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;
55 static struct extent_map_ops btrfs_extent_map_ops;
57 static struct kmem_cache *btrfs_inode_cachep;
58 struct kmem_cache *btrfs_trans_handle_cachep;
59 struct kmem_cache *btrfs_transaction_cachep;
60 struct kmem_cache *btrfs_bit_radix_cachep;
61 struct kmem_cache *btrfs_path_cachep;
64 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
65 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
66 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
67 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
68 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
69 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
70 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
71 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
74 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
76 struct btrfs_root *root = BTRFS_I(inode)->root;
77 struct btrfs_trans_handle *trans;
82 u64 blocksize = root->sectorsize;
84 mutex_lock(&root->fs_info->fs_mutex);
85 trans = btrfs_start_transaction(root, 1);
86 btrfs_set_trans_block_group(trans, inode);
88 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
89 ret = btrfs_drop_extents(trans, root, inode,
90 start, start + num_bytes, start, &alloc_hint);
92 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 0,
93 alloc_hint, (u64)-1, &ins, 1);
98 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
99 start, ins.objectid, ins.offset,
102 btrfs_end_transaction(trans, root);
103 mutex_unlock(&root->fs_info->fs_mutex);
107 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
109 struct inode *inode = page->mapping->host;
110 struct btrfs_root *root = BTRFS_I(inode)->root;
111 struct btrfs_trans_handle *trans;
114 u64 page_start = page->index << PAGE_CACHE_SHIFT;
115 size_t offset = start - page_start;
117 mutex_lock(&root->fs_info->fs_mutex);
118 trans = btrfs_start_transaction(root, 1);
119 btrfs_set_trans_block_group(trans, inode);
121 btrfs_csum_file_block(trans, root, inode, inode->i_ino,
122 start, kaddr + offset, end - start + 1);
124 ret = btrfs_end_transaction(trans, root);
126 mutex_unlock(&root->fs_info->fs_mutex);
130 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
133 struct inode *inode = page->mapping->host;
134 struct btrfs_root *root = BTRFS_I(inode)->root;
135 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
136 struct btrfs_csum_item *item;
137 struct btrfs_path *path = NULL;
140 mutex_lock(&root->fs_info->fs_mutex);
141 path = btrfs_alloc_path();
142 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
145 /* a csum that isn't present is a preallocated region. */
146 if (ret == -ENOENT || ret == -EFBIG)
151 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
153 set_state_private(em_tree, start, csum);
156 btrfs_free_path(path);
157 mutex_unlock(&root->fs_info->fs_mutex);
161 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
163 size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
164 struct inode *inode = page->mapping->host;
165 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
169 struct btrfs_root *root = BTRFS_I(inode)->root;
173 ret = get_state_private(em_tree, start, &private);
174 local_irq_save(flags);
175 kaddr = kmap_atomic(page, KM_IRQ0);
179 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
180 btrfs_csum_final(csum, (char *)&csum);
181 if (csum != private) {
184 kunmap_atomic(kaddr, KM_IRQ0);
185 local_irq_restore(flags);
189 printk("btrfs csum failed ino %lu off %llu\n",
190 page->mapping->host->i_ino, (unsigned long long)start);
191 memset(kaddr + offset, 1, end - start + 1);
192 flush_dcache_page(page);
193 kunmap_atomic(kaddr, KM_IRQ0);
194 local_irq_restore(flags);
198 void btrfs_read_locked_inode(struct inode *inode)
200 struct btrfs_path *path;
201 struct extent_buffer *leaf;
202 struct btrfs_inode_item *inode_item;
203 struct btrfs_inode_timespec *tspec;
204 struct btrfs_root *root = BTRFS_I(inode)->root;
205 struct btrfs_key location;
206 u64 alloc_group_block;
210 path = btrfs_alloc_path();
212 mutex_lock(&root->fs_info->fs_mutex);
214 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
215 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
219 leaf = path->nodes[0];
220 inode_item = btrfs_item_ptr(leaf, path->slots[0],
221 struct btrfs_inode_item);
223 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
224 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
225 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
226 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
227 inode->i_size = btrfs_inode_size(leaf, inode_item);
229 tspec = btrfs_inode_atime(inode_item);
230 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
231 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
233 tspec = btrfs_inode_mtime(inode_item);
234 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
235 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
237 tspec = btrfs_inode_ctime(inode_item);
238 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
239 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
241 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
242 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
244 rdev = btrfs_inode_rdev(leaf, inode_item);
246 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
247 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
250 btrfs_free_path(path);
253 mutex_unlock(&root->fs_info->fs_mutex);
255 switch (inode->i_mode & S_IFMT) {
257 inode->i_mapping->a_ops = &btrfs_aops;
258 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
259 inode->i_fop = &btrfs_file_operations;
260 inode->i_op = &btrfs_file_inode_operations;
263 inode->i_fop = &btrfs_dir_file_operations;
264 if (root == root->fs_info->tree_root)
265 inode->i_op = &btrfs_dir_ro_inode_operations;
267 inode->i_op = &btrfs_dir_inode_operations;
270 inode->i_op = &btrfs_symlink_inode_operations;
271 inode->i_mapping->a_ops = &btrfs_symlink_aops;
274 init_special_inode(inode, inode->i_mode, rdev);
280 btrfs_release_path(root, path);
281 btrfs_free_path(path);
282 mutex_unlock(&root->fs_info->fs_mutex);
283 make_bad_inode(inode);
286 static void fill_inode_item(struct extent_buffer *leaf,
287 struct btrfs_inode_item *item,
290 btrfs_set_inode_uid(leaf, item, inode->i_uid);
291 btrfs_set_inode_gid(leaf, item, inode->i_gid);
292 btrfs_set_inode_size(leaf, item, inode->i_size);
293 btrfs_set_inode_mode(leaf, item, inode->i_mode);
294 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
296 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
297 inode->i_atime.tv_sec);
298 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
299 inode->i_atime.tv_nsec);
301 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
302 inode->i_mtime.tv_sec);
303 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
304 inode->i_mtime.tv_nsec);
306 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
307 inode->i_ctime.tv_sec);
308 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
309 inode->i_ctime.tv_nsec);
311 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
312 btrfs_set_inode_generation(leaf, item, inode->i_generation);
313 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
314 btrfs_set_inode_block_group(leaf, item,
315 BTRFS_I(inode)->block_group->key.objectid);
318 int btrfs_update_inode(struct btrfs_trans_handle *trans,
319 struct btrfs_root *root,
322 struct btrfs_inode_item *inode_item;
323 struct btrfs_path *path;
324 struct extent_buffer *leaf;
327 path = btrfs_alloc_path();
