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/buffer_head.h>
21 #include <linux/pagemap.h>
22 #include <linux/highmem.h>
23 #include <linux/time.h>
24 #include <linux/init.h>
25 #include <linux/string.h>
26 #include <linux/smp_lock.h>
27 #include <linux/backing-dev.h>
28 #include <linux/mpage.h>
29 #include <linux/swap.h>
30 #include <linux/writeback.h>
31 #include <linux/statfs.h>
32 #include <linux/compat.h>
33 #include <linux/bit_spinlock.h>
34 #include <linux/version.h>
37 #include "transaction.h"
38 #include "btrfs_inode.h"
40 #include "print-tree.h"
42 struct btrfs_iget_args {
44 struct btrfs_root *root;
47 static struct inode_operations btrfs_dir_inode_operations;
48 static struct inode_operations btrfs_symlink_inode_operations;
49 static struct inode_operations btrfs_dir_ro_inode_operations;
50 static struct inode_operations btrfs_special_inode_operations;
51 static struct inode_operations btrfs_file_inode_operations;
52 static struct address_space_operations btrfs_aops;
53 static struct address_space_operations btrfs_symlink_aops;
54 static struct file_operations btrfs_dir_file_operations;
56 static struct kmem_cache *btrfs_inode_cachep;
57 struct kmem_cache *btrfs_trans_handle_cachep;
58 struct kmem_cache *btrfs_transaction_cachep;
59 struct kmem_cache *btrfs_bit_radix_cachep;
60 struct kmem_cache *btrfs_path_cachep;
63 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
64 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
65 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
66 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
67 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
68 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
69 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
70 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
73 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
75 struct btrfs_root *root = BTRFS_I(inode)->root;
76 struct btrfs_trans_handle *trans;
81 u64 blocksize = 1 << inode->i_blkbits;
83 mutex_lock(&root->fs_info->fs_mutex);
84 trans = btrfs_start_transaction(root, 1);
85 btrfs_set_trans_block_group(trans, inode);
87 num_blocks = (end - start + blocksize) & ~(blocksize - 1);
88 ret = btrfs_drop_extents(trans, root, inode,
89 start, start + num_blocks, &alloc_hint);
90 num_blocks = num_blocks >> inode->i_blkbits;
91 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0,
92 alloc_hint, (u64)-1, &ins, 1);
97 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
98 start, ins.objectid, ins.offset,
101 btrfs_end_transaction(trans, root);
102 mutex_unlock(&root->fs_info->fs_mutex);
107 void btrfs_read_locked_inode(struct inode *inode)
109 struct btrfs_path *path;
110 struct btrfs_inode_item *inode_item;
111 struct btrfs_root *root = BTRFS_I(inode)->root;
112 struct btrfs_key location;
113 u64 alloc_group_block;
117 path = btrfs_alloc_path();
119 mutex_lock(&root->fs_info->fs_mutex);
121 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
122 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
124 btrfs_free_path(path);
127 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
129 struct btrfs_inode_item);
131 inode->i_mode = btrfs_inode_mode(inode_item);
132 inode->i_nlink = btrfs_inode_nlink(inode_item);
133 inode->i_uid = btrfs_inode_uid(inode_item);
134 inode->i_gid = btrfs_inode_gid(inode_item);
135 inode->i_size = btrfs_inode_size(inode_item);
136 inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
137 inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
138 inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
139 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
140 inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
141 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
142 inode->i_blocks = btrfs_inode_nblocks(inode_item);
143 inode->i_generation = btrfs_inode_generation(inode_item);
145 rdev = btrfs_inode_rdev(inode_item);
146 alloc_group_block = btrfs_inode_block_group(inode_item);
147 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
150 btrfs_free_path(path);
153 mutex_unlock(&root->fs_info->fs_mutex);
155 switch (inode->i_mode & S_IFMT) {
157 inode->i_mapping->a_ops = &btrfs_aops;
158 BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
159 inode->i_fop = &btrfs_file_operations;
160 inode->i_op = &btrfs_file_inode_operations;
163 inode->i_fop = &btrfs_dir_file_operations;
164 if (root == root->fs_info->tree_root)
165 inode->i_op = &btrfs_dir_ro_inode_operations;
167 inode->i_op = &btrfs_dir_inode_operations;
170 inode->i_op = &btrfs_symlink_inode_operations;
171 inode->i_mapping->a_ops = &btrfs_symlink_aops;
174 init_special_inode(inode, inode->i_mode, rdev);
180 btrfs_release_path(root, path);
181 btrfs_free_path(path);
182 mutex_unlock(&root->fs_info->fs_mutex);
183 make_bad_inode(inode);
186 static void fill_inode_item(struct btrfs_inode_item *item,
189 btrfs_set_inode_uid(item, inode->i_uid);
190 btrfs_set_inode_gid(item, inode->i_gid);
191 btrfs_set_inode_size(item, inode->i_size);
192 btrfs_set_inode_mode(item, inode->i_mode);
193 btrfs_set_inode_nlink(item, inode->i_nlink);
194 btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
195 btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
196 btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
197 btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
198 btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
199 btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
200 btrfs_set_inode_nblocks(item, inode->i_blocks);
201 btrfs_set_inode_generation(item, inode->i_generation);
202 btrfs_set_inode_rdev(item, inode->i_rdev);
203 btrfs_set_inode_block_group(item,
204 BTRFS_I(inode)->block_group->key.objectid);
207 int btrfs_update_inode(struct btrfs_trans_handle *trans,
208 struct btrfs_root *root,
211 struct btrfs_inode_item *inode_item;
212 struct btrfs_path *path;
215 path = btrfs_alloc_path();
217 ret = btrfs_lookup_inode(trans, root, path,
218 &BTRFS_I(inode)->location, 1);
225 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
227 struct btrfs_inode_item);
229 fill_inode_item(inode_item, inode);
230 btrfs_mark_buffer_dirty(path->nodes[0]);
231 btrfs_set_inode_last_trans(trans, inode);
234 btrfs_release_path(root, path);
235 btrfs_free_path(path);
240 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
241 struct btrfs_root *root,
243 struct dentry *dentry)
245 struct btrfs_path *path;
246 const char *name = dentry->d_name.name;
247 int name_len = dentry->d_name.len;
250 struct btrfs_dir_item *di;
252 path = btrfs_alloc_path();
258 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
268 objectid = btrfs_disk_key_objectid(&di->location);
269 ret = btrfs_delete_one_dir_name(trans, root, path, di);
