1 #include <linux/module.h>
2 #include <linux/buffer_head.h>
4 #include <linux/pagemap.h>
5 #include <linux/highmem.h>
6 #include <linux/time.h>
7 #include <linux/init.h>
8 #include <linux/string.h>
9 #include <linux/smp_lock.h>
10 #include <linux/backing-dev.h>
11 #include <linux/mpage.h>
12 #include <linux/swap.h>
13 #include <linux/writeback.h>
16 #include "transaction.h"
17 #include "btrfs_inode.h"
19 #define BTRFS_SUPER_MAGIC 0x9123682E
21 static struct inode_operations btrfs_dir_inode_operations;
22 static struct super_operations btrfs_super_ops;
23 static struct file_operations btrfs_dir_file_operations;
24 static struct inode_operations btrfs_file_inode_operations;
25 static struct address_space_operations btrfs_aops;
26 static struct file_operations btrfs_file_operations;
28 static int check_inode(struct inode *inode)
30 struct btrfs_inode *ei = BTRFS_I(inode);
31 WARN_ON(ei->magic != 0xDEADBEEF);
32 WARN_ON(ei->magic2 != 0xDEADBEAF);
36 static void btrfs_read_locked_inode(struct inode *inode)
38 struct btrfs_path path;
39 struct btrfs_inode_item *inode_item;
40 struct btrfs_root *root = btrfs_sb(inode->i_sb);
43 btrfs_init_path(&path);
44 mutex_lock(&root->fs_info->fs_mutex);
47 ret = btrfs_lookup_inode(NULL, root, &path, inode->i_ino, 0);
49 btrfs_release_path(root, &path);
50 mutex_unlock(&root->fs_info->fs_mutex);
51 make_bad_inode(inode);
55 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
57 struct btrfs_inode_item);
59 inode->i_mode = btrfs_inode_mode(inode_item);
60 inode->i_nlink = btrfs_inode_nlink(inode_item);
61 inode->i_uid = btrfs_inode_uid(inode_item);
62 inode->i_gid = btrfs_inode_gid(inode_item);
63 inode->i_size = btrfs_inode_size(inode_item);
64 inode->i_atime.tv_sec = btrfs_timespec_sec(&inode_item->atime);
65 inode->i_atime.tv_nsec = btrfs_timespec_nsec(&inode_item->atime);
66 inode->i_mtime.tv_sec = btrfs_timespec_sec(&inode_item->mtime);
67 inode->i_mtime.tv_nsec = btrfs_timespec_nsec(&inode_item->mtime);
68 inode->i_ctime.tv_sec = btrfs_timespec_sec(&inode_item->ctime);
69 inode->i_ctime.tv_nsec = btrfs_timespec_nsec(&inode_item->ctime);
70 inode->i_blocks = btrfs_inode_nblocks(inode_item);
71 inode->i_generation = btrfs_inode_generation(inode_item);
72 btrfs_release_path(root, &path);
73 mutex_unlock(&root->fs_info->fs_mutex);
75 switch (inode->i_mode & S_IFMT) {
78 init_special_inode(inode, inode->i_mode,
79 btrfs_inode_rdev(inode_item));
83 inode->i_mapping->a_ops = &btrfs_aops;
84 inode->i_fop = &btrfs_file_operations;
85 inode->i_op = &btrfs_file_inode_operations;
88 inode->i_op = &btrfs_dir_inode_operations;
89 inode->i_fop = &btrfs_dir_file_operations;
92 // inode->i_op = &page_symlink_inode_operations;
99 static int btrfs_unlink_trans(struct btrfs_trans_handle *trans,
100 struct btrfs_root *root,
102 struct dentry *dentry)
104 struct btrfs_path path;
105 const char *name = dentry->d_name.name;
106 int name_len = dentry->d_name.len;
109 struct btrfs_dir_item *di;
111 btrfs_init_path(&path);
112 ret = btrfs_lookup_dir_item(trans, root, &path, dir->i_ino,
120 di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
121 struct btrfs_dir_item);
122 objectid = btrfs_dir_objectid(di);
124 ret = btrfs_del_item(trans, root, &path);
126 dentry->d_inode->i_ctime = dir->i_ctime;
128 btrfs_release_path(root, &path);
130 inode_dec_link_count(dentry->d_inode);
134 static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
136 struct btrfs_root *root;
137 struct btrfs_trans_handle *trans;
140 root = btrfs_sb(dir->i_sb);
141 mutex_lock(&root->fs_info->fs_mutex);
142 trans = btrfs_start_transaction(root, 1);
143 ret = btrfs_unlink_trans(trans, root, dir, dentry);
144 btrfs_end_transaction(trans, root);
145 mutex_unlock(&root->fs_info->fs_mutex);
149 static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
151 struct inode *inode = dentry->d_inode;
154 struct btrfs_root *root = btrfs_sb(dir->i_sb);
155 struct btrfs_path path;
