xref: /linux/fs/btrfs/ioctl.c (revision 492c826b9facefa84995f4dea917e301b5ee0884)
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 
54 /* Mask out flags that are inappropriate for the given type of inode. */
55 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
56 {
57 	if (S_ISDIR(mode))
58 		return flags;
59 	else if (S_ISREG(mode))
60 		return flags & ~FS_DIRSYNC_FL;
61 	else
62 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
63 }
64 
65 /*
66  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
67  */
68 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
69 {
70 	unsigned int iflags = 0;
71 
72 	if (flags & BTRFS_INODE_SYNC)
73 		iflags |= FS_SYNC_FL;
74 	if (flags & BTRFS_INODE_IMMUTABLE)
75 		iflags |= FS_IMMUTABLE_FL;
76 	if (flags & BTRFS_INODE_APPEND)
77 		iflags |= FS_APPEND_FL;
78 	if (flags & BTRFS_INODE_NODUMP)
79 		iflags |= FS_NODUMP_FL;
80 	if (flags & BTRFS_INODE_NOATIME)
81 		iflags |= FS_NOATIME_FL;
82 	if (flags & BTRFS_INODE_DIRSYNC)
83 		iflags |= FS_DIRSYNC_FL;
84 	if (flags & BTRFS_INODE_NODATACOW)
85 		iflags |= FS_NOCOW_FL;
86 
87 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
88 		iflags |= FS_COMPR_FL;
89 	else if (flags & BTRFS_INODE_NOCOMPRESS)
90 		iflags |= FS_NOCOMP_FL;
91 
92 	return iflags;
93 }
94 
95 /*
96  * Update inode->i_flags based on the btrfs internal flags.
97  */
98 void btrfs_update_iflags(struct inode *inode)
99 {
100 	struct btrfs_inode *ip = BTRFS_I(inode);
101 
102 	inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
103 
104 	if (ip->flags & BTRFS_INODE_SYNC)
105 		inode->i_flags |= S_SYNC;
106 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
107 		inode->i_flags |= S_IMMUTABLE;
108 	if (ip->flags & BTRFS_INODE_APPEND)
109 		inode->i_flags |= S_APPEND;
110 	if (ip->flags & BTRFS_INODE_NOATIME)
111 		inode->i_flags |= S_NOATIME;
112 	if (ip->flags & BTRFS_INODE_DIRSYNC)
113 		inode->i_flags |= S_DIRSYNC;
114 }
115 
116 /*
117  * Inherit flags from the parent inode.
118  *
119  * Unlike extN we don't have any flags we don't want to inherit currently.
120  */
121 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
122 {
123 	unsigned int flags;
124 
125 	if (!dir)
126 		return;
127 
128 	flags = BTRFS_I(dir)->flags;
129 
130 	if (S_ISREG(inode->i_mode))
131 		flags &= ~BTRFS_INODE_DIRSYNC;
132 	else if (!S_ISDIR(inode->i_mode))
133 		flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
134 
135 	BTRFS_I(inode)->flags = flags;
136 	btrfs_update_iflags(inode);
137 }
138 
139 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
140 {
141 	struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
142 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
143 
144 	if (copy_to_user(arg, &flags, sizeof(flags)))
145 		return -EFAULT;
146 	return 0;
147 }
148 
149 static int check_flags(unsigned int flags)
150 {
151 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
152 		      FS_NOATIME_FL | FS_NODUMP_FL | \
153 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
154 		      FS_NOCOMP_FL | FS_COMPR_FL |
155 		      FS_NOCOW_FL))
156 		return -EOPNOTSUPP;
157 
158 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
159 		return -EINVAL;
160 
161 	return 0;
162 }
163 
164 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
165 {
166 	struct inode *inode = file->f_path.dentry->d_inode;
167 	struct btrfs_inode *ip = BTRFS_I(inode);
168 	struct btrfs_root *root = ip->root;
169 	struct btrfs_trans_handle *trans;
170 	unsigned int flags, oldflags;
171 	int ret;
172 
173 	if (btrfs_root_readonly(root))
174 		return -EROFS;
175 
176 	if (copy_from_user(&flags, arg, sizeof(flags)))
177 		return -EFAULT;
178 
179 	ret = check_flags(flags);
180 	if (ret)
181 		return ret;
182 
183 	if (!inode_owner_or_capable(inode))
184 		return -EACCES;
185 
186 	mutex_lock(&inode->i_mutex);
187 
188 	flags = btrfs_mask_flags(inode->i_mode, flags);
189 	oldflags = btrfs_flags_to_ioctl(ip->flags);
190 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
191 		if (!capable(CAP_LINUX_IMMUTABLE)) {
192 			ret = -EPERM;
193 			goto out_unlock;
194 		}
195 	}
196 
197 	ret = mnt_want_write(file->f_path.mnt);
198 	if (ret)
199 		goto out_unlock;
200 
201 	if (flags & FS_SYNC_FL)
202 		ip->flags |= BTRFS_INODE_SYNC;
203 	else
204 		ip->flags &= ~BTRFS_INODE_SYNC;
205 	if (flags & FS_IMMUTABLE_FL)
206 		ip->flags |= BTRFS_INODE_IMMUTABLE;
207 	else
208 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
209 	if (flags & FS_APPEND_FL)
210 		ip->flags |= BTRFS_INODE_APPEND;
211 	else
212 		ip->flags &= ~BTRFS_INODE_APPEND;
213 	if (flags & FS_NODUMP_FL)
214 		ip->flags |= BTRFS_INODE_NODUMP;
215 	else
216 		ip->flags &= ~BTRFS_INODE_NODUMP;
217 	if (flags & FS_NOATIME_FL)
218 		ip->flags |= BTRFS_INODE_NOATIME;
219 	else
220 		ip->flags &= ~BTRFS_INODE_NOATIME;
221 	if (flags & FS_DIRSYNC_FL)
222 		ip->flags |= BTRFS_INODE_DIRSYNC;
223 	else
224 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
225 	if (flags & FS_NOCOW_FL)
226 		ip->flags |= BTRFS_INODE_NODATACOW;
227 	else
228 		ip->flags &= ~BTRFS_INODE_NODATACOW;
229 
230 	/*
231 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
232 	 * flag may be changed automatically if compression code won't make
233 	 * things smaller.
