xref: /linux/fs/btrfs/ioctl.c (revision 61d0b5a4b2777dcf5daef245e212b3c1fa8091ca)
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 <linux/btrfs.h>
46 #include "compat.h"
47 #include "ctree.h"
48 #include "disk-io.h"
49 #include "transaction.h"
50 #include "btrfs_inode.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 #include "dev-replace.h"
59 
60 /* Mask out flags that are inappropriate for the given type of inode. */
61 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
62 {
63 	if (S_ISDIR(mode))
64 		return flags;
65 	else if (S_ISREG(mode))
66 		return flags & ~FS_DIRSYNC_FL;
67 	else
68 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
69 }
70 
71 /*
72  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
73  */
74 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
75 {
76 	unsigned int iflags = 0;
77 
78 	if (flags & BTRFS_INODE_SYNC)
79 		iflags |= FS_SYNC_FL;
80 	if (flags & BTRFS_INODE_IMMUTABLE)
81 		iflags |= FS_IMMUTABLE_FL;
82 	if (flags & BTRFS_INODE_APPEND)
83 		iflags |= FS_APPEND_FL;
84 	if (flags & BTRFS_INODE_NODUMP)
85 		iflags |= FS_NODUMP_FL;
86 	if (flags & BTRFS_INODE_NOATIME)
87 		iflags |= FS_NOATIME_FL;
88 	if (flags & BTRFS_INODE_DIRSYNC)
89 		iflags |= FS_DIRSYNC_FL;
90 	if (flags & BTRFS_INODE_NODATACOW)
91 		iflags |= FS_NOCOW_FL;
92 
93 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
94 		iflags |= FS_COMPR_FL;
95 	else if (flags & BTRFS_INODE_NOCOMPRESS)
96 		iflags |= FS_NOCOMP_FL;
97 
98 	return iflags;
99 }
100 
101 /*
102  * Update inode->i_flags based on the btrfs internal flags.
103  */
104 void btrfs_update_iflags(struct inode *inode)
105 {
106 	struct btrfs_inode *ip = BTRFS_I(inode);
107 
108 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
109 
110 	if (ip->flags & BTRFS_INODE_SYNC)
111 		inode->i_flags |= S_SYNC;
112 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
113 		inode->i_flags |= S_IMMUTABLE;
114 	if (ip->flags & BTRFS_INODE_APPEND)
115 		inode->i_flags |= S_APPEND;
116 	if (ip->flags & BTRFS_INODE_NOATIME)
117 		inode->i_flags |= S_NOATIME;
118 	if (ip->flags & BTRFS_INODE_DIRSYNC)
119 		inode->i_flags |= S_DIRSYNC;
120 }
121 
122 /*
123  * Inherit flags from the parent inode.
124  *
125  * Currently only the compression flags and the cow flags are inherited.
126  */
127 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
128 {
129 	unsigned int flags;
130 
131 	if (!dir)
132 		return;
133 
134 	flags = BTRFS_I(dir)->flags;
135 
136 	if (flags & BTRFS_INODE_NOCOMPRESS) {
137 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
138 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
139 	} else if (flags & BTRFS_INODE_COMPRESS) {
140 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
141 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
142 	}
143 
144 	if (flags & BTRFS_INODE_NODATACOW) {
145 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
146 		if (S_ISREG(inode->i_mode))
147 			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
148 	}
149 
150 	btrfs_update_iflags(inode);
151 }
152 
153 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
154 {
155 	struct btrfs_inode *ip = BTRFS_I(file_inode(file));
156 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
157 
158 	if (copy_to_user(arg, &flags, sizeof(flags)))
159 		return -EFAULT;
160 	return 0;
161 }
162 
163 static int check_flags(unsigned int flags)
164 {
165 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
166 		      FS_NOATIME_FL | FS_NODUMP_FL | \
167 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
168 		      FS_NOCOMP_FL | FS_COMPR_FL |
169 		      FS_NOCOW_FL))
170 		return -EOPNOTSUPP;
171 
172 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
173 		return -EINVAL;
174 
175 	return 0;
176 }
177 
178 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
179 {
180 	struct inode *inode = file_inode(file);
181 	struct btrfs_inode *ip = BTRFS_I(inode);
182 	struct btrfs_root *root = ip->root;
183 	struct btrfs_trans_handle *trans;
184 	unsigned int flags, oldflags;
185 	int ret;
186 	u64 ip_oldflags;
187 	unsigned int i_oldflags;
188 	umode_t mode;
189 
190 	if (btrfs_root_readonly(root))
191 		return -EROFS;
192 
193 	if (copy_from_user(&flags, arg, sizeof(flags)))
194 		return -EFAULT;
195 
196 	ret = check_flags(flags);
197 	if (ret)
198 		return ret;
199 
200 	if (!inode_owner_or_capable(inode))
201 		return -EACCES;
202 
203 	ret = mnt_want_write_file(file);
204 	if (ret)
205 		return ret;
206 
207 	mutex_lock(&inode->i_mutex);
208 
209 	ip_oldflags = ip->flags;
210 	i_oldflags = inode->i_flags;
211 	mode = inode->i_mode;
212 
213 	flags = btrfs_mask_flags(inode->i_mode, flags);
214 	oldflags = btrfs_flags_to_ioctl(ip->flags);
215 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
216 		if (!capable(CAP_LINUX_IMMUTABLE)) {
217 			ret = -EPERM;
218 			goto out_unlock;
219 		}
220 	}
221 
222 	if (flags & FS_SYNC_FL)
223 		ip->flags |= BTRFS_INODE_SYNC;
224 	else
225 		ip->flags &= ~BTRFS_INODE_SYNC;
226 	if (flags & FS_IMMUTABLE_FL)
227 		ip->flags |= BTRFS_INODE_IMMUTABLE;
228 	else
229 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
230 	if (flags & FS_APPEND_FL)
231 		ip->flags |= BTRFS_INODE_APPEND;
232 	else
233 		ip->flags &= ~BTRFS_INODE_APPEND;
234 	if (flags & FS_NODUMP_FL)
235 		ip->flags |= BTRFS_INODE_NODUMP;
236 	else
237 		ip->flags &= ~BTRFS_INODE_NODUMP;
238 	if (flags & FS_NOATIME_FL)
239 		ip->flags |= BTRFS_INODE_NOATIME;
240 	else
241 		ip->flags &= ~BTRFS_INODE_NOATIME;
242 	if (flags & FS_DIRSYNC_FL)
243 		ip->flags |= BTRFS_INODE_DIRSYNC;
244 	else
245 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
246 	if (flags & FS_NOCOW_FL) {
247 		if (S_ISREG(mode)) {
248 			/*
249 			 * It's safe to turn csums off here, no extents exist.
250 			 * Otherwise we want the flag to reflect the real COW
251 			 * status of the file and will not set it.
252 			 */
253 			if (inode->i_size == 0)
254 				ip->flags |= BTRFS_INODE_NODATACOW
255 					   | BTRFS_INODE_NODATASUM;
256 		} else {
257 			ip->flags |= BTRFS_INODE_NODATACOW;
258 		}
259 	} else {
260 		/*
261 		 * Revert back under same assuptions as above
262 		 */
263 		if (S_ISREG(mode)) {
264 			if (inode->i_size == 0)
265 				ip->flags &= ~(BTRFS_INODE_NODATACOW
266 				             | BTRFS_INODE_NODATASUM);
267 		} else {
268 			ip->flags &= ~BTRFS_INODE_NODATACOW;
269 		}
270 	}
271 
272 	/*
273 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
274 	 * flag may be changed automatically if compression code won't make
275 	 * things smaller.
276 	 */
277 	if (flags & FS_NOCOMP_FL) {
278 		ip->flags &= ~BTRFS_INODE_COMPRESS;
279 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
280 	} else if (flags & FS_COMPR_FL) {
281 		ip->flags |= BTRFS_INODE_COMPRESS;
282 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
283 	} else {
284 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
285 	}
286 
287 	trans = btrfs_start_transaction(root, 1);
288 	if (IS_ERR(trans)) {
289 		ret = PTR_ERR(trans);
290 		goto out_drop;
291 	}
292 
293 	btrfs_update_iflags(inode);
294 	inode_inc_iversion(inode);
295 	inode->i_ctime = CURRENT_TIME;
296 	ret = btrfs_update_inode(trans, root, inode);
297 
298 	btrfs_end_transaction(trans, root);
299  out_drop:
300 	if (ret) {
301 		ip->flags = ip_oldflags;
302 		inode->i_flags = i_oldflags;
303 	}
304 
305  out_unlock:
306 	mutex_unlock(&inode->i_mutex);
307 	mnt_drop_write_file(file);
308 	return ret;
309 }
310 
311 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
312 {
313 	struct inode *inode = file_inode(file);
314 
315 	return put_user(inode->i_generation, arg);
316 }
317 
318 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
319 {
320 	struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
321 	struct btrfs_device *device;
322 	struct request_queue *q;
323 	struct fstrim_range range;
324 	u64 minlen = ULLONG_MAX;
325 	u64 num_devices = 0;
326 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
327 	int ret;
328 
329 	if (!capable(CAP_SYS_ADMIN))
330 		return -EPERM;
331 
332 	rcu_read_lock();
333 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
334 				dev_list) {
335 		if (!device->bdev)
336 			continue;
337 		q = bdev_get_queue(device->bdev);
338 		if (blk_queue_discard(q)) {
339 			num_devices++;
340 			minlen = min((u64)q->limits.discard_granularity,
341 				     minlen);
342 		}
343 	}
344 	rcu_read_unlock();
345 
346 	if (!num_devices)
347 		return -EOPNOTSUPP;
348 	if (copy_from_user(&range, arg, sizeof(range)))
349 		return -EFAULT;
350 	if (range.start > total_bytes ||
351 	    range.len < fs_info->sb->s_blocksize)
352 		return -EINVAL;
353 
354 	range.len = min(range.len, total_bytes - range.start);
355 	range.minlen = max(range.minlen, minlen);
356 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
357 	if (ret < 0)
358 		return ret;
359 
360 	if (copy_to_user(arg, &range, sizeof(range)))
361 		return -EFAULT;
362 
363 	return 0;
364 }
365 
366 static noinline int create_subvol(struct inode *dir,
367 				  struct dentry *dentry,
368 				  char *name, int namelen,
369 				  u64 *async_transid,
370 				  struct btrfs_qgroup_inherit *inherit)
371 {
372 	struct btrfs_trans_handle *trans;
373 	struct btrfs_key key;
374 	struct btrfs_root_item root_item;
375 	struct btrfs_inode_item *inode_item;
376 	struct extent_buffer *leaf;
377 	struct btrfs_root *root = BTRFS_I(dir)->root;
378 	struct btrfs_root *new_root;
379 	struct btrfs_block_rsv block_rsv;
380 	struct timespec cur_time = CURRENT_TIME;
381 	int ret;
382 	int err;
383 	u64 objectid;
384 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
385 	u64 index = 0;
386 	u64 qgroup_reserved;
387 	uuid_le new_uuid;
388 
389 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
390 	if (ret)
391 		return ret;
392 
393 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
394 	/*
395 	 * The same as the snapshot creation, please see the comment
396 	 * of create_snapshot().
397 	 */
398 	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
399 					       7, &qgroup_reserved);
400 	if (ret)
401 		return ret;
402 
403 	trans = btrfs_start_transaction(root, 0);
404 	if (IS_ERR(trans)) {
405 		ret = PTR_ERR(trans);
406 		goto out;
407 	}
408 	trans->block_rsv = &block_rsv;
409 	trans->bytes_reserved = block_rsv.size;
410 
411 	ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
412 	if (ret)
413 		goto fail;
414 
415 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
416 				      0, objectid, NULL, 0, 0, 0);
417 	if (IS_ERR(leaf)) {
418 		ret = PTR_ERR(leaf);
419 		goto fail;
420 	}
421 
422 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
423 	btrfs_set_header_bytenr(leaf, leaf->start);
424 	btrfs_set_header_generation(leaf, trans->transid);
425 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
426 	btrfs_set_header_owner(leaf, objectid);
427 
428 	write_extent_buffer(leaf, root->fs_info->fsid,
429 			    (unsigned long)btrfs_header_fsid(leaf),
430 			    BTRFS_FSID_SIZE);
431 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
432 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
433 			    BTRFS_UUID_SIZE);
434 	btrfs_mark_buffer_dirty(leaf);
435 
436 	memset(&root_item, 0, sizeof(root_item));
437 
438 	inode_item = &root_item.inode;
439 	inode_item->generation = cpu_to_le64(1);
440 	inode_item->size = cpu_to_le64(3);
441 	inode_item->nlink = cpu_to_le32(1);
442 	inode_item->nbytes = cpu_to_le64(root->leafsize);
443 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
444 
445 	root_item.flags = 0;
446 	root_item.byte_limit = 0;
447 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
448 
449 	btrfs_set_root_bytenr(&root_item, leaf->start);
450 	btrfs_set_root_generation(&root_item, trans->transid);
451 	btrfs_set_root_level(&root_item, 0);
452 	btrfs_set_root_refs(&root_item, 1);
453 	btrfs_set_root_used(&root_item, leaf->len);
454 	btrfs_set_root_last_snapshot(&root_item, 0);
455 
456 	btrfs_set_root_generation_v2(&root_item,
457 			btrfs_root_generation(&root_item));
458 	uuid_le_gen(&new_uuid);
459 	memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
460 	root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
461 	root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
462 	root_item.ctime = root_item.otime;
463 	btrfs_set_root_ctransid(&root_item, trans->transid);
464 	btrfs_set_root_otransid(&root_item, trans->transid);
465 
466 	btrfs_tree_unlock(leaf);
467 	free_extent_buffer(leaf);
468 	leaf = NULL;
469 
470 	btrfs_set_root_dirid(&root_item, new_dirid);
471 
472 	key.objectid = objectid;
473 	key.offset = 0;
474 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
475 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
476 				&root_item);
477 	if (ret)
478 		goto fail;
479 
480 	key.offset = (u64)-1;
481 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
482 	if (IS_ERR(new_root)) {
483 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
484 		ret = PTR_ERR(new_root);
485 		goto fail;
486 	}
487 
488 	btrfs_record_root_in_trans(trans, new_root);
489 
490 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
491 	if (ret) {
492 		/* We potentially lose an unused inode item here */
493 		btrfs_abort_transaction(trans, root, ret);
494 		goto fail;
495 	}
496 
497 	/*
498 	 * insert the directory item
499 	 */
500 	ret = btrfs_set_inode_index(dir, &index);
501 	if (ret) {
502 		btrfs_abort_transaction(trans, root, ret);
503 		goto fail;
504 	}
505 
506 	ret = btrfs_insert_dir_item(trans, root,
507 				    name, namelen, dir, &key,
508 				    BTRFS_FT_DIR, index);
509 	if (ret) {
510 		btrfs_abort_transaction(trans, root, ret);
511 		goto fail;
512 	}
513 
514 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
515 	ret = btrfs_update_inode(trans, root, dir);
516 	BUG_ON(ret);
517 
518 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
519 				 objectid, root->root_key.objectid,
520 				 btrfs_ino(dir), index, name, namelen);
521 
522 	BUG_ON(ret);
523 
524 fail:
525 	trans->block_rsv = NULL;
526 	trans->bytes_reserved = 0;
527 	if (async_transid) {
528 		*async_transid = trans->transid;
529 		err = btrfs_commit_transaction_async(trans, root, 1);
530 	} else {
531 		err = btrfs_commit_transaction(trans, root);
532 	}
533 	if (err && !ret)
534 		ret = err;
535 
536 	if (!ret)
537 		d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
538 out:
539 	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
540 	return ret;
541 }
542 
543 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
544 			   struct dentry *dentry, char *name, int namelen,
545 			   u64 *async_transid, bool readonly,
546 			   struct btrfs_qgroup_inherit *inherit)
547 {
548 	struct inode *inode;
549 	struct btrfs_pending_snapshot *pending_snapshot;
550 	struct btrfs_trans_handle *trans;
551 	int ret;
552 
553 	if (!root->ref_cows)
554 		return -EINVAL;
555 
556 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
557 	if (!pending_snapshot)
558 		return -ENOMEM;
559 
560 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
561 			     BTRFS_BLOCK_RSV_TEMP);
562 	/*
563 	 * 1 - parent dir inode
564 	 * 2 - dir entries
565 	 * 1 - root item
566 	 * 2 - root ref/backref
567 	 * 1 - root of snapshot
568 	 */
569 	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
570 					&pending_snapshot->block_rsv, 7,
571 					&pending_snapshot->qgroup_reserved);
572 	if (ret)
573 		goto out;
574 
575 	pending_snapshot->dentry = dentry;
576 	pending_snapshot->root = root;
577 	pending_snapshot->readonly = readonly;
578 	pending_snapshot->dir = dir;
579 	pending_snapshot->inherit = inherit;
580 
581 	trans = btrfs_start_transaction(root, 0);
582 	if (IS_ERR(trans)) {
583 		ret = PTR_ERR(trans);
584 		goto fail;
585 	}
586 
587 	spin_lock(&root->fs_info->trans_lock);
588 	list_add(&pending_snapshot->list,
589 		 &trans->transaction->pending_snapshots);
590 	spin_unlock(&root->fs_info->trans_lock);
591 	if (async_transid) {
592 		*async_transid = trans->transid;
593 		ret = btrfs_commit_transaction_async(trans,
594 				     root->fs_info->extent_root, 1);
595 	} else {
596 		ret = btrfs_commit_transaction(trans,
597 					       root->fs_info->extent_root);
598 	}
599 	if (ret) {
600 		/* cleanup_transaction has freed this for us */
601 		if (trans->aborted)
602 			pending_snapshot = NULL;
603 		goto fail;
604 	}
605 
606 	ret = pending_snapshot->error;
607 	if (ret)
608 		goto fail;
609 
610 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
611 	if (ret)
612 		goto fail;
613 
614 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
615 	if (IS_ERR(inode)) {
616 		ret = PTR_ERR(inode);
617 		goto fail;
618 	}
619 	BUG_ON(!inode);
620 	d_instantiate(dentry, inode);
621 	ret = 0;
622 fail:
623 	btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
624 					 &pending_snapshot->block_rsv,
625 					 pending_snapshot->qgroup_reserved);
626 out:
627 	kfree(pending_snapshot);
628 	return ret;
629 }
630 
631 /*  copy of check_sticky in fs/namei.c()
632 * It's inline, so penalty for filesystems that don't use sticky bit is
633 * minimal.
