xref: /openbmc/linux/fs/btrfs/ioctl.c (revision 63dc02bd)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include "compat.h"
45 #include "ctree.h"
46 #include "disk-io.h"
47 #include "transaction.h"
48 #include "btrfs_inode.h"
49 #include "ioctl.h"
50 #include "print-tree.h"
51 #include "volumes.h"
52 #include "locking.h"
53 #include "inode-map.h"
54 #include "backref.h"
55 
56 /* Mask out flags that are inappropriate for the given type of inode. */
57 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
58 {
59 	if (S_ISDIR(mode))
60 		return flags;
61 	else if (S_ISREG(mode))
62 		return flags & ~FS_DIRSYNC_FL;
63 	else
64 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
65 }
66 
67 /*
68  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
69  */
70 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
71 {
72 	unsigned int iflags = 0;
73 
74 	if (flags & BTRFS_INODE_SYNC)
75 		iflags |= FS_SYNC_FL;
76 	if (flags & BTRFS_INODE_IMMUTABLE)
77 		iflags |= FS_IMMUTABLE_FL;
78 	if (flags & BTRFS_INODE_APPEND)
79 		iflags |= FS_APPEND_FL;
80 	if (flags & BTRFS_INODE_NODUMP)
81 		iflags |= FS_NODUMP_FL;
82 	if (flags & BTRFS_INODE_NOATIME)
83 		iflags |= FS_NOATIME_FL;
84 	if (flags & BTRFS_INODE_DIRSYNC)
85 		iflags |= FS_DIRSYNC_FL;
86 	if (flags & BTRFS_INODE_NODATACOW)
87 		iflags |= FS_NOCOW_FL;
88 
89 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
90 		iflags |= FS_COMPR_FL;
91 	else if (flags & BTRFS_INODE_NOCOMPRESS)
92 		iflags |= FS_NOCOMP_FL;
93 
94 	return iflags;
95 }
96 
97 /*
98  * Update inode->i_flags based on the btrfs internal flags.
99  */
100 void btrfs_update_iflags(struct inode *inode)
101 {
102 	struct btrfs_inode *ip = BTRFS_I(inode);
103 
104 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
105 
106 	if (ip->flags & BTRFS_INODE_SYNC)
107 		inode->i_flags |= S_SYNC;
108 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
109 		inode->i_flags |= S_IMMUTABLE;
110 	if (ip->flags & BTRFS_INODE_APPEND)
111 		inode->i_flags |= S_APPEND;
112 	if (ip->flags & BTRFS_INODE_NOATIME)
113 		inode->i_flags |= S_NOATIME;
114 	if (ip->flags & BTRFS_INODE_DIRSYNC)
115 		inode->i_flags |= S_DIRSYNC;
116 }
117 
118 /*
119  * Inherit flags from the parent inode.
120  *
121  * Currently only the compression flags and the cow flags are inherited.
122  */
123 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
124 {
125 	unsigned int flags;
126 
127 	if (!dir)
128 		return;
129 
130 	flags = BTRFS_I(dir)->flags;
131 
132 	if (flags & BTRFS_INODE_NOCOMPRESS) {
133 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
134 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
135 	} else if (flags & BTRFS_INODE_COMPRESS) {
136 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
137 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
138 	}
139 
140 	if (flags & BTRFS_INODE_NODATACOW)
141 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
142 
143 	btrfs_update_iflags(inode);
144 }
145 
146 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
147 {
148 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
149 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
150 
151 	if (copy_to_user(arg, &flags, sizeof(flags)))
152 		return -EFAULT;
153 	return 0;
154 }
155 
156 static int check_flags(unsigned int flags)
157 {
158 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
159 		      FS_NOATIME_FL | FS_NODUMP_FL | \
160 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
161 		      FS_NOCOMP_FL | FS_COMPR_FL |
162 		      FS_NOCOW_FL))
163 		return -EOPNOTSUPP;
164 
165 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
166 		return -EINVAL;
167 
168 	return 0;
169 }
170 
171 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
172 {
173 	struct inode *inode = file->f_path.dentry->d_inode;
174 	struct btrfs_inode *ip = BTRFS_I(inode);
175 	struct btrfs_root *root = ip->root;
176 	struct btrfs_trans_handle *trans;
177 	unsigned int flags, oldflags;
178 	int ret;
179 	u64 ip_oldflags;
180 	unsigned int i_oldflags;
181 
182 	if (btrfs_root_readonly(root))
183 		return -EROFS;
184 
185 	if (copy_from_user(&flags, arg, sizeof(flags)))
186 		return -EFAULT;
187 
188 	ret = check_flags(flags);
189 	if (ret)
190 		return ret;
191 
192 	if (!inode_owner_or_capable(inode))
193 		return -EACCES;
194 
195 	mutex_lock(&inode->i_mutex);
196 
197 	ip_oldflags = ip->flags;
198 	i_oldflags = inode->i_flags;
199 
200 	flags = btrfs_mask_flags(inode->i_mode, flags);
201 	oldflags = btrfs_flags_to_ioctl(ip->flags);
202 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
203 		if (!capable(CAP_LINUX_IMMUTABLE)) {
204 			ret = -EPERM;
205 			goto out_unlock;
206 		}
207 	}
208 
209 	ret = mnt_want_write_file(file);
210 	if (ret)
211 		goto out_unlock;
212 
213 	if (flags & FS_SYNC_FL)
214 		ip->flags |= BTRFS_INODE_SYNC;
215 	else
216 		ip->flags &= ~BTRFS_INODE_SYNC;
217 	if (flags & FS_IMMUTABLE_FL)
218 		ip->flags |= BTRFS_INODE_IMMUTABLE;
219 	else
220 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
221 	if (flags & FS_APPEND_FL)
222 		ip->flags |= BTRFS_INODE_APPEND;
223 	else
224 		ip->flags &= ~BTRFS_INODE_APPEND;
225 	if (flags & FS_NODUMP_FL)
226 		ip->flags |= BTRFS_INODE_NODUMP;
227 	else
228 		ip->flags &= ~BTRFS_INODE_NODUMP;
229 	if (flags & FS_NOATIME_FL)
230 		ip->flags |= BTRFS_INODE_NOATIME;
231 	else
232 		ip->flags &= ~BTRFS_INODE_NOATIME;
233 	if (flags & FS_DIRSYNC_FL)
234 		ip->flags |= BTRFS_INODE_DIRSYNC;
235 	else
236 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
237 	if (flags & FS_NOCOW_FL)
238 		ip->flags |= BTRFS_INODE_NODATACOW;
239 	else
240 		ip->flags &= ~BTRFS_INODE_NODATACOW;
241 
242 	/*
243 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
244 	 * flag may be changed automatically if compression code won't make
245 	 * things smaller.
246 	 */
247 	if (flags & FS_NOCOMP_FL) {
248 		ip->flags &= ~BTRFS_INODE_COMPRESS;
249 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
250 	} else if (flags & FS_COMPR_FL) {
251 		ip->flags |= BTRFS_INODE_COMPRESS;
252 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
253 	} else {
254 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
255 	}
256 
257 	trans = btrfs_start_transaction(root, 1);
258 	if (IS_ERR(trans)) {
259 		ret = PTR_ERR(trans);
260 		goto out_drop;
261 	}
262 
263 	btrfs_update_iflags(inode);
264 	inode->i_ctime = CURRENT_TIME;
265 	ret = btrfs_update_inode(trans, root, inode);
266 
267 	btrfs_end_transaction(trans, root);
268  out_drop:
269 	if (ret) {
270 		ip->flags = ip_oldflags;
271 		inode->i_flags = i_oldflags;
272 	}
273 
274 	mnt_drop_write_file(file);
275  out_unlock:
276 	mutex_unlock(&inode->i_mutex);
277 	return ret;
278 }
279 
280 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
281 {
282 	struct inode *inode = file->f_path.dentry->d_inode;
283 
284 	return put_user(inode->i_generation, arg);
285 }
286 
287 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
288 {
289 	struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
290 	struct btrfs_device *device;
291 	struct request_queue *q;
292 	struct fstrim_range range;
293 	u64 minlen = ULLONG_MAX;
294 	u64 num_devices = 0;
295 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
296 	int ret;
297 
298 	if (!capable(CAP_SYS_ADMIN))
299 		return -EPERM;
300 
301 	rcu_read_lock();
302 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
303 				dev_list) {
304 		if (!device->bdev)
305 			continue;
306 		q = bdev_get_queue(device->bdev);
307 		if (blk_queue_discard(q)) {
308 			num_devices++;
309 			minlen = min((u64)q->limits.discard_granularity,
310 				     minlen);
311 		}
312 	}
313 	rcu_read_unlock();
314 
315 	if (!num_devices)
316 		return -EOPNOTSUPP;
317 	if (copy_from_user(&range, arg, sizeof(range)))
318 		return -EFAULT;
319 	if (range.start > total_bytes)
320 		return -EINVAL;
321 
322 	range.len = min(range.len, total_bytes - range.start);
323 	range.minlen = max(range.minlen, minlen);
324 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
325 	if (ret < 0)
326 		return ret;
327 
328 	if (copy_to_user(arg, &range, sizeof(range)))
329 		return -EFAULT;
330 
331 	return 0;
332 }
333 
334 static noinline int create_subvol(struct btrfs_root *root,
335 				  struct dentry *dentry,
336 				  char *name, int namelen,
337 				  u64 *async_transid)
338 {
339 	struct btrfs_trans_handle *trans;
340 	struct btrfs_key key;
341 	struct btrfs_root_item root_item;
342 	struct btrfs_inode_item *inode_item;
343 	struct extent_buffer *leaf;
344 	struct btrfs_root *new_root;
345 	struct dentry *parent = dentry->d_parent;
346 	struct inode *dir;
347 	int ret;
348 	int err;
349 	u64 objectid;
350 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
351 	u64 index = 0;
352 
353 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
354 	if (ret)
355 		return ret;
356 
357 	dir = parent->d_inode;
358 
359 	/*
360 	 * 1 - inode item
361 	 * 2 - refs
362 	 * 1 - root item
363 	 * 2 - dir items
364 	 */
365 	trans = btrfs_start_transaction(root, 6);
366 	if (IS_ERR(trans))
367 		return PTR_ERR(trans);
368 
369 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
370 				      0, objectid, NULL, 0, 0, 0, 0);
371 	if (IS_ERR(leaf)) {
372 		ret = PTR_ERR(leaf);
373 		goto fail;
374 	}
375 
376 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
377 	btrfs_set_header_bytenr(leaf, leaf->start);
378 	btrfs_set_header_generation(leaf, trans->transid);
379 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
380 	btrfs_set_header_owner(leaf, objectid);
381 
382 	write_extent_buffer(leaf, root->fs_info->fsid,
383 			    (unsigned long)btrfs_header_fsid(leaf),
384 			    BTRFS_FSID_SIZE);
385 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
386 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
387 			    BTRFS_UUID_SIZE);
388 	btrfs_mark_buffer_dirty(leaf);
389 
390 	inode_item = &root_item.inode;
391 	memset(inode_item, 0, sizeof(*inode_item));
392 	inode_item->generation = cpu_to_le64(1);
393 	inode_item->size = cpu_to_le64(3);
394 	inode_item->nlink = cpu_to_le32(1);
395 	inode_item->nbytes = cpu_to_le64(root->leafsize);
396 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
397 
398 	root_item.flags = 0;
399 	root_item.byte_limit = 0;
400 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
401 
402 	btrfs_set_root_bytenr(&root_item, leaf->start);
403 	btrfs_set_root_generation(&root_item, trans->transid);
404 	btrfs_set_root_level(&root_item, 0);
405 	btrfs_set_root_refs(&root_item, 1);
406 	btrfs_set_root_used(&root_item, leaf->len);
407 	btrfs_set_root_last_snapshot(&root_item, 0);
408 
409 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
410 	root_item.drop_level = 0;
411 
412 	btrfs_tree_unlock(leaf);
413 	free_extent_buffer(leaf);
414 	leaf = NULL;
415 
416 	btrfs_set_root_dirid(&root_item, new_dirid);
417 
418 	key.objectid = objectid;
419 	key.offset = 0;
420 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
421 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
422 				&root_item);
423 	if (ret)
424 		goto fail;
425 
426 	key.offset = (u64)-1;
427 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
428 	if (IS_ERR(new_root)) {
429 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
430 		ret = PTR_ERR(new_root);
431 		goto fail;
432 	}
433 
434 	btrfs_record_root_in_trans(trans, new_root);
435 
436 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
437 	if (ret) {
438 		/* We potentially lose an unused inode item here */
439 		btrfs_abort_transaction(trans, root, ret);
440 		goto fail;
441 	}
442 
443 	/*
444 	 * insert the directory item
445 	 */
446 	ret = btrfs_set_inode_index(dir, &index);
447 	if (ret) {
448 		btrfs_abort_transaction(trans, root, ret);
449 		goto fail;
450 	}
451 
452 	ret = btrfs_insert_dir_item(trans, root,
453 				    name, namelen, dir, &key,
454 				    BTRFS_FT_DIR, index);
455 	if (ret) {
456 		btrfs_abort_transaction(trans, root, ret);
457 		goto fail;
458 	}
459 
460 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
461 	ret = btrfs_update_inode(trans, root, dir);
462 	BUG_ON(ret);
463 
464 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
465 				 objectid, root->root_key.objectid,
466 				 btrfs_ino(dir), index, name, namelen);
467 
468 	BUG_ON(ret);
469 
470 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
471 fail:
472 	if (async_transid) {
473 		*async_transid = trans->transid;
474 		err = btrfs_commit_transaction_async(trans, root, 1);
475 	} else {
476 		err = btrfs_commit_transaction(trans, root);
477 	}
478 	if (err && !ret)
479 		ret = err;
480 	return ret;
481 }
482 
483 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
484 			   char *name, int namelen, u64 *async_transid,
485 			   bool readonly)
486 {
487 	struct inode *inode;
488 	struct btrfs_pending_snapshot *pending_snapshot;
489 	struct btrfs_trans_handle *trans;
490 	int ret;
491 
492 	if (!root->ref_cows)
493 		return -EINVAL;
494 
495 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
496 	if (!pending_snapshot)
497 		return -ENOMEM;
498 
499 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
500 	pending_snapshot->dentry = dentry;
501 	pending_snapshot->root = root;
502 	pending_snapshot->readonly = readonly;
503 
504 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
505 	if (IS_ERR(trans)) {
506 		ret = PTR_ERR(trans);
507 		goto fail;
508 	}
509 
510 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
511 	BUG_ON(ret);
512 
513 	spin_lock(&root->fs_info->trans_lock);
514 	list_add(&pending_snapshot->list,
515 		 &trans->transaction->pending_snapshots);
516 	spin_unlock(&root->fs_info->trans_lock);
517 	if (async_transid) {
518 		*async_transid = trans->transid;
519 		ret = btrfs_commit_transaction_async(trans,
520 				     root->fs_info->extent_root, 1);
521 	} else {
522 		ret = btrfs_commit_transaction(trans,
523 					       root->fs_info->extent_root);
524 	}
525 	BUG_ON(ret);
526 
527 	ret = pending_snapshot->error;
528 	if (ret)
529 		goto fail;
530 
531 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
532 	if (ret)
533 		goto fail;
534 
535 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
536 	if (IS_ERR(inode)) {
537 		ret = PTR_ERR(inode);
538 		goto fail;
539 	}
540 	BUG_ON(!inode);
541 	d_instantiate(dentry, inode);
542 	ret = 0;
543 fail:
544 	kfree(pending_snapshot);
545 	return ret;
546 }
547 
548 /*  copy of check_sticky in fs/namei.c()
549 * It's inline, so penalty for filesystems that don't use sticky bit is
550 * minimal.
