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>
10 #include "transaction.h"
11 #include "btrfs_inode.h"
16 struct block_device *bdev;
19 int btrfs_insert_dev_radix(struct btrfs_root *root,
20 struct block_device *bdev,
24 struct dev_lookup *lookup;
25 char b[BDEVNAME_SIZE];
28 lookup = kmalloc(sizeof(*lookup), GFP_NOFS);
31 lookup->block_start = block_start;
32 lookup->num_blocks = num_blocks;
34 printk("inserting %s into dev radix %Lu %Lu\n", bdevname(bdev, b), block_start, num_blocks);
36 ret = radix_tree_insert(&root->fs_info->dev_radix, block_start +
37 num_blocks - 1, lookup);
41 u64 bh_blocknr(struct buffer_head *bh)
43 int blkbits = bh->b_page->mapping->host->i_blkbits;
44 u64 blocknr = bh->b_page->index << (PAGE_CACHE_SHIFT - blkbits);
47 if (PageHighMem(bh->b_page))
48 offset = (unsigned long)bh->b_data;
50 offset = bh->b_data - (char *)page_address(bh->b_page);
51 blocknr += offset >> (PAGE_CACHE_SHIFT - blkbits);
55 static int check_tree_block(struct btrfs_root *root, struct buffer_head *buf)
57 struct btrfs_node *node = btrfs_buffer_node(buf);
58 if (bh_blocknr(buf) != btrfs_header_blocknr(&node->header)) {
59 printk(KERN_CRIT "bh_blocknr(buf) is %Lu, header is %Lu\n",
60 bh_blocknr(buf), btrfs_header_blocknr(&node->header));
66 struct buffer_head *btrfs_find_tree_block(struct btrfs_root *root, u64 blocknr)
68 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
69 int blockbits = root->fs_info->sb->s_blocksize_bits;
70 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
72 struct buffer_head *bh;
73 struct buffer_head *head;
74 struct buffer_head *ret = NULL;
77 page = find_lock_page(mapping, index);
81 if (!page_has_buffers(page))
84 head = page_buffers(page);
87 if (buffer_mapped(bh) && bh_blocknr(bh) == blocknr) {
99 page_cache_release(page);
103 int btrfs_map_bh_to_logical(struct btrfs_root *root, struct buffer_head *bh,
106 struct dev_lookup *lookup[2];
110 root = root->fs_info->dev_root;
111 ret = radix_tree_gang_lookup(&root->fs_info->dev_radix,
113 (unsigned long)logical,
115 if (ret == 0 || lookup[0]->block_start > logical ||
116 lookup[0]->block_start + lookup[0]->num_blocks <= logical) {
120 bh->b_bdev = lookup[0]->bdev;
121 bh->b_blocknr = logical - lookup[0]->block_start;
122 set_buffer_mapped(bh);
128 struct buffer_head *btrfs_find_create_tree_block(struct btrfs_root *root,
131 struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
132 int blockbits = root->fs_info->sb->s_blocksize_bits;
133 unsigned long index = blocknr >> (PAGE_CACHE_SHIFT - blockbits);
135 struct buffer_head *bh;
136 struct buffer_head *head;
137 struct buffer_head *ret = NULL;
139 u64 first_block = index << (PAGE_CACHE_SHIFT - blockbits);
141 page = grab_cache_page(mapping, index);
145 if (!page_has_buffers(page))
146 create_empty_buffers(page, root->fs_info->sb->s_blocksize, 0);
147 head = page_buffers(page);
150 if (!buffer_mapped(bh)) {
151 err = btrfs_map_bh_to_logical(root, bh, first_block);
154 if (bh_blocknr(bh) == blocknr) {
159 bh = bh->b_this_page;
161 } while (bh != head);
166 page_cache_release(page);
170 static int btree_get_block(struct inode *inode, sector_t iblock,
171 struct buffer_head *bh, int create)