156 struct btrfs_key key;
157 struct btrfs_trans_handle *trans;
158 struct btrfs_disk_key *found_key;
159 struct btrfs_leaf *leaf;
161 btrfs_init_path(&path);
162 mutex_lock(&root->fs_info->fs_mutex);
163 trans = btrfs_start_transaction(root, 1);
164 key.objectid = inode->i_ino;
165 key.offset = (u64)-1;
167 btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
168 ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
175 BUG_ON(path.slots[0] == 0);
177 leaf = btrfs_buffer_leaf(path.nodes[0]);
178 found_key = &leaf->items[path.slots[0]].key;
179 if (btrfs_disk_key_objectid(found_key) != inode->i_ino) {
183 if (btrfs_disk_key_type(found_key) != BTRFS_DIR_ITEM_KEY ||
184 btrfs_disk_key_offset(found_key) != 2) {
188 ret = btrfs_del_item(trans, root, &path);
190 btrfs_release_path(root, &path);
192 ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
201 ret = btrfs_del_item(trans, root, &path);
206 btrfs_release_path(root, &path);
208 /* now the directory is empty */
209 err = btrfs_unlink_trans(trans, root, dir, dentry);
214 mutex_unlock(&root->fs_info->fs_mutex);
215 ret = btrfs_end_transaction(trans, root);
221 static int btrfs_free_inode(struct btrfs_trans_handle *trans,
222 struct btrfs_root *root,
225 u64 objectid = inode->i_ino;
226 struct btrfs_path path;
227 struct btrfs_inode_map_item *map;
228 struct btrfs_key stat_data_key;
231 btrfs_init_path(&path);
232 ret = btrfs_lookup_inode_map(trans, root, &path, objectid, -1);
236 btrfs_release_path(root, &path);
239 map = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
240 struct btrfs_inode_map_item);
241 btrfs_disk_key_to_cpu(&stat_data_key, &map->key);
242 ret = btrfs_del_item(trans, root->fs_info->inode_root, &path);
244 btrfs_release_path(root, &path);
245 btrfs_init_path(&path);
247 ret = btrfs_lookup_inode(trans, root, &path, objectid, -1);
249 ret = btrfs_del_item(trans, root, &path);
251 btrfs_release_path(root, &path);
256 static int btrfs_truncate_in_trans(struct btrfs_trans_handle *trans,
257 struct btrfs_root *root,
261 struct btrfs_path path;
262 struct btrfs_key key;
263 struct btrfs_disk_key *found_key;
264 struct btrfs_leaf *leaf;
265 struct btrfs_file_extent_item *fi = NULL;
266 u64 extent_start = 0;
267 u64 extent_num_blocks = 0;
270 /* FIXME, add redo link to tree so we don't leak on crash */
271 key.objectid = inode->i_ino;
272 key.offset = (u64)-1;
274 btrfs_set_key_type(&key, BTRFS_CSUM_ITEM_KEY);
276 btrfs_init_path(&path);
277 ret = btrfs_search_slot(trans, root, &key, &path, -1, 1);
279 btrfs_release_path(root, &path);
283 BUG_ON(path.slots[0] == 0);
286 leaf = btrfs_buffer_leaf(path.nodes[0]);
287 found_key = &leaf->items[path.slots[0]].key;
288 if (btrfs_disk_key_objectid(found_key) != inode->i_ino)
290 if (btrfs_disk_key_type(found_key) != BTRFS_CSUM_ITEM_KEY &&
291 btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY)
293 if (btrfs_disk_key_offset(found_key) < inode->i_size)
295 if (btrfs_disk_key_type(found_key) == BTRFS_EXTENT_DATA_KEY) {
296 fi = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
298 struct btrfs_file_extent_item);
299 extent_start = btrfs_file_extent_disk_blocknr(fi);
301 btrfs_file_extent_disk_num_blocks(fi);
303 btrfs_file_extent_num_blocks(fi) >> 9;
308 ret = btrfs_del_item(trans, root, &path);
310 btrfs_release_path(root, &path);
312 ret = btrfs_free_extent(trans, root, extent_start,
313 extent_num_blocks, 0);
317 btrfs_release_path(root, &path);
323 static void btrfs_delete_inode(struct inode *inode)
325 struct btrfs_trans_handle *trans;
326 struct btrfs_root *root = btrfs_sb(inode->i_sb);
329 truncate_inode_pages(&inode->i_data, 0);
330 if (is_bad_inode(inode)) {
334 mutex_lock(&root->fs_info->fs_mutex);
335 trans = btrfs_start_transaction(root, 1);
336 if (S_ISREG(inode->i_mode)) {
337 ret = btrfs_truncate_in_trans(trans, root, inode);
340 btrfs_free_inode(trans, root, inode);
341 btrfs_end_transaction(trans, root);
