xref: /linux/fs/btrfs/ioctl.c (revision 1525e06e4414dbe053ea10cbd8cc03bb150e8e27)
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_inc_iversion(inode);
265 	inode->i_ctime = CURRENT_TIME;
266 	ret = btrfs_update_inode(trans, root, inode);
267 
268 	btrfs_end_transaction(trans, root);
269  out_drop:
270 	if (ret) {
271 		ip->flags = ip_oldflags;
272 		inode->i_flags = i_oldflags;
273 	}
274 
275 	mnt_drop_write_file(file);
276  out_unlock:
277 	mutex_unlock(&inode->i_mutex);
278 	return ret;
279 }
280 
281 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
282 {
283 	struct inode *inode = file->f_path.dentry->d_inode;
284 
285 	return put_user(inode->i_generation, arg);
286 }
287 
288 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
289 {
290 	struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
291 	struct btrfs_device *device;
292 	struct request_queue *q;
293 	struct fstrim_range range;
294 	u64 minlen = ULLONG_MAX;
295 	u64 num_devices = 0;
296 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
297 	int ret;
298 
299 	if (!capable(CAP_SYS_ADMIN))
300 		return -EPERM;
301 
302 	rcu_read_lock();
303 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
304 				dev_list) {
305 		if (!device->bdev)
306 			continue;
307 		q = bdev_get_queue(device->bdev);
308 		if (blk_queue_discard(q)) {
309 			num_devices++;
310 			minlen = min((u64)q->limits.discard_granularity,
311 				     minlen);
312 		}
313 	}
314 	rcu_read_unlock();
315 
316 	if (!num_devices)
317 		return -EOPNOTSUPP;
318 	if (copy_from_user(&range, arg, sizeof(range)))
319 		return -EFAULT;
320 	if (range.start > total_bytes)
321 		return -EINVAL;
322 
323 	range.len = min(range.len, total_bytes - range.start);
324 	range.minlen = max(range.minlen, minlen);
325 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
326 	if (ret < 0)
327 		return ret;
328 
329 	if (copy_to_user(arg, &range, sizeof(range)))
330 		return -EFAULT;
331 
332 	return 0;
333 }
334 
335 static noinline int create_subvol(struct btrfs_root *root,
336 				  struct dentry *dentry,
337 				  char *name, int namelen,
338 				  u64 *async_transid)
339 {
340 	struct btrfs_trans_handle *trans;
341 	struct btrfs_key key;
342 	struct btrfs_root_item root_item;
343 	struct btrfs_inode_item *inode_item;
344 	struct extent_buffer *leaf;
345 	struct btrfs_root *new_root;
346 	struct dentry *parent = dentry->d_parent;
347 	struct inode *dir;
348 	int ret;
349 	int err;
350 	u64 objectid;
351 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
352 	u64 index = 0;
353 
354 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
355 	if (ret)
356 		return ret;
357 
358 	dir = parent->d_inode;
359 
360 	/*
361 	 * 1 - inode item
362 	 * 2 - refs
363 	 * 1 - root item
364 	 * 2 - dir items
365 	 */
366 	trans = btrfs_start_transaction(root, 6);
367 	if (IS_ERR(trans))
368 		return PTR_ERR(trans);
369 
370 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
371 				      0, objectid, NULL, 0, 0, 0);
372 	if (IS_ERR(leaf)) {
373 		ret = PTR_ERR(leaf);
374 		goto fail;
375 	}
376 
377 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
378 	btrfs_set_header_bytenr(leaf, leaf->start);
379 	btrfs_set_header_generation(leaf, trans->transid);
380 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
381 	btrfs_set_header_owner(leaf, objectid);
382 
383 	write_extent_buffer(leaf, root->fs_info->fsid,
384 			    (unsigned long)btrfs_header_fsid(leaf),
385 			    BTRFS_FSID_SIZE);
386 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
387 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
388 			    BTRFS_UUID_SIZE);
389 	btrfs_mark_buffer_dirty(leaf);
390 
391 	inode_item = &root_item.inode;
392 	memset(inode_item, 0, sizeof(*inode_item));
393 	inode_item->generation = cpu_to_le64(1);
394 	inode_item->size = cpu_to_le64(3);
395 	inode_item->nlink = cpu_to_le32(1);
396 	inode_item->nbytes = cpu_to_le64(root->leafsize);
397 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
398 
399 	root_item.flags = 0;
400 	root_item.byte_limit = 0;
401 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
402 
403 	btrfs_set_root_bytenr(&root_item, leaf->start);
404 	btrfs_set_root_generation(&root_item, trans->transid);
405 	btrfs_set_root_level(&root_item, 0);
406 	btrfs_set_root_refs(&root_item, 1);
407 	btrfs_set_root_used(&root_item, leaf->len);
408 	btrfs_set_root_last_snapshot(&root_item, 0);
409 
410 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
411 	root_item.drop_level = 0;
412 
413 	btrfs_tree_unlock(leaf);
414 	free_extent_buffer(leaf);
415 	leaf = NULL;
416 
417 	btrfs_set_root_dirid(&root_item, new_dirid);
418 
419 	key.objectid = objectid;
420 	key.offset = 0;
421 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
422 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
423 				&root_item);
424 	if (ret)
425 		goto fail;
426 
427 	key.offset = (u64)-1;
428 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
429 	if (IS_ERR(new_root)) {
430 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
431 		ret = PTR_ERR(new_root);
432 		goto fail;
433 	}
434 
435 	btrfs_record_root_in_trans(trans, new_root);
436 
437 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
438 	if (ret) {
439 		/* We potentially lose an unused inode item here */
440 		btrfs_abort_transaction(trans, root, ret);
441 		goto fail;
442 	}
443 
444 	/*
445 	 * insert the directory item
446 	 */
447 	ret = btrfs_set_inode_index(dir, &index);
448 	if (ret) {
449 		btrfs_abort_transaction(trans, root, ret);
450 		goto fail;
451 	}
452 
453 	ret = btrfs_insert_dir_item(trans, root,
454 				    name, namelen, dir, &key,
455 				    BTRFS_FT_DIR, index);
456 	if (ret) {
457 		btrfs_abort_transaction(trans, root, ret);
458 		goto fail;
459 	}
460 
461 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
462 	ret = btrfs_update_inode(trans, root, dir);
463 	BUG_ON(ret);
464 
465 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
466 				 objectid, root->root_key.objectid,
467 				 btrfs_ino(dir), index, name, namelen);
468 
469 	BUG_ON(ret);
470 
471 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
472 fail:
473 	if (async_transid) {
474 		*async_transid = trans->transid;
475 		err = btrfs_commit_transaction_async(trans, root, 1);
476 	} else {
477 		err = btrfs_commit_transaction(trans, root);
478 	}
479 	if (err && !ret)
480 		ret = err;
481 	return ret;
482 }
483 
484 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
485 			   char *name, int namelen, u64 *async_transid,
486 			   bool readonly)
487 {
488 	struct inode *inode;
489 	struct btrfs_pending_snapshot *pending_snapshot;
490 	struct btrfs_trans_handle *trans;
491 	int ret;
492 
493 	if (!root->ref_cows)
494 		return -EINVAL;
495 
496 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
497 	if (!pending_snapshot)
498 		return -ENOMEM;
499 
500 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
501 	pending_snapshot->dentry = dentry;
502 	pending_snapshot->root = root;
503 	pending_snapshot->readonly = readonly;
504 
505 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
506 	if (IS_ERR(trans)) {
507 		ret = PTR_ERR(trans);
508 		goto fail;
509 	}
510 
511 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
512 	BUG_ON(ret);
513 
514 	spin_lock(&root->fs_info->trans_lock);
515 	list_add(&pending_snapshot->list,
516 		 &trans->transaction->pending_snapshots);
517 	spin_unlock(&root->fs_info->trans_lock);
518 	if (async_transid) {
519 		*async_transid = trans->transid;
520 		ret = btrfs_commit_transaction_async(trans,
521 				     root->fs_info->extent_root, 1);
522 	} else {
523 		ret = btrfs_commit_transaction(trans,
524 					       root->fs_info->extent_root);
525 	}
526 	BUG_ON(ret);
527 
528 	ret = pending_snapshot->error;
529 	if (ret)
530 		goto fail;
531 
532 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
533 	if (ret)
534 		goto fail;
535 
536 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
537 	if (IS_ERR(inode)) {
538 		ret = PTR_ERR(inode);
539 		goto fail;
540 	}
541 	BUG_ON(!inode);
542 	d_instantiate(dentry, inode);
543 	ret = 0;
544 fail:
545 	kfree(pending_snapshot);
546 	return ret;
547 }
548 
549 /*  copy of check_sticky in fs/namei.c()
550 * It's inline, so penalty for filesystems that don't use sticky bit is
551 * minimal.
552 */
553 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
554 {
555 	uid_t fsuid = current_fsuid();
556 
557 	if (!(dir->i_mode & S_ISVTX))
558 		return 0;
559 	if (inode->i_uid == fsuid)
560 		return 0;
561 	if (dir->i_uid == fsuid)
562 		return 0;
563 	return !capable(CAP_FOWNER);
564 }
565 
566 /*  copy of may_delete in fs/namei.c()
567  *	Check whether we can remove a link victim from directory dir, check
568  *  whether the type of victim is right.
569  *  1. We can't do it if dir is read-only (done in permission())
570  *  2. We should have write and exec permissions on dir
571  *  3. We can't remove anything from append-only dir
572  *  4. We can't do anything with immutable dir (done in permission())
573  *  5. If the sticky bit on dir is set we should either
574  *	a. be owner of dir, or
575  *	b. be owner of victim, or
576  *	c. have CAP_FOWNER capability
577  *  6. If the victim is append-only or immutable we can't do antyhing with
578  *     links pointing to it.
579  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
580  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
581  *  9. We can't remove a root or mountpoint.
582  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
583  *     nfs_async_unlink().
584  */
585 
586 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
587 {
588 	int error;
589 
590 	if (!victim->d_inode)
591 		return -ENOENT;
592 
593 	BUG_ON(victim->d_parent->d_inode != dir);
594 	audit_inode_child(victim, dir);
595 
596 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
597 	if (error)
598 		return error;
599 	if (IS_APPEND(dir))
600 		return -EPERM;
601 	if (btrfs_check_sticky(dir, victim->d_inode)||
602 		IS_APPEND(victim->d_inode)||
603 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
604 		return -EPERM;
605 	if (isdir) {
606 		if (!S_ISDIR(victim->d_inode->i_mode))
607 			return -ENOTDIR;
608 		if (IS_ROOT(victim))
609 			return -EBUSY;
610 	} else if (S_ISDIR(victim->d_inode->i_mode))
611 		return -EISDIR;
612 	if (IS_DEADDIR(dir))
613 		return -ENOENT;
614 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
615 		return -EBUSY;
616 	return 0;
617 }
618 
619 /* copy of may_create in fs/namei.c() */
620 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
621 {
622 	if (child->d_inode)
623 		return -EEXIST;
624 	if (IS_DEADDIR(dir))
625 		return -ENOENT;
626 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
627 }
628 
629 /*
630  * Create a new subvolume below @parent.  This is largely modeled after
631  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
632  * inside this filesystem so it's quite a bit simpler.
633  */
634 static noinline int btrfs_mksubvol(struct path *parent,
635 				   char *name, int namelen,
636 				   struct btrfs_root *snap_src,
637 				   u64 *async_transid, bool readonly)
638 {
639 	struct inode *dir  = parent->dentry->d_inode;
640 	struct dentry *dentry;
641 	int error;
642 
643 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
644 
645 	dentry = lookup_one_len(name, parent->dentry, namelen);
646 	error = PTR_ERR(dentry);
647 	if (IS_ERR(dentry))
648 		goto out_unlock;
649 
650 	error = -EEXIST;
651 	if (dentry->d_inode)
652 		goto out_dput;
653 
654 	error = mnt_want_write(parent->mnt);
655 	if (error)
656 		goto out_dput;
657 
658 	error = btrfs_may_create(dir, dentry);
659 	if (error)
660 		goto out_drop_write;
661 
662 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
663 
664 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
665 		goto out_up_read;
666 
667 	if (snap_src) {
668 		error = create_snapshot(snap_src, dentry,
669 					name, namelen, async_transid, readonly);
670 	} else {
671 		error = create_subvol(BTRFS_I(dir)->root, dentry,
672 				      name, namelen, async_transid);
673 	}
674 	if (!error)
675 		fsnotify_mkdir(dir, dentry);
676 out_up_read:
677 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
678 out_drop_write:
679 	mnt_drop_write(parent->mnt);
680 out_dput:
681 	dput(dentry);
682 out_unlock:
683 	mutex_unlock(&dir->i_mutex);
684 	return error;
685 }
686 
687 /*
688  * When we're defragging a range, we don't want to kick it off again
689  * if it is really just waiting for delalloc to send it down.
