->bv_offset = offset; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 785 | bio->bi_vcnt++; |
| 786 | bio->bi_phys_segments++; |
| 787 | bio->bi_iter.bi_size += len; |
| 788 | |
| 789 | /* |
| 790 | * Perform a recount if the number of segments is greater |
| 791 | * than queue_max_segments(q). |
| 792 | */ |
| 793 | |
| 794 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
| 795 | |
| 796 | if (retried_segments) |
| 797 | goto failed; |
| 798 | |
| 799 | retried_segments = 1; |
| 800 | blk_recount_segments(q, bio); |
| 801 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 802 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 803 | /* If we may be able to merge these biovecs, force a recount */ |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 804 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 805 | bio_clear_flag(bio, BIO_SEG_VALID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 806 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 807 | done: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 808 | return len; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 809 | |
| 810 | failed: |
| 811 | bvec->bv_page = NULL; |
| 812 | bvec->bv_len = 0; |
| 813 | bvec->bv_offset = 0; |
| 814 | bio->bi_vcnt--; |
| 815 | bio->bi_iter.bi_size -= len; |
| 816 | blk_recount_segments(q, bio); |
| 817 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 818 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 819 | EXPORT_SYMBOL(bio_add_pc_page); |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 820 | |
| 821 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 822 | * bio_add_page - attempt to add page to bio |
| 823 | * @bio: destination bio |
| 824 | * @page: page to add |
| 825 | * @len: vec entry length |
| 826 | * @offset: vec entry offset |
| 827 | * |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 828 | * Attempt to add a page to the bio_vec maplist. This will only fail |
| 829 | * 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] | 830 | */ |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 831 | int bio_add_page(struct bio *bio, struct page *page, |
| 832 | unsigned int len, unsigned int offset) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 833 | { |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 834 | struct bio_vec *bv; |
Jens Axboe | 762380a | 2014-06-05 13:38:39 -0600 | [diff] [blame] | 835 | |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 836 | /* |
| 837 | * cloned bio must not modify vec list |
| 838 | */ |
| 839 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
| 840 | return 0; |
Jens Axboe | 58a4915 | 2014-06-10 12:53:56 -0600 | [diff] [blame] | 841 | |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 842 | /* |
| 843 | * For filesystems with a blocksize smaller than the pagesize |
| 844 | * we will often be called with the same page as last time and |
| 845 | * a consecutive offset. Optimize this special case. |
| 846 | */ |
| 847 | if (bio->bi_vcnt > 0) { |
| 848 | bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 849 | |
| 850 | if (page == bv->bv_page && |
| 851 | offset == bv->bv_offset + bv->bv_len) { |
| 852 | bv->bv_len += len; |
| 853 | goto done; |
| 854 | } |
| 855 | } |
| 856 | |
| 857 | if (bio->bi_vcnt >= bio->bi_max_vecs) |
| 858 | return 0; |
| 859 | |
| 860 | bv = &bio->bi_io_vec[bio->bi_vcnt]; |
| 861 | bv->bv_page = page; |
| 862 | bv->bv_len = len; |
| 863 | bv->bv_offset = offset; |
| 864 | |
| 865 | bio->bi_vcnt++; |
| 866 | done: |
| 867 | bio->bi_iter.bi_size += len; |
| 868 | return len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 869 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 870 | EXPORT_SYMBOL(bio_add_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 871 | |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 872 | struct submit_bio_ret { |
| 873 | struct completion event; |
| 874 | int error; |
| 875 | }; |
| 876 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 877 | static void submit_bio_wait_endio(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 878 | { |
| 879 | struct submit_bio_ret *ret = bio->bi_private; |
| 880 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 881 | ret->error = bio->bi_error; |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 882 | complete(&ret->event); |
| 883 | } |
| 884 | |
| 885 | /** |
| 886 | * submit_bio_wait - submit a bio, and wait until it completes |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 887 | * @bio: The &struct bio which describes the I/O |
| 888 | * |
| 889 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from |
| 890 | * bio_endio() on failure. |
| 891 | */ |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 892 | int submit_bio_wait(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 893 | { |
| 894 | struct submit_bio_ret ret; |
| 895 | |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 896 | init_completion(&ret.event); |
| 897 | bio->bi_private = &ret; |
| 898 | bio->bi_end_io = submit_bio_wait_endio; |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 899 | bio->bi_opf |= REQ_SYNC; |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 900 | submit_bio(bio); |
Stephane Gasparini | d57d611 | 2016-02-09 17:07:38 +0100 | [diff] [blame] | 901 | wait_for_completion_io(&ret.event); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 902 | |
| 903 | return ret.error; |
| 904 | } |
| 905 | EXPORT_SYMBOL(submit_bio_wait); |
| 906 | |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 907 | /** |
| 908 | * bio_advance - increment/complete a bio by some number of bytes |
| 909 | * @bio: bio to advance |
| 910 | * @bytes: number of bytes to complete |
| 911 | * |
| 912 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to |
| 913 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will |
| 914 | * be updated on the last bvec as well. |
| 915 | * |
| 916 | * @bio will then represent the remaining, uncompleted portion of the io. |
| 917 | */ |
| 918 | void bio_advance(struct bio *bio, unsigned bytes) |
| 919 | { |
| 920 | if (bio_integrity(bio)) |
| 921 | bio_integrity_advance(bio, bytes); |
| 922 | |
Kent Overstreet | 4550dd6 | 2013-08-07 14:26:21 -0700 | [diff] [blame] | 923 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 924 | } |
| 925 | EXPORT_SYMBOL(bio_advance); |
| 926 | |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 927 | /** |
Kent Overstreet | a078760 | 2012-09-10 14:03:28 -0700 | [diff] [blame] | 928 | * bio_alloc_pages - allocates a single page for each bvec in a bio |
| 929 | * @bio: bio to allocate pages for |
| 930 | * @gfp_mask: flags for allocation |
| 931 | * |
| 932 | * Allocates pages up to @bio->bi_vcnt. |
| 933 | * |
| 934 | * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are |
| 935 | * freed. |
| 936 | */ |
| 937 | int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask) |
| 938 | { |
| 939 | int i; |
| 940 | struct bio_vec *bv; |
| 941 | |
| 942 | bio_for_each_segment_all(bv, bio, i) { |
| 943 | bv->bv_page = alloc_page(gfp_mask); |
| 944 | if (!bv->bv_page) { |
| 945 | while (--bv >= bio->bi_io_vec) |
| 946 | __free_page(bv->bv_page); |
| 947 | return -ENOMEM; |
| 948 | } |
| 949 | } |
| 950 | |
| 951 | return 0; |
| 952 | } |
| 953 | EXPORT_SYMBOL(bio_alloc_pages); |
| 954 | |
| 955 | /** |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 956 | * bio_copy_data - copy contents of data buffers from one chain of bios to |
| 957 | * another |
| 958 | * @src: source bio list |
| 959 | * @dst: destination bio list |
| 960 | * |
| 961 | * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats |
| 962 | * @src and @dst as linked lists of bios. |
| 963 | * |
| 964 | * Stops when it reaches the end of either @src or @dst - that is, copies |
| 965 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). |
