xref: /linux/fs/btrfs/ioctl.c (revision 26b0d14106954ae46d2f4f7eec3481828a210f7d)
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 #include "rcu-string.h"
56 
57 /* Mask out flags that are inappropriate for the given type of inode. */
58 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
59 {
60 	if (S_ISDIR(mode))
61 		return flags;
62 	else if (S_ISREG(mode))
63 		return flags & ~FS_DIRSYNC_FL;
64 	else
65 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
66 }
67 
68 /*
69  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
70  */
71 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
72 {
73 	unsigned int iflags = 0;
74 
75 	if (flags & BTRFS_INODE_SYNC)
76 		iflags |= FS_SYNC_FL;
77 	if (flags & BTRFS_INODE_IMMUTABLE)
78 		iflags |= FS_IMMUTABLE_FL;
79 	if (flags & BTRFS_INODE_APPEND)
80 		iflags |= FS_APPEND_FL;
81 	if (flags & BTRFS_INODE_NODUMP)
82 		iflags |= FS_NODUMP_FL;
83 	if (flags & BTRFS_INODE_NOATIME)
84 		iflags |= FS_NOATIME_FL;
85 	if (flags & BTRFS_INODE_DIRSYNC)
86 		iflags |= FS_DIRSYNC_FL;
87 	if (flags & BTRFS_INODE_NODATACOW)
88 		iflags |= FS_NOCOW_FL;
89 
90 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
91 		iflags |= FS_COMPR_FL;
92 	else if (flags & BTRFS_INODE_NOCOMPRESS)
93 		iflags |= FS_NOCOMP_FL;
94 
95 	return iflags;
96 }
97 
98 /*
99  * Update inode->i_flags based on the btrfs internal flags.
100  */
101 void btrfs_update_iflags(struct inode *inode)
102 {
103 	struct btrfs_inode *ip = BTRFS_I(inode);
104 
105 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
106 
107 	if (ip->flags & BTRFS_INODE_SYNC)
108 		inode->i_flags |= S_SYNC;
109 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
110 		inode->i_flags |= S_IMMUTABLE;
111 	if (ip->flags & BTRFS_INODE_APPEND)
112 		inode->i_flags |= S_APPEND;
113 	if (ip->flags & BTRFS_INODE_NOATIME)
114 		inode->i_flags |= S_NOATIME;
115 	if (ip->flags & BTRFS_INODE_DIRSYNC)
116 		inode->i_flags |= S_DIRSYNC;
117 }
118 
119 /*
120  * Inherit flags from the parent inode.
121  *
122  * Currently only the compression flags and the cow flags are inherited.
123  */
124 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
125 {
126 	unsigned int flags;
127 
128 	if (!dir)
129 		return;
130 
131 	flags = BTRFS_I(dir)->flags;
132 
133 	if (flags & BTRFS_INODE_NOCOMPRESS) {
134 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
135 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
136 	} else if (flags & BTRFS_INODE_COMPRESS) {
137 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
138 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
139 	}
140 
141 	if (flags & BTRFS_INODE_NODATACOW)
142 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
143 
144 	btrfs_update_iflags(inode);
145 }
146 
147 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
148 {
149 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
150 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
151 
152 	if (copy_to_user(arg, &flags, sizeof(flags)))
153 		return -EFAULT;
154 	return 0;
155 }
156 
157 static int check_flags(unsigned int flags)
158 {
159 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
160 		      FS_NOATIME_FL | FS_NODUMP_FL | \
161 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
162 		      FS_NOCOMP_FL | FS_COMPR_FL |
163 		      FS_NOCOW_FL))
164 		return -EOPNOTSUPP;
165 
166 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
167 		return -EINVAL;
168 
169 	return 0;
170 }
171 
172 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
173 {
174 	struct inode *inode = file->f_path.dentry->d_inode;
175 	struct btrfs_inode *ip = BTRFS_I(inode);
176 	struct btrfs_root *root = ip->root;
177 	struct btrfs_trans_handle *trans;
178 	unsigned int flags, oldflags;
179 	int ret;
180 	u64 ip_oldflags;
181 	unsigned int i_oldflags;
182 
183 	if (btrfs_root_readonly(root))
184 		return -EROFS;
185 
186 	if (copy_from_user(&flags, arg, sizeof(flags)))
187 		return -EFAULT;
188 
189 	ret = check_flags(flags);
190 	if (ret)
191 		return ret;
192 
193 	if (!inode_owner_or_capable(inode))
194 		return -EACCES;
195 
196 	mutex_lock(&inode->i_mutex);
197 
198 	ip_oldflags = ip->flags;
199 	i_oldflags = inode->i_flags;
200 
201 	flags = btrfs_mask_flags(inode->i_mode, flags);
202 	oldflags = btrfs_flags_to_ioctl(ip->flags);
203 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
204 		if (!capable(CAP_LINUX_IMMUTABLE)) {
205 			ret = -EPERM;
206 			goto out_unlock;
207 		}
208 	}
209 
210 	ret = mnt_want_write_file(file);
211 	if (ret)
212 		goto out_unlock;
213 
214 	if (flags & FS_SYNC_FL)
215 		ip->flags |= BTRFS_INODE_SYNC;
216 	else
217 		ip->flags &= ~BTRFS_INODE_SYNC;
218 	if (flags & FS_IMMUTABLE_FL)
219 		ip->flags |= BTRFS_INODE_IMMUTABLE;
220 	else
221 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
222 	if (flags & FS_APPEND_FL)
223 		ip->flags |= BTRFS_INODE_APPEND;
224 	else
225 		ip->flags &= ~BTRFS_INODE_APPEND;
226 	if (flags & FS_NODUMP_FL)
227 		ip->flags |= BTRFS_INODE_NODUMP;
228 	else
229 		ip->flags &= ~BTRFS_INODE_NODUMP;
230 	if (flags & FS_NOATIME_FL)
231 		ip->flags |= BTRFS_INODE_NOATIME;
232 	else
233 		ip->flags &= ~BTRFS_INODE_NOATIME;
234 	if (flags & FS_DIRSYNC_FL)
235 		ip->flags |= BTRFS_INODE_DIRSYNC;
236 	else
237 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
238 	if (flags & FS_NOCOW_FL)
239 		ip->flags |= BTRFS_INODE_NODATACOW;
240 	else
241 		ip->flags &= ~BTRFS_INODE_NODATACOW;
242 
243 	/*
244 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
245 	 * flag may be changed automatically if compression code won't make
246 	 * things smaller.
247 	 */
248 	if (flags & FS_NOCOMP_FL) {
249 		ip->flags &= ~BTRFS_INODE_COMPRESS;
250 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
251 	} else if (flags & FS_COMPR_FL) {
252 		ip->flags |= BTRFS_INODE_COMPRESS;
253 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
254 	} else {
255 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
256 	}
257 
258 	trans = btrfs_start_transaction(root, 1);
259 	if (IS_ERR(trans)) {
260 		ret = PTR_ERR(trans);
261 		goto out_drop;
262 	}
263 
264 	btrfs_update_iflags(inode);
265 	inode_inc_iversion(inode);
266 	inode->i_ctime = CURRENT_TIME;
267 	ret = btrfs_update_inode(trans, root, inode);
268 
269 	btrfs_end_transaction(trans, root);
270  out_drop:
271 	if (ret) {
272 		ip->flags = ip_oldflags;
273 		inode->i_flags = i_oldflags;
274 	}
275 
276 	mnt_drop_write_file(file);
277  out_unlock:
278 	mutex_unlock(&inode->i_mutex);
279 	return ret;
280 }
281 
282 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
283 {
284 	struct inode *inode = file->f_path.dentry->d_inode;
285 
286 	return put_user(inode->i_generation, arg);
287 }
288 
289 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
290 {
291 	struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
292 	struct btrfs_device *device;
293 	struct request_queue *q;
294 	struct fstrim_range range;
295 	u64 minlen = ULLONG_MAX;
296 	u64 num_devices = 0;
297 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
298 	int ret;
299 
300 	if (!capable(CAP_SYS_ADMIN))
301 		return -EPERM;
302 
303 	rcu_read_lock();
304 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
305 				dev_list) {
306 		if (!device->bdev)
307 			continue;
308 		q = bdev_get_queue(device->bdev);
309 		if (blk_queue_discard(q)) {
310 			num_devices++;
311 			minlen = min((u64)q->limits.discard_granularity,
312 				     minlen);
313 		}
314 	}
315 	rcu_read_unlock();
316 
317 	if (!num_devices)
318 		return -EOPNOTSUPP;
319 	if (copy_from_user(&range, arg, sizeof(range)))
320 		return -EFAULT;
321 	if (range.start > total_bytes)
322 		return -EINVAL;
323 
324 	range.len = min(range.len, total_bytes - range.start);
325 	range.minlen = max(range.minlen, minlen);
326 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
327 	if (ret < 0)
328 		return ret;
329 
330 	if (copy_to_user(arg, &range, sizeof(range)))
331 		return -EFAULT;
332 
333 	return 0;
334 }
335 
336 static noinline int create_subvol(struct btrfs_root *root,
337 				  struct dentry *dentry,
338 				  char *name, int namelen,
339 				  u64 *async_transid)
340 {
341 	struct btrfs_trans_handle *trans;
342 	struct btrfs_key key;
343 	struct btrfs_root_item root_item;
344 	struct btrfs_inode_item *inode_item;
345 	struct extent_buffer *leaf;
346 	struct btrfs_root *new_root;
347 	struct dentry *parent = dentry->d_parent;
348 	struct inode *dir;
349 	int ret;
350 	int err;
351 	u64 objectid;
352 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
353 	u64 index = 0;
354 
355 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
356 	if (ret)
357 		return ret;
358 
359 	dir = parent->d_inode;
360 
361 	/*
362 	 * 1 - inode item
363 	 * 2 - refs
364 	 * 1 - root item
365 	 * 2 - dir items
366 	 */
367 	trans = btrfs_start_transaction(root, 6);
368 	if (IS_ERR(trans))
369 		return PTR_ERR(trans);
370 
371 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
372 				      0, objectid, NULL, 0, 0, 0);
373 	if (IS_ERR(leaf)) {
374 		ret = PTR_ERR(leaf);
375 		goto fail;
376 	}
377 
378 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
379 	btrfs_set_header_bytenr(leaf, leaf->start);
380 	btrfs_set_header_generation(leaf, trans->transid);
381 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
382 	btrfs_set_header_owner(leaf, objectid);
383 
384 	write_extent_buffer(leaf, root->fs_info->fsid,
385 			    (unsigned long)btrfs_header_fsid(leaf),
386 			    BTRFS_FSID_SIZE);
387 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
388 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
389 			    BTRFS_UUID_SIZE);
390 	btrfs_mark_buffer_dirty(leaf);
391 
392 	inode_item = &root_item.inode;
393 	memset(inode_item, 0, sizeof(*inode_item));
394 	inode_item->generation = cpu_to_le64(1);
395 	inode_item->size = cpu_to_le64(3);
396 	inode_item->nlink = cpu_to_le32(1);
397 	inode_item->nbytes = cpu_to_le64(root->leafsize);
398 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
399 
400 	root_item.flags = 0;
401 	root_item.byte_limit = 0;
402 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
403 
404 	btrfs_set_root_bytenr(&root_item, leaf->start);
405 	btrfs_set_root_generation(&root_item, trans->transid);
406 	btrfs_set_root_level(&root_item, 0);
407 	btrfs_set_root_refs(&root_item, 1);
408 	btrfs_set_root_used(&root_item, leaf->len);
409 	btrfs_set_root_last_snapshot(&root_item, 0);
410 
411 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
412 	root_item.drop_level = 0;
413 
414 	btrfs_tree_unlock(leaf);
415 	free_extent_buffer(leaf);
416 	leaf = NULL;
417 
418 	btrfs_set_root_dirid(&root_item, new_dirid);
419 
420 	key.objectid = objectid;
421 	key.offset = 0;
422 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
423 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
424 				&root_item);
425 	if (ret)
426 		goto fail;
427 
428 	key.offset = (u64)-1;
429 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
430 	if (IS_ERR(new_root)) {
431 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
432 		ret = PTR_ERR(new_root);
433 		goto fail;
434 	}
435 
436 	btrfs_record_root_in_trans(trans, new_root);
437 
438 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
439 	if (ret) {
440 		/* We potentially lose an unused inode item here */
441 		btrfs_abort_transaction(trans, root, ret);
442 		goto fail;
443 	}
444 
445 	/*
446 	 * insert the directory item
447 	 */
448 	ret = btrfs_set_inode_index(dir, &index);
449 	if (ret) {
450 		btrfs_abort_transaction(trans, root, ret);
451 		goto fail;
452 	}
453 
454 	ret = btrfs_insert_dir_item(trans, root,
455 				    name, namelen, dir, &key,
456 				    BTRFS_FT_DIR, index);
457 	if (ret) {
458 		btrfs_abort_transaction(trans, root, ret);
459 		goto fail;
460 	}
461 
462 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
463 	ret = btrfs_update_inode(trans, root, dir);
464 	BUG_ON(ret);
465 
466 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
467 				 objectid, root->root_key.objectid,
468 				 btrfs_ino(dir), index, name, namelen);
469 
470 	BUG_ON(ret);
471 
472 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
473 fail:
474 	if (async_transid) {
475 		*async_transid = trans->transid;
476 		err = btrfs_commit_transaction_async(trans, root, 1);
477 	} else {
478 		err = btrfs_commit_transaction(trans, root);
479 	}
480 	if (err && !ret)
481 		ret = err;
482 	return ret;
483 }
484 
485 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
486 			   char *name, int namelen, u64 *async_transid,
487 			   bool readonly)
488 {
489 	struct inode *inode;
490 	struct btrfs_pending_snapshot *pending_snapshot;
491 	struct btrfs_trans_handle *trans;
492 	int ret;
493 
494 	if (!root->ref_cows)
495 		return -EINVAL;
496 
497 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
498 	if (!pending_snapshot)
499 		return -ENOMEM;
500 
501 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
502 	pending_snapshot->dentry = dentry;
503 	pending_snapshot->root = root;
504 	pending_snapshot->readonly = readonly;
505 
506 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
507 	if (IS_ERR(trans)) {
508 		ret = PTR_ERR(trans);
509 		goto fail;
510 	}
511 
512 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
513 	BUG_ON(ret);
514 
515 	spin_lock(&root->fs_info->trans_lock);
516 	list_add(&pending_snapshot->list,
517 		 &trans->transaction->pending_snapshots);
518 	spin_unlock(&root->fs_info->trans_lock);
519 	if (async_transid) {
520 		*async_transid = trans->transid;
521 		ret = btrfs_commit_transaction_async(trans,
522 				     root->fs_info->extent_root, 1);
523 	} else {
524 		ret = btrfs_commit_transaction(trans,
525 					       root->fs_info->extent_root);
526 	}
527 	BUG_ON(ret);
528 
529 	ret = pending_snapshot->error;
530 	if (ret)
531 		goto fail;
532 
533 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
534 	if (ret)
535 		goto fail;
536 
537 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
538 	if (IS_ERR(inode)) {
539 		ret = PTR_ERR(inode);
540 		goto fail;
541 	}
542 	BUG_ON(!inode);
543 	d_instantiate(dentry, inode);
544 	ret = 0;
545 fail:
546 	kfree(pending_snapshot);
547 	return ret;
548 }
549 
550 /*  copy of check_sticky in fs/namei.c()
551 * It's inline, so penalty for filesystems that don't use sticky bit is
552 * minimal.