272 btrfs_release_path(root, path);
274 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
275 objectid, name, name_len, -1);
284 ret = btrfs_delete_one_dir_name(trans, root, path, di);
286 dentry->d_inode->i_ctime = dir->i_ctime;
288 btrfs_free_path(path);
290 dir->i_size -= name_len * 2;
291 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
292 btrfs_update_inode(trans, root, dir);
293 drop_nlink(dentry->d_inode);
294 ret = btrfs_update_inode(trans, root, dentry->d_inode);
295 dir->i_sb->s_dirt = 1;
300 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
302 struct btrfs_root *root;
303 struct btrfs_trans_handle *trans;
306 root = BTRFS_I(dir)->root;
307 mutex_lock(&root->fs_info->fs_mutex);
308 trans = btrfs_start_transaction(root, 1);
309 btrfs_set_trans_block_group(trans, dir);
310 ret = btrfs_unlink_trans(trans, root, dir, dentry);
311 btrfs_end_transaction(trans, root);
312 mutex_unlock(&root->fs_info->fs_mutex);
313 btrfs_btree_balance_dirty(root);
317 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
319 struct inode *inode = dentry->d_inode;
322 struct btrfs_root *root = BTRFS_I(dir)->root;
323 struct btrfs_path *path;
324 struct btrfs_key key;
325 struct btrfs_trans_handle *trans;
326 struct btrfs_key found_key;
328 struct btrfs_leaf *leaf;
329 char *goodnames = "..";
331 path = btrfs_alloc_path();
333 mutex_lock(&root->fs_info->fs_mutex);
334 trans = btrfs_start_transaction(root, 1);
335 btrfs_set_trans_block_group(trans, dir);
336 key.objectid = inode->i_ino;
337 key.offset = (u64)-1;
340 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
346 if (path->slots[0] == 0) {
351 leaf = btrfs_buffer_leaf(path->nodes[0]);
352 btrfs_disk_key_to_cpu(&found_key,
353 &leaf->items[path->slots[0]].key);
354 found_type = btrfs_key_type(&found_key);
355 if (found_key.objectid != inode->i_ino) {
359 if ((found_type != BTRFS_DIR_ITEM_KEY &&
360 found_type != BTRFS_DIR_INDEX_KEY) ||
361 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
362 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
366 ret = btrfs_del_item(trans, root, path);
369 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
371 btrfs_release_path(root, path);
374 btrfs_release_path(root, path);
376 /* now the directory is empty */
377 err = btrfs_unlink_trans(trans, root, dir, dentry);
382 btrfs_release_path(root, path);
383 btrfs_free_path(path);
384 mutex_unlock(&root->fs_info->fs_mutex);
385 ret = btrfs_end_transaction(trans, root);
386 btrfs_btree_balance_dirty(root);
392 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
393 struct btrfs_root *root,
396 struct btrfs_path *path;
401 path = btrfs_alloc_path();
403 ret = btrfs_lookup_inode(trans, root, path,
404 &BTRFS_I(inode)->location, -1);
408 ret = btrfs_del_item(trans, root, path);
409 btrfs_free_path(path);
414 * this can truncate away extent items, csum items and directory items.
415 * It starts at a high offset and removes keys until it can't find
416 * any higher than i_size.
418 * csum items that cross the new i_size are truncated to the new size
421 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
422 struct btrfs_root *root,
426 struct btrfs_path *path;
427 struct btrfs_key key;
428 struct btrfs_disk_key *found_key;
430 struct btrfs_leaf *leaf;
431 struct btrfs_file_extent_item *fi;
432 u64 extent_start = 0;
433 u64 extent_num_blocks = 0;
438 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
439 path = btrfs_alloc_path();
442 /* FIXME, add redo link to tree so we don't leak on crash */
443 key.objectid = inode->i_ino;
444 key.offset = (u64)-1;
447 btrfs_init_path(path);
449 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
454 BUG_ON(path->slots[0] == 0);
457 leaf = btrfs_buffer_leaf(path->nodes[0]);
458 found_key = &leaf->items[path->slots[0]].key;
459 found_type = btrfs_disk_key_type(found_key);
461 if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
463 if (found_type != BTRFS_CSUM_ITEM_KEY &&
464 found_type != BTRFS_DIR_ITEM_KEY &&
465 found_type != BTRFS_DIR_INDEX_KEY &&
466 found_type != BTRFS_EXTENT_DATA_KEY)
469 item_end = btrfs_disk_key_offset(found_key);
470 if (found_type == BTRFS_EXTENT_DATA_KEY) {
471 fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
473 struct btrfs_file_extent_item);
474 if (btrfs_file_extent_type(fi) !=
475 BTRFS_FILE_EXTENT_INLINE) {
476 item_end += btrfs_file_extent_num_blocks(fi) <<
480 if (found_type == BTRFS_CSUM_ITEM_KEY) {
481 ret = btrfs_csum_truncate(trans, root, path,
485 if (item_end < inode->i_size) {
486 if (found_type == BTRFS_DIR_ITEM_KEY) {
487 found_type = BTRFS_INODE_ITEM_KEY;
488 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
489 found_type = BTRFS_CSUM_ITEM_KEY;
490 } else if (found_type) {
495 btrfs_set_key_type(&key, found_type - 1);
498 if (btrfs_disk_key_offset(found_key) >= inode->i_size)
504 /* FIXME, shrink the extent if the ref count is only 1 */
505 if (found_type == BTRFS_EXTENT_DATA_KEY &&
506 btrfs_file_extent_type(fi) !=
507 BTRFS_FILE_EXTENT_INLINE) {
510 u64 orig_num_blocks =
511 btrfs_file_extent_num_blocks(fi);
512 extent_num_blocks = inode->i_size -
513 btrfs_disk_key_offset(found_key) +
515 extent_num_blocks >>= inode->i_blkbits;
516 btrfs_set_file_extent_num_blocks(fi,
518 inode->i_blocks -= (orig_num_blocks -
519 extent_num_blocks) << 3;
520 btrfs_mark_buffer_dirty(path->nodes[0]);
523 btrfs_file_extent_disk_blocknr(fi);
525 btrfs_file_extent_disk_num_blocks(fi);
526 /* FIXME blocksize != 4096 */
527 num_dec = btrfs_file_extent_num_blocks(fi) << 3;
528 if (extent_start != 0) {
530 inode->i_blocks -= num_dec;
535 ret = btrfs_del_item(trans, root, path);
541 btrfs_release_path(root, path);
543 ret = btrfs_free_extent(trans, root, extent_start,
544 extent_num_blocks, 0);
550 btrfs_release_path(root, path);
551 btrfs_free_path(path);
552 inode->i_sb->s_dirt = 1;
556 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
561 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
562 u64 page_start = page->index << PAGE_CACHE_SHIFT;
563 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
565 if (!PagePrivate(page)) {
566 SetPagePrivate(page);
567 set_page_private(page, 1);
568 WARN_ON(!page->mapping->a_ops->invalidatepage);
569 page_cache_get(page);
572 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
573 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
575 if (zero_start != PAGE_CACHE_SIZE) {
577 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
578 flush_dcache_page(page);
581 set_page_dirty(page);
582 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
588 * taken from block_truncate_page, but does cow as it zeros out
589 * any bytes left in the last page in the file.
