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 = root->sectorsize;
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_bytes = (end - start + blocksize) & ~(blocksize - 1);
89 ret = btrfs_drop_extents(trans, root, inode,
90 start, start + num_bytes, start, &alloc_hint);
92 ret = btrfs_alloc_extent(trans, root, inode->i_ino, num_bytes, 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, 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 read_extent_buffer(path->nodes[0], &csum, (unsigned long)item,
153 set_state_private(em_tree, start, csum);
156 btrfs_free_path(path);
157 mutex_unlock(&root->fs_info->fs_mutex);
161 int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end)
163 size_t offset = start - (page->index << PAGE_CACHE_SHIFT);
164 struct inode *inode = page->mapping->host;
165 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
169 struct btrfs_root *root = BTRFS_I(inode)->root;
173 ret = get_state_private(em_tree, start, &private);
174 local_irq_save(flags);
175 kaddr = kmap_atomic(page, KM_IRQ0);
179 csum = btrfs_csum_data(root, kaddr + offset, csum, end - start + 1);
180 btrfs_csum_final(csum, (char *)&csum);
181 if (csum != private) {
184 kunmap_atomic(kaddr, KM_IRQ0);
185 local_irq_restore(flags);
189 printk("btrfs csum failed ino %lu off %llu\n",
190 page->mapping->host->i_ino, (unsigned long long)start);
191 memset(kaddr + offset, 1, end - start + 1);
192 flush_dcache_page(page);
193 kunmap_atomic(kaddr, KM_IRQ0);
194 local_irq_restore(flags);
198 void btrfs_read_locked_inode(struct inode *inode)
200 struct btrfs_path *path;
201 struct extent_buffer *leaf;
202 struct btrfs_inode_item *inode_item;
203 struct btrfs_inode_timespec *tspec;
204 struct btrfs_root *root = BTRFS_I(inode)->root;
205 struct btrfs_key location;
206 u64 alloc_group_block;
210 path = btrfs_alloc_path();
212 mutex_lock(&root->fs_info->fs_mutex);
214 memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
215 ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
219 leaf = path->nodes[0];
220 inode_item = btrfs_item_ptr(leaf, path->slots[0],
221 struct btrfs_inode_item);
223 inode->i_mode = btrfs_inode_mode(leaf, inode_item);
224 inode->i_nlink = btrfs_inode_nlink(leaf, inode_item);
225 inode->i_uid = btrfs_inode_uid(leaf, inode_item);
226 inode->i_gid = btrfs_inode_gid(leaf, inode_item);
227 inode->i_size = btrfs_inode_size(leaf, inode_item);
229 tspec = btrfs_inode_atime(inode_item);
230 inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
231 inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
233 tspec = btrfs_inode_mtime(inode_item);
234 inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
235 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
237 tspec = btrfs_inode_ctime(inode_item);
238 inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
239 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);
241 inode->i_blocks = btrfs_inode_nblocks(leaf, inode_item);
242 inode->i_generation = btrfs_inode_generation(leaf, inode_item);
244 rdev = btrfs_inode_rdev(leaf, inode_item);
246 alloc_group_block = btrfs_inode_block_group(leaf, inode_item);
247 BTRFS_I(inode)->block_group = btrfs_lookup_block_group(root->fs_info,
250 btrfs_free_path(path);
253 mutex_unlock(&root->fs_info->fs_mutex);
255 switch (inode->i_mode & S_IFMT) {
257 inode->i_mapping->a_ops = &btrfs_aops;
258 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
259 inode->i_fop = &btrfs_file_operations;
260 inode->i_op = &btrfs_file_inode_operations;
263 inode->i_fop = &btrfs_dir_file_operations;
264 if (root == root->fs_info->tree_root)
265 inode->i_op = &btrfs_dir_ro_inode_operations;
267 inode->i_op = &btrfs_dir_inode_operations;
270 inode->i_op = &btrfs_symlink_inode_operations;
271 inode->i_mapping->a_ops = &btrfs_symlink_aops;
274 init_special_inode(inode, inode->i_mode, rdev);
280 btrfs_release_path(root, path);
281 btrfs_free_path(path);
282 mutex_unlock(&root->fs_info->fs_mutex);
283 make_bad_inode(inode);
286 static void fill_inode_item(struct extent_buffer *leaf,
287 struct btrfs_inode_item *item,
290 btrfs_set_inode_uid(leaf, item, inode->i_uid);
291 btrfs_set_inode_gid(leaf, item, inode->i_gid);
292 btrfs_set_inode_size(leaf, item, inode->i_size);
293 btrfs_set_inode_mode(leaf, item, inode->i_mode);
294 btrfs_set_inode_nlink(leaf, item, inode->i_nlink);
296 btrfs_set_timespec_sec(leaf, btrfs_inode_atime(item),
297 inode->i_atime.tv_sec);
298 btrfs_set_timespec_nsec(leaf, btrfs_inode_atime(item),
299 inode->i_atime.tv_nsec);
301 btrfs_set_timespec_sec(leaf, btrfs_inode_mtime(item),
302 inode->i_mtime.tv_sec);
303 btrfs_set_timespec_nsec(leaf, btrfs_inode_mtime(item),
304 inode->i_mtime.tv_nsec);
306 btrfs_set_timespec_sec(leaf, btrfs_inode_ctime(item),
307 inode->i_ctime.tv_sec);
308 btrfs_set_timespec_nsec(leaf, btrfs_inode_ctime(item),
309 inode->i_ctime.tv_nsec);
311 btrfs_set_inode_nblocks(leaf, item, inode->i_blocks);
312 btrfs_set_inode_generation(leaf, item, inode->i_generation);
313 btrfs_set_inode_rdev(leaf, item, inode->i_rdev);
314 btrfs_set_inode_block_group(leaf, item,
315 BTRFS_I(inode)->block_group->key.objectid);
318 int btrfs_update_inode(struct btrfs_trans_handle *trans,
319 struct btrfs_root *root,
322 struct btrfs_inode_item *inode_item;
323 struct btrfs_path *path;
324 struct extent_buffer *leaf;
327 path = btrfs_alloc_path();
329 ret = btrfs_lookup_inode(trans, root, path,
330 &BTRFS_I(inode)->location, 1);
337 leaf = path->nodes[0];
338 inode_item = btrfs_item_ptr(leaf, path->slots[0],
339 struct btrfs_inode_item);
341 fill_inode_item(leaf, inode_item, inode);
342 btrfs_mark_buffer_dirty(leaf);
343 btrfs_set_inode_last_trans(trans, inode);
346 btrfs_release_path(root, path);
347 btrfs_free_path(path);
352 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
353 struct btrfs_root *root,
355 struct dentry *dentry)
357 struct btrfs_path *path;
358 const char *name = dentry->d_name.name;
359 int name_len = dentry->d_name.len;
361 struct extent_buffer *leaf;
362 struct btrfs_dir_item *di;
363 struct btrfs_key key;
365 path = btrfs_alloc_path();
371 di = btrfs_lookup_dir_item(trans, root, path, dir->i_ino,
381 leaf = path->nodes[0];
382 btrfs_dir_item_key_to_cpu(leaf, di, &key);
383 ret = btrfs_delete_one_dir_name(trans, root, path, di);
386 btrfs_release_path(root, path);
388 di = btrfs_lookup_dir_index_item(trans, root, path, dir->i_ino,
389 key.objectid, name, name_len, -1);
