xref: /openbmc/linux/fs/btrfs/ioctl.c (revision 425d17a290c0c63785ec65db154a95c6337aeefa)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include "compat.h"
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "ioctl.h"
51 #include "print-tree.h"
52 #include "volumes.h"
53 #include "locking.h"
54 #include "inode-map.h"
55 #include "backref.h"
56 #include "rcu-string.h"
57 #include "send.h"
58 
59 /* Mask out flags that are inappropriate for the given type of inode. */
60 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
61 {
62 	if (S_ISDIR(mode))
63 		return flags;
64 	else if (S_ISREG(mode))
65 		return flags & ~FS_DIRSYNC_FL;
66 	else
67 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
68 }
69 
70 /*
71  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
72  */
73 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
74 {
75 	unsigned int iflags = 0;
76 
77 	if (flags & BTRFS_INODE_SYNC)
78 		iflags |= FS_SYNC_FL;
79 	if (flags & BTRFS_INODE_IMMUTABLE)
80 		iflags |= FS_IMMUTABLE_FL;
81 	if (flags & BTRFS_INODE_APPEND)
82 		iflags |= FS_APPEND_FL;
83 	if (flags & BTRFS_INODE_NODUMP)
84 		iflags |= FS_NODUMP_FL;
85 	if (flags & BTRFS_INODE_NOATIME)
86 		iflags |= FS_NOATIME_FL;
87 	if (flags & BTRFS_INODE_DIRSYNC)
88 		iflags |= FS_DIRSYNC_FL;
89 	if (flags & BTRFS_INODE_NODATACOW)
90 		iflags |= FS_NOCOW_FL;
91 
92 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
93 		iflags |= FS_COMPR_FL;
94 	else if (flags & BTRFS_INODE_NOCOMPRESS)
95 		iflags |= FS_NOCOMP_FL;
96 
97 	return iflags;
98 }
99 
100 /*
101  * Update inode->i_flags based on the btrfs internal flags.
102  */
103 void btrfs_update_iflags(struct inode *inode)
104 {
105 	struct btrfs_inode *ip = BTRFS_I(inode);
106 
107 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
108 
109 	if (ip->flags & BTRFS_INODE_SYNC)
110 		inode->i_flags |= S_SYNC;
111 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
112 		inode->i_flags |= S_IMMUTABLE;
113 	if (ip->flags & BTRFS_INODE_APPEND)
114 		inode->i_flags |= S_APPEND;
115 	if (ip->flags & BTRFS_INODE_NOATIME)
116 		inode->i_flags |= S_NOATIME;
117 	if (ip->flags & BTRFS_INODE_DIRSYNC)
118 		inode->i_flags |= S_DIRSYNC;
119 }
120 
121 /*
122  * Inherit flags from the parent inode.
123  *
124  * Currently only the compression flags and the cow flags are inherited.
125  */
126 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
127 {
128 	unsigned int flags;
129 
130 	if (!dir)
131 		return;
132 
133 	flags = BTRFS_I(dir)->flags;
134 
135 	if (flags & BTRFS_INODE_NOCOMPRESS) {
136 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
137 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
138 	} else if (flags & BTRFS_INODE_COMPRESS) {
139 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
140 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
141 	}
142 
143 	if (flags & BTRFS_INODE_NODATACOW)
144 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
145 
146 	btrfs_update_iflags(inode);
147 }
148 
149 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
150 {
151 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
152 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
153 
154 	if (copy_to_user(arg, &flags, sizeof(flags)))
155 		return -EFAULT;
156 	return 0;
157 }
158 
159 static int check_flags(unsigned int flags)
160 {
161 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
162 		      FS_NOATIME_FL | FS_NODUMP_FL | \
163 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
164 		      FS_NOCOMP_FL | FS_COMPR_FL |
165 		      FS_NOCOW_FL))
166 		return -EOPNOTSUPP;
167 
168 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
169 		return -EINVAL;
170 
171 	return 0;
172 }
173 
174 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
175 {
176 	struct inode *inode = file->f_path.dentry->d_inode;
177 	struct btrfs_inode *ip = BTRFS_I(inode);
178 	struct btrfs_root *root = ip->root;
179 	struct btrfs_trans_handle *trans;
180 	unsigned int flags, oldflags;
181 	int ret;
182 	u64 ip_oldflags;
183 	unsigned int i_oldflags;
184 
185 	if (btrfs_root_readonly(root))
186 		return -EROFS;
187 
188 	if (copy_from_user(&flags, arg, sizeof(flags)))
189 		return -EFAULT;
190 
191 	ret = check_flags(flags);
192 	if (ret)
193 		return ret;
194 
195 	if (!inode_owner_or_capable(inode))
196 		return -EACCES;
197 
198 	ret = mnt_want_write_file(file);
199 	if (ret)
200 		return ret;
201 
202 	mutex_lock(&inode->i_mutex);
203 
204 	ip_oldflags = ip->flags;
205 	i_oldflags = inode->i_flags;
206 
207 	flags = btrfs_mask_flags(inode->i_mode, flags);
208 	oldflags = btrfs_flags_to_ioctl(ip->flags);
209 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
210 		if (!capable(CAP_LINUX_IMMUTABLE)) {
211 			ret = -EPERM;
212 			goto out_unlock;
213 		}
214 	}
215 
216 	if (flags & FS_SYNC_FL)
217 		ip->flags |= BTRFS_INODE_SYNC;
218 	else
219 		ip->flags &= ~BTRFS_INODE_SYNC;
220 	if (flags & FS_IMMUTABLE_FL)
221 		ip->flags |= BTRFS_INODE_IMMUTABLE;
222 	else
223 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
224 	if (flags & FS_APPEND_FL)
225 		ip->flags |= BTRFS_INODE_APPEND;
226 	else
227 		ip->flags &= ~BTRFS_INODE_APPEND;
228 	if (flags & FS_NODUMP_FL)
229 		ip->flags |= BTRFS_INODE_NODUMP;
230 	else
231 		ip->flags &= ~BTRFS_INODE_NODUMP;
232 	if (flags & FS_NOATIME_FL)
233 		ip->flags |= BTRFS_INODE_NOATIME;
234 	else
235 		ip->flags &= ~BTRFS_INODE_NOATIME;
236 	if (flags & FS_DIRSYNC_FL)
237 		ip->flags |= BTRFS_INODE_DIRSYNC;
238 	else
239 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
240 	if (flags & FS_NOCOW_FL)
241 		ip->flags |= BTRFS_INODE_NODATACOW;
242 	else
243 		ip->flags &= ~BTRFS_INODE_NODATACOW;
244 
245 	/*
246 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
247 	 * flag may be changed automatically if compression code won't make
248 	 * things smaller.
249 	 */
250 	if (flags & FS_NOCOMP_FL) {
251 		ip->flags &= ~BTRFS_INODE_COMPRESS;
252 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
253 	} else if (flags & FS_COMPR_FL) {
254 		ip->flags |= BTRFS_INODE_COMPRESS;
255 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
256 	} else {
257 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
258 	}
259 
260 	trans = btrfs_start_transaction(root, 1);
261 	if (IS_ERR(trans)) {
262 		ret = PTR_ERR(trans);
263 		goto out_drop;
264 	}
265 
266 	btrfs_update_iflags(inode);
267 	inode_inc_iversion(inode);
268 	inode->i_ctime = CURRENT_TIME;
269 	ret = btrfs_update_inode(trans, root, inode);
270 
271 	btrfs_end_transaction(trans, root);
272  out_drop:
273 	if (ret) {
274 		ip->flags = ip_oldflags;
275 		inode->i_flags = i_oldflags;
276 	}
277 
278  out_unlock:
279 	mutex_unlock(&inode->i_mutex);
280 	mnt_drop_write_file(file);
281 	return ret;
282 }
283 
284 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
285 {
286 	struct inode *inode = file->f_path.dentry->d_inode;
287 
288 	return put_user(inode->i_generation, arg);
289 }
290 
291 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
292 {
293 	struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
294 	struct btrfs_device *device;
295 	struct request_queue *q;
296 	struct fstrim_range range;
297 	u64 minlen = ULLONG_MAX;
298 	u64 num_devices = 0;
299 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
300 	int ret;
301 
302 	if (!capable(CAP_SYS_ADMIN))
303 		return -EPERM;
304 
305 	rcu_read_lock();
306 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
307 				dev_list) {
308 		if (!device->bdev)
309 			continue;
310 		q = bdev_get_queue(device->bdev);
311 		if (blk_queue_discard(q)) {
312 			num_devices++;
313 			minlen = min((u64)q->limits.discard_granularity,
314 				     minlen);
315 		}
316 	}
317 	rcu_read_unlock();
318 
319 	if (!num_devices)
320 		return -EOPNOTSUPP;
321 	if (copy_from_user(&range, arg, sizeof(range)))
322 		return -EFAULT;
323 	if (range.start > total_bytes)
324 		return -EINVAL;
325 
326 	range.len = min(range.len, total_bytes - range.start);
327 	range.minlen = max(range.minlen, minlen);
328 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
329 	if (ret < 0)
330 		return ret;
331 
332 	if (copy_to_user(arg, &range, sizeof(range)))
333 		return -EFAULT;
334 
335 	return 0;
336 }
337 
338 static noinline int create_subvol(struct btrfs_root *root,
339 				  struct dentry *dentry,
340 				  char *name, int namelen,
341 				  u64 *async_transid,
342 				  struct btrfs_qgroup_inherit **inherit)
343 {
344 	struct btrfs_trans_handle *trans;
345 	struct btrfs_key key;
346 	struct btrfs_root_item root_item;
347 	struct btrfs_inode_item *inode_item;
348 	struct extent_buffer *leaf;
349 	struct btrfs_root *new_root;
350 	struct dentry *parent = dentry->d_parent;
351 	struct inode *dir;
352 	struct timespec cur_time = CURRENT_TIME;
353 	int ret;
354 	int err;
355 	u64 objectid;
356 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
357 	u64 index = 0;
358 	uuid_le new_uuid;
359 
360 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
361 	if (ret)
362 		return ret;
363 
364 	dir = parent->d_inode;
365 
366 	/*
367 	 * 1 - inode item
368 	 * 2 - refs
369 	 * 1 - root item
370 	 * 2 - dir items
371 	 */
372 	trans = btrfs_start_transaction(root, 6);
373 	if (IS_ERR(trans))
374 		return PTR_ERR(trans);
375 
376 	ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid,
377 				   inherit ? *inherit : NULL);
378 	if (ret)
379 		goto fail;
380 
381 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
382 				      0, objectid, NULL, 0, 0, 0);
383 	if (IS_ERR(leaf)) {
384 		ret = PTR_ERR(leaf);
385 		goto fail;
386 	}
387 
388 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
389 	btrfs_set_header_bytenr(leaf, leaf->start);
390 	btrfs_set_header_generation(leaf, trans->transid);
391 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
392 	btrfs_set_header_owner(leaf, objectid);
393 
394 	write_extent_buffer(leaf, root->fs_info->fsid,
395 			    (unsigned long)btrfs_header_fsid(leaf),
396 			    BTRFS_FSID_SIZE);
397 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
398 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
399 			    BTRFS_UUID_SIZE);
400 	btrfs_mark_buffer_dirty(leaf);
401 
402 	memset(&root_item, 0, sizeof(root_item));
403 
404 	inode_item = &root_item.inode;
405 	inode_item->generation = cpu_to_le64(1);
406 	inode_item->size = cpu_to_le64(3);
407 	inode_item->nlink = cpu_to_le32(1);
408 	inode_item->nbytes = cpu_to_le64(root->leafsize);
409 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
410 
411 	root_item.flags = 0;
412 	root_item.byte_limit = 0;
413 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
414 
415 	btrfs_set_root_bytenr(&root_item, leaf->start);
416 	btrfs_set_root_generation(&root_item, trans->transid);
417 	btrfs_set_root_level(&root_item, 0);
418 	btrfs_set_root_refs(&root_item, 1);
419 	btrfs_set_root_used(&root_item, leaf->len);
420 	btrfs_set_root_last_snapshot(&root_item, 0);
421 
422 	btrfs_set_root_generation_v2(&root_item,
423 			btrfs_root_generation(&root_item));
424 	uuid_le_gen(&new_uuid);
425 	memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
426 	root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
427 	root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
428 	root_item.ctime = root_item.otime;
429 	btrfs_set_root_ctransid(&root_item, trans->transid);
430 	btrfs_set_root_otransid(&root_item, trans->transid);
431 
432 	btrfs_tree_unlock(leaf);
433 	free_extent_buffer(leaf);
434 	leaf = NULL;
435 
436 	btrfs_set_root_dirid(&root_item, new_dirid);
437 
438 	key.objectid = objectid;
439 	key.offset = 0;
440 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
441 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
442 				&root_item);
443 	if (ret)
444 		goto fail;
445 
446 	key.offset = (u64)-1;
447 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
448 	if (IS_ERR(new_root)) {
449 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
450 		ret = PTR_ERR(new_root);
451 		goto fail;
452 	}
453 
454 	btrfs_record_root_in_trans(trans, new_root);
455 
456 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
457 	if (ret) {
458 		/* We potentially lose an unused inode item here */
459 		btrfs_abort_transaction(trans, root, ret);
460 		goto fail;
461 	}
462 
463 	/*
464 	 * insert the directory item
465 	 */
466 	ret = btrfs_set_inode_index(dir, &index);
467 	if (ret) {
468 		btrfs_abort_transaction(trans, root, ret);
469 		goto fail;
470 	}
471 
472 	ret = btrfs_insert_dir_item(trans, root,
473 				    name, namelen, dir, &key,
474 				    BTRFS_FT_DIR, index);
475 	if (ret) {
476 		btrfs_abort_transaction(trans, root, ret);
477 		goto fail;
478 	}
479 
480 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
481 	ret = btrfs_update_inode(trans, root, dir);
482 	BUG_ON(ret);
483 
484 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
485 				 objectid, root->root_key.objectid,
486 				 btrfs_ino(dir), index, name, namelen);
487 
488 	BUG_ON(ret);
489 
490 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
491 fail:
492 	if (async_transid) {
493 		*async_transid = trans->transid;
494 		err = btrfs_commit_transaction_async(trans, root, 1);
495 	} else {
496 		err = btrfs_commit_transaction(trans, root);
497 	}
498 	if (err && !ret)
499 		ret = err;
500 	return ret;
501 }
502 
503 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
504 			   char *name, int namelen, u64 *async_transid,
505 			   bool readonly, struct btrfs_qgroup_inherit **inherit)
506 {
507 	struct inode *inode;
508 	struct btrfs_pending_snapshot *pending_snapshot;
509 	struct btrfs_trans_handle *trans;
510 	int ret;
511 
512 	if (!root->ref_cows)
513 		return -EINVAL;
514 
515 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
516 	if (!pending_snapshot)
517 		return -ENOMEM;
518 
519 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
520 			     BTRFS_BLOCK_RSV_TEMP);
521 	pending_snapshot->dentry = dentry;
522 	pending_snapshot->root = root;
523 	pending_snapshot->readonly = readonly;
524 	if (inherit) {
525 		pending_snapshot->inherit = *inherit;
526 		*inherit = NULL;	/* take responsibility to free it */
527 	}
528 
529 	trans = btrfs_start_transaction(root->fs_info->extent_root, 6);
530 	if (IS_ERR(trans)) {
531 		ret = PTR_ERR(trans);
532 		goto fail;
533 	}
534 
535 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
536 	BUG_ON(ret);
537 
538 	spin_lock(&root->fs_info->trans_lock);
539 	list_add(&pending_snapshot->list,
540 		 &trans->transaction->pending_snapshots);
541 	spin_unlock(&root->fs_info->trans_lock);
542 	if (async_transid) {
543 		*async_transid = trans->transid;
544 		ret = btrfs_commit_transaction_async(trans,
545 				     root->fs_info->extent_root, 1);
546 	} else {
547 		ret = btrfs_commit_transaction(trans,
548 					       root->fs_info->extent_root);
549 	}
550 	BUG_ON(ret);
551 
552 	ret = pending_snapshot->error;
553 	if (ret)
554 		goto fail;
555 
556 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
557 	if (ret)
558 		goto fail;
559 
560 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
561 	if (IS_ERR(inode)) {
562 		ret = PTR_ERR(inode);
563 		goto fail;
564 	}
565 	BUG_ON(!inode);
566 	d_instantiate(dentry, inode);
567 	ret = 0;
568 fail:
569 	kfree(pending_snapshot);
570 	return ret;
571 }
572 
573 /*  copy of check_sticky in fs/namei.c()
574 * It's inline, so penalty for filesystems that don't use sticky bit is
575 * minimal.
