1 #include <linux/module.h>
3 #include <linux/blkdev.h>
4 #include <linux/crypto.h>
5 #include <linux/scatterlist.h>
6 #include <linux/swap.h>
7 #include <linux/radix-tree.h>
8 #include <linux/writeback.h>
11 #include "transaction.h"
12 #include "btrfs_inode.h"
14 u64 bh_blocknr(struct buffer_head *bh)
19 static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
21 struct btrfs_node *node = btrfs_buffer_node(buf);
22 if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
23 printk(KERN_CRIT "bh_blocknr(buf) is %llu, header is %llu\n",
24 (unsigned long long)bh_blocknr(buf),
25 (unsigned long long)btrfs_header_blocknr(&node->header));
31 struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
33 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
34 int blockbits = root->fs_info->sb->s_blocksize_bits;
35 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
37 struct buffer_head *bh;
38 struct buffer_head *head;
39 struct buffer_head *ret = NULL;
42 page = find_lock_page(mapping, index);
46 if (!page_has_buffers(page))
49 head = page_buffers(page);
52 if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) {
61 page_cache_release(page);
65 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
71 set_buffer_mapped(bh);
73 map_bh(bh, root->fs_info->sb, logical);
78 struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
81 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
82 int blockbits = root->fs_info->sb->s_blocksize_bits;
83 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
85 struct buffer_head *bh;
86 struct buffer_head *head;
87 struct buffer_head *ret = NULL;
89 u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
91 page = grab_cache_page(mapping, index);
95 if (!page_has_buffers(page))
96 create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
97 head = page_buffers(page);
100 if (!buffer_mapped(bh)) {
101 err = btrfs_map_bh_to_logical(root, bh, first_block);
104 if (bh_blocknr(bh) == blocknr) {
109 bh = bh->b_this_page;
111 } while (bh != head);
116 page_cache_release(page);
120 static int btree_get_block(struct inode *inode, sector_t iblock,
121 struct buffer_head *bh, int create)
124 struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
125 err = btrfs_map_bh_to_logical(root, bh, iblock);
129 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
132 struct scatterlist sg;
133 struct crypto_hash *tfm = root->fs_info->hash_tfm;
134 struct hash_desc desc;
139 sg_init_one(&sg, data, len);
140 spin_lock(&root->fs_info->hash_lock);
141 ret = crypto_hash_digest(&desc, &sg, 1, result);
142 spin_unlock(&root->fs_info->hash_lock);
144 printk("digest failed\n");
148 static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
151 char result[BTRFS_CRC32_SIZE];
153 struct btrfs_node *node;
155 ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
156 bh->b_size - BTRFS_CSUM_SIZE, result);
160 if (memcmp(bh->b_data, result, BTRFS_CRC32_SIZE)) {
161 printk("btrfs: %s checksum verify failed on %llu\n",
162 root->fs_info->sb->s_id,
163 (unsigned long long)bh_blocknr(bh));
167 node = btrfs_buffer_node(bh);
168 memcpy(node->header.csum, result, BTRFS_CRC32_SIZE);
173 static int btree_writepage(struct page *page, struct writeback_control *wbc)
175 struct buffer_head *bh;
176 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
177 struct buffer_head *head;
178 if (!page_has_buffers(page)) {
179 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
180 (1 << BH_Dirty)|(1 << BH_Uptodate));
182 head = page_buffers(page);
185 if (buffer_dirty(bh))
186 csum_tree_block(root, bh, 0);
187 bh = bh->b_this_page;
188 } while (bh != head);
189 return block_write_full_page(page, btree_get_block, wbc);
192 static int btree_readpage(struct file * file, struct page * page)
194 return block_read_full_page(page, btree_get_block);