398 ret = btrfs_delete_one_dir_name(trans, root, path, di);
400 dentry->d_inode->i_ctime = dir->i_ctime;
402 btrfs_free_path(path);
404 dir->i_size -= name_len * 2;
405 dir->i_mtime = dir->i_ctime = CURRENT_TIME;
406 btrfs_update_inode(trans, root, dir);
407 drop_nlink(dentry->d_inode);
408 ret = btrfs_update_inode(trans, root, dentry->d_inode);
409 dir->i_sb->s_dirt = 1;
414 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
416 struct btrfs_root *root;
417 struct btrfs_trans_handle *trans;
421 root = BTRFS_I(dir)->root;
422 mutex_lock(&root->fs_info->fs_mutex);
423 trans = btrfs_start_transaction(root, 1);
425 btrfs_set_trans_block_group(trans, dir);
426 ret = btrfs_unlink_trans(trans, root, dir, dentry);
427 nr = trans->blocks_used;
429 btrfs_end_transaction(trans, root);
430 mutex_unlock(&root->fs_info->fs_mutex);
431 btrfs_btree_balance_dirty(root, nr);
436 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
438 struct inode *inode = dentry->d_inode;
441 struct btrfs_root *root = BTRFS_I(dir)->root;
442 struct btrfs_path *path;
443 struct btrfs_key key;
444 struct btrfs_trans_handle *trans;
445 struct btrfs_key found_key;
447 struct extent_buffer *leaf;
448 char *goodnames = "..";
451 path = btrfs_alloc_path();
453 mutex_lock(&root->fs_info->fs_mutex);
454 trans = btrfs_start_transaction(root, 1);
456 btrfs_set_trans_block_group(trans, dir);
457 key.objectid = inode->i_ino;
458 key.offset = (u64)-1;
461 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
467 if (path->slots[0] == 0) {
472 leaf = path->nodes[0];
473 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
474 found_type = btrfs_key_type(&found_key);
475 if (found_key.objectid != inode->i_ino) {
479 if ((found_type != BTRFS_DIR_ITEM_KEY &&
480 found_type != BTRFS_DIR_INDEX_KEY) ||
481 (!btrfs_match_dir_item_name(root, path, goodnames, 2) &&
482 !btrfs_match_dir_item_name(root, path, goodnames, 1))) {
486 ret = btrfs_del_item(trans, root, path);
489 if (found_type == BTRFS_DIR_ITEM_KEY && found_key.offset == 1)
491 btrfs_release_path(root, path);
494 btrfs_release_path(root, path);
496 /* now the directory is empty */
497 err = btrfs_unlink_trans(trans, root, dir, dentry);
502 btrfs_release_path(root, path);
503 btrfs_free_path(path);
504 mutex_unlock(&root->fs_info->fs_mutex);
505 nr = trans->blocks_used;
506 ret = btrfs_end_transaction(trans, root);
507 btrfs_btree_balance_dirty(root, nr);
513 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
514 struct btrfs_root *root,
517 struct btrfs_path *path;
522 path = btrfs_alloc_path();
524 ret = btrfs_lookup_inode(trans, root, path,
525 &BTRFS_I(inode)->location, -1);
529 ret = btrfs_del_item(trans, root, path);
530 btrfs_free_path(path);
535 * this can truncate away extent items, csum items and directory items.
536 * It starts at a high offset and removes keys until it can't find
537 * any higher than i_size.
539 * csum items that cross the new i_size are truncated to the new size
542 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
543 struct btrfs_root *root,
547 struct btrfs_path *path;
548 struct btrfs_key key;
549 struct btrfs_key found_key;
551 struct extent_buffer *leaf;
552 struct btrfs_file_extent_item *fi;
553 u64 extent_start = 0;
554 u64 extent_num_bytes = 0;
559 btrfs_drop_extent_cache(inode, inode->i_size, (u64)-1);
560 path = btrfs_alloc_path();
564 /* FIXME, add redo link to tree so we don't leak on crash */
565 key.objectid = inode->i_ino;
566 key.offset = (u64)-1;
570 btrfs_init_path(path);
572 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
577 BUG_ON(path->slots[0] == 0);
580 leaf = path->nodes[0];
581 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
582 found_type = btrfs_key_type(&found_key);
584 if (found_key.objectid != inode->i_ino)
587 if (found_type != BTRFS_CSUM_ITEM_KEY &&
588 found_type != BTRFS_DIR_ITEM_KEY &&
589 found_type != BTRFS_DIR_INDEX_KEY &&
590 found_type != BTRFS_EXTENT_DATA_KEY)
593 item_end = found_key.offset;
594 if (found_type == BTRFS_EXTENT_DATA_KEY) {
595 fi = btrfs_item_ptr(leaf, path->slots[0],
596 struct btrfs_file_extent_item);
597 if (btrfs_file_extent_type(leaf, fi) !=
598 BTRFS_FILE_EXTENT_INLINE) {
600 btrfs_file_extent_num_bytes(leaf, fi);
603 if (found_type == BTRFS_CSUM_ITEM_KEY) {
604 ret = btrfs_csum_truncate(trans, root, path,
608 if (item_end < inode->i_size) {
609 if (found_type == BTRFS_DIR_ITEM_KEY) {
610 found_type = BTRFS_INODE_ITEM_KEY;
611 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
612 found_type = BTRFS_CSUM_ITEM_KEY;
613 } else if (found_type) {
618 btrfs_set_key_type(&key, found_type);
619 btrfs_release_path(root, path);
622 if (found_key.offset >= inode->i_size)
628 /* FIXME, shrink the extent if the ref count is only 1 */
629 if (found_type == BTRFS_EXTENT_DATA_KEY &&
630 btrfs_file_extent_type(leaf, fi) !=
631 BTRFS_FILE_EXTENT_INLINE) {
633 extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
636 btrfs_file_extent_num_bytes(leaf, fi);
637 extent_num_bytes = inode->i_size -
638 found_key.offset + root->sectorsize - 1;
639 btrfs_set_file_extent_num_bytes(leaf, fi,
641 num_dec = (orig_num_bytes -
642 extent_num_bytes) >> 9;
643 if (extent_start != 0) {
644 inode->i_blocks -= num_dec;
646 btrfs_mark_buffer_dirty(leaf);
649 btrfs_file_extent_disk_num_bytes(leaf,
651 /* FIXME blocksize != 4096 */
652 num_dec = btrfs_file_extent_num_bytes(leaf,
654 if (extent_start != 0) {
656 inode->i_blocks -= num_dec;
661 ret = btrfs_del_item(trans, root, path);
667 btrfs_release_path(root, path);
669 ret = btrfs_free_extent(trans, root, extent_start,
670 extent_num_bytes, 0);
676 btrfs_release_path(root, path);
677 btrfs_free_path(path);
678 inode->i_sb->s_dirt = 1;
682 static int btrfs_cow_one_page(struct inode *inode, struct page *page,
687 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
688 u64 page_start = page->index << PAGE_CACHE_SHIFT;
689 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
691 set_page_extent_mapped(page);
693 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
694 set_extent_delalloc(&BTRFS_I(inode)->extent_tree, page_start,
696 if (zero_start != PAGE_CACHE_SIZE) {
698 memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
699 flush_dcache_page(page);
702 set_page_dirty(page);
703 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
709 * taken from block_truncate_page, but does cow as it zeros out
710 * any bytes left in the last page in the file.
