xref: /linux/fs/btrfs/ioctl.c (revision d229807f669ba3dea9f64467ee965051c4366aed)
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 = dentry->d_parent;
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 		return ret;
337 
338 	dir = parent->d_inode;
339 
340 	/*
341 	 * 1 - inode item
342 	 * 2 - refs
343 	 * 1 - root item
344 	 * 2 - dir items
345 	 */
346 	trans = btrfs_start_transaction(root, 6);
347 	if (IS_ERR(trans))
348 		return PTR_ERR(trans);
349 
350 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
351 				      0, objectid, NULL, 0, 0, 0);
352 	if (IS_ERR(leaf)) {
353 		ret = PTR_ERR(leaf);
354 		goto fail;
355 	}
356 
357 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
358 	btrfs_set_header_bytenr(leaf, leaf->start);
359 	btrfs_set_header_generation(leaf, trans->transid);
360 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
361 	btrfs_set_header_owner(leaf, objectid);
362 
363 	write_extent_buffer(leaf, root->fs_info->fsid,
364 			    (unsigned long)btrfs_header_fsid(leaf),
365 			    BTRFS_FSID_SIZE);
366 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
367 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
368 			    BTRFS_UUID_SIZE);
369 	btrfs_mark_buffer_dirty(leaf);
370 
371 	inode_item = &root_item.inode;
372 	memset(inode_item, 0, sizeof(*inode_item));
373 	inode_item->generation = cpu_to_le64(1);
374 	inode_item->size = cpu_to_le64(3);
375 	inode_item->nlink = cpu_to_le32(1);
376 	inode_item->nbytes = cpu_to_le64(root->leafsize);
377 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
378 
379 	root_item.flags = 0;
380 	root_item.byte_limit = 0;
381 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
382 
383 	btrfs_set_root_bytenr(&root_item, leaf->start);
384 	btrfs_set_root_generation(&root_item, trans->transid);
385 	btrfs_set_root_level(&root_item, 0);
386 	btrfs_set_root_refs(&root_item, 1);
387 	btrfs_set_root_used(&root_item, leaf->len);
388 	btrfs_set_root_last_snapshot(&root_item, 0);
389 
390 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
391 	root_item.drop_level = 0;
392 
393 	btrfs_tree_unlock(leaf);
394 	free_extent_buffer(leaf);
395 	leaf = NULL;
396 
397 	btrfs_set_root_dirid(&root_item, new_dirid);
398 
399 	key.objectid = objectid;
400 	key.offset = 0;
401 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
402 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
403 				&root_item);
404 	if (ret)
405 		goto fail;
406 
407 	key.offset = (u64)-1;
408 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
409 	BUG_ON(IS_ERR(new_root));
410 
411 	btrfs_record_root_in_trans(trans, new_root);
412 
413 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
414 	/*
415 	 * insert the directory item
416 	 */
417 	ret = btrfs_set_inode_index(dir, &index);
418 	BUG_ON(ret);
419 
420 	ret = btrfs_insert_dir_item(trans, root,
421 				    name, namelen, dir, &key,
422 				    BTRFS_FT_DIR, index);
423 	if (ret)
424 		goto fail;
425 
426 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
427 	ret = btrfs_update_inode(trans, root, dir);
428 	BUG_ON(ret);
429 
430 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
431 				 objectid, root->root_key.objectid,
432 				 btrfs_ino(dir), index, name, namelen);
433 
434 	BUG_ON(ret);
435 
436 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
437 fail:
438 	if (async_transid) {
439 		*async_transid = trans->transid;
440 		err = btrfs_commit_transaction_async(trans, root, 1);
441 	} else {
442 		err = btrfs_commit_transaction(trans, root);
443 	}
444 	if (err && !ret)
445 		ret = err;
446 	return ret;
447 }
448 
449 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
450 			   char *name, int namelen, u64 *async_transid,
451 			   bool readonly)
452 {
453 	struct inode *inode;
454 	struct btrfs_pending_snapshot *pending_snapshot;
455 	struct btrfs_trans_handle *trans;
456 	int ret;
457 
458 	if (!root->ref_cows)
459 		return -EINVAL;
460 
461 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
462 	if (!pending_snapshot)
463 		return -ENOMEM;
464 
465 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
466 	pending_snapshot->dentry = dentry;
467 	pending_snapshot->root = root;
468 	pending_snapshot->readonly = readonly;
469 
470 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
471 	if (IS_ERR(trans)) {
472 		ret = PTR_ERR(trans);
473 		goto fail;
474 	}
475 
476 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
477 	BUG_ON(ret);
478 
479 	spin_lock(&root->fs_info->trans_lock);
480 	list_add(&pending_snapshot->list,
481 		 &trans->transaction->pending_snapshots);
482 	spin_unlock(&root->fs_info->trans_lock);
483 	if (async_transid) {
484 		*async_transid = trans->transid;
485 		ret = btrfs_commit_transaction_async(trans,
486 				     root->fs_info->extent_root, 1);
487 	} else {
488 		ret = btrfs_commit_transaction(trans,
489 					       root->fs_info->extent_root);
490 	}
491 	BUG_ON(ret);
492 
493 	ret = pending_snapshot->error;
494 	if (ret)
495 		goto fail;
496 
497 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
498 	if (ret)
499 		goto fail;
500 
501 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
502 	if (IS_ERR(inode)) {
503 		ret = PTR_ERR(inode);
504 		goto fail;
505 	}
506 	BUG_ON(!inode);
507 	d_instantiate(dentry, inode);
508 	ret = 0;
509 fail:
510 	kfree(pending_snapshot);
511 	return ret;
512 }
513 
514 /*  copy of check_sticky in fs/namei.c()
515 * It's inline, so penalty for filesystems that don't use sticky bit is
516 * minimal.
517 */
518 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
519 {
520 	uid_t fsuid = current_fsuid();
521 
522 	if (!(dir->i_mode & S_ISVTX))
523 		return 0;
524 	if (inode->i_uid == fsuid)
525 		return 0;
526 	if (dir->i_uid == fsuid)
527 		return 0;
528 	return !capable(CAP_FOWNER);
529 }
530 
531 /*  copy of may_delete in fs/namei.c()
532  *	Check whether we can remove a link victim from directory dir, check
533  *  whether the type of victim is right.
534  *  1. We can't do it if dir is read-only (done in permission())
535  *  2. We should have write and exec permissions on dir
536  *  3. We can't remove anything from append-only dir
537  *  4. We can't do anything with immutable dir (done in permission())
538  *  5. If the sticky bit on dir is set we should either
539  *	a. be owner of dir, or
540  *	b. be owner of victim, or
541  *	c. have CAP_FOWNER capability
542  *  6. If the victim is append-only or immutable we can't do antyhing with
543  *     links pointing to it.
544  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
545  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
546  *  9. We can't remove a root or mountpoint.
547  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
548  *     nfs_async_unlink().
549  */
550 
551 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
552 {
553 	int error;
554 
555 	if (!victim->d_inode)
556 		return -ENOENT;
557 
558 	BUG_ON(victim->d_parent->d_inode != dir);
559 	audit_inode_child(victim, dir);
560 
561 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
562 	if (error)
563 		return error;
564 	if (IS_APPEND(dir))
565 		return -EPERM;
566 	if (btrfs_check_sticky(dir, victim->d_inode)||
567 		IS_APPEND(victim->d_inode)||
568 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
569 		return -EPERM;
570 	if (isdir) {
571 		if (!S_ISDIR(victim->d_inode->i_mode))
572 			return -ENOTDIR;
573 		if (IS_ROOT(victim))
574 			return -EBUSY;
575 	} else if (S_ISDIR(victim->d_inode->i_mode))
576 		return -EISDIR;
577 	if (IS_DEADDIR(dir))
578 		return -ENOENT;
579 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
580 		return -EBUSY;
581 	return 0;
582 }
583 
584 /* copy of may_create in fs/namei.c() */
585 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
586 {
587 	if (child->d_inode)
588 		return -EEXIST;
589 	if (IS_DEADDIR(dir))
590 		return -ENOENT;
591 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
592 }
593 
594 /*
595  * Create a new subvolume below @parent.  This is largely modeled after
596  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
597  * inside this filesystem so it's quite a bit simpler.
598  */
599 static noinline int btrfs_mksubvol(struct path *parent,
600 				   char *name, int namelen,
601 				   struct btrfs_root *snap_src,
602 				   u64 *async_transid, bool readonly)
603 {
604 	struct inode *dir  = parent->dentry->d_inode;
605 	struct dentry *dentry;
606 	int error;
607 
608 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
609 
610 	dentry = lookup_one_len(name, parent->dentry, namelen);
611 	error = PTR_ERR(dentry);
612 	if (IS_ERR(dentry))
613 		goto out_unlock;
614 
615 	error = -EEXIST;
616 	if (dentry->d_inode)
617 		goto out_dput;
618 
619 	error = mnt_want_write(parent->mnt);
620 	if (error)
621 		goto out_dput;
622 
623 	error = btrfs_may_create(dir, dentry);
624 	if (error)
625 		goto out_drop_write;
626 
627 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
628 
629 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
630 		goto out_up_read;
631 
632 	if (snap_src) {
633 		error = create_snapshot(snap_src, dentry,
634 					name, namelen, async_transid, readonly);
635 	} else {
636 		error = create_subvol(BTRFS_I(dir)->root, dentry,
637 				      name, namelen, async_transid);
638 	}
639 	if (!error)
640 		fsnotify_mkdir(dir, dentry);
641 out_up_read:
642 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
643 out_drop_write:
644 	mnt_drop_write(parent->mnt);
645 out_dput:
646 	dput(dentry);
647 out_unlock:
648 	mutex_unlock(&dir->i_mutex);
649 	return error;
650 }
651 
652 /*
653  * When we're defragging a range, we don't want to kick it off again
654  * if it is really just waiting for delalloc to send it down.
655  * If we find a nice big extent or delalloc range for the bytes in the
656  * file you want to defrag, we return 0 to let you know to skip this
657  * part of the file
658  */
659 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
660 {
661 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
662 	struct extent_map *em = NULL;
663 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
664 	u64 end;
665 
666 	read_lock(&em_tree->lock);
667 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
668 	read_unlock(&em_tree->lock);
669 
670 	if (em) {
671 		end = extent_map_end(em);
672 		free_extent_map(em);
673 		if (end - offset > thresh)
674 			return 0;
675 	}
676 	/* if we already have a nice delalloc here, just stop */
677 	thresh /= 2;
678 	end = count_range_bits(io_tree, &offset, offset + thresh,
679 			       thresh, EXTENT_DELALLOC, 1);
680 	if (end >= thresh)
681 		return 0;
682 	return 1;
683 }
684 
685 /*
686  * helper function to walk through a file and find extents
687  * newer than a specific transid, and smaller than thresh.
