Btrfs: still corruption hunting
[deliverable/linux.git] / fs / btrfs / disk-io.c
CommitLineData
e20d96d6
CM
1#include <linux/module.h>
2#include <linux/fs.h>
d98237b3 3#include <linux/blkdev.h>
87cbda5c
CM
4#include <linux/crypto.h>
5#include <linux/scatterlist.h>
22b0ebda 6#include <linux/swap.h>
eb60ceac
CM
7#include "ctree.h"
8#include "disk-io.h"
e089f05c 9#include "transaction.h"
eb60ceac 10
e20d96d6 11static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
eb60ceac 12{
e20d96d6 13 struct btrfs_node *node = btrfs_buffer_node(buf);
d98237b3 14 if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
9a8dd150 15 BUG();
d98237b3 16 }
e20d96d6 17 if (root->node && btrfs_header_parentid(&node->header) !=
df2ce34c 18 btrfs_header_parentid(btrfs_buffer_header(root->node))) {
7f5c1516 19 BUG();
df2ce34c 20 }
9a8dd150 21 return 0;
eb60ceac
CM
22}
23
d98237b3
CM
24struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
25{
2c90e5d6
CM
26 return sb_find_get_block(root->fs_info->sb, blocknr);
27#if 0
d98237b3
CM
28 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
29 int blockbits = root->fs_info->sb->s_blocksize_bits;
30 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
31 struct page *page;
32 struct buffer_head *bh;
33 struct buffer_head *head;
34 struct buffer_head *ret = NULL;
35
2c90e5d6 36
d98237b3
CM
37 page = find_lock_page(mapping, index);
38 if (!page)
39 return NULL;
40
41 if (!page_has_buffers(page))
42 goto out_unlock;
43
44 head = page_buffers(page);
45 bh = head;
46 do {
47 if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
48 ret = bh;
49 get_bh(bh);
50 goto out_unlock;
51 }
52 bh = bh->b_this_page;
53 } while (bh != head);
54out_unlock:
55 unlock_page(page);
d6025579 56 if (ret) {
22b0ebda 57 touch_buffer(ret);
d6025579 58 }
d98237b3
CM
59 page_cache_release(page);
60 return ret;
2c90e5d6 61#endif
d98237b3
CM
62}
63
64struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
65 u64 blocknr)
66{
2c90e5d6
CM
67 return sb_getblk(root->fs_info->sb, blocknr);
68#if 0
d98237b3
CM
69 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
70 int blockbits = root->fs_info->sb->s_blocksize_bits;
71 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
72 struct page *page;
73 struct buffer_head *bh;
74 struct buffer_head *head;
75 struct buffer_head *ret = NULL;
76 u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
22b0ebda 77
d98237b3
CM
78 page = grab_cache_page(mapping, index);
79 if (!page)
80 return NULL;
81
d98237b3
CM
82 if (!page_has_buffers(page))
83 create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
84 head = page_buffers(page);
85 bh = head;
86 do {
87 if (!buffer_mapped(bh)) {
88 bh->b_bdev = root->fs_info->sb->s_bdev;
89 bh->b_blocknr = first_block;
90 set_buffer_mapped(bh);
91 }
92 if (bh->b_blocknr == blocknr) {
93 ret = bh;
94 get_bh(bh);
95 goto out_unlock;
96 }
97 bh = bh->b_this_page;
98 first_block++;
99 } while (bh != head);
100out_unlock:
101 unlock_page(page);
22b0ebda
CM
102 if (ret)
103 touch_buffer(ret);
d98237b3
CM
104 page_cache_release(page);
105 return ret;
2c90e5d6 106#endif
d98237b3
CM
107}
108
109static sector_t max_block(struct block_device *bdev)
110{
111 sector_t retval = ~((sector_t)0);
112 loff_t sz = i_size_read(bdev->bd_inode);
113
114 if (sz) {
115 unsigned int size = block_size(bdev);
116 unsigned int sizebits = blksize_bits(size);
117 retval = (sz >> sizebits);
118 }
119 return retval;
120}
121
122static int btree_get_block(struct inode *inode, sector_t iblock,
123 struct buffer_head *bh, int create)
124{
125 if (iblock >= max_block(inode->i_sb->s_bdev)) {
126 if (create)
127 return -EIO;
128
129 /*
130 * for reads, we're just trying to fill a partial page.
