Btrfs: hunting slab corruption
[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
d98237b3 11
e20d96d6 12static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
eb60ceac 13{
e20d96d6 14 struct btrfs_node *node = btrfs_buffer_node(buf);
d98237b3 15 if (buf->b_blocknr != btrfs_header_blocknr(&node->header)) {
9a8dd150 16 BUG();
d98237b3 17 }
e20d96d6 18 if (root->node && btrfs_header_parentid(&node->header) !=
df2ce34c 19 btrfs_header_parentid(btrfs_buffer_header(root->node))) {
7f5c1516 20 BUG();
df2ce34c 21 }
9a8dd150 22 return 0;
eb60ceac
CM
23}
24
d98237b3
CM
25struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
26{
27 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
28 int blockbits = root->fs_info->sb->s_blocksize_bits;
29 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
30 struct page *page;
31 struct buffer_head *bh;
32 struct buffer_head *head;
33 struct buffer_head *ret = NULL;
34
35 page = find_lock_page(mapping, index);
36 if (!page)
37 return NULL;
38
39 if (!page_has_buffers(page))
40 goto out_unlock;
41
42 head = page_buffers(page);
43 bh = head;
44 do {
45 if (buffer_mapped(bh) && bh->b_blocknr == blocknr) {
46 ret = bh;
47 get_bh(bh);
48 goto out_unlock;
49 }
50 bh = bh->b_this_page;
51 } while (bh != head);
52out_unlock:
53 unlock_page(page);
22b0ebda
CM
54 if (ret)
55 touch_buffer(ret);
d98237b3
CM
56 page_cache_release(page);
57 return ret;
58}
59
60struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
61 u64 blocknr)
62{
63 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
64 int blockbits = root->fs_info->sb->s_blocksize_bits;
65 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
66 struct page *page;
67 struct buffer_head *bh;
68 struct buffer_head *head;
69 struct buffer_head *ret = NULL;
70 u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
22b0ebda 71
d98237b3
CM
72 page = grab_cache_page(mapping, index);
73 if (!page)
74 return NULL;
75
d98237b3
CM
76 if (!page_has_buffers(page))
77 create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
78 head = page_buffers(page);
79 bh = head;
80 do {
81 if (!buffer_mapped(bh)) {
82 bh->b_bdev = root->fs_info->sb->s_bdev;
83 bh->b_blocknr = first_block;
84 set_buffer_mapped(bh);
85 }
86 if (bh->b_blocknr == blocknr) {
87 ret = bh;
88 get_bh(bh);
89 goto out_unlock;
90 }
91 bh = bh->b_this_page;
92 first_block++;
93 } while (bh != head);
94out_unlock:
95 unlock_page(page);
22b0ebda
CM
96 if (ret)
97 touch_buffer(ret);
d98237b3
CM
98 page_cache_release(page);
99 return ret;
100}
101
102static sector_t max_block(struct block_device *bdev)
103{
104 sector_t retval = ~((sector_t)0);
105 loff_t sz = i_size_read(bdev->bd_inode);
106
107 if (sz) {
108 unsigned int size = block_size(bdev);
109 unsigned int sizebits = blksize_bits(size);
110 retval = (sz >> sizebits);
111 }
112 return retval;
113}
114
115static int btree_get_block(struct inode *inode, sector_t iblock,
116 struct buffer_head *bh, int create)
117{
118 if (iblock >= max_block(inode->i_sb->s_bdev)) {
119 if (create)
120 return -EIO;
121
122 /*
123 * for reads, we're just trying to fill a partial page.
