Eric Lee / smarc-fsl-linux-kernel

Download as
Browse Code ยป

Commit 1d6165851cd8e3f919d446cd6da35dee44e8837e

Authored by Dmitry Monakhov
Committed by Jens Axboe
1 parent c84a301d4a
Exists in master and in 41 other branches 8mp-imx_5.4.70_2.3.0, 8qm-imx_5.4.70_2.3.0, emb-lf-6.6.52-2.2.0, emb_imx_lf-5.15.y, emb_lf-6.1.y, emb_lf-6.6.y, imx_3.0.35_4.1.0, imx_3.10.17_1.0.1_ga, imx_3.10.53_1.1.0_ga, imx_3.14.28_1.0.0_ga, imx_4.1.15_1.0.0_ga, pitx_8mp_lf-5.10.y, rt-smarc-imx_4.1.15_1.0.0_ga, rt_linux_5.15.71, smarc-8m-android-11.0.0_2.0.0, smarc-imx6_4.14.98_2.0.0_ga, smarc-imx6_4.9.88_2.0.0_ga, smarc-imx7_4.14.98_2.0.0_ga, smarc-imx7_4.9.11_1.0.0_ga, smarc-imx7_4.9.88_2.0.0_ga, smarc-imx_3.10.53_1.1.0_ga, smarc-imx_3.14.28_1.0.0_ga, smarc-imx_4.1.15_1.0.0_ga, smarc-imx_4.9.11_1.0.0_ga, smarc-imx_4.9.51_imx8m_ga, smarc-imx_4.9.88_2.0.0_ga, smarc-m6.0.1_2.1.0-ga, smarc-n7.1.2_2.0.0-ga, smarc-rel_imx_4.1.15_1.2.0_ga, smarc_8m_00d0_imx_4.14.98_2.0.0_ga, smarc_8m_imx_4.14.78_1.0.0_ga, smarc_8m_imx_4.14.98_2.0.0_ga, smarc_8m_imx_4.19.35_1.1.0, smarc_8mm_imx_4.14.78_1.0.0_ga, smarc_8mm_imx_4.14.98_2.0.0_ga, smarc_8mm_imx_4.19.35_1.1.0, smarc_8mm_imx_5.4.24_2.1.0, smarc_8mp_lf-5.10.y, smarc_8mq_imx_5.4.24_2.1.0, smarc_8mq_lf-5.10.y, smarc_imx_lf-5.15.y

block: fix bio_add_page for non trivial merge_bvec_fn case 

We have to properly decrease bi_size in order to merge_bvec_fn return
right result.  Otherwise this result in false merge rejects for two
absolutely valid bio_vecs.  This may cause significant performance
penalty for example fs_block_size == 1k and block device is raid0 with
small chunk_size = 8k. Then it is impossible to merge 7-th fs-block in
to bio which already has 6 fs-blocks.

Cc: <stable@kernel.org>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Showing 1 changed file with 6 additions and 1 deletions Inline Diff

1 /* 1 /*
2 * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk> 2 * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk>
3 * 3 *
4 * This program is free software; you can redistribute it and/or modify 4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as 5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation. 6 * published by the Free Software Foundation.
7 * 7 *
8 * This program is distributed in the hope that it will be useful, 8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details. 11 * GNU General Public License for more details.
12 * 12 *
13 * You should have received a copy of the GNU General Public Licens 13 * You should have received a copy of the GNU General Public Licens
14 * along with this program; if not, write to the Free Software 14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111- 15 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
16 * 16 *
17 */ 17 */
18 #include <linux/mm.h> 18 #include <linux/mm.h>
19 #include <linux/swap.h> 19 #include <linux/swap.h>
20 #include <linux/bio.h> 20 #include <linux/bio.h>
21 #include <linux/blkdev.h> 21 #include <linux/blkdev.h>
22 #include <linux/slab.h> 22 #include <linux/slab.h>
23 #include <linux/init.h> 23 #include <linux/init.h>
24 #include <linux/kernel.h> 24 #include <linux/kernel.h>
25 #include <linux/module.h> 25 #include <linux/module.h>
26 #include <linux/mempool.h> 26 #include <linux/mempool.h>
27 #include <linux/workqueue.h> 27 #include <linux/workqueue.h>
28 #include <scsi/sg.h> /* for struct sg_iovec */ 28 #include <scsi/sg.h> /* for struct sg_iovec */
29 29
30 #include <trace/events/block.h> 30 #include <trace/events/block.h>
31 31
32 /* 32 /*
33 * Test patch to inline a certain number of bi_io_vec's inside the bio 33 * Test patch to inline a certain number of bi_io_vec's inside the bio
34 * itself, to shrink a bio data allocation from two mempool calls to one 34 * itself, to shrink a bio data allocation from two mempool calls to one
35 */ 35 */
36 #define BIO_INLINE_VECS 4 36 #define BIO_INLINE_VECS 4
37 37
38 static mempool_t *bio_split_pool __read_mostly; 38 static mempool_t *bio_split_pool __read_mostly;
39 39
40 /* 40 /*
41 * if you change this list, also change bvec_alloc or things will 41 * if you change this list, also change bvec_alloc or things will
42 * break badly! cannot be bigger than what you can fit into an 42 * break badly! cannot be bigger than what you can fit into an
43 * unsigned short 43 * unsigned short
44 */ 44 */
45 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) } 45 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
46 struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = { 46 struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
47 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES), 47 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES),
48 }; 48 };
49 #undef BV 49 #undef BV
50 50
51 /* 51 /*
52 * fs_bio_set is the bio_set containing bio and iovec memory pools used by 52 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
53 * IO code that does not need private memory pools. 53 * IO code that does not need private memory pools.
54 */ 54 */
55 struct bio_set *fs_bio_set; 55 struct bio_set *fs_bio_set;
56 56
57 /* 57 /*
58 * Our slab pool management 58 * Our slab pool management
59 */ 59 */
60 struct bio_slab { 60 struct bio_slab {
61 struct kmem_cache *slab; 61 struct kmem_cache *slab;
62 unsigned int slab_ref; 62 unsigned int slab_ref;
63 unsigned int slab_size; 63 unsigned int slab_size;
64 char name[8]; 64 char name[8];
65 }; 65 };
66 static DEFINE_MUTEX(bio_slab_lock); 66 static DEFINE_MUTEX(bio_slab_lock);
67 static struct bio_slab *bio_slabs; 67 static struct bio_slab *bio_slabs;
68 static unsigned int bio_slab_nr, bio_slab_max; 68 static unsigned int bio_slab_nr, bio_slab_max;
69 69
70 static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size) 70 static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size)
71 { 71 {
72 unsigned int sz = sizeof(struct bio) + extra_size; 72 unsigned int sz = sizeof(struct bio) + extra_size;
73 struct kmem_cache *slab = NULL; 73 struct kmem_cache *slab = NULL;
74 struct bio_slab *bslab; 74 struct bio_slab *bslab;
75 unsigned int i, entry = -1; 75 unsigned int i, entry = -1;
76 76
77 mutex_lock(&bio_slab_lock); 77 mutex_lock(&bio_slab_lock);
78 78
79 i = 0; 79 i = 0;
80 while (i < bio_slab_nr) { 80 while (i < bio_slab_nr) {
81 bslab = &bio_slabs[i]; 81 bslab = &bio_slabs[i];
82 82
83 if (!bslab->slab && entry == -1) 83 if (!bslab->slab && entry == -1)
84 entry = i; 84 entry = i;
85 else if (bslab->slab_size == sz) { 85 else if (bslab->slab_size == sz) {
86 slab = bslab->slab; 86 slab = bslab->slab;
87 bslab->slab_ref++; 87 bslab->slab_ref++;
88 break; 88 break;
89 } 89 }
90 i++; 90 i++;
91 } 91 }
92 92
93 if (slab) 93 if (slab)
94 goto out_unlock; 94 goto out_unlock;
95 95
96 if (bio_slab_nr == bio_slab_max && entry == -1) { 96 if (bio_slab_nr == bio_slab_max && entry == -1) {
97 bio_slab_max <<= 1; 97 bio_slab_max <<= 1;
98 bio_slabs = krealloc(bio_slabs, 98 bio_slabs = krealloc(bio_slabs,
99 bio_slab_max * sizeof(struct bio_slab), 99 bio_slab_max * sizeof(struct bio_slab),
100 GFP_KERNEL); 100 GFP_KERNEL);
101 if (!bio_slabs) 101 if (!bio_slabs)
102 goto out_unlock; 102 goto out_unlock;
103 } 103 }
104 if (entry == -1) 104 if (entry == -1)
105 entry = bio_slab_nr++; 105 entry = bio_slab_nr++;
106 106
107 bslab = &bio_slabs[entry]; 107 bslab = &bio_slabs[entry];
108 108
109 snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry); 109 snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry);
110 slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL); 110 slab = kmem_cache_create(bslab->name, sz, 0, SLAB_HWCACHE_ALIGN, NULL);
111 if (!slab) 111 if (!slab)
112 goto out_unlock; 112 goto out_unlock;
113 113
114 printk("bio: create slab <%s> at %d\n", bslab->name, entry); 114 printk("bio: create slab <%s> at %d\n", bslab->name, entry);
115 bslab->slab = slab; 115 bslab->slab = slab;
116 bslab->slab_ref = 1; 116 bslab->slab_ref = 1;
117 bslab->slab_size = sz; 117 bslab->slab_size = sz;
118 out_unlock: 118 out_unlock:
119 mutex_unlock(&bio_slab_lock); 119 mutex_unlock(&bio_slab_lock);
120 return slab; 120 return slab;
121 } 121 }
122 122
123 static void bio_put_slab(struct bio_set *bs) 123 static void bio_put_slab(struct bio_set *bs)
124 { 124 {
125 struct bio_slab *bslab = NULL; 125 struct bio_slab *bslab = NULL;
126 unsigned int i; 126 unsigned int i;
127 127
128 mutex_lock(&bio_slab_lock); 128 mutex_lock(&bio_slab_lock);
129 129
130 for (i = 0; i < bio_slab_nr; i++) { 130 for (i = 0; i < bio_slab_nr; i++) {
131 if (bs->bio_slab == bio_slabs[i].slab) { 131 if (bs->bio_slab == bio_slabs[i].slab) {
132 bslab = &bio_slabs[i]; 132 bslab = &bio_slabs[i];
133 break; 133 break;
134 } 134 }
135 } 135 }
136 136
137 if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n")) 137 if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n"))
138 goto out; 138 goto out;
139 139
140 WARN_ON(!bslab->slab_ref); 140 WARN_ON(!bslab->slab_ref);
141 141
142 if (--bslab->slab_ref) 142 if (--bslab->slab_ref)
143 goto out; 143 goto out;
144 144
145 kmem_cache_destroy(bslab->slab); 145 kmem_cache_destroy(bslab->slab);
146 bslab->slab = NULL; 146 bslab->slab = NULL;
147 147
148 out: 148 out:
149 mutex_unlock(&bio_slab_lock); 149 mutex_unlock(&bio_slab_lock);
150 } 150 }
151 151
152 unsigned int bvec_nr_vecs(unsigned short idx) 152 unsigned int bvec_nr_vecs(unsigned short idx)
153 { 153 {
154 return bvec_slabs[idx].nr_vecs; 154 return bvec_slabs[idx].nr_vecs;
155 } 155 }
156 156
157 void bvec_free_bs(struct bio_set *bs, struct bio_vec *bv, unsigned int idx) 157 void bvec_free_bs(struct bio_set *bs, struct bio_vec *bv, unsigned int idx)
158 { 158 {
159 BIO_BUG_ON(idx >= BIOVEC_NR_POOLS); 159 BIO_BUG_ON(idx >= BIOVEC_NR_POOLS);
160 160
161 if (idx == BIOVEC_MAX_IDX) 161 if (idx == BIOVEC_MAX_IDX)
162 mempool_free(bv, bs->bvec_pool); 162 mempool_free(bv, bs->bvec_pool);
163 else { 163 else {
164 struct biovec_slab *bvs = bvec_slabs + idx; 164 struct biovec_slab *bvs = bvec_slabs + idx;
165 165
166 kmem_cache_free(bvs->slab, bv); 166 kmem_cache_free(bvs->slab, bv);
167 } 167 }
168 } 168 }
169 169
170 struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx, 170 struct bio_vec *bvec_alloc_bs(gfp_t gfp_mask, int nr, unsigned long *idx,
171 struct bio_set *bs) 171 struct bio_set *bs)
172 { 172 {
173 struct bio_vec *bvl; 173 struct bio_vec *bvl;
174 174
175 /* 175 /*
176 * see comment near bvec_array define! 176 * see comment near bvec_array define!
177 */ 177 */
178 switch (nr) { 178 switch (nr) {
179 case 1: 179 case 1:
180 *idx = 0; 180 *idx = 0;
181 break; 181 break;
182 case 2 ... 4: 182 case 2 ... 4:
183 *idx = 1; 183 *idx = 1;
184 break; 184 break;
185 case 5 ... 16: 185 case 5 ... 16:
186 *idx = 2; 186 *idx = 2;
187 break; 187 break;
188 case 17 ... 64: 188 case 17 ... 64:
189 *idx = 3; 189 *idx = 3;
190 break; 190 break;
191 case 65 ... 128: 191 case 65 ... 128:
192 *idx = 4; 192 *idx = 4;
193 break; 193 break;
194 case 129 ... BIO_MAX_PAGES: 194 case 129 ... BIO_MAX_PAGES:
195 *idx = 5; 195 *idx = 5;
196 break; 196 break;
197 default: 197 default:
198 return NULL; 198 return NULL;
199 } 199 }
200 200
201 /* 201 /*
202 * idx now points to the pool we want to allocate from. only the 202 * idx now points to the pool we want to allocate from. only the
203 * 1-vec entry pool is mempool backed. 203 * 1-vec entry pool is mempool backed.