| 966 | */ |
| 967 | void bio_copy_data(struct bio *dst, struct bio *src) |
| 968 | { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 969 | struct bvec_iter src_iter, dst_iter; |
| 970 | struct bio_vec src_bv, dst_bv; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 971 | void *src_p, *dst_p; |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 972 | unsigned bytes; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 973 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 974 | src_iter = src->bi_iter; |
| 975 | dst_iter = dst->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 976 | |
| 977 | while (1) { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 978 | if (!src_iter.bi_size) { |
| 979 | src = src->bi_next; |
| 980 | if (!src) |
| 981 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 982 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 983 | src_iter = src->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 984 | } |
| 985 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 986 | if (!dst_iter.bi_size) { |
| 987 | dst = dst->bi_next; |
| 988 | if (!dst) |
| 989 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 990 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 991 | dst_iter = dst->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 992 | } |
| 993 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 994 | src_bv = bio_iter_iovec(src, src_iter); |
| 995 | dst_bv = bio_iter_iovec(dst, dst_iter); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 996 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 997 | bytes = min(src_bv.bv_len, dst_bv.bv_len); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 998 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 999 | src_p = kmap_atomic(src_bv.bv_page); |
| 1000 | dst_p = kmap_atomic(dst_bv.bv_page); |
| 1001 | |
| 1002 | memcpy(dst_p + dst_bv.bv_offset, |
| 1003 | src_p + src_bv.bv_offset, |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1004 | bytes); |
| 1005 | |
| 1006 | kunmap_atomic(dst_p); |
| 1007 | kunmap_atomic(src_p); |
| 1008 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1009 | bio_advance_iter(src, &src_iter, bytes); |
| 1010 | bio_advance_iter(dst, &dst_iter, bytes); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1011 | } |
| 1012 | } |
| 1013 | EXPORT_SYMBOL(bio_copy_data); |
| 1014 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1015 | struct bio_map_data { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1016 | int is_our_pages; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1017 | struct iov_iter iter; |
| 1018 | struct iovec iov[]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1019 | }; |
| 1020 | |
Fabian Frederick | 7410b3c | 2014-04-22 15:09:07 -0600 | [diff] [blame] | 1021 | 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] | 1022 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1023 | { |
Jens Axboe | f3f63c1 | 2010-10-29 11:46:56 -0600 | [diff] [blame] | 1024 | if (iov_count > UIO_MAXIOV) |
| 1025 | return NULL; |
| 1026 | |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1027 | return kmalloc(sizeof(struct bio_map_data) + |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1028 | sizeof(struct iovec) * iov_count, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1029 | } |
| 1030 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1031 | /** |
| 1032 | * bio_copy_from_iter - copy all pages from iov_iter to bio |
| 1033 | * @bio: The &struct bio which describes the I/O as destination |
| 1034 | * @iter: iov_iter as source |
| 1035 | * |
| 1036 | * Copy all pages from iov_iter to bio. |
| 1037 | * Returns 0 on success, or error on failure. |
| 1038 | */ |
| 1039 | 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] | 1040 | { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1041 | int i; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1042 | struct bio_vec *bvec; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1043 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1044 | bio_for_each_segment_all(bvec, bio, i) { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1045 | ssize_t ret; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1046 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1047 | ret = copy_page_from_iter(bvec->bv_page, |
| 1048 | bvec->bv_offset, |
| 1049 | bvec->bv_len, |
| 1050 | &iter); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1051 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1052 | if (!iov_iter_count(&iter)) |
| 1053 | break; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1054 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1055 | if (ret < bvec->bv_len) |
| 1056 | return -EFAULT; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1057 | } |
| 1058 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1059 | return 0; |
| 1060 | } |
| 1061 | |
| 1062 | /** |
| 1063 | * bio_copy_to_iter - copy all pages from bio to iov_iter |
| 1064 | * @bio: The &struct bio which describes the I/O as source |
| 1065 | * @iter: iov_iter as destination |
| 1066 | * |
| 1067 | * Copy all pages from bio to iov_iter. |
| 1068 | * Returns 0 on success, or error on failure. |
| 1069 | */ |
| 1070 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) |
| 1071 | { |
| 1072 | int i; |
| 1073 | struct bio_vec *bvec; |
| 1074 | |
| 1075 | bio_for_each_segment_all(bvec, bio, i) { |
| 1076 | ssize_t ret; |
| 1077 | |
| 1078 | ret = copy_page_to_iter(bvec->bv_page, |
| 1079 | bvec->bv_offset, |
| 1080 | bvec->bv_len, |
| 1081 | &iter); |
| 1082 | |
| 1083 | if (!iov_iter_count(&iter)) |
| 1084 | break; |
| 1085 | |
| 1086 | if (ret < bvec->bv_len) |
| 1087 | return -EFAULT; |
| 1088 | } |
| 1089 | |
| 1090 | return 0; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1091 | } |
| 1092 | |
Guoqing Jiang | 491221f | 2016-09-22 03:10:01 -0400 | [diff] [blame] | 1093 | void bio_free_pages(struct bio *bio) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1094 | { |
| 1095 | struct bio_vec *bvec; |
| 1096 | int i; |
| 1097 | |
| 1098 | bio_for_each_segment_all(bvec, bio, i) |
| 1099 | __free_page(bvec->bv_page); |
| 1100 | } |
Guoqing Jiang | 491221f | 2016-09-22 03:10:01 -0400 | [diff] [blame] | 1101 | EXPORT_SYMBOL(bio_free_pages); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1102 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1103 | /** |
| 1104 | * bio_uncopy_user - finish previously mapped bio |
| 1105 | * @bio: bio being terminated |
| 1106 | * |
Christoph Hellwig | ddad8dd | 2015-01-18 16:16:29 +0100 | [diff] [blame] | 1107 | * Free pages allocated from bio_copy_user_iov() and write back data |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1108 | * to user space in case of a read. |
| 1109 | */ |
| 1110 | int bio_uncopy_user(struct bio *bio) |
| 1111 | { |
| 1112 | struct bio_map_data *bmd = bio->bi_private; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1113 | int ret = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1114 | |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1115 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
| 1116 | /* |
| 1117 | * if we're in a workqueue, the request is orphaned, so |
Hannes Reinecke | 2d99b55 | 2016-02-12 09:39:15 +0100 | [diff] [blame] | 1118 | * don't copy into a random user address space, just free |
| 1119 | * and return -EINTR so user space doesn't expect any data. |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1120 | */ |
Hannes Reinecke | 2d99b55 | 2016-02-12 09:39:15 +0100 | [diff] [blame] | 1121 | if (!current->mm) |
| 1122 | ret = -EINTR; |
| 1123 | else if (bio_data_dir(bio) == READ) |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1124 | ret = bio_copy_to_iter(bio, bmd->iter); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1125 | if (bmd->is_our_pages) |