553 */
554 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
555 {
556 	uid_t fsuid = current_fsuid();
557 
558 	if (!(dir->i_mode & S_ISVTX))
559 		return 0;
560 	if (inode->i_uid == fsuid)
561 		return 0;
562 	if (dir->i_uid == fsuid)
563 		return 0;
564 	return !capable(CAP_FOWNER);
565 }
566 
567 /*  copy of may_delete in fs/namei.c()
568  *	Check whether we can remove a link victim from directory dir, check
569  *  whether the type of victim is right.
570  *  1. We can't do it if dir is read-only (done in permission())
571  *  2. We should have write and exec permissions on dir
572  *  3. We can't remove anything from append-only dir
573  *  4. We can't do anything with immutable dir (done in permission())
574  *  5. If the sticky bit on dir is set we should either
575  *	a. be owner of dir, or
576  *	b. be owner of victim, or
577  *	c. have CAP_FOWNER capability
578  *  6. If the victim is append-only or immutable we can't do antyhing with
579  *     links pointing to it.
580  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
581  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
582  *  9. We can't remove a root or mountpoint.
583  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
584  *     nfs_async_unlink().
585  */
586 
587 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
588 {
589 	int error;
590 
591 	if (!victim->d_inode)
592 		return -ENOENT;
593 
594 	BUG_ON(victim->d_parent->d_inode != dir);
595 	audit_inode_child(victim, dir);
596 
597 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
598 	if (error)
599 		return error;
600 	if (IS_APPEND(dir))
601 		return -EPERM;
602 	if (btrfs_check_sticky(dir, victim->d_inode)||
603 		IS_APPEND(victim->d_inode)||
604 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
605 		return -EPERM;
606 	if (isdir) {
607 		if (!S_ISDIR(victim->d_inode->i_mode))
608 			return -ENOTDIR;
609 		if (IS_ROOT(victim))
610 			return -EBUSY;
611 	} else if (S_ISDIR(victim->d_inode->i_mode))
612 		return -EISDIR;
613 	if (IS_DEADDIR(dir))
614 		return -ENOENT;
615 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
616 		return -EBUSY;
617 	return 0;
618 }
619 
620 /* copy of may_create in fs/namei.c() */
621 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
622 {
623 	if (child->d_inode)
624 		return -EEXIST;
625 	if (IS_DEADDIR(dir))
626 		return -ENOENT;
627 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
628 }
629 
630 /*
631  * Create a new subvolume below @parent.  This is largely modeled after
632  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
633  * inside this filesystem so it's quite a bit simpler.
634  */
635 static noinline int btrfs_mksubvol(struct path *parent,
636 				   char *name, int namelen,
637 				   struct btrfs_root *snap_src,
638 				   u64 *async_transid, bool readonly)
639 {
640 	struct inode *dir  = parent->dentry->d_inode;
641 	struct dentry *dentry;
642 	int error;
643 
644 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
645 
646 	dentry = lookup_one_len(name, parent->dentry, namelen);
647 	error = PTR_ERR(dentry);
648 	if (IS_ERR(dentry))
649 		goto out_unlock;
650 
651 	error = -EEXIST;
652 	if (dentry->d_inode)
653 		goto out_dput;
654 
655 	error = mnt_want_write(parent->mnt);
656 	if (error)
657 		goto out_dput;
658 
659 	error = btrfs_may_create(dir, dentry);
660 	if (error)
661 		goto out_drop_write;
662 
663 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
664 
665 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
666 		goto out_up_read;
667 
668 	if (snap_src) {
669 		error = create_snapshot(snap_src, dentry,
670 					name, namelen, async_transid, readonly);
671 	} else {
672 		error = create_subvol(BTRFS_I(dir)->root, dentry,
673 				      name, namelen, async_transid);
674 	}
675 	if (!error)
676 		fsnotify_mkdir(dir, dentry);
677 out_up_read:
678 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
679 out_drop_write:
680 	mnt_drop_write(parent->mnt);
681 out_dput:
682 	dput(dentry);
683 out_unlock:
684 	mutex_unlock(&dir->i_mutex);
685 	return error;
686 }
687 
688 /*
689  * When we're defragging a range, we don't want to kick it off again
690  * if it is really just waiting for delalloc to send it down.
691  * If we find a nice big extent or delalloc range for the bytes in the
692  * file you want to defrag, we return 0 to let you know to skip this
693  * part of the file
694  */
695 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
696 {
697 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
698 	struct extent_map *em = NULL;
699 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
700 	u64 end;
701 
702 	read_lock(&em_tree->lock);
703 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
704 	read_unlock(&em_tree->lock);
705 
706 	if (em) {
707 		end = extent_map_end(em);
708 		free_extent_map(em);
709 		if (end - offset > thresh)
710 			return 0;
711 	}
712 	/* if we already have a nice delalloc here, just stop */
713 	thresh /= 2;
714 	end = count_range_bits(io_tree, &offset, offset + thresh,
715 			       thresh, EXTENT_DELALLOC, 1);
716 	if (end >= thresh)
717 		return 0;
718 	return 1;
719 }
720 
721 /*
722  * helper function to walk through a file and find extents
723  * newer than a specific transid, and smaller than thresh.
724  *
725  * This is used by the defragging code to find new and small
726  * extents
727  */
728 static int find_new_extents(struct btrfs_root *root,
729 			    struct inode *inode, u64 newer_than,
730 			    u64 *off, int thresh)
731 {
732 	struct btrfs_path *path;
733 	struct btrfs_key min_key;
734 	struct btrfs_key max_key;
735 	struct extent_buffer *leaf;
736 	struct btrfs_file_extent_item *extent;
737 	int type;
738 	int ret;
739 	u64 ino = btrfs_ino(inode);
740 
741 	path = btrfs_alloc_path();
742 	if (!path)
743 		return -ENOMEM;
744 
745 	min_key.objectid = ino;
746 	min_key.type = BTRFS_EXTENT_DATA_KEY;
747 	min_key.offset = *off;
748 
749 	max_key.objectid = ino;
750 	max_key.type = (u8)-1;
751 	max_key.offset = (u64)-1;
752 
753 	path->keep_locks = 1;
754 
755 	while(1) {
756 		ret = btrfs_search_forward(root, &min_key, &max_key,
757 					   path, 0, newer_than);
758 		if (ret != 0)
759 			goto none;
760 		if (min_key.objectid != ino)
761 			goto none;
762 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
763 			goto none;
764 
765 		leaf = path->nodes[0];
766 		extent = btrfs_item_ptr(leaf, path->slots[0],
767 					struct btrfs_file_extent_item);
768 
769 		type = btrfs_file_extent_type(leaf, extent);
770 		if (type == BTRFS_FILE_EXTENT_REG &&
771 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
772 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
773 			*off = min_key.offset;
774 			btrfs_free_path(path);
775 			return 0;
776 		}
777 
778 		if (min_key.offset == (u64)-1)
779 			goto none;
780 
781 		min_key.offset++;
782 		btrfs_release_path(path);
783 	}
784 none:
785 	btrfs_free_path(path);
786 	return -ENOENT;
787 }
788 
789 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
790 {
791 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
792 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
793 	struct extent_map *em;
794 	u64 len = PAGE_CACHE_SIZE;
795 
796 	/*
797 	 * hopefully we have this extent in the tree already, try without
798 	 * the full extent lock
799 	 */
800 	read_lock(&em_tree->lock);
801 	em = lookup_extent_mapping(em_tree, start, len);
802 	read_unlock(&em_tree->lock);
803 
804 	if (!em) {
805 		/* get the big lock and read metadata off disk */
806 		lock_extent(io_tree, start, start + len - 1);
807 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
808 		unlock_extent(io_tree, start, start + len - 1);
809 
810 		if (IS_ERR(em))
811 			return NULL;
812 	}
813 
814 	return em;
815 }
816 
817 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
818 {
819 	struct extent_map *next;
820 	bool ret = true;
821 
822 	/* this is the last extent */
823 	if (em->start + em->len >= i_size_read(inode))
824 		return false;
825 
826 	next = defrag_lookup_extent(inode, em->start + em->len);
827 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
828 		ret = false;
829 
830 	free_extent_map(next);
831 	return ret;
832 }
833 
834 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
835 			       u64 *last_len, u64 *skip, u64 *defrag_end)
836 {
837 	struct extent_map *em;
838 	int ret = 1;
839 	bool next_mergeable = true;
840 
841 	/*
842 	 * make sure that once we start defragging an extent, we keep on
843 	 * defragging it
844 	 */
845 	if (start < *defrag_end)
846 		return 1;
847 
848 	*skip = 0;
849 
850 	em = defrag_lookup_extent(inode, start);
851 	if (!em)
852 		return 0;
853 
854 	/* this will cover holes, and inline extents */
855 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
856 		ret = 0;
857 		goto out;
858 	}
859 
860 	next_mergeable = defrag_check_next_extent(inode, em);
861 
862 	/*
863 	 * we hit a real extent, if it is big or the next extent is not a
864 	 * real extent, don't bother defragging it
865 	 */
866 	if ((*last_len == 0 || *last_len >= thresh) &&
867 	    (em->len >= thresh || !next_mergeable))
868 		ret = 0;
869 out:
870 	/*
871 	 * last_len ends up being a counter of how many bytes we've defragged.
872 	 * every time we choose not to defrag an extent, we reset *last_len
873 	 * so that the next tiny extent will force a defrag.
874 	 *
875 	 * The end result of this is that tiny extents before a single big
876 	 * extent will force at least part of that big extent to be defragged.
877 	 */
878 	if (ret) {
879 		*defrag_end = extent_map_end(em);
880 	} else {
881 		*last_len = 0;
882 		*skip = extent_map_end(em);
883 		*defrag_end = 0;
884 	}
885 
886 	free_extent_map(em);
887 	return ret;
888 }
889 
890 /*
891  * it doesn't do much good to defrag one or two pages
892  * at a time.  This pulls in a nice chunk of pages
893  * to COW and defrag.
894  *
895  * It also makes sure the delalloc code has enough
896  * dirty data to avoid making new small extents as part
897  * of the defrag
898  *
899  * It's a good idea to start RA on this range
900  * before calling this.
901  */
902 static int cluster_pages_for_defrag(struct inode *inode,
903 				    struct page **pages,
904 				    unsigned long start_index,
905 				    int num_pages)
906 {
907 	unsigned long file_end;
908 	u64 isize = i_size_read(inode);
909 	u64 page_start;
910 	u64 page_end;
911 	u64 page_cnt;
912 	int ret;
913 	int i;
914 	int i_done;
915 	struct btrfs_ordered_extent *ordered;
916 	struct extent_state *cached_state = NULL;
917 	struct extent_io_tree *tree;
918 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
919 
920 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
921 	if (!isize || start_index > file_end)
922 		return 0;
923 
924 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
925 
926 	ret = btrfs_delalloc_reserve_space(inode,
927 					   page_cnt << PAGE_CACHE_SHIFT);
928 	if (ret)
929 		return ret;
930 	i_done = 0;
931 	tree = &BTRFS_I(inode)->io_tree;
932 
933 	/* step one, lock all the pages */
934 	for (i = 0; i < page_cnt; i++) {
935 		struct page *page;
936 again:
937 		page = find_or_create_page(inode->i_mapping,
938 					   start_index + i, mask);
939 		if (!page)
940 			break;
941 
942 		page_start = page_offset(page);
943 		page_end = page_start + PAGE_CACHE_SIZE - 1;
944 		while (1) {
945 			lock_extent(tree, page_start, page_end);
946 			ordered = btrfs_lookup_ordered_extent(inode,
947 							      page_start);
948 			unlock_extent(tree, page_start, page_end);
949 			if (!ordered)
950 				break;
951 
952 			unlock_page(page);
953 			btrfs_start_ordered_extent(inode, ordered, 1);
954 			btrfs_put_ordered_extent(ordered);
955 			lock_page(page);
956 			/*
957 			 * we unlocked the page above, so we need check if
958 			 * it was released or not.