576 */
577 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
578 {
579 	uid_t fsuid = current_fsuid();
580 
581 	if (!(dir->i_mode & S_ISVTX))
582 		return 0;
583 	if (inode->i_uid == fsuid)
584 		return 0;
585 	if (dir->i_uid == fsuid)
586 		return 0;
587 	return !capable(CAP_FOWNER);
588 }
589 
590 /*  copy of may_delete in fs/namei.c()
591  *	Check whether we can remove a link victim from directory dir, check
592  *  whether the type of victim is right.
593  *  1. We can't do it if dir is read-only (done in permission())
594  *  2. We should have write and exec permissions on dir
595  *  3. We can't remove anything from append-only dir
596  *  4. We can't do anything with immutable dir (done in permission())
597  *  5. If the sticky bit on dir is set we should either
598  *	a. be owner of dir, or
599  *	b. be owner of victim, or
600  *	c. have CAP_FOWNER capability
601  *  6. If the victim is append-only or immutable we can't do antyhing with
602  *     links pointing to it.
603  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
604  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
605  *  9. We can't remove a root or mountpoint.
606  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
607  *     nfs_async_unlink().
608  */
609 
610 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
611 {
612 	int error;
613 
614 	if (!victim->d_inode)
615 		return -ENOENT;
616 
617 	BUG_ON(victim->d_parent->d_inode != dir);
618 	audit_inode_child(victim, dir);
619 
620 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
621 	if (error)
622 		return error;
623 	if (IS_APPEND(dir))
624 		return -EPERM;
625 	if (btrfs_check_sticky(dir, victim->d_inode)||
626 		IS_APPEND(victim->d_inode)||
627 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
628 		return -EPERM;
629 	if (isdir) {
630 		if (!S_ISDIR(victim->d_inode->i_mode))
631 			return -ENOTDIR;
632 		if (IS_ROOT(victim))
633 			return -EBUSY;
634 	} else if (S_ISDIR(victim->d_inode->i_mode))
635 		return -EISDIR;
636 	if (IS_DEADDIR(dir))
637 		return -ENOENT;
638 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
639 		return -EBUSY;
640 	return 0;
641 }
642 
643 /* copy of may_create in fs/namei.c() */
644 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
645 {
646 	if (child->d_inode)
647 		return -EEXIST;
648 	if (IS_DEADDIR(dir))
649 		return -ENOENT;
650 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
651 }
652 
653 /*
654  * Create a new subvolume below @parent.  This is largely modeled after
655  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
656  * inside this filesystem so it's quite a bit simpler.
657  */
658 static noinline int btrfs_mksubvol(struct path *parent,
659 				   char *name, int namelen,
660 				   struct btrfs_root *snap_src,
661 				   u64 *async_transid, bool readonly,
662 				   struct btrfs_qgroup_inherit **inherit)
663 {
664 	struct inode *dir  = parent->dentry->d_inode;
665 	struct dentry *dentry;
666 	int error;
667 
668 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
669 
670 	dentry = lookup_one_len(name, parent->dentry, namelen);
671 	error = PTR_ERR(dentry);
672 	if (IS_ERR(dentry))
673 		goto out_unlock;
674 
675 	error = -EEXIST;
676 	if (dentry->d_inode)
677 		goto out_dput;
678 
679 	error = btrfs_may_create(dir, dentry);
680 	if (error)
681 		goto out_dput;
682 
683 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
684 
685 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
686 		goto out_up_read;
687 
688 	if (snap_src) {
689 		error = create_snapshot(snap_src, dentry, name, namelen,
690 					async_transid, readonly, inherit);
691 	} else {
692 		error = create_subvol(BTRFS_I(dir)->root, dentry,
693 				      name, namelen, async_transid, inherit);
694 	}
695 	if (!error)
696 		fsnotify_mkdir(dir, dentry);
697 out_up_read:
698 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
699 out_dput:
700 	dput(dentry);
701 out_unlock:
702 	mutex_unlock(&dir->i_mutex);
703 	return error;
704 }
705 
706 /*
707  * When we're defragging a range, we don't want to kick it off again
708  * if it is really just waiting for delalloc to send it down.
709  * If we find a nice big extent or delalloc range for the bytes in the
710  * file you want to defrag, we return 0 to let you know to skip this
711  * part of the file
712  */
713 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
714 {
715 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
716 	struct extent_map *em = NULL;
717 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
718 	u64 end;
719 
720 	read_lock(&em_tree->lock);
721 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
722 	read_unlock(&em_tree->lock);
723 
724 	if (em) {
725 		end = extent_map_end(em);
726 		free_extent_map(em);
727 		if (end - offset > thresh)
728 			return 0;
729 	}
730 	/* if we already have a nice delalloc here, just stop */
731 	thresh /= 2;
732 	end = count_range_bits(io_tree, &offset, offset + thresh,
733 			       thresh, EXTENT_DELALLOC, 1);
734 	if (end >= thresh)
735 		return 0;
736 	return 1;
737 }
738 
739 /*
740  * helper function to walk through a file and find extents
741  * newer than a specific transid, and smaller than thresh.
742  *
743  * This is used by the defragging code to find new and small
744  * extents
745  */
746 static int find_new_extents(struct btrfs_root *root,
747 			    struct inode *inode, u64 newer_than,
748 			    u64 *off, int thresh)
749 {
750 	struct btrfs_path *path;
751 	struct btrfs_key min_key;
752 	struct btrfs_key max_key;
753 	struct extent_buffer *leaf;
754 	struct btrfs_file_extent_item *extent;
755 	int type;
756 	int ret;
757 	u64 ino = btrfs_ino(inode);
758 
759 	path = btrfs_alloc_path();
760 	if (!path)
761 		return -ENOMEM;
762 
763 	min_key.objectid = ino;
764 	min_key.type = BTRFS_EXTENT_DATA_KEY;
765 	min_key.offset = *off;
766 
767 	max_key.objectid = ino;
768 	max_key.type = (u8)-1;
769 	max_key.offset = (u64)-1;
770 
771 	path->keep_locks = 1;
772 
773 	while(1) {
774 		ret = btrfs_search_forward(root, &min_key, &max_key,
775 					   path, 0, newer_than);
776 		if (ret != 0)
777 			goto none;
778 		if (min_key.objectid != ino)
779 			goto none;
780 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
781 			goto none;
782 
783 		leaf = path->nodes[0];
784 		extent = btrfs_item_ptr(leaf, path->slots[0],
785 					struct btrfs_file_extent_item);
786 
787 		type = btrfs_file_extent_type(leaf, extent);
788 		if (type == BTRFS_FILE_EXTENT_REG &&
789 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
790 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
791 			*off = min_key.offset;
792 			btrfs_free_path(path);
793 			return 0;
794 		}
795 
796 		if (min_key.offset == (u64)-1)
797 			goto none;
798 
799 		min_key.offset++;
800 		btrfs_release_path(path);
801 	}
802 none:
803 	btrfs_free_path(path);
804 	return -ENOENT;
805 }
806 
807 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
808 {
809 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
810 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
811 	struct extent_map *em;
812 	u64 len = PAGE_CACHE_SIZE;
813 
814 	/*
815 	 * hopefully we have this extent in the tree already, try without
816 	 * the full extent lock
817 	 */
818 	read_lock(&em_tree->lock);
819 	em = lookup_extent_mapping(em_tree, start, len);
820 	read_unlock(&em_tree->lock);
821 
822 	if (!em) {
823 		/* get the big lock and read metadata off disk */
824 		lock_extent(io_tree, start, start + len - 1);
825 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
826 		unlock_extent(io_tree, start, start + len - 1);
827 
828 		if (IS_ERR(em))
829 			return NULL;
830 	}
831 
832 	return em;
833 }
834 
835 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
836 {
837 	struct extent_map *next;
838 	bool ret = true;
839 
840 	/* this is the last extent */
841 	if (em->start + em->len >= i_size_read(inode))
842 		return false;
843 
844 	next = defrag_lookup_extent(inode, em->start + em->len);
845 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
846 		ret = false;
847 
848 	free_extent_map(next);
849 	return ret;
850 }
851 
852 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
853 			       u64 *last_len, u64 *skip, u64 *defrag_end,
854 			       int compress)
855 {
856 	struct extent_map *em;
857 	int ret = 1;
858 	bool next_mergeable = true;
859 
860 	/*
861 	 * make sure that once we start defragging an extent, we keep on
862 	 * defragging it
863 	 */
864 	if (start < *defrag_end)
865 		return 1;
866 
867 	*skip = 0;
868 
869 	em = defrag_lookup_extent(inode, start);
870 	if (!em)
871 		return 0;
872 
873 	/* this will cover holes, and inline extents */
874 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
875 		ret = 0;
876 		goto out;
877 	}
878 
879 	next_mergeable = defrag_check_next_extent(inode, em);
880 
881 	/*
882 	 * we hit a real extent, if it is big or the next extent is not a
883 	 * real extent, don't bother defragging it
884 	 */
885 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
886 	    (em->len >= thresh || !next_mergeable))
887 		ret = 0;
888 out:
889 	/*
890 	 * last_len ends up being a counter of how many bytes we've defragged.
891 	 * every time we choose not to defrag an extent, we reset *last_len
892 	 * so that the next tiny extent will force a defrag.
893 	 *
894 	 * The end result of this is that tiny extents before a single big
895 	 * extent will force at least part of that big extent to be defragged.
896 	 */
897 	if (ret) {
898 		*defrag_end = extent_map_end(em);
899 	} else {
900 		*last_len = 0;
901 		*skip = extent_map_end(em);
902 		*defrag_end = 0;
903 	}
904 
905 	free_extent_map(em);
906 	return ret;
907 }
908 
909 /*
910  * it doesn't do much good to defrag one or two pages
911  * at a time.  This pulls in a nice chunk of pages
912  * to COW and defrag.
913  *
914  * It also makes sure the delalloc code has enough
915  * dirty data to avoid making new small extents as part
916  * of the defrag
917  *
918  * It's a good idea to start RA on this range
919  * before calling this.
920  */
921 static int cluster_pages_for_defrag(struct inode *inode,
922 				    struct page **pages,
923 				    unsigned long start_index,
924 				    int num_pages)
925 {
926 	unsigned long file_end;
927 	u64 isize = i_size_read(inode);
928 	u64 page_start;
929 	u64 page_end;
930 	u64 page_cnt;
931 	int ret;
932 	int i;
933 	int i_done;
934 	struct btrfs_ordered_extent *ordered;
935 	struct extent_state *cached_state = NULL;
936 	struct extent_io_tree *tree;
937 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
938 
939 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
940 	if (!isize || start_index > file_end)
941 		return 0;
942 
943 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
944 
945 	ret = btrfs_delalloc_reserve_space(inode,
946 					   page_cnt << PAGE_CACHE_SHIFT);
947 	if (ret)
948 		return ret;
949 	i_done = 0;
950 	tree = &BTRFS_I(inode)->io_tree;
951 
952 	/* step one, lock all the pages */
953 	for (i = 0; i < page_cnt; i++) {
954 		struct page *page;
955 again:
956 		page = find_or_create_page(inode->i_mapping,
957 					   start_index + i, mask);
958 		if (!page)
959 			break;
960 
961 		page_start = page_offset(page);
962 		page_end = page_start + PAGE_CACHE_SIZE - 1;
963 		while (1) {
964 			lock_extent(tree, page_start, page_end);
965 			ordered = btrfs_lookup_ordered_extent(inode,
966 							      page_start);
967 			unlock_extent(tree, page_start, page_end);
968 			if (!ordered)
969 				break;
970 
971 			unlock_page(page);
972 			btrfs_start_ordered_extent(inode, ordered, 1);
973 			btrfs_put_ordered_extent(ordered);
974 			lock_page(page);
975 			/*
976 			 * we unlocked the page above, so we need check if
977 			 * it was released or not.