| 1126 | bio_free_pages(bio); |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1127 | } |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1128 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1129 | bio_put(bio); |
| 1130 | return ret; |
| 1131 | } |
| 1132 | |
| 1133 | /** |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1134 | * bio_copy_user_iov - copy user data to bio |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1135 | * @q: destination block queue |
| 1136 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
| 1137 | * @iter: iovec iterator |
| 1138 | * @gfp_mask: memory allocation flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1139 | * |
| 1140 | * Prepares and returns a bio for indirect user io, bouncing data |
| 1141 | * to/from kernel pages as necessary. Must be paired with |
| 1142 | * call bio_uncopy_user() on io completion. |
| 1143 | */ |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1144 | struct bio *bio_copy_user_iov(struct request_queue *q, |
| 1145 | struct rq_map_data *map_data, |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1146 | const struct iov_iter *iter, |
| 1147 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1148 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1149 | struct bio_map_data *bmd; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1150 | struct page *page; |
| 1151 | struct bio *bio; |
| 1152 | int i, ret; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1153 | int nr_pages = 0; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1154 | unsigned int len = iter->count; |
Geliang Tang | bd5cece | 2015-11-21 17:27:31 +0800 | [diff] [blame] | 1155 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1156 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1157 | for (i = 0; i < iter->nr_segs; i++) { |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1158 | unsigned long uaddr; |
| 1159 | unsigned long end; |
| 1160 | unsigned long start; |
| 1161 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1162 | uaddr = (unsigned long) iter->iov[i].iov_base; |
| 1163 | end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1) |
| 1164 | >> PAGE_SHIFT; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1165 | start = uaddr >> PAGE_SHIFT; |
| 1166 | |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1167 | /* |
| 1168 | * Overflow, abort |
| 1169 | */ |
| 1170 | if (end < start) |
| 1171 | return ERR_PTR(-EINVAL); |
| 1172 | |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1173 | nr_pages += end - start; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1174 | } |
| 1175 | |
FUJITA Tomonori | 6983872 | 2009-04-28 20:24:29 +0200 | [diff] [blame] | 1176 | if (offset) |
| 1177 | nr_pages++; |
| 1178 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1179 | bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1180 | if (!bmd) |
| 1181 | return ERR_PTR(-ENOMEM); |
| 1182 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1183 | /* |
| 1184 | * We need to do a deep copy of the iov_iter including the iovecs. |
| 1185 | * The caller provided iov might point to an on-stack or otherwise |
| 1186 | * shortlived one. |
| 1187 | */ |
| 1188 | bmd->is_our_pages = map_data ? 0 : 1; |
| 1189 | memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs); |
Al Viro | ed35ded | 2017-09-24 10:21:15 -0400 | [diff] [blame] | 1190 | bmd->iter = *iter; |
| 1191 | bmd->iter.iov = bmd->iov; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1192 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1193 | ret = -ENOMEM; |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1194 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1195 | if (!bio) |
| 1196 | goto out_bmd; |
| 1197 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1198 | if (iter->type & WRITE) |
Mike Christie | 95fe6c1 | 2016-06-05 14:31:48 -0500 | [diff] [blame] | 1199 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1200 | |
| 1201 | ret = 0; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1202 | |
| 1203 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1204 | nr_pages = 1 << map_data->page_order; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1205 | i = map_data->offset / PAGE_SIZE; |
| 1206 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1207 | while (len) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1208 | unsigned int bytes = PAGE_SIZE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1209 | |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1210 | bytes -= offset; |
| 1211 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1212 | if (bytes > len) |
| 1213 | bytes = len; |
| 1214 | |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1215 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1216 | if (i == map_data->nr_entries * nr_pages) { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1217 | ret = -ENOMEM; |
| 1218 | break; |
| 1219 | } |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1220 | |
| 1221 | page = map_data->pages[i / nr_pages]; |
| 1222 | page += (i % nr_pages); |
| 1223 | |
| 1224 | i++; |
| 1225 | } else { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1226 | page = alloc_page(q->bounce_gfp | gfp_mask); |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1227 | if (!page) { |
| 1228 | ret = -ENOMEM; |
| 1229 | break; |
| 1230 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1231 | } |
| 1232 | |
Jérôme Glisse | 056066d | 2019-04-10 16:27:51 -0400 | [diff] [blame] | 1233 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) { |
| 1234 | if (!map_data) |
| 1235 | __free_page(page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1236 | break; |
Jérôme Glisse | 056066d | 2019-04-10 16:27:51 -0400 | [diff] [blame] | 1237 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1238 | |
| 1239 | len -= bytes; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1240 | offset = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1241 | } |
| 1242 | |
| 1243 | if (ret) |
| 1244 | goto cleanup; |
| 1245 | |
| 1246 | /* |
| 1247 | * success |
| 1248 | */ |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1249 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
FUJITA Tomonori | ecb554a | 2009-07-09 14:46:53 +0200 | [diff] [blame] | 1250 | (map_data && map_data->from_user)) { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1251 | ret = bio_copy_from_iter(bio, *iter); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1252 | if (ret) |
| 1253 | goto cleanup; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1254 | } |
| 1255 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1256 | bio->bi_private = bmd; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1257 | return bio; |
| 1258 | cleanup: |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1259 | if (!map_data) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1260 | bio_free_pages(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1261 | bio_put(bio); |
| 1262 | out_bmd: |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1263 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1264 | return ERR_PTR(ret); |
| 1265 | } |
| 1266 | |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1267 | /** |
| 1268 | * bio_map_user_iov - map user iovec into bio |
| 1269 | * @q: the struct request_queue for the bio |
| 1270 | * @iter: iovec iterator |
| 1271 | * @gfp_mask: memory allocation flags |
| 1272 | * |
| 1273 | * Map the user space address into a bio suitable for io to a block |
| 1274 | * device. Returns an error pointer in case of error. |
| 1275 | */ |
| 1276 | struct bio *bio_map_user_iov(struct request_queue *q, |
| 1277 | const struct iov_iter *iter, |