124 * return a hole, they will have to call get_block again
125 * before they can fill it, and they will get -EIO at that
126 * time
127 */
128 return 0;
129 }
130 bh->b_bdev = inode->i_sb->s_bdev;
131 bh->b_blocknr = iblock;
132 set_buffer_mapped(bh);
133 return 0;
134}
135
f254e52c
CM
136int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
137 char *result)
87cbda5c 138{
87cbda5c
CM
139 struct scatterlist sg;
140 struct crypto_hash *tfm = root->fs_info->hash_tfm;
141 struct hash_desc desc;
142 int ret;
87cbda5c
CM
143
144 desc.tfm = tfm;
145 desc.flags = 0;
f254e52c 146 sg_init_one(&sg, data, len);
87cbda5c 147 spin_lock(&root->fs_info->hash_lock);
22b0ebda 148 ret = crypto_hash_digest(&desc, &sg, 1, result);
87cbda5c
CM
149 spin_unlock(&root->fs_info->hash_lock);
150 if (ret) {
151 printk("sha256 digest failed\n");
152 }
f254e52c
CM
153 return ret;
154}
155static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
156 int verify)
157{
158 char result[BTRFS_CSUM_SIZE];
159 int ret;
160 struct btrfs_node *node;
161
22b0ebda 162 return 0;
f254e52c
CM
163 ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
164 bh->b_size - BTRFS_CSUM_SIZE, result);
165 if (ret)
166 return ret;
87cbda5c 167 if (verify) {
f254e52c
CM
168 if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
169 printk("checksum verify failed on %lu\n",
170 bh->b_blocknr);
171 return 1;
172 }
173 } else {
174 node = btrfs_buffer_node(bh);
22b0ebda 175 memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
f254e52c 176 }
87cbda5c
CM
177 return 0;
178}
179
d98237b3 180static int btree_writepage(struct page *page, struct writeback_control *wbc)
ed2ff2cb 181{
22b0ebda 182#if 0
87cbda5c
CM
183 struct buffer_head *bh;
184 struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
185 struct buffer_head *head;
87cbda5c
CM
186 if (!page_has_buffers(page)) {
187 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
188 (1 << BH_Dirty)|(1 << BH_Uptodate));
189 }
190 head = page_buffers(page);
191 bh = head;
192 do {
193 if (buffer_dirty(bh))
194 csum_tree_block(root, bh, 0);
195 bh = bh->b_this_page;
196 } while (bh != head);
22b0ebda 197#endif
d98237b3 198 return block_write_full_page(page, btree_get_block, wbc);
ed2ff2cb
CM
199}
200
d98237b3 201static int btree_readpage(struct file * file, struct page * page)
eb60ceac 202{
d98237b3 203 return block_read_full_page(page, btree_get_block);
eb60ceac
CM
204}
205
d98237b3
CM
206static struct address_space_operations btree_aops = {
207 .readpage = btree_readpage,
208 .writepage = btree_writepage,
209 .sync_page = block_sync_page,
210};
211
e20d96d6 212struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
eb60ceac 213{
d98237b3 214 struct buffer_head *bh = NULL;
eb60ceac 215
d98237b3
CM
216 bh = btrfs_find_create_tree_block(root, blocknr);
217 if (!bh)
218 return bh;
219 lock_buffer(bh);
220 if (!buffer_uptodate(bh)) {
221 get_bh(bh);
222 bh->b_end_io = end_buffer_read_sync;
223 submit_bh(READ, bh);
224 wait_on_buffer(bh);
225 if (!buffer_uptodate(bh))
226 goto fail;
87cbda5c 227 csum_tree_block(root, bh, 1);
d98237b3
CM
228 } else {
229 unlock_buffer(bh);
230 }
231 if (check_tree_block(root, bh))
cfaa7295 232 BUG();
d98237b3
CM
233 return bh;
234fail:
235 brelse(bh);
236 return NULL;
237
eb60ceac
CM
238}
239
e089f05c 240int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
e20d96d6 241 struct buffer_head *buf)
ed2ff2cb 242{
e20d96d6 243 mark_buffer_dirty(buf);
ed2ff2cb
CM
244 return 0;
