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_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
301 struct btrfs_root *root = BTRFS_I(inode)->root;
302 struct btrfs_trans_handle *trans;
309 if (btrfs_test_opt(root, NODATASUM) ||
310 btrfs_test_flag(inode, NODATASUM)) {
314 mutex_lock(&root->fs_info->fs_mutex);
315 trans = btrfs_start_transaction(root, 1);
316 btrfs_set_trans_block_group(trans, inode);
317 btrfs_csum_file_blocks(trans, root, inode, bio);
318 ret = btrfs_end_transaction(trans, root);
320 mutex_unlock(&root->fs_info->fs_mutex);
322 return btrfs_map_bio(root, rw, bio);
325 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
328 struct inode *inode = page->mapping->host;
329 struct btrfs_root *root = BTRFS_I(inode)->root;
330 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
331 struct btrfs_csum_item *item;
332 struct btrfs_path *path = NULL;
334 if (btrfs_test_opt(root, NODATASUM) ||
335 btrfs_test_flag(inode, NODATASUM))
337 mutex_lock(&root->fs_info->fs_mutex);
338 path = btrfs_alloc_path();
339 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
342 /* a csum that isn't present is a preallocated region. */
343 if (ret == -ENOENT || ret == -EFBIG)
346 printk("no csum found for inode %lu start %Lu\n", inode->i_ino, start);
349 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
351 set_state_private(io_tree, start, csum);
354 btrfs_free_path(path);
355 mutex_unlock(&root->fs_info->fs_mutex);
359 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
360 struct extent_state *state)
362 size_t offset = start - ((u64)page->index << PAGE_CACHE_SHIFT);
363 struct inode *inode = page->mapping->host;
364 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
366 u64 private = ~(u32)0;
368 struct btrfs_root *root = BTRFS_I(inode)->root;
372 if (btrfs_test_opt(root, NODATASUM) ||
373 btrfs_test_flag(inode, NODATASUM))
375 if (state && state->start == start) {
376 private = state->private;
379 ret = get_state_private(io_tree, start, &private);
381 local_irq_save(flags);
382 kaddr = kmap_atomic(page, KM_IRQ0);
386 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
387 btrfs_csum_final(csum, (char *)&csum);
388 if (csum != private) {
391 kunmap_atomic(kaddr, KM_IRQ0);
392 local_irq_restore(flags);
396 printk("btrfs csum failed ino %lu off %llu csum %u private %Lu\n",
397 page->mapping->host->i_ino, (unsigned long long)start, csum,
399 memset(kaddr + offset, 1, end - start + 1);
400 flush_dcache_page(page);
401 kunmap_atomic(kaddr, KM_IRQ0);
402 local_irq_restore(flags);
406 void btrfs_read_locked_inode(struct inode *inode)
408 struct btrfs_path *path;
409 struct extent_buffer *leaf;
410 struct btrfs_inode_item *inode_item;
411 struct btrfs_timespec *tspec;
412 struct btrfs_root *root = BTRFS_I(inode)->root;
413 struct btrfs_key location;
414 u64 alloc_group_block;
418 path = btrfs_alloc_path();
420 mutex_lock(&root->fs_info->fs_mutex);
421 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
423 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
427 leaf = path->nodes[0];
428 inode_item = btrfs_item_ptr(leaf, path->slots[0],
429 struct btrfs_inode_item);
431 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
432 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
433 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
434 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
435 inode->i_size = btrfs_inode_size(leaf, inode_item);
437 tspec = btrfs_inode_atime(inode_item);
438 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
439 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
441 tspec = btrfs_inode_mtime(inode_item);
442 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
443 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
445 tspec = btrfs_inode_ctime(inode_item);
446 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
447 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
449 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
450 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
452 rdev = btrfs_inode_rdev(leaf, inode_item);
454 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
455 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
457 BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
458 if (!BTRFS_I(inode)->block_group) {
459 BTRFS_I(inode)->block_group = btrfs_find_block_group(root,
461 BTRFS_BLOCK_GROUP_METADATA, 0);
463 btrfs_free_path(path);
466 mutex_unlock(&root->fs_info->fs_mutex);
468 switch (inode->i_mode & S_IFMT) {
470 inode->i_mapping->a_ops = &btrfs_aops;
471 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
472 inode->i_fop = &btrfs_file_operations;
473 inode->i_op = &btrfs_file_inode_operations;
476 inode->i_fop = &btrfs_dir_file_operations;
477 if (root == root->fs_info->tree_root)
478 inode->i_op = &btrfs_dir_ro_inode_operations;
480 inode->i_op = &btrfs_dir_inode_operations;
483 inode->i_op = &btrfs_symlink_inode_operations;
484 inode->i_mapping->a_ops = &btrfs_symlink_aops;
487 init_special_inode(inode, inode->i_mode, rdev);
493 btrfs_release_path(root, path);
494 btrfs_free_path(path);
495 mutex_unlock(&root->fs_info->fs_mutex);
496 make_bad_inode(inode);
499 static void fill_inode_item(struct extent_buffer *leaf,
500 struct btrfs_inode_item *item,
503 btrfs_set_inode_uid(leaf, item, inode->i_uid);
504 btrfs_set_inode_gid(leaf, item, inode->i_gid);
505 btrfs_set_inode_size(leaf, item, inode->i_size);
506 btrfs_set_inode_mode(leaf, item, inode->i_mode);
507 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
509 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
510 inode->i_atime.tv_sec);
511 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
512 inode->i_atime.tv_nsec);
514 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
515 inode->i_mtime.tv_sec);
516 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
517 inode->i_mtime.tv_nsec);
519 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
520 inode->i_ctime.tv_sec);
521 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
522 inode->i_ctime.tv_nsec);
524 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
525 btrfs_set_inode_generation(leaf, item, inode->i_generation);
526 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
527 btrfs_set_inode_flags(leaf, item, BTRFS_I(inode)->flags);
528 btrfs_set_inode_block_group(leaf, item,
529 BTRFS_I(inode)->block_group->key.objectid);
532 int btrfs_update_inode(struct btrfs_trans_handle *trans,
533 struct btrfs_root *root,
536 struct btrfs_inode_item *inode_item;
537 struct btrfs_path *path;
538 struct extent_buffer *leaf;
541 path = btrfs_alloc_path();
543 ret = btrfs_lookup_inode(trans, root, path,
544 &BTRFS_I(inode)->location, 1);
551 leaf = path->nodes[0];
552 inode_item = btrfs_item_ptr(leaf, path->slots[0],
553 struct btrfs_inode_item);
555 fill_inode_item(leaf, inode_item, inode);
556 btrfs_mark_buffer_dirty(leaf);
557 btrfs_set_inode_last_trans(trans, inode);
560 btrfs_release_path(root, path);
561 btrfs_free_path(path);
566 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
567 struct btrfs_root *root,
569 struct dentry *dentry)
571 struct btrfs_path *path;
572 const char *name = dentry->d_name.name;
573 int name_len = dentry->d_name.len;
575 struct extent_buffer *leaf;
576 struct btrfs_dir_item *di;
577 struct btrfs_key key;
579 path = btrfs_alloc_path();
585 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
595 leaf = path->nodes[0];
596 btrfs_dir_item_key_to_cpu(leaf, di, &key);
597 ret = btrfs_delete_one_dir_name(trans, root, path, di);
600 btrfs_release_path(root, path);
602 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
603 key.objectid, name, name_len, -1);
612 ret = btrfs_delete_one_dir_name(trans, root, path, di);
614 dentry->d_inode->i_ctime = dir->i_ctime;
615 ret = btrfs_del_inode_ref(trans, root, name, name_len,
616 dentry->d_inode->i_ino,
617 dentry->d_parent->d_inode->i_ino);
619 printk("failed to delete reference to %.*s, "
620 "inode %lu parent %lu\n", name_len, name,
621 dentry->d_inode->i_ino,
622 dentry->d_parent->d_inode->i_ino);
625 btrfs_free_path(path);
627 dir->i_size -= name_len * 2;
628 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
629 btrfs_update_inode(trans, root, dir);
630 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
631 dentry->d_inode->i_nlink--;
633 drop_nlink(dentry->d_inode);
635 ret = btrfs_update_inode(trans, root, dentry->d_inode);
636 dir->i_sb->s_dirt = 1;
641 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
643 struct btrfs_root *root;
644 struct btrfs_trans_handle *trans;
645 struct inode *inode = dentry->d_inode;
647 unsigned long nr = 0;
649 root = BTRFS_I(dir)->root;
650 mutex_lock(&root->fs_info->fs_mutex);
652 ret = btrfs_check_free_space(root, 1, 1);
656 trans = btrfs_start_transaction(root, 1);
658 btrfs_set_trans_block_group(trans, dir);
659 ret = btrfs_unlink_trans(trans, root, dir, dentry);
660 nr = trans->blocks_used;
662 if (inode->i_nlink == 0) {
664 /* if the inode isn't linked anywhere,
665 * we don't need to worry about
668 found = btrfs_del_ordered_inode(inode);
670 atomic_dec(&inode->i_count);
674 btrfs_end_transaction(trans, root);
676 mutex_unlock(&root->fs_info->fs_mutex);
677 btrfs_btree_balance_dirty(root, nr);
678 btrfs_throttle(root);
682 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
684 struct inode *inode = dentry->d_inode;
687 struct btrfs_root *root = BTRFS_I(dir)->root;
688 struct btrfs_trans_handle *trans;
689 unsigned long nr = 0;
691 if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
694 mutex_lock(&root->fs_info->fs_mutex);
695 ret = btrfs_check_free_space(root, 1, 1);
699 trans = btrfs_start_transaction(root, 1);
700 btrfs_set_trans_block_group(trans, dir);
702 /* now the directory is empty */
703 err = btrfs_unlink_trans(trans, root, dir, dentry);
708 nr = trans->blocks_used;
709 ret = btrfs_end_transaction(trans, root);
711 mutex_unlock(&root->fs_info->fs_mutex);
712 btrfs_btree_balance_dirty(root, nr);
713 btrfs_throttle(root);
721 * this can truncate away extent items, csum items and directory items.
