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>
35 #include <linux/xattr.h>
38 #include "transaction.h"
39 #include "btrfs_inode.h"
41 #include "print-tree.h"
43 struct btrfs_iget_args {
45 struct btrfs_root *root;
48 static struct inode_operations btrfs_dir_inode_operations;
49 static struct inode_operations btrfs_symlink_inode_operations;
50 static struct inode_operations btrfs_dir_ro_inode_operations;
51 static struct inode_operations btrfs_special_inode_operations;
52 static struct inode_operations btrfs_file_inode_operations;
53 static struct address_space_operations btrfs_aops;
54 static struct address_space_operations btrfs_symlink_aops;
55 static struct file_operations btrfs_dir_file_operations;
56 static struct extent_map_ops btrfs_extent_map_ops;
58 static struct kmem_cache *btrfs_inode_cachep;
59 struct kmem_cache *btrfs_trans_handle_cachep;
60 struct kmem_cache *btrfs_transaction_cachep;
61 struct kmem_cache *btrfs_bit_radix_cachep;
62 struct kmem_cache *btrfs_path_cachep;
65 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
66 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
67 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
68 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
69 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
70 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
71 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
72 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
75 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
78 u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
79 u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
90 spin_lock(&root->fs_info->delalloc_lock);
91 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
93 spin_unlock(&root->fs_info->delalloc_lock);
97 static int cow_file_range(struct inode *inode, u64 start, u64 end)
99 struct btrfs_root *root = BTRFS_I(inode)->root;
100 struct btrfs_trans_handle *trans;
104 u64 blocksize = root->sectorsize;
105 struct btrfs_key ins;
108 trans = btrfs_start_transaction(root, 1);
110 btrfs_set_trans_block_group(trans, inode);
112 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
113 num_bytes = max(blocksize, num_bytes);
114 ret = btrfs_drop_extents(trans, root, inode,
115 start, start + num_bytes, start, &alloc_hint);
117 if (alloc_hint == EXTENT_MAP_INLINE)
120 while(num_bytes > 0) {
121 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
122 ret = btrfs_alloc_extent(trans, root, cur_alloc_size,
123 root->root_key.objectid,
125 inode->i_ino, start, 0,
126 alloc_hint, (u64)-1, &ins, 1);
131 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
132 start, ins.objectid, ins.offset,
134 num_bytes -= cur_alloc_size;
135 alloc_hint = ins.objectid + ins.offset;
136 start += cur_alloc_size;
138 btrfs_add_ordered_inode(inode);
140 btrfs_end_transaction(trans, root);
144 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
152 struct btrfs_root *root = BTRFS_I(inode)->root;
153 struct extent_buffer *leaf;
155 struct btrfs_path *path;
156 struct btrfs_file_extent_item *item;
159 struct btrfs_key found_key;
161 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
162 path = btrfs_alloc_path();
165 ret = btrfs_lookup_file_extent(NULL, root, path,
166 inode->i_ino, start, 0);
168 btrfs_free_path(path);
174 if (path->slots[0] == 0)
179 leaf = path->nodes[0];
180 item = btrfs_item_ptr(leaf, path->slots[0],
181 struct btrfs_file_extent_item);
183 /* are we inside the extent that was found? */
184 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
185 found_type = btrfs_key_type(&found_key);
186 if (found_key.objectid != inode->i_ino ||
187 found_type != BTRFS_EXTENT_DATA_KEY) {
191 found_type = btrfs_file_extent_type(leaf, item);
192 extent_start = found_key.offset;
193 if (found_type == BTRFS_FILE_EXTENT_REG) {
194 u64 extent_num_bytes;
196 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
197 extent_end = extent_start + extent_num_bytes;
200 if (loops && start != extent_start)
203 if (start < extent_start || start >= extent_end)
206 cow_end = min(end, extent_end - 1);
207 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
212 * we may be called by the resizer, make sure we're inside
213 * the limits of the FS
215 if (bytenr + extent_num_bytes > total_fs_bytes)
218 if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) {
228 btrfs_free_path(path);
231 btrfs_release_path(root, path);
236 cow_file_range(inode, start, cow_end);
241 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
243 struct btrfs_root *root = BTRFS_I(inode)->root;
246 mutex_lock(&root->fs_info->fs_mutex);
247 if (btrfs_test_opt(root, NODATACOW) ||
248 btrfs_test_flag(inode, NODATACOW))
249 ret = run_delalloc_nocow(inode, start, end);
251 ret = cow_file_range(inode, start, end);
253 spin_lock(&root->fs_info->delalloc_lock);
254 num_bytes = end + 1 - start;
255 if (root->fs_info->delalloc_bytes < num_bytes) {
256 printk("delalloc accounting error total %llu sub %llu\n",
257 root->fs_info->delalloc_bytes, num_bytes);
259 root->fs_info->delalloc_bytes -= num_bytes;
261 spin_unlock(&root->fs_info->delalloc_lock);
263 mutex_unlock(&root->fs_info->fs_mutex);
267 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
269 struct inode *inode = page->mapping->host;
270 struct btrfs_root *root = BTRFS_I(inode)->root;
271 struct btrfs_trans_handle *trans;
274 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
275 size_t offset = start - page_start;
276 if (btrfs_test_opt(root, NODATASUM) ||
277 btrfs_test_flag(inode, NODATASUM))
279 mutex_lock(&root->fs_info->fs_mutex);
280 trans = btrfs_start_transaction(root, 1);
281 btrfs_set_trans_block_group(trans, inode);
283 btrfs_csum_file_block(trans, root, inode, inode->i_ino,
284 start, kaddr + offset, end - start + 1);
286 ret = btrfs_end_transaction(trans, root);
288 mutex_unlock(&root->fs_info->fs_mutex);
292 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
295 struct inode *inode = page->mapping->host;
296 struct btrfs_root *root = BTRFS_I(inode)->root;
297 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
298 struct btrfs_csum_item *item;
299 struct btrfs_path *path = NULL;
301 if (btrfs_test_opt(root, NODATASUM) ||
302 btrfs_test_flag(inode, NODATASUM))
304 mutex_lock(&root->fs_info->fs_mutex);
305 path = btrfs_alloc_path();
306 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
309 /* a csum that isn't present is a preallocated region. */
310 if (ret == -ENOENT || ret == -EFBIG)
315 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
317 set_state_private(em_tree, start, csum);
320 btrfs_free_path(path);
321 mutex_unlock(&root->fs_info->fs_mutex);
325 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
327 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
328 struct inode *inode = page->mapping->host;
329 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
333 struct btrfs_root *root = BTRFS_I(inode)->root;
336 if (btrfs_test_opt(root, NODATASUM) ||
337 btrfs_test_flag(inode, NODATASUM))
339 ret = get_state_private(em_tree, start, &private);
340 local_irq_save(flags);
341 kaddr = kmap_atomic(page, KM_IRQ0);
345 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
346 btrfs_csum_final(csum, (char *)&csum);
347 if (csum != private) {
350 kunmap_atomic(kaddr, KM_IRQ0);
351 local_irq_restore(flags);
355 printk("btrfs csum failed ino %lu off %llu\n",
356 page->mapping->host->i_ino, (unsigned long long)start);
357 memset(kaddr + offset, 1, end - start + 1);
358 flush_dcache_page(page);
359 kunmap_atomic(kaddr, KM_IRQ0);
360 local_irq_restore(flags);
364 void btrfs_read_locked_inode(struct inode *inode)
366 struct btrfs_path *path;
367 struct extent_buffer *leaf;
368 struct btrfs_inode_item *inode_item;
369 struct btrfs_inode_timespec *tspec;
370 struct btrfs_root *root = BTRFS_I(inode)->root;
371 struct btrfs_key location;
372 u64 alloc_group_block;
376 path = btrfs_alloc_path();
378 mutex_lock(&root->fs_info->fs_mutex);
379 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
381 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
385 leaf = path->nodes[0];
386 inode_item = btrfs_item_ptr(leaf, path->slots[0],
387 struct btrfs_inode_item);
389 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
390 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
391 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
392 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
393 inode->i_size = btrfs_inode_size(leaf, inode_item);
395 tspec = btrfs_inode_atime(inode_item);
396 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
397 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
399 tspec = btrfs_inode_mtime(inode_item);
400 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
401 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
403 tspec = btrfs_inode_ctime(inode_item);
404 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
405 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
407 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
408 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
410 rdev = btrfs_inode_rdev(leaf, inode_item);
412 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
413 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
415 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
416 if (!BTRFS_I(inode)->block_group) {
417 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
420 btrfs_free_path(path);
423 mutex_unlock(&root->fs_info->fs_mutex);
425 switch (inode->i_mode & S_IFMT) {
427 inode->i_mapping->a_ops = &btrfs_aops;
428 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
429 inode->i_fop = &btrfs_file_operations;
430 inode->i_op = &btrfs_file_inode_operations;
433 inode->i_fop = &btrfs_dir_file_operations;
434 if (root == root->fs_info->tree_root)
435 inode->i_op = &btrfs_dir_ro_inode_operations;
437 inode->i_op = &btrfs_dir_inode_operations;
