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_io_ops btrfs_extent_io_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 u64 orig_start = start;
107 struct btrfs_key ins;
110 trans = btrfs_start_transaction(root, 1);
112 btrfs_set_trans_block_group(trans, inode);
114 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
115 num_bytes = max(blocksize, num_bytes);
116 ret = btrfs_drop_extents(trans, root, inode,
117 start, start + num_bytes, start, &alloc_hint);
118 orig_num_bytes = num_bytes;
120 if (alloc_hint == EXTENT_MAP_INLINE)
123 while(num_bytes > 0) {
124 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
125 ret = btrfs_alloc_extent(trans, root, cur_alloc_size,
126 root->root_key.objectid,
128 inode->i_ino, start, 0,
129 alloc_hint, (u64)-1, &ins, 1);
134 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
135 start, ins.objectid, ins.offset,
137 btrfs_check_file(root, inode);
138 num_bytes -= cur_alloc_size;
139 alloc_hint = ins.objectid + ins.offset;
140 start += cur_alloc_size;
142 btrfs_drop_extent_cache(inode, orig_start,
143 orig_start + orig_num_bytes - 1);
144 btrfs_add_ordered_inode(inode);
146 btrfs_end_transaction(trans, root);
150 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
158 struct btrfs_root *root = BTRFS_I(inode)->root;
159 struct extent_buffer *leaf;
161 struct btrfs_path *path;
162 struct btrfs_file_extent_item *item;
165 struct btrfs_key found_key;
167 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
168 path = btrfs_alloc_path();
171 ret = btrfs_lookup_file_extent(NULL, root, path,
172 inode->i_ino, start, 0);
174 btrfs_free_path(path);
180 if (path->slots[0] == 0)
185 leaf = path->nodes[0];
186 item = btrfs_item_ptr(leaf, path->slots[0],
187 struct btrfs_file_extent_item);
189 /* are we inside the extent that was found? */
190 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
191 found_type = btrfs_key_type(&found_key);
192 if (found_key.objectid != inode->i_ino ||
193 found_type != BTRFS_EXTENT_DATA_KEY) {
197 found_type = btrfs_file_extent_type(leaf, item);
198 extent_start = found_key.offset;
199 if (found_type == BTRFS_FILE_EXTENT_REG) {
200 u64 extent_num_bytes;
202 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
203 extent_end = extent_start + extent_num_bytes;
206 if (loops && start != extent_start)
209 if (start < extent_start || start >= extent_end)
212 cow_end = min(end, extent_end - 1);
213 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
218 * we may be called by the resizer, make sure we're inside
219 * the limits of the FS
221 if (bytenr + extent_num_bytes > total_fs_bytes)
224 if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) {
234 btrfs_free_path(path);
237 btrfs_release_path(root, path);
242 cow_file_range(inode, start, cow_end);
247 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
249 struct btrfs_root *root = BTRFS_I(inode)->root;
251 mutex_lock(&root->fs_info->fs_mutex);
252 if (btrfs_test_opt(root, NODATACOW) ||
253 btrfs_test_flag(inode, NODATACOW))
254 ret = run_delalloc_nocow(inode, start, end);
256 ret = cow_file_range(inode, start, end);
258 mutex_unlock(&root->fs_info->fs_mutex);
262 int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
265 if ((bits & EXTENT_DELALLOC)) {
266 struct btrfs_root *root = BTRFS_I(inode)->root;
267 spin_lock(&root->fs_info->delalloc_lock);
268 root->fs_info->delalloc_bytes += end - start + 1;
269 spin_unlock(&root->fs_info->delalloc_lock);
274 int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
277 if ((bits & EXTENT_DELALLOC)) {
278 struct btrfs_root *root = BTRFS_I(inode)->root;
279 spin_lock(&root->fs_info->delalloc_lock);
280 root->fs_info->delalloc_bytes -= end - start + 1;
281 spin_unlock(&root->fs_info->delalloc_lock);
286 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
288 struct inode *inode = page->mapping->host;
289 struct btrfs_root *root = BTRFS_I(inode)->root;
290 struct btrfs_trans_handle *trans;
293 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
294 size_t offset = start - page_start;
295 if (btrfs_test_opt(root, NODATASUM) ||
296 btrfs_test_flag(inode, NODATASUM))
298 mutex_lock(&root->fs_info->fs_mutex);
299 trans = btrfs_start_transaction(root, 1);
300 btrfs_set_trans_block_group(trans, inode);
302 btrfs_csum_file_block(trans, root, inode, inode->i_ino,
303 start, kaddr + offset, end - start + 1);
305 ret = btrfs_end_transaction(trans, root);
307 mutex_unlock(&root->fs_info->fs_mutex);
311 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
314 struct inode *inode = page->mapping->host;
315 struct btrfs_root *root = BTRFS_I(inode)->root;
316 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
317 struct btrfs_csum_item *item;
318 struct btrfs_path *path = NULL;
320 if (btrfs_test_opt(root, NODATASUM) ||
321 btrfs_test_flag(inode, NODATASUM))
323 mutex_lock(&root->fs_info->fs_mutex);
324 path = btrfs_alloc_path();
325 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
328 /* a csum that isn't present is a preallocated region. */
329 if (ret == -ENOENT || ret == -EFBIG)
332 printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start);
335 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
337 set_state_private(io_tree, start, csum);
340 btrfs_free_path(path);
341 mutex_unlock(&root->fs_info->fs_mutex);
345 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
346 struct extent_state *state)
348 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
349 struct inode *inode = page->mapping->host;
350 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
352 u64 private = ~(u32)0;
354 struct btrfs_root *root = BTRFS_I(inode)->root;
358 if (btrfs_test_opt(root, NODATASUM) ||
359 btrfs_test_flag(inode, NODATASUM))
362 if (state->start == start) {
363 private = state->private;
366 ret = get_state_private(io_tree, start, &private);
368 local_irq_save(flags);
369 kaddr = kmap_atomic(page, KM_IRQ0);
373 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
374 btrfs_csum_final(csum, (char *)&csum);
375 if (csum != private) {
378 kunmap_atomic(kaddr, KM_IRQ0);
379 local_irq_restore(flags);
383 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
384 page->mapping->host->i_ino, (unsigned long long)start, csum,
386 memset(kaddr + offset, 1, end - start + 1);
387 flush_dcache_page(page);
388 kunmap_atomic(kaddr, KM_IRQ0);
389 local_irq_restore(flags);
393 void btrfs_read_locked_inode(struct inode *inode)
395 struct btrfs_path *path;
396 struct extent_buffer *leaf;
397 struct btrfs_inode_item *inode_item;
398 struct btrfs_inode_timespec *tspec;
399 struct btrfs_root *root = BTRFS_I(inode)->root;
400 struct btrfs_key location;
401 u64 alloc_group_block;
405 path = btrfs_alloc_path();
407 mutex_lock(&root->fs_info->fs_mutex);
408 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
410 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
414 leaf = path->nodes[0];
415 inode_item = btrfs_item_ptr(leaf, path->slots[0],
416 struct btrfs_inode_item);
418 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
419 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
420 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
421 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
422 inode->i_size = btrfs_inode_size(leaf, inode_item);
424 tspec = btrfs_inode_atime(inode_item);
425 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
426 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
428 tspec = btrfs_inode_mtime(inode_item);
429 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
430 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
432 tspec = btrfs_inode_ctime(inode_item);
433 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
434 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
436 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
437 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
439 rdev = btrfs_inode_rdev(leaf, inode_item);
441 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
442 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
444 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
445 if (!BTRFS_I(inode)->block_group) {
446 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
449 btrfs_free_path(path);
452 mutex_unlock(&root->fs_info->fs_mutex);
454 switch (inode->i_mode & S_IFMT) {
456 inode->i_mapping->a_ops = &btrfs_aops;
457 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
458 inode->i_fop = &btrfs_file_operations;
459 inode->i_op = &btrfs_file_inode_operations;
462 inode->i_fop = &btrfs_dir_file_operations;
463 if (root == root->fs_info->tree_root)
464 inode->i_op = &btrfs_dir_ro_inode_operations;
466 inode->i_op = &btrfs_dir_inode_operations;
469 inode->i_op = &btrfs_symlink_inode_operations;
470 inode->i_mapping->a_ops = &btrfs_symlink_aops;
473 init_special_inode(inode, inode->i_mode, rdev);
479 btrfs_release_path(root, path);
480 btrfs_free_path(path);
481 mutex_unlock(&root->fs_info->fs_mutex);
482 make_bad_inode(inode);
485 static void fill_inode_item(struct extent_buffer *leaf,
486 struct btrfs_inode_item *item,
489 btrfs_set_inode_uid(leaf, item, inode->i_uid);
490 btrfs_set_inode_gid(leaf, item, inode->i_gid);
491 btrfs_set_inode_size(leaf, item, inode->i_size);
492 btrfs_set_inode_mode(leaf, item, inode->i_mode);
493 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
495 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
496 inode->i_atime.tv_sec);
