Christoph Hellwig | 8c16567 | 2019-04-30 14:42:39 -0400 | [diff] [blame] | 1 | // SPDX-License-Identifier: GPL-2.0 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2 | /* |
Jens Axboe | 0fe2347 | 2006-09-04 15:41:16 +0200 | [diff] [blame] | 3 | * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | */ |
| 5 | #include <linux/mm.h> |
| 6 | #include <linux/swap.h> |
| 7 | #include <linux/bio.h> |
| 8 | #include <linux/blkdev.h> |
Kent Overstreet | a27bb33 | 2013-05-07 16:19:08 -0700 | [diff] [blame] | 9 | #include <linux/uio.h> |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 10 | #include <linux/iocontext.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 11 | #include <linux/slab.h> |
| 12 | #include <linux/init.h> |
| 13 | #include <linux/kernel.h> |
Paul Gortmaker | 630d9c4 | 2011-11-16 23:57:37 -0500 | [diff] [blame] | 14 | #include <linux/export.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 15 | #include <linux/mempool.h> |
| 16 | #include <linux/workqueue.h> |
Tejun Heo | 852c788 | 2012-03-05 13:15:27 -0800 | [diff] [blame] | 17 | #include <linux/cgroup.h> |
Damien Le Moal | b4c5875 | 2019-07-01 14:09:15 +0900 | [diff] [blame] | 18 | #include <linux/highmem.h> |
Satya Tangirala | a892c8d | 2020-05-14 00:37:18 +0000 | [diff] [blame] | 19 | #include <linux/blk-crypto.h> |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 20 | #include <linux/xarray.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 21 | |
Li Zefan | 5578213 | 2009-06-09 13:43:05 +0800 | [diff] [blame] | 22 | #include <trace/events/block.h> |
Shaohua Li | 9e234ee | 2017-03-27 10:51:41 -0700 | [diff] [blame] | 23 | #include "blk.h" |
Josef Bacik | 67b42d0 | 2018-07-03 11:15:00 -0400 | [diff] [blame] | 24 | #include "blk-rq-qos.h" |
Ming Lei | 672fdcf | 2022-02-11 18:11:49 +0800 | [diff] [blame] | 25 | #include "blk-cgroup.h" |
Ingo Molnar | 0bfc245 | 2008-11-26 11:59:56 +0100 | [diff] [blame] | 26 | |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 27 | #define ALLOC_CACHE_THRESHOLD 16 |
Pavel Begunkov | 42b2b2f | 2022-11-02 15:18:23 +0000 | [diff] [blame] | 28 | #define ALLOC_CACHE_MAX 256 |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 29 | |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 30 | struct bio_alloc_cache { |
Jens Axboe | fcade2c | 2021-12-01 16:19:18 -0700 | [diff] [blame] | 31 | struct bio *free_list; |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 32 | struct bio *free_list_irq; |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 33 | unsigned int nr; |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 34 | unsigned int nr_irq; |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 35 | }; |
| 36 | |
Christoph Hellwig | de76fd8 | 2021-02-02 18:19:23 +0100 | [diff] [blame] | 37 | static struct biovec_slab { |
Christoph Hellwig | 6ac0b71 | 2021-02-02 18:19:20 +0100 | [diff] [blame] | 38 | int nr_vecs; |
| 39 | char *name; |
| 40 | struct kmem_cache *slab; |
Christoph Hellwig | de76fd8 | 2021-02-02 18:19:23 +0100 | [diff] [blame] | 41 | } bvec_slabs[] __read_mostly = { |
| 42 | { .nr_vecs = 16, .name = "biovec-16" }, |
| 43 | { .nr_vecs = 64, .name = "biovec-64" }, |
| 44 | { .nr_vecs = 128, .name = "biovec-128" }, |
Christoph Hellwig | a8affc0 | 2021-03-11 12:01:37 +0100 | [diff] [blame] | 45 | { .nr_vecs = BIO_MAX_VECS, .name = "biovec-max" }, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 46 | }; |
Christoph Hellwig | 6ac0b71 | 2021-02-02 18:19:20 +0100 | [diff] [blame] | 47 | |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 48 | static struct biovec_slab *biovec_slab(unsigned short nr_vecs) |
| 49 | { |
| 50 | switch (nr_vecs) { |
| 51 | /* smaller bios use inline vecs */ |
| 52 | case 5 ... 16: |
| 53 | return &bvec_slabs[0]; |
| 54 | case 17 ... 64: |
| 55 | return &bvec_slabs[1]; |
| 56 | case 65 ... 128: |
| 57 | return &bvec_slabs[2]; |
Christoph Hellwig | a8affc0 | 2021-03-11 12:01:37 +0100 | [diff] [blame] | 58 | case 129 ... BIO_MAX_VECS: |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 59 | return &bvec_slabs[3]; |
| 60 | default: |
| 61 | BUG(); |
| 62 | return NULL; |
| 63 | } |
| 64 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 65 | |
| 66 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 67 | * fs_bio_set is the bio_set containing bio and iovec memory pools used by |
| 68 | * IO code that does not need private memory pools. |
| 69 | */ |
Kent Overstreet | f4f8154 | 2018-05-08 21:33:52 -0400 | [diff] [blame] | 70 | struct bio_set fs_bio_set; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 71 | EXPORT_SYMBOL(fs_bio_set); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 73 | /* |
| 74 | * Our slab pool management |
| 75 | */ |
| 76 | struct bio_slab { |
| 77 | struct kmem_cache *slab; |
| 78 | unsigned int slab_ref; |
| 79 | unsigned int slab_size; |
| 80 | char name[8]; |
| 81 | }; |
| 82 | static DEFINE_MUTEX(bio_slab_lock); |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 83 | static DEFINE_XARRAY(bio_slabs); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 84 | |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 85 | static struct bio_slab *create_bio_slab(unsigned int size) |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 86 | { |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 87 | struct bio_slab *bslab = kzalloc(sizeof(*bslab), GFP_KERNEL); |
| 88 | |
| 89 | if (!bslab) |
| 90 | return NULL; |
| 91 | |
| 92 | snprintf(bslab->name, sizeof(bslab->name), "bio-%d", size); |
| 93 | bslab->slab = kmem_cache_create(bslab->name, size, |
Christoph Hellwig | 1a7e76e | 2021-10-12 13:12:22 +0200 | [diff] [blame] | 94 | ARCH_KMALLOC_MINALIGN, |
| 95 | SLAB_HWCACHE_ALIGN | SLAB_TYPESAFE_BY_RCU, NULL); |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 96 | if (!bslab->slab) |
| 97 | goto fail_alloc_slab; |
| 98 | |
| 99 | bslab->slab_ref = 1; |
| 100 | bslab->slab_size = size; |
| 101 | |
| 102 | if (!xa_err(xa_store(&bio_slabs, size, bslab, GFP_KERNEL))) |
| 103 | return bslab; |
| 104 | |
| 105 | kmem_cache_destroy(bslab->slab); |
| 106 | |
| 107 | fail_alloc_slab: |
| 108 | kfree(bslab); |
| 109 | return NULL; |
| 110 | } |
| 111 | |
| 112 | static inline unsigned int bs_bio_slab_size(struct bio_set *bs) |
| 113 | { |
Ming Lei | 9f180e3 | 2021-01-11 11:05:54 +0800 | [diff] [blame] | 114 | return bs->front_pad + sizeof(struct bio) + bs->back_pad; |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 115 | } |
| 116 | |
| 117 | static struct kmem_cache *bio_find_or_create_slab(struct bio_set *bs) |
| 118 | { |
| 119 | unsigned int size = bs_bio_slab_size(bs); |
| 120 | struct bio_slab *bslab; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 121 | |
| 122 | mutex_lock(&bio_slab_lock); |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 123 | bslab = xa_load(&bio_slabs, size); |
| 124 | if (bslab) |
| 125 | bslab->slab_ref++; |
| 126 | else |
| 127 | bslab = create_bio_slab(size); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 128 | mutex_unlock(&bio_slab_lock); |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 129 | |
| 130 | if (bslab) |
| 131 | return bslab->slab; |
| 132 | return NULL; |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 133 | } |
| 134 | |
| 135 | static void bio_put_slab(struct bio_set *bs) |
| 136 | { |
| 137 | struct bio_slab *bslab = NULL; |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 138 | unsigned int slab_size = bs_bio_slab_size(bs); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 139 | |
| 140 | mutex_lock(&bio_slab_lock); |
| 141 | |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 142 | bslab = xa_load(&bio_slabs, slab_size); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 143 | if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) |
| 144 | goto out; |
| 145 | |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 146 | WARN_ON_ONCE(bslab->slab != bs->bio_slab); |
| 147 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 148 | WARN_ON(!bslab->slab_ref); |
| 149 | |
| 150 | if (--bslab->slab_ref) |
| 151 | goto out; |
| 152 | |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 153 | xa_erase(&bio_slabs, slab_size); |
| 154 | |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 155 | kmem_cache_destroy(bslab->slab); |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 156 | kfree(bslab); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 157 | |
| 158 | out: |
| 159 | mutex_unlock(&bio_slab_lock); |
| 160 | } |
| 161 | |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 162 | void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned short nr_vecs) |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 163 | { |
Christoph Hellwig | 9e8c0d0 | 2021-10-12 18:17:57 +0200 | [diff] [blame] | 164 | BUG_ON(nr_vecs > BIO_MAX_VECS); |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 165 | |
Christoph Hellwig | a8affc0 | 2021-03-11 12:01:37 +0100 | [diff] [blame] | 166 | if (nr_vecs == BIO_MAX_VECS) |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 167 | mempool_free(bv, pool); |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 168 | else if (nr_vecs > BIO_INLINE_VECS) |
| 169 | kmem_cache_free(biovec_slab(nr_vecs)->slab, bv); |
Jens Axboe | bb799ca | 2008-12-10 15:35:05 +0100 | [diff] [blame] | 170 | } |
| 171 | |
Christoph Hellwig | f2c3eb9 | 2021-02-02 18:19:21 +0100 | [diff] [blame] | 172 | /* |
| 173 | * Make the first allocation restricted and don't dump info on allocation |
| 174 | * failures, since we'll fall back to the mempool in case of failure. |
| 175 | */ |
| 176 | static inline gfp_t bvec_alloc_gfp(gfp_t gfp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 177 | { |
Christoph Hellwig | f2c3eb9 | 2021-02-02 18:19:21 +0100 | [diff] [blame] | 178 | return (gfp & ~(__GFP_DIRECT_RECLAIM | __GFP_IO)) | |
| 179 | __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; |
| 180 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 181 | |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 182 | struct bio_vec *bvec_alloc(mempool_t *pool, unsigned short *nr_vecs, |
| 183 | gfp_t gfp_mask) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 184 | { |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 185 | struct biovec_slab *bvs = biovec_slab(*nr_vecs); |
| 186 | |
| 187 | if (WARN_ON_ONCE(!bvs)) |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 188 | return NULL; |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 189 | |
| 190 | /* |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 191 | * Upgrade the nr_vecs request to take full advantage of the allocation. |
| 192 | * We also rely on this in the bvec_free path. |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 193 | */ |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 194 | *nr_vecs = bvs->nr_vecs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 195 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 196 | /* |
Christoph Hellwig | f007a3d | 2021-02-02 18:19:22 +0100 | [diff] [blame] | 197 | * Try a slab allocation first for all smaller allocations. If that |
| 198 | * fails and __GFP_DIRECT_RECLAIM is set retry with the mempool. |
Christoph Hellwig | a8affc0 | 2021-03-11 12:01:37 +0100 | [diff] [blame] | 199 | * The mempool is sized to handle up to BIO_MAX_VECS entries. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 200 | */ |
Christoph Hellwig | a8affc0 | 2021-03-11 12:01:37 +0100 | [diff] [blame] | 201 | if (*nr_vecs < BIO_MAX_VECS) { |
Christoph Hellwig | f007a3d | 2021-02-02 18:19:22 +0100 | [diff] [blame] | 202 | struct bio_vec *bvl; |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 203 | |
Christoph Hellwig | f2c3eb9 | 2021-02-02 18:19:21 +0100 | [diff] [blame] | 204 | bvl = kmem_cache_alloc(bvs->slab, bvec_alloc_gfp(gfp_mask)); |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 205 | if (likely(bvl) || !(gfp_mask & __GFP_DIRECT_RECLAIM)) |
Christoph Hellwig | f007a3d | 2021-02-02 18:19:22 +0100 | [diff] [blame] | 206 | return bvl; |
Christoph Hellwig | a8affc0 | 2021-03-11 12:01:37 +0100 | [diff] [blame] | 207 | *nr_vecs = BIO_MAX_VECS; |
Jens Axboe | 7ff9345 | 2008-12-11 11:53:43 +0100 | [diff] [blame] | 208 | } |
| 209 | |
Christoph Hellwig | f007a3d | 2021-02-02 18:19:22 +0100 | [diff] [blame] | 210 | return mempool_alloc(pool, gfp_mask); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 211 | } |
| 212 | |
Jens Axboe | 9ae3b3f5 | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 213 | void bio_uninit(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 214 | { |
Christoph Hellwig | db9819c | 2020-06-27 09:31:47 +0200 | [diff] [blame] | 215 | #ifdef CONFIG_BLK_CGROUP |
| 216 | if (bio->bi_blkg) { |
| 217 | blkg_put(bio->bi_blkg); |
| 218 | bio->bi_blkg = NULL; |
| 219 | } |
| 220 | #endif |
Justin Tee | ece841a | 2019-12-05 10:09:01 +0800 | [diff] [blame] | 221 | if (bio_integrity(bio)) |
| 222 | bio_integrity_free(bio); |
Satya Tangirala | a892c8d | 2020-05-14 00:37:18 +0000 | [diff] [blame] | 223 | |
| 224 | bio_crypt_free_ctx(bio); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 225 | } |
Jens Axboe | 9ae3b3f5 | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 226 | EXPORT_SYMBOL(bio_uninit); |
Martin K. Petersen | 7ba1ba1 | 2008-06-30 20:04:41 +0200 | [diff] [blame] | 227 | |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 228 | static void bio_free(struct bio *bio) |
| 229 | { |
| 230 | struct bio_set *bs = bio->bi_pool; |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 231 | void *p = bio; |
| 232 | |
| 233 | WARN_ON_ONCE(!bs); |
Kent Overstreet | 4254bba | 2012-09-06 15:35:00 -0700 | [diff] [blame] | 234 | |
Jens Axboe | 9ae3b3f5 | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 235 | bio_uninit(bio); |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 236 | bvec_free(&bs->bvec_pool, bio->bi_io_vec, bio->bi_max_vecs); |
| 237 | mempool_free(p - bs->front_pad, &bs->bio_pool); |
Peter Osterlund | 3676347 | 2005-09-06 15:16:42 -0700 | [diff] [blame] | 238 | } |
| 239 | |
Jens Axboe | 9ae3b3f5 | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 240 | /* |