329 ret = btrfs_lookup_inode(trans, root, path,
330 &BTRFS_I(inode)->location, 1);
337 leaf = path->nodes[0];
338 inode_item = btrfs_item_ptr(leaf, path->slots[0],
339 struct btrfs_inode_item);
341 fill_inode_item(leaf, inode_item, inode);
342 btrfs_mark_buffer_dirty(leaf);
343 btrfs_set_inode_last_trans(trans, inode);
346 btrfs_release_path(root, path);
347 btrfs_free_path(path);
352 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
353 struct btrfs_root *root,
355 struct dentry *dentry)
357 struct btrfs_path *path;
358 const char *name = dentry->d_name.name;
359 int name_len = dentry->d_name.len;
361 struct extent_buffer *leaf;
362 struct btrfs_dir_item *di;
363 struct btrfs_key key;
365 path = btrfs_alloc_path();
371 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
381 leaf = path->nodes[0];
382 btrfs_dir_item_key_to_cpu(leaf, di, &key);
383 ret = btrfs_delete_one_dir_name(trans, root, path, di);
386 btrfs_release_path(root, path);
388 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
389 key.objectid, name, name_len, -1);
398 ret = btrfs_delete_one_dir_name(trans, root, path, di);
400 dentry->d_inode->i_ctime = dir->i_ctime;
402 btrfs_free_path(path);
404 dir->i_size -= name_len * 2;
405 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
406 btrfs_update_inode(trans, root, dir);
407 drop_nlink(dentry->d_inode);
408 ret = btrfs_update_inode(trans, root, dentry->d_inode);
409 dir->i_sb->s_dirt = 1;
414 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
416 struct btrfs_root *root;
417 struct btrfs_trans_handle *trans;
421 root = BTRFS_I(dir)->root;
422 mutex_lock(&root->fs_info->fs_mutex);
423 trans = btrfs_start_transaction(root, 1);
425 btrfs_set_trans_block_group(trans, dir);
426 ret = btrfs_unlink_trans(trans, root, dir, dentry);
427 nr = trans->blocks_used;
429 btrfs_end_transaction(trans, root);
430 mutex_unlock(&root->fs_info->fs_mutex);
431 btrfs_btree_balance_dirty(root, nr);
436 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
438 struct inode *inode = dentry->d_inode;
441 struct btrfs_root *root = BTRFS_I(dir)->root;
442 struct btrfs_path *path;
443 struct btrfs_key key;
444 struct btrfs_trans_handle *trans;
445 struct btrfs_key found_key;
447 struct extent_buffer *leaf;
448 char *goodnames = "..";
451 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
454 path = btrfs_alloc_path();
456 mutex_lock(&root->fs_info->fs_mutex);
457 trans = btrfs_start_transaction(root, 1);
459 btrfs_set_trans_block_group(trans, dir);
460 key.objectid = inode->i_ino;
461 key.offset = (u64)-1;
464 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
470 if (path->slots[0] == 0) {
475 leaf = path->nodes[0];
476 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
477 found_type = btrfs_key_type(&found_key);
478 if (found_key.objectid != inode->i_ino) {
482 if ((found_type != BTRFS_DIR_ITEM_KEY &&
483 found_type != BTRFS_DIR_INDEX_KEY) ||
484 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
485 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
489 ret = btrfs_del_item(trans, root, path);
492 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
494 btrfs_release_path(root, path);
497 btrfs_release_path(root, path);
499 /* now the directory is empty */
500 err = btrfs_unlink_trans(trans, root, dir, dentry);
505 btrfs_release_path(root, path);
506 btrfs_free_path(path);
507 nr = trans->blocks_used;
508 ret = btrfs_end_transaction(trans, root);
509 mutex_unlock(&root->fs_info->fs_mutex);
510 btrfs_btree_balance_dirty(root, nr);
516 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
517 struct btrfs_root *root,
520 struct btrfs_path *path;
525 path = btrfs_alloc_path();
527 ret = btrfs_lookup_inode(trans, root, path,
528 &BTRFS_I(inode)->location, -1);
532 ret = btrfs_del_item(trans, root, path);
533 btrfs_free_path(path);
538 * this can truncate away extent items, csum items and directory items.
539 * It starts at a high offset and removes keys until it can't find
540 * any higher than i_size.
542 * csum items that cross the new i_size are truncated to the new size
545 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
546 struct btrfs_root *root,
550 struct btrfs_path *path;
551 struct btrfs_key key;
552 struct btrfs_key found_key;
554 struct extent_buffer *leaf;
555 struct btrfs_file_extent_item *fi;
556 u64 extent_start = 0;
557 u64 extent_num_bytes = 0;
562 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
563 path = btrfs_alloc_path();
567 /* FIXME, add redo link to tree so we don't leak on crash */
568 key.objectid = inode->i_ino;
569 key.offset = (u64)-1;
573 btrfs_init_path(path);
575 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
580 BUG_ON(path->slots[0] == 0);
583 leaf = path->nodes[0];
584 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
585 found_type = btrfs_key_type(&found_key);
587 if (found_key.objectid != inode->i_ino)
590 if (found_type != BTRFS_CSUM_ITEM_KEY &&
591 found_type != BTRFS_DIR_ITEM_KEY &&
592 found_type != BTRFS_DIR_INDEX_KEY &&
593 found_type != BTRFS_EXTENT_DATA_KEY)
596 item_end = found_key.offset;
597 if (found_type == BTRFS_EXTENT_DATA_KEY) {
598 fi = btrfs_item_ptr(leaf, path->slots[0],
599 struct btrfs_file_extent_item);
600 if (btrfs_file_extent_type(leaf, fi) !=
601 BTRFS_FILE_EXTENT_INLINE) {
603 btrfs_file_extent_num_bytes(leaf, fi);
606 if (found_type == BTRFS_CSUM_ITEM_KEY) {
607 ret = btrfs_csum_truncate(trans, root, path,
611 if (item_end < inode->i_size) {
612 if (found_type == BTRFS_DIR_ITEM_KEY) {
613 found_type = BTRFS_INODE_ITEM_KEY;
614 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
615 found_type = BTRFS_CSUM_ITEM_KEY;
616 } else if (found_type) {
621 btrfs_set_key_type(&key, found_type);
622 btrfs_release_path(root, path);
625 if (found_key.offset >= inode->i_size)
631 /* FIXME, shrink the extent if the ref count is only 1 */
632 if (found_type == BTRFS_EXTENT_DATA_KEY &&
633 btrfs_file_extent_type(leaf, fi) !=
634 BTRFS_FILE_EXTENT_INLINE) {
636 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
639 btrfs_file_extent_num_bytes(leaf, fi);
640 extent_num_bytes = inode->i_size -
641 found_key.offset + root->sectorsize - 1;
642 btrfs_set_file_extent_num_bytes(leaf, fi,
644 num_dec = (orig_num_bytes -
645 extent_num_bytes) >> 9;
646 if (extent_start != 0) {
647 inode->i_blocks -= num_dec;
649 btrfs_mark_buffer_dirty(leaf);
652 btrfs_file_extent_disk_num_bytes(leaf,
654 /* FIXME blocksize != 4096 */
655 num_dec = btrfs_file_extent_num_bytes(leaf,
657 if (extent_start != 0) {
659 inode->i_blocks -= num_dec;
664 ret = btrfs_del_item(trans, root, path);
670 btrfs_release_path(root, path);
672 ret = btrfs_free_extent(trans, root, extent_start,
673 extent_num_bytes, 0);
679 btrfs_release_path(root, path);
680 btrfs_free_path(path);
681 inode->i_sb->s_dirt = 1;
685 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
690 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
691 u64 page_start = page->index << PAGE_CACHE_SHIFT;
692 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
694 set_page_extent_mapped(page);
696 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
697 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
699 if (zero_start != PAGE_CACHE_SIZE) {
701 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
702 flush_dcache_page(page);
705 set_page_dirty(page);
706 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
712 * taken from block_truncate_page, but does cow as it zeros out
713 * any bytes left in the last page in the file.