978 			 */
979 			if (page->mapping != inode->i_mapping) {
980 				unlock_page(page);
981 				page_cache_release(page);
982 				goto again;
983 			}
984 		}
985 
986 		if (!PageUptodate(page)) {
987 			btrfs_readpage(NULL, page);
988 			lock_page(page);
989 			if (!PageUptodate(page)) {
990 				unlock_page(page);
991 				page_cache_release(page);
992 				ret = -EIO;
993 				break;
994 			}
995 		}
996 
997 		if (page->mapping != inode->i_mapping) {
998 			unlock_page(page);
999 			page_cache_release(page);
1000 			goto again;
1001 		}
1002 
1003 		pages[i] = page;
1004 		i_done++;
1005 	}
1006 	if (!i_done || ret)
1007 		goto out;
1008 
1009 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
1010 		goto out;
1011 
1012 	/*
1013 	 * so now we have a nice long stream of locked
1014 	 * and up to date pages, lets wait on them
1015 	 */
1016 	for (i = 0; i < i_done; i++)
1017 		wait_on_page_writeback(pages[i]);
1018 
1019 	page_start = page_offset(pages[0]);
1020 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1021 
1022 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1023 			 page_start, page_end - 1, 0, &cached_state);
1024 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1025 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1026 			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1027 			  &cached_state, GFP_NOFS);
1028 
1029 	if (i_done != page_cnt) {
1030 		spin_lock(&BTRFS_I(inode)->lock);
1031 		BTRFS_I(inode)->outstanding_extents++;
1032 		spin_unlock(&BTRFS_I(inode)->lock);
1033 		btrfs_delalloc_release_space(inode,
1034 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1035 	}
1036 
1037 
1038 	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1039 			  &cached_state, GFP_NOFS);
1040 
1041 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1042 			     page_start, page_end - 1, &cached_state,
1043 			     GFP_NOFS);
1044 
1045 	for (i = 0; i < i_done; i++) {
1046 		clear_page_dirty_for_io(pages[i]);
1047 		ClearPageChecked(pages[i]);
1048 		set_page_extent_mapped(pages[i]);
1049 		set_page_dirty(pages[i]);
1050 		unlock_page(pages[i]);
1051 		page_cache_release(pages[i]);
1052 	}
1053 	return i_done;
1054 out:
1055 	for (i = 0; i < i_done; i++) {
1056 		unlock_page(pages[i]);
1057 		page_cache_release(pages[i]);
1058 	}
1059 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1060 	return ret;
1061 
1062 }
1063 
1064 int btrfs_defrag_file(struct inode *inode, struct file *file,
1065 		      struct btrfs_ioctl_defrag_range_args *range,
1066 		      u64 newer_than, unsigned long max_to_defrag)
1067 {
1068 	struct btrfs_root *root = BTRFS_I(inode)->root;
1069 	struct file_ra_state *ra = NULL;
1070 	unsigned long last_index;
1071 	u64 isize = i_size_read(inode);
1072 	u64 last_len = 0;
1073 	u64 skip = 0;
1074 	u64 defrag_end = 0;
1075 	u64 newer_off = range->start;
1076 	unsigned long i;
1077 	unsigned long ra_index = 0;
1078 	int ret;
1079 	int defrag_count = 0;
1080 	int compress_type = BTRFS_COMPRESS_ZLIB;
1081 	int extent_thresh = range->extent_thresh;
1082 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1083 	int cluster = max_cluster;
1084 	u64 new_align = ~((u64)128 * 1024 - 1);
1085 	struct page **pages = NULL;
1086 
1087 	if (extent_thresh == 0)
1088 		extent_thresh = 256 * 1024;
1089 
1090 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1091 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1092 			return -EINVAL;
1093 		if (range->compress_type)
1094 			compress_type = range->compress_type;
1095 	}
1096 
1097 	if (isize == 0)
1098 		return 0;
1099 
1100 	/*
1101 	 * if we were not given a file, allocate a readahead
1102 	 * context
1103 	 */
1104 	if (!file) {
1105 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1106 		if (!ra)
1107 			return -ENOMEM;
1108 		file_ra_state_init(ra, inode->i_mapping);
1109 	} else {
1110 		ra = &file->f_ra;
1111 	}
1112 
1113 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1114 			GFP_NOFS);
1115 	if (!pages) {
1116 		ret = -ENOMEM;
1117 		goto out_ra;
1118 	}
1119 
1120 	/* find the last page to defrag */
1121 	if (range->start + range->len > range->start) {
1122 		last_index = min_t(u64, isize - 1,
1123 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1124 	} else {
1125 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1126 	}
1127 
1128 	if (newer_than) {
1129 		ret = find_new_extents(root, inode, newer_than,
1130 				       &newer_off, 64 * 1024);
1131 		if (!ret) {
1132 			range->start = newer_off;
1133 			/*
1134 			 * we always align our defrag to help keep
1135 			 * the extents in the file evenly spaced
1136 			 */
1137 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1138 		} else
1139 			goto out_ra;
1140 	} else {
1141 		i = range->start >> PAGE_CACHE_SHIFT;
1142 	}
1143 	if (!max_to_defrag)
1144 		max_to_defrag = last_index + 1;
1145 
1146 	/*
1147 	 * make writeback starts from i, so the defrag range can be
1148 	 * written sequentially.
1149 	 */
1150 	if (i < inode->i_mapping->writeback_index)
1151 		inode->i_mapping->writeback_index = i;
1152 
1153 	while (i <= last_index && defrag_count < max_to_defrag &&
1154 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1155 		PAGE_CACHE_SHIFT)) {
1156 		/*
1157 		 * make sure we stop running if someone unmounts
1158 		 * the FS
1159 		 */
1160 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1161 			break;
1162 
1163 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1164 					 extent_thresh, &last_len, &skip,
1165 					 &defrag_end, range->flags &
1166 					 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1167 			unsigned long next;
1168 			/*
1169 			 * the should_defrag function tells us how much to skip
1170 			 * bump our counter by the suggested amount
1171 			 */
1172 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1173 			i = max(i + 1, next);
1174 			continue;
1175 		}
1176 
1177 		if (!newer_than) {
1178 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1179 				   PAGE_CACHE_SHIFT) - i;
1180 			cluster = min(cluster, max_cluster);
1181 		} else {
1182 			cluster = max_cluster;
1183 		}
1184 
1185 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1186 			BTRFS_I(inode)->force_compress = compress_type;
1187 
1188 		if (i + cluster > ra_index) {
1189 			ra_index = max(i, ra_index);
1190 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1191 				       cluster);
1192 			ra_index += max_cluster;
1193 		}
1194 
1195 		mutex_lock(&inode->i_mutex);
1196 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1197 		if (ret < 0) {
1198 			mutex_unlock(&inode->i_mutex);
1199 			goto out_ra;
1200 		}
1201 
1202 		defrag_count += ret;
1203 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1204 		mutex_unlock(&inode->i_mutex);
1205 
1206 		if (newer_than) {
1207 			if (newer_off == (u64)-1)
1208 				break;
1209 
1210 			if (ret > 0)
1211 				i += ret;
1212 
1213 			newer_off = max(newer_off + 1,
1214 					(u64)i << PAGE_CACHE_SHIFT);
1215 
1216 			ret = find_new_extents(root, inode,
1217 					       newer_than, &newer_off,
1218 					       64 * 1024);
1219 			if (!ret) {
1220 				range->start = newer_off;
1221 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1222 			} else {
1223 				break;
1224 			}
1225 		} else {
1226 			if (ret > 0) {
1227 				i += ret;
1228 				last_len += ret << PAGE_CACHE_SHIFT;
1229 			} else {
1230 				i++;
1231 				last_len = 0;
1232 			}
1233 		}
1234 	}
1235 
1236 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1237 		filemap_flush(inode->i_mapping);
1238 
1239 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1240 		/* the filemap_flush will queue IO into the worker threads, but
1241 		 * we have to make sure the IO is actually started and that
1242 		 * ordered extents get created before we return
1243 		 */
1244 		atomic_inc(&root->fs_info->async_submit_draining);
1245 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1246 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1247 			wait_event(root->fs_info->async_submit_wait,
1248 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1249 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1250 		}
1251 		atomic_dec(&root->fs_info->async_submit_draining);
1252 
1253 		mutex_lock(&inode->i_mutex);
1254 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1255 		mutex_unlock(&inode->i_mutex);
1256 	}
1257 
1258 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1259 		btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1260 	}
1261 
1262 	ret = defrag_count;
1263 
1264 out_ra:
1265 	if (!file)
1266 		kfree(ra);
1267 	kfree(pages);
1268 	return ret;
1269 }
1270 
1271 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1272 					void __user *arg)
1273 {
1274 	u64 new_size;
1275 	u64 old_size;
1276 	u64 devid = 1;
1277 	struct btrfs_ioctl_vol_args *vol_args;
1278 	struct btrfs_trans_handle *trans;
1279 	struct btrfs_device *device = NULL;
1280 	char *sizestr;
1281 	char *devstr = NULL;
1282 	int ret = 0;
1283 	int mod = 0;
1284 
1285 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1286 		return -EROFS;
1287 
1288 	if (!capable(CAP_SYS_ADMIN))
1289 		return -EPERM;
1290 
1291 	mutex_lock(&root->fs_info->volume_mutex);
1292 	if (root->fs_info->balance_ctl) {
1293 		printk(KERN_INFO "btrfs: balance in progress\n");
1294 		ret = -EINVAL;
1295 		goto out;
1296 	}
1297 
1298 	vol_args = memdup_user(arg, sizeof(*vol_args));
1299 	if (IS_ERR(vol_args)) {
1300 		ret = PTR_ERR(vol_args);
1301 		goto out;
1302 	}
1303 
1304 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1305 
1306 	sizestr = vol_args->name;
1307 	devstr = strchr(sizestr, ':');
1308 	if (devstr) {
1309 		char *end;
1310 		sizestr = devstr + 1;
1311 		*devstr = '\0';
1312 		devstr = vol_args->name;
1313 		devid = simple_strtoull(devstr, &end, 10);
1314 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1315 		       (unsigned long long)devid);
1316 	}
1317 	device = btrfs_find_device(root, devid, NULL, NULL);
1318 	if (!device) {
1319 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1320 		       (unsigned long long)devid);
1321 		ret = -EINVAL;
1322 		goto out_free;
1323 	}
1324 	if (device->fs_devices && device->fs_devices->seeding) {
1325 		printk(KERN_INFO "btrfs: resizer unable to apply on "
1326 		       "seeding device %llu\n",
1327 		       (unsigned long long)devid);
1328 		ret = -EINVAL;
1329 		goto out_free;
1330 	}
1331 
1332 	if (!strcmp(sizestr, "max"))
1333 		new_size = device->bdev->bd_inode->i_size;
1334 	else {
1335 		if (sizestr[0] == '-') {
1336 			mod = -1;
1337 			sizestr++;
1338 		} else if (sizestr[0] == '+') {
1339 			mod = 1;
1340 			sizestr++;
1341 		}
1342 		new_size = memparse(sizestr, NULL);
1343 		if (new_size == 0) {
1344 			ret = -EINVAL;
1345 			goto out_free;
1346 		}
1347 	}
1348 
1349 	old_size = device->total_bytes;
1350 
1351 	if (mod < 0) {
1352 		if (new_size > old_size) {
1353 			ret = -EINVAL;
1354 			goto out_free;
1355 		}
1356 		new_size = old_size - new_size;
1357 	} else if (mod > 0) {
1358 		new_size = old_size + new_size;
1359 	}
1360 
1361 	if (new_size < 256 * 1024 * 1024) {
1362 		ret = -EINVAL;
1363 		goto out_free;
1364 	}
1365 	if (new_size > device->bdev->bd_inode->i_size) {
1366 		ret = -EFBIG;
1367 		goto out_free;
1368 	}
1369 
1370 	do_div(new_size, root->sectorsize);
1371 	new_size *= root->sectorsize;
1372 
1373 	printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1374 		      rcu_str_deref(device->name),
1375 		      (unsigned long long)new_size);
1376 
1377 	if (new_size > old_size) {
1378 		trans = btrfs_start_transaction(root, 0);
1379 		if (IS_ERR(trans)) {
1380 			ret = PTR_ERR(trans);
1381 			goto out_free;
1382 		}
1383 		ret = btrfs_grow_device(trans, device, new_size);
1384 		btrfs_commit_transaction(trans, root);
1385 	} else if (new_size < old_size) {
1386 		ret = btrfs_shrink_device(device, new_size);
1387 	}
1388 
1389 out_free:
1390 	kfree(vol_args);
1391 out:
1392 	mutex_unlock(&root->fs_info->volume_mutex);
1393 	return ret;
1394 }
1395 
1396 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1397 				char *name, unsigned long fd, int subvol,
1398 				u64 *transid, bool readonly,
1399 				struct btrfs_qgroup_inherit **inherit)
1400 {
1401 	struct file *src_file;
1402 	int namelen;
1403 	int ret = 0;
1404 
1405 	ret = mnt_want_write_file(file);
1406 	if (ret)
1407 		goto out;
1408 
1409 	namelen = strlen(name);
1410 	if (strchr(name, '/')) {
1411 		ret = -EINVAL;
1412 		goto out_drop_write;
1413 	}
1414 
1415 	if (name[0] == '.' &&
1416 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1417 		ret = -EEXIST;
1418 		goto out_drop_write;
1419 	}
1420 
1421 	if (subvol) {
1422 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1423 				     NULL, transid, readonly, inherit);
1424 	} else {
1425 		struct inode *src_inode;
1426 		src_file = fget(fd);
1427 		if (!src_file) {
1428 			ret = -EINVAL;
1429 			goto out_drop_write;
1430 		}
1431 
1432 		src_inode = src_file->f_path.dentry->d_inode;
1433 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1434 			printk(KERN_INFO "btrfs: Snapshot src from "
1435 			       "another FS\n");
1436 			ret = -EINVAL;
1437 			fput(src_file);
1438 			goto out_drop_write;
1439 		}
1440 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1441 				     BTRFS_I(src_inode)->root,
1442 				     transid, readonly, inherit);
1443 		fput(src_file);
1444 	}
1445 out_drop_write:
1446 	mnt_drop_write_file(file);
1447 out:
1448 	return ret;
1449 }
1450 
1451 static noinline int btrfs_ioctl_snap_create(struct file *file,
1452 					    void __user *arg, int subvol)
1453 {
1454 	struct btrfs_ioctl_vol_args *vol_args;
1455 	int ret;
1456 
1457 	vol_args = memdup_user(arg, sizeof(*vol_args));
1458 	if (IS_ERR(vol_args))
1459 		return PTR_ERR(vol_args);
1460 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1461 
1462 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1463 					      vol_args->fd, subvol,
1464 					      NULL, false, NULL);
1465 
1466 	kfree(vol_args);
1467 	return ret;
1468 }
1469 
1470 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1471 					       void __user *arg, int subvol)
1472 {
1473 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1474 	int ret;
1475 	u64 transid = 0;
1476 	u64 *ptr = NULL;
1477 	bool readonly = false;
1478 	struct btrfs_qgroup_inherit *inherit = NULL;
1479 
1480 	vol_args = memdup_user(arg, sizeof(*vol_args));
1481 	if (IS_ERR(vol_args))
1482 		return PTR_ERR(vol_args);