| 241 | * Users of this function have their own bio allocation. Subsequently, |
| 242 | * they must remember to pair any call to bio_init() with bio_uninit() |
| 243 | * when IO has completed, or when the bio is released. |
| 244 | */ |
Christoph Hellwig | 49add49 | 2022-01-24 10:11:06 +0100 | [diff] [blame] | 245 | void bio_init(struct bio *bio, struct block_device *bdev, struct bio_vec *table, |
Bart Van Assche | 16458cf | 2022-07-14 11:06:32 -0700 | [diff] [blame] | 246 | unsigned short max_vecs, blk_opf_t opf) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 247 | { |
Jens Axboe | da52162 | 2021-08-11 09:20:04 -0600 | [diff] [blame] | 248 | bio->bi_next = NULL; |
Christoph Hellwig | 49add49 | 2022-01-24 10:11:06 +0100 | [diff] [blame] | 249 | bio->bi_bdev = bdev; |
| 250 | bio->bi_opf = opf; |
Jens Axboe | da52162 | 2021-08-11 09:20:04 -0600 | [diff] [blame] | 251 | bio->bi_flags = 0; |
| 252 | bio->bi_ioprio = 0; |
Bart Van Assche | 4498135 | 2024-02-02 12:39:25 -0800 | [diff] [blame] | 253 | bio->bi_write_hint = 0; |
Jens Axboe | da52162 | 2021-08-11 09:20:04 -0600 | [diff] [blame] | 254 | bio->bi_status = 0; |
| 255 | bio->bi_iter.bi_sector = 0; |
| 256 | bio->bi_iter.bi_size = 0; |
| 257 | bio->bi_iter.bi_idx = 0; |
| 258 | bio->bi_iter.bi_bvec_done = 0; |
| 259 | bio->bi_end_io = NULL; |
| 260 | bio->bi_private = NULL; |
| 261 | #ifdef CONFIG_BLK_CGROUP |
| 262 | bio->bi_blkg = NULL; |
| 263 | bio->bi_issue.value = 0; |
Christoph Hellwig | 49add49 | 2022-01-24 10:11:06 +0100 | [diff] [blame] | 264 | if (bdev) |
| 265 | bio_associate_blkg(bio); |
Jens Axboe | da52162 | 2021-08-11 09:20:04 -0600 | [diff] [blame] | 266 | #ifdef CONFIG_BLK_CGROUP_IOCOST |
| 267 | bio->bi_iocost_cost = 0; |
| 268 | #endif |
| 269 | #endif |
| 270 | #ifdef CONFIG_BLK_INLINE_ENCRYPTION |
| 271 | bio->bi_crypt_context = NULL; |
| 272 | #endif |
| 273 | #ifdef CONFIG_BLK_DEV_INTEGRITY |
| 274 | bio->bi_integrity = NULL; |
| 275 | #endif |
| 276 | bio->bi_vcnt = 0; |
| 277 | |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 278 | atomic_set(&bio->__bi_remaining, 1); |
Jens Axboe | dac5621 | 2015-04-17 16:23:59 -0600 | [diff] [blame] | 279 | atomic_set(&bio->__bi_cnt, 1); |
Christoph Hellwig | 3e08773 | 2021-10-12 13:12:24 +0200 | [diff] [blame] | 280 | bio->bi_cookie = BLK_QC_T_NONE; |
Ming Lei | 3a83f46 | 2016-11-22 08:57:21 -0700 | [diff] [blame] | 281 | |
Ming Lei | 3a83f46 | 2016-11-22 08:57:21 -0700 | [diff] [blame] | 282 | bio->bi_max_vecs = max_vecs; |
Jens Axboe | da52162 | 2021-08-11 09:20:04 -0600 | [diff] [blame] | 283 | bio->bi_io_vec = table; |
| 284 | bio->bi_pool = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 285 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 286 | EXPORT_SYMBOL(bio_init); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 287 | |
| 288 | /** |
Kent Overstreet | f44b48c7 | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 289 | * bio_reset - reinitialize a bio |
| 290 | * @bio: bio to reset |
Christoph Hellwig | a7c50c9 | 2022-01-24 10:11:07 +0100 | [diff] [blame] | 291 | * @bdev: block device to use the bio for |
| 292 | * @opf: operation and flags for bio |
Kent Overstreet | f44b48c7 | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 293 | * |
| 294 | * Description: |
| 295 | * After calling bio_reset(), @bio will be in the same state as a freshly |
| 296 | * allocated bio returned bio bio_alloc_bioset() - the only fields that are |
| 297 | * preserved are the ones that are initialized by bio_alloc_bioset(). See |
| 298 | * comment in struct bio. |
| 299 | */ |
Bart Van Assche | 16458cf | 2022-07-14 11:06:32 -0700 | [diff] [blame] | 300 | void bio_reset(struct bio *bio, struct block_device *bdev, blk_opf_t opf) |
Kent Overstreet | f44b48c7 | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 301 | { |
Jens Axboe | 9ae3b3f5 | 2017-06-28 15:30:13 -0600 | [diff] [blame] | 302 | bio_uninit(bio); |
Kent Overstreet | f44b48c7 | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 303 | memset(bio, 0, BIO_RESET_BYTES); |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 304 | atomic_set(&bio->__bi_remaining, 1); |
Christoph Hellwig | a7c50c9 | 2022-01-24 10:11:07 +0100 | [diff] [blame] | 305 | bio->bi_bdev = bdev; |
Christoph Hellwig | 78e3437 | 2022-02-04 08:19:34 +0100 | [diff] [blame] | 306 | if (bio->bi_bdev) |
| 307 | bio_associate_blkg(bio); |
Christoph Hellwig | a7c50c9 | 2022-01-24 10:11:07 +0100 | [diff] [blame] | 308 | bio->bi_opf = opf; |
Kent Overstreet | f44b48c7 | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 309 | } |
| 310 | EXPORT_SYMBOL(bio_reset); |
| 311 | |
Christoph Hellwig | 38f8baa | 2016-03-11 17:34:51 +0100 | [diff] [blame] | 312 | static struct bio *__bio_chain_endio(struct bio *bio) |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 313 | { |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 314 | struct bio *parent = bio->bi_private; |
| 315 | |
Yufen Yu | 3edf534 | 2021-03-31 07:53:59 -0400 | [diff] [blame] | 316 | if (bio->bi_status && !parent->bi_status) |
Christoph Hellwig | 4e4cbee | 2017-06-03 09:38:06 +0200 | [diff] [blame] | 317 | parent->bi_status = bio->bi_status; |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 318 | bio_put(bio); |
Christoph Hellwig | 38f8baa | 2016-03-11 17:34:51 +0100 | [diff] [blame] | 319 | return parent; |
| 320 | } |
| 321 | |
| 322 | static void bio_chain_endio(struct bio *bio) |
| 323 | { |
| 324 | bio_endio(__bio_chain_endio(bio)); |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 325 | } |
| 326 | |
| 327 | /** |
| 328 | * bio_chain - chain bio completions |
Randy Dunlap | 1051a90 | 2014-04-20 16:03:31 -0700 | [diff] [blame] | 329 | * @bio: the target bio |
Mauro Carvalho Chehab | 5b874af | 2020-08-27 07:51:44 +0200 | [diff] [blame] | 330 | * @parent: the parent bio of @bio |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 331 | * |
| 332 | * The caller won't have a bi_end_io called when @bio completes - instead, |
| 333 | * @parent's bi_end_io won't be called until both @parent and @bio have |
| 334 | * completed; the chained bio will also be freed when it completes. |
| 335 | * |
| 336 | * The caller must not set bi_private or bi_end_io in @bio. |
| 337 | */ |
| 338 | void bio_chain(struct bio *bio, struct bio *parent) |
| 339 | { |
| 340 | BUG_ON(bio->bi_private || bio->bi_end_io); |
| 341 | |
| 342 | bio->bi_private = parent; |
| 343 | bio->bi_end_io = bio_chain_endio; |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 344 | bio_inc_remaining(parent); |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 345 | } |
| 346 | EXPORT_SYMBOL(bio_chain); |
| 347 | |
Christoph Hellwig | 81c2168 | 2024-05-06 06:20:24 +0200 | [diff] [blame] | 348 | /** |
| 349 | * bio_chain_and_submit - submit a bio after chaining it to another one |
| 350 | * @prev: bio to chain and submit |
| 351 | * @new: bio to chain to |
| 352 | * |
| 353 | * If @prev is non-NULL, chain it to @new and submit it. |
| 354 | * |
| 355 | * Return: @new. |
| 356 | */ |
| 357 | struct bio *bio_chain_and_submit(struct bio *prev, struct bio *new) |
| 358 | { |
| 359 | if (prev) { |
| 360 | bio_chain(prev, new); |
| 361 | submit_bio(prev); |
| 362 | } |
| 363 | return new; |
| 364 | } |
| 365 | |
Chaitanya Kulkarni | 0a3140e | 2022-01-24 10:11:02 +0100 | [diff] [blame] | 366 | struct bio *blk_next_bio(struct bio *bio, struct block_device *bdev, |
Bart Van Assche | 16458cf | 2022-07-14 11:06:32 -0700 | [diff] [blame] | 367 | unsigned int nr_pages, blk_opf_t opf, gfp_t gfp) |
Christoph Hellwig | 3b005bf | 2022-01-24 10:11:01 +0100 | [diff] [blame] | 368 | { |
Christoph Hellwig | 81c2168 | 2024-05-06 06:20:24 +0200 | [diff] [blame] | 369 | return bio_chain_and_submit(bio, bio_alloc(bdev, nr_pages, opf, gfp)); |
Christoph Hellwig | 3b005bf | 2022-01-24 10:11:01 +0100 | [diff] [blame] | 370 | } |
| 371 | EXPORT_SYMBOL_GPL(blk_next_bio); |
| 372 | |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 373 | static void bio_alloc_rescue(struct work_struct *work) |
| 374 | { |
| 375 | struct bio_set *bs = container_of(work, struct bio_set, rescue_work); |
| 376 | struct bio *bio; |
| 377 | |
| 378 | while (1) { |
| 379 | spin_lock(&bs->rescue_lock); |
| 380 | bio = bio_list_pop(&bs->rescue_list); |
| 381 | spin_unlock(&bs->rescue_lock); |
| 382 | |
| 383 | if (!bio) |
| 384 | break; |
| 385 | |
Christoph Hellwig | ed00aab | 2020-07-01 10:59:44 +0200 | [diff] [blame] | 386 | submit_bio_noacct(bio); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 387 | } |
| 388 | } |
| 389 | |
| 390 | static void punt_bios_to_rescuer(struct bio_set *bs) |
| 391 | { |
| 392 | struct bio_list punt, nopunt; |
| 393 | struct bio *bio; |
| 394 | |
NeilBrown | 47e0fb4 | 2017-06-18 14:38:57 +1000 | [diff] [blame] | 395 | if (WARN_ON_ONCE(!bs->rescue_workqueue)) |
| 396 | return; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 397 | /* |
| 398 | * In order to guarantee forward progress we must punt only bios that |
| 399 | * were allocated from this bio_set; otherwise, if there was a bio on |
| 400 | * there for a stacking driver higher up in the stack, processing it |
| 401 | * could require allocating bios from this bio_set, and doing that from |
| 402 | * our own rescuer would be bad. |
| 403 | * |
| 404 | * Since bio lists are singly linked, pop them all instead of trying to |
| 405 | * remove from the middle of the list: |
| 406 | */ |
| 407 | |
| 408 | bio_list_init(&punt); |
| 409 | bio_list_init(&nopunt); |
| 410 | |
NeilBrown | f5fe1b5 | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 411 | while ((bio = bio_list_pop(¤t->bio_list[0]))) |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 412 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
NeilBrown | f5fe1b5 | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 413 | current->bio_list[0] = nopunt; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 414 | |
NeilBrown | f5fe1b5 | 2017-03-10 17:00:47 +1100 | [diff] [blame] | 415 | bio_list_init(&nopunt); |
| 416 | while ((bio = bio_list_pop(¤t->bio_list[1]))) |
| 417 | bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio); |
| 418 | current->bio_list[1] = nopunt; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 419 | |
| 420 | spin_lock(&bs->rescue_lock); |
| 421 | bio_list_merge(&bs->rescue_list, &punt); |
| 422 | spin_unlock(&bs->rescue_lock); |
| 423 | |
| 424 | queue_work(bs->rescue_workqueue, &bs->rescue_work); |
| 425 | } |
| 426 | |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 427 | static void bio_alloc_irq_cache_splice(struct bio_alloc_cache *cache) |
| 428 | { |
| 429 | unsigned long flags; |
| 430 | |
| 431 | /* cache->free_list must be empty */ |
| 432 | if (WARN_ON_ONCE(cache->free_list)) |
| 433 | return; |
| 434 | |
| 435 | local_irq_save(flags); |
| 436 | cache->free_list = cache->free_list_irq; |
| 437 | cache->free_list_irq = NULL; |
| 438 | cache->nr += cache->nr_irq; |
| 439 | cache->nr_irq = 0; |
| 440 | local_irq_restore(flags); |
| 441 | } |
| 442 | |
Mike Snitzer | 0df7165 | 2022-03-24 16:35:24 -0400 | [diff] [blame] | 443 | static struct bio *bio_alloc_percpu_cache(struct block_device *bdev, |
Bart Van Assche | 16458cf | 2022-07-14 11:06:32 -0700 | [diff] [blame] | 444 | unsigned short nr_vecs, blk_opf_t opf, gfp_t gfp, |
Mike Snitzer | 0df7165 | 2022-03-24 16:35:24 -0400 | [diff] [blame] | 445 | struct bio_set *bs) |
| 446 | { |
| 447 | struct bio_alloc_cache *cache; |
| 448 | struct bio *bio; |
| 449 | |
| 450 | cache = per_cpu_ptr(bs->cache, get_cpu()); |
| 451 | if (!cache->free_list) { |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 452 | if (READ_ONCE(cache->nr_irq) >= ALLOC_CACHE_THRESHOLD) |
| 453 | bio_alloc_irq_cache_splice(cache); |
| 454 | if (!cache->free_list) { |
| 455 | put_cpu(); |
| 456 | return NULL; |
| 457 | } |
Mike Snitzer | 0df7165 | 2022-03-24 16:35:24 -0400 | [diff] [blame] | 458 | } |
| 459 | bio = cache->free_list; |
| 460 | cache->free_list = bio->bi_next; |
| 461 | cache->nr--; |
| 462 | put_cpu(); |
| 463 | |
| 464 | bio_init(bio, bdev, nr_vecs ? bio->bi_inline_vecs : NULL, nr_vecs, opf); |
| 465 | bio->bi_pool = bs; |
| 466 | return bio; |
| 467 | } |
| 468 | |
Kent Overstreet | f44b48c7 | 2012-09-06 15:34:58 -0700 | [diff] [blame] | 469 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 470 | * bio_alloc_bioset - allocate a bio for I/O |
Christoph Hellwig | 609be10 | 2022-01-24 10:11:03 +0100 | [diff] [blame] | 471 | * @bdev: block device to allocate the bio for (can be %NULL) |
| 472 | * @nr_vecs: number of bvecs to pre-allocate |
| 473 | * @opf: operation and flags for bio |
Randy Dunlap | 519c8e9 | 2017-10-16 11:01:00 -0700 | [diff] [blame] | 474 | * @gfp_mask: the GFP_* mask given to the slab allocator |
Jaak Ristioja | db18efa | 2010-01-15 12:05:07 +0200 | [diff] [blame] | 475 | * @bs: the bio_set to allocate from. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 476 | * |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 477 | * Allocate a bio from the mempools in @bs. |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 478 | * |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 479 | * If %__GFP_DIRECT_RECLAIM is set then bio_alloc will always be able to |
| 480 | * allocate a bio. This is due to the mempool guarantees. To make this work, |
| 481 | * callers must never allocate more than 1 bio at a time from the general pool. |
| 482 | * Callers that need to allocate more than 1 bio must always submit the |
| 483 | * previously allocated bio for IO before attempting to allocate a new one. |
| 484 | * Failure to do so can cause deadlocks under memory pressure. |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 485 | * |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 486 | * Note that when running under submit_bio_noacct() (i.e. any block driver), |
| 487 | * bios are not submitted until after you return - see the code in |
| 488 | * submit_bio_noacct() that converts recursion into iteration, to prevent |