591 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
593 struct inode *inode = mapping->host;
594 unsigned blocksize = 1 << inode->i_blkbits;
595 pgoff_t index = from >> PAGE_CACHE_SHIFT;
596 unsigned offset = from & (PAGE_CACHE_SIZE-1);
601 if ((offset & (blocksize - 1)) == 0)
605 page = grab_cache_page(mapping, index);
608 if (!PageUptodate(page)) {
609 ret = btrfs_readpage(NULL, page);
611 if (!PageUptodate(page)) {
616 page_start = page->index << PAGE_CACHE_SHIFT;
618 ret = btrfs_cow_one_page(inode, page, offset);
621 page_cache_release(page);
626 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
628 struct inode *inode = dentry->d_inode;
631 err = inode_change_ok(inode, attr);
635 if (S_ISREG(inode->i_mode) &&
636 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
637 struct btrfs_trans_handle *trans;
638 struct btrfs_root *root = BTRFS_I(inode)->root;
639 u64 mask = root->blocksize - 1;
640 u64 pos = (inode->i_size + mask) & ~mask;
643 if (attr->ia_size <= pos)
646 btrfs_truncate_page(inode->i_mapping, inode->i_size);
648 hole_size = (attr->ia_size - pos + mask) & ~mask;
649 hole_size >>= inode->i_blkbits;
651 mutex_lock(&root->fs_info->fs_mutex);
652 trans = btrfs_start_transaction(root, 1);
653 btrfs_set_trans_block_group(trans, inode);
654 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
655 pos, 0, 0, hole_size);
656 btrfs_end_transaction(trans, root);
657 mutex_unlock(&root->fs_info->fs_mutex);
662 err = inode_setattr(inode, attr);
666 void btrfs_delete_inode(struct inode *inode)
668 struct btrfs_trans_handle *trans;
669 struct btrfs_root *root = BTRFS_I(inode)->root;
672 truncate_inode_pages(&inode->i_data, 0);
673 if (is_bad_inode(inode)) {
677 mutex_lock(&root->fs_info->fs_mutex);
678 trans = btrfs_start_transaction(root, 1);
679 btrfs_set_trans_block_group(trans, inode);
680 ret = btrfs_truncate_in_trans(trans, root, inode);
683 ret = btrfs_free_inode(trans, root, inode);
686 btrfs_end_transaction(trans, root);
687 mutex_unlock(&root->fs_info->fs_mutex);
688 btrfs_btree_balance_dirty(root);
692 btrfs_end_transaction(trans, root);
693 mutex_unlock(&root->fs_info->fs_mutex);
694 btrfs_btree_balance_dirty(root);
700 * this returns the key found in the dir entry in the location pointer.
701 * If no dir entries were found, location->objectid is 0.
703 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
704 struct btrfs_key *location)
706 const char *name = dentry->d_name.name;
707 int namelen = dentry->d_name.len;
708 struct btrfs_dir_item *di;
709 struct btrfs_path *path;
710 struct btrfs_root *root = BTRFS_I(dir)->root;
713 path = btrfs_alloc_path();
715 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
717 if (!di || IS_ERR(di)) {
718 location->objectid = 0;
722 btrfs_disk_key_to_cpu(location, &di->location);
724 btrfs_release_path(root, path);
725 btrfs_free_path(path);
730 * when we hit a tree root in a directory, the btrfs part of the inode
731 * needs to be changed to reflect the root directory of the tree root. This
732 * is kind of like crossing a mount point.
734 static int fixup_tree_root_location(struct btrfs_root *root,
735 struct btrfs_key *location,
736 struct btrfs_root **sub_root,
737 struct dentry *dentry)
739 struct btrfs_path *path;
740 struct btrfs_root_item *ri;
742 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
744 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
747 path = btrfs_alloc_path();
749 mutex_lock(&root->fs_info->fs_mutex);
751 *sub_root = btrfs_read_fs_root(root->fs_info, location,
754 if (IS_ERR(*sub_root))
755 return PTR_ERR(*sub_root);
757 ri = &(*sub_root)->root_item;
758 location->objectid = btrfs_root_dirid(ri);
760 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
761 location->offset = 0;
763 btrfs_free_path(path);
764 mutex_unlock(&root->fs_info->fs_mutex);
768 static int btrfs_init_locked_inode(struct inode *inode, void *p)
770 struct btrfs_iget_args *args = p;
771 inode->i_ino = args->ino;
772 BTRFS_I(inode)->root = args->root;
773 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
774 inode->i_mapping, GFP_NOFS);
778 static int btrfs_find_actor(struct inode *inode, void *opaque)
780 struct btrfs_iget_args *args = opaque;
781 return (args->ino == inode->i_ino &&
782 args->root == BTRFS_I(inode)->root);
785 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
786 struct btrfs_root *root)
789 struct btrfs_iget_args args;
793 inode = iget5_locked(s, objectid, btrfs_find_actor,
794 btrfs_init_locked_inode,
799 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
800 struct nameidata *nd)
802 struct inode * inode;
803 struct btrfs_inode *bi = BTRFS_I(dir);
804 struct btrfs_root *root = bi->root;
805 struct btrfs_root *sub_root = root;
806 struct btrfs_key location;
809 if (dentry->d_name.len > BTRFS_NAME_LEN)
810 return ERR_PTR(-ENAMETOOLONG);
811 mutex_lock(&root->fs_info->fs_mutex);
812 ret = btrfs_inode_by_name(dir, dentry, &location);
813 mutex_unlock(&root->fs_info->fs_mutex);
817 if (location.objectid) {
818 ret = fixup_tree_root_location(root, &location, &sub_root,
823 return ERR_PTR(-ENOENT);
824 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
827 return ERR_PTR(-EACCES);
828 if (inode->i_state & I_NEW) {
829 /* the inode and parent dir are two different roots */
830 if (sub_root != root) {
832 sub_root->inode = inode;
834 BTRFS_I(inode)->root = sub_root;
835 memcpy(&BTRFS_I(inode)->location, &location,
837 btrfs_read_locked_inode(inode);
838 unlock_new_inode(inode);
841 return d_splice_alias(inode, dentry);
844 static unsigned char btrfs_filetype_table[] = {
845 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
848 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
850 struct inode *inode = filp->f_path.dentry->d_inode;
851 struct btrfs_root *root = BTRFS_I(inode)->root;
852 struct btrfs_item *item;
853 struct btrfs_dir_item *di;
854 struct btrfs_key key;
855 struct btrfs_path *path;
858 struct btrfs_leaf *leaf;
861 unsigned char d_type;
866 int key_type = BTRFS_DIR_INDEX_KEY;
868 /* FIXME, use a real flag for deciding about the key type */
869 if (root->fs_info->tree_root == root)
870 key_type = BTRFS_DIR_ITEM_KEY;
871 mutex_lock(&root->fs_info->fs_mutex);
872 key.objectid = inode->i_ino;
874 btrfs_set_key_type(&key, key_type);
875 key.offset = filp->f_pos;
876 path = btrfs_alloc_path();
878 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
883 leaf = btrfs_buffer_leaf(path->nodes[0]);
884 nritems = btrfs_header_nritems(&leaf->header);
885 slot = path->slots[0];
886 if (advance || slot >= nritems) {
887 if (slot >= nritems -1) {
888 ret = btrfs_next_leaf(root, path);
891 leaf = btrfs_buffer_leaf(path->nodes[0]);
892 nritems = btrfs_header_nritems(&leaf->header);
893 slot = path->slots[0];
900 item = leaf->items + slot;
901 if (btrfs_disk_key_objectid(&item->key) != key.objectid)
903 if (btrfs_disk_key_type(&item->key) != key_type)
905 if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
907 filp->f_pos = btrfs_disk_key_offset(&item->key);
909 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
911 di_total = btrfs_item_size(leaf->items + slot);
912 while(di_cur < di_total) {
913 d_type = btrfs_filetype_table[btrfs_dir_type(di)];
914 over = filldir(dirent, (const char *)(di + 1),
915 btrfs_dir_name_len(di),
916 btrfs_disk_key_offset(&item->key),
917 btrfs_disk_key_objectid(&di->location),
921 di_len = btrfs_dir_name_len(di) + sizeof(*di);
923 di = (struct btrfs_dir_item *)((char *)di + di_len);
930 btrfs_release_path(root, path);
931 btrfs_free_path(path);
932 mutex_unlock(&root->fs_info->fs_mutex);
936 int btrfs_write_inode(struct inode *inode, int wait)
938 struct btrfs_root *root = BTRFS_I(inode)->root;
939 struct btrfs_trans_handle *trans;
943 mutex_lock(&root->fs_info->fs_mutex);
944 trans = btrfs_start_transaction(root, 1);
945 btrfs_set_trans_block_group(trans, inode);
946 ret = btrfs_commit_transaction(trans, root);
947 mutex_unlock(&root->fs_info->fs_mutex);
953 * This is somewhat expensive, updating the tree every time the
954 * inode changes. But, it is most likely to find the inode in cache.
