xref: /linux/fs/btrfs/ioctl.c (revision 90ab5ee94171b3e28de6bb42ee30b527014e0be7)
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 #include "backref.h"
55 
56 /* Mask out flags that are inappropriate for the given type of inode. */
57 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
58 {
59 	if (S_ISDIR(mode))
60 		return flags;
61 	else if (S_ISREG(mode))
62 		return flags & ~FS_DIRSYNC_FL;
63 	else
64 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
65 }
66 
67 /*
68  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
69  */
70 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
71 {
72 	unsigned int iflags = 0;
73 
74 	if (flags & BTRFS_INODE_SYNC)
75 		iflags |= FS_SYNC_FL;
76 	if (flags & BTRFS_INODE_IMMUTABLE)
77 		iflags |= FS_IMMUTABLE_FL;
78 	if (flags & BTRFS_INODE_APPEND)
79 		iflags |= FS_APPEND_FL;
80 	if (flags & BTRFS_INODE_NODUMP)
81 		iflags |= FS_NODUMP_FL;
82 	if (flags & BTRFS_INODE_NOATIME)
83 		iflags |= FS_NOATIME_FL;
84 	if (flags & BTRFS_INODE_DIRSYNC)
85 		iflags |= FS_DIRSYNC_FL;
86 	if (flags & BTRFS_INODE_NODATACOW)
87 		iflags |= FS_NOCOW_FL;
88 
89 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
90 		iflags |= FS_COMPR_FL;
91 	else if (flags & BTRFS_INODE_NOCOMPRESS)
92 		iflags |= FS_NOCOMP_FL;
93 
94 	return iflags;
95 }
96 
97 /*
98  * Update inode->i_flags based on the btrfs internal flags.
99  */
100 void btrfs_update_iflags(struct inode *inode)
101 {
102 	struct btrfs_inode *ip = BTRFS_I(inode);
103 
104 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
105 
106 	if (ip->flags & BTRFS_INODE_SYNC)
107 		inode->i_flags |= S_SYNC;
108 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
109 		inode->i_flags |= S_IMMUTABLE;
110 	if (ip->flags & BTRFS_INODE_APPEND)
111 		inode->i_flags |= S_APPEND;
112 	if (ip->flags & BTRFS_INODE_NOATIME)
113 		inode->i_flags |= S_NOATIME;
114 	if (ip->flags & BTRFS_INODE_DIRSYNC)
115 		inode->i_flags |= S_DIRSYNC;
116 }
117 
118 /*
119  * Inherit flags from the parent inode.
120  *
121  * Currently only the compression flags and the cow flags are inherited.
122  */
123 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
124 {
125 	unsigned int flags;
126 
127 	if (!dir)
128 		return;
129 
130 	flags = BTRFS_I(dir)->flags;
131 
132 	if (flags & BTRFS_INODE_NOCOMPRESS) {
133 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
134 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
135 	} else if (flags & BTRFS_INODE_COMPRESS) {
136 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
137 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
138 	}
139 
140 	if (flags & BTRFS_INODE_NODATACOW)
141 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
142 
143 	btrfs_update_iflags(inode);
144 }
145 
146 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
147 {
148 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
149 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
150 
151 	if (copy_to_user(arg, &flags, sizeof(flags)))
152 		return -EFAULT;
153 	return 0;
154 }
155 
156 static int check_flags(unsigned int flags)
157 {
158 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
159 		      FS_NOATIME_FL | FS_NODUMP_FL | \
160 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
161 		      FS_NOCOMP_FL | FS_COMPR_FL |
162 		      FS_NOCOW_FL))
163 		return -EOPNOTSUPP;
164 
165 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
166 		return -EINVAL;
167 
168 	return 0;
169 }
170 
171 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
172 {
173 	struct inode *inode = file->f_path.dentry->d_inode;
174 	struct btrfs_inode *ip = BTRFS_I(inode);
175 	struct btrfs_root *root = ip->root;
176 	struct btrfs_trans_handle *trans;
177 	unsigned int flags, oldflags;
178 	int ret;
179 
180 	if (btrfs_root_readonly(root))
181 		return -EROFS;
182 
183 	if (copy_from_user(&flags, arg, sizeof(flags)))
184 		return -EFAULT;
185 
186 	ret = check_flags(flags);
187 	if (ret)
188 		return ret;
189 
190 	if (!inode_owner_or_capable(inode))
191 		return -EACCES;
192 
193 	mutex_lock(&inode->i_mutex);
194 
195 	flags = btrfs_mask_flags(inode->i_mode, flags);
196 	oldflags = btrfs_flags_to_ioctl(ip->flags);
197 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
198 		if (!capable(CAP_LINUX_IMMUTABLE)) {
199 			ret = -EPERM;
200 			goto out_unlock;
201 		}
202 	}
203 
204 	ret = mnt_want_write_file(file);
205 	if (ret)
206 		goto out_unlock;
207 
208 	if (flags & FS_SYNC_FL)
209 		ip->flags |= BTRFS_INODE_SYNC;
210 	else
211 		ip->flags &= ~BTRFS_INODE_SYNC;
212 	if (flags & FS_IMMUTABLE_FL)
213 		ip->flags |= BTRFS_INODE_IMMUTABLE;
214 	else
215 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
216 	if (flags & FS_APPEND_FL)
217 		ip->flags |= BTRFS_INODE_APPEND;
218 	else
219 		ip->flags &= ~BTRFS_INODE_APPEND;
220 	if (flags & FS_NODUMP_FL)
221 		ip->flags |= BTRFS_INODE_NODUMP;
222 	else
223 		ip->flags &= ~BTRFS_INODE_NODUMP;
224 	if (flags & FS_NOATIME_FL)
225 		ip->flags |= BTRFS_INODE_NOATIME;
226 	else
227 		ip->flags &= ~BTRFS_INODE_NOATIME;
228 	if (flags & FS_DIRSYNC_FL)
229 		ip->flags |= BTRFS_INODE_DIRSYNC;
230 	else
231 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
232 	if (flags & FS_NOCOW_FL)
233 		ip->flags |= BTRFS_INODE_NODATACOW;
234 	else
235 		ip->flags &= ~BTRFS_INODE_NODATACOW;
236 
237 	/*
238 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
239 	 * flag may be changed automatically if compression code won't make
240 	 * things smaller.
241 	 */
242 	if (flags & FS_NOCOMP_FL) {
243 		ip->flags &= ~BTRFS_INODE_COMPRESS;
244 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
245 	} else if (flags & FS_COMPR_FL) {
246 		ip->flags |= BTRFS_INODE_COMPRESS;
247 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
248 	} else {
249 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
250 	}
251 
252 	trans = btrfs_join_transaction(root);
253 	BUG_ON(IS_ERR(trans));
254 
255 	btrfs_update_iflags(inode);
256 	inode->i_ctime = CURRENT_TIME;
257 	ret = btrfs_update_inode(trans, root, inode);
258 	BUG_ON(ret);
259 
260 	btrfs_end_transaction(trans, root);
261 
262 	mnt_drop_write_file(file);
263 
264 	ret = 0;
265  out_unlock:
266 	mutex_unlock(&inode->i_mutex);
267 	return ret;
268 }
269 
270 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
271 {
272 	struct inode *inode = file->f_path.dentry->d_inode;
273 
274 	return put_user(inode->i_generation, arg);
275 }
276 
277 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
278 {
279 	struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
280 	struct btrfs_fs_info *fs_info = root->fs_info;
281 	struct btrfs_device *device;
282 	struct request_queue *q;
283 	struct fstrim_range range;
284 	u64 minlen = ULLONG_MAX;
285 	u64 num_devices = 0;
286 	u64 total_bytes = btrfs_super_total_bytes(root->fs_info->super_copy);
287 	int ret;
288 
289 	if (!capable(CAP_SYS_ADMIN))
290 		return -EPERM;
291 
292 	rcu_read_lock();
293 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
294 				dev_list) {
295 		if (!device->bdev)
296 			continue;
297 		q = bdev_get_queue(device->bdev);
298 		if (blk_queue_discard(q)) {
299 			num_devices++;
300 			minlen = min((u64)q->limits.discard_granularity,
301 				     minlen);
302 		}
303 	}
304 	rcu_read_unlock();
305 
306 	if (!num_devices)
307 		return -EOPNOTSUPP;
308 	if (copy_from_user(&range, arg, sizeof(range)))
309 		return -EFAULT;
310 	if (range.start > total_bytes)
311 		return -EINVAL;
312 
313 	range.len = min(range.len, total_bytes - range.start);
314 	range.minlen = max(range.minlen, minlen);
315 	ret = btrfs_trim_fs(root, &range);
316 	if (ret < 0)
317 		return ret;
318 
319 	if (copy_to_user(arg, &range, sizeof(range)))
320 		return -EFAULT;
321 
322 	return 0;
323 }
324 
325 static noinline int create_subvol(struct btrfs_root *root,
326 				  struct dentry *dentry,
327 				  char *name, int namelen,
328 				  u64 *async_transid)
329 {
330 	struct btrfs_trans_handle *trans;
331 	struct btrfs_key key;
332 	struct btrfs_root_item root_item;
333 	struct btrfs_inode_item *inode_item;
334 	struct extent_buffer *leaf;
335 	struct btrfs_root *new_root;
336 	struct dentry *parent = dentry->d_parent;
337 	struct inode *dir;
338 	int ret;
339 	int err;
340 	u64 objectid;
341 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
342 	u64 index = 0;
343 
344 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
345 	if (ret)
346 		return ret;
347 
348 	dir = parent->d_inode;
349 
350 	/*
351 	 * 1 - inode item
352 	 * 2 - refs
353 	 * 1 - root item
354 	 * 2 - dir items
355 	 */
356 	trans = btrfs_start_transaction(root, 6);
357 	if (IS_ERR(trans))
358 		return PTR_ERR(trans);
359 
360 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
361 				      0, objectid, NULL, 0, 0, 0);
362 	if (IS_ERR(leaf)) {
363 		ret = PTR_ERR(leaf);
364 		goto fail;
365 	}
366 
367 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
368 	btrfs_set_header_bytenr(leaf, leaf->start);
369 	btrfs_set_header_generation(leaf, trans->transid);
370 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
371 	btrfs_set_header_owner(leaf, objectid);
372 
373 	write_extent_buffer(leaf, root->fs_info->fsid,
374 			    (unsigned long)btrfs_header_fsid(leaf),
375 			    BTRFS_FSID_SIZE);
376 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
377 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
378 			    BTRFS_UUID_SIZE);
379 	btrfs_mark_buffer_dirty(leaf);
380 
381 	inode_item = &root_item.inode;
382 	memset(inode_item, 0, sizeof(*inode_item));
383 	inode_item->generation = cpu_to_le64(1);
384 	inode_item->size = cpu_to_le64(3);
385 	inode_item->nlink = cpu_to_le32(1);
386 	inode_item->nbytes = cpu_to_le64(root->leafsize);
387 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
388 
389 	root_item.flags = 0;
390 	root_item.byte_limit = 0;
391 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
392 
393 	btrfs_set_root_bytenr(&root_item, leaf->start);
394 	btrfs_set_root_generation(&root_item, trans->transid);
395 	btrfs_set_root_level(&root_item, 0);
396 	btrfs_set_root_refs(&root_item, 1);
397 	btrfs_set_root_used(&root_item, leaf->len);
398 	btrfs_set_root_last_snapshot(&root_item, 0);
399 
400 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
401 	root_item.drop_level = 0;
402 
403 	btrfs_tree_unlock(leaf);
404 	free_extent_buffer(leaf);
405 	leaf = NULL;
406 
407 	btrfs_set_root_dirid(&root_item, new_dirid);
408 
409 	key.objectid = objectid;
410 	key.offset = 0;
411 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
412 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
413 				&root_item);
414 	if (ret)
415 		goto fail;
416 
417 	key.offset = (u64)-1;
418 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
419 	BUG_ON(IS_ERR(new_root));
420 
421 	btrfs_record_root_in_trans(trans, new_root);
422 
423 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
424 	/*
425 	 * insert the directory item
426 	 */
427 	ret = btrfs_set_inode_index(dir, &index);
428 	BUG_ON(ret);
429 
430 	ret = btrfs_insert_dir_item(trans, root,
431 				    name, namelen, dir, &key,
432 				    BTRFS_FT_DIR, index);
433 	if (ret)
434 		goto fail;
435 
436 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
437 	ret = btrfs_update_inode(trans, root, dir);
438 	BUG_ON(ret);
439 
440 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
441 				 objectid, root->root_key.objectid,
442 				 btrfs_ino(dir), index, name, namelen);
443 
444 	BUG_ON(ret);
445 
446 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
447 fail:
448 	if (async_transid) {
449 		*async_transid = trans->transid;
450 		err = btrfs_commit_transaction_async(trans, root, 1);
451 	} else {
452 		err = btrfs_commit_transaction(trans, root);
453 	}
454 	if (err && !ret)
455 		ret = err;
456 	return ret;
457 }
458 
459 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
460 			   char *name, int namelen, u64 *async_transid,
461 			   bool readonly)
462 {
463 	struct inode *inode;
464 	struct btrfs_pending_snapshot *pending_snapshot;
465 	struct btrfs_trans_handle *trans;
466 	int ret;
467 
468 	if (!root->ref_cows)
469 		return -EINVAL;
470 
471 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
472 	if (!pending_snapshot)
473 		return -ENOMEM;
474 
475 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
476 	pending_snapshot->dentry = dentry;
477 	pending_snapshot->root = root;
478 	pending_snapshot->readonly = readonly;
479 
480 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
481 	if (IS_ERR(trans)) {
482 		ret = PTR_ERR(trans);
483 		goto fail;
484 	}
485 
486 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
487 	BUG_ON(ret);
488 
489 	spin_lock(&root->fs_info->trans_lock);
490 	list_add(&pending_snapshot->list,
491 		 &trans->transaction->pending_snapshots);
492 	spin_unlock(&root->fs_info->trans_lock);
493 	if (async_transid) {
494 		*async_transid = trans->transid;
495 		ret = btrfs_commit_transaction_async(trans,
496 				     root->fs_info->extent_root, 1);
497 	} else {
498 		ret = btrfs_commit_transaction(trans,
499 					       root->fs_info->extent_root);
500 	}
501 	BUG_ON(ret);
502 
503 	ret = pending_snapshot->error;
504 	if (ret)
505 		goto fail;
506 
507 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
508 	if (ret)
509 		goto fail;
510 
511 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
512 	if (IS_ERR(inode)) {
513 		ret = PTR_ERR(inode);
514 		goto fail;
515 	}
516 	BUG_ON(!inode);
517 	d_instantiate(dentry, inode);
518 	ret = 0;
519 fail:
520 	kfree(pending_snapshot);
521 	return ret;
522 }
523 
524 /*  copy of check_sticky in fs/namei.c()
525 * It's inline, so penalty for filesystems that don't use sticky bit is
526 * minimal.