551 */
552 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
553 {
554 	uid_t fsuid = current_fsuid();
555 
556 	if (!(dir->i_mode & S_ISVTX))
557 		return 0;
558 	if (inode->i_uid == fsuid)
559 		return 0;
560 	if (dir->i_uid == fsuid)
561 		return 0;
562 	return !capable(CAP_FOWNER);
563 }
564 
565 /*  copy of may_delete in fs/namei.c()
566  *	Check whether we can remove a link victim from directory dir, check
567  *  whether the type of victim is right.
568  *  1. We can't do it if dir is read-only (done in permission())
569  *  2. We should have write and exec permissions on dir
570  *  3. We can't remove anything from append-only dir
571  *  4. We can't do anything with immutable dir (done in permission())
572  *  5. If the sticky bit on dir is set we should either
573  *	a. be owner of dir, or
574  *	b. be owner of victim, or
575  *	c. have CAP_FOWNER capability
576  *  6. If the victim is append-only or immutable we can't do antyhing with
577  *     links pointing to it.
578  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
579  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
580  *  9. We can't remove a root or mountpoint.
581  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
582  *     nfs_async_unlink().
583  */
584 
585 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
586 {
587 	int error;
588 
589 	if (!victim->d_inode)
590 		return -ENOENT;
591 
592 	BUG_ON(victim->d_parent->d_inode != dir);
593 	audit_inode_child(victim, dir);
594 
595 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
596 	if (error)
597 		return error;
598 	if (IS_APPEND(dir))
599 		return -EPERM;
600 	if (btrfs_check_sticky(dir, victim->d_inode)||
601 		IS_APPEND(victim->d_inode)||
602 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
603 		return -EPERM;
604 	if (isdir) {
605 		if (!S_ISDIR(victim->d_inode->i_mode))
606 			return -ENOTDIR;
607 		if (IS_ROOT(victim))
608 			return -EBUSY;
609 	} else if (S_ISDIR(victim->d_inode->i_mode))
610 		return -EISDIR;
611 	if (IS_DEADDIR(dir))
612 		return -ENOENT;
613 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
614 		return -EBUSY;
615 	return 0;
616 }
617 
618 /* copy of may_create in fs/namei.c() */
619 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
620 {
621 	if (child->d_inode)
622 		return -EEXIST;
623 	if (IS_DEADDIR(dir))
624 		return -ENOENT;
625 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
626 }
627 
628 /*
629  * Create a new subvolume below @parent.  This is largely modeled after
630  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
631  * inside this filesystem so it's quite a bit simpler.
632  */
633 static noinline int btrfs_mksubvol(struct path *parent,
634 				   char *name, int namelen,
635 				   struct btrfs_root *snap_src,
636 				   u64 *async_transid, bool readonly)
637 {
638 	struct inode *dir  = parent->dentry->d_inode;
639 	struct dentry *dentry;
640 	int error;
641 
642 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
643 
644 	dentry = lookup_one_len(name, parent->dentry, namelen);
645 	error = PTR_ERR(dentry);
646 	if (IS_ERR(dentry))
647 		goto out_unlock;
648 
649 	error = -EEXIST;
650 	if (dentry->d_inode)
651 		goto out_dput;
652 
653 	error = mnt_want_write(parent->mnt);
654 	if (error)
655 		goto out_dput;
656 
657 	error = btrfs_may_create(dir, dentry);
658 	if (error)
659 		goto out_drop_write;
660 
661 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
662 
663 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
664 		goto out_up_read;
665 
666 	if (snap_src) {
667 		error = create_snapshot(snap_src, dentry,
668 					name, namelen, async_transid, readonly);
669 	} else {
670 		error = create_subvol(BTRFS_I(dir)->root, dentry,
671 				      name, namelen, async_transid);
672 	}
673 	if (!error)
674 		fsnotify_mkdir(dir, dentry);
675 out_up_read:
676 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
677 out_drop_write:
678 	mnt_drop_write(parent->mnt);
679 out_dput:
680 	dput(dentry);
681 out_unlock:
682 	mutex_unlock(&dir->i_mutex);
683 	return error;
684 }
685 
686 /*
687  * When we're defragging a range, we don't want to kick it off again
688  * if it is really just waiting for delalloc to send it down.
689  * If we find a nice big extent or delalloc range for the bytes in the
690  * file you want to defrag, we return 0 to let you know to skip this
691  * part of the file
692  */
693 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
694 {
695 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
696 	struct extent_map *em = NULL;
697 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
698 	u64 end;
699 
700 	read_lock(&em_tree->lock);
701 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
702 	read_unlock(&em_tree->lock);
703 
704 	if (em) {
705 		end = extent_map_end(em);
706 		free_extent_map(em);
707 		if (end - offset > thresh)
708 			return 0;
709 	}
710 	/* if we already have a nice delalloc here, just stop */
711 	thresh /= 2;
712 	end = count_range_bits(io_tree, &offset, offset + thresh,
713 			       thresh, EXTENT_DELALLOC, 1);
714 	if (end >= thresh)
715 		return 0;
716 	return 1;
717 }
718 
719 /*
720  * helper function to walk through a file and find extents
721  * newer than a specific transid, and smaller than thresh.
722  *
723  * This is used by the defragging code to find new and small
724  * extents
725  */
726 static int find_new_extents(struct btrfs_root *root,
727 			    struct inode *inode, u64 newer_than,
728 			    u64 *off, int thresh)
729 {
730 	struct btrfs_path *path;
731 	struct btrfs_key min_key;
732 	struct btrfs_key max_key;
733 	struct extent_buffer *leaf;
734 	struct btrfs_file_extent_item *extent;
735 	int type;
736 	int ret;
737 	u64 ino = btrfs_ino(inode);
738 
739 	path = btrfs_alloc_path();
740 	if (!path)
741 		return -ENOMEM;
742 
743 	min_key.objectid = ino;
744 	min_key.type = BTRFS_EXTENT_DATA_KEY;
745 	min_key.offset = *off;
746 
747 	max_key.objectid = ino;
748 	max_key.type = (u8)-1;
749 	max_key.offset = (u64)-1;
750 
751 	path->keep_locks = 1;
752 
753 	while(1) {
754 		ret = btrfs_search_forward(root, &min_key, &max_key,
755 					   path, 0, newer_than);
756 		if (ret != 0)
757 			goto none;
758 		if (min_key.objectid != ino)
759 			goto none;
760 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
761 			goto none;
762 
763 		leaf = path->nodes[0];
764 		extent = btrfs_item_ptr(leaf, path->slots[0],
765 					struct btrfs_file_extent_item);
766 
767 		type = btrfs_file_extent_type(leaf, extent);
768 		if (type == BTRFS_FILE_EXTENT_REG &&
769 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
770 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
771 			*off = min_key.offset;
772 			btrfs_free_path(path);
773 			return 0;
774 		}
775 
776 		if (min_key.offset == (u64)-1)
777 			goto none;
778 
779 		min_key.offset++;
780 		btrfs_release_path(path);
781 	}
782 none:
783 	btrfs_free_path(path);
784 	return -ENOENT;
785 }
786 
787 /*
788  * Validaty check of prev em and next em:
789  * 1) no prev/next em
790  * 2) prev/next em is an hole/inline extent
791  */
792 static int check_adjacent_extents(struct inode *inode, struct extent_map *em)
793 {
794 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
795 	struct extent_map *prev = NULL, *next = NULL;
796 	int ret = 0;
797 
798 	read_lock(&em_tree->lock);
799 	prev = lookup_extent_mapping(em_tree, em->start - 1, (u64)-1);
800 	next = lookup_extent_mapping(em_tree, em->start + em->len, (u64)-1);
801 	read_unlock(&em_tree->lock);
802 
803 	if ((!prev || prev->block_start >= EXTENT_MAP_LAST_BYTE) &&
804 	    (!next || next->block_start >= EXTENT_MAP_LAST_BYTE))
805 		ret = 1;
806 	free_extent_map(prev);
807 	free_extent_map(next);
808 
809 	return ret;
810 }
811 
812 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
813 			       int thresh, u64 *last_len, u64 *skip,
814 			       u64 *defrag_end)
815 {
816 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
817 	struct extent_map *em = NULL;
818 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
819 	int ret = 1;
820 
821 	/*
822 	 * make sure that once we start defragging an extent, we keep on
823 	 * defragging it
824 	 */
825 	if (start < *defrag_end)
826 		return 1;
827 
828 	*skip = 0;
829 
830 	/*
831 	 * hopefully we have this extent in the tree already, try without
832 	 * the full extent lock
833 	 */
834 	read_lock(&em_tree->lock);
835 	em = lookup_extent_mapping(em_tree, start, len);
836 	read_unlock(&em_tree->lock);
837 
838 	if (!em) {
839 		/* get the big lock and read metadata off disk */
840 		lock_extent(io_tree, start, start + len - 1);
841 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
842 		unlock_extent(io_tree, start, start + len - 1);
843 
844 		if (IS_ERR(em))
845 			return 0;
846 	}
847 
848 	/* this will cover holes, and inline extents */
849 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
850 		ret = 0;
851 		goto out;
852 	}
853 
854 	/* If we have nothing to merge with us, just skip. */
855 	if (check_adjacent_extents(inode, em)) {
856 		ret = 0;
857 		goto out;
858 	}
859 
860 	/*
861 	 * we hit a real extent, if it is big don't bother defragging it again
862 	 */
863 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
864 		ret = 0;
865 
866 out:
867 	/*
868 	 * last_len ends up being a counter of how many bytes we've defragged.
869 	 * every time we choose not to defrag an extent, we reset *last_len
870 	 * so that the next tiny extent will force a defrag.
871 	 *
872 	 * The end result of this is that tiny extents before a single big
873 	 * extent will force at least part of that big extent to be defragged.
874 	 */
875 	if (ret) {
876 		*defrag_end = extent_map_end(em);
877 	} else {
878 		*last_len = 0;
879 		*skip = extent_map_end(em);
880 		*defrag_end = 0;
881 	}
882 
883 	free_extent_map(em);
884 	return ret;
885 }
886 
887 /*
888  * it doesn't do much good to defrag one or two pages
889  * at a time.  This pulls in a nice chunk of pages
890  * to COW and defrag.
891  *
892  * It also makes sure the delalloc code has enough
893  * dirty data to avoid making new small extents as part
894  * of the defrag
895  *
896  * It's a good idea to start RA on this range
897  * before calling this.
