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/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
45 #include "transaction.h"
46 #include "btrfs_inode.h"
48 #include "print-tree.h"
53 /* Mask out flags that are inappropriate for the given type of inode. */
54 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
58 else if (S_ISREG(mode))
59 return flags & ~FS_DIRSYNC_FL;
61 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
65 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
67 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
69 unsigned int iflags = 0;
71 if (flags & BTRFS_INODE_SYNC)
73 if (flags & BTRFS_INODE_IMMUTABLE)
74 iflags |= FS_IMMUTABLE_FL;
75 if (flags & BTRFS_INODE_APPEND)
76 iflags |= FS_APPEND_FL;
77 if (flags & BTRFS_INODE_NODUMP)
78 iflags |= FS_NODUMP_FL;
79 if (flags & BTRFS_INODE_NOATIME)
80 iflags |= FS_NOATIME_FL;
81 if (flags & BTRFS_INODE_DIRSYNC)
82 iflags |= FS_DIRSYNC_FL;
88 * Update inode->i_flags based on the btrfs internal flags.
90 void btrfs_update_iflags(struct inode *inode)
92 struct btrfs_inode *ip = BTRFS_I(inode);
94 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
96 if (ip->flags & BTRFS_INODE_SYNC)
97 inode->i_flags |= S_SYNC;
98 if (ip->flags & BTRFS_INODE_IMMUTABLE)
99 inode->i_flags |= S_IMMUTABLE;
100 if (ip->flags & BTRFS_INODE_APPEND)
101 inode->i_flags |= S_APPEND;
102 if (ip->flags & BTRFS_INODE_NOATIME)
103 inode->i_flags |= S_NOATIME;
104 if (ip->flags & BTRFS_INODE_DIRSYNC)
105 inode->i_flags |= S_DIRSYNC;
109 * Inherit flags from the parent inode.
111 * Unlike extN we don't have any flags we don't want to inherit currently.
113 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
120 flags = BTRFS_I(dir)->flags;
122 if (S_ISREG(inode->i_mode))
123 flags &= ~BTRFS_INODE_DIRSYNC;
124 else if (!S_ISDIR(inode->i_mode))
125 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
127 BTRFS_I(inode)->flags = flags;
128 btrfs_update_iflags(inode);
131 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
133 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
134 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
136 if (copy_to_user(arg, &flags, sizeof(flags)))
141 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
143 struct inode *inode = file->f_path.dentry->d_inode;
144 struct btrfs_inode *ip = BTRFS_I(inode);
145 struct btrfs_root *root = ip->root;
146 struct btrfs_trans_handle *trans;
147 unsigned int flags, oldflags;
150 if (copy_from_user(&flags, arg, sizeof(flags)))
153 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
154 FS_NOATIME_FL | FS_NODUMP_FL | \
155 FS_SYNC_FL | FS_DIRSYNC_FL))
158 if (!is_owner_or_cap(inode))
161 mutex_lock(&inode->i_mutex);
163 flags = btrfs_mask_flags(inode->i_mode, flags);
164 oldflags = btrfs_flags_to_ioctl(ip->flags);
165 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
166 if (!capable(CAP_LINUX_IMMUTABLE)) {
172 ret = mnt_want_write(file->f_path.mnt);
176 if (flags & FS_SYNC_FL)
177 ip->flags |= BTRFS_INODE_SYNC;
179 ip->flags &= ~BTRFS_INODE_SYNC;
180 if (flags & FS_IMMUTABLE_FL)
181 ip->flags |= BTRFS_INODE_IMMUTABLE;
183 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
184 if (flags & FS_APPEND_FL)
185 ip->flags |= BTRFS_INODE_APPEND;
187 ip->flags &= ~BTRFS_INODE_APPEND;
188 if (flags & FS_NODUMP_FL)
189 ip->flags |= BTRFS_INODE_NODUMP;
191 ip->flags &= ~BTRFS_INODE_NODUMP;
192 if (flags & FS_NOATIME_FL)
193 ip->flags |= BTRFS_INODE_NOATIME;
195 ip->flags &= ~BTRFS_INODE_NOATIME;
196 if (flags & FS_DIRSYNC_FL)
197 ip->flags |= BTRFS_INODE_DIRSYNC;
199 ip->flags &= ~BTRFS_INODE_DIRSYNC;
202 trans = btrfs_join_transaction(root, 1);
205 ret = btrfs_update_inode(trans, root, inode);
208 btrfs_update_iflags(inode);
209 inode->i_ctime = CURRENT_TIME;
210 btrfs_end_transaction(trans, root);
212 mnt_drop_write(file->f_path.mnt);
214 mutex_unlock(&inode->i_mutex);
218 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
220 struct inode *inode = file->f_path.dentry->d_inode;
222 return put_user(inode->i_generation, arg);
225 static noinline int create_subvol(struct btrfs_root *root,
226 struct dentry *dentry,
227 char *name, int namelen)
229 struct btrfs_trans_handle *trans;
