Eric Lee / linux-smarc-t335x-v3.2

Download as
Browse Code ยป

Commit d3f761104b097738932afcc310fbbbbfb007ef92

Authored by Jens Axboe
1 parent f735b5eeb9
Exists in master and in 4 other branches v3.2_AM335xPSP_04.06.00.11, v3.2_NEWT335xPSP_04.06.00.11, v3.2_SBC_SMARTMEN, v3.2_SMARCT335xPSP_04.06.00.11

bio: get rid of bio_vec clearing 

We don't need to clear the memory used for adding bio_vec entries,
since nobody should be looking at members unitialized. Any valid
use should be below bio->bi_vcnt, and that members up until that count
must be valid since they were added through bio_add_page().

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

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