xref: /linux/fs/btrfs/ioctl.c (revision 9429ec96c2718c0d1e3317cf60a87a0405223814)
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 <linux/uuid.h>
45 #include "compat.h"
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "ioctl.h"
51 #include "print-tree.h"
52 #include "volumes.h"
53 #include "locking.h"
54 #include "inode-map.h"
55 #include "backref.h"
56 #include "rcu-string.h"
57 #include "send.h"
58 
59 /* Mask out flags that are inappropriate for the given type of inode. */
60 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
61 {
62 	if (S_ISDIR(mode))
63 		return flags;
64 	else if (S_ISREG(mode))
65 		return flags & ~FS_DIRSYNC_FL;
66 	else
67 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
68 }
69 
70 /*
71  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
72  */
73 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
74 {
75 	unsigned int iflags = 0;
76 
77 	if (flags & BTRFS_INODE_SYNC)
78 		iflags |= FS_SYNC_FL;
79 	if (flags & BTRFS_INODE_IMMUTABLE)
80 		iflags |= FS_IMMUTABLE_FL;
81 	if (flags & BTRFS_INODE_APPEND)
82 		iflags |= FS_APPEND_FL;
83 	if (flags & BTRFS_INODE_NODUMP)
84 		iflags |= FS_NODUMP_FL;
85 	if (flags & BTRFS_INODE_NOATIME)
86 		iflags |= FS_NOATIME_FL;
87 	if (flags & BTRFS_INODE_DIRSYNC)
88 		iflags |= FS_DIRSYNC_FL;
89 	if (flags & BTRFS_INODE_NODATACOW)
90 		iflags |= FS_NOCOW_FL;
91 
92 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
93 		iflags |= FS_COMPR_FL;
94 	else if (flags & BTRFS_INODE_NOCOMPRESS)
95 		iflags |= FS_NOCOMP_FL;
96 
97 	return iflags;
98 }
99 
100 /*
101  * Update inode->i_flags based on the btrfs internal flags.
102  */
103 void btrfs_update_iflags(struct inode *inode)
104 {
105 	struct btrfs_inode *ip = BTRFS_I(inode);
106 
107 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
108 
109 	if (ip->flags & BTRFS_INODE_SYNC)
110 		inode->i_flags |= S_SYNC;
111 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
112 		inode->i_flags |= S_IMMUTABLE;
113 	if (ip->flags & BTRFS_INODE_APPEND)
114 		inode->i_flags |= S_APPEND;
115 	if (ip->flags & BTRFS_INODE_NOATIME)
116 		inode->i_flags |= S_NOATIME;
117 	if (ip->flags & BTRFS_INODE_DIRSYNC)
118 		inode->i_flags |= S_DIRSYNC;
119 }
120 
121 /*
122  * Inherit flags from the parent inode.
123  *
124  * Currently only the compression flags and the cow flags are inherited.
125  */
126 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
127 {
128 	unsigned int flags;
129 
130 	if (!dir)
131 		return;
132 
133 	flags = BTRFS_I(dir)->flags;
134 
135 	if (flags & BTRFS_INODE_NOCOMPRESS) {
136 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
137 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
138 	} else if (flags & BTRFS_INODE_COMPRESS) {
139 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
140 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
141 	}
142 
143 	if (flags & BTRFS_INODE_NODATACOW)
144 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
145 
146 	btrfs_update_iflags(inode);
147 }
148 
149 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
150 {
151 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
152 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
153 
154 	if (copy_to_user(arg, &flags, sizeof(flags)))
155 		return -EFAULT;
156 	return 0;
157 }
158 
159 static int check_flags(unsigned int flags)
160 {
161 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
162 		      FS_NOATIME_FL | FS_NODUMP_FL | \
163 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
164 		      FS_NOCOMP_FL | FS_COMPR_FL |
165 		      FS_NOCOW_FL))
166 		return -EOPNOTSUPP;
167 
168 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
169 		return -EINVAL;
170 
171 	return 0;
172 }
173 
174 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
175 {
176 	struct inode *inode = file->f_path.dentry->d_inode;
177 	struct btrfs_inode *ip = BTRFS_I(inode);
178 	struct btrfs_root *root = ip->root;
179 	struct btrfs_trans_handle *trans;
180 	unsigned int flags, oldflags;
181 	int ret;
182 	u64 ip_oldflags;
183 	unsigned int i_oldflags;
184 
185 	if (btrfs_root_readonly(root))
186 		return -EROFS;
187 
188 	if (copy_from_user(&flags, arg, sizeof(flags)))
189 		return -EFAULT;
190 
191 	ret = check_flags(flags);
192 	if (ret)
193 		return ret;
194 
195 	if (!inode_owner_or_capable(inode))
196 		return -EACCES;
197 
198 	ret = mnt_want_write_file(file);
199 	if (ret)
200 		return ret;
201 
202 	mutex_lock(&inode->i_mutex);
203 
204 	ip_oldflags = ip->flags;
205 	i_oldflags = inode->i_flags;
206 
207 	flags = btrfs_mask_flags(inode->i_mode, flags);
208 	oldflags = btrfs_flags_to_ioctl(ip->flags);
209 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
210 		if (!capable(CAP_LINUX_IMMUTABLE)) {
211 			ret = -EPERM;
212 			goto out_unlock;
213 		}
214 	}
215 
216 	if (flags & FS_SYNC_FL)
217 		ip->flags |= BTRFS_INODE_SYNC;
218 	else
219 		ip->flags &= ~BTRFS_INODE_SYNC;
220 	if (flags & FS_IMMUTABLE_FL)
221 		ip->flags |= BTRFS_INODE_IMMUTABLE;
222 	else
223 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
224 	if (flags & FS_APPEND_FL)
225 		ip->flags |= BTRFS_INODE_APPEND;
226 	else
227 		ip->flags &= ~BTRFS_INODE_APPEND;
228 	if (flags & FS_NODUMP_FL)
229 		ip->flags |= BTRFS_INODE_NODUMP;
230 	else
231 		ip->flags &= ~BTRFS_INODE_NODUMP;
232 	if (flags & FS_NOATIME_FL)
233 		ip->flags |= BTRFS_INODE_NOATIME;
234 	else
235 		ip->flags &= ~BTRFS_INODE_NOATIME;
236 	if (flags & FS_DIRSYNC_FL)
237 		ip->flags |= BTRFS_INODE_DIRSYNC;
238 	else
239 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
240 	if (flags & FS_NOCOW_FL)
241 		ip->flags |= BTRFS_INODE_NODATACOW;
242 	else
243 		ip->flags &= ~BTRFS_INODE_NODATACOW;
244 
245 	/*
246 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
247 	 * flag may be changed automatically if compression code won't make
248 	 * things smaller.
249 	 */
250 	if (flags & FS_NOCOMP_FL) {
251 		ip->flags &= ~BTRFS_INODE_COMPRESS;
252 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
253 	} else if (flags & FS_COMPR_FL) {
254 		ip->flags |= BTRFS_INODE_COMPRESS;
255 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
256 	} else {
257 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
258 	}
259 
260 	trans = btrfs_start_transaction(root, 1);
261 	if (IS_ERR(trans)) {
262 		ret = PTR_ERR(trans);
263 		goto out_drop;
264 	}
265 
266 	btrfs_update_iflags(inode);
267 	inode_inc_iversion(inode);
268 	inode->i_ctime = CURRENT_TIME;
269 	ret = btrfs_update_inode(trans, root, inode);
270 
271 	btrfs_end_transaction(trans, root);
272  out_drop:
273 	if (ret) {
274 		ip->flags = ip_oldflags;
275 		inode->i_flags = i_oldflags;
276 	}
277 
278  out_unlock:
279 	mutex_unlock(&inode->i_mutex);
280 	mnt_drop_write_file(file);
281 	return ret;
282 }
283 
284 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
285 {
286 	struct inode *inode = file->f_path.dentry->d_inode;
287 
288 	return put_user(inode->i_generation, arg);
289 }
290 
291 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
292 {
293 	struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
294 	struct btrfs_device *device;
295 	struct request_queue *q;
296 	struct fstrim_range range;
297 	u64 minlen = ULLONG_MAX;
298 	u64 num_devices = 0;
299 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
300 	int ret;
301 
302 	if (!capable(CAP_SYS_ADMIN))
303 		return -EPERM;
304 
305 	rcu_read_lock();
306 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
307 				dev_list) {
308 		if (!device->bdev)
309 			continue;
310 		q = bdev_get_queue(device->bdev);
311 		if (blk_queue_discard(q)) {
312 			num_devices++;
313 			minlen = min((u64)q->limits.discard_granularity,
314 				     minlen);
315 		}
316 	}
317 	rcu_read_unlock();
318 
319 	if (!num_devices)
320 		return -EOPNOTSUPP;
321 	if (copy_from_user(&range, arg, sizeof(range)))
322 		return -EFAULT;
323 	if (range.start > total_bytes)
324 		return -EINVAL;
325 
326 	range.len = min(range.len, total_bytes - range.start);
327 	range.minlen = max(range.minlen, minlen);
328 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
329 	if (ret < 0)
330 		return ret;
331 
332 	if (copy_to_user(arg, &range, sizeof(range)))
333 		return -EFAULT;
334 
335 	return 0;
336 }
337 
338 static noinline int create_subvol(struct btrfs_root *root,
339 				  struct dentry *dentry,
340 				  char *name, int namelen,
341 				  u64 *async_transid,
342 				  struct btrfs_qgroup_inherit **inherit)
343 {
344 	struct btrfs_trans_handle *trans;
345 	struct btrfs_key key;
346 	struct btrfs_root_item root_item;
347 	struct btrfs_inode_item *inode_item;
348 	struct extent_buffer *leaf;
349 	struct btrfs_root *new_root;
350 	struct dentry *parent = dentry->d_parent;
351 	struct inode *dir;
352 	struct timespec cur_time = CURRENT_TIME;
353 	int ret;
354 	int err;
355 	u64 objectid;
356 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
357 	u64 index = 0;
358 	uuid_le new_uuid;
359 
360 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
361 	if (ret)
362 		return ret;
363 
364 	dir = parent->d_inode;
365 
366 	/*
367 	 * 1 - inode item
368 	 * 2 - refs
369 	 * 1 - root item
370 	 * 2 - dir items
371 	 */
372 	trans = btrfs_start_transaction(root, 6);
373 	if (IS_ERR(trans))
374 		return PTR_ERR(trans);
375 
376 	ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid,
377 				   inherit ? *inherit : NULL);
378 	if (ret)
379 		goto fail;
380 
381 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
382 				      0, objectid, NULL, 0, 0, 0);
383 	if (IS_ERR(leaf)) {
384 		ret = PTR_ERR(leaf);
385 		goto fail;
386 	}
387 
388 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
389 	btrfs_set_header_bytenr(leaf, leaf->start);
390 	btrfs_set_header_generation(leaf, trans->transid);
391 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
392 	btrfs_set_header_owner(leaf, objectid);
393 
394 	write_extent_buffer(leaf, root->fs_info->fsid,
395 			    (unsigned long)btrfs_header_fsid(leaf),
396 			    BTRFS_FSID_SIZE);
397 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
398 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
399 			    BTRFS_UUID_SIZE);
400 	btrfs_mark_buffer_dirty(leaf);
401 
402 	memset(&root_item, 0, sizeof(root_item));
403 
404 	inode_item = &root_item.inode;
405 	inode_item->generation = cpu_to_le64(1);
406 	inode_item->size = cpu_to_le64(3);
407 	inode_item->nlink = cpu_to_le32(1);
408 	inode_item->nbytes = cpu_to_le64(root->leafsize);
409 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
410 
411 	root_item.flags = 0;
412 	root_item.byte_limit = 0;
413 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
414 
415 	btrfs_set_root_bytenr(&root_item, leaf->start);
416 	btrfs_set_root_generation(&root_item, trans->transid);
417 	btrfs_set_root_level(&root_item, 0);
418 	btrfs_set_root_refs(&root_item, 1);
419 	btrfs_set_root_used(&root_item, leaf->len);
420 	btrfs_set_root_last_snapshot(&root_item, 0);
421 
422 	btrfs_set_root_generation_v2(&root_item,
423 			btrfs_root_generation(&root_item));
424 	uuid_le_gen(&new_uuid);
425 	memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
426 	root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
427 	root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
428 	root_item.ctime = root_item.otime;
429 	btrfs_set_root_ctransid(&root_item, trans->transid);
430 	btrfs_set_root_otransid(&root_item, trans->transid);
431 
432 	btrfs_tree_unlock(leaf);
433 	free_extent_buffer(leaf);
434 	leaf = NULL;
435 
436 	btrfs_set_root_dirid(&root_item, new_dirid);
437 
438 	key.objectid = objectid;
439 	key.offset = 0;
440 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
441 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
442 				&root_item);
443 	if (ret)
444 		goto fail;
445 
446 	key.offset = (u64)-1;
447 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
448 	if (IS_ERR(new_root)) {
449 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
450 		ret = PTR_ERR(new_root);
451 		goto fail;
452 	}
453 
454 	btrfs_record_root_in_trans(trans, new_root);
455 
456 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
457 	if (ret) {
458 		/* We potentially lose an unused inode item here */
459 		btrfs_abort_transaction(trans, root, ret);
460 		goto fail;
461 	}
462 
463 	/*
464 	 * insert the directory item
465 	 */
466 	ret = btrfs_set_inode_index(dir, &index);
467 	if (ret) {
468 		btrfs_abort_transaction(trans, root, ret);
469 		goto fail;
470 	}
471 
472 	ret = btrfs_insert_dir_item(trans, root,
473 				    name, namelen, dir, &key,
474 				    BTRFS_FT_DIR, index);
475 	if (ret) {
476 		btrfs_abort_transaction(trans, root, ret);
477 		goto fail;
478 	}
479 
480 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
481 	ret = btrfs_update_inode(trans, root, dir);
482 	BUG_ON(ret);
483 
484 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
485 				 objectid, root->root_key.objectid,
486 				 btrfs_ino(dir), index, name, namelen);
487 
488 	BUG_ON(ret);
489 
490 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
491 fail:
492 	if (async_transid) {
493 		*async_transid = trans->transid;
494 		err = btrfs_commit_transaction_async(trans, root, 1);
495 	} else {
496 		err = btrfs_commit_transaction(trans, root);
497 	}
498 	if (err && !ret)
499 		ret = err;
500 	return ret;
501 }
502 
503 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
504 			   char *name, int namelen, u64 *async_transid,
505 			   bool readonly, struct btrfs_qgroup_inherit **inherit)
506 {
507 	struct inode *inode;
508 	struct btrfs_pending_snapshot *pending_snapshot;
509 	struct btrfs_trans_handle *trans;
510 	int ret;
511 
512 	if (!root->ref_cows)
513 		return -EINVAL;
514 
515 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
516 	if (!pending_snapshot)
517 		return -ENOMEM;
518 
519 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
520 	pending_snapshot->dentry = dentry;
521 	pending_snapshot->root = root;
522 	pending_snapshot->readonly = readonly;
523 	if (inherit) {
524 		pending_snapshot->inherit = *inherit;
525 		*inherit = NULL;	/* take responsibility to free it */
526 	}
527 
528 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
529 	if (IS_ERR(trans)) {
530 		ret = PTR_ERR(trans);
531 		goto fail;
532 	}
533 
534 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
535 	BUG_ON(ret);
536 
537 	spin_lock(&root->fs_info->trans_lock);
538 	list_add(&pending_snapshot->list,
539 		 &trans->transaction->pending_snapshots);
540 	spin_unlock(&root->fs_info->trans_lock);
541 	if (async_transid) {
542 		*async_transid = trans->transid;
543 		ret = btrfs_commit_transaction_async(trans,
544 				     root->fs_info->extent_root, 1);
545 	} else {
546 		ret = btrfs_commit_transaction(trans,
547 					       root->fs_info->extent_root);
548 	}
549 	BUG_ON(ret);
550 
551 	ret = pending_snapshot->error;
552 	if (ret)
553 		goto fail;
554 
555 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
556 	if (ret)
557 		goto fail;
558 
559 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
560 	if (IS_ERR(inode)) {
561 		ret = PTR_ERR(inode);
562 		goto fail;
563 	}
564 	BUG_ON(!inode);
565 	d_instantiate(dentry, inode);
566 	ret = 0;
567 fail:
568 	kfree(pending_snapshot);
569 	return ret;
570 }
571 
572 /*  copy of check_sticky in fs/namei.c()
573 * It's inline, so penalty for filesystems that don't use sticky bit is
574 * minimal.
