xref: /linux/fs/btrfs/ioctl.c (revision 394d83c17fac2b7bcf05cb99d1e945135767bb6b)
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 "compat.h"
44 #include "ctree.h"
45 #include "disk-io.h"
46 #include "transaction.h"
47 #include "btrfs_inode.h"
48 #include "ioctl.h"
49 #include "print-tree.h"
50 #include "volumes.h"
51 #include "locking.h"
52 
53 /* Mask out flags that are inappropriate for the given type of inode. */
54 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
55 {
56 	if (S_ISDIR(mode))
57 		return flags;
58 	else if (S_ISREG(mode))
59 		return flags & ~FS_DIRSYNC_FL;
60 	else
61 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
62 }
63 
64 /*
65  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
66  */
67 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
68 {
69 	unsigned int iflags = 0;
70 
71 	if (flags & BTRFS_INODE_SYNC)
72 		iflags |= FS_SYNC_FL;
73 	if (flags & BTRFS_INODE_IMMUTABLE)
74 		iflags |= FS_IMMUTABLE_FL;
75 	if (flags & BTRFS_INODE_APPEND)
76 		iflags |= FS_APPEND_FL;
77 	if (flags & BTRFS_INODE_NODUMP)
78 		iflags |= FS_NODUMP_FL;
79 	if (flags & BTRFS_INODE_NOATIME)
80 		iflags |= FS_NOATIME_FL;
81 	if (flags & BTRFS_INODE_DIRSYNC)
82 		iflags |= FS_DIRSYNC_FL;
83 
84 	return iflags;
85 }
86 
87 /*
88  * Update inode->i_flags based on the btrfs internal flags.
89  */
90 void btrfs_update_iflags(struct inode *inode)
91 {
92 	struct btrfs_inode *ip = BTRFS_I(inode);
93 
94 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
95 
96 	if (ip->flags & BTRFS_INODE_SYNC)
97 		inode->i_flags |= S_SYNC;
98 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
99 		inode->i_flags |= S_IMMUTABLE;
100 	if (ip->flags & BTRFS_INODE_APPEND)
101 		inode->i_flags |= S_APPEND;
102 	if (ip->flags & BTRFS_INODE_NOATIME)
103 		inode->i_flags |= S_NOATIME;
104 	if (ip->flags & BTRFS_INODE_DIRSYNC)
105 		inode->i_flags |= S_DIRSYNC;
106 }
107 
108 /*
109  * Inherit flags from the parent inode.
110  *
111  * Unlike extN we don't have any flags we don't want to inherit currently.
112  */
113 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
114 {
115 	unsigned int flags;
116 
117 	if (!dir)
118 		return;
119 
120 	flags = BTRFS_I(dir)->flags;
121 
122 	if (S_ISREG(inode->i_mode))
123 		flags &= ~BTRFS_INODE_DIRSYNC;
124 	else if (!S_ISDIR(inode->i_mode))
125 		flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
126 
127 	BTRFS_I(inode)->flags = flags;
128 	btrfs_update_iflags(inode);
129 }
130 
131 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
132 {
133 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
134 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
135 
136 	if (copy_to_user(arg, &flags, sizeof(flags)))
137 		return -EFAULT;
138 	return 0;
139 }
140 
141 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
142 {
143 	struct inode *inode = file->f_path.dentry->d_inode;
144 	struct btrfs_inode *ip = BTRFS_I(inode);
145 	struct btrfs_root *root = ip->root;
146 	struct btrfs_trans_handle *trans;
147 	unsigned int flags, oldflags;
148 	int ret;
149 
150 	if (copy_from_user(&flags, arg, sizeof(flags)))
151 		return -EFAULT;
152 
153 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
154 		      FS_NOATIME_FL | FS_NODUMP_FL | \
155 		      FS_SYNC_FL | FS_DIRSYNC_FL))
156 		return -EOPNOTSUPP;
157 
158 	if (!is_owner_or_cap(inode))
159 		return -EACCES;
160 
161 	mutex_lock(&inode->i_mutex);
162 
163 	flags = btrfs_mask_flags(inode->i_mode, flags);
164 	oldflags = btrfs_flags_to_ioctl(ip->flags);
165 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
166 		if (!capable(CAP_LINUX_IMMUTABLE)) {
167 			ret = -EPERM;
168 			goto out_unlock;
169 		}
170 	}
171 
172 	ret = mnt_want_write(file->f_path.mnt);
173 	if (ret)
174 		goto out_unlock;
175 
176 	if (flags & FS_SYNC_FL)
177 		ip->flags |= BTRFS_INODE_SYNC;
178 	else
179 		ip->flags &= ~BTRFS_INODE_SYNC;
180 	if (flags & FS_IMMUTABLE_FL)
181 		ip->flags |= BTRFS_INODE_IMMUTABLE;
182 	else
183 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
184 	if (flags & FS_APPEND_FL)
185 		ip->flags |= BTRFS_INODE_APPEND;
186 	else
187 		ip->flags &= ~BTRFS_INODE_APPEND;
188 	if (flags & FS_NODUMP_FL)
189 		ip->flags |= BTRFS_INODE_NODUMP;
190 	else
191 		ip->flags &= ~BTRFS_INODE_NODUMP;
192 	if (flags & FS_NOATIME_FL)
193 		ip->flags |= BTRFS_INODE_NOATIME;
194 	else
195 		ip->flags &= ~BTRFS_INODE_NOATIME;
196 	if (flags & FS_DIRSYNC_FL)
197 		ip->flags |= BTRFS_INODE_DIRSYNC;
198 	else
199 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
200 
201 
202 	trans = btrfs_join_transaction(root, 1);
203 	BUG_ON(!trans);
204 
205 	ret = btrfs_update_inode(trans, root, inode);
206 	BUG_ON(ret);
207 
208 	btrfs_update_iflags(inode);
209 	inode->i_ctime = CURRENT_TIME;
210 	btrfs_end_transaction(trans, root);
211 
212 	mnt_drop_write(file->f_path.mnt);
213  out_unlock:
214 	mutex_unlock(&inode->i_mutex);
215 	return 0;
216 }
217 
218 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
219 {
220 	struct inode *inode = file->f_path.dentry->d_inode;
221 
222 	return put_user(inode->i_generation, arg);
223 }
224 
225 static noinline int create_subvol(struct btrfs_root *root,
226 				  struct dentry *dentry,
227 				  char *name, int namelen,
228 				  u64 *async_transid)
229 {
230 	struct btrfs_trans_handle *trans;
231 	struct btrfs_key key;
232 	struct btrfs_root_item root_item;
233 	struct btrfs_inode_item *inode_item;
234 	struct extent_buffer *leaf;
235 	struct btrfs_root *new_root;
236 	struct dentry *parent = dget_parent(dentry);
237 	struct inode *dir;
238 	int ret;
239 	int err;
240 	u64 objectid;
241 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
242 	u64 index = 0;
243 
244 	ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root,
245 				       0, &objectid);
246 	if (ret) {
247 		dput(parent);
248 		return ret;
249 	}
250 
251 	dir = parent->d_inode;
252 
253 	/*
254 	 * 1 - inode item
255 	 * 2 - refs
256 	 * 1 - root item
257 	 * 2 - dir items
258 	 */
259 	trans = btrfs_start_transaction(root, 6);
260 	if (IS_ERR(trans)) {
261 		dput(parent);
262 		return PTR_ERR(trans);
263 	}
264 
265 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
266 				      0, objectid, NULL, 0, 0, 0);
267 	if (IS_ERR(leaf)) {
268 		ret = PTR_ERR(leaf);
269 		goto fail;
270 	}
271 
272 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
273 	btrfs_set_header_bytenr(leaf, leaf->start);
274 	btrfs_set_header_generation(leaf, trans->transid);
275 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
276 	btrfs_set_header_owner(leaf, objectid);
277 
278 	write_extent_buffer(leaf, root->fs_info->fsid,
279 			    (unsigned long)btrfs_header_fsid(leaf),
280 			    BTRFS_FSID_SIZE);
281 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
282 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
283 			    BTRFS_UUID_SIZE);
284 	btrfs_mark_buffer_dirty(leaf);
285 
286 	inode_item = &root_item.inode;
287 	memset(inode_item, 0, sizeof(*inode_item));
288 	inode_item->generation = cpu_to_le64(1);
289 	inode_item->size = cpu_to_le64(3);
290 	inode_item->nlink = cpu_to_le32(1);
291 	inode_item->nbytes = cpu_to_le64(root->leafsize);
292 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
293 
294 	btrfs_set_root_bytenr(&root_item, leaf->start);
295 	btrfs_set_root_generation(&root_item, trans->transid);
296 	btrfs_set_root_level(&root_item, 0);
297 	btrfs_set_root_refs(&root_item, 1);
298 	btrfs_set_root_used(&root_item, leaf->len);
299 	btrfs_set_root_last_snapshot(&root_item, 0);
300 
301 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
302 	root_item.drop_level = 0;
303 
304 	btrfs_tree_unlock(leaf);
305 	free_extent_buffer(leaf);
306 	leaf = NULL;
307 
308 	btrfs_set_root_dirid(&root_item, new_dirid);
309 
310 	key.objectid = objectid;
311 	key.offset = 0;
312 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
313 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
314 				&root_item);
315 	if (ret)
316 		goto fail;
317 
318 	key.offset = (u64)-1;
319 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
320 	BUG_ON(IS_ERR(new_root));
321 
322 	btrfs_record_root_in_trans(trans, new_root);
323 
324 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
325 				       BTRFS_I(dir)->block_group);
326 	/*
327 	 * insert the directory item
328 	 */
329 	ret = btrfs_set_inode_index(dir, &index);
330 	BUG_ON(ret);
331 
332 	ret = btrfs_insert_dir_item(trans, root,