440 inode->i_op = &btrfs_symlink_inode_operations;
441 inode->i_mapping->a_ops = &btrfs_symlink_aops;
444 init_special_inode(inode, inode->i_mode, rdev);
450 btrfs_release_path(root, path);
451 btrfs_free_path(path);
452 mutex_unlock(&root->fs_info->fs_mutex);
453 make_bad_inode(inode);
456 static void fill_inode_item(struct extent_buffer *leaf,
457 struct btrfs_inode_item *item,
460 btrfs_set_inode_uid(leaf, item, inode->i_uid);
461 btrfs_set_inode_gid(leaf, item, inode->i_gid);
462 btrfs_set_inode_size(leaf, item, inode->i_size);
463 btrfs_set_inode_mode(leaf, item, inode->i_mode);
464 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
466 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
467 inode->i_atime.tv_sec);
468 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
469 inode->i_atime.tv_nsec);
471 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
472 inode->i_mtime.tv_sec);
473 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
474 inode->i_mtime.tv_nsec);
476 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
477 inode->i_ctime.tv_sec);
478 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
479 inode->i_ctime.tv_nsec);
481 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
482 btrfs_set_inode_generation(leaf, item, inode->i_generation);
483 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
484 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
485 btrfs_set_inode_block_group(leaf, item,
486 BTRFS_I(inode)->block_group->key.objectid);
489 int btrfs_update_inode(struct btrfs_trans_handle *trans,
490 struct btrfs_root *root,
493 struct btrfs_inode_item *inode_item;
494 struct btrfs_path *path;
495 struct extent_buffer *leaf;
498 path = btrfs_alloc_path();
500 ret = btrfs_lookup_inode(trans, root, path,
501 &BTRFS_I(inode)->location, 1);
508 leaf = path->nodes[0];
509 inode_item = btrfs_item_ptr(leaf, path->slots[0],
510 struct btrfs_inode_item);
512 fill_inode_item(leaf, inode_item, inode);
513 btrfs_mark_buffer_dirty(leaf);
514 btrfs_set_inode_last_trans(trans, inode);
517 btrfs_release_path(root, path);
518 btrfs_free_path(path);
523 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
524 struct btrfs_root *root,
526 struct dentry *dentry)
528 struct btrfs_path *path;
529 const char *name = dentry->d_name.name;
530 int name_len = dentry->d_name.len;
532 struct extent_buffer *leaf;
533 struct btrfs_dir_item *di;
534 struct btrfs_key key;
536 path = btrfs_alloc_path();
542 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
552 leaf = path->nodes[0];
553 btrfs_dir_item_key_to_cpu(leaf, di, &key);
554 ret = btrfs_delete_one_dir_name(trans, root, path, di);
557 btrfs_release_path(root, path);
559 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
560 key.objectid, name, name_len, -1);
569 ret = btrfs_delete_one_dir_name(trans, root, path, di);
571 dentry->d_inode->i_ctime = dir->i_ctime;
572 ret = btrfs_del_inode_ref(trans, root, name, name_len,
573 dentry->d_inode->i_ino,
574 dentry->d_parent->d_inode->i_ino);
576 printk("failed to delete reference to %.*s, "
577 "inode %lu parent %lu\n", name_len, name,
578 dentry->d_inode->i_ino,
579 dentry->d_parent->d_inode->i_ino);
582 btrfs_free_path(path);
584 dir->i_size -= name_len * 2;
585 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
586 btrfs_update_inode(trans, root, dir);
587 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
588 dentry->d_inode->i_nlink--;
590 drop_nlink(dentry->d_inode);
592 ret = btrfs_update_inode(trans, root, dentry->d_inode);
593 dir->i_sb->s_dirt = 1;
598 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
600 struct btrfs_root *root;
601 struct btrfs_trans_handle *trans;
603 unsigned long nr = 0;
605 root = BTRFS_I(dir)->root;
606 mutex_lock(&root->fs_info->fs_mutex);
608 ret = btrfs_check_free_space(root, 1, 1);
612 trans = btrfs_start_transaction(root, 1);
614 btrfs_set_trans_block_group(trans, dir);
615 ret = btrfs_unlink_trans(trans, root, dir, dentry);
616 nr = trans->blocks_used;
618 btrfs_end_transaction(trans, root);
620 mutex_unlock(&root->fs_info->fs_mutex);
621 btrfs_btree_balance_dirty(root, nr);
622 btrfs_throttle(root);
626 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
628 struct inode *inode = dentry->d_inode;
631 struct btrfs_root *root = BTRFS_I(dir)->root;
632 struct btrfs_trans_handle *trans;
633 unsigned long nr = 0;
635 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
638 mutex_lock(&root->fs_info->fs_mutex);
639 ret = btrfs_check_free_space(root, 1, 1);
643 trans = btrfs_start_transaction(root, 1);
644 btrfs_set_trans_block_group(trans, dir);
646 /* now the directory is empty */
647 err = btrfs_unlink_trans(trans, root, dir, dentry);
652 nr = trans->blocks_used;
653 ret = btrfs_end_transaction(trans, root);
655 mutex_unlock(&root->fs_info->fs_mutex);
656 btrfs_btree_balance_dirty(root, nr);
657 btrfs_throttle(root);
664 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
665 struct btrfs_root *root,
668 struct btrfs_path *path;
673 path = btrfs_alloc_path();
675 ret = btrfs_lookup_inode(trans, root, path,
676 &BTRFS_I(inode)->location, -1);
680 ret = btrfs_del_item(trans, root, path);
681 btrfs_free_path(path);
686 * this can truncate away extent items, csum items and directory items.
687 * It starts at a high offset and removes keys until it can't find
688 * any higher than i_size.
690 * csum items that cross the new i_size are truncated to the new size
693 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
694 struct btrfs_root *root,
698 struct btrfs_path *path;
699 struct btrfs_key key;
700 struct btrfs_key found_key;
702 struct extent_buffer *leaf;
703 struct btrfs_file_extent_item *fi;
704 u64 extent_start = 0;
705 u64 extent_num_bytes = 0;
711 int extent_type = -1;
713 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
714 path = btrfs_alloc_path();
718 /* FIXME, add redo link to tree so we don't leak on crash */
719 key.objectid = inode->i_ino;
720 key.offset = (u64)-1;
724 btrfs_init_path(path);
726 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
731 BUG_ON(path->slots[0] == 0);
734 leaf = path->nodes[0];
735 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
736 found_type = btrfs_key_type(&found_key);
738 if (found_key.objectid != inode->i_ino)
741 if (found_type != BTRFS_CSUM_ITEM_KEY &&
742 found_type != BTRFS_DIR_ITEM_KEY &&
743 found_type != BTRFS_DIR_INDEX_KEY &&
744 found_type != BTRFS_EXTENT_DATA_KEY)
747 item_end = found_key.offset;
748 if (found_type == BTRFS_EXTENT_DATA_KEY) {
749 fi = btrfs_item_ptr(leaf, path->slots[0],
750 struct btrfs_file_extent_item);
751 extent_type = btrfs_file_extent_type(leaf, fi);
752 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
754 btrfs_file_extent_num_bytes(leaf, fi);
755 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
756 struct btrfs_item *item = btrfs_item_nr(leaf,
758 item_end += btrfs_file_extent_inline_len(leaf,
763 if (found_type == BTRFS_CSUM_ITEM_KEY) {
764 ret = btrfs_csum_truncate(trans, root, path,
768 if (item_end < inode->i_size) {
769 if (found_type == BTRFS_DIR_ITEM_KEY) {
770 found_type = BTRFS_INODE_ITEM_KEY;
771 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
772 found_type = BTRFS_CSUM_ITEM_KEY;
773 } else if (found_type) {
778 btrfs_set_key_type(&key, found_type);
779 btrfs_release_path(root, path);
782 if (found_key.offset >= inode->i_size)
788 /* FIXME, shrink the extent if the ref count is only 1 */
789 if (found_type != BTRFS_EXTENT_DATA_KEY)
792 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
794 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
797 btrfs_file_extent_num_bytes(leaf, fi);
798 extent_num_bytes = inode->i_size -
799 found_key.offset + root->sectorsize - 1;
800 btrfs_set_file_extent_num_bytes(leaf, fi,
802 num_dec = (orig_num_bytes -
803 extent_num_bytes) >> 9;
804 if (extent_start != 0) {
805 inode->i_blocks -= num_dec;
807 btrfs_mark_buffer_dirty(leaf);
810 btrfs_file_extent_disk_num_bytes(leaf,
812 /* FIXME blocksize != 4096 */
813 num_dec = btrfs_file_extent_num_bytes(leaf,
815 if (extent_start != 0) {
817 inode->i_blocks -= num_dec;
819 root_gen = btrfs_header_generation(leaf);
820 root_owner = btrfs_header_owner(leaf);
822 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE &&
824 u32 newsize = inode->i_size - found_key.offset;
825 newsize = btrfs_file_extent_calc_inline_size(newsize);
826 ret = btrfs_truncate_item(trans, root, path,
832 ret = btrfs_del_item(trans, root, path);
838 btrfs_release_path(root, path);
840 ret = btrfs_free_extent(trans, root, extent_start,
843 root_gen, inode->i_ino,
844 found_key.offset, 0);
850 btrfs_release_path(root, path);
851 btrfs_free_path(path);
852 inode->i_sb->s_dirt = 1;
856 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
860 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
861 struct btrfs_root *root = BTRFS_I(inode)->root;
862 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
863 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
864 u64 existing_delalloc;
868 WARN_ON(!PageLocked(page));
869 set_page_extent_mapped(page);
871 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
872 delalloc_start = page_start;
873 existing_delalloc = count_range_bits(&BTRFS_I(inode)->extent_tree,
874 &delalloc_start, page_end,
875 PAGE_CACHE_SIZE, EXTENT_DELALLOC);
876 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
879 spin_lock(&root->fs_info->delalloc_lock);
880 root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE - existing_delalloc;
881 spin_unlock(&root->fs_info->delalloc_lock);
883 if (zero_start != PAGE_CACHE_SIZE) {
885 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
886 flush_dcache_page(page);
889 set_page_dirty(page);
890 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
896 * taken from block_truncate_page, but does cow as it zeros out
897 * any bytes left in the last page in the file.