1483 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1484 
1485 	if (vol_args->flags &
1486 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1487 	      BTRFS_SUBVOL_QGROUP_INHERIT)) {
1488 		ret = -EOPNOTSUPP;
1489 		goto out;
1490 	}
1491 
1492 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1493 		ptr = &transid;
1494 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1495 		readonly = true;
1496 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1497 		if (vol_args->size > PAGE_CACHE_SIZE) {
1498 			ret = -EINVAL;
1499 			goto out;
1500 		}
1501 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1502 		if (IS_ERR(inherit)) {
1503 			ret = PTR_ERR(inherit);
1504 			goto out;
1505 		}
1506 	}
1507 
1508 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1509 					      vol_args->fd, subvol, ptr,
1510 					      readonly, &inherit);
1511 
1512 	if (ret == 0 && ptr &&
1513 	    copy_to_user(arg +
1514 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1515 				  transid), ptr, sizeof(*ptr)))
1516 		ret = -EFAULT;
1517 out:
1518 	kfree(vol_args);
1519 	kfree(inherit);
1520 	return ret;
1521 }
1522 
1523 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1524 						void __user *arg)
1525 {
1526 	struct inode *inode = fdentry(file)->d_inode;
1527 	struct btrfs_root *root = BTRFS_I(inode)->root;
1528 	int ret = 0;
1529 	u64 flags = 0;
1530 
1531 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1532 		return -EINVAL;
1533 
1534 	down_read(&root->fs_info->subvol_sem);
1535 	if (btrfs_root_readonly(root))
1536 		flags |= BTRFS_SUBVOL_RDONLY;
1537 	up_read(&root->fs_info->subvol_sem);
1538 
1539 	if (copy_to_user(arg, &flags, sizeof(flags)))
1540 		ret = -EFAULT;
1541 
1542 	return ret;
1543 }
1544 
1545 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1546 					      void __user *arg)
1547 {
1548 	struct inode *inode = fdentry(file)->d_inode;
1549 	struct btrfs_root *root = BTRFS_I(inode)->root;
1550 	struct btrfs_trans_handle *trans;
1551 	u64 root_flags;
1552 	u64 flags;
1553 	int ret = 0;
1554 
1555 	ret = mnt_want_write_file(file);
1556 	if (ret)
1557 		goto out;
1558 
1559 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1560 		ret = -EINVAL;
1561 		goto out_drop_write;
1562 	}
1563 
1564 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1565 		ret = -EFAULT;
1566 		goto out_drop_write;
1567 	}
1568 
1569 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1570 		ret = -EINVAL;
1571 		goto out_drop_write;
1572 	}
1573 
1574 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1575 		ret = -EOPNOTSUPP;
1576 		goto out_drop_write;
1577 	}
1578 
1579 	if (!inode_owner_or_capable(inode)) {
1580 		ret = -EACCES;
1581 		goto out_drop_write;
1582 	}
1583 
1584 	down_write(&root->fs_info->subvol_sem);
1585 
1586 	/* nothing to do */
1587 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1588 		goto out_drop_sem;
1589 
1590 	root_flags = btrfs_root_flags(&root->root_item);
1591 	if (flags & BTRFS_SUBVOL_RDONLY)
1592 		btrfs_set_root_flags(&root->root_item,
1593 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1594 	else
1595 		btrfs_set_root_flags(&root->root_item,
1596 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1597 
1598 	trans = btrfs_start_transaction(root, 1);
1599 	if (IS_ERR(trans)) {
1600 		ret = PTR_ERR(trans);
1601 		goto out_reset;
1602 	}
1603 
1604 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1605 				&root->root_key, &root->root_item);
1606 
1607 	btrfs_commit_transaction(trans, root);
1608 out_reset:
1609 	if (ret)
1610 		btrfs_set_root_flags(&root->root_item, root_flags);
1611 out_drop_sem:
1612 	up_write(&root->fs_info->subvol_sem);
1613 out_drop_write:
1614 	mnt_drop_write_file(file);
1615 out:
1616 	return ret;
1617 }
1618 
1619 /*
1620  * helper to check if the subvolume references other subvolumes
1621  */
1622 static noinline int may_destroy_subvol(struct btrfs_root *root)
1623 {
1624 	struct btrfs_path *path;
1625 	struct btrfs_key key;
1626 	int ret;
1627 
1628 	path = btrfs_alloc_path();
1629 	if (!path)
1630 		return -ENOMEM;
1631 
1632 	key.objectid = root->root_key.objectid;
1633 	key.type = BTRFS_ROOT_REF_KEY;
1634 	key.offset = (u64)-1;
1635 
1636 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1637 				&key, path, 0, 0);
1638 	if (ret < 0)
1639 		goto out;
1640 	BUG_ON(ret == 0);
1641 
1642 	ret = 0;
1643 	if (path->slots[0] > 0) {
1644 		path->slots[0]--;
1645 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1646 		if (key.objectid == root->root_key.objectid &&
1647 		    key.type == BTRFS_ROOT_REF_KEY)
1648 			ret = -ENOTEMPTY;
1649 	}
1650 out:
1651 	btrfs_free_path(path);
1652 	return ret;
1653 }
1654 
1655 static noinline int key_in_sk(struct btrfs_key *key,
1656 			      struct btrfs_ioctl_search_key *sk)
1657 {
1658 	struct btrfs_key test;
1659 	int ret;
1660 
1661 	test.objectid = sk->min_objectid;
1662 	test.type = sk->min_type;
1663 	test.offset = sk->min_offset;
1664 
1665 	ret = btrfs_comp_cpu_keys(key, &test);
1666 	if (ret < 0)
1667 		return 0;
1668 
1669 	test.objectid = sk->max_objectid;
1670 	test.type = sk->max_type;
1671 	test.offset = sk->max_offset;
1672 
1673 	ret = btrfs_comp_cpu_keys(key, &test);
1674 	if (ret > 0)
1675 		return 0;
1676 	return 1;
1677 }
1678 
1679 static noinline int copy_to_sk(struct btrfs_root *root,
1680 			       struct btrfs_path *path,
1681 			       struct btrfs_key *key,
1682 			       struct btrfs_ioctl_search_key *sk,
1683 			       char *buf,
1684 			       unsigned long *sk_offset,
1685 			       int *num_found)
1686 {
1687 	u64 found_transid;
1688 	struct extent_buffer *leaf;
1689 	struct btrfs_ioctl_search_header sh;
1690 	unsigned long item_off;
1691 	unsigned long item_len;
1692 	int nritems;
1693 	int i;
1694 	int slot;
1695 	int ret = 0;
1696 
1697 	leaf = path->nodes[0];
1698 	slot = path->slots[0];
1699 	nritems = btrfs_header_nritems(leaf);
1700 
1701 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1702 		i = nritems;
1703 		goto advance_key;
1704 	}
1705 	found_transid = btrfs_header_generation(leaf);
1706 
1707 	for (i = slot; i < nritems; i++) {
1708 		item_off = btrfs_item_ptr_offset(leaf, i);
1709 		item_len = btrfs_item_size_nr(leaf, i);
1710 
1711 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1712 			item_len = 0;
1713 
1714 		if (sizeof(sh) + item_len + *sk_offset >
1715 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1716 			ret = 1;
1717 			goto overflow;
1718 		}
1719 
1720 		btrfs_item_key_to_cpu(leaf, key, i);
1721 		if (!key_in_sk(key, sk))
1722 			continue;
1723 
1724 		sh.objectid = key->objectid;
1725 		sh.offset = key->offset;
1726 		sh.type = key->type;
1727 		sh.len = item_len;
1728 		sh.transid = found_transid;
1729 
1730 		/* copy search result header */
1731 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1732 		*sk_offset += sizeof(sh);
1733 
1734 		if (item_len) {
1735 			char *p = buf + *sk_offset;
1736 			/* copy the item */
1737 			read_extent_buffer(leaf, p,
1738 					   item_off, item_len);
1739 			*sk_offset += item_len;
1740 		}
1741 		(*num_found)++;
1742 
1743 		if (*num_found >= sk->nr_items)
1744 			break;
1745 	}
1746 advance_key:
1747 	ret = 0;
1748 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1749 		key->offset++;
1750 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1751 		key->offset = 0;
1752 		key->type++;
1753 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1754 		key->offset = 0;
1755 		key->type = 0;
1756 		key->objectid++;
1757 	} else
1758 		ret = 1;
1759 overflow:
1760 	return ret;
1761 }
1762 
1763 static noinline int search_ioctl(struct inode *inode,
1764 				 struct btrfs_ioctl_search_args *args)
1765 {
1766 	struct btrfs_root *root;
1767 	struct btrfs_key key;
1768 	struct btrfs_key max_key;
1769 	struct btrfs_path *path;
1770 	struct btrfs_ioctl_search_key *sk = &args->key;
1771 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1772 	int ret;
1773 	int num_found = 0;
1774 	unsigned long sk_offset = 0;
1775 
1776 	path = btrfs_alloc_path();
1777 	if (!path)
1778 		return -ENOMEM;
1779 
1780 	if (sk->tree_id == 0) {
1781 		/* search the root of the inode that was passed */
1782 		root = BTRFS_I(inode)->root;
1783 	} else {
1784 		key.objectid = sk->tree_id;
1785 		key.type = BTRFS_ROOT_ITEM_KEY;
1786 		key.offset = (u64)-1;
1787 		root = btrfs_read_fs_root_no_name(info, &key);
1788 		if (IS_ERR(root)) {
1789 			printk(KERN_ERR "could not find root %llu\n",
1790 			       sk->tree_id);
1791 			btrfs_free_path(path);
1792 			return -ENOENT;
1793 		}
1794 	}
1795 
1796 	key.objectid = sk->min_objectid;
1797 	key.type = sk->min_type;
1798 	key.offset = sk->min_offset;
1799 
1800 	max_key.objectid = sk->max_objectid;
1801 	max_key.type = sk->max_type;
1802 	max_key.offset = sk->max_offset;
1803 
1804 	path->keep_locks = 1;
1805 
1806 	while(1) {
1807 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1808 					   sk->min_transid);
1809 		if (ret != 0) {
1810 			if (ret > 0)
1811 				ret = 0;
1812 			goto err;
1813 		}
1814 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1815 				 &sk_offset, &num_found);
1816 		btrfs_release_path(path);
1817 		if (ret || num_found >= sk->nr_items)
1818 			break;
1819 
1820 	}
1821 	ret = 0;
1822 err:
1823 	sk->nr_items = num_found;
1824 	btrfs_free_path(path);
1825 	return ret;
1826 }
1827 
1828 static noinline int btrfs_ioctl_tree_search(struct file *file,
1829 					   void __user *argp)
1830 {
1831 	 struct btrfs_ioctl_search_args *args;
1832 	 struct inode *inode;
1833 	 int ret;
1834 
1835 	if (!capable(CAP_SYS_ADMIN))
1836 		return -EPERM;
1837 
1838 	args = memdup_user(argp, sizeof(*args));
1839 	if (IS_ERR(args))
1840 		return PTR_ERR(args);
1841 
1842 	inode = fdentry(file)->d_inode;
1843 	ret = search_ioctl(inode, args);
1844 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1845 		ret = -EFAULT;
1846 	kfree(args);
1847 	return ret;
1848 }
1849 
1850 /*
1851  * Search INODE_REFs to identify path name of 'dirid' directory
1852  * in a 'tree_id' tree. and sets path name to 'name'.
1853  */
1854 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1855 				u64 tree_id, u64 dirid, char *name)
1856 {
1857 	struct btrfs_root *root;
1858 	struct btrfs_key key;
1859 	char *ptr;
1860 	int ret = -1;
1861 	int slot;
1862 	int len;
1863 	int total_len = 0;
1864 	struct btrfs_inode_ref *iref;
1865 	struct extent_buffer *l;
1866 	struct btrfs_path *path;
1867 
1868 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1869 		name[0]='\0';
1870 		return 0;
1871 	}
1872 
1873 	path = btrfs_alloc_path();
1874 	if (!path)
1875 		return -ENOMEM;
1876 
1877 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1878 
1879 	key.objectid = tree_id;
1880 	key.type = BTRFS_ROOT_ITEM_KEY;
1881 	key.offset = (u64)-1;
1882 	root = btrfs_read_fs_root_no_name(info, &key);
1883 	if (IS_ERR(root)) {
1884 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1885 		ret = -ENOENT;
1886 		goto out;
1887 	}
1888 
1889 	key.objectid = dirid;
1890 	key.type = BTRFS_INODE_REF_KEY;
1891 	key.offset = (u64)-1;
1892 
1893 	while(1) {
1894 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1895 		if (ret < 0)
1896 			goto out;
1897 
1898 		l = path->nodes[0];
1899 		slot = path->slots[0];
1900 		if (ret > 0 && slot > 0)
1901 			slot--;
1902 		btrfs_item_key_to_cpu(l, &key, slot);
1903 
1904 		if (ret > 0 && (key.objectid != dirid ||
1905 				key.type != BTRFS_INODE_REF_KEY)) {
1906 			ret = -ENOENT;
1907 			goto out;
1908 		}
1909 
1910 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1911 		len = btrfs_inode_ref_name_len(l, iref);
1912 		ptr -= len + 1;
1913 		total_len += len + 1;
1914 		if (ptr < name)
1915 			goto out;
1916 
1917 		*(ptr + len) = '/';
1918 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1919 
1920 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1921 			break;
1922 
1923 		btrfs_release_path(path);
1924 		key.objectid = key.offset;
1925 		key.offset = (u64)-1;
1926 		dirid = key.objectid;
1927 	}
1928 	if (ptr < name)
1929 		goto out;
1930 	memmove(name, ptr, total_len);
1931 	name[total_len]='\0';
1932 	ret = 0;
1933 out:
1934 	btrfs_free_path(path);
1935 	return ret;
1936 }
1937 
1938 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1939 					   void __user *argp)
1940 {
1941 	 struct btrfs_ioctl_ino_lookup_args *args;
1942 	 struct inode *inode;
1943 	 int ret;
1944 
1945 	if (!capable(CAP_SYS_ADMIN))
1946 		return -EPERM;
1947 
1948 	args = memdup_user(argp, sizeof(*args));
1949 	if (IS_ERR(args))
1950 		return PTR_ERR(args);
1951 
1952 	inode = fdentry(file)->d_inode;
1953 
1954 	if (args->treeid == 0)
1955 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1956 
1957 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1958 					args->treeid, args->objectid,
1959 					args->name);
1960 
1961 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1962 		ret = -EFAULT;
1963 
1964 	kfree(args);
1965 	return ret;
1966 }
1967 
1968 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1969 					     void __user *arg)
1970 {
1971 	struct dentry *parent = fdentry(file);
1972 	struct dentry *dentry;
1973 	struct inode *dir = parent->d_inode;
1974 	struct inode *inode;
1975 	struct btrfs_root *root = BTRFS_I(dir)->root;
1976 	struct btrfs_root *dest = NULL;
1977 	struct btrfs_ioctl_vol_args *vol_args;
1978 	struct btrfs_trans_handle *trans;
1979 	int namelen;
1980 	int ret;
1981 	int err = 0;
1982 
1983 	vol_args = memdup_user(arg, sizeof(*vol_args));
1984 	if (IS_ERR(vol_args))
1985 		return PTR_ERR(vol_args);
1986 
1987 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1988 	namelen = strlen(vol_args->name);
1989 	if (strchr(vol_args->name, '/') ||
1990 	    strncmp(vol_args->name, "..", namelen) == 0) {
1991 		err = -EINVAL;
1992 		goto out;
1993 	}
1994 
1995 	err = mnt_want_write_file(file);
1996 	if (err)
1997 		goto out;
1998 
1999 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2000 	dentry = lookup_one_len(vol_args->name, parent, namelen);
2001 	if (IS_ERR(dentry)) {
2002 		err = PTR_ERR(dentry);
2003 		goto out_unlock_dir;
2004 	}
2005 
2006 	if (!dentry->d_inode) {
2007 		err = -ENOENT;
2008 		goto out_dput;
2009 	}
2010 
2011 	inode = dentry->d_inode;
2012 	dest = BTRFS_I(inode)->root;
2013 	if (!capable(CAP_SYS_ADMIN)){
2014 		/*
2015 		 * Regular user.  Only allow this with a special mount
2016 		 * option, when the user has write+exec access to the
2017 		 * subvol root, and when rmdir(2) would have been
2018 		 * allowed.