342 mutex_unlock(&root->fs_info->fs_mutex);
348 static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
351 const char *name = dentry->d_name.name;
352 int namelen = dentry->d_name.len;
353 struct btrfs_dir_item *di;
354 struct btrfs_path path;
355 struct btrfs_root *root = btrfs_sb(dir->i_sb);
358 btrfs_init_path(&path);
359 ret = btrfs_lookup_dir_item(NULL, root, &path, dir->i_ino, name,
361 if (ret || !btrfs_match_dir_item_name(root, &path, name, namelen)) {
366 di = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
367 struct btrfs_dir_item);
368 *ino = btrfs_dir_objectid(di);
370 btrfs_release_path(root, &path);
375 static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
376 struct nameidata *nd)
378 struct inode * inode;
379 struct btrfs_root *root = btrfs_sb(dir->i_sb);
383 if (dentry->d_name.len > BTRFS_NAME_LEN)
384 return ERR_PTR(-ENAMETOOLONG);
385 mutex_lock(&root->fs_info->fs_mutex);
386 ret = btrfs_inode_by_name(dir, dentry, &ino);
387 mutex_unlock(&root->fs_info->fs_mutex);
392 inode = iget(dir->i_sb, ino);
394 return ERR_PTR(-EACCES);
398 return d_splice_alias(inode, dentry);
401 static int btrfs_readdir(struct file *filp, void *dirent, filldir_t filldir)
403 struct inode *inode = filp->f_path.dentry->d_inode;
404 struct btrfs_root *root = btrfs_sb(inode->i_sb);
405 struct btrfs_item *item;
406 struct btrfs_dir_item *di;
407 struct btrfs_key key;
408 struct btrfs_path path;
411 struct btrfs_leaf *leaf;
414 unsigned char d_type = DT_UNKNOWN;
417 mutex_lock(&root->fs_info->fs_mutex);
418 key.objectid = inode->i_ino;
420 btrfs_set_key_type(&key, BTRFS_DIR_ITEM_KEY);
421 key.offset = filp->f_pos;
422 btrfs_init_path(&path);
423 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
429 leaf = btrfs_buffer_leaf(path.nodes[0]);
430 nritems = btrfs_header_nritems(&leaf->header);
431 slot = path.slots[0];
432 if (advance || slot >= nritems) {
433 if (slot >= nritems -1) {
434 ret = btrfs_next_leaf(root, &path);
437 leaf = btrfs_buffer_leaf(path.nodes[0]);
438 nritems = btrfs_header_nritems(&leaf->header);
439 slot = path.slots[0];
446 item = leaf->items + slot;
447 if (btrfs_disk_key_objectid(&item->key) != key.objectid)
449 if (btrfs_disk_key_type(&item->key) != BTRFS_DIR_ITEM_KEY)
451 if (btrfs_disk_key_offset(&item->key) < filp->f_pos)
455 di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
456 over = filldir(dirent, (const char *)(di + 1),
457 btrfs_dir_name_len(di),
458 btrfs_disk_key_offset(&item->key),
459 btrfs_dir_objectid(di), d_type);
461 filp->f_pos = btrfs_disk_key_offset(&item->key);
464 filp->f_pos = btrfs_disk_key_offset(&item->key) + 1;
468 btrfs_release_path(root, &path);
469 mutex_unlock(&root->fs_info->fs_mutex);
473 static void btrfs_put_super (struct super_block * sb)
475 struct btrfs_root *root = btrfs_sb(sb);
478 ret = close_ctree(root);
480 printk("close ctree returns %d\n", ret);
482 sb->s_fs_info = NULL;
485 static int btrfs_fill_super(struct super_block * sb, void * data, int silent)
487 struct inode * inode;
488 struct dentry * root_dentry;
489 struct btrfs_super_block *disk_super;
490 struct btrfs_root *root;
492 sb->s_maxbytes = MAX_LFS_FILESIZE;
493 sb->s_magic = BTRFS_SUPER_MAGIC;
494 sb->s_op = &btrfs_super_ops;
497 root = open_ctree(sb);
500 printk("btrfs: open_ctree failed\n");
503 sb->s_fs_info = root;
504 disk_super = root->fs_info->disk_super;
505 printk("read in super total blocks %Lu root %Lu\n",
506 btrfs_super_total_blocks(disk_super),
507 btrfs_super_root_dir(disk_super));
509 inode = iget_locked(sb, btrfs_super_root_dir(disk_super));
512 if (inode->i_state & I_NEW) {
513 btrfs_read_locked_inode(inode);
514 unlock_new_inode(inode);
517 root_dentry = d_alloc_root(inode);
522 sb->s_root = root_dentry;
527 static void fill_inode_item(struct btrfs_inode_item *item,
530 btrfs_set_inode_uid(item, inode->i_uid);
531 btrfs_set_inode_gid(item, inode->i_gid);
532 btrfs_set_inode_size(item, inode->i_size);