899 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
901 struct inode *inode = mapping->host;
902 struct btrfs_root *root = BTRFS_I(inode)->root;
903 u32 blocksize = root->sectorsize;
904 pgoff_t index = from >> PAGE_CACHE_SHIFT;
905 unsigned offset = from & (PAGE_CACHE_SIZE-1);
910 if ((offset & (blocksize - 1)) == 0)
914 page = grab_cache_page(mapping, index);
917 if (!PageUptodate(page)) {
918 ret = btrfs_readpage(NULL, page);
920 if (!PageUptodate(page)) {
925 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
927 ret = btrfs_cow_one_page(inode, page, offset);
930 page_cache_release(page);
935 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
937 struct inode *inode = dentry->d_inode;
940 err = inode_change_ok(inode, attr);
944 if (S_ISREG(inode->i_mode) &&
945 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
946 struct btrfs_trans_handle *trans;
947 struct btrfs_root *root = BTRFS_I(inode)->root;
948 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
950 u64 mask = root->sectorsize - 1;
951 u64 pos = (inode->i_size + mask) & ~mask;
952 u64 block_end = attr->ia_size | mask;
956 if (attr->ia_size <= pos)
959 mutex_lock(&root->fs_info->fs_mutex);
960 err = btrfs_check_free_space(root, 1, 0);
961 mutex_unlock(&root->fs_info->fs_mutex);
965 btrfs_truncate_page(inode->i_mapping, inode->i_size);
967 lock_extent(em_tree, pos, block_end, GFP_NOFS);
968 hole_size = (attr->ia_size - pos + mask) & ~mask;
970 mutex_lock(&root->fs_info->fs_mutex);
971 trans = btrfs_start_transaction(root, 1);
972 btrfs_set_trans_block_group(trans, inode);
973 err = btrfs_drop_extents(trans, root, inode,
974 pos, pos + hole_size, pos,
977 if (alloc_hint != EXTENT_MAP_INLINE) {
978 err = btrfs_insert_file_extent(trans, root,
980 pos, 0, 0, hole_size);
982 btrfs_end_transaction(trans, root);
983 mutex_unlock(&root->fs_info->fs_mutex);
984 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
989 err = inode_setattr(inode, attr);
994 void btrfs_drop_inode(struct inode *inode)
996 struct btrfs_trans_handle *trans;
997 struct btrfs_root *root = BTRFS_I(inode)->root;
999 if (!BTRFS_I(inode)->ordered_trans) {
1000 generic_drop_inode(inode);
1003 /* nasty, but it prevents a deadlock with data=ordered by preventing
1004 * a commit until after this inode is done
1006 trans = btrfs_start_transaction(root, 1);
1007 generic_drop_inode(inode);
1008 /* note, the inode is now untouchable */
1009 btrfs_end_transaction(trans, root);
1012 void btrfs_delete_inode(struct inode *inode)
1014 struct btrfs_trans_handle *trans;
1015 struct btrfs_root *root = BTRFS_I(inode)->root;
1019 truncate_inode_pages(&inode->i_data, 0);
1020 if (is_bad_inode(inode)) {
1025 mutex_lock(&root->fs_info->fs_mutex);
1026 trans = btrfs_start_transaction(root, 1);
1028 btrfs_set_trans_block_group(trans, inode);
1029 ret = btrfs_truncate_in_trans(trans, root, inode);
1031 goto no_delete_lock;
1032 ret = btrfs_delete_xattrs(trans, root, inode);
1034 goto no_delete_lock;
1035 ret = btrfs_free_inode(trans, root, inode);
1037 goto no_delete_lock;
1038 nr = trans->blocks_used;
1040 btrfs_end_transaction(trans, root);
1041 mutex_unlock(&root->fs_info->fs_mutex);
1042 btrfs_btree_balance_dirty(root, nr);
1043 btrfs_throttle(root);
1047 nr = trans->blocks_used;
1048 btrfs_end_transaction(trans, root);
1049 mutex_unlock(&root->fs_info->fs_mutex);
1050 btrfs_btree_balance_dirty(root, nr);
1051 btrfs_throttle(root);
1057 * this returns the key found in the dir entry in the location pointer.
1058 * If no dir entries were found, location->objectid is 0.
1060 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1061 struct btrfs_key *location)
1063 const char *name = dentry->d_name.name;
1064 int namelen = dentry->d_name.len;
1065 struct btrfs_dir_item *di;
1066 struct btrfs_path *path;
1067 struct btrfs_root *root = BTRFS_I(dir)->root;
1070 if (namelen == 1 && strcmp(name, ".") == 0) {
1071 location->objectid = dir->i_ino;
1072 location->type = BTRFS_INODE_ITEM_KEY;
1073 location->offset = 0;
1076 path = btrfs_alloc_path();
1079 if (namelen == 2 && strcmp(name, "..") == 0) {
1080 struct btrfs_key key;
1081 struct extent_buffer *leaf;
1085 key.objectid = dir->i_ino;
1086 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1088 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1092 leaf = path->nodes[0];
1093 slot = path->slots[0];
1094 nritems = btrfs_header_nritems(leaf);
1095 if (slot >= nritems)
1098 btrfs_item_key_to_cpu(leaf, &key, slot);
1099 if (key.objectid != dir->i_ino ||
1100 key.type != BTRFS_INODE_REF_KEY) {
1103 location->objectid = key.offset;
1104 location->type = BTRFS_INODE_ITEM_KEY;
1105 location->offset = 0;
1109 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1113 if (!di || IS_ERR(di)) {
1116 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1118 btrfs_free_path(path);
1121 location->objectid = 0;
1126 * when we hit a tree root in a directory, the btrfs part of the inode
1127 * needs to be changed to reflect the root directory of the tree root. This
1128 * is kind of like crossing a mount point.