527 */
528 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
529 {
530 	uid_t fsuid = current_fsuid();
531 
532 	if (!(dir->i_mode & S_ISVTX))
533 		return 0;
534 	if (inode->i_uid == fsuid)
535 		return 0;
536 	if (dir->i_uid == fsuid)
537 		return 0;
538 	return !capable(CAP_FOWNER);
539 }
540 
541 /*  copy of may_delete in fs/namei.c()
542  *	Check whether we can remove a link victim from directory dir, check
543  *  whether the type of victim is right.
544  *  1. We can't do it if dir is read-only (done in permission())
545  *  2. We should have write and exec permissions on dir
546  *  3. We can't remove anything from append-only dir
547  *  4. We can't do anything with immutable dir (done in permission())
548  *  5. If the sticky bit on dir is set we should either
549  *	a. be owner of dir, or
550  *	b. be owner of victim, or
551  *	c. have CAP_FOWNER capability
552  *  6. If the victim is append-only or immutable we can't do antyhing with
553  *     links pointing to it.
554  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
555  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
556  *  9. We can't remove a root or mountpoint.
557  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
558  *     nfs_async_unlink().
559  */
560 
561 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
562 {
563 	int error;
564 
565 	if (!victim->d_inode)
566 		return -ENOENT;
567 
568 	BUG_ON(victim->d_parent->d_inode != dir);
569 	audit_inode_child(victim, dir);
570 
571 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
572 	if (error)
573 		return error;
574 	if (IS_APPEND(dir))
575 		return -EPERM;
576 	if (btrfs_check_sticky(dir, victim->d_inode)||
577 		IS_APPEND(victim->d_inode)||
578 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
579 		return -EPERM;
580 	if (isdir) {
581 		if (!S_ISDIR(victim->d_inode->i_mode))
582 			return -ENOTDIR;
583 		if (IS_ROOT(victim))
584 			return -EBUSY;
585 	} else if (S_ISDIR(victim->d_inode->i_mode))
586 		return -EISDIR;
587 	if (IS_DEADDIR(dir))
588 		return -ENOENT;
589 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
590 		return -EBUSY;
591 	return 0;
592 }
593 
594 /* copy of may_create in fs/namei.c() */
595 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
596 {
597 	if (child->d_inode)
598 		return -EEXIST;
599 	if (IS_DEADDIR(dir))
600 		return -ENOENT;
601 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
602 }
603 
604 /*
605  * Create a new subvolume below @parent.  This is largely modeled after
606  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
607  * inside this filesystem so it's quite a bit simpler.
608  */
609 static noinline int btrfs_mksubvol(struct path *parent,
610 				   char *name, int namelen,
611 				   struct btrfs_root *snap_src,
612 				   u64 *async_transid, bool readonly)
613 {
614 	struct inode *dir  = parent->dentry->d_inode;
615 	struct dentry *dentry;
616 	int error;
617 
618 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
619 
620 	dentry = lookup_one_len(name, parent->dentry, namelen);
621 	error = PTR_ERR(dentry);
622 	if (IS_ERR(dentry))
623 		goto out_unlock;
624 
625 	error = -EEXIST;
626 	if (dentry->d_inode)
627 		goto out_dput;
628 
629 	error = mnt_want_write(parent->mnt);
630 	if (error)
631 		goto out_dput;
632 
633 	error = btrfs_may_create(dir, dentry);
634 	if (error)
635 		goto out_drop_write;
636 
637 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
638 
639 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
640 		goto out_up_read;
641 
642 	if (snap_src) {
643 		error = create_snapshot(snap_src, dentry,
644 					name, namelen, async_transid, readonly);
645 	} else {
646 		error = create_subvol(BTRFS_I(dir)->root, dentry,
647 				      name, namelen, async_transid);
648 	}
649 	if (!error)
650 		fsnotify_mkdir(dir, dentry);
651 out_up_read:
652 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
653 out_drop_write:
654 	mnt_drop_write(parent->mnt);
655 out_dput:
656 	dput(dentry);
657 out_unlock:
658 	mutex_unlock(&dir->i_mutex);
659 	return error;
660 }
661 
662 /*
663  * When we're defragging a range, we don't want to kick it off again
664  * if it is really just waiting for delalloc to send it down.
665  * If we find a nice big extent or delalloc range for the bytes in the
666  * file you want to defrag, we return 0 to let you know to skip this
667  * part of the file
668  */
669 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
670 {
671 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
672 	struct extent_map *em = NULL;
673 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
674 	u64 end;
675 
676 	read_lock(&em_tree->lock);
677 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
678 	read_unlock(&em_tree->lock);
679 
680 	if (em) {
681 		end = extent_map_end(em);
682 		free_extent_map(em);
683 		if (end - offset > thresh)
684 			return 0;
685 	}
686 	/* if we already have a nice delalloc here, just stop */
687 	thresh /= 2;
688 	end = count_range_bits(io_tree, &offset, offset + thresh,
689 			       thresh, EXTENT_DELALLOC, 1);
690 	if (end >= thresh)
691 		return 0;
692 	return 1;
693 }
694 
695 /*
696  * helper function to walk through a file and find extents
697  * newer than a specific transid, and smaller than thresh.
698  *
699  * This is used by the defragging code to find new and small
700  * extents
701  */
702 static int find_new_extents(struct btrfs_root *root,
703 			    struct inode *inode, u64 newer_than,
704 			    u64 *off, int thresh)
705 {
706 	struct btrfs_path *path;
707 	struct btrfs_key min_key;
708 	struct btrfs_key max_key;
709 	struct extent_buffer *leaf;
710 	struct btrfs_file_extent_item *extent;
711 	int type;
712 	int ret;
713 	u64 ino = btrfs_ino(inode);
714 
715 	path = btrfs_alloc_path();
716 	if (!path)
717 		return -ENOMEM;
718 
719 	min_key.objectid = ino;
720 	min_key.type = BTRFS_EXTENT_DATA_KEY;
721 	min_key.offset = *off;
722 
723 	max_key.objectid = ino;
724 	max_key.type = (u8)-1;
725 	max_key.offset = (u64)-1;
726 
727 	path->keep_locks = 1;
728 
729 	while(1) {
730 		ret = btrfs_search_forward(root, &min_key, &max_key,
731 					   path, 0, newer_than);
732 		if (ret != 0)
733 			goto none;
734 		if (min_key.objectid != ino)
735 			goto none;
736 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
737 			goto none;
738 
739 		leaf = path->nodes[0];
740 		extent = btrfs_item_ptr(leaf, path->slots[0],
741 					struct btrfs_file_extent_item);
742 
743 		type = btrfs_file_extent_type(leaf, extent);
744 		if (type == BTRFS_FILE_EXTENT_REG &&
745 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
746 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
747 			*off = min_key.offset;
748 			btrfs_free_path(path);
749 			return 0;
750 		}
751 
752 		if (min_key.offset == (u64)-1)
753 			goto none;
754 
755 		min_key.offset++;
756 		btrfs_release_path(path);
757 	}
758 none:
759 	btrfs_free_path(path);
760 	return -ENOENT;
761 }
762 
763 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
764 			       int thresh, u64 *last_len, u64 *skip,
765 			       u64 *defrag_end)
766 {
767 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
768 	struct extent_map *em = NULL;
769 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
770 	int ret = 1;
771 
772 	/*
773 	 * make sure that once we start defragging an extent, we keep on
774 	 * defragging it
775 	 */
776 	if (start < *defrag_end)
777 		return 1;
778 
779 	*skip = 0;
780 
781 	/*
782 	 * hopefully we have this extent in the tree already, try without
783 	 * the full extent lock
784 	 */
785 	read_lock(&em_tree->lock);
786 	em = lookup_extent_mapping(em_tree, start, len);
787 	read_unlock(&em_tree->lock);
788 
789 	if (!em) {
790 		/* get the big lock and read metadata off disk */
791 		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
792 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
793 		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
794 
795 		if (IS_ERR(em))
796 			return 0;
797 	}
798 
799 	/* this will cover holes, and inline extents */
800 	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
801 		ret = 0;
802 
803 	/*
804 	 * we hit a real extent, if it is big don't bother defragging it again
805 	 */
806 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
807 		ret = 0;
808 
809 	/*
810 	 * last_len ends up being a counter of how many bytes we've defragged.
811 	 * every time we choose not to defrag an extent, we reset *last_len
812 	 * so that the next tiny extent will force a defrag.
813 	 *
814 	 * The end result of this is that tiny extents before a single big
815 	 * extent will force at least part of that big extent to be defragged.
816 	 */
817 	if (ret) {
818 		*defrag_end = extent_map_end(em);
819 	} else {
820 		*last_len = 0;
821 		*skip = extent_map_end(em);
822 		*defrag_end = 0;
823 	}
824 
825 	free_extent_map(em);
826 	return ret;
827 }
828 
829 /*
830  * it doesn't do much good to defrag one or two pages
831  * at a time.  This pulls in a nice chunk of pages
832  * to COW and defrag.
833  *
834  * It also makes sure the delalloc code has enough
835  * dirty data to avoid making new small extents as part
836  * of the defrag
837  *
838  * It's a good idea to start RA on this range
839  * before calling this.