715 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
717 struct inode *inode = mapping->host;
718 struct btrfs_root *root = BTRFS_I(inode)->root;
719 u32 blocksize = root->sectorsize;
720 pgoff_t index = from >> PAGE_CACHE_SHIFT;
721 unsigned offset = from & (PAGE_CACHE_SIZE-1);
726 if ((offset & (blocksize - 1)) == 0)
729 down_read(&root->snap_sem);
731 page = grab_cache_page(mapping, index);
734 if (!PageUptodate(page)) {
735 ret = btrfs_readpage(NULL, page);
737 if (!PageUptodate(page)) {
742 page_start = page->index << PAGE_CACHE_SHIFT;
744 ret = btrfs_cow_one_page(inode, page, offset);
747 page_cache_release(page);
748 up_read(&BTRFS_I(inode)->root->snap_sem);
753 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
755 struct inode *inode = dentry->d_inode;
758 err = inode_change_ok(inode, attr);
762 if (S_ISREG(inode->i_mode) &&
763 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
764 struct btrfs_trans_handle *trans;
765 struct btrfs_root *root = BTRFS_I(inode)->root;
766 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
768 u64 mask = root->sectorsize - 1;
769 u64 pos = (inode->i_size + mask) & ~mask;
770 u64 block_end = attr->ia_size | mask;
774 if (attr->ia_size <= pos)
777 btrfs_truncate_page(inode->i_mapping, inode->i_size);
779 lock_extent(em_tree, pos, block_end, GFP_NOFS);
780 hole_size = (attr->ia_size - pos + mask) & ~mask;
782 mutex_lock(&root->fs_info->fs_mutex);
783 trans = btrfs_start_transaction(root, 1);
784 btrfs_set_trans_block_group(trans, inode);
785 err = btrfs_drop_extents(trans, root, inode,
786 pos, pos + hole_size, pos,
789 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
790 pos, 0, 0, hole_size);
791 btrfs_end_transaction(trans, root);
792 mutex_unlock(&root->fs_info->fs_mutex);
793 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
798 err = inode_setattr(inode, attr);
802 void btrfs_delete_inode(struct inode *inode)
804 struct btrfs_trans_handle *trans;
805 struct btrfs_root *root = BTRFS_I(inode)->root;
809 truncate_inode_pages(&inode->i_data, 0);
810 if (is_bad_inode(inode)) {
815 mutex_lock(&root->fs_info->fs_mutex);
816 trans = btrfs_start_transaction(root, 1);
818 btrfs_set_trans_block_group(trans, inode);
819 ret = btrfs_truncate_in_trans(trans, root, inode);
822 ret = btrfs_free_inode(trans, root, inode);
825 nr = trans->blocks_used;
827 btrfs_end_transaction(trans, root);
828 mutex_unlock(&root->fs_info->fs_mutex);
829 btrfs_btree_balance_dirty(root, nr);
833 nr = trans->blocks_used;
834 btrfs_end_transaction(trans, root);
835 mutex_unlock(&root->fs_info->fs_mutex);
836 btrfs_btree_balance_dirty(root, nr);
842 * this returns the key found in the dir entry in the location pointer.
843 * If no dir entries were found, location->objectid is 0.
845 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
846 struct btrfs_key *location)
848 const char *name = dentry->d_name.name;
849 int namelen = dentry->d_name.len;
850 struct btrfs_dir_item *di;
851 struct btrfs_path *path;
852 struct btrfs_root *root = BTRFS_I(dir)->root;
855 path = btrfs_alloc_path();
857 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
861 if (!di || IS_ERR(di)) {
862 location->objectid = 0;
865 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
867 btrfs_release_path(root, path);
868 btrfs_free_path(path);
873 * when we hit a tree root in a directory, the btrfs part of the inode
874 * needs to be changed to reflect the root directory of the tree root. This
875 * is kind of like crossing a mount point.
877 static int fixup_tree_root_location(struct btrfs_root *root,
878 struct btrfs_key *location,
879 struct btrfs_root **sub_root,
880 struct dentry *dentry)
882 struct btrfs_path *path;
883 struct btrfs_root_item *ri;
885 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
887 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
890 path = btrfs_alloc_path();
892 mutex_lock(&root->fs_info->fs_mutex);
894 *sub_root = btrfs_read_fs_root(root->fs_info, location,
897 if (IS_ERR(*sub_root))
898 return PTR_ERR(*sub_root);
900 ri = &(*sub_root)->root_item;
901 location->objectid = btrfs_root_dirid(ri);
902 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
903 location->offset = 0;
905 btrfs_free_path(path);
906 mutex_unlock(&root->fs_info->fs_mutex);
910 static int btrfs_init_locked_inode(struct inode *inode, void *p)
912 struct btrfs_iget_args *args = p;
913 inode->i_ino = args->ino;
914 BTRFS_I(inode)->root = args->root;
915 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
916 inode->i_mapping, GFP_NOFS);
920 static int btrfs_find_actor(struct inode *inode, void *opaque)
922 struct btrfs_iget_args *args = opaque;
923 return (args->ino == inode->i_ino &&
924 args->root == BTRFS_I(inode)->root);
927 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
928 struct btrfs_root *root)
931 struct btrfs_iget_args args;
935 inode = iget5_locked(s, objectid, btrfs_find_actor,
936 btrfs_init_locked_inode,
941 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
942 struct nameidata *nd)
944 struct inode * inode;
945 struct btrfs_inode *bi = BTRFS_I(dir);
946 struct btrfs_root *root = bi->root;
947 struct btrfs_root *sub_root = root;
948 struct btrfs_key location;
951 if (dentry->d_name.len > BTRFS_NAME_LEN)
952 return ERR_PTR(-ENAMETOOLONG);
954 mutex_lock(&root->fs_info->fs_mutex);
955 ret = btrfs_inode_by_name(dir, dentry, &location);
956 mutex_unlock(&root->fs_info->fs_mutex);
962 if (location.objectid) {
963 ret = fixup_tree_root_location(root, &location, &sub_root,
968 return ERR_PTR(-ENOENT);
969 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
972 return ERR_PTR(-EACCES);
973 if (inode->i_state & I_NEW) {
974 /* the inode and parent dir are two different roots */
975 if (sub_root != root) {
977 sub_root->inode = inode;
979 BTRFS_I(inode)->root = sub_root;
980 memcpy(&BTRFS_I(inode)->location, &location,
982 btrfs_read_locked_inode(inode);
983 unlock_new_inode(inode);
986 return d_splice_alias(inode, dentry);
989 static unsigned char btrfs_filetype_table[] = {
990 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
993 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
995 struct inode *inode = filp->f_path.dentry->d_inode;
996 struct btrfs_root *root = BTRFS_I(inode)->root;
997 struct btrfs_item *item;
998 struct btrfs_dir_item *di;
999 struct btrfs_key key;
1000 struct btrfs_key found_key;
1001 struct btrfs_path *path;
1004 struct extent_buffer *leaf;
1007 unsigned char d_type;
1012 int key_type = BTRFS_DIR_INDEX_KEY;
1017 /* FIXME, use a real flag for deciding about the key type */
1018 if (root->fs_info->tree_root == root)
1019 key_type = BTRFS_DIR_ITEM_KEY;
1021 mutex_lock(&root->fs_info->fs_mutex);
