2 * Copyright (C) 2007 Oracle. All rights reserved.
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.
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.
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.
19 #include <linux/module.h>
21 #include <linux/blkdev.h>
22 #include <linux/crc32c.h>
23 #include <linux/scatterlist.h>
24 #include <linux/swap.h>
25 #include <linux/radix-tree.h>
26 #include <linux/writeback.h>
29 #include "transaction.h"
30 #include "btrfs_inode.h"
32 u64 bh_blocknr(struct buffer_head *bh)
37 static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
39 struct btrfs_node *node = btrfs_buffer_node(buf);
40 if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
41 printk(KERN_CRIT "bh_blocknr(buf) is %llu, header is %llu\n",
42 (unsigned long long)bh_blocknr(buf),
43 (unsigned long long)btrfs_header_blocknr(&node->header));
49 struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
51 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
52 int blockbits = root->fs_info->sb->s_blocksize_bits;
53 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
55 struct buffer_head *bh;
56 struct buffer_head *head;
57 struct buffer_head *ret = NULL;
60 page = find_lock_page(mapping, index);
64 if (!page_has_buffers(page))
67 head = page_buffers(page);
70 if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) {
79 page_cache_release(page);
83 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
89 set_buffer_mapped(bh);
91 map_bh(bh, root->fs_info->sb, logical);
96 struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
99 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
100 int blockbits = root->fs_info->sb->s_blocksize_bits;
101 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
103 struct buffer_head *bh;
104 struct buffer_head *head;
105 struct buffer_head *ret = NULL;
107 u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
109 page = find_or_create_page(mapping, index, GFP_NOFS);
113 if (!page_has_buffers(page))
114 create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
115 head = page_buffers(page);
118 if (!buffer_mapped(bh)) {
119 err = btrfs_map_bh_to_logical(root, bh, first_block);
122 if (bh_blocknr(bh) == blocknr) {
127 bh = bh->b_this_page;
129 } while (bh != head);
134 page_cache_release(page);
138 static int btree_get_block(struct inode *inode, sector_t iblock,
139 struct buffer_head *bh, int create)
142 struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
143 err = btrfs_map_bh_to_logical(root, bh, iblock);
147 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
151 crc = crc32c(0, data, len);
152 memcpy(result, &crc, BTRFS_CRC32_SIZE);
156 static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
159 char result[BTRFS_CRC32_SIZE];
161 struct btrfs_node *node;
163 ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
164 bh->b_size - BTRFS_CSUM_SIZE, result);
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));
175 node = btrfs_buffer_node(bh);
176 memcpy(node->header.csum, result, BTRFS_CRC32_SIZE);
181 static int btree_writepage(struct page *page, struct writeback_control *wbc)
183 struct buffer_head *bh;
184 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
185 struct buffer_head *head;
186 if (!page_has_buffers(page)) {
187 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
188 (1 << BH_Dirty)|(1 << BH_Uptodate));
190 head = page_buffers(page);
193 if (buffer_dirty(bh))
194 csum_tree_block(root, bh, 0);
195 bh = bh->b_this_page;
196 } while (bh != head);
197 return block_write_full_page(page, btree_get_block, wbc);
200 static int btree_readpage(struct file * file, struct page * page)
202 return block_read_full_page(page, btree_get_block);
205 static struct address_space_operations btree_aops = {
206 .readpage = btree_readpage,
207 .writepage = btree_writepage,
208 .sync_page = block_sync_page,
211 int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
213 struct buffer_head *bh = NULL;
216 bh = btrfs_find_create_tree_block(root, blocknr);
219 if (buffer_uptodate(bh)) {
223 if (test_set_buffer_locked(bh)) {