234 	 */
235 	if (flags & FS_NOCOMP_FL) {
236 		ip->flags &= ~BTRFS_INODE_COMPRESS;
237 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
238 	} else if (flags & FS_COMPR_FL) {
239 		ip->flags |= BTRFS_INODE_COMPRESS;
240 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
241 	} else {
242 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
243 	}
244 
245 	trans = btrfs_join_transaction(root, 1);
246 	BUG_ON(IS_ERR(trans));
247 
248 	ret = btrfs_update_inode(trans, root, inode);
249 	BUG_ON(ret);
250 
251 	btrfs_update_iflags(inode);
252 	inode->i_ctime = CURRENT_TIME;
253 	btrfs_end_transaction(trans, root);
254 
255 	mnt_drop_write(file->f_path.mnt);
256 
257 	ret = 0;
258  out_unlock:
259 	mutex_unlock(&inode->i_mutex);
260 	return ret;
261 }
262 
263 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
264 {
265 	struct inode *inode = file->f_path.dentry->d_inode;
266 
267 	return put_user(inode->i_generation, arg);
268 }
269 
270 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
271 {
272 	struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
273 	struct btrfs_fs_info *fs_info = root->fs_info;
274 	struct btrfs_device *device;
275 	struct request_queue *q;
276 	struct fstrim_range range;
277 	u64 minlen = ULLONG_MAX;
278 	u64 num_devices = 0;
279 	int ret;
280 
281 	if (!capable(CAP_SYS_ADMIN))
282 		return -EPERM;
283 
284 	mutex_lock(&fs_info->fs_devices->device_list_mutex);
285 	list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
286 		if (!device->bdev)
287 			continue;
288 		q = bdev_get_queue(device->bdev);
289 		if (blk_queue_discard(q)) {
290 			num_devices++;
291 			minlen = min((u64)q->limits.discard_granularity,
292 				     minlen);
293 		}
294 	}
295 	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
296 	if (!num_devices)
297 		return -EOPNOTSUPP;
298 
299 	if (copy_from_user(&range, arg, sizeof(range)))
300 		return -EFAULT;
301 
302 	range.minlen = max(range.minlen, minlen);
303 	ret = btrfs_trim_fs(root, &range);
304 	if (ret < 0)
305 		return ret;
306 
307 	if (copy_to_user(arg, &range, sizeof(range)))
308 		return -EFAULT;
309 
310 	return 0;
311 }
312 
313 static noinline int create_subvol(struct btrfs_root *root,
314 				  struct dentry *dentry,
315 				  char *name, int namelen,
316 				  u64 *async_transid)
317 {
318 	struct btrfs_trans_handle *trans;
319 	struct btrfs_key key;
320 	struct btrfs_root_item root_item;
321 	struct btrfs_inode_item *inode_item;
322 	struct extent_buffer *leaf;
323 	struct btrfs_root *new_root;
324 	struct dentry *parent = dget_parent(dentry);
325 	struct inode *dir;
326 	int ret;
327 	int err;
328 	u64 objectid;
329 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
330 	u64 index = 0;
331 
332 	ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root,
333 				       0, &objectid);
334 	if (ret) {
335 		dput(parent);
336 		return ret;
337 	}
338 
339 	dir = parent->d_inode;
340 
341 	/*
342 	 * 1 - inode item
343 	 * 2 - refs
344 	 * 1 - root item
345 	 * 2 - dir items
346 	 */
347 	trans = btrfs_start_transaction(root, 6);
348 	if (IS_ERR(trans)) {
349 		dput(parent);
350 		return PTR_ERR(trans);
351 	}
352 
353 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
354 				      0, objectid, NULL, 0, 0, 0);
355 	if (IS_ERR(leaf)) {
356 		ret = PTR_ERR(leaf);
357 		goto fail;
358 	}
359 
360 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
361 	btrfs_set_header_bytenr(leaf, leaf->start);
362 	btrfs_set_header_generation(leaf, trans->transid);
363 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
364 	btrfs_set_header_owner(leaf, objectid);
365 
366 	write_extent_buffer(leaf, root->fs_info->fsid,
367 			    (unsigned long)btrfs_header_fsid(leaf),
368 			    BTRFS_FSID_SIZE);
369 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
370 			    (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
371 			    BTRFS_UUID_SIZE);
372 	btrfs_mark_buffer_dirty(leaf);
373 
374 	inode_item = &root_item.inode;
375 	memset(inode_item, 0, sizeof(*inode_item));
376 	inode_item->generation = cpu_to_le64(1);
377 	inode_item->size = cpu_to_le64(3);
378 	inode_item->nlink = cpu_to_le32(1);
379 	inode_item->nbytes = cpu_to_le64(root->leafsize);
380 	inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
381 
382 	root_item.flags = 0;
383 	root_item.byte_limit = 0;
384 	inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
385 
386 	btrfs_set_root_bytenr(&root_item, leaf->start);
387 	btrfs_set_root_generation(&root_item, trans->transid);
388 	btrfs_set_root_level(&root_item, 0);
389 	btrfs_set_root_refs(&root_item, 1);
390 	btrfs_set_root_used(&root_item, leaf->len);
391 	btrfs_set_root_last_snapshot(&root_item, 0);
392 
393 	memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
394 	root_item.drop_level = 0;
395 
396 	btrfs_tree_unlock(leaf);
397 	free_extent_buffer(leaf);
398 	leaf = NULL;
399 
400 	btrfs_set_root_dirid(&root_item, new_dirid);
401 
402 	key.objectid = objectid;
403 	key.offset = 0;
404 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
405 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
406 				&root_item);
407 	if (ret)
408 		goto fail;
409 
410 	key.offset = (u64)-1;
411 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
412 	BUG_ON(IS_ERR(new_root));
413 
414 	btrfs_record_root_in_trans(trans, new_root);
415 
416 	ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
417 				       BTRFS_I(dir)->block_group);
418 	/*
