xref: /openbmc/linux/fs/btrfs/ioctl.c (revision 81d67439)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include "compat.h"
45 #include "ctree.h"
46 #include "disk-io.h"
47 #include "transaction.h"
48 #include "btrfs_inode.h"
49 #include "ioctl.h"
50 #include "print-tree.h"
51 #include "volumes.h"
52 #include "locking.h"
53 #include "inode-map.h"
54 
55 /* Mask out flags that are inappropriate for the given type of inode. */
56 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
57 {
58 	if (S_ISDIR(mode))
59 		return flags;
60 	else if (S_ISREG(mode))
61 		return flags & ~FS_DIRSYNC_FL;
62 	else
63 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
64 }
65 
66 /*
67  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
68  */
69 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
70 {
71 	unsigned int iflags = 0;
72 
73 	if (flags & BTRFS_INODE_SYNC)
74 		iflags |= FS_SYNC_FL;
75 	if (flags & BTRFS_INODE_IMMUTABLE)
76 		iflags |= FS_IMMUTABLE_FL;
77 	if (flags & BTRFS_INODE_APPEND)
78 		iflags |= FS_APPEND_FL;
79 	if (flags & BTRFS_INODE_NODUMP)
80 		iflags |= FS_NODUMP_FL;
81 	if (flags & BTRFS_INODE_NOATIME)
82 		iflags |= FS_NOATIME_FL;
83 	if (flags & BTRFS_INODE_DIRSYNC)
84 		iflags |= FS_DIRSYNC_FL;
85 	if (flags & BTRFS_INODE_NODATACOW)
86 		iflags |= FS_NOCOW_FL;
87 
88 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
89 		iflags |= FS_COMPR_FL;
90 	else if (flags & BTRFS_INODE_NOCOMPRESS)
91 		iflags |= FS_NOCOMP_FL;
92 
93 	return iflags;
94 }
95 
96 /*
97  * Update inode->i_flags based on the btrfs internal flags.
98  */
99 void btrfs_update_iflags(struct inode *inode)
100 {
101 	struct btrfs_inode *ip = BTRFS_I(inode);
102 
103 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
104 
105 	if (ip->flags & BTRFS_INODE_SYNC)
106 		inode->i_flags |= S_SYNC;
107 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
108 		inode->i_flags |= S_IMMUTABLE;
109 	if (ip->flags & BTRFS_INODE_APPEND)
110 		inode->i_flags |= S_APPEND;
111 	if (ip->flags & BTRFS_INODE_NOATIME)
112 		inode->i_flags |= S_NOATIME;
113 	if (ip->flags & BTRFS_INODE_DIRSYNC)
114 		inode->i_flags |= S_DIRSYNC;
115 }
116 
117 /*
118  * Inherit flags from the parent inode.
119  *
120  * Unlike extN we don't have any flags we don't want to inherit currently.
121  */
122 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
123 {
124 	unsigned int flags;
125 
126 	if (!dir)
127 		return;
128 
129 	flags = BTRFS_I(dir)->flags;
130 
131 	if (S_ISREG(inode->i_mode))
132 		flags &= ~BTRFS_INODE_DIRSYNC;
133 	else if (!S_ISDIR(inode->i_mode))
134 		flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
135 
136 	BTRFS_I(inode)->flags = flags;
137 	btrfs_update_iflags(inode);
138 }
139 
140 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
141 {
142 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
143 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
144 
145 	if (copy_to_user(arg, &flags, sizeof(flags)))
146 		return -EFAULT;
147 	return 0;
148 }
149 
150 static int check_flags(unsigned int flags)
151 {
152 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
153 		      FS_NOATIME_FL | FS_NODUMP_FL | \
154 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
155 		      FS_NOCOMP_FL | FS_COMPR_FL |
156 		      FS_NOCOW_FL))
157 		return -EOPNOTSUPP;
158 
159 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
160 		return -EINVAL;
161 
162 	return 0;
163 }
164 
165 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
166 {
167 	struct inode *inode = file->f_path.dentry->d_inode;
168 	struct btrfs_inode *ip = BTRFS_I(inode);
169 	struct btrfs_root *root = ip->root;
170 	struct btrfs_trans_handle *trans;
171 	unsigned int flags, oldflags;
172 	int ret;
173 
174 	if (btrfs_root_readonly(root))
175 		return -EROFS;
176 
177 	if (copy_from_user(&flags, arg, sizeof(flags)))
178 		return -EFAULT;
179 
180 	ret = check_flags(flags);
181 	if (ret)
182 		return ret;
183 
184 	if (!inode_owner_or_capable(inode))
185 		return -EACCES;
186 
187 	mutex_lock(&inode->i_mutex);
188 
189 	flags = btrfs_mask_flags(inode->i_mode, flags);
190 	oldflags = btrfs_flags_to_ioctl(ip->flags);
191 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
192 		if (!capable(CAP_LINUX_IMMUTABLE)) {
193 			ret = -EPERM;
194 			goto out_unlock;
195 		}
196 	}
197 
198 	ret = mnt_want_write(file->f_path.mnt);
199 	if (ret)
200 		goto out_unlock;
201 
202 	if (flags & FS_SYNC_FL)
203 		ip->flags |= BTRFS_INODE_SYNC;
204 	else
205 		ip->flags &= ~BTRFS_INODE_SYNC;
206 	if (flags & FS_IMMUTABLE_FL)
207 		ip->flags |= BTRFS_INODE_IMMUTABLE;
208 	else
209 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
210 	if (flags & FS_APPEND_FL)
211 		ip->flags |= BTRFS_INODE_APPEND;
212 	else
213 		ip->flags &= ~BTRFS_INODE_APPEND;
214 	if (flags & FS_NODUMP_FL)
215 		ip->flags |= BTRFS_INODE_NODUMP;
216 	else
217 		ip->flags &= ~BTRFS_INODE_NODUMP;
218 	if (flags & FS_NOATIME_FL)
219 		ip->flags |= BTRFS_INODE_NOATIME;
220 	else
221 		ip->flags &= ~BTRFS_INODE_NOATIME;
222 	if (flags & FS_DIRSYNC_FL)
223 		ip->flags |= BTRFS_INODE_DIRSYNC;
224 	else
225 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
226 	if (flags & FS_NOCOW_FL)
227 		ip->flags |= BTRFS_INODE_NODATACOW;
228 	else
229 		ip->flags &= ~BTRFS_INODE_NODATACOW;
230 
231 	/*
232 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
233 	 * flag may be changed automatically if compression code won't make
234 	 * things smaller.
235 	 */
236 	if (flags & FS_NOCOMP_FL) {
237 		ip->flags &= ~BTRFS_INODE_COMPRESS;
238 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
239 	} else if (flags & FS_COMPR_FL) {
240 		ip->flags |= BTRFS_INODE_COMPRESS;
241 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
242 	} else {
243 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
244 	}
245 
246 	trans = btrfs_join_transaction(root);
247 	BUG_ON(IS_ERR(trans));
248 
249 	ret = btrfs_update_inode(trans, root, inode);
250 	BUG_ON(ret);
251 
252 	btrfs_update_iflags(inode);
253 	inode->i_ctime = CURRENT_TIME;
254 	btrfs_end_transaction(trans, root);
255 
256 	mnt_drop_write(file->f_path.mnt);
257 
258 	ret = 0;
259  out_unlock:
260 	mutex_unlock(&inode->i_mutex);
261 	return ret;
262 }
263 
264 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
265 {
266 	struct inode *inode = file->f_path.dentry->d_inode;
267 
268 	return put_user(inode->i_generation, arg);
269 }
270 
271 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
272 {
273 	struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
274 	struct btrfs_fs_info *fs_info = root->fs_info;
275 	struct btrfs_device *device;
276 	struct request_queue *q;
277 	struct fstrim_range range;
278 	u64 minlen = ULLONG_MAX;
279 	u64 num_devices = 0;
280 	int ret;
281 
282 	if (!capable(CAP_SYS_ADMIN))
283 		return -EPERM;
284 
285 	rcu_read_lock();
286 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
287 				dev_list) {
288 		if (!device->bdev)
289 			continue;
290 		q = bdev_get_queue(device->bdev);
291 		if (blk_queue_discard(q)) {
292 			num_devices++;
293 			minlen = min((u64)q->limits.discard_granularity,
294 				     minlen);
295 		}
296 	}
297 	rcu_read_unlock();
298 	if (!num_devices)
299 		return -EOPNOTSUPP;
300 
301 	if (copy_from_user(&range, arg, sizeof(range)))
302 		return -EFAULT;
303 
304 	range.minlen = max(range.minlen, minlen);
305 	ret = btrfs_trim_fs(root, &range);
306 	if (ret < 0)
307 		return ret;
308 
309 	if (copy_to_user(arg, &range, sizeof(range)))
310 		return -EFAULT;
311 
312 	return 0;
313 }
314 
315 static noinline int create_subvol(struct btrfs_root *root,
316 				  struct dentry *dentry,
317 				  char *name, int namelen,
318 				  u64 *async_transid)
319 {
320 	struct btrfs_trans_handle *trans;
321 	struct btrfs_key key;
322 	struct btrfs_root_item root_item;
323 	struct btrfs_inode_item *inode_item;
324 	struct extent_buffer *leaf;
325 	struct btrfs_root *new_root;
326 	struct dentry *parent = dentry->d_parent;
327 	struct inode *dir;
328 	int ret;
329 	int err;
330 	u64 objectid;
331 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
332 	u64 index = 0;
333 
334 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
335 	if (ret)
336 		return ret;
337 
338 	dir = parent->d_inode;
339 
340 	/*
341 	 * 1 - inode item
342 	 * 2 - refs
343 	 * 1 - root item
344 	 * 2 - dir items
345 	 */
346 	trans = btrfs_start_transaction(root, 6);
347 	if (IS_ERR(trans))
348 		return PTR_ERR(trans);
349 
350 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
351 				      0, objectid, NULL, 0, 0, 0);
352 	if (IS_ERR(leaf)) {
353 		ret = PTR_ERR(leaf);
354 		goto fail;
355 	}
356 
357 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
358 	btrfs_set_header_bytenr(leaf, leaf->start);
359 	btrfs_set_header_generation(leaf, trans->transid);
360 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
361 	btrfs_set_header_owner(leaf, objectid);
362 
363 	write_extent_buffer(leaf, root->fs_info->fsid,
364 			    (unsigned long)btrfs_header_fsid(leaf),
365 			    BTRFS_FSID_SIZE);
366 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
367 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
368 			    BTRFS_UUID_SIZE);
369 	btrfs_mark_buffer_dirty(leaf);
370 
371 	inode_item = &root_item.inode;
372 	memset(inode_item, 0, sizeof(*inode_item));
373 	inode_item->generation = cpu_to_le64(1);
374 	inode_item->size = cpu_to_le64(3);
375 	inode_item->nlink = cpu_to_le32(1);
376 	inode_item->nbytes = cpu_to_le64(root->leafsize);
377 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
378 
379 	root_item.flags = 0;
380 	root_item.byte_limit = 0;
381 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
382 
383 	btrfs_set_root_bytenr(&root_item, leaf->start);
384 	btrfs_set_root_generation(&root_item, trans->transid);
385 	btrfs_set_root_level(&root_item, 0);
386 	btrfs_set_root_refs(&root_item, 1);
387 	btrfs_set_root_used(&root_item, leaf->len);
388 	btrfs_set_root_last_snapshot(&root_item, 0);
389 
390 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
391 	root_item.drop_level = 0;
392 
393 	btrfs_tree_unlock(leaf);
394 	free_extent_buffer(leaf);
395 	leaf = NULL;
396 
397 	btrfs_set_root_dirid(&root_item, new_dirid);
398 
399 	key.objectid = objectid;
400 	key.offset = 0;
401 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
402 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
403 				&root_item);
404 	if (ret)
405 		goto fail;
406 
407 	key.offset = (u64)-1;
408 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
409 	BUG_ON(IS_ERR(new_root));
410 
411 	btrfs_record_root_in_trans(trans, new_root);
412 
413 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
414 	/*
415 	 * insert the directory item
416 	 */
417 	ret = btrfs_set_inode_index(dir, &index);
418 	BUG_ON(ret);
419 
420 	ret = btrfs_insert_dir_item(trans, root,
421 				    name, namelen, dir, &key,
422 				    BTRFS_FT_DIR, index);
423 	if (ret)
424 		goto fail;