575 */
576 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
577 {
578 	uid_t fsuid = current_fsuid();
579 
580 	if (!(dir->i_mode & S_ISVTX))
581 		return 0;
582 	if (inode->i_uid == fsuid)
583 		return 0;
584 	if (dir->i_uid == fsuid)
585 		return 0;
586 	return !capable(CAP_FOWNER);
587 }
588 
589 /*  copy of may_delete in fs/namei.c()
590  *	Check whether we can remove a link victim from directory dir, check
591  *  whether the type of victim is right.
592  *  1. We can't do it if dir is read-only (done in permission())
593  *  2. We should have write and exec permissions on dir
594  *  3. We can't remove anything from append-only dir
595  *  4. We can't do anything with immutable dir (done in permission())
596  *  5. If the sticky bit on dir is set we should either
597  *	a. be owner of dir, or
598  *	b. be owner of victim, or
599  *	c. have CAP_FOWNER capability
600  *  6. If the victim is append-only or immutable we can't do antyhing with
601  *     links pointing to it.
602  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
603  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
604  *  9. We can't remove a root or mountpoint.
605  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
606  *     nfs_async_unlink().
607  */
608 
609 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
610 {
611 	int error;
612 
613 	if (!victim->d_inode)
614 		return -ENOENT;
615 
616 	BUG_ON(victim->d_parent->d_inode != dir);
617 	audit_inode_child(victim, dir);
618 
619 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
620 	if (error)
621 		return error;
622 	if (IS_APPEND(dir))
623 		return -EPERM;
624 	if (btrfs_check_sticky(dir, victim->d_inode)||
625 		IS_APPEND(victim->d_inode)||
626 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
627 		return -EPERM;
628 	if (isdir) {
629 		if (!S_ISDIR(victim->d_inode->i_mode))
630 			return -ENOTDIR;
631 		if (IS_ROOT(victim))
632 			return -EBUSY;
633 	} else if (S_ISDIR(victim->d_inode->i_mode))
634 		return -EISDIR;
635 	if (IS_DEADDIR(dir))
636 		return -ENOENT;
637 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
638 		return -EBUSY;
639 	return 0;
640 }
641 
642 /* copy of may_create in fs/namei.c() */
643 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
644 {
645 	if (child->d_inode)
646 		return -EEXIST;
647 	if (IS_DEADDIR(dir))
648 		return -ENOENT;
649 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
650 }
651 
652 /*
653  * Create a new subvolume below @parent.  This is largely modeled after
654  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
655  * inside this filesystem so it's quite a bit simpler.
656  */
657 static noinline int btrfs_mksubvol(struct path *parent,
658 				   char *name, int namelen,
659 				   struct btrfs_root *snap_src,
660 				   u64 *async_transid, bool readonly,
661 				   struct btrfs_qgroup_inherit **inherit)
662 {
663 	struct inode *dir  = parent->dentry->d_inode;
664 	struct dentry *dentry;
665 	int error;
666 
667 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
668 
669 	dentry = lookup_one_len(name, parent->dentry, namelen);
670 	error = PTR_ERR(dentry);
671 	if (IS_ERR(dentry))
672 		goto out_unlock;
673 
674 	error = -EEXIST;
675 	if (dentry->d_inode)
676 		goto out_dput;
677 
678 	error = btrfs_may_create(dir, dentry);
679 	if (error)
680 		goto out_dput;
681 
682 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
683 
684 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
685 		goto out_up_read;
686 
687 	if (snap_src) {
688 		error = create_snapshot(snap_src, dentry, name, namelen,
689 					async_transid, readonly, inherit);
690 	} else {
691 		error = create_subvol(BTRFS_I(dir)->root, dentry,
692 				      name, namelen, async_transid, inherit);
693 	}
694 	if (!error)
695 		fsnotify_mkdir(dir, dentry);
696 out_up_read:
697 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
698 out_dput:
699 	dput(dentry);
700 out_unlock:
701 	mutex_unlock(&dir->i_mutex);
702 	return error;
703 }
704 
705 /*
706  * When we're defragging a range, we don't want to kick it off again
707  * if it is really just waiting for delalloc to send it down.
708  * If we find a nice big extent or delalloc range for the bytes in the
709  * file you want to defrag, we return 0 to let you know to skip this
710  * part of the file
711  */
712 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
713 {
714 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
715 	struct extent_map *em = NULL;
716 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
717 	u64 end;
718 
719 	read_lock(&em_tree->lock);
720 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
721 	read_unlock(&em_tree->lock);
722 
723 	if (em) {
724 		end = extent_map_end(em);
725 		free_extent_map(em);
726 		if (end - offset > thresh)
727 			return 0;
728 	}
729 	/* if we already have a nice delalloc here, just stop */
730 	thresh /= 2;
731 	end = count_range_bits(io_tree, &offset, offset + thresh,
732 			       thresh, EXTENT_DELALLOC, 1);
733 	if (end >= thresh)
734 		return 0;
735 	return 1;
736 }
737 
738 /*
739  * helper function to walk through a file and find extents
740  * newer than a specific transid, and smaller than thresh.
741  *
742  * This is used by the defragging code to find new and small
743  * extents
744  */
745 static int find_new_extents(struct btrfs_root *root,
746 			    struct inode *inode, u64 newer_than,
747 			    u64 *off, int thresh)
748 {
749 	struct btrfs_path *path;
750 	struct btrfs_key min_key;
751 	struct btrfs_key max_key;
752 	struct extent_buffer *leaf;
753 	struct btrfs_file_extent_item *extent;
754 	int type;
755 	int ret;
756 	u64 ino = btrfs_ino(inode);
757 
758 	path = btrfs_alloc_path();
759 	if (!path)
760 		return -ENOMEM;
761 
762 	min_key.objectid = ino;
763 	min_key.type = BTRFS_EXTENT_DATA_KEY;
764 	min_key.offset = *off;
765 
766 	max_key.objectid = ino;
767 	max_key.type = (u8)-1;
768 	max_key.offset = (u64)-1;
769 
770 	path->keep_locks = 1;
771 
772 	while(1) {
773 		ret = btrfs_search_forward(root, &min_key, &max_key,
774 					   path, 0, newer_than);
775 		if (ret != 0)
776 			goto none;
777 		if (min_key.objectid != ino)
778 			goto none;
779 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
780 			goto none;
781 
782 		leaf = path->nodes[0];
783 		extent = btrfs_item_ptr(leaf, path->slots[0],
784 					struct btrfs_file_extent_item);
785 
786 		type = btrfs_file_extent_type(leaf, extent);
787 		if (type == BTRFS_FILE_EXTENT_REG &&
788 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
789 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
790 			*off = min_key.offset;
791 			btrfs_free_path(path);
792 			return 0;
793 		}
794 
795 		if (min_key.offset == (u64)-1)
796 			goto none;
797 
798 		min_key.offset++;
799 		btrfs_release_path(path);
800 	}
801 none:
802 	btrfs_free_path(path);
803 	return -ENOENT;
804 }
805 
806 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
807 {
808 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
809 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
810 	struct extent_map *em;
811 	u64 len = PAGE_CACHE_SIZE;
812 
813 	/*
814 	 * hopefully we have this extent in the tree already, try without
815 	 * the full extent lock
816 	 */
817 	read_lock(&em_tree->lock);
818 	em = lookup_extent_mapping(em_tree, start, len);
819 	read_unlock(&em_tree->lock);
820 
821 	if (!em) {
822 		/* get the big lock and read metadata off disk */
823 		lock_extent(io_tree, start, start + len - 1);
824 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
825 		unlock_extent(io_tree, start, start + len - 1);
826 
827 		if (IS_ERR(em))
828 			return NULL;
829 	}
830 
831 	return em;
832 }
833 
834 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
835 {
836 	struct extent_map *next;
837 	bool ret = true;
838 
839 	/* this is the last extent */
840 	if (em->start + em->len >= i_size_read(inode))
841 		return false;
842 
843 	next = defrag_lookup_extent(inode, em->start + em->len);
844 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
845 		ret = false;
846 
847 	free_extent_map(next);
848 	return ret;
849 }
850 
851 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
852 			       u64 *last_len, u64 *skip, u64 *defrag_end,
853 			       int compress)
854 {
855 	struct extent_map *em;
856 	int ret = 1;
857 	bool next_mergeable = true;
858 
859 	/*
860 	 * make sure that once we start defragging an extent, we keep on
861 	 * defragging it
862 	 */
863 	if (start < *defrag_end)
864 		return 1;
865 
866 	*skip = 0;
867 
868 	em = defrag_lookup_extent(inode, start);
869 	if (!em)
870 		return 0;
871 
872 	/* this will cover holes, and inline extents */
873 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
874 		ret = 0;
875 		goto out;
876 	}
877 
878 	next_mergeable = defrag_check_next_extent(inode, em);
879 
880 	/*
881 	 * we hit a real extent, if it is big or the next extent is not a
882 	 * real extent, don't bother defragging it
883 	 */
884 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
885 	    (em->len >= thresh || !next_mergeable))
886 		ret = 0;
887 out:
888 	/*
889 	 * last_len ends up being a counter of how many bytes we've defragged.
890 	 * every time we choose not to defrag an extent, we reset *last_len
891 	 * so that the next tiny extent will force a defrag.
892 	 *
893 	 * The end result of this is that tiny extents before a single big
894 	 * extent will force at least part of that big extent to be defragged.
895 	 */
896 	if (ret) {
897 		*defrag_end = extent_map_end(em);
898 	} else {
899 		*last_len = 0;
900 		*skip = extent_map_end(em);
901 		*defrag_end = 0;
902 	}
903 
904 	free_extent_map(em);
905 	return ret;
906 }
907 
908 /*
909  * it doesn't do much good to defrag one or two pages
910  * at a time.  This pulls in a nice chunk of pages
911  * to COW and defrag.
912  *
913  * It also makes sure the delalloc code has enough
914  * dirty data to avoid making new small extents as part
915  * of the defrag
916  *
917  * It's a good idea to start RA on this range
918  * before calling this.
919  */
920 static int cluster_pages_for_defrag(struct inode *inode,
921 				    struct page **pages,
922 				    unsigned long start_index,
923 				    int num_pages)
924 {
925 	unsigned long file_end;
926 	u64 isize = i_size_read(inode);
927 	u64 page_start;
928 	u64 page_end;
929 	u64 page_cnt;
930 	int ret;
931 	int i;
932 	int i_done;
933 	struct btrfs_ordered_extent *ordered;
934 	struct extent_state *cached_state = NULL;
935 	struct extent_io_tree *tree;
936 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
937 
938 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
939 	if (!isize || start_index > file_end)
940 		return 0;
941 
942 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
943 
944 	ret = btrfs_delalloc_reserve_space(inode,
945 					   page_cnt << PAGE_CACHE_SHIFT);
946 	if (ret)
947 		return ret;
948 	i_done = 0;
949 	tree = &BTRFS_I(inode)->io_tree;
950 
951 	/* step one, lock all the pages */
952 	for (i = 0; i < page_cnt; i++) {
953 		struct page *page;
954 again:
955 		page = find_or_create_page(inode->i_mapping,
956 					   start_index + i, mask);
957 		if (!page)
958 			break;
959 
960 		page_start = page_offset(page);
961 		page_end = page_start + PAGE_CACHE_SIZE - 1;
962 		while (1) {
963 			lock_extent(tree, page_start, page_end);
964 			ordered = btrfs_lookup_ordered_extent(inode,
965 							      page_start);
966 			unlock_extent(tree, page_start, page_end);
967 			if (!ordered)
968 				break;
969 
970 			unlock_page(page);
971 			btrfs_start_ordered_extent(inode, ordered, 1);
972 			btrfs_put_ordered_extent(ordered);
973 			lock_page(page);
974 			/*
975 			 * we unlocked the page above, so we need check if
976 			 * it was released or not.