| 489 | * stack overflows. |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 490 | * |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 491 | * This would normally mean allocating multiple bios under submit_bio_noacct() |
| 492 | * would be susceptible to deadlocks, but we have |
| 493 | * deadlock avoidance code that resubmits any blocked bios from a rescuer |
| 494 | * thread. |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 495 | * |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 496 | * However, we do not guarantee forward progress for allocations from other |
| 497 | * mempools. Doing multiple allocations from the same mempool under |
| 498 | * submit_bio_noacct() should be avoided - instead, use bio_set's front_pad |
| 499 | * for per bio allocations. |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 500 | * |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 501 | * Returns: Pointer to new bio on success, NULL on failure. |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 502 | */ |
Christoph Hellwig | 609be10 | 2022-01-24 10:11:03 +0100 | [diff] [blame] | 503 | struct bio *bio_alloc_bioset(struct block_device *bdev, unsigned short nr_vecs, |
Bart Van Assche | 16458cf | 2022-07-14 11:06:32 -0700 | [diff] [blame] | 504 | blk_opf_t opf, gfp_t gfp_mask, |
Dan Carpenter | 7a88fa1 | 2017-03-23 13:24:55 +0300 | [diff] [blame] | 505 | struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 506 | { |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 507 | gfp_t saved_gfp = gfp_mask; |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 508 | struct bio *bio; |
| 509 | void *p; |
Jens Axboe | 0a0d96b | 2008-09-11 13:17:37 +0200 | [diff] [blame] | 510 | |
Christoph Hellwig | 609be10 | 2022-01-24 10:11:03 +0100 | [diff] [blame] | 511 | /* should not use nobvec bioset for nr_vecs > 0 */ |
| 512 | if (WARN_ON_ONCE(!mempool_initialized(&bs->bvec_pool) && nr_vecs > 0)) |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 513 | return NULL; |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 514 | |
Mike Snitzer | 0df7165 | 2022-03-24 16:35:24 -0400 | [diff] [blame] | 515 | if (opf & REQ_ALLOC_CACHE) { |
| 516 | if (bs->cache && nr_vecs <= BIO_INLINE_VECS) { |
| 517 | bio = bio_alloc_percpu_cache(bdev, nr_vecs, opf, |
| 518 | gfp_mask, bs); |
| 519 | if (bio) |
| 520 | return bio; |
| 521 | /* |
| 522 | * No cached bio available, bio returned below marked with |
| 523 | * REQ_ALLOC_CACHE to particpate in per-cpu alloc cache. |
| 524 | */ |
| 525 | } else { |
| 526 | opf &= ~REQ_ALLOC_CACHE; |
| 527 | } |
| 528 | } |
| 529 | |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 530 | /* |
| 531 | * submit_bio_noacct() converts recursion to iteration; this means if |
| 532 | * we're running beneath it, any bios we allocate and submit will not be |
| 533 | * submitted (and thus freed) until after we return. |
| 534 | * |
| 535 | * This exposes us to a potential deadlock if we allocate multiple bios |
| 536 | * from the same bio_set() while running underneath submit_bio_noacct(). |
| 537 | * If we were to allocate multiple bios (say a stacking block driver |
| 538 | * that was splitting bios), we would deadlock if we exhausted the |
| 539 | * mempool's reserve. |
| 540 | * |
| 541 | * We solve this, and guarantee forward progress, with a rescuer |
| 542 | * workqueue per bio_set. If we go to allocate and there are bios on |
| 543 | * current->bio_list, we first try the allocation without |
| 544 | * __GFP_DIRECT_RECLAIM; if that fails, we punt those bios we would be |
| 545 | * blocking to the rescuer workqueue before we retry with the original |
| 546 | * gfp_flags. |
| 547 | */ |
| 548 | if (current->bio_list && |
| 549 | (!bio_list_empty(¤t->bio_list[0]) || |
| 550 | !bio_list_empty(¤t->bio_list[1])) && |
| 551 | bs->rescue_workqueue) |
| 552 | gfp_mask &= ~__GFP_DIRECT_RECLAIM; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 553 | |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 554 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
| 555 | if (!p && gfp_mask != saved_gfp) { |
| 556 | punt_bios_to_rescuer(bs); |
| 557 | gfp_mask = saved_gfp; |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 558 | p = mempool_alloc(&bs->bio_pool, gfp_mask); |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 559 | } |
Tejun Heo | 451a9eb | 2009-04-15 19:50:51 +0200 | [diff] [blame] | 560 | if (unlikely(!p)) |
| 561 | return NULL; |
Pavel Begunkov | 759aa12 | 2022-11-02 15:18:20 +0000 | [diff] [blame] | 562 | if (!mempool_is_saturated(&bs->bio_pool)) |
| 563 | opf &= ~REQ_ALLOC_CACHE; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 564 | |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 565 | bio = p + bs->front_pad; |
Christoph Hellwig | 609be10 | 2022-01-24 10:11:03 +0100 | [diff] [blame] | 566 | if (nr_vecs > BIO_INLINE_VECS) { |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 567 | struct bio_vec *bvl = NULL; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 568 | |
Christoph Hellwig | 609be10 | 2022-01-24 10:11:03 +0100 | [diff] [blame] | 569 | bvl = bvec_alloc(&bs->bvec_pool, &nr_vecs, gfp_mask); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 570 | if (!bvl && gfp_mask != saved_gfp) { |
| 571 | punt_bios_to_rescuer(bs); |
| 572 | gfp_mask = saved_gfp; |
Christoph Hellwig | 609be10 | 2022-01-24 10:11:03 +0100 | [diff] [blame] | 573 | bvl = bvec_alloc(&bs->bvec_pool, &nr_vecs, gfp_mask); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 574 | } |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 575 | if (unlikely(!bvl)) |
| 576 | goto err_free; |
Kent Overstreet | a38352e | 2012-05-25 13:03:11 -0700 | [diff] [blame] | 577 | |
Christoph Hellwig | 49add49 | 2022-01-24 10:11:06 +0100 | [diff] [blame] | 578 | bio_init(bio, bdev, bvl, nr_vecs, opf); |
Christoph Hellwig | 609be10 | 2022-01-24 10:11:03 +0100 | [diff] [blame] | 579 | } else if (nr_vecs) { |
Christoph Hellwig | 49add49 | 2022-01-24 10:11:06 +0100 | [diff] [blame] | 580 | bio_init(bio, bdev, bio->bi_inline_vecs, BIO_INLINE_VECS, opf); |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 581 | } else { |
Christoph Hellwig | 49add49 | 2022-01-24 10:11:06 +0100 | [diff] [blame] | 582 | bio_init(bio, bdev, NULL, 0, opf); |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 583 | } |
Kent Overstreet | 3f86a82 | 2012-09-06 15:35:01 -0700 | [diff] [blame] | 584 | |
| 585 | bio->bi_pool = bs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 586 | return bio; |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 587 | |
| 588 | err_free: |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 589 | mempool_free(p, &bs->bio_pool); |
Ingo Molnar | 3405397 | 2009-02-21 11:16:36 +0100 | [diff] [blame] | 590 | return NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 591 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 592 | EXPORT_SYMBOL(bio_alloc_bioset); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 593 | |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 594 | /** |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 595 | * bio_kmalloc - kmalloc a bio |
| 596 | * @nr_vecs: number of bio_vecs to allocate |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 597 | * @gfp_mask: the GFP_* mask given to the slab allocator |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 598 | * |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 599 | * Use kmalloc to allocate a bio (including bvecs). The bio must be initialized |
| 600 | * using bio_init() before use. To free a bio returned from this function use |
| 601 | * kfree() after calling bio_uninit(). A bio returned from this function can |
| 602 | * be reused by calling bio_uninit() before calling bio_init() again. |
| 603 | * |
| 604 | * Note that unlike bio_alloc() or bio_alloc_bioset() allocations from this |
Deming Wang | 340e134 | 2022-10-06 04:44:50 -0400 | [diff] [blame] | 605 | * function are not backed by a mempool can fail. Do not use this function |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 606 | * for allocations in the file system I/O path. |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 607 | * |
| 608 | * Returns: Pointer to new bio on success, NULL on failure. |
| 609 | */ |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 610 | struct bio *bio_kmalloc(unsigned short nr_vecs, gfp_t gfp_mask) |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 611 | { |
| 612 | struct bio *bio; |
| 613 | |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 614 | if (nr_vecs > UIO_MAXIOV) |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 615 | return NULL; |
Christoph Hellwig | 066ff57 | 2022-04-06 08:12:27 +0200 | [diff] [blame] | 616 | return kmalloc(struct_size(bio, bi_inline_vecs, nr_vecs), gfp_mask); |
Christoph Hellwig | 3175199 | 2021-01-26 15:52:34 +0100 | [diff] [blame] | 617 | } |
| 618 | EXPORT_SYMBOL(bio_kmalloc); |
| 619 | |
Kent Overstreet | 649f070 | 2023-08-13 14:26:36 -0400 | [diff] [blame] | 620 | void zero_fill_bio_iter(struct bio *bio, struct bvec_iter start) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 621 | { |
Kent Overstreet | 7988613 | 2013-11-23 17:19:00 -0800 | [diff] [blame] | 622 | struct bio_vec bv; |
| 623 | struct bvec_iter iter; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 624 | |
Kent Overstreet | 649f070 | 2023-08-13 14:26:36 -0400 | [diff] [blame] | 625 | __bio_for_each_segment(bv, bio, iter, start) |
Christoph Hellwig | ab6c340 | 2021-07-27 07:56:36 +0200 | [diff] [blame] | 626 | memzero_bvec(&bv); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 627 | } |
Kent Overstreet | 649f070 | 2023-08-13 14:26:36 -0400 | [diff] [blame] | 628 | EXPORT_SYMBOL(zero_fill_bio_iter); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 629 | |
Ming Lei | 83c9c54 | 2020-01-05 09:41:14 +0800 | [diff] [blame] | 630 | /** |
| 631 | * bio_truncate - truncate the bio to small size of @new_size |
| 632 | * @bio: the bio to be truncated |
| 633 | * @new_size: new size for truncating the bio |
| 634 | * |
| 635 | * Description: |
| 636 | * Truncate the bio to new size of @new_size. If bio_op(bio) is |
| 637 | * REQ_OP_READ, zero the truncated part. This function should only |
| 638 | * be used for handling corner cases, such as bio eod. |
| 639 | */ |
Christoph Hellwig | 4f7ab09 | 2021-10-12 18:18:04 +0200 | [diff] [blame] | 640 | static void bio_truncate(struct bio *bio, unsigned new_size) |
Ming Lei | 85a8ce6 | 2019-12-28 07:05:48 +0800 | [diff] [blame] | 641 | { |
| 642 | struct bio_vec bv; |
| 643 | struct bvec_iter iter; |
| 644 | unsigned int done = 0; |
| 645 | bool truncated = false; |
| 646 | |
| 647 | if (new_size >= bio->bi_iter.bi_size) |
| 648 | return; |
| 649 | |
Ming Lei | 83c9c54 | 2020-01-05 09:41:14 +0800 | [diff] [blame] | 650 | if (bio_op(bio) != REQ_OP_READ) |
Ming Lei | 85a8ce6 | 2019-12-28 07:05:48 +0800 | [diff] [blame] | 651 | goto exit; |
| 652 | |
| 653 | bio_for_each_segment(bv, bio, iter) { |
| 654 | if (done + bv.bv_len > new_size) { |
| 655 | unsigned offset; |
| 656 | |
| 657 | if (!truncated) |
| 658 | offset = new_size - done; |
| 659 | else |
| 660 | offset = 0; |
OGAWA Hirofumi | 3ee859e | 2022-01-09 18:36:43 +0900 | [diff] [blame] | 661 | zero_user(bv.bv_page, bv.bv_offset + offset, |
| 662 | bv.bv_len - offset); |
Ming Lei | 85a8ce6 | 2019-12-28 07:05:48 +0800 | [diff] [blame] | 663 | truncated = true; |
| 664 | } |
| 665 | done += bv.bv_len; |
| 666 | } |
| 667 | |
| 668 | exit: |
| 669 | /* |
| 670 | * Don't touch bvec table here and make it really immutable, since |
| 671 | * fs bio user has to retrieve all pages via bio_for_each_segment_all |
| 672 | * in its .end_bio() callback. |
| 673 | * |
| 674 | * It is enough to truncate bio by updating .bi_size since we can make |
| 675 | * correct bvec with the updated .bi_size for drivers. |
| 676 | */ |
| 677 | bio->bi_iter.bi_size = new_size; |
| 678 | } |
| 679 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 680 | /** |
Christoph Hellwig | 29125ed | 2020-03-25 16:48:40 +0100 | [diff] [blame] | 681 | * guard_bio_eod - truncate a BIO to fit the block device |
| 682 | * @bio: bio to truncate |
| 683 | * |
| 684 | * This allows us to do IO even on the odd last sectors of a device, even if the |
| 685 | * block size is some multiple of the physical sector size. |
| 686 | * |
| 687 | * We'll just truncate the bio to the size of the device, and clear the end of |
| 688 | * the buffer head manually. Truly out-of-range accesses will turn into actual |
| 689 | * I/O errors, this only handles the "we need to be able to do I/O at the final |
| 690 | * sector" case. |
| 691 | */ |
| 692 | void guard_bio_eod(struct bio *bio) |
| 693 | { |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 694 | sector_t maxsector = bdev_nr_sectors(bio->bi_bdev); |
Christoph Hellwig | 29125ed | 2020-03-25 16:48:40 +0100 | [diff] [blame] | 695 | |
| 696 | if (!maxsector) |
| 697 | return; |
| 698 | |
| 699 | /* |
| 700 | * If the *whole* IO is past the end of the device, |
| 701 | * let it through, and the IO layer will turn it into |
| 702 | * an EIO. |
| 703 | */ |
| 704 | if (unlikely(bio->bi_iter.bi_sector >= maxsector)) |
| 705 | return; |
| 706 | |
| 707 | maxsector -= bio->bi_iter.bi_sector; |
| 708 | if (likely((bio->bi_iter.bi_size >> 9) <= maxsector)) |
| 709 | return; |
| 710 | |
| 711 | bio_truncate(bio, maxsector << 9); |
| 712 | } |
| 713 | |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 714 | static int __bio_alloc_cache_prune(struct bio_alloc_cache *cache, |
| 715 | unsigned int nr) |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 716 | { |
| 717 | unsigned int i = 0; |
| 718 | struct bio *bio; |
| 719 | |
Jens Axboe | fcade2c | 2021-12-01 16:19:18 -0700 | [diff] [blame] | 720 | while ((bio = cache->free_list) != NULL) { |
| 721 | cache->free_list = bio->bi_next; |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 722 | cache->nr--; |
| 723 | bio_free(bio); |
| 724 | if (++i == nr) |
| 725 | break; |
| 726 | } |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 727 | return i; |