840  */
841 static int cluster_pages_for_defrag(struct inode *inode,
842 				    struct page **pages,
843 				    unsigned long start_index,
844 				    int num_pages)
845 {
846 	unsigned long file_end;
847 	u64 isize = i_size_read(inode);
848 	u64 page_start;
849 	u64 page_end;
850 	int ret;
851 	int i;
852 	int i_done;
853 	struct btrfs_ordered_extent *ordered;
854 	struct extent_state *cached_state = NULL;
855 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
856 
857 	if (isize == 0)
858 		return 0;
859 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
860 
861 	mutex_lock(&inode->i_mutex);
862 	ret = btrfs_delalloc_reserve_space(inode,
863 					   num_pages << PAGE_CACHE_SHIFT);
864 	mutex_unlock(&inode->i_mutex);
865 	if (ret)
866 		return ret;
867 again:
868 	ret = 0;
869 	i_done = 0;
870 
871 	/* step one, lock all the pages */
872 	for (i = 0; i < num_pages; i++) {
873 		struct page *page;
874 		page = find_or_create_page(inode->i_mapping,
875 					    start_index + i, mask);
876 		if (!page)
877 			break;
878 
879 		if (!PageUptodate(page)) {
880 			btrfs_readpage(NULL, page);
881 			lock_page(page);
882 			if (!PageUptodate(page)) {
883 				unlock_page(page);
884 				page_cache_release(page);
885 				ret = -EIO;
886 				break;
887 			}
888 		}
889 		isize = i_size_read(inode);
890 		file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
891 		if (!isize || page->index > file_end ||
892 		    page->mapping != inode->i_mapping) {
893 			/* whoops, we blew past eof, skip this page */
894 			unlock_page(page);
895 			page_cache_release(page);
896 			break;
897 		}
898 		pages[i] = page;
899 		i_done++;
900 	}
901 	if (!i_done || ret)
902 		goto out;
903 
904 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
905 		goto out;
906 
907 	/*
908 	 * so now we have a nice long stream of locked
909 	 * and up to date pages, lets wait on them
910 	 */
911 	for (i = 0; i < i_done; i++)
912 		wait_on_page_writeback(pages[i]);
913 
914 	page_start = page_offset(pages[0]);
915 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
916 
917 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
918 			 page_start, page_end - 1, 0, &cached_state,
919 			 GFP_NOFS);
920 	ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
921 	if (ordered &&
922 	    ordered->file_offset + ordered->len > page_start &&
923 	    ordered->file_offset < page_end) {
924 		btrfs_put_ordered_extent(ordered);
925 		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
926 				     page_start, page_end - 1,
927 				     &cached_state, GFP_NOFS);
928 		for (i = 0; i < i_done; i++) {
929 			unlock_page(pages[i]);
930 			page_cache_release(pages[i]);
931 		}
932 		btrfs_wait_ordered_range(inode, page_start,
933 					 page_end - page_start);
934 		goto again;
935 	}
936 	if (ordered)
937 		btrfs_put_ordered_extent(ordered);
938 
939 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
940 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
941 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
942 			  GFP_NOFS);
943 
944 	if (i_done != num_pages) {
945 		spin_lock(&BTRFS_I(inode)->lock);
946 		BTRFS_I(inode)->outstanding_extents++;
947 		spin_unlock(&BTRFS_I(inode)->lock);
948 		btrfs_delalloc_release_space(inode,
949 				     (num_pages - i_done) << PAGE_CACHE_SHIFT);
950 	}
951 
952 
953 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
954 				  &cached_state);
955 
956 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
957 			     page_start, page_end - 1, &cached_state,
958 			     GFP_NOFS);
959 
960 	for (i = 0; i < i_done; i++) {
961 		clear_page_dirty_for_io(pages[i]);
962 		ClearPageChecked(pages[i]);
963 		set_page_extent_mapped(pages[i]);
964 		set_page_dirty(pages[i]);
965 		unlock_page(pages[i]);
966 		page_cache_release(pages[i]);
967 	}
968 	return i_done;
969 out:
970 	for (i = 0; i < i_done; i++) {
971 		unlock_page(pages[i]);
972 		page_cache_release(pages[i]);
973 	}
974 	btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
975 	return ret;
976 
977 }
978 
979 int btrfs_defrag_file(struct inode *inode, struct file *file,
980 		      struct btrfs_ioctl_defrag_range_args *range,
981 		      u64 newer_than, unsigned long max_to_defrag)
982 {
983 	struct btrfs_root *root = BTRFS_I(inode)->root;
984 	struct btrfs_super_block *disk_super;
985 	struct file_ra_state *ra = NULL;
986 	unsigned long last_index;
987 	u64 isize = i_size_read(inode);
988 	u64 features;
989 	u64 last_len = 0;
990 	u64 skip = 0;
991 	u64 defrag_end = 0;
992 	u64 newer_off = range->start;
993 	unsigned long i;
994 	unsigned long ra_index = 0;
995 	int ret;
996 	int defrag_count = 0;
997 	int compress_type = BTRFS_COMPRESS_ZLIB;
998 	int extent_thresh = range->extent_thresh;
999 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1000 	int cluster = max_cluster;
1001 	u64 new_align = ~((u64)128 * 1024 - 1);
1002 	struct page **pages = NULL;
1003 
1004 	if (extent_thresh == 0)
1005 		extent_thresh = 256 * 1024;
1006 
1007 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1008 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1009 			return -EINVAL;
1010 		if (range->compress_type)
1011 			compress_type = range->compress_type;
1012 	}
1013 
1014 	if (isize == 0)
1015 		return 0;
1016 
1017 	/*
1018 	 * if we were not given a file, allocate a readahead
1019 	 * context
1020 	 */
1021 	if (!file) {
1022 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1023 		if (!ra)
1024 			return -ENOMEM;
1025 		file_ra_state_init(ra, inode->i_mapping);
1026 	} else {
1027 		ra = &file->f_ra;
1028 	}
1029 
1030 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1031 			GFP_NOFS);
1032 	if (!pages) {
1033 		ret = -ENOMEM;
1034 		goto out_ra;
1035 	}
1036 
1037 	/* find the last page to defrag */
1038 	if (range->start + range->len > range->start) {
1039 		last_index = min_t(u64, isize - 1,
1040 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1041 	} else {
1042 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1043 	}
1044 
1045 	if (newer_than) {
1046 		ret = find_new_extents(root, inode, newer_than,
1047 				       &newer_off, 64 * 1024);
1048 		if (!ret) {
1049 			range->start = newer_off;
1050 			/*
1051 			 * we always align our defrag to help keep
1052 			 * the extents in the file evenly spaced
1053 			 */
1054 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1055 		} else
1056 			goto out_ra;
1057 	} else {
1058 		i = range->start >> PAGE_CACHE_SHIFT;
1059 	}
1060 	if (!max_to_defrag)
1061 		max_to_defrag = last_index;
1062 
1063 	/*
1064 	 * make writeback starts from i, so the defrag range can be
1065 	 * written sequentially.
1066 	 */
1067 	if (i < inode->i_mapping->writeback_index)
1068 		inode->i_mapping->writeback_index = i;
1069 
1070 	while (i <= last_index && defrag_count < max_to_defrag &&
1071 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1072 		PAGE_CACHE_SHIFT)) {
1073 		/*
1074 		 * make sure we stop running if someone unmounts
1075 		 * the FS
1076 		 */
1077 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1078 			break;
1079 
1080 		if (!newer_than &&
1081 		    !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1082 					PAGE_CACHE_SIZE,
1083 					extent_thresh,
1084 					&last_len, &skip,
1085 					&defrag_end)) {
1086 			unsigned long next;
1087 			/*
1088 			 * the should_defrag function tells us how much to skip
1089 			 * bump our counter by the suggested amount
1090 			 */
1091 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1092 			i = max(i + 1, next);
1093 			continue;
1094 		}
1095 
1096 		if (!newer_than) {
1097 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1098 				   PAGE_CACHE_SHIFT) - i;
1099 			cluster = min(cluster, max_cluster);
1100 		} else {
1101 			cluster = max_cluster;
1102 		}
1103 
1104 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1105 			BTRFS_I(inode)->force_compress = compress_type;
1106 
1107 		if (i + cluster > ra_index) {
1108 			ra_index = max(i, ra_index);
1109 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1110 				       cluster);
1111 			ra_index += max_cluster;
1112 		}
1113 
1114 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1115 		if (ret < 0)
1116 			goto out_ra;
1117 
1118 		defrag_count += ret;
1119 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1120 
1121 		if (newer_than) {
1122 			if (newer_off == (u64)-1)
1123 				break;
1124 
1125 			newer_off = max(newer_off + 1,
1126 					(u64)i << PAGE_CACHE_SHIFT);
1127 
1128 			ret = find_new_extents(root, inode,
1129 					       newer_than, &newer_off,
1130 					       64 * 1024);
1131 			if (!ret) {
1132 				range->start = newer_off;
1133 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1134 			} else {
1135 				break;
1136 			}
1137 		} else {
1138 			if (ret > 0) {
1139 				i += ret;
1140 				last_len += ret << PAGE_CACHE_SHIFT;
1141 			} else {
1142 				i++;
1143 				last_len = 0;
1144 			}
1145 		}
1146 	}
1147 
1148 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1149 		filemap_flush(inode->i_mapping);
1150 
1151 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1152 		/* the filemap_flush will queue IO into the worker threads, but
1153 		 * we have to make sure the IO is actually started and that
1154 		 * ordered extents get created before we return
1155 		 */
1156 		atomic_inc(&root->fs_info->async_submit_draining);
1157 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1158 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1159 			wait_event(root->fs_info->async_submit_wait,
1160 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1161 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1162 		}
1163 		atomic_dec(&root->fs_info->async_submit_draining);
1164 
1165 		mutex_lock(&inode->i_mutex);
1166 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1167 		mutex_unlock(&inode->i_mutex);
1168 	}
1169 
1170 	disk_super = root->fs_info->super_copy;
1171 	features = btrfs_super_incompat_flags(disk_super);
1172 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1173 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1174 		btrfs_set_super_incompat_flags(disk_super, features);
1175 	}
1176 
1177 	ret = defrag_count;
1178 
1179 out_ra:
1180 	if (!file)
1181 		kfree(ra);
1182 	kfree(pages);
1183 	return ret;
1184 }
1185 
1186 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1187 					void __user *arg)
1188 {
1189 	u64 new_size;
1190 	u64 old_size;
1191 	u64 devid = 1;
1192 	struct btrfs_ioctl_vol_args *vol_args;
1193 	struct btrfs_trans_handle *trans;
1194 	struct btrfs_device *device = NULL;
1195 	char *sizestr;
1196 	char *devstr = NULL;
1197 	int ret = 0;
1198 	int mod = 0;
1199 
1200 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1201 		return -EROFS;
1202 
1203 	if (!capable(CAP_SYS_ADMIN))
1204 		return -EPERM;
1205 
1206 	vol_args = memdup_user(arg, sizeof(*vol_args));
1207 	if (IS_ERR(vol_args))
1208 		return PTR_ERR(vol_args);
1209 
1210 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1211 
1212 	mutex_lock(&root->fs_info->volume_mutex);
1213 	sizestr = vol_args->name;
1214 	devstr = strchr(sizestr, ':');
1215 	if (devstr) {
1216 		char *end;
1217 		sizestr = devstr + 1;
1218 		*devstr = '\0';
1219 		devstr = vol_args->name;
1220 		devid = simple_strtoull(devstr, &end, 10);
1221 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1222 		       (unsigned long long)devid);
1223 	}
1224 	device = btrfs_find_device(root, devid, NULL, NULL);
1225 	if (!device) {
1226 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1227 		       (unsigned long long)devid);
1228 		ret = -EINVAL;
1229 		goto out_unlock;
1230 	}
1231 	if (!strcmp(sizestr, "max"))
1232 		new_size = device->bdev->bd_inode->i_size;
1233 	else {
1234 		if (sizestr[0] == '-') {
1235 			mod = -1;
1236 			sizestr++;
1237 		} else if (sizestr[0] == '+') {
1238 			mod = 1;
1239 			sizestr++;
1240 		}
1241 		new_size = memparse(sizestr, NULL);
1242 		if (new_size == 0) {
1243 			ret = -EINVAL;
1244 			goto out_unlock;
1245 		}
1246 	}
1247 
1248 	old_size = device->total_bytes;
1249 
1250 	if (mod < 0) {
1251 		if (new_size > old_size) {
1252 			ret = -EINVAL;
1253 			goto out_unlock;
1254 		}
1255 		new_size = old_size - new_size;
1256 	} else if (mod > 0) {
1257 		new_size = old_size + new_size;
1258 	}
1259 
1260 	if (new_size < 256 * 1024 * 1024) {
1261 		ret = -EINVAL;
1262 		goto out_unlock;
1263 	}
1264 	if (new_size > device->bdev->bd_inode->i_size) {