690  * If we find a nice big extent or delalloc range for the bytes in the
691  * file you want to defrag, we return 0 to let you know to skip this
692  * part of the file
693  */
694 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
695 {
696 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
697 	struct extent_map *em = NULL;
698 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
699 	u64 end;
700 
701 	read_lock(&em_tree->lock);
702 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
703 	read_unlock(&em_tree->lock);
704 
705 	if (em) {
706 		end = extent_map_end(em);
707 		free_extent_map(em);
708 		if (end - offset > thresh)
709 			return 0;
710 	}
711 	/* if we already have a nice delalloc here, just stop */
712 	thresh /= 2;
713 	end = count_range_bits(io_tree, &offset, offset + thresh,
714 			       thresh, EXTENT_DELALLOC, 1);
715 	if (end >= thresh)
716 		return 0;
717 	return 1;
718 }
719 
720 /*
721  * helper function to walk through a file and find extents
722  * newer than a specific transid, and smaller than thresh.
723  *
724  * This is used by the defragging code to find new and small
725  * extents
726  */
727 static int find_new_extents(struct btrfs_root *root,
728 			    struct inode *inode, u64 newer_than,
729 			    u64 *off, int thresh)
730 {
731 	struct btrfs_path *path;
732 	struct btrfs_key min_key;
733 	struct btrfs_key max_key;
734 	struct extent_buffer *leaf;
735 	struct btrfs_file_extent_item *extent;
736 	int type;
737 	int ret;
738 	u64 ino = btrfs_ino(inode);
739 
740 	path = btrfs_alloc_path();
741 	if (!path)
742 		return -ENOMEM;
743 
744 	min_key.objectid = ino;
745 	min_key.type = BTRFS_EXTENT_DATA_KEY;
746 	min_key.offset = *off;
747 
748 	max_key.objectid = ino;
749 	max_key.type = (u8)-1;
750 	max_key.offset = (u64)-1;
751 
752 	path->keep_locks = 1;
753 
754 	while(1) {
755 		ret = btrfs_search_forward(root, &min_key, &max_key,
756 					   path, 0, newer_than);
757 		if (ret != 0)
758 			goto none;
759 		if (min_key.objectid != ino)
760 			goto none;
761 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
762 			goto none;
763 
764 		leaf = path->nodes[0];
765 		extent = btrfs_item_ptr(leaf, path->slots[0],
766 					struct btrfs_file_extent_item);
767 
768 		type = btrfs_file_extent_type(leaf, extent);
769 		if (type == BTRFS_FILE_EXTENT_REG &&
770 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
771 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
772 			*off = min_key.offset;
773 			btrfs_free_path(path);
774 			return 0;
775 		}
776 
777 		if (min_key.offset == (u64)-1)
778 			goto none;
779 
780 		min_key.offset++;
781 		btrfs_release_path(path);
782 	}
783 none:
784 	btrfs_free_path(path);
785 	return -ENOENT;
786 }
787 
788 /*
789  * Validaty check of prev em and next em:
790  * 1) no prev/next em
791  * 2) prev/next em is an hole/inline extent
792  */
793 static int check_adjacent_extents(struct inode *inode, struct extent_map *em)
794 {
795 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
796 	struct extent_map *prev = NULL, *next = NULL;
797 	int ret = 0;
798 
799 	read_lock(&em_tree->lock);
800 	prev = lookup_extent_mapping(em_tree, em->start - 1, (u64)-1);
801 	next = lookup_extent_mapping(em_tree, em->start + em->len, (u64)-1);
802 	read_unlock(&em_tree->lock);
803 
804 	if ((!prev || prev->block_start >= EXTENT_MAP_LAST_BYTE) &&
805 	    (!next || next->block_start >= EXTENT_MAP_LAST_BYTE))
806 		ret = 1;
807 	free_extent_map(prev);
808 	free_extent_map(next);
809 
810 	return ret;
811 }
812 
813 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
814 			       int thresh, u64 *last_len, u64 *skip,
815 			       u64 *defrag_end)
816 {
817 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
818 	struct extent_map *em = NULL;
819 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
820 	int ret = 1;
821 
822 	/*
823 	 * make sure that once we start defragging an extent, we keep on
824 	 * defragging it
825 	 */
826 	if (start < *defrag_end)
827 		return 1;
828 
829 	*skip = 0;
830 
831 	/*
832 	 * hopefully we have this extent in the tree already, try without
833 	 * the full extent lock
834 	 */
835 	read_lock(&em_tree->lock);
836 	em = lookup_extent_mapping(em_tree, start, len);
837 	read_unlock(&em_tree->lock);
838 
839 	if (!em) {
840 		/* get the big lock and read metadata off disk */
841 		lock_extent(io_tree, start, start + len - 1);
842 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
843 		unlock_extent(io_tree, start, start + len - 1);
844 
845 		if (IS_ERR(em))
846 			return 0;
847 	}
848 
849 	/* this will cover holes, and inline extents */
850 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
851 		ret = 0;
852 		goto out;
853 	}
854 
855 	/* If we have nothing to merge with us, just skip. */
856 	if (check_adjacent_extents(inode, em)) {
857 		ret = 0;
858 		goto out;
859 	}
860 
861 	/*
862 	 * we hit a real extent, if it is big don't bother defragging it again
863 	 */
864 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
865 		ret = 0;
866 
867 out:
868 	/*
869 	 * last_len ends up being a counter of how many bytes we've defragged.
870 	 * every time we choose not to defrag an extent, we reset *last_len
871 	 * so that the next tiny extent will force a defrag.
872 	 *
873 	 * The end result of this is that tiny extents before a single big
874 	 * extent will force at least part of that big extent to be defragged.
875 	 */
876 	if (ret) {
877 		*defrag_end = extent_map_end(em);
878 	} else {
879 		*last_len = 0;
880 		*skip = extent_map_end(em);
881 		*defrag_end = 0;
882 	}
883 
884 	free_extent_map(em);
885 	return ret;
886 }
887 
888 /*
889  * it doesn't do much good to defrag one or two pages
890  * at a time.  This pulls in a nice chunk of pages
891  * to COW and defrag.
892  *
893  * It also makes sure the delalloc code has enough
894  * dirty data to avoid making new small extents as part
895  * of the defrag
896  *
897  * It's a good idea to start RA on this range
898  * before calling this.
899  */
900 static int cluster_pages_for_defrag(struct inode *inode,
901 				    struct page **pages,
902 				    unsigned long start_index,
903 				    int num_pages)
904 {
905 	unsigned long file_end;
906 	u64 isize = i_size_read(inode);
907 	u64 page_start;
908 	u64 page_end;
909 	u64 page_cnt;
910 	int ret;
911 	int i;
912 	int i_done;
913 	struct btrfs_ordered_extent *ordered;
914 	struct extent_state *cached_state = NULL;
915 	struct extent_io_tree *tree;
916 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
917 
918 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
919 	if (!isize || start_index > file_end)
920 		return 0;
921 
922 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
923 
924 	ret = btrfs_delalloc_reserve_space(inode,
925 					   page_cnt << PAGE_CACHE_SHIFT);
926 	if (ret)
927 		return ret;
928 	i_done = 0;
929 	tree = &BTRFS_I(inode)->io_tree;
930 
931 	/* step one, lock all the pages */
932 	for (i = 0; i < page_cnt; i++) {
933 		struct page *page;
934 again:
935 		page = find_or_create_page(inode->i_mapping,
936 					   start_index + i, mask);
937 		if (!page)
938 			break;
939 
940 		page_start = page_offset(page);
941 		page_end = page_start + PAGE_CACHE_SIZE - 1;
942 		while (1) {
943 			lock_extent(tree, page_start, page_end);
944 			ordered = btrfs_lookup_ordered_extent(inode,
945 							      page_start);
946 			unlock_extent(tree, page_start, page_end);
947 			if (!ordered)
948 				break;
949 
950 			unlock_page(page);
951 			btrfs_start_ordered_extent(inode, ordered, 1);
952 			btrfs_put_ordered_extent(ordered);
953 			lock_page(page);
954 			/*
955 			 * we unlocked the page above, so we need check if
956 			 * it was released or not.
957 			 */
958 			if (page->mapping != inode->i_mapping) {
959 				unlock_page(page);
960 				page_cache_release(page);
961 				goto again;
962 			}
963 		}
964 
965 		if (!PageUptodate(page)) {
966 			btrfs_readpage(NULL, page);
967 			lock_page(page);
968 			if (!PageUptodate(page)) {
969 				unlock_page(page);
970 				page_cache_release(page);
971 				ret = -EIO;
972 				break;
973 			}
974 		}
975 
976 		if (page->mapping != inode->i_mapping) {
977 			unlock_page(page);
978 			page_cache_release(page);
979 			goto again;
980 		}
981 
982 		pages[i] = page;
983 		i_done++;
984 	}
985 	if (!i_done || ret)
986 		goto out;
987 
988 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
989 		goto out;
990 
991 	/*
992 	 * so now we have a nice long stream of locked
993 	 * and up to date pages, lets wait on them
994 	 */
995 	for (i = 0; i < i_done; i++)
996 		wait_on_page_writeback(pages[i]);
997 
998 	page_start = page_offset(pages[0]);
999 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1000 
1001 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1002 			 page_start, page_end - 1, 0, &cached_state);
1003 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1004 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1005 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1006 			  GFP_NOFS);
1007 
1008 	if (i_done != page_cnt) {
1009 		spin_lock(&BTRFS_I(inode)->lock);
1010 		BTRFS_I(inode)->outstanding_extents++;
1011 		spin_unlock(&BTRFS_I(inode)->lock);
1012 		btrfs_delalloc_release_space(inode,
1013 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1014 	}
1015 
1016 
1017 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1018 				  &cached_state);
1019 
1020 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1021 			     page_start, page_end - 1, &cached_state,
1022 			     GFP_NOFS);
1023 
1024 	for (i = 0; i < i_done; i++) {
1025 		clear_page_dirty_for_io(pages[i]);
1026 		ClearPageChecked(pages[i]);
1027 		set_page_extent_mapped(pages[i]);
1028 		set_page_dirty(pages[i]);
1029 		unlock_page(pages[i]);
1030 		page_cache_release(pages[i]);
1031 	}
1032 	return i_done;
1033 out:
1034 	for (i = 0; i < i_done; i++) {
1035 		unlock_page(pages[i]);
1036 		page_cache_release(pages[i]);
1037 	}
1038 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1039 	return ret;
1040 
1041 }
1042 
1043 int btrfs_defrag_file(struct inode *inode, struct file *file,
1044 		      struct btrfs_ioctl_defrag_range_args *range,
1045 		      u64 newer_than, unsigned long max_to_defrag)
1046 {
1047 	struct btrfs_root *root = BTRFS_I(inode)->root;
1048 	struct btrfs_super_block *disk_super;
1049 	struct file_ra_state *ra = NULL;
1050 	unsigned long last_index;
1051 	u64 isize = i_size_read(inode);
1052 	u64 features;
1053 	u64 last_len = 0;
1054 	u64 skip = 0;
1055 	u64 defrag_end = 0;
1056 	u64 newer_off = range->start;
1057 	unsigned long i;
1058 	unsigned long ra_index = 0;
1059 	int ret;
1060 	int defrag_count = 0;
1061 	int compress_type = BTRFS_COMPRESS_ZLIB;
1062 	int extent_thresh = range->extent_thresh;
1063 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1064 	int cluster = max_cluster;
1065 	u64 new_align = ~((u64)128 * 1024 - 1);
1066 	struct page **pages = NULL;
1067 
1068 	if (extent_thresh == 0)
1069 		extent_thresh = 256 * 1024;
1070 
1071 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1072 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1073 			return -EINVAL;
1074 		if (range->compress_type)
1075 			compress_type = range->compress_type;
1076 	}
1077 
1078 	if (isize == 0)
1079 		return 0;
1080 
1081 	/*
1082 	 * if we were not given a file, allocate a readahead
1083 	 * context
1084 	 */
1085 	if (!file) {
1086 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1087 		if (!ra)
1088 			return -ENOMEM;
1089 		file_ra_state_init(ra, inode->i_mapping);
1090 	} else {
1091 		ra = &file->f_ra;
1092 	}
1093 
1094 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1095 			GFP_NOFS);
1096 	if (!pages) {
1097 		ret = -ENOMEM;
1098 		goto out_ra;
1099 	}
1100 
1101 	/* find the last page to defrag */
1102 	if (range->start + range->len > range->start) {
1103 		last_index = min_t(u64, isize - 1,
1104 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1105 	} else {
1106 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1107 	}
1108 
1109 	if (newer_than) {
1110 		ret = find_new_extents(root, inode, newer_than,
1111 				       &newer_off, 64 * 1024);
1112 		if (!ret) {
1113 			range->start = newer_off;
1114 			/*
1115 			 * we always align our defrag to help keep
1116 			 * the extents in the file evenly spaced
1117 			 */
1118 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1119 		} else
1120 			goto out_ra;
1121 	} else {
1122 		i = range->start >> PAGE_CACHE_SHIFT;
1123 	}
1124 	if (!max_to_defrag)
1125 		max_to_defrag = last_index + 1;
1126 
1127 	/*
1128 	 * make writeback starts from i, so the defrag range can be
1129 	 * written sequentially.