959 			 */
960 			if (page->mapping != inode->i_mapping) {
961 				unlock_page(page);
962 				page_cache_release(page);
963 				goto again;
964 			}
965 		}
966 
967 		if (!PageUptodate(page)) {
968 			btrfs_readpage(NULL, page);
969 			lock_page(page);
970 			if (!PageUptodate(page)) {
971 				unlock_page(page);
972 				page_cache_release(page);
973 				ret = -EIO;
974 				break;
975 			}
976 		}
977 
978 		if (page->mapping != inode->i_mapping) {
979 			unlock_page(page);
980 			page_cache_release(page);
981 			goto again;
982 		}
983 
984 		pages[i] = page;
985 		i_done++;
986 	}
987 	if (!i_done || ret)
988 		goto out;
989 
990 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
991 		goto out;
992 
993 	/*
994 	 * so now we have a nice long stream of locked
995 	 * and up to date pages, lets wait on them
996 	 */
997 	for (i = 0; i < i_done; i++)
998 		wait_on_page_writeback(pages[i]);
999 
1000 	page_start = page_offset(pages[0]);
1001 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1002 
1003 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1004 			 page_start, page_end - 1, 0, &cached_state);
1005 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1006 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1007 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1008 			  GFP_NOFS);
1009 
1010 	if (i_done != page_cnt) {
1011 		spin_lock(&BTRFS_I(inode)->lock);
1012 		BTRFS_I(inode)->outstanding_extents++;
1013 		spin_unlock(&BTRFS_I(inode)->lock);
1014 		btrfs_delalloc_release_space(inode,
1015 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1016 	}
1017 
1018 
1019 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1020 				  &cached_state);
1021 
1022 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1023 			     page_start, page_end - 1, &cached_state,
1024 			     GFP_NOFS);
1025 
1026 	for (i = 0; i < i_done; i++) {
1027 		clear_page_dirty_for_io(pages[i]);
1028 		ClearPageChecked(pages[i]);
1029 		set_page_extent_mapped(pages[i]);
1030 		set_page_dirty(pages[i]);
1031 		unlock_page(pages[i]);
1032 		page_cache_release(pages[i]);
1033 	}
1034 	return i_done;
1035 out:
1036 	for (i = 0; i < i_done; i++) {
1037 		unlock_page(pages[i]);
1038 		page_cache_release(pages[i]);
1039 	}
1040 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1041 	return ret;
1042 
1043 }
1044 
1045 int btrfs_defrag_file(struct inode *inode, struct file *file,
1046 		      struct btrfs_ioctl_defrag_range_args *range,
1047 		      u64 newer_than, unsigned long max_to_defrag)
1048 {
1049 	struct btrfs_root *root = BTRFS_I(inode)->root;
1050 	struct btrfs_super_block *disk_super;
1051 	struct file_ra_state *ra = NULL;
1052 	unsigned long last_index;
1053 	u64 isize = i_size_read(inode);
1054 	u64 features;
1055 	u64 last_len = 0;
1056 	u64 skip = 0;
1057 	u64 defrag_end = 0;
1058 	u64 newer_off = range->start;
1059 	unsigned long i;
1060 	unsigned long ra_index = 0;
1061 	int ret;
1062 	int defrag_count = 0;
1063 	int compress_type = BTRFS_COMPRESS_ZLIB;
1064 	int extent_thresh = range->extent_thresh;
1065 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1066 	int cluster = max_cluster;
1067 	u64 new_align = ~((u64)128 * 1024 - 1);
1068 	struct page **pages = NULL;
1069 
1070 	if (extent_thresh == 0)
1071 		extent_thresh = 256 * 1024;
1072 
1073 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1074 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1075 			return -EINVAL;
1076 		if (range->compress_type)
1077 			compress_type = range->compress_type;
1078 	}
1079 
1080 	if (isize == 0)
1081 		return 0;
1082 
1083 	/*
1084 	 * if we were not given a file, allocate a readahead
1085 	 * context
1086 	 */
1087 	if (!file) {
1088 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1089 		if (!ra)
1090 			return -ENOMEM;
1091 		file_ra_state_init(ra, inode->i_mapping);
1092 	} else {
1093 		ra = &file->f_ra;
1094 	}
1095 
1096 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1097 			GFP_NOFS);
1098 	if (!pages) {
1099 		ret = -ENOMEM;
1100 		goto out_ra;
1101 	}
1102 
1103 	/* find the last page to defrag */
1104 	if (range->start + range->len > range->start) {
1105 		last_index = min_t(u64, isize - 1,
1106 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1107 	} else {
1108 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1109 	}
1110 
1111 	if (newer_than) {
1112 		ret = find_new_extents(root, inode, newer_than,
1113 				       &newer_off, 64 * 1024);
1114 		if (!ret) {
1115 			range->start = newer_off;
1116 			/*
1117 			 * we always align our defrag to help keep
1118 			 * the extents in the file evenly spaced
1119 			 */
1120 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1121 		} else
1122 			goto out_ra;
1123 	} else {
1124 		i = range->start >> PAGE_CACHE_SHIFT;
1125 	}
1126 	if (!max_to_defrag)
1127 		max_to_defrag = last_index + 1;
1128 
1129 	/*
1130 	 * make writeback starts from i, so the defrag range can be
1131 	 * written sequentially.
1132 	 */
1133 	if (i < inode->i_mapping->writeback_index)
1134 		inode->i_mapping->writeback_index = i;
1135 
1136 	while (i <= last_index && defrag_count < max_to_defrag &&
1137 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1138 		PAGE_CACHE_SHIFT)) {
1139 		/*
1140 		 * make sure we stop running if someone unmounts
1141 		 * the FS
1142 		 */
1143 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1144 			break;
1145 
1146 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1147 					 extent_thresh, &last_len, &skip,
1148 					 &defrag_end)) {
1149 			unsigned long next;
1150 			/*
1151 			 * the should_defrag function tells us how much to skip
1152 			 * bump our counter by the suggested amount
1153 			 */
1154 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1155 			i = max(i + 1, next);
1156 			continue;
1157 		}
1158 
1159 		if (!newer_than) {
1160 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1161 				   PAGE_CACHE_SHIFT) - i;
1162 			cluster = min(cluster, max_cluster);
1163 		} else {
1164 			cluster = max_cluster;
1165 		}
1166 
1167 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1168 			BTRFS_I(inode)->force_compress = compress_type;
1169 
1170 		if (i + cluster > ra_index) {
1171 			ra_index = max(i, ra_index);
1172 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1173 				       cluster);
1174 			ra_index += max_cluster;
1175 		}
1176 
1177 		mutex_lock(&inode->i_mutex);
1178 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1179 		if (ret < 0) {
1180 			mutex_unlock(&inode->i_mutex);
1181 			goto out_ra;
1182 		}
1183 
1184 		defrag_count += ret;
1185 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1186 		mutex_unlock(&inode->i_mutex);
1187 
1188 		if (newer_than) {
1189 			if (newer_off == (u64)-1)
1190 				break;
1191 
1192 			if (ret > 0)
1193 				i += ret;
1194 
1195 			newer_off = max(newer_off + 1,
1196 					(u64)i << PAGE_CACHE_SHIFT);
1197 
1198 			ret = find_new_extents(root, inode,
1199 					       newer_than, &newer_off,
1200 					       64 * 1024);
1201 			if (!ret) {
1202 				range->start = newer_off;
1203 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1204 			} else {
1205 				break;
1206 			}
1207 		} else {
1208 			if (ret > 0) {
1209 				i += ret;
1210 				last_len += ret << PAGE_CACHE_SHIFT;
1211 			} else {
1212 				i++;
1213 				last_len = 0;
1214 			}
1215 		}
1216 	}
1217 
1218 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1219 		filemap_flush(inode->i_mapping);
1220 
1221 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1222 		/* the filemap_flush will queue IO into the worker threads, but
1223 		 * we have to make sure the IO is actually started and that
1224 		 * ordered extents get created before we return
1225 		 */
1226 		atomic_inc(&root->fs_info->async_submit_draining);
1227 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1228 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1229 			wait_event(root->fs_info->async_submit_wait,
1230 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1231 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1232 		}
1233 		atomic_dec(&root->fs_info->async_submit_draining);
1234 
1235 		mutex_lock(&inode->i_mutex);
1236 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1237 		mutex_unlock(&inode->i_mutex);
1238 	}
1239 
1240 	disk_super = root->fs_info->super_copy;
1241 	features = btrfs_super_incompat_flags(disk_super);
1242 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1243 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1244 		btrfs_set_super_incompat_flags(disk_super, features);
1245 	}
1246 
1247 	ret = defrag_count;
1248 
1249 out_ra:
1250 	if (!file)
1251 		kfree(ra);
1252 	kfree(pages);
1253 	return ret;
1254 }
1255 
1256 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1257 					void __user *arg)
1258 {
1259 	u64 new_size;
1260 	u64 old_size;
1261 	u64 devid = 1;
1262 	struct btrfs_ioctl_vol_args *vol_args;
1263 	struct btrfs_trans_handle *trans;
1264 	struct btrfs_device *device = NULL;
1265 	char *sizestr;
1266 	char *devstr = NULL;
1267 	int ret = 0;
1268 	int mod = 0;
1269 
1270 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1271 		return -EROFS;
1272 
1273 	if (!capable(CAP_SYS_ADMIN))
1274 		return -EPERM;
1275 
1276 	mutex_lock(&root->fs_info->volume_mutex);
1277 	if (root->fs_info->balance_ctl) {
1278 		printk(KERN_INFO "btrfs: balance in progress\n");
1279 		ret = -EINVAL;
1280 		goto out;
1281 	}
1282 
1283 	vol_args = memdup_user(arg, sizeof(*vol_args));
1284 	if (IS_ERR(vol_args)) {
1285 		ret = PTR_ERR(vol_args);
1286 		goto out;
1287 	}
1288 
1289 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1290 
1291 	sizestr = vol_args->name;
1292 	devstr = strchr(sizestr, ':');
1293 	if (devstr) {
1294 		char *end;
1295 		sizestr = devstr + 1;
1296 		*devstr = '\0';
1297 		devstr = vol_args->name;
1298 		devid = simple_strtoull(devstr, &end, 10);
1299 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1300 		       (unsigned long long)devid);
1301 	}
1302 	device = btrfs_find_device(root, devid, NULL, NULL);
1303 	if (!device) {
1304 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1305 		       (unsigned long long)devid);
1306 		ret = -EINVAL;
1307 		goto out_free;
1308 	}
1309 	if (device->fs_devices && device->fs_devices->seeding) {
1310 		printk(KERN_INFO "btrfs: resizer unable to apply on "
1311 		       "seeding device %llu\n",
1312 		       (unsigned long long)devid);
1313 		ret = -EINVAL;
1314 		goto out_free;
1315 	}
1316 
1317 	if (!strcmp(sizestr, "max"))
1318 		new_size = device->bdev->bd_inode->i_size;
1319 	else {
1320 		if (sizestr[0] == '-') {
1321 			mod = -1;
1322 			sizestr++;
1323 		} else if (sizestr[0] == '+') {
1324 			mod = 1;
1325 			sizestr++;
1326 		}
1327 		new_size = memparse(sizestr, NULL);
1328 		if (new_size == 0) {
1329 			ret = -EINVAL;
1330 			goto out_free;
1331 		}
1332 	}
1333 
1334 	old_size = device->total_bytes;
1335 
1336 	if (mod < 0) {
1337 		if (new_size > old_size) {
1338 			ret = -EINVAL;
1339 			goto out_free;
1340 		}
1341 		new_size = old_size - new_size;
1342 	} else if (mod > 0) {
1343 		new_size = old_size + new_size;
1344 	}
1345 
1346 	if (new_size < 256 * 1024 * 1024) {
1347 		ret = -EINVAL;
1348 		goto out_free;
1349 	}
1350 	if (new_size > device->bdev->bd_inode->i_size) {
1351 		ret = -EFBIG;
1352 		goto out_free;
1353 	}
1354 
1355 	do_div(new_size, root->sectorsize);
1356 	new_size *= root->sectorsize;
1357 
1358 	printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1359 		      rcu_str_deref(device->name),
1360 		      (unsigned long long)new_size);
1361 
1362 	if (new_size > old_size) {
1363 		trans = btrfs_start_transaction(root, 0);
1364 		if (IS_ERR(trans)) {
1365 			ret = PTR_ERR(trans);
1366 			goto out_free;
1367 		}
1368 		ret = btrfs_grow_device(trans, device, new_size);
1369 		btrfs_commit_transaction(trans, root);
1370 	} else if (new_size < old_size) {
1371 		ret = btrfs_shrink_device(device, new_size);