1022 key.objectid = inode->i_ino;
1023 btrfs_set_key_type(&key, key_type);
1024 key.offset = filp->f_pos;
1026 path = btrfs_alloc_path();
1028 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1033 leaf = path->nodes[0];
1034 nritems = btrfs_header_nritems(leaf);
1035 slot = path->slots[0];
1036 if (advance || slot >= nritems) {
1037 if (slot >= nritems -1) {
1038 ret = btrfs_next_leaf(root, path);
1041 leaf = path->nodes[0];
1042 nritems = btrfs_header_nritems(leaf);
1043 slot = path->slots[0];
1050 item = btrfs_item_nr(leaf, slot);
1051 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1053 if (found_key.objectid != key.objectid)
1055 if (btrfs_key_type(&found_key) != key_type)
1057 if (found_key.offset < filp->f_pos)
1060 filp->f_pos = found_key.offset;
1062 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1064 di_total = btrfs_item_size(leaf, item);
1065 while(di_cur < di_total) {
1066 struct btrfs_key location;
1068 name_len = btrfs_dir_name_len(leaf, di);
1069 if (name_len < 32) {
1070 name_ptr = tmp_name;
1072 name_ptr = kmalloc(name_len, GFP_NOFS);
1075 read_extent_buffer(leaf, name_ptr,
1076 (unsigned long)(di + 1), name_len);
1078 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1079 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1081 over = filldir(dirent, name_ptr, name_len,
1086 if (name_ptr != tmp_name)
1091 di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
1093 di = (struct btrfs_dir_item *)((char *)di + di_len);
1100 btrfs_release_path(root, path);
1101 btrfs_free_path(path);
1102 mutex_unlock(&root->fs_info->fs_mutex);
1106 int btrfs_write_inode(struct inode *inode, int wait)
1108 struct btrfs_root *root = BTRFS_I(inode)->root;
1109 struct btrfs_trans_handle *trans;
1113 mutex_lock(&root->fs_info->fs_mutex);
1114 trans = btrfs_start_transaction(root, 1);
1115 btrfs_set_trans_block_group(trans, inode);
1116 ret = btrfs_commit_transaction(trans, root);
1117 mutex_unlock(&root->fs_info->fs_mutex);
1123 * This is somewhat expensive, updating the tree every time the
1124 * inode changes. But, it is most likely to find the inode in cache.
1125 * FIXME, needs more benchmarking...there are no reasons other than performance
1126 * to keep or drop this code.
1128 void btrfs_dirty_inode(struct inode *inode)
1130 struct btrfs_root *root = BTRFS_I(inode)->root;
1131 struct btrfs_trans_handle *trans;
1133 mutex_lock(&root->fs_info->fs_mutex);
1134 trans = btrfs_start_transaction(root, 1);
1135 btrfs_set_trans_block_group(trans, inode);
1136 btrfs_update_inode(trans, root, inode);
1137 btrfs_end_transaction(trans, root);
1138 mutex_unlock(&root->fs_info->fs_mutex);
1141 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1142 struct btrfs_root *root,
1144 struct btrfs_block_group_cache *group,
1147 struct inode *inode;
1148 struct btrfs_inode_item *inode_item;
1149 struct btrfs_key *location;
1150 struct btrfs_path *path;
1154 path = btrfs_alloc_path();
1157 inode = new_inode(root->fs_info->sb);
1159 return ERR_PTR(-ENOMEM);
1161 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1162 inode->i_mapping, GFP_NOFS);
1163 BTRFS_I(inode)->root = root;
1169 group = btrfs_find_block_group(root, group, 0, 0, owner);
1170 BTRFS_I(inode)->block_group = group;
1172 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1176 inode->i_uid = current->fsuid;
1177 inode->i_gid = current->fsgid;
1178 inode->i_mode = mode;
1179 inode->i_ino = objectid;
1180 inode->i_blocks = 0;
1181 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1182 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1183 struct btrfs_inode_item);
1184 fill_inode_item(path->nodes[0], inode_item, inode);
1185 btrfs_mark_buffer_dirty(path->nodes[0]);
1186 btrfs_free_path(path);
1188 location = &BTRFS_I(inode)->location;
1189 location->objectid = objectid;
1190 location->offset = 0;
1191 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1193 insert_inode_hash(inode);
1196 btrfs_free_path(path);
1197 return ERR_PTR(ret);
1200 static inline u8 btrfs_inode_type(struct inode *inode)
1202 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1205 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1206 struct dentry *dentry, struct inode *inode)
1209 struct btrfs_key key;
1210 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1211 struct inode *parent_inode;
1213 key.objectid = inode->i_ino;
1214 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1217 ret = btrfs_insert_dir_item(trans, root,
1218 dentry->d_name.name, dentry->d_name.len,
1219 dentry->d_parent->d_inode->i_ino,
1220 &key, btrfs_inode_type(inode));
1222 parent_inode = dentry->d_parent->d_inode;
1223 parent_inode->i_size += dentry->d_name.len * 2;
1224 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1225 ret = btrfs_update_inode(trans, root,
1226 dentry->d_parent->d_inode);
1231 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1232 struct dentry *dentry, struct inode *inode)
1234 int err = btrfs_add_link(trans, dentry, inode);
1236 d_instantiate(dentry, inode);
1244 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1245 int mode, dev_t rdev)
1247 struct btrfs_trans_handle *trans;
1248 struct btrfs_root *root = BTRFS_I(dir)->root;
1249 struct inode *inode;
1255 if (!new_valid_dev(rdev))
1258 mutex_lock(&root->fs_info->fs_mutex);
1259 trans = btrfs_start_transaction(root, 1);
1260 btrfs_set_trans_block_group(trans, dir);
1262 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1268 inode = btrfs_new_inode(trans, root, objectid,
1269 BTRFS_I(dir)->block_group, mode);
1270 err = PTR_ERR(inode);
1274 btrfs_set_trans_block_group(trans, inode);
1275 err = btrfs_add_nondir(trans, dentry, inode);
1279 inode->i_op = &btrfs_special_inode_operations;
1280 init_special_inode(inode, inode->i_mode, rdev);
1281 btrfs_update_inode(trans, root, inode);
1283 dir->i_sb->s_dirt = 1;
1284 btrfs_update_inode_block_group(trans, inode);
1285 btrfs_update_inode_block_group(trans, dir);
1287 nr = trans->blocks_used;
1288 btrfs_end_transaction(trans, root);
1289 mutex_unlock(&root->fs_info->fs_mutex);
1292 inode_dec_link_count(inode);
1295 btrfs_btree_balance_dirty(root, nr);
1299 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1300 int mode, struct nameidata *nd)
1302 struct btrfs_trans_handle *trans;
1303 struct btrfs_root *root = BTRFS_I(dir)->root;
1304 struct inode *inode;
1310 mutex_lock(&root->fs_info->fs_mutex);
1311 trans = btrfs_start_transaction(root, 1);
1312 btrfs_set_trans_block_group(trans, dir);
1314 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1320 inode = btrfs_new_inode(trans, root, objectid,
1321 BTRFS_I(dir)->block_group, mode);
1322 err = PTR_ERR(inode);
1326 btrfs_set_trans_block_group(trans, inode);