688  *
689  * This is used by the defragging code to find new and small
690  * extents
691  */
692 static int find_new_extents(struct btrfs_root *root,
693 			    struct inode *inode, u64 newer_than,
694 			    u64 *off, int thresh)
695 {
696 	struct btrfs_path *path;
697 	struct btrfs_key min_key;
698 	struct btrfs_key max_key;
699 	struct extent_buffer *leaf;
700 	struct btrfs_file_extent_item *extent;
701 	int type;
702 	int ret;
703 	u64 ino = btrfs_ino(inode);
704 
705 	path = btrfs_alloc_path();
706 	if (!path)
707 		return -ENOMEM;
708 
709 	min_key.objectid = ino;
710 	min_key.type = BTRFS_EXTENT_DATA_KEY;
711 	min_key.offset = *off;
712 
713 	max_key.objectid = ino;
714 	max_key.type = (u8)-1;
715 	max_key.offset = (u64)-1;
716 
717 	path->keep_locks = 1;
718 
719 	while(1) {
720 		ret = btrfs_search_forward(root, &min_key, &max_key,
721 					   path, 0, newer_than);
722 		if (ret != 0)
723 			goto none;
724 		if (min_key.objectid != ino)
725 			goto none;
726 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
727 			goto none;
728 
729 		leaf = path->nodes[0];
730 		extent = btrfs_item_ptr(leaf, path->slots[0],
731 					struct btrfs_file_extent_item);
732 
733 		type = btrfs_file_extent_type(leaf, extent);
734 		if (type == BTRFS_FILE_EXTENT_REG &&
735 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
736 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
737 			*off = min_key.offset;
738 			btrfs_free_path(path);
739 			return 0;
740 		}
741 
742 		if (min_key.offset == (u64)-1)
743 			goto none;
744 
745 		min_key.offset++;
746 		btrfs_release_path(path);
747 	}
748 none:
749 	btrfs_free_path(path);
750 	return -ENOENT;
751 }
752 
753 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
754 			       int thresh, u64 *last_len, u64 *skip,
755 			       u64 *defrag_end)
756 {
757 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
758 	struct extent_map *em = NULL;
759 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
760 	int ret = 1;
761 
762 	/*
763 	 * make sure that once we start defragging and extent, we keep on
764 	 * defragging it
765 	 */
766 	if (start < *defrag_end)
767 		return 1;
768 
769 	*skip = 0;
770 
771 	/*
772 	 * hopefully we have this extent in the tree already, try without
773 	 * the full extent lock
774 	 */
775 	read_lock(&em_tree->lock);
776 	em = lookup_extent_mapping(em_tree, start, len);
777 	read_unlock(&em_tree->lock);
778 
779 	if (!em) {
780 		/* get the big lock and read metadata off disk */
781 		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
782 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
783 		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
784 
785 		if (IS_ERR(em))
786 			return 0;
787 	}
788 
789 	/* this will cover holes, and inline extents */
790 	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
791 		ret = 0;
792 
793 	/*
794 	 * we hit a real extent, if it is big don't bother defragging it again
795 	 */
796 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
797 		ret = 0;
798 
799 	/*
800 	 * last_len ends up being a counter of how many bytes we've defragged.
801 	 * every time we choose not to defrag an extent, we reset *last_len
802 	 * so that the next tiny extent will force a defrag.
803 	 *
804 	 * The end result of this is that tiny extents before a single big
805 	 * extent will force at least part of that big extent to be defragged.
806 	 */
807 	if (ret) {
808 		*last_len += len;
809 		*defrag_end = extent_map_end(em);
810 	} else {
811 		*last_len = 0;
812 		*skip = extent_map_end(em);
813 		*defrag_end = 0;
814 	}
815 
816 	free_extent_map(em);
817 	return ret;
818 }
819 
820 /*
821  * it doesn't do much good to defrag one or two pages
822  * at a time.  This pulls in a nice chunk of pages
823  * to COW and defrag.
824  *
825  * It also makes sure the delalloc code has enough
826  * dirty data to avoid making new small extents as part
827  * of the defrag
828  *
829  * It's a good idea to start RA on this range
830  * before calling this.
831  */
832 static int cluster_pages_for_defrag(struct inode *inode,
833 				    struct page **pages,
834 				    unsigned long start_index,
835 				    int num_pages)
836 {
837 	unsigned long file_end;
838 	u64 isize = i_size_read(inode);
839 	u64 page_start;
840 	u64 page_end;
841 	int ret;
842 	int i;
843 	int i_done;
844 	struct btrfs_ordered_extent *ordered;
845 	struct extent_state *cached_state = NULL;
846 
847 	if (isize == 0)
848 		return 0;
849 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
850 
851 	ret = btrfs_delalloc_reserve_space(inode,
852 					   num_pages << PAGE_CACHE_SHIFT);
853 	if (ret)
854 		return ret;
855 again:
856 	ret = 0;
857 	i_done = 0;
858 
859 	/* step one, lock all the pages */
860 	for (i = 0; i < num_pages; i++) {
861 		struct page *page;
862 		page = find_or_create_page(inode->i_mapping,
863 					    start_index + i, GFP_NOFS);
864 		if (!page)
865 			break;
866 
867 		if (!PageUptodate(page)) {
868 			btrfs_readpage(NULL, page);
869 			lock_page(page);
870 			if (!PageUptodate(page)) {
871 				unlock_page(page);
872 				page_cache_release(page);
873 				ret = -EIO;
874 				break;
875 			}
876 		}
877 		isize = i_size_read(inode);
878 		file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
879 		if (!isize || page->index > file_end ||
880 		    page->mapping != inode->i_mapping) {
881 			/* whoops, we blew past eof, skip this page */
882 			unlock_page(page);
883 			page_cache_release(page);
884 			break;
885 		}
886 		pages[i] = page;
887 		i_done++;
888 	}
889 	if (!i_done || ret)
890 		goto out;
891 
892 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
893 		goto out;
894 
895 	/*
896 	 * so now we have a nice long stream of locked
897 	 * and up to date pages, lets wait on them
898 	 */
899 	for (i = 0; i < i_done; i++)
900 		wait_on_page_writeback(pages[i]);
901 
902 	page_start = page_offset(pages[0]);
903 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
904 
905 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
906 			 page_start, page_end - 1, 0, &cached_state,
907 			 GFP_NOFS);
908 	ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
909 	if (ordered &&
910 	    ordered->file_offset + ordered->len > page_start &&
911 	    ordered->file_offset < page_end) {
912 		btrfs_put_ordered_extent(ordered);
913 		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
914 				     page_start, page_end - 1,
915 				     &cached_state, GFP_NOFS);
916 		for (i = 0; i < i_done; i++) {
917 			unlock_page(pages[i]);
918 			page_cache_release(pages[i]);
919 		}
920 		btrfs_wait_ordered_range(inode, page_start,
921 					 page_end - page_start);
922 		goto again;
923 	}
924 	if (ordered)
925 		btrfs_put_ordered_extent(ordered);
926 
927 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
928 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
929 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
930 			  GFP_NOFS);
931 
932 	if (i_done != num_pages) {
933 		spin_lock(&BTRFS_I(inode)->lock);
934 		BTRFS_I(inode)->outstanding_extents++;
935 		spin_unlock(&BTRFS_I(inode)->lock);
936 		btrfs_delalloc_release_space(inode,
937 				     (num_pages - i_done) << PAGE_CACHE_SHIFT);
938 	}
939 
940 
941 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
942 				  &cached_state);
943 
944 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
945 			     page_start, page_end - 1, &cached_state,
946 			     GFP_NOFS);
947 
948 	for (i = 0; i < i_done; i++) {
949 		clear_page_dirty_for_io(pages[i]);
950 		ClearPageChecked(pages[i]);
951 		set_page_extent_mapped(pages[i]);
952 		set_page_dirty(pages[i]);
953 		unlock_page(pages[i]);
954 		page_cache_release(pages[i]);
955 	}
956 	return i_done;
957 out:
958 	for (i = 0; i < i_done; i++) {
959 		unlock_page(pages[i]);
960 		page_cache_release(pages[i]);
961 	}
962 	btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
963 	return ret;
964 
965 }
966 
967 int btrfs_defrag_file(struct inode *inode, struct file *file,
968 		      struct btrfs_ioctl_defrag_range_args *range,
969 		      u64 newer_than, unsigned long max_to_defrag)
970 {
971 	struct btrfs_root *root = BTRFS_I(inode)->root;
972 	struct btrfs_super_block *disk_super;
973 	struct file_ra_state *ra = NULL;
974 	unsigned long last_index;
975 	u64 features;
976 	u64 last_len = 0;
977 	u64 skip = 0;
978 	u64 defrag_end = 0;
979 	u64 newer_off = range->start;
980 	int newer_left = 0;
981 	unsigned long i;
982 	int ret;
983 	int defrag_count = 0;
984 	int compress_type = BTRFS_COMPRESS_ZLIB;
985 	int extent_thresh = range->extent_thresh;
986 	int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
987 	u64 new_align = ~((u64)128 * 1024 - 1);
988 	struct page **pages = NULL;
989 
990 	if (extent_thresh == 0)
991 		extent_thresh = 256 * 1024;
992 
993 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
994 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
995 			return -EINVAL;
996 		if (range->compress_type)
997 			compress_type = range->compress_type;
998 	}
999 
1000 	if (inode->i_size == 0)
1001 		return 0;
1002 
1003 	/*
1004 	 * if we were not given a file, allocate a readahead
1005 	 * context
1006 	 */
1007 	if (!file) {
1008 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1009 		if (!ra)
1010 			return -ENOMEM;
1011 		file_ra_state_init(ra, inode->i_mapping);
1012 	} else {
1013 		ra = &file->f_ra;
1014 	}
1015 
1016 	pages = kmalloc(sizeof(struct page *) * newer_cluster,
1017 			GFP_NOFS);
1018 	if (!pages) {
1019 		ret = -ENOMEM;
1020 		goto out_ra;
1021 	}
1022 
1023 	/* find the last page to defrag */
1024 	if (range->start + range->len > range->start) {
1025 		last_index = min_t(u64, inode->i_size - 1,
1026 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1027 	} else {
1028 		last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
1029 	}
1030 
1031 	if (newer_than) {
1032 		ret = find_new_extents(root, inode, newer_than,
1033 				       &newer_off, 64 * 1024);
1034 		if (!ret) {
1035 			range->start = newer_off;
1036 			/*
1037 			 * we always align our defrag to help keep
1038 			 * the extents in the file evenly spaced
1039 			 */
1040 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1041 			newer_left = newer_cluster;
1042 		} else
1043 			goto out_ra;
1044 	} else {
1045 		i = range->start >> PAGE_CACHE_SHIFT;
1046 	}
1047 	if (!max_to_defrag)
1048 		max_to_defrag = last_index - 1;
1049 
1050 	/*
1051 	 * make writeback starts from i, so the defrag range can be
1052 	 * written sequentially.