419 	 * insert the directory item
420 	 */
421 	ret = btrfs_set_inode_index(dir, &index);
422 	BUG_ON(ret);
423 
424 	ret = btrfs_insert_dir_item(trans, root,
425 				    name, namelen, dir->i_ino, &key,
426 				    BTRFS_FT_DIR, index);
427 	if (ret)
428 		goto fail;
429 
430 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
431 	ret = btrfs_update_inode(trans, root, dir);
432 	BUG_ON(ret);
433 
434 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
435 				 objectid, root->root_key.objectid,
436 				 dir->i_ino, index, name, namelen);
437 
438 	BUG_ON(ret);
439 
440 	d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
441 fail:
442 	dput(parent);
443 	if (async_transid) {
444 		*async_transid = trans->transid;
445 		err = btrfs_commit_transaction_async(trans, root, 1);
446 	} else {
447 		err = btrfs_commit_transaction(trans, root);
448 	}
449 	if (err && !ret)
450 		ret = err;
451 	return ret;
452 }
453 
454 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
455 			   char *name, int namelen, u64 *async_transid,
456 			   bool readonly)
457 {
458 	struct inode *inode;
459 	struct dentry *parent;
460 	struct btrfs_pending_snapshot *pending_snapshot;
461 	struct btrfs_trans_handle *trans;
462 	int ret;
463 
464 	if (!root->ref_cows)
465 		return -EINVAL;
466 
467 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
468 	if (!pending_snapshot)
469 		return -ENOMEM;
470 
471 	btrfs_init_block_rsv(&pending_snapshot->block_rsv);
472 	pending_snapshot->dentry = dentry;
473 	pending_snapshot->root = root;
474 	pending_snapshot->readonly = readonly;
475 
476 	trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
477 	if (IS_ERR(trans)) {
478 		ret = PTR_ERR(trans);
479 		goto fail;
480 	}
481 
482 	ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
483 	BUG_ON(ret);
484 
485 	list_add(&pending_snapshot->list,
486 		 &trans->transaction->pending_snapshots);
487 	if (async_transid) {
488 		*async_transid = trans->transid;
489 		ret = btrfs_commit_transaction_async(trans,
490 				     root->fs_info->extent_root, 1);
491 	} else {
492 		ret = btrfs_commit_transaction(trans,
493 					       root->fs_info->extent_root);
494 	}
495 	BUG_ON(ret);
496 
497 	ret = pending_snapshot->error;
498 	if (ret)
499 		goto fail;
500 
501 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
502 	if (ret)
503 		goto fail;
504 
505 	parent = dget_parent(dentry);
506 	inode = btrfs_lookup_dentry(parent->d_inode, dentry);
507 	dput(parent);
508 	if (IS_ERR(inode)) {
509 		ret = PTR_ERR(inode);
510 		goto fail;
511 	}
512 	BUG_ON(!inode);
513 	d_instantiate(dentry, inode);
514 	ret = 0;
515 fail:
516 	kfree(pending_snapshot);
517 	return ret;
518 }
519 
520 /*  copy of check_sticky in fs/namei.c()
521 * It's inline, so penalty for filesystems that don't use sticky bit is
522 * minimal.
523 */
524 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
525 {
526 	uid_t fsuid = current_fsuid();
527 
528 	if (!(dir->i_mode & S_ISVTX))
529 		return 0;
530 	if (inode->i_uid == fsuid)
531 		return 0;
532 	if (dir->i_uid == fsuid)
533 		return 0;
534 	return !capable(CAP_FOWNER);
535 }
536 
537 /*  copy of may_delete in fs/namei.c()
538  *	Check whether we can remove a link victim from directory dir, check
539  *  whether the type of victim is right.
540  *  1. We can't do it if dir is read-only (done in permission())
541  *  2. We should have write and exec permissions on dir
542  *  3. We can't remove anything from append-only dir
543  *  4. We can't do anything with immutable dir (done in permission())
544  *  5. If the sticky bit on dir is set we should either
545  *	a. be owner of dir, or
546  *	b. be owner of victim, or
547  *	c. have CAP_FOWNER capability
548  *  6. If the victim is append-only or immutable we can't do antyhing with
549  *     links pointing to it.
550  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
551  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
552  *  9. We can't remove a root or mountpoint.
553  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
554  *     nfs_async_unlink().
555  */
556 
557 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
558 {
559 	int error;
560 
561 	if (!victim->d_inode)
562 		return -ENOENT;
563 
564 	BUG_ON(victim->d_parent->d_inode != dir);
565 	audit_inode_child(victim, dir);
566 
567 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
568 	if (error)
569 		return error;
570 	if (IS_APPEND(dir))
571 		return -EPERM;
572 	if (btrfs_check_sticky(dir, victim->d_inode)||
573 		IS_APPEND(victim->d_inode)||
574 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
575 		return -EPERM;
576 	if (isdir) {
577 		if (!S_ISDIR(victim->d_inode->i_mode))
578 			return -ENOTDIR;
579 		if (IS_ROOT(victim))
580 			return -EBUSY;
581 	} else if (S_ISDIR(victim->d_inode->i_mode))
582 		return -EISDIR;
583 	if (IS_DEADDIR(dir))
584 		return -ENOENT;
585 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
586 		return -EBUSY;
587 	return 0;
588 }
589 
590 /* copy of may_create in fs/namei.c() */
591 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
592 {
593 	if (child->d_inode)
594 		return -EEXIST;
595 	if (IS_DEADDIR(dir))
596 		return -ENOENT;
597 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
598 }
599 
600 /*
601  * Create a new subvolume below @parent.  This is largely modeled after
602  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
603  * inside this filesystem so it's quite a bit simpler.