977 			 */
978 			if (page->mapping != inode->i_mapping) {
979 				unlock_page(page);
980 				page_cache_release(page);
981 				goto again;
982 			}
983 		}
984 
985 		if (!PageUptodate(page)) {
986 			btrfs_readpage(NULL, page);
987 			lock_page(page);
988 			if (!PageUptodate(page)) {
989 				unlock_page(page);
990 				page_cache_release(page);
991 				ret = -EIO;
992 				break;
993 			}
994 		}
995 
996 		if (page->mapping != inode->i_mapping) {
997 			unlock_page(page);
998 			page_cache_release(page);
999 			goto again;
1000 		}
1001 
1002 		pages[i] = page;
1003 		i_done++;
1004 	}
1005 	if (!i_done || ret)
1006 		goto out;
1007 
1008 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
1009 		goto out;
1010 
1011 	/*
1012 	 * so now we have a nice long stream of locked
1013 	 * and up to date pages, lets wait on them
1014 	 */
1015 	for (i = 0; i < i_done; i++)
1016 		wait_on_page_writeback(pages[i]);
1017 
1018 	page_start = page_offset(pages[0]);
1019 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1020 
1021 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1022 			 page_start, page_end - 1, 0, &cached_state);
1023 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1024 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1025 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1026 			  GFP_NOFS);
1027 
1028 	if (i_done != page_cnt) {
1029 		spin_lock(&BTRFS_I(inode)->lock);
1030 		BTRFS_I(inode)->outstanding_extents++;
1031 		spin_unlock(&BTRFS_I(inode)->lock);
1032 		btrfs_delalloc_release_space(inode,
1033 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1034 	}
1035 
1036 
1037 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1038 				  &cached_state);
1039 
1040 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1041 			     page_start, page_end - 1, &cached_state,
1042 			     GFP_NOFS);
1043 
1044 	for (i = 0; i < i_done; i++) {
1045 		clear_page_dirty_for_io(pages[i]);
1046 		ClearPageChecked(pages[i]);
1047 		set_page_extent_mapped(pages[i]);
1048 		set_page_dirty(pages[i]);
1049 		unlock_page(pages[i]);
1050 		page_cache_release(pages[i]);
1051 	}
1052 	return i_done;
1053 out:
1054 	for (i = 0; i < i_done; i++) {
1055 		unlock_page(pages[i]);
1056 		page_cache_release(pages[i]);
1057 	}
1058 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1059 	return ret;
1060 
1061 }
1062 
1063 int btrfs_defrag_file(struct inode *inode, struct file *file,
1064 		      struct btrfs_ioctl_defrag_range_args *range,
1065 		      u64 newer_than, unsigned long max_to_defrag)
1066 {
1067 	struct btrfs_root *root = BTRFS_I(inode)->root;
1068 	struct file_ra_state *ra = NULL;
1069 	unsigned long last_index;
1070 	u64 isize = i_size_read(inode);
1071 	u64 last_len = 0;
1072 	u64 skip = 0;
1073 	u64 defrag_end = 0;
1074 	u64 newer_off = range->start;
1075 	unsigned long i;
1076 	unsigned long ra_index = 0;
1077 	int ret;
1078 	int defrag_count = 0;
1079 	int compress_type = BTRFS_COMPRESS_ZLIB;
1080 	int extent_thresh = range->extent_thresh;
1081 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1082 	int cluster = max_cluster;
1083 	u64 new_align = ~((u64)128 * 1024 - 1);
1084 	struct page **pages = NULL;
1085 
1086 	if (extent_thresh == 0)
1087 		extent_thresh = 256 * 1024;
1088 
1089 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1090 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1091 			return -EINVAL;
1092 		if (range->compress_type)
1093 			compress_type = range->compress_type;
1094 	}
1095 
1096 	if (isize == 0)
1097 		return 0;
1098 
1099 	/*
1100 	 * if we were not given a file, allocate a readahead
1101 	 * context
1102 	 */
1103 	if (!file) {
1104 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1105 		if (!ra)
1106 			return -ENOMEM;
1107 		file_ra_state_init(ra, inode->i_mapping);
1108 	} else {
1109 		ra = &file->f_ra;
1110 	}
1111 
1112 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1113 			GFP_NOFS);
1114 	if (!pages) {
1115 		ret = -ENOMEM;
1116 		goto out_ra;
1117 	}
1118 
1119 	/* find the last page to defrag */
1120 	if (range->start + range->len > range->start) {
1121 		last_index = min_t(u64, isize - 1,
1122 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1123 	} else {
1124 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1125 	}
1126 
1127 	if (newer_than) {
1128 		ret = find_new_extents(root, inode, newer_than,
1129 				       &newer_off, 64 * 1024);
1130 		if (!ret) {
1131 			range->start = newer_off;
1132 			/*
1133 			 * we always align our defrag to help keep
1134 			 * the extents in the file evenly spaced
1135 			 */
1136 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1137 		} else
1138 			goto out_ra;
1139 	} else {
1140 		i = range->start >> PAGE_CACHE_SHIFT;
1141 	}
1142 	if (!max_to_defrag)
1143 		max_to_defrag = last_index + 1;
1144 
1145 	/*
1146 	 * make writeback starts from i, so the defrag range can be
1147 	 * written sequentially.
1148 	 */
1149 	if (i < inode->i_mapping->writeback_index)
1150 		inode->i_mapping->writeback_index = i;
1151 
1152 	while (i <= last_index && defrag_count < max_to_defrag &&
1153 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1154 		PAGE_CACHE_SHIFT)) {
1155 		/*
1156 		 * make sure we stop running if someone unmounts
1157 		 * the FS
1158 		 */
1159 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1160 			break;
1161 
1162 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1163 					 extent_thresh, &last_len, &skip,
1164 					 &defrag_end, range->flags &
1165 					 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1166 			unsigned long next;
1167 			/*
1168 			 * the should_defrag function tells us how much to skip
1169 			 * bump our counter by the suggested amount
1170 			 */
1171 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1172 			i = max(i + 1, next);
1173 			continue;
1174 		}
1175 
1176 		if (!newer_than) {
1177 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1178 				   PAGE_CACHE_SHIFT) - i;
1179 			cluster = min(cluster, max_cluster);
1180 		} else {
1181 			cluster = max_cluster;
1182 		}
1183 
1184 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1185 			BTRFS_I(inode)->force_compress = compress_type;
1186 
1187 		if (i + cluster > ra_index) {
1188 			ra_index = max(i, ra_index);
1189 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1190 				       cluster);
1191 			ra_index += max_cluster;
1192 		}
1193 
1194 		mutex_lock(&inode->i_mutex);
1195 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1196 		if (ret < 0) {
1197 			mutex_unlock(&inode->i_mutex);
1198 			goto out_ra;
1199 		}
1200 
1201 		defrag_count += ret;
1202 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1203 		mutex_unlock(&inode->i_mutex);
1204 
1205 		if (newer_than) {
1206 			if (newer_off == (u64)-1)
1207 				break;
1208 
1209 			if (ret > 0)
1210 				i += ret;
1211 
1212 			newer_off = max(newer_off + 1,
1213 					(u64)i << PAGE_CACHE_SHIFT);
1214 
1215 			ret = find_new_extents(root, inode,
1216 					       newer_than, &newer_off,
1217 					       64 * 1024);
1218 			if (!ret) {
1219 				range->start = newer_off;
1220 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1221 			} else {
1222 				break;
1223 			}
1224 		} else {
1225 			if (ret > 0) {
1226 				i += ret;
1227 				last_len += ret << PAGE_CACHE_SHIFT;
1228 			} else {
1229 				i++;
1230 				last_len = 0;
1231 			}
1232 		}
1233 	}
1234 
1235 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1236 		filemap_flush(inode->i_mapping);
1237 
1238 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1239 		/* the filemap_flush will queue IO into the worker threads, but
1240 		 * we have to make sure the IO is actually started and that
1241 		 * ordered extents get created before we return
1242 		 */
1243 		atomic_inc(&root->fs_info->async_submit_draining);
1244 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1245 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1246 			wait_event(root->fs_info->async_submit_wait,
1247 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1248 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1249 		}
1250 		atomic_dec(&root->fs_info->async_submit_draining);
1251 
1252 		mutex_lock(&inode->i_mutex);
1253 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1254 		mutex_unlock(&inode->i_mutex);
1255 	}
1256 
1257 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1258 		btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1259 	}
1260 
1261 	ret = defrag_count;
1262 
1263 out_ra:
1264 	if (!file)
1265 		kfree(ra);
1266 	kfree(pages);
1267 	return ret;
1268 }
1269 
1270 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1271 					void __user *arg)
1272 {
1273 	u64 new_size;
1274 	u64 old_size;
1275 	u64 devid = 1;
1276 	struct btrfs_ioctl_vol_args *vol_args;
1277 	struct btrfs_trans_handle *trans;
1278 	struct btrfs_device *device = NULL;
1279 	char *sizestr;
1280 	char *devstr = NULL;
1281 	int ret = 0;
1282 	int mod = 0;
1283 
1284 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1285 		return -EROFS;
1286 
1287 	if (!capable(CAP_SYS_ADMIN))
1288 		return -EPERM;
1289 
1290 	mutex_lock(&root->fs_info->volume_mutex);
1291 	if (root->fs_info->balance_ctl) {
1292 		printk(KERN_INFO "btrfs: balance in progress\n");
1293 		ret = -EINVAL;
1294 		goto out;
1295 	}
1296 
1297 	vol_args = memdup_user(arg, sizeof(*vol_args));
1298 	if (IS_ERR(vol_args)) {
1299 		ret = PTR_ERR(vol_args);
1300 		goto out;
1301 	}
1302 
1303 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1304 
1305 	sizestr = vol_args->name;
1306 	devstr = strchr(sizestr, ':');
1307 	if (devstr) {
1308 		char *end;
1309 		sizestr = devstr + 1;
1310 		*devstr = '\0';
1311 		devstr = vol_args->name;
1312 		devid = simple_strtoull(devstr, &end, 10);
1313 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1314 		       (unsigned long long)devid);
1315 	}
1316 	device = btrfs_find_device(root, devid, NULL, NULL);
1317 	if (!device) {
1318 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1319 		       (unsigned long long)devid);
1320 		ret = -EINVAL;
1321 		goto out_free;
1322 	}
1323 	if (device->fs_devices && device->fs_devices->seeding) {
1324 		printk(KERN_INFO "btrfs: resizer unable to apply on "
1325 		       "seeding device %llu\n",
1326 		       (unsigned long long)devid);
1327 		ret = -EINVAL;
1328 		goto out_free;
1329 	}
1330 
1331 	if (!strcmp(sizestr, "max"))
1332 		new_size = device->bdev->bd_inode->i_size;
1333 	else {
1334 		if (sizestr[0] == '-') {
1335 			mod = -1;
1336 			sizestr++;
1337 		} else if (sizestr[0] == '+') {
1338 			mod = 1;
1339 			sizestr++;
1340 		}
1341 		new_size = memparse(sizestr, NULL);
1342 		if (new_size == 0) {
1343 			ret = -EINVAL;
1344 			goto out_free;
1345 		}
1346 	}
1347 
1348 	old_size = device->total_bytes;
1349 
1350 	if (mod < 0) {
1351 		if (new_size > old_size) {
1352 			ret = -EINVAL;
1353 			goto out_free;
1354 		}
1355 		new_size = old_size - new_size;
1356 	} else if (mod > 0) {
1357 		new_size = old_size + new_size;
1358 	}
1359 
1360 	if (new_size < 256 * 1024 * 1024) {
1361 		ret = -EINVAL;
1362 		goto out_free;
1363 	}
1364 	if (new_size > device->bdev->bd_inode->i_size) {
1365 		ret = -EFBIG;
1366 		goto out_free;
1367 	}
1368 
1369 	do_div(new_size, root->sectorsize);
1370 	new_size *= root->sectorsize;
1371 
1372 	printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1373 		      rcu_str_deref(device->name),
1374 		      (unsigned long long)new_size);
1375 
1376 	if (new_size > old_size) {
1377 		trans = btrfs_start_transaction(root, 0);
1378 		if (IS_ERR(trans)) {
1379 			ret = PTR_ERR(trans);
1380 			goto out_free;
1381 		}
1382 		ret = btrfs_grow_device(trans, device, new_size);
1383 		btrfs_commit_transaction(trans, root);
1384 	} else if (new_size < old_size) {
1385 		ret = btrfs_shrink_device(device, new_size);
1386 	}
1387 
1388 out_free:
1389 	kfree(vol_args);
1390 out:
1391 	mutex_unlock(&root->fs_info->volume_mutex);
1392 	return ret;
1393 }
1394 
1395 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1396 				char *name, unsigned long fd, int subvol,
1397 				u64 *transid, bool readonly,
1398 				struct btrfs_qgroup_inherit **inherit)
1399 {
1400 	struct file *src_file;
1401 	int namelen;
1402 	int ret = 0;
1403 
1404 	ret = mnt_want_write_file(file);
1405 	if (ret)
1406 		goto out;
1407 
1408 	namelen = strlen(name);
1409 	if (strchr(name, '/')) {
1410 		ret = -EINVAL;
1411 		goto out_drop_write;
1412 	}
1413 
1414 	if (name[0] == '.' &&
1415 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1416 		ret = -EEXIST;
1417 		goto out_drop_write;
1418 	}
1419 
1420 	if (subvol) {
1421 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1422 				     NULL, transid, readonly, inherit);
1423 	} else {
1424 		struct inode *src_inode;
1425 		src_file = fget(fd);
1426 		if (!src_file) {
1427 			ret = -EINVAL;
1428 			goto out_drop_write;
1429 		}
1430 
1431 		src_inode = src_file->f_path.dentry->d_inode;
1432 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1433 			printk(KERN_INFO "btrfs: Snapshot src from "
1434 			       "another FS\n");
1435 			ret = -EINVAL;
1436 			fput(src_file);
1437 			goto out_drop_write;
1438 		}
1439 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1440 				     BTRFS_I(src_inode)->root,
1441 				     transid, readonly, inherit);
1442 		fput(src_file);
1443 	}
1444 out_drop_write:
1445 	mnt_drop_write_file(file);
1446 out:
1447 	return ret;
1448 }
1449 
1450 static noinline int btrfs_ioctl_snap_create(struct file *file,
1451 					    void __user *arg, int subvol)
1452 {
1453 	struct btrfs_ioctl_vol_args *vol_args;
1454 	int ret;
1455 
1456 	vol_args = memdup_user(arg, sizeof(*vol_args));
1457 	if (IS_ERR(vol_args))
1458 		return PTR_ERR(vol_args);
1459 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1460 
1461 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1462 					      vol_args->fd, subvol,
1463 					      NULL, false, NULL);
1464 
1465 	kfree(vol_args);
1466 	return ret;
1467 }
1468 
1469 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1470 					       void __user *arg, int subvol)
1471 {
1472 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1473 	int ret;
1474 	u64 transid = 0;
1475 	u64 *ptr = NULL;
1476 	bool readonly = false;
1477 	struct btrfs_qgroup_inherit *inherit = NULL;
1478 
1479 	vol_args = memdup_user(arg, sizeof(*vol_args));
1480 	if (IS_ERR(vol_args))
1481 		return PTR_ERR(vol_args);
1482 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1483 