634 */
635 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
636 {
637 	kuid_t fsuid = current_fsuid();
638 
639 	if (!(dir->i_mode & S_ISVTX))
640 		return 0;
641 	if (uid_eq(inode->i_uid, fsuid))
642 		return 0;
643 	if (uid_eq(dir->i_uid, fsuid))
644 		return 0;
645 	return !capable(CAP_FOWNER);
646 }
647 
648 /*  copy of may_delete in fs/namei.c()
649  *	Check whether we can remove a link victim from directory dir, check
650  *  whether the type of victim is right.
651  *  1. We can't do it if dir is read-only (done in permission())
652  *  2. We should have write and exec permissions on dir
653  *  3. We can't remove anything from append-only dir
654  *  4. We can't do anything with immutable dir (done in permission())
655  *  5. If the sticky bit on dir is set we should either
656  *	a. be owner of dir, or
657  *	b. be owner of victim, or
658  *	c. have CAP_FOWNER capability
659  *  6. If the victim is append-only or immutable we can't do antyhing with
660  *     links pointing to it.
661  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
662  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
663  *  9. We can't remove a root or mountpoint.
664  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
665  *     nfs_async_unlink().
666  */
667 
668 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
669 {
670 	int error;
671 
672 	if (!victim->d_inode)
673 		return -ENOENT;
674 
675 	BUG_ON(victim->d_parent->d_inode != dir);
676 	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
677 
678 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
679 	if (error)
680 		return error;
681 	if (IS_APPEND(dir))
682 		return -EPERM;
683 	if (btrfs_check_sticky(dir, victim->d_inode)||
684 		IS_APPEND(victim->d_inode)||
685 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
686 		return -EPERM;
687 	if (isdir) {
688 		if (!S_ISDIR(victim->d_inode->i_mode))
689 			return -ENOTDIR;
690 		if (IS_ROOT(victim))
691 			return -EBUSY;
692 	} else if (S_ISDIR(victim->d_inode->i_mode))
693 		return -EISDIR;
694 	if (IS_DEADDIR(dir))
695 		return -ENOENT;
696 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
697 		return -EBUSY;
698 	return 0;
699 }
700 
701 /* copy of may_create in fs/namei.c() */
702 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
703 {
704 	if (child->d_inode)
705 		return -EEXIST;
706 	if (IS_DEADDIR(dir))
707 		return -ENOENT;
708 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
709 }
710 
711 /*
712  * Create a new subvolume below @parent.  This is largely modeled after
713  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
714  * inside this filesystem so it's quite a bit simpler.
715  */
716 static noinline int btrfs_mksubvol(struct path *parent,
717 				   char *name, int namelen,
718 				   struct btrfs_root *snap_src,
719 				   u64 *async_transid, bool readonly,
720 				   struct btrfs_qgroup_inherit *inherit)
721 {
722 	struct inode *dir  = parent->dentry->d_inode;
723 	struct dentry *dentry;
724 	int error;
725 
726 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
727 
728 	dentry = lookup_one_len(name, parent->dentry, namelen);
729 	error = PTR_ERR(dentry);
730 	if (IS_ERR(dentry))
731 		goto out_unlock;
732 
733 	error = -EEXIST;
734 	if (dentry->d_inode)
735 		goto out_dput;
736 
737 	error = btrfs_may_create(dir, dentry);
738 	if (error)
739 		goto out_dput;
740 
741 	/*
742 	 * even if this name doesn't exist, we may get hash collisions.
743 	 * check for them now when we can safely fail
744 	 */
745 	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
746 					       dir->i_ino, name,
747 					       namelen);
748 	if (error)
749 		goto out_dput;
750 
751 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
752 
753 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
754 		goto out_up_read;
755 
756 	if (snap_src) {
757 		error = create_snapshot(snap_src, dir, dentry, name, namelen,
758 					async_transid, readonly, inherit);
759 	} else {
760 		error = create_subvol(dir, dentry, name, namelen,
761 				      async_transid, inherit);
762 	}
763 	if (!error)
764 		fsnotify_mkdir(dir, dentry);
765 out_up_read:
766 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
767 out_dput:
768 	dput(dentry);
769 out_unlock:
770 	mutex_unlock(&dir->i_mutex);
771 	return error;
772 }
773 
774 /*
775  * When we're defragging a range, we don't want to kick it off again
776  * if it is really just waiting for delalloc to send it down.
777  * If we find a nice big extent or delalloc range for the bytes in the
778  * file you want to defrag, we return 0 to let you know to skip this
779  * part of the file
780  */
781 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
782 {
783 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
784 	struct extent_map *em = NULL;
785 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
786 	u64 end;
787 
788 	read_lock(&em_tree->lock);
789 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
790 	read_unlock(&em_tree->lock);
791 
792 	if (em) {
793 		end = extent_map_end(em);
794 		free_extent_map(em);
795 		if (end - offset > thresh)
796 			return 0;
797 	}
798 	/* if we already have a nice delalloc here, just stop */
799 	thresh /= 2;
800 	end = count_range_bits(io_tree, &offset, offset + thresh,
801 			       thresh, EXTENT_DELALLOC, 1);
802 	if (end >= thresh)
803 		return 0;
804 	return 1;
805 }
806 
807 /*
808  * helper function to walk through a file and find extents
809  * newer than a specific transid, and smaller than thresh.
810  *
811  * This is used by the defragging code to find new and small
812  * extents
813  */
814 static int find_new_extents(struct btrfs_root *root,
815 			    struct inode *inode, u64 newer_than,
816 			    u64 *off, int thresh)
817 {
818 	struct btrfs_path *path;
819 	struct btrfs_key min_key;
820 	struct btrfs_key max_key;
821 	struct extent_buffer *leaf;
822 	struct btrfs_file_extent_item *extent;
823 	int type;
824 	int ret;
825 	u64 ino = btrfs_ino(inode);
826 
827 	path = btrfs_alloc_path();
828 	if (!path)
829 		return -ENOMEM;
830 
831 	min_key.objectid = ino;
832 	min_key.type = BTRFS_EXTENT_DATA_KEY;
833 	min_key.offset = *off;
834 
835 	max_key.objectid = ino;
836 	max_key.type = (u8)-1;
837 	max_key.offset = (u64)-1;
838 
839 	path->keep_locks = 1;
840 
841 	while(1) {
842 		ret = btrfs_search_forward(root, &min_key, &max_key,
843 					   path, newer_than);
844 		if (ret != 0)
845 			goto none;
846 		if (min_key.objectid != ino)
847 			goto none;
848 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
849 			goto none;
850 
851 		leaf = path->nodes[0];
852 		extent = btrfs_item_ptr(leaf, path->slots[0],
853 					struct btrfs_file_extent_item);
854 
855 		type = btrfs_file_extent_type(leaf, extent);
856 		if (type == BTRFS_FILE_EXTENT_REG &&
857 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
858 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
859 			*off = min_key.offset;
860 			btrfs_free_path(path);
861 			return 0;
862 		}
863 
864 		if (min_key.offset == (u64)-1)
865 			goto none;
866 
867 		min_key.offset++;
868 		btrfs_release_path(path);
869 	}
870 none:
871 	btrfs_free_path(path);
872 	return -ENOENT;
873 }
874 
875 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
876 {
877 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
878 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
879 	struct extent_map *em;
880 	u64 len = PAGE_CACHE_SIZE;
881 
882 	/*
883 	 * hopefully we have this extent in the tree already, try without
884 	 * the full extent lock
885 	 */
886 	read_lock(&em_tree->lock);
887 	em = lookup_extent_mapping(em_tree, start, len);
888 	read_unlock(&em_tree->lock);
889 
890 	if (!em) {
891 		/* get the big lock and read metadata off disk */
892 		lock_extent(io_tree, start, start + len - 1);
893 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
894 		unlock_extent(io_tree, start, start + len - 1);
895 
896 		if (IS_ERR(em))
897 			return NULL;
898 	}
899 
900 	return em;
901 }
902 
903 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
904 {
905 	struct extent_map *next;
906 	bool ret = true;
907 
908 	/* this is the last extent */
909 	if (em->start + em->len >= i_size_read(inode))
910 		return false;
911 
912 	next = defrag_lookup_extent(inode, em->start + em->len);
913 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
914 		ret = false;
915 
916 	free_extent_map(next);
917 	return ret;
918 }
919 
920 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
921 			       u64 *last_len, u64 *skip, u64 *defrag_end,
922 			       int compress)
923 {
924 	struct extent_map *em;
925 	int ret = 1;
926 	bool next_mergeable = true;
927 
928 	/*
929 	 * make sure that once we start defragging an extent, we keep on
930 	 * defragging it
931 	 */
932 	if (start < *defrag_end)
933 		return 1;
934 
935 	*skip = 0;
936 
937 	em = defrag_lookup_extent(inode, start);
938 	if (!em)
939 		return 0;
940 
941 	/* this will cover holes, and inline extents */
942 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
943 		ret = 0;
944 		goto out;
945 	}
946 
947 	next_mergeable = defrag_check_next_extent(inode, em);
948 
949 	/*
950 	 * we hit a real extent, if it is big or the next extent is not a
951 	 * real extent, don't bother defragging it
952 	 */
953 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
954 	    (em->len >= thresh || !next_mergeable))
955 		ret = 0;
956 out:
957 	/*
958 	 * last_len ends up being a counter of how many bytes we've defragged.
959 	 * every time we choose not to defrag an extent, we reset *last_len
960 	 * so that the next tiny extent will force a defrag.
961 	 *
962 	 * The end result of this is that tiny extents before a single big
963 	 * extent will force at least part of that big extent to be defragged.
964 	 */
965 	if (ret) {
966 		*defrag_end = extent_map_end(em);
967 	} else {
968 		*last_len = 0;
969 		*skip = extent_map_end(em);
970 		*defrag_end = 0;
971 	}
972 
973 	free_extent_map(em);
974 	return ret;
975 }
976 
977 /*
978  * it doesn't do much good to defrag one or two pages
979  * at a time.  This pulls in a nice chunk of pages
980  * to COW and defrag.
981  *
982  * It also makes sure the delalloc code has enough
983  * dirty data to avoid making new small extents as part
984  * of the defrag
985  *
986  * It's a good idea to start RA on this range
987  * before calling this.
988  */
989 static int cluster_pages_for_defrag(struct inode *inode,
990 				    struct page **pages,
991 				    unsigned long start_index,
992 				    int num_pages)
993 {
994 	unsigned long file_end;
995 	u64 isize = i_size_read(inode);
996 	u64 page_start;
997 	u64 page_end;
998 	u64 page_cnt;
999 	int ret;
1000 	int i;
1001 	int i_done;
1002 	struct btrfs_ordered_extent *ordered;
1003 	struct extent_state *cached_state = NULL;
1004 	struct extent_io_tree *tree;
1005 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1006 
1007 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1008 	if (!isize || start_index > file_end)
1009 		return 0;
1010 
1011 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1012 
1013 	ret = btrfs_delalloc_reserve_space(inode,
1014 					   page_cnt << PAGE_CACHE_SHIFT);
1015 	if (ret)
1016 		return ret;
1017 	i_done = 0;
1018 	tree = &BTRFS_I(inode)->io_tree;
1019 
1020 	/* step one, lock all the pages */
1021 	for (i = 0; i < page_cnt; i++) {
1022 		struct page *page;
1023 again:
1024 		page = find_or_create_page(inode->i_mapping,
1025 					   start_index + i, mask);
1026 		if (!page)
1027 			break;
1028 
1029 		page_start = page_offset(page);
1030 		page_end = page_start + PAGE_CACHE_SIZE - 1;
1031 		while (1) {
1032 			lock_extent(tree, page_start, page_end);
1033 			ordered = btrfs_lookup_ordered_extent(inode,
1034 							      page_start);
1035 			unlock_extent(tree, page_start, page_end);
1036 			if (!ordered)
1037 				break;
1038 
1039 			unlock_page(page);
1040 			btrfs_start_ordered_extent(inode, ordered, 1);
1041 			btrfs_put_ordered_extent(ordered);
1042 			lock_page(page);
1043 			/*
1044 			 * we unlocked the page above, so we need check if
1045 			 * it was released or not.