| 1278 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1279 | { |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1280 | int j; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1281 | int nr_pages = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1282 | struct page **pages; |
| 1283 | struct bio *bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1284 | int cur_page = 0; |
| 1285 | int ret, offset; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1286 | struct iov_iter i; |
| 1287 | struct iovec iov; |
Al Viro | e67dfe7 | 2017-09-23 15:51:23 -0400 | [diff] [blame] | 1288 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1289 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1290 | iov_for_each(iov, i, *iter) { |
| 1291 | unsigned long uaddr = (unsigned long) iov.iov_base; |
| 1292 | unsigned long len = iov.iov_len; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1293 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1294 | unsigned long start = uaddr >> PAGE_SHIFT; |
| 1295 | |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1296 | /* |
| 1297 | * Overflow, abort |
| 1298 | */ |
| 1299 | if (end < start) |
| 1300 | return ERR_PTR(-EINVAL); |
| 1301 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1302 | nr_pages += end - start; |
| 1303 | /* |
Linus Walleij | a441b0d | 2016-09-14 14:32:52 +0200 | [diff] [blame] | 1304 | * buffer must be aligned to at least logical block size for now |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1305 | */ |
Mike Christie | ad2d722 | 2006-12-01 10:40:20 +0100 | [diff] [blame] | 1306 | if (uaddr & queue_dma_alignment(q)) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1307 | return ERR_PTR(-EINVAL); |
| 1308 | } |
| 1309 | |
| 1310 | if (!nr_pages) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1311 | return ERR_PTR(-EINVAL); |
| 1312 | |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1313 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1314 | if (!bio) |
| 1315 | return ERR_PTR(-ENOMEM); |
| 1316 | |
| 1317 | ret = -ENOMEM; |
FUJITA Tomonori | a3bce90 | 2008-08-28 16:17:05 +0900 | [diff] [blame] | 1318 | pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1319 | if (!pages) |
| 1320 | goto out; |
| 1321 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1322 | iov_for_each(iov, i, *iter) { |
| 1323 | unsigned long uaddr = (unsigned long) iov.iov_base; |
| 1324 | unsigned long len = iov.iov_len; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1325 | unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1326 | unsigned long start = uaddr >> PAGE_SHIFT; |
| 1327 | const int local_nr_pages = end - start; |
| 1328 | const int page_limit = cur_page + local_nr_pages; |
Jens Axboe | cb4644c | 2010-11-10 14:36:25 +0100 | [diff] [blame] | 1329 | |
Nick Piggin | f5dd33c | 2008-07-25 19:45:25 -0700 | [diff] [blame] | 1330 | ret = get_user_pages_fast(uaddr, local_nr_pages, |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1331 | (iter->type & WRITE) != WRITE, |
| 1332 | &pages[cur_page]); |
Al Viro | e67dfe7 | 2017-09-23 15:51:23 -0400 | [diff] [blame] | 1333 | if (unlikely(ret < local_nr_pages)) { |
| 1334 | for (j = cur_page; j < page_limit; j++) { |
| 1335 | if (!pages[j]) |
| 1336 | break; |
| 1337 | put_page(pages[j]); |
| 1338 | } |
Jens Axboe | 9917215 | 2006-06-16 13:02:29 +0200 | [diff] [blame] | 1339 | ret = -EFAULT; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1340 | goto out_unmap; |
Jens Axboe | 9917215 | 2006-06-16 13:02:29 +0200 | [diff] [blame] | 1341 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1342 | |
Geliang Tang | bd5cece | 2015-11-21 17:27:31 +0800 | [diff] [blame] | 1343 | offset = offset_in_page(uaddr); |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1344 | for (j = cur_page; j < page_limit; j++) { |
| 1345 | unsigned int bytes = PAGE_SIZE - offset; |
Vitaly Mayatskikh | 5444d8a | 2017-09-22 01:18:39 -0400 | [diff] [blame] | 1346 | unsigned short prev_bi_vcnt = bio->bi_vcnt; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1347 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1348 | if (len <= 0) |
| 1349 | break; |
| 1350 | |
| 1351 | if (bytes > len) |
| 1352 | bytes = len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1353 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1354 | /* |
| 1355 | * sorry... |
| 1356 | */ |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 1357 | if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < |
| 1358 | bytes) |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1359 | break; |
| 1360 | |
Vitaly Mayatskikh | 5444d8a | 2017-09-22 01:18:39 -0400 | [diff] [blame] | 1361 | /* |
| 1362 | * check if vector was merged with previous |
| 1363 | * drop page reference if needed |
| 1364 | */ |
| 1365 | if (bio->bi_vcnt == prev_bi_vcnt) |
| 1366 | put_page(pages[j]); |
| 1367 | |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1368 | len -= bytes; |
| 1369 | offset = 0; |
| 1370 | } |
| 1371 | |
| 1372 | cur_page = j; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1373 | /* |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1374 | * release the pages we didn't map into the bio, if any |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1375 | */ |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1376 | while (j < page_limit) |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1377 | put_page(pages[j++]); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1378 | } |
| 1379 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1380 | kfree(pages); |
| 1381 | |
| 1382 | /* |
| 1383 | * set data direction, and check if mapped pages need bouncing |
| 1384 | */ |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1385 | if (iter->type & WRITE) |
Mike Christie | 95fe6c1 | 2016-06-05 14:31:48 -0500 | [diff] [blame] | 1386 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1387 | |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1388 | bio_set_flag(bio, BIO_USER_MAPPED); |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1389 | |
| 1390 | /* |
| 1391 | * subtle -- if __bio_map_user() ended up bouncing a bio, |
| 1392 | * it would normally disappear when its bi_end_io is run. |
| 1393 | * however, we need it for the unmap, so grab an extra |
| 1394 | * reference to it |
| 1395 | */ |
| 1396 | bio_get(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1397 | return bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1398 | |
| 1399 | out_unmap: |
Al Viro | e67dfe7 | 2017-09-23 15:51:23 -0400 | [diff] [blame] | 1400 | bio_for_each_segment_all(bvec, bio, j) { |
| 1401 | put_page(bvec->bv_page); |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1402 | } |
| 1403 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1404 | kfree(pages); |
| 1405 | bio_put(bio); |
| 1406 | return ERR_PTR(ret); |
| 1407 | } |
| 1408 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1409 | static void __bio_unmap_user(struct bio *bio) |
| 1410 | { |
| 1411 | struct bio_vec *bvec; |
| 1412 | int i; |
| 1413 | |
| 1414 | /* |
| 1415 | * make sure we dirty pages we wrote to |
| 1416 | */ |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1417 | bio_for_each_segment_all(bvec, bio, i) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1418 | if (bio_data_dir(bio) == READ) |
| 1419 | set_page_dirty_lock(bvec->bv_page); |
| 1420 | |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1421 | put_page(bvec->bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1422 | } |
| 1423 | |
| 1424 | bio_put(bio); |
| 1425 | } |
| 1426 | |
| 1427 | /** |
| 1428 | * bio_unmap_user - unmap a bio |
| 1429 | * @bio: the bio being unmapped |
| 1430 | * |
| 1431 | * Unmap a bio previously mapped by bio_map_user(). Must be called with |
| 1432 | * a process context. |
| 1433 | * |
| 1434 | * bio_unmap_user() may sleep. |
| 1435 | */ |
| 1436 | void bio_unmap_user(struct bio *bio) |