955 * FIXME, needs more benchmarking...there are no reasons other than performance
956 * to keep or drop this code.
958 void btrfs_dirty_inode(struct inode *inode)
960 struct btrfs_root *root = BTRFS_I(inode)->root;
961 struct btrfs_trans_handle *trans;
963 mutex_lock(&root->fs_info->fs_mutex);
964 trans = btrfs_start_transaction(root, 1);
965 btrfs_set_trans_block_group(trans, inode);
966 btrfs_update_inode(trans, root, inode);
967 btrfs_end_transaction(trans, root);
968 mutex_unlock(&root->fs_info->fs_mutex);
971 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
972 struct btrfs_root *root,
974 struct btrfs_block_group_cache *group,
978 struct btrfs_inode_item inode_item;
979 struct btrfs_key *location;
983 inode = new_inode(root->fs_info->sb);
985 return ERR_PTR(-ENOMEM);
987 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
988 inode->i_mapping, GFP_NOFS);
989 BTRFS_I(inode)->root = root;
995 group = btrfs_find_block_group(root, group, 0, 0, owner);
996 BTRFS_I(inode)->block_group = group;
998 inode->i_uid = current->fsuid;
999 inode->i_gid = current->fsgid;
1000 inode->i_mode = mode;
1001 inode->i_ino = objectid;
1002 inode->i_blocks = 0;
1003 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1004 fill_inode_item(&inode_item, inode);
1005 location = &BTRFS_I(inode)->location;
1006 location->objectid = objectid;
1007 location->flags = 0;
1008 location->offset = 0;
1009 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1011 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1013 return ERR_PTR(ret);
1014 insert_inode_hash(inode);
1018 static inline u8 btrfs_inode_type(struct inode *inode)
1020 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1023 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1024 struct dentry *dentry, struct inode *inode)
1027 struct btrfs_key key;
1028 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1029 struct inode *parent_inode;
1030 key.objectid = inode->i_ino;
1032 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1035 ret = btrfs_insert_dir_item(trans, root,
1036 dentry->d_name.name, dentry->d_name.len,
1037 dentry->d_parent->d_inode->i_ino,
1038 &key, btrfs_inode_type(inode));
1040 parent_inode = dentry->d_parent->d_inode;
1041 parent_inode->i_size += dentry->d_name.len * 2;
1042 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1043 ret = btrfs_update_inode(trans, root,
1044 dentry->d_parent->d_inode);
1049 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1050 struct dentry *dentry, struct inode *inode)
1052 int err = btrfs_add_link(trans, dentry, inode);
1054 d_instantiate(dentry, inode);
1062 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1063 int mode, dev_t rdev)
1065 struct btrfs_trans_handle *trans;
1066 struct btrfs_root *root = BTRFS_I(dir)->root;
1067 struct inode *inode;
1072 if (!new_valid_dev(rdev))
1075 mutex_lock(&root->fs_info->fs_mutex);
1076 trans = btrfs_start_transaction(root, 1);
1077 btrfs_set_trans_block_group(trans, dir);
1079 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1085 inode = btrfs_new_inode(trans, root, objectid,
1086 BTRFS_I(dir)->block_group, mode);
1087 err = PTR_ERR(inode);
1091 btrfs_set_trans_block_group(trans, inode);
1092 err = btrfs_add_nondir(trans, dentry, inode);
1096 inode->i_op = &btrfs_special_inode_operations;
1097 init_special_inode(inode, inode->i_mode, rdev);
1098 btrfs_update_inode(trans, root, inode);
1100 dir->i_sb->s_dirt = 1;
1101 btrfs_update_inode_block_group(trans, inode);
1102 btrfs_update_inode_block_group(trans, dir);
1104 btrfs_end_transaction(trans, root);
1105 mutex_unlock(&root->fs_info->fs_mutex);
1108 inode_dec_link_count(inode);
1111 btrfs_btree_balance_dirty(root);
1115 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1116 int mode, struct nameidata *nd)
1118 struct btrfs_trans_handle *trans;
1119 struct btrfs_root *root = BTRFS_I(dir)->root;
1120 struct inode *inode;
1125 mutex_lock(&root->fs_info->fs_mutex);
1126 trans = btrfs_start_transaction(root, 1);
1127 btrfs_set_trans_block_group(trans, dir);
1129 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1135 inode = btrfs_new_inode(trans, root, objectid,
1136 BTRFS_I(dir)->block_group, mode);
1137 err = PTR_ERR(inode);
1141 btrfs_set_trans_block_group(trans, inode);
1142 err = btrfs_add_nondir(trans, dentry, inode);
1146 inode->i_mapping->a_ops = &btrfs_aops;
1147 inode->i_fop = &btrfs_file_operations;
1148 inode->i_op = &btrfs_file_inode_operations;
1149 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1150 inode->i_mapping, GFP_NOFS);
1151 BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
1153 dir->i_sb->s_dirt = 1;
1154 btrfs_update_inode_block_group(trans, inode);
1155 btrfs_update_inode_block_group(trans, dir);
1157 btrfs_end_transaction(trans, root);
1158 mutex_unlock(&root->fs_info->fs_mutex);
1161 inode_dec_link_count(inode);
1164 btrfs_btree_balance_dirty(root);
1168 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1169 struct dentry *dentry)
1171 struct btrfs_trans_handle *trans;
1172 struct btrfs_root *root = BTRFS_I(dir)->root;
1173 struct inode *inode = old_dentry->d_inode;
1177 if (inode->i_nlink == 0)
1181 mutex_lock(&root->fs_info->fs_mutex);
1182 trans = btrfs_start_transaction(root, 1);
1183 btrfs_set_trans_block_group(trans, dir);
1184 atomic_inc(&inode->i_count);
1185 err = btrfs_add_nondir(trans, dentry, inode);
1188 dir->i_sb->s_dirt = 1;
1189 btrfs_update_inode_block_group(trans, dir);
1190 err = btrfs_update_inode(trans, root, inode);