230 struct btrfs_key key;
231 struct btrfs_root_item root_item;
232 struct btrfs_inode_item *inode_item;
233 struct extent_buffer *leaf;
234 struct btrfs_root *new_root;
235 struct inode *dir = dentry->d_parent->d_inode;
239 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
248 ret = btrfs_reserve_metadata_space(root, 6);
252 trans = btrfs_start_transaction(root, 1);
255 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
260 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
261 0, objectid, NULL, 0, 0, 0);
267 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
268 btrfs_set_header_bytenr(leaf, leaf->start);
269 btrfs_set_header_generation(leaf, trans->transid);
270 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
271 btrfs_set_header_owner(leaf, objectid);
273 write_extent_buffer(leaf, root->fs_info->fsid,
274 (unsigned long)btrfs_header_fsid(leaf),
276 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
277 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
279 btrfs_mark_buffer_dirty(leaf);
281 inode_item = &root_item.inode;
282 memset(inode_item, 0, sizeof(*inode_item));
283 inode_item->generation = cpu_to_le64(1);
284 inode_item->size = cpu_to_le64(3);
285 inode_item->nlink = cpu_to_le32(1);
286 inode_item->nbytes = cpu_to_le64(root->leafsize);
287 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
289 btrfs_set_root_bytenr(&root_item, leaf->start);
290 btrfs_set_root_generation(&root_item, trans->transid);
291 btrfs_set_root_level(&root_item, 0);
292 btrfs_set_root_refs(&root_item, 1);
293 btrfs_set_root_used(&root_item, leaf->len);
294 btrfs_set_root_last_snapshot(&root_item, 0);
296 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
297 root_item.drop_level = 0;
299 btrfs_tree_unlock(leaf);
300 free_extent_buffer(leaf);
303 btrfs_set_root_dirid(&root_item, new_dirid);
305 key.objectid = objectid;
307 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
308 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
313 key.offset = (u64)-1;
314 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
315 BUG_ON(IS_ERR(new_root));
317 btrfs_record_root_in_trans(trans, new_root);
319 ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
320 BTRFS_I(dir)->block_group);
322 * insert the directory item
324 ret = btrfs_set_inode_index(dir, &index);
327 ret = btrfs_insert_dir_item(trans, root,
328 name, namelen, dir->i_ino, &key,
329 BTRFS_FT_DIR, index);
333 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
334 ret = btrfs_update_inode(trans, root, dir);
337 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
338 objectid, root->root_key.objectid,
339 dir->i_ino, index, name, namelen);
343 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
345 err = btrfs_commit_transaction(trans, root);
349 btrfs_unreserve_metadata_space(root, 6);
353 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
354 char *name, int namelen)
357 struct btrfs_pending_snapshot *pending_snapshot;
358 struct btrfs_trans_handle *trans;
370 ret = btrfs_reserve_metadata_space(root, 6);
374 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
375 if (!pending_snapshot) {
377 btrfs_unreserve_metadata_space(root, 6);
380 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
381 if (!pending_snapshot->name) {
383 kfree(pending_snapshot);
384 btrfs_unreserve_metadata_space(root, 6);
387 memcpy(pending_snapshot->name, name, namelen);
388 pending_snapshot->name[namelen] = '\0';
389 pending_snapshot->dentry = dentry;
390 trans = btrfs_start_transaction(root, 1);
392 pending_snapshot->root = root;
393 list_add(&pending_snapshot->list,
394 &trans->transaction->pending_snapshots);
395 ret = btrfs_commit_transaction(trans, root);
397 btrfs_unreserve_metadata_space(root, 6);
399 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
401 ret = PTR_ERR(inode);
405 d_instantiate(dentry, inode);
411 /* copy of may_create in fs/namei.c() */
412 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
418 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
422 * Create a new subvolume below @parent. This is largely modeled after
423 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
424 * inside this filesystem so it's quite a bit simpler.
