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/bio.h>
20 #include <linux/buffer_head.h>
22 #include <linux/pagemap.h>
23 #include <linux/highmem.h>
24 #include <linux/time.h>
25 #include <linux/init.h>
26 #include <linux/string.h>
27 #include <linux/smp_lock.h>
28 #include <linux/backing-dev.h>
29 #include <linux/mpage.h>
30 #include <linux/swap.h>
31 #include <linux/writeback.h>
32 #include <linux/statfs.h>
33 #include <linux/compat.h>
34 #include <linux/bit_spinlock.h>
35 #include <linux/version.h>
36 #include <linux/xattr.h>
39 #include "transaction.h"
40 #include "btrfs_inode.h"
42 #include "print-tree.h"
45 struct btrfs_iget_args {
47 struct btrfs_root *root;
50 static struct inode_operations btrfs_dir_inode_operations;
51 static struct inode_operations btrfs_symlink_inode_operations;
52 static struct inode_operations btrfs_dir_ro_inode_operations;
53 static struct inode_operations btrfs_special_inode_operations;
54 static struct inode_operations btrfs_file_inode_operations;
55 static struct address_space_operations btrfs_aops;
56 static struct address_space_operations btrfs_symlink_aops;
57 static struct file_operations btrfs_dir_file_operations;
58 static struct extent_io_ops btrfs_extent_io_ops;
60 static struct kmem_cache *btrfs_inode_cachep;
61 struct kmem_cache *btrfs_trans_handle_cachep;
62 struct kmem_cache *btrfs_transaction_cachep;
63 struct kmem_cache *btrfs_bit_radix_cachep;
64 struct kmem_cache *btrfs_path_cachep;
67 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
68 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
69 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
70 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
71 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
72 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
73 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
74 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
77 int btrfs_check_free_space(struct btrfs_root *root, u64 num_required,
80 u64 total = btrfs_super_total_bytes(&root->fs_info->super_copy);
81 u64 used = btrfs_super_bytes_used(&root->fs_info->super_copy);
92 spin_lock(&root->fs_info->delalloc_lock);
93 if (used + root->fs_info->delalloc_bytes + num_required > thresh)
95 spin_unlock(&root->fs_info->delalloc_lock);
99 static int cow_file_range(struct inode *inode, u64 start, u64 end)
101 struct btrfs_root *root = BTRFS_I(inode)->root;
102 struct btrfs_trans_handle *trans;
106 u64 blocksize = root->sectorsize;
107 u64 orig_start = start;
109 struct btrfs_key ins;
112 trans = btrfs_start_transaction(root, 1);
114 btrfs_set_trans_block_group(trans, inode);
116 num_bytes = (end - start + blocksize) & ~(blocksize - 1);
117 num_bytes = max(blocksize, num_bytes);
118 ret = btrfs_drop_extents(trans, root, inode,
119 start, start + num_bytes, start, &alloc_hint);
120 orig_num_bytes = num_bytes;
122 if (alloc_hint == EXTENT_MAP_INLINE)
125 while(num_bytes > 0) {
126 cur_alloc_size = min(num_bytes, root->fs_info->max_extent);
127 ret = btrfs_alloc_extent(trans, root, cur_alloc_size,
128 root->root_key.objectid,
130 inode->i_ino, start, 0,
131 alloc_hint, (u64)-1, &ins, 1);
136 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
137 start, ins.objectid, ins.offset,
139 inode->i_blocks += ins.offset >> 9;
140 btrfs_check_file(root, inode);
141 num_bytes -= cur_alloc_size;
142 alloc_hint = ins.objectid + ins.offset;
143 start += cur_alloc_size;
145 btrfs_drop_extent_cache(inode, orig_start,
146 orig_start + orig_num_bytes - 1);
147 btrfs_add_ordered_inode(inode);
148 btrfs_update_inode(trans, root, inode);
150 btrfs_end_transaction(trans, root);
154 static int run_delalloc_nocow(struct inode *inode, u64 start, u64 end)
162 struct btrfs_root *root = BTRFS_I(inode)->root;
163 struct extent_buffer *leaf;
165 struct btrfs_path *path;
166 struct btrfs_file_extent_item *item;
169 struct btrfs_key found_key;
171 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
172 path = btrfs_alloc_path();
175 ret = btrfs_lookup_file_extent(NULL, root, path,
176 inode->i_ino, start, 0);
178 btrfs_free_path(path);
184 if (path->slots[0] == 0)
189 leaf = path->nodes[0];
190 item = btrfs_item_ptr(leaf, path->slots[0],
191 struct btrfs_file_extent_item);
193 /* are we inside the extent that was found? */
194 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
195 found_type = btrfs_key_type(&found_key);
196 if (found_key.objectid != inode->i_ino ||
197 found_type != BTRFS_EXTENT_DATA_KEY) {
201 found_type = btrfs_file_extent_type(leaf, item);
202 extent_start = found_key.offset;
203 if (found_type == BTRFS_FILE_EXTENT_REG) {
204 u64 extent_num_bytes;
206 extent_num_bytes = btrfs_file_extent_num_bytes(leaf, item);
207 extent_end = extent_start + extent_num_bytes;
210 if (loops && start != extent_start)
213 if (start < extent_start || start >= extent_end)
216 cow_end = min(end, extent_end - 1);
217 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
222 * we may be called by the resizer, make sure we're inside
223 * the limits of the FS
225 if (bytenr + extent_num_bytes > total_fs_bytes)
228 if (btrfs_count_snapshots_in_path(root, path, bytenr) != 1) {
238 btrfs_free_path(path);
241 btrfs_release_path(root, path);
246 cow_file_range(inode, start, cow_end);
251 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
253 struct btrfs_root *root = BTRFS_I(inode)->root;
255 mutex_lock(&root->fs_info->fs_mutex);
256 if (btrfs_test_opt(root, NODATACOW) ||
257 btrfs_test_flag(inode, NODATACOW))
258 ret = run_delalloc_nocow(inode, start, end);
260 ret = cow_file_range(inode, start, end);
262 mutex_unlock(&root->fs_info->fs_mutex);
266 int btrfs_set_bit_hook(struct inode *inode, u64 start, u64 end,
267 unsigned long old, unsigned long bits)
269 if (!(old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
270 struct btrfs_root *root = BTRFS_I(inode)->root;
271 spin_lock(&root->fs_info->delalloc_lock);
272 BTRFS_I(inode)->delalloc_bytes += end - start + 1;
273 root->fs_info->delalloc_bytes += end - start + 1;
274 spin_unlock(&root->fs_info->delalloc_lock);
279 int btrfs_clear_bit_hook(struct inode *inode, u64 start, u64 end,
280 unsigned long old, unsigned long bits)
282 if ((old & EXTENT_DELALLOC) && (bits & EXTENT_DELALLOC)) {
283 struct btrfs_root *root = BTRFS_I(inode)->root;
284 spin_lock(&root->fs_info->delalloc_lock);
285 if (end - start + 1 > root->fs_info->delalloc_bytes) {
286 printk("warning: delalloc account %Lu %Lu\n",
287 end - start + 1, root->fs_info->delalloc_bytes);
288 root->fs_info->delalloc_bytes = 0;
289 BTRFS_I(inode)->delalloc_bytes = 0;
291 root->fs_info->delalloc_bytes -= end - start + 1;
292 BTRFS_I(inode)->delalloc_bytes -= end - start + 1;
294 spin_unlock(&root->fs_info->delalloc_lock);
299 int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
300 size_t size, struct bio *bio)
302 struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
303 struct btrfs_mapping_tree *map_tree;
304 u64 logical = bio->bi_sector << 9;
307 struct bio_vec *bvec;
311 bio_for_each_segment(bvec, bio, i) {
312 length += bvec->bv_len;
314 map_tree = &root->fs_info->mapping_tree;
316 ret = btrfs_map_block(map_tree, READ, logical,
317 &map_length, NULL, 0);
319 if (map_length < length + size) {
325 int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
328 struct btrfs_root *root = BTRFS_I(inode)->root;
329 struct btrfs_trans_handle *trans;
332 if (!(rw & (1 << BIO_RW))) {
333 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
338 if (btrfs_test_opt(root, NODATASUM) ||
339 btrfs_test_flag(inode, NODATASUM)) {
343 mutex_lock(&root->fs_info->fs_mutex);
344 trans = btrfs_start_transaction(root, 1);
345 btrfs_set_trans_block_group(trans, inode);
346 btrfs_csum_file_blocks(trans, root, inode, bio);
347 ret = btrfs_end_transaction(trans, root);
349 mutex_unlock(&root->fs_info->fs_mutex);
351 return btrfs_map_bio(root, rw, bio, mirror_num);
354 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
357 struct inode *inode = page->mapping->host;
358 struct btrfs_root *root = BTRFS_I(inode)->root;
359 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
360 struct btrfs_csum_item *item;
361 struct btrfs_path *path = NULL;
363 if (btrfs_test_opt(root, NODATASUM) ||
364 btrfs_test_flag(inode, NODATASUM))
366 mutex_lock(&root->fs_info->fs_mutex);
367 path = btrfs_alloc_path();
368 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
371 /* a csum that isn't present is a preallocated region. */
372 if (ret == -ENOENT || ret == -EFBIG)
375 printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start);
378 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
380 set_state_private(io_tree, start, csum);
383 btrfs_free_path(path);
384 mutex_unlock(&root->fs_info->fs_mutex);