204 */ 204 */
205 if (*idx == BIOVEC_MAX_IDX) { 205 if (*idx == BIOVEC_MAX_IDX) {
206 fallback: 206 fallback:
207 bvl = mempool_alloc(bs->bvec_pool, gfp_mask); 207 bvl = mempool_alloc(bs->bvec_pool, gfp_mask);
208 } else { 208 } else {
209 struct biovec_slab *bvs = bvec_slabs + *idx; 209 struct biovec_slab *bvs = bvec_slabs + *idx;
210 gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO); 210 gfp_t __gfp_mask = gfp_mask & ~(__GFP_WAIT | __GFP_IO);
211 211
212 /* 212 /*
213 * Make this allocation restricted and don't dump info on 213 * Make this allocation restricted and don't dump info on
214 * allocation failures, since we'll fallback to the mempool 214 * allocation failures, since we'll fallback to the mempool
215 * in case of failure. 215 * in case of failure.
216 */ 216 */
217 __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN; 217 __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN;
218 218
219 /* 219 /*
220 * Try a slab allocation. If this fails and __GFP_WAIT 220 * Try a slab allocation. If this fails and __GFP_WAIT
221 * is set, retry with the 1-entry mempool 221 * is set, retry with the 1-entry mempool
222 */ 222 */
223 bvl = kmem_cache_alloc(bvs->slab, __gfp_mask); 223 bvl = kmem_cache_alloc(bvs->slab, __gfp_mask);
224 if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) { 224 if (unlikely(!bvl && (gfp_mask & __GFP_WAIT))) {
225 *idx = BIOVEC_MAX_IDX; 225 *idx = BIOVEC_MAX_IDX;
226 goto fallback; 226 goto fallback;
227 } 227 }
228 } 228 }
229 229
230 return bvl; 230 return bvl;
231 } 231 }
232 232
233 void bio_free(struct bio *bio, struct bio_set *bs) 233 void bio_free(struct bio *bio, struct bio_set *bs)
234 { 234 {
235 void *p; 235 void *p;
236 236
237 if (bio_has_allocated_vec(bio)) 237 if (bio_has_allocated_vec(bio))
238 bvec_free_bs(bs, bio->bi_io_vec, BIO_POOL_IDX(bio)); 238 bvec_free_bs(bs, bio->bi_io_vec, BIO_POOL_IDX(bio));
239 239
240 if (bio_integrity(bio)) 240 if (bio_integrity(bio))
241 bio_integrity_free(bio, bs); 241 bio_integrity_free(bio, bs);
242 242
243 /* 243 /*
244 * If we have front padding, adjust the bio pointer before freeing 244 * If we have front padding, adjust the bio pointer before freeing
245 */ 245 */
246 p = bio; 246 p = bio;
247 if (bs->front_pad) 247 if (bs->front_pad)
248 p -= bs->front_pad; 248 p -= bs->front_pad;
249 249
250 mempool_free(p, bs->bio_pool); 250 mempool_free(p, bs->bio_pool);
251 } 251 }
252 EXPORT_SYMBOL(bio_free); 252 EXPORT_SYMBOL(bio_free);
253 253
254 void bio_init(struct bio *bio) 254 void bio_init(struct bio *bio)
255 { 255 {
256 memset(bio, 0, sizeof(*bio)); 256 memset(bio, 0, sizeof(*bio));
257 bio->bi_flags = 1 << BIO_UPTODATE; 257 bio->bi_flags = 1 << BIO_UPTODATE;
258 bio->bi_comp_cpu = -1; 258 bio->bi_comp_cpu = -1;
259 atomic_set(&bio->bi_cnt, 1); 259 atomic_set(&bio->bi_cnt, 1);
260 } 260 }
261 EXPORT_SYMBOL(bio_init); 261 EXPORT_SYMBOL(bio_init);
262 262
263 /** 263 /**
264 * bio_alloc_bioset - allocate a bio for I/O 264 * bio_alloc_bioset - allocate a bio for I/O
265 * @gfp_mask: the GFP_ mask given to the slab allocator 265 * @gfp_mask: the GFP_ mask given to the slab allocator
266 * @nr_iovecs: number of iovecs to pre-allocate 266 * @nr_iovecs: number of iovecs to pre-allocate
267 * @bs: the bio_set to allocate from. If %NULL, just use kmalloc 267 * @bs: the bio_set to allocate from. If %NULL, just use kmalloc
268 * 268 *
269 * Description: 269 * Description:
270 * bio_alloc_bioset will first try its own mempool to satisfy the allocation. 270 * bio_alloc_bioset will first try its own mempool to satisfy the allocation.
271 * If %__GFP_WAIT is set then we will block on the internal pool waiting 271 * If %__GFP_WAIT is set then we will block on the internal pool waiting
272 * for a &struct bio to become free. If a %NULL @bs is passed in, we will 272 * for a &struct bio to become free. If a %NULL @bs is passed in, we will
273 * fall back to just using @kmalloc to allocate the required memory. 273 * fall back to just using @kmalloc to allocate the required memory.
274 * 274 *
275 * Note that the caller must set ->bi_destructor on successful return 275 * Note that the caller must set ->bi_destructor on successful return
276 * of a bio, to do the appropriate freeing of the bio once the reference 276 * of a bio, to do the appropriate freeing of the bio once the reference
277 * count drops to zero. 277 * count drops to zero.
278 **/ 278 **/
279 struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs) 279 struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
280 { 280 {
281 unsigned long idx = BIO_POOL_NONE; 281 unsigned long idx = BIO_POOL_NONE;
282 struct bio_vec *bvl = NULL; 282 struct bio_vec *bvl = NULL;
283 struct bio *bio; 283 struct bio *bio;
284 void *p; 284 void *p;
285 285
286 p = mempool_alloc(bs->bio_pool, gfp_mask); 286 p = mempool_alloc(bs->bio_pool, gfp_mask);
287 if (unlikely(!p)) 287 if (unlikely(!p))
288 return NULL; 288 return NULL;
289 bio = p + bs->front_pad; 289 bio = p + bs->front_pad;
290 290
291 bio_init(bio); 291 bio_init(bio);
292 292
293 if (unlikely(!nr_iovecs)) 293 if (unlikely(!nr_iovecs))
294 goto out_set; 294 goto out_set;
295 295
296 if (nr_iovecs <= BIO_INLINE_VECS) { 296 if (nr_iovecs <= BIO_INLINE_VECS) {
297 bvl = bio->bi_inline_vecs; 297 bvl = bio->bi_inline_vecs;
298 nr_iovecs = BIO_INLINE_VECS; 298 nr_iovecs = BIO_INLINE_VECS;
299 } else { 299 } else {
300 bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs); 300 bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs);
301 if (unlikely(!bvl)) 301 if (unlikely(!bvl))
302 goto err_free; 302 goto err_free;
303 303
304 nr_iovecs = bvec_nr_vecs(idx); 304 nr_iovecs = bvec_nr_vecs(idx);
305 } 305 }
306 out_set: 306 out_set:
307 bio->bi_flags |= idx << BIO_POOL_OFFSET; 307 bio->bi_flags |= idx << BIO_POOL_OFFSET;
308 bio->bi_max_vecs = nr_iovecs; 308 bio->bi_max_vecs = nr_iovecs;
309 bio->bi_io_vec = bvl; 309 bio->bi_io_vec = bvl;
310 return bio; 310 return bio;
311 311
312 err_free: 312 err_free:
313 mempool_free(p, bs->bio_pool); 313 mempool_free(p, bs->bio_pool);
314 return NULL; 314 return NULL;
315 } 315 }
316 EXPORT_SYMBOL(bio_alloc_bioset); 316 EXPORT_SYMBOL(bio_alloc_bioset);
317 317
318 static void bio_fs_destructor(struct bio *bio) 318 static void bio_fs_destructor(struct bio *bio)
319 { 319 {
320 bio_free(bio, fs_bio_set); 320 bio_free(bio, fs_bio_set);
321 } 321 }
322 322
323 /** 323 /**
324 * bio_alloc - allocate a new bio, memory pool backed 324 * bio_alloc - allocate a new bio, memory pool backed
325 * @gfp_mask: allocation mask to use 325 * @gfp_mask: allocation mask to use
326 * @nr_iovecs: number of iovecs 326 * @nr_iovecs: number of iovecs
327 * 327 *
328 * bio_alloc will allocate a bio and associated bio_vec array that can hold 328 * bio_alloc will allocate a bio and associated bio_vec array that can hold
329 * at least @nr_iovecs entries. Allocations will be done from the 329 * at least @nr_iovecs entries. Allocations will be done from the
330 * fs_bio_set. Also see @bio_alloc_bioset and @bio_kmalloc. 330 * fs_bio_set. Also see @bio_alloc_bioset and @bio_kmalloc.
331 * 331 *
332 * If %__GFP_WAIT is set, then bio_alloc will always be able to allocate 332 * If %__GFP_WAIT is set, then bio_alloc will always be able to allocate
333 * a bio. This is due to the mempool guarantees. To make this work, callers 333 * a bio. This is due to the mempool guarantees. To make this work, callers
334 * must never allocate more than 1 bio at a time from this pool. Callers 334 * must never allocate more than 1 bio at a time from this pool. Callers
335 * that need to allocate more than 1 bio must always submit the previously 335 * that need to allocate more than 1 bio must always submit the previously
336 * allocated bio for IO before attempting to allocate a new one. Failure to 336 * allocated bio for IO before attempting to allocate a new one. Failure to
337 * do so can cause livelocks under memory pressure. 337 * do so can cause livelocks under memory pressure.
338 * 338 *
339 * RETURNS: 339 * RETURNS:
340 * Pointer to new bio on success, NULL on failure. 340 * Pointer to new bio on success, NULL on failure.
341 */ 341 */
342 struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs) 342 struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs)
343 { 343 {
344 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set); 344 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
345 345
346 if (bio) 346 if (bio)
347 bio->bi_destructor = bio_fs_destructor; 347 bio->bi_destructor = bio_fs_destructor;
348 348
349 return bio; 349 return bio;
350 } 350 }
351 EXPORT_SYMBOL(bio_alloc); 351 EXPORT_SYMBOL(bio_alloc);
352 352
353 static void bio_kmalloc_destructor(struct bio *bio) 353 static void bio_kmalloc_destructor(struct bio *bio)
354 { 354 {
355 if (bio_integrity(bio)) 355 if (bio_integrity(bio))
356 bio_integrity_free(bio, fs_bio_set); 356 bio_integrity_free(bio, fs_bio_set);
357 kfree(bio); 357 kfree(bio);
358 } 358 }
359 359
360 /** 360 /**
361 * bio_kmalloc - allocate a bio for I/O using kmalloc() 361 * bio_kmalloc - allocate a bio for I/O using kmalloc()
362 * @gfp_mask: the GFP_ mask given to the slab allocator 362 * @gfp_mask: the GFP_ mask given to the slab allocator
363 * @nr_iovecs: number of iovecs to pre-allocate 363 * @nr_iovecs: number of iovecs to pre-allocate
364 * 364 *
365 * Description: 365 * Description:
366 * Allocate a new bio with @nr_iovecs bvecs. If @gfp_mask contains 366 * Allocate a new bio with @nr_iovecs bvecs. If @gfp_mask contains
367 * %__GFP_WAIT, the allocation is guaranteed to succeed. 367 * %__GFP_WAIT, the allocation is guaranteed to succeed.
368 * 368 *
369 **/ 369 **/
370 struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs) 370 struct bio *bio_kmalloc(gfp_t gfp_mask, int nr_iovecs)
371 { 371 {
372 struct bio *bio; 372 struct bio *bio;
373 373
374 bio = kmalloc(sizeof(struct bio) + nr_iovecs * sizeof(struct bio_vec), 374 bio = kmalloc(sizeof(struct bio) + nr_iovecs * sizeof(struct bio_vec),
375 gfp_mask); 375 gfp_mask);
376 if (unlikely(!bio)) 376 if (unlikely(!bio))
377 return NULL; 377 return NULL;
378 378
379 bio_init(bio); 379 bio_init(bio);
380 bio->bi_flags |= BIO_POOL_NONE << BIO_POOL_OFFSET; 380 bio->bi_flags |= BIO_POOL_NONE << BIO_POOL_OFFSET;
381 bio->bi_max_vecs = nr_iovecs; 381 bio->bi_max_vecs = nr_iovecs;
382 bio->bi_io_vec = bio->bi_inline_vecs; 382 bio->bi_io_vec = bio->bi_inline_vecs;
383 bio->bi_destructor = bio_kmalloc_destructor; 383 bio->bi_destructor = bio_kmalloc_destructor;
384 384
385 return bio; 385 return bio;
386 } 386 }
387 EXPORT_SYMBOL(bio_kmalloc); 387 EXPORT_SYMBOL(bio_kmalloc);
388 388
389 void zero_fill_bio(struct bio *bio) 389 void zero_fill_bio(struct bio *bio)
390 { 390 {
391 unsigned long flags; 391 unsigned long flags;
392 struct bio_vec *bv; 392 struct bio_vec *bv;
393 int i; 393 int i;
394 394
395 bio_for_each_segment(bv, bio, i) { 395 bio_for_each_segment(bv, bio, i) {
396 char *data = bvec_kmap_irq(bv, &flags); 396 char *data = bvec_kmap_irq(bv, &flags);
397 memset(data, 0, bv->bv_len); 397 memset(data, 0, bv->bv_len);
398 flush_dcache_page(bv->bv_page); 398 flush_dcache_page(bv->bv_page);
399 bvec_kunmap_irq(data, &flags); 399 bvec_kunmap_irq(data, &flags);
400 } 400 }
401 } 401 }
402 EXPORT_SYMBOL(zero_fill_bio); 402 EXPORT_SYMBOL(zero_fill_bio);
403 403
404 /** 404 /**
405 * bio_put - release a reference to a bio 405 * bio_put - release a reference to a bio
406 * @bio: bio to release reference to 406 * @bio: bio to release reference to
407 * 407 *
408 * Description: 408 * Description:
409 * Put a reference to a &struct bio, either one you have gotten with 409 * Put a reference to a &struct bio, either one you have gotten with
410 * bio_alloc, bio_get or bio_clone. The last put of a bio will free it. 410 * bio_alloc, bio_get or bio_clone. The last put of a bio will free it.