604  */
605 static noinline int btrfs_mksubvol(struct path *parent,
606 				   char *name, int namelen,
607 				   struct btrfs_root *snap_src,
608 				   u64 *async_transid, bool readonly)
609 {
610 	struct inode *dir  = parent->dentry->d_inode;
611 	struct dentry *dentry;
612 	int error;
613 
614 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
615 
616 	dentry = lookup_one_len(name, parent->dentry, namelen);
617 	error = PTR_ERR(dentry);
618 	if (IS_ERR(dentry))
619 		goto out_unlock;
620 
621 	error = -EEXIST;
622 	if (dentry->d_inode)
623 		goto out_dput;
624 
625 	error = mnt_want_write(parent->mnt);
626 	if (error)
627 		goto out_dput;
628 
629 	error = btrfs_may_create(dir, dentry);
630 	if (error)
631 		goto out_drop_write;
632 
633 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
634 
635 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
636 		goto out_up_read;
637 
638 	if (snap_src) {
639 		error = create_snapshot(snap_src, dentry,
640 					name, namelen, async_transid, readonly);
641 	} else {
642 		error = create_subvol(BTRFS_I(dir)->root, dentry,
643 				      name, namelen, async_transid);
644 	}
645 	if (!error)
646 		fsnotify_mkdir(dir, dentry);
647 out_up_read:
648 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
649 out_drop_write:
650 	mnt_drop_write(parent->mnt);
651 out_dput:
652 	dput(dentry);
653 out_unlock:
654 	mutex_unlock(&dir->i_mutex);
655 	return error;
656 }
657 
658 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
659 			       int thresh, u64 *last_len, u64 *skip,
660 			       u64 *defrag_end)
661 {
662 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
663 	struct extent_map *em = NULL;
664 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
665 	int ret = 1;
666 
667 
668 	if (thresh == 0)
669 		thresh = 256 * 1024;
670 
671 	/*
672 	 * make sure that once we start defragging and extent, we keep on
673 	 * defragging it
674 	 */
675 	if (start < *defrag_end)
676 		return 1;
677 
678 	*skip = 0;
679 
680 	/*
681 	 * hopefully we have this extent in the tree already, try without
682 	 * the full extent lock
683 	 */
684 	read_lock(&em_tree->lock);
685 	em = lookup_extent_mapping(em_tree, start, len);
686 	read_unlock(&em_tree->lock);
687 
688 	if (!em) {
689 		/* get the big lock and read metadata off disk */
690 		lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
691 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
692 		unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
693 
694 		if (IS_ERR(em))
695 			return 0;
696 	}
697 
698 	/* this will cover holes, and inline extents */
699 	if (em->block_start >= EXTENT_MAP_LAST_BYTE)
700 		ret = 0;
701 
702 	/*
703 	 * we hit a real extent, if it is big don't bother defragging it again
704 	 */
705 	if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
706 		ret = 0;
707 
708 	/*
709 	 * last_len ends up being a counter of how many bytes we've defragged.
710 	 * every time we choose not to defrag an extent, we reset *last_len
711 	 * so that the next tiny extent will force a defrag.
712 	 *
713 	 * The end result of this is that tiny extents before a single big
714 	 * extent will force at least part of that big extent to be defragged.
715 	 */
716 	if (ret) {
717 		*last_len += len;
718 		*defrag_end = extent_map_end(em);
719 	} else {
720 		*last_len = 0;
721 		*skip = extent_map_end(em);
722 		*defrag_end = 0;
723 	}
724 
725 	free_extent_map(em);
726 	return ret;
727 }
728 
729 static int btrfs_defrag_file(struct file *file,
730 			     struct btrfs_ioctl_defrag_range_args *range)
731 {
732 	struct inode *inode = fdentry(file)->d_inode;
733 	struct btrfs_root *root = BTRFS_I(inode)->root;
734 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
735 	struct btrfs_ordered_extent *ordered;
736 	struct page *page;
737 	struct btrfs_super_block *disk_super;
738 	unsigned long last_index;
739 	unsigned long ra_pages = root->fs_info->bdi.ra_pages;
740 	unsigned long total_read = 0;
741 	u64 features;
742 	u64 page_start;
743 	u64 page_end;
744 	u64 last_len = 0;
745 	u64 skip = 0;
746 	u64 defrag_end = 0;
747 	unsigned long i;
748 	int ret;
749 	int compress_type = BTRFS_COMPRESS_ZLIB;
750 
751 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
752 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
753 			return -EINVAL;
754 		if (range->compress_type)
755 			compress_type = range->compress_type;
756 	}
757 
758 	if (inode->i_size == 0)
759 		return 0;
760 
761 	if (range->start + range->len > range->start) {
762 		last_index = min_t(u64, inode->i_size - 1,
763 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
764 	} else {
765 		last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
766 	}
767 
768 	i = range->start >> PAGE_CACHE_SHIFT;
769 	while (i <= last_index) {
770 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
771 					PAGE_CACHE_SIZE,
772 					range->extent_thresh,
773 					&last_len, &skip,
774 					&defrag_end)) {
775 			unsigned long next;
776 			/*
777 			 * the should_defrag function tells us how much to skip
778 			 * bump our counter by the suggested amount
779 			 */
780 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
781 			i = max(i + 1, next);
782 			continue;
783 		}
784 
785 		if (total_read % ra_pages == 0) {
786 			btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
787 				       min(last_index, i + ra_pages - 1));
788 		}
789 		total_read++;
790 		mutex_lock(&inode->i_mutex);
791 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
792 			BTRFS_I(inode)->force_compress = compress_type;
793 
794 		ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
795 		if (ret)
796 			goto err_unlock;
797 again:
798 		if (inode->i_size == 0 ||
799 		    i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) {
800 			ret = 0;
801 			goto err_reservations;
802 		}
803 
804 		page = grab_cache_page(inode->i_mapping, i);
805 		if (!page) {
