Btrfs: Fix extent_buffer and extent_state leaks
[deliverable/linux.git] / fs / btrfs / disk-io.c
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <linux/fs.h>
20 #include <linux/blkdev.h>
21 #include <linux/crc32c.h>
22 #include <linux/scatterlist.h>
23 #include <linux/swap.h>
24 #include <linux/radix-tree.h>
25 #include <linux/writeback.h>
26 #include <linux/buffer_head.h> // for block_sync_page
27 #include "ctree.h"
28 #include "disk-io.h"
29 #include "transaction.h"
30 #include "btrfs_inode.h"
31
32 #if 0
33 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
34 {
35 if (extent_buffer_blocknr(buf) != btrfs_header_blocknr(buf)) {
36 printk(KERN_CRIT "buf blocknr(buf) is %llu, header is %llu\n",
37 (unsigned long long)extent_buffer_blocknr(buf),
38 (unsigned long long)btrfs_header_blocknr(buf));
39 return 1;
40 }
41 return 0;
42 }
43 #endif
44
45 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
46 u64 blocknr)
47 {
48 struct inode *btree_inode = root->fs_info->btree_inode;
49 struct extent_buffer *eb;
50 eb = find_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
51 blocknr * root->sectorsize,
52 root->sectorsize, GFP_NOFS);
53 if (eb)
54 eb->alloc_addr = (unsigned long)__builtin_return_address(0);
55 return eb;
56 }
57
58 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
59 u64 blocknr)
60 {
61 struct inode *btree_inode = root->fs_info->btree_inode;
62 struct extent_buffer *eb;
63 eb = alloc_extent_buffer(&BTRFS_I(btree_inode)->extent_tree,
64 blocknr * root->sectorsize,
65 root->sectorsize, GFP_NOFS);
66 eb->alloc_addr = (unsigned long)__builtin_return_address(0);
67 return eb;
68 }
69
70 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
71 size_t page_offset, u64 start, u64 end,
72 int create)
73 {
74 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
75 struct extent_map *em;
76 int ret;
77
78 again:
79 em = lookup_extent_mapping(em_tree, start, end);
80 if (em) {
81 goto out;
82 }
83 em = alloc_extent_map(GFP_NOFS);
84 if (!em) {
85 em = ERR_PTR(-ENOMEM);
86 goto out;
87 }
88 em->start = 0;
89 em->end = (i_size_read(inode) & ~((u64)PAGE_CACHE_SIZE -1)) - 1;
90 em->block_start = 0;
91 em->block_end = em->end;
92 em->bdev = inode->i_sb->s_bdev;
93 ret = add_extent_mapping(em_tree, em);
94 if (ret == -EEXIST) {
95 free_extent_map(em);
96 em = NULL;
97 goto again;
98 } else if (ret) {
99 em = ERR_PTR(ret);
100 }
101 out:
102 return em;
103 }
104
105 static int btree_writepage(struct page *page, struct writeback_control *wbc)
106 {
107 struct extent_map_tree *tree;
108 tree = &BTRFS_I(page->mapping->host)->extent_tree;
109 return extent_write_full_page(tree, page, btree_get_extent, wbc);
110 }
111 int btree_readpage(struct file *file, struct page *page)
112 {
113 struct extent_map_tree *tree;
114 tree = &BTRFS_I(page->mapping->host)->extent_tree;
115 return extent_read_full_page(tree, page, btree_get_extent);
116 }
117
118 static int btree_releasepage(struct page *page, gfp_t unused_gfp_flags)
119 {
120 struct extent_map_tree *tree;
121 int ret;
122
123 BUG_ON(page->private != 1);
124 tree = &BTRFS_I(page->mapping->host)->extent_tree;
125 ret = try_release_extent_mapping(tree, page);
126 if (ret == 1) {
127 ClearPagePrivate(page);
128 set_page_private(page, 0);
129 page_cache_release(page);
130 }
131 return ret;
132 }
133
134 static void btree_invalidatepage(struct page *page, unsigned long offset)
135 {
136 struct extent_map_tree *tree;
137 tree = &BTRFS_I(page->mapping->host)->extent_tree;
138 extent_invalidatepage(tree, page, offset);
139 btree_releasepage(page, GFP_NOFS);
140 }
141
142 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
143 char *result)
144 {