411 **/ 411 **/
412 void bio_put(struct bio *bio) 412 void bio_put(struct bio *bio)
413 { 413 {
414 BIO_BUG_ON(!atomic_read(&bio->bi_cnt)); 414 BIO_BUG_ON(!atomic_read(&bio->bi_cnt));
415 415
416 /* 416 /*
417 * last put frees it 417 * last put frees it
418 */ 418 */
419 if (atomic_dec_and_test(&bio->bi_cnt)) { 419 if (atomic_dec_and_test(&bio->bi_cnt)) {
420 bio->bi_next = NULL; 420 bio->bi_next = NULL;
421 bio->bi_destructor(bio); 421 bio->bi_destructor(bio);
422 } 422 }
423 } 423 }
424 EXPORT_SYMBOL(bio_put); 424 EXPORT_SYMBOL(bio_put);
425 425
426 inline int bio_phys_segments(struct request_queue *q, struct bio *bio) 426 inline int bio_phys_segments(struct request_queue *q, struct bio *bio)
427 { 427 {
428 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID))) 428 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
429 blk_recount_segments(q, bio); 429 blk_recount_segments(q, bio);
430 430
431 return bio->bi_phys_segments; 431 return bio->bi_phys_segments;
432 } 432 }
433 EXPORT_SYMBOL(bio_phys_segments); 433 EXPORT_SYMBOL(bio_phys_segments);
434 434
435 /** 435 /**
436 * __bio_clone - clone a bio 436 * __bio_clone - clone a bio
437 * @bio: destination bio 437 * @bio: destination bio
438 * @bio_src: bio to clone 438 * @bio_src: bio to clone
439 * 439 *
440 * Clone a &bio. Caller will own the returned bio, but not 440 * Clone a &bio. Caller will own the returned bio, but not
441 * the actual data it points to. Reference count of returned 441 * the actual data it points to. Reference count of returned
442 * bio will be one. 442 * bio will be one.
443 */ 443 */
444 void __bio_clone(struct bio *bio, struct bio *bio_src) 444 void __bio_clone(struct bio *bio, struct bio *bio_src)
445 { 445 {
446 memcpy(bio->bi_io_vec, bio_src->bi_io_vec, 446 memcpy(bio->bi_io_vec, bio_src->bi_io_vec,
447 bio_src->bi_max_vecs * sizeof(struct bio_vec)); 447 bio_src->bi_max_vecs * sizeof(struct bio_vec));
448 448
449 /* 449 /*
450 * most users will be overriding ->bi_bdev with a new target, 450 * most users will be overriding ->bi_bdev with a new target,
451 * so we don't set nor calculate new physical/hw segment counts here 451 * so we don't set nor calculate new physical/hw segment counts here
452 */ 452 */
453 bio->bi_sector = bio_src->bi_sector; 453 bio->bi_sector = bio_src->bi_sector;
454 bio->bi_bdev = bio_src->bi_bdev; 454 bio->bi_bdev = bio_src->bi_bdev;
455 bio->bi_flags |= 1 << BIO_CLONED; 455 bio->bi_flags |= 1 << BIO_CLONED;
456 bio->bi_rw = bio_src->bi_rw; 456 bio->bi_rw = bio_src->bi_rw;
457 bio->bi_vcnt = bio_src->bi_vcnt; 457 bio->bi_vcnt = bio_src->bi_vcnt;
458 bio->bi_size = bio_src->bi_size; 458 bio->bi_size = bio_src->bi_size;
459 bio->bi_idx = bio_src->bi_idx; 459 bio->bi_idx = bio_src->bi_idx;
460 } 460 }
461 EXPORT_SYMBOL(__bio_clone); 461 EXPORT_SYMBOL(__bio_clone);
462 462
463 /** 463 /**
464 * bio_clone - clone a bio 464 * bio_clone - clone a bio
465 * @bio: bio to clone 465 * @bio: bio to clone
466 * @gfp_mask: allocation priority 466 * @gfp_mask: allocation priority
467 * 467 *
468 * Like __bio_clone, only also allocates the returned bio 468 * Like __bio_clone, only also allocates the returned bio
469 */ 469 */
470 struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask) 470 struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask)
471 { 471 {
472 struct bio *b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, fs_bio_set); 472 struct bio *b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, fs_bio_set);
473 473
474 if (!b) 474 if (!b)
475 return NULL; 475 return NULL;
476 476
477 b->bi_destructor = bio_fs_destructor; 477 b->bi_destructor = bio_fs_destructor;
478 __bio_clone(b, bio); 478 __bio_clone(b, bio);
479 479
480 if (bio_integrity(bio)) { 480 if (bio_integrity(bio)) {
481 int ret; 481 int ret;
482 482
483 ret = bio_integrity_clone(b, bio, gfp_mask, fs_bio_set); 483 ret = bio_integrity_clone(b, bio, gfp_mask, fs_bio_set);
484 484
485 if (ret < 0) { 485 if (ret < 0) {
486 bio_put(b); 486 bio_put(b);
487 return NULL; 487 return NULL;
488 } 488 }
489 } 489 }
490 490
491 return b; 491 return b;
492 } 492 }
493 EXPORT_SYMBOL(bio_clone); 493 EXPORT_SYMBOL(bio_clone);
494 494
495 /** 495 /**
496 * bio_get_nr_vecs - return approx number of vecs 496 * bio_get_nr_vecs - return approx number of vecs
497 * @bdev: I/O target 497 * @bdev: I/O target
498 * 498 *
499 * Return the approximate number of pages we can send to this target. 499 * Return the approximate number of pages we can send to this target.
500 * There's no guarantee that you will be able to fit this number of pages 500 * There's no guarantee that you will be able to fit this number of pages
501 * into a bio, it does not account for dynamic restrictions that vary 501 * into a bio, it does not account for dynamic restrictions that vary
502 * on offset. 502 * on offset.
503 */ 503 */
504 int bio_get_nr_vecs(struct block_device *bdev) 504 int bio_get_nr_vecs(struct block_device *bdev)
505 { 505 {
506 struct request_queue *q = bdev_get_queue(bdev); 506 struct request_queue *q = bdev_get_queue(bdev);
507 int nr_pages; 507 int nr_pages;
508 508
509 nr_pages = ((queue_max_sectors(q) << 9) + PAGE_SIZE - 1) >> PAGE_SHIFT; 509 nr_pages = ((queue_max_sectors(q) << 9) + PAGE_SIZE - 1) >> PAGE_SHIFT;
510 if (nr_pages > queue_max_phys_segments(q)) 510 if (nr_pages > queue_max_phys_segments(q))
511 nr_pages = queue_max_phys_segments(q); 511 nr_pages = queue_max_phys_segments(q);
512 if (nr_pages > queue_max_hw_segments(q)) 512 if (nr_pages > queue_max_hw_segments(q))
513 nr_pages = queue_max_hw_segments(q); 513 nr_pages = queue_max_hw_segments(q);
514 514
515 return nr_pages; 515 return nr_pages;
516 } 516 }
517 EXPORT_SYMBOL(bio_get_nr_vecs); 517 EXPORT_SYMBOL(bio_get_nr_vecs);
518 518
519 static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page 519 static int __bio_add_page(struct request_queue *q, struct bio *bio, struct page
520 *page, unsigned int len, unsigned int offset, 520 *page, unsigned int len, unsigned int offset,
521 unsigned short max_sectors) 521 unsigned short max_sectors)
522 { 522 {
523 int retried_segments = 0; 523 int retried_segments = 0;
524 struct bio_vec *bvec; 524 struct bio_vec *bvec;
525 525
526 /* 526 /*
527 * cloned bio must not modify vec list 527 * cloned bio must not modify vec list
528 */ 528 */
529 if (unlikely(bio_flagged(bio, BIO_CLONED))) 529 if (unlikely(bio_flagged(bio, BIO_CLONED)))
530 return 0; 530 return 0;
531 531
532 if (((bio->bi_size + len) >> 9) > max_sectors) 532 if (((bio->bi_size + len) >> 9) > max_sectors)
533 return 0; 533 return 0;
534 534
535 /* 535 /*
536 * For filesystems with a blocksize smaller than the pagesize 536 * For filesystems with a blocksize smaller than the pagesize
537 * we will often be called with the same page as last time and 537 * we will often be called with the same page as last time and
538 * a consecutive offset. Optimize this special case. 538 * a consecutive offset. Optimize this special case.
539 */ 539 */
540 if (bio->bi_vcnt > 0) { 540 if (bio->bi_vcnt > 0) {
541 struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1]; 541 struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1];
542 542
543 if (page == prev->bv_page && 543 if (page == prev->bv_page &&
544 offset == prev->bv_offset + prev->bv_len) { 544 offset == prev->bv_offset + prev->bv_len) {
545 unsigned int prev_bv_len = prev->bv_len;
545 prev->bv_len += len; 546 prev->bv_len += len;
546 547
547 if (q->merge_bvec_fn) { 548 if (q->merge_bvec_fn) {
548 struct bvec_merge_data bvm = { 549 struct bvec_merge_data bvm = {
550 /* prev_bvec is already charged in
551 bi_size, discharge it in order to
552 simulate merging updated prev_bvec
553 as new bvec. */
549 .bi_bdev = bio->bi_bdev, 554 .bi_bdev = bio->bi_bdev,
550 .bi_sector = bio->bi_sector, 555 .bi_sector = bio->bi_sector,
551 .bi_size = bio->bi_size, 556 .bi_size = bio->bi_size - prev_bv_len,
552 .bi_rw = bio->bi_rw, 557 .bi_rw = bio->bi_rw,
553 }; 558 };
554 559
555 if (q->merge_bvec_fn(q, &bvm, prev) < len) { 560 if (q->merge_bvec_fn(q, &bvm, prev) < len) {
556 prev->bv_len -= len; 561 prev->bv_len -= len;
557 return 0; 562 return 0;
558 } 563 }
559 } 564 }
560 565
561 goto done; 566 goto done;
562 } 567 }
563 } 568 }
564 569
565 if (bio->bi_vcnt >= bio->bi_max_vecs) 570 if (bio->bi_vcnt >= bio->bi_max_vecs)
566 return 0; 571 return 0;
567 572
568 /* 573 /*
569 * we might lose a segment or two here, but rather that than 574 * we might lose a segment or two here, but rather that than
570 * make this too complex. 575 * make this too complex.