1327 err = btrfs_add_nondir(trans, dentry, inode);
1331 inode->i_mapping->a_ops = &btrfs_aops;
1332 inode->i_fop = &btrfs_file_operations;
1333 inode->i_op = &btrfs_file_inode_operations;
1334 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1335 inode->i_mapping, GFP_NOFS);
1336 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
1338 dir->i_sb->s_dirt = 1;
1339 btrfs_update_inode_block_group(trans, inode);
1340 btrfs_update_inode_block_group(trans, dir);
1342 nr = trans->blocks_used;
1343 btrfs_end_transaction(trans, root);
1344 mutex_unlock(&root->fs_info->fs_mutex);
1347 inode_dec_link_count(inode);
1350 btrfs_btree_balance_dirty(root, nr);
1354 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1355 struct dentry *dentry)
1357 struct btrfs_trans_handle *trans;
1358 struct btrfs_root *root = BTRFS_I(dir)->root;
1359 struct inode *inode = old_dentry->d_inode;
1364 if (inode->i_nlink == 0)
1368 mutex_lock(&root->fs_info->fs_mutex);
1369 trans = btrfs_start_transaction(root, 1);
1371 btrfs_set_trans_block_group(trans, dir);
1372 atomic_inc(&inode->i_count);
1373 err = btrfs_add_nondir(trans, dentry, inode);
1378 dir->i_sb->s_dirt = 1;
1379 btrfs_update_inode_block_group(trans, dir);
1380 err = btrfs_update_inode(trans, root, inode);
1385 nr = trans->blocks_used;
1386 btrfs_end_transaction(trans, root);
1387 mutex_unlock(&root->fs_info->fs_mutex);
1390 inode_dec_link_count(inode);
1393 btrfs_btree_balance_dirty(root, nr);
1397 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1398 struct btrfs_root *root,
1399 u64 objectid, u64 dirid)
1403 struct btrfs_key key;
1408 key.objectid = objectid;
1410 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1412 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1413 &key, BTRFS_FT_DIR);
1417 key.objectid = dirid;
1418 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1419 &key, BTRFS_FT_DIR);
1426 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1428 struct inode *inode;
1429 struct btrfs_trans_handle *trans;
1430 struct btrfs_root *root = BTRFS_I(dir)->root;
1432 int drop_on_err = 0;
1434 unsigned long nr = 1;
1436 mutex_lock(&root->fs_info->fs_mutex);
1437 trans = btrfs_start_transaction(root, 1);
1438 btrfs_set_trans_block_group(trans, dir);
1440 if (IS_ERR(trans)) {
1441 err = PTR_ERR(trans);
1445 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1451 inode = btrfs_new_inode(trans, root, objectid,
1452 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1453 if (IS_ERR(inode)) {
1454 err = PTR_ERR(inode);
1459 inode->i_op = &btrfs_dir_inode_operations;
1460 inode->i_fop = &btrfs_dir_file_operations;
1461 btrfs_set_trans_block_group(trans, inode);
1463 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1468 err = btrfs_update_inode(trans, root, inode);
1472 err = btrfs_add_link(trans, dentry, inode);
1476 d_instantiate(dentry, inode);
1478 dir->i_sb->s_dirt = 1;
1479 btrfs_update_inode_block_group(trans, inode);
1480 btrfs_update_inode_block_group(trans, dir);
1483 nr = trans->blocks_used;
1484 btrfs_end_transaction(trans, root);
1487 mutex_unlock(&root->fs_info->fs_mutex);
1490 btrfs_btree_balance_dirty(root, nr);
1494 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1495 size_t page_offset, u64 start, u64 end,
1501 u64 extent_start = 0;
1503 u64 objectid = inode->i_ino;
1505 int failed_insert = 0;
1506 struct btrfs_path *path;
1507 struct btrfs_root *root = BTRFS_I(inode)->root;
1508 struct btrfs_file_extent_item *item;
1509 struct extent_buffer *leaf;
1510 struct btrfs_key found_key;
1511 struct extent_map *em = NULL;
1512 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1513 struct btrfs_trans_handle *trans = NULL;
1515 path = btrfs_alloc_path();
1517 mutex_lock(&root->fs_info->fs_mutex);
1520 em = lookup_extent_mapping(em_tree, start, end);
1525 em = alloc_extent_map(GFP_NOFS);
1530 em->start = EXTENT_MAP_HOLE;
1531 em->end = EXTENT_MAP_HOLE;
1533 em->bdev = inode->i_sb->s_bdev;
1534 ret = btrfs_lookup_file_extent(NULL, root, path,
1535 objectid, start, 0);
1542 if (path->slots[0] == 0)
1547 leaf = path->nodes[0];
1548 item = btrfs_item_ptr(leaf, path->slots[0],
1549 struct btrfs_file_extent_item);
1550 /* are we inside the extent that was found? */
1551 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1552 found_type = btrfs_key_type(&found_key);
1553 if (found_key.objectid != objectid ||
1554 found_type != BTRFS_EXTENT_DATA_KEY) {
1558 found_type = btrfs_file_extent_type(leaf, item);
1559 extent_start = found_key.offset;
1560 if (found_type == BTRFS_FILE_EXTENT_REG) {
1561 extent_end = extent_start +
1562 btrfs_file_extent_num_bytes(leaf, item);
1564 if (start < extent_start || start >= extent_end) {
1566 if (start < extent_start) {
1567 if (end < extent_start)
1569 em->end = extent_end - 1;
1575 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1577 em->start = extent_start;
1578 em->end = extent_end - 1;
1579 em->block_start = EXTENT_MAP_HOLE;
1580 em->block_end = EXTENT_MAP_HOLE;
1583 bytenr += btrfs_file_extent_offset(leaf, item);
1584 em->block_start = bytenr;
1585 em->block_end = em->block_start +
1586 btrfs_file_extent_num_bytes(leaf, item) - 1;
1587 em->start = extent_start;
1588 em->end = extent_end - 1;
1590 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1594 size_t extent_offset;
1597 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1599 extent_end = (extent_start + size - 1) |
1600 ((u64)root->sectorsize - 1);
1601 if (start < extent_start || start >= extent_end) {
1603 if (start < extent_start) {
1604 if (end < extent_start)
1606 em->end = extent_end;
1612 em->block_start = EXTENT_MAP_INLINE;
1613 em->block_end = EXTENT_MAP_INLINE;
1616 em->start = extent_start;
1617 em->end = extent_start + size - 1;
1621 extent_offset = (page->index << PAGE_CACHE_SHIFT) -
1622 extent_start + page_offset;
1623 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
1624 size - extent_offset);
1625 em->start = extent_start + extent_offset;
1626 em->end = (em->start + copy_size -1) |
1627 ((u64)root->sectorsize -1);
1629 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
1630 read_extent_buffer(leaf, map + page_offset, ptr, copy_size);
1632 if (em->start + copy_size <= em->end) {
1633 size = min_t(u64, em->end + 1 - em->start,
1634 PAGE_CACHE_SIZE - page_offset) - copy_size;
1635 memset(map + page_offset + copy_size, 0, size);
1638 flush_dcache_page(page);
1640 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
1643 printk("unkknown found_type %d\n", found_type);
1650 em->block_start = EXTENT_MAP_HOLE;
1651 em->block_end = EXTENT_MAP_HOLE;
1653 btrfs_release_path(root, path);
1654 if (em->start > start || em->end < start) {