| 728 | } |
| 729 | |
| 730 | static void bio_alloc_cache_prune(struct bio_alloc_cache *cache, |
| 731 | unsigned int nr) |
| 732 | { |
| 733 | nr -= __bio_alloc_cache_prune(cache, nr); |
| 734 | if (!READ_ONCE(cache->free_list)) { |
| 735 | bio_alloc_irq_cache_splice(cache); |
| 736 | __bio_alloc_cache_prune(cache, nr); |
| 737 | } |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 738 | } |
| 739 | |
| 740 | static int bio_cpu_dead(unsigned int cpu, struct hlist_node *node) |
| 741 | { |
| 742 | struct bio_set *bs; |
| 743 | |
| 744 | bs = hlist_entry_safe(node, struct bio_set, cpuhp_dead); |
| 745 | if (bs->cache) { |
| 746 | struct bio_alloc_cache *cache = per_cpu_ptr(bs->cache, cpu); |
| 747 | |
| 748 | bio_alloc_cache_prune(cache, -1U); |
| 749 | } |
| 750 | return 0; |
| 751 | } |
| 752 | |
| 753 | static void bio_alloc_cache_destroy(struct bio_set *bs) |
| 754 | { |
| 755 | int cpu; |
| 756 | |
| 757 | if (!bs->cache) |
| 758 | return; |
| 759 | |
| 760 | cpuhp_state_remove_instance_nocalls(CPUHP_BIO_DEAD, &bs->cpuhp_dead); |
| 761 | for_each_possible_cpu(cpu) { |
| 762 | struct bio_alloc_cache *cache; |
| 763 | |
| 764 | cache = per_cpu_ptr(bs->cache, cpu); |
| 765 | bio_alloc_cache_prune(cache, -1U); |
| 766 | } |
| 767 | free_percpu(bs->cache); |
Jens Axboe | 605f741 | 2022-05-29 07:13:09 -0600 | [diff] [blame] | 768 | bs->cache = NULL; |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 769 | } |
| 770 | |
Pavel Begunkov | f25cf75 | 2022-11-02 15:18:21 +0000 | [diff] [blame] | 771 | static inline void bio_put_percpu_cache(struct bio *bio) |
| 772 | { |
| 773 | struct bio_alloc_cache *cache; |
| 774 | |
| 775 | cache = per_cpu_ptr(bio->bi_pool->cache, get_cpu()); |
Pavel Begunkov | e516c3f | 2024-02-07 14:14:29 +0000 | [diff] [blame] | 776 | if (READ_ONCE(cache->nr_irq) + cache->nr > ALLOC_CACHE_MAX) |
| 777 | goto out_free; |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 778 | |
Pavel Begunkov | c9f5f3a | 2024-02-07 14:14:28 +0000 | [diff] [blame] | 779 | if (in_task()) { |
Pavel Begunkov | e516c3f | 2024-02-07 14:14:29 +0000 | [diff] [blame] | 780 | bio_uninit(bio); |
Pavel Begunkov | f25cf75 | 2022-11-02 15:18:21 +0000 | [diff] [blame] | 781 | bio->bi_next = cache->free_list; |
Pavel Begunkov | c9f5f3a | 2024-02-07 14:14:28 +0000 | [diff] [blame] | 782 | /* Not necessary but helps not to iopoll already freed bios */ |
Jens Axboe | 11eb695 | 2023-02-24 09:59:44 -0700 | [diff] [blame] | 783 | bio->bi_bdev = NULL; |
Pavel Begunkov | f25cf75 | 2022-11-02 15:18:21 +0000 | [diff] [blame] | 784 | cache->free_list = bio; |
| 785 | cache->nr++; |
Pavel Begunkov | e516c3f | 2024-02-07 14:14:29 +0000 | [diff] [blame] | 786 | } else if (in_hardirq()) { |
| 787 | lockdep_assert_irqs_disabled(); |
Pavel Begunkov | f25cf75 | 2022-11-02 15:18:21 +0000 | [diff] [blame] | 788 | |
Pavel Begunkov | e516c3f | 2024-02-07 14:14:29 +0000 | [diff] [blame] | 789 | bio_uninit(bio); |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 790 | bio->bi_next = cache->free_list_irq; |
| 791 | cache->free_list_irq = bio; |
| 792 | cache->nr_irq++; |
Pavel Begunkov | e516c3f | 2024-02-07 14:14:29 +0000 | [diff] [blame] | 793 | } else { |
| 794 | goto out_free; |
Pavel Begunkov | b99182c | 2022-11-02 15:18:22 +0000 | [diff] [blame] | 795 | } |
Pavel Begunkov | f25cf75 | 2022-11-02 15:18:21 +0000 | [diff] [blame] | 796 | put_cpu(); |
Pavel Begunkov | e516c3f | 2024-02-07 14:14:29 +0000 | [diff] [blame] | 797 | return; |
| 798 | out_free: |
| 799 | put_cpu(); |
| 800 | bio_free(bio); |
Pavel Begunkov | f25cf75 | 2022-11-02 15:18:21 +0000 | [diff] [blame] | 801 | } |
| 802 | |
Christoph Hellwig | 29125ed | 2020-03-25 16:48:40 +0100 | [diff] [blame] | 803 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 804 | * bio_put - release a reference to a bio |
| 805 | * @bio: bio to release reference to |
| 806 | * |
| 807 | * Description: |
| 808 | * Put a reference to a &struct bio, either one you have gotten with |
NeilBrown | 9b10f6a | 2017-06-18 14:38:59 +1000 | [diff] [blame] | 809 | * 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] | 810 | **/ |
| 811 | void bio_put(struct bio *bio) |
| 812 | { |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 813 | if (unlikely(bio_flagged(bio, BIO_REFFED))) { |
Christoph Hellwig | 9e8c0d0 | 2021-10-12 18:17:57 +0200 | [diff] [blame] | 814 | BUG_ON(!atomic_read(&bio->__bi_cnt)); |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 815 | if (!atomic_dec_and_test(&bio->__bi_cnt)) |
| 816 | return; |
| 817 | } |
Pavel Begunkov | f25cf75 | 2022-11-02 15:18:21 +0000 | [diff] [blame] | 818 | if (bio->bi_opf & REQ_ALLOC_CACHE) |
| 819 | bio_put_percpu_cache(bio); |
| 820 | else |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 821 | bio_free(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 822 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 823 | EXPORT_SYMBOL(bio_put); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 824 | |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 825 | static int __bio_clone(struct bio *bio, struct bio *bio_src, gfp_t gfp) |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 826 | { |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 827 | bio_set_flag(bio, BIO_CLONED); |
Hannes Reinecke | ca474b7 | 2018-11-12 10:35:25 -0700 | [diff] [blame] | 828 | bio->bi_ioprio = bio_src->bi_ioprio; |
Bart Van Assche | 4498135 | 2024-02-02 12:39:25 -0800 | [diff] [blame] | 829 | bio->bi_write_hint = bio_src->bi_write_hint; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 830 | bio->bi_iter = bio_src->bi_iter; |
Paolo Valente | 20bd723 | 2016-07-27 07:22:05 +0200 | [diff] [blame] | 831 | |
Christoph Hellwig | 7ecc56c | 2022-05-04 07:29:50 -0700 | [diff] [blame] | 832 | if (bio->bi_bdev) { |
| 833 | if (bio->bi_bdev == bio_src->bi_bdev && |
| 834 | bio_flagged(bio_src, BIO_REMAPPED)) |
| 835 | bio_set_flag(bio, BIO_REMAPPED); |
| 836 | bio_clone_blkg_association(bio, bio_src); |
| 837 | } |
Christoph Hellwig | 56b4b5a | 2022-02-02 17:01:06 +0100 | [diff] [blame] | 838 | |
| 839 | if (bio_crypt_clone(bio, bio_src, gfp) < 0) |
| 840 | return -ENOMEM; |
| 841 | if (bio_integrity(bio_src) && |
| 842 | bio_integrity_clone(bio, bio_src, gfp) < 0) |
| 843 | return -ENOMEM; |
| 844 | return 0; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 845 | } |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 846 | |
| 847 | /** |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 848 | * bio_alloc_clone - clone a bio that shares the original bio's biovec |
| 849 | * @bdev: block_device to clone onto |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 850 | * @bio_src: bio to clone from |
| 851 | * @gfp: allocation priority |
| 852 | * @bs: bio_set to allocate from |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 853 | * |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 854 | * Allocate a new bio that is a clone of @bio_src. The caller owns the returned |
| 855 | * bio, but not the actual data it points to. |
| 856 | * |
| 857 | * The caller must ensure that the return bio is not freed before @bio_src. |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 858 | */ |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 859 | struct bio *bio_alloc_clone(struct block_device *bdev, struct bio *bio_src, |
| 860 | gfp_t gfp, struct bio_set *bs) |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 861 | { |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 862 | struct bio *bio; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 863 | |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 864 | bio = bio_alloc_bioset(bdev, 0, bio_src->bi_opf, gfp, bs); |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 865 | if (!bio) |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 866 | return NULL; |
| 867 | |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 868 | if (__bio_clone(bio, bio_src, gfp) < 0) { |
| 869 | bio_put(bio); |
Christoph Hellwig | 56b4b5a | 2022-02-02 17:01:06 +0100 | [diff] [blame] | 870 | return NULL; |
| 871 | } |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 872 | bio->bi_io_vec = bio_src->bi_io_vec; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 873 | |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 874 | return bio; |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 875 | } |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 876 | EXPORT_SYMBOL(bio_alloc_clone); |
Kent Overstreet | 59d276f | 2013-11-23 18:19:27 -0800 | [diff] [blame] | 877 | |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 878 | /** |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 879 | * bio_init_clone - clone a bio that shares the original bio's biovec |
| 880 | * @bdev: block_device to clone onto |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 881 | * @bio: bio to clone into |
| 882 | * @bio_src: bio to clone from |
| 883 | * @gfp: allocation priority |
| 884 | * |
| 885 | * Initialize a new bio in caller provided memory that is a clone of @bio_src. |
| 886 | * The caller owns the returned bio, but not the actual data it points to. |
| 887 | * |
| 888 | * The caller must ensure that @bio_src is not freed before @bio. |
| 889 | */ |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 890 | int bio_init_clone(struct block_device *bdev, struct bio *bio, |
| 891 | struct bio *bio_src, gfp_t gfp) |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 892 | { |
| 893 | int ret; |
| 894 | |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 895 | bio_init(bio, bdev, bio_src->bi_io_vec, 0, bio_src->bi_opf); |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 896 | ret = __bio_clone(bio, bio_src, gfp); |
| 897 | if (ret) |
| 898 | bio_uninit(bio); |
| 899 | return ret; |
| 900 | } |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 901 | EXPORT_SYMBOL(bio_init_clone); |
Christoph Hellwig | a0e8de7 | 2022-02-02 17:01:08 +0100 | [diff] [blame] | 902 | |
Christoph Hellwig | 9a6083b | 2021-10-12 18:18:01 +0200 | [diff] [blame] | 903 | /** |
| 904 | * bio_full - check if the bio is full |
| 905 | * @bio: bio to check |
| 906 | * @len: length of one segment to be added |
| 907 | * |
| 908 | * Return true if @bio is full and one segment with @len bytes can't be |
| 909 | * added to the bio, otherwise return false |
| 910 | */ |
| 911 | static inline bool bio_full(struct bio *bio, unsigned len) |
| 912 | { |
| 913 | if (bio->bi_vcnt >= bio->bi_max_vecs) |
| 914 | return true; |
| 915 | if (bio->bi_iter.bi_size > UINT_MAX - len) |
| 916 | return true; |
| 917 | return false; |
| 918 | } |
| 919 | |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 920 | static bool bvec_try_merge_page(struct bio_vec *bv, struct page *page, |
| 921 | unsigned int len, unsigned int off, bool *same_page) |
Ming Lei | 5919482 | 2019-03-17 18:01:07 +0800 | [diff] [blame] | 922 | { |
Matthew Wilcox (Oracle) | d816651 | 2020-08-17 20:52:06 +0100 | [diff] [blame] | 923 | size_t bv_end = bv->bv_offset + bv->bv_len; |
| 924 | phys_addr_t vec_end_addr = page_to_phys(bv->bv_page) + bv_end - 1; |
Ming Lei | 5919482 | 2019-03-17 18:01:07 +0800 | [diff] [blame] | 925 | phys_addr_t page_addr = page_to_phys(page); |
| 926 | |
| 927 | if (vec_end_addr + 1 != page_addr + off) |
| 928 | return false; |
| 929 | if (xen_domain() && !xen_biovec_phys_mergeable(bv, page)) |
| 930 | return false; |
Logan Gunthorpe | 49580e69 | 2022-10-21 11:41:11 -0600 | [diff] [blame] | 931 | if (!zone_device_pages_have_same_pgmap(bv->bv_page, page)) |
| 932 | return false; |
Christoph Hellwig | 52d52d1 | 2019-04-11 08:23:31 +0200 | [diff] [blame] | 933 | |
Christoph Hellwig | ff89673 | 2019-06-17 11:14:11 +0200 | [diff] [blame] | 934 | *same_page = ((vec_end_addr & PAGE_MASK) == page_addr); |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 935 | if (!*same_page) { |
| 936 | if (IS_ENABLED(CONFIG_KMSAN)) |
| 937 | return false; |
| 938 | if (bv->bv_page + bv_end / PAGE_SIZE != page + off / PAGE_SIZE) |
| 939 | return false; |
| 940 | } |
Ming Lei | 5919482 | 2019-03-17 18:01:07 +0800 | [diff] [blame] | 941 | |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 942 | bv->bv_len += len; |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 943 | return true; |
Christoph Hellwig | 9774b39 | 2021-10-12 18:18:02 +0200 | [diff] [blame] | 944 | } |
| 945 | |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 946 | /* |
| 947 | * Try to merge a page into a segment, while obeying the hardware segment |
| 948 | * size limit. This is not for normal read/write bios, but for passthrough |
| 949 | * or Zone Append operations that we can't split. |
| 950 | */ |
Jinyoung Choi | 7c8998f | 2023-08-03 11:48:27 +0900 | [diff] [blame] | 951 | bool bvec_try_merge_hw_page(struct request_queue *q, struct bio_vec *bv, |
Christoph Hellwig | ae42f0b | 2023-07-24 09:54:33 -0700 | [diff] [blame] | 952 | struct page *page, unsigned len, unsigned offset, |
| 953 | bool *same_page) |
Ming Lei | 489fbbc | 2019-03-29 15:08:00 +0800 | [diff] [blame] | 954 | { |
| 955 | unsigned long mask = queue_segment_boundary(q); |
| 956 | phys_addr_t addr1 = page_to_phys(bv->bv_page) + bv->bv_offset; |
| 957 | phys_addr_t addr2 = page_to_phys(page) + offset + len - 1; |
| 958 | |
| 959 | if ((addr1 | mask) != (addr2 | mask)) |
| 960 | return false; |
Christoph Hellwig | 3f034c3 | 2023-12-04 18:34:18 +0100 | [diff] [blame] | 961 | if (len > queue_max_segment_size(q) - bv->bv_len) |
Ming Lei | 489fbbc | 2019-03-29 15:08:00 +0800 | [diff] [blame] | 962 | return false; |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 963 | return bvec_try_merge_page(bv, page, len, offset, same_page); |
Ming Lei | 489fbbc | 2019-03-29 15:08:00 +0800 | [diff] [blame] | 964 | } |
| 965 | |
Shaohua Li | f459587 | 2017-03-24 10:34:43 -0700 | [diff] [blame] | 966 | /** |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 967 | * bio_add_hw_page - attempt to add a page to a bio with hw constraints |
| 968 | * @q: the target queue |