571 */ 576 */
572 577
573 while (bio->bi_phys_segments >= queue_max_phys_segments(q) 578 while (bio->bi_phys_segments >= queue_max_phys_segments(q)
574 || bio->bi_phys_segments >= queue_max_hw_segments(q)) { 579 || bio->bi_phys_segments >= queue_max_hw_segments(q)) {
575 580
576 if (retried_segments) 581 if (retried_segments)
577 return 0; 582 return 0;
578 583
579 retried_segments = 1; 584 retried_segments = 1;
580 blk_recount_segments(q, bio); 585 blk_recount_segments(q, bio);
581 } 586 }
582 587
583 /* 588 /*
584 * setup the new entry, we might clear it again later if we 589 * setup the new entry, we might clear it again later if we
585 * cannot add the page 590 * cannot add the page
586 */ 591 */
587 bvec = &bio->bi_io_vec[bio->bi_vcnt]; 592 bvec = &bio->bi_io_vec[bio->bi_vcnt];
588 bvec->bv_page = page; 593 bvec->bv_page = page;
589 bvec->bv_len = len; 594 bvec->bv_len = len;
590 bvec->bv_offset = offset; 595 bvec->bv_offset = offset;
591 596
592 /* 597 /*
593 * if queue has other restrictions (eg varying max sector size 598 * if queue has other restrictions (eg varying max sector size
594 * depending on offset), it can specify a merge_bvec_fn in the 599 * depending on offset), it can specify a merge_bvec_fn in the
595 * queue to get further control 600 * queue to get further control
596 */ 601 */
597 if (q->merge_bvec_fn) { 602 if (q->merge_bvec_fn) {
598 struct bvec_merge_data bvm = { 603 struct bvec_merge_data bvm = {
599 .bi_bdev = bio->bi_bdev, 604 .bi_bdev = bio->bi_bdev,
600 .bi_sector = bio->bi_sector, 605 .bi_sector = bio->bi_sector,
601 .bi_size = bio->bi_size, 606 .bi_size = bio->bi_size,
602 .bi_rw = bio->bi_rw, 607 .bi_rw = bio->bi_rw,
603 }; 608 };
604 609
605 /* 610 /*
606 * merge_bvec_fn() returns number of bytes it can accept 611 * merge_bvec_fn() returns number of bytes it can accept
607 * at this offset 612 * at this offset
608 */ 613 */
609 if (q->merge_bvec_fn(q, &bvm, bvec) < len) { 614 if (q->merge_bvec_fn(q, &bvm, bvec) < len) {
610 bvec->bv_page = NULL; 615 bvec->bv_page = NULL;
611 bvec->bv_len = 0; 616 bvec->bv_len = 0;
612 bvec->bv_offset = 0; 617 bvec->bv_offset = 0;
613 return 0; 618 return 0;
614 } 619 }
615 } 620 }
616 621
617 /* If we may be able to merge these biovecs, force a recount */ 622 /* If we may be able to merge these biovecs, force a recount */
618 if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec))) 623 if (bio->bi_vcnt && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec)))
619 bio->bi_flags &= ~(1 << BIO_SEG_VALID); 624 bio->bi_flags &= ~(1 << BIO_SEG_VALID);
620 625
621 bio->bi_vcnt++; 626 bio->bi_vcnt++;
622 bio->bi_phys_segments++; 627 bio->bi_phys_segments++;
623 done: 628 done:
624 bio->bi_size += len; 629 bio->bi_size += len;
625 return len; 630 return len;
626 } 631 }
627 632
628 /** 633 /**
629 * bio_add_pc_page - attempt to add page to bio 634 * bio_add_pc_page - attempt to add page to bio
630 * @q: the target queue 635 * @q: the target queue
631 * @bio: destination bio 636 * @bio: destination bio
632 * @page: page to add 637 * @page: page to add
633 * @len: vec entry length 638 * @len: vec entry length
634 * @offset: vec entry offset 639 * @offset: vec entry offset
635 * 640 *
636 * Attempt to add a page to the bio_vec maplist. This can fail for a 641 * Attempt to add a page to the bio_vec maplist. This can fail for a
637 * number of reasons, such as the bio being full or target block 642 * number of reasons, such as the bio being full or target block
638 * device limitations. The target block device must allow bio's 643 * device limitations. The target block device must allow bio's
639 * smaller than PAGE_SIZE, so it is always possible to add a single 644 * smaller than PAGE_SIZE, so it is always possible to add a single
640 * page to an empty bio. This should only be used by REQ_PC bios. 645 * page to an empty bio. This should only be used by REQ_PC bios.
641 */ 646 */
642 int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page, 647 int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page *page,
643 unsigned int len, unsigned int offset) 648 unsigned int len, unsigned int offset)
644 { 649 {
645 return __bio_add_page(q, bio, page, len, offset, 650 return __bio_add_page(q, bio, page, len, offset,
646 queue_max_hw_sectors(q)); 651 queue_max_hw_sectors(q));
647 } 652 }
648 EXPORT_SYMBOL(bio_add_pc_page); 653 EXPORT_SYMBOL(bio_add_pc_page);
649 654
650 /** 655 /**
651 * bio_add_page - attempt to add page to bio 656 * bio_add_page - attempt to add page to bio
652 * @bio: destination bio 657 * @bio: destination bio
653 * @page: page to add 658 * @page: page to add
654 * @len: vec entry length 659 * @len: vec entry length
655 * @offset: vec entry offset 660 * @offset: vec entry offset
656 * 661 *
657 * Attempt to add a page to the bio_vec maplist. This can fail for a 662 * Attempt to add a page to the bio_vec maplist. This can fail for a
658 * number of reasons, such as the bio being full or target block 663 * number of reasons, such as the bio being full or target block
659 * device limitations. The target block device must allow bio's 664 * device limitations. The target block device must allow bio's
660 * smaller than PAGE_SIZE, so it is always possible to add a single 665 * smaller than PAGE_SIZE, so it is always possible to add a single
661 * page to an empty bio. 666 * page to an empty bio.
662 */ 667 */
663 int bio_add_page(struct bio *bio, struct page *page, unsigned int len, 668 int bio_add_page(struct bio *bio, struct page *page, unsigned int len,
664 unsigned int offset) 669 unsigned int offset)
665 { 670 {
666 struct request_queue *q = bdev_get_queue(bio->bi_bdev); 671 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
667 return __bio_add_page(q, bio, page, len, offset, queue_max_sectors(q)); 672 return __bio_add_page(q, bio, page, len, offset, queue_max_sectors(q));
668 } 673 }
669 EXPORT_SYMBOL(bio_add_page); 674 EXPORT_SYMBOL(bio_add_page);
670 675
671 struct bio_map_data { 676 struct bio_map_data {
672 struct bio_vec *iovecs; 677 struct bio_vec *iovecs;
673 struct sg_iovec *sgvecs; 678 struct sg_iovec *sgvecs;
674 int nr_sgvecs; 679 int nr_sgvecs;
675 int is_our_pages; 680 int is_our_pages;
676 }; 681 };
677 682
678 static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio, 683 static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio,
679 struct sg_iovec *iov, int iov_count, 684 struct sg_iovec *iov, int iov_count,
680 int is_our_pages) 685 int is_our_pages)
681 { 686 {
682 memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt); 687 memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt);
683 memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count); 688 memcpy(bmd->sgvecs, iov, sizeof(struct sg_iovec) * iov_count);
684 bmd->nr_sgvecs = iov_count; 689 bmd->nr_sgvecs = iov_count;
685 bmd->is_our_pages = is_our_pages; 690 bmd->is_our_pages = is_our_pages;
686 bio->bi_private = bmd; 691 bio->bi_private = bmd;
687 } 692 }
688 693
689 static void bio_free_map_data(struct bio_map_data *bmd) 694 static void bio_free_map_data(struct bio_map_data *bmd)
690 { 695 {
691 kfree(bmd->iovecs); 696 kfree(bmd->iovecs);
692 kfree(bmd->sgvecs); 697 kfree(bmd->sgvecs);
693 kfree(bmd); 698 kfree(bmd);
694 } 699 }
695 700
696 static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count, 701 static struct bio_map_data *bio_alloc_map_data(int nr_segs, int iov_count,
697 gfp_t gfp_mask) 702 gfp_t gfp_mask)
698 { 703 {
699 struct bio_map_data *bmd = kmalloc(sizeof(*bmd), gfp_mask); 704 struct bio_map_data *bmd = kmalloc(sizeof(*bmd), gfp_mask);
700 705
701 if (!bmd) 706 if (!bmd)
702 return NULL; 707 return NULL;
703 708
704 bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask); 709 bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, gfp_mask);
705 if (!bmd->iovecs) { 710 if (!bmd->iovecs) {
706 kfree(bmd); 711 kfree(bmd);
707 return NULL; 712 return NULL;
708 } 713 }
709 714
710 bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask); 715 bmd->sgvecs = kmalloc(sizeof(struct sg_iovec) * iov_count, gfp_mask);
711 if (bmd->sgvecs) 716 if (bmd->sgvecs)
712 return bmd; 717 return bmd;
713 718
714 kfree(bmd->iovecs); 719 kfree(bmd->iovecs);
715 kfree(bmd); 720 kfree(bmd);
716 return NULL; 721 return NULL;
717 } 722 }
718 723
719 static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs, 724 static int __bio_copy_iov(struct bio *bio, struct bio_vec *iovecs,
720 struct sg_iovec *iov, int iov_count, 725 struct sg_iovec *iov, int iov_count,
721 int to_user, int from_user, int do_free_page) 726 int to_user, int from_user, int do_free_page)
722 { 727 {
723 int ret = 0, i; 728 int ret = 0, i;
724 struct bio_vec *bvec; 729 struct bio_vec *bvec;
725 int iov_idx = 0; 730 int iov_idx = 0;
726 unsigned int iov_off = 0; 731 unsigned int iov_off = 0;
727 732
728 __bio_for_each_segment(bvec, bio, i, 0) { 733 __bio_for_each_segment(bvec, bio, i, 0) {
729 char *bv_addr = page_address(bvec->bv_page); 734 char *bv_addr = page_address(bvec->bv_page);
730 unsigned int bv_len = iovecs[i].bv_len; 735 unsigned int bv_len = iovecs[i].bv_len;
731 736
732 while (bv_len && iov_idx < iov_count) { 737 while (bv_len && iov_idx < iov_count) {
733 unsigned int bytes; 738 unsigned int bytes;
734 char __user *iov_addr; 739 char __user *iov_addr;
735 740
736 bytes = min_t(unsigned int, 741 bytes = min_t(unsigned int,
737 iov[iov_idx].iov_len - iov_off, bv_len); 742 iov[iov_idx].iov_len - iov_off, bv_len);
738 iov_addr = iov[iov_idx].iov_base + iov_off; 743 iov_addr = iov[iov_idx].iov_base + iov_off;
739 744
740 if (!ret) { 745 if (!ret) {
741 if (to_user) 746 if (to_user)
742 ret = copy_to_user(iov_addr, bv_addr, 747 ret = copy_to_user(iov_addr, bv_addr,
743 bytes); 748 bytes);
744 749
745 if (from_user) 750 if (from_user)
746 ret = copy_from_user(bv_addr, iov_addr, 751 ret = copy_from_user(bv_addr, iov_addr,
747 bytes); 752 bytes);
748 753
749 if (ret) 754 if (ret)
750 ret = -EFAULT; 755 ret = -EFAULT;
751 } 756 }
752 757
753 bv_len -= bytes; 758 bv_len -= bytes;
754 bv_addr += bytes; 759 bv_addr += bytes;
755 iov_addr += bytes; 760 iov_addr += bytes;
756 iov_off += bytes; 761 iov_off += bytes;
757 762
758 if (iov[iov_idx].iov_len == iov_off) { 763 if (iov[iov_idx].iov_len == iov_off) {
759 iov_idx++; 764 iov_idx++;
760 iov_off = 0; 765 iov_off = 0;
761 } 766 }
762 } 767 }
763 768
764 if (do_free_page) 769 if (do_free_page)
765 __free_page(bvec->bv_page); 770 __free_page(bvec->bv_page);
766 } 771 }
767 772
768 return ret; 773 return ret;
769 } 774 }
770 775
771 /** 776 /**
772 * bio_uncopy_user - finish previously mapped bio 777 * bio_uncopy_user - finish previously mapped bio
773 * @bio: bio being terminated 778 * @bio: bio being terminated
774 * 779 *
775 * Free pages allocated from bio_copy_user() and write back data 780 * Free pages allocated from bio_copy_user() and write back data
776 * to user space in case of a read. 781 * to user space in case of a read.
777 */ 782 */
778 int bio_uncopy_user(struct bio *bio) 783 int bio_uncopy_user(struct bio *bio)
779 { 784 {
780 struct bio_map_data *bmd = bio->bi_private; 785 struct bio_map_data *bmd = bio->bi_private;
781 int ret = 0; 786 int ret = 0;
782 787
783 if (!bio_flagged(bio, BIO_NULL_MAPPED)) 788 if (!bio_flagged(bio, BIO_NULL_MAPPED))
784 ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs, 789 ret = __bio_copy_iov(bio, bmd->iovecs, bmd->sgvecs,
785 bmd->nr_sgvecs, bio_data_dir(bio) == READ, 790 bmd->nr_sgvecs, bio_data_dir(bio) == READ,
786 0, bmd->is_our_pages); 791 0, bmd->is_our_pages);
787 bio_free_map_data(bmd); 792 bio_free_map_data(bmd);
788 bio_put(bio); 793 bio_put(bio);
789 return ret; 794 return ret;
790 } 795 }
791 EXPORT_SYMBOL(bio_uncopy_user); 796 EXPORT_SYMBOL(bio_uncopy_user);
792 797
793 /** 798 /**
794 * bio_copy_user_iov - copy user data to bio 799 * bio_copy_user_iov - copy user data to bio
795 * @q: destination block queue 800 * @q: destination block queue
796 * @map_data: pointer to the rq_map_data holding pages (if necessary) 801 * @map_data: pointer to the rq_map_data holding pages (if necessary)
797 * @iov: the iovec. 802 * @iov: the iovec.
798 * @iov_count: number of elements in the iovec 803 * @iov_count: number of elements in the iovec
799 * @write_to_vm: bool indicating writing to pages or not 804 * @write_to_vm: bool indicating writing to pages or not
800 * @gfp_mask: memory allocation flags 805 * @gfp_mask: memory allocation flags
801 * 806 *
802 * Prepares and returns a bio for indirect user io, bouncing data 807 * Prepares and returns a bio for indirect user io, bouncing data
803 * to/from kernel pages as necessary. Must be paired with 808 * to/from kernel pages as necessary. Must be paired with
804 * call bio_uncopy_user() on io completion. 809 * call bio_uncopy_user() on io completion.