145 return 0;
146 #if 0
147 u32 crc;
148 crc = crc32c(0, data, len);
149 memcpy(result, &crc, BTRFS_CRC32_SIZE);
150 return 0;
151 #endif
152 }
153
154 #if 0
155 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
156 int verify)
157 {
158 return 0;
159 char result[BTRFS_CRC32_SIZE];
160 int ret;
161 struct btrfs_node *node;
162
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;
167 if (verify) {
168 if (memcmp(bh->b_data, result, BTRFS_CRC32_SIZE)) {
169 printk("btrfs: %s checksum verify failed on %llu\n",
170 root->fs_info->sb->s_id,
171 (unsigned long long)bh_blocknr(bh));
172 return 1;
173 }
174 } else {
175 node = btrfs_buffer_node(bh);
176 memcpy(node->header.csum, result, BTRFS_CRC32_SIZE);
177 }
178 return 0;
179 }
180 #endif
181
182 #if 0
183 static int btree_writepage(struct page *page, struct writeback_control *wbc)
184 {
185 struct buffer_head *bh;
186 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
187 struct buffer_head *head;
188 if (!page_has_buffers(page)) {
189 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
190 (1 << BH_Dirty)|(1 << BH_Uptodate));
191 }
192 head = page_buffers(page);
193 bh = head;
194 do {
195 if (buffer_dirty(bh))
196 csum_tree_block(root, bh, 0);
197 bh = bh->b_this_page;
198 } while (bh != head);
199 return block_write_full_page(page, btree_get_block, wbc);
200 }
201 #endif
202
203 static struct address_space_operations btree_aops = {
204 .readpage = btree_readpage,
205 .writepage = btree_writepage,
206 .releasepage = btree_releasepage,
207 .invalidatepage = btree_invalidatepage,
208 .sync_page = block_sync_page,
209 };
210
211 int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
212 {
213 struct extent_buffer *buf = NULL;
214 struct inode *btree_inode = root->fs_info->btree_inode;
215 int ret = 0;
216
217 buf = btrfs_find_create_tree_block(root, blocknr);
218 if (!buf)
219 return 0;
220 read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
221 buf, 0);
222 free_extent_buffer(buf);
223 return ret;
224 }
225
226 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 blocknr)
227 {
228 struct extent_buffer *buf = NULL;
229 struct inode *btree_inode = root->fs_info->btree_inode;
230
231 buf = btrfs_find_create_tree_block(root, blocknr);
232 if (!buf)
233 return NULL;
234 read_extent_buffer_pages(&BTRFS_I(btree_inode)->extent_tree,
235 buf, 1);
236 buf->alloc_addr = (unsigned long)__builtin_return_address(0);
237 return buf;
238 }
239
240 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
241 struct extent_buffer *buf)
242 {
243 struct inode *btree_inode = root->fs_info->btree_inode;
244 clear_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
245 return 0;
246 }
247
248 int wait_on_tree_block_writeback(struct btrfs_root *root,
249 struct extent_buffer *buf)
250 {
251 struct inode *btree_inode = root->fs_info->btree_inode;
252 wait_on_extent_buffer_writeback(&BTRFS_I(btree_inode)->extent_tree,
253 buf);
254 return 0;
255 }
256
257 int set_tree_block_dirty(struct btrfs_root *root, struct extent_buffer *buf)
258 {
259 struct inode *btree_inode = root->fs_info->btree_inode;
260 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
261 return 0;
262 }
263
264 static int __setup_root(int blocksize,
265 struct btrfs_root *root,
266 struct btrfs_fs_info *fs_info,
267 u64 objectid)
268 {
269 root->node = NULL;
270 root->inode = NULL;
271 root->commit_root = NULL;
272 root->sectorsize = blocksize;
273 root->nodesize = blocksize;
274 root->leafsize = blocksize;
275 root->ref_cows = 0;
276 root->fs_info = fs_info;
277 root->objectid = objectid;
278 root->last_trans = 0;
279 root->highest_inode = 0;
280 root->last_inode_alloc = 0;