722 * It starts at a high offset and removes keys until it can't find
723 * any higher than i_size.
725 * csum items that cross the new i_size are truncated to the new size
728 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
729 struct btrfs_root *root,
734 struct btrfs_path *path;
735 struct btrfs_key key;
736 struct btrfs_key found_key;
738 struct extent_buffer *leaf;
739 struct btrfs_file_extent_item *fi;
740 u64 extent_start = 0;
741 u64 extent_num_bytes = 0;
747 int pending_del_nr = 0;
748 int pending_del_slot = 0;
749 int extent_type = -1;
751 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
752 path = btrfs_alloc_path();
756 /* FIXME, add redo link to tree so we don't leak on crash */
757 key.objectid = inode->i_ino;
758 key.offset = (u64)-1;
761 btrfs_init_path(path);
763 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
768 BUG_ON(path->slots[0] == 0);
774 leaf = path->nodes[0];
775 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
776 found_type = btrfs_key_type(&found_key);
778 if (found_key.objectid != inode->i_ino)
781 if (found_type < min_type)
784 item_end = found_key.offset;
785 if (found_type == BTRFS_EXTENT_DATA_KEY) {
786 fi = btrfs_item_ptr(leaf, path->slots[0],
787 struct btrfs_file_extent_item);
788 extent_type = btrfs_file_extent_type(leaf, fi);
789 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
791 btrfs_file_extent_num_bytes(leaf, fi);
792 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
793 struct btrfs_item *item = btrfs_item_nr(leaf,
795 item_end += btrfs_file_extent_inline_len(leaf,
800 if (found_type == BTRFS_CSUM_ITEM_KEY) {
801 ret = btrfs_csum_truncate(trans, root, path,
805 if (item_end < inode->i_size) {
806 if (found_type == BTRFS_DIR_ITEM_KEY) {
807 found_type = BTRFS_INODE_ITEM_KEY;
808 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
809 found_type = BTRFS_CSUM_ITEM_KEY;
810 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
811 found_type = BTRFS_XATTR_ITEM_KEY;
812 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
813 found_type = BTRFS_INODE_REF_KEY;
814 } else if (found_type) {
819 btrfs_set_key_type(&key, found_type);
822 if (found_key.offset >= inode->i_size)
828 /* FIXME, shrink the extent if the ref count is only 1 */
829 if (found_type != BTRFS_EXTENT_DATA_KEY)
832 if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
834 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
837 btrfs_file_extent_num_bytes(leaf, fi);
838 extent_num_bytes = inode->i_size -
839 found_key.offset + root->sectorsize - 1;
840 extent_num_bytes = extent_num_bytes &
841 ~((u64)root->sectorsize - 1);
842 btrfs_set_file_extent_num_bytes(leaf, fi,
844 num_dec = (orig_num_bytes -
846 if (extent_start != 0)
847 dec_i_blocks(inode, num_dec);
848 btrfs_mark_buffer_dirty(leaf);
851 btrfs_file_extent_disk_num_bytes(leaf,
853 /* FIXME blocksize != 4096 */
854 num_dec = btrfs_file_extent_num_bytes(leaf, fi);
855 if (extent_start != 0) {
857 dec_i_blocks(inode, num_dec);
859 root_gen = btrfs_header_generation(leaf);
860 root_owner = btrfs_header_owner(leaf);
862 } else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
864 u32 newsize = inode->i_size - found_key.offset;
865 dec_i_blocks(inode, item_end + 1 -
866 found_key.offset - newsize);
868 btrfs_file_extent_calc_inline_size(newsize);
869 ret = btrfs_truncate_item(trans, root, path,
873 dec_i_blocks(inode, item_end + 1 -
879 if (!pending_del_nr) {
880 /* no pending yet, add ourselves */
881 pending_del_slot = path->slots[0];
883 } else if (pending_del_nr &&
884 path->slots[0] + 1 == pending_del_slot) {
885 /* hop on the pending chunk */
887 pending_del_slot = path->slots[0];
889 printk("bad pending slot %d pending_del_nr %d pending_del_slot %d\n", path->slots[0], pending_del_nr, pending_del_slot);
895 ret = btrfs_free_extent(trans, root, extent_start,
898 root_gen, inode->i_ino,
899 found_key.offset, 0);
903 if (path->slots[0] == 0) {
906 btrfs_release_path(root, path);
911 if (pending_del_nr &&
912 path->slots[0] + 1 != pending_del_slot) {
913 struct btrfs_key debug;
915 btrfs_item_key_to_cpu(path->nodes[0], &debug,
917 ret = btrfs_del_items(trans, root, path,
922 btrfs_release_path(root, path);
928 if (pending_del_nr) {
929 ret = btrfs_del_items(trans, root, path, pending_del_slot,
932 btrfs_release_path(root, path);
933 btrfs_free_path(path);
934 inode->i_sb->s_dirt = 1;
938 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
942 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
943 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
944 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
947 WARN_ON(!PageLocked(page));
948 set_page_extent_mapped(page);
950 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
951 set_extent_delalloc(&BTRFS_I(inode)->io_tree, page_start,
954 if (zero_start != PAGE_CACHE_SIZE) {
956 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
957 flush_dcache_page(page);
960 set_page_dirty(page);
961 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
967 * taken from block_truncate_page, but does cow as it zeros out
968 * any bytes left in the last page in the file.