1655 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1659 ret = add_extent_mapping(em_tree, em);
1660 if (ret == -EEXIST) {
1661 free_extent_map(em);
1664 if (failed_insert > 5) {
1665 printk("failing to insert %Lu %Lu\n", start, end);
1673 btrfs_free_path(path);
1675 ret = btrfs_end_transaction(trans, root);
1679 mutex_unlock(&root->fs_info->fs_mutex);
1681 free_extent_map(em);
1683 return ERR_PTR(err);
1688 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
1690 return extent_bmap(mapping, iblock, btrfs_get_extent);
1693 static int btrfs_prepare_write(struct file *file, struct page *page,
1694 unsigned from, unsigned to)
1696 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1697 page->mapping->host, page, from, to,
1701 int btrfs_readpage(struct file *file, struct page *page)
1703 struct extent_map_tree *tree;
1704 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1705 return extent_read_full_page(tree, page, btrfs_get_extent);
1707 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1709 struct extent_map_tree *tree;
1712 if (current->flags & PF_MEMALLOC) {
1713 redirty_page_for_writepage(wbc, page);
1717 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1718 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
1721 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1723 struct extent_map_tree *tree;
1726 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1727 ret = try_release_extent_mapping(tree, page);
1729 ClearPagePrivate(page);
1730 set_page_private(page, 0);
1731 page_cache_release(page);
1736 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1738 struct extent_map_tree *tree;
1740 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1741 extent_invalidatepage(tree, page, offset);
1742 btrfs_releasepage(page, GFP_NOFS);
1746 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1747 * called from a page fault handler when a page is first dirtied. Hence we must
1748 * be careful to check for EOF conditions here. We set the page up correctly
1749 * for a written page which means we get ENOSPC checking when writing into
1750 * holes and correct delalloc and unwritten extent mapping on filesystems that
1751 * support these features.
1753 * We are not allowed to take the i_mutex here so we have to play games to
1754 * protect against truncate races as the page could now be beyond EOF. Because
1755 * vmtruncate() writes the inode size before removing pages, once we have the
1756 * page lock we can determine safely if the page is beyond EOF. If it is not
1757 * beyond EOF, then the page is guaranteed safe against truncation until we
1760 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1762 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1768 down_read(&BTRFS_I(inode)->root->snap_sem);
1770 wait_on_page_writeback(page);
1771 size = i_size_read(inode);
1772 page_start = page->index << PAGE_CACHE_SHIFT;
1774 if ((page->mapping != inode->i_mapping) ||
1775 (page_start > size)) {
1776 /* page got truncated out from underneath us */
1780 /* page is wholly or partially inside EOF */
1781 if (page_start + PAGE_CACHE_SIZE > size)
1782 end = size & ~PAGE_CACHE_MASK;
1784 end = PAGE_CACHE_SIZE;
1786 ret = btrfs_cow_one_page(inode, page, end);
1789 up_read(&BTRFS_I(inode)->root->snap_sem);
1794 static void btrfs_truncate(struct inode *inode)
1796 struct btrfs_root *root = BTRFS_I(inode)->root;
1798 struct btrfs_trans_handle *trans;
1801 if (!S_ISREG(inode->i_mode))
1803 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1806 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1808 mutex_lock(&root->fs_info->fs_mutex);
1809 trans = btrfs_start_transaction(root, 1);
1810 btrfs_set_trans_block_group(trans, inode);
1812 /* FIXME, add redo link to tree so we don't leak on crash */
1813 ret = btrfs_truncate_in_trans(trans, root, inode);
1814 btrfs_update_inode(trans, root, inode);
1815 nr = trans->blocks_used;
1817 ret = btrfs_end_transaction(trans, root);
1819 mutex_unlock(&root->fs_info->fs_mutex);
1820 btrfs_btree_balance_dirty(root, nr);
1823 int btrfs_commit_write(struct file *file, struct page *page,
1824 unsigned from, unsigned to)
1826 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1827 page->mapping->host, page, from, to);
1830 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1832 struct btrfs_trans_handle *trans;
1833 struct btrfs_key key;
1834 struct btrfs_root_item root_item;
1835 struct btrfs_inode_item *inode_item;
1836 struct extent_buffer *leaf;
1837 struct btrfs_root *new_root;
1838 struct inode *inode;
1843 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1844 unsigned long nr = 1;
1846 mutex_lock(&root->fs_info->fs_mutex);
1847 trans = btrfs_start_transaction(root, 1);
1850 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
1852 return PTR_ERR(leaf);
1854 btrfs_set_header_nritems(leaf, 0);
1855 btrfs_set_header_level(leaf, 0);
1856 btrfs_set_header_bytenr(leaf, leaf->start);
1857 btrfs_set_header_generation(leaf, trans->transid);
1858 btrfs_set_header_owner(leaf, root->root_key.objectid);
1859 write_extent_buffer(leaf, root->fs_info->fsid,
1860 (unsigned long)btrfs_header_fsid(leaf),
1862 btrfs_mark_buffer_dirty(leaf);
1864 inode_item = &root_item.inode;
1865 memset(inode_item, 0, sizeof(*inode_item));
1866 inode_item->generation = cpu_to_le64(1);
1867 inode_item->size = cpu_to_le64(3);
1868 inode_item->nlink = cpu_to_le32(1);
1869 inode_item->nblocks = cpu_to_le64(1);
1870 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1872 btrfs_set_root_bytenr(&root_item, leaf->start);
1873 btrfs_set_root_level(&root_item, 0);
1874 btrfs_set_root_refs(&root_item, 1);
1875 btrfs_set_root_used(&root_item, 0);
1877 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1878 root_item.drop_level = 0;
1880 free_extent_buffer(leaf);
1883 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1888 btrfs_set_root_dirid(&root_item, new_dirid);
1890 key.objectid = objectid;
1892 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1893 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1899 * insert the directory item
1901 key.offset = (u64)-1;
1902 dir = root->fs_info->sb->s_root->d_inode;
1903 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1904 name, namelen, dir->i_ino, &key,
1909 ret = btrfs_commit_transaction(trans, root);
1913 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1916 trans = btrfs_start_transaction(new_root, 1);
1919 inode = btrfs_new_inode(trans, new_root, new_dirid,
1920 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
1923 inode->i_op = &btrfs_dir_inode_operations;
1924 inode->i_fop = &btrfs_dir_file_operations;
1925 new_root->inode = inode;
1927 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1933 ret = btrfs_update_inode(trans, new_root, inode);