426 static noinline int btrfs_mksubvol(struct path *parent,
427 char *name, int namelen,
428 struct btrfs_root *snap_src)
430 struct inode *dir = parent->dentry->d_inode;
431 struct dentry *dentry;
434 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
436 dentry = lookup_one_len(name, parent->dentry, namelen);
437 error = PTR_ERR(dentry);
445 error = mnt_want_write(parent->mnt);
449 error = btrfs_may_create(dir, dentry);
453 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
455 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
459 error = create_snapshot(snap_src, dentry,
462 error = create_subvol(BTRFS_I(dir)->root, dentry,
466 fsnotify_mkdir(dir, dentry);
468 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
470 mnt_drop_write(parent->mnt);
474 mutex_unlock(&dir->i_mutex);
478 static int btrfs_defrag_file(struct file *file)
480 struct inode *inode = fdentry(file)->d_inode;
481 struct btrfs_root *root = BTRFS_I(inode)->root;
482 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
483 struct btrfs_ordered_extent *ordered;
485 unsigned long last_index;
486 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
487 unsigned long total_read = 0;
493 ret = btrfs_check_data_free_space(root, inode, inode->i_size);
497 mutex_lock(&inode->i_mutex);
498 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
499 for (i = 0; i <= last_index; i++) {
500 if (total_read % ra_pages == 0) {
501 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
502 min(last_index, i + ra_pages - 1));
506 page = grab_cache_page(inode->i_mapping, i);
509 if (!PageUptodate(page)) {
510 btrfs_readpage(NULL, page);
512 if (!PageUptodate(page)) {
514 page_cache_release(page);
519 wait_on_page_writeback(page);
521 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
522 page_end = page_start + PAGE_CACHE_SIZE - 1;
523 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
525 ordered = btrfs_lookup_ordered_extent(inode, page_start);
527 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
529 page_cache_release(page);
530 btrfs_start_ordered_extent(inode, ordered, 1);
531 btrfs_put_ordered_extent(ordered);
534 set_page_extent_mapped(page);
537 * this makes sure page_mkwrite is called on the
538 * page if it is dirtied again later
540 clear_page_dirty_for_io(page);
542 btrfs_set_extent_delalloc(inode, page_start, page_end);
543 set_page_dirty(page);
544 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
546 page_cache_release(page);
547 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
551 mutex_unlock(&inode->i_mutex);
555 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
561 struct btrfs_ioctl_vol_args *vol_args;
562 struct btrfs_trans_handle *trans;
563 struct btrfs_device *device = NULL;
570 if (root->fs_info->sb->s_flags & MS_RDONLY)
573 if (!capable(CAP_SYS_ADMIN))
576 vol_args = memdup_user(arg, sizeof(*vol_args));
577 if (IS_ERR(vol_args))
578 return PTR_ERR(vol_args);
580 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
581 namelen = strlen(vol_args->name);
583 mutex_lock(&root->fs_info->volume_mutex);
584 sizestr = vol_args->name;
585 devstr = strchr(sizestr, ':');
588 sizestr = devstr + 1;
590 devstr = vol_args->name;
591 devid = simple_strtoull(devstr, &end, 10);
592 printk(KERN_INFO "resizing devid %llu\n",
593 (unsigned long long)devid);
595 device = btrfs_find_device(root, devid, NULL, NULL);
597 printk(KERN_INFO "resizer unable to find device %llu\n",
598 (unsigned long long)devid);
602 if (!strcmp(sizestr, "max"))
603 new_size = device->bdev->bd_inode->i_size;
605 if (sizestr[0] == '-') {
608 } else if (sizestr[0] == '+') {
612 new_size = btrfs_parse_size(sizestr);
619 old_size = device->total_bytes;
622 if (new_size > old_size) {
626 new_size = old_size - new_size;
627 } else if (mod > 0) {
628 new_size = old_size + new_size;
631 if (new_size < 256 * 1024 * 1024) {
635 if (new_size > device->bdev->bd_inode->i_size) {
640 do_div(new_size, root->sectorsize);
641 new_size *= root->sectorsize;
643 printk(KERN_INFO "new size for %s is %llu\n",
644 device->name, (unsigned long long)new_size);
646 if (new_size > old_size) {
647 trans = btrfs_start_transaction(root, 1);
648 ret = btrfs_grow_device(trans, device, new_size);
649 btrfs_commit_transaction(trans, root);
651 ret = btrfs_shrink_device(device, new_size);
655 mutex_unlock(&root->fs_info->volume_mutex);
660 static noinline int btrfs_ioctl_snap_create(struct file *file,
661 void __user *arg, int subvol)
663 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
664 struct btrfs_ioctl_vol_args *vol_args;
665 struct file *src_file;
