xref: /linux/fs/btrfs/ioctl.c (revision f2ee442115c9b6219083c019939a9cc0c9abb2f8)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
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.
7  *
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.
12  *
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.
17  */
18 
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.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>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include "compat.h"
45 #include "ctree.h"
46 #include "disk-io.h"
47 #include "transaction.h"
48 #include "btrfs_inode.h"
49 #include "ioctl.h"
50 #include "print-tree.h"
51 #include "volumes.h"
52 #include "locking.h"
53 #include "inode-map.h"
54 #include "backref.h"
55 
56 /* Mask out flags that are inappropriate for the given type of inode. */
57 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
58 {
59 	if (S_ISDIR(mode))
60 		return flags;
61 	else if (S_ISREG(mode))
62 		return flags & ~FS_DIRSYNC_FL;
63 	else
64 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
65 }
66 
67 /*
68  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
69  */
70 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
71 {
72 	unsigned int iflags = 0;
73 
74 	if (flags & BTRFS_INODE_SYNC)
75 		iflags |= FS_SYNC_FL;
76 	if (flags & BTRFS_INODE_IMMUTABLE)
77 		iflags |= FS_IMMUTABLE_FL;
78 	if (flags & BTRFS_INODE_APPEND)
79 		iflags |= FS_APPEND_FL;
80 	if (flags & BTRFS_INODE_NODUMP)
81 		iflags |= FS_NODUMP_FL;
82 	if (flags & BTRFS_INODE_NOATIME)
83 		iflags |= FS_NOATIME_FL;
84 	if (flags & BTRFS_INODE_DIRSYNC)
85 		iflags |= FS_DIRSYNC_FL;
86 	if (flags & BTRFS_INODE_NODATACOW)
87 		iflags |= FS_NOCOW_FL;
88 
89 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
90 		iflags |= FS_COMPR_FL;
91 	else if (flags & BTRFS_INODE_NOCOMPRESS)
92 		iflags |= FS_NOCOMP_FL;
93 
94 	return iflags;
95 }
96 
97 /*
98  * Update inode->i_flags based on the btrfs internal flags.
99  */
100 void btrfs_update_iflags(struct inode *inode)
101 {
102 	struct btrfs_inode *ip = BTRFS_I(inode);
103 
104 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
105 
106 	if (ip->flags & BTRFS_INODE_SYNC)
107 		inode->i_flags |= S_SYNC;
108 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
109 		inode->i_flags |= S_IMMUTABLE;
110 	if (ip->flags & BTRFS_INODE_APPEND)
111 		inode->i_flags |= S_APPEND;
112 	if (ip->flags & BTRFS_INODE_NOATIME)
113 		inode->i_flags |= S_NOATIME;
114 	if (ip->flags & BTRFS_INODE_DIRSYNC)
115 		inode->i_flags |= S_DIRSYNC;
116 }
117 
118 /*
119  * Inherit flags from the parent inode.
120  *
121  * Currently only the compression flags and the cow flags are inherited.
122  */
123 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
124 {
125 	unsigned int flags;
126 
127 	if (!dir)
128 		return;
129 
130 	flags = BTRFS_I(dir)->flags;
131 
132 	if (flags & BTRFS_INODE_NOCOMPRESS) {
133 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
134 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
135 	} else if (flags & BTRFS_INODE_COMPRESS) {
136 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
137 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
138 	}
139 
140 	if (flags & BTRFS_INODE_NODATACOW)
141 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
142 
143 	btrfs_update_iflags(inode);
144 }
145 
146 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
147 {
148 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
149 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
150 
151 	if (copy_to_user(arg, &flags, sizeof(flags)))
152 		return -EFAULT;
153 	return 0;
154 }
155 
156 static int check_flags(unsigned int flags)
157 {
158 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
159 		      FS_NOATIME_FL | FS_NODUMP_FL | \
160 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
161 		      FS_NOCOMP_FL | FS_COMPR_FL |
162 		      FS_NOCOW_FL))
163 		return -EOPNOTSUPP;
164 
165 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
166 		return -EINVAL;
167 
168 	return 0;
169 }
170 
171 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
172 {
173 	struct inode *inode = file->f_path.dentry->d_inode;
174 	struct btrfs_inode *ip = BTRFS_I(inode);
175 	struct btrfs_root *root = ip->root;
176 	struct btrfs_trans_handle *trans;
177 	unsigned int flags, oldflags;
178 	int ret;
179 
180 	if (btrfs_root_readonly(root))
181 		return -EROFS;
182 
183 	if (copy_from_user(&flags, arg, sizeof(flags)))
184 		return -EFAULT;
185 
186 	ret = check_flags(flags);
187 	if (ret)
188 		return ret;
189 
190 	if (!inode_owner_or_capable(inode))
191 		return -EACCES;
192 
193 	mutex_lock(&inode->i_mutex);
194 
195 	flags = btrfs_mask_flags(inode->i_mode, flags);
196 	oldflags = btrfs_flags_to_ioctl(ip->flags);
197 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
198 		if (!capable(CAP_LINUX_IMMUTABLE)) {
199 			ret = -EPERM;
200 			goto out_unlock;
201 		}
202 	}
203 
204 	ret = mnt_want_write(file->f_path.mnt);
205 	if (ret)
206 		goto out_unlock;
207 
208 	if (flags & FS_SYNC_FL)
209 		ip->flags |= BTRFS_INODE_SYNC;
210 	else
211 		ip->flags &= ~BTRFS_INODE_SYNC;
212 	if (flags & FS_IMMUTABLE_FL)
213 		ip->flags |= BTRFS_INODE_IMMUTABLE;
214 	else
215 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
216 	if (flags & FS_APPEND_FL)
217 		ip->flags |= BTRFS_INODE_APPEND;
218 	else
219 		ip->flags &= ~BTRFS_INODE_APPEND;
220 	if (flags & FS_NODUMP_FL)
221 		ip->flags |= BTRFS_INODE_NODUMP;
222 	else
223 		ip->flags &= ~BTRFS_INODE_NODUMP;
224 	if (flags & FS_NOATIME_FL)
225 		ip->flags |= BTRFS_INODE_NOATIME;
226 	else
227 		ip->flags &= ~BTRFS_INODE_NOATIME;
228 	if (flags & FS_DIRSYNC_FL)
229 		ip->flags |= BTRFS_INODE_DIRSYNC;
230 	else
231 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
232 	if (flags & FS_NOCOW_FL)
233 		ip->flags |= BTRFS_INODE_NODATACOW;
234 	else
235 		ip->flags &= ~BTRFS_INODE_NODATACOW;
236 
237 	/*
238 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
239 	 * flag may be changed automatically if compression code won't make
240 	 * things smaller.
241 	 */
242 	if (flags & FS_NOCOMP_FL) {
243 		ip->flags &= ~BTRFS_INODE_COMPRESS;
244 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
245 	} else if (flags & FS_COMPR_FL) {
246 		ip->flags |= BTRFS_INODE_COMPRESS;
247 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
248 	} else {
249 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
250 	}
251 
252 	trans = btrfs_join_transaction(root);
253 	BUG_ON(IS_ERR(trans));
254 
255 	ret = btrfs_update_inode(trans, root, inode);
256 	BUG_ON(ret);
257 
258 	btrfs_update_iflags(inode);
259 	inode->i_ctime = CURRENT_TIME;
260 	btrfs_end_transaction(trans, root);
261 
262 	mnt_drop_write(file->f_path.mnt);
263 
264 	ret = 0;
265  out_unlock:
266 	mutex_unlock(&inode->i_mutex);
267 	return ret;
268 }
269 
270 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
271 {
272 	struct inode *inode = file->f_path.dentry->d_inode;
273 
274 	return put_user(inode->i_generation, arg);
275 }
276 
277 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
278 {
279 	struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
280 	struct btrfs_fs_info *fs_info = root->fs_info;
281 	struct btrfs_device *device;
282 	struct request_queue *q;
283 	struct fstrim_range range;
284 	u64 minlen = ULLONG_MAX;
285 	u64 num_devices = 0;
286 	u64 total_bytes = btrfs_super_total_bytes(root->fs_info->super_copy);
287 	int ret;
288 
289 	if (!capable(CAP_SYS_ADMIN))
290 		return -EPERM;
291 
292 	rcu_read_lock();
293 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
294 				dev_list) {
295 		if (!device->bdev)
296 			continue;
297 		q = bdev_get_queue(device->bdev);
298 		if (blk_queue_discard(q)) {
299 			num_devices++;
300 			minlen = min((u64)q->limits.discard_granularity,
301 				     minlen);
302 		}
303 	}
304 	rcu_read_unlock();
305 
306 	if (!num_devices)
307 		return -EOPNOTSUPP;
308 	if (copy_from_user(&range, arg, sizeof(range)))
309 		return -EFAULT;
310 	if (range.start > total_bytes)
311 		return -EINVAL;
312 
313 	range.len = min(range.len, total_bytes - range.start);
314 	range.minlen = max(range.minlen, minlen);
315 	ret = btrfs_trim_fs(root, &range);
316 	if (ret < 0)
317 		return ret;
318 
319 	if (copy_to_user(arg, &range, sizeof(range)))
320 		return -EFAULT;
321 
322 	return 0;
323 }
324 
325 static noinline int create_subvol(struct btrfs_root *root,
326 				  struct dentry *dentry,
327 				  char *name, int namelen,
328 				  u64 *async_transid)
329 {
330 	struct btrfs_trans_handle *trans;
331 	struct btrfs_key key;
332 	struct btrfs_root_item root_item;
333 	struct btrfs_inode_item *inode_item;
334 	struct extent_buffer *leaf;
335 	struct btrfs_root *new_root;
336 	struct dentry *parent = dentry->d_parent;
337 	struct inode *dir;
338 	int ret;
339 	int err;
340 	u64 objectid;
341 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
342 	u64 index = 0;
343 
344 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
345 	if (ret)
346 		return ret;
347 
348 	dir = parent->d_inode;
349 
350 	/*
351 	 * 1 - inode item
352 	 * 2 - refs
353 	 * 1 - root item
354 	 * 2 - dir items
355 	 */
356 	trans = btrfs_start_transaction(root, 6);
357 	if (IS_ERR(trans))
358 		return PTR_ERR(trans);
359 
360 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
361 				      0, objectid, NULL, 0, 0, 0);
362 	if (IS_ERR(leaf)) {
363 		ret = PTR_ERR(leaf);
364 		goto fail;
365 	}
366 
367 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
368 	btrfs_set_header_bytenr(leaf, leaf->start);
369 	btrfs_set_header_generation(leaf, trans->transid);
370 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
371 	btrfs_set_header_owner(leaf, objectid);
372 
373 	write_extent_buffer(leaf, root->fs_info->fsid,
374 			    (unsigned long)btrfs_header_fsid(leaf),
375 			    BTRFS_FSID_SIZE);
376 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
377 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
378 			    BTRFS_UUID_SIZE);
379 	btrfs_mark_buffer_dirty(leaf);
380 
381 	inode_item = &root_item.inode;
382 	memset(inode_item, 0, sizeof(*inode_item));
383 	inode_item->generation = cpu_to_le64(1);
384 	inode_item->size = cpu_to_le64(3);
385 	inode_item->nlink = cpu_to_le32(1);
386 	inode_item->nbytes = cpu_to_le64(root->leafsize);
387 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
388 
389 	root_item.flags = 0;
390 	root_item.byte_limit = 0;
391 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
392 
393 	btrfs_set_root_bytenr(&root_item, leaf->start);
394 	btrfs_set_root_generation(&root_item, trans->transid);
395 	btrfs_set_root_level(&root_item, 0);
396 	btrfs_set_root_refs(&root_item, 1);
397 	btrfs_set_root_used(&root_item, leaf->len);
398 	btrfs_set_root_last_snapshot(&root_item, 0);
399 
400 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
401 	root_item.drop_level = 0;
402 
403 	btrfs_tree_unlock(leaf);
404 	free_extent_buffer(leaf);
405 	leaf = NULL;
406 
407 	btrfs_set_root_dirid(&root_item, new_dirid);
408 
409 	key.objectid = objectid;
410 	key.offset = 0;
411 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
412 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
413 				&root_item);
414 	if (ret)
415 		goto fail;
416 
417 	key.offset = (u64)-1;
418 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
419 	BUG_ON(IS_ERR(new_root));
420 
421 	btrfs_record_root_in_trans(trans, new_root);
422 
423 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
424 	/*
425 	 * insert the directory item
426 	 */
427 	ret = btrfs_set_inode_index(dir, &index);
428 	BUG_ON(ret);
429 
430 	ret = btrfs_insert_dir_item(trans, root,
431 				    name, namelen, dir, &key,
432 				    BTRFS_FT_DIR, index);
433 	if (ret)
434 		goto fail;
435 
436 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
437 	ret = btrfs_update_inode(trans, root, dir);
438 	BUG_ON(ret);
439 
440 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
441 				 objectid, root->root_key.objectid,
442 				 btrfs_ino(dir), index, name, namelen);
443 
444 	BUG_ON(ret);
445 
446 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
447 fail:
448 	if (async_transid) {
449 		*async_transid = trans->transid;
450 		err = btrfs_commit_transaction_async(trans, root, 1);
451 	} else {
452 		err = btrfs_commit_transaction(trans, root);
453 	}
454 	if (err && !ret)
455 		ret = err;
456 	return ret;
457 }
458 
459 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
460 			   char *name, int namelen, u64 *async_transid,
461 			   bool readonly)
462 {
463 	struct inode *inode;
464 	struct btrfs_pending_snapshot *pending_snapshot;
465 	struct btrfs_trans_handle *trans;
466 	int ret;
467 
468 	if (!root->ref_cows)
469 		return -EINVAL;
470 
471 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
472 	if (!pending_snapshot)
473 		return -ENOMEM;
474 
475 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
476 	pending_snapshot->dentry = dentry;
477 	pending_snapshot->root = root;
478 	pending_snapshot->readonly = readonly;
479 
480 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
481 	if (IS_ERR(trans)) {
482 		ret = PTR_ERR(trans);
483 		goto fail;
484 	}
485 
486 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
487 	BUG_ON(ret);
488 
489 	spin_lock(&root->fs_info->trans_lock);
490 	list_add(&pending_snapshot->list,
491 		 &trans->transaction->pending_snapshots);
492 	spin_unlock(&root->fs_info->trans_lock);
493 	if (async_transid) {
494 		*async_transid = trans->transid;
495 		ret = btrfs_commit_transaction_async(trans,
496 				     root->fs_info->extent_root, 1);
497 	} else {
498 		ret = btrfs_commit_transaction(trans,
499 					       root->fs_info->extent_root);
500 	}
501 	BUG_ON(ret);
502 
503 	ret = pending_snapshot->error;
504 	if (ret)
505 		goto fail;
506 
507 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
508 	if (ret)
509 		goto fail;
510 
511 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
512 	if (IS_ERR(inode)) {
513 		ret = PTR_ERR(inode);
514 		goto fail;
515 	}
516 	BUG_ON(!inode);
517 	d_instantiate(dentry, inode);
518 	ret = 0;
519 fail:
520 	kfree(pending_snapshot);
521 	return ret;
522 }
523 
524 /*  copy of check_sticky in fs/namei.c()
525 * It's inline, so penalty for filesystems that don't use sticky bit is
526 * minimal.