970 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
972 struct inode *inode = mapping->host;
973 struct btrfs_root *root = BTRFS_I(inode)->root;
974 u32 blocksize = root->sectorsize;
975 pgoff_t index = from >> PAGE_CACHE_SHIFT;
976 unsigned offset = from & (PAGE_CACHE_SIZE-1);
981 if ((offset & (blocksize - 1)) == 0)
985 page = grab_cache_page(mapping, index);
988 if (!PageUptodate(page)) {
989 ret = btrfs_readpage(NULL, page);
991 if (!PageUptodate(page)) {
996 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
998 ret = btrfs_cow_one_page(inode, page, offset);
1001 page_cache_release(page);
1006 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
1008 struct inode *inode = dentry->d_inode;
1011 err = inode_change_ok(inode, attr);
1015 if (S_ISREG(inode->i_mode) &&
1016 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
1017 struct btrfs_trans_handle *trans;
1018 struct btrfs_root *root = BTRFS_I(inode)->root;
1019 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1021 u64 mask = root->sectorsize - 1;
1022 u64 hole_start = (inode->i_size + mask) & ~mask;
1023 u64 block_end = (attr->ia_size + mask) & ~mask;
1027 if (attr->ia_size <= hole_start)
1030 mutex_lock(&root->fs_info->fs_mutex);
1031 err = btrfs_check_free_space(root, 1, 0);
1032 mutex_unlock(&root->fs_info->fs_mutex);
1036 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1038 lock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1039 hole_size = block_end - hole_start;
1041 mutex_lock(&root->fs_info->fs_mutex);
1042 trans = btrfs_start_transaction(root, 1);
1043 btrfs_set_trans_block_group(trans, inode);
1044 err = btrfs_drop_extents(trans, root, inode,
1045 hole_start, block_end, hole_start,
1048 if (alloc_hint != EXTENT_MAP_INLINE) {
1049 err = btrfs_insert_file_extent(trans, root,
1053 btrfs_drop_extent_cache(inode, hole_start,
1055 btrfs_check_file(root, inode);
1057 btrfs_end_transaction(trans, root);
1058 mutex_unlock(&root->fs_info->fs_mutex);
1059 unlock_extent(io_tree, hole_start, block_end - 1, GFP_NOFS);
1064 err = inode_setattr(inode, attr);
1069 void btrfs_put_inode(struct inode *inode)
1073 if (!BTRFS_I(inode)->ordered_trans) {
1077 if (mapping_tagged(inode->i_mapping, PAGECACHE_TAG_DIRTY) ||
1078 mapping_tagged(inode->i_mapping, PAGECACHE_TAG_WRITEBACK))
1081 ret = btrfs_del_ordered_inode(inode);
1083 atomic_dec(&inode->i_count);
1087 void btrfs_delete_inode(struct inode *inode)
1089 struct btrfs_trans_handle *trans;
1090 struct btrfs_root *root = BTRFS_I(inode)->root;
1094 truncate_inode_pages(&inode->i_data, 0);
1095 if (is_bad_inode(inode)) {
1100 mutex_lock(&root->fs_info->fs_mutex);
1101 trans = btrfs_start_transaction(root, 1);
1103 btrfs_set_trans_block_group(trans, inode);
1104 ret = btrfs_truncate_in_trans(trans, root, inode, 0);
1106 goto no_delete_lock;
1108 nr = trans->blocks_used;
1111 btrfs_end_transaction(trans, root);
1112 mutex_unlock(&root->fs_info->fs_mutex);
1113 btrfs_btree_balance_dirty(root, nr);
1114 btrfs_throttle(root);
1118 nr = trans->blocks_used;
1119 btrfs_end_transaction(trans, root);
1120 mutex_unlock(&root->fs_info->fs_mutex);
1121 btrfs_btree_balance_dirty(root, nr);
1122 btrfs_throttle(root);
1128 * this returns the key found in the dir entry in the location pointer.
1129 * If no dir entries were found, location->objectid is 0.
1131 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
1132 struct btrfs_key *location)
1134 const char *name = dentry->d_name.name;
1135 int namelen = dentry->d_name.len;
1136 struct btrfs_dir_item *di;
1137 struct btrfs_path *path;
1138 struct btrfs_root *root = BTRFS_I(dir)->root;
1141 if (namelen == 1 && strcmp(name, ".") == 0) {
1142 location->objectid = dir->i_ino;
1143 location->type = BTRFS_INODE_ITEM_KEY;
1144 location->offset = 0;
1147 path = btrfs_alloc_path();
1150 if (namelen == 2 && strcmp(name, "..") == 0) {
1151 struct btrfs_key key;
1152 struct extent_buffer *leaf;
1156 key.objectid = dir->i_ino;
1157 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1159 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1163 leaf = path->nodes[0];
1164 slot = path->slots[0];
1165 nritems = btrfs_header_nritems(leaf);
1166 if (slot >= nritems)
1169 btrfs_item_key_to_cpu(leaf, &key, slot);
1170 if (key.objectid != dir->i_ino ||
1171 key.type != BTRFS_INODE_REF_KEY) {
1174 location->objectid = key.offset;
1175 location->type = BTRFS_INODE_ITEM_KEY;
1176 location->offset = 0;
1180 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
1184 if (!di || IS_ERR(di)) {
1187 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
1189 btrfs_free_path(path);
1192 location->objectid = 0;
1197 * when we hit a tree root in a directory, the btrfs part of the inode
1198 * needs to be changed to reflect the root directory of the tree root. This
1199 * is kind of like crossing a mount point.
1201 static int fixup_tree_root_location(struct btrfs_root *root,
1202 struct btrfs_key *location,
1203 struct btrfs_root **sub_root,
1204 struct dentry *dentry)
1206 struct btrfs_path *path;
1207 struct btrfs_root_item *ri;
1209 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
1211 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
1214 path = btrfs_alloc_path();
1216 mutex_lock(&root->fs_info->fs_mutex);
1218 *sub_root = btrfs_read_fs_root(root->fs_info, location,
1219 dentry->d_name.name,
1220 dentry->d_name.len);
1221 if (IS_ERR(*sub_root))
1222 return PTR_ERR(*sub_root);
1224 ri = &(*sub_root)->root_item;
1225 location->objectid = btrfs_root_dirid(ri);
1226 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1227 location->offset = 0;
1229 btrfs_free_path(path);
1230 mutex_unlock(&root->fs_info->fs_mutex);
1234 static int btrfs_init_locked_inode(struct inode *inode, void *p)
1236 struct btrfs_iget_args *args = p;
1237 inode->i_ino = args->ino;
1238 BTRFS_I(inode)->root = args->root;
1239 BTRFS_I(inode)->delalloc_bytes = 0;
1240 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1241 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1242 inode->i_mapping, GFP_NOFS);
1246 static int btrfs_find_actor(struct inode *inode, void *opaque)
1248 struct btrfs_iget_args *args = opaque;
1249 return (args->ino == inode->i_ino &&
1250 args->root == BTRFS_I(inode)->root);
1253 struct inode *btrfs_ilookup(struct super_block *s, u64 objectid,
1256 struct btrfs_iget_args args;
1257 args.ino = objectid;
1258 args.root = btrfs_lookup_fs_root(btrfs_sb(s)->fs_info, root_objectid);
1263 return ilookup5(s, objectid, btrfs_find_actor, (void *)&args);
1266 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
1267 struct btrfs_root *root)
1269 struct inode *inode;
1270 struct btrfs_iget_args args;
1271 args.ino = objectid;
1274 inode = iget5_locked(s, objectid, btrfs_find_actor,
1275 btrfs_init_locked_inode,
1280 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
1281 struct nameidata *nd)
1283 struct inode * inode;
1284 struct btrfs_inode *bi = BTRFS_I(dir);
1285 struct btrfs_root *root = bi->root;
1286 struct btrfs_root *sub_root = root;
1287 struct btrfs_key location;
1290 if (dentry->d_name.len > BTRFS_NAME_LEN)
1291 return ERR_PTR(-ENAMETOOLONG);
1293 mutex_lock(&root->fs_info->fs_mutex);
1294 ret = btrfs_inode_by_name(dir, dentry, &location);
1295 mutex_unlock(&root->fs_info->fs_mutex);
1298 return ERR_PTR(ret);
1301 if (location.objectid) {
1302 ret = fixup_tree_root_location(root, &location, &sub_root,
1305 return ERR_PTR(ret);
1307 return ERR_PTR(-ENOENT);
1308 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
1311 return ERR_PTR(-EACCES);
1312 if (inode->i_state & I_NEW) {
1313 /* the inode and parent dir are two different roots */
1314 if (sub_root != root) {
1316 sub_root->inode = inode;
1318 BTRFS_I(inode)->root = sub_root;
1319 memcpy(&BTRFS_I(inode)->location, &location,
1321 btrfs_read_locked_inode(inode);
1322 unlock_new_inode(inode);
1325 return d_splice_alias(inode, dentry);