388 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
389 struct extent_state *state)
391 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
392 struct inode *inode = page->mapping->host;
393 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
395 u64 private = ~(u32)0;
397 struct btrfs_root *root = BTRFS_I(inode)->root;
401 if (btrfs_test_opt(root, NODATASUM) ||
402 btrfs_test_flag(inode, NODATASUM))
404 if (state && state->start == start) {
405 private = state->private;
408 ret = get_state_private(io_tree, start, &private);
410 local_irq_save(flags);
411 kaddr = kmap_atomic(page, KM_IRQ0);
415 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
416 btrfs_csum_final(csum, (char *)&csum);
417 if (csum != private) {
420 kunmap_atomic(kaddr, KM_IRQ0);
421 local_irq_restore(flags);
425 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
426 page->mapping->host->i_ino, (unsigned long long)start, csum,
428 memset(kaddr + offset, 1, end - start + 1);
429 flush_dcache_page(page);
430 kunmap_atomic(kaddr, KM_IRQ0);
431 local_irq_restore(flags);
435 void btrfs_read_locked_inode(struct inode *inode)
437 struct btrfs_path *path;
438 struct extent_buffer *leaf;
439 struct btrfs_inode_item *inode_item;
440 struct btrfs_timespec *tspec;
441 struct btrfs_root *root = BTRFS_I(inode)->root;
442 struct btrfs_key location;
443 u64 alloc_group_block;
447 path = btrfs_alloc_path();
449 mutex_lock(&root->fs_info->fs_mutex);
450 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
452 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
456 leaf = path->nodes[0];
457 inode_item = btrfs_item_ptr(leaf, path->slots[0],
458 struct btrfs_inode_item);
460 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
461 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
462 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
463 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
464 inode->i_size = btrfs_inode_size(leaf, inode_item);
466 tspec = btrfs_inode_atime(inode_item);
467 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
468 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
470 tspec = btrfs_inode_mtime(inode_item);
471 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
472 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
474 tspec = btrfs_inode_ctime(inode_item);
475 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
476 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
478 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
479 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
481 rdev = btrfs_inode_rdev(leaf, inode_item);
483 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
484 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
486 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
487 if (!BTRFS_I(inode)->block_group) {
488 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
490 BTRFS_BLOCK_GROUP_METADATA, 0);
492 btrfs_free_path(path);
495 mutex_unlock(&root->fs_info->fs_mutex);
497 switch (inode->i_mode & S_IFMT) {
499 inode->i_mapping->a_ops = &btrfs_aops;
500 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
501 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
502 inode->i_fop = &btrfs_file_operations;
503 inode->i_op = &btrfs_file_inode_operations;
506 inode->i_fop = &btrfs_dir_file_operations;
507 if (root == root->fs_info->tree_root)
508 inode->i_op = &btrfs_dir_ro_inode_operations;
510 inode->i_op = &btrfs_dir_inode_operations;
513 inode->i_op = &btrfs_symlink_inode_operations;
514 inode->i_mapping->a_ops = &btrfs_symlink_aops;
515 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
518 init_special_inode(inode, inode->i_mode, rdev);
524 btrfs_release_path(root, path);
525 btrfs_free_path(path);
526 mutex_unlock(&root->fs_info->fs_mutex);
527 make_bad_inode(inode);
530 static void fill_inode_item(struct extent_buffer *leaf,
531 struct btrfs_inode_item *item,
534 btrfs_set_inode_uid(leaf, item, inode->i_uid);
535 btrfs_set_inode_gid(leaf, item, inode->i_gid);
536 btrfs_set_inode_size(leaf, item, inode->i_size);
537 btrfs_set_inode_mode(leaf, item, inode->i_mode);
538 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
540 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
541 inode->i_atime.tv_sec);
542 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
543 inode->i_atime.tv_nsec);
545 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
546 inode->i_mtime.tv_sec);
547 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
548 inode->i_mtime.tv_nsec);
550 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
551 inode->i_ctime.tv_sec);
552 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
553 inode->i_ctime.tv_nsec);
555 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
556 btrfs_set_inode_generation(leaf, item, inode->i_generation);
557 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
558 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
559 btrfs_set_inode_block_group(leaf, item,
560 BTRFS_I(inode)->block_group->key.objectid);
563 int btrfs_update_inode(struct btrfs_trans_handle *trans,
564 struct btrfs_root *root,
567 struct btrfs_inode_item *inode_item;
568 struct btrfs_path *path;
569 struct extent_buffer *leaf;
572 path = btrfs_alloc_path();
574 ret = btrfs_lookup_inode(trans, root, path,
575 &BTRFS_I(inode)->location, 1);
582 leaf = path->nodes[0];
583 inode_item = btrfs_item_ptr(leaf, path->slots[0],
584 struct btrfs_inode_item);
586 fill_inode_item(leaf, inode_item, inode);
587 btrfs_mark_buffer_dirty(leaf);
588 btrfs_set_inode_last_trans(trans, inode);
591 btrfs_release_path(root, path);
592 btrfs_free_path(path);
597 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
598 struct btrfs_root *root,
600 struct dentry *dentry)
602 struct btrfs_path *path;
603 const char *name = dentry->d_name.name;
604 int name_len = dentry->d_name.len;
606 struct extent_buffer *leaf;
607 struct btrfs_dir_item *di;
608 struct btrfs_key key;
610 path = btrfs_alloc_path();
616 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
626 leaf = path->nodes[0];
627 btrfs_dir_item_key_to_cpu(leaf, di, &key);
628 ret = btrfs_delete_one_dir_name(trans, root, path, di);
631 btrfs_release_path(root, path);
633 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
634 key.objectid, name, name_len, -1);
643 ret = btrfs_delete_one_dir_name(trans, root, path, di);
645 dentry->d_inode->i_ctime = dir->i_ctime;
646 ret = btrfs_del_inode_ref(trans, root, name, name_len,
647 dentry->d_inode->i_ino,
648 dentry->d_parent->d_inode->i_ino);
650 printk("failed to delete reference to %.*s, "
651 "inode %lu parent %lu\n", name_len, name,
652 dentry->d_inode->i_ino,
653 dentry->d_parent->d_inode->i_ino);
656 btrfs_free_path(path);
658 dir->i_size -= name_len * 2;
659 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
660 btrfs_update_inode(trans, root, dir);
661 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
662 dentry->d_inode->i_nlink--;
664 drop_nlink(dentry->d_inode);
666 ret = btrfs_update_inode(trans, root, dentry->d_inode);
667 dir->i_sb->s_dirt = 1;
672 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
674 struct btrfs_root *root;
675 struct btrfs_trans_handle *trans;
676 struct inode *inode = dentry->d_inode;
678 unsigned long nr = 0;
680 root = BTRFS_I(dir)->root;
681 mutex_lock(&root->fs_info->fs_mutex);
683 ret = btrfs_check_free_space(root, 1, 1);
687 trans = btrfs_start_transaction(root, 1);
689 btrfs_set_trans_block_group(trans, dir);
690 ret = btrfs_unlink_trans(trans, root, dir, dentry);
691 nr = trans->blocks_used;
693 if (inode->i_nlink == 0) {
695 /* if the inode isn't linked anywhere,
696 * we don't need to worry about
699 found = btrfs_del_ordered_inode(inode);
701 atomic_dec(&inode->i_count);
705 btrfs_end_transaction(trans, root);
707 mutex_unlock(&root->fs_info->fs_mutex);
708 btrfs_btree_balance_dirty(root, nr);
709 btrfs_throttle(root);
713 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
715 struct inode *inode = dentry->d_inode;
718 struct btrfs_root *root = BTRFS_I(dir)->root;
719 struct btrfs_trans_handle *trans;
720 unsigned long nr = 0;
722 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
725 mutex_lock(&root->fs_info->fs_mutex);
726 ret = btrfs_check_free_space(root, 1, 1);
730 trans = btrfs_start_transaction(root, 1);
731 btrfs_set_trans_block_group(trans, dir);
733 /* now the directory is empty */
734 err = btrfs_unlink_trans(trans, root, dir, dentry);
739 nr = trans->blocks_used;
740 ret = btrfs_end_transaction(trans, root);
742 mutex_unlock(&root->fs_info->fs_mutex);
743 btrfs_btree_balance_dirty(root, nr);
744 btrfs_throttle(root);
752 * this can truncate away extent items, csum items and directory items.
753 * It starts at a high offset and removes keys until it can't find
754 * any higher than i_size.