533 btrfs_set_inode_mode(item, inode->i_mode);
534 btrfs_set_inode_nlink(item, inode->i_nlink);
535 btrfs_set_timespec_sec(&item->atime, inode->i_atime.tv_sec);
536 btrfs_set_timespec_nsec(&item->atime, inode->i_atime.tv_nsec);
537 btrfs_set_timespec_sec(&item->mtime, inode->i_mtime.tv_sec);
538 btrfs_set_timespec_nsec(&item->mtime, inode->i_mtime.tv_nsec);
539 btrfs_set_timespec_sec(&item->ctime, inode->i_ctime.tv_sec);
540 btrfs_set_timespec_nsec(&item->ctime, inode->i_ctime.tv_nsec);
541 btrfs_set_inode_nblocks(item, inode->i_blocks);
542 btrfs_set_inode_generation(item, inode->i_generation);
546 static int btrfs_update_inode(struct btrfs_trans_handle *trans,
547 struct btrfs_root *root,
550 struct btrfs_inode_item *inode_item;
551 struct btrfs_path path;
554 btrfs_init_path(&path);
556 ret = btrfs_lookup_inode(trans, root, &path, inode->i_ino, 1);
563 inode_item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]),
565 struct btrfs_inode_item);
567 fill_inode_item(inode_item, inode);
568 btrfs_mark_buffer_dirty(path.nodes[0]);
570 btrfs_release_path(root, &path);
575 static int btrfs_write_inode(struct inode *inode, int wait)
577 struct btrfs_root *root = btrfs_sb(inode->i_sb);
578 struct btrfs_trans_handle *trans;
581 mutex_lock(&root->fs_info->fs_mutex);
582 trans = btrfs_start_transaction(root, 1);
583 ret = btrfs_update_inode(trans, root, inode);
585 btrfs_commit_transaction(trans, root);
587 btrfs_end_transaction(trans, root);
588 mutex_unlock(&root->fs_info->fs_mutex);
593 static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
594 struct inode *dir, int mode)
597 struct btrfs_inode_item inode_item;
598 struct btrfs_root *root = btrfs_sb(dir->i_sb);
599 struct btrfs_key key;
603 inode = new_inode(dir->i_sb);
605 return ERR_PTR(-ENOMEM);
608 ret = btrfs_find_free_objectid(trans, root, dir->i_ino, &objectid);
611 inode->i_uid = current->fsuid;
612 inode->i_gid = current->fsgid;
613 inode->i_mode = mode;
614 inode->i_ino = objectid;
616 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME_SEC;
617 fill_inode_item(&inode_item, inode);
619 key.objectid = objectid;
622 btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
623 ret = btrfs_insert_inode_map(trans, root, objectid, &key);
626 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
629 insert_inode_hash(inode);
635 static int btrfs_add_link(struct btrfs_trans_handle *trans,
636 struct dentry *dentry, struct inode *inode)
639 ret = btrfs_insert_dir_item(trans, btrfs_sb(inode->i_sb),
640 dentry->d_name.name, dentry->d_name.len,
641 dentry->d_parent->d_inode->i_ino,
644 dentry->d_parent->d_inode->i_size += dentry->d_name.len;
645 ret = btrfs_update_inode(trans, btrfs_sb(inode->i_sb),
646 dentry->d_parent->d_inode);
649 check_inode(dentry->d_parent->d_inode);
653 static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
654 struct dentry *dentry, struct inode *inode)
656 int err = btrfs_add_link(trans, dentry, inode);
658 d_instantiate(dentry, inode);
667 static int btrfs_create(struct inode *dir, struct dentry *dentry,
668 int mode, struct nameidata *nd)
670 struct btrfs_trans_handle *trans;
671 struct btrfs_root *root = btrfs_sb(dir->i_sb);
676 mutex_lock(&root->fs_info->fs_mutex);
677 trans = btrfs_start_transaction(root, 1);
678 inode = btrfs_new_inode(trans, dir, mode);
679 err = PTR_ERR(inode);
682 // FIXME mark the inode dirty
683 err = btrfs_add_nondir(trans, dentry, inode);
687 inode->i_mapping->a_ops = &btrfs_aops;
688 inode->i_fop = &btrfs_file_operations;
689 inode->i_op = &btrfs_file_inode_operations;
691 dir->i_sb->s_dirt = 1;
693 btrfs_end_transaction(trans, root);
694 mutex_unlock(&root->fs_info->fs_mutex);
699 inode_dec_link_count(inode);
705 static int btrfs_make_empty_dir(struct btrfs_trans_handle *trans,
706 struct inode *inode, struct inode *dir)
708 struct btrfs_root *root = btrfs_sb(inode->i_sb);
714 ret = btrfs_insert_dir_item(trans, root, buf, 1, inode->i_ino,