712 static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
714 struct inode *inode = mapping->host;
715 struct btrfs_root *root = BTRFS_I(inode)->root;
716 u32 blocksize = root->sectorsize;
717 pgoff_t index = from >> PAGE_CACHE_SHIFT;
718 unsigned offset = from & (PAGE_CACHE_SIZE-1);
723 if ((offset & (blocksize - 1)) == 0)
726 down_read(&root->snap_sem);
728 page = grab_cache_page(mapping, index);
731 if (!PageUptodate(page)) {
732 ret = btrfs_readpage(NULL, page);
734 if (!PageUptodate(page)) {
739 page_start = page->index << PAGE_CACHE_SHIFT;
741 ret = btrfs_cow_one_page(inode, page, offset);
744 page_cache_release(page);
745 up_read(&BTRFS_I(inode)->root->snap_sem);
750 static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
752 struct inode *inode = dentry->d_inode;
755 err = inode_change_ok(inode, attr);
759 if (S_ISREG(inode->i_mode) &&
760 attr->ia_valid & ATTR_SIZE && attr->ia_size > inode->i_size) {
761 struct btrfs_trans_handle *trans;
762 struct btrfs_root *root = BTRFS_I(inode)->root;
763 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
765 u64 mask = root->sectorsize - 1;
766 u64 pos = (inode->i_size + mask) & ~mask;
767 u64 block_end = attr->ia_size | mask;
771 if (attr->ia_size <= pos)
774 btrfs_truncate_page(inode->i_mapping, inode->i_size);
776 lock_extent(em_tree, pos, block_end, GFP_NOFS);
777 hole_size = (attr->ia_size - pos + mask) & ~mask;
779 mutex_lock(&root->fs_info->fs_mutex);
780 trans = btrfs_start_transaction(root, 1);
781 btrfs_set_trans_block_group(trans, inode);
782 err = btrfs_drop_extents(trans, root, inode,
783 pos, pos + hole_size, pos,
786 err = btrfs_insert_file_extent(trans, root, inode->i_ino,
787 pos, 0, 0, hole_size);
788 btrfs_end_transaction(trans, root);
789 mutex_unlock(&root->fs_info->fs_mutex);
790 unlock_extent(em_tree, pos, block_end, GFP_NOFS);
795 err = inode_setattr(inode, attr);
799 void btrfs_delete_inode(struct inode *inode)
801 struct btrfs_trans_handle *trans;
802 struct btrfs_root *root = BTRFS_I(inode)->root;
806 truncate_inode_pages(&inode->i_data, 0);
807 if (is_bad_inode(inode)) {
812 mutex_lock(&root->fs_info->fs_mutex);
813 trans = btrfs_start_transaction(root, 1);
815 btrfs_set_trans_block_group(trans, inode);
816 ret = btrfs_truncate_in_trans(trans, root, inode);
819 ret = btrfs_free_inode(trans, root, inode);
822 nr = trans->blocks_used;
824 btrfs_end_transaction(trans, root);
825 mutex_unlock(&root->fs_info->fs_mutex);
826 btrfs_btree_balance_dirty(root, nr);
830 nr = trans->blocks_used;
831 btrfs_end_transaction(trans, root);
832 mutex_unlock(&root->fs_info->fs_mutex);
833 btrfs_btree_balance_dirty(root, nr);
839 * this returns the key found in the dir entry in the location pointer.
840 * If no dir entries were found, location->objectid is 0.
842 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
843 struct btrfs_key *location)
845 const char *name = dentry->d_name.name;
846 int namelen = dentry->d_name.len;
847 struct btrfs_dir_item *di;
848 struct btrfs_path *path;
849 struct btrfs_root *root = BTRFS_I(dir)->root;
852 path = btrfs_alloc_path();
854 di = btrfs_lookup_dir_item(NULL, root, path, dir->i_ino, name,
858 if (!di || IS_ERR(di)) {
859 location->objectid = 0;
862 btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
864 btrfs_release_path(root, path);
865 btrfs_free_path(path);
870 * when we hit a tree root in a directory, the btrfs part of the inode
871 * needs to be changed to reflect the root directory of the tree root. This
872 * is kind of like crossing a mount point.
874 static int fixup_tree_root_location(struct btrfs_root *root,
875 struct btrfs_key *location,
876 struct btrfs_root **sub_root,
877 struct dentry *dentry)
879 struct btrfs_path *path;
880 struct btrfs_root_item *ri;
882 if (btrfs_key_type(location) != BTRFS_ROOT_ITEM_KEY)
884 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
887 path = btrfs_alloc_path();
889 mutex_lock(&root->fs_info->fs_mutex);
891 *sub_root = btrfs_read_fs_root(root->fs_info, location,
894 if (IS_ERR(*sub_root))
895 return PTR_ERR(*sub_root);
897 ri = &(*sub_root)->root_item;
898 location->objectid = btrfs_root_dirid(ri);
899 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
900 location->offset = 0;
902 btrfs_free_path(path);
903 mutex_unlock(&root->fs_info->fs_mutex);
907 static int btrfs_init_locked_inode(struct inode *inode, void *p)
909 struct btrfs_iget_args *args = p;
910 inode->i_ino = args->ino;
911 BTRFS_I(inode)->root = args->root;
912 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
913 inode->i_mapping, GFP_NOFS);
917 static int btrfs_find_actor(struct inode *inode, void *opaque)
919 struct btrfs_iget_args *args = opaque;
920 return (args->ino == inode->i_ino &&
921 args->root == BTRFS_I(inode)->root);
924 struct inode *btrfs_iget_locked(struct super_block *s, u64 objectid,
925 struct btrfs_root *root)
928 struct btrfs_iget_args args;
932 inode = iget5_locked(s, objectid, btrfs_find_actor,
933 btrfs_init_locked_inode,
938 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
939 struct nameidata *nd)
941 struct inode * inode;
942 struct btrfs_inode *bi = BTRFS_I(dir);
943 struct btrfs_root *root = bi->root;
944 struct btrfs_root *sub_root = root;
945 struct btrfs_key location;
948 if (dentry->d_name.len > BTRFS_NAME_LEN)
949 return ERR_PTR(-ENAMETOOLONG);
951 mutex_lock(&root->fs_info->fs_mutex);
952 ret = btrfs_inode_by_name(dir, dentry, &location);
953 mutex_unlock(&root->fs_info->fs_mutex);
959 if (location.objectid) {
960 ret = fixup_tree_root_location(root, &location, &sub_root,
965 return ERR_PTR(-ENOENT);
966 inode = btrfs_iget_locked(dir->i_sb, location.objectid,
969 return ERR_PTR(-EACCES);
970 if (inode->i_state & I_NEW) {
971 /* the inode and parent dir are two different roots */
972 if (sub_root != root) {
974 sub_root->inode = inode;
976 BTRFS_I(inode)->root = sub_root;
977 memcpy(&BTRFS_I(inode)->location, &location,
979 btrfs_read_locked_inode(inode);
980 unlock_new_inode(inode);
983 return d_splice_alias(inode, dentry);
986 static unsigned char btrfs_filetype_table[] = {
987 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
990 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
992 struct inode *inode = filp->f_path.dentry->d_inode;
993 struct btrfs_root *root = BTRFS_I(inode)->root;
994 struct btrfs_item *item;
995 struct btrfs_dir_item *di;
996 struct btrfs_key key;
997 struct btrfs_key found_key;
998 struct btrfs_path *path;
1001 struct extent_buffer *leaf;
1004 unsigned char d_type;
1009 int key_type = BTRFS_DIR_INDEX_KEY;
1014 /* FIXME, use a real flag for deciding about the key type */
1015 if (root->fs_info->tree_root == root)
1016 key_type = BTRFS_DIR_ITEM_KEY;
1018 mutex_lock(&root->fs_info->fs_mutex);
1019 key.objectid = inode->i_ino;
1020 btrfs_set_key_type(&key, key_type);
1021 key.offset = filp->f_pos;
1023 path = btrfs_alloc_path();
1025 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1030 leaf = path->nodes[0];
1031 nritems = btrfs_header_nritems(leaf);
1032 slot = path->slots[0];
1033 if (advance || slot >= nritems) {
1034 if (slot >= nritems -1) {
1035 ret = btrfs_next_leaf(root, path);
1038 leaf = path->nodes[0];
1039 nritems = btrfs_header_nritems(leaf);
1040 slot = path->slots[0];
1047 item = btrfs_item_nr(leaf, slot);
1048 btrfs_item_key_to_cpu(leaf, &found_key, slot);
1050 if (found_key.objectid != key.objectid)
1052 if (btrfs_key_type(&found_key) != key_type)
1054 if (found_key.offset < filp->f_pos)
1057 filp->f_pos = found_key.offset;
1059 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
1061 di_total = btrfs_item_size(leaf, item);
1062 while(di_cur < di_total) {
1063 struct btrfs_key location;
1065 name_len = btrfs_dir_name_len(leaf, di);
1066 if (name_len < 32) {
1067 name_ptr = tmp_name;
1069 name_ptr = kmalloc(name_len, GFP_NOFS);
1072 read_extent_buffer(leaf, name_ptr,
1073 (unsigned long)(di + 1), name_len);
1075 d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
1076 btrfs_dir_item_key_to_cpu(leaf, di, &location);
1078 over = filldir(dirent, name_ptr, name_len,
1083 if (name_ptr != tmp_name)
1088 di_len = btrfs_dir_name_len(leaf, di) + sizeof(*di);
1090 di = (struct btrfs_dir_item *)((char *)di + di_len);
1097 btrfs_release_path(root, path);
1098 btrfs_free_path(path);
1099 mutex_unlock(&root->fs_info->fs_mutex);
1103 int btrfs_write_inode(struct inode *inode, int wait)
1105 struct btrfs_root *root = BTRFS_I(inode)->root;
1106 struct btrfs_trans_handle *trans;
1110 mutex_lock(&root->fs_info->fs_mutex);
1111 trans = btrfs_start_transaction(root, 1);
1112 btrfs_set_trans_block_group(trans, inode);
1113 ret = btrfs_commit_transaction(trans, root);
1114 mutex_unlock(&root->fs_info->fs_mutex);
1120 * This is somewhat expensive, updating the tree every time the
1121 * inode changes. But, it is most likely to find the inode in cache.