527 */
528 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
529 {
530 	uid_t fsuid = current_fsuid();
531 
532 	if (!(dir->i_mode & S_ISVTX))
533 		return 0;
534 	if (inode->i_uid == fsuid)
535 		return 0;
536 	if (dir->i_uid == fsuid)
537 		return 0;
538 	return !capable(CAP_FOWNER);
539 }
540 
541 /*  copy of may_delete in fs/namei.c()
542  *	Check whether we can remove a link victim from directory dir, check
543  *  whether the type of victim is right.
544  *  1. We can't do it if dir is read-only (done in permission())
545  *  2. We should have write and exec permissions on dir
546  *  3. We can't remove anything from append-only dir
547  *  4. We can't do anything with immutable dir (done in permission())
548  *  5. If the sticky bit on dir is set we should either
549  *	a. be owner of dir, or
550  *	b. be owner of victim, or
551  *	c. have CAP_FOWNER capability
552  *  6. If the victim is append-only or immutable we can't do antyhing with
553  *     links pointing to it.
554  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
555  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
556  *  9. We can't remove a root or mountpoint.
557  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
558  *     nfs_async_unlink().
559  */
560 
561 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
562 {
563 	int error;
564 
565 	if (!victim->d_inode)
566 		return -ENOENT;
567 
568 	BUG_ON(victim->d_parent->d_inode != dir);
569 	audit_inode_child(victim, dir);
570 
571 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
572 	if (error)
573 		return error;
574 	if (IS_APPEND(dir))
575 		return -EPERM;
576 	if (btrfs_check_sticky(dir, victim->d_inode)||
577 		IS_APPEND(victim->d_inode)||
578 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
579 		return -EPERM;
580 	if (isdir) {
581 		if (!S_ISDIR(victim->d_inode->i_mode))
582 			return -ENOTDIR;
583 		if (IS_ROOT(victim))
584 			return -EBUSY;
585 	} else if (S_ISDIR(victim->d_inode->i_mode))
586 		return -EISDIR;
587 	if (IS_DEADDIR(dir))
588 		return -ENOENT;
589 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
590 		return -EBUSY;
591 	return 0;
592 }
593 
594 /* copy of may_create in fs/namei.c() */
595 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
596 {
597 	if (child->d_inode)
598 		return -EEXIST;
599 	if (IS_DEADDIR(dir))
600 		return -ENOENT;
601 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
602 }
603 
604 /*
605  * Create a new subvolume below @parent.  This is largely modeled after
606  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
607  * inside this filesystem so it's quite a bit simpler.
608  */
609 static noinline int btrfs_mksubvol(struct path *parent,
610 				   char *name, int namelen,
611 				   struct btrfs_root *snap_src,
612 				   u64 *async_transid, bool readonly)
613 {
614 	struct inode *dir  = parent->dentry->d_inode;
615 	struct dentry *dentry;
616 	int error;
617 
618 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
619 
620 	dentry = lookup_one_len(name, parent->dentry, namelen);
621 	error = PTR_ERR(dentry);
622 	if (IS_ERR(dentry))
623 		goto out_unlock;
624 
625 	error = -EEXIST;
626 	if (dentry->d_inode)
627 		goto out_dput;
628 
629 	error = mnt_want_write(parent->mnt);
630 	if (error)
631 		goto out_dput;
632 
633 	error = btrfs_may_create(dir, dentry);
634 	if (error)
635 		goto out_drop_write;
636 
637 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
638 
639 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
640 		goto out_up_read;
641 
642 	if (snap_src) {
643 		error = create_snapshot(snap_src, dentry,
644 					name, namelen, async_transid, readonly);
645 	} else {
646 		error = create_subvol(BTRFS_I(dir)->root, dentry,
647 				      name, namelen, async_transid);
648 	}
649 	if (!error)
650 		fsnotify_mkdir(dir, dentry);
651 out_up_read:
652 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
653 out_drop_write:
654 	mnt_drop_write(parent->mnt);
655 out_dput:
656 	dput(dentry);
657 out_unlock:
658 	mutex_unlock(&dir->i_mutex);
659 	return error;
660 }
661 
662 /*
663  * When we're defragging a range, we don't want to kick it off again
664  * if it is really just waiting for delalloc to send it down.
665  * If we find a nice big extent or delalloc range for the bytes in the
666  * file you want to defrag, we return 0 to let you know to skip this
667  * part of the file
668  */
669 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
670 {
671 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
672 	struct extent_map *em = NULL;
673 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
674 	u64 end;
675 
676 	read_lock(&em_tree->lock);
677 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
678 	read_unlock(&em_tree->lock);
679 
680 	if (em) {
681 		end = extent_map_end(em);
682 		free_extent_map(em);
683 		if (end - offset > thresh)
684 			return 0;
685 	}
686 	/* if we already have a nice delalloc here, just stop */
687 	thresh /= 2;
688 	end = count_range_bits(io_tree, &offset, offset + thresh,
689 			       thresh, EXTENT_DELALLOC, 1);
690 	if (end >= thresh)
691 		return 0;
692 	return 1;
693 }
694 
695 /*
696  * helper function to walk through a file and find extents
697  * newer than a specific transid, and smaller than thresh.
698  *
699  * This is used by the defragging code to find new and small
700  * extents
701  */
702 static int find_new_extents(struct btrfs_root *root,
703 			    struct inode *inode, u64 newer_than,
704 			    u64 *off, int thresh)
705 {
706 	struct btrfs_path *path;
707 	struct btrfs_key min_key;
708 	struct btrfs_key max_key;
709 	struct extent_buffer *leaf;
710 	struct btrfs_file_extent_item *extent;
711 	int type;
712 	int ret;
713 	u64 ino = btrfs_ino(inode);
714 
715 	path = btrfs_alloc_path();
716 	if (!path)
717 		return -ENOMEM;
718 
719 	min_key.objectid = ino;
720 	min_key.type = BTRFS_EXTENT_DATA_KEY;
721 	min_key.offset = *off;
722 
723 	max_key.objectid = ino;
724 	max_key.type = (u8)-1;
725 	max_key.offset = (u64)-1;
726 
727 	path->keep_locks = 1;
728 
729 	while(1) {
730 		ret = btrfs_search_forward(root, &min_key, &max_key,
731 					   path, 0, newer_than);
732 		if (ret != 0)
733 			goto none;
734 		if (min_key.objectid != ino)
735 			goto none;
736 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
737 			goto none;
738 
739 		leaf = path->nodes[0];
740 		extent = btrfs_item_ptr(leaf, path->slots[0],
741 					struct btrfs_file_extent_item);
742 
743 		type = btrfs_file_extent_type(leaf, extent);
744 		if (type == BTRFS_FILE_EXTENT_REG &&
745 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
746 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
747 			*off = min_key.offset;
748 			btrfs_free_path(path);
749 			return 0;
750 		}
751 
752 		if (min_key.offset == (u64)-1)
753 			goto none;
754 
755 		min_key.offset++;
756 		btrfs_release_path(path);
757 	}
758 none:
759 	btrfs_free_path(path);
760 	return -ENOENT;
761 }
762 
763 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
764 			       int thresh, u64 *last_len, u64 *skip,
765 			       u64 *defrag_end)
766 {
767 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
768 	struct extent_map *em = NULL;
769 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
770 	int ret = 1;
771 
772 	/*
773 	 * make sure that once we start defragging an extent, we keep on
774 	 * defragging it
775 	 */
776 	if (start < *defrag_end)
777 		return 1;
778 
779 	*skip = 0;
780 
781 	/*
782 	 * hopefully we have this extent in the tree already, try without
783 	 * the full extent lock
784 	 */
785 	read_lock(&em_tree->lock);
786 	em = lookup_extent_mapping(em_tree, start, len);
787 	read_unlock(&em_tree->lock);
788 
789 	if (!em) {
790 		/* get the big lock and read metadata off disk */
791 		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
792 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
793 		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
794 
795 		if (IS_ERR(em))
796 			return 0;
797 	}
798 
799 	/* this will cover holes, and inline extents */
800 	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
801 		ret = 0;
802 
803 	/*
804 	 * we hit a real extent, if it is big don't bother defragging it again
805 	 */
806 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
807 		ret = 0;
808 
809 	/*
810 	 * last_len ends up being a counter of how many bytes we've defragged.
811 	 * every time we choose not to defrag an extent, we reset *last_len
812 	 * so that the next tiny extent will force a defrag.
813 	 *
814 	 * The end result of this is that tiny extents before a single big
815 	 * extent will force at least part of that big extent to be defragged.
816 	 */
817 	if (ret) {
818 		*defrag_end = extent_map_end(em);
819 	} else {
820 		*last_len = 0;
821 		*skip = extent_map_end(em);
822 		*defrag_end = 0;
823 	}
824 
825 	free_extent_map(em);
826 	return ret;
827 }
828 
829 /*
830  * it doesn't do much good to defrag one or two pages
831  * at a time.  This pulls in a nice chunk of pages
832  * to COW and defrag.