1053 	 */
1054 	if (i < inode->i_mapping->writeback_index)
1055 		inode->i_mapping->writeback_index = i;
1056 
1057 	while (i <= last_index && defrag_count < max_to_defrag &&
1058 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1059 		PAGE_CACHE_SHIFT)) {
1060 		/*
1061 		 * make sure we stop running if someone unmounts
1062 		 * the FS
1063 		 */
1064 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1065 			break;
1066 
1067 		if (!newer_than &&
1068 		    !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1069 					PAGE_CACHE_SIZE,
1070 					extent_thresh,
1071 					&last_len, &skip,
1072 					&defrag_end)) {
1073 			unsigned long next;
1074 			/*
1075 			 * the should_defrag function tells us how much to skip
1076 			 * bump our counter by the suggested amount
1077 			 */
1078 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1079 			i = max(i + 1, next);
1080 			continue;
1081 		}
1082 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1083 			BTRFS_I(inode)->force_compress = compress_type;
1084 
1085 		btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
1086 
1087 		ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1088 		if (ret < 0)
1089 			goto out_ra;
1090 
1091 		defrag_count += ret;
1092 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1093 		i += ret;
1094 
1095 		if (newer_than) {
1096 			if (newer_off == (u64)-1)
1097 				break;
1098 
1099 			newer_off = max(newer_off + 1,
1100 					(u64)i << PAGE_CACHE_SHIFT);
1101 
1102 			ret = find_new_extents(root, inode,
1103 					       newer_than, &newer_off,
1104 					       64 * 1024);
1105 			if (!ret) {
1106 				range->start = newer_off;
1107 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1108 				newer_left = newer_cluster;
1109 			} else {
1110 				break;
1111 			}
1112 		} else {
1113 			i++;
1114 		}
1115 	}
1116 
1117 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1118 		filemap_flush(inode->i_mapping);
1119 
1120 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1121 		/* the filemap_flush will queue IO into the worker threads, but
1122 		 * we have to make sure the IO is actually started and that
1123 		 * ordered extents get created before we return
1124 		 */
1125 		atomic_inc(&root->fs_info->async_submit_draining);
1126 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1127 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1128 			wait_event(root->fs_info->async_submit_wait,
1129 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1130 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1131 		}
1132 		atomic_dec(&root->fs_info->async_submit_draining);
1133 
1134 		mutex_lock(&inode->i_mutex);
1135 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1136 		mutex_unlock(&inode->i_mutex);
1137 	}
1138 
1139 	disk_super = &root->fs_info->super_copy;
1140 	features = btrfs_super_incompat_flags(disk_super);
1141 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1142 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1143 		btrfs_set_super_incompat_flags(disk_super, features);
1144 	}
1145 
1146 	if (!file)
1147 		kfree(ra);
1148 	return defrag_count;
1149 
1150 out_ra:
1151 	if (!file)
1152 		kfree(ra);
1153 	kfree(pages);
1154 	return ret;
1155 }
1156 
1157 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1158 					void __user *arg)
1159 {
1160 	u64 new_size;
1161 	u64 old_size;
1162 	u64 devid = 1;
1163 	struct btrfs_ioctl_vol_args *vol_args;
1164 	struct btrfs_trans_handle *trans;
1165 	struct btrfs_device *device = NULL;
1166 	char *sizestr;
1167 	char *devstr = NULL;
1168 	int ret = 0;
1169 	int mod = 0;
1170 
1171 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1172 		return -EROFS;
1173 
1174 	if (!capable(CAP_SYS_ADMIN))
1175 		return -EPERM;
1176 
1177 	vol_args = memdup_user(arg, sizeof(*vol_args));
1178 	if (IS_ERR(vol_args))
1179 		return PTR_ERR(vol_args);
1180 
1181 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1182 
1183 	mutex_lock(&root->fs_info->volume_mutex);
1184 	sizestr = vol_args->name;
1185 	devstr = strchr(sizestr, ':');
1186 	if (devstr) {
1187 		char *end;
1188 		sizestr = devstr + 1;
1189 		*devstr = '\0';
1190 		devstr = vol_args->name;
1191 		devid = simple_strtoull(devstr, &end, 10);
1192 		printk(KERN_INFO "resizing devid %llu\n",
1193 		       (unsigned long long)devid);
1194 	}
1195 	device = btrfs_find_device(root, devid, NULL, NULL);
1196 	if (!device) {
1197 		printk(KERN_INFO "resizer unable to find device %llu\n",
1198 		       (unsigned long long)devid);
1199 		ret = -EINVAL;
1200 		goto out_unlock;
1201 	}
1202 	if (!strcmp(sizestr, "max"))
1203 		new_size = device->bdev->bd_inode->i_size;
1204 	else {
1205 		if (sizestr[0] == '-') {
1206 			mod = -1;
1207 			sizestr++;
1208 		} else if (sizestr[0] == '+') {
1209 			mod = 1;
1210 			sizestr++;
1211 		}
1212 		new_size = memparse(sizestr, NULL);
1213 		if (new_size == 0) {
1214 			ret = -EINVAL;
1215 			goto out_unlock;
1216 		}
1217 	}
1218 
1219 	old_size = device->total_bytes;
1220 
1221 	if (mod < 0) {
1222 		if (new_size > old_size) {
1223 			ret = -EINVAL;
1224 			goto out_unlock;
1225 		}
1226 		new_size = old_size - new_size;
1227 	} else if (mod > 0) {
1228 		new_size = old_size + new_size;
1229 	}
1230 
1231 	if (new_size < 256 * 1024 * 1024) {
1232 		ret = -EINVAL;
1233 		goto out_unlock;
1234 	}
1235 	if (new_size > device->bdev->bd_inode->i_size) {
1236 		ret = -EFBIG;
1237 		goto out_unlock;
1238 	}
1239 
1240 	do_div(new_size, root->sectorsize);
1241 	new_size *= root->sectorsize;
1242 
1243 	printk(KERN_INFO "new size for %s is %llu\n",
1244 		device->name, (unsigned long long)new_size);
1245 
1246 	if (new_size > old_size) {
1247 		trans = btrfs_start_transaction(root, 0);
1248 		if (IS_ERR(trans)) {
1249 			ret = PTR_ERR(trans);
1250 			goto out_unlock;
1251 		}
1252 		ret = btrfs_grow_device(trans, device, new_size);
1253 		btrfs_commit_transaction(trans, root);
1254 	} else {
1255 		ret = btrfs_shrink_device(device, new_size);
1256 	}
1257 
1258 out_unlock:
1259 	mutex_unlock(&root->fs_info->volume_mutex);
1260 	kfree(vol_args);
1261 	return ret;
1262 }
1263 
1264 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1265 						    char *name,
1266 						    unsigned long fd,
1267 						    int subvol,
1268 						    u64 *transid,
1269 						    bool readonly)
1270 {
1271 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1272 	struct file *src_file;
1273 	int namelen;
1274 	int ret = 0;
1275 
1276 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1277 		return -EROFS;
1278 
1279 	namelen = strlen(name);
1280 	if (strchr(name, '/')) {
1281 		ret = -EINVAL;
1282 		goto out;
1283 	}
1284 
1285 	if (subvol) {
1286 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1287 				     NULL, transid, readonly);
1288 	} else {
1289 		struct inode *src_inode;
1290 		src_file = fget(fd);
1291 		if (!src_file) {
1292 			ret = -EINVAL;
1293 			goto out;
1294 		}
1295 
1296 		src_inode = src_file->f_path.dentry->d_inode;
1297 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1298 			printk(KERN_INFO "btrfs: Snapshot src from "
1299 			       "another FS\n");
1300 			ret = -EINVAL;
1301 			fput(src_file);
1302 			goto out;
1303 		}
1304 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1305 				     BTRFS_I(src_inode)->root,
1306 				     transid, readonly);
1307 		fput(src_file);
1308 	}
1309 out:
1310 	return ret;
1311 }
1312 
1313 static noinline int btrfs_ioctl_snap_create(struct file *file,
1314 					    void __user *arg, int subvol)
1315 {
1316 	struct btrfs_ioctl_vol_args *vol_args;
1317 	int ret;
1318 
1319 	vol_args = memdup_user(arg, sizeof(*vol_args));
1320 	if (IS_ERR(vol_args))
1321 		return PTR_ERR(vol_args);
1322 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1323 
1324 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1325 					      vol_args->fd, subvol,
1326 					      NULL, false);
1327 
1328 	kfree(vol_args);
1329 	return ret;
1330 }
1331 
1332 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1333 					       void __user *arg, int subvol)
1334 {
1335 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1336 	int ret;
1337 	u64 transid = 0;
1338 	u64 *ptr = NULL;
1339 	bool readonly = false;
1340 
1341 	vol_args = memdup_user(arg, sizeof(*vol_args));
1342 	if (IS_ERR(vol_args))
1343 		return PTR_ERR(vol_args);
1344 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1345 
1346 	if (vol_args->flags &
1347 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1348 		ret = -EOPNOTSUPP;
1349 		goto out;
1350 	}
1351 
1352 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1353 		ptr = &transid;
1354 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1355 		readonly = true;
1356 
1357 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1358 					      vol_args->fd, subvol,
1359 					      ptr, readonly);
1360 
1361 	if (ret == 0 && ptr &&