333 				    name, namelen, dir->i_ino, &key,
334 				    BTRFS_FT_DIR, index);
335 	if (ret)
336 		goto fail;
337 
338 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
339 	ret = btrfs_update_inode(trans, root, dir);
340 	BUG_ON(ret);
341 
342 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
343 				 objectid, root->root_key.objectid,
344 				 dir->i_ino, index, name, namelen);
345 
346 	BUG_ON(ret);
347 
348 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
349 fail:
350 	dput(parent);
351 	if (async_transid) {
352 		*async_transid = trans->transid;
353 		err = btrfs_commit_transaction_async(trans, root, 1);
354 	} else {
355 		err = btrfs_commit_transaction(trans, root);
356 	}
357 	if (err && !ret)
358 		ret = err;
359 	return ret;
360 }
361 
362 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
363 			   char *name, int namelen, u64 *async_transid)
364 {
365 	struct inode *inode;
366 	struct dentry *parent;
367 	struct btrfs_pending_snapshot *pending_snapshot;
368 	struct btrfs_trans_handle *trans;
369 	int ret;
370 
371 	if (!root->ref_cows)
372 		return -EINVAL;
373 
374 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
375 	if (!pending_snapshot)
376 		return -ENOMEM;
377 
378 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
379 	pending_snapshot->dentry = dentry;
380 	pending_snapshot->root = root;
381 
382 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
383 	if (IS_ERR(trans)) {
384 		ret = PTR_ERR(trans);
385 		goto fail;
386 	}
387 
388 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
389 	BUG_ON(ret);
390 
391 	list_add(&pending_snapshot->list,
392 		 &trans->transaction->pending_snapshots);
393 	if (async_transid) {
394 		*async_transid = trans->transid;
395 		ret = btrfs_commit_transaction_async(trans,
396 				     root->fs_info->extent_root, 1);
397 	} else {
398 		ret = btrfs_commit_transaction(trans,
399 					       root->fs_info->extent_root);
400 	}
401 	BUG_ON(ret);
402 
403 	ret = pending_snapshot->error;
404 	if (ret)
405 		goto fail;
406 
407 	btrfs_orphan_cleanup(pending_snapshot->snap);
408 
409 	parent = dget_parent(dentry);
410 	inode = btrfs_lookup_dentry(parent->d_inode, dentry);
411 	dput(parent);
412 	if (IS_ERR(inode)) {
413 		ret = PTR_ERR(inode);
414 		goto fail;
415 	}
416 	BUG_ON(!inode);
417 	d_instantiate(dentry, inode);
418 	ret = 0;
419 fail:
420 	kfree(pending_snapshot);
421 	return ret;
422 }
423 
424 /*  copy of check_sticky in fs/namei.c()
425 * It's inline, so penalty for filesystems that don't use sticky bit is
426 * minimal.
427 */
428 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
429 {
430 	uid_t fsuid = current_fsuid();
431 
432 	if (!(dir->i_mode & S_ISVTX))
433 		return 0;
434 	if (inode->i_uid == fsuid)
435 		return 0;
436 	if (dir->i_uid == fsuid)
437 		return 0;
438 	return !capable(CAP_FOWNER);
439 }
440 
441 /*  copy of may_delete in fs/namei.c()
442  *	Check whether we can remove a link victim from directory dir, check
443  *  whether the type of victim is right.
444  *  1. We can't do it if dir is read-only (done in permission())
445  *  2. We should have write and exec permissions on dir
446  *  3. We can't remove anything from append-only dir
447  *  4. We can't do anything with immutable dir (done in permission())
448  *  5. If the sticky bit on dir is set we should either
449  *	a. be owner of dir, or
450  *	b. be owner of victim, or
451  *	c. have CAP_FOWNER capability
452  *  6. If the victim is append-only or immutable we can't do antyhing with
453  *     links pointing to it.
454  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
455  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
456  *  9. We can't remove a root or mountpoint.
457  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
458  *     nfs_async_unlink().
459  */
460 
461 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
462 {
463 	int error;
464 
465 	if (!victim->d_inode)
466 		return -ENOENT;
467 
468 	BUG_ON(victim->d_parent->d_inode != dir);
469 	audit_inode_child(victim, dir);
470 
471 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
472 	if (error)
473 		return error;
474 	if (IS_APPEND(dir))
475 		return -EPERM;
476 	if (btrfs_check_sticky(dir, victim->d_inode)||
477 		IS_APPEND(victim->d_inode)||
478 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
479 		return -EPERM;
480 	if (isdir) {
481 		if (!S_ISDIR(victim->d_inode->i_mode))
482 			return -ENOTDIR;
483 		if (IS_ROOT(victim))
484 			return -EBUSY;
485 	} else if (S_ISDIR(victim->d_inode->i_mode))
486 		return -EISDIR;
487 	if (IS_DEADDIR(dir))
488 		return -ENOENT;
489 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
490 		return -EBUSY;
491 	return 0;
492 }
493 
494 /* copy of may_create in fs/namei.c() */
495 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
496 {
497 	if (child->d_inode)
498 		return -EEXIST;
499 	if (IS_DEADDIR(dir))
500 		return -ENOENT;
501 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
502 }
503 
504 /*
505  * Create a new subvolume below @parent.  This is largely modeled after
506  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
507  * inside this filesystem so it's quite a bit simpler.
508  */
509 static noinline int btrfs_mksubvol(struct path *parent,
510 				   char *name, int namelen,
511 				   struct btrfs_root *snap_src,
512 				   u64 *async_transid)
513 {
514 	struct inode *dir  = parent->dentry->d_inode;
515 	struct dentry *dentry;
516 	int error;
517 
518 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
519 
520 	dentry = lookup_one_len(name, parent->dentry, namelen);
521 	error = PTR_ERR(dentry);
522 	if (IS_ERR(dentry))
523 		goto out_unlock;
524 
525 	error = -EEXIST;
526 	if (dentry->d_inode)
527 		goto out_dput;
528 
529 	error = mnt_want_write(parent->mnt);
530 	if (error)
531 		goto out_dput;
532 
533 	error = btrfs_may_create(dir, dentry);
534 	if (error)
535 		goto out_drop_write;
536 
537 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
538 
539 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
540 		goto out_up_read;
541 
542 	if (snap_src) {
543 		error = create_snapshot(snap_src, dentry,
544 					name, namelen, async_transid);
545 	} else {
546 		error = create_subvol(BTRFS_I(dir)->root, dentry,
547 				      name, namelen, async_transid);
548 	}
549 	if (!error)
550 		fsnotify_mkdir(dir, dentry);
551 out_up_read:
552 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
553 out_drop_write:
554 	mnt_drop_write(parent->mnt);
555 out_dput:
556 	dput(dentry);
557 out_unlock:
558 	mutex_unlock(&dir->i_mutex);
559 	return error;
560 }
561 
562 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
563 			       int thresh, u64 *last_len, u64 *skip,
564 			       u64 *defrag_end)
565 {
566 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
567 	struct extent_map *em = NULL;
568 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
569 	int ret = 1;
570 
571 
572 	if (thresh == 0)
573 		thresh = 256 * 1024;
574 
575 	/*
576 	 * make sure that once we start defragging and extent, we keep on
577 	 * defragging it
578 	 */
579 	if (start < *defrag_end)
580 		return 1;
581 
582 	*skip = 0;
583 
584 	/*
585 	 * hopefully we have this extent in the tree already, try without
586 	 * the full extent lock
587 	 */
588 	read_lock(&em_tree->lock);
589 	em = lookup_extent_mapping(em_tree, start, len);
590 	read_unlock(&em_tree->lock);
591 
592 	if (!em) {
593 		/* get the big lock and read metadata off disk */
594 		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
595 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
596 		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
597 
598 		if (IS_ERR(em))
599 			return 0;
600 	}
601 
602 	/* this will cover holes, and inline extents */
603 	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
604 		ret = 0;
605 
606 	/*
607 	 * we hit a real extent, if it is big don't bother defragging it again
608 	 */
609 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
610 		ret = 0;
611 
612 	/*
613 	 * last_len ends up being a counter of how many bytes we've defragged.
614 	 * every time we choose not to defrag an extent, we reset *last_len
615 	 * so that the next tiny extent will force a defrag.
616 	 *
617 	 * The end result of this is that tiny extents before a single big
618 	 * extent will force at least part of that big extent to be defragged.