833  *
834  * It also makes sure the delalloc code has enough
835  * dirty data to avoid making new small extents as part
836  * of the defrag
837  *
838  * It's a good idea to start RA on this range
839  * before calling this.
840  */
841 static int cluster_pages_for_defrag(struct inode *inode,
842 				    struct page **pages,
843 				    unsigned long start_index,
844 				    int num_pages)
845 {
846 	unsigned long file_end;
847 	u64 isize = i_size_read(inode);
848 	u64 page_start;
849 	u64 page_end;
850 	int ret;
851 	int i;
852 	int i_done;
853 	struct btrfs_ordered_extent *ordered;
854 	struct extent_state *cached_state = NULL;
855 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
856 
857 	if (isize == 0)
858 		return 0;
859 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
860 
861 	ret = btrfs_delalloc_reserve_space(inode,
862 					   num_pages << PAGE_CACHE_SHIFT);
863 	if (ret)
864 		return ret;
865 again:
866 	ret = 0;
867 	i_done = 0;
868 
869 	/* step one, lock all the pages */
870 	for (i = 0; i < num_pages; i++) {
871 		struct page *page;
872 		page = find_or_create_page(inode->i_mapping,
873 					    start_index + i, mask);
874 		if (!page)
875 			break;
876 
877 		if (!PageUptodate(page)) {
878 			btrfs_readpage(NULL, page);
879 			lock_page(page);
880 			if (!PageUptodate(page)) {
881 				unlock_page(page);
882 				page_cache_release(page);
883 				ret = -EIO;
884 				break;
885 			}
886 		}
887 		isize = i_size_read(inode);
888 		file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
889 		if (!isize || page->index > file_end ||
890 		    page->mapping != inode->i_mapping) {
891 			/* whoops, we blew past eof, skip this page */
892 			unlock_page(page);
893 			page_cache_release(page);
894 			break;
895 		}
896 		pages[i] = page;
897 		i_done++;
898 	}
899 	if (!i_done || ret)
900 		goto out;
901 
902 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
903 		goto out;
904 
905 	/*
906 	 * so now we have a nice long stream of locked
907 	 * and up to date pages, lets wait on them
908 	 */
909 	for (i = 0; i < i_done; i++)
910 		wait_on_page_writeback(pages[i]);
911 
912 	page_start = page_offset(pages[0]);
913 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
914 
915 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
916 			 page_start, page_end - 1, 0, &cached_state,
917 			 GFP_NOFS);
918 	ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
919 	if (ordered &&
920 	    ordered->file_offset + ordered->len > page_start &&
921 	    ordered->file_offset < page_end) {
922 		btrfs_put_ordered_extent(ordered);
923 		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
924 				     page_start, page_end - 1,
925 				     &cached_state, GFP_NOFS);
926 		for (i = 0; i < i_done; i++) {
927 			unlock_page(pages[i]);
928 			page_cache_release(pages[i]);
929 		}
930 		btrfs_wait_ordered_range(inode, page_start,
931 					 page_end - page_start);
932 		goto again;
933 	}
934 	if (ordered)
935 		btrfs_put_ordered_extent(ordered);
936 
937 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
938 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
939 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
940 			  GFP_NOFS);
941 
942 	if (i_done != num_pages) {
943 		spin_lock(&BTRFS_I(inode)->lock);
944 		BTRFS_I(inode)->outstanding_extents++;
945 		spin_unlock(&BTRFS_I(inode)->lock);
946 		btrfs_delalloc_release_space(inode,
947 				     (num_pages - i_done) << PAGE_CACHE_SHIFT);
948 	}
949 
950 
951 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
952 				  &cached_state);
953 
954 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
955 			     page_start, page_end - 1, &cached_state,
956 			     GFP_NOFS);
957 
958 	for (i = 0; i < i_done; i++) {
959 		clear_page_dirty_for_io(pages[i]);
960 		ClearPageChecked(pages[i]);
961 		set_page_extent_mapped(pages[i]);
962 		set_page_dirty(pages[i]);
963 		unlock_page(pages[i]);
964 		page_cache_release(pages[i]);
965 	}
966 	return i_done;
967 out:
968 	for (i = 0; i < i_done; i++) {
969 		unlock_page(pages[i]);
970 		page_cache_release(pages[i]);
971 	}
972 	btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
973 	return ret;
974 
975 }
976 
977 int btrfs_defrag_file(struct inode *inode, struct file *file,
978 		      struct btrfs_ioctl_defrag_range_args *range,
979 		      u64 newer_than, unsigned long max_to_defrag)
980 {
981 	struct btrfs_root *root = BTRFS_I(inode)->root;
982 	struct btrfs_super_block *disk_super;
983 	struct file_ra_state *ra = NULL;
984 	unsigned long last_index;
985 	u64 isize = i_size_read(inode);
986 	u64 features;
987 	u64 last_len = 0;
988 	u64 skip = 0;
989 	u64 defrag_end = 0;
990 	u64 newer_off = range->start;
991 	unsigned long i;
992 	unsigned long ra_index = 0;
993 	int ret;
994 	int defrag_count = 0;
995 	int compress_type = BTRFS_COMPRESS_ZLIB;
996 	int extent_thresh = range->extent_thresh;
997 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
998 	int cluster = max_cluster;
999 	u64 new_align = ~((u64)128 * 1024 - 1);
1000 	struct page **pages = NULL;
1001 
1002 	if (extent_thresh == 0)
1003 		extent_thresh = 256 * 1024;
1004 
1005 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1006 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1007 			return -EINVAL;
1008 		if (range->compress_type)
1009 			compress_type = range->compress_type;
1010 	}
1011 
1012 	if (isize == 0)
1013 		return 0;
1014 
1015 	/*
1016 	 * if we were not given a file, allocate a readahead
1017 	 * context
1018 	 */
1019 	if (!file) {
1020 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1021 		if (!ra)
1022 			return -ENOMEM;
1023 		file_ra_state_init(ra, inode->i_mapping);
1024 	} else {
1025 		ra = &file->f_ra;
1026 	}
1027 
1028 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1029 			GFP_NOFS);
1030 	if (!pages) {
1031 		ret = -ENOMEM;
1032 		goto out_ra;
1033 	}
1034 
1035 	/* find the last page to defrag */
1036 	if (range->start + range->len > range->start) {
1037 		last_index = min_t(u64, isize - 1,
1038 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1039 	} else {
1040 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1041 	}
1042 
1043 	if (newer_than) {
1044 		ret = find_new_extents(root, inode, newer_than,
1045 				       &newer_off, 64 * 1024);
1046 		if (!ret) {
1047 			range->start = newer_off;
1048 			/*
1049 			 * we always align our defrag to help keep
1050 			 * the extents in the file evenly spaced
1051 			 */
1052 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1053 		} else
1054 			goto out_ra;
1055 	} else {
1056 		i = range->start >> PAGE_CACHE_SHIFT;
1057 	}
1058 	if (!max_to_defrag)
1059 		max_to_defrag = last_index;
1060 
1061 	/*
1062 	 * make writeback starts from i, so the defrag range can be
1063 	 * written sequentially.
1064 	 */
1065 	if (i < inode->i_mapping->writeback_index)
1066 		inode->i_mapping->writeback_index = i;
1067 
1068 	while (i <= last_index && defrag_count < max_to_defrag &&
1069 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1070 		PAGE_CACHE_SHIFT)) {
1071 		/*
1072 		 * make sure we stop running if someone unmounts
1073 		 * the FS
1074 		 */
1075 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1076 			break;
1077 
1078 		if (!newer_than &&
1079 		    !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1080 					PAGE_CACHE_SIZE,
1081 					extent_thresh,
1082 					&last_len, &skip,
1083 					&defrag_end)) {
1084 			unsigned long next;
1085 			/*
1086 			 * the should_defrag function tells us how much to skip
1087 			 * bump our counter by the suggested amount
1088 			 */
1089 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1090 			i = max(i + 1, next);
1091 			continue;
1092 		}
1093 
1094 		if (!newer_than) {
1095 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1096 				   PAGE_CACHE_SHIFT) - i;
1097 			cluster = min(cluster, max_cluster);
1098 		} else {
1099 			cluster = max_cluster;
1100 		}
1101 
1102 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1103 			BTRFS_I(inode)->force_compress = compress_type;
1104 
1105 		if (i + cluster > ra_index) {
1106 			ra_index = max(i, ra_index);
1107 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1108 				       cluster);
1109 			ra_index += max_cluster;
1110 		}
1111 
1112 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1113 		if (ret < 0)
1114 			goto out_ra;
1115 
1116 		defrag_count += ret;
1117 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1118 
1119 		if (newer_than) {
1120 			if (newer_off == (u64)-1)
1121 				break;
1122 
1123 			newer_off = max(newer_off + 1,
1124 					(u64)i << PAGE_CACHE_SHIFT);
1125 
1126 			ret = find_new_extents(root, inode,
1127 					       newer_than, &newer_off,
1128 					       64 * 1024);
1129 			if (!ret) {
1130 				range->start = newer_off;
1131 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1132 			} else {
1133 				break;
1134 			}
1135 		} else {
1136 			if (ret > 0) {
1137 				i += ret;
1138 				last_len += ret << PAGE_CACHE_SHIFT;
1139 			} else {
1140 				i++;
1141 				last_len = 0;
1142 			}
1143 		}
1144 	}
1145 
1146 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1147 		filemap_flush(inode->i_mapping);
1148 
1149 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1150 		/* the filemap_flush will queue IO into the worker threads, but
1151 		 * we have to make sure the IO is actually started and that
1152 		 * ordered extents get created before we return
1153 		 */
1154 		atomic_inc(&root->fs_info->async_submit_draining);
1155 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1156 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1157 			wait_event(root->fs_info->async_submit_wait,
1158 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1159 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1160 		}
1161 		atomic_dec(&root->fs_info->async_submit_draining);
1162 
1163 		mutex_lock(&inode->i_mutex);
1164 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1165 		mutex_unlock(&inode->i_mutex);
1166 	}
1167 
1168 	disk_super = root->fs_info->super_copy;
1169 	features = btrfs_super_incompat_flags(disk_super);
1170 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1171 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1172 		btrfs_set_super_incompat_flags(disk_super, features);
1173 	}
1174 
1175 	ret = defrag_count;
1176 
1177 out_ra:
1178 	if (!file)
1179 		kfree(ra);
1180 	kfree(pages);
1181 	return ret;
1182 }
1183 
1184 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1185 					void __user *arg)
1186 {
1187 	u64 new_size;
1188 	u64 old_size;
1189 	u64 devid = 1;
1190 	struct btrfs_ioctl_vol_args *vol_args;
1191 	struct btrfs_trans_handle *trans;
1192 	struct btrfs_device *device = NULL;
1193 	char *sizestr;
1194 	char *devstr = NULL;
1195 	int ret = 0;
1196 	int mod = 0;
1197 
1198 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1199 		return -EROFS;
1200 
1201 	if (!capable(CAP_SYS_ADMIN))
1202 		return -EPERM;
1203 
1204 	vol_args = memdup_user(arg, sizeof(*vol_args));
1205 	if (IS_ERR(vol_args))
1206 		return PTR_ERR(vol_args);
1207 
1208 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1209 
1210 	mutex_lock(&root->fs_info->volume_mutex);
1211 	sizestr = vol_args->name;
1212 	devstr = strchr(sizestr, ':');
1213 	if (devstr) {
1214 		char *end;
1215 		sizestr = devstr + 1;
1216 		*devstr = '\0';
1217 		devstr = vol_args->name;
1218 		devid = simple_strtoull(devstr, &end, 10);
1219 		printk(KERN_INFO "resizing devid %llu\n",
1220 		       (unsigned long long)devid);
1221 	}
1222 	device = btrfs_find_device(root, devid, NULL, NULL);
1223 	if (!device) {
1224 		printk(KERN_INFO "resizer unable to find device %llu\n",
1225 		       (unsigned long long)devid);
1226 		ret = -EINVAL;
1227 		goto out_unlock;
1228 	}
1229 	if (!strcmp(sizestr, "max"))
1230 		new_size = device->bdev->bd_inode->i_size;
1231 	else {
1232 		if (sizestr[0] == '-') {
1233 			mod = -1;
1234 			sizestr++;
1235 		} else if (sizestr[0] == '+') {
1236 			mod = 1;
1237 			sizestr++;
1238 		}
1239 		new_size = memparse(sizestr, NULL);
1240 		if (new_size == 0) {
1241 			ret = -EINVAL;
1242 			goto out_unlock;
1243 		}
1244 	}
1245 
1246 	old_size = device->total_bytes;
1247 
1248 	if (mod < 0) {
1249 		if (new_size > old_size) {
1250 			ret = -EINVAL;
1251 			goto out_unlock;
1252 		}
1253 		new_size = old_size - new_size;
1254 	} else if (mod > 0) {
1255 		new_size = old_size + new_size;
1256 	}
1257 
1258 	if (new_size < 256 * 1024 * 1024) {