1194 btrfs_end_transaction(trans, root);
1195 mutex_unlock(&root->fs_info->fs_mutex);
1198 inode_dec_link_count(inode);
1201 btrfs_btree_balance_dirty(root);
1205 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1206 struct btrfs_root *root,
1207 u64 objectid, u64 dirid)
1211 struct btrfs_key key;
1216 key.objectid = objectid;
1219 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1221 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1222 &key, BTRFS_FT_DIR);
1225 key.objectid = dirid;
1226 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1227 &key, BTRFS_FT_DIR);
1234 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1236 struct inode *inode;
1237 struct btrfs_trans_handle *trans;
1238 struct btrfs_root *root = BTRFS_I(dir)->root;
1240 int drop_on_err = 0;
1243 mutex_lock(&root->fs_info->fs_mutex);
1244 trans = btrfs_start_transaction(root, 1);
1245 btrfs_set_trans_block_group(trans, dir);
1246 if (IS_ERR(trans)) {
1247 err = PTR_ERR(trans);
1251 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1257 inode = btrfs_new_inode(trans, root, objectid,
1258 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1259 if (IS_ERR(inode)) {
1260 err = PTR_ERR(inode);
1264 inode->i_op = &btrfs_dir_inode_operations;
1265 inode->i_fop = &btrfs_dir_file_operations;
1266 btrfs_set_trans_block_group(trans, inode);
1268 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1273 err = btrfs_update_inode(trans, root, inode);
1276 err = btrfs_add_link(trans, dentry, inode);
1279 d_instantiate(dentry, inode);
1281 dir->i_sb->s_dirt = 1;
1282 btrfs_update_inode_block_group(trans, inode);
1283 btrfs_update_inode_block_group(trans, dir);
1286 btrfs_end_transaction(trans, root);
1288 mutex_unlock(&root->fs_info->fs_mutex);
1291 btrfs_btree_balance_dirty(root);
1295 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1296 size_t page_offset, u64 start, u64 end,
1302 u64 extent_start = 0;
1304 u64 objectid = inode->i_ino;
1306 int failed_insert = 0;
1307 struct btrfs_path *path;
1308 struct btrfs_root *root = BTRFS_I(inode)->root;
1309 struct btrfs_file_extent_item *item;
1310 struct btrfs_leaf *leaf;
1311 struct btrfs_disk_key *found_key;
1312 struct extent_map *em = NULL;
1313 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1314 struct btrfs_trans_handle *trans = NULL;
1316 path = btrfs_alloc_path();
1318 mutex_lock(&root->fs_info->fs_mutex);
1321 em = lookup_extent_mapping(em_tree, start, end);
1326 em = alloc_extent_map(GFP_NOFS);
1334 em->bdev = inode->i_sb->s_bdev;
1335 ret = btrfs_lookup_file_extent(NULL, root, path,
1336 objectid, start, 0);
1343 if (path->slots[0] == 0)
1348 item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1349 struct btrfs_file_extent_item);
1350 leaf = btrfs_buffer_leaf(path->nodes[0]);
1351 blocknr = btrfs_file_extent_disk_blocknr(item);
1352 blocknr += btrfs_file_extent_offset(item);
1354 /* are we inside the extent that was found? */
1355 found_key = &leaf->items[path->slots[0]].key;
1356 found_type = btrfs_disk_key_type(found_key);
1357 if (btrfs_disk_key_objectid(found_key) != objectid ||
1358 found_type != BTRFS_EXTENT_DATA_KEY) {
1362 found_type = btrfs_file_extent_type(item);
1363 extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1364 if (found_type == BTRFS_FILE_EXTENT_REG) {
1365 extent_end = extent_start +
1366 (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
1368 if (start < extent_start || start >= extent_end) {
1370 if (start < extent_start) {
1371 if (end < extent_start)
1373 em->end = extent_end - 1;
1379 if (btrfs_file_extent_disk_blocknr(item) == 0) {
1380 em->start = extent_start;
1381 em->end = extent_end - 1;
1382 em->block_start = 0;
1386 em->block_start = blocknr << inode->i_blkbits;
1387 em->block_end = em->block_start +
1388 (btrfs_file_extent_num_blocks(item) <<
1389 inode->i_blkbits) - 1;
1390 em->start = extent_start;
1391 em->end = extent_end - 1;
1393 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1398 size = btrfs_file_extent_inline_len(leaf->items +
1400 extent_end = extent_start + size;
1401 if (start < extent_start || start >= extent_end) {
1403 if (start < extent_start) {
1404 if (end < extent_start)
1406 em->end = extent_end - 1;
1412 em->block_start = EXTENT_MAP_INLINE;
1413 em->block_end = EXTENT_MAP_INLINE;
1414 em->start = extent_start;
1415 em->end = extent_end - 1;
1419 ptr = btrfs_file_extent_inline_start(item);
1421 memcpy(map + page_offset, ptr, size);
1422 flush_dcache_page(result->b_page);
1424 set_extent_uptodate(em_tree, extent_start,
1425 extent_end, GFP_NOFS);
1428 printk("unkknown found_type %d\n", found_type);
1435 em->block_start = 0;
1438 btrfs_release_path(root, path);
1439 if (em->start > start || em->end < start) {
1440 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1444 ret = add_extent_mapping(em_tree, em);
1445 if (ret == -EEXIST) {
1446 free_extent_map(em);
1448 if (failed_insert > 5) {
1449 printk("failing to insert %Lu %Lu\n", start, end);
1458 btrfs_free_path(path);
1460 ret = btrfs_end_transaction(trans, root);
1464 mutex_unlock(&root->fs_info->fs_mutex);
1466 free_extent_map(em);
1468 return ERR_PTR(err);
1475 * FIBMAP and others want to pass in a fake buffer head. They need to
1476 * use BTRFS_GET_BLOCK_NO_DIRECT to make sure we don't try to memcpy
1477 * any packed file data into the fake bh
1479 #define BTRFS_GET_BLOCK_NO_CREATE 0
1480 #define BTRFS_GET_BLOCK_CREATE 1
1481 #define BTRFS_GET_BLOCK_NO_DIRECT 2
1484 * FIXME create==1 doe not work.