425 
426 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
427 	ret = btrfs_update_inode(trans, root, dir);
428 	BUG_ON(ret);
429 
430 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
431 				 objectid, root->root_key.objectid,
432 				 btrfs_ino(dir), index, name, namelen);
433 
434 	BUG_ON(ret);
435 
436 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
437 fail:
438 	if (async_transid) {
439 		*async_transid = trans->transid;
440 		err = btrfs_commit_transaction_async(trans, root, 1);
441 	} else {
442 		err = btrfs_commit_transaction(trans, root);
443 	}
444 	if (err && !ret)
445 		ret = err;
446 	return ret;
447 }
448 
449 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
450 			   char *name, int namelen, u64 *async_transid,
451 			   bool readonly)
452 {
453 	struct inode *inode;
454 	struct btrfs_pending_snapshot *pending_snapshot;
455 	struct btrfs_trans_handle *trans;
456 	int ret;
457 
458 	if (!root->ref_cows)
459 		return -EINVAL;
460 
461 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
462 	if (!pending_snapshot)
463 		return -ENOMEM;
464 
465 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
466 	pending_snapshot->dentry = dentry;
467 	pending_snapshot->root = root;
468 	pending_snapshot->readonly = readonly;
469 
470 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
471 	if (IS_ERR(trans)) {
472 		ret = PTR_ERR(trans);
473 		goto fail;
474 	}
475 
476 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
477 	BUG_ON(ret);
478 
479 	spin_lock(&root->fs_info->trans_lock);
480 	list_add(&pending_snapshot->list,
481 		 &trans->transaction->pending_snapshots);
482 	spin_unlock(&root->fs_info->trans_lock);
483 	if (async_transid) {
484 		*async_transid = trans->transid;
485 		ret = btrfs_commit_transaction_async(trans,
486 				     root->fs_info->extent_root, 1);
487 	} else {
488 		ret = btrfs_commit_transaction(trans,
489 					       root->fs_info->extent_root);
490 	}
491 	BUG_ON(ret);
492 
493 	ret = pending_snapshot->error;
494 	if (ret)
495 		goto fail;
496 
497 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
498 	if (ret)
499 		goto fail;
500 
501 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
502 	if (IS_ERR(inode)) {
503 		ret = PTR_ERR(inode);
504 		goto fail;
505 	}
506 	BUG_ON(!inode);
507 	d_instantiate(dentry, inode);
508 	ret = 0;
509 fail:
510 	kfree(pending_snapshot);
511 	return ret;
512 }
513 
514 /*  copy of check_sticky in fs/namei.c()
515 * It's inline, so penalty for filesystems that don't use sticky bit is
516 * minimal.
517 */
518 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
519 {
520 	uid_t fsuid = current_fsuid();
521 
522 	if (!(dir->i_mode & S_ISVTX))
523 		return 0;
524 	if (inode->i_uid == fsuid)
525 		return 0;
526 	if (dir->i_uid == fsuid)
527 		return 0;
528 	return !capable(CAP_FOWNER);
529 }
530 
531 /*  copy of may_delete in fs/namei.c()
532  *	Check whether we can remove a link victim from directory dir, check
533  *  whether the type of victim is right.
534  *  1. We can't do it if dir is read-only (done in permission())
535  *  2. We should have write and exec permissions on dir
536  *  3. We can't remove anything from append-only dir
537  *  4. We can't do anything with immutable dir (done in permission())
538  *  5. If the sticky bit on dir is set we should either
539  *	a. be owner of dir, or
540  *	b. be owner of victim, or
541  *	c. have CAP_FOWNER capability
542  *  6. If the victim is append-only or immutable we can't do antyhing with
543  *     links pointing to it.
544  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
545  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
546  *  9. We can't remove a root or mountpoint.
547  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
548  *     nfs_async_unlink().
549  */
550 
551 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
552 {
553 	int error;
554 
555 	if (!victim->d_inode)
556 		return -ENOENT;
557 
558 	BUG_ON(victim->d_parent->d_inode != dir);
559 	audit_inode_child(victim, dir);
560 
561 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
562 	if (error)
563 		return error;
564 	if (IS_APPEND(dir))
565 		return -EPERM;
566 	if (btrfs_check_sticky(dir, victim->d_inode)||
567 		IS_APPEND(victim->d_inode)||
568 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
569 		return -EPERM;
570 	if (isdir) {
571 		if (!S_ISDIR(victim->d_inode->i_mode))
572 			return -ENOTDIR;
573 		if (IS_ROOT(victim))
574 			return -EBUSY;
575 	} else if (S_ISDIR(victim->d_inode->i_mode))
576 		return -EISDIR;
577 	if (IS_DEADDIR(dir))
578 		return -ENOENT;
579 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
580 		return -EBUSY;
581 	return 0;
582 }
583 
584 /* copy of may_create in fs/namei.c() */
585 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
586 {
587 	if (child->d_inode)
588 		return -EEXIST;
589 	if (IS_DEADDIR(dir))
590 		return -ENOENT;
591 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
592 }
593 
594 /*
595  * Create a new subvolume below @parent.  This is largely modeled after
596  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
597  * inside this filesystem so it's quite a bit simpler.
598  */
599 static noinline int btrfs_mksubvol(struct path *parent,
600 				   char *name, int namelen,
601 				   struct btrfs_root *snap_src,
602 				   u64 *async_transid, bool readonly)
603 {
604 	struct inode *dir  = parent->dentry->d_inode;
605 	struct dentry *dentry;
606 	int error;
607 
608 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
609 
610 	dentry = lookup_one_len(name, parent->dentry, namelen);
611 	error = PTR_ERR(dentry);
612 	if (IS_ERR(dentry))
613 		goto out_unlock;
614 
615 	error = -EEXIST;
616 	if (dentry->d_inode)
617 		goto out_dput;
618 
619 	error = mnt_want_write(parent->mnt);
620 	if (error)
621 		goto out_dput;
622 
623 	error = btrfs_may_create(dir, dentry);
624 	if (error)
625 		goto out_drop_write;
626 
627 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
628 
629 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
630 		goto out_up_read;
631 
632 	if (snap_src) {
633 		error = create_snapshot(snap_src, dentry,
634 					name, namelen, async_transid, readonly);
635 	} else {
636 		error = create_subvol(BTRFS_I(dir)->root, dentry,
637 				      name, namelen, async_transid);
638 	}
639 	if (!error)
640 		fsnotify_mkdir(dir, dentry);
641 out_up_read:
642 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
643 out_drop_write:
644 	mnt_drop_write(parent->mnt);
645 out_dput:
646 	dput(dentry);
647 out_unlock:
648 	mutex_unlock(&dir->i_mutex);
649 	return error;
650 }
651 
652 /*
653  * When we're defragging a range, we don't want to kick it off again
654  * if it is really just waiting for delalloc to send it down.
655  * If we find a nice big extent or delalloc range for the bytes in the
656  * file you want to defrag, we return 0 to let you know to skip this
657  * part of the file
658  */
659 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
660 {
661 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
662 	struct extent_map *em = NULL;
663 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
664 	u64 end;
665 
666 	read_lock(&em_tree->lock);
667 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
668 	read_unlock(&em_tree->lock);
669 
670 	if (em) {
671 		end = extent_map_end(em);
672 		free_extent_map(em);
673 		if (end - offset > thresh)
674 			return 0;
675 	}
676 	/* if we already have a nice delalloc here, just stop */
677 	thresh /= 2;
678 	end = count_range_bits(io_tree, &offset, offset + thresh,
679 			       thresh, EXTENT_DELALLOC, 1);
680 	if (end >= thresh)
681 		return 0;
682 	return 1;
683 }
684 
685 /*
686  * helper function to walk through a file and find extents
687  * newer than a specific transid, and smaller than thresh.
688  *
689  * This is used by the defragging code to find new and small
690  * extents
691  */
692 static int find_new_extents(struct btrfs_root *root,
693 			    struct inode *inode, u64 newer_than,
694 			    u64 *off, int thresh)
695 {
696 	struct btrfs_path *path;
697 	struct btrfs_key min_key;
698 	struct btrfs_key max_key;
699 	struct extent_buffer *leaf;
700 	struct btrfs_file_extent_item *extent;
701 	int type;
702 	int ret;
703 	u64 ino = btrfs_ino(inode);
704 
705 	path = btrfs_alloc_path();
706 	if (!path)
707 		return -ENOMEM;
708 
709 	min_key.objectid = ino;
710 	min_key.type = BTRFS_EXTENT_DATA_KEY;
711 	min_key.offset = *off;
712 
713 	max_key.objectid = ino;
714 	max_key.type = (u8)-1;
715 	max_key.offset = (u64)-1;
716 
717 	path->keep_locks = 1;
718 
719 	while(1) {
720 		ret = btrfs_search_forward(root, &min_key, &max_key,
721 					   path, 0, newer_than);
722 		if (ret != 0)
723 			goto none;
724 		if (min_key.objectid != ino)
725 			goto none;
726 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
727 			goto none;
728 
729 		leaf = path->nodes[0];
730 		extent = btrfs_item_ptr(leaf, path->slots[0],
731 					struct btrfs_file_extent_item);
732 
733 		type = btrfs_file_extent_type(leaf, extent);
734 		if (type == BTRFS_FILE_EXTENT_REG &&
735 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
736 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
737 			*off = min_key.offset;
738 			btrfs_free_path(path);
739 			return 0;
740 		}
741 
742 		if (min_key.offset == (u64)-1)
743 			goto none;
744 
745 		min_key.offset++;
746 		btrfs_release_path(path);
747 	}
748 none:
749 	btrfs_free_path(path);
750 	return -ENOENT;
751 }
752 
753 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
754 			       int thresh, u64 *last_len, u64 *skip,
755 			       u64 *defrag_end)
756 {
757 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
758 	struct extent_map *em = NULL;
759 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
760 	int ret = 1;
761 
762 	/*
763 	 * make sure that once we start defragging and extent, we keep on
764 	 * defragging it
765 	 */
766 	if (start < *defrag_end)
767 		return 1;
768 
769 	*skip = 0;
770 
771 	/*
772 	 * hopefully we have this extent in the tree already, try without
773 	 * the full extent lock
774 	 */
775 	read_lock(&em_tree->lock);
776 	em = lookup_extent_mapping(em_tree, start, len);
777 	read_unlock(&em_tree->lock);
778 
779 	if (!em) {
780 		/* get the big lock and read metadata off disk */
781 		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
782 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
783 		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
784 
785 		if (IS_ERR(em))
786 			return 0;
787 	}
788 
789 	/* this will cover holes, and inline extents */
790 	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
791 		ret = 0;
792 
793 	/*
794 	 * we hit a real extent, if it is big don't bother defragging it again
795 	 */
796 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
797 		ret = 0;
798 
799 	/*
800 	 * last_len ends up being a counter of how many bytes we've defragged.