1259 		ret = -EINVAL;
1260 		goto out_unlock;
1261 	}
1262 	if (new_size > device->bdev->bd_inode->i_size) {
1263 		ret = -EFBIG;
1264 		goto out_unlock;
1265 	}
1266 
1267 	do_div(new_size, root->sectorsize);
1268 	new_size *= root->sectorsize;
1269 
1270 	printk(KERN_INFO "new size for %s is %llu\n",
1271 		device->name, (unsigned long long)new_size);
1272 
1273 	if (new_size > old_size) {
1274 		trans = btrfs_start_transaction(root, 0);
1275 		if (IS_ERR(trans)) {
1276 			ret = PTR_ERR(trans);
1277 			goto out_unlock;
1278 		}
1279 		ret = btrfs_grow_device(trans, device, new_size);
1280 		btrfs_commit_transaction(trans, root);
1281 	} else {
1282 		ret = btrfs_shrink_device(device, new_size);
1283 	}
1284 
1285 out_unlock:
1286 	mutex_unlock(&root->fs_info->volume_mutex);
1287 	kfree(vol_args);
1288 	return ret;
1289 }
1290 
1291 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1292 						    char *name,
1293 						    unsigned long fd,
1294 						    int subvol,
1295 						    u64 *transid,
1296 						    bool readonly)
1297 {
1298 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1299 	struct file *src_file;
1300 	int namelen;
1301 	int ret = 0;
1302 
1303 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1304 		return -EROFS;
1305 
1306 	namelen = strlen(name);
1307 	if (strchr(name, '/')) {
1308 		ret = -EINVAL;
1309 		goto out;
1310 	}
1311 
1312 	if (subvol) {
1313 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1314 				     NULL, transid, readonly);
1315 	} else {
1316 		struct inode *src_inode;
1317 		src_file = fget(fd);
1318 		if (!src_file) {
1319 			ret = -EINVAL;
1320 			goto out;
1321 		}
1322 
1323 		src_inode = src_file->f_path.dentry->d_inode;
1324 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1325 			printk(KERN_INFO "btrfs: Snapshot src from "
1326 			       "another FS\n");
1327 			ret = -EINVAL;
1328 			fput(src_file);
1329 			goto out;
1330 		}
1331 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1332 				     BTRFS_I(src_inode)->root,
1333 				     transid, readonly);
1334 		fput(src_file);
1335 	}
1336 out:
1337 	return ret;
1338 }
1339 
1340 static noinline int btrfs_ioctl_snap_create(struct file *file,
1341 					    void __user *arg, int subvol)
1342 {
1343 	struct btrfs_ioctl_vol_args *vol_args;
1344 	int ret;
1345 
1346 	vol_args = memdup_user(arg, sizeof(*vol_args));
1347 	if (IS_ERR(vol_args))
1348 		return PTR_ERR(vol_args);
1349 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1350 
1351 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1352 					      vol_args->fd, subvol,
1353 					      NULL, false);
1354 
1355 	kfree(vol_args);
1356 	return ret;
1357 }
1358 
1359 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1360 					       void __user *arg, int subvol)
1361 {
1362 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1363 	int ret;
1364 	u64 transid = 0;
1365 	u64 *ptr = NULL;
1366 	bool readonly = false;
1367 
1368 	vol_args = memdup_user(arg, sizeof(*vol_args));
1369 	if (IS_ERR(vol_args))
1370 		return PTR_ERR(vol_args);
1371 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1372 
1373 	if (vol_args->flags &
1374 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1375 		ret = -EOPNOTSUPP;
1376 		goto out;
1377 	}
1378 
1379 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1380 		ptr = &transid;
1381 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1382 		readonly = true;
1383 
1384 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1385 					      vol_args->fd, subvol,
1386 					      ptr, readonly);
1387 
1388 	if (ret == 0 && ptr &&
1389 	    copy_to_user(arg +
1390 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1391 				  transid), ptr, sizeof(*ptr)))
1392 		ret = -EFAULT;
1393 out:
1394 	kfree(vol_args);
1395 	return ret;
1396 }
1397 
1398 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1399 						void __user *arg)
1400 {
1401 	struct inode *inode = fdentry(file)->d_inode;
1402 	struct btrfs_root *root = BTRFS_I(inode)->root;
1403 	int ret = 0;
1404 	u64 flags = 0;
1405 
1406 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1407 		return -EINVAL;
1408 
1409 	down_read(&root->fs_info->subvol_sem);
1410 	if (btrfs_root_readonly(root))
1411 		flags |= BTRFS_SUBVOL_RDONLY;
1412 	up_read(&root->fs_info->subvol_sem);
1413 
1414 	if (copy_to_user(arg, &flags, sizeof(flags)))
1415 		ret = -EFAULT;
1416 
1417 	return ret;
1418 }
1419 
1420 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1421 					      void __user *arg)
1422 {
1423 	struct inode *inode = fdentry(file)->d_inode;
1424 	struct btrfs_root *root = BTRFS_I(inode)->root;
1425 	struct btrfs_trans_handle *trans;
1426 	u64 root_flags;
1427 	u64 flags;
1428 	int ret = 0;
1429 
1430 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1431 		return -EROFS;
1432 
1433 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1434 		return -EINVAL;
1435 
1436 	if (copy_from_user(&flags, arg, sizeof(flags)))
1437 		return -EFAULT;
1438 
1439 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1440 		return -EINVAL;
1441 
1442 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1443 		return -EOPNOTSUPP;
1444 
1445 	if (!inode_owner_or_capable(inode))
1446 		return -EACCES;
1447 
1448 	down_write(&root->fs_info->subvol_sem);
1449 
1450 	/* nothing to do */
1451 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1452 		goto out;
1453 
1454 	root_flags = btrfs_root_flags(&root->root_item);
1455 	if (flags & BTRFS_SUBVOL_RDONLY)
1456 		btrfs_set_root_flags(&root->root_item,
1457 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1458 	else
1459 		btrfs_set_root_flags(&root->root_item,
1460 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1461 
1462 	trans = btrfs_start_transaction(root, 1);
1463 	if (IS_ERR(trans)) {
1464 		ret = PTR_ERR(trans);
1465 		goto out_reset;
1466 	}
1467 
1468 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1469 				&root->root_key, &root->root_item);
1470 
1471 	btrfs_commit_transaction(trans, root);
1472 out_reset:
1473 	if (ret)
1474 		btrfs_set_root_flags(&root->root_item, root_flags);
1475 out:
1476 	up_write(&root->fs_info->subvol_sem);
1477 	return ret;
1478 }
1479 
1480 /*
1481  * helper to check if the subvolume references other subvolumes
1482  */
1483 static noinline int may_destroy_subvol(struct btrfs_root *root)
1484 {
1485 	struct btrfs_path *path;
1486 	struct btrfs_key key;
1487 	int ret;
1488 
1489 	path = btrfs_alloc_path();
1490 	if (!path)
1491 		return -ENOMEM;
1492 
1493 	key.objectid = root->root_key.objectid;
1494 	key.type = BTRFS_ROOT_REF_KEY;
1495 	key.offset = (u64)-1;
1496 
1497 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1498 				&key, path, 0, 0);
1499 	if (ret < 0)
1500 		goto out;
1501 	BUG_ON(ret == 0);
1502 
1503 	ret = 0;
1504 	if (path->slots[0] > 0) {
1505 		path->slots[0]--;
1506 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1507 		if (key.objectid == root->root_key.objectid &&
1508 		    key.type == BTRFS_ROOT_REF_KEY)
1509 			ret = -ENOTEMPTY;
1510 	}
1511 out:
1512 	btrfs_free_path(path);
1513 	return ret;
1514 }
1515 
1516 static noinline int key_in_sk(struct btrfs_key *key,
1517 			      struct btrfs_ioctl_search_key *sk)
1518 {
1519 	struct btrfs_key test;
1520 	int ret;
1521 
1522 	test.objectid = sk->min_objectid;
1523 	test.type = sk->min_type;
1524 	test.offset = sk->min_offset;
1525 
1526 	ret = btrfs_comp_cpu_keys(key, &test);
1527 	if (ret < 0)
1528 		return 0;
1529 
1530 	test.objectid = sk->max_objectid;
1531 	test.type = sk->max_type;
1532 	test.offset = sk->max_offset;
1533 
1534 	ret = btrfs_comp_cpu_keys(key, &test);
1535 	if (ret > 0)
1536 		return 0;
1537 	return 1;
1538 }
1539 
1540 static noinline int copy_to_sk(struct btrfs_root *root,
1541 			       struct btrfs_path *path,
1542 			       struct btrfs_key *key,
1543 			       struct btrfs_ioctl_search_key *sk,
1544 			       char *buf,
1545 			       unsigned long *sk_offset,
1546 			       int *num_found)
1547 {
1548 	u64 found_transid;
1549 	struct extent_buffer *leaf;
1550 	struct btrfs_ioctl_search_header sh;
1551 	unsigned long item_off;
1552 	unsigned long item_len;
1553 	int nritems;
1554 	int i;
1555 	int slot;
1556 	int ret = 0;
1557 
1558 	leaf = path->nodes[0];
1559 	slot = path->slots[0];
1560 	nritems = btrfs_header_nritems(leaf);
1561 
1562 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1563 		i = nritems;
1564 		goto advance_key;
1565 	}
1566 	found_transid = btrfs_header_generation(leaf);
1567 
1568 	for (i = slot; i < nritems; i++) {
1569 		item_off = btrfs_item_ptr_offset(leaf, i);
1570 		item_len = btrfs_item_size_nr(leaf, i);
1571 
1572 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1573 			item_len = 0;
1574 
1575 		if (sizeof(sh) + item_len + *sk_offset >
1576 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1577 			ret = 1;
1578 			goto overflow;
1579 		}
1580 
1581 		btrfs_item_key_to_cpu(leaf, key, i);
1582 		if (!key_in_sk(key, sk))
1583 			continue;
1584 
1585 		sh.objectid = key->objectid;
1586 		sh.offset = key->offset;
1587 		sh.type = key->type;
1588 		sh.len = item_len;
1589 		sh.transid = found_transid;
1590 
1591 		/* copy search result header */
1592 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1593 		*sk_offset += sizeof(sh);
1594 
1595 		if (item_len) {
1596 			char *p = buf + *sk_offset;
1597 			/* copy the item */
1598 			read_extent_buffer(leaf, p,
1599 					   item_off, item_len);
1600 			*sk_offset += item_len;
1601 		}
1602 		(*num_found)++;
1603 
1604 		if (*num_found >= sk->nr_items)
1605 			break;
1606 	}
1607 advance_key:
1608 	ret = 0;
1609 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1610 		key->offset++;
1611 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1612 		key->offset = 0;
1613 		key->type++;
1614 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1615 		key->offset = 0;
1616 		key->type = 0;
1617 		key->objectid++;
1618 	} else
1619 		ret = 1;
1620 overflow:
1621 	return ret;
1622 }
1623 
1624 static noinline int search_ioctl(struct inode *inode,
1625 				 struct btrfs_ioctl_search_args *args)
1626 {
1627 	struct btrfs_root *root;
1628 	struct btrfs_key key;
1629 	struct btrfs_key max_key;
1630 	struct btrfs_path *path;
1631 	struct btrfs_ioctl_search_key *sk = &args->key;
1632 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1633 	int ret;
1634 	int num_found = 0;
1635 	unsigned long sk_offset = 0;
1636 
1637 	path = btrfs_alloc_path();
1638 	if (!path)
1639 		return -ENOMEM;
1640 
1641 	if (sk->tree_id == 0) {
1642 		/* search the root of the inode that was passed */
1643 		root = BTRFS_I(inode)->root;
1644 	} else {
1645 		key.objectid = sk->tree_id;
1646 		key.type = BTRFS_ROOT_ITEM_KEY;
1647 		key.offset = (u64)-1;
1648 		root = btrfs_read_fs_root_no_name(info, &key);
1649 		if (IS_ERR(root)) {
1650 			printk(KERN_ERR "could not find root %llu\n",
1651 			       sk->tree_id);
1652 			btrfs_free_path(path);
1653 			return -ENOENT;
1654 		}
1655 	}
1656 
1657 	key.objectid = sk->min_objectid;
1658 	key.type = sk->min_type;
1659 	key.offset = sk->min_offset;
1660 
1661 	max_key.objectid = sk->max_objectid;
1662 	max_key.type = sk->max_type;
1663 	max_key.offset = sk->max_offset;
1664 
1665 	path->keep_locks = 1;
1666 
1667 	while(1) {
1668 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1669 					   sk->min_transid);
1670 		if (ret != 0) {
1671 			if (ret > 0)
1672 				ret = 0;
1673 			goto err;
1674 		}
1675 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1676 				 &sk_offset, &num_found);
1677 		btrfs_release_path(path);
1678 		if (ret || num_found >= sk->nr_items)
1679 			break;
1680 
1681 	}
1682 	ret = 0;
1683 err:
1684 	sk->nr_items = num_found;
1685 	btrfs_free_path(path);
1686 	return ret;
1687 }
1688 
1689 static noinline int btrfs_ioctl_tree_search(struct file *file,
1690 					   void __user *argp)
1691 {
1692 	 struct btrfs_ioctl_search_args *args;
1693 	 struct inode *inode;
1694 	 int ret;
1695 
1696 	if (!capable(CAP_SYS_ADMIN))
1697 		return -EPERM;
1698 
1699 	args = memdup_user(argp, sizeof(*args));
1700 	if (IS_ERR(args))
1701 		return PTR_ERR(args);
1702 
1703 	inode = fdentry(file)->d_inode;
1704 	ret = search_ioctl(inode, args);
1705 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1706 		ret = -EFAULT;
1707 	kfree(args);
1708 	return ret;
1709 }
1710 
1711 /*
1712  * Search INODE_REFs to identify path name of 'dirid' directory
1713  * in a 'tree_id' tree. and sets path name to 'name'.