1122 * FIXME, needs more benchmarking...there are no reasons other than performance
1123 * to keep or drop this code.
1125 void btrfs_dirty_inode(struct inode *inode)
1127 struct btrfs_root *root = BTRFS_I(inode)->root;
1128 struct btrfs_trans_handle *trans;
1130 mutex_lock(&root->fs_info->fs_mutex);
1131 trans = btrfs_start_transaction(root, 1);
1132 btrfs_set_trans_block_group(trans, inode);
1133 btrfs_update_inode(trans, root, inode);
1134 btrfs_end_transaction(trans, root);
1135 mutex_unlock(&root->fs_info->fs_mutex);
1138 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
1139 struct btrfs_root *root,
1141 struct btrfs_block_group_cache *group,
1144 struct inode *inode;
1145 struct btrfs_inode_item *inode_item;
1146 struct btrfs_key *location;
1147 struct btrfs_path *path;
1151 path = btrfs_alloc_path();
1154 inode = new_inode(root->fs_info->sb);
1156 return ERR_PTR(-ENOMEM);
1158 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1159 inode->i_mapping, GFP_NOFS);
1160 BTRFS_I(inode)->root = root;
1166 group = btrfs_find_block_group(root, group, 0, 0, owner);
1167 BTRFS_I(inode)->block_group = group;
1169 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
1173 inode->i_uid = current->fsuid;
1174 inode->i_gid = current->fsgid;
1175 inode->i_mode = mode;
1176 inode->i_ino = objectid;
1177 inode->i_blocks = 0;
1178 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1179 inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
1180 struct btrfs_inode_item);
1181 fill_inode_item(path->nodes[0], inode_item, inode);
1182 btrfs_mark_buffer_dirty(path->nodes[0]);
1183 btrfs_free_path(path);
1185 location = &BTRFS_I(inode)->location;
1186 location->objectid = objectid;
1187 location->offset = 0;
1188 btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);
1190 insert_inode_hash(inode);
1193 btrfs_free_path(path);
1194 return ERR_PTR(ret);
1197 static inline u8 btrfs_inode_type(struct inode *inode)
1199 return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
1202 static int btrfs_add_link(struct btrfs_trans_handle *trans,
1203 struct dentry *dentry, struct inode *inode)
1206 struct btrfs_key key;
1207 struct btrfs_root *root = BTRFS_I(dentry->d_parent->d_inode)->root;
1208 struct inode *parent_inode;
1210 key.objectid = inode->i_ino;
1211 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1214 ret = btrfs_insert_dir_item(trans, root,
1215 dentry->d_name.name, dentry->d_name.len,
1216 dentry->d_parent->d_inode->i_ino,
1217 &key, btrfs_inode_type(inode));
1219 parent_inode = dentry->d_parent->d_inode;
1220 parent_inode->i_size += dentry->d_name.len * 2;
1221 parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
1222 ret = btrfs_update_inode(trans, root,
1223 dentry->d_parent->d_inode);
1228 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
1229 struct dentry *dentry, struct inode *inode)
1231 int err = btrfs_add_link(trans, dentry, inode);
1233 d_instantiate(dentry, inode);
1241 static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
1242 int mode, dev_t rdev)
1244 struct btrfs_trans_handle *trans;
1245 struct btrfs_root *root = BTRFS_I(dir)->root;
1246 struct inode *inode;
1252 if (!new_valid_dev(rdev))
1255 mutex_lock(&root->fs_info->fs_mutex);
1256 trans = btrfs_start_transaction(root, 1);
1257 btrfs_set_trans_block_group(trans, dir);
1259 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1265 inode = btrfs_new_inode(trans, root, objectid,
1266 BTRFS_I(dir)->block_group, mode);
1267 err = PTR_ERR(inode);
1271 btrfs_set_trans_block_group(trans, inode);
1272 err = btrfs_add_nondir(trans, dentry, inode);
1276 inode->i_op = &btrfs_special_inode_operations;
1277 init_special_inode(inode, inode->i_mode, rdev);
1278 btrfs_update_inode(trans, root, inode);
1280 dir->i_sb->s_dirt = 1;
1281 btrfs_update_inode_block_group(trans, inode);
1282 btrfs_update_inode_block_group(trans, dir);
1284 nr = trans->blocks_used;
1285 btrfs_end_transaction(trans, root);
1286 mutex_unlock(&root->fs_info->fs_mutex);
1289 inode_dec_link_count(inode);
1292 btrfs_btree_balance_dirty(root, nr);
1296 static int btrfs_create(struct inode *dir, struct dentry *dentry,
1297 int mode, struct nameidata *nd)
1299 struct btrfs_trans_handle *trans;
1300 struct btrfs_root *root = BTRFS_I(dir)->root;
1301 struct inode *inode;
1307 mutex_lock(&root->fs_info->fs_mutex);
1308 trans = btrfs_start_transaction(root, 1);
1309 btrfs_set_trans_block_group(trans, dir);
1311 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1317 inode = btrfs_new_inode(trans, root, objectid,
1318 BTRFS_I(dir)->block_group, mode);
1319 err = PTR_ERR(inode);
1323 btrfs_set_trans_block_group(trans, inode);
1324 err = btrfs_add_nondir(trans, dentry, inode);
1328 inode->i_mapping->a_ops = &btrfs_aops;
1329 inode->i_fop = &btrfs_file_operations;
1330 inode->i_op = &btrfs_file_inode_operations;
1331 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
1332 inode->i_mapping, GFP_NOFS);
1333 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
1335 dir->i_sb->s_dirt = 1;
1336 btrfs_update_inode_block_group(trans, inode);
1337 btrfs_update_inode_block_group(trans, dir);
1339 nr = trans->blocks_used;
1340 btrfs_end_transaction(trans, root);
1341 mutex_unlock(&root->fs_info->fs_mutex);
1344 inode_dec_link_count(inode);
1347 btrfs_btree_balance_dirty(root, nr);
1351 static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
1352 struct dentry *dentry)
1354 struct btrfs_trans_handle *trans;
1355 struct btrfs_root *root = BTRFS_I(dir)->root;
1356 struct inode *inode = old_dentry->d_inode;
1361 if (inode->i_nlink == 0)
1365 mutex_lock(&root->fs_info->fs_mutex);
1366 trans = btrfs_start_transaction(root, 1);
1368 btrfs_set_trans_block_group(trans, dir);
1369 atomic_inc(&inode->i_count);
1370 err = btrfs_add_nondir(trans, dentry, inode);
1375 dir->i_sb->s_dirt = 1;
1376 btrfs_update_inode_block_group(trans, dir);
1377 err = btrfs_update_inode(trans, root, inode);
1382 nr = trans->blocks_used;
1383 btrfs_end_transaction(trans, root);
1384 mutex_unlock(&root->fs_info->fs_mutex);
1387 inode_dec_link_count(inode);
1390 btrfs_btree_balance_dirty(root, nr);
1394 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
1395 struct btrfs_root *root,
1396 u64 objectid, u64 dirid)
1400 struct btrfs_key key;
1405 key.objectid = objectid;
1407 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
1409 ret = btrfs_insert_dir_item(trans, root, buf, 1, objectid,
1410 &key, BTRFS_FT_DIR);
1414 key.objectid = dirid;
1415 ret = btrfs_insert_dir_item(trans, root, buf, 2, objectid,
1416 &key, BTRFS_FT_DIR);
1423 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
1425 struct inode *inode;
1426 struct btrfs_trans_handle *trans;
1427 struct btrfs_root *root = BTRFS_I(dir)->root;