| 1437 | { |
| 1438 | __bio_unmap_user(bio); |
| 1439 | bio_put(bio); |
| 1440 | } |
| 1441 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1442 | static void bio_map_kern_endio(struct bio *bio) |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1443 | { |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1444 | bio_put(bio); |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1445 | } |
| 1446 | |
Christoph Hellwig | 75c72b8 | 2015-01-18 16:16:32 +0100 | [diff] [blame] | 1447 | /** |
| 1448 | * bio_map_kern - map kernel address into bio |
| 1449 | * @q: the struct request_queue for the bio |
| 1450 | * @data: pointer to buffer to map |
| 1451 | * @len: length in bytes |
| 1452 | * @gfp_mask: allocation flags for bio allocation |
| 1453 | * |
| 1454 | * Map the kernel address into a bio suitable for io to a block |
| 1455 | * device. Returns an error pointer in case of error. |
| 1456 | */ |
| 1457 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
| 1458 | gfp_t gfp_mask) |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1459 | { |
| 1460 | unsigned long kaddr = (unsigned long)data; |
| 1461 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1462 | unsigned long start = kaddr >> PAGE_SHIFT; |
| 1463 | const int nr_pages = end - start; |
| 1464 | int offset, i; |
| 1465 | struct bio *bio; |
| 1466 | |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1467 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1468 | if (!bio) |
| 1469 | return ERR_PTR(-ENOMEM); |
| 1470 | |
| 1471 | offset = offset_in_page(kaddr); |
| 1472 | for (i = 0; i < nr_pages; i++) { |
| 1473 | unsigned int bytes = PAGE_SIZE - offset; |
| 1474 | |
| 1475 | if (len <= 0) |
| 1476 | break; |
| 1477 | |
| 1478 | if (bytes > len) |
| 1479 | bytes = len; |
| 1480 | |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 1481 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
Christoph Hellwig | 75c72b8 | 2015-01-18 16:16:32 +0100 | [diff] [blame] | 1482 | offset) < bytes) { |
| 1483 | /* we don't support partial mappings */ |
| 1484 | bio_put(bio); |
| 1485 | return ERR_PTR(-EINVAL); |
| 1486 | } |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1487 | |
| 1488 | data += bytes; |
| 1489 | len -= bytes; |
| 1490 | offset = 0; |
| 1491 | } |
| 1492 | |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1493 | bio->bi_end_io = bio_map_kern_endio; |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1494 | return bio; |
| 1495 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1496 | EXPORT_SYMBOL(bio_map_kern); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1497 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1498 | static void bio_copy_kern_endio(struct bio *bio) |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1499 | { |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1500 | bio_free_pages(bio); |
| 1501 | bio_put(bio); |
| 1502 | } |
| 1503 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1504 | static void bio_copy_kern_endio_read(struct bio *bio) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1505 | { |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1506 | char *p = bio->bi_private; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1507 | struct bio_vec *bvec; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1508 | int i; |
| 1509 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1510 | bio_for_each_segment_all(bvec, bio, i) { |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1511 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1512 | p += bvec->bv_len; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1513 | } |
| 1514 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1515 | bio_copy_kern_endio(bio); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1516 | } |
| 1517 | |
| 1518 | /** |
| 1519 | * bio_copy_kern - copy kernel address into bio |
| 1520 | * @q: the struct request_queue for the bio |
| 1521 | * @data: pointer to buffer to copy |
| 1522 | * @len: length in bytes |
| 1523 | * @gfp_mask: allocation flags for bio and page allocation |
Randy Dunlap | ffee025 | 2008-04-30 09:08:54 +0200 | [diff] [blame] | 1524 | * @reading: data direction is READ |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1525 | * |
| 1526 | * copy the kernel address into a bio suitable for io to a block |
| 1527 | * device. Returns an error pointer in case of error. |
| 1528 | */ |
| 1529 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, |
| 1530 | gfp_t gfp_mask, int reading) |
| 1531 | { |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1532 | unsigned long kaddr = (unsigned long)data; |
| 1533 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1534 | unsigned long start = kaddr >> PAGE_SHIFT; |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1535 | struct bio *bio; |
| 1536 | void *p = data; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1537 | int nr_pages = 0; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1538 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1539 | /* |
| 1540 | * Overflow, abort |
| 1541 | */ |
| 1542 | if (end < start) |
| 1543 | return ERR_PTR(-EINVAL); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1544 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1545 | nr_pages = end - start; |
| 1546 | bio = bio_kmalloc(gfp_mask, nr_pages); |
| 1547 | if (!bio) |
| 1548 | return ERR_PTR(-ENOMEM); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1549 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1550 | while (len) { |
| 1551 | struct page *page; |
| 1552 | unsigned int bytes = PAGE_SIZE; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1553 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1554 | if (bytes > len) |
| 1555 | bytes = len; |
| 1556 | |
| 1557 | page = alloc_page(q->bounce_gfp | gfp_mask); |
| 1558 | if (!page) |
| 1559 | goto cleanup; |
| 1560 | |
| 1561 | if (!reading) |
| 1562 | memcpy(page_address(page), p, bytes); |
| 1563 | |
| 1564 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) |
| 1565 | break; |
| 1566 | |
| 1567 | len -= bytes; |
| 1568 | p += bytes; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1569 | } |
| 1570 | |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1571 | if (reading) { |
| 1572 | bio->bi_end_io = bio_copy_kern_endio_read; |
| 1573 | bio->bi_private = data; |
| 1574 | } else { |
| 1575 | bio->bi_end_io = bio_copy_kern_endio; |
Mike Christie | 95fe6c1 | 2016-06-05 14:31:48 -0500 | [diff] [blame] | 1576 | bio_set_op_attrs(bio, REQ_OP_WRITE, 0); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1577 | } |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1578 | |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1579 | return bio; |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1580 | |
| 1581 | cleanup: |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1582 | bio_free_pages(bio); |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1583 | bio_put(bio); |
| 1584 | return ERR_PTR(-ENOMEM); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1585 | } |
| 1586 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1587 | /* |
| 1588 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions |
| 1589 | * for performing direct-IO in BIOs. |
| 1590 | * |
| 1591 | * The problem is that we cannot run set_page_dirty() from interrupt context |
| 1592 | * because the required locks are not interrupt-safe. So what we can do is to |
| 1593 | * mark the pages dirty _before_ performing IO. And in interrupt context, |
| 1594 | * check that the pages are still dirty. If so, fine. If not, redirty them |
| 1595 | * in process context. |
| 1596 | * |
| 1597 | * We special-case compound pages here: normally this means reads into hugetlb |
| 1598 | * pages. The logic in here doesn't really work right for compound pages |