1714  */
1715 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1716 				u64 tree_id, u64 dirid, char *name)
1717 {
1718 	struct btrfs_root *root;
1719 	struct btrfs_key key;
1720 	char *ptr;
1721 	int ret = -1;
1722 	int slot;
1723 	int len;
1724 	int total_len = 0;
1725 	struct btrfs_inode_ref *iref;
1726 	struct extent_buffer *l;
1727 	struct btrfs_path *path;
1728 
1729 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1730 		name[0]='\0';
1731 		return 0;
1732 	}
1733 
1734 	path = btrfs_alloc_path();
1735 	if (!path)
1736 		return -ENOMEM;
1737 
1738 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1739 
1740 	key.objectid = tree_id;
1741 	key.type = BTRFS_ROOT_ITEM_KEY;
1742 	key.offset = (u64)-1;
1743 	root = btrfs_read_fs_root_no_name(info, &key);
1744 	if (IS_ERR(root)) {
1745 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1746 		ret = -ENOENT;
1747 		goto out;
1748 	}
1749 
1750 	key.objectid = dirid;
1751 	key.type = BTRFS_INODE_REF_KEY;
1752 	key.offset = (u64)-1;
1753 
1754 	while(1) {
1755 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1756 		if (ret < 0)
1757 			goto out;
1758 
1759 		l = path->nodes[0];
1760 		slot = path->slots[0];
1761 		if (ret > 0 && slot > 0)
1762 			slot--;
1763 		btrfs_item_key_to_cpu(l, &key, slot);
1764 
1765 		if (ret > 0 && (key.objectid != dirid ||
1766 				key.type != BTRFS_INODE_REF_KEY)) {
1767 			ret = -ENOENT;
1768 			goto out;
1769 		}
1770 
1771 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1772 		len = btrfs_inode_ref_name_len(l, iref);
1773 		ptr -= len + 1;
1774 		total_len += len + 1;
1775 		if (ptr < name)
1776 			goto out;
1777 
1778 		*(ptr + len) = '/';
1779 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1780 
1781 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1782 			break;
1783 
1784 		btrfs_release_path(path);
1785 		key.objectid = key.offset;
1786 		key.offset = (u64)-1;
1787 		dirid = key.objectid;
1788 	}
1789 	if (ptr < name)
1790 		goto out;
1791 	memmove(name, ptr, total_len);
1792 	name[total_len]='\0';
1793 	ret = 0;
1794 out:
1795 	btrfs_free_path(path);
1796 	return ret;
1797 }
1798 
1799 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1800 					   void __user *argp)
1801 {
1802 	 struct btrfs_ioctl_ino_lookup_args *args;
1803 	 struct inode *inode;
1804 	 int ret;
1805 
1806 	if (!capable(CAP_SYS_ADMIN))
1807 		return -EPERM;
1808 
1809 	args = memdup_user(argp, sizeof(*args));
1810 	if (IS_ERR(args))
1811 		return PTR_ERR(args);
1812 
1813 	inode = fdentry(file)->d_inode;
1814 
1815 	if (args->treeid == 0)
1816 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1817 
1818 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1819 					args->treeid, args->objectid,
1820 					args->name);
1821 
1822 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1823 		ret = -EFAULT;
1824 
1825 	kfree(args);
1826 	return ret;
1827 }
1828 
1829 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1830 					     void __user *arg)
1831 {
1832 	struct dentry *parent = fdentry(file);
1833 	struct dentry *dentry;
1834 	struct inode *dir = parent->d_inode;
1835 	struct inode *inode;
1836 	struct btrfs_root *root = BTRFS_I(dir)->root;
1837 	struct btrfs_root *dest = NULL;
1838 	struct btrfs_ioctl_vol_args *vol_args;
1839 	struct btrfs_trans_handle *trans;
1840 	int namelen;
1841 	int ret;
1842 	int err = 0;
1843 
1844 	vol_args = memdup_user(arg, sizeof(*vol_args));
1845 	if (IS_ERR(vol_args))
1846 		return PTR_ERR(vol_args);
1847 
1848 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1849 	namelen = strlen(vol_args->name);
1850 	if (strchr(vol_args->name, '/') ||
1851 	    strncmp(vol_args->name, "..", namelen) == 0) {
1852 		err = -EINVAL;
1853 		goto out;
1854 	}
1855 
1856 	err = mnt_want_write(file->f_path.mnt);
1857 	if (err)
1858 		goto out;
1859 
1860 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1861 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1862 	if (IS_ERR(dentry)) {
1863 		err = PTR_ERR(dentry);
1864 		goto out_unlock_dir;
1865 	}
1866 
1867 	if (!dentry->d_inode) {
1868 		err = -ENOENT;
1869 		goto out_dput;
1870 	}
1871 
1872 	inode = dentry->d_inode;
1873 	dest = BTRFS_I(inode)->root;
1874 	if (!capable(CAP_SYS_ADMIN)){
1875 		/*
1876 		 * Regular user.  Only allow this with a special mount
1877 		 * option, when the user has write+exec access to the
1878 		 * subvol root, and when rmdir(2) would have been
1879 		 * allowed.
1880 		 *
1881 		 * Note that this is _not_ check that the subvol is
1882 		 * empty or doesn't contain data that we wouldn't
1883 		 * otherwise be able to delete.
1884 		 *
1885 		 * Users who want to delete empty subvols should try
1886 		 * rmdir(2).
1887 		 */
1888 		err = -EPERM;
1889 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1890 			goto out_dput;
1891 
1892 		/*
1893 		 * Do not allow deletion if the parent dir is the same
1894 		 * as the dir to be deleted.  That means the ioctl
1895 		 * must be called on the dentry referencing the root
1896 		 * of the subvol, not a random directory contained
1897 		 * within it.
1898 		 */
1899 		err = -EINVAL;
1900 		if (root == dest)
1901 			goto out_dput;
1902 
1903 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1904 		if (err)
1905 			goto out_dput;
1906 
1907 		/* check if subvolume may be deleted by a non-root user */
1908 		err = btrfs_may_delete(dir, dentry, 1);
1909 		if (err)
1910 			goto out_dput;
1911 	}
1912 
1913 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1914 		err = -EINVAL;
1915 		goto out_dput;
1916 	}
1917 
1918 	mutex_lock(&inode->i_mutex);
1919 	err = d_invalidate(dentry);
1920 	if (err)
1921 		goto out_unlock;
1922 
1923 	down_write(&root->fs_info->subvol_sem);
1924 
1925 	err = may_destroy_subvol(dest);
1926 	if (err)
1927 		goto out_up_write;
1928 
1929 	trans = btrfs_start_transaction(root, 0);
1930 	if (IS_ERR(trans)) {
1931 		err = PTR_ERR(trans);
1932 		goto out_up_write;
1933 	}
1934 	trans->block_rsv = &root->fs_info->global_block_rsv;
1935 
1936 	ret = btrfs_unlink_subvol(trans, root, dir,
1937 				dest->root_key.objectid,
1938 				dentry->d_name.name,
1939 				dentry->d_name.len);
1940 	BUG_ON(ret);
1941 
1942 	btrfs_record_root_in_trans(trans, dest);
1943 
1944 	memset(&dest->root_item.drop_progress, 0,
1945 		sizeof(dest->root_item.drop_progress));
1946 	dest->root_item.drop_level = 0;
1947 	btrfs_set_root_refs(&dest->root_item, 0);
1948 
1949 	if (!xchg(&dest->orphan_item_inserted, 1)) {
1950 		ret = btrfs_insert_orphan_item(trans,
1951 					root->fs_info->tree_root,
1952 					dest->root_key.objectid);
1953 		BUG_ON(ret);
1954 	}
1955 
1956 	ret = btrfs_end_transaction(trans, root);
1957 	BUG_ON(ret);
1958 	inode->i_flags |= S_DEAD;
1959 out_up_write:
1960 	up_write(&root->fs_info->subvol_sem);
1961 out_unlock:
1962 	mutex_unlock(&inode->i_mutex);
1963 	if (!err) {
1964 		shrink_dcache_sb(root->fs_info->sb);
1965 		btrfs_invalidate_inodes(dest);
1966 		d_delete(dentry);
1967 	}
1968 out_dput:
1969 	dput(dentry);
1970 out_unlock_dir:
1971 	mutex_unlock(&dir->i_mutex);
1972 	mnt_drop_write(file->f_path.mnt);
1973 out:
1974 	kfree(vol_args);
1975 	return err;
1976 }
1977 
1978 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1979 {
1980 	struct inode *inode = fdentry(file)->d_inode;
1981 	struct btrfs_root *root = BTRFS_I(inode)->root;
1982 	struct btrfs_ioctl_defrag_range_args *range;
1983 	int ret;
1984 
1985 	if (btrfs_root_readonly(root))
1986 		return -EROFS;
1987 
1988 	ret = mnt_want_write(file->f_path.mnt);
1989 	if (ret)
1990 		return ret;
1991 
1992 	switch (inode->i_mode & S_IFMT) {
1993 	case S_IFDIR:
1994 		if (!capable(CAP_SYS_ADMIN)) {
1995 			ret = -EPERM;
1996 			goto out;
1997 		}
1998 		ret = btrfs_defrag_root(root, 0);
1999 		if (ret)
2000 			goto out;
2001 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2002 		break;
2003 	case S_IFREG:
2004 		if (!(file->f_mode & FMODE_WRITE)) {
2005 			ret = -EINVAL;
2006 			goto out;
2007 		}
2008 
2009 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2010 		if (!range) {
2011 			ret = -ENOMEM;
2012 			goto out;
2013 		}
2014 
2015 		if (argp) {
2016 			if (copy_from_user(range, argp,
2017 					   sizeof(*range))) {
2018 				ret = -EFAULT;
2019 				kfree(range);
2020 				goto out;
2021 			}
2022 			/* compression requires us to start the IO */
2023 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2024 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2025 				range->extent_thresh = (u32)-1;
2026 			}
2027 		} else {
2028 			/* the rest are all set to zero by kzalloc */
2029 			range->len = (u64)-1;
2030 		}
2031 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2032 					range, 0, 0);
2033 		if (ret > 0)
2034 			ret = 0;
2035 		kfree(range);
2036 		break;
2037 	default:
2038 		ret = -EINVAL;
2039 	}
2040 out:
2041 	mnt_drop_write(file->f_path.mnt);
2042 	return ret;
2043 }
2044 
2045 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2046 {
2047 	struct btrfs_ioctl_vol_args *vol_args;
2048 	int ret;
2049 
2050 	if (!capable(CAP_SYS_ADMIN))
2051 		return -EPERM;
2052 
2053 	vol_args = memdup_user(arg, sizeof(*vol_args));
2054 	if (IS_ERR(vol_args))
2055 		return PTR_ERR(vol_args);
2056 
2057 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2058 	ret = btrfs_init_new_device(root, vol_args->name);
2059 
2060 	kfree(vol_args);
2061 	return ret;
2062 }
2063 
2064 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2065 {
2066 	struct btrfs_ioctl_vol_args *vol_args;
2067 	int ret;
2068 
2069 	if (!capable(CAP_SYS_ADMIN))
2070 		return -EPERM;
2071 
2072 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2073 		return -EROFS;
2074 
2075 	vol_args = memdup_user(arg, sizeof(*vol_args));
2076 	if (IS_ERR(vol_args))
2077 		return PTR_ERR(vol_args);
2078 
2079 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2080 	ret = btrfs_rm_device(root, vol_args->name);
2081 
2082 	kfree(vol_args);
2083 	return ret;
2084 }
2085 
2086 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2087 {
2088 	struct btrfs_ioctl_fs_info_args *fi_args;
2089 	struct btrfs_device *device;
2090 	struct btrfs_device *next;
2091 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2092 	int ret = 0;
2093 
2094 	if (!capable(CAP_SYS_ADMIN))