756 * csum items that cross the new i_size are truncated to the new size
759 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
760 struct btrfs_root *root,
765 struct btrfs_path *path;
766 struct btrfs_key key;
767 struct btrfs_key found_key;
769 struct extent_buffer *leaf;
770 struct btrfs_file_extent_item *fi;
771 u64 extent_start = 0;
772 u64 extent_num_bytes = 0;
778 int pending_del_nr = 0;
779 int pending_del_slot = 0;
780 int extent_type = -1;
782 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
783 path = btrfs_alloc_path();
787 /* FIXME, add redo link to tree so we don't leak on crash */
788 key.objectid = inode->i_ino;
789 key.offset = (u64)-1;
792 btrfs_init_path(path);
794 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
799 BUG_ON(path->slots[0] == 0);
805 leaf = path->nodes[0];
806 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
807 found_type = btrfs_key_type(&found_key);
809 if (found_key.objectid != inode->i_ino)
812 if (found_type < min_type)
815 item_end = found_key.offset;
816 if (found_type == BTRFS_EXTENT_DATA_KEY) {
817 fi = btrfs_item_ptr(leaf, path->slots[0],
818 struct btrfs_file_extent_item);
819 extent_type = btrfs_file_extent_type(leaf, fi);
820 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
822 btrfs_file_extent_num_bytes(leaf, fi);
823 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
824 struct btrfs_item *item = btrfs_item_nr(leaf,
826 item_end += btrfs_file_extent_inline_len(leaf,
831 if (found_type == BTRFS_CSUM_ITEM_KEY) {
832 ret = btrfs_csum_truncate(trans, root, path,
836 if (item_end < inode->i_size) {
837 if (found_type == BTRFS_DIR_ITEM_KEY) {
838 found_type = BTRFS_INODE_ITEM_KEY;
839 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
840 found_type = BTRFS_CSUM_ITEM_KEY;
841 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
842 found_type = BTRFS_XATTR_ITEM_KEY;
843 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
844 found_type = BTRFS_INODE_REF_KEY;
845 } else if (found_type) {
850 btrfs_set_key_type(&key, found_type);
853 if (found_key.offset >= inode->i_size)
859 /* FIXME, shrink the extent if the ref count is only 1 */
860 if (found_type != BTRFS_EXTENT_DATA_KEY)
863 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
865 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
868 btrfs_file_extent_num_bytes(leaf, fi);
869 extent_num_bytes = inode->i_size -
870 found_key.offset + root->sectorsize - 1;
871 extent_num_bytes = extent_num_bytes &
872 ~((u64)root->sectorsize - 1);
873 btrfs_set_file_extent_num_bytes(leaf, fi,
875 num_dec = (orig_num_bytes -
877 if (extent_start != 0)
878 dec_i_blocks(inode, num_dec);
879 btrfs_mark_buffer_dirty(leaf);
882 btrfs_file_extent_disk_num_bytes(leaf,
884 /* FIXME blocksize != 4096 */
885 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
886 if (extent_start != 0) {
888 dec_i_blocks(inode, num_dec);
890 root_gen = btrfs_header_generation(leaf);
891 root_owner = btrfs_header_owner(leaf);
893 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
895 u32 newsize = inode->i_size - found_key.offset;
896 dec_i_blocks(inode, item_end + 1 -
897 found_key.offset - newsize);
899 btrfs_file_extent_calc_inline_size(newsize);
900 ret = btrfs_truncate_item(trans, root, path,
904 dec_i_blocks(inode, item_end + 1 -
910 if (!pending_del_nr) {
911 /* no pending yet, add ourselves */
912 pending_del_slot = path->slots[0];
914 } else if (pending_del_nr &&
915 path->slots[0] + 1 == pending_del_slot) {
916 /* hop on the pending chunk */
918 pending_del_slot = path->slots[0];
920 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
926 ret = btrfs_free_extent(trans, root, extent_start,
929 root_gen, inode->i_ino,
930 found_key.offset, 0);
934 if (path->slots[0] == 0) {
937 btrfs_release_path(root, path);
942 if (pending_del_nr &&
943 path->slots[0] + 1 != pending_del_slot) {
944 struct btrfs_key debug;
946 btrfs_item_key_to_cpu(path->nodes[0], &debug,
948 ret = btrfs_del_items(trans, root, path,
953 btrfs_release_path(root, path);
959 if (pending_del_nr) {
960 ret = btrfs_del_items(trans, root, path, pending_del_slot,
963 btrfs_release_path(root, path);
964 btrfs_free_path(path);
965 inode->i_sb->s_dirt = 1;
969 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
973 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
974 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
975 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
978 WARN_ON(!PageLocked(page));
979 set_page_extent_mapped(page);
981 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
982 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
985 if (zero_start != PAGE_CACHE_SIZE) {
987 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
988 flush_dcache_page(page);
991 set_page_dirty(page);
992 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
998 * taken from block_truncate_page, but does cow as it zeros out
999 * any bytes left in the last page in the file.
1001 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
1003 struct inode *inode = mapping->host;
1004 struct btrfs_root *root = BTRFS_I(inode)->root;
1005 u32 blocksize = root->sectorsize;
1006 pgoff_t index = from >> PAGE_CACHE_SHIFT;
1007 unsigned offset = from & (PAGE_CACHE_SIZE-1);
1012 if ((offset & (blocksize - 1)) == 0)
1016 page = grab_cache_page(mapping, index);
1019 if (!PageUptodate(page)) {
1020 ret = btrfs_readpage(NULL, page);
1022 if (!PageUptodate(page)) {
1027 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1029 ret = btrfs_cow_one_page(inode, page, offset);
1032 page_cache_release(page);
1037 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1039 struct inode *inode = dentry->d_inode;
1042 err = inode_change_ok(inode, attr);
1046 if (S_ISREG(inode->i_mode) &&
1047 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1048 struct btrfs_trans_handle *trans;
1049 struct btrfs_root *root = BTRFS_I(inode)->root;
1050 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1052 u64 mask = root->sectorsize - 1;
1053 u64 hole_start = (inode->i_size + mask) & ~mask;
1054 u64 block_end = (attr->ia_size + mask) & ~mask;
1058 if (attr->ia_size <= hole_start)
1061 mutex_lock(&root->fs_info->fs_mutex);
1062 err = btrfs_check_free_space(root, 1, 0);
1063 mutex_unlock(&root->fs_info->fs_mutex);
1067 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1069 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1070 hole_size = block_end - hole_start;
1072 mutex_lock(&root->fs_info->fs_mutex);
1073 trans = btrfs_start_transaction(root, 1);
1074 btrfs_set_trans_block_group(trans, inode);
1075 err = btrfs_drop_extents(trans, root, inode,
1076 hole_start, block_end, hole_start,
1079 if (alloc_hint != EXTENT_MAP_INLINE) {
1080 err = btrfs_insert_file_extent(trans, root,
1084 btrfs_drop_extent_cache(inode, hole_start,
1086 btrfs_check_file(root, inode);
1088 btrfs_end_transaction(trans, root);
1089 mutex_unlock(&root->fs_info->fs_mutex);
1090 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1095 err = inode_setattr(inode, attr);
1100 void btrfs_put_inode(struct inode *inode)
1104 if (!BTRFS_I(inode)->ordered_trans) {
1108 if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) ||
1109 mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))
1112 ret = btrfs_del_ordered_inode(inode);
1114 atomic_dec(&inode->i_count);
1118 void btrfs_delete_inode(struct inode *inode)
1120 struct btrfs_trans_handle *trans;
1121 struct btrfs_root *root = BTRFS_I(inode)->root;
1125 truncate_inode_pages(&inode->i_data, 0);
1126 if (is_bad_inode(inode)) {
1131 mutex_lock(&root->fs_info->fs_mutex);
1132 trans = btrfs_start_transaction(root, 1);
1134 btrfs_set_trans_block_group(trans, inode);
1135 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1137 goto no_delete_lock;
1139 nr = trans->blocks_used;
1142 btrfs_end_transaction(trans, root);
1143 mutex_unlock(&root->fs_info->fs_mutex);
1144 btrfs_btree_balance_dirty(root, nr);
1145 btrfs_throttle(root);
1149 nr = trans->blocks_used;
1150 btrfs_end_transaction(trans, root);
1151 mutex_unlock(&root->fs_info->fs_mutex);
1152 btrfs_btree_balance_dirty(root, nr);
1153 btrfs_throttle(root);
1159 * this returns the key found in the dir entry in the location pointer.
1160 * If no dir entries were found, location->objectid is 0.
1162 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1163 struct btrfs_key *location)
1165 const char *name = dentry->d_name.name;
1166 int namelen = dentry->d_name.len;
1167 struct btrfs_dir_item *di;
1168 struct btrfs_path *path;
1169 struct btrfs_root *root = BTRFS_I(dir)->root;
1172 if (namelen == 1 && strcmp(name, ".") == 0) {
1173 location->objectid = dir->i_ino;
1174 location->type = BTRFS_INODE_ITEM_KEY;
1175 location->offset = 0;
1178 path = btrfs_alloc_path();
1181 if (namelen == 2 && strcmp(name, "..") == 0) {
1182 struct btrfs_key key;
1183 struct extent_buffer *leaf;
1187 key.objectid = dir->i_ino;
1188 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1190 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1194 leaf = path->nodes[0];
1195 slot = path->slots[0];
1196 nritems = btrfs_header_nritems(leaf);
1197 if (slot >= nritems)
1200 btrfs_item_key_to_cpu(leaf, &key, slot);
1201 if (key.objectid != dir->i_ino ||
1202 key.type != BTRFS_INODE_REF_KEY) {
1205 location->objectid = key.offset;
1206 location->type = BTRFS_INODE_ITEM_KEY;
1207 location->offset = 0;
1211 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1215 if (!di || IS_ERR(di)) {
1218 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1220 btrfs_free_path(path);
1223 location->objectid = 0;
1228 * when we hit a tree root in a directory, the btrfs part of the inode
1229 * needs to be changed to reflect the root directory of the tree root. This
1230 * is kind of like crossing a mount point.
