xref: /openbmc/linux/fs/btrfs/ioctl.c (revision cdfce539)
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 		if (err)
531 			err = btrfs_commit_transaction(trans, root);
532 	} else {
533 		err = btrfs_commit_transaction(trans, root);
534 	}
535 	if (err && !ret)
536 		ret = err;
537 
538 	if (!ret)
539 		d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
540 out:
541 	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
542 	return ret;
543 }
544 
545 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
546 			   struct dentry *dentry, char *name, int namelen,
547 			   u64 *async_transid, bool readonly,
548 			   struct btrfs_qgroup_inherit *inherit)
549 {
550 	struct inode *inode;
551 	struct btrfs_pending_snapshot *pending_snapshot;
552 	struct btrfs_trans_handle *trans;
553 	int ret;
554 
555 	if (!root->ref_cows)
556 		return -EINVAL;
557 
558 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
559 	if (!pending_snapshot)
560 		return -ENOMEM;
561 
562 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
563 			     BTRFS_BLOCK_RSV_TEMP);
564 	/*
565 	 * 1 - parent dir inode
566 	 * 2 - dir entries
567 	 * 1 - root item
568 	 * 2 - root ref/backref
569 	 * 1 - root of snapshot
570 	 */
571 	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
572 					&pending_snapshot->block_rsv, 7,
573 					&pending_snapshot->qgroup_reserved);
574 	if (ret)
575 		goto out;
576 
577 	pending_snapshot->dentry = dentry;
578 	pending_snapshot->root = root;
579 	pending_snapshot->readonly = readonly;
580 	pending_snapshot->dir = dir;
581 	pending_snapshot->inherit = inherit;
582 
583 	trans = btrfs_start_transaction(root, 0);
584 	if (IS_ERR(trans)) {
585 		ret = PTR_ERR(trans);
586 		goto fail;
587 	}
588 
589 	spin_lock(&root->fs_info->trans_lock);
590 	list_add(&pending_snapshot->list,
591 		 &trans->transaction->pending_snapshots);
592 	spin_unlock(&root->fs_info->trans_lock);
593 	if (async_transid) {
594 		*async_transid = trans->transid;
595 		ret = btrfs_commit_transaction_async(trans,
596 				     root->fs_info->extent_root, 1);
597 		if (ret)
598 			ret = btrfs_commit_transaction(trans, root);
599 	} else {
600 		ret = btrfs_commit_transaction(trans,
601 					       root->fs_info->extent_root);
602 	}
603 	if (ret)
604 		goto fail;
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 	error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
727 	if (error == -EINTR)
728 		return error;
729 
730 	dentry = lookup_one_len(name, parent->dentry, namelen);
731 	error = PTR_ERR(dentry);
732 	if (IS_ERR(dentry))
733 		goto out_unlock;
734 
735 	error = -EEXIST;
736 	if (dentry->d_inode)
737 		goto out_dput;
738 
739 	error = btrfs_may_create(dir, dentry);
740 	if (error)
741 		goto out_dput;
742 
743 	/*
744 	 * even if this name doesn't exist, we may get hash collisions.
745 	 * check for them now when we can safely fail
746 	 */
747 	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
748 					       dir->i_ino, name,
749 					       namelen);
750 	if (error)
751 		goto out_dput;
752 
753 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
754 
755 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
756 		goto out_up_read;
757 
758 	if (snap_src) {
759 		error = create_snapshot(snap_src, dir, dentry, name, namelen,
760 					async_transid, readonly, inherit);
761 	} else {
762 		error = create_subvol(dir, dentry, name, namelen,
763 				      async_transid, inherit);
764 	}
765 	if (!error)
766 		fsnotify_mkdir(dir, dentry);
767 out_up_read:
768 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
769 out_dput:
770 	dput(dentry);
771 out_unlock:
772 	mutex_unlock(&dir->i_mutex);
773 	return error;
774 }
775 
776 /*
777  * When we're defragging a range, we don't want to kick it off again
778  * if it is really just waiting for delalloc to send it down.
779  * If we find a nice big extent or delalloc range for the bytes in the
780  * file you want to defrag, we return 0 to let you know to skip this
781  * part of the file
782  */
783 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
784 {
785 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
786 	struct extent_map *em = NULL;
787 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
788 	u64 end;
789 
790 	read_lock(&em_tree->lock);
791 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
792 	read_unlock(&em_tree->lock);
793 
794 	if (em) {
795 		end = extent_map_end(em);
796 		free_extent_map(em);
797 		if (end - offset > thresh)
798 			return 0;
799 	}
800 	/* if we already have a nice delalloc here, just stop */
801 	thresh /= 2;
802 	end = count_range_bits(io_tree, &offset, offset + thresh,
803 			       thresh, EXTENT_DELALLOC, 1);
804 	if (end >= thresh)
805 		return 0;
806 	return 1;
807 }
808 
809 /*
810  * helper function to walk through a file and find extents
811  * newer than a specific transid, and smaller than thresh.
812  *
813  * This is used by the defragging code to find new and small
814  * extents
815  */
816 static int find_new_extents(struct btrfs_root *root,
817 			    struct inode *inode, u64 newer_than,
818 			    u64 *off, int thresh)
819 {
820 	struct btrfs_path *path;
821 	struct btrfs_key min_key;
822 	struct btrfs_key max_key;
823 	struct extent_buffer *leaf;
824 	struct btrfs_file_extent_item *extent;
825 	int type;
826 	int ret;
827 	u64 ino = btrfs_ino(inode);
828 
829 	path = btrfs_alloc_path();
830 	if (!path)
831 		return -ENOMEM;
832 
833 	min_key.objectid = ino;
834 	min_key.type = BTRFS_EXTENT_DATA_KEY;
835 	min_key.offset = *off;
836 
837 	max_key.objectid = ino;
838 	max_key.type = (u8)-1;
839 	max_key.offset = (u64)-1;
840 
841 	path->keep_locks = 1;
842 
843 	while(1) {
844 		ret = btrfs_search_forward(root, &min_key, &max_key,
845 					   path, newer_than);
846 		if (ret != 0)
847 			goto none;
848 		if (min_key.objectid != ino)
849 			goto none;
850 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
851 			goto none;
852 
853 		leaf = path->nodes[0];
854 		extent = btrfs_item_ptr(leaf, path->slots[0],
855 					struct btrfs_file_extent_item);
856 
857 		type = btrfs_file_extent_type(leaf, extent);
858 		if (type == BTRFS_FILE_EXTENT_REG &&
859 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
860 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
861 			*off = min_key.offset;
862 			btrfs_free_path(path);
863 			return 0;
864 		}
865 
866 		if (min_key.offset == (u64)-1)
867 			goto none;
868 
869 		min_key.offset++;
870 		btrfs_release_path(path);
871 	}
872 none:
873 	btrfs_free_path(path);
874 	return -ENOENT;
875 }
876 
877 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
878 {
879 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
880 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
881 	struct extent_map *em;
882 	u64 len = PAGE_CACHE_SIZE;
883 
884 	/*
885 	 * hopefully we have this extent in the tree already, try without
886 	 * the full extent lock
887 	 */
888 	read_lock(&em_tree->lock);
889 	em = lookup_extent_mapping(em_tree, start, len);
890 	read_unlock(&em_tree->lock);
891 
892 	if (!em) {
893 		/* get the big lock and read metadata off disk */
894 		lock_extent(io_tree, start, start + len - 1);
895 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
896 		unlock_extent(io_tree, start, start + len - 1);
897 
898 		if (IS_ERR(em))
899 			return NULL;
900 	}
901 
902 	return em;
903 }
904 
905 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
906 {
907 	struct extent_map *next;
908 	bool ret = true;
909 
910 	/* this is the last extent */
911 	if (em->start + em->len >= i_size_read(inode))
912 		return false;
913 
914 	next = defrag_lookup_extent(inode, em->start + em->len);
915 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
916 		ret = false;
917 
918 	free_extent_map(next);
919 	return ret;
920 }
921 
922 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
923 			       u64 *last_len, u64 *skip, u64 *defrag_end,
924 			       int compress)
925 {
926 	struct extent_map *em;
927 	int ret = 1;
928 	bool next_mergeable = true;
929 
930 	/*
931 	 * make sure that once we start defragging an extent, we keep on
932 	 * defragging it
933 	 */
934 	if (start < *defrag_end)
935 		return 1;
936 
937 	*skip = 0;
938 
939 	em = defrag_lookup_extent(inode, start);
940 	if (!em)
941 		return 0;
942 
943 	/* this will cover holes, and inline extents */
944 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
945 		ret = 0;
946 		goto out;
947 	}
948 
949 	next_mergeable = defrag_check_next_extent(inode, em);
950 
951 	/*
952 	 * we hit a real extent, if it is big or the next extent is not a
953 	 * real extent, don't bother defragging it
954 	 */
955 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
956 	    (em->len >= thresh || !next_mergeable))
957 		ret = 0;
958 out:
959 	/*
960 	 * last_len ends up being a counter of how many bytes we've defragged.
961 	 * every time we choose not to defrag an extent, we reset *last_len
962 	 * so that the next tiny extent will force a defrag.
963 	 *
964 	 * The end result of this is that tiny extents before a single big
965 	 * extent will force at least part of that big extent to be defragged.
966 	 */
967 	if (ret) {
968 		*defrag_end = extent_map_end(em);
969 	} else {
970 		*last_len = 0;
971 		*skip = extent_map_end(em);
972 		*defrag_end = 0;
973 	}
974 
975 	free_extent_map(em);
976 	return ret;
977 }
978 
979 /*
980  * it doesn't do much good to defrag one or two pages
981  * at a time.  This pulls in a nice chunk of pages
982  * to COW and defrag.
983  *
984  * It also makes sure the delalloc code has enough
985  * dirty data to avoid making new small extents as part
986  * of the defrag
987  *
988  * It's a good idea to start RA on this range
989  * before calling this.
990  */
991 static int cluster_pages_for_defrag(struct inode *inode,
992 				    struct page **pages,
993 				    unsigned long start_index,
994 				    int num_pages)
995 {
996 	unsigned long file_end;
997 	u64 isize = i_size_read(inode);
998 	u64 page_start;
999 	u64 page_end;
1000 	u64 page_cnt;
1001 	int ret;
1002 	int i;
1003 	int i_done;
1004 	struct btrfs_ordered_extent *ordered;
1005 	struct extent_state *cached_state = NULL;
1006 	struct extent_io_tree *tree;
1007 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1008 
1009 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1010 	if (!isize || start_index > file_end)
1011 		return 0;
1012 
1013 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1014 
1015 	ret = btrfs_delalloc_reserve_space(inode,
1016 					   page_cnt << PAGE_CACHE_SHIFT);
1017 	if (ret)
1018 		return ret;
1019 	i_done = 0;
1020 	tree = &BTRFS_I(inode)->io_tree;
1021 
1022 	/* step one, lock all the pages */
1023 	for (i = 0; i < page_cnt; i++) {
1024 		struct page *page;
1025 again:
1026 		page = find_or_create_page(inode->i_mapping,
1027 					   start_index + i, mask);
1028 		if (!page)
1029 			break;
1030 
1031 		page_start = page_offset(page);
1032 		page_end = page_start + PAGE_CACHE_SIZE - 1;
1033 		while (1) {
1034 			lock_extent(tree, page_start, page_end);
1035 			ordered = btrfs_lookup_ordered_extent(inode,
1036 							      page_start);
1037 			unlock_extent(tree, page_start, page_end);
1038 			if (!ordered)
1039 				break;
1040 
1041 			unlock_page(page);
1042 			btrfs_start_ordered_extent(inode, ordered, 1);
1043 			btrfs_put_ordered_extent(ordered);
1044 			lock_page(page);
1045 			/*
1046 			 * we unlocked the page above, so we need check if
1047 			 * it was released or not.
