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> |
Josef Bacik | 08e18ea | 2018-07-03 11:14:50 -0400 | [diff] [blame] | 31 | #include <linux/blk-cgroup.h> |
Blagovest Kolenichev | 9469959 | 2020-03-27 03:19:47 -0700 | [diff] [blame] | 32 | #include <linux/blk-crypto.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 33 | |
Li Zefan | 5578213 | 2009-06-09 13:43:05 +0800 | [diff] [blame] | 34 | #include <trace/events/block.h> |
Shaohua Li | 9e234ee | 2017-03-27 10:51:41 -0700 | [diff] [blame] | 35 | #include "blk.h" |
Josef Bacik | 67b42d0 | 2018-07-03 11:15:00 -0400 | [diff] [blame] | 36 | #include "blk-rq-qos.h" |
Ingo Molnar | 0bfc245 | 2008-11-26 11:59:56 +0100 | [diff] [blame] | 37 | |
Jens Axboe | 392ddc3 | 2008-12-23 12:42:54 +0100 | [diff] [blame] | 38 | /* |
| 39 | * Test patch to inline a certain number of bi_io_vec's inside the bio |
| 40 | * itself, to shrink a bio data allocation from two mempool calls to one |
| 41 | */ |
| 42 | #define BIO_INLINE_VECS 4 |
| 43 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 44 | /* |
| 45 | * if you change this list, also change bvec_alloc or things will |
| 46 | * break badly! cannot be bigger than what you can fit into an |
| 47 | * unsigned short |
| 48 | */ |
Mikulas Patocka | bd5c4fa | 2018-03-21 12:49:29 -0400 | [diff] [blame] | 49 | #define BV(x, n) { .nr_vecs = x, .name = "biovec-"#n } |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 50 | static struct biovec_slab bvec_slabs[BVEC_POOL_NR] __read_mostly = { |
Mikulas Patocka | bd5c4fa | 2018-03-21 12:49:29 -0400 | [diff] [blame] | 51 | 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] | 52 | }; |
| 53 | #undef BV |
| 54 | |
| 55 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 56 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by |
| 57 | * IO code that does not need private memory pools. |
| 58 | */ |
Kent Overstreet | f4f8154 | 2018-05-08 21:33:52 -0400 | [diff] [blame] | 59 | struct bio_set fs_bio_set; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 60 | EXPORT_SYMBOL(fs_bio_set); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 61 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 62 | /* |
| 63 | * Our slab pool management |
| 64 | */ |
| 65 | struct bio_slab { |
| 66 | struct kmem_cache *slab; |
| 67 | unsigned int slab_ref; |
| 68 | unsigned int slab_size; |
| 69 | char name[8]; |
| 70 | }; |
| 71 | static DEFINE_MUTEX(bio_slab_lock); |
| 72 | static struct bio_slab *bio_slabs; |
| 73 | static unsigned int bio_slab_nr, bio_slab_max; |
| 74 | |
| 75 | static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) |
| 76 | { |
| 77 | unsigned int sz = sizeof(struct bio) + extra_size; |
| 78 | struct kmem_cache *slab = NULL; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 79 | struct bio_slab *bslab, *new_bio_slabs; |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 80 | unsigned int new_bio_slab_max; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 81 | unsigned int i, entry = -1; |
| 82 | |
| 83 | mutex_lock(&bio_slab_lock); |
| 84 | |
| 85 | i = 0; |
| 86 | while (i < bio_slab_nr) { |
Thiago Farina | f06f135 | 2010-01-19 14:07:09 +0100 | [diff] [blame] | 87 | bslab = &bio_slabs[i]; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 88 | |
| 89 | if (!bslab->slab && entry == -1) |
| 90 | entry = i; |
| 91 | else if (bslab->slab_size == sz) { |
| 92 | slab = bslab->slab; |
| 93 | bslab->slab_ref++; |
| 94 | break; |
| 95 | } |
| 96 | i++; |
| 97 | } |
| 98 | |
| 99 | if (slab) |
| 100 | goto out_unlock; |
| 101 | |
| 102 | if (bio_slab_nr == bio_slab_max && entry == -1) { |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 103 | new_bio_slab_max = bio_slab_max << 1; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 104 | new_bio_slabs = krealloc(bio_slabs, |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 105 | new_bio_slab_max * sizeof(struct bio_slab), |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 106 | GFP_KERNEL); |
| 107 | if (!new_bio_slabs) |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 108 | goto out_unlock; |
Anna Leuschner | 386bc35 | 2012-10-22 21:53:36 +0200 | [diff] [blame] | 109 | bio_slab_max = new_bio_slab_max; |
Alexey Khoroshilov | 389d7b2 | 2012-08-09 15:19:25 +0200 | [diff] [blame] | 110 | bio_slabs = new_bio_slabs; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 111 | } |
| 112 | if (entry == -1) |
| 113 | entry = bio_slab_nr++; |
| 114 | |
| 115 | bslab = &bio_slabs[entry]; |
| 116 | |
| 117 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); |
Mikulas Patocka | 6a24148 | 2014-03-28 15:51:55 -0400 | [diff] [blame] | 118 | slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN, |
| 119 | SLAB_HWCACHE_ALIGN, NULL); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 120 | if (!slab) |
| 121 | goto out_unlock; |
| 122 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 123 | bslab->slab = slab; |
| 124 | bslab->slab_ref = 1; |
| 125 | bslab->slab_size = sz; |
| 126 | out_unlock: |
| 127 | mutex_unlock(&bio_slab_lock); |
| 128 | return slab; |
| 129 | } |
| 130 | |
| 131 | static void bio_put_slab(struct bio_set *bs) |
| 132 | { |
| 133 | struct bio_slab *bslab = NULL; |
| 134 | unsigned int i; |
| 135 | |
| 136 | mutex_lock(&bio_slab_lock); |
| 137 | |
| 138 | for (i = 0; i < bio_slab_nr; i++) { |
| 139 | if (bs->bio_slab == bio_slabs[i].slab) { |
| 140 | bslab = &bio_slabs[i]; |
| 141 | break; |
| 142 | } |
| 143 | } |
| 144 | |
| 145 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) |
| 146 | goto out; |
| 147 | |
| 148 | WARN_ON(!bslab->slab_ref); |
| 149 | |
| 150 | if (--bslab->slab_ref) |
| 151 | goto out; |
| 152 | |
| 153 | kmem_cache_destroy(bslab->slab); |
| 154 | bslab->slab = NULL; |
| 155 | |
| 156 | out: |
| 157 | mutex_unlock(&bio_slab_lock); |
| 158 | } |
| 159 | |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 160 | unsigned int bvec_nr_vecs(unsigned short idx) |
| 161 | { |
Greg Edwards | d6c02a9 | 2018-08-08 13:27:53 -0600 | [diff] [blame] | 162 | return bvec_slabs[--idx].nr_vecs; |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 163 | } |
| 164 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 165 | 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] | 166 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 167 | if (!idx) |
| 168 | return; |
| 169 | idx--; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 170 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 171 | BIO_BUG_ON(idx >= BVEC_POOL_NR); |
| 172 | |
| 173 | if (idx == BVEC_POOL_MAX) { |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 174 | mempool_free(bv, pool); |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 175 | } else { |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 176 | struct biovec_slab *bvs = bvec_slabs + idx; |
| 177 | |
| 178 | kmem_cache_free(bvs->slab, bv); |
| 179 | } |
| 180 | } |
| 181 | |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 182 | struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx, |
| 183 | mempool_t *pool) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 184 | { |
| 185 | struct bio_vec *bvl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 186 | |
| 187 | /* |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 188 | * see comment near bvec_array define! |
| 189 | */ |
| 190 | switch (nr) { |
| 191 | case 1: |
| 192 | *idx = 0; |
| 193 | break; |
| 194 | case 2 ... 4: |
| 195 | *idx = 1; |
| 196 | break; |
| 197 | case 5 ... 16: |
| 198 | *idx = 2; |
| 199 | break; |
| 200 | case 17 ... 64: |
| 201 | *idx = 3; |
| 202 | break; |
| 203 | case 65 ... 128: |
| 204 | *idx = 4; |
| 205 | break; |
| 206 | case 129 ... BIO_MAX_PAGES: |
| 207 | *idx = 5; |
| 208 | break; |
| 209 | default: |
| 210 | return NULL; |
| 211 | } |
| 212 | |
| 213 | /* |
| 214 | * idx now points to the pool we want to allocate from. only the |
| 215 | * 1-vec entry pool is mempool backed. |
| 216 | */ |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 217 | if (*idx == BVEC_POOL_MAX) { |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 218 | fallback: |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 219 | bvl = mempool_alloc(pool, gfp_mask); |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 220 | } else { |
| 221 | struct biovec_slab *bvs = bvec_slabs + *idx; |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 222 | gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 223 | |
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 | * Make this allocation restricted and don't dump info on |
| 226 | * allocation failures, since we'll fallback to the mempool |
| 227 | * in case of failure. |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 228 | */ |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 229 | __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
| 230 | |
| 231 | /* |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 232 | * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 233 | * is set, retry with the 1-entry mempool |
| 234 | */ |
| 235 | bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 236 | if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 237 | *idx = BVEC_POOL_MAX; |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 238 | goto fallback; |
| 239 | } |
| 240 | } |
| 241 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 242 | (*idx)++; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 243 | return bvl; |
| 244 | } |
| 245 | |
Jens Axboe | 9ae3b3f | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 246 | void bio_uninit(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 247 | { |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 248 | bio_disassociate_task(bio); |
Blagovest Kolenichev | 4987567 | 2020-04-20 07:29:59 -0700 | [diff] [blame] | 249 | |
| 250 | bio_crypt_free_ctx(bio); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 251 | } |
Jens Axboe | 9ae3b3f | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 252 | EXPORT_SYMBOL(bio_uninit); |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 253 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 254 | static void bio_free(struct bio *bio) |
| 255 | { |
| 256 | struct bio_set *bs = bio->bi_pool; |
| 257 | void *p; |
| 258 | |
Jens Axboe | 9ae3b3f | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 259 | bio_uninit(bio); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 260 | |
| 261 | if (bs) { |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 262 | 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] | 263 | |
| 264 | /* |
| 265 | * If we have front padding, adjust the bio pointer before freeing |
| 266 | */ |
| 267 | p = bio; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 268 | p -= bs->front_pad; |
| 269 | |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 270 | mempool_free(p, &bs->bio_pool); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 271 | } else { |
| 272 | /* Bio was allocated by bio_kmalloc() */ |
| 273 | kfree(bio); |
| 274 | } |
Peter Osterlund | 3676347 | 2005-09-06 15:16:42 -0700 | [diff] [blame] | 275 | } |
| 276 | |
Jens Axboe | 9ae3b3f | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 277 | /* |
| 278 | * Users of this function have their own bio allocation. Subsequently, |
| 279 | * they must remember to pair any call to bio_init() with bio_uninit() |
| 280 | * when IO has completed, or when the bio is released. |
| 281 | */ |
Ming Lei | 3a83f46 | 2016-11-22 08:57:21 -0700 | [diff] [blame] | 282 | void bio_init(struct bio *bio, struct bio_vec *table, |
| 283 | unsigned short max_vecs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 284 | { |
Jens Axboe | 2b94de5 | 2007-07-18 13:14:03 +0200 | [diff] [blame] | 285 | memset(bio, 0, sizeof(*bio)); |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 286 | atomic_set(&bio->__bi_remaining, 1); |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 287 | atomic_set(&bio->__bi_cnt, 1); |
Ming Lei | 3a83f46 | 2016-11-22 08:57:21 -0700 | [diff] [blame] | 288 | |
| 289 | bio->bi_io_vec = table; |
| 290 | bio->bi_max_vecs = max_vecs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 291 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 292 | EXPORT_SYMBOL(bio_init); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 293 | |
| 294 | /** |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 295 | * bio_reset - reinitialize a bio |
| 296 | * @bio: bio to reset |
| 297 | * |
| 298 | * Description: |
| 299 | * After calling bio_reset(), @bio will be in the same state as a freshly |
| 300 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are |
| 301 | * preserved are the ones that are initialized by bio_alloc_bioset(). See |
| 302 | * comment in struct bio. |
| 303 | */ |
| 304 | void bio_reset(struct bio *bio) |
| 305 | { |
| 306 | unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS); |
| 307 | |
Jens Axboe | 9ae3b3f | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 308 | bio_uninit(bio); |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 309 | |
| 310 | memset(bio, 0, BIO_RESET_BYTES); |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 311 | bio->bi_flags = flags; |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 312 | atomic_set(&bio->__bi_remaining, 1); |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 313 | } |
| 314 | EXPORT_SYMBOL(bio_reset); |
| 315 | |
Christoph Hellwig | 38f8baa | 2016-03-11 17:34:51 +0100 | [diff] [blame] | 316 | static struct bio *__bio_chain_endio(struct bio *bio) |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 317 | { |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 318 | struct bio *parent = bio->bi_private; |