1130 static int fixup_tree_root_location(struct btrfs_root *root,
1131 struct btrfs_key *location,
1132 struct btrfs_root **sub_root,
1133 struct dentry *dentry)
1135 struct btrfs_path *path;
1136 struct btrfs_root_item *ri;
1138 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1140 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1143 path = btrfs_alloc_path();
1145 mutex_lock(&root->fs_info->fs_mutex);
1147 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1148 dentry->d_name.name,
1149 dentry->d_name.len);
1150 if (IS_ERR(*sub_root))
1151 return PTR_ERR(*sub_root);
1153 ri = &(*sub_root)->root_item;
1154 location->objectid = btrfs_root_dirid(ri);
1155 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1156 location->offset = 0;
1158 btrfs_free_path(path);
1159 mutex_unlock(&root->fs_info->fs_mutex);
1163 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1165 struct btrfs_iget_args *args = p;
1166 inode->i_ino = args->ino;
1167 BTRFS_I(inode)->root = args->root;
1168 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1169 inode->i_mapping, GFP_NOFS);
1173 static int btrfs_find_actor(struct inode *inode, void *opaque)
1175 struct btrfs_iget_args *args = opaque;
1176 return (args->ino == inode->i_ino &&
1177 args->root == BTRFS_I(inode)->root);
1180 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1183 struct btrfs_iget_args args;
1184 args.ino = objectid;
1185 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1190 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1193 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1194 struct btrfs_root *root)
1196 struct inode *inode;
1197 struct btrfs_iget_args args;
1198 args.ino = objectid;
1201 inode = iget5_locked(s, objectid, btrfs_find_actor,
1202 btrfs_init_locked_inode,
1207 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1208 struct nameidata *nd)
1210 struct inode * inode;
1211 struct btrfs_inode *bi = BTRFS_I(dir);
1212 struct btrfs_root *root = bi->root;
1213 struct btrfs_root *sub_root = root;
1214 struct btrfs_key location;
1217 if (dentry->d_name.len > BTRFS_NAME_LEN)
1218 return ERR_PTR(-ENAMETOOLONG);
1220 mutex_lock(&root->fs_info->fs_mutex);
1221 ret = btrfs_inode_by_name(dir, dentry, &location);
1222 mutex_unlock(&root->fs_info->fs_mutex);
1225 return ERR_PTR(ret);
1228 if (location.objectid) {
1229 ret = fixup_tree_root_location(root, &location, &sub_root,
1232 return ERR_PTR(ret);
1234 return ERR_PTR(-ENOENT);
1235 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1238 return ERR_PTR(-EACCES);
1239 if (inode->i_state & I_NEW) {
1240 /* the inode and parent dir are two different roots */
1241 if (sub_root != root) {
1243 sub_root->inode = inode;
1245 BTRFS_I(inode)->root = sub_root;
1246 memcpy(&BTRFS_I(inode)->location, &location,
1248 btrfs_read_locked_inode(inode);
1249 unlock_new_inode(inode);
1252 return d_splice_alias(inode, dentry);
1255 static unsigned char btrfs_filetype_table[] = {
1256 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1259 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1261 struct inode *inode = filp->f_dentry->d_inode;
1262 struct btrfs_root *root = BTRFS_I(inode)->root;
1263 struct btrfs_item *item;
1264 struct btrfs_dir_item *di;
1265 struct btrfs_key key;
1266 struct btrfs_key found_key;
1267 struct btrfs_path *path;
1270 struct extent_buffer *leaf;
1273 unsigned char d_type;
1278 int key_type = BTRFS_DIR_INDEX_KEY;
1283 /* FIXME, use a real flag for deciding about the key type */
1284 if (root->fs_info->tree_root == root)
1285 key_type = BTRFS_DIR_ITEM_KEY;
1287 /* special case for "." */
1288 if (filp->f_pos == 0) {
1289 over = filldir(dirent, ".", 1,
1297 mutex_lock(&root->fs_info->fs_mutex);
1298 key.objectid = inode->i_ino;
1299 path = btrfs_alloc_path();
1302 /* special case for .., just use the back ref */
1303 if (filp->f_pos == 1) {
1304 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1306 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1308 leaf = path->nodes[0];
1309 slot = path->slots[0];
1310 nritems = btrfs_header_nritems(leaf);
1311 if (slot >= nritems) {
1312 btrfs_release_path(root, path);
1313 goto read_dir_items;
1315 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1316 btrfs_release_path(root, path);
1317 if (found_key.objectid != key.objectid ||
1318 found_key.type != BTRFS_INODE_REF_KEY)
1319 goto read_dir_items;
1320 over = filldir(dirent, "..", 2,
1321 2, found_key.offset, DT_DIR);
1328 btrfs_set_key_type(&key, key_type);
1329 key.offset = filp->f_pos;
1331 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1336 leaf = path->nodes[0];
1337 nritems = btrfs_header_nritems(leaf);
1338 slot = path->slots[0];
1339 if (advance || slot >= nritems) {
1340 if (slot >= nritems -1) {
1341 ret = btrfs_next_leaf(root, path);
1344 leaf = path->nodes[0];
1345 nritems = btrfs_header_nritems(leaf);
1346 slot = path->slots[0];
1353 item = btrfs_item_nr(leaf, slot);
1354 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1356 if (found_key.objectid != key.objectid)
1358 if (btrfs_key_type(&found_key) != key_type)
1360 if (found_key.offset < filp->f_pos)
1363 filp->f_pos = found_key.offset;
1365 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1367 di_total = btrfs_item_size(leaf, item);
1368 while(di_cur < di_total) {
1369 struct btrfs_key location;
1371 name_len = btrfs_dir_name_len(leaf, di);
1372 if (name_len < 32) {
1373 name_ptr = tmp_name;
1375 name_ptr = kmalloc(name_len, GFP_NOFS);
1378 read_extent_buffer(leaf, name_ptr,
1379 (unsigned long)(di + 1), name_len);
1381 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1382 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1383 over = filldir(dirent, name_ptr, name_len,
1388 if (name_ptr != tmp_name)
1393 di_len = btrfs_dir_name_len(leaf, di) +
1394 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1396 di = (struct btrfs_dir_item *)((char *)di + di_len);
1403 btrfs_release_path(root, path);
1404 btrfs_free_path(path);
1405 mutex_unlock(&root->fs_info->fs_mutex);
1409 int btrfs_write_inode(struct inode *inode, int wait)
1411 struct btrfs_root *root = BTRFS_I(inode)->root;
1412 struct btrfs_trans_handle *trans;
1416 mutex_lock(&root->fs_info->fs_mutex);
1417 trans = btrfs_start_transaction(root, 1);
1418 btrfs_set_trans_block_group(trans, inode);
1419 ret = btrfs_commit_transaction(trans, root);
1420 mutex_unlock(&root->fs_info->fs_mutex);
1426 * This is somewhat expensive, updating the tree every time the
1427 * inode changes. But, it is most likely to find the inode in cache.
1428 * FIXME, needs more benchmarking...there are no reasons other than performance
1429 * to keep or drop this code.