718 ret = btrfs_insert_dir_item(trans, root, buf, 2, inode->i_ino,
723 ret = btrfs_update_inode(trans, root, inode);
728 static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
731 struct btrfs_trans_handle *trans;
732 struct btrfs_root *root = btrfs_sb(dir->i_sb);
736 mutex_lock(&root->fs_info->fs_mutex);
737 trans = btrfs_start_transaction(root, 1);
739 err = PTR_ERR(trans);
742 inode = btrfs_new_inode(trans, dir, S_IFDIR | mode);
744 err = PTR_ERR(inode);
748 inode->i_op = &btrfs_dir_inode_operations;
749 inode->i_fop = &btrfs_dir_file_operations;
751 err = btrfs_make_empty_dir(trans, inode, dir);
754 err = btrfs_add_link(trans, dentry, inode);
757 d_instantiate(dentry, inode);
761 btrfs_end_transaction(trans, root);
763 mutex_unlock(&root->fs_info->fs_mutex);
769 static int btrfs_sync_fs(struct super_block *sb, int wait)
771 struct btrfs_trans_handle *trans;
772 struct btrfs_root *root;
778 // filemap_flush(root->fs_info->btree_inode->i_mapping);
779 filemap_flush(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
782 filemap_write_and_wait(root->fs_info->sb->s_bdev->bd_inode->i_mapping);
783 mutex_lock(&root->fs_info->fs_mutex);
784 trans = btrfs_start_transaction(root, 1);
785 ret = btrfs_commit_transaction(trans, root);
788 printk("btrfs sync_fs\n");
789 mutex_unlock(&root->fs_info->fs_mutex);
793 static int btrfs_get_block_lock(struct inode *inode, sector_t iblock,
794 struct buffer_head *result, int create)
799 u64 extent_start = 0;
801 u64 objectid = inode->i_ino;
802 struct btrfs_path path;
803 struct btrfs_root *root = btrfs_sb(inode->i_sb);
804 struct btrfs_trans_handle *trans = NULL;
805 struct btrfs_file_extent_item *item;
806 struct btrfs_leaf *leaf;
807 struct btrfs_disk_key *found_key;
809 btrfs_init_path(&path);
811 trans = btrfs_start_transaction(root, 1);
814 ret = btrfs_lookup_file_extent(trans, root, &path,
816 iblock << inode->i_blkbits, 0);
818 btrfs_release_path(root, &path);
824 if (path.slots[0] == 0) {
825 btrfs_release_path(root, &path);
831 item = btrfs_item_ptr(btrfs_buffer_leaf(path.nodes[0]), path.slots[0],
832 struct btrfs_file_extent_item);
833 leaf = btrfs_buffer_leaf(path.nodes[0]);
834 blocknr = btrfs_file_extent_disk_blocknr(item);
835 blocknr += btrfs_file_extent_offset(item);
837 /* exact match found, use it */
840 map_bh(result, inode->i_sb, blocknr);
841 btrfs_release_path(root, &path);
845 /* are we inside the extent that was found? */
846 found_key = &leaf->items[path.slots[0]].key;
847 if (btrfs_disk_key_objectid(found_key) != objectid ||
848 btrfs_disk_key_type(found_key) != BTRFS_EXTENT_DATA_KEY) {
851 btrfs_release_path(root, &path);
855 extent_start = btrfs_disk_key_offset(&leaf->items[path.slots[0]].key);
856 extent_start = extent_start >> inode->i_blkbits;
857 extent_start += btrfs_file_extent_offset(item);
858 extent_end = extent_start + btrfs_file_extent_num_blocks(item);
859 btrfs_release_path(root, &path);
860 if (iblock >= extent_start && iblock < extent_end) {
862 map_bh(result, inode->i_sb, blocknr + iblock - extent_start);
866 /* ok, create a new extent */
871 ret = btrfs_alloc_file_extent(trans, root, objectid,
872 iblock << inode->i_blkbits,
873 1, extent_end, &blocknr);
878 inode->i_blocks += inode->i_sb->s_blocksize >> 9;
879 set_buffer_new(result);
880 map_bh(result, inode->i_sb, blocknr);
884 btrfs_end_transaction(trans, root);
888 static int btrfs_get_block(struct inode *inode, sector_t iblock,
889 struct buffer_head *result, int create)
892 struct btrfs_root *root = btrfs_sb(inode->i_sb);
893 mutex_lock(&root->fs_info->fs_mutex);
894 err = btrfs_get_block_lock(inode, iblock, result, create);
895 mutex_unlock(&root->fs_info->fs_mutex);
899 static int btrfs_prepare_write(struct file *file, struct page *page,
900 unsigned from, unsigned to)
903 return nobh_prepare_write(page, from, to, btrfs_get_block);
905 static int btrfs_commit_write(struct file *file, struct page *page,
906 unsigned from, unsigned to)