1486 static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
1487 struct buffer_head *result, int create)
1492 u64 extent_start = 0;
1494 u64 objectid = inode->i_ino;
1497 struct btrfs_path *path;
1498 struct btrfs_root *root = BTRFS_I(inode)->root;
1499 struct btrfs_file_extent_item *item;
1500 struct btrfs_leaf *leaf;
1501 struct btrfs_disk_key *found_key;
1502 struct btrfs_trans_handle *trans = NULL;
1504 path = btrfs_alloc_path();
1506 if (create & BTRFS_GET_BLOCK_CREATE) {
1508 * danger!, this only works if the page is properly up
1511 trans = btrfs_start_transaction(root, 1);
1516 ret = btrfs_drop_extents(trans, root, inode,
1517 iblock << inode->i_blkbits,
1518 (iblock + 1) << inode->i_blkbits,
1523 ret = btrfs_lookup_file_extent(NULL, root, path,
1525 iblock << inode->i_blkbits, 0);
1532 if (path->slots[0] == 0) {
1533 btrfs_release_path(root, path);
1539 item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1540 struct btrfs_file_extent_item);
1541 leaf = btrfs_buffer_leaf(path->nodes[0]);
1542 blocknr = btrfs_file_extent_disk_blocknr(item);
1543 blocknr += btrfs_file_extent_offset(item);
1545 /* are we inside the extent that was found? */
1546 found_key = &leaf->items[path->slots[0]].key;
1547 found_type = btrfs_disk_key_type(found_key);
1548 if (btrfs_disk_key_objectid(found_key) != objectid ||
1549 found_type != BTRFS_EXTENT_DATA_KEY) {
1554 found_type = btrfs_file_extent_type(item);
1555 extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1556 if (found_type == BTRFS_FILE_EXTENT_REG) {
1557 extent_start = extent_start >> inode->i_blkbits;
1558 extent_end = extent_start + btrfs_file_extent_num_blocks(item);
1560 if (btrfs_file_extent_disk_blocknr(item) == 0)
1562 if (iblock >= extent_start && iblock < extent_end) {
1563 btrfs_map_bh_to_logical(root, result, blocknr +
1564 iblock - extent_start);
1567 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1572 if (create & BTRFS_GET_BLOCK_NO_DIRECT) {
1576 size = btrfs_file_extent_inline_len(leaf->items +
1578 extent_end = (extent_start + size) >> inode->i_blkbits;
1579 extent_start >>= inode->i_blkbits;
1580 if (iblock < extent_start || iblock > extent_end) {
1583 ptr = btrfs_file_extent_inline_start(item);
1584 map = kmap(result->b_page);
1585 memcpy(map, ptr, size);
1586 memset(map + size, 0, PAGE_CACHE_SIZE - size);
1587 flush_dcache_page(result->b_page);
1588 kunmap(result->b_page);
1589 set_buffer_uptodate(result);
1590 SetPageChecked(result->b_page);
1591 btrfs_map_bh_to_logical(root, result, 0);
1594 if (create & BTRFS_GET_BLOCK_CREATE) {
1595 struct btrfs_key ins;
1596 ret = btrfs_alloc_extent(trans, root, inode->i_ino,
1597 1, 0, alloc_hint, (u64)-1,
1603 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
1604 iblock << inode->i_blkbits,
1605 ins.objectid, ins.offset,
1611 btrfs_map_bh_to_logical(root, result, ins.objectid);
1615 ret = btrfs_end_transaction(trans, root);
1619 btrfs_free_path(path);
1623 int btrfs_get_block(struct inode *inode, sector_t iblock,
1624 struct buffer_head *result, int create)
1627 struct btrfs_root *root = BTRFS_I(inode)->root;
1628 mutex_lock(&root->fs_info->fs_mutex);
1629 err = btrfs_get_block_lock(inode, iblock, result, create);
1630 mutex_unlock(&root->fs_info->fs_mutex);
1634 static int btrfs_get_block_bmap(struct inode *inode, sector_t iblock,
1635 struct buffer_head *result, int create)
1637 struct btrfs_root *root = BTRFS_I(inode)->root;
1638 u64 start = iblock << inode->i_blkbits;
1639 u64 end = start + root->blocksize -1;
1640 struct extent_map *em;
1642 em = btrfs_get_extent(inode, NULL, 0, start, end, 0);
1643 if (em && !IS_ERR(em) && em->block_start != EXTENT_MAP_INLINE &&
1644 em->block_start != 0) {
1646 offset = start - em->start;
1647 start = (em->block_start + offset) >> inode->i_blkbits;
1648 btrfs_map_bh_to_logical(root, result, start);
1653 static sector_t btrfs_bmap(struct address_space *as, sector_t block)
1655 return generic_block_bmap(as, block, btrfs_get_block_bmap);
1658 static int btrfs_prepare_write(struct file *file, struct page *page,
1659 unsigned from, unsigned to)
1661 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1662 page->mapping->host, page, from, to,
1666 int btrfs_readpage(struct file *file, struct page *page)
1668 struct extent_map_tree *tree;
1669 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1670 return extent_read_full_page(tree, page, btrfs_get_extent);
1672 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1674 struct extent_map_tree *tree;
1677 if (current->flags & PF_MEMALLOC) {
1678 redirty_page_for_writepage(wbc, page);
1682 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1683 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
1686 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1688 struct extent_map_tree *tree;
1691 if (page->private != 1) {
1693 return try_to_free_buffers(page);
1695 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1696 ret = try_release_extent_mapping(tree, page);
1698 ClearPagePrivate(page);
1699 set_page_private(page, 0);
1700 page_cache_release(page);
1705 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1707 struct extent_map_tree *tree;
1709 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1710 extent_invalidatepage(tree, page, offset);
1711 btrfs_releasepage(page, GFP_NOFS);
1715 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1716 * called from a page fault handler when a page is first dirtied. Hence we must
1717 * be careful to check for EOF conditions here. We set the page up correctly
1718 * for a written page which means we get ENOSPC checking when writing into
1719 * holes and correct delalloc and unwritten extent mapping on filesystems that
1720 * support these features.
1722 * We are not allowed to take the i_mutex here so we have to play games to
1723 * protect against truncate races as the page could now be beyond EOF. Because
1724 * vmtruncate() writes the inode size before removing pages, once we have the
1725 * page lock we can determine safely if the page is beyond EOF. If it is not
1726 * beyond EOF, then the page is guaranteed safe against truncation until we
1729 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1731 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1738 wait_on_page_writeback(page);
1739 size = i_size_read(inode);
1740 page_start = page->index << PAGE_CACHE_SHIFT;
1742 if ((page->mapping != inode->i_mapping) ||
1743 (page_start > size)) {
1744 /* page got truncated out from underneath us */
1748 /* page is wholly or partially inside EOF */
1749 if (page_start + PAGE_CACHE_SIZE > size)
1750 end = size & ~PAGE_CACHE_MASK;
1752 end = PAGE_CACHE_SIZE;
1754 ret = btrfs_cow_one_page(inode, page, end);
1761 static void btrfs_truncate(struct inode *inode)
1763 struct btrfs_root *root = BTRFS_I(inode)->root;
1765 struct btrfs_trans_handle *trans;
1767 if (!S_ISREG(inode->i_mode))
1769 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1772 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1774 mutex_lock(&root->fs_info->fs_mutex);