1362 	    copy_to_user(arg +
1363 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1364 				  transid), ptr, sizeof(*ptr)))
1365 		ret = -EFAULT;
1366 out:
1367 	kfree(vol_args);
1368 	return ret;
1369 }
1370 
1371 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1372 						void __user *arg)
1373 {
1374 	struct inode *inode = fdentry(file)->d_inode;
1375 	struct btrfs_root *root = BTRFS_I(inode)->root;
1376 	int ret = 0;
1377 	u64 flags = 0;
1378 
1379 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1380 		return -EINVAL;
1381 
1382 	down_read(&root->fs_info->subvol_sem);
1383 	if (btrfs_root_readonly(root))
1384 		flags |= BTRFS_SUBVOL_RDONLY;
1385 	up_read(&root->fs_info->subvol_sem);
1386 
1387 	if (copy_to_user(arg, &flags, sizeof(flags)))
1388 		ret = -EFAULT;
1389 
1390 	return ret;
1391 }
1392 
1393 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1394 					      void __user *arg)
1395 {
1396 	struct inode *inode = fdentry(file)->d_inode;
1397 	struct btrfs_root *root = BTRFS_I(inode)->root;
1398 	struct btrfs_trans_handle *trans;
1399 	u64 root_flags;
1400 	u64 flags;
1401 	int ret = 0;
1402 
1403 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1404 		return -EROFS;
1405 
1406 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1407 		return -EINVAL;
1408 
1409 	if (copy_from_user(&flags, arg, sizeof(flags)))
1410 		return -EFAULT;
1411 
1412 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1413 		return -EINVAL;
1414 
1415 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1416 		return -EOPNOTSUPP;
1417 
1418 	if (!inode_owner_or_capable(inode))
1419 		return -EACCES;
1420 
1421 	down_write(&root->fs_info->subvol_sem);
1422 
1423 	/* nothing to do */
1424 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1425 		goto out;
1426 
1427 	root_flags = btrfs_root_flags(&root->root_item);
1428 	if (flags & BTRFS_SUBVOL_RDONLY)
1429 		btrfs_set_root_flags(&root->root_item,
1430 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1431 	else
1432 		btrfs_set_root_flags(&root->root_item,
1433 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1434 
1435 	trans = btrfs_start_transaction(root, 1);
1436 	if (IS_ERR(trans)) {
1437 		ret = PTR_ERR(trans);
1438 		goto out_reset;
1439 	}
1440 
1441 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1442 				&root->root_key, &root->root_item);
1443 
1444 	btrfs_commit_transaction(trans, root);
1445 out_reset:
1446 	if (ret)
1447 		btrfs_set_root_flags(&root->root_item, root_flags);
1448 out:
1449 	up_write(&root->fs_info->subvol_sem);
1450 	return ret;
1451 }
1452 
1453 /*
1454  * helper to check if the subvolume references other subvolumes
1455  */
1456 static noinline int may_destroy_subvol(struct btrfs_root *root)
1457 {
1458 	struct btrfs_path *path;
1459 	struct btrfs_key key;
1460 	int ret;
1461 
1462 	path = btrfs_alloc_path();
1463 	if (!path)
1464 		return -ENOMEM;
1465 
1466 	key.objectid = root->root_key.objectid;
1467 	key.type = BTRFS_ROOT_REF_KEY;
1468 	key.offset = (u64)-1;
1469 
1470 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1471 				&key, path, 0, 0);
1472 	if (ret < 0)
1473 		goto out;
1474 	BUG_ON(ret == 0);
1475 
1476 	ret = 0;
1477 	if (path->slots[0] > 0) {
1478 		path->slots[0]--;
1479 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1480 		if (key.objectid == root->root_key.objectid &&
1481 		    key.type == BTRFS_ROOT_REF_KEY)
1482 			ret = -ENOTEMPTY;
1483 	}
1484 out:
1485 	btrfs_free_path(path);
1486 	return ret;
1487 }
1488 
1489 static noinline int key_in_sk(struct btrfs_key *key,
1490 			      struct btrfs_ioctl_search_key *sk)
1491 {
1492 	struct btrfs_key test;
1493 	int ret;
1494 
1495 	test.objectid = sk->min_objectid;
1496 	test.type = sk->min_type;
1497 	test.offset = sk->min_offset;
1498 
1499 	ret = btrfs_comp_cpu_keys(key, &test);
1500 	if (ret < 0)
1501 		return 0;
1502 
1503 	test.objectid = sk->max_objectid;
1504 	test.type = sk->max_type;
1505 	test.offset = sk->max_offset;
1506 
1507 	ret = btrfs_comp_cpu_keys(key, &test);
1508 	if (ret > 0)
1509 		return 0;
1510 	return 1;
1511 }
1512 
1513 static noinline int copy_to_sk(struct btrfs_root *root,
1514 			       struct btrfs_path *path,
1515 			       struct btrfs_key *key,
1516 			       struct btrfs_ioctl_search_key *sk,
1517 			       char *buf,
1518 			       unsigned long *sk_offset,
1519 			       int *num_found)
1520 {
1521 	u64 found_transid;
1522 	struct extent_buffer *leaf;
1523 	struct btrfs_ioctl_search_header sh;
1524 	unsigned long item_off;
1525 	unsigned long item_len;
1526 	int nritems;
1527 	int i;
1528 	int slot;
1529 	int ret = 0;
1530 
1531 	leaf = path->nodes[0];
1532 	slot = path->slots[0];
1533 	nritems = btrfs_header_nritems(leaf);
1534 
1535 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1536 		i = nritems;
1537 		goto advance_key;
1538 	}
1539 	found_transid = btrfs_header_generation(leaf);
1540 
1541 	for (i = slot; i < nritems; i++) {
1542 		item_off = btrfs_item_ptr_offset(leaf, i);
1543 		item_len = btrfs_item_size_nr(leaf, i);
1544 
1545 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1546 			item_len = 0;
1547 
1548 		if (sizeof(sh) + item_len + *sk_offset >
1549 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1550 			ret = 1;
1551 			goto overflow;
1552 		}
1553 
1554 		btrfs_item_key_to_cpu(leaf, key, i);
1555 		if (!key_in_sk(key, sk))
1556 			continue;
1557 
1558 		sh.objectid = key->objectid;
1559 		sh.offset = key->offset;
1560 		sh.type = key->type;
1561 		sh.len = item_len;
1562 		sh.transid = found_transid;
1563 
1564 		/* copy search result header */
1565 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1566 		*sk_offset += sizeof(sh);
1567 
1568 		if (item_len) {
1569 			char *p = buf + *sk_offset;
1570 			/* copy the item */
1571 			read_extent_buffer(leaf, p,
1572 					   item_off, item_len);
1573 			*sk_offset += item_len;
1574 		}
1575 		(*num_found)++;
1576 
1577 		if (*num_found >= sk->nr_items)
1578 			break;
1579 	}
1580 advance_key:
1581 	ret = 0;
1582 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1583 		key->offset++;
1584 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1585 		key->offset = 0;
1586 		key->type++;
1587 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1588 		key->offset = 0;
1589 		key->type = 0;
1590 		key->objectid++;
1591 	} else
1592 		ret = 1;
1593 overflow:
1594 	return ret;
1595 }
1596 
1597 static noinline int search_ioctl(struct inode *inode,
1598 				 struct btrfs_ioctl_search_args *args)
1599 {
1600 	struct btrfs_root *root;
1601 	struct btrfs_key key;
1602 	struct btrfs_key max_key;
1603 	struct btrfs_path *path;
1604 	struct btrfs_ioctl_search_key *sk = &args->key;
1605 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1606 	int ret;
1607 	int num_found = 0;
1608 	unsigned long sk_offset = 0;
1609 
1610 	path = btrfs_alloc_path();
1611 	if (!path)
1612 		return -ENOMEM;
1613 
1614 	if (sk->tree_id == 0) {
1615 		/* search the root of the inode that was passed */
1616 		root = BTRFS_I(inode)->root;
1617 	} else {
1618 		key.objectid = sk->tree_id;
1619 		key.type = BTRFS_ROOT_ITEM_KEY;
1620 		key.offset = (u64)-1;
1621 		root = btrfs_read_fs_root_no_name(info, &key);
1622 		if (IS_ERR(root)) {
1623 			printk(KERN_ERR "could not find root %llu\n",
1624 			       sk->tree_id);
1625 			btrfs_free_path(path);
1626 			return -ENOENT;
1627 		}
1628 	}
1629 
1630 	key.objectid = sk->min_objectid;
1631 	key.type = sk->min_type;
1632 	key.offset = sk->min_offset;
1633 
1634 	max_key.objectid = sk->max_objectid;
1635 	max_key.type = sk->max_type;
1636 	max_key.offset = sk->max_offset;
1637 
1638 	path->keep_locks = 1;
1639 
1640 	while(1) {
1641 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1642 					   sk->min_transid);
1643 		if (ret != 0) {
1644 			if (ret > 0)
1645 				ret = 0;
1646 			goto err;
1647 		}
1648 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1649 				 &sk_offset, &num_found);
1650 		btrfs_release_path(path);
1651 		if (ret || num_found >= sk->nr_items)
1652 			break;
1653 
1654 	}
1655 	ret = 0;
1656 err:
1657 	sk->nr_items = num_found;
1658 	btrfs_free_path(path);
1659 	return ret;
1660 }
1661 
1662 static noinline int btrfs_ioctl_tree_search(struct file *file,
1663 					   void __user *argp)
1664 {
1665 	 struct btrfs_ioctl_search_args *args;
1666 	 struct inode *inode;
1667 	 int ret;
1668 
1669 	if (!capable(CAP_SYS_ADMIN))
1670 		return -EPERM;
1671 
1672 	args = memdup_user(argp, sizeof(*args));
1673 	if (IS_ERR(args))
1674 		return PTR_ERR(args);
1675 
1676 	inode = fdentry(file)->d_inode;
1677 	ret = search_ioctl(inode, args);
1678 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1679 		ret = -EFAULT;
1680 	kfree(args);
1681 	return ret;
1682 }
1683 
1684 /*
1685  * Search INODE_REFs to identify path name of 'dirid' directory
1686  * in a 'tree_id' tree. and sets path name to 'name'.