1429 int drop_on_err = 0;
1431 unsigned long nr = 1;
1433 mutex_lock(&root->fs_info->fs_mutex);
1434 trans = btrfs_start_transaction(root, 1);
1435 btrfs_set_trans_block_group(trans, dir);
1437 if (IS_ERR(trans)) {
1438 err = PTR_ERR(trans);
1442 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
1448 inode = btrfs_new_inode(trans, root, objectid,
1449 BTRFS_I(dir)->block_group, S_IFDIR | mode);
1450 if (IS_ERR(inode)) {
1451 err = PTR_ERR(inode);
1456 inode->i_op = &btrfs_dir_inode_operations;
1457 inode->i_fop = &btrfs_dir_file_operations;
1458 btrfs_set_trans_block_group(trans, inode);
1460 err = btrfs_make_empty_dir(trans, root, inode->i_ino, dir->i_ino);
1465 err = btrfs_update_inode(trans, root, inode);
1469 err = btrfs_add_link(trans, dentry, inode);
1473 d_instantiate(dentry, inode);
1475 dir->i_sb->s_dirt = 1;
1476 btrfs_update_inode_block_group(trans, inode);
1477 btrfs_update_inode_block_group(trans, dir);
1480 nr = trans->blocks_used;
1481 btrfs_end_transaction(trans, root);
1484 mutex_unlock(&root->fs_info->fs_mutex);
1487 btrfs_btree_balance_dirty(root, nr);
1491 struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
1492 size_t page_offset, u64 start, u64 end,
1498 u64 extent_start = 0;
1500 u64 objectid = inode->i_ino;
1502 int failed_insert = 0;
1503 struct btrfs_path *path;
1504 struct btrfs_root *root = BTRFS_I(inode)->root;
1505 struct btrfs_file_extent_item *item;
1506 struct extent_buffer *leaf;
1507 struct btrfs_key found_key;
1508 struct extent_map *em = NULL;
1509 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1510 struct btrfs_trans_handle *trans = NULL;
1512 path = btrfs_alloc_path();
1514 mutex_lock(&root->fs_info->fs_mutex);
1517 em = lookup_extent_mapping(em_tree, start, end);
1522 em = alloc_extent_map(GFP_NOFS);
1527 em->start = EXTENT_MAP_HOLE;
1528 em->end = EXTENT_MAP_HOLE;
1530 em->bdev = inode->i_sb->s_bdev;
1531 ret = btrfs_lookup_file_extent(NULL, root, path,
1532 objectid, start, 0);
1539 if (path->slots[0] == 0)
1544 leaf = path->nodes[0];
1545 item = btrfs_item_ptr(leaf, path->slots[0],
1546 struct btrfs_file_extent_item);
1547 /* are we inside the extent that was found? */
1548 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1549 found_type = btrfs_key_type(&found_key);
1550 if (found_key.objectid != objectid ||
1551 found_type != BTRFS_EXTENT_DATA_KEY) {
1555 found_type = btrfs_file_extent_type(leaf, item);
1556 extent_start = found_key.offset;
1557 if (found_type == BTRFS_FILE_EXTENT_REG) {
1558 extent_end = extent_start +
1559 btrfs_file_extent_num_bytes(leaf, item);
1561 if (start < extent_start || start >= extent_end) {
1563 if (start < extent_start) {
1564 if (end < extent_start)
1566 em->end = extent_end - 1;
1572 bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
1574 em->start = extent_start;
1575 em->end = extent_end - 1;
1576 em->block_start = EXTENT_MAP_HOLE;
1577 em->block_end = EXTENT_MAP_HOLE;
1580 bytenr += btrfs_file_extent_offset(leaf, item);
1581 em->block_start = bytenr;
1582 em->block_end = em->block_start +
1583 btrfs_file_extent_num_bytes(leaf, item) - 1;
1584 em->start = extent_start;
1585 em->end = extent_end - 1;
1587 } else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
1591 size_t extent_offset;
1594 size = btrfs_file_extent_inline_len(leaf, btrfs_item_nr(leaf,
1597 extent_end = (extent_start + size) |
1598 ((u64)root->sectorsize - 1);
1599 if (start < extent_start || start >= extent_end) {
1601 if (start < extent_start) {
1602 if (end < extent_start)
1604 em->end = extent_end;
1611 extent_offset = (page->index << PAGE_CACHE_SHIFT) -
1613 ptr = btrfs_file_extent_inline_start(item) + extent_offset;
1615 copy_size = min_t(u64, PAGE_CACHE_SIZE - page_offset,
1616 size - extent_offset);
1618 em->block_start = EXTENT_MAP_INLINE;
1619 em->block_end = EXTENT_MAP_INLINE;
1620 em->start = extent_start + extent_offset;
1621 em->end = (em->start + copy_size -1) |
1622 ((u64)root->sectorsize -1);
1628 read_extent_buffer(leaf, map + page_offset, ptr, copy_size);
1630 memset(map + page_offset + copy_size, 0,
1631 PAGE_CACHE_SIZE - copy_size - page_offset);
1633 flush_dcache_page(page);
1635 set_extent_uptodate(em_tree, em->start, em->end, GFP_NOFS);
1638 printk("unkknown found_type %d\n", found_type);
1645 em->block_start = EXTENT_MAP_HOLE;
1646 em->block_end = EXTENT_MAP_HOLE;
1648 btrfs_release_path(root, path);
1649 if (em->start > start || em->end < start) {
1650 printk("bad extent! em: [%Lu %Lu] passed [%Lu %Lu]\n", em->start, em->end, start, end);
1654 ret = add_extent_mapping(em_tree, em);
1655 if (ret == -EEXIST) {
1656 free_extent_map(em);
1659 if (failed_insert > 5) {
1660 printk("failing to insert %Lu %Lu\n", start, end);
1668 btrfs_free_path(path);
1670 ret = btrfs_end_transaction(trans, root);
1674 mutex_unlock(&root->fs_info->fs_mutex);
1676 free_extent_map(em);
1678 return ERR_PTR(err);
1683 static sector_t btrfs_bmap(struct address_space *mapping, sector_t iblock)
1685 return extent_bmap(mapping, iblock, btrfs_get_extent);
1688 static int btrfs_prepare_write(struct file *file, struct page *page,
1689 unsigned from, unsigned to)
1691 return extent_prepare_write(&BTRFS_I(page->mapping->host)->extent_tree,
1692 page->mapping->host, page, from, to,
1696 int btrfs_readpage(struct file *file, struct page *page)
1698 struct extent_map_tree *tree;
1699 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1700 return extent_read_full_page(tree, page, btrfs_get_extent);
1702 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
1704 struct extent_map_tree *tree;
1707 if (current->flags & PF_MEMALLOC) {
1708 redirty_page_for_writepage(wbc, page);
1712 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1713 return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
1716 static int btrfs_releasepage(struct page *page, gfp_t unused_gfp_flags)
1718 struct extent_map_tree *tree;
1721 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1722 ret = try_release_extent_mapping(tree, page);
1724 ClearPagePrivate(page);
1725 set_page_private(page, 0);
1726 page_cache_release(page);
1731 static void btrfs_invalidatepage(struct page *page, unsigned long offset)
1733 struct extent_map_tree *tree;
1735 tree = &BTRFS_I(page->mapping->host)->extent_tree;
1736 extent_invalidatepage(tree, page, offset);
1737 btrfs_releasepage(page, GFP_NOFS);
1741 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
1742 * called from a page fault handler when a page is first dirtied. Hence we must
1743 * be careful to check for EOF conditions here. We set the page up correctly
1744 * for a written page which means we get ENOSPC checking when writing into
1745 * holes and correct delalloc and unwritten extent mapping on filesystems that
1746 * support these features.