1265 		ret = -EFBIG;
1266 		goto out_unlock;
1267 	}
1268 
1269 	do_div(new_size, root->sectorsize);
1270 	new_size *= root->sectorsize;
1271 
1272 	printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1273 		device->name, (unsigned long long)new_size);
1274 
1275 	if (new_size > old_size) {
1276 		trans = btrfs_start_transaction(root, 0);
1277 		if (IS_ERR(trans)) {
1278 			ret = PTR_ERR(trans);
1279 			goto out_unlock;
1280 		}
1281 		ret = btrfs_grow_device(trans, device, new_size);
1282 		btrfs_commit_transaction(trans, root);
1283 	} else if (new_size < old_size) {
1284 		ret = btrfs_shrink_device(device, new_size);
1285 	}
1286 
1287 out_unlock:
1288 	mutex_unlock(&root->fs_info->volume_mutex);
1289 	kfree(vol_args);
1290 	return ret;
1291 }
1292 
1293 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1294 						    char *name,
1295 						    unsigned long fd,
1296 						    int subvol,
1297 						    u64 *transid,
1298 						    bool readonly)
1299 {
1300 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1301 	struct file *src_file;
1302 	int namelen;
1303 	int ret = 0;
1304 
1305 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1306 		return -EROFS;
1307 
1308 	namelen = strlen(name);
1309 	if (strchr(name, '/')) {
1310 		ret = -EINVAL;
1311 		goto out;
1312 	}
1313 
1314 	if (subvol) {
1315 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1316 				     NULL, transid, readonly);
1317 	} else {
1318 		struct inode *src_inode;
1319 		src_file = fget(fd);
1320 		if (!src_file) {
1321 			ret = -EINVAL;
1322 			goto out;
1323 		}
1324 
1325 		src_inode = src_file->f_path.dentry->d_inode;
1326 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1327 			printk(KERN_INFO "btrfs: Snapshot src from "
1328 			       "another FS\n");
1329 			ret = -EINVAL;
1330 			fput(src_file);
1331 			goto out;
1332 		}
1333 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1334 				     BTRFS_I(src_inode)->root,
1335 				     transid, readonly);
1336 		fput(src_file);
1337 	}
1338 out:
1339 	return ret;
1340 }
1341 
1342 static noinline int btrfs_ioctl_snap_create(struct file *file,
1343 					    void __user *arg, int subvol)
1344 {
1345 	struct btrfs_ioctl_vol_args *vol_args;
1346 	int ret;
1347 
1348 	vol_args = memdup_user(arg, sizeof(*vol_args));
1349 	if (IS_ERR(vol_args))
1350 		return PTR_ERR(vol_args);
1351 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1352 
1353 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1354 					      vol_args->fd, subvol,
1355 					      NULL, false);
1356 
1357 	kfree(vol_args);
1358 	return ret;
1359 }
1360 
1361 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1362 					       void __user *arg, int subvol)
1363 {
1364 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1365 	int ret;
1366 	u64 transid = 0;
1367 	u64 *ptr = NULL;
1368 	bool readonly = false;
1369 
1370 	vol_args = memdup_user(arg, sizeof(*vol_args));
1371 	if (IS_ERR(vol_args))
1372 		return PTR_ERR(vol_args);
1373 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1374 
1375 	if (vol_args->flags &
1376 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1377 		ret = -EOPNOTSUPP;
1378 		goto out;
1379 	}
1380 
1381 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1382 		ptr = &transid;
1383 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1384 		readonly = true;
1385 
1386 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1387 					      vol_args->fd, subvol,
1388 					      ptr, readonly);
1389 
1390 	if (ret == 0 && ptr &&
1391 	    copy_to_user(arg +
1392 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1393 				  transid), ptr, sizeof(*ptr)))
1394 		ret = -EFAULT;
1395 out:
1396 	kfree(vol_args);
1397 	return ret;
1398 }
1399 
1400 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1401 						void __user *arg)
1402 {
1403 	struct inode *inode = fdentry(file)->d_inode;
1404 	struct btrfs_root *root = BTRFS_I(inode)->root;
1405 	int ret = 0;
1406 	u64 flags = 0;
1407 
1408 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1409 		return -EINVAL;
1410 
1411 	down_read(&root->fs_info->subvol_sem);
1412 	if (btrfs_root_readonly(root))
1413 		flags |= BTRFS_SUBVOL_RDONLY;
1414 	up_read(&root->fs_info->subvol_sem);
1415 
1416 	if (copy_to_user(arg, &flags, sizeof(flags)))
1417 		ret = -EFAULT;
1418 
1419 	return ret;
1420 }
1421 
1422 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1423 					      void __user *arg)
1424 {
1425 	struct inode *inode = fdentry(file)->d_inode;
1426 	struct btrfs_root *root = BTRFS_I(inode)->root;
1427 	struct btrfs_trans_handle *trans;
1428 	u64 root_flags;
1429 	u64 flags;
1430 	int ret = 0;
1431 
1432 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1433 		return -EROFS;
1434 
1435 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1436 		return -EINVAL;
1437 
1438 	if (copy_from_user(&flags, arg, sizeof(flags)))
1439 		return -EFAULT;
1440 
1441 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1442 		return -EINVAL;
1443 
1444 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1445 		return -EOPNOTSUPP;
1446 
1447 	if (!inode_owner_or_capable(inode))
1448 		return -EACCES;
1449 
1450 	down_write(&root->fs_info->subvol_sem);
1451 
1452 	/* nothing to do */
1453 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1454 		goto out;
1455 
1456 	root_flags = btrfs_root_flags(&root->root_item);
1457 	if (flags & BTRFS_SUBVOL_RDONLY)
1458 		btrfs_set_root_flags(&root->root_item,
1459 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1460 	else
1461 		btrfs_set_root_flags(&root->root_item,
1462 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1463 
1464 	trans = btrfs_start_transaction(root, 1);
1465 	if (IS_ERR(trans)) {
1466 		ret = PTR_ERR(trans);
1467 		goto out_reset;
1468 	}
1469 
1470 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1471 				&root->root_key, &root->root_item);
1472 
1473 	btrfs_commit_transaction(trans, root);
1474 out_reset:
1475 	if (ret)
1476 		btrfs_set_root_flags(&root->root_item, root_flags);
1477 out:
1478 	up_write(&root->fs_info->subvol_sem);
1479 	return ret;
1480 }
1481 
1482 /*
1483  * helper to check if the subvolume references other subvolumes
1484  */
1485 static noinline int may_destroy_subvol(struct btrfs_root *root)
1486 {
1487 	struct btrfs_path *path;
1488 	struct btrfs_key key;
1489 	int ret;
1490 
1491 	path = btrfs_alloc_path();
1492 	if (!path)
1493 		return -ENOMEM;
1494 
1495 	key.objectid = root->root_key.objectid;
1496 	key.type = BTRFS_ROOT_REF_KEY;
1497 	key.offset = (u64)-1;
1498 
1499 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1500 				&key, path, 0, 0);
1501 	if (ret < 0)
1502 		goto out;
1503 	BUG_ON(ret == 0);
1504 
1505 	ret = 0;
1506 	if (path->slots[0] > 0) {
1507 		path->slots[0]--;
1508 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1509 		if (key.objectid == root->root_key.objectid &&
1510 		    key.type == BTRFS_ROOT_REF_KEY)
1511 			ret = -ENOTEMPTY;
1512 	}
1513 out:
1514 	btrfs_free_path(path);
1515 	return ret;
1516 }
1517 
1518 static noinline int key_in_sk(struct btrfs_key *key,
1519 			      struct btrfs_ioctl_search_key *sk)
1520 {
1521 	struct btrfs_key test;
1522 	int ret;
1523 
1524 	test.objectid = sk->min_objectid;
1525 	test.type = sk->min_type;
1526 	test.offset = sk->min_offset;
1527 
1528 	ret = btrfs_comp_cpu_keys(key, &test);
1529 	if (ret < 0)
1530 		return 0;
1531 
1532 	test.objectid = sk->max_objectid;
1533 	test.type = sk->max_type;
1534 	test.offset = sk->max_offset;
1535 
1536 	ret = btrfs_comp_cpu_keys(key, &test);
1537 	if (ret > 0)
1538 		return 0;
1539 	return 1;
1540 }
1541 
1542 static noinline int copy_to_sk(struct btrfs_root *root,
1543 			       struct btrfs_path *path,
1544 			       struct btrfs_key *key,
1545 			       struct btrfs_ioctl_search_key *sk,
1546 			       char *buf,
1547 			       unsigned long *sk_offset,
1548 			       int *num_found)
1549 {
1550 	u64 found_transid;
1551 	struct extent_buffer *leaf;
1552 	struct btrfs_ioctl_search_header sh;
1553 	unsigned long item_off;
1554 	unsigned long item_len;
1555 	int nritems;
1556 	int i;
1557 	int slot;
1558 	int ret = 0;
1559 
1560 	leaf = path->nodes[0];
1561 	slot = path->slots[0];
1562 	nritems = btrfs_header_nritems(leaf);
1563 
1564 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1565 		i = nritems;
1566 		goto advance_key;
1567 	}
1568 	found_transid = btrfs_header_generation(leaf);
1569 
1570 	for (i = slot; i < nritems; i++) {
1571 		item_off = btrfs_item_ptr_offset(leaf, i);
1572 		item_len = btrfs_item_size_nr(leaf, i);
1573 
1574 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1575 			item_len = 0;
1576 
1577 		if (sizeof(sh) + item_len + *sk_offset >
1578 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1579 			ret = 1;
1580 			goto overflow;
1581 		}
1582 
1583 		btrfs_item_key_to_cpu(leaf, key, i);
1584 		if (!key_in_sk(key, sk))
1585 			continue;
1586 
1587 		sh.objectid = key->objectid;
1588 		sh.offset = key->offset;
1589 		sh.type = key->type;
1590 		sh.len = item_len;
1591 		sh.transid = found_transid;
1592 
1593 		/* copy search result header */
1594 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1595 		*sk_offset += sizeof(sh);
1596 
1597 		if (item_len) {
1598 			char *p = buf + *sk_offset;
1599 			/* copy the item */
1600 			read_extent_buffer(leaf, p,
1601 					   item_off, item_len);
1602 			*sk_offset += item_len;
1603 		}
1604 		(*num_found)++;
1605 
1606 		if (*num_found >= sk->nr_items)
1607 			break;
1608 	}
1609 advance_key:
1610 	ret = 0;
1611 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1612 		key->offset++;
1613 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1614 		key->offset = 0;
1615 		key->type++;
1616 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1617 		key->offset = 0;
1618 		key->type = 0;
1619 		key->objectid++;
1620 	} else
1621 		ret = 1;
1622 overflow:
1623 	return ret;
1624 }
1625 
1626 static noinline int search_ioctl(struct inode *inode,
1627 				 struct btrfs_ioctl_search_args *args)
1628 {
1629 	struct btrfs_root *root;
1630 	struct btrfs_key key;
1631 	struct btrfs_key max_key;
1632 	struct btrfs_path *path;
1633 	struct btrfs_ioctl_search_key *sk = &args->key;
1634 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1635 	int ret;
1636 	int num_found = 0;
1637 	unsigned long sk_offset = 0;
1638 
1639 	path = btrfs_alloc_path();
1640 	if (!path)
1641 		return -ENOMEM;
1642 
1643 	if (sk->tree_id == 0) {
1644 		/* search the root of the inode that was passed */
1645 		root = BTRFS_I(inode)->root;
1646 	} else {
1647 		key.objectid = sk->tree_id;
1648 		key.type = BTRFS_ROOT_ITEM_KEY;
1649 		key.offset = (u64)-1;
1650 		root = btrfs_read_fs_root_no_name(info, &key);
1651 		if (IS_ERR(root)) {
1652 			printk(KERN_ERR "could not find root %llu\n",
1653 			       sk->tree_id);
1654 			btrfs_free_path(path);
1655 			return -ENOENT;
1656 		}
1657 	}
1658 
1659 	key.objectid = sk->min_objectid;
1660 	key.type = sk->min_type;
1661 	key.offset = sk->min_offset;
1662 
1663 	max_key.objectid = sk->max_objectid;
1664 	max_key.type = sk->max_type;
1665 	max_key.offset = sk->max_offset;
1666 
1667 	path->keep_locks = 1;
1668 
1669 	while(1) {
1670 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1671 					   sk->min_transid);
1672 		if (ret != 0) {
1673 			if (ret > 0)
1674 				ret = 0;
1675 			goto err;
1676 		}
1677 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1678 				 &sk_offset, &num_found);
1679 		btrfs_release_path(path);
1680 		if (ret || num_found >= sk->nr_items)
1681 			break;
1682 
1683 	}
1684 	ret = 0;
1685 err:
1686 	sk->nr_items = num_found;
1687 	btrfs_free_path(path);
1688 	return ret;
1689 }
1690 
1691 static noinline int btrfs_ioctl_tree_search(struct file *file,
1692 					   void __user *argp)
1693 {
1694 	 struct btrfs_ioctl_search_args *args;
1695 	 struct inode *inode;
1696 	 int ret;
1697 
1698 	if (!capable(CAP_SYS_ADMIN))
1699 		return -EPERM;
1700 
1701 	args = memdup_user(argp, sizeof(*args));
1702 	if (IS_ERR(args))
1703 		return PTR_ERR(args);
1704 
1705 	inode = fdentry(file)->d_inode;
1706 	ret = search_ioctl(inode, args);
1707 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1708 		ret = -EFAULT;
1709 	kfree(args);
1710 	return ret;
1711 }
1712 
1713 /*
1714  * Search INODE_REFs to identify path name of 'dirid' directory
1715  * in a 'tree_id' tree. and sets path name to 'name'.