2095 		return -EPERM;
2096 
2097 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2098 	if (!fi_args)
2099 		return -ENOMEM;
2100 
2101 	fi_args->num_devices = fs_devices->num_devices;
2102 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2103 
2104 	mutex_lock(&fs_devices->device_list_mutex);
2105 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2106 		if (device->devid > fi_args->max_id)
2107 			fi_args->max_id = device->devid;
2108 	}
2109 	mutex_unlock(&fs_devices->device_list_mutex);
2110 
2111 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2112 		ret = -EFAULT;
2113 
2114 	kfree(fi_args);
2115 	return ret;
2116 }
2117 
2118 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2119 {
2120 	struct btrfs_ioctl_dev_info_args *di_args;
2121 	struct btrfs_device *dev;
2122 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2123 	int ret = 0;
2124 	char *s_uuid = NULL;
2125 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2126 
2127 	if (!capable(CAP_SYS_ADMIN))
2128 		return -EPERM;
2129 
2130 	di_args = memdup_user(arg, sizeof(*di_args));
2131 	if (IS_ERR(di_args))
2132 		return PTR_ERR(di_args);
2133 
2134 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2135 		s_uuid = di_args->uuid;
2136 
2137 	mutex_lock(&fs_devices->device_list_mutex);
2138 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2139 	mutex_unlock(&fs_devices->device_list_mutex);
2140 
2141 	if (!dev) {
2142 		ret = -ENODEV;
2143 		goto out;
2144 	}
2145 
2146 	di_args->devid = dev->devid;
2147 	di_args->bytes_used = dev->bytes_used;
2148 	di_args->total_bytes = dev->total_bytes;
2149 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2150 	strncpy(di_args->path, dev->name, sizeof(di_args->path));
2151 
2152 out:
2153 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2154 		ret = -EFAULT;
2155 
2156 	kfree(di_args);
2157 	return ret;
2158 }
2159 
2160 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2161 				       u64 off, u64 olen, u64 destoff)
2162 {
2163 	struct inode *inode = fdentry(file)->d_inode;
2164 	struct btrfs_root *root = BTRFS_I(inode)->root;
2165 	struct file *src_file;
2166 	struct inode *src;
2167 	struct btrfs_trans_handle *trans;
2168 	struct btrfs_path *path;
2169 	struct extent_buffer *leaf;
2170 	char *buf;
2171 	struct btrfs_key key;
2172 	u32 nritems;
2173 	int slot;
2174 	int ret;
2175 	u64 len = olen;
2176 	u64 bs = root->fs_info->sb->s_blocksize;
2177 	u64 hint_byte;
2178 
2179 	/*
2180 	 * TODO:
2181 	 * - split compressed inline extents.  annoying: we need to
2182 	 *   decompress into destination's address_space (the file offset
2183 	 *   may change, so source mapping won't do), then recompress (or
2184 	 *   otherwise reinsert) a subrange.
2185 	 * - allow ranges within the same file to be cloned (provided
2186 	 *   they don't overlap)?
2187 	 */
2188 
2189 	/* the destination must be opened for writing */
2190 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2191 		return -EINVAL;
2192 
2193 	if (btrfs_root_readonly(root))
2194 		return -EROFS;
2195 
2196 	ret = mnt_want_write(file->f_path.mnt);
2197 	if (ret)
2198 		return ret;
2199 
2200 	src_file = fget(srcfd);
2201 	if (!src_file) {
2202 		ret = -EBADF;
2203 		goto out_drop_write;
2204 	}
2205 
2206 	src = src_file->f_dentry->d_inode;
2207 
2208 	ret = -EINVAL;
2209 	if (src == inode)
2210 		goto out_fput;
2211 
2212 	/* the src must be open for reading */
2213 	if (!(src_file->f_mode & FMODE_READ))
2214 		goto out_fput;
2215 
2216 	/* don't make the dst file partly checksummed */
2217 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2218 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2219 		goto out_fput;
2220 
2221 	ret = -EISDIR;
2222 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2223 		goto out_fput;
2224 
2225 	ret = -EXDEV;
2226 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2227 		goto out_fput;
2228 
2229 	ret = -ENOMEM;
2230 	buf = vmalloc(btrfs_level_size(root, 0));
2231 	if (!buf)
2232 		goto out_fput;
2233 
2234 	path = btrfs_alloc_path();
2235 	if (!path) {
2236 		vfree(buf);
2237 		goto out_fput;
2238 	}
2239 	path->reada = 2;
2240 
2241 	if (inode < src) {
2242 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2243 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2244 	} else {
2245 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2246 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2247 	}
2248 
2249 	/* determine range to clone */
2250 	ret = -EINVAL;
2251 	if (off + len > src->i_size || off + len < off)
2252 		goto out_unlock;
2253 	if (len == 0)
2254 		olen = len = src->i_size - off;
2255 	/* if we extend to eof, continue to block boundary */
2256 	if (off + len == src->i_size)
2257 		len = ALIGN(src->i_size, bs) - off;
2258 
2259 	/* verify the end result is block aligned */
2260 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2261 	    !IS_ALIGNED(destoff, bs))
2262 		goto out_unlock;
2263 
2264 	if (destoff > inode->i_size) {
2265 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2266 		if (ret)
2267 			goto out_unlock;
2268 	}
2269 
2270 	/* truncate page cache pages from target inode range */
2271 	truncate_inode_pages_range(&inode->i_data, destoff,
2272 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2273 
2274 	/* do any pending delalloc/csum calc on src, one way or
2275 	   another, and lock file content */
2276 	while (1) {
2277 		struct btrfs_ordered_extent *ordered;
2278 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2279 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2280 		if (!ordered &&
2281 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2282 				   EXTENT_DELALLOC, 0, NULL))
2283 			break;
2284 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2285 		if (ordered)
2286 			btrfs_put_ordered_extent(ordered);
2287 		btrfs_wait_ordered_range(src, off, len);
2288 	}
2289 
2290 	/* clone data */
2291 	key.objectid = btrfs_ino(src);
2292 	key.type = BTRFS_EXTENT_DATA_KEY;
2293 	key.offset = 0;
2294 
2295 	while (1) {
2296 		/*
2297 		 * note the key will change type as we walk through the
2298 		 * tree.
2299 		 */
2300 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2301 		if (ret < 0)
2302 			goto out;
2303 
2304 		nritems = btrfs_header_nritems(path->nodes[0]);
2305 		if (path->slots[0] >= nritems) {
2306 			ret = btrfs_next_leaf(root, path);
2307 			if (ret < 0)
2308 				goto out;
2309 			if (ret > 0)
2310 				break;
2311 			nritems = btrfs_header_nritems(path->nodes[0]);
2312 		}
2313 		leaf = path->nodes[0];
2314 		slot = path->slots[0];
2315 
2316 		btrfs_item_key_to_cpu(leaf, &key, slot);
2317 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2318 		    key.objectid != btrfs_ino(src))
2319 			break;
2320 
2321 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2322 			struct btrfs_file_extent_item *extent;
2323 			int type;
2324 			u32 size;
2325 			struct btrfs_key new_key;
2326 			u64 disko = 0, diskl = 0;
2327 			u64 datao = 0, datal = 0;
2328 			u8 comp;
2329 			u64 endoff;
2330 
2331 			size = btrfs_item_size_nr(leaf, slot);
2332 			read_extent_buffer(leaf, buf,
2333 					   btrfs_item_ptr_offset(leaf, slot),
2334 					   size);
2335 
2336 			extent = btrfs_item_ptr(leaf, slot,
2337 						struct btrfs_file_extent_item);
2338 			comp = btrfs_file_extent_compression(leaf, extent);
2339 			type = btrfs_file_extent_type(leaf, extent);
2340 			if (type == BTRFS_FILE_EXTENT_REG ||
2341 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2342 				disko = btrfs_file_extent_disk_bytenr(leaf,
2343 								      extent);
2344 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2345 								 extent);
2346 				datao = btrfs_file_extent_offset(leaf, extent);
2347 				datal = btrfs_file_extent_num_bytes(leaf,
2348 								    extent);
2349 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2350 				/* take upper bound, may be compressed */
2351 				datal = btrfs_file_extent_ram_bytes(leaf,
2352 								    extent);
2353 			}
2354 			btrfs_release_path(path);
2355 
2356 			if (key.offset + datal <= off ||
2357 			    key.offset >= off+len)
2358 				goto next;
2359 
2360 			memcpy(&new_key, &key, sizeof(new_key));
2361 			new_key.objectid = btrfs_ino(inode);
2362 			if (off <= key.offset)
2363 				new_key.offset = key.offset + destoff - off;
2364 			else
2365 				new_key.offset = destoff;
2366 
2367 			/*
2368 			 * 1 - adjusting old extent (we may have to split it)
2369 			 * 1 - add new extent
2370 			 * 1 - inode update
2371 			 */
2372 			trans = btrfs_start_transaction(root, 3);
2373 			if (IS_ERR(trans)) {
2374 				ret = PTR_ERR(trans);
2375 				goto out;
2376 			}
2377 
2378 			if (type == BTRFS_FILE_EXTENT_REG ||
2379 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2380 				/*
2381 				 *    a  | --- range to clone ---|  b
2382 				 * | ------------- extent ------------- |
2383 				 */
2384 
2385 				/* substract range b */
2386 				if (key.offset + datal > off + len)
2387 					datal = off + len - key.offset;
2388 
2389 				/* substract range a */
2390 				if (off > key.offset) {
2391 					datao += off - key.offset;
2392 					datal -= off - key.offset;
2393 				}
2394 
2395 				ret = btrfs_drop_extents(trans, inode,
2396 							 new_key.offset,
2397 							 new_key.offset + datal,
2398 							 &hint_byte, 1);
2399 				BUG_ON(ret);
2400 
2401 				ret = btrfs_insert_empty_item(trans, root, path,
2402 							      &new_key, size);
2403 				BUG_ON(ret);
2404 
2405 				leaf = path->nodes[0];
2406 				slot = path->slots[0];
2407 				write_extent_buffer(leaf, buf,
2408 					    btrfs_item_ptr_offset(leaf, slot),
2409 					    size);
2410 
2411 				extent = btrfs_item_ptr(leaf, slot,
2412 						struct btrfs_file_extent_item);
2413 
2414 				/* disko == 0 means it's a hole */
2415 				if (!disko)
2416 					datao = 0;
2417 
2418 				btrfs_set_file_extent_offset(leaf, extent,
2419 							     datao);
2420 				btrfs_set_file_extent_num_bytes(leaf, extent,
2421 								datal);
2422 				if (disko) {
2423 					inode_add_bytes(inode, datal);
2424 					ret = btrfs_inc_extent_ref(trans, root,
2425 							disko, diskl, 0,
2426 							root->root_key.objectid,
2427 							btrfs_ino(inode),
2428 							new_key.offset - datao);
2429 					BUG_ON(ret);
2430 				}