174 struct btrfs_root *root = BTRFS_I(bh->b_page->mapping->host)->root;
175 err = btrfs_map_bh_to_logical(root, bh, iblock);
179 int btrfs_csum_data(struct btrfs_root * root, char *data, size_t len,
182 struct scatterlist sg;
183 struct crypto_hash *tfm = root->fs_info->hash_tfm;
184 struct hash_desc desc;
189 sg_init_one(&sg, data, len);
190 spin_lock(&root->fs_info->hash_lock);
191 ret = crypto_hash_digest(&desc, &sg, 1, result);
192 spin_unlock(&root->fs_info->hash_lock);
194 printk("sha256 digest failed\n");
198 static int csum_tree_block(struct btrfs_root *root, struct buffer_head *bh,
201 char result[BTRFS_CSUM_SIZE];
203 struct btrfs_node *node;
205 ret = btrfs_csum_data(root, bh->b_data + BTRFS_CSUM_SIZE,
206 bh->b_size - BTRFS_CSUM_SIZE, result);
210 if (memcmp(bh->b_data, result, BTRFS_CSUM_SIZE)) {
211 printk("checksum verify failed on %Lu\n",
216 node = btrfs_buffer_node(bh);
217 memcpy(node->header.csum, result, BTRFS_CSUM_SIZE);
222 static int btree_writepage(struct page *page, struct writeback_control *wbc)
224 struct buffer_head *bh;
225 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
226 struct buffer_head *head;
227 if (!page_has_buffers(page)) {
228 create_empty_buffers(page, root->fs_info->sb->s_blocksize,
229 (1 << BH_Dirty)|(1 << BH_Uptodate));
231 head = page_buffers(page);
234 if (buffer_dirty(bh))
235 csum_tree_block(root, bh, 0);
236 bh = bh->b_this_page;
237 } while (bh != head);
238 return block_write_full_page(page, btree_get_block, wbc);
241 static int btree_readpage(struct file * file, struct page * page)
243 return block_read_full_page(page, btree_get_block);
246 static struct address_space_operations btree_aops = {
247 .readpage = btree_readpage,
248 .writepage = btree_writepage,
249 .sync_page = block_sync_page,
252 struct buffer_head *read_tree_block(struct btrfs_root *root, u64 blocknr)
254 struct buffer_head *bh = NULL;
256 bh = btrfs_find_create_tree_block(root, blocknr);
259 if (buffer_uptodate(bh))
262 if (!buffer_uptodate(bh)) {
264 bh->b_end_io = end_buffer_read_sync;
267 if (!buffer_uptodate(bh))
269 csum_tree_block(root, bh, 1);
274 if (check_tree_block(root, bh))
282 int dirty_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
283 struct buffer_head *buf)
285 WARN_ON(atomic_read(&buf->b_count) == 0);
286 mark_buffer_dirty(buf);
290 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
291 struct buffer_head *buf)
293 WARN_ON(atomic_read(&buf->b_count) == 0);
294 clear_buffer_dirty(buf);
298 static int __setup_root(int blocksize,
299 struct btrfs_root *root,
300 struct btrfs_fs_info *fs_info,
305 root->commit_root = NULL;
306 root->blocksize = blocksize;
308 root->fs_info = fs_info;
309 root->objectid = objectid;
310 root->last_trans = 0;
311 root->highest_inode = 0;
312 root->last_inode_alloc = 0;
313 memset(&root->root_key, 0, sizeof(root->root_key));
314 memset(&root->root_item, 0, sizeof(root->root_item));
318 static int find_and_setup_root(int blocksize,
319 struct btrfs_root *tree_root,
320 struct btrfs_fs_info *fs_info,
322 struct btrfs_root *root)
326 __setup_root(blocksize, root, fs_info, objectid);
327 ret = btrfs_find_last_root(tree_root, objectid,
328 &root->root_item, &root->root_key);
