]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - block/bounce.c
nfc: llcp: fix nfc_llcp_setsockopt() unsafe copies
[thirdparty/kernel/linux.git] / block / bounce.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
831058de
DH
2/* bounce buffer handling for block devices
3 *
4 * - Split from highmem.c
5 */
6
b1de0d13
MH
7#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
8
831058de 9#include <linux/mm.h>
b95f1b31 10#include <linux/export.h>
831058de 11#include <linux/swap.h>
5a0e3ad6 12#include <linux/gfp.h>
831058de
DH
13#include <linux/bio.h>
14#include <linux/pagemap.h>
15#include <linux/mempool.h>
16#include <linux/blkdev.h>
66114cad 17#include <linux/backing-dev.h>
831058de
DH
18#include <linux/init.h>
19#include <linux/hash.h>
20#include <linux/highmem.h>
b1de0d13 21#include <linux/printk.h>
831058de
DH
22#include <asm/tlbflush.h>
23
55782138 24#include <trace/events/block.h>
3bce016a 25#include "blk.h"
672fdcf0 26#include "blk-cgroup.h"
55782138 27
831058de
DH
28#define POOL_SIZE 64
29#define ISA_POOL_SIZE 16
30
338aa96d 31static struct bio_set bounce_bio_set, bounce_bio_split;
ce288e05 32static mempool_t page_pool;
831058de 33
52990a5f
JA
34static void init_bounce_bioset(void)
35{
36 static bool bounce_bs_setup;
37 int ret;
38
39 if (bounce_bs_setup)
40 return;
41
42 ret = bioset_init(&bounce_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
43 BUG_ON(ret);
44 if (bioset_integrity_create(&bounce_bio_set, BIO_POOL_SIZE))
45 BUG_ON(1);
46
47 ret = bioset_init(&bounce_bio_split, BIO_POOL_SIZE, 0, 0);
48 BUG_ON(ret);
49 bounce_bs_setup = true;
50}
51
831058de
DH
52static __init int init_emergency_pool(void)
53{
338aa96d 54 int ret;
9bb33f24
CH
55
56#ifndef CONFIG_MEMORY_HOTPLUG
3bcfeaf9 57 if (max_pfn <= max_low_pfn)
831058de 58 return 0;
3bcfeaf9 59#endif
831058de 60
338aa96d
KO
61 ret = mempool_init_page_pool(&page_pool, POOL_SIZE, 0);
62 BUG_ON(ret);
b1de0d13 63 pr_info("pool size: %d pages\n", POOL_SIZE);
831058de 64
52990a5f 65 init_bounce_bioset();
831058de
DH
66 return 0;
67}
68
69__initcall(init_emergency_pool);
70
831058de
DH
71/*
72 * Simple bounce buffer support for highmem pages. Depending on the
73 * queue gfp mask set, *to may or may not be a highmem page. kmap it
74 * always, it will do the Right Thing
75 */
76static void copy_to_high_bio_irq(struct bio *to, struct bio *from)
77{
3c892a09 78 struct bio_vec tovec, fromvec;
7988613b 79 struct bvec_iter iter;
3c892a09
ML
80 /*
81 * The bio of @from is created by bounce, so we can iterate
82 * its bvec from start to end, but the @from->bi_iter can't be
83 * trusted because it might be changed by splitting.