| 1599 | * because the VM does not uniformly chase down the head page in all cases. |
| 1600 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't |
| 1601 | * handle them at all. So we skip compound pages here at an early stage. |
| 1602 | * |
| 1603 | * Note that this code is very hard to test under normal circumstances because |
| 1604 | * direct-io pins the pages with get_user_pages(). This makes |
| 1605 | * 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] | 1606 | * 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] | 1607 | * pagecache. |
| 1608 | * |
| 1609 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the |
| 1610 | * deferred bio dirtying paths. |
| 1611 | */ |
| 1612 | |
| 1613 | /* |
| 1614 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. |
| 1615 | */ |
| 1616 | void bio_set_pages_dirty(struct bio *bio) |
| 1617 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1618 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1619 | int i; |
| 1620 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1621 | bio_for_each_segment_all(bvec, bio, i) { |
| 1622 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1623 | |
| 1624 | if (page && !PageCompound(page)) |
| 1625 | set_page_dirty_lock(page); |
| 1626 | } |
| 1627 | } |
| 1628 | |
Adrian Bunk | 86b6c7a | 2008-02-18 13:48:32 +0100 | [diff] [blame] | 1629 | static void bio_release_pages(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1630 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1631 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1632 | int i; |
| 1633 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1634 | bio_for_each_segment_all(bvec, bio, i) { |
| 1635 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1636 | |
| 1637 | if (page) |
| 1638 | put_page(page); |
| 1639 | } |
| 1640 | } |
| 1641 | |
| 1642 | /* |
| 1643 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. |
| 1644 | * If they are, then fine. If, however, some pages are clean then they must |
| 1645 | * have been written out during the direct-IO read. So we take another ref on |
| 1646 | * the BIO and the offending pages and re-dirty the pages in process context. |
| 1647 | * |
| 1648 | * 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] | 1649 | * here on. It will run one put_page() against each page and will run one |
| 1650 | * bio_put() against the BIO. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1651 | */ |
| 1652 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1653 | static void bio_dirty_fn(struct work_struct *work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1654 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1655 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1656 | static DEFINE_SPINLOCK(bio_dirty_lock); |
| 1657 | static struct bio *bio_dirty_list; |
| 1658 | |
| 1659 | /* |
| 1660 | * This runs in process context |
| 1661 | */ |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1662 | static void bio_dirty_fn(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1663 | { |
| 1664 | unsigned long flags; |
| 1665 | struct bio *bio; |
| 1666 | |
| 1667 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1668 | bio = bio_dirty_list; |
| 1669 | bio_dirty_list = NULL; |
| 1670 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1671 | |
| 1672 | while (bio) { |
| 1673 | struct bio *next = bio->bi_private; |
| 1674 | |
| 1675 | bio_set_pages_dirty(bio); |
| 1676 | bio_release_pages(bio); |
| 1677 | bio_put(bio); |
| 1678 | bio = next; |
| 1679 | } |
| 1680 | } |
| 1681 | |
| 1682 | void bio_check_pages_dirty(struct bio *bio) |
| 1683 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1684 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1685 | int nr_clean_pages = 0; |
| 1686 | int i; |
| 1687 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1688 | bio_for_each_segment_all(bvec, bio, i) { |
| 1689 | struct page *page = bvec->bv_page; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1690 | |
| 1691 | if (PageDirty(page) || PageCompound(page)) { |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1692 | put_page(page); |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1693 | bvec->bv_page = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1694 | } else { |
| 1695 | nr_clean_pages++; |
| 1696 | } |
| 1697 | } |
| 1698 | |
| 1699 | if (nr_clean_pages) { |
| 1700 | unsigned long flags; |
| 1701 | |
| 1702 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1703 | bio->bi_private = bio_dirty_list; |
| 1704 | bio_dirty_list = bio; |
| 1705 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1706 | schedule_work(&bio_dirty_work); |
| 1707 | } else { |
| 1708 | bio_put(bio); |
| 1709 | } |
| 1710 | } |
| 1711 | |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1712 | void generic_start_io_acct(int rw, unsigned long sectors, |
| 1713 | struct hd_struct *part) |
| 1714 | { |
| 1715 | int cpu = part_stat_lock(); |
| 1716 | |
| 1717 | part_round_stats(cpu, part); |
| 1718 | part_stat_inc(cpu, part, ios[rw]); |
| 1719 | part_stat_add(cpu, part, sectors[rw], sectors); |
| 1720 | part_inc_in_flight(part, rw); |
| 1721 | |
| 1722 | part_stat_unlock(); |
| 1723 | } |
| 1724 | EXPORT_SYMBOL(generic_start_io_acct); |
| 1725 | |
| 1726 | void generic_end_io_acct(int rw, struct hd_struct *part, |
| 1727 | unsigned long start_time) |
| 1728 | { |
| 1729 | unsigned long duration = jiffies - start_time; |
| 1730 | int cpu = part_stat_lock(); |
| 1731 | |
| 1732 | part_stat_add(cpu, part, ticks[rw], duration); |
| 1733 | part_round_stats(cpu, part); |
| 1734 | part_dec_in_flight(part, rw); |
| 1735 | |
| 1736 | part_stat_unlock(); |
| 1737 | } |
| 1738 | EXPORT_SYMBOL(generic_end_io_acct); |
| 1739 | |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1740 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
| 1741 | void bio_flush_dcache_pages(struct bio *bi) |
| 1742 | { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1743 | struct bio_vec bvec; |
| 1744 | struct bvec_iter iter; |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1745 | |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1746 | bio_for_each_segment(bvec, bi, iter) |
| 1747 | flush_dcache_page(bvec.bv_page); |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1748 | } |
| 1749 | EXPORT_SYMBOL(bio_flush_dcache_pages); |
| 1750 | #endif |
| 1751 | |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1752 | static inline bool bio_remaining_done(struct bio *bio) |
| 1753 | { |
| 1754 | /* |
| 1755 | * If we're not chaining, then ->__bi_remaining is always 1 and |
| 1756 | * we always end io on the first invocation. |
| 1757 | */ |
| 1758 | if (!bio_flagged(bio, BIO_CHAIN)) |
| 1759 | return true; |
| 1760 | |
| 1761 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); |
| 1762 | |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1763 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1764 | bio_clear_flag(bio, BIO_CHAIN); |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1765 | return true; |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1766 | } |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1767 | |
| 1768 | return false; |
| 1769 | } |
| 1770 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1771 | /** |
| 1772 | * bio_endio - end I/O on a bio |
| 1773 | * @bio: bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1774 | * |
| 1775 | * Description: |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1776 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
| 1777 | * way to end I/O on a bio. No one should call bi_end_io() directly on a |
| 1778 | * 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] | 1779 | **/ |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1780 | void bio_endio(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1781 | { |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1782 | again: |