245}
246
e089f05c 247int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
e20d96d6 248 struct buffer_head *buf)
ed2ff2cb 249{
e20d96d6 250 clear_buffer_dirty(buf);
ed2ff2cb
CM
251 return 0;
252}
253
123abc88 254static int __setup_root(struct btrfs_super_block *super,
9f5fae2f
CM
255 struct btrfs_root *root,
256 struct btrfs_fs_info *fs_info,
e20d96d6 257 u64 objectid)
d97e63b6 258{
cfaa7295 259 root->node = NULL;
a28ec197 260 root->commit_root = NULL;
123abc88
CM
261 root->blocksize = btrfs_super_blocksize(super);
262 root->ref_cows = 0;
9f5fae2f 263 root->fs_info = fs_info;
3768f368
CM
264 memset(&root->root_key, 0, sizeof(root->root_key));
265 memset(&root->root_item, 0, sizeof(root->root_item));
266 return 0;
267}
268
123abc88 269static int find_and_setup_root(struct btrfs_super_block *super,
9f5fae2f
CM
270 struct btrfs_root *tree_root,
271 struct btrfs_fs_info *fs_info,
272 u64 objectid,
e20d96d6 273 struct btrfs_root *root)
3768f368
CM
274{
275 int ret;
276
e20d96d6 277 __setup_root(super, root, fs_info, objectid);
3768f368
CM
278 ret = btrfs_find_last_root(tree_root, objectid,
279 &root->root_item, &root->root_key);
280 BUG_ON(ret);
281
282 root->node = read_tree_block(root,
283 btrfs_root_blocknr(&root->root_item));
3768f368 284 BUG_ON(!root->node);
d97e63b6
CM
285 return 0;
286}
287
e20d96d6
CM
288struct btrfs_root *open_ctree(struct super_block *sb,
289 struct buffer_head *sb_buffer,
290 struct btrfs_super_block *disk_super)
2e635a27 291{
e20d96d6
CM
292 struct btrfs_root *root = kmalloc(sizeof(struct btrfs_root),
293 GFP_NOFS);
294 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
295 GFP_NOFS);
296 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
297 GFP_NOFS);
298 struct btrfs_root *inode_root = kmalloc(sizeof(struct btrfs_root),
299 GFP_NOFS);
300 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
301 GFP_NOFS);
eb60ceac
CM
302 int ret;
303
87cbda5c 304 if (!btrfs_super_root(disk_super)) {
e20d96d6 305 return NULL;
87cbda5c 306 }
8ef97622
CM
307 init_bit_radix(&fs_info->pinned_radix);
308 init_bit_radix(&fs_info->pending_del_radix);
d98237b3 309 sb_set_blocksize(sb, sb_buffer->b_size);
9f5fae2f
CM
310 fs_info->running_transaction = NULL;
311 fs_info->fs_root = root;
312 fs_info->tree_root = tree_root;
313 fs_info->extent_root = extent_root;
314 fs_info->inode_root = inode_root;
315 fs_info->last_inode_alloc = 0;
316 fs_info->last_inode_alloc_dirid = 0;
e20d96d6 317 fs_info->disk_super = disk_super;
e20d96d6 318 fs_info->sb = sb;
d98237b3
CM
319 fs_info->btree_inode = new_inode(sb);
320 fs_info->btree_inode->i_ino = 1;
321 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
322 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
22b0ebda
CM
323 insert_inode_hash(fs_info->btree_inode);
324
d98237b3 325 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
87cbda5c 326 fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
30ae8467
CM
327 spin_lock_init(&fs_info->hash_lock);
328
329 if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
87cbda5c
CM
330 printk("failed to allocate sha256 hash\n");
331 return NULL;
332 }
d98237b3 333
79154b1b 334 mutex_init(&fs_info->trans_mutex);
d561c025 335 mutex_init(&fs_info->fs_mutex);
9f5fae2f
CM
336 memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
337 memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
3768f368 338
e20d96d6 339 __setup_root(disk_super, tree_root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
d98237b3