1130 	 */
1131 	if (i < inode->i_mapping->writeback_index)
1132 		inode->i_mapping->writeback_index = i;
1133 
1134 	while (i <= last_index && defrag_count < max_to_defrag &&
1135 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1136 		PAGE_CACHE_SHIFT)) {
1137 		/*
1138 		 * make sure we stop running if someone unmounts
1139 		 * the FS
1140 		 */
1141 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1142 			break;
1143 
1144 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1145 					 PAGE_CACHE_SIZE, extent_thresh,
1146 					 &last_len, &skip, &defrag_end)) {
1147 			unsigned long next;
1148 			/*
1149 			 * the should_defrag function tells us how much to skip
1150 			 * bump our counter by the suggested amount
1151 			 */
1152 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1153 			i = max(i + 1, next);
1154 			continue;
1155 		}
1156 
1157 		if (!newer_than) {
1158 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1159 				   PAGE_CACHE_SHIFT) - i;
1160 			cluster = min(cluster, max_cluster);
1161 		} else {
1162 			cluster = max_cluster;
1163 		}
1164 
1165 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1166 			BTRFS_I(inode)->force_compress = compress_type;
1167 
1168 		if (i + cluster > ra_index) {
1169 			ra_index = max(i, ra_index);
1170 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1171 				       cluster);
1172 			ra_index += max_cluster;
1173 		}
1174 
1175 		mutex_lock(&inode->i_mutex);
1176 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1177 		if (ret < 0) {
1178 			mutex_unlock(&inode->i_mutex);
1179 			goto out_ra;
1180 		}
1181 
1182 		defrag_count += ret;
1183 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1184 		mutex_unlock(&inode->i_mutex);
1185 
1186 		if (newer_than) {
1187 			if (newer_off == (u64)-1)
1188 				break;
1189 
1190 			if (ret > 0)
1191 				i += ret;
1192 
1193 			newer_off = max(newer_off + 1,
1194 					(u64)i << PAGE_CACHE_SHIFT);
1195 
1196 			ret = find_new_extents(root, inode,
1197 					       newer_than, &newer_off,
1198 					       64 * 1024);
1199 			if (!ret) {
1200 				range->start = newer_off;
1201 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1202 			} else {
1203 				break;
1204 			}
1205 		} else {
1206 			if (ret > 0) {
1207 				i += ret;
1208 				last_len += ret << PAGE_CACHE_SHIFT;
1209 			} else {
1210 				i++;
1211 				last_len = 0;
1212 			}
1213 		}
1214 	}
1215 
1216 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1217 		filemap_flush(inode->i_mapping);
1218 
1219 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1220 		/* the filemap_flush will queue IO into the worker threads, but
1221 		 * we have to make sure the IO is actually started and that
1222 		 * ordered extents get created before we return
1223 		 */
1224 		atomic_inc(&root->fs_info->async_submit_draining);
1225 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1226 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1227 			wait_event(root->fs_info->async_submit_wait,
1228 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1229 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1230 		}
1231 		atomic_dec(&root->fs_info->async_submit_draining);
1232 
1233 		mutex_lock(&inode->i_mutex);
1234 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1235 		mutex_unlock(&inode->i_mutex);
1236 	}
1237 
1238 	disk_super = root->fs_info->super_copy;
1239 	features = btrfs_super_incompat_flags(disk_super);
1240 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1241 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1242 		btrfs_set_super_incompat_flags(disk_super, features);
1243 	}
1244 
1245 	ret = defrag_count;
1246 
1247 out_ra:
1248 	if (!file)
1249 		kfree(ra);
1250 	kfree(pages);
1251 	return ret;
1252 }
1253 
1254 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1255 					void __user *arg)
1256 {
1257 	u64 new_size;
1258 	u64 old_size;
1259 	u64 devid = 1;
1260 	struct btrfs_ioctl_vol_args *vol_args;
1261 	struct btrfs_trans_handle *trans;
1262 	struct btrfs_device *device = NULL;
1263 	char *sizestr;
1264 	char *devstr = NULL;
1265 	int ret = 0;
1266 	int mod = 0;
1267 
1268 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1269 		return -EROFS;
1270 
1271 	if (!capable(CAP_SYS_ADMIN))
1272 		return -EPERM;
1273 
1274 	mutex_lock(&root->fs_info->volume_mutex);
1275 	if (root->fs_info->balance_ctl) {
1276 		printk(KERN_INFO "btrfs: balance in progress\n");
1277 		ret = -EINVAL;
1278 		goto out;
1279 	}
1280 
1281 	vol_args = memdup_user(arg, sizeof(*vol_args));
1282 	if (IS_ERR(vol_args)) {
1283 		ret = PTR_ERR(vol_args);
1284 		goto out;
1285 	}
1286 
1287 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1288 
1289 	sizestr = vol_args->name;
1290 	devstr = strchr(sizestr, ':');
1291 	if (devstr) {
1292 		char *end;
1293 		sizestr = devstr + 1;
1294 		*devstr = '\0';
1295 		devstr = vol_args->name;
1296 		devid = simple_strtoull(devstr, &end, 10);
1297 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1298 		       (unsigned long long)devid);
1299 	}
1300 	device = btrfs_find_device(root, devid, NULL, NULL);
1301 	if (!device) {
1302 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1303 		       (unsigned long long)devid);
1304 		ret = -EINVAL;
1305 		goto out_free;
1306 	}
1307 	if (!strcmp(sizestr, "max"))
1308 		new_size = device->bdev->bd_inode->i_size;
1309 	else {
1310 		if (sizestr[0] == '-') {
1311 			mod = -1;
1312 			sizestr++;
1313 		} else if (sizestr[0] == '+') {
1314 			mod = 1;
1315 			sizestr++;
1316 		}
1317 		new_size = memparse(sizestr, NULL);
1318 		if (new_size == 0) {
1319 			ret = -EINVAL;
1320 			goto out_free;
1321 		}
1322 	}
1323 
1324 	old_size = device->total_bytes;
1325 
1326 	if (mod < 0) {
1327 		if (new_size > old_size) {
1328 			ret = -EINVAL;
1329 			goto out_free;
1330 		}
1331 		new_size = old_size - new_size;
1332 	} else if (mod > 0) {
1333 		new_size = old_size + new_size;
1334 	}
1335 
1336 	if (new_size < 256 * 1024 * 1024) {
1337 		ret = -EINVAL;
1338 		goto out_free;
1339 	}
1340 	if (new_size > device->bdev->bd_inode->i_size) {
1341 		ret = -EFBIG;
1342 		goto out_free;
1343 	}
1344 
1345 	do_div(new_size, root->sectorsize);
1346 	new_size *= root->sectorsize;
1347 
1348 	printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1349 		device->name, (unsigned long long)new_size);
1350 
1351 	if (new_size > old_size) {
1352 		trans = btrfs_start_transaction(root, 0);
1353 		if (IS_ERR(trans)) {
1354 			ret = PTR_ERR(trans);
1355 			goto out_free;
1356 		}
1357 		ret = btrfs_grow_device(trans, device, new_size);
1358 		btrfs_commit_transaction(trans, root);
1359 	} else if (new_size < old_size) {
1360 		ret = btrfs_shrink_device(device, new_size);
1361 	}
1362 
1363 out_free:
1364 	kfree(vol_args);
1365 out:
1366 	mutex_unlock(&root->fs_info->volume_mutex);
1367 	return ret;
1368 }
1369 
1370 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1371 						    char *name,
1372 						    unsigned long fd,
1373 						    int subvol,
1374 						    u64 *transid,
1375 						    bool readonly)
1376 {
1377 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1378 	struct file *src_file;
1379 	int namelen;
1380 	int ret = 0;
1381 
1382 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1383 		return -EROFS;
1384 
1385 	namelen = strlen(name);
1386 	if (strchr(name, '/')) {
1387 		ret = -EINVAL;
1388 		goto out;
1389 	}
1390 
1391 	if (name[0] == '.' &&
1392 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1393 		ret = -EEXIST;
1394 		goto out;
1395 	}
1396 
1397 	if (subvol) {
1398 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1399 				     NULL, transid, readonly);
1400 	} else {
1401 		struct inode *src_inode;
1402 		src_file = fget(fd);
1403 		if (!src_file) {
1404 			ret = -EINVAL;
1405 			goto out;
1406 		}
1407 
1408 		src_inode = src_file->f_path.dentry->d_inode;
1409 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1410 			printk(KERN_INFO "btrfs: Snapshot src from "
1411 			       "another FS\n");
1412 			ret = -EINVAL;
1413 			fput(src_file);
1414 			goto out;
1415 		}
1416 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1417 				     BTRFS_I(src_inode)->root,
1418 				     transid, readonly);
1419 		fput(src_file);
1420 	}
1421 out:
1422 	return ret;
1423 }
1424 
1425 static noinline int btrfs_ioctl_snap_create(struct file *file,
1426 					    void __user *arg, int subvol)
1427 {
1428 	struct btrfs_ioctl_vol_args *vol_args;
1429 	int ret;
1430 
1431 	vol_args = memdup_user(arg, sizeof(*vol_args));
1432 	if (IS_ERR(vol_args))
1433 		return PTR_ERR(vol_args);
1434 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1435 
1436 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1437 					      vol_args->fd, subvol,
1438 					      NULL, false);
1439 
1440 	kfree(vol_args);
1441 	return ret;
1442 }
1443 
1444 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1445 					       void __user *arg, int subvol)
1446 {
1447 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1448 	int ret;
1449 	u64 transid = 0;
1450 	u64 *ptr = NULL;
1451 	bool readonly = false;
1452 
1453 	vol_args = memdup_user(arg, sizeof(*vol_args));
1454 	if (IS_ERR(vol_args))
1455 		return PTR_ERR(vol_args);
1456 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1457 
1458 	if (vol_args->flags &
1459 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1460 		ret = -EOPNOTSUPP;
1461 		goto out;
1462 	}
1463 
1464 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1465 		ptr = &transid;
1466 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1467 		readonly = true;
1468 
1469 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1470 					      vol_args->fd, subvol,
1471 					      ptr, readonly);
1472 
1473 	if (ret == 0 && ptr &&
1474 	    copy_to_user(arg +
1475 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1476 				  transid), ptr, sizeof(*ptr)))
1477 		ret = -EFAULT;
1478 out:
1479 	kfree(vol_args);
1480 	return ret;
1481 }
1482 
1483 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1484 						void __user *arg)
1485 {
1486 	struct inode *inode = fdentry(file)->d_inode;
1487 	struct btrfs_root *root = BTRFS_I(inode)->root;
1488 	int ret = 0;
1489 	u64 flags = 0;
1490 
1491 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1492 		return -EINVAL;
1493 
1494 	down_read(&root->fs_info->subvol_sem);
1495 	if (btrfs_root_readonly(root))
1496 		flags |= BTRFS_SUBVOL_RDONLY;
1497 	up_read(&root->fs_info->subvol_sem);
1498 
1499 	if (copy_to_user(arg, &flags, sizeof(flags)))
1500 		ret = -EFAULT;
1501 
1502 	return ret;
1503 }
1504 
1505 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1506 					      void __user *arg)
1507 {
1508 	struct inode *inode = fdentry(file)->d_inode;
1509 	struct btrfs_root *root = BTRFS_I(inode)->root;
1510 	struct btrfs_trans_handle *trans;
1511 	u64 root_flags;
1512 	u64 flags;
1513 	int ret = 0;
1514 
1515 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1516 		return -EROFS;
1517 
1518 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1519 		return -EINVAL;
1520 
1521 	if (copy_from_user(&flags, arg, sizeof(flags)))
1522 		return -EFAULT;
1523 
1524 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1525 		return -EINVAL;
1526 
1527 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1528 		return -EOPNOTSUPP;
1529 
1530 	if (!inode_owner_or_capable(inode))
1531 		return -EACCES;
1532 
1533 	down_write(&root->fs_info->subvol_sem);
1534 
1535 	/* nothing to do */
1536 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1537 		goto out;