1484 	if (vol_args->flags &
1485 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1486 	      BTRFS_SUBVOL_QGROUP_INHERIT)) {
1487 		ret = -EOPNOTSUPP;
1488 		goto out;
1489 	}
1490 
1491 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1492 		ptr = &transid;
1493 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1494 		readonly = true;
1495 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1496 		if (vol_args->size > PAGE_CACHE_SIZE) {
1497 			ret = -EINVAL;
1498 			goto out;
1499 		}
1500 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1501 		if (IS_ERR(inherit)) {
1502 			ret = PTR_ERR(inherit);
1503 			goto out;
1504 		}
1505 	}
1506 
1507 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1508 					      vol_args->fd, subvol, ptr,
1509 					      readonly, &inherit);
1510 
1511 	if (ret == 0 && ptr &&
1512 	    copy_to_user(arg +
1513 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1514 				  transid), ptr, sizeof(*ptr)))
1515 		ret = -EFAULT;
1516 out:
1517 	kfree(vol_args);
1518 	kfree(inherit);
1519 	return ret;
1520 }
1521 
1522 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1523 						void __user *arg)
1524 {
1525 	struct inode *inode = fdentry(file)->d_inode;
1526 	struct btrfs_root *root = BTRFS_I(inode)->root;
1527 	int ret = 0;
1528 	u64 flags = 0;
1529 
1530 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1531 		return -EINVAL;
1532 
1533 	down_read(&root->fs_info->subvol_sem);
1534 	if (btrfs_root_readonly(root))
1535 		flags |= BTRFS_SUBVOL_RDONLY;
1536 	up_read(&root->fs_info->subvol_sem);
1537 
1538 	if (copy_to_user(arg, &flags, sizeof(flags)))
1539 		ret = -EFAULT;
1540 
1541 	return ret;
1542 }
1543 
1544 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1545 					      void __user *arg)
1546 {
1547 	struct inode *inode = fdentry(file)->d_inode;
1548 	struct btrfs_root *root = BTRFS_I(inode)->root;
1549 	struct btrfs_trans_handle *trans;
1550 	u64 root_flags;
1551 	u64 flags;
1552 	int ret = 0;
1553 
1554 	ret = mnt_want_write_file(file);
1555 	if (ret)
1556 		goto out;
1557 
1558 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1559 		ret = -EINVAL;
1560 		goto out_drop_write;
1561 	}
1562 
1563 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1564 		ret = -EFAULT;
1565 		goto out_drop_write;
1566 	}
1567 
1568 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1569 		ret = -EINVAL;
1570 		goto out_drop_write;
1571 	}
1572 
1573 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1574 		ret = -EOPNOTSUPP;
1575 		goto out_drop_write;
1576 	}
1577 
1578 	if (!inode_owner_or_capable(inode)) {
1579 		ret = -EACCES;
1580 		goto out_drop_write;
1581 	}
1582 
1583 	down_write(&root->fs_info->subvol_sem);
1584 
1585 	/* nothing to do */
1586 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1587 		goto out_drop_sem;
1588 
1589 	root_flags = btrfs_root_flags(&root->root_item);
1590 	if (flags & BTRFS_SUBVOL_RDONLY)
1591 		btrfs_set_root_flags(&root->root_item,
1592 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1593 	else
1594 		btrfs_set_root_flags(&root->root_item,
1595 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1596 
1597 	trans = btrfs_start_transaction(root, 1);
1598 	if (IS_ERR(trans)) {
1599 		ret = PTR_ERR(trans);
1600 		goto out_reset;
1601 	}
1602 
1603 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1604 				&root->root_key, &root->root_item);
1605 
1606 	btrfs_commit_transaction(trans, root);
1607 out_reset:
1608 	if (ret)
1609 		btrfs_set_root_flags(&root->root_item, root_flags);
1610 out_drop_sem:
1611 	up_write(&root->fs_info->subvol_sem);
1612 out_drop_write:
1613 	mnt_drop_write_file(file);
1614 out:
1615 	return ret;
1616 }
1617 
1618 /*
1619  * helper to check if the subvolume references other subvolumes
1620  */
1621 static noinline int may_destroy_subvol(struct btrfs_root *root)
1622 {
1623 	struct btrfs_path *path;
1624 	struct btrfs_key key;
1625 	int ret;
1626 
1627 	path = btrfs_alloc_path();
1628 	if (!path)
1629 		return -ENOMEM;
1630 
1631 	key.objectid = root->root_key.objectid;
1632 	key.type = BTRFS_ROOT_REF_KEY;
1633 	key.offset = (u64)-1;
1634 
1635 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1636 				&key, path, 0, 0);
1637 	if (ret < 0)
1638 		goto out;
1639 	BUG_ON(ret == 0);
1640 
1641 	ret = 0;
1642 	if (path->slots[0] > 0) {
1643 		path->slots[0]--;
1644 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1645 		if (key.objectid == root->root_key.objectid &&
1646 		    key.type == BTRFS_ROOT_REF_KEY)
1647 			ret = -ENOTEMPTY;
1648 	}
1649 out:
1650 	btrfs_free_path(path);
1651 	return ret;
1652 }
1653 
1654 static noinline int key_in_sk(struct btrfs_key *key,
1655 			      struct btrfs_ioctl_search_key *sk)
1656 {
1657 	struct btrfs_key test;
1658 	int ret;
1659 
1660 	test.objectid = sk->min_objectid;
1661 	test.type = sk->min_type;
1662 	test.offset = sk->min_offset;
1663 
1664 	ret = btrfs_comp_cpu_keys(key, &test);
1665 	if (ret < 0)
1666 		return 0;
1667 
1668 	test.objectid = sk->max_objectid;
1669 	test.type = sk->max_type;
1670 	test.offset = sk->max_offset;
1671 
1672 	ret = btrfs_comp_cpu_keys(key, &test);
1673 	if (ret > 0)
1674 		return 0;
1675 	return 1;
1676 }
1677 
1678 static noinline int copy_to_sk(struct btrfs_root *root,
1679 			       struct btrfs_path *path,
1680 			       struct btrfs_key *key,
1681 			       struct btrfs_ioctl_search_key *sk,
1682 			       char *buf,
1683 			       unsigned long *sk_offset,
1684 			       int *num_found)
1685 {
1686 	u64 found_transid;
1687 	struct extent_buffer *leaf;
1688 	struct btrfs_ioctl_search_header sh;
1689 	unsigned long item_off;
1690 	unsigned long item_len;
1691 	int nritems;
1692 	int i;
1693 	int slot;
1694 	int ret = 0;
1695 
1696 	leaf = path->nodes[0];
1697 	slot = path->slots[0];
1698 	nritems = btrfs_header_nritems(leaf);
1699 
1700 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1701 		i = nritems;
1702 		goto advance_key;
1703 	}
1704 	found_transid = btrfs_header_generation(leaf);
1705 
1706 	for (i = slot; i < nritems; i++) {
1707 		item_off = btrfs_item_ptr_offset(leaf, i);
1708 		item_len = btrfs_item_size_nr(leaf, i);
1709 
1710 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1711 			item_len = 0;
1712 
1713 		if (sizeof(sh) + item_len + *sk_offset >
1714 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1715 			ret = 1;
1716 			goto overflow;
1717 		}
1718 
1719 		btrfs_item_key_to_cpu(leaf, key, i);
1720 		if (!key_in_sk(key, sk))
1721 			continue;
1722 
1723 		sh.objectid = key->objectid;
1724 		sh.offset = key->offset;
1725 		sh.type = key->type;
1726 		sh.len = item_len;
1727 		sh.transid = found_transid;
1728 
1729 		/* copy search result header */
1730 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1731 		*sk_offset += sizeof(sh);
1732 
1733 		if (item_len) {
1734 			char *p = buf + *sk_offset;
1735 			/* copy the item */
1736 			read_extent_buffer(leaf, p,
1737 					   item_off, item_len);
1738 			*sk_offset += item_len;
1739 		}
1740 		(*num_found)++;
1741 
1742 		if (*num_found >= sk->nr_items)
1743 			break;
1744 	}
1745 advance_key:
1746 	ret = 0;
1747 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1748 		key->offset++;
1749 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1750 		key->offset = 0;
1751 		key->type++;
1752 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1753 		key->offset = 0;
1754 		key->type = 0;
1755 		key->objectid++;
1756 	} else
1757 		ret = 1;
1758 overflow:
1759 	return ret;
1760 }
1761 
1762 static noinline int search_ioctl(struct inode *inode,
1763 				 struct btrfs_ioctl_search_args *args)
1764 {
1765 	struct btrfs_root *root;
1766 	struct btrfs_key key;
1767 	struct btrfs_key max_key;
1768 	struct btrfs_path *path;
1769 	struct btrfs_ioctl_search_key *sk = &args->key;
1770 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1771 	int ret;
1772 	int num_found = 0;
1773 	unsigned long sk_offset = 0;
1774 
1775 	path = btrfs_alloc_path();
1776 	if (!path)
1777 		return -ENOMEM;
1778 
1779 	if (sk->tree_id == 0) {
1780 		/* search the root of the inode that was passed */
1781 		root = BTRFS_I(inode)->root;
1782 	} else {
1783 		key.objectid = sk->tree_id;
1784 		key.type = BTRFS_ROOT_ITEM_KEY;
1785 		key.offset = (u64)-1;
1786 		root = btrfs_read_fs_root_no_name(info, &key);
1787 		if (IS_ERR(root)) {
1788 			printk(KERN_ERR "could not find root %llu\n",
1789 			       sk->tree_id);
1790 			btrfs_free_path(path);
1791 			return -ENOENT;
1792 		}
1793 	}
1794 
1795 	key.objectid = sk->min_objectid;
1796 	key.type = sk->min_type;
1797 	key.offset = sk->min_offset;
1798 
1799 	max_key.objectid = sk->max_objectid;
1800 	max_key.type = sk->max_type;
1801 	max_key.offset = sk->max_offset;
1802 
1803 	path->keep_locks = 1;
1804 
1805 	while(1) {
1806 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1807 					   sk->min_transid);
1808 		if (ret != 0) {
1809 			if (ret > 0)
1810 				ret = 0;
1811 			goto err;
1812 		}
1813 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1814 				 &sk_offset, &num_found);
1815 		btrfs_release_path(path);
1816 		if (ret || num_found >= sk->nr_items)
1817 			break;
1818 
1819 	}
1820 	ret = 0;
1821 err:
1822 	sk->nr_items = num_found;
1823 	btrfs_free_path(path);
1824 	return ret;
1825 }
1826 
1827 static noinline int btrfs_ioctl_tree_search(struct file *file,
1828 					   void __user *argp)
1829 {
1830 	 struct btrfs_ioctl_search_args *args;
1831 	 struct inode *inode;
1832 	 int ret;
1833 
1834 	if (!capable(CAP_SYS_ADMIN))
1835 		return -EPERM;
1836 
1837 	args = memdup_user(argp, sizeof(*args));
1838 	if (IS_ERR(args))
1839 		return PTR_ERR(args);
1840 
1841 	inode = fdentry(file)->d_inode;
1842 	ret = search_ioctl(inode, args);
1843 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1844 		ret = -EFAULT;
1845 	kfree(args);
1846 	return ret;
1847 }
1848 
1849 /*
1850  * Search INODE_REFs to identify path name of 'dirid' directory
1851  * in a 'tree_id' tree. and sets path name to 'name'.
1852  */
1853 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1854 				u64 tree_id, u64 dirid, char *name)
1855 {
1856 	struct btrfs_root *root;
1857 	struct btrfs_key key;
1858 	char *ptr;
1859 	int ret = -1;
1860 	int slot;
1861 	int len;
1862 	int total_len = 0;
1863 	struct btrfs_inode_ref *iref;
1864 	struct extent_buffer *l;
1865 	struct btrfs_path *path;
1866 
1867 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1868 		name[0]='\0';
1869 		return 0;
1870 	}
1871 
1872 	path = btrfs_alloc_path();
1873 	if (!path)
1874 		return -ENOMEM;
1875 
1876 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1877 
1878 	key.objectid = tree_id;
1879 	key.type = BTRFS_ROOT_ITEM_KEY;
1880 	key.offset = (u64)-1;
1881 	root = btrfs_read_fs_root_no_name(info, &key);
1882 	if (IS_ERR(root)) {
1883 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1884 		ret = -ENOENT;
1885 		goto out;
1886 	}
1887 
1888 	key.objectid = dirid;
1889 	key.type = BTRFS_INODE_REF_KEY;
1890 	key.offset = (u64)-1;
1891 
1892 	while(1) {
1893 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1894 		if (ret < 0)
1895 			goto out;
1896 
1897 		l = path->nodes[0];
1898 		slot = path->slots[0];
1899 		if (ret > 0 && slot > 0)
1900 			slot--;
1901 		btrfs_item_key_to_cpu(l, &key, slot);
1902 
1903 		if (ret > 0 && (key.objectid != dirid ||
1904 				key.type != BTRFS_INODE_REF_KEY)) {
1905 			ret = -ENOENT;
1906 			goto out;
1907 		}
1908 
1909 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1910 		len = btrfs_inode_ref_name_len(l, iref);
1911 		ptr -= len + 1;
1912 		total_len += len + 1;
1913 		if (ptr < name)
1914 			goto out;
1915 
1916 		*(ptr + len) = '/';
1917 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1918 
1919 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1920 			break;
1921 
1922 		btrfs_release_path(path);
1923 		key.objectid = key.offset;
1924 		key.offset = (u64)-1;
1925 		dirid = key.objectid;
1926 	}
1927 	if (ptr < name)
1928 		goto out;
1929 	memmove(name, ptr, total_len);
1930 	name[total_len]='\0';
1931 	ret = 0;
1932 out:
1933 	btrfs_free_path(path);
1934 	return ret;
1935 }
1936 
1937 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1938 					   void __user *argp)
1939 {
1940 	 struct btrfs_ioctl_ino_lookup_args *args;
1941 	 struct inode *inode;
1942 	 int ret;
1943 
1944 	if (!capable(CAP_SYS_ADMIN))
1945 		return -EPERM;
1946 
1947 	args = memdup_user(argp, sizeof(*args));
1948 	if (IS_ERR(args))
1949 		return PTR_ERR(args);
1950 
1951 	inode = fdentry(file)->d_inode;
1952 
1953 	if (args->treeid == 0)
1954 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1955 
1956 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1957 					args->treeid, args->objectid,
1958 					args->name);
1959 
1960 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1961 		ret = -EFAULT;
1962 
1963 	kfree(args);
1964 	return ret;
1965 }
1966 
1967 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1968 					     void __user *arg)
1969 {
1970 	struct dentry *parent = fdentry(file);
1971 	struct dentry *dentry;
1972 	struct inode *dir = parent->d_inode;
1973 	struct inode *inode;
1974 	struct btrfs_root *root = BTRFS_I(dir)->root;
1975 	struct btrfs_root *dest = NULL;
1976 	struct btrfs_ioctl_vol_args *vol_args;
1977 	struct btrfs_trans_handle *trans;
1978 	int namelen;
1979 	int ret;
1980 	int err = 0;
1981 
1982 	vol_args = memdup_user(arg, sizeof(*vol_args));
1983 	if (IS_ERR(vol_args))
1984 		return PTR_ERR(vol_args);
1985 
1986 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1987 	namelen = strlen(vol_args->name);
1988 	if (strchr(vol_args->name, '/') ||
1989 	    strncmp(vol_args->name, "..", namelen) == 0) {
1990 		err = -EINVAL;
1991 		goto out;
1992 	}
1993 
1994 	err = mnt_want_write_file(file);
1995 	if (err)
1996 		goto out;
1997 
1998 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1999 	dentry = lookup_one_len(vol_args->name, parent, namelen);
2000 	if (IS_ERR(dentry)) {
2001 		err = PTR_ERR(dentry);
2002 		goto out_unlock_dir;
2003 	}
2004 
2005 	if (!dentry->d_inode) {
2006 		err = -ENOENT;
2007 		goto out_dput;
2008 	}
2009 
2010 	inode = dentry->d_inode;
2011 	dest = BTRFS_I(inode)->root;
2012 	if (!capable(CAP_SYS_ADMIN)){
2013 		/*
2014 		 * Regular user.  Only allow this with a special mount
2015 		 * option, when the user has write+exec access to the
2016 		 * subvol root, and when rmdir(2) would have been
2017 		 * allowed.