801 	 * every time we choose not to defrag an extent, we reset *last_len
802 	 * so that the next tiny extent will force a defrag.
803 	 *
804 	 * The end result of this is that tiny extents before a single big
805 	 * extent will force at least part of that big extent to be defragged.
806 	 */
807 	if (ret) {
808 		*last_len += len;
809 		*defrag_end = extent_map_end(em);
810 	} else {
811 		*last_len = 0;
812 		*skip = extent_map_end(em);
813 		*defrag_end = 0;
814 	}
815 
816 	free_extent_map(em);
817 	return ret;
818 }
819 
820 /*
821  * it doesn't do much good to defrag one or two pages
822  * at a time.  This pulls in a nice chunk of pages
823  * to COW and defrag.
824  *
825  * It also makes sure the delalloc code has enough
826  * dirty data to avoid making new small extents as part
827  * of the defrag
828  *
829  * It's a good idea to start RA on this range
830  * before calling this.
831  */
832 static int cluster_pages_for_defrag(struct inode *inode,
833 				    struct page **pages,
834 				    unsigned long start_index,
835 				    int num_pages)
836 {
837 	unsigned long file_end;
838 	u64 isize = i_size_read(inode);
839 	u64 page_start;
840 	u64 page_end;
841 	int ret;
842 	int i;
843 	int i_done;
844 	struct btrfs_ordered_extent *ordered;
845 	struct extent_state *cached_state = NULL;
846 
847 	if (isize == 0)
848 		return 0;
849 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
850 
851 	ret = btrfs_delalloc_reserve_space(inode,
852 					   num_pages << PAGE_CACHE_SHIFT);
853 	if (ret)
854 		return ret;
855 again:
856 	ret = 0;
857 	i_done = 0;
858 
859 	/* step one, lock all the pages */
860 	for (i = 0; i < num_pages; i++) {
861 		struct page *page;
862 		page = grab_cache_page(inode->i_mapping,
863 					    start_index + i);
864 		if (!page)
865 			break;
866 
867 		if (!PageUptodate(page)) {
868 			btrfs_readpage(NULL, page);
869 			lock_page(page);
870 			if (!PageUptodate(page)) {
871 				unlock_page(page);
872 				page_cache_release(page);
873 				ret = -EIO;
874 				break;
875 			}
876 		}
877 		isize = i_size_read(inode);
878 		file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
879 		if (!isize || page->index > file_end ||
880 		    page->mapping != inode->i_mapping) {
881 			/* whoops, we blew past eof, skip this page */
882 			unlock_page(page);
883 			page_cache_release(page);
884 			break;
885 		}
886 		pages[i] = page;
887 		i_done++;
888 	}
889 	if (!i_done || ret)
890 		goto out;
891 
892 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
893 		goto out;
894 
895 	/*
896 	 * so now we have a nice long stream of locked
897 	 * and up to date pages, lets wait on them
898 	 */
899 	for (i = 0; i < i_done; i++)
900 		wait_on_page_writeback(pages[i]);
901 
902 	page_start = page_offset(pages[0]);
903 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
904 
905 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
906 			 page_start, page_end - 1, 0, &cached_state,
907 			 GFP_NOFS);
908 	ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
909 	if (ordered &&
910 	    ordered->file_offset + ordered->len > page_start &&
911 	    ordered->file_offset < page_end) {
912 		btrfs_put_ordered_extent(ordered);
913 		unlock_extent_cached(&BTRFS_I(inode)->io_tree,
914 				     page_start, page_end - 1,
915 				     &cached_state, GFP_NOFS);
916 		for (i = 0; i < i_done; i++) {
917 			unlock_page(pages[i]);
918 			page_cache_release(pages[i]);
919 		}
920 		btrfs_wait_ordered_range(inode, page_start,
921 					 page_end - page_start);
922 		goto again;
923 	}
924 	if (ordered)
925 		btrfs_put_ordered_extent(ordered);
926 
927 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
928 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
929 			  EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
930 			  GFP_NOFS);
931 
932 	if (i_done != num_pages) {
933 		atomic_inc(&BTRFS_I(inode)->outstanding_extents);
934 		btrfs_delalloc_release_space(inode,
935 				     (num_pages - i_done) << PAGE_CACHE_SHIFT);
936 	}
937 
938 
939 	btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
940 				  &cached_state);
941 
942 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
943 			     page_start, page_end - 1, &cached_state,
944 			     GFP_NOFS);
945 
946 	for (i = 0; i < i_done; i++) {
947 		clear_page_dirty_for_io(pages[i]);
948 		ClearPageChecked(pages[i]);
949 		set_page_extent_mapped(pages[i]);
950 		set_page_dirty(pages[i]);
951 		unlock_page(pages[i]);
952 		page_cache_release(pages[i]);
953 	}
954 	return i_done;
955 out:
956 	for (i = 0; i < i_done; i++) {
957 		unlock_page(pages[i]);
958 		page_cache_release(pages[i]);
959 	}
960 	btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
961 	return ret;
962 
963 }
964 
965 int btrfs_defrag_file(struct inode *inode, struct file *file,
966 		      struct btrfs_ioctl_defrag_range_args *range,
967 		      u64 newer_than, unsigned long max_to_defrag)
968 {
969 	struct btrfs_root *root = BTRFS_I(inode)->root;
970 	struct btrfs_super_block *disk_super;
971 	struct file_ra_state *ra = NULL;
972 	unsigned long last_index;
973 	u64 features;
974 	u64 last_len = 0;
975 	u64 skip = 0;
976 	u64 defrag_end = 0;
977 	u64 newer_off = range->start;
978 	int newer_left = 0;
979 	unsigned long i;
980 	int ret;
981 	int defrag_count = 0;
982 	int compress_type = BTRFS_COMPRESS_ZLIB;
983 	int extent_thresh = range->extent_thresh;
984 	int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
985 	u64 new_align = ~((u64)128 * 1024 - 1);
986 	struct page **pages = NULL;
987 
988 	if (extent_thresh == 0)
989 		extent_thresh = 256 * 1024;
990 
991 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
992 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
993 			return -EINVAL;
994 		if (range->compress_type)
995 			compress_type = range->compress_type;
996 	}
997 
998 	if (inode->i_size == 0)
999 		return 0;
1000 
1001 	/*
1002 	 * if we were not given a file, allocate a readahead
1003 	 * context
1004 	 */
1005 	if (!file) {
1006 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1007 		if (!ra)
1008 			return -ENOMEM;
1009 		file_ra_state_init(ra, inode->i_mapping);
1010 	} else {
1011 		ra = &file->f_ra;
1012 	}
1013 
1014 	pages = kmalloc(sizeof(struct page *) * newer_cluster,
1015 			GFP_NOFS);
1016 	if (!pages) {
1017 		ret = -ENOMEM;
1018 		goto out_ra;
1019 	}
1020 
1021 	/* find the last page to defrag */
1022 	if (range->start + range->len > range->start) {
1023 		last_index = min_t(u64, inode->i_size - 1,
1024 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1025 	} else {
1026 		last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
1027 	}
1028 
1029 	if (newer_than) {
1030 		ret = find_new_extents(root, inode, newer_than,
1031 				       &newer_off, 64 * 1024);
1032 		if (!ret) {
1033 			range->start = newer_off;
1034 			/*
1035 			 * we always align our defrag to help keep
1036 			 * the extents in the file evenly spaced
1037 			 */
1038 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1039 			newer_left = newer_cluster;
1040 		} else
1041 			goto out_ra;
1042 	} else {
1043 		i = range->start >> PAGE_CACHE_SHIFT;
1044 	}
1045 	if (!max_to_defrag)
1046 		max_to_defrag = last_index - 1;
1047 
1048 	while (i <= last_index && defrag_count < max_to_defrag) {
1049 		/*
1050 		 * make sure we stop running if someone unmounts
1051 		 * the FS
1052 		 */
1053 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1054 			break;
1055 
1056 		if (!newer_than &&
1057 		    !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1058 					PAGE_CACHE_SIZE,
1059 					extent_thresh,
1060 					&last_len, &skip,
1061 					&defrag_end)) {
1062 			unsigned long next;
1063 			/*
1064 			 * the should_defrag function tells us how much to skip
1065 			 * bump our counter by the suggested amount
1066 			 */
1067 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1068 			i = max(i + 1, next);
1069 			continue;
1070 		}
1071 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1072 			BTRFS_I(inode)->force_compress = compress_type;
1073 
1074 		btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
1075 
1076 		ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1077 		if (ret < 0)
1078 			goto out_ra;
1079 
1080 		defrag_count += ret;
1081 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1082 		i += ret;
1083 
1084 		if (newer_than) {
1085 			if (newer_off == (u64)-1)
1086 				break;
1087 
1088 			newer_off = max(newer_off + 1,
1089 					(u64)i << PAGE_CACHE_SHIFT);
1090 
1091 			ret = find_new_extents(root, inode,
1092 					       newer_than, &newer_off,
1093 					       64 * 1024);
1094 			if (!ret) {
1095 				range->start = newer_off;
1096 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1097 				newer_left = newer_cluster;
1098 			} else {
1099 				break;
1100 			}
1101 		} else {
1102 			i++;
1103 		}
1104 	}
1105 
1106 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1107 		filemap_flush(inode->i_mapping);
1108 
1109 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1110 		/* the filemap_flush will queue IO into the worker threads, but
1111 		 * we have to make sure the IO is actually started and that
1112 		 * ordered extents get created before we return
1113 		 */
1114 		atomic_inc(&root->fs_info->async_submit_draining);
1115 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1116 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1117 			wait_event(root->fs_info->async_submit_wait,
1118 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1119 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1120 		}
1121 		atomic_dec(&root->fs_info->async_submit_draining);
1122 
1123 		mutex_lock(&inode->i_mutex);
1124 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1125 		mutex_unlock(&inode->i_mutex);
1126 	}
1127 
1128 	disk_super = &root->fs_info->super_copy;
1129 	features = btrfs_super_incompat_flags(disk_super);