1046 			 */
1047 			if (page->mapping != inode->i_mapping) {
1048 				unlock_page(page);
1049 				page_cache_release(page);
1050 				goto again;
1051 			}
1052 		}
1053 
1054 		if (!PageUptodate(page)) {
1055 			btrfs_readpage(NULL, page);
1056 			lock_page(page);
1057 			if (!PageUptodate(page)) {
1058 				unlock_page(page);
1059 				page_cache_release(page);
1060 				ret = -EIO;
1061 				break;
1062 			}
1063 		}
1064 
1065 		if (page->mapping != inode->i_mapping) {
1066 			unlock_page(page);
1067 			page_cache_release(page);
1068 			goto again;
1069 		}
1070 
1071 		pages[i] = page;
1072 		i_done++;
1073 	}
1074 	if (!i_done || ret)
1075 		goto out;
1076 
1077 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
1078 		goto out;
1079 
1080 	/*
1081 	 * so now we have a nice long stream of locked
1082 	 * and up to date pages, lets wait on them
1083 	 */
1084 	for (i = 0; i < i_done; i++)
1085 		wait_on_page_writeback(pages[i]);
1086 
1087 	page_start = page_offset(pages[0]);
1088 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1089 
1090 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1091 			 page_start, page_end - 1, 0, &cached_state);
1092 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1093 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1094 			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1095 			  &cached_state, GFP_NOFS);
1096 
1097 	if (i_done != page_cnt) {
1098 		spin_lock(&BTRFS_I(inode)->lock);
1099 		BTRFS_I(inode)->outstanding_extents++;
1100 		spin_unlock(&BTRFS_I(inode)->lock);
1101 		btrfs_delalloc_release_space(inode,
1102 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1103 	}
1104 
1105 
1106 	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1107 			  &cached_state, GFP_NOFS);
1108 
1109 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1110 			     page_start, page_end - 1, &cached_state,
1111 			     GFP_NOFS);
1112 
1113 	for (i = 0; i < i_done; i++) {
1114 		clear_page_dirty_for_io(pages[i]);
1115 		ClearPageChecked(pages[i]);
1116 		set_page_extent_mapped(pages[i]);
1117 		set_page_dirty(pages[i]);
1118 		unlock_page(pages[i]);
1119 		page_cache_release(pages[i]);
1120 	}
1121 	return i_done;
1122 out:
1123 	for (i = 0; i < i_done; i++) {
1124 		unlock_page(pages[i]);
1125 		page_cache_release(pages[i]);
1126 	}
1127 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1128 	return ret;
1129 
1130 }
1131 
1132 int btrfs_defrag_file(struct inode *inode, struct file *file,
1133 		      struct btrfs_ioctl_defrag_range_args *range,
1134 		      u64 newer_than, unsigned long max_to_defrag)
1135 {
1136 	struct btrfs_root *root = BTRFS_I(inode)->root;
1137 	struct file_ra_state *ra = NULL;
1138 	unsigned long last_index;
1139 	u64 isize = i_size_read(inode);
1140 	u64 last_len = 0;
1141 	u64 skip = 0;
1142 	u64 defrag_end = 0;
1143 	u64 newer_off = range->start;
1144 	unsigned long i;
1145 	unsigned long ra_index = 0;
1146 	int ret;
1147 	int defrag_count = 0;
1148 	int compress_type = BTRFS_COMPRESS_ZLIB;
1149 	int extent_thresh = range->extent_thresh;
1150 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1151 	int cluster = max_cluster;
1152 	u64 new_align = ~((u64)128 * 1024 - 1);
1153 	struct page **pages = NULL;
1154 
1155 	if (extent_thresh == 0)
1156 		extent_thresh = 256 * 1024;
1157 
1158 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1159 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1160 			return -EINVAL;
1161 		if (range->compress_type)
1162 			compress_type = range->compress_type;
1163 	}
1164 
1165 	if (isize == 0)
1166 		return 0;
1167 
1168 	/*
1169 	 * if we were not given a file, allocate a readahead
1170 	 * context
1171 	 */
1172 	if (!file) {
1173 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1174 		if (!ra)
1175 			return -ENOMEM;
1176 		file_ra_state_init(ra, inode->i_mapping);
1177 	} else {
1178 		ra = &file->f_ra;
1179 	}
1180 
1181 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1182 			GFP_NOFS);
1183 	if (!pages) {
1184 		ret = -ENOMEM;
1185 		goto out_ra;
1186 	}
1187 
1188 	/* find the last page to defrag */
1189 	if (range->start + range->len > range->start) {
1190 		last_index = min_t(u64, isize - 1,
1191 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1192 	} else {
1193 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1194 	}
1195 
1196 	if (newer_than) {
1197 		ret = find_new_extents(root, inode, newer_than,
1198 				       &newer_off, 64 * 1024);
1199 		if (!ret) {
1200 			range->start = newer_off;
1201 			/*
1202 			 * we always align our defrag to help keep
1203 			 * the extents in the file evenly spaced
1204 			 */
1205 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1206 		} else
1207 			goto out_ra;
1208 	} else {
1209 		i = range->start >> PAGE_CACHE_SHIFT;
1210 	}
1211 	if (!max_to_defrag)
1212 		max_to_defrag = last_index + 1;
1213 
1214 	/*
1215 	 * make writeback starts from i, so the defrag range can be
1216 	 * written sequentially.
1217 	 */
1218 	if (i < inode->i_mapping->writeback_index)
1219 		inode->i_mapping->writeback_index = i;
1220 
1221 	while (i <= last_index && defrag_count < max_to_defrag &&
1222 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1223 		PAGE_CACHE_SHIFT)) {
1224 		/*
1225 		 * make sure we stop running if someone unmounts
1226 		 * the FS
1227 		 */
1228 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1229 			break;
1230 
1231 		if (btrfs_defrag_cancelled(root->fs_info)) {
1232 			printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1233 			ret = -EAGAIN;
1234 			break;
1235 		}
1236 
1237 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1238 					 extent_thresh, &last_len, &skip,
1239 					 &defrag_end, range->flags &
1240 					 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1241 			unsigned long next;
1242 			/*
1243 			 * the should_defrag function tells us how much to skip
1244 			 * bump our counter by the suggested amount
1245 			 */
1246 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1247 			i = max(i + 1, next);
1248 			continue;
1249 		}
1250 
1251 		if (!newer_than) {
1252 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1253 				   PAGE_CACHE_SHIFT) - i;
1254 			cluster = min(cluster, max_cluster);
1255 		} else {
1256 			cluster = max_cluster;
1257 		}
1258 
1259 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1260 			BTRFS_I(inode)->force_compress = compress_type;
1261 
1262 		if (i + cluster > ra_index) {
1263 			ra_index = max(i, ra_index);
1264 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1265 				       cluster);
1266 			ra_index += max_cluster;
1267 		}
1268 
1269 		mutex_lock(&inode->i_mutex);
1270 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1271 		if (ret < 0) {
1272 			mutex_unlock(&inode->i_mutex);
1273 			goto out_ra;
1274 		}
1275 
1276 		defrag_count += ret;
1277 		balance_dirty_pages_ratelimited(inode->i_mapping);
1278 		mutex_unlock(&inode->i_mutex);
1279 
1280 		if (newer_than) {
1281 			if (newer_off == (u64)-1)
1282 				break;
1283 
1284 			if (ret > 0)
1285 				i += ret;
1286 
1287 			newer_off = max(newer_off + 1,
1288 					(u64)i << PAGE_CACHE_SHIFT);
1289 
1290 			ret = find_new_extents(root, inode,
1291 					       newer_than, &newer_off,
1292 					       64 * 1024);
1293 			if (!ret) {
1294 				range->start = newer_off;
1295 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1296 			} else {
1297 				break;
1298 			}
1299 		} else {
1300 			if (ret > 0) {
1301 				i += ret;
1302 				last_len += ret << PAGE_CACHE_SHIFT;
1303 			} else {
1304 				i++;
1305 				last_len = 0;
1306 			}
1307 		}
1308 	}
1309 
1310 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1311 		filemap_flush(inode->i_mapping);
1312 
1313 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1314 		/* the filemap_flush will queue IO into the worker threads, but
1315 		 * we have to make sure the IO is actually started and that
1316 		 * ordered extents get created before we return
1317 		 */
1318 		atomic_inc(&root->fs_info->async_submit_draining);
1319 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1320 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1321 			wait_event(root->fs_info->async_submit_wait,
1322 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1323 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1324 		}
1325 		atomic_dec(&root->fs_info->async_submit_draining);
1326 
1327 		mutex_lock(&inode->i_mutex);
1328 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1329 		mutex_unlock(&inode->i_mutex);
1330 	}
1331 
1332 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1333 		btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1334 	}
1335 
1336 	ret = defrag_count;
1337 
1338 out_ra:
1339 	if (!file)
1340 		kfree(ra);
1341 	kfree(pages);
1342 	return ret;
1343 }
1344 
1345 static noinline int btrfs_ioctl_resize(struct file *file,
1346 					void __user *arg)
1347 {
1348 	u64 new_size;
1349 	u64 old_size;
1350 	u64 devid = 1;
1351 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1352 	struct btrfs_ioctl_vol_args *vol_args;
1353 	struct btrfs_trans_handle *trans;
1354 	struct btrfs_device *device = NULL;
1355 	char *sizestr;
1356 	char *devstr = NULL;
1357 	int ret = 0;
1358 	int mod = 0;
1359 
1360 	if (!capable(CAP_SYS_ADMIN))
1361 		return -EPERM;
1362 
1363 	ret = mnt_want_write_file(file);
1364 	if (ret)
1365 		return ret;
1366 
1367 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1368 			1)) {
1369 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1370 		mnt_drop_write_file(file);
1371 		return -EINVAL;
1372 	}
1373 
1374 	mutex_lock(&root->fs_info->volume_mutex);
1375 	vol_args = memdup_user(arg, sizeof(*vol_args));
1376 	if (IS_ERR(vol_args)) {
1377 		ret = PTR_ERR(vol_args);
1378 		goto out;
1379 	}
1380 
1381 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1382 
1383 	sizestr = vol_args->name;
1384 	devstr = strchr(sizestr, ':');
1385 	if (devstr) {
1386 		char *end;
1387 		sizestr = devstr + 1;
1388 		*devstr = '\0';
1389 		devstr = vol_args->name;
1390 		devid = simple_strtoull(devstr, &end, 10);
1391 		if (!devid) {
1392 			ret = -EINVAL;
1393 			goto out_free;
1394 		}
1395 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1396 		       (unsigned long long)devid);
1397 	}
1398 
1399 	device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1400 	if (!device) {
1401 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1402 		       (unsigned long long)devid);
1403 		ret = -ENODEV;
1404 		goto out_free;
1405 	}
1406 
1407 	if (!device->writeable) {
1408 		printk(KERN_INFO "btrfs: resizer unable to apply on "
1409 		       "readonly device %llu\n",
1410 		       (unsigned long long)devid);
1411 		ret = -EPERM;
1412 		goto out_free;
1413 	}
1414 
1415 	if (!strcmp(sizestr, "max"))
1416 		new_size = device->bdev->bd_inode->i_size;
1417 	else {
1418 		if (sizestr[0] == '-') {
1419 			mod = -1;
1420 			sizestr++;
1421 		} else if (sizestr[0] == '+') {
1422 			mod = 1;
1423 			sizestr++;
1424 		}
1425 		new_size = memparse(sizestr, NULL);
1426 		if (new_size == 0) {
1427 			ret = -EINVAL;
1428 			goto out_free;
1429 		}
1430 	}
1431 
1432 	if (device->is_tgtdev_for_dev_replace) {
1433 		ret = -EPERM;
1434 		goto out_free;
1435 	}
1436 
1437 	old_size = device->total_bytes;
1438 
1439 	if (mod < 0) {
1440 		if (new_size > old_size) {
1441 			ret = -EINVAL;
1442 			goto out_free;
1443 		}
1444 		new_size = old_size - new_size;
1445 	} else if (mod > 0) {
1446 		new_size = old_size + new_size;
1447 	}
1448 
1449 	if (new_size < 256 * 1024 * 1024) {
1450 		ret = -EINVAL;
1451 		goto out_free;
1452 	}
1453 	if (new_size > device->bdev->bd_inode->i_size) {
1454 		ret = -EFBIG;
1455 		goto out_free;
1456 	}
1457 
1458 	do_div(new_size, root->sectorsize);
1459 	new_size *= root->sectorsize;
1460 
1461 	printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1462 		      rcu_str_deref(device->name),
1463 		      (unsigned long long)new_size);
1464 
1465 	if (new_size > old_size) {
1466 		trans = btrfs_start_transaction(root, 0);
1467 		if (IS_ERR(trans)) {
1468 			ret = PTR_ERR(trans);
1469 			goto out_free;
1470 		}
1471 		ret = btrfs_grow_device(trans, device, new_size);
1472 		btrfs_commit_transaction(trans, root);
1473 	} else if (new_size < old_size) {
1474 		ret = btrfs_shrink_device(device, new_size);
1475 	} /* equal, nothing need to do */
1476 
1477 out_free:
1478 	kfree(vol_args);
1479 out:
1480 	mutex_unlock(&root->fs_info->volume_mutex);
1481 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1482 	mnt_drop_write_file(file);
1483 	return ret;
1484 }
1485 
1486 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1487 				char *name, unsigned long fd, int subvol,
1488 				u64 *transid, bool readonly,
1489 				struct btrfs_qgroup_inherit *inherit)
1490 {
1491 	int namelen;
1492 	int ret = 0;
1493 
1494 	ret = mnt_want_write_file(file);
1495 	if (ret)
1496 		goto out;
1497 
1498 	namelen = strlen(name);
1499 	if (strchr(name, '/')) {
1500 		ret = -EINVAL;
1501 		goto out_drop_write;
1502 	}
1503 
1504 	if (name[0] == '.' &&
1505 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1506 		ret = -EEXIST;
1507 		goto out_drop_write;
1508 	}
1509 
1510 	if (subvol) {
1511 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1512 				     NULL, transid, readonly, inherit);
1513 	} else {
1514 		struct fd src = fdget(fd);
1515 		struct inode *src_inode;
1516 		if (!src.file) {
1517 			ret = -EINVAL;
1518 			goto out_drop_write;
1519 		}
1520 
1521 		src_inode = file_inode(src.file);
1522 		if (src_inode->i_sb != file_inode(file)->i_sb) {
1523 			printk(KERN_INFO "btrfs: Snapshot src from "
1524 			       "another FS\n");
1525 			ret = -EINVAL;
1526 		} else {
1527 			ret = btrfs_mksubvol(&file->f_path, name, namelen,
1528 					     BTRFS_I(src_inode)->root,
1529 					     transid, readonly, inherit);
1530 		}
1531 		fdput(src);
1532 	}
1533 out_drop_write:
1534 	mnt_drop_write_file(file);
1535 out:
1536 	return ret;
1537 }
1538 
1539 static noinline int btrfs_ioctl_snap_create(struct file *file,
1540 					    void __user *arg, int subvol)
1541 {
1542 	struct btrfs_ioctl_vol_args *vol_args;
1543 	int ret;
1544 
1545 	vol_args = memdup_user(arg, sizeof(*vol_args));
1546 	if (IS_ERR(vol_args))
1547 		return PTR_ERR(vol_args);
1548 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1549 
1550 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1551 					      vol_args->fd, subvol,
1552 					      NULL, false, NULL);
1553 
1554 	kfree(vol_args);
1555 	return ret;
1556 }
1557 
1558 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1559 					       void __user *arg, int subvol)
1560 {
1561 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1562 	int ret;
1563 	u64 transid = 0;
1564 	u64 *ptr = NULL;
1565 	bool readonly = false;
1566 	struct btrfs_qgroup_inherit *inherit = NULL;
1567 
1568 	vol_args = memdup_user(arg, sizeof(*vol_args));
1569 	if (IS_ERR(vol_args))
1570 		return PTR_ERR(vol_args);
1571 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1572 
1573 	if (vol_args->flags &
1574 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1575 	      BTRFS_SUBVOL_QGROUP_INHERIT)) {
1576 		ret = -EOPNOTSUPP;
1577 		goto out;
1578 	}
1579 
1580 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1581 		ptr = &transid;
1582 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1583 		readonly = true;
1584 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1585 		if (vol_args->size > PAGE_CACHE_SIZE) {
1586 			ret = -EINVAL;