| 319 | |
Christoph Hellwig | 4e4cbee | 2017-06-03 09:38:06 +0200 | [diff] [blame] | 320 | if (!parent->bi_status) |
| 321 | parent->bi_status = bio->bi_status; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 322 | bio_put(bio); |
Christoph Hellwig | 38f8baa | 2016-03-11 17:34:51 +0100 | [diff] [blame] | 323 | return parent; |
| 324 | } |
| 325 | |
| 326 | static void bio_chain_endio(struct bio *bio) |
| 327 | { |
| 328 | bio_endio(__bio_chain_endio(bio)); |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 329 | } |
| 330 | |
| 331 | /** |
| 332 | * bio_chain - chain bio completions |
Randy Dunlap | 1051a90 | 2014-04-20 16:03:31 -0700 | [diff] [blame] | 333 | * @bio: the target bio |
| 334 | * @parent: the @bio's parent bio |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 335 | * |
| 336 | * The caller won't have a bi_end_io called when @bio completes - instead, |
| 337 | * @parent's bi_end_io won't be called until both @parent and @bio have |
| 338 | * completed; the chained bio will also be freed when it completes. |
| 339 | * |
| 340 | * The caller must not set bi_private or bi_end_io in @bio. |
| 341 | */ |
| 342 | void bio_chain(struct bio *bio, struct bio *parent) |
| 343 | { |
| 344 | BUG_ON(bio->bi_private || bio->bi_end_io); |
| 345 | |
| 346 | bio->bi_private = parent; |
| 347 | bio->bi_end_io = bio_chain_endio; |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 348 | bio_inc_remaining(parent); |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 349 | } |
| 350 | EXPORT_SYMBOL(bio_chain); |
| 351 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 352 | static void bio_alloc_rescue(struct work_struct *work) |
| 353 | { |
| 354 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); |
| 355 | struct bio *bio; |
| 356 | |
| 357 | while (1) { |
| 358 | spin_lock(&bs->rescue_lock); |
| 359 | bio = bio_list_pop(&bs->rescue_list); |
| 360 | spin_unlock(&bs->rescue_lock); |
| 361 | |
| 362 | if (!bio) |
| 363 | break; |
| 364 | |
| 365 | generic_make_request(bio); |
| 366 | } |
| 367 | } |
| 368 | |
| 369 | static void punt_bios_to_rescuer(struct bio_set *bs) |
| 370 | { |
| 371 | struct bio_list punt, nopunt; |
| 372 | struct bio *bio; |
| 373 | |
NeilBrown | 47e0fb4 | 2017-06-18 14:38:57 +1000 | [diff] [blame] | 374 | if (WARN_ON_ONCE(!bs->rescue_workqueue)) |
| 375 | return; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 376 | /* |
| 377 | * In order to guarantee forward progress we must punt only bios that |
| 378 | * were allocated from this bio_set; otherwise, if there was a bio on |
| 379 | * there for a stacking driver higher up in the stack, processing it |
| 380 | * could require allocating bios from this bio_set, and doing that from |
| 381 | * our own rescuer would be bad. |
| 382 | * |
| 383 | * Since bio lists are singly linked, pop them all instead of trying to |
| 384 | * remove from the middle of the list: |
| 385 | */ |
| 386 | |
| 387 | bio_list_init(&punt); |
| 388 | bio_list_init(&nopunt); |
| 389 | |
NeilBrown | f5fe1b5 | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 390 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 391 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
NeilBrown | f5fe1b5 | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 392 | current->bio_list[0] = nopunt; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 393 | |
NeilBrown | f5fe1b5 | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 394 | bio_list_init(&nopunt); |
| 395 | while ((bio = bio_list_pop(¤t->bio_list[1]))) |
| 396 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
| 397 | current->bio_list[1] = nopunt; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 398 | |
| 399 | spin_lock(&bs->rescue_lock); |
| 400 | bio_list_merge(&bs->rescue_list, &punt); |
| 401 | spin_unlock(&bs->rescue_lock); |
| 402 | |
| 403 | queue_work(bs->rescue_workqueue, &bs->rescue_work); |
| 404 | } |
| 405 | |
Kent Overstreet | f44b48c | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 406 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 407 | * bio_alloc_bioset - allocate a bio for I/O |
Randy Dunlap | 519c8e9 | 2017-10-16 11:01:00 -0700 | [diff] [blame] | 408 | * @gfp_mask: the GFP_* mask given to the slab allocator |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 409 | * @nr_iovecs: number of iovecs to pre-allocate |
Jaak Ristioja | db18efa | 2010-01-15 12:05:07 +0200 | [diff] [blame] | 410 | * @bs: the bio_set to allocate from. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 411 | * |
| 412 | * Description: |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 413 | * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is |
| 414 | * backed by the @bs's mempool. |
| 415 | * |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 416 | * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will |
| 417 | * always be able to allocate a bio. This is due to the mempool guarantees. |
| 418 | * To make this work, callers must never allocate more than 1 bio at a time |
| 419 | * from this pool. Callers that need to allocate more than 1 bio must always |
| 420 | * submit the previously allocated bio for IO before attempting to allocate |
| 421 | * 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] | 422 | * |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 423 | * Note that when running under generic_make_request() (i.e. any block |
| 424 | * driver), bios are not submitted until after you return - see the code in |
| 425 | * generic_make_request() that converts recursion into iteration, to prevent |
| 426 | * stack overflows. |
| 427 | * |
| 428 | * This would normally mean allocating multiple bios under |
| 429 | * generic_make_request() would be susceptible to deadlocks, but we have |
| 430 | * deadlock avoidance code that resubmits any blocked bios from a rescuer |
| 431 | * thread. |
| 432 | * |
| 433 | * However, we do not guarantee forward progress for allocations from other |
| 434 | * mempools. Doing multiple allocations from the same mempool under |
| 435 | * generic_make_request() should be avoided - instead, use bio_set's front_pad |
| 436 | * for per bio allocations. |
| 437 | * |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 438 | * RETURNS: |
| 439 | * Pointer to new bio on success, NULL on failure. |
| 440 | */ |
Dan Carpenter | 7a88fa1 | 2017-03-23 13:24:55 +0300 | [diff] [blame] | 441 | struct bio *bio_alloc_bioset(gfp_t gfp_mask, unsigned int nr_iovecs, |
| 442 | struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 443 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 444 | gfp_t saved_gfp = gfp_mask; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 445 | unsigned front_pad; |
| 446 | unsigned inline_vecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 447 | struct bio_vec *bvl = NULL; |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 448 | struct bio *bio; |
| 449 | void *p; |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 450 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 451 | if (!bs) { |
| 452 | if (nr_iovecs > UIO_MAXIOV) |
| 453 | return NULL; |
| 454 | |
| 455 | p = kmalloc(sizeof(struct bio) + |
| 456 | nr_iovecs * sizeof(struct bio_vec), |
| 457 | gfp_mask); |
| 458 | front_pad = 0; |
| 459 | inline_vecs = nr_iovecs; |
| 460 | } else { |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 461 | /* should not use nobvec bioset for nr_iovecs > 0 */ |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 462 | if (WARN_ON_ONCE(!mempool_initialized(&bs->bvec_pool) && |
| 463 | nr_iovecs > 0)) |
Junichi Nomura | d8f429e | 2014-10-03 17:27:12 -0400 | [diff] [blame] | 464 | return NULL; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 465 | /* |
| 466 | * generic_make_request() converts recursion to iteration; this |
| 467 | * means if we're running beneath it, any bios we allocate and |
| 468 | * submit will not be submitted (and thus freed) until after we |
| 469 | * return. |
| 470 | * |
| 471 | * This exposes us to a potential deadlock if we allocate |
| 472 | * multiple bios from the same bio_set() while running |
| 473 | * underneath generic_make_request(). If we were to allocate |
| 474 | * multiple bios (say a stacking block driver that was splitting |
| 475 | * bios), we would deadlock if we exhausted the mempool's |
| 476 | * reserve. |
| 477 | * |
| 478 | * We solve this, and guarantee forward progress, with a rescuer |
| 479 | * workqueue per bio_set. If we go to allocate and there are |
| 480 | * bios on current->bio_list, we first try the allocation |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 481 | * without __GFP_DIRECT_RECLAIM; if that fails, we punt those |
| 482 | * bios we would be blocking to the rescuer workqueue before |
| 483 | * we retry with the original gfp_flags. |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 484 | */ |
| 485 | |
NeilBrown | f5fe1b5 | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 486 | if (current->bio_list && |
| 487 | (!bio_list_empty(¤t->bio_list[0]) || |
NeilBrown | 47e0fb4 | 2017-06-18 14:38:57 +1000 | [diff] [blame] | 488 | !bio_list_empty(¤t->bio_list[1])) && |
| 489 | bs->rescue_workqueue) |
Mel Gorman | d0164ad | 2015-11-06 16:28:21 -0800 | [diff] [blame] | 490 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 491 | |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 492 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 493 | if (!p && gfp_mask != saved_gfp) { |
| 494 | punt_bios_to_rescuer(bs); |
| 495 | gfp_mask = saved_gfp; |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 496 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 497 | } |
| 498 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 499 | front_pad = bs->front_pad; |
| 500 | inline_vecs = BIO_INLINE_VECS; |
| 501 | } |
| 502 | |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 503 | if (unlikely(!p)) |
| 504 | return NULL; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 505 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 506 | bio = p + front_pad; |
Ming Lei | 3a83f46 | 2016-11-22 08:57:21 -0700 | [diff] [blame] | 507 | bio_init(bio, NULL, 0); |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 508 | |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 509 | if (nr_iovecs > inline_vecs) { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 510 | unsigned long idx = 0; |
| 511 | |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 512 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, &bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 513 | if (!bvl && gfp_mask != saved_gfp) { |
| 514 | punt_bios_to_rescuer(bs); |
| 515 | gfp_mask = saved_gfp; |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 516 | bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, &bs->bvec_pool); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 517 | } |
| 518 | |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 519 | if (unlikely(!bvl)) |
| 520 | goto err_free; |
Kent Overstreet | a38352e | 2012-05-25 13:03:11 -0700 | [diff] [blame] | 521 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 522 | bio->bi_flags |= idx << BVEC_POOL_OFFSET; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 523 | } else if (nr_iovecs) { |
| 524 | bvl = bio->bi_inline_vecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 525 | } |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 526 | |
| 527 | bio->bi_pool = bs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 528 | bio->bi_max_vecs = nr_iovecs; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 529 | bio->bi_io_vec = bvl; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 530 | return bio; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 531 | |
| 532 | err_free: |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 533 | mempool_free(p, &bs->bio_pool); |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 534 | return NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 535 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 536 | EXPORT_SYMBOL(bio_alloc_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 537 | |
Kent Overstreet | 38a72da | 2018-05-08 21:33:53 -0400 | [diff] [blame] | 538 | void zero_fill_bio_iter(struct bio *bio, struct bvec_iter start) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 539 | { |
| 540 | unsigned long flags; |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 541 | struct bio_vec bv; |
| 542 | struct bvec_iter iter; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 543 | |
Kent Overstreet | 38a72da | 2018-05-08 21:33:53 -0400 | [diff] [blame] | 544 | __bio_for_each_segment(bv, bio, iter, start) { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 545 | char *data = bvec_kmap_irq(&bv, &flags); |
| 546 | memset(data, 0, bv.bv_len); |
| 547 | flush_dcache_page(bv.bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 548 | bvec_kunmap_irq(data, &flags); |
| 549 | } |
| 550 | } |
Kent Overstreet | 38a72da | 2018-05-08 21:33:53 -0400 | [diff] [blame] | 551 | EXPORT_SYMBOL(zero_fill_bio_iter); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 552 | |
| 553 | /** |
| 554 | * bio_put - release a reference to a bio |
| 555 | * @bio: bio to release reference to |
| 556 | * |
| 557 | * Description: |
| 558 | * Put a reference to a &struct bio, either one you have gotten with |
