2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/buffer_head.h>
21 #include <linux/pagemap.h>
22 #include <linux/highmem.h>
23 #include <linux/time.h>
24 #include <linux/init.h>
25 #include <linux/string.h>
26 #include <linux/smp_lock.h>
27 #include <linux/backing-dev.h>
28 #include <linux/mpage.h>
29 #include <linux/swap.h>
30 #include <linux/writeback.h>
31 #include <linux/statfs.h>
32 #include <linux/compat.h>
33 #include <linux/bit_spinlock.h>
34 #include <linux/version.h>
37 #include "transaction.h"
38 #include "btrfs_inode.h"
40 #include "print-tree.h"
42 struct btrfs_iget_args {
44 struct btrfs_root *root;
47 static struct inode_operations btrfs_dir_inode_operations;
48 static struct inode_operations btrfs_symlink_inode_operations;
49 static struct inode_operations btrfs_dir_ro_inode_operations;
50 static struct inode_operations btrfs_special_inode_operations;
51 static struct inode_operations btrfs_file_inode_operations;
52 static struct address_space_operations btrfs_aops;
53 static struct address_space_operations btrfs_symlink_aops;
54 static struct file_operations btrfs_dir_file_operations;
55 static struct extent_map_ops btrfs_extent_map_ops;
57 static struct kmem_cache *btrfs_inode_cachep;
58 struct kmem_cache *btrfs_trans_handle_cachep;
59 struct kmem_cache *btrfs_transaction_cachep;
60 struct kmem_cache *btrfs_bit_radix_cachep;
61 struct kmem_cache *btrfs_path_cachep;
64 static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
65 [S_IFREG >> S_SHIFT] = BTRFS_FT_REG_FILE,
66 [S_IFDIR >> S_SHIFT] = BTRFS_FT_DIR,
67 [S_IFCHR >> S_SHIFT] = BTRFS_FT_CHRDEV,
68 [S_IFBLK >> S_SHIFT] = BTRFS_FT_BLKDEV,
69 [S_IFIFO >> S_SHIFT] = BTRFS_FT_FIFO,
70 [S_IFSOCK >> S_SHIFT] = BTRFS_FT_SOCK,
71 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
74 static int run_delalloc_range(struct inode *inode, u64 start, u64 end)
76 struct btrfs_root *root = BTRFS_I(inode)->root;
77 struct btrfs_trans_handle *trans;
82 u64 blocksize = 1 << inode->i_blkbits;
84 mutex_lock(&root->fs_info->fs_mutex);
85 trans = btrfs_start_transaction(root, 1);
86 btrfs_set_trans_block_group(trans, inode);
88 num_blocks = (end - start + blocksize) & ~(blocksize - 1);
89 ret = btrfs_drop_extents(trans, root, inode,
90 start, start + num_blocks, &alloc_hint);
91 num_blocks = num_blocks >> inode->i_blkbits;
92 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_blocks, 0,
93 alloc_hint, (u64)-1, &ins, 1);
98 ret = btrfs_insert_file_extent(trans, root, inode->i_ino,
99 start, ins.objectid, ins.offset,
102 btrfs_end_transaction(trans, root);
103 mutex_unlock(&root->fs_info->fs_mutex);
107 int btrfs_writepage_io_hook(struct page *page, u64 start, u64 end)
109 struct inode *inode = page->mapping->host;
110 struct btrfs_root *root = BTRFS_I(inode)->root;
111 struct btrfs_trans_handle *trans;
114 u64 page_start = page->index << PAGE_CACHE_SHIFT;
115 size_t offset = start - page_start;
117 mutex_lock(&root->fs_info->fs_mutex);
118 trans = btrfs_start_transaction(root, 1);
119 btrfs_set_trans_block_group(trans, inode);
121 btrfs_csum_file_block(trans, root, inode->i_ino,
122 start, kaddr + offset, end - start + 1);
124 ret = btrfs_end_transaction(trans, root);
126 mutex_unlock(&root->fs_info->fs_mutex);
130 int btrfs_readpage_io_hook(struct page *page, u64 start, u64 end)
133 struct inode *inode = page->mapping->host;
134 struct btrfs_root *root = BTRFS_I(inode)->root;
135 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
136 struct btrfs_csum_item *item;
137 struct btrfs_path *path = NULL;
140 mutex_lock(&root->fs_info->fs_mutex);
141 path = btrfs_alloc_path();
142 item = btrfs_lookup_csum(NULL, root, path, inode->i_ino, start, 0);
145 /* a csum that isn't present is a preallocated region. */
146 if (ret == -ENOENT || ret == -EFBIG)
151 memcpy((char *)&private, &item->csum, BTRFS_CRC32_SIZE);
152 set_state_private(em_tree, start, private);
155 btrfs_free_path(path);
156 mutex_unlock(&root->fs_info->fs_mutex);
160 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
162 char csum[BTRFS_CRC32_SIZE];
163 size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
164 struct inode *inode = page->mapping->host;
165 struct btrfs_root *root = BTRFS_I(inode)->root;
166 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
171 ret = get_state_private(em_tree, start, &private);
172 kaddr = kmap_atomic(page, KM_IRQ0);
176 ret = btrfs_csum_data(root, kaddr + offset, end - start + 1, csum);
178 if (memcmp(csum, &private, BTRFS_CRC32_SIZE)) {
181 kunmap_atomic(kaddr, KM_IRQ0);
185 printk("btrfs csum failed ino %lu off %llu\n",
186 page->mapping->host->i_ino, (unsigned long long)start);
187 memset(kaddr + offset, 1, end - start + 1); flush_dcache_page(page);
188 kunmap_atomic(kaddr, KM_IRQ0);
192 void btrfs_read_locked_inode(struct inode *inode)
194 struct btrfs_path *path;
195 struct btrfs_inode_item *inode_item;
196 struct btrfs_root *root = BTRFS_I(inode)->root;
197 struct btrfs_key location;
198 u64 alloc_group_block;
202 path = btrfs_alloc_path();
204 mutex_lock(&root->fs_info->fs_mutex);
206 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
207 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
209 btrfs_free_path(path);
212 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
214 struct btrfs_inode_item);
216 inode->i_mode = btrfs_inode_mode(inode_item);
217 inode->i_nlink = btrfs_inode_nlink(inode_item);
218 inode->i_uid = btrfs_inode_uid(inode_item);
219 inode->i_gid = btrfs_inode_gid(inode_item);
220 inode->i_size = btrfs_inode_size(inode_item);
221 inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
222 inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
223 inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
224 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
225 inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
226 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
227 inode->i_blocks = btrfs_inode_nblocks(inode_item);