1538 
1539 	root_flags = btrfs_root_flags(&root->root_item);
1540 	if (flags & BTRFS_SUBVOL_RDONLY)
1541 		btrfs_set_root_flags(&root->root_item,
1542 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1543 	else
1544 		btrfs_set_root_flags(&root->root_item,
1545 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1546 
1547 	trans = btrfs_start_transaction(root, 1);
1548 	if (IS_ERR(trans)) {
1549 		ret = PTR_ERR(trans);
1550 		goto out_reset;
1551 	}
1552 
1553 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1554 				&root->root_key, &root->root_item);
1555 
1556 	btrfs_commit_transaction(trans, root);
1557 out_reset:
1558 	if (ret)
1559 		btrfs_set_root_flags(&root->root_item, root_flags);
1560 out:
1561 	up_write(&root->fs_info->subvol_sem);
1562 	return ret;
1563 }
1564 
1565 /*
1566  * helper to check if the subvolume references other subvolumes
1567  */
1568 static noinline int may_destroy_subvol(struct btrfs_root *root)
1569 {
1570 	struct btrfs_path *path;
1571 	struct btrfs_key key;
1572 	int ret;
1573 
1574 	path = btrfs_alloc_path();
1575 	if (!path)
1576 		return -ENOMEM;
1577 
1578 	key.objectid = root->root_key.objectid;
1579 	key.type = BTRFS_ROOT_REF_KEY;
1580 	key.offset = (u64)-1;
1581 
1582 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1583 				&key, path, 0, 0);
1584 	if (ret < 0)
1585 		goto out;
1586 	BUG_ON(ret == 0);
1587 
1588 	ret = 0;
1589 	if (path->slots[0] > 0) {
1590 		path->slots[0]--;
1591 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1592 		if (key.objectid == root->root_key.objectid &&
1593 		    key.type == BTRFS_ROOT_REF_KEY)
1594 			ret = -ENOTEMPTY;
1595 	}
1596 out:
1597 	btrfs_free_path(path);
1598 	return ret;
1599 }
1600 
1601 static noinline int key_in_sk(struct btrfs_key *key,
1602 			      struct btrfs_ioctl_search_key *sk)
1603 {
1604 	struct btrfs_key test;
1605 	int ret;
1606 
1607 	test.objectid = sk->min_objectid;
1608 	test.type = sk->min_type;
1609 	test.offset = sk->min_offset;
1610 
1611 	ret = btrfs_comp_cpu_keys(key, &test);
1612 	if (ret < 0)
1613 		return 0;
1614 
1615 	test.objectid = sk->max_objectid;
1616 	test.type = sk->max_type;
1617 	test.offset = sk->max_offset;
1618 
1619 	ret = btrfs_comp_cpu_keys(key, &test);
1620 	if (ret > 0)
1621 		return 0;
1622 	return 1;
1623 }
1624 
1625 static noinline int copy_to_sk(struct btrfs_root *root,
1626 			       struct btrfs_path *path,
1627 			       struct btrfs_key *key,
1628 			       struct btrfs_ioctl_search_key *sk,
1629 			       char *buf,
1630 			       unsigned long *sk_offset,
1631 			       int *num_found)
1632 {
1633 	u64 found_transid;
1634 	struct extent_buffer *leaf;
1635 	struct btrfs_ioctl_search_header sh;
1636 	unsigned long item_off;
1637 	unsigned long item_len;
1638 	int nritems;
1639 	int i;
1640 	int slot;
1641 	int ret = 0;
1642 
1643 	leaf = path->nodes[0];
1644 	slot = path->slots[0];
1645 	nritems = btrfs_header_nritems(leaf);
1646 
1647 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1648 		i = nritems;
1649 		goto advance_key;
1650 	}
1651 	found_transid = btrfs_header_generation(leaf);
1652 
1653 	for (i = slot; i < nritems; i++) {
1654 		item_off = btrfs_item_ptr_offset(leaf, i);
1655 		item_len = btrfs_item_size_nr(leaf, i);
1656 
1657 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1658 			item_len = 0;
1659 
1660 		if (sizeof(sh) + item_len + *sk_offset >
1661 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1662 			ret = 1;
1663 			goto overflow;
1664 		}
1665 
1666 		btrfs_item_key_to_cpu(leaf, key, i);
1667 		if (!key_in_sk(key, sk))
1668 			continue;
1669 
1670 		sh.objectid = key->objectid;
1671 		sh.offset = key->offset;
1672 		sh.type = key->type;
1673 		sh.len = item_len;
1674 		sh.transid = found_transid;
1675 
1676 		/* copy search result header */
1677 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1678 		*sk_offset += sizeof(sh);
1679 
1680 		if (item_len) {
1681 			char *p = buf + *sk_offset;
1682 			/* copy the item */
1683 			read_extent_buffer(leaf, p,
1684 					   item_off, item_len);
1685 			*sk_offset += item_len;
1686 		}
1687 		(*num_found)++;
1688 
1689 		if (*num_found >= sk->nr_items)
1690 			break;
1691 	}
1692 advance_key:
1693 	ret = 0;
1694 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1695 		key->offset++;
1696 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1697 		key->offset = 0;
1698 		key->type++;
1699 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1700 		key->offset = 0;
1701 		key->type = 0;
1702 		key->objectid++;
1703 	} else
1704 		ret = 1;
1705 overflow:
1706 	return ret;
1707 }
1708 
1709 static noinline int search_ioctl(struct inode *inode,
1710 				 struct btrfs_ioctl_search_args *args)
1711 {
1712 	struct btrfs_root *root;
1713 	struct btrfs_key key;
1714 	struct btrfs_key max_key;
1715 	struct btrfs_path *path;
1716 	struct btrfs_ioctl_search_key *sk = &args->key;
1717 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1718 	int ret;
1719 	int num_found = 0;
1720 	unsigned long sk_offset = 0;
1721 
1722 	path = btrfs_alloc_path();
1723 	if (!path)
1724 		return -ENOMEM;
1725 
1726 	if (sk->tree_id == 0) {
1727 		/* search the root of the inode that was passed */
1728 		root = BTRFS_I(inode)->root;
1729 	} else {
1730 		key.objectid = sk->tree_id;
1731 		key.type = BTRFS_ROOT_ITEM_KEY;
1732 		key.offset = (u64)-1;
1733 		root = btrfs_read_fs_root_no_name(info, &key);
1734 		if (IS_ERR(root)) {
1735 			printk(KERN_ERR "could not find root %llu\n",
1736 			       sk->tree_id);
1737 			btrfs_free_path(path);
1738 			return -ENOENT;
1739 		}
1740 	}
1741 
1742 	key.objectid = sk->min_objectid;
1743 	key.type = sk->min_type;
1744 	key.offset = sk->min_offset;
1745 
1746 	max_key.objectid = sk->max_objectid;
1747 	max_key.type = sk->max_type;
1748 	max_key.offset = sk->max_offset;
1749 
1750 	path->keep_locks = 1;
1751 
1752 	while(1) {
1753 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1754 					   sk->min_transid);
1755 		if (ret != 0) {
1756 			if (ret > 0)
1757 				ret = 0;
1758 			goto err;
1759 		}
1760 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1761 				 &sk_offset, &num_found);
1762 		btrfs_release_path(path);
1763 		if (ret || num_found >= sk->nr_items)
1764 			break;
1765 
1766 	}
1767 	ret = 0;
1768 err:
1769 	sk->nr_items = num_found;
1770 	btrfs_free_path(path);
1771 	return ret;
1772 }
1773 
1774 static noinline int btrfs_ioctl_tree_search(struct file *file,
1775 					   void __user *argp)
1776 {
1777 	 struct btrfs_ioctl_search_args *args;
1778 	 struct inode *inode;
1779 	 int ret;
1780 
1781 	if (!capable(CAP_SYS_ADMIN))
1782 		return -EPERM;
1783 
1784 	args = memdup_user(argp, sizeof(*args));
1785 	if (IS_ERR(args))
1786 		return PTR_ERR(args);
1787 
1788 	inode = fdentry(file)->d_inode;
1789 	ret = search_ioctl(inode, args);
1790 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1791 		ret = -EFAULT;
1792 	kfree(args);
1793 	return ret;
1794 }
1795 
1796 /*
1797  * Search INODE_REFs to identify path name of 'dirid' directory
1798  * in a 'tree_id' tree. and sets path name to 'name'.
1799  */
1800 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1801 				u64 tree_id, u64 dirid, char *name)
1802 {
1803 	struct btrfs_root *root;
1804 	struct btrfs_key key;
1805 	char *ptr;
1806 	int ret = -1;
1807 	int slot;
1808 	int len;
1809 	int total_len = 0;
1810 	struct btrfs_inode_ref *iref;
1811 	struct extent_buffer *l;
1812 	struct btrfs_path *path;
1813 
1814 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1815 		name[0]='\0';
1816 		return 0;
1817 	}
1818 
1819 	path = btrfs_alloc_path();
1820 	if (!path)
1821 		return -ENOMEM;
1822 
1823 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1824 
1825 	key.objectid = tree_id;
1826 	key.type = BTRFS_ROOT_ITEM_KEY;
1827 	key.offset = (u64)-1;
1828 	root = btrfs_read_fs_root_no_name(info, &key);
1829 	if (IS_ERR(root)) {
1830 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1831 		ret = -ENOENT;
1832 		goto out;
1833 	}
1834 
1835 	key.objectid = dirid;
1836 	key.type = BTRFS_INODE_REF_KEY;
1837 	key.offset = (u64)-1;
1838 
1839 	while(1) {
1840 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1841 		if (ret < 0)
1842 			goto out;
1843 
1844 		l = path->nodes[0];
1845 		slot = path->slots[0];
1846 		if (ret > 0 && slot > 0)
1847 			slot--;
1848 		btrfs_item_key_to_cpu(l, &key, slot);
1849 
1850 		if (ret > 0 && (key.objectid != dirid ||
1851 				key.type != BTRFS_INODE_REF_KEY)) {
1852 			ret = -ENOENT;
1853 			goto out;
1854 		}
1855 
1856 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1857 		len = btrfs_inode_ref_name_len(l, iref);
1858 		ptr -= len + 1;
1859 		total_len += len + 1;
1860 		if (ptr < name)
1861 			goto out;
1862 
1863 		*(ptr + len) = '/';
1864 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1865 
1866 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1867 			break;
1868 
1869 		btrfs_release_path(path);
1870 		key.objectid = key.offset;
1871 		key.offset = (u64)-1;
1872 		dirid = key.objectid;
1873 	}
1874 	if (ptr < name)
1875 		goto out;
1876 	memmove(name, ptr, total_len);
1877 	name[total_len]='\0';
1878 	ret = 0;
1879 out:
1880 	btrfs_free_path(path);
1881 	return ret;
1882 }
1883 
1884 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1885 					   void __user *argp)
1886 {
1887 	 struct btrfs_ioctl_ino_lookup_args *args;
1888 	 struct inode *inode;
1889 	 int ret;
1890 
1891 	if (!capable(CAP_SYS_ADMIN))
1892 		return -EPERM;
1893 
1894 	args = memdup_user(argp, sizeof(*args));
1895 	if (IS_ERR(args))
1896 		return PTR_ERR(args);
1897 
1898 	inode = fdentry(file)->d_inode;
1899 
1900 	if (args->treeid == 0)
1901 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1902 
1903 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1904 					args->treeid, args->objectid,
1905 					args->name);
1906 
1907 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1908 		ret = -EFAULT;
1909 
1910 	kfree(args);
1911 	return ret;
1912 }
1913 
1914 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1915 					     void __user *arg)
1916 {
1917 	struct dentry *parent = fdentry(file);
1918 	struct dentry *dentry;
1919 	struct inode *dir = parent->d_inode;
1920 	struct inode *inode;
1921 	struct btrfs_root *root = BTRFS_I(dir)->root;
1922 	struct btrfs_root *dest = NULL;
1923 	struct btrfs_ioctl_vol_args *vol_args;
1924 	struct btrfs_trans_handle *trans;
1925 	int namelen;
1926 	int ret;
1927 	int err = 0;
1928 
1929 	vol_args = memdup_user(arg, sizeof(*vol_args));
1930 	if (IS_ERR(vol_args))
1931 		return PTR_ERR(vol_args);
1932 
1933 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1934 	namelen = strlen(vol_args->name);
1935 	if (strchr(vol_args->name, '/') ||
1936 	    strncmp(vol_args->name, "..", namelen) == 0) {
1937 		err = -EINVAL;
1938 		goto out;
1939 	}
1940 
1941 	err = mnt_want_write_file(file);
1942 	if (err)
1943 		goto out;
1944 
1945 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1946 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1947 	if (IS_ERR(dentry)) {
1948 		err = PTR_ERR(dentry);
1949 		goto out_unlock_dir;
1950 	}
1951 
1952 	if (!dentry->d_inode) {
1953 		err = -ENOENT;
1954 		goto out_dput;
1955 	}
1956 
1957 	inode = dentry->d_inode;
1958 	dest = BTRFS_I(inode)->root;
1959 	if (!capable(CAP_SYS_ADMIN)){
1960 		/*
1961 		 * Regular user.  Only allow this with a special mount
1962 		 * option, when the user has write+exec access to the
1963 		 * subvol root, and when rmdir(2) would have been
1964 		 * allowed.