2019 		 *
2020 		 * Note that this is _not_ check that the subvol is
2021 		 * empty or doesn't contain data that we wouldn't
2022 		 * otherwise be able to delete.
2023 		 *
2024 		 * Users who want to delete empty subvols should try
2025 		 * rmdir(2).
2026 		 */
2027 		err = -EPERM;
2028 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2029 			goto out_dput;
2030 
2031 		/*
2032 		 * Do not allow deletion if the parent dir is the same
2033 		 * as the dir to be deleted.  That means the ioctl
2034 		 * must be called on the dentry referencing the root
2035 		 * of the subvol, not a random directory contained
2036 		 * within it.
2037 		 */
2038 		err = -EINVAL;
2039 		if (root == dest)
2040 			goto out_dput;
2041 
2042 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2043 		if (err)
2044 			goto out_dput;
2045 
2046 		/* check if subvolume may be deleted by a non-root user */
2047 		err = btrfs_may_delete(dir, dentry, 1);
2048 		if (err)
2049 			goto out_dput;
2050 	}
2051 
2052 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2053 		err = -EINVAL;
2054 		goto out_dput;
2055 	}
2056 
2057 	mutex_lock(&inode->i_mutex);
2058 	err = d_invalidate(dentry);
2059 	if (err)
2060 		goto out_unlock;
2061 
2062 	down_write(&root->fs_info->subvol_sem);
2063 
2064 	err = may_destroy_subvol(dest);
2065 	if (err)
2066 		goto out_up_write;
2067 
2068 	trans = btrfs_start_transaction(root, 0);
2069 	if (IS_ERR(trans)) {
2070 		err = PTR_ERR(trans);
2071 		goto out_up_write;
2072 	}
2073 	trans->block_rsv = &root->fs_info->global_block_rsv;
2074 
2075 	ret = btrfs_unlink_subvol(trans, root, dir,
2076 				dest->root_key.objectid,
2077 				dentry->d_name.name,
2078 				dentry->d_name.len);
2079 	if (ret) {
2080 		err = ret;
2081 		btrfs_abort_transaction(trans, root, ret);
2082 		goto out_end_trans;
2083 	}
2084 
2085 	btrfs_record_root_in_trans(trans, dest);
2086 
2087 	memset(&dest->root_item.drop_progress, 0,
2088 		sizeof(dest->root_item.drop_progress));
2089 	dest->root_item.drop_level = 0;
2090 	btrfs_set_root_refs(&dest->root_item, 0);
2091 
2092 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2093 		ret = btrfs_insert_orphan_item(trans,
2094 					root->fs_info->tree_root,
2095 					dest->root_key.objectid);
2096 		if (ret) {
2097 			btrfs_abort_transaction(trans, root, ret);
2098 			err = ret;
2099 			goto out_end_trans;
2100 		}
2101 	}
2102 out_end_trans:
2103 	ret = btrfs_end_transaction(trans, root);
2104 	if (ret && !err)
2105 		err = ret;
2106 	inode->i_flags |= S_DEAD;
2107 out_up_write:
2108 	up_write(&root->fs_info->subvol_sem);
2109 out_unlock:
2110 	mutex_unlock(&inode->i_mutex);
2111 	if (!err) {
2112 		shrink_dcache_sb(root->fs_info->sb);
2113 		btrfs_invalidate_inodes(dest);
2114 		d_delete(dentry);
2115 	}
2116 out_dput:
2117 	dput(dentry);
2118 out_unlock_dir:
2119 	mutex_unlock(&dir->i_mutex);
2120 	mnt_drop_write_file(file);
2121 out:
2122 	kfree(vol_args);
2123 	return err;
2124 }
2125 
2126 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2127 {
2128 	struct inode *inode = fdentry(file)->d_inode;
2129 	struct btrfs_root *root = BTRFS_I(inode)->root;
2130 	struct btrfs_ioctl_defrag_range_args *range;
2131 	int ret;
2132 
2133 	if (btrfs_root_readonly(root))
2134 		return -EROFS;
2135 
2136 	ret = mnt_want_write_file(file);
2137 	if (ret)
2138 		return ret;
2139 
2140 	switch (inode->i_mode & S_IFMT) {
2141 	case S_IFDIR:
2142 		if (!capable(CAP_SYS_ADMIN)) {
2143 			ret = -EPERM;
2144 			goto out;
2145 		}
2146 		ret = btrfs_defrag_root(root, 0);
2147 		if (ret)
2148 			goto out;
2149 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2150 		break;
2151 	case S_IFREG:
2152 		if (!(file->f_mode & FMODE_WRITE)) {
2153 			ret = -EINVAL;
2154 			goto out;
2155 		}
2156 
2157 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2158 		if (!range) {
2159 			ret = -ENOMEM;
2160 			goto out;
2161 		}
2162 
2163 		if (argp) {
2164 			if (copy_from_user(range, argp,
2165 					   sizeof(*range))) {
2166 				ret = -EFAULT;
2167 				kfree(range);
2168 				goto out;
2169 			}
2170 			/* compression requires us to start the IO */
2171 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2172 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2173 				range->extent_thresh = (u32)-1;
2174 			}
2175 		} else {
2176 			/* the rest are all set to zero by kzalloc */
2177 			range->len = (u64)-1;
2178 		}
2179 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2180 					range, 0, 0);
2181 		if (ret > 0)
2182 			ret = 0;
2183 		kfree(range);
2184 		break;
2185 	default:
2186 		ret = -EINVAL;
2187 	}
2188 out:
2189 	mnt_drop_write_file(file);
2190 	return ret;
2191 }
2192 
2193 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2194 {
2195 	struct btrfs_ioctl_vol_args *vol_args;
2196 	int ret;
2197 
2198 	if (!capable(CAP_SYS_ADMIN))
2199 		return -EPERM;
2200 
2201 	mutex_lock(&root->fs_info->volume_mutex);
2202 	if (root->fs_info->balance_ctl) {
2203 		printk(KERN_INFO "btrfs: balance in progress\n");
2204 		ret = -EINVAL;
2205 		goto out;
2206 	}
2207 
2208 	vol_args = memdup_user(arg, sizeof(*vol_args));
2209 	if (IS_ERR(vol_args)) {
2210 		ret = PTR_ERR(vol_args);
2211 		goto out;
2212 	}
2213 
2214 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2215 	ret = btrfs_init_new_device(root, vol_args->name);
2216 
2217 	kfree(vol_args);
2218 out:
2219 	mutex_unlock(&root->fs_info->volume_mutex);
2220 	return ret;
2221 }
2222 
2223 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2224 {
2225 	struct btrfs_ioctl_vol_args *vol_args;
2226 	int ret;
2227 
2228 	if (!capable(CAP_SYS_ADMIN))
2229 		return -EPERM;
2230 
2231 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2232 		return -EROFS;
2233 
2234 	mutex_lock(&root->fs_info->volume_mutex);
2235 	if (root->fs_info->balance_ctl) {
2236 		printk(KERN_INFO "btrfs: balance in progress\n");
2237 		ret = -EINVAL;
2238 		goto out;
2239 	}
2240 
2241 	vol_args = memdup_user(arg, sizeof(*vol_args));
2242 	if (IS_ERR(vol_args)) {
2243 		ret = PTR_ERR(vol_args);
2244 		goto out;
2245 	}
2246 
2247 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2248 	ret = btrfs_rm_device(root, vol_args->name);
2249 
2250 	kfree(vol_args);
2251 out:
2252 	mutex_unlock(&root->fs_info->volume_mutex);
2253 	return ret;
2254 }
2255 
2256 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2257 {
2258 	struct btrfs_ioctl_fs_info_args *fi_args;
2259 	struct btrfs_device *device;
2260 	struct btrfs_device *next;
2261 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2262 	int ret = 0;
2263 
2264 	if (!capable(CAP_SYS_ADMIN))
2265 		return -EPERM;
2266 
2267 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2268 	if (!fi_args)
2269 		return -ENOMEM;
2270 
2271 	fi_args->num_devices = fs_devices->num_devices;
2272 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2273 
2274 	mutex_lock(&fs_devices->device_list_mutex);
2275 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2276 		if (device->devid > fi_args->max_id)
2277 			fi_args->max_id = device->devid;
2278 	}
2279 	mutex_unlock(&fs_devices->device_list_mutex);
2280 
2281 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2282 		ret = -EFAULT;
2283 
2284 	kfree(fi_args);
2285 	return ret;
2286 }
2287 
2288 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2289 {
2290 	struct btrfs_ioctl_dev_info_args *di_args;
2291 	struct btrfs_device *dev;
2292 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2293 	int ret = 0;
2294 	char *s_uuid = NULL;
2295 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2296 
2297 	if (!capable(CAP_SYS_ADMIN))
2298 		return -EPERM;
2299 
2300 	di_args = memdup_user(arg, sizeof(*di_args));
2301 	if (IS_ERR(di_args))
2302 		return PTR_ERR(di_args);
2303 
2304 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2305 		s_uuid = di_args->uuid;
2306 
2307 	mutex_lock(&fs_devices->device_list_mutex);
2308 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2309 	mutex_unlock(&fs_devices->device_list_mutex);
2310 
2311 	if (!dev) {
2312 		ret = -ENODEV;
2313 		goto out;
2314 	}
2315 
2316 	di_args->devid = dev->devid;
2317 	di_args->bytes_used = dev->bytes_used;
2318 	di_args->total_bytes = dev->total_bytes;
2319 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2320 	if (dev->name) {
2321 		struct rcu_string *name;
2322 
2323 		rcu_read_lock();
2324 		name = rcu_dereference(dev->name);
2325 		strncpy(di_args->path, name->str, sizeof(di_args->path));
2326 		rcu_read_unlock();
2327 		di_args->path[sizeof(di_args->path) - 1] = 0;
2328 	} else {
2329 		di_args->path[0] = '\0';
2330 	}
2331 
2332 out:
2333 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2334 		ret = -EFAULT;
2335 
2336 	kfree(di_args);
2337 	return ret;
2338 }
2339 
2340 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2341 				       u64 off, u64 olen, u64 destoff)
2342 {
2343 	struct inode *inode = fdentry(file)->d_inode;
2344 	struct btrfs_root *root = BTRFS_I(inode)->root;
2345 	struct file *src_file;
2346 	struct inode *src;
2347 	struct btrfs_trans_handle *trans;
2348 	struct btrfs_path *path;
2349 	struct extent_buffer *leaf;
2350 	char *buf;
2351 	struct btrfs_key key;
2352 	u32 nritems;
2353 	int slot;
2354 	int ret;
2355 	u64 len = olen;
2356 	u64 bs = root->fs_info->sb->s_blocksize;
2357 
2358 	/*
2359 	 * TODO:
2360 	 * - split compressed inline extents.  annoying: we need to
2361 	 *   decompress into destination's address_space (the file offset
2362 	 *   may change, so source mapping won't do), then recompress (or
2363 	 *   otherwise reinsert) a subrange.
2364 	 * - allow ranges within the same file to be cloned (provided
2365 	 *   they don't overlap)?
2366 	 */
2367 
2368 	/* the destination must be opened for writing */
2369 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2370 		return -EINVAL;
2371 
2372 	if (btrfs_root_readonly(root))
2373 		return -EROFS;
2374 
2375 	ret = mnt_want_write_file(file);
2376 	if (ret)
2377 		return ret;
2378 
2379 	src_file = fget(srcfd);
2380 	if (!src_file) {
2381 		ret = -EBADF;
2382 		goto out_drop_write;
2383 	}
2384 
2385 	ret = -EXDEV;
2386 	if (src_file->f_path.mnt != file->f_path.mnt)
2387 		goto out_fput;
2388 
2389 	src = src_file->f_dentry->d_inode;
2390 
2391 	ret = -EINVAL;
2392 	if (src == inode)
2393 		goto out_fput;
2394 
2395 	/* the src must be open for reading */
2396 	if (!(src_file->f_mode & FMODE_READ))
2397 		goto out_fput;
2398 
2399 	/* don't make the dst file partly checksummed */
2400 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2401 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2402 		goto out_fput;
2403 
2404 	ret = -EISDIR;
2405 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2406 		goto out_fput;
2407 
2408 	ret = -EXDEV;
2409 	if (src->i_sb != inode->i_sb)
2410 		goto out_fput;
2411 
2412 	ret = -ENOMEM;
2413 	buf = vmalloc(btrfs_level_size(root, 0));
2414 	if (!buf)
2415 		goto out_fput;
2416 
2417 	path = btrfs_alloc_path();
2418 	if (!path) {
2419 		vfree(buf);
2420 		goto out_fput;
2421 	}
2422 	path->reada = 2;
2423 
2424 	if (inode < src) {
2425 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2426 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2427 	} else {
2428 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2429 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2430 	}
2431 
2432 	/* determine range to clone */
2433 	ret = -EINVAL;
2434 	if (off + len > src->i_size || off + len < off)
2435 		goto out_unlock;
2436 	if (len == 0)
2437 		olen = len = src->i_size - off;
2438 	/* if we extend to eof, continue to block boundary */
2439 	if (off + len == src->i_size)
2440 		len = ALIGN(src->i_size, bs) - off;
2441 
2442 	/* verify the end result is block aligned */
2443 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2444 	    !IS_ALIGNED(destoff, bs))
2445 		goto out_unlock;
2446 
2447 	if (destoff > inode->i_size) {
2448 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2449 		if (ret)
2450 			goto out_unlock;
2451 	}
2452 
2453 	/* truncate page cache pages from target inode range */
2454 	truncate_inode_pages_range(&inode->i_data, destoff,
2455 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2456 
2457 	/* do any pending delalloc/csum calc on src, one way or
2458 	   another, and lock file content */
2459 	while (1) {
2460 		struct btrfs_ordered_extent *ordered;
2461 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2462 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2463 		if (!ordered &&
2464 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2465 				   EXTENT_DELALLOC, 0, NULL))
2466 			break;
2467 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2468 		if (ordered)
2469 			btrfs_put_ordered_extent(ordered);
2470 		btrfs_wait_ordered_range(src, off, len);
2471 	}
2472 
2473 	/* clone data */
2474 	key.objectid = btrfs_ino(src);
2475 	key.type = BTRFS_EXTENT_DATA_KEY;
2476 	key.offset = 0;
2477 
2478 	while (1) {
2479 		/*
2480 		 * note the key will change type as we walk through the
2481 		 * tree.