497 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
498 inode->i_atime.tv_nsec);
500 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
501 inode->i_mtime.tv_sec);
502 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
503 inode->i_mtime.tv_nsec);
505 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
506 inode->i_ctime.tv_sec);
507 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
508 inode->i_ctime.tv_nsec);
510 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
511 btrfs_set_inode_generation(leaf, item, inode->i_generation);
512 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
513 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
514 btrfs_set_inode_block_group(leaf, item,
515 BTRFS_I(inode)->block_group->key.objectid);
518 int btrfs_update_inode(struct btrfs_trans_handle *trans,
519 struct btrfs_root *root,
522 struct btrfs_inode_item *inode_item;
523 struct btrfs_path *path;
524 struct extent_buffer *leaf;
527 path = btrfs_alloc_path();
529 ret = btrfs_lookup_inode(trans, root, path,
530 &BTRFS_I(inode)->location, 1);
537 leaf = path->nodes[0];
538 inode_item = btrfs_item_ptr(leaf, path->slots[0],
539 struct btrfs_inode_item);
541 fill_inode_item(leaf, inode_item, inode);
542 btrfs_mark_buffer_dirty(leaf);
543 btrfs_set_inode_last_trans(trans, inode);
546 btrfs_release_path(root, path);
547 btrfs_free_path(path);
552 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
553 struct btrfs_root *root,
555 struct dentry *dentry)
557 struct btrfs_path *path;
558 const char *name = dentry->d_name.name;
559 int name_len = dentry->d_name.len;
561 struct extent_buffer *leaf;
562 struct btrfs_dir_item *di;
563 struct btrfs_key key;
565 path = btrfs_alloc_path();
571 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
581 leaf = path->nodes[0];
582 btrfs_dir_item_key_to_cpu(leaf, di, &key);
583 ret = btrfs_delete_one_dir_name(trans, root, path, di);
586 btrfs_release_path(root, path);
588 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
589 key.objectid, name, name_len, -1);
598 ret = btrfs_delete_one_dir_name(trans, root, path, di);
600 dentry->d_inode->i_ctime = dir->i_ctime;
601 ret = btrfs_del_inode_ref(trans, root, name, name_len,
602 dentry->d_inode->i_ino,
603 dentry->d_parent->d_inode->i_ino);
605 printk("failed to delete reference to %.*s, "
606 "inode %lu parent %lu\n", name_len, name,
607 dentry->d_inode->i_ino,
608 dentry->d_parent->d_inode->i_ino);
611 btrfs_free_path(path);
613 dir->i_size -= name_len * 2;
614 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
615 btrfs_update_inode(trans, root, dir);
616 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
617 dentry->d_inode->i_nlink--;
619 drop_nlink(dentry->d_inode);
621 ret = btrfs_update_inode(trans, root, dentry->d_inode);
622 dir->i_sb->s_dirt = 1;
627 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
629 struct btrfs_root *root;
630 struct btrfs_trans_handle *trans;
631 struct inode *inode = dentry->d_inode;
633 unsigned long nr = 0;
635 root = BTRFS_I(dir)->root;
636 mutex_lock(&root->fs_info->fs_mutex);
638 ret = btrfs_check_free_space(root, 1, 1);
642 trans = btrfs_start_transaction(root, 1);
644 btrfs_set_trans_block_group(trans, dir);
645 ret = btrfs_unlink_trans(trans, root, dir, dentry);
646 nr = trans->blocks_used;
648 if (inode->i_nlink == 0) {
650 /* if the inode isn't linked anywhere,
651 * we don't need to worry about
654 found = btrfs_del_ordered_inode(inode);
656 atomic_dec(&inode->i_count);
660 btrfs_end_transaction(trans, root);
662 mutex_unlock(&root->fs_info->fs_mutex);
663 btrfs_btree_balance_dirty(root, nr);
664 btrfs_throttle(root);
668 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
670 struct inode *inode = dentry->d_inode;
673 struct btrfs_root *root = BTRFS_I(dir)->root;
674 struct btrfs_trans_handle *trans;
675 unsigned long nr = 0;
677 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
680 mutex_lock(&root->fs_info->fs_mutex);
681 ret = btrfs_check_free_space(root, 1, 1);
685 trans = btrfs_start_transaction(root, 1);
686 btrfs_set_trans_block_group(trans, dir);
688 /* now the directory is empty */
689 err = btrfs_unlink_trans(trans, root, dir, dentry);
694 nr = trans->blocks_used;
695 ret = btrfs_end_transaction(trans, root);
697 mutex_unlock(&root->fs_info->fs_mutex);
698 btrfs_btree_balance_dirty(root, nr);
699 btrfs_throttle(root);
707 * this can truncate away extent items, csum items and directory items.
708 * It starts at a high offset and removes keys until it can't find
709 * any higher than i_size.
711 * csum items that cross the new i_size are truncated to the new size
714 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
715 struct btrfs_root *root,
720 struct btrfs_path *path;
721 struct btrfs_key key;
722 struct btrfs_key found_key;
724 struct extent_buffer *leaf;
725 struct btrfs_file_extent_item *fi;
726 u64 extent_start = 0;
727 u64 extent_num_bytes = 0;
733 int pending_del_nr = 0;
734 int pending_del_slot = 0;
735 int extent_type = -1;
737 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
738 path = btrfs_alloc_path();
742 /* FIXME, add redo link to tree so we don't leak on crash */
743 key.objectid = inode->i_ino;
744 key.offset = (u64)-1;
747 btrfs_init_path(path);
749 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
754 BUG_ON(path->slots[0] == 0);
760 leaf = path->nodes[0];
761 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
762 found_type = btrfs_key_type(&found_key);
764 if (found_key.objectid != inode->i_ino)
767 if (found_type < min_type)
770 item_end = found_key.offset;
771 if (found_type == BTRFS_EXTENT_DATA_KEY) {
772 fi = btrfs_item_ptr(leaf, path->slots[0],
773 struct btrfs_file_extent_item);
774 extent_type = btrfs_file_extent_type(leaf, fi);
775 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
777 btrfs_file_extent_num_bytes(leaf, fi);
778 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
779 struct btrfs_item *item = btrfs_item_nr(leaf,
781 item_end += btrfs_file_extent_inline_len(leaf,
786 if (found_type == BTRFS_CSUM_ITEM_KEY) {
787 ret = btrfs_csum_truncate(trans, root, path,
791 if (item_end < inode->i_size) {
792 if (found_type == BTRFS_DIR_ITEM_KEY) {
793 found_type = BTRFS_INODE_ITEM_KEY;
794 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
795 found_type = BTRFS_CSUM_ITEM_KEY;
796 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
797 found_type = BTRFS_XATTR_ITEM_KEY;
798 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
799 found_type = BTRFS_INODE_REF_KEY;
800 } else if (found_type) {
805 btrfs_set_key_type(&key, found_type);
808 if (found_key.offset >= inode->i_size)
814 /* FIXME, shrink the extent if the ref count is only 1 */
815 if (found_type != BTRFS_EXTENT_DATA_KEY)
818 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
820 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
823 btrfs_file_extent_num_bytes(leaf, fi);
824 extent_num_bytes = inode->i_size -
825 found_key.offset + root->sectorsize - 1;
826 extent_num_bytes = extent_num_bytes &
827 ~((u64)root->sectorsize - 1);
828 btrfs_set_file_extent_num_bytes(leaf, fi,
830 num_dec = (orig_num_bytes -
831 extent_num_bytes) >> 9;
832 if (extent_start != 0) {
833 inode->i_blocks -= num_dec;
835 btrfs_mark_buffer_dirty(leaf);
838 btrfs_file_extent_disk_num_bytes(leaf,
840 /* FIXME blocksize != 4096 */
841 num_dec = btrfs_file_extent_num_bytes(leaf,
843 if (extent_start != 0) {
845 inode->i_blocks -= num_dec;
847 root_gen = btrfs_header_generation(leaf);
848 root_owner = btrfs_header_owner(leaf);
850 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE &&
852 u32 newsize = inode->i_size - found_key.offset;
853 newsize = btrfs_file_extent_calc_inline_size(newsize);
854 ret = btrfs_truncate_item(trans, root, path,
860 if (!pending_del_nr) {
861 /* no pending yet, add ourselves */
862 pending_del_slot = path->slots[0];
864 } else if (pending_del_nr &&
865 path->slots[0] + 1 == pending_del_slot) {
866 /* hop on the pending chunk */
868 pending_del_slot = path->slots[0];
870 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
876 ret = btrfs_free_extent(trans, root, extent_start,
879 root_gen, inode->i_ino,
880 found_key.offset, 0);
884 if (path->slots[0] == 0) {
887 btrfs_release_path(root, path);
892 if (pending_del_nr &&
893 path->slots[0] + 1 != pending_del_slot) {
894 struct btrfs_key debug;
896 btrfs_item_key_to_cpu(path->nodes[0], &debug,
898 ret = btrfs_del_items(trans, root, path,
903 btrfs_release_path(root, path);
909 if (pending_del_nr) {
910 ret = btrfs_del_items(trans, root, path, pending_del_slot,
913 btrfs_release_path(root, path);
914 btrfs_free_path(path);
915 inode->i_sb->s_dirt = 1;
919 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
923 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
924 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
925 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
928 WARN_ON(!PageLocked(page));
929 set_page_extent_mapped(page);
931 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
932 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
935 if (zero_start != PAGE_CACHE_SIZE) {
937 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
938 flush_dcache_page(page);
941 set_page_dirty(page);
942 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
948 * taken from block_truncate_page, but does cow as it zeros out
949 * any bytes left in the last page in the file.