1431 void btrfs_dirty_inode(struct inode *inode)
1433 struct btrfs_root *root = BTRFS_I(inode)->root;
1434 struct btrfs_trans_handle *trans;
1436 mutex_lock(&root->fs_info->fs_mutex);
1437 trans = btrfs_start_transaction(root, 1);
1438 btrfs_set_trans_block_group(trans, inode);
1439 btrfs_update_inode(trans, root, inode);
1440 btrfs_end_transaction(trans, root);
1441 mutex_unlock(&root->fs_info->fs_mutex);
1444 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1445 struct btrfs_root *root,
1447 struct btrfs_block_group_cache *group,
1450 struct inode *inode;
1451 struct btrfs_inode_item *inode_item;
1452 struct btrfs_key *location;
1453 struct btrfs_path *path;
1457 path = btrfs_alloc_path();
1460 inode = new_inode(root->fs_info->sb);
1462 return ERR_PTR(-ENOMEM);
1464 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1465 inode->i_mapping, GFP_NOFS);
1466 BTRFS_I(inode)->root = root;
1472 group = btrfs_find_block_group(root, group, 0, 0, owner);
1473 BTRFS_I(inode)->block_group = group;
1474 BTRFS_I(inode)->flags = 0;
1475 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1479 inode->i_uid = current->fsuid;
1480 inode->i_gid = current->fsgid;
1481 inode->i_mode = mode;
1482 inode->i_ino = objectid;
1483 inode->i_blocks = 0;
1484 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1485 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1486 struct btrfs_inode_item);
1487 fill_inode_item(path->nodes[0], inode_item, inode);
1488 btrfs_mark_buffer_dirty(path->nodes[0]);
1489 btrfs_free_path(path);
1491 location = &BTRFS_I(inode)->location;
1492 location->objectid = objectid;
1493 location->offset = 0;
1494 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1496 insert_inode_hash(inode);
1499 btrfs_free_path(path);
1500 return ERR_PTR(ret);
1503 static inline u8 btrfs_inode_type(struct inode *inode)
1505 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1508 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1509 struct dentry *dentry, struct inode *inode)
1512 struct btrfs_key key;
1513 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1514 struct inode *parent_inode;
1516 key.objectid = inode->i_ino;
1517 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1520 ret = btrfs_insert_dir_item(trans, root,
1521 dentry->d_name.name, dentry->d_name.len,
1522 dentry->d_parent->d_inode->i_ino,
1523 &key, btrfs_inode_type(inode));
1525 ret = btrfs_insert_inode_ref(trans, root,
1526 dentry->d_name.name,
1529 dentry->d_parent->d_inode->i_ino);
1530 parent_inode = dentry->d_parent->d_inode;
1531 parent_inode->i_size += dentry->d_name.len * 2;
1532 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1533 ret = btrfs_update_inode(trans, root,
1534 dentry->d_parent->d_inode);
1539 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1540 struct dentry *dentry, struct inode *inode)
1542 int err = btrfs_add_link(trans, dentry, inode);
1544 d_instantiate(dentry, inode);
1552 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1553 int mode, dev_t rdev)
1555 struct btrfs_trans_handle *trans;
1556 struct btrfs_root *root = BTRFS_I(dir)->root;
1557 struct inode *inode = NULL;
1561 unsigned long nr = 0;
1563 if (!new_valid_dev(rdev))
1566 mutex_lock(&root->fs_info->fs_mutex);
1567 err = btrfs_check_free_space(root, 1, 0);
1571 trans = btrfs_start_transaction(root, 1);
1572 btrfs_set_trans_block_group(trans, dir);
1574 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1580 inode = btrfs_new_inode(trans, root, objectid,
1581 BTRFS_I(dir)->block_group, mode);
1582 err = PTR_ERR(inode);
1586 btrfs_set_trans_block_group(trans, inode);
1587 err = btrfs_add_nondir(trans, dentry, inode);
1591 inode->i_op = &btrfs_special_inode_operations;
1592 init_special_inode(inode, inode->i_mode, rdev);
1593 btrfs_update_inode(trans, root, inode);
1595 dir->i_sb->s_dirt = 1;
1596 btrfs_update_inode_block_group(trans, inode);
1597 btrfs_update_inode_block_group(trans, dir);
1599 nr = trans->blocks_used;
1600 btrfs_end_transaction(trans, root);
1602 mutex_unlock(&root->fs_info->fs_mutex);
1605 inode_dec_link_count(inode);
1608 btrfs_btree_balance_dirty(root, nr);
1609 btrfs_throttle(root);
1613 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1614 int mode, struct nameidata *nd)
1616 struct btrfs_trans_handle *trans;
1617 struct btrfs_root *root = BTRFS_I(dir)->root;
1618 struct inode *inode = NULL;
1621 unsigned long nr = 0;
1624 mutex_lock(&root->fs_info->fs_mutex);
1625 err = btrfs_check_free_space(root, 1, 0);
1628 trans = btrfs_start_transaction(root, 1);
1629 btrfs_set_trans_block_group(trans, dir);
1631 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1637 inode = btrfs_new_inode(trans, root, objectid,
1638 BTRFS_I(dir)->block_group, mode);
1639 err = PTR_ERR(inode);
1643 btrfs_set_trans_block_group(trans, inode);
1644 err = btrfs_add_nondir(trans, dentry, inode);
1648 inode->i_mapping->a_ops = &btrfs_aops;
1649 inode->i_fop = &btrfs_file_operations;
1650 inode->i_op = &btrfs_file_inode_operations;
1651 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1652 inode->i_mapping, GFP_NOFS);
1653 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
1655 dir->i_sb->s_dirt = 1;
1656 btrfs_update_inode_block_group(trans, inode);
1657 btrfs_update_inode_block_group(trans, dir);
1659 nr = trans->blocks_used;
1660 btrfs_end_transaction(trans, root);
1662 mutex_unlock(&root->fs_info->fs_mutex);
1665 inode_dec_link_count(inode);
1668 btrfs_btree_balance_dirty(root, nr);
1669 btrfs_throttle(root);
1673 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1674 struct dentry *dentry)
1676 struct btrfs_trans_handle *trans;
1677 struct btrfs_root *root = BTRFS_I(dir)->root;
1678 struct inode *inode = old_dentry->d_inode;
1679 unsigned long nr = 0;
1683 if (inode->i_nlink == 0)
1686 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1691 mutex_lock(&root->fs_info->fs_mutex);
1692 err = btrfs_check_free_space(root, 1, 0);
1695 trans = btrfs_start_transaction(root, 1);
1697 btrfs_set_trans_block_group(trans, dir);
1698 atomic_inc(&inode->i_count);
1699 err = btrfs_add_nondir(trans, dentry, inode);
1704 dir->i_sb->s_dirt = 1;
1705 btrfs_update_inode_block_group(trans, dir);
1706 err = btrfs_update_inode(trans, root, inode);
1711 nr = trans->blocks_used;
1712 btrfs_end_transaction(trans, root);
1714 mutex_unlock(&root->fs_info->fs_mutex);
1717 inode_dec_link_count(inode);
1720 btrfs_btree_balance_dirty(root, nr);
1721 btrfs_throttle(root);
1725 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1727 struct inode *inode;
1728 struct btrfs_trans_handle *trans;
1729 struct btrfs_root *root = BTRFS_I(dir)->root;
1731 int drop_on_err = 0;
1733 unsigned long nr = 1;
1735 mutex_lock(&root->fs_info->fs_mutex);
1736 err = btrfs_check_free_space(root, 1, 0);
1740 trans = btrfs_start_transaction(root, 1);
1741 btrfs_set_trans_block_group(trans, dir);
1743 if (IS_ERR(trans)) {
1744 err = PTR_ERR(trans);
1748 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1754 inode = btrfs_new_inode(trans, root, objectid,
1755 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1756 if (IS_ERR(inode)) {
1757 err = PTR_ERR(inode);
1762 inode->i_op = &btrfs_dir_inode_operations;
1763 inode->i_fop = &btrfs_dir_file_operations;
1764 btrfs_set_trans_block_group(trans, inode);
1767 err = btrfs_update_inode(trans, root, inode);
1771 err = btrfs_add_link(trans, dentry, inode);
1775 d_instantiate(dentry, inode);
1777 dir->i_sb->s_dirt = 1;
1778 btrfs_update_inode_block_group(trans, inode);
1779 btrfs_update_inode_block_group(trans, dir);
1782 nr = trans->blocks_used;
1783 btrfs_end_transaction(trans, root);
1786 mutex_unlock(&root->fs_info->fs_mutex);
1789 btrfs_btree_balance_dirty(root, nr);
1790 btrfs_throttle(root);
1794 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1795 size_t page_offset, u64 start, u64 end,
1801 u64 extent_start = 0;
1803 u64 objectid = inode->i_ino;
1805 int failed_insert = 0;
1806 struct btrfs_path *path;
1807 struct btrfs_root *root = BTRFS_I(inode)->root;
1808 struct btrfs_file_extent_item *item;
1809 struct extent_buffer *leaf;
1810 struct btrfs_key found_key;
1811 struct extent_map *em = NULL;
1812 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1813 struct btrfs_trans_handle *trans = NULL;
1815 path = btrfs_alloc_path();
1817 mutex_lock(&root->fs_info->fs_mutex);
1820 em = lookup_extent_mapping(em_tree, start, end);
1822 if (em->start > start) {
1823 printk("get_extent start %Lu em start %Lu\n",
1830 em = alloc_extent_map(GFP_NOFS);