197 static struct address_space_operations btree_aops = {
198 .readpage = btree_readpage,
199 .writepage = btree_writepage,
200 .sync_page = block_sync_page,
203 int readahead_tree_block(struct btrfs_root *root, u64 blocknr)
205 struct buffer_head *bh = NULL;
208 bh = btrfs_find_create_tree_block(root, blocknr);
211 if (buffer_uptodate(bh)) {
215 if (test_set_buffer_locked(bh)) {
219 if (!buffer_uptodate(bh)) {
221 bh->b_end_io = end_buffer_read_sync;
232 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
234 struct buffer_head *bh = NULL;
236 bh = btrfs_find_create_tree_block(root, blocknr);
239 if (buffer_uptodate(bh))
242 if (!buffer_uptodate(bh)) {
244 bh->b_end_io = end_buffer_read_sync;
247 if (!buffer_uptodate(bh))
253 if (!buffer_checked(bh)) {
254 csum_tree_block(root, bh, 1);
255 set_buffer_checked(bh);
257 if (check_tree_block(root, bh))
265 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
266 struct buffer_head *buf)
268 WARN_ON(atomic_read(&buf->b_count) == 0);
269 mark_buffer_dirty(buf);
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(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 = radix_tree_lookup(&fs_info->fs_roots_radix,
332 (unsigned long)location->objectid);
335 root = kmalloc(sizeof(*root), GFP_NOFS);
337 return ERR_PTR(-ENOMEM);
338 if (location->offset == (u64)-1) {
339 ret = find_and_setup_root(fs_info->sb->s_blocksize,
340 fs_info->tree_root, fs_info,
341 location->objectid, root);
349 __setup_root(fs_info->sb->s_blocksize, root, fs_info,
352 path = btrfs_alloc_path();
354 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
360 l = btrfs_buffer_leaf(path->nodes[0]);
361 memcpy(&root->root_item,
362 btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
363 sizeof(root->root_item));
364 memcpy(&root->root_key, location, sizeof(*location));
367 btrfs_release_path(root, path);
368 btrfs_free_path(path);
373 root->node = read_tree_block(root,
374 btrfs_root_blocknr(&root->root_item));
378 ret = radix_tree_insert(&fs_info->fs_roots_radix,
379 (unsigned long)root->root_key.objectid,
386 ret = btrfs_find_highest_inode(root, &highest_inode);
388 root->highest_inode = highest_inode;
389 root->last_inode_alloc = highest_inode;
394 struct btrfs_root *open_ctree(struct super_block *sb)
396 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
398 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
400 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
404 struct btrfs_super_block *disk_super;
406 if (!extent_root || !tree_root || !fs_info) {
410 init_bit_radix(&fs_info->pinned_radix);
411 init_bit_radix(&fs_info->pending_del_radix);
412 init_bit_radix(&fs_info->extent_map_radix);
413 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
414 INIT_RADIX_TREE(&fs_info->block_group_radix, GFP_KERNEL);
415 INIT_RADIX_TREE(&fs_info->block_group_data_radix, GFP_KERNEL);
416 INIT_LIST_HEAD(&fs_info->trans_list);
417 INIT_LIST_HEAD(&fs_info->dead_roots);
418 sb_set_blocksize(sb, 4096);
419 fs_info->running_transaction = NULL;
420 fs_info->tree_root = tree_root;
421 fs_info->extent_root = extent_root;
423 fs_info->btree_inode = new_inode(sb);
424 fs_info->btree_inode->i_ino = 1;
425 fs_info->btree_inode->i_nlink = 1;
426 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
427 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
428 fs_info->do_barriers = 1;
429 fs_info->extent_tree_insert_nr = 0;
430 fs_info->extent_tree_prealloc_nr = 0;
431 fs_info->closing = 0;
433 INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
434 BTRFS_I(fs_info->btree_inode)->root = tree_root;
435 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
436 sizeof(struct btrfs_key));
437 insert_inode_hash(fs_info->btree_inode);
438 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
439 fs_info->hash_tfm = crypto_alloc_hash("crc32c", 0, CRYPTO_ALG_ASYNC);