Christoph Hellwig | 2b88551 | 2016-03-11 17:34:53 +0100 | [diff] [blame] | 1783 | if (!bio_remaining_done(bio)) |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1784 | return; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1785 | |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1786 | /* |
| 1787 | * Need to have a real endio function for chained bios, otherwise |
| 1788 | * various corner cases will break (like stacking block devices that |
| 1789 | * save/restore bi_end_io) - however, we want to avoid unbounded |
| 1790 | * recursion and blowing the stack. Tail call optimization would |
| 1791 | * handle this, but compiling with frame pointers also disables |
| 1792 | * gcc's sibling call optimization. |
| 1793 | */ |
| 1794 | if (bio->bi_end_io == bio_chain_endio) { |
| 1795 | bio = __bio_chain_endio(bio); |
| 1796 | goto again; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1797 | } |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1798 | |
| 1799 | if (bio->bi_end_io) |
| 1800 | bio->bi_end_io(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1801 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1802 | EXPORT_SYMBOL(bio_endio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1803 | |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1804 | /** |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1805 | * bio_split - split a bio |
| 1806 | * @bio: bio to split |
| 1807 | * @sectors: number of sectors to split from the front of @bio |
| 1808 | * @gfp: gfp mask |
| 1809 | * @bs: bio set to allocate from |
| 1810 | * |
| 1811 | * Allocates and returns a new bio which represents @sectors from the start of |
| 1812 | * @bio, and updates @bio to represent the remaining sectors. |
| 1813 | * |
Martin K. Petersen | f3f5da6 | 2015-07-22 07:57:12 -0400 | [diff] [blame] | 1814 | * Unless this is a discard request the newly allocated bio will point |
| 1815 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that |
| 1816 | * @bio is not freed before the split. |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1817 | */ |
| 1818 | struct bio *bio_split(struct bio *bio, int sectors, |
| 1819 | gfp_t gfp, struct bio_set *bs) |
| 1820 | { |
| 1821 | struct bio *split = NULL; |
| 1822 | |
| 1823 | BUG_ON(sectors <= 0); |
| 1824 | BUG_ON(sectors >= bio_sectors(bio)); |
| 1825 | |
Martin K. Petersen | f3f5da6 | 2015-07-22 07:57:12 -0400 | [diff] [blame] | 1826 | /* |
| 1827 | * Discards need a mutable bio_vec to accommodate the payload |
| 1828 | * required by the DSM TRIM and UNMAP commands. |
| 1829 | */ |
Adrian Hunter | 7afafc8 | 2016-08-16 10:59:35 +0300 | [diff] [blame] | 1830 | 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] | 1831 | split = bio_clone_bioset(bio, gfp, bs); |
| 1832 | else |
| 1833 | split = bio_clone_fast(bio, gfp, bs); |
| 1834 | |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1835 | if (!split) |
| 1836 | return NULL; |
| 1837 | |
| 1838 | split->bi_iter.bi_size = sectors << 9; |
| 1839 | |
| 1840 | if (bio_integrity(split)) |
| 1841 | bio_integrity_trim(split, 0, sectors); |
| 1842 | |
| 1843 | bio_advance(bio, split->bi_iter.bi_size); |
| 1844 | |
| 1845 | return split; |
| 1846 | } |
| 1847 | EXPORT_SYMBOL(bio_split); |
| 1848 | |
Martin K. Petersen | ad3316b | 2008-10-01 22:42:53 -0400 | [diff] [blame] | 1849 | /** |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1850 | * bio_trim - trim a bio |
| 1851 | * @bio: bio to trim |
| 1852 | * @offset: number of sectors to trim from the front of @bio |
| 1853 | * @size: size we want to trim @bio to, in sectors |
| 1854 | */ |
| 1855 | void bio_trim(struct bio *bio, int offset, int size) |
| 1856 | { |
| 1857 | /* 'bio' is a cloned bio which we need to trim to match |
| 1858 | * the given offset and size. |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1859 | */ |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1860 | |
| 1861 | size <<= 9; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1862 | if (offset == 0 && size == bio->bi_iter.bi_size) |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1863 | return; |
| 1864 | |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1865 | bio_clear_flag(bio, BIO_SEG_VALID); |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1866 | |
| 1867 | bio_advance(bio, offset << 9); |
| 1868 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1869 | bio->bi_iter.bi_size = size; |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1870 | } |
| 1871 | EXPORT_SYMBOL_GPL(bio_trim); |
| 1872 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1873 | /* |
| 1874 | * create memory pools for biovec's in a bio_set. |
| 1875 | * use the global biovec slabs created for general use. |
| 1876 | */ |
Fabian Frederick | a6c39cb4f | 2014-04-22 15:09:05 -0600 | [diff] [blame] | 1877 | mempool_t *biovec_create_pool(int pool_entries) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1878 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 1879 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1880 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1881 | return mempool_create_slab_pool(pool_entries, bp->slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1882 | } |
| 1883 | |
| 1884 | void bioset_free(struct bio_set *bs) |
| 1885 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1886 | if (bs->rescue_workqueue) |
| 1887 | destroy_workqueue(bs->rescue_workqueue); |
| 1888 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1889 | if (bs->bio_pool) |
| 1890 | mempool_destroy(bs->bio_pool); |
| 1891 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1892 | if (bs->bvec_pool) |
| 1893 | mempool_destroy(bs->bvec_pool); |
| 1894 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 1895 | bioset_integrity_free(bs); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1896 | bio_put_slab(bs); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1897 | |
| 1898 | kfree(bs); |
| 1899 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1900 | EXPORT_SYMBOL(bioset_free); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1901 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1902 | static struct bio_set *__bioset_create(unsigned int pool_size, |
| 1903 | unsigned int front_pad, |
| 1904 | bool create_bvec_pool) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1905 | { |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 1906 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1907 | struct bio_set *bs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1908 | |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1909 | bs = kzalloc(sizeof(*bs), GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1910 | if (!bs) |
| 1911 | return NULL; |
| 1912 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1913 | bs->front_pad = front_pad; |
Jens Axboe | 1b43449 | 2008-10-22 20:32:58 +0200 | [diff] [blame] | 1914 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1915 | spin_lock_init(&bs->rescue_lock); |
| 1916 | bio_list_init(&bs->rescue_list); |
| 1917 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); |
| 1918 | |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 1919 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 1920 | if (!bs->bio_slab) { |
| 1921 | kfree(bs); |
| 1922 | return NULL; |
| 1923 | } |
| 1924 | |
| 1925 | bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1926 | if (!bs->bio_pool) |
| 1927 | goto bad; |
| 1928 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1929 | if (create_bvec_pool) { |
| 1930 | bs->bvec_pool = biovec_create_pool(pool_size); |
| 1931 | if (!bs->bvec_pool) |
| 1932 | goto bad; |