1587 			goto out;
1588 		}
1589 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1590 		if (IS_ERR(inherit)) {
1591 			ret = PTR_ERR(inherit);
1592 			goto out;
1593 		}
1594 	}
1595 
1596 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1597 					      vol_args->fd, subvol, ptr,
1598 					      readonly, inherit);
1599 
1600 	if (ret == 0 && ptr &&
1601 	    copy_to_user(arg +
1602 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1603 				  transid), ptr, sizeof(*ptr)))
1604 		ret = -EFAULT;
1605 out:
1606 	kfree(vol_args);
1607 	kfree(inherit);
1608 	return ret;
1609 }
1610 
1611 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1612 						void __user *arg)
1613 {
1614 	struct inode *inode = file_inode(file);
1615 	struct btrfs_root *root = BTRFS_I(inode)->root;
1616 	int ret = 0;
1617 	u64 flags = 0;
1618 
1619 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1620 		return -EINVAL;
1621 
1622 	down_read(&root->fs_info->subvol_sem);
1623 	if (btrfs_root_readonly(root))
1624 		flags |= BTRFS_SUBVOL_RDONLY;
1625 	up_read(&root->fs_info->subvol_sem);
1626 
1627 	if (copy_to_user(arg, &flags, sizeof(flags)))
1628 		ret = -EFAULT;
1629 
1630 	return ret;
1631 }
1632 
1633 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1634 					      void __user *arg)
1635 {
1636 	struct inode *inode = file_inode(file);
1637 	struct btrfs_root *root = BTRFS_I(inode)->root;
1638 	struct btrfs_trans_handle *trans;
1639 	u64 root_flags;
1640 	u64 flags;
1641 	int ret = 0;
1642 
1643 	ret = mnt_want_write_file(file);
1644 	if (ret)
1645 		goto out;
1646 
1647 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1648 		ret = -EINVAL;
1649 		goto out_drop_write;
1650 	}
1651 
1652 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1653 		ret = -EFAULT;
1654 		goto out_drop_write;
1655 	}
1656 
1657 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1658 		ret = -EINVAL;
1659 		goto out_drop_write;
1660 	}
1661 
1662 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1663 		ret = -EOPNOTSUPP;
1664 		goto out_drop_write;
1665 	}
1666 
1667 	if (!inode_owner_or_capable(inode)) {
1668 		ret = -EACCES;
1669 		goto out_drop_write;
1670 	}
1671 
1672 	down_write(&root->fs_info->subvol_sem);
1673 
1674 	/* nothing to do */
1675 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1676 		goto out_drop_sem;
1677 
1678 	root_flags = btrfs_root_flags(&root->root_item);
1679 	if (flags & BTRFS_SUBVOL_RDONLY)
1680 		btrfs_set_root_flags(&root->root_item,
1681 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1682 	else
1683 		btrfs_set_root_flags(&root->root_item,
1684 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1685 
1686 	trans = btrfs_start_transaction(root, 1);
1687 	if (IS_ERR(trans)) {
1688 		ret = PTR_ERR(trans);
1689 		goto out_reset;
1690 	}
1691 
1692 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1693 				&root->root_key, &root->root_item);
1694 
1695 	btrfs_commit_transaction(trans, root);
1696 out_reset:
1697 	if (ret)
1698 		btrfs_set_root_flags(&root->root_item, root_flags);
1699 out_drop_sem:
1700 	up_write(&root->fs_info->subvol_sem);
1701 out_drop_write:
1702 	mnt_drop_write_file(file);
1703 out:
1704 	return ret;
1705 }
1706 
1707 /*
1708  * helper to check if the subvolume references other subvolumes
1709  */
1710 static noinline int may_destroy_subvol(struct btrfs_root *root)
1711 {
1712 	struct btrfs_path *path;
1713 	struct btrfs_key key;
1714 	int ret;
1715 
1716 	path = btrfs_alloc_path();
1717 	if (!path)
1718 		return -ENOMEM;
1719 
1720 	key.objectid = root->root_key.objectid;
1721 	key.type = BTRFS_ROOT_REF_KEY;
1722 	key.offset = (u64)-1;
1723 
1724 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1725 				&key, path, 0, 0);
1726 	if (ret < 0)
1727 		goto out;
1728 	BUG_ON(ret == 0);
1729 
1730 	ret = 0;
1731 	if (path->slots[0] > 0) {
1732 		path->slots[0]--;
1733 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1734 		if (key.objectid == root->root_key.objectid &&
1735 		    key.type == BTRFS_ROOT_REF_KEY)
1736 			ret = -ENOTEMPTY;
1737 	}
1738 out:
1739 	btrfs_free_path(path);
1740 	return ret;
1741 }
1742 
1743 static noinline int key_in_sk(struct btrfs_key *key,
1744 			      struct btrfs_ioctl_search_key *sk)
1745 {
1746 	struct btrfs_key test;
1747 	int ret;
1748 
1749 	test.objectid = sk->min_objectid;
1750 	test.type = sk->min_type;
1751 	test.offset = sk->min_offset;
1752 
1753 	ret = btrfs_comp_cpu_keys(key, &test);
1754 	if (ret < 0)
1755 		return 0;
1756 
1757 	test.objectid = sk->max_objectid;
1758 	test.type = sk->max_type;
1759 	test.offset = sk->max_offset;
1760 
1761 	ret = btrfs_comp_cpu_keys(key, &test);
1762 	if (ret > 0)
1763 		return 0;
1764 	return 1;
1765 }
1766 
1767 static noinline int copy_to_sk(struct btrfs_root *root,
1768 			       struct btrfs_path *path,
1769 			       struct btrfs_key *key,
1770 			       struct btrfs_ioctl_search_key *sk,
1771 			       char *buf,
1772 			       unsigned long *sk_offset,
1773 			       int *num_found)
1774 {
1775 	u64 found_transid;
1776 	struct extent_buffer *leaf;
1777 	struct btrfs_ioctl_search_header sh;
1778 	unsigned long item_off;
1779 	unsigned long item_len;
1780 	int nritems;
1781 	int i;
1782 	int slot;
1783 	int ret = 0;
1784 
1785 	leaf = path->nodes[0];
1786 	slot = path->slots[0];
1787 	nritems = btrfs_header_nritems(leaf);
1788 
1789 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1790 		i = nritems;
1791 		goto advance_key;
1792 	}
1793 	found_transid = btrfs_header_generation(leaf);
1794 
1795 	for (i = slot; i < nritems; i++) {
1796 		item_off = btrfs_item_ptr_offset(leaf, i);
1797 		item_len = btrfs_item_size_nr(leaf, i);
1798 
1799 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1800 			item_len = 0;
1801 
1802 		if (sizeof(sh) + item_len + *sk_offset >
1803 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1804 			ret = 1;
1805 			goto overflow;
1806 		}
1807 
1808 		btrfs_item_key_to_cpu(leaf, key, i);
1809 		if (!key_in_sk(key, sk))
1810 			continue;
1811 
1812 		sh.objectid = key->objectid;
1813 		sh.offset = key->offset;
1814 		sh.type = key->type;
1815 		sh.len = item_len;
1816 		sh.transid = found_transid;
1817 
1818 		/* copy search result header */
1819 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1820 		*sk_offset += sizeof(sh);
1821 
1822 		if (item_len) {
1823 			char *p = buf + *sk_offset;
1824 			/* copy the item */
1825 			read_extent_buffer(leaf, p,
1826 					   item_off, item_len);
1827 			*sk_offset += item_len;
1828 		}
1829 		(*num_found)++;
1830 
1831 		if (*num_found >= sk->nr_items)
1832 			break;
1833 	}
1834 advance_key:
1835 	ret = 0;
1836 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1837 		key->offset++;
1838 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1839 		key->offset = 0;
1840 		key->type++;
1841 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1842 		key->offset = 0;
1843 		key->type = 0;
1844 		key->objectid++;
1845 	} else
1846 		ret = 1;
1847 overflow:
1848 	return ret;
1849 }
1850 
1851 static noinline int search_ioctl(struct inode *inode,
1852 				 struct btrfs_ioctl_search_args *args)
1853 {
1854 	struct btrfs_root *root;
1855 	struct btrfs_key key;
1856 	struct btrfs_key max_key;
1857 	struct btrfs_path *path;
1858 	struct btrfs_ioctl_search_key *sk = &args->key;
1859 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1860 	int ret;
1861 	int num_found = 0;
1862 	unsigned long sk_offset = 0;
1863 
1864 	path = btrfs_alloc_path();
1865 	if (!path)
1866 		return -ENOMEM;
1867 
1868 	if (sk->tree_id == 0) {
1869 		/* search the root of the inode that was passed */
1870 		root = BTRFS_I(inode)->root;
1871 	} else {
1872 		key.objectid = sk->tree_id;
1873 		key.type = BTRFS_ROOT_ITEM_KEY;
1874 		key.offset = (u64)-1;
1875 		root = btrfs_read_fs_root_no_name(info, &key);
1876 		if (IS_ERR(root)) {
1877 			printk(KERN_ERR "could not find root %llu\n",
1878 			       sk->tree_id);
1879 			btrfs_free_path(path);
1880 			return -ENOENT;
1881 		}
1882 	}
1883 
1884 	key.objectid = sk->min_objectid;
1885 	key.type = sk->min_type;
1886 	key.offset = sk->min_offset;
1887 
1888 	max_key.objectid = sk->max_objectid;
1889 	max_key.type = sk->max_type;
1890 	max_key.offset = sk->max_offset;
1891 
1892 	path->keep_locks = 1;
1893 
1894 	while(1) {
1895 		ret = btrfs_search_forward(root, &key, &max_key, path,
1896 					   sk->min_transid);
1897 		if (ret != 0) {
1898 			if (ret > 0)
1899 				ret = 0;
1900 			goto err;
1901 		}
1902 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1903 				 &sk_offset, &num_found);
1904 		btrfs_release_path(path);
1905 		if (ret || num_found >= sk->nr_items)
1906 			break;
1907 
1908 	}
1909 	ret = 0;
1910 err:
1911 	sk->nr_items = num_found;
1912 	btrfs_free_path(path);
1913 	return ret;
1914 }
1915 
1916 static noinline int btrfs_ioctl_tree_search(struct file *file,
1917 					   void __user *argp)
1918 {
1919 	 struct btrfs_ioctl_search_args *args;
1920 	 struct inode *inode;
1921 	 int ret;
1922 
1923 	if (!capable(CAP_SYS_ADMIN))
1924 		return -EPERM;
1925 
1926 	args = memdup_user(argp, sizeof(*args));
1927 	if (IS_ERR(args))
1928 		return PTR_ERR(args);
1929 
1930 	inode = file_inode(file);
1931 	ret = search_ioctl(inode, args);
1932 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1933 		ret = -EFAULT;
1934 	kfree(args);
1935 	return ret;
1936 }
1937 
1938 /*
1939  * Search INODE_REFs to identify path name of 'dirid' directory
1940  * in a 'tree_id' tree. and sets path name to 'name'.
1941  */
1942 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1943 				u64 tree_id, u64 dirid, char *name)
1944 {
1945 	struct btrfs_root *root;
1946 	struct btrfs_key key;
1947 	char *ptr;
1948 	int ret = -1;
1949 	int slot;
1950 	int len;
1951 	int total_len = 0;
1952 	struct btrfs_inode_ref *iref;
1953 	struct extent_buffer *l;
1954 	struct btrfs_path *path;
1955 
1956 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1957 		name[0]='\0';
1958 		return 0;
1959 	}
1960 
1961 	path = btrfs_alloc_path();
1962 	if (!path)
1963 		return -ENOMEM;
1964 
1965 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1966 
1967 	key.objectid = tree_id;
1968 	key.type = BTRFS_ROOT_ITEM_KEY;
1969 	key.offset = (u64)-1;
1970 	root = btrfs_read_fs_root_no_name(info, &key);
1971 	if (IS_ERR(root)) {
1972 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1973 		ret = -ENOENT;
1974 		goto out;
1975 	}
1976 
1977 	key.objectid = dirid;
1978 	key.type = BTRFS_INODE_REF_KEY;
1979 	key.offset = (u64)-1;
1980 
1981 	while(1) {
1982 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1983 		if (ret < 0)
1984 			goto out;
1985 
1986 		l = path->nodes[0];
1987 		slot = path->slots[0];
1988 		if (ret > 0 && slot > 0)
1989 			slot--;
1990 		btrfs_item_key_to_cpu(l, &key, slot);
1991 
1992 		if (ret > 0 && (key.objectid != dirid ||
1993 				key.type != BTRFS_INODE_REF_KEY)) {
1994 			ret = -ENOENT;
1995 			goto out;
1996 		}
1997 
1998 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1999 		len = btrfs_inode_ref_name_len(l, iref);
2000 		ptr -= len + 1;
2001 		total_len += len + 1;
2002 		if (ptr < name)
2003 			goto out;
2004 
2005 		*(ptr + len) = '/';
2006 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2007 
2008 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2009 			break;
2010 
2011 		btrfs_release_path(path);
2012 		key.objectid = key.offset;
2013 		key.offset = (u64)-1;
2014 		dirid = key.objectid;
2015 	}
2016 	if (ptr < name)
2017 		goto out;
2018 	memmove(name, ptr, total_len);
2019 	name[total_len]='\0';
2020 	ret = 0;
2021 out:
2022 	btrfs_free_path(path);
2023 	return ret;
2024 }
2025 
2026 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2027 					   void __user *argp)
2028 {
2029 	 struct btrfs_ioctl_ino_lookup_args *args;
2030 	 struct inode *inode;
2031 	 int ret;
2032 
2033 	if (!capable(CAP_SYS_ADMIN))
2034 		return -EPERM;
2035 
2036 	args = memdup_user(argp, sizeof(*args));
2037 	if (IS_ERR(args))
2038 		return PTR_ERR(args);
2039 
2040 	inode = file_inode(file);
2041 
2042 	if (args->treeid == 0)
2043 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2044 
2045 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2046 					args->treeid, args->objectid,
2047 					args->name);
2048 
2049 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2050 		ret = -EFAULT;
2051 
2052 	kfree(args);
2053 	return ret;
2054 }
2055 
2056 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2057 					     void __user *arg)
2058 {
2059 	struct dentry *parent = fdentry(file);
2060 	struct dentry *dentry;
2061 	struct inode *dir = parent->d_inode;
2062 	struct inode *inode;
2063 	struct btrfs_root *root = BTRFS_I(dir)->root;
2064 	struct btrfs_root *dest = NULL;
2065 	struct btrfs_ioctl_vol_args *vol_args;
2066 	struct btrfs_trans_handle *trans;
2067 	struct btrfs_block_rsv block_rsv;
2068 	u64 qgroup_reserved;
2069 	int namelen;
2070 	int ret;
2071 	int err = 0;
2072 
2073 	vol_args = memdup_user(arg, sizeof(*vol_args));
2074 	if (IS_ERR(vol_args))
2075 		return PTR_ERR(vol_args);
2076 
2077 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2078 	namelen = strlen(vol_args->name);
2079 	if (strchr(vol_args->name, '/') ||
2080 	    strncmp(vol_args->name, "..", namelen) == 0) {
2081 		err = -EINVAL;
2082 		goto out;
2083 	}
2084 
2085 	err = mnt_want_write_file(file);
2086 	if (err)
2087 		goto out;
2088 
2089 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2090 	dentry = lookup_one_len(vol_args->name, parent, namelen);
2091 	if (IS_ERR(dentry)) {
2092 		err = PTR_ERR(dentry);
2093 		goto out_unlock_dir;
2094 	}
2095 
2096 	if (!dentry->d_inode) {
2097 		err = -ENOENT;
2098 		goto out_dput;
2099 	}
2100 
2101 	inode = dentry->d_inode;
2102 	dest = BTRFS_I(inode)->root;
2103 	if (!capable(CAP_SYS_ADMIN)){
2104 		/*
2105 		 * Regular user.  Only allow this with a special mount
2106 		 * option, when the user has write+exec access to the
2107 		 * subvol root, and when rmdir(2) would have been
2108 		 * allowed.
2109 		 *
2110 		 * Note that this is _not_ check that the subvol is
2111 		 * empty or doesn't contain data that we wouldn't
2112 		 * otherwise be able to delete.
2113 		 *
2114 		 * Users who want to delete empty subvols should try
2115 		 * rmdir(2).
2116 		 */
2117 		err = -EPERM;
2118 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2119 			goto out_dput;
2120 
2121 		/*
2122 		 * Do not allow deletion if the parent dir is the same
2123 		 * as the dir to be deleted.  That means the ioctl
2124 		 * must be called on the dentry referencing the root
2125 		 * of the subvol, not a random directory contained
2126 		 * within it.
2127 		 */
2128 		err = -EINVAL;
2129 		if (root == dest)
2130 			goto out_dput;
2131 
2132 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2133 		if (err)
2134 			goto out_dput;
2135 	}
2136 
2137 	/* check if subvolume may be deleted by a user */
2138 	err = btrfs_may_delete(dir, dentry, 1);
2139 	if (err)
2140 		goto out_dput;
2141 
2142 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2143 		err = -EINVAL;
2144 		goto out_dput;
2145 	}
2146 
2147 	mutex_lock(&inode->i_mutex);
2148 	err = d_invalidate(dentry);
2149 	if (err)
2150 		goto out_unlock;
2151 
2152 	down_write(&root->fs_info->subvol_sem);
2153 
2154 	err = may_destroy_subvol(dest);
2155 	if (err)
2156 		goto out_up_write;
2157 
2158 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2159 	/*
2160 	 * One for dir inode, two for dir entries, two for root
2161 	 * ref/backref.