2431 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2432 				u64 skip = 0;
2433 				u64 trim = 0;
2434 				if (off > key.offset) {
2435 					skip = off - key.offset;
2436 					new_key.offset += skip;
2437 				}
2438 
2439 				if (key.offset + datal > off+len)
2440 					trim = key.offset + datal - (off+len);
2441 
2442 				if (comp && (skip || trim)) {
2443 					ret = -EINVAL;
2444 					btrfs_end_transaction(trans, root);
2445 					goto out;
2446 				}
2447 				size -= skip + trim;
2448 				datal -= skip + trim;
2449 
2450 				ret = btrfs_drop_extents(trans, inode,
2451 							 new_key.offset,
2452 							 new_key.offset + datal,
2453 							 &hint_byte, 1);
2454 				BUG_ON(ret);
2455 
2456 				ret = btrfs_insert_empty_item(trans, root, path,
2457 							      &new_key, size);
2458 				BUG_ON(ret);
2459 
2460 				if (skip) {
2461 					u32 start =
2462 					  btrfs_file_extent_calc_inline_size(0);
2463 					memmove(buf+start, buf+start+skip,
2464 						datal);
2465 				}
2466 
2467 				leaf = path->nodes[0];
2468 				slot = path->slots[0];
2469 				write_extent_buffer(leaf, buf,
2470 					    btrfs_item_ptr_offset(leaf, slot),
2471 					    size);
2472 				inode_add_bytes(inode, datal);
2473 			}
2474 
2475 			btrfs_mark_buffer_dirty(leaf);
2476 			btrfs_release_path(path);
2477 
2478 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2479 
2480 			/*
2481 			 * we round up to the block size at eof when
2482 			 * determining which extents to clone above,
2483 			 * but shouldn't round up the file size
2484 			 */
2485 			endoff = new_key.offset + datal;
2486 			if (endoff > destoff+olen)
2487 				endoff = destoff+olen;
2488 			if (endoff > inode->i_size)
2489 				btrfs_i_size_write(inode, endoff);
2490 
2491 			ret = btrfs_update_inode(trans, root, inode);
2492 			BUG_ON(ret);
2493 			btrfs_end_transaction(trans, root);
2494 		}
2495 next:
2496 		btrfs_release_path(path);
2497 		key.offset++;
2498 	}
2499 	ret = 0;
2500 out:
2501 	btrfs_release_path(path);
2502 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2503 out_unlock:
2504 	mutex_unlock(&src->i_mutex);
2505 	mutex_unlock(&inode->i_mutex);
2506 	vfree(buf);
2507 	btrfs_free_path(path);
2508 out_fput:
2509 	fput(src_file);
2510 out_drop_write:
2511 	mnt_drop_write(file->f_path.mnt);
2512 	return ret;
2513 }
2514 
2515 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2516 {
2517 	struct btrfs_ioctl_clone_range_args args;
2518 
2519 	if (copy_from_user(&args, argp, sizeof(args)))
2520 		return -EFAULT;
2521 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2522 				 args.src_length, args.dest_offset);
2523 }
2524 
2525 /*
2526  * there are many ways the trans_start and trans_end ioctls can lead
2527  * to deadlocks.  They should only be used by applications that
2528  * basically own the machine, and have a very in depth understanding
2529  * of all the possible deadlocks and enospc problems.
2530  */
2531 static long btrfs_ioctl_trans_start(struct file *file)
2532 {
2533 	struct inode *inode = fdentry(file)->d_inode;
2534 	struct btrfs_root *root = BTRFS_I(inode)->root;
2535 	struct btrfs_trans_handle *trans;
2536 	int ret;
2537 
2538 	ret = -EPERM;
2539 	if (!capable(CAP_SYS_ADMIN))
2540 		goto out;
2541 
2542 	ret = -EINPROGRESS;
2543 	if (file->private_data)
2544 		goto out;
2545 
2546 	ret = -EROFS;
2547 	if (btrfs_root_readonly(root))
2548 		goto out;
2549 
2550 	ret = mnt_want_write(file->f_path.mnt);
2551 	if (ret)
2552 		goto out;
2553 
2554 	atomic_inc(&root->fs_info->open_ioctl_trans);
2555 
2556 	ret = -ENOMEM;
2557 	trans = btrfs_start_ioctl_transaction(root);
2558 	if (IS_ERR(trans))
2559 		goto out_drop;
2560 
2561 	file->private_data = trans;
2562 	return 0;
2563 
2564 out_drop:
2565 	atomic_dec(&root->fs_info->open_ioctl_trans);
2566 	mnt_drop_write(file->f_path.mnt);
2567 out:
2568 	return ret;
2569 }
2570 
2571 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2572 {
2573 	struct inode *inode = fdentry(file)->d_inode;
2574 	struct btrfs_root *root = BTRFS_I(inode)->root;
2575 	struct btrfs_root *new_root;
2576 	struct btrfs_dir_item *di;
2577 	struct btrfs_trans_handle *trans;
2578 	struct btrfs_path *path;
2579 	struct btrfs_key location;
2580 	struct btrfs_disk_key disk_key;
2581 	struct btrfs_super_block *disk_super;
2582 	u64 features;
2583 	u64 objectid = 0;
2584 	u64 dir_id;
2585 
2586 	if (!capable(CAP_SYS_ADMIN))
2587 		return -EPERM;
2588 
2589 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2590 		return -EFAULT;
2591 
2592 	if (!objectid)
2593 		objectid = root->root_key.objectid;
2594 
2595 	location.objectid = objectid;
2596 	location.type = BTRFS_ROOT_ITEM_KEY;
2597 	location.offset = (u64)-1;
2598 
2599 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2600 	if (IS_ERR(new_root))
2601 		return PTR_ERR(new_root);
2602 
2603 	if (btrfs_root_refs(&new_root->root_item) == 0)
2604 		return -ENOENT;
2605 
2606 	path = btrfs_alloc_path();
2607 	if (!path)
2608 		return -ENOMEM;
2609 	path->leave_spinning = 1;
2610 
2611 	trans = btrfs_start_transaction(root, 1);
2612 	if (IS_ERR(trans)) {
2613 		btrfs_free_path(path);
2614 		return PTR_ERR(trans);
2615 	}
2616 
2617 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2618 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2619 				   dir_id, "default", 7, 1);
2620 	if (IS_ERR_OR_NULL(di)) {
2621 		btrfs_free_path(path);
2622 		btrfs_end_transaction(trans, root);
2623 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2624 		       "this isn't going to work\n");
2625 		return -ENOENT;
2626 	}
2627 
2628 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2629 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2630 	btrfs_mark_buffer_dirty(path->nodes[0]);
2631 	btrfs_free_path(path);
2632 
2633 	disk_super = root->fs_info->super_copy;
2634 	features = btrfs_super_incompat_flags(disk_super);
2635 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2636 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2637 		btrfs_set_super_incompat_flags(disk_super, features);
2638 	}
2639 	btrfs_end_transaction(trans, root);
2640 
2641 	return 0;
2642 }
2643 
2644 static void get_block_group_info(struct list_head *groups_list,
2645 				 struct btrfs_ioctl_space_info *space)
2646 {
2647 	struct btrfs_block_group_cache *block_group;
2648 
2649 	space->total_bytes = 0;
2650 	space->used_bytes = 0;
2651 	space->flags = 0;
2652 	list_for_each_entry(block_group, groups_list, list) {
2653 		space->flags = block_group->flags;
2654 		space->total_bytes += block_group->key.offset;
2655 		space->used_bytes +=
2656 			btrfs_block_group_used(&block_group->item);
2657 	}
2658 }
2659 
2660 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2661 {
2662 	struct btrfs_ioctl_space_args space_args;
2663 	struct btrfs_ioctl_space_info space;
2664 	struct btrfs_ioctl_space_info *dest;
2665 	struct btrfs_ioctl_space_info *dest_orig;
2666 	struct btrfs_ioctl_space_info __user *user_dest;
2667 	struct btrfs_space_info *info;
2668 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2669 		       BTRFS_BLOCK_GROUP_SYSTEM,
2670 		       BTRFS_BLOCK_GROUP_METADATA,
2671 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2672 	int num_types = 4;
2673 	int alloc_size;
2674 	int ret = 0;
2675 	u64 slot_count = 0;
2676 	int i, c;
2677 
2678 	if (copy_from_user(&space_args,
2679 			   (struct btrfs_ioctl_space_args __user *)arg,
2680 			   sizeof(space_args)))
2681 		return -EFAULT;
2682 
2683 	for (i = 0; i < num_types; i++) {
2684 		struct btrfs_space_info *tmp;
2685 
2686 		info = NULL;
2687 		rcu_read_lock();
2688 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2689 					list) {
2690 			if (tmp->flags == types[i]) {
2691 				info = tmp;
2692 				break;
2693 			}
2694 		}
2695 		rcu_read_unlock();
2696 
2697 		if (!info)
2698 			continue;
2699 
2700 		down_read(&info->groups_sem);
2701 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2702 			if (!list_empty(&info->block_groups[c]))
2703 				slot_count++;
2704 		}
2705 		up_read(&info->groups_sem);
2706 	}
2707 
2708 	/* space_slots == 0 means they are asking for a count */
2709 	if (space_args.space_slots == 0) {
2710 		space_args.total_spaces = slot_count;
2711 		goto out;
2712 	}
2713 
2714 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2715 
2716 	alloc_size = sizeof(*dest) * slot_count;
2717 
2718 	/* we generally have at most 6 or so space infos, one for each raid
2719 	 * level.  So, a whole page should be more than enough for everyone
2720 	 */
2721 	if (alloc_size > PAGE_CACHE_SIZE)
2722 		return -ENOMEM;
2723 
2724 	space_args.total_spaces = 0;
2725 	dest = kmalloc(alloc_size, GFP_NOFS);
2726 	if (!dest)
2727 		return -ENOMEM;
2728 	dest_orig = dest;
2729 
2730 	/* now we have a buffer to copy into */
2731 	for (i = 0; i < num_types; i++) {
2732 		struct btrfs_space_info *tmp;
2733 
2734 		if (!slot_count)
2735 			break;
2736 
2737 		info = NULL;
2738 		rcu_read_lock();
2739 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2740 					list) {
2741 			if (tmp->flags == types[i]) {
2742 				info = tmp;
2743 				break;
2744 			}
2745 		}
2746 		rcu_read_unlock();
2747 
2748 		if (!info)
2749 			continue;
2750 		down_read(&info->groups_sem);
2751 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2752 			if (!list_empty(&info->block_groups[c])) {
2753 				get_block_group_info(&info->block_groups[c],
2754 						     &space);
2755 				memcpy(dest, &space, sizeof(space));
2756 				dest++;
2757 				space_args.total_spaces++;
2758 				slot_count--;
2759 			}
2760 			if (!slot_count)
2761 				break;
2762 		}
2763 		up_read(&info->groups_sem);
2764 	}
2765 
2766 	user_dest = (struct btrfs_ioctl_space_info *)
2767 		(arg + sizeof(struct btrfs_ioctl_space_args));
2768 
2769 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2770 		ret = -EFAULT;
2771 
2772 	kfree(dest_orig);
2773 out:
2774 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2775 		ret = -EFAULT;
2776 
2777 	return ret;
2778 }
2779 
2780 /*
2781  * there are many ways the trans_start and trans_end ioctls can lead
2782  * to deadlocks.  They should only be used by applications that
2783  * basically own the machine, and have a very in depth understanding
2784  * of all the possible deadlocks and enospc problems.