669 if (root->fs_info->sb->s_flags & MS_RDONLY)
672 vol_args = memdup_user(arg, sizeof(*vol_args));
673 if (IS_ERR(vol_args))
674 return PTR_ERR(vol_args);
676 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
677 namelen = strlen(vol_args->name);
678 if (strchr(vol_args->name, '/')) {
684 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
687 struct inode *src_inode;
688 src_file = fget(vol_args->fd);
694 src_inode = src_file->f_path.dentry->d_inode;
695 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
696 printk(KERN_INFO "btrfs: Snapshot src from "
702 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
703 BTRFS_I(src_inode)->root);
712 * helper to check if the subvolume references other subvolumes
714 static noinline int may_destroy_subvol(struct btrfs_root *root)
716 struct btrfs_path *path;
717 struct btrfs_key key;
720 path = btrfs_alloc_path();
724 key.objectid = root->root_key.objectid;
725 key.type = BTRFS_ROOT_REF_KEY;
726 key.offset = (u64)-1;
728 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
735 if (path->slots[0] > 0) {
737 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
738 if (key.objectid == root->root_key.objectid &&
739 key.type == BTRFS_ROOT_REF_KEY)
743 btrfs_free_path(path);
747 static noinline int key_in_sk(struct btrfs_key *key,
748 struct btrfs_ioctl_search_key *sk)
750 if (key->objectid < sk->min_objectid)
752 if (key->offset < sk->min_offset)
754 if (key->type < sk->min_type)
756 if (key->objectid > sk->max_objectid)
758 if (key->type > sk->max_type)
760 if (key->offset > sk->max_offset)
765 static noinline int copy_to_sk(struct btrfs_root *root,
766 struct btrfs_path *path,
767 struct btrfs_key *key,
768 struct btrfs_ioctl_search_key *sk,
770 unsigned long *sk_offset,
774 struct extent_buffer *leaf;
775 struct btrfs_ioctl_search_header sh;
776 unsigned long item_off;
777 unsigned long item_len;
784 leaf = path->nodes[0];
785 slot = path->slots[0];
786 nritems = btrfs_header_nritems(leaf);
788 if (btrfs_header_generation(leaf) > sk->max_transid) {
792 found_transid = btrfs_header_generation(leaf);
794 for (i = slot; i < nritems; i++) {
795 item_off = btrfs_item_ptr_offset(leaf, i);
796 item_len = btrfs_item_size_nr(leaf, i);
798 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
801 if (sizeof(sh) + item_len + *sk_offset >
802 BTRFS_SEARCH_ARGS_BUFSIZE) {
807 btrfs_item_key_to_cpu(leaf, key, i);
808 if (!key_in_sk(key, sk))
811 sh.objectid = key->objectid;
812 sh.offset = key->offset;
815 sh.transid = found_transid;
817 /* copy search result header */
818 memcpy(buf + *sk_offset, &sh, sizeof(sh));
819 *sk_offset += sizeof(sh);
822 char *p = buf + *sk_offset;
824 read_extent_buffer(leaf, p,
826 *sk_offset += item_len;
830 if (*num_found >= sk->nr_items)
834 if (key->offset < (u64)-1)
836 else if (key->type < (u64)-1)
838 else if (key->objectid < (u64)-1)
846 static noinline int search_ioctl(struct inode *inode,
847 struct btrfs_ioctl_search_args *args)
849 struct btrfs_root *root;
850 struct btrfs_key key;
851 struct btrfs_key max_key;
852 struct btrfs_path *path;
853 struct btrfs_ioctl_search_key *sk = &args->key;
854 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
857 unsigned long sk_offset = 0;
859 path = btrfs_alloc_path();
863 if (sk->tree_id == 0) {
864 /* search the root of the inode that was passed */
865 root = BTRFS_I(inode)->root;
867 key.objectid = sk->tree_id;
868 key.type = BTRFS_ROOT_ITEM_KEY;
869 key.offset = (u64)-1;
870 root = btrfs_read_fs_root_no_name(info, &key);
872 printk(KERN_ERR "could not find root %llu\n",
874 btrfs_free_path(path);
879 key.objectid = sk->min_objectid;
880 key.type = sk->min_type;
881 key.offset = sk->min_offset;
883 max_key.objectid = sk->max_objectid;
884 max_key.type = sk->max_type;
885 max_key.offset = sk->max_offset;
887 path->keep_locks = 1;
890 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
897 ret = copy_to_sk(root, path, &key, sk, args->buf,
898 &sk_offset, &num_found);
899 btrfs_release_path(root, path);
900 if (ret || num_found >= sk->nr_items)
906 sk->nr_items = num_found;
907 btrfs_free_path(path);
911 static noinline int btrfs_ioctl_tree_search(struct file *file,
914 struct btrfs_ioctl_search_args *args;
918 if (!capable(CAP_SYS_ADMIN))
921 args = kmalloc(sizeof(*args), GFP_KERNEL);
925 if (copy_from_user(args, argp, sizeof(*args))) {
929 inode = fdentry(file)->d_inode;
930 ret = search_ioctl(inode, args);
931 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
938 * Search INODE_REFs to identify path name of 'dirid' directory
939 * in a 'tree_id' tree. and sets path name to 'name'.