951 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
953 struct inode *inode = mapping->host;
954 struct btrfs_root *root = BTRFS_I(inode)->root;
955 u32 blocksize = root->sectorsize;
956 pgoff_t index = from >> PAGE_CACHE_SHIFT;
957 unsigned offset = from & (PAGE_CACHE_SIZE-1);
962 if ((offset & (blocksize - 1)) == 0)
966 page = grab_cache_page(mapping, index);
969 if (!PageUptodate(page)) {
970 ret = btrfs_readpage(NULL, page);
972 if (!PageUptodate(page)) {
977 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
979 ret = btrfs_cow_one_page(inode, page, offset);
982 page_cache_release(page);
987 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
989 struct inode *inode = dentry->d_inode;
992 err = inode_change_ok(inode, attr);
996 if (S_ISREG(inode->i_mode) &&
997 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
998 struct btrfs_trans_handle *trans;
999 struct btrfs_root *root = BTRFS_I(inode)->root;
1000 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1002 u64 mask = root->sectorsize - 1;
1003 u64 pos = (inode->i_size + mask) & ~mask;
1004 u64 block_end = (attr->ia_size + mask) & ~mask;
1009 if (attr->ia_size <= pos)
1012 if (pos != inode->i_size)
1013 hole_start = pos + root->sectorsize;
1017 mutex_lock(&root->fs_info->fs_mutex);
1018 err = btrfs_check_free_space(root, 1, 0);
1019 mutex_unlock(&root->fs_info->fs_mutex);
1023 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1025 lock_extent(io_tree, pos, block_end - 1, GFP_NOFS);
1026 hole_size = block_end - hole_start;
1028 mutex_lock(&root->fs_info->fs_mutex);
1029 trans = btrfs_start_transaction(root, 1);
1030 btrfs_set_trans_block_group(trans, inode);
1031 err = btrfs_drop_extents(trans, root, inode,
1032 pos, block_end, pos,
1035 if (alloc_hint != EXTENT_MAP_INLINE) {
1036 err = btrfs_insert_file_extent(trans, root,
1040 btrfs_drop_extent_cache(inode, hole_start,
1042 btrfs_check_file(root, inode);
1044 btrfs_end_transaction(trans, root);
1045 mutex_unlock(&root->fs_info->fs_mutex);
1046 unlock_extent(io_tree, pos, block_end - 1, GFP_NOFS);
1051 err = inode_setattr(inode, attr);
1056 void btrfs_put_inode(struct inode *inode)
1060 if (!BTRFS_I(inode)->ordered_trans) {
1064 if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) ||
1065 mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))
1068 ret = btrfs_del_ordered_inode(inode);
1070 atomic_dec(&inode->i_count);
1074 void btrfs_delete_inode(struct inode *inode)
1076 struct btrfs_trans_handle *trans;
1077 struct btrfs_root *root = BTRFS_I(inode)->root;
1081 truncate_inode_pages(&inode->i_data, 0);
1082 if (is_bad_inode(inode)) {
1087 mutex_lock(&root->fs_info->fs_mutex);
1088 trans = btrfs_start_transaction(root, 1);
1090 btrfs_set_trans_block_group(trans, inode);
1091 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1093 goto no_delete_lock;
1095 nr = trans->blocks_used;
1098 btrfs_end_transaction(trans, root);
1099 mutex_unlock(&root->fs_info->fs_mutex);
1100 btrfs_btree_balance_dirty(root, nr);
1101 btrfs_throttle(root);
1105 nr = trans->blocks_used;
1106 btrfs_end_transaction(trans, root);
1107 mutex_unlock(&root->fs_info->fs_mutex);
1108 btrfs_btree_balance_dirty(root, nr);
1109 btrfs_throttle(root);
1115 * this returns the key found in the dir entry in the location pointer.
1116 * If no dir entries were found, location->objectid is 0.
1118 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1119 struct btrfs_key *location)
1121 const char *name = dentry->d_name.name;
1122 int namelen = dentry->d_name.len;
1123 struct btrfs_dir_item *di;
1124 struct btrfs_path *path;
1125 struct btrfs_root *root = BTRFS_I(dir)->root;
1128 if (namelen == 1 && strcmp(name, ".") == 0) {
1129 location->objectid = dir->i_ino;
1130 location->type = BTRFS_INODE_ITEM_KEY;
1131 location->offset = 0;
1134 path = btrfs_alloc_path();
1137 if (namelen == 2 && strcmp(name, "..") == 0) {
1138 struct btrfs_key key;
1139 struct extent_buffer *leaf;
1143 key.objectid = dir->i_ino;
1144 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1146 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1150 leaf = path->nodes[0];
1151 slot = path->slots[0];
1152 nritems = btrfs_header_nritems(leaf);
1153 if (slot >= nritems)
1156 btrfs_item_key_to_cpu(leaf, &key, slot);
1157 if (key.objectid != dir->i_ino ||
1158 key.type != BTRFS_INODE_REF_KEY) {
1161 location->objectid = key.offset;
1162 location->type = BTRFS_INODE_ITEM_KEY;
1163 location->offset = 0;
1167 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1171 if (!di || IS_ERR(di)) {
1174 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1176 btrfs_free_path(path);
1179 location->objectid = 0;
1184 * when we hit a tree root in a directory, the btrfs part of the inode
1185 * needs to be changed to reflect the root directory of the tree root. This
1186 * is kind of like crossing a mount point.