2162 	 */
2163 	err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2164 					       5, &qgroup_reserved);
2165 	if (err)
2166 		goto out_up_write;
2167 
2168 	trans = btrfs_start_transaction(root, 0);
2169 	if (IS_ERR(trans)) {
2170 		err = PTR_ERR(trans);
2171 		goto out_release;
2172 	}
2173 	trans->block_rsv = &block_rsv;
2174 	trans->bytes_reserved = block_rsv.size;
2175 
2176 	ret = btrfs_unlink_subvol(trans, root, dir,
2177 				dest->root_key.objectid,
2178 				dentry->d_name.name,
2179 				dentry->d_name.len);
2180 	if (ret) {
2181 		err = ret;
2182 		btrfs_abort_transaction(trans, root, ret);
2183 		goto out_end_trans;
2184 	}
2185 
2186 	btrfs_record_root_in_trans(trans, dest);
2187 
2188 	memset(&dest->root_item.drop_progress, 0,
2189 		sizeof(dest->root_item.drop_progress));
2190 	dest->root_item.drop_level = 0;
2191 	btrfs_set_root_refs(&dest->root_item, 0);
2192 
2193 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2194 		ret = btrfs_insert_orphan_item(trans,
2195 					root->fs_info->tree_root,
2196 					dest->root_key.objectid);
2197 		if (ret) {
2198 			btrfs_abort_transaction(trans, root, ret);
2199 			err = ret;
2200 			goto out_end_trans;
2201 		}
2202 	}
2203 out_end_trans:
2204 	trans->block_rsv = NULL;
2205 	trans->bytes_reserved = 0;
2206 	ret = btrfs_end_transaction(trans, root);
2207 	if (ret && !err)
2208 		err = ret;
2209 	inode->i_flags |= S_DEAD;
2210 out_release:
2211 	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2212 out_up_write:
2213 	up_write(&root->fs_info->subvol_sem);
2214 out_unlock:
2215 	mutex_unlock(&inode->i_mutex);
2216 	if (!err) {
2217 		shrink_dcache_sb(root->fs_info->sb);
2218 		btrfs_invalidate_inodes(dest);
2219 		d_delete(dentry);
2220 
2221 		/* the last ref */
2222 		if (dest->cache_inode) {
2223 			iput(dest->cache_inode);
2224 			dest->cache_inode = NULL;
2225 		}
2226 	}
2227 out_dput:
2228 	dput(dentry);
2229 out_unlock_dir:
2230 	mutex_unlock(&dir->i_mutex);
2231 	mnt_drop_write_file(file);
2232 out:
2233 	kfree(vol_args);
2234 	return err;
2235 }
2236 
2237 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2238 {
2239 	struct inode *inode = file_inode(file);
2240 	struct btrfs_root *root = BTRFS_I(inode)->root;
2241 	struct btrfs_ioctl_defrag_range_args *range;
2242 	int ret;
2243 
2244 	ret = mnt_want_write_file(file);
2245 	if (ret)
2246 		return ret;
2247 
2248 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2249 			1)) {
2250 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2251 		mnt_drop_write_file(file);
2252 		return -EINVAL;
2253 	}
2254 
2255 	if (btrfs_root_readonly(root)) {
2256 		ret = -EROFS;
2257 		goto out;
2258 	}
2259 
2260 	switch (inode->i_mode & S_IFMT) {
2261 	case S_IFDIR:
2262 		if (!capable(CAP_SYS_ADMIN)) {
2263 			ret = -EPERM;
2264 			goto out;
2265 		}
2266 		ret = btrfs_defrag_root(root);
2267 		if (ret)
2268 			goto out;
2269 		ret = btrfs_defrag_root(root->fs_info->extent_root);
2270 		break;
2271 	case S_IFREG:
2272 		if (!(file->f_mode & FMODE_WRITE)) {
2273 			ret = -EINVAL;
2274 			goto out;
2275 		}
2276 
2277 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2278 		if (!range) {
2279 			ret = -ENOMEM;
2280 			goto out;
2281 		}
2282 
2283 		if (argp) {
2284 			if (copy_from_user(range, argp,
2285 					   sizeof(*range))) {
2286 				ret = -EFAULT;
2287 				kfree(range);
2288 				goto out;
2289 			}
2290 			/* compression requires us to start the IO */
2291 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2292 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2293 				range->extent_thresh = (u32)-1;
2294 			}
2295 		} else {
2296 			/* the rest are all set to zero by kzalloc */
2297 			range->len = (u64)-1;
2298 		}
2299 		ret = btrfs_defrag_file(file_inode(file), file,
2300 					range, 0, 0);
2301 		if (ret > 0)
2302 			ret = 0;
2303 		kfree(range);
2304 		break;
2305 	default:
2306 		ret = -EINVAL;
2307 	}
2308 out:
2309 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2310 	mnt_drop_write_file(file);
2311 	return ret;
2312 }
2313 
2314 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2315 {
2316 	struct btrfs_ioctl_vol_args *vol_args;
2317 	int ret;
2318 
2319 	if (!capable(CAP_SYS_ADMIN))
2320 		return -EPERM;
2321 
2322 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2323 			1)) {
2324 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2325 		return -EINVAL;
2326 	}
2327 
2328 	mutex_lock(&root->fs_info->volume_mutex);
2329 	vol_args = memdup_user(arg, sizeof(*vol_args));
2330 	if (IS_ERR(vol_args)) {
2331 		ret = PTR_ERR(vol_args);
2332 		goto out;
2333 	}
2334 
2335 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2336 	ret = btrfs_init_new_device(root, vol_args->name);
2337 
2338 	kfree(vol_args);
2339 out:
2340 	mutex_unlock(&root->fs_info->volume_mutex);
2341 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2342 	return ret;
2343 }
2344 
2345 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2346 {
2347 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2348 	struct btrfs_ioctl_vol_args *vol_args;
2349 	int ret;
2350 
2351 	if (!capable(CAP_SYS_ADMIN))
2352 		return -EPERM;
2353 
2354 	ret = mnt_want_write_file(file);
2355 	if (ret)
2356 		return ret;
2357 
2358 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2359 			1)) {
2360 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2361 		mnt_drop_write_file(file);
2362 		return -EINVAL;
2363 	}
2364 
2365 	mutex_lock(&root->fs_info->volume_mutex);
2366 	vol_args = memdup_user(arg, sizeof(*vol_args));
2367 	if (IS_ERR(vol_args)) {
2368 		ret = PTR_ERR(vol_args);
2369 		goto out;
2370 	}
2371 
2372 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2373 	ret = btrfs_rm_device(root, vol_args->name);
2374 
2375 	kfree(vol_args);
2376 out:
2377 	mutex_unlock(&root->fs_info->volume_mutex);
2378 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2379 	mnt_drop_write_file(file);
2380 	return ret;
2381 }
2382 
2383 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2384 {
2385 	struct btrfs_ioctl_fs_info_args *fi_args;
2386 	struct btrfs_device *device;
2387 	struct btrfs_device *next;
2388 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2389 	int ret = 0;
2390 
2391 	if (!capable(CAP_SYS_ADMIN))
2392 		return -EPERM;
2393 
2394 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2395 	if (!fi_args)
2396 		return -ENOMEM;
2397 
2398 	fi_args->num_devices = fs_devices->num_devices;
2399 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2400 
2401 	mutex_lock(&fs_devices->device_list_mutex);
2402 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2403 		if (device->devid > fi_args->max_id)
2404 			fi_args->max_id = device->devid;
2405 	}
2406 	mutex_unlock(&fs_devices->device_list_mutex);
2407 
2408 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2409 		ret = -EFAULT;
2410 
2411 	kfree(fi_args);
2412 	return ret;
2413 }
2414 
2415 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2416 {
2417 	struct btrfs_ioctl_dev_info_args *di_args;
2418 	struct btrfs_device *dev;
2419 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2420 	int ret = 0;
2421 	char *s_uuid = NULL;
2422 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2423 
2424 	if (!capable(CAP_SYS_ADMIN))
2425 		return -EPERM;
2426 
2427 	di_args = memdup_user(arg, sizeof(*di_args));
2428 	if (IS_ERR(di_args))
2429 		return PTR_ERR(di_args);
2430 
2431 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2432 		s_uuid = di_args->uuid;
2433 
2434 	mutex_lock(&fs_devices->device_list_mutex);
2435 	dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2436 	mutex_unlock(&fs_devices->device_list_mutex);
2437 
2438 	if (!dev) {
2439 		ret = -ENODEV;
2440 		goto out;
2441 	}
2442 
2443 	di_args->devid = dev->devid;
2444 	di_args->bytes_used = dev->bytes_used;
2445 	di_args->total_bytes = dev->total_bytes;
2446 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2447 	if (dev->name) {
2448 		struct rcu_string *name;
2449 
2450 		rcu_read_lock();
2451 		name = rcu_dereference(dev->name);
2452 		strncpy(di_args->path, name->str, sizeof(di_args->path));
2453 		rcu_read_unlock();
2454 		di_args->path[sizeof(di_args->path) - 1] = 0;
2455 	} else {
2456 		di_args->path[0] = '\0';
2457 	}
2458 
2459 out:
2460 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2461 		ret = -EFAULT;
2462 
2463 	kfree(di_args);
2464 	return ret;
2465 }
2466 
2467 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2468 				       u64 off, u64 olen, u64 destoff)
2469 {
2470 	struct inode *inode = file_inode(file);
2471 	struct btrfs_root *root = BTRFS_I(inode)->root;
2472 	struct fd src_file;
2473 	struct inode *src;
2474 	struct btrfs_trans_handle *trans;
2475 	struct btrfs_path *path;
2476 	struct extent_buffer *leaf;
2477 	char *buf;
2478 	struct btrfs_key key;
2479 	u32 nritems;
2480 	int slot;
2481 	int ret;
2482 	u64 len = olen;
2483 	u64 bs = root->fs_info->sb->s_blocksize;
2484 
2485 	/*
2486 	 * TODO:
2487 	 * - split compressed inline extents.  annoying: we need to
2488 	 *   decompress into destination's address_space (the file offset
2489 	 *   may change, so source mapping won't do), then recompress (or
2490 	 *   otherwise reinsert) a subrange.
2491 	 * - allow ranges within the same file to be cloned (provided
2492 	 *   they don't overlap)?
2493 	 */
2494 
2495 	/* the destination must be opened for writing */
2496 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2497 		return -EINVAL;
2498 
2499 	if (btrfs_root_readonly(root))
2500 		return -EROFS;
2501 
2502 	ret = mnt_want_write_file(file);
2503 	if (ret)
2504 		return ret;
2505 
2506 	src_file = fdget(srcfd);
2507 	if (!src_file.file) {
2508 		ret = -EBADF;
2509 		goto out_drop_write;
2510 	}
2511 
2512 	ret = -EXDEV;
2513 	if (src_file.file->f_path.mnt != file->f_path.mnt)
2514 		goto out_fput;
2515 
2516 	src = file_inode(src_file.file);
2517 
2518 	ret = -EINVAL;
2519 	if (src == inode)
2520 		goto out_fput;
2521 
2522 	/* the src must be open for reading */
2523 	if (!(src_file.file->f_mode & FMODE_READ))
2524 		goto out_fput;
2525 
2526 	/* don't make the dst file partly checksummed */
2527 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2528 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2529 		goto out_fput;
2530 
2531 	ret = -EISDIR;
2532 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2533 		goto out_fput;
2534 
2535 	ret = -EXDEV;
2536 	if (src->i_sb != inode->i_sb)
2537 		goto out_fput;
2538 
2539 	ret = -ENOMEM;
2540 	buf = vmalloc(btrfs_level_size(root, 0));
2541 	if (!buf)
2542 		goto out_fput;
2543 
2544 	path = btrfs_alloc_path();
2545 	if (!path) {
2546 		vfree(buf);
2547 		goto out_fput;
2548 	}
2549 	path->reada = 2;
2550 
2551 	if (inode < src) {
2552 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2553 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2554 	} else {
2555 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2556 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2557 	}
2558 
2559 	/* determine range to clone */
2560 	ret = -EINVAL;
2561 	if (off + len > src->i_size || off + len < off)
2562 		goto out_unlock;
2563 	if (len == 0)
2564 		olen = len = src->i_size - off;
2565 	/* if we extend to eof, continue to block boundary */
2566 	if (off + len == src->i_size)
2567 		len = ALIGN(src->i_size, bs) - off;
2568 
2569 	/* verify the end result is block aligned */
2570 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2571 	    !IS_ALIGNED(destoff, bs))
2572 		goto out_unlock;
2573 
2574 	if (destoff > inode->i_size) {
2575 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2576 		if (ret)
2577 			goto out_unlock;
2578 	}
2579 
2580 	/* truncate page cache pages from target inode range */
2581 	truncate_inode_pages_range(&inode->i_data, destoff,
2582 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2583 
2584 	/* do any pending delalloc/csum calc on src, one way or
2585 	   another, and lock file content */
2586 	while (1) {
2587 		struct btrfs_ordered_extent *ordered;
2588 		lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2589 		ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2590 		if (!ordered &&
2591 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2592 				    EXTENT_DELALLOC, 0, NULL))
2593 			break;
2594 		unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2595 		if (ordered)
2596 			btrfs_put_ordered_extent(ordered);
2597 		btrfs_wait_ordered_range(src, off, len);
2598 	}
2599 
2600 	/* clone data */
2601 	key.objectid = btrfs_ino(src);
2602 	key.type = BTRFS_EXTENT_DATA_KEY;
2603 	key.offset = 0;
2604 
2605 	while (1) {
2606 		/*
2607 		 * note the key will change type as we walk through the
2608 		 * tree.