228 inode->i_generation = btrfs_inode_generation(inode_item);
230 rdev = btrfs_inode_rdev(inode_item);
231 alloc_group_block = btrfs_inode_block_group(inode_item);
232 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
235 btrfs_free_path(path);
238 mutex_unlock(&root->fs_info->fs_mutex);
240 switch (inode->i_mode & S_IFMT) {
242 inode->i_mapping->a_ops = &btrfs_aops;
243 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
244 inode->i_fop = &btrfs_file_operations;
245 inode->i_op = &btrfs_file_inode_operations;
248 inode->i_fop = &btrfs_dir_file_operations;
249 if (root == root->fs_info->tree_root)
250 inode->i_op = &btrfs_dir_ro_inode_operations;
252 inode->i_op = &btrfs_dir_inode_operations;
255 inode->i_op = &btrfs_symlink_inode_operations;
256 inode->i_mapping->a_ops = &btrfs_symlink_aops;
259 init_special_inode(inode, inode->i_mode, rdev);
265 btrfs_release_path(root, path);
266 btrfs_free_path(path);
267 mutex_unlock(&root->fs_info->fs_mutex);
268 make_bad_inode(inode);
271 static void fill_inode_item(struct btrfs_inode_item *item,
274 btrfs_set_inode_uid(item, inode->i_uid);
275 btrfs_set_inode_gid(item, inode->i_gid);
276 btrfs_set_inode_size(item, inode->i_size);
277 btrfs_set_inode_mode(item, inode->i_mode);
278 btrfs_set_inode_nlink(item, inode->i_nlink);
279 btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
280 btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
281 btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
282 btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
283 btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
284 btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
285 btrfs_set_inode_nblocks(item, inode->i_blocks);
286 btrfs_set_inode_generation(item, inode->i_generation);
287 btrfs_set_inode_rdev(item, inode->i_rdev);
288 btrfs_set_inode_block_group(item,
289 BTRFS_I(inode)->block_group->key.objectid);
292 int btrfs_update_inode(struct btrfs_trans_handle *trans,
293 struct btrfs_root *root,
296 struct btrfs_inode_item *inode_item;
297 struct btrfs_path *path;
300 path = btrfs_alloc_path();
302 ret = btrfs_lookup_inode(trans, root, path,
303 &BTRFS_I(inode)->location, 1);
310 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
312 struct btrfs_inode_item);
314 fill_inode_item(inode_item, inode);
315 btrfs_mark_buffer_dirty(path->nodes[0]);
316 btrfs_set_inode_last_trans(trans, inode);
319 btrfs_release_path(root, path);
320 btrfs_free_path(path);
325 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
326 struct btrfs_root *root,
328 struct dentry *dentry)
330 struct btrfs_path *path;
331 const char *name = dentry->d_name.name;
332 int name_len = dentry->d_name.len;
335 struct btrfs_dir_item *di;
337 path = btrfs_alloc_path();
343 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
353 objectid = btrfs_disk_key_objectid(&di->location);
354 ret = btrfs_delete_one_dir_name(trans, root, path, di);
357 btrfs_release_path(root, path);
359 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
360 objectid, name, name_len, -1);
369 ret = btrfs_delete_one_dir_name(trans, root, path, di);
371 dentry->d_inode->i_ctime = dir->i_ctime;
373 btrfs_free_path(path);
375 dir->i_size -= name_len * 2;
376 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
377 btrfs_update_inode(trans, root, dir);
378 drop_nlink(dentry->d_inode);
379 ret = btrfs_update_inode(trans, root, dentry->d_inode);
380 dir->i_sb->s_dirt = 1;
385 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
387 struct btrfs_root *root;
388 struct btrfs_trans_handle *trans;
391 root = BTRFS_I(dir)->root;
392 mutex_lock(&root->fs_info->fs_mutex);
393 trans = btrfs_start_transaction(root, 1);
394 btrfs_set_trans_block_group(trans, dir);
395 ret = btrfs_unlink_trans(trans, root, dir, dentry);
396 btrfs_end_transaction(trans, root);
397 mutex_unlock(&root->fs_info->fs_mutex);
398 btrfs_btree_balance_dirty(root);
402 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
404 struct inode *inode = dentry->d_inode;
407 struct btrfs_root *root = BTRFS_I(dir)->root;
408 struct btrfs_path *path;
409 struct btrfs_key key;
410 struct btrfs_trans_handle *trans;
411 struct btrfs_key found_key;
413 struct btrfs_leaf *leaf;
414 char *goodnames = "..";
416 path = btrfs_alloc_path();
418 mutex_lock(&root->fs_info->fs_mutex);
419 trans = btrfs_start_transaction(root, 1);
420 btrfs_set_trans_block_group(trans, dir);
421 key.objectid = inode->i_ino;
422 key.offset = (u64)-1;
425 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
431 if (path->slots[0] == 0) {
436 leaf = btrfs_buffer_leaf(path->nodes[0]);
437 btrfs_disk_key_to_cpu(&found_key,
438 &leaf->items[path->slots[0]].key);
439 found_type = btrfs_key_type(&found_key);
440 if (found_key.objectid != inode->i_ino) {
444 if ((found_type != BTRFS_DIR_ITEM_KEY &&
445 found_type != BTRFS_DIR_INDEX_KEY) ||
446 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
447 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
451 ret = btrfs_del_item(trans, root, path);
454 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
456 btrfs_release_path(root, path);
459 btrfs_release_path(root, path);
461 /* now the directory is empty */
462 err = btrfs_unlink_trans(trans, root, dir, dentry);
467 btrfs_release_path(root, path);
468 btrfs_free_path(path);
469 mutex_unlock(&root->fs_info->fs_mutex);
470 ret = btrfs_end_transaction(trans, root);
471 btrfs_btree_balance_dirty(root);
477 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
478 struct btrfs_root *root,
481 struct btrfs_path *path;
486 path = btrfs_alloc_path();
488 ret = btrfs_lookup_inode(trans, root, path,
489 &BTRFS_I(inode)->location, -1);
493 ret = btrfs_del_item(trans, root, path);
494 btrfs_free_path(path);
499 * this can truncate away extent items, csum items and directory items.
500 * It starts at a high offset and removes keys until it can't find
501 * any higher than i_size.