227 if (!buffer_uptodate(bh)) {
229 bh->b_end_io = end_buffer_read_sync;
240 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
242 struct buffer_head *bh = NULL;
244 bh = btrfs_find_create_tree_block(root, blocknr);
247 if (buffer_uptodate(bh))
250 if (!buffer_uptodate(bh)) {
252 bh->b_end_io = end_buffer_read_sync;
255 if (!buffer_uptodate(bh))
261 if (!buffer_checked(bh)) {
262 csum_tree_block(root, bh, 1);
263 set_buffer_checked(bh);
265 if (check_tree_block(root, bh))
273 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
274 struct buffer_head *buf)
276 WARN_ON(atomic_read(&buf->b_count) == 0);
277 mark_buffer_dirty(buf);
281 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
282 struct buffer_head *buf)
284 WARN_ON(atomic_read(&buf->b_count) == 0);
285 clear_buffer_dirty(buf);
289 static int __setup_root(int blocksize,
290 struct btrfs_root *root,
291 struct btrfs_fs_info *fs_info,
296 root->commit_root = NULL;
297 root->blocksize = blocksize;
299 root->fs_info = fs_info;
300 root->objectid = objectid;
301 root->last_trans = 0;
302 root->highest_inode = 0;
303 root->last_inode_alloc = 0;
304 memset(&root->root_key, 0, sizeof(root->root_key));
305 memset(&root->root_item, 0, sizeof(root->root_item));
306 root->root_key.objectid = objectid;
310 static int find_and_setup_root(int blocksize,
311 struct btrfs_root *tree_root,
312 struct btrfs_fs_info *fs_info,
314 struct btrfs_root *root)
318 __setup_root(blocksize, root, fs_info, objectid);
319 ret = btrfs_find_last_root(tree_root, objectid,
320 &root->root_item, &root->root_key);
323 root->node = read_tree_block(root,
324 btrfs_root_blocknr(&root->root_item));
329 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
330 struct btrfs_key *location)
332 struct btrfs_root *root;
333 struct btrfs_root *tree_root = fs_info->tree_root;
334 struct btrfs_path *path;
335 struct btrfs_leaf *l;
339 root = radix_tree_lookup(&fs_info->fs_roots_radix,
340 (unsigned long)location->objectid);
343 root = kmalloc(sizeof(*root), GFP_NOFS);
345 return ERR_PTR(-ENOMEM);
346 if (location->offset == (u64)-1) {
347 ret = find_and_setup_root(fs_info->sb->s_blocksize,
348 fs_info->tree_root, fs_info,
349 location->objectid, root);
357 __setup_root(fs_info->sb->s_blocksize, root, fs_info,
360 path = btrfs_alloc_path();
362 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
368 l = btrfs_buffer_leaf(path->nodes[0]);
369 memcpy(&root->root_item,
370 btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
371 sizeof(root->root_item));
372 memcpy(&root->root_key, location, sizeof(*location));
375 btrfs_release_path(root, path);
376 btrfs_free_path(path);
381 root->node = read_tree_block(root,
382 btrfs_root_blocknr(&root->root_item));
386 ret = radix_tree_insert(&fs_info->fs_roots_radix,
387 (unsigned long)root->root_key.objectid,
394 ret = btrfs_find_highest_inode(root, &highest_inode);
396 root->highest_inode = highest_inode;
397 root->last_inode_alloc = highest_inode;
402 struct btrfs_root *open_ctree(struct super_block *sb)
404 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
406 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
408 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
412 struct btrfs_super_block *disk_super;
414 if (!extent_root || !tree_root || !fs_info) {
418 init_bit_radix(&fs_info->pinned_radix);
419 init_bit_radix(&fs_info->pending_del_radix);
420 init_bit_radix(&fs_info->extent_map_radix);
421 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
422 INIT_RADIX_TREE(&fs_info->block_group_radix, GFP_KERNEL);
423 INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
424 INIT_LIST_HEAD(&fs_info->trans_list);
425 INIT_LIST_HEAD(&fs_info->dead_roots);
426 sb_set_blocksize(sb, 4096);
427 fs_info->running_transaction = NULL;
428 fs_info->tree_root = tree_root;
429 fs_info->extent_root = extent_root;
431 fs_info->btree_inode = new_inode(sb);
432 fs_info->btree_inode->i_ino = 1;
433 fs_info->btree_inode->i_nlink = 1;
434 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
435 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