1188 static int fixup_tree_root_location(struct btrfs_root *root,
1189 struct btrfs_key *location,
1190 struct btrfs_root **sub_root,
1191 struct dentry *dentry)
1193 struct btrfs_path *path;
1194 struct btrfs_root_item *ri;
1196 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1198 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1201 path = btrfs_alloc_path();
1203 mutex_lock(&root->fs_info->fs_mutex);
1205 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1206 dentry->d_name.name,
1207 dentry->d_name.len);
1208 if (IS_ERR(*sub_root))
1209 return PTR_ERR(*sub_root);
1211 ri = &(*sub_root)->root_item;
1212 location->objectid = btrfs_root_dirid(ri);
1213 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1214 location->offset = 0;
1216 btrfs_free_path(path);
1217 mutex_unlock(&root->fs_info->fs_mutex);
1221 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1223 struct btrfs_iget_args *args = p;
1224 inode->i_ino = args->ino;
1225 BTRFS_I(inode)->root = args->root;
1226 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1227 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1228 inode->i_mapping, GFP_NOFS);
1232 static int btrfs_find_actor(struct inode *inode, void *opaque)
1234 struct btrfs_iget_args *args = opaque;
1235 return (args->ino == inode->i_ino &&
1236 args->root == BTRFS_I(inode)->root);
1239 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1242 struct btrfs_iget_args args;
1243 args.ino = objectid;
1244 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1249 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1252 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1253 struct btrfs_root *root)
1255 struct inode *inode;
1256 struct btrfs_iget_args args;
1257 args.ino = objectid;
1260 inode = iget5_locked(s, objectid, btrfs_find_actor,
1261 btrfs_init_locked_inode,
1266 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1267 struct nameidata *nd)
1269 struct inode * inode;
1270 struct btrfs_inode *bi = BTRFS_I(dir);
1271 struct btrfs_root *root = bi->root;
1272 struct btrfs_root *sub_root = root;
1273 struct btrfs_key location;
1276 if (dentry->d_name.len > BTRFS_NAME_LEN)
1277 return ERR_PTR(-ENAMETOOLONG);
1279 mutex_lock(&root->fs_info->fs_mutex);
1280 ret = btrfs_inode_by_name(dir, dentry, &location);
1281 mutex_unlock(&root->fs_info->fs_mutex);
1284 return ERR_PTR(ret);
1287 if (location.objectid) {
1288 ret = fixup_tree_root_location(root, &location, &sub_root,
1291 return ERR_PTR(ret);
1293 return ERR_PTR(-ENOENT);
1294 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1297 return ERR_PTR(-EACCES);
1298 if (inode->i_state & I_NEW) {
1299 /* the inode and parent dir are two different roots */
1300 if (sub_root != root) {
1302 sub_root->inode = inode;
1304 BTRFS_I(inode)->root = sub_root;
1305 memcpy(&BTRFS_I(inode)->location, &location,
1307 btrfs_read_locked_inode(inode);
1308 unlock_new_inode(inode);
1311 return d_splice_alias(inode, dentry);
1314 static unsigned char btrfs_filetype_table[] = {
1315 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1318 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1320 struct inode *inode = filp->f_dentry->d_inode;
1321 struct btrfs_root *root = BTRFS_I(inode)->root;
1322 struct btrfs_item *item;
1323 struct btrfs_dir_item *di;
1324 struct btrfs_key key;
1325 struct btrfs_key found_key;
1326 struct btrfs_path *path;
1329 struct extent_buffer *leaf;
1332 unsigned char d_type;
1337 int key_type = BTRFS_DIR_INDEX_KEY;
1342 /* FIXME, use a real flag for deciding about the key type */
1343 if (root->fs_info->tree_root == root)
1344 key_type = BTRFS_DIR_ITEM_KEY;
1346 /* special case for "." */
1347 if (filp->f_pos == 0) {
1348 over = filldir(dirent, ".", 1,
1356 mutex_lock(&root->fs_info->fs_mutex);
1357 key.objectid = inode->i_ino;
1358 path = btrfs_alloc_path();
1361 /* special case for .., just use the back ref */
1362 if (filp->f_pos == 1) {
1363 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1365 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1367 leaf = path->nodes[0];
1368 slot = path->slots[0];
1369 nritems = btrfs_header_nritems(leaf);
1370 if (slot >= nritems) {
1371 btrfs_release_path(root, path);
1372 goto read_dir_items;
1374 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1375 btrfs_release_path(root, path);
1376 if (found_key.objectid != key.objectid ||
1377 found_key.type != BTRFS_INODE_REF_KEY)
1378 goto read_dir_items;
1379 over = filldir(dirent, "..", 2,
1380 2, found_key.offset, DT_DIR);
1387 btrfs_set_key_type(&key, key_type);
1388 key.offset = filp->f_pos;
1390 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1395 leaf = path->nodes[0];
1396 nritems = btrfs_header_nritems(leaf);
1397 slot = path->slots[0];
1398 if (advance || slot >= nritems) {
1399 if (slot >= nritems -1) {
1400 ret = btrfs_next_leaf(root, path);
1403 leaf = path->nodes[0];
1404 nritems = btrfs_header_nritems(leaf);
1405 slot = path->slots[0];
1412 item = btrfs_item_nr(leaf, slot);
1413 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1415 if (found_key.objectid != key.objectid)
1417 if (btrfs_key_type(&found_key) != key_type)
1419 if (found_key.offset < filp->f_pos)
1422 filp->f_pos = found_key.offset;
1424 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1426 di_total = btrfs_item_size(leaf, item);
1427 while(di_cur < di_total) {
1428 struct btrfs_key location;
1430 name_len = btrfs_dir_name_len(leaf, di);
1431 if (name_len < 32) {
1432 name_ptr = tmp_name;
1434 name_ptr = kmalloc(name_len, GFP_NOFS);
1437 read_extent_buffer(leaf, name_ptr,
1438 (unsigned long)(di + 1), name_len);
1440 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1441 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1442 over = filldir(dirent, name_ptr, name_len,
1447 if (name_ptr != tmp_name)
1452 di_len = btrfs_dir_name_len(leaf, di) +
1453 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1455 di = (struct btrfs_dir_item *)((char *)di + di_len);
1458 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1462 btrfs_release_path(root, path);
1463 btrfs_free_path(path);
1464 mutex_unlock(&root->fs_info->fs_mutex);
1468 int btrfs_write_inode(struct inode *inode, int wait)
1470 struct btrfs_root *root = BTRFS_I(inode)->root;
1471 struct btrfs_trans_handle *trans;
1475 mutex_lock(&root->fs_info->fs_mutex);
1476 trans = btrfs_start_transaction(root, 1);
1477 btrfs_set_trans_block_group(trans, inode);
1478 ret = btrfs_commit_transaction(trans, root);
1479 mutex_unlock(&root->fs_info->fs_mutex);
1485 * This is somewhat expensive, updating the tree every time the
1486 * inode changes. But, it is most likely to find the inode in cache.
1487 * FIXME, needs more benchmarking...there are no reasons other than performance
1488 * to keep or drop this code.
1490 void btrfs_dirty_inode(struct inode *inode)
1492 struct btrfs_root *root = BTRFS_I(inode)->root;
1493 struct btrfs_trans_handle *trans;
1495 mutex_lock(&root->fs_info->fs_mutex);
1496 trans = btrfs_start_transaction(root, 1);
1497 btrfs_set_trans_block_group(trans, inode);
1498 btrfs_update_inode(trans, root, inode);
1499 btrfs_end_transaction(trans, root);
1500 mutex_unlock(&root->fs_info->fs_mutex);
1503 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1504 struct btrfs_root *root,
1505 const char *name, int name_len,
1508 struct btrfs_block_group_cache *group,
1511 struct inode *inode;
1512 struct btrfs_inode_item *inode_item;
1513 struct btrfs_key *location;
1514 struct btrfs_path *path;
1515 struct btrfs_inode_ref *ref;
1516 struct btrfs_key key[2];
1522 path = btrfs_alloc_path();
1525 inode = new_inode(root->fs_info->sb);
1527 return ERR_PTR(-ENOMEM);
1529 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1530 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1531 inode->i_mapping, GFP_NOFS);
1532 BTRFS_I(inode)->root = root;
1538 group = btrfs_find_block_group(root, group, 0, 0, owner);
1539 BTRFS_I(inode)->block_group = group;
1540 BTRFS_I(inode)->flags = 0;
1542 key[0].objectid = objectid;
1543 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1546 key[1].objectid = objectid;
1547 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1548 key[1].offset = ref_objectid;
1550 sizes[0] = sizeof(struct btrfs_inode_item);
1551 sizes[1] = name_len + sizeof(*ref);
1553 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1557 if (objectid > root->highest_inode)