NeilBrown | 9b10f6a | 2017-06-18 14:38:59 +1000 | [diff] [blame] | 559 | * 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] | 560 | **/ |
| 561 | void bio_put(struct bio *bio) |
| 562 | { |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 563 | if (!bio_flagged(bio, BIO_REFFED)) |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 564 | bio_free(bio); |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 565 | else { |
| 566 | BIO_BUG_ON(!atomic_read(&bio->__bi_cnt)); |
| 567 | |
| 568 | /* |
| 569 | * last put frees it |
| 570 | */ |
| 571 | if (atomic_dec_and_test(&bio->__bi_cnt)) |
| 572 | bio_free(bio); |
| 573 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 574 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 575 | EXPORT_SYMBOL(bio_put); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 576 | |
Jens Axboe | 165125e | 2007-07-24 09:28:11 +0200 | [diff] [blame] | 577 | inline int bio_phys_segments(struct request_queue *q, struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 578 | { |
| 579 | if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) |
| 580 | blk_recount_segments(q, bio); |
| 581 | |
| 582 | return bio->bi_phys_segments; |
| 583 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 584 | EXPORT_SYMBOL(bio_phys_segments); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 585 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 586 | /** |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 587 | * __bio_clone_fast - clone a bio that shares the original bio's biovec |
| 588 | * @bio: destination bio |
| 589 | * @bio_src: bio to clone |
| 590 | * |
| 591 | * Clone a &bio. Caller will own the returned bio, but not |
| 592 | * the actual data it points to. Reference count of returned |
| 593 | * bio will be one. |
| 594 | * |
| 595 | * Caller must ensure that @bio_src is not freed before @bio. |
| 596 | */ |
| 597 | void __bio_clone_fast(struct bio *bio, struct bio *bio_src) |
| 598 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 599 | BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio)); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 600 | |
| 601 | /* |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 602 | * most users will be overriding ->bi_disk with a new target, |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 603 | * so we don't set nor calculate new physical/hw segment counts here |
| 604 | */ |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 605 | bio->bi_disk = bio_src->bi_disk; |
Michael Lyle | 62530ed | 2017-11-16 23:47:25 -0800 | [diff] [blame] | 606 | bio->bi_partno = bio_src->bi_partno; |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 607 | bio_set_flag(bio, BIO_CLONED); |
Shaohua Li | 111be88 | 2017-12-20 11:10:17 -0700 | [diff] [blame] | 608 | if (bio_flagged(bio_src, BIO_THROTTLED)) |
| 609 | bio_set_flag(bio, BIO_THROTTLED); |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 610 | bio->bi_opf = bio_src->bi_opf; |
Hannes Reinecke | 487d58a | 2018-11-12 10:35:25 -0700 | [diff] [blame] | 611 | bio->bi_ioprio = bio_src->bi_ioprio; |
Jens Axboe | cb6934f | 2017-06-27 09:22:02 -0600 | [diff] [blame] | 612 | bio->bi_write_hint = bio_src->bi_write_hint; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 613 | bio->bi_iter = bio_src->bi_iter; |
| 614 | bio->bi_io_vec = bio_src->bi_io_vec; |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 615 | |
| 616 | bio_clone_blkcg_association(bio, bio_src); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 617 | } |
| 618 | EXPORT_SYMBOL(__bio_clone_fast); |
| 619 | |
| 620 | /** |
| 621 | * bio_clone_fast - clone a bio that shares the original bio's biovec |
| 622 | * @bio: bio to clone |
| 623 | * @gfp_mask: allocation priority |
| 624 | * @bs: bio_set to allocate from |
| 625 | * |
| 626 | * Like __bio_clone_fast, only also allocates the returned bio |
| 627 | */ |
| 628 | struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs) |
| 629 | { |
| 630 | struct bio *b; |
| 631 | |
| 632 | b = bio_alloc_bioset(gfp_mask, 0, bs); |
| 633 | if (!b) |
| 634 | return NULL; |
| 635 | |
| 636 | __bio_clone_fast(b, bio); |
| 637 | |
Blagovest Kolenichev | 4987567 | 2020-04-20 07:29:59 -0700 | [diff] [blame] | 638 | bio_crypt_clone(b, bio, gfp_mask); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 639 | |
Blagovest Kolenichev | 9469959 | 2020-03-27 03:19:47 -0700 | [diff] [blame] | 640 | if (bio_integrity(bio) && |
| 641 | bio_integrity_clone(b, bio, gfp_mask) < 0) { |
| 642 | bio_put(b); |
| 643 | return NULL; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 644 | } |
| 645 | |
| 646 | return b; |
| 647 | } |
| 648 | EXPORT_SYMBOL(bio_clone_fast); |
| 649 | |
Shaohua Li | f459587 | 2017-03-24 10:34:43 -0700 | [diff] [blame] | 650 | /** |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 651 | * bio_add_pc_page - attempt to add page to bio |
| 652 | * @q: the target queue |
| 653 | * @bio: destination bio |
| 654 | * @page: page to add |
| 655 | * @len: vec entry length |
| 656 | * @offset: vec entry offset |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 657 | * |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 658 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
| 659 | * number of reasons, such as the bio being full or target block device |
| 660 | * limitations. The target block device must allow bio's up to PAGE_SIZE, |
| 661 | * so it is always possible to add a single page to an empty bio. |
| 662 | * |
| 663 | * This should only be used by REQ_PC bios. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 664 | */ |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 665 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page |
| 666 | *page, unsigned int len, unsigned int offset) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 667 | { |
| 668 | int retried_segments = 0; |
| 669 | struct bio_vec *bvec; |
| 670 | |
| 671 | /* |
| 672 | * cloned bio must not modify vec list |
| 673 | */ |
| 674 | if (unlikely(bio_flagged(bio, BIO_CLONED))) |
| 675 | return 0; |
| 676 | |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 677 | 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] | 678 | return 0; |
| 679 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 680 | /* |
| 681 | * For filesystems with a blocksize smaller than the pagesize |
| 682 | * we will often be called with the same page as last time and |
| 683 | * a consecutive offset. Optimize this special case. |
| 684 | */ |
| 685 | if (bio->bi_vcnt > 0) { |
| 686 | struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 687 | |
| 688 | if (page == prev->bv_page && |
| 689 | offset == prev->bv_offset + prev->bv_len) { |
| 690 | prev->bv_len += len; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 691 | bio->bi_iter.bi_size += len; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 692 | goto done; |
| 693 | } |
Jens Axboe | 66cb45a | 2014-06-24 16:22:24 -0600 | [diff] [blame] | 694 | |
| 695 | /* |
| 696 | * If the queue doesn't support SG gaps and adding this |
| 697 | * offset would create a gap, disallow it. |
| 698 | */ |
Keith Busch | 03100aa | 2015-08-19 14:24:05 -0700 | [diff] [blame] | 699 | if (bvec_gap_to_prev(q, prev, offset)) |
Jens Axboe | 66cb45a | 2014-06-24 16:22:24 -0600 | [diff] [blame] | 700 | return 0; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 701 | } |
| 702 | |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 703 | if (bio_full(bio)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 704 | return 0; |
| 705 | |
| 706 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 707 | * setup the new entry, we might clear it again later if we |
| 708 | * cannot add the page |
| 709 | */ |
| 710 | bvec = &bio->bi_io_vec[bio->bi_vcnt]; |
| 711 | bvec->bv_page = page; |
| 712 | bvec->bv_len = len; |
| 713 | bvec->bv_offset = offset; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 714 | bio->bi_vcnt++; |
| 715 | bio->bi_phys_segments++; |
| 716 | bio->bi_iter.bi_size += len; |
| 717 | |
| 718 | /* |
| 719 | * Perform a recount if the number of segments is greater |
| 720 | * than queue_max_segments(q). |
| 721 | */ |
| 722 | |
| 723 | while (bio->bi_phys_segments > queue_max_segments(q)) { |
| 724 | |
| 725 | if (retried_segments) |
| 726 | goto failed; |
| 727 | |
| 728 | retried_segments = 1; |
| 729 | blk_recount_segments(q, bio); |
| 730 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 731 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 732 | /* If we may be able to merge these biovecs, force a recount */ |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 733 | if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 734 | bio_clear_flag(bio, BIO_SEG_VALID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 735 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 736 | done: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 737 | return len; |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 738 | |
| 739 | failed: |
| 740 | bvec->bv_page = NULL; |
| 741 | bvec->bv_len = 0; |
| 742 | bvec->bv_offset = 0; |
| 743 | bio->bi_vcnt--; |
| 744 | bio->bi_iter.bi_size -= len; |
| 745 | blk_recount_segments(q, bio); |
| 746 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 747 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 748 | EXPORT_SYMBOL(bio_add_pc_page); |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 749 | |
| 750 | /** |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 751 | * __bio_try_merge_page - try appending data to an existing bvec. |
| 752 | * @bio: destination bio |
| 753 | * @page: page to add |
| 754 | * @len: length of the data to add |
| 755 | * @off: offset of the data in @page |
| 756 | * |
| 757 | * Try to add the data at @page + @off to the last bvec of @bio. This is a |
| 758 | * a useful optimisation for file systems with a block size smaller than the |
| 759 | * page size. |
| 760 | * |
| 761 | * Return %true on success or %false on failure. |
| 762 | */ |
| 763 | bool __bio_try_merge_page(struct bio *bio, struct page *page, |
| 764 | unsigned int len, unsigned int off) |
| 765 | { |
| 766 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
| 767 | return false; |
| 768 | |
| 769 | if (bio->bi_vcnt > 0) { |
| 770 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 771 | |
| 772 | if (page == bv->bv_page && off == bv->bv_offset + bv->bv_len) { |
| 773 | bv->bv_len += len; |
| 774 | bio->bi_iter.bi_size += len; |
| 775 | return true; |
| 776 | } |
| 777 | } |
| 778 | return false; |
| 779 | } |
| 780 | EXPORT_SYMBOL_GPL(__bio_try_merge_page); |
| 781 | |
| 782 | /** |
| 783 | * __bio_add_page - add page to a bio in a new segment |
| 784 | * @bio: destination bio |
| 785 | * @page: page to add |
| 786 | * @len: length of the data to add |
| 787 | * @off: offset of the data in @page |
| 788 | * |
| 789 | * Add the data at @page + @off to @bio as a new bvec. The caller must ensure |
| 790 | * that @bio has space for another bvec. |
| 791 | */ |
| 792 | void __bio_add_page(struct bio *bio, struct page *page, |
| 793 | unsigned int len, unsigned int off) |
| 794 | { |
| 795 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt]; |
| 796 | |
| 797 | WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)); |
| 798 | WARN_ON_ONCE(bio_full(bio)); |
| 799 | |
| 800 | bv->bv_page = page; |
| 801 | bv->bv_offset = off; |
| 802 | bv->bv_len = len; |
| 803 | |
| 804 | bio->bi_iter.bi_size += len; |
| 805 | bio->bi_vcnt++; |
Johannes Weiner | 1bae217 | 2019-08-08 15:03:00 -0400 | [diff] [blame] | 806 | |
| 807 | if (!bio_flagged(bio, BIO_WORKINGSET) && unlikely(PageWorkingset(page))) |
| 808 | bio_set_flag(bio, BIO_WORKINGSET); |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 809 | } |
| 810 | EXPORT_SYMBOL_GPL(__bio_add_page); |
| 811 | |
| 812 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 813 | * bio_add_page - attempt to add page to bio |
| 814 | * @bio: destination bio |
| 815 | * @page: page to add |
| 816 | * @len: vec entry length |
| 817 | * @offset: vec entry offset |
| 818 | * |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 819 | * Attempt to add a page to the bio_vec maplist. This will only fail |
| 820 | * 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] | 821 | */ |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 822 | int bio_add_page(struct bio *bio, struct page *page, |
| 823 | unsigned int len, unsigned int offset) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 824 | { |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 825 | if (!__bio_try_merge_page(bio, page, len, offset)) { |
| 826 | if (bio_full(bio)) |
| 827 | return 0; |
| 828 | __bio_add_page(bio, page, len, offset); |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 829 | } |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 830 | return len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 831 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 832 | EXPORT_SYMBOL(bio_add_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 833 | |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 834 | /** |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 835 | * __bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 836 | * @bio: bio to add pages to |
| 837 | * @iter: iov iterator describing the region to be mapped |
| 838 | * |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 839 | * Pins pages from *iter and appends them to @bio's bvec array. The |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 840 | * pages will have to be released using put_page() when done. |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 841 | * For multi-segment *iter, this function only adds pages from the |