1835 em->start = EXTENT_MAP_HOLE;
1836 em->end = EXTENT_MAP_HOLE;
1838 em->bdev = inode->i_sb->s_bdev;
1839 ret = btrfs_lookup_file_extent(trans, root, path,
1840 objectid, start, trans != NULL);
1847 if (path->slots[0] == 0)
1852 leaf = path->nodes[0];
1853 item = btrfs_item_ptr(leaf, path->slots[0],
1854 struct btrfs_file_extent_item);
1855 /* are we inside the extent that was found? */
1856 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1857 found_type = btrfs_key_type(&found_key);
1858 if (found_key.objectid != objectid ||
1859 found_type != BTRFS_EXTENT_DATA_KEY) {
1863 found_type = btrfs_file_extent_type(leaf, item);
1864 extent_start = found_key.offset;
1865 if (found_type == BTRFS_FILE_EXTENT_REG) {
1866 extent_end = extent_start +
1867 btrfs_file_extent_num_bytes(leaf, item);
1869 if (start < extent_start || start >= extent_end) {
1871 if (start < extent_start) {
1872 if (end < extent_start)
1874 em->end = extent_end - 1;
1880 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1882 em->start = extent_start;
1883 em->end = extent_end - 1;
1884 em->block_start = EXTENT_MAP_HOLE;
1885 em->block_end = EXTENT_MAP_HOLE;
1888 bytenr += btrfs_file_extent_offset(leaf, item);
1889 em->block_start = bytenr;
1890 em->block_end = em->block_start +
1891 btrfs_file_extent_num_bytes(leaf, item) - 1;
1892 em->start = extent_start;
1893 em->end = extent_end - 1;
1895 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1899 size_t extent_offset;
1902 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1904 extent_end = (extent_start + size - 1) |
1905 ((u64)root->sectorsize - 1);
1906 if (start < extent_start || start >= extent_end) {
1908 if (start < extent_start) {
1909 if (end < extent_start)
1911 em->end = extent_end;
1917 em->block_start = EXTENT_MAP_INLINE;
1918 em->block_end = EXTENT_MAP_INLINE;
1921 em->start = extent_start;
1922 em->end = extent_start + size - 1;
1926 extent_offset = ((u64)page->index << PAGE_CACHE_SHIFT) -
1927 extent_start + page_offset;
1928 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
1929 size - extent_offset);
1930 em->start = extent_start + extent_offset;
1931 em->end = (em->start + copy_size -1) |
1932 ((u64)root->sectorsize -1);
1934 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
1935 if (create == 0 && !PageUptodate(page)) {
1936 read_extent_buffer(leaf, map + page_offset, ptr,
1938 flush_dcache_page(page);
1939 } else if (create && PageUptodate(page)) {
1942 free_extent_map(em);
1944 btrfs_release_path(root, path);
1945 trans = btrfs_start_transaction(root, 1);
1948 write_extent_buffer(leaf, map + page_offset, ptr,
1950 btrfs_mark_buffer_dirty(leaf);
1953 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
1956 printk("unkknown found_type %d\n", found_type);
1963 em->block_start = EXTENT_MAP_HOLE;
1964 em->block_end = EXTENT_MAP_HOLE;
1966 btrfs_release_path(root, path);
1967 if (em->start > start || em->end < start) {
1968 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1972 ret = add_extent_mapping(em_tree, em);
1973 if (ret == -EEXIST) {
1974 free_extent_map(em);
1976 if (0 && failed_insert == 1) {
1977 btrfs_drop_extent_cache(inode, start, end);
1980 if (failed_insert > 5) {
1981 printk("failing to insert %Lu %Lu\n", start, end);
1989 btrfs_free_path(path);
1991 ret = btrfs_end_transaction(trans, root);
1995 mutex_unlock(&root->fs_info->fs_mutex);
1997 free_extent_map(em);
1999 return ERR_PTR(err);
2004 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2006 return extent_bmap(mapping, iblock, btrfs_get_extent);
2009 int btrfs_readpage(struct file *file, struct page *page)
2011 struct extent_map_tree *tree;
2012 tree = &BTRFS_I(page->mapping->host)->extent_tree;
2013 return extent_read_full_page(tree, page, btrfs_get_extent);
2016 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2018 struct extent_map_tree *tree;
2021 if (current->flags & PF_MEMALLOC) {
2022 redirty_page_for_writepage(wbc, page);
2026 tree = &BTRFS_I(page->mapping->host)->extent_tree;
2027 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2030 static int btrfs_writepages(struct address_space *mapping,
2031 struct writeback_control *wbc)
2033 struct extent_map_tree *tree;
2034 tree = &BTRFS_I(mapping->host)->extent_tree;
2035 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2039 btrfs_readpages(struct file *file, struct address_space *mapping,
2040 struct list_head *pages, unsigned nr_pages)
2042 struct extent_map_tree *tree;
2043 tree = &BTRFS_I(mapping->host)->extent_tree;
2044 return extent_readpages(tree, mapping, pages, nr_pages,
2048 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
2050 struct extent_map_tree *tree;
2053 tree = &BTRFS_I(page->mapping->host)->extent_tree;
2054 ret = try_release_extent_mapping(tree, page);
2056 ClearPagePrivate(page);
2057 set_page_private(page, 0);
2058 page_cache_release(page);
2063 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2065 struct extent_map_tree *tree;
2067 tree = &BTRFS_I(page->mapping->host)->extent_tree;
2068 extent_invalidatepage(tree, page, offset);
2069 btrfs_releasepage(page, GFP_NOFS);
2073 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2074 * called from a page fault handler when a page is first dirtied. Hence we must
2075 * be careful to check for EOF conditions here. We set the page up correctly
2076 * for a written page which means we get ENOSPC checking when writing into
2077 * holes and correct delalloc and unwritten extent mapping on filesystems that
2078 * support these features.
2080 * We are not allowed to take the i_mutex here so we have to play games to
2081 * protect against truncate races as the page could now be beyond EOF. Because
2082 * vmtruncate() writes the inode size before removing pages, once we have the
2083 * page lock we can determine safely if the page is beyond EOF. If it is not
2084 * beyond EOF, then the page is guaranteed safe against truncation until we
2087 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2089 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2090 struct btrfs_root *root = BTRFS_I(inode)->root;
2096 mutex_lock(&root->fs_info->fs_mutex);
2097 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2098 mutex_unlock(&root->fs_info->fs_mutex);
2105 wait_on_page_writeback(page);
2106 size = i_size_read(inode);
2107 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2109 if ((page->mapping != inode->i_mapping) ||
2110 (page_start > size)) {
2111 /* page got truncated out from underneath us */
2115 /* page is wholly or partially inside EOF */
2116 if (page_start + PAGE_CACHE_SIZE > size)
2117 end = size & ~PAGE_CACHE_MASK;
2119 end = PAGE_CACHE_SIZE;
2121 ret = btrfs_cow_one_page(inode, page, end);
2129 static void btrfs_truncate(struct inode *inode)
2131 struct btrfs_root *root = BTRFS_I(inode)->root;
2133 struct btrfs_trans_handle *trans;
2136 if (!S_ISREG(inode->i_mode))
2138 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2141 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2143 mutex_lock(&root->fs_info->fs_mutex);
2144 trans = btrfs_start_transaction(root, 1);
2145 btrfs_set_trans_block_group(trans, inode);
2147 /* FIXME, add redo link to tree so we don't leak on crash */
2148 ret = btrfs_truncate_in_trans(trans, root, inode);
2149 btrfs_update_inode(trans, root, inode);
2150 nr = trans->blocks_used;
2152 ret = btrfs_end_transaction(trans, root);
2154 mutex_unlock(&root->fs_info->fs_mutex);
2155 btrfs_btree_balance_dirty(root, nr);
2156 btrfs_throttle(root);
2159 static int noinline create_subvol(struct btrfs_root *root, char *name,
2162 struct btrfs_trans_handle *trans;
2163 struct btrfs_key key;
2164 struct btrfs_root_item root_item;
2165 struct btrfs_inode_item *inode_item;
2166 struct extent_buffer *leaf;
2167 struct btrfs_root *new_root = root;
2168 struct inode *inode;
2173 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
2174 unsigned long nr = 1;
2176 mutex_lock(&root->fs_info->fs_mutex);
2177 ret = btrfs_check_free_space(root, 1, 0);
2181 trans = btrfs_start_transaction(root, 1);
2184 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2189 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
2190 objectid, trans->transid, 0, 0,
2193 return PTR_ERR(leaf);
2195 btrfs_set_header_nritems(leaf, 0);
2196 btrfs_set_header_level(leaf, 0);
2197 btrfs_set_header_bytenr(leaf, leaf->start);
2198 btrfs_set_header_generation(leaf, trans->transid);
2199 btrfs_set_header_owner(leaf, objectid);