1130 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1131 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1132 		btrfs_set_super_incompat_flags(disk_super, features);
1133 	}
1134 
1135 	if (!file)
1136 		kfree(ra);
1137 	return defrag_count;
1138 
1139 out_ra:
1140 	if (!file)
1141 		kfree(ra);
1142 	kfree(pages);
1143 	return ret;
1144 }
1145 
1146 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1147 					void __user *arg)
1148 {
1149 	u64 new_size;
1150 	u64 old_size;
1151 	u64 devid = 1;
1152 	struct btrfs_ioctl_vol_args *vol_args;
1153 	struct btrfs_trans_handle *trans;
1154 	struct btrfs_device *device = NULL;
1155 	char *sizestr;
1156 	char *devstr = NULL;
1157 	int ret = 0;
1158 	int mod = 0;
1159 
1160 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1161 		return -EROFS;
1162 
1163 	if (!capable(CAP_SYS_ADMIN))
1164 		return -EPERM;
1165 
1166 	vol_args = memdup_user(arg, sizeof(*vol_args));
1167 	if (IS_ERR(vol_args))
1168 		return PTR_ERR(vol_args);
1169 
1170 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1171 
1172 	mutex_lock(&root->fs_info->volume_mutex);
1173 	sizestr = vol_args->name;
1174 	devstr = strchr(sizestr, ':');
1175 	if (devstr) {
1176 		char *end;
1177 		sizestr = devstr + 1;
1178 		*devstr = '\0';
1179 		devstr = vol_args->name;
1180 		devid = simple_strtoull(devstr, &end, 10);
1181 		printk(KERN_INFO "resizing devid %llu\n",
1182 		       (unsigned long long)devid);
1183 	}
1184 	device = btrfs_find_device(root, devid, NULL, NULL);
1185 	if (!device) {
1186 		printk(KERN_INFO "resizer unable to find device %llu\n",
1187 		       (unsigned long long)devid);
1188 		ret = -EINVAL;
1189 		goto out_unlock;
1190 	}
1191 	if (!strcmp(sizestr, "max"))
1192 		new_size = device->bdev->bd_inode->i_size;
1193 	else {
1194 		if (sizestr[0] == '-') {
1195 			mod = -1;
1196 			sizestr++;
1197 		} else if (sizestr[0] == '+') {
1198 			mod = 1;
1199 			sizestr++;
1200 		}
1201 		new_size = memparse(sizestr, NULL);
1202 		if (new_size == 0) {
1203 			ret = -EINVAL;
1204 			goto out_unlock;
1205 		}
1206 	}
1207 
1208 	old_size = device->total_bytes;
1209 
1210 	if (mod < 0) {
1211 		if (new_size > old_size) {
1212 			ret = -EINVAL;
1213 			goto out_unlock;
1214 		}
1215 		new_size = old_size - new_size;
1216 	} else if (mod > 0) {
1217 		new_size = old_size + new_size;
1218 	}
1219 
1220 	if (new_size < 256 * 1024 * 1024) {
1221 		ret = -EINVAL;
1222 		goto out_unlock;
1223 	}
1224 	if (new_size > device->bdev->bd_inode->i_size) {
1225 		ret = -EFBIG;
1226 		goto out_unlock;
1227 	}
1228 
1229 	do_div(new_size, root->sectorsize);
1230 	new_size *= root->sectorsize;
1231 
1232 	printk(KERN_INFO "new size for %s is %llu\n",
1233 		device->name, (unsigned long long)new_size);
1234 
1235 	if (new_size > old_size) {
1236 		trans = btrfs_start_transaction(root, 0);
1237 		if (IS_ERR(trans)) {
1238 			ret = PTR_ERR(trans);
1239 			goto out_unlock;
1240 		}
1241 		ret = btrfs_grow_device(trans, device, new_size);
1242 		btrfs_commit_transaction(trans, root);
1243 	} else {
1244 		ret = btrfs_shrink_device(device, new_size);
1245 	}
1246 
1247 out_unlock:
1248 	mutex_unlock(&root->fs_info->volume_mutex);
1249 	kfree(vol_args);
1250 	return ret;
1251 }
1252 
1253 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1254 						    char *name,
1255 						    unsigned long fd,
1256 						    int subvol,
1257 						    u64 *transid,
1258 						    bool readonly)
1259 {
1260 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1261 	struct file *src_file;
1262 	int namelen;
1263 	int ret = 0;
1264 
1265 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1266 		return -EROFS;
1267 
1268 	namelen = strlen(name);
1269 	if (strchr(name, '/')) {
1270 		ret = -EINVAL;
1271 		goto out;
1272 	}
1273 
1274 	if (subvol) {
1275 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1276 				     NULL, transid, readonly);
1277 	} else {
1278 		struct inode *src_inode;
1279 		src_file = fget(fd);
1280 		if (!src_file) {
1281 			ret = -EINVAL;
1282 			goto out;
1283 		}
1284 
1285 		src_inode = src_file->f_path.dentry->d_inode;
1286 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1287 			printk(KERN_INFO "btrfs: Snapshot src from "
1288 			       "another FS\n");
1289 			ret = -EINVAL;
1290 			fput(src_file);
1291 			goto out;
1292 		}
1293 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1294 				     BTRFS_I(src_inode)->root,
1295 				     transid, readonly);
1296 		fput(src_file);
1297 	}
1298 out:
1299 	return ret;
1300 }
1301 
1302 static noinline int btrfs_ioctl_snap_create(struct file *file,
1303 					    void __user *arg, int subvol)
1304 {
1305 	struct btrfs_ioctl_vol_args *vol_args;
1306 	int ret;
1307 
1308 	vol_args = memdup_user(arg, sizeof(*vol_args));
1309 	if (IS_ERR(vol_args))
1310 		return PTR_ERR(vol_args);
1311 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1312 
1313 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1314 					      vol_args->fd, subvol,
1315 					      NULL, false);
1316 
1317 	kfree(vol_args);
1318 	return ret;
1319 }
1320 
1321 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1322 					       void __user *arg, int subvol)
1323 {
1324 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1325 	int ret;
1326 	u64 transid = 0;
1327 	u64 *ptr = NULL;
1328 	bool readonly = false;
1329 
1330 	vol_args = memdup_user(arg, sizeof(*vol_args));
1331 	if (IS_ERR(vol_args))
1332 		return PTR_ERR(vol_args);
1333 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1334 
1335 	if (vol_args->flags &
1336 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1337 		ret = -EOPNOTSUPP;
1338 		goto out;
1339 	}
1340 
1341 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1342 		ptr = &transid;
1343 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1344 		readonly = true;
1345 
1346 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1347 					      vol_args->fd, subvol,
1348 					      ptr, readonly);
1349 
1350 	if (ret == 0 && ptr &&
1351 	    copy_to_user(arg +
1352 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1353 				  transid), ptr, sizeof(*ptr)))
1354 		ret = -EFAULT;
1355 out:
1356 	kfree(vol_args);
1357 	return ret;
1358 }
1359 
1360 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1361 						void __user *arg)
1362 {
1363 	struct inode *inode = fdentry(file)->d_inode;
1364 	struct btrfs_root *root = BTRFS_I(inode)->root;
1365 	int ret = 0;
1366 	u64 flags = 0;
1367 
1368 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1369 		return -EINVAL;
1370 
1371 	down_read(&root->fs_info->subvol_sem);
1372 	if (btrfs_root_readonly(root))
1373 		flags |= BTRFS_SUBVOL_RDONLY;
1374 	up_read(&root->fs_info->subvol_sem);
1375 
1376 	if (copy_to_user(arg, &flags, sizeof(flags)))
1377 		ret = -EFAULT;
1378 
1379 	return ret;
1380 }
1381 
1382 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1383 					      void __user *arg)
1384 {
1385 	struct inode *inode = fdentry(file)->d_inode;
1386 	struct btrfs_root *root = BTRFS_I(inode)->root;
1387 	struct btrfs_trans_handle *trans;
1388 	u64 root_flags;
1389 	u64 flags;
1390 	int ret = 0;
1391 
1392 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1393 		return -EROFS;
1394 
1395 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1396 		return -EINVAL;
1397 
1398 	if (copy_from_user(&flags, arg, sizeof(flags)))
1399 		return -EFAULT;
1400 
1401 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1402 		return -EINVAL;
1403 
1404 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1405 		return -EOPNOTSUPP;
1406 
1407 	if (!inode_owner_or_capable(inode))
1408 		return -EACCES;
1409 
1410 	down_write(&root->fs_info->subvol_sem);
1411 
1412 	/* nothing to do */
1413 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1414 		goto out;
1415 
1416 	root_flags = btrfs_root_flags(&root->root_item);
1417 	if (flags & BTRFS_SUBVOL_RDONLY)
1418 		btrfs_set_root_flags(&root->root_item,
1419 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1420 	else
1421 		btrfs_set_root_flags(&root->root_item,
1422 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1423 
1424 	trans = btrfs_start_transaction(root, 1);
1425 	if (IS_ERR(trans)) {
1426 		ret = PTR_ERR(trans);
1427 		goto out_reset;
1428 	}
1429 
1430 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1431 				&root->root_key, &root->root_item);
1432 
1433 	btrfs_commit_transaction(trans, root);
1434 out_reset:
1435 	if (ret)
1436 		btrfs_set_root_flags(&root->root_item, root_flags);
1437 out:
1438 	up_write(&root->fs_info->subvol_sem);
1439 	return ret;
1440 }
1441 
1442 /*
1443  * helper to check if the subvolume references other subvolumes
1444  */
1445 static noinline int may_destroy_subvol(struct btrfs_root *root)
1446 {
1447 	struct btrfs_path *path;
1448 	struct btrfs_key key;
1449 	int ret;
1450 
1451 	path = btrfs_alloc_path();
1452 	if (!path)
1453 		return -ENOMEM;
1454 
1455 	key.objectid = root->root_key.objectid;
1456 	key.type = BTRFS_ROOT_REF_KEY;
1457 	key.offset = (u64)-1;
1458 
1459 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1460 				&key, path, 0, 0);
1461 	if (ret < 0)
1462 		goto out;
1463 	BUG_ON(ret == 0);
1464 
1465 	ret = 0;
1466 	if (path->slots[0] > 0) {
1467 		path->slots[0]--;
1468 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1469 		if (key.objectid == root->root_key.objectid &&
1470 		    key.type == BTRFS_ROOT_REF_KEY)
1471 			ret = -ENOTEMPTY;
1472 	}
1473 out:
1474 	btrfs_free_path(path);
1475 	return ret;
1476 }
1477 
1478 static noinline int key_in_sk(struct btrfs_key *key,
1479 			      struct btrfs_ioctl_search_key *sk)
1480 {
1481 	struct btrfs_key test;
1482 	int ret;
1483 
1484 	test.objectid = sk->min_objectid;
1485 	test.type = sk->min_type;
1486 	test.offset = sk->min_offset;
1487 
1488 	ret = btrfs_comp_cpu_keys(key, &test);
1489 	if (ret < 0)
1490 		return 0;
1491 
1492 	test.objectid = sk->max_objectid;
1493 	test.type = sk->max_type;
1494 	test.offset = sk->max_offset;
1495 
1496 	ret = btrfs_comp_cpu_keys(key, &test);
1497 	if (ret > 0)
1498 		return 0;
1499 	return 1;
1500 }
1501 
1502 static noinline int copy_to_sk(struct btrfs_root *root,
1503 			       struct btrfs_path *path,
1504 			       struct btrfs_key *key,
1505 			       struct btrfs_ioctl_search_key *sk,
1506 			       char *buf,
1507 			       unsigned long *sk_offset,
1508 			       int *num_found)
1509 {
1510 	u64 found_transid;
1511 	struct extent_buffer *leaf;
1512 	struct btrfs_ioctl_search_header sh;
1513 	unsigned long item_off;
1514 	unsigned long item_len;
1515 	int nritems;
1516 	int i;
1517 	int slot;
1518 	int ret = 0;
1519 
1520 	leaf = path->nodes[0];
1521 	slot = path->slots[0];
1522 	nritems = btrfs_header_nritems(leaf);
1523 
1524 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1525 		i = nritems;
1526 		goto advance_key;
1527 	}
1528 	found_transid = btrfs_header_generation(leaf);
1529 
1530 	for (i = slot; i < nritems; i++) {
1531 		item_off = btrfs_item_ptr_offset(leaf, i);
1532 		item_len = btrfs_item_size_nr(leaf, i);
1533 
1534 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1535 			item_len = 0;
1536 
1537 		if (sizeof(sh) + item_len + *sk_offset >
1538 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1539 			ret = 1;
1540 			goto overflow;
1541 		}
1542 
1543 		btrfs_item_key_to_cpu(leaf, key, i);
1544 		if (!key_in_sk(key, sk))
1545 			continue;
1546 
1547 		sh.objectid = key->objectid;
1548 		sh.offset = key->offset;
1549 		sh.type = key->type;
1550 		sh.len = item_len;
1551 		sh.transid = found_transid;
1552 
1553 		/* copy search result header */
1554 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1555 		*sk_offset += sizeof(sh);
1556 
1557 		if (item_len) {
1558 			char *p = buf + *sk_offset;
1559 			/* copy the item */
1560 			read_extent_buffer(leaf, p,
1561 					   item_off, item_len);
1562 			*sk_offset += item_len;
1563 		}
1564 		(*num_found)++;
1565 
1566 		if (*num_found >= sk->nr_items)
1567 			break;
1568 	}
1569 advance_key:
1570 	ret = 0;
1571 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1572 		key->offset++;
1573 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1574 		key->offset = 0;
1575 		key->type++;
1576 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1577 		key->offset = 0;
1578 		key->type = 0;
1579 		key->objectid++;
1580 	} else
1581 		ret = 1;
1582 overflow:
1583 	return ret;
1584 }
1585 
1586 static noinline int search_ioctl(struct inode *inode,
1587 				 struct btrfs_ioctl_search_args *args)
1588 {
1589 	struct btrfs_root *root;
1590 	struct btrfs_key key;
1591 	struct btrfs_key max_key;
1592 	struct btrfs_path *path;
1593 	struct btrfs_ioctl_search_key *sk = &args->key;
1594 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1595 	int ret;
1596 	int num_found = 0;
1597 	unsigned long sk_offset = 0;
1598 
1599 	path = btrfs_alloc_path();
1600 	if (!path)
1601 		return -ENOMEM;
1602 
1603 	if (sk->tree_id == 0) {
1604 		/* search the root of the inode that was passed */
1605 		root = BTRFS_I(inode)->root;
1606 	} else {
1607 		key.objectid = sk->tree_id;
1608 		key.type = BTRFS_ROOT_ITEM_KEY;
1609 		key.offset = (u64)-1;
1610 		root = btrfs_read_fs_root_no_name(info, &key);
1611 		if (IS_ERR(root)) {
1612 			printk(KERN_ERR "could not find root %llu\n",
1613 			       sk->tree_id);
1614 			btrfs_free_path(path);
1615 			return -ENOENT;
1616 		}
1617 	}
1618 
1619 	key.objectid = sk->min_objectid;
1620 	key.type = sk->min_type;
1621 	key.offset = sk->min_offset;
1622 
1623 	max_key.objectid = sk->max_objectid;
1624 	max_key.type = sk->max_type;
1625 	max_key.offset = sk->max_offset;
1626 
1627 	path->keep_locks = 1;
1628 
1629 	while(1) {
1630 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1631 					   sk->min_transid);
1632 		if (ret != 0) {
1633 			if (ret > 0)
1634 				ret = 0;
1635 			goto err;
1636 		}
1637 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1638 				 &sk_offset, &num_found);
1639 		btrfs_release_path(path);
1640 		if (ret || num_found >= sk->nr_items)
1641 			break;
1642 
1643 	}
1644 	ret = 0;
1645 err:
1646 	sk->nr_items = num_found;
1647 	btrfs_free_path(path);
1648 	return ret;
1649 }
1650 
1651 static noinline int btrfs_ioctl_tree_search(struct file *file,
1652 					   void __user *argp)
1653 {
1654 	 struct btrfs_ioctl_search_args *args;
1655 	 struct inode *inode;
1656 	 int ret;
1657 
1658 	if (!capable(CAP_SYS_ADMIN))
1659 		return -EPERM;
1660 
1661 	args = memdup_user(argp, sizeof(*args));
1662 	if (IS_ERR(args))
1663 		return PTR_ERR(args);
1664 
1665 	inode = fdentry(file)->d_inode;
1666 	ret = search_ioctl(inode, args);
1667 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1668 		ret = -EFAULT;
1669 	kfree(args);
1670 	return ret;
1671 }
1672 
1673 /*
1674  * Search INODE_REFs to identify path name of 'dirid' directory
1675  * in a 'tree_id' tree. and sets path name to 'name'.
1676  */
1677 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1678 				u64 tree_id, u64 dirid, char *name)
1679 {
1680 	struct btrfs_root *root;
1681 	struct btrfs_key key;
1682 	char *ptr;
1683 	int ret = -1;
1684 	int slot;
1685 	int len;
1686 	int total_len = 0;
1687 	struct btrfs_inode_ref *iref;
1688 	struct extent_buffer *l;
1689 	struct btrfs_path *path;
1690 
1691 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1692 		name[0]='\0';
1693 		return 0;
1694 	}
1695 
1696 	path = btrfs_alloc_path();
1697 	if (!path)
1698 		return -ENOMEM;
1699 
1700 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1701 
1702 	key.objectid = tree_id;
1703 	key.type = BTRFS_ROOT_ITEM_KEY;
1704 	key.offset = (u64)-1;
1705 	root = btrfs_read_fs_root_no_name(info, &key);
1706 	if (IS_ERR(root)) {
1707 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1708 		ret = -ENOENT;
1709 		goto out;
1710 	}
1711 
1712 	key.objectid = dirid;
1713 	key.type = BTRFS_INODE_REF_KEY;
1714 	key.offset = (u64)-1;
1715 
1716 	while(1) {
1717 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1718 		if (ret < 0)
1719 			goto out;
1720 
1721 		l = path->nodes[0];
1722 		slot = path->slots[0];
1723 		if (ret > 0 && slot > 0)
1724 			slot--;
1725 		btrfs_item_key_to_cpu(l, &key, slot);
1726 
1727 		if (ret > 0 && (key.objectid != dirid ||
1728 				key.type != BTRFS_INODE_REF_KEY)) {
1729 			ret = -ENOENT;
1730 			goto out;
1731 		}
1732 
1733 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1734 		len = btrfs_inode_ref_name_len(l, iref);
1735 		ptr -= len + 1;
1736 		total_len += len + 1;
1737 		if (ptr < name)
1738 			goto out;
1739 
1740 		*(ptr + len) = '/';
1741 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1742 
1743 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1744 			break;
1745 
1746 		btrfs_release_path(path);
1747 		key.objectid = key.offset;
1748 		key.offset = (u64)-1;
1749 		dirid = key.objectid;
1750 
1751 	}
1752 	if (ptr < name)
1753 		goto out;
1754 	memcpy(name, ptr, total_len);
1755 	name[total_len]='\0';
1756 	ret = 0;
1757 out:
1758 	btrfs_free_path(path);
1759 	return ret;
1760 }
1761 
1762 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1763 					   void __user *argp)
1764 {
1765 	 struct btrfs_ioctl_ino_lookup_args *args;
1766 	 struct inode *inode;
1767 	 int ret;
1768 
1769 	if (!capable(CAP_SYS_ADMIN))
1770 		return -EPERM;
1771 
1772 	args = memdup_user(argp, sizeof(*args));
1773 	if (IS_ERR(args))
1774 		return PTR_ERR(args);
1775 
1776 	inode = fdentry(file)->d_inode;
1777 
1778 	if (args->treeid == 0)
1779 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1780 
1781 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1782 					args->treeid, args->objectid,
1783 					args->name);
1784 
1785 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1786 		ret = -EFAULT;
1787 
1788 	kfree(args);
1789 	return ret;
1790 }
1791 
1792 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1793 					     void __user *arg)
1794 {
1795 	struct dentry *parent = fdentry(file);
1796 	struct dentry *dentry;
1797 	struct inode *dir = parent->d_inode;
1798 	struct inode *inode;
1799 	struct btrfs_root *root = BTRFS_I(dir)->root;
1800 	struct btrfs_root *dest = NULL;
1801 	struct btrfs_ioctl_vol_args *vol_args;
1802 	struct btrfs_trans_handle *trans;
1803 	int namelen;
1804 	int ret;
1805 	int err = 0;
1806 
1807 	vol_args = memdup_user(arg, sizeof(*vol_args));
1808 	if (IS_ERR(vol_args))
1809 		return PTR_ERR(vol_args);
1810 
1811 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1812 	namelen = strlen(vol_args->name);
1813 	if (strchr(vol_args->name, '/') ||
1814 	    strncmp(vol_args->name, "..", namelen) == 0) {
1815 		err = -EINVAL;
1816 		goto out;
1817 	}
1818 
1819 	err = mnt_want_write(file->f_path.mnt);
1820 	if (err)
1821 		goto out;
1822 
1823 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1824 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1825 	if (IS_ERR(dentry)) {
1826 		err = PTR_ERR(dentry);
1827 		goto out_unlock_dir;
1828 	}
1829 
1830 	if (!dentry->d_inode) {
1831 		err = -ENOENT;
1832 		goto out_dput;
1833 	}
1834 
1835 	inode = dentry->d_inode;
1836 	dest = BTRFS_I(inode)->root;
1837 	if (!capable(CAP_SYS_ADMIN)){
1838 		/*
1839 		 * Regular user.  Only allow this with a special mount
1840 		 * option, when the user has write+exec access to the
1841 		 * subvol root, and when rmdir(2) would have been
1842 		 * allowed.
1843 		 *
1844 		 * Note that this is _not_ check that the subvol is
1845 		 * empty or doesn't contain data that we wouldn't
1846 		 * otherwise be able to delete.
1847 		 *
1848 		 * Users who want to delete empty subvols should try
1849 		 * rmdir(2).
1850 		 */
1851 		err = -EPERM;
1852 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1853 			goto out_dput;
1854 
1855 		/*
1856 		 * Do not allow deletion if the parent dir is the same
1857 		 * as the dir to be deleted.  That means the ioctl
1858 		 * must be called on the dentry referencing the root
1859 		 * of the subvol, not a random directory contained
1860 		 * within it.
1861 		 */
1862 		err = -EINVAL;
1863 		if (root == dest)
1864 			goto out_dput;
1865 
1866 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1867 		if (err)
1868 			goto out_dput;
1869 
1870 		/* check if subvolume may be deleted by a non-root user */
1871 		err = btrfs_may_delete(dir, dentry, 1);
1872 		if (err)
1873 			goto out_dput;
1874 	}
1875 
1876 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1877 		err = -EINVAL;
1878 		goto out_dput;
1879 	}
1880 
1881 	mutex_lock(&inode->i_mutex);
1882 	err = d_invalidate(dentry);
1883 	if (err)
1884 		goto out_unlock;
1885 
1886 	down_write(&root->fs_info->subvol_sem);
1887 
1888 	err = may_destroy_subvol(dest);
1889 	if (err)
1890 		goto out_up_write;
1891 
1892 	trans = btrfs_start_transaction(root, 0);
1893 	if (IS_ERR(trans)) {
1894 		err = PTR_ERR(trans);
1895 		goto out_up_write;
1896 	}
1897 	trans->block_rsv = &root->fs_info->global_block_rsv;
1898 
1899 	ret = btrfs_unlink_subvol(trans, root, dir,
1900 				dest->root_key.objectid,
1901 				dentry->d_name.name,
1902 				dentry->d_name.len);
1903 	BUG_ON(ret);
1904 
1905 	btrfs_record_root_in_trans(trans, dest);
1906 
1907 	memset(&dest->root_item.drop_progress, 0,
1908 		sizeof(dest->root_item.drop_progress));
1909 	dest->root_item.drop_level = 0;
1910 	btrfs_set_root_refs(&dest->root_item, 0);
1911 
1912 	if (!xchg(&dest->orphan_item_inserted, 1)) {
1913 		ret = btrfs_insert_orphan_item(trans,
1914 					root->fs_info->tree_root,
1915 					dest->root_key.objectid);
1916 		BUG_ON(ret);
1917 	}
1918 
1919 	ret = btrfs_end_transaction(trans, root);
1920 	BUG_ON(ret);
1921 	inode->i_flags |= S_DEAD;
1922 out_up_write:
1923 	up_write(&root->fs_info->subvol_sem);
1924 out_unlock:
1925 	mutex_unlock(&inode->i_mutex);
1926 	if (!err) {
1927 		shrink_dcache_sb(root->fs_info->sb);
1928 		btrfs_invalidate_inodes(dest);
1929 		d_delete(dentry);
1930 	}
1931 out_dput:
1932 	dput(dentry);
1933 out_unlock_dir:
1934 	mutex_unlock(&dir->i_mutex);
1935 	mnt_drop_write(file->f_path.mnt);
1936 out:
1937 	kfree(vol_args);
1938 	return err;
1939 }
1940 
1941 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1942 {
1943 	struct inode *inode = fdentry(file)->d_inode;
1944 	struct btrfs_root *root = BTRFS_I(inode)->root;
1945 	struct btrfs_ioctl_defrag_range_args *range;
1946 	int ret;
1947 
1948 	if (btrfs_root_readonly(root))
1949 		return -EROFS;
1950 
1951 	ret = mnt_want_write(file->f_path.mnt);
1952 	if (ret)
1953 		return ret;
1954 
1955 	switch (inode->i_mode & S_IFMT) {
1956 	case S_IFDIR:
1957 		if (!capable(CAP_SYS_ADMIN)) {
1958 			ret = -EPERM;
1959 			goto out;
1960 		}
1961 		ret = btrfs_defrag_root(root, 0);
1962 		if (ret)
1963 			goto out;
1964 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1965 		break;
1966 	case S_IFREG:
1967 		if (!(file->f_mode & FMODE_WRITE)) {
1968 			ret = -EINVAL;
1969 			goto out;
1970 		}
1971 
1972 		range = kzalloc(sizeof(*range), GFP_KERNEL);
1973 		if (!range) {
1974 			ret = -ENOMEM;
1975 			goto out;
1976 		}
1977 
1978 		if (argp) {
1979 			if (copy_from_user(range, argp,
1980 					   sizeof(*range))) {
1981 				ret = -EFAULT;
1982 				kfree(range);
1983 				goto out;
1984 			}
1985 			/* compression requires us to start the IO */
1986 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1987 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1988 				range->extent_thresh = (u32)-1;
1989 			}
1990 		} else {
1991 			/* the rest are all set to zero by kzalloc */
1992 			range->len = (u64)-1;
1993 		}
1994 		ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
1995 					range, 0, 0);
1996 		if (ret > 0)
1997 			ret = 0;
1998 		kfree(range);
1999 		break;
2000 	default:
2001 		ret = -EINVAL;
2002 	}
2003 out:
2004 	mnt_drop_write(file->f_path.mnt);
2005 	return ret;
2006 }
2007 
2008 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2009 {
2010 	struct btrfs_ioctl_vol_args *vol_args;
2011 	int ret;
2012 
2013 	if (!capable(CAP_SYS_ADMIN))
2014 		return -EPERM;
2015 
2016 	vol_args = memdup_user(arg, sizeof(*vol_args));
2017 	if (IS_ERR(vol_args))
2018 		return PTR_ERR(vol_args);
2019 
2020 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2021 	ret = btrfs_init_new_device(root, vol_args->name);
2022 
2023 	kfree(vol_args);
2024 	return ret;
2025 }
2026 
2027 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2028 {
2029 	struct btrfs_ioctl_vol_args *vol_args;
2030 	int ret;
2031 
2032 	if (!capable(CAP_SYS_ADMIN))
2033 		return -EPERM;
2034 
2035 	if (root->fs_info->sb->s_flags & MS_RDONLY)
2036 		return -EROFS;
2037 
2038 	vol_args = memdup_user(arg, sizeof(*vol_args));
2039 	if (IS_ERR(vol_args))
2040 		return PTR_ERR(vol_args);
2041 
2042 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2043 	ret = btrfs_rm_device(root, vol_args->name);
2044 
2045 	kfree(vol_args);
2046 	return ret;
2047 }
2048 
2049 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2050 {
2051 	struct btrfs_ioctl_fs_info_args *fi_args;
2052 	struct btrfs_device *device;
2053 	struct btrfs_device *next;
2054 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2055 	int ret = 0;
2056 
2057 	if (!capable(CAP_SYS_ADMIN))
2058 		return -EPERM;
2059 
2060 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2061 	if (!fi_args)
2062 		return -ENOMEM;
2063 
2064 	fi_args->num_devices = fs_devices->num_devices;
2065 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2066 
2067 	mutex_lock(&fs_devices->device_list_mutex);
2068 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2069 		if (device->devid > fi_args->max_id)
2070 			fi_args->max_id = device->devid;
2071 	}
2072 	mutex_unlock(&fs_devices->device_list_mutex);
2073 
2074 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2075 		ret = -EFAULT;
2076 
2077 	kfree(fi_args);
2078 	return ret;
2079 }
2080 
2081 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2082 {
2083 	struct btrfs_ioctl_dev_info_args *di_args;
2084 	struct btrfs_device *dev;
2085 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2086 	int ret = 0;
2087 	char *s_uuid = NULL;
2088 	char empty_uuid[BTRFS_UUID_SIZE] = {0};
2089 
2090 	if (!capable(CAP_SYS_ADMIN))
2091 		return -EPERM;
2092 
2093 	di_args = memdup_user(arg, sizeof(*di_args));
2094 	if (IS_ERR(di_args))
2095 		return PTR_ERR(di_args);
2096 
2097 	if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2098 		s_uuid = di_args->uuid;
2099 
2100 	mutex_lock(&fs_devices->device_list_mutex);
2101 	dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2102 	mutex_unlock(&fs_devices->device_list_mutex);
2103 
2104 	if (!dev) {
2105 		ret = -ENODEV;
2106 		goto out;
2107 	}
2108 
2109 	di_args->devid = dev->devid;
2110 	di_args->bytes_used = dev->bytes_used;
2111 	di_args->total_bytes = dev->total_bytes;
2112 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2113 	strncpy(di_args->path, dev->name, sizeof(di_args->path));
2114 
2115 out:
2116 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2117 		ret = -EFAULT;
2118 
2119 	kfree(di_args);
2120 	return ret;
2121 }
2122 
2123 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2124 				       u64 off, u64 olen, u64 destoff)
2125 {
2126 	struct inode *inode = fdentry(file)->d_inode;
2127 	struct btrfs_root *root = BTRFS_I(inode)->root;
2128 	struct file *src_file;
2129 	struct inode *src;
2130 	struct btrfs_trans_handle *trans;
2131 	struct btrfs_path *path;
2132 	struct extent_buffer *leaf;
2133 	char *buf;
2134 	struct btrfs_key key;
2135 	u32 nritems;
2136 	int slot;
2137 	int ret;
2138 	u64 len = olen;
2139 	u64 bs = root->fs_info->sb->s_blocksize;
2140 	u64 hint_byte;
2141 
2142 	/*
2143 	 * TODO:
2144 	 * - split compressed inline extents.  annoying: we need to
2145 	 *   decompress into destination's address_space (the file offset
2146 	 *   may change, so source mapping won't do), then recompress (or
2147 	 *   otherwise reinsert) a subrange.
2148 	 * - allow ranges within the same file to be cloned (provided
2149 	 *   they don't overlap)?
2150 	 */
2151 
2152 	/* the destination must be opened for writing */
2153 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2154 		return -EINVAL;
2155 
2156 	if (btrfs_root_readonly(root))
2157 		return -EROFS;
2158 
2159 	ret = mnt_want_write(file->f_path.mnt);
2160 	if (ret)
2161 		return ret;
2162 
2163 	src_file = fget(srcfd);
2164 	if (!src_file) {
2165 		ret = -EBADF;
2166 		goto out_drop_write;
2167 	}
2168 
2169 	src = src_file->f_dentry->d_inode;
2170 
2171 	ret = -EINVAL;
2172 	if (src == inode)
2173 		goto out_fput;
2174 
2175 	/* the src must be open for reading */
2176 	if (!(src_file->f_mode & FMODE_READ))
2177 		goto out_fput;
2178 
2179 	ret = -EISDIR;
2180 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2181 		goto out_fput;
2182 
2183 	ret = -EXDEV;
2184 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2185 		goto out_fput;
2186 
2187 	ret = -ENOMEM;
2188 	buf = vmalloc(btrfs_level_size(root, 0));
2189 	if (!buf)
2190 		goto out_fput;
2191 
2192 	path = btrfs_alloc_path();
2193 	if (!path) {
2194 		vfree(buf);
2195 		goto out_fput;
2196 	}
2197 	path->reada = 2;
2198 
2199 	if (inode < src) {
2200 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2201 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2202 	} else {
2203 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2204 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2205 	}
2206 
2207 	/* determine range to clone */
2208 	ret = -EINVAL;
2209 	if (off + len > src->i_size || off + len < off)
2210 		goto out_unlock;
2211 	if (len == 0)
2212 		olen = len = src->i_size - off;
2213 	/* if we extend to eof, continue to block boundary */
2214 	if (off + len == src->i_size)
2215 		len = ALIGN(src->i_size, bs) - off;
2216 
2217 	/* verify the end result is block aligned */
2218 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2219 	    !IS_ALIGNED(destoff, bs))
2220 		goto out_unlock;
2221 
2222 	/* do any pending delalloc/csum calc on src, one way or
2223 	   another, and lock file content */
2224 	while (1) {
2225 		struct btrfs_ordered_extent *ordered;
2226 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2227 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2228 		if (!ordered &&
2229 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2230 				   EXTENT_DELALLOC, 0, NULL))
2231 			break;
2232 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2233 		if (ordered)
2234 			btrfs_put_ordered_extent(ordered);
2235 		btrfs_wait_ordered_range(src, off, len);
2236 	}
2237 
2238 	/* clone data */
2239 	key.objectid = btrfs_ino(src);
2240 	key.type = BTRFS_EXTENT_DATA_KEY;
2241 	key.offset = 0;
2242 
2243 	while (1) {
2244 		/*
2245 		 * note the key will change type as we walk through the
2246 		 * tree.