1232 static int fixup_tree_root_location(struct btrfs_root *root,
1233 struct btrfs_key *location,
1234 struct btrfs_root **sub_root,
1235 struct dentry *dentry)
1237 struct btrfs_path *path;
1238 struct btrfs_root_item *ri;
1240 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1242 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1245 path = btrfs_alloc_path();
1247 mutex_lock(&root->fs_info->fs_mutex);
1249 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1250 dentry->d_name.name,
1251 dentry->d_name.len);
1252 if (IS_ERR(*sub_root))
1253 return PTR_ERR(*sub_root);
1255 ri = &(*sub_root)->root_item;
1256 location->objectid = btrfs_root_dirid(ri);
1257 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1258 location->offset = 0;
1260 btrfs_free_path(path);
1261 mutex_unlock(&root->fs_info->fs_mutex);
1265 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1267 struct btrfs_iget_args *args = p;
1268 inode->i_ino = args->ino;
1269 BTRFS_I(inode)->root = args->root;
1270 BTRFS_I(inode)->delalloc_bytes = 0;
1271 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1272 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1273 inode->i_mapping, GFP_NOFS);
1277 static int btrfs_find_actor(struct inode *inode, void *opaque)
1279 struct btrfs_iget_args *args = opaque;
1280 return (args->ino == inode->i_ino &&
1281 args->root == BTRFS_I(inode)->root);
1284 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1287 struct btrfs_iget_args args;
1288 args.ino = objectid;
1289 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1294 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1297 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1298 struct btrfs_root *root)
1300 struct inode *inode;
1301 struct btrfs_iget_args args;
1302 args.ino = objectid;
1305 inode = iget5_locked(s, objectid, btrfs_find_actor,
1306 btrfs_init_locked_inode,
1311 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1312 struct nameidata *nd)
1314 struct inode * inode;
1315 struct btrfs_inode *bi = BTRFS_I(dir);
1316 struct btrfs_root *root = bi->root;
1317 struct btrfs_root *sub_root = root;
1318 struct btrfs_key location;
1321 if (dentry->d_name.len > BTRFS_NAME_LEN)
1322 return ERR_PTR(-ENAMETOOLONG);
1324 mutex_lock(&root->fs_info->fs_mutex);
1325 ret = btrfs_inode_by_name(dir, dentry, &location);
1326 mutex_unlock(&root->fs_info->fs_mutex);
1329 return ERR_PTR(ret);
1332 if (location.objectid) {
1333 ret = fixup_tree_root_location(root, &location, &sub_root,
1336 return ERR_PTR(ret);
1338 return ERR_PTR(-ENOENT);
1339 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1342 return ERR_PTR(-EACCES);
1343 if (inode->i_state & I_NEW) {
1344 /* the inode and parent dir are two different roots */
1345 if (sub_root != root) {
1347 sub_root->inode = inode;
1349 BTRFS_I(inode)->root = sub_root;
1350 memcpy(&BTRFS_I(inode)->location, &location,
1352 btrfs_read_locked_inode(inode);
1353 unlock_new_inode(inode);
1356 return d_splice_alias(inode, dentry);
1359 static unsigned char btrfs_filetype_table[] = {
1360 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1363 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1365 struct inode *inode = filp->f_dentry->d_inode;
1366 struct btrfs_root *root = BTRFS_I(inode)->root;
1367 struct btrfs_item *item;
1368 struct btrfs_dir_item *di;
1369 struct btrfs_key key;
1370 struct btrfs_key found_key;
1371 struct btrfs_path *path;
1374 struct extent_buffer *leaf;
1377 unsigned char d_type;
1382 int key_type = BTRFS_DIR_INDEX_KEY;
1387 /* FIXME, use a real flag for deciding about the key type */
1388 if (root->fs_info->tree_root == root)
1389 key_type = BTRFS_DIR_ITEM_KEY;
1391 /* special case for "." */
1392 if (filp->f_pos == 0) {
1393 over = filldir(dirent, ".", 1,
1401 mutex_lock(&root->fs_info->fs_mutex);
1402 key.objectid = inode->i_ino;
1403 path = btrfs_alloc_path();
1406 /* special case for .., just use the back ref */
1407 if (filp->f_pos == 1) {
1408 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1410 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1412 leaf = path->nodes[0];
1413 slot = path->slots[0];
1414 nritems = btrfs_header_nritems(leaf);
1415 if (slot >= nritems) {
1416 btrfs_release_path(root, path);
1417 goto read_dir_items;
1419 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1420 btrfs_release_path(root, path);
1421 if (found_key.objectid != key.objectid ||
1422 found_key.type != BTRFS_INODE_REF_KEY)
1423 goto read_dir_items;
1424 over = filldir(dirent, "..", 2,
1425 2, found_key.offset, DT_DIR);
1432 btrfs_set_key_type(&key, key_type);
1433 key.offset = filp->f_pos;
1435 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1440 leaf = path->nodes[0];
1441 nritems = btrfs_header_nritems(leaf);
1442 slot = path->slots[0];
1443 if (advance || slot >= nritems) {
1444 if (slot >= nritems -1) {
1445 ret = btrfs_next_leaf(root, path);
1448 leaf = path->nodes[0];
1449 nritems = btrfs_header_nritems(leaf);
1450 slot = path->slots[0];
1457 item = btrfs_item_nr(leaf, slot);
1458 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1460 if (found_key.objectid != key.objectid)
1462 if (btrfs_key_type(&found_key) != key_type)
1464 if (found_key.offset < filp->f_pos)
1467 filp->f_pos = found_key.offset;
1469 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1471 di_total = btrfs_item_size(leaf, item);
1472 while(di_cur < di_total) {
1473 struct btrfs_key location;
1475 name_len = btrfs_dir_name_len(leaf, di);
1476 if (name_len < 32) {
1477 name_ptr = tmp_name;
1479 name_ptr = kmalloc(name_len, GFP_NOFS);
1482 read_extent_buffer(leaf, name_ptr,
1483 (unsigned long)(di + 1), name_len);
1485 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1486 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1487 over = filldir(dirent, name_ptr, name_len,
1492 if (name_ptr != tmp_name)
1497 di_len = btrfs_dir_name_len(leaf, di) +
1498 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1500 di = (struct btrfs_dir_item *)((char *)di + di_len);
1503 if (key_type == BTRFS_DIR_INDEX_KEY)
1504 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1510 btrfs_release_path(root, path);
1511 btrfs_free_path(path);
1512 mutex_unlock(&root->fs_info->fs_mutex);
1516 int btrfs_write_inode(struct inode *inode, int wait)
1518 struct btrfs_root *root = BTRFS_I(inode)->root;
1519 struct btrfs_trans_handle *trans;
1523 mutex_lock(&root->fs_info->fs_mutex);
1524 trans = btrfs_start_transaction(root, 1);
1525 btrfs_set_trans_block_group(trans, inode);
1526 ret = btrfs_commit_transaction(trans, root);
1527 mutex_unlock(&root->fs_info->fs_mutex);
1533 * This is somewhat expensive, updating the tree every time the
1534 * inode changes. But, it is most likely to find the inode in cache.
1535 * FIXME, needs more benchmarking...there are no reasons other than performance
1536 * to keep or drop this code.