1372 	}
1373 
1374 out_free:
1375 	kfree(vol_args);
1376 out:
1377 	mutex_unlock(&root->fs_info->volume_mutex);
1378 	return ret;
1379 }
1380 
1381 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1382 						    char *name,
1383 						    unsigned long fd,
1384 						    int subvol,
1385 						    u64 *transid,
1386 						    bool readonly)
1387 {
1388 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1389 	struct file *src_file;
1390 	int namelen;
1391 	int ret = 0;
1392 
1393 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1394 		return -EROFS;
1395 
1396 	namelen = strlen(name);
1397 	if (strchr(name, '/')) {
1398 		ret = -EINVAL;
1399 		goto out;
1400 	}
1401 
1402 	if (name[0] == '.' &&
1403 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1404 		ret = -EEXIST;
1405 		goto out;
1406 	}
1407 
1408 	if (subvol) {
1409 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1410 				     NULL, transid, readonly);
1411 	} else {
1412 		struct inode *src_inode;
1413 		src_file = fget(fd);
1414 		if (!src_file) {
1415 			ret = -EINVAL;
1416 			goto out;
1417 		}
1418 
1419 		src_inode = src_file->f_path.dentry->d_inode;
1420 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1421 			printk(KERN_INFO "btrfs: Snapshot src from "
1422 			       "another FS\n");
1423 			ret = -EINVAL;
1424 			fput(src_file);
1425 			goto out;
1426 		}
1427 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1428 				     BTRFS_I(src_inode)->root,
1429 				     transid, readonly);
1430 		fput(src_file);
1431 	}
1432 out:
1433 	return ret;
1434 }
1435 
1436 static noinline int btrfs_ioctl_snap_create(struct file *file,
1437 					    void __user *arg, int subvol)
1438 {
1439 	struct btrfs_ioctl_vol_args *vol_args;
1440 	int ret;
1441 
1442 	vol_args = memdup_user(arg, sizeof(*vol_args));
1443 	if (IS_ERR(vol_args))
1444 		return PTR_ERR(vol_args);
1445 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1446 
1447 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1448 					      vol_args->fd, subvol,
1449 					      NULL, false);
1450 
1451 	kfree(vol_args);
1452 	return ret;
1453 }
1454 
1455 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1456 					       void __user *arg, int subvol)
1457 {
1458 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1459 	int ret;
1460 	u64 transid = 0;
1461 	u64 *ptr = NULL;
1462 	bool readonly = false;
1463 
1464 	vol_args = memdup_user(arg, sizeof(*vol_args));
1465 	if (IS_ERR(vol_args))
1466 		return PTR_ERR(vol_args);
1467 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1468 
1469 	if (vol_args->flags &
1470 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1471 		ret = -EOPNOTSUPP;
1472 		goto out;
1473 	}
1474 
1475 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1476 		ptr = &transid;
1477 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1478 		readonly = true;
1479 
1480 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1481 					      vol_args->fd, subvol,
1482 					      ptr, readonly);
1483 
1484 	if (ret == 0 && ptr &&
1485 	    copy_to_user(arg +
1486 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1487 				  transid), ptr, sizeof(*ptr)))
1488 		ret = -EFAULT;
1489 out:
1490 	kfree(vol_args);
1491 	return ret;
1492 }
1493 
1494 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1495 						void __user *arg)
1496 {
1497 	struct inode *inode = fdentry(file)->d_inode;
1498 	struct btrfs_root *root = BTRFS_I(inode)->root;
1499 	int ret = 0;
1500 	u64 flags = 0;
1501 
1502 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1503 		return -EINVAL;
1504 
1505 	down_read(&root->fs_info->subvol_sem);
1506 	if (btrfs_root_readonly(root))
1507 		flags |= BTRFS_SUBVOL_RDONLY;
1508 	up_read(&root->fs_info->subvol_sem);
1509 
1510 	if (copy_to_user(arg, &flags, sizeof(flags)))
1511 		ret = -EFAULT;
1512 
1513 	return ret;
1514 }
1515 
1516 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1517 					      void __user *arg)
1518 {
1519 	struct inode *inode = fdentry(file)->d_inode;
1520 	struct btrfs_root *root = BTRFS_I(inode)->root;
1521 	struct btrfs_trans_handle *trans;
1522 	u64 root_flags;
1523 	u64 flags;
1524 	int ret = 0;
1525 
1526 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1527 		return -EROFS;
1528 
1529 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1530 		return -EINVAL;
1531 
1532 	if (copy_from_user(&flags, arg, sizeof(flags)))
1533 		return -EFAULT;
1534 
1535 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1536 		return -EINVAL;
1537 
1538 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1539 		return -EOPNOTSUPP;
1540 
1541 	if (!inode_owner_or_capable(inode))
1542 		return -EACCES;
1543 
1544 	down_write(&root->fs_info->subvol_sem);
1545 
1546 	/* nothing to do */
1547 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1548 		goto out;
1549 
1550 	root_flags = btrfs_root_flags(&root->root_item);
1551 	if (flags & BTRFS_SUBVOL_RDONLY)
1552 		btrfs_set_root_flags(&root->root_item,
1553 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1554 	else
1555 		btrfs_set_root_flags(&root->root_item,
1556 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1557 
1558 	trans = btrfs_start_transaction(root, 1);
1559 	if (IS_ERR(trans)) {
1560 		ret = PTR_ERR(trans);
1561 		goto out_reset;
1562 	}
1563 
1564 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1565 				&root->root_key, &root->root_item);
1566 
1567 	btrfs_commit_transaction(trans, root);
1568 out_reset:
1569 	if (ret)
1570 		btrfs_set_root_flags(&root->root_item, root_flags);
1571 out:
1572 	up_write(&root->fs_info->subvol_sem);
1573 	return ret;
1574 }
1575 
1576 /*
1577  * helper to check if the subvolume references other subvolumes
1578  */
1579 static noinline int may_destroy_subvol(struct btrfs_root *root)
1580 {
1581 	struct btrfs_path *path;
1582 	struct btrfs_key key;
1583 	int ret;
1584 
1585 	path = btrfs_alloc_path();
1586 	if (!path)
1587 		return -ENOMEM;
1588 
1589 	key.objectid = root->root_key.objectid;
1590 	key.type = BTRFS_ROOT_REF_KEY;
1591 	key.offset = (u64)-1;
1592 
1593 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1594 				&key, path, 0, 0);
1595 	if (ret < 0)
1596 		goto out;
1597 	BUG_ON(ret == 0);
1598 
1599 	ret = 0;
1600 	if (path->slots[0] > 0) {
1601 		path->slots[0]--;
1602 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1603 		if (key.objectid == root->root_key.objectid &&
1604 		    key.type == BTRFS_ROOT_REF_KEY)
1605 			ret = -ENOTEMPTY;
1606 	}
1607 out:
1608 	btrfs_free_path(path);
1609 	return ret;
1610 }
1611 
1612 static noinline int key_in_sk(struct btrfs_key *key,
1613 			      struct btrfs_ioctl_search_key *sk)
1614 {
1615 	struct btrfs_key test;
1616 	int ret;
1617 
1618 	test.objectid = sk->min_objectid;
1619 	test.type = sk->min_type;
1620 	test.offset = sk->min_offset;
1621 
1622 	ret = btrfs_comp_cpu_keys(key, &test);
1623 	if (ret < 0)
1624 		return 0;
1625 
1626 	test.objectid = sk->max_objectid;
1627 	test.type = sk->max_type;
1628 	test.offset = sk->max_offset;
1629 
1630 	ret = btrfs_comp_cpu_keys(key, &test);
1631 	if (ret > 0)
1632 		return 0;
1633 	return 1;
1634 }
1635 
1636 static noinline int copy_to_sk(struct btrfs_root *root,
1637 			       struct btrfs_path *path,
1638 			       struct btrfs_key *key,
1639 			       struct btrfs_ioctl_search_key *sk,
1640 			       char *buf,
1641 			       unsigned long *sk_offset,
1642 			       int *num_found)
1643 {
1644 	u64 found_transid;
1645 	struct extent_buffer *leaf;
1646 	struct btrfs_ioctl_search_header sh;
1647 	unsigned long item_off;
1648 	unsigned long item_len;
1649 	int nritems;
1650 	int i;
1651 	int slot;
1652 	int ret = 0;
1653 
1654 	leaf = path->nodes[0];
1655 	slot = path->slots[0];
1656 	nritems = btrfs_header_nritems(leaf);
1657 
1658 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1659 		i = nritems;
1660 		goto advance_key;
1661 	}
1662 	found_transid = btrfs_header_generation(leaf);
1663 
1664 	for (i = slot; i < nritems; i++) {
1665 		item_off = btrfs_item_ptr_offset(leaf, i);
1666 		item_len = btrfs_item_size_nr(leaf, i);
1667 
1668 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1669 			item_len = 0;
1670 
1671 		if (sizeof(sh) + item_len + *sk_offset >
1672 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1673 			ret = 1;
1674 			goto overflow;
1675 		}
1676 
1677 		btrfs_item_key_to_cpu(leaf, key, i);
1678 		if (!key_in_sk(key, sk))
1679 			continue;
1680 
1681 		sh.objectid = key->objectid;
1682 		sh.offset = key->offset;
1683 		sh.type = key->type;
1684 		sh.len = item_len;
1685 		sh.transid = found_transid;
1686 
1687 		/* copy search result header */
1688 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1689 		*sk_offset += sizeof(sh);
1690 
1691 		if (item_len) {
1692 			char *p = buf + *sk_offset;
1693 			/* copy the item */
1694 			read_extent_buffer(leaf, p,
1695 					   item_off, item_len);
1696 			*sk_offset += item_len;
1697 		}
1698 		(*num_found)++;
1699 
1700 		if (*num_found >= sk->nr_items)
1701 			break;
1702 	}
1703 advance_key:
1704 	ret = 0;
1705 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1706 		key->offset++;
1707 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1708 		key->offset = 0;
1709 		key->type++;
1710 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1711 		key->offset = 0;
1712 		key->type = 0;
1713 		key->objectid++;
1714 	} else
1715 		ret = 1;
1716 overflow:
1717 	return ret;
1718 }
1719 
1720 static noinline int search_ioctl(struct inode *inode,
1721 				 struct btrfs_ioctl_search_args *args)
1722 {
1723 	struct btrfs_root *root;
1724 	struct btrfs_key key;
1725 	struct btrfs_key max_key;
1726 	struct btrfs_path *path;
1727 	struct btrfs_ioctl_search_key *sk = &args->key;
1728 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1729 	int ret;
1730 	int num_found = 0;
1731 	unsigned long sk_offset = 0;
1732 
1733 	path = btrfs_alloc_path();
1734 	if (!path)
1735 		return -ENOMEM;
1736 
1737 	if (sk->tree_id == 0) {
1738 		/* search the root of the inode that was passed */
1739 		root = BTRFS_I(inode)->root;
1740 	} else {
1741 		key.objectid = sk->tree_id;
1742 		key.type = BTRFS_ROOT_ITEM_KEY;
1743 		key.offset = (u64)-1;
1744 		root = btrfs_read_fs_root_no_name(info, &key);
1745 		if (IS_ERR(root)) {
1746 			printk(KERN_ERR "could not find root %llu\n",
1747 			       sk->tree_id);
1748 			btrfs_free_path(path);
1749 			return -ENOENT;
1750 		}
1751 	}
1752 
1753 	key.objectid = sk->min_objectid;
1754 	key.type = sk->min_type;
1755 	key.offset = sk->min_offset;
1756 
1757 	max_key.objectid = sk->max_objectid;
1758 	max_key.type = sk->max_type;
1759 	max_key.offset = sk->max_offset;
1760 
1761 	path->keep_locks = 1;
1762 
1763 	while(1) {
1764 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1765 					   sk->min_transid);
1766 		if (ret != 0) {
1767 			if (ret > 0)
1768 				ret = 0;
1769 			goto err;
1770 		}
1771 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1772 				 &sk_offset, &num_found);
1773 		btrfs_release_path(path);
1774 		if (ret || num_found >= sk->nr_items)
1775 			break;
1776 
1777 	}
1778 	ret = 0;
1779 err:
1780 	sk->nr_items = num_found;
1781 	btrfs_free_path(path);
1782 	return ret;
1783 }
1784 
1785 static noinline int btrfs_ioctl_tree_search(struct file *file,
1786 					   void __user *argp)
1787 {
1788 	 struct btrfs_ioctl_search_args *args;
1789 	 struct inode *inode;
1790 	 int ret;
1791 
1792 	if (!capable(CAP_SYS_ADMIN))
1793 		return -EPERM;
1794 
1795 	args = memdup_user(argp, sizeof(*args));
1796 	if (IS_ERR(args))
1797 		return PTR_ERR(args);
1798 
1799 	inode = fdentry(file)->d_inode;
1800 	ret = search_ioctl(inode, args);
1801 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1802 		ret = -EFAULT;
1803 	kfree(args);
1804 	return ret;
1805 }
1806 
1807 /*
1808  * Search INODE_REFs to identify path name of 'dirid' directory
1809  * in a 'tree_id' tree. and sets path name to 'name'.