131 * return a hole, they will have to call get_block again
132 * before they can fill it, and they will get -EIO at that
133 * time
134 */
135 return 0;
136 }
137 bh->b_bdev = inode->i_sb->s_bdev;
138 bh->b_blocknr = iblock;
139 set_buffer_mapped(bh);
140 return 0;
141}
142
f254e52c
CM
143int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
144 char *result)
87cbda5c 145{
87cbda5c
CM
146 struct scatterlist sg;
147 struct crypto_hash *tfm = root->fs_info->hash_tfm;
148 struct hash_desc desc;
149 int ret;
87cbda5c
CM
150
151 desc.tfm = tfm;
152 desc.flags = 0;
f254e52c 153 sg_init_one(&sg, data, len);
87cbda5c 154 spin_lock(&root->fs_info->hash_lock);
22b0ebda 155 ret = crypto_hash_digest(&desc, &sg, 1, result);
87cbda5c
CM
156 spin_unlock(&root->fs_info->hash_lock);
157 if (ret) {
158 printk("sha256 digest failed\n");
159 }
f254e52c
CM
160 return ret;
161}
162static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
163 int verify)
164{
165 char result[BTRFS_CSUM_SIZE];
166 int ret;
167 struct btrfs_node *node;
168
22b0ebda 169 return 0;
f254e52c
CM
170 ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
171 bh->b_size - BTRFS_CSUM_SIZE, result);
172 if (ret)
173 return ret;
87cbda5c 174 if (verify) {
f254e52c
CM
175 if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
176 printk("checksum verify failed on %lu\n",
177 bh->b_blocknr);
178 return 1;
179 }
180 } else {
181 node = btrfs_buffer_node(bh);
22b0ebda 182 memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
f254e52c 183 }
87cbda5c
CM
184 return 0;
185}
186
d98237b3 187static int btree_writepage(struct page *page, struct writeback_control *wbc)
ed2ff2cb 188{
22b0ebda 189#if 0
87cbda5c
CM
190 struct buffer_head *bh;
191 struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
192 struct buffer_head *head;
87cbda5c
CM
193 if (!page_has_buffers(page)) {
194 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
195 (1 << BH_Dirty)|(1 << BH_Uptodate));
196 }
197 head = page_buffers(page);
198 bh = head;
199 do {
200 if (buffer_dirty(bh))
201 csum_tree_block(root, bh, 0);
202 bh = bh->b_this_page;
203 } while (bh != head);
22b0ebda 204#endif
d98237b3 205 return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb
CM
206}
207
d98237b3 208static int btree_readpage(struct file * file, struct page * page)
eb60ceac 209{
d98237b3 210 return block_read_full_page(page, btree_get_block);
eb60ceac
CM
211}
212
d98237b3
CM
213static struct address_space_operations btree_aops = {
214 .readpage = btree_readpage,
215 .writepage = btree_writepage,
216 .sync_page = block_sync_page,
217};
218
e20d96d6 219struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
eb60ceac 220{
2c90e5d6
CM
221 return sb_bread(root->fs_info->sb, blocknr);
222#if 0
d98237b3 223 struct buffer_head *bh = NULL;
eb60ceac 224
d98237b3
CM
225 bh = btrfs_find_create_tree_block(root, blocknr);
226 if (!bh)
227 return bh;
228 lock_buffer(bh);
229 if (!buffer_uptodate(bh)) {
230 get_bh(bh);
231 bh->b_end_io = end_buffer_read_sync;
232 submit_bh(READ, bh);
233 wait_on_buffer(bh);
234 if (!buffer_uptodate(bh))
235 goto fail;
87cbda5c 236 csum_tree_block(root, bh, 1);
d98237b3
CM
237 } else {
238 unlock_buffer(bh);
239 }
240 if (check_tree_block(root, bh))
cfaa7295 241 BUG();
d98237b3
CM
242 return bh;
243fail:
244 brelse(bh);
245 return NULL;
246
2c90e5d6 247#endif
eb60ceac
CM
248}
249
e089f05c 250int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
e20d96d6 251 struct buffer_head *buf)
ed2ff2cb 252{
d6025579 253 WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6 254 mark_buffer_dirty(buf);
ed2ff2cb
CM