1558 root->highest_inode = objectid;
1560 inode->i_uid = current->fsuid;
1561 inode->i_gid = current->fsgid;
1562 inode->i_mode = mode;
1563 inode->i_ino = objectid;
1564 inode->i_blocks = 0;
1565 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1566 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1567 struct btrfs_inode_item);
1568 fill_inode_item(path->nodes[0], inode_item, inode);
1570 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1571 struct btrfs_inode_ref);
1572 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1573 ptr = (unsigned long)(ref + 1);
1574 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1576 btrfs_mark_buffer_dirty(path->nodes[0]);
1577 btrfs_free_path(path);
1579 location = &BTRFS_I(inode)->location;
1580 location->objectid = objectid;
1581 location->offset = 0;
1582 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1584 insert_inode_hash(inode);
1587 btrfs_free_path(path);
1588 return ERR_PTR(ret);
1591 static inline u8 btrfs_inode_type(struct inode *inode)
1593 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1596 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1597 struct dentry *dentry, struct inode *inode,
1601 struct btrfs_key key;
1602 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1603 struct inode *parent_inode;
1605 key.objectid = inode->i_ino;
1606 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1609 ret = btrfs_insert_dir_item(trans, root,
1610 dentry->d_name.name, dentry->d_name.len,
1611 dentry->d_parent->d_inode->i_ino,
1612 &key, btrfs_inode_type(inode));
1615 ret = btrfs_insert_inode_ref(trans, root,
1616 dentry->d_name.name,
1619 dentry->d_parent->d_inode->i_ino);
1621 parent_inode = dentry->d_parent->d_inode;
1622 parent_inode->i_size += dentry->d_name.len * 2;
1623 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1624 ret = btrfs_update_inode(trans, root,
1625 dentry->d_parent->d_inode);
1630 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1631 struct dentry *dentry, struct inode *inode,
1634 int err = btrfs_add_link(trans, dentry, inode, backref);
1636 d_instantiate(dentry, inode);
1644 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1645 int mode, dev_t rdev)
1647 struct btrfs_trans_handle *trans;
1648 struct btrfs_root *root = BTRFS_I(dir)->root;
1649 struct inode *inode = NULL;
1653 unsigned long nr = 0;
1655 if (!new_valid_dev(rdev))
1658 mutex_lock(&root->fs_info->fs_mutex);
1659 err = btrfs_check_free_space(root, 1, 0);
1663 trans = btrfs_start_transaction(root, 1);
1664 btrfs_set_trans_block_group(trans, dir);
1666 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1672 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1674 dentry->d_parent->d_inode->i_ino, objectid,
1675 BTRFS_I(dir)->block_group, mode);
1676 err = PTR_ERR(inode);
1680 btrfs_set_trans_block_group(trans, inode);
1681 err = btrfs_add_nondir(trans, dentry, inode, 0);
1685 inode->i_op = &btrfs_special_inode_operations;
1686 init_special_inode(inode, inode->i_mode, rdev);
1687 btrfs_update_inode(trans, root, inode);
1689 dir->i_sb->s_dirt = 1;
1690 btrfs_update_inode_block_group(trans, inode);
1691 btrfs_update_inode_block_group(trans, dir);
1693 nr = trans->blocks_used;
1694 btrfs_end_transaction(trans, root);
1696 mutex_unlock(&root->fs_info->fs_mutex);
1699 inode_dec_link_count(inode);
1702 btrfs_btree_balance_dirty(root, nr);
1703 btrfs_throttle(root);
1707 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1708 int mode, struct nameidata *nd)
1710 struct btrfs_trans_handle *trans;
1711 struct btrfs_root *root = BTRFS_I(dir)->root;
1712 struct inode *inode = NULL;
1715 unsigned long nr = 0;
1718 mutex_lock(&root->fs_info->fs_mutex);
1719 err = btrfs_check_free_space(root, 1, 0);
1722 trans = btrfs_start_transaction(root, 1);
1723 btrfs_set_trans_block_group(trans, dir);
1725 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1731 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1733 dentry->d_parent->d_inode->i_ino,
1734 objectid, BTRFS_I(dir)->block_group, mode);
1735 err = PTR_ERR(inode);
1739 btrfs_set_trans_block_group(trans, inode);
1740 err = btrfs_add_nondir(trans, dentry, inode, 0);
1744 inode->i_mapping->a_ops = &btrfs_aops;
1745 inode->i_fop = &btrfs_file_operations;
1746 inode->i_op = &btrfs_file_inode_operations;
1747 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1748 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1749 inode->i_mapping, GFP_NOFS);
1750 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
1752 dir->i_sb->s_dirt = 1;
1753 btrfs_update_inode_block_group(trans, inode);
1754 btrfs_update_inode_block_group(trans, dir);
1756 nr = trans->blocks_used;
1757 btrfs_end_transaction(trans, root);
1759 mutex_unlock(&root->fs_info->fs_mutex);
1762 inode_dec_link_count(inode);
1765 btrfs_btree_balance_dirty(root, nr);
1766 btrfs_throttle(root);
1770 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1771 struct dentry *dentry)
1773 struct btrfs_trans_handle *trans;
1774 struct btrfs_root *root = BTRFS_I(dir)->root;
1775 struct inode *inode = old_dentry->d_inode;
1776 unsigned long nr = 0;
1780 if (inode->i_nlink == 0)
1783 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1788 mutex_lock(&root->fs_info->fs_mutex);
1789 err = btrfs_check_free_space(root, 1, 0);
1792 trans = btrfs_start_transaction(root, 1);
1794 btrfs_set_trans_block_group(trans, dir);
1795 atomic_inc(&inode->i_count);
1796 err = btrfs_add_nondir(trans, dentry, inode, 1);
1801 dir->i_sb->s_dirt = 1;
1802 btrfs_update_inode_block_group(trans, dir);
1803 err = btrfs_update_inode(trans, root, inode);
1808 nr = trans->blocks_used;
1809 btrfs_end_transaction(trans, root);
1811 mutex_unlock(&root->fs_info->fs_mutex);
1814 inode_dec_link_count(inode);
1817 btrfs_btree_balance_dirty(root, nr);
1818 btrfs_throttle(root);
1822 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1824 struct inode *inode;
1825 struct btrfs_trans_handle *trans;
1826 struct btrfs_root *root = BTRFS_I(dir)->root;
1828 int drop_on_err = 0;
1830 unsigned long nr = 1;
1832 mutex_lock(&root->fs_info->fs_mutex);
1833 err = btrfs_check_free_space(root, 1, 0);
1837 trans = btrfs_start_transaction(root, 1);
1838 btrfs_set_trans_block_group(trans, dir);
1840 if (IS_ERR(trans)) {
1841 err = PTR_ERR(trans);
1845 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1851 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1853 dentry->d_parent->d_inode->i_ino, objectid,
1854 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1855 if (IS_ERR(inode)) {
1856 err = PTR_ERR(inode);
1861 inode->i_op = &btrfs_dir_inode_operations;
1862 inode->i_fop = &btrfs_dir_file_operations;
1863 btrfs_set_trans_block_group(trans, inode);
1866 err = btrfs_update_inode(trans, root, inode);
1870 err = btrfs_add_link(trans, dentry, inode, 0);
1874 d_instantiate(dentry, inode);
1876 dir->i_sb->s_dirt = 1;
1877 btrfs_update_inode_block_group(trans, inode);
1878 btrfs_update_inode_block_group(trans, dir);
1881 nr = trans->blocks_used;
1882 btrfs_end_transaction(trans, root);
1885 mutex_unlock(&root->fs_info->fs_mutex);
1888 btrfs_btree_balance_dirty(root, nr);
1889 btrfs_throttle(root);
1893 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1894 size_t pg_offset, u64 start, u64 len,
1900 u64 extent_start = 0;
1902 u64 objectid = inode->i_ino;
1904 struct btrfs_path *path;
1905 struct btrfs_root *root = BTRFS_I(inode)->root;
1906 struct btrfs_file_extent_item *item;
1907 struct extent_buffer *leaf;
1908 struct btrfs_key found_key;
1909 struct extent_map *em = NULL;
1910 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1911 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1912 struct btrfs_trans_handle *trans = NULL;
1914 path = btrfs_alloc_path();
1916 mutex_lock(&root->fs_info->fs_mutex);
1919 spin_lock(&em_tree->lock);
1920 em = lookup_extent_mapping(em_tree, start, len);
1921 spin_unlock(&em_tree->lock);
1924 if (em->start > start) {
1925 printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n",
1926 start, len, em->start, em->len);
1929 if (em->block_start == EXTENT_MAP_INLINE && page)
1930 free_extent_map(em);
1934 em = alloc_extent_map(GFP_NOFS);
1940 em->start = EXTENT_MAP_HOLE;
1942 em->bdev = inode->i_sb->s_bdev;
1943 ret = btrfs_lookup_file_extent(trans, root, path,
1944 objectid, start, trans != NULL);
1951 if (path->slots[0] == 0)