| 842 | * the next non-empty segment of the iov iterator. |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 843 | */ |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 844 | static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 845 | { |
Martin Wilck | b403ea2 | 2018-07-25 23:15:07 +0200 | [diff] [blame] | 846 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt, idx; |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 847 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; |
| 848 | struct page **pages = (struct page **)bv; |
Martin Wilck | b403ea2 | 2018-07-25 23:15:07 +0200 | [diff] [blame] | 849 | size_t offset; |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 850 | ssize_t size; |
| 851 | |
| 852 | size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset); |
| 853 | if (unlikely(size <= 0)) |
| 854 | return size ? size : -EFAULT; |
Martin Wilck | b403ea2 | 2018-07-25 23:15:07 +0200 | [diff] [blame] | 855 | idx = nr_pages = (size + offset + PAGE_SIZE - 1) / PAGE_SIZE; |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 856 | |
| 857 | /* |
| 858 | * Deep magic below: We need to walk the pinned pages backwards |
| 859 | * because we are abusing the space allocated for the bio_vecs |
| 860 | * for the page array. Because the bio_vecs are larger than the |
| 861 | * page pointers by definition this will always work. But it also |
| 862 | * means we can't use bio_add_page, so any changes to it's semantics |
| 863 | * need to be reflected here as well. |
| 864 | */ |
| 865 | bio->bi_iter.bi_size += size; |
| 866 | bio->bi_vcnt += nr_pages; |
| 867 | |
Martin Wilck | b403ea2 | 2018-07-25 23:15:07 +0200 | [diff] [blame] | 868 | while (idx--) { |
| 869 | bv[idx].bv_page = pages[idx]; |
| 870 | bv[idx].bv_len = PAGE_SIZE; |
| 871 | bv[idx].bv_offset = 0; |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 872 | } |
| 873 | |
| 874 | bv[0].bv_offset += offset; |
| 875 | bv[0].bv_len -= offset; |
Martin Wilck | b403ea2 | 2018-07-25 23:15:07 +0200 | [diff] [blame] | 876 | bv[nr_pages - 1].bv_len -= nr_pages * PAGE_SIZE - offset - size; |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 877 | |
| 878 | iov_iter_advance(iter, size); |
| 879 | return 0; |
| 880 | } |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 881 | |
| 882 | /** |
| 883 | * bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio |
| 884 | * @bio: bio to add pages to |
| 885 | * @iter: iov iterator describing the region to be mapped |
| 886 | * |
| 887 | * Pins pages from *iter and appends them to @bio's bvec array. The |
| 888 | * pages will have to be released using put_page() when done. |
| 889 | * The function tries, but does not guarantee, to pin as many pages as |
| 890 | * fit into the bio, or are requested in *iter, whatever is smaller. |
| 891 | * If MM encounters an error pinning the requested pages, it stops. |
| 892 | * Error is returned only if 0 pages could be pinned. |
| 893 | */ |
| 894 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) |
| 895 | { |
| 896 | unsigned short orig_vcnt = bio->bi_vcnt; |
| 897 | |
| 898 | do { |
| 899 | int ret = __bio_iov_iter_get_pages(bio, iter); |
| 900 | |
| 901 | if (unlikely(ret)) |
| 902 | return bio->bi_vcnt > orig_vcnt ? 0 : ret; |
| 903 | |
| 904 | } while (iov_iter_count(iter) && !bio_full(bio)); |
| 905 | |
| 906 | return 0; |
| 907 | } |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 908 | EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages); |
| 909 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 910 | static void submit_bio_wait_endio(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 911 | { |
Christoph Hellwig | 65e53aa | 2017-10-25 17:55:57 +0900 | [diff] [blame] | 912 | complete(bio->bi_private); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 913 | } |
| 914 | |
| 915 | /** |
| 916 | * submit_bio_wait - submit a bio, and wait until it completes |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 917 | * @bio: The &struct bio which describes the I/O |
| 918 | * |
| 919 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from |
| 920 | * bio_endio() on failure. |
Jan Kara | 3d289d6 | 2017-08-02 10:25:21 +0200 | [diff] [blame] | 921 | * |
| 922 | * WARNING: Unlike to how submit_bio() is usually used, this function does not |
| 923 | * result in bio reference to be consumed. The caller must drop the reference |
| 924 | * on his own. |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 925 | */ |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 926 | int submit_bio_wait(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 927 | { |
Byungchul Park | e319e1f | 2017-10-25 17:56:05 +0900 | [diff] [blame] | 928 | DECLARE_COMPLETION_ONSTACK_MAP(done, bio->bi_disk->lockdep_map); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 929 | |
Christoph Hellwig | 65e53aa | 2017-10-25 17:55:57 +0900 | [diff] [blame] | 930 | bio->bi_private = &done; |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 931 | bio->bi_end_io = submit_bio_wait_endio; |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 932 | bio->bi_opf |= REQ_SYNC; |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 933 | submit_bio(bio); |
Christoph Hellwig | 65e53aa | 2017-10-25 17:55:57 +0900 | [diff] [blame] | 934 | wait_for_completion_io(&done); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 935 | |
Christoph Hellwig | 65e53aa | 2017-10-25 17:55:57 +0900 | [diff] [blame] | 936 | return blk_status_to_errno(bio->bi_status); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 937 | } |
| 938 | EXPORT_SYMBOL(submit_bio_wait); |
| 939 | |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 940 | /** |
| 941 | * bio_advance - increment/complete a bio by some number of bytes |
| 942 | * @bio: bio to advance |
| 943 | * @bytes: number of bytes to complete |
| 944 | * |
| 945 | * This updates bi_sector, bi_size and bi_idx; if the number of bytes to |
| 946 | * complete doesn't align with a bvec boundary, then bv_len and bv_offset will |
| 947 | * be updated on the last bvec as well. |
| 948 | * |
| 949 | * @bio will then represent the remaining, uncompleted portion of the io. |
| 950 | */ |
| 951 | void bio_advance(struct bio *bio, unsigned bytes) |
| 952 | { |
| 953 | if (bio_integrity(bio)) |
| 954 | bio_integrity_advance(bio, bytes); |
| 955 | |
Blagovest Kolenichev | 9469959 | 2020-03-27 03:19:47 -0700 | [diff] [blame] | 956 | bio_crypt_advance(bio, bytes); |
Kent Overstreet | 4550dd6 | 2013-08-07 14:26:21 -0700 | [diff] [blame] | 957 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 958 | } |
| 959 | EXPORT_SYMBOL(bio_advance); |
| 960 | |
Kent Overstreet | 45db54d | 2018-05-08 21:33:54 -0400 | [diff] [blame] | 961 | void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter, |
| 962 | struct bio *src, struct bvec_iter *src_iter) |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 963 | { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 964 | struct bio_vec src_bv, dst_bv; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 965 | void *src_p, *dst_p; |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 966 | unsigned bytes; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 967 | |
Kent Overstreet | 45db54d | 2018-05-08 21:33:54 -0400 | [diff] [blame] | 968 | while (src_iter->bi_size && dst_iter->bi_size) { |
| 969 | src_bv = bio_iter_iovec(src, *src_iter); |
| 970 | dst_bv = bio_iter_iovec(dst, *dst_iter); |
| 971 | |
| 972 | bytes = min(src_bv.bv_len, dst_bv.bv_len); |
| 973 | |
| 974 | src_p = kmap_atomic(src_bv.bv_page); |
| 975 | dst_p = kmap_atomic(dst_bv.bv_page); |
| 976 | |
| 977 | memcpy(dst_p + dst_bv.bv_offset, |
| 978 | src_p + src_bv.bv_offset, |
| 979 | bytes); |
| 980 | |
| 981 | kunmap_atomic(dst_p); |
| 982 | kunmap_atomic(src_p); |
| 983 | |
Kent Overstreet | 6e6e811 | 2018-05-08 21:33:55 -0400 | [diff] [blame] | 984 | flush_dcache_page(dst_bv.bv_page); |
| 985 | |
Kent Overstreet | 45db54d | 2018-05-08 21:33:54 -0400 | [diff] [blame] | 986 | bio_advance_iter(src, src_iter, bytes); |
| 987 | bio_advance_iter(dst, dst_iter, bytes); |
| 988 | } |
| 989 | } |
| 990 | EXPORT_SYMBOL(bio_copy_data_iter); |
| 991 | |
| 992 | /** |
| 993 | * bio_copy_data - copy contents of data buffers from one bio to another |
| 994 | * @src: source bio |
| 995 | * @dst: destination bio |
| 996 | * |
| 997 | * Stops when it reaches the end of either @src or @dst - that is, copies |
| 998 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). |
| 999 | */ |
| 1000 | void bio_copy_data(struct bio *dst, struct bio *src) |
| 1001 | { |
| 1002 | struct bvec_iter src_iter = src->bi_iter; |
| 1003 | struct bvec_iter dst_iter = dst->bi_iter; |
| 1004 | |
| 1005 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); |
| 1006 | } |
| 1007 | EXPORT_SYMBOL(bio_copy_data); |
| 1008 | |
| 1009 | /** |
| 1010 | * bio_list_copy_data - copy contents of data buffers from one chain of bios to |
| 1011 | * another |
| 1012 | * @src: source bio list |
| 1013 | * @dst: destination bio list |
| 1014 | * |
| 1015 | * Stops when it reaches the end of either the @src list or @dst list - that is, |
| 1016 | * copies min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of |
| 1017 | * bios). |
| 1018 | */ |
| 1019 | void bio_list_copy_data(struct bio *dst, struct bio *src) |
| 1020 | { |
| 1021 | struct bvec_iter src_iter = src->bi_iter; |
| 1022 | struct bvec_iter dst_iter = dst->bi_iter; |
| 1023 | |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1024 | while (1) { |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1025 | if (!src_iter.bi_size) { |
| 1026 | src = src->bi_next; |
| 1027 | if (!src) |
| 1028 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1029 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1030 | src_iter = src->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1031 | } |
| 1032 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1033 | if (!dst_iter.bi_size) { |
| 1034 | dst = dst->bi_next; |
| 1035 | if (!dst) |
| 1036 | break; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1037 | |
Kent Overstreet | 1cb9dda | 2013-08-07 14:26:39 -0700 | [diff] [blame] | 1038 | dst_iter = dst->bi_iter; |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1039 | } |
| 1040 | |
Kent Overstreet | 45db54d | 2018-05-08 21:33:54 -0400 | [diff] [blame] | 1041 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1042 | } |
| 1043 | } |
Kent Overstreet | 45db54d | 2018-05-08 21:33:54 -0400 | [diff] [blame] | 1044 | EXPORT_SYMBOL(bio_list_copy_data); |
Kent Overstreet | 16ac3d6 | 2012-09-10 13:57:51 -0700 | [diff] [blame] | 1045 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1046 | struct bio_map_data { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1047 | int is_our_pages; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1048 | struct iov_iter iter; |
| 1049 | struct iovec iov[]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1050 | }; |
| 1051 | |
Al Viro | 0e5b935 | 2017-09-24 13:14:35 -0400 | [diff] [blame] | 1052 | static struct bio_map_data *bio_alloc_map_data(struct iov_iter *data, |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1053 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1054 | { |
Al Viro | 0e5b935 | 2017-09-24 13:14:35 -0400 | [diff] [blame] | 1055 | struct bio_map_data *bmd; |
| 1056 | if (data->nr_segs > UIO_MAXIOV) |
Jens Axboe | f3f63c1 | 2010-10-29 11:46:56 -0600 | [diff] [blame] | 1057 | return NULL; |
| 1058 | |
Al Viro | 0e5b935 | 2017-09-24 13:14:35 -0400 | [diff] [blame] | 1059 | bmd = kmalloc(sizeof(struct bio_map_data) + |
| 1060 | sizeof(struct iovec) * data->nr_segs, gfp_mask); |
| 1061 | if (!bmd) |
| 1062 | return NULL; |
| 1063 | memcpy(bmd->iov, data->iov, sizeof(struct iovec) * data->nr_segs); |
| 1064 | bmd->iter = *data; |
| 1065 | bmd->iter.iov = bmd->iov; |
| 1066 | return bmd; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1067 | } |
| 1068 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1069 | /** |
| 1070 | * bio_copy_from_iter - copy all pages from iov_iter to bio |
| 1071 | * @bio: The &struct bio which describes the I/O as destination |
| 1072 | * @iter: iov_iter as source |
| 1073 | * |
| 1074 | * Copy all pages from iov_iter to bio. |
| 1075 | * Returns 0 on success, or error on failure. |
| 1076 | */ |
Al Viro | 98a09d6 | 2017-09-24 12:14:36 -0400 | [diff] [blame] | 1077 | 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] | 1078 | { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1079 | int i; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1080 | struct bio_vec *bvec; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1081 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1082 | bio_for_each_segment_all(bvec, bio, i) { |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1083 | ssize_t ret; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1084 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1085 | ret = copy_page_from_iter(bvec->bv_page, |
| 1086 | bvec->bv_offset, |
| 1087 | bvec->bv_len, |
Al Viro | 98a09d6 | 2017-09-24 12:14:36 -0400 | [diff] [blame] | 1088 | iter); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1089 | |
Al Viro | 98a09d6 | 2017-09-24 12:14:36 -0400 | [diff] [blame] | 1090 | if (!iov_iter_count(iter)) |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1091 | break; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1092 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1093 | if (ret < bvec->bv_len) |