1748 * We are not allowed to take the i_mutex here so we have to play games to
1749 * protect against truncate races as the page could now be beyond EOF. Because
1750 * vmtruncate() writes the inode size before removing pages, once we have the
1751 * page lock we can determine safely if the page is beyond EOF. If it is not
1752 * beyond EOF, then the page is guaranteed safe against truncation until we
1755 int btrfs_page_mkwrite(struct vm_area_struct *vma, struct page *page)
1757 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
1763 down_read(&BTRFS_I(inode)->root->snap_sem);
1765 wait_on_page_writeback(page);
1766 size = i_size_read(inode);
1767 page_start = page->index << PAGE_CACHE_SHIFT;
1769 if ((page->mapping != inode->i_mapping) ||
1770 (page_start > size)) {
1771 /* page got truncated out from underneath us */
1775 /* page is wholly or partially inside EOF */
1776 if (page_start + PAGE_CACHE_SIZE > size)
1777 end = size & ~PAGE_CACHE_MASK;
1779 end = PAGE_CACHE_SIZE;
1781 ret = btrfs_cow_one_page(inode, page, end);
1784 up_read(&BTRFS_I(inode)->root->snap_sem);
1789 static void btrfs_truncate(struct inode *inode)
1791 struct btrfs_root *root = BTRFS_I(inode)->root;
1793 struct btrfs_trans_handle *trans;
1796 if (!S_ISREG(inode->i_mode))
1798 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
1801 btrfs_truncate_page(inode->i_mapping, inode->i_size);
1803 mutex_lock(&root->fs_info->fs_mutex);
1804 trans = btrfs_start_transaction(root, 1);
1805 btrfs_set_trans_block_group(trans, inode);
1807 /* FIXME, add redo link to tree so we don't leak on crash */
1808 ret = btrfs_truncate_in_trans(trans, root, inode);
1809 btrfs_update_inode(trans, root, inode);
1810 nr = trans->blocks_used;
1812 ret = btrfs_end_transaction(trans, root);
1814 mutex_unlock(&root->fs_info->fs_mutex);
1815 btrfs_btree_balance_dirty(root, nr);
1818 int btrfs_commit_write(struct file *file, struct page *page,
1819 unsigned from, unsigned to)
1821 return extent_commit_write(&BTRFS_I(page->mapping->host)->extent_tree,
1822 page->mapping->host, page, from, to);
1825 static int create_subvol(struct btrfs_root *root, char *name, int namelen)
1827 struct btrfs_trans_handle *trans;
1828 struct btrfs_key key;
1829 struct btrfs_root_item root_item;
1830 struct btrfs_inode_item *inode_item;
1831 struct extent_buffer *leaf;
1832 struct btrfs_root *new_root;
1833 struct inode *inode;
1838 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
1839 unsigned long nr = 1;
1841 mutex_lock(&root->fs_info->fs_mutex);
1842 trans = btrfs_start_transaction(root, 1);
1845 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0, 0);
1847 return PTR_ERR(leaf);
1849 btrfs_set_header_nritems(leaf, 0);
1850 btrfs_set_header_level(leaf, 0);
1851 btrfs_set_header_bytenr(leaf, leaf->start);
1852 btrfs_set_header_generation(leaf, trans->transid);
1853 btrfs_set_header_owner(leaf, root->root_key.objectid);
1854 write_extent_buffer(leaf, root->fs_info->fsid,
1855 (unsigned long)btrfs_header_fsid(leaf),
1857 btrfs_mark_buffer_dirty(leaf);
1859 inode_item = &root_item.inode;
1860 memset(inode_item, 0, sizeof(*inode_item));
1861 inode_item->generation = cpu_to_le64(1);
1862 inode_item->size = cpu_to_le64(3);
1863 inode_item->nlink = cpu_to_le32(1);
1864 inode_item->nblocks = cpu_to_le64(1);
1865 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
1867 btrfs_set_root_bytenr(&root_item, leaf->start);
1868 btrfs_set_root_level(&root_item, 0);
1869 btrfs_set_root_refs(&root_item, 1);
1870 btrfs_set_root_used(&root_item, 0);
1872 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
1873 root_item.drop_level = 0;
1875 free_extent_buffer(leaf);
1878 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1883 btrfs_set_root_dirid(&root_item, new_dirid);
1885 key.objectid = objectid;
1887 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1888 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1894 * insert the directory item
1896 key.offset = (u64)-1;
1897 dir = root->fs_info->sb->s_root->d_inode;
1898 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1899 name, namelen, dir->i_ino, &key,
1904 ret = btrfs_commit_transaction(trans, root);
1908 new_root = btrfs_read_fs_root(root->fs_info, &key, name, namelen);
1911 trans = btrfs_start_transaction(new_root, 1);
1914 inode = btrfs_new_inode(trans, new_root, new_dirid,
1915 BTRFS_I(dir)->block_group, S_IFDIR | 0700);
1918 inode->i_op = &btrfs_dir_inode_operations;
1919 inode->i_fop = &btrfs_dir_file_operations;
1920 new_root->inode = inode;
1922 ret = btrfs_make_empty_dir(trans, new_root, new_dirid, new_dirid);
1928 ret = btrfs_update_inode(trans, new_root, inode);
1932 nr = trans->blocks_used;
1933 err = btrfs_commit_transaction(trans, root);
1937 mutex_unlock(&root->fs_info->fs_mutex);
1938 btrfs_btree_balance_dirty(root, nr);
1942 static int create_snapshot(struct btrfs_root *root, char *name, int namelen)
1944 struct btrfs_trans_handle *trans;
1945 struct btrfs_key key;
1946 struct btrfs_root_item new_root_item;
1947 struct extent_buffer *tmp;
1953 if (!root->ref_cows)
1956 down_write(&root->snap_sem);
1957 freeze_bdev(root->fs_info->sb->s_bdev);
1958 thaw_bdev(root->fs_info->sb->s_bdev, root->fs_info->sb);
1960 mutex_lock(&root->fs_info->fs_mutex);
1961 trans = btrfs_start_transaction(root, 1);
1964 ret = btrfs_update_inode(trans, root, root->inode);
1968 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
1973 memcpy(&new_root_item, &root->root_item,
1974 sizeof(new_root_item));
1976 key.objectid = objectid;
1978 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
1980 btrfs_cow_block(trans, root, root->node, NULL, 0, &tmp);
1981 btrfs_set_root_bytenr(&new_root_item, root->node->start);
1982 btrfs_set_root_level(&new_root_item, btrfs_header_level(root->node));
1984 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
1990 * insert the directory item
1992 key.offset = (u64)-1;
1993 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
1995 root->fs_info->sb->s_root->d_inode->i_ino,
1996 &key, BTRFS_FT_DIR);