503 * csum items that cross the new i_size are truncated to the new size
506 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
507 struct btrfs_root *root,
511 struct btrfs_path *path;
512 struct btrfs_key key;
513 struct btrfs_disk_key *found_key;
515 struct btrfs_leaf *leaf;
516 struct btrfs_file_extent_item *fi;
517 u64 extent_start = 0;
518 u64 extent_num_blocks = 0;
523 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
524 path = btrfs_alloc_path();
527 /* FIXME, add redo link to tree so we don't leak on crash */
528 key.objectid = inode->i_ino;
529 key.offset = (u64)-1;
532 btrfs_init_path(path);
534 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
539 BUG_ON(path->slots[0] == 0);
542 leaf = btrfs_buffer_leaf(path->nodes[0]);
543 found_key = &leaf->items[path->slots[0]].key;
544 found_type = btrfs_disk_key_type(found_key);
546 if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
548 if (found_type != BTRFS_CSUM_ITEM_KEY &&
549 found_type != BTRFS_DIR_ITEM_KEY &&
550 found_type != BTRFS_DIR_INDEX_KEY &&
551 found_type != BTRFS_EXTENT_DATA_KEY)
554 item_end = btrfs_disk_key_offset(found_key);
555 if (found_type == BTRFS_EXTENT_DATA_KEY) {
556 fi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
558 struct btrfs_file_extent_item);
559 if (btrfs_file_extent_type(fi) !=
560 BTRFS_FILE_EXTENT_INLINE) {
561 item_end += btrfs_file_extent_num_blocks(fi) <<
565 if (found_type == BTRFS_CSUM_ITEM_KEY) {
566 ret = btrfs_csum_truncate(trans, root, path,
570 if (item_end < inode->i_size) {
571 if (found_type == BTRFS_DIR_ITEM_KEY) {
572 found_type = BTRFS_INODE_ITEM_KEY;
573 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
574 found_type = BTRFS_CSUM_ITEM_KEY;
575 } else if (found_type) {
580 btrfs_set_key_type(&key, found_type - 1);
583 if (btrfs_disk_key_offset(found_key) >= inode->i_size)
589 /* FIXME, shrink the extent if the ref count is only 1 */
590 if (found_type == BTRFS_EXTENT_DATA_KEY &&
591 btrfs_file_extent_type(fi) !=
592 BTRFS_FILE_EXTENT_INLINE) {
595 u64 orig_num_blocks =
596 btrfs_file_extent_num_blocks(fi);
597 extent_num_blocks = inode->i_size -
598 btrfs_disk_key_offset(found_key) +
600 extent_num_blocks >>= inode->i_blkbits;
601 btrfs_set_file_extent_num_blocks(fi,
603 inode->i_blocks -= (orig_num_blocks -
604 extent_num_blocks) << 3;
605 btrfs_mark_buffer_dirty(path->nodes[0]);
608 btrfs_file_extent_disk_blocknr(fi);
610 btrfs_file_extent_disk_num_blocks(fi);
611 /* FIXME blocksize != 4096 */
612 num_dec = btrfs_file_extent_num_blocks(fi) << 3;
613 if (extent_start != 0) {
615 inode->i_blocks -= num_dec;
620 ret = btrfs_del_item(trans, root, path);
626 btrfs_release_path(root, path);
628 ret = btrfs_free_extent(trans, root, extent_start,
629 extent_num_blocks, 0);
635 btrfs_release_path(root, path);
636 btrfs_free_path(path);
637 inode->i_sb->s_dirt = 1;
641 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
646 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
647 u64 page_start = page->index << PAGE_CACHE_SHIFT;
648 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
650 if (!PagePrivate(page)) {
651 SetPagePrivate(page);
652 set_page_private(page, 1);
653 WARN_ON(!page->mapping->a_ops->invalidatepage);
654 page_cache_get(page);
657 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
658 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
660 if (zero_start != PAGE_CACHE_SIZE) {
662 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
663 flush_dcache_page(page);
666 set_page_dirty(page);
667 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
673 * taken from block_truncate_page, but does cow as it zeros out
674 * any bytes left in the last page in the file.
676 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
678 struct inode *inode = mapping->host;
679 unsigned blocksize = 1 << inode->i_blkbits;
680 pgoff_t index = from >> PAGE_CACHE_SHIFT;
681 unsigned offset = from & (PAGE_CACHE_SIZE-1);
686 if ((offset & (blocksize - 1)) == 0)
689 down_read(&BTRFS_I(inode)->root->snap_sem);
691 page = grab_cache_page(mapping, index);
694 if (!PageUptodate(page)) {
695 ret = btrfs_readpage(NULL, page);
697 if (!PageUptodate(page)) {
702 page_start = page->index << PAGE_CACHE_SHIFT;
704 ret = btrfs_cow_one_page(inode, page, offset);
707 page_cache_release(page);
708 up_read(&BTRFS_I(inode)->root->snap_sem);
713 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
715 struct inode *inode = dentry->d_inode;
718 err = inode_change_ok(inode, attr);
722 if (S_ISREG(inode->i_mode) &&
723 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
724 struct btrfs_trans_handle *trans;
725 struct btrfs_root *root = BTRFS_I(inode)->root;
726 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
728 u64 mask = root->blocksize - 1;
729 u64 pos = (inode->i_size + mask) & ~mask;
730 u64 block_end = attr->ia_size | mask;
734 if (attr->ia_size <= pos)
737 btrfs_truncate_page(inode->i_mapping, inode->i_size);
739 lock_extent(em_tree, pos, block_end, GFP_NOFS);
740 hole_size = (attr->ia_size - pos + mask) & ~mask;
742 mutex_lock(&root->fs_info->fs_mutex);
743 trans = btrfs_start_transaction(root, 1);
744 btrfs_set_trans_block_group(trans, inode);
745 err = btrfs_drop_extents(trans, root, inode,
746 pos, pos + hole_size, &alloc_hint);
748 hole_size >>= inode->i_blkbits;
750 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
751 pos, 0, 0, hole_size);
752 btrfs_end_transaction(trans, root);
753 mutex_unlock(&root->fs_info->fs_mutex);
754 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
759 err = inode_setattr(inode, attr);
763 void btrfs_delete_inode(struct inode *inode)
765 struct btrfs_trans_handle *trans;
766 struct btrfs_root *root = BTRFS_I(inode)->root;
769 truncate_inode_pages(&inode->i_data, 0);
770 if (is_bad_inode(inode)) {
774 mutex_lock(&root->fs_info->fs_mutex);
775 trans = btrfs_start_transaction(root, 1);
776 btrfs_set_trans_block_group(trans, inode);
777 ret = btrfs_truncate_in_trans(trans, root, inode);
780 ret = btrfs_free_inode(trans, root, inode);
783 btrfs_end_transaction(trans, root);
784 mutex_unlock(&root->fs_info->fs_mutex);
785 btrfs_btree_balance_dirty(root);
789 btrfs_end_transaction(trans, root);
790 mutex_unlock(&root->fs_info->fs_mutex);
791 btrfs_btree_balance_dirty(root);
797 * this returns the key found in the dir entry in the location pointer.
798 * If no dir entries were found, location->objectid is 0.
800 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
801 struct btrfs_key *location)
803 const char *name = dentry->d_name.name;
804 int namelen = dentry->d_name.len;
805 struct btrfs_dir_item *di;
806 struct btrfs_path *path;
807 struct btrfs_root *root = BTRFS_I(dir)->root;
810 path = btrfs_alloc_path();
812 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
814 if (!di || IS_ERR(di)) {
815 location->objectid = 0;
819 btrfs_disk_key_to_cpu(location, &di->location);
821 btrfs_release_path(root, path);
822 btrfs_free_path(path);
827 * when we hit a tree root in a directory, the btrfs part of the inode
828 * needs to be changed to reflect the root directory of the tree root. This
829 * is kind of like crossing a mount point.