2201 write_extent_buffer(leaf, root->fs_info->fsid,
2202 (unsigned long)btrfs_header_fsid(leaf),
2204 btrfs_mark_buffer_dirty(leaf);
2206 inode_item = &root_item.inode;
2207 memset(inode_item, 0, sizeof(*inode_item));
2208 inode_item->generation = cpu_to_le64(1);
2209 inode_item->size = cpu_to_le64(3);
2210 inode_item->nlink = cpu_to_le32(1);
2211 inode_item->nblocks = cpu_to_le64(1);
2212 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
2214 btrfs_set_root_bytenr(&root_item, leaf->start);
2215 btrfs_set_root_level(&root_item, 0);
2216 btrfs_set_root_refs(&root_item, 1);
2217 btrfs_set_root_used(&root_item, 0);
2219 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
2220 root_item.drop_level = 0;
2222 free_extent_buffer(leaf);
2225 btrfs_set_root_dirid(&root_item, new_dirid);
2227 key.objectid = objectid;
2229 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2230 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2236 * insert the directory item
2238 key.offset = (u64)-1;
2239 dir = root->fs_info->sb->s_root->d_inode;
2240 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2241 name, namelen, dir->i_ino, &key,
2246 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2247 name, namelen, objectid,
2248 root->fs_info->sb->s_root->d_inode->i_ino);
2252 ret = btrfs_commit_transaction(trans, root);
2256 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
2259 trans = btrfs_start_transaction(new_root, 1);
2262 inode = btrfs_new_inode(trans, new_root, new_dirid,
2263 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2266 inode->i_op = &btrfs_dir_inode_operations;
2267 inode->i_fop = &btrfs_dir_file_operations;
2268 new_root->inode = inode;
2270 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2274 ret = btrfs_update_inode(trans, new_root, inode);
2278 nr = trans->blocks_used;
2279 err = btrfs_commit_transaction(trans, new_root);
2283 mutex_unlock(&root->fs_info->fs_mutex);
2284 btrfs_btree_balance_dirty(root, nr);
2285 btrfs_throttle(root);
2289 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2291 struct btrfs_pending_snapshot *pending_snapshot;
2292 struct btrfs_trans_handle *trans;
2295 unsigned long nr = 0;
2297 if (!root->ref_cows)
2300 mutex_lock(&root->fs_info->fs_mutex);
2301 ret = btrfs_check_free_space(root, 1, 0);
2305 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
2306 if (!pending_snapshot) {
2310 pending_snapshot->name = kstrndup(name, namelen, GFP_NOFS);
2311 if (!pending_snapshot->name) {
2313 kfree(pending_snapshot);
2316 trans = btrfs_start_transaction(root, 1);
2319 pending_snapshot->root = root;
2320 list_add(&pending_snapshot->list,
2321 &trans->transaction->pending_snapshots);
2322 ret = btrfs_update_inode(trans, root, root->inode);
2323 err = btrfs_commit_transaction(trans, root);
2326 mutex_unlock(&root->fs_info->fs_mutex);
2327 btrfs_btree_balance_dirty(root, nr);
2328 btrfs_throttle(root);
2332 unsigned long btrfs_force_ra(struct address_space *mapping,
2333 struct file_ra_state *ra, struct file *file,
2334 pgoff_t offset, pgoff_t last_index)
2338 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2339 req_size = last_index - offset + 1;
2340 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2343 req_size = min(last_index - offset + 1, (pgoff_t)128);
2344 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2345 return offset + req_size;
2349 int btrfs_defrag_file(struct file *file) {
2350 struct inode *inode = fdentry(file)->d_inode;
2351 struct btrfs_root *root = BTRFS_I(inode)->root;
2352 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2354 unsigned long last_index;
2355 unsigned long ra_index = 0;
2359 u64 existing_delalloc;
2363 mutex_lock(&root->fs_info->fs_mutex);
2364 ret = btrfs_check_free_space(root, inode->i_size, 0);
2365 mutex_unlock(&root->fs_info->fs_mutex);
2369 mutex_lock(&inode->i_mutex);
2370 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2371 for (i = 0; i <= last_index; i++) {
2372 if (i == ra_index) {
2373 ra_index = btrfs_force_ra(inode->i_mapping,
2375 file, ra_index, last_index);
2377 page = grab_cache_page(inode->i_mapping, i);
2380 if (!PageUptodate(page)) {
2381 btrfs_readpage(NULL, page);
2383 if (!PageUptodate(page)) {
2385 page_cache_release(page);
2389 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2390 page_end = page_start + PAGE_CACHE_SIZE - 1;
2392 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2393 delalloc_start = page_start;
2395 count_range_bits(&BTRFS_I(inode)->extent_tree,
2396 &delalloc_start, page_end,
2397 PAGE_CACHE_SIZE, EXTENT_DELALLOC);
2398 set_extent_delalloc(em_tree, page_start,
2399 page_end, GFP_NOFS);
2401 spin_lock(&root->fs_info->delalloc_lock);
2402 root->fs_info->delalloc_bytes += PAGE_CACHE_SIZE -
2404 spin_unlock(&root->fs_info->delalloc_lock);
2406 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2407 set_page_dirty(page);
2409 page_cache_release(page);
2410 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2414 mutex_unlock(&inode->i_mutex);
2418 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
2422 struct btrfs_ioctl_vol_args *vol_args;
2423 struct btrfs_trans_handle *trans;
2429 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2434 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2438 namelen = strlen(vol_args->name);
2439 if (namelen > BTRFS_VOL_NAME_MAX) {
2444 sizestr = vol_args->name;
2445 if (!strcmp(sizestr, "max"))
2446 new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
2448 if (sizestr[0] == '-') {
2451 } else if (sizestr[0] == '+') {
2455 new_size = btrfs_parse_size(sizestr);
2456 if (new_size == 0) {
2462 mutex_lock(&root->fs_info->fs_mutex);
2463 old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
2466 if (new_size > old_size) {
2470 new_size = old_size - new_size;
2471 } else if (mod > 0) {
2472 new_size = old_size + new_size;
2475 if (new_size < 256 * 1024 * 1024) {
2479 if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
2484 do_div(new_size, root->sectorsize);
2485 new_size *= root->sectorsize;
2487 printk("new size is %Lu\n", new_size);
2488 if (new_size > old_size) {
2489 trans = btrfs_start_transaction(root, 1);
2490 ret = btrfs_grow_extent_tree(trans, root, new_size);
2491 btrfs_commit_transaction(trans, root);
2493 ret = btrfs_shrink_extent_tree(root, new_size);
2497 mutex_unlock(&root->fs_info->fs_mutex);
2503 static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root,
2506 struct btrfs_ioctl_vol_args *vol_args;
2507 struct btrfs_dir_item *di;
2508 struct btrfs_path *path;
2513 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2518 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2523 namelen = strlen(vol_args->name);
2524 if (namelen > BTRFS_VOL_NAME_MAX) {
2528 if (strchr(vol_args->name, '/')) {
2533 path = btrfs_alloc_path();
2539 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2540 mutex_lock(&root->fs_info->fs_mutex);
2541 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2543 vol_args->name, namelen, 0);
2544 mutex_unlock(&root->fs_info->fs_mutex);
2545 btrfs_free_path(path);
2547 if (di && !IS_ERR(di)) {
2557 if (root == root->fs_info->tree_root)
2558 ret = create_subvol(root, vol_args->name, namelen);
2560 ret = create_snapshot(root, vol_args->name, namelen);
2566 static int btrfs_ioctl_defrag(struct file *file)
2568 struct inode *inode = fdentry(file)->d_inode;
2569 struct btrfs_root *root = BTRFS_I(inode)->root;
2571 switch (inode->i_mode & S_IFMT) {
2573 mutex_lock(&root->fs_info->fs_mutex);
2574 btrfs_defrag_root(root, 0);
2575 btrfs_defrag_root(root->fs_info->extent_root, 0);
2576 mutex_unlock(&root->fs_info->fs_mutex);
2579 btrfs_defrag_file(file);
2586 long btrfs_ioctl(struct file *file, unsigned int
2587 cmd, unsigned long arg)
2589 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2592 case BTRFS_IOC_SNAP_CREATE:
2593 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2594 case BTRFS_IOC_DEFRAG:
2595 return btrfs_ioctl_defrag(file);
2596 case BTRFS_IOC_RESIZE:
2597 return btrfs_ioctl_resize(root, (void __user *)arg);
2604 * Called inside transaction, so use GFP_NOFS
2606 struct inode *btrfs_alloc_inode(struct super_block *sb)
2608 struct btrfs_inode *ei;
2610 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2614 ei->ordered_trans = 0;
2615 return &ei->vfs_inode;
2618 void btrfs_destroy_inode(struct inode *inode)
2620 WARN_ON(!list_empty(&inode->i_dentry));
2621 WARN_ON(inode->i_data.nrpages);