898  */
899 static int cluster_pages_for_defrag(struct inode *inode,
900 				    struct page **pages,
901 				    unsigned long start_index,
902 				    int num_pages)
903 {
904 	unsigned long file_end;
905 	u64 isize = i_size_read(inode);
906 	u64 page_start;
907 	u64 page_end;
908 	u64 page_cnt;
909 	int ret;
910 	int i;
911 	int i_done;
912 	struct btrfs_ordered_extent *ordered;
913 	struct extent_state *cached_state = NULL;
914 	struct extent_io_tree *tree;
915 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
916 
917 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
918 	if (!isize || start_index > file_end)
919 		return 0;
920 
921 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
922 
923 	ret = btrfs_delalloc_reserve_space(inode,
924 					   page_cnt << PAGE_CACHE_SHIFT);
925 	if (ret)
926 		return ret;
927 	i_done = 0;
928 	tree = &BTRFS_I(inode)->io_tree;
929 
930 	/* step one, lock all the pages */
931 	for (i = 0; i < page_cnt; i++) {
932 		struct page *page;
933 again:
934 		page = find_or_create_page(inode->i_mapping,
935 					   start_index + i, mask);
936 		if (!page)
937 			break;
938 
939 		page_start = page_offset(page);
940 		page_end = page_start + PAGE_CACHE_SIZE - 1;
941 		while (1) {
942 			lock_extent(tree, page_start, page_end);
943 			ordered = btrfs_lookup_ordered_extent(inode,
944 							      page_start);
945 			unlock_extent(tree, page_start, page_end);
946 			if (!ordered)
947 				break;
948 
949 			unlock_page(page);
950 			btrfs_start_ordered_extent(inode, ordered, 1);
951 			btrfs_put_ordered_extent(ordered);
952 			lock_page(page);
953 			/*
954 			 * we unlocked the page above, so we need check if
955 			 * it was released or not.
956 			 */
957 			if (page->mapping != inode->i_mapping) {
958 				unlock_page(page);
959 				page_cache_release(page);
960 				goto again;
961 			}
962 		}
963 
964 		if (!PageUptodate(page)) {
965 			btrfs_readpage(NULL, page);
966 			lock_page(page);
967 			if (!PageUptodate(page)) {
968 				unlock_page(page);
969 				page_cache_release(page);
970 				ret = -EIO;
971 				break;
972 			}
973 		}
974 
975 		if (page->mapping != inode->i_mapping) {
976 			unlock_page(page);
977 			page_cache_release(page);
978 			goto again;
979 		}
980 
981 		pages[i] = page;
982 		i_done++;
983 	}
984 	if (!i_done || ret)
985 		goto out;
986 
987 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
988 		goto out;
989 
990 	/*
991 	 * so now we have a nice long stream of locked
992 	 * and up to date pages, lets wait on them
993 	 */
994 	for (i = 0; i < i_done; i++)
995 		wait_on_page_writeback(pages[i]);
996 
997 	page_start = page_offset(pages[0]);
998 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
999 
1000 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1001 			 page_start, page_end - 1, 0, &cached_state);
1002 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1003 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1004 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1005 			  GFP_NOFS);
1006 
1007 	if (i_done != page_cnt) {
1008 		spin_lock(&BTRFS_I(inode)->lock);
1009 		BTRFS_I(inode)->outstanding_extents++;
1010 		spin_unlock(&BTRFS_I(inode)->lock);
1011 		btrfs_delalloc_release_space(inode,
1012 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1013 	}
1014 
1015 
1016 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1017 				  &cached_state);
1018 
1019 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1020 			     page_start, page_end - 1, &cached_state,
1021 			     GFP_NOFS);
1022 
1023 	for (i = 0; i < i_done; i++) {
1024 		clear_page_dirty_for_io(pages[i]);
1025 		ClearPageChecked(pages[i]);
1026 		set_page_extent_mapped(pages[i]);
1027 		set_page_dirty(pages[i]);
1028 		unlock_page(pages[i]);
1029 		page_cache_release(pages[i]);
1030 	}
1031 	return i_done;
1032 out:
1033 	for (i = 0; i < i_done; i++) {
1034 		unlock_page(pages[i]);
1035 		page_cache_release(pages[i]);
1036 	}
1037 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1038 	return ret;
1039 
1040 }
1041 
1042 int btrfs_defrag_file(struct inode *inode, struct file *file,
1043 		      struct btrfs_ioctl_defrag_range_args *range,
1044 		      u64 newer_than, unsigned long max_to_defrag)
1045 {
1046 	struct btrfs_root *root = BTRFS_I(inode)->root;
1047 	struct btrfs_super_block *disk_super;
1048 	struct file_ra_state *ra = NULL;
1049 	unsigned long last_index;
1050 	u64 isize = i_size_read(inode);
1051 	u64 features;
1052 	u64 last_len = 0;
1053 	u64 skip = 0;
1054 	u64 defrag_end = 0;
1055 	u64 newer_off = range->start;
1056 	unsigned long i;
1057 	unsigned long ra_index = 0;
1058 	int ret;
1059 	int defrag_count = 0;
1060 	int compress_type = BTRFS_COMPRESS_ZLIB;
1061 	int extent_thresh = range->extent_thresh;
1062 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1063 	int cluster = max_cluster;
1064 	u64 new_align = ~((u64)128 * 1024 - 1);
1065 	struct page **pages = NULL;
1066 
1067 	if (extent_thresh == 0)
1068 		extent_thresh = 256 * 1024;
1069 
1070 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1071 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1072 			return -EINVAL;
1073 		if (range->compress_type)
1074 			compress_type = range->compress_type;
1075 	}
1076 
1077 	if (isize == 0)
1078 		return 0;
1079 
1080 	/*
1081 	 * if we were not given a file, allocate a readahead
1082 	 * context
1083 	 */
1084 	if (!file) {
1085 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1086 		if (!ra)
1087 			return -ENOMEM;
1088 		file_ra_state_init(ra, inode->i_mapping);
1089 	} else {
1090 		ra = &file->f_ra;
1091 	}
1092 
1093 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1094 			GFP_NOFS);
1095 	if (!pages) {
1096 		ret = -ENOMEM;
1097 		goto out_ra;
1098 	}
1099 
1100 	/* find the last page to defrag */
1101 	if (range->start + range->len > range->start) {
1102 		last_index = min_t(u64, isize - 1,
1103 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1104 	} else {
1105 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1106 	}
1107 
1108 	if (newer_than) {
1109 		ret = find_new_extents(root, inode, newer_than,
1110 				       &newer_off, 64 * 1024);
1111 		if (!ret) {
1112 			range->start = newer_off;
1113 			/*
1114 			 * we always align our defrag to help keep
1115 			 * the extents in the file evenly spaced
1116 			 */
1117 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1118 		} else
1119 			goto out_ra;
1120 	} else {
1121 		i = range->start >> PAGE_CACHE_SHIFT;
1122 	}
1123 	if (!max_to_defrag)
1124 		max_to_defrag = last_index + 1;
1125 
1126 	/*
1127 	 * make writeback starts from i, so the defrag range can be
1128 	 * written sequentially.
1129 	 */
1130 	if (i < inode->i_mapping->writeback_index)
1131 		inode->i_mapping->writeback_index = i;
1132 
1133 	while (i <= last_index && defrag_count < max_to_defrag &&
1134 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1135 		PAGE_CACHE_SHIFT)) {
1136 		/*
1137 		 * make sure we stop running if someone unmounts
1138 		 * the FS
1139 		 */
1140 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1141 			break;
1142 
1143 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1144 					 PAGE_CACHE_SIZE, extent_thresh,
1145 					 &last_len, &skip, &defrag_end)) {
1146 			unsigned long next;
1147 			/*
1148 			 * the should_defrag function tells us how much to skip
1149 			 * bump our counter by the suggested amount
1150 			 */
1151 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1152 			i = max(i + 1, next);
1153 			continue;
1154 		}
1155 
1156 		if (!newer_than) {
1157 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1158 				   PAGE_CACHE_SHIFT) - i;
1159 			cluster = min(cluster, max_cluster);
1160 		} else {
1161 			cluster = max_cluster;
1162 		}
1163 
1164 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1165 			BTRFS_I(inode)->force_compress = compress_type;
1166 
1167 		if (i + cluster > ra_index) {
1168 			ra_index = max(i, ra_index);
1169 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1170 				       cluster);
1171 			ra_index += max_cluster;
1172 		}
1173 
1174 		mutex_lock(&inode->i_mutex);
1175 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1176 		if (ret < 0) {
1177 			mutex_unlock(&inode->i_mutex);
1178 			goto out_ra;
1179 		}
1180 
1181 		defrag_count += ret;
1182 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1183 		mutex_unlock(&inode->i_mutex);
1184 
1185 		if (newer_than) {
1186 			if (newer_off == (u64)-1)
1187 				break;
1188 
1189 			if (ret > 0)
1190 				i += ret;
1191 
1192 			newer_off = max(newer_off + 1,
1193 					(u64)i << PAGE_CACHE_SHIFT);
1194 
1195 			ret = find_new_extents(root, inode,
1196 					       newer_than, &newer_off,
1197 					       64 * 1024);
1198 			if (!ret) {
1199 				range->start = newer_off;
1200 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1201 			} else {
1202 				break;
1203 			}
1204 		} else {
1205 			if (ret > 0) {
1206 				i += ret;
1207 				last_len += ret << PAGE_CACHE_SHIFT;
1208 			} else {
1209 				i++;
1210 				last_len = 0;
1211 			}
1212 		}
1213 	}
1214 
1215 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1216 		filemap_flush(inode->i_mapping);
1217 
1218 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1219 		/* the filemap_flush will queue IO into the worker threads, but
1220 		 * we have to make sure the IO is actually started and that
1221 		 * ordered extents get created before we return
1222 		 */
1223 		atomic_inc(&root->fs_info->async_submit_draining);
1224 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1225 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1226 			wait_event(root->fs_info->async_submit_wait,
1227 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1228 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1229 		}
1230 		atomic_dec(&root->fs_info->async_submit_draining);
1231 
1232 		mutex_lock(&inode->i_mutex);
1233 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1234 		mutex_unlock(&inode->i_mutex);
1235 	}
1236 
1237 	disk_super = root->fs_info->super_copy;
1238 	features = btrfs_super_incompat_flags(disk_super);
1239 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1240 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1241 		btrfs_set_super_incompat_flags(disk_super, features);
1242 	}
1243 
1244 	ret = defrag_count;
1245 
1246 out_ra:
1247 	if (!file)
1248 		kfree(ra);
1249 	kfree(pages);
1250 	return ret;
1251 }
1252 
1253 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1254 					void __user *arg)
1255 {
1256 	u64 new_size;
1257 	u64 old_size;
1258 	u64 devid = 1;
1259 	struct btrfs_ioctl_vol_args *vol_args;
1260 	struct btrfs_trans_handle *trans;
1261 	struct btrfs_device *device = NULL;
1262 	char *sizestr;
1263 	char *devstr = NULL;
1264 	int ret = 0;
1265 	int mod = 0;
1266 
1267 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1268 		return -EROFS;
1269 
1270 	if (!capable(CAP_SYS_ADMIN))
1271 		return -EPERM;
1272 
1273 	mutex_lock(&root->fs_info->volume_mutex);
1274 	if (root->fs_info->balance_ctl) {
1275 		printk(KERN_INFO "btrfs: balance in progress\n");
1276 		ret = -EINVAL;
1277 		goto out;
1278 	}
1279 
1280 	vol_args = memdup_user(arg, sizeof(*vol_args));
1281 	if (IS_ERR(vol_args)) {
1282 		ret = PTR_ERR(vol_args);
1283 		goto out;
1284 	}
1285 
1286 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1287 
1288 	sizestr = vol_args->name;
1289 	devstr = strchr(sizestr, ':');
1290 	if (devstr) {
1291 		char *end;
1292 		sizestr = devstr + 1;
1293 		*devstr = '\0';
1294 		devstr = vol_args->name;
1295 		devid = simple_strtoull(devstr, &end, 10);
1296 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1297 		       (unsigned long long)devid);
1298 	}
1299 	device = btrfs_find_device(root, devid, NULL, NULL);
1300 	if (!device) {
1301 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1302 		       (unsigned long long)devid);
1303 		ret = -EINVAL;
1304 		goto out_free;
1305 	}
1306 	if (!strcmp(sizestr, "max"))
1307 		new_size = device->bdev->bd_inode->i_size;
1308 	else {
1309 		if (sizestr[0] == '-') {
1310 			mod = -1;
1311 			sizestr++;
1312 		} else if (sizestr[0] == '+') {
1313 			mod = 1;
1314 			sizestr++;
1315 		}
1316 		new_size = memparse(sizestr, NULL);
1317 		if (new_size == 0) {
1318 			ret = -EINVAL;
1319 			goto out_free;
1320 		}
1321 	}
1322 
1323 	old_size = device->total_bytes;
1324 
1325 	if (mod < 0) {
1326 		if (new_size > old_size) {
1327 			ret = -EINVAL;
1328 			goto out_free;
1329 		}
1330 		new_size = old_size - new_size;
1331 	} else if (mod > 0) {
1332 		new_size = old_size + new_size;
1333 	}
1334 
1335 	if (new_size < 256 * 1024 * 1024) {