331 root->node = read_tree_block(root,
332 btrfs_root_blocknr(&root->root_item));
337 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
338 struct btrfs_key *location)
340 struct btrfs_root *root;
341 struct btrfs_root *tree_root = fs_info->tree_root;
342 struct btrfs_path *path;
343 struct btrfs_leaf *l;
347 printk("read_fs_root looking for %Lu %Lu %u\n", location->objectid, location->offset, location->flags);
348 root = radix_tree_lookup(&fs_info->fs_roots_radix,
349 (unsigned long)location->objectid);
351 printk("found %p in cache\n", root);
354 root = kmalloc(sizeof(*root), GFP_NOFS);
357 return ERR_PTR(-ENOMEM);
359 if (location->offset == (u64)-1) {
360 ret = find_and_setup_root(fs_info->sb->s_blocksize,
361 fs_info->tree_root, fs_info,
362 location->objectid, root);
371 __setup_root(fs_info->sb->s_blocksize, root, fs_info,
374 path = btrfs_alloc_path();
376 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
378 printk("internal search_slot gives us %d\n", ret);
383 l = btrfs_buffer_leaf(path->nodes[0]);
384 memcpy(&root->root_item,
385 btrfs_item_ptr(l, path->slots[0], struct btrfs_root_item),
386 sizeof(root->root_item));
387 memcpy(&root->root_key, location, sizeof(*location));
390 btrfs_release_path(root, path);
391 btrfs_free_path(path);
396 root->node = read_tree_block(root,
397 btrfs_root_blocknr(&root->root_item));
400 printk("inserting %p\n", root);
402 ret = radix_tree_insert(&fs_info->fs_roots_radix,
403 (unsigned long)root->root_key.objectid,
406 printk("radix_tree_insert gives us %d\n", ret);
411 ret = btrfs_find_highest_inode(root, &highest_inode);
413 root->highest_inode = highest_inode;
414 root->last_inode_alloc = highest_inode;
415 printk("highest inode is %Lu\n", highest_inode);
417 printk("all worked\n");
421 int btrfs_open_disk(struct btrfs_root *root, u64 block_start, u64 num_blocks,
422 char *filename, int name_len)
425 struct block_device *bdev;
428 if (block_start == 0) {
429 printk("skipping disk with block_start == 0\n");
432 null_filename = kmalloc(name_len + 1, GFP_NOFS);
435 memcpy(null_filename, filename, name_len);
436 null_filename[name_len] = '\0';
438 bdev = open_bdev_excl(null_filename, O_RDWR, root->fs_info->sb);
443 set_blocksize(bdev, root->fs_info->sb->s_blocksize);
444 ret = btrfs_insert_dev_radix(root, bdev, block_start, num_blocks);
448 kfree(null_filename);
452 static int read_device_info(struct btrfs_root *root)
454 struct btrfs_path *path;
456 struct btrfs_key key;
457 struct btrfs_leaf *leaf;
458 struct btrfs_device_item *dev_item;
462 root = root->fs_info->dev_root;
464 path = btrfs_alloc_path();
470 btrfs_set_key_type(&key, BTRFS_DEV_ITEM_KEY);
472 mutex_lock(&root->fs_info->fs_mutex);
473 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
474 leaf = btrfs_buffer_leaf(path->nodes[0]);
475 nritems = btrfs_header_nritems(&leaf->header);
477 slot = path->slots[0];
478 if (slot >= nritems) {
479 ret = btrfs_next_leaf(root, path);
482 leaf = btrfs_buffer_leaf(path->nodes[0]);
483 nritems = btrfs_header_nritems(&leaf->header);
484 slot = path->slots[0];
486 btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
487 if (btrfs_key_type(&key) != BTRFS_DEV_ITEM_KEY) {