831 static int fixup_tree_root_location(struct btrfs_root *root,
832 struct btrfs_key *location,
833 struct btrfs_root **sub_root,
834 struct dentry *dentry)
836 struct btrfs_path *path;
837 struct btrfs_root_item *ri;
839 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
841 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
844 path = btrfs_alloc_path();
846 mutex_lock(&root->fs_info->fs_mutex);
848 *sub_root = btrfs_read_fs_root(root->fs_info, location,
851 if (IS_ERR(*sub_root))
852 return PTR_ERR(*sub_root);
854 ri = &(*sub_root)->root_item;
855 location->objectid = btrfs_root_dirid(ri);
857 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
858 location->offset = 0;
860 btrfs_free_path(path);
861 mutex_unlock(&root->fs_info->fs_mutex);
865 static int btrfs_init_locked_inode(struct inode *inode, void *p)
867 struct btrfs_iget_args *args = p;
868 inode->i_ino = args->ino;
869 BTRFS_I(inode)->root = args->root;
870 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
871 inode->i_mapping, GFP_NOFS);
875 static int btrfs_find_actor(struct inode *inode, void *opaque)
877 struct btrfs_iget_args *args = opaque;
878 return (args->ino == inode->i_ino &&
879 args->root == BTRFS_I(inode)->root);
882 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
883 struct btrfs_root *root)
886 struct btrfs_iget_args args;
890 inode = iget5_locked(s, objectid, btrfs_find_actor,
891 btrfs_init_locked_inode,
896 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
897 struct nameidata *nd)
899 struct inode * inode;
900 struct btrfs_inode *bi = BTRFS_I(dir);
901 struct btrfs_root *root = bi->root;
902 struct btrfs_root *sub_root = root;
903 struct btrfs_key location;
906 if (dentry->d_name.len > BTRFS_NAME_LEN)
907 return ERR_PTR(-ENAMETOOLONG);
908 mutex_lock(&root->fs_info->fs_mutex);
909 ret = btrfs_inode_by_name(dir, dentry, &location);
910 mutex_unlock(&root->fs_info->fs_mutex);
914 if (location.objectid) {
915 ret = fixup_tree_root_location(root, &location, &sub_root,
920 return ERR_PTR(-ENOENT);
921 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
924 return ERR_PTR(-EACCES);
925 if (inode->i_state & I_NEW) {
926 /* the inode and parent dir are two different roots */
927 if (sub_root != root) {
929 sub_root->inode = inode;
931 BTRFS_I(inode)->root = sub_root;
932 memcpy(&BTRFS_I(inode)->location, &location,
934 btrfs_read_locked_inode(inode);
935 unlock_new_inode(inode);
938 return d_splice_alias(inode, dentry);
941 static unsigned char btrfs_filetype_table[] = {
942 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
945 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
947 struct inode *inode = filp->f_path.dentry->d_inode;
948 struct btrfs_root *root = BTRFS_I(inode)->root;
949 struct btrfs_item *item;
950 struct btrfs_dir_item *di;
951 struct btrfs_key key;
952 struct btrfs_path *path;
955 struct btrfs_leaf *leaf;
958 unsigned char d_type;
963 int key_type = BTRFS_DIR_INDEX_KEY;
965 /* FIXME, use a real flag for deciding about the key type */
966 if (root->fs_info->tree_root == root)
967 key_type = BTRFS_DIR_ITEM_KEY;
968 mutex_lock(&root->fs_info->fs_mutex);
969 key.objectid = inode->i_ino;
971 btrfs_set_key_type(&key, key_type);
972 key.offset = filp->f_pos;
973 path = btrfs_alloc_path();
975 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
980 leaf = btrfs_buffer_leaf(path->nodes[0]);
981 nritems = btrfs_header_nritems(&leaf->header);
982 slot = path->slots[0];
983 if (advance || slot >= nritems) {
984 if (slot >= nritems -1) {
985 ret = btrfs_next_leaf(root, path);
988 leaf = btrfs_buffer_leaf(path->nodes[0]);
989 nritems = btrfs_header_nritems(&leaf->header);
990 slot = path->slots[0];
997 item = leaf->items + slot;
998 if (btrfs_disk_key_objectid(&item->key) != key.objectid)
1000 if (btrfs_disk_key_type(&item->key) != key_type)
1002 if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
1004 filp->f_pos = btrfs_disk_key_offset(&item->key);
1006 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1008 di_total = btrfs_item_size(leaf->items + slot);
1009 while(di_cur < di_total) {
1010 d_type = btrfs_filetype_table[btrfs_dir_type(di)];
1011 over = filldir(dirent, (const char *)(di + 1),
1012 btrfs_dir_name_len(di),
1013 btrfs_disk_key_offset(&item->key),
1014 btrfs_disk_key_objectid(&di->location),
1018 di_len = btrfs_dir_name_len(di) + sizeof(*di);
1020 di = (struct btrfs_dir_item *)((char *)di + di_len);
1027 btrfs_release_path(root, path);
1028 btrfs_free_path(path);
1029 mutex_unlock(&root->fs_info->fs_mutex);
1033 int btrfs_write_inode(struct inode *inode, int wait)
1035 struct btrfs_root *root = BTRFS_I(inode)->root;
1036 struct btrfs_trans_handle *trans;
1040 mutex_lock(&root->fs_info->fs_mutex);
1041 trans = btrfs_start_transaction(root, 1);
1042 btrfs_set_trans_block_group(trans, inode);
1043 ret = btrfs_commit_transaction(trans, root);
1044 mutex_unlock(&root->fs_info->fs_mutex);
1050 * This is somewhat expensive, updating the tree every time the
1051 * inode changes. But, it is most likely to find the inode in cache.
1052 * FIXME, needs more benchmarking...there are no reasons other than performance
1053 * to keep or drop this code.
1055 void btrfs_dirty_inode(struct inode *inode)
1057 struct btrfs_root *root = BTRFS_I(inode)->root;
1058 struct btrfs_trans_handle *trans;
1060 mutex_lock(&root->fs_info->fs_mutex);
1061 trans = btrfs_start_transaction(root, 1);
1062 btrfs_set_trans_block_group(trans, inode);
1063 btrfs_update_inode(trans, root, inode);
1064 btrfs_end_transaction(trans, root);
1065 mutex_unlock(&root->fs_info->fs_mutex);
1068 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1069 struct btrfs_root *root,
1071 struct btrfs_block_group_cache *group,
1074 struct inode *inode;
1075 struct btrfs_inode_item inode_item;
1076 struct btrfs_key *location;
1080 inode = new_inode(root->fs_info->sb);
1082 return ERR_PTR(-ENOMEM);
1084 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1085 inode->i_mapping, GFP_NOFS);
1086 BTRFS_I(inode)->root = root;
1092 group = btrfs_find_block_group(root, group, 0, 0, owner);
1093 BTRFS_I(inode)->block_group = group;
1095 inode->i_uid = current->fsuid;
1096 inode->i_gid = current->fsgid;
1097 inode->i_mode = mode;
1098 inode->i_ino = objectid;
1099 inode->i_blocks = 0;
1100 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1101 fill_inode_item(&inode_item, inode);
1102 location = &BTRFS_I(inode)->location;
1103 location->objectid = objectid;
1104 location->flags = 0;
1105 location->offset = 0;
1106 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1108 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1110 return ERR_PTR(ret);
1111 insert_inode_hash(inode);
1115 static inline u8 btrfs_inode_type(struct inode *inode)
1117 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1120 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1121 struct dentry *dentry, struct inode *inode)
1124 struct btrfs_key key;