| 1094 | return -EFAULT; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1095 | } |
| 1096 | |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1097 | return 0; |
| 1098 | } |
| 1099 | |
| 1100 | /** |
| 1101 | * bio_copy_to_iter - copy all pages from bio to iov_iter |
| 1102 | * @bio: The &struct bio which describes the I/O as source |
| 1103 | * @iter: iov_iter as destination |
| 1104 | * |
| 1105 | * Copy all pages from bio to iov_iter. |
| 1106 | * Returns 0 on success, or error on failure. |
| 1107 | */ |
| 1108 | static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter) |
| 1109 | { |
| 1110 | int i; |
| 1111 | struct bio_vec *bvec; |
| 1112 | |
| 1113 | bio_for_each_segment_all(bvec, bio, i) { |
| 1114 | ssize_t ret; |
| 1115 | |
| 1116 | ret = copy_page_to_iter(bvec->bv_page, |
| 1117 | bvec->bv_offset, |
| 1118 | bvec->bv_len, |
| 1119 | &iter); |
| 1120 | |
| 1121 | if (!iov_iter_count(&iter)) |
| 1122 | break; |
| 1123 | |
| 1124 | if (ret < bvec->bv_len) |
| 1125 | return -EFAULT; |
| 1126 | } |
| 1127 | |
| 1128 | return 0; |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1129 | } |
| 1130 | |
Guoqing Jiang | 491221f | 2016-09-22 03:10:01 -0400 | [diff] [blame] | 1131 | void bio_free_pages(struct bio *bio) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1132 | { |
| 1133 | struct bio_vec *bvec; |
| 1134 | int i; |
| 1135 | |
| 1136 | bio_for_each_segment_all(bvec, bio, i) |
| 1137 | __free_page(bvec->bv_page); |
| 1138 | } |
Guoqing Jiang | 491221f | 2016-09-22 03:10:01 -0400 | [diff] [blame] | 1139 | EXPORT_SYMBOL(bio_free_pages); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1140 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1141 | /** |
| 1142 | * bio_uncopy_user - finish previously mapped bio |
| 1143 | * @bio: bio being terminated |
| 1144 | * |
Christoph Hellwig | ddad8dd | 2015-01-18 16:16:29 +0100 | [diff] [blame] | 1145 | * Free pages allocated from bio_copy_user_iov() and write back data |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1146 | * to user space in case of a read. |
| 1147 | */ |
| 1148 | int bio_uncopy_user(struct bio *bio) |
| 1149 | { |
| 1150 | struct bio_map_data *bmd = bio->bi_private; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1151 | int ret = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1152 | |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1153 | if (!bio_flagged(bio, BIO_NULL_MAPPED)) { |
| 1154 | /* |
| 1155 | * if we're in a workqueue, the request is orphaned, so |
Hannes Reinecke | 2d99b55 | 2016-02-12 09:39:15 +0100 | [diff] [blame] | 1156 | * don't copy into a random user address space, just free |
| 1157 | * and return -EINTR so user space doesn't expect any data. |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1158 | */ |
Hannes Reinecke | 2d99b55 | 2016-02-12 09:39:15 +0100 | [diff] [blame] | 1159 | if (!current->mm) |
| 1160 | ret = -EINTR; |
| 1161 | else if (bio_data_dir(bio) == READ) |
Dongsu Park | 9124d3f | 2015-01-18 16:16:34 +0100 | [diff] [blame] | 1162 | ret = bio_copy_to_iter(bio, bmd->iter); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1163 | if (bmd->is_our_pages) |
| 1164 | bio_free_pages(bio); |
Roland Dreier | 35dc248 | 2013-08-05 17:55:01 -0700 | [diff] [blame] | 1165 | } |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1166 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1167 | bio_put(bio); |
| 1168 | return ret; |
| 1169 | } |
| 1170 | |
| 1171 | /** |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1172 | * bio_copy_user_iov - copy user data to bio |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1173 | * @q: destination block queue |
| 1174 | * @map_data: pointer to the rq_map_data holding pages (if necessary) |
| 1175 | * @iter: iovec iterator |
| 1176 | * @gfp_mask: memory allocation flags |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1177 | * |
| 1178 | * Prepares and returns a bio for indirect user io, bouncing data |
| 1179 | * to/from kernel pages as necessary. Must be paired with |
| 1180 | * call bio_uncopy_user() on io completion. |
| 1181 | */ |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1182 | struct bio *bio_copy_user_iov(struct request_queue *q, |
| 1183 | struct rq_map_data *map_data, |
Al Viro | e81cef5 | 2017-09-24 09:25:39 -0400 | [diff] [blame] | 1184 | struct iov_iter *iter, |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1185 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1186 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1187 | struct bio_map_data *bmd; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1188 | struct page *page; |
| 1189 | struct bio *bio; |
Al Viro | d16d44e | 2017-09-24 13:09:18 -0400 | [diff] [blame] | 1190 | int i = 0, ret; |
| 1191 | int nr_pages; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1192 | unsigned int len = iter->count; |
Geliang Tang | bd5cece | 2015-11-21 17:27:31 +0800 | [diff] [blame] | 1193 | unsigned int offset = map_data ? offset_in_page(map_data->offset) : 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1194 | |
Al Viro | 0e5b935 | 2017-09-24 13:14:35 -0400 | [diff] [blame] | 1195 | bmd = bio_alloc_map_data(iter, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1196 | if (!bmd) |
| 1197 | return ERR_PTR(-ENOMEM); |
| 1198 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1199 | /* |
| 1200 | * We need to do a deep copy of the iov_iter including the iovecs. |
| 1201 | * The caller provided iov might point to an on-stack or otherwise |
| 1202 | * shortlived one. |
| 1203 | */ |
| 1204 | bmd->is_our_pages = map_data ? 0 : 1; |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1205 | |
Al Viro | d16d44e | 2017-09-24 13:09:18 -0400 | [diff] [blame] | 1206 | nr_pages = DIV_ROUND_UP(offset + len, PAGE_SIZE); |
| 1207 | if (nr_pages > BIO_MAX_PAGES) |
| 1208 | nr_pages = BIO_MAX_PAGES; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1209 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1210 | ret = -ENOMEM; |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1211 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1212 | if (!bio) |
| 1213 | goto out_bmd; |
| 1214 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1215 | ret = 0; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1216 | |
| 1217 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1218 | nr_pages = 1 << map_data->page_order; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1219 | i = map_data->offset / PAGE_SIZE; |
| 1220 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1221 | while (len) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1222 | unsigned int bytes = PAGE_SIZE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1223 | |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1224 | bytes -= offset; |
| 1225 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1226 | if (bytes > len) |
| 1227 | bytes = len; |
| 1228 | |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1229 | if (map_data) { |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1230 | if (i == map_data->nr_entries * nr_pages) { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1231 | ret = -ENOMEM; |
| 1232 | break; |
| 1233 | } |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1234 | |
| 1235 | page = map_data->pages[i / nr_pages]; |
| 1236 | page += (i % nr_pages); |
| 1237 | |
| 1238 | i++; |
| 1239 | } else { |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1240 | page = alloc_page(q->bounce_gfp | gfp_mask); |
FUJITA Tomonori | e623ddb | 2008-12-18 14:49:36 +0900 | [diff] [blame] | 1241 | if (!page) { |
| 1242 | ret = -ENOMEM; |
| 1243 | break; |
| 1244 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1245 | } |
| 1246 | |
Jérôme Glisse | 2591bfc | 2019-04-10 16:27:51 -0400 | [diff] [blame] | 1247 | if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) { |
| 1248 | if (!map_data) |
| 1249 | __free_page(page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1250 | break; |
Jérôme Glisse | 2591bfc | 2019-04-10 16:27:51 -0400 | [diff] [blame] | 1251 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1252 | |
| 1253 | len -= bytes; |
FUJITA Tomonori | 56c451f | 2008-12-18 14:49:37 +0900 | [diff] [blame] | 1254 | offset = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1255 | } |
| 1256 | |
| 1257 | if (ret) |
| 1258 | goto cleanup; |
| 1259 | |
Al Viro | 2884d0b | 2017-09-24 12:09:21 -0400 | [diff] [blame] | 1260 | if (map_data) |
| 1261 | map_data->offset += bio->bi_iter.bi_size; |
| 1262 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1263 | /* |
| 1264 | * success |
| 1265 | */ |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1266 | if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) || |
FUJITA Tomonori | ecb554a | 2009-07-09 14:46:53 +0200 | [diff] [blame] | 1267 | (map_data && map_data->from_user)) { |
Al Viro | 98a09d6 | 2017-09-24 12:14:36 -0400 | [diff] [blame] | 1268 | ret = bio_copy_from_iter(bio, iter); |
FUJITA Tomonori | c5dec1c | 2008-04-11 12:56:49 +0200 | [diff] [blame] | 1269 | if (ret) |
| 1270 | goto cleanup; |
Al Viro | 98a09d6 | 2017-09-24 12:14:36 -0400 | [diff] [blame] | 1271 | } else { |
Keith Busch | 7290c71 | 2018-12-10 08:44:42 -0700 | [diff] [blame] | 1272 | if (bmd->is_our_pages) |
| 1273 | zero_fill_bio(bio); |
Al Viro | 98a09d6 | 2017-09-24 12:14:36 -0400 | [diff] [blame] | 1274 | iov_iter_advance(iter, bio->bi_iter.bi_size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1275 | } |
| 1276 | |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1277 | bio->bi_private = bmd; |
Al Viro | 2884d0b | 2017-09-24 12:09:21 -0400 | [diff] [blame] | 1278 | if (map_data && map_data->null_mapped) |
| 1279 | bio_set_flag(bio, BIO_NULL_MAPPED); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1280 | return bio; |
| 1281 | cleanup: |
FUJITA Tomonori | 152e283 | 2008-08-28 16:17:06 +0900 | [diff] [blame] | 1282 | if (!map_data) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1283 | bio_free_pages(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1284 | bio_put(bio); |
| 1285 | out_bmd: |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1286 | kfree(bmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1287 | return ERR_PTR(ret); |
| 1288 | } |
| 1289 | |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1290 | /** |
| 1291 | * bio_map_user_iov - map user iovec into bio |
| 1292 | * @q: the struct request_queue for the bio |
| 1293 | * @iter: iovec iterator |
| 1294 | * @gfp_mask: memory allocation flags |
| 1295 | * |
| 1296 | * Map the user space address into a bio suitable for io to a block |
| 1297 | * device. Returns an error pointer in case of error. |
| 1298 | */ |
| 1299 | struct bio *bio_map_user_iov(struct request_queue *q, |
Al Viro | e81cef5 | 2017-09-24 09:25:39 -0400 | [diff] [blame] | 1300 | struct iov_iter *iter, |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1301 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1302 | { |
Kent Overstreet | 26e49cf | 2015-01-18 16:16:31 +0100 | [diff] [blame] | 1303 | int j; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1304 | struct bio *bio; |
Al Viro | 076098e | 2017-09-23 16:08:57 -0400 | [diff] [blame] | 1305 | int ret; |
Al Viro | 2b04e8f | 2017-09-23 15:51:23 -0400 | [diff] [blame] | 1306 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1307 | |
Al Viro | b282cc7 | 2017-09-23 16:24:59 -0400 | [diff] [blame] | 1308 | if (!iov_iter_count(iter)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1309 | return ERR_PTR(-EINVAL); |
| 1310 | |
Al Viro | b282cc7 | 2017-09-23 16:24:59 -0400 | [diff] [blame] | 1311 | bio = bio_kmalloc(gfp_mask, iov_iter_npages(iter, BIO_MAX_PAGES)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1312 | if (!bio) |
| 1313 | return ERR_PTR(-ENOMEM); |
| 1314 | |
Al Viro | 0a0f151 | 2017-09-24 12:30:17 -0400 | [diff] [blame] | 1315 | while (iov_iter_count(iter)) { |
Al Viro | 629e42b | 2017-09-23 16:13:10 -0400 | [diff] [blame] | 1316 | struct page **pages; |
Al Viro | 076098e | 2017-09-23 16:08:57 -0400 | [diff] [blame] | 1317 | ssize_t bytes; |
| 1318 | size_t offs, added = 0; |
| 1319 | int npages; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1320 | |
Al Viro | 0a0f151 | 2017-09-24 12:30:17 -0400 | [diff] [blame] | 1321 | bytes = iov_iter_get_pages_alloc(iter, &pages, LONG_MAX, &offs); |
Al Viro | 076098e | 2017-09-23 16:08:57 -0400 | [diff] [blame] | 1322 | if (unlikely(bytes <= 0)) { |
| 1323 | ret = bytes ? bytes : -EFAULT; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1324 | goto out_unmap; |
Jens Axboe | 9917215 | 2006-06-16 13:02:29 +0200 | [diff] [blame] | 1325 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1326 | |
Al Viro | 076098e | 2017-09-23 16:08:57 -0400 | [diff] [blame] | 1327 | npages = DIV_ROUND_UP(offs + bytes, PAGE_SIZE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1328 | |
Al Viro | 98f0bc9 | 2017-09-23 16:23:18 -0400 | [diff] [blame] | 1329 | if (unlikely(offs & queue_dma_alignment(q))) { |
| 1330 | ret = -EINVAL; |
| 1331 | j = 0; |
| 1332 | } else { |
| 1333 | for (j = 0; j < npages; j++) { |
| 1334 | struct page *page = pages[j]; |
| 1335 | unsigned int n = PAGE_SIZE - offs; |
| 1336 | unsigned short prev_bi_vcnt = bio->bi_vcnt; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1337 | |
Al Viro | 98f0bc9 | 2017-09-23 16:23:18 -0400 | [diff] [blame] | 1338 | if (n > bytes) |
| 1339 | n = bytes; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1340 | |