2001 ret = btrfs_inc_root_ref(trans, root);
2005 nr = trans->blocks_used;
2006 err = btrfs_commit_transaction(trans, root);
2011 mutex_unlock(&root->fs_info->fs_mutex);
2012 up_write(&root->snap_sem);
2013 btrfs_btree_balance_dirty(root, nr);
2017 static unsigned long force_ra(struct address_space *mapping,
2018 struct file_ra_state *ra, struct file *file,
2019 pgoff_t offset, pgoff_t last_index)
2023 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2024 req_size = last_index - offset + 1;
2025 offset = page_cache_readahead(mapping, ra, file, offset, req_size);
2028 req_size = min(last_index - offset + 1, (pgoff_t)128);
2029 page_cache_sync_readahead(mapping, ra, file, offset, req_size);
2030 return offset + req_size;
2034 int btrfs_defrag_file(struct file *file) {
2035 struct inode *inode = file->f_path.dentry->d_inode;
2036 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2038 unsigned long last_index;
2039 unsigned long ra_index = 0;
2044 mutex_lock(&inode->i_mutex);
2045 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
2046 for (i = 0; i <= last_index; i++) {
2047 if (i == ra_index) {
2048 ra_index = force_ra(inode->i_mapping, &file->f_ra,
2049 file, ra_index, last_index);
2051 page = grab_cache_page(inode->i_mapping, i);
2054 if (!PageUptodate(page)) {
2055 btrfs_readpage(NULL, page);
2057 if (!PageUptodate(page)) {
2059 page_cache_release(page);
2063 page_start = page->index << PAGE_CACHE_SHIFT;
2064 page_end = page_start + PAGE_CACHE_SIZE - 1;
2066 lock_extent(em_tree, page_start, page_end, GFP_NOFS);
2067 set_extent_delalloc(em_tree, page_start,
2068 page_end, GFP_NOFS);
2069 unlock_extent(em_tree, page_start, page_end, GFP_NOFS);
2070 set_page_dirty(page);
2072 page_cache_release(page);
2073 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
2077 mutex_unlock(&inode->i_mutex);
2081 static int btrfs_ioctl_snap_create(struct btrfs_root *root, void __user *arg)
2083 struct btrfs_ioctl_vol_args vol_args;
2084 struct btrfs_dir_item *di;
2085 struct btrfs_path *path;
2089 if (copy_from_user(&vol_args, arg, sizeof(vol_args)))
2092 namelen = strlen(vol_args.name);
2093 if (namelen > BTRFS_VOL_NAME_MAX)
2095 if (strchr(vol_args.name, '/'))
2098 path = btrfs_alloc_path();
2102 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
2103 mutex_lock(&root->fs_info->fs_mutex);
2104 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
2106 vol_args.name, namelen, 0);
2107 mutex_unlock(&root->fs_info->fs_mutex);
2108 btrfs_free_path(path);
2109 if (di && !IS_ERR(di))
2114 if (root == root->fs_info->tree_root)
2115 return create_subvol(root, vol_args.name, namelen);
2116 return create_snapshot(root, vol_args.name, namelen);
2119 static int btrfs_ioctl_defrag(struct file *file)
2121 struct inode *inode = file->f_path.dentry->d_inode;
2122 struct btrfs_root *root = BTRFS_I(inode)->root;
2124 switch (inode->i_mode & S_IFMT) {
2126 mutex_lock(&root->fs_info->fs_mutex);
2127 btrfs_defrag_root(root, 0);
2128 btrfs_defrag_root(root->fs_info->extent_root, 0);
2129 mutex_unlock(&root->fs_info->fs_mutex);
2132 btrfs_defrag_file(file);
2139 long btrfs_ioctl(struct file *file, unsigned int
2140 cmd, unsigned long arg)
2142 struct btrfs_root *root = BTRFS_I(file->f_path.dentry->d_inode)->root;
2145 case BTRFS_IOC_SNAP_CREATE:
2146 return btrfs_ioctl_snap_create(root, (void __user *)arg);
2147 case BTRFS_IOC_DEFRAG:
2148 return btrfs_ioctl_defrag(file);
2155 * Called inside transaction, so use GFP_NOFS
2157 struct inode *btrfs_alloc_inode(struct super_block *sb)
2159 struct btrfs_inode *ei;
2161 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
2165 return &ei->vfs_inode;
2168 void btrfs_destroy_inode(struct inode *inode)
2170 WARN_ON(!list_empty(&inode->i_dentry));
2171 WARN_ON(inode->i_data.nrpages);
2173 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
2176 static void init_once(void * foo, struct kmem_cache * cachep,
2177 unsigned long flags)
2179 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
2181 inode_init_once(&ei->vfs_inode);
2184 void btrfs_destroy_cachep(void)
2186 if (btrfs_inode_cachep)
2187 kmem_cache_destroy(btrfs_inode_cachep);
2188 if (btrfs_trans_handle_cachep)
2189 kmem_cache_destroy(btrfs_trans_handle_cachep);
2190 if (btrfs_transaction_cachep)
2191 kmem_cache_destroy(btrfs_transaction_cachep);
2192 if (btrfs_bit_radix_cachep)
2193 kmem_cache_destroy(btrfs_bit_radix_cachep);
2194 if (btrfs_path_cachep)
2195 kmem_cache_destroy(btrfs_path_cachep);
2198 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
2199 unsigned long extra_flags,
2200 void (*ctor)(void *, struct kmem_cache *,
2203 return kmem_cache_create(name, size, 0, (SLAB_RECLAIM_ACCOUNT |
2204 SLAB_MEM_SPREAD | extra_flags), ctor
2205 #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,23)
2211 int btrfs_init_cachep(void)
2213 btrfs_inode_cachep = btrfs_cache_create("btrfs_inode_cache",
2214 sizeof(struct btrfs_inode),
2216 if (!btrfs_inode_cachep)
2218 btrfs_trans_handle_cachep =
2219 btrfs_cache_create("btrfs_trans_handle_cache",
2220 sizeof(struct btrfs_trans_handle),
2222 if (!btrfs_trans_handle_cachep)
2224 btrfs_transaction_cachep = btrfs_cache_create("btrfs_transaction_cache",
2225 sizeof(struct btrfs_transaction),
2227 if (!btrfs_transaction_cachep)
2229 btrfs_path_cachep = btrfs_cache_create("btrfs_path_cache",
2230 sizeof(struct btrfs_path),
2232 if (!btrfs_path_cachep)
2234 btrfs_bit_radix_cachep = btrfs_cache_create("btrfs_radix", 256,
2235 SLAB_DESTROY_BY_RCU, NULL);
2236 if (!btrfs_bit_radix_cachep)
2240 btrfs_destroy_cachep();
2244 static int btrfs_getattr(struct vfsmount *mnt,
2245 struct dentry *dentry, struct kstat *stat)
2247 struct inode *inode = dentry->d_inode;
2248 generic_fillattr(inode, stat);
2249 stat->blksize = 256 * 1024;
2253 static int btrfs_rename(struct inode * old_dir, struct dentry *old_dentry,
2254 struct inode * new_dir,struct dentry *new_dentry)
2256 struct btrfs_trans_handle *trans;
2257 struct btrfs_root *root = BTRFS_I(old_dir)->root;