1965 		 *
1966 		 * Note that this is _not_ check that the subvol is
1967 		 * empty or doesn't contain data that we wouldn't
1968 		 * otherwise be able to delete.
1969 		 *
1970 		 * Users who want to delete empty subvols should try
1971 		 * rmdir(2).
1972 		 */
1973 		err = -EPERM;
1974 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1975 			goto out_dput;
1976 
1977 		/*
1978 		 * Do not allow deletion if the parent dir is the same
1979 		 * as the dir to be deleted.  That means the ioctl
1980 		 * must be called on the dentry referencing the root
1981 		 * of the subvol, not a random directory contained
1982 		 * within it.
1983 		 */
1984 		err = -EINVAL;
1985 		if (root == dest)
1986 			goto out_dput;
1987 
1988 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1989 		if (err)
1990 			goto out_dput;
1991 
1992 		/* check if subvolume may be deleted by a non-root user */
1993 		err = btrfs_may_delete(dir, dentry, 1);
1994 		if (err)
1995 			goto out_dput;
1996 	}
1997 
1998 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1999 		err = -EINVAL;
2000 		goto out_dput;
2001 	}
2002 
2003 	mutex_lock(&inode->i_mutex);
2004 	err = d_invalidate(dentry);
2005 	if (err)
2006 		goto out_unlock;
2007 
2008 	down_write(&root->fs_info->subvol_sem);
2009 
2010 	err = may_destroy_subvol(dest);
2011 	if (err)
2012 		goto out_up_write;
2013 
2014 	trans = btrfs_start_transaction(root, 0);
2015 	if (IS_ERR(trans)) {
2016 		err = PTR_ERR(trans);
2017 		goto out_up_write;
2018 	}
2019 	trans->block_rsv = &root->fs_info->global_block_rsv;
2020 
2021 	ret = btrfs_unlink_subvol(trans, root, dir,
2022 				dest->root_key.objectid,
2023 				dentry->d_name.name,
2024 				dentry->d_name.len);
2025 	if (ret) {
2026 		err = ret;
2027 		btrfs_abort_transaction(trans, root, ret);
2028 		goto out_end_trans;
2029 	}
2030 
2031 	btrfs_record_root_in_trans(trans, dest);
2032 
2033 	memset(&dest->root_item.drop_progress, 0,
2034 		sizeof(dest->root_item.drop_progress));
2035 	dest->root_item.drop_level = 0;
2036 	btrfs_set_root_refs(&dest->root_item, 0);
2037 
2038 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2039 		ret = btrfs_insert_orphan_item(trans,
2040 					root->fs_info->tree_root,
2041 					dest->root_key.objectid);
2042 		if (ret) {
2043 			btrfs_abort_transaction(trans, root, ret);
2044 			err = ret;
2045 			goto out_end_trans;
2046 		}
2047 	}
2048 out_end_trans:
2049 	ret = btrfs_end_transaction(trans, root);
2050 	if (ret && !err)
2051 		err = ret;
2052 	inode->i_flags |= S_DEAD;
2053 out_up_write:
2054 	up_write(&root->fs_info->subvol_sem);
2055 out_unlock:
2056 	mutex_unlock(&inode->i_mutex);
2057 	if (!err) {
2058 		shrink_dcache_sb(root->fs_info->sb);
2059 		btrfs_invalidate_inodes(dest);
2060 		d_delete(dentry);
2061 	}
2062 out_dput:
2063 	dput(dentry);
2064 out_unlock_dir:
2065 	mutex_unlock(&dir->i_mutex);
2066 	mnt_drop_write_file(file);
2067 out:
2068 	kfree(vol_args);
2069 	return err;
2070 }
2071 
2072 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2073 {
2074 	struct inode *inode = fdentry(file)->d_inode;
2075 	struct btrfs_root *root = BTRFS_I(inode)->root;
2076 	struct btrfs_ioctl_defrag_range_args *range;
2077 	int ret;
2078 
2079 	if (btrfs_root_readonly(root))
2080 		return -EROFS;
2081 
2082 	ret = mnt_want_write_file(file);
2083 	if (ret)
2084 		return ret;
2085 
2086 	switch (inode->i_mode & S_IFMT) {
2087 	case S_IFDIR:
2088 		if (!capable(CAP_SYS_ADMIN)) {
2089 			ret = -EPERM;
2090 			goto out;
2091 		}
2092 		ret = btrfs_defrag_root(root, 0);
2093 		if (ret)
2094 			goto out;
2095 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2096 		break;
2097 	case S_IFREG:
2098 		if (!(file->f_mode & FMODE_WRITE)) {
2099 			ret = -EINVAL;
2100 			goto out;
2101 		}
2102 
2103 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2104 		if (!range) {
2105 			ret = -ENOMEM;
2106 			goto out;
2107 		}
2108 
2109 		if (argp) {
2110 			if (copy_from_user(range, argp,
2111 					   sizeof(*range))) {
2112 				ret = -EFAULT;
2113 				kfree(range);
2114 				goto out;
2115 			}
2116 			/* compression requires us to start the IO */
2117 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2118 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2119 				range->extent_thresh = (u32)-1;
2120 			}
2121 		} else {
2122 			/* the rest are all set to zero by kzalloc */
2123 			range->len = (u64)-1;
2124 		}
2125 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2126 					range, 0, 0);
2127 		if (ret > 0)
2128 			ret = 0;
2129 		kfree(range);
2130 		break;
2131 	default:
2132 		ret = -EINVAL;
2133 	}
2134 out:
2135 	mnt_drop_write_file(file);
2136 	return ret;
2137 }
2138 
2139 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2140 {
2141 	struct btrfs_ioctl_vol_args *vol_args;
2142 	int ret;
2143 
2144 	if (!capable(CAP_SYS_ADMIN))
2145 		return -EPERM;
2146 
2147 	mutex_lock(&root->fs_info->volume_mutex);
2148 	if (root->fs_info->balance_ctl) {
2149 		printk(KERN_INFO "btrfs: balance in progress\n");
2150 		ret = -EINVAL;
2151 		goto out;
2152 	}
2153 
2154 	vol_args = memdup_user(arg, sizeof(*vol_args));
2155 	if (IS_ERR(vol_args)) {
2156 		ret = PTR_ERR(vol_args);
2157 		goto out;
2158 	}
2159 
2160 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2161 	ret = btrfs_init_new_device(root, vol_args->name);
2162 
2163 	kfree(vol_args);
2164 out:
2165 	mutex_unlock(&root->fs_info->volume_mutex);
2166 	return ret;
2167 }
2168 
2169 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2170 {
2171 	struct btrfs_ioctl_vol_args *vol_args;
2172 	int ret;
2173 
2174 	if (!capable(CAP_SYS_ADMIN))
2175 		return -EPERM;
2176 
2177 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2178 		return -EROFS;
2179 
2180 	mutex_lock(&root->fs_info->volume_mutex);
2181 	if (root->fs_info->balance_ctl) {
2182 		printk(KERN_INFO "btrfs: balance in progress\n");
2183 		ret = -EINVAL;
2184 		goto out;
2185 	}
2186 
2187 	vol_args = memdup_user(arg, sizeof(*vol_args));
2188 	if (IS_ERR(vol_args)) {
2189 		ret = PTR_ERR(vol_args);
2190 		goto out;
2191 	}
2192 
2193 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2194 	ret = btrfs_rm_device(root, vol_args->name);
2195 
2196 	kfree(vol_args);
2197 out:
2198 	mutex_unlock(&root->fs_info->volume_mutex);
2199 	return ret;
2200 }
2201 
2202 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2203 {
2204 	struct btrfs_ioctl_fs_info_args *fi_args;
2205 	struct btrfs_device *device;
2206 	struct btrfs_device *next;
2207 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2208 	int ret = 0;
2209 
2210 	if (!capable(CAP_SYS_ADMIN))
2211 		return -EPERM;
2212 
2213 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2214 	if (!fi_args)
2215 		return -ENOMEM;
2216 
2217 	fi_args->num_devices = fs_devices->num_devices;
2218 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2219 
2220 	mutex_lock(&fs_devices->device_list_mutex);
2221 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2222 		if (device->devid > fi_args->max_id)
2223 			fi_args->max_id = device->devid;
2224 	}
2225 	mutex_unlock(&fs_devices->device_list_mutex);
2226 
2227 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2228 		ret = -EFAULT;
2229 
2230 	kfree(fi_args);
2231 	return ret;
2232 }
2233 
2234 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2235 {
2236 	struct btrfs_ioctl_dev_info_args *di_args;
2237 	struct btrfs_device *dev;
2238 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2239 	int ret = 0;
2240 	char *s_uuid = NULL;
2241 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2242 
2243 	if (!capable(CAP_SYS_ADMIN))
2244 		return -EPERM;
2245 
2246 	di_args = memdup_user(arg, sizeof(*di_args));
2247 	if (IS_ERR(di_args))
2248 		return PTR_ERR(di_args);
2249 
2250 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2251 		s_uuid = di_args->uuid;
2252 
2253 	mutex_lock(&fs_devices->device_list_mutex);
2254 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2255 	mutex_unlock(&fs_devices->device_list_mutex);
2256 
2257 	if (!dev) {
2258 		ret = -ENODEV;
2259 		goto out;
2260 	}
2261 
2262 	di_args->devid = dev->devid;
2263 	di_args->bytes_used = dev->bytes_used;
2264 	di_args->total_bytes = dev->total_bytes;
2265 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2266 	if (dev->name) {
2267 		strncpy(di_args->path, dev->name, sizeof(di_args->path));
2268 		di_args->path[sizeof(di_args->path) - 1] = 0;
2269 	} else {
2270 		di_args->path[0] = '\0';
2271 	}
2272 
2273 out:
2274 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2275 		ret = -EFAULT;
2276 
2277 	kfree(di_args);
2278 	return ret;
2279 }
2280 
2281 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2282 				       u64 off, u64 olen, u64 destoff)
2283 {
2284 	struct inode *inode = fdentry(file)->d_inode;
2285 	struct btrfs_root *root = BTRFS_I(inode)->root;
2286 	struct file *src_file;
2287 	struct inode *src;
2288 	struct btrfs_trans_handle *trans;
2289 	struct btrfs_path *path;
2290 	struct extent_buffer *leaf;
2291 	char *buf;
2292 	struct btrfs_key key;
2293 	u32 nritems;
2294 	int slot;
2295 	int ret;
2296 	u64 len = olen;
2297 	u64 bs = root->fs_info->sb->s_blocksize;
2298 	u64 hint_byte;
2299 
2300 	/*
2301 	 * TODO:
2302 	 * - split compressed inline extents.  annoying: we need to
2303 	 *   decompress into destination's address_space (the file offset
2304 	 *   may change, so source mapping won't do), then recompress (or
2305 	 *   otherwise reinsert) a subrange.
2306 	 * - allow ranges within the same file to be cloned (provided
2307 	 *   they don't overlap)?