440 spin_lock_init(&fs_info->hash_lock);
442 if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
443 printk("btrfs: failed hash setup, modprobe cryptomgr?\n");
447 mutex_init(&fs_info->trans_mutex);
448 mutex_init(&fs_info->fs_mutex);
450 __setup_root(sb->s_blocksize, tree_root,
451 fs_info, BTRFS_ROOT_TREE_OBJECTID);
453 fs_info->sb_buffer = read_tree_block(tree_root,
454 BTRFS_SUPER_INFO_OFFSET /
457 if (!fs_info->sb_buffer)
459 disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
461 if (!btrfs_super_root(disk_super))
464 i_size_write(fs_info->btree_inode,
465 btrfs_super_total_blocks(disk_super) <<
466 fs_info->btree_inode->i_blkbits);
468 fs_info->disk_super = disk_super;
470 if (strncmp((char *)(&disk_super->magic), BTRFS_MAGIC,
471 sizeof(disk_super->magic))) {
472 printk("btrfs: valid FS not found on %s\n", sb->s_id);
475 tree_root->node = read_tree_block(tree_root,
476 btrfs_super_root(disk_super));
477 if (!tree_root->node)
480 mutex_lock(&fs_info->fs_mutex);
481 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
482 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
484 mutex_unlock(&fs_info->fs_mutex);
488 btrfs_read_block_groups(extent_root);
490 fs_info->generation = btrfs_super_generation(disk_super) + 1;
491 mutex_unlock(&fs_info->fs_mutex);
495 btrfs_block_release(tree_root, tree_root->node);
497 btrfs_block_release(tree_root, fs_info->sb_buffer);
499 iput(fs_info->btree_inode);
507 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
511 struct buffer_head *bh = root->fs_info->sb_buffer;
513 btrfs_set_super_root(root->fs_info->disk_super,
514 bh_blocknr(root->fs_info->tree_root->node));
516 WARN_ON(atomic_read(&bh->b_count) < 1);
517 clear_buffer_dirty(bh);
518 csum_tree_block(root, bh, 0);
519 bh->b_end_io = end_buffer_write_sync;
521 if (root->fs_info->do_barriers)
522 ret = submit_bh(WRITE_BARRIER, bh);
524 ret = submit_bh(WRITE, bh);
525 if (ret == -EOPNOTSUPP) {
526 set_buffer_uptodate(bh);
527 root->fs_info->do_barriers = 0;
528 ret = submit_bh(WRITE, bh);
531 if (!buffer_uptodate(bh)) {
538 static int free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
540 radix_tree_delete(&fs_info->fs_roots_radix,
541 (unsigned long)root->root_key.objectid);
546 if (root->commit_root)
547 brelse(root->commit_root);
552 static int del_fs_roots(struct btrfs_fs_info *fs_info)
555 struct btrfs_root *gang[8];
559 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
564 for (i = 0; i < ret; i++)
565 free_fs_root(fs_info, gang[i]);
570 int close_ctree(struct btrfs_root *root)
573 struct btrfs_trans_handle *trans;
574 struct btrfs_fs_info *fs_info = root->fs_info;
576 fs_info->closing = 1;
577 btrfs_transaction_flush_work(root);
578 mutex_lock(&fs_info->fs_mutex);
579 trans = btrfs_start_transaction(root, 1);
580 btrfs_commit_transaction(trans, root);
581 /* run commit again to drop the original snapshot */
582 trans = btrfs_start_transaction(root, 1);
583 btrfs_commit_transaction(trans, root);
584 ret = btrfs_write_and_wait_transaction(NULL, root);
586 write_ctree_super(NULL, root);
587 mutex_unlock(&fs_info->fs_mutex);
589 if (fs_info->extent_root->node)
590 btrfs_block_release(fs_info->extent_root,
591 fs_info->extent_root->node);
592 if (fs_info->tree_root->node)
593 btrfs_block_release(fs_info->tree_root,
594 fs_info->tree_root->node);
595 btrfs_block_release(root, fs_info->sb_buffer);
596 crypto_free_hash(fs_info->hash_tfm);
597 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
598 iput(fs_info->btree_inode);
600 btrfs_free_block_groups(root->fs_info);
601 del_fs_roots(fs_info);
602 kfree(fs_info->extent_root);
603 kfree(fs_info->tree_root);
607 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)
612 void btrfs_btree_balance_dirty(struct btrfs_root *root)
614 balance_dirty_pages_ratelimited(root->fs_info->btree_inode->i_mapping);