491 dev_item = btrfs_item_ptr(leaf, slot, struct btrfs_device_item);
492 printk("found key %Lu %Lu\n", key.objectid, key.offset);
493 ret = btrfs_open_disk(root, key.objectid, key.offset,
494 (char *)(dev_item + 1),
495 btrfs_device_pathlen(dev_item));
499 btrfs_free_path(path);
500 mutex_unlock(&root->fs_info->fs_mutex);
504 struct btrfs_root *open_ctree(struct super_block *sb)
506 struct btrfs_root *extent_root = kmalloc(sizeof(struct btrfs_root),
508 struct btrfs_root *dev_root = kmalloc(sizeof(struct btrfs_root),
510 struct btrfs_root *tree_root = kmalloc(sizeof(struct btrfs_root),
512 struct btrfs_fs_info *fs_info = kmalloc(sizeof(*fs_info),
515 struct btrfs_super_block *disk_super;
516 struct dev_lookup *dev_lookup;
518 init_bit_radix(&fs_info->pinned_radix);
519 init_bit_radix(&fs_info->pending_del_radix);
520 INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
521 INIT_RADIX_TREE(&fs_info->dev_radix, GFP_NOFS);
522 sb_set_blocksize(sb, 4096);
523 fs_info->running_transaction = NULL;
524 fs_info->tree_root = tree_root;
525 fs_info->extent_root = extent_root;
526 fs_info->dev_root = dev_root;
528 fs_info->btree_inode = new_inode(sb);
529 fs_info->btree_inode->i_ino = 1;
530 fs_info->btree_inode->i_nlink = 1;
531 fs_info->btree_inode->i_size = sb->s_bdev->bd_inode->i_size;
532 fs_info->btree_inode->i_mapping->a_ops = &btree_aops;
533 BTRFS_I(fs_info->btree_inode)->root = tree_root;
534 memset(&BTRFS_I(fs_info->btree_inode)->location, 0,
535 sizeof(struct btrfs_key));
536 insert_inode_hash(fs_info->btree_inode);
537 mapping_set_gfp_mask(fs_info->btree_inode->i_mapping, GFP_NOFS);
538 fs_info->hash_tfm = crypto_alloc_hash("sha256", 0, CRYPTO_ALG_ASYNC);
539 spin_lock_init(&fs_info->hash_lock);
540 if (!fs_info->hash_tfm || IS_ERR(fs_info->hash_tfm)) {
541 printk("failed to allocate sha256 hash\n");
544 mutex_init(&fs_info->trans_mutex);
545 mutex_init(&fs_info->fs_mutex);
546 memset(&fs_info->current_insert, 0, sizeof(fs_info->current_insert));
547 memset(&fs_info->last_insert, 0, sizeof(fs_info->last_insert));
549 __setup_root(sb->s_blocksize, dev_root,
550 fs_info, BTRFS_DEV_TREE_OBJECTID);
552 __setup_root(sb->s_blocksize, tree_root,
553 fs_info, BTRFS_ROOT_TREE_OBJECTID);
555 dev_lookup = kmalloc(sizeof(*dev_lookup), GFP_NOFS);
556 dev_lookup->block_start = 0;
557 dev_lookup->num_blocks = (u32)-2;
558 dev_lookup->bdev = sb->s_bdev;
559 ret = radix_tree_insert(&fs_info->dev_radix, (u32)-2, dev_lookup);
561 fs_info->sb_buffer = read_tree_block(tree_root,
562 BTRFS_SUPER_INFO_OFFSET /
565 if (!fs_info->sb_buffer)
567 disk_super = (struct btrfs_super_block *)fs_info->sb_buffer->b_data;
568 if (!btrfs_super_root(disk_super))
571 i_size_write(fs_info->btree_inode,
572 btrfs_super_total_blocks(disk_super) <<
573 fs_info->btree_inode->i_blkbits);
575 radix_tree_delete(&fs_info->dev_radix, (u32)-2);
576 dev_lookup->block_start = btrfs_super_device_block_start(disk_super);
577 dev_lookup->num_blocks = btrfs_super_device_num_blocks(disk_super);
578 ret = radix_tree_insert(&fs_info->dev_radix,
579 dev_lookup->block_start +
580 dev_lookup->num_blocks - 1, dev_lookup);
583 fs_info->disk_super = disk_super;
585 dev_root->node = read_tree_block(tree_root,
586 btrfs_super_device_root(disk_super));