805 */ 810 */
806 struct bio *bio_copy_user_iov(struct request_queue *q, 811 struct bio *bio_copy_user_iov(struct request_queue *q,
807 struct rq_map_data *map_data, 812 struct rq_map_data *map_data,
808 struct sg_iovec *iov, int iov_count, 813 struct sg_iovec *iov, int iov_count,
809 int write_to_vm, gfp_t gfp_mask) 814 int write_to_vm, gfp_t gfp_mask)
810 { 815 {
811 struct bio_map_data *bmd; 816 struct bio_map_data *bmd;
812 struct bio_vec *bvec; 817 struct bio_vec *bvec;
813 struct page *page; 818 struct page *page;
814 struct bio *bio; 819 struct bio *bio;
815 int i, ret; 820 int i, ret;
816 int nr_pages = 0; 821 int nr_pages = 0;
817 unsigned int len = 0; 822 unsigned int len = 0;
818 unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0; 823 unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0;
819 824
820 for (i = 0; i < iov_count; i++) { 825 for (i = 0; i < iov_count; i++) {
821 unsigned long uaddr; 826 unsigned long uaddr;
822 unsigned long end; 827 unsigned long end;
823 unsigned long start; 828 unsigned long start;
824 829
825 uaddr = (unsigned long)iov[i].iov_base; 830 uaddr = (unsigned long)iov[i].iov_base;
826 end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT; 831 end = (uaddr + iov[i].iov_len + PAGE_SIZE - 1) >> PAGE_SHIFT;
827 start = uaddr >> PAGE_SHIFT; 832 start = uaddr >> PAGE_SHIFT;
828 833
829 nr_pages += end - start; 834 nr_pages += end - start;
830 len += iov[i].iov_len; 835 len += iov[i].iov_len;
831 } 836 }
832 837
833 if (offset) 838 if (offset)
834 nr_pages++; 839 nr_pages++;
835 840
836 bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask); 841 bmd = bio_alloc_map_data(nr_pages, iov_count, gfp_mask);
837 if (!bmd) 842 if (!bmd)
838 return ERR_PTR(-ENOMEM); 843 return ERR_PTR(-ENOMEM);
839 844
840 ret = -ENOMEM; 845 ret = -ENOMEM;
841 bio = bio_kmalloc(gfp_mask, nr_pages); 846 bio = bio_kmalloc(gfp_mask, nr_pages);
842 if (!bio) 847 if (!bio)
843 goto out_bmd; 848 goto out_bmd;
844 849
845 bio->bi_rw |= (!write_to_vm << BIO_RW); 850 bio->bi_rw |= (!write_to_vm << BIO_RW);
846 851
847 ret = 0; 852 ret = 0;
848 853
849 if (map_data) { 854 if (map_data) {
850 nr_pages = 1 << map_data->page_order; 855 nr_pages = 1 << map_data->page_order;
851 i = map_data->offset / PAGE_SIZE; 856 i = map_data->offset / PAGE_SIZE;
852 } 857 }
853 while (len) { 858 while (len) {
854 unsigned int bytes = PAGE_SIZE; 859 unsigned int bytes = PAGE_SIZE;
855 860
856 bytes -= offset; 861 bytes -= offset;
857 862
858 if (bytes > len) 863 if (bytes > len)
859 bytes = len; 864 bytes = len;
860 865
861 if (map_data) { 866 if (map_data) {
862 if (i == map_data->nr_entries * nr_pages) { 867 if (i == map_data->nr_entries * nr_pages) {
863 ret = -ENOMEM; 868 ret = -ENOMEM;
864 break; 869 break;
865 } 870 }
866 871
867 page = map_data->pages[i / nr_pages]; 872 page = map_data->pages[i / nr_pages];
868 page += (i % nr_pages); 873 page += (i % nr_pages);
869 874
870 i++; 875 i++;
871 } else { 876 } else {
872 page = alloc_page(q->bounce_gfp | gfp_mask); 877 page = alloc_page(q->bounce_gfp | gfp_mask);
873 if (!page) { 878 if (!page) {
874 ret = -ENOMEM; 879 ret = -ENOMEM;
875 break; 880 break;
876 } 881 }
877 } 882 }
878 883
879 if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) 884 if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes)
880 break; 885 break;
881 886
882 len -= bytes; 887 len -= bytes;
883 offset = 0; 888 offset = 0;
884 } 889 }
885 890
886 if (ret) 891 if (ret)
887 goto cleanup; 892 goto cleanup;
888 893
889 /* 894 /*
890 * success 895 * success
891 */ 896 */
892 if ((!write_to_vm && (!map_data || !map_data->null_mapped)) || 897 if ((!write_to_vm && (!map_data || !map_data->null_mapped)) ||
893 (map_data && map_data->from_user)) { 898 (map_data && map_data->from_user)) {
894 ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 1, 0); 899 ret = __bio_copy_iov(bio, bio->bi_io_vec, iov, iov_count, 0, 1, 0);
895 if (ret) 900 if (ret)
896 goto cleanup; 901 goto cleanup;
897 } 902 }
898 903
899 bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1); 904 bio_set_map_data(bmd, bio, iov, iov_count, map_data ? 0 : 1);
900 return bio; 905 return bio;
901 cleanup: 906 cleanup:
902 if (!map_data) 907 if (!map_data)
903 bio_for_each_segment(bvec, bio, i) 908 bio_for_each_segment(bvec, bio, i)
904 __free_page(bvec->bv_page); 909 __free_page(bvec->bv_page);
905 910
906 bio_put(bio); 911 bio_put(bio);
907 out_bmd: 912 out_bmd:
908 bio_free_map_data(bmd); 913 bio_free_map_data(bmd);
909 return ERR_PTR(ret); 914 return ERR_PTR(ret);
910 } 915 }
911 916
912 /** 917 /**
913 * bio_copy_user - copy user data to bio 918 * bio_copy_user - copy user data to bio
914 * @q: destination block queue 919 * @q: destination block queue
915 * @map_data: pointer to the rq_map_data holding pages (if necessary) 920 * @map_data: pointer to the rq_map_data holding pages (if necessary)
916 * @uaddr: start of user address 921 * @uaddr: start of user address
917 * @len: length in bytes 922 * @len: length in bytes
918 * @write_to_vm: bool indicating writing to pages or not 923 * @write_to_vm: bool indicating writing to pages or not
919 * @gfp_mask: memory allocation flags 924 * @gfp_mask: memory allocation flags
920 * 925 *
921 * Prepares and returns a bio for indirect user io, bouncing data 926 * Prepares and returns a bio for indirect user io, bouncing data
922 * to/from kernel pages as necessary. Must be paired with 927 * to/from kernel pages as necessary. Must be paired with
923 * call bio_uncopy_user() on io completion. 928 * call bio_uncopy_user() on io completion.
924 */ 929 */
925 struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data, 930 struct bio *bio_copy_user(struct request_queue *q, struct rq_map_data *map_data,
926 unsigned long uaddr, unsigned int len, 931 unsigned long uaddr, unsigned int len,
927 int write_to_vm, gfp_t gfp_mask) 932 int write_to_vm, gfp_t gfp_mask)
928 { 933 {
929 struct sg_iovec iov; 934 struct sg_iovec iov;
930 935
931 iov.iov_base = (void __user *)uaddr; 936 iov.iov_base = (void __user *)uaddr;
932 iov.iov_len = len; 937 iov.iov_len = len;
933 938
934 return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask); 939 return bio_copy_user_iov(q, map_data, &iov, 1, write_to_vm, gfp_mask);
935 } 940 }
936 EXPORT_SYMBOL(bio_copy_user); 941 EXPORT_SYMBOL(bio_copy_user);
937 942
938 static struct bio *__bio_map_user_iov(struct request_queue *q, 943 static struct bio *__bio_map_user_iov(struct request_queue *q,
939 struct block_device *bdev, 944 struct block_device *bdev,
940 struct sg_iovec *iov, int iov_count, 945 struct sg_iovec *iov, int iov_count,
941 int write_to_vm, gfp_t gfp_mask) 946 int write_to_vm, gfp_t gfp_mask)
942 { 947 {
943 int i, j; 948 int i, j;
944 int nr_pages = 0; 949 int nr_pages = 0;
945 struct page **pages; 950 struct page **pages;
946 struct bio *bio; 951 struct bio *bio;
947 int cur_page = 0; 952 int cur_page = 0;
948 int ret, offset; 953 int ret, offset;
949 954
950 for (i = 0; i < iov_count; i++) { 955 for (i = 0; i < iov_count; i++) {
951 unsigned long uaddr = (unsigned long)iov[i].iov_base; 956 unsigned long uaddr = (unsigned long)iov[i].iov_base;
952 unsigned long len = iov[i].iov_len; 957 unsigned long len = iov[i].iov_len;
953 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; 958 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
954 unsigned long start = uaddr >> PAGE_SHIFT; 959 unsigned long start = uaddr >> PAGE_SHIFT;
955 960
956 nr_pages += end - start; 961 nr_pages += end - start;
957 /* 962 /*
958 * buffer must be aligned to at least hardsector size for now 963 * buffer must be aligned to at least hardsector size for now
959 */ 964 */
960 if (uaddr & queue_dma_alignment(q)) 965 if (uaddr & queue_dma_alignment(q))
961 return ERR_PTR(-EINVAL); 966 return ERR_PTR(-EINVAL);
962 } 967 }
963 968
964 if (!nr_pages) 969 if (!nr_pages)
965 return ERR_PTR(-EINVAL); 970 return ERR_PTR(-EINVAL);
966 971
967 bio = bio_kmalloc(gfp_mask, nr_pages); 972 bio = bio_kmalloc(gfp_mask, nr_pages);
968 if (!bio) 973 if (!bio)
969 return ERR_PTR(-ENOMEM); 974 return ERR_PTR(-ENOMEM);
970 975
971 ret = -ENOMEM; 976 ret = -ENOMEM;
972 pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask); 977 pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask);
973 if (!pages) 978 if (!pages)
974 goto out; 979 goto out;
975 980
976 for (i = 0; i < iov_count; i++) { 981 for (i = 0; i < iov_count; i++) {
977 unsigned long uaddr = (unsigned long)iov[i].iov_base; 982 unsigned long uaddr = (unsigned long)iov[i].iov_base;
978 unsigned long len = iov[i].iov_len; 983 unsigned long len = iov[i].iov_len;
979 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; 984 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
980 unsigned long start = uaddr >> PAGE_SHIFT; 985 unsigned long start = uaddr >> PAGE_SHIFT;
981 const int local_nr_pages = end - start; 986 const int local_nr_pages = end - start;
982 const int page_limit = cur_page + local_nr_pages; 987 const int page_limit = cur_page + local_nr_pages;
983 988
984 ret = get_user_pages_fast(uaddr, local_nr_pages, 989 ret = get_user_pages_fast(uaddr, local_nr_pages,
985 write_to_vm, &pages[cur_page]); 990 write_to_vm, &pages[cur_page]);
986 if (ret < local_nr_pages) { 991 if (ret < local_nr_pages) {
987 ret = -EFAULT; 992 ret = -EFAULT;
988 goto out_unmap; 993 goto out_unmap;
989 } 994 }
990 995
991 offset = uaddr & ~PAGE_MASK; 996 offset = uaddr & ~PAGE_MASK;
992 for (j = cur_page; j < page_limit; j++) { 997 for (j = cur_page; j < page_limit; j++) {
993 unsigned int bytes = PAGE_SIZE - offset; 998 unsigned int bytes = PAGE_SIZE - offset;
994 999
995 if (len <= 0) 1000 if (len <= 0)
996 break; 1001 break;
997 1002
998 if (bytes > len) 1003 if (bytes > len)
999 bytes = len; 1004 bytes = len;
1000 1005
1001 /* 1006 /*
1002 * sorry... 1007 * sorry...
1003 */ 1008 */
1004 if (bio_add_pc_page(q, bio, pages[j], bytes, offset) < 1009 if (bio_add_pc_page(q, bio, pages[j], bytes, offset) <
1005 bytes) 1010 bytes)
1006 break; 1011 break;
1007 1012
1008 len -= bytes; 1013 len -= bytes;
1009 offset = 0; 1014 offset = 0;
1010 } 1015 }
1011 1016
1012 cur_page = j; 1017 cur_page = j;
1013 /* 1018 /*
1014 * release the pages we didn't map into the bio, if any 1019 * release the pages we didn't map into the bio, if any
1015 */ 1020 */
1016 while (j < page_limit) 1021 while (j < page_limit)
1017 page_cache_release(pages[j++]); 1022 page_cache_release(pages[j++]);
1018 } 1023 }
1019 1024
1020 kfree(pages); 1025 kfree(pages);
1021 1026
1022 /* 1027 /*
1023 * set data direction, and check if mapped pages need bouncing 1028 * set data direction, and check if mapped pages need bouncing
1024 */ 1029 */
1025 if (!write_to_vm) 1030 if (!write_to_vm)
1026 bio->bi_rw |= (1 << BIO_RW); 1031 bio->bi_rw |= (1 << BIO_RW);
1027 1032
1028 bio->bi_bdev = bdev; 1033 bio->bi_bdev = bdev;
1029 bio->bi_flags |= (1 << BIO_USER_MAPPED); 1034 bio->bi_flags |= (1 << BIO_USER_MAPPED);
1030 return bio; 1035 return bio;
1031 1036
1032 out_unmap: 1037 out_unmap:
1033 for (i = 0; i < nr_pages; i++) { 1038 for (i = 0; i < nr_pages; i++) {
1034 if(!pages[i]) 1039 if(!pages[i])
1035 break; 1040 break;
1036 page_cache_release(pages[i]); 1041 page_cache_release(pages[i]);
1037 } 1042 }
1038 out: 1043 out:
1039 kfree(pages); 1044 kfree(pages);
1040 bio_put(bio); 1045 bio_put(bio);
1041 return ERR_PTR(ret); 1046 return ERR_PTR(ret);
1042 } 1047 }
1043 1048
1044 /** 1049 /**
1045 * bio_map_user - map user address into bio 1050 * bio_map_user - map user address into bio
1046 * @q: the struct request_queue for the bio 1051 * @q: the struct request_queue for the bio
1047 * @bdev: destination block device 1052 * @bdev: destination block device
1048 * @uaddr: start of user address 1053 * @uaddr: start of user address
1049 * @len: length in bytes 1054 * @len: length in bytes
1050 * @write_to_vm: bool indicating writing to pages or not 1055 * @write_to_vm: bool indicating writing to pages or not
1051 * @gfp_mask: memory allocation flags 1056 * @gfp_mask: memory allocation flags
1052 * 1057 *
1053 * Map the user space address into a bio suitable for io to a block 1058 * Map the user space address into a bio suitable for io to a block
1054 * device. Returns an error pointer in case of error. 1059 * device. Returns an error pointer in case of error.