| 969 | * @bio: destination bio |
| 970 | * @page: page to add |
| 971 | * @len: vec entry length |
| 972 | * @offset: vec entry offset |
| 973 | * @max_sectors: maximum number of sectors that can be added |
| 974 | * @same_page: return if the segment has been merged inside the same page |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 975 | * |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 976 | * Add a page to a bio while respecting the hardware max_sectors, max_segment |
| 977 | * and gap limitations. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 978 | */ |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 979 | int bio_add_hw_page(struct request_queue *q, struct bio *bio, |
Ming Lei | 1904708 | 2019-03-17 18:01:08 +0800 | [diff] [blame] | 980 | struct page *page, unsigned int len, unsigned int offset, |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 981 | unsigned int max_sectors, bool *same_page) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 982 | { |
Christoph Hellwig | 6ef02df | 2023-12-04 18:34:19 +0100 | [diff] [blame] | 983 | unsigned int max_size = max_sectors << SECTOR_SHIFT; |
| 984 | |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 985 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 986 | return 0; |
| 987 | |
Christoph Hellwig | 6ef02df | 2023-12-04 18:34:19 +0100 | [diff] [blame] | 988 | len = min3(len, max_size, queue_max_segment_size(q)); |
| 989 | if (len > max_size - bio->bi_iter.bi_size) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 990 | return 0; |
| 991 | |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 992 | if (bio->bi_vcnt > 0) { |
Christoph Hellwig | ae42f0b | 2023-07-24 09:54:33 -0700 | [diff] [blame] | 993 | struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1]; |
| 994 | |
| 995 | if (bvec_try_merge_hw_page(q, bv, page, len, offset, |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 996 | same_page)) { |
| 997 | bio->bi_iter.bi_size += len; |
Christoph Hellwig | 384209cd | 2019-08-12 17:39:57 +0200 | [diff] [blame] | 998 | return len; |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 999 | } |
Christoph Hellwig | 320ea86 | 2019-08-12 17:39:56 +0200 | [diff] [blame] | 1000 | |
Christoph Hellwig | cd1d83e | 2023-07-24 09:54:26 -0700 | [diff] [blame] | 1001 | if (bio->bi_vcnt >= |
| 1002 | min(bio->bi_max_vecs, queue_max_segments(q))) |
| 1003 | return 0; |
| 1004 | |
Christoph Hellwig | 320ea86 | 2019-08-12 17:39:56 +0200 | [diff] [blame] | 1005 | /* |
| 1006 | * If the queue doesn't support SG gaps and adding this segment |
| 1007 | * would create a gap, disallow it. |
| 1008 | */ |
Christoph Hellwig | ae42f0b | 2023-07-24 09:54:33 -0700 | [diff] [blame] | 1009 | if (bvec_gap_to_prev(&q->limits, bv, offset)) |
Christoph Hellwig | 320ea86 | 2019-08-12 17:39:56 +0200 | [diff] [blame] | 1010 | return 0; |
Jens Axboe | 80cfd54 | 2006-01-06 09:43:28 +0100 | [diff] [blame] | 1011 | } |
| 1012 | |
Christoph Hellwig | d58cdfa | 2023-02-03 16:06:12 +0100 | [diff] [blame] | 1013 | bvec_set_page(&bio->bi_io_vec[bio->bi_vcnt], page, len, offset); |
Maurizio Lombardi | fcbf6a0 | 2014-12-10 14:16:53 -0800 | [diff] [blame] | 1014 | bio->bi_vcnt++; |
Christoph Hellwig | dcdca75 | 2019-04-25 09:04:35 +0200 | [diff] [blame] | 1015 | bio->bi_iter.bi_size += len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1016 | return len; |
| 1017 | } |
Ming Lei | 1904708 | 2019-03-17 18:01:08 +0800 | [diff] [blame] | 1018 | |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 1019 | /** |
| 1020 | * bio_add_pc_page - attempt to add page to passthrough bio |
| 1021 | * @q: the target queue |
| 1022 | * @bio: destination bio |
| 1023 | * @page: page to add |
| 1024 | * @len: vec entry length |
| 1025 | * @offset: vec entry offset |
| 1026 | * |
| 1027 | * Attempt to add a page to the bio_vec maplist. This can fail for a |
| 1028 | * number of reasons, such as the bio being full or target block device |
| 1029 | * limitations. The target block device must allow bio's up to PAGE_SIZE, |
| 1030 | * so it is always possible to add a single page to an empty bio. |
| 1031 | * |
| 1032 | * This should only be used by passthrough bios. |
| 1033 | */ |
Ming Lei | 1904708 | 2019-03-17 18:01:08 +0800 | [diff] [blame] | 1034 | int bio_add_pc_page(struct request_queue *q, struct bio *bio, |
| 1035 | struct page *page, unsigned int len, unsigned int offset) |
| 1036 | { |
Christoph Hellwig | d1916c8 | 2019-08-12 17:39:58 +0200 | [diff] [blame] | 1037 | bool same_page = false; |
Christoph Hellwig | e458110 | 2020-05-12 17:55:46 +0900 | [diff] [blame] | 1038 | return bio_add_hw_page(q, bio, page, len, offset, |
| 1039 | queue_max_hw_sectors(q), &same_page); |
Ming Lei | 1904708 | 2019-03-17 18:01:08 +0800 | [diff] [blame] | 1040 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1041 | EXPORT_SYMBOL(bio_add_pc_page); |
Mike Christie | 6e68af6 | 2005-11-11 05:30:27 -0600 | [diff] [blame] | 1042 | |
| 1043 | /** |
Johannes Thumshirn | ae29333 | 2021-02-04 19:21:40 +0900 | [diff] [blame] | 1044 | * bio_add_zone_append_page - attempt to add page to zone-append bio |
| 1045 | * @bio: destination bio |
| 1046 | * @page: page to add |
| 1047 | * @len: vec entry length |
| 1048 | * @offset: vec entry offset |
| 1049 | * |
| 1050 | * Attempt to add a page to the bio_vec maplist of a bio that will be submitted |
| 1051 | * for a zone-append request. This can fail for a number of reasons, such as the |
| 1052 | * bio being full or the target block device is not a zoned block device or |
| 1053 | * other limitations of the target block device. The target block device must |
| 1054 | * allow bio's up to PAGE_SIZE, so it is always possible to add a single page |
| 1055 | * to an empty bio. |
| 1056 | * |
| 1057 | * Returns: number of bytes added to the bio, or 0 in case of a failure. |
| 1058 | */ |
| 1059 | int bio_add_zone_append_page(struct bio *bio, struct page *page, |
| 1060 | unsigned int len, unsigned int offset) |
| 1061 | { |
Pavel Begunkov | 3caee463 | 2021-10-14 15:03:28 +0100 | [diff] [blame] | 1062 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
Johannes Thumshirn | ae29333 | 2021-02-04 19:21:40 +0900 | [diff] [blame] | 1063 | bool same_page = false; |
| 1064 | |
| 1065 | if (WARN_ON_ONCE(bio_op(bio) != REQ_OP_ZONE_APPEND)) |
| 1066 | return 0; |
| 1067 | |
Christoph Hellwig | edd1dbc | 2022-07-06 09:03:37 +0200 | [diff] [blame] | 1068 | if (WARN_ON_ONCE(!bdev_is_zoned(bio->bi_bdev))) |
Johannes Thumshirn | ae29333 | 2021-02-04 19:21:40 +0900 | [diff] [blame] | 1069 | return 0; |
| 1070 | |
| 1071 | return bio_add_hw_page(q, bio, page, len, offset, |
| 1072 | queue_max_zone_append_sectors(q), &same_page); |
| 1073 | } |
| 1074 | EXPORT_SYMBOL_GPL(bio_add_zone_append_page); |
| 1075 | |
| 1076 | /** |
Ming Lei | 551879a | 2019-04-23 10:51:04 +0800 | [diff] [blame] | 1077 | * __bio_add_page - add page(s) to a bio in a new segment |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 1078 | * @bio: destination bio |
Ming Lei | 551879a | 2019-04-23 10:51:04 +0800 | [diff] [blame] | 1079 | * @page: start page to add |
| 1080 | * @len: length of the data to add, may cross pages |
| 1081 | * @off: offset of the data relative to @page, may cross pages |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 1082 | * |
| 1083 | * Add the data at @page + @off to @bio as a new bvec. The caller must ensure |
| 1084 | * that @bio has space for another bvec. |
| 1085 | */ |
| 1086 | void __bio_add_page(struct bio *bio, struct page *page, |
| 1087 | unsigned int len, unsigned int off) |
| 1088 | { |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 1089 | WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)); |
Ming Lei | 79d08f8 | 2019-07-01 15:14:46 +0800 | [diff] [blame] | 1090 | WARN_ON_ONCE(bio_full(bio, len)); |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 1091 | |
Christoph Hellwig | d58cdfa | 2023-02-03 16:06:12 +0100 | [diff] [blame] | 1092 | bvec_set_page(&bio->bi_io_vec[bio->bi_vcnt], page, len, off); |
Christoph Hellwig | 0aa69fd | 2018-06-01 09:03:05 -0700 | [diff] [blame] | 1093 | bio->bi_iter.bi_size += len; |
| 1094 | bio->bi_vcnt++; |
| 1095 | } |
| 1096 | EXPORT_SYMBOL_GPL(__bio_add_page); |
| 1097 | |
| 1098 | /** |
Ming Lei | 551879a | 2019-04-23 10:51:04 +0800 | [diff] [blame] | 1099 | * bio_add_page - attempt to add page(s) to bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1100 | * @bio: destination bio |
Ming Lei | 551879a | 2019-04-23 10:51:04 +0800 | [diff] [blame] | 1101 | * @page: start page to add |
| 1102 | * @len: vec entry length, may cross pages |
| 1103 | * @offset: vec entry offset relative to @page, may cross pages |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1104 | * |
Ming Lei | 551879a | 2019-04-23 10:51:04 +0800 | [diff] [blame] | 1105 | * Attempt to add page(s) to the bio_vec maplist. This will only fail |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 1106 | * 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] | 1107 | */ |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 1108 | int bio_add_page(struct bio *bio, struct page *page, |
| 1109 | unsigned int len, unsigned int offset) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1110 | { |
Christoph Hellwig | ff89673 | 2019-06-17 11:14:11 +0200 | [diff] [blame] | 1111 | bool same_page = false; |
| 1112 | |
Christoph Hellwig | 939e1a3 | 2023-07-24 09:54:28 -0700 | [diff] [blame] | 1113 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
| 1114 | return 0; |
Christoph Hellwig | 6136990 | 2023-07-24 09:54:30 -0700 | [diff] [blame] | 1115 | if (bio->bi_iter.bi_size > UINT_MAX - len) |
| 1116 | return 0; |
Christoph Hellwig | 939e1a3 | 2023-07-24 09:54:28 -0700 | [diff] [blame] | 1117 | |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 1118 | if (bio->bi_vcnt > 0 && |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 1119 | bvec_try_merge_page(&bio->bi_io_vec[bio->bi_vcnt - 1], |
| 1120 | page, len, offset, &same_page)) { |
| 1121 | bio->bi_iter.bi_size += len; |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 1122 | return len; |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 1123 | } |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 1124 | |
Christoph Hellwig | 80232b5 | 2023-07-24 09:54:31 -0700 | [diff] [blame] | 1125 | if (bio->bi_vcnt >= bio->bi_max_vecs) |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 1126 | return 0; |
| 1127 | __bio_add_page(bio, page, len, offset); |
Kent Overstreet | c66a14d | 2013-11-23 22:30:22 -0800 | [diff] [blame] | 1128 | return len; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1129 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1130 | EXPORT_SYMBOL(bio_add_page); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1131 | |
Johannes Thumshirn | 7a150f1 | 2023-05-31 04:50:41 -0700 | [diff] [blame] | 1132 | void bio_add_folio_nofail(struct bio *bio, struct folio *folio, size_t len, |
| 1133 | size_t off) |
| 1134 | { |
| 1135 | WARN_ON_ONCE(len > UINT_MAX); |
| 1136 | WARN_ON_ONCE(off > UINT_MAX); |
| 1137 | __bio_add_page(bio, &folio->page, len, off); |
| 1138 | } |
Matthew Wilcox (Oracle) | 8fde439 | 2024-04-25 17:37:56 +0100 | [diff] [blame] | 1139 | EXPORT_SYMBOL_GPL(bio_add_folio_nofail); |
Johannes Thumshirn | 7a150f1 | 2023-05-31 04:50:41 -0700 | [diff] [blame] | 1140 | |
Matthew Wilcox (Oracle) | 85f5a74 | 2021-04-08 01:33:45 -0400 | [diff] [blame] | 1141 | /** |
| 1142 | * bio_add_folio - Attempt to add part of a folio to a bio. |
| 1143 | * @bio: BIO to add to. |
| 1144 | * @folio: Folio to add. |
| 1145 | * @len: How many bytes from the folio to add. |
| 1146 | * @off: First byte in this folio to add. |
| 1147 | * |
| 1148 | * Filesystems that use folios can call this function instead of calling |
| 1149 | * bio_add_page() for each page in the folio. If @off is bigger than |
| 1150 | * PAGE_SIZE, this function can create a bio_vec that starts in a page |
| 1151 | * after the bv_page. BIOs do not support folios that are 4GiB or larger. |
| 1152 | * |
| 1153 | * Return: Whether the addition was successful. |
| 1154 | */ |
| 1155 | bool bio_add_folio(struct bio *bio, struct folio *folio, size_t len, |
| 1156 | size_t off) |
| 1157 | { |
| 1158 | if (len > UINT_MAX || off > UINT_MAX) |
Jiapeng Chong | 455a844 | 2022-01-28 12:34:54 +0800 | [diff] [blame] | 1159 | return false; |
Matthew Wilcox (Oracle) | 85f5a74 | 2021-04-08 01:33:45 -0400 | [diff] [blame] | 1160 | return bio_add_page(bio, &folio->page, len, off) > 0; |
| 1161 | } |
Matthew Wilcox | cd57b77 | 2023-03-24 18:01:03 +0000 | [diff] [blame] | 1162 | EXPORT_SYMBOL(bio_add_folio); |
Matthew Wilcox (Oracle) | 85f5a74 | 2021-04-08 01:33:45 -0400 | [diff] [blame] | 1163 | |
Pavel Begunkov | c809084 | 2021-10-19 22:24:14 +0100 | [diff] [blame] | 1164 | void __bio_release_pages(struct bio *bio, bool mark_dirty) |
Christoph Hellwig | 7321ecb | 2019-04-11 08:23:30 +0200 | [diff] [blame] | 1165 | { |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1166 | struct folio_iter fi; |
Christoph Hellwig | 7321ecb | 2019-04-11 08:23:30 +0200 | [diff] [blame] | 1167 | |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1168 | bio_for_each_folio_all(fi, bio) { |
| 1169 | struct page *page; |
Tony Battersby | 38b4353 | 2024-02-29 13:08:09 -0500 | [diff] [blame] | 1170 | size_t nr_pages; |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1171 | |
| 1172 | if (mark_dirty) { |
| 1173 | folio_lock(fi.folio); |
| 1174 | folio_mark_dirty(fi.folio); |
| 1175 | folio_unlock(fi.folio); |
| 1176 | } |
| 1177 | page = folio_page(fi.folio, fi.offset / PAGE_SIZE); |
Tony Battersby | 38b4353 | 2024-02-29 13:08:09 -0500 | [diff] [blame] | 1178 | nr_pages = (fi.offset + fi.length - 1) / PAGE_SIZE - |
| 1179 | fi.offset / PAGE_SIZE + 1; |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1180 | do { |
| 1181 | bio_release_page(bio, page++); |
Tony Battersby | 38b4353 | 2024-02-29 13:08:09 -0500 | [diff] [blame] | 1182 | } while (--nr_pages != 0); |