1956 leaf = path->nodes[0];
1957 item = btrfs_item_ptr(leaf, path->slots[0],
1958 struct btrfs_file_extent_item);
1959 /* are we inside the extent that was found? */
1960 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1961 found_type = btrfs_key_type(&found_key);
1962 if (found_key.objectid != objectid ||
1963 found_type != BTRFS_EXTENT_DATA_KEY) {
1967 found_type = btrfs_file_extent_type(leaf, item);
1968 extent_start = found_key.offset;
1969 if (found_type == BTRFS_FILE_EXTENT_REG) {
1970 extent_end = extent_start +
1971 btrfs_file_extent_num_bytes(leaf, item);
1973 if (start < extent_start || start >= extent_end) {
1975 if (start < extent_start) {
1976 if (start + len <= extent_start)
1978 em->len = extent_end - extent_start;
1984 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1986 em->start = extent_start;
1987 em->len = extent_end - extent_start;
1988 em->block_start = EXTENT_MAP_HOLE;
1991 bytenr += btrfs_file_extent_offset(leaf, item);
1992 em->block_start = bytenr;
1993 em->start = extent_start;
1994 em->len = extent_end - extent_start;
1996 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2001 size_t extent_offset;
2004 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2006 extent_end = (extent_start + size + root->sectorsize - 1) &
2007 ~((u64)root->sectorsize - 1);
2008 if (start < extent_start || start >= extent_end) {
2010 if (start < extent_start) {
2011 if (start + len <= extent_start)
2013 em->len = extent_end - extent_start;
2019 em->block_start = EXTENT_MAP_INLINE;
2022 em->start = extent_start;
2027 page_start = page_offset(page) + pg_offset;
2028 extent_offset = page_start - extent_start;
2029 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2030 size - extent_offset);
2031 em->start = extent_start + extent_offset;
2032 em->len = (copy_size + root->sectorsize - 1) &
2033 ~((u64)root->sectorsize - 1);
2035 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2036 if (create == 0 && !PageUptodate(page)) {
2037 read_extent_buffer(leaf, map + pg_offset, ptr,
2039 flush_dcache_page(page);
2040 } else if (create && PageUptodate(page)) {
2043 free_extent_map(em);
2045 btrfs_release_path(root, path);
2046 trans = btrfs_start_transaction(root, 1);
2049 write_extent_buffer(leaf, map + pg_offset, ptr,
2051 btrfs_mark_buffer_dirty(leaf);
2054 set_extent_uptodate(io_tree, em->start,
2055 extent_map_end(em) - 1, GFP_NOFS);
2058 printk("unkknown found_type %d\n", found_type);
2065 em->block_start = EXTENT_MAP_HOLE;
2067 btrfs_release_path(root, path);
2068 if (em->start > start || extent_map_end(em) <= start) {
2069 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2075 spin_lock(&em_tree->lock);
2076 ret = add_extent_mapping(em_tree, em);
2077 if (ret == -EEXIST) {
2078 free_extent_map(em);
2079 em = lookup_extent_mapping(em_tree, start, len);
2082 printk("failing to insert %Lu %Lu\n", start, len);
2085 spin_unlock(&em_tree->lock);
2087 btrfs_free_path(path);
2089 ret = btrfs_end_transaction(trans, root);
2093 mutex_unlock(&root->fs_info->fs_mutex);
2095 free_extent_map(em);
2097 return ERR_PTR(err);
2102 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2104 return extent_bmap(mapping, iblock, btrfs_get_extent);
2107 int btrfs_readpage(struct file *file, struct page *page)
2109 struct extent_io_tree *tree;
2110 tree = &BTRFS_I(page->mapping->host)->io_tree;
2111 return extent_read_full_page(tree, page, btrfs_get_extent);
2114 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2116 struct extent_io_tree *tree;
2119 if (current->flags & PF_MEMALLOC) {
2120 redirty_page_for_writepage(wbc, page);
2124 tree = &BTRFS_I(page->mapping->host)->io_tree;
2125 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2128 static int btrfs_writepages(struct address_space *mapping,
2129 struct writeback_control *wbc)
2131 struct extent_io_tree *tree;
2132 tree = &BTRFS_I(mapping->host)->io_tree;
2133 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2137 btrfs_readpages(struct file *file, struct address_space *mapping,
2138 struct list_head *pages, unsigned nr_pages)
2140 struct extent_io_tree *tree;
2141 tree = &BTRFS_I(mapping->host)->io_tree;
2142 return extent_readpages(tree, mapping, pages, nr_pages,
2146 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2148 struct extent_io_tree *tree;
2149 struct extent_map_tree *map;
2152 tree = &BTRFS_I(page->mapping->host)->io_tree;
2153 map = &BTRFS_I(page->mapping->host)->extent_tree;
2154 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2156 ClearPagePrivate(page);
2157 set_page_private(page, 0);
2158 page_cache_release(page);
2163 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2165 struct extent_io_tree *tree;
2167 tree = &BTRFS_I(page->mapping->host)->io_tree;
2168 extent_invalidatepage(tree, page, offset);
2169 btrfs_releasepage(page, GFP_NOFS);
2173 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2174 * called from a page fault handler when a page is first dirtied. Hence we must
2175 * be careful to check for EOF conditions here. We set the page up correctly
2176 * for a written page which means we get ENOSPC checking when writing into
2177 * holes and correct delalloc and unwritten extent mapping on filesystems that
2178 * support these features.
2180 * We are not allowed to take the i_mutex here so we have to play games to
2181 * protect against truncate races as the page could now be beyond EOF. Because
2182 * vmtruncate() writes the inode size before removing pages, once we have the
2183 * page lock we can determine safely if the page is beyond EOF. If it is not
2184 * beyond EOF, then the page is guaranteed safe against truncation until we
2187 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2189 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2190 struct btrfs_root *root = BTRFS_I(inode)->root;
2196 mutex_lock(&root->fs_info->fs_mutex);
2197 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2198 mutex_unlock(&root->fs_info->fs_mutex);
2205 wait_on_page_writeback(page);
2206 size = i_size_read(inode);
2207 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2209 if ((page->mapping != inode->i_mapping) ||
2210 (page_start > size)) {
2211 /* page got truncated out from underneath us */
2215 /* page is wholly or partially inside EOF */
2216 if (page_start + PAGE_CACHE_SIZE > size)
2217 end = size & ~PAGE_CACHE_MASK;
2219 end = PAGE_CACHE_SIZE;
2221 ret = btrfs_cow_one_page(inode, page, end);
2229 static void btrfs_truncate(struct inode *inode)
2231 struct btrfs_root *root = BTRFS_I(inode)->root;
2233 struct btrfs_trans_handle *trans;
2236 if (!S_ISREG(inode->i_mode))
2238 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2241 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2243 mutex_lock(&root->fs_info->fs_mutex);
2244 trans = btrfs_start_transaction(root, 1);
2245 btrfs_set_trans_block_group(trans, inode);
2247 /* FIXME, add redo link to tree so we don't leak on crash */
2248 ret = btrfs_truncate_in_trans(trans, root, inode,
2249 BTRFS_EXTENT_DATA_KEY);
2250 btrfs_update_inode(trans, root, inode);
2251 nr = trans->blocks_used;
2253 ret = btrfs_end_transaction(trans, root);
2255 mutex_unlock(&root->fs_info->fs_mutex);
2256 btrfs_btree_balance_dirty(root, nr);
2257 btrfs_throttle(root);
2260 static int noinline create_subvol(struct btrfs_root *root, char *name,
2263 struct btrfs_trans_handle *trans;
2264 struct btrfs_key key;
2265 struct btrfs_root_item root_item;
2266 struct btrfs_inode_item *inode_item;
2267 struct extent_buffer *leaf;
2268 struct btrfs_root *new_root = root;
2269 struct inode *inode;
2274 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
2275 unsigned long nr = 1;
2277 mutex_lock(&root->fs_info->fs_mutex);
2278 ret = btrfs_check_free_space(root, 1, 0);
2282 trans = btrfs_start_transaction(root, 1);
2285 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2290 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
2291 objectid, trans->transid, 0, 0,
2294 return PTR_ERR(leaf);
2296 btrfs_set_header_nritems(leaf, 0);
2297 btrfs_set_header_level(leaf, 0);
2298 btrfs_set_header_bytenr(leaf, leaf->start);
2299 btrfs_set_header_generation(leaf, trans->transid);
2300 btrfs_set_header_owner(leaf, objectid);
2302 write_extent_buffer(leaf, root->fs_info->fsid,
2303 (unsigned long)btrfs_header_fsid(leaf),
2305 btrfs_mark_buffer_dirty(leaf);