281 root->name = NULL;
282 memset(&root->root_key, 0, sizeof(root->root_key));
283 memset(&root->root_item, 0, sizeof(root->root_item));
284 memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
285 memset(&root->root_kobj, 0, sizeof(root->root_kobj));
286 init_completion(&root->kobj_unregister);
287 init_rwsem(&root->snap_sem);
288 root->defrag_running = 0;
289 root->defrag_level = 0;
290 root->root_key.objectid = objectid;
291 return 0;
292 }
293
294 static int find_and_setup_root(int blocksize,
295 struct btrfs_root *tree_root,
296 struct btrfs_fs_info *fs_info,
297 u64 objectid,
298 struct btrfs_root *root)
299 {
300 int ret;
301
302 __setup_root(blocksize, root, fs_info, objectid);
303 ret = btrfs_find_last_root(tree_root, objectid,
304 &root->root_item, &root->root_key);
305 BUG_ON(ret);
306
307 root->node = read_tree_block(root,
308 btrfs_root_blocknr(&root->root_item));
309 BUG_ON(!root->node);
310 return 0;
311 }
312
313 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
314 struct btrfs_key *location)
315 {
316 struct btrfs_root *root;
317 struct btrfs_root *tree_root = fs_info->tree_root;
318 struct btrfs_path *path;
319 struct extent_buffer *l;
320 u64 highest_inode;
321 int ret = 0;
322
323 root = kzalloc(sizeof(*root), GFP_NOFS);
324 if (!root)
325 return ERR_PTR(-ENOMEM);
326 if (location->offset == (u64)-1) {
327 ret = find_and_setup_root(fs_info->sb->s_blocksize,
328 fs_info->tree_root, fs_info,
329 location->objectid, root);
330 if (ret) {
331 kfree(root);
332 return ERR_PTR(ret);
333 }
334 goto insert;
335 }
336
337 __setup_root(fs_info->sb->s_blocksize, root, fs_info,
338 location->objectid);
339
340 path = btrfs_alloc_path();
341 BUG_ON(!path);
342 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
343 if (ret != 0) {
344 if (ret > 0)
345 ret = -ENOENT;
346 goto out;
347 }
348 l = path->nodes[0];
349 read_extent_buffer(l, &root->root_item,
350 btrfs_item_ptr_offset(l, path->slots[0]),
351 sizeof(root->root_item));
352 ret = 0;
353 out:
354 btrfs_release_path(root, path);
355 btrfs_free_path(path);
356 if (ret) {
357 kfree(root);
358 return ERR_PTR(ret);
359 }
360 root->node = read_tree_block(root,
361 btrfs_root_blocknr(&root->root_item));
362 BUG_ON(!root->node);
363 insert:
364 root->ref_cows = 1;
365 ret = btrfs_find_highest_inode(root, &highest_inode);
366 if (ret == 0) {
367 root->highest_inode = highest_inode;
368 root->last_inode_alloc = highest_inode;
369 }
370 return root;
371 }
372
373 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
374 struct btrfs_key *location,
375 const char *name, int namelen)
376 {
377 struct btrfs_root *root;
378 int ret;
379
380 root = radix_tree_lookup(&fs_info->fs_roots_radix,
381 (unsigned long)location->objectid);
382 if (root)
383 return root;
384
385 root = btrfs_read_fs_root_no_radix(fs_info, location);
386 if (IS_ERR(root))
387 return root;
388 ret = radix_tree_insert(&fs_info->fs_roots_radix,
389 (unsigned long)root->root_key.objectid,
390 root);
391 if (ret) {
392 free_extent_buffer(root->node);
393 kfree(root);
394 return ERR_PTR(ret);
395 }
396
397 ret = btrfs_set_root_name(root, name, namelen);
398 if (ret) {
399 free_extent_buffer(root->node);
400 kfree(root);
401 return ERR_PTR(ret);
402 }
403
404 ret = btrfs_sysfs_add_root(root);
405 if (ret) {
406 free_extent_buffer(root->node);
407 kfree(root->name);
408 kfree(root);
409 return ERR_PTR(ret);
410 }
411
412 ret = btrfs_find_dead_roots(fs_info->tree_root,
413 root->root_key.objectid, root);
414 BUG_ON(ret);
415
416 return root;
417 }
418
419 struct btrfs_root *open_ctree(struct super_block *sb)
420 {
421 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