2258 struct inode *new_inode = new_dentry->d_inode;
2259 struct inode *old_inode = old_dentry->d_inode;
2260 struct timespec ctime = CURRENT_TIME;
2261 struct btrfs_path *path;
2262 struct btrfs_dir_item *di;
2265 if (S_ISDIR(old_inode->i_mode) && new_inode &&
2266 new_inode->i_size > BTRFS_EMPTY_DIR_SIZE) {
2270 mutex_lock(&root->fs_info->fs_mutex);
2271 trans = btrfs_start_transaction(root, 1);
2273 btrfs_set_trans_block_group(trans, new_dir);
2274 path = btrfs_alloc_path();
2280 old_dentry->d_inode->i_nlink++;
2281 old_dir->i_ctime = old_dir->i_mtime = ctime;
2282 new_dir->i_ctime = new_dir->i_mtime = ctime;
2283 old_inode->i_ctime = ctime;
2285 if (S_ISDIR(old_inode->i_mode) && old_dir != new_dir) {
2286 struct btrfs_key *location = &BTRFS_I(new_dir)->location;
2287 struct btrfs_key old_parent_key;
2288 di = btrfs_lookup_dir_item(trans, root, path, old_inode->i_ino,
2298 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &old_parent_key);
2299 ret = btrfs_del_item(trans, root, path);
2303 btrfs_release_path(root, path);
2305 di = btrfs_lookup_dir_index_item(trans, root, path,
2307 old_parent_key.objectid,
2317 ret = btrfs_del_item(trans, root, path);
2321 btrfs_release_path(root, path);
2323 ret = btrfs_insert_dir_item(trans, root, "..", 2,
2324 old_inode->i_ino, location,
2331 ret = btrfs_unlink_trans(trans, root, old_dir, old_dentry);
2336 new_inode->i_ctime = CURRENT_TIME;
2337 ret = btrfs_unlink_trans(trans, root, new_dir, new_dentry);
2341 ret = btrfs_add_link(trans, new_dentry, old_inode);
2346 btrfs_free_path(path);
2347 btrfs_end_transaction(trans, root);
2348 mutex_unlock(&root->fs_info->fs_mutex);
2352 static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
2353 const char *symname)
2355 struct btrfs_trans_handle *trans;
2356 struct btrfs_root *root = BTRFS_I(dir)->root;
2357 struct btrfs_path *path;
2358 struct btrfs_key key;
2359 struct inode *inode;
2366 struct btrfs_file_extent_item *ei;
2367 struct extent_buffer *leaf;
2370 name_len = strlen(symname) + 1;
2371 if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
2372 return -ENAMETOOLONG;
2373 mutex_lock(&root->fs_info->fs_mutex);
2374 trans = btrfs_start_transaction(root, 1);
2375 btrfs_set_trans_block_group(trans, dir);
2377 err = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
2383 inode = btrfs_new_inode(trans, root, objectid,
2384 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO);
2385 err = PTR_ERR(inode);
2389 btrfs_set_trans_block_group(trans, inode);
2390 err = btrfs_add_nondir(trans, dentry, inode);
2394 inode->i_mapping->a_ops = &btrfs_aops;
2395 inode->i_fop = &btrfs_file_operations;
2396 inode->i_op = &btrfs_file_inode_operations;
2397 extent_map_tree_init(&BTRFS_I(inode)->extent_tree,
2398 inode->i_mapping, GFP_NOFS);
2399 BTRFS_I(inode)->extent_tree.ops = &btrfs_extent_map_ops;
2401 dir->i_sb->s_dirt = 1;
2402 btrfs_update_inode_block_group(trans, inode);
2403 btrfs_update_inode_block_group(trans, dir);
2407 path = btrfs_alloc_path();
2409 key.objectid = inode->i_ino;
2411 btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
2412 datasize = btrfs_file_extent_calc_inline_size(name_len);
2413 err = btrfs_insert_empty_item(trans, root, path, &key,
2419 leaf = path->nodes[0];
2420 ei = btrfs_item_ptr(leaf, path->slots[0],
2421 struct btrfs_file_extent_item);
2422 btrfs_set_file_extent_generation(leaf, ei, trans->transid);
2423 btrfs_set_file_extent_type(leaf, ei,
2424 BTRFS_FILE_EXTENT_INLINE);
2425 ptr = btrfs_file_extent_inline_start(ei);
2426 write_extent_buffer(leaf, symname, ptr, name_len);
2427 btrfs_mark_buffer_dirty(leaf);
2428 btrfs_free_path(path);
2430 inode->i_op = &btrfs_symlink_inode_operations;
2431 inode->i_mapping->a_ops = &btrfs_symlink_aops;
2432 inode->i_size = name_len - 1;
2433 err = btrfs_update_inode(trans, root, inode);
2438 nr = trans->blocks_used;
2439 btrfs_end_transaction(trans, root);
2440 mutex_unlock(&root->fs_info->fs_mutex);
2442 inode_dec_link_count(inode);
2445 btrfs_btree_balance_dirty(root, nr);
2449 static struct inode_operations btrfs_dir_inode_operations = {
2450 .lookup = btrfs_lookup,
2451 .create = btrfs_create,
2452 .unlink = btrfs_unlink,
2454 .mkdir = btrfs_mkdir,
2455 .rmdir = btrfs_rmdir,
2456 .rename = btrfs_rename,
2457 .symlink = btrfs_symlink,
2458 .setattr = btrfs_setattr,
2459 .mknod = btrfs_mknod,
2462 static struct inode_operations btrfs_dir_ro_inode_operations = {
2463 .lookup = btrfs_lookup,
2466 static struct file_operations btrfs_dir_file_operations = {
2467 .llseek = generic_file_llseek,
2468 .read = generic_read_dir,
2469 .readdir = btrfs_readdir,
2470 .unlocked_ioctl = btrfs_ioctl,
2471 #ifdef CONFIG_COMPAT
2472 .compat_ioctl = btrfs_ioctl,
2476 static struct extent_map_ops btrfs_extent_map_ops = {
2477 .fill_delalloc = run_delalloc_range,
2478 .writepage_io_hook = btrfs_writepage_io_hook,
2479 .readpage_io_hook = btrfs_readpage_io_hook,
2480 .readpage_end_io_hook = btrfs_readpage_end_io_hook,
2483 static struct address_space_operations btrfs_aops = {
2484 .readpage = btrfs_readpage,
2485 .writepage = btrfs_writepage,
2486 .sync_page = block_sync_page,
2487 .prepare_write = btrfs_prepare_write,
2488 .commit_write = btrfs_commit_write,
2490 .invalidatepage = btrfs_invalidatepage,
2491 .releasepage = btrfs_releasepage,
2492 .set_page_dirty = __set_page_dirty_nobuffers,
2495 static struct address_space_operations btrfs_symlink_aops = {
2496 .readpage = btrfs_readpage,
2497 .writepage = btrfs_writepage,
2498 .invalidatepage = btrfs_invalidatepage,
2499 .releasepage = btrfs_releasepage,
2502 static struct inode_operations btrfs_file_inode_operations = {
2503 .truncate = btrfs_truncate,
2504 .getattr = btrfs_getattr,
2505 .setattr = btrfs_setattr,
2508 static struct inode_operations btrfs_special_inode_operations = {
2509 .getattr = btrfs_getattr,
2510 .setattr = btrfs_setattr,
2513 static struct inode_operations btrfs_symlink_inode_operations = {
2514 .readlink = generic_readlink,
2515 .follow_link = page_follow_link_light,
2516 .put_link = page_put_link,