1125 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1126 struct inode *parent_inode;
1127 key.objectid = inode->i_ino;
1129 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1132 ret = btrfs_insert_dir_item(trans, root,
1133 dentry->d_name.name, dentry->d_name.len,
1134 dentry->d_parent->d_inode->i_ino,
1135 &key, btrfs_inode_type(inode));
1137 parent_inode = dentry->d_parent->d_inode;
1138 parent_inode->i_size += dentry->d_name.len * 2;
1139 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1140 ret = btrfs_update_inode(trans, root,
1141 dentry->d_parent->d_inode);
1146 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1147 struct dentry *dentry, struct inode *inode)
1149 int err = btrfs_add_link(trans, dentry, inode);
1151 d_instantiate(dentry, inode);
1159 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1160 int mode, dev_t rdev)
1162 struct btrfs_trans_handle *trans;
1163 struct btrfs_root *root = BTRFS_I(dir)->root;
1164 struct inode *inode;
1169 if (!new_valid_dev(rdev))
1172 mutex_lock(&root->fs_info->fs_mutex);
1173 trans = btrfs_start_transaction(root, 1);
1174 btrfs_set_trans_block_group(trans, dir);
1176 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1182 inode = btrfs_new_inode(trans, root, objectid,
1183 BTRFS_I(dir)->block_group, mode);
1184 err = PTR_ERR(inode);
1188 btrfs_set_trans_block_group(trans, inode);
1189 err = btrfs_add_nondir(trans, dentry, inode);
1193 inode->i_op = &btrfs_special_inode_operations;
1194 init_special_inode(inode, inode->i_mode, rdev);
1195 btrfs_update_inode(trans, root, inode);
1197 dir->i_sb->s_dirt = 1;
1198 btrfs_update_inode_block_group(trans, inode);
1199 btrfs_update_inode_block_group(trans, dir);
1201 btrfs_end_transaction(trans, root);
1202 mutex_unlock(&root->fs_info->fs_mutex);
1205 inode_dec_link_count(inode);
1208 btrfs_btree_balance_dirty(root);
1212 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1213 int mode, struct nameidata *nd)
1215 struct btrfs_trans_handle *trans;
1216 struct btrfs_root *root = BTRFS_I(dir)->root;
1217 struct inode *inode;
1222 mutex_lock(&root->fs_info->fs_mutex);
1223 trans = btrfs_start_transaction(root, 1);
1224 btrfs_set_trans_block_group(trans, dir);
1226 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1232 inode = btrfs_new_inode(trans, root, objectid,
1233 BTRFS_I(dir)->block_group, mode);
1234 err = PTR_ERR(inode);
1238 btrfs_set_trans_block_group(trans, inode);
1239 err = btrfs_add_nondir(trans, dentry, inode);
1243 inode->i_mapping->a_ops = &btrfs_aops;
1244 inode->i_fop = &btrfs_file_operations;
1245 inode->i_op = &btrfs_file_inode_operations;
1246 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1247 inode->i_mapping, GFP_NOFS);
1248 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
1250 dir->i_sb->s_dirt = 1;
1251 btrfs_update_inode_block_group(trans, inode);
1252 btrfs_update_inode_block_group(trans, dir);
1254 btrfs_end_transaction(trans, root);
1255 mutex_unlock(&root->fs_info->fs_mutex);
1258 inode_dec_link_count(inode);
1261 btrfs_btree_balance_dirty(root);
1265 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1266 struct dentry *dentry)
1268 struct btrfs_trans_handle *trans;
1269 struct btrfs_root *root = BTRFS_I(dir)->root;
1270 struct inode *inode = old_dentry->d_inode;
1274 if (inode->i_nlink == 0)
1278 mutex_lock(&root->fs_info->fs_mutex);
1279 trans = btrfs_start_transaction(root, 1);
1280 btrfs_set_trans_block_group(trans, dir);
1281 atomic_inc(&inode->i_count);
1282 err = btrfs_add_nondir(trans, dentry, inode);
1285 dir->i_sb->s_dirt = 1;
1286 btrfs_update_inode_block_group(trans, dir);
1287 err = btrfs_update_inode(trans, root, inode);
1291 btrfs_end_transaction(trans, root);
1292 mutex_unlock(&root->fs_info->fs_mutex);
1295 inode_dec_link_count(inode);
1298 btrfs_btree_balance_dirty(root);
1302 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1303 struct btrfs_root *root,
1304 u64 objectid, u64 dirid)
1308 struct btrfs_key key;
1313 key.objectid = objectid;
1316 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1318 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1319 &key, BTRFS_FT_DIR);
1322 key.objectid = dirid;
1323 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1324 &key, BTRFS_FT_DIR);
1331 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1333 struct inode *inode;
1334 struct btrfs_trans_handle *trans;
1335 struct btrfs_root *root = BTRFS_I(dir)->root;
1337 int drop_on_err = 0;
1340 mutex_lock(&root->fs_info->fs_mutex);
1341 trans = btrfs_start_transaction(root, 1);
1342 btrfs_set_trans_block_group(trans, dir);
1343 if (IS_ERR(trans)) {
1344 err = PTR_ERR(trans);
1348 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1354 inode = btrfs_new_inode(trans, root, objectid,
1355 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1356 if (IS_ERR(inode)) {
1357 err = PTR_ERR(inode);
1361 inode->i_op = &btrfs_dir_inode_operations;
1362 inode->i_fop = &btrfs_dir_file_operations;
1363 btrfs_set_trans_block_group(trans, inode);
1365 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1370 err = btrfs_update_inode(trans, root, inode);
1373 err = btrfs_add_link(trans, dentry, inode);
1376 d_instantiate(dentry, inode);
1378 dir->i_sb->s_dirt = 1;
1379 btrfs_update_inode_block_group(trans, inode);
1380 btrfs_update_inode_block_group(trans, dir);
1383 btrfs_end_transaction(trans, root);
1385 mutex_unlock(&root->fs_info->fs_mutex);
1388 btrfs_btree_balance_dirty(root);
1392 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1393 size_t page_offset, u64 start, u64 end,
1399 u64 extent_start = 0;
1401 u64 objectid = inode->i_ino;
1403 int failed_insert = 0;
1404 struct btrfs_path *path;
1405 struct btrfs_root *root = BTRFS_I(inode)->root;
1406 struct btrfs_file_extent_item *item;
1407 struct btrfs_leaf *leaf;
1408 struct btrfs_disk_key *found_key;
1409 struct extent_map *em = NULL;
1410 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1411 struct btrfs_trans_handle *trans = NULL;
1413 path = btrfs_alloc_path();
1415 mutex_lock(&root->fs_info->fs_mutex);
1418 em = lookup_extent_mapping(em_tree, start, end);
1423 em = alloc_extent_map(GFP_NOFS);
1431 em->bdev = inode->i_sb->s_bdev;
1432 ret = btrfs_lookup_file_extent(NULL, root, path,
1433 objectid, start, 0);
1440 if (path->slots[0] == 0)
1445 item = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
1446 struct btrfs_file_extent_item);
1447 leaf = btrfs_buffer_leaf(path->nodes[0]);
1448 blocknr = btrfs_file_extent_disk_blocknr(item);
1449 blocknr += btrfs_file_extent_offset(item);
1451 /* are we inside the extent that was found? */
1452 found_key = &leaf->items[path->slots[0]].key;
1453 found_type = btrfs_disk_key_type(found_key);
1454 if (btrfs_disk_key_objectid(found_key) != objectid ||
1455 found_type != BTRFS_EXTENT_DATA_KEY) {
1459 found_type = btrfs_file_extent_type(item);
1460 extent_start = btrfs_disk_key_offset(&leaf->items[path->slots[0]].key);
1461 if (found_type == BTRFS_FILE_EXTENT_REG) {
1462 extent_end = extent_start +