619 	 */
620 	if (ret) {
621 		*last_len += len;
622 		*defrag_end = extent_map_end(em);
623 	} else {
624 		*last_len = 0;
625 		*skip = extent_map_end(em);
626 		*defrag_end = 0;
627 	}
628 
629 	free_extent_map(em);
630 	return ret;
631 }
632 
633 static int btrfs_defrag_file(struct file *file,
634 			     struct btrfs_ioctl_defrag_range_args *range)
635 {
636 	struct inode *inode = fdentry(file)->d_inode;
637 	struct btrfs_root *root = BTRFS_I(inode)->root;
638 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
639 	struct btrfs_ordered_extent *ordered;
640 	struct page *page;
641 	unsigned long last_index;
642 	unsigned long ra_pages = root->fs_info->bdi.ra_pages;
643 	unsigned long total_read = 0;
644 	u64 page_start;
645 	u64 page_end;
646 	u64 last_len = 0;
647 	u64 skip = 0;
648 	u64 defrag_end = 0;
649 	unsigned long i;
650 	int ret;
651 
652 	if (inode->i_size == 0)
653 		return 0;
654 
655 	if (range->start + range->len > range->start) {
656 		last_index = min_t(u64, inode->i_size - 1,
657 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
658 	} else {
659 		last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
660 	}
661 
662 	i = range->start >> PAGE_CACHE_SHIFT;
663 	while (i <= last_index) {
664 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
665 					PAGE_CACHE_SIZE,
666 					range->extent_thresh,
667 					&last_len, &skip,
668 					&defrag_end)) {
669 			unsigned long next;
670 			/*
671 			 * the should_defrag function tells us how much to skip
672 			 * bump our counter by the suggested amount
673 			 */
674 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
675 			i = max(i + 1, next);
676 			continue;
677 		}
678 
679 		if (total_read % ra_pages == 0) {
680 			btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
681 				       min(last_index, i + ra_pages - 1));
682 		}
683 		total_read++;
684 		mutex_lock(&inode->i_mutex);
685 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
686 			BTRFS_I(inode)->force_compress = 1;
687 
688 		ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
689 		if (ret)
690 			goto err_unlock;
691 again:
692 		if (inode->i_size == 0 ||
693 		    i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) {
694 			ret = 0;
695 			goto err_reservations;
696 		}
697 
698 		page = grab_cache_page(inode->i_mapping, i);
699 		if (!page) {
700 			ret = -ENOMEM;
701 			goto err_reservations;
702 		}
703 
704 		if (!PageUptodate(page)) {
705 			btrfs_readpage(NULL, page);
706 			lock_page(page);
707 			if (!PageUptodate(page)) {
708 				unlock_page(page);
709 				page_cache_release(page);
710 				ret = -EIO;
711 				goto err_reservations;
712 			}
713 		}
714 
715 		if (page->mapping != inode->i_mapping) {
716 			unlock_page(page);
717 			page_cache_release(page);
718 			goto again;
719 		}
720 
721 		wait_on_page_writeback(page);
722 
723 		if (PageDirty(page)) {
724 			btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
725 			goto loop_unlock;
726 		}
727 
728 		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
729 		page_end = page_start + PAGE_CACHE_SIZE - 1;
730 		lock_extent(io_tree, page_start, page_end, GFP_NOFS);
731 
732 		ordered = btrfs_lookup_ordered_extent(inode, page_start);
733 		if (ordered) {
734 			unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
735 			unlock_page(page);
736 			page_cache_release(page);
737 			btrfs_start_ordered_extent(inode, ordered, 1);
738 			btrfs_put_ordered_extent(ordered);
739 			goto again;
740 		}
741 		set_page_extent_mapped(page);
742 
743 		/*
744 		 * this makes sure page_mkwrite is called on the
745 		 * page if it is dirtied again later
746 		 */
747 		clear_page_dirty_for_io(page);
748 		clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start,
749 				  page_end, EXTENT_DIRTY | EXTENT_DELALLOC |
750 				  EXTENT_DO_ACCOUNTING, GFP_NOFS);
751 
752 		btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
753 		ClearPageChecked(page);
754 		set_page_dirty(page);
755 		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
756 
757 loop_unlock:
758 		unlock_page(page);
759 		page_cache_release(page);
760 		mutex_unlock(&inode->i_mutex);
761 
762 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
763 		i++;
764 	}
765 
766 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
767 		filemap_flush(inode->i_mapping);
768 
769 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
770 		/* the filemap_flush will queue IO into the worker threads, but
771 		 * we have to make sure the IO is actually started and that
772 		 * ordered extents get created before we return
773 		 */
774 		atomic_inc(&root->fs_info->async_submit_draining);
775 		while (atomic_read(&root->fs_info->nr_async_submits) ||
776 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
777 			wait_event(root->fs_info->async_submit_wait,
778 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
779 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
780 		}
781 		atomic_dec(&root->fs_info->async_submit_draining);
782 
783 		mutex_lock(&inode->i_mutex);
784 		BTRFS_I(inode)->force_compress = 0;
785 		mutex_unlock(&inode->i_mutex);
786 	}
787 
788 	return 0;
789 
790 err_reservations:
791 	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
792 err_unlock:
793 	mutex_unlock(&inode->i_mutex);
794 	return ret;
795 }
796 
797 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
798 					void __user *arg)
799 {
800 	u64 new_size;
801 	u64 old_size;
802 	u64 devid = 1;
803 	struct btrfs_ioctl_vol_args *vol_args;
804 	struct btrfs_trans_handle *trans;
805 	struct btrfs_device *device = NULL;
806 	char *sizestr;
807 	char *devstr = NULL;
808 	int ret = 0;
809 	int mod = 0;
810 
811 	if (root->fs_info->sb->s_flags & MS_RDONLY)
812 		return -EROFS;
813 
814 	if (!capable(CAP_SYS_ADMIN))
815 		return -EPERM;
816 
817 	vol_args = memdup_user(arg, sizeof(*vol_args));
818 	if (IS_ERR(vol_args))
819 		return PTR_ERR(vol_args);
820 
821 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
822 
823 	mutex_lock(&root->fs_info->volume_mutex);
824 	sizestr = vol_args->name;
825 	devstr = strchr(sizestr, ':');
826 	if (devstr) {
827 		char *end;
828 		sizestr = devstr + 1;
829 		*devstr = '\0';
830 		devstr = vol_args->name;
831 		devid = simple_strtoull(devstr, &end, 10);
832 		printk(KERN_INFO "resizing devid %llu\n",
833 		       (unsigned long long)devid);
834 	}
835 	device = btrfs_find_device(root, devid, NULL, NULL);
836 	if (!device) {
837 		printk(KERN_INFO "resizer unable to find device %llu\n",
838 		       (unsigned long long)devid);
839 		ret = -EINVAL;
840 		goto out_unlock;
841 	}
842 	if (!strcmp(sizestr, "max"))
843 		new_size = device->bdev->bd_inode->i_size;
844 	else {
845 		if (sizestr[0] == '-') {
846 			mod = -1;
847 			sizestr++;
848 		} else if (sizestr[0] == '+') {
849 			mod = 1;
850 			sizestr++;
851 		}
852 		new_size = memparse(sizestr, NULL);
853 		if (new_size == 0) {
854 			ret = -EINVAL;
855 			goto out_unlock;
856 		}
857 	}
858 
859 	old_size = device->total_bytes;
860 
861 	if (mod < 0) {
862 		if (new_size > old_size) {
863 			ret = -EINVAL;
864 			goto out_unlock;
865 		}
866 		new_size = old_size - new_size;
867 	} else if (mod > 0) {
868 		new_size = old_size + new_size;
869 	}
870 
871 	if (new_size < 256 * 1024 * 1024) {
872 		ret = -EINVAL;
873 		goto out_unlock;
874 	}
875 	if (new_size > device->bdev->bd_inode->i_size) {
876 		ret = -EFBIG;
877 		goto out_unlock;
878 	}
879 
880 	do_div(new_size, root->sectorsize);
881 	new_size *= root->sectorsize;
882 
883 	printk(KERN_INFO "new size for %s is %llu\n",
884 		device->name, (unsigned long long)new_size);
885 
886 	if (new_size > old_size) {
887 		trans = btrfs_start_transaction(root, 0);
888 		ret = btrfs_grow_device(trans, device, new_size);
889 		btrfs_commit_transaction(trans, root);
890 	} else {
891 		ret = btrfs_shrink_device(device, new_size);
892 	}
893 
894 out_unlock:
895 	mutex_unlock(&root->fs_info->volume_mutex);
896 	kfree(vol_args);
897 	return ret;
898 }
899 
900 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
901 						    char *name,
902 						    unsigned long fd,
903 						    int subvol,
904 						    u64 *transid)
905 {
906 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
907 	struct file *src_file;
908 	int namelen;
909 	int ret = 0;
910 
911 	if (root->fs_info->sb->s_flags & MS_RDONLY)
912 		return -EROFS;
913 
914 	namelen = strlen(name);
915 	if (strchr(name, '/')) {
916 		ret = -EINVAL;
917 		goto out;
918 	}
919 
920 	if (subvol) {
921 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