1810  */
1811 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1812 				u64 tree_id, u64 dirid, char *name)
1813 {
1814 	struct btrfs_root *root;
1815 	struct btrfs_key key;
1816 	char *ptr;
1817 	int ret = -1;
1818 	int slot;
1819 	int len;
1820 	int total_len = 0;
1821 	struct btrfs_inode_ref *iref;
1822 	struct extent_buffer *l;
1823 	struct btrfs_path *path;
1824 
1825 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1826 		name[0]='\0';
1827 		return 0;
1828 	}
1829 
1830 	path = btrfs_alloc_path();
1831 	if (!path)
1832 		return -ENOMEM;
1833 
1834 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1835 
1836 	key.objectid = tree_id;
1837 	key.type = BTRFS_ROOT_ITEM_KEY;
1838 	key.offset = (u64)-1;
1839 	root = btrfs_read_fs_root_no_name(info, &key);
1840 	if (IS_ERR(root)) {
1841 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1842 		ret = -ENOENT;
1843 		goto out;
1844 	}
1845 
1846 	key.objectid = dirid;
1847 	key.type = BTRFS_INODE_REF_KEY;
1848 	key.offset = (u64)-1;
1849 
1850 	while(1) {
1851 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1852 		if (ret < 0)
1853 			goto out;
1854 
1855 		l = path->nodes[0];
1856 		slot = path->slots[0];
1857 		if (ret > 0 && slot > 0)
1858 			slot--;
1859 		btrfs_item_key_to_cpu(l, &key, slot);
1860 
1861 		if (ret > 0 && (key.objectid != dirid ||
1862 				key.type != BTRFS_INODE_REF_KEY)) {
1863 			ret = -ENOENT;
1864 			goto out;
1865 		}
1866 
1867 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1868 		len = btrfs_inode_ref_name_len(l, iref);
1869 		ptr -= len + 1;
1870 		total_len += len + 1;
1871 		if (ptr < name)
1872 			goto out;
1873 
1874 		*(ptr + len) = '/';
1875 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1876 
1877 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1878 			break;
1879 
1880 		btrfs_release_path(path);
1881 		key.objectid = key.offset;
1882 		key.offset = (u64)-1;
1883 		dirid = key.objectid;
1884 	}
1885 	if (ptr < name)
1886 		goto out;
1887 	memmove(name, ptr, total_len);
1888 	name[total_len]='\0';
1889 	ret = 0;
1890 out:
1891 	btrfs_free_path(path);
1892 	return ret;
1893 }
1894 
1895 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1896 					   void __user *argp)
1897 {
1898 	 struct btrfs_ioctl_ino_lookup_args *args;
1899 	 struct inode *inode;
1900 	 int ret;
1901 
1902 	if (!capable(CAP_SYS_ADMIN))
1903 		return -EPERM;
1904 
1905 	args = memdup_user(argp, sizeof(*args));
1906 	if (IS_ERR(args))
1907 		return PTR_ERR(args);
1908 
1909 	inode = fdentry(file)->d_inode;
1910 
1911 	if (args->treeid == 0)
1912 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1913 
1914 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1915 					args->treeid, args->objectid,
1916 					args->name);
1917 
1918 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1919 		ret = -EFAULT;
1920 
1921 	kfree(args);
1922 	return ret;
1923 }
1924 
1925 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1926 					     void __user *arg)
1927 {
1928 	struct dentry *parent = fdentry(file);
1929 	struct dentry *dentry;
1930 	struct inode *dir = parent->d_inode;
1931 	struct inode *inode;
1932 	struct btrfs_root *root = BTRFS_I(dir)->root;
1933 	struct btrfs_root *dest = NULL;
1934 	struct btrfs_ioctl_vol_args *vol_args;
1935 	struct btrfs_trans_handle *trans;
1936 	int namelen;
1937 	int ret;
1938 	int err = 0;
1939 
1940 	vol_args = memdup_user(arg, sizeof(*vol_args));
1941 	if (IS_ERR(vol_args))
1942 		return PTR_ERR(vol_args);
1943 
1944 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1945 	namelen = strlen(vol_args->name);
1946 	if (strchr(vol_args->name, '/') ||
1947 	    strncmp(vol_args->name, "..", namelen) == 0) {
1948 		err = -EINVAL;
1949 		goto out;
1950 	}
1951 
1952 	err = mnt_want_write_file(file);
1953 	if (err)
1954 		goto out;
1955 
1956 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1957 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1958 	if (IS_ERR(dentry)) {
1959 		err = PTR_ERR(dentry);
1960 		goto out_unlock_dir;
1961 	}
1962 
1963 	if (!dentry->d_inode) {
1964 		err = -ENOENT;
1965 		goto out_dput;
1966 	}
1967 
1968 	inode = dentry->d_inode;
1969 	dest = BTRFS_I(inode)->root;
1970 	if (!capable(CAP_SYS_ADMIN)){
1971 		/*
1972 		 * Regular user.  Only allow this with a special mount
1973 		 * option, when the user has write+exec access to the
1974 		 * subvol root, and when rmdir(2) would have been
1975 		 * allowed.
1976 		 *
1977 		 * Note that this is _not_ check that the subvol is
1978 		 * empty or doesn't contain data that we wouldn't
1979 		 * otherwise be able to delete.
1980 		 *
1981 		 * Users who want to delete empty subvols should try
1982 		 * rmdir(2).
1983 		 */
1984 		err = -EPERM;
1985 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1986 			goto out_dput;
1987 
1988 		/*
1989 		 * Do not allow deletion if the parent dir is the same
1990 		 * as the dir to be deleted.  That means the ioctl
1991 		 * must be called on the dentry referencing the root
1992 		 * of the subvol, not a random directory contained
1993 		 * within it.
1994 		 */
1995 		err = -EINVAL;
1996 		if (root == dest)
1997 			goto out_dput;
1998 
1999 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2000 		if (err)
2001 			goto out_dput;
2002 
2003 		/* check if subvolume may be deleted by a non-root user */
2004 		err = btrfs_may_delete(dir, dentry, 1);
2005 		if (err)
2006 			goto out_dput;
2007 	}
2008 
2009 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2010 		err = -EINVAL;
2011 		goto out_dput;
2012 	}
2013 
2014 	mutex_lock(&inode->i_mutex);
2015 	err = d_invalidate(dentry);
2016 	if (err)
2017 		goto out_unlock;
2018 
2019 	down_write(&root->fs_info->subvol_sem);
2020 
2021 	err = may_destroy_subvol(dest);
2022 	if (err)
2023 		goto out_up_write;
2024 
2025 	trans = btrfs_start_transaction(root, 0);
2026 	if (IS_ERR(trans)) {
2027 		err = PTR_ERR(trans);
2028 		goto out_up_write;
2029 	}
2030 	trans->block_rsv = &root->fs_info->global_block_rsv;
2031 
2032 	ret = btrfs_unlink_subvol(trans, root, dir,
2033 				dest->root_key.objectid,
2034 				dentry->d_name.name,
2035 				dentry->d_name.len);
2036 	if (ret) {
2037 		err = ret;
2038 		btrfs_abort_transaction(trans, root, ret);
2039 		goto out_end_trans;
2040 	}
2041 
2042 	btrfs_record_root_in_trans(trans, dest);
2043 
2044 	memset(&dest->root_item.drop_progress, 0,
2045 		sizeof(dest->root_item.drop_progress));
2046 	dest->root_item.drop_level = 0;
2047 	btrfs_set_root_refs(&dest->root_item, 0);
2048 
2049 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2050 		ret = btrfs_insert_orphan_item(trans,
2051 					root->fs_info->tree_root,
2052 					dest->root_key.objectid);
2053 		if (ret) {
2054 			btrfs_abort_transaction(trans, root, ret);
2055 			err = ret;
2056 			goto out_end_trans;
2057 		}
2058 	}
2059 out_end_trans:
2060 	ret = btrfs_end_transaction(trans, root);
2061 	if (ret && !err)
2062 		err = ret;
2063 	inode->i_flags |= S_DEAD;
2064 out_up_write:
2065 	up_write(&root->fs_info->subvol_sem);
2066 out_unlock:
2067 	mutex_unlock(&inode->i_mutex);
2068 	if (!err) {
2069 		shrink_dcache_sb(root->fs_info->sb);
2070 		btrfs_invalidate_inodes(dest);
2071 		d_delete(dentry);
2072 	}
2073 out_dput:
2074 	dput(dentry);
2075 out_unlock_dir:
2076 	mutex_unlock(&dir->i_mutex);
2077 	mnt_drop_write_file(file);
2078 out:
2079 	kfree(vol_args);
2080 	return err;
2081 }
2082 
2083 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2084 {
2085 	struct inode *inode = fdentry(file)->d_inode;
2086 	struct btrfs_root *root = BTRFS_I(inode)->root;
2087 	struct btrfs_ioctl_defrag_range_args *range;
2088 	int ret;
2089 
2090 	if (btrfs_root_readonly(root))
2091 		return -EROFS;
2092 
2093 	ret = mnt_want_write_file(file);
2094 	if (ret)
2095 		return ret;
2096 
2097 	switch (inode->i_mode & S_IFMT) {
2098 	case S_IFDIR:
2099 		if (!capable(CAP_SYS_ADMIN)) {
2100 			ret = -EPERM;
2101 			goto out;
2102 		}
2103 		ret = btrfs_defrag_root(root, 0);
2104 		if (ret)
2105 			goto out;
2106 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2107 		break;
2108 	case S_IFREG:
2109 		if (!(file->f_mode & FMODE_WRITE)) {
2110 			ret = -EINVAL;
2111 			goto out;
2112 		}
2113 
2114 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2115 		if (!range) {
2116 			ret = -ENOMEM;
2117 			goto out;
2118 		}
2119 
2120 		if (argp) {
2121 			if (copy_from_user(range, argp,
2122 					   sizeof(*range))) {
2123 				ret = -EFAULT;
2124 				kfree(range);
2125 				goto out;
2126 			}
2127 			/* compression requires us to start the IO */
2128 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2129 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2130 				range->extent_thresh = (u32)-1;
2131 			}
2132 		} else {
2133 			/* the rest are all set to zero by kzalloc */
2134 			range->len = (u64)-1;
2135 		}
2136 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2137 					range, 0, 0);
2138 		if (ret > 0)
2139 			ret = 0;
2140 		kfree(range);
2141 		break;
2142 	default:
2143 		ret = -EINVAL;
2144 	}
2145 out:
2146 	mnt_drop_write_file(file);
2147 	return ret;
2148 }
2149 
2150 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2151 {
2152 	struct btrfs_ioctl_vol_args *vol_args;
2153 	int ret;
2154 
2155 	if (!capable(CAP_SYS_ADMIN))
2156 		return -EPERM;
2157 
2158 	mutex_lock(&root->fs_info->volume_mutex);
2159 	if (root->fs_info->balance_ctl) {
2160 		printk(KERN_INFO "btrfs: balance in progress\n");
2161 		ret = -EINVAL;
2162 		goto out;
2163 	}
2164 
2165 	vol_args = memdup_user(arg, sizeof(*vol_args));
2166 	if (IS_ERR(vol_args)) {
2167 		ret = PTR_ERR(vol_args);
2168 		goto out;
2169 	}
2170 
2171 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2172 	ret = btrfs_init_new_device(root, vol_args->name);
2173 
2174 	kfree(vol_args);
2175 out:
2176 	mutex_unlock(&root->fs_info->volume_mutex);
2177 	return ret;
2178 }
2179 
2180 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2181 {
2182 	struct btrfs_ioctl_vol_args *vol_args;
2183 	int ret;
2184 
2185 	if (!capable(CAP_SYS_ADMIN))
2186 		return -EPERM;
2187 
2188 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2189 		return -EROFS;
2190 
2191 	mutex_lock(&root->fs_info->volume_mutex);
2192 	if (root->fs_info->balance_ctl) {
2193 		printk(KERN_INFO "btrfs: balance in progress\n");
2194 		ret = -EINVAL;
2195 		goto out;
2196 	}
2197 
2198 	vol_args = memdup_user(arg, sizeof(*vol_args));
2199 	if (IS_ERR(vol_args)) {
2200 		ret = PTR_ERR(vol_args);
2201 		goto out;
2202 	}
2203 
2204 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2205 	ret = btrfs_rm_device(root, vol_args->name);
2206 
2207 	kfree(vol_args);
2208 out:
2209 	mutex_unlock(&root->fs_info->volume_mutex);
2210 	return ret;
2211 }
2212 
2213 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2214 {
2215 	struct btrfs_ioctl_fs_info_args *fi_args;
2216 	struct btrfs_device *device;
2217 	struct btrfs_device *next;
2218 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2219 	int ret = 0;
2220 
2221 	if (!capable(CAP_SYS_ADMIN))
2222 		return -EPERM;
2223 
2224 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2225 	if (!fi_args)
2226 		return -ENOMEM;
2227 
2228 	fi_args->num_devices = fs_devices->num_devices;
2229 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2230 
2231 	mutex_lock(&fs_devices->device_list_mutex);
2232 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2233 		if (device->devid > fi_args->max_id)
2234 			fi_args->max_id = device->devid;
2235 	}
2236 	mutex_unlock(&fs_devices->device_list_mutex);
2237 
2238 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2239 		ret = -EFAULT;
2240 
2241 	kfree(fi_args);
2242 	return ret;
2243 }
2244 
2245 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2246 {
2247 	struct btrfs_ioctl_dev_info_args *di_args;
2248 	struct btrfs_device *dev;
2249 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2250 	int ret = 0;
2251 	char *s_uuid = NULL;
2252 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2253 
2254 	if (!capable(CAP_SYS_ADMIN))
2255 		return -EPERM;
2256 
2257 	di_args = memdup_user(arg, sizeof(*di_args));
2258 	if (IS_ERR(di_args))
2259 		return PTR_ERR(di_args);
2260 
2261 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2262 		s_uuid = di_args->uuid;
2263 
2264 	mutex_lock(&fs_devices->device_list_mutex);
2265 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2266 	mutex_unlock(&fs_devices->device_list_mutex);
2267 
2268 	if (!dev) {
2269 		ret = -ENODEV;
2270 		goto out;
2271 	}
2272 
2273 	di_args->devid = dev->devid;
2274 	di_args->bytes_used = dev->bytes_used;
2275 	di_args->total_bytes = dev->total_bytes;
2276 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2277 	if (dev->name) {
2278 		struct rcu_string *name;
2279 
2280 		rcu_read_lock();
2281 		name = rcu_dereference(dev->name);
2282 		strncpy(di_args->path, name->str, sizeof(di_args->path));
2283 		rcu_read_unlock();
2284 		di_args->path[sizeof(di_args->path) - 1] = 0;
2285 	} else {
2286 		di_args->path[0] = '\0';
2287 	}
2288 
2289 out:
2290 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2291 		ret = -EFAULT;
2292 
2293 	kfree(di_args);
2294 	return ret;
2295 }
2296 
2297 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2298 				       u64 off, u64 olen, u64 destoff)
2299 {
2300 	struct inode *inode = fdentry(file)->d_inode;
2301 	struct btrfs_root *root = BTRFS_I(inode)->root;
2302 	struct file *src_file;
2303 	struct inode *src;
2304 	struct btrfs_trans_handle *trans;
2305 	struct btrfs_path *path;
2306 	struct extent_buffer *leaf;
2307 	char *buf;
2308 	struct btrfs_key key;
2309 	u32 nritems;
2310 	int slot;
2311 	int ret;
2312 	u64 len = olen;
2313 	u64 bs = root->fs_info->sb->s_blocksize;
2314 	u64 hint_byte;
2315 
2316 	/*
2317 	 * TODO:
2318 	 * - split compressed inline extents.  annoying: we need to
2319 	 *   decompress into destination's address_space (the file offset
2320 	 *   may change, so source mapping won't do), then recompress (or
2321 	 *   otherwise reinsert) a subrange.