1716  */
1717 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1718 				u64 tree_id, u64 dirid, char *name)
1719 {
1720 	struct btrfs_root *root;
1721 	struct btrfs_key key;
1722 	char *ptr;
1723 	int ret = -1;
1724 	int slot;
1725 	int len;
1726 	int total_len = 0;
1727 	struct btrfs_inode_ref *iref;
1728 	struct extent_buffer *l;
1729 	struct btrfs_path *path;
1730 
1731 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1732 		name[0]='\0';
1733 		return 0;
1734 	}
1735 
1736 	path = btrfs_alloc_path();
1737 	if (!path)
1738 		return -ENOMEM;
1739 
1740 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1741 
1742 	key.objectid = tree_id;
1743 	key.type = BTRFS_ROOT_ITEM_KEY;
1744 	key.offset = (u64)-1;
1745 	root = btrfs_read_fs_root_no_name(info, &key);
1746 	if (IS_ERR(root)) {
1747 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1748 		ret = -ENOENT;
1749 		goto out;
1750 	}
1751 
1752 	key.objectid = dirid;
1753 	key.type = BTRFS_INODE_REF_KEY;
1754 	key.offset = (u64)-1;
1755 
1756 	while(1) {
1757 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1758 		if (ret < 0)
1759 			goto out;
1760 
1761 		l = path->nodes[0];
1762 		slot = path->slots[0];
1763 		if (ret > 0 && slot > 0)
1764 			slot--;
1765 		btrfs_item_key_to_cpu(l, &key, slot);
1766 
1767 		if (ret > 0 && (key.objectid != dirid ||
1768 				key.type != BTRFS_INODE_REF_KEY)) {
1769 			ret = -ENOENT;
1770 			goto out;
1771 		}
1772 
1773 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1774 		len = btrfs_inode_ref_name_len(l, iref);
1775 		ptr -= len + 1;
1776 		total_len += len + 1;
1777 		if (ptr < name)
1778 			goto out;
1779 
1780 		*(ptr + len) = '/';
1781 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1782 
1783 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1784 			break;
1785 
1786 		btrfs_release_path(path);
1787 		key.objectid = key.offset;
1788 		key.offset = (u64)-1;
1789 		dirid = key.objectid;
1790 	}
1791 	if (ptr < name)
1792 		goto out;
1793 	memmove(name, ptr, total_len);
1794 	name[total_len]='\0';
1795 	ret = 0;
1796 out:
1797 	btrfs_free_path(path);
1798 	return ret;
1799 }
1800 
1801 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1802 					   void __user *argp)
1803 {
1804 	 struct btrfs_ioctl_ino_lookup_args *args;
1805 	 struct inode *inode;
1806 	 int ret;
1807 
1808 	if (!capable(CAP_SYS_ADMIN))
1809 		return -EPERM;
1810 
1811 	args = memdup_user(argp, sizeof(*args));
1812 	if (IS_ERR(args))
1813 		return PTR_ERR(args);
1814 
1815 	inode = fdentry(file)->d_inode;
1816 
1817 	if (args->treeid == 0)
1818 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1819 
1820 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1821 					args->treeid, args->objectid,
1822 					args->name);
1823 
1824 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1825 		ret = -EFAULT;
1826 
1827 	kfree(args);
1828 	return ret;
1829 }
1830 
1831 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1832 					     void __user *arg)
1833 {
1834 	struct dentry *parent = fdentry(file);
1835 	struct dentry *dentry;
1836 	struct inode *dir = parent->d_inode;
1837 	struct inode *inode;
1838 	struct btrfs_root *root = BTRFS_I(dir)->root;
1839 	struct btrfs_root *dest = NULL;
1840 	struct btrfs_ioctl_vol_args *vol_args;
1841 	struct btrfs_trans_handle *trans;
1842 	int namelen;
1843 	int ret;
1844 	int err = 0;
1845 
1846 	vol_args = memdup_user(arg, sizeof(*vol_args));
1847 	if (IS_ERR(vol_args))
1848 		return PTR_ERR(vol_args);
1849 
1850 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1851 	namelen = strlen(vol_args->name);
1852 	if (strchr(vol_args->name, '/') ||
1853 	    strncmp(vol_args->name, "..", namelen) == 0) {
1854 		err = -EINVAL;
1855 		goto out;
1856 	}
1857 
1858 	err = mnt_want_write_file(file);
1859 	if (err)
1860 		goto out;
1861 
1862 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1863 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1864 	if (IS_ERR(dentry)) {
1865 		err = PTR_ERR(dentry);
1866 		goto out_unlock_dir;
1867 	}
1868 
1869 	if (!dentry->d_inode) {
1870 		err = -ENOENT;
1871 		goto out_dput;
1872 	}
1873 
1874 	inode = dentry->d_inode;
1875 	dest = BTRFS_I(inode)->root;
1876 	if (!capable(CAP_SYS_ADMIN)){
1877 		/*
1878 		 * Regular user.  Only allow this with a special mount
1879 		 * option, when the user has write+exec access to the
1880 		 * subvol root, and when rmdir(2) would have been
1881 		 * allowed.
1882 		 *
1883 		 * Note that this is _not_ check that the subvol is
1884 		 * empty or doesn't contain data that we wouldn't
1885 		 * otherwise be able to delete.
1886 		 *
1887 		 * Users who want to delete empty subvols should try
1888 		 * rmdir(2).
1889 		 */
1890 		err = -EPERM;
1891 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1892 			goto out_dput;
1893 
1894 		/*
1895 		 * Do not allow deletion if the parent dir is the same
1896 		 * as the dir to be deleted.  That means the ioctl
1897 		 * must be called on the dentry referencing the root
1898 		 * of the subvol, not a random directory contained
1899 		 * within it.
1900 		 */
1901 		err = -EINVAL;
1902 		if (root == dest)
1903 			goto out_dput;
1904 
1905 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1906 		if (err)
1907 			goto out_dput;
1908 
1909 		/* check if subvolume may be deleted by a non-root user */
1910 		err = btrfs_may_delete(dir, dentry, 1);
1911 		if (err)
1912 			goto out_dput;
1913 	}
1914 
1915 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1916 		err = -EINVAL;
1917 		goto out_dput;
1918 	}
1919 
1920 	mutex_lock(&inode->i_mutex);
1921 	err = d_invalidate(dentry);
1922 	if (err)
1923 		goto out_unlock;
1924 
1925 	down_write(&root->fs_info->subvol_sem);
1926 
1927 	err = may_destroy_subvol(dest);
1928 	if (err)
1929 		goto out_up_write;
1930 
1931 	trans = btrfs_start_transaction(root, 0);
1932 	if (IS_ERR(trans)) {
1933 		err = PTR_ERR(trans);
1934 		goto out_up_write;
1935 	}
1936 	trans->block_rsv = &root->fs_info->global_block_rsv;
1937 
1938 	ret = btrfs_unlink_subvol(trans, root, dir,
1939 				dest->root_key.objectid,
1940 				dentry->d_name.name,
1941 				dentry->d_name.len);
1942 	BUG_ON(ret);
1943 
1944 	btrfs_record_root_in_trans(trans, dest);
1945 
1946 	memset(&dest->root_item.drop_progress, 0,
1947 		sizeof(dest->root_item.drop_progress));
1948 	dest->root_item.drop_level = 0;
1949 	btrfs_set_root_refs(&dest->root_item, 0);
1950 
1951 	if (!xchg(&dest->orphan_item_inserted, 1)) {
1952 		ret = btrfs_insert_orphan_item(trans,
1953 					root->fs_info->tree_root,
1954 					dest->root_key.objectid);
1955 		BUG_ON(ret);
1956 	}
1957 
1958 	ret = btrfs_end_transaction(trans, root);
1959 	BUG_ON(ret);
1960 	inode->i_flags |= S_DEAD;
1961 out_up_write:
1962 	up_write(&root->fs_info->subvol_sem);
1963 out_unlock:
1964 	mutex_unlock(&inode->i_mutex);
1965 	if (!err) {
1966 		shrink_dcache_sb(root->fs_info->sb);
1967 		btrfs_invalidate_inodes(dest);
1968 		d_delete(dentry);
1969 	}
1970 out_dput:
1971 	dput(dentry);
1972 out_unlock_dir:
1973 	mutex_unlock(&dir->i_mutex);
1974 	mnt_drop_write_file(file);
1975 out:
1976 	kfree(vol_args);
1977 	return err;
1978 }
1979 
1980 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1981 {
1982 	struct inode *inode = fdentry(file)->d_inode;
1983 	struct btrfs_root *root = BTRFS_I(inode)->root;
1984 	struct btrfs_ioctl_defrag_range_args *range;
1985 	int ret;
1986 
1987 	if (btrfs_root_readonly(root))
1988 		return -EROFS;
1989 
1990 	ret = mnt_want_write_file(file);
1991 	if (ret)
1992 		return ret;
1993 
1994 	switch (inode->i_mode & S_IFMT) {
1995 	case S_IFDIR:
1996 		if (!capable(CAP_SYS_ADMIN)) {
1997 			ret = -EPERM;
1998 			goto out;
1999 		}
2000 		ret = btrfs_defrag_root(root, 0);
2001 		if (ret)
2002 			goto out;
2003 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2004 		break;
2005 	case S_IFREG:
2006 		if (!(file->f_mode & FMODE_WRITE)) {
2007 			ret = -EINVAL;
2008 			goto out;
2009 		}
2010 
2011 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2012 		if (!range) {
2013 			ret = -ENOMEM;
2014 			goto out;
2015 		}
2016 
2017 		if (argp) {
2018 			if (copy_from_user(range, argp,
2019 					   sizeof(*range))) {
2020 				ret = -EFAULT;
2021 				kfree(range);
2022 				goto out;
2023 			}
2024 			/* compression requires us to start the IO */
2025 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2026 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2027 				range->extent_thresh = (u32)-1;
2028 			}
2029 		} else {
2030 			/* the rest are all set to zero by kzalloc */
2031 			range->len = (u64)-1;
2032 		}
2033 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2034 					range, 0, 0);
2035 		if (ret > 0)
2036 			ret = 0;
2037 		kfree(range);
2038 		break;
2039 	default:
2040 		ret = -EINVAL;
2041 	}
2042 out:
2043 	mnt_drop_write_file(file);
2044 	return ret;
2045 }
2046 
2047 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2048 {
2049 	struct btrfs_ioctl_vol_args *vol_args;
2050 	int ret;
2051 
2052 	if (!capable(CAP_SYS_ADMIN))
2053 		return -EPERM;
2054 
2055 	vol_args = memdup_user(arg, sizeof(*vol_args));
2056 	if (IS_ERR(vol_args))
2057 		return PTR_ERR(vol_args);
2058 
2059 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2060 	ret = btrfs_init_new_device(root, vol_args->name);
2061 
2062 	kfree(vol_args);
2063 	return ret;
2064 }
2065 
2066 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2067 {
2068 	struct btrfs_ioctl_vol_args *vol_args;
2069 	int ret;
2070 
2071 	if (!capable(CAP_SYS_ADMIN))
2072 		return -EPERM;
2073 
2074 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2075 		return -EROFS;
2076 
2077 	vol_args = memdup_user(arg, sizeof(*vol_args));
2078 	if (IS_ERR(vol_args))
2079 		return PTR_ERR(vol_args);
2080 
2081 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2082 	ret = btrfs_rm_device(root, vol_args->name);
2083 
2084 	kfree(vol_args);
2085 	return ret;
2086 }
2087 
2088 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2089 {
2090 	struct btrfs_ioctl_fs_info_args *fi_args;
2091 	struct btrfs_device *device;
2092 	struct btrfs_device *next;
2093 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2094 	int ret = 0;
2095 
2096 	if (!capable(CAP_SYS_ADMIN))
2097 		return -EPERM;