922 				     NULL, transid);
923 	} else {
924 		struct inode *src_inode;
925 		src_file = fget(fd);
926 		if (!src_file) {
927 			ret = -EINVAL;
928 			goto out;
929 		}
930 
931 		src_inode = src_file->f_path.dentry->d_inode;
932 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
933 			printk(KERN_INFO "btrfs: Snapshot src from "
934 			       "another FS\n");
935 			ret = -EINVAL;
936 			fput(src_file);
937 			goto out;
938 		}
939 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
940 				     BTRFS_I(src_inode)->root,
941 				     transid);
942 		fput(src_file);
943 	}
944 out:
945 	return ret;
946 }
947 
948 static noinline int btrfs_ioctl_snap_create(struct file *file,
949 					    void __user *arg, int subvol,
950 					    int v2)
951 {
952 	struct btrfs_ioctl_vol_args *vol_args = NULL;
953 	struct btrfs_ioctl_vol_args_v2 *vol_args_v2 = NULL;
954 	char *name;
955 	u64 fd;
956 	int ret;
957 
958 	if (v2) {
959 		u64 transid = 0;
960 		u64 *ptr = NULL;
961 
962 		vol_args_v2 = memdup_user(arg, sizeof(*vol_args_v2));
963 		if (IS_ERR(vol_args_v2))
964 			return PTR_ERR(vol_args_v2);
965 
966 		if (vol_args_v2->flags & ~BTRFS_SUBVOL_CREATE_ASYNC) {
967 			ret = -EINVAL;
968 			goto out;
969 		}
970 
971 		name = vol_args_v2->name;
972 		fd = vol_args_v2->fd;
973 		vol_args_v2->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
974 
975 		if (vol_args_v2->flags & BTRFS_SUBVOL_CREATE_ASYNC)
976 			ptr = &transid;
977 
978 		ret = btrfs_ioctl_snap_create_transid(file, name, fd,
979 						      subvol, ptr);
980 
981 		if (ret == 0 && ptr &&
982 		    copy_to_user(arg +
983 				 offsetof(struct btrfs_ioctl_vol_args_v2,
984 					  transid), ptr, sizeof(*ptr)))
985 			ret = -EFAULT;
986 	} else {
987 		vol_args = memdup_user(arg, sizeof(*vol_args));
988 		if (IS_ERR(vol_args))
989 			return PTR_ERR(vol_args);
990 		name = vol_args->name;
991 		fd = vol_args->fd;
992 		vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
993 
994 		ret = btrfs_ioctl_snap_create_transid(file, name, fd,
995 						      subvol, NULL);
996 	}
997 out:
998 	kfree(vol_args);
999 	kfree(vol_args_v2);
1000 
1001 	return ret;
1002 }
1003 
1004 /*
1005  * helper to check if the subvolume references other subvolumes
1006  */
1007 static noinline int may_destroy_subvol(struct btrfs_root *root)
1008 {
1009 	struct btrfs_path *path;
1010 	struct btrfs_key key;
1011 	int ret;
1012 
1013 	path = btrfs_alloc_path();
1014 	if (!path)
1015 		return -ENOMEM;
1016 
1017 	key.objectid = root->root_key.objectid;
1018 	key.type = BTRFS_ROOT_REF_KEY;
1019 	key.offset = (u64)-1;
1020 
1021 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1022 				&key, path, 0, 0);
1023 	if (ret < 0)
1024 		goto out;
1025 	BUG_ON(ret == 0);
1026 
1027 	ret = 0;
1028 	if (path->slots[0] > 0) {
1029 		path->slots[0]--;
1030 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1031 		if (key.objectid == root->root_key.objectid &&
1032 		    key.type == BTRFS_ROOT_REF_KEY)
1033 			ret = -ENOTEMPTY;
1034 	}
1035 out:
1036 	btrfs_free_path(path);
1037 	return ret;
1038 }
1039 
1040 static noinline int key_in_sk(struct btrfs_key *key,
1041 			      struct btrfs_ioctl_search_key *sk)
1042 {
1043 	struct btrfs_key test;
1044 	int ret;
1045 
1046 	test.objectid = sk->min_objectid;
1047 	test.type = sk->min_type;
1048 	test.offset = sk->min_offset;
1049 
1050 	ret = btrfs_comp_cpu_keys(key, &test);
1051 	if (ret < 0)
1052 		return 0;
1053 
1054 	test.objectid = sk->max_objectid;
1055 	test.type = sk->max_type;
1056 	test.offset = sk->max_offset;
1057 
1058 	ret = btrfs_comp_cpu_keys(key, &test);
1059 	if (ret > 0)
1060 		return 0;
1061 	return 1;
1062 }
1063 
1064 static noinline int copy_to_sk(struct btrfs_root *root,
1065 			       struct btrfs_path *path,
1066 			       struct btrfs_key *key,
1067 			       struct btrfs_ioctl_search_key *sk,
1068 			       char *buf,
1069 			       unsigned long *sk_offset,
1070 			       int *num_found)
1071 {
1072 	u64 found_transid;
1073 	struct extent_buffer *leaf;
1074 	struct btrfs_ioctl_search_header sh;
1075 	unsigned long item_off;
1076 	unsigned long item_len;
1077 	int nritems;
1078 	int i;
1079 	int slot;
1080 	int found = 0;
1081 	int ret = 0;
1082 
1083 	leaf = path->nodes[0];
1084 	slot = path->slots[0];
1085 	nritems = btrfs_header_nritems(leaf);
1086 
1087 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1088 		i = nritems;
1089 		goto advance_key;
1090 	}
1091 	found_transid = btrfs_header_generation(leaf);
1092 
1093 	for (i = slot; i < nritems; i++) {
1094 		item_off = btrfs_item_ptr_offset(leaf, i);
1095 		item_len = btrfs_item_size_nr(leaf, i);
1096 
1097 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1098 			item_len = 0;
1099 
1100 		if (sizeof(sh) + item_len + *sk_offset >
1101 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1102 			ret = 1;
1103 			goto overflow;
1104 		}
1105 
1106 		btrfs_item_key_to_cpu(leaf, key, i);
1107 		if (!key_in_sk(key, sk))
1108 			continue;
1109 
1110 		sh.objectid = key->objectid;
1111 		sh.offset = key->offset;
1112 		sh.type = key->type;
1113 		sh.len = item_len;
1114 		sh.transid = found_transid;
1115 
1116 		/* copy search result header */
1117 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1118 		*sk_offset += sizeof(sh);
1119 
1120 		if (item_len) {
1121 			char *p = buf + *sk_offset;
1122 			/* copy the item */
1123 			read_extent_buffer(leaf, p,
1124 					   item_off, item_len);
1125 			*sk_offset += item_len;
1126 		}
1127 		found++;
1128 
1129 		if (*num_found >= sk->nr_items)
1130 			break;
1131 	}
1132 advance_key:
1133 	ret = 0;
1134 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1135 		key->offset++;
1136 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1137 		key->offset = 0;
1138 		key->type++;
1139 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1140 		key->offset = 0;
1141 		key->type = 0;
1142 		key->objectid++;
1143 	} else
1144 		ret = 1;
1145 overflow:
1146 	*num_found += found;
1147 	return ret;
1148 }
1149 
1150 static noinline int search_ioctl(struct inode *inode,
1151 				 struct btrfs_ioctl_search_args *args)
1152 {
1153 	struct btrfs_root *root;
1154 	struct btrfs_key key;
1155 	struct btrfs_key max_key;
1156 	struct btrfs_path *path;
1157 	struct btrfs_ioctl_search_key *sk = &args->key;
1158 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1159 	int ret;
1160 	int num_found = 0;
1161 	unsigned long sk_offset = 0;
1162 
1163 	path = btrfs_alloc_path();
1164 	if (!path)
1165 		return -ENOMEM;
1166 
1167 	if (sk->tree_id == 0) {
1168 		/* search the root of the inode that was passed */
1169 		root = BTRFS_I(inode)->root;
1170 	} else {
1171 		key.objectid = sk->tree_id;
1172 		key.type = BTRFS_ROOT_ITEM_KEY;
1173 		key.offset = (u64)-1;
1174 		root = btrfs_read_fs_root_no_name(info, &key);
1175 		if (IS_ERR(root)) {
1176 			printk(KERN_ERR "could not find root %llu\n",
1177 			       sk->tree_id);
1178 			btrfs_free_path(path);
1179 			return -ENOENT;
1180 		}
1181 	}
1182 
1183 	key.objectid = sk->min_objectid;
1184 	key.type = sk->min_type;
1185 	key.offset = sk->min_offset;
1186 
1187 	max_key.objectid = sk->max_objectid;
1188 	max_key.type = sk->max_type;
1189 	max_key.offset = sk->max_offset;
1190 
1191 	path->keep_locks = 1;
1192 
1193 	while(1) {
1194 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1195 					   sk->min_transid);
1196 		if (ret != 0) {
1197 			if (ret > 0)
1198 				ret = 0;
1199 			goto err;
1200 		}
1201 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1202 				 &sk_offset, &num_found);
1203 		btrfs_release_path(root, path);
1204 		if (ret || num_found >= sk->nr_items)
1205 			break;
1206 
1207 	}
1208 	ret = 0;
1209 err:
1210 	sk->nr_items = num_found;
1211 	btrfs_free_path(path);
1212 	return ret;
1213 }
1214 
1215 static noinline int btrfs_ioctl_tree_search(struct file *file,
1216 					   void __user *argp)
1217 {
1218 	 struct btrfs_ioctl_search_args *args;
1219 	 struct inode *inode;
1220 	 int ret;
1221 
1222 	if (!capable(CAP_SYS_ADMIN))
1223 		return -EPERM;
1224 
1225 	args = memdup_user(argp, sizeof(*args));
1226 	if (IS_ERR(args))
1227 		return PTR_ERR(args);
1228 
1229 	inode = fdentry(file)->d_inode;
1230 	ret = search_ioctl(inode, args);
1231 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1232 		ret = -EFAULT;
1233 	kfree(args);
1234 	return ret;
1235 }
1236 
1237 /*
1238  * Search INODE_REFs to identify path name of 'dirid' directory
1239  * in a 'tree_id' tree. and sets path name to 'name'.
1240  */
1241 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1242 				u64 tree_id, u64 dirid, char *name)
1243 {
1244 	struct btrfs_root *root;
1245 	struct btrfs_key key;
1246 	char *ptr;
1247 	int ret = -1;
1248 	int slot;
1249 	int len;
1250 	int total_len = 0;
1251 	struct btrfs_inode_ref *iref;
1252 	struct extent_buffer *l;
1253 	struct btrfs_path *path;
1254 
1255 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1256 		name[0]='\0';
1257 		return 0;
1258 	}
1259 
1260 	path = btrfs_alloc_path();
1261 	if (!path)
1262 		return -ENOMEM;
1263 
1264 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1265 
1266 	key.objectid = tree_id;
1267 	key.type = BTRFS_ROOT_ITEM_KEY;
1268 	key.offset = (u64)-1;
1269 	root = btrfs_read_fs_root_no_name(info, &key);
1270 	if (IS_ERR(root)) {
1271 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1272 		ret = -ENOENT;
1273 		goto out;
1274 	}
1275 
1276 	key.objectid = dirid;
1277 	key.type = BTRFS_INODE_REF_KEY;
1278 	key.offset = (u64)-1;
1279 
1280 	while(1) {
1281 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1282 		if (ret < 0)
1283 			goto out;
1284 
1285 		l = path->nodes[0];
1286 		slot = path->slots[0];
1287 		if (ret > 0 && slot > 0)
1288 			slot--;
1289 		btrfs_item_key_to_cpu(l, &key, slot);
1290 
1291 		if (ret > 0 && (key.objectid != dirid ||
1292 				key.type != BTRFS_INODE_REF_KEY)) {
1293 			ret = -ENOENT;
1294 			goto out;
1295 		}
1296 
1297 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1298 		len = btrfs_inode_ref_name_len(l, iref);
1299 		ptr -= len + 1;
1300 		total_len += len + 1;
1301 		if (ptr < name)
1302 			goto out;
1303 
1304 		*(ptr + len) = '/';
1305 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1306 
1307 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1308 			break;
1309 
1310 		btrfs_release_path(root, path);
1311 		key.objectid = key.offset;
1312 		key.offset = (u64)-1;
1313 		dirid = key.objectid;
1314 
1315 	}
1316 	if (ptr < name)
1317 		goto out;
1318 	memcpy(name, ptr, total_len);
1319 	name[total_len]='\0';
1320 	ret = 0;
1321 out:
1322 	btrfs_free_path(path);
1323 	return ret;
1324 }
1325 
1326 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1327 					   void __user *argp)
1328 {
1329 	 struct btrfs_ioctl_ino_lookup_args *args;
1330 	 struct inode *inode;
1331 	 int ret;
1332 
1333 	if (!capable(CAP_SYS_ADMIN))
1334 		return -EPERM;
1335 
1336 	args = memdup_user(argp, sizeof(*args));
1337 	if (IS_ERR(args))
1338 		return PTR_ERR(args);
1339 
1340 	inode = fdentry(file)->d_inode;
1341 
1342 	if (args->treeid == 0)
1343 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1344 
1345 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1346 					args->treeid, args->objectid,
1347 					args->name);
1348 
1349 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1350 		ret = -EFAULT;
1351 
1352 	kfree(args);
1353 	return ret;
1354 }
1355 
1356 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1357 					     void __user *arg)
1358 {
1359 	struct dentry *parent = fdentry(file);
1360 	struct dentry *dentry;
1361 	struct inode *dir = parent->d_inode;
1362 	struct inode *inode;
1363 	struct btrfs_root *root = BTRFS_I(dir)->root;
1364 	struct btrfs_root *dest = NULL;
1365 	struct btrfs_ioctl_vol_args *vol_args;
1366 	struct btrfs_trans_handle *trans;
1367 	int namelen;
1368 	int ret;
1369 	int err = 0;
1370 
1371 	vol_args = memdup_user(arg, sizeof(*vol_args));
1372 	if (IS_ERR(vol_args))
1373 		return PTR_ERR(vol_args);
1374 
1375 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1376 	namelen = strlen(vol_args->name);
1377 	if (strchr(vol_args->name, '/') ||
1378 	    strncmp(vol_args->name, "..", namelen) == 0) {
1379 		err = -EINVAL;
1380 		goto out;
1381 	}
1382 
1383 	err = mnt_want_write(file->f_path.mnt);
1384 	if (err)
1385 		goto out;
1386 
1387 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1388 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1389 	if (IS_ERR(dentry)) {
1390 		err = PTR_ERR(dentry);
1391 		goto out_unlock_dir;
1392 	}
1393 
1394 	if (!dentry->d_inode) {
1395 		err = -ENOENT;
1396 		goto out_dput;
1397 	}
1398 
1399 	inode = dentry->d_inode;
1400 	dest = BTRFS_I(inode)->root;
1401 	if (!capable(CAP_SYS_ADMIN)){
1402 		/*
1403 		 * Regular user.  Only allow this with a special mount
1404 		 * option, when the user has write+exec access to the
1405 		 * subvol root, and when rmdir(2) would have been
1406 		 * allowed.
1407 		 *
1408 		 * Note that this is _not_ check that the subvol is
1409 		 * empty or doesn't contain data that we wouldn't
1410 		 * otherwise be able to delete.
1411 		 *
1412 		 * Users who want to delete empty subvols should try
1413 		 * rmdir(2).
1414 		 */
1415 		err = -EPERM;
1416 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1417 			goto out_dput;
1418 
1419 		/*
1420 		 * Do not allow deletion if the parent dir is the same
1421 		 * as the dir to be deleted.  That means the ioctl
1422 		 * must be called on the dentry referencing the root
1423 		 * of the subvol, not a random directory contained
1424 		 * within it.
1425 		 */
1426 		err = -EINVAL;
1427 		if (root == dest)
1428 			goto out_dput;
1429 
1430 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1431 		if (err)
1432 			goto out_dput;
1433 
1434 		/* check if subvolume may be deleted by a non-root user */
1435 		err = btrfs_may_delete(dir, dentry, 1);
1436 		if (err)
1437 			goto out_dput;
1438 	}
1439 
1440 	if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
1441 		err = -EINVAL;
1442 		goto out_dput;
1443 	}
1444 
1445 	mutex_lock(&inode->i_mutex);
1446 	err = d_invalidate(dentry);
1447 	if (err)
1448 		goto out_unlock;
1449 
1450 	down_write(&root->fs_info->subvol_sem);
1451 
1452 	err = may_destroy_subvol(dest);
1453 	if (err)
1454 		goto out_up_write;
1455 
1456 	trans = btrfs_start_transaction(root, 0);
1457 	if (IS_ERR(trans)) {
1458 		err = PTR_ERR(trans);
1459 		goto out_up_write;
1460 	}
1461 	trans->block_rsv = &root->fs_info->global_block_rsv;
1462 
1463 	ret = btrfs_unlink_subvol(trans, root, dir,
1464 				dest->root_key.objectid,
1465 				dentry->d_name.name,
1466 				dentry->d_name.len);
1467 	BUG_ON(ret);
1468 
1469 	btrfs_record_root_in_trans(trans, dest);
1470 
1471 	memset(&dest->root_item.drop_progress, 0,
1472 		sizeof(dest->root_item.drop_progress));
1473 	dest->root_item.drop_level = 0;
1474 	btrfs_set_root_refs(&dest->root_item, 0);
1475 
1476 	if (!xchg(&dest->orphan_item_inserted, 1)) {
1477 		ret = btrfs_insert_orphan_item(trans,
1478 					root->fs_info->tree_root,
1479 					dest->root_key.objectid);
1480 		BUG_ON(ret);
1481 	}
1482 
1483 	ret = btrfs_end_transaction(trans, root);
1484 	BUG_ON(ret);
1485 	inode->i_flags |= S_DEAD;
1486 out_up_write:
1487 	up_write(&root->fs_info->subvol_sem);
1488 out_unlock:
1489 	mutex_unlock(&inode->i_mutex);
1490 	if (!err) {
1491 		shrink_dcache_sb(root->fs_info->sb);
1492 		btrfs_invalidate_inodes(dest);
1493 		d_delete(dentry);
1494 	}
1495 out_dput:
1496 	dput(dentry);
1497 out_unlock_dir:
1498 	mutex_unlock(&dir->i_mutex);
1499 	mnt_drop_write(file->f_path.mnt);
1500 out:
1501 	kfree(vol_args);
1502 	return err;
1503 }
1504 
1505 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1506 {
1507 	struct inode *inode = fdentry(file)->d_inode;
1508 	struct btrfs_root *root = BTRFS_I(inode)->root;
1509 	struct btrfs_ioctl_defrag_range_args *range;
1510 	int ret;
1511 
1512 	ret = mnt_want_write(file->f_path.mnt);
1513 	if (ret)
1514 		return ret;
1515 
1516 	switch (inode->i_mode & S_IFMT) {
1517 	case S_IFDIR:
1518 		if (!capable(CAP_SYS_ADMIN)) {
1519 			ret = -EPERM;
1520 			goto out;
1521 		}
1522 		ret = btrfs_defrag_root(root, 0);
1523 		if (ret)
1524 			goto out;
1525 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1526 		break;
1527 	case S_IFREG:
1528 		if (!(file->f_mode & FMODE_WRITE)) {
1529 			ret = -EINVAL;
1530 			goto out;
1531 		}
1532 
1533 		range = kzalloc(sizeof(*range), GFP_KERNEL);
1534 		if (!range) {
1535 			ret = -ENOMEM;
1536 			goto out;
1537 		}
1538 
1539 		if (argp) {
1540 			if (copy_from_user(range, argp,
1541 					   sizeof(*range))) {
1542 				ret = -EFAULT;
1543 				kfree(range);
1544 				goto out;
1545 			}
1546 			/* compression requires us to start the IO */
1547 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1548 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1549 				range->extent_thresh = (u32)-1;
1550 			}
1551 		} else {
1552 			/* the rest are all set to zero by kzalloc */
1553 			range->len = (u64)-1;
1554 		}
1555 		ret = btrfs_defrag_file(file, range);
1556 		kfree(range);
1557 		break;
1558 	default:
1559 		ret = -EINVAL;
1560 	}
1561 out:
1562 	mnt_drop_write(file->f_path.mnt);
1563 	return ret;
1564 }
1565 
1566 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
1567 {
1568 	struct btrfs_ioctl_vol_args *vol_args;
1569 	int ret;
1570 
1571 	if (!capable(CAP_SYS_ADMIN))
1572 		return -EPERM;
1573 
1574 	vol_args = memdup_user(arg, sizeof(*vol_args));
1575 	if (IS_ERR(vol_args))
1576 		return PTR_ERR(vol_args);
1577 
1578 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1579 	ret = btrfs_init_new_device(root, vol_args->name);
1580 
1581 	kfree(vol_args);
1582 	return ret;
1583 }
1584 
1585 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
1586 {
1587 	struct btrfs_ioctl_vol_args *vol_args;
1588 	int ret;
1589 
1590 	if (!capable(CAP_SYS_ADMIN))
1591 		return -EPERM;
1592 
1593 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1594 		return -EROFS;
1595 
1596 	vol_args = memdup_user(arg, sizeof(*vol_args));
1597 	if (IS_ERR(vol_args))
1598 		return PTR_ERR(vol_args);
1599 
1600 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1601 	ret = btrfs_rm_device(root, vol_args->name);
1602 
1603 	kfree(vol_args);
1604 	return ret;
1605 }
1606 
1607 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1608 				       u64 off, u64 olen, u64 destoff)
1609 {
1610 	struct inode *inode = fdentry(file)->d_inode;
1611 	struct btrfs_root *root = BTRFS_I(inode)->root;
1612 	struct file *src_file;
1613 	struct inode *src;
1614 	struct btrfs_trans_handle *trans;
1615 	struct btrfs_path *path;
1616 	struct extent_buffer *leaf;
1617 	char *buf;
1618 	struct btrfs_key key;
1619 	u32 nritems;
1620 	int slot;
1621 	int ret;
1622 	u64 len = olen;
1623 	u64 bs = root->fs_info->sb->s_blocksize;
1624 	u64 hint_byte;
1625 
1626 	/*
1627 	 * TODO:
1628 	 * - split compressed inline extents.  annoying: we need to
1629 	 *   decompress into destination's address_space (the file offset
1630 	 *   may change, so source mapping won't do), then recompress (or
1631 	 *   otherwise reinsert) a subrange.
1632 	 * - allow ranges within the same file to be cloned (provided
1633 	 *   they don't overlap)?
1634 	 */
1635 
1636 	/* the destination must be opened for writing */
1637 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
1638 		return -EINVAL;
1639 
1640 	ret = mnt_want_write(file->f_path.mnt);
1641 	if (ret)
1642 		return ret;
1643 
1644 	src_file = fget(srcfd);
1645 	if (!src_file) {
1646 		ret = -EBADF;
1647 		goto out_drop_write;
1648 	}
1649 
1650 	src = src_file->f_dentry->d_inode;
1651 
1652 	ret = -EINVAL;
1653 	if (src == inode)
1654 		goto out_fput;
1655 
1656 	/* the src must be open for reading */
1657 	if (!(src_file->f_mode & FMODE_READ))
1658 		goto out_fput;
1659 
1660 	ret = -EISDIR;
1661 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
1662 		goto out_fput;
1663 
1664 	ret = -EXDEV;
1665 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
1666 		goto out_fput;
1667 
1668 	ret = -ENOMEM;
1669 	buf = vmalloc(btrfs_level_size(root, 0));
1670 	if (!buf)
1671 		goto out_fput;
1672 
1673 	path = btrfs_alloc_path();
1674 	if (!path) {
1675 		vfree(buf);
1676 		goto out_fput;
1677 	}
1678 	path->reada = 2;
1679 
1680 	if (inode < src) {
1681 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
1682 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
1683 	} else {
1684 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
1685 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
1686 	}
1687 
1688 	/* determine range to clone */
1689 	ret = -EINVAL;
1690 	if (off + len > src->i_size || off + len < off)
1691 		goto out_unlock;
1692 	if (len == 0)
1693 		olen = len = src->i_size - off;
1694 	/* if we extend to eof, continue to block boundary */
1695 	if (off + len == src->i_size)
1696 		len = ALIGN(src->i_size, bs) - off;
1697 
1698 	/* verify the end result is block aligned */
1699 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
1700 	    !IS_ALIGNED(destoff, bs))
1701 		goto out_unlock;
1702 
1703 	/* do any pending delalloc/csum calc on src, one way or
1704 	   another, and lock file content */
1705 	while (1) {
1706 		struct btrfs_ordered_extent *ordered;
1707 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1708 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
1709 		if (!ordered &&
1710 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
1711 				   EXTENT_DELALLOC, 0, NULL))
1712 			break;
1713 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1714 		if (ordered)
1715 			btrfs_put_ordered_extent(ordered);
1716 		btrfs_wait_ordered_range(src, off, len);
1717 	}
1718 
1719 	/* clone data */
1720 	key.objectid = src->i_ino;
1721 	key.type = BTRFS_EXTENT_DATA_KEY;
1722 	key.offset = 0;
1723 
1724 	while (1) {
1725 		/*
1726 		 * note the key will change type as we walk through the
1727 		 * tree.
1728 		 */
1729 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1730 		if (ret < 0)
1731 			goto out;
1732 
1733 		nritems = btrfs_header_nritems(path->nodes[0]);
1734 		if (path->slots[0] >= nritems) {
1735 			ret = btrfs_next_leaf(root, path);
1736 			if (ret < 0)
1737 				goto out;
1738 			if (ret > 0)
1739 				break;
1740 			nritems = btrfs_header_nritems(path->nodes[0]);
1741 		}
1742 		leaf = path->nodes[0];
1743 		slot = path->slots[0];
1744 
1745 		btrfs_item_key_to_cpu(leaf, &key, slot);
1746 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1747 		    key.objectid != src->i_ino)
1748 			break;
1749 
1750 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1751 			struct btrfs_file_extent_item *extent;
1752 			int type;
1753 			u32 size;
1754 			struct btrfs_key new_key;
1755 			u64 disko = 0, diskl = 0;
1756 			u64 datao = 0, datal = 0;
1757 			u8 comp;
1758 			u64 endoff;
1759 
1760 			size = btrfs_item_size_nr(leaf, slot);
1761 			read_extent_buffer(leaf, buf,
1762 					   btrfs_item_ptr_offset(leaf, slot),
1763 					   size);
1764 
1765 			extent = btrfs_item_ptr(leaf, slot,
1766 						struct btrfs_file_extent_item);
1767 			comp = btrfs_file_extent_compression(leaf, extent);
1768 			type = btrfs_file_extent_type(leaf, extent);
1769 			if (type == BTRFS_FILE_EXTENT_REG ||
1770 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
1771 				disko = btrfs_file_extent_disk_bytenr(leaf,
1772 								      extent);
1773 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
1774 								 extent);
1775 				datao = btrfs_file_extent_offset(leaf, extent);
1776 				datal = btrfs_file_extent_num_bytes(leaf,
1777 								    extent);
1778 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
1779 				/* take upper bound, may be compressed */
1780 				datal = btrfs_file_extent_ram_bytes(leaf,
1781 								    extent);
1782 			}
1783 			btrfs_release_path(root, path);
1784 
1785 			if (key.offset + datal <= off ||
1786 			    key.offset >= off+len)
1787 				goto next;
1788 
1789 			memcpy(&new_key, &key, sizeof(new_key));
1790 			new_key.objectid = inode->i_ino;
1791 			new_key.offset = key.offset + destoff - off;
1792 
1793 			trans = btrfs_start_transaction(root, 1);
1794 			if (IS_ERR(trans)) {
1795 				ret = PTR_ERR(trans);
1796 				goto out;
1797 			}
1798 
1799 			if (type == BTRFS_FILE_EXTENT_REG ||
1800 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
1801 				if (off > key.offset) {
1802 					datao += off - key.offset;
1803 					datal -= off - key.offset;
1804 				}
1805 
1806 				if (key.offset + datal > off + len)
1807 					datal = off + len - key.offset;
1808 
1809 				ret = btrfs_drop_extents(trans, inode,
1810 							 new_key.offset,
1811 							 new_key.offset + datal,
1812 							 &hint_byte, 1);
1813 				BUG_ON(ret);
1814 
1815 				ret = btrfs_insert_empty_item(trans, root, path,
1816 							      &new_key, size);
1817 				BUG_ON(ret);
1818 
1819 				leaf = path->nodes[0];
1820 				slot = path->slots[0];
1821 				write_extent_buffer(leaf, buf,
1822 					    btrfs_item_ptr_offset(leaf, slot),
1823 					    size);
1824 
1825 				extent = btrfs_item_ptr(leaf, slot,
1826 						struct btrfs_file_extent_item);
1827 
1828 				/* disko == 0 means it's a hole */
1829 				if (!disko)
1830 					datao = 0;
1831 
1832 				btrfs_set_file_extent_offset(leaf, extent,
1833 							     datao);
1834 				btrfs_set_file_extent_num_bytes(leaf, extent,
1835 								datal);
1836 				if (disko) {
1837 					inode_add_bytes(inode, datal);
1838 					ret = btrfs_inc_extent_ref(trans, root,
1839 							disko, diskl, 0,
1840 							root->root_key.objectid,
1841 							inode->i_ino,
1842 							new_key.offset - datao);
1843 					BUG_ON(ret);
1844 				}
1845 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
1846 				u64 skip = 0;
1847 				u64 trim = 0;
1848 				if (off > key.offset) {
1849 					skip = off - key.offset;
1850 					new_key.offset += skip;
1851 				}
1852 
1853 				if (key.offset + datal > off+len)
1854 					trim = key.offset + datal - (off+len);
1855 
1856 				if (comp && (skip || trim)) {
1857 					ret = -EINVAL;
1858 					btrfs_end_transaction(trans, root);
1859 					goto out;
1860 				}
1861 				size -= skip + trim;
1862 				datal -= skip + trim;
1863 
1864 				ret = btrfs_drop_extents(trans, inode,
1865 							 new_key.offset,
1866 							 new_key.offset + datal,
1867 							 &hint_byte, 1);
1868 				BUG_ON(ret);
1869 
1870 				ret = btrfs_insert_empty_item(trans, root, path,
1871 							      &new_key, size);
1872 				BUG_ON(ret);
1873 
1874 				if (skip) {
1875 					u32 start =
1876 					  btrfs_file_extent_calc_inline_size(0);
1877 					memmove(buf+start, buf+start+skip,
1878 						datal);
1879 				}
1880 
1881 				leaf = path->nodes[0];
1882 				slot = path->slots[0];
1883 				write_extent_buffer(leaf, buf,
1884 					    btrfs_item_ptr_offset(leaf, slot),
1885 					    size);
1886 				inode_add_bytes(inode, datal);
1887 			}
1888 
1889 			btrfs_mark_buffer_dirty(leaf);
1890 			btrfs_release_path(root, path);
1891 
1892 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1893 
1894 			/*
1895 			 * we round up to the block size at eof when
1896 			 * determining which extents to clone above,
1897 			 * but shouldn't round up the file size
1898 			 */
1899 			endoff = new_key.offset + datal;
1900 			if (endoff > destoff+olen)
1901 				endoff = destoff+olen;
1902 			if (endoff > inode->i_size)
1903 				btrfs_i_size_write(inode, endoff);
1904 
1905 			BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1906 			ret = btrfs_update_inode(trans, root, inode);
1907 			BUG_ON(ret);
1908 			btrfs_end_transaction(trans, root);
1909 		}
1910 next:
1911 		btrfs_release_path(root, path);
1912 		key.offset++;
1913 	}
1914 	ret = 0;
1915 out:
1916 	btrfs_release_path(root, path);
1917 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1918 out_unlock:
1919 	mutex_unlock(&src->i_mutex);
1920 	mutex_unlock(&inode->i_mutex);
1921 	vfree(buf);
1922 	btrfs_free_path(path);
1923 out_fput:
1924 	fput(src_file);
1925 out_drop_write:
1926 	mnt_drop_write(file->f_path.mnt);
1927 	return ret;
1928 }
1929 
1930 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
1931 {
1932 	struct btrfs_ioctl_clone_range_args args;
1933 
1934 	if (copy_from_user(&args, argp, sizeof(args)))
1935 		return -EFAULT;
1936 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1937 				 args.src_length, args.dest_offset);
1938 }
1939 
1940 /*
1941  * there are many ways the trans_start and trans_end ioctls can lead
1942  * to deadlocks.  They should only be used by applications that
1943  * basically own the machine, and have a very in depth understanding
1944  * of all the possible deadlocks and enospc problems.
1945  */
1946 static long btrfs_ioctl_trans_start(struct file *file)
1947 {
1948 	struct inode *inode = fdentry(file)->d_inode;
1949 	struct btrfs_root *root = BTRFS_I(inode)->root;
1950 	struct btrfs_trans_handle *trans;
1951 	int ret;
1952 
1953 	ret = -EPERM;
1954 	if (!capable(CAP_SYS_ADMIN))
1955 		goto out;
1956 
1957 	ret = -EINPROGRESS;
1958 	if (file->private_data)
1959 		goto out;
1960 
1961 	ret = mnt_want_write(file->f_path.mnt);
1962 	if (ret)
1963 		goto out;
1964 
1965 	mutex_lock(&root->fs_info->trans_mutex);
1966 	root->fs_info->open_ioctl_trans++;
1967 	mutex_unlock(&root->fs_info->trans_mutex);
1968 
1969 	ret = -ENOMEM;
1970 	trans = btrfs_start_ioctl_transaction(root, 0);
1971 	if (!trans)
1972 		goto out_drop;
1973 
1974 	file->private_data = trans;
1975 	return 0;
1976 
1977 out_drop:
1978 	mutex_lock(&root->fs_info->trans_mutex);
1979 	root->fs_info->open_ioctl_trans--;
1980 	mutex_unlock(&root->fs_info->trans_mutex);
1981 	mnt_drop_write(file->f_path.mnt);
1982 out:
1983 	return ret;
1984 }
1985 
1986 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
1987 {
1988 	struct inode *inode = fdentry(file)->d_inode;
1989 	struct btrfs_root *root = BTRFS_I(inode)->root;
1990 	struct btrfs_root *new_root;
1991 	struct btrfs_dir_item *di;
1992 	struct btrfs_trans_handle *trans;
1993 	struct btrfs_path *path;
1994 	struct btrfs_key location;
1995 	struct btrfs_disk_key disk_key;
1996 	struct btrfs_super_block *disk_super;
1997 	u64 features;
1998 	u64 objectid = 0;
1999 	u64 dir_id;
2000 
2001 	if (!capable(CAP_SYS_ADMIN))
2002 		return -EPERM;
2003 
2004 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2005 		return -EFAULT;
2006 
2007 	if (!objectid)
2008 		objectid = root->root_key.objectid;
2009 
2010 	location.objectid = objectid;
2011 	location.type = BTRFS_ROOT_ITEM_KEY;
2012 	location.offset = (u64)-1;
2013 
2014 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2015 	if (IS_ERR(new_root))
2016 		return PTR_ERR(new_root);
2017 
2018 	if (btrfs_root_refs(&new_root->root_item) == 0)
2019 		return -ENOENT;
2020 
2021 	path = btrfs_alloc_path();
2022 	if (!path)
2023 		return -ENOMEM;