2018 		 *
2019 		 * Note that this is _not_ check that the subvol is
2020 		 * empty or doesn't contain data that we wouldn't
2021 		 * otherwise be able to delete.
2022 		 *
2023 		 * Users who want to delete empty subvols should try
2024 		 * rmdir(2).
2025 		 */
2026 		err = -EPERM;
2027 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2028 			goto out_dput;
2029 
2030 		/*
2031 		 * Do not allow deletion if the parent dir is the same
2032 		 * as the dir to be deleted.  That means the ioctl
2033 		 * must be called on the dentry referencing the root
2034 		 * of the subvol, not a random directory contained
2035 		 * within it.
2036 		 */
2037 		err = -EINVAL;
2038 		if (root == dest)
2039 			goto out_dput;
2040 
2041 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2042 		if (err)
2043 			goto out_dput;
2044 
2045 		/* check if subvolume may be deleted by a non-root user */
2046 		err = btrfs_may_delete(dir, dentry, 1);
2047 		if (err)
2048 			goto out_dput;
2049 	}
2050 
2051 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2052 		err = -EINVAL;
2053 		goto out_dput;
2054 	}
2055 
2056 	mutex_lock(&inode->i_mutex);
2057 	err = d_invalidate(dentry);
2058 	if (err)
2059 		goto out_unlock;
2060 
2061 	down_write(&root->fs_info->subvol_sem);
2062 
2063 	err = may_destroy_subvol(dest);
2064 	if (err)
2065 		goto out_up_write;
2066 
2067 	trans = btrfs_start_transaction(root, 0);
2068 	if (IS_ERR(trans)) {
2069 		err = PTR_ERR(trans);
2070 		goto out_up_write;
2071 	}
2072 	trans->block_rsv = &root->fs_info->global_block_rsv;
2073 
2074 	ret = btrfs_unlink_subvol(trans, root, dir,
2075 				dest->root_key.objectid,
2076 				dentry->d_name.name,
2077 				dentry->d_name.len);
2078 	if (ret) {
2079 		err = ret;
2080 		btrfs_abort_transaction(trans, root, ret);
2081 		goto out_end_trans;
2082 	}
2083 
2084 	btrfs_record_root_in_trans(trans, dest);
2085 
2086 	memset(&dest->root_item.drop_progress, 0,
2087 		sizeof(dest->root_item.drop_progress));
2088 	dest->root_item.drop_level = 0;
2089 	btrfs_set_root_refs(&dest->root_item, 0);
2090 
2091 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2092 		ret = btrfs_insert_orphan_item(trans,
2093 					root->fs_info->tree_root,
2094 					dest->root_key.objectid);
2095 		if (ret) {
2096 			btrfs_abort_transaction(trans, root, ret);
2097 			err = ret;
2098 			goto out_end_trans;
2099 		}
2100 	}
2101 out_end_trans:
2102 	ret = btrfs_end_transaction(trans, root);
2103 	if (ret && !err)
2104 		err = ret;
2105 	inode->i_flags |= S_DEAD;
2106 out_up_write:
2107 	up_write(&root->fs_info->subvol_sem);
2108 out_unlock:
2109 	mutex_unlock(&inode->i_mutex);
2110 	if (!err) {
2111 		shrink_dcache_sb(root->fs_info->sb);
2112 		btrfs_invalidate_inodes(dest);
2113 		d_delete(dentry);
2114 	}
2115 out_dput:
2116 	dput(dentry);
2117 out_unlock_dir:
2118 	mutex_unlock(&dir->i_mutex);
2119 	mnt_drop_write_file(file);
2120 out:
2121 	kfree(vol_args);
2122 	return err;
2123 }
2124 
2125 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2126 {
2127 	struct inode *inode = fdentry(file)->d_inode;
2128 	struct btrfs_root *root = BTRFS_I(inode)->root;
2129 	struct btrfs_ioctl_defrag_range_args *range;
2130 	int ret;
2131 
2132 	if (btrfs_root_readonly(root))
2133 		return -EROFS;
2134 
2135 	ret = mnt_want_write_file(file);
2136 	if (ret)
2137 		return ret;
2138 
2139 	switch (inode->i_mode & S_IFMT) {
2140 	case S_IFDIR:
2141 		if (!capable(CAP_SYS_ADMIN)) {
2142 			ret = -EPERM;
2143 			goto out;
2144 		}
2145 		ret = btrfs_defrag_root(root, 0);
2146 		if (ret)
2147 			goto out;
2148 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2149 		break;
2150 	case S_IFREG:
2151 		if (!(file->f_mode & FMODE_WRITE)) {
2152 			ret = -EINVAL;
2153 			goto out;
2154 		}
2155 
2156 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2157 		if (!range) {
2158 			ret = -ENOMEM;
2159 			goto out;
2160 		}
2161 
2162 		if (argp) {
2163 			if (copy_from_user(range, argp,
2164 					   sizeof(*range))) {
2165 				ret = -EFAULT;
2166 				kfree(range);
2167 				goto out;
2168 			}
2169 			/* compression requires us to start the IO */
2170 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2171 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2172 				range->extent_thresh = (u32)-1;
2173 			}
2174 		} else {
2175 			/* the rest are all set to zero by kzalloc */
2176 			range->len = (u64)-1;
2177 		}
2178 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2179 					range, 0, 0);
2180 		if (ret > 0)
2181 			ret = 0;
2182 		kfree(range);
2183 		break;
2184 	default:
2185 		ret = -EINVAL;
2186 	}
2187 out:
2188 	mnt_drop_write_file(file);
2189 	return ret;
2190 }
2191 
2192 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2193 {
2194 	struct btrfs_ioctl_vol_args *vol_args;
2195 	int ret;
2196 
2197 	if (!capable(CAP_SYS_ADMIN))
2198 		return -EPERM;
2199 
2200 	mutex_lock(&root->fs_info->volume_mutex);
2201 	if (root->fs_info->balance_ctl) {
2202 		printk(KERN_INFO "btrfs: balance in progress\n");
2203 		ret = -EINVAL;
2204 		goto out;
2205 	}
2206 
2207 	vol_args = memdup_user(arg, sizeof(*vol_args));
2208 	if (IS_ERR(vol_args)) {
2209 		ret = PTR_ERR(vol_args);
2210 		goto out;
2211 	}
2212 
2213 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2214 	ret = btrfs_init_new_device(root, vol_args->name);
2215 
2216 	kfree(vol_args);
2217 out:
2218 	mutex_unlock(&root->fs_info->volume_mutex);
2219 	return ret;
2220 }
2221 
2222 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2223 {
2224 	struct btrfs_ioctl_vol_args *vol_args;
2225 	int ret;
2226 
2227 	if (!capable(CAP_SYS_ADMIN))
2228 		return -EPERM;
2229 
2230 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2231 		return -EROFS;
2232 
2233 	mutex_lock(&root->fs_info->volume_mutex);
2234 	if (root->fs_info->balance_ctl) {
2235 		printk(KERN_INFO "btrfs: balance in progress\n");
2236 		ret = -EINVAL;
2237 		goto out;
2238 	}
2239 
2240 	vol_args = memdup_user(arg, sizeof(*vol_args));
2241 	if (IS_ERR(vol_args)) {
2242 		ret = PTR_ERR(vol_args);
2243 		goto out;
2244 	}
2245 
2246 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2247 	ret = btrfs_rm_device(root, vol_args->name);
2248 
2249 	kfree(vol_args);
2250 out:
2251 	mutex_unlock(&root->fs_info->volume_mutex);
2252 	return ret;
2253 }
2254 
2255 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2256 {
2257 	struct btrfs_ioctl_fs_info_args *fi_args;
2258 	struct btrfs_device *device;
2259 	struct btrfs_device *next;
2260 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2261 	int ret = 0;
2262 
2263 	if (!capable(CAP_SYS_ADMIN))
2264 		return -EPERM;
2265 
2266 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2267 	if (!fi_args)
2268 		return -ENOMEM;
2269 
2270 	fi_args->num_devices = fs_devices->num_devices;
2271 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2272 
2273 	mutex_lock(&fs_devices->device_list_mutex);
2274 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2275 		if (device->devid > fi_args->max_id)
2276 			fi_args->max_id = device->devid;
2277 	}
2278 	mutex_unlock(&fs_devices->device_list_mutex);
2279 
2280 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2281 		ret = -EFAULT;
2282 
2283 	kfree(fi_args);
2284 	return ret;
2285 }
2286 
2287 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2288 {
2289 	struct btrfs_ioctl_dev_info_args *di_args;
2290 	struct btrfs_device *dev;
2291 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2292 	int ret = 0;
2293 	char *s_uuid = NULL;
2294 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2295 
2296 	if (!capable(CAP_SYS_ADMIN))
2297 		return -EPERM;
2298 
2299 	di_args = memdup_user(arg, sizeof(*di_args));
2300 	if (IS_ERR(di_args))
2301 		return PTR_ERR(di_args);
2302 
2303 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2304 		s_uuid = di_args->uuid;
2305 
2306 	mutex_lock(&fs_devices->device_list_mutex);
2307 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2308 	mutex_unlock(&fs_devices->device_list_mutex);
2309 
2310 	if (!dev) {
2311 		ret = -ENODEV;
2312 		goto out;
2313 	}
2314 
2315 	di_args->devid = dev->devid;
2316 	di_args->bytes_used = dev->bytes_used;
2317 	di_args->total_bytes = dev->total_bytes;
2318 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2319 	if (dev->name) {
2320 		struct rcu_string *name;
2321 
2322 		rcu_read_lock();
2323 		name = rcu_dereference(dev->name);
2324 		strncpy(di_args->path, name->str, sizeof(di_args->path));
2325 		rcu_read_unlock();
2326 		di_args->path[sizeof(di_args->path) - 1] = 0;
2327 	} else {
2328 		di_args->path[0] = '\0';
2329 	}
2330 
2331 out:
2332 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2333 		ret = -EFAULT;
2334 
2335 	kfree(di_args);
2336 	return ret;
2337 }
2338 
2339 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2340 				       u64 off, u64 olen, u64 destoff)
2341 {
2342 	struct inode *inode = fdentry(file)->d_inode;
2343 	struct btrfs_root *root = BTRFS_I(inode)->root;
2344 	struct file *src_file;
2345 	struct inode *src;
2346 	struct btrfs_trans_handle *trans;
2347 	struct btrfs_path *path;
2348 	struct extent_buffer *leaf;
2349 	char *buf;
2350 	struct btrfs_key key;
2351 	u32 nritems;
2352 	int slot;
2353 	int ret;
2354 	u64 len = olen;
2355 	u64 bs = root->fs_info->sb->s_blocksize;
2356 	u64 hint_byte;
2357 
2358 	/*
2359 	 * TODO:
2360 	 * - split compressed inline extents.  annoying: we need to
2361 	 *   decompress into destination's address_space (the file offset
2362 	 *   may change, so source mapping won't do), then recompress (or
2363 	 *   otherwise reinsert) a subrange.
2364 	 * - allow ranges within the same file to be cloned (provided
2365 	 *   they don't overlap)?
2366 	 */
2367 
2368 	/* the destination must be opened for writing */
2369 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2370 		return -EINVAL;
2371 
2372 	if (btrfs_root_readonly(root))
2373 		return -EROFS;
2374 
2375 	ret = mnt_want_write_file(file);
2376 	if (ret)
2377 		return ret;
2378 
2379 	src_file = fget(srcfd);
2380 	if (!src_file) {
2381 		ret = -EBADF;
2382 		goto out_drop_write;
2383 	}
2384 
2385 	ret = -EXDEV;
2386 	if (src_file->f_path.mnt != file->f_path.mnt)
2387 		goto out_fput;
2388 
2389 	src = src_file->f_dentry->d_inode;
2390 
2391 	ret = -EINVAL;
2392 	if (src == inode)
2393 		goto out_fput;
2394 
2395 	/* the src must be open for reading */
2396 	if (!(src_file->f_mode & FMODE_READ))
2397 		goto out_fput;
2398 
2399 	/* don't make the dst file partly checksummed */
2400 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2401 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2402 		goto out_fput;
2403 
2404 	ret = -EISDIR;
2405 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2406 		goto out_fput;
2407 
2408 	ret = -EXDEV;
2409 	if (src->i_sb != inode->i_sb)
2410 		goto out_fput;
2411 
2412 	ret = -ENOMEM;
2413 	buf = vmalloc(btrfs_level_size(root, 0));
2414 	if (!buf)
2415 		goto out_fput;
2416 
2417 	path = btrfs_alloc_path();
2418 	if (!path) {
2419 		vfree(buf);
2420 		goto out_fput;
2421 	}
2422 	path->reada = 2;
2423 
2424 	if (inode < src) {
2425 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2426 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2427 	} else {
2428 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2429 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2430 	}
2431 
2432 	/* determine range to clone */
2433 	ret = -EINVAL;
2434 	if (off + len > src->i_size || off + len < off)
2435 		goto out_unlock;
2436 	if (len == 0)
2437 		olen = len = src->i_size - off;
2438 	/* if we extend to eof, continue to block boundary */
2439 	if (off + len == src->i_size)
2440 		len = ALIGN(src->i_size, bs) - off;
2441 
2442 	/* verify the end result is block aligned */
2443 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2444 	    !IS_ALIGNED(destoff, bs))
2445 		goto out_unlock;
2446 
2447 	if (destoff > inode->i_size) {
2448 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2449 		if (ret)
2450 			goto out_unlock;
2451 	}
2452 
2453 	/* truncate page cache pages from target inode range */
2454 	truncate_inode_pages_range(&inode->i_data, destoff,
2455 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2456 
2457 	/* do any pending delalloc/csum calc on src, one way or
2458 	   another, and lock file content */
2459 	while (1) {
2460 		struct btrfs_ordered_extent *ordered;
2461 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2462 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2463 		if (!ordered &&
2464 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2465 				   EXTENT_DELALLOC, 0, NULL))
2466 			break;
2467 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2468 		if (ordered)
2469 			btrfs_put_ordered_extent(ordered);
2470 		btrfs_wait_ordered_range(src, off, len);
2471 	}
2472 
2473 	/* clone data */
2474 	key.objectid = btrfs_ino(src);
2475 	key.type = BTRFS_EXTENT_DATA_KEY;
2476 	key.offset = 0;
2477 
2478 	while (1) {
2479 		/*
2480 		 * note the key will change type as we walk through the
2481 		 * tree.