2247 		 */
2248 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2249 		if (ret < 0)
2250 			goto out;
2251 
2252 		nritems = btrfs_header_nritems(path->nodes[0]);
2253 		if (path->slots[0] >= nritems) {
2254 			ret = btrfs_next_leaf(root, path);
2255 			if (ret < 0)
2256 				goto out;
2257 			if (ret > 0)
2258 				break;
2259 			nritems = btrfs_header_nritems(path->nodes[0]);
2260 		}
2261 		leaf = path->nodes[0];
2262 		slot = path->slots[0];
2263 
2264 		btrfs_item_key_to_cpu(leaf, &key, slot);
2265 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2266 		    key.objectid != btrfs_ino(src))
2267 			break;
2268 
2269 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2270 			struct btrfs_file_extent_item *extent;
2271 			int type;
2272 			u32 size;
2273 			struct btrfs_key new_key;
2274 			u64 disko = 0, diskl = 0;
2275 			u64 datao = 0, datal = 0;
2276 			u8 comp;
2277 			u64 endoff;
2278 
2279 			size = btrfs_item_size_nr(leaf, slot);
2280 			read_extent_buffer(leaf, buf,
2281 					   btrfs_item_ptr_offset(leaf, slot),
2282 					   size);
2283 
2284 			extent = btrfs_item_ptr(leaf, slot,
2285 						struct btrfs_file_extent_item);
2286 			comp = btrfs_file_extent_compression(leaf, extent);
2287 			type = btrfs_file_extent_type(leaf, extent);
2288 			if (type == BTRFS_FILE_EXTENT_REG ||
2289 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2290 				disko = btrfs_file_extent_disk_bytenr(leaf,
2291 								      extent);
2292 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
2293 								 extent);
2294 				datao = btrfs_file_extent_offset(leaf, extent);
2295 				datal = btrfs_file_extent_num_bytes(leaf,
2296 								    extent);
2297 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2298 				/* take upper bound, may be compressed */
2299 				datal = btrfs_file_extent_ram_bytes(leaf,
2300 								    extent);
2301 			}
2302 			btrfs_release_path(path);
2303 
2304 			if (key.offset + datal <= off ||
2305 			    key.offset >= off+len)
2306 				goto next;
2307 
2308 			memcpy(&new_key, &key, sizeof(new_key));
2309 			new_key.objectid = btrfs_ino(inode);
2310 			if (off <= key.offset)
2311 				new_key.offset = key.offset + destoff - off;
2312 			else
2313 				new_key.offset = destoff;
2314 
2315 			trans = btrfs_start_transaction(root, 1);
2316 			if (IS_ERR(trans)) {
2317 				ret = PTR_ERR(trans);
2318 				goto out;
2319 			}
2320 
2321 			if (type == BTRFS_FILE_EXTENT_REG ||
2322 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2323 				if (off > key.offset) {
2324 					datao += off - key.offset;
2325 					datal -= off - key.offset;
2326 				}
2327 
2328 				if (key.offset + datal > off + len)
2329 					datal = off + len - key.offset;
2330 
2331 				ret = btrfs_drop_extents(trans, inode,
2332 							 new_key.offset,
2333 							 new_key.offset + datal,
2334 							 &hint_byte, 1);
2335 				BUG_ON(ret);
2336 
2337 				ret = btrfs_insert_empty_item(trans, root, path,
2338 							      &new_key, size);
2339 				BUG_ON(ret);
2340 
2341 				leaf = path->nodes[0];
2342 				slot = path->slots[0];
2343 				write_extent_buffer(leaf, buf,
2344 					    btrfs_item_ptr_offset(leaf, slot),
2345 					    size);
2346 
2347 				extent = btrfs_item_ptr(leaf, slot,
2348 						struct btrfs_file_extent_item);
2349 
2350 				/* disko == 0 means it's a hole */
2351 				if (!disko)
2352 					datao = 0;
2353 
2354 				btrfs_set_file_extent_offset(leaf, extent,
2355 							     datao);
2356 				btrfs_set_file_extent_num_bytes(leaf, extent,
2357 								datal);
2358 				if (disko) {
2359 					inode_add_bytes(inode, datal);
2360 					ret = btrfs_inc_extent_ref(trans, root,
2361 							disko, diskl, 0,
2362 							root->root_key.objectid,
2363 							btrfs_ino(inode),
2364 							new_key.offset - datao);
2365 					BUG_ON(ret);
2366 				}
2367 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2368 				u64 skip = 0;
2369 				u64 trim = 0;
2370 				if (off > key.offset) {
2371 					skip = off - key.offset;
2372 					new_key.offset += skip;
2373 				}
2374 
2375 				if (key.offset + datal > off+len)
2376 					trim = key.offset + datal - (off+len);
2377 
2378 				if (comp && (skip || trim)) {
2379 					ret = -EINVAL;
2380 					btrfs_end_transaction(trans, root);
2381 					goto out;
2382 				}
2383 				size -= skip + trim;
2384 				datal -= skip + trim;
2385 
2386 				ret = btrfs_drop_extents(trans, inode,
2387 							 new_key.offset,
2388 							 new_key.offset + datal,
2389 							 &hint_byte, 1);
2390 				BUG_ON(ret);
2391 
2392 				ret = btrfs_insert_empty_item(trans, root, path,
2393 							      &new_key, size);
2394 				BUG_ON(ret);
2395 
2396 				if (skip) {
2397 					u32 start =
2398 					  btrfs_file_extent_calc_inline_size(0);
2399 					memmove(buf+start, buf+start+skip,
2400 						datal);
2401 				}
2402 
2403 				leaf = path->nodes[0];
2404 				slot = path->slots[0];
2405 				write_extent_buffer(leaf, buf,
2406 					    btrfs_item_ptr_offset(leaf, slot),
2407 					    size);
2408 				inode_add_bytes(inode, datal);
2409 			}
2410 
2411 			btrfs_mark_buffer_dirty(leaf);
2412 			btrfs_release_path(path);
2413 
2414 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2415 
2416 			/*
2417 			 * we round up to the block size at eof when
2418 			 * determining which extents to clone above,
2419 			 * but shouldn't round up the file size
2420 			 */
2421 			endoff = new_key.offset + datal;
2422 			if (endoff > destoff+olen)
2423 				endoff = destoff+olen;
2424 			if (endoff > inode->i_size)
2425 				btrfs_i_size_write(inode, endoff);
2426 
2427 			BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
2428 			ret = btrfs_update_inode(trans, root, inode);
2429 			BUG_ON(ret);
2430 			btrfs_end_transaction(trans, root);
2431 		}
2432 next:
2433 		btrfs_release_path(path);
2434 		key.offset++;
2435 	}
2436 	ret = 0;
2437 out:
2438 	btrfs_release_path(path);
2439 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2440 out_unlock:
2441 	mutex_unlock(&src->i_mutex);
2442 	mutex_unlock(&inode->i_mutex);
2443 	vfree(buf);
2444 	btrfs_free_path(path);
2445 out_fput:
2446 	fput(src_file);
2447 out_drop_write:
2448 	mnt_drop_write(file->f_path.mnt);
2449 	return ret;
2450 }
2451 
2452 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2453 {
2454 	struct btrfs_ioctl_clone_range_args args;
2455 
2456 	if (copy_from_user(&args, argp, sizeof(args)))
2457 		return -EFAULT;
2458 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2459 				 args.src_length, args.dest_offset);
2460 }
2461 
2462 /*
2463  * there are many ways the trans_start and trans_end ioctls can lead
2464  * to deadlocks.  They should only be used by applications that
2465  * basically own the machine, and have a very in depth understanding
2466  * of all the possible deadlocks and enospc problems.
2467  */
2468 static long btrfs_ioctl_trans_start(struct file *file)
2469 {
2470 	struct inode *inode = fdentry(file)->d_inode;
2471 	struct btrfs_root *root = BTRFS_I(inode)->root;
2472 	struct btrfs_trans_handle *trans;
2473 	int ret;
2474 
2475 	ret = -EPERM;
2476 	if (!capable(CAP_SYS_ADMIN))
2477 		goto out;
2478 
2479 	ret = -EINPROGRESS;
2480 	if (file->private_data)
2481 		goto out;
2482 
2483 	ret = -EROFS;
2484 	if (btrfs_root_readonly(root))
2485 		goto out;
2486 
2487 	ret = mnt_want_write(file->f_path.mnt);
2488 	if (ret)
2489 		goto out;
2490 
2491 	atomic_inc(&root->fs_info->open_ioctl_trans);
2492 
2493 	ret = -ENOMEM;
2494 	trans = btrfs_start_ioctl_transaction(root);
2495 	if (IS_ERR(trans))
2496 		goto out_drop;
2497 
2498 	file->private_data = trans;
2499 	return 0;
2500 
2501 out_drop:
2502 	atomic_dec(&root->fs_info->open_ioctl_trans);
2503 	mnt_drop_write(file->f_path.mnt);
2504 out:
2505 	return ret;
2506 }
2507 
2508 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2509 {
2510 	struct inode *inode = fdentry(file)->d_inode;
2511 	struct btrfs_root *root = BTRFS_I(inode)->root;
2512 	struct btrfs_root *new_root;
2513 	struct btrfs_dir_item *di;
2514 	struct btrfs_trans_handle *trans;
2515 	struct btrfs_path *path;
2516 	struct btrfs_key location;
2517 	struct btrfs_disk_key disk_key;
2518 	struct btrfs_super_block *disk_super;
2519 	u64 features;
2520 	u64 objectid = 0;
2521 	u64 dir_id;
2522 
2523 	if (!capable(CAP_SYS_ADMIN))
2524 		return -EPERM;
2525 
2526 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2527 		return -EFAULT;
2528 
2529 	if (!objectid)
2530 		objectid = root->root_key.objectid;
2531 
2532 	location.objectid = objectid;
2533 	location.type = BTRFS_ROOT_ITEM_KEY;
2534 	location.offset = (u64)-1;
2535 
2536 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2537 	if (IS_ERR(new_root))
2538 		return PTR_ERR(new_root);
2539 
2540 	if (btrfs_root_refs(&new_root->root_item) == 0)
2541 		return -ENOENT;
2542 
2543 	path = btrfs_alloc_path();
2544 	if (!path)
2545 		return -ENOMEM;
2546 	path->leave_spinning = 1;
2547 
2548 	trans = btrfs_start_transaction(root, 1);
2549 	if (IS_ERR(trans)) {
2550 		btrfs_free_path(path);
2551 		return PTR_ERR(trans);
2552 	}
2553 
2554 	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2555 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2556 				   dir_id, "default", 7, 1);
2557 	if (IS_ERR_OR_NULL(di)) {
2558 		btrfs_free_path(path);
2559 		btrfs_end_transaction(trans, root);
2560 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2561 		       "this isn't going to work\n");
2562 		return -ENOENT;
2563 	}
2564 
2565 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2566 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2567 	btrfs_mark_buffer_dirty(path->nodes[0]);
2568 	btrfs_free_path(path);
2569 
2570 	disk_super = &root->fs_info->super_copy;
2571 	features = btrfs_super_incompat_flags(disk_super);
2572 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2573 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2574 		btrfs_set_super_incompat_flags(disk_super, features);