1055 */ 1060 */
1056 struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev, 1061 struct bio *bio_map_user(struct request_queue *q, struct block_device *bdev,
1057 unsigned long uaddr, unsigned int len, int write_to_vm, 1062 unsigned long uaddr, unsigned int len, int write_to_vm,
1058 gfp_t gfp_mask) 1063 gfp_t gfp_mask)
1059 { 1064 {
1060 struct sg_iovec iov; 1065 struct sg_iovec iov;
1061 1066
1062 iov.iov_base = (void __user *)uaddr; 1067 iov.iov_base = (void __user *)uaddr;
1063 iov.iov_len = len; 1068 iov.iov_len = len;
1064 1069
1065 return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask); 1070 return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm, gfp_mask);
1066 } 1071 }
1067 EXPORT_SYMBOL(bio_map_user); 1072 EXPORT_SYMBOL(bio_map_user);
1068 1073
1069 /** 1074 /**
1070 * bio_map_user_iov - map user sg_iovec table into bio 1075 * bio_map_user_iov - map user sg_iovec table into bio
1071 * @q: the struct request_queue for the bio 1076 * @q: the struct request_queue for the bio
1072 * @bdev: destination block device 1077 * @bdev: destination block device
1073 * @iov: the iovec. 1078 * @iov: the iovec.
1074 * @iov_count: number of elements in the iovec 1079 * @iov_count: number of elements in the iovec
1075 * @write_to_vm: bool indicating writing to pages or not 1080 * @write_to_vm: bool indicating writing to pages or not
1076 * @gfp_mask: memory allocation flags 1081 * @gfp_mask: memory allocation flags
1077 * 1082 *
1078 * Map the user space address into a bio suitable for io to a block 1083 * Map the user space address into a bio suitable for io to a block
1079 * device. Returns an error pointer in case of error. 1084 * device. Returns an error pointer in case of error.
1080 */ 1085 */
1081 struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev, 1086 struct bio *bio_map_user_iov(struct request_queue *q, struct block_device *bdev,
1082 struct sg_iovec *iov, int iov_count, 1087 struct sg_iovec *iov, int iov_count,
1083 int write_to_vm, gfp_t gfp_mask) 1088 int write_to_vm, gfp_t gfp_mask)
1084 { 1089 {
1085 struct bio *bio; 1090 struct bio *bio;
1086 1091
1087 bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm, 1092 bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm,
1088 gfp_mask); 1093 gfp_mask);
1089 if (IS_ERR(bio)) 1094 if (IS_ERR(bio))
1090 return bio; 1095 return bio;
1091 1096
1092 /* 1097 /*
1093 * subtle -- if __bio_map_user() ended up bouncing a bio, 1098 * subtle -- if __bio_map_user() ended up bouncing a bio,
1094 * it would normally disappear when its bi_end_io is run. 1099 * it would normally disappear when its bi_end_io is run.
1095 * however, we need it for the unmap, so grab an extra 1100 * however, we need it for the unmap, so grab an extra
1096 * reference to it 1101 * reference to it
1097 */ 1102 */
1098 bio_get(bio); 1103 bio_get(bio);
1099 1104
1100 return bio; 1105 return bio;
1101 } 1106 }
1102 1107
1103 static void __bio_unmap_user(struct bio *bio) 1108 static void __bio_unmap_user(struct bio *bio)
1104 { 1109 {
1105 struct bio_vec *bvec; 1110 struct bio_vec *bvec;
1106 int i; 1111 int i;
1107 1112
1108 /* 1113 /*
1109 * make sure we dirty pages we wrote to 1114 * make sure we dirty pages we wrote to
1110 */ 1115 */
1111 __bio_for_each_segment(bvec, bio, i, 0) { 1116 __bio_for_each_segment(bvec, bio, i, 0) {
1112 if (bio_data_dir(bio) == READ) 1117 if (bio_data_dir(bio) == READ)
1113 set_page_dirty_lock(bvec->bv_page); 1118 set_page_dirty_lock(bvec->bv_page);
1114 1119
1115 page_cache_release(bvec->bv_page); 1120 page_cache_release(bvec->bv_page);
1116 } 1121 }
1117 1122
1118 bio_put(bio); 1123 bio_put(bio);
1119 } 1124 }
1120 1125
1121 /** 1126 /**
1122 * bio_unmap_user - unmap a bio 1127 * bio_unmap_user - unmap a bio
1123 * @bio: the bio being unmapped 1128 * @bio: the bio being unmapped
1124 * 1129 *
1125 * Unmap a bio previously mapped by bio_map_user(). Must be called with 1130 * Unmap a bio previously mapped by bio_map_user(). Must be called with
1126 * a process context. 1131 * a process context.
1127 * 1132 *
1128 * bio_unmap_user() may sleep. 1133 * bio_unmap_user() may sleep.
1129 */ 1134 */
1130 void bio_unmap_user(struct bio *bio) 1135 void bio_unmap_user(struct bio *bio)
1131 { 1136 {
1132 __bio_unmap_user(bio); 1137 __bio_unmap_user(bio);
1133 bio_put(bio); 1138 bio_put(bio);
1134 } 1139 }
1135 EXPORT_SYMBOL(bio_unmap_user); 1140 EXPORT_SYMBOL(bio_unmap_user);
1136 1141
1137 static void bio_map_kern_endio(struct bio *bio, int err) 1142 static void bio_map_kern_endio(struct bio *bio, int err)
1138 { 1143 {
1139 bio_put(bio); 1144 bio_put(bio);
1140 } 1145 }
1141 1146
1142 static struct bio *__bio_map_kern(struct request_queue *q, void *data, 1147 static struct bio *__bio_map_kern(struct request_queue *q, void *data,
1143 unsigned int len, gfp_t gfp_mask) 1148 unsigned int len, gfp_t gfp_mask)
1144 { 1149 {
1145 unsigned long kaddr = (unsigned long)data; 1150 unsigned long kaddr = (unsigned long)data;
1146 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT; 1151 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1147 unsigned long start = kaddr >> PAGE_SHIFT; 1152 unsigned long start = kaddr >> PAGE_SHIFT;
1148 const int nr_pages = end - start; 1153 const int nr_pages = end - start;
1149 int offset, i; 1154 int offset, i;
1150 struct bio *bio; 1155 struct bio *bio;
1151 1156
1152 bio = bio_kmalloc(gfp_mask, nr_pages); 1157 bio = bio_kmalloc(gfp_mask, nr_pages);
1153 if (!bio) 1158 if (!bio)
1154 return ERR_PTR(-ENOMEM); 1159 return ERR_PTR(-ENOMEM);
1155 1160
1156 offset = offset_in_page(kaddr); 1161 offset = offset_in_page(kaddr);
1157 for (i = 0; i < nr_pages; i++) { 1162 for (i = 0; i < nr_pages; i++) {
1158 unsigned int bytes = PAGE_SIZE - offset; 1163 unsigned int bytes = PAGE_SIZE - offset;
1159 1164
1160 if (len <= 0) 1165 if (len <= 0)
1161 break; 1166 break;
1162 1167
1163 if (bytes > len) 1168 if (bytes > len)
1164 bytes = len; 1169 bytes = len;
1165 1170
1166 if (bio_add_pc_page(q, bio, virt_to_page(data), bytes, 1171 if (bio_add_pc_page(q, bio, virt_to_page(data), bytes,
1167 offset) < bytes) 1172 offset) < bytes)
1168 break; 1173 break;
1169 1174
1170 data += bytes; 1175 data += bytes;
1171 len -= bytes; 1176 len -= bytes;
1172 offset = 0; 1177 offset = 0;
1173 } 1178 }
1174 1179
1175 bio->bi_end_io = bio_map_kern_endio; 1180 bio->bi_end_io = bio_map_kern_endio;
1176 return bio; 1181 return bio;
1177 } 1182 }
1178 1183
1179 /** 1184 /**
1180 * bio_map_kern - map kernel address into bio 1185 * bio_map_kern - map kernel address into bio
1181 * @q: the struct request_queue for the bio 1186 * @q: the struct request_queue for the bio
1182 * @data: pointer to buffer to map 1187 * @data: pointer to buffer to map
1183 * @len: length in bytes 1188 * @len: length in bytes
1184 * @gfp_mask: allocation flags for bio allocation 1189 * @gfp_mask: allocation flags for bio allocation
1185 * 1190 *
1186 * Map the kernel address into a bio suitable for io to a block 1191 * Map the kernel address into a bio suitable for io to a block
1187 * device. Returns an error pointer in case of error. 1192 * device. Returns an error pointer in case of error.
1188 */ 1193 */
1189 struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len, 1194 struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len,
1190 gfp_t gfp_mask) 1195 gfp_t gfp_mask)
1191 { 1196 {
1192 struct bio *bio; 1197 struct bio *bio;
1193 1198
1194 bio = __bio_map_kern(q, data, len, gfp_mask); 1199 bio = __bio_map_kern(q, data, len, gfp_mask);
1195 if (IS_ERR(bio)) 1200 if (IS_ERR(bio))
1196 return bio; 1201 return bio;
1197 1202
1198 if (bio->bi_size == len) 1203 if (bio->bi_size == len)
1199 return bio; 1204 return bio;
1200 1205
1201 /* 1206 /*
1202 * Don't support partial mappings. 1207 * Don't support partial mappings.
1203 */ 1208 */
1204 bio_put(bio); 1209 bio_put(bio);
1205 return ERR_PTR(-EINVAL); 1210 return ERR_PTR(-EINVAL);
1206 } 1211 }
1207 EXPORT_SYMBOL(bio_map_kern); 1212 EXPORT_SYMBOL(bio_map_kern);
1208 1213
1209 static void bio_copy_kern_endio(struct bio *bio, int err) 1214 static void bio_copy_kern_endio(struct bio *bio, int err)
1210 { 1215 {
1211 struct bio_vec *bvec; 1216 struct bio_vec *bvec;
1212 const int read = bio_data_dir(bio) == READ; 1217 const int read = bio_data_dir(bio) == READ;
1213 struct bio_map_data *bmd = bio->bi_private; 1218 struct bio_map_data *bmd = bio->bi_private;
1214 int i; 1219 int i;
1215 char *p = bmd->sgvecs[0].iov_base; 1220 char *p = bmd->sgvecs[0].iov_base;
1216 1221
1217 __bio_for_each_segment(bvec, bio, i, 0) { 1222 __bio_for_each_segment(bvec, bio, i, 0) {
1218 char *addr = page_address(bvec->bv_page); 1223 char *addr = page_address(bvec->bv_page);
1219 int len = bmd->iovecs[i].bv_len; 1224 int len = bmd->iovecs[i].bv_len;
1220 1225
1221 if (read) 1226 if (read)
1222 memcpy(p, addr, len); 1227 memcpy(p, addr, len);
1223 1228
1224 __free_page(bvec->bv_page); 1229 __free_page(bvec->bv_page);
1225 p += len; 1230 p += len;
1226 } 1231 }
1227 1232
1228 bio_free_map_data(bmd); 1233 bio_free_map_data(bmd);
1229 bio_put(bio); 1234 bio_put(bio);
1230 } 1235 }
1231 1236
1232 /** 1237 /**
1233 * bio_copy_kern - copy kernel address into bio 1238 * bio_copy_kern - copy kernel address into bio
1234 * @q: the struct request_queue for the bio 1239 * @q: the struct request_queue for the bio
1235 * @data: pointer to buffer to copy 1240 * @data: pointer to buffer to copy
1236 * @len: length in bytes 1241 * @len: length in bytes
1237 * @gfp_mask: allocation flags for bio and page allocation 1242 * @gfp_mask: allocation flags for bio and page allocation
1238 * @reading: data direction is READ 1243 * @reading: data direction is READ
1239 * 1244 *
1240 * copy the kernel address into a bio suitable for io to a block 1245 * copy the kernel address into a bio suitable for io to a block
1241 * device. Returns an error pointer in case of error. 1246 * device. Returns an error pointer in case of error.
1242 */ 1247 */
1243 struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len, 1248 struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len,
1244 gfp_t gfp_mask, int reading) 1249 gfp_t gfp_mask, int reading)
1245 { 1250 {
1246 struct bio *bio; 1251 struct bio *bio;
1247 struct bio_vec *bvec; 1252 struct bio_vec *bvec;
1248 int i; 1253 int i;
1249 1254
1250 bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask); 1255 bio = bio_copy_user(q, NULL, (unsigned long)data, len, 1, gfp_mask);
1251 if (IS_ERR(bio)) 1256 if (IS_ERR(bio))
1252 return bio; 1257 return bio;
1253 1258
1254 if (!reading) { 1259 if (!reading) {
1255 void *p = data; 1260 void *p = data;
1256 1261
1257 bio_for_each_segment(bvec, bio, i) { 1262 bio_for_each_segment(bvec, bio, i) {
1258 char *addr = page_address(bvec->bv_page); 1263 char *addr = page_address(bvec->bv_page);
1259 1264
1260 memcpy(addr, p, bvec->bv_len); 1265 memcpy(addr, p, bvec->bv_len);
1261 p += bvec->bv_len; 1266 p += bvec->bv_len;
1262 } 1267 }
1263 } 1268 }
1264 1269
1265 bio->bi_end_io = bio_copy_kern_endio; 1270 bio->bi_end_io = bio_copy_kern_endio;
1266 1271
1267 return bio; 1272 return bio;
1268 } 1273 }
1269 EXPORT_SYMBOL(bio_copy_kern); 1274 EXPORT_SYMBOL(bio_copy_kern);
1270 1275
1271 /* 1276 /*
1272 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions 1277 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1273 * for performing direct-IO in BIOs. 1278 * for performing direct-IO in BIOs.