2308 	 */
2309 
2310 	/* the destination must be opened for writing */
2311 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2312 		return -EINVAL;
2313 
2314 	if (btrfs_root_readonly(root))
2315 		return -EROFS;
2316 
2317 	ret = mnt_want_write_file(file);
2318 	if (ret)
2319 		return ret;
2320 
2321 	src_file = fget(srcfd);
2322 	if (!src_file) {
2323 		ret = -EBADF;
2324 		goto out_drop_write;
2325 	}
2326 
2327 	src = src_file->f_dentry->d_inode;
2328 
2329 	ret = -EINVAL;
2330 	if (src == inode)
2331 		goto out_fput;
2332 
2333 	/* the src must be open for reading */
2334 	if (!(src_file->f_mode & FMODE_READ))
2335 		goto out_fput;
2336 
2337 	/* don't make the dst file partly checksummed */
2338 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2339 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2340 		goto out_fput;
2341 
2342 	ret = -EISDIR;
2343 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2344 		goto out_fput;
2345 
2346 	ret = -EXDEV;
2347 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2348 		goto out_fput;
2349 
2350 	ret = -ENOMEM;
2351 	buf = vmalloc(btrfs_level_size(root, 0));
2352 	if (!buf)
2353 		goto out_fput;
2354 
2355 	path = btrfs_alloc_path();
2356 	if (!path) {
2357 		vfree(buf);
2358 		goto out_fput;
2359 	}
2360 	path->reada = 2;
2361 
2362 	if (inode < src) {
2363 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2364 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2365 	} else {
2366 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2367 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2368 	}
2369 
2370 	/* determine range to clone */
2371 	ret = -EINVAL;
2372 	if (off + len > src->i_size || off + len < off)
2373 		goto out_unlock;
2374 	if (len == 0)
2375 		olen = len = src->i_size - off;
2376 	/* if we extend to eof, continue to block boundary */
2377 	if (off + len == src->i_size)
2378 		len = ALIGN(src->i_size, bs) - off;
2379 
2380 	/* verify the end result is block aligned */
2381 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2382 	    !IS_ALIGNED(destoff, bs))
2383 		goto out_unlock;
2384 
2385 	if (destoff > inode->i_size) {
2386 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2387 		if (ret)
2388 			goto out_unlock;
2389 	}
2390 
2391 	/* truncate page cache pages from target inode range */
2392 	truncate_inode_pages_range(&inode->i_data, destoff,
2393 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2394 
2395 	/* do any pending delalloc/csum calc on src, one way or
2396 	   another, and lock file content */
2397 	while (1) {
2398 		struct btrfs_ordered_extent *ordered;
2399 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2400 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2401 		if (!ordered &&
2402 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2403 				   EXTENT_DELALLOC, 0, NULL))
2404 			break;
2405 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2406 		if (ordered)
2407 			btrfs_put_ordered_extent(ordered);
2408 		btrfs_wait_ordered_range(src, off, len);
2409 	}
2410 
2411 	/* clone data */
2412 	key.objectid = btrfs_ino(src);
2413 	key.type = BTRFS_EXTENT_DATA_KEY;
2414 	key.offset = 0;
2415 
2416 	while (1) {
2417 		/*
2418 		 * note the key will change type as we walk through the
2419 		 * tree.
2420 		 */
2421 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2422 		if (ret < 0)
2423 			goto out;
2424 
2425 		nritems = btrfs_header_nritems(path->nodes[0]);
2426 		if (path->slots[0] >= nritems) {
2427 			ret = btrfs_next_leaf(root, path);
2428 			if (ret < 0)
2429 				goto out;
2430 			if (ret > 0)
2431 				break;
2432 			nritems = btrfs_header_nritems(path->nodes[0]);
2433 		}
2434 		leaf = path->nodes[0];
2435 		slot = path->slots[0];
2436 
2437 		btrfs_item_key_to_cpu(leaf, &key, slot);
2438 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2439 		    key.objectid != btrfs_ino(src))
2440 			break;
2441 
2442 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2443 			struct btrfs_file_extent_item *extent;
2444 			int type;
2445 			u32 size;
2446 			struct btrfs_key new_key;
2447 			u64 disko = 0, diskl = 0;
2448 			u64 datao = 0, datal = 0;
2449 			u8 comp;
2450 			u64 endoff;
2451 
2452 			size = btrfs_item_size_nr(leaf, slot);
2453 			read_extent_buffer(leaf, buf,
2454 					   btrfs_item_ptr_offset(leaf, slot),
2455 					   size);
2456 
2457 			extent = btrfs_item_ptr(leaf, slot,
2458 						struct btrfs_file_extent_item);
2459 			comp = btrfs_file_extent_compression(leaf, extent);
2460 			type = btrfs_file_extent_type(leaf, extent);
2461 			if (type == BTRFS_FILE_EXTENT_REG ||
2462 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2463 				disko = btrfs_file_extent_disk_bytenr(leaf,
2464 								      extent);
2465 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2466 								 extent);
2467 				datao = btrfs_file_extent_offset(leaf, extent);
2468 				datal = btrfs_file_extent_num_bytes(leaf,
2469 								    extent);
2470 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2471 				/* take upper bound, may be compressed */
2472 				datal = btrfs_file_extent_ram_bytes(leaf,
2473 								    extent);
2474 			}
2475 			btrfs_release_path(path);
2476 
2477 			if (key.offset + datal <= off ||
2478 			    key.offset >= off+len)
2479 				goto next;
2480 
2481 			memcpy(&new_key, &key, sizeof(new_key));
2482 			new_key.objectid = btrfs_ino(inode);
2483 			if (off <= key.offset)
2484 				new_key.offset = key.offset + destoff - off;
2485 			else
2486 				new_key.offset = destoff;
2487 
2488 			/*
2489 			 * 1 - adjusting old extent (we may have to split it)
2490 			 * 1 - add new extent
2491 			 * 1 - inode update
2492 			 */
2493 			trans = btrfs_start_transaction(root, 3);
2494 			if (IS_ERR(trans)) {
2495 				ret = PTR_ERR(trans);
2496 				goto out;
2497 			}
2498 
2499 			if (type == BTRFS_FILE_EXTENT_REG ||
2500 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2501 				/*
2502 				 *    a  | --- range to clone ---|  b
2503 				 * | ------------- extent ------------- |
2504 				 */
2505 
2506 				/* substract range b */
2507 				if (key.offset + datal > off + len)
2508 					datal = off + len - key.offset;
2509 
2510 				/* substract range a */
2511 				if (off > key.offset) {
2512 					datao += off - key.offset;
2513 					datal -= off - key.offset;
2514 				}
2515 
2516 				ret = btrfs_drop_extents(trans, inode,
2517 							 new_key.offset,
2518 							 new_key.offset + datal,
2519 							 &hint_byte, 1);
2520 				if (ret) {
2521 					btrfs_abort_transaction(trans, root,
2522 								ret);
2523 					btrfs_end_transaction(trans, root);
2524 					goto out;
2525 				}
2526 
2527 				ret = btrfs_insert_empty_item(trans, root, path,
2528 							      &new_key, size);
2529 				if (ret) {
2530 					btrfs_abort_transaction(trans, root,
2531 								ret);
2532 					btrfs_end_transaction(trans, root);
2533 					goto out;
2534 				}
2535 
2536 				leaf = path->nodes[0];
2537 				slot = path->slots[0];
2538 				write_extent_buffer(leaf, buf,
2539 					    btrfs_item_ptr_offset(leaf, slot),
2540 					    size);
2541 
2542 				extent = btrfs_item_ptr(leaf, slot,
2543 						struct btrfs_file_extent_item);
2544 
2545 				/* disko == 0 means it's a hole */
2546 				if (!disko)
2547 					datao = 0;
2548 
2549 				btrfs_set_file_extent_offset(leaf, extent,
2550 							     datao);
2551 				btrfs_set_file_extent_num_bytes(leaf, extent,
2552 								datal);
2553 				if (disko) {
2554 					inode_add_bytes(inode, datal);
2555 					ret = btrfs_inc_extent_ref(trans, root,
2556 							disko, diskl, 0,
2557 							root->root_key.objectid,
2558 							btrfs_ino(inode),
2559 							new_key.offset - datao,
2560 							0);
2561 					if (ret) {
2562 						btrfs_abort_transaction(trans,
2563 									root,
2564 									ret);
2565 						btrfs_end_transaction(trans,
2566 								      root);
2567 						goto out;
2568 
2569 					}
2570 				}
2571 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2572 				u64 skip = 0;
2573 				u64 trim = 0;
2574 				if (off > key.offset) {
2575 					skip = off - key.offset;
2576 					new_key.offset += skip;
2577 				}
2578 
2579 				if (key.offset + datal > off+len)
2580 					trim = key.offset + datal - (off+len);
2581 
2582 				if (comp && (skip || trim)) {
2583 					ret = -EINVAL;
2584 					btrfs_end_transaction(trans, root);
2585 					goto out;
2586 				}
2587 				size -= skip + trim;
2588 				datal -= skip + trim;
2589 
2590 				ret = btrfs_drop_extents(trans, inode,
2591 							 new_key.offset,
2592 							 new_key.offset + datal,
2593 							 &hint_byte, 1);
2594 				if (ret) {
2595 					btrfs_abort_transaction(trans, root,
2596 								ret);
2597 					btrfs_end_transaction(trans, root);
2598 					goto out;
2599 				}
2600 
2601 				ret = btrfs_insert_empty_item(trans, root, path,
2602 							      &new_key, size);
2603 				if (ret) {
2604 					btrfs_abort_transaction(trans, root,
2605 								ret);
2606 					btrfs_end_transaction(trans, root);
2607 					goto out;
2608 				}
2609 
2610 				if (skip) {
2611 					u32 start =
2612 					  btrfs_file_extent_calc_inline_size(0);
2613 					memmove(buf+start, buf+start+skip,
2614 						datal);
2615 				}
2616 
2617 				leaf = path->nodes[0];
2618 				slot = path->slots[0];
2619 				write_extent_buffer(leaf, buf,
2620 					    btrfs_item_ptr_offset(leaf, slot),
2621 					    size);
2622 				inode_add_bytes(inode, datal);
2623 			}
2624 
2625 			btrfs_mark_buffer_dirty(leaf);
2626 			btrfs_release_path(path);
2627 
2628 			inode_inc_iversion(inode);
2629 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2630 
2631 			/*
2632 			 * we round up to the block size at eof when
2633 			 * determining which extents to clone above,
2634 			 * but shouldn't round up the file size
2635 			 */
2636 			endoff = new_key.offset + datal;
2637 			if (endoff > destoff+olen)
2638 				endoff = destoff+olen;
2639 			if (endoff > inode->i_size)
2640 				btrfs_i_size_write(inode, endoff);
2641 
2642 			ret = btrfs_update_inode(trans, root, inode);
2643 			if (ret) {
2644 				btrfs_abort_transaction(trans, root, ret);
2645 				btrfs_end_transaction(trans, root);
2646 				goto out;
2647 			}
2648 			ret = btrfs_end_transaction(trans, root);
2649 		}
2650 next:
2651 		btrfs_release_path(path);
2652 		key.offset++;
2653 	}
2654 	ret = 0;
2655 out:
2656 	btrfs_release_path(path);
2657 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2658 out_unlock:
2659 	mutex_unlock(&src->i_mutex);
2660 	mutex_unlock(&inode->i_mutex);
2661 	vfree(buf);
2662 	btrfs_free_path(path);
2663 out_fput:
2664 	fput(src_file);
2665 out_drop_write:
2666 	mnt_drop_write_file(file);
2667 	return ret;
2668 }
2669 
2670 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2671 {
2672 	struct btrfs_ioctl_clone_range_args args;
2673 
2674 	if (copy_from_user(&args, argp, sizeof(args)))
2675 		return -EFAULT;
2676 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2677 				 args.src_length, args.dest_offset);
2678 }
2679 
2680 /*
2681  * there are many ways the trans_start and trans_end ioctls can lead
2682  * to deadlocks.  They should only be used by applications that
2683  * basically own the machine, and have a very in depth understanding
2684  * of all the possible deadlocks and enospc problems.