422 GFP_NOFS);
423 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
424 GFP_NOFS);
425 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
426 GFP_NOFS);
427 int ret;
428 int err = -EIO;
429 struct btrfs_super_block *disk_super;
430
431 if (!extent_root || !tree_root || !fs_info) {
432 err = -ENOMEM;
433 goto fail;
434 }
435 init_bit_radix(&fs_info->pinned_radix);
436 init_bit_radix(&fs_info->pending_del_radix);
437 init_bit_radix(&fs_info->extent_ins_radix);
438 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
439 INIT_RADIX_TREE(&fs_info->block_group_radix, GFP_KERNEL);
440 INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
441 INIT_LIST_HEAD(&fs_info->trans_list);
442 INIT_LIST_HEAD(&fs_info->dead_roots);
443 memset(&fs_info->super_kobj, 0, sizeof(fs_info->super_kobj));
444 init_completion(&fs_info->kobj_unregister);
445 sb_set_blocksize(sb, 4096);
446 fs_info->running_transaction = NULL;
447 fs_info->last_trans_committed = 0;
448 fs_info->tree_root = tree_root;
449 fs_info->extent_root = extent_root;
450 fs_info->sb = sb;
451 fs_info->btree_inode = new_inode(sb);
452 fs_info->btree_inode->i_ino = 1;
453 fs_info->btree_inode->i_nlink = 1;
454 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
455 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
456 extent_map_tree_init(&BTRFS_I(fs_info->btree_inode)->extent_tree,
457 fs_info->btree_inode->i_mapping,
458 GFP_NOFS);
459 extent_map_tree_init(&fs_info->free_space_cache,
460 fs_info->btree_inode->i_mapping, GFP_NOFS);
461 fs_info->do_barriers = 1;
462 fs_info->closing = 0;
463
464 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
465 BTRFS_I(fs_info->btree_inode)->root = tree_root;
466 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
467 sizeof(struct btrfs_key));
468 insert_inode_hash(fs_info->btree_inode);
469 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
470
471 mutex_init(&fs_info->trans_mutex);
472 mutex_init(&fs_info->fs_mutex);
473
474 __setup_root(sb->s_blocksize, tree_root,
475 fs_info, BTRFS_ROOT_TREE_OBJECTID);
476
477 fs_info->sb_buffer = read_tree_block(tree_root,
478 BTRFS_SUPER_INFO_OFFSET /
479 sb->s_blocksize);
480
481 if (!fs_info->sb_buffer)
482 goto fail_iput;
483
484 read_extent_buffer(fs_info->sb_buffer, &fs_info->super_copy, 0,
485 sizeof(fs_info->super_copy));
486
487 read_extent_buffer(fs_info->sb_buffer, fs_info->fsid,
488 (unsigned long)btrfs_super_fsid(fs_info->sb_buffer),
489 BTRFS_FSID_SIZE);
490 disk_super = &fs_info->super_copy;
491 if (!btrfs_super_root(disk_super))
492 goto fail_sb_buffer;
493
494 i_size_write(fs_info->btree_inode,
495 btrfs_super_total_blocks(disk_super) <<
496 fs_info->btree_inode->i_blkbits);
497
498
499 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
500 sizeof(disk_super->magic))) {
501 printk("btrfs: valid FS not found on %s\n", sb->s_id);
502 goto fail_sb_buffer;
503 }
504 tree_root->node = read_tree_block(tree_root,
505 btrfs_super_root(disk_super));
506 if (!tree_root->node)
507 goto fail_sb_buffer;
508
509 mutex_lock(&fs_info->fs_mutex);
510 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
511 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
512 if (ret) {
513 mutex_unlock(&fs_info->fs_mutex);
514 goto fail_tree_root;
515 }
516
517 btrfs_read_block_groups(extent_root);
518
519 fs_info->generation = btrfs_super_generation(disk_super) + 1;
520 mutex_unlock(&fs_info->fs_mutex);
521 return tree_root;
522
523 fail_tree_root:
524 free_extent_buffer(tree_root->node);
525 fail_sb_buffer:
526 free_extent_buffer(fs_info->sb_buffer);
527 fail_iput:
528 iput(fs_info->btree_inode);
529 fail:
530 kfree(extent_root);
531 kfree(tree_root);