806 			ret = -ENOMEM;
807 			goto err_reservations;
808 		}
809 
810 		if (!PageUptodate(page)) {
811 			btrfs_readpage(NULL, page);
812 			lock_page(page);
813 			if (!PageUptodate(page)) {
814 				unlock_page(page);
815 				page_cache_release(page);
816 				ret = -EIO;
817 				goto err_reservations;
818 			}
819 		}
820 
821 		if (page->mapping != inode->i_mapping) {
822 			unlock_page(page);
823 			page_cache_release(page);
824 			goto again;
825 		}
826 
827 		wait_on_page_writeback(page);
828 
829 		if (PageDirty(page)) {
830 			btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
831 			goto loop_unlock;
832 		}
833 
834 		page_start = (u64)page->index << PAGE_CACHE_SHIFT;
835 		page_end = page_start + PAGE_CACHE_SIZE - 1;
836 		lock_extent(io_tree, page_start, page_end, GFP_NOFS);
837 
838 		ordered = btrfs_lookup_ordered_extent(inode, page_start);
839 		if (ordered) {
840 			unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
841 			unlock_page(page);
842 			page_cache_release(page);
843 			btrfs_start_ordered_extent(inode, ordered, 1);
844 			btrfs_put_ordered_extent(ordered);
845 			goto again;
846 		}
847 		set_page_extent_mapped(page);
848 
849 		/*
850 		 * this makes sure page_mkwrite is called on the
851 		 * page if it is dirtied again later
852 		 */
853 		clear_page_dirty_for_io(page);
854 		clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start,
855 				  page_end, EXTENT_DIRTY | EXTENT_DELALLOC |
856 				  EXTENT_DO_ACCOUNTING, GFP_NOFS);
857 
858 		btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
859 		ClearPageChecked(page);
860 		set_page_dirty(page);
861 		unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
862 
863 loop_unlock:
864 		unlock_page(page);
865 		page_cache_release(page);
866 		mutex_unlock(&inode->i_mutex);
867 
868 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
869 		i++;
870 	}
871 
872 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
873 		filemap_flush(inode->i_mapping);
874 
875 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
876 		/* the filemap_flush will queue IO into the worker threads, but
877 		 * we have to make sure the IO is actually started and that
878 		 * ordered extents get created before we return
879 		 */
880 		atomic_inc(&root->fs_info->async_submit_draining);
881 		while (atomic_read(&root->fs_info->nr_async_submits) ||
882 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
883 			wait_event(root->fs_info->async_submit_wait,
884 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
885 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
886 		}
887 		atomic_dec(&root->fs_info->async_submit_draining);
888 
889 		mutex_lock(&inode->i_mutex);
890 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
891 		mutex_unlock(&inode->i_mutex);
892 	}
893 
894 	disk_super = &root->fs_info->super_copy;
895 	features = btrfs_super_incompat_flags(disk_super);
896 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
897 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
898 		btrfs_set_super_incompat_flags(disk_super, features);
899 	}
900 
901 	return 0;
902 
903 err_reservations:
904 	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
905 err_unlock:
906 	mutex_unlock(&inode->i_mutex);
907 	return ret;
908 }
909 
910 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
911 					void __user *arg)
912 {
913 	u64 new_size;
914 	u64 old_size;
915 	u64 devid = 1;
916 	struct btrfs_ioctl_vol_args *vol_args;
917 	struct btrfs_trans_handle *trans;
918 	struct btrfs_device *device = NULL;
919 	char *sizestr;
920 	char *devstr = NULL;
921 	int ret = 0;
922 	int mod = 0;
923 
924 	if (root->fs_info->sb->s_flags & MS_RDONLY)
925 		return -EROFS;
926 
927 	if (!capable(CAP_SYS_ADMIN))
928 		return -EPERM;
929 
930 	vol_args = memdup_user(arg, sizeof(*vol_args));
931 	if (IS_ERR(vol_args))
932 		return PTR_ERR(vol_args);
933 
934 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
935 
936 	mutex_lock(&root->fs_info->volume_mutex);
937 	sizestr = vol_args->name;
938 	devstr = strchr(sizestr, ':');
939 	if (devstr) {
940 		char *end;
941 		sizestr = devstr + 1;
942 		*devstr = '\0';
943 		devstr = vol_args->name;
944 		devid = simple_strtoull(devstr, &end, 10);
945 		printk(KERN_INFO "resizing devid %llu\n",
946 		       (unsigned long long)devid);
947 	}
948 	device = btrfs_find_device(root, devid, NULL, NULL);
949 	if (!device) {
950 		printk(KERN_INFO "resizer unable to find device %llu\n",
951 		       (unsigned long long)devid);
952 		ret = -EINVAL;
953 		goto out_unlock;
954 	}
955 	if (!strcmp(sizestr, "max"))
956 		new_size = device->bdev->bd_inode->i_size;
957 	else {
958 		if (sizestr[0] == '-') {
959 			mod = -1;
960 			sizestr++;
961 		} else if (sizestr[0] == '+') {
962 			mod = 1;
963 			sizestr++;
964 		}
965 		new_size = memparse(sizestr, NULL);
966 		if (new_size == 0) {
967 			ret = -EINVAL;
968 			goto out_unlock;
969 		}
970 	}
971 
972 	old_size = device->total_bytes;
973 
974 	if (mod < 0) {
975 		if (new_size > old_size) {
976 			ret = -EINVAL;
977 			goto out_unlock;
978 		}
979 		new_size = old_size - new_size;
980 	} else if (mod > 0) {
981 		new_size = old_size + new_size;
982 	}
983 
984 	if (new_size < 256 * 1024 * 1024) {
985 		ret = -EINVAL;
986 		goto out_unlock;
987 	}
988 	if (new_size > device->bdev->bd_inode->i_size) {
989 		ret = -EFBIG;
990 		goto out_unlock;
991 	}
992 
993 	do_div(new_size, root->sectorsize);
994 	new_size *= root->sectorsize;
995 
996 	printk(KERN_INFO "new size for %s is %llu\n",
997 		device->name, (unsigned long long)new_size);
998 
999 	if (new_size > old_size) {
1000 		trans = btrfs_start_transaction(root, 0);
1001 		if (IS_ERR(trans)) {
1002 			ret = PTR_ERR(trans);
1003 			goto out_unlock;
1004 		}
1005 		ret = btrfs_grow_device(trans, device, new_size);
1006 		btrfs_commit_transaction(trans, root);
1007 	} else {
1008 		ret = btrfs_shrink_device(device, new_size);