255 return 0;
256}
257
e089f05c 258int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
e20d96d6 259 struct buffer_head *buf)
ed2ff2cb 260{
d6025579 261 WARN_ON(atomic_read(&buf->b_count) == 0);
e20d96d6 262 clear_buffer_dirty(buf);
ed2ff2cb
CM
263 return 0;
264}
265
2c90e5d6 266static int __setup_root(int blocksize,
9f5fae2f
CM
267 struct btrfs_root *root,
268 struct btrfs_fs_info *fs_info,
e20d96d6 269 u64 objectid)
d97e63b6 270{
cfaa7295 271 root->node = NULL;
a28ec197 272 root->commit_root = NULL;
2c90e5d6 273 root->blocksize = blocksize;
123abc88 274 root->ref_cows = 0;
9f5fae2f 275 root->fs_info = fs_info;
3768f368
CM
276 memset(&root->root_key, 0, sizeof(root->root_key));
277 memset(&root->root_item, 0, sizeof(root->root_item));
278 return 0;
279}
280
2c90e5d6 281static int find_and_setup_root(int blocksize,
9f5fae2f
CM
282 struct btrfs_root *tree_root,
283 struct btrfs_fs_info *fs_info,
284 u64 objectid,
e20d96d6 285 struct btrfs_root *root)
3768f368
CM
286{
287 int ret;
288
2c90e5d6 289 __setup_root(blocksize, root, fs_info, objectid);
3768f368
CM
290 ret = btrfs_find_last_root(tree_root, objectid,
291 &root->root_item, &root->root_key);
292 BUG_ON(ret);
293
294 root->node = read_tree_block(root,
295 btrfs_root_blocknr(&root->root_item));
3768f368 296 BUG_ON(!root->node);
d97e63b6
CM
297 return 0;
298}
299
2c90e5d6 300struct btrfs_root *open_ctree(struct super_block *sb)
2e635a27 301{
e20d96d6
CM
302 struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
303 GFP_NOFS);
304 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
305 GFP_NOFS);
306 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
307 GFP_NOFS);
308 struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
309 GFP_NOFS);
310 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
311 GFP_NOFS);
eb60ceac 312 int ret;
2c90e5d6 313 struct btrfs_super_block *disk_super;
eb60ceac 314
8ef97622
CM
315 init_bit_radix(&fs_info->pinned_radix);
316 init_bit_radix(&fs_info->pending_del_radix);
2c90e5d6 317 sb_set_blocksize(sb, 4096);
9f5fae2f
CM
318 fs_info->running_transaction = NULL;
319 fs_info->fs_root = root;
320 fs_info->tree_root = tree_root;
321 fs_info->extent_root = extent_root;
322 fs_info->inode_root = inode_root;
323 fs_info->last_inode_alloc = 0;
324 fs_info->last_inode_alloc_dirid = 0;
e20d96d6 325 fs_info->sb = sb;
2c90e5d6
CM
326 fs_info->btree_inode = NULL;
327#if 0
d98237b3
CM
328 fs_info->btree_inode = new_inode(sb);
329 fs_info->btree_inode->i_ino = 1;
2c90e5d6 330 fs_info->btree_inode->i_nlink = 1;
d98237b3
CM
331 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
332 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
22b0ebda 333 insert_inode_hash(fs_info->btree_inode);
d98237b3 334 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
2c90e5d6 335#endif
87cbda5c 336 fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
30ae8467 337 spin_lock_init(&fs_info->hash_lock);
30ae8467 338 if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
87cbda5c
CM
339 printk("failed to allocate sha256 hash\n");
340 return NULL;
341 }
79154b1b 342 mutex_init(&fs_info->trans_mutex);
d561c025 343 mutex_init(&fs_info->fs_mutex);
9f5fae2f
CM
344 memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
345 memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
3768f368 346
2c90e5d6
CM
347 __setup_root(sb->s_blocksize, tree_root,
348 fs_info, BTRFS_ROOT_TREE_OBJECTID);
349 fs_info->sb_buffer = read_tree_block(tree_root,
350 BTRFS_SUPER_INFO_OFFSET /
351 sb->s_blocksize);
d98237b3 352
87cbda5c
CM
353 if (!fs_info->sb_buffer) {
354printk("failed2\n");