1274 * 1279 *
1275 * The problem is that we cannot run set_page_dirty() from interrupt context 1280 * The problem is that we cannot run set_page_dirty() from interrupt context
1276 * because the required locks are not interrupt-safe. So what we can do is to 1281 * because the required locks are not interrupt-safe. So what we can do is to
1277 * mark the pages dirty _before_ performing IO. And in interrupt context, 1282 * mark the pages dirty _before_ performing IO. And in interrupt context,
1278 * check that the pages are still dirty. If so, fine. If not, redirty them 1283 * check that the pages are still dirty. If so, fine. If not, redirty them
1279 * in process context. 1284 * in process context.
1280 * 1285 *
1281 * We special-case compound pages here: normally this means reads into hugetlb 1286 * We special-case compound pages here: normally this means reads into hugetlb
1282 * pages. The logic in here doesn't really work right for compound pages 1287 * pages. The logic in here doesn't really work right for compound pages
1283 * because the VM does not uniformly chase down the head page in all cases. 1288 * because the VM does not uniformly chase down the head page in all cases.
1284 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't 1289 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1285 * handle them at all. So we skip compound pages here at an early stage. 1290 * handle them at all. So we skip compound pages here at an early stage.
1286 * 1291 *
1287 * Note that this code is very hard to test under normal circumstances because 1292 * Note that this code is very hard to test under normal circumstances because
1288 * direct-io pins the pages with get_user_pages(). This makes 1293 * direct-io pins the pages with get_user_pages(). This makes
1289 * is_page_cache_freeable return false, and the VM will not clean the pages. 1294 * is_page_cache_freeable return false, and the VM will not clean the pages.
1290 * But other code (eg, pdflush) could clean the pages if they are mapped 1295 * But other code (eg, pdflush) could clean the pages if they are mapped
1291 * pagecache. 1296 * pagecache.
1292 * 1297 *
1293 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the 1298 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1294 * deferred bio dirtying paths. 1299 * deferred bio dirtying paths.
1295 */ 1300 */
1296 1301
1297 /* 1302 /*
1298 * bio_set_pages_dirty() will mark all the bio's pages as dirty. 1303 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1299 */ 1304 */
1300 void bio_set_pages_dirty(struct bio *bio) 1305 void bio_set_pages_dirty(struct bio *bio)
1301 { 1306 {
1302 struct bio_vec *bvec = bio->bi_io_vec; 1307 struct bio_vec *bvec = bio->bi_io_vec;
1303 int i; 1308 int i;
1304 1309
1305 for (i = 0; i < bio->bi_vcnt; i++) { 1310 for (i = 0; i < bio->bi_vcnt; i++) {
1306 struct page *page = bvec[i].bv_page; 1311 struct page *page = bvec[i].bv_page;
1307 1312
1308 if (page && !PageCompound(page)) 1313 if (page && !PageCompound(page))
1309 set_page_dirty_lock(page); 1314 set_page_dirty_lock(page);
1310 } 1315 }
1311 } 1316 }
1312 1317
1313 static void bio_release_pages(struct bio *bio) 1318 static void bio_release_pages(struct bio *bio)
1314 { 1319 {
1315 struct bio_vec *bvec = bio->bi_io_vec; 1320 struct bio_vec *bvec = bio->bi_io_vec;
1316 int i; 1321 int i;
1317 1322
1318 for (i = 0; i < bio->bi_vcnt; i++) { 1323 for (i = 0; i < bio->bi_vcnt; i++) {
1319 struct page *page = bvec[i].bv_page; 1324 struct page *page = bvec[i].bv_page;
1320 1325
1321 if (page) 1326 if (page)
1322 put_page(page); 1327 put_page(page);
1323 } 1328 }
1324 } 1329 }
1325 1330
1326 /* 1331 /*
1327 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty. 1332 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1328 * If they are, then fine. If, however, some pages are clean then they must 1333 * If they are, then fine. If, however, some pages are clean then they must
1329 * have been written out during the direct-IO read. So we take another ref on 1334 * have been written out during the direct-IO read. So we take another ref on
1330 * the BIO and the offending pages and re-dirty the pages in process context. 1335 * the BIO and the offending pages and re-dirty the pages in process context.
1331 * 1336 *
1332 * It is expected that bio_check_pages_dirty() will wholly own the BIO from 1337 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1333 * here on. It will run one page_cache_release() against each page and will 1338 * here on. It will run one page_cache_release() against each page and will
1334 * run one bio_put() against the BIO. 1339 * run one bio_put() against the BIO.
1335 */ 1340 */
1336 1341
1337 static void bio_dirty_fn(struct work_struct *work); 1342 static void bio_dirty_fn(struct work_struct *work);
1338 1343
1339 static DECLARE_WORK(bio_dirty_work, bio_dirty_fn); 1344 static DECLARE_WORK(bio_dirty_work, bio_dirty_fn);
1340 static DEFINE_SPINLOCK(bio_dirty_lock); 1345 static DEFINE_SPINLOCK(bio_dirty_lock);
1341 static struct bio *bio_dirty_list; 1346 static struct bio *bio_dirty_list;
1342 1347
1343 /* 1348 /*
1344 * This runs in process context 1349 * This runs in process context
1345 */ 1350 */
1346 static void bio_dirty_fn(struct work_struct *work) 1351 static void bio_dirty_fn(struct work_struct *work)
1347 { 1352 {
1348 unsigned long flags; 1353 unsigned long flags;
1349 struct bio *bio; 1354 struct bio *bio;
1350 1355
1351 spin_lock_irqsave(&bio_dirty_lock, flags); 1356 spin_lock_irqsave(&bio_dirty_lock, flags);
1352 bio = bio_dirty_list; 1357 bio = bio_dirty_list;
1353 bio_dirty_list = NULL; 1358 bio_dirty_list = NULL;
1354 spin_unlock_irqrestore(&bio_dirty_lock, flags); 1359 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1355 1360
1356 while (bio) { 1361 while (bio) {
1357 struct bio *next = bio->bi_private; 1362 struct bio *next = bio->bi_private;
1358 1363
1359 bio_set_pages_dirty(bio); 1364 bio_set_pages_dirty(bio);
1360 bio_release_pages(bio); 1365 bio_release_pages(bio);
1361 bio_put(bio); 1366 bio_put(bio);
1362 bio = next; 1367 bio = next;
1363 } 1368 }
1364 } 1369 }
1365 1370
1366 void bio_check_pages_dirty(struct bio *bio) 1371 void bio_check_pages_dirty(struct bio *bio)
1367 { 1372 {
1368 struct bio_vec *bvec = bio->bi_io_vec; 1373 struct bio_vec *bvec = bio->bi_io_vec;
1369 int nr_clean_pages = 0; 1374 int nr_clean_pages = 0;
1370 int i; 1375 int i;
1371 1376
1372 for (i = 0; i < bio->bi_vcnt; i++) { 1377 for (i = 0; i < bio->bi_vcnt; i++) {
1373 struct page *page = bvec[i].bv_page; 1378 struct page *page = bvec[i].bv_page;
1374 1379
1375 if (PageDirty(page) || PageCompound(page)) { 1380 if (PageDirty(page) || PageCompound(page)) {
1376 page_cache_release(page); 1381 page_cache_release(page);
1377 bvec[i].bv_page = NULL; 1382 bvec[i].bv_page = NULL;
1378 } else { 1383 } else {
1379 nr_clean_pages++; 1384 nr_clean_pages++;
1380 } 1385 }
1381 } 1386 }
1382 1387
1383 if (nr_clean_pages) { 1388 if (nr_clean_pages) {
1384 unsigned long flags; 1389 unsigned long flags;
1385 1390
1386 spin_lock_irqsave(&bio_dirty_lock, flags); 1391 spin_lock_irqsave(&bio_dirty_lock, flags);
1387 bio->bi_private = bio_dirty_list; 1392 bio->bi_private = bio_dirty_list;
1388 bio_dirty_list = bio; 1393 bio_dirty_list = bio;
1389 spin_unlock_irqrestore(&bio_dirty_lock, flags); 1394 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1390 schedule_work(&bio_dirty_work); 1395 schedule_work(&bio_dirty_work);
1391 } else { 1396 } else {
1392 bio_put(bio); 1397 bio_put(bio);
1393 } 1398 }
1394 } 1399 }
1395 1400
1396 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1401 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1397 void bio_flush_dcache_pages(struct bio *bi) 1402 void bio_flush_dcache_pages(struct bio *bi)
1398 { 1403 {
1399 int i; 1404 int i;
1400 struct bio_vec *bvec; 1405 struct bio_vec *bvec;
1401 1406
1402 bio_for_each_segment(bvec, bi, i) 1407 bio_for_each_segment(bvec, bi, i)
1403 flush_dcache_page(bvec->bv_page); 1408 flush_dcache_page(bvec->bv_page);
1404 } 1409 }
1405 EXPORT_SYMBOL(bio_flush_dcache_pages); 1410 EXPORT_SYMBOL(bio_flush_dcache_pages);
1406 #endif 1411 #endif
1407 1412
1408 /** 1413 /**
1409 * bio_endio - end I/O on a bio 1414 * bio_endio - end I/O on a bio
1410 * @bio: bio 1415 * @bio: bio
1411 * @error: error, if any 1416 * @error: error, if any
1412 * 1417 *
1413 * Description: 1418 * Description:
1414 * bio_endio() will end I/O on the whole bio. bio_endio() is the 1419 * bio_endio() will end I/O on the whole bio. bio_endio() is the
1415 * preferred way to end I/O on a bio, it takes care of clearing 1420 * preferred way to end I/O on a bio, it takes care of clearing
1416 * BIO_UPTODATE on error. @error is 0 on success, and and one of the 1421 * BIO_UPTODATE on error. @error is 0 on success, and and one of the
1417 * established -Exxxx (-EIO, for instance) error values in case 1422 * established -Exxxx (-EIO, for instance) error values in case
1418 * something went wrong. Noone should call bi_end_io() directly on a 1423 * something went wrong. Noone should call bi_end_io() directly on a
1419 * bio unless they own it and thus know that it has an end_io 1424 * bio unless they own it and thus know that it has an end_io
1420 * function. 1425 * function.