1463 (btrfs_file_extent_num_blocks(item) << inode->i_blkbits);
1465 if (start < extent_start || start >= extent_end) {
1467 if (start < extent_start) {
1468 if (end < extent_start)
1470 em->end = extent_end - 1;
1476 if (btrfs_file_extent_disk_blocknr(item) == 0) {
1477 em->start = extent_start;
1478 em->end = extent_end - 1;
1479 em->block_start = 0;
1483 em->block_start = blocknr << inode->i_blkbits;
1484 em->block_end = em->block_start +
1485 (btrfs_file_extent_num_blocks(item) <<
1486 inode->i_blkbits) - 1;
1487 em->start = extent_start;
1488 em->end = extent_end - 1;
1490 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1495 size = btrfs_file_extent_inline_len(leaf->items +
1497 extent_end = extent_start + size;
1498 if (start < extent_start || start >= extent_end) {
1500 if (start < extent_start) {
1501 if (end < extent_start)
1503 em->end = extent_end - 1;
1509 em->block_start = EXTENT_MAP_INLINE;
1510 em->block_end = EXTENT_MAP_INLINE;
1511 em->start = extent_start;
1512 em->end = extent_end - 1;
1516 ptr = btrfs_file_extent_inline_start(item);
1518 memcpy(map + page_offset, ptr, size);
1519 flush_dcache_page(result->b_page);
1521 set_extent_uptodate(em_tree, extent_start,
1522 extent_end, GFP_NOFS);
1525 printk("unkknown found_type %d\n", found_type);
1532 em->block_start = 0;
1535 btrfs_release_path(root, path);
1536 if (em->start > start || em->end < start) {
1537 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1541 ret = add_extent_mapping(em_tree, em);
1542 if (ret == -EEXIST) {
1543 free_extent_map(em);
1546 if (failed_insert > 5) {
1547 printk("failing to insert %Lu %Lu\n", start, end);
1555 btrfs_free_path(path);
1557 ret = btrfs_end_transaction(trans, root);
1561 mutex_unlock(&root->fs_info->fs_mutex);
1563 free_extent_map(em);
1565 return ERR_PTR(err);
1570 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
1572 return extent_bmap(mapping, iblock, btrfs_get_extent);
1575 static int btrfs_prepare_write(struct file *file, struct page *page,
1576 unsigned from, unsigned to)
1578 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1579 page->mapping->host, page, from, to,
1583 int btrfs_readpage(struct file *file, struct page *page)
1585 struct extent_map_tree *tree;
1586 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1587 return extent_read_full_page(tree, page, btrfs_get_extent);
1589 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1591 struct extent_map_tree *tree;
1594 if (current->flags & PF_MEMALLOC) {
1595 redirty_page_for_writepage(wbc, page);
1599 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1600 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
1603 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1605 struct extent_map_tree *tree;
1608 if (page->private != 1) {
1610 return try_to_free_buffers(page);
1612 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1613 ret = try_release_extent_mapping(tree, page);
1615 ClearPagePrivate(page);
1616 set_page_private(page, 0);
1617 page_cache_release(page);
1622 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1624 struct extent_map_tree *tree;
1626 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1627 extent_invalidatepage(tree, page, offset);
1628 btrfs_releasepage(page, GFP_NOFS);
1632 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1633 * called from a page fault handler when a page is first dirtied. Hence we must
1634 * be careful to check for EOF conditions here. We set the page up correctly
1635 * for a written page which means we get ENOSPC checking when writing into
1636 * holes and correct delalloc and unwritten extent mapping on filesystems that
1637 * support these features.
1639 * We are not allowed to take the i_mutex here so we have to play games to
1640 * protect against truncate races as the page could now be beyond EOF. Because
1641 * vmtruncate() writes the inode size before removing pages, once we have the
1642 * page lock we can determine safely if the page is beyond EOF. If it is not
1643 * beyond EOF, then the page is guaranteed safe against truncation until we
1646 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1648 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1654 down_read(&BTRFS_I(inode)->root->snap_sem);
1656 wait_on_page_writeback(page);
1657 size = i_size_read(inode);
1658 page_start = page->index << PAGE_CACHE_SHIFT;
1660 if ((page->mapping != inode->i_mapping) ||
1661 (page_start > size)) {
1662 /* page got truncated out from underneath us */
1666 /* page is wholly or partially inside EOF */
1667 if (page_start + PAGE_CACHE_SIZE > size)
1668 end = size & ~PAGE_CACHE_MASK;
1670 end = PAGE_CACHE_SIZE;
1672 ret = btrfs_cow_one_page(inode, page, end);
1675 up_read(&BTRFS_I(inode)->root->snap_sem);
1680 static void btrfs_truncate(struct inode *inode)
1682 struct btrfs_root *root = BTRFS_I(inode)->root;
1684 struct btrfs_trans_handle *trans;
1686 if (!S_ISREG(inode->i_mode))
1688 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1691 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1693 mutex_lock(&root->fs_info->fs_mutex);
1694 trans = btrfs_start_transaction(root, 1);
1695 btrfs_set_trans_block_group(trans, inode);
1697 /* FIXME, add redo link to tree so we don't leak on crash */
1698 ret = btrfs_truncate_in_trans(trans, root, inode);
1699 btrfs_update_inode(trans, root, inode);
1700 ret = btrfs_end_transaction(trans, root);
1702 mutex_unlock(&root->fs_info->fs_mutex);
1703 btrfs_btree_balance_dirty(root);
1706 int btrfs_commit_write(struct file *file, struct page *page,
1707 unsigned from, unsigned to)
1709 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1710 page->mapping->host, page, from, to);
1713 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1715 struct btrfs_trans_handle *trans;
1716 struct btrfs_key key;
1717 struct btrfs_root_item root_item;
1718 struct btrfs_inode_item *inode_item;
1719 struct buffer_head *subvol;
1720 struct btrfs_leaf *leaf;
1721 struct btrfs_root *new_root;
1722 struct inode *inode;
1727 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1729 mutex_lock(&root->fs_info->fs_mutex);
1730 trans = btrfs_start_transaction(root, 1);
1733 subvol = btrfs_alloc_free_block(trans, root, 0, 0);
1735 return PTR_ERR(subvol);
1736 leaf = btrfs_buffer_leaf(subvol);
1737 btrfs_set_header_nritems(&leaf->header, 0);
1738 btrfs_set_header_level(&leaf->header, 0);
1739 btrfs_set_header_blocknr(&leaf->header, bh_blocknr(subvol));
1740 btrfs_set_header_generation(&leaf->header, trans->transid);
1741 btrfs_set_header_owner(&leaf->header, root->root_key.objectid);
1742 memcpy(leaf->header.fsid, root->fs_info->disk_super->fsid,
1743 sizeof(leaf->header.fsid));
1744 btrfs_mark_buffer_dirty(subvol);
1746 inode_item = &root_item.inode;
1747 memset(inode_item, 0, sizeof(*inode_item));
1748 btrfs_set_inode_generation(inode_item, 1);
1749 btrfs_set_inode_size(inode_item, 3);
1750 btrfs_set_inode_nlink(inode_item, 1);
1751 btrfs_set_inode_nblocks(inode_item, 1);