1538 void btrfs_dirty_inode(struct inode *inode)
1540 struct btrfs_root *root = BTRFS_I(inode)->root;
1541 struct btrfs_trans_handle *trans;
1543 mutex_lock(&root->fs_info->fs_mutex);
1544 trans = btrfs_start_transaction(root, 1);
1545 btrfs_set_trans_block_group(trans, inode);
1546 btrfs_update_inode(trans, root, inode);
1547 btrfs_end_transaction(trans, root);
1548 mutex_unlock(&root->fs_info->fs_mutex);
1551 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1552 struct btrfs_root *root,
1553 const char *name, int name_len,
1556 struct btrfs_block_group_cache *group,
1559 struct inode *inode;
1560 struct btrfs_inode_item *inode_item;
1561 struct btrfs_block_group_cache *new_inode_group;
1562 struct btrfs_key *location;
1563 struct btrfs_path *path;
1564 struct btrfs_inode_ref *ref;
1565 struct btrfs_key key[2];
1571 path = btrfs_alloc_path();
1574 inode = new_inode(root->fs_info->sb);
1576 return ERR_PTR(-ENOMEM);
1578 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1579 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1580 inode->i_mapping, GFP_NOFS);
1581 BTRFS_I(inode)->delalloc_bytes = 0;
1582 BTRFS_I(inode)->root = root;
1588 new_inode_group = btrfs_find_block_group(root, group, 0,
1589 BTRFS_BLOCK_GROUP_METADATA, owner);
1590 if (!new_inode_group) {
1591 printk("find_block group failed\n");
1592 new_inode_group = group;
1594 BTRFS_I(inode)->block_group = new_inode_group;
1595 BTRFS_I(inode)->flags = 0;
1597 key[0].objectid = objectid;
1598 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1601 key[1].objectid = objectid;
1602 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1603 key[1].offset = ref_objectid;
1605 sizes[0] = sizeof(struct btrfs_inode_item);
1606 sizes[1] = name_len + sizeof(*ref);
1608 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1612 if (objectid > root->highest_inode)
1613 root->highest_inode = objectid;
1615 inode->i_uid = current->fsuid;
1616 inode->i_gid = current->fsgid;
1617 inode->i_mode = mode;
1618 inode->i_ino = objectid;
1619 inode->i_blocks = 0;
1620 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1621 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1622 struct btrfs_inode_item);
1623 fill_inode_item(path->nodes[0], inode_item, inode);
1625 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1626 struct btrfs_inode_ref);
1627 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1628 ptr = (unsigned long)(ref + 1);
1629 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1631 btrfs_mark_buffer_dirty(path->nodes[0]);
1632 btrfs_free_path(path);
1634 location = &BTRFS_I(inode)->location;
1635 location->objectid = objectid;
1636 location->offset = 0;
1637 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1639 insert_inode_hash(inode);
1642 btrfs_free_path(path);
1643 return ERR_PTR(ret);
1646 static inline u8 btrfs_inode_type(struct inode *inode)
1648 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1651 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1652 struct dentry *dentry, struct inode *inode,
1656 struct btrfs_key key;
1657 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1658 struct inode *parent_inode;
1660 key.objectid = inode->i_ino;
1661 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1664 ret = btrfs_insert_dir_item(trans, root,
1665 dentry->d_name.name, dentry->d_name.len,
1666 dentry->d_parent->d_inode->i_ino,
1667 &key, btrfs_inode_type(inode));
1670 ret = btrfs_insert_inode_ref(trans, root,
1671 dentry->d_name.name,
1674 dentry->d_parent->d_inode->i_ino);
1676 parent_inode = dentry->d_parent->d_inode;
1677 parent_inode->i_size += dentry->d_name.len * 2;
1678 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1679 ret = btrfs_update_inode(trans, root,
1680 dentry->d_parent->d_inode);
1685 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1686 struct dentry *dentry, struct inode *inode,
1689 int err = btrfs_add_link(trans, dentry, inode, backref);
1691 d_instantiate(dentry, inode);
1699 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1700 int mode, dev_t rdev)
1702 struct btrfs_trans_handle *trans;
1703 struct btrfs_root *root = BTRFS_I(dir)->root;
1704 struct inode *inode = NULL;
1708 unsigned long nr = 0;
1710 if (!new_valid_dev(rdev))
1713 mutex_lock(&root->fs_info->fs_mutex);
1714 err = btrfs_check_free_space(root, 1, 0);
1718 trans = btrfs_start_transaction(root, 1);
1719 btrfs_set_trans_block_group(trans, dir);
1721 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1727 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1729 dentry->d_parent->d_inode->i_ino, objectid,
1730 BTRFS_I(dir)->block_group, mode);
1731 err = PTR_ERR(inode);
1735 btrfs_set_trans_block_group(trans, inode);
1736 err = btrfs_add_nondir(trans, dentry, inode, 0);
1740 inode->i_op = &btrfs_special_inode_operations;
1741 init_special_inode(inode, inode->i_mode, rdev);
1742 btrfs_update_inode(trans, root, inode);
1744 dir->i_sb->s_dirt = 1;
1745 btrfs_update_inode_block_group(trans, inode);
1746 btrfs_update_inode_block_group(trans, dir);
1748 nr = trans->blocks_used;
1749 btrfs_end_transaction(trans, root);
1751 mutex_unlock(&root->fs_info->fs_mutex);
1754 inode_dec_link_count(inode);
1757 btrfs_btree_balance_dirty(root, nr);
1758 btrfs_throttle(root);
1762 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1763 int mode, struct nameidata *nd)
1765 struct btrfs_trans_handle *trans;
1766 struct btrfs_root *root = BTRFS_I(dir)->root;
1767 struct inode *inode = NULL;
1770 unsigned long nr = 0;
1773 mutex_lock(&root->fs_info->fs_mutex);
1774 err = btrfs_check_free_space(root, 1, 0);
1777 trans = btrfs_start_transaction(root, 1);
1778 btrfs_set_trans_block_group(trans, dir);
1780 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1786 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1788 dentry->d_parent->d_inode->i_ino,
1789 objectid, BTRFS_I(dir)->block_group, mode);
1790 err = PTR_ERR(inode);
1794 btrfs_set_trans_block_group(trans, inode);
1795 err = btrfs_add_nondir(trans, dentry, inode, 0);
1799 inode->i_mapping->a_ops = &btrfs_aops;
1800 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
1801 inode->i_fop = &btrfs_file_operations;
1802 inode->i_op = &btrfs_file_inode_operations;
1803 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1804 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1805 inode->i_mapping, GFP_NOFS);
1806 BTRFS_I(inode)->delalloc_bytes = 0;
1807 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
1809 dir->i_sb->s_dirt = 1;
1810 btrfs_update_inode_block_group(trans, inode);
1811 btrfs_update_inode_block_group(trans, dir);
1813 nr = trans->blocks_used;
1814 btrfs_end_transaction(trans, root);
1816 mutex_unlock(&root->fs_info->fs_mutex);
1819 inode_dec_link_count(inode);
1822 btrfs_btree_balance_dirty(root, nr);
1823 btrfs_throttle(root);
1827 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1828 struct dentry *dentry)
1830 struct btrfs_trans_handle *trans;
1831 struct btrfs_root *root = BTRFS_I(dir)->root;
1832 struct inode *inode = old_dentry->d_inode;
1833 unsigned long nr = 0;
1837 if (inode->i_nlink == 0)
1840 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1845 mutex_lock(&root->fs_info->fs_mutex);
1846 err = btrfs_check_free_space(root, 1, 0);
1849 trans = btrfs_start_transaction(root, 1);
1851 btrfs_set_trans_block_group(trans, dir);
1852 atomic_inc(&inode->i_count);
1853 err = btrfs_add_nondir(trans, dentry, inode, 1);
1858 dir->i_sb->s_dirt = 1;
1859 btrfs_update_inode_block_group(trans, dir);
1860 err = btrfs_update_inode(trans, root, inode);
1865 nr = trans->blocks_used;
1866 btrfs_end_transaction(trans, root);
1868 mutex_unlock(&root->fs_info->fs_mutex);
1871 inode_dec_link_count(inode);
1874 btrfs_btree_balance_dirty(root, nr);
1875 btrfs_throttle(root);
1879 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1881 struct inode *inode;
1882 struct btrfs_trans_handle *trans;
1883 struct btrfs_root *root = BTRFS_I(dir)->root;
1885 int drop_on_err = 0;
1887 unsigned long nr = 1;
1889 mutex_lock(&root->fs_info->fs_mutex);
1890 err = btrfs_check_free_space(root, 1, 0);
1894 trans = btrfs_start_transaction(root, 1);
1895 btrfs_set_trans_block_group(trans, dir);
1897 if (IS_ERR(trans)) {
1898 err = PTR_ERR(trans);
1902 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1908 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1910 dentry->d_parent->d_inode->i_ino, objectid,
1911 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1912 if (IS_ERR(inode)) {
1913 err = PTR_ERR(inode);
1918 inode->i_op = &btrfs_dir_inode_operations;
1919 inode->i_fop = &btrfs_dir_file_operations;
1920 btrfs_set_trans_block_group(trans, inode);
1923 err = btrfs_update_inode(trans, root, inode);
1927 err = btrfs_add_link(trans, dentry, inode, 0);
1931 d_instantiate(dentry, inode);
1933 dir->i_sb->s_dirt = 1;
1934 btrfs_update_inode_block_group(trans, inode);
1935 btrfs_update_inode_block_group(trans, dir);