1336 		ret = -EINVAL;
1337 		goto out_free;
1338 	}
1339 	if (new_size > device->bdev->bd_inode->i_size) {
1340 		ret = -EFBIG;
1341 		goto out_free;
1342 	}
1343 
1344 	do_div(new_size, root->sectorsize);
1345 	new_size *= root->sectorsize;
1346 
1347 	printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1348 		device->name, (unsigned long long)new_size);
1349 
1350 	if (new_size > old_size) {
1351 		trans = btrfs_start_transaction(root, 0);
1352 		if (IS_ERR(trans)) {
1353 			ret = PTR_ERR(trans);
1354 			goto out_free;
1355 		}
1356 		ret = btrfs_grow_device(trans, device, new_size);
1357 		btrfs_commit_transaction(trans, root);
1358 	} else if (new_size < old_size) {
1359 		ret = btrfs_shrink_device(device, new_size);
1360 	}
1361 
1362 out_free:
1363 	kfree(vol_args);
1364 out:
1365 	mutex_unlock(&root->fs_info->volume_mutex);
1366 	return ret;
1367 }
1368 
1369 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1370 						    char *name,
1371 						    unsigned long fd,
1372 						    int subvol,
1373 						    u64 *transid,
1374 						    bool readonly)
1375 {
1376 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1377 	struct file *src_file;
1378 	int namelen;
1379 	int ret = 0;
1380 
1381 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1382 		return -EROFS;
1383 
1384 	namelen = strlen(name);
1385 	if (strchr(name, '/')) {
1386 		ret = -EINVAL;
1387 		goto out;
1388 	}
1389 
1390 	if (name[0] == '.' &&
1391 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1392 		ret = -EEXIST;
1393 		goto out;
1394 	}
1395 
1396 	if (subvol) {
1397 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1398 				     NULL, transid, readonly);
1399 	} else {
1400 		struct inode *src_inode;
1401 		src_file = fget(fd);
1402 		if (!src_file) {
1403 			ret = -EINVAL;
1404 			goto out;
1405 		}
1406 
1407 		src_inode = src_file->f_path.dentry->d_inode;
1408 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1409 			printk(KERN_INFO "btrfs: Snapshot src from "
1410 			       "another FS\n");
1411 			ret = -EINVAL;
1412 			fput(src_file);
1413 			goto out;
1414 		}
1415 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1416 				     BTRFS_I(src_inode)->root,
1417 				     transid, readonly);
1418 		fput(src_file);
1419 	}
1420 out:
1421 	return ret;
1422 }
1423 
1424 static noinline int btrfs_ioctl_snap_create(struct file *file,
1425 					    void __user *arg, int subvol)
1426 {
1427 	struct btrfs_ioctl_vol_args *vol_args;
1428 	int ret;
1429 
1430 	vol_args = memdup_user(arg, sizeof(*vol_args));
1431 	if (IS_ERR(vol_args))
1432 		return PTR_ERR(vol_args);
1433 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1434 
1435 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1436 					      vol_args->fd, subvol,
1437 					      NULL, false);
1438 
1439 	kfree(vol_args);
1440 	return ret;
1441 }
1442 
1443 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1444 					       void __user *arg, int subvol)
1445 {
1446 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1447 	int ret;
1448 	u64 transid = 0;
1449 	u64 *ptr = NULL;
1450 	bool readonly = false;
1451 
1452 	vol_args = memdup_user(arg, sizeof(*vol_args));
1453 	if (IS_ERR(vol_args))
1454 		return PTR_ERR(vol_args);
1455 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1456 
1457 	if (vol_args->flags &
1458 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1459 		ret = -EOPNOTSUPP;
1460 		goto out;
1461 	}
1462 
1463 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1464 		ptr = &transid;
1465 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1466 		readonly = true;
1467 
1468 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1469 					      vol_args->fd, subvol,
1470 					      ptr, readonly);
1471 
1472 	if (ret == 0 && ptr &&
1473 	    copy_to_user(arg +
1474 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1475 				  transid), ptr, sizeof(*ptr)))
1476 		ret = -EFAULT;
1477 out:
1478 	kfree(vol_args);
1479 	return ret;
1480 }
1481 
1482 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1483 						void __user *arg)
1484 {
1485 	struct inode *inode = fdentry(file)->d_inode;
1486 	struct btrfs_root *root = BTRFS_I(inode)->root;
1487 	int ret = 0;
1488 	u64 flags = 0;
1489 
1490 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1491 		return -EINVAL;
1492 
1493 	down_read(&root->fs_info->subvol_sem);
1494 	if (btrfs_root_readonly(root))
1495 		flags |= BTRFS_SUBVOL_RDONLY;
1496 	up_read(&root->fs_info->subvol_sem);
1497 
1498 	if (copy_to_user(arg, &flags, sizeof(flags)))
1499 		ret = -EFAULT;
1500 
1501 	return ret;
1502 }
1503 
1504 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1505 					      void __user *arg)
1506 {
1507 	struct inode *inode = fdentry(file)->d_inode;
1508 	struct btrfs_root *root = BTRFS_I(inode)->root;
1509 	struct btrfs_trans_handle *trans;
1510 	u64 root_flags;
1511 	u64 flags;
1512 	int ret = 0;
1513 
1514 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1515 		return -EROFS;
1516 
1517 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1518 		return -EINVAL;
1519 
1520 	if (copy_from_user(&flags, arg, sizeof(flags)))
1521 		return -EFAULT;
1522 
1523 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1524 		return -EINVAL;
1525 
1526 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1527 		return -EOPNOTSUPP;
1528 
1529 	if (!inode_owner_or_capable(inode))
1530 		return -EACCES;
1531 
1532 	down_write(&root->fs_info->subvol_sem);
1533 
1534 	/* nothing to do */
1535 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1536 		goto out;
1537 
1538 	root_flags = btrfs_root_flags(&root->root_item);
1539 	if (flags & BTRFS_SUBVOL_RDONLY)
1540 		btrfs_set_root_flags(&root->root_item,
1541 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1542 	else
1543 		btrfs_set_root_flags(&root->root_item,
1544 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1545 
1546 	trans = btrfs_start_transaction(root, 1);
1547 	if (IS_ERR(trans)) {
1548 		ret = PTR_ERR(trans);
1549 		goto out_reset;
1550 	}
1551 
1552 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1553 				&root->root_key, &root->root_item);
1554 
1555 	btrfs_commit_transaction(trans, root);
1556 out_reset:
1557 	if (ret)
1558 		btrfs_set_root_flags(&root->root_item, root_flags);
1559 out:
1560 	up_write(&root->fs_info->subvol_sem);
1561 	return ret;
1562 }
1563 
1564 /*
1565  * helper to check if the subvolume references other subvolumes
1566  */
1567 static noinline int may_destroy_subvol(struct btrfs_root *root)
1568 {
1569 	struct btrfs_path *path;
1570 	struct btrfs_key key;
1571 	int ret;
1572 
1573 	path = btrfs_alloc_path();
1574 	if (!path)
1575 		return -ENOMEM;
1576 
1577 	key.objectid = root->root_key.objectid;
1578 	key.type = BTRFS_ROOT_REF_KEY;
1579 	key.offset = (u64)-1;
1580 
1581 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1582 				&key, path, 0, 0);
1583 	if (ret < 0)
1584 		goto out;
1585 	BUG_ON(ret == 0);
1586 
1587 	ret = 0;
1588 	if (path->slots[0] > 0) {
1589 		path->slots[0]--;
1590 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1591 		if (key.objectid == root->root_key.objectid &&
1592 		    key.type == BTRFS_ROOT_REF_KEY)
1593 			ret = -ENOTEMPTY;
1594 	}
1595 out:
1596 	btrfs_free_path(path);
1597 	return ret;
1598 }
1599 
1600 static noinline int key_in_sk(struct btrfs_key *key,
1601 			      struct btrfs_ioctl_search_key *sk)
1602 {
1603 	struct btrfs_key test;
1604 	int ret;
1605 
1606 	test.objectid = sk->min_objectid;
1607 	test.type = sk->min_type;
1608 	test.offset = sk->min_offset;
1609 
1610 	ret = btrfs_comp_cpu_keys(key, &test);
1611 	if (ret < 0)
1612 		return 0;
1613 
1614 	test.objectid = sk->max_objectid;
1615 	test.type = sk->max_type;
1616 	test.offset = sk->max_offset;
1617 
1618 	ret = btrfs_comp_cpu_keys(key, &test);
1619 	if (ret > 0)
1620 		return 0;
1621 	return 1;
1622 }
1623 
1624 static noinline int copy_to_sk(struct btrfs_root *root,
1625 			       struct btrfs_path *path,
1626 			       struct btrfs_key *key,
1627 			       struct btrfs_ioctl_search_key *sk,
1628 			       char *buf,
1629 			       unsigned long *sk_offset,
1630 			       int *num_found)
1631 {
1632 	u64 found_transid;
1633 	struct extent_buffer *leaf;
1634 	struct btrfs_ioctl_search_header sh;
1635 	unsigned long item_off;
1636 	unsigned long item_len;
1637 	int nritems;
1638 	int i;
1639 	int slot;
1640 	int ret = 0;
1641 
1642 	leaf = path->nodes[0];
1643 	slot = path->slots[0];
1644 	nritems = btrfs_header_nritems(leaf);
1645 
1646 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1647 		i = nritems;
1648 		goto advance_key;
1649 	}
1650 	found_transid = btrfs_header_generation(leaf);
1651 
1652 	for (i = slot; i < nritems; i++) {
1653 		item_off = btrfs_item_ptr_offset(leaf, i);
1654 		item_len = btrfs_item_size_nr(leaf, i);
1655 
1656 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1657 			item_len = 0;
1658 
1659 		if (sizeof(sh) + item_len + *sk_offset >
1660 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1661 			ret = 1;
1662 			goto overflow;
1663 		}
1664 
1665 		btrfs_item_key_to_cpu(leaf, key, i);
1666 		if (!key_in_sk(key, sk))
1667 			continue;
1668 
1669 		sh.objectid = key->objectid;
1670 		sh.offset = key->offset;
1671 		sh.type = key->type;
1672 		sh.len = item_len;
1673 		sh.transid = found_transid;
1674 
1675 		/* copy search result header */
1676 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1677 		*sk_offset += sizeof(sh);
1678 
1679 		if (item_len) {
1680 			char *p = buf + *sk_offset;
1681 			/* copy the item */
1682 			read_extent_buffer(leaf, p,
1683 					   item_off, item_len);
1684 			*sk_offset += item_len;
1685 		}
1686 		(*num_found)++;
1687 
1688 		if (*num_found >= sk->nr_items)
1689 			break;
1690 	}
1691 advance_key:
1692 	ret = 0;
1693 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1694 		key->offset++;
1695 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1696 		key->offset = 0;
1697 		key->type++;
1698 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1699 		key->offset = 0;
1700 		key->type = 0;
1701 		key->objectid++;
1702 	} else
1703 		ret = 1;
1704 overflow:
1705 	return ret;
1706 }
1707 
1708 static noinline int search_ioctl(struct inode *inode,
1709 				 struct btrfs_ioctl_search_args *args)
1710 {
1711 	struct btrfs_root *root;
1712 	struct btrfs_key key;
1713 	struct btrfs_key max_key;
1714 	struct btrfs_path *path;
1715 	struct btrfs_ioctl_search_key *sk = &args->key;
1716 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1717 	int ret;
1718 	int num_found = 0;
1719 	unsigned long sk_offset = 0;
1720 
1721 	path = btrfs_alloc_path();
1722 	if (!path)
1723 		return -ENOMEM;
1724 
1725 	if (sk->tree_id == 0) {
1726 		/* search the root of the inode that was passed */
1727 		root = BTRFS_I(inode)->root;
1728 	} else {
1729 		key.objectid = sk->tree_id;
1730 		key.type = BTRFS_ROOT_ITEM_KEY;
1731 		key.offset = (u64)-1;
1732 		root = btrfs_read_fs_root_no_name(info, &key);
1733 		if (IS_ERR(root)) {
1734 			printk(KERN_ERR "could not find root %llu\n",
1735 			       sk->tree_id);
1736 			btrfs_free_path(path);
1737 			return -ENOENT;
1738 		}
1739 	}
1740 
1741 	key.objectid = sk->min_objectid;
1742 	key.type = sk->min_type;
1743 	key.offset = sk->min_offset;
1744 
1745 	max_key.objectid = sk->max_objectid;
1746 	max_key.type = sk->max_type;
1747 	max_key.offset = sk->max_offset;
1748 
1749 	path->keep_locks = 1;
1750 
1751 	while(1) {
1752 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1753 					   sk->min_transid);
1754 		if (ret != 0) {
1755 			if (ret > 0)
1756 				ret = 0;
1757 			goto err;
1758 		}
1759 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1760 				 &sk_offset, &num_found);
1761 		btrfs_release_path(path);
1762 		if (ret || num_found >= sk->nr_items)
1763 			break;
1764 
1765 	}
1766 	ret = 0;
1767 err:
1768 	sk->nr_items = num_found;
1769 	btrfs_free_path(path);
1770 	return ret;
1771 }
1772 
1773 static noinline int btrfs_ioctl_tree_search(struct file *file,
1774 					   void __user *argp)
1775 {
1776 	 struct btrfs_ioctl_search_args *args;
1777 	 struct inode *inode;
1778 	 int ret;
1779 
1780 	if (!capable(CAP_SYS_ADMIN))
1781 		return -EPERM;
1782 
1783 	args = memdup_user(argp, sizeof(*args));
1784 	if (IS_ERR(args))
1785 		return PTR_ERR(args);
1786 
1787 	inode = fdentry(file)->d_inode;
1788 	ret = search_ioctl(inode, args);
1789 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1790 		ret = -EFAULT;
1791 	kfree(args);
1792 	return ret;
1793 }
1794 
1795 /*
1796  * Search INODE_REFs to identify path name of 'dirid' directory
1797  * in a 'tree_id' tree. and sets path name to 'name'.