1687  */
1688 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1689 				u64 tree_id, u64 dirid, char *name)
1690 {
1691 	struct btrfs_root *root;
1692 	struct btrfs_key key;
1693 	char *ptr;
1694 	int ret = -1;
1695 	int slot;
1696 	int len;
1697 	int total_len = 0;
1698 	struct btrfs_inode_ref *iref;
1699 	struct extent_buffer *l;
1700 	struct btrfs_path *path;
1701 
1702 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1703 		name[0]='\0';
1704 		return 0;
1705 	}
1706 
1707 	path = btrfs_alloc_path();
1708 	if (!path)
1709 		return -ENOMEM;
1710 
1711 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1712 
1713 	key.objectid = tree_id;
1714 	key.type = BTRFS_ROOT_ITEM_KEY;
1715 	key.offset = (u64)-1;
1716 	root = btrfs_read_fs_root_no_name(info, &key);
1717 	if (IS_ERR(root)) {
1718 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1719 		ret = -ENOENT;
1720 		goto out;
1721 	}
1722 
1723 	key.objectid = dirid;
1724 	key.type = BTRFS_INODE_REF_KEY;
1725 	key.offset = (u64)-1;
1726 
1727 	while(1) {
1728 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1729 		if (ret < 0)
1730 			goto out;
1731 
1732 		l = path->nodes[0];
1733 		slot = path->slots[0];
1734 		if (ret > 0 && slot > 0)
1735 			slot--;
1736 		btrfs_item_key_to_cpu(l, &key, slot);
1737 
1738 		if (ret > 0 && (key.objectid != dirid ||
1739 				key.type != BTRFS_INODE_REF_KEY)) {
1740 			ret = -ENOENT;
1741 			goto out;
1742 		}
1743 
1744 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1745 		len = btrfs_inode_ref_name_len(l, iref);
1746 		ptr -= len + 1;
1747 		total_len += len + 1;
1748 		if (ptr < name)
1749 			goto out;
1750 
1751 		*(ptr + len) = '/';
1752 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1753 
1754 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1755 			break;
1756 
1757 		btrfs_release_path(path);
1758 		key.objectid = key.offset;
1759 		key.offset = (u64)-1;
1760 		dirid = key.objectid;
1761 	}
1762 	if (ptr < name)
1763 		goto out;
1764 	memmove(name, ptr, total_len);
1765 	name[total_len]='\0';
1766 	ret = 0;
1767 out:
1768 	btrfs_free_path(path);
1769 	return ret;
1770 }
1771 
1772 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1773 					   void __user *argp)
1774 {
1775 	 struct btrfs_ioctl_ino_lookup_args *args;
1776 	 struct inode *inode;
1777 	 int ret;
1778 
1779 	if (!capable(CAP_SYS_ADMIN))
1780 		return -EPERM;
1781 
1782 	args = memdup_user(argp, sizeof(*args));
1783 	if (IS_ERR(args))
1784 		return PTR_ERR(args);
1785 
1786 	inode = fdentry(file)->d_inode;
1787 
1788 	if (args->treeid == 0)
1789 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1790 
1791 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1792 					args->treeid, args->objectid,
1793 					args->name);
1794 
1795 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1796 		ret = -EFAULT;
1797 
1798 	kfree(args);
1799 	return ret;
1800 }
1801 
1802 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1803 					     void __user *arg)
1804 {
1805 	struct dentry *parent = fdentry(file);
1806 	struct dentry *dentry;
1807 	struct inode *dir = parent->d_inode;
1808 	struct inode *inode;
1809 	struct btrfs_root *root = BTRFS_I(dir)->root;
1810 	struct btrfs_root *dest = NULL;
1811 	struct btrfs_ioctl_vol_args *vol_args;
1812 	struct btrfs_trans_handle *trans;
1813 	int namelen;
1814 	int ret;
1815 	int err = 0;
1816 
1817 	vol_args = memdup_user(arg, sizeof(*vol_args));
1818 	if (IS_ERR(vol_args))
1819 		return PTR_ERR(vol_args);
1820 
1821 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1822 	namelen = strlen(vol_args->name);
1823 	if (strchr(vol_args->name, '/') ||
1824 	    strncmp(vol_args->name, "..", namelen) == 0) {
1825 		err = -EINVAL;
1826 		goto out;
1827 	}
1828 
1829 	err = mnt_want_write(file->f_path.mnt);
1830 	if (err)
1831 		goto out;
1832 
1833 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1834 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1835 	if (IS_ERR(dentry)) {
1836 		err = PTR_ERR(dentry);
1837 		goto out_unlock_dir;
1838 	}
1839 
1840 	if (!dentry->d_inode) {
1841 		err = -ENOENT;
1842 		goto out_dput;
1843 	}
1844 
1845 	inode = dentry->d_inode;
1846 	dest = BTRFS_I(inode)->root;
1847 	if (!capable(CAP_SYS_ADMIN)){
1848 		/*
1849 		 * Regular user.  Only allow this with a special mount
1850 		 * option, when the user has write+exec access to the
1851 		 * subvol root, and when rmdir(2) would have been
1852 		 * allowed.
1853 		 *
1854 		 * Note that this is _not_ check that the subvol is
1855 		 * empty or doesn't contain data that we wouldn't
1856 		 * otherwise be able to delete.
1857 		 *
1858 		 * Users who want to delete empty subvols should try
1859 		 * rmdir(2).
1860 		 */
1861 		err = -EPERM;
1862 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1863 			goto out_dput;
1864 
1865 		/*
1866 		 * Do not allow deletion if the parent dir is the same
1867 		 * as the dir to be deleted.  That means the ioctl
1868 		 * must be called on the dentry referencing the root
1869 		 * of the subvol, not a random directory contained
1870 		 * within it.
1871 		 */
1872 		err = -EINVAL;
1873 		if (root == dest)
1874 			goto out_dput;
1875 
1876 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1877 		if (err)
1878 			goto out_dput;
1879 
1880 		/* check if subvolume may be deleted by a non-root user */
1881 		err = btrfs_may_delete(dir, dentry, 1);
1882 		if (err)
1883 			goto out_dput;
1884 	}
1885 
1886 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1887 		err = -EINVAL;
1888 		goto out_dput;
1889 	}
1890 
1891 	mutex_lock(&inode->i_mutex);
1892 	err = d_invalidate(dentry);
1893 	if (err)
1894 		goto out_unlock;
1895 
1896 	down_write(&root->fs_info->subvol_sem);
1897 
1898 	err = may_destroy_subvol(dest);
1899 	if (err)
1900 		goto out_up_write;
1901 
1902 	trans = btrfs_start_transaction(root, 0);
1903 	if (IS_ERR(trans)) {
1904 		err = PTR_ERR(trans);
1905 		goto out_up_write;
1906 	}
1907 	trans->block_rsv = &root->fs_info->global_block_rsv;
1908 
1909 	ret = btrfs_unlink_subvol(trans, root, dir,
1910 				dest->root_key.objectid,
1911 				dentry->d_name.name,
1912 				dentry->d_name.len);
1913 	BUG_ON(ret);
1914 
1915 	btrfs_record_root_in_trans(trans, dest);
1916 
1917 	memset(&dest->root_item.drop_progress, 0,
1918 		sizeof(dest->root_item.drop_progress));
1919 	dest->root_item.drop_level = 0;
1920 	btrfs_set_root_refs(&dest->root_item, 0);
1921 
1922 	if (!xchg(&dest->orphan_item_inserted, 1)) {
1923 		ret = btrfs_insert_orphan_item(trans,
1924 					root->fs_info->tree_root,
1925 					dest->root_key.objectid);
1926 		BUG_ON(ret);
1927 	}
1928 
1929 	ret = btrfs_end_transaction(trans, root);
1930 	BUG_ON(ret);
1931 	inode->i_flags |= S_DEAD;
1932 out_up_write:
1933 	up_write(&root->fs_info->subvol_sem);
1934 out_unlock:
1935 	mutex_unlock(&inode->i_mutex);
1936 	if (!err) {
1937 		shrink_dcache_sb(root->fs_info->sb);
1938 		btrfs_invalidate_inodes(dest);
1939 		d_delete(dentry);
1940 	}
1941 out_dput:
1942 	dput(dentry);
1943 out_unlock_dir:
1944 	mutex_unlock(&dir->i_mutex);
1945 	mnt_drop_write(file->f_path.mnt);
1946 out:
1947 	kfree(vol_args);
1948 	return err;
1949 }
1950 
1951 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1952 {
1953 	struct inode *inode = fdentry(file)->d_inode;
1954 	struct btrfs_root *root = BTRFS_I(inode)->root;
1955 	struct btrfs_ioctl_defrag_range_args *range;
1956 	int ret;
1957 
1958 	if (btrfs_root_readonly(root))
1959 		return -EROFS;
1960 
1961 	ret = mnt_want_write(file->f_path.mnt);
1962 	if (ret)
1963 		return ret;
1964 
1965 	switch (inode->i_mode & S_IFMT) {
1966 	case S_IFDIR:
1967 		if (!capable(CAP_SYS_ADMIN)) {
1968 			ret = -EPERM;
1969 			goto out;
1970 		}
1971 		ret = btrfs_defrag_root(root, 0);
1972 		if (ret)
1973 			goto out;
1974 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1975 		break;
1976 	case S_IFREG:
1977 		if (!(file->f_mode & FMODE_WRITE)) {
1978 			ret = -EINVAL;
1979 			goto out;
1980 		}
1981 
1982 		range = kzalloc(sizeof(*range), GFP_KERNEL);
1983 		if (!range) {
1984 			ret = -ENOMEM;
1985 			goto out;
1986 		}
1987 
1988 		if (argp) {
1989 			if (copy_from_user(range, argp,