2322 	 * - allow ranges within the same file to be cloned (provided
2323 	 *   they don't overlap)?
2324 	 */
2325 
2326 	/* the destination must be opened for writing */
2327 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2328 		return -EINVAL;
2329 
2330 	if (btrfs_root_readonly(root))
2331 		return -EROFS;
2332 
2333 	ret = mnt_want_write_file(file);
2334 	if (ret)
2335 		return ret;
2336 
2337 	src_file = fget(srcfd);
2338 	if (!src_file) {
2339 		ret = -EBADF;
2340 		goto out_drop_write;
2341 	}
2342 
2343 	src = src_file->f_dentry->d_inode;
2344 
2345 	ret = -EINVAL;
2346 	if (src == inode)
2347 		goto out_fput;
2348 
2349 	/* the src must be open for reading */
2350 	if (!(src_file->f_mode & FMODE_READ))
2351 		goto out_fput;
2352 
2353 	/* don't make the dst file partly checksummed */
2354 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2355 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2356 		goto out_fput;
2357 
2358 	ret = -EISDIR;
2359 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2360 		goto out_fput;
2361 
2362 	ret = -EXDEV;
2363 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2364 		goto out_fput;
2365 
2366 	ret = -ENOMEM;
2367 	buf = vmalloc(btrfs_level_size(root, 0));
2368 	if (!buf)
2369 		goto out_fput;
2370 
2371 	path = btrfs_alloc_path();
2372 	if (!path) {
2373 		vfree(buf);
2374 		goto out_fput;
2375 	}
2376 	path->reada = 2;
2377 
2378 	if (inode < src) {
2379 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2380 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2381 	} else {
2382 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2383 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2384 	}
2385 
2386 	/* determine range to clone */
2387 	ret = -EINVAL;
2388 	if (off + len > src->i_size || off + len < off)
2389 		goto out_unlock;
2390 	if (len == 0)
2391 		olen = len = src->i_size - off;
2392 	/* if we extend to eof, continue to block boundary */
2393 	if (off + len == src->i_size)
2394 		len = ALIGN(src->i_size, bs) - off;
2395 
2396 	/* verify the end result is block aligned */
2397 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2398 	    !IS_ALIGNED(destoff, bs))
2399 		goto out_unlock;
2400 
2401 	if (destoff > inode->i_size) {
2402 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2403 		if (ret)
2404 			goto out_unlock;
2405 	}
2406 
2407 	/* truncate page cache pages from target inode range */
2408 	truncate_inode_pages_range(&inode->i_data, destoff,
2409 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2410 
2411 	/* do any pending delalloc/csum calc on src, one way or
2412 	   another, and lock file content */
2413 	while (1) {
2414 		struct btrfs_ordered_extent *ordered;
2415 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2416 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2417 		if (!ordered &&
2418 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2419 				   EXTENT_DELALLOC, 0, NULL))
2420 			break;
2421 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2422 		if (ordered)
2423 			btrfs_put_ordered_extent(ordered);
2424 		btrfs_wait_ordered_range(src, off, len);
2425 	}
2426 
2427 	/* clone data */
2428 	key.objectid = btrfs_ino(src);
2429 	key.type = BTRFS_EXTENT_DATA_KEY;
2430 	key.offset = 0;
2431 
2432 	while (1) {
2433 		/*
2434 		 * note the key will change type as we walk through the
2435 		 * tree.
2436 		 */
2437 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2438 		if (ret < 0)
2439 			goto out;
2440 
2441 		nritems = btrfs_header_nritems(path->nodes[0]);
2442 		if (path->slots[0] >= nritems) {
2443 			ret = btrfs_next_leaf(root, path);
2444 			if (ret < 0)
2445 				goto out;
2446 			if (ret > 0)
2447 				break;
2448 			nritems = btrfs_header_nritems(path->nodes[0]);
2449 		}
2450 		leaf = path->nodes[0];
2451 		slot = path->slots[0];
2452 
2453 		btrfs_item_key_to_cpu(leaf, &key, slot);
2454 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2455 		    key.objectid != btrfs_ino(src))
2456 			break;
2457 
2458 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2459 			struct btrfs_file_extent_item *extent;
2460 			int type;
2461 			u32 size;
2462 			struct btrfs_key new_key;
2463 			u64 disko = 0, diskl = 0;
2464 			u64 datao = 0, datal = 0;
2465 			u8 comp;
2466 			u64 endoff;
2467 
2468 			size = btrfs_item_size_nr(leaf, slot);
2469 			read_extent_buffer(leaf, buf,
2470 					   btrfs_item_ptr_offset(leaf, slot),
2471 					   size);
2472 
2473 			extent = btrfs_item_ptr(leaf, slot,
2474 						struct btrfs_file_extent_item);
2475 			comp = btrfs_file_extent_compression(leaf, extent);
2476 			type = btrfs_file_extent_type(leaf, extent);
2477 			if (type == BTRFS_FILE_EXTENT_REG ||
2478 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2479 				disko = btrfs_file_extent_disk_bytenr(leaf,
2480 								      extent);
2481 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2482 								 extent);
2483 				datao = btrfs_file_extent_offset(leaf, extent);
2484 				datal = btrfs_file_extent_num_bytes(leaf,
2485 								    extent);
2486 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2487 				/* take upper bound, may be compressed */
2488 				datal = btrfs_file_extent_ram_bytes(leaf,
2489 								    extent);
2490 			}
2491 			btrfs_release_path(path);
2492 
2493 			if (key.offset + datal <= off ||
2494 			    key.offset >= off+len)
2495 				goto next;
2496 
2497 			memcpy(&new_key, &key, sizeof(new_key));
2498 			new_key.objectid = btrfs_ino(inode);
2499 			if (off <= key.offset)
2500 				new_key.offset = key.offset + destoff - off;
2501 			else
2502 				new_key.offset = destoff;
2503 
2504 			/*
2505 			 * 1 - adjusting old extent (we may have to split it)
2506 			 * 1 - add new extent
2507 			 * 1 - inode update
2508 			 */
2509 			trans = btrfs_start_transaction(root, 3);
2510 			if (IS_ERR(trans)) {
2511 				ret = PTR_ERR(trans);
2512 				goto out;
2513 			}
2514 
2515 			if (type == BTRFS_FILE_EXTENT_REG ||
2516 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2517 				/*
2518 				 *    a  | --- range to clone ---|  b
2519 				 * | ------------- extent ------------- |
2520 				 */
2521 
2522 				/* substract range b */
2523 				if (key.offset + datal > off + len)
2524 					datal = off + len - key.offset;
2525 
2526 				/* substract range a */
2527 				if (off > key.offset) {
2528 					datao += off - key.offset;
2529 					datal -= off - key.offset;
2530 				}
2531 
2532 				ret = btrfs_drop_extents(trans, inode,
2533 							 new_key.offset,
2534 							 new_key.offset + datal,
2535 							 &hint_byte, 1);
2536 				if (ret) {
2537 					btrfs_abort_transaction(trans, root,
2538 								ret);
2539 					btrfs_end_transaction(trans, root);
2540 					goto out;
2541 				}
2542 
2543 				ret = btrfs_insert_empty_item(trans, root, path,
2544 							      &new_key, size);
2545 				if (ret) {
2546 					btrfs_abort_transaction(trans, root,
2547 								ret);
2548 					btrfs_end_transaction(trans, root);
2549 					goto out;
2550 				}
2551 
2552 				leaf = path->nodes[0];
2553 				slot = path->slots[0];
2554 				write_extent_buffer(leaf, buf,
2555 					    btrfs_item_ptr_offset(leaf, slot),
2556 					    size);
2557 
2558 				extent = btrfs_item_ptr(leaf, slot,
2559 						struct btrfs_file_extent_item);
2560 
2561 				/* disko == 0 means it's a hole */
2562 				if (!disko)
2563 					datao = 0;
2564 
2565 				btrfs_set_file_extent_offset(leaf, extent,
2566 							     datao);
2567 				btrfs_set_file_extent_num_bytes(leaf, extent,
2568 								datal);
2569 				if (disko) {
2570 					inode_add_bytes(inode, datal);
2571 					ret = btrfs_inc_extent_ref(trans, root,
2572 							disko, diskl, 0,
2573 							root->root_key.objectid,
2574 							btrfs_ino(inode),
2575 							new_key.offset - datao,
2576 							0);
2577 					if (ret) {
2578 						btrfs_abort_transaction(trans,
2579 									root,
2580 									ret);
2581 						btrfs_end_transaction(trans,
2582 								      root);
2583 						goto out;
2584 
2585 					}
2586 				}
2587 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2588 				u64 skip = 0;
2589 				u64 trim = 0;
2590 				if (off > key.offset) {
2591 					skip = off - key.offset;
2592 					new_key.offset += skip;
2593 				}
2594 
2595 				if (key.offset + datal > off+len)
2596 					trim = key.offset + datal - (off+len);
2597 
2598 				if (comp && (skip || trim)) {
2599 					ret = -EINVAL;
2600 					btrfs_end_transaction(trans, root);
2601 					goto out;
2602 				}
2603 				size -= skip + trim;
2604 				datal -= skip + trim;
2605 
2606 				ret = btrfs_drop_extents(trans, inode,
2607 							 new_key.offset,
2608 							 new_key.offset + datal,
2609 							 &hint_byte, 1);
2610 				if (ret) {
2611 					btrfs_abort_transaction(trans, root,
2612 								ret);
2613 					btrfs_end_transaction(trans, root);
2614 					goto out;
2615 				}
2616 
2617 				ret = btrfs_insert_empty_item(trans, root, path,
2618 							      &new_key, size);
2619 				if (ret) {
2620 					btrfs_abort_transaction(trans, root,
2621 								ret);
2622 					btrfs_end_transaction(trans, root);
2623 					goto out;
2624 				}
2625 
2626 				if (skip) {
2627 					u32 start =
2628 					  btrfs_file_extent_calc_inline_size(0);
2629 					memmove(buf+start, buf+start+skip,
2630 						datal);
2631 				}
2632 
2633 				leaf = path->nodes[0];
2634 				slot = path->slots[0];
2635 				write_extent_buffer(leaf, buf,
2636 					    btrfs_item_ptr_offset(leaf, slot),
2637 					    size);
2638 				inode_add_bytes(inode, datal);
2639 			}
2640 
2641 			btrfs_mark_buffer_dirty(leaf);
2642 			btrfs_release_path(path);
2643 
2644 			inode_inc_iversion(inode);
2645 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2646 
2647 			/*
2648 			 * we round up to the block size at eof when
2649 			 * determining which extents to clone above,
2650 			 * but shouldn't round up the file size
2651 			 */
2652 			endoff = new_key.offset + datal;
2653 			if (endoff > destoff+olen)
2654 				endoff = destoff+olen;
2655 			if (endoff > inode->i_size)
2656 				btrfs_i_size_write(inode, endoff);
2657 
2658 			ret = btrfs_update_inode(trans, root, inode);
2659 			if (ret) {
2660 				btrfs_abort_transaction(trans, root, ret);
2661 				btrfs_end_transaction(trans, root);
2662 				goto out;
2663 			}
2664 			ret = btrfs_end_transaction(trans, root);
2665 		}
2666 next:
2667 		btrfs_release_path(path);
2668 		key.offset++;
2669 	}
2670 	ret = 0;
2671 out:
2672 	btrfs_release_path(path);
2673 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2674 out_unlock:
2675 	mutex_unlock(&src->i_mutex);
2676 	mutex_unlock(&inode->i_mutex);
2677 	vfree(buf);
2678 	btrfs_free_path(path);
2679 out_fput:
2680 	fput(src_file);
2681 out_drop_write:
2682 	mnt_drop_write_file(file);
2683 	return ret;
2684 }
2685 
2686 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2687 {
2688 	struct btrfs_ioctl_clone_range_args args;
2689 
2690 	if (copy_from_user(&args, argp, sizeof(args)))
2691 		return -EFAULT;
2692 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2693 				 args.src_length, args.dest_offset);
2694 }
2695 
2696 /*
2697  * there are many ways the trans_start and trans_end ioctls can lead
2698  * to deadlocks.  They should only be used by applications that
2699  * basically own the machine, and have a very in depth understanding
2700  * of all the possible deadlocks and enospc problems.