1938 nr = trans->blocks_used;
1939 btrfs_end_transaction(trans, root);
1942 mutex_unlock(&root->fs_info->fs_mutex);
1945 btrfs_btree_balance_dirty(root, nr);
1946 btrfs_throttle(root);
1950 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1951 size_t pg_offset, u64 start, u64 len,
1957 u64 extent_start = 0;
1959 u64 objectid = inode->i_ino;
1961 struct btrfs_path *path;
1962 struct btrfs_root *root = BTRFS_I(inode)->root;
1963 struct btrfs_file_extent_item *item;
1964 struct extent_buffer *leaf;
1965 struct btrfs_key found_key;
1966 struct extent_map *em = NULL;
1967 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1968 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1969 struct btrfs_trans_handle *trans = NULL;
1971 path = btrfs_alloc_path();
1973 mutex_lock(&root->fs_info->fs_mutex);
1976 spin_lock(&em_tree->lock);
1977 em = lookup_extent_mapping(em_tree, start, len);
1978 spin_unlock(&em_tree->lock);
1981 if (em->start > start) {
1982 printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n",
1983 start, len, em->start, em->len);
1986 if (em->block_start == EXTENT_MAP_INLINE && page)
1987 free_extent_map(em);
1991 em = alloc_extent_map(GFP_NOFS);
1997 em->start = EXTENT_MAP_HOLE;
1999 em->bdev = inode->i_sb->s_bdev;
2000 ret = btrfs_lookup_file_extent(trans, root, path,
2001 objectid, start, trans != NULL);
2008 if (path->slots[0] == 0)
2013 leaf = path->nodes[0];
2014 item = btrfs_item_ptr(leaf, path->slots[0],
2015 struct btrfs_file_extent_item);
2016 /* are we inside the extent that was found? */
2017 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2018 found_type = btrfs_key_type(&found_key);
2019 if (found_key.objectid != objectid ||
2020 found_type != BTRFS_EXTENT_DATA_KEY) {
2024 found_type = btrfs_file_extent_type(leaf, item);
2025 extent_start = found_key.offset;
2026 if (found_type == BTRFS_FILE_EXTENT_REG) {
2027 extent_end = extent_start +
2028 btrfs_file_extent_num_bytes(leaf, item);
2030 if (start < extent_start || start >= extent_end) {
2032 if (start < extent_start) {
2033 if (start + len <= extent_start)
2035 em->len = extent_end - extent_start;
2041 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2043 em->start = extent_start;
2044 em->len = extent_end - extent_start;
2045 em->block_start = EXTENT_MAP_HOLE;
2048 bytenr += btrfs_file_extent_offset(leaf, item);
2049 em->block_start = bytenr;
2050 em->start = extent_start;
2051 em->len = extent_end - extent_start;
2053 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2058 size_t extent_offset;
2061 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2063 extent_end = (extent_start + size + root->sectorsize - 1) &
2064 ~((u64)root->sectorsize - 1);
2065 if (start < extent_start || start >= extent_end) {
2067 if (start < extent_start) {
2068 if (start + len <= extent_start)
2070 em->len = extent_end - extent_start;
2076 em->block_start = EXTENT_MAP_INLINE;
2079 em->start = extent_start;
2084 page_start = page_offset(page) + pg_offset;
2085 extent_offset = page_start - extent_start;
2086 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2087 size - extent_offset);
2088 em->start = extent_start + extent_offset;
2089 em->len = (copy_size + root->sectorsize - 1) &
2090 ~((u64)root->sectorsize - 1);
2092 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2093 if (create == 0 && !PageUptodate(page)) {
2094 read_extent_buffer(leaf, map + pg_offset, ptr,
2096 flush_dcache_page(page);
2097 } else if (create && PageUptodate(page)) {
2100 free_extent_map(em);
2102 btrfs_release_path(root, path);
2103 trans = btrfs_start_transaction(root, 1);
2106 write_extent_buffer(leaf, map + pg_offset, ptr,
2108 btrfs_mark_buffer_dirty(leaf);
2111 set_extent_uptodate(io_tree, em->start,
2112 extent_map_end(em) - 1, GFP_NOFS);
2115 printk("unkknown found_type %d\n", found_type);
2122 em->block_start = EXTENT_MAP_HOLE;
2124 btrfs_release_path(root, path);
2125 if (em->start > start || extent_map_end(em) <= start) {
2126 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2132 spin_lock(&em_tree->lock);
2133 ret = add_extent_mapping(em_tree, em);
2134 if (ret == -EEXIST) {
2135 free_extent_map(em);
2136 em = lookup_extent_mapping(em_tree, start, len);
2139 printk("failing to insert %Lu %Lu\n", start, len);
2142 spin_unlock(&em_tree->lock);
2144 btrfs_free_path(path);
2146 ret = btrfs_end_transaction(trans, root);
2150 mutex_unlock(&root->fs_info->fs_mutex);
2152 free_extent_map(em);
2154 return ERR_PTR(err);
2159 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2161 return extent_bmap(mapping, iblock, btrfs_get_extent);
2164 int btrfs_readpage(struct file *file, struct page *page)
2166 struct extent_io_tree *tree;
2167 tree = &BTRFS_I(page->mapping->host)->io_tree;
2168 return extent_read_full_page(tree, page, btrfs_get_extent);
2171 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2173 struct extent_io_tree *tree;
2176 if (current->flags & PF_MEMALLOC) {
2177 redirty_page_for_writepage(wbc, page);
2181 tree = &BTRFS_I(page->mapping->host)->io_tree;
2182 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2185 static int btrfs_writepages(struct address_space *mapping,
2186 struct writeback_control *wbc)
2188 struct extent_io_tree *tree;
2189 tree = &BTRFS_I(mapping->host)->io_tree;
2190 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2194 btrfs_readpages(struct file *file, struct address_space *mapping,
2195 struct list_head *pages, unsigned nr_pages)
2197 struct extent_io_tree *tree;
2198 tree = &BTRFS_I(mapping->host)->io_tree;
2199 return extent_readpages(tree, mapping, pages, nr_pages,
2203 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2205 struct extent_io_tree *tree;
2206 struct extent_map_tree *map;
2209 tree = &BTRFS_I(page->mapping->host)->io_tree;
2210 map = &BTRFS_I(page->mapping->host)->extent_tree;
2211 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2213 ClearPagePrivate(page);
2214 set_page_private(page, 0);
2215 page_cache_release(page);
2220 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2222 struct extent_io_tree *tree;
2224 tree = &BTRFS_I(page->mapping->host)->io_tree;
2225 extent_invalidatepage(tree, page, offset);
2226 btrfs_releasepage(page, GFP_NOFS);
2230 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2231 * called from a page fault handler when a page is first dirtied. Hence we must
2232 * be careful to check for EOF conditions here. We set the page up correctly
2233 * for a written page which means we get ENOSPC checking when writing into
2234 * holes and correct delalloc and unwritten extent mapping on filesystems that
2235 * support these features.
2237 * We are not allowed to take the i_mutex here so we have to play games to
2238 * protect against truncate races as the page could now be beyond EOF. Because
2239 * vmtruncate() writes the inode size before removing pages, once we have the
2240 * page lock we can determine safely if the page is beyond EOF. If it is not
2241 * beyond EOF, then the page is guaranteed safe against truncation until we
2244 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2246 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2247 struct btrfs_root *root = BTRFS_I(inode)->root;
2253 mutex_lock(&root->fs_info->fs_mutex);
2254 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2255 mutex_unlock(&root->fs_info->fs_mutex);
2262 wait_on_page_writeback(page);
2263 size = i_size_read(inode);
2264 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2266 if ((page->mapping != inode->i_mapping) ||
2267 (page_start > size)) {
2268 /* page got truncated out from underneath us */
2272 /* page is wholly or partially inside EOF */
2273 if (page_start + PAGE_CACHE_SIZE > size)
2274 end = size & ~PAGE_CACHE_MASK;
2276 end = PAGE_CACHE_SIZE;
2278 ret = btrfs_cow_one_page(inode, page, end);
2286 static void btrfs_truncate(struct inode *inode)
2288 struct btrfs_root *root = BTRFS_I(inode)->root;
2290 struct btrfs_trans_handle *trans;
2293 if (!S_ISREG(inode->i_mode))
2295 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2298 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2300 mutex_lock(&root->fs_info->fs_mutex);
2301 trans = btrfs_start_transaction(root, 1);
2302 btrfs_set_trans_block_group(trans, inode);
2304 /* FIXME, add redo link to tree so we don't leak on crash */
2305 ret = btrfs_truncate_in_trans(trans, root, inode,
2306 BTRFS_EXTENT_DATA_KEY);
2307 btrfs_update_inode(trans, root, inode);
2308 nr = trans->blocks_used;
2310 ret = btrfs_end_transaction(trans, root);
2312 mutex_unlock(&root->fs_info->fs_mutex);
2313 btrfs_btree_balance_dirty(root, nr);
2314 btrfs_throttle(root);
2317 static int noinline create_subvol(struct btrfs_root *root, char *name,
2320 struct btrfs_trans_handle *trans;
2321 struct btrfs_key key;
2322 struct btrfs_root_item root_item;
2323 struct btrfs_inode_item *inode_item;
2324 struct extent_buffer *leaf;
2325 struct btrfs_root *new_root = root;