Christoph Hellwig | d241a95 | 2019-06-26 15:49:21 +0200 | [diff] [blame] | 1183 | } |
Christoph Hellwig | 7321ecb | 2019-04-11 08:23:30 +0200 | [diff] [blame] | 1184 | } |
Pavel Begunkov | c809084 | 2021-10-19 22:24:14 +0100 | [diff] [blame] | 1185 | EXPORT_SYMBOL_GPL(__bio_release_pages); |
Christoph Hellwig | 7321ecb | 2019-04-11 08:23:30 +0200 | [diff] [blame] | 1186 | |
Pavel Begunkov | 1bb6b81 | 2021-10-27 13:21:07 +0100 | [diff] [blame] | 1187 | void bio_iov_bvec_set(struct bio *bio, struct iov_iter *iter) |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1188 | { |
Pavel Begunkov | fa5fa8e | 2021-10-23 17:21:33 +0100 | [diff] [blame] | 1189 | size_t size = iov_iter_count(iter); |
| 1190 | |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 1191 | WARN_ON_ONCE(bio->bi_max_vecs); |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1192 | |
Pavel Begunkov | fa5fa8e | 2021-10-23 17:21:33 +0100 | [diff] [blame] | 1193 | if (bio_op(bio) == REQ_OP_ZONE_APPEND) { |
| 1194 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
| 1195 | size_t max_sectors = queue_max_zone_append_sectors(q); |
| 1196 | |
| 1197 | size = min(size, max_sectors << SECTOR_SHIFT); |
| 1198 | } |
| 1199 | |
Pavel Begunkov | c42bca9 | 2021-01-09 16:03:03 +0000 | [diff] [blame] | 1200 | bio->bi_vcnt = iter->nr_segs; |
Pavel Begunkov | c42bca9 | 2021-01-09 16:03:03 +0000 | [diff] [blame] | 1201 | bio->bi_io_vec = (struct bio_vec *)iter->bvec; |
| 1202 | bio->bi_iter.bi_bvec_done = iter->iov_offset; |
Pavel Begunkov | fa5fa8e | 2021-10-23 17:21:33 +0100 | [diff] [blame] | 1203 | bio->bi_iter.bi_size = size; |
Christoph Hellwig | 977be01 | 2021-02-02 18:19:27 +0100 | [diff] [blame] | 1204 | bio_set_flag(bio, BIO_CLONED); |
Johannes Thumshirn | 7de55b7 | 2021-03-25 00:47:26 +0900 | [diff] [blame] | 1205 | } |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1206 | |
Keith Busch | c58c007 | 2022-06-10 12:58:21 -0700 | [diff] [blame] | 1207 | static int bio_iov_add_page(struct bio *bio, struct page *page, |
| 1208 | unsigned int len, unsigned int offset) |
| 1209 | { |
| 1210 | bool same_page = false; |
| 1211 | |
Christoph Hellwig | 6136990 | 2023-07-24 09:54:30 -0700 | [diff] [blame] | 1212 | if (WARN_ON_ONCE(bio->bi_iter.bi_size > UINT_MAX - len)) |
| 1213 | return -EIO; |
| 1214 | |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 1215 | if (bio->bi_vcnt > 0 && |
Christoph Hellwig | 858c708 | 2023-07-24 09:54:32 -0700 | [diff] [blame] | 1216 | bvec_try_merge_page(&bio->bi_io_vec[bio->bi_vcnt - 1], |
| 1217 | page, len, offset, &same_page)) { |
| 1218 | bio->bi_iter.bi_size += len; |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 1219 | if (same_page) |
| 1220 | bio_release_page(bio, page); |
Keith Busch | c58c007 | 2022-06-10 12:58:21 -0700 | [diff] [blame] | 1221 | return 0; |
| 1222 | } |
Christoph Hellwig | 0eca8b6 | 2023-07-24 09:54:29 -0700 | [diff] [blame] | 1223 | __bio_add_page(bio, page, len, offset); |
Keith Busch | c58c007 | 2022-06-10 12:58:21 -0700 | [diff] [blame] | 1224 | return 0; |
| 1225 | } |
| 1226 | |
| 1227 | static int bio_iov_add_zone_append_page(struct bio *bio, struct page *page, |
| 1228 | unsigned int len, unsigned int offset) |
| 1229 | { |
| 1230 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
| 1231 | bool same_page = false; |
| 1232 | |
| 1233 | if (bio_add_hw_page(q, bio, page, len, offset, |
| 1234 | queue_max_zone_append_sectors(q), &same_page) != len) |
| 1235 | return -EINVAL; |
| 1236 | if (same_page) |
David Howells | a7e689d | 2023-05-22 21:57:43 +0100 | [diff] [blame] | 1237 | bio_release_page(bio, page); |
Keith Busch | c58c007 | 2022-06-10 12:58:21 -0700 | [diff] [blame] | 1238 | return 0; |
| 1239 | } |
| 1240 | |
Christoph Hellwig | 576ed91 | 2018-09-20 08:28:21 +0200 | [diff] [blame] | 1241 | #define PAGE_PTRS_PER_BVEC (sizeof(struct bio_vec) / sizeof(struct page *)) |
| 1242 | |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 1243 | /** |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1244 | * __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] | 1245 | * @bio: bio to add pages to |
| 1246 | * @iter: iov iterator describing the region to be mapped |
| 1247 | * |
David Howells | a7e689d | 2023-05-22 21:57:43 +0100 | [diff] [blame] | 1248 | * Extracts pages from *iter and appends them to @bio's bvec array. The pages |
| 1249 | * will have to be cleaned up in the way indicated by the BIO_PAGE_PINNED flag. |
| 1250 | * For a multi-segment *iter, this function only adds pages from the next |
| 1251 | * non-empty segment of the iov iterator. |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 1252 | */ |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1253 | 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] | 1254 | { |
David Howells | f62e52d | 2023-01-19 12:47:23 +0000 | [diff] [blame] | 1255 | iov_iter_extraction_t extraction_flags = 0; |
Christoph Hellwig | 576ed91 | 2018-09-20 08:28:21 +0200 | [diff] [blame] | 1256 | unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt; |
| 1257 | unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt; |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 1258 | struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt; |
| 1259 | struct page **pages = (struct page **)bv; |
Christoph Hellwig | 576ed91 | 2018-09-20 08:28:21 +0200 | [diff] [blame] | 1260 | ssize_t size, left; |
Keith Busch | e97424f | 2022-07-12 08:32:56 -0700 | [diff] [blame] | 1261 | unsigned len, i = 0; |
Kent Overstreet | 168145f | 2023-08-13 14:26:35 -0400 | [diff] [blame] | 1262 | size_t offset; |
Keith Busch | 325347d | 2022-07-12 08:32:54 -0700 | [diff] [blame] | 1263 | int ret = 0; |
Christoph Hellwig | 576ed91 | 2018-09-20 08:28:21 +0200 | [diff] [blame] | 1264 | |
| 1265 | /* |
| 1266 | * Move page array up in the allocated memory for the bio vecs as far as |
| 1267 | * possible so that we can start filling biovecs from the beginning |
| 1268 | * without overwriting the temporary page array. |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 1269 | */ |
| 1270 | BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2); |
| 1271 | pages += entries_left * (PAGE_PTRS_PER_BVEC - 1); |
| 1272 | |
Logan Gunthorpe | 5e3e3f2 | 2022-10-21 11:41:13 -0600 | [diff] [blame] | 1273 | if (bio->bi_bdev && blk_queue_pci_p2pdma(bio->bi_bdev->bd_disk->queue)) |
David Howells | f62e52d | 2023-01-19 12:47:23 +0000 | [diff] [blame] | 1274 | extraction_flags |= ITER_ALLOW_P2PDMA; |
Logan Gunthorpe | 5e3e3f2 | 2022-10-21 11:41:13 -0600 | [diff] [blame] | 1275 | |
Keith Busch | b1a000d | 2022-06-10 12:58:29 -0700 | [diff] [blame] | 1276 | /* |
| 1277 | * Each segment in the iov is required to be a block size multiple. |
| 1278 | * However, we may not be able to get the entire segment if it spans |
| 1279 | * more pages than bi_max_vecs allows, so we have to ALIGN_DOWN the |
| 1280 | * result to ensure the bio's total size is correct. The remainder of |
| 1281 | * the iov data will be picked up in the next bio iteration. |
| 1282 | */ |
David Howells | a7e689d | 2023-05-22 21:57:43 +0100 | [diff] [blame] | 1283 | size = iov_iter_extract_pages(iter, &pages, |
| 1284 | UINT_MAX - bio->bi_iter.bi_size, |
| 1285 | nr_pages, extraction_flags, &offset); |
Al Viro | 480cb84 | 2022-06-09 10:37:57 -0400 | [diff] [blame] | 1286 | if (unlikely(size <= 0)) |
| 1287 | return size ? size : -EFAULT; |
Keith Busch | e97424f | 2022-07-12 08:32:56 -0700 | [diff] [blame] | 1288 | |
Al Viro | 480cb84 | 2022-06-09 10:37:57 -0400 | [diff] [blame] | 1289 | nr_pages = DIV_ROUND_UP(offset + size, PAGE_SIZE); |
| 1290 | |
Kent Overstreet | 168145f | 2023-08-13 14:26:35 -0400 | [diff] [blame] | 1291 | if (bio->bi_bdev) { |
| 1292 | size_t trim = size & (bdev_logical_block_size(bio->bi_bdev) - 1); |
| 1293 | iov_iter_revert(iter, trim); |
| 1294 | size -= trim; |
| 1295 | } |
Al Viro | 480cb84 | 2022-06-09 10:37:57 -0400 | [diff] [blame] | 1296 | |
Al Viro | 480cb84 | 2022-06-09 10:37:57 -0400 | [diff] [blame] | 1297 | if (unlikely(!size)) { |
| 1298 | ret = -EFAULT; |
Keith Busch | e97424f | 2022-07-12 08:32:56 -0700 | [diff] [blame] | 1299 | goto out; |
| 1300 | } |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 1301 | |
| 1302 | for (left = size, i = 0; left > 0; left -= len, i++) { |
| 1303 | struct page *page = pages[i]; |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 1304 | |
| 1305 | len = min_t(size_t, PAGE_SIZE - offset, left); |
Keith Busch | 34cdb8c | 2022-07-12 08:32:55 -0700 | [diff] [blame] | 1306 | if (bio_op(bio) == REQ_OP_ZONE_APPEND) { |
Keith Busch | c58c007 | 2022-06-10 12:58:21 -0700 | [diff] [blame] | 1307 | ret = bio_iov_add_zone_append_page(bio, page, len, |
| 1308 | offset); |
Keith Busch | e97424f | 2022-07-12 08:32:56 -0700 | [diff] [blame] | 1309 | if (ret) |
Keith Busch | 34cdb8c | 2022-07-12 08:32:55 -0700 | [diff] [blame] | 1310 | break; |
Keith Busch | 34cdb8c | 2022-07-12 08:32:55 -0700 | [diff] [blame] | 1311 | } else |
| 1312 | bio_iov_add_page(bio, page, len, offset); |
Keith Busch | c58c007 | 2022-06-10 12:58:21 -0700 | [diff] [blame] | 1313 | |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 1314 | offset = 0; |
| 1315 | } |
| 1316 | |
Al Viro | 480cb84 | 2022-06-09 10:37:57 -0400 | [diff] [blame] | 1317 | iov_iter_revert(iter, left); |
Keith Busch | e97424f | 2022-07-12 08:32:56 -0700 | [diff] [blame] | 1318 | out: |
| 1319 | while (i < nr_pages) |
David Howells | a7e689d | 2023-05-22 21:57:43 +0100 | [diff] [blame] | 1320 | bio_release_page(bio, pages[i++]); |
Keith Busch | e97424f | 2022-07-12 08:32:56 -0700 | [diff] [blame] | 1321 | |
Keith Busch | 325347d | 2022-07-12 08:32:54 -0700 | [diff] [blame] | 1322 | return ret; |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 1323 | } |
| 1324 | |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1325 | /** |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1326 | * bio_iov_iter_get_pages - add user or kernel pages to a bio |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1327 | * @bio: bio to add pages to |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1328 | * @iter: iov iterator describing the region to be added |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1329 | * |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1330 | * This takes either an iterator pointing to user memory, or one pointing to |
| 1331 | * kernel pages (BVEC iterator). If we're adding user pages, we pin them and |
| 1332 | * map them into the kernel. On IO completion, the caller should put those |
Pavel Begunkov | c42bca9 | 2021-01-09 16:03:03 +0000 | [diff] [blame] | 1333 | * pages. For bvec based iterators bio_iov_iter_get_pages() uses the provided |
| 1334 | * bvecs rather than copying them. Hence anyone issuing kiocb based IO needs |
| 1335 | * to ensure the bvecs and pages stay referenced until the submitted I/O is |
| 1336 | * completed by a call to ->ki_complete() or returns with an error other than |
| 1337 | * -EIOCBQUEUED. The caller needs to check if the bio is flagged BIO_NO_PAGE_REF |
| 1338 | * on IO completion. If it isn't, then pages should be released. |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1339 | * |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1340 | * The function tries, but does not guarantee, to pin as many pages as |
Mauro Carvalho Chehab | 5cd3ddc | 2020-09-09 11:44:33 +0200 | [diff] [blame] | 1341 | * fit into the bio, or are requested in @iter, whatever is smaller. If |
Jens Axboe | 6d0c48a | 2018-11-30 08:47:03 -0700 | [diff] [blame] | 1342 | * MM encounters an error pinning the requested pages, it stops. Error |
| 1343 | * is returned only if 0 pages could be pinned. |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1344 | */ |
| 1345 | int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter) |
| 1346 | { |
Pavel Begunkov | c42bca9 | 2021-01-09 16:03:03 +0000 | [diff] [blame] | 1347 | int ret = 0; |
Christoph Hellwig | 14eacf1 | 2019-04-11 08:23:29 +0200 | [diff] [blame] | 1348 | |
Christoph Hellwig | 939e1a3 | 2023-07-24 09:54:28 -0700 | [diff] [blame] | 1349 | if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED))) |
| 1350 | return -EIO; |
| 1351 | |
Pavel Begunkov | c42bca9 | 2021-01-09 16:03:03 +0000 | [diff] [blame] | 1352 | if (iov_iter_is_bvec(iter)) { |
Pavel Begunkov | fa5fa8e | 2021-10-23 17:21:33 +0100 | [diff] [blame] | 1353 | bio_iov_bvec_set(bio, iter); |
| 1354 | iov_iter_advance(iter, bio->bi_iter.bi_size); |
| 1355 | return 0; |
Pavel Begunkov | c42bca9 | 2021-01-09 16:03:03 +0000 | [diff] [blame] | 1356 | } |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1357 | |
David Howells | a7e689d | 2023-05-22 21:57:43 +0100 | [diff] [blame] | 1358 | if (iov_iter_extract_will_pin(iter)) |
| 1359 | bio_set_flag(bio, BIO_PAGE_PINNED); |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1360 | do { |
Keith Busch | c58c007 | 2022-06-10 12:58:21 -0700 | [diff] [blame] | 1361 | ret = __bio_iov_iter_get_pages(bio, iter); |
Ming Lei | 79d08f8 | 2019-07-01 15:14:46 +0800 | [diff] [blame] | 1362 | } while (!ret && iov_iter_count(iter) && !bio_full(bio, 0)); |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1363 | |
Christoph Hellwig | 14eacf1 | 2019-04-11 08:23:29 +0200 | [diff] [blame] | 1364 | return bio->bi_vcnt ? 0 : ret; |
Martin Wilck | 17d51b1 | 2018-07-25 23:15:09 +0200 | [diff] [blame] | 1365 | } |
Johannes Thumshirn | 29b2a3a | 2020-05-12 17:55:53 +0900 | [diff] [blame] | 1366 | EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages); |
Kent Overstreet | 2cefe4d | 2016-10-31 11:59:24 -0600 | [diff] [blame] | 1367 | |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1368 | static void submit_bio_wait_endio(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1369 | { |
Christoph Hellwig | 65e53aa | 2017-10-25 17:55:57 +0900 | [diff] [blame] | 1370 | complete(bio->bi_private); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1371 | } |
| 1372 | |
| 1373 | /** |
| 1374 | * submit_bio_wait - submit a bio, and wait until it completes |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1375 | * @bio: The &struct bio which describes the I/O |