1328 static unsigned char btrfs_filetype_table[] = {
1329 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
1332 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
1334 struct inode *inode = filp->f_dentry->d_inode;
1335 struct btrfs_root *root = BTRFS_I(inode)->root;
1336 struct btrfs_item *item;
1337 struct btrfs_dir_item *di;
1338 struct btrfs_key key;
1339 struct btrfs_key found_key;
1340 struct btrfs_path *path;
1343 struct extent_buffer *leaf;
1346 unsigned char d_type;
1351 int key_type = BTRFS_DIR_INDEX_KEY;
1356 /* FIXME, use a real flag for deciding about the key type */
1357 if (root->fs_info->tree_root == root)
1358 key_type = BTRFS_DIR_ITEM_KEY;
1360 /* special case for "." */
1361 if (filp->f_pos == 0) {
1362 over = filldir(dirent, ".", 1,
1370 mutex_lock(&root->fs_info->fs_mutex);
1371 key.objectid = inode->i_ino;
1372 path = btrfs_alloc_path();
1375 /* special case for .., just use the back ref */
1376 if (filp->f_pos == 1) {
1377 btrfs_set_key_type(&key, BTRFS_INODE_REF_KEY);
1379 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1381 leaf = path->nodes[0];
1382 slot = path->slots[0];
1383 nritems = btrfs_header_nritems(leaf);
1384 if (slot >= nritems) {
1385 btrfs_release_path(root, path);
1386 goto read_dir_items;
1388 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1389 btrfs_release_path(root, path);
1390 if (found_key.objectid != key.objectid ||
1391 found_key.type != BTRFS_INODE_REF_KEY)
1392 goto read_dir_items;
1393 over = filldir(dirent, "..", 2,
1394 2, found_key.offset, DT_DIR);
1401 btrfs_set_key_type(&key, key_type);
1402 key.offset = filp->f_pos;
1404 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1409 leaf = path->nodes[0];
1410 nritems = btrfs_header_nritems(leaf);
1411 slot = path->slots[0];
1412 if (advance || slot >= nritems) {
1413 if (slot >= nritems -1) {
1414 ret = btrfs_next_leaf(root, path);
1417 leaf = path->nodes[0];
1418 nritems = btrfs_header_nritems(leaf);
1419 slot = path->slots[0];
1426 item = btrfs_item_nr(leaf, slot);
1427 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1429 if (found_key.objectid != key.objectid)
1431 if (btrfs_key_type(&found_key) != key_type)
1433 if (found_key.offset < filp->f_pos)
1436 filp->f_pos = found_key.offset;
1438 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1440 di_total = btrfs_item_size(leaf, item);
1441 while(di_cur < di_total) {
1442 struct btrfs_key location;
1444 name_len = btrfs_dir_name_len(leaf, di);
1445 if (name_len < 32) {
1446 name_ptr = tmp_name;
1448 name_ptr = kmalloc(name_len, GFP_NOFS);
1451 read_extent_buffer(leaf, name_ptr,
1452 (unsigned long)(di + 1), name_len);
1454 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1455 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1456 over = filldir(dirent, name_ptr, name_len,
1461 if (name_ptr != tmp_name)
1466 di_len = btrfs_dir_name_len(leaf, di) +
1467 btrfs_dir_data_len(leaf, di) +sizeof(*di);
1469 di = (struct btrfs_dir_item *)((char *)di + di_len);
1472 if (key_type == BTRFS_DIR_INDEX_KEY)
1473 filp->f_pos = INT_LIMIT(typeof(filp->f_pos));
1479 btrfs_release_path(root, path);
1480 btrfs_free_path(path);
1481 mutex_unlock(&root->fs_info->fs_mutex);
1485 int btrfs_write_inode(struct inode *inode, int wait)
1487 struct btrfs_root *root = BTRFS_I(inode)->root;
1488 struct btrfs_trans_handle *trans;
1492 mutex_lock(&root->fs_info->fs_mutex);
1493 trans = btrfs_start_transaction(root, 1);
1494 btrfs_set_trans_block_group(trans, inode);
1495 ret = btrfs_commit_transaction(trans, root);
1496 mutex_unlock(&root->fs_info->fs_mutex);
1502 * This is somewhat expensive, updating the tree every time the
1503 * inode changes. But, it is most likely to find the inode in cache.
1504 * FIXME, needs more benchmarking...there are no reasons other than performance
1505 * to keep or drop this code.
1507 void btrfs_dirty_inode(struct inode *inode)
1509 struct btrfs_root *root = BTRFS_I(inode)->root;
1510 struct btrfs_trans_handle *trans;
1512 mutex_lock(&root->fs_info->fs_mutex);
1513 trans = btrfs_start_transaction(root, 1);
1514 btrfs_set_trans_block_group(trans, inode);
1515 btrfs_update_inode(trans, root, inode);
1516 btrfs_end_transaction(trans, root);
1517 mutex_unlock(&root->fs_info->fs_mutex);
1520 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1521 struct btrfs_root *root,
1522 const char *name, int name_len,
1525 struct btrfs_block_group_cache *group,
1528 struct inode *inode;
1529 struct btrfs_inode_item *inode_item;
1530 struct btrfs_block_group_cache *new_inode_group;
1531 struct btrfs_key *location;
1532 struct btrfs_path *path;
1533 struct btrfs_inode_ref *ref;
1534 struct btrfs_key key[2];
1540 path = btrfs_alloc_path();
1543 inode = new_inode(root->fs_info->sb);
1545 return ERR_PTR(-ENOMEM);
1547 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1548 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1549 inode->i_mapping, GFP_NOFS);
1550 BTRFS_I(inode)->delalloc_bytes = 0;
1551 BTRFS_I(inode)->root = root;
1557 new_inode_group = btrfs_find_block_group(root, group, 0,
1558 BTRFS_BLOCK_GROUP_METADATA, owner);
1559 if (!new_inode_group) {
1560 printk("find_block group failed\n");
1561 new_inode_group = group;
1563 BTRFS_I(inode)->block_group = new_inode_group;
1564 BTRFS_I(inode)->flags = 0;
1566 key[0].objectid = objectid;
1567 btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
1570 key[1].objectid = objectid;
1571 btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
1572 key[1].offset = ref_objectid;
1574 sizes[0] = sizeof(struct btrfs_inode_item);
1575 sizes[1] = name_len + sizeof(*ref);
1577 ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
1581 if (objectid > root->highest_inode)
1582 root->highest_inode = objectid;
1584 inode->i_uid = current->fsuid;
1585 inode->i_gid = current->fsgid;
1586 inode->i_mode = mode;
1587 inode->i_ino = objectid;
1588 inode->i_blocks = 0;
1589 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1590 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1591 struct btrfs_inode_item);
1592 fill_inode_item(path->nodes[0], inode_item, inode);
1594 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
1595 struct btrfs_inode_ref);
1596 btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
1597 ptr = (unsigned long)(ref + 1);
1598 write_extent_buffer(path->nodes[0], name, ptr, name_len);
1600 btrfs_mark_buffer_dirty(path->nodes[0]);
1601 btrfs_free_path(path);
1603 location = &BTRFS_I(inode)->location;
1604 location->objectid = objectid;
1605 location->offset = 0;
1606 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1608 insert_inode_hash(inode);
1611 btrfs_free_path(path);
1612 return ERR_PTR(ret);
1615 static inline u8 btrfs_inode_type(struct inode *inode)
1617 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1620 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1621 struct dentry *dentry, struct inode *inode,
1625 struct btrfs_key key;
1626 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1627 struct inode *parent_inode;
1629 key.objectid = inode->i_ino;
1630 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1633 ret = btrfs_insert_dir_item(trans, root,
1634 dentry->d_name.name, dentry->d_name.len,
1635 dentry->d_parent->d_inode->i_ino,
1636 &key, btrfs_inode_type(inode));
1639 ret = btrfs_insert_inode_ref(trans, root,
1640 dentry->d_name.name,
1643 dentry->d_parent->d_inode->i_ino);
1645 parent_inode = dentry->d_parent->d_inode;
1646 parent_inode->i_size += dentry->d_name.len * 2;
1647 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1648 ret = btrfs_update_inode(trans, root,
1649 dentry->d_parent->d_inode);
1654 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1655 struct dentry *dentry, struct inode *inode,
1658 int err = btrfs_add_link(trans, dentry, inode, backref);
1660 d_instantiate(dentry, inode);
1668 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1669 int mode, dev_t rdev)