2785  */
2786 long btrfs_ioctl_trans_end(struct file *file)
2787 {
2788 	struct inode *inode = fdentry(file)->d_inode;
2789 	struct btrfs_root *root = BTRFS_I(inode)->root;
2790 	struct btrfs_trans_handle *trans;
2791 
2792 	trans = file->private_data;
2793 	if (!trans)
2794 		return -EINVAL;
2795 	file->private_data = NULL;
2796 
2797 	btrfs_end_transaction(trans, root);
2798 
2799 	atomic_dec(&root->fs_info->open_ioctl_trans);
2800 
2801 	mnt_drop_write(file->f_path.mnt);
2802 	return 0;
2803 }
2804 
2805 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2806 {
2807 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2808 	struct btrfs_trans_handle *trans;
2809 	u64 transid;
2810 	int ret;
2811 
2812 	trans = btrfs_start_transaction(root, 0);
2813 	if (IS_ERR(trans))
2814 		return PTR_ERR(trans);
2815 	transid = trans->transid;
2816 	ret = btrfs_commit_transaction_async(trans, root, 0);
2817 	if (ret) {
2818 		btrfs_end_transaction(trans, root);
2819 		return ret;
2820 	}
2821 
2822 	if (argp)
2823 		if (copy_to_user(argp, &transid, sizeof(transid)))
2824 			return -EFAULT;
2825 	return 0;
2826 }
2827 
2828 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2829 {
2830 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2831 	u64 transid;
2832 
2833 	if (argp) {
2834 		if (copy_from_user(&transid, argp, sizeof(transid)))
2835 			return -EFAULT;
2836 	} else {
2837 		transid = 0;  /* current trans */
2838 	}
2839 	return btrfs_wait_for_commit(root, transid);
2840 }
2841 
2842 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2843 {
2844 	int ret;
2845 	struct btrfs_ioctl_scrub_args *sa;
2846 
2847 	if (!capable(CAP_SYS_ADMIN))
2848 		return -EPERM;
2849 
2850 	sa = memdup_user(arg, sizeof(*sa));
2851 	if (IS_ERR(sa))
2852 		return PTR_ERR(sa);
2853 
2854 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2855 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2856 
2857 	if (copy_to_user(arg, sa, sizeof(*sa)))
2858 		ret = -EFAULT;
2859 
2860 	kfree(sa);
2861 	return ret;
2862 }
2863 
2864 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2865 {
2866 	if (!capable(CAP_SYS_ADMIN))
2867 		return -EPERM;
2868 
2869 	return btrfs_scrub_cancel(root);
2870 }
2871 
2872 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2873 				       void __user *arg)
2874 {
2875 	struct btrfs_ioctl_scrub_args *sa;
2876 	int ret;
2877 
2878 	if (!capable(CAP_SYS_ADMIN))
2879 		return -EPERM;
2880 
2881 	sa = memdup_user(arg, sizeof(*sa));
2882 	if (IS_ERR(sa))
2883 		return PTR_ERR(sa);
2884 
2885 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2886 
2887 	if (copy_to_user(arg, sa, sizeof(*sa)))
2888 		ret = -EFAULT;
2889 
2890 	kfree(sa);
2891 	return ret;
2892 }
2893 
2894 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
2895 {
2896 	int ret = 0;
2897 	int i;
2898 	u64 rel_ptr;
2899 	int size;
2900 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
2901 	struct inode_fs_paths *ipath = NULL;
2902 	struct btrfs_path *path;
2903 
2904 	if (!capable(CAP_SYS_ADMIN))
2905 		return -EPERM;
2906 
2907 	path = btrfs_alloc_path();
2908 	if (!path) {
2909 		ret = -ENOMEM;
2910 		goto out;
2911 	}
2912 
2913 	ipa = memdup_user(arg, sizeof(*ipa));
2914 	if (IS_ERR(ipa)) {
2915 		ret = PTR_ERR(ipa);
2916 		ipa = NULL;
2917 		goto out;
2918 	}
2919 
2920 	size = min_t(u32, ipa->size, 4096);
2921 	ipath = init_ipath(size, root, path);
2922 	if (IS_ERR(ipath)) {
2923 		ret = PTR_ERR(ipath);
2924 		ipath = NULL;
2925 		goto out;
2926 	}
2927 
2928 	ret = paths_from_inode(ipa->inum, ipath);
2929 	if (ret < 0)
2930 		goto out;
2931 
2932 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
2933 		rel_ptr = ipath->fspath->val[i] - (u64)ipath->fspath->val;
2934 		ipath->fspath->val[i] = rel_ptr;
2935 	}
2936 
2937 	ret = copy_to_user((void *)ipa->fspath, (void *)ipath->fspath, size);
2938 	if (ret) {
2939 		ret = -EFAULT;
2940 		goto out;
2941 	}
2942 
2943 out:
2944 	btrfs_free_path(path);
2945 	free_ipath(ipath);
2946 	kfree(ipa);
2947 
2948 	return ret;
2949 }
2950 
2951 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
2952 {
2953 	struct btrfs_data_container *inodes = ctx;
2954 	const size_t c = 3 * sizeof(u64);
2955 
2956 	if (inodes->bytes_left >= c) {
2957 		inodes->bytes_left -= c;
2958 		inodes->val[inodes->elem_cnt] = inum;
2959 		inodes->val[inodes->elem_cnt + 1] = offset;
2960 		inodes->val[inodes->elem_cnt + 2] = root;
2961 		inodes->elem_cnt += 3;
2962 	} else {
2963 		inodes->bytes_missing += c - inodes->bytes_left;
2964 		inodes->bytes_left = 0;
2965 		inodes->elem_missed += 3;
2966 	}
2967 
2968 	return 0;
2969 }
2970 
2971 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
2972 					void __user *arg)
2973 {
2974 	int ret = 0;
2975 	int size;
2976 	u64 extent_offset;
2977 	struct btrfs_ioctl_logical_ino_args *loi;
2978 	struct btrfs_data_container *inodes = NULL;
2979 	struct btrfs_path *path = NULL;
2980 	struct btrfs_key key;
2981 
2982 	if (!capable(CAP_SYS_ADMIN))
2983 		return -EPERM;
2984 
2985 	loi = memdup_user(arg, sizeof(*loi));
2986 	if (IS_ERR(loi)) {
2987 		ret = PTR_ERR(loi);
2988 		loi = NULL;
2989 		goto out;
2990 	}
2991 
2992 	path = btrfs_alloc_path();
2993 	if (!path) {
2994 		ret = -ENOMEM;
2995 		goto out;
2996 	}
2997 
2998 	size = min_t(u32, loi->size, 4096);
2999 	inodes = init_data_container(size);
3000 	if (IS_ERR(inodes)) {
3001 		ret = PTR_ERR(inodes);
3002 		inodes = NULL;
3003 		goto out;
3004 	}
3005 
3006 	ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3007 
3008 	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3009 		ret = -ENOENT;
3010 	if (ret < 0)
3011 		goto out;
3012 
3013 	extent_offset = loi->logical - key.objectid;
3014 	ret = iterate_extent_inodes(root->fs_info, path, key.objectid,
3015 					extent_offset, build_ino_list, inodes);
3016 
3017 	if (ret < 0)
3018 		goto out;
3019 
3020 	ret = copy_to_user((void *)loi->inodes, (void *)inodes, size);
3021 	if (ret)
3022 		ret = -EFAULT;
3023 
3024 out:
3025 	btrfs_free_path(path);
3026 	kfree(inodes);
3027 	kfree(loi);
3028 
3029 	return ret;
3030 }
3031 
3032 long btrfs_ioctl(struct file *file, unsigned int
3033 		cmd, unsigned long arg)
3034 {
3035 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3036 	void __user *argp = (void __user *)arg;
3037 
3038 	switch (cmd) {
3039 	case FS_IOC_GETFLAGS:
3040 		return btrfs_ioctl_getflags(file, argp);
3041 	case FS_IOC_SETFLAGS:
3042 		return btrfs_ioctl_setflags(file, argp);
3043 	case FS_IOC_GETVERSION:
3044 		return btrfs_ioctl_getversion(file, argp);
3045 	case FITRIM:
3046 		return btrfs_ioctl_fitrim(file, argp);
3047 	case BTRFS_IOC_SNAP_CREATE:
3048 		return btrfs_ioctl_snap_create(file, argp, 0);
3049 	case BTRFS_IOC_SNAP_CREATE_V2:
3050 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3051 	case BTRFS_IOC_SUBVOL_CREATE:
3052 		return btrfs_ioctl_snap_create(file, argp, 1);
3053 	case BTRFS_IOC_SNAP_DESTROY:
3054 		return btrfs_ioctl_snap_destroy(file, argp);
3055 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3056 		return btrfs_ioctl_subvol_getflags(file, argp);
3057 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3058 		return btrfs_ioctl_subvol_setflags(file, argp);
3059 	case BTRFS_IOC_DEFAULT_SUBVOL:
3060 		return btrfs_ioctl_default_subvol(file, argp);
3061 	case BTRFS_IOC_DEFRAG:
3062 		return btrfs_ioctl_defrag(file, NULL);
3063 	case BTRFS_IOC_DEFRAG_RANGE:
3064 		return btrfs_ioctl_defrag(file, argp);
3065 	case BTRFS_IOC_RESIZE:
3066 		return btrfs_ioctl_resize(root, argp);
3067 	case BTRFS_IOC_ADD_DEV:
3068 		return btrfs_ioctl_add_dev(root, argp);
3069 	case BTRFS_IOC_RM_DEV:
3070 		return btrfs_ioctl_rm_dev(root, argp);
3071 	case BTRFS_IOC_FS_INFO:
3072 		return btrfs_ioctl_fs_info(root, argp);
3073 	case BTRFS_IOC_DEV_INFO:
3074 		return btrfs_ioctl_dev_info(root, argp);
3075 	case BTRFS_IOC_BALANCE:
3076 		return btrfs_balance(root->fs_info->dev_root);
3077 	case BTRFS_IOC_CLONE:
3078 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3079 	case BTRFS_IOC_CLONE_RANGE:
3080 		return btrfs_ioctl_clone_range(file, argp);
3081 	case BTRFS_IOC_TRANS_START:
3082 		return btrfs_ioctl_trans_start(file);
3083 	case BTRFS_IOC_TRANS_END:
3084 		return btrfs_ioctl_trans_end(file);
3085 	case BTRFS_IOC_TREE_SEARCH:
3086 		return btrfs_ioctl_tree_search(file, argp);
3087 	case BTRFS_IOC_INO_LOOKUP:
3088 		return btrfs_ioctl_ino_lookup(file, argp);
3089 	case BTRFS_IOC_INO_PATHS:
3090 		return btrfs_ioctl_ino_to_path(root, argp);
3091 	case BTRFS_IOC_LOGICAL_INO:
3092 		return btrfs_ioctl_logical_to_ino(root, argp);
3093 	case BTRFS_IOC_SPACE_INFO:
3094 		return btrfs_ioctl_space_info(root, argp);
3095 	case BTRFS_IOC_SYNC:
3096 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3097 		return 0;
3098 	case BTRFS_IOC_START_SYNC:
3099 		return btrfs_ioctl_start_sync(file, argp);
3100 	case BTRFS_IOC_WAIT_SYNC:
3101 		return btrfs_ioctl_wait_sync(file, argp);
3102 	case BTRFS_IOC_SCRUB:
3103 		return btrfs_ioctl_scrub(root, argp);
3104 	case BTRFS_IOC_SCRUB_CANCEL:
3105 		return btrfs_ioctl_scrub_cancel(root, argp);
3106 	case BTRFS_IOC_SCRUB_PROGRESS:
3107 		return btrfs_ioctl_scrub_progress(root, argp);
3108 	}
3109 
3110 	return -ENOTTY;
3111 }
3112