2609 		 */
2610 		ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2611 				0, 0);
2612 		if (ret < 0)
2613 			goto out;
2614 
2615 		nritems = btrfs_header_nritems(path->nodes[0]);
2616 		if (path->slots[0] >= nritems) {
2617 			ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2618 			if (ret < 0)
2619 				goto out;
2620 			if (ret > 0)
2621 				break;
2622 			nritems = btrfs_header_nritems(path->nodes[0]);
2623 		}
2624 		leaf = path->nodes[0];
2625 		slot = path->slots[0];
2626 
2627 		btrfs_item_key_to_cpu(leaf, &key, slot);
2628 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2629 		    key.objectid != btrfs_ino(src))
2630 			break;
2631 
2632 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2633 			struct btrfs_file_extent_item *extent;
2634 			int type;
2635 			u32 size;
2636 			struct btrfs_key new_key;
2637 			u64 disko = 0, diskl = 0;
2638 			u64 datao = 0, datal = 0;
2639 			u8 comp;
2640 			u64 endoff;
2641 
2642 			size = btrfs_item_size_nr(leaf, slot);
2643 			read_extent_buffer(leaf, buf,
2644 					   btrfs_item_ptr_offset(leaf, slot),
2645 					   size);
2646 
2647 			extent = btrfs_item_ptr(leaf, slot,
2648 						struct btrfs_file_extent_item);
2649 			comp = btrfs_file_extent_compression(leaf, extent);
2650 			type = btrfs_file_extent_type(leaf, extent);
2651 			if (type == BTRFS_FILE_EXTENT_REG ||
2652 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2653 				disko = btrfs_file_extent_disk_bytenr(leaf,
2654 								      extent);
2655 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2656 								 extent);
2657 				datao = btrfs_file_extent_offset(leaf, extent);
2658 				datal = btrfs_file_extent_num_bytes(leaf,
2659 								    extent);
2660 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2661 				/* take upper bound, may be compressed */
2662 				datal = btrfs_file_extent_ram_bytes(leaf,
2663 								    extent);
2664 			}
2665 			btrfs_release_path(path);
2666 
2667 			if (key.offset + datal <= off ||
2668 			    key.offset >= off + len - 1)
2669 				goto next;
2670 
2671 			memcpy(&new_key, &key, sizeof(new_key));
2672 			new_key.objectid = btrfs_ino(inode);
2673 			if (off <= key.offset)
2674 				new_key.offset = key.offset + destoff - off;
2675 			else
2676 				new_key.offset = destoff;
2677 
2678 			/*
2679 			 * 1 - adjusting old extent (we may have to split it)
2680 			 * 1 - add new extent
2681 			 * 1 - inode update
2682 			 */
2683 			trans = btrfs_start_transaction(root, 3);
2684 			if (IS_ERR(trans)) {
2685 				ret = PTR_ERR(trans);
2686 				goto out;
2687 			}
2688 
2689 			if (type == BTRFS_FILE_EXTENT_REG ||
2690 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2691 				/*
2692 				 *    a  | --- range to clone ---|  b
2693 				 * | ------------- extent ------------- |
2694 				 */
2695 
2696 				/* substract range b */
2697 				if (key.offset + datal > off + len)
2698 					datal = off + len - key.offset;
2699 
2700 				/* substract range a */
2701 				if (off > key.offset) {
2702 					datao += off - key.offset;
2703 					datal -= off - key.offset;
2704 				}
2705 
2706 				ret = btrfs_drop_extents(trans, root, inode,
2707 							 new_key.offset,
2708 							 new_key.offset + datal,
2709 							 1);
2710 				if (ret) {
2711 					btrfs_abort_transaction(trans, root,
2712 								ret);
2713 					btrfs_end_transaction(trans, root);
2714 					goto out;
2715 				}
2716 
2717 				ret = btrfs_insert_empty_item(trans, root, path,
2718 							      &new_key, size);
2719 				if (ret) {
2720 					btrfs_abort_transaction(trans, root,
2721 								ret);
2722 					btrfs_end_transaction(trans, root);
2723 					goto out;
2724 				}
2725 
2726 				leaf = path->nodes[0];
2727 				slot = path->slots[0];
2728 				write_extent_buffer(leaf, buf,
2729 					    btrfs_item_ptr_offset(leaf, slot),
2730 					    size);
2731 
2732 				extent = btrfs_item_ptr(leaf, slot,
2733 						struct btrfs_file_extent_item);
2734 
2735 				/* disko == 0 means it's a hole */
2736 				if (!disko)
2737 					datao = 0;
2738 
2739 				btrfs_set_file_extent_offset(leaf, extent,
2740 							     datao);
2741 				btrfs_set_file_extent_num_bytes(leaf, extent,
2742 								datal);
2743 				if (disko) {
2744 					inode_add_bytes(inode, datal);
2745 					ret = btrfs_inc_extent_ref(trans, root,
2746 							disko, diskl, 0,
2747 							root->root_key.objectid,
2748 							btrfs_ino(inode),
2749 							new_key.offset - datao,
2750 							0);
2751 					if (ret) {
2752 						btrfs_abort_transaction(trans,
2753 									root,
2754 									ret);
2755 						btrfs_end_transaction(trans,
2756 								      root);
2757 						goto out;
2758 
2759 					}
2760 				}
2761 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2762 				u64 skip = 0;
2763 				u64 trim = 0;
2764 				if (off > key.offset) {
2765 					skip = off - key.offset;
2766 					new_key.offset += skip;
2767 				}
2768 
2769 				if (key.offset + datal > off + len)
2770 					trim = key.offset + datal - (off + len);
2771 
2772 				if (comp && (skip || trim)) {
2773 					ret = -EINVAL;
2774 					btrfs_end_transaction(trans, root);
2775 					goto out;
2776 				}
2777 				size -= skip + trim;
2778 				datal -= skip + trim;
2779 
2780 				ret = btrfs_drop_extents(trans, root, inode,
2781 							 new_key.offset,
2782 							 new_key.offset + datal,
2783 							 1);
2784 				if (ret) {
2785 					btrfs_abort_transaction(trans, root,
2786 								ret);
2787 					btrfs_end_transaction(trans, root);
2788 					goto out;
2789 				}
2790 
2791 				ret = btrfs_insert_empty_item(trans, root, path,
2792 							      &new_key, size);
2793 				if (ret) {
2794 					btrfs_abort_transaction(trans, root,
2795 								ret);
2796 					btrfs_end_transaction(trans, root);
2797 					goto out;
2798 				}
2799 
2800 				if (skip) {
2801 					u32 start =
2802 					  btrfs_file_extent_calc_inline_size(0);
2803 					memmove(buf+start, buf+start+skip,
2804 						datal);
2805 				}
2806 
2807 				leaf = path->nodes[0];
2808 				slot = path->slots[0];
2809 				write_extent_buffer(leaf, buf,
2810 					    btrfs_item_ptr_offset(leaf, slot),
2811 					    size);
2812 				inode_add_bytes(inode, datal);
2813 			}
2814 
2815 			btrfs_mark_buffer_dirty(leaf);
2816 			btrfs_release_path(path);
2817 
2818 			inode_inc_iversion(inode);
2819 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2820 
2821 			/*
2822 			 * we round up to the block size at eof when
2823 			 * determining which extents to clone above,
2824 			 * but shouldn't round up the file size
2825 			 */
2826 			endoff = new_key.offset + datal;
2827 			if (endoff > destoff+olen)
2828 				endoff = destoff+olen;
2829 			if (endoff > inode->i_size)
2830 				btrfs_i_size_write(inode, endoff);
2831 
2832 			ret = btrfs_update_inode(trans, root, inode);
2833 			if (ret) {
2834 				btrfs_abort_transaction(trans, root, ret);
2835 				btrfs_end_transaction(trans, root);
2836 				goto out;
2837 			}
2838 			ret = btrfs_end_transaction(trans, root);
2839 		}
2840 next:
2841 		btrfs_release_path(path);
2842 		key.offset++;
2843 	}
2844 	ret = 0;
2845 out:
2846 	btrfs_release_path(path);
2847 	unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2848 out_unlock:
2849 	mutex_unlock(&src->i_mutex);
2850 	mutex_unlock(&inode->i_mutex);
2851 	vfree(buf);
2852 	btrfs_free_path(path);
2853 out_fput:
2854 	fdput(src_file);
2855 out_drop_write:
2856 	mnt_drop_write_file(file);
2857 	return ret;
2858 }
2859 
2860 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2861 {
2862 	struct btrfs_ioctl_clone_range_args args;
2863 
2864 	if (copy_from_user(&args, argp, sizeof(args)))
2865 		return -EFAULT;
2866 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2867 				 args.src_length, args.dest_offset);
2868 }
2869 
2870 /*
2871  * there are many ways the trans_start and trans_end ioctls can lead
2872  * to deadlocks.  They should only be used by applications that
2873  * basically own the machine, and have a very in depth understanding
2874  * of all the possible deadlocks and enospc problems.
2875  */
2876 static long btrfs_ioctl_trans_start(struct file *file)
2877 {
2878 	struct inode *inode = file_inode(file);
2879 	struct btrfs_root *root = BTRFS_I(inode)->root;
2880 	struct btrfs_trans_handle *trans;
2881 	int ret;
2882 
2883 	ret = -EPERM;
2884 	if (!capable(CAP_SYS_ADMIN))
2885 		goto out;
2886 
2887 	ret = -EINPROGRESS;
2888 	if (file->private_data)
2889 		goto out;
2890 
2891 	ret = -EROFS;
2892 	if (btrfs_root_readonly(root))
2893 		goto out;
2894 
2895 	ret = mnt_want_write_file(file);
2896 	if (ret)
2897 		goto out;
2898 
2899 	atomic_inc(&root->fs_info->open_ioctl_trans);
2900 
2901 	ret = -ENOMEM;
2902 	trans = btrfs_start_ioctl_transaction(root);
2903 	if (IS_ERR(trans))
2904 		goto out_drop;
2905 
2906 	file->private_data = trans;
2907 	return 0;
2908 
2909 out_drop:
2910 	atomic_dec(&root->fs_info->open_ioctl_trans);
2911 	mnt_drop_write_file(file);
2912 out:
2913 	return ret;
2914 }
2915 
2916 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2917 {
2918 	struct inode *inode = file_inode(file);
2919 	struct btrfs_root *root = BTRFS_I(inode)->root;
2920 	struct btrfs_root *new_root;
2921 	struct btrfs_dir_item *di;
2922 	struct btrfs_trans_handle *trans;
2923 	struct btrfs_path *path;
2924 	struct btrfs_key location;
2925 	struct btrfs_disk_key disk_key;
2926 	u64 objectid = 0;
2927 	u64 dir_id;
2928 	int ret;
2929 
2930 	if (!capable(CAP_SYS_ADMIN))
2931 		return -EPERM;
2932 
2933 	ret = mnt_want_write_file(file);
2934 	if (ret)
2935 		return ret;
2936 
2937 	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2938 		ret = -EFAULT;
2939 		goto out;
2940 	}
2941 
2942 	if (!objectid)
2943 		objectid = root->root_key.objectid;
2944 
2945 	location.objectid = objectid;
2946 	location.type = BTRFS_ROOT_ITEM_KEY;
2947 	location.offset = (u64)-1;
2948 
2949 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2950 	if (IS_ERR(new_root)) {
2951 		ret = PTR_ERR(new_root);
2952 		goto out;
2953 	}
2954 
2955 	if (btrfs_root_refs(&new_root->root_item) == 0) {
2956 		ret = -ENOENT;
2957 		goto out;
2958 	}
2959 
2960 	path = btrfs_alloc_path();
2961 	if (!path) {
2962 		ret = -ENOMEM;
2963 		goto out;
2964 	}
2965 	path->leave_spinning = 1;
2966 
2967 	trans = btrfs_start_transaction(root, 1);
2968 	if (IS_ERR(trans)) {
2969 		btrfs_free_path(path);
2970 		ret = PTR_ERR(trans);
2971 		goto out;
2972 	}
2973 
2974 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2975 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2976 				   dir_id, "default", 7, 1);
2977 	if (IS_ERR_OR_NULL(di)) {
2978 		btrfs_free_path(path);
2979 		btrfs_end_transaction(trans, root);
2980 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2981 		       "this isn't going to work\n");
2982 		ret = -ENOENT;
2983 		goto out;
2984 	}
2985 
2986 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2987 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2988 	btrfs_mark_buffer_dirty(path->nodes[0]);
2989 	btrfs_free_path(path);
2990 
2991 	btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2992 	btrfs_end_transaction(trans, root);
2993 out:
2994 	mnt_drop_write_file(file);
2995 	return ret;
2996 }
2997 
2998 void btrfs_get_block_group_info(struct list_head *groups_list,
2999 				struct btrfs_ioctl_space_info *space)
3000 {
3001 	struct btrfs_block_group_cache *block_group;
3002 
3003 	space->total_bytes = 0;
3004 	space->used_bytes = 0;
3005 	space->flags = 0;
3006 	list_for_each_entry(block_group, groups_list, list) {
3007 		space->flags = block_group->flags;
3008 		space->total_bytes += block_group->key.offset;
3009 		space->used_bytes +=
3010 			btrfs_block_group_used(&block_group->item);
3011 	}
3012 }
3013 
3014 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3015 {
3016 	struct btrfs_ioctl_space_args space_args;
3017 	struct btrfs_ioctl_space_info space;
3018 	struct btrfs_ioctl_space_info *dest;
3019 	struct btrfs_ioctl_space_info *dest_orig;
3020 	struct btrfs_ioctl_space_info __user *user_dest;
3021 	struct btrfs_space_info *info;
3022 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3023 		       BTRFS_BLOCK_GROUP_SYSTEM,
3024 		       BTRFS_BLOCK_GROUP_METADATA,
3025 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3026 	int num_types = 4;
3027 	int alloc_size;
3028 	int ret = 0;
3029 	u64 slot_count = 0;
3030 	int i, c;
3031 
3032 	if (copy_from_user(&space_args,
3033 			   (struct btrfs_ioctl_space_args __user *)arg,
3034 			   sizeof(space_args)))
3035 		return -EFAULT;
3036 
3037 	for (i = 0; i < num_types; i++) {
3038 		struct btrfs_space_info *tmp;
3039 
3040 		info = NULL;
3041 		rcu_read_lock();
3042 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3043 					list) {
3044 			if (tmp->flags == types[i]) {
3045 				info = tmp;
3046 				break;
3047 			}
3048 		}
3049 		rcu_read_unlock();
3050 
3051 		if (!info)
3052 			continue;
3053 
3054 		down_read(&info->groups_sem);
3055 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3056 			if (!list_empty(&info->block_groups[c]))
3057 				slot_count++;
3058 		}
3059 		up_read(&info->groups_sem);
3060 	}
3061 
3062 	/* space_slots == 0 means they are asking for a count */
3063 	if (space_args.space_slots == 0) {
3064 		space_args.total_spaces = slot_count;
3065 		goto out;
3066 	}
3067 
3068 	slot_count = min_t(u64, space_args.space_slots, slot_count);
3069 
3070 	alloc_size = sizeof(*dest) * slot_count;
3071 
3072 	/* we generally have at most 6 or so space infos, one for each raid
3073 	 * level.  So, a whole page should be more than enough for everyone
3074 	 */
3075 	if (alloc_size > PAGE_CACHE_SIZE)
3076 		return -ENOMEM;
3077 
3078 	space_args.total_spaces = 0;
3079 	dest = kmalloc(alloc_size, GFP_NOFS);
3080 	if (!dest)
3081 		return -ENOMEM;
3082 	dest_orig = dest;
3083 
3084 	/* now we have a buffer to copy into */
3085 	for (i = 0; i < num_types; i++) {
3086 		struct btrfs_space_info *tmp;
3087 
3088 		if (!slot_count)
3089 			break;
3090 
3091 		info = NULL;
3092 		rcu_read_lock();
3093 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3094 					list) {
3095 			if (tmp->flags == types[i]) {
3096 				info = tmp;
3097 				break;
3098 			}
3099 		}
3100 		rcu_read_unlock();
3101 
3102 		if (!info)
3103 			continue;
3104 		down_read(&info->groups_sem);
3105 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3106 			if (!list_empty(&info->block_groups[c])) {
3107 				btrfs_get_block_group_info(
3108 					&info->block_groups[c], &space);
3109 				memcpy(dest, &space, sizeof(space));
3110 				dest++;
3111 				space_args.total_spaces++;
3112 				slot_count--;
3113 			}
3114 			if (!slot_count)
3115 				break;
3116 		}
3117 		up_read(&info->groups_sem);
3118 	}
3119 
3120 	user_dest = (struct btrfs_ioctl_space_info __user *)
3121 		(arg + sizeof(struct btrfs_ioctl_space_args));
3122 
3123 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3124 		ret = -EFAULT;
3125 
3126 	kfree(dest_orig);
3127 out:
3128 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3129 		ret = -EFAULT;
3130 
3131 	return ret;
3132 }
3133 
3134 /*
3135  * there are many ways the trans_start and trans_end ioctls can lead
3136  * to deadlocks.  They should only be used by applications that
3137  * basically own the machine, and have a very in depth understanding
3138  * of all the possible deadlocks and enospc problems.