1775 trans = btrfs_start_transaction(root, 1);
1776 btrfs_set_trans_block_group(trans, inode);
1778 /* FIXME, add redo link to tree so we don't leak on crash */
1779 ret = btrfs_truncate_in_trans(trans, root, inode);
1780 btrfs_update_inode(trans, root, inode);
1781 ret = btrfs_end_transaction(trans, root);
1783 mutex_unlock(&root->fs_info->fs_mutex);
1784 btrfs_btree_balance_dirty(root);
1787 int btrfs_commit_write(struct file *file, struct page *page,
1788 unsigned from, unsigned to)
1790 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1791 page->mapping->host, page, from, to);
1794 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1796 struct btrfs_trans_handle *trans;
1797 struct btrfs_key key;
1798 struct btrfs_root_item root_item;
1799 struct btrfs_inode_item *inode_item;
1800 struct buffer_head *subvol;
1801 struct btrfs_leaf *leaf;
1802 struct btrfs_root *new_root;
1803 struct inode *inode;
1808 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1810 mutex_lock(&root->fs_info->fs_mutex);
1811 trans = btrfs_start_transaction(root, 1);
1814 subvol = btrfs_alloc_free_block(trans, root, 0, 0);
1816 return PTR_ERR(subvol);
1817 leaf = btrfs_buffer_leaf(subvol);
1818 btrfs_set_header_nritems(&leaf->header, 0);
1819 btrfs_set_header_level(&leaf->header, 0);
1820 btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
1821 btrfs_set_header_generation(&leaf->header, trans->transid);
1822 btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
1823 memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
1824 sizeof(leaf->header.fsid));
1825 btrfs_mark_buffer_dirty(subvol);
1827 inode_item = &root_item.inode;
1828 memset(inode_item, 0, sizeof(*inode_item));
1829 btrfs_set_inode_generation(inode_item, 1);
1830 btrfs_set_inode_size(inode_item, 3);
1831 btrfs_set_inode_nlink(inode_item, 1);
1832 btrfs_set_inode_nblocks(inode_item, 1);
1833 btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
1835 btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
1836 btrfs_set_root_refs(&root_item, 1);
1837 btrfs_set_root_blocks_used(&root_item, 0);
1838 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1839 root_item.drop_level = 0;
1843 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1848 btrfs_set_root_dirid(&root_item, new_dirid);
1850 key.objectid = objectid;
1853 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1854 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1860 * insert the directory item
1862 key.offset = (u64)-1;
1863 dir = root->fs_info->sb->s_root->d_inode;
1864 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1865 name, namelen, dir->i_ino, &key,
1870 ret = btrfs_commit_transaction(trans, root);
1874 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1877 trans = btrfs_start_transaction(new_root, 1);
1880 inode = btrfs_new_inode(trans, new_root, new_dirid,
1881 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
1884 inode->i_op = &btrfs_dir_inode_operations;
1885 inode->i_fop = &btrfs_dir_file_operations;
1886 new_root->inode = inode;
1888 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1894 ret = btrfs_update_inode(trans, new_root, inode);
1898 err = btrfs_commit_transaction(trans, root);
1902 mutex_unlock(&root->fs_info->fs_mutex);
1903 btrfs_btree_balance_dirty(root);
1907 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1909 struct btrfs_trans_handle *trans;
1910 struct btrfs_key key;
1911 struct btrfs_root_item new_root_item;
1912 struct buffer_head *tmp;
1917 if (!root->ref_cows)
1920 mutex_lock(&root->fs_info->fs_mutex);
1921 trans = btrfs_start_transaction(root, 1);
1924 ret = btrfs_update_inode(trans, root, root->inode);
1928 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1933 memcpy(&new_root_item, &root->root_item,
1934 sizeof(new_root_item));
1936 key.objectid = objectid;
1939 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1940 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
1941 btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
1943 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1949 * insert the directory item
1951 key.offset = (u64)-1;
1952 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1954 root->fs_info->sb->s_root->d_inode->i_ino,
1955 &key, BTRFS_FT_DIR);
1960 ret = btrfs_inc_root_ref(trans, root);
1965 err = btrfs_commit_transaction(trans, root);
1968 mutex_unlock(&root->fs_info->fs_mutex);
1969 btrfs_btree_balance_dirty(root);
1973 int btrfs_ioctl(struct inode *inode, struct file *filp, unsigned int
1974 cmd, unsigned long arg)
1976 struct btrfs_root *root = BTRFS_I(inode)->root;
1977 struct btrfs_ioctl_vol_args vol_args;
1979 struct btrfs_dir_item *di;
1981 struct btrfs_path *path;
1985 case BTRFS_IOC_SNAP_CREATE:
1986 if (copy_from_user(&vol_args,
1987 (struct btrfs_ioctl_vol_args __user *)arg,
1990 namelen = strlen(vol_args.name);
1991 if (namelen > BTRFS_VOL_NAME_MAX)
1993 if (strchr(vol_args.name, '/'))
1995 path = btrfs_alloc_path();
1998 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
1999 mutex_lock(&root->fs_info->fs_mutex);
2000 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2002 vol_args.name, namelen, 0);
2003 mutex_unlock(&root->fs_info->fs_mutex);
2004 btrfs_free_path(path);
2005 if (di && !IS_ERR(di))
2010 if (root == root->fs_info->tree_root)
2011 ret = create_subvol(root, vol_args.name, namelen);
2013 ret = create_snapshot(root, vol_args.name, namelen);
2016 case BTRFS_IOC_DEFRAG:
2017 mutex_lock(&root->fs_info->fs_mutex);
2018 btrfs_defrag_root(root, 0);
2019 btrfs_defrag_root(root->fs_info->extent_root, 0);
2020 mutex_unlock(&root->fs_info->fs_mutex);
2029 #ifdef CONFIG_COMPAT
2030 long btrfs_compat_ioctl(struct file *file, unsigned int cmd,
2033 struct inode *inode = file->f_path.dentry->d_inode;
2036 ret = btrfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg));
2044 * Called inside transaction, so use GFP_NOFS
2046 struct inode *btrfs_alloc_inode(struct super_block *sb)
2048 struct btrfs_inode *ei;
2050 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2054 return &ei->vfs_inode;
2057 void btrfs_destroy_inode(struct inode *inode)
2059 WARN_ON(!list_empty(&inode->i_dentry));
2060 WARN_ON(inode->i_data.nrpages);
2062 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2065 static void init_once(void * foo, struct kmem_cache * cachep,
2066 unsigned long flags)
2068 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2070 inode_init_once(&ei->vfs_inode);
2073 void btrfs_destroy_cachep(void)
2075 if (btrfs_inode_cachep)
2076 kmem_cache_destroy(btrfs_inode_cachep);
2077 if (btrfs_trans_handle_cachep)
2078 kmem_cache_destroy(btrfs_trans_handle_cachep);
2079 if (btrfs_transaction_cachep)
2080 kmem_cache_destroy(btrfs_transaction_cachep);
2081 if (btrfs_bit_radix_cachep)
2082 kmem_cache_destroy(btrfs_bit_radix_cachep);
2083 if (btrfs_path_cachep)
2084 kmem_cache_destroy(btrfs_path_cachep);
2087 static struct kmem_cache *cache_create(const char *name, size_t size,
2088 unsigned long extra_flags,
2089 void (*ctor)(void *, struct kmem_cache *,