1990 					   sizeof(*range))) {
1991 				ret = -EFAULT;
1992 				kfree(range);
1993 				goto out;
1994 			}
1995 			/* compression requires us to start the IO */
1996 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1997 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1998 				range->extent_thresh = (u32)-1;
1999 			}
2000 		} else {
2001 			/* the rest are all set to zero by kzalloc */
2002 			range->len = (u64)-1;
2003 		}
2004 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2005 					range, 0, 0);
2006 		if (ret > 0)
2007 			ret = 0;
2008 		kfree(range);
2009 		break;
2010 	default:
2011 		ret = -EINVAL;
2012 	}
2013 out:
2014 	mnt_drop_write(file->f_path.mnt);
2015 	return ret;
2016 }
2017 
2018 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2019 {
2020 	struct btrfs_ioctl_vol_args *vol_args;
2021 	int ret;
2022 
2023 	if (!capable(CAP_SYS_ADMIN))
2024 		return -EPERM;
2025 
2026 	vol_args = memdup_user(arg, sizeof(*vol_args));
2027 	if (IS_ERR(vol_args))
2028 		return PTR_ERR(vol_args);
2029 
2030 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2031 	ret = btrfs_init_new_device(root, vol_args->name);
2032 
2033 	kfree(vol_args);
2034 	return ret;
2035 }
2036 
2037 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2038 {
2039 	struct btrfs_ioctl_vol_args *vol_args;
2040 	int ret;
2041 
2042 	if (!capable(CAP_SYS_ADMIN))
2043 		return -EPERM;
2044 
2045 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2046 		return -EROFS;
2047 
2048 	vol_args = memdup_user(arg, sizeof(*vol_args));
2049 	if (IS_ERR(vol_args))
2050 		return PTR_ERR(vol_args);
2051 
2052 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2053 	ret = btrfs_rm_device(root, vol_args->name);
2054 
2055 	kfree(vol_args);
2056 	return ret;
2057 }
2058 
2059 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2060 {
2061 	struct btrfs_ioctl_fs_info_args *fi_args;
2062 	struct btrfs_device *device;
2063 	struct btrfs_device *next;
2064 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2065 	int ret = 0;
2066 
2067 	if (!capable(CAP_SYS_ADMIN))
2068 		return -EPERM;
2069 
2070 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2071 	if (!fi_args)
2072 		return -ENOMEM;
2073 
2074 	fi_args->num_devices = fs_devices->num_devices;
2075 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2076 
2077 	mutex_lock(&fs_devices->device_list_mutex);
2078 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2079 		if (device->devid > fi_args->max_id)
2080 			fi_args->max_id = device->devid;
2081 	}
2082 	mutex_unlock(&fs_devices->device_list_mutex);
2083 
2084 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2085 		ret = -EFAULT;
2086 
2087 	kfree(fi_args);
2088 	return ret;
2089 }
2090 
2091 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2092 {
2093 	struct btrfs_ioctl_dev_info_args *di_args;
2094 	struct btrfs_device *dev;
2095 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2096 	int ret = 0;
2097 	char *s_uuid = NULL;
2098 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2099 
2100 	if (!capable(CAP_SYS_ADMIN))
2101 		return -EPERM;
2102 
2103 	di_args = memdup_user(arg, sizeof(*di_args));
2104 	if (IS_ERR(di_args))
2105 		return PTR_ERR(di_args);
2106 
2107 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2108 		s_uuid = di_args->uuid;
2109 
2110 	mutex_lock(&fs_devices->device_list_mutex);
2111 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2112 	mutex_unlock(&fs_devices->device_list_mutex);
2113 
2114 	if (!dev) {
2115 		ret = -ENODEV;
2116 		goto out;
2117 	}
2118 
2119 	di_args->devid = dev->devid;
2120 	di_args->bytes_used = dev->bytes_used;
2121 	di_args->total_bytes = dev->total_bytes;
2122 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2123 	strncpy(di_args->path, dev->name, sizeof(di_args->path));
2124 
2125 out:
2126 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2127 		ret = -EFAULT;
2128 
2129 	kfree(di_args);
2130 	return ret;
2131 }
2132 
2133 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2134 				       u64 off, u64 olen, u64 destoff)
2135 {
2136 	struct inode *inode = fdentry(file)->d_inode;
2137 	struct btrfs_root *root = BTRFS_I(inode)->root;
2138 	struct file *src_file;
2139 	struct inode *src;
2140 	struct btrfs_trans_handle *trans;
2141 	struct btrfs_path *path;
2142 	struct extent_buffer *leaf;
2143 	char *buf;
2144 	struct btrfs_key key;
2145 	u32 nritems;
2146 	int slot;
2147 	int ret;
2148 	u64 len = olen;
2149 	u64 bs = root->fs_info->sb->s_blocksize;
2150 	u64 hint_byte;
2151 
2152 	/*
2153 	 * TODO:
2154 	 * - split compressed inline extents.  annoying: we need to
2155 	 *   decompress into destination's address_space (the file offset
2156 	 *   may change, so source mapping won't do), then recompress (or
2157 	 *   otherwise reinsert) a subrange.
2158 	 * - allow ranges within the same file to be cloned (provided
2159 	 *   they don't overlap)?
2160 	 */
2161 
2162 	/* the destination must be opened for writing */
2163 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2164 		return -EINVAL;
2165 
2166 	if (btrfs_root_readonly(root))
2167 		return -EROFS;
2168 
2169 	ret = mnt_want_write(file->f_path.mnt);
2170 	if (ret)
2171 		return ret;
2172 
2173 	src_file = fget(srcfd);
2174 	if (!src_file) {
2175 		ret = -EBADF;
2176 		goto out_drop_write;
2177 	}
2178 
2179 	src = src_file->f_dentry->d_inode;
2180 
2181 	ret = -EINVAL;
2182 	if (src == inode)
2183 		goto out_fput;
2184 
2185 	/* the src must be open for reading */
2186 	if (!(src_file->f_mode & FMODE_READ))
2187 		goto out_fput;
2188 
2189 	/* don't make the dst file partly checksummed */
2190 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2191 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2192 		goto out_fput;
2193 
2194 	ret = -EISDIR;
2195 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2196 		goto out_fput;
2197 
2198 	ret = -EXDEV;
2199 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2200 		goto out_fput;
2201 
2202 	ret = -ENOMEM;
2203 	buf = vmalloc(btrfs_level_size(root, 0));
2204 	if (!buf)
2205 		goto out_fput;
2206 
2207 	path = btrfs_alloc_path();
2208 	if (!path) {
2209 		vfree(buf);
2210 		goto out_fput;
2211 	}
2212 	path->reada = 2;
2213 
2214 	if (inode < src) {
2215 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2216 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2217 	} else {
2218 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2219 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2220 	}
2221 
2222 	/* determine range to clone */
2223 	ret = -EINVAL;
2224 	if (off + len > src->i_size || off + len < off)
2225 		goto out_unlock;
2226 	if (len == 0)
2227 		olen = len = src->i_size - off;
2228 	/* if we extend to eof, continue to block boundary */
2229 	if (off + len == src->i_size)
2230 		len = ALIGN(src->i_size, bs) - off;
2231 
2232 	/* verify the end result is block aligned */
2233 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2234 	    !IS_ALIGNED(destoff, bs))
2235 		goto out_unlock;
2236 
2237 	if (destoff > inode->i_size) {
2238 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2239 		if (ret)
2240 			goto out_unlock;
2241 	}
2242 
2243 	/* truncate page cache pages from target inode range */
2244 	truncate_inode_pages_range(&inode->i_data, destoff,
2245 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2246 
2247 	/* do any pending delalloc/csum calc on src, one way or
2248 	   another, and lock file content */
2249 	while (1) {
2250 		struct btrfs_ordered_extent *ordered;
2251 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2252 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2253 		if (!ordered &&
2254 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2255 				   EXTENT_DELALLOC, 0, NULL))
2256 			break;
2257 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2258 		if (ordered)
2259 			btrfs_put_ordered_extent(ordered);
2260 		btrfs_wait_ordered_range(src, off, len);
2261 	}
2262 
2263 	/* clone data */
2264 	key.objectid = btrfs_ino(src);
2265 	key.type = BTRFS_EXTENT_DATA_KEY;
2266 	key.offset = 0;
2267 
2268 	while (1) {
2269 		/*
2270 		 * note the key will change type as we walk through the
2271 		 * tree.