1798  */
1799 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1800 				u64 tree_id, u64 dirid, char *name)
1801 {
1802 	struct btrfs_root *root;
1803 	struct btrfs_key key;
1804 	char *ptr;
1805 	int ret = -1;
1806 	int slot;
1807 	int len;
1808 	int total_len = 0;
1809 	struct btrfs_inode_ref *iref;
1810 	struct extent_buffer *l;
1811 	struct btrfs_path *path;
1812 
1813 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1814 		name[0]='\0';
1815 		return 0;
1816 	}
1817 
1818 	path = btrfs_alloc_path();
1819 	if (!path)
1820 		return -ENOMEM;
1821 
1822 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1823 
1824 	key.objectid = tree_id;
1825 	key.type = BTRFS_ROOT_ITEM_KEY;
1826 	key.offset = (u64)-1;
1827 	root = btrfs_read_fs_root_no_name(info, &key);
1828 	if (IS_ERR(root)) {
1829 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1830 		ret = -ENOENT;
1831 		goto out;
1832 	}
1833 
1834 	key.objectid = dirid;
1835 	key.type = BTRFS_INODE_REF_KEY;
1836 	key.offset = (u64)-1;
1837 
1838 	while(1) {
1839 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1840 		if (ret < 0)
1841 			goto out;
1842 
1843 		l = path->nodes[0];
1844 		slot = path->slots[0];
1845 		if (ret > 0 && slot > 0)
1846 			slot--;
1847 		btrfs_item_key_to_cpu(l, &key, slot);
1848 
1849 		if (ret > 0 && (key.objectid != dirid ||
1850 				key.type != BTRFS_INODE_REF_KEY)) {
1851 			ret = -ENOENT;
1852 			goto out;
1853 		}
1854 
1855 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1856 		len = btrfs_inode_ref_name_len(l, iref);
1857 		ptr -= len + 1;
1858 		total_len += len + 1;
1859 		if (ptr < name)
1860 			goto out;
1861 
1862 		*(ptr + len) = '/';
1863 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1864 
1865 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1866 			break;
1867 
1868 		btrfs_release_path(path);
1869 		key.objectid = key.offset;
1870 		key.offset = (u64)-1;
1871 		dirid = key.objectid;
1872 	}
1873 	if (ptr < name)
1874 		goto out;
1875 	memmove(name, ptr, total_len);
1876 	name[total_len]='\0';
1877 	ret = 0;
1878 out:
1879 	btrfs_free_path(path);
1880 	return ret;
1881 }
1882 
1883 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1884 					   void __user *argp)
1885 {
1886 	 struct btrfs_ioctl_ino_lookup_args *args;
1887 	 struct inode *inode;
1888 	 int ret;
1889 
1890 	if (!capable(CAP_SYS_ADMIN))
1891 		return -EPERM;
1892 
1893 	args = memdup_user(argp, sizeof(*args));
1894 	if (IS_ERR(args))
1895 		return PTR_ERR(args);
1896 
1897 	inode = fdentry(file)->d_inode;
1898 
1899 	if (args->treeid == 0)
1900 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1901 
1902 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1903 					args->treeid, args->objectid,
1904 					args->name);
1905 
1906 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1907 		ret = -EFAULT;
1908 
1909 	kfree(args);
1910 	return ret;
1911 }
1912 
1913 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1914 					     void __user *arg)
1915 {
1916 	struct dentry *parent = fdentry(file);
1917 	struct dentry *dentry;
1918 	struct inode *dir = parent->d_inode;
1919 	struct inode *inode;
1920 	struct btrfs_root *root = BTRFS_I(dir)->root;
1921 	struct btrfs_root *dest = NULL;
1922 	struct btrfs_ioctl_vol_args *vol_args;
1923 	struct btrfs_trans_handle *trans;
1924 	int namelen;
1925 	int ret;
1926 	int err = 0;
1927 
1928 	vol_args = memdup_user(arg, sizeof(*vol_args));
1929 	if (IS_ERR(vol_args))
1930 		return PTR_ERR(vol_args);
1931 
1932 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1933 	namelen = strlen(vol_args->name);
1934 	if (strchr(vol_args->name, '/') ||
1935 	    strncmp(vol_args->name, "..", namelen) == 0) {
1936 		err = -EINVAL;
1937 		goto out;
1938 	}
1939 
1940 	err = mnt_want_write_file(file);
1941 	if (err)
1942 		goto out;
1943 
1944 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1945 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1946 	if (IS_ERR(dentry)) {
1947 		err = PTR_ERR(dentry);
1948 		goto out_unlock_dir;
1949 	}
1950 
1951 	if (!dentry->d_inode) {
1952 		err = -ENOENT;
1953 		goto out_dput;
1954 	}
1955 
1956 	inode = dentry->d_inode;
1957 	dest = BTRFS_I(inode)->root;
1958 	if (!capable(CAP_SYS_ADMIN)){
1959 		/*
1960 		 * Regular user.  Only allow this with a special mount
1961 		 * option, when the user has write+exec access to the
1962 		 * subvol root, and when rmdir(2) would have been
1963 		 * allowed.
1964 		 *
1965 		 * Note that this is _not_ check that the subvol is
1966 		 * empty or doesn't contain data that we wouldn't
1967 		 * otherwise be able to delete.
1968 		 *
1969 		 * Users who want to delete empty subvols should try
1970 		 * rmdir(2).
1971 		 */
1972 		err = -EPERM;
1973 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1974 			goto out_dput;
1975 
1976 		/*
1977 		 * Do not allow deletion if the parent dir is the same
1978 		 * as the dir to be deleted.  That means the ioctl
1979 		 * must be called on the dentry referencing the root
1980 		 * of the subvol, not a random directory contained
1981 		 * within it.
1982 		 */
1983 		err = -EINVAL;
1984 		if (root == dest)
1985 			goto out_dput;
1986 
1987 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1988 		if (err)
1989 			goto out_dput;
1990 
1991 		/* check if subvolume may be deleted by a non-root user */
1992 		err = btrfs_may_delete(dir, dentry, 1);
1993 		if (err)
1994 			goto out_dput;
1995 	}
1996 
1997 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1998 		err = -EINVAL;
1999 		goto out_dput;
2000 	}
2001 
2002 	mutex_lock(&inode->i_mutex);
2003 	err = d_invalidate(dentry);
2004 	if (err)
2005 		goto out_unlock;
2006 
2007 	down_write(&root->fs_info->subvol_sem);
2008 
2009 	err = may_destroy_subvol(dest);
2010 	if (err)
2011 		goto out_up_write;
2012 
2013 	trans = btrfs_start_transaction(root, 0);
2014 	if (IS_ERR(trans)) {
2015 		err = PTR_ERR(trans);
2016 		goto out_up_write;
2017 	}
2018 	trans->block_rsv = &root->fs_info->global_block_rsv;
2019 
2020 	ret = btrfs_unlink_subvol(trans, root, dir,
2021 				dest->root_key.objectid,
2022 				dentry->d_name.name,
2023 				dentry->d_name.len);
2024 	if (ret) {
2025 		err = ret;
2026 		btrfs_abort_transaction(trans, root, ret);
2027 		goto out_end_trans;
2028 	}
2029 
2030 	btrfs_record_root_in_trans(trans, dest);
2031 
2032 	memset(&dest->root_item.drop_progress, 0,
2033 		sizeof(dest->root_item.drop_progress));
2034 	dest->root_item.drop_level = 0;
2035 	btrfs_set_root_refs(&dest->root_item, 0);
2036 
2037 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2038 		ret = btrfs_insert_orphan_item(trans,
2039 					root->fs_info->tree_root,
2040 					dest->root_key.objectid);
2041 		if (ret) {
2042 			btrfs_abort_transaction(trans, root, ret);
2043 			err = ret;
2044 			goto out_end_trans;
2045 		}
2046 	}
2047 out_end_trans:
2048 	ret = btrfs_end_transaction(trans, root);
2049 	if (ret && !err)
2050 		err = ret;
2051 	inode->i_flags |= S_DEAD;
2052 out_up_write:
2053 	up_write(&root->fs_info->subvol_sem);
2054 out_unlock:
2055 	mutex_unlock(&inode->i_mutex);
2056 	if (!err) {
2057 		shrink_dcache_sb(root->fs_info->sb);
2058 		btrfs_invalidate_inodes(dest);
2059 		d_delete(dentry);
2060 	}
2061 out_dput:
2062 	dput(dentry);
2063 out_unlock_dir:
2064 	mutex_unlock(&dir->i_mutex);
2065 	mnt_drop_write_file(file);
2066 out:
2067 	kfree(vol_args);
2068 	return err;
2069 }
2070 
2071 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2072 {
2073 	struct inode *inode = fdentry(file)->d_inode;
2074 	struct btrfs_root *root = BTRFS_I(inode)->root;
2075 	struct btrfs_ioctl_defrag_range_args *range;
2076 	int ret;
2077 
2078 	if (btrfs_root_readonly(root))
2079 		return -EROFS;
2080 
2081 	ret = mnt_want_write_file(file);
2082 	if (ret)
2083 		return ret;
2084 
2085 	switch (inode->i_mode & S_IFMT) {
2086 	case S_IFDIR:
2087 		if (!capable(CAP_SYS_ADMIN)) {
2088 			ret = -EPERM;
2089 			goto out;
2090 		}
2091 		ret = btrfs_defrag_root(root, 0);
2092 		if (ret)
2093 			goto out;
2094 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2095 		break;
2096 	case S_IFREG:
2097 		if (!(file->f_mode & FMODE_WRITE)) {
2098 			ret = -EINVAL;
2099 			goto out;
2100 		}
2101 
2102 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2103 		if (!range) {
2104 			ret = -ENOMEM;
2105 			goto out;
2106 		}
2107 
2108 		if (argp) {
2109 			if (copy_from_user(range, argp,
2110 					   sizeof(*range))) {
2111 				ret = -EFAULT;
2112 				kfree(range);
2113 				goto out;
2114 			}
2115 			/* compression requires us to start the IO */
2116 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2117 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2118 				range->extent_thresh = (u32)-1;
2119 			}
2120 		} else {
2121 			/* the rest are all set to zero by kzalloc */
2122 			range->len = (u64)-1;
2123 		}
2124 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2125 					range, 0, 0);
2126 		if (ret > 0)
2127 			ret = 0;
2128 		kfree(range);
2129 		break;
2130 	default:
2131 		ret = -EINVAL;
2132 	}
2133 out:
2134 	mnt_drop_write_file(file);
2135 	return ret;
2136 }
2137 
2138 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2139 {
2140 	struct btrfs_ioctl_vol_args *vol_args;
2141 	int ret;
2142 
2143 	if (!capable(CAP_SYS_ADMIN))
2144 		return -EPERM;
2145 
2146 	mutex_lock(&root->fs_info->volume_mutex);
2147 	if (root->fs_info->balance_ctl) {
2148 		printk(KERN_INFO "btrfs: balance in progress\n");
2149 		ret = -EINVAL;
2150 		goto out;
2151 	}
2152 
2153 	vol_args = memdup_user(arg, sizeof(*vol_args));
2154 	if (IS_ERR(vol_args)) {
2155 		ret = PTR_ERR(vol_args);
2156 		goto out;
2157 	}
2158 
2159 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2160 	ret = btrfs_init_new_device(root, vol_args->name);
2161 
2162 	kfree(vol_args);
2163 out:
2164 	mutex_unlock(&root->fs_info->volume_mutex);
2165 	return ret;
2166 }
2167 
2168 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2169 {
2170 	struct btrfs_ioctl_vol_args *vol_args;
2171 	int ret;
2172 
2173 	if (!capable(CAP_SYS_ADMIN))
2174 		return -EPERM;
2175 
2176 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2177 		return -EROFS;
2178 
2179 	mutex_lock(&root->fs_info->volume_mutex);
2180 	if (root->fs_info->balance_ctl) {
2181 		printk(KERN_INFO "btrfs: balance in progress\n");
2182 		ret = -EINVAL;
2183 		goto out;
2184 	}
2185 
2186 	vol_args = memdup_user(arg, sizeof(*vol_args));
2187 	if (IS_ERR(vol_args)) {
2188 		ret = PTR_ERR(vol_args);
2189 		goto out;
2190 	}
2191 
2192 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2193 	ret = btrfs_rm_device(root, vol_args->name);
2194 
2195 	kfree(vol_args);
2196 out:
2197 	mutex_unlock(&root->fs_info->volume_mutex);
2198 	return ret;
2199 }
2200 
2201 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2202 {
2203 	struct btrfs_ioctl_fs_info_args *fi_args;
2204 	struct btrfs_device *device;
2205 	struct btrfs_device *next;
2206 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2207 	int ret = 0;
2208 
2209 	if (!capable(CAP_SYS_ADMIN))
2210 		return -EPERM;
2211 
2212 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2213 	if (!fi_args)
2214 		return -ENOMEM;
2215 
2216 	fi_args->num_devices = fs_devices->num_devices;
2217 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2218 
2219 	mutex_lock(&fs_devices->device_list_mutex);
2220 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2221 		if (device->devid > fi_args->max_id)
2222 			fi_args->max_id = device->devid;
2223 	}
2224 	mutex_unlock(&fs_devices->device_list_mutex);
2225 
2226 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2227 		ret = -EFAULT;
2228 
2229 	kfree(fi_args);
2230 	return ret;
2231 }
2232 
2233 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2234 {
2235 	struct btrfs_ioctl_dev_info_args *di_args;
2236 	struct btrfs_device *dev;
2237 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2238 	int ret = 0;
2239 	char *s_uuid = NULL;
2240 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2241 
2242 	if (!capable(CAP_SYS_ADMIN))
2243 		return -EPERM;
2244 
2245 	di_args = memdup_user(arg, sizeof(*di_args));
2246 	if (IS_ERR(di_args))
2247 		return PTR_ERR(di_args);
2248 
2249 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2250 		s_uuid = di_args->uuid;
2251 
2252 	mutex_lock(&fs_devices->device_list_mutex);
2253 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2254 	mutex_unlock(&fs_devices->device_list_mutex);
2255 
2256 	if (!dev) {
2257 		ret = -ENODEV;
2258 		goto out;
2259 	}
2260 
2261 	di_args->devid = dev->devid;
2262 	di_args->bytes_used = dev->bytes_used;
2263 	di_args->total_bytes = dev->total_bytes;
2264 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2265 	if (dev->name)
2266 		strncpy(di_args->path, dev->name, sizeof(di_args->path));
2267 	else
2268 		di_args->path[0] = '\0';
2269 
2270 out:
2271 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2272 		ret = -EFAULT;
2273 
2274 	kfree(di_args);
2275 	return ret;
2276 }
2277 
2278 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2279 				       u64 off, u64 olen, u64 destoff)
2280 {
2281 	struct inode *inode = fdentry(file)->d_inode;
2282 	struct btrfs_root *root = BTRFS_I(inode)->root;
2283 	struct file *src_file;
2284 	struct inode *src;
2285 	struct btrfs_trans_handle *trans;
2286 	struct btrfs_path *path;
2287 	struct extent_buffer *leaf;
2288 	char *buf;
2289 	struct btrfs_key key;
2290 	u32 nritems;
2291 	int slot;
2292 	int ret;
2293 	u64 len = olen;
2294 	u64 bs = root->fs_info->sb->s_blocksize;
2295 	u64 hint_byte;
2296 
2297 	/*
2298 	 * TODO:
2299 	 * - split compressed inline extents.  annoying: we need to
2300 	 *   decompress into destination's address_space (the file offset
2301 	 *   may change, so source mapping won't do), then recompress (or
2302 	 *   otherwise reinsert) a subrange.