1671 struct btrfs_trans_handle *trans;
1672 struct btrfs_root *root = BTRFS_I(dir)->root;
1673 struct inode *inode = NULL;
1677 unsigned long nr = 0;
1679 if (!new_valid_dev(rdev))
1682 mutex_lock(&root->fs_info->fs_mutex);
1683 err = btrfs_check_free_space(root, 1, 0);
1687 trans = btrfs_start_transaction(root, 1);
1688 btrfs_set_trans_block_group(trans, dir);
1690 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1696 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1698 dentry->d_parent->d_inode->i_ino, objectid,
1699 BTRFS_I(dir)->block_group, mode);
1700 err = PTR_ERR(inode);
1704 btrfs_set_trans_block_group(trans, inode);
1705 err = btrfs_add_nondir(trans, dentry, inode, 0);
1709 inode->i_op = &btrfs_special_inode_operations;
1710 init_special_inode(inode, inode->i_mode, rdev);
1711 btrfs_update_inode(trans, root, inode);
1713 dir->i_sb->s_dirt = 1;
1714 btrfs_update_inode_block_group(trans, inode);
1715 btrfs_update_inode_block_group(trans, dir);
1717 nr = trans->blocks_used;
1718 btrfs_end_transaction(trans, root);
1720 mutex_unlock(&root->fs_info->fs_mutex);
1723 inode_dec_link_count(inode);
1726 btrfs_btree_balance_dirty(root, nr);
1727 btrfs_throttle(root);
1731 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1732 int mode, struct nameidata *nd)
1734 struct btrfs_trans_handle *trans;
1735 struct btrfs_root *root = BTRFS_I(dir)->root;
1736 struct inode *inode = NULL;
1739 unsigned long nr = 0;
1742 mutex_lock(&root->fs_info->fs_mutex);
1743 err = btrfs_check_free_space(root, 1, 0);
1746 trans = btrfs_start_transaction(root, 1);
1747 btrfs_set_trans_block_group(trans, dir);
1749 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1755 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1757 dentry->d_parent->d_inode->i_ino,
1758 objectid, BTRFS_I(dir)->block_group, mode);
1759 err = PTR_ERR(inode);
1763 btrfs_set_trans_block_group(trans, inode);
1764 err = btrfs_add_nondir(trans, dentry, inode, 0);
1768 inode->i_mapping->a_ops = &btrfs_aops;
1769 inode->i_fop = &btrfs_file_operations;
1770 inode->i_op = &btrfs_file_inode_operations;
1771 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
1772 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
1773 inode->i_mapping, GFP_NOFS);
1774 BTRFS_I(inode)->delalloc_bytes = 0;
1775 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
1777 dir->i_sb->s_dirt = 1;
1778 btrfs_update_inode_block_group(trans, inode);
1779 btrfs_update_inode_block_group(trans, dir);
1781 nr = trans->blocks_used;
1782 btrfs_end_transaction(trans, root);
1784 mutex_unlock(&root->fs_info->fs_mutex);
1787 inode_dec_link_count(inode);
1790 btrfs_btree_balance_dirty(root, nr);
1791 btrfs_throttle(root);
1795 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1796 struct dentry *dentry)
1798 struct btrfs_trans_handle *trans;
1799 struct btrfs_root *root = BTRFS_I(dir)->root;
1800 struct inode *inode = old_dentry->d_inode;
1801 unsigned long nr = 0;
1805 if (inode->i_nlink == 0)
1808 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
1813 mutex_lock(&root->fs_info->fs_mutex);
1814 err = btrfs_check_free_space(root, 1, 0);
1817 trans = btrfs_start_transaction(root, 1);
1819 btrfs_set_trans_block_group(trans, dir);
1820 atomic_inc(&inode->i_count);
1821 err = btrfs_add_nondir(trans, dentry, inode, 1);
1826 dir->i_sb->s_dirt = 1;
1827 btrfs_update_inode_block_group(trans, dir);
1828 err = btrfs_update_inode(trans, root, inode);
1833 nr = trans->blocks_used;
1834 btrfs_end_transaction(trans, root);
1836 mutex_unlock(&root->fs_info->fs_mutex);
1839 inode_dec_link_count(inode);
1842 btrfs_btree_balance_dirty(root, nr);
1843 btrfs_throttle(root);
1847 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1849 struct inode *inode;
1850 struct btrfs_trans_handle *trans;
1851 struct btrfs_root *root = BTRFS_I(dir)->root;
1853 int drop_on_err = 0;
1855 unsigned long nr = 1;
1857 mutex_lock(&root->fs_info->fs_mutex);
1858 err = btrfs_check_free_space(root, 1, 0);
1862 trans = btrfs_start_transaction(root, 1);
1863 btrfs_set_trans_block_group(trans, dir);
1865 if (IS_ERR(trans)) {
1866 err = PTR_ERR(trans);
1870 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1876 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
1878 dentry->d_parent->d_inode->i_ino, objectid,
1879 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1880 if (IS_ERR(inode)) {
1881 err = PTR_ERR(inode);
1886 inode->i_op = &btrfs_dir_inode_operations;
1887 inode->i_fop = &btrfs_dir_file_operations;
1888 btrfs_set_trans_block_group(trans, inode);
1891 err = btrfs_update_inode(trans, root, inode);
1895 err = btrfs_add_link(trans, dentry, inode, 0);
1899 d_instantiate(dentry, inode);
1901 dir->i_sb->s_dirt = 1;
1902 btrfs_update_inode_block_group(trans, inode);
1903 btrfs_update_inode_block_group(trans, dir);
1906 nr = trans->blocks_used;
1907 btrfs_end_transaction(trans, root);
1910 mutex_unlock(&root->fs_info->fs_mutex);
1913 btrfs_btree_balance_dirty(root, nr);
1914 btrfs_throttle(root);
1918 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1919 size_t pg_offset, u64 start, u64 len,
1925 u64 extent_start = 0;
1927 u64 objectid = inode->i_ino;
1929 struct btrfs_path *path;
1930 struct btrfs_root *root = BTRFS_I(inode)->root;
1931 struct btrfs_file_extent_item *item;
1932 struct extent_buffer *leaf;
1933 struct btrfs_key found_key;
1934 struct extent_map *em = NULL;
1935 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1936 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1937 struct btrfs_trans_handle *trans = NULL;
1939 path = btrfs_alloc_path();
1941 mutex_lock(&root->fs_info->fs_mutex);
1944 spin_lock(&em_tree->lock);
1945 em = lookup_extent_mapping(em_tree, start, len);
1946 spin_unlock(&em_tree->lock);
1949 if (em->start > start) {
1950 printk("get_extent lookup [%Lu %Lu] em [%Lu %Lu]\n",
1951 start, len, em->start, em->len);
1954 if (em->block_start == EXTENT_MAP_INLINE && page)
1955 free_extent_map(em);
1959 em = alloc_extent_map(GFP_NOFS);
1965 em->start = EXTENT_MAP_HOLE;
1967 em->bdev = inode->i_sb->s_bdev;
1968 ret = btrfs_lookup_file_extent(trans, root, path,
1969 objectid, start, trans != NULL);
1976 if (path->slots[0] == 0)
1981 leaf = path->nodes[0];
1982 item = btrfs_item_ptr(leaf, path->slots[0],
1983 struct btrfs_file_extent_item);
1984 /* are we inside the extent that was found? */
1985 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1986 found_type = btrfs_key_type(&found_key);
1987 if (found_key.objectid != objectid ||
1988 found_type != BTRFS_EXTENT_DATA_KEY) {
1992 found_type = btrfs_file_extent_type(leaf, item);
1993 extent_start = found_key.offset;
1994 if (found_type == BTRFS_FILE_EXTENT_REG) {
1995 extent_end = extent_start +
1996 btrfs_file_extent_num_bytes(leaf, item);
1998 if (start < extent_start || start >= extent_end) {
2000 if (start < extent_start) {
2001 if (start + len <= extent_start)
2003 em->len = extent_end - extent_start;
2009 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
2011 em->start = extent_start;
2012 em->len = extent_end - extent_start;
2013 em->block_start = EXTENT_MAP_HOLE;
2016 bytenr += btrfs_file_extent_offset(leaf, item);
2017 em->block_start = bytenr;
2018 em->start = extent_start;
2019 em->len = extent_end - extent_start;
2021 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
2026 size_t extent_offset;
2029 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
2031 extent_end = (extent_start + size + root->sectorsize - 1) &
2032 ~((u64)root->sectorsize - 1);
2033 if (start < extent_start || start >= extent_end) {
2035 if (start < extent_start) {
2036 if (start + len <= extent_start)
2038 em->len = extent_end - extent_start;
2044 em->block_start = EXTENT_MAP_INLINE;
2047 em->start = extent_start;
2052 page_start = page_offset(page) + pg_offset;
2053 extent_offset = page_start - extent_start;
2054 copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