2482 		 */
2483 		ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2484 				0, 0);
2485 		if (ret < 0)
2486 			goto out;
2487 
2488 		nritems = btrfs_header_nritems(path->nodes[0]);
2489 		if (path->slots[0] >= nritems) {
2490 			ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2491 			if (ret < 0)
2492 				goto out;
2493 			if (ret > 0)
2494 				break;
2495 			nritems = btrfs_header_nritems(path->nodes[0]);
2496 		}
2497 		leaf = path->nodes[0];
2498 		slot = path->slots[0];
2499 
2500 		btrfs_item_key_to_cpu(leaf, &key, slot);
2501 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2502 		    key.objectid != btrfs_ino(src))
2503 			break;
2504 
2505 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2506 			struct btrfs_file_extent_item *extent;
2507 			int type;
2508 			u32 size;
2509 			struct btrfs_key new_key;
2510 			u64 disko = 0, diskl = 0;
2511 			u64 datao = 0, datal = 0;
2512 			u8 comp;
2513 			u64 endoff;
2514 
2515 			size = btrfs_item_size_nr(leaf, slot);
2516 			read_extent_buffer(leaf, buf,
2517 					   btrfs_item_ptr_offset(leaf, slot),
2518 					   size);
2519 
2520 			extent = btrfs_item_ptr(leaf, slot,
2521 						struct btrfs_file_extent_item);
2522 			comp = btrfs_file_extent_compression(leaf, extent);
2523 			type = btrfs_file_extent_type(leaf, extent);
2524 			if (type == BTRFS_FILE_EXTENT_REG ||
2525 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2526 				disko = btrfs_file_extent_disk_bytenr(leaf,
2527 								      extent);
2528 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2529 								 extent);
2530 				datao = btrfs_file_extent_offset(leaf, extent);
2531 				datal = btrfs_file_extent_num_bytes(leaf,
2532 								    extent);
2533 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2534 				/* take upper bound, may be compressed */
2535 				datal = btrfs_file_extent_ram_bytes(leaf,
2536 								    extent);
2537 			}
2538 			btrfs_release_path(path);
2539 
2540 			if (key.offset + datal <= off ||
2541 			    key.offset >= off+len)
2542 				goto next;
2543 
2544 			memcpy(&new_key, &key, sizeof(new_key));
2545 			new_key.objectid = btrfs_ino(inode);
2546 			if (off <= key.offset)
2547 				new_key.offset = key.offset + destoff - off;
2548 			else
2549 				new_key.offset = destoff;
2550 
2551 			/*
2552 			 * 1 - adjusting old extent (we may have to split it)
2553 			 * 1 - add new extent
2554 			 * 1 - inode update
2555 			 */
2556 			trans = btrfs_start_transaction(root, 3);
2557 			if (IS_ERR(trans)) {
2558 				ret = PTR_ERR(trans);
2559 				goto out;
2560 			}
2561 
2562 			if (type == BTRFS_FILE_EXTENT_REG ||
2563 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2564 				/*
2565 				 *    a  | --- range to clone ---|  b
2566 				 * | ------------- extent ------------- |
2567 				 */
2568 
2569 				/* substract range b */
2570 				if (key.offset + datal > off + len)
2571 					datal = off + len - key.offset;
2572 
2573 				/* substract range a */
2574 				if (off > key.offset) {
2575 					datao += off - key.offset;
2576 					datal -= off - key.offset;
2577 				}
2578 
2579 				ret = btrfs_drop_extents(trans, inode,
2580 							 new_key.offset,
2581 							 new_key.offset + datal,
2582 							 &hint_byte, 1);
2583 				if (ret) {
2584 					btrfs_abort_transaction(trans, root,
2585 								ret);
2586 					btrfs_end_transaction(trans, root);
2587 					goto out;
2588 				}
2589 
2590 				ret = btrfs_insert_empty_item(trans, root, path,
2591 							      &new_key, size);
2592 				if (ret) {
2593 					btrfs_abort_transaction(trans, root,
2594 								ret);
2595 					btrfs_end_transaction(trans, root);
2596 					goto out;
2597 				}
2598 
2599 				leaf = path->nodes[0];
2600 				slot = path->slots[0];
2601 				write_extent_buffer(leaf, buf,
2602 					    btrfs_item_ptr_offset(leaf, slot),
2603 					    size);
2604 
2605 				extent = btrfs_item_ptr(leaf, slot,
2606 						struct btrfs_file_extent_item);
2607 
2608 				/* disko == 0 means it's a hole */
2609 				if (!disko)
2610 					datao = 0;
2611 
2612 				btrfs_set_file_extent_offset(leaf, extent,
2613 							     datao);
2614 				btrfs_set_file_extent_num_bytes(leaf, extent,
2615 								datal);
2616 				if (disko) {
2617 					inode_add_bytes(inode, datal);
2618 					ret = btrfs_inc_extent_ref(trans, root,
2619 							disko, diskl, 0,
2620 							root->root_key.objectid,
2621 							btrfs_ino(inode),
2622 							new_key.offset - datao,
2623 							0);
2624 					if (ret) {
2625 						btrfs_abort_transaction(trans,
2626 									root,
2627 									ret);
2628 						btrfs_end_transaction(trans,
2629 								      root);
2630 						goto out;
2631 
2632 					}
2633 				}
2634 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2635 				u64 skip = 0;
2636 				u64 trim = 0;
2637 				if (off > key.offset) {
2638 					skip = off - key.offset;
2639 					new_key.offset += skip;
2640 				}
2641 
2642 				if (key.offset + datal > off+len)
2643 					trim = key.offset + datal - (off+len);
2644 
2645 				if (comp && (skip || trim)) {
2646 					ret = -EINVAL;
2647 					btrfs_end_transaction(trans, root);
2648 					goto out;
2649 				}
2650 				size -= skip + trim;
2651 				datal -= skip + trim;
2652 
2653 				ret = btrfs_drop_extents(trans, inode,
2654 							 new_key.offset,
2655 							 new_key.offset + datal,
2656 							 &hint_byte, 1);
2657 				if (ret) {
2658 					btrfs_abort_transaction(trans, root,
2659 								ret);
2660 					btrfs_end_transaction(trans, root);
2661 					goto out;
2662 				}
2663 
2664 				ret = btrfs_insert_empty_item(trans, root, path,
2665 							      &new_key, size);
2666 				if (ret) {
2667 					btrfs_abort_transaction(trans, root,
2668 								ret);
2669 					btrfs_end_transaction(trans, root);
2670 					goto out;
2671 				}
2672 
2673 				if (skip) {
2674 					u32 start =
2675 					  btrfs_file_extent_calc_inline_size(0);
2676 					memmove(buf+start, buf+start+skip,
2677 						datal);
2678 				}
2679 
2680 				leaf = path->nodes[0];
2681 				slot = path->slots[0];
2682 				write_extent_buffer(leaf, buf,
2683 					    btrfs_item_ptr_offset(leaf, slot),
2684 					    size);
2685 				inode_add_bytes(inode, datal);
2686 			}
2687 
2688 			btrfs_mark_buffer_dirty(leaf);
2689 			btrfs_release_path(path);
2690 
2691 			inode_inc_iversion(inode);
2692 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2693 
2694 			/*
2695 			 * we round up to the block size at eof when
2696 			 * determining which extents to clone above,
2697 			 * but shouldn't round up the file size
2698 			 */
2699 			endoff = new_key.offset + datal;
2700 			if (endoff > destoff+olen)
2701 				endoff = destoff+olen;
2702 			if (endoff > inode->i_size)
2703 				btrfs_i_size_write(inode, endoff);
2704 
2705 			ret = btrfs_update_inode(trans, root, inode);
2706 			if (ret) {
2707 				btrfs_abort_transaction(trans, root, ret);
2708 				btrfs_end_transaction(trans, root);
2709 				goto out;
2710 			}
2711 			ret = btrfs_end_transaction(trans, root);
2712 		}
2713 next:
2714 		btrfs_release_path(path);
2715 		key.offset++;
2716 	}
2717 	ret = 0;
2718 out:
2719 	btrfs_release_path(path);
2720 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2721 out_unlock:
2722 	mutex_unlock(&src->i_mutex);
2723 	mutex_unlock(&inode->i_mutex);
2724 	vfree(buf);
2725 	btrfs_free_path(path);
2726 out_fput:
2727 	fput(src_file);
2728 out_drop_write:
2729 	mnt_drop_write_file(file);
2730 	return ret;
2731 }
2732 
2733 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2734 {
2735 	struct btrfs_ioctl_clone_range_args args;
2736 
2737 	if (copy_from_user(&args, argp, sizeof(args)))
2738 		return -EFAULT;
2739 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2740 				 args.src_length, args.dest_offset);
2741 }
2742 
2743 /*
2744  * there are many ways the trans_start and trans_end ioctls can lead
2745  * to deadlocks.  They should only be used by applications that
2746  * basically own the machine, and have a very in depth understanding
2747  * of all the possible deadlocks and enospc problems.
2748  */
2749 static long btrfs_ioctl_trans_start(struct file *file)
2750 {
2751 	struct inode *inode = fdentry(file)->d_inode;
2752 	struct btrfs_root *root = BTRFS_I(inode)->root;
2753 	struct btrfs_trans_handle *trans;
2754 	int ret;
2755 
2756 	ret = -EPERM;
2757 	if (!capable(CAP_SYS_ADMIN))
2758 		goto out;
2759 
2760 	ret = -EINPROGRESS;
2761 	if (file->private_data)
2762 		goto out;
2763 
2764 	ret = -EROFS;
2765 	if (btrfs_root_readonly(root))
2766 		goto out;
2767 
2768 	ret = mnt_want_write_file(file);
2769 	if (ret)
2770 		goto out;
2771 
2772 	atomic_inc(&root->fs_info->open_ioctl_trans);
2773 
2774 	ret = -ENOMEM;
2775 	trans = btrfs_start_ioctl_transaction(root);
2776 	if (IS_ERR(trans))
2777 		goto out_drop;
2778 
2779 	file->private_data = trans;
2780 	return 0;
2781 
2782 out_drop:
2783 	atomic_dec(&root->fs_info->open_ioctl_trans);
2784 	mnt_drop_write_file(file);
2785 out:
2786 	return ret;
2787 }
2788 
2789 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2790 {
2791 	struct inode *inode = fdentry(file)->d_inode;
2792 	struct btrfs_root *root = BTRFS_I(inode)->root;
2793 	struct btrfs_root *new_root;
2794 	struct btrfs_dir_item *di;
2795 	struct btrfs_trans_handle *trans;
2796 	struct btrfs_path *path;
2797 	struct btrfs_key location;
2798 	struct btrfs_disk_key disk_key;
2799 	u64 objectid = 0;
2800 	u64 dir_id;
2801 
2802 	if (!capable(CAP_SYS_ADMIN))
2803 		return -EPERM;
2804 
2805 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2806 		return -EFAULT;
2807 
2808 	if (!objectid)
2809 		objectid = root->root_key.objectid;
2810 
2811 	location.objectid = objectid;
2812 	location.type = BTRFS_ROOT_ITEM_KEY;
2813 	location.offset = (u64)-1;
2814 
2815 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2816 	if (IS_ERR(new_root))
2817 		return PTR_ERR(new_root);
2818 
2819 	if (btrfs_root_refs(&new_root->root_item) == 0)
2820 		return -ENOENT;
2821 
2822 	path = btrfs_alloc_path();
2823 	if (!path)
2824 		return -ENOMEM;
2825 	path->leave_spinning = 1;
2826 
2827 	trans = btrfs_start_transaction(root, 1);
2828 	if (IS_ERR(trans)) {
2829 		btrfs_free_path(path);
2830 		return PTR_ERR(trans);
2831 	}
2832 
2833 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2834 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2835 				   dir_id, "default", 7, 1);
2836 	if (IS_ERR_OR_NULL(di)) {
2837 		btrfs_free_path(path);
2838 		btrfs_end_transaction(trans, root);
2839 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2840 		       "this isn't going to work\n");
2841 		return -ENOENT;
2842 	}
2843 
2844 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2845 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2846 	btrfs_mark_buffer_dirty(path->nodes[0]);
2847 	btrfs_free_path(path);
2848 
2849 	btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2850 	btrfs_end_transaction(trans, root);
2851 
2852 	return 0;
2853 }
2854 
2855 static void get_block_group_info(struct list_head *groups_list,
2856 				 struct btrfs_ioctl_space_info *space)
2857 {
2858 	struct btrfs_block_group_cache *block_group;
2859 
2860 	space->total_bytes = 0;
2861 	space->used_bytes = 0;
2862 	space->flags = 0;
2863 	list_for_each_entry(block_group, groups_list, list) {
2864 		space->flags = block_group->flags;
2865 		space->total_bytes += block_group->key.offset;
2866 		space->used_bytes +=
2867 			btrfs_block_group_used(&block_group->item);
2868 	}
2869 }
2870 
2871 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2872 {
2873 	struct btrfs_ioctl_space_args space_args;
2874 	struct btrfs_ioctl_space_info space;
2875 	struct btrfs_ioctl_space_info *dest;
2876 	struct btrfs_ioctl_space_info *dest_orig;
2877 	struct btrfs_ioctl_space_info __user *user_dest;
2878 	struct btrfs_space_info *info;
2879 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2880 		       BTRFS_BLOCK_GROUP_SYSTEM,
2881 		       BTRFS_BLOCK_GROUP_METADATA,
2882 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2883 	int num_types = 4;
2884 	int alloc_size;
2885 	int ret = 0;
2886 	u64 slot_count = 0;
2887 	int i, c;
2888 
2889 	if (copy_from_user(&space_args,
2890 			   (struct btrfs_ioctl_space_args __user *)arg,
2891 			   sizeof(space_args)))
2892 		return -EFAULT;
2893 
2894 	for (i = 0; i < num_types; i++) {
2895 		struct btrfs_space_info *tmp;
2896 
2897 		info = NULL;
2898 		rcu_read_lock();
2899 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2900 					list) {
2901 			if (tmp->flags == types[i]) {
2902 				info = tmp;
2903 				break;
2904 			}
2905 		}
2906 		rcu_read_unlock();
2907 
2908 		if (!info)
2909 			continue;
2910 
2911 		down_read(&info->groups_sem);
2912 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2913 			if (!list_empty(&info->block_groups[c]))
2914 				slot_count++;
2915 		}
2916 		up_read(&info->groups_sem);
2917 	}
2918 
2919 	/* space_slots == 0 means they are asking for a count */
2920 	if (space_args.space_slots == 0) {
2921 		space_args.total_spaces = slot_count;
2922 		goto out;
2923 	}
2924 
2925 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2926 
2927 	alloc_size = sizeof(*dest) * slot_count;
2928 
2929 	/* we generally have at most 6 or so space infos, one for each raid
2930 	 * level.  So, a whole page should be more than enough for everyone
2931 	 */
2932 	if (alloc_size > PAGE_CACHE_SIZE)
2933 		return -ENOMEM;
2934 
2935 	space_args.total_spaces = 0;
2936 	dest = kmalloc(alloc_size, GFP_NOFS);
2937 	if (!dest)
2938 		return -ENOMEM;
2939 	dest_orig = dest;
2940 
2941 	/* now we have a buffer to copy into */
2942 	for (i = 0; i < num_types; i++) {
2943 		struct btrfs_space_info *tmp;
2944 
2945 		if (!slot_count)
2946 			break;
2947 
2948 		info = NULL;
2949 		rcu_read_lock();
2950 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2951 					list) {
2952 			if (tmp->flags == types[i]) {
2953 				info = tmp;
2954 				break;
2955 			}
2956 		}
2957 		rcu_read_unlock();
2958 
2959 		if (!info)
2960 			continue;
2961 		down_read(&info->groups_sem);
2962 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2963 			if (!list_empty(&info->block_groups[c])) {
2964 				get_block_group_info(&info->block_groups[c],
2965 						     &space);
2966 				memcpy(dest, &space, sizeof(space));
2967 				dest++;
2968 				space_args.total_spaces++;
2969 				slot_count--;
2970 			}
2971 			if (!slot_count)
2972 				break;
2973 		}
2974 		up_read(&info->groups_sem);
2975 	}
2976 
2977 	user_dest = (struct btrfs_ioctl_space_info __user *)
2978 		(arg + sizeof(struct btrfs_ioctl_space_args));
2979 
2980 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2981 		ret = -EFAULT;
2982 
2983 	kfree(dest_orig);
2984 out:
2985 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2986 		ret = -EFAULT;
2987 
2988 	return ret;
2989 }
2990 
2991 /*
2992  * there are many ways the trans_start and trans_end ioctls can lead
2993  * to deadlocks.  They should only be used by applications that
2994  * basically own the machine, and have a very in depth understanding
2995  * of all the possible deadlocks and enospc problems.