3139  */
3140 long btrfs_ioctl_trans_end(struct file *file)
3141 {
3142 	struct inode *inode = file_inode(file);
3143 	struct btrfs_root *root = BTRFS_I(inode)->root;
3144 	struct btrfs_trans_handle *trans;
3145 
3146 	trans = file->private_data;
3147 	if (!trans)
3148 		return -EINVAL;
3149 	file->private_data = NULL;
3150 
3151 	btrfs_end_transaction(trans, root);
3152 
3153 	atomic_dec(&root->fs_info->open_ioctl_trans);
3154 
3155 	mnt_drop_write_file(file);
3156 	return 0;
3157 }
3158 
3159 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3160 					    void __user *argp)
3161 {
3162 	struct btrfs_trans_handle *trans;
3163 	u64 transid;
3164 	int ret;
3165 
3166 	trans = btrfs_attach_transaction_barrier(root);
3167 	if (IS_ERR(trans)) {
3168 		if (PTR_ERR(trans) != -ENOENT)
3169 			return PTR_ERR(trans);
3170 
3171 		/* No running transaction, don't bother */
3172 		transid = root->fs_info->last_trans_committed;
3173 		goto out;
3174 	}
3175 	transid = trans->transid;
3176 	ret = btrfs_commit_transaction_async(trans, root, 0);
3177 	if (ret) {
3178 		btrfs_end_transaction(trans, root);
3179 		return ret;
3180 	}
3181 out:
3182 	if (argp)
3183 		if (copy_to_user(argp, &transid, sizeof(transid)))
3184 			return -EFAULT;
3185 	return 0;
3186 }
3187 
3188 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3189 					   void __user *argp)
3190 {
3191 	u64 transid;
3192 
3193 	if (argp) {
3194 		if (copy_from_user(&transid, argp, sizeof(transid)))
3195 			return -EFAULT;
3196 	} else {
3197 		transid = 0;  /* current trans */
3198 	}
3199 	return btrfs_wait_for_commit(root, transid);
3200 }
3201 
3202 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3203 {
3204 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3205 	struct btrfs_ioctl_scrub_args *sa;
3206 	int ret;
3207 
3208 	if (!capable(CAP_SYS_ADMIN))
3209 		return -EPERM;
3210 
3211 	sa = memdup_user(arg, sizeof(*sa));
3212 	if (IS_ERR(sa))
3213 		return PTR_ERR(sa);
3214 
3215 	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3216 		ret = mnt_want_write_file(file);
3217 		if (ret)
3218 			goto out;
3219 	}
3220 
3221 	ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3222 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3223 			      0);
3224 
3225 	if (copy_to_user(arg, sa, sizeof(*sa)))
3226 		ret = -EFAULT;
3227 
3228 	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3229 		mnt_drop_write_file(file);
3230 out:
3231 	kfree(sa);
3232 	return ret;
3233 }
3234 
3235 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3236 {
3237 	if (!capable(CAP_SYS_ADMIN))
3238 		return -EPERM;
3239 
3240 	return btrfs_scrub_cancel(root->fs_info);
3241 }
3242 
3243 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3244 				       void __user *arg)
3245 {
3246 	struct btrfs_ioctl_scrub_args *sa;
3247 	int ret;
3248 
3249 	if (!capable(CAP_SYS_ADMIN))
3250 		return -EPERM;
3251 
3252 	sa = memdup_user(arg, sizeof(*sa));
3253 	if (IS_ERR(sa))
3254 		return PTR_ERR(sa);
3255 
3256 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3257 
3258 	if (copy_to_user(arg, sa, sizeof(*sa)))
3259 		ret = -EFAULT;
3260 
3261 	kfree(sa);
3262 	return ret;
3263 }
3264 
3265 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3266 				      void __user *arg)
3267 {
3268 	struct btrfs_ioctl_get_dev_stats *sa;
3269 	int ret;
3270 
3271 	sa = memdup_user(arg, sizeof(*sa));
3272 	if (IS_ERR(sa))
3273 		return PTR_ERR(sa);
3274 
3275 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3276 		kfree(sa);
3277 		return -EPERM;
3278 	}
3279 
3280 	ret = btrfs_get_dev_stats(root, sa);
3281 
3282 	if (copy_to_user(arg, sa, sizeof(*sa)))
3283 		ret = -EFAULT;
3284 
3285 	kfree(sa);
3286 	return ret;
3287 }
3288 
3289 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3290 {
3291 	struct btrfs_ioctl_dev_replace_args *p;
3292 	int ret;
3293 
3294 	if (!capable(CAP_SYS_ADMIN))
3295 		return -EPERM;
3296 
3297 	p = memdup_user(arg, sizeof(*p));
3298 	if (IS_ERR(p))
3299 		return PTR_ERR(p);
3300 
3301 	switch (p->cmd) {
3302 	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3303 		if (atomic_xchg(
3304 			&root->fs_info->mutually_exclusive_operation_running,
3305 			1)) {
3306 			pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3307 			ret = -EINPROGRESS;
3308 		} else {
3309 			ret = btrfs_dev_replace_start(root, p);
3310 			atomic_set(
3311 			 &root->fs_info->mutually_exclusive_operation_running,
3312 			 0);
3313 		}
3314 		break;
3315 	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3316 		btrfs_dev_replace_status(root->fs_info, p);
3317 		ret = 0;
3318 		break;
3319 	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3320 		ret = btrfs_dev_replace_cancel(root->fs_info, p);
3321 		break;
3322 	default:
3323 		ret = -EINVAL;
3324 		break;
3325 	}
3326 
3327 	if (copy_to_user(arg, p, sizeof(*p)))
3328 		ret = -EFAULT;
3329 
3330 	kfree(p);
3331 	return ret;
3332 }
3333 
3334 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3335 {
3336 	int ret = 0;
3337 	int i;
3338 	u64 rel_ptr;
3339 	int size;
3340 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3341 	struct inode_fs_paths *ipath = NULL;
3342 	struct btrfs_path *path;
3343 
3344 	if (!capable(CAP_DAC_READ_SEARCH))
3345 		return -EPERM;
3346 
3347 	path = btrfs_alloc_path();
3348 	if (!path) {
3349 		ret = -ENOMEM;
3350 		goto out;
3351 	}
3352 
3353 	ipa = memdup_user(arg, sizeof(*ipa));
3354 	if (IS_ERR(ipa)) {
3355 		ret = PTR_ERR(ipa);
3356 		ipa = NULL;
3357 		goto out;
3358 	}
3359 
3360 	size = min_t(u32, ipa->size, 4096);
3361 	ipath = init_ipath(size, root, path);
3362 	if (IS_ERR(ipath)) {
3363 		ret = PTR_ERR(ipath);
3364 		ipath = NULL;
3365 		goto out;
3366 	}
3367 
3368 	ret = paths_from_inode(ipa->inum, ipath);
3369 	if (ret < 0)
3370 		goto out;
3371 
3372 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3373 		rel_ptr = ipath->fspath->val[i] -
3374 			  (u64)(unsigned long)ipath->fspath->val;
3375 		ipath->fspath->val[i] = rel_ptr;
3376 	}
3377 
3378 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3379 			   (void *)(unsigned long)ipath->fspath, size);
3380 	if (ret) {
3381 		ret = -EFAULT;
3382 		goto out;
3383 	}
3384 
3385 out:
3386 	btrfs_free_path(path);
3387 	free_ipath(ipath);
3388 	kfree(ipa);
3389 
3390 	return ret;
3391 }
3392 
3393 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3394 {
3395 	struct btrfs_data_container *inodes = ctx;
3396 	const size_t c = 3 * sizeof(u64);
3397 
3398 	if (inodes->bytes_left >= c) {
3399 		inodes->bytes_left -= c;
3400 		inodes->val[inodes->elem_cnt] = inum;
3401 		inodes->val[inodes->elem_cnt + 1] = offset;
3402 		inodes->val[inodes->elem_cnt + 2] = root;
3403 		inodes->elem_cnt += 3;
3404 	} else {
3405 		inodes->bytes_missing += c - inodes->bytes_left;
3406 		inodes->bytes_left = 0;
3407 		inodes->elem_missed += 3;
3408 	}
3409 
3410 	return 0;
3411 }
3412 
3413 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3414 					void __user *arg)
3415 {
3416 	int ret = 0;
3417 	int size;
3418 	struct btrfs_ioctl_logical_ino_args *loi;
3419 	struct btrfs_data_container *inodes = NULL;
3420 	struct btrfs_path *path = NULL;
3421 
3422 	if (!capable(CAP_SYS_ADMIN))
3423 		return -EPERM;
3424 
3425 	loi = memdup_user(arg, sizeof(*loi));
3426 	if (IS_ERR(loi)) {
3427 		ret = PTR_ERR(loi);
3428 		loi = NULL;
3429 		goto out;
3430 	}
3431 
3432 	path = btrfs_alloc_path();
3433 	if (!path) {
3434 		ret = -ENOMEM;
3435 		goto out;
3436 	}
3437 
3438 	size = min_t(u32, loi->size, 64 * 1024);
3439 	inodes = init_data_container(size);
3440 	if (IS_ERR(inodes)) {
3441 		ret = PTR_ERR(inodes);
3442 		inodes = NULL;
3443 		goto out;
3444 	}
3445 
3446 	ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3447 					  build_ino_list, inodes);
3448 	if (ret == -EINVAL)
3449 		ret = -ENOENT;
3450 	if (ret < 0)
3451 		goto out;
3452 
3453 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3454 			   (void *)(unsigned long)inodes, size);
3455 	if (ret)
3456 		ret = -EFAULT;
3457 
3458 out:
3459 	btrfs_free_path(path);
3460 	vfree(inodes);
3461 	kfree(loi);
3462 
3463 	return ret;
3464 }
3465 
3466 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3467 			       struct btrfs_ioctl_balance_args *bargs)
3468 {
3469 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3470 
3471 	bargs->flags = bctl->flags;
3472 
3473 	if (atomic_read(&fs_info->balance_running))
3474 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3475 	if (atomic_read(&fs_info->balance_pause_req))
3476 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3477 	if (atomic_read(&fs_info->balance_cancel_req))
3478 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3479 
3480 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3481 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3482 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3483 
3484 	if (lock) {
3485 		spin_lock(&fs_info->balance_lock);
3486 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3487 		spin_unlock(&fs_info->balance_lock);
3488 	} else {
3489 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3490 	}
3491 }
3492 
3493 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3494 {
3495 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3496 	struct btrfs_fs_info *fs_info = root->fs_info;
3497 	struct btrfs_ioctl_balance_args *bargs;
3498 	struct btrfs_balance_control *bctl;
3499 	bool need_unlock; /* for mut. excl. ops lock */
3500 	int ret;
3501 
3502 	if (!capable(CAP_SYS_ADMIN))
3503 		return -EPERM;
3504 
3505 	ret = mnt_want_write_file(file);
3506 	if (ret)
3507 		return ret;
3508 
3509 again:
3510 	if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3511 		mutex_lock(&fs_info->volume_mutex);
3512 		mutex_lock(&fs_info->balance_mutex);
3513 		need_unlock = true;
3514 		goto locked;
3515 	}
3516 
3517 	/*
3518 	 * mut. excl. ops lock is locked.  Three possibilites:
3519 	 *   (1) some other op is running
3520 	 *   (2) balance is running
3521 	 *   (3) balance is paused -- special case (think resume)
3522 	 */
3523 	mutex_lock(&fs_info->balance_mutex);
3524 	if (fs_info->balance_ctl) {
3525 		/* this is either (2) or (3) */
3526 		if (!atomic_read(&fs_info->balance_running)) {
3527 			mutex_unlock(&fs_info->balance_mutex);
3528 			if (!mutex_trylock(&fs_info->volume_mutex))
3529 				goto again;
3530 			mutex_lock(&fs_info->balance_mutex);
3531 
3532 			if (fs_info->balance_ctl &&
3533 			    !atomic_read(&fs_info->balance_running)) {
3534 				/* this is (3) */
3535 				need_unlock = false;
3536 				goto locked;
3537 			}
3538 
3539 			mutex_unlock(&fs_info->balance_mutex);
3540 			mutex_unlock(&fs_info->volume_mutex);
3541 			goto again;
3542 		} else {
3543 			/* this is (2) */
3544 			mutex_unlock(&fs_info->balance_mutex);
3545 			ret = -EINPROGRESS;
3546 			goto out;
3547 		}
3548 	} else {
3549 		/* this is (1) */
3550 		mutex_unlock(&fs_info->balance_mutex);
3551 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3552 		ret = -EINVAL;
3553 		goto out;
3554 	}
3555 
3556 locked:
3557 	BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3558 
3559 	if (arg) {
3560 		bargs = memdup_user(arg, sizeof(*bargs));
3561 		if (IS_ERR(bargs)) {
3562 			ret = PTR_ERR(bargs);
3563 			goto out_unlock;
3564 		}
3565 
3566 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3567 			if (!fs_info->balance_ctl) {
3568 				ret = -ENOTCONN;
3569 				goto out_bargs;
3570 			}
3571 
3572 			bctl = fs_info->balance_ctl;
3573 			spin_lock(&fs_info->balance_lock);
3574 			bctl->flags |= BTRFS_BALANCE_RESUME;
3575 			spin_unlock(&fs_info->balance_lock);
3576 
3577 			goto do_balance;
3578 		}
3579 	} else {
3580 		bargs = NULL;
3581 	}
3582 
3583 	if (fs_info->balance_ctl) {
3584 		ret = -EINPROGRESS;
3585 		goto out_bargs;
3586 	}
3587 
3588 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3589 	if (!bctl) {
3590 		ret = -ENOMEM;
3591 		goto out_bargs;
3592 	}
3593 
3594 	bctl->fs_info = fs_info;
3595 	if (arg) {
3596 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3597 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3598 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3599 
3600 		bctl->flags = bargs->flags;
3601 	} else {
3602 		/* balance everything - no filters */
3603 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3604 	}
3605 
3606 do_balance:
3607 	/*
3608 	 * Ownership of bctl and mutually_exclusive_operation_running
3609 	 * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
3610 	 * or, if restriper was paused all the way until unmount, in
3611 	 * free_fs_info.  mutually_exclusive_operation_running is
3612 	 * cleared in __cancel_balance.