2685  */
2686 static long btrfs_ioctl_trans_start(struct file *file)
2687 {
2688 	struct inode *inode = fdentry(file)->d_inode;
2689 	struct btrfs_root *root = BTRFS_I(inode)->root;
2690 	struct btrfs_trans_handle *trans;
2691 	int ret;
2692 
2693 	ret = -EPERM;
2694 	if (!capable(CAP_SYS_ADMIN))
2695 		goto out;
2696 
2697 	ret = -EINPROGRESS;
2698 	if (file->private_data)
2699 		goto out;
2700 
2701 	ret = -EROFS;
2702 	if (btrfs_root_readonly(root))
2703 		goto out;
2704 
2705 	ret = mnt_want_write_file(file);
2706 	if (ret)
2707 		goto out;
2708 
2709 	atomic_inc(&root->fs_info->open_ioctl_trans);
2710 
2711 	ret = -ENOMEM;
2712 	trans = btrfs_start_ioctl_transaction(root);
2713 	if (IS_ERR(trans))
2714 		goto out_drop;
2715 
2716 	file->private_data = trans;
2717 	return 0;
2718 
2719 out_drop:
2720 	atomic_dec(&root->fs_info->open_ioctl_trans);
2721 	mnt_drop_write_file(file);
2722 out:
2723 	return ret;
2724 }
2725 
2726 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2727 {
2728 	struct inode *inode = fdentry(file)->d_inode;
2729 	struct btrfs_root *root = BTRFS_I(inode)->root;
2730 	struct btrfs_root *new_root;
2731 	struct btrfs_dir_item *di;
2732 	struct btrfs_trans_handle *trans;
2733 	struct btrfs_path *path;
2734 	struct btrfs_key location;
2735 	struct btrfs_disk_key disk_key;
2736 	struct btrfs_super_block *disk_super;
2737 	u64 features;
2738 	u64 objectid = 0;
2739 	u64 dir_id;
2740 
2741 	if (!capable(CAP_SYS_ADMIN))
2742 		return -EPERM;
2743 
2744 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2745 		return -EFAULT;
2746 
2747 	if (!objectid)
2748 		objectid = root->root_key.objectid;
2749 
2750 	location.objectid = objectid;
2751 	location.type = BTRFS_ROOT_ITEM_KEY;
2752 	location.offset = (u64)-1;
2753 
2754 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2755 	if (IS_ERR(new_root))
2756 		return PTR_ERR(new_root);
2757 
2758 	if (btrfs_root_refs(&new_root->root_item) == 0)
2759 		return -ENOENT;
2760 
2761 	path = btrfs_alloc_path();
2762 	if (!path)
2763 		return -ENOMEM;
2764 	path->leave_spinning = 1;
2765 
2766 	trans = btrfs_start_transaction(root, 1);
2767 	if (IS_ERR(trans)) {
2768 		btrfs_free_path(path);
2769 		return PTR_ERR(trans);
2770 	}
2771 
2772 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2773 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2774 				   dir_id, "default", 7, 1);
2775 	if (IS_ERR_OR_NULL(di)) {
2776 		btrfs_free_path(path);
2777 		btrfs_end_transaction(trans, root);
2778 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2779 		       "this isn't going to work\n");
2780 		return -ENOENT;
2781 	}
2782 
2783 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2784 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2785 	btrfs_mark_buffer_dirty(path->nodes[0]);
2786 	btrfs_free_path(path);
2787 
2788 	disk_super = root->fs_info->super_copy;
2789 	features = btrfs_super_incompat_flags(disk_super);
2790 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2791 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2792 		btrfs_set_super_incompat_flags(disk_super, features);
2793 	}
2794 	btrfs_end_transaction(trans, root);
2795 
2796 	return 0;
2797 }
2798 
2799 static void get_block_group_info(struct list_head *groups_list,
2800 				 struct btrfs_ioctl_space_info *space)
2801 {
2802 	struct btrfs_block_group_cache *block_group;
2803 
2804 	space->total_bytes = 0;
2805 	space->used_bytes = 0;
2806 	space->flags = 0;
2807 	list_for_each_entry(block_group, groups_list, list) {
2808 		space->flags = block_group->flags;
2809 		space->total_bytes += block_group->key.offset;
2810 		space->used_bytes +=
2811 			btrfs_block_group_used(&block_group->item);
2812 	}
2813 }
2814 
2815 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2816 {
2817 	struct btrfs_ioctl_space_args space_args;
2818 	struct btrfs_ioctl_space_info space;
2819 	struct btrfs_ioctl_space_info *dest;
2820 	struct btrfs_ioctl_space_info *dest_orig;
2821 	struct btrfs_ioctl_space_info __user *user_dest;
2822 	struct btrfs_space_info *info;
2823 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2824 		       BTRFS_BLOCK_GROUP_SYSTEM,
2825 		       BTRFS_BLOCK_GROUP_METADATA,
2826 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2827 	int num_types = 4;
2828 	int alloc_size;
2829 	int ret = 0;
2830 	u64 slot_count = 0;
2831 	int i, c;
2832 
2833 	if (copy_from_user(&space_args,
2834 			   (struct btrfs_ioctl_space_args __user *)arg,
2835 			   sizeof(space_args)))
2836 		return -EFAULT;
2837 
2838 	for (i = 0; i < num_types; i++) {
2839 		struct btrfs_space_info *tmp;
2840 
2841 		info = NULL;
2842 		rcu_read_lock();
2843 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2844 					list) {
2845 			if (tmp->flags == types[i]) {
2846 				info = tmp;
2847 				break;
2848 			}
2849 		}
2850 		rcu_read_unlock();
2851 
2852 		if (!info)
2853 			continue;
2854 
2855 		down_read(&info->groups_sem);
2856 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2857 			if (!list_empty(&info->block_groups[c]))
2858 				slot_count++;
2859 		}
2860 		up_read(&info->groups_sem);
2861 	}
2862 
2863 	/* space_slots == 0 means they are asking for a count */
2864 	if (space_args.space_slots == 0) {
2865 		space_args.total_spaces = slot_count;
2866 		goto out;
2867 	}
2868 
2869 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2870 
2871 	alloc_size = sizeof(*dest) * slot_count;
2872 
2873 	/* we generally have at most 6 or so space infos, one for each raid
2874 	 * level.  So, a whole page should be more than enough for everyone
2875 	 */
2876 	if (alloc_size > PAGE_CACHE_SIZE)
2877 		return -ENOMEM;
2878 
2879 	space_args.total_spaces = 0;
2880 	dest = kmalloc(alloc_size, GFP_NOFS);
2881 	if (!dest)
2882 		return -ENOMEM;
2883 	dest_orig = dest;
2884 
2885 	/* now we have a buffer to copy into */
2886 	for (i = 0; i < num_types; i++) {
2887 		struct btrfs_space_info *tmp;
2888 
2889 		if (!slot_count)
2890 			break;
2891 
2892 		info = NULL;
2893 		rcu_read_lock();
2894 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2895 					list) {
2896 			if (tmp->flags == types[i]) {
2897 				info = tmp;
2898 				break;
2899 			}
2900 		}
2901 		rcu_read_unlock();
2902 
2903 		if (!info)
2904 			continue;
2905 		down_read(&info->groups_sem);
2906 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2907 			if (!list_empty(&info->block_groups[c])) {
2908 				get_block_group_info(&info->block_groups[c],
2909 						     &space);
2910 				memcpy(dest, &space, sizeof(space));
2911 				dest++;
2912 				space_args.total_spaces++;
2913 				slot_count--;
2914 			}
2915 			if (!slot_count)
2916 				break;
2917 		}
2918 		up_read(&info->groups_sem);
2919 	}
2920 
2921 	user_dest = (struct btrfs_ioctl_space_info __user *)
2922 		(arg + sizeof(struct btrfs_ioctl_space_args));
2923 
2924 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2925 		ret = -EFAULT;
2926 
2927 	kfree(dest_orig);
2928 out:
2929 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2930 		ret = -EFAULT;
2931 
2932 	return ret;
2933 }
2934 
2935 /*
2936  * there are many ways the trans_start and trans_end ioctls can lead
2937  * to deadlocks.  They should only be used by applications that
2938  * basically own the machine, and have a very in depth understanding
2939  * of all the possible deadlocks and enospc problems.
2940  */
2941 long btrfs_ioctl_trans_end(struct file *file)
2942 {
2943 	struct inode *inode = fdentry(file)->d_inode;
2944 	struct btrfs_root *root = BTRFS_I(inode)->root;
2945 	struct btrfs_trans_handle *trans;
2946 
2947 	trans = file->private_data;
2948 	if (!trans)
2949 		return -EINVAL;
2950 	file->private_data = NULL;
2951 
2952 	btrfs_end_transaction(trans, root);
2953 
2954 	atomic_dec(&root->fs_info->open_ioctl_trans);
2955 
2956 	mnt_drop_write_file(file);
2957 	return 0;
2958 }
2959 
2960 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2961 {
2962 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2963 	struct btrfs_trans_handle *trans;
2964 	u64 transid;
2965 	int ret;
2966 
2967 	trans = btrfs_start_transaction(root, 0);
2968 	if (IS_ERR(trans))
2969 		return PTR_ERR(trans);
2970 	transid = trans->transid;
2971 	ret = btrfs_commit_transaction_async(trans, root, 0);
2972 	if (ret) {
2973 		btrfs_end_transaction(trans, root);
2974 		return ret;
2975 	}
2976 
2977 	if (argp)
2978 		if (copy_to_user(argp, &transid, sizeof(transid)))
2979 			return -EFAULT;
2980 	return 0;
2981 }
2982 
2983 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2984 {
2985 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2986 	u64 transid;
2987 
2988 	if (argp) {
2989 		if (copy_from_user(&transid, argp, sizeof(transid)))
2990 			return -EFAULT;
2991 	} else {
2992 		transid = 0;  /* current trans */
2993 	}
2994 	return btrfs_wait_for_commit(root, transid);
2995 }
2996 
2997 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2998 {
2999 	int ret;
3000 	struct btrfs_ioctl_scrub_args *sa;
3001 
3002 	if (!capable(CAP_SYS_ADMIN))
3003 		return -EPERM;
3004 
3005 	sa = memdup_user(arg, sizeof(*sa));
3006 	if (IS_ERR(sa))
3007 		return PTR_ERR(sa);
3008 
3009 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3010 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3011 
3012 	if (copy_to_user(arg, sa, sizeof(*sa)))
3013 		ret = -EFAULT;
3014 
3015 	kfree(sa);
3016 	return ret;
3017 }
3018 
3019 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3020 {
3021 	if (!capable(CAP_SYS_ADMIN))
3022 		return -EPERM;
3023 
3024 	return btrfs_scrub_cancel(root);
3025 }
3026 
3027 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3028 				       void __user *arg)
3029 {
3030 	struct btrfs_ioctl_scrub_args *sa;
3031 	int ret;
3032 
3033 	if (!capable(CAP_SYS_ADMIN))
3034 		return -EPERM;
3035 
3036 	sa = memdup_user(arg, sizeof(*sa));
3037 	if (IS_ERR(sa))
3038 		return PTR_ERR(sa);
3039 
3040 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3041 
3042 	if (copy_to_user(arg, sa, sizeof(*sa)))
3043 		ret = -EFAULT;
3044 
3045 	kfree(sa);
3046 	return ret;
3047 }
3048 
3049 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3050 				      void __user *arg, int reset_after_read)
3051 {
3052 	struct btrfs_ioctl_get_dev_stats *sa;
3053 	int ret;
3054 
3055 	if (reset_after_read && !capable(CAP_SYS_ADMIN))
3056 		return -EPERM;
3057 
3058 	sa = memdup_user(arg, sizeof(*sa));
3059 	if (IS_ERR(sa))
3060 		return PTR_ERR(sa);
3061 
3062 	ret = btrfs_get_dev_stats(root, sa, reset_after_read);
3063 
3064 	if (copy_to_user(arg, sa, sizeof(*sa)))
3065 		ret = -EFAULT;
3066 
3067 	kfree(sa);
3068 	return ret;
3069 }
3070 
3071 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3072 {
3073 	int ret = 0;
3074 	int i;
3075 	u64 rel_ptr;
3076 	int size;
3077 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3078 	struct inode_fs_paths *ipath = NULL;
3079 	struct btrfs_path *path;
3080 
3081 	if (!capable(CAP_SYS_ADMIN))
3082 		return -EPERM;
3083 
3084 	path = btrfs_alloc_path();
3085 	if (!path) {
3086 		ret = -ENOMEM;
3087 		goto out;
3088 	}
3089 
3090 	ipa = memdup_user(arg, sizeof(*ipa));