2326 struct inode *inode;
2331 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
2332 unsigned long nr = 1;
2334 mutex_lock(&root->fs_info->fs_mutex);
2335 ret = btrfs_check_free_space(root, 1, 0);
2339 trans = btrfs_start_transaction(root, 1);
2342 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2347 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
2348 objectid, trans->transid, 0, 0,
2351 return PTR_ERR(leaf);
2353 btrfs_set_header_nritems(leaf, 0);
2354 btrfs_set_header_level(leaf, 0);
2355 btrfs_set_header_bytenr(leaf, leaf->start);
2356 btrfs_set_header_generation(leaf, trans->transid);
2357 btrfs_set_header_owner(leaf, objectid);
2359 write_extent_buffer(leaf, root->fs_info->fsid,
2360 (unsigned long)btrfs_header_fsid(leaf),
2362 btrfs_mark_buffer_dirty(leaf);
2364 inode_item = &root_item.inode;
2365 memset(inode_item, 0, sizeof(*inode_item));
2366 inode_item->generation = cpu_to_le64(1);
2367 inode_item->size = cpu_to_le64(3);
2368 inode_item->nlink = cpu_to_le32(1);
2369 inode_item->nblocks = cpu_to_le64(1);
2370 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
2372 btrfs_set_root_bytenr(&root_item, leaf->start);
2373 btrfs_set_root_level(&root_item, 0);
2374 btrfs_set_root_refs(&root_item, 1);
2375 btrfs_set_root_used(&root_item, 0);
2377 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
2378 root_item.drop_level = 0;
2380 free_extent_buffer(leaf);
2383 btrfs_set_root_dirid(&root_item, new_dirid);
2385 key.objectid = objectid;
2387 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2388 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2394 * insert the directory item
2396 key.offset = (u64)-1;
2397 dir = root->fs_info->sb->s_root->d_inode;
2398 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2399 name, namelen, dir->i_ino, &key,
2404 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2405 name, namelen, objectid,
2406 root->fs_info->sb->s_root->d_inode->i_ino);
2410 ret = btrfs_commit_transaction(trans, root);
2414 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
2417 trans = btrfs_start_transaction(new_root, 1);
2420 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2422 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2425 inode->i_op = &btrfs_dir_inode_operations;
2426 inode->i_fop = &btrfs_dir_file_operations;
2427 new_root->inode = inode;
2429 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2433 ret = btrfs_update_inode(trans, new_root, inode);
2437 nr = trans->blocks_used;
2438 err = btrfs_commit_transaction(trans, new_root);
2442 mutex_unlock(&root->fs_info->fs_mutex);
2443 btrfs_btree_balance_dirty(root, nr);
2444 btrfs_throttle(root);
2448 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2450 struct btrfs_pending_snapshot *pending_snapshot;
2451 struct btrfs_trans_handle *trans;
2454 unsigned long nr = 0;
2456 if (!root->ref_cows)
2459 mutex_lock(&root->fs_info->fs_mutex);
2460 ret = btrfs_check_free_space(root, 1, 0);
2464 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
2465 if (!pending_snapshot) {
2469 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
2470 if (!pending_snapshot->name) {
2472 kfree(pending_snapshot);
2475 memcpy(pending_snapshot->name, name, namelen);
2476 pending_snapshot->name[namelen] = '\0';
2477 trans = btrfs_start_transaction(root, 1);
2479 pending_snapshot->root = root;
2480 list_add(&pending_snapshot->list,
2481 &trans->transaction->pending_snapshots);
2482 ret = btrfs_update_inode(trans, root, root->inode);
2483 err = btrfs_commit_transaction(trans, root);
2486 mutex_unlock(&root->fs_info->fs_mutex);
2487 btrfs_btree_balance_dirty(root, nr);
2488 btrfs_throttle(root);
2492 unsigned long btrfs_force_ra(struct address_space *mapping,
2493 struct file_ra_state *ra, struct file *file,
2494 pgoff_t offset, pgoff_t last_index)
2498 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2499 req_size = last_index - offset + 1;
2500 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2503 req_size = min(last_index - offset + 1, (pgoff_t)128);
2504 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2505 return offset + req_size;
2509 int btrfs_defrag_file(struct file *file) {
2510 struct inode *inode = fdentry(file)->d_inode;
2511 struct btrfs_root *root = BTRFS_I(inode)->root;
2512 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2514 unsigned long last_index;
2515 unsigned long ra_index = 0;
2521 mutex_lock(&root->fs_info->fs_mutex);
2522 ret = btrfs_check_free_space(root, inode->i_size, 0);
2523 mutex_unlock(&root->fs_info->fs_mutex);
2527 mutex_lock(&inode->i_mutex);
2528 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2529 for (i = 0; i <= last_index; i++) {
2530 if (i == ra_index) {
2531 ra_index = btrfs_force_ra(inode->i_mapping,
2533 file, ra_index, last_index);
2535 page = grab_cache_page(inode->i_mapping, i);
2538 if (!PageUptodate(page)) {
2539 btrfs_readpage(NULL, page);
2541 if (!PageUptodate(page)) {
2543 page_cache_release(page);
2547 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2548 page_end = page_start + PAGE_CACHE_SIZE - 1;
2550 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2551 set_extent_delalloc(io_tree, page_start,
2552 page_end, GFP_NOFS);
2554 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2555 set_page_dirty(page);
2557 page_cache_release(page);
2558 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2562 mutex_unlock(&inode->i_mutex);
2566 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
2570 struct btrfs_ioctl_vol_args *vol_args;
2571 struct btrfs_trans_handle *trans;
2577 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2582 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2586 namelen = strlen(vol_args->name);
2587 if (namelen > BTRFS_VOL_NAME_MAX) {
2592 sizestr = vol_args->name;
2593 if (!strcmp(sizestr, "max"))
2594 new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
2596 if (sizestr[0] == '-') {
2599 } else if (sizestr[0] == '+') {
2603 new_size = btrfs_parse_size(sizestr);
2604 if (new_size == 0) {
2610 mutex_lock(&root->fs_info->fs_mutex);
2611 old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
2614 if (new_size > old_size) {
2618 new_size = old_size - new_size;
2619 } else if (mod > 0) {
2620 new_size = old_size + new_size;
2623 if (new_size < 256 * 1024 * 1024) {
2627 if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
2632 do_div(new_size, root->sectorsize);
2633 new_size *= root->sectorsize;
2635 printk("new size is %Lu\n", new_size);
2636 if (new_size > old_size) {
2637 trans = btrfs_start_transaction(root, 1);
2638 ret = btrfs_grow_extent_tree(trans, root, new_size);
2639 btrfs_commit_transaction(trans, root);
2641 ret = btrfs_shrink_extent_tree(root, new_size);
2645 mutex_unlock(&root->fs_info->fs_mutex);
2651 static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root,
2654 struct btrfs_ioctl_vol_args *vol_args;
2655 struct btrfs_dir_item *di;
2656 struct btrfs_path *path;
2661 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2666 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2671 namelen = strlen(vol_args->name);
2672 if (namelen > BTRFS_VOL_NAME_MAX) {
2676 if (strchr(vol_args->name, '/')) {
2681 path = btrfs_alloc_path();
2687 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2688 mutex_lock(&root->fs_info->fs_mutex);
2689 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2691 vol_args->name, namelen, 0);
2692 mutex_unlock(&root->fs_info->fs_mutex);
2693 btrfs_free_path(path);
2695 if (di && !IS_ERR(di)) {
2705 if (root == root->fs_info->tree_root)
2706 ret = create_subvol(root, vol_args->name, namelen);
2708 ret = create_snapshot(root, vol_args->name, namelen);
2714 static int btrfs_ioctl_defrag(struct file *file)
2716 struct inode *inode = fdentry(file)->d_inode;
2717 struct btrfs_root *root = BTRFS_I(inode)->root;
2719 switch (inode->i_mode & S_IFMT) {
2721 mutex_lock(&root->fs_info->fs_mutex);
2722 btrfs_defrag_root(root, 0);
2723 btrfs_defrag_root(root->fs_info->extent_root, 0);
2724 mutex_unlock(&root->fs_info->fs_mutex);
2727 btrfs_defrag_file(file);
2734 long btrfs_ioctl(struct file *file, unsigned int
2735 cmd, unsigned long arg)
2737 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2740 case BTRFS_IOC_SNAP_CREATE:
2741 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2742 case BTRFS_IOC_DEFRAG:
2743 return btrfs_ioctl_defrag(file);
2744 case BTRFS_IOC_RESIZE:
2745 return btrfs_ioctl_resize(root, (void __user *)arg);
2752 * Called inside transaction, so use GFP_NOFS
2754 struct inode *btrfs_alloc_inode(struct super_block *sb)
2756 struct btrfs_inode *ei;
2758 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2762 ei->ordered_trans = 0;
2763 return &ei->vfs_inode;
2766 void btrfs_destroy_inode(struct inode *inode)
2768 WARN_ON(!list_empty(&inode->i_dentry));
2769 WARN_ON(inode->i_data.nrpages);