2055 size - extent_offset);
2056 em->start = extent_start + extent_offset;
2057 em->len = (copy_size + root->sectorsize - 1) &
2058 ~((u64)root->sectorsize - 1);
2060 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
2061 if (create == 0 && !PageUptodate(page)) {
2062 read_extent_buffer(leaf, map + pg_offset, ptr,
2064 flush_dcache_page(page);
2065 } else if (create && PageUptodate(page)) {
2068 free_extent_map(em);
2070 btrfs_release_path(root, path);
2071 trans = btrfs_start_transaction(root, 1);
2074 write_extent_buffer(leaf, map + pg_offset, ptr,
2076 btrfs_mark_buffer_dirty(leaf);
2079 set_extent_uptodate(io_tree, em->start,
2080 extent_map_end(em) - 1, GFP_NOFS);
2083 printk("unkknown found_type %d\n", found_type);
2090 em->block_start = EXTENT_MAP_HOLE;
2092 btrfs_release_path(root, path);
2093 if (em->start > start || extent_map_end(em) <= start) {
2094 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->len, start, len);
2100 spin_lock(&em_tree->lock);
2101 ret = add_extent_mapping(em_tree, em);
2102 if (ret == -EEXIST) {
2103 free_extent_map(em);
2104 em = lookup_extent_mapping(em_tree, start, len);
2107 printk("failing to insert %Lu %Lu\n", start, len);
2110 spin_unlock(&em_tree->lock);
2112 btrfs_free_path(path);
2114 ret = btrfs_end_transaction(trans, root);
2118 mutex_unlock(&root->fs_info->fs_mutex);
2120 free_extent_map(em);
2122 return ERR_PTR(err);
2127 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
2129 return extent_bmap(mapping, iblock, btrfs_get_extent);
2132 int btrfs_readpage(struct file *file, struct page *page)
2134 struct extent_io_tree *tree;
2135 tree = &BTRFS_I(page->mapping->host)->io_tree;
2136 return extent_read_full_page(tree, page, btrfs_get_extent);
2139 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
2141 struct extent_io_tree *tree;
2144 if (current->flags & PF_MEMALLOC) {
2145 redirty_page_for_writepage(wbc, page);
2149 tree = &BTRFS_I(page->mapping->host)->io_tree;
2150 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
2153 static int btrfs_writepages(struct address_space *mapping,
2154 struct writeback_control *wbc)
2156 struct extent_io_tree *tree;
2157 tree = &BTRFS_I(mapping->host)->io_tree;
2158 return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
2162 btrfs_readpages(struct file *file, struct address_space *mapping,
2163 struct list_head *pages, unsigned nr_pages)
2165 struct extent_io_tree *tree;
2166 tree = &BTRFS_I(mapping->host)->io_tree;
2167 return extent_readpages(tree, mapping, pages, nr_pages,
2171 static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
2173 struct extent_io_tree *tree;
2174 struct extent_map_tree *map;
2177 tree = &BTRFS_I(page->mapping->host)->io_tree;
2178 map = &BTRFS_I(page->mapping->host)->extent_tree;
2179 ret = try_release_extent_mapping(map, tree, page, gfp_flags);
2181 ClearPagePrivate(page);
2182 set_page_private(page, 0);
2183 page_cache_release(page);
2188 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
2190 struct extent_io_tree *tree;
2192 tree = &BTRFS_I(page->mapping->host)->io_tree;
2193 extent_invalidatepage(tree, page, offset);
2194 btrfs_releasepage(page, GFP_NOFS);
2198 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
2199 * called from a page fault handler when a page is first dirtied. Hence we must
2200 * be careful to check for EOF conditions here. We set the page up correctly
2201 * for a written page which means we get ENOSPC checking when writing into
2202 * holes and correct delalloc and unwritten extent mapping on filesystems that
2203 * support these features.
2205 * We are not allowed to take the i_mutex here so we have to play games to
2206 * protect against truncate races as the page could now be beyond EOF. Because
2207 * vmtruncate() writes the inode size before removing pages, once we have the
2208 * page lock we can determine safely if the page is beyond EOF. If it is not
2209 * beyond EOF, then the page is guaranteed safe against truncation until we
2212 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
2214 struct inode *inode = fdentry(vma->vm_file)->d_inode;
2215 struct btrfs_root *root = BTRFS_I(inode)->root;
2221 mutex_lock(&root->fs_info->fs_mutex);
2222 ret = btrfs_check_free_space(root, PAGE_CACHE_SIZE, 0);
2223 mutex_unlock(&root->fs_info->fs_mutex);
2230 wait_on_page_writeback(page);
2231 size = i_size_read(inode);
2232 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2234 if ((page->mapping != inode->i_mapping) ||
2235 (page_start > size)) {
2236 /* page got truncated out from underneath us */
2240 /* page is wholly or partially inside EOF */
2241 if (page_start + PAGE_CACHE_SIZE > size)
2242 end = size & ~PAGE_CACHE_MASK;
2244 end = PAGE_CACHE_SIZE;
2246 ret = btrfs_cow_one_page(inode, page, end);
2254 static void btrfs_truncate(struct inode *inode)
2256 struct btrfs_root *root = BTRFS_I(inode)->root;
2258 struct btrfs_trans_handle *trans;
2261 if (!S_ISREG(inode->i_mode))
2263 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
2266 btrfs_truncate_page(inode->i_mapping, inode->i_size);
2268 mutex_lock(&root->fs_info->fs_mutex);
2269 trans = btrfs_start_transaction(root, 1);
2270 btrfs_set_trans_block_group(trans, inode);
2272 /* FIXME, add redo link to tree so we don't leak on crash */
2273 ret = btrfs_truncate_in_trans(trans, root, inode,
2274 BTRFS_EXTENT_DATA_KEY);
2275 btrfs_update_inode(trans, root, inode);
2276 nr = trans->blocks_used;
2278 ret = btrfs_end_transaction(trans, root);
2280 mutex_unlock(&root->fs_info->fs_mutex);
2281 btrfs_btree_balance_dirty(root, nr);
2282 btrfs_throttle(root);
2285 static int noinline create_subvol(struct btrfs_root *root, char *name,
2288 struct btrfs_trans_handle *trans;
2289 struct btrfs_key key;
2290 struct btrfs_root_item root_item;
2291 struct btrfs_inode_item *inode_item;
2292 struct extent_buffer *leaf;
2293 struct btrfs_root *new_root = root;
2294 struct inode *inode;
2299 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
2300 unsigned long nr = 1;
2302 mutex_lock(&root->fs_info->fs_mutex);
2303 ret = btrfs_check_free_space(root, 1, 0);
2307 trans = btrfs_start_transaction(root, 1);
2310 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
2315 leaf = __btrfs_alloc_free_block(trans, root, root->leafsize,
2316 objectid, trans->transid, 0, 0,
2319 return PTR_ERR(leaf);
2321 btrfs_set_header_nritems(leaf, 0);
2322 btrfs_set_header_level(leaf, 0);
2323 btrfs_set_header_bytenr(leaf, leaf->start);
2324 btrfs_set_header_generation(leaf, trans->transid);
2325 btrfs_set_header_owner(leaf, objectid);
2327 write_extent_buffer(leaf, root->fs_info->fsid,
2328 (unsigned long)btrfs_header_fsid(leaf),
2330 btrfs_mark_buffer_dirty(leaf);
2332 inode_item = &root_item.inode;
2333 memset(inode_item, 0, sizeof(*inode_item));
2334 inode_item->generation = cpu_to_le64(1);
2335 inode_item->size = cpu_to_le64(3);
2336 inode_item->nlink = cpu_to_le32(1);
2337 inode_item->nblocks = cpu_to_le64(1);
2338 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
2340 btrfs_set_root_bytenr(&root_item, leaf->start);
2341 btrfs_set_root_level(&root_item, 0);
2342 btrfs_set_root_refs(&root_item, 1);
2343 btrfs_set_root_used(&root_item, 0);
2345 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
2346 root_item.drop_level = 0;
2348 free_extent_buffer(leaf);
2351 btrfs_set_root_dirid(&root_item, new_dirid);
2353 key.objectid = objectid;
2355 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
2356 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
2362 * insert the directory item
2364 key.offset = (u64)-1;
2365 dir = root->fs_info->sb->s_root->d_inode;
2366 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
2367 name, namelen, dir->i_ino, &key,
2372 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
2373 name, namelen, objectid,
2374 root->fs_info->sb->s_root->d_inode->i_ino);
2378 ret = btrfs_commit_transaction(trans, root);