84 */
85 struct bvec_iter from_iter = BVEC_ITER_ALL_INIT;
7988613b
KO
86
87 bio_for_each_segment(tovec, to, iter) {
3c892a09
ML
88 fromvec = bio_iter_iovec(from, from_iter);
89 if (tovec.bv_page != fromvec.bv_page) {
7988613b
KO
90 /*
91 * fromvec->bv_offset and fromvec->bv_len might have
92 * been modified by the block layer, so use the original
93 * copy, bounce_copy_vec already uses tovec->bv_len
94 */
f434cdc7
CH
95 memcpy_to_bvec(&tovec, page_address(fromvec.bv_page) +
96 tovec.bv_offset);
7988613b 97 }
3c892a09 98 bio_advance_iter(from, &from_iter, tovec.bv_len);
831058de
DH
99 }
100}
101
ce288e05 102static void bounce_end_io(struct bio *bio)
831058de
DH
103{
104 struct bio *bio_orig = bio->bi_private;
7891f05c 105 struct bio_vec *bvec, orig_vec;
7891f05c 106 struct bvec_iter orig_iter = bio_orig->bi_iter;
6dc4f100 107 struct bvec_iter_all iter_all;
831058de 108
831058de
DH
109 /*
110 * free up bounce indirect pages used
111 */
2b070cfe 112 bio_for_each_segment_all(bvec, bio, iter_all) {
7891f05c
ML
113 orig_vec = bio_iter_iovec(bio_orig, orig_iter);
114 if (bvec->bv_page != orig_vec.bv_page) {
115 dec_zone_page_state(bvec->bv_page, NR_BOUNCE);
ce288e05 116 mempool_free(bvec->bv_page, &page_pool);
7891f05c
ML
117 }
118 bio_advance_iter(bio_orig, &orig_iter, orig_vec.bv_len);
831058de
DH
119 }
120
4e4cbee9 121 bio_orig->bi_status = bio->bi_status;
4246a0b6 122 bio_endio(bio_orig);
831058de
DH
123 bio_put(bio);
124}
125
4246a0b6 126static void bounce_end_io_write(struct bio *bio)
831058de 127{
ce288e05 128 bounce_end_io(bio);
831058de
DH
129}
130
ce288e05 131static void bounce_end_io_read(struct bio *bio)
831058de
DH
132{
133 struct bio *bio_orig = bio->bi_private;
134
4e4cbee9 135 if (!bio->bi_status)
831058de
DH
136 copy_to_high_bio_irq(bio_orig, bio);
137
ce288e05 138 bounce_end_io(bio);
831058de
DH
139}
140
ebfe4183 141static struct bio *bounce_clone_bio(struct bio *bio_src)
c55183c9
CH
142{
143 struct bvec_iter iter;
144 struct bio_vec bv;
145 struct bio *bio;
146
147 /*
148 * Pre immutable biovecs, __bio_clone() used to just do a memcpy from
149 * bio_src->bi_io_vec to bio->bi_io_vec.
150 *
151 * We can't do that anymore, because:
152 *
153 * - The point of cloning the biovec is to produce a bio with a biovec
154 * the caller can modify: bi_idx and bi_bvec_done should be 0.
155 *
a8affc03 156 * - The original bio could've had more than BIO_MAX_VECS biovecs; if
c55183c9
CH
157 * we tried to clone the whole thing bio_alloc_bioset() would fail.
158 * But the clone should succeed as long as the number of biovecs we
a8affc03 159 * actually need to allocate is fewer than BIO_MAX_VECS.
c55183c9
CH
160 *
161 * - Lastly, bi_vcnt should not be looked at or relied upon by code
162 * that does not own the bio - reason being drivers don't use it for
163 * iterating over the biovec anymore, so expecting it to be kept up
164 * to date (i.e. for clones that share the parent biovec) is just
abfc426d 165 * asking for trouble and would force extra work.