909 return nobh_commit_write(file, page, from, to);
912 static void btrfs_write_super(struct super_block *sb)
914 btrfs_sync_fs(sb, 1);
917 static int btrfs_readpage(struct file *file, struct page *page)
919 return mpage_readpage(page, btrfs_get_block);
922 static int btrfs_readpages(struct file *file, struct address_space *mapping,
923 struct list_head *pages, unsigned nr_pages)
925 return mpage_readpages(mapping, pages, nr_pages, btrfs_get_block);
928 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
930 return nobh_writepage(page, btrfs_get_block, wbc);
933 static void btrfs_truncate(struct inode *inode)
935 struct btrfs_root *root = btrfs_sb(inode->i_sb);
937 struct btrfs_trans_handle *trans;
939 if (!S_ISREG(inode->i_mode))
941 if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
944 nobh_truncate_page(inode->i_mapping, inode->i_size);
946 /* FIXME, add redo link to tree so we don't leak on crash */
947 mutex_lock(&root->fs_info->fs_mutex);
948 trans = btrfs_start_transaction(root, 1);
949 ret = btrfs_truncate_in_trans(trans, root, inode);
951 ret = btrfs_end_transaction(trans, root);
953 mutex_unlock(&root->fs_info->fs_mutex);
954 mark_inode_dirty(inode);
957 static int btrfs_copy_from_user(loff_t pos, int num_pages, int write_bytes,
958 struct page **prepared_pages,
959 const char __user * buf)
963 int offset = pos & (PAGE_CACHE_SIZE - 1);
965 for (i = 0; i < num_pages && write_bytes > 0; i++, offset = 0) {
966 size_t count = min_t(size_t,
967 PAGE_CACHE_SIZE - offset, write_bytes);
968 struct page *page = prepared_pages[i];
969 fault_in_pages_readable(buf, count);
971 /* Copy data from userspace to the current page */
973 page_fault = __copy_from_user(page_address(page) + offset,
975 /* Flush processor's dcache for this page */
976 flush_dcache_page(page);
979 write_bytes -= count;
984 return page_fault ? -EFAULT : 0;
987 static void btrfs_drop_pages(struct page **pages, size_t num_pages)
990 for (i = 0; i < num_pages; i++) {
993 unlock_page(pages[i]);
994 mark_page_accessed(pages[i]);
995 page_cache_release(pages[i]);
998 static int dirty_and_release_pages(struct btrfs_trans_handle *trans,
999 struct btrfs_root *root,
1001 struct page **pages,
1011 struct inode *inode = file->f_path.dentry->d_inode;
1013 for (i = 0; i < num_pages; i++) {
1014 offset = pos & (PAGE_CACHE_SIZE -1);
1015 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
1016 /* FIXME, one block at a time */
1018 mutex_lock(&root->fs_info->fs_mutex);
1019 trans = btrfs_start_transaction(root, 1);
1020 btrfs_csum_file_block(trans, root, inode->i_ino,
1021 pages[i]->index << PAGE_CACHE_SHIFT,
1022 kmap(pages[i]), PAGE_CACHE_SIZE);
1024 SetPageChecked(pages[i]);
1025 ret = btrfs_end_transaction(trans, root);
1027 mutex_unlock(&root->fs_info->fs_mutex);
1029 ret = nobh_commit_write(file, pages[i], offset,
1030 offset + this_write);
1036 WARN_ON(this_write > write_bytes);
1037 write_bytes -= this_write;
1043 static int prepare_pages(struct btrfs_trans_handle *trans,
1044 struct btrfs_root *root,
1046 struct page **pages,
1052 unsigned long index = pos >> PAGE_CACHE_SHIFT;
1053 struct inode *inode = file->f_path.dentry->d_inode;
1058 loff_t isize = i_size_read(inode);
1060 memset(pages, 0, num_pages * sizeof(struct page *));
1062 for (i = 0; i < num_pages; i++) {
1063 pages[i] = grab_cache_page(inode->i_mapping, index + i);
1066 goto failed_release;
1068 offset = pos & (PAGE_CACHE_SIZE -1);
1069 this_write = min(PAGE_CACHE_SIZE - offset, write_bytes);
1070 ret = nobh_prepare_write(pages[i], offset,
1071 offset + this_write,
1076 goto failed_truncate;
1078 WARN_ON(this_write > write_bytes);
1079 write_bytes -= this_write;
1084 btrfs_drop_pages(pages, num_pages);
1088 btrfs_drop_pages(pages, num_pages);
1090 vmtruncate(inode, isize);
1094 static ssize_t btrfs_file_write(struct file *file, const char __user *buf,
1095 size_t count, loff_t *ppos)
1098 size_t num_written = 0;