2307 inode_item = &root_item.inode;
2308 memset(inode_item, 0, sizeof(*inode_item));
2309 inode_item->generation = cpu_to_le64(1);
2310 inode_item->size = cpu_to_le64(3);
2311 inode_item->nlink = cpu_to_le32(1);
2312 inode_item->nblocks = cpu_to_le64(1);
2313 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
2315 btrfs_set_root_bytenr(&root_item, leaf->start);
2316 btrfs_set_root_level(&root_item, 0);
2317 btrfs_set_root_refs(&root_item, 1);
2318 btrfs_set_root_used(&root_item, 0);
2320 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
2321 root_item.drop_level = 0;
2323 free_extent_buffer(leaf);
2326 btrfs_set_root_dirid(&root_item, new_dirid);
2328 key.objectid = objectid;
2330 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2331 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2337 * insert the directory item
2339 key.offset = (u64)-1;
2340 dir = root->fs_info->sb->s_root->d_inode;
2341 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2342 name, namelen, dir->i_ino, &key,
2347 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2348 name, namelen, objectid,
2349 root->fs_info->sb->s_root->d_inode->i_ino);
2353 ret = btrfs_commit_transaction(trans, root);
2357 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
2360 trans = btrfs_start_transaction(new_root, 1);
2363 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2365 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2368 inode->i_op = &btrfs_dir_inode_operations;
2369 inode->i_fop = &btrfs_dir_file_operations;
2370 new_root->inode = inode;
2372 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2376 ret = btrfs_update_inode(trans, new_root, inode);
2380 nr = trans->blocks_used;
2381 err = btrfs_commit_transaction(trans, new_root);
2385 mutex_unlock(&root->fs_info->fs_mutex);
2386 btrfs_btree_balance_dirty(root, nr);
2387 btrfs_throttle(root);
2391 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2393 struct btrfs_pending_snapshot *pending_snapshot;
2394 struct btrfs_trans_handle *trans;
2397 unsigned long nr = 0;
2399 if (!root->ref_cows)
2402 mutex_lock(&root->fs_info->fs_mutex);
2403 ret = btrfs_check_free_space(root, 1, 0);
2407 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
2408 if (!pending_snapshot) {
2412 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
2413 if (!pending_snapshot->name) {
2415 kfree(pending_snapshot);
2418 memcpy(pending_snapshot->name, name, namelen);
2419 pending_snapshot->name[namelen] = '\0';
2420 trans = btrfs_start_transaction(root, 1);
2422 pending_snapshot->root = root;
2423 list_add(&pending_snapshot->list,
2424 &trans->transaction->pending_snapshots);
2425 ret = btrfs_update_inode(trans, root, root->inode);
2426 err = btrfs_commit_transaction(trans, root);
2429 mutex_unlock(&root->fs_info->fs_mutex);
2430 btrfs_btree_balance_dirty(root, nr);
2431 btrfs_throttle(root);
2435 unsigned long btrfs_force_ra(struct address_space *mapping,
2436 struct file_ra_state *ra, struct file *file,
2437 pgoff_t offset, pgoff_t last_index)
2441 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2442 req_size = last_index - offset + 1;
2443 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2446 req_size = min(last_index - offset + 1, (pgoff_t)128);
2447 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2448 return offset + req_size;
2452 int btrfs_defrag_file(struct file *file) {
2453 struct inode *inode = fdentry(file)->d_inode;
2454 struct btrfs_root *root = BTRFS_I(inode)->root;
2455 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2457 unsigned long last_index;
2458 unsigned long ra_index = 0;
2464 mutex_lock(&root->fs_info->fs_mutex);
2465 ret = btrfs_check_free_space(root, inode->i_size, 0);
2466 mutex_unlock(&root->fs_info->fs_mutex);
2470 mutex_lock(&inode->i_mutex);
2471 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2472 for (i = 0; i <= last_index; i++) {
2473 if (i == ra_index) {
2474 ra_index = btrfs_force_ra(inode->i_mapping,
2476 file, ra_index, last_index);
2478 page = grab_cache_page(inode->i_mapping, i);
2481 if (!PageUptodate(page)) {
2482 btrfs_readpage(NULL, page);
2484 if (!PageUptodate(page)) {
2486 page_cache_release(page);
2490 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2491 page_end = page_start + PAGE_CACHE_SIZE - 1;
2493 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2494 set_extent_delalloc(io_tree, page_start,
2495 page_end, GFP_NOFS);
2497 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2498 set_page_dirty(page);
2500 page_cache_release(page);
2501 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2505 mutex_unlock(&inode->i_mutex);
2509 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
2513 struct btrfs_ioctl_vol_args *vol_args;
2514 struct btrfs_trans_handle *trans;
2520 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2525 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2529 namelen = strlen(vol_args->name);
2530 if (namelen > BTRFS_VOL_NAME_MAX) {
2535 sizestr = vol_args->name;
2536 if (!strcmp(sizestr, "max"))
2537 new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
2539 if (sizestr[0] == '-') {
2542 } else if (sizestr[0] == '+') {
2546 new_size = btrfs_parse_size(sizestr);
2547 if (new_size == 0) {
2553 mutex_lock(&root->fs_info->fs_mutex);
2554 old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
2557 if (new_size > old_size) {
2561 new_size = old_size - new_size;
2562 } else if (mod > 0) {
2563 new_size = old_size + new_size;
2566 if (new_size < 256 * 1024 * 1024) {
2570 if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
2575 do_div(new_size, root->sectorsize);
2576 new_size *= root->sectorsize;
2578 printk("new size is %Lu\n", new_size);
2579 if (new_size > old_size) {
2580 trans = btrfs_start_transaction(root, 1);
2581 ret = btrfs_grow_extent_tree(trans, root, new_size);
2582 btrfs_commit_transaction(trans, root);
2584 ret = btrfs_shrink_extent_tree(root, new_size);
2588 mutex_unlock(&root->fs_info->fs_mutex);
2594 static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root,
2597 struct btrfs_ioctl_vol_args *vol_args;
2598 struct btrfs_dir_item *di;
2599 struct btrfs_path *path;
2604 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2609 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2614 namelen = strlen(vol_args->name);
2615 if (namelen > BTRFS_VOL_NAME_MAX) {
2619 if (strchr(vol_args->name, '/')) {
2624 path = btrfs_alloc_path();
2630 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2631 mutex_lock(&root->fs_info->fs_mutex);
2632 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2634 vol_args->name, namelen, 0);
2635 mutex_unlock(&root->fs_info->fs_mutex);
2636 btrfs_free_path(path);
2638 if (di && !IS_ERR(di)) {
2648 if (root == root->fs_info->tree_root)
2649 ret = create_subvol(root, vol_args->name, namelen);
2651 ret = create_snapshot(root, vol_args->name, namelen);
2657 static int btrfs_ioctl_defrag(struct file *file)
2659 struct inode *inode = fdentry(file)->d_inode;
2660 struct btrfs_root *root = BTRFS_I(inode)->root;
2662 switch (inode->i_mode & S_IFMT) {
2664 mutex_lock(&root->fs_info->fs_mutex);
2665 btrfs_defrag_root(root, 0);
2666 btrfs_defrag_root(root->fs_info->extent_root, 0);
2667 mutex_unlock(&root->fs_info->fs_mutex);
2670 btrfs_defrag_file(file);
2677 long btrfs_ioctl(struct file *file, unsigned int
2678 cmd, unsigned long arg)
2680 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2683 case BTRFS_IOC_SNAP_CREATE:
2684 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2685 case BTRFS_IOC_DEFRAG:
2686 return btrfs_ioctl_defrag(file);
2687 case BTRFS_IOC_RESIZE:
2688 return btrfs_ioctl_resize(root, (void __user *)arg);
2695 * Called inside transaction, so use GFP_NOFS
2697 struct inode *btrfs_alloc_inode(struct super_block *sb)
2699 struct btrfs_inode *ei;
2701 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2705 ei->ordered_trans = 0;
2706 return &ei->vfs_inode;
2709 void btrfs_destroy_inode(struct inode *inode)
2711 WARN_ON(!list_empty(&inode->i_dentry));
2712 WARN_ON(inode->i_data.nrpages);
2714 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2715 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2718 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2719 static void init_once(struct kmem_cache * cachep, void *foo)
2721 static void init_once(void * foo, struct kmem_cache * cachep,
2722 unsigned long flags)