2098 
2099 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2100 	if (!fi_args)
2101 		return -ENOMEM;
2102 
2103 	fi_args->num_devices = fs_devices->num_devices;
2104 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2105 
2106 	mutex_lock(&fs_devices->device_list_mutex);
2107 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2108 		if (device->devid > fi_args->max_id)
2109 			fi_args->max_id = device->devid;
2110 	}
2111 	mutex_unlock(&fs_devices->device_list_mutex);
2112 
2113 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2114 		ret = -EFAULT;
2115 
2116 	kfree(fi_args);
2117 	return ret;
2118 }
2119 
2120 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2121 {
2122 	struct btrfs_ioctl_dev_info_args *di_args;
2123 	struct btrfs_device *dev;
2124 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2125 	int ret = 0;
2126 	char *s_uuid = NULL;
2127 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2128 
2129 	if (!capable(CAP_SYS_ADMIN))
2130 		return -EPERM;
2131 
2132 	di_args = memdup_user(arg, sizeof(*di_args));
2133 	if (IS_ERR(di_args))
2134 		return PTR_ERR(di_args);
2135 
2136 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2137 		s_uuid = di_args->uuid;
2138 
2139 	mutex_lock(&fs_devices->device_list_mutex);
2140 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2141 	mutex_unlock(&fs_devices->device_list_mutex);
2142 
2143 	if (!dev) {
2144 		ret = -ENODEV;
2145 		goto out;
2146 	}
2147 
2148 	di_args->devid = dev->devid;
2149 	di_args->bytes_used = dev->bytes_used;
2150 	di_args->total_bytes = dev->total_bytes;
2151 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2152 	strncpy(di_args->path, dev->name, sizeof(di_args->path));
2153 
2154 out:
2155 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2156 		ret = -EFAULT;
2157 
2158 	kfree(di_args);
2159 	return ret;
2160 }
2161 
2162 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2163 				       u64 off, u64 olen, u64 destoff)
2164 {
2165 	struct inode *inode = fdentry(file)->d_inode;
2166 	struct btrfs_root *root = BTRFS_I(inode)->root;
2167 	struct file *src_file;
2168 	struct inode *src;
2169 	struct btrfs_trans_handle *trans;
2170 	struct btrfs_path *path;
2171 	struct extent_buffer *leaf;
2172 	char *buf;
2173 	struct btrfs_key key;
2174 	u32 nritems;
2175 	int slot;
2176 	int ret;
2177 	u64 len = olen;
2178 	u64 bs = root->fs_info->sb->s_blocksize;
2179 	u64 hint_byte;
2180 
2181 	/*
2182 	 * TODO:
2183 	 * - split compressed inline extents.  annoying: we need to
2184 	 *   decompress into destination's address_space (the file offset
2185 	 *   may change, so source mapping won't do), then recompress (or
2186 	 *   otherwise reinsert) a subrange.
2187 	 * - allow ranges within the same file to be cloned (provided
2188 	 *   they don't overlap)?
2189 	 */
2190 
2191 	/* the destination must be opened for writing */
2192 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2193 		return -EINVAL;
2194 
2195 	if (btrfs_root_readonly(root))
2196 		return -EROFS;
2197 
2198 	ret = mnt_want_write_file(file);
2199 	if (ret)
2200 		return ret;
2201 
2202 	src_file = fget(srcfd);
2203 	if (!src_file) {
2204 		ret = -EBADF;
2205 		goto out_drop_write;
2206 	}
2207 
2208 	src = src_file->f_dentry->d_inode;
2209 
2210 	ret = -EINVAL;
2211 	if (src == inode)
2212 		goto out_fput;
2213 
2214 	/* the src must be open for reading */
2215 	if (!(src_file->f_mode & FMODE_READ))
2216 		goto out_fput;
2217 
2218 	/* don't make the dst file partly checksummed */
2219 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2220 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2221 		goto out_fput;
2222 
2223 	ret = -EISDIR;
2224 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2225 		goto out_fput;
2226 
2227 	ret = -EXDEV;
2228 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2229 		goto out_fput;
2230 
2231 	ret = -ENOMEM;
2232 	buf = vmalloc(btrfs_level_size(root, 0));
2233 	if (!buf)
2234 		goto out_fput;
2235 
2236 	path = btrfs_alloc_path();
2237 	if (!path) {
2238 		vfree(buf);
2239 		goto out_fput;
2240 	}
2241 	path->reada = 2;
2242 
2243 	if (inode < src) {
2244 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2245 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2246 	} else {
2247 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2248 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2249 	}
2250 
2251 	/* determine range to clone */
2252 	ret = -EINVAL;
2253 	if (off + len > src->i_size || off + len < off)
2254 		goto out_unlock;
2255 	if (len == 0)
2256 		olen = len = src->i_size - off;
2257 	/* if we extend to eof, continue to block boundary */
2258 	if (off + len == src->i_size)
2259 		len = ALIGN(src->i_size, bs) - off;
2260 
2261 	/* verify the end result is block aligned */
2262 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2263 	    !IS_ALIGNED(destoff, bs))
2264 		goto out_unlock;
2265 
2266 	if (destoff > inode->i_size) {
2267 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2268 		if (ret)
2269 			goto out_unlock;
2270 	}
2271 
2272 	/* truncate page cache pages from target inode range */
2273 	truncate_inode_pages_range(&inode->i_data, destoff,
2274 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2275 
2276 	/* do any pending delalloc/csum calc on src, one way or
2277 	   another, and lock file content */
2278 	while (1) {
2279 		struct btrfs_ordered_extent *ordered;
2280 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2281 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2282 		if (!ordered &&
2283 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2284 				   EXTENT_DELALLOC, 0, NULL))
2285 			break;
2286 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2287 		if (ordered)
2288 			btrfs_put_ordered_extent(ordered);
2289 		btrfs_wait_ordered_range(src, off, len);
2290 	}
2291 
2292 	/* clone data */
2293 	key.objectid = btrfs_ino(src);
2294 	key.type = BTRFS_EXTENT_DATA_KEY;
2295 	key.offset = 0;
2296 
2297 	while (1) {
2298 		/*
2299 		 * note the key will change type as we walk through the
2300 		 * tree.
2301 		 */
2302 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2303 		if (ret < 0)
2304 			goto out;
2305 
2306 		nritems = btrfs_header_nritems(path->nodes[0]);
2307 		if (path->slots[0] >= nritems) {
2308 			ret = btrfs_next_leaf(root, path);
2309 			if (ret < 0)
2310 				goto out;
2311 			if (ret > 0)
2312 				break;
2313 			nritems = btrfs_header_nritems(path->nodes[0]);
2314 		}
2315 		leaf = path->nodes[0];
2316 		slot = path->slots[0];
2317 
2318 		btrfs_item_key_to_cpu(leaf, &key, slot);
2319 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2320 		    key.objectid != btrfs_ino(src))
2321 			break;
2322 
2323 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2324 			struct btrfs_file_extent_item *extent;
2325 			int type;
2326 			u32 size;
2327 			struct btrfs_key new_key;
2328 			u64 disko = 0, diskl = 0;
2329 			u64 datao = 0, datal = 0;
2330 			u8 comp;
2331 			u64 endoff;
2332 
2333 			size = btrfs_item_size_nr(leaf, slot);
2334 			read_extent_buffer(leaf, buf,
2335 					   btrfs_item_ptr_offset(leaf, slot),
2336 					   size);
2337 
2338 			extent = btrfs_item_ptr(leaf, slot,
2339 						struct btrfs_file_extent_item);
2340 			comp = btrfs_file_extent_compression(leaf, extent);
2341 			type = btrfs_file_extent_type(leaf, extent);
2342 			if (type == BTRFS_FILE_EXTENT_REG ||
2343 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2344 				disko = btrfs_file_extent_disk_bytenr(leaf,
2345 								      extent);
2346 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2347 								 extent);
2348 				datao = btrfs_file_extent_offset(leaf, extent);
2349 				datal = btrfs_file_extent_num_bytes(leaf,
2350 								    extent);
2351 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2352 				/* take upper bound, may be compressed */
2353 				datal = btrfs_file_extent_ram_bytes(leaf,
2354 								    extent);
2355 			}
2356 			btrfs_release_path(path);
2357 
2358 			if (key.offset + datal <= off ||
2359 			    key.offset >= off+len)
2360 				goto next;
2361 
2362 			memcpy(&new_key, &key, sizeof(new_key));
2363 			new_key.objectid = btrfs_ino(inode);
2364 			if (off <= key.offset)
2365 				new_key.offset = key.offset + destoff - off;
2366 			else
2367 				new_key.offset = destoff;
2368 
2369 			/*
2370 			 * 1 - adjusting old extent (we may have to split it)
2371 			 * 1 - add new extent
2372 			 * 1 - inode update
2373 			 */
2374 			trans = btrfs_start_transaction(root, 3);
2375 			if (IS_ERR(trans)) {
2376 				ret = PTR_ERR(trans);
2377 				goto out;
2378 			}
2379 
2380 			if (type == BTRFS_FILE_EXTENT_REG ||
2381 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2382 				/*
2383 				 *    a  | --- range to clone ---|  b
2384 				 * | ------------- extent ------------- |
2385 				 */
2386 
2387 				/* substract range b */
2388 				if (key.offset + datal > off + len)
2389 					datal = off + len - key.offset;
2390 
2391 				/* substract range a */
2392 				if (off > key.offset) {
2393 					datao += off - key.offset;
2394 					datal -= off - key.offset;
2395 				}
2396 
2397 				ret = btrfs_drop_extents(trans, inode,
2398 							 new_key.offset,
2399 							 new_key.offset + datal,
2400 							 &hint_byte, 1);
2401 				BUG_ON(ret);
2402 
2403 				ret = btrfs_insert_empty_item(trans, root, path,
2404 							      &new_key, size);
2405 				BUG_ON(ret);
2406 
2407 				leaf = path->nodes[0];
2408 				slot = path->slots[0];
2409 				write_extent_buffer(leaf, buf,
2410 					    btrfs_item_ptr_offset(leaf, slot),
2411 					    size);
2412 
2413 				extent = btrfs_item_ptr(leaf, slot,
2414 						struct btrfs_file_extent_item);
2415 
2416 				/* disko == 0 means it's a hole */
2417 				if (!disko)
2418 					datao = 0;
2419 
2420 				btrfs_set_file_extent_offset(leaf, extent,
2421 							     datao);
2422 				btrfs_set_file_extent_num_bytes(leaf, extent,
2423 								datal);
2424 				if (disko) {
2425 					inode_add_bytes(inode, datal);
2426 					ret = btrfs_inc_extent_ref(trans, root,
2427 							disko, diskl, 0,
2428 							root->root_key.objectid,
2429 							btrfs_ino(inode),
2430 							new_key.offset - datao);
2431 					BUG_ON(ret);
2432 				}