c55183c9 166 */
609be106
CH
167 bio = bio_alloc_bioset(bio_src->bi_bdev, bio_segments(bio_src),
168 bio_src->bi_opf, GFP_NOIO, &bounce_bio_set);
46bbf653
CH
169 if (bio_flagged(bio_src, BIO_REMAPPED))
170 bio_set_flag(bio, BIO_REMAPPED);
ca474b73 171 bio->bi_ioprio = bio_src->bi_ioprio;
44981351 172 bio->bi_write_hint = bio_src->bi_write_hint;
c55183c9
CH
173 bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector;
174 bio->bi_iter.bi_size = bio_src->bi_iter.bi_size;
175
176 switch (bio_op(bio)) {
177 case REQ_OP_DISCARD:
178 case REQ_OP_SECURE_ERASE:
179 case REQ_OP_WRITE_ZEROES:
180 break;
c55183c9
CH
181 default:
182 bio_for_each_segment(bv, bio_src, iter)
183 bio->bi_io_vec[bio->bi_vcnt++] = bv;
184 break;
185 }
186
ebfe4183 187 if (bio_crypt_clone(bio, bio_src, GFP_NOIO) < 0)
07560151 188 goto err_put;
a892c8d5 189
07560151 190 if (bio_integrity(bio_src) &&
ebfe4183 191 bio_integrity_clone(bio, bio_src, GFP_NOIO) < 0)
07560151 192 goto err_put;
c55183c9 193
db6638d7 194 bio_clone_blkg_association(bio, bio_src);
5bf9a1f3 195
c55183c9 196 return bio;
07560151
EB
197
198err_put:
199 bio_put(bio);
200 return NULL;
c55183c9
CH
201}
202
51d798cd 203struct bio *__blk_queue_bounce(struct bio *bio_orig, struct request_queue *q)
831058de 204{
6bc454d1 205 struct bio *bio;
51d798cd 206 int rw = bio_data_dir(bio_orig);
7988613b
KO
207 struct bio_vec *to, from;
208 struct bvec_iter iter;
9cfe3dde 209 unsigned i = 0, bytes = 0;
a8821f3f 210 bool bounce = false;
9cfe3dde 211 int sectors;
831058de 212
51d798cd 213 bio_for_each_segment(from, bio_orig, iter) {
a8affc03 214 if (i++ < BIO_MAX_VECS)
9cfe3dde 215 bytes += from.bv_len;
9bb33f24 216 if (PageHighMem(from.bv_page))
a8821f3f
N
217 bounce = true;
218 }
219 if (!bounce)
51d798cd 220 return bio_orig;
a8821f3f 221
9cfe3dde
KB
222 /*
223 * Individual bvecs might not be logical block aligned. Round down
224 * the split size so that each bio is properly block size aligned,
225 * even if we do not use the full hardware limits.
226 */
227 sectors = ALIGN_DOWN(bytes, queue_logical_block_size(q)) >>
228 SECTOR_SHIFT;
51d798cd
CH
229 if (sectors < bio_sectors(bio_orig)) {
230 bio = bio_split(bio_orig, sectors, GFP_NOIO, &bounce_bio_split);
231 bio_chain(bio, bio_orig);
232 submit_bio_noacct(bio_orig);
233 bio_orig = bio;
a8821f3f 234 }
51d798cd 235 bio = bounce_clone_bio(bio_orig);
831058de 236
8f4e80da
ML
237 /*
238 * Bvec table can't be updated by bio_for_each_segment_all(),
239 * so retrieve bvec from the table directly. This way is safe
240 * because the 'bio' is single-page bvec.
241 */
242 for (i = 0, to = bio->bi_io_vec; i < bio->bi_vcnt; to++, i++) {
4aebe859 243 struct page *bounce_page;
f735b5ee 244
4aebe859 245 if (!PageHighMem(to->bv_page))
6bc454d1 246 continue;
831058de 247
4aebe859
CH
248 bounce_page = mempool_alloc(&page_pool, GFP_NOIO);
249 inc_zone_page_state(bounce_page, NR_BOUNCE);
831058de
DH
250
251 if (rw == WRITE) {
4aebe859
CH
252 flush_dcache_page(to->bv_page);
253 memcpy_from_bvec(page_address(bounce_page), to);
831058de 254 }
4aebe859 255 to->bv_page = bounce_page;
831058de
DH
256 }
257
51d798cd 258 trace_block_bio_bounce(bio_orig);
c43a5082 259
831058de 260 bio->bi_flags |= (1 << BIO_BOUNCED);
831058de 261
ce288e05
CH
262 if (rw == READ)
263 bio->bi_end_io = bounce_end_io_read;
264 else
831058de 265 bio->bi_end_io = bounce_end_io_write;
831058de 266
51d798cd
CH
267 bio->bi_private = bio_orig;
268 return bio;
831058de 269}