Al Viro | 98f0bc9 | 2017-09-23 16:23:18 -0400 | [diff] [blame] | 1341 | if (!bio_add_pc_page(q, bio, page, n, offs)) |
| 1342 | break; |
Vitaly Mayatskikh | 95d78c2 | 2017-09-22 01:18:39 -0400 | [diff] [blame] | 1343 | |
Al Viro | 98f0bc9 | 2017-09-23 16:23:18 -0400 | [diff] [blame] | 1344 | /* |
| 1345 | * check if vector was merged with previous |
| 1346 | * drop page reference if needed |
| 1347 | */ |
| 1348 | if (bio->bi_vcnt == prev_bi_vcnt) |
| 1349 | put_page(page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1350 | |
Al Viro | 98f0bc9 | 2017-09-23 16:23:18 -0400 | [diff] [blame] | 1351 | added += n; |
| 1352 | bytes -= n; |
| 1353 | offs = 0; |
| 1354 | } |
Al Viro | 0a0f151 | 2017-09-24 12:30:17 -0400 | [diff] [blame] | 1355 | iov_iter_advance(iter, added); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1356 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1357 | /* |
| 1358 | * release the pages we didn't map into the bio, if any |
| 1359 | */ |
Al Viro | 629e42b | 2017-09-23 16:13:10 -0400 | [diff] [blame] | 1360 | while (j < npages) |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1361 | put_page(pages[j++]); |
Al Viro | 629e42b | 2017-09-23 16:13:10 -0400 | [diff] [blame] | 1362 | kvfree(pages); |
Al Viro | e2e115d | 2017-09-23 16:16:06 -0400 | [diff] [blame] | 1363 | /* couldn't stuff something into bio? */ |
| 1364 | if (bytes) |
| 1365 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1366 | } |
| 1367 | |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1368 | bio_set_flag(bio, BIO_USER_MAPPED); |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1369 | |
| 1370 | /* |
Bart Van Assche | 5fad1b6 | 2017-02-01 08:20:08 -0800 | [diff] [blame] | 1371 | * subtle -- if bio_map_user_iov() ended up bouncing a bio, |
Christoph Hellwig | 37f19e5 | 2015-01-18 16:16:33 +0100 | [diff] [blame] | 1372 | * it would normally disappear when its bi_end_io is run. |
| 1373 | * however, we need it for the unmap, so grab an extra |
| 1374 | * reference to it |
| 1375 | */ |
| 1376 | bio_get(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1377 | return bio; |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1378 | |
| 1379 | out_unmap: |
Al Viro | 2b04e8f | 2017-09-23 15:51:23 -0400 | [diff] [blame] | 1380 | bio_for_each_segment_all(bvec, bio, j) { |
| 1381 | put_page(bvec->bv_page); |
James Bottomley | f1970ba | 2005-06-20 14:06:52 +0200 | [diff] [blame] | 1382 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1383 | bio_put(bio); |
| 1384 | return ERR_PTR(ret); |
| 1385 | } |
| 1386 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1387 | static void __bio_unmap_user(struct bio *bio) |
| 1388 | { |
| 1389 | struct bio_vec *bvec; |
| 1390 | int i; |
| 1391 | |
| 1392 | /* |
| 1393 | * make sure we dirty pages we wrote to |
| 1394 | */ |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1395 | bio_for_each_segment_all(bvec, bio, i) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1396 | if (bio_data_dir(bio) == READ) |
| 1397 | set_page_dirty_lock(bvec->bv_page); |
| 1398 | |
Kirill A. Shutemov | 09cbfea | 2016-04-01 15:29:47 +0300 | [diff] [blame] | 1399 | put_page(bvec->bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1400 | } |
| 1401 | |
| 1402 | bio_put(bio); |
| 1403 | } |
| 1404 | |
| 1405 | /** |
| 1406 | * bio_unmap_user - unmap a bio |
| 1407 | * @bio: the bio being unmapped |
| 1408 | * |
Bart Van Assche | 5fad1b6 | 2017-02-01 08:20:08 -0800 | [diff] [blame] | 1409 | * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from |
| 1410 | * process context. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1411 | * |
| 1412 | * bio_unmap_user() may sleep. |
| 1413 | */ |
| 1414 | void bio_unmap_user(struct bio *bio) |
| 1415 | { |
| 1416 | __bio_unmap_user(bio); |
| 1417 | bio_put(bio); |
| 1418 | } |
| 1419 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1420 | static void bio_map_kern_endio(struct bio *bio) |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1421 | { |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1422 | bio_put(bio); |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1423 | } |
| 1424 | |
Christoph Hellwig | 75c72b8 | 2015-01-18 16:16:32 +0100 | [diff] [blame] | 1425 | /** |
| 1426 | * bio_map_kern - map kernel address into bio |
| 1427 | * @q: the struct request_queue for the bio |
| 1428 | * @data: pointer to buffer to map |
| 1429 | * @len: length in bytes |
| 1430 | * @gfp_mask: allocation flags for bio allocation |
| 1431 | * |
| 1432 | * Map the kernel address into a bio suitable for io to a block |
| 1433 | * device. Returns an error pointer in case of error. |
| 1434 | */ |
| 1435 | struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, |
| 1436 | gfp_t gfp_mask) |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1437 | { |
| 1438 | unsigned long kaddr = (unsigned long)data; |
| 1439 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1440 | unsigned long start = kaddr >> PAGE_SHIFT; |
| 1441 | const int nr_pages = end - start; |
| 1442 | int offset, i; |
| 1443 | struct bio *bio; |
| 1444 | |
Tejun Heo | a9e9dc2 | 2009-04-15 22:10:27 +0900 | [diff] [blame] | 1445 | bio = bio_kmalloc(gfp_mask, nr_pages); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1446 | if (!bio) |
| 1447 | return ERR_PTR(-ENOMEM); |
| 1448 | |
| 1449 | offset = offset_in_page(kaddr); |
| 1450 | for (i = 0; i < nr_pages; i++) { |
| 1451 | unsigned int bytes = PAGE_SIZE - offset; |
| 1452 | |
| 1453 | if (len <= 0) |
| 1454 | break; |
| 1455 | |
| 1456 | if (bytes > len) |
| 1457 | bytes = len; |
| 1458 | |
Mike Christie | defd94b | 2005-12-05 02:37:06 -0600 | [diff] [blame] | 1459 | if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, |
Christoph Hellwig | 75c72b8 | 2015-01-18 16:16:32 +0100 | [diff] [blame] | 1460 | offset) < bytes) { |
| 1461 | /* we don't support partial mappings */ |
| 1462 | bio_put(bio); |
| 1463 | return ERR_PTR(-EINVAL); |
| 1464 | } |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1465 | |
| 1466 | data += bytes; |
| 1467 | len -= bytes; |
| 1468 | offset = 0; |
| 1469 | } |
| 1470 | |
Jens Axboe | b823825 | 2005-06-20 14:05:27 +0200 | [diff] [blame] | 1471 | bio->bi_end_io = bio_map_kern_endio; |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1472 | return bio; |
| 1473 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1474 | EXPORT_SYMBOL(bio_map_kern); |
Mike Christie | df46b9a | 2005-06-20 14:04:44 +0200 | [diff] [blame] | 1475 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1476 | static void bio_copy_kern_endio(struct bio *bio) |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1477 | { |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1478 | bio_free_pages(bio); |
| 1479 | bio_put(bio); |
| 1480 | } |
| 1481 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1482 | static void bio_copy_kern_endio_read(struct bio *bio) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1483 | { |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1484 | char *p = bio->bi_private; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1485 | struct bio_vec *bvec; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1486 | int i; |
| 1487 | |
Kent Overstreet | d74c6d5 | 2013-02-06 12:23:11 -0800 | [diff] [blame] | 1488 | bio_for_each_segment_all(bvec, bio, i) { |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1489 | memcpy(p, page_address(bvec->bv_page), bvec->bv_len); |
Kent Overstreet | c8db444 | 2013-11-22 19:39:06 -0800 | [diff] [blame] | 1490 | p += bvec->bv_len; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1491 | } |
| 1492 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1493 | bio_copy_kern_endio(bio); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1494 | } |
| 1495 | |
| 1496 | /** |
| 1497 | * bio_copy_kern - copy kernel address into bio |
| 1498 | * @q: the struct request_queue for the bio |
| 1499 | * @data: pointer to buffer to copy |
| 1500 | * @len: length in bytes |
| 1501 | * @gfp_mask: allocation flags for bio and page allocation |
Randy Dunlap | ffee025 | 2008-04-30 09:08:54 +0200 | [diff] [blame] | 1502 | * @reading: data direction is READ |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1503 | * |
| 1504 | * copy the kernel address into a bio suitable for io to a block |
| 1505 | * device. Returns an error pointer in case of error. |
| 1506 | */ |
| 1507 | struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, |
| 1508 | gfp_t gfp_mask, int reading) |
| 1509 | { |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1510 | unsigned long kaddr = (unsigned long)data; |
| 1511 | unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; |
| 1512 | unsigned long start = kaddr >> PAGE_SHIFT; |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1513 | struct bio *bio; |
| 1514 | void *p = data; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1515 | int nr_pages = 0; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1516 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1517 | /* |
| 1518 | * Overflow, abort |
| 1519 | */ |
| 1520 | if (end < start) |
| 1521 | return ERR_PTR(-EINVAL); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1522 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1523 | nr_pages = end - start; |
| 1524 | bio = bio_kmalloc(gfp_mask, nr_pages); |
| 1525 | if (!bio) |
| 1526 | return ERR_PTR(-ENOMEM); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1527 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1528 | while (len) { |
| 1529 | struct page *page; |
| 1530 | unsigned int bytes = PAGE_SIZE; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1531 | |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1532 | if (bytes > len) |
| 1533 | bytes = len; |
| 1534 | |
| 1535 | page = alloc_page(q->bounce_gfp | gfp_mask); |
| 1536 | if (!page) |
| 1537 | goto cleanup; |
| 1538 | |
| 1539 | if (!reading) |
| 1540 | memcpy(page_address(page), p, bytes); |
| 1541 | |
| 1542 | if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) |
| 1543 | break; |
| 1544 | |
| 1545 | len -= bytes; |
| 1546 | p += bytes; |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1547 | } |
| 1548 | |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1549 | if (reading) { |
| 1550 | bio->bi_end_io = bio_copy_kern_endio_read; |
| 1551 | bio->bi_private = data; |
| 1552 | } else { |
| 1553 | bio->bi_end_io = bio_copy_kern_endio; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1554 | } |
FUJITA Tomonori | 76029ff | 2008-08-25 20:36:08 +0200 | [diff] [blame] | 1555 | |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1556 | return bio; |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1557 | |
| 1558 | cleanup: |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1559 | bio_free_pages(bio); |
Christoph Hellwig | 42d2683 | 2015-01-18 16:16:28 +0100 | [diff] [blame] | 1560 | bio_put(bio); |
| 1561 | return ERR_PTR(-ENOMEM); |
FUJITA Tomonori | 68154e9 | 2008-04-25 12:47:50 +0200 | [diff] [blame] | 1562 | } |
| 1563 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1564 | /* |
| 1565 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions |
| 1566 | * for performing direct-IO in BIOs. |
| 1567 | * |
| 1568 | * The problem is that we cannot run set_page_dirty() from interrupt context |
| 1569 | * because the required locks are not interrupt-safe. So what we can do is to |
| 1570 | * mark the pages dirty _before_ performing IO. And in interrupt context, |
| 1571 | * check that the pages are still dirty. If so, fine. If not, redirty them |
| 1572 | * in process context. |
| 1573 | * |
| 1574 | * We special-case compound pages here: normally this means reads into hugetlb |
| 1575 | * pages. The logic in here doesn't really work right for compound pages |
| 1576 | * because the VM does not uniformly chase down the head page in all cases. |
| 1577 | * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't |
| 1578 | * handle them at all. So we skip compound pages here at an early stage. |
| 1579 | * |
| 1580 | * Note that this code is very hard to test under normal circumstances because |
| 1581 | * direct-io pins the pages with get_user_pages(). This makes |
| 1582 | * 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] | 1583 | * 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] | 1584 | * pagecache. |
| 1585 | * |
| 1586 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the |
| 1587 | * deferred bio dirtying paths. |
| 1588 | */ |
| 1589 | |
| 1590 | /* |
| 1591 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. |
| 1592 | */ |
| 1593 | void bio_set_pages_dirty(struct bio *bio) |
| 1594 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1595 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1596 | int i; |
| 1597 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1598 | bio_for_each_segment_all(bvec, bio, i) { |
Christoph Hellwig | 3bb5098 | 2018-07-24 14:04:13 +0200 | [diff] [blame] | 1599 | if (!PageCompound(bvec->bv_page)) |
| 1600 | set_page_dirty_lock(bvec->bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1601 | } |
| 1602 | } |
Kent Overstreet | 1900fcc | 2018-05-08 21:33:57 -0400 | [diff] [blame] | 1603 | EXPORT_SYMBOL_GPL(bio_set_pages_dirty); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1604 | |
Adrian Bunk | 86b6c7a | 2008-02-18 13:48:32 +0100 | [diff] [blame] | 1605 | static void bio_release_pages(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1606 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1607 | struct bio_vec *bvec; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1608 | int i; |