2433 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2434 				u64 skip = 0;
2435 				u64 trim = 0;
2436 				if (off > key.offset) {
2437 					skip = off - key.offset;
2438 					new_key.offset += skip;
2439 				}
2440 
2441 				if (key.offset + datal > off+len)
2442 					trim = key.offset + datal - (off+len);
2443 
2444 				if (comp && (skip || trim)) {
2445 					ret = -EINVAL;
2446 					btrfs_end_transaction(trans, root);
2447 					goto out;
2448 				}
2449 				size -= skip + trim;
2450 				datal -= skip + trim;
2451 
2452 				ret = btrfs_drop_extents(trans, inode,
2453 							 new_key.offset,
2454 							 new_key.offset + datal,
2455 							 &hint_byte, 1);
2456 				BUG_ON(ret);
2457 
2458 				ret = btrfs_insert_empty_item(trans, root, path,
2459 							      &new_key, size);
2460 				BUG_ON(ret);
2461 
2462 				if (skip) {
2463 					u32 start =
2464 					  btrfs_file_extent_calc_inline_size(0);
2465 					memmove(buf+start, buf+start+skip,
2466 						datal);
2467 				}
2468 
2469 				leaf = path->nodes[0];
2470 				slot = path->slots[0];
2471 				write_extent_buffer(leaf, buf,
2472 					    btrfs_item_ptr_offset(leaf, slot),
2473 					    size);
2474 				inode_add_bytes(inode, datal);
2475 			}
2476 
2477 			btrfs_mark_buffer_dirty(leaf);
2478 			btrfs_release_path(path);
2479 
2480 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2481 
2482 			/*
2483 			 * we round up to the block size at eof when
2484 			 * determining which extents to clone above,
2485 			 * but shouldn't round up the file size
2486 			 */
2487 			endoff = new_key.offset + datal;
2488 			if (endoff > destoff+olen)
2489 				endoff = destoff+olen;
2490 			if (endoff > inode->i_size)
2491 				btrfs_i_size_write(inode, endoff);
2492 
2493 			ret = btrfs_update_inode(trans, root, inode);
2494 			BUG_ON(ret);
2495 			btrfs_end_transaction(trans, root);
2496 		}
2497 next:
2498 		btrfs_release_path(path);
2499 		key.offset++;
2500 	}
2501 	ret = 0;
2502 out:
2503 	btrfs_release_path(path);
2504 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2505 out_unlock:
2506 	mutex_unlock(&src->i_mutex);
2507 	mutex_unlock(&inode->i_mutex);
2508 	vfree(buf);
2509 	btrfs_free_path(path);
2510 out_fput:
2511 	fput(src_file);
2512 out_drop_write:
2513 	mnt_drop_write_file(file);
2514 	return ret;
2515 }
2516 
2517 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2518 {
2519 	struct btrfs_ioctl_clone_range_args args;
2520 
2521 	if (copy_from_user(&args, argp, sizeof(args)))
2522 		return -EFAULT;
2523 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2524 				 args.src_length, args.dest_offset);
2525 }
2526 
2527 /*
2528  * there are many ways the trans_start and trans_end ioctls can lead
2529  * to deadlocks.  They should only be used by applications that
2530  * basically own the machine, and have a very in depth understanding
2531  * of all the possible deadlocks and enospc problems.
2532  */
2533 static long btrfs_ioctl_trans_start(struct file *file)
2534 {
2535 	struct inode *inode = fdentry(file)->d_inode;
2536 	struct btrfs_root *root = BTRFS_I(inode)->root;
2537 	struct btrfs_trans_handle *trans;
2538 	int ret;
2539 
2540 	ret = -EPERM;
2541 	if (!capable(CAP_SYS_ADMIN))
2542 		goto out;
2543 
2544 	ret = -EINPROGRESS;
2545 	if (file->private_data)
2546 		goto out;
2547 
2548 	ret = -EROFS;
2549 	if (btrfs_root_readonly(root))
2550 		goto out;
2551 
2552 	ret = mnt_want_write_file(file);
2553 	if (ret)
2554 		goto out;
2555 
2556 	atomic_inc(&root->fs_info->open_ioctl_trans);
2557 
2558 	ret = -ENOMEM;
2559 	trans = btrfs_start_ioctl_transaction(root);
2560 	if (IS_ERR(trans))
2561 		goto out_drop;
2562 
2563 	file->private_data = trans;
2564 	return 0;
2565 
2566 out_drop:
2567 	atomic_dec(&root->fs_info->open_ioctl_trans);
2568 	mnt_drop_write_file(file);
2569 out:
2570 	return ret;
2571 }
2572 
2573 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2574 {
2575 	struct inode *inode = fdentry(file)->d_inode;
2576 	struct btrfs_root *root = BTRFS_I(inode)->root;
2577 	struct btrfs_root *new_root;
2578 	struct btrfs_dir_item *di;
2579 	struct btrfs_trans_handle *trans;
2580 	struct btrfs_path *path;
2581 	struct btrfs_key location;
2582 	struct btrfs_disk_key disk_key;
2583 	struct btrfs_super_block *disk_super;
2584 	u64 features;
2585 	u64 objectid = 0;
2586 	u64 dir_id;
2587 
2588 	if (!capable(CAP_SYS_ADMIN))
2589 		return -EPERM;
2590 
2591 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2592 		return -EFAULT;
2593 
2594 	if (!objectid)
2595 		objectid = root->root_key.objectid;
2596 
2597 	location.objectid = objectid;
2598 	location.type = BTRFS_ROOT_ITEM_KEY;
2599 	location.offset = (u64)-1;
2600 
2601 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2602 	if (IS_ERR(new_root))
2603 		return PTR_ERR(new_root);
2604 
2605 	if (btrfs_root_refs(&new_root->root_item) == 0)
2606 		return -ENOENT;
2607 
2608 	path = btrfs_alloc_path();
2609 	if (!path)
2610 		return -ENOMEM;
2611 	path->leave_spinning = 1;
2612 
2613 	trans = btrfs_start_transaction(root, 1);
2614 	if (IS_ERR(trans)) {
2615 		btrfs_free_path(path);
2616 		return PTR_ERR(trans);
2617 	}
2618 
2619 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2620 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2621 				   dir_id, "default", 7, 1);
2622 	if (IS_ERR_OR_NULL(di)) {
2623 		btrfs_free_path(path);
2624 		btrfs_end_transaction(trans, root);
2625 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2626 		       "this isn't going to work\n");
2627 		return -ENOENT;
2628 	}
2629 
2630 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2631 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2632 	btrfs_mark_buffer_dirty(path->nodes[0]);
2633 	btrfs_free_path(path);
2634 
2635 	disk_super = root->fs_info->super_copy;
2636 	features = btrfs_super_incompat_flags(disk_super);
2637 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2638 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2639 		btrfs_set_super_incompat_flags(disk_super, features);
2640 	}
2641 	btrfs_end_transaction(trans, root);
2642 
2643 	return 0;
2644 }
2645 
2646 static void get_block_group_info(struct list_head *groups_list,
2647 				 struct btrfs_ioctl_space_info *space)
2648 {
2649 	struct btrfs_block_group_cache *block_group;
2650 
2651 	space->total_bytes = 0;
2652 	space->used_bytes = 0;
2653 	space->flags = 0;
2654 	list_for_each_entry(block_group, groups_list, list) {
2655 		space->flags = block_group->flags;
2656 		space->total_bytes += block_group->key.offset;
2657 		space->used_bytes +=
2658 			btrfs_block_group_used(&block_group->item);
2659 	}
2660 }
2661 
2662 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2663 {
2664 	struct btrfs_ioctl_space_args space_args;
2665 	struct btrfs_ioctl_space_info space;
2666 	struct btrfs_ioctl_space_info *dest;
2667 	struct btrfs_ioctl_space_info *dest_orig;
2668 	struct btrfs_ioctl_space_info __user *user_dest;
2669 	struct btrfs_space_info *info;
2670 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2671 		       BTRFS_BLOCK_GROUP_SYSTEM,
2672 		       BTRFS_BLOCK_GROUP_METADATA,
2673 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2674 	int num_types = 4;
2675 	int alloc_size;
2676 	int ret = 0;
2677 	u64 slot_count = 0;
2678 	int i, c;
2679 
2680 	if (copy_from_user(&space_args,
2681 			   (struct btrfs_ioctl_space_args __user *)arg,
2682 			   sizeof(space_args)))
2683 		return -EFAULT;
2684 
2685 	for (i = 0; i < num_types; i++) {
2686 		struct btrfs_space_info *tmp;
2687 
2688 		info = NULL;
2689 		rcu_read_lock();
2690 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2691 					list) {
2692 			if (tmp->flags == types[i]) {
2693 				info = tmp;
2694 				break;
2695 			}
2696 		}
2697 		rcu_read_unlock();
2698 
2699 		if (!info)
2700 			continue;
2701 
2702 		down_read(&info->groups_sem);
2703 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2704 			if (!list_empty(&info->block_groups[c]))
2705 				slot_count++;
2706 		}
2707 		up_read(&info->groups_sem);
2708 	}
2709 
2710 	/* space_slots == 0 means they are asking for a count */
2711 	if (space_args.space_slots == 0) {
2712 		space_args.total_spaces = slot_count;
2713 		goto out;
2714 	}
2715 
2716 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2717 
2718 	alloc_size = sizeof(*dest) * slot_count;
2719 
2720 	/* we generally have at most 6 or so space infos, one for each raid
2721 	 * level.  So, a whole page should be more than enough for everyone
2722 	 */
2723 	if (alloc_size > PAGE_CACHE_SIZE)
2724 		return -ENOMEM;
2725 
2726 	space_args.total_spaces = 0;
2727 	dest = kmalloc(alloc_size, GFP_NOFS);
2728 	if (!dest)
2729 		return -ENOMEM;
2730 	dest_orig = dest;
2731 
2732 	/* now we have a buffer to copy into */
2733 	for (i = 0; i < num_types; i++) {
2734 		struct btrfs_space_info *tmp;
2735 
2736 		if (!slot_count)
2737 			break;
2738 
2739 		info = NULL;
2740 		rcu_read_lock();
2741 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2742 					list) {
2743 			if (tmp->flags == types[i]) {
2744 				info = tmp;
2745 				break;
2746 			}
2747 		}
2748 		rcu_read_unlock();
2749 
2750 		if (!info)
2751 			continue;
2752 		down_read(&info->groups_sem);
2753 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2754 			if (!list_empty(&info->block_groups[c])) {
2755 				get_block_group_info(&info->block_groups[c],
2756 						     &space);
2757 				memcpy(dest, &space, sizeof(space));
2758 				dest++;
2759 				space_args.total_spaces++;
2760 				slot_count--;
2761 			}
2762 			if (!slot_count)
2763 				break;
2764 		}
2765 		up_read(&info->groups_sem);
2766 	}
2767 
2768 	user_dest = (struct btrfs_ioctl_space_info *)
2769 		(arg + sizeof(struct btrfs_ioctl_space_args));
2770 
2771 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2772 		ret = -EFAULT;
2773 
2774 	kfree(dest_orig);
2775 out:
2776 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2777 		ret = -EFAULT;
2778 
2779 	return ret;
2780 }
2781 
2782 /*
2783  * there are many ways the trans_start and trans_end ioctls can lead
2784  * to deadlocks.  They should only be used by applications that
2785  * basically own the machine, and have a very in depth understanding
2786  * of all the possible deadlocks and enospc problems.