2303 	 * - allow ranges within the same file to be cloned (provided
2304 	 *   they don't overlap)?
2305 	 */
2306 
2307 	/* the destination must be opened for writing */
2308 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2309 		return -EINVAL;
2310 
2311 	if (btrfs_root_readonly(root))
2312 		return -EROFS;
2313 
2314 	ret = mnt_want_write_file(file);
2315 	if (ret)
2316 		return ret;
2317 
2318 	src_file = fget(srcfd);
2319 	if (!src_file) {
2320 		ret = -EBADF;
2321 		goto out_drop_write;
2322 	}
2323 
2324 	src = src_file->f_dentry->d_inode;
2325 
2326 	ret = -EINVAL;
2327 	if (src == inode)
2328 		goto out_fput;
2329 
2330 	/* the src must be open for reading */
2331 	if (!(src_file->f_mode & FMODE_READ))
2332 		goto out_fput;
2333 
2334 	/* don't make the dst file partly checksummed */
2335 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2336 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2337 		goto out_fput;
2338 
2339 	ret = -EISDIR;
2340 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2341 		goto out_fput;
2342 
2343 	ret = -EXDEV;
2344 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2345 		goto out_fput;
2346 
2347 	ret = -ENOMEM;
2348 	buf = vmalloc(btrfs_level_size(root, 0));
2349 	if (!buf)
2350 		goto out_fput;
2351 
2352 	path = btrfs_alloc_path();
2353 	if (!path) {
2354 		vfree(buf);
2355 		goto out_fput;
2356 	}
2357 	path->reada = 2;
2358 
2359 	if (inode < src) {
2360 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2361 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2362 	} else {
2363 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2364 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2365 	}
2366 
2367 	/* determine range to clone */
2368 	ret = -EINVAL;
2369 	if (off + len > src->i_size || off + len < off)
2370 		goto out_unlock;
2371 	if (len == 0)
2372 		olen = len = src->i_size - off;
2373 	/* if we extend to eof, continue to block boundary */
2374 	if (off + len == src->i_size)
2375 		len = ALIGN(src->i_size, bs) - off;
2376 
2377 	/* verify the end result is block aligned */
2378 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2379 	    !IS_ALIGNED(destoff, bs))
2380 		goto out_unlock;
2381 
2382 	if (destoff > inode->i_size) {
2383 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2384 		if (ret)
2385 			goto out_unlock;
2386 	}
2387 
2388 	/* truncate page cache pages from target inode range */
2389 	truncate_inode_pages_range(&inode->i_data, destoff,
2390 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2391 
2392 	/* do any pending delalloc/csum calc on src, one way or
2393 	   another, and lock file content */
2394 	while (1) {
2395 		struct btrfs_ordered_extent *ordered;
2396 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2397 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2398 		if (!ordered &&
2399 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2400 				   EXTENT_DELALLOC, 0, NULL))
2401 			break;
2402 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2403 		if (ordered)
2404 			btrfs_put_ordered_extent(ordered);
2405 		btrfs_wait_ordered_range(src, off, len);
2406 	}
2407 
2408 	/* clone data */
2409 	key.objectid = btrfs_ino(src);
2410 	key.type = BTRFS_EXTENT_DATA_KEY;
2411 	key.offset = 0;
2412 
2413 	while (1) {
2414 		/*
2415 		 * note the key will change type as we walk through the
2416 		 * tree.
2417 		 */
2418 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2419 		if (ret < 0)
2420 			goto out;
2421 
2422 		nritems = btrfs_header_nritems(path->nodes[0]);
2423 		if (path->slots[0] >= nritems) {
2424 			ret = btrfs_next_leaf(root, path);
2425 			if (ret < 0)
2426 				goto out;
2427 			if (ret > 0)
2428 				break;
2429 			nritems = btrfs_header_nritems(path->nodes[0]);
2430 		}
2431 		leaf = path->nodes[0];
2432 		slot = path->slots[0];
2433 
2434 		btrfs_item_key_to_cpu(leaf, &key, slot);
2435 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2436 		    key.objectid != btrfs_ino(src))
2437 			break;
2438 
2439 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2440 			struct btrfs_file_extent_item *extent;
2441 			int type;
2442 			u32 size;
2443 			struct btrfs_key new_key;
2444 			u64 disko = 0, diskl = 0;
2445 			u64 datao = 0, datal = 0;
2446 			u8 comp;
2447 			u64 endoff;
2448 
2449 			size = btrfs_item_size_nr(leaf, slot);
2450 			read_extent_buffer(leaf, buf,
2451 					   btrfs_item_ptr_offset(leaf, slot),
2452 					   size);
2453 
2454 			extent = btrfs_item_ptr(leaf, slot,
2455 						struct btrfs_file_extent_item);
2456 			comp = btrfs_file_extent_compression(leaf, extent);
2457 			type = btrfs_file_extent_type(leaf, extent);
2458 			if (type == BTRFS_FILE_EXTENT_REG ||
2459 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2460 				disko = btrfs_file_extent_disk_bytenr(leaf,
2461 								      extent);
2462 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2463 								 extent);
2464 				datao = btrfs_file_extent_offset(leaf, extent);
2465 				datal = btrfs_file_extent_num_bytes(leaf,
2466 								    extent);
2467 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2468 				/* take upper bound, may be compressed */
2469 				datal = btrfs_file_extent_ram_bytes(leaf,
2470 								    extent);
2471 			}
2472 			btrfs_release_path(path);
2473 
2474 			if (key.offset + datal <= off ||
2475 			    key.offset >= off+len)
2476 				goto next;
2477 
2478 			memcpy(&new_key, &key, sizeof(new_key));
2479 			new_key.objectid = btrfs_ino(inode);
2480 			if (off <= key.offset)
2481 				new_key.offset = key.offset + destoff - off;
2482 			else
2483 				new_key.offset = destoff;
2484 
2485 			/*
2486 			 * 1 - adjusting old extent (we may have to split it)
2487 			 * 1 - add new extent
2488 			 * 1 - inode update
2489 			 */
2490 			trans = btrfs_start_transaction(root, 3);
2491 			if (IS_ERR(trans)) {
2492 				ret = PTR_ERR(trans);
2493 				goto out;
2494 			}
2495 
2496 			if (type == BTRFS_FILE_EXTENT_REG ||
2497 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2498 				/*
2499 				 *    a  | --- range to clone ---|  b
2500 				 * | ------------- extent ------------- |
2501 				 */
2502 
2503 				/* substract range b */
2504 				if (key.offset + datal > off + len)
2505 					datal = off + len - key.offset;
2506 
2507 				/* substract range a */
2508 				if (off > key.offset) {
2509 					datao += off - key.offset;
2510 					datal -= off - key.offset;
2511 				}
2512 
2513 				ret = btrfs_drop_extents(trans, inode,
2514 							 new_key.offset,
2515 							 new_key.offset + datal,
2516 							 &hint_byte, 1);
2517 				if (ret) {
2518 					btrfs_abort_transaction(trans, root,
2519 								ret);
2520 					btrfs_end_transaction(trans, root);
2521 					goto out;
2522 				}
2523 
2524 				ret = btrfs_insert_empty_item(trans, root, path,
2525 							      &new_key, size);
2526 				if (ret) {
2527 					btrfs_abort_transaction(trans, root,
2528 								ret);
2529 					btrfs_end_transaction(trans, root);
2530 					goto out;
2531 				}
2532 
2533 				leaf = path->nodes[0];
2534 				slot = path->slots[0];
2535 				write_extent_buffer(leaf, buf,
2536 					    btrfs_item_ptr_offset(leaf, slot),
2537 					    size);
2538 
2539 				extent = btrfs_item_ptr(leaf, slot,
2540 						struct btrfs_file_extent_item);
2541 
2542 				/* disko == 0 means it's a hole */
2543 				if (!disko)
2544 					datao = 0;
2545 
2546 				btrfs_set_file_extent_offset(leaf, extent,
2547 							     datao);
2548 				btrfs_set_file_extent_num_bytes(leaf, extent,
2549 								datal);
2550 				if (disko) {
2551 					inode_add_bytes(inode, datal);
2552 					ret = btrfs_inc_extent_ref(trans, root,
2553 							disko, diskl, 0,
2554 							root->root_key.objectid,
2555 							btrfs_ino(inode),
2556 							new_key.offset - datao,
2557 							0);
2558 					if (ret) {
2559 						btrfs_abort_transaction(trans,
2560 									root,
2561 									ret);
2562 						btrfs_end_transaction(trans,
2563 								      root);
2564 						goto out;
2565 
2566 					}
2567 				}
2568 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2569 				u64 skip = 0;
2570 				u64 trim = 0;
2571 				if (off > key.offset) {
2572 					skip = off - key.offset;
2573 					new_key.offset += skip;
2574 				}
2575 
2576 				if (key.offset + datal > off+len)
2577 					trim = key.offset + datal - (off+len);
2578 
2579 				if (comp && (skip || trim)) {
2580 					ret = -EINVAL;
2581 					btrfs_end_transaction(trans, root);
2582 					goto out;
2583 				}
2584 				size -= skip + trim;
2585 				datal -= skip + trim;
2586 
2587 				ret = btrfs_drop_extents(trans, inode,
2588 							 new_key.offset,
2589 							 new_key.offset + datal,
2590 							 &hint_byte, 1);
2591 				if (ret) {
2592 					btrfs_abort_transaction(trans, root,
2593 								ret);
2594 					btrfs_end_transaction(trans, root);
2595 					goto out;
2596 				}
2597 
2598 				ret = btrfs_insert_empty_item(trans, root, path,
2599 							      &new_key, size);
2600 				if (ret) {
2601 					btrfs_abort_transaction(trans, root,
2602 								ret);
2603 					btrfs_end_transaction(trans, root);
2604 					goto out;
2605 				}
2606 
2607 				if (skip) {
2608 					u32 start =
2609 					  btrfs_file_extent_calc_inline_size(0);
2610 					memmove(buf+start, buf+start+skip,
2611 						datal);
2612 				}
2613 
2614 				leaf = path->nodes[0];
2615 				slot = path->slots[0];
2616 				write_extent_buffer(leaf, buf,
2617 					    btrfs_item_ptr_offset(leaf, slot),
2618 					    size);
2619 				inode_add_bytes(inode, datal);
2620 			}
2621 
2622 			btrfs_mark_buffer_dirty(leaf);
2623 			btrfs_release_path(path);
2624 
2625 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2626 
2627 			/*
2628 			 * we round up to the block size at eof when
2629 			 * determining which extents to clone above,
2630 			 * but shouldn't round up the file size
2631 			 */
2632 			endoff = new_key.offset + datal;
2633 			if (endoff > destoff+olen)
2634 				endoff = destoff+olen;
2635 			if (endoff > inode->i_size)
2636 				btrfs_i_size_write(inode, endoff);
2637 
2638 			ret = btrfs_update_inode(trans, root, inode);
2639 			if (ret) {
2640 				btrfs_abort_transaction(trans, root, ret);
2641 				btrfs_end_transaction(trans, root);
2642 				goto out;
2643 			}
2644 			ret = btrfs_end_transaction(trans, root);
2645 		}
2646 next:
2647 		btrfs_release_path(path);
2648 		key.offset++;
2649 	}
2650 	ret = 0;
2651 out:
2652 	btrfs_release_path(path);
2653 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2654 out_unlock:
2655 	mutex_unlock(&src->i_mutex);
2656 	mutex_unlock(&inode->i_mutex);
2657 	vfree(buf);
2658 	btrfs_free_path(path);
2659 out_fput:
2660 	fput(src_file);
2661 out_drop_write:
2662 	mnt_drop_write_file(file);
2663 	return ret;
2664 }
2665 
2666 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2667 {
2668 	struct btrfs_ioctl_clone_range_args args;
2669 
2670 	if (copy_from_user(&args, argp, sizeof(args)))
2671 		return -EFAULT;
2672 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2673 				 args.src_length, args.dest_offset);
2674 }
2675 
2676 /*
2677  * there are many ways the trans_start and trans_end ioctls can lead
2678  * to deadlocks.  They should only be used by applications that
2679  * basically own the machine, and have a very in depth understanding
2680  * of all the possible deadlocks and enospc problems.
