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 clean_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 clear_buffer_dirty(buf);
281 static int __setup_root(int blocksize,
282 struct btrfs_root *root,
283 struct btrfs_fs_info *fs_info,
288 root->commit_root = NULL;
289 root->blocksize = blocksize;
291 root->fs_info = fs_info;
292 root->objectid = objectid;
293 root->last_trans = 0;
294 root->highest_inode = 0;
295 root->last_inode_alloc = 0;
296 memset(&root->root_key, 0, sizeof(root->root_key));
297 memset(&root->root_item, 0, sizeof(root->root_item));
298 root->root_key.objectid = objectid;
302 static int find_and_setup_root(int blocksize,
303 struct btrfs_root *tree_root,
304 struct btrfs_fs_info *fs_info,
306 struct btrfs_root *root)
310 __setup_root(blocksize, root, fs_info, objectid);
311 ret = btrfs_find_last_root(tree_root, objectid,
312 &root->root_item, &root->root_key);
315 root->node = read_tree_block(root,
316 btrfs_root_blocknr(&root->root_item));
321 struct btrfs_root *btrfs_read_fs_root_no_radix(struct btrfs_fs_info *fs_info,
322 struct btrfs_key *location)
324 struct btrfs_root *root;
325 struct btrfs_root *tree_root = fs_info->tree_root;
326 struct btrfs_path *path;
327 struct btrfs_leaf *l;
331 root = kzalloc(sizeof(*root), GFP_NOFS);
333 return ERR_PTR(-ENOMEM);
334 if (location->offset == (u64)-1) {
335 ret = find_and_setup_root(fs_info->sb->s_blocksize,
336 fs_info->tree_root, fs_info,
337 location->objectid, root);
345 __setup_root(fs_info->sb->s_blocksize, root, fs_info,
348 path = btrfs_alloc_path();
350 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
356 l = btrfs_buffer_leaf(path->nodes[0]);
357 memcpy(&root->root_item,
358 btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
359 sizeof(root->root_item));
360 memcpy(&root->root_key, location, sizeof(*location));
363 btrfs_release_path(root, path);
364 btrfs_free_path(path);
369 root->node = read_tree_block(root,
370 btrfs_root_blocknr(&root->root_item));
374 ret = btrfs_find_highest_inode(root, &highest_inode);
376 root->highest_inode = highest_inode;
377 root->last_inode_alloc = highest_inode;
382 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
383 struct btrfs_key *location)
385 struct btrfs_root *root;
388 root = radix_tree_lookup(&fs_info->fs_roots_radix,
389 (unsigned long)location->objectid);
393 root = btrfs_read_fs_root_no_radix(fs_info, location);
396 ret = radix_tree_insert(&fs_info->fs_roots_radix,
397 (unsigned long)root->root_key.objectid,
407 struct btrfs_root *open_ctree(struct super_block *sb)
409 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
411 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
413 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
417 struct btrfs_super_block *disk_super;
419 if (!extent_root || !tree_root || !fs_info) {
423 init_bit_radix(&fs_info->pinned_radix);
424 init_bit_radix(&fs_info->pending_del_radix);
425 init_bit_radix(&fs_info->extent_map_radix);
426 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
427 INIT_RADIX_TREE(&fs_info->block_group_radix, GFP_KERNEL);
428 INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
429 INIT_LIST_HEAD(&fs_info->trans_list);
430 INIT_LIST_HEAD(&fs_info->dead_roots);
431 sb_set_blocksize(sb, 4096);
432 fs_info->running_transaction = NULL;
433 fs_info->tree_root = tree_root;
434 fs_info->extent_root = extent_root;
436 fs_info->btree_inode = new_inode(sb);
437 fs_info->btree_inode->i_ino = 1;
438 fs_info->btree_inode->i_nlink = 1;
439 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
440 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
441 fs_info->do_barriers = 1;
442 fs_info->extent_tree_insert_nr = 0;
443 fs_info->extent_tree_prealloc_nr = 0;
444 fs_info->closing = 0;
446 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
447 BTRFS_I(fs_info->btree_inode)->root = tree_root;
448 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
449 sizeof(struct btrfs_key));
450 insert_inode_hash(fs_info->btree_inode);
451 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
453 mutex_init(&fs_info->trans_mutex);