| 1376 | * |
| 1377 | * Simple wrapper around submit_bio(). Returns 0 on success, or the error from |
| 1378 | * bio_endio() on failure. |
Jan Kara | 3d289d6 | 2017-08-02 10:25:21 +0200 | [diff] [blame] | 1379 | * |
| 1380 | * WARNING: Unlike to how submit_bio() is usually used, this function does not |
| 1381 | * result in bio reference to be consumed. The caller must drop the reference |
| 1382 | * on his own. |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1383 | */ |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 1384 | int submit_bio_wait(struct bio *bio) |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1385 | { |
Christoph Hellwig | 309dca30 | 2021-01-24 11:02:34 +0100 | [diff] [blame] | 1386 | DECLARE_COMPLETION_ONSTACK_MAP(done, |
| 1387 | bio->bi_bdev->bd_disk->lockdep_map); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1388 | |
Christoph Hellwig | 65e53aa | 2017-10-25 17:55:57 +0900 | [diff] [blame] | 1389 | bio->bi_private = &done; |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1390 | bio->bi_end_io = submit_bio_wait_endio; |
Jens Axboe | 1eff9d3 | 2016-08-05 15:35:16 -0600 | [diff] [blame] | 1391 | bio->bi_opf |= REQ_SYNC; |
Mike Christie | 4e49ea4 | 2016-06-05 14:31:41 -0500 | [diff] [blame] | 1392 | submit_bio(bio); |
Keith Busch | 0eb4db4 | 2024-02-23 07:59:09 -0800 | [diff] [blame] | 1393 | blk_wait_io(&done); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1394 | |
Christoph Hellwig | 65e53aa | 2017-10-25 17:55:57 +0900 | [diff] [blame] | 1395 | return blk_status_to_errno(bio->bi_status); |
Kent Overstreet | 9e88224 | 2012-09-10 14:41:12 -0700 | [diff] [blame] | 1396 | } |
| 1397 | EXPORT_SYMBOL(submit_bio_wait); |
| 1398 | |
Keith Busch | 0f8e9ec | 2024-05-06 06:20:26 +0200 | [diff] [blame] | 1399 | static void bio_wait_end_io(struct bio *bio) |
| 1400 | { |
| 1401 | complete(bio->bi_private); |
| 1402 | bio_put(bio); |
| 1403 | } |
| 1404 | |
| 1405 | /* |
| 1406 | * bio_await_chain - ends @bio and waits for every chained bio to complete |
| 1407 | */ |
| 1408 | void bio_await_chain(struct bio *bio) |
| 1409 | { |
| 1410 | DECLARE_COMPLETION_ONSTACK_MAP(done, |
| 1411 | bio->bi_bdev->bd_disk->lockdep_map); |
| 1412 | |
| 1413 | bio->bi_private = &done; |
| 1414 | bio->bi_end_io = bio_wait_end_io; |
| 1415 | bio_endio(bio); |
| 1416 | blk_wait_io(&done); |
| 1417 | } |
| 1418 | |
Jens Axboe | d4aa57a | 2021-10-13 09:01:43 -0600 | [diff] [blame] | 1419 | void __bio_advance(struct bio *bio, unsigned bytes) |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 1420 | { |
| 1421 | if (bio_integrity(bio)) |
| 1422 | bio_integrity_advance(bio, bytes); |
| 1423 | |
Satya Tangirala | a892c8d | 2020-05-14 00:37:18 +0000 | [diff] [blame] | 1424 | bio_crypt_advance(bio, bytes); |
Kent Overstreet | 4550dd6 | 2013-08-07 14:26:21 -0700 | [diff] [blame] | 1425 | bio_advance_iter(bio, &bio->bi_iter, bytes); |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 1426 | } |
Jens Axboe | d4aa57a | 2021-10-13 09:01:43 -0600 | [diff] [blame] | 1427 | EXPORT_SYMBOL(__bio_advance); |
Kent Overstreet | 054bdf6 | 2012-09-28 13:17:55 -0700 | [diff] [blame] | 1428 | |
Jens Axboe | ee4b4e2 | 2023-01-04 14:43:27 -0700 | [diff] [blame] | 1429 | void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter, |
| 1430 | struct bio *src, struct bvec_iter *src_iter) |
| 1431 | { |
| 1432 | while (src_iter->bi_size && dst_iter->bi_size) { |
| 1433 | struct bio_vec src_bv = bio_iter_iovec(src, *src_iter); |
| 1434 | struct bio_vec dst_bv = bio_iter_iovec(dst, *dst_iter); |
| 1435 | unsigned int bytes = min(src_bv.bv_len, dst_bv.bv_len); |
| 1436 | void *src_buf = bvec_kmap_local(&src_bv); |
| 1437 | void *dst_buf = bvec_kmap_local(&dst_bv); |
| 1438 | |
| 1439 | memcpy(dst_buf, src_buf, bytes); |
| 1440 | |
| 1441 | kunmap_local(dst_buf); |
| 1442 | kunmap_local(src_buf); |
| 1443 | |
| 1444 | bio_advance_iter_single(src, src_iter, bytes); |
| 1445 | bio_advance_iter_single(dst, dst_iter, bytes); |
| 1446 | } |
| 1447 | } |
| 1448 | EXPORT_SYMBOL(bio_copy_data_iter); |
| 1449 | |
Kent Overstreet | 45db54d | 2018-05-08 21:33:54 -0400 | [diff] [blame] | 1450 | /** |
| 1451 | * bio_copy_data - copy contents of data buffers from one bio to another |
| 1452 | * @src: source bio |
| 1453 | * @dst: destination bio |
| 1454 | * |
| 1455 | * Stops when it reaches the end of either @src or @dst - that is, copies |
| 1456 | * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios). |
| 1457 | */ |
| 1458 | void bio_copy_data(struct bio *dst, struct bio *src) |
| 1459 | { |
| 1460 | struct bvec_iter src_iter = src->bi_iter; |
| 1461 | struct bvec_iter dst_iter = dst->bi_iter; |
| 1462 | |
Jens Axboe | ee4b4e2 | 2023-01-04 14:43:27 -0700 | [diff] [blame] | 1463 | bio_copy_data_iter(dst, &dst_iter, src, &src_iter); |
Kent Overstreet | 45db54d | 2018-05-08 21:33:54 -0400 | [diff] [blame] | 1464 | } |
| 1465 | EXPORT_SYMBOL(bio_copy_data); |
| 1466 | |
Guoqing Jiang | 491221f | 2016-09-22 03:10:01 -0400 | [diff] [blame] | 1467 | void bio_free_pages(struct bio *bio) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1468 | { |
| 1469 | struct bio_vec *bvec; |
Ming Lei | 6dc4f10 | 2019-02-15 19:13:19 +0800 | [diff] [blame] | 1470 | struct bvec_iter_all iter_all; |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1471 | |
Christoph Hellwig | 2b070cf | 2019-04-25 09:03:00 +0200 | [diff] [blame] | 1472 | bio_for_each_segment_all(bvec, bio, iter_all) |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1473 | __free_page(bvec->bv_page); |
| 1474 | } |
Guoqing Jiang | 491221f | 2016-09-22 03:10:01 -0400 | [diff] [blame] | 1475 | EXPORT_SYMBOL(bio_free_pages); |
Christoph Hellwig | 1dfa0f6 | 2015-01-18 16:16:30 +0100 | [diff] [blame] | 1476 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1477 | /* |
| 1478 | * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions |
| 1479 | * for performing direct-IO in BIOs. |
| 1480 | * |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1481 | * The problem is that we cannot run folio_mark_dirty() from interrupt context |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1482 | * because the required locks are not interrupt-safe. So what we can do is to |
| 1483 | * mark the pages dirty _before_ performing IO. And in interrupt context, |
| 1484 | * check that the pages are still dirty. If so, fine. If not, redirty them |
| 1485 | * in process context. |
| 1486 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1487 | * Note that this code is very hard to test under normal circumstances because |
| 1488 | * direct-io pins the pages with get_user_pages(). This makes |
| 1489 | * 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] | 1490 | * 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] | 1491 | * pagecache. |
| 1492 | * |
| 1493 | * Simply disabling the call to bio_set_pages_dirty() is a good way to test the |
| 1494 | * deferred bio dirtying paths. |
| 1495 | */ |
| 1496 | |
| 1497 | /* |
| 1498 | * bio_set_pages_dirty() will mark all the bio's pages as dirty. |
| 1499 | */ |
| 1500 | void bio_set_pages_dirty(struct bio *bio) |
| 1501 | { |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1502 | struct folio_iter fi; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1503 | |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1504 | bio_for_each_folio_all(fi, bio) { |
| 1505 | folio_lock(fi.folio); |
| 1506 | folio_mark_dirty(fi.folio); |
| 1507 | folio_unlock(fi.folio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1508 | } |
| 1509 | } |
Kent Overstreet | 7ba3792 | 2023-08-13 14:26:34 -0400 | [diff] [blame] | 1510 | EXPORT_SYMBOL_GPL(bio_set_pages_dirty); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1511 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1512 | /* |
| 1513 | * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. |
| 1514 | * If they are, then fine. If, however, some pages are clean then they must |
| 1515 | * 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] | 1516 | * the BIO and re-dirty the pages in process context. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1517 | * |
| 1518 | * It is expected that bio_check_pages_dirty() will wholly own the BIO from |
David Howells | fd36324 | 2023-05-22 21:57:42 +0100 | [diff] [blame] | 1519 | * here on. It will unpin each page and will run one bio_put() against the |
| 1520 | * BIO. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1521 | */ |
| 1522 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1523 | static void bio_dirty_fn(struct work_struct *work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1524 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1525 | static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1526 | static DEFINE_SPINLOCK(bio_dirty_lock); |
| 1527 | static struct bio *bio_dirty_list; |
| 1528 | |
| 1529 | /* |
| 1530 | * This runs in process context |
| 1531 | */ |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 1532 | static void bio_dirty_fn(struct work_struct *work) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1533 | { |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1534 | struct bio *bio, *next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1535 | |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1536 | spin_lock_irq(&bio_dirty_lock); |
| 1537 | next = bio_dirty_list; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1538 | bio_dirty_list = NULL; |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1539 | spin_unlock_irq(&bio_dirty_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1540 | |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1541 | while ((bio = next) != NULL) { |
| 1542 | next = bio->bi_private; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1543 | |
Christoph Hellwig | d241a95 | 2019-06-26 15:49:21 +0200 | [diff] [blame] | 1544 | bio_release_pages(bio, true); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1545 | bio_put(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1546 | } |
| 1547 | } |
| 1548 | |
| 1549 | void bio_check_pages_dirty(struct bio *bio) |
| 1550 | { |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1551 | struct folio_iter fi; |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1552 | unsigned long flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1553 | |
Matthew Wilcox (Oracle) | 1b151e2 | 2023-08-14 15:41:00 +0100 | [diff] [blame] | 1554 | bio_for_each_folio_all(fi, bio) { |
| 1555 | if (!folio_test_dirty(fi.folio)) |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1556 | goto defer; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1557 | } |
| 1558 | |
Christoph Hellwig | d241a95 | 2019-06-26 15:49:21 +0200 | [diff] [blame] | 1559 | bio_release_pages(bio, false); |
Christoph Hellwig | 24d5493 | 2018-07-24 14:04:12 +0200 | [diff] [blame] | 1560 | bio_put(bio); |
| 1561 | return; |
| 1562 | defer: |
| 1563 | spin_lock_irqsave(&bio_dirty_lock, flags); |
| 1564 | bio->bi_private = bio_dirty_list; |
| 1565 | bio_dirty_list = bio; |
| 1566 | spin_unlock_irqrestore(&bio_dirty_lock, flags); |
| 1567 | schedule_work(&bio_dirty_work); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1568 | } |
Kent Overstreet | 7ba3792 | 2023-08-13 14:26:34 -0400 | [diff] [blame] | 1569 | EXPORT_SYMBOL_GPL(bio_check_pages_dirty); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1570 | |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1571 | static inline bool bio_remaining_done(struct bio *bio) |
| 1572 | { |
| 1573 | /* |
| 1574 | * If we're not chaining, then ->__bi_remaining is always 1 and |
| 1575 | * we always end io on the first invocation. |
| 1576 | */ |
| 1577 | if (!bio_flagged(bio, BIO_CHAIN)) |
| 1578 | return true; |
| 1579 | |
| 1580 | BUG_ON(atomic_read(&bio->__bi_remaining) <= 0); |
| 1581 | |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1582 | if (atomic_dec_and_test(&bio->__bi_remaining)) { |
Jens Axboe | b7c44ed | 2015-07-24 12:37:59 -0600 | [diff] [blame] | 1583 | bio_clear_flag(bio, BIO_CHAIN); |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1584 | return true; |
Mike Snitzer | 326e1db | 2015-05-22 09:14:03 -0400 | [diff] [blame] | 1585 | } |
Jens Axboe | c4cf526 | 2015-04-17 16:15:18 -0600 | [diff] [blame] | 1586 | |
| 1587 | return false; |
| 1588 | } |
| 1589 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1590 | /** |
| 1591 | * bio_endio - end I/O on a bio |
| 1592 | * @bio: bio |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1593 | * |
| 1594 | * Description: |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1595 | * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred |
| 1596 | * way to end I/O on a bio. No one should call bi_end_io() directly on a |
| 1597 | * 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] | 1598 | * |
| 1599 | * bio_endio() can be called several times on a bio that has been chained |
| 1600 | * using bio_chain(). The ->bi_end_io() function will only be called the |
Edward Hsieh | 60b6a7e | 2021-06-24 20:30:30 +0800 | [diff] [blame] | 1601 | * last time. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1602 | **/ |
Christoph Hellwig | 4246a0b | 2015-07-20 15:29:37 +0200 | [diff] [blame] | 1603 | void bio_endio(struct bio *bio) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1604 | { |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1605 | again: |
Christoph Hellwig | 2b88551 | 2016-03-11 17:34:53 +0100 | [diff] [blame] | 1606 | if (!bio_remaining_done(bio)) |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1607 | return; |
Christoph Hellwig | 7c20f11 | 2017-07-03 16:58:43 -0600 | [diff] [blame] | 1608 | if (!bio_integrity_endio(bio)) |
| 1609 | return; |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1610 | |
Damien Le Moal | dd291d7 | 2024-04-08 10:41:07 +0900 | [diff] [blame] | 1611 | blk_zone_bio_endio(bio); |
| 1612 | |
Tejun Heo | aa1b46d | 2022-03-13 21:15:02 -1000 | [diff] [blame] | 1613 | rq_qos_done_bio(bio); |
Josef Bacik | 67b42d0 | 2018-07-03 11:15:00 -0400 | [diff] [blame] | 1614 | |
Edward Hsieh | 60b6a7e | 2021-06-24 20:30:30 +0800 | [diff] [blame] | 1615 | if (bio->bi_bdev && bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
Pavel Begunkov | 3caee463 | 2021-10-14 15:03:28 +0100 | [diff] [blame] | 1616 | trace_block_bio_complete(bdev_get_queue(bio->bi_bdev), bio); |
Edward Hsieh | 60b6a7e | 2021-06-24 20:30:30 +0800 | [diff] [blame] | 1617 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); |
| 1618 | } |
| 1619 | |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1620 | /* |
| 1621 | * Need to have a real endio function for chained bios, otherwise |
| 1622 | * various corner cases will break (like stacking block devices that |
| 1623 | * save/restore bi_end_io) - however, we want to avoid unbounded |
| 1624 | * recursion and blowing the stack. Tail call optimization would |
| 1625 | * handle this, but compiling with frame pointers also disables |
| 1626 | * gcc's sibling call optimization. |
| 1627 | */ |
| 1628 | if (bio->bi_end_io == bio_chain_endio) { |
| 1629 | bio = __bio_chain_endio(bio); |
| 1630 | goto again; |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1631 | } |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1632 | |
Shaohua Li | b222dd2 | 2017-07-10 11:40:17 -0700 | [diff] [blame] | 1633 | /* release cgroup info */ |
| 1634 | bio_uninit(bio); |
Christoph Hellwig | ba8c696 | 2016-03-11 17:34:52 +0100 | [diff] [blame] | 1635 | if (bio->bi_end_io) |
| 1636 | bio->bi_end_io(bio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1637 | } |
H Hartley Sweeten | a112a71 | 2009-09-26 16:19:21 +0200 | [diff] [blame] | 1638 | EXPORT_SYMBOL(bio_endio); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1639 | |
Kent Overstreet | 196d38bc | 2013-11-23 18:34:15 -0800 | [diff] [blame] | 1640 | /** |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1641 | * bio_split - split a bio |
| 1642 | * @bio: bio to split |
| 1643 | * @sectors: number of sectors to split from the front of @bio |
| 1644 | * @gfp: gfp mask |
| 1645 | * @bs: bio set to allocate from |
| 1646 | * |
| 1647 | * Allocates and returns a new bio which represents @sectors from the start of |
| 1648 | * @bio, and updates @bio to represent the remaining sectors. |
| 1649 | * |
Martin K. Petersen | f3f5da6 | 2015-07-22 07:57:12 -0400 | [diff] [blame] | 1650 | * Unless this is a discard request the newly allocated bio will point |
Bart Van Assche | dad7758 | 2019-08-01 15:50:41 -0700 | [diff] [blame] | 1651 | * to @bio's bi_io_vec. It is the caller's responsibility to ensure that |
| 1652 | * neither @bio nor @bs are freed before the split bio. |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1653 | */ |
| 1654 | struct bio *bio_split(struct bio *bio, int sectors, |
| 1655 | gfp_t gfp, struct bio_set *bs) |
| 1656 | { |
Mikulas Patocka | f341a4d | 2017-11-22 13:18:05 -0500 | [diff] [blame] | 1657 | struct bio *split; |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1658 | |
| 1659 | BUG_ON(sectors <= 0); |
| 1660 | BUG_ON(sectors >= bio_sectors(bio)); |
| 1661 | |
Keith Busch | 0512a75 | 2020-05-12 17:55:47 +0900 | [diff] [blame] | 1662 | /* Zone append commands cannot be split */ |
| 1663 | if (WARN_ON_ONCE(bio_op(bio) == REQ_OP_ZONE_APPEND)) |
| 1664 | return NULL; |
| 1665 | |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 1666 | split = bio_alloc_clone(bio->bi_bdev, bio, gfp, bs); |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1667 | if (!split) |
| 1668 | return NULL; |
| 1669 | |
| 1670 | split->bi_iter.bi_size = sectors << 9; |
| 1671 | |
| 1672 | if (bio_integrity(split)) |
Dmitry Monakhov | fbd08e7 | 2017-06-29 11:31:10 -0700 | [diff] [blame] | 1673 | bio_integrity_trim(split); |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1674 | |
| 1675 | bio_advance(bio, split->bi_iter.bi_size); |
| 1676 | |
NeilBrown | fbbaf70 | 2017-04-07 09:40:52 -0600 | [diff] [blame] | 1677 | if (bio_flagged(bio, BIO_TRACE_COMPLETION)) |
Goldwyn Rodrigues | 20d5902 | 2018-01-23 09:10:19 -0700 | [diff] [blame] | 1678 | bio_set_flag(split, BIO_TRACE_COMPLETION); |
NeilBrown | fbbaf70 | 2017-04-07 09:40:52 -0600 | [diff] [blame] | 1679 | |
Kent Overstreet | 20d0189 | 2013-11-23 18:21:01 -0800 | [diff] [blame] | 1680 | return split; |
| 1681 | } |
| 1682 | EXPORT_SYMBOL(bio_split); |
| 1683 | |
Martin K. Petersen | ad3316b | 2008-10-01 22:42:53 -0400 | [diff] [blame] | 1684 | /** |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1685 | * bio_trim - trim a bio |
| 1686 | * @bio: bio to trim |
| 1687 | * @offset: number of sectors to trim from the front of @bio |
| 1688 | * @size: size we want to trim @bio to, in sectors |
Chaitanya Kulkarni | e83502c | 2021-07-21 21:43:32 +0900 | [diff] [blame] | 1689 | * |
| 1690 | * This function is typically used for bios that are cloned and submitted |
| 1691 | * to the underlying device in parts. |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1692 | */ |
Chaitanya Kulkarni | e83502c | 2021-07-21 21:43:32 +0900 | [diff] [blame] | 1693 | void bio_trim(struct bio *bio, sector_t offset, sector_t size) |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1694 | { |
Chaitanya Kulkarni | e83502c | 2021-07-21 21:43:32 +0900 | [diff] [blame] | 1695 | if (WARN_ON_ONCE(offset > BIO_MAX_SECTORS || size > BIO_MAX_SECTORS || |
Ming Lei | 8535c01 | 2022-04-14 16:44:43 +0800 | [diff] [blame] | 1696 | offset + size > bio_sectors(bio))) |
Chaitanya Kulkarni | e83502c | 2021-07-21 21:43:32 +0900 | [diff] [blame] | 1697 | return; |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1698 | |
| 1699 | size <<= 9; |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1700 | if (offset == 0 && size == bio->bi_iter.bi_size) |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1701 | return; |
| 1702 | |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1703 | bio_advance(bio, offset << 9); |
Kent Overstreet | 4f024f3 | 2013-10-11 15:44:27 -0700 | [diff] [blame] | 1704 | bio->bi_iter.bi_size = size; |
Dmitry Monakhov | 376a78a | 2017-06-29 11:31:08 -0700 | [diff] [blame] | 1705 | |
| 1706 | if (bio_integrity(bio)) |
Dmitry Monakhov | fbd08e7 | 2017-06-29 11:31:10 -0700 | [diff] [blame] | 1707 | bio_integrity_trim(bio); |
Kent Overstreet | 6678d83 | 2013-08-07 11:14:32 -0700 | [diff] [blame] | 1708 | } |
| 1709 | EXPORT_SYMBOL_GPL(bio_trim); |
| 1710 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1711 | /* |
| 1712 | * create memory pools for biovec's in a bio_set. |
| 1713 | * use the global biovec slabs created for general use. |
| 1714 | */ |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 1715 | int biovec_init_pool(mempool_t *pool, int pool_entries) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1716 | { |
Christoph Hellwig | 7a800a2 | 2021-02-02 18:19:29 +0100 | [diff] [blame] | 1717 | struct biovec_slab *bp = bvec_slabs + ARRAY_SIZE(bvec_slabs) - 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1718 | |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 1719 | return mempool_init_slab_pool(pool, pool_entries, bp->slab); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1720 | } |
| 1721 | |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1722 | /* |
| 1723 | * bioset_exit - exit a bioset initialized with bioset_init() |
| 1724 | * |
| 1725 | * May be called on a zeroed but uninitialized bioset (i.e. allocated with |
| 1726 | * kzalloc()). |
| 1727 | */ |
| 1728 | void bioset_exit(struct bio_set *bs) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1729 | { |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 1730 | bio_alloc_cache_destroy(bs); |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1731 | if (bs->rescue_workqueue) |
| 1732 | destroy_workqueue(bs->rescue_workqueue); |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1733 | bs->rescue_workqueue = NULL; |
Kent Overstreet | df2cb6d | 2012-09-10 14:33:46 -0700 | [diff] [blame] | 1734 | |
Kent Overstreet | 8aa6ba2 | 2018-05-08 21:33:50 -0400 | [diff] [blame] | 1735 | mempool_exit(&bs->bio_pool); |
| 1736 | mempool_exit(&bs->bvec_pool); |
Kent Overstreet | 9f060e2 | 2012-10-12 15:29:33 -0700 | [diff] [blame] | 1737 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 1738 | bioset_integrity_free(bs); |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1739 | if (bs->bio_slab) |
| 1740 | bio_put_slab(bs); |
| 1741 | bs->bio_slab = NULL; |
| 1742 | } |
| 1743 | EXPORT_SYMBOL(bioset_exit); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1744 | |
NeilBrown | 011067b | 2017-06-18 14:38:57 +1000 | [diff] [blame] | 1745 | /** |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1746 | * bioset_init - Initialize a bio_set |
Kent Overstreet | dad0852 | 2018-05-20 18:25:58 -0400 | [diff] [blame] | 1747 | * @bs: pool to initialize |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1748 | * @pool_size: Number of bio and bio_vecs to cache in the mempool |
| 1749 | * @front_pad: Number of bytes to allocate in front of the returned bio |
| 1750 | * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS |
| 1751 | * and %BIOSET_NEED_RESCUER |
| 1752 | * |
Kent Overstreet | dad0852 | 2018-05-20 18:25:58 -0400 | [diff] [blame] | 1753 | * Description: |
| 1754 | * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller |
| 1755 | * to ask for a number of bytes to be allocated in front of the bio. |
| 1756 | * Front pad allocation is useful for embedding the bio inside |
| 1757 | * another structure, to avoid allocating extra data to go with the bio. |
| 1758 | * Note that the bio must be embedded at the END of that structure always, |
| 1759 | * or things will break badly. |
| 1760 | * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated |
Christoph Hellwig | abfc426 | 2022-02-02 17:01:09 +0100 | [diff] [blame] | 1761 | * for allocating iovecs. This pool is not needed e.g. for bio_init_clone(). |
| 1762 | * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used |
| 1763 | * to dispatch queued requests when the mempool runs out of space. |
Kent Overstreet | dad0852 | 2018-05-20 18:25:58 -0400 | [diff] [blame] | 1764 | * |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1765 | */ |
| 1766 | int bioset_init(struct bio_set *bs, |
| 1767 | unsigned int pool_size, |
| 1768 | unsigned int front_pad, |
| 1769 | int flags) |
| 1770 | { |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1771 | bs->front_pad = front_pad; |
Ming Lei | 9f180e3 | 2021-01-11 11:05:54 +0800 | [diff] [blame] | 1772 | if (flags & BIOSET_NEED_BVECS) |
| 1773 | bs->back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); |
| 1774 | else |
| 1775 | bs->back_pad = 0; |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1776 | |
| 1777 | spin_lock_init(&bs->rescue_lock); |
| 1778 | bio_list_init(&bs->rescue_list); |
| 1779 | INIT_WORK(&bs->rescue_work, bio_alloc_rescue); |
| 1780 | |
Ming Lei | 49d1ec8 | 2021-01-11 11:05:52 +0800 | [diff] [blame] | 1781 | bs->bio_slab = bio_find_or_create_slab(bs); |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1782 | if (!bs->bio_slab) |
| 1783 | return -ENOMEM; |
| 1784 | |
| 1785 | if (mempool_init_slab_pool(&bs->bio_pool, pool_size, bs->bio_slab)) |
| 1786 | goto bad; |
| 1787 | |
| 1788 | if ((flags & BIOSET_NEED_BVECS) && |
| 1789 | biovec_init_pool(&bs->bvec_pool, pool_size)) |
| 1790 | goto bad; |
| 1791 | |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 1792 | if (flags & BIOSET_NEED_RESCUER) { |
| 1793 | bs->rescue_workqueue = alloc_workqueue("bioset", |
| 1794 | WQ_MEM_RECLAIM, 0); |
| 1795 | if (!bs->rescue_workqueue) |
| 1796 | goto bad; |
| 1797 | } |
| 1798 | if (flags & BIOSET_PERCPU_CACHE) { |
| 1799 | bs->cache = alloc_percpu(struct bio_alloc_cache); |
| 1800 | if (!bs->cache) |
| 1801 | goto bad; |
| 1802 | cpuhp_state_add_instance_nocalls(CPUHP_BIO_DEAD, &bs->cpuhp_dead); |
| 1803 | } |
Kent Overstreet | 917a38c | 2018-05-08 21:33:51 -0400 | [diff] [blame] | 1804 | |
| 1805 | return 0; |
| 1806 | bad: |
| 1807 | bioset_exit(bs); |
| 1808 | return -ENOMEM; |
| 1809 | } |
| 1810 | EXPORT_SYMBOL(bioset_init); |
| 1811 | |
Christoph Hellwig | de76fd8 | 2021-02-02 18:19:23 +0100 | [diff] [blame] | 1812 | static int __init init_bio(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1813 | { |
| 1814 | int i; |
| 1815 | |
Jens Axboe | a3df2e4 | 2023-01-09 10:45:56 -0700 | [diff] [blame] | 1816 | BUILD_BUG_ON(BIO_FLAG_LAST > 8 * sizeof_field(struct bio, bi_flags)); |
| 1817 | |
Martin K. Petersen | 7878cba | 2009-06-26 15:37:49 +0200 | [diff] [blame] | 1818 | bio_integrity_init(); |
Christoph Hellwig | de76fd8 | 2021-02-02 18:19:23 +0100 | [diff] [blame] | 1819 | |
| 1820 | for (i = 0; i < ARRAY_SIZE(bvec_slabs); i++) { |
Jens Axboe | a7fcd37 | 2008-12-05 16:10:29 +0100 | [diff] [blame] | 1821 | struct biovec_slab *bvs = bvec_slabs + i; |
| 1822 | |
Christoph Hellwig | de76fd8 | 2021-02-02 18:19:23 +0100 | [diff] [blame] | 1823 | bvs->slab = kmem_cache_create(bvs->name, |
| 1824 | bvs->nr_vecs * sizeof(struct bio_vec), 0, |
| 1825 | SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1826 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1827 | |
Jens Axboe | be4d234 | 2021-03-08 11:37:47 -0700 | [diff] [blame] | 1828 | cpuhp_setup_state_multi(CPUHP_BIO_DEAD, "block/bio:dead", NULL, |
| 1829 | bio_cpu_dead); |
| 1830 | |
Jens Axboe | 12c5b70 | 2022-09-02 10:40:11 -0600 | [diff] [blame] | 1831 | if (bioset_init(&fs_bio_set, BIO_POOL_SIZE, 0, |
| 1832 | BIOSET_NEED_BVECS | BIOSET_PERCPU_CACHE)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1833 | panic("bio: can't allocate bios\n"); |
| 1834 | |
Kent Overstreet | f4f8154 | 2018-05-08 21:33:52 -0400 | [diff] [blame] | 1835 | if (bioset_integrity_create(&fs_bio_set, BIO_POOL_SIZE)) |
Martin K. Petersen | a91a278 | 2011-03-17 11:11:05 +0100 | [diff] [blame] | 1836 | panic("bio: can't create integrity pool\n"); |
| 1837 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1838 | return 0; |
| 1839 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1840 | subsys_initcall(init_bio); |