3091 	if (IS_ERR(ipa)) {
3092 		ret = PTR_ERR(ipa);
3093 		ipa = NULL;
3094 		goto out;
3095 	}
3096 
3097 	size = min_t(u32, ipa->size, 4096);
3098 	ipath = init_ipath(size, root, path);
3099 	if (IS_ERR(ipath)) {
3100 		ret = PTR_ERR(ipath);
3101 		ipath = NULL;
3102 		goto out;
3103 	}
3104 
3105 	ret = paths_from_inode(ipa->inum, ipath);
3106 	if (ret < 0)
3107 		goto out;
3108 
3109 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3110 		rel_ptr = ipath->fspath->val[i] -
3111 			  (u64)(unsigned long)ipath->fspath->val;
3112 		ipath->fspath->val[i] = rel_ptr;
3113 	}
3114 
3115 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3116 			   (void *)(unsigned long)ipath->fspath, size);
3117 	if (ret) {
3118 		ret = -EFAULT;
3119 		goto out;
3120 	}
3121 
3122 out:
3123 	btrfs_free_path(path);
3124 	free_ipath(ipath);
3125 	kfree(ipa);
3126 
3127 	return ret;
3128 }
3129 
3130 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3131 {
3132 	struct btrfs_data_container *inodes = ctx;
3133 	const size_t c = 3 * sizeof(u64);
3134 
3135 	if (inodes->bytes_left >= c) {
3136 		inodes->bytes_left -= c;
3137 		inodes->val[inodes->elem_cnt] = inum;
3138 		inodes->val[inodes->elem_cnt + 1] = offset;
3139 		inodes->val[inodes->elem_cnt + 2] = root;
3140 		inodes->elem_cnt += 3;
3141 	} else {
3142 		inodes->bytes_missing += c - inodes->bytes_left;
3143 		inodes->bytes_left = 0;
3144 		inodes->elem_missed += 3;
3145 	}
3146 
3147 	return 0;
3148 }
3149 
3150 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3151 					void __user *arg)
3152 {
3153 	int ret = 0;
3154 	int size;
3155 	u64 extent_item_pos;
3156 	struct btrfs_ioctl_logical_ino_args *loi;
3157 	struct btrfs_data_container *inodes = NULL;
3158 	struct btrfs_path *path = NULL;
3159 	struct btrfs_key key;
3160 
3161 	if (!capable(CAP_SYS_ADMIN))
3162 		return -EPERM;
3163 
3164 	loi = memdup_user(arg, sizeof(*loi));
3165 	if (IS_ERR(loi)) {
3166 		ret = PTR_ERR(loi);
3167 		loi = NULL;
3168 		goto out;
3169 	}
3170 
3171 	path = btrfs_alloc_path();
3172 	if (!path) {
3173 		ret = -ENOMEM;
3174 		goto out;
3175 	}
3176 
3177 	size = min_t(u32, loi->size, 4096);
3178 	inodes = init_data_container(size);
3179 	if (IS_ERR(inodes)) {
3180 		ret = PTR_ERR(inodes);
3181 		inodes = NULL;
3182 		goto out;
3183 	}
3184 
3185 	ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3186 	btrfs_release_path(path);
3187 
3188 	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3189 		ret = -ENOENT;
3190 	if (ret < 0)
3191 		goto out;
3192 
3193 	extent_item_pos = loi->logical - key.objectid;
3194 	ret = iterate_extent_inodes(root->fs_info, key.objectid,
3195 					extent_item_pos, 0, build_ino_list,
3196 					inodes);
3197 
3198 	if (ret < 0)
3199 		goto out;
3200 
3201 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3202 			   (void *)(unsigned long)inodes, size);
3203 	if (ret)
3204 		ret = -EFAULT;
3205 
3206 out:
3207 	btrfs_free_path(path);
3208 	kfree(inodes);
3209 	kfree(loi);
3210 
3211 	return ret;
3212 }
3213 
3214 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3215 			       struct btrfs_ioctl_balance_args *bargs)
3216 {
3217 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3218 
3219 	bargs->flags = bctl->flags;
3220 
3221 	if (atomic_read(&fs_info->balance_running))
3222 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3223 	if (atomic_read(&fs_info->balance_pause_req))
3224 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3225 	if (atomic_read(&fs_info->balance_cancel_req))
3226 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3227 
3228 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3229 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3230 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3231 
3232 	if (lock) {
3233 		spin_lock(&fs_info->balance_lock);
3234 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3235 		spin_unlock(&fs_info->balance_lock);
3236 	} else {
3237 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3238 	}
3239 }
3240 
3241 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3242 {
3243 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3244 	struct btrfs_fs_info *fs_info = root->fs_info;
3245 	struct btrfs_ioctl_balance_args *bargs;
3246 	struct btrfs_balance_control *bctl;
3247 	int ret;
3248 
3249 	if (!capable(CAP_SYS_ADMIN))
3250 		return -EPERM;
3251 
3252 	if (fs_info->sb->s_flags & MS_RDONLY)
3253 		return -EROFS;
3254 
3255 	ret = mnt_want_write(file->f_path.mnt);
3256 	if (ret)
3257 		return ret;
3258 
3259 	mutex_lock(&fs_info->volume_mutex);
3260 	mutex_lock(&fs_info->balance_mutex);
3261 
3262 	if (arg) {
3263 		bargs = memdup_user(arg, sizeof(*bargs));
3264 		if (IS_ERR(bargs)) {
3265 			ret = PTR_ERR(bargs);
3266 			goto out;
3267 		}
3268 
3269 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3270 			if (!fs_info->balance_ctl) {
3271 				ret = -ENOTCONN;
3272 				goto out_bargs;
3273 			}
3274 
3275 			bctl = fs_info->balance_ctl;
3276 			spin_lock(&fs_info->balance_lock);
3277 			bctl->flags |= BTRFS_BALANCE_RESUME;
3278 			spin_unlock(&fs_info->balance_lock);
3279 
3280 			goto do_balance;
3281 		}
3282 	} else {
3283 		bargs = NULL;
3284 	}
3285 
3286 	if (fs_info->balance_ctl) {
3287 		ret = -EINPROGRESS;
3288 		goto out_bargs;
3289 	}
3290 
3291 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3292 	if (!bctl) {
3293 		ret = -ENOMEM;
3294 		goto out_bargs;
3295 	}
3296 
3297 	bctl->fs_info = fs_info;
3298 	if (arg) {
3299 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3300 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3301 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3302 
3303 		bctl->flags = bargs->flags;
3304 	} else {
3305 		/* balance everything - no filters */
3306 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3307 	}
3308 
3309 do_balance:
3310 	ret = btrfs_balance(bctl, bargs);
3311 	/*
3312 	 * bctl is freed in __cancel_balance or in free_fs_info if
3313 	 * restriper was paused all the way until unmount
3314 	 */
3315 	if (arg) {
3316 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3317 			ret = -EFAULT;
3318 	}
3319 
3320 out_bargs:
3321 	kfree(bargs);
3322 out:
3323 	mutex_unlock(&fs_info->balance_mutex);
3324 	mutex_unlock(&fs_info->volume_mutex);
3325 	mnt_drop_write(file->f_path.mnt);
3326 	return ret;
3327 }
3328 
3329 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3330 {
3331 	if (!capable(CAP_SYS_ADMIN))
3332 		return -EPERM;
3333 
3334 	switch (cmd) {
3335 	case BTRFS_BALANCE_CTL_PAUSE:
3336 		return btrfs_pause_balance(root->fs_info);
3337 	case BTRFS_BALANCE_CTL_CANCEL:
3338 		return btrfs_cancel_balance(root->fs_info);
3339 	}
3340 
3341 	return -EINVAL;
3342 }
3343 
3344 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3345 					 void __user *arg)
3346 {
3347 	struct btrfs_fs_info *fs_info = root->fs_info;
3348 	struct btrfs_ioctl_balance_args *bargs;
3349 	int ret = 0;
3350 
3351 	if (!capable(CAP_SYS_ADMIN))
3352 		return -EPERM;
3353 
3354 	mutex_lock(&fs_info->balance_mutex);
3355 	if (!fs_info->balance_ctl) {
3356 		ret = -ENOTCONN;
3357 		goto out;
3358 	}
3359 
3360 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3361 	if (!bargs) {
3362 		ret = -ENOMEM;
3363 		goto out;
3364 	}
3365 
3366 	update_ioctl_balance_args(fs_info, 1, bargs);
3367 
3368 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3369 		ret = -EFAULT;
3370 
3371 	kfree(bargs);
3372 out:
3373 	mutex_unlock(&fs_info->balance_mutex);
3374 	return ret;
3375 }
3376 
3377 long btrfs_ioctl(struct file *file, unsigned int
3378 		cmd, unsigned long arg)
3379 {
3380 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3381 	void __user *argp = (void __user *)arg;
3382 
3383 	switch (cmd) {
3384 	case FS_IOC_GETFLAGS:
3385 		return btrfs_ioctl_getflags(file, argp);
3386 	case FS_IOC_SETFLAGS:
3387 		return btrfs_ioctl_setflags(file, argp);
3388 	case FS_IOC_GETVERSION:
3389 		return btrfs_ioctl_getversion(file, argp);
3390 	case FITRIM:
3391 		return btrfs_ioctl_fitrim(file, argp);
3392 	case BTRFS_IOC_SNAP_CREATE:
3393 		return btrfs_ioctl_snap_create(file, argp, 0);
3394 	case BTRFS_IOC_SNAP_CREATE_V2:
3395 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3396 	case BTRFS_IOC_SUBVOL_CREATE:
3397 		return btrfs_ioctl_snap_create(file, argp, 1);
3398 	case BTRFS_IOC_SNAP_DESTROY:
3399 		return btrfs_ioctl_snap_destroy(file, argp);
3400 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3401 		return btrfs_ioctl_subvol_getflags(file, argp);
3402 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3403 		return btrfs_ioctl_subvol_setflags(file, argp);
3404 	case BTRFS_IOC_DEFAULT_SUBVOL:
3405 		return btrfs_ioctl_default_subvol(file, argp);
3406 	case BTRFS_IOC_DEFRAG:
3407 		return btrfs_ioctl_defrag(file, NULL);
3408 	case BTRFS_IOC_DEFRAG_RANGE:
3409 		return btrfs_ioctl_defrag(file, argp);
3410 	case BTRFS_IOC_RESIZE:
3411 		return btrfs_ioctl_resize(root, argp);
3412 	case BTRFS_IOC_ADD_DEV:
3413 		return btrfs_ioctl_add_dev(root, argp);
3414 	case BTRFS_IOC_RM_DEV:
3415 		return btrfs_ioctl_rm_dev(root, argp);
3416 	case BTRFS_IOC_FS_INFO:
3417 		return btrfs_ioctl_fs_info(root, argp);
3418 	case BTRFS_IOC_DEV_INFO:
3419 		return btrfs_ioctl_dev_info(root, argp);
3420 	case BTRFS_IOC_BALANCE:
3421 		return btrfs_ioctl_balance(file, NULL);
3422 	case BTRFS_IOC_CLONE:
3423 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3424 	case BTRFS_IOC_CLONE_RANGE:
3425 		return btrfs_ioctl_clone_range(file, argp);
3426 	case BTRFS_IOC_TRANS_START:
3427 		return btrfs_ioctl_trans_start(file);
3428 	case BTRFS_IOC_TRANS_END:
3429 		return btrfs_ioctl_trans_end(file);
3430 	case BTRFS_IOC_TREE_SEARCH:
3431 		return btrfs_ioctl_tree_search(file, argp);
3432 	case BTRFS_IOC_INO_LOOKUP:
3433 		return btrfs_ioctl_ino_lookup(file, argp);
3434 	case BTRFS_IOC_INO_PATHS:
3435 		return btrfs_ioctl_ino_to_path(root, argp);
3436 	case BTRFS_IOC_LOGICAL_INO:
3437 		return btrfs_ioctl_logical_to_ino(root, argp);
3438 	case BTRFS_IOC_SPACE_INFO:
3439 		return btrfs_ioctl_space_info(root, argp);
3440 	case BTRFS_IOC_SYNC:
3441 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3442 		return 0;
3443 	case BTRFS_IOC_START_SYNC:
3444 		return btrfs_ioctl_start_sync(file, argp);
3445 	case BTRFS_IOC_WAIT_SYNC:
3446 		return btrfs_ioctl_wait_sync(file, argp);
3447 	case BTRFS_IOC_SCRUB:
3448 		return btrfs_ioctl_scrub(root, argp);
3449 	case BTRFS_IOC_SCRUB_CANCEL:
3450 		return btrfs_ioctl_scrub_cancel(root, argp);
3451 	case BTRFS_IOC_SCRUB_PROGRESS:
3452 		return btrfs_ioctl_scrub_progress(root, argp);
3453 	case BTRFS_IOC_BALANCE_V2:
3454 		return btrfs_ioctl_balance(file, argp);
3455 	case BTRFS_IOC_BALANCE_CTL:
3456 		return btrfs_ioctl_balance_ctl(root, arg);
3457 	case BTRFS_IOC_BALANCE_PROGRESS:
3458 		return btrfs_ioctl_balance_progress(root, argp);
3459 	case BTRFS_IOC_GET_DEV_STATS:
3460 		return btrfs_ioctl_get_dev_stats(root, argp, 0);
3461 	case BTRFS_IOC_GET_AND_RESET_DEV_STATS:
3462 		return btrfs_ioctl_get_dev_stats(root, argp, 1);
3463 	}
3464 
3465 	return -ENOTTY;
3466 }
3467