CM
340
341 fs_info->sb_buffer = read_tree_block(tree_root, sb_buffer->b_blocknr);
342
87cbda5c
CM
343 if (!fs_info->sb_buffer) {
344printk("failed2\n");
d98237b3 345 return NULL;
87cbda5c 346 }
d98237b3
CM
347 brelse(sb_buffer);
348 sb_buffer = NULL;
349 disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
350 fs_info->disk_super = disk_super;
351
e20d96d6
CM
352 tree_root->node = read_tree_block(tree_root,
353 btrfs_super_root(disk_super));
3768f368
CM
354 BUG_ON(!tree_root->node);
355
e20d96d6
CM
356 ret = find_and_setup_root(disk_super, tree_root, fs_info,
357 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
3768f368
CM
358 BUG_ON(ret);
359
e20d96d6
CM
360 ret = find_and_setup_root(disk_super, tree_root, fs_info,
361 BTRFS_INODE_MAP_OBJECTID, inode_root);
9f5fae2f
CM
362 BUG_ON(ret);
363
e20d96d6
CM
364 ret = find_and_setup_root(disk_super, tree_root, fs_info,
365 BTRFS_FS_TREE_OBJECTID, root);
3768f368 366 BUG_ON(ret);
a28ec197 367 root->commit_root = root->node;
e20d96d6 368 get_bh(root->node);
3768f368 369 root->ref_cows = 1;
293ffd5f 370 root->fs_info->generation = root->root_key.offset + 1;
eb60ceac
CM
371 return root;
372}
373
e089f05c 374int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
79154b1b 375 *root)
eb60ceac 376{
d5719762
CM
377 struct buffer_head *bh = root->fs_info->sb_buffer;
378 btrfs_set_super_root(root->fs_info->disk_super,
379 root->fs_info->tree_root->node->b_blocknr);
380 lock_buffer(bh);
381 clear_buffer_dirty(bh);
87cbda5c 382 csum_tree_block(root, bh, 0);
d5719762
CM
383 bh->b_end_io = end_buffer_write_sync;
384 get_bh(bh);
385 submit_bh(WRITE, bh);
386 wait_on_buffer(bh);
387 if (!buffer_uptodate(bh)) {
388 WARN_ON(1);
389 return -EIO;
cfaa7295
CM
390 }
391 return 0;
392}
393
e20d96d6 394int close_ctree(struct btrfs_root *root)
cfaa7295 395{
3768f368 396 int ret;
e089f05c
CM
397 struct btrfs_trans_handle *trans;
398
79154b1b
CM
399 trans = btrfs_start_transaction(root, 1);
400 btrfs_commit_transaction(trans, root);
401 /* run commit again to drop the original snapshot */
402 trans = btrfs_start_transaction(root, 1);
403 btrfs_commit_transaction(trans, root);
404 ret = btrfs_write_and_wait_transaction(NULL, root);
3768f368 405 BUG_ON(ret);
79154b1b 406 write_ctree_super(NULL, root);
ed2ff2cb 407
cfaa7295 408 if (root->node)
234b63a0 409 btrfs_block_release(root, root->node);
9f5fae2f
CM
410 if (root->fs_info->extent_root->node)
411 btrfs_block_release(root->fs_info->extent_root,
412 root->fs_info->extent_root->node);
413 if (root->fs_info->inode_root->node)
414 btrfs_block_release(root->fs_info->inode_root,
415 root->fs_info->inode_root->node);
416 if (root->fs_info->tree_root->node)
417 btrfs_block_release(root->fs_info->tree_root,
418 root->fs_info->tree_root->node);
234b63a0 419 btrfs_block_release(root, root->commit_root);
e20d96d6 420 btrfs_block_release(root, root->fs_info->sb_buffer);
87cbda5c 421 crypto_free_hash(root->fs_info->hash_tfm);
30ae8467 422 truncate_inode_pages(root->fs_info->btree_inode->i_mapping, 0);
d98237b3 423 iput(root->fs_info->btree_inode);
e20d96d6
CM
424 kfree(root->fs_info->extent_root);
425 kfree(root->fs_info->inode_root);
426 kfree(root->fs_info->tree_root);
427 kfree(root->fs_info);
428 kfree(root);
eb60ceac
CM
429 return 0;
430}
431
e20d96d6 432void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
eb60ceac 433{
e20d96d6 434 brelse(buf);
eb60ceac
CM
435}
436
This page took 0.048587 seconds and 5 git commands to generate.