454 mutex_init(&fs_info->fs_mutex);
456 __setup_root(sb->s_blocksize, tree_root,
457 fs_info, BTRFS_ROOT_TREE_OBJECTID);
459 fs_info->sb_buffer = read_tree_block(tree_root,
460 BTRFS_SUPER_INFO_OFFSET /
463 if (!fs_info->sb_buffer)
465 disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
466 fs_info->disk_super = disk_super;
467 memcpy(&fs_info->super_copy, disk_super, sizeof(fs_info->super_copy));
469 if (!btrfs_super_root(disk_super))
472 i_size_write(fs_info->btree_inode,
473 btrfs_super_total_blocks(disk_super) <<
474 fs_info->btree_inode->i_blkbits);
477 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
478 sizeof(disk_super->magic))) {
479 printk("btrfs: valid FS not found on %s\n", sb->s_id);
482 tree_root->node = read_tree_block(tree_root,
483 btrfs_super_root(disk_super));
484 if (!tree_root->node)
487 mutex_lock(&fs_info->fs_mutex);
488 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
489 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
491 mutex_unlock(&fs_info->fs_mutex);
495 btrfs_read_block_groups(extent_root);
497 fs_info->generation = btrfs_super_generation(disk_super) + 1;
498 ret = btrfs_find_dead_roots(tree_root);
501 mutex_unlock(&fs_info->fs_mutex);
505 btrfs_block_release(tree_root, tree_root->node);
507 btrfs_block_release(tree_root, fs_info->sb_buffer);
509 iput(fs_info->btree_inode);
517 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
521 struct buffer_head *bh = root->fs_info->sb_buffer;
524 WARN_ON(atomic_read(&bh->b_count) < 1);
525 clear_buffer_dirty(bh);
526 csum_tree_block(root, bh, 0);
527 bh->b_end_io = end_buffer_write_sync;
529 if (root->fs_info->do_barriers)
530 ret = submit_bh(WRITE_BARRIER, bh);
532 ret = submit_bh(WRITE, bh);
533 if (ret == -EOPNOTSUPP) {
536 set_buffer_uptodate(bh);
537 root->fs_info->do_barriers = 0;
538 ret = submit_bh(WRITE, bh);
541 if (!buffer_uptodate(bh)) {
548 int btrfs_free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
550 radix_tree_delete(&fs_info->fs_roots_radix,
551 (unsigned long)root->root_key.objectid);
556 if (root->commit_root)
557 brelse(root->commit_root);
562 static int del_fs_roots(struct btrfs_fs_info *fs_info)
565 struct btrfs_root *gang[8];
569 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
574 for (i = 0; i < ret; i++)
575 btrfs_free_fs_root(fs_info, gang[i]);
580 int close_ctree(struct btrfs_root *root)
583 struct btrfs_trans_handle *trans;
584 struct btrfs_fs_info *fs_info = root->fs_info;
586 fs_info->closing = 1;
587 btrfs_transaction_flush_work(root);
588 mutex_lock(&fs_info->fs_mutex);
589 trans = btrfs_start_transaction(root, 1);
590 ret = btrfs_commit_transaction(trans, root);
591 /* run commit again to drop the original snapshot */
592 trans = btrfs_start_transaction(root, 1);
593 btrfs_commit_transaction(trans, root);
594 ret = btrfs_write_and_wait_transaction(NULL, root);
596 write_ctree_super(NULL, root);
597 mutex_unlock(&fs_info->fs_mutex);
599 if (fs_info->extent_root->node)
600 btrfs_block_release(fs_info->extent_root,
601 fs_info->extent_root->node);
602 if (fs_info->tree_root->node)
603 btrfs_block_release(fs_info->tree_root,
604 fs_info->tree_root->node);
605 btrfs_block_release(root, fs_info->sb_buffer);
606 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
607 iput(fs_info->btree_inode);
609 btrfs_free_block_groups(root->fs_info);
610 del_fs_roots(fs_info);
611 kfree(fs_info->extent_root);
612 kfree(fs_info->tree_root);
616 void btrfs_mark_buffer_dirty(struct buffer_head *bh)
618 struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
619 u64 transid = btrfs_header_generation(btrfs_buffer_header(bh));
620 WARN_ON(!atomic_read(&bh->b_count));
621 if (transid != root->fs_info->generation) {
622 printk(KERN_CRIT "transid mismatch buffer %llu, found %Lu running %Lu\n",
623 (unsigned long long)bh->b_blocknr,
624 transid, root->fs_info->generation);
627 mark_buffer_dirty(bh);
630 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
635 void btrfs_btree_balance_dirty(struct btrfs_root *root)
637 balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping);