xref: /linux/fs/btrfs/ioctl.c (revision 7056741fd9fc14a65608549a4657cf5178f05f63)
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 	umode_t mode;
185 
186 	if (btrfs_root_readonly(root))
187 		return -EROFS;
188 
189 	if (copy_from_user(&flags, arg, sizeof(flags)))
190 		return -EFAULT;
191 
192 	ret = check_flags(flags);
193 	if (ret)
194 		return ret;
195 
196 	if (!inode_owner_or_capable(inode))
197 		return -EACCES;
198 
199 	ret = mnt_want_write_file(file);
200 	if (ret)
201 		return ret;
202 
203 	mutex_lock(&inode->i_mutex);
204 
205 	ip_oldflags = ip->flags;
206 	i_oldflags = inode->i_flags;
207 	mode = inode->i_mode;
208 
209 	flags = btrfs_mask_flags(inode->i_mode, flags);
210 	oldflags = btrfs_flags_to_ioctl(ip->flags);
211 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
212 		if (!capable(CAP_LINUX_IMMUTABLE)) {
213 			ret = -EPERM;
214 			goto out_unlock;
215 		}
216 	}
217 
218 	if (flags & FS_SYNC_FL)
219 		ip->flags |= BTRFS_INODE_SYNC;
220 	else
221 		ip->flags &= ~BTRFS_INODE_SYNC;
222 	if (flags & FS_IMMUTABLE_FL)
223 		ip->flags |= BTRFS_INODE_IMMUTABLE;
224 	else
225 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
226 	if (flags & FS_APPEND_FL)
227 		ip->flags |= BTRFS_INODE_APPEND;
228 	else
229 		ip->flags &= ~BTRFS_INODE_APPEND;
230 	if (flags & FS_NODUMP_FL)
231 		ip->flags |= BTRFS_INODE_NODUMP;
232 	else
233 		ip->flags &= ~BTRFS_INODE_NODUMP;
234 	if (flags & FS_NOATIME_FL)
235 		ip->flags |= BTRFS_INODE_NOATIME;
236 	else
237 		ip->flags &= ~BTRFS_INODE_NOATIME;
238 	if (flags & FS_DIRSYNC_FL)
239 		ip->flags |= BTRFS_INODE_DIRSYNC;
240 	else
241 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
242 	if (flags & FS_NOCOW_FL) {
243 		if (S_ISREG(mode)) {
244 			/*
245 			 * It's safe to turn csums off here, no extents exist.
246 			 * Otherwise we want the flag to reflect the real COW
247 			 * status of the file and will not set it.
248 			 */
249 			if (inode->i_size == 0)
250 				ip->flags |= BTRFS_INODE_NODATACOW
251 					   | BTRFS_INODE_NODATASUM;
252 		} else {
253 			ip->flags |= BTRFS_INODE_NODATACOW;
254 		}
255 	} else {
256 		/*
257 		 * Revert back under same assuptions as above
258 		 */
259 		if (S_ISREG(mode)) {
260 			if (inode->i_size == 0)
261 				ip->flags &= ~(BTRFS_INODE_NODATACOW
262 				             | BTRFS_INODE_NODATASUM);
263 		} else {
264 			ip->flags &= ~BTRFS_INODE_NODATACOW;
265 		}
266 	}
267 
268 	/*
269 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
270 	 * flag may be changed automatically if compression code won't make
271 	 * things smaller.
272 	 */
273 	if (flags & FS_NOCOMP_FL) {
274 		ip->flags &= ~BTRFS_INODE_COMPRESS;
275 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
276 	} else if (flags & FS_COMPR_FL) {
277 		ip->flags |= BTRFS_INODE_COMPRESS;
278 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
279 	} else {
280 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
281 	}
282 
283 	trans = btrfs_start_transaction(root, 1);
284 	if (IS_ERR(trans)) {
285 		ret = PTR_ERR(trans);
286 		goto out_drop;
287 	}
288 
289 	btrfs_update_iflags(inode);
290 	inode_inc_iversion(inode);
291 	inode->i_ctime = CURRENT_TIME;
292 	ret = btrfs_update_inode(trans, root, inode);
293 
294 	btrfs_end_transaction(trans, root);
295  out_drop:
296 	if (ret) {
297 		ip->flags = ip_oldflags;
298 		inode->i_flags = i_oldflags;
299 	}
300 
301  out_unlock:
302 	mutex_unlock(&inode->i_mutex);
303 	mnt_drop_write_file(file);
304 	return ret;
305 }
306 
307 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
308 {
309 	struct inode *inode = file->f_path.dentry->d_inode;
310 
311 	return put_user(inode->i_generation, arg);
312 }
313 
314 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
315 {
316 	struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
317 	struct btrfs_device *device;
318 	struct request_queue *q;
319 	struct fstrim_range range;
320 	u64 minlen = ULLONG_MAX;
321 	u64 num_devices = 0;
322 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
323 	int ret;
324 
325 	if (!capable(CAP_SYS_ADMIN))
326 		return -EPERM;
327 
328 	rcu_read_lock();
329 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
330 				dev_list) {
331 		if (!device->bdev)
332 			continue;
333 		q = bdev_get_queue(device->bdev);
334 		if (blk_queue_discard(q)) {
335 			num_devices++;
336 			minlen = min((u64)q->limits.discard_granularity,
337 				     minlen);
338 		}
339 	}
340 	rcu_read_unlock();
341 
342 	if (!num_devices)
343 		return -EOPNOTSUPP;
344 	if (copy_from_user(&range, arg, sizeof(range)))
345 		return -EFAULT;
346 	if (range.start > total_bytes)
347 		return -EINVAL;
348 
349 	range.len = min(range.len, total_bytes - range.start);
350 	range.minlen = max(range.minlen, minlen);
351 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
352 	if (ret < 0)
353 		return ret;
354 
355 	if (copy_to_user(arg, &range, sizeof(range)))
356 		return -EFAULT;
357 
358 	return 0;
359 }
360 
361 static noinline int create_subvol(struct btrfs_root *root,
362 				  struct dentry *dentry,
363 				  char *name, int namelen,
364 				  u64 *async_transid,
365 				  struct btrfs_qgroup_inherit **inherit)
366 {
367 	struct btrfs_trans_handle *trans;
368 	struct btrfs_key key;
369 	struct btrfs_root_item root_item;
370 	struct btrfs_inode_item *inode_item;
371 	struct extent_buffer *leaf;
372 	struct btrfs_root *new_root;
373 	struct dentry *parent = dentry->d_parent;
374 	struct inode *dir;
375 	struct timespec cur_time = CURRENT_TIME;
376 	int ret;
377 	int err;
378 	u64 objectid;
379 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
380 	u64 index = 0;
381 	uuid_le new_uuid;
382 
383 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
384 	if (ret)
385 		return ret;
386 
387 	dir = parent->d_inode;
388 
389 	/*
390 	 * 1 - inode item
391 	 * 2 - refs
392 	 * 1 - root item
393 	 * 2 - dir items
394 	 */
395 	trans = btrfs_start_transaction(root, 6);
396 	if (IS_ERR(trans))
397 		return PTR_ERR(trans);
398 
399 	ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid,
400 				   inherit ? *inherit : NULL);
401 	if (ret)
402 		goto fail;
403 
404 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
405 				      0, objectid, NULL, 0, 0, 0);
406 	if (IS_ERR(leaf)) {
407 		ret = PTR_ERR(leaf);
408 		goto fail;
409 	}
410 
411 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
412 	btrfs_set_header_bytenr(leaf, leaf->start);
413 	btrfs_set_header_generation(leaf, trans->transid);
414 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
415 	btrfs_set_header_owner(leaf, objectid);
416 
417 	write_extent_buffer(leaf, root->fs_info->fsid,
418 			    (unsigned long)btrfs_header_fsid(leaf),
419 			    BTRFS_FSID_SIZE);
420 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
421 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
422 			    BTRFS_UUID_SIZE);
423 	btrfs_mark_buffer_dirty(leaf);
424 
425 	memset(&root_item, 0, sizeof(root_item));
426 
427 	inode_item = &root_item.inode;
428 	inode_item->generation = cpu_to_le64(1);
429 	inode_item->size = cpu_to_le64(3);
430 	inode_item->nlink = cpu_to_le32(1);
431 	inode_item->nbytes = cpu_to_le64(root->leafsize);
432 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
433 
434 	root_item.flags = 0;
435 	root_item.byte_limit = 0;
436 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
437 
438 	btrfs_set_root_bytenr(&root_item, leaf->start);
439 	btrfs_set_root_generation(&root_item, trans->transid);
440 	btrfs_set_root_level(&root_item, 0);
441 	btrfs_set_root_refs(&root_item, 1);
442 	btrfs_set_root_used(&root_item, leaf->len);
443 	btrfs_set_root_last_snapshot(&root_item, 0);
444 
445 	btrfs_set_root_generation_v2(&root_item,
446 			btrfs_root_generation(&root_item));
447 	uuid_le_gen(&new_uuid);
448 	memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
449 	root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
450 	root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
451 	root_item.ctime = root_item.otime;
452 	btrfs_set_root_ctransid(&root_item, trans->transid);
453 	btrfs_set_root_otransid(&root_item, trans->transid);
454 
455 	btrfs_tree_unlock(leaf);
456 	free_extent_buffer(leaf);
457 	leaf = NULL;
458 
459 	btrfs_set_root_dirid(&root_item, new_dirid);
460 
461 	key.objectid = objectid;
462 	key.offset = 0;
463 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
464 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
465 				&root_item);
466 	if (ret)
467 		goto fail;
468 
469 	key.offset = (u64)-1;
470 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
471 	if (IS_ERR(new_root)) {
472 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
473 		ret = PTR_ERR(new_root);
474 		goto fail;
475 	}
476 
477 	btrfs_record_root_in_trans(trans, new_root);
478 
479 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
480 	if (ret) {
481 		/* We potentially lose an unused inode item here */
482 		btrfs_abort_transaction(trans, root, ret);
483 		goto fail;
484 	}
485 
486 	/*
487 	 * insert the directory item
488 	 */
489 	ret = btrfs_set_inode_index(dir, &index);
490 	if (ret) {
491 		btrfs_abort_transaction(trans, root, ret);
492 		goto fail;
493 	}
494 
495 	ret = btrfs_insert_dir_item(trans, root,
496 				    name, namelen, dir, &key,
497 				    BTRFS_FT_DIR, index);
498 	if (ret) {
499 		btrfs_abort_transaction(trans, root, ret);
500 		goto fail;
501 	}
502 
503 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
504 	ret = btrfs_update_inode(trans, root, dir);
505 	BUG_ON(ret);
506 
507 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
508 				 objectid, root->root_key.objectid,
509 				 btrfs_ino(dir), index, name, namelen);
510 
511 	BUG_ON(ret);
512 
513 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
514 fail:
515 	if (async_transid) {
516 		*async_transid = trans->transid;
517 		err = btrfs_commit_transaction_async(trans, root, 1);
518 	} else {
519 		err = btrfs_commit_transaction(trans, root);
520 	}
521 	if (err && !ret)
522 		ret = err;
523 	return ret;
524 }
525 
526 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
527 			   char *name, int namelen, u64 *async_transid,
528 			   bool readonly, struct btrfs_qgroup_inherit **inherit)
529 {
530 	struct inode *inode;
531 	struct btrfs_pending_snapshot *pending_snapshot;
532 	struct btrfs_trans_handle *trans;
533 	int ret;
534 
535 	if (!root->ref_cows)
536 		return -EINVAL;
537 
538 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
539 	if (!pending_snapshot)
540 		return -ENOMEM;
541 
542 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
543 			     BTRFS_BLOCK_RSV_TEMP);
544 	pending_snapshot->dentry = dentry;
545 	pending_snapshot->root = root;
546 	pending_snapshot->readonly = readonly;
547 	if (inherit) {
548 		pending_snapshot->inherit = *inherit;
549 		*inherit = NULL;	/* take responsibility to free it */
550 	}
551 
552 	trans = btrfs_start_transaction(root->fs_info->extent_root, 6);
553 	if (IS_ERR(trans)) {
554 		ret = PTR_ERR(trans);
555 		goto fail;
556 	}
557 
558 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
559 	BUG_ON(ret);
560 
561 	spin_lock(&root->fs_info->trans_lock);
562 	list_add(&pending_snapshot->list,
563 		 &trans->transaction->pending_snapshots);
564 	spin_unlock(&root->fs_info->trans_lock);
565 	if (async_transid) {
566 		*async_transid = trans->transid;
567 		ret = btrfs_commit_transaction_async(trans,
568 				     root->fs_info->extent_root, 1);
569 	} else {
570 		ret = btrfs_commit_transaction(trans,
571 					       root->fs_info->extent_root);
572 	}
573 	BUG_ON(ret);
574 
575 	ret = pending_snapshot->error;
576 	if (ret)
577 		goto fail;
578 
579 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
580 	if (ret)
581 		goto fail;
582 
583 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
584 	if (IS_ERR(inode)) {
585 		ret = PTR_ERR(inode);
586 		goto fail;
587 	}
588 	BUG_ON(!inode);
589 	d_instantiate(dentry, inode);
590 	ret = 0;
591 fail:
592 	kfree(pending_snapshot);
593 	return ret;
594 }
595 
596 /*  copy of check_sticky in fs/namei.c()
597 * It's inline, so penalty for filesystems that don't use sticky bit is
598 * minimal.
599 */
600 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
601 {
602 	kuid_t fsuid = current_fsuid();
603 
604 	if (!(dir->i_mode & S_ISVTX))
605 		return 0;
606 	if (uid_eq(inode->i_uid, fsuid))
607 		return 0;
608 	if (uid_eq(dir->i_uid, fsuid))
609 		return 0;
610 	return !capable(CAP_FOWNER);
611 }
612 
613 /*  copy of may_delete in fs/namei.c()
614  *	Check whether we can remove a link victim from directory dir, check
615  *  whether the type of victim is right.
616  *  1. We can't do it if dir is read-only (done in permission())
617  *  2. We should have write and exec permissions on dir
618  *  3. We can't remove anything from append-only dir
619  *  4. We can't do anything with immutable dir (done in permission())
620  *  5. If the sticky bit on dir is set we should either
621  *	a. be owner of dir, or
622  *	b. be owner of victim, or
623  *	c. have CAP_FOWNER capability
624  *  6. If the victim is append-only or immutable we can't do antyhing with
625  *     links pointing to it.
626  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
627  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
628  *  9. We can't remove a root or mountpoint.
629  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
630  *     nfs_async_unlink().
631  */
632 
633 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
634 {
635 	int error;
636 
637 	if (!victim->d_inode)
638 		return -ENOENT;
639 
640 	BUG_ON(victim->d_parent->d_inode != dir);
641 	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
642 
643 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
644 	if (error)
645 		return error;
646 	if (IS_APPEND(dir))
647 		return -EPERM;
648 	if (btrfs_check_sticky(dir, victim->d_inode)||
649 		IS_APPEND(victim->d_inode)||
650 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
651 		return -EPERM;
652 	if (isdir) {
653 		if (!S_ISDIR(victim->d_inode->i_mode))
654 			return -ENOTDIR;
655 		if (IS_ROOT(victim))
656 			return -EBUSY;
657 	} else if (S_ISDIR(victim->d_inode->i_mode))
658 		return -EISDIR;
659 	if (IS_DEADDIR(dir))
660 		return -ENOENT;
661 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
662 		return -EBUSY;
663 	return 0;
664 }
665 
666 /* copy of may_create in fs/namei.c() */
667 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
668 {
669 	if (child->d_inode)
670 		return -EEXIST;
671 	if (IS_DEADDIR(dir))
672 		return -ENOENT;
673 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
674 }
675 
676 /*
677  * Create a new subvolume below @parent.  This is largely modeled after
678  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
679  * inside this filesystem so it's quite a bit simpler.
680  */
681 static noinline int btrfs_mksubvol(struct path *parent,
682 				   char *name, int namelen,
683 				   struct btrfs_root *snap_src,
684 				   u64 *async_transid, bool readonly,
685 				   struct btrfs_qgroup_inherit **inherit)
686 {
687 	struct inode *dir  = parent->dentry->d_inode;
688 	struct dentry *dentry;
689 	int error;
690 
691 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
692 
693 	dentry = lookup_one_len(name, parent->dentry, namelen);
694 	error = PTR_ERR(dentry);
695 	if (IS_ERR(dentry))
696 		goto out_unlock;
697 
698 	error = -EEXIST;
699 	if (dentry->d_inode)
700 		goto out_dput;
701 
702 	error = btrfs_may_create(dir, dentry);
703 	if (error)
704 		goto out_dput;
705 
706 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
707 
708 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
709 		goto out_up_read;
710 
711 	if (snap_src) {
712 		error = create_snapshot(snap_src, dentry, name, namelen,
713 					async_transid, readonly, inherit);
714 	} else {
715 		error = create_subvol(BTRFS_I(dir)->root, dentry,
716 				      name, namelen, async_transid, inherit);
717 	}
718 	if (!error)
719 		fsnotify_mkdir(dir, dentry);
720 out_up_read:
721 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
722 out_dput:
723 	dput(dentry);
724 out_unlock:
725 	mutex_unlock(&dir->i_mutex);
726 	return error;
727 }
728 
729 /*
730  * When we're defragging a range, we don't want to kick it off again
731  * if it is really just waiting for delalloc to send it down.
732  * If we find a nice big extent or delalloc range for the bytes in the
733  * file you want to defrag, we return 0 to let you know to skip this
734  * part of the file
735  */
736 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
737 {
738 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
739 	struct extent_map *em = NULL;
740 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
741 	u64 end;
742 
743 	read_lock(&em_tree->lock);
744 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
745 	read_unlock(&em_tree->lock);
746 
747 	if (em) {
748 		end = extent_map_end(em);
749 		free_extent_map(em);
750 		if (end - offset > thresh)
751 			return 0;
752 	}
753 	/* if we already have a nice delalloc here, just stop */
754 	thresh /= 2;
755 	end = count_range_bits(io_tree, &offset, offset + thresh,
756 			       thresh, EXTENT_DELALLOC, 1);
757 	if (end >= thresh)
758 		return 0;
759 	return 1;
760 }
761 
762 /*
763  * helper function to walk through a file and find extents
764  * newer than a specific transid, and smaller than thresh.
765  *
766  * This is used by the defragging code to find new and small
767  * extents
768  */
769 static int find_new_extents(struct btrfs_root *root,
770 			    struct inode *inode, u64 newer_than,
771 			    u64 *off, int thresh)
772 {
773 	struct btrfs_path *path;
774 	struct btrfs_key min_key;
775 	struct btrfs_key max_key;
776 	struct extent_buffer *leaf;
777 	struct btrfs_file_extent_item *extent;
778 	int type;
779 	int ret;
780 	u64 ino = btrfs_ino(inode);
781 
782 	path = btrfs_alloc_path();
783 	if (!path)
784 		return -ENOMEM;
785 
786 	min_key.objectid = ino;
787 	min_key.type = BTRFS_EXTENT_DATA_KEY;
788 	min_key.offset = *off;
789 
790 	max_key.objectid = ino;
791 	max_key.type = (u8)-1;
792 	max_key.offset = (u64)-1;
793 
794 	path->keep_locks = 1;
795 
796 	while(1) {
797 		ret = btrfs_search_forward(root, &min_key, &max_key,
798 					   path, 0, newer_than);
799 		if (ret != 0)
800 			goto none;
801 		if (min_key.objectid != ino)
802 			goto none;
803 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
804 			goto none;
805 
806 		leaf = path->nodes[0];
807 		extent = btrfs_item_ptr(leaf, path->slots[0],
808 					struct btrfs_file_extent_item);
809 
810 		type = btrfs_file_extent_type(leaf, extent);
811 		if (type == BTRFS_FILE_EXTENT_REG &&
812 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
813 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
814 			*off = min_key.offset;
815 			btrfs_free_path(path);
816 			return 0;
817 		}
818 
819 		if (min_key.offset == (u64)-1)
820 			goto none;
821 
822 		min_key.offset++;
823 		btrfs_release_path(path);
824 	}
825 none:
826 	btrfs_free_path(path);
827 	return -ENOENT;
828 }
829 
830 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
831 {
832 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
833 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
834 	struct extent_map *em;
835 	u64 len = PAGE_CACHE_SIZE;
836 
837 	/*
838 	 * hopefully we have this extent in the tree already, try without
839 	 * the full extent lock
840 	 */
841 	read_lock(&em_tree->lock);
842 	em = lookup_extent_mapping(em_tree, start, len);
843 	read_unlock(&em_tree->lock);
844 
845 	if (!em) {
846 		/* get the big lock and read metadata off disk */
847 		lock_extent(io_tree, start, start + len - 1);
848 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
849 		unlock_extent(io_tree, start, start + len - 1);
850 
851 		if (IS_ERR(em))
852 			return NULL;
853 	}
854 
855 	return em;
856 }
857 
858 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
859 {
860 	struct extent_map *next;
861 	bool ret = true;
862 
863 	/* this is the last extent */
864 	if (em->start + em->len >= i_size_read(inode))
865 		return false;
866 
867 	next = defrag_lookup_extent(inode, em->start + em->len);
868 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
869 		ret = false;
870 
871 	free_extent_map(next);
872 	return ret;
873 }
874 
875 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
876 			       u64 *last_len, u64 *skip, u64 *defrag_end,
877 			       int compress)
878 {
879 	struct extent_map *em;
880 	int ret = 1;
881 	bool next_mergeable = true;
882 
883 	/*
884 	 * make sure that once we start defragging an extent, we keep on
885 	 * defragging it
886 	 */
887 	if (start < *defrag_end)
888 		return 1;
889 
890 	*skip = 0;
891 
892 	em = defrag_lookup_extent(inode, start);
893 	if (!em)
894 		return 0;
895 
896 	/* this will cover holes, and inline extents */
897 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
898 		ret = 0;
899 		goto out;
900 	}
901 
902 	next_mergeable = defrag_check_next_extent(inode, em);
903 
904 	/*
905 	 * we hit a real extent, if it is big or the next extent is not a
906 	 * real extent, don't bother defragging it
907 	 */
908 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
909 	    (em->len >= thresh || !next_mergeable))
910 		ret = 0;
911 out:
912 	/*
913 	 * last_len ends up being a counter of how many bytes we've defragged.
914 	 * every time we choose not to defrag an extent, we reset *last_len
915 	 * so that the next tiny extent will force a defrag.
916 	 *
917 	 * The end result of this is that tiny extents before a single big
918 	 * extent will force at least part of that big extent to be defragged.
919 	 */
920 	if (ret) {
921 		*defrag_end = extent_map_end(em);
922 	} else {
923 		*last_len = 0;
924 		*skip = extent_map_end(em);
925 		*defrag_end = 0;
926 	}
927 
928 	free_extent_map(em);
929 	return ret;
930 }
931 
932 /*
933  * it doesn't do much good to defrag one or two pages
934  * at a time.  This pulls in a nice chunk of pages
935  * to COW and defrag.
936  *
937  * It also makes sure the delalloc code has enough
938  * dirty data to avoid making new small extents as part
939  * of the defrag
940  *
941  * It's a good idea to start RA on this range
942  * before calling this.
943  */
944 static int cluster_pages_for_defrag(struct inode *inode,
945 				    struct page **pages,
946 				    unsigned long start_index,
947 				    int num_pages)
948 {
949 	unsigned long file_end;
950 	u64 isize = i_size_read(inode);
951 	u64 page_start;
952 	u64 page_end;
953 	u64 page_cnt;
954 	int ret;
955 	int i;
956 	int i_done;
957 	struct btrfs_ordered_extent *ordered;
958 	struct extent_state *cached_state = NULL;
959 	struct extent_io_tree *tree;
960 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
961 
962 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
963 	if (!isize || start_index > file_end)
964 		return 0;
965 
966 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
967 
968 	ret = btrfs_delalloc_reserve_space(inode,
969 					   page_cnt << PAGE_CACHE_SHIFT);
970 	if (ret)
971 		return ret;
972 	i_done = 0;
973 	tree = &BTRFS_I(inode)->io_tree;
974 
975 	/* step one, lock all the pages */
976 	for (i = 0; i < page_cnt; i++) {
977 		struct page *page;
978 again:
979 		page = find_or_create_page(inode->i_mapping,
980 					   start_index + i, mask);
981 		if (!page)
982 			break;
983 
984 		page_start = page_offset(page);
985 		page_end = page_start + PAGE_CACHE_SIZE - 1;
986 		while (1) {
987 			lock_extent(tree, page_start, page_end);
988 			ordered = btrfs_lookup_ordered_extent(inode,
989 							      page_start);
990 			unlock_extent(tree, page_start, page_end);
991 			if (!ordered)
992 				break;
993 
994 			unlock_page(page);
995 			btrfs_start_ordered_extent(inode, ordered, 1);
996 			btrfs_put_ordered_extent(ordered);
997 			lock_page(page);
998 			/*
999 			 * we unlocked the page above, so we need check if
1000 			 * it was released or not.
1001 			 */
1002 			if (page->mapping != inode->i_mapping) {
1003 				unlock_page(page);
1004 				page_cache_release(page);
1005 				goto again;
1006 			}
1007 		}
1008 
1009 		if (!PageUptodate(page)) {
1010 			btrfs_readpage(NULL, page);
1011 			lock_page(page);
1012 			if (!PageUptodate(page)) {
1013 				unlock_page(page);
1014 				page_cache_release(page);
1015 				ret = -EIO;
1016 				break;
1017 			}
1018 		}
1019 
1020 		if (page->mapping != inode->i_mapping) {
1021 			unlock_page(page);
1022 			page_cache_release(page);
1023 			goto again;
1024 		}
1025 
1026 		pages[i] = page;
1027 		i_done++;
1028 	}
1029 	if (!i_done || ret)
1030 		goto out;
1031 
1032 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
1033 		goto out;
1034 
1035 	/*
1036 	 * so now we have a nice long stream of locked
1037 	 * and up to date pages, lets wait on them
1038 	 */
1039 	for (i = 0; i < i_done; i++)
1040 		wait_on_page_writeback(pages[i]);
1041 
1042 	page_start = page_offset(pages[0]);
1043 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1044 
1045 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1046 			 page_start, page_end - 1, 0, &cached_state);
1047 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1048 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1049 			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1050 			  &cached_state, GFP_NOFS);
1051 
1052 	if (i_done != page_cnt) {
1053 		spin_lock(&BTRFS_I(inode)->lock);
1054 		BTRFS_I(inode)->outstanding_extents++;
1055 		spin_unlock(&BTRFS_I(inode)->lock);
1056 		btrfs_delalloc_release_space(inode,
1057 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1058 	}
1059 
1060 
1061 	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1062 			  &cached_state, GFP_NOFS);
1063 
1064 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1065 			     page_start, page_end - 1, &cached_state,
1066 			     GFP_NOFS);
1067 
1068 	for (i = 0; i < i_done; i++) {
1069 		clear_page_dirty_for_io(pages[i]);
1070 		ClearPageChecked(pages[i]);
1071 		set_page_extent_mapped(pages[i]);
1072 		set_page_dirty(pages[i]);
1073 		unlock_page(pages[i]);
1074 		page_cache_release(pages[i]);
1075 	}
1076 	return i_done;
1077 out:
1078 	for (i = 0; i < i_done; i++) {
1079 		unlock_page(pages[i]);
1080 		page_cache_release(pages[i]);
1081 	}
1082 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1083 	return ret;
1084 
1085 }
1086 
1087 int btrfs_defrag_file(struct inode *inode, struct file *file,
1088 		      struct btrfs_ioctl_defrag_range_args *range,
1089 		      u64 newer_than, unsigned long max_to_defrag)
1090 {
1091 	struct btrfs_root *root = BTRFS_I(inode)->root;
1092 	struct file_ra_state *ra = NULL;
1093 	unsigned long last_index;
1094 	u64 isize = i_size_read(inode);
1095 	u64 last_len = 0;
1096 	u64 skip = 0;
1097 	u64 defrag_end = 0;
1098 	u64 newer_off = range->start;
1099 	unsigned long i;
1100 	unsigned long ra_index = 0;
1101 	int ret;
1102 	int defrag_count = 0;
1103 	int compress_type = BTRFS_COMPRESS_ZLIB;
1104 	int extent_thresh = range->extent_thresh;
1105 	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1106 	int cluster = max_cluster;
1107 	u64 new_align = ~((u64)128 * 1024 - 1);
1108 	struct page **pages = NULL;
1109 
1110 	if (extent_thresh == 0)
1111 		extent_thresh = 256 * 1024;
1112 
1113 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1114 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1115 			return -EINVAL;
1116 		if (range->compress_type)
1117 			compress_type = range->compress_type;
1118 	}
1119 
1120 	if (isize == 0)
1121 		return 0;
1122 
1123 	/*
1124 	 * if we were not given a file, allocate a readahead
1125 	 * context
1126 	 */
1127 	if (!file) {
1128 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1129 		if (!ra)
1130 			return -ENOMEM;
1131 		file_ra_state_init(ra, inode->i_mapping);
1132 	} else {
1133 		ra = &file->f_ra;
1134 	}
1135 
1136 	pages = kmalloc(sizeof(struct page *) * max_cluster,
1137 			GFP_NOFS);
1138 	if (!pages) {
1139 		ret = -ENOMEM;
1140 		goto out_ra;
1141 	}
1142 
1143 	/* find the last page to defrag */
1144 	if (range->start + range->len > range->start) {
1145 		last_index = min_t(u64, isize - 1,
1146 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1147 	} else {
1148 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1149 	}
1150 
1151 	if (newer_than) {
1152 		ret = find_new_extents(root, inode, newer_than,
1153 				       &newer_off, 64 * 1024);
1154 		if (!ret) {
1155 			range->start = newer_off;
1156 			/*
1157 			 * we always align our defrag to help keep
1158 			 * the extents in the file evenly spaced
1159 			 */
1160 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1161 		} else
1162 			goto out_ra;
1163 	} else {
1164 		i = range->start >> PAGE_CACHE_SHIFT;
1165 	}
1166 	if (!max_to_defrag)
1167 		max_to_defrag = last_index + 1;
1168 
1169 	/*
1170 	 * make writeback starts from i, so the defrag range can be
1171 	 * written sequentially.
1172 	 */
1173 	if (i < inode->i_mapping->writeback_index)
1174 		inode->i_mapping->writeback_index = i;
1175 
1176 	while (i <= last_index && defrag_count < max_to_defrag &&
1177 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1178 		PAGE_CACHE_SHIFT)) {
1179 		/*
1180 		 * make sure we stop running if someone unmounts
1181 		 * the FS
1182 		 */
1183 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1184 			break;
1185 
1186 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1187 					 extent_thresh, &last_len, &skip,
1188 					 &defrag_end, range->flags &
1189 					 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1190 			unsigned long next;
1191 			/*
1192 			 * the should_defrag function tells us how much to skip
1193 			 * bump our counter by the suggested amount
1194 			 */
1195 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1196 			i = max(i + 1, next);
1197 			continue;
1198 		}
1199 
1200 		if (!newer_than) {
1201 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1202 				   PAGE_CACHE_SHIFT) - i;
1203 			cluster = min(cluster, max_cluster);
1204 		} else {
1205 			cluster = max_cluster;
1206 		}
1207 
1208 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1209 			BTRFS_I(inode)->force_compress = compress_type;
1210 
1211 		if (i + cluster > ra_index) {
1212 			ra_index = max(i, ra_index);
1213 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1214 				       cluster);
1215 			ra_index += max_cluster;
1216 		}
1217 
1218 		mutex_lock(&inode->i_mutex);
1219 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1220 		if (ret < 0) {
1221 			mutex_unlock(&inode->i_mutex);
1222 			goto out_ra;
1223 		}
1224 
1225 		defrag_count += ret;
1226 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1227 		mutex_unlock(&inode->i_mutex);
1228 
1229 		if (newer_than) {
1230 			if (newer_off == (u64)-1)
1231 				break;
1232 
1233 			if (ret > 0)
1234 				i += ret;
1235 
1236 			newer_off = max(newer_off + 1,
1237 					(u64)i << PAGE_CACHE_SHIFT);
1238 
1239 			ret = find_new_extents(root, inode,
1240 					       newer_than, &newer_off,
1241 					       64 * 1024);
1242 			if (!ret) {
1243 				range->start = newer_off;
1244 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1245 			} else {
1246 				break;
1247 			}
1248 		} else {
1249 			if (ret > 0) {
1250 				i += ret;
1251 				last_len += ret << PAGE_CACHE_SHIFT;
1252 			} else {
1253 				i++;
1254 				last_len = 0;
1255 			}
1256 		}
1257 	}
1258 
1259 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1260 		filemap_flush(inode->i_mapping);
1261 
1262 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1263 		/* the filemap_flush will queue IO into the worker threads, but
1264 		 * we have to make sure the IO is actually started and that
1265 		 * ordered extents get created before we return
1266 		 */
1267 		atomic_inc(&root->fs_info->async_submit_draining);
1268 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1269 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1270 			wait_event(root->fs_info->async_submit_wait,
1271 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1272 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1273 		}
1274 		atomic_dec(&root->fs_info->async_submit_draining);
1275 
1276 		mutex_lock(&inode->i_mutex);
1277 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1278 		mutex_unlock(&inode->i_mutex);
1279 	}
1280 
1281 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1282 		btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1283 	}
1284 
1285 	ret = defrag_count;
1286 
1287 out_ra:
1288 	if (!file)
1289 		kfree(ra);
1290 	kfree(pages);
1291 	return ret;
1292 }
1293 
1294 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1295 					void __user *arg)
1296 {
1297 	u64 new_size;
1298 	u64 old_size;
1299 	u64 devid = 1;
1300 	struct btrfs_ioctl_vol_args *vol_args;
1301 	struct btrfs_trans_handle *trans;
1302 	struct btrfs_device *device = NULL;
1303 	char *sizestr;
1304 	char *devstr = NULL;
1305 	int ret = 0;
1306 	int mod = 0;
1307 
1308 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1309 		return -EROFS;
1310 
1311 	if (!capable(CAP_SYS_ADMIN))
1312 		return -EPERM;
1313 
1314 	mutex_lock(&root->fs_info->volume_mutex);
1315 	if (root->fs_info->balance_ctl) {
1316 		printk(KERN_INFO "btrfs: balance in progress\n");
1317 		ret = -EINVAL;
1318 		goto out;
1319 	}
1320 
1321 	vol_args = memdup_user(arg, sizeof(*vol_args));
1322 	if (IS_ERR(vol_args)) {
1323 		ret = PTR_ERR(vol_args);
1324 		goto out;
1325 	}
1326 
1327 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1328 
1329 	sizestr = vol_args->name;
1330 	devstr = strchr(sizestr, ':');
1331 	if (devstr) {
1332 		char *end;
1333 		sizestr = devstr + 1;
1334 		*devstr = '\0';
1335 		devstr = vol_args->name;
1336 		devid = simple_strtoull(devstr, &end, 10);
1337 		printk(KERN_INFO "btrfs: resizing devid %llu\n",
1338 		       (unsigned long long)devid);
1339 	}
1340 	device = btrfs_find_device(root, devid, NULL, NULL);
1341 	if (!device) {
1342 		printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1343 		       (unsigned long long)devid);
1344 		ret = -EINVAL;
1345 		goto out_free;
1346 	}
1347 	if (device->fs_devices && device->fs_devices->seeding) {
1348 		printk(KERN_INFO "btrfs: resizer unable to apply on "
1349 		       "seeding device %llu\n",
1350 		       (unsigned long long)devid);
1351 		ret = -EINVAL;
1352 		goto out_free;
1353 	}
1354 
1355 	if (!strcmp(sizestr, "max"))
1356 		new_size = device->bdev->bd_inode->i_size;
1357 	else {
1358 		if (sizestr[0] == '-') {
1359 			mod = -1;
1360 			sizestr++;
1361 		} else if (sizestr[0] == '+') {
1362 			mod = 1;
1363 			sizestr++;
1364 		}
1365 		new_size = memparse(sizestr, NULL);
1366 		if (new_size == 0) {
1367 			ret = -EINVAL;
1368 			goto out_free;
1369 		}
1370 	}
1371 
1372 	old_size = device->total_bytes;
1373 
1374 	if (mod < 0) {
1375 		if (new_size > old_size) {
1376 			ret = -EINVAL;
1377 			goto out_free;
1378 		}
1379 		new_size = old_size - new_size;
1380 	} else if (mod > 0) {
1381 		new_size = old_size + new_size;
1382 	}
1383 
1384 	if (new_size < 256 * 1024 * 1024) {
1385 		ret = -EINVAL;
1386 		goto out_free;
1387 	}
1388 	if (new_size > device->bdev->bd_inode->i_size) {
1389 		ret = -EFBIG;
1390 		goto out_free;
1391 	}
1392 
1393 	do_div(new_size, root->sectorsize);
1394 	new_size *= root->sectorsize;
1395 
1396 	printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1397 		      rcu_str_deref(device->name),
1398 		      (unsigned long long)new_size);
1399 
1400 	if (new_size > old_size) {
1401 		trans = btrfs_start_transaction(root, 0);
1402 		if (IS_ERR(trans)) {
1403 			ret = PTR_ERR(trans);
1404 			goto out_free;
1405 		}
1406 		ret = btrfs_grow_device(trans, device, new_size);
1407 		btrfs_commit_transaction(trans, root);
1408 	} else if (new_size < old_size) {
1409 		ret = btrfs_shrink_device(device, new_size);
1410 	}
1411 
1412 out_free:
1413 	kfree(vol_args);
1414 out:
1415 	mutex_unlock(&root->fs_info->volume_mutex);
1416 	return ret;
1417 }
1418 
1419 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1420 				char *name, unsigned long fd, int subvol,
1421 				u64 *transid, bool readonly,
1422 				struct btrfs_qgroup_inherit **inherit)
1423 {
1424 	int namelen;
1425 	int ret = 0;
1426 
1427 	ret = mnt_want_write_file(file);
1428 	if (ret)
1429 		goto out;
1430 
1431 	namelen = strlen(name);
1432 	if (strchr(name, '/')) {
1433 		ret = -EINVAL;
1434 		goto out_drop_write;
1435 	}
1436 
1437 	if (name[0] == '.' &&
1438 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1439 		ret = -EEXIST;
1440 		goto out_drop_write;
1441 	}
1442 
1443 	if (subvol) {
1444 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1445 				     NULL, transid, readonly, inherit);
1446 	} else {
1447 		struct fd src = fdget(fd);
1448 		struct inode *src_inode;
1449 		if (!src.file) {
1450 			ret = -EINVAL;
1451 			goto out_drop_write;
1452 		}
1453 
1454 		src_inode = src.file->f_path.dentry->d_inode;
1455 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1456 			printk(KERN_INFO "btrfs: Snapshot src from "
1457 			       "another FS\n");
1458 			ret = -EINVAL;
1459 		} else {
1460 			ret = btrfs_mksubvol(&file->f_path, name, namelen,
1461 					     BTRFS_I(src_inode)->root,
1462 					     transid, readonly, inherit);
1463 		}
1464 		fdput(src);
1465 	}
1466 out_drop_write:
1467 	mnt_drop_write_file(file);
1468 out:
1469 	return ret;
1470 }
1471 
1472 static noinline int btrfs_ioctl_snap_create(struct file *file,
1473 					    void __user *arg, int subvol)
1474 {
1475 	struct btrfs_ioctl_vol_args *vol_args;
1476 	int ret;
1477 
1478 	vol_args = memdup_user(arg, sizeof(*vol_args));
1479 	if (IS_ERR(vol_args))
1480 		return PTR_ERR(vol_args);
1481 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1482 
1483 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1484 					      vol_args->fd, subvol,
1485 					      NULL, false, NULL);
1486 
1487 	kfree(vol_args);
1488 	return ret;
1489 }
1490 
1491 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1492 					       void __user *arg, int subvol)
1493 {
1494 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1495 	int ret;
1496 	u64 transid = 0;
1497 	u64 *ptr = NULL;
1498 	bool readonly = false;
1499 	struct btrfs_qgroup_inherit *inherit = NULL;
1500 
1501 	vol_args = memdup_user(arg, sizeof(*vol_args));
1502 	if (IS_ERR(vol_args))
1503 		return PTR_ERR(vol_args);
1504 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1505 
1506 	if (vol_args->flags &
1507 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1508 	      BTRFS_SUBVOL_QGROUP_INHERIT)) {
1509 		ret = -EOPNOTSUPP;
1510 		goto out;
1511 	}
1512 
1513 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1514 		ptr = &transid;
1515 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1516 		readonly = true;
1517 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1518 		if (vol_args->size > PAGE_CACHE_SIZE) {
1519 			ret = -EINVAL;
1520 			goto out;
1521 		}
1522 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1523 		if (IS_ERR(inherit)) {
1524 			ret = PTR_ERR(inherit);
1525 			goto out;
1526 		}
1527 	}
1528 
1529 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1530 					      vol_args->fd, subvol, ptr,
1531 					      readonly, &inherit);
1532 
1533 	if (ret == 0 && ptr &&
1534 	    copy_to_user(arg +
1535 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1536 				  transid), ptr, sizeof(*ptr)))
1537 		ret = -EFAULT;
1538 out:
1539 	kfree(vol_args);
1540 	kfree(inherit);
1541 	return ret;
1542 }
1543 
1544 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1545 						void __user *arg)
1546 {
1547 	struct inode *inode = fdentry(file)->d_inode;
1548 	struct btrfs_root *root = BTRFS_I(inode)->root;
1549 	int ret = 0;
1550 	u64 flags = 0;
1551 
1552 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1553 		return -EINVAL;
1554 
1555 	down_read(&root->fs_info->subvol_sem);
1556 	if (btrfs_root_readonly(root))
1557 		flags |= BTRFS_SUBVOL_RDONLY;
1558 	up_read(&root->fs_info->subvol_sem);
1559 
1560 	if (copy_to_user(arg, &flags, sizeof(flags)))
1561 		ret = -EFAULT;
1562 
1563 	return ret;
1564 }
1565 
1566 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1567 					      void __user *arg)
1568 {
1569 	struct inode *inode = fdentry(file)->d_inode;
1570 	struct btrfs_root *root = BTRFS_I(inode)->root;
1571 	struct btrfs_trans_handle *trans;
1572 	u64 root_flags;
1573 	u64 flags;
1574 	int ret = 0;
1575 
1576 	ret = mnt_want_write_file(file);
1577 	if (ret)
1578 		goto out;
1579 
1580 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1581 		ret = -EINVAL;
1582 		goto out_drop_write;
1583 	}
1584 
1585 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1586 		ret = -EFAULT;
1587 		goto out_drop_write;
1588 	}
1589 
1590 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1591 		ret = -EINVAL;
1592 		goto out_drop_write;
1593 	}
1594 
1595 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1596 		ret = -EOPNOTSUPP;
1597 		goto out_drop_write;
1598 	}
1599 
1600 	if (!inode_owner_or_capable(inode)) {
1601 		ret = -EACCES;
1602 		goto out_drop_write;
1603 	}
1604 
1605 	down_write(&root->fs_info->subvol_sem);
1606 
1607 	/* nothing to do */
1608 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1609 		goto out_drop_sem;
1610 
1611 	root_flags = btrfs_root_flags(&root->root_item);
1612 	if (flags & BTRFS_SUBVOL_RDONLY)
1613 		btrfs_set_root_flags(&root->root_item,
1614 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1615 	else
1616 		btrfs_set_root_flags(&root->root_item,
1617 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1618 
1619 	trans = btrfs_start_transaction(root, 1);
1620 	if (IS_ERR(trans)) {
1621 		ret = PTR_ERR(trans);
1622 		goto out_reset;
1623 	}
1624 
1625 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1626 				&root->root_key, &root->root_item);
1627 
1628 	btrfs_commit_transaction(trans, root);
1629 out_reset:
1630 	if (ret)
1631 		btrfs_set_root_flags(&root->root_item, root_flags);
1632 out_drop_sem:
1633 	up_write(&root->fs_info->subvol_sem);
1634 out_drop_write:
1635 	mnt_drop_write_file(file);
1636 out:
1637 	return ret;
1638 }
1639 
1640 /*
1641  * helper to check if the subvolume references other subvolumes
1642  */
1643 static noinline int may_destroy_subvol(struct btrfs_root *root)
1644 {
1645 	struct btrfs_path *path;
1646 	struct btrfs_key key;
1647 	int ret;
1648 
1649 	path = btrfs_alloc_path();
1650 	if (!path)
1651 		return -ENOMEM;
1652 
1653 	key.objectid = root->root_key.objectid;
1654 	key.type = BTRFS_ROOT_REF_KEY;
1655 	key.offset = (u64)-1;
1656 
1657 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1658 				&key, path, 0, 0);
1659 	if (ret < 0)
1660 		goto out;
1661 	BUG_ON(ret == 0);
1662 
1663 	ret = 0;
1664 	if (path->slots[0] > 0) {
1665 		path->slots[0]--;
1666 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1667 		if (key.objectid == root->root_key.objectid &&
1668 		    key.type == BTRFS_ROOT_REF_KEY)
1669 			ret = -ENOTEMPTY;
1670 	}
1671 out:
1672 	btrfs_free_path(path);
1673 	return ret;
1674 }
1675 
1676 static noinline int key_in_sk(struct btrfs_key *key,
1677 			      struct btrfs_ioctl_search_key *sk)
1678 {
1679 	struct btrfs_key test;
1680 	int ret;
1681 
1682 	test.objectid = sk->min_objectid;
1683 	test.type = sk->min_type;
1684 	test.offset = sk->min_offset;
1685 
1686 	ret = btrfs_comp_cpu_keys(key, &test);
1687 	if (ret < 0)
1688 		return 0;
1689 
1690 	test.objectid = sk->max_objectid;
1691 	test.type = sk->max_type;
1692 	test.offset = sk->max_offset;
1693 
1694 	ret = btrfs_comp_cpu_keys(key, &test);
1695 	if (ret > 0)
1696 		return 0;
1697 	return 1;
1698 }
1699 
1700 static noinline int copy_to_sk(struct btrfs_root *root,
1701 			       struct btrfs_path *path,
1702 			       struct btrfs_key *key,
1703 			       struct btrfs_ioctl_search_key *sk,
1704 			       char *buf,
1705 			       unsigned long *sk_offset,
1706 			       int *num_found)
1707 {
1708 	u64 found_transid;
1709 	struct extent_buffer *leaf;
1710 	struct btrfs_ioctl_search_header sh;
1711 	unsigned long item_off;
1712 	unsigned long item_len;
1713 	int nritems;
1714 	int i;
1715 	int slot;
1716 	int ret = 0;
1717 
1718 	leaf = path->nodes[0];
1719 	slot = path->slots[0];
1720 	nritems = btrfs_header_nritems(leaf);
1721 
1722 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1723 		i = nritems;
1724 		goto advance_key;
1725 	}
1726 	found_transid = btrfs_header_generation(leaf);
1727 
1728 	for (i = slot; i < nritems; i++) {
1729 		item_off = btrfs_item_ptr_offset(leaf, i);
1730 		item_len = btrfs_item_size_nr(leaf, i);
1731 
1732 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1733 			item_len = 0;
1734 
1735 		if (sizeof(sh) + item_len + *sk_offset >
1736 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1737 			ret = 1;
1738 			goto overflow;
1739 		}
1740 
1741 		btrfs_item_key_to_cpu(leaf, key, i);
1742 		if (!key_in_sk(key, sk))
1743 			continue;
1744 
1745 		sh.objectid = key->objectid;
1746 		sh.offset = key->offset;
1747 		sh.type = key->type;
1748 		sh.len = item_len;
1749 		sh.transid = found_transid;
1750 
1751 		/* copy search result header */
1752 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1753 		*sk_offset += sizeof(sh);
1754 
1755 		if (item_len) {
1756 			char *p = buf + *sk_offset;
1757 			/* copy the item */
1758 			read_extent_buffer(leaf, p,
1759 					   item_off, item_len);
1760 			*sk_offset += item_len;
1761 		}
1762 		(*num_found)++;
1763 
1764 		if (*num_found >= sk->nr_items)
1765 			break;
1766 	}
1767 advance_key:
1768 	ret = 0;
1769 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1770 		key->offset++;
1771 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1772 		key->offset = 0;
1773 		key->type++;
1774 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1775 		key->offset = 0;
1776 		key->type = 0;
1777 		key->objectid++;
1778 	} else
1779 		ret = 1;
1780 overflow:
1781 	return ret;
1782 }
1783 
1784 static noinline int search_ioctl(struct inode *inode,
1785 				 struct btrfs_ioctl_search_args *args)
1786 {
1787 	struct btrfs_root *root;
1788 	struct btrfs_key key;
1789 	struct btrfs_key max_key;
1790 	struct btrfs_path *path;
1791 	struct btrfs_ioctl_search_key *sk = &args->key;
1792 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1793 	int ret;
1794 	int num_found = 0;
1795 	unsigned long sk_offset = 0;
1796 
1797 	path = btrfs_alloc_path();
1798 	if (!path)
1799 		return -ENOMEM;
1800 
1801 	if (sk->tree_id == 0) {
1802 		/* search the root of the inode that was passed */
1803 		root = BTRFS_I(inode)->root;
1804 	} else {
1805 		key.objectid = sk->tree_id;
1806 		key.type = BTRFS_ROOT_ITEM_KEY;
1807 		key.offset = (u64)-1;
1808 		root = btrfs_read_fs_root_no_name(info, &key);
1809 		if (IS_ERR(root)) {
1810 			printk(KERN_ERR "could not find root %llu\n",
1811 			       sk->tree_id);
1812 			btrfs_free_path(path);
1813 			return -ENOENT;
1814 		}
1815 	}
1816 
1817 	key.objectid = sk->min_objectid;
1818 	key.type = sk->min_type;
1819 	key.offset = sk->min_offset;
1820 
1821 	max_key.objectid = sk->max_objectid;
1822 	max_key.type = sk->max_type;
1823 	max_key.offset = sk->max_offset;
1824 
1825 	path->keep_locks = 1;
1826 
1827 	while(1) {
1828 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1829 					   sk->min_transid);
1830 		if (ret != 0) {
1831 			if (ret > 0)
1832 				ret = 0;
1833 			goto err;
1834 		}
1835 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1836 				 &sk_offset, &num_found);
1837 		btrfs_release_path(path);
1838 		if (ret || num_found >= sk->nr_items)
1839 			break;
1840 
1841 	}
1842 	ret = 0;
1843 err:
1844 	sk->nr_items = num_found;
1845 	btrfs_free_path(path);
1846 	return ret;
1847 }
1848 
1849 static noinline int btrfs_ioctl_tree_search(struct file *file,
1850 					   void __user *argp)
1851 {
1852 	 struct btrfs_ioctl_search_args *args;
1853 	 struct inode *inode;
1854 	 int ret;
1855 
1856 	if (!capable(CAP_SYS_ADMIN))
1857 		return -EPERM;
1858 
1859 	args = memdup_user(argp, sizeof(*args));
1860 	if (IS_ERR(args))
1861 		return PTR_ERR(args);
1862 
1863 	inode = fdentry(file)->d_inode;
1864 	ret = search_ioctl(inode, args);
1865 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1866 		ret = -EFAULT;
1867 	kfree(args);
1868 	return ret;
1869 }
1870 
1871 /*
1872  * Search INODE_REFs to identify path name of 'dirid' directory
1873  * in a 'tree_id' tree. and sets path name to 'name'.
1874  */
1875 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1876 				u64 tree_id, u64 dirid, char *name)
1877 {
1878 	struct btrfs_root *root;
1879 	struct btrfs_key key;
1880 	char *ptr;
1881 	int ret = -1;
1882 	int slot;
1883 	int len;
1884 	int total_len = 0;
1885 	struct btrfs_inode_ref *iref;
1886 	struct extent_buffer *l;
1887 	struct btrfs_path *path;
1888 
1889 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1890 		name[0]='\0';
1891 		return 0;
1892 	}
1893 
1894 	path = btrfs_alloc_path();
1895 	if (!path)
1896 		return -ENOMEM;
1897 
1898 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1899 
1900 	key.objectid = tree_id;
1901 	key.type = BTRFS_ROOT_ITEM_KEY;
1902 	key.offset = (u64)-1;
1903 	root = btrfs_read_fs_root_no_name(info, &key);
1904 	if (IS_ERR(root)) {
1905 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1906 		ret = -ENOENT;
1907 		goto out;
1908 	}
1909 
1910 	key.objectid = dirid;
1911 	key.type = BTRFS_INODE_REF_KEY;
1912 	key.offset = (u64)-1;
1913 
1914 	while(1) {
1915 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1916 		if (ret < 0)
1917 			goto out;
1918 
1919 		l = path->nodes[0];
1920 		slot = path->slots[0];
1921 		if (ret > 0 && slot > 0)
1922 			slot--;
1923 		btrfs_item_key_to_cpu(l, &key, slot);
1924 
1925 		if (ret > 0 && (key.objectid != dirid ||
1926 				key.type != BTRFS_INODE_REF_KEY)) {
1927 			ret = -ENOENT;
1928 			goto out;
1929 		}
1930 
1931 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1932 		len = btrfs_inode_ref_name_len(l, iref);
1933 		ptr -= len + 1;
1934 		total_len += len + 1;
1935 		if (ptr < name)
1936 			goto out;
1937 
1938 		*(ptr + len) = '/';
1939 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1940 
1941 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1942 			break;
1943 
1944 		btrfs_release_path(path);
1945 		key.objectid = key.offset;
1946 		key.offset = (u64)-1;
1947 		dirid = key.objectid;
1948 	}
1949 	if (ptr < name)
1950 		goto out;
1951 	memmove(name, ptr, total_len);
1952 	name[total_len]='\0';
1953 	ret = 0;
1954 out:
1955 	btrfs_free_path(path);
1956 	return ret;
1957 }
1958 
1959 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1960 					   void __user *argp)
1961 {
1962 	 struct btrfs_ioctl_ino_lookup_args *args;
1963 	 struct inode *inode;
1964 	 int ret;
1965 
1966 	if (!capable(CAP_SYS_ADMIN))
1967 		return -EPERM;
1968 
1969 	args = memdup_user(argp, sizeof(*args));
1970 	if (IS_ERR(args))
1971 		return PTR_ERR(args);
1972 
1973 	inode = fdentry(file)->d_inode;
1974 
1975 	if (args->treeid == 0)
1976 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1977 
1978 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1979 					args->treeid, args->objectid,
1980 					args->name);
1981 
1982 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1983 		ret = -EFAULT;
1984 
1985 	kfree(args);
1986 	return ret;
1987 }
1988 
1989 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1990 					     void __user *arg)
1991 {
1992 	struct dentry *parent = fdentry(file);
1993 	struct dentry *dentry;
1994 	struct inode *dir = parent->d_inode;
1995 	struct inode *inode;
1996 	struct btrfs_root *root = BTRFS_I(dir)->root;
1997 	struct btrfs_root *dest = NULL;
1998 	struct btrfs_ioctl_vol_args *vol_args;
1999 	struct btrfs_trans_handle *trans;
2000 	int namelen;
2001 	int ret;
2002 	int err = 0;
2003 
2004 	vol_args = memdup_user(arg, sizeof(*vol_args));
2005 	if (IS_ERR(vol_args))
2006 		return PTR_ERR(vol_args);
2007 
2008 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2009 	namelen = strlen(vol_args->name);
2010 	if (strchr(vol_args->name, '/') ||
2011 	    strncmp(vol_args->name, "..", namelen) == 0) {
2012 		err = -EINVAL;
2013 		goto out;
2014 	}
2015 
2016 	err = mnt_want_write_file(file);
2017 	if (err)
2018 		goto out;
2019 
2020 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2021 	dentry = lookup_one_len(vol_args->name, parent, namelen);
2022 	if (IS_ERR(dentry)) {
2023 		err = PTR_ERR(dentry);
2024 		goto out_unlock_dir;
2025 	}
2026 
2027 	if (!dentry->d_inode) {
2028 		err = -ENOENT;
2029 		goto out_dput;
2030 	}
2031 
2032 	inode = dentry->d_inode;
2033 	dest = BTRFS_I(inode)->root;
2034 	if (!capable(CAP_SYS_ADMIN)){
2035 		/*
2036 		 * Regular user.  Only allow this with a special mount
2037 		 * option, when the user has write+exec access to the
2038 		 * subvol root, and when rmdir(2) would have been
2039 		 * allowed.
2040 		 *
2041 		 * Note that this is _not_ check that the subvol is
2042 		 * empty or doesn't contain data that we wouldn't
2043 		 * otherwise be able to delete.
2044 		 *
2045 		 * Users who want to delete empty subvols should try
2046 		 * rmdir(2).
2047 		 */
2048 		err = -EPERM;
2049 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2050 			goto out_dput;
2051 
2052 		/*
2053 		 * Do not allow deletion if the parent dir is the same
2054 		 * as the dir to be deleted.  That means the ioctl
2055 		 * must be called on the dentry referencing the root
2056 		 * of the subvol, not a random directory contained
2057 		 * within it.
2058 		 */
2059 		err = -EINVAL;
2060 		if (root == dest)
2061 			goto out_dput;
2062 
2063 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2064 		if (err)
2065 			goto out_dput;
2066 
2067 		/* check if subvolume may be deleted by a non-root user */
2068 		err = btrfs_may_delete(dir, dentry, 1);
2069 		if (err)
2070 			goto out_dput;
2071 	}
2072 
2073 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2074 		err = -EINVAL;
2075 		goto out_dput;
2076 	}
2077 
2078 	mutex_lock(&inode->i_mutex);
2079 	err = d_invalidate(dentry);
2080 	if (err)
2081 		goto out_unlock;
2082 
2083 	down_write(&root->fs_info->subvol_sem);
2084 
2085 	err = may_destroy_subvol(dest);
2086 	if (err)
2087 		goto out_up_write;
2088 
2089 	trans = btrfs_start_transaction(root, 0);
2090 	if (IS_ERR(trans)) {
2091 		err = PTR_ERR(trans);
2092 		goto out_up_write;
2093 	}
2094 	trans->block_rsv = &root->fs_info->global_block_rsv;
2095 
2096 	ret = btrfs_unlink_subvol(trans, root, dir,
2097 				dest->root_key.objectid,
2098 				dentry->d_name.name,
2099 				dentry->d_name.len);
2100 	if (ret) {
2101 		err = ret;
2102 		btrfs_abort_transaction(trans, root, ret);
2103 		goto out_end_trans;
2104 	}
2105 
2106 	btrfs_record_root_in_trans(trans, dest);
2107 
2108 	memset(&dest->root_item.drop_progress, 0,
2109 		sizeof(dest->root_item.drop_progress));
2110 	dest->root_item.drop_level = 0;
2111 	btrfs_set_root_refs(&dest->root_item, 0);
2112 
2113 	if (!xchg(&dest->orphan_item_inserted, 1)) {
2114 		ret = btrfs_insert_orphan_item(trans,
2115 					root->fs_info->tree_root,
2116 					dest->root_key.objectid);
2117 		if (ret) {
2118 			btrfs_abort_transaction(trans, root, ret);
2119 			err = ret;
2120 			goto out_end_trans;
2121 		}
2122 	}
2123 out_end_trans:
2124 	ret = btrfs_end_transaction(trans, root);
2125 	if (ret && !err)
2126 		err = ret;
2127 	inode->i_flags |= S_DEAD;
2128 out_up_write:
2129 	up_write(&root->fs_info->subvol_sem);
2130 out_unlock:
2131 	mutex_unlock(&inode->i_mutex);
2132 	if (!err) {
2133 		shrink_dcache_sb(root->fs_info->sb);
2134 		btrfs_invalidate_inodes(dest);
2135 		d_delete(dentry);
2136 	}
2137 out_dput:
2138 	dput(dentry);
2139 out_unlock_dir:
2140 	mutex_unlock(&dir->i_mutex);
2141 	mnt_drop_write_file(file);
2142 out:
2143 	kfree(vol_args);
2144 	return err;
2145 }
2146 
2147 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2148 {
2149 	struct inode *inode = fdentry(file)->d_inode;
2150 	struct btrfs_root *root = BTRFS_I(inode)->root;
2151 	struct btrfs_ioctl_defrag_range_args *range;
2152 	int ret;
2153 
2154 	if (btrfs_root_readonly(root))
2155 		return -EROFS;
2156 
2157 	ret = mnt_want_write_file(file);
2158 	if (ret)
2159 		return ret;
2160 
2161 	switch (inode->i_mode & S_IFMT) {
2162 	case S_IFDIR:
2163 		if (!capable(CAP_SYS_ADMIN)) {
2164 			ret = -EPERM;
2165 			goto out;
2166 		}
2167 		ret = btrfs_defrag_root(root, 0);
2168 		if (ret)
2169 			goto out;
2170 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2171 		break;
2172 	case S_IFREG:
2173 		if (!(file->f_mode & FMODE_WRITE)) {
2174 			ret = -EINVAL;
2175 			goto out;
2176 		}
2177 
2178 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2179 		if (!range) {
2180 			ret = -ENOMEM;
2181 			goto out;
2182 		}
2183 
2184 		if (argp) {
2185 			if (copy_from_user(range, argp,
2186 					   sizeof(*range))) {
2187 				ret = -EFAULT;
2188 				kfree(range);
2189 				goto out;
2190 			}
2191 			/* compression requires us to start the IO */
2192 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2193 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2194 				range->extent_thresh = (u32)-1;
2195 			}
2196 		} else {
2197 			/* the rest are all set to zero by kzalloc */
2198 			range->len = (u64)-1;
2199 		}
2200 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2201 					range, 0, 0);
2202 		if (ret > 0)
2203 			ret = 0;
2204 		kfree(range);
2205 		break;
2206 	default:
2207 		ret = -EINVAL;
2208 	}
2209 out:
2210 	mnt_drop_write_file(file);
2211 	return ret;
2212 }
2213 
2214 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2215 {
2216 	struct btrfs_ioctl_vol_args *vol_args;
2217 	int ret;
2218 
2219 	if (!capable(CAP_SYS_ADMIN))
2220 		return -EPERM;
2221 
2222 	mutex_lock(&root->fs_info->volume_mutex);
2223 	if (root->fs_info->balance_ctl) {
2224 		printk(KERN_INFO "btrfs: balance in progress\n");
2225 		ret = -EINVAL;
2226 		goto out;
2227 	}
2228 
2229 	vol_args = memdup_user(arg, sizeof(*vol_args));
2230 	if (IS_ERR(vol_args)) {
2231 		ret = PTR_ERR(vol_args);
2232 		goto out;
2233 	}
2234 
2235 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2236 	ret = btrfs_init_new_device(root, vol_args->name);
2237 
2238 	kfree(vol_args);
2239 out:
2240 	mutex_unlock(&root->fs_info->volume_mutex);
2241 	return ret;
2242 }
2243 
2244 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2245 {
2246 	struct btrfs_ioctl_vol_args *vol_args;
2247 	int ret;
2248 
2249 	if (!capable(CAP_SYS_ADMIN))
2250 		return -EPERM;
2251 
2252 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2253 		return -EROFS;
2254 
2255 	mutex_lock(&root->fs_info->volume_mutex);
2256 	if (root->fs_info->balance_ctl) {
2257 		printk(KERN_INFO "btrfs: balance in progress\n");
2258 		ret = -EINVAL;
2259 		goto out;
2260 	}
2261 
2262 	vol_args = memdup_user(arg, sizeof(*vol_args));
2263 	if (IS_ERR(vol_args)) {
2264 		ret = PTR_ERR(vol_args);
2265 		goto out;
2266 	}
2267 
2268 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2269 	ret = btrfs_rm_device(root, vol_args->name);
2270 
2271 	kfree(vol_args);
2272 out:
2273 	mutex_unlock(&root->fs_info->volume_mutex);
2274 	return ret;
2275 }
2276 
2277 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2278 {
2279 	struct btrfs_ioctl_fs_info_args *fi_args;
2280 	struct btrfs_device *device;
2281 	struct btrfs_device *next;
2282 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2283 	int ret = 0;
2284 
2285 	if (!capable(CAP_SYS_ADMIN))
2286 		return -EPERM;
2287 
2288 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2289 	if (!fi_args)
2290 		return -ENOMEM;
2291 
2292 	fi_args->num_devices = fs_devices->num_devices;
2293 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2294 
2295 	mutex_lock(&fs_devices->device_list_mutex);
2296 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2297 		if (device->devid > fi_args->max_id)
2298 			fi_args->max_id = device->devid;
2299 	}
2300 	mutex_unlock(&fs_devices->device_list_mutex);
2301 
2302 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2303 		ret = -EFAULT;
2304 
2305 	kfree(fi_args);
2306 	return ret;
2307 }
2308 
2309 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2310 {
2311 	struct btrfs_ioctl_dev_info_args *di_args;
2312 	struct btrfs_device *dev;
2313 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2314 	int ret = 0;
2315 	char *s_uuid = NULL;
2316 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2317 
2318 	if (!capable(CAP_SYS_ADMIN))
2319 		return -EPERM;
2320 
2321 	di_args = memdup_user(arg, sizeof(*di_args));
2322 	if (IS_ERR(di_args))
2323 		return PTR_ERR(di_args);
2324 
2325 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2326 		s_uuid = di_args->uuid;
2327 
2328 	mutex_lock(&fs_devices->device_list_mutex);
2329 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2330 	mutex_unlock(&fs_devices->device_list_mutex);
2331 
2332 	if (!dev) {
2333 		ret = -ENODEV;
2334 		goto out;
2335 	}
2336 
2337 	di_args->devid = dev->devid;
2338 	di_args->bytes_used = dev->bytes_used;
2339 	di_args->total_bytes = dev->total_bytes;
2340 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2341 	if (dev->name) {
2342 		struct rcu_string *name;
2343 
2344 		rcu_read_lock();
2345 		name = rcu_dereference(dev->name);
2346 		strncpy(di_args->path, name->str, sizeof(di_args->path));
2347 		rcu_read_unlock();
2348 		di_args->path[sizeof(di_args->path) - 1] = 0;
2349 	} else {
2350 		di_args->path[0] = '\0';
2351 	}
2352 
2353 out:
2354 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2355 		ret = -EFAULT;
2356 
2357 	kfree(di_args);
2358 	return ret;
2359 }
2360 
2361 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2362 				       u64 off, u64 olen, u64 destoff)
2363 {
2364 	struct inode *inode = fdentry(file)->d_inode;
2365 	struct btrfs_root *root = BTRFS_I(inode)->root;
2366 	struct fd src_file;
2367 	struct inode *src;
2368 	struct btrfs_trans_handle *trans;
2369 	struct btrfs_path *path;
2370 	struct extent_buffer *leaf;
2371 	char *buf;
2372 	struct btrfs_key key;
2373 	u32 nritems;
2374 	int slot;
2375 	int ret;
2376 	u64 len = olen;
2377 	u64 bs = root->fs_info->sb->s_blocksize;
2378 
2379 	/*
2380 	 * TODO:
2381 	 * - split compressed inline extents.  annoying: we need to
2382 	 *   decompress into destination's address_space (the file offset
2383 	 *   may change, so source mapping won't do), then recompress (or
2384 	 *   otherwise reinsert) a subrange.
2385 	 * - allow ranges within the same file to be cloned (provided
2386 	 *   they don't overlap)?
2387 	 */
2388 
2389 	/* the destination must be opened for writing */
2390 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2391 		return -EINVAL;
2392 
2393 	if (btrfs_root_readonly(root))
2394 		return -EROFS;
2395 
2396 	ret = mnt_want_write_file(file);
2397 	if (ret)
2398 		return ret;
2399 
2400 	src_file = fdget(srcfd);
2401 	if (!src_file.file) {
2402 		ret = -EBADF;
2403 		goto out_drop_write;
2404 	}
2405 
2406 	ret = -EXDEV;
2407 	if (src_file.file->f_path.mnt != file->f_path.mnt)
2408 		goto out_fput;
2409 
2410 	src = src_file.file->f_dentry->d_inode;
2411 
2412 	ret = -EINVAL;
2413 	if (src == inode)
2414 		goto out_fput;
2415 
2416 	/* the src must be open for reading */
2417 	if (!(src_file.file->f_mode & FMODE_READ))
2418 		goto out_fput;
2419 
2420 	/* don't make the dst file partly checksummed */
2421 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2422 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2423 		goto out_fput;
2424 
2425 	ret = -EISDIR;
2426 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2427 		goto out_fput;
2428 
2429 	ret = -EXDEV;
2430 	if (src->i_sb != inode->i_sb)
2431 		goto out_fput;
2432 
2433 	ret = -ENOMEM;
2434 	buf = vmalloc(btrfs_level_size(root, 0));
2435 	if (!buf)
2436 		goto out_fput;
2437 
2438 	path = btrfs_alloc_path();
2439 	if (!path) {
2440 		vfree(buf);
2441 		goto out_fput;
2442 	}
2443 	path->reada = 2;
2444 
2445 	if (inode < src) {
2446 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2447 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2448 	} else {
2449 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2450 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2451 	}
2452 
2453 	/* determine range to clone */
2454 	ret = -EINVAL;
2455 	if (off + len > src->i_size || off + len < off)
2456 		goto out_unlock;
2457 	if (len == 0)
2458 		olen = len = src->i_size - off;
2459 	/* if we extend to eof, continue to block boundary */
2460 	if (off + len == src->i_size)
2461 		len = ALIGN(src->i_size, bs) - off;
2462 
2463 	/* verify the end result is block aligned */
2464 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2465 	    !IS_ALIGNED(destoff, bs))
2466 		goto out_unlock;
2467 
2468 	if (destoff > inode->i_size) {
2469 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2470 		if (ret)
2471 			goto out_unlock;
2472 	}
2473 
2474 	/* truncate page cache pages from target inode range */
2475 	truncate_inode_pages_range(&inode->i_data, destoff,
2476 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
2477 
2478 	/* do any pending delalloc/csum calc on src, one way or
2479 	   another, and lock file content */
2480 	while (1) {
2481 		struct btrfs_ordered_extent *ordered;
2482 		lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2483 		ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2484 		if (!ordered &&
2485 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2486 				    EXTENT_DELALLOC, 0, NULL))
2487 			break;
2488 		unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2489 		if (ordered)
2490 			btrfs_put_ordered_extent(ordered);
2491 		btrfs_wait_ordered_range(src, off, len);
2492 	}
2493 
2494 	/* clone data */
2495 	key.objectid = btrfs_ino(src);
2496 	key.type = BTRFS_EXTENT_DATA_KEY;
2497 	key.offset = 0;
2498 
2499 	while (1) {
2500 		/*
2501 		 * note the key will change type as we walk through the
2502 		 * tree.
2503 		 */
2504 		ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2505 				0, 0);
2506 		if (ret < 0)
2507 			goto out;
2508 
2509 		nritems = btrfs_header_nritems(path->nodes[0]);
2510 		if (path->slots[0] >= nritems) {
2511 			ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2512 			if (ret < 0)
2513 				goto out;
2514 			if (ret > 0)
2515 				break;
2516 			nritems = btrfs_header_nritems(path->nodes[0]);
2517 		}
2518 		leaf = path->nodes[0];
2519 		slot = path->slots[0];
2520 
2521 		btrfs_item_key_to_cpu(leaf, &key, slot);
2522 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2523 		    key.objectid != btrfs_ino(src))
2524 			break;
2525 
2526 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2527 			struct btrfs_file_extent_item *extent;
2528 			int type;
2529 			u32 size;
2530 			struct btrfs_key new_key;
2531 			u64 disko = 0, diskl = 0;
2532 			u64 datao = 0, datal = 0;
2533 			u8 comp;
2534 			u64 endoff;
2535 
2536 			size = btrfs_item_size_nr(leaf, slot);
2537 			read_extent_buffer(leaf, buf,
2538 					   btrfs_item_ptr_offset(leaf, slot),
2539 					   size);
2540 
2541 			extent = btrfs_item_ptr(leaf, slot,
2542 						struct btrfs_file_extent_item);
2543 			comp = btrfs_file_extent_compression(leaf, extent);
2544 			type = btrfs_file_extent_type(leaf, extent);
2545 			if (type == BTRFS_FILE_EXTENT_REG ||
2546 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2547 				disko = btrfs_file_extent_disk_bytenr(leaf,
2548 								      extent);
2549 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2550 								 extent);
2551 				datao = btrfs_file_extent_offset(leaf, extent);
2552 				datal = btrfs_file_extent_num_bytes(leaf,
2553 								    extent);
2554 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2555 				/* take upper bound, may be compressed */
2556 				datal = btrfs_file_extent_ram_bytes(leaf,
2557 								    extent);
2558 			}
2559 			btrfs_release_path(path);
2560 
2561 			if (key.offset + datal <= off ||
2562 			    key.offset >= off + len - 1)
2563 				goto next;
2564 
2565 			memcpy(&new_key, &key, sizeof(new_key));
2566 			new_key.objectid = btrfs_ino(inode);
2567 			if (off <= key.offset)
2568 				new_key.offset = key.offset + destoff - off;
2569 			else
2570 				new_key.offset = destoff;
2571 
2572 			/*
2573 			 * 1 - adjusting old extent (we may have to split it)
2574 			 * 1 - add new extent
2575 			 * 1 - inode update
2576 			 */
2577 			trans = btrfs_start_transaction(root, 3);
2578 			if (IS_ERR(trans)) {
2579 				ret = PTR_ERR(trans);
2580 				goto out;
2581 			}
2582 
2583 			if (type == BTRFS_FILE_EXTENT_REG ||
2584 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2585 				/*
2586 				 *    a  | --- range to clone ---|  b
2587 				 * | ------------- extent ------------- |
2588 				 */
2589 
2590 				/* substract range b */
2591 				if (key.offset + datal > off + len)
2592 					datal = off + len - key.offset;
2593 
2594 				/* substract range a */
2595 				if (off > key.offset) {
2596 					datao += off - key.offset;
2597 					datal -= off - key.offset;
2598 				}
2599 
2600 				ret = btrfs_drop_extents(trans, root, inode,
2601 							 new_key.offset,
2602 							 new_key.offset + datal,
2603 							 1);
2604 				if (ret) {
2605 					btrfs_abort_transaction(trans, root,
2606 								ret);
2607 					btrfs_end_transaction(trans, root);
2608 					goto out;
2609 				}
2610 
2611 				ret = btrfs_insert_empty_item(trans, root, path,
2612 							      &new_key, size);
2613 				if (ret) {
2614 					btrfs_abort_transaction(trans, root,
2615 								ret);
2616 					btrfs_end_transaction(trans, root);
2617 					goto out;
2618 				}
2619 
2620 				leaf = path->nodes[0];
2621 				slot = path->slots[0];
2622 				write_extent_buffer(leaf, buf,
2623 					    btrfs_item_ptr_offset(leaf, slot),
2624 					    size);
2625 
2626 				extent = btrfs_item_ptr(leaf, slot,
2627 						struct btrfs_file_extent_item);
2628 
2629 				/* disko == 0 means it's a hole */
2630 				if (!disko)
2631 					datao = 0;
2632 
2633 				btrfs_set_file_extent_offset(leaf, extent,
2634 							     datao);
2635 				btrfs_set_file_extent_num_bytes(leaf, extent,
2636 								datal);
2637 				if (disko) {
2638 					inode_add_bytes(inode, datal);
2639 					ret = btrfs_inc_extent_ref(trans, root,
2640 							disko, diskl, 0,
2641 							root->root_key.objectid,
2642 							btrfs_ino(inode),
2643 							new_key.offset - datao,
2644 							0);
2645 					if (ret) {
2646 						btrfs_abort_transaction(trans,
2647 									root,
2648 									ret);
2649 						btrfs_end_transaction(trans,
2650 								      root);
2651 						goto out;
2652 
2653 					}
2654 				}
2655 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2656 				u64 skip = 0;
2657 				u64 trim = 0;
2658 				if (off > key.offset) {
2659 					skip = off - key.offset;
2660 					new_key.offset += skip;
2661 				}
2662 
2663 				if (key.offset + datal > off + len)
2664 					trim = key.offset + datal - (off + len);
2665 
2666 				if (comp && (skip || trim)) {
2667 					ret = -EINVAL;
2668 					btrfs_end_transaction(trans, root);
2669 					goto out;
2670 				}
2671 				size -= skip + trim;
2672 				datal -= skip + trim;
2673 
2674 				ret = btrfs_drop_extents(trans, root, inode,
2675 							 new_key.offset,
2676 							 new_key.offset + datal,
2677 							 1);
2678 				if (ret) {
2679 					btrfs_abort_transaction(trans, root,
2680 								ret);
2681 					btrfs_end_transaction(trans, root);
2682 					goto out;
2683 				}
2684 
2685 				ret = btrfs_insert_empty_item(trans, root, path,
2686 							      &new_key, size);
2687 				if (ret) {
2688 					btrfs_abort_transaction(trans, root,
2689 								ret);
2690 					btrfs_end_transaction(trans, root);
2691 					goto out;
2692 				}
2693 
2694 				if (skip) {
2695 					u32 start =
2696 					  btrfs_file_extent_calc_inline_size(0);
2697 					memmove(buf+start, buf+start+skip,
2698 						datal);
2699 				}
2700 
2701 				leaf = path->nodes[0];
2702 				slot = path->slots[0];
2703 				write_extent_buffer(leaf, buf,
2704 					    btrfs_item_ptr_offset(leaf, slot),
2705 					    size);
2706 				inode_add_bytes(inode, datal);
2707 			}
2708 
2709 			btrfs_mark_buffer_dirty(leaf);
2710 			btrfs_release_path(path);
2711 
2712 			inode_inc_iversion(inode);
2713 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2714 
2715 			/*
2716 			 * we round up to the block size at eof when
2717 			 * determining which extents to clone above,
2718 			 * but shouldn't round up the file size
2719 			 */
2720 			endoff = new_key.offset + datal;
2721 			if (endoff > destoff+olen)
2722 				endoff = destoff+olen;
2723 			if (endoff > inode->i_size)
2724 				btrfs_i_size_write(inode, endoff);
2725 
2726 			ret = btrfs_update_inode(trans, root, inode);
2727 			if (ret) {
2728 				btrfs_abort_transaction(trans, root, ret);
2729 				btrfs_end_transaction(trans, root);
2730 				goto out;
2731 			}
2732 			ret = btrfs_end_transaction(trans, root);
2733 		}
2734 next:
2735 		btrfs_release_path(path);
2736 		key.offset++;
2737 	}
2738 	ret = 0;
2739 out:
2740 	btrfs_release_path(path);
2741 	unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2742 out_unlock:
2743 	mutex_unlock(&src->i_mutex);
2744 	mutex_unlock(&inode->i_mutex);
2745 	vfree(buf);
2746 	btrfs_free_path(path);
2747 out_fput:
2748 	fdput(src_file);
2749 out_drop_write:
2750 	mnt_drop_write_file(file);
2751 	return ret;
2752 }
2753 
2754 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2755 {
2756 	struct btrfs_ioctl_clone_range_args args;
2757 
2758 	if (copy_from_user(&args, argp, sizeof(args)))
2759 		return -EFAULT;
2760 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2761 				 args.src_length, args.dest_offset);
2762 }
2763 
2764 /*
2765  * there are many ways the trans_start and trans_end ioctls can lead
2766  * to deadlocks.  They should only be used by applications that
2767  * basically own the machine, and have a very in depth understanding
2768  * of all the possible deadlocks and enospc problems.
2769  */
2770 static long btrfs_ioctl_trans_start(struct file *file)
2771 {
2772 	struct inode *inode = fdentry(file)->d_inode;
2773 	struct btrfs_root *root = BTRFS_I(inode)->root;
2774 	struct btrfs_trans_handle *trans;
2775 	int ret;
2776 
2777 	ret = -EPERM;
2778 	if (!capable(CAP_SYS_ADMIN))
2779 		goto out;
2780 
2781 	ret = -EINPROGRESS;
2782 	if (file->private_data)
2783 		goto out;
2784 
2785 	ret = -EROFS;
2786 	if (btrfs_root_readonly(root))
2787 		goto out;
2788 
2789 	ret = mnt_want_write_file(file);
2790 	if (ret)
2791 		goto out;
2792 
2793 	atomic_inc(&root->fs_info->open_ioctl_trans);
2794 
2795 	ret = -ENOMEM;
2796 	trans = btrfs_start_ioctl_transaction(root);
2797 	if (IS_ERR(trans))
2798 		goto out_drop;
2799 
2800 	file->private_data = trans;
2801 	return 0;
2802 
2803 out_drop:
2804 	atomic_dec(&root->fs_info->open_ioctl_trans);
2805 	mnt_drop_write_file(file);
2806 out:
2807 	return ret;
2808 }
2809 
2810 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2811 {
2812 	struct inode *inode = fdentry(file)->d_inode;
2813 	struct btrfs_root *root = BTRFS_I(inode)->root;
2814 	struct btrfs_root *new_root;
2815 	struct btrfs_dir_item *di;
2816 	struct btrfs_trans_handle *trans;
2817 	struct btrfs_path *path;
2818 	struct btrfs_key location;
2819 	struct btrfs_disk_key disk_key;
2820 	u64 objectid = 0;
2821 	u64 dir_id;
2822 
2823 	if (!capable(CAP_SYS_ADMIN))
2824 		return -EPERM;
2825 
2826 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2827 		return -EFAULT;
2828 
2829 	if (!objectid)
2830 		objectid = root->root_key.objectid;
2831 
2832 	location.objectid = objectid;
2833 	location.type = BTRFS_ROOT_ITEM_KEY;
2834 	location.offset = (u64)-1;
2835 
2836 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2837 	if (IS_ERR(new_root))
2838 		return PTR_ERR(new_root);
2839 
2840 	if (btrfs_root_refs(&new_root->root_item) == 0)
2841 		return -ENOENT;
2842 
2843 	path = btrfs_alloc_path();
2844 	if (!path)
2845 		return -ENOMEM;
2846 	path->leave_spinning = 1;
2847 
2848 	trans = btrfs_start_transaction(root, 1);
2849 	if (IS_ERR(trans)) {
2850 		btrfs_free_path(path);
2851 		return PTR_ERR(trans);
2852 	}
2853 
2854 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2855 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2856 				   dir_id, "default", 7, 1);
2857 	if (IS_ERR_OR_NULL(di)) {
2858 		btrfs_free_path(path);
2859 		btrfs_end_transaction(trans, root);
2860 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2861 		       "this isn't going to work\n");
2862 		return -ENOENT;
2863 	}
2864 
2865 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2866 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2867 	btrfs_mark_buffer_dirty(path->nodes[0]);
2868 	btrfs_free_path(path);
2869 
2870 	btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2871 	btrfs_end_transaction(trans, root);
2872 
2873 	return 0;
2874 }
2875 
2876 void btrfs_get_block_group_info(struct list_head *groups_list,
2877 				struct btrfs_ioctl_space_info *space)
2878 {
2879 	struct btrfs_block_group_cache *block_group;
2880 
2881 	space->total_bytes = 0;
2882 	space->used_bytes = 0;
2883 	space->flags = 0;
2884 	list_for_each_entry(block_group, groups_list, list) {
2885 		space->flags = block_group->flags;
2886 		space->total_bytes += block_group->key.offset;
2887 		space->used_bytes +=
2888 			btrfs_block_group_used(&block_group->item);
2889 	}
2890 }
2891 
2892 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2893 {
2894 	struct btrfs_ioctl_space_args space_args;
2895 	struct btrfs_ioctl_space_info space;
2896 	struct btrfs_ioctl_space_info *dest;
2897 	struct btrfs_ioctl_space_info *dest_orig;
2898 	struct btrfs_ioctl_space_info __user *user_dest;
2899 	struct btrfs_space_info *info;
2900 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2901 		       BTRFS_BLOCK_GROUP_SYSTEM,
2902 		       BTRFS_BLOCK_GROUP_METADATA,
2903 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2904 	int num_types = 4;
2905 	int alloc_size;
2906 	int ret = 0;
2907 	u64 slot_count = 0;
2908 	int i, c;
2909 
2910 	if (copy_from_user(&space_args,
2911 			   (struct btrfs_ioctl_space_args __user *)arg,
2912 			   sizeof(space_args)))
2913 		return -EFAULT;
2914 
2915 	for (i = 0; i < num_types; i++) {
2916 		struct btrfs_space_info *tmp;
2917 
2918 		info = NULL;
2919 		rcu_read_lock();
2920 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2921 					list) {
2922 			if (tmp->flags == types[i]) {
2923 				info = tmp;
2924 				break;
2925 			}
2926 		}
2927 		rcu_read_unlock();
2928 
2929 		if (!info)
2930 			continue;
2931 
2932 		down_read(&info->groups_sem);
2933 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2934 			if (!list_empty(&info->block_groups[c]))
2935 				slot_count++;
2936 		}
2937 		up_read(&info->groups_sem);
2938 	}
2939 
2940 	/* space_slots == 0 means they are asking for a count */
2941 	if (space_args.space_slots == 0) {
2942 		space_args.total_spaces = slot_count;
2943 		goto out;
2944 	}
2945 
2946 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2947 
2948 	alloc_size = sizeof(*dest) * slot_count;
2949 
2950 	/* we generally have at most 6 or so space infos, one for each raid
2951 	 * level.  So, a whole page should be more than enough for everyone
2952 	 */
2953 	if (alloc_size > PAGE_CACHE_SIZE)
2954 		return -ENOMEM;
2955 
2956 	space_args.total_spaces = 0;
2957 	dest = kmalloc(alloc_size, GFP_NOFS);
2958 	if (!dest)
2959 		return -ENOMEM;
2960 	dest_orig = dest;
2961 
2962 	/* now we have a buffer to copy into */
2963 	for (i = 0; i < num_types; i++) {
2964 		struct btrfs_space_info *tmp;
2965 
2966 		if (!slot_count)
2967 			break;
2968 
2969 		info = NULL;
2970 		rcu_read_lock();
2971 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2972 					list) {
2973 			if (tmp->flags == types[i]) {
2974 				info = tmp;
2975 				break;
2976 			}
2977 		}
2978 		rcu_read_unlock();
2979 
2980 		if (!info)
2981 			continue;
2982 		down_read(&info->groups_sem);
2983 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2984 			if (!list_empty(&info->block_groups[c])) {
2985 				btrfs_get_block_group_info(
2986 					&info->block_groups[c], &space);
2987 				memcpy(dest, &space, sizeof(space));
2988 				dest++;
2989 				space_args.total_spaces++;
2990 				slot_count--;
2991 			}
2992 			if (!slot_count)
2993 				break;
2994 		}
2995 		up_read(&info->groups_sem);
2996 	}
2997 
2998 	user_dest = (struct btrfs_ioctl_space_info __user *)
2999 		(arg + sizeof(struct btrfs_ioctl_space_args));
3000 
3001 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3002 		ret = -EFAULT;
3003 
3004 	kfree(dest_orig);
3005 out:
3006 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3007 		ret = -EFAULT;
3008 
3009 	return ret;
3010 }
3011 
3012 /*
3013  * there are many ways the trans_start and trans_end ioctls can lead
3014  * to deadlocks.  They should only be used by applications that
3015  * basically own the machine, and have a very in depth understanding
3016  * of all the possible deadlocks and enospc problems.
3017  */
3018 long btrfs_ioctl_trans_end(struct file *file)
3019 {
3020 	struct inode *inode = fdentry(file)->d_inode;
3021 	struct btrfs_root *root = BTRFS_I(inode)->root;
3022 	struct btrfs_trans_handle *trans;
3023 
3024 	trans = file->private_data;
3025 	if (!trans)
3026 		return -EINVAL;
3027 	file->private_data = NULL;
3028 
3029 	btrfs_end_transaction(trans, root);
3030 
3031 	atomic_dec(&root->fs_info->open_ioctl_trans);
3032 
3033 	mnt_drop_write_file(file);
3034 	return 0;
3035 }
3036 
3037 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
3038 {
3039 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3040 	struct btrfs_trans_handle *trans;
3041 	u64 transid;
3042 	int ret;
3043 
3044 	trans = btrfs_start_transaction(root, 0);
3045 	if (IS_ERR(trans))
3046 		return PTR_ERR(trans);
3047 	transid = trans->transid;
3048 	ret = btrfs_commit_transaction_async(trans, root, 0);
3049 	if (ret) {
3050 		btrfs_end_transaction(trans, root);
3051 		return ret;
3052 	}
3053 
3054 	if (argp)
3055 		if (copy_to_user(argp, &transid, sizeof(transid)))
3056 			return -EFAULT;
3057 	return 0;
3058 }
3059 
3060 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3061 {
3062 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3063 	u64 transid;
3064 
3065 	if (argp) {
3066 		if (copy_from_user(&transid, argp, sizeof(transid)))
3067 			return -EFAULT;
3068 	} else {
3069 		transid = 0;  /* current trans */
3070 	}
3071 	return btrfs_wait_for_commit(root, transid);
3072 }
3073 
3074 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3075 {
3076 	int ret;
3077 	struct btrfs_ioctl_scrub_args *sa;
3078 
3079 	if (!capable(CAP_SYS_ADMIN))
3080 		return -EPERM;
3081 
3082 	sa = memdup_user(arg, sizeof(*sa));
3083 	if (IS_ERR(sa))
3084 		return PTR_ERR(sa);
3085 
3086 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3087 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3088 
3089 	if (copy_to_user(arg, sa, sizeof(*sa)))
3090 		ret = -EFAULT;
3091 
3092 	kfree(sa);
3093 	return ret;
3094 }
3095 
3096 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3097 {
3098 	if (!capable(CAP_SYS_ADMIN))
3099 		return -EPERM;
3100 
3101 	return btrfs_scrub_cancel(root);
3102 }
3103 
3104 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3105 				       void __user *arg)
3106 {
3107 	struct btrfs_ioctl_scrub_args *sa;
3108 	int ret;
3109 
3110 	if (!capable(CAP_SYS_ADMIN))
3111 		return -EPERM;
3112 
3113 	sa = memdup_user(arg, sizeof(*sa));
3114 	if (IS_ERR(sa))
3115 		return PTR_ERR(sa);
3116 
3117 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3118 
3119 	if (copy_to_user(arg, sa, sizeof(*sa)))
3120 		ret = -EFAULT;
3121 
3122 	kfree(sa);
3123 	return ret;
3124 }
3125 
3126 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3127 				      void __user *arg)
3128 {
3129 	struct btrfs_ioctl_get_dev_stats *sa;
3130 	int ret;
3131 
3132 	sa = memdup_user(arg, sizeof(*sa));
3133 	if (IS_ERR(sa))
3134 		return PTR_ERR(sa);
3135 
3136 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3137 		kfree(sa);
3138 		return -EPERM;
3139 	}
3140 
3141 	ret = btrfs_get_dev_stats(root, sa);
3142 
3143 	if (copy_to_user(arg, sa, sizeof(*sa)))
3144 		ret = -EFAULT;
3145 
3146 	kfree(sa);
3147 	return ret;
3148 }
3149 
3150 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3151 {
3152 	int ret = 0;
3153 	int i;
3154 	u64 rel_ptr;
3155 	int size;
3156 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3157 	struct inode_fs_paths *ipath = NULL;
3158 	struct btrfs_path *path;
3159 
3160 	if (!capable(CAP_SYS_ADMIN))
3161 		return -EPERM;
3162 
3163 	path = btrfs_alloc_path();
3164 	if (!path) {
3165 		ret = -ENOMEM;
3166 		goto out;
3167 	}
3168 
3169 	ipa = memdup_user(arg, sizeof(*ipa));
3170 	if (IS_ERR(ipa)) {
3171 		ret = PTR_ERR(ipa);
3172 		ipa = NULL;
3173 		goto out;
3174 	}
3175 
3176 	size = min_t(u32, ipa->size, 4096);
3177 	ipath = init_ipath(size, root, path);
3178 	if (IS_ERR(ipath)) {
3179 		ret = PTR_ERR(ipath);
3180 		ipath = NULL;
3181 		goto out;
3182 	}
3183 
3184 	ret = paths_from_inode(ipa->inum, ipath);
3185 	if (ret < 0)
3186 		goto out;
3187 
3188 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3189 		rel_ptr = ipath->fspath->val[i] -
3190 			  (u64)(unsigned long)ipath->fspath->val;
3191 		ipath->fspath->val[i] = rel_ptr;
3192 	}
3193 
3194 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3195 			   (void *)(unsigned long)ipath->fspath, size);
3196 	if (ret) {
3197 		ret = -EFAULT;
3198 		goto out;
3199 	}
3200 
3201 out:
3202 	btrfs_free_path(path);
3203 	free_ipath(ipath);
3204 	kfree(ipa);
3205 
3206 	return ret;
3207 }
3208 
3209 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3210 {
3211 	struct btrfs_data_container *inodes = ctx;
3212 	const size_t c = 3 * sizeof(u64);
3213 
3214 	if (inodes->bytes_left >= c) {
3215 		inodes->bytes_left -= c;
3216 		inodes->val[inodes->elem_cnt] = inum;
3217 		inodes->val[inodes->elem_cnt + 1] = offset;
3218 		inodes->val[inodes->elem_cnt + 2] = root;
3219 		inodes->elem_cnt += 3;
3220 	} else {
3221 		inodes->bytes_missing += c - inodes->bytes_left;
3222 		inodes->bytes_left = 0;
3223 		inodes->elem_missed += 3;
3224 	}
3225 
3226 	return 0;
3227 }
3228 
3229 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3230 					void __user *arg)
3231 {
3232 	int ret = 0;
3233 	int size;
3234 	struct btrfs_ioctl_logical_ino_args *loi;
3235 	struct btrfs_data_container *inodes = NULL;
3236 	struct btrfs_path *path = NULL;
3237 
3238 	if (!capable(CAP_SYS_ADMIN))
3239 		return -EPERM;
3240 
3241 	loi = memdup_user(arg, sizeof(*loi));
3242 	if (IS_ERR(loi)) {
3243 		ret = PTR_ERR(loi);
3244 		loi = NULL;
3245 		goto out;
3246 	}
3247 
3248 	path = btrfs_alloc_path();
3249 	if (!path) {
3250 		ret = -ENOMEM;
3251 		goto out;
3252 	}
3253 
3254 	size = min_t(u32, loi->size, 64 * 1024);
3255 	inodes = init_data_container(size);
3256 	if (IS_ERR(inodes)) {
3257 		ret = PTR_ERR(inodes);
3258 		inodes = NULL;
3259 		goto out;
3260 	}
3261 
3262 	ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3263 					  build_ino_list, inodes);
3264 	if (ret == -EINVAL)
3265 		ret = -ENOENT;
3266 	if (ret < 0)
3267 		goto out;
3268 
3269 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
3270 			   (void *)(unsigned long)inodes, size);
3271 	if (ret)
3272 		ret = -EFAULT;
3273 
3274 out:
3275 	btrfs_free_path(path);
3276 	vfree(inodes);
3277 	kfree(loi);
3278 
3279 	return ret;
3280 }
3281 
3282 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3283 			       struct btrfs_ioctl_balance_args *bargs)
3284 {
3285 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3286 
3287 	bargs->flags = bctl->flags;
3288 
3289 	if (atomic_read(&fs_info->balance_running))
3290 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3291 	if (atomic_read(&fs_info->balance_pause_req))
3292 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3293 	if (atomic_read(&fs_info->balance_cancel_req))
3294 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3295 
3296 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3297 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3298 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3299 
3300 	if (lock) {
3301 		spin_lock(&fs_info->balance_lock);
3302 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3303 		spin_unlock(&fs_info->balance_lock);
3304 	} else {
3305 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3306 	}
3307 }
3308 
3309 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3310 {
3311 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3312 	struct btrfs_fs_info *fs_info = root->fs_info;
3313 	struct btrfs_ioctl_balance_args *bargs;
3314 	struct btrfs_balance_control *bctl;
3315 	int ret;
3316 
3317 	if (!capable(CAP_SYS_ADMIN))
3318 		return -EPERM;
3319 
3320 	ret = mnt_want_write_file(file);
3321 	if (ret)
3322 		return ret;
3323 
3324 	mutex_lock(&fs_info->volume_mutex);
3325 	mutex_lock(&fs_info->balance_mutex);
3326 
3327 	if (arg) {
3328 		bargs = memdup_user(arg, sizeof(*bargs));
3329 		if (IS_ERR(bargs)) {
3330 			ret = PTR_ERR(bargs);
3331 			goto out;
3332 		}
3333 
3334 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3335 			if (!fs_info->balance_ctl) {
3336 				ret = -ENOTCONN;
3337 				goto out_bargs;
3338 			}
3339 
3340 			bctl = fs_info->balance_ctl;
3341 			spin_lock(&fs_info->balance_lock);
3342 			bctl->flags |= BTRFS_BALANCE_RESUME;
3343 			spin_unlock(&fs_info->balance_lock);
3344 
3345 			goto do_balance;
3346 		}
3347 	} else {
3348 		bargs = NULL;
3349 	}
3350 
3351 	if (fs_info->balance_ctl) {
3352 		ret = -EINPROGRESS;
3353 		goto out_bargs;
3354 	}
3355 
3356 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3357 	if (!bctl) {
3358 		ret = -ENOMEM;
3359 		goto out_bargs;
3360 	}
3361 
3362 	bctl->fs_info = fs_info;
3363 	if (arg) {
3364 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3365 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3366 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3367 
3368 		bctl->flags = bargs->flags;
3369 	} else {
3370 		/* balance everything - no filters */
3371 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3372 	}
3373 
3374 do_balance:
3375 	ret = btrfs_balance(bctl, bargs);
3376 	/*
3377 	 * bctl is freed in __cancel_balance or in free_fs_info if
3378 	 * restriper was paused all the way until unmount
3379 	 */
3380 	if (arg) {
3381 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3382 			ret = -EFAULT;
3383 	}
3384 
3385 out_bargs:
3386 	kfree(bargs);
3387 out:
3388 	mutex_unlock(&fs_info->balance_mutex);
3389 	mutex_unlock(&fs_info->volume_mutex);
3390 	mnt_drop_write_file(file);
3391 	return ret;
3392 }
3393 
3394 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3395 {
3396 	if (!capable(CAP_SYS_ADMIN))
3397 		return -EPERM;
3398 
3399 	switch (cmd) {
3400 	case BTRFS_BALANCE_CTL_PAUSE:
3401 		return btrfs_pause_balance(root->fs_info);
3402 	case BTRFS_BALANCE_CTL_CANCEL:
3403 		return btrfs_cancel_balance(root->fs_info);
3404 	}
3405 
3406 	return -EINVAL;
3407 }
3408 
3409 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3410 					 void __user *arg)
3411 {
3412 	struct btrfs_fs_info *fs_info = root->fs_info;
3413 	struct btrfs_ioctl_balance_args *bargs;
3414 	int ret = 0;
3415 
3416 	if (!capable(CAP_SYS_ADMIN))
3417 		return -EPERM;
3418 
3419 	mutex_lock(&fs_info->balance_mutex);
3420 	if (!fs_info->balance_ctl) {
3421 		ret = -ENOTCONN;
3422 		goto out;
3423 	}
3424 
3425 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3426 	if (!bargs) {
3427 		ret = -ENOMEM;
3428 		goto out;
3429 	}
3430 
3431 	update_ioctl_balance_args(fs_info, 1, bargs);
3432 
3433 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3434 		ret = -EFAULT;
3435 
3436 	kfree(bargs);
3437 out:
3438 	mutex_unlock(&fs_info->balance_mutex);
3439 	return ret;
3440 }
3441 
3442 static long btrfs_ioctl_quota_ctl(struct btrfs_root *root, void __user *arg)
3443 {
3444 	struct btrfs_ioctl_quota_ctl_args *sa;
3445 	struct btrfs_trans_handle *trans = NULL;
3446 	int ret;
3447 	int err;
3448 
3449 	if (!capable(CAP_SYS_ADMIN))
3450 		return -EPERM;
3451 
3452 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3453 		return -EROFS;
3454 
3455 	sa = memdup_user(arg, sizeof(*sa));
3456 	if (IS_ERR(sa))
3457 		return PTR_ERR(sa);
3458 
3459 	if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3460 		trans = btrfs_start_transaction(root, 2);
3461 		if (IS_ERR(trans)) {
3462 			ret = PTR_ERR(trans);
3463 			goto out;
3464 		}
3465 	}
3466 
3467 	switch (sa->cmd) {
3468 	case BTRFS_QUOTA_CTL_ENABLE:
3469 		ret = btrfs_quota_enable(trans, root->fs_info);
3470 		break;
3471 	case BTRFS_QUOTA_CTL_DISABLE:
3472 		ret = btrfs_quota_disable(trans, root->fs_info);
3473 		break;
3474 	case BTRFS_QUOTA_CTL_RESCAN:
3475 		ret = btrfs_quota_rescan(root->fs_info);
3476 		break;
3477 	default:
3478 		ret = -EINVAL;
3479 		break;
3480 	}
3481 
3482 	if (copy_to_user(arg, sa, sizeof(*sa)))
3483 		ret = -EFAULT;
3484 
3485 	if (trans) {
3486 		err = btrfs_commit_transaction(trans, root);
3487 		if (err && !ret)
3488 			ret = err;
3489 	}
3490 
3491 out:
3492 	kfree(sa);
3493 	return ret;
3494 }
3495 
3496 static long btrfs_ioctl_qgroup_assign(struct btrfs_root *root, void __user *arg)
3497 {
3498 	struct btrfs_ioctl_qgroup_assign_args *sa;
3499 	struct btrfs_trans_handle *trans;
3500 	int ret;
3501 	int err;
3502 
3503 	if (!capable(CAP_SYS_ADMIN))
3504 		return -EPERM;
3505 
3506 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3507 		return -EROFS;
3508 
3509 	sa = memdup_user(arg, sizeof(*sa));
3510 	if (IS_ERR(sa))
3511 		return PTR_ERR(sa);
3512 
3513 	trans = btrfs_join_transaction(root);
3514 	if (IS_ERR(trans)) {
3515 		ret = PTR_ERR(trans);
3516 		goto out;
3517 	}
3518 
3519 	/* FIXME: check if the IDs really exist */
3520 	if (sa->assign) {
3521 		ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3522 						sa->src, sa->dst);
3523 	} else {
3524 		ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3525 						sa->src, sa->dst);
3526 	}
3527 
3528 	err = btrfs_end_transaction(trans, root);
3529 	if (err && !ret)
3530 		ret = err;
3531 
3532 out:
3533 	kfree(sa);
3534 	return ret;
3535 }
3536 
3537 static long btrfs_ioctl_qgroup_create(struct btrfs_root *root, void __user *arg)
3538 {
3539 	struct btrfs_ioctl_qgroup_create_args *sa;
3540 	struct btrfs_trans_handle *trans;
3541 	int ret;
3542 	int err;
3543 
3544 	if (!capable(CAP_SYS_ADMIN))
3545 		return -EPERM;
3546 
3547 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3548 		return -EROFS;
3549 
3550 	sa = memdup_user(arg, sizeof(*sa));
3551 	if (IS_ERR(sa))
3552 		return PTR_ERR(sa);
3553 
3554 	trans = btrfs_join_transaction(root);
3555 	if (IS_ERR(trans)) {
3556 		ret = PTR_ERR(trans);
3557 		goto out;
3558 	}
3559 
3560 	/* FIXME: check if the IDs really exist */
3561 	if (sa->create) {
3562 		ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3563 					  NULL);
3564 	} else {
3565 		ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3566 	}
3567 
3568 	err = btrfs_end_transaction(trans, root);
3569 	if (err && !ret)
3570 		ret = err;
3571 
3572 out:
3573 	kfree(sa);
3574 	return ret;
3575 }
3576 
3577 static long btrfs_ioctl_qgroup_limit(struct btrfs_root *root, void __user *arg)
3578 {
3579 	struct btrfs_ioctl_qgroup_limit_args *sa;
3580 	struct btrfs_trans_handle *trans;
3581 	int ret;
3582 	int err;
3583 	u64 qgroupid;
3584 
3585 	if (!capable(CAP_SYS_ADMIN))
3586 		return -EPERM;
3587 
3588 	if (root->fs_info->sb->s_flags & MS_RDONLY)
3589 		return -EROFS;
3590 
3591 	sa = memdup_user(arg, sizeof(*sa));
3592 	if (IS_ERR(sa))
3593 		return PTR_ERR(sa);
3594 
3595 	trans = btrfs_join_transaction(root);
3596 	if (IS_ERR(trans)) {
3597 		ret = PTR_ERR(trans);
3598 		goto out;
3599 	}
3600 
3601 	qgroupid = sa->qgroupid;
3602 	if (!qgroupid) {
3603 		/* take the current subvol as qgroup */
3604 		qgroupid = root->root_key.objectid;
3605 	}
3606 
3607 	/* FIXME: check if the IDs really exist */
3608 	ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3609 
3610 	err = btrfs_end_transaction(trans, root);
3611 	if (err && !ret)
3612 		ret = err;
3613 
3614 out:
3615 	kfree(sa);
3616 	return ret;
3617 }
3618 
3619 static long btrfs_ioctl_set_received_subvol(struct file *file,
3620 					    void __user *arg)
3621 {
3622 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
3623 	struct inode *inode = fdentry(file)->d_inode;
3624 	struct btrfs_root *root = BTRFS_I(inode)->root;
3625 	struct btrfs_root_item *root_item = &root->root_item;
3626 	struct btrfs_trans_handle *trans;
3627 	struct timespec ct = CURRENT_TIME;
3628 	int ret = 0;
3629 
3630 	ret = mnt_want_write_file(file);
3631 	if (ret < 0)
3632 		return ret;
3633 
3634 	down_write(&root->fs_info->subvol_sem);
3635 
3636 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3637 		ret = -EINVAL;
3638 		goto out;
3639 	}
3640 
3641 	if (btrfs_root_readonly(root)) {
3642 		ret = -EROFS;
3643 		goto out;
3644 	}
3645 
3646 	if (!inode_owner_or_capable(inode)) {
3647 		ret = -EACCES;
3648 		goto out;
3649 	}
3650 
3651 	sa = memdup_user(arg, sizeof(*sa));
3652 	if (IS_ERR(sa)) {
3653 		ret = PTR_ERR(sa);
3654 		sa = NULL;
3655 		goto out;
3656 	}
3657 
3658 	trans = btrfs_start_transaction(root, 1);
3659 	if (IS_ERR(trans)) {
3660 		ret = PTR_ERR(trans);
3661 		trans = NULL;
3662 		goto out;
3663 	}
3664 
3665 	sa->rtransid = trans->transid;
3666 	sa->rtime.sec = ct.tv_sec;
3667 	sa->rtime.nsec = ct.tv_nsec;
3668 
3669 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3670 	btrfs_set_root_stransid(root_item, sa->stransid);
3671 	btrfs_set_root_rtransid(root_item, sa->rtransid);
3672 	root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3673 	root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3674 	root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3675 	root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3676 
3677 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
3678 				&root->root_key, &root->root_item);
3679 	if (ret < 0) {
3680 		btrfs_end_transaction(trans, root);
3681 		trans = NULL;
3682 		goto out;
3683 	} else {
3684 		ret = btrfs_commit_transaction(trans, root);
3685 		if (ret < 0)
3686 			goto out;
3687 	}
3688 
3689 	ret = copy_to_user(arg, sa, sizeof(*sa));
3690 	if (ret)
3691 		ret = -EFAULT;
3692 
3693 out:
3694 	kfree(sa);
3695 	up_write(&root->fs_info->subvol_sem);
3696 	mnt_drop_write_file(file);
3697 	return ret;
3698 }
3699 
3700 long btrfs_ioctl(struct file *file, unsigned int
3701 		cmd, unsigned long arg)
3702 {
3703 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3704 	void __user *argp = (void __user *)arg;
3705 
3706 	switch (cmd) {
3707 	case FS_IOC_GETFLAGS:
3708 		return btrfs_ioctl_getflags(file, argp);
3709 	case FS_IOC_SETFLAGS:
3710 		return btrfs_ioctl_setflags(file, argp);
3711 	case FS_IOC_GETVERSION:
3712 		return btrfs_ioctl_getversion(file, argp);
3713 	case FITRIM:
3714 		return btrfs_ioctl_fitrim(file, argp);
3715 	case BTRFS_IOC_SNAP_CREATE:
3716 		return btrfs_ioctl_snap_create(file, argp, 0);
3717 	case BTRFS_IOC_SNAP_CREATE_V2:
3718 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
3719 	case BTRFS_IOC_SUBVOL_CREATE:
3720 		return btrfs_ioctl_snap_create(file, argp, 1);
3721 	case BTRFS_IOC_SUBVOL_CREATE_V2:
3722 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
3723 	case BTRFS_IOC_SNAP_DESTROY:
3724 		return btrfs_ioctl_snap_destroy(file, argp);
3725 	case BTRFS_IOC_SUBVOL_GETFLAGS:
3726 		return btrfs_ioctl_subvol_getflags(file, argp);
3727 	case BTRFS_IOC_SUBVOL_SETFLAGS:
3728 		return btrfs_ioctl_subvol_setflags(file, argp);
3729 	case BTRFS_IOC_DEFAULT_SUBVOL:
3730 		return btrfs_ioctl_default_subvol(file, argp);
3731 	case BTRFS_IOC_DEFRAG:
3732 		return btrfs_ioctl_defrag(file, NULL);
3733 	case BTRFS_IOC_DEFRAG_RANGE:
3734 		return btrfs_ioctl_defrag(file, argp);
3735 	case BTRFS_IOC_RESIZE:
3736 		return btrfs_ioctl_resize(root, argp);
3737 	case BTRFS_IOC_ADD_DEV:
3738 		return btrfs_ioctl_add_dev(root, argp);
3739 	case BTRFS_IOC_RM_DEV:
3740 		return btrfs_ioctl_rm_dev(root, argp);
3741 	case BTRFS_IOC_FS_INFO:
3742 		return btrfs_ioctl_fs_info(root, argp);
3743 	case BTRFS_IOC_DEV_INFO:
3744 		return btrfs_ioctl_dev_info(root, argp);
3745 	case BTRFS_IOC_BALANCE:
3746 		return btrfs_ioctl_balance(file, NULL);
3747 	case BTRFS_IOC_CLONE:
3748 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3749 	case BTRFS_IOC_CLONE_RANGE:
3750 		return btrfs_ioctl_clone_range(file, argp);
3751 	case BTRFS_IOC_TRANS_START:
3752 		return btrfs_ioctl_trans_start(file);
3753 	case BTRFS_IOC_TRANS_END:
3754 		return btrfs_ioctl_trans_end(file);
3755 	case BTRFS_IOC_TREE_SEARCH:
3756 		return btrfs_ioctl_tree_search(file, argp);
3757 	case BTRFS_IOC_INO_LOOKUP:
3758 		return btrfs_ioctl_ino_lookup(file, argp);
3759 	case BTRFS_IOC_INO_PATHS:
3760 		return btrfs_ioctl_ino_to_path(root, argp);
3761 	case BTRFS_IOC_LOGICAL_INO:
3762 		return btrfs_ioctl_logical_to_ino(root, argp);
3763 	case BTRFS_IOC_SPACE_INFO:
3764 		return btrfs_ioctl_space_info(root, argp);
3765 	case BTRFS_IOC_SYNC:
3766 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
3767 		return 0;
3768 	case BTRFS_IOC_START_SYNC:
3769 		return btrfs_ioctl_start_sync(file, argp);
3770 	case BTRFS_IOC_WAIT_SYNC:
3771 		return btrfs_ioctl_wait_sync(file, argp);
3772 	case BTRFS_IOC_SCRUB:
3773 		return btrfs_ioctl_scrub(root, argp);
3774 	case BTRFS_IOC_SCRUB_CANCEL:
3775 		return btrfs_ioctl_scrub_cancel(root, argp);
3776 	case BTRFS_IOC_SCRUB_PROGRESS:
3777 		return btrfs_ioctl_scrub_progress(root, argp);
3778 	case BTRFS_IOC_BALANCE_V2:
3779 		return btrfs_ioctl_balance(file, argp);
3780 	case BTRFS_IOC_BALANCE_CTL:
3781 		return btrfs_ioctl_balance_ctl(root, arg);
3782 	case BTRFS_IOC_BALANCE_PROGRESS:
3783 		return btrfs_ioctl_balance_progress(root, argp);
3784 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
3785 		return btrfs_ioctl_set_received_subvol(file, argp);
3786 	case BTRFS_IOC_SEND:
3787 		return btrfs_ioctl_send(file, argp);
3788 	case BTRFS_IOC_GET_DEV_STATS:
3789 		return btrfs_ioctl_get_dev_stats(root, argp);
3790 	case BTRFS_IOC_QUOTA_CTL:
3791 		return btrfs_ioctl_quota_ctl(root, argp);
3792 	case BTRFS_IOC_QGROUP_ASSIGN:
3793 		return btrfs_ioctl_qgroup_assign(root, argp);
3794 	case BTRFS_IOC_QGROUP_CREATE:
3795 		return btrfs_ioctl_qgroup_create(root, argp);
3796 	case BTRFS_IOC_QGROUP_LIMIT:
3797 		return btrfs_ioctl_qgroup_limit(root, argp);
3798 	}
3799 
3800 	return -ENOTTY;
3801 }
3802