2681  */
2682 static long btrfs_ioctl_trans_start(struct file *file)
2683 {
2684 	struct inode *inode = fdentry(file)->d_inode;
2685 	struct btrfs_root *root = BTRFS_I(inode)->root;
2686 	struct btrfs_trans_handle *trans;
2687 	int ret;
2688 
2689 	ret = -EPERM;
2690 	if (!capable(CAP_SYS_ADMIN))
2691 		goto out;
2692 
2693 	ret = -EINPROGRESS;
2694 	if (file->private_data)
2695 		goto out;
2696 
2697 	ret = -EROFS;
2698 	if (btrfs_root_readonly(root))
2699 		goto out;
2700 
2701 	ret = mnt_want_write_file(file);
2702 	if (ret)
2703 		goto out;
2704 
2705 	atomic_inc(&root->fs_info->open_ioctl_trans);
2706 
2707 	ret = -ENOMEM;
2708 	trans = btrfs_start_ioctl_transaction(root);
2709 	if (IS_ERR(trans))
2710 		goto out_drop;
2711 
2712 	file->private_data = trans;
2713 	return 0;
2714 
2715 out_drop:
2716 	atomic_dec(&root->fs_info->open_ioctl_trans);
2717 	mnt_drop_write_file(file);
2718 out:
2719 	return ret;
2720 }
2721 
2722 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2723 {
2724 	struct inode *inode = fdentry(file)->d_inode;
2725 	struct btrfs_root *root = BTRFS_I(inode)->root;
2726 	struct btrfs_root *new_root;
2727 	struct btrfs_dir_item *di;
2728 	struct btrfs_trans_handle *trans;
2729 	struct btrfs_path *path;
2730 	struct btrfs_key location;
2731 	struct btrfs_disk_key disk_key;
2732 	struct btrfs_super_block *disk_super;
2733 	u64 features;
2734 	u64 objectid = 0;
2735 	u64 dir_id;
2736 
2737 	if (!capable(CAP_SYS_ADMIN))
2738 		return -EPERM;
2739 
2740 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2741 		return -EFAULT;
2742 
2743 	if (!objectid)
2744 		objectid = root->root_key.objectid;
2745 
2746 	location.objectid = objectid;
2747 	location.type = BTRFS_ROOT_ITEM_KEY;
2748 	location.offset = (u64)-1;
2749 
2750 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2751 	if (IS_ERR(new_root))
2752 		return PTR_ERR(new_root);
2753 
2754 	if (btrfs_root_refs(&new_root->root_item) == 0)
2755 		return -ENOENT;
2756 
2757 	path = btrfs_alloc_path();
2758 	if (!path)
2759 		return -ENOMEM;
2760 	path->leave_spinning = 1;
2761 
2762 	trans = btrfs_start_transaction(root, 1);
2763 	if (IS_ERR(trans)) {
2764 		btrfs_free_path(path);
2765 		return PTR_ERR(trans);
2766 	}
2767 
2768 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2769 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2770 				   dir_id, "default", 7, 1);
2771 	if (IS_ERR_OR_NULL(di)) {
2772 		btrfs_free_path(path);
2773 		btrfs_end_transaction(trans, root);
2774 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2775 		       "this isn't going to work\n");
2776 		return -ENOENT;
2777 	}
2778 
2779 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2780 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2781 	btrfs_mark_buffer_dirty(path->nodes[0]);
2782 	btrfs_free_path(path);
2783 
2784 	disk_super = root->fs_info->super_copy;
2785 	features = btrfs_super_incompat_flags(disk_super);
2786 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2787 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2788 		btrfs_set_super_incompat_flags(disk_super, features);
2789 	}
2790 	btrfs_end_transaction(trans, root);
2791 
2792 	return 0;
2793 }
2794 
2795 static void get_block_group_info(struct list_head *groups_list,
2796 				 struct btrfs_ioctl_space_info *space)
2797 {
2798 	struct btrfs_block_group_cache *block_group;
2799 
2800 	space->total_bytes = 0;
2801 	space->used_bytes = 0;
2802 	space->flags = 0;
2803 	list_for_each_entry(block_group, groups_list, list) {
2804 		space->flags = block_group->flags;
2805 		space->total_bytes += block_group->key.offset;
2806 		space->used_bytes +=
2807 			btrfs_block_group_used(&block_group->item);
2808 	}
2809 }
2810 
2811 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2812 {
2813 	struct btrfs_ioctl_space_args space_args;
2814 	struct btrfs_ioctl_space_info space;
2815 	struct btrfs_ioctl_space_info *dest;
2816 	struct btrfs_ioctl_space_info *dest_orig;
2817 	struct btrfs_ioctl_space_info __user *user_dest;
2818 	struct btrfs_space_info *info;
2819 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2820 		       BTRFS_BLOCK_GROUP_SYSTEM,
2821 		       BTRFS_BLOCK_GROUP_METADATA,
2822 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2823 	int num_types = 4;
2824 	int alloc_size;
2825 	int ret = 0;
2826 	u64 slot_count = 0;
2827 	int i, c;
2828 
2829 	if (copy_from_user(&space_args,
2830 			   (struct btrfs_ioctl_space_args __user *)arg,
2831 			   sizeof(space_args)))
2832 		return -EFAULT;
2833 
2834 	for (i = 0; i < num_types; i++) {
2835 		struct btrfs_space_info *tmp;
2836 
2837 		info = NULL;
2838 		rcu_read_lock();
2839 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2840 					list) {
2841 			if (tmp->flags == types[i]) {
2842 				info = tmp;
2843 				break;
2844 			}
2845 		}
2846 		rcu_read_unlock();
2847 
2848 		if (!info)
2849 			continue;
2850 
2851 		down_read(&info->groups_sem);
2852 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2853 			if (!list_empty(&info->block_groups[c]))
2854 				slot_count++;
2855 		}
2856 		up_read(&info->groups_sem);
2857 	}
2858 
2859 	/* space_slots == 0 means they are asking for a count */
2860 	if (space_args.space_slots == 0) {
2861 		space_args.total_spaces = slot_count;
2862 		goto out;
2863 	}
2864 
2865 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2866 
2867 	alloc_size = sizeof(*dest) * slot_count;
2868 
2869 	/* we generally have at most 6 or so space infos, one for each raid
2870 	 * level.  So, a whole page should be more than enough for everyone
2871 	 */
2872 	if (alloc_size > PAGE_CACHE_SIZE)
2873 		return -ENOMEM;
2874 
2875 	space_args.total_spaces = 0;
2876 	dest = kmalloc(alloc_size, GFP_NOFS);
2877 	if (!dest)
2878 		return -ENOMEM;
2879 	dest_orig = dest;
2880 
2881 	/* now we have a buffer to copy into */
2882 	for (i = 0; i < num_types; i++) {
2883 		struct btrfs_space_info *tmp;
2884 
2885 		if (!slot_count)
2886 			break;
2887 
2888 		info = NULL;
2889 		rcu_read_lock();
2890 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2891 					list) {
2892 			if (tmp->flags == types[i]) {
2893 				info = tmp;
2894 				break;
2895 			}
2896 		}
2897 		rcu_read_unlock();
2898 
2899 		if (!info)
2900 			continue;
2901 		down_read(&info->groups_sem);
2902 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2903 			if (!list_empty(&info->block_groups[c])) {
2904 				get_block_group_info(&info->block_groups[c],
2905 						     &space);
2906 				memcpy(dest, &space, sizeof(space));
2907 				dest++;
2908 				space_args.total_spaces++;
2909 				slot_count--;
2910 			}
2911 			if (!slot_count)
2912 				break;
2913 		}
2914 		up_read(&info->groups_sem);
2915 	}
2916 
2917 	user_dest = (struct btrfs_ioctl_space_info *)
2918 		(arg + sizeof(struct btrfs_ioctl_space_args));
2919 
2920 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2921 		ret = -EFAULT;
2922 
2923 	kfree(dest_orig);
2924 out:
2925 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2926 		ret = -EFAULT;
2927 
2928 	return ret;
2929 }
2930 
2931 /*
2932  * there are many ways the trans_start and trans_end ioctls can lead
2933  * to deadlocks.  They should only be used by applications that
2934  * basically own the machine, and have a very in depth understanding
2935  * of all the possible deadlocks and enospc problems.