2382 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
2385 trans = btrfs_start_transaction(new_root, 1);
2388 inode = btrfs_new_inode(trans, new_root, "..", 2, new_dirid,
2390 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
2393 inode->i_op = &btrfs_dir_inode_operations;
2394 inode->i_fop = &btrfs_dir_file_operations;
2395 new_root->inode = inode;
2397 ret = btrfs_insert_inode_ref(trans, new_root, "..", 2, new_dirid,
2401 ret = btrfs_update_inode(trans, new_root, inode);
2405 nr = trans->blocks_used;
2406 err = btrfs_commit_transaction(trans, new_root);
2410 mutex_unlock(&root->fs_info->fs_mutex);
2411 btrfs_btree_balance_dirty(root, nr);
2412 btrfs_throttle(root);
2416 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
2418 struct btrfs_pending_snapshot *pending_snapshot;
2419 struct btrfs_trans_handle *trans;
2422 unsigned long nr = 0;
2424 if (!root->ref_cows)
2427 mutex_lock(&root->fs_info->fs_mutex);
2428 ret = btrfs_check_free_space(root, 1, 0);
2432 pending_snapshot = kmalloc(sizeof(*pending_snapshot), GFP_NOFS);
2433 if (!pending_snapshot) {
2437 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
2438 if (!pending_snapshot->name) {
2440 kfree(pending_snapshot);
2443 memcpy(pending_snapshot->name, name, namelen);
2444 pending_snapshot->name[namelen] = '\0';
2445 trans = btrfs_start_transaction(root, 1);
2447 pending_snapshot->root = root;
2448 list_add(&pending_snapshot->list,
2449 &trans->transaction->pending_snapshots);
2450 ret = btrfs_update_inode(trans, root, root->inode);
2451 err = btrfs_commit_transaction(trans, root);
2454 mutex_unlock(&root->fs_info->fs_mutex);
2455 btrfs_btree_balance_dirty(root, nr);
2456 btrfs_throttle(root);
2460 unsigned long btrfs_force_ra(struct address_space *mapping,
2461 struct file_ra_state *ra, struct file *file,
2462 pgoff_t offset, pgoff_t last_index)
2466 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2467 req_size = last_index - offset + 1;
2468 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2471 req_size = min(last_index - offset + 1, (pgoff_t)128);
2472 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2473 return offset + req_size;
2477 int btrfs_defrag_file(struct file *file) {
2478 struct inode *inode = fdentry(file)->d_inode;
2479 struct btrfs_root *root = BTRFS_I(inode)->root;
2480 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2482 unsigned long last_index;
2483 unsigned long ra_index = 0;
2489 mutex_lock(&root->fs_info->fs_mutex);
2490 ret = btrfs_check_free_space(root, inode->i_size, 0);
2491 mutex_unlock(&root->fs_info->fs_mutex);
2495 mutex_lock(&inode->i_mutex);
2496 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2497 for (i = 0; i <= last_index; i++) {
2498 if (i == ra_index) {
2499 ra_index = btrfs_force_ra(inode->i_mapping,
2501 file, ra_index, last_index);
2503 page = grab_cache_page(inode->i_mapping, i);
2506 if (!PageUptodate(page)) {
2507 btrfs_readpage(NULL, page);
2509 if (!PageUptodate(page)) {
2511 page_cache_release(page);
2515 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2516 page_end = page_start + PAGE_CACHE_SIZE - 1;
2518 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2519 set_extent_delalloc(io_tree, page_start,
2520 page_end, GFP_NOFS);
2522 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2523 set_page_dirty(page);
2525 page_cache_release(page);
2526 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2530 mutex_unlock(&inode->i_mutex);
2534 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
2538 struct btrfs_ioctl_vol_args *vol_args;
2539 struct btrfs_trans_handle *trans;
2545 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2550 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2554 namelen = strlen(vol_args->name);
2555 if (namelen > BTRFS_VOL_NAME_MAX) {
2560 sizestr = vol_args->name;
2561 if (!strcmp(sizestr, "max"))
2562 new_size = root->fs_info->sb->s_bdev->bd_inode->i_size;
2564 if (sizestr[0] == '-') {
2567 } else if (sizestr[0] == '+') {
2571 new_size = btrfs_parse_size(sizestr);
2572 if (new_size == 0) {
2578 mutex_lock(&root->fs_info->fs_mutex);
2579 old_size = btrfs_super_total_bytes(&root->fs_info->super_copy);
2582 if (new_size > old_size) {
2586 new_size = old_size - new_size;
2587 } else if (mod > 0) {
2588 new_size = old_size + new_size;
2591 if (new_size < 256 * 1024 * 1024) {
2595 if (new_size > root->fs_info->sb->s_bdev->bd_inode->i_size) {
2600 do_div(new_size, root->sectorsize);
2601 new_size *= root->sectorsize;
2603 printk("new size is %Lu\n", new_size);
2604 if (new_size > old_size) {
2605 trans = btrfs_start_transaction(root, 1);
2606 ret = btrfs_grow_extent_tree(trans, root, new_size);
2607 btrfs_commit_transaction(trans, root);
2609 ret = btrfs_shrink_extent_tree(root, new_size);
2613 mutex_unlock(&root->fs_info->fs_mutex);
2619 static int noinline btrfs_ioctl_snap_create(struct btrfs_root *root,
2622 struct btrfs_ioctl_vol_args *vol_args;
2623 struct btrfs_dir_item *di;
2624 struct btrfs_path *path;
2629 vol_args = kmalloc(sizeof(*vol_args), GFP_NOFS);
2634 if (copy_from_user(vol_args, arg, sizeof(*vol_args))) {
2639 namelen = strlen(vol_args->name);
2640 if (namelen > BTRFS_VOL_NAME_MAX) {
2644 if (strchr(vol_args->name, '/')) {
2649 path = btrfs_alloc_path();
2655 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2656 mutex_lock(&root->fs_info->fs_mutex);
2657 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2659 vol_args->name, namelen, 0);
2660 mutex_unlock(&root->fs_info->fs_mutex);
2661 btrfs_free_path(path);
2663 if (di && !IS_ERR(di)) {
2673 if (root == root->fs_info->tree_root)
2674 ret = create_subvol(root, vol_args->name, namelen);
2676 ret = create_snapshot(root, vol_args->name, namelen);
2682 static int btrfs_ioctl_defrag(struct file *file)
2684 struct inode *inode = fdentry(file)->d_inode;
2685 struct btrfs_root *root = BTRFS_I(inode)->root;
2687 switch (inode->i_mode & S_IFMT) {
2689 mutex_lock(&root->fs_info->fs_mutex);
2690 btrfs_defrag_root(root, 0);
2691 btrfs_defrag_root(root->fs_info->extent_root, 0);
2692 mutex_unlock(&root->fs_info->fs_mutex);
2695 btrfs_defrag_file(file);
2702 long btrfs_ioctl(struct file *file, unsigned int
2703 cmd, unsigned long arg)
2705 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2708 case BTRFS_IOC_SNAP_CREATE:
2709 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2710 case BTRFS_IOC_DEFRAG:
2711 return btrfs_ioctl_defrag(file);
2712 case BTRFS_IOC_RESIZE:
2713 return btrfs_ioctl_resize(root, (void __user *)arg);
2720 * Called inside transaction, so use GFP_NOFS
2722 struct inode *btrfs_alloc_inode(struct super_block *sb)
2724 struct btrfs_inode *ei;
2726 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2730 ei->ordered_trans = 0;
2731 return &ei->vfs_inode;
2734 void btrfs_destroy_inode(struct inode *inode)
2736 WARN_ON(!list_empty(&inode->i_dentry));
2737 WARN_ON(inode->i_data.nrpages);
2739 btrfs_drop_extent_cache(inode, 0, (u64)-1);
2740 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2743 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2744 static void init_once(struct kmem_cache * cachep, void *foo)
2746 static void init_once(void * foo, struct kmem_cache * cachep,
2747 unsigned long flags)
2750 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2752 inode_init_once(&ei->vfs_inode);
2755 void btrfs_destroy_cachep(void)
2757 if (btrfs_inode_cachep)
2758 kmem_cache_destroy(btrfs_inode_cachep);
2759 if (btrfs_trans_handle_cachep)
2760 kmem_cache_destroy(btrfs_trans_handle_cachep);
2761 if (btrfs_transaction_cachep)
2762 kmem_cache_destroy(btrfs_transaction_cachep);
2763 if (btrfs_bit_radix_cachep)
2764 kmem_cache_destroy(btrfs_bit_radix_cachep);
2765 if (btrfs_path_cachep)
2766 kmem_cache_destroy(btrfs_path_cachep);
2769 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2770 unsigned long extra_flags,
2771 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