| 1933 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1934 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1935 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); |
| 1936 | if (!bs->rescue_workqueue) |
| 1937 | goto bad; |
| 1938 | |
| 1939 | return bs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1940 | bad: |
| 1941 | bioset_free(bs); |
| 1942 | return NULL; |
| 1943 | } |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1944 | |
| 1945 | /** |
| 1946 | * bioset_create - Create a bio_set |
| 1947 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
| 1948 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1949 | * |
| 1950 | * Description: |
| 1951 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller |
| 1952 | * to ask for a number of bytes to be allocated in front of the bio. |
| 1953 | * Front pad allocation is useful for embedding the bio inside |
| 1954 | * another structure, to avoid allocating extra data to go with the bio. |
| 1955 | * Note that the bio must be embedded at the END of that structure always, |
| 1956 | * or things will break badly. |
| 1957 | */ |
| 1958 | struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) |
| 1959 | { |
| 1960 | return __bioset_create(pool_size, front_pad, true); |
| 1961 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1962 | EXPORT_SYMBOL(bioset_create); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1963 | |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 1964 | /** |
| 1965 | * bioset_create_nobvec - Create a bio_set without bio_vec mempool |
| 1966 | * @pool_size: Number of bio to cache in the mempool |
| 1967 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1968 | * |
| 1969 | * Description: |
| 1970 | * Same functionality as bioset_create() except that mempool is not |
| 1971 | * created for bio_vecs. Saving some memory for bio_clone_fast() users. |
| 1972 | */ |
| 1973 | struct bio_set *bioset_create_nobvec(unsigned int pool_size, unsigned int front_pad) |
| 1974 | { |
| 1975 | return __bioset_create(pool_size, front_pad, false); |
| 1976 | } |
| 1977 | EXPORT_SYMBOL(bioset_create_nobvec); |
| 1978 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1979 | #ifdef CONFIG_BLK_CGROUP |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 1980 | |
| 1981 | /** |
| 1982 | * bio_associate_blkcg - associate a bio with the specified blkcg |
| 1983 | * @bio: target bio |
| 1984 | * @blkcg_css: css of the blkcg to associate |
| 1985 | * |
| 1986 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will |
| 1987 | * treat @bio as if it were issued by a task which belongs to the blkcg. |
| 1988 | * |
| 1989 | * This function takes an extra reference of @blkcg_css which will be put |
| 1990 | * when @bio is released. The caller must own @bio and is responsible for |
| 1991 | * synchronizing calls to this function. |
| 1992 | */ |
| 1993 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) |
| 1994 | { |
| 1995 | if (unlikely(bio->bi_css)) |
| 1996 | return -EBUSY; |
| 1997 | css_get(blkcg_css); |
| 1998 | bio->bi_css = blkcg_css; |
| 1999 | return 0; |
| 2000 | } |
Tejun Heo | 5aa2a96 | 2015-07-23 14:27:09 -0400 | [diff] [blame] | 2001 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 2002 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2003 | /** |
| 2004 | * bio_associate_current - associate a bio with %current |
| 2005 | * @bio: target bio |
| 2006 | * |
| 2007 | * Associate @bio with %current if it hasn't been associated yet. Block |
| 2008 | * layer will treat @bio as if it were issued by %current no matter which |
| 2009 | * task actually issues it. |
| 2010 | * |
| 2011 | * This function takes an extra reference of @task's io_context and blkcg |
| 2012 | * which will be put when @bio is released. The caller must own @bio, |
| 2013 | * ensure %current->io_context exists, and is responsible for synchronizing |
| 2014 | * calls to this function. |
| 2015 | */ |
| 2016 | int bio_associate_current(struct bio *bio) |
| 2017 | { |
| 2018 | struct io_context *ioc; |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2019 | |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 2020 | if (bio->bi_css) |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2021 | return -EBUSY; |
| 2022 | |
| 2023 | ioc = current->io_context; |
| 2024 | if (!ioc) |
| 2025 | return -ENOENT; |
| 2026 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2027 | get_io_context_active(ioc); |
| 2028 | bio->bi_ioc = ioc; |
Tejun Heo | c165b3e | 2015-08-18 14:55:29 -0700 | [diff] [blame] | 2029 | bio->bi_css = task_get_css(current, io_cgrp_id); |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2030 | return 0; |
| 2031 | } |
Tejun Heo | 5aa2a96 | 2015-07-23 14:27:09 -0400 | [diff] [blame] | 2032 | EXPORT_SYMBOL_GPL(bio_associate_current); |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2033 | |
| 2034 | /** |
| 2035 | * bio_disassociate_task - undo bio_associate_current() |
| 2036 | * @bio: target bio |
| 2037 | */ |
| 2038 | void bio_disassociate_task(struct bio *bio) |
| 2039 | { |
| 2040 | if (bio->bi_ioc) { |
| 2041 | put_io_context(bio->bi_ioc); |
| 2042 | bio->bi_ioc = NULL; |
| 2043 | } |
| 2044 | if (bio->bi_css) { |
| 2045 | css_put(bio->bi_css); |
| 2046 | bio->bi_css = NULL; |
| 2047 | } |
| 2048 | } |
| 2049 | |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 2050 | /** |
| 2051 | * bio_clone_blkcg_association - clone blkcg association from src to dst bio |
| 2052 | * @dst: destination bio |
| 2053 | * @src: source bio |
| 2054 | */ |
| 2055 | void bio_clone_blkcg_association(struct bio *dst, struct bio *src) |
| 2056 | { |
| 2057 | if (src->bi_css) |
| 2058 | WARN_ON(bio_associate_blkcg(dst, src->bi_css)); |
| 2059 | } |
| 2060 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2061 | #endif /* CONFIG_BLK_CGROUP */ |
| 2062 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2063 | static void __init biovec_init_slabs(void) |
| 2064 | { |
| 2065 | int i; |
| 2066 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 2067 | for (i = 0; i < BVEC_POOL_NR; i++) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2068 | int size; |
| 2069 | struct biovec_slab *bvs = bvec_slabs + i; |
| 2070 | |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2071 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
| 2072 | bvs->slab = NULL; |
| 2073 | continue; |
| 2074 | } |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2075 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2076 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
| 2077 | bvs->slab = kmem_cache_create(bvs->name, size, 0, |
Paul Mundt | 20c2df8 | 2007-07-20 10:11:58 +0900 | [diff] [blame] | 2078 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2079 | } |
| 2080 | } |
| 2081 | |
| 2082 | static int __init init_bio(void) |
| 2083 | { |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2084 | bio_slab_max = 2; |
| 2085 | bio_slab_nr = 0; |
| 2086 | bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); |
| 2087 | if (!bio_slabs) |
| 2088 | panic("bio: can't allocate bios\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2089 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 2090 | bio_integrity_init(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2091 | biovec_init_slabs(); |
| 2092 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2093 | fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2094 | if (!fs_bio_set) |
| 2095 | panic("bio: can't allocate bios\n"); |
| 2096 | |
Martin K. Petersen | a91a278 | 2011-03-17 11:11:05 +0100 | [diff] [blame] | 2097 | if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE)) |
| 2098 | panic("bio: can't create integrity pool\n"); |
| 2099 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2100 | return 0; |
| 2101 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2102 | subsys_initcall(init_bio); |