2725 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2727 inode_init_once(&ei->vfs_inode);
2730 void btrfs_destroy_cachep(void)
2732 if (btrfs_inode_cachep)
2733 kmem_cache_destroy(btrfs_inode_cachep);
2734 if (btrfs_trans_handle_cachep)
2735 kmem_cache_destroy(btrfs_trans_handle_cachep);
2736 if (btrfs_transaction_cachep)
2737 kmem_cache_destroy(btrfs_transaction_cachep);
2738 if (btrfs_bit_radix_cachep)
2739 kmem_cache_destroy(btrfs_bit_radix_cachep);
2740 if (btrfs_path_cachep)
2741 kmem_cache_destroy(btrfs_path_cachep);
2744 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2745 unsigned long extra_flags,
2746 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2747 void (*ctor)(struct kmem_cache *, void *)
2749 void (*ctor)(void *, struct kmem_cache *,
2754 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2755 SLAB_MEM_SPREAD | extra_flags), ctor
2756 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2762 int btrfs_init_cachep(void)
2764 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2765 sizeof(struct btrfs_inode),
2767 if (!btrfs_inode_cachep)
2769 btrfs_trans_handle_cachep =
2770 btrfs_cache_create("btrfs_trans_handle_cache",
2771 sizeof(struct btrfs_trans_handle),
2773 if (!btrfs_trans_handle_cachep)
2775 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2776 sizeof(struct btrfs_transaction),
2778 if (!btrfs_transaction_cachep)
2780 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2781 sizeof(struct btrfs_path),
2783 if (!btrfs_path_cachep)
2785 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2786 SLAB_DESTROY_BY_RCU, NULL);
2787 if (!btrfs_bit_radix_cachep)
2791 btrfs_destroy_cachep();
2795 static int btrfs_getattr(struct vfsmount *mnt,
2796 struct dentry *dentry, struct kstat *stat)
2798 struct inode *inode = dentry->d_inode;
2799 generic_fillattr(inode, stat);
2800 stat->blksize = PAGE_CACHE_SIZE;
2804 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2805 struct inode * new_dir,struct dentry *new_dentry)
2807 struct btrfs_trans_handle *trans;
2808 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2809 struct inode *new_inode = new_dentry->d_inode;
2810 struct inode *old_inode = old_dentry->d_inode;
2811 struct timespec ctime = CURRENT_TIME;
2812 struct btrfs_path *path;
2815 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2816 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2820 mutex_lock(&root->fs_info->fs_mutex);
2821 ret = btrfs_check_free_space(root, 1, 0);
2825 trans = btrfs_start_transaction(root, 1);
2827 btrfs_set_trans_block_group(trans, new_dir);
2828 path = btrfs_alloc_path();
2834 old_dentry->d_inode->i_nlink++;
2835 old_dir->i_ctime = old_dir->i_mtime = ctime;
2836 new_dir->i_ctime = new_dir->i_mtime = ctime;
2837 old_inode->i_ctime = ctime;
2839 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2844 new_inode->i_ctime = CURRENT_TIME;
2845 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2849 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
2854 btrfs_free_path(path);
2855 btrfs_end_transaction(trans, root);
2857 mutex_unlock(&root->fs_info->fs_mutex);
2861 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2862 const char *symname)
2864 struct btrfs_trans_handle *trans;
2865 struct btrfs_root *root = BTRFS_I(dir)->root;
2866 struct btrfs_path *path;
2867 struct btrfs_key key;
2868 struct inode *inode = NULL;
2875 struct btrfs_file_extent_item *ei;
2876 struct extent_buffer *leaf;
2877 unsigned long nr = 0;
2879 name_len = strlen(symname) + 1;
2880 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2881 return -ENAMETOOLONG;
2883 mutex_lock(&root->fs_info->fs_mutex);
2884 err = btrfs_check_free_space(root, 1, 0);
2888 trans = btrfs_start_transaction(root, 1);
2889 btrfs_set_trans_block_group(trans, dir);
2891 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2897 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2899 dentry->d_parent->d_inode->i_ino, objectid,
2900 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2901 err = PTR_ERR(inode);
2905 btrfs_set_trans_block_group(trans, inode);
2906 err = btrfs_add_nondir(trans, dentry, inode, 0);
2910 inode->i_mapping->a_ops = &btrfs_aops;
2911 inode->i_fop = &btrfs_file_operations;
2912 inode->i_op = &btrfs_file_inode_operations;
2913 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2914 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2915 inode->i_mapping, GFP_NOFS);
2916 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2918 dir->i_sb->s_dirt = 1;
2919 btrfs_update_inode_block_group(trans, inode);
2920 btrfs_update_inode_block_group(trans, dir);
2924 path = btrfs_alloc_path();
2926 key.objectid = inode->i_ino;
2928 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2929 datasize = btrfs_file_extent_calc_inline_size(name_len);
2930 err = btrfs_insert_empty_item(trans, root, path, &key,
2936 leaf = path->nodes[0];
2937 ei = btrfs_item_ptr(leaf, path->slots[0],
2938 struct btrfs_file_extent_item);
2939 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2940 btrfs_set_file_extent_type(leaf, ei,
2941 BTRFS_FILE_EXTENT_INLINE);
2942 ptr = btrfs_file_extent_inline_start(ei);
2943 write_extent_buffer(leaf, symname, ptr, name_len);
2944 btrfs_mark_buffer_dirty(leaf);
2945 btrfs_free_path(path);
2947 inode->i_op = &btrfs_symlink_inode_operations;
2948 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2949 inode->i_size = name_len - 1;
2950 err = btrfs_update_inode(trans, root, inode);
2955 nr = trans->blocks_used;
2956 btrfs_end_transaction(trans, root);
2958 mutex_unlock(&root->fs_info->fs_mutex);
2960 inode_dec_link_count(inode);
2963 btrfs_btree_balance_dirty(root, nr);
2964 btrfs_throttle(root);
2967 static int btrfs_permission(struct inode *inode, int mask,
2968 struct nameidata *nd)
2970 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
2972 return generic_permission(inode, mask, NULL);
2975 static struct inode_operations btrfs_dir_inode_operations = {
2976 .lookup = btrfs_lookup,
2977 .create = btrfs_create,
2978 .unlink = btrfs_unlink,
2980 .mkdir = btrfs_mkdir,
2981 .rmdir = btrfs_rmdir,
2982 .rename = btrfs_rename,
2983 .symlink = btrfs_symlink,
2984 .setattr = btrfs_setattr,
2985 .mknod = btrfs_mknod,
2986 .setxattr = generic_setxattr,
2987 .getxattr = generic_getxattr,
2988 .listxattr = btrfs_listxattr,
2989 .removexattr = generic_removexattr,
2990 .permission = btrfs_permission,
2992 static struct inode_operations btrfs_dir_ro_inode_operations = {
2993 .lookup = btrfs_lookup,
2994 .permission = btrfs_permission,
2996 static struct file_operations btrfs_dir_file_operations = {
2997 .llseek = generic_file_llseek,
2998 .read = generic_read_dir,
2999 .readdir = btrfs_readdir,
3000 .unlocked_ioctl = btrfs_ioctl,
3001 #ifdef CONFIG_COMPAT
3002 .compat_ioctl = btrfs_ioctl,
3006 static struct extent_io_ops btrfs_extent_io_ops = {
3007 .fill_delalloc = run_delalloc_range,
3008 .writepage_io_hook = btrfs_writepage_io_hook,
3009 .readpage_io_hook = btrfs_readpage_io_hook,
3010 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3013 static struct address_space_operations btrfs_aops = {
3014 .readpage = btrfs_readpage,
3015 .writepage = btrfs_writepage,
3016 .writepages = btrfs_writepages,
3017 .readpages = btrfs_readpages,
3018 .sync_page = block_sync_page,
3020 .invalidatepage = btrfs_invalidatepage,
3021 .releasepage = btrfs_releasepage,
3022 .set_page_dirty = __set_page_dirty_nobuffers,
3025 static struct address_space_operations btrfs_symlink_aops = {
3026 .readpage = btrfs_readpage,
3027 .writepage = btrfs_writepage,
3028 .invalidatepage = btrfs_invalidatepage,
3029 .releasepage = btrfs_releasepage,
3032 static struct inode_operations btrfs_file_inode_operations = {
3033 .truncate = btrfs_truncate,
3034 .getattr = btrfs_getattr,
3035 .setattr = btrfs_setattr,
3036 .setxattr = generic_setxattr,
3037 .getxattr = generic_getxattr,
3038 .listxattr = btrfs_listxattr,
3039 .removexattr = generic_removexattr,
3040 .permission = btrfs_permission,
3042 static struct inode_operations btrfs_special_inode_operations = {
3043 .getattr = btrfs_getattr,
3044 .setattr = btrfs_setattr,
3045 .permission = btrfs_permission,
3047 static struct inode_operations btrfs_symlink_inode_operations = {
3048 .readlink = generic_readlink,
3049 .follow_link = page_follow_link_light,
3050 .put_link = page_put_link,
3051 .permission = btrfs_permission,