2996  */
2997 long btrfs_ioctl_trans_end(struct file *file)
2998 {
2999 	struct inode *inode = fdentry(file)->d_inode;
3000 	struct btrfs_root *root = BTRFS_I(inode)->root;
3001 	struct btrfs_trans_handle *trans;
3002 
3003 	trans = file->private_data;
3004 	if (!trans)
3005 		return -EINVAL;
3006 	file->private_data = NULL;
3007 
3008 	btrfs_end_transaction(trans, root);
3009 
3010 	atomic_dec(&root->fs_info->open_ioctl_trans);
3011 
3012 	mnt_drop_write_file(file);
3013 	return 0;
3014 }
3015 
3016 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
3017 {
3018 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3019 	struct btrfs_trans_handle *trans;
3020 	u64 transid;
3021 	int ret;
3022 
3023 	trans = btrfs_start_transaction(root, 0);
3024 	if (IS_ERR(trans))
3025 		return PTR_ERR(trans);
3026 	transid = trans->transid;
3027 	ret = btrfs_commit_transaction_async(trans, root, 0);
3028 	if (ret) {
3029 		btrfs_end_transaction(trans, root);
3030 		return ret;
3031 	}
3032 
3033 	if (argp)
3034 		if (copy_to_user(argp, &transid, sizeof(transid)))
3035 			return -EFAULT;
3036 	return 0;
3037 }
3038 
3039 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3040 {
3041 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3042 	u64 transid;
3043 
3044 	if (argp) {
3045 		if (copy_from_user(&transid, argp, sizeof(transid)))
3046 			return -EFAULT;
3047 	} else {
3048 		transid = 0;  /* current trans */
3049 	}
3050 	return btrfs_wait_for_commit(root, transid);
3051 }
3052 
3053 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3054 {
3055 	int ret;
3056 	struct btrfs_ioctl_scrub_args *sa;
3057 
3058 	if (!capable(CAP_SYS_ADMIN))
3059 		return -EPERM;
3060 
3061 	sa = memdup_user(arg, sizeof(*sa));
3062 	if (IS_ERR(sa))
3063 		return PTR_ERR(sa);
3064 
3065 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3066 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3067 
3068 	if (copy_to_user(arg, sa, sizeof(*sa)))
3069 		ret = -EFAULT;
3070 
3071 	kfree(sa);
3072 	return ret;
3073 }
3074 
3075 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3076 {
3077 	if (!capable(CAP_SYS_ADMIN))
3078 		return -EPERM;
3079 
3080 	return btrfs_scrub_cancel(root);
3081 }
3082 
3083 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3084 				       void __user *arg)
3085 {
3086 	struct btrfs_ioctl_scrub_args *sa;
3087 	int ret;
3088 
3089 	if (!capable(CAP_SYS_ADMIN))
3090 		return -EPERM;
3091 
3092 	sa = memdup_user(arg, sizeof(*sa));
3093 	if (IS_ERR(sa))
3094 		return PTR_ERR(sa);
3095 
3096 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3097 
3098 	if (copy_to_user(arg, sa, sizeof(*sa)))
3099 		ret = -EFAULT;
3100 
3101 	kfree(sa);
3102 	return ret;
3103 }
3104 
3105 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3106 				      void __user *arg)
3107 {
3108 	struct btrfs_ioctl_get_dev_stats *sa;
3109 	int ret;
3110 
3111 	sa = memdup_user(arg, sizeof(*sa));
3112 	if (IS_ERR(sa))
3113 		return PTR_ERR(sa);
3114 
3115 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3116 		kfree(sa);
3117 		return -EPERM;
3118 	}
3119 
3120 	ret = btrfs_get_dev_stats(root, sa);
3121 
3122 	if (copy_to_user(arg, sa, sizeof(*sa)))
3123 		ret = -EFAULT;
3124 
3125 	kfree(sa);
3126 	return ret;
3127 }
3128 
3129 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3130 {
3131 	int ret = 0;
3132 	int i;
3133 	u64 rel_ptr;
3134 	int size;
3135 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3136 	struct inode_fs_paths *ipath = NULL;
3137 	struct btrfs_path *path;
3138 
3139 	if (!capable(CAP_SYS_ADMIN))
3140 		return -EPERM;
3141 
3142 	path = btrfs_alloc_path();
3143 	if (!path) {
3144 		ret = -ENOMEM;
3145 		goto out;
3146 	}
3147 
3148 	ipa = memdup_user(arg, sizeof(*ipa));
3149 	if (IS_ERR(ipa)) {
3150 		ret = PTR_ERR(ipa);
3151 		ipa = NULL;
3152 		goto out;
3153 	}
3154 
3155 	size = min_t(u32, ipa->size, 4096);
3156 	ipath = init_ipath(size, root, path);
3157 	if (IS_ERR(ipath)) {
3158 		ret = PTR_ERR(ipath);
3159 		ipath = NULL;
3160 		goto out;
3161 	}
3162 
3163 	ret = paths_from_inode(ipa->inum, ipath);
3164 	if (ret < 0)
3165 		goto out;
3166 
3167 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3168 		rel_ptr = ipath->fspath->val[i] -
3169 			  (u64)(unsigned long)ipath->fspath->val;
3170 		ipath->fspath->val[i] = rel_ptr;
3171 	}
3172 
3173 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3174 			   (void *)(unsigned long)ipath->fspath, size);
3175 	if (ret) {
3176 		ret = -EFAULT;
3177 		goto out;
3178 	}
3179 
3180 out:
3181 	btrfs_free_path(path);
3182 	free_ipath(ipath);
3183 	kfree(ipa);
3184 
3185 	return ret;
3186 }
3187 
3188 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3189 {
3190 	struct btrfs_data_container *inodes = ctx;
3191 	const size_t c = 3 * sizeof(u64);
3192 
3193 	if (inodes->bytes_left >= c) {
3194 		inodes->bytes_left -= c;
3195 		inodes->val[inodes->elem_cnt] = inum;
3196 		inodes->val[inodes->elem_cnt + 1] = offset;
3197 		inodes->val[inodes->elem_cnt + 2] = root;
3198 		inodes->elem_cnt += 3;
3199 	} else {
3200 		inodes->bytes_missing += c - inodes->bytes_left;
3201 		inodes->bytes_left = 0;
3202 		inodes->elem_missed += 3;
3203 	}
3204 
3205 	return 0;
3206 }
3207 
3208 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3209 					void __user *arg)
3210 {
3211 	int ret = 0;
3212 	int size;
3213 	u64 extent_item_pos;
3214 	struct btrfs_ioctl_logical_ino_args *loi;
3215 	struct btrfs_data_container *inodes = NULL;
3216 	struct btrfs_path *path = NULL;
3217 	struct btrfs_key key;
3218 
3219 	if (!capable(CAP_SYS_ADMIN))
3220 		return -EPERM;
3221 
3222 	loi = memdup_user(arg, sizeof(*loi));
3223 	if (IS_ERR(loi)) {
3224 		ret = PTR_ERR(loi);
3225 		loi = NULL;
3226 		goto out;
3227 	}
3228 
3229 	path = btrfs_alloc_path();
3230 	if (!path) {
3231 		ret = -ENOMEM;
3232 		goto out;
3233 	}
3234 
3235 	size = min_t(u32, loi->size, 4096);
3236 	inodes = init_data_container(size);
3237 	if (IS_ERR(inodes)) {
3238 		ret = PTR_ERR(inodes);
3239 		inodes = NULL;
3240 		goto out;
3241 	}
3242 
3243 	ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3244 	btrfs_release_path(path);
3245 
3246 	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3247 		ret = -ENOENT;
3248 	if (ret < 0)
3249 		goto out;
3250 
3251 	extent_item_pos = loi->logical - key.objectid;
3252 	ret = iterate_extent_inodes(root->fs_info, key.objectid,
3253 					extent_item_pos, 0, build_ino_list,
3254 					inodes);
3255 
3256 	if (ret < 0)
3257 		goto out;
3258 
3259 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3260 			   (void *)(unsigned long)inodes, size);
3261 	if (ret)
3262 		ret = -EFAULT;
3263 
3264 out:
3265 	btrfs_free_path(path);
3266 	kfree(inodes);
3267 	kfree(loi);
3268 
3269 	return ret;
3270 }
3271 
3272 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3273 			       struct btrfs_ioctl_balance_args *bargs)
3274 {
3275 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3276 
3277 	bargs->flags = bctl->flags;
3278 
3279 	if (atomic_read(&fs_info->balance_running))
3280 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3281 	if (atomic_read(&fs_info->balance_pause_req))
3282 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3283 	if (atomic_read(&fs_info->balance_cancel_req))
3284 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3285 
3286 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3287 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3288 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3289 
3290 	if (lock) {
3291 		spin_lock(&fs_info->balance_lock);
3292 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3293 		spin_unlock(&fs_info->balance_lock);
3294 	} else {
3295 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3296 	}
3297 }
3298 
3299 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3300 {
3301 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3302 	struct btrfs_fs_info *fs_info = root->fs_info;
3303 	struct btrfs_ioctl_balance_args *bargs;
3304 	struct btrfs_balance_control *bctl;
3305 	int ret;
3306 
3307 	if (!capable(CAP_SYS_ADMIN))
3308 		return -EPERM;
3309 
3310 	ret = mnt_want_write_file(file);
3311 	if (ret)
3312 		return ret;
3313 
3314 	mutex_lock(&fs_info->volume_mutex);
3315 	mutex_lock(&fs_info->balance_mutex);
3316 
3317 	if (arg) {
3318 		bargs = memdup_user(arg, sizeof(*bargs));
3319 		if (IS_ERR(bargs)) {
3320 			ret = PTR_ERR(bargs);
3321 			goto out;
3322 		}
3323 
3324 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3325 			if (!fs_info->balance_ctl) {
3326 				ret = -ENOTCONN;
3327 				goto out_bargs;
3328 			}
3329 
3330 			bctl = fs_info->balance_ctl;
3331 			spin_lock(&fs_info->balance_lock);
3332 			bctl->flags |= BTRFS_BALANCE_RESUME;
3333 			spin_unlock(&fs_info->balance_lock);
3334 
3335 			goto do_balance;
3336 		}
3337 	} else {
3338 		bargs = NULL;
3339 	}
3340 
3341 	if (fs_info->balance_ctl) {
3342 		ret = -EINPROGRESS;
3343 		goto out_bargs;
3344 	}
3345 
3346 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3347 	if (!bctl) {
3348 		ret = -ENOMEM;
3349 		goto out_bargs;
3350 	}
3351 
3352 	bctl->fs_info = fs_info;
3353 	if (arg) {
3354 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3355 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3356 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3357 
3358 		bctl->flags = bargs->flags;
3359 	} else {
3360 		/* balance everything - no filters */
3361 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3362 	}
3363 
3364 do_balance:
3365 	ret = btrfs_balance(bctl, bargs);
3366 	/*
3367 	 * bctl is freed in __cancel_balance or in free_fs_info if
3368 	 * restriper was paused all the way until unmount
3369 	 */
3370 	if (arg) {
3371 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3372 			ret = -EFAULT;
3373 	}
3374 
3375 out_bargs:
3376 	kfree(bargs);
3377 out:
3378 	mutex_unlock(&fs_info->balance_mutex);
3379 	mutex_unlock(&fs_info->volume_mutex);
3380 	mnt_drop_write_file(file);
3381 	return ret;
3382 }
3383 
3384 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3385 {
3386 	if (!capable(CAP_SYS_ADMIN))
3387 		return -EPERM;
3388 
3389 	switch (cmd) {
3390 	case BTRFS_BALANCE_CTL_PAUSE:
3391 		return btrfs_pause_balance(root->fs_info);
3392 	case BTRFS_BALANCE_CTL_CANCEL:
3393 		return btrfs_cancel_balance(root->fs_info);
3394 	}
3395 
3396 	return -EINVAL;
3397 }
3398 
3399 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3400 					 void __user *arg)
3401 {
3402 	struct btrfs_fs_info *fs_info = root->fs_info;
3403 	struct btrfs_ioctl_balance_args *bargs;
3404 	int ret = 0;
3405 
3406 	if (!capable(CAP_SYS_ADMIN))
3407 		return -EPERM;
3408 
3409 	mutex_lock(&fs_info->balance_mutex);
3410 	if (!fs_info->balance_ctl) {
3411 		ret = -ENOTCONN;
3412 		goto out;
3413 	}
3414 
3415 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3416 	if (!bargs) {
3417 		ret = -ENOMEM;
3418 		goto out;
3419 	}
3420 
3421 	update_ioctl_balance_args(fs_info, 1, bargs);