1048 			 */
1049 			if (page->mapping != inode->i_mapping) {
1050 				unlock_page(page);
1051 				page_cache_release(page);
1052 				goto again;
1053 			}
1054 		}
1055 
1056 		if (!PageUptodate(page)) {
1057 			btrfs_readpage(NULL, page);
1058 			lock_page(page);
1059 			if (!PageUptodate(page)) {
1060 				unlock_page(page);
1061 				page_cache_release(page);
1062 				ret = -EIO;
1063 				break;
1064 			}
1065 		}
1066 
1067 		if (page->mapping != inode->i_mapping) {
1068 			unlock_page(page);
1069 			page_cache_release(page);
1070 			goto again;
1071 		}
1072 
1073 		pages[i] = page;
1074 		i_done++;
1075 	}
1076 	if (!i_done || ret)
1077 		goto out;
1078 
1079 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
1080 		goto out;
1081 
1082 	/*
1083 	 * so now we have a nice long stream of locked
1084 	 * and up to date pages, lets wait on them
1085 	 */
1086 	for (i = 0; i < i_done; i++)
1087 		wait_on_page_writeback(pages[i]);
1088 
1089 	page_start = page_offset(pages[0]);
1090 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1091 
1092 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1093 			 page_start, page_end - 1, 0, &cached_state);
1094 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1095 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1096 			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1097 			  &cached_state, GFP_NOFS);
1098 
1099 	if (i_done != page_cnt) {
1100 		spin_lock(&BTRFS_I(inode)->lock);
1101 		BTRFS_I(inode)->outstanding_extents++;
1102 		spin_unlock(&BTRFS_I(inode)->lock);
1103 		btrfs_delalloc_release_space(inode,
1104 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1105 	}
1106 
1107 
1108 	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1109 			  &cached_state, GFP_NOFS);
1110 
1111 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1112 			     page_start, page_end - 1, &cached_state,
1113 			     GFP_NOFS);
1114 
1115 	for (i = 0; i < i_done; i++) {
1116 		clear_page_dirty_for_io(pages[i]);
1117 		ClearPageChecked(pages[i]);
1118 		set_page_extent_mapped(pages[i]);
1119 		set_page_dirty(pages[i]);
1120 		unlock_page(pages[i]);
1121 		page_cache_release(pages[i]);
1122 	}
1123 	return i_done;
1124 out:
1125 	for (i = 0; i < i_done; i++) {
1126 		unlock_page(pages[i]);
1127 		page_cache_release(pages[i]);
1128 	}
1129 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1130 	return ret;
1131 
1132 }
1133 
1134 int btrfs_defrag_file(struct inode *inode, struct file *file,
1135 		      struct btrfs_ioctl_defrag_range_args *range,
1136 		      u64 newer_than, unsigned long max_to_defrag)
1137 {
1138 	struct btrfs_root *root = BTRFS_I(inode)->root;
1139 	struct file_ra_state *ra = NULL;
1140 	unsigned long last_index;
1141 	u64 isize = i_size_read(inode);
1142 	u64 last_len = 0;
1143 	u64 skip = 0;
1144 	u64 defrag_end = 0;
1145 	u64 newer_off = range->start;
1146 	unsigned long i;
1147 	unsigned long ra_index = 0;
1148 	int ret;
1149 	int defrag_count = 0;
1150 	int compress_type = BTRFS_COMPRESS_ZLIB;
1151 	int extent_thresh = range->extent_thresh;
1152 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1153 	int cluster = max_cluster;
1154 	u64 new_align = ~((u64)128 * 1024 - 1);
1155 	struct page **pages = NULL;
1156 
1157 	if (isize == 0)
1158 		return 0;
1159 
1160 	if (range->start >= isize)
1161 		return -EINVAL;
1162 
1163 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1164 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1165 			return -EINVAL;
1166 		if (range->compress_type)
1167 			compress_type = range->compress_type;
1168 	}
1169 
1170 	if (extent_thresh == 0)
1171 		extent_thresh = 256 * 1024;
1172 
1173 	/*
1174 	 * if we were not given a file, allocate a readahead
1175 	 * context
1176 	 */
1177 	if (!file) {
1178 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1179 		if (!ra)
1180 			return -ENOMEM;
1181 		file_ra_state_init(ra, inode->i_mapping);
1182 	} else {
1183 		ra = &file->f_ra;
1184 	}
1185 
1186 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1187 			GFP_NOFS);
1188 	if (!pages) {
1189 		ret = -ENOMEM;
1190 		goto out_ra;
1191 	}
1192 
1193 	/* find the last page to defrag */
1194 	if (range->start + range->len > range->start) {
1195 		last_index = min_t(u64, isize - 1,
1196 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1197 	} else {
1198 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1199 	}
1200 
1201 	if (newer_than) {
1202 		ret = find_new_extents(root, inode, newer_than,
1203 				       &newer_off, 64 * 1024);
1204 		if (!ret) {
1205 			range->start = newer_off;
1206 			/*
1207 			 * we always align our defrag to help keep
1208 			 * the extents in the file evenly spaced
1209 			 */
1210 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1211 		} else
1212 			goto out_ra;
1213 	} else {
1214 		i = range->start >> PAGE_CACHE_SHIFT;
1215 	}
1216 	if (!max_to_defrag)
1217 		max_to_defrag = last_index + 1;
1218 
1219 	/*
1220 	 * make writeback starts from i, so the defrag range can be
1221 	 * written sequentially.
1222 	 */
1223 	if (i < inode->i_mapping->writeback_index)
1224 		inode->i_mapping->writeback_index = i;
1225 
1226 	while (i <= last_index && defrag_count < max_to_defrag &&
1227 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1228 		PAGE_CACHE_SHIFT)) {
1229 		/*
1230 		 * make sure we stop running if someone unmounts
1231 		 * the FS
1232 		 */
1233 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1234 			break;
1235 
1236 		if (btrfs_defrag_cancelled(root->fs_info)) {
1237 			printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1238 			ret = -EAGAIN;
1239 			break;
1240 		}
1241 
1242 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1243 					 extent_thresh, &last_len, &skip,
1244 					 &defrag_end, range->flags &
1245 					 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1246 			unsigned long next;
1247 			/*
1248 			 * the should_defrag function tells us how much to skip
1249 			 * bump our counter by the suggested amount
1250 			 */
1251 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1252 			i = max(i + 1, next);
1253 			continue;
1254 		}
1255 
1256 		if (!newer_than) {
1257 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1258 				   PAGE_CACHE_SHIFT) - i;
1259 			cluster = min(cluster, max_cluster);
1260 		} else {
1261 			cluster = max_cluster;
1262 		}
1263 
1264 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1265 			BTRFS_I(inode)->force_compress = compress_type;
1266 
1267 		if (i + cluster > ra_index) {
1268 			ra_index = max(i, ra_index);
1269 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1270 				       cluster);
1271 			ra_index += max_cluster;
1272 		}
1273 
1274 		mutex_lock(&inode->i_mutex);
1275 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1276 		if (ret < 0) {
1277 			mutex_unlock(&inode->i_mutex);
1278 			goto out_ra;
1279 		}
1280 
1281 		defrag_count += ret;
1282 		balance_dirty_pages_ratelimited(inode->i_mapping);
1283 		mutex_unlock(&inode->i_mutex);
1284 
1285 		if (newer_than) {
1286 			if (newer_off == (u64)-1)
1287 				break;
1288 
1289 			if (ret > 0)
1290 				i += ret;
1291 
1292 			newer_off = max(newer_off + 1,
1293 					(u64)i << PAGE_CACHE_SHIFT);
1294 
1295 			ret = find_new_extents(root, inode,
1296 					       newer_than, &newer_off,
1297 					       64 * 1024);
1298 			if (!ret) {
1299 				range->start = newer_off;
1300 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1301 			} else {
1302 				break;
1303 			}
1304 		} else {
1305 			if (ret > 0) {
1306 				i += ret;
1307 				last_len += ret << PAGE_CACHE_SHIFT;
1308 			} else {
1309 				i++;
1310 				last_len = 0;
1311 			}
1312 		}
1313 	}
1314 
1315 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1316 		filemap_flush(inode->i_mapping);
1317 
1318 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1319 		/* the filemap_flush will queue IO into the worker threads, but
1320 		 * we have to make sure the IO is actually started and that
1321 		 * ordered extents get created before we return
1322 		 */
1323 		atomic_inc(&root->fs_info->async_submit_draining);
1324 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1325 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1326 			wait_event(root->fs_info->async_submit_wait,
1327 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1328 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1329 		}
1330 		atomic_dec(&root->fs_info->async_submit_draining);
1331 
1332 		mutex_lock(&inode->i_mutex);
1333 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1334 		mutex_unlock(&inode->i_mutex);
1335 	}
1336 
1337 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1338 		btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1339 	}
1340 
1341 	ret = defrag_count;
1342 
1343 out_ra:
1344 	if (!file)
1345 		kfree(ra);
1346 	kfree(pages);
1347 	return ret;
1348 }
1349 
1350 static noinline int btrfs_ioctl_resize(struct file *file,
1351 					void __user *arg)
1352 {
1353 	u64 new_size;
1354 	u64 old_size;
1355 	u64 devid = 1;
1356 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1357 	struct btrfs_ioctl_vol_args *vol_args;
1358 	struct btrfs_trans_handle *trans;
1359 	struct btrfs_device *device = NULL;
1360 	char *sizestr;
1361 	char *devstr = NULL;
1362 	int ret = 0;
1363 	int mod = 0;
1364 
1365 	if (!capable(CAP_SYS_ADMIN))
1366 		return -EPERM;
1367 
1368 	ret = mnt_want_write_file(file);
1369 	if (ret)
1370 		return ret;
1371 
1372 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1373 			1)) {
1374 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1375 		mnt_drop_write_file(file);
1376 		return -EINVAL;
1377 	}
1378 
1379 	mutex_lock(&root->fs_info->volume_mutex);
1380 	vol_args = memdup_user(arg, sizeof(*vol_args));
1381 	if (IS_ERR(vol_args)) {
1382 		ret = PTR_ERR(vol_args);
1383 		goto out;
1384 	}
1385 
1386 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1387 
1388 	sizestr = vol_args->name;
1389 	devstr = strchr(sizestr, ':');
1390 	if (devstr) {
1391 		char *end;
1392 		sizestr = devstr + 1;
1393 		*devstr = '\0';
1394 		devstr = vol_args->name;
1395 		devid = simple_strtoull(devstr, &end, 10);
1396 		if (!devid) {
1397 			ret = -EINVAL;
1398 			goto out_free;
1399 		}
1400 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1401 		       (unsigned long long)devid);
1402 	}
1403 
1404 	device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1405 	if (!device) {
1406 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1407 		       (unsigned long long)devid);
1408 		ret = -ENODEV;
1409 		goto out_free;
1410 	}
1411 
1412 	if (!device->writeable) {
1413 		printk(KERN_INFO "btrfs: resizer unable to apply on "
1414 		       "readonly device %llu\n",
1415 		       (unsigned long long)devid);
1416 		ret = -EPERM;
1417 		goto out_free;
1418 	}
1419 
1420 	if (!strcmp(sizestr, "max"))
1421 		new_size = device->bdev->bd_inode->i_size;
1422 	else {
1423 		if (sizestr[0] == '-') {
1424 			mod = -1;
1425 			sizestr++;
1426 		} else if (sizestr[0] == '+') {
1427 			mod = 1;
1428 			sizestr++;
1429 		}
1430 		new_size = memparse(sizestr, NULL);
1431 		if (new_size == 0) {
1432 			ret = -EINVAL;
1433 			goto out_free;
1434 		}
1435 	}
1436 
1437 	if (device->is_tgtdev_for_dev_replace) {
1438 		ret = -EPERM;
1439 		goto out_free;
1440 	}
1441 
1442 	old_size = device->total_bytes;
1443 
1444 	if (mod < 0) {
1445 		if (new_size > old_size) {
1446 			ret = -EINVAL;
1447 			goto out_free;
1448 		}
1449 		new_size = old_size - new_size;
1450 	} else if (mod > 0) {
1451 		new_size = old_size + new_size;
1452 	}
1453 
1454 	if (new_size < 256 * 1024 * 1024) {
1455 		ret = -EINVAL;
1456 		goto out_free;
1457 	}
1458 	if (new_size > device->bdev->bd_inode->i_size) {
1459 		ret = -EFBIG;
1460 		goto out_free;
1461 	}
1462 
1463 	do_div(new_size, root->sectorsize);
1464 	new_size *= root->sectorsize;
1465 
1466 	printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1467 		      rcu_str_deref(device->name),
1468 		      (unsigned long long)new_size);
1469 
1470 	if (new_size > old_size) {
1471 		trans = btrfs_start_transaction(root, 0);
1472 		if (IS_ERR(trans)) {
1473 			ret = PTR_ERR(trans);
1474 			goto out_free;
1475 		}
1476 		ret = btrfs_grow_device(trans, device, new_size);
1477 		btrfs_commit_transaction(trans, root);
1478 	} else if (new_size < old_size) {
1479 		ret = btrfs_shrink_device(device, new_size);
1480 	} /* equal, nothing need to do */
1481 
1482 out_free:
1483 	kfree(vol_args);
1484 out:
1485 	mutex_unlock(&root->fs_info->volume_mutex);
1486 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1487 	mnt_drop_write_file(file);
1488 	return ret;
1489 }
1490 
1491 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1492 				char *name, unsigned long fd, int subvol,
1493 				u64 *transid, bool readonly,
1494 				struct btrfs_qgroup_inherit *inherit)
1495 {
1496 	int namelen;
1497 	int ret = 0;
1498 
1499 	ret = mnt_want_write_file(file);
1500 	if (ret)
1501 		goto out;
1502 
1503 	namelen = strlen(name);
1504 	if (strchr(name, '/')) {
1505 		ret = -EINVAL;
1506 		goto out_drop_write;
1507 	}
1508 
1509 	if (name[0] == '.' &&
1510 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1511 		ret = -EEXIST;
1512 		goto out_drop_write;
1513 	}
1514 
1515 	if (subvol) {
1516 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1517 				     NULL, transid, readonly, inherit);
1518 	} else {
1519 		struct fd src = fdget(fd);
1520 		struct inode *src_inode;
1521 		if (!src.file) {
1522 			ret = -EINVAL;
1523 			goto out_drop_write;
1524 		}
1525 
1526 		src_inode = file_inode(src.file);
1527 		if (src_inode->i_sb != file_inode(file)->i_sb) {
1528 			printk(KERN_INFO "btrfs: Snapshot src from "
1529 			       "another FS\n");
1530 			ret = -EINVAL;
1531 		} else {
1532 			ret = btrfs_mksubvol(&file->f_path, name, namelen,
1533 					     BTRFS_I(src_inode)->root,
1534 					     transid, readonly, inherit);
1535 		}
1536 		fdput(src);
1537 	}
1538 out_drop_write:
1539 	mnt_drop_write_file(file);
1540 out:
1541 	return ret;
1542 }
1543 
1544 static noinline int btrfs_ioctl_snap_create(struct file *file,
1545 					    void __user *arg, int subvol)
1546 {
1547 	struct btrfs_ioctl_vol_args *vol_args;
1548 	int ret;
1549 
1550 	vol_args = memdup_user(arg, sizeof(*vol_args));
1551 	if (IS_ERR(vol_args))
1552 		return PTR_ERR(vol_args);
1553 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1554 
1555 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1556 					      vol_args->fd, subvol,
1557 					      NULL, false, NULL);
1558 
1559 	kfree(vol_args);
1560 	return ret;
1561 }
1562 
1563 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1564 					       void __user *arg, int subvol)
1565 {
1566 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1567 	int ret;
1568 	u64 transid = 0;
1569 	u64 *ptr = NULL;
1570 	bool readonly = false;
1571 	struct btrfs_qgroup_inherit *inherit = NULL;
1572 
1573 	vol_args = memdup_user(arg, sizeof(*vol_args));
1574 	if (IS_ERR(vol_args))
1575 		return PTR_ERR(vol_args);
1576 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1577 
1578 	if (vol_args->flags &
1579 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1580 	      BTRFS_SUBVOL_QGROUP_INHERIT)) {
1581 		ret = -EOPNOTSUPP;
1582 		goto out;
1583 	}
1584 
1585 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1586 		ptr = &transid;
1587 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1588 		readonly = true;
1589 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1590 		if (vol_args->size > PAGE_CACHE_SIZE) {
1591 			ret = -EINVAL;
1592 			goto out;
1593 		}
1594 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1595 		if (IS_ERR(inherit)) {
1596 			ret = PTR_ERR(inherit);
1597 			goto out;
1598 		}
1599 	}
1600 
1601 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1602 					      vol_args->fd, subvol, ptr,
1603 					      readonly, inherit);
1604 
1605 	if (ret == 0 && ptr &&
1606 	    copy_to_user(arg +
1607 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1608 				  transid), ptr, sizeof(*ptr)))
1609 		ret = -EFAULT;
1610 out:
1611 	kfree(vol_args);
1612 	kfree(inherit);
1613 	return ret;
1614 }
1615 
1616 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1617 						void __user *arg)
1618 {
1619 	struct inode *inode = file_inode(file);
1620 	struct btrfs_root *root = BTRFS_I(inode)->root;
1621 	int ret = 0;
1622 	u64 flags = 0;
1623 
1624 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1625 		return -EINVAL;
1626 
1627 	down_read(&root->fs_info->subvol_sem);
1628 	if (btrfs_root_readonly(root))
1629 		flags |= BTRFS_SUBVOL_RDONLY;
1630 	up_read(&root->fs_info->subvol_sem);
1631 
1632 	if (copy_to_user(arg, &flags, sizeof(flags)))
1633 		ret = -EFAULT;
1634 
1635 	return ret;
1636 }
1637 
1638 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1639 					      void __user *arg)
1640 {
1641 	struct inode *inode = file_inode(file);
1642 	struct btrfs_root *root = BTRFS_I(inode)->root;
1643 	struct btrfs_trans_handle *trans;
1644 	u64 root_flags;
1645 	u64 flags;
1646 	int ret = 0;
1647 
1648 	ret = mnt_want_write_file(file);
1649 	if (ret)
1650 		goto out;
1651 
1652 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1653 		ret = -EINVAL;
1654 		goto out_drop_write;
1655 	}
1656 
1657 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1658 		ret = -EFAULT;
1659 		goto out_drop_write;
1660 	}
1661 
1662 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1663 		ret = -EINVAL;
1664 		goto out_drop_write;
1665 	}
1666 
1667 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1668 		ret = -EOPNOTSUPP;
1669 		goto out_drop_write;
1670 	}
1671 
1672 	if (!inode_owner_or_capable(inode)) {
1673 		ret = -EACCES;
1674 		goto out_drop_write;
1675 	}
1676 
1677 	down_write(&root->fs_info->subvol_sem);
1678 
1679 	/* nothing to do */
1680 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1681 		goto out_drop_sem;
1682 
1683 	root_flags = btrfs_root_flags(&root->root_item);
1684 	if (flags & BTRFS_SUBVOL_RDONLY)
1685 		btrfs_set_root_flags(&root->root_item,
1686 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1687 	else
1688 		btrfs_set_root_flags(&root->root_item,
1689 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1690 
1691 	trans = btrfs_start_transaction(root, 1);
1692 	if (IS_ERR(trans)) {
1693 		ret = PTR_ERR(trans);
1694 		goto out_reset;
1695 	}
1696 
1697 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1698 				&root->root_key, &root->root_item);
1699 
1700 	btrfs_commit_transaction(trans, root);
1701 out_reset:
1702 	if (ret)
1703 		btrfs_set_root_flags(&root->root_item, root_flags);
1704 out_drop_sem:
1705 	up_write(&root->fs_info->subvol_sem);
1706 out_drop_write:
1707 	mnt_drop_write_file(file);
1708 out:
1709 	return ret;
1710 }
1711 
1712 /*
1713  * helper to check if the subvolume references other subvolumes
1714  */
1715 static noinline int may_destroy_subvol(struct btrfs_root *root)
1716 {
1717 	struct btrfs_path *path;
1718 	struct btrfs_key key;
1719 	int ret;
1720 
1721 	path = btrfs_alloc_path();
1722 	if (!path)
1723 		return -ENOMEM;
1724 
1725 	key.objectid = root->root_key.objectid;
1726 	key.type = BTRFS_ROOT_REF_KEY;
1727 	key.offset = (u64)-1;
1728 
1729 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1730 				&key, path, 0, 0);
1731 	if (ret < 0)
1732 		goto out;
1733 	BUG_ON(ret == 0);
1734 
1735 	ret = 0;
1736 	if (path->slots[0] > 0) {
1737 		path->slots[0]--;
1738 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1739 		if (key.objectid == root->root_key.objectid &&
1740 		    key.type == BTRFS_ROOT_REF_KEY)
1741 			ret = -ENOTEMPTY;
1742 	}
1743 out:
1744 	btrfs_free_path(path);
1745 	return ret;
1746 }
1747 
1748 static noinline int key_in_sk(struct btrfs_key *key,
1749 			      struct btrfs_ioctl_search_key *sk)
1750 {
1751 	struct btrfs_key test;
1752 	int ret;
1753 
1754 	test.objectid = sk->min_objectid;
1755 	test.type = sk->min_type;
1756 	test.offset = sk->min_offset;
1757 
1758 	ret = btrfs_comp_cpu_keys(key, &test);
1759 	if (ret < 0)
1760 		return 0;
1761 
1762 	test.objectid = sk->max_objectid;
1763 	test.type = sk->max_type;
1764 	test.offset = sk->max_offset;
1765 
1766 	ret = btrfs_comp_cpu_keys(key, &test);
1767 	if (ret > 0)
1768 		return 0;
1769 	return 1;
1770 }
1771 
1772 static noinline int copy_to_sk(struct btrfs_root *root,
1773 			       struct btrfs_path *path,
1774 			       struct btrfs_key *key,
1775 			       struct btrfs_ioctl_search_key *sk,
1776 			       char *buf,
1777 			       unsigned long *sk_offset,
1778 			       int *num_found)
1779 {
1780 	u64 found_transid;
1781 	struct extent_buffer *leaf;
1782 	struct btrfs_ioctl_search_header sh;
1783 	unsigned long item_off;
1784 	unsigned long item_len;
1785 	int nritems;
1786 	int i;
1787 	int slot;
1788 	int ret = 0;
1789 
1790 	leaf = path->nodes[0];
1791 	slot = path->slots[0];
1792 	nritems = btrfs_header_nritems(leaf);
1793 
1794 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1795 		i = nritems;
1796 		goto advance_key;
1797 	}
1798 	found_transid = btrfs_header_generation(leaf);
1799 
1800 	for (i = slot; i < nritems; i++) {
1801 		item_off = btrfs_item_ptr_offset(leaf, i);
1802 		item_len = btrfs_item_size_nr(leaf, i);
1803 
1804 		btrfs_item_key_to_cpu(leaf, key, i);
1805 		if (!key_in_sk(key, sk))
1806 			continue;
1807 
1808 		if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1809 			item_len = 0;
1810 
1811 		if (sizeof(sh) + item_len + *sk_offset >
1812 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1813 			ret = 1;
1814 			goto overflow;
1815 		}
1816 
1817 		sh.objectid = key->objectid;
1818 		sh.offset = key->offset;
1819 		sh.type = key->type;
1820 		sh.len = item_len;
1821 		sh.transid = found_transid;
1822 
1823 		/* copy search result header */
1824 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1825 		*sk_offset += sizeof(sh);
1826 
1827 		if (item_len) {
1828 			char *p = buf + *sk_offset;
1829 			/* copy the item */
1830 			read_extent_buffer(leaf, p,
1831 					   item_off, item_len);
1832 			*sk_offset += item_len;
1833 		}
1834 		(*num_found)++;
1835 
1836 		if (*num_found >= sk->nr_items)
1837 			break;
1838 	}
1839 advance_key:
1840 	ret = 0;
1841 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1842 		key->offset++;
1843 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1844 		key->offset = 0;
1845 		key->type++;
1846 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1847 		key->offset = 0;
1848 		key->type = 0;
1849 		key->objectid++;
1850 	} else
1851 		ret = 1;
1852 overflow:
1853 	return ret;
1854 }
1855 
1856 static noinline int search_ioctl(struct inode *inode,
1857 				 struct btrfs_ioctl_search_args *args)
1858 {
1859 	struct btrfs_root *root;
1860 	struct btrfs_key key;
1861 	struct btrfs_key max_key;
1862 	struct btrfs_path *path;
1863 	struct btrfs_ioctl_search_key *sk = &args->key;
1864 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1865 	int ret;
1866 	int num_found = 0;
1867 	unsigned long sk_offset = 0;
1868 
1869 	path = btrfs_alloc_path();
1870 	if (!path)
1871 		return -ENOMEM;
1872 
1873 	if (sk->tree_id == 0) {
1874 		/* search the root of the inode that was passed */
1875 		root = BTRFS_I(inode)->root;
1876 	} else {
1877 		key.objectid = sk->tree_id;
1878 		key.type = BTRFS_ROOT_ITEM_KEY;
1879 		key.offset = (u64)-1;
1880 		root = btrfs_read_fs_root_no_name(info, &key);
1881 		if (IS_ERR(root)) {
1882 			printk(KERN_ERR "could not find root %llu\n",
1883 			       sk->tree_id);
1884 			btrfs_free_path(path);
1885 			return -ENOENT;
1886 		}
1887 	}
1888 
1889 	key.objectid = sk->min_objectid;
1890 	key.type = sk->min_type;
1891 	key.offset = sk->min_offset;
1892 
1893 	max_key.objectid = sk->max_objectid;
1894 	max_key.type = sk->max_type;
1895 	max_key.offset = sk->max_offset;
1896 
1897 	path->keep_locks = 1;
1898 
1899 	while(1) {
1900 		ret = btrfs_search_forward(root, &key, &max_key, path,
1901 					   sk->min_transid);
1902 		if (ret != 0) {
1903 			if (ret > 0)
1904 				ret = 0;
1905 			goto err;
1906 		}
1907 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1908 				 &sk_offset, &num_found);
1909 		btrfs_release_path(path);
1910 		if (ret || num_found >= sk->nr_items)
1911 			break;
1912 
1913 	}
1914 	ret = 0;
1915 err:
1916 	sk->nr_items = num_found;
1917 	btrfs_free_path(path);
1918 	return ret;
1919 }
1920 
1921 static noinline int btrfs_ioctl_tree_search(struct file *file,
1922 					   void __user *argp)
1923 {
1924 	 struct btrfs_ioctl_search_args *args;
1925 	 struct inode *inode;
1926 	 int ret;
1927 
1928 	if (!capable(CAP_SYS_ADMIN))
1929 		return -EPERM;
1930 
1931 	args = memdup_user(argp, sizeof(*args));
1932 	if (IS_ERR(args))
1933 		return PTR_ERR(args);
1934 
1935 	inode = file_inode(file);
1936 	ret = search_ioctl(inode, args);
1937 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1938 		ret = -EFAULT;
1939 	kfree(args);
1940 	return ret;
1941 }
1942 
1943 /*
1944  * Search INODE_REFs to identify path name of 'dirid' directory
1945  * in a 'tree_id' tree. and sets path name to 'name'.
1946  */
1947 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1948 				u64 tree_id, u64 dirid, char *name)
1949 {
1950 	struct btrfs_root *root;
1951 	struct btrfs_key key;
1952 	char *ptr;
1953 	int ret = -1;
1954 	int slot;
1955 	int len;
1956 	int total_len = 0;
1957 	struct btrfs_inode_ref *iref;
1958 	struct extent_buffer *l;
1959 	struct btrfs_path *path;
1960 
1961 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1962 		name[0]='\0';
1963 		return 0;
1964 	}
1965 
1966 	path = btrfs_alloc_path();
1967 	if (!path)
1968 		return -ENOMEM;
1969 
1970 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1971 
1972 	key.objectid = tree_id;
1973 	key.type = BTRFS_ROOT_ITEM_KEY;
1974 	key.offset = (u64)-1;
1975 	root = btrfs_read_fs_root_no_name(info, &key);
1976 	if (IS_ERR(root)) {
1977 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1978 		ret = -ENOENT;
1979 		goto out;
1980 	}
1981 
1982 	key.objectid = dirid;
1983 	key.type = BTRFS_INODE_REF_KEY;
1984 	key.offset = (u64)-1;
1985 
1986 	while(1) {
1987 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1988 		if (ret < 0)
1989 			goto out;
1990 
1991 		l = path->nodes[0];
1992 		slot = path->slots[0];
1993 		if (ret > 0 && slot > 0)
1994 			slot--;
1995 		btrfs_item_key_to_cpu(l, &key, slot);
1996 
1997 		if (ret > 0 && (key.objectid != dirid ||
1998 				key.type != BTRFS_INODE_REF_KEY)) {
1999 			ret = -ENOENT;
2000 			goto out;
2001 		}
2002 
2003 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2004 		len = btrfs_inode_ref_name_len(l, iref);
2005 		ptr -= len + 1;
2006 		total_len += len + 1;
2007 		if (ptr < name)
2008 			goto out;
2009 
2010 		*(ptr + len) = '/';
2011 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2012 
2013 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2014 			break;
2015 
2016 		btrfs_release_path(path);
2017 		key.objectid = key.offset;
2018 		key.offset = (u64)-1;
2019 		dirid = key.objectid;
2020 	}
2021 	if (ptr < name)
2022 		goto out;
2023 	memmove(name, ptr, total_len);
2024 	name[total_len]='\0';
2025 	ret = 0;
2026 out:
2027 	btrfs_free_path(path);
2028 	return ret;
2029 }
2030 
2031 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2032 					   void __user *argp)
2033 {
2034 	 struct btrfs_ioctl_ino_lookup_args *args;
2035 	 struct inode *inode;
2036 	 int ret;
2037 
2038 	if (!capable(CAP_SYS_ADMIN))
2039 		return -EPERM;
2040 
2041 	args = memdup_user(argp, sizeof(*args));
2042 	if (IS_ERR(args))
2043 		return PTR_ERR(args);
2044 
2045 	inode = file_inode(file);
2046 
2047 	if (args->treeid == 0)
2048 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2049 
2050 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2051 					args->treeid, args->objectid,
2052 					args->name);
2053 
2054 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2055 		ret = -EFAULT;
2056 
2057 	kfree(args);
2058 	return ret;
2059 }
2060 
2061 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2062 					     void __user *arg)
2063 {
2064 	struct dentry *parent = fdentry(file);
2065 	struct dentry *dentry;
2066 	struct inode *dir = parent->d_inode;
2067 	struct inode *inode;
2068 	struct btrfs_root *root = BTRFS_I(dir)->root;
2069 	struct btrfs_root *dest = NULL;
2070 	struct btrfs_ioctl_vol_args *vol_args;
2071 	struct btrfs_trans_handle *trans;
2072 	struct btrfs_block_rsv block_rsv;
2073 	u64 qgroup_reserved;
2074 	int namelen;
2075 	int ret;
2076 	int err = 0;
2077 
2078 	vol_args = memdup_user(arg, sizeof(*vol_args));
2079 	if (IS_ERR(vol_args))
2080 		return PTR_ERR(vol_args);
2081 
2082 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2083 	namelen = strlen(vol_args->name);
2084 	if (strchr(vol_args->name, '/') ||
2085 	    strncmp(vol_args->name, "..", namelen) == 0) {
2086 		err = -EINVAL;
2087 		goto out;
2088 	}
2089 
2090 	err = mnt_want_write_file(file);
2091 	if (err)
2092 		goto out;
2093 
2094 	err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2095 	if (err == -EINTR)
2096 		goto out;
2097 	dentry = lookup_one_len(vol_args->name, parent, namelen);
2098 	if (IS_ERR(dentry)) {
2099 		err = PTR_ERR(dentry);
2100 		goto out_unlock_dir;
2101 	}
2102 
2103 	if (!dentry->d_inode) {
2104 		err = -ENOENT;
2105 		goto out_dput;
2106 	}
2107 
2108 	inode = dentry->d_inode;
2109 	dest = BTRFS_I(inode)->root;
2110 	if (!capable(CAP_SYS_ADMIN)){
2111 		/*
2112 		 * Regular user.  Only allow this with a special mount
2113 		 * option, when the user has write+exec access to the
2114 		 * subvol root, and when rmdir(2) would have been
2115 		 * allowed.
2116 		 *
2117 		 * Note that this is _not_ check that the subvol is
2118 		 * empty or doesn't contain data that we wouldn't
2119 		 * otherwise be able to delete.
2120 		 *
2121 		 * Users who want to delete empty subvols should try
2122 		 * rmdir(2).
2123 		 */
2124 		err = -EPERM;
2125 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2126 			goto out_dput;
2127 
2128 		/*
2129 		 * Do not allow deletion if the parent dir is the same
2130 		 * as the dir to be deleted.  That means the ioctl
2131 		 * must be called on the dentry referencing the root
2132 		 * of the subvol, not a random directory contained
2133 		 * within it.
2134 		 */
2135 		err = -EINVAL;
2136 		if (root == dest)
2137 			goto out_dput;
2138 
2139 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2140 		if (err)
2141 			goto out_dput;
2142 	}
2143 
2144 	/* check if subvolume may be deleted by a user */
2145 	err = btrfs_may_delete(dir, dentry, 1);
2146 	if (err)
2147 		goto out_dput;
2148 
2149 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2150 		err = -EINVAL;
2151 		goto out_dput;
2152 	}
2153 
2154 	mutex_lock(&inode->i_mutex);
2155 	err = d_invalidate(dentry);
2156 	if (err)
2157 		goto out_unlock;
2158 
2159 	down_write(&root->fs_info->subvol_sem);
2160 
2161 	err = may_destroy_subvol(dest);
2162 	if (err)
2163 		goto out_up_write;
2164 
2165 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2166 	/*
2167 	 * One for dir inode, two for dir entries, two for root
2168 	 * ref/backref.
2169 	 */
2170 	err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2171 					       5, &qgroup_reserved);
2172 	if (err)
2173 		goto out_up_write;
2174 
2175 	trans = btrfs_start_transaction(root, 0);
2176 	if (IS_ERR(trans)) {
2177 		err = PTR_ERR(trans);
2178 		goto out_release;
2179 	}
2180 	trans->block_rsv = &block_rsv;
2181 	trans->bytes_reserved = block_rsv.size;
2182 
2183 	ret = btrfs_unlink_subvol(trans, root, dir,
2184 				dest->root_key.objectid,
2185 				dentry->d_name.name,
2186 				dentry->d_name.len);
2187 	if (ret) {
2188 		err = ret;
2189 		btrfs_abort_transaction(trans, root, ret);
2190 		goto out_end_trans;
2191 	}
2192 
2193 	btrfs_record_root_in_trans(trans, dest);
2194 
2195 	memset(&dest->root_item.drop_progress, 0,
2196 		sizeof(dest->root_item.drop_progress));
2197 	dest->root_item.drop_level = 0;
2198 	btrfs_set_root_refs(&dest->root_item, 0);
2199 
2200 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2201 		ret = btrfs_insert_orphan_item(trans,
2202 					root->fs_info->tree_root,
2203 					dest->root_key.objectid);
2204 		if (ret) {
2205 			btrfs_abort_transaction(trans, root, ret);
2206 			err = ret;
2207 			goto out_end_trans;
2208 		}
2209 	}
2210 out_end_trans:
2211 	trans->block_rsv = NULL;
2212 	trans->bytes_reserved = 0;
2213 	ret = btrfs_end_transaction(trans, root);
2214 	if (ret && !err)
2215 		err = ret;
2216 	inode->i_flags |= S_DEAD;
2217 out_release:
2218 	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2219 out_up_write:
2220 	up_write(&root->fs_info->subvol_sem);
2221 out_unlock:
2222 	mutex_unlock(&inode->i_mutex);
2223 	if (!err) {
2224 		shrink_dcache_sb(root->fs_info->sb);
2225 		btrfs_invalidate_inodes(dest);
2226 		d_delete(dentry);
2227 
2228 		/* the last ref */
2229 		if (dest->cache_inode) {
2230 			iput(dest->cache_inode);
2231 			dest->cache_inode = NULL;
2232 		}
2233 	}
2234 out_dput:
2235 	dput(dentry);
2236 out_unlock_dir:
2237 	mutex_unlock(&dir->i_mutex);
2238 	mnt_drop_write_file(file);
2239 out:
2240 	kfree(vol_args);
2241 	return err;
2242 }
2243 
2244 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2245 {
2246 	struct inode *inode = file_inode(file);
2247 	struct btrfs_root *root = BTRFS_I(inode)->root;
2248 	struct btrfs_ioctl_defrag_range_args *range;
2249 	int ret;
2250 
2251 	ret = mnt_want_write_file(file);
2252 	if (ret)
2253 		return ret;
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 	mnt_drop_write_file(file);
2310 	return ret;
2311 }
2312 
2313 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2314 {
2315 	struct btrfs_ioctl_vol_args *vol_args;
2316 	int ret;
2317 
2318 	if (!capable(CAP_SYS_ADMIN))
2319 		return -EPERM;
2320 
2321 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2322 			1)) {
2323 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2324 		return -EINVAL;
2325 	}
2326 
2327 	mutex_lock(&root->fs_info->volume_mutex);
2328 	vol_args = memdup_user(arg, sizeof(*vol_args));
2329 	if (IS_ERR(vol_args)) {
2330 		ret = PTR_ERR(vol_args);
2331 		goto out;
2332 	}
2333 
2334 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2335 	ret = btrfs_init_new_device(root, vol_args->name);
2336 
2337 	kfree(vol_args);
2338 out:
2339 	mutex_unlock(&root->fs_info->volume_mutex);
2340 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2341 	return ret;
2342 }
2343 
2344 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2345 {
2346 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2347 	struct btrfs_ioctl_vol_args *vol_args;
2348 	int ret;
2349 
2350 	if (!capable(CAP_SYS_ADMIN))
2351 		return -EPERM;
2352 
2353 	ret = mnt_want_write_file(file);
2354 	if (ret)
2355 		return ret;
2356 
2357 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2358 			1)) {
2359 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2360 		mnt_drop_write_file(file);
2361 		return -EINVAL;
2362 	}
2363 
2364 	mutex_lock(&root->fs_info->volume_mutex);
2365 	vol_args = memdup_user(arg, sizeof(*vol_args));
2366 	if (IS_ERR(vol_args)) {
2367 		ret = PTR_ERR(vol_args);
2368 		goto out;
2369 	}
2370 
2371 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2372 	ret = btrfs_rm_device(root, vol_args->name);
2373 
2374 	kfree(vol_args);
2375 out:
2376 	mutex_unlock(&root->fs_info->volume_mutex);
2377 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2378 	mnt_drop_write_file(file);
2379 	return ret;
2380 }
2381 
2382 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2383 {
2384 	struct btrfs_ioctl_fs_info_args *fi_args;
2385 	struct btrfs_device *device;
2386 	struct btrfs_device *next;
2387 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2388 	int ret = 0;
2389 
2390 	if (!capable(CAP_SYS_ADMIN))
2391 		return -EPERM;
2392 
2393 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2394 	if (!fi_args)
2395 		return -ENOMEM;
2396 
2397 	fi_args->num_devices = fs_devices->num_devices;
2398 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2399 
2400 	mutex_lock(&fs_devices->device_list_mutex);
2401 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2402 		if (device->devid > fi_args->max_id)
2403 			fi_args->max_id = device->devid;
2404 	}
2405 	mutex_unlock(&fs_devices->device_list_mutex);
2406 
2407 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2408 		ret = -EFAULT;
2409 
2410 	kfree(fi_args);
2411 	return ret;
2412 }
2413 
2414 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2415 {
2416 	struct btrfs_ioctl_dev_info_args *di_args;
2417 	struct btrfs_device *dev;
2418 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2419 	int ret = 0;
2420 	char *s_uuid = NULL;
2421 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2422 
2423 	if (!capable(CAP_SYS_ADMIN))
2424 		return -EPERM;
2425 
2426 	di_args = memdup_user(arg, sizeof(*di_args));
2427 	if (IS_ERR(di_args))
2428 		return PTR_ERR(di_args);
2429 
2430 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2431 		s_uuid = di_args->uuid;
2432 
2433 	mutex_lock(&fs_devices->device_list_mutex);
2434 	dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2435 
2436 	if (!dev) {
2437 		ret = -ENODEV;
2438 		goto out;
2439 	}
2440 
2441 	di_args->devid = dev->devid;
2442 	di_args->bytes_used = dev->bytes_used;
2443 	di_args->total_bytes = dev->total_bytes;
2444 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2445 	if (dev->name) {
2446 		struct rcu_string *name;
2447 
2448 		rcu_read_lock();
2449 		name = rcu_dereference(dev->name);
2450 		strncpy(di_args->path, name->str, sizeof(di_args->path));
2451 		rcu_read_unlock();
2452 		di_args->path[sizeof(di_args->path) - 1] = 0;
2453 	} else {
2454 		di_args->path[0] = '\0';
2455 	}
2456 
2457 out:
2458 	mutex_unlock(&fs_devices->device_list_mutex);
2459 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2460 		ret = -EFAULT;
2461 
2462 	kfree(di_args);
2463 	return ret;
2464 }
2465 
2466 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2467 				       u64 off, u64 olen, u64 destoff)
2468 {
2469 	struct inode *inode = file_inode(file);
2470 	struct btrfs_root *root = BTRFS_I(inode)->root;
2471 	struct fd src_file;
2472 	struct inode *src;
2473 	struct btrfs_trans_handle *trans;
2474 	struct btrfs_path *path;
2475 	struct extent_buffer *leaf;
2476 	char *buf;
2477 	struct btrfs_key key;
2478 	u32 nritems;
2479 	int slot;
2480 	int ret;
2481 	u64 len = olen;
2482 	u64 bs = root->fs_info->sb->s_blocksize;
2483 
2484 	/*
2485 	 * TODO:
2486 	 * - split compressed inline extents.  annoying: we need to
2487 	 *   decompress into destination's address_space (the file offset
2488 	 *   may change, so source mapping won't do), then recompress (or
2489 	 *   otherwise reinsert) a subrange.
2490 	 * - allow ranges within the same file to be cloned (provided
2491 	 *   they don't overlap)?
2492 	 */
2493 
2494 	/* the destination must be opened for writing */
2495 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2496 		return -EINVAL;
2497 
2498 	if (btrfs_root_readonly(root))
2499 		return -EROFS;
2500 
2501 	ret = mnt_want_write_file(file);
2502 	if (ret)
2503 		return ret;
2504 
2505 	src_file = fdget(srcfd);
2506 	if (!src_file.file) {
2507 		ret = -EBADF;
2508 		goto out_drop_write;
2509 	}
2510 
2511 	ret = -EXDEV;
2512 	if (src_file.file->f_path.mnt != file->f_path.mnt)
2513 		goto out_fput;
2514 
2515 	src = file_inode(src_file.file);
2516 
2517 	ret = -EINVAL;
2518 	if (src == inode)
2519 		goto out_fput;
2520 
2521 	/* the src must be open for reading */
2522 	if (!(src_file.file->f_mode & FMODE_READ))
2523 		goto out_fput;
2524 
2525 	/* don't make the dst file partly checksummed */
2526 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2527 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2528 		goto out_fput;
2529 
2530 	ret = -EISDIR;
2531 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2532 		goto out_fput;
2533 
2534 	ret = -EXDEV;
2535 	if (src->i_sb != inode->i_sb)
2536 		goto out_fput;
2537 
2538 	ret = -ENOMEM;
2539 	buf = vmalloc(btrfs_level_size(root, 0));
2540 	if (!buf)
2541 		goto out_fput;
2542 
2543 	path = btrfs_alloc_path();
2544 	if (!path) {
2545 		vfree(buf);
2546 		goto out_fput;
2547 	}
2548 	path->reada = 2;
2549 
2550 	if (inode < src) {
2551 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2552 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2553 	} else {
2554 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2555 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2556 	}
2557 
2558 	/* determine range to clone */
2559 	ret = -EINVAL;
2560 	if (off + len > src->i_size || off + len < off)
2561 		goto out_unlock;
2562 	if (len == 0)
2563 		olen = len = src->i_size - off;
2564 	/* if we extend to eof, continue to block boundary */
2565 	if (off + len == src->i_size)
2566 		len = ALIGN(src->i_size, bs) - off;
2567 
2568 	/* verify the end result is block aligned */
2569 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2570 	    !IS_ALIGNED(destoff, bs))
2571 		goto out_unlock;
2572 
2573 	if (destoff > inode->i_size) {
2574 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2575 		if (ret)
2576 			goto out_unlock;
2577 	}
2578 
2579 	/* truncate page cache pages from target inode range */
2580 	truncate_inode_pages_range(&inode->i_data, destoff,
2581 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2582 
2583 	/* do any pending delalloc/csum calc on src, one way or
2584 	   another, and lock file content */
2585 	while (1) {
2586 		struct btrfs_ordered_extent *ordered;
2587 		lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2588 		ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2589 		if (!ordered &&
2590 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2591 				    EXTENT_DELALLOC, 0, NULL))
2592 			break;
2593 		unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2594 		if (ordered)
2595 			btrfs_put_ordered_extent(ordered);
2596 		btrfs_wait_ordered_range(src, off, len);
2597 	}
2598 
2599 	/* clone data */
2600 	key.objectid = btrfs_ino(src);
2601 	key.type = BTRFS_EXTENT_DATA_KEY;
2602 	key.offset = 0;
2603 
2604 	while (1) {
2605 		/*
2606 		 * note the key will change type as we walk through the
2607 		 * tree.
2608 		 */
2609 		ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2610 				0, 0);
2611 		if (ret < 0)
2612 			goto out;
2613 
2614 		nritems = btrfs_header_nritems(path->nodes[0]);
2615 		if (path->slots[0] >= nritems) {
2616 			ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2617 			if (ret < 0)
2618 				goto out;
2619 			if (ret > 0)
2620 				break;
2621 			nritems = btrfs_header_nritems(path->nodes[0]);
2622 		}
2623 		leaf = path->nodes[0];
2624 		slot = path->slots[0];
2625 
2626 		btrfs_item_key_to_cpu(leaf, &key, slot);
2627 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2628 		    key.objectid != btrfs_ino(src))
2629 			break;
2630 
2631 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2632 			struct btrfs_file_extent_item *extent;
2633 			int type;
2634 			u32 size;
2635 			struct btrfs_key new_key;
2636 			u64 disko = 0, diskl = 0;
2637 			u64 datao = 0, datal = 0;
2638 			u8 comp;
2639 			u64 endoff;
2640 
2641 			size = btrfs_item_size_nr(leaf, slot);
2642 			read_extent_buffer(leaf, buf,
2643 					   btrfs_item_ptr_offset(leaf, slot),
2644 					   size);
2645 
2646 			extent = btrfs_item_ptr(leaf, slot,
2647 						struct btrfs_file_extent_item);
2648 			comp = btrfs_file_extent_compression(leaf, extent);
2649 			type = btrfs_file_extent_type(leaf, extent);
2650 			if (type == BTRFS_FILE_EXTENT_REG ||
2651 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2652 				disko = btrfs_file_extent_disk_bytenr(leaf,
2653 								      extent);
2654 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2655 								 extent);
2656 				datao = btrfs_file_extent_offset(leaf, extent);
2657 				datal = btrfs_file_extent_num_bytes(leaf,
2658 								    extent);
2659 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2660 				/* take upper bound, may be compressed */
2661 				datal = btrfs_file_extent_ram_bytes(leaf,
2662 								    extent);
2663 			}
2664 			btrfs_release_path(path);
2665 
2666 			if (key.offset + datal <= off ||
2667 			    key.offset >= off + len - 1)
2668 				goto next;
2669 
2670 			memcpy(&new_key, &key, sizeof(new_key));
2671 			new_key.objectid = btrfs_ino(inode);
2672 			if (off <= key.offset)
2673 				new_key.offset = key.offset + destoff - off;
2674 			else
2675 				new_key.offset = destoff;
2676 
2677 			/*
2678 			 * 1 - adjusting old extent (we may have to split it)
2679 			 * 1 - add new extent
2680 			 * 1 - inode update
2681 			 */
2682 			trans = btrfs_start_transaction(root, 3);
2683 			if (IS_ERR(trans)) {
2684 				ret = PTR_ERR(trans);
2685 				goto out;
2686 			}
2687 
2688 			if (type == BTRFS_FILE_EXTENT_REG ||
2689 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2690 				/*
2691 				 *    a  | --- range to clone ---|  b
2692 				 * | ------------- extent ------------- |
2693 				 */
2694 
2695 				/* substract range b */
2696 				if (key.offset + datal > off + len)
2697 					datal = off + len - key.offset;
2698 
2699 				/* substract range a */
2700 				if (off > key.offset) {
2701 					datao += off - key.offset;
2702 					datal -= off - key.offset;
2703 				}
2704 
2705 				ret = btrfs_drop_extents(trans, root, inode,
2706 							 new_key.offset,
2707 							 new_key.offset + datal,
2708 							 1);
2709 				if (ret) {
2710 					btrfs_abort_transaction(trans, root,
2711 								ret);
2712 					btrfs_end_transaction(trans, root);
2713 					goto out;
2714 				}
2715 
2716 				ret = btrfs_insert_empty_item(trans, root, path,
2717 							      &new_key, size);
2718 				if (ret) {
2719 					btrfs_abort_transaction(trans, root,
2720 								ret);
2721 					btrfs_end_transaction(trans, root);
2722 					goto out;
2723 				}
2724 
2725 				leaf = path->nodes[0];
2726 				slot = path->slots[0];
2727 				write_extent_buffer(leaf, buf,
2728 					    btrfs_item_ptr_offset(leaf, slot),
2729 					    size);
2730 
2731 				extent = btrfs_item_ptr(leaf, slot,
2732 						struct btrfs_file_extent_item);
2733 
2734 				/* disko == 0 means it's a hole */
2735 				if (!disko)
2736 					datao = 0;
2737 
2738 				btrfs_set_file_extent_offset(leaf, extent,
2739 							     datao);
2740 				btrfs_set_file_extent_num_bytes(leaf, extent,
2741 								datal);
2742 				if (disko) {
2743 					inode_add_bytes(inode, datal);
2744 					ret = btrfs_inc_extent_ref(trans, root,
2745 							disko, diskl, 0,
2746 							root->root_key.objectid,
2747 							btrfs_ino(inode),
2748 							new_key.offset - datao,
2749 							0);
2750 					if (ret) {
2751 						btrfs_abort_transaction(trans,
2752 									root,
2753 									ret);
2754 						btrfs_end_transaction(trans,
2755 								      root);
2756 						goto out;
2757 
2758 					}
2759 				}
2760 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2761 				u64 skip = 0;
2762 				u64 trim = 0;
2763 				if (off > key.offset) {
2764 					skip = off - key.offset;
2765 					new_key.offset += skip;
2766 				}
2767 
2768 				if (key.offset + datal > off + len)
2769 					trim = key.offset + datal - (off + len);
2770 
2771 				if (comp && (skip || trim)) {
2772 					ret = -EINVAL;
2773 					btrfs_end_transaction(trans, root);
2774 					goto out;
2775 				}
2776 				size -= skip + trim;
2777 				datal -= skip + trim;
2778 
2779 				ret = btrfs_drop_extents(trans, root, inode,
2780 							 new_key.offset,
2781 							 new_key.offset + datal,
2782 							 1);
2783 				if (ret) {
2784 					btrfs_abort_transaction(trans, root,
2785 								ret);
2786 					btrfs_end_transaction(trans, root);
2787 					goto out;
2788 				}
2789 
2790 				ret = btrfs_insert_empty_item(trans, root, path,
2791 							      &new_key, size);
2792 				if (ret) {
2793 					btrfs_abort_transaction(trans, root,
2794 								ret);
2795 					btrfs_end_transaction(trans, root);
2796 					goto out;
2797 				}
2798 
2799 				if (skip) {
2800 					u32 start =
2801 					  btrfs_file_extent_calc_inline_size(0);
2802 					memmove(buf+start, buf+start+skip,
2803 						datal);
2804 				}
2805 
2806 				leaf = path->nodes[0];
2807 				slot = path->slots[0];
2808 				write_extent_buffer(leaf, buf,
2809 					    btrfs_item_ptr_offset(leaf, slot),
2810 					    size);
2811 				inode_add_bytes(inode, datal);
2812 			}
2813 
2814 			btrfs_mark_buffer_dirty(leaf);
2815 			btrfs_release_path(path);
2816 
2817 			inode_inc_iversion(inode);
2818 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2819 
2820 			/*
2821 			 * we round up to the block size at eof when
2822 			 * determining which extents to clone above,
2823 			 * but shouldn't round up the file size
2824 			 */
2825 			endoff = new_key.offset + datal;
2826 			if (endoff > destoff+olen)
2827 				endoff = destoff+olen;
2828 			if (endoff > inode->i_size)
2829 				btrfs_i_size_write(inode, endoff);
2830 
2831 			ret = btrfs_update_inode(trans, root, inode);
2832 			if (ret) {
2833 				btrfs_abort_transaction(trans, root, ret);
2834 				btrfs_end_transaction(trans, root);
2835 				goto out;
2836 			}
2837 			ret = btrfs_end_transaction(trans, root);
2838 		}
2839 next:
2840 		btrfs_release_path(path);
2841 		key.offset++;
2842 	}
2843 	ret = 0;
2844 out:
2845 	btrfs_release_path(path);
2846 	unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2847 out_unlock:
2848 	mutex_unlock(&src->i_mutex);
2849 	mutex_unlock(&inode->i_mutex);
2850 	vfree(buf);
2851 	btrfs_free_path(path);
2852 out_fput:
2853 	fdput(src_file);
2854 out_drop_write:
2855 	mnt_drop_write_file(file);
2856 	return ret;
2857 }
2858 
2859 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2860 {
2861 	struct btrfs_ioctl_clone_range_args args;
2862 
2863 	if (copy_from_user(&args, argp, sizeof(args)))
2864 		return -EFAULT;
2865 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2866 				 args.src_length, args.dest_offset);
2867 }
2868 
2869 /*
2870  * there are many ways the trans_start and trans_end ioctls can lead
2871  * to deadlocks.  They should only be used by applications that
2872  * basically own the machine, and have a very in depth understanding
2873  * of all the possible deadlocks and enospc problems.
2874  */
2875 static long btrfs_ioctl_trans_start(struct file *file)
2876 {
2877 	struct inode *inode = file_inode(file);
2878 	struct btrfs_root *root = BTRFS_I(inode)->root;
2879 	struct btrfs_trans_handle *trans;
2880 	int ret;
2881 
2882 	ret = -EPERM;
2883 	if (!capable(CAP_SYS_ADMIN))
2884 		goto out;
2885 
2886 	ret = -EINPROGRESS;
2887 	if (file->private_data)
2888 		goto out;
2889 
2890 	ret = -EROFS;
2891 	if (btrfs_root_readonly(root))
2892 		goto out;
2893 
2894 	ret = mnt_want_write_file(file);
2895 	if (ret)
2896 		goto out;
2897 
2898 	atomic_inc(&root->fs_info->open_ioctl_trans);
2899 
2900 	ret = -ENOMEM;
2901 	trans = btrfs_start_ioctl_transaction(root);
2902 	if (IS_ERR(trans))
2903 		goto out_drop;
2904 
2905 	file->private_data = trans;
2906 	return 0;
2907 
2908 out_drop:
2909 	atomic_dec(&root->fs_info->open_ioctl_trans);
2910 	mnt_drop_write_file(file);
2911 out:
2912 	return ret;
2913 }
2914 
2915 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2916 {
2917 	struct inode *inode = file_inode(file);
2918 	struct btrfs_root *root = BTRFS_I(inode)->root;
2919 	struct btrfs_root *new_root;
2920 	struct btrfs_dir_item *di;
2921 	struct btrfs_trans_handle *trans;
2922 	struct btrfs_path *path;
2923 	struct btrfs_key location;
2924 	struct btrfs_disk_key disk_key;
2925 	u64 objectid = 0;
2926 	u64 dir_id;
2927 	int ret;
2928 
2929 	if (!capable(CAP_SYS_ADMIN))
2930 		return -EPERM;
2931 
2932 	ret = mnt_want_write_file(file);
2933 	if (ret)
2934 		return ret;
2935 
2936 	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2937 		ret = -EFAULT;
2938 		goto out;
2939 	}
2940 
2941 	if (!objectid)
2942 		objectid = root->root_key.objectid;
2943 
2944 	location.objectid = objectid;
2945 	location.type = BTRFS_ROOT_ITEM_KEY;
2946 	location.offset = (u64)-1;
2947 
2948 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2949 	if (IS_ERR(new_root)) {
2950 		ret = PTR_ERR(new_root);
2951 		goto out;
2952 	}
2953 
2954 	if (btrfs_root_refs(&new_root->root_item) == 0) {
2955 		ret = -ENOENT;
2956 		goto out;
2957 	}
2958 
2959 	path = btrfs_alloc_path();
2960 	if (!path) {
2961 		ret = -ENOMEM;
2962 		goto out;
2963 	}
2964 	path->leave_spinning = 1;
2965 
2966 	trans = btrfs_start_transaction(root, 1);
2967 	if (IS_ERR(trans)) {
2968 		btrfs_free_path(path);
2969 		ret = PTR_ERR(trans);
2970 		goto out;
2971 	}
2972 
2973 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2974 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2975 				   dir_id, "default", 7, 1);
2976 	if (IS_ERR_OR_NULL(di)) {
2977 		btrfs_free_path(path);
2978 		btrfs_end_transaction(trans, root);
2979 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2980 		       "this isn't going to work\n");
2981 		ret = -ENOENT;
2982 		goto out;
2983 	}
2984 
2985 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2986 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2987 	btrfs_mark_buffer_dirty(path->nodes[0]);
2988 	btrfs_free_path(path);
2989 
2990 	btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2991 	btrfs_end_transaction(trans, root);
2992 out:
2993 	mnt_drop_write_file(file);
2994 	return ret;
2995 }
2996 
2997 void btrfs_get_block_group_info(struct list_head *groups_list,
2998 				struct btrfs_ioctl_space_info *space)
2999 {
3000 	struct btrfs_block_group_cache *block_group;
3001 
3002 	space->total_bytes = 0;
3003 	space->used_bytes = 0;
3004 	space->flags = 0;
3005 	list_for_each_entry(block_group, groups_list, list) {
3006 		space->flags = block_group->flags;
3007 		space->total_bytes += block_group->key.offset;
3008 		space->used_bytes +=
3009 			btrfs_block_group_used(&block_group->item);
3010 	}
3011 }
3012 
3013 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3014 {
3015 	struct btrfs_ioctl_space_args space_args;
3016 	struct btrfs_ioctl_space_info space;
3017 	struct btrfs_ioctl_space_info *dest;
3018 	struct btrfs_ioctl_space_info *dest_orig;
3019 	struct btrfs_ioctl_space_info __user *user_dest;
3020 	struct btrfs_space_info *info;
3021 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3022 		       BTRFS_BLOCK_GROUP_SYSTEM,
3023 		       BTRFS_BLOCK_GROUP_METADATA,
3024 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3025 	int num_types = 4;
3026 	int alloc_size;
3027 	int ret = 0;
3028 	u64 slot_count = 0;
3029 	int i, c;
3030 
3031 	if (copy_from_user(&space_args,
3032 			   (struct btrfs_ioctl_space_args __user *)arg,
3033 			   sizeof(space_args)))
3034 		return -EFAULT;
3035 
3036 	for (i = 0; i < num_types; i++) {
3037 		struct btrfs_space_info *tmp;
3038 
3039 		info = NULL;
3040 		rcu_read_lock();
3041 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3042 					list) {
3043 			if (tmp->flags == types[i]) {
3044 				info = tmp;
3045 				break;
3046 			}
3047 		}
3048 		rcu_read_unlock();
3049 
3050 		if (!info)
3051 			continue;
3052 
3053 		down_read(&info->groups_sem);
3054 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3055 			if (!list_empty(&info->block_groups[c]))
3056 				slot_count++;
3057 		}
3058 		up_read(&info->groups_sem);
3059 	}
3060 
3061 	/* space_slots == 0 means they are asking for a count */
3062 	if (space_args.space_slots == 0) {
3063 		space_args.total_spaces = slot_count;
3064 		goto out;
3065 	}
3066 
3067 	slot_count = min_t(u64, space_args.space_slots, slot_count);
3068 
3069 	alloc_size = sizeof(*dest) * slot_count;
3070 
3071 	/* we generally have at most 6 or so space infos, one for each raid
3072 	 * level.  So, a whole page should be more than enough for everyone
3073 	 */
3074 	if (alloc_size > PAGE_CACHE_SIZE)
3075 		return -ENOMEM;
3076 
3077 	space_args.total_spaces = 0;
3078 	dest = kmalloc(alloc_size, GFP_NOFS);
3079 	if (!dest)
3080 		return -ENOMEM;
3081 	dest_orig = dest;
3082 
3083 	/* now we have a buffer to copy into */
3084 	for (i = 0; i < num_types; i++) {
3085 		struct btrfs_space_info *tmp;
3086 
3087 		if (!slot_count)
3088 			break;
3089 
3090 		info = NULL;
3091 		rcu_read_lock();
3092 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3093 					list) {
3094 			if (tmp->flags == types[i]) {
3095 				info = tmp;
3096 				break;
3097 			}
3098 		}
3099 		rcu_read_unlock();
3100 
3101 		if (!info)
3102 			continue;
3103 		down_read(&info->groups_sem);
3104 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3105 			if (!list_empty(&info->block_groups[c])) {
3106 				btrfs_get_block_group_info(
3107 					&info->block_groups[c], &space);
3108 				memcpy(dest, &space, sizeof(space));
3109 				dest++;
3110 				space_args.total_spaces++;
3111 				slot_count--;
3112 			}
3113 			if (!slot_count)
3114 				break;
3115 		}
3116 		up_read(&info->groups_sem);
3117 	}
3118 
3119 	user_dest = (struct btrfs_ioctl_space_info __user *)
3120 		(arg + sizeof(struct btrfs_ioctl_space_args));
3121 
3122 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3123 		ret = -EFAULT;
3124 
3125 	kfree(dest_orig);
3126 out:
3127 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3128 		ret = -EFAULT;
3129 
3130 	return ret;
3131 }
3132 
3133 /*
3134  * there are many ways the trans_start and trans_end ioctls can lead
3135  * to deadlocks.  They should only be used by applications that
3136  * basically own the machine, and have a very in depth understanding
3137  * of all the possible deadlocks and enospc problems.
3138  */
3139 long btrfs_ioctl_trans_end(struct file *file)
3140 {
3141 	struct inode *inode = file_inode(file);
3142 	struct btrfs_root *root = BTRFS_I(inode)->root;
3143 	struct btrfs_trans_handle *trans;
3144 
3145 	trans = file->private_data;
3146 	if (!trans)
3147 		return -EINVAL;
3148 	file->private_data = NULL;
3149 
3150 	btrfs_end_transaction(trans, root);
3151 
3152 	atomic_dec(&root->fs_info->open_ioctl_trans);
3153 
3154 	mnt_drop_write_file(file);
3155 	return 0;
3156 }
3157 
3158 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3159 					    void __user *argp)
3160 {
3161 	struct btrfs_trans_handle *trans;
3162 	u64 transid;
3163 	int ret;
3164 
3165 	trans = btrfs_attach_transaction_barrier(root);
3166 	if (IS_ERR(trans)) {
3167 		if (PTR_ERR(trans) != -ENOENT)
3168 			return PTR_ERR(trans);
3169 
3170 		/* No running transaction, don't bother */
3171 		transid = root->fs_info->last_trans_committed;
3172 		goto out;
3173 	}
3174 	transid = trans->transid;
3175 	ret = btrfs_commit_transaction_async(trans, root, 0);
3176 	if (ret) {
3177 		btrfs_end_transaction(trans, root);
3178 		return ret;
3179 	}
3180 out:
3181 	if (argp)
3182 		if (copy_to_user(argp, &transid, sizeof(transid)))
3183 			return -EFAULT;
3184 	return 0;
3185 }
3186 
3187 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3188 					   void __user *argp)
3189 {
3190 	u64 transid;
3191 
3192 	if (argp) {
3193 		if (copy_from_user(&transid, argp, sizeof(transid)))
3194 			return -EFAULT;
3195 	} else {
3196 		transid = 0;  /* current trans */
3197 	}
3198 	return btrfs_wait_for_commit(root, transid);
3199 }
3200 
3201 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3202 {
3203 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3204 	struct btrfs_ioctl_scrub_args *sa;
3205 	int ret;
3206 
3207 	if (!capable(CAP_SYS_ADMIN))
3208 		return -EPERM;
3209 
3210 	sa = memdup_user(arg, sizeof(*sa));
3211 	if (IS_ERR(sa))
3212 		return PTR_ERR(sa);
3213 
3214 	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3215 		ret = mnt_want_write_file(file);
3216 		if (ret)
3217 			goto out;
3218 	}
3219 
3220 	ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3221 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3222 			      0);
3223 
3224 	if (copy_to_user(arg, sa, sizeof(*sa)))
3225 		ret = -EFAULT;
3226 
3227 	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3228 		mnt_drop_write_file(file);
3229 out:
3230 	kfree(sa);
3231 	return ret;
3232 }
3233 
3234 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3235 {
3236 	if (!capable(CAP_SYS_ADMIN))
3237 		return -EPERM;
3238 
3239 	return btrfs_scrub_cancel(root->fs_info);
3240 }
3241 
3242 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3243 				       void __user *arg)
3244 {
3245 	struct btrfs_ioctl_scrub_args *sa;
3246 	int ret;
3247 
3248 	if (!capable(CAP_SYS_ADMIN))
3249 		return -EPERM;
3250 
3251 	sa = memdup_user(arg, sizeof(*sa));
3252 	if (IS_ERR(sa))
3253 		return PTR_ERR(sa);
3254 
3255 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3256 
3257 	if (copy_to_user(arg, sa, sizeof(*sa)))
3258 		ret = -EFAULT;
3259 
3260 	kfree(sa);
3261 	return ret;
3262 }
3263 
3264 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3265 				      void __user *arg)
3266 {
3267 	struct btrfs_ioctl_get_dev_stats *sa;
3268 	int ret;
3269 
3270 	sa = memdup_user(arg, sizeof(*sa));
3271 	if (IS_ERR(sa))
3272 		return PTR_ERR(sa);
3273 
3274 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3275 		kfree(sa);
3276 		return -EPERM;
3277 	}
3278 
3279 	ret = btrfs_get_dev_stats(root, sa);
3280 
3281 	if (copy_to_user(arg, sa, sizeof(*sa)))
3282 		ret = -EFAULT;
3283 
3284 	kfree(sa);
3285 	return ret;
3286 }
3287 
3288 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3289 {
3290 	struct btrfs_ioctl_dev_replace_args *p;
3291 	int ret;
3292 
3293 	if (!capable(CAP_SYS_ADMIN))
3294 		return -EPERM;
3295 
3296 	p = memdup_user(arg, sizeof(*p));
3297 	if (IS_ERR(p))
3298 		return PTR_ERR(p);
3299 
3300 	switch (p->cmd) {
3301 	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3302 		if (atomic_xchg(
3303 			&root->fs_info->mutually_exclusive_operation_running,
3304 			1)) {
3305 			pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3306 			ret = -EINPROGRESS;
3307 		} else {
3308 			ret = btrfs_dev_replace_start(root, p);
3309 			atomic_set(
3310 			 &root->fs_info->mutually_exclusive_operation_running,
3311 			 0);
3312 		}
3313 		break;
3314 	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3315 		btrfs_dev_replace_status(root->fs_info, p);
3316 		ret = 0;
3317 		break;
3318 	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3319 		ret = btrfs_dev_replace_cancel(root->fs_info, p);
3320 		break;
3321 	default:
3322 		ret = -EINVAL;
3323 		break;
3324 	}
3325 
3326 	if (copy_to_user(arg, p, sizeof(*p)))
3327 		ret = -EFAULT;
3328 
3329 	kfree(p);
3330 	return ret;
3331 }
3332 
3333 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3334 {
3335 	int ret = 0;
3336 	int i;
3337 	u64 rel_ptr;
3338 	int size;
3339 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3340 	struct inode_fs_paths *ipath = NULL;
3341 	struct btrfs_path *path;
3342 
3343 	if (!capable(CAP_DAC_READ_SEARCH))
3344 		return -EPERM;
3345 
3346 	path = btrfs_alloc_path();
3347 	if (!path) {
3348 		ret = -ENOMEM;
3349 		goto out;
3350 	}
3351 
3352 	ipa = memdup_user(arg, sizeof(*ipa));
3353 	if (IS_ERR(ipa)) {
3354 		ret = PTR_ERR(ipa);
3355 		ipa = NULL;
3356 		goto out;
3357 	}
3358 
3359 	size = min_t(u32, ipa->size, 4096);
3360 	ipath = init_ipath(size, root, path);
3361 	if (IS_ERR(ipath)) {
3362 		ret = PTR_ERR(ipath);
3363 		ipath = NULL;
3364 		goto out;
3365 	}
3366 
3367 	ret = paths_from_inode(ipa->inum, ipath);
3368 	if (ret < 0)
3369 		goto out;
3370 
3371 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3372 		rel_ptr = ipath->fspath->val[i] -
3373 			  (u64)(unsigned long)ipath->fspath->val;
3374 		ipath->fspath->val[i] = rel_ptr;
3375 	}
3376 
3377 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3378 			   (void *)(unsigned long)ipath->fspath, size);
3379 	if (ret) {
3380 		ret = -EFAULT;
3381 		goto out;
3382 	}
3383 
3384 out:
3385 	btrfs_free_path(path);
3386 	free_ipath(ipath);
3387 	kfree(ipa);
3388 
3389 	return ret;
3390 }
3391 
3392 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3393 {
3394 	struct btrfs_data_container *inodes = ctx;
3395 	const size_t c = 3 * sizeof(u64);
3396 
3397 	if (inodes->bytes_left >= c) {
3398 		inodes->bytes_left -= c;
3399 		inodes->val[inodes->elem_cnt] = inum;
3400 		inodes->val[inodes->elem_cnt + 1] = offset;
3401 		inodes->val[inodes->elem_cnt + 2] = root;
3402 		inodes->elem_cnt += 3;
3403 	} else {
3404 		inodes->bytes_missing += c - inodes->bytes_left;
3405 		inodes->bytes_left = 0;
3406 		inodes->elem_missed += 3;
3407 	}
3408 
3409 	return 0;
3410 }
3411 
3412 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3413 					void __user *arg)
3414 {
3415 	int ret = 0;
3416 	int size;
3417 	struct btrfs_ioctl_logical_ino_args *loi;
3418 	struct btrfs_data_container *inodes = NULL;
3419 	struct btrfs_path *path = NULL;
3420 
3421 	if (!capable(CAP_SYS_ADMIN))
3422 		return -EPERM;
3423 
3424 	loi = memdup_user(arg, sizeof(*loi));
3425 	if (IS_ERR(loi)) {
3426 		ret = PTR_ERR(loi);
3427 		loi = NULL;
3428 		goto out;
3429 	}
3430 
3431 	path = btrfs_alloc_path();
3432 	if (!path) {
3433 		ret = -ENOMEM;
3434 		goto out;
3435 	}
3436 
3437 	size = min_t(u32, loi->size, 64 * 1024);
3438 	inodes = init_data_container(size);
3439 	if (IS_ERR(inodes)) {
3440 		ret = PTR_ERR(inodes);
3441 		inodes = NULL;
3442 		goto out;
3443 	}
3444 
3445 	ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3446 					  build_ino_list, inodes);
3447 	if (ret == -EINVAL)
3448 		ret = -ENOENT;
3449 	if (ret < 0)
3450 		goto out;
3451 
3452 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3453 			   (void *)(unsigned long)inodes, size);
3454 	if (ret)
3455 		ret = -EFAULT;
3456 
3457 out:
3458 	btrfs_free_path(path);
3459 	vfree(inodes);
3460 	kfree(loi);
3461 
3462 	return ret;
3463 }
3464 
3465 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3466 			       struct btrfs_ioctl_balance_args *bargs)
3467 {
3468 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3469 
3470 	bargs->flags = bctl->flags;
3471 
3472 	if (atomic_read(&fs_info->balance_running))
3473 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3474 	if (atomic_read(&fs_info->balance_pause_req))
3475 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3476 	if (atomic_read(&fs_info->balance_cancel_req))
3477 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3478 
3479 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3480 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3481 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3482 
3483 	if (lock) {
3484 		spin_lock(&fs_info->balance_lock);
3485 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3486 		spin_unlock(&fs_info->balance_lock);
3487 	} else {
3488 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3489 	}
3490 }
3491 
3492 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3493 {
3494 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3495 	struct btrfs_fs_info *fs_info = root->fs_info;
3496 	struct btrfs_ioctl_balance_args *bargs;
3497 	struct btrfs_balance_control *bctl;
3498 	bool need_unlock; /* for mut. excl. ops lock */
3499 	int ret;
3500 
3501 	if (!capable(CAP_SYS_ADMIN))
3502 		return -EPERM;
3503 
3504 	ret = mnt_want_write_file(file);
3505 	if (ret)
3506 		return ret;
3507 
3508 again:
3509 	if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3510 		mutex_lock(&fs_info->volume_mutex);
3511 		mutex_lock(&fs_info->balance_mutex);
3512 		need_unlock = true;
3513 		goto locked;
3514 	}
3515 
3516 	/*
3517 	 * mut. excl. ops lock is locked.  Three possibilites:
3518 	 *   (1) some other op is running
3519 	 *   (2) balance is running
3520 	 *   (3) balance is paused -- special case (think resume)
3521 	 */
3522 	mutex_lock(&fs_info->balance_mutex);
3523 	if (fs_info->balance_ctl) {
3524 		/* this is either (2) or (3) */
3525 		if (!atomic_read(&fs_info->balance_running)) {
3526 			mutex_unlock(&fs_info->balance_mutex);
3527 			if (!mutex_trylock(&fs_info->volume_mutex))
3528 				goto again;
3529 			mutex_lock(&fs_info->balance_mutex);
3530 
3531 			if (fs_info->balance_ctl &&
3532 			    !atomic_read(&fs_info->balance_running)) {
3533 				/* this is (3) */
3534 				need_unlock = false;
3535 				goto locked;
3536 			}
3537 
3538 			mutex_unlock(&fs_info->balance_mutex);
3539 			mutex_unlock(&fs_info->volume_mutex);
3540 			goto again;
3541 		} else {
3542 			/* this is (2) */
3543 			mutex_unlock(&fs_info->balance_mutex);
3544 			ret = -EINPROGRESS;
3545 			goto out;
3546 		}
3547 	} else {
3548 		/* this is (1) */
3549 		mutex_unlock(&fs_info->balance_mutex);
3550 		pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3551 		ret = -EINVAL;
3552 		goto out;
3553 	}
3554 
3555 locked:
3556 	BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3557 
3558 	if (arg) {
3559 		bargs = memdup_user(arg, sizeof(*bargs));
3560 		if (IS_ERR(bargs)) {
3561 			ret = PTR_ERR(bargs);
3562 			goto out_unlock;
3563 		}
3564 
3565 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3566 			if (!fs_info->balance_ctl) {
3567 				ret = -ENOTCONN;
3568 				goto out_bargs;
3569 			}
3570 
3571 			bctl = fs_info->balance_ctl;
3572 			spin_lock(&fs_info->balance_lock);
3573 			bctl->flags |= BTRFS_BALANCE_RESUME;
3574 			spin_unlock(&fs_info->balance_lock);
3575 
3576 			goto do_balance;
3577 		}
3578 	} else {
3579 		bargs = NULL;
3580 	}
3581 
3582 	if (fs_info->balance_ctl) {
3583 		ret = -EINPROGRESS;
3584 		goto out_bargs;
3585 	}
3586 
3587 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3588 	if (!bctl) {
3589 		ret = -ENOMEM;
3590 		goto out_bargs;
3591 	}
3592 
3593 	bctl->fs_info = fs_info;
3594 	if (arg) {
3595 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3596 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3597 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3598 
3599 		bctl->flags = bargs->flags;
3600 	} else {
3601 		/* balance everything - no filters */
3602 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3603 	}
3604 
3605 do_balance:
3606 	/*
3607 	 * Ownership of bctl and mutually_exclusive_operation_running
3608 	 * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
3609 	 * or, if restriper was paused all the way until unmount, in
3610 	 * free_fs_info.  mutually_exclusive_operation_running is
3611 	 * cleared in __cancel_balance.
3612 	 */
3613 	need_unlock = false;
3614 
3615 	ret = btrfs_balance(bctl, bargs);
3616 
3617 	if (arg) {
3618 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3619 			ret = -EFAULT;
3620 	}
3621 
3622 out_bargs:
3623 	kfree(bargs);
3624 out_unlock:
3625 	mutex_unlock(&fs_info->balance_mutex);
3626 	mutex_unlock(&fs_info->volume_mutex);
3627 	if (need_unlock)
3628 		atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3629 out:
3630 	mnt_drop_write_file(file);
3631 	return ret;
3632 }
3633 
3634 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3635 {
3636 	if (!capable(CAP_SYS_ADMIN))
3637 		return -EPERM;
3638 
3639 	switch (cmd) {
3640 	case BTRFS_BALANCE_CTL_PAUSE:
3641 		return btrfs_pause_balance(root->fs_info);
3642 	case BTRFS_BALANCE_CTL_CANCEL:
3643 		return btrfs_cancel_balance(root->fs_info);
3644 	}
3645 
3646 	return -EINVAL;
3647 }
3648 
3649 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3650 					 void __user *arg)
3651 {
3652 	struct btrfs_fs_info *fs_info = root->fs_info;
3653 	struct btrfs_ioctl_balance_args *bargs;
3654 	int ret = 0;
3655 
3656 	if (!capable(CAP_SYS_ADMIN))
3657 		return -EPERM;
3658 
3659 	mutex_lock(&fs_info->balance_mutex);
3660 	if (!fs_info->balance_ctl) {
3661 		ret = -ENOTCONN;
3662 		goto out;
3663 	}
3664 
3665 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3666 	if (!bargs) {
3667 		ret = -ENOMEM;
3668 		goto out;
3669 	}
3670 
3671 	update_ioctl_balance_args(fs_info, 1, bargs);
3672 
3673 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3674 		ret = -EFAULT;
3675 
3676 	kfree(bargs);
3677 out:
3678 	mutex_unlock(&fs_info->balance_mutex);
3679 	return ret;
3680 }
3681 
3682 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3683 {
3684 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3685 	struct btrfs_ioctl_quota_ctl_args *sa;
3686 	struct btrfs_trans_handle *trans = NULL;
3687 	int ret;
3688 	int err;
3689 
3690 	if (!capable(CAP_SYS_ADMIN))
3691 		return -EPERM;
3692 
3693 	ret = mnt_want_write_file(file);
3694 	if (ret)
3695 		return ret;
3696 
3697 	sa = memdup_user(arg, sizeof(*sa));
3698 	if (IS_ERR(sa)) {
3699 		ret = PTR_ERR(sa);
3700 		goto drop_write;
3701 	}
3702 
3703 	down_write(&root->fs_info->subvol_sem);
3704 	trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
3705 	if (IS_ERR(trans)) {
3706 		ret = PTR_ERR(trans);
3707 		goto out;
3708 	}
3709 
3710 	switch (sa->cmd) {
3711 	case BTRFS_QUOTA_CTL_ENABLE:
3712 		ret = btrfs_quota_enable(trans, root->fs_info);
3713 		break;
3714 	case BTRFS_QUOTA_CTL_DISABLE:
3715 		ret = btrfs_quota_disable(trans, root->fs_info);
3716 		break;
3717 	default:
3718 		ret = -EINVAL;
3719 		break;
3720 	}
3721 
3722 	if (copy_to_user(arg, sa, sizeof(*sa)))
3723 		ret = -EFAULT;
3724 
3725 	err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
3726 	if (err && !ret)
3727 		ret = err;
3728 out:
3729 	kfree(sa);
3730 	up_write(&root->fs_info->subvol_sem);
3731 drop_write:
3732 	mnt_drop_write_file(file);
3733 	return ret;
3734 }
3735 
3736 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3737 {
3738 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3739 	struct btrfs_ioctl_qgroup_assign_args *sa;
3740 	struct btrfs_trans_handle *trans;
3741 	int ret;
3742 	int err;
3743 
3744 	if (!capable(CAP_SYS_ADMIN))
3745 		return -EPERM;
3746 
3747 	ret = mnt_want_write_file(file);
3748 	if (ret)
3749 		return ret;
3750 
3751 	sa = memdup_user(arg, sizeof(*sa));
3752 	if (IS_ERR(sa)) {
3753 		ret = PTR_ERR(sa);
3754 		goto drop_write;
3755 	}
3756 
3757 	trans = btrfs_join_transaction(root);
3758 	if (IS_ERR(trans)) {
3759 		ret = PTR_ERR(trans);
3760 		goto out;
3761 	}
3762 
3763 	/* FIXME: check if the IDs really exist */
3764 	if (sa->assign) {
3765 		ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3766 						sa->src, sa->dst);
3767 	} else {
3768 		ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3769 						sa->src, sa->dst);
3770 	}
3771 
3772 	err = btrfs_end_transaction(trans, root);
3773 	if (err && !ret)
3774 		ret = err;
3775 
3776 out:
3777 	kfree(sa);
3778 drop_write:
3779 	mnt_drop_write_file(file);
3780 	return ret;
3781 }
3782 
3783 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3784 {
3785 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3786 	struct btrfs_ioctl_qgroup_create_args *sa;
3787 	struct btrfs_trans_handle *trans;
3788 	int ret;
3789 	int err;
3790 
3791 	if (!capable(CAP_SYS_ADMIN))
3792 		return -EPERM;
3793 
3794 	ret = mnt_want_write_file(file);
3795 	if (ret)
3796 		return ret;
3797 
3798 	sa = memdup_user(arg, sizeof(*sa));
3799 	if (IS_ERR(sa)) {
3800 		ret = PTR_ERR(sa);
3801 		goto drop_write;
3802 	}
3803 
3804 	if (!sa->qgroupid) {
3805 		ret = -EINVAL;
3806 		goto out;
3807 	}
3808 
3809 	trans = btrfs_join_transaction(root);
3810 	if (IS_ERR(trans)) {
3811 		ret = PTR_ERR(trans);
3812 		goto out;
3813 	}
3814 
3815 	/* FIXME: check if the IDs really exist */
3816 	if (sa->create) {
3817 		ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3818 					  NULL);
3819 	} else {
3820 		ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3821 	}
3822 
3823 	err = btrfs_end_transaction(trans, root);
3824 	if (err && !ret)
3825 		ret = err;
3826 
3827 out:
3828 	kfree(sa);
3829 drop_write:
3830 	mnt_drop_write_file(file);
3831 	return ret;
3832 }
3833 
3834 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3835 {
3836 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3837 	struct btrfs_ioctl_qgroup_limit_args *sa;
3838 	struct btrfs_trans_handle *trans;
3839 	int ret;
3840 	int err;
3841 	u64 qgroupid;
3842 
3843 	if (!capable(CAP_SYS_ADMIN))
3844 		return -EPERM;
3845 
3846 	ret = mnt_want_write_file(file);
3847 	if (ret)
3848 		return ret;
3849 
3850 	sa = memdup_user(arg, sizeof(*sa));
3851 	if (IS_ERR(sa)) {
3852 		ret = PTR_ERR(sa);
3853 		goto drop_write;
3854 	}
3855 
3856 	trans = btrfs_join_transaction(root);
3857 	if (IS_ERR(trans)) {
3858 		ret = PTR_ERR(trans);
3859 		goto out;
3860 	}
3861 
3862 	qgroupid = sa->qgroupid;
3863 	if (!qgroupid) {
3864 		/* take the current subvol as qgroup */
3865 		qgroupid = root->root_key.objectid;
3866 	}
3867 
3868 	/* FIXME: check if the IDs really exist */
3869 	ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3870 
3871 	err = btrfs_end_transaction(trans, root);
3872 	if (err && !ret)
3873 		ret = err;
3874 
3875 out:
3876 	kfree(sa);
3877 drop_write:
3878 	mnt_drop_write_file(file);
3879 	return ret;
3880 }
3881 
3882 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3883 {
3884 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3885 	struct btrfs_ioctl_quota_rescan_args *qsa;
3886 	int ret;
3887 
3888 	if (!capable(CAP_SYS_ADMIN))
3889 		return -EPERM;
3890 
3891 	ret = mnt_want_write_file(file);
3892 	if (ret)
3893 		return ret;
3894 
3895 	qsa = memdup_user(arg, sizeof(*qsa));
3896 	if (IS_ERR(qsa)) {
3897 		ret = PTR_ERR(qsa);
3898 		goto drop_write;
3899 	}
3900 
3901 	if (qsa->flags) {
3902 		ret = -EINVAL;
3903 		goto out;
3904 	}
3905 
3906 	ret = btrfs_qgroup_rescan(root->fs_info);
3907 
3908 out:
3909 	kfree(qsa);
3910 drop_write:
3911 	mnt_drop_write_file(file);
3912 	return ret;
3913 }
3914 
3915 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
3916 {
3917 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3918 	struct btrfs_ioctl_quota_rescan_args *qsa;
3919 	int ret = 0;
3920 
3921 	if (!capable(CAP_SYS_ADMIN))
3922 		return -EPERM;
3923 
3924 	qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
3925 	if (!qsa)
3926 		return -ENOMEM;
3927 
3928 	if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3929 		qsa->flags = 1;
3930 		qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
3931 	}
3932 
3933 	if (copy_to_user(arg, qsa, sizeof(*qsa)))
3934 		ret = -EFAULT;
3935 
3936 	kfree(qsa);
3937 	return ret;
3938 }
3939 
3940 static long btrfs_ioctl_set_received_subvol(struct file *file,
3941 					    void __user *arg)
3942 {
3943 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
3944 	struct inode *inode = file_inode(file);
3945 	struct btrfs_root *root = BTRFS_I(inode)->root;
3946 	struct btrfs_root_item *root_item = &root->root_item;
3947 	struct btrfs_trans_handle *trans;
3948 	struct timespec ct = CURRENT_TIME;
3949 	int ret = 0;
3950 
3951 	ret = mnt_want_write_file(file);
3952 	if (ret < 0)
3953 		return ret;
3954 
3955 	down_write(&root->fs_info->subvol_sem);
3956 
3957 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3958 		ret = -EINVAL;
3959 		goto out;
3960 	}
3961 
3962 	if (btrfs_root_readonly(root)) {
3963 		ret = -EROFS;
3964 		goto out;
3965 	}
3966 
3967 	if (!inode_owner_or_capable(inode)) {
3968 		ret = -EACCES;
3969 		goto out;
3970 	}
3971 
3972 	sa = memdup_user(arg, sizeof(*sa));
3973 	if (IS_ERR(sa)) {
3974 		ret = PTR_ERR(sa);
3975 		sa = NULL;
3976 		goto out;
3977 	}
3978 
3979 	trans = btrfs_start_transaction(root, 1);
3980 	if (IS_ERR(trans)) {
3981 		ret = PTR_ERR(trans);
3982 		trans = NULL;
3983 		goto out;
3984 	}
3985 
3986 	sa->rtransid = trans->transid;
3987 	sa->rtime.sec = ct.tv_sec;
3988 	sa->rtime.nsec = ct.tv_nsec;
3989 
3990 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3991 	btrfs_set_root_stransid(root_item, sa->stransid);
3992 	btrfs_set_root_rtransid(root_item, sa->rtransid);
3993 	root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3994 	root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3995 	root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3996 	root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3997 
3998 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
3999 				&root->root_key, &root->root_item);
4000 	if (ret < 0) {
4001 		btrfs_end_transaction(trans, root);
4002 		trans = NULL;
4003 		goto out;
4004 	} else {
4005 		ret = btrfs_commit_transaction(trans, root);
4006 		if (ret < 0)
4007 			goto out;
4008 	}
4009 
4010 	ret = copy_to_user(arg, sa, sizeof(*sa));
4011 	if (ret)
4012 		ret = -EFAULT;
4013 
4014 out:
4015 	kfree(sa);
4016 	up_write(&root->fs_info->subvol_sem);
4017 	mnt_drop_write_file(file);
4018 	return ret;
4019 }
4020 
4021 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4022 {
4023 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4024 	const char *label = root->fs_info->super_copy->label;
4025 	size_t len = strnlen(label, BTRFS_LABEL_SIZE);
4026 	int ret;
4027 
4028 	if (len == BTRFS_LABEL_SIZE) {
4029 		pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4030 			--len);
4031 	}
4032 
4033 	mutex_lock(&root->fs_info->volume_mutex);
4034 	ret = copy_to_user(arg, label, len);
4035 	mutex_unlock(&root->fs_info->volume_mutex);
4036 
4037 	return ret ? -EFAULT : 0;
4038 }
4039 
4040 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4041 {
4042 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4043 	struct btrfs_super_block *super_block = root->fs_info->super_copy;
4044 	struct btrfs_trans_handle *trans;
4045 	char label[BTRFS_LABEL_SIZE];
4046 	int ret;
4047 
4048 	if (!capable(CAP_SYS_ADMIN))
4049 		return -EPERM;
4050 
4051 	if (copy_from_user(label, arg, sizeof(label)))
4052 		return -EFAULT;
4053 
4054 	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4055 		pr_err("btrfs: unable to set label with more than %d bytes\n",
4056 		       BTRFS_LABEL_SIZE - 1);
4057 		return -EINVAL;
4058 	}
4059 
4060 	ret = mnt_want_write_file(file);
4061 	if (ret)
4062 		return ret;
4063 
4064 	mutex_lock(&root->fs_info->volume_mutex);
4065 	trans = btrfs_start_transaction(root, 0);
4066 	if (IS_ERR(trans)) {
4067 		ret = PTR_ERR(trans);
4068 		goto out_unlock;
4069 	}
4070 
4071 	strcpy(super_block->label, label);
4072 	ret = btrfs_end_transaction(trans, root);
4073 
4074 out_unlock:
4075 	mutex_unlock(&root->fs_info->volume_mutex);
4076 	mnt_drop_write_file(file);
4077 	return ret;
4078 }
4079 
4080 long btrfs_ioctl(struct file *file, unsigned int
4081 		cmd, unsigned long arg)
4082 {
4083 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4084 	void __user *argp = (void __user *)arg;
4085 
4086 	switch (cmd) {
4087 	case FS_IOC_GETFLAGS:
4088 		return btrfs_ioctl_getflags(file, argp);
4089 	case FS_IOC_SETFLAGS:
4090 		return btrfs_ioctl_setflags(file, argp);
4091 	case FS_IOC_GETVERSION:
4092 		return btrfs_ioctl_getversion(file, argp);
4093 	case FITRIM:
4094 		return btrfs_ioctl_fitrim(file, argp);
4095 	case BTRFS_IOC_SNAP_CREATE:
4096 		return btrfs_ioctl_snap_create(file, argp, 0);
4097 	case BTRFS_IOC_SNAP_CREATE_V2:
4098 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
4099 	case BTRFS_IOC_SUBVOL_CREATE:
4100 		return btrfs_ioctl_snap_create(file, argp, 1);
4101 	case BTRFS_IOC_SUBVOL_CREATE_V2:
4102 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
4103 	case BTRFS_IOC_SNAP_DESTROY:
4104 		return btrfs_ioctl_snap_destroy(file, argp);
4105 	case BTRFS_IOC_SUBVOL_GETFLAGS:
4106 		return btrfs_ioctl_subvol_getflags(file, argp);
4107 	case BTRFS_IOC_SUBVOL_SETFLAGS:
4108 		return btrfs_ioctl_subvol_setflags(file, argp);
4109 	case BTRFS_IOC_DEFAULT_SUBVOL:
4110 		return btrfs_ioctl_default_subvol(file, argp);
4111 	case BTRFS_IOC_DEFRAG:
4112 		return btrfs_ioctl_defrag(file, NULL);
4113 	case BTRFS_IOC_DEFRAG_RANGE:
4114 		return btrfs_ioctl_defrag(file, argp);
4115 	case BTRFS_IOC_RESIZE:
4116 		return btrfs_ioctl_resize(file, argp);
4117 	case BTRFS_IOC_ADD_DEV:
4118 		return btrfs_ioctl_add_dev(root, argp);
4119 	case BTRFS_IOC_RM_DEV:
4120 		return btrfs_ioctl_rm_dev(file, argp);
4121 	case BTRFS_IOC_FS_INFO:
4122 		return btrfs_ioctl_fs_info(root, argp);
4123 	case BTRFS_IOC_DEV_INFO:
4124 		return btrfs_ioctl_dev_info(root, argp);
4125 	case BTRFS_IOC_BALANCE:
4126 		return btrfs_ioctl_balance(file, NULL);
4127 	case BTRFS_IOC_CLONE:
4128 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4129 	case BTRFS_IOC_CLONE_RANGE:
4130 		return btrfs_ioctl_clone_range(file, argp);
4131 	case BTRFS_IOC_TRANS_START:
4132 		return btrfs_ioctl_trans_start(file);
4133 	case BTRFS_IOC_TRANS_END:
4134 		return btrfs_ioctl_trans_end(file);
4135 	case BTRFS_IOC_TREE_SEARCH:
4136 		return btrfs_ioctl_tree_search(file, argp);
4137 	case BTRFS_IOC_INO_LOOKUP:
4138 		return btrfs_ioctl_ino_lookup(file, argp);
4139 	case BTRFS_IOC_INO_PATHS:
4140 		return btrfs_ioctl_ino_to_path(root, argp);
4141 	case BTRFS_IOC_LOGICAL_INO:
4142 		return btrfs_ioctl_logical_to_ino(root, argp);
4143 	case BTRFS_IOC_SPACE_INFO:
4144 		return btrfs_ioctl_space_info(root, argp);
4145 	case BTRFS_IOC_SYNC:
4146 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
4147 		return 0;
4148 	case BTRFS_IOC_START_SYNC:
4149 		return btrfs_ioctl_start_sync(root, argp);
4150 	case BTRFS_IOC_WAIT_SYNC:
4151 		return btrfs_ioctl_wait_sync(root, argp);
4152 	case BTRFS_IOC_SCRUB:
4153 		return btrfs_ioctl_scrub(file, argp);
4154 	case BTRFS_IOC_SCRUB_CANCEL:
4155 		return btrfs_ioctl_scrub_cancel(root, argp);
4156 	case BTRFS_IOC_SCRUB_PROGRESS:
4157 		return btrfs_ioctl_scrub_progress(root, argp);
4158 	case BTRFS_IOC_BALANCE_V2:
4159 		return btrfs_ioctl_balance(file, argp);
4160 	case BTRFS_IOC_BALANCE_CTL:
4161 		return btrfs_ioctl_balance_ctl(root, arg);
4162 	case BTRFS_IOC_BALANCE_PROGRESS:
4163 		return btrfs_ioctl_balance_progress(root, argp);
4164 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4165 		return btrfs_ioctl_set_received_subvol(file, argp);
4166 	case BTRFS_IOC_SEND:
4167 		return btrfs_ioctl_send(file, argp);
4168 	case BTRFS_IOC_GET_DEV_STATS:
4169 		return btrfs_ioctl_get_dev_stats(root, argp);
4170 	case BTRFS_IOC_QUOTA_CTL:
4171 		return btrfs_ioctl_quota_ctl(file, argp);
4172 	case BTRFS_IOC_QGROUP_ASSIGN:
4173 		return btrfs_ioctl_qgroup_assign(file, argp);
4174 	case BTRFS_IOC_QGROUP_CREATE:
4175 		return btrfs_ioctl_qgroup_create(file, argp);
4176 	case BTRFS_IOC_QGROUP_LIMIT:
4177 		return btrfs_ioctl_qgroup_limit(file, argp);
4178 	case BTRFS_IOC_QUOTA_RESCAN:
4179 		return btrfs_ioctl_quota_rescan(file, argp);
4180 	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4181 		return btrfs_ioctl_quota_rescan_status(file, argp);
4182 	case BTRFS_IOC_DEV_REPLACE:
4183 		return btrfs_ioctl_dev_replace(root, argp);
4184 	case BTRFS_IOC_GET_FSLABEL:
4185 		return btrfs_ioctl_get_fslabel(file, argp);
4186 	case BTRFS_IOC_SET_FSLABEL:
4187 		return btrfs_ioctl_set_fslabel(file, argp);
4188 	}
4189 
4190 	return -ENOTTY;
4191 }
4192