2482 		 */
2483 		ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2484 				0, 0);
2485 		if (ret < 0)
2486 			goto out;
2487 
2488 		nritems = btrfs_header_nritems(path->nodes[0]);
2489 		if (path->slots[0] >= nritems) {
2490 			ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2491 			if (ret < 0)
2492 				goto out;
2493 			if (ret > 0)
2494 				break;
2495 			nritems = btrfs_header_nritems(path->nodes[0]);
2496 		}
2497 		leaf = path->nodes[0];
2498 		slot = path->slots[0];
2499 
2500 		btrfs_item_key_to_cpu(leaf, &key, slot);
2501 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2502 		    key.objectid != btrfs_ino(src))
2503 			break;
2504 
2505 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2506 			struct btrfs_file_extent_item *extent;
2507 			int type;
2508 			u32 size;
2509 			struct btrfs_key new_key;
2510 			u64 disko = 0, diskl = 0;
2511 			u64 datao = 0, datal = 0;
2512 			u8 comp;
2513 			u64 endoff;
2514 
2515 			size = btrfs_item_size_nr(leaf, slot);
2516 			read_extent_buffer(leaf, buf,
2517 					   btrfs_item_ptr_offset(leaf, slot),
2518 					   size);
2519 
2520 			extent = btrfs_item_ptr(leaf, slot,
2521 						struct btrfs_file_extent_item);
2522 			comp = btrfs_file_extent_compression(leaf, extent);
2523 			type = btrfs_file_extent_type(leaf, extent);
2524 			if (type == BTRFS_FILE_EXTENT_REG ||
2525 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2526 				disko = btrfs_file_extent_disk_bytenr(leaf,
2527 								      extent);
2528 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2529 								 extent);
2530 				datao = btrfs_file_extent_offset(leaf, extent);
2531 				datal = btrfs_file_extent_num_bytes(leaf,
2532 								    extent);
2533 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2534 				/* take upper bound, may be compressed */
2535 				datal = btrfs_file_extent_ram_bytes(leaf,
2536 								    extent);
2537 			}
2538 			btrfs_release_path(path);
2539 
2540 			if (key.offset + datal <= off ||
2541 			    key.offset >= off+len)
2542 				goto next;
2543 
2544 			memcpy(&new_key, &key, sizeof(new_key));
2545 			new_key.objectid = btrfs_ino(inode);
2546 			if (off <= key.offset)
2547 				new_key.offset = key.offset + destoff - off;
2548 			else
2549 				new_key.offset = destoff;
2550 
2551 			/*
2552 			 * 1 - adjusting old extent (we may have to split it)
2553 			 * 1 - add new extent
2554 			 * 1 - inode update
2555 			 */
2556 			trans = btrfs_start_transaction(root, 3);
2557 			if (IS_ERR(trans)) {
2558 				ret = PTR_ERR(trans);
2559 				goto out;
2560 			}
2561 
2562 			if (type == BTRFS_FILE_EXTENT_REG ||
2563 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2564 				/*
2565 				 *    a  | --- range to clone ---|  b
2566 				 * | ------------- extent ------------- |
2567 				 */
2568 
2569 				/* substract range b */
2570 				if (key.offset + datal > off + len)
2571 					datal = off + len - key.offset;
2572 
2573 				/* substract range a */
2574 				if (off > key.offset) {
2575 					datao += off - key.offset;
2576 					datal -= off - key.offset;
2577 				}
2578 
2579 				ret = btrfs_drop_extents(trans, root, inode,
2580 							 new_key.offset,
2581 							 new_key.offset + datal,
2582 							 1);
2583 				if (ret) {
2584 					btrfs_abort_transaction(trans, root,
2585 								ret);
2586 					btrfs_end_transaction(trans, root);
2587 					goto out;
2588 				}
2589 
2590 				ret = btrfs_insert_empty_item(trans, root, path,
2591 							      &new_key, size);
2592 				if (ret) {
2593 					btrfs_abort_transaction(trans, root,
2594 								ret);
2595 					btrfs_end_transaction(trans, root);
2596 					goto out;
2597 				}
2598 
2599 				leaf = path->nodes[0];
2600 				slot = path->slots[0];
2601 				write_extent_buffer(leaf, buf,
2602 					    btrfs_item_ptr_offset(leaf, slot),
2603 					    size);
2604 
2605 				extent = btrfs_item_ptr(leaf, slot,
2606 						struct btrfs_file_extent_item);
2607 
2608 				/* disko == 0 means it's a hole */
2609 				if (!disko)
2610 					datao = 0;
2611 
2612 				btrfs_set_file_extent_offset(leaf, extent,
2613 							     datao);
2614 				btrfs_set_file_extent_num_bytes(leaf, extent,
2615 								datal);
2616 				if (disko) {
2617 					inode_add_bytes(inode, datal);
2618 					ret = btrfs_inc_extent_ref(trans, root,
2619 							disko, diskl, 0,
2620 							root->root_key.objectid,
2621 							btrfs_ino(inode),
2622 							new_key.offset - datao,
2623 							0);
2624 					if (ret) {
2625 						btrfs_abort_transaction(trans,
2626 									root,
2627 									ret);
2628 						btrfs_end_transaction(trans,
2629 								      root);
2630 						goto out;
2631 
2632 					}
2633 				}
2634 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2635 				u64 skip = 0;
2636 				u64 trim = 0;
2637 				if (off > key.offset) {
2638 					skip = off - key.offset;
2639 					new_key.offset += skip;
2640 				}
2641 
2642 				if (key.offset + datal > off+len)
2643 					trim = key.offset + datal - (off+len);
2644 
2645 				if (comp && (skip || trim)) {
2646 					ret = -EINVAL;
2647 					btrfs_end_transaction(trans, root);
2648 					goto out;
2649 				}
2650 				size -= skip + trim;
2651 				datal -= skip + trim;
2652 
2653 				ret = btrfs_drop_extents(trans, root, inode,
2654 							 new_key.offset,
2655 							 new_key.offset + datal,
2656 							 1);
2657 				if (ret) {
2658 					btrfs_abort_transaction(trans, root,
2659 								ret);
2660 					btrfs_end_transaction(trans, root);
2661 					goto out;
2662 				}
2663 
2664 				ret = btrfs_insert_empty_item(trans, root, path,
2665 							      &new_key, size);
2666 				if (ret) {
2667 					btrfs_abort_transaction(trans, root,
2668 								ret);
2669 					btrfs_end_transaction(trans, root);
2670 					goto out;
2671 				}
2672 
2673 				if (skip) {
2674 					u32 start =
2675 					  btrfs_file_extent_calc_inline_size(0);
2676 					memmove(buf+start, buf+start+skip,
2677 						datal);
2678 				}
2679 
2680 				leaf = path->nodes[0];
2681 				slot = path->slots[0];
2682 				write_extent_buffer(leaf, buf,
2683 					    btrfs_item_ptr_offset(leaf, slot),
2684 					    size);
2685 				inode_add_bytes(inode, datal);
2686 			}
2687 
2688 			btrfs_mark_buffer_dirty(leaf);
2689 			btrfs_release_path(path);
2690 
2691 			inode_inc_iversion(inode);
2692 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2693 
2694 			/*
2695 			 * we round up to the block size at eof when
2696 			 * determining which extents to clone above,
2697 			 * but shouldn't round up the file size
2698 			 */
2699 			endoff = new_key.offset + datal;
2700 			if (endoff > destoff+olen)
2701 				endoff = destoff+olen;
2702 			if (endoff > inode->i_size)
2703 				btrfs_i_size_write(inode, endoff);
2704 
2705 			ret = btrfs_update_inode(trans, root, inode);
2706 			if (ret) {
2707 				btrfs_abort_transaction(trans, root, ret);
2708 				btrfs_end_transaction(trans, root);
2709 				goto out;
2710 			}
2711 			ret = btrfs_end_transaction(trans, root);
2712 		}
2713 next:
2714 		btrfs_release_path(path);
2715 		key.offset++;
2716 	}
2717 	ret = 0;
2718 out:
2719 	btrfs_release_path(path);
2720 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2721 out_unlock:
2722 	mutex_unlock(&src->i_mutex);
2723 	mutex_unlock(&inode->i_mutex);
2724 	vfree(buf);
2725 	btrfs_free_path(path);
2726 out_fput:
2727 	fput(src_file);
2728 out_drop_write:
2729 	mnt_drop_write_file(file);
2730 	return ret;
2731 }
2732 
2733 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2734 {
2735 	struct btrfs_ioctl_clone_range_args args;
2736 
2737 	if (copy_from_user(&args, argp, sizeof(args)))
2738 		return -EFAULT;
2739 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2740 				 args.src_length, args.dest_offset);
2741 }
2742 
2743 /*
2744  * there are many ways the trans_start and trans_end ioctls can lead
2745  * to deadlocks.  They should only be used by applications that
2746  * basically own the machine, and have a very in depth understanding
2747  * of all the possible deadlocks and enospc problems.
2748  */
2749 static long btrfs_ioctl_trans_start(struct file *file)
2750 {
2751 	struct inode *inode = fdentry(file)->d_inode;
2752 	struct btrfs_root *root = BTRFS_I(inode)->root;
2753 	struct btrfs_trans_handle *trans;
2754 	int ret;
2755 
2756 	ret = -EPERM;
2757 	if (!capable(CAP_SYS_ADMIN))
2758 		goto out;
2759 
2760 	ret = -EINPROGRESS;
2761 	if (file->private_data)
2762 		goto out;
2763 
2764 	ret = -EROFS;
2765 	if (btrfs_root_readonly(root))
2766 		goto out;
2767 
2768 	ret = mnt_want_write_file(file);
2769 	if (ret)
2770 		goto out;
2771 
2772 	atomic_inc(&root->fs_info->open_ioctl_trans);
2773 
2774 	ret = -ENOMEM;
2775 	trans = btrfs_start_ioctl_transaction(root);
2776 	if (IS_ERR(trans))
2777 		goto out_drop;
2778 
2779 	file->private_data = trans;
2780 	return 0;
2781 
2782 out_drop:
2783 	atomic_dec(&root->fs_info->open_ioctl_trans);
2784 	mnt_drop_write_file(file);
2785 out:
2786 	return ret;
2787 }
2788 
2789 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2790 {
2791 	struct inode *inode = fdentry(file)->d_inode;
2792 	struct btrfs_root *root = BTRFS_I(inode)->root;
2793 	struct btrfs_root *new_root;
2794 	struct btrfs_dir_item *di;
2795 	struct btrfs_trans_handle *trans;
2796 	struct btrfs_path *path;
2797 	struct btrfs_key location;
2798 	struct btrfs_disk_key disk_key;
2799 	u64 objectid = 0;
2800 	u64 dir_id;
2801 
2802 	if (!capable(CAP_SYS_ADMIN))
2803 		return -EPERM;
2804 
2805 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2806 		return -EFAULT;
2807 
2808 	if (!objectid)
2809 		objectid = root->root_key.objectid;
2810 
2811 	location.objectid = objectid;
2812 	location.type = BTRFS_ROOT_ITEM_KEY;
2813 	location.offset = (u64)-1;
2814 
2815 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2816 	if (IS_ERR(new_root))
2817 		return PTR_ERR(new_root);
2818 
2819 	if (btrfs_root_refs(&new_root->root_item) == 0)
2820 		return -ENOENT;
2821 
2822 	path = btrfs_alloc_path();
2823 	if (!path)
2824 		return -ENOMEM;
2825 	path->leave_spinning = 1;
2826 
2827 	trans = btrfs_start_transaction(root, 1);
2828 	if (IS_ERR(trans)) {
2829 		btrfs_free_path(path);
2830 		return PTR_ERR(trans);
2831 	}
2832 
2833 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2834 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2835 				   dir_id, "default", 7, 1);
2836 	if (IS_ERR_OR_NULL(di)) {
2837 		btrfs_free_path(path);
2838 		btrfs_end_transaction(trans, root);
2839 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2840 		       "this isn't going to work\n");
2841 		return -ENOENT;
2842 	}
2843 
2844 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2845 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2846 	btrfs_mark_buffer_dirty(path->nodes[0]);
2847 	btrfs_free_path(path);
2848 
2849 	btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2850 	btrfs_end_transaction(trans, root);
2851 
2852 	return 0;
2853 }
2854 
2855 static void get_block_group_info(struct list_head *groups_list,
2856 				 struct btrfs_ioctl_space_info *space)
2857 {
2858 	struct btrfs_block_group_cache *block_group;
2859 
2860 	space->total_bytes = 0;
2861 	space->used_bytes = 0;
2862 	space->flags = 0;
2863 	list_for_each_entry(block_group, groups_list, list) {
2864 		space->flags = block_group->flags;
2865 		space->total_bytes += block_group->key.offset;
2866 		space->used_bytes +=
2867 			btrfs_block_group_used(&block_group->item);
2868 	}
2869 }
2870 
2871 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2872 {
2873 	struct btrfs_ioctl_space_args space_args;
2874 	struct btrfs_ioctl_space_info space;
2875 	struct btrfs_ioctl_space_info *dest;
2876 	struct btrfs_ioctl_space_info *dest_orig;
2877 	struct btrfs_ioctl_space_info __user *user_dest;
2878 	struct btrfs_space_info *info;
2879 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2880 		       BTRFS_BLOCK_GROUP_SYSTEM,
2881 		       BTRFS_BLOCK_GROUP_METADATA,
2882 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2883 	int num_types = 4;
2884 	int alloc_size;
2885 	int ret = 0;
2886 	u64 slot_count = 0;
2887 	int i, c;
2888 
2889 	if (copy_from_user(&space_args,
2890 			   (struct btrfs_ioctl_space_args __user *)arg,
2891 			   sizeof(space_args)))
2892 		return -EFAULT;
2893 
2894 	for (i = 0; i < num_types; i++) {
2895 		struct btrfs_space_info *tmp;
2896 
2897 		info = NULL;
2898 		rcu_read_lock();
2899 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2900 					list) {
2901 			if (tmp->flags == types[i]) {
2902 				info = tmp;
2903 				break;
2904 			}
2905 		}
2906 		rcu_read_unlock();
2907 
2908 		if (!info)
2909 			continue;
2910 
2911 		down_read(&info->groups_sem);
2912 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2913 			if (!list_empty(&info->block_groups[c]))
2914 				slot_count++;
2915 		}
2916 		up_read(&info->groups_sem);
2917 	}
2918 
2919 	/* space_slots == 0 means they are asking for a count */
2920 	if (space_args.space_slots == 0) {
2921 		space_args.total_spaces = slot_count;
2922 		goto out;
2923 	}
2924 
2925 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2926 
2927 	alloc_size = sizeof(*dest) * slot_count;
2928 
2929 	/* we generally have at most 6 or so space infos, one for each raid
2930 	 * level.  So, a whole page should be more than enough for everyone
2931 	 */
2932 	if (alloc_size > PAGE_CACHE_SIZE)
2933 		return -ENOMEM;
2934 
2935 	space_args.total_spaces = 0;
2936 	dest = kmalloc(alloc_size, GFP_NOFS);
2937 	if (!dest)
2938 		return -ENOMEM;
2939 	dest_orig = dest;
2940 
2941 	/* now we have a buffer to copy into */
2942 	for (i = 0; i < num_types; i++) {
2943 		struct btrfs_space_info *tmp;
2944 
2945 		if (!slot_count)
2946 			break;
2947 
2948 		info = NULL;
2949 		rcu_read_lock();
2950 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2951 					list) {
2952 			if (tmp->flags == types[i]) {
2953 				info = tmp;
2954 				break;
2955 			}
2956 		}
2957 		rcu_read_unlock();
2958 
2959 		if (!info)
2960 			continue;
2961 		down_read(&info->groups_sem);
2962 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2963 			if (!list_empty(&info->block_groups[c])) {
2964 				get_block_group_info(&info->block_groups[c],
2965 						     &space);
2966 				memcpy(dest, &space, sizeof(space));
2967 				dest++;
2968 				space_args.total_spaces++;
2969 				slot_count--;
2970 			}
2971 			if (!slot_count)
2972 				break;
2973 		}
2974 		up_read(&info->groups_sem);
2975 	}
2976 
2977 	user_dest = (struct btrfs_ioctl_space_info __user *)
2978 		(arg + sizeof(struct btrfs_ioctl_space_args));
2979 
2980 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2981 		ret = -EFAULT;
2982 
2983 	kfree(dest_orig);
2984 out:
2985 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2986 		ret = -EFAULT;
2987 
2988 	return ret;
2989 }
2990 
2991 /*
2992  * there are many ways the trans_start and trans_end ioctls can lead
2993  * to deadlocks.  They should only be used by applications that
2994  * basically own the machine, and have a very in depth understanding
2995  * of all the possible deadlocks and enospc problems.