941 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
942 u64 tree_id, u64 dirid, char *name)
944 struct btrfs_root *root;
945 struct btrfs_key key;
951 struct btrfs_inode_ref *iref;
952 struct extent_buffer *l;
953 struct btrfs_path *path;
955 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
960 path = btrfs_alloc_path();
964 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
966 key.objectid = tree_id;
967 key.type = BTRFS_ROOT_ITEM_KEY;
968 key.offset = (u64)-1;
969 root = btrfs_read_fs_root_no_name(info, &key);
971 printk(KERN_ERR "could not find root %llu\n", tree_id);
975 key.objectid = dirid;
976 key.type = BTRFS_INODE_REF_KEY;
980 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
985 slot = path->slots[0];
986 btrfs_item_key_to_cpu(l, &key, slot);
988 if (ret > 0 && (key.objectid != dirid ||
989 key.type != BTRFS_INODE_REF_KEY)) {
994 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
995 len = btrfs_inode_ref_name_len(l, iref);
997 total_len += len + 1;
1002 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1004 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1007 btrfs_release_path(root, path);
1008 key.objectid = key.offset;
1010 dirid = key.objectid;
1015 memcpy(name, ptr, total_len);
1016 name[total_len]='\0';
1019 btrfs_free_path(path);
1023 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1026 struct btrfs_ioctl_ino_lookup_args *args;
1027 struct inode *inode;
1030 if (!capable(CAP_SYS_ADMIN))
1033 args = kmalloc(sizeof(*args), GFP_KERNEL);
1034 if (copy_from_user(args, argp, sizeof(*args))) {
1038 inode = fdentry(file)->d_inode;
1040 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1041 args->treeid, args->objectid,
1044 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1051 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1054 struct dentry *parent = fdentry(file);
1055 struct dentry *dentry;
1056 struct inode *dir = parent->d_inode;
1057 struct inode *inode;
1058 struct btrfs_root *root = BTRFS_I(dir)->root;
1059 struct btrfs_root *dest = NULL;
1060 struct btrfs_ioctl_vol_args *vol_args;
1061 struct btrfs_trans_handle *trans;
1066 if (!capable(CAP_SYS_ADMIN))
1069 vol_args = memdup_user(arg, sizeof(*vol_args));
1070 if (IS_ERR(vol_args))
1071 return PTR_ERR(vol_args);
1073 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1074 namelen = strlen(vol_args->name);
1075 if (strchr(vol_args->name, '/') ||
1076 strncmp(vol_args->name, "..", namelen) == 0) {
1081 err = mnt_want_write(file->f_path.mnt);
1085 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1086 dentry = lookup_one_len(vol_args->name, parent, namelen);
1087 if (IS_ERR(dentry)) {
1088 err = PTR_ERR(dentry);
1089 goto out_unlock_dir;
1092 if (!dentry->d_inode) {
1097 inode = dentry->d_inode;
1098 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
1103 dest = BTRFS_I(inode)->root;
1105 mutex_lock(&inode->i_mutex);
1106 err = d_invalidate(dentry);
1110 down_write(&root->fs_info->subvol_sem);
1112 err = may_destroy_subvol(dest);
1116 trans = btrfs_start_transaction(root, 1);
1117 ret = btrfs_unlink_subvol(trans, root, dir,
1118 dest->root_key.objectid,
1119 dentry->d_name.name,
1120 dentry->d_name.len);
1123 btrfs_record_root_in_trans(trans, dest);
1125 memset(&dest->root_item.drop_progress, 0,
1126 sizeof(dest->root_item.drop_progress));
1127 dest->root_item.drop_level = 0;
1128 btrfs_set_root_refs(&dest->root_item, 0);
1130 ret = btrfs_insert_orphan_item(trans,
1131 root->fs_info->tree_root,
1132 dest->root_key.objectid);
1135 ret = btrfs_commit_transaction(trans, root);
1137 inode->i_flags |= S_DEAD;
1139 up_write(&root->fs_info->subvol_sem);
1141 mutex_unlock(&inode->i_mutex);
1143 shrink_dcache_sb(root->fs_info->sb);
1144 btrfs_invalidate_inodes(dest);
1150 mutex_unlock(&dir->i_mutex);
1151 mnt_drop_write(file->f_path.mnt);
1157 static int btrfs_ioctl_defrag(struct file *file)