532 kfree(fs_info);
533 return ERR_PTR(err);
534 }
535
536 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
537 *root)
538 {
539 int ret;
540 struct extent_buffer *super = root->fs_info->sb_buffer;
541 struct inode *btree_inode = root->fs_info->btree_inode;
542
543 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, super);
544 ret = sync_page_range_nolock(btree_inode, btree_inode->i_mapping,
545 super->start, super->len);
546 return ret;
547 }
548
549 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
550 {
551 radix_tree_delete(&fs_info->fs_roots_radix,
552 (unsigned long)root->root_key.objectid);
553 btrfs_sysfs_del_root(root);
554 if (root->inode)
555 iput(root->inode);
556 if (root->node)
557 free_extent_buffer(root->node);
558 if (root->commit_root)
559 free_extent_buffer(root->commit_root);
560 if (root->name)
561 kfree(root->name);
562 kfree(root);
563 return 0;
564 }
565
566 static int del_fs_roots(struct btrfs_fs_info *fs_info)
567 {
568 int ret;
569 struct btrfs_root *gang[8];
570 int i;
571
572 while(1) {
573 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
574 (void **)gang, 0,
575 ARRAY_SIZE(gang));
576 if (!ret)
577 break;
578 for (i = 0; i < ret; i++)
579 btrfs_free_fs_root(fs_info, gang[i]);
580 }
581 return 0;
582 }
583
584 int close_ctree(struct btrfs_root *root)
585 {
586 int ret;
587 struct btrfs_trans_handle *trans;
588 struct btrfs_fs_info *fs_info = root->fs_info;
589
590 fs_info->closing = 1;
591 btrfs_transaction_flush_work(root);
592 mutex_lock(&fs_info->fs_mutex);
593 btrfs_defrag_dirty_roots(root->fs_info);
594 trans = btrfs_start_transaction(root, 1);
595 ret = btrfs_commit_transaction(trans, root);
596 /* run commit again to drop the original snapshot */
597 trans = btrfs_start_transaction(root, 1);
598 btrfs_commit_transaction(trans, root);
599 ret = btrfs_write_and_wait_transaction(NULL, root);
600 BUG_ON(ret);
601 write_ctree_super(NULL, root);
602 mutex_unlock(&fs_info->fs_mutex);
603
604 if (fs_info->extent_root->node)
605 free_extent_buffer(fs_info->extent_root->node);
606
607 if (fs_info->tree_root->node)
608 free_extent_buffer(fs_info->tree_root->node);
609
610 free_extent_buffer(fs_info->sb_buffer);
611 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
612 iput(fs_info->btree_inode);
613
614 btrfs_free_block_groups(root->fs_info);
615 del_fs_roots(fs_info);
616 kfree(fs_info->extent_root);
617 kfree(fs_info->tree_root);
618 return 0;
619 }
620
621 int btrfs_buffer_uptodate(struct extent_buffer *buf)
622 {
623 struct inode *btree_inode = buf->first_page->mapping->host;
624 return extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree, buf);
625 }
626
627 int btrfs_set_buffer_uptodate(struct extent_buffer *buf)
628 {
629 struct inode *btree_inode = buf->first_page->mapping->host;
630 return set_extent_buffer_uptodate(&BTRFS_I(btree_inode)->extent_tree,
631 buf);
632 }
633
634 void btrfs_mark_buffer_dirty(struct extent_buffer *buf)
635 {
636 struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
637 u64 transid = btrfs_header_generation(buf);
638 struct inode *btree_inode = root->fs_info->btree_inode;
639
640 if (transid != root->fs_info->generation) {
641 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
642 (unsigned long long)extent_buffer_blocknr(buf),
643 transid, root->fs_info->generation);
644 WARN_ON(1);
645 }
646 set_extent_buffer_dirty(&BTRFS_I(btree_inode)->extent_tree, buf);
647 }
648
649 void btrfs_btree_balance_dirty(struct btrfs_root *root, unsigned long nr)
650 {
651 balance_dirty_pages_ratelimited_nr(
652 root->fs_info->btree_inode->i_mapping, nr);
653 }
This page took 0.059172 seconds and 6 git commands to generate.