2996  */
2997 long btrfs_ioctl_trans_end(struct file *file)
2998 {
2999 	struct inode *inode = fdentry(file)->d_inode;
3000 	struct btrfs_root *root = BTRFS_I(inode)->root;
3001 	struct btrfs_trans_handle *trans;
3002 
3003 	trans = file->private_data;
3004 	if (!trans)
3005 		return -EINVAL;
3006 	file->private_data = NULL;
3007 
3008 	btrfs_end_transaction(trans, root);
3009 
3010 	atomic_dec(&root->fs_info->open_ioctl_trans);
3011 
3012 	mnt_drop_write_file(file);
3013 	return 0;
3014 }
3015 
3016 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
3017 {
3018 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3019 	struct btrfs_trans_handle *trans;
3020 	u64 transid;
3021 	int ret;
3022 
3023 	trans = btrfs_start_transaction(root, 0);
3024 	if (IS_ERR(trans))
3025 		return PTR_ERR(trans);
3026 	transid = trans->transid;
3027 	ret = btrfs_commit_transaction_async(trans, root, 0);
3028 	if (ret) {
3029 		btrfs_end_transaction(trans, root);
3030 		return ret;
3031 	}
3032 
3033 	if (argp)
3034 		if (copy_to_user(argp, &transid, sizeof(transid)))
3035 			return -EFAULT;
3036 	return 0;
3037 }
3038 
3039 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3040 {
3041 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3042 	u64 transid;
3043 
3044 	if (argp) {
3045 		if (copy_from_user(&transid, argp, sizeof(transid)))
3046 			return -EFAULT;
3047 	} else {
3048 		transid = 0;  /* current trans */
3049 	}
3050 	return btrfs_wait_for_commit(root, transid);
3051 }
3052 
3053 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3054 {
3055 	int ret;
3056 	struct btrfs_ioctl_scrub_args *sa;
3057 
3058 	if (!capable(CAP_SYS_ADMIN))
3059 		return -EPERM;
3060 
3061 	sa = memdup_user(arg, sizeof(*sa));
3062 	if (IS_ERR(sa))
3063 		return PTR_ERR(sa);
3064 
3065 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3066 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3067 
3068 	if (copy_to_user(arg, sa, sizeof(*sa)))
3069 		ret = -EFAULT;
3070 
3071 	kfree(sa);
3072 	return ret;
3073 }
3074 
3075 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3076 {
3077 	if (!capable(CAP_SYS_ADMIN))
3078 		return -EPERM;
3079 
3080 	return btrfs_scrub_cancel(root);
3081 }
3082 
3083 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3084 				       void __user *arg)
3085 {
3086 	struct btrfs_ioctl_scrub_args *sa;
3087 	int ret;
3088 
3089 	if (!capable(CAP_SYS_ADMIN))
3090 		return -EPERM;
3091 
3092 	sa = memdup_user(arg, sizeof(*sa));
3093 	if (IS_ERR(sa))
3094 		return PTR_ERR(sa);
3095 
3096 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3097 
3098 	if (copy_to_user(arg, sa, sizeof(*sa)))
3099 		ret = -EFAULT;
3100 
3101 	kfree(sa);
3102 	return ret;
3103 }
3104 
3105 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3106 				      void __user *arg)
3107 {
3108 	struct btrfs_ioctl_get_dev_stats *sa;
3109 	int ret;
3110 
3111 	sa = memdup_user(arg, sizeof(*sa));
3112 	if (IS_ERR(sa))
3113 		return PTR_ERR(sa);
3114 
3115 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3116 		kfree(sa);
3117 		return -EPERM;
3118 	}
3119 
3120 	ret = btrfs_get_dev_stats(root, sa);
3121 
3122 	if (copy_to_user(arg, sa, sizeof(*sa)))
3123 		ret = -EFAULT;
3124 
3125 	kfree(sa);
3126 	return ret;
3127 }
3128 
3129 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3130 {
3131 	int ret = 0;
3132 	int i;
3133 	u64 rel_ptr;
3134 	int size;
3135 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3136 	struct inode_fs_paths *ipath = NULL;
3137 	struct btrfs_path *path;
3138 
3139 	if (!capable(CAP_SYS_ADMIN))
3140 		return -EPERM;
3141 
3142 	path = btrfs_alloc_path();
3143 	if (!path) {
3144 		ret = -ENOMEM;
3145 		goto out;
3146 	}
3147 
3148 	ipa = memdup_user(arg, sizeof(*ipa));
3149 	if (IS_ERR(ipa)) {
3150 		ret = PTR_ERR(ipa);
3151 		ipa = NULL;
3152 		goto out;
3153 	}
3154 
3155 	size = min_t(u32, ipa->size, 4096);
3156 	ipath = init_ipath(size, root, path);
3157 	if (IS_ERR(ipath)) {
3158 		ret = PTR_ERR(ipath);
3159 		ipath = NULL;
3160 		goto out;
3161 	}
3162 
3163 	ret = paths_from_inode(ipa->inum, ipath);
3164 	if (ret < 0)
3165 		goto out;
3166 
3167 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3168 		rel_ptr = ipath->fspath->val[i] -
3169 			  (u64)(unsigned long)ipath->fspath->val;
3170 		ipath->fspath->val[i] = rel_ptr;
3171 	}
3172 
3173 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3174 			   (void *)(unsigned long)ipath->fspath, size);
3175 	if (ret) {
3176 		ret = -EFAULT;
3177 		goto out;
3178 	}
3179 
3180 out:
3181 	btrfs_free_path(path);
3182 	free_ipath(ipath);
3183 	kfree(ipa);
3184 
3185 	return ret;
3186 }
3187 
3188 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3189 {
3190 	struct btrfs_data_container *inodes = ctx;
3191 	const size_t c = 3 * sizeof(u64);
3192 
3193 	if (inodes->bytes_left >= c) {
3194 		inodes->bytes_left -= c;
3195 		inodes->val[inodes->elem_cnt] = inum;
3196 		inodes->val[inodes->elem_cnt + 1] = offset;
3197 		inodes->val[inodes->elem_cnt + 2] = root;
3198 		inodes->elem_cnt += 3;
3199 	} else {
3200 		inodes->bytes_missing += c - inodes->bytes_left;
3201 		inodes->bytes_left = 0;
3202 		inodes->elem_missed += 3;
3203 	}
3204 
3205 	return 0;
3206 }
3207 
3208 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3209 					void __user *arg)
3210 {
3211 	int ret = 0;
3212 	int size;
3213 	struct btrfs_ioctl_logical_ino_args *loi;
3214 	struct btrfs_data_container *inodes = NULL;
3215 	struct btrfs_path *path = NULL;
3216 
3217 	if (!capable(CAP_SYS_ADMIN))
3218 		return -EPERM;
3219 
3220 	loi = memdup_user(arg, sizeof(*loi));
3221 	if (IS_ERR(loi)) {
3222 		ret = PTR_ERR(loi);
3223 		loi = NULL;
3224 		goto out;
3225 	}
3226 
3227 	path = btrfs_alloc_path();
3228 	if (!path) {
3229 		ret = -ENOMEM;
3230 		goto out;
3231 	}
3232 
3233 	size = min_t(u32, loi->size, 64 * 1024);
3234 	inodes = init_data_container(size);
3235 	if (IS_ERR(inodes)) {
3236 		ret = PTR_ERR(inodes);
3237 		inodes = NULL;
3238 		goto out;
3239 	}
3240 
3241 	ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3242 					  build_ino_list, inodes);
3243 	if (ret == -EINVAL)
3244 		ret = -ENOENT;
3245 	if (ret < 0)
3246 		goto out;
3247 
3248 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3249 			   (void *)(unsigned long)inodes, size);
3250 	if (ret)
3251 		ret = -EFAULT;
3252 
3253 out:
3254 	btrfs_free_path(path);
3255 	vfree(inodes);
3256 	kfree(loi);
3257 
3258 	return ret;
3259 }
3260 
3261 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3262 			       struct btrfs_ioctl_balance_args *bargs)
3263 {
3264 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3265 
3266 	bargs->flags = bctl->flags;
3267 
3268 	if (atomic_read(&fs_info->balance_running))
3269 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3270 	if (atomic_read(&fs_info->balance_pause_req))
3271 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3272 	if (atomic_read(&fs_info->balance_cancel_req))
3273 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3274 
3275 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3276 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3277 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3278 
3279 	if (lock) {
3280 		spin_lock(&fs_info->balance_lock);
3281 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3282 		spin_unlock(&fs_info->balance_lock);
3283 	} else {
3284 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3285 	}
3286 }
3287 
3288 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3289 {
3290 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3291 	struct btrfs_fs_info *fs_info = root->fs_info;
3292 	struct btrfs_ioctl_balance_args *bargs;
3293 	struct btrfs_balance_control *bctl;
3294 	int ret;
3295 
3296 	if (!capable(CAP_SYS_ADMIN))
3297 		return -EPERM;
3298 
3299 	ret = mnt_want_write_file(file);
3300 	if (ret)
3301 		return ret;
3302 
3303 	mutex_lock(&fs_info->volume_mutex);
3304 	mutex_lock(&fs_info->balance_mutex);
3305 
3306 	if (arg) {
3307 		bargs = memdup_user(arg, sizeof(*bargs));
3308 		if (IS_ERR(bargs)) {
3309 			ret = PTR_ERR(bargs);
3310 			goto out;
3311 		}
3312 
3313 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3314 			if (!fs_info->balance_ctl) {
3315 				ret = -ENOTCONN;
3316 				goto out_bargs;
3317 			}
3318 
3319 			bctl = fs_info->balance_ctl;
3320 			spin_lock(&fs_info->balance_lock);
3321 			bctl->flags |= BTRFS_BALANCE_RESUME;
3322 			spin_unlock(&fs_info->balance_lock);
3323 
3324 			goto do_balance;
3325 		}
3326 	} else {
3327 		bargs = NULL;
3328 	}
3329 
3330 	if (fs_info->balance_ctl) {
3331 		ret = -EINPROGRESS;
3332 		goto out_bargs;
3333 	}
3334 
3335 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3336 	if (!bctl) {
3337 		ret = -ENOMEM;
3338 		goto out_bargs;
3339 	}
3340 
3341 	bctl->fs_info = fs_info;
3342 	if (arg) {
3343 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3344 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3345 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3346 
3347 		bctl->flags = bargs->flags;
3348 	} else {
3349 		/* balance everything - no filters */
3350 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3351 	}
3352 
3353 do_balance:
3354 	ret = btrfs_balance(bctl, bargs);
3355 	/*
3356 	 * bctl is freed in __cancel_balance or in free_fs_info if
3357 	 * restriper was paused all the way until unmount
3358 	 */
3359 	if (arg) {
3360 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3361 			ret = -EFAULT;
3362 	}
3363 
3364 out_bargs:
3365 	kfree(bargs);
3366 out:
3367 	mutex_unlock(&fs_info->balance_mutex);
3368 	mutex_unlock(&fs_info->volume_mutex);
3369 	mnt_drop_write_file(file);
3370 	return ret;
3371 }
3372 
3373 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3374 {
3375 	if (!capable(CAP_SYS_ADMIN))
3376 		return -EPERM;
3377 
3378 	switch (cmd) {
3379 	case BTRFS_BALANCE_CTL_PAUSE:
3380 		return btrfs_pause_balance(root->fs_info);
3381 	case BTRFS_BALANCE_CTL_CANCEL:
3382 		return btrfs_cancel_balance(root->fs_info);
3383 	}
3384 
3385 	return -EINVAL;
3386 }
3387 
3388 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3389 					 void __user *arg)
3390 {
3391 	struct btrfs_fs_info *fs_info = root->fs_info;
3392 	struct btrfs_ioctl_balance_args *bargs;
3393 	int ret = 0;
3394 
3395 	if (!capable(CAP_SYS_ADMIN))
3396 		return -EPERM;
3397 
3398 	mutex_lock(&fs_info->balance_mutex);
3399 	if (!fs_info->balance_ctl) {
3400 		ret = -ENOTCONN;
3401 		goto out;
3402 	}
3403 
3404 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3405 	if (!bargs) {
3406 		ret = -ENOMEM;
3407 		goto out;
3408 	}
3409 
3410 	update_ioctl_balance_args(fs_info, 1, bargs);
3411 
3412 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3413 		ret = -EFAULT;
3414 
3415 	kfree(bargs);
3416 out:
3417 	mutex_unlock(&fs_info->balance_mutex);