d98237b3 355 return NULL;
87cbda5c 356 }
d98237b3 357 disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
2c90e5d6
CM
358 if (!btrfs_super_root(disk_super)) {
359 return NULL;
360 }
d98237b3 361 fs_info->disk_super = disk_super;
e20d96d6
CM
362 tree_root->node = read_tree_block(tree_root,
363 btrfs_super_root(disk_super));
3768f368
CM
364 BUG_ON(!tree_root->node);
365
2c90e5d6
CM
366 mutex_lock(&fs_info->fs_mutex);
367 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6 368 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
3768f368
CM
369 BUG_ON(ret);
370
2c90e5d6 371 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6 372 BTRFS_INODE_MAP_OBJECTID, inode_root);
9f5fae2f
CM
373 BUG_ON(ret);
374
2c90e5d6 375 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
e20d96d6 376 BTRFS_FS_TREE_OBJECTID, root);
2c90e5d6 377 mutex_unlock(&fs_info->fs_mutex);
3768f368 378 BUG_ON(ret);
a28ec197 379 root->commit_root = root->node;
e20d96d6 380 get_bh(root->node);
3768f368 381 root->ref_cows = 1;
293ffd5f 382 root->fs_info->generation = root->root_key.offset + 1;
eb60ceac
CM
383 return root;
384}
385
e089f05c 386int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
79154b1b 387 *root)
eb60ceac 388{
d5719762 389 struct buffer_head *bh = root->fs_info->sb_buffer;
2c90e5d6 390
d5719762
CM
391 btrfs_set_super_root(root->fs_info->disk_super,
392 root->fs_info->tree_root->node->b_blocknr);
393 lock_buffer(bh);
2c90e5d6 394 WARN_ON(atomic_read(&bh->b_count) < 1);
d5719762 395 clear_buffer_dirty(bh);
87cbda5c 396 csum_tree_block(root, bh, 0);
d5719762
CM
397 bh->b_end_io = end_buffer_write_sync;
398 get_bh(bh);
399 submit_bh(WRITE, bh);
400 wait_on_buffer(bh);
401 if (!buffer_uptodate(bh)) {
402 WARN_ON(1);
403 return -EIO;
cfaa7295
CM
404 }
405 return 0;
406}
407
e20d96d6 408int close_ctree(struct btrfs_root *root)
cfaa7295 409{
3768f368 410 int ret;
e089f05c
CM
411 struct btrfs_trans_handle *trans;
412
2c90e5d6 413 mutex_lock(&root->fs_info->fs_mutex);
79154b1b
CM
414 trans = btrfs_start_transaction(root, 1);
415 btrfs_commit_transaction(trans, root);
416 /* run commit again to drop the original snapshot */
417 trans = btrfs_start_transaction(root, 1);
418 btrfs_commit_transaction(trans, root);
419 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 420 BUG_ON(ret);
79154b1b 421 write_ctree_super(NULL, root);
2c90e5d6 422 mutex_unlock(&root->fs_info->fs_mutex);
ed2ff2cb 423
cfaa7295 424 if (root->node)
234b63a0 425 btrfs_block_release(root, root->node);
9f5fae2f
CM
426 if (root->fs_info->extent_root->node)
427 btrfs_block_release(root->fs_info->extent_root,
428 root->fs_info->extent_root->node);
429 if (root->fs_info->inode_root->node)
430 btrfs_block_release(root->fs_info->inode_root,
431 root->fs_info->inode_root->node);
432 if (root->fs_info->tree_root->node)
433 btrfs_block_release(root->fs_info->tree_root,
434 root->fs_info->tree_root->node);
234b63a0 435 btrfs_block_release(root, root->commit_root);
e20d96d6 436 btrfs_block_release(root, root->fs_info->sb_buffer);
87cbda5c 437 crypto_free_hash(root->fs_info->hash_tfm);
2c90e5d6
CM
438 // truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
439 // iput(root->fs_info->btree_inode);
e20d96d6
CM
440 kfree(root->fs_info->extent_root);
441 kfree(root->fs_info->inode_root);
442 kfree(root->fs_info->tree_root);
443 kfree(root->fs_info);
444 kfree(root);
eb60ceac
CM
445 return 0;
446}
447
e20d96d6 448void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
eb60ceac 449{
2c90e5d6 450 // brelse(buf);
eb60ceac
CM
451}
452
This page took 0.050853 seconds and 5 git commands to generate.