2701  */
2702 static long btrfs_ioctl_trans_start(struct file *file)
2703 {
2704 	struct inode *inode = fdentry(file)->d_inode;
2705 	struct btrfs_root *root = BTRFS_I(inode)->root;
2706 	struct btrfs_trans_handle *trans;
2707 	int ret;
2708 
2709 	ret = -EPERM;
2710 	if (!capable(CAP_SYS_ADMIN))
2711 		goto out;
2712 
2713 	ret = -EINPROGRESS;
2714 	if (file->private_data)
2715 		goto out;
2716 
2717 	ret = -EROFS;
2718 	if (btrfs_root_readonly(root))
2719 		goto out;
2720 
2721 	ret = mnt_want_write_file(file);
2722 	if (ret)
2723 		goto out;
2724 
2725 	atomic_inc(&root->fs_info->open_ioctl_trans);
2726 
2727 	ret = -ENOMEM;
2728 	trans = btrfs_start_ioctl_transaction(root);
2729 	if (IS_ERR(trans))
2730 		goto out_drop;
2731 
2732 	file->private_data = trans;
2733 	return 0;
2734 
2735 out_drop:
2736 	atomic_dec(&root->fs_info->open_ioctl_trans);
2737 	mnt_drop_write_file(file);
2738 out:
2739 	return ret;
2740 }
2741 
2742 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2743 {
2744 	struct inode *inode = fdentry(file)->d_inode;
2745 	struct btrfs_root *root = BTRFS_I(inode)->root;
2746 	struct btrfs_root *new_root;
2747 	struct btrfs_dir_item *di;
2748 	struct btrfs_trans_handle *trans;
2749 	struct btrfs_path *path;
2750 	struct btrfs_key location;
2751 	struct btrfs_disk_key disk_key;
2752 	struct btrfs_super_block *disk_super;
2753 	u64 features;
2754 	u64 objectid = 0;
2755 	u64 dir_id;
2756 
2757 	if (!capable(CAP_SYS_ADMIN))
2758 		return -EPERM;
2759 
2760 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2761 		return -EFAULT;
2762 
2763 	if (!objectid)
2764 		objectid = root->root_key.objectid;
2765 
2766 	location.objectid = objectid;
2767 	location.type = BTRFS_ROOT_ITEM_KEY;
2768 	location.offset = (u64)-1;
2769 
2770 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2771 	if (IS_ERR(new_root))
2772 		return PTR_ERR(new_root);
2773 
2774 	if (btrfs_root_refs(&new_root->root_item) == 0)
2775 		return -ENOENT;
2776 
2777 	path = btrfs_alloc_path();
2778 	if (!path)
2779 		return -ENOMEM;
2780 	path->leave_spinning = 1;
2781 
2782 	trans = btrfs_start_transaction(root, 1);
2783 	if (IS_ERR(trans)) {
2784 		btrfs_free_path(path);
2785 		return PTR_ERR(trans);
2786 	}
2787 
2788 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2789 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2790 				   dir_id, "default", 7, 1);
2791 	if (IS_ERR_OR_NULL(di)) {
2792 		btrfs_free_path(path);
2793 		btrfs_end_transaction(trans, root);
2794 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2795 		       "this isn't going to work\n");
2796 		return -ENOENT;
2797 	}
2798 
2799 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2800 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2801 	btrfs_mark_buffer_dirty(path->nodes[0]);
2802 	btrfs_free_path(path);
2803 
2804 	disk_super = root->fs_info->super_copy;
2805 	features = btrfs_super_incompat_flags(disk_super);
2806 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2807 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2808 		btrfs_set_super_incompat_flags(disk_super, features);
2809 	}
2810 	btrfs_end_transaction(trans, root);
2811 
2812 	return 0;
2813 }
2814 
2815 static void get_block_group_info(struct list_head *groups_list,
2816 				 struct btrfs_ioctl_space_info *space)
2817 {
2818 	struct btrfs_block_group_cache *block_group;
2819 
2820 	space->total_bytes = 0;
2821 	space->used_bytes = 0;
2822 	space->flags = 0;
2823 	list_for_each_entry(block_group, groups_list, list) {
2824 		space->flags = block_group->flags;
2825 		space->total_bytes += block_group->key.offset;
2826 		space->used_bytes +=
2827 			btrfs_block_group_used(&block_group->item);
2828 	}
2829 }
2830 
2831 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2832 {
2833 	struct btrfs_ioctl_space_args space_args;
2834 	struct btrfs_ioctl_space_info space;
2835 	struct btrfs_ioctl_space_info *dest;
2836 	struct btrfs_ioctl_space_info *dest_orig;
2837 	struct btrfs_ioctl_space_info __user *user_dest;
2838 	struct btrfs_space_info *info;
2839 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2840 		       BTRFS_BLOCK_GROUP_SYSTEM,
2841 		       BTRFS_BLOCK_GROUP_METADATA,
2842 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2843 	int num_types = 4;
2844 	int alloc_size;
2845 	int ret = 0;
2846 	u64 slot_count = 0;
2847 	int i, c;
2848 
2849 	if (copy_from_user(&space_args,
2850 			   (struct btrfs_ioctl_space_args __user *)arg,
2851 			   sizeof(space_args)))
2852 		return -EFAULT;
2853 
2854 	for (i = 0; i < num_types; i++) {
2855 		struct btrfs_space_info *tmp;
2856 
2857 		info = NULL;
2858 		rcu_read_lock();
2859 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2860 					list) {
2861 			if (tmp->flags == types[i]) {
2862 				info = tmp;
2863 				break;
2864 			}
2865 		}
2866 		rcu_read_unlock();
2867 
2868 		if (!info)
2869 			continue;
2870 
2871 		down_read(&info->groups_sem);
2872 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2873 			if (!list_empty(&info->block_groups[c]))
2874 				slot_count++;
2875 		}
2876 		up_read(&info->groups_sem);
2877 	}
2878 
2879 	/* space_slots == 0 means they are asking for a count */
2880 	if (space_args.space_slots == 0) {
2881 		space_args.total_spaces = slot_count;
2882 		goto out;
2883 	}
2884 
2885 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2886 
2887 	alloc_size = sizeof(*dest) * slot_count;
2888 
2889 	/* we generally have at most 6 or so space infos, one for each raid
2890 	 * level.  So, a whole page should be more than enough for everyone
2891 	 */
2892 	if (alloc_size > PAGE_CACHE_SIZE)
2893 		return -ENOMEM;
2894 
2895 	space_args.total_spaces = 0;
2896 	dest = kmalloc(alloc_size, GFP_NOFS);
2897 	if (!dest)
2898 		return -ENOMEM;
2899 	dest_orig = dest;
2900 
2901 	/* now we have a buffer to copy into */
2902 	for (i = 0; i < num_types; i++) {
2903 		struct btrfs_space_info *tmp;
2904 
2905 		if (!slot_count)
2906 			break;
2907 
2908 		info = NULL;
2909 		rcu_read_lock();
2910 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2911 					list) {
2912 			if (tmp->flags == types[i]) {
2913 				info = tmp;
2914 				break;
2915 			}
2916 		}
2917 		rcu_read_unlock();
2918 
2919 		if (!info)
2920 			continue;
2921 		down_read(&info->groups_sem);
2922 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2923 			if (!list_empty(&info->block_groups[c])) {
2924 				get_block_group_info(&info->block_groups[c],
2925 						     &space);
2926 				memcpy(dest, &space, sizeof(space));
2927 				dest++;
2928 				space_args.total_spaces++;
2929 				slot_count--;
2930 			}
2931 			if (!slot_count)
2932 				break;
2933 		}
2934 		up_read(&info->groups_sem);
2935 	}
2936 
2937 	user_dest = (struct btrfs_ioctl_space_info __user *)
2938 		(arg + sizeof(struct btrfs_ioctl_space_args));
2939 
2940 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2941 		ret = -EFAULT;
2942 
2943 	kfree(dest_orig);
2944 out:
2945 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2946 		ret = -EFAULT;
2947 
2948 	return ret;
2949 }
2950 
2951 /*
2952  * there are many ways the trans_start and trans_end ioctls can lead
2953  * to deadlocks.  They should only be used by applications that
2954  * basically own the machine, and have a very in depth understanding
2955  * of all the possible deadlocks and enospc problems.
2956  */
2957 long btrfs_ioctl_trans_end(struct file *file)
2958 {
2959 	struct inode *inode = fdentry(file)->d_inode;
2960 	struct btrfs_root *root = BTRFS_I(inode)->root;
2961 	struct btrfs_trans_handle *trans;
2962 
2963 	trans = file->private_data;
2964 	if (!trans)
2965 		return -EINVAL;
2966 	file->private_data = NULL;
2967 
2968 	btrfs_end_transaction(trans, root);
2969 
2970 	atomic_dec(&root->fs_info->open_ioctl_trans);
2971 
2972 	mnt_drop_write_file(file);
2973 	return 0;
2974 }
2975 
2976 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2977 {
2978 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2979 	struct btrfs_trans_handle *trans;
2980 	u64 transid;
2981 	int ret;
2982 
2983 	trans = btrfs_start_transaction(root, 0);
2984 	if (IS_ERR(trans))
2985 		return PTR_ERR(trans);
2986 	transid = trans->transid;
2987 	ret = btrfs_commit_transaction_async(trans, root, 0);
2988 	if (ret) {
2989 		btrfs_end_transaction(trans, root);
2990 		return ret;
2991 	}
2992 
2993 	if (argp)
2994 		if (copy_to_user(argp, &transid, sizeof(transid)))
2995 			return -EFAULT;
2996 	return 0;
2997 }
2998 
2999 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3000 {
3001 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3002 	u64 transid;
3003 
3004 	if (argp) {
3005 		if (copy_from_user(&transid, argp, sizeof(transid)))
3006 			return -EFAULT;
3007 	} else {
3008 		transid = 0;  /* current trans */
3009 	}
3010 	return btrfs_wait_for_commit(root, transid);
3011 }
3012 
3013 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3014 {
3015 	int ret;
3016 	struct btrfs_ioctl_scrub_args *sa;
3017 
3018 	if (!capable(CAP_SYS_ADMIN))
3019 		return -EPERM;
3020 
3021 	sa = memdup_user(arg, sizeof(*sa));
3022 	if (IS_ERR(sa))
3023 		return PTR_ERR(sa);
3024 
3025 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3026 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3027 
3028 	if (copy_to_user(arg, sa, sizeof(*sa)))
3029 		ret = -EFAULT;
3030 
3031 	kfree(sa);
3032 	return ret;
3033 }
3034 
3035 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3036 {
3037 	if (!capable(CAP_SYS_ADMIN))
3038 		return -EPERM;
3039 
3040 	return btrfs_scrub_cancel(root);
3041 }
3042 
3043 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3044 				       void __user *arg)
3045 {
3046 	struct btrfs_ioctl_scrub_args *sa;
3047 	int ret;
3048 
3049 	if (!capable(CAP_SYS_ADMIN))
3050 		return -EPERM;
3051 
3052 	sa = memdup_user(arg, sizeof(*sa));
3053 	if (IS_ERR(sa))
3054 		return PTR_ERR(sa);
3055 
3056 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3057 
3058 	if (copy_to_user(arg, sa, sizeof(*sa)))
3059 		ret = -EFAULT;
3060 
3061 	kfree(sa);
3062 	return ret;
3063 }
3064 
3065 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3066 				      void __user *arg, int reset_after_read)
3067 {
3068 	struct btrfs_ioctl_get_dev_stats *sa;
3069 	int ret;
3070 
3071 	if (reset_after_read && !capable(CAP_SYS_ADMIN))
3072 		return -EPERM;
3073 
3074 	sa = memdup_user(arg, sizeof(*sa));
3075 	if (IS_ERR(sa))
3076 		return PTR_ERR(sa);
3077 
3078 	ret = btrfs_get_dev_stats(root, sa, reset_after_read);
3079 
3080 	if (copy_to_user(arg, sa, sizeof(*sa)))
3081 		ret = -EFAULT;
3082 
3083 	kfree(sa);
3084 	return ret;
3085 }
3086 
3087 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3088 {
3089 	int ret = 0;
3090 	int i;
3091 	u64 rel_ptr;
3092 	int size;
3093 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3094 	struct inode_fs_paths *ipath = NULL;
3095 	struct btrfs_path *path;
3096 
3097 	if (!capable(CAP_SYS_ADMIN))
3098 		return -EPERM;
3099 
3100 	path = btrfs_alloc_path();
3101 	if (!path) {
3102 		ret = -ENOMEM;
3103 		goto out;
3104 	}
3105 
3106 	ipa = memdup_user(arg, sizeof(*ipa));