| 1609 | |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1610 | bio_for_each_segment_all(bvec, bio, i) |
| 1611 | put_page(bvec->bv_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1612 | } |
| 1613 | |
| 1614 | /* |
| 1615 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. |
| 1616 | * If they are, then fine. If, however, some pages are clean then they must |
| 1617 | * have been written out during the direct-IO read. So we take another ref on |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1618 | * the BIO and re-dirty the pages in process context. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1619 | * |
| 1620 | * 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] | 1621 | * here on. It will run one put_page() against each page and will run one |
| 1622 | * bio_put() against the BIO. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1623 | */ |
| 1624 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1625 | static void bio_dirty_fn(struct work_struct *work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1626 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1627 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1628 | static DEFINE_SPINLOCK(bio_dirty_lock); |
| 1629 | static struct bio *bio_dirty_list; |
| 1630 | |
| 1631 | /* |
| 1632 | * This runs in process context |
| 1633 | */ |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1634 | static void bio_dirty_fn(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1635 | { |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1636 | struct bio *bio, *next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1637 | |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1638 | spin_lock_irq(&bio_dirty_lock); |
| 1639 | next = bio_dirty_list; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1640 | bio_dirty_list = NULL; |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1641 | spin_unlock_irq(&bio_dirty_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1642 | |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1643 | while ((bio = next) != NULL) { |
| 1644 | next = bio->bi_private; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1645 | |
| 1646 | bio_set_pages_dirty(bio); |
| 1647 | bio_release_pages(bio); |
| 1648 | bio_put(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1649 | } |
| 1650 | } |
| 1651 | |
| 1652 | void bio_check_pages_dirty(struct bio *bio) |
| 1653 | { |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1654 | struct bio_vec *bvec; |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1655 | unsigned long flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1656 | int i; |
| 1657 | |
Kent Overstreet | cb34e05 | 2012-09-05 15:22:02 -0700 | [diff] [blame] | 1658 | bio_for_each_segment_all(bvec, bio, i) { |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1659 | if (!PageDirty(bvec->bv_page) && !PageCompound(bvec->bv_page)) |
| 1660 | goto defer; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1661 | } |
| 1662 | |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1663 | bio_release_pages(bio); |
| 1664 | bio_put(bio); |
| 1665 | return; |
| 1666 | defer: |
| 1667 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1668 | bio->bi_private = bio_dirty_list; |
| 1669 | bio_dirty_list = bio; |
| 1670 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1671 | schedule_work(&bio_dirty_work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1672 | } |
Kent Overstreet | 1900fcc | 2018-05-08 21:33:57 -0400 | [diff] [blame] | 1673 | EXPORT_SYMBOL_GPL(bio_check_pages_dirty); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1674 | |
Michael Callahan | ddcf35d | 2018-07-18 04:47:39 -0700 | [diff] [blame] | 1675 | void generic_start_io_acct(struct request_queue *q, int op, |
Jens Axboe | d62e26b | 2017-06-30 21:55:08 -0600 | [diff] [blame] | 1676 | unsigned long sectors, struct hd_struct *part) |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1677 | { |
Michael Callahan | ddcf35d | 2018-07-18 04:47:39 -0700 | [diff] [blame] | 1678 | const int sgrp = op_stat_group(op); |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1679 | int cpu = part_stat_lock(); |
| 1680 | |
Jens Axboe | d62e26b | 2017-06-30 21:55:08 -0600 | [diff] [blame] | 1681 | part_round_stats(q, cpu, part); |
Michael Callahan | ddcf35d | 2018-07-18 04:47:39 -0700 | [diff] [blame] | 1682 | part_stat_inc(cpu, part, ios[sgrp]); |
| 1683 | part_stat_add(cpu, part, sectors[sgrp], sectors); |
| 1684 | part_inc_in_flight(q, part, op_is_write(op)); |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1685 | |
| 1686 | part_stat_unlock(); |
| 1687 | } |
| 1688 | EXPORT_SYMBOL(generic_start_io_acct); |
| 1689 | |
Michael Callahan | ddcf35d | 2018-07-18 04:47:39 -0700 | [diff] [blame] | 1690 | void generic_end_io_acct(struct request_queue *q, int req_op, |
Jens Axboe | d62e26b | 2017-06-30 21:55:08 -0600 | [diff] [blame] | 1691 | struct hd_struct *part, unsigned long start_time) |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1692 | { |
| 1693 | unsigned long duration = jiffies - start_time; |
Michael Callahan | ddcf35d | 2018-07-18 04:47:39 -0700 | [diff] [blame] | 1694 | const int sgrp = op_stat_group(req_op); |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1695 | int cpu = part_stat_lock(); |
| 1696 | |
Omar Sandoval | b57e99b | 2018-09-21 16:44:34 -0700 | [diff] [blame] | 1697 | part_stat_add(cpu, part, nsecs[sgrp], jiffies_to_nsecs(duration)); |
Jens Axboe | d62e26b | 2017-06-30 21:55:08 -0600 | [diff] [blame] | 1698 | part_round_stats(q, cpu, part); |
Michael Callahan | ddcf35d | 2018-07-18 04:47:39 -0700 | [diff] [blame] | 1699 | part_dec_in_flight(q, part, op_is_write(req_op)); |
Gu Zheng | 394ffa5 | 2014-11-24 11:05:22 +0800 | [diff] [blame] | 1700 | |
| 1701 | part_stat_unlock(); |
| 1702 | } |
| 1703 | EXPORT_SYMBOL(generic_end_io_acct); |
| 1704 | |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1705 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
| 1706 | void bio_flush_dcache_pages(struct bio *bi) |
| 1707 | { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1708 | struct bio_vec bvec; |
| 1709 | struct bvec_iter iter; |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1710 | |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 1711 | bio_for_each_segment(bvec, bi, iter) |
| 1712 | flush_dcache_page(bvec.bv_page); |
Ilya Loginov | 2d4dc89 | 2009-11-26 09:16:19 +0100 | [diff] [blame] | 1713 | } |
| 1714 | EXPORT_SYMBOL(bio_flush_dcache_pages); |
| 1715 | #endif |
| 1716 | |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1717 | static inline bool bio_remaining_done(struct bio *bio) |
| 1718 | { |
| 1719 | /* |
| 1720 | * If we're not chaining, then ->__bi_remaining is always 1 and |
| 1721 | * we always end io on the first invocation. |
| 1722 | */ |
| 1723 | if (!bio_flagged(bio, BIO_CHAIN)) |
| 1724 | return true; |
| 1725 | |
| 1726 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); |
| 1727 | |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1728 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1729 | bio_clear_flag(bio, BIO_CHAIN); |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1730 | return true; |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1731 | } |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1732 | |
| 1733 | return false; |
| 1734 | } |
| 1735 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1736 | /** |
| 1737 | * bio_endio - end I/O on a bio |
| 1738 | * @bio: bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1739 | * |
| 1740 | * Description: |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1741 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
| 1742 | * way to end I/O on a bio. No one should call bi_end_io() directly on a |
| 1743 | * bio unless they own it and thus know that it has an end_io function. |
NeilBrown | fbbaf70 | 2017-04-07 09:40:52 -0600 | [diff] [blame] | 1744 | * |
| 1745 | * bio_endio() can be called several times on a bio that has been chained |
| 1746 | * using bio_chain(). The ->bi_end_io() function will only be called the |
| 1747 | * last time. At this point the BLK_TA_COMPLETE tracing event will be |
| 1748 | * generated if BIO_TRACE_COMPLETION is set. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1749 | **/ |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1750 | void bio_endio(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1751 | { |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1752 | again: |
Christoph Hellwig | 2b88551 | 2016-03-11 17:34:53 +0100 | [diff] [blame] | 1753 | if (!bio_remaining_done(bio)) |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1754 | return; |
Blagovest Kolenichev | 9469959 | 2020-03-27 03:19:47 -0700 | [diff] [blame] | 1755 | |
| 1756 | if (!blk_crypto_endio(bio)) |
| 1757 | return; |
| 1758 | |
Christoph Hellwig | 7c20f11 | 2017-07-03 16:58:43 -0600 | [diff] [blame] | 1759 | if (!bio_integrity_endio(bio)) |
| 1760 | return; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1761 | |
Josef Bacik | 67b42d0 | 2018-07-03 11:15:00 -0400 | [diff] [blame] | 1762 | if (bio->bi_disk) |
| 1763 | rq_qos_done_bio(bio->bi_disk->queue, bio); |
| 1764 | |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1765 | /* |
| 1766 | * Need to have a real endio function for chained bios, otherwise |
| 1767 | * various corner cases will break (like stacking block devices that |
| 1768 | * save/restore bi_end_io) - however, we want to avoid unbounded |
| 1769 | * recursion and blowing the stack. Tail call optimization would |
| 1770 | * handle this, but compiling with frame pointers also disables |
| 1771 | * gcc's sibling call optimization. |
| 1772 | */ |
| 1773 | if (bio->bi_end_io == bio_chain_endio) { |
| 1774 | bio = __bio_chain_endio(bio); |
| 1775 | goto again; |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1776 | } |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1777 | |
Christoph Hellwig | 74d4699 | 2017-08-23 19:10:32 +0200 | [diff] [blame] | 1778 | if (bio->bi_disk && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
| 1779 | trace_block_bio_complete(bio->bi_disk->queue, bio, |
Bart Van Assche | a462b95 | 2017-06-13 08:07:33 -0700 | [diff] [blame] | 1780 | blk_status_to_errno(bio->bi_status)); |
NeilBrown | fbbaf70 | 2017-04-07 09:40:52 -0600 | [diff] [blame] | 1781 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
| 1782 | } |
| 1783 | |
Shaohua Li | 9e234ee | 2017-03-27 10:51:41 -0700 | [diff] [blame] | 1784 | blk_throtl_bio_endio(bio); |
Shaohua Li | b222dd2 | 2017-07-10 11:40:17 -0700 | [diff] [blame] | 1785 | /* release cgroup info */ |
| 1786 | bio_uninit(bio); |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1787 | if (bio->bi_end_io) |
| 1788 | bio->bi_end_io(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1789 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1790 | EXPORT_SYMBOL(bio_endio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1791 | |
Kent Overstreet | 196d38b | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1792 | /** |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1793 | * bio_split - split a bio |
| 1794 | * @bio: bio to split |
| 1795 | * @sectors: number of sectors to split from the front of @bio |
| 1796 | * @gfp: gfp mask |
| 1797 | * @bs: bio set to allocate from |
| 1798 | * |
| 1799 | * Allocates and returns a new bio which represents @sectors from the start of |
| 1800 | * @bio, and updates @bio to represent the remaining sectors. |
| 1801 | * |
Martin K. Petersen | f3f5da6 | 2015-07-22 07:57:12 -0400 | [diff] [blame] | 1802 | * Unless this is a discard request the newly allocated bio will point |
| 1803 | * to @bio's bi_io_vec; it is the caller's responsibility to ensure that |
| 1804 | * @bio is not freed before the split. |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1805 | */ |
| 1806 | struct bio *bio_split(struct bio *bio, int sectors, |
| 1807 | gfp_t gfp, struct bio_set *bs) |
| 1808 | { |
Mikulas Patocka | f341a4d | 2017-11-22 13:18:05 -0500 | [diff] [blame] | 1809 | struct bio *split; |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1810 | |
| 1811 | BUG_ON(sectors <= 0); |
| 1812 | BUG_ON(sectors >= bio_sectors(bio)); |
| 1813 | |
Christoph Hellwig | f9d03f9 | 2016-12-08 15:20:32 -0700 | [diff] [blame] | 1814 | split = bio_clone_fast(bio, gfp, bs); |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1815 | if (!split) |
| 1816 | return NULL; |
| 1817 | |
| 1818 | split->bi_iter.bi_size = sectors << 9; |
| 1819 | |
| 1820 | if (bio_integrity(split)) |
Dmitry Monakhov | fbd08e7 | 2017-06-29 11:31:10 -0700 | [diff] [blame] | 1821 | bio_integrity_trim(split); |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1822 | |
| 1823 | bio_advance(bio, split->bi_iter.bi_size); |
Greg Edwards | 5151842 | 2018-07-26 14:39:37 -0400 | [diff] [blame] | 1824 | bio->bi_iter.bi_done = 0; |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1825 | |
NeilBrown | fbbaf70 | 2017-04-07 09:40:52 -0600 | [diff] [blame] | 1826 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
Goldwyn Rodrigues | 20d5902 | 2018-01-23 09:10:19 -0700 | [diff] [blame] | 1827 | bio_set_flag(split, BIO_TRACE_COMPLETION); |
NeilBrown | fbbaf70 | 2017-04-07 09:40:52 -0600 | [diff] [blame] | 1828 | |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1829 | return split; |
| 1830 | } |
| 1831 | EXPORT_SYMBOL(bio_split); |
| 1832 | |