2772 void (*ctor)(struct kmem_cache *, void *)
2774 void (*ctor)(void *, struct kmem_cache *,
2779 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2780 SLAB_MEM_SPREAD | extra_flags), ctor
2781 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2787 int btrfs_init_cachep(void)
2789 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2790 sizeof(struct btrfs_inode),
2792 if (!btrfs_inode_cachep)
2794 btrfs_trans_handle_cachep =
2795 btrfs_cache_create("btrfs_trans_handle_cache",
2796 sizeof(struct btrfs_trans_handle),
2798 if (!btrfs_trans_handle_cachep)
2800 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2801 sizeof(struct btrfs_transaction),
2803 if (!btrfs_transaction_cachep)
2805 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2806 sizeof(struct btrfs_path),
2808 if (!btrfs_path_cachep)
2810 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2811 SLAB_DESTROY_BY_RCU, NULL);
2812 if (!btrfs_bit_radix_cachep)
2816 btrfs_destroy_cachep();
2820 static int btrfs_getattr(struct vfsmount *mnt,
2821 struct dentry *dentry, struct kstat *stat)
2823 struct inode *inode = dentry->d_inode;
2824 generic_fillattr(inode, stat);
2825 stat->blksize = PAGE_CACHE_SIZE;
2826 stat->blocks = inode->i_blocks + (BTRFS_I(inode)->delalloc_bytes >> 9);
2830 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2831 struct inode * new_dir,struct dentry *new_dentry)
2833 struct btrfs_trans_handle *trans;
2834 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2835 struct inode *new_inode = new_dentry->d_inode;
2836 struct inode *old_inode = old_dentry->d_inode;
2837 struct timespec ctime = CURRENT_TIME;
2838 struct btrfs_path *path;
2841 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2842 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2846 mutex_lock(&root->fs_info->fs_mutex);
2847 ret = btrfs_check_free_space(root, 1, 0);
2851 trans = btrfs_start_transaction(root, 1);
2853 btrfs_set_trans_block_group(trans, new_dir);
2854 path = btrfs_alloc_path();
2860 old_dentry->d_inode->i_nlink++;
2861 old_dir->i_ctime = old_dir->i_mtime = ctime;
2862 new_dir->i_ctime = new_dir->i_mtime = ctime;
2863 old_inode->i_ctime = ctime;
2865 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2870 new_inode->i_ctime = CURRENT_TIME;
2871 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2875 ret = btrfs_add_link(trans, new_dentry, old_inode, 1);
2880 btrfs_free_path(path);
2881 btrfs_end_transaction(trans, root);
2883 mutex_unlock(&root->fs_info->fs_mutex);
2887 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2888 const char *symname)
2890 struct btrfs_trans_handle *trans;
2891 struct btrfs_root *root = BTRFS_I(dir)->root;
2892 struct btrfs_path *path;
2893 struct btrfs_key key;
2894 struct inode *inode = NULL;
2901 struct btrfs_file_extent_item *ei;
2902 struct extent_buffer *leaf;
2903 unsigned long nr = 0;
2905 name_len = strlen(symname) + 1;
2906 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2907 return -ENAMETOOLONG;
2909 mutex_lock(&root->fs_info->fs_mutex);
2910 err = btrfs_check_free_space(root, 1, 0);
2914 trans = btrfs_start_transaction(root, 1);
2915 btrfs_set_trans_block_group(trans, dir);
2917 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2923 inode = btrfs_new_inode(trans, root, dentry->d_name.name,
2925 dentry->d_parent->d_inode->i_ino, objectid,
2926 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2927 err = PTR_ERR(inode);
2931 btrfs_set_trans_block_group(trans, inode);
2932 err = btrfs_add_nondir(trans, dentry, inode, 0);
2936 inode->i_mapping->a_ops = &btrfs_aops;
2937 inode->i_fop = &btrfs_file_operations;
2938 inode->i_op = &btrfs_file_inode_operations;
2939 extent_map_tree_init(&BTRFS_I(inode)->extent_tree, GFP_NOFS);
2940 extent_io_tree_init(&BTRFS_I(inode)->io_tree,
2941 inode->i_mapping, GFP_NOFS);
2942 BTRFS_I(inode)->delalloc_bytes = 0;
2943 BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
2945 dir->i_sb->s_dirt = 1;
2946 btrfs_update_inode_block_group(trans, inode);
2947 btrfs_update_inode_block_group(trans, dir);
2951 path = btrfs_alloc_path();
2953 key.objectid = inode->i_ino;
2955 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2956 datasize = btrfs_file_extent_calc_inline_size(name_len);
2957 err = btrfs_insert_empty_item(trans, root, path, &key,
2963 leaf = path->nodes[0];
2964 ei = btrfs_item_ptr(leaf, path->slots[0],
2965 struct btrfs_file_extent_item);
2966 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2967 btrfs_set_file_extent_type(leaf, ei,
2968 BTRFS_FILE_EXTENT_INLINE);
2969 ptr = btrfs_file_extent_inline_start(ei);
2970 write_extent_buffer(leaf, symname, ptr, name_len);
2971 btrfs_mark_buffer_dirty(leaf);
2972 btrfs_free_path(path);
2974 inode->i_op = &btrfs_symlink_inode_operations;
2975 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2976 inode->i_size = name_len - 1;
2977 err = btrfs_update_inode(trans, root, inode);
2982 nr = trans->blocks_used;
2983 btrfs_end_transaction(trans, root);
2985 mutex_unlock(&root->fs_info->fs_mutex);
2987 inode_dec_link_count(inode);
2990 btrfs_btree_balance_dirty(root, nr);
2991 btrfs_throttle(root);
2994 static int btrfs_permission(struct inode *inode, int mask,
2995 struct nameidata *nd)
2997 if (btrfs_test_flag(inode, READONLY) && (mask & MAY_WRITE))
2999 return generic_permission(inode, mask, NULL);
3002 static struct inode_operations btrfs_dir_inode_operations = {
3003 .lookup = btrfs_lookup,
3004 .create = btrfs_create,
3005 .unlink = btrfs_unlink,
3007 .mkdir = btrfs_mkdir,
3008 .rmdir = btrfs_rmdir,
3009 .rename = btrfs_rename,
3010 .symlink = btrfs_symlink,
3011 .setattr = btrfs_setattr,
3012 .mknod = btrfs_mknod,
3013 .setxattr = generic_setxattr,
3014 .getxattr = generic_getxattr,
3015 .listxattr = btrfs_listxattr,
3016 .removexattr = generic_removexattr,
3017 .permission = btrfs_permission,
3019 static struct inode_operations btrfs_dir_ro_inode_operations = {
3020 .lookup = btrfs_lookup,
3021 .permission = btrfs_permission,
3023 static struct file_operations btrfs_dir_file_operations = {
3024 .llseek = generic_file_llseek,
3025 .read = generic_read_dir,
3026 .readdir = btrfs_readdir,
3027 .unlocked_ioctl = btrfs_ioctl,
3028 #ifdef CONFIG_COMPAT
3029 .compat_ioctl = btrfs_ioctl,
3033 static struct extent_io_ops btrfs_extent_io_ops = {
3034 .fill_delalloc = run_delalloc_range,
3035 .submit_bio_hook = btrfs_submit_bio_hook,
3036 .readpage_io_hook = btrfs_readpage_io_hook,
3037 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
3038 .set_bit_hook = btrfs_set_bit_hook,
3039 .clear_bit_hook = btrfs_clear_bit_hook,
3042 static struct address_space_operations btrfs_aops = {
3043 .readpage = btrfs_readpage,
3044 .writepage = btrfs_writepage,
3045 .writepages = btrfs_writepages,
3046 .readpages = btrfs_readpages,
3047 .sync_page = block_sync_page,
3049 .invalidatepage = btrfs_invalidatepage,
3050 .releasepage = btrfs_releasepage,
3051 .set_page_dirty = __set_page_dirty_nobuffers,
3054 static struct address_space_operations btrfs_symlink_aops = {
3055 .readpage = btrfs_readpage,
3056 .writepage = btrfs_writepage,
3057 .invalidatepage = btrfs_invalidatepage,
3058 .releasepage = btrfs_releasepage,
3061 static struct inode_operations btrfs_file_inode_operations = {
3062 .truncate = btrfs_truncate,
3063 .getattr = btrfs_getattr,
3064 .setattr = btrfs_setattr,
3065 .setxattr = generic_setxattr,
3066 .getxattr = generic_getxattr,
3067 .listxattr = btrfs_listxattr,
3068 .removexattr = generic_removexattr,
3069 .permission = btrfs_permission,
3071 static struct inode_operations btrfs_special_inode_operations = {
3072 .getattr = btrfs_getattr,
3073 .setattr = btrfs_setattr,
3074 .permission = btrfs_permission,
3076 static struct inode_operations btrfs_symlink_inode_operations = {
3077 .readlink = generic_readlink,
3078 .follow_link = page_follow_link_light,
3079 .put_link = page_put_link,
3080 .permission = btrfs_permission,