2575 	}
2576 	btrfs_end_transaction(trans, root);
2577 
2578 	return 0;
2579 }
2580 
2581 static void get_block_group_info(struct list_head *groups_list,
2582 				 struct btrfs_ioctl_space_info *space)
2583 {
2584 	struct btrfs_block_group_cache *block_group;
2585 
2586 	space->total_bytes = 0;
2587 	space->used_bytes = 0;
2588 	space->flags = 0;
2589 	list_for_each_entry(block_group, groups_list, list) {
2590 		space->flags = block_group->flags;
2591 		space->total_bytes += block_group->key.offset;
2592 		space->used_bytes +=
2593 			btrfs_block_group_used(&block_group->item);
2594 	}
2595 }
2596 
2597 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2598 {
2599 	struct btrfs_ioctl_space_args space_args;
2600 	struct btrfs_ioctl_space_info space;
2601 	struct btrfs_ioctl_space_info *dest;
2602 	struct btrfs_ioctl_space_info *dest_orig;
2603 	struct btrfs_ioctl_space_info __user *user_dest;
2604 	struct btrfs_space_info *info;
2605 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2606 		       BTRFS_BLOCK_GROUP_SYSTEM,
2607 		       BTRFS_BLOCK_GROUP_METADATA,
2608 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2609 	int num_types = 4;
2610 	int alloc_size;
2611 	int ret = 0;
2612 	u64 slot_count = 0;
2613 	int i, c;
2614 
2615 	if (copy_from_user(&space_args,
2616 			   (struct btrfs_ioctl_space_args __user *)arg,
2617 			   sizeof(space_args)))
2618 		return -EFAULT;
2619 
2620 	for (i = 0; i < num_types; i++) {
2621 		struct btrfs_space_info *tmp;
2622 
2623 		info = NULL;
2624 		rcu_read_lock();
2625 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2626 					list) {
2627 			if (tmp->flags == types[i]) {
2628 				info = tmp;
2629 				break;
2630 			}
2631 		}
2632 		rcu_read_unlock();
2633 
2634 		if (!info)
2635 			continue;
2636 
2637 		down_read(&info->groups_sem);
2638 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2639 			if (!list_empty(&info->block_groups[c]))
2640 				slot_count++;
2641 		}
2642 		up_read(&info->groups_sem);
2643 	}
2644 
2645 	/* space_slots == 0 means they are asking for a count */
2646 	if (space_args.space_slots == 0) {
2647 		space_args.total_spaces = slot_count;
2648 		goto out;
2649 	}
2650 
2651 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2652 
2653 	alloc_size = sizeof(*dest) * slot_count;
2654 
2655 	/* we generally have at most 6 or so space infos, one for each raid
2656 	 * level.  So, a whole page should be more than enough for everyone
2657 	 */
2658 	if (alloc_size > PAGE_CACHE_SIZE)
2659 		return -ENOMEM;
2660 
2661 	space_args.total_spaces = 0;
2662 	dest = kmalloc(alloc_size, GFP_NOFS);
2663 	if (!dest)
2664 		return -ENOMEM;
2665 	dest_orig = dest;
2666 
2667 	/* now we have a buffer to copy into */
2668 	for (i = 0; i < num_types; i++) {
2669 		struct btrfs_space_info *tmp;
2670 
2671 		if (!slot_count)
2672 			break;
2673 
2674 		info = NULL;
2675 		rcu_read_lock();
2676 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2677 					list) {
2678 			if (tmp->flags == types[i]) {
2679 				info = tmp;
2680 				break;
2681 			}
2682 		}
2683 		rcu_read_unlock();
2684 
2685 		if (!info)
2686 			continue;
2687 		down_read(&info->groups_sem);
2688 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2689 			if (!list_empty(&info->block_groups[c])) {
2690 				get_block_group_info(&info->block_groups[c],
2691 						     &space);
2692 				memcpy(dest, &space, sizeof(space));
2693 				dest++;
2694 				space_args.total_spaces++;
2695 				slot_count--;
2696 			}
2697 			if (!slot_count)
2698 				break;
2699 		}
2700 		up_read(&info->groups_sem);
2701 	}
2702 
2703 	user_dest = (struct btrfs_ioctl_space_info *)
2704 		(arg + sizeof(struct btrfs_ioctl_space_args));
2705 
2706 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2707 		ret = -EFAULT;
2708 
2709 	kfree(dest_orig);
2710 out:
2711 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2712 		ret = -EFAULT;
2713 
2714 	return ret;
2715 }
2716 
2717 /*
2718  * there are many ways the trans_start and trans_end ioctls can lead
2719  * to deadlocks.  They should only be used by applications that
2720  * basically own the machine, and have a very in depth understanding
2721  * of all the possible deadlocks and enospc problems.
2722  */
2723 long btrfs_ioctl_trans_end(struct file *file)
2724 {
2725 	struct inode *inode = fdentry(file)->d_inode;
2726 	struct btrfs_root *root = BTRFS_I(inode)->root;
2727 	struct btrfs_trans_handle *trans;
2728 
2729 	trans = file->private_data;
2730 	if (!trans)
2731 		return -EINVAL;
2732 	file->private_data = NULL;
2733 
2734 	btrfs_end_transaction(trans, root);
2735 
2736 	atomic_dec(&root->fs_info->open_ioctl_trans);
2737 
2738 	mnt_drop_write(file->f_path.mnt);
2739 	return 0;
2740 }
2741 
2742 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2743 {
2744 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2745 	struct btrfs_trans_handle *trans;
2746 	u64 transid;
2747 	int ret;
2748 
2749 	trans = btrfs_start_transaction(root, 0);
2750 	if (IS_ERR(trans))
2751 		return PTR_ERR(trans);
2752 	transid = trans->transid;
2753 	ret = btrfs_commit_transaction_async(trans, root, 0);
2754 	if (ret) {
2755 		btrfs_end_transaction(trans, root);
2756 		return ret;
2757 	}
2758 
2759 	if (argp)
2760 		if (copy_to_user(argp, &transid, sizeof(transid)))
2761 			return -EFAULT;
2762 	return 0;
2763 }
2764 
2765 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2766 {
2767 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2768 	u64 transid;
2769 
2770 	if (argp) {
2771 		if (copy_from_user(&transid, argp, sizeof(transid)))
2772 			return -EFAULT;
2773 	} else {
2774 		transid = 0;  /* current trans */
2775 	}
2776 	return btrfs_wait_for_commit(root, transid);
2777 }
2778 
2779 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2780 {
2781 	int ret;
2782 	struct btrfs_ioctl_scrub_args *sa;
2783 
2784 	if (!capable(CAP_SYS_ADMIN))
2785 		return -EPERM;
2786 
2787 	sa = memdup_user(arg, sizeof(*sa));
2788 	if (IS_ERR(sa))
2789 		return PTR_ERR(sa);
2790 
2791 	ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2792 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2793 
2794 	if (copy_to_user(arg, sa, sizeof(*sa)))
2795 		ret = -EFAULT;
2796 
2797 	kfree(sa);
2798 	return ret;
2799 }
2800 
2801 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2802 {
2803 	if (!capable(CAP_SYS_ADMIN))
2804 		return -EPERM;
2805 
2806 	return btrfs_scrub_cancel(root);
2807 }
2808 
2809 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2810 				       void __user *arg)
2811 {
2812 	struct btrfs_ioctl_scrub_args *sa;
2813 	int ret;
2814 
2815 	if (!capable(CAP_SYS_ADMIN))
2816 		return -EPERM;
2817 
2818 	sa = memdup_user(arg, sizeof(*sa));
2819 	if (IS_ERR(sa))
2820 		return PTR_ERR(sa);
2821 
2822 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2823 
2824 	if (copy_to_user(arg, sa, sizeof(*sa)))
2825 		ret = -EFAULT;
2826 
2827 	kfree(sa);
2828 	return ret;
2829 }
2830 
2831 long btrfs_ioctl(struct file *file, unsigned int
2832 		cmd, unsigned long arg)
2833 {
2834 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2835 	void __user *argp = (void __user *)arg;
2836 
2837 	switch (cmd) {
2838 	case FS_IOC_GETFLAGS:
2839 		return btrfs_ioctl_getflags(file, argp);
2840 	case FS_IOC_SETFLAGS:
2841 		return btrfs_ioctl_setflags(file, argp);
2842 	case FS_IOC_GETVERSION:
2843 		return btrfs_ioctl_getversion(file, argp);
2844 	case FITRIM:
2845 		return btrfs_ioctl_fitrim(file, argp);
2846 	case BTRFS_IOC_SNAP_CREATE:
2847 		return btrfs_ioctl_snap_create(file, argp, 0);
2848 	case BTRFS_IOC_SNAP_CREATE_V2:
2849 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
2850 	case BTRFS_IOC_SUBVOL_CREATE:
2851 		return btrfs_ioctl_snap_create(file, argp, 1);
2852 	case BTRFS_IOC_SNAP_DESTROY:
2853 		return btrfs_ioctl_snap_destroy(file, argp);
2854 	case BTRFS_IOC_SUBVOL_GETFLAGS:
2855 		return btrfs_ioctl_subvol_getflags(file, argp);
2856 	case BTRFS_IOC_SUBVOL_SETFLAGS:
2857 		return btrfs_ioctl_subvol_setflags(file, argp);
2858 	case BTRFS_IOC_DEFAULT_SUBVOL:
2859 		return btrfs_ioctl_default_subvol(file, argp);
2860 	case BTRFS_IOC_DEFRAG:
2861 		return btrfs_ioctl_defrag(file, NULL);
2862 	case BTRFS_IOC_DEFRAG_RANGE:
2863 		return btrfs_ioctl_defrag(file, argp);
2864 	case BTRFS_IOC_RESIZE:
2865 		return btrfs_ioctl_resize(root, argp);
2866 	case BTRFS_IOC_ADD_DEV:
2867 		return btrfs_ioctl_add_dev(root, argp);
2868 	case BTRFS_IOC_RM_DEV:
2869 		return btrfs_ioctl_rm_dev(root, argp);
2870 	case BTRFS_IOC_FS_INFO:
2871 		return btrfs_ioctl_fs_info(root, argp);
2872 	case BTRFS_IOC_DEV_INFO:
2873 		return btrfs_ioctl_dev_info(root, argp);
2874 	case BTRFS_IOC_BALANCE:
2875 		return btrfs_balance(root->fs_info->dev_root);
2876 	case BTRFS_IOC_CLONE:
2877 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2878 	case BTRFS_IOC_CLONE_RANGE:
2879 		return btrfs_ioctl_clone_range(file, argp);
2880 	case BTRFS_IOC_TRANS_START:
2881 		return btrfs_ioctl_trans_start(file);
2882 	case BTRFS_IOC_TRANS_END:
2883 		return btrfs_ioctl_trans_end(file);
2884 	case BTRFS_IOC_TREE_SEARCH:
2885 		return btrfs_ioctl_tree_search(file, argp);
2886 	case BTRFS_IOC_INO_LOOKUP:
2887 		return btrfs_ioctl_ino_lookup(file, argp);
2888 	case BTRFS_IOC_SPACE_INFO:
2889 		return btrfs_ioctl_space_info(root, argp);
2890 	case BTRFS_IOC_SYNC:
2891 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
2892 		return 0;
2893 	case BTRFS_IOC_START_SYNC:
2894 		return btrfs_ioctl_start_sync(file, argp);
2895 	case BTRFS_IOC_WAIT_SYNC:
2896 		return btrfs_ioctl_wait_sync(file, argp);
2897 	case BTRFS_IOC_SCRUB:
2898 		return btrfs_ioctl_scrub(root, argp);
2899 	case BTRFS_IOC_SCRUB_CANCEL:
2900 		return btrfs_ioctl_scrub_cancel(root, argp);
2901 	case BTRFS_IOC_SCRUB_PROGRESS:
2902 		return btrfs_ioctl_scrub_progress(root, argp);
2903 	}
2904 
2905 	return -ENOTTY;
2906 }
2907