3418 	return ret;
3419 }
3420 
3421 static long btrfs_ioctl_quota_ctl(struct btrfs_root *root, void __user *arg)
3422 {
3423 	struct btrfs_ioctl_quota_ctl_args *sa;
3424 	struct btrfs_trans_handle *trans = NULL;
3425 	int ret;
3426 	int err;
3427 
3428 	if (!capable(CAP_SYS_ADMIN))
3429 		return -EPERM;
3430 
3431 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3432 		return -EROFS;
3433 
3434 	sa = memdup_user(arg, sizeof(*sa));
3435 	if (IS_ERR(sa))
3436 		return PTR_ERR(sa);
3437 
3438 	if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3439 		trans = btrfs_start_transaction(root, 2);
3440 		if (IS_ERR(trans)) {
3441 			ret = PTR_ERR(trans);
3442 			goto out;
3443 		}
3444 	}
3445 
3446 	switch (sa->cmd) {
3447 	case BTRFS_QUOTA_CTL_ENABLE:
3448 		ret = btrfs_quota_enable(trans, root->fs_info);
3449 		break;
3450 	case BTRFS_QUOTA_CTL_DISABLE:
3451 		ret = btrfs_quota_disable(trans, root->fs_info);
3452 		break;
3453 	case BTRFS_QUOTA_CTL_RESCAN:
3454 		ret = btrfs_quota_rescan(root->fs_info);
3455 		break;
3456 	default:
3457 		ret = -EINVAL;
3458 		break;
3459 	}
3460 
3461 	if (copy_to_user(arg, sa, sizeof(*sa)))
3462 		ret = -EFAULT;
3463 
3464 	if (trans) {
3465 		err = btrfs_commit_transaction(trans, root);
3466 		if (err && !ret)
3467 			ret = err;
3468 	}
3469 
3470 out:
3471 	kfree(sa);
3472 	return ret;
3473 }
3474 
3475 static long btrfs_ioctl_qgroup_assign(struct btrfs_root *root, void __user *arg)
3476 {
3477 	struct btrfs_ioctl_qgroup_assign_args *sa;
3478 	struct btrfs_trans_handle *trans;
3479 	int ret;
3480 	int err;
3481 
3482 	if (!capable(CAP_SYS_ADMIN))
3483 		return -EPERM;
3484 
3485 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3486 		return -EROFS;
3487 
3488 	sa = memdup_user(arg, sizeof(*sa));
3489 	if (IS_ERR(sa))
3490 		return PTR_ERR(sa);
3491 
3492 	trans = btrfs_join_transaction(root);
3493 	if (IS_ERR(trans)) {
3494 		ret = PTR_ERR(trans);
3495 		goto out;
3496 	}
3497 
3498 	/* FIXME: check if the IDs really exist */
3499 	if (sa->assign) {
3500 		ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3501 						sa->src, sa->dst);
3502 	} else {
3503 		ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3504 						sa->src, sa->dst);
3505 	}
3506 
3507 	err = btrfs_end_transaction(trans, root);
3508 	if (err && !ret)
3509 		ret = err;
3510 
3511 out:
3512 	kfree(sa);
3513 	return ret;
3514 }
3515 
3516 static long btrfs_ioctl_qgroup_create(struct btrfs_root *root, void __user *arg)
3517 {
3518 	struct btrfs_ioctl_qgroup_create_args *sa;
3519 	struct btrfs_trans_handle *trans;
3520 	int ret;
3521 	int err;
3522 
3523 	if (!capable(CAP_SYS_ADMIN))
3524 		return -EPERM;
3525 
3526 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3527 		return -EROFS;
3528 
3529 	sa = memdup_user(arg, sizeof(*sa));
3530 	if (IS_ERR(sa))
3531 		return PTR_ERR(sa);
3532 
3533 	trans = btrfs_join_transaction(root);
3534 	if (IS_ERR(trans)) {
3535 		ret = PTR_ERR(trans);
3536 		goto out;
3537 	}
3538 
3539 	/* FIXME: check if the IDs really exist */
3540 	if (sa->create) {
3541 		ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3542 					  NULL);
3543 	} else {
3544 		ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3545 	}
3546 
3547 	err = btrfs_end_transaction(trans, root);
3548 	if (err && !ret)
3549 		ret = err;
3550 
3551 out:
3552 	kfree(sa);
3553 	return ret;
3554 }
3555 
3556 static long btrfs_ioctl_qgroup_limit(struct btrfs_root *root, void __user *arg)
3557 {
3558 	struct btrfs_ioctl_qgroup_limit_args *sa;
3559 	struct btrfs_trans_handle *trans;
3560 	int ret;
3561 	int err;
3562 	u64 qgroupid;
3563 
3564 	if (!capable(CAP_SYS_ADMIN))
3565 		return -EPERM;
3566 
3567 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3568 		return -EROFS;
3569 
3570 	sa = memdup_user(arg, sizeof(*sa));
3571 	if (IS_ERR(sa))
3572 		return PTR_ERR(sa);
3573 
3574 	trans = btrfs_join_transaction(root);
3575 	if (IS_ERR(trans)) {
3576 		ret = PTR_ERR(trans);
3577 		goto out;
3578 	}
3579 
3580 	qgroupid = sa->qgroupid;
3581 	if (!qgroupid) {
3582 		/* take the current subvol as qgroup */
3583 		qgroupid = root->root_key.objectid;
3584 	}
3585 
3586 	/* FIXME: check if the IDs really exist */
3587 	ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3588 
3589 	err = btrfs_end_transaction(trans, root);
3590 	if (err && !ret)
3591 		ret = err;
3592 
3593 out:
3594 	kfree(sa);
3595 	return ret;
3596 }
3597 
3598 static long btrfs_ioctl_set_received_subvol(struct file *file,
3599 					    void __user *arg)
3600 {
3601 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
3602 	struct inode *inode = fdentry(file)->d_inode;
3603 	struct btrfs_root *root = BTRFS_I(inode)->root;
3604 	struct btrfs_root_item *root_item = &root->root_item;
3605 	struct btrfs_trans_handle *trans;
3606 	struct timespec ct = CURRENT_TIME;
3607 	int ret = 0;
3608 
3609 	ret = mnt_want_write_file(file);
3610 	if (ret < 0)
3611 		return ret;
3612 
3613 	down_write(&root->fs_info->subvol_sem);
3614 
3615 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3616 		ret = -EINVAL;
3617 		goto out;
3618 	}
3619 
3620 	if (btrfs_root_readonly(root)) {
3621 		ret = -EROFS;
3622 		goto out;
3623 	}
3624 
3625 	if (!inode_owner_or_capable(inode)) {
3626 		ret = -EACCES;
3627 		goto out;
3628 	}
3629 
3630 	sa = memdup_user(arg, sizeof(*sa));
3631 	if (IS_ERR(sa)) {
3632 		ret = PTR_ERR(sa);
3633 		sa = NULL;
3634 		goto out;
3635 	}
3636 
3637 	trans = btrfs_start_transaction(root, 1);
3638 	if (IS_ERR(trans)) {
3639 		ret = PTR_ERR(trans);
3640 		trans = NULL;
3641 		goto out;
3642 	}
3643 
3644 	sa->rtransid = trans->transid;
3645 	sa->rtime.sec = ct.tv_sec;
3646 	sa->rtime.nsec = ct.tv_nsec;
3647 
3648 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3649 	btrfs_set_root_stransid(root_item, sa->stransid);
3650 	btrfs_set_root_rtransid(root_item, sa->rtransid);
3651 	root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3652 	root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3653 	root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3654 	root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3655 
3656 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
3657 				&root->root_key, &root->root_item);
3658 	if (ret < 0) {
3659 		btrfs_end_transaction(trans, root);
3660 		trans = NULL;
3661 		goto out;
3662 	} else {
3663 		ret = btrfs_commit_transaction(trans, root);
3664 		if (ret < 0)
3665 			goto out;
3666 	}
3667 
3668 	ret = copy_to_user(arg, sa, sizeof(*sa));
3669 	if (ret)
3670 		ret = -EFAULT;
3671 
3672 out:
3673 	kfree(sa);
3674 	up_write(&root->fs_info->subvol_sem);
3675 	mnt_drop_write_file(file);
3676 	return ret;
3677 }
3678 
3679 long btrfs_ioctl(struct file *file, unsigned int
3680 		cmd, unsigned long arg)
3681 {
3682 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3683 	void __user *argp = (void __user *)arg;
3684 
3685 	switch (cmd) {
3686 	case FS_IOC_GETFLAGS:
3687 		return btrfs_ioctl_getflags(file, argp);
3688 	case FS_IOC_SETFLAGS:
3689 		return btrfs_ioctl_setflags(file, argp);
3690 	case FS_IOC_GETVERSION:
3691 		return btrfs_ioctl_getversion(file, argp);
3692 	case FITRIM:
3693 		return btrfs_ioctl_fitrim(file, argp);
3694 	case BTRFS_IOC_SNAP_CREATE:
3695 		return btrfs_ioctl_snap_create(file, argp, 0);
3696 	case BTRFS_IOC_SNAP_CREATE_V2:
3697 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3698 	case BTRFS_IOC_SUBVOL_CREATE:
3699 		return btrfs_ioctl_snap_create(file, argp, 1);
3700 	case BTRFS_IOC_SUBVOL_CREATE_V2:
3701 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
3702 	case BTRFS_IOC_SNAP_DESTROY:
3703 		return btrfs_ioctl_snap_destroy(file, argp);
3704 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3705 		return btrfs_ioctl_subvol_getflags(file, argp);
3706 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3707 		return btrfs_ioctl_subvol_setflags(file, argp);
3708 	case BTRFS_IOC_DEFAULT_SUBVOL:
3709 		return btrfs_ioctl_default_subvol(file, argp);
3710 	case BTRFS_IOC_DEFRAG:
3711 		return btrfs_ioctl_defrag(file, NULL);
3712 	case BTRFS_IOC_DEFRAG_RANGE:
3713 		return btrfs_ioctl_defrag(file, argp);
3714 	case BTRFS_IOC_RESIZE:
3715 		return btrfs_ioctl_resize(root, argp);
3716 	case BTRFS_IOC_ADD_DEV:
3717 		return btrfs_ioctl_add_dev(root, argp);
3718 	case BTRFS_IOC_RM_DEV:
3719 		return btrfs_ioctl_rm_dev(root, argp);
3720 	case BTRFS_IOC_FS_INFO:
3721 		return btrfs_ioctl_fs_info(root, argp);
3722 	case BTRFS_IOC_DEV_INFO:
3723 		return btrfs_ioctl_dev_info(root, argp);
3724 	case BTRFS_IOC_BALANCE:
3725 		return btrfs_ioctl_balance(file, NULL);
3726 	case BTRFS_IOC_CLONE:
3727 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3728 	case BTRFS_IOC_CLONE_RANGE:
3729 		return btrfs_ioctl_clone_range(file, argp);
3730 	case BTRFS_IOC_TRANS_START:
3731 		return btrfs_ioctl_trans_start(file);
3732 	case BTRFS_IOC_TRANS_END:
3733 		return btrfs_ioctl_trans_end(file);
3734 	case BTRFS_IOC_TREE_SEARCH:
3735 		return btrfs_ioctl_tree_search(file, argp);
3736 	case BTRFS_IOC_INO_LOOKUP:
3737 		return btrfs_ioctl_ino_lookup(file, argp);
3738 	case BTRFS_IOC_INO_PATHS:
3739 		return btrfs_ioctl_ino_to_path(root, argp);
3740 	case BTRFS_IOC_LOGICAL_INO:
3741 		return btrfs_ioctl_logical_to_ino(root, argp);
3742 	case BTRFS_IOC_SPACE_INFO:
3743 		return btrfs_ioctl_space_info(root, argp);
3744 	case BTRFS_IOC_SYNC:
3745 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3746 		return 0;
3747 	case BTRFS_IOC_START_SYNC:
3748 		return btrfs_ioctl_start_sync(file, argp);
3749 	case BTRFS_IOC_WAIT_SYNC:
3750 		return btrfs_ioctl_wait_sync(file, argp);
3751 	case BTRFS_IOC_SCRUB:
3752 		return btrfs_ioctl_scrub(root, argp);
3753 	case BTRFS_IOC_SCRUB_CANCEL:
3754 		return btrfs_ioctl_scrub_cancel(root, argp);
3755 	case BTRFS_IOC_SCRUB_PROGRESS:
3756 		return btrfs_ioctl_scrub_progress(root, argp);
3757 	case BTRFS_IOC_BALANCE_V2:
3758 		return btrfs_ioctl_balance(file, argp);
3759 	case BTRFS_IOC_BALANCE_CTL:
3760 		return btrfs_ioctl_balance_ctl(root, arg);
3761 	case BTRFS_IOC_BALANCE_PROGRESS:
3762 		return btrfs_ioctl_balance_progress(root, argp);
3763 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
3764 		return btrfs_ioctl_set_received_subvol(file, argp);
3765 	case BTRFS_IOC_SEND:
3766 		return btrfs_ioctl_send(file, argp);
3767 	case BTRFS_IOC_GET_DEV_STATS:
3768 		return btrfs_ioctl_get_dev_stats(root, argp);
3769 	case BTRFS_IOC_QUOTA_CTL:
3770 		return btrfs_ioctl_quota_ctl(root, argp);
3771 	case BTRFS_IOC_QGROUP_ASSIGN:
3772 		return btrfs_ioctl_qgroup_assign(root, argp);
3773 	case BTRFS_IOC_QGROUP_CREATE:
3774 		return btrfs_ioctl_qgroup_create(root, argp);
3775 	case BTRFS_IOC_QGROUP_LIMIT:
3776 		return btrfs_ioctl_qgroup_limit(root, argp);
3777 	}
3778 
3779 	return -ENOTTY;
3780 }
3781