2024 	path->leave_spinning = 1;
2025 
2026 	trans = btrfs_start_transaction(root, 1);
2027 	if (!trans) {
2028 		btrfs_free_path(path);
2029 		return -ENOMEM;
2030 	}
2031 
2032 	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2033 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2034 				   dir_id, "default", 7, 1);
2035 	if (IS_ERR_OR_NULL(di)) {
2036 		btrfs_free_path(path);
2037 		btrfs_end_transaction(trans, root);
2038 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2039 		       "this isn't going to work\n");
2040 		return -ENOENT;
2041 	}
2042 
2043 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2044 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2045 	btrfs_mark_buffer_dirty(path->nodes[0]);
2046 	btrfs_free_path(path);
2047 
2048 	disk_super = &root->fs_info->super_copy;
2049 	features = btrfs_super_incompat_flags(disk_super);
2050 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2051 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2052 		btrfs_set_super_incompat_flags(disk_super, features);
2053 	}
2054 	btrfs_end_transaction(trans, root);
2055 
2056 	return 0;
2057 }
2058 
2059 static void get_block_group_info(struct list_head *groups_list,
2060 				 struct btrfs_ioctl_space_info *space)
2061 {
2062 	struct btrfs_block_group_cache *block_group;
2063 
2064 	space->total_bytes = 0;
2065 	space->used_bytes = 0;
2066 	space->flags = 0;
2067 	list_for_each_entry(block_group, groups_list, list) {
2068 		space->flags = block_group->flags;
2069 		space->total_bytes += block_group->key.offset;
2070 		space->used_bytes +=
2071 			btrfs_block_group_used(&block_group->item);
2072 	}
2073 }
2074 
2075 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2076 {
2077 	struct btrfs_ioctl_space_args space_args;
2078 	struct btrfs_ioctl_space_info space;
2079 	struct btrfs_ioctl_space_info *dest;
2080 	struct btrfs_ioctl_space_info *dest_orig;
2081 	struct btrfs_ioctl_space_info *user_dest;
2082 	struct btrfs_space_info *info;
2083 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2084 		       BTRFS_BLOCK_GROUP_SYSTEM,
2085 		       BTRFS_BLOCK_GROUP_METADATA,
2086 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2087 	int num_types = 4;
2088 	int alloc_size;
2089 	int ret = 0;
2090 	int slot_count = 0;
2091 	int i, c;
2092 
2093 	if (copy_from_user(&space_args,
2094 			   (struct btrfs_ioctl_space_args __user *)arg,
2095 			   sizeof(space_args)))
2096 		return -EFAULT;
2097 
2098 	for (i = 0; i < num_types; i++) {
2099 		struct btrfs_space_info *tmp;
2100 
2101 		info = NULL;
2102 		rcu_read_lock();
2103 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2104 					list) {
2105 			if (tmp->flags == types[i]) {
2106 				info = tmp;
2107 				break;
2108 			}
2109 		}
2110 		rcu_read_unlock();
2111 
2112 		if (!info)
2113 			continue;
2114 
2115 		down_read(&info->groups_sem);
2116 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2117 			if (!list_empty(&info->block_groups[c]))
2118 				slot_count++;
2119 		}
2120 		up_read(&info->groups_sem);
2121 	}
2122 
2123 	/* space_slots == 0 means they are asking for a count */
2124 	if (space_args.space_slots == 0) {
2125 		space_args.total_spaces = slot_count;
2126 		goto out;
2127 	}
2128 
2129 	slot_count = min_t(int, space_args.space_slots, slot_count);
2130 
2131 	alloc_size = sizeof(*dest) * slot_count;
2132 
2133 	/* we generally have at most 6 or so space infos, one for each raid
2134 	 * level.  So, a whole page should be more than enough for everyone
2135 	 */
2136 	if (alloc_size > PAGE_CACHE_SIZE)
2137 		return -ENOMEM;
2138 
2139 	space_args.total_spaces = 0;
2140 	dest = kmalloc(alloc_size, GFP_NOFS);
2141 	if (!dest)
2142 		return -ENOMEM;
2143 	dest_orig = dest;
2144 
2145 	/* now we have a buffer to copy into */
2146 	for (i = 0; i < num_types; i++) {
2147 		struct btrfs_space_info *tmp;
2148 
2149 		info = NULL;
2150 		rcu_read_lock();
2151 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2152 					list) {
2153 			if (tmp->flags == types[i]) {
2154 				info = tmp;
2155 				break;
2156 			}
2157 		}
2158 		rcu_read_unlock();
2159 
2160 		if (!info)
2161 			continue;
2162 		down_read(&info->groups_sem);
2163 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2164 			if (!list_empty(&info->block_groups[c])) {
2165 				get_block_group_info(&info->block_groups[c],
2166 						     &space);
2167 				memcpy(dest, &space, sizeof(space));
2168 				dest++;
2169 				space_args.total_spaces++;
2170 			}
2171 		}
2172 		up_read(&info->groups_sem);
2173 	}
2174 
2175 	user_dest = (struct btrfs_ioctl_space_info *)
2176 		(arg + sizeof(struct btrfs_ioctl_space_args));
2177 
2178 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2179 		ret = -EFAULT;
2180 
2181 	kfree(dest_orig);
2182 out:
2183 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2184 		ret = -EFAULT;
2185 
2186 	return ret;
2187 }
2188 
2189 /*
2190  * there are many ways the trans_start and trans_end ioctls can lead
2191  * to deadlocks.  They should only be used by applications that
2192  * basically own the machine, and have a very in depth understanding
2193  * of all the possible deadlocks and enospc problems.
2194  */
2195 long btrfs_ioctl_trans_end(struct file *file)
2196 {
2197 	struct inode *inode = fdentry(file)->d_inode;
2198 	struct btrfs_root *root = BTRFS_I(inode)->root;
2199 	struct btrfs_trans_handle *trans;
2200 
2201 	trans = file->private_data;
2202 	if (!trans)
2203 		return -EINVAL;
2204 	file->private_data = NULL;
2205 
2206 	btrfs_end_transaction(trans, root);
2207 
2208 	mutex_lock(&root->fs_info->trans_mutex);
2209 	root->fs_info->open_ioctl_trans--;
2210 	mutex_unlock(&root->fs_info->trans_mutex);
2211 
2212 	mnt_drop_write(file->f_path.mnt);
2213 	return 0;
2214 }
2215 
2216 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2217 {
2218 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2219 	struct btrfs_trans_handle *trans;
2220 	u64 transid;
2221 
2222 	trans = btrfs_start_transaction(root, 0);
2223 	transid = trans->transid;
2224 	btrfs_commit_transaction_async(trans, root, 0);
2225 
2226 	if (argp)
2227 		if (copy_to_user(argp, &transid, sizeof(transid)))
2228 			return -EFAULT;
2229 	return 0;
2230 }
2231 
2232 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2233 {
2234 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2235 	u64 transid;
2236 
2237 	if (argp) {
2238 		if (copy_from_user(&transid, argp, sizeof(transid)))
2239 			return -EFAULT;
2240 	} else {
2241 		transid = 0;  /* current trans */
2242 	}
2243 	return btrfs_wait_for_commit(root, transid);
2244 }
2245 
2246 long btrfs_ioctl(struct file *file, unsigned int
2247 		cmd, unsigned long arg)
2248 {
2249 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2250 	void __user *argp = (void __user *)arg;
2251 
2252 	switch (cmd) {
2253 	case FS_IOC_GETFLAGS:
2254 		return btrfs_ioctl_getflags(file, argp);
2255 	case FS_IOC_SETFLAGS:
2256 		return btrfs_ioctl_setflags(file, argp);
2257 	case FS_IOC_GETVERSION:
2258 		return btrfs_ioctl_getversion(file, argp);
2259 	case BTRFS_IOC_SNAP_CREATE:
2260 		return btrfs_ioctl_snap_create(file, argp, 0, 0);
2261 	case BTRFS_IOC_SNAP_CREATE_V2:
2262 		return btrfs_ioctl_snap_create(file, argp, 0, 1);
2263 	case BTRFS_IOC_SUBVOL_CREATE:
2264 		return btrfs_ioctl_snap_create(file, argp, 1, 0);
2265 	case BTRFS_IOC_SNAP_DESTROY:
2266 		return btrfs_ioctl_snap_destroy(file, argp);
2267 	case BTRFS_IOC_DEFAULT_SUBVOL:
2268 		return btrfs_ioctl_default_subvol(file, argp);
2269 	case BTRFS_IOC_DEFRAG:
2270 		return btrfs_ioctl_defrag(file, NULL);
2271 	case BTRFS_IOC_DEFRAG_RANGE:
2272 		return btrfs_ioctl_defrag(file, argp);
2273 	case BTRFS_IOC_RESIZE:
2274 		return btrfs_ioctl_resize(root, argp);
2275 	case BTRFS_IOC_ADD_DEV:
2276 		return btrfs_ioctl_add_dev(root, argp);
2277 	case BTRFS_IOC_RM_DEV:
2278 		return btrfs_ioctl_rm_dev(root, argp);
2279 	case BTRFS_IOC_BALANCE:
2280 		return btrfs_balance(root->fs_info->dev_root);
2281 	case BTRFS_IOC_CLONE:
2282 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2283 	case BTRFS_IOC_CLONE_RANGE:
2284 		return btrfs_ioctl_clone_range(file, argp);
2285 	case BTRFS_IOC_TRANS_START:
2286 		return btrfs_ioctl_trans_start(file);
2287 	case BTRFS_IOC_TRANS_END:
2288 		return btrfs_ioctl_trans_end(file);
2289 	case BTRFS_IOC_TREE_SEARCH:
2290 		return btrfs_ioctl_tree_search(file, argp);
2291 	case BTRFS_IOC_INO_LOOKUP:
2292 		return btrfs_ioctl_ino_lookup(file, argp);
2293 	case BTRFS_IOC_SPACE_INFO:
2294 		return btrfs_ioctl_space_info(root, argp);
2295 	case BTRFS_IOC_SYNC:
2296 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
2297 		return 0;
2298 	case BTRFS_IOC_START_SYNC:
2299 		return btrfs_ioctl_start_sync(file, argp);
2300 	case BTRFS_IOC_WAIT_SYNC:
2301 		return btrfs_ioctl_wait_sync(file, argp);
2302 	}
2303 
2304 	return -ENOTTY;
2305 }
2306