1009 	}
1010 
1011 out_unlock:
1012 	mutex_unlock(&root->fs_info->volume_mutex);
1013 	kfree(vol_args);
1014 	return ret;
1015 }
1016 
1017 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1018 						    char *name,
1019 						    unsigned long fd,
1020 						    int subvol,
1021 						    u64 *transid,
1022 						    bool readonly)
1023 {
1024 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1025 	struct file *src_file;
1026 	int namelen;
1027 	int ret = 0;
1028 
1029 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1030 		return -EROFS;
1031 
1032 	namelen = strlen(name);
1033 	if (strchr(name, '/')) {
1034 		ret = -EINVAL;
1035 		goto out;
1036 	}
1037 
1038 	if (subvol) {
1039 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1040 				     NULL, transid, readonly);
1041 	} else {
1042 		struct inode *src_inode;
1043 		src_file = fget(fd);
1044 		if (!src_file) {
1045 			ret = -EINVAL;
1046 			goto out;
1047 		}
1048 
1049 		src_inode = src_file->f_path.dentry->d_inode;
1050 		if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1051 			printk(KERN_INFO "btrfs: Snapshot src from "
1052 			       "another FS\n");
1053 			ret = -EINVAL;
1054 			fput(src_file);
1055 			goto out;
1056 		}
1057 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1058 				     BTRFS_I(src_inode)->root,
1059 				     transid, readonly);
1060 		fput(src_file);
1061 	}
1062 out:
1063 	return ret;
1064 }
1065 
1066 static noinline int btrfs_ioctl_snap_create(struct file *file,
1067 					    void __user *arg, int subvol)
1068 {
1069 	struct btrfs_ioctl_vol_args *vol_args;
1070 	int ret;
1071 
1072 	vol_args = memdup_user(arg, sizeof(*vol_args));
1073 	if (IS_ERR(vol_args))
1074 		return PTR_ERR(vol_args);
1075 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1076 
1077 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1078 					      vol_args->fd, subvol,
1079 					      NULL, false);
1080 
1081 	kfree(vol_args);
1082 	return ret;
1083 }
1084 
1085 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1086 					       void __user *arg, int subvol)
1087 {
1088 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1089 	int ret;
1090 	u64 transid = 0;
1091 	u64 *ptr = NULL;
1092 	bool readonly = false;
1093 
1094 	vol_args = memdup_user(arg, sizeof(*vol_args));
1095 	if (IS_ERR(vol_args))
1096 		return PTR_ERR(vol_args);
1097 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1098 
1099 	if (vol_args->flags &
1100 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1101 		ret = -EOPNOTSUPP;
1102 		goto out;
1103 	}
1104 
1105 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1106 		ptr = &transid;
1107 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1108 		readonly = true;
1109 
1110 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1111 					      vol_args->fd, subvol,
1112 					      ptr, readonly);
1113 
1114 	if (ret == 0 && ptr &&
1115 	    copy_to_user(arg +
1116 			 offsetof(struct btrfs_ioctl_vol_args_v2,
1117 				  transid), ptr, sizeof(*ptr)))
1118 		ret = -EFAULT;
1119 out:
1120 	kfree(vol_args);
1121 	return ret;
1122 }
1123 
1124 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1125 						void __user *arg)
1126 {
1127 	struct inode *inode = fdentry(file)->d_inode;
1128 	struct btrfs_root *root = BTRFS_I(inode)->root;
1129 	int ret = 0;
1130 	u64 flags = 0;
1131 
1132 	if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID)
1133 		return -EINVAL;
1134 
1135 	down_read(&root->fs_info->subvol_sem);
1136 	if (btrfs_root_readonly(root))
1137 		flags |= BTRFS_SUBVOL_RDONLY;
1138 	up_read(&root->fs_info->subvol_sem);
1139 
1140 	if (copy_to_user(arg, &flags, sizeof(flags)))
1141 		ret = -EFAULT;
1142 
1143 	return ret;
1144 }
1145 
1146 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1147 					      void __user *arg)
1148 {
1149 	struct inode *inode = fdentry(file)->d_inode;
1150 	struct btrfs_root *root = BTRFS_I(inode)->root;
1151 	struct btrfs_trans_handle *trans;
1152 	u64 root_flags;
1153 	u64 flags;
1154 	int ret = 0;
1155 
1156 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1157 		return -EROFS;
1158 
1159 	if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID)
1160 		return -EINVAL;
1161 
1162 	if (copy_from_user(&flags, arg, sizeof(flags)))
1163 		return -EFAULT;
1164 
1165 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1166 		return -EINVAL;
1167 
1168 	if (flags & ~BTRFS_SUBVOL_RDONLY)
1169 		return -EOPNOTSUPP;
1170 
1171 	if (!inode_owner_or_capable(inode))
1172 		return -EACCES;
1173 
1174 	down_write(&root->fs_info->subvol_sem);
1175 
1176 	/* nothing to do */
1177 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1178 		goto out;
1179 
1180 	root_flags = btrfs_root_flags(&root->root_item);
1181 	if (flags & BTRFS_SUBVOL_RDONLY)
1182 		btrfs_set_root_flags(&root->root_item,
1183 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1184 	else
1185 		btrfs_set_root_flags(&root->root_item,
1186 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1187 
1188 	trans = btrfs_start_transaction(root, 1);
1189 	if (IS_ERR(trans)) {
1190 		ret = PTR_ERR(trans);
1191 		goto out_reset;
1192 	}
1193 
1194 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1195 				&root->root_key, &root->root_item);
1196 
1197 	btrfs_commit_transaction(trans, root);
1198 out_reset:
1199 	if (ret)
1200 		btrfs_set_root_flags(&root->root_item, root_flags);
1201 out:
1202 	up_write(&root->fs_info->subvol_sem);
1203 	return ret;
1204 }
1205 
1206 /*
1207  * helper to check if the subvolume references other subvolumes
1208  */
1209 static noinline int may_destroy_subvol(struct btrfs_root *root)
1210 {
1211 	struct btrfs_path *path;
1212 	struct btrfs_key key;
1213 	int ret;
1214 
1215 	path = btrfs_alloc_path();
1216 	if (!path)
1217 		return -ENOMEM;
1218 
1219 	key.objectid = root->root_key.objectid;
1220 	key.type = BTRFS_ROOT_REF_KEY;
1221 	key.offset = (u64)-1;
1222 
1223 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1224 				&key, path, 0, 0);
1225 	if (ret < 0)
1226 		goto out;
1227 	BUG_ON(ret == 0);
1228 
1229 	ret = 0;
1230 	if (path->slots[0] > 0) {
1231 		path->slots[0]--;
1232 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1233 		if (key.objectid == root->root_key.objectid &&
1234 		    key.type == BTRFS_ROOT_REF_KEY)
1235 			ret = -ENOTEMPTY;
1236 	}
1237 out:
1238 	btrfs_free_path(path);
1239 	return ret;
1240 }
1241 
1242 static noinline int key_in_sk(struct btrfs_key *key,
1243 			      struct btrfs_ioctl_search_key *sk)
1244 {
1245 	struct btrfs_key test;
1246 	int ret;
1247 
1248 	test.objectid = sk->min_objectid;
1249 	test.type = sk->min_type;
1250 	test.offset = sk->min_offset;
1251 
1252 	ret = btrfs_comp_cpu_keys(key, &test);
1253 	if (ret < 0)
1254 		return 0;
1255 
1256 	test.objectid = sk->max_objectid;
1257 	test.type = sk->max_type;
1258 	test.offset = sk->max_offset;
1259 
1260 	ret = btrfs_comp_cpu_keys(key, &test);
1261 	if (ret > 0)
1262 		return 0;
1263 	return 1;
1264 }
1265 
1266 static noinline int copy_to_sk(struct btrfs_root *root,
1267 			       struct btrfs_path *path,
1268 			       struct btrfs_key *key,
1269 			       struct btrfs_ioctl_search_key *sk,
1270 			       char *buf,
1271 			       unsigned long *sk_offset,
1272 			       int *num_found)
1273 {
1274 	u64 found_transid;
1275 	struct extent_buffer *leaf;
1276 	struct btrfs_ioctl_search_header sh;
1277 	unsigned long item_off;
1278 	unsigned long item_len;
1279 	int nritems;
1280 	int i;
1281 	int slot;
1282 	int found = 0;
1283 	int ret = 0;
1284 
1285 	leaf = path->nodes[0];
1286 	slot = path->slots[0];
1287 	nritems = btrfs_header_nritems(leaf);
1288 
1289 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1290 		i = nritems;
1291 		goto advance_key;
1292 	}
1293 	found_transid = btrfs_header_generation(leaf);
1294 
1295 	for (i = slot; i < nritems; i++) {
1296 		item_off = btrfs_item_ptr_offset(leaf, i);
1297 		item_len = btrfs_item_size_nr(leaf, i);
1298 
1299 		if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1300 			item_len = 0;
1301 
1302 		if (sizeof(sh) + item_len + *sk_offset >
1303 		    BTRFS_SEARCH_ARGS_BUFSIZE) {
1304 			ret = 1;
1305 			goto overflow;
1306 		}
1307 
1308 		btrfs_item_key_to_cpu(leaf, key, i);
1309 		if (!key_in_sk(key, sk))
1310 			continue;
1311 
1312 		sh.objectid = key->objectid;
1313 		sh.offset = key->offset;
1314 		sh.type = key->type;
1315 		sh.len = item_len;
1316 		sh.transid = found_transid;
1317 
1318 		/* copy search result header */
1319 		memcpy(buf + *sk_offset, &sh, sizeof(sh));
1320 		*sk_offset += sizeof(sh);
1321 
1322 		if (item_len) {
1323 			char *p = buf + *sk_offset;
1324 			/* copy the item */
1325 			read_extent_buffer(leaf, p,
1326 					   item_off, item_len);
1327 			*sk_offset += item_len;
1328 		}
1329 		found++;
1330 
1331 		if (*num_found >= sk->nr_items)
1332 			break;
1333 	}
1334 advance_key:
1335 	ret = 0;
1336 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1337 		key->offset++;
1338 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
1339 		key->offset = 0;
1340 		key->type++;
1341 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1342 		key->offset = 0;
1343 		key->type = 0;
1344 		key->objectid++;
1345 	} else
1346 		ret = 1;
1347 overflow:
1348 	*num_found += found;
1349 	return ret;
1350 }
1351 
1352 static noinline int search_ioctl(struct inode *inode,
1353 				 struct btrfs_ioctl_search_args *args)
1354 {
1355 	struct btrfs_root *root;
1356 	struct btrfs_key key;
1357 	struct btrfs_key max_key;
1358 	struct btrfs_path *path;
1359 	struct btrfs_ioctl_search_key *sk = &args->key;
1360 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1361 	int ret;
1362 	int num_found = 0;
1363 	unsigned long sk_offset = 0;
1364 
1365 	path = btrfs_alloc_path();
1366 	if (!path)
1367 		return -ENOMEM;
1368 
1369 	if (sk->tree_id == 0) {
1370 		/* search the root of the inode that was passed */
1371 		root = BTRFS_I(inode)->root;
1372 	} else {
1373 		key.objectid = sk->tree_id;
1374 		key.type = BTRFS_ROOT_ITEM_KEY;
1375 		key.offset = (u64)-1;
1376 		root = btrfs_read_fs_root_no_name(info, &key);
1377 		if (IS_ERR(root)) {
1378 			printk(KERN_ERR "could not find root %llu\n",
1379 			       sk->tree_id);
1380 			btrfs_free_path(path);
1381 			return -ENOENT;
1382 		}
1383 	}
1384 
1385 	key.objectid = sk->min_objectid;
1386 	key.type = sk->min_type;
1387 	key.offset = sk->min_offset;
1388 
1389 	max_key.objectid = sk->max_objectid;
1390 	max_key.type = sk->max_type;
1391 	max_key.offset = sk->max_offset;
1392 
1393 	path->keep_locks = 1;
1394 
1395 	while(1) {
1396 		ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1397 					   sk->min_transid);
1398 		if (ret != 0) {
1399 			if (ret > 0)
1400 				ret = 0;
1401 			goto err;
1402 		}
1403 		ret = copy_to_sk(root, path, &key, sk, args->buf,
1404 				 &sk_offset, &num_found);
1405 		btrfs_release_path(root, path);
1406 		if (ret || num_found >= sk->nr_items)
1407 			break;
1408 
1409 	}
1410 	ret = 0;
1411 err:
1412 	sk->nr_items = num_found;
1413 	btrfs_free_path(path);
1414 	return ret;
1415 }
1416 
1417 static noinline int btrfs_ioctl_tree_search(struct file *file,
1418 					   void __user *argp)
1419 {
1420 	 struct btrfs_ioctl_search_args *args;
1421 	 struct inode *inode;
1422 	 int ret;
1423 
1424 	if (!capable(CAP_SYS_ADMIN))
1425 		return -EPERM;
1426 
1427 	args = memdup_user(argp, sizeof(*args));
1428 	if (IS_ERR(args))
1429 		return PTR_ERR(args);
1430 
1431 	inode = fdentry(file)->d_inode;
1432 	ret = search_ioctl(inode, args);
1433 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1434 		ret = -EFAULT;
1435 	kfree(args);
1436 	return ret;
1437 }
1438 
1439 /*
1440  * Search INODE_REFs to identify path name of 'dirid' directory
1441  * in a 'tree_id' tree. and sets path name to 'name'.