2272 		 */
2273 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2274 		if (ret < 0)
2275 			goto out;
2276 
2277 		nritems = btrfs_header_nritems(path->nodes[0]);
2278 		if (path->slots[0] >= nritems) {
2279 			ret = btrfs_next_leaf(root, path);
2280 			if (ret < 0)
2281 				goto out;
2282 			if (ret > 0)
2283 				break;
2284 			nritems = btrfs_header_nritems(path->nodes[0]);
2285 		}
2286 		leaf = path->nodes[0];
2287 		slot = path->slots[0];
2288 
2289 		btrfs_item_key_to_cpu(leaf, &key, slot);
2290 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2291 		    key.objectid != btrfs_ino(src))
2292 			break;
2293 
2294 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2295 			struct btrfs_file_extent_item *extent;
2296 			int type;
2297 			u32 size;
2298 			struct btrfs_key new_key;
2299 			u64 disko = 0, diskl = 0;
2300 			u64 datao = 0, datal = 0;
2301 			u8 comp;
2302 			u64 endoff;
2303 
2304 			size = btrfs_item_size_nr(leaf, slot);
2305 			read_extent_buffer(leaf, buf,
2306 					   btrfs_item_ptr_offset(leaf, slot),
2307 					   size);
2308 
2309 			extent = btrfs_item_ptr(leaf, slot,
2310 						struct btrfs_file_extent_item);
2311 			comp = btrfs_file_extent_compression(leaf, extent);
2312 			type = btrfs_file_extent_type(leaf, extent);
2313 			if (type == BTRFS_FILE_EXTENT_REG ||
2314 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2315 				disko = btrfs_file_extent_disk_bytenr(leaf,
2316 								      extent);
2317 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2318 								 extent);
2319 				datao = btrfs_file_extent_offset(leaf, extent);
2320 				datal = btrfs_file_extent_num_bytes(leaf,
2321 								    extent);
2322 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2323 				/* take upper bound, may be compressed */
2324 				datal = btrfs_file_extent_ram_bytes(leaf,
2325 								    extent);
2326 			}
2327 			btrfs_release_path(path);
2328 
2329 			if (key.offset + datal <= off ||
2330 			    key.offset >= off+len)
2331 				goto next;
2332 
2333 			memcpy(&new_key, &key, sizeof(new_key));
2334 			new_key.objectid = btrfs_ino(inode);
2335 			if (off <= key.offset)
2336 				new_key.offset = key.offset + destoff - off;
2337 			else
2338 				new_key.offset = destoff;
2339 
2340 			/*
2341 			 * 1 - adjusting old extent (we may have to split it)
2342 			 * 1 - add new extent
2343 			 * 1 - inode update
2344 			 */
2345 			trans = btrfs_start_transaction(root, 3);
2346 			if (IS_ERR(trans)) {
2347 				ret = PTR_ERR(trans);
2348 				goto out;
2349 			}
2350 
2351 			if (type == BTRFS_FILE_EXTENT_REG ||
2352 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2353 				/*
2354 				 *    a  | --- range to clone ---|  b
2355 				 * | ------------- extent ------------- |
2356 				 */
2357 
2358 				/* substract range b */
2359 				if (key.offset + datal > off + len)
2360 					datal = off + len - key.offset;
2361 
2362 				/* substract range a */
2363 				if (off > key.offset) {
2364 					datao += off - key.offset;
2365 					datal -= off - key.offset;
2366 				}
2367 
2368 				ret = btrfs_drop_extents(trans, inode,
2369 							 new_key.offset,
2370 							 new_key.offset + datal,
2371 							 &hint_byte, 1);
2372 				BUG_ON(ret);
2373 
2374 				ret = btrfs_insert_empty_item(trans, root, path,
2375 							      &new_key, size);
2376 				BUG_ON(ret);
2377 
2378 				leaf = path->nodes[0];
2379 				slot = path->slots[0];
2380 				write_extent_buffer(leaf, buf,
2381 					    btrfs_item_ptr_offset(leaf, slot),
2382 					    size);
2383 
2384 				extent = btrfs_item_ptr(leaf, slot,
2385 						struct btrfs_file_extent_item);
2386 
2387 				/* disko == 0 means it's a hole */
2388 				if (!disko)
2389 					datao = 0;
2390 
2391 				btrfs_set_file_extent_offset(leaf, extent,
2392 							     datao);
2393 				btrfs_set_file_extent_num_bytes(leaf, extent,
2394 								datal);
2395 				if (disko) {
2396 					inode_add_bytes(inode, datal);
2397 					ret = btrfs_inc_extent_ref(trans, root,
2398 							disko, diskl, 0,
2399 							root->root_key.objectid,
2400 							btrfs_ino(inode),
2401 							new_key.offset - datao);
2402 					BUG_ON(ret);
2403 				}
2404 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2405 				u64 skip = 0;
2406 				u64 trim = 0;
2407 				if (off > key.offset) {
2408 					skip = off - key.offset;
2409 					new_key.offset += skip;
2410 				}
2411 
2412 				if (key.offset + datal > off+len)
2413 					trim = key.offset + datal - (off+len);
2414 
2415 				if (comp && (skip || trim)) {
2416 					ret = -EINVAL;
2417 					btrfs_end_transaction(trans, root);
2418 					goto out;
2419 				}
2420 				size -= skip + trim;
2421 				datal -= skip + trim;
2422 
2423 				ret = btrfs_drop_extents(trans, inode,
2424 							 new_key.offset,
2425 							 new_key.offset + datal,
2426 							 &hint_byte, 1);
2427 				BUG_ON(ret);
2428 
2429 				ret = btrfs_insert_empty_item(trans, root, path,
2430 							      &new_key, size);
2431 				BUG_ON(ret);
2432 
2433 				if (skip) {
2434 					u32 start =
2435 					  btrfs_file_extent_calc_inline_size(0);
2436 					memmove(buf+start, buf+start+skip,
2437 						datal);
2438 				}
2439 
2440 				leaf = path->nodes[0];
2441 				slot = path->slots[0];
2442 				write_extent_buffer(leaf, buf,
2443 					    btrfs_item_ptr_offset(leaf, slot),
2444 					    size);
2445 				inode_add_bytes(inode, datal);
2446 			}
2447 
2448 			btrfs_mark_buffer_dirty(leaf);
2449 			btrfs_release_path(path);
2450 
2451 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2452 
2453 			/*
2454 			 * we round up to the block size at eof when
2455 			 * determining which extents to clone above,
2456 			 * but shouldn't round up the file size
2457 			 */
2458 			endoff = new_key.offset + datal;
2459 			if (endoff > destoff+olen)
2460 				endoff = destoff+olen;
2461 			if (endoff > inode->i_size)
2462 				btrfs_i_size_write(inode, endoff);
2463 
2464 			ret = btrfs_update_inode(trans, root, inode);
2465 			BUG_ON(ret);
2466 			btrfs_end_transaction(trans, root);
2467 		}
2468 next:
2469 		btrfs_release_path(path);
2470 		key.offset++;
2471 	}
2472 	ret = 0;
2473 out:
2474 	btrfs_release_path(path);
2475 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2476 out_unlock:
2477 	mutex_unlock(&src->i_mutex);
2478 	mutex_unlock(&inode->i_mutex);
2479 	vfree(buf);
2480 	btrfs_free_path(path);
2481 out_fput:
2482 	fput(src_file);
2483 out_drop_write:
2484 	mnt_drop_write(file->f_path.mnt);
2485 	return ret;
2486 }
2487 
2488 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2489 {
2490 	struct btrfs_ioctl_clone_range_args args;
2491 
2492 	if (copy_from_user(&args, argp, sizeof(args)))
2493 		return -EFAULT;
2494 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2495 				 args.src_length, args.dest_offset);
2496 }
2497 
2498 /*
2499  * there are many ways the trans_start and trans_end ioctls can lead
2500  * to deadlocks.  They should only be used by applications that
2501  * basically own the machine, and have a very in depth understanding
2502  * of all the possible deadlocks and enospc problems.
2503  */
2504 static long btrfs_ioctl_trans_start(struct file *file)
2505 {
2506 	struct inode *inode = fdentry(file)->d_inode;
2507 	struct btrfs_root *root = BTRFS_I(inode)->root;
2508 	struct btrfs_trans_handle *trans;
2509 	int ret;
2510 
2511 	ret = -EPERM;
2512 	if (!capable(CAP_SYS_ADMIN))
2513 		goto out;
2514 
2515 	ret = -EINPROGRESS;
2516 	if (file->private_data)
2517 		goto out;
2518 
2519 	ret = -EROFS;
2520 	if (btrfs_root_readonly(root))
2521 		goto out;
2522 
2523 	ret = mnt_want_write(file->f_path.mnt);
2524 	if (ret)
2525 		goto out;
2526 
2527 	atomic_inc(&root->fs_info->open_ioctl_trans);
2528 
2529 	ret = -ENOMEM;
2530 	trans = btrfs_start_ioctl_transaction(root);
2531 	if (IS_ERR(trans))
2532 		goto out_drop;
2533 
2534 	file->private_data = trans;
2535 	return 0;
2536 
2537 out_drop:
2538 	atomic_dec(&root->fs_info->open_ioctl_trans);
2539 	mnt_drop_write(file->f_path.mnt);
2540 out:
2541 	return ret;
2542 }
2543 
2544 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2545 {
2546 	struct inode *inode = fdentry(file)->d_inode;
2547 	struct btrfs_root *root = BTRFS_I(inode)->root;
2548 	struct btrfs_root *new_root;
2549 	struct btrfs_dir_item *di;
2550 	struct btrfs_trans_handle *trans;
2551 	struct btrfs_path *path;
2552 	struct btrfs_key location;
2553 	struct btrfs_disk_key disk_key;
2554 	struct btrfs_super_block *disk_super;
2555 	u64 features;
2556 	u64 objectid = 0;
2557 	u64 dir_id;
2558 
2559 	if (!capable(CAP_SYS_ADMIN))
2560 		return -EPERM;
2561 
2562 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2563 		return -EFAULT;
2564 
2565 	if (!objectid)
2566 		objectid = root->root_key.objectid;
2567 
2568 	location.objectid = objectid;
2569 	location.type = BTRFS_ROOT_ITEM_KEY;
2570 	location.offset = (u64)-1;
2571 
2572 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2573 	if (IS_ERR(new_root))
2574 		return PTR_ERR(new_root);
2575 
2576 	if (btrfs_root_refs(&new_root->root_item) == 0)
2577 		return -ENOENT;
2578 
2579 	path = btrfs_alloc_path();
2580 	if (!path)
2581 		return -ENOMEM;
2582 	path->leave_spinning = 1;
2583 
2584 	trans = btrfs_start_transaction(root, 1);
2585 	if (IS_ERR(trans)) {
2586 		btrfs_free_path(path);
2587 		return PTR_ERR(trans);
2588 	}
2589 
2590 	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2591 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2592 				   dir_id, "default", 7, 1);
2593 	if (IS_ERR_OR_NULL(di)) {
2594 		btrfs_free_path(path);
2595 		btrfs_end_transaction(trans, root);
2596 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2597 		       "this isn't going to work\n");
2598 		return -ENOENT;
2599 	}
2600 
2601 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2602 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2603 	btrfs_mark_buffer_dirty(path->nodes[0]);
2604 	btrfs_free_path(path);
2605 
2606 	disk_super = &root->fs_info->super_copy;
2607 	features = btrfs_super_incompat_flags(disk_super);
2608 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2609 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2610 		btrfs_set_super_incompat_flags(disk_super, features);
2611 	}
2612 	btrfs_end_transaction(trans, root);
2613 
2614 	return 0;
2615 }
2616 
2617 static void get_block_group_info(struct list_head *groups_list,
2618 				 struct btrfs_ioctl_space_info *space)
2619 {
2620 	struct btrfs_block_group_cache *block_group;
2621 
2622 	space->total_bytes = 0;
2623 	space->used_bytes = 0;
2624 	space->flags = 0;
2625 	list_for_each_entry(block_group, groups_list, list) {
2626 		space->flags = block_group->flags;
2627 		space->total_bytes += block_group->key.offset;
2628 		space->used_bytes +=
2629 			btrfs_block_group_used(&block_group->item);
2630 	}
2631 }
2632 
2633 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2634 {
2635 	struct btrfs_ioctl_space_args space_args;
2636 	struct btrfs_ioctl_space_info space;
2637 	struct btrfs_ioctl_space_info *dest;
2638 	struct btrfs_ioctl_space_info *dest_orig;
2639 	struct btrfs_ioctl_space_info __user *user_dest;
2640 	struct btrfs_space_info *info;
2641 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2642 		       BTRFS_BLOCK_GROUP_SYSTEM,
2643 		       BTRFS_BLOCK_GROUP_METADATA,
2644 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2645 	int num_types = 4;
2646 	int alloc_size;
2647 	int ret = 0;
2648 	u64 slot_count = 0;
2649 	int i, c;
2650 
2651 	if (copy_from_user(&space_args,
2652 			   (struct btrfs_ioctl_space_args __user *)arg,
2653 			   sizeof(space_args)))
2654 		return -EFAULT;
2655 
2656 	for (i = 0; i < num_types; i++) {
2657 		struct btrfs_space_info *tmp;
2658 
2659 		info = NULL;
2660 		rcu_read_lock();
2661 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2662 					list) {
2663 			if (tmp->flags == types[i]) {
2664 				info = tmp;
2665 				break;
2666 			}
2667 		}
2668 		rcu_read_unlock();
2669 
2670 		if (!info)
2671 			continue;
2672 
2673 		down_read(&info->groups_sem);
2674 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2675 			if (!list_empty(&info->block_groups[c]))
2676 				slot_count++;
2677 		}
2678 		up_read(&info->groups_sem);
2679 	}
2680 
2681 	/* space_slots == 0 means they are asking for a count */
2682 	if (space_args.space_slots == 0) {
2683 		space_args.total_spaces = slot_count;
2684 		goto out;
2685 	}
2686 
2687 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2688 
2689 	alloc_size = sizeof(*dest) * slot_count;
2690 
2691 	/* we generally have at most 6 or so space infos, one for each raid
2692 	 * level.  So, a whole page should be more than enough for everyone
2693 	 */
2694 	if (alloc_size > PAGE_CACHE_SIZE)
2695 		return -ENOMEM;
2696 
2697 	space_args.total_spaces = 0;
2698 	dest = kmalloc(alloc_size, GFP_NOFS);
2699 	if (!dest)
2700 		return -ENOMEM;
2701 	dest_orig = dest;
2702 
2703 	/* now we have a buffer to copy into */
2704 	for (i = 0; i < num_types; i++) {
2705 		struct btrfs_space_info *tmp;
2706 
2707 		if (!slot_count)
2708 			break;
2709 
2710 		info = NULL;
2711 		rcu_read_lock();
2712 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2713 					list) {
2714 			if (tmp->flags == types[i]) {
2715 				info = tmp;
2716 				break;
2717 			}
2718 		}
2719 		rcu_read_unlock();
2720 
2721 		if (!info)
2722 			continue;
2723 		down_read(&info->groups_sem);
2724 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2725 			if (!list_empty(&info->block_groups[c])) {
2726 				get_block_group_info(&info->block_groups[c],
2727 						     &space);
2728 				memcpy(dest, &space, sizeof(space));
2729 				dest++;
2730 				space_args.total_spaces++;
2731 				slot_count--;
2732 			}
2733 			if (!slot_count)
2734 				break;
2735 		}
2736 		up_read(&info->groups_sem);
2737 	}
2738 
2739 	user_dest = (struct btrfs_ioctl_space_info *)
2740 		(arg + sizeof(struct btrfs_ioctl_space_args));
2741 
2742 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2743 		ret = -EFAULT;
2744 
2745 	kfree(dest_orig);
2746 out:
2747 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2748 		ret = -EFAULT;
2749 
2750 	return ret;
2751 }
2752 
2753 /*
2754  * there are many ways the trans_start and trans_end ioctls can lead
2755  * to deadlocks.  They should only be used by applications that
2756  * basically own the machine, and have a very in depth understanding
2757  * of all the possible deadlocks and enospc problems.