3613 	 */
3614 	need_unlock = false;
3615 
3616 	ret = btrfs_balance(bctl, bargs);
3617 
3618 	if (arg) {
3619 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3620 			ret = -EFAULT;
3621 	}
3622 
3623 out_bargs:
3624 	kfree(bargs);
3625 out_unlock:
3626 	mutex_unlock(&fs_info->balance_mutex);
3627 	mutex_unlock(&fs_info->volume_mutex);
3628 	if (need_unlock)
3629 		atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3630 out:
3631 	mnt_drop_write_file(file);
3632 	return ret;
3633 }
3634 
3635 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3636 {
3637 	if (!capable(CAP_SYS_ADMIN))
3638 		return -EPERM;
3639 
3640 	switch (cmd) {
3641 	case BTRFS_BALANCE_CTL_PAUSE:
3642 		return btrfs_pause_balance(root->fs_info);
3643 	case BTRFS_BALANCE_CTL_CANCEL:
3644 		return btrfs_cancel_balance(root->fs_info);
3645 	}
3646 
3647 	return -EINVAL;
3648 }
3649 
3650 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3651 					 void __user *arg)
3652 {
3653 	struct btrfs_fs_info *fs_info = root->fs_info;
3654 	struct btrfs_ioctl_balance_args *bargs;
3655 	int ret = 0;
3656 
3657 	if (!capable(CAP_SYS_ADMIN))
3658 		return -EPERM;
3659 
3660 	mutex_lock(&fs_info->balance_mutex);
3661 	if (!fs_info->balance_ctl) {
3662 		ret = -ENOTCONN;
3663 		goto out;
3664 	}
3665 
3666 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3667 	if (!bargs) {
3668 		ret = -ENOMEM;
3669 		goto out;
3670 	}
3671 
3672 	update_ioctl_balance_args(fs_info, 1, bargs);
3673 
3674 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3675 		ret = -EFAULT;
3676 
3677 	kfree(bargs);
3678 out:
3679 	mutex_unlock(&fs_info->balance_mutex);
3680 	return ret;
3681 }
3682 
3683 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3684 {
3685 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3686 	struct btrfs_ioctl_quota_ctl_args *sa;
3687 	struct btrfs_trans_handle *trans = NULL;
3688 	int ret;
3689 	int err;
3690 
3691 	if (!capable(CAP_SYS_ADMIN))
3692 		return -EPERM;
3693 
3694 	ret = mnt_want_write_file(file);
3695 	if (ret)
3696 		return ret;
3697 
3698 	sa = memdup_user(arg, sizeof(*sa));
3699 	if (IS_ERR(sa)) {
3700 		ret = PTR_ERR(sa);
3701 		goto drop_write;
3702 	}
3703 
3704 	if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3705 		trans = btrfs_start_transaction(root, 2);
3706 		if (IS_ERR(trans)) {
3707 			ret = PTR_ERR(trans);
3708 			goto out;
3709 		}
3710 	}
3711 
3712 	switch (sa->cmd) {
3713 	case BTRFS_QUOTA_CTL_ENABLE:
3714 		ret = btrfs_quota_enable(trans, root->fs_info);
3715 		break;
3716 	case BTRFS_QUOTA_CTL_DISABLE:
3717 		ret = btrfs_quota_disable(trans, root->fs_info);
3718 		break;
3719 	case BTRFS_QUOTA_CTL_RESCAN:
3720 		ret = btrfs_quota_rescan(root->fs_info);
3721 		break;
3722 	default:
3723 		ret = -EINVAL;
3724 		break;
3725 	}
3726 
3727 	if (copy_to_user(arg, sa, sizeof(*sa)))
3728 		ret = -EFAULT;
3729 
3730 	if (trans) {
3731 		err = btrfs_commit_transaction(trans, root);
3732 		if (err && !ret)
3733 			ret = err;
3734 	}
3735 out:
3736 	kfree(sa);
3737 drop_write:
3738 	mnt_drop_write_file(file);
3739 	return ret;
3740 }
3741 
3742 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3743 {
3744 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3745 	struct btrfs_ioctl_qgroup_assign_args *sa;
3746 	struct btrfs_trans_handle *trans;
3747 	int ret;
3748 	int err;
3749 
3750 	if (!capable(CAP_SYS_ADMIN))
3751 		return -EPERM;
3752 
3753 	ret = mnt_want_write_file(file);
3754 	if (ret)
3755 		return ret;
3756 
3757 	sa = memdup_user(arg, sizeof(*sa));
3758 	if (IS_ERR(sa)) {
3759 		ret = PTR_ERR(sa);
3760 		goto drop_write;
3761 	}
3762 
3763 	trans = btrfs_join_transaction(root);
3764 	if (IS_ERR(trans)) {
3765 		ret = PTR_ERR(trans);
3766 		goto out;
3767 	}
3768 
3769 	/* FIXME: check if the IDs really exist */
3770 	if (sa->assign) {
3771 		ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3772 						sa->src, sa->dst);
3773 	} else {
3774 		ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3775 						sa->src, sa->dst);
3776 	}
3777 
3778 	err = btrfs_end_transaction(trans, root);
3779 	if (err && !ret)
3780 		ret = err;
3781 
3782 out:
3783 	kfree(sa);
3784 drop_write:
3785 	mnt_drop_write_file(file);
3786 	return ret;
3787 }
3788 
3789 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3790 {
3791 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3792 	struct btrfs_ioctl_qgroup_create_args *sa;
3793 	struct btrfs_trans_handle *trans;
3794 	int ret;
3795 	int err;
3796 
3797 	if (!capable(CAP_SYS_ADMIN))
3798 		return -EPERM;
3799 
3800 	ret = mnt_want_write_file(file);
3801 	if (ret)
3802 		return ret;
3803 
3804 	sa = memdup_user(arg, sizeof(*sa));
3805 	if (IS_ERR(sa)) {
3806 		ret = PTR_ERR(sa);
3807 		goto drop_write;
3808 	}
3809 
3810 	if (!sa->qgroupid) {
3811 		ret = -EINVAL;
3812 		goto out;
3813 	}
3814 
3815 	trans = btrfs_join_transaction(root);
3816 	if (IS_ERR(trans)) {
3817 		ret = PTR_ERR(trans);
3818 		goto out;
3819 	}
3820 
3821 	/* FIXME: check if the IDs really exist */
3822 	if (sa->create) {
3823 		ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3824 					  NULL);
3825 	} else {
3826 		ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3827 	}
3828 
3829 	err = btrfs_end_transaction(trans, root);
3830 	if (err && !ret)
3831 		ret = err;
3832 
3833 out:
3834 	kfree(sa);
3835 drop_write:
3836 	mnt_drop_write_file(file);
3837 	return ret;
3838 }
3839 
3840 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3841 {
3842 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3843 	struct btrfs_ioctl_qgroup_limit_args *sa;
3844 	struct btrfs_trans_handle *trans;
3845 	int ret;
3846 	int err;
3847 	u64 qgroupid;
3848 
3849 	if (!capable(CAP_SYS_ADMIN))
3850 		return -EPERM;
3851 
3852 	ret = mnt_want_write_file(file);
3853 	if (ret)
3854 		return ret;
3855 
3856 	sa = memdup_user(arg, sizeof(*sa));
3857 	if (IS_ERR(sa)) {
3858 		ret = PTR_ERR(sa);
3859 		goto drop_write;
3860 	}
3861 
3862 	trans = btrfs_join_transaction(root);
3863 	if (IS_ERR(trans)) {
3864 		ret = PTR_ERR(trans);
3865 		goto out;
3866 	}
3867 
3868 	qgroupid = sa->qgroupid;
3869 	if (!qgroupid) {
3870 		/* take the current subvol as qgroup */
3871 		qgroupid = root->root_key.objectid;
3872 	}
3873 
3874 	/* FIXME: check if the IDs really exist */
3875 	ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3876 
3877 	err = btrfs_end_transaction(trans, root);
3878 	if (err && !ret)
3879 		ret = err;
3880 
3881 out:
3882 	kfree(sa);
3883 drop_write:
3884 	mnt_drop_write_file(file);
3885 	return ret;
3886 }
3887 
3888 static long btrfs_ioctl_set_received_subvol(struct file *file,
3889 					    void __user *arg)
3890 {
3891 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
3892 	struct inode *inode = file_inode(file);
3893 	struct btrfs_root *root = BTRFS_I(inode)->root;
3894 	struct btrfs_root_item *root_item = &root->root_item;
3895 	struct btrfs_trans_handle *trans;
3896 	struct timespec ct = CURRENT_TIME;
3897 	int ret = 0;
3898 
3899 	ret = mnt_want_write_file(file);
3900 	if (ret < 0)
3901 		return ret;
3902 
3903 	down_write(&root->fs_info->subvol_sem);
3904 
3905 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3906 		ret = -EINVAL;
3907 		goto out;
3908 	}
3909 
3910 	if (btrfs_root_readonly(root)) {
3911 		ret = -EROFS;
3912 		goto out;
3913 	}
3914 
3915 	if (!inode_owner_or_capable(inode)) {
3916 		ret = -EACCES;
3917 		goto out;
3918 	}
3919 
3920 	sa = memdup_user(arg, sizeof(*sa));
3921 	if (IS_ERR(sa)) {
3922 		ret = PTR_ERR(sa);
3923 		sa = NULL;
3924 		goto out;
3925 	}
3926 
3927 	trans = btrfs_start_transaction(root, 1);
3928 	if (IS_ERR(trans)) {
3929 		ret = PTR_ERR(trans);
3930 		trans = NULL;
3931 		goto out;
3932 	}
3933 
3934 	sa->rtransid = trans->transid;
3935 	sa->rtime.sec = ct.tv_sec;
3936 	sa->rtime.nsec = ct.tv_nsec;
3937 
3938 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3939 	btrfs_set_root_stransid(root_item, sa->stransid);
3940 	btrfs_set_root_rtransid(root_item, sa->rtransid);
3941 	root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3942 	root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3943 	root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3944 	root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3945 
3946 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
3947 				&root->root_key, &root->root_item);
3948 	if (ret < 0) {
3949 		btrfs_end_transaction(trans, root);
3950 		trans = NULL;
3951 		goto out;
3952 	} else {
3953 		ret = btrfs_commit_transaction(trans, root);
3954 		if (ret < 0)
3955 			goto out;
3956 	}
3957 
3958 	ret = copy_to_user(arg, sa, sizeof(*sa));
3959 	if (ret)
3960 		ret = -EFAULT;
3961 
3962 out:
3963 	kfree(sa);
3964 	up_write(&root->fs_info->subvol_sem);
3965 	mnt_drop_write_file(file);
3966 	return ret;
3967 }
3968 
3969 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
3970 {
3971 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3972 	const char *label = root->fs_info->super_copy->label;
3973 	size_t len = strnlen(label, BTRFS_LABEL_SIZE);
3974 	int ret;
3975 
3976 	if (len == BTRFS_LABEL_SIZE) {
3977 		pr_warn("btrfs: label is too long, return the first %zu bytes\n",
3978 			--len);
3979 	}
3980 
3981 	mutex_lock(&root->fs_info->volume_mutex);
3982 	ret = copy_to_user(arg, label, len);
3983 	mutex_unlock(&root->fs_info->volume_mutex);
3984 
3985 	return ret ? -EFAULT : 0;
3986 }
3987 
3988 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
3989 {
3990 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3991 	struct btrfs_super_block *super_block = root->fs_info->super_copy;
3992 	struct btrfs_trans_handle *trans;
3993 	char label[BTRFS_LABEL_SIZE];
3994 	int ret;
3995 
3996 	if (!capable(CAP_SYS_ADMIN))
3997 		return -EPERM;
3998 
3999 	if (copy_from_user(label, arg, sizeof(label)))
4000 		return -EFAULT;
4001 
4002 	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4003 		pr_err("btrfs: unable to set label with more than %d bytes\n",
4004 		       BTRFS_LABEL_SIZE - 1);
4005 		return -EINVAL;
4006 	}
4007 
4008 	ret = mnt_want_write_file(file);
4009 	if (ret)
4010 		return ret;
4011 
4012 	mutex_lock(&root->fs_info->volume_mutex);
4013 	trans = btrfs_start_transaction(root, 0);
4014 	if (IS_ERR(trans)) {
4015 		ret = PTR_ERR(trans);
4016 		goto out_unlock;
4017 	}
4018 
4019 	strcpy(super_block->label, label);
4020 	ret = btrfs_end_transaction(trans, root);
4021 
4022 out_unlock:
4023 	mutex_unlock(&root->fs_info->volume_mutex);
4024 	mnt_drop_write_file(file);
4025 	return ret;
4026 }
4027 
4028 long btrfs_ioctl(struct file *file, unsigned int
4029 		cmd, unsigned long arg)
4030 {
4031 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4032 	void __user *argp = (void __user *)arg;
4033 
4034 	switch (cmd) {
4035 	case FS_IOC_GETFLAGS:
4036 		return btrfs_ioctl_getflags(file, argp);
4037 	case FS_IOC_SETFLAGS:
4038 		return btrfs_ioctl_setflags(file, argp);
4039 	case FS_IOC_GETVERSION:
4040 		return btrfs_ioctl_getversion(file, argp);
4041 	case FITRIM:
4042 		return btrfs_ioctl_fitrim(file, argp);
4043 	case BTRFS_IOC_SNAP_CREATE:
4044 		return btrfs_ioctl_snap_create(file, argp, 0);
4045 	case BTRFS_IOC_SNAP_CREATE_V2:
4046 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
4047 	case BTRFS_IOC_SUBVOL_CREATE:
4048 		return btrfs_ioctl_snap_create(file, argp, 1);
4049 	case BTRFS_IOC_SUBVOL_CREATE_V2:
4050 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
4051 	case BTRFS_IOC_SNAP_DESTROY:
4052 		return btrfs_ioctl_snap_destroy(file, argp);
4053 	case BTRFS_IOC_SUBVOL_GETFLAGS:
4054 		return btrfs_ioctl_subvol_getflags(file, argp);
4055 	case BTRFS_IOC_SUBVOL_SETFLAGS:
4056 		return btrfs_ioctl_subvol_setflags(file, argp);
4057 	case BTRFS_IOC_DEFAULT_SUBVOL:
4058 		return btrfs_ioctl_default_subvol(file, argp);
4059 	case BTRFS_IOC_DEFRAG:
4060 		return btrfs_ioctl_defrag(file, NULL);
4061 	case BTRFS_IOC_DEFRAG_RANGE:
4062 		return btrfs_ioctl_defrag(file, argp);
4063 	case BTRFS_IOC_RESIZE:
4064 		return btrfs_ioctl_resize(file, argp);
4065 	case BTRFS_IOC_ADD_DEV:
4066 		return btrfs_ioctl_add_dev(root, argp);
4067 	case BTRFS_IOC_RM_DEV:
4068 		return btrfs_ioctl_rm_dev(file, argp);
4069 	case BTRFS_IOC_FS_INFO:
4070 		return btrfs_ioctl_fs_info(root, argp);
4071 	case BTRFS_IOC_DEV_INFO:
4072 		return btrfs_ioctl_dev_info(root, argp);
4073 	case BTRFS_IOC_BALANCE:
4074 		return btrfs_ioctl_balance(file, NULL);
4075 	case BTRFS_IOC_CLONE:
4076 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4077 	case BTRFS_IOC_CLONE_RANGE:
4078 		return btrfs_ioctl_clone_range(file, argp);
4079 	case BTRFS_IOC_TRANS_START:
4080 		return btrfs_ioctl_trans_start(file);
4081 	case BTRFS_IOC_TRANS_END:
4082 		return btrfs_ioctl_trans_end(file);
4083 	case BTRFS_IOC_TREE_SEARCH:
4084 		return btrfs_ioctl_tree_search(file, argp);
4085 	case BTRFS_IOC_INO_LOOKUP:
4086 		return btrfs_ioctl_ino_lookup(file, argp);
4087 	case BTRFS_IOC_INO_PATHS:
4088 		return btrfs_ioctl_ino_to_path(root, argp);
4089 	case BTRFS_IOC_LOGICAL_INO:
4090 		return btrfs_ioctl_logical_to_ino(root, argp);
4091 	case BTRFS_IOC_SPACE_INFO:
4092 		return btrfs_ioctl_space_info(root, argp);
4093 	case BTRFS_IOC_SYNC:
4094 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
4095 		return 0;
4096 	case BTRFS_IOC_START_SYNC:
4097 		return btrfs_ioctl_start_sync(root, argp);
4098 	case BTRFS_IOC_WAIT_SYNC:
4099 		return btrfs_ioctl_wait_sync(root, argp);
4100 	case BTRFS_IOC_SCRUB:
4101 		return btrfs_ioctl_scrub(file, argp);
4102 	case BTRFS_IOC_SCRUB_CANCEL:
4103 		return btrfs_ioctl_scrub_cancel(root, argp);
4104 	case BTRFS_IOC_SCRUB_PROGRESS:
4105 		return btrfs_ioctl_scrub_progress(root, argp);
4106 	case BTRFS_IOC_BALANCE_V2:
4107 		return btrfs_ioctl_balance(file, argp);
4108 	case BTRFS_IOC_BALANCE_CTL:
4109 		return btrfs_ioctl_balance_ctl(root, arg);
4110 	case BTRFS_IOC_BALANCE_PROGRESS:
4111 		return btrfs_ioctl_balance_progress(root, argp);
4112 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4113 		return btrfs_ioctl_set_received_subvol(file, argp);
4114 	case BTRFS_IOC_SEND:
4115 		return btrfs_ioctl_send(file, argp);
4116 	case BTRFS_IOC_GET_DEV_STATS:
4117 		return btrfs_ioctl_get_dev_stats(root, argp);
4118 	case BTRFS_IOC_QUOTA_CTL:
4119 		return btrfs_ioctl_quota_ctl(file, argp);
4120 	case BTRFS_IOC_QGROUP_ASSIGN:
4121 		return btrfs_ioctl_qgroup_assign(file, argp);
4122 	case BTRFS_IOC_QGROUP_CREATE:
4123 		return btrfs_ioctl_qgroup_create(file, argp);
4124 	case BTRFS_IOC_QGROUP_LIMIT:
4125 		return btrfs_ioctl_qgroup_limit(file, argp);
4126 	case BTRFS_IOC_DEV_REPLACE:
4127 		return btrfs_ioctl_dev_replace(root, argp);
4128 	case BTRFS_IOC_GET_FSLABEL:
4129 		return btrfs_ioctl_get_fslabel(file, argp);
4130 	case BTRFS_IOC_SET_FSLABEL:
4131 		return btrfs_ioctl_set_fslabel(file, argp);
4132 	}
4133 
4134 	return -ENOTTY;
4135 }
4136