Martin K. Petersen | ad3316b | 2008-10-01 22:42:53 -0400 | [diff] [blame] | 1833 | /** |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1834 | * bio_trim - trim a bio |
| 1835 | * @bio: bio to trim |
| 1836 | * @offset: number of sectors to trim from the front of @bio |
| 1837 | * @size: size we want to trim @bio to, in sectors |
| 1838 | */ |
| 1839 | void bio_trim(struct bio *bio, int offset, int size) |
| 1840 | { |
| 1841 | /* 'bio' is a cloned bio which we need to trim to match |
| 1842 | * the given offset and size. |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1843 | */ |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1844 | |
| 1845 | size <<= 9; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1846 | if (offset == 0 && size == bio->bi_iter.bi_size) |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1847 | return; |
| 1848 | |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1849 | bio_clear_flag(bio, BIO_SEG_VALID); |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1850 | |
| 1851 | bio_advance(bio, offset << 9); |
| 1852 | |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1853 | bio->bi_iter.bi_size = size; |
Dmitry Monakhov | 376a78a | 2017-06-29 11:31:08 -0700 | [diff] [blame] | 1854 | |
| 1855 | if (bio_integrity(bio)) |
Dmitry Monakhov | fbd08e7 | 2017-06-29 11:31:10 -0700 | [diff] [blame] | 1856 | bio_integrity_trim(bio); |
Dmitry Monakhov | 376a78a | 2017-06-29 11:31:08 -0700 | [diff] [blame] | 1857 | |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1858 | } |
| 1859 | EXPORT_SYMBOL_GPL(bio_trim); |
| 1860 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1861 | /* |
| 1862 | * create memory pools for biovec's in a bio_set. |
| 1863 | * use the global biovec slabs created for general use. |
| 1864 | */ |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 1865 | int biovec_init_pool(mempool_t *pool, int pool_entries) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1866 | { |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 1867 | struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1868 | |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 1869 | return mempool_init_slab_pool(pool, pool_entries, bp->slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1870 | } |
| 1871 | |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1872 | /* |
| 1873 | * bioset_exit - exit a bioset initialized with bioset_init() |
| 1874 | * |
| 1875 | * May be called on a zeroed but uninitialized bioset (i.e. allocated with |
| 1876 | * kzalloc()). |
| 1877 | */ |
| 1878 | void bioset_exit(struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1879 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1880 | if (bs->rescue_workqueue) |
| 1881 | destroy_workqueue(bs->rescue_workqueue); |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1882 | bs->rescue_workqueue = NULL; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1883 | |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 1884 | mempool_exit(&bs->bio_pool); |
| 1885 | mempool_exit(&bs->bvec_pool); |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1886 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 1887 | bioset_integrity_free(bs); |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1888 | if (bs->bio_slab) |
| 1889 | bio_put_slab(bs); |
| 1890 | bs->bio_slab = NULL; |
| 1891 | } |
| 1892 | EXPORT_SYMBOL(bioset_exit); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1893 | |
NeilBrown | 011067b | 2017-06-18 14:38:57 +1000 | [diff] [blame] | 1894 | /** |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1895 | * bioset_init - Initialize a bio_set |
Kent Overstreet | dad0852 | 2018-05-20 18:25:58 -0400 | [diff] [blame] | 1896 | * @bs: pool to initialize |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1897 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
| 1898 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1899 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS |
| 1900 | * and %BIOSET_NEED_RESCUER |
| 1901 | * |
Kent Overstreet | dad0852 | 2018-05-20 18:25:58 -0400 | [diff] [blame] | 1902 | * Description: |
| 1903 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller |
| 1904 | * to ask for a number of bytes to be allocated in front of the bio. |
| 1905 | * Front pad allocation is useful for embedding the bio inside |
| 1906 | * another structure, to avoid allocating extra data to go with the bio. |
| 1907 | * Note that the bio must be embedded at the END of that structure always, |
| 1908 | * or things will break badly. |
| 1909 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated |
| 1910 | * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast(). |
| 1911 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to |
| 1912 | * dispatch queued requests when the mempool runs out of space. |
| 1913 | * |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1914 | */ |
| 1915 | int bioset_init(struct bio_set *bs, |
| 1916 | unsigned int pool_size, |
| 1917 | unsigned int front_pad, |
| 1918 | int flags) |
| 1919 | { |
| 1920 | unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
| 1921 | |
| 1922 | bs->front_pad = front_pad; |
| 1923 | |
| 1924 | spin_lock_init(&bs->rescue_lock); |
| 1925 | bio_list_init(&bs->rescue_list); |
| 1926 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); |
| 1927 | |
| 1928 | bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); |
| 1929 | if (!bs->bio_slab) |
| 1930 | return -ENOMEM; |
| 1931 | |
| 1932 | if (mempool_init_slab_pool(&bs->bio_pool, pool_size, bs->bio_slab)) |
| 1933 | goto bad; |
| 1934 | |
| 1935 | if ((flags & BIOSET_NEED_BVECS) && |
| 1936 | biovec_init_pool(&bs->bvec_pool, pool_size)) |
| 1937 | goto bad; |
| 1938 | |
| 1939 | if (!(flags & BIOSET_NEED_RESCUER)) |
| 1940 | return 0; |
| 1941 | |
| 1942 | bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0); |
| 1943 | if (!bs->rescue_workqueue) |
| 1944 | goto bad; |
| 1945 | |
| 1946 | return 0; |
| 1947 | bad: |
| 1948 | bioset_exit(bs); |
| 1949 | return -ENOMEM; |
| 1950 | } |
| 1951 | EXPORT_SYMBOL(bioset_init); |
| 1952 | |
Jens Axboe | 28e89fd9 | 2018-06-07 14:42:05 -0600 | [diff] [blame] | 1953 | /* |
| 1954 | * Initialize and setup a new bio_set, based on the settings from |
| 1955 | * another bio_set. |
| 1956 | */ |
| 1957 | int bioset_init_from_src(struct bio_set *bs, struct bio_set *src) |
| 1958 | { |
| 1959 | int flags; |
| 1960 | |
| 1961 | flags = 0; |
| 1962 | if (src->bvec_pool.min_nr) |
| 1963 | flags |= BIOSET_NEED_BVECS; |
| 1964 | if (src->rescue_workqueue) |
| 1965 | flags |= BIOSET_NEED_RESCUER; |
| 1966 | |
| 1967 | return bioset_init(bs, src->bio_pool.min_nr, src->front_pad, flags); |
| 1968 | } |
| 1969 | EXPORT_SYMBOL(bioset_init_from_src); |
| 1970 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 1971 | #ifdef CONFIG_BLK_CGROUP |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 1972 | |
Tejun Heo | 0d3bd88 | 2018-07-03 11:14:54 -0400 | [diff] [blame] | 1973 | #ifdef CONFIG_MEMCG |
| 1974 | /** |
| 1975 | * bio_associate_blkcg_from_page - associate a bio with the page's blkcg |
| 1976 | * @bio: target bio |
| 1977 | * @page: the page to lookup the blkcg from |
| 1978 | * |
| 1979 | * Associate @bio with the blkcg from @page's owning memcg. This works like |
| 1980 | * every other associate function wrt references. |
| 1981 | */ |
| 1982 | int bio_associate_blkcg_from_page(struct bio *bio, struct page *page) |
| 1983 | { |
| 1984 | struct cgroup_subsys_state *blkcg_css; |
| 1985 | |
| 1986 | if (unlikely(bio->bi_css)) |
| 1987 | return -EBUSY; |
| 1988 | if (!page->mem_cgroup) |
| 1989 | return 0; |
| 1990 | blkcg_css = cgroup_get_e_css(page->mem_cgroup->css.cgroup, |
| 1991 | &io_cgrp_subsys); |
| 1992 | bio->bi_css = blkcg_css; |
| 1993 | return 0; |
| 1994 | } |
| 1995 | #endif /* CONFIG_MEMCG */ |
| 1996 | |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 1997 | /** |
| 1998 | * bio_associate_blkcg - associate a bio with the specified blkcg |
| 1999 | * @bio: target bio |
| 2000 | * @blkcg_css: css of the blkcg to associate |
| 2001 | * |
| 2002 | * Associate @bio with the blkcg specified by @blkcg_css. Block layer will |
| 2003 | * treat @bio as if it were issued by a task which belongs to the blkcg. |
| 2004 | * |
| 2005 | * This function takes an extra reference of @blkcg_css which will be put |
| 2006 | * when @bio is released. The caller must own @bio and is responsible for |
| 2007 | * synchronizing calls to this function. |
| 2008 | */ |
| 2009 | int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css) |
| 2010 | { |
| 2011 | if (unlikely(bio->bi_css)) |
| 2012 | return -EBUSY; |
| 2013 | css_get(blkcg_css); |
| 2014 | bio->bi_css = blkcg_css; |
| 2015 | return 0; |
| 2016 | } |
Tejun Heo | 5aa2a96 | 2015-07-23 14:27:09 -0400 | [diff] [blame] | 2017 | EXPORT_SYMBOL_GPL(bio_associate_blkcg); |
Tejun Heo | 1d933cf | 2015-05-22 17:13:24 -0400 | [diff] [blame] | 2018 | |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2019 | /** |
Josef Bacik | 08e18ea | 2018-07-03 11:14:50 -0400 | [diff] [blame] | 2020 | * bio_associate_blkg - associate a bio with the specified blkg |
| 2021 | * @bio: target bio |
| 2022 | * @blkg: the blkg to associate |
| 2023 | * |
| 2024 | * Associate @bio with the blkg specified by @blkg. This is the queue specific |
| 2025 | * blkcg information associated with the @bio, a reference will be taken on the |
| 2026 | * @blkg and will be freed when the bio is freed. |
| 2027 | */ |
| 2028 | int bio_associate_blkg(struct bio *bio, struct blkcg_gq *blkg) |
| 2029 | { |
| 2030 | if (unlikely(bio->bi_blkg)) |
| 2031 | return -EBUSY; |
Dennis Zhou (Facebook) | 3111885 | 2018-08-31 16:22:44 -0400 | [diff] [blame] | 2032 | if (!blkg_try_get(blkg)) |
| 2033 | return -ENODEV; |
Josef Bacik | 08e18ea | 2018-07-03 11:14:50 -0400 | [diff] [blame] | 2034 | bio->bi_blkg = blkg; |
| 2035 | return 0; |
| 2036 | } |
| 2037 | |
| 2038 | /** |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2039 | * bio_disassociate_task - undo bio_associate_current() |
| 2040 | * @bio: target bio |
| 2041 | */ |
| 2042 | void bio_disassociate_task(struct bio *bio) |
| 2043 | { |
| 2044 | if (bio->bi_ioc) { |
| 2045 | put_io_context(bio->bi_ioc); |
| 2046 | bio->bi_ioc = NULL; |
| 2047 | } |
| 2048 | if (bio->bi_css) { |
| 2049 | css_put(bio->bi_css); |
| 2050 | bio->bi_css = NULL; |
| 2051 | } |
Josef Bacik | 08e18ea | 2018-07-03 11:14:50 -0400 | [diff] [blame] | 2052 | if (bio->bi_blkg) { |
| 2053 | blkg_put(bio->bi_blkg); |
| 2054 | bio->bi_blkg = NULL; |
| 2055 | } |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2056 | } |
| 2057 | |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 2058 | /** |
| 2059 | * bio_clone_blkcg_association - clone blkcg association from src to dst bio |
| 2060 | * @dst: destination bio |
| 2061 | * @src: source bio |
| 2062 | */ |
| 2063 | void bio_clone_blkcg_association(struct bio *dst, struct bio *src) |
| 2064 | { |
| 2065 | if (src->bi_css) |
| 2066 | WARN_ON(bio_associate_blkcg(dst, src->bi_css)); |
| 2067 | } |
Shaohua Li | 8a8e6f8 | 2017-08-18 10:27:59 -0700 | [diff] [blame] | 2068 | EXPORT_SYMBOL_GPL(bio_clone_blkcg_association); |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 2069 | #endif /* CONFIG_BLK_CGROUP */ |
| 2070 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2071 | static void __init biovec_init_slabs(void) |
| 2072 | { |
| 2073 | int i; |
| 2074 | |
Christoph Hellwig | ed996a5 | 2016-07-19 11:28:42 +0200 | [diff] [blame] | 2075 | for (i = 0; i < BVEC_POOL_NR; i++) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2076 | int size; |
| 2077 | struct biovec_slab *bvs = bvec_slabs + i; |
| 2078 | |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2079 | if (bvs->nr_vecs <= BIO_INLINE_VECS) { |
| 2080 | bvs->slab = NULL; |
| 2081 | continue; |
| 2082 | } |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 2083 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2084 | size = bvs->nr_vecs * sizeof(struct bio_vec); |
| 2085 | bvs->slab = kmem_cache_create(bvs->name, size, 0, |
Paul Mundt | 20c2df8 | 2007-07-20 10:11:58 +0900 | [diff] [blame] | 2086 | SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2087 | } |
| 2088 | } |
| 2089 | |
| 2090 | static int __init init_bio(void) |
| 2091 | { |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2092 | bio_slab_max = 2; |
| 2093 | bio_slab_nr = 0; |
Kees Cook | 6396bb2 | 2018-06-12 14:03:40 -0700 | [diff] [blame] | 2094 | bio_slabs = kcalloc(bio_slab_max, sizeof(struct bio_slab), |
| 2095 | GFP_KERNEL); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 2096 | if (!bio_slabs) |
| 2097 | panic("bio: can't allocate bios\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2098 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 2099 | bio_integrity_init(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2100 | biovec_init_slabs(); |
| 2101 | |
Kent Overstreet | f4f8154 | 2018-05-08 21:33:52 -0400 | [diff] [blame] | 2102 | if (bioset_init(&fs_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2103 | panic("bio: can't allocate bios\n"); |
| 2104 | |
Kent Overstreet | f4f8154 | 2018-05-08 21:33:52 -0400 | [diff] [blame] | 2105 | if (bioset_integrity_create(&fs_bio_set, BIO_POOL_SIZE)) |
Martin K. Petersen | a91a278 | 2011-03-17 11:11:05 +0100 | [diff] [blame] | 2106 | panic("bio: can't create integrity pool\n"); |
| 2107 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2108 | return 0; |
| 2109 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2110 | subsys_initcall(init_bio); |