1752 btrfs_set_inode_mode(inode_item, S_IFDIR | 0755);
1754 btrfs_set_root_blocknr(&root_item, bh_blocknr(subvol));
1755 btrfs_set_root_refs(&root_item, 1);
1756 btrfs_set_root_blocks_used(&root_item, 0);
1757 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1758 root_item.drop_level = 0;
1762 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1767 btrfs_set_root_dirid(&root_item, new_dirid);
1769 key.objectid = objectid;
1772 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1773 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1779 * insert the directory item
1781 key.offset = (u64)-1;
1782 dir = root->fs_info->sb->s_root->d_inode;
1783 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1784 name, namelen, dir->i_ino, &key,
1789 ret = btrfs_commit_transaction(trans, root);
1793 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1796 trans = btrfs_start_transaction(new_root, 1);
1799 inode = btrfs_new_inode(trans, new_root, new_dirid,
1800 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
1803 inode->i_op = &btrfs_dir_inode_operations;
1804 inode->i_fop = &btrfs_dir_file_operations;
1805 new_root->inode = inode;
1807 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1813 ret = btrfs_update_inode(trans, new_root, inode);
1817 err = btrfs_commit_transaction(trans, root);
1821 mutex_unlock(&root->fs_info->fs_mutex);
1822 btrfs_btree_balance_dirty(root);
1826 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1828 struct btrfs_trans_handle *trans;
1829 struct btrfs_key key;
1830 struct btrfs_root_item new_root_item;
1831 struct buffer_head *tmp;
1836 if (!root->ref_cows)
1839 down_write(&root->snap_sem);
1840 freeze_bdev(root->fs_info->sb->s_bdev);
1841 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
1843 mutex_lock(&root->fs_info->fs_mutex);
1844 trans = btrfs_start_transaction(root, 1);
1847 ret = btrfs_update_inode(trans, root, root->inode);
1851 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1856 memcpy(&new_root_item, &root->root_item,
1857 sizeof(new_root_item));
1859 key.objectid = objectid;
1862 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1863 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
1864 btrfs_set_root_blocknr(&new_root_item, bh_blocknr(root->node));
1866 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1872 * insert the directory item
1874 key.offset = (u64)-1;
1875 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1877 root->fs_info->sb->s_root->d_inode->i_ino,
1878 &key, BTRFS_FT_DIR);
1883 ret = btrfs_inc_root_ref(trans, root);
1887 err = btrfs_commit_transaction(trans, root);
1890 mutex_unlock(&root->fs_info->fs_mutex);
1891 up_write(&root->snap_sem);
1892 btrfs_btree_balance_dirty(root);
1896 static unsigned long force_ra(struct address_space *mapping,
1897 struct file_ra_state *ra, struct file *file,
1898 pgoff_t offset, pgoff_t last_index)
1902 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
1903 req_size = last_index - offset + 1;
1904 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
1907 req_size = min(last_index - offset + 1, (pgoff_t)128);
1908 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
1909 return offset + req_size;
1913 int btrfs_defrag_file(struct file *file) {
1914 struct inode *inode = file->f_path.dentry->d_inode;
1915 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1917 unsigned long last_index;
1918 unsigned long ra_index = 0;
1923 mutex_lock(&inode->i_mutex);
1924 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
1925 for (i = 0; i <= last_index; i++) {
1926 if (i == ra_index) {
1927 ra_index = force_ra(inode->i_mapping, &file->f_ra,
1928 file, ra_index, last_index);
1930 page = grab_cache_page(inode->i_mapping, i);
1933 if (!PageUptodate(page)) {
1934 btrfs_readpage(NULL, page);
1936 if (!PageUptodate(page)) {
1938 page_cache_release(page);
1942 page_start = page->index << PAGE_CACHE_SHIFT;
1943 page_end = page_start + PAGE_CACHE_SIZE - 1;
1945 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
1946 set_extent_delalloc(em_tree, page_start,
1947 page_end, GFP_NOFS);
1948 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
1949 set_page_dirty(page);
1951 page_cache_release(page);
1952 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
1956 mutex_unlock(&inode->i_mutex);
1960 long btrfs_ioctl(struct file *file, unsigned int
1961 cmd, unsigned long arg)
1963 struct inode *inode = file->f_path.dentry->d_inode;
1964 struct btrfs_root *root = BTRFS_I(inode)->root;
1965 struct btrfs_ioctl_vol_args vol_args;
1967 struct btrfs_dir_item *di;
1969 struct btrfs_path *path;
1973 case BTRFS_IOC_SNAP_CREATE:
1974 if (copy_from_user(&vol_args,
1975 (struct btrfs_ioctl_vol_args __user *)arg,
1978 namelen = strlen(vol_args.name);
1979 if (namelen > BTRFS_VOL_NAME_MAX)
1981 if (strchr(vol_args.name, '/'))
1983 path = btrfs_alloc_path();
1986 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
1987 mutex_lock(&root->fs_info->fs_mutex);
1988 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
1990 vol_args.name, namelen, 0);
1991 mutex_unlock(&root->fs_info->fs_mutex);
1992 btrfs_free_path(path);
1993 if (di && !IS_ERR(di))
1998 if (root == root->fs_info->tree_root)
1999 ret = create_subvol(root, vol_args.name, namelen);
2001 ret = create_snapshot(root, vol_args.name, namelen);
2004 case BTRFS_IOC_DEFRAG:
2005 if (S_ISDIR(inode->i_mode)) {
2006 mutex_lock(&root->fs_info->fs_mutex);
2007 btrfs_defrag_root(root, 0);
2008 btrfs_defrag_root(root->fs_info->extent_root, 0);
2009 mutex_unlock(&root->fs_info->fs_mutex);
2010 } else if (S_ISREG(inode->i_mode)) {
2011 btrfs_defrag_file(file);
2022 * Called inside transaction, so use GFP_NOFS
2024 struct inode *btrfs_alloc_inode(struct super_block *sb)
2026 struct btrfs_inode *ei;
2028 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2032 return &ei->vfs_inode;
2035 void btrfs_destroy_inode(struct inode *inode)
2037 WARN_ON(!list_empty(&inode->i_dentry));
2038 WARN_ON(inode->i_data.nrpages);
2040 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2043 static void init_once(void * foo, struct kmem_cache * cachep,
2044 unsigned long flags)
2046 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2048 inode_init_once(&ei->vfs_inode);
2051 void btrfs_destroy_cachep(void)
2053 if (btrfs_inode_cachep)
2054 kmem_cache_destroy(btrfs_inode_cachep);
2055 if (btrfs_trans_handle_cachep)
2056 kmem_cache_destroy(btrfs_trans_handle_cachep);
2057 if (btrfs_transaction_cachep)
2058 kmem_cache_destroy(btrfs_transaction_cachep);
2059 if (btrfs_bit_radix_cachep)
2060 kmem_cache_destroy(btrfs_bit_radix_cachep);
2061 if (btrfs_path_cachep)
2062 kmem_cache_destroy(btrfs_path_cachep);
2065 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2066 unsigned long extra_flags,
2067 void (*ctor)(void *, struct kmem_cache *,
2070 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2071 SLAB_MEM_SPREAD | extra_flags), ctor
2072 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2078 int btrfs_init_cachep(void)