436 fs_info->do_barriers = 1;
437 fs_info->extent_tree_insert_nr = 0;
438 fs_info->extent_tree_prealloc_nr = 0;
439 fs_info->closing = 0;
441 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
442 BTRFS_I(fs_info->btree_inode)->root = tree_root;
443 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
444 sizeof(struct btrfs_key));
445 insert_inode_hash(fs_info->btree_inode);
446 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
448 mutex_init(&fs_info->trans_mutex);
449 mutex_init(&fs_info->fs_mutex);
451 __setup_root(sb->s_blocksize, tree_root,
452 fs_info, BTRFS_ROOT_TREE_OBJECTID);
454 fs_info->sb_buffer = read_tree_block(tree_root,
455 BTRFS_SUPER_INFO_OFFSET /
458 if (!fs_info->sb_buffer)
460 disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
462 if (!btrfs_super_root(disk_super))
465 i_size_write(fs_info->btree_inode,
466 btrfs_super_total_blocks(disk_super) <<
467 fs_info->btree_inode->i_blkbits);
469 fs_info->disk_super = disk_super;
471 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
472 sizeof(disk_super->magic))) {
473 printk("btrfs: valid FS not found on %s\n", sb->s_id);
476 tree_root->node = read_tree_block(tree_root,
477 btrfs_super_root(disk_super));
478 if (!tree_root->node)
481 mutex_lock(&fs_info->fs_mutex);
482 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
483 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
485 mutex_unlock(&fs_info->fs_mutex);
489 btrfs_read_block_groups(extent_root);
491 fs_info->generation = btrfs_super_generation(disk_super) + 1;
492 mutex_unlock(&fs_info->fs_mutex);
496 btrfs_block_release(tree_root, tree_root->node);
498 btrfs_block_release(tree_root, fs_info->sb_buffer);
500 iput(fs_info->btree_inode);
508 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
512 struct buffer_head *bh = root->fs_info->sb_buffer;
514 btrfs_set_super_root(root->fs_info->disk_super,
515 bh_blocknr(root->fs_info->tree_root->node));
517 WARN_ON(atomic_read(&bh->b_count) < 1);
518 clear_buffer_dirty(bh);
519 csum_tree_block(root, bh, 0);
520 bh->b_end_io = end_buffer_write_sync;
522 if (root->fs_info->do_barriers)
523 ret = submit_bh(WRITE_BARRIER, bh);
525 ret = submit_bh(WRITE, bh);
526 if (ret == -EOPNOTSUPP) {
529 set_buffer_uptodate(bh);
530 root->fs_info->do_barriers = 0;
531 ret = submit_bh(WRITE, bh);
534 if (!buffer_uptodate(bh)) {
541 static int free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
543 radix_tree_delete(&fs_info->fs_roots_radix,
544 (unsigned long)root->root_key.objectid);
549 if (root->commit_root)
550 brelse(root->commit_root);
555 static int del_fs_roots(struct btrfs_fs_info *fs_info)
558 struct btrfs_root *gang[8];
562 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
567 for (i = 0; i < ret; i++)
568 free_fs_root(fs_info, gang[i]);
573 int close_ctree(struct btrfs_root *root)
576 struct btrfs_trans_handle *trans;
577 struct btrfs_fs_info *fs_info = root->fs_info;
579 fs_info->closing = 1;
580 btrfs_transaction_flush_work(root);
581 mutex_lock(&fs_info->fs_mutex);
582 trans = btrfs_start_transaction(root, 1);
583 btrfs_commit_transaction(trans, root);
584 /* run commit again to drop the original snapshot */
585 trans = btrfs_start_transaction(root, 1);
586 btrfs_commit_transaction(trans, root);
587 ret = btrfs_write_and_wait_transaction(NULL, root);
589 write_ctree_super(NULL, root);
590 mutex_unlock(&fs_info->fs_mutex);
592 if (fs_info->extent_root->node)
593 btrfs_block_release(fs_info->extent_root,
594 fs_info->extent_root->node);
595 if (fs_info->tree_root->node)
596 btrfs_block_release(fs_info->tree_root,
597 fs_info->tree_root->node);
598 btrfs_block_release(root, fs_info->sb_buffer);
599 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
600 iput(fs_info->btree_inode);
602 btrfs_free_block_groups(root->fs_info);
603 del_fs_roots(fs_info);
604 kfree(fs_info->extent_root);
605 kfree(fs_info->tree_root);
609 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
614 void btrfs_btree_balance_dirty(struct btrfs_root *root)
616 balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping);