xref: /linux/fs/btrfs/ioctl.c (revision 12871a0bd67dd4db4418e1daafcd46e9d329ef10)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include "compat.h"
45 #include "ctree.h"
46 #include "disk-io.h"
47 #include "transaction.h"
48 #include "btrfs_inode.h"
49 #include "ioctl.h"
50 #include "print-tree.h"
51 #include "volumes.h"
52 #include "locking.h"
53 #include "inode-map.h"
54 
55 /* Mask out flags that are inappropriate for the given type of inode. */
56 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
57 {
58 	if (S_ISDIR(mode))
59 		return flags;
60 	else if (S_ISREG(mode))
61 		return flags & ~FS_DIRSYNC_FL;
62 	else
63 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
64 }
65 
66 /*
67  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
68  */
69 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
70 {
71 	unsigned int iflags = 0;
72 
73 	if (flags & BTRFS_INODE_SYNC)
74 		iflags |= FS_SYNC_FL;
75 	if (flags & BTRFS_INODE_IMMUTABLE)
76 		iflags |= FS_IMMUTABLE_FL;
77 	if (flags & BTRFS_INODE_APPEND)
78 		iflags |= FS_APPEND_FL;
79 	if (flags & BTRFS_INODE_NODUMP)
80 		iflags |= FS_NODUMP_FL;
81 	if (flags & BTRFS_INODE_NOATIME)
82 		iflags |= FS_NOATIME_FL;
83 	if (flags & BTRFS_INODE_DIRSYNC)
84 		iflags |= FS_DIRSYNC_FL;
85 	if (flags & BTRFS_INODE_NODATACOW)
86 		iflags |= FS_NOCOW_FL;
87 
88 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
89 		iflags |= FS_COMPR_FL;
90 	else if (flags & BTRFS_INODE_NOCOMPRESS)
91 		iflags |= FS_NOCOMP_FL;
92 
93 	return iflags;
94 }
95 
96 /*
97  * Update inode->i_flags based on the btrfs internal flags.
98  */
99 void btrfs_update_iflags(struct inode *inode)
100 {
101 	struct btrfs_inode *ip = BTRFS_I(inode);
102 
103 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
104 
105 	if (ip->flags & BTRFS_INODE_SYNC)
106 		inode->i_flags |= S_SYNC;
107 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
108 		inode->i_flags |= S_IMMUTABLE;
109 	if (ip->flags & BTRFS_INODE_APPEND)
110 		inode->i_flags |= S_APPEND;
111 	if (ip->flags & BTRFS_INODE_NOATIME)
112 		inode->i_flags |= S_NOATIME;
113 	if (ip->flags & BTRFS_INODE_DIRSYNC)
114 		inode->i_flags |= S_DIRSYNC;
115 }
116 
117 /*
118  * Inherit flags from the parent inode.
119  *
120  * Unlike extN we don't have any flags we don't want to inherit currently.
121  */
122 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
123 {
124 	unsigned int flags;
125 
126 	if (!dir)
127 		return;
128 
129 	flags = BTRFS_I(dir)->flags;
130 
131 	if (S_ISREG(inode->i_mode))
132 		flags &= ~BTRFS_INODE_DIRSYNC;
133 	else if (!S_ISDIR(inode->i_mode))
134 		flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
135 
136 	BTRFS_I(inode)->flags = flags;
137 	btrfs_update_iflags(inode);
138 }
139 
140 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
141 {
142 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
143 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
144 
145 	if (copy_to_user(arg, &flags, sizeof(flags)))
146 		return -EFAULT;
147 	return 0;
148 }
149 
150 static int check_flags(unsigned int flags)
151 {
152 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
153 		      FS_NOATIME_FL | FS_NODUMP_FL | \
154 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
155 		      FS_NOCOMP_FL | FS_COMPR_FL |
156 		      FS_NOCOW_FL))
157 		return -EOPNOTSUPP;
158 
159 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
160 		return -EINVAL;
161 
162 	return 0;
163 }
164 
165 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
166 {
167 	struct inode *inode = file->f_path.dentry->d_inode;
168 	struct btrfs_inode *ip = BTRFS_I(inode);
169 	struct btrfs_root *root = ip->root;
170 	struct btrfs_trans_handle *trans;
171 	unsigned int flags, oldflags;
172 	int ret;
173 
174 	if (btrfs_root_readonly(root))
175 		return -EROFS;
176 
177 	if (copy_from_user(&flags, arg, sizeof(flags)))
178 		return -EFAULT;
179 
180 	ret = check_flags(flags);
181 	if (ret)
182 		return ret;
183 
184 	if (!inode_owner_or_capable(inode))
185 		return -EACCES;
186 
187 	mutex_lock(&inode->i_mutex);
188 
189 	flags = btrfs_mask_flags(inode->i_mode, flags);
190 	oldflags = btrfs_flags_to_ioctl(ip->flags);
191 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
192 		if (!capable(CAP_LINUX_IMMUTABLE)) {
193 			ret = -EPERM;
194 			goto out_unlock;
195 		}
196 	}
197 
198 	ret = mnt_want_write(file->f_path.mnt);
199 	if (ret)
200 		goto out_unlock;
201 
202 	if (flags & FS_SYNC_FL)
203 		ip->flags |= BTRFS_INODE_SYNC;
204 	else
205 		ip->flags &= ~BTRFS_INODE_SYNC;
206 	if (flags & FS_IMMUTABLE_FL)
207 		ip->flags |= BTRFS_INODE_IMMUTABLE;
208 	else
209 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
210 	if (flags & FS_APPEND_FL)
211 		ip->flags |= BTRFS_INODE_APPEND;
212 	else
213 		ip->flags &= ~BTRFS_INODE_APPEND;
214 	if (flags & FS_NODUMP_FL)
215 		ip->flags |= BTRFS_INODE_NODUMP;
216 	else
217 		ip->flags &= ~BTRFS_INODE_NODUMP;
218 	if (flags & FS_NOATIME_FL)
219 		ip->flags |= BTRFS_INODE_NOATIME;
220 	else
221 		ip->flags &= ~BTRFS_INODE_NOATIME;
222 	if (flags & FS_DIRSYNC_FL)
223 		ip->flags |= BTRFS_INODE_DIRSYNC;
224 	else
225 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
226 	if (flags & FS_NOCOW_FL)
227 		ip->flags |= BTRFS_INODE_NODATACOW;
228 	else
229 		ip->flags &= ~BTRFS_INODE_NODATACOW;
230 
231 	/*
232 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
233 	 * flag may be changed automatically if compression code won't make
234 	 * things smaller.
235 	 */
236 	if (flags & FS_NOCOMP_FL) {
237 		ip->flags &= ~BTRFS_INODE_COMPRESS;
238 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
239 	} else if (flags & FS_COMPR_FL) {
240 		ip->flags |= BTRFS_INODE_COMPRESS;
241 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
242 	} else {
243 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
244 	}
245 
246 	trans = btrfs_join_transaction(root);
247 	BUG_ON(IS_ERR(trans));
248 
249 	ret = btrfs_update_inode(trans, root, inode);
250 	BUG_ON(ret);
251 
252 	btrfs_update_iflags(inode);
253 	inode->i_ctime = CURRENT_TIME;
254 	btrfs_end_transaction(trans, root);
255 
256 	mnt_drop_write(file->f_path.mnt);
257 
258 	ret = 0;
259  out_unlock:
260 	mutex_unlock(&inode->i_mutex);
261 	return ret;
262 }
263 
264 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
265 {
266 	struct inode *inode = file->f_path.dentry->d_inode;
267 
268 	return put_user(inode->i_generation, arg);
269 }
270 
271 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
272 {
273 	struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
274 	struct btrfs_fs_info *fs_info = root->fs_info;
275 	struct btrfs_device *device;
276 	struct request_queue *q;
277 	struct fstrim_range range;
278 	u64 minlen = ULLONG_MAX;
279 	u64 num_devices = 0;
280 	int ret;
281 
282 	if (!capable(CAP_SYS_ADMIN))
283 		return -EPERM;
284 
285 	rcu_read_lock();
286 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
287 				dev_list) {
288 		if (!device->bdev)
289 			continue;
290 		q = bdev_get_queue(device->bdev);
291 		if (blk_queue_discard(q)) {
292 			num_devices++;
293 			minlen = min((u64)q->limits.discard_granularity,
294 				     minlen);
295 		}
296 	}
297 	rcu_read_unlock();
298 	if (!num_devices)
299 		return -EOPNOTSUPP;
300 
301 	if (copy_from_user(&range, arg, sizeof(range)))
302 		return -EFAULT;
303 
304 	range.minlen = max(range.minlen, minlen);
305 	ret = btrfs_trim_fs(root, &range);
306 	if (ret < 0)
307 		return ret;
308 
309 	if (copy_to_user(arg, &range, sizeof(range)))
310 		return -EFAULT;
311 
312 	return 0;
313 }
314 
315 static noinline int create_subvol(struct btrfs_root *root,
316 				  struct dentry *dentry,
317 				  char *name, int namelen,
318 				  u64 *async_transid)
319 {
320 	struct btrfs_trans_handle *trans;
321 	struct btrfs_key key;
322 	struct btrfs_root_item root_item;
323 	struct btrfs_inode_item *inode_item;
324 	struct extent_buffer *leaf;
325 	struct btrfs_root *new_root;
326 	struct dentry *parent = dget_parent(dentry);
327 	struct inode *dir;
328 	int ret;
329 	int err;
330 	u64 objectid;
331 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
332 	u64 index = 0;
333 
334 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
335 	if (ret) {
336 		dput(parent);
337 		return ret;
338 	}
339 
340 	dir = parent->d_inode;
341 
342 	/*
343 	 * 1 - inode item
344 	 * 2 - refs
345 	 * 1 - root item
346 	 * 2 - dir items
347 	 */
348 	trans = btrfs_start_transaction(root, 6);
349 	if (IS_ERR(trans)) {
350 		dput(parent);
351 		return PTR_ERR(trans);
352 	}
353 
354 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
355 				      0, objectid, NULL, 0, 0, 0);
356 	if (IS_ERR(leaf)) {
357 		ret = PTR_ERR(leaf);
358 		goto fail;
359 	}
360 
361 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
362 	btrfs_set_header_bytenr(leaf, leaf->start);
363 	btrfs_set_header_generation(leaf, trans->transid);
364 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
365 	btrfs_set_header_owner(leaf, objectid);
366 
367 	write_extent_buffer(leaf, root->fs_info->fsid,
368 			    (unsigned long)btrfs_header_fsid(leaf),
369 			    BTRFS_FSID_SIZE);
370 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
371 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
372 			    BTRFS_UUID_SIZE);
373 	btrfs_mark_buffer_dirty(leaf);
374 
375 	inode_item = &root_item.inode;
376 	memset(inode_item, 0, sizeof(*inode_item));
377 	inode_item->generation = cpu_to_le64(1);
378 	inode_item->size = cpu_to_le64(3);
379 	inode_item->nlink = cpu_to_le32(1);
380 	inode_item->nbytes = cpu_to_le64(root->leafsize);
381 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
382 
383 	root_item.flags = 0;
384 	root_item.byte_limit = 0;
385 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
386 
387 	btrfs_set_root_bytenr(&root_item, leaf->start);
388 	btrfs_set_root_generation(&root_item, trans->transid);
389 	btrfs_set_root_level(&root_item, 0);
390 	btrfs_set_root_refs(&root_item, 1);
391 	btrfs_set_root_used(&root_item, leaf->len);
392 	btrfs_set_root_last_snapshot(&root_item, 0);
393 
394 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
395 	root_item.drop_level = 0;
396 
397 	btrfs_tree_unlock(leaf);
398 	free_extent_buffer(leaf);
399 	leaf = NULL;
400 
401 	btrfs_set_root_dirid(&root_item, new_dirid);
402 
403 	key.objectid = objectid;
404 	key.offset = 0;
405 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
406 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
407 				&root_item);
408 	if (ret)
409 		goto fail;
410 
411 	key.offset = (u64)-1;
412 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
413 	BUG_ON(IS_ERR(new_root));
414 
415 	btrfs_record_root_in_trans(trans, new_root);
416 
417 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
418 	/*
419 	 * insert the directory item
420 	 */
421 	ret = btrfs_set_inode_index(dir, &index);
422 	BUG_ON(ret);
423 
424 	ret = btrfs_insert_dir_item(trans, root,
425 				    name, namelen, dir, &key,
426 				    BTRFS_FT_DIR, index);
427 	if (ret)
428 		goto fail;
429 
430 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
431 	ret = btrfs_update_inode(trans, root, dir);
432 	BUG_ON(ret);
433 
434 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
435 				 objectid, root->root_key.objectid,
436 				 btrfs_ino(dir), index, name, namelen);
437 
438 	BUG_ON(ret);
439 
440 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
441 fail:
442 	dput(parent);
443 	if (async_transid) {
444 		*async_transid = trans->transid;
445 		err = btrfs_commit_transaction_async(trans, root, 1);
446 	} else {
447 		err = btrfs_commit_transaction(trans, root);
448 	}
449 	if (err && !ret)
450 		ret = err;
451 	return ret;
452 }
453 
454 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
455 			   char *name, int namelen, u64 *async_transid,
456 			   bool readonly)
457 {
458 	struct inode *inode;
459 	struct dentry *parent;
460 	struct btrfs_pending_snapshot *pending_snapshot;
461 	struct btrfs_trans_handle *trans;
462 	int ret;
463 
464 	if (!root->ref_cows)
465 		return -EINVAL;
466 
467 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
468 	if (!pending_snapshot)
469 		return -ENOMEM;
470 
471 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
472 	pending_snapshot->dentry = dentry;
473 	pending_snapshot->root = root;
474 	pending_snapshot->readonly = readonly;
475 
476 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
477 	if (IS_ERR(trans)) {
478 		ret = PTR_ERR(trans);
479 		goto fail;
480 	}
481 
482 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
483 	BUG_ON(ret);
484 
485 	list_add(&pending_snapshot->list,
486 		 &trans->transaction->pending_snapshots);
487 	if (async_transid) {
488 		*async_transid = trans->transid;
489 		ret = btrfs_commit_transaction_async(trans,
490 				     root->fs_info->extent_root, 1);
491 	} else {
492 		ret = btrfs_commit_transaction(trans,
493 					       root->fs_info->extent_root);
494 	}
495 	BUG_ON(ret);
496 
497 	ret = pending_snapshot->error;
498 	if (ret)
499 		goto fail;
500 
501 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
502 	if (ret)
503 		goto fail;
504 
505 	parent = dget_parent(dentry);
506 	inode = btrfs_lookup_dentry(parent->d_inode, dentry);
507 	dput(parent);
508 	if (IS_ERR(inode)) {
509 		ret = PTR_ERR(inode);
510 		goto fail;
511 	}
512 	BUG_ON(!inode);
513 	d_instantiate(dentry, inode);
514 	ret = 0;
515 fail:
516 	kfree(pending_snapshot);
517 	return ret;
518 }
519 
520 /*  copy of check_sticky in fs/namei.c()
521 * It's inline, so penalty for filesystems that don't use sticky bit is
522 * minimal.
523 */
524 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
525 {
526 	uid_t fsuid = current_fsuid();
527 
528 	if (!(dir->i_mode & S_ISVTX))
529 		return 0;
530 	if (inode->i_uid == fsuid)
531 		return 0;
532 	if (dir->i_uid == fsuid)
533 		return 0;
534 	return !capable(CAP_FOWNER);
535 }
536 
537 /*  copy of may_delete in fs/namei.c()
538  *	Check whether we can remove a link victim from directory dir, check
539  *  whether the type of victim is right.
540  *  1. We can't do it if dir is read-only (done in permission())
541  *  2. We should have write and exec permissions on dir
542  *  3. We can't remove anything from append-only dir
543  *  4. We can't do anything with immutable dir (done in permission())
544  *  5. If the sticky bit on dir is set we should either
545  *	a. be owner of dir, or
546  *	b. be owner of victim, or
547  *	c. have CAP_FOWNER capability
548  *  6. If the victim is append-only or immutable we can't do antyhing with
549  *     links pointing to it.
550  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
551  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
552  *  9. We can't remove a root or mountpoint.
553  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
554  *     nfs_async_unlink().
555  */
556 
557 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
558 {
559 	int error;
560 
561 	if (!victim->d_inode)
562 		return -ENOENT;
563 
564 	BUG_ON(victim->d_parent->d_inode != dir);
565 	audit_inode_child(victim, dir);
566 
567 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
568 	if (error)
569 		return error;
570 	if (IS_APPEND(dir))
571 		return -EPERM;
572 	if (btrfs_check_sticky(dir, victim->d_inode)||
573 		IS_APPEND(victim->d_inode)||
574 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
575 		return -EPERM;
576 	if (isdir) {
577 		if (!S_ISDIR(victim->d_inode->i_mode))
578 			return -ENOTDIR;
579 		if (IS_ROOT(victim))
580 			return -EBUSY;
581 	} else if (S_ISDIR(victim->d_inode->i_mode))
582 		return -EISDIR;
583 	if (IS_DEADDIR(dir))
584 		return -ENOENT;
585 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
586 		return -EBUSY;
587 	return 0;
588 }
589 
590 /* copy of may_create in fs/namei.c() */
591 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
592 {
593 	if (child->d_inode)
594 		return -EEXIST;
595 	if (IS_DEADDIR(dir))
596 		return -ENOENT;
597 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
598 }
599 
600 /*
601  * Create a new subvolume below @parent.  This is largely modeled after
602  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
603  * inside this filesystem so it's quite a bit simpler.
604  */
605 static noinline int btrfs_mksubvol(struct path *parent,
606 				   char *name, int namelen,
607 				   struct btrfs_root *snap_src,
608 				   u64 *async_transid, bool readonly)
609 {
610 	struct inode *dir  = parent->dentry->d_inode;
611 	struct dentry *dentry;
612 	int error;
613 
614 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
615 
616 	dentry = lookup_one_len(name, parent->dentry, namelen);
617 	error = PTR_ERR(dentry);
618 	if (IS_ERR(dentry))
619 		goto out_unlock;
620 
621 	error = -EEXIST;
622 	if (dentry->d_inode)
623 		goto out_dput;
624 
625 	error = mnt_want_write(parent->mnt);
626 	if (error)
627 		goto out_dput;
628 
629 	error = btrfs_may_create(dir, dentry);
630 	if (error)
631 		goto out_drop_write;
632 
633 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
634 
635 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
636 		goto out_up_read;
637 
638 	if (snap_src) {
639 		error = create_snapshot(snap_src, dentry,
640 					name, namelen, async_transid, readonly);
641 	} else {
642 		error = create_subvol(BTRFS_I(dir)->root, dentry,
643 				      name, namelen, async_transid);
644 	}
645 	if (!error)
646 		fsnotify_mkdir(dir, dentry);
647 out_up_read:
648 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
649 out_drop_write:
650 	mnt_drop_write(parent->mnt);
651 out_dput:
652 	dput(dentry);
653 out_unlock:
654 	mutex_unlock(&dir->i_mutex);
655 	return error;
656 }
657 
658 /*
659  * When we're defragging a range, we don't want to kick it off again
660  * if it is really just waiting for delalloc to send it down.
661  * If we find a nice big extent or delalloc range for the bytes in the
662  * file you want to defrag, we return 0 to let you know to skip this
663  * part of the file
664  */
665 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
666 {
667 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
668 	struct extent_map *em = NULL;
669 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
670 	u64 end;
671 
672 	read_lock(&em_tree->lock);
673 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
674 	read_unlock(&em_tree->lock);
675 
676 	if (em) {
677 		end = extent_map_end(em);
678 		free_extent_map(em);
679 		if (end - offset > thresh)
680 			return 0;
681 	}
682 	/* if we already have a nice delalloc here, just stop */
683 	thresh /= 2;
684 	end = count_range_bits(io_tree, &offset, offset + thresh,
685 			       thresh, EXTENT_DELALLOC, 1);
686 	if (end >= thresh)
687 		return 0;
688 	return 1;
689 }
690 
691 /*
692  * helper function to walk through a file and find extents
693  * newer than a specific transid, and smaller than thresh.
694  *
695  * This is used by the defragging code to find new and small
696  * extents
697  */
698 static int find_new_extents(struct btrfs_root *root,
699 			    struct inode *inode, u64 newer_than,
700 			    u64 *off, int thresh)
701 {
702 	struct btrfs_path *path;
703 	struct btrfs_key min_key;
704 	struct btrfs_key max_key;
705 	struct extent_buffer *leaf;
706 	struct btrfs_file_extent_item *extent;
707 	int type;
708 	int ret;
709 	u64 ino = btrfs_ino(inode);
710 
711 	path = btrfs_alloc_path();
712 	if (!path)
713 		return -ENOMEM;
714 
715 	min_key.objectid = ino;
716 	min_key.type = BTRFS_EXTENT_DATA_KEY;
717 	min_key.offset = *off;
718 
719 	max_key.objectid = ino;
720 	max_key.type = (u8)-1;
721 	max_key.offset = (u64)-1;
722 
723 	path->keep_locks = 1;
724 
725 	while(1) {
726 		ret = btrfs_search_forward(root, &min_key, &max_key,
727 					   path, 0, newer_than);
728 		if (ret != 0)
729 			goto none;
730 		if (min_key.objectid != ino)
731 			goto none;
732 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
733 			goto none;
734 
735 		leaf = path->nodes[0];
736 		extent = btrfs_item_ptr(leaf, path->slots[0],
737 					struct btrfs_file_extent_item);
738 
739 		type = btrfs_file_extent_type(leaf, extent);
740 		if (type == BTRFS_FILE_EXTENT_REG &&
741 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
742 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
743 			*off = min_key.offset;
744 			btrfs_free_path(path);
745 			return 0;
746 		}
747 
748 		if (min_key.offset == (u64)-1)
749 			goto none;
750 
751 		min_key.offset++;
752 		btrfs_release_path(path);
753 	}
754 none:
755 	btrfs_free_path(path);
756 	return -ENOENT;
757 }
758 
759 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
760 			       int thresh, u64 *last_len, u64 *skip,
761 			       u64 *defrag_end)
762 {
763 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
764 	struct extent_map *em = NULL;
765 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
766 	int ret = 1;
767 
768 	/*
769 	 * make sure that once we start defragging and extent, we keep on
770 	 * defragging it
771 	 */
772 	if (start < *defrag_end)
773 		return 1;
774 
775 	*skip = 0;
776 
777 	/*
778 	 * hopefully we have this extent in the tree already, try without
779 	 * the full extent lock
780 	 */
781 	read_lock(&em_tree->lock);
782 	em = lookup_extent_mapping(em_tree, start, len);
783 	read_unlock(&em_tree->lock);
784 
785 	if (!em) {
786 		/* get the big lock and read metadata off disk */
787 		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
788 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
789 		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
790 
791 		if (IS_ERR(em))
792 			return 0;
793 	}
794 
795 	/* this will cover holes, and inline extents */
796 	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
797 		ret = 0;
798 
799 	/*
800 	 * we hit a real extent, if it is big don't bother defragging it again
801 	 */
802 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
803 		ret = 0;
804 
805 	/*
806 	 * last_len ends up being a counter of how many bytes we've defragged.
807 	 * every time we choose not to defrag an extent, we reset *last_len
808 	 * so that the next tiny extent will force a defrag.
809 	 *
810 	 * The end result of this is that tiny extents before a single big
811 	 * extent will force at least part of that big extent to be defragged.
812 	 */
813 	if (ret) {
814 		*last_len += len;
815 		*defrag_end = extent_map_end(em);
816 	} else {
817 		*last_len = 0;
818 		*skip = extent_map_end(em);
819 		*defrag_end = 0;
820 	}
821 
822 	free_extent_map(em);
823 	return ret;
824 }
825 
826 /*
827  * it doesn't do much good to defrag one or two pages
828  * at a time.  This pulls in a nice chunk of pages
829  * to COW and defrag.
830  *
831  * It also makes sure the delalloc code has enough
832  * dirty data to avoid making new small extents as part
833  * of the defrag
834  *
835  * It's a good idea to start RA on this range
836  * before calling this.
837  */
838 static int cluster_pages_for_defrag(struct inode *inode,
839 				    struct page **pages,
840 				    unsigned long start_index,
841 				    int num_pages)
842 {
843 	unsigned long file_end;
844 	u64 isize = i_size_read(inode);
845 	u64 page_start;
846 	u64 page_end;
847 	int ret;
848 	int i;
849 	int i_done;
850 	struct btrfs_ordered_extent *ordered;
851 	struct extent_state *cached_state = NULL;
852 
853 	if (isize == 0)
854 		return 0;
855 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
856 
857 	ret = btrfs_delalloc_reserve_space(inode,
858 					   num_pages << PAGE_CACHE_SHIFT);
859 	if (ret)
860 		return ret;
861 again:
862 	ret = 0;
863 	i_done = 0;
864 
865 	/* step one, lock all the pages */
866 	for (i = 0; i < num_pages; i++) {
867 		struct page *page;
868 		page = grab_cache_page(inode->i_mapping,
869 					    start_index + i);
870 		if (!page)
871 			break;
872 
873 		if (!PageUptodate(page)) {
874 			btrfs_readpage(NULL, page);
875 			lock_page(page);
876 			if (!PageUptodate(page)) {
877 				unlock_page(page);
878 				page_cache_release(page);
879 				ret = -EIO;
880 				break;
881 			}
882 		}
883 		isize = i_size_read(inode);
884 		file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
885 		if (!isize || page->index > file_end ||
886 		    page->mapping != inode->i_mapping) {
887 			/* whoops, we blew past eof, skip this page */
888 			unlock_page(page);
889 			page_cache_release(page);
890 			break;
891 		}
892 		pages[i] = page;
893 		i_done++;
894 	}
895 	if (!i_done || ret)
896 		goto out;
897 
898 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
899 		goto out;
900 
901 	/*
902 	 * so now we have a nice long stream of locked
903 	 * and up to date pages, lets wait on them
904 	 */
905 	for (i = 0; i < i_done; i++)
906 		wait_on_page_writeback(pages[i]);
907 
908 	page_start = page_offset(pages[0]);
909 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
910 
911 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
912 			 page_start, page_end - 1, 0, &cached_state,
913 			 GFP_NOFS);
914 	ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
915 	if (ordered &&
916 	    ordered->file_offset + ordered->len > page_start &&
917 	    ordered->file_offset < page_end) {
918 		btrfs_put_ordered_extent(ordered);
919 		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
920 				     page_start, page_end - 1,
921 				     &cached_state, GFP_NOFS);
922 		for (i = 0; i < i_done; i++) {
923 			unlock_page(pages[i]);
924 			page_cache_release(pages[i]);
925 		}
926 		btrfs_wait_ordered_range(inode, page_start,
927 					 page_end - page_start);
928 		goto again;
929 	}
930 	if (ordered)
931 		btrfs_put_ordered_extent(ordered);
932 
933 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
934 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
935 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
936 			  GFP_NOFS);
937 
938 	if (i_done != num_pages) {
939 		atomic_inc(&BTRFS_I(inode)->outstanding_extents);
940 		btrfs_delalloc_release_space(inode,
941 				     (num_pages - i_done) << PAGE_CACHE_SHIFT);
942 	}
943 
944 
945 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
946 				  &cached_state);
947 
948 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
949 			     page_start, page_end - 1, &cached_state,
950 			     GFP_NOFS);
951 
952 	for (i = 0; i < i_done; i++) {
953 		clear_page_dirty_for_io(pages[i]);
954 		ClearPageChecked(pages[i]);
955 		set_page_extent_mapped(pages[i]);
956 		set_page_dirty(pages[i]);
957 		unlock_page(pages[i]);
958 		page_cache_release(pages[i]);
959 	}
960 	return i_done;
961 out:
962 	for (i = 0; i < i_done; i++) {
963 		unlock_page(pages[i]);
964 		page_cache_release(pages[i]);
965 	}
966 	btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
967 	return ret;
968 
969 }
970 
971 int btrfs_defrag_file(struct inode *inode, struct file *file,
972 		      struct btrfs_ioctl_defrag_range_args *range,
973 		      u64 newer_than, unsigned long max_to_defrag)
974 {
975 	struct btrfs_root *root = BTRFS_I(inode)->root;
976 	struct btrfs_super_block *disk_super;
977 	struct file_ra_state *ra = NULL;
978 	unsigned long last_index;
979 	u64 features;
980 	u64 last_len = 0;
981 	u64 skip = 0;
982 	u64 defrag_end = 0;
983 	u64 newer_off = range->start;
984 	int newer_left = 0;
985 	unsigned long i;
986 	int ret;
987 	int defrag_count = 0;
988 	int compress_type = BTRFS_COMPRESS_ZLIB;
989 	int extent_thresh = range->extent_thresh;
990 	int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
991 	u64 new_align = ~((u64)128 * 1024 - 1);
992 	struct page **pages = NULL;
993 
994 	if (extent_thresh == 0)
995 		extent_thresh = 256 * 1024;
996 
997 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
998 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
999 			return -EINVAL;
1000 		if (range->compress_type)
1001 			compress_type = range->compress_type;
1002 	}
1003 
1004 	if (inode->i_size == 0)
1005 		return 0;
1006 
1007 	/*
1008 	 * if we were not given a file, allocate a readahead
1009 	 * context
1010 	 */
1011 	if (!file) {
1012 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1013 		if (!ra)
1014 			return -ENOMEM;
1015 		file_ra_state_init(ra, inode->i_mapping);
1016 	} else {
1017 		ra = &file->f_ra;
1018 	}
1019 
1020 	pages = kmalloc(sizeof(struct page *) * newer_cluster,
1021 			GFP_NOFS);
1022 	if (!pages) {
1023 		ret = -ENOMEM;
1024 		goto out_ra;
1025 	}
1026 
1027 	/* find the last page to defrag */
1028 	if (range->start + range->len > range->start) {
1029 		last_index = min_t(u64, inode->i_size - 1,
1030 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1031 	} else {
1032 		last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
1033 	}
1034 
1035 	if (newer_than) {
1036 		ret = find_new_extents(root, inode, newer_than,
1037 				       &newer_off, 64 * 1024);
1038 		if (!ret) {
1039 			range->start = newer_off;
1040 			/*
1041 			 * we always align our defrag to help keep
1042 			 * the extents in the file evenly spaced
1043 			 */
1044 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1045 			newer_left = newer_cluster;
1046 		} else
1047 			goto out_ra;
1048 	} else {
1049 		i = range->start >> PAGE_CACHE_SHIFT;
1050 	}
1051 	if (!max_to_defrag)
1052 		max_to_defrag = last_index - 1;
1053 
1054 	while (i <= last_index && defrag_count < max_to_defrag) {
1055 		/*
1056 		 * make sure we stop running if someone unmounts
1057 		 * the FS
1058 		 */
1059 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1060 			break;
1061 
1062 		if (!newer_than &&
1063 		    !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1064 					PAGE_CACHE_SIZE,
1065 					extent_thresh,
1066 					&last_len, &skip,
1067 					&defrag_end)) {
1068 			unsigned long next;
1069 			/*
1070 			 * the should_defrag function tells us how much to skip
1071 			 * bump our counter by the suggested amount
1072 			 */
1073 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1074 			i = max(i + 1, next);
1075 			continue;
1076 		}
1077 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1078 			BTRFS_I(inode)->force_compress = compress_type;
1079 
1080 		btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
1081 
1082 		ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1083 		if (ret < 0)
1084 			goto out_ra;
1085 
1086 		defrag_count += ret;
1087 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1088 		i += ret;
1089 
1090 		if (newer_than) {
1091 			if (newer_off == (u64)-1)
1092 				break;
1093 
1094 			newer_off = max(newer_off + 1,
1095 					(u64)i << PAGE_CACHE_SHIFT);
1096 
1097 			ret = find_new_extents(root, inode,
1098 					       newer_than, &newer_off,
1099 					       64 * 1024);
1100 			if (!ret) {
1101 				range->start = newer_off;
1102 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1103 				newer_left = newer_cluster;
1104 			} else {
1105 				break;
1106 			}
1107 		} else {
1108 			i++;
1109 		}
1110 	}
1111 
1112 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1113 		filemap_flush(inode->i_mapping);
1114 
1115 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1116 		/* the filemap_flush will queue IO into the worker threads, but
1117 		 * we have to make sure the IO is actually started and that
1118 		 * ordered extents get created before we return
1119 		 */
1120 		atomic_inc(&root->fs_info->async_submit_draining);
1121 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1122 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1123 			wait_event(root->fs_info->async_submit_wait,
1124 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1125 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1126 		}
1127 		atomic_dec(&root->fs_info->async_submit_draining);
1128 
1129 		mutex_lock(&inode->i_mutex);
1130 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1131 		mutex_unlock(&inode->i_mutex);
1132 	}
1133 
1134 	disk_super = &root->fs_info->super_copy;
1135 	features = btrfs_super_incompat_flags(disk_super);
1136 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1137 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1138 		btrfs_set_super_incompat_flags(disk_super, features);
1139 	}
1140 
1141 	if (!file)
1142 		kfree(ra);
1143 	return defrag_count;
1144 
1145 out_ra:
1146 	if (!file)
1147 		kfree(ra);
1148 	kfree(pages);
1149 	return ret;
1150 }
1151 
1152 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1153 					void __user *arg)
1154 {
1155 	u64 new_size;
1156 	u64 old_size;
1157 	u64 devid = 1;
1158 	struct btrfs_ioctl_vol_args *vol_args;
1159 	struct btrfs_trans_handle *trans;
1160 	struct btrfs_device *device = NULL;
1161 	char *sizestr;
1162 	char *devstr = NULL;
1163 	int ret = 0;
1164 	int mod = 0;
1165 
1166 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1167 		return -EROFS;
1168 
1169 	if (!capable(CAP_SYS_ADMIN))
1170 		return -EPERM;
1171 
1172 	vol_args = memdup_user(arg, sizeof(*vol_args));
1173 	if (IS_ERR(vol_args))
1174 		return PTR_ERR(vol_args);
1175 
1176 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1177 
1178 	mutex_lock(&root->fs_info->volume_mutex);
1179 	sizestr = vol_args->name;
1180 	devstr = strchr(sizestr, ':');
1181 	if (devstr) {
1182 		char *end;
1183 		sizestr = devstr + 1;
1184 		*devstr = '\0';
1185 		devstr = vol_args->name;
1186 		devid = simple_strtoull(devstr, &end, 10);
1187 		printk(KERN_INFO "resizing devid %llu\n",
1188 		       (unsigned long long)devid);
1189 	}
1190 	device = btrfs_find_device(root, devid, NULL, NULL);
1191 	if (!device) {
1192 		printk(KERN_INFO "resizer unable to find device %llu\n",
1193 		       (unsigned long long)devid);
1194 		ret = -EINVAL;
1195 		goto out_unlock;
1196 	}
1197 	if (!strcmp(sizestr, "max"))
1198 		new_size = device->bdev->bd_inode->i_size;
1199 	else {
1200 		if (sizestr[0] == '-') {
1201 			mod = -1;
1202 			sizestr++;
1203 		} else if (sizestr[0] == '+') {
1204 			mod = 1;
1205 			sizestr++;
1206 		}
1207 		new_size = memparse(sizestr, NULL);
1208 		if (new_size == 0) {
1209 			ret = -EINVAL;
1210 			goto out_unlock;
1211 		}
1212 	}
1213 
1214 	old_size = device->total_bytes;
1215 
1216 	if (mod < 0) {
1217 		if (new_size > old_size) {
1218 			ret = -EINVAL;
1219 			goto out_unlock;
1220 		}
1221 		new_size = old_size - new_size;
1222 	} else if (mod > 0) {
1223 		new_size = old_size + new_size;
1224 	}
1225 
1226 	if (new_size < 256 * 1024 * 1024) {
1227 		ret = -EINVAL;
1228 		goto out_unlock;
1229 	}
1230 	if (new_size > device->bdev->bd_inode->i_size) {
1231 		ret = -EFBIG;
1232 		goto out_unlock;
1233 	}
1234 
1235 	do_div(new_size, root->sectorsize);
1236 	new_size *= root->sectorsize;
1237 
1238 	printk(KERN_INFO "new size for %s is %llu\n",
1239 		device->name, (unsigned long long)new_size);
1240 
1241 	if (new_size > old_size) {
1242 		trans = btrfs_start_transaction(root, 0);
1243 		if (IS_ERR(trans)) {
1244 			ret = PTR_ERR(trans);
1245 			goto out_unlock;
1246 		}
1247 		ret = btrfs_grow_device(trans, device, new_size);
1248 		btrfs_commit_transaction(trans, root);
1249 	} else {
1250 		ret = btrfs_shrink_device(device, new_size);
1251 	}
1252 
1253 out_unlock:
1254 	mutex_unlock(&root->fs_info->volume_mutex);
1255 	kfree(vol_args);
1256 	return ret;
1257 }
1258 
1259 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1260 						    char *name,
1261 						    unsigned long fd,
1262 						    int subvol,
1263 						    u64 *transid,
1264 						    bool readonly)
1265 {
1266 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1267 	struct file *src_file;
1268 	int namelen;
1269 	int ret = 0;
1270 
1271 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1272 		return -EROFS;
1273 
1274 	namelen = strlen(name);
1275 	if (strchr(name, '/')) {
1276 		ret = -EINVAL;
1277 		goto out;
1278 	}
1279 
1280 	if (subvol) {
1281 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1282 				     NULL, transid, readonly);
1283 	} else {
1284 		struct inode *src_inode;
1285 		src_file = fget(fd);
1286 		if (!src_file) {
1287 			ret = -EINVAL;
1288 			goto out;
1289 		}
1290 
1291 		src_inode = src_file->f_path.dentry->d_inode;
1292 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1293 			printk(KERN_INFO "btrfs: Snapshot src from "
1294 			       "another FS\n");
1295 			ret = -EINVAL;
1296 			fput(src_file);
1297 			goto out;
1298 		}
1299 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1300 				     BTRFS_I(src_inode)->root,
1301 				     transid, readonly);
1302 		fput(src_file);
1303 	}
1304 out:
1305 	return ret;
1306 }
1307 
1308 static noinline int btrfs_ioctl_snap_create(struct file *file,
1309 					    void __user *arg, int subvol)
1310 {
1311 	struct btrfs_ioctl_vol_args *vol_args;
1312 	int ret;
1313 
1314 	vol_args = memdup_user(arg, sizeof(*vol_args));
1315 	if (IS_ERR(vol_args))
1316 		return PTR_ERR(vol_args);
1317 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1318 
1319 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1320 					      vol_args->fd, subvol,
1321 					      NULL, false);
1322 
1323 	kfree(vol_args);
1324 	return ret;
1325 }
1326 
1327 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1328 					       void __user *arg, int subvol)
1329 {
1330 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1331 	int ret;
1332 	u64 transid = 0;
1333 	u64 *ptr = NULL;
1334 	bool readonly = false;
1335 
1336 	vol_args = memdup_user(arg, sizeof(*vol_args));
1337 	if (IS_ERR(vol_args))
1338 		return PTR_ERR(vol_args);
1339 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1340 
1341 	if (vol_args->flags &
1342 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1343 		ret = -EOPNOTSUPP;
1344 		goto out;
1345 	}
1346 
1347 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1348 		ptr = &transid;
1349 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1350 		readonly = true;
1351 
1352 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1353 					      vol_args->fd, subvol,
1354 					      ptr, readonly);
1355 
1356 	if (ret == 0 && ptr &&
1357 	    copy_to_user(arg +
1358 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1359 				  transid), ptr, sizeof(*ptr)))
1360 		ret = -EFAULT;
1361 out:
1362 	kfree(vol_args);
1363 	return ret;
1364 }
1365 
1366 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1367 						void __user *arg)
1368 {
1369 	struct inode *inode = fdentry(file)->d_inode;
1370 	struct btrfs_root *root = BTRFS_I(inode)->root;
1371 	int ret = 0;
1372 	u64 flags = 0;
1373 
1374 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1375 		return -EINVAL;
1376 
1377 	down_read(&root->fs_info->subvol_sem);
1378 	if (btrfs_root_readonly(root))
1379 		flags |= BTRFS_SUBVOL_RDONLY;
1380 	up_read(&root->fs_info->subvol_sem);
1381 
1382 	if (copy_to_user(arg, &flags, sizeof(flags)))
1383 		ret = -EFAULT;
1384 
1385 	return ret;
1386 }
1387 
1388 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1389 					      void __user *arg)
1390 {
1391 	struct inode *inode = fdentry(file)->d_inode;
1392 	struct btrfs_root *root = BTRFS_I(inode)->root;
1393 	struct btrfs_trans_handle *trans;
1394 	u64 root_flags;
1395 	u64 flags;
1396 	int ret = 0;
1397 
1398 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1399 		return -EROFS;
1400 
1401 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1402 		return -EINVAL;
1403 
1404 	if (copy_from_user(&flags, arg, sizeof(flags)))
1405 		return -EFAULT;
1406 
1407 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1408 		return -EINVAL;
1409 
1410 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1411 		return -EOPNOTSUPP;
1412 
1413 	if (!inode_owner_or_capable(inode))
1414 		return -EACCES;
1415 
1416 	down_write(&root->fs_info->subvol_sem);
1417 
1418 	/* nothing to do */
1419 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1420 		goto out;
1421 
1422 	root_flags = btrfs_root_flags(&root->root_item);
1423 	if (flags & BTRFS_SUBVOL_RDONLY)
1424 		btrfs_set_root_flags(&root->root_item,
1425 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1426 	else
1427 		btrfs_set_root_flags(&root->root_item,
1428 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1429 
1430 	trans = btrfs_start_transaction(root, 1);
1431 	if (IS_ERR(trans)) {
1432 		ret = PTR_ERR(trans);
1433 		goto out_reset;
1434 	}
1435 
1436 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1437 				&root->root_key, &root->root_item);
1438 
1439 	btrfs_commit_transaction(trans, root);
1440 out_reset:
1441 	if (ret)
1442 		btrfs_set_root_flags(&root->root_item, root_flags);
1443 out:
1444 	up_write(&root->fs_info->subvol_sem);
1445 	return ret;
1446 }
1447 
1448 /*
1449  * helper to check if the subvolume references other subvolumes
1450  */
1451 static noinline int may_destroy_subvol(struct btrfs_root *root)
1452 {
1453 	struct btrfs_path *path;
1454 	struct btrfs_key key;
1455 	int ret;
1456 
1457 	path = btrfs_alloc_path();
1458 	if (!path)
1459 		return -ENOMEM;
1460 
1461 	key.objectid = root->root_key.objectid;
1462 	key.type = BTRFS_ROOT_REF_KEY;
1463 	key.offset = (u64)-1;
1464 
1465 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1466 				&key, path, 0, 0);
1467 	if (ret < 0)
1468 		goto out;
1469 	BUG_ON(ret == 0);
1470 
1471 	ret = 0;
1472 	if (path->slots[0] > 0) {
1473 		path->slots[0]--;
1474 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1475 		if (key.objectid == root->root_key.objectid &&
1476 		    key.type == BTRFS_ROOT_REF_KEY)
1477 			ret = -ENOTEMPTY;
1478 	}
1479 out:
1480 	btrfs_free_path(path);
1481 	return ret;
1482 }
1483 
1484 static noinline int key_in_sk(struct btrfs_key *key,
1485 			      struct btrfs_ioctl_search_key *sk)
1486 {
1487 	struct btrfs_key test;
1488 	int ret;
1489 
1490 	test.objectid = sk->min_objectid;
1491 	test.type = sk->min_type;
1492 	test.offset = sk->min_offset;
1493 
1494 	ret = btrfs_comp_cpu_keys(key, &test);
1495 	if (ret < 0)
1496 		return 0;
1497 
1498 	test.objectid = sk->max_objectid;
1499 	test.type = sk->max_type;
1500 	test.offset = sk->max_offset;
1501 
1502 	ret = btrfs_comp_cpu_keys(key, &test);
1503 	if (ret > 0)
1504 		return 0;
1505 	return 1;
1506 }
1507 
1508 static noinline int copy_to_sk(struct btrfs_root *root,
1509 			       struct btrfs_path *path,
1510 			       struct btrfs_key *key,
1511 			       struct btrfs_ioctl_search_key *sk,
1512 			       char *buf,
1513 			       unsigned long *sk_offset,
1514 			       int *num_found)
1515 {
1516 	u64 found_transid;
1517 	struct extent_buffer *leaf;
1518 	struct btrfs_ioctl_search_header sh;
1519 	unsigned long item_off;
1520 	unsigned long item_len;
1521 	int nritems;
1522 	int i;
1523 	int slot;
1524 	int ret = 0;
1525 
1526 	leaf = path->nodes[0];
1527 	slot = path->slots[0];
1528 	nritems = btrfs_header_nritems(leaf);
1529 
1530 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1531 		i = nritems;
1532 		goto advance_key;
1533 	}
1534 	found_transid = btrfs_header_generation(leaf);
1535 
1536 	for (i = slot; i < nritems; i++) {
1537 		item_off = btrfs_item_ptr_offset(leaf, i);
1538 		item_len = btrfs_item_size_nr(leaf, i);
1539 
1540 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1541 			item_len = 0;
1542 
1543 		if (sizeof(sh) + item_len + *sk_offset >
1544 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1545 			ret = 1;
1546 			goto overflow;
1547 		}
1548 
1549 		btrfs_item_key_to_cpu(leaf, key, i);
1550 		if (!key_in_sk(key, sk))
1551 			continue;
1552 
1553 		sh.objectid = key->objectid;
1554 		sh.offset = key->offset;
1555 		sh.type = key->type;
1556 		sh.len = item_len;
1557 		sh.transid = found_transid;
1558 
1559 		/* copy search result header */
1560 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1561 		*sk_offset += sizeof(sh);
1562 
1563 		if (item_len) {
1564 			char *p = buf + *sk_offset;
1565 			/* copy the item */
1566 			read_extent_buffer(leaf, p,
1567 					   item_off, item_len);
1568 			*sk_offset += item_len;
1569 		}
1570 		(*num_found)++;
1571 
1572 		if (*num_found >= sk->nr_items)
1573 			break;
1574 	}
1575 advance_key:
1576 	ret = 0;
1577 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1578 		key->offset++;
1579 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1580 		key->offset = 0;
1581 		key->type++;
1582 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1583 		key->offset = 0;
1584 		key->type = 0;
1585 		key->objectid++;
1586 	} else
1587 		ret = 1;
1588 overflow:
1589 	return ret;
1590 }
1591 
1592 static noinline int search_ioctl(struct inode *inode,
1593 				 struct btrfs_ioctl_search_args *args)
1594 {
1595 	struct btrfs_root *root;
1596 	struct btrfs_key key;
1597 	struct btrfs_key max_key;
1598 	struct btrfs_path *path;
1599 	struct btrfs_ioctl_search_key *sk = &args->key;
1600 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1601 	int ret;
1602 	int num_found = 0;
1603 	unsigned long sk_offset = 0;
1604 
1605 	path = btrfs_alloc_path();
1606 	if (!path)
1607 		return -ENOMEM;
1608 
1609 	if (sk->tree_id == 0) {
1610 		/* search the root of the inode that was passed */
1611 		root = BTRFS_I(inode)->root;
1612 	} else {
1613 		key.objectid = sk->tree_id;
1614 		key.type = BTRFS_ROOT_ITEM_KEY;
1615 		key.offset = (u64)-1;
1616 		root = btrfs_read_fs_root_no_name(info, &key);
1617 		if (IS_ERR(root)) {
1618 			printk(KERN_ERR "could not find root %llu\n",
1619 			       sk->tree_id);
1620 			btrfs_free_path(path);
1621 			return -ENOENT;
1622 		}
1623 	}
1624 
1625 	key.objectid = sk->min_objectid;
1626 	key.type = sk->min_type;
1627 	key.offset = sk->min_offset;
1628 
1629 	max_key.objectid = sk->max_objectid;
1630 	max_key.type = sk->max_type;
1631 	max_key.offset = sk->max_offset;
1632 
1633 	path->keep_locks = 1;
1634 
1635 	while(1) {
1636 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1637 					   sk->min_transid);
1638 		if (ret != 0) {
1639 			if (ret > 0)
1640 				ret = 0;
1641 			goto err;
1642 		}
1643 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1644 				 &sk_offset, &num_found);
1645 		btrfs_release_path(path);
1646 		if (ret || num_found >= sk->nr_items)
1647 			break;
1648 
1649 	}
1650 	ret = 0;
1651 err:
1652 	sk->nr_items = num_found;
1653 	btrfs_free_path(path);
1654 	return ret;
1655 }
1656 
1657 static noinline int btrfs_ioctl_tree_search(struct file *file,
1658 					   void __user *argp)
1659 {
1660 	 struct btrfs_ioctl_search_args *args;
1661 	 struct inode *inode;
1662 	 int ret;
1663 
1664 	if (!capable(CAP_SYS_ADMIN))
1665 		return -EPERM;
1666 
1667 	args = memdup_user(argp, sizeof(*args));
1668 	if (IS_ERR(args))
1669 		return PTR_ERR(args);
1670 
1671 	inode = fdentry(file)->d_inode;
1672 	ret = search_ioctl(inode, args);
1673 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1674 		ret = -EFAULT;
1675 	kfree(args);
1676 	return ret;
1677 }
1678 
1679 /*
1680  * Search INODE_REFs to identify path name of 'dirid' directory
1681  * in a 'tree_id' tree. and sets path name to 'name'.
1682  */
1683 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1684 				u64 tree_id, u64 dirid, char *name)
1685 {
1686 	struct btrfs_root *root;
1687 	struct btrfs_key key;
1688 	char *ptr;
1689 	int ret = -1;
1690 	int slot;
1691 	int len;
1692 	int total_len = 0;
1693 	struct btrfs_inode_ref *iref;
1694 	struct extent_buffer *l;
1695 	struct btrfs_path *path;
1696 
1697 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1698 		name[0]='\0';
1699 		return 0;
1700 	}
1701 
1702 	path = btrfs_alloc_path();
1703 	if (!path)
1704 		return -ENOMEM;
1705 
1706 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1707 
1708 	key.objectid = tree_id;
1709 	key.type = BTRFS_ROOT_ITEM_KEY;
1710 	key.offset = (u64)-1;
1711 	root = btrfs_read_fs_root_no_name(info, &key);
1712 	if (IS_ERR(root)) {
1713 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1714 		ret = -ENOENT;
1715 		goto out;
1716 	}
1717 
1718 	key.objectid = dirid;
1719 	key.type = BTRFS_INODE_REF_KEY;
1720 	key.offset = (u64)-1;
1721 
1722 	while(1) {
1723 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1724 		if (ret < 0)
1725 			goto out;
1726 
1727 		l = path->nodes[0];
1728 		slot = path->slots[0];
1729 		if (ret > 0 && slot > 0)
1730 			slot--;
1731 		btrfs_item_key_to_cpu(l, &key, slot);
1732 
1733 		if (ret > 0 && (key.objectid != dirid ||
1734 				key.type != BTRFS_INODE_REF_KEY)) {
1735 			ret = -ENOENT;
1736 			goto out;
1737 		}
1738 
1739 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1740 		len = btrfs_inode_ref_name_len(l, iref);
1741 		ptr -= len + 1;
1742 		total_len += len + 1;
1743 		if (ptr < name)
1744 			goto out;
1745 
1746 		*(ptr + len) = '/';
1747 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1748 
1749 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1750 			break;
1751 
1752 		btrfs_release_path(path);
1753 		key.objectid = key.offset;
1754 		key.offset = (u64)-1;
1755 		dirid = key.objectid;
1756 
1757 	}
1758 	if (ptr < name)
1759 		goto out;
1760 	memcpy(name, ptr, total_len);
1761 	name[total_len]='\0';
1762 	ret = 0;
1763 out:
1764 	btrfs_free_path(path);
1765 	return ret;
1766 }
1767 
1768 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1769 					   void __user *argp)
1770 {
1771 	 struct btrfs_ioctl_ino_lookup_args *args;
1772 	 struct inode *inode;
1773 	 int ret;
1774 
1775 	if (!capable(CAP_SYS_ADMIN))
1776 		return -EPERM;
1777 
1778 	args = memdup_user(argp, sizeof(*args));
1779 	if (IS_ERR(args))
1780 		return PTR_ERR(args);
1781 
1782 	inode = fdentry(file)->d_inode;
1783 
1784 	if (args->treeid == 0)
1785 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1786 
1787 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1788 					args->treeid, args->objectid,
1789 					args->name);
1790 
1791 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1792 		ret = -EFAULT;
1793 
1794 	kfree(args);
1795 	return ret;
1796 }
1797 
1798 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1799 					     void __user *arg)
1800 {
1801 	struct dentry *parent = fdentry(file);
1802 	struct dentry *dentry;
1803 	struct inode *dir = parent->d_inode;
1804 	struct inode *inode;
1805 	struct btrfs_root *root = BTRFS_I(dir)->root;
1806 	struct btrfs_root *dest = NULL;
1807 	struct btrfs_ioctl_vol_args *vol_args;
1808 	struct btrfs_trans_handle *trans;
1809 	int namelen;
1810 	int ret;
1811 	int err = 0;
1812 
1813 	vol_args = memdup_user(arg, sizeof(*vol_args));
1814 	if (IS_ERR(vol_args))
1815 		return PTR_ERR(vol_args);
1816 
1817 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1818 	namelen = strlen(vol_args->name);
1819 	if (strchr(vol_args->name, '/') ||
1820 	    strncmp(vol_args->name, "..", namelen) == 0) {
1821 		err = -EINVAL;
1822 		goto out;
1823 	}
1824 
1825 	err = mnt_want_write(file->f_path.mnt);
1826 	if (err)
1827 		goto out;
1828 
1829 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1830 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1831 	if (IS_ERR(dentry)) {
1832 		err = PTR_ERR(dentry);
1833 		goto out_unlock_dir;
1834 	}
1835 
1836 	if (!dentry->d_inode) {
1837 		err = -ENOENT;
1838 		goto out_dput;
1839 	}
1840 
1841 	inode = dentry->d_inode;
1842 	dest = BTRFS_I(inode)->root;
1843 	if (!capable(CAP_SYS_ADMIN)){
1844 		/*
1845 		 * Regular user.  Only allow this with a special mount
1846 		 * option, when the user has write+exec access to the
1847 		 * subvol root, and when rmdir(2) would have been
1848 		 * allowed.
1849 		 *
1850 		 * Note that this is _not_ check that the subvol is
1851 		 * empty or doesn't contain data that we wouldn't
1852 		 * otherwise be able to delete.
1853 		 *
1854 		 * Users who want to delete empty subvols should try
1855 		 * rmdir(2).
1856 		 */
1857 		err = -EPERM;
1858 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1859 			goto out_dput;
1860 
1861 		/*
1862 		 * Do not allow deletion if the parent dir is the same
1863 		 * as the dir to be deleted.  That means the ioctl
1864 		 * must be called on the dentry referencing the root
1865 		 * of the subvol, not a random directory contained
1866 		 * within it.
1867 		 */
1868 		err = -EINVAL;
1869 		if (root == dest)
1870 			goto out_dput;
1871 
1872 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1873 		if (err)
1874 			goto out_dput;
1875 
1876 		/* check if subvolume may be deleted by a non-root user */
1877 		err = btrfs_may_delete(dir, dentry, 1);
1878 		if (err)
1879 			goto out_dput;
1880 	}
1881 
1882 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1883 		err = -EINVAL;
1884 		goto out_dput;
1885 	}
1886 
1887 	mutex_lock(&inode->i_mutex);
1888 	err = d_invalidate(dentry);
1889 	if (err)
1890 		goto out_unlock;
1891 
1892 	down_write(&root->fs_info->subvol_sem);
1893 
1894 	err = may_destroy_subvol(dest);
1895 	if (err)
1896 		goto out_up_write;
1897 
1898 	trans = btrfs_start_transaction(root, 0);
1899 	if (IS_ERR(trans)) {
1900 		err = PTR_ERR(trans);
1901 		goto out_up_write;
1902 	}
1903 	trans->block_rsv = &root->fs_info->global_block_rsv;
1904 
1905 	ret = btrfs_unlink_subvol(trans, root, dir,
1906 				dest->root_key.objectid,
1907 				dentry->d_name.name,
1908 				dentry->d_name.len);
1909 	BUG_ON(ret);
1910 
1911 	btrfs_record_root_in_trans(trans, dest);
1912 
1913 	memset(&dest->root_item.drop_progress, 0,
1914 		sizeof(dest->root_item.drop_progress));
1915 	dest->root_item.drop_level = 0;
1916 	btrfs_set_root_refs(&dest->root_item, 0);
1917 
1918 	if (!xchg(&dest->orphan_item_inserted, 1)) {
1919 		ret = btrfs_insert_orphan_item(trans,
1920 					root->fs_info->tree_root,
1921 					dest->root_key.objectid);
1922 		BUG_ON(ret);
1923 	}
1924 
1925 	ret = btrfs_end_transaction(trans, root);
1926 	BUG_ON(ret);
1927 	inode->i_flags |= S_DEAD;
1928 out_up_write:
1929 	up_write(&root->fs_info->subvol_sem);
1930 out_unlock:
1931 	mutex_unlock(&inode->i_mutex);
1932 	if (!err) {
1933 		shrink_dcache_sb(root->fs_info->sb);
1934 		btrfs_invalidate_inodes(dest);
1935 		d_delete(dentry);
1936 	}
1937 out_dput:
1938 	dput(dentry);
1939 out_unlock_dir:
1940 	mutex_unlock(&dir->i_mutex);
1941 	mnt_drop_write(file->f_path.mnt);
1942 out:
1943 	kfree(vol_args);
1944 	return err;
1945 }
1946 
1947 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1948 {
1949 	struct inode *inode = fdentry(file)->d_inode;
1950 	struct btrfs_root *root = BTRFS_I(inode)->root;
1951 	struct btrfs_ioctl_defrag_range_args *range;
1952 	int ret;
1953 
1954 	if (btrfs_root_readonly(root))
1955 		return -EROFS;
1956 
1957 	ret = mnt_want_write(file->f_path.mnt);
1958 	if (ret)
1959 		return ret;
1960 
1961 	switch (inode->i_mode & S_IFMT) {
1962 	case S_IFDIR:
1963 		if (!capable(CAP_SYS_ADMIN)) {
1964 			ret = -EPERM;
1965 			goto out;
1966 		}
1967 		ret = btrfs_defrag_root(root, 0);
1968 		if (ret)
1969 			goto out;
1970 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1971 		break;
1972 	case S_IFREG:
1973 		if (!(file->f_mode & FMODE_WRITE)) {
1974 			ret = -EINVAL;
1975 			goto out;
1976 		}
1977 
1978 		range = kzalloc(sizeof(*range), GFP_KERNEL);
1979 		if (!range) {
1980 			ret = -ENOMEM;
1981 			goto out;
1982 		}
1983 
1984 		if (argp) {
1985 			if (copy_from_user(range, argp,
1986 					   sizeof(*range))) {
1987 				ret = -EFAULT;
1988 				kfree(range);
1989 				goto out;
1990 			}
1991 			/* compression requires us to start the IO */
1992 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1993 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1994 				range->extent_thresh = (u32)-1;
1995 			}
1996 		} else {
1997 			/* the rest are all set to zero by kzalloc */
1998 			range->len = (u64)-1;
1999 		}
2000 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2001 					range, 0, 0);
2002 		if (ret > 0)
2003 			ret = 0;
2004 		kfree(range);
2005 		break;
2006 	default:
2007 		ret = -EINVAL;
2008 	}
2009 out:
2010 	mnt_drop_write(file->f_path.mnt);
2011 	return ret;
2012 }
2013 
2014 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2015 {
2016 	struct btrfs_ioctl_vol_args *vol_args;
2017 	int ret;
2018 
2019 	if (!capable(CAP_SYS_ADMIN))
2020 		return -EPERM;
2021 
2022 	vol_args = memdup_user(arg, sizeof(*vol_args));
2023 	if (IS_ERR(vol_args))
2024 		return PTR_ERR(vol_args);
2025 
2026 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2027 	ret = btrfs_init_new_device(root, vol_args->name);
2028 
2029 	kfree(vol_args);
2030 	return ret;
2031 }
2032 
2033 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2034 {
2035 	struct btrfs_ioctl_vol_args *vol_args;
2036 	int ret;
2037 
2038 	if (!capable(CAP_SYS_ADMIN))
2039 		return -EPERM;
2040 
2041 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2042 		return -EROFS;
2043 
2044 	vol_args = memdup_user(arg, sizeof(*vol_args));
2045 	if (IS_ERR(vol_args))
2046 		return PTR_ERR(vol_args);
2047 
2048 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2049 	ret = btrfs_rm_device(root, vol_args->name);
2050 
2051 	kfree(vol_args);
2052 	return ret;
2053 }
2054 
2055 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2056 {
2057 	struct btrfs_ioctl_fs_info_args fi_args;
2058 	struct btrfs_device *device;
2059 	struct btrfs_device *next;
2060 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2061 
2062 	if (!capable(CAP_SYS_ADMIN))
2063 		return -EPERM;
2064 
2065 	fi_args.num_devices = fs_devices->num_devices;
2066 	fi_args.max_id = 0;
2067 	memcpy(&fi_args.fsid, root->fs_info->fsid, sizeof(fi_args.fsid));
2068 
2069 	mutex_lock(&fs_devices->device_list_mutex);
2070 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2071 		if (device->devid > fi_args.max_id)
2072 			fi_args.max_id = device->devid;
2073 	}
2074 	mutex_unlock(&fs_devices->device_list_mutex);
2075 
2076 	if (copy_to_user(arg, &fi_args, sizeof(fi_args)))
2077 		return -EFAULT;
2078 
2079 	return 0;
2080 }
2081 
2082 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2083 {
2084 	struct btrfs_ioctl_dev_info_args *di_args;
2085 	struct btrfs_device *dev;
2086 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2087 	int ret = 0;
2088 	char *s_uuid = NULL;
2089 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2090 
2091 	if (!capable(CAP_SYS_ADMIN))
2092 		return -EPERM;
2093 
2094 	di_args = memdup_user(arg, sizeof(*di_args));
2095 	if (IS_ERR(di_args))
2096 		return PTR_ERR(di_args);
2097 
2098 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2099 		s_uuid = di_args->uuid;
2100 
2101 	mutex_lock(&fs_devices->device_list_mutex);
2102 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2103 	mutex_unlock(&fs_devices->device_list_mutex);
2104 
2105 	if (!dev) {
2106 		ret = -ENODEV;
2107 		goto out;
2108 	}
2109 
2110 	di_args->devid = dev->devid;
2111 	di_args->bytes_used = dev->bytes_used;
2112 	di_args->total_bytes = dev->total_bytes;
2113 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2114 	strncpy(di_args->path, dev->name, sizeof(di_args->path));
2115 
2116 out:
2117 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2118 		ret = -EFAULT;
2119 
2120 	kfree(di_args);
2121 	return ret;
2122 }
2123 
2124 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2125 				       u64 off, u64 olen, u64 destoff)
2126 {
2127 	struct inode *inode = fdentry(file)->d_inode;
2128 	struct btrfs_root *root = BTRFS_I(inode)->root;
2129 	struct file *src_file;
2130 	struct inode *src;
2131 	struct btrfs_trans_handle *trans;
2132 	struct btrfs_path *path;
2133 	struct extent_buffer *leaf;
2134 	char *buf;
2135 	struct btrfs_key key;
2136 	u32 nritems;
2137 	int slot;
2138 	int ret;
2139 	u64 len = olen;
2140 	u64 bs = root->fs_info->sb->s_blocksize;
2141 	u64 hint_byte;
2142 
2143 	/*
2144 	 * TODO:
2145 	 * - split compressed inline extents.  annoying: we need to
2146 	 *   decompress into destination's address_space (the file offset
2147 	 *   may change, so source mapping won't do), then recompress (or
2148 	 *   otherwise reinsert) a subrange.
2149 	 * - allow ranges within the same file to be cloned (provided
2150 	 *   they don't overlap)?
2151 	 */
2152 
2153 	/* the destination must be opened for writing */
2154 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2155 		return -EINVAL;
2156 
2157 	if (btrfs_root_readonly(root))
2158 		return -EROFS;
2159 
2160 	ret = mnt_want_write(file->f_path.mnt);
2161 	if (ret)
2162 		return ret;
2163 
2164 	src_file = fget(srcfd);
2165 	if (!src_file) {
2166 		ret = -EBADF;
2167 		goto out_drop_write;
2168 	}
2169 
2170 	src = src_file->f_dentry->d_inode;
2171 
2172 	ret = -EINVAL;
2173 	if (src == inode)
2174 		goto out_fput;
2175 
2176 	/* the src must be open for reading */
2177 	if (!(src_file->f_mode & FMODE_READ))
2178 		goto out_fput;
2179 
2180 	ret = -EISDIR;
2181 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2182 		goto out_fput;
2183 
2184 	ret = -EXDEV;
2185 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2186 		goto out_fput;
2187 
2188 	ret = -ENOMEM;
2189 	buf = vmalloc(btrfs_level_size(root, 0));
2190 	if (!buf)
2191 		goto out_fput;
2192 
2193 	path = btrfs_alloc_path();
2194 	if (!path) {
2195 		vfree(buf);
2196 		goto out_fput;
2197 	}
2198 	path->reada = 2;
2199 
2200 	if (inode < src) {
2201 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2202 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2203 	} else {
2204 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2205 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2206 	}
2207 
2208 	/* determine range to clone */
2209 	ret = -EINVAL;
2210 	if (off + len > src->i_size || off + len < off)
2211 		goto out_unlock;
2212 	if (len == 0)
2213 		olen = len = src->i_size - off;
2214 	/* if we extend to eof, continue to block boundary */
2215 	if (off + len == src->i_size)
2216 		len = ALIGN(src->i_size, bs) - off;
2217 
2218 	/* verify the end result is block aligned */
2219 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2220 	    !IS_ALIGNED(destoff, bs))
2221 		goto out_unlock;
2222 
2223 	/* do any pending delalloc/csum calc on src, one way or
2224 	   another, and lock file content */
2225 	while (1) {
2226 		struct btrfs_ordered_extent *ordered;
2227 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2228 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2229 		if (!ordered &&
2230 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2231 				   EXTENT_DELALLOC, 0, NULL))
2232 			break;
2233 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2234 		if (ordered)
2235 			btrfs_put_ordered_extent(ordered);
2236 		btrfs_wait_ordered_range(src, off, len);
2237 	}
2238 
2239 	/* clone data */
2240 	key.objectid = btrfs_ino(src);
2241 	key.type = BTRFS_EXTENT_DATA_KEY;
2242 	key.offset = 0;
2243 
2244 	while (1) {
2245 		/*
2246 		 * note the key will change type as we walk through the
2247 		 * tree.
2248 		 */
2249 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2250 		if (ret < 0)
2251 			goto out;
2252 
2253 		nritems = btrfs_header_nritems(path->nodes[0]);
2254 		if (path->slots[0] >= nritems) {
2255 			ret = btrfs_next_leaf(root, path);
2256 			if (ret < 0)
2257 				goto out;
2258 			if (ret > 0)
2259 				break;
2260 			nritems = btrfs_header_nritems(path->nodes[0]);
2261 		}
2262 		leaf = path->nodes[0];
2263 		slot = path->slots[0];
2264 
2265 		btrfs_item_key_to_cpu(leaf, &key, slot);
2266 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2267 		    key.objectid != btrfs_ino(src))
2268 			break;
2269 
2270 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2271 			struct btrfs_file_extent_item *extent;
2272 			int type;
2273 			u32 size;
2274 			struct btrfs_key new_key;
2275 			u64 disko = 0, diskl = 0;
2276 			u64 datao = 0, datal = 0;
2277 			u8 comp;
2278 			u64 endoff;
2279 
2280 			size = btrfs_item_size_nr(leaf, slot);
2281 			read_extent_buffer(leaf, buf,
2282 					   btrfs_item_ptr_offset(leaf, slot),
2283 					   size);
2284 
2285 			extent = btrfs_item_ptr(leaf, slot,
2286 						struct btrfs_file_extent_item);
2287 			comp = btrfs_file_extent_compression(leaf, extent);
2288 			type = btrfs_file_extent_type(leaf, extent);
2289 			if (type == BTRFS_FILE_EXTENT_REG ||
2290 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2291 				disko = btrfs_file_extent_disk_bytenr(leaf,
2292 								      extent);
2293 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2294 								 extent);
2295 				datao = btrfs_file_extent_offset(leaf, extent);
2296 				datal = btrfs_file_extent_num_bytes(leaf,
2297 								    extent);
2298 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2299 				/* take upper bound, may be compressed */
2300 				datal = btrfs_file_extent_ram_bytes(leaf,
2301 								    extent);
2302 			}
2303 			btrfs_release_path(path);
2304 
2305 			if (key.offset + datal <= off ||
2306 			    key.offset >= off+len)
2307 				goto next;
2308 
2309 			memcpy(&new_key, &key, sizeof(new_key));
2310 			new_key.objectid = btrfs_ino(inode);
2311 			if (off <= key.offset)
2312 				new_key.offset = key.offset + destoff - off;
2313 			else
2314 				new_key.offset = destoff;
2315 
2316 			trans = btrfs_start_transaction(root, 1);
2317 			if (IS_ERR(trans)) {
2318 				ret = PTR_ERR(trans);
2319 				goto out;
2320 			}
2321 
2322 			if (type == BTRFS_FILE_EXTENT_REG ||
2323 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2324 				if (off > key.offset) {
2325 					datao += off - key.offset;
2326 					datal -= off - key.offset;
2327 				}
2328 
2329 				if (key.offset + datal > off + len)
2330 					datal = off + len - key.offset;
2331 
2332 				ret = btrfs_drop_extents(trans, inode,
2333 							 new_key.offset,
2334 							 new_key.offset + datal,
2335 							 &hint_byte, 1);
2336 				BUG_ON(ret);
2337 
2338 				ret = btrfs_insert_empty_item(trans, root, path,
2339 							      &new_key, size);
2340 				BUG_ON(ret);
2341 
2342 				leaf = path->nodes[0];
2343 				slot = path->slots[0];
2344 				write_extent_buffer(leaf, buf,
2345 					    btrfs_item_ptr_offset(leaf, slot),
2346 					    size);
2347 
2348 				extent = btrfs_item_ptr(leaf, slot,
2349 						struct btrfs_file_extent_item);
2350 
2351 				/* disko == 0 means it's a hole */
2352 				if (!disko)
2353 					datao = 0;
2354 
2355 				btrfs_set_file_extent_offset(leaf, extent,
2356 							     datao);
2357 				btrfs_set_file_extent_num_bytes(leaf, extent,
2358 								datal);
2359 				if (disko) {
2360 					inode_add_bytes(inode, datal);
2361 					ret = btrfs_inc_extent_ref(trans, root,
2362 							disko, diskl, 0,
2363 							root->root_key.objectid,
2364 							btrfs_ino(inode),
2365 							new_key.offset - datao);
2366 					BUG_ON(ret);
2367 				}
2368 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2369 				u64 skip = 0;
2370 				u64 trim = 0;
2371 				if (off > key.offset) {
2372 					skip = off - key.offset;
2373 					new_key.offset += skip;
2374 				}
2375 
2376 				if (key.offset + datal > off+len)
2377 					trim = key.offset + datal - (off+len);
2378 
2379 				if (comp && (skip || trim)) {
2380 					ret = -EINVAL;
2381 					btrfs_end_transaction(trans, root);
2382 					goto out;
2383 				}
2384 				size -= skip + trim;
2385 				datal -= skip + trim;
2386 
2387 				ret = btrfs_drop_extents(trans, inode,
2388 							 new_key.offset,
2389 							 new_key.offset + datal,
2390 							 &hint_byte, 1);
2391 				BUG_ON(ret);
2392 
2393 				ret = btrfs_insert_empty_item(trans, root, path,
2394 							      &new_key, size);
2395 				BUG_ON(ret);
2396 
2397 				if (skip) {
2398 					u32 start =
2399 					  btrfs_file_extent_calc_inline_size(0);
2400 					memmove(buf+start, buf+start+skip,
2401 						datal);
2402 				}
2403 
2404 				leaf = path->nodes[0];
2405 				slot = path->slots[0];
2406 				write_extent_buffer(leaf, buf,
2407 					    btrfs_item_ptr_offset(leaf, slot),
2408 					    size);
2409 				inode_add_bytes(inode, datal);
2410 			}
2411 
2412 			btrfs_mark_buffer_dirty(leaf);
2413 			btrfs_release_path(path);
2414 
2415 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2416 
2417 			/*
2418 			 * we round up to the block size at eof when
2419 			 * determining which extents to clone above,
2420 			 * but shouldn't round up the file size
2421 			 */
2422 			endoff = new_key.offset + datal;
2423 			if (endoff > destoff+olen)
2424 				endoff = destoff+olen;
2425 			if (endoff > inode->i_size)
2426 				btrfs_i_size_write(inode, endoff);
2427 
2428 			BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
2429 			ret = btrfs_update_inode(trans, root, inode);
2430 			BUG_ON(ret);
2431 			btrfs_end_transaction(trans, root);
2432 		}
2433 next:
2434 		btrfs_release_path(path);
2435 		key.offset++;
2436 	}
2437 	ret = 0;
2438 out:
2439 	btrfs_release_path(path);
2440 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2441 out_unlock:
2442 	mutex_unlock(&src->i_mutex);
2443 	mutex_unlock(&inode->i_mutex);
2444 	vfree(buf);
2445 	btrfs_free_path(path);
2446 out_fput:
2447 	fput(src_file);
2448 out_drop_write:
2449 	mnt_drop_write(file->f_path.mnt);
2450 	return ret;
2451 }
2452 
2453 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2454 {
2455 	struct btrfs_ioctl_clone_range_args args;
2456 
2457 	if (copy_from_user(&args, argp, sizeof(args)))
2458 		return -EFAULT;
2459 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2460 				 args.src_length, args.dest_offset);
2461 }
2462 
2463 /*
2464  * there are many ways the trans_start and trans_end ioctls can lead
2465  * to deadlocks.  They should only be used by applications that
2466  * basically own the machine, and have a very in depth understanding
2467  * of all the possible deadlocks and enospc problems.
2468  */
2469 static long btrfs_ioctl_trans_start(struct file *file)
2470 {
2471 	struct inode *inode = fdentry(file)->d_inode;
2472 	struct btrfs_root *root = BTRFS_I(inode)->root;
2473 	struct btrfs_trans_handle *trans;
2474 	int ret;
2475 
2476 	ret = -EPERM;
2477 	if (!capable(CAP_SYS_ADMIN))
2478 		goto out;
2479 
2480 	ret = -EINPROGRESS;
2481 	if (file->private_data)
2482 		goto out;
2483 
2484 	ret = -EROFS;
2485 	if (btrfs_root_readonly(root))
2486 		goto out;
2487 
2488 	ret = mnt_want_write(file->f_path.mnt);
2489 	if (ret)
2490 		goto out;
2491 
2492 	atomic_inc(&root->fs_info->open_ioctl_trans);
2493 
2494 	ret = -ENOMEM;
2495 	trans = btrfs_start_ioctl_transaction(root);
2496 	if (IS_ERR(trans))
2497 		goto out_drop;
2498 
2499 	file->private_data = trans;
2500 	return 0;
2501 
2502 out_drop:
2503 	atomic_dec(&root->fs_info->open_ioctl_trans);
2504 	mnt_drop_write(file->f_path.mnt);
2505 out:
2506 	return ret;
2507 }
2508 
2509 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2510 {
2511 	struct inode *inode = fdentry(file)->d_inode;
2512 	struct btrfs_root *root = BTRFS_I(inode)->root;
2513 	struct btrfs_root *new_root;
2514 	struct btrfs_dir_item *di;
2515 	struct btrfs_trans_handle *trans;
2516 	struct btrfs_path *path;
2517 	struct btrfs_key location;
2518 	struct btrfs_disk_key disk_key;
2519 	struct btrfs_super_block *disk_super;
2520 	u64 features;
2521 	u64 objectid = 0;
2522 	u64 dir_id;
2523 
2524 	if (!capable(CAP_SYS_ADMIN))
2525 		return -EPERM;
2526 
2527 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2528 		return -EFAULT;
2529 
2530 	if (!objectid)
2531 		objectid = root->root_key.objectid;
2532 
2533 	location.objectid = objectid;
2534 	location.type = BTRFS_ROOT_ITEM_KEY;
2535 	location.offset = (u64)-1;
2536 
2537 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2538 	if (IS_ERR(new_root))
2539 		return PTR_ERR(new_root);
2540 
2541 	if (btrfs_root_refs(&new_root->root_item) == 0)
2542 		return -ENOENT;
2543 
2544 	path = btrfs_alloc_path();
2545 	if (!path)
2546 		return -ENOMEM;
2547 	path->leave_spinning = 1;
2548 
2549 	trans = btrfs_start_transaction(root, 1);
2550 	if (IS_ERR(trans)) {
2551 		btrfs_free_path(path);
2552 		return PTR_ERR(trans);
2553 	}
2554 
2555 	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2556 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2557 				   dir_id, "default", 7, 1);
2558 	if (IS_ERR_OR_NULL(di)) {
2559 		btrfs_free_path(path);
2560 		btrfs_end_transaction(trans, root);
2561 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2562 		       "this isn't going to work\n");
2563 		return -ENOENT;
2564 	}
2565 
2566 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2567 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2568 	btrfs_mark_buffer_dirty(path->nodes[0]);
2569 	btrfs_free_path(path);
2570 
2571 	disk_super = &root->fs_info->super_copy;
2572 	features = btrfs_super_incompat_flags(disk_super);
2573 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2574 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2575 		btrfs_set_super_incompat_flags(disk_super, features);
2576 	}
2577 	btrfs_end_transaction(trans, root);
2578 
2579 	return 0;
2580 }
2581 
2582 static void get_block_group_info(struct list_head *groups_list,
2583 				 struct btrfs_ioctl_space_info *space)
2584 {
2585 	struct btrfs_block_group_cache *block_group;
2586 
2587 	space->total_bytes = 0;
2588 	space->used_bytes = 0;
2589 	space->flags = 0;
2590 	list_for_each_entry(block_group, groups_list, list) {
2591 		space->flags = block_group->flags;
2592 		space->total_bytes += block_group->key.offset;
2593 		space->used_bytes +=
2594 			btrfs_block_group_used(&block_group->item);
2595 	}
2596 }
2597 
2598 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2599 {
2600 	struct btrfs_ioctl_space_args space_args;
2601 	struct btrfs_ioctl_space_info space;
2602 	struct btrfs_ioctl_space_info *dest;
2603 	struct btrfs_ioctl_space_info *dest_orig;
2604 	struct btrfs_ioctl_space_info __user *user_dest;
2605 	struct btrfs_space_info *info;
2606 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2607 		       BTRFS_BLOCK_GROUP_SYSTEM,
2608 		       BTRFS_BLOCK_GROUP_METADATA,
2609 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2610 	int num_types = 4;
2611 	int alloc_size;
2612 	int ret = 0;
2613 	u64 slot_count = 0;
2614 	int i, c;
2615 
2616 	if (copy_from_user(&space_args,
2617 			   (struct btrfs_ioctl_space_args __user *)arg,
2618 			   sizeof(space_args)))
2619 		return -EFAULT;
2620 
2621 	for (i = 0; i < num_types; i++) {
2622 		struct btrfs_space_info *tmp;
2623 
2624 		info = NULL;
2625 		rcu_read_lock();
2626 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2627 					list) {
2628 			if (tmp->flags == types[i]) {
2629 				info = tmp;
2630 				break;
2631 			}
2632 		}
2633 		rcu_read_unlock();
2634 
2635 		if (!info)
2636 			continue;
2637 
2638 		down_read(&info->groups_sem);
2639 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2640 			if (!list_empty(&info->block_groups[c]))
2641 				slot_count++;
2642 		}
2643 		up_read(&info->groups_sem);
2644 	}
2645 
2646 	/* space_slots == 0 means they are asking for a count */
2647 	if (space_args.space_slots == 0) {
2648 		space_args.total_spaces = slot_count;
2649 		goto out;
2650 	}
2651 
2652 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2653 
2654 	alloc_size = sizeof(*dest) * slot_count;
2655 
2656 	/* we generally have at most 6 or so space infos, one for each raid
2657 	 * level.  So, a whole page should be more than enough for everyone
2658 	 */
2659 	if (alloc_size > PAGE_CACHE_SIZE)
2660 		return -ENOMEM;
2661 
2662 	space_args.total_spaces = 0;
2663 	dest = kmalloc(alloc_size, GFP_NOFS);
2664 	if (!dest)
2665 		return -ENOMEM;
2666 	dest_orig = dest;
2667 
2668 	/* now we have a buffer to copy into */
2669 	for (i = 0; i < num_types; i++) {
2670 		struct btrfs_space_info *tmp;
2671 
2672 		if (!slot_count)
2673 			break;
2674 
2675 		info = NULL;
2676 		rcu_read_lock();
2677 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2678 					list) {
2679 			if (tmp->flags == types[i]) {
2680 				info = tmp;
2681 				break;
2682 			}
2683 		}
2684 		rcu_read_unlock();
2685 
2686 		if (!info)
2687 			continue;
2688 		down_read(&info->groups_sem);
2689 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2690 			if (!list_empty(&info->block_groups[c])) {
2691 				get_block_group_info(&info->block_groups[c],
2692 						     &space);
2693 				memcpy(dest, &space, sizeof(space));
2694 				dest++;
2695 				space_args.total_spaces++;
2696 				slot_count--;
2697 			}
2698 			if (!slot_count)
2699 				break;
2700 		}
2701 		up_read(&info->groups_sem);
2702 	}
2703 
2704 	user_dest = (struct btrfs_ioctl_space_info *)
2705 		(arg + sizeof(struct btrfs_ioctl_space_args));
2706 
2707 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2708 		ret = -EFAULT;
2709 
2710 	kfree(dest_orig);
2711 out:
2712 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2713 		ret = -EFAULT;
2714 
2715 	return ret;
2716 }
2717 
2718 /*
2719  * there are many ways the trans_start and trans_end ioctls can lead
2720  * to deadlocks.  They should only be used by applications that
2721  * basically own the machine, and have a very in depth understanding
2722  * of all the possible deadlocks and enospc problems.
2723  */
2724 long btrfs_ioctl_trans_end(struct file *file)
2725 {
2726 	struct inode *inode = fdentry(file)->d_inode;
2727 	struct btrfs_root *root = BTRFS_I(inode)->root;
2728 	struct btrfs_trans_handle *trans;
2729 
2730 	trans = file->private_data;
2731 	if (!trans)
2732 		return -EINVAL;
2733 	file->private_data = NULL;
2734 
2735 	btrfs_end_transaction(trans, root);
2736 
2737 	atomic_dec(&root->fs_info->open_ioctl_trans);
2738 
2739 	mnt_drop_write(file->f_path.mnt);
2740 	return 0;
2741 }
2742 
2743 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2744 {
2745 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2746 	struct btrfs_trans_handle *trans;
2747 	u64 transid;
2748 	int ret;
2749 
2750 	trans = btrfs_start_transaction(root, 0);
2751 	if (IS_ERR(trans))
2752 		return PTR_ERR(trans);
2753 	transid = trans->transid;
2754 	ret = btrfs_commit_transaction_async(trans, root, 0);
2755 	if (ret) {
2756 		btrfs_end_transaction(trans, root);
2757 		return ret;
2758 	}
2759 
2760 	if (argp)
2761 		if (copy_to_user(argp, &transid, sizeof(transid)))
2762 			return -EFAULT;
2763 	return 0;
2764 }
2765 
2766 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2767 {
2768 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2769 	u64 transid;
2770 
2771 	if (argp) {
2772 		if (copy_from_user(&transid, argp, sizeof(transid)))
2773 			return -EFAULT;
2774 	} else {
2775 		transid = 0;  /* current trans */
2776 	}
2777 	return btrfs_wait_for_commit(root, transid);
2778 }
2779 
2780 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2781 {
2782 	int ret;
2783 	struct btrfs_ioctl_scrub_args *sa;
2784 
2785 	if (!capable(CAP_SYS_ADMIN))
2786 		return -EPERM;
2787 
2788 	sa = memdup_user(arg, sizeof(*sa));
2789 	if (IS_ERR(sa))
2790 		return PTR_ERR(sa);
2791 
2792 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2793 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2794 
2795 	if (copy_to_user(arg, sa, sizeof(*sa)))
2796 		ret = -EFAULT;
2797 
2798 	kfree(sa);
2799 	return ret;
2800 }
2801 
2802 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2803 {
2804 	if (!capable(CAP_SYS_ADMIN))
2805 		return -EPERM;
2806 
2807 	return btrfs_scrub_cancel(root);
2808 }
2809 
2810 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2811 				       void __user *arg)
2812 {
2813 	struct btrfs_ioctl_scrub_args *sa;
2814 	int ret;
2815 
2816 	if (!capable(CAP_SYS_ADMIN))
2817 		return -EPERM;
2818 
2819 	sa = memdup_user(arg, sizeof(*sa));
2820 	if (IS_ERR(sa))
2821 		return PTR_ERR(sa);
2822 
2823 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2824 
2825 	if (copy_to_user(arg, sa, sizeof(*sa)))
2826 		ret = -EFAULT;
2827 
2828 	kfree(sa);
2829 	return ret;
2830 }
2831 
2832 long btrfs_ioctl(struct file *file, unsigned int
2833 		cmd, unsigned long arg)
2834 {
2835 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2836 	void __user *argp = (void __user *)arg;
2837 
2838 	switch (cmd) {
2839 	case FS_IOC_GETFLAGS:
2840 		return btrfs_ioctl_getflags(file, argp);
2841 	case FS_IOC_SETFLAGS:
2842 		return btrfs_ioctl_setflags(file, argp);
2843 	case FS_IOC_GETVERSION:
2844 		return btrfs_ioctl_getversion(file, argp);
2845 	case FITRIM:
2846 		return btrfs_ioctl_fitrim(file, argp);
2847 	case BTRFS_IOC_SNAP_CREATE:
2848 		return btrfs_ioctl_snap_create(file, argp, 0);
2849 	case BTRFS_IOC_SNAP_CREATE_V2:
2850 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
2851 	case BTRFS_IOC_SUBVOL_CREATE:
2852 		return btrfs_ioctl_snap_create(file, argp, 1);
2853 	case BTRFS_IOC_SNAP_DESTROY:
2854 		return btrfs_ioctl_snap_destroy(file, argp);
2855 	case BTRFS_IOC_SUBVOL_GETFLAGS:
2856 		return btrfs_ioctl_subvol_getflags(file, argp);
2857 	case BTRFS_IOC_SUBVOL_SETFLAGS:
2858 		return btrfs_ioctl_subvol_setflags(file, argp);
2859 	case BTRFS_IOC_DEFAULT_SUBVOL:
2860 		return btrfs_ioctl_default_subvol(file, argp);
2861 	case BTRFS_IOC_DEFRAG:
2862 		return btrfs_ioctl_defrag(file, NULL);
2863 	case BTRFS_IOC_DEFRAG_RANGE:
2864 		return btrfs_ioctl_defrag(file, argp);
2865 	case BTRFS_IOC_RESIZE:
2866 		return btrfs_ioctl_resize(root, argp);
2867 	case BTRFS_IOC_ADD_DEV:
2868 		return btrfs_ioctl_add_dev(root, argp);
2869 	case BTRFS_IOC_RM_DEV:
2870 		return btrfs_ioctl_rm_dev(root, argp);
2871 	case BTRFS_IOC_FS_INFO:
2872 		return btrfs_ioctl_fs_info(root, argp);
2873 	case BTRFS_IOC_DEV_INFO:
2874 		return btrfs_ioctl_dev_info(root, argp);
2875 	case BTRFS_IOC_BALANCE:
2876 		return btrfs_balance(root->fs_info->dev_root);
2877 	case BTRFS_IOC_CLONE:
2878 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2879 	case BTRFS_IOC_CLONE_RANGE:
2880 		return btrfs_ioctl_clone_range(file, argp);
2881 	case BTRFS_IOC_TRANS_START:
2882 		return btrfs_ioctl_trans_start(file);
2883 	case BTRFS_IOC_TRANS_END:
2884 		return btrfs_ioctl_trans_end(file);
2885 	case BTRFS_IOC_TREE_SEARCH:
2886 		return btrfs_ioctl_tree_search(file, argp);
2887 	case BTRFS_IOC_INO_LOOKUP:
2888 		return btrfs_ioctl_ino_lookup(file, argp);
2889 	case BTRFS_IOC_SPACE_INFO:
2890 		return btrfs_ioctl_space_info(root, argp);
2891 	case BTRFS_IOC_SYNC:
2892 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
2893 		return 0;
2894 	case BTRFS_IOC_START_SYNC:
2895 		return btrfs_ioctl_start_sync(file, argp);
2896 	case BTRFS_IOC_WAIT_SYNC:
2897 		return btrfs_ioctl_wait_sync(file, argp);
2898 	case BTRFS_IOC_SCRUB:
2899 		return btrfs_ioctl_scrub(root, argp);
2900 	case BTRFS_IOC_SCRUB_CANCEL:
2901 		return btrfs_ioctl_scrub_cancel(root, argp);
2902 	case BTRFS_IOC_SCRUB_PROGRESS:
2903 		return btrfs_ioctl_scrub_progress(root, argp);
2904 	}
2905 
2906 	return -ENOTTY;
2907 }
2908