1159 struct inode *inode = fdentry(file)->d_inode;
1160 struct btrfs_root *root = BTRFS_I(inode)->root;
1163 ret = mnt_want_write(file->f_path.mnt);
1167 switch (inode->i_mode & S_IFMT) {
1169 if (!capable(CAP_SYS_ADMIN)) {
1173 btrfs_defrag_root(root, 0);
1174 btrfs_defrag_root(root->fs_info->extent_root, 0);
1177 if (!(file->f_mode & FMODE_WRITE)) {
1181 btrfs_defrag_file(file);
1185 mnt_drop_write(file->f_path.mnt);
1189 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
1191 struct btrfs_ioctl_vol_args *vol_args;
1194 if (!capable(CAP_SYS_ADMIN))
1197 vol_args = memdup_user(arg, sizeof(*vol_args));
1198 if (IS_ERR(vol_args))
1199 return PTR_ERR(vol_args);
1201 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1202 ret = btrfs_init_new_device(root, vol_args->name);
1208 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
1210 struct btrfs_ioctl_vol_args *vol_args;
1213 if (!capable(CAP_SYS_ADMIN))
1216 if (root->fs_info->sb->s_flags & MS_RDONLY)
1219 vol_args = memdup_user(arg, sizeof(*vol_args));
1220 if (IS_ERR(vol_args))
1221 return PTR_ERR(vol_args);
1223 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1224 ret = btrfs_rm_device(root, vol_args->name);
1230 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1231 u64 off, u64 olen, u64 destoff)
1233 struct inode *inode = fdentry(file)->d_inode;
1234 struct btrfs_root *root = BTRFS_I(inode)->root;
1235 struct file *src_file;
1237 struct btrfs_trans_handle *trans;
1238 struct btrfs_path *path;
1239 struct extent_buffer *leaf;
1241 struct btrfs_key key;
1246 u64 bs = root->fs_info->sb->s_blocksize;
1251 * - split compressed inline extents. annoying: we need to
1252 * decompress into destination's address_space (the file offset
1253 * may change, so source mapping won't do), then recompress (or
1254 * otherwise reinsert) a subrange.
1255 * - allow ranges within the same file to be cloned (provided
1256 * they don't overlap)?
1259 /* the destination must be opened for writing */
1260 if (!(file->f_mode & FMODE_WRITE))
1263 ret = mnt_want_write(file->f_path.mnt);
1267 src_file = fget(srcfd);
1270 goto out_drop_write;
1272 src = src_file->f_dentry->d_inode;
1279 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
1283 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
1287 buf = vmalloc(btrfs_level_size(root, 0));
1291 path = btrfs_alloc_path();
1299 mutex_lock(&inode->i_mutex);
1300 mutex_lock(&src->i_mutex);
1302 mutex_lock(&src->i_mutex);
1303 mutex_lock(&inode->i_mutex);
1306 /* determine range to clone */
1308 if (off >= src->i_size || off + len > src->i_size)
1311 olen = len = src->i_size - off;
1312 /* if we extend to eof, continue to block boundary */
1313 if (off + len == src->i_size)
1314 len = ((src->i_size + bs-1) & ~(bs-1))
1317 /* verify the end result is block aligned */
1318 if ((off & (bs-1)) ||
1319 ((off + len) & (bs-1)))
1322 /* do any pending delalloc/csum calc on src, one way or
1323 another, and lock file content */
1325 struct btrfs_ordered_extent *ordered;
1326 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1327 ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
1328 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
1330 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1332 btrfs_put_ordered_extent(ordered);
1333 btrfs_wait_ordered_range(src, off, off+len);
1336 trans = btrfs_start_transaction(root, 1);
1339 /* punch hole in destination first */
1340 btrfs_drop_extents(trans, inode, off, off + len, &hint_byte, 1);
1343 key.objectid = src->i_ino;
1344 key.type = BTRFS_EXTENT_DATA_KEY;
1349 * note the key will change type as we walk through the
1352 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
1356 nritems = btrfs_header_nritems(path->nodes[0]);
1357 if (path->slots[0] >= nritems) {
1358 ret = btrfs_next_leaf(root, path);
1363 nritems = btrfs_header_nritems(path->nodes[0]);
1365 leaf = path->nodes[0];
1366 slot = path->slots[0];