2080 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2081 sizeof(struct btrfs_inode),
2083 if (!btrfs_inode_cachep)
2085 btrfs_trans_handle_cachep =
2086 btrfs_cache_create("btrfs_trans_handle_cache",
2087 sizeof(struct btrfs_trans_handle),
2089 if (!btrfs_trans_handle_cachep)
2091 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2092 sizeof(struct btrfs_transaction),
2094 if (!btrfs_transaction_cachep)
2096 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2097 sizeof(struct btrfs_transaction),
2099 if (!btrfs_path_cachep)
2101 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2102 SLAB_DESTROY_BY_RCU, NULL);
2103 if (!btrfs_bit_radix_cachep)
2107 btrfs_destroy_cachep();
2111 static int btrfs_getattr(struct vfsmount *mnt,
2112 struct dentry *dentry, struct kstat *stat)
2114 struct inode *inode = dentry->d_inode;
2115 generic_fillattr(inode, stat);
2116 stat->blksize = 256 * 1024;
2120 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2121 struct inode * new_dir,struct dentry *new_dentry)
2123 struct btrfs_trans_handle *trans;
2124 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2125 struct inode *new_inode = new_dentry->d_inode;
2126 struct inode *old_inode = old_dentry->d_inode;
2127 struct timespec ctime = CURRENT_TIME;
2128 struct btrfs_path *path;
2129 struct btrfs_dir_item *di;
2132 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2133 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2136 mutex_lock(&root->fs_info->fs_mutex);
2137 trans = btrfs_start_transaction(root, 1);
2138 btrfs_set_trans_block_group(trans, new_dir);
2139 path = btrfs_alloc_path();
2145 old_dentry->d_inode->i_nlink++;
2146 old_dir->i_ctime = old_dir->i_mtime = ctime;
2147 new_dir->i_ctime = new_dir->i_mtime = ctime;
2148 old_inode->i_ctime = ctime;
2149 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2150 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2152 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2162 old_parent_oid = btrfs_disk_key_objectid(&di->location);
2163 ret = btrfs_del_item(trans, root, path);
2167 btrfs_release_path(root, path);
2169 di = btrfs_lookup_dir_index_item(trans, root, path,
2181 ret = btrfs_del_item(trans, root, path);
2185 btrfs_release_path(root, path);
2187 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2188 old_inode->i_ino, location,
2195 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2200 new_inode->i_ctime = CURRENT_TIME;
2201 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2205 ret = btrfs_add_link(trans, new_dentry, old_inode);
2210 btrfs_free_path(path);
2211 btrfs_end_transaction(trans, root);
2212 mutex_unlock(&root->fs_info->fs_mutex);
2216 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2217 const char *symname)
2219 struct btrfs_trans_handle *trans;
2220 struct btrfs_root *root = BTRFS_I(dir)->root;
2221 struct btrfs_path *path;
2222 struct btrfs_key key;
2223 struct inode *inode;
2230 struct btrfs_file_extent_item *ei;
2232 name_len = strlen(symname) + 1;
2233 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2234 return -ENAMETOOLONG;
2235 mutex_lock(&root->fs_info->fs_mutex);
2236 trans = btrfs_start_transaction(root, 1);
2237 btrfs_set_trans_block_group(trans, dir);
2239 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2245 inode = btrfs_new_inode(trans, root, objectid,
2246 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2247 err = PTR_ERR(inode);
2251 btrfs_set_trans_block_group(trans, inode);
2252 err = btrfs_add_nondir(trans, dentry, inode);
2256 inode->i_mapping->a_ops = &btrfs_aops;
2257 inode->i_fop = &btrfs_file_operations;
2258 inode->i_op = &btrfs_file_inode_operations;
2259 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2260 inode->i_mapping, GFP_NOFS);
2261 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2263 dir->i_sb->s_dirt = 1;
2264 btrfs_update_inode_block_group(trans, inode);
2265 btrfs_update_inode_block_group(trans, dir);
2269 path = btrfs_alloc_path();
2271 key.objectid = inode->i_ino;
2274 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2275 datasize = btrfs_file_extent_calc_inline_size(name_len);
2276 err = btrfs_insert_empty_item(trans, root, path, &key,
2282 ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]),
2283 path->slots[0], struct btrfs_file_extent_item);
2284 btrfs_set_file_extent_generation(ei, trans->transid);
2285 btrfs_set_file_extent_type(ei,
2286 BTRFS_FILE_EXTENT_INLINE);
2287 ptr = btrfs_file_extent_inline_start(ei);
2288 btrfs_memcpy(root, path->nodes[0]->b_data,
2289 ptr, symname, name_len);
2290 btrfs_mark_buffer_dirty(path->nodes[0]);
2291 btrfs_free_path(path);
2292 inode->i_op = &btrfs_symlink_inode_operations;
2293 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2294 inode->i_size = name_len - 1;
2295 err = btrfs_update_inode(trans, root, inode);
2300 btrfs_end_transaction(trans, root);
2301 mutex_unlock(&root->fs_info->fs_mutex);
2303 inode_dec_link_count(inode);
2306 btrfs_btree_balance_dirty(root);
2310 static struct inode_operations btrfs_dir_inode_operations = {
2311 .lookup = btrfs_lookup,
2312 .create = btrfs_create,
2313 .unlink = btrfs_unlink,
2315 .mkdir = btrfs_mkdir,
2316 .rmdir = btrfs_rmdir,
2317 .rename = btrfs_rename,
2318 .symlink = btrfs_symlink,
2319 .setattr = btrfs_setattr,
2320 .mknod = btrfs_mknod,
2323 static struct inode_operations btrfs_dir_ro_inode_operations = {
2324 .lookup = btrfs_lookup,
2327 static struct file_operations btrfs_dir_file_operations = {
2328 .llseek = generic_file_llseek,
2329 .read = generic_read_dir,
2330 .readdir = btrfs_readdir,
2331 .unlocked_ioctl = btrfs_ioctl,
2332 #ifdef CONFIG_COMPAT
2333 .compat_ioctl = btrfs_ioctl,
2337 static struct extent_map_ops btrfs_extent_map_ops = {
2338 .fill_delalloc = run_delalloc_range,
2339 .writepage_io_hook = btrfs_writepage_io_hook,
2340 .readpage_io_hook = btrfs_readpage_io_hook,
2341 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2344 static struct address_space_operations btrfs_aops = {
2345 .readpage = btrfs_readpage,
2346 .writepage = btrfs_writepage,
2347 .sync_page = block_sync_page,
2348 .prepare_write = btrfs_prepare_write,
2349 .commit_write = btrfs_commit_write,
2351 .invalidatepage = btrfs_invalidatepage,
2352 .releasepage = btrfs_releasepage,
2353 .set_page_dirty = __set_page_dirty_nobuffers,
2356 static struct address_space_operations btrfs_symlink_aops = {
2357 .readpage = btrfs_readpage,
2358 .writepage = btrfs_writepage,
2359 .invalidatepage = btrfs_invalidatepage,
2360 .releasepage = btrfs_releasepage,
2363 static struct inode_operations btrfs_file_inode_operations = {
2364 .truncate = btrfs_truncate,
2365 .getattr = btrfs_getattr,
2366 .setattr = btrfs_setattr,
2369 static struct inode_operations btrfs_special_inode_operations = {
2370 .getattr = btrfs_getattr,
2371 .setattr = btrfs_setattr,
2374 static struct inode_operations btrfs_symlink_inode_operations = {
2375 .readlink = generic_readlink,
2376 .follow_link = page_follow_link_light,
2377 .put_link = page_put_link,