588 ret = read_device_info(dev_root);
591 tree_root->node = read_tree_block(tree_root,
592 btrfs_super_root(disk_super));
593 BUG_ON(!tree_root->node);
595 mutex_lock(&fs_info->fs_mutex);
596 ret = find_and_setup_root(sb->s_blocksize, tree_root, fs_info,
597 BTRFS_EXTENT_TREE_OBJECTID, extent_root);
600 fs_info->generation = btrfs_super_generation(disk_super) + 1;
601 memset(&fs_info->kobj, 0, sizeof(fs_info->kobj));
602 kobj_set_kset_s(fs_info, btrfs_subsys);
603 kobject_set_name(&fs_info->kobj, "%s", sb->s_id);
604 kobject_register(&fs_info->kobj);
605 mutex_unlock(&fs_info->fs_mutex);
609 int write_ctree_super(struct btrfs_trans_handle *trans, struct btrfs_root
612 struct buffer_head *bh = root->fs_info->sb_buffer;
614 btrfs_set_super_root(root->fs_info->disk_super,
615 bh_blocknr(root->fs_info->tree_root->node));
617 WARN_ON(atomic_read(&bh->b_count) < 1);
618 clear_buffer_dirty(bh);
619 csum_tree_block(root, bh, 0);
620 bh->b_end_io = end_buffer_write_sync;
622 submit_bh(WRITE, bh);
624 if (!buffer_uptodate(bh)) {
631 static int free_fs_root(struct btrfs_fs_info *fs_info, struct btrfs_root *root)
633 radix_tree_delete(&fs_info->fs_roots_radix,
634 (unsigned long)root->root_key.objectid);
639 if (root->commit_root)
640 brelse(root->commit_root);
645 int del_fs_roots(struct btrfs_fs_info *fs_info)
648 struct btrfs_root *gang[8];
652 ret = radix_tree_gang_lookup(&fs_info->fs_roots_radix,
657 for (i = 0; i < ret; i++)
658 free_fs_root(fs_info, gang[i]);
662 static int free_dev_radix(struct btrfs_fs_info *fs_info)
664 struct dev_lookup *lookup[8];
665 struct block_device *super_bdev = fs_info->sb->s_bdev;
669 ret = radix_tree_gang_lookup(&fs_info->dev_radix,
674 for (i = 0; i < ret; i++) {
675 if (lookup[i]->bdev != super_bdev)
676 close_bdev_excl(lookup[i]->bdev);
677 radix_tree_delete(&fs_info->dev_radix,
678 lookup[i]->block_start +
679 lookup[i]->num_blocks - 1);
686 int close_ctree(struct btrfs_root *root)
689 struct btrfs_trans_handle *trans;
690 struct btrfs_fs_info *fs_info = root->fs_info;
692 mutex_lock(&fs_info->fs_mutex);
693 trans = btrfs_start_transaction(root, 1);
694 btrfs_commit_transaction(trans, root);
695 /* run commit again to drop the original snapshot */
696 trans = btrfs_start_transaction(root, 1);
697 btrfs_commit_transaction(trans, root);
698 ret = btrfs_write_and_wait_transaction(NULL, root);
700 write_ctree_super(NULL, root);
701 mutex_unlock(&fs_info->fs_mutex);
703 if (fs_info->extent_root->node)
704 btrfs_block_release(fs_info->extent_root,
705 fs_info->extent_root->node);
706 if (fs_info->dev_root->node)
707 btrfs_block_release(fs_info->dev_root,
708 fs_info->dev_root->node);
709 if (fs_info->tree_root->node)
710 btrfs_block_release(fs_info->tree_root,
711 fs_info->tree_root->node);
712 btrfs_block_release(root, fs_info->sb_buffer);
713 crypto_free_hash(fs_info->hash_tfm);
714 truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
715 iput(fs_info->btree_inode);
717 free_dev_radix(fs_info);
718 del_fs_roots(fs_info);
719 kfree(fs_info->extent_root);
720 kfree(fs_info->tree_root);
721 kobject_unregister(&fs_info->kobj);
725 void btrfs_block_release(struct btrfs_root *root, struct buffer_head *buf)