1937 nr = trans->blocks_used;
1938 err = btrfs_commit_transaction(trans, root);
1942 mutex_unlock(&root->fs_info->fs_mutex);
1943 btrfs_btree_balance_dirty(root, nr);
1947 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1949 struct btrfs_trans_handle *trans;
1950 struct btrfs_key key;
1951 struct btrfs_root_item new_root_item;
1952 struct extent_buffer *tmp;
1958 if (!root->ref_cows)
1961 down_write(&root->snap_sem);
1962 freeze_bdev(root->fs_info->sb->s_bdev);
1963 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
1965 mutex_lock(&root->fs_info->fs_mutex);
1966 trans = btrfs_start_transaction(root, 1);
1969 ret = btrfs_update_inode(trans, root, root->inode);
1973 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1978 memcpy(&new_root_item, &root->root_item,
1979 sizeof(new_root_item));
1981 key.objectid = objectid;
1983 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1985 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
1986 btrfs_set_root_bytenr(&new_root_item, root->node->start);
1987 btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
1989 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1995 * insert the directory item
1997 key.offset = (u64)-1;
1998 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2000 root->fs_info->sb->s_root->d_inode->i_ino,
2001 &key, BTRFS_FT_DIR);
2006 ret = btrfs_inc_root_ref(trans, root);
2010 nr = trans->blocks_used;
2011 err = btrfs_commit_transaction(trans, root);
2016 mutex_unlock(&root->fs_info->fs_mutex);
2017 up_write(&root->snap_sem);
2018 btrfs_btree_balance_dirty(root, nr);
2022 static unsigned long force_ra(struct address_space *mapping,
2023 struct file_ra_state *ra, struct file *file,
2024 pgoff_t offset, pgoff_t last_index)
2028 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2029 req_size = last_index - offset + 1;
2030 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2033 req_size = min(last_index - offset + 1, (pgoff_t)128);
2034 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2035 return offset + req_size;
2039 int btrfs_defrag_file(struct file *file) {
2040 struct inode *inode = file->f_path.dentry->d_inode;
2041 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2043 unsigned long last_index;
2044 unsigned long ra_index = 0;
2049 mutex_lock(&inode->i_mutex);
2050 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2051 for (i = 0; i <= last_index; i++) {
2052 if (i == ra_index) {
2053 ra_index = force_ra(inode->i_mapping, &file->f_ra,
2054 file, ra_index, last_index);
2056 page = grab_cache_page(inode->i_mapping, i);
2059 if (!PageUptodate(page)) {
2060 btrfs_readpage(NULL, page);
2062 if (!PageUptodate(page)) {
2064 page_cache_release(page);
2068 page_start = page->index << PAGE_CACHE_SHIFT;
2069 page_end = page_start + PAGE_CACHE_SIZE - 1;
2071 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2072 set_extent_delalloc(em_tree, page_start,
2073 page_end, GFP_NOFS);
2074 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2075 set_page_dirty(page);
2077 page_cache_release(page);
2078 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2082 mutex_unlock(&inode->i_mutex);
2086 static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
2088 struct btrfs_ioctl_vol_args vol_args;
2089 struct btrfs_dir_item *di;
2090 struct btrfs_path *path;
2094 if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
2097 namelen = strlen(vol_args.name);
2098 if (namelen > BTRFS_VOL_NAME_MAX)
2100 if (strchr(vol_args.name, '/'))
2103 path = btrfs_alloc_path();
2107 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2108 mutex_lock(&root->fs_info->fs_mutex);
2109 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2111 vol_args.name, namelen, 0);
2112 mutex_unlock(&root->fs_info->fs_mutex);
2113 btrfs_free_path(path);
2114 if (di && !IS_ERR(di))
2119 if (root == root->fs_info->tree_root)
2120 return create_subvol(root, vol_args.name, namelen);
2121 return create_snapshot(root, vol_args.name, namelen);
2124 static int btrfs_ioctl_defrag(struct file *file)
2126 struct inode *inode = file->f_path.dentry->d_inode;
2127 struct btrfs_root *root = BTRFS_I(inode)->root;
2129 switch (inode->i_mode & S_IFMT) {
2131 mutex_lock(&root->fs_info->fs_mutex);
2132 btrfs_defrag_root(root, 0);
2133 btrfs_defrag_root(root->fs_info->extent_root, 0);
2134 mutex_unlock(&root->fs_info->fs_mutex);
2137 btrfs_defrag_file(file);
2144 long btrfs_ioctl(struct file *file, unsigned int
2145 cmd, unsigned long arg)
2147 struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root;
2150 case BTRFS_IOC_SNAP_CREATE:
2151 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2152 case BTRFS_IOC_DEFRAG:
2153 return btrfs_ioctl_defrag(file);
2160 * Called inside transaction, so use GFP_NOFS
2162 struct inode *btrfs_alloc_inode(struct super_block *sb)
2164 struct btrfs_inode *ei;
2166 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2170 return &ei->vfs_inode;
2173 void btrfs_destroy_inode(struct inode *inode)
2175 WARN_ON(!list_empty(&inode->i_dentry));
2176 WARN_ON(inode->i_data.nrpages);
2178 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2181 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2182 static void init_once(struct kmem_cache * cachep, void *foo)
2184 static void init_once(void * foo, struct kmem_cache * cachep,
2185 unsigned long flags)
2188 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2190 inode_init_once(&ei->vfs_inode);
2193 void btrfs_destroy_cachep(void)
2195 if (btrfs_inode_cachep)
2196 kmem_cache_destroy(btrfs_inode_cachep);
2197 if (btrfs_trans_handle_cachep)
2198 kmem_cache_destroy(btrfs_trans_handle_cachep);
2199 if (btrfs_transaction_cachep)
2200 kmem_cache_destroy(btrfs_transaction_cachep);
2201 if (btrfs_bit_radix_cachep)
2202 kmem_cache_destroy(btrfs_bit_radix_cachep);
2203 if (btrfs_path_cachep)
2204 kmem_cache_destroy(btrfs_path_cachep);
2207 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2208 unsigned long extra_flags,
2209 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2210 void (*ctor)(struct kmem_cache *, void *)
2212 void (*ctor)(void *, struct kmem_cache *,
2217 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2218 SLAB_MEM_SPREAD | extra_flags), ctor
2219 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2225 int btrfs_init_cachep(void)
2227 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2228 sizeof(struct btrfs_inode),
2230 if (!btrfs_inode_cachep)
2232 btrfs_trans_handle_cachep =
2233 btrfs_cache_create("btrfs_trans_handle_cache",
2234 sizeof(struct btrfs_trans_handle),
2236 if (!btrfs_trans_handle_cachep)
2238 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2239 sizeof(struct btrfs_transaction),