1368 btrfs_item_key_to_cpu(leaf, &key, slot);
1369 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1370 key.objectid != src->i_ino)
1373 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1374 struct btrfs_file_extent_item *extent;
1377 struct btrfs_key new_key;
1378 u64 disko = 0, diskl = 0;
1379 u64 datao = 0, datal = 0;
1382 size = btrfs_item_size_nr(leaf, slot);
1383 read_extent_buffer(leaf, buf,
1384 btrfs_item_ptr_offset(leaf, slot),
1387 extent = btrfs_item_ptr(leaf, slot,
1388 struct btrfs_file_extent_item);
1389 comp = btrfs_file_extent_compression(leaf, extent);
1390 type = btrfs_file_extent_type(leaf, extent);
1391 if (type == BTRFS_FILE_EXTENT_REG ||
1392 type == BTRFS_FILE_EXTENT_PREALLOC) {
1393 disko = btrfs_file_extent_disk_bytenr(leaf,
1395 diskl = btrfs_file_extent_disk_num_bytes(leaf,
1397 datao = btrfs_file_extent_offset(leaf, extent);
1398 datal = btrfs_file_extent_num_bytes(leaf,
1400 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1401 /* take upper bound, may be compressed */
1402 datal = btrfs_file_extent_ram_bytes(leaf,
1405 btrfs_release_path(root, path);
1407 if (key.offset + datal < off ||
1408 key.offset >= off+len)
1411 memcpy(&new_key, &key, sizeof(new_key));
1412 new_key.objectid = inode->i_ino;
1413 new_key.offset = key.offset + destoff - off;
1415 if (type == BTRFS_FILE_EXTENT_REG ||
1416 type == BTRFS_FILE_EXTENT_PREALLOC) {
1417 ret = btrfs_insert_empty_item(trans, root, path,
1422 leaf = path->nodes[0];
1423 slot = path->slots[0];
1424 write_extent_buffer(leaf, buf,
1425 btrfs_item_ptr_offset(leaf, slot),
1428 extent = btrfs_item_ptr(leaf, slot,
1429 struct btrfs_file_extent_item);
1431 if (off > key.offset) {
1432 datao += off - key.offset;
1433 datal -= off - key.offset;
1436 if (key.offset + datal > off + len)
1437 datal = off + len - key.offset;
1439 /* disko == 0 means it's a hole */
1443 btrfs_set_file_extent_offset(leaf, extent,
1445 btrfs_set_file_extent_num_bytes(leaf, extent,
1448 inode_add_bytes(inode, datal);
1449 ret = btrfs_inc_extent_ref(trans, root,
1451 root->root_key.objectid,
1453 new_key.offset - datao);
1456 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1459 if (off > key.offset) {
1460 skip = off - key.offset;
1461 new_key.offset += skip;
1464 if (key.offset + datal > off+len)
1465 trim = key.offset + datal - (off+len);
1467 if (comp && (skip || trim)) {
1471 size -= skip + trim;
1472 datal -= skip + trim;
1473 ret = btrfs_insert_empty_item(trans, root, path,
1480 btrfs_file_extent_calc_inline_size(0);
1481 memmove(buf+start, buf+start+skip,
1485 leaf = path->nodes[0];
1486 slot = path->slots[0];
1487 write_extent_buffer(leaf, buf,
1488 btrfs_item_ptr_offset(leaf, slot),
1490 inode_add_bytes(inode, datal);
1493 btrfs_mark_buffer_dirty(leaf);
1497 btrfs_release_path(root, path);
1502 btrfs_release_path(root, path);
1504 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1505 if (destoff + olen > inode->i_size)
1506 btrfs_i_size_write(inode, destoff + olen);
1507 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1508 ret = btrfs_update_inode(trans, root, inode);
1510 btrfs_end_transaction(trans, root);
1511 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1513 vmtruncate(inode, 0);
1515 mutex_unlock(&src->i_mutex);
1516 mutex_unlock(&inode->i_mutex);
1518 btrfs_free_path(path);
1522 mnt_drop_write(file->f_path.mnt);
1526 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
1528 struct btrfs_ioctl_clone_range_args args;
1530 if (copy_from_user(&args, argp, sizeof(args)))
1532 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1533 args.src_length, args.dest_offset);
1537 * there are many ways the trans_start and trans_end ioctls can lead
1538 * to deadlocks. They should only be used by applications that
1539 * basically own the machine, and have a very in depth understanding
1540 * of all the possible deadlocks and enospc problems.