2771 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2772 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2775 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2776 static void init_once(struct kmem_cache * cachep, void *foo)
2778 static void init_once(void * foo, struct kmem_cache * cachep,
2779 unsigned long flags)
2782 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2784 inode_init_once(&ei->vfs_inode);
2787 void btrfs_destroy_cachep(void)
2789 if (btrfs_inode_cachep)
2790 kmem_cache_destroy(btrfs_inode_cachep);
2791 if (btrfs_trans_handle_cachep)
2792 kmem_cache_destroy(btrfs_trans_handle_cachep);
2793 if (btrfs_transaction_cachep)
2794 kmem_cache_destroy(btrfs_transaction_cachep);
2795 if (btrfs_bit_radix_cachep)
2796 kmem_cache_destroy(btrfs_bit_radix_cachep);
2797 if (btrfs_path_cachep)
2798 kmem_cache_destroy(btrfs_path_cachep);
2801 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2802 unsigned long extra_flags,
2803 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2804 void (*ctor)(struct kmem_cache *, void *)
2806 void (*ctor)(void *, struct kmem_cache *,
2811 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2812 SLAB_MEM_SPREAD | extra_flags), ctor
2813 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2819 int btrfs_init_cachep(void)
2821 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2822 sizeof(struct btrfs_inode),
2824 if (!btrfs_inode_cachep)
2826 btrfs_trans_handle_cachep =
2827 btrfs_cache_create("btrfs_trans_handle_cache",
2828 sizeof(struct btrfs_trans_handle),
2830 if (!btrfs_trans_handle_cachep)
2832 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2833 sizeof(struct btrfs_transaction),
2835 if (!btrfs_transaction_cachep)
2837 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2838 sizeof(struct btrfs_path),
2840 if (!btrfs_path_cachep)
2842 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2843 SLAB_DESTROY_BY_RCU, NULL);
2844 if (!btrfs_bit_radix_cachep)
2848 btrfs_destroy_cachep();
2852 static int btrfs_getattr(struct vfsmount *mnt,
2853 struct dentry *dentry, struct kstat *stat)
2855 struct inode *inode = dentry->d_inode;
2856 generic_fillattr(inode, stat);
2857 stat->blksize = PAGE_CACHE_SIZE;
2858 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
2862 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2863 struct inode * new_dir,struct dentry *new_dentry)
2865 struct btrfs_trans_handle *trans;
2866 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2867 struct inode *new_inode = new_dentry->d_inode;
2868 struct inode *old_inode = old_dentry->d_inode;
2869 struct timespec ctime = CURRENT_TIME;
2870 struct btrfs_path *path;
2873 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2874 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2878 mutex_lock(&root->fs_info->fs_mutex);
2879 ret = btrfs_check_free_space(root, 1, 0);
2883 trans = btrfs_start_transaction(root, 1);
2885 btrfs_set_trans_block_group(trans, new_dir);
2886 path = btrfs_alloc_path();
2892 old_dentry->d_inode->i_nlink++;
2893 old_dir->i_ctime = old_dir->i_mtime = ctime;
2894 new_dir->i_ctime = new_dir->i_mtime = ctime;
2895 old_inode->i_ctime = ctime;
2897 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2902 new_inode->i_ctime = CURRENT_TIME;
2903 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2907 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
2912 btrfs_free_path(path);
2913 btrfs_end_transaction(trans, root);
2915 mutex_unlock(&root->fs_info->fs_mutex);
2919 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2920 const char *symname)
2922 struct btrfs_trans_handle *trans;
2923 struct btrfs_root *root = BTRFS_I(dir)->root;
2924 struct btrfs_path *path;
2925 struct btrfs_key key;
2926 struct inode *inode = NULL;
2933 struct btrfs_file_extent_item *ei;
2934 struct extent_buffer *leaf;
2935 unsigned long nr = 0;
2937 name_len = strlen(symname) + 1;
2938 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2939 return -ENAMETOOLONG;
2941 mutex_lock(&root->fs_info->fs_mutex);
2942 err = btrfs_check_free_space(root, 1, 0);
2946 trans = btrfs_start_transaction(root, 1);
2947 btrfs_set_trans_block_group(trans, dir);
2949 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2955 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2957 dentry->d_parent->d_inode->i_ino, objectid,
2958 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2959 err = PTR_ERR(inode);
2963 btrfs_set_trans_block_group(trans, inode);
2964 err = btrfs_add_nondir(trans, dentry, inode, 0);
2968 inode->i_mapping->a_ops = &btrfs_aops;
2969 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
2970 inode->i_fop = &btrfs_file_operations;
2971 inode->i_op = &btrfs_file_inode_operations;
2972 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2973 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2974 inode->i_mapping, GFP_NOFS);
2975 BTRFS_I(inode)->delalloc_bytes = 0;
2976 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2978 dir->i_sb->s_dirt = 1;
2979 btrfs_update_inode_block_group(trans, inode);
2980 btrfs_update_inode_block_group(trans, dir);
2984 path = btrfs_alloc_path();
2986 key.objectid = inode->i_ino;
2988 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2989 datasize = btrfs_file_extent_calc_inline_size(name_len);
2990 err = btrfs_insert_empty_item(trans, root, path, &key,
2996 leaf = path->nodes[0];
2997 ei = btrfs_item_ptr(leaf, path->slots[0],
2998 struct btrfs_file_extent_item);
2999 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
3000 btrfs_set_file_extent_type(leaf, ei,
3001 BTRFS_FILE_EXTENT_INLINE);
3002 ptr = btrfs_file_extent_inline_start(ei);
3003 write_extent_buffer(leaf, symname, ptr, name_len);
3004 btrfs_mark_buffer_dirty(leaf);
3005 btrfs_free_path(path);
3007 inode->i_op = &btrfs_symlink_inode_operations;
3008 inode->i_mapping->a_ops = &btrfs_symlink_aops;
3009 inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3010 inode->i_size = name_len - 1;
3011 err = btrfs_update_inode(trans, root, inode);
3016 nr = trans->blocks_used;
3017 btrfs_end_transaction(trans, root);
3019 mutex_unlock(&root->fs_info->fs_mutex);
3021 inode_dec_link_count(inode);
3024 btrfs_btree_balance_dirty(root, nr);
3025 btrfs_throttle(root);
3028 static int btrfs_permission(struct inode *inode, int mask,
3029 struct nameidata *nd)
3031 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
3033 return generic_permission(inode, mask, NULL);
3036 static struct inode_operations btrfs_dir_inode_operations = {
3037 .lookup = btrfs_lookup,
3038 .create = btrfs_create,
3039 .unlink = btrfs_unlink,
3041 .mkdir = btrfs_mkdir,
3042 .rmdir = btrfs_rmdir,
3043 .rename = btrfs_rename,
3044 .symlink = btrfs_symlink,
3045 .setattr = btrfs_setattr,
3046 .mknod = btrfs_mknod,
3047 .setxattr = generic_setxattr,
3048 .getxattr = generic_getxattr,
3049 .listxattr = btrfs_listxattr,
3050 .removexattr = generic_removexattr,
3051 .permission = btrfs_permission,
3053 static struct inode_operations btrfs_dir_ro_inode_operations = {
3054 .lookup = btrfs_lookup,
3055 .permission = btrfs_permission,
3057 static struct file_operations btrfs_dir_file_operations = {
3058 .llseek = generic_file_llseek,
3059 .read = generic_read_dir,
3060 .readdir = btrfs_readdir,
3061 .unlocked_ioctl = btrfs_ioctl,
3062 #ifdef CONFIG_COMPAT
3063 .compat_ioctl = btrfs_ioctl,
3067 static struct extent_io_ops btrfs_extent_io_ops = {
3068 .fill_delalloc = run_delalloc_range,
3069 .submit_bio_hook = btrfs_submit_bio_hook,
3070 .merge_bio_hook = btrfs_merge_bio_hook,
3071 .readpage_io_hook = btrfs_readpage_io_hook,
3072 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3073 .set_bit_hook = btrfs_set_bit_hook,
3074 .clear_bit_hook = btrfs_clear_bit_hook,
3077 static struct address_space_operations btrfs_aops = {
3078 .readpage = btrfs_readpage,
3079 .writepage = btrfs_writepage,
3080 .writepages = btrfs_writepages,
3081 .readpages = btrfs_readpages,
3082 .sync_page = block_sync_page,
3084 .invalidatepage = btrfs_invalidatepage,
3085 .releasepage = btrfs_releasepage,
3086 .set_page_dirty = __set_page_dirty_nobuffers,
3089 static struct address_space_operations btrfs_symlink_aops = {
3090 .readpage = btrfs_readpage,
3091 .writepage = btrfs_writepage,
3092 .invalidatepage = btrfs_invalidatepage,
3093 .releasepage = btrfs_releasepage,
3096 static struct inode_operations btrfs_file_inode_operations = {
3097 .truncate = btrfs_truncate,
3098 .getattr = btrfs_getattr,
3099 .setattr = btrfs_setattr,
3100 .setxattr = generic_setxattr,
3101 .getxattr = generic_getxattr,
3102 .listxattr = btrfs_listxattr,
3103 .removexattr = generic_removexattr,
3104 .permission = btrfs_permission,
3106 static struct inode_operations btrfs_special_inode_operations = {
3107 .getattr = btrfs_getattr,
3108 .setattr = btrfs_setattr,
3109 .permission = btrfs_permission,
3111 static struct inode_operations btrfs_symlink_inode_operations = {
3112 .readlink = generic_readlink,
3113 .follow_link = page_follow_link_light,
3114 .put_link = page_put_link,
3115 .permission = btrfs_permission,