3422 
3423 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3424 		ret = -EFAULT;
3425 
3426 	kfree(bargs);
3427 out:
3428 	mutex_unlock(&fs_info->balance_mutex);
3429 	return ret;
3430 }
3431 
3432 static long btrfs_ioctl_quota_ctl(struct btrfs_root *root, void __user *arg)
3433 {
3434 	struct btrfs_ioctl_quota_ctl_args *sa;
3435 	struct btrfs_trans_handle *trans = NULL;
3436 	int ret;
3437 	int err;
3438 
3439 	if (!capable(CAP_SYS_ADMIN))
3440 		return -EPERM;
3441 
3442 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3443 		return -EROFS;
3444 
3445 	sa = memdup_user(arg, sizeof(*sa));
3446 	if (IS_ERR(sa))
3447 		return PTR_ERR(sa);
3448 
3449 	if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3450 		trans = btrfs_start_transaction(root, 2);
3451 		if (IS_ERR(trans)) {
3452 			ret = PTR_ERR(trans);
3453 			goto out;
3454 		}
3455 	}
3456 
3457 	switch (sa->cmd) {
3458 	case BTRFS_QUOTA_CTL_ENABLE:
3459 		ret = btrfs_quota_enable(trans, root->fs_info);
3460 		break;
3461 	case BTRFS_QUOTA_CTL_DISABLE:
3462 		ret = btrfs_quota_disable(trans, root->fs_info);
3463 		break;
3464 	case BTRFS_QUOTA_CTL_RESCAN:
3465 		ret = btrfs_quota_rescan(root->fs_info);
3466 		break;
3467 	default:
3468 		ret = -EINVAL;
3469 		break;
3470 	}
3471 
3472 	if (copy_to_user(arg, sa, sizeof(*sa)))
3473 		ret = -EFAULT;
3474 
3475 	if (trans) {
3476 		err = btrfs_commit_transaction(trans, root);
3477 		if (err && !ret)
3478 			ret = err;
3479 	}
3480 
3481 out:
3482 	kfree(sa);
3483 	return ret;
3484 }
3485 
3486 static long btrfs_ioctl_qgroup_assign(struct btrfs_root *root, void __user *arg)
3487 {
3488 	struct btrfs_ioctl_qgroup_assign_args *sa;
3489 	struct btrfs_trans_handle *trans;
3490 	int ret;
3491 	int err;
3492 
3493 	if (!capable(CAP_SYS_ADMIN))
3494 		return -EPERM;
3495 
3496 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3497 		return -EROFS;
3498 
3499 	sa = memdup_user(arg, sizeof(*sa));
3500 	if (IS_ERR(sa))
3501 		return PTR_ERR(sa);
3502 
3503 	trans = btrfs_join_transaction(root);
3504 	if (IS_ERR(trans)) {
3505 		ret = PTR_ERR(trans);
3506 		goto out;
3507 	}
3508 
3509 	/* FIXME: check if the IDs really exist */
3510 	if (sa->assign) {
3511 		ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3512 						sa->src, sa->dst);
3513 	} else {
3514 		ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3515 						sa->src, sa->dst);
3516 	}
3517 
3518 	err = btrfs_end_transaction(trans, root);
3519 	if (err && !ret)
3520 		ret = err;
3521 
3522 out:
3523 	kfree(sa);
3524 	return ret;
3525 }
3526 
3527 static long btrfs_ioctl_qgroup_create(struct btrfs_root *root, void __user *arg)
3528 {
3529 	struct btrfs_ioctl_qgroup_create_args *sa;
3530 	struct btrfs_trans_handle *trans;
3531 	int ret;
3532 	int err;
3533 
3534 	if (!capable(CAP_SYS_ADMIN))
3535 		return -EPERM;
3536 
3537 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3538 		return -EROFS;
3539 
3540 	sa = memdup_user(arg, sizeof(*sa));
3541 	if (IS_ERR(sa))
3542 		return PTR_ERR(sa);
3543 
3544 	trans = btrfs_join_transaction(root);
3545 	if (IS_ERR(trans)) {
3546 		ret = PTR_ERR(trans);
3547 		goto out;
3548 	}
3549 
3550 	/* FIXME: check if the IDs really exist */
3551 	if (sa->create) {
3552 		ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3553 					  NULL);
3554 	} else {
3555 		ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3556 	}
3557 
3558 	err = btrfs_end_transaction(trans, root);
3559 	if (err && !ret)
3560 		ret = err;
3561 
3562 out:
3563 	kfree(sa);
3564 	return ret;
3565 }
3566 
3567 static long btrfs_ioctl_qgroup_limit(struct btrfs_root *root, void __user *arg)
3568 {
3569 	struct btrfs_ioctl_qgroup_limit_args *sa;
3570 	struct btrfs_trans_handle *trans;
3571 	int ret;
3572 	int err;
3573 	u64 qgroupid;
3574 
3575 	if (!capable(CAP_SYS_ADMIN))
3576 		return -EPERM;
3577 
3578 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3579 		return -EROFS;
3580 
3581 	sa = memdup_user(arg, sizeof(*sa));
3582 	if (IS_ERR(sa))
3583 		return PTR_ERR(sa);
3584 
3585 	trans = btrfs_join_transaction(root);
3586 	if (IS_ERR(trans)) {
3587 		ret = PTR_ERR(trans);
3588 		goto out;
3589 	}
3590 
3591 	qgroupid = sa->qgroupid;
3592 	if (!qgroupid) {
3593 		/* take the current subvol as qgroup */
3594 		qgroupid = root->root_key.objectid;
3595 	}
3596 
3597 	/* FIXME: check if the IDs really exist */
3598 	ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3599 
3600 	err = btrfs_end_transaction(trans, root);
3601 	if (err && !ret)
3602 		ret = err;
3603 
3604 out:
3605 	kfree(sa);
3606 	return ret;
3607 }
3608 
3609 static long btrfs_ioctl_set_received_subvol(struct file *file,
3610 					    void __user *arg)
3611 {
3612 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
3613 	struct inode *inode = fdentry(file)->d_inode;
3614 	struct btrfs_root *root = BTRFS_I(inode)->root;
3615 	struct btrfs_root_item *root_item = &root->root_item;
3616 	struct btrfs_trans_handle *trans;
3617 	struct timespec ct = CURRENT_TIME;
3618 	int ret = 0;
3619 
3620 	ret = mnt_want_write_file(file);
3621 	if (ret < 0)
3622 		return ret;
3623 
3624 	down_write(&root->fs_info->subvol_sem);
3625 
3626 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3627 		ret = -EINVAL;
3628 		goto out;
3629 	}
3630 
3631 	if (btrfs_root_readonly(root)) {
3632 		ret = -EROFS;
3633 		goto out;
3634 	}
3635 
3636 	if (!inode_owner_or_capable(inode)) {
3637 		ret = -EACCES;
3638 		goto out;
3639 	}
3640 
3641 	sa = memdup_user(arg, sizeof(*sa));
3642 	if (IS_ERR(sa)) {
3643 		ret = PTR_ERR(sa);
3644 		sa = NULL;
3645 		goto out;
3646 	}
3647 
3648 	trans = btrfs_start_transaction(root, 1);
3649 	if (IS_ERR(trans)) {
3650 		ret = PTR_ERR(trans);
3651 		trans = NULL;
3652 		goto out;
3653 	}
3654 
3655 	sa->rtransid = trans->transid;
3656 	sa->rtime.sec = ct.tv_sec;
3657 	sa->rtime.nsec = ct.tv_nsec;
3658 
3659 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3660 	btrfs_set_root_stransid(root_item, sa->stransid);
3661 	btrfs_set_root_rtransid(root_item, sa->rtransid);
3662 	root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3663 	root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3664 	root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3665 	root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3666 
3667 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
3668 				&root->root_key, &root->root_item);
3669 	if (ret < 0) {
3670 		btrfs_end_transaction(trans, root);
3671 		trans = NULL;
3672 		goto out;
3673 	} else {
3674 		ret = btrfs_commit_transaction(trans, root);
3675 		if (ret < 0)
3676 			goto out;
3677 	}
3678 
3679 	ret = copy_to_user(arg, sa, sizeof(*sa));
3680 	if (ret)
3681 		ret = -EFAULT;
3682 
3683 out:
3684 	kfree(sa);
3685 	up_write(&root->fs_info->subvol_sem);
3686 	mnt_drop_write_file(file);
3687 	return ret;
3688 }
3689 
3690 long btrfs_ioctl(struct file *file, unsigned int
3691 		cmd, unsigned long arg)
3692 {
3693 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3694 	void __user *argp = (void __user *)arg;
3695 
3696 	switch (cmd) {
3697 	case FS_IOC_GETFLAGS:
3698 		return btrfs_ioctl_getflags(file, argp);
3699 	case FS_IOC_SETFLAGS:
3700 		return btrfs_ioctl_setflags(file, argp);
3701 	case FS_IOC_GETVERSION:
3702 		return btrfs_ioctl_getversion(file, argp);
3703 	case FITRIM:
3704 		return btrfs_ioctl_fitrim(file, argp);
3705 	case BTRFS_IOC_SNAP_CREATE:
3706 		return btrfs_ioctl_snap_create(file, argp, 0);
3707 	case BTRFS_IOC_SNAP_CREATE_V2:
3708 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3709 	case BTRFS_IOC_SUBVOL_CREATE:
3710 		return btrfs_ioctl_snap_create(file, argp, 1);
3711 	case BTRFS_IOC_SUBVOL_CREATE_V2:
3712 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
3713 	case BTRFS_IOC_SNAP_DESTROY:
3714 		return btrfs_ioctl_snap_destroy(file, argp);
3715 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3716 		return btrfs_ioctl_subvol_getflags(file, argp);
3717 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3718 		return btrfs_ioctl_subvol_setflags(file, argp);
3719 	case BTRFS_IOC_DEFAULT_SUBVOL:
3720 		return btrfs_ioctl_default_subvol(file, argp);
3721 	case BTRFS_IOC_DEFRAG:
3722 		return btrfs_ioctl_defrag(file, NULL);
3723 	case BTRFS_IOC_DEFRAG_RANGE:
3724 		return btrfs_ioctl_defrag(file, argp);
3725 	case BTRFS_IOC_RESIZE:
3726 		return btrfs_ioctl_resize(root, argp);
3727 	case BTRFS_IOC_ADD_DEV:
3728 		return btrfs_ioctl_add_dev(root, argp);
3729 	case BTRFS_IOC_RM_DEV:
3730 		return btrfs_ioctl_rm_dev(root, argp);
3731 	case BTRFS_IOC_FS_INFO:
3732 		return btrfs_ioctl_fs_info(root, argp);
3733 	case BTRFS_IOC_DEV_INFO:
3734 		return btrfs_ioctl_dev_info(root, argp);
3735 	case BTRFS_IOC_BALANCE:
3736 		return btrfs_ioctl_balance(file, NULL);
3737 	case BTRFS_IOC_CLONE:
3738 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3739 	case BTRFS_IOC_CLONE_RANGE:
3740 		return btrfs_ioctl_clone_range(file, argp);
3741 	case BTRFS_IOC_TRANS_START:
3742 		return btrfs_ioctl_trans_start(file);
3743 	case BTRFS_IOC_TRANS_END:
3744 		return btrfs_ioctl_trans_end(file);
3745 	case BTRFS_IOC_TREE_SEARCH:
3746 		return btrfs_ioctl_tree_search(file, argp);
3747 	case BTRFS_IOC_INO_LOOKUP:
3748 		return btrfs_ioctl_ino_lookup(file, argp);
3749 	case BTRFS_IOC_INO_PATHS:
3750 		return btrfs_ioctl_ino_to_path(root, argp);
3751 	case BTRFS_IOC_LOGICAL_INO:
3752 		return btrfs_ioctl_logical_to_ino(root, argp);
3753 	case BTRFS_IOC_SPACE_INFO:
3754 		return btrfs_ioctl_space_info(root, argp);
3755 	case BTRFS_IOC_SYNC:
3756 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3757 		return 0;
3758 	case BTRFS_IOC_START_SYNC:
3759 		return btrfs_ioctl_start_sync(file, argp);
3760 	case BTRFS_IOC_WAIT_SYNC:
3761 		return btrfs_ioctl_wait_sync(file, argp);
3762 	case BTRFS_IOC_SCRUB:
3763 		return btrfs_ioctl_scrub(root, argp);
3764 	case BTRFS_IOC_SCRUB_CANCEL:
3765 		return btrfs_ioctl_scrub_cancel(root, argp);
3766 	case BTRFS_IOC_SCRUB_PROGRESS:
3767 		return btrfs_ioctl_scrub_progress(root, argp);
3768 	case BTRFS_IOC_BALANCE_V2:
3769 		return btrfs_ioctl_balance(file, argp);
3770 	case BTRFS_IOC_BALANCE_CTL:
3771 		return btrfs_ioctl_balance_ctl(root, arg);
3772 	case BTRFS_IOC_BALANCE_PROGRESS:
3773 		return btrfs_ioctl_balance_progress(root, argp);
3774 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
3775 		return btrfs_ioctl_set_received_subvol(file, argp);
3776 	case BTRFS_IOC_SEND:
3777 		return btrfs_ioctl_send(file, argp);
3778 	case BTRFS_IOC_GET_DEV_STATS:
3779 		return btrfs_ioctl_get_dev_stats(root, argp);
3780 	case BTRFS_IOC_QUOTA_CTL:
3781 		return btrfs_ioctl_quota_ctl(root, argp);
3782 	case BTRFS_IOC_QGROUP_ASSIGN:
3783 		return btrfs_ioctl_qgroup_assign(root, argp);
3784 	case BTRFS_IOC_QGROUP_CREATE:
3785 		return btrfs_ioctl_qgroup_create(root, argp);
3786 	case BTRFS_IOC_QGROUP_LIMIT:
3787 		return btrfs_ioctl_qgroup_limit(root, argp);
3788 	}
3789 
3790 	return -ENOTTY;
3791 }
3792