1101 struct inode *inode = file->f_path.dentry->d_inode;
1102 struct btrfs_root *root = btrfs_sb(inode->i_sb);
1103 struct page *pages[1];
1105 if (file->f_flags & O_DIRECT)
1109 vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
1110 current->backing_dev_info = inode->i_mapping->backing_dev_info;
1111 err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
1116 err = remove_suid(file->f_path.dentry);
1119 file_update_time(file);
1120 mutex_lock(&inode->i_mutex);
1122 size_t offset = pos & (PAGE_CACHE_SIZE - 1);
1123 size_t write_bytes = min(count, PAGE_CACHE_SIZE - offset);
1124 size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
1126 ret = prepare_pages(NULL, root, file, pages, num_pages,
1129 ret = btrfs_copy_from_user(pos, num_pages,
1130 write_bytes, pages, buf);
1133 ret = dirty_and_release_pages(NULL, root, file, pages,
1134 num_pages, pos, write_bytes);
1136 btrfs_drop_pages(pages, num_pages);
1139 count -= write_bytes;
1141 num_written += write_bytes;
1143 balance_dirty_pages_ratelimited(inode->i_mapping);
1146 mutex_unlock(&inode->i_mutex);
1149 current->backing_dev_info = NULL;
1150 return num_written ? num_written : err;
1153 static int btrfs_read_actor(read_descriptor_t *desc, struct page *page,
1154 unsigned long offset, unsigned long size)
1157 unsigned long left, count = desc->count;
1162 if (!PageChecked(page)) {
1163 /* FIXME, do it per block */
1164 struct btrfs_root *root = btrfs_sb(page->mapping->host->i_sb);
1165 int ret = btrfs_csum_verify_file_block(root,
1166 page->mapping->host->i_ino,
1167 page->index << PAGE_CACHE_SHIFT,
1168 kmap(page), PAGE_CACHE_SIZE);
1170 printk("failed to verify ino %lu page %lu\n",
1171 page->mapping->host->i_ino,
1173 memset(page_address(page), 0, PAGE_CACHE_SIZE);
1175 SetPageChecked(page);
1179 * Faults on the destination of a read are common, so do it before
1182 if (!fault_in_pages_writeable(desc->arg.buf, size)) {
1183 kaddr = kmap_atomic(page, KM_USER0);
1184 left = __copy_to_user_inatomic(desc->arg.buf,
1185 kaddr + offset, size);
1186 kunmap_atomic(kaddr, KM_USER0);
1191 /* Do it the slow way */
1193 left = __copy_to_user(desc->arg.buf, kaddr + offset, size);
1198 desc->error = -EFAULT;
1201 desc->count = count - size;
1202 desc->written += size;
1203 desc->arg.buf += size;
1208 * btrfs_file_aio_read - filesystem read routine
1209 * @iocb: kernel I/O control block
1210 * @iov: io vector request
1211 * @nr_segs: number of segments in the iovec
1212 * @pos: current file position
1214 static ssize_t btrfs_file_aio_read(struct kiocb *iocb, const struct iovec *iov,
1215 unsigned long nr_segs, loff_t pos)
1217 struct file *filp = iocb->ki_filp;
1221 loff_t *ppos = &iocb->ki_pos;
1224 for (seg = 0; seg < nr_segs; seg++) {
1225 const struct iovec *iv = &iov[seg];
1228 * If any segment has a negative length, or the cumulative
1229 * length ever wraps negative then return -EINVAL.
1231 count += iv->iov_len;
1232 if (unlikely((ssize_t)(count|iv->iov_len) < 0))
1234 if (access_ok(VERIFY_WRITE, iv->iov_base, iv->iov_len))
1239 count -= iv->iov_len; /* This segment is no good */
1244 for (seg = 0; seg < nr_segs; seg++) {
1245 read_descriptor_t desc;
1248 desc.arg.buf = iov[seg].iov_base;
1249 desc.count = iov[seg].iov_len;
1250 if (desc.count == 0)
1253 do_generic_file_read(filp, ppos, &desc,
1255 retval += desc.written;
1257 retval = retval ?: desc.error;
1265 static struct kmem_cache *btrfs_inode_cachep;
1266 struct kmem_cache *btrfs_trans_handle_cachep;
1267 struct kmem_cache *btrfs_transaction_cachep;
1268 struct kmem_cache *btrfs_bit_radix_cachep;
1269 struct kmem_cache *btrfs_path_cachep;
1272 * Called inside transaction, so use GFP_NOFS
1274 static struct inode *btrfs_alloc_inode(struct super_block *sb)
1276 struct btrfs_inode *ei;
1278 ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
1281 ei->magic = 0xDEADBEEF;
1282 ei->magic2 = 0xDEADBEAF;
1283 return &ei->vfs_inode;
1286 static void btrfs_destroy_inode(struct inode *inode)