2241 if (!btrfs_transaction_cachep)
2243 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2244 sizeof(struct btrfs_path),
2246 if (!btrfs_path_cachep)
2248 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2249 SLAB_DESTROY_BY_RCU, NULL);
2250 if (!btrfs_bit_radix_cachep)
2254 btrfs_destroy_cachep();
2258 static int btrfs_getattr(struct vfsmount *mnt,
2259 struct dentry *dentry, struct kstat *stat)
2261 struct inode *inode = dentry->d_inode;
2262 generic_fillattr(inode, stat);
2263 stat->blksize = 256 * 1024;
2267 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2268 struct inode * new_dir,struct dentry *new_dentry)
2270 struct btrfs_trans_handle *trans;
2271 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2272 struct inode *new_inode = new_dentry->d_inode;
2273 struct inode *old_inode = old_dentry->d_inode;
2274 struct timespec ctime = CURRENT_TIME;
2275 struct btrfs_path *path;
2276 struct btrfs_dir_item *di;
2279 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2280 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2284 mutex_lock(&root->fs_info->fs_mutex);
2285 trans = btrfs_start_transaction(root, 1);
2287 btrfs_set_trans_block_group(trans, new_dir);
2288 path = btrfs_alloc_path();
2294 old_dentry->d_inode->i_nlink++;
2295 old_dir->i_ctime = old_dir->i_mtime = ctime;
2296 new_dir->i_ctime = new_dir->i_mtime = ctime;
2297 old_inode->i_ctime = ctime;
2299 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2300 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2301 struct btrfs_key old_parent_key;
2302 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2312 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
2313 ret = btrfs_del_item(trans, root, path);
2317 btrfs_release_path(root, path);
2319 di = btrfs_lookup_dir_index_item(trans, root, path,
2321 old_parent_key.objectid,
2331 ret = btrfs_del_item(trans, root, path);
2335 btrfs_release_path(root, path);
2337 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2338 old_inode->i_ino, location,
2345 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2350 new_inode->i_ctime = CURRENT_TIME;
2351 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2355 ret = btrfs_add_link(trans, new_dentry, old_inode);
2360 btrfs_free_path(path);
2361 btrfs_end_transaction(trans, root);
2362 mutex_unlock(&root->fs_info->fs_mutex);
2366 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2367 const char *symname)
2369 struct btrfs_trans_handle *trans;
2370 struct btrfs_root *root = BTRFS_I(dir)->root;
2371 struct btrfs_path *path;
2372 struct btrfs_key key;
2373 struct inode *inode;
2380 struct btrfs_file_extent_item *ei;
2381 struct extent_buffer *leaf;
2384 name_len = strlen(symname) + 1;
2385 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2386 return -ENAMETOOLONG;
2387 mutex_lock(&root->fs_info->fs_mutex);
2388 trans = btrfs_start_transaction(root, 1);
2389 btrfs_set_trans_block_group(trans, dir);
2391 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2397 inode = btrfs_new_inode(trans, root, objectid,
2398 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2399 err = PTR_ERR(inode);
2403 btrfs_set_trans_block_group(trans, inode);
2404 err = btrfs_add_nondir(trans, dentry, inode);
2408 inode->i_mapping->a_ops = &btrfs_aops;
2409 inode->i_fop = &btrfs_file_operations;
2410 inode->i_op = &btrfs_file_inode_operations;
2411 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2412 inode->i_mapping, GFP_NOFS);
2413 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2415 dir->i_sb->s_dirt = 1;
2416 btrfs_update_inode_block_group(trans, inode);
2417 btrfs_update_inode_block_group(trans, dir);
2421 path = btrfs_alloc_path();
2423 key.objectid = inode->i_ino;
2425 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2426 datasize = btrfs_file_extent_calc_inline_size(name_len);
2427 err = btrfs_insert_empty_item(trans, root, path, &key,
2433 leaf = path->nodes[0];
2434 ei = btrfs_item_ptr(leaf, path->slots[0],
2435 struct btrfs_file_extent_item);
2436 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2437 btrfs_set_file_extent_type(leaf, ei,
2438 BTRFS_FILE_EXTENT_INLINE);
2439 ptr = btrfs_file_extent_inline_start(ei);
2440 write_extent_buffer(leaf, symname, ptr, name_len);
2441 btrfs_mark_buffer_dirty(leaf);
2442 btrfs_free_path(path);
2444 inode->i_op = &btrfs_symlink_inode_operations;
2445 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2446 inode->i_size = name_len - 1;
2447 err = btrfs_update_inode(trans, root, inode);
2452 nr = trans->blocks_used;
2453 btrfs_end_transaction(trans, root);
2454 mutex_unlock(&root->fs_info->fs_mutex);
2456 inode_dec_link_count(inode);
2459 btrfs_btree_balance_dirty(root, nr);
2463 static struct inode_operations btrfs_dir_inode_operations = {
2464 .lookup = btrfs_lookup,
2465 .create = btrfs_create,
2466 .unlink = btrfs_unlink,
2468 .mkdir = btrfs_mkdir,
2469 .rmdir = btrfs_rmdir,
2470 .rename = btrfs_rename,
2471 .symlink = btrfs_symlink,
2472 .setattr = btrfs_setattr,
2473 .mknod = btrfs_mknod,
2476 static struct inode_operations btrfs_dir_ro_inode_operations = {
2477 .lookup = btrfs_lookup,
2480 static struct file_operations btrfs_dir_file_operations = {
2481 .llseek = generic_file_llseek,
2482 .read = generic_read_dir,
2483 .readdir = btrfs_readdir,
2484 .unlocked_ioctl = btrfs_ioctl,
2485 #ifdef CONFIG_COMPAT
2486 .compat_ioctl = btrfs_ioctl,
2490 static struct extent_map_ops btrfs_extent_map_ops = {
2491 .fill_delalloc = run_delalloc_range,
2492 .writepage_io_hook = btrfs_writepage_io_hook,
2493 .readpage_io_hook = btrfs_readpage_io_hook,
2494 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2497 static struct address_space_operations btrfs_aops = {
2498 .readpage = btrfs_readpage,
2499 .writepage = btrfs_writepage,
2500 .sync_page = block_sync_page,
2501 .prepare_write = btrfs_prepare_write,
2502 .commit_write = btrfs_commit_write,
2504 .invalidatepage = btrfs_invalidatepage,
2505 .releasepage = btrfs_releasepage,
2506 .set_page_dirty = __set_page_dirty_nobuffers,
2509 static struct address_space_operations btrfs_symlink_aops = {
2510 .readpage = btrfs_readpage,
2511 .writepage = btrfs_writepage,
2512 .invalidatepage = btrfs_invalidatepage,
2513 .releasepage = btrfs_releasepage,
2516 static struct inode_operations btrfs_file_inode_operations = {
2517 .truncate = btrfs_truncate,
2518 .getattr = btrfs_getattr,
2519 .setattr = btrfs_setattr,
2522 static struct inode_operations btrfs_special_inode_operations = {
2523 .getattr = btrfs_getattr,
2524 .setattr = btrfs_setattr,
2527 static struct inode_operations btrfs_symlink_inode_operations = {
2528 .readlink = generic_readlink,
2529 .follow_link = page_follow_link_light,
2530 .put_link = page_put_link,