2787  */
2788 long btrfs_ioctl_trans_end(struct file *file)
2789 {
2790 	struct inode *inode = fdentry(file)->d_inode;
2791 	struct btrfs_root *root = BTRFS_I(inode)->root;
2792 	struct btrfs_trans_handle *trans;
2793 
2794 	trans = file->private_data;
2795 	if (!trans)
2796 		return -EINVAL;
2797 	file->private_data = NULL;
2798 
2799 	btrfs_end_transaction(trans, root);
2800 
2801 	atomic_dec(&root->fs_info->open_ioctl_trans);
2802 
2803 	mnt_drop_write_file(file);
2804 	return 0;
2805 }
2806 
2807 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2808 {
2809 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2810 	struct btrfs_trans_handle *trans;
2811 	u64 transid;
2812 	int ret;
2813 
2814 	trans = btrfs_start_transaction(root, 0);
2815 	if (IS_ERR(trans))
2816 		return PTR_ERR(trans);
2817 	transid = trans->transid;
2818 	ret = btrfs_commit_transaction_async(trans, root, 0);
2819 	if (ret) {
2820 		btrfs_end_transaction(trans, root);
2821 		return ret;
2822 	}
2823 
2824 	if (argp)
2825 		if (copy_to_user(argp, &transid, sizeof(transid)))
2826 			return -EFAULT;
2827 	return 0;
2828 }
2829 
2830 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2831 {
2832 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2833 	u64 transid;
2834 
2835 	if (argp) {
2836 		if (copy_from_user(&transid, argp, sizeof(transid)))
2837 			return -EFAULT;
2838 	} else {
2839 		transid = 0;  /* current trans */
2840 	}
2841 	return btrfs_wait_for_commit(root, transid);
2842 }
2843 
2844 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2845 {
2846 	int ret;
2847 	struct btrfs_ioctl_scrub_args *sa;
2848 
2849 	if (!capable(CAP_SYS_ADMIN))
2850 		return -EPERM;
2851 
2852 	sa = memdup_user(arg, sizeof(*sa));
2853 	if (IS_ERR(sa))
2854 		return PTR_ERR(sa);
2855 
2856 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2857 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2858 
2859 	if (copy_to_user(arg, sa, sizeof(*sa)))
2860 		ret = -EFAULT;
2861 
2862 	kfree(sa);
2863 	return ret;
2864 }
2865 
2866 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2867 {
2868 	if (!capable(CAP_SYS_ADMIN))
2869 		return -EPERM;
2870 
2871 	return btrfs_scrub_cancel(root);
2872 }
2873 
2874 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2875 				       void __user *arg)
2876 {
2877 	struct btrfs_ioctl_scrub_args *sa;
2878 	int ret;
2879 
2880 	if (!capable(CAP_SYS_ADMIN))
2881 		return -EPERM;
2882 
2883 	sa = memdup_user(arg, sizeof(*sa));
2884 	if (IS_ERR(sa))
2885 		return PTR_ERR(sa);
2886 
2887 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2888 
2889 	if (copy_to_user(arg, sa, sizeof(*sa)))
2890 		ret = -EFAULT;
2891 
2892 	kfree(sa);
2893 	return ret;
2894 }
2895 
2896 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
2897 {
2898 	int ret = 0;
2899 	int i;
2900 	u64 rel_ptr;
2901 	int size;
2902 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
2903 	struct inode_fs_paths *ipath = NULL;
2904 	struct btrfs_path *path;
2905 
2906 	if (!capable(CAP_SYS_ADMIN))
2907 		return -EPERM;
2908 
2909 	path = btrfs_alloc_path();
2910 	if (!path) {
2911 		ret = -ENOMEM;
2912 		goto out;
2913 	}
2914 
2915 	ipa = memdup_user(arg, sizeof(*ipa));
2916 	if (IS_ERR(ipa)) {
2917 		ret = PTR_ERR(ipa);
2918 		ipa = NULL;
2919 		goto out;
2920 	}
2921 
2922 	size = min_t(u32, ipa->size, 4096);
2923 	ipath = init_ipath(size, root, path);
2924 	if (IS_ERR(ipath)) {
2925 		ret = PTR_ERR(ipath);
2926 		ipath = NULL;
2927 		goto out;
2928 	}
2929 
2930 	ret = paths_from_inode(ipa->inum, ipath);
2931 	if (ret < 0)
2932 		goto out;
2933 
2934 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
2935 		rel_ptr = ipath->fspath->val[i] -
2936 			  (u64)(unsigned long)ipath->fspath->val;
2937 		ipath->fspath->val[i] = rel_ptr;
2938 	}
2939 
2940 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
2941 			   (void *)(unsigned long)ipath->fspath, size);
2942 	if (ret) {
2943 		ret = -EFAULT;
2944 		goto out;
2945 	}
2946 
2947 out:
2948 	btrfs_free_path(path);
2949 	free_ipath(ipath);
2950 	kfree(ipa);
2951 
2952 	return ret;
2953 }
2954 
2955 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
2956 {
2957 	struct btrfs_data_container *inodes = ctx;
2958 	const size_t c = 3 * sizeof(u64);
2959 
2960 	if (inodes->bytes_left >= c) {
2961 		inodes->bytes_left -= c;
2962 		inodes->val[inodes->elem_cnt] = inum;
2963 		inodes->val[inodes->elem_cnt + 1] = offset;
2964 		inodes->val[inodes->elem_cnt + 2] = root;
2965 		inodes->elem_cnt += 3;
2966 	} else {
2967 		inodes->bytes_missing += c - inodes->bytes_left;
2968 		inodes->bytes_left = 0;
2969 		inodes->elem_missed += 3;
2970 	}
2971 
2972 	return 0;
2973 }
2974 
2975 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
2976 					void __user *arg)
2977 {
2978 	int ret = 0;
2979 	int size;
2980 	u64 extent_offset;
2981 	struct btrfs_ioctl_logical_ino_args *loi;
2982 	struct btrfs_data_container *inodes = NULL;
2983 	struct btrfs_path *path = NULL;
2984 	struct btrfs_key key;
2985 
2986 	if (!capable(CAP_SYS_ADMIN))
2987 		return -EPERM;
2988 
2989 	loi = memdup_user(arg, sizeof(*loi));
2990 	if (IS_ERR(loi)) {
2991 		ret = PTR_ERR(loi);
2992 		loi = NULL;
2993 		goto out;
2994 	}
2995 
2996 	path = btrfs_alloc_path();
2997 	if (!path) {
2998 		ret = -ENOMEM;
2999 		goto out;
3000 	}
3001 
3002 	size = min_t(u32, loi->size, 4096);
3003 	inodes = init_data_container(size);
3004 	if (IS_ERR(inodes)) {
3005 		ret = PTR_ERR(inodes);
3006 		inodes = NULL;
3007 		goto out;
3008 	}
3009 
3010 	ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3011 
3012 	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3013 		ret = -ENOENT;
3014 	if (ret < 0)
3015 		goto out;
3016 
3017 	extent_offset = loi->logical - key.objectid;
3018 	ret = iterate_extent_inodes(root->fs_info, path, key.objectid,
3019 					extent_offset, build_ino_list, inodes);
3020 
3021 	if (ret < 0)
3022 		goto out;
3023 
3024 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3025 			   (void *)(unsigned long)inodes, size);
3026 	if (ret)
3027 		ret = -EFAULT;
3028 
3029 out:
3030 	btrfs_free_path(path);
3031 	kfree(inodes);
3032 	kfree(loi);
3033 
3034 	return ret;
3035 }
3036 
3037 long btrfs_ioctl(struct file *file, unsigned int
3038 		cmd, unsigned long arg)
3039 {
3040 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3041 	void __user *argp = (void __user *)arg;
3042 
3043 	switch (cmd) {
3044 	case FS_IOC_GETFLAGS:
3045 		return btrfs_ioctl_getflags(file, argp);
3046 	case FS_IOC_SETFLAGS:
3047 		return btrfs_ioctl_setflags(file, argp);
3048 	case FS_IOC_GETVERSION:
3049 		return btrfs_ioctl_getversion(file, argp);
3050 	case FITRIM:
3051 		return btrfs_ioctl_fitrim(file, argp);
3052 	case BTRFS_IOC_SNAP_CREATE:
3053 		return btrfs_ioctl_snap_create(file, argp, 0);
3054 	case BTRFS_IOC_SNAP_CREATE_V2:
3055 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3056 	case BTRFS_IOC_SUBVOL_CREATE:
3057 		return btrfs_ioctl_snap_create(file, argp, 1);
3058 	case BTRFS_IOC_SNAP_DESTROY:
3059 		return btrfs_ioctl_snap_destroy(file, argp);
3060 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3061 		return btrfs_ioctl_subvol_getflags(file, argp);
3062 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3063 		return btrfs_ioctl_subvol_setflags(file, argp);
3064 	case BTRFS_IOC_DEFAULT_SUBVOL:
3065 		return btrfs_ioctl_default_subvol(file, argp);
3066 	case BTRFS_IOC_DEFRAG:
3067 		return btrfs_ioctl_defrag(file, NULL);
3068 	case BTRFS_IOC_DEFRAG_RANGE:
3069 		return btrfs_ioctl_defrag(file, argp);
3070 	case BTRFS_IOC_RESIZE:
3071 		return btrfs_ioctl_resize(root, argp);
3072 	case BTRFS_IOC_ADD_DEV:
3073 		return btrfs_ioctl_add_dev(root, argp);
3074 	case BTRFS_IOC_RM_DEV:
3075 		return btrfs_ioctl_rm_dev(root, argp);
3076 	case BTRFS_IOC_FS_INFO:
3077 		return btrfs_ioctl_fs_info(root, argp);
3078 	case BTRFS_IOC_DEV_INFO:
3079 		return btrfs_ioctl_dev_info(root, argp);
3080 	case BTRFS_IOC_BALANCE:
3081 		return btrfs_balance(root->fs_info->dev_root);
3082 	case BTRFS_IOC_CLONE:
3083 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3084 	case BTRFS_IOC_CLONE_RANGE:
3085 		return btrfs_ioctl_clone_range(file, argp);
3086 	case BTRFS_IOC_TRANS_START:
3087 		return btrfs_ioctl_trans_start(file);
3088 	case BTRFS_IOC_TRANS_END:
3089 		return btrfs_ioctl_trans_end(file);
3090 	case BTRFS_IOC_TREE_SEARCH:
3091 		return btrfs_ioctl_tree_search(file, argp);
3092 	case BTRFS_IOC_INO_LOOKUP:
3093 		return btrfs_ioctl_ino_lookup(file, argp);
3094 	case BTRFS_IOC_INO_PATHS:
3095 		return btrfs_ioctl_ino_to_path(root, argp);
3096 	case BTRFS_IOC_LOGICAL_INO:
3097 		return btrfs_ioctl_logical_to_ino(root, argp);
3098 	case BTRFS_IOC_SPACE_INFO:
3099 		return btrfs_ioctl_space_info(root, argp);
3100 	case BTRFS_IOC_SYNC:
3101 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3102 		return 0;
3103 	case BTRFS_IOC_START_SYNC:
3104 		return btrfs_ioctl_start_sync(file, argp);
3105 	case BTRFS_IOC_WAIT_SYNC:
3106 		return btrfs_ioctl_wait_sync(file, argp);
3107 	case BTRFS_IOC_SCRUB:
3108 		return btrfs_ioctl_scrub(root, argp);
3109 	case BTRFS_IOC_SCRUB_CANCEL:
3110 		return btrfs_ioctl_scrub_cancel(root, argp);
3111 	case BTRFS_IOC_SCRUB_PROGRESS:
3112 		return btrfs_ioctl_scrub_progress(root, argp);
3113 	}
3114 
3115 	return -ENOTTY;
3116 }
3117