1421 **/ 1426 **/
1422 void bio_endio(struct bio *bio, int error) 1427 void bio_endio(struct bio *bio, int error)
1423 { 1428 {
1424 if (error) 1429 if (error)
1425 clear_bit(BIO_UPTODATE, &bio->bi_flags); 1430 clear_bit(BIO_UPTODATE, &bio->bi_flags);
1426 else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) 1431 else if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
1427 error = -EIO; 1432 error = -EIO;
1428 1433
1429 if (bio->bi_end_io) 1434 if (bio->bi_end_io)
1430 bio->bi_end_io(bio, error); 1435 bio->bi_end_io(bio, error);
1431 } 1436 }
1432 EXPORT_SYMBOL(bio_endio); 1437 EXPORT_SYMBOL(bio_endio);
1433 1438
1434 void bio_pair_release(struct bio_pair *bp) 1439 void bio_pair_release(struct bio_pair *bp)
1435 { 1440 {
1436 if (atomic_dec_and_test(&bp->cnt)) { 1441 if (atomic_dec_and_test(&bp->cnt)) {
1437 struct bio *master = bp->bio1.bi_private; 1442 struct bio *master = bp->bio1.bi_private;
1438 1443
1439 bio_endio(master, bp->error); 1444 bio_endio(master, bp->error);
1440 mempool_free(bp, bp->bio2.bi_private); 1445 mempool_free(bp, bp->bio2.bi_private);
1441 } 1446 }
1442 } 1447 }
1443 EXPORT_SYMBOL(bio_pair_release); 1448 EXPORT_SYMBOL(bio_pair_release);
1444 1449
1445 static void bio_pair_end_1(struct bio *bi, int err) 1450 static void bio_pair_end_1(struct bio *bi, int err)
1446 { 1451 {
1447 struct bio_pair *bp = container_of(bi, struct bio_pair, bio1); 1452 struct bio_pair *bp = container_of(bi, struct bio_pair, bio1);
1448 1453
1449 if (err) 1454 if (err)
1450 bp->error = err; 1455 bp->error = err;
1451 1456
1452 bio_pair_release(bp); 1457 bio_pair_release(bp);
1453 } 1458 }
1454 1459
1455 static void bio_pair_end_2(struct bio *bi, int err) 1460 static void bio_pair_end_2(struct bio *bi, int err)
1456 { 1461 {
1457 struct bio_pair *bp = container_of(bi, struct bio_pair, bio2); 1462 struct bio_pair *bp = container_of(bi, struct bio_pair, bio2);
1458 1463
1459 if (err) 1464 if (err)
1460 bp->error = err; 1465 bp->error = err;
1461 1466
1462 bio_pair_release(bp); 1467 bio_pair_release(bp);
1463 } 1468 }
1464 1469
1465 /* 1470 /*
1466 * split a bio - only worry about a bio with a single page in its iovec 1471 * split a bio - only worry about a bio with a single page in its iovec
1467 */ 1472 */
1468 struct bio_pair *bio_split(struct bio *bi, int first_sectors) 1473 struct bio_pair *bio_split(struct bio *bi, int first_sectors)
1469 { 1474 {
1470 struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO); 1475 struct bio_pair *bp = mempool_alloc(bio_split_pool, GFP_NOIO);
1471 1476
1472 if (!bp) 1477 if (!bp)
1473 return bp; 1478 return bp;
1474 1479
1475 trace_block_split(bdev_get_queue(bi->bi_bdev), bi, 1480 trace_block_split(bdev_get_queue(bi->bi_bdev), bi,
1476 bi->bi_sector + first_sectors); 1481 bi->bi_sector + first_sectors);
1477 1482
1478 BUG_ON(bi->bi_vcnt != 1); 1483 BUG_ON(bi->bi_vcnt != 1);
1479 BUG_ON(bi->bi_idx != 0); 1484 BUG_ON(bi->bi_idx != 0);
1480 atomic_set(&bp->cnt, 3); 1485 atomic_set(&bp->cnt, 3);
1481 bp->error = 0; 1486 bp->error = 0;
1482 bp->bio1 = *bi; 1487 bp->bio1 = *bi;
1483 bp->bio2 = *bi; 1488 bp->bio2 = *bi;
1484 bp->bio2.bi_sector += first_sectors; 1489 bp->bio2.bi_sector += first_sectors;
1485 bp->bio2.bi_size -= first_sectors << 9; 1490 bp->bio2.bi_size -= first_sectors << 9;
1486 bp->bio1.bi_size = first_sectors << 9; 1491 bp->bio1.bi_size = first_sectors << 9;
1487 1492
1488 bp->bv1 = bi->bi_io_vec[0]; 1493 bp->bv1 = bi->bi_io_vec[0];
1489 bp->bv2 = bi->bi_io_vec[0]; 1494 bp->bv2 = bi->bi_io_vec[0];
1490 bp->bv2.bv_offset += first_sectors << 9; 1495 bp->bv2.bv_offset += first_sectors << 9;
1491 bp->bv2.bv_len -= first_sectors << 9; 1496 bp->bv2.bv_len -= first_sectors << 9;
1492 bp->bv1.bv_len = first_sectors << 9; 1497 bp->bv1.bv_len = first_sectors << 9;
1493 1498
1494 bp->bio1.bi_io_vec = &bp->bv1; 1499 bp->bio1.bi_io_vec = &bp->bv1;
1495 bp->bio2.bi_io_vec = &bp->bv2; 1500 bp->bio2.bi_io_vec = &bp->bv2;
1496 1501
1497 bp->bio1.bi_max_vecs = 1; 1502 bp->bio1.bi_max_vecs = 1;
1498 bp->bio2.bi_max_vecs = 1; 1503 bp->bio2.bi_max_vecs = 1;
1499 1504
1500 bp->bio1.bi_end_io = bio_pair_end_1; 1505 bp->bio1.bi_end_io = bio_pair_end_1;
1501 bp->bio2.bi_end_io = bio_pair_end_2; 1506 bp->bio2.bi_end_io = bio_pair_end_2;
1502 1507
1503 bp->bio1.bi_private = bi; 1508 bp->bio1.bi_private = bi;
1504 bp->bio2.bi_private = bio_split_pool; 1509 bp->bio2.bi_private = bio_split_pool;
1505 1510
1506 if (bio_integrity(bi)) 1511 if (bio_integrity(bi))
1507 bio_integrity_split(bi, bp, first_sectors); 1512 bio_integrity_split(bi, bp, first_sectors);
1508 1513
1509 return bp; 1514 return bp;
1510 } 1515 }
1511 EXPORT_SYMBOL(bio_split); 1516 EXPORT_SYMBOL(bio_split);
1512 1517
1513 /** 1518 /**
1514 * bio_sector_offset - Find hardware sector offset in bio 1519 * bio_sector_offset - Find hardware sector offset in bio
1515 * @bio: bio to inspect 1520 * @bio: bio to inspect
1516 * @index: bio_vec index 1521 * @index: bio_vec index
1517 * @offset: offset in bv_page 1522 * @offset: offset in bv_page
1518 * 1523 *
1519 * Return the number of hardware sectors between beginning of bio 1524 * Return the number of hardware sectors between beginning of bio
1520 * and an end point indicated by a bio_vec index and an offset 1525 * and an end point indicated by a bio_vec index and an offset
1521 * within that vector's page. 1526 * within that vector's page.
1522 */ 1527 */
1523 sector_t bio_sector_offset(struct bio *bio, unsigned short index, 1528 sector_t bio_sector_offset(struct bio *bio, unsigned short index,
1524 unsigned int offset) 1529 unsigned int offset)
1525 { 1530 {
1526 unsigned int sector_sz; 1531 unsigned int sector_sz;
1527 struct bio_vec *bv; 1532 struct bio_vec *bv;
1528 sector_t sectors; 1533 sector_t sectors;
1529 int i; 1534 int i;
1530 1535
1531 sector_sz = queue_logical_block_size(bio->bi_bdev->bd_disk->queue); 1536 sector_sz = queue_logical_block_size(bio->bi_bdev->bd_disk->queue);
1532 sectors = 0; 1537 sectors = 0;
1533 1538
1534 if (index >= bio->bi_idx) 1539 if (index >= bio->bi_idx)
1535 index = bio->bi_vcnt - 1; 1540 index = bio->bi_vcnt - 1;
1536 1541
1537 __bio_for_each_segment(bv, bio, i, 0) { 1542 __bio_for_each_segment(bv, bio, i, 0) {
1538 if (i == index) { 1543 if (i == index) {
1539 if (offset > bv->bv_offset) 1544 if (offset > bv->bv_offset)
1540 sectors += (offset - bv->bv_offset) / sector_sz; 1545 sectors += (offset - bv->bv_offset) / sector_sz;
1541 break; 1546 break;
1542 } 1547 }
1543 1548
1544 sectors += bv->bv_len / sector_sz; 1549 sectors += bv->bv_len / sector_sz;
1545 } 1550 }
1546 1551
1547 return sectors; 1552 return sectors;
1548 } 1553 }
1549 EXPORT_SYMBOL(bio_sector_offset); 1554 EXPORT_SYMBOL(bio_sector_offset);
1550 1555
1551 /* 1556 /*
1552 * create memory pools for biovec's in a bio_set. 1557 * create memory pools for biovec's in a bio_set.
1553 * use the global biovec slabs created for general use. 1558 * use the global biovec slabs created for general use.
1554 */ 1559 */
1555 static int biovec_create_pools(struct bio_set *bs, int pool_entries) 1560 static int biovec_create_pools(struct bio_set *bs, int pool_entries)
1556 { 1561 {
1557 struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX; 1562 struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX;
1558 1563
1559 bs->bvec_pool = mempool_create_slab_pool(pool_entries, bp->slab); 1564 bs->bvec_pool = mempool_create_slab_pool(pool_entries, bp->slab);
1560 if (!bs->bvec_pool) 1565 if (!bs->bvec_pool)
1561 return -ENOMEM; 1566 return -ENOMEM;
1562 1567
1563 return 0; 1568 return 0;
1564 } 1569 }
1565 1570
1566 static void biovec_free_pools(struct bio_set *bs) 1571 static void biovec_free_pools(struct bio_set *bs)
1567 { 1572 {
1568 mempool_destroy(bs->bvec_pool); 1573 mempool_destroy(bs->bvec_pool);
1569 } 1574 }
1570 1575
1571 void bioset_free(struct bio_set *bs) 1576 void bioset_free(struct bio_set *bs)
1572 { 1577 {
1573 if (bs->bio_pool) 1578 if (bs->bio_pool)
1574 mempool_destroy(bs->bio_pool); 1579 mempool_destroy(bs->bio_pool);
1575 1580
1576 bioset_integrity_free(bs); 1581 bioset_integrity_free(bs);
1577 biovec_free_pools(bs); 1582 biovec_free_pools(bs);
1578 bio_put_slab(bs); 1583 bio_put_slab(bs);
1579 1584
1580 kfree(bs); 1585 kfree(bs);
1581 } 1586 }
1582 EXPORT_SYMBOL(bioset_free); 1587 EXPORT_SYMBOL(bioset_free);
1583 1588
1584 /** 1589 /**
1585 * bioset_create - Create a bio_set 1590 * bioset_create - Create a bio_set
1586 * @pool_size: Number of bio and bio_vecs to cache in the mempool 1591 * @pool_size: Number of bio and bio_vecs to cache in the mempool
1587 * @front_pad: Number of bytes to allocate in front of the returned bio 1592 * @front_pad: Number of bytes to allocate in front of the returned bio
1588 * 1593 *
1589 * Description: 1594 * Description:
1590 * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller 1595 * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller
1591 * to ask for a number of bytes to be allocated in front of the bio. 1596 * to ask for a number of bytes to be allocated in front of the bio.
1592 * Front pad allocation is useful for embedding the bio inside 1597 * Front pad allocation is useful for embedding the bio inside
1593 * another structure, to avoid allocating extra data to go with the bio. 1598 * another structure, to avoid allocating extra data to go with the bio.
1594 * Note that the bio must be embedded at the END of that structure always, 1599 * Note that the bio must be embedded at the END of that structure always,
1595 * or things will break badly. 1600 * or things will break badly.
1596 */ 1601 */
1597 struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad) 1602 struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad)
1598 { 1603 {
1599 unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec); 1604 unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec);
1600 struct bio_set *bs; 1605 struct bio_set *bs;
1601 1606
1602 bs = kzalloc(sizeof(*bs), GFP_KERNEL); 1607 bs = kzalloc(sizeof(*bs), GFP_KERNEL);
1603 if (!bs) 1608 if (!bs)
1604 return NULL; 1609 return NULL;
1605 1610
1606 bs->front_pad = front_pad; 1611 bs->front_pad = front_pad;
1607 1612
1608 bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad); 1613 bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad);
1609 if (!bs->bio_slab) { 1614 if (!bs->bio_slab) {
1610 kfree(bs); 1615 kfree(bs);
1611 return NULL; 1616 return NULL;
1612 } 1617 }
1613 1618
1614 bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab); 1619 bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab);
1615 if (!bs->bio_pool) 1620 if (!bs->bio_pool)
1616 goto bad; 1621 goto bad;
1617 1622
1618 if (bioset_integrity_create(bs, pool_size)) 1623 if (bioset_integrity_create(bs, pool_size))
1619 goto bad; 1624 goto bad;
1620 1625
1621 if (!biovec_create_pools(bs, pool_size)) 1626 if (!biovec_create_pools(bs, pool_size))
1622 return bs; 1627 return bs;
1623 1628
1624 bad: 1629 bad:
1625 bioset_free(bs); 1630 bioset_free(bs);
1626 return NULL; 1631 return NULL;
1627 } 1632 }
1628 EXPORT_SYMBOL(bioset_create); 1633 EXPORT_SYMBOL(bioset_create);
1629 1634
1630 static void __init biovec_init_slabs(void) 1635 static void __init biovec_init_slabs(void)
1631 { 1636 {
1632 int i; 1637 int i;
1633 1638
1634 for (i = 0; i < BIOVEC_NR_POOLS; i++) { 1639 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1635 int size; 1640 int size;
1636 struct biovec_slab *bvs = bvec_slabs + i; 1641 struct biovec_slab *bvs = bvec_slabs + i;
1637 1642
1638 #ifndef CONFIG_BLK_DEV_INTEGRITY 1643 #ifndef CONFIG_BLK_DEV_INTEGRITY
1639 if (bvs->nr_vecs <= BIO_INLINE_VECS) { 1644 if (bvs->nr_vecs <= BIO_INLINE_VECS) {
1640 bvs->slab = NULL; 1645 bvs->slab = NULL;
1641 continue; 1646 continue;
1642 } 1647 }
1643 #endif 1648 #endif
1644 1649
1645 size = bvs->nr_vecs * sizeof(struct bio_vec); 1650 size = bvs->nr_vecs * sizeof(struct bio_vec);
1646 bvs->slab = kmem_cache_create(bvs->name, size, 0, 1651 bvs->slab = kmem_cache_create(bvs->name, size, 0,
1647 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL); 1652 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
1648 } 1653 }
1649 } 1654 }
1650 1655
1651 static int __init init_bio(void) 1656 static int __init init_bio(void)
1652 { 1657 {
1653 bio_slab_max = 2; 1658 bio_slab_max = 2;
1654 bio_slab_nr = 0; 1659 bio_slab_nr = 0;
1655 bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL); 1660 bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL);
1656 if (!bio_slabs) 1661 if (!bio_slabs)
1657 panic("bio: can't allocate bios\n"); 1662 panic("bio: can't allocate bios\n");
1658 1663
1659 bio_integrity_init(); 1664 bio_integrity_init();
1660 biovec_init_slabs(); 1665 biovec_init_slabs();
1661 1666
1662 fs_bio_set = bioset_create(BIO_POOL_SIZE, 0); 1667 fs_bio_set = bioset_create(BIO_POOL_SIZE, 0);
1663 if (!fs_bio_set) 1668 if (!fs_bio_set)
1664 panic("bio: can't allocate bios\n"); 1669 panic("bio: can't allocate bios\n");
1665 1670
1666 bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES, 1671 bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES,
1667 sizeof(struct bio_pair)); 1672 sizeof(struct bio_pair));
1668 if (!bio_split_pool) 1673 if (!bio_split_pool)
1669 panic("bio: can't create split pool\n"); 1674 panic("bio: can't create split pool\n");
1670 1675
1671 return 0; 1676 return 0;
1672 } 1677 }
1673 subsys_initcall(init_bio); 1678 subsys_initcall(init_bio);
1674 1679