2092 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2093 SLAB_MEM_SPREAD | extra_flags), ctor
2094 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2100 int btrfs_init_cachep(void)
2102 btrfs_inode_cachep = cache_create("btrfs_inode_cache",
2103 sizeof(struct btrfs_inode),
2105 if (!btrfs_inode_cachep)
2107 btrfs_trans_handle_cachep = cache_create("btrfs_trans_handle_cache",
2108 sizeof(struct btrfs_trans_handle),
2110 if (!btrfs_trans_handle_cachep)
2112 btrfs_transaction_cachep = cache_create("btrfs_transaction_cache",
2113 sizeof(struct btrfs_transaction),
2115 if (!btrfs_transaction_cachep)
2117 btrfs_path_cachep = cache_create("btrfs_path_cache",
2118 sizeof(struct btrfs_transaction),
2120 if (!btrfs_path_cachep)
2122 btrfs_bit_radix_cachep = cache_create("btrfs_radix", 256,
2123 SLAB_DESTROY_BY_RCU, NULL);
2124 if (!btrfs_bit_radix_cachep)
2128 btrfs_destroy_cachep();
2132 static int btrfs_getattr(struct vfsmount *mnt,
2133 struct dentry *dentry, struct kstat *stat)
2135 struct inode *inode = dentry->d_inode;
2136 generic_fillattr(inode, stat);
2137 stat->blksize = 256 * 1024;
2141 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2142 struct inode * new_dir,struct dentry *new_dentry)
2144 struct btrfs_trans_handle *trans;
2145 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2146 struct inode *new_inode = new_dentry->d_inode;
2147 struct inode *old_inode = old_dentry->d_inode;
2148 struct timespec ctime = CURRENT_TIME;
2149 struct btrfs_path *path;
2150 struct btrfs_dir_item *di;
2153 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2154 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2157 mutex_lock(&root->fs_info->fs_mutex);
2158 trans = btrfs_start_transaction(root, 1);
2159 btrfs_set_trans_block_group(trans, new_dir);
2160 path = btrfs_alloc_path();
2166 old_dentry->d_inode->i_nlink++;
2167 old_dir->i_ctime = old_dir->i_mtime = ctime;
2168 new_dir->i_ctime = new_dir->i_mtime = ctime;
2169 old_inode->i_ctime = ctime;
2170 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2171 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2173 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2183 old_parent_oid = btrfs_disk_key_objectid(&di->location);
2184 ret = btrfs_del_item(trans, root, path);
2188 btrfs_release_path(root, path);
2190 di = btrfs_lookup_dir_index_item(trans, root, path,
2202 ret = btrfs_del_item(trans, root, path);
2206 btrfs_release_path(root, path);
2208 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2209 old_inode->i_ino, location,
2216 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2221 new_inode->i_ctime = CURRENT_TIME;
2222 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2225 if (S_ISDIR(new_inode->i_mode))
2226 clear_nlink(new_inode);
2228 drop_nlink(new_inode);
2229 ret = btrfs_update_inode(trans, root, new_inode);
2233 ret = btrfs_add_link(trans, new_dentry, old_inode);
2238 btrfs_free_path(path);
2239 btrfs_end_transaction(trans, root);
2240 mutex_unlock(&root->fs_info->fs_mutex);
2244 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2245 const char *symname)
2247 struct btrfs_trans_handle *trans;
2248 struct btrfs_root *root = BTRFS_I(dir)->root;
2249 struct btrfs_path *path;
2250 struct btrfs_key key;
2251 struct inode *inode;
2258 struct btrfs_file_extent_item *ei;
2260 name_len = strlen(symname) + 1;
2261 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2262 return -ENAMETOOLONG;
2263 mutex_lock(&root->fs_info->fs_mutex);
2264 trans = btrfs_start_transaction(root, 1);
2265 btrfs_set_trans_block_group(trans, dir);
2267 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2273 inode = btrfs_new_inode(trans, root, objectid,
2274 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2275 err = PTR_ERR(inode);
2279 btrfs_set_trans_block_group(trans, inode);
2280 err = btrfs_add_nondir(trans, dentry, inode);
2284 inode->i_mapping->a_ops = &btrfs_aops;
2285 inode->i_fop = &btrfs_file_operations;
2286 inode->i_op = &btrfs_file_inode_operations;
2287 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2288 inode->i_mapping, GFP_NOFS);
2289 BTRFS_I(inode)->extent_tree.fill_delalloc = run_delalloc_range;
2291 dir->i_sb->s_dirt = 1;
2292 btrfs_update_inode_block_group(trans, inode);
2293 btrfs_update_inode_block_group(trans, dir);
2297 path = btrfs_alloc_path();
2299 key.objectid = inode->i_ino;
2302 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2303 datasize = btrfs_file_extent_calc_inline_size(name_len);
2304 err = btrfs_insert_empty_item(trans, root, path, &key,
2310 ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
2311 path->slots[0], struct btrfs_file_extent_item);
2312 btrfs_set_file_extent_generation(ei, trans->transid);
2313 btrfs_set_file_extent_type(ei,
2314 BTRFS_FILE_EXTENT_INLINE);
2315 ptr = btrfs_file_extent_inline_start(ei);
2316 btrfs_memcpy(root, path->nodes[0]->b_data,
2317 ptr, symname, name_len);
2318 btrfs_mark_buffer_dirty(path->nodes[0]);
2319 btrfs_free_path(path);
2320 inode->i_op = &btrfs_symlink_inode_operations;
2321 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2322 inode->i_size = name_len - 1;
2323 err = btrfs_update_inode(trans, root, inode);
2328 btrfs_end_transaction(trans, root);
2329 mutex_unlock(&root->fs_info->fs_mutex);
2331 inode_dec_link_count(inode);
2334 btrfs_btree_balance_dirty(root);
2338 static struct inode_operations btrfs_dir_inode_operations = {
2339 .lookup = btrfs_lookup,
2340 .create = btrfs_create,
2341 .unlink = btrfs_unlink,
2343 .mkdir = btrfs_mkdir,
2344 .rmdir = btrfs_rmdir,
2345 .rename = btrfs_rename,
2346 .symlink = btrfs_symlink,
2347 .setattr = btrfs_setattr,
2348 .mknod = btrfs_mknod,
2351 static struct inode_operations btrfs_dir_ro_inode_operations = {
2352 .lookup = btrfs_lookup,
2355 static struct file_operations btrfs_dir_file_operations = {
2356 .llseek = generic_file_llseek,
2357 .read = generic_read_dir,
2358 .readdir = btrfs_readdir,
2359 .ioctl = btrfs_ioctl,
2360 #ifdef CONFIG_COMPAT
2361 .compat_ioctl = btrfs_compat_ioctl,
2365 static struct address_space_operations btrfs_aops = {
2366 .readpage = btrfs_readpage,
2367 .writepage = btrfs_writepage,
2368 .sync_page = block_sync_page,
2369 .prepare_write = btrfs_prepare_write,
2370 .commit_write = btrfs_commit_write,
2372 .invalidatepage = btrfs_invalidatepage,
2373 .releasepage = btrfs_releasepage,
2374 .set_page_dirty = __set_page_dirty_nobuffers,
2377 static struct address_space_operations btrfs_symlink_aops = {
2378 .readpage = btrfs_readpage,
2379 .writepage = btrfs_writepage,
2382 static struct inode_operations btrfs_file_inode_operations = {
2383 .truncate = btrfs_truncate,
2384 .getattr = btrfs_getattr,
2385 .setattr = btrfs_setattr,
2388 static struct inode_operations btrfs_special_inode_operations = {
2389 .getattr = btrfs_getattr,
2390 .setattr = btrfs_setattr,
2393 static struct inode_operations btrfs_symlink_inode_operations = {
2394 .readlink = generic_readlink,
2395 .follow_link = page_follow_link_light,
2396 .put_link = page_put_link,