3107 	if (IS_ERR(ipa)) {
3108 		ret = PTR_ERR(ipa);
3109 		ipa = NULL;
3110 		goto out;
3111 	}
3112 
3113 	size = min_t(u32, ipa->size, 4096);
3114 	ipath = init_ipath(size, root, path);
3115 	if (IS_ERR(ipath)) {
3116 		ret = PTR_ERR(ipath);
3117 		ipath = NULL;
3118 		goto out;
3119 	}
3120 
3121 	ret = paths_from_inode(ipa->inum, ipath);
3122 	if (ret < 0)
3123 		goto out;
3124 
3125 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3126 		rel_ptr = ipath->fspath->val[i] -
3127 			  (u64)(unsigned long)ipath->fspath->val;
3128 		ipath->fspath->val[i] = rel_ptr;
3129 	}
3130 
3131 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3132 			   (void *)(unsigned long)ipath->fspath, size);
3133 	if (ret) {
3134 		ret = -EFAULT;
3135 		goto out;
3136 	}
3137 
3138 out:
3139 	btrfs_free_path(path);
3140 	free_ipath(ipath);
3141 	kfree(ipa);
3142 
3143 	return ret;
3144 }
3145 
3146 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3147 {
3148 	struct btrfs_data_container *inodes = ctx;
3149 	const size_t c = 3 * sizeof(u64);
3150 
3151 	if (inodes->bytes_left >= c) {
3152 		inodes->bytes_left -= c;
3153 		inodes->val[inodes->elem_cnt] = inum;
3154 		inodes->val[inodes->elem_cnt + 1] = offset;
3155 		inodes->val[inodes->elem_cnt + 2] = root;
3156 		inodes->elem_cnt += 3;
3157 	} else {
3158 		inodes->bytes_missing += c - inodes->bytes_left;
3159 		inodes->bytes_left = 0;
3160 		inodes->elem_missed += 3;
3161 	}
3162 
3163 	return 0;
3164 }
3165 
3166 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3167 					void __user *arg)
3168 {
3169 	int ret = 0;
3170 	int size;
3171 	u64 extent_item_pos;
3172 	struct btrfs_ioctl_logical_ino_args *loi;
3173 	struct btrfs_data_container *inodes = NULL;
3174 	struct btrfs_path *path = NULL;
3175 	struct btrfs_key key;
3176 
3177 	if (!capable(CAP_SYS_ADMIN))
3178 		return -EPERM;
3179 
3180 	loi = memdup_user(arg, sizeof(*loi));
3181 	if (IS_ERR(loi)) {
3182 		ret = PTR_ERR(loi);
3183 		loi = NULL;
3184 		goto out;
3185 	}
3186 
3187 	path = btrfs_alloc_path();
3188 	if (!path) {
3189 		ret = -ENOMEM;
3190 		goto out;
3191 	}
3192 
3193 	size = min_t(u32, loi->size, 4096);
3194 	inodes = init_data_container(size);
3195 	if (IS_ERR(inodes)) {
3196 		ret = PTR_ERR(inodes);
3197 		inodes = NULL;
3198 		goto out;
3199 	}
3200 
3201 	ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3202 	btrfs_release_path(path);
3203 
3204 	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3205 		ret = -ENOENT;
3206 	if (ret < 0)
3207 		goto out;
3208 
3209 	extent_item_pos = loi->logical - key.objectid;
3210 	ret = iterate_extent_inodes(root->fs_info, key.objectid,
3211 					extent_item_pos, 0, build_ino_list,
3212 					inodes);
3213 
3214 	if (ret < 0)
3215 		goto out;
3216 
3217 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3218 			   (void *)(unsigned long)inodes, size);
3219 	if (ret)
3220 		ret = -EFAULT;
3221 
3222 out:
3223 	btrfs_free_path(path);
3224 	kfree(inodes);
3225 	kfree(loi);
3226 
3227 	return ret;
3228 }
3229 
3230 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3231 			       struct btrfs_ioctl_balance_args *bargs)
3232 {
3233 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3234 
3235 	bargs->flags = bctl->flags;
3236 
3237 	if (atomic_read(&fs_info->balance_running))
3238 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3239 	if (atomic_read(&fs_info->balance_pause_req))
3240 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3241 	if (atomic_read(&fs_info->balance_cancel_req))
3242 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3243 
3244 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3245 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3246 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3247 
3248 	if (lock) {
3249 		spin_lock(&fs_info->balance_lock);
3250 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3251 		spin_unlock(&fs_info->balance_lock);
3252 	} else {
3253 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3254 	}
3255 }
3256 
3257 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3258 {
3259 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3260 	struct btrfs_fs_info *fs_info = root->fs_info;
3261 	struct btrfs_ioctl_balance_args *bargs;
3262 	struct btrfs_balance_control *bctl;
3263 	int ret;
3264 
3265 	if (!capable(CAP_SYS_ADMIN))
3266 		return -EPERM;
3267 
3268 	if (fs_info->sb->s_flags & MS_RDONLY)
3269 		return -EROFS;
3270 
3271 	ret = mnt_want_write(file->f_path.mnt);
3272 	if (ret)
3273 		return ret;
3274 
3275 	mutex_lock(&fs_info->volume_mutex);
3276 	mutex_lock(&fs_info->balance_mutex);
3277 
3278 	if (arg) {
3279 		bargs = memdup_user(arg, sizeof(*bargs));
3280 		if (IS_ERR(bargs)) {
3281 			ret = PTR_ERR(bargs);
3282 			goto out;
3283 		}
3284 
3285 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3286 			if (!fs_info->balance_ctl) {
3287 				ret = -ENOTCONN;
3288 				goto out_bargs;
3289 			}
3290 
3291 			bctl = fs_info->balance_ctl;
3292 			spin_lock(&fs_info->balance_lock);
3293 			bctl->flags |= BTRFS_BALANCE_RESUME;
3294 			spin_unlock(&fs_info->balance_lock);
3295 
3296 			goto do_balance;
3297 		}
3298 	} else {
3299 		bargs = NULL;
3300 	}
3301 
3302 	if (fs_info->balance_ctl) {
3303 		ret = -EINPROGRESS;
3304 		goto out_bargs;
3305 	}
3306 
3307 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3308 	if (!bctl) {
3309 		ret = -ENOMEM;
3310 		goto out_bargs;
3311 	}
3312 
3313 	bctl->fs_info = fs_info;
3314 	if (arg) {
3315 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3316 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3317 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3318 
3319 		bctl->flags = bargs->flags;
3320 	} else {
3321 		/* balance everything - no filters */
3322 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3323 	}
3324 
3325 do_balance:
3326 	ret = btrfs_balance(bctl, bargs);
3327 	/*
3328 	 * bctl is freed in __cancel_balance or in free_fs_info if
3329 	 * restriper was paused all the way until unmount
3330 	 */
3331 	if (arg) {
3332 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3333 			ret = -EFAULT;
3334 	}
3335 
3336 out_bargs:
3337 	kfree(bargs);
3338 out:
3339 	mutex_unlock(&fs_info->balance_mutex);
3340 	mutex_unlock(&fs_info->volume_mutex);
3341 	mnt_drop_write(file->f_path.mnt);
3342 	return ret;
3343 }
3344 
3345 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3346 {
3347 	if (!capable(CAP_SYS_ADMIN))
3348 		return -EPERM;
3349 
3350 	switch (cmd) {
3351 	case BTRFS_BALANCE_CTL_PAUSE:
3352 		return btrfs_pause_balance(root->fs_info);
3353 	case BTRFS_BALANCE_CTL_CANCEL:
3354 		return btrfs_cancel_balance(root->fs_info);
3355 	}
3356 
3357 	return -EINVAL;
3358 }
3359 
3360 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3361 					 void __user *arg)
3362 {
3363 	struct btrfs_fs_info *fs_info = root->fs_info;
3364 	struct btrfs_ioctl_balance_args *bargs;
3365 	int ret = 0;
3366 
3367 	if (!capable(CAP_SYS_ADMIN))
3368 		return -EPERM;
3369 
3370 	mutex_lock(&fs_info->balance_mutex);
3371 	if (!fs_info->balance_ctl) {
3372 		ret = -ENOTCONN;
3373 		goto out;
3374 	}
3375 
3376 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3377 	if (!bargs) {
3378 		ret = -ENOMEM;
3379 		goto out;
3380 	}
3381 
3382 	update_ioctl_balance_args(fs_info, 1, bargs);
3383 
3384 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3385 		ret = -EFAULT;
3386 
3387 	kfree(bargs);
3388 out:
3389 	mutex_unlock(&fs_info->balance_mutex);
3390 	return ret;
3391 }
3392 
3393 long btrfs_ioctl(struct file *file, unsigned int
3394 		cmd, unsigned long arg)
3395 {
3396 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3397 	void __user *argp = (void __user *)arg;
3398 
3399 	switch (cmd) {
3400 	case FS_IOC_GETFLAGS:
3401 		return btrfs_ioctl_getflags(file, argp);
3402 	case FS_IOC_SETFLAGS:
3403 		return btrfs_ioctl_setflags(file, argp);
3404 	case FS_IOC_GETVERSION:
3405 		return btrfs_ioctl_getversion(file, argp);
3406 	case FITRIM:
3407 		return btrfs_ioctl_fitrim(file, argp);
3408 	case BTRFS_IOC_SNAP_CREATE:
3409 		return btrfs_ioctl_snap_create(file, argp, 0);
3410 	case BTRFS_IOC_SNAP_CREATE_V2:
3411 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3412 	case BTRFS_IOC_SUBVOL_CREATE:
3413 		return btrfs_ioctl_snap_create(file, argp, 1);
3414 	case BTRFS_IOC_SNAP_DESTROY:
3415 		return btrfs_ioctl_snap_destroy(file, argp);
3416 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3417 		return btrfs_ioctl_subvol_getflags(file, argp);
3418 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3419 		return btrfs_ioctl_subvol_setflags(file, argp);
3420 	case BTRFS_IOC_DEFAULT_SUBVOL:
3421 		return btrfs_ioctl_default_subvol(file, argp);
3422 	case BTRFS_IOC_DEFRAG:
3423 		return btrfs_ioctl_defrag(file, NULL);
3424 	case BTRFS_IOC_DEFRAG_RANGE:
3425 		return btrfs_ioctl_defrag(file, argp);
3426 	case BTRFS_IOC_RESIZE:
3427 		return btrfs_ioctl_resize(root, argp);
3428 	case BTRFS_IOC_ADD_DEV:
3429 		return btrfs_ioctl_add_dev(root, argp);
3430 	case BTRFS_IOC_RM_DEV:
3431 		return btrfs_ioctl_rm_dev(root, argp);
3432 	case BTRFS_IOC_FS_INFO:
3433 		return btrfs_ioctl_fs_info(root, argp);
3434 	case BTRFS_IOC_DEV_INFO:
3435 		return btrfs_ioctl_dev_info(root, argp);
3436 	case BTRFS_IOC_BALANCE:
3437 		return btrfs_ioctl_balance(file, NULL);
3438 	case BTRFS_IOC_CLONE:
3439 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3440 	case BTRFS_IOC_CLONE_RANGE:
3441 		return btrfs_ioctl_clone_range(file, argp);
3442 	case BTRFS_IOC_TRANS_START:
3443 		return btrfs_ioctl_trans_start(file);
3444 	case BTRFS_IOC_TRANS_END:
3445 		return btrfs_ioctl_trans_end(file);
3446 	case BTRFS_IOC_TREE_SEARCH:
3447 		return btrfs_ioctl_tree_search(file, argp);
3448 	case BTRFS_IOC_INO_LOOKUP:
3449 		return btrfs_ioctl_ino_lookup(file, argp);
3450 	case BTRFS_IOC_INO_PATHS:
3451 		return btrfs_ioctl_ino_to_path(root, argp);
3452 	case BTRFS_IOC_LOGICAL_INO:
3453 		return btrfs_ioctl_logical_to_ino(root, argp);
3454 	case BTRFS_IOC_SPACE_INFO:
3455 		return btrfs_ioctl_space_info(root, argp);
3456 	case BTRFS_IOC_SYNC:
3457 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3458 		return 0;
3459 	case BTRFS_IOC_START_SYNC:
3460 		return btrfs_ioctl_start_sync(file, argp);
3461 	case BTRFS_IOC_WAIT_SYNC:
3462 		return btrfs_ioctl_wait_sync(file, argp);
3463 	case BTRFS_IOC_SCRUB:
3464 		return btrfs_ioctl_scrub(root, argp);
3465 	case BTRFS_IOC_SCRUB_CANCEL:
3466 		return btrfs_ioctl_scrub_cancel(root, argp);
3467 	case BTRFS_IOC_SCRUB_PROGRESS:
3468 		return btrfs_ioctl_scrub_progress(root, argp);
3469 	case BTRFS_IOC_BALANCE_V2:
3470 		return btrfs_ioctl_balance(file, argp);
3471 	case BTRFS_IOC_BALANCE_CTL:
3472 		return btrfs_ioctl_balance_ctl(root, arg);
3473 	case BTRFS_IOC_BALANCE_PROGRESS:
3474 		return btrfs_ioctl_balance_progress(root, argp);
3475 	case BTRFS_IOC_GET_DEV_STATS:
3476 		return btrfs_ioctl_get_dev_stats(root, argp, 0);
3477 	case BTRFS_IOC_GET_AND_RESET_DEV_STATS:
3478 		return btrfs_ioctl_get_dev_stats(root, argp, 1);
3479 	}
3480 
3481 	return -ENOTTY;
3482 }
3483