2758  */
2759 long btrfs_ioctl_trans_end(struct file *file)
2760 {
2761 	struct inode *inode = fdentry(file)->d_inode;
2762 	struct btrfs_root *root = BTRFS_I(inode)->root;
2763 	struct btrfs_trans_handle *trans;
2764 
2765 	trans = file->private_data;
2766 	if (!trans)
2767 		return -EINVAL;
2768 	file->private_data = NULL;
2769 
2770 	btrfs_end_transaction(trans, root);
2771 
2772 	atomic_dec(&root->fs_info->open_ioctl_trans);
2773 
2774 	mnt_drop_write(file->f_path.mnt);
2775 	return 0;
2776 }
2777 
2778 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2779 {
2780 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2781 	struct btrfs_trans_handle *trans;
2782 	u64 transid;
2783 	int ret;
2784 
2785 	trans = btrfs_start_transaction(root, 0);
2786 	if (IS_ERR(trans))
2787 		return PTR_ERR(trans);
2788 	transid = trans->transid;
2789 	ret = btrfs_commit_transaction_async(trans, root, 0);
2790 	if (ret) {
2791 		btrfs_end_transaction(trans, root);
2792 		return ret;
2793 	}
2794 
2795 	if (argp)
2796 		if (copy_to_user(argp, &transid, sizeof(transid)))
2797 			return -EFAULT;
2798 	return 0;
2799 }
2800 
2801 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2802 {
2803 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2804 	u64 transid;
2805 
2806 	if (argp) {
2807 		if (copy_from_user(&transid, argp, sizeof(transid)))
2808 			return -EFAULT;
2809 	} else {
2810 		transid = 0;  /* current trans */
2811 	}
2812 	return btrfs_wait_for_commit(root, transid);
2813 }
2814 
2815 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2816 {
2817 	int ret;
2818 	struct btrfs_ioctl_scrub_args *sa;
2819 
2820 	if (!capable(CAP_SYS_ADMIN))
2821 		return -EPERM;
2822 
2823 	sa = memdup_user(arg, sizeof(*sa));
2824 	if (IS_ERR(sa))
2825 		return PTR_ERR(sa);
2826 
2827 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2828 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2829 
2830 	if (copy_to_user(arg, sa, sizeof(*sa)))
2831 		ret = -EFAULT;
2832 
2833 	kfree(sa);
2834 	return ret;
2835 }
2836 
2837 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2838 {
2839 	if (!capable(CAP_SYS_ADMIN))
2840 		return -EPERM;
2841 
2842 	return btrfs_scrub_cancel(root);
2843 }
2844 
2845 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2846 				       void __user *arg)
2847 {
2848 	struct btrfs_ioctl_scrub_args *sa;
2849 	int ret;
2850 
2851 	if (!capable(CAP_SYS_ADMIN))
2852 		return -EPERM;
2853 
2854 	sa = memdup_user(arg, sizeof(*sa));
2855 	if (IS_ERR(sa))
2856 		return PTR_ERR(sa);
2857 
2858 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2859 
2860 	if (copy_to_user(arg, sa, sizeof(*sa)))
2861 		ret = -EFAULT;
2862 
2863 	kfree(sa);
2864 	return ret;
2865 }
2866 
2867 long btrfs_ioctl(struct file *file, unsigned int
2868 		cmd, unsigned long arg)
2869 {
2870 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2871 	void __user *argp = (void __user *)arg;
2872 
2873 	switch (cmd) {
2874 	case FS_IOC_GETFLAGS:
2875 		return btrfs_ioctl_getflags(file, argp);
2876 	case FS_IOC_SETFLAGS:
2877 		return btrfs_ioctl_setflags(file, argp);
2878 	case FS_IOC_GETVERSION:
2879 		return btrfs_ioctl_getversion(file, argp);
2880 	case FITRIM:
2881 		return btrfs_ioctl_fitrim(file, argp);
2882 	case BTRFS_IOC_SNAP_CREATE:
2883 		return btrfs_ioctl_snap_create(file, argp, 0);
2884 	case BTRFS_IOC_SNAP_CREATE_V2:
2885 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
2886 	case BTRFS_IOC_SUBVOL_CREATE:
2887 		return btrfs_ioctl_snap_create(file, argp, 1);
2888 	case BTRFS_IOC_SNAP_DESTROY:
2889 		return btrfs_ioctl_snap_destroy(file, argp);
2890 	case BTRFS_IOC_SUBVOL_GETFLAGS:
2891 		return btrfs_ioctl_subvol_getflags(file, argp);
2892 	case BTRFS_IOC_SUBVOL_SETFLAGS:
2893 		return btrfs_ioctl_subvol_setflags(file, argp);
2894 	case BTRFS_IOC_DEFAULT_SUBVOL:
2895 		return btrfs_ioctl_default_subvol(file, argp);
2896 	case BTRFS_IOC_DEFRAG:
2897 		return btrfs_ioctl_defrag(file, NULL);
2898 	case BTRFS_IOC_DEFRAG_RANGE:
2899 		return btrfs_ioctl_defrag(file, argp);
2900 	case BTRFS_IOC_RESIZE:
2901 		return btrfs_ioctl_resize(root, argp);
2902 	case BTRFS_IOC_ADD_DEV:
2903 		return btrfs_ioctl_add_dev(root, argp);
2904 	case BTRFS_IOC_RM_DEV:
2905 		return btrfs_ioctl_rm_dev(root, argp);
2906 	case BTRFS_IOC_FS_INFO:
2907 		return btrfs_ioctl_fs_info(root, argp);
2908 	case BTRFS_IOC_DEV_INFO:
2909 		return btrfs_ioctl_dev_info(root, argp);
2910 	case BTRFS_IOC_BALANCE:
2911 		return btrfs_balance(root->fs_info->dev_root);
2912 	case BTRFS_IOC_CLONE:
2913 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2914 	case BTRFS_IOC_CLONE_RANGE:
2915 		return btrfs_ioctl_clone_range(file, argp);
2916 	case BTRFS_IOC_TRANS_START:
2917 		return btrfs_ioctl_trans_start(file);
2918 	case BTRFS_IOC_TRANS_END:
2919 		return btrfs_ioctl_trans_end(file);
2920 	case BTRFS_IOC_TREE_SEARCH:
2921 		return btrfs_ioctl_tree_search(file, argp);
2922 	case BTRFS_IOC_INO_LOOKUP:
2923 		return btrfs_ioctl_ino_lookup(file, argp);
2924 	case BTRFS_IOC_SPACE_INFO:
2925 		return btrfs_ioctl_space_info(root, argp);
2926 	case BTRFS_IOC_SYNC:
2927 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
2928 		return 0;
2929 	case BTRFS_IOC_START_SYNC:
2930 		return btrfs_ioctl_start_sync(file, argp);
2931 	case BTRFS_IOC_WAIT_SYNC:
2932 		return btrfs_ioctl_wait_sync(file, argp);
2933 	case BTRFS_IOC_SCRUB:
2934 		return btrfs_ioctl_scrub(root, argp);
2935 	case BTRFS_IOC_SCRUB_CANCEL:
2936 		return btrfs_ioctl_scrub_cancel(root, argp);
2937 	case BTRFS_IOC_SCRUB_PROGRESS:
2938 		return btrfs_ioctl_scrub_progress(root, argp);
2939 	}
2940 
2941 	return -ENOTTY;
2942 }
2943