1542 static long btrfs_ioctl_trans_start(struct file *file)
1544 struct inode *inode = fdentry(file)->d_inode;
1545 struct btrfs_root *root = BTRFS_I(inode)->root;
1546 struct btrfs_trans_handle *trans;
1550 if (!capable(CAP_SYS_ADMIN))
1554 if (file->private_data)
1557 ret = mnt_want_write(file->f_path.mnt);
1561 mutex_lock(&root->fs_info->trans_mutex);
1562 root->fs_info->open_ioctl_trans++;
1563 mutex_unlock(&root->fs_info->trans_mutex);
1566 trans = btrfs_start_ioctl_transaction(root, 0);
1570 file->private_data = trans;
1574 mutex_lock(&root->fs_info->trans_mutex);
1575 root->fs_info->open_ioctl_trans--;
1576 mutex_unlock(&root->fs_info->trans_mutex);
1577 mnt_drop_write(file->f_path.mnt);
1583 * there are many ways the trans_start and trans_end ioctls can lead
1584 * to deadlocks. They should only be used by applications that
1585 * basically own the machine, and have a very in depth understanding
1586 * of all the possible deadlocks and enospc problems.
1588 long btrfs_ioctl_trans_end(struct file *file)
1590 struct inode *inode = fdentry(file)->d_inode;
1591 struct btrfs_root *root = BTRFS_I(inode)->root;
1592 struct btrfs_trans_handle *trans;
1594 trans = file->private_data;
1597 file->private_data = NULL;
1599 btrfs_end_transaction(trans, root);
1601 mutex_lock(&root->fs_info->trans_mutex);
1602 root->fs_info->open_ioctl_trans--;
1603 mutex_unlock(&root->fs_info->trans_mutex);
1605 mnt_drop_write(file->f_path.mnt);
1609 long btrfs_ioctl(struct file *file, unsigned int
1610 cmd, unsigned long arg)
1612 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1613 void __user *argp = (void __user *)arg;
1616 case FS_IOC_GETFLAGS:
1617 return btrfs_ioctl_getflags(file, argp);
1618 case FS_IOC_SETFLAGS:
1619 return btrfs_ioctl_setflags(file, argp);
1620 case FS_IOC_GETVERSION:
1621 return btrfs_ioctl_getversion(file, argp);
1622 case BTRFS_IOC_SNAP_CREATE:
1623 return btrfs_ioctl_snap_create(file, argp, 0);
1624 case BTRFS_IOC_SUBVOL_CREATE:
1625 return btrfs_ioctl_snap_create(file, argp, 1);
1626 case BTRFS_IOC_SNAP_DESTROY:
1627 return btrfs_ioctl_snap_destroy(file, argp);
1628 case BTRFS_IOC_DEFRAG:
1629 return btrfs_ioctl_defrag(file);
1630 case BTRFS_IOC_RESIZE:
1631 return btrfs_ioctl_resize(root, argp);
1632 case BTRFS_IOC_ADD_DEV:
1633 return btrfs_ioctl_add_dev(root, argp);
1634 case BTRFS_IOC_RM_DEV:
1635 return btrfs_ioctl_rm_dev(root, argp);
1636 case BTRFS_IOC_BALANCE:
1637 return btrfs_balance(root->fs_info->dev_root);
1638 case BTRFS_IOC_CLONE:
1639 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
1640 case BTRFS_IOC_CLONE_RANGE:
1641 return btrfs_ioctl_clone_range(file, argp);
1642 case BTRFS_IOC_TRANS_START:
1643 return btrfs_ioctl_trans_start(file);
1644 case BTRFS_IOC_TRANS_END:
1645 return btrfs_ioctl_trans_end(file);
1646 case BTRFS_IOC_TREE_SEARCH:
1647 return btrfs_ioctl_tree_search(file, argp);
1648 case BTRFS_IOC_INO_LOOKUP:
1649 return btrfs_ioctl_ino_lookup(file, argp);
1650 case BTRFS_IOC_SYNC:
1651 btrfs_sync_fs(file->f_dentry->d_sb, 1);