1288 struct btrfs_inode *ei = BTRFS_I(inode);
1289 WARN_ON(ei->magic != 0xDEADBEEF);
1290 WARN_ON(ei->magic2 != 0xDEADBEAF);
1291 WARN_ON(!list_empty(&inode->i_dentry));
1292 WARN_ON(inode->i_ino == 1);
1293 WARN_ON(inode->i_data.nrpages);
1297 kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
1300 static void init_once(void * foo, struct kmem_cache * cachep,
1301 unsigned long flags)
1303 struct btrfs_inode *ei = (struct btrfs_inode *) foo;
1305 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
1306 SLAB_CTOR_CONSTRUCTOR) {
1307 inode_init_once(&ei->vfs_inode);
1311 static int init_inodecache(void)
1313 btrfs_inode_cachep = kmem_cache_create("btrfs_inode_cache",
1314 sizeof(struct btrfs_inode),
1315 0, (SLAB_RECLAIM_ACCOUNT|
1318 btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle_cache",
1319 sizeof(struct btrfs_trans_handle),
1320 0, (SLAB_RECLAIM_ACCOUNT|
1323 btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction_cache",
1324 sizeof(struct btrfs_transaction),
1325 0, (SLAB_RECLAIM_ACCOUNT|
1328 btrfs_path_cachep = kmem_cache_create("btrfs_path_cache",
1329 sizeof(struct btrfs_transaction),
1330 0, (SLAB_RECLAIM_ACCOUNT|
1333 btrfs_bit_radix_cachep = kmem_cache_create("btrfs_radix",
1335 0, (SLAB_RECLAIM_ACCOUNT|
1337 SLAB_DESTROY_BY_RCU),
1339 if (btrfs_inode_cachep == NULL || btrfs_trans_handle_cachep == NULL ||
1340 btrfs_transaction_cachep == NULL || btrfs_bit_radix_cachep == NULL)
1345 static void destroy_inodecache(void)
1347 kmem_cache_destroy(btrfs_inode_cachep);
1348 kmem_cache_destroy(btrfs_trans_handle_cachep);
1349 kmem_cache_destroy(btrfs_transaction_cachep);
1350 kmem_cache_destroy(btrfs_bit_radix_cachep);
1351 kmem_cache_destroy(btrfs_path_cachep);
1354 static int btrfs_get_sb(struct file_system_type *fs_type,
1355 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1357 return get_sb_bdev(fs_type, flags, dev_name, data,
1358 btrfs_fill_super, mnt);
1361 static struct file_system_type btrfs_fs_type = {
1362 .owner = THIS_MODULE,
1364 .get_sb = btrfs_get_sb,
1365 .kill_sb = kill_block_super,
1366 .fs_flags = FS_REQUIRES_DEV,
1369 static struct super_operations btrfs_super_ops = {
1370 .statfs = simple_statfs,
1371 .delete_inode = btrfs_delete_inode,
1372 .put_super = btrfs_put_super,
1373 .read_inode = btrfs_read_locked_inode,
1374 .write_super = btrfs_write_super,
1375 .sync_fs = btrfs_sync_fs,
1376 .write_inode = btrfs_write_inode,
1377 .alloc_inode = btrfs_alloc_inode,
1378 .destroy_inode = btrfs_destroy_inode,
1381 static struct inode_operations btrfs_dir_inode_operations = {
1382 .lookup = btrfs_lookup,
1383 .create = btrfs_create,
1384 .unlink = btrfs_unlink,
1385 .mkdir = btrfs_mkdir,
1386 .rmdir = btrfs_rmdir,
1389 static struct file_operations btrfs_dir_file_operations = {
1390 .llseek = generic_file_llseek,
1391 .read = generic_read_dir,
1392 .readdir = btrfs_readdir,
1395 static struct address_space_operations btrfs_aops = {
1396 .readpage = btrfs_readpage,
1397 .readpages = btrfs_readpages,
1398 .writepage = btrfs_writepage,
1399 .sync_page = block_sync_page,
1400 .prepare_write = btrfs_prepare_write,
1401 .commit_write = btrfs_commit_write,
1404 static struct inode_operations btrfs_file_inode_operations = {
1405 .truncate = btrfs_truncate,
1408 static struct file_operations btrfs_file_operations = {
1409 .llseek = generic_file_llseek,
1410 .read = do_sync_read,
1411 .aio_read = btrfs_file_aio_read,
1412 .write = btrfs_file_write,
1413 .mmap = generic_file_mmap,
1414 .open = generic_file_open,
1417 static int __init init_btrfs_fs(void)
1420 printk("btrfs loaded!\n");
1421 err = init_inodecache();
1424 return register_filesystem(&btrfs_fs_type);
1427 static void __exit exit_btrfs_fs(void)
1429 destroy_inodecache();
1430 unregister_filesystem(&btrfs_fs_type);
1431 printk("btrfs unloaded\n");
1434 module_init(init_btrfs_fs)
1435 module_exit(exit_btrfs_fs)
1437 MODULE_LICENSE("GPL");