2936  */
2937 long btrfs_ioctl_trans_end(struct file *file)
2938 {
2939 	struct inode *inode = fdentry(file)->d_inode;
2940 	struct btrfs_root *root = BTRFS_I(inode)->root;
2941 	struct btrfs_trans_handle *trans;
2942 
2943 	trans = file->private_data;
2944 	if (!trans)
2945 		return -EINVAL;
2946 	file->private_data = NULL;
2947 
2948 	btrfs_end_transaction(trans, root);
2949 
2950 	atomic_dec(&root->fs_info->open_ioctl_trans);
2951 
2952 	mnt_drop_write_file(file);
2953 	return 0;
2954 }
2955 
2956 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2957 {
2958 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2959 	struct btrfs_trans_handle *trans;
2960 	u64 transid;
2961 	int ret;
2962 
2963 	trans = btrfs_start_transaction(root, 0);
2964 	if (IS_ERR(trans))
2965 		return PTR_ERR(trans);
2966 	transid = trans->transid;
2967 	ret = btrfs_commit_transaction_async(trans, root, 0);
2968 	if (ret) {
2969 		btrfs_end_transaction(trans, root);
2970 		return ret;
2971 	}
2972 
2973 	if (argp)
2974 		if (copy_to_user(argp, &transid, sizeof(transid)))
2975 			return -EFAULT;
2976 	return 0;
2977 }
2978 
2979 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2980 {
2981 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2982 	u64 transid;
2983 
2984 	if (argp) {
2985 		if (copy_from_user(&transid, argp, sizeof(transid)))
2986 			return -EFAULT;
2987 	} else {
2988 		transid = 0;  /* current trans */
2989 	}
2990 	return btrfs_wait_for_commit(root, transid);
2991 }
2992 
2993 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2994 {
2995 	int ret;
2996 	struct btrfs_ioctl_scrub_args *sa;
2997 
2998 	if (!capable(CAP_SYS_ADMIN))
2999 		return -EPERM;
3000 
3001 	sa = memdup_user(arg, sizeof(*sa));
3002 	if (IS_ERR(sa))
3003 		return PTR_ERR(sa);
3004 
3005 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3006 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3007 
3008 	if (copy_to_user(arg, sa, sizeof(*sa)))
3009 		ret = -EFAULT;
3010 
3011 	kfree(sa);
3012 	return ret;
3013 }
3014 
3015 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3016 {
3017 	if (!capable(CAP_SYS_ADMIN))
3018 		return -EPERM;
3019 
3020 	return btrfs_scrub_cancel(root);
3021 }
3022 
3023 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3024 				       void __user *arg)
3025 {
3026 	struct btrfs_ioctl_scrub_args *sa;
3027 	int ret;
3028 
3029 	if (!capable(CAP_SYS_ADMIN))
3030 		return -EPERM;
3031 
3032 	sa = memdup_user(arg, sizeof(*sa));
3033 	if (IS_ERR(sa))
3034 		return PTR_ERR(sa);
3035 
3036 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3037 
3038 	if (copy_to_user(arg, sa, sizeof(*sa)))
3039 		ret = -EFAULT;
3040 
3041 	kfree(sa);
3042 	return ret;
3043 }
3044 
3045 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3046 {
3047 	int ret = 0;
3048 	int i;
3049 	u64 rel_ptr;
3050 	int size;
3051 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3052 	struct inode_fs_paths *ipath = NULL;
3053 	struct btrfs_path *path;
3054 
3055 	if (!capable(CAP_SYS_ADMIN))
3056 		return -EPERM;
3057 
3058 	path = btrfs_alloc_path();
3059 	if (!path) {
3060 		ret = -ENOMEM;
3061 		goto out;
3062 	}
3063 
3064 	ipa = memdup_user(arg, sizeof(*ipa));
3065 	if (IS_ERR(ipa)) {
3066 		ret = PTR_ERR(ipa);
3067 		ipa = NULL;
3068 		goto out;
3069 	}
3070 
3071 	size = min_t(u32, ipa->size, 4096);
3072 	ipath = init_ipath(size, root, path);
3073 	if (IS_ERR(ipath)) {
3074 		ret = PTR_ERR(ipath);
3075 		ipath = NULL;
3076 		goto out;
3077 	}
3078 
3079 	ret = paths_from_inode(ipa->inum, ipath);
3080 	if (ret < 0)
3081 		goto out;
3082 
3083 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3084 		rel_ptr = ipath->fspath->val[i] -
3085 			  (u64)(unsigned long)ipath->fspath->val;
3086 		ipath->fspath->val[i] = rel_ptr;
3087 	}
3088 
3089 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3090 			   (void *)(unsigned long)ipath->fspath, size);
3091 	if (ret) {
3092 		ret = -EFAULT;
3093 		goto out;
3094 	}
3095 
3096 out:
3097 	btrfs_free_path(path);
3098 	free_ipath(ipath);
3099 	kfree(ipa);
3100 
3101 	return ret;
3102 }
3103 
3104 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3105 {
3106 	struct btrfs_data_container *inodes = ctx;
3107 	const size_t c = 3 * sizeof(u64);
3108 
3109 	if (inodes->bytes_left >= c) {
3110 		inodes->bytes_left -= c;
3111 		inodes->val[inodes->elem_cnt] = inum;
3112 		inodes->val[inodes->elem_cnt + 1] = offset;
3113 		inodes->val[inodes->elem_cnt + 2] = root;
3114 		inodes->elem_cnt += 3;
3115 	} else {
3116 		inodes->bytes_missing += c - inodes->bytes_left;
3117 		inodes->bytes_left = 0;
3118 		inodes->elem_missed += 3;
3119 	}
3120 
3121 	return 0;
3122 }
3123 
3124 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3125 					void __user *arg)
3126 {
3127 	int ret = 0;
3128 	int size;
3129 	u64 extent_item_pos;
3130 	struct btrfs_ioctl_logical_ino_args *loi;
3131 	struct btrfs_data_container *inodes = NULL;
3132 	struct btrfs_path *path = NULL;
3133 	struct btrfs_key key;
3134 
3135 	if (!capable(CAP_SYS_ADMIN))
3136 		return -EPERM;
3137 
3138 	loi = memdup_user(arg, sizeof(*loi));
3139 	if (IS_ERR(loi)) {
3140 		ret = PTR_ERR(loi);
3141 		loi = NULL;
3142 		goto out;
3143 	}
3144 
3145 	path = btrfs_alloc_path();
3146 	if (!path) {
3147 		ret = -ENOMEM;
3148 		goto out;
3149 	}
3150 
3151 	size = min_t(u32, loi->size, 4096);
3152 	inodes = init_data_container(size);
3153 	if (IS_ERR(inodes)) {
3154 		ret = PTR_ERR(inodes);
3155 		inodes = NULL;
3156 		goto out;
3157 	}
3158 
3159 	ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3160 	btrfs_release_path(path);
3161 
3162 	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3163 		ret = -ENOENT;
3164 	if (ret < 0)
3165 		goto out;
3166 
3167 	extent_item_pos = loi->logical - key.objectid;
3168 	ret = iterate_extent_inodes(root->fs_info, key.objectid,
3169 					extent_item_pos, 0, build_ino_list,
3170 					inodes);
3171 
3172 	if (ret < 0)
3173 		goto out;
3174 
3175 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3176 			   (void *)(unsigned long)inodes, size);
3177 	if (ret)
3178 		ret = -EFAULT;
3179 
3180 out:
3181 	btrfs_free_path(path);
3182 	kfree(inodes);
3183 	kfree(loi);
3184 
3185 	return ret;
3186 }
3187 
3188 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3189 			       struct btrfs_ioctl_balance_args *bargs)
3190 {
3191 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3192 
3193 	bargs->flags = bctl->flags;
3194 
3195 	if (atomic_read(&fs_info->balance_running))
3196 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3197 	if (atomic_read(&fs_info->balance_pause_req))
3198 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3199 	if (atomic_read(&fs_info->balance_cancel_req))
3200 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3201 
3202 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3203 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3204 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3205 
3206 	if (lock) {
3207 		spin_lock(&fs_info->balance_lock);
3208 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3209 		spin_unlock(&fs_info->balance_lock);
3210 	} else {
3211 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3212 	}
3213 }
3214 
3215 static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3216 {
3217 	struct btrfs_fs_info *fs_info = root->fs_info;
3218 	struct btrfs_ioctl_balance_args *bargs;
3219 	struct btrfs_balance_control *bctl;
3220 	int ret;
3221 
3222 	if (!capable(CAP_SYS_ADMIN))
3223 		return -EPERM;
3224 
3225 	if (fs_info->sb->s_flags & MS_RDONLY)
3226 		return -EROFS;
3227 
3228 	mutex_lock(&fs_info->volume_mutex);
3229 	mutex_lock(&fs_info->balance_mutex);
3230 
3231 	if (arg) {
3232 		bargs = memdup_user(arg, sizeof(*bargs));
3233 		if (IS_ERR(bargs)) {
3234 			ret = PTR_ERR(bargs);
3235 			goto out;
3236 		}
3237 
3238 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3239 			if (!fs_info->balance_ctl) {
3240 				ret = -ENOTCONN;
3241 				goto out_bargs;
3242 			}
3243 
3244 			bctl = fs_info->balance_ctl;
3245 			spin_lock(&fs_info->balance_lock);
3246 			bctl->flags |= BTRFS_BALANCE_RESUME;
3247 			spin_unlock(&fs_info->balance_lock);
3248 
3249 			goto do_balance;
3250 		}
3251 	} else {
3252 		bargs = NULL;
3253 	}
3254 
3255 	if (fs_info->balance_ctl) {
3256 		ret = -EINPROGRESS;
3257 		goto out_bargs;
3258 	}
3259 
3260 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3261 	if (!bctl) {
3262 		ret = -ENOMEM;
3263 		goto out_bargs;
3264 	}
3265 
3266 	bctl->fs_info = fs_info;
3267 	if (arg) {
3268 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3269 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3270 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3271 
3272 		bctl->flags = bargs->flags;
3273 	} else {
3274 		/* balance everything - no filters */
3275 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3276 	}
3277 
3278 do_balance:
3279 	ret = btrfs_balance(bctl, bargs);
3280 	/*
3281 	 * bctl is freed in __cancel_balance or in free_fs_info if
3282 	 * restriper was paused all the way until unmount
3283 	 */
3284 	if (arg) {
3285 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3286 			ret = -EFAULT;
3287 	}
3288 
3289 out_bargs:
3290 	kfree(bargs);
3291 out:
3292 	mutex_unlock(&fs_info->balance_mutex);
3293 	mutex_unlock(&fs_info->volume_mutex);
3294 	return ret;
3295 }
3296 
3297 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3298 {
3299 	if (!capable(CAP_SYS_ADMIN))
3300 		return -EPERM;
3301 
3302 	switch (cmd) {
3303 	case BTRFS_BALANCE_CTL_PAUSE:
3304 		return btrfs_pause_balance(root->fs_info);
3305 	case BTRFS_BALANCE_CTL_CANCEL:
3306 		return btrfs_cancel_balance(root->fs_info);
3307 	}
3308 
3309 	return -EINVAL;
3310 }
3311 
3312 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3313 					 void __user *arg)
3314 {
3315 	struct btrfs_fs_info *fs_info = root->fs_info;
3316 	struct btrfs_ioctl_balance_args *bargs;
3317 	int ret = 0;
3318 
3319 	if (!capable(CAP_SYS_ADMIN))
3320 		return -EPERM;
3321 
3322 	mutex_lock(&fs_info->balance_mutex);
3323 	if (!fs_info->balance_ctl) {
3324 		ret = -ENOTCONN;
3325 		goto out;
3326 	}
3327 
3328 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3329 	if (!bargs) {
3330 		ret = -ENOMEM;
3331 		goto out;
3332 	}
3333 
3334 	update_ioctl_balance_args(fs_info, 1, bargs);
3335 
3336 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3337 		ret = -EFAULT;
3338 
3339 	kfree(bargs);
3340 out:
3341 	mutex_unlock(&fs_info->balance_mutex);
3342 	return ret;
3343 }
3344 
3345 long btrfs_ioctl(struct file *file, unsigned int
3346 		cmd, unsigned long arg)
3347 {
3348 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3349 	void __user *argp = (void __user *)arg;
3350 
3351 	switch (cmd) {
3352 	case FS_IOC_GETFLAGS:
3353 		return btrfs_ioctl_getflags(file, argp);
3354 	case FS_IOC_SETFLAGS:
3355 		return btrfs_ioctl_setflags(file, argp);
3356 	case FS_IOC_GETVERSION:
3357 		return btrfs_ioctl_getversion(file, argp);
3358 	case FITRIM:
3359 		return btrfs_ioctl_fitrim(file, argp);
3360 	case BTRFS_IOC_SNAP_CREATE:
3361 		return btrfs_ioctl_snap_create(file, argp, 0);
3362 	case BTRFS_IOC_SNAP_CREATE_V2:
3363 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3364 	case BTRFS_IOC_SUBVOL_CREATE:
3365 		return btrfs_ioctl_snap_create(file, argp, 1);
3366 	case BTRFS_IOC_SNAP_DESTROY:
3367 		return btrfs_ioctl_snap_destroy(file, argp);
3368 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3369 		return btrfs_ioctl_subvol_getflags(file, argp);
3370 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3371 		return btrfs_ioctl_subvol_setflags(file, argp);
3372 	case BTRFS_IOC_DEFAULT_SUBVOL:
3373 		return btrfs_ioctl_default_subvol(file, argp);
3374 	case BTRFS_IOC_DEFRAG:
3375 		return btrfs_ioctl_defrag(file, NULL);
3376 	case BTRFS_IOC_DEFRAG_RANGE:
3377 		return btrfs_ioctl_defrag(file, argp);
3378 	case BTRFS_IOC_RESIZE:
3379 		return btrfs_ioctl_resize(root, argp);
3380 	case BTRFS_IOC_ADD_DEV:
3381 		return btrfs_ioctl_add_dev(root, argp);
3382 	case BTRFS_IOC_RM_DEV:
3383 		return btrfs_ioctl_rm_dev(root, argp);
3384 	case BTRFS_IOC_FS_INFO:
3385 		return btrfs_ioctl_fs_info(root, argp);
3386 	case BTRFS_IOC_DEV_INFO:
3387 		return btrfs_ioctl_dev_info(root, argp);
3388 	case BTRFS_IOC_BALANCE:
3389 		return btrfs_ioctl_balance(root, NULL);
3390 	case BTRFS_IOC_CLONE:
3391 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3392 	case BTRFS_IOC_CLONE_RANGE:
3393 		return btrfs_ioctl_clone_range(file, argp);
3394 	case BTRFS_IOC_TRANS_START:
3395 		return btrfs_ioctl_trans_start(file);
3396 	case BTRFS_IOC_TRANS_END:
3397 		return btrfs_ioctl_trans_end(file);
3398 	case BTRFS_IOC_TREE_SEARCH:
3399 		return btrfs_ioctl_tree_search(file, argp);
3400 	case BTRFS_IOC_INO_LOOKUP:
3401 		return btrfs_ioctl_ino_lookup(file, argp);
3402 	case BTRFS_IOC_INO_PATHS:
3403 		return btrfs_ioctl_ino_to_path(root, argp);
3404 	case BTRFS_IOC_LOGICAL_INO:
3405 		return btrfs_ioctl_logical_to_ino(root, argp);
3406 	case BTRFS_IOC_SPACE_INFO:
3407 		return btrfs_ioctl_space_info(root, argp);
3408 	case BTRFS_IOC_SYNC:
3409 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3410 		return 0;
3411 	case BTRFS_IOC_START_SYNC:
3412 		return btrfs_ioctl_start_sync(file, argp);
3413 	case BTRFS_IOC_WAIT_SYNC:
3414 		return btrfs_ioctl_wait_sync(file, argp);
3415 	case BTRFS_IOC_SCRUB:
3416 		return btrfs_ioctl_scrub(root, argp);
3417 	case BTRFS_IOC_SCRUB_CANCEL:
3418 		return btrfs_ioctl_scrub_cancel(root, argp);
3419 	case BTRFS_IOC_SCRUB_PROGRESS:
3420 		return btrfs_ioctl_scrub_progress(root, argp);
3421 	case BTRFS_IOC_BALANCE_V2:
3422 		return btrfs_ioctl_balance(root, argp);
3423 	case BTRFS_IOC_BALANCE_CTL:
3424 		return btrfs_ioctl_balance_ctl(root, arg);
3425 	case BTRFS_IOC_BALANCE_PROGRESS:
3426 		return btrfs_ioctl_balance_progress(root, argp);
3427 	}
3428 
3429 	return -ENOTTY;
3430 }
3431