2623 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2624 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2627 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2628 static void init_once(struct kmem_cache * cachep, void *foo)
2630 static void init_once(void * foo, struct kmem_cache * cachep,
2631 unsigned long flags)
2634 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2636 inode_init_once(&ei->vfs_inode);
2639 void btrfs_destroy_cachep(void)
2641 if (btrfs_inode_cachep)
2642 kmem_cache_destroy(btrfs_inode_cachep);
2643 if (btrfs_trans_handle_cachep)
2644 kmem_cache_destroy(btrfs_trans_handle_cachep);
2645 if (btrfs_transaction_cachep)
2646 kmem_cache_destroy(btrfs_transaction_cachep);
2647 if (btrfs_bit_radix_cachep)
2648 kmem_cache_destroy(btrfs_bit_radix_cachep);
2649 if (btrfs_path_cachep)
2650 kmem_cache_destroy(btrfs_path_cachep);
2653 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2654 unsigned long extra_flags,
2655 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2656 void (*ctor)(struct kmem_cache *, void *)
2658 void (*ctor)(void *, struct kmem_cache *,
2663 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2664 SLAB_MEM_SPREAD | extra_flags), ctor
2665 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2671 int btrfs_init_cachep(void)
2673 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2674 sizeof(struct btrfs_inode),
2676 if (!btrfs_inode_cachep)
2678 btrfs_trans_handle_cachep =
2679 btrfs_cache_create("btrfs_trans_handle_cache",
2680 sizeof(struct btrfs_trans_handle),
2682 if (!btrfs_trans_handle_cachep)
2684 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2685 sizeof(struct btrfs_transaction),
2687 if (!btrfs_transaction_cachep)
2689 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2690 sizeof(struct btrfs_path),
2692 if (!btrfs_path_cachep)
2694 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2695 SLAB_DESTROY_BY_RCU, NULL);
2696 if (!btrfs_bit_radix_cachep)
2700 btrfs_destroy_cachep();
2704 static int btrfs_getattr(struct vfsmount *mnt,
2705 struct dentry *dentry, struct kstat *stat)
2707 struct inode *inode = dentry->d_inode;
2708 generic_fillattr(inode, stat);
2709 stat->blksize = PAGE_CACHE_SIZE;
2713 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2714 struct inode * new_dir,struct dentry *new_dentry)
2716 struct btrfs_trans_handle *trans;
2717 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2718 struct inode *new_inode = new_dentry->d_inode;
2719 struct inode *old_inode = old_dentry->d_inode;
2720 struct timespec ctime = CURRENT_TIME;
2721 struct btrfs_path *path;
2724 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2725 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2729 mutex_lock(&root->fs_info->fs_mutex);
2730 ret = btrfs_check_free_space(root, 1, 0);
2734 trans = btrfs_start_transaction(root, 1);
2736 btrfs_set_trans_block_group(trans, new_dir);
2737 path = btrfs_alloc_path();
2743 old_dentry->d_inode->i_nlink++;
2744 old_dir->i_ctime = old_dir->i_mtime = ctime;
2745 new_dir->i_ctime = new_dir->i_mtime = ctime;
2746 old_inode->i_ctime = ctime;
2748 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2753 new_inode->i_ctime = CURRENT_TIME;
2754 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2758 ret = btrfs_add_link(trans, new_dentry, old_inode);
2763 btrfs_free_path(path);
2764 btrfs_end_transaction(trans, root);
2766 mutex_unlock(&root->fs_info->fs_mutex);
2770 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2771 const char *symname)
2773 struct btrfs_trans_handle *trans;
2774 struct btrfs_root *root = BTRFS_I(dir)->root;
2775 struct btrfs_path *path;
2776 struct btrfs_key key;
2777 struct inode *inode = NULL;
2784 struct btrfs_file_extent_item *ei;
2785 struct extent_buffer *leaf;
2786 unsigned long nr = 0;
2788 name_len = strlen(symname) + 1;
2789 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2790 return -ENAMETOOLONG;
2792 mutex_lock(&root->fs_info->fs_mutex);
2793 err = btrfs_check_free_space(root, 1, 0);
2797 trans = btrfs_start_transaction(root, 1);
2798 btrfs_set_trans_block_group(trans, dir);
2800 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2806 inode = btrfs_new_inode(trans, root, objectid,
2807 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2808 err = PTR_ERR(inode);
2812 btrfs_set_trans_block_group(trans, inode);
2813 err = btrfs_add_nondir(trans, dentry, inode);
2817 inode->i_mapping->a_ops = &btrfs_aops;
2818 inode->i_fop = &btrfs_file_operations;
2819 inode->i_op = &btrfs_file_inode_operations;
2820 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2821 inode->i_mapping, GFP_NOFS);
2822 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2824 dir->i_sb->s_dirt = 1;
2825 btrfs_update_inode_block_group(trans, inode);
2826 btrfs_update_inode_block_group(trans, dir);
2830 path = btrfs_alloc_path();
2832 key.objectid = inode->i_ino;
2834 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2835 datasize = btrfs_file_extent_calc_inline_size(name_len);
2836 err = btrfs_insert_empty_item(trans, root, path, &key,
2842 leaf = path->nodes[0];
2843 ei = btrfs_item_ptr(leaf, path->slots[0],
2844 struct btrfs_file_extent_item);
2845 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2846 btrfs_set_file_extent_type(leaf, ei,
2847 BTRFS_FILE_EXTENT_INLINE);
2848 ptr = btrfs_file_extent_inline_start(ei);
2849 write_extent_buffer(leaf, symname, ptr, name_len);
2850 btrfs_mark_buffer_dirty(leaf);
2851 btrfs_free_path(path);
2853 inode->i_op = &btrfs_symlink_inode_operations;
2854 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2855 inode->i_size = name_len - 1;
2856 err = btrfs_update_inode(trans, root, inode);
2861 nr = trans->blocks_used;
2862 btrfs_end_transaction(trans, root);
2864 mutex_unlock(&root->fs_info->fs_mutex);
2866 inode_dec_link_count(inode);
2869 btrfs_btree_balance_dirty(root, nr);
2870 btrfs_throttle(root);
2873 static int btrfs_permission(struct inode *inode, int mask,
2874 struct nameidata *nd)
2876 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
2878 return generic_permission(inode, mask, NULL);
2881 static struct inode_operations btrfs_dir_inode_operations = {
2882 .lookup = btrfs_lookup,
2883 .create = btrfs_create,
2884 .unlink = btrfs_unlink,
2886 .mkdir = btrfs_mkdir,
2887 .rmdir = btrfs_rmdir,
2888 .rename = btrfs_rename,
2889 .symlink = btrfs_symlink,
2890 .setattr = btrfs_setattr,
2891 .mknod = btrfs_mknod,
2892 .setxattr = generic_setxattr,
2893 .getxattr = generic_getxattr,
2894 .listxattr = btrfs_listxattr,
2895 .removexattr = generic_removexattr,
2896 .permission = btrfs_permission,
2898 static struct inode_operations btrfs_dir_ro_inode_operations = {
2899 .lookup = btrfs_lookup,
2900 .permission = btrfs_permission,
2902 static struct file_operations btrfs_dir_file_operations = {
2903 .llseek = generic_file_llseek,
2904 .read = generic_read_dir,
2905 .readdir = btrfs_readdir,
2906 .unlocked_ioctl = btrfs_ioctl,
2907 #ifdef CONFIG_COMPAT
2908 .compat_ioctl = btrfs_ioctl,
2912 static struct extent_map_ops btrfs_extent_map_ops = {
2913 .fill_delalloc = run_delalloc_range,
2914 .writepage_io_hook = btrfs_writepage_io_hook,
2915 .readpage_io_hook = btrfs_readpage_io_hook,
2916 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2919 static struct address_space_operations btrfs_aops = {
2920 .readpage = btrfs_readpage,
2921 .writepage = btrfs_writepage,
2922 .writepages = btrfs_writepages,
2923 .readpages = btrfs_readpages,
2924 .sync_page = block_sync_page,
2926 .invalidatepage = btrfs_invalidatepage,
2927 .releasepage = btrfs_releasepage,
2928 .set_page_dirty = __set_page_dirty_nobuffers,
2931 static struct address_space_operations btrfs_symlink_aops = {
2932 .readpage = btrfs_readpage,
2933 .writepage = btrfs_writepage,
2934 .invalidatepage = btrfs_invalidatepage,
2935 .releasepage = btrfs_releasepage,
2938 static struct inode_operations btrfs_file_inode_operations = {
2939 .truncate = btrfs_truncate,
2940 .getattr = btrfs_getattr,
2941 .setattr = btrfs_setattr,
2942 .setxattr = generic_setxattr,
2943 .getxattr = generic_getxattr,
2944 .listxattr = btrfs_listxattr,
2945 .removexattr = generic_removexattr,
2946 .permission = btrfs_permission,
2948 static struct inode_operations btrfs_special_inode_operations = {
2949 .getattr = btrfs_getattr,
2950 .setattr = btrfs_setattr,
2951 .permission = btrfs_permission,
2953 static struct inode_operations btrfs_symlink_inode_operations = {
2954 .readlink = generic_readlink,
2955 .follow_link = page_follow_link_light,
2956 .put_link = page_put_link,
2957 .permission = btrfs_permission,