1442  */
1443 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1444 				u64 tree_id, u64 dirid, char *name)
1445 {
1446 	struct btrfs_root *root;
1447 	struct btrfs_key key;
1448 	char *ptr;
1449 	int ret = -1;
1450 	int slot;
1451 	int len;
1452 	int total_len = 0;
1453 	struct btrfs_inode_ref *iref;
1454 	struct extent_buffer *l;
1455 	struct btrfs_path *path;
1456 
1457 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1458 		name[0]='\0';
1459 		return 0;
1460 	}
1461 
1462 	path = btrfs_alloc_path();
1463 	if (!path)
1464 		return -ENOMEM;
1465 
1466 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1467 
1468 	key.objectid = tree_id;
1469 	key.type = BTRFS_ROOT_ITEM_KEY;
1470 	key.offset = (u64)-1;
1471 	root = btrfs_read_fs_root_no_name(info, &key);
1472 	if (IS_ERR(root)) {
1473 		printk(KERN_ERR "could not find root %llu\n", tree_id);
1474 		ret = -ENOENT;
1475 		goto out;
1476 	}
1477 
1478 	key.objectid = dirid;
1479 	key.type = BTRFS_INODE_REF_KEY;
1480 	key.offset = (u64)-1;
1481 
1482 	while(1) {
1483 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1484 		if (ret < 0)
1485 			goto out;
1486 
1487 		l = path->nodes[0];
1488 		slot = path->slots[0];
1489 		if (ret > 0 && slot > 0)
1490 			slot--;
1491 		btrfs_item_key_to_cpu(l, &key, slot);
1492 
1493 		if (ret > 0 && (key.objectid != dirid ||
1494 				key.type != BTRFS_INODE_REF_KEY)) {
1495 			ret = -ENOENT;
1496 			goto out;
1497 		}
1498 
1499 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1500 		len = btrfs_inode_ref_name_len(l, iref);
1501 		ptr -= len + 1;
1502 		total_len += len + 1;
1503 		if (ptr < name)
1504 			goto out;
1505 
1506 		*(ptr + len) = '/';
1507 		read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1508 
1509 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1510 			break;
1511 
1512 		btrfs_release_path(root, path);
1513 		key.objectid = key.offset;
1514 		key.offset = (u64)-1;
1515 		dirid = key.objectid;
1516 
1517 	}
1518 	if (ptr < name)
1519 		goto out;
1520 	memcpy(name, ptr, total_len);
1521 	name[total_len]='\0';
1522 	ret = 0;
1523 out:
1524 	btrfs_free_path(path);
1525 	return ret;
1526 }
1527 
1528 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1529 					   void __user *argp)
1530 {
1531 	 struct btrfs_ioctl_ino_lookup_args *args;
1532 	 struct inode *inode;
1533 	 int ret;
1534 
1535 	if (!capable(CAP_SYS_ADMIN))
1536 		return -EPERM;
1537 
1538 	args = memdup_user(argp, sizeof(*args));
1539 	if (IS_ERR(args))
1540 		return PTR_ERR(args);
1541 
1542 	inode = fdentry(file)->d_inode;
1543 
1544 	if (args->treeid == 0)
1545 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1546 
1547 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1548 					args->treeid, args->objectid,
1549 					args->name);
1550 
1551 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1552 		ret = -EFAULT;
1553 
1554 	kfree(args);
1555 	return ret;
1556 }
1557 
1558 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1559 					     void __user *arg)
1560 {
1561 	struct dentry *parent = fdentry(file);
1562 	struct dentry *dentry;
1563 	struct inode *dir = parent->d_inode;
1564 	struct inode *inode;
1565 	struct btrfs_root *root = BTRFS_I(dir)->root;
1566 	struct btrfs_root *dest = NULL;
1567 	struct btrfs_ioctl_vol_args *vol_args;
1568 	struct btrfs_trans_handle *trans;
1569 	int namelen;
1570 	int ret;
1571 	int err = 0;
1572 
1573 	vol_args = memdup_user(arg, sizeof(*vol_args));
1574 	if (IS_ERR(vol_args))
1575 		return PTR_ERR(vol_args);
1576 
1577 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1578 	namelen = strlen(vol_args->name);
1579 	if (strchr(vol_args->name, '/') ||
1580 	    strncmp(vol_args->name, "..", namelen) == 0) {
1581 		err = -EINVAL;
1582 		goto out;
1583 	}
1584 
1585 	err = mnt_want_write(file->f_path.mnt);
1586 	if (err)
1587 		goto out;
1588 
1589 	mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1590 	dentry = lookup_one_len(vol_args->name, parent, namelen);
1591 	if (IS_ERR(dentry)) {
1592 		err = PTR_ERR(dentry);
1593 		goto out_unlock_dir;
1594 	}
1595 
1596 	if (!dentry->d_inode) {
1597 		err = -ENOENT;
1598 		goto out_dput;
1599 	}
1600 
1601 	inode = dentry->d_inode;
1602 	dest = BTRFS_I(inode)->root;
1603 	if (!capable(CAP_SYS_ADMIN)){
1604 		/*
1605 		 * Regular user.  Only allow this with a special mount
1606 		 * option, when the user has write+exec access to the
1607 		 * subvol root, and when rmdir(2) would have been
1608 		 * allowed.
1609 		 *
1610 		 * Note that this is _not_ check that the subvol is
1611 		 * empty or doesn't contain data that we wouldn't
1612 		 * otherwise be able to delete.
1613 		 *
1614 		 * Users who want to delete empty subvols should try
1615 		 * rmdir(2).
1616 		 */
1617 		err = -EPERM;
1618 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1619 			goto out_dput;
1620 
1621 		/*
1622 		 * Do not allow deletion if the parent dir is the same
1623 		 * as the dir to be deleted.  That means the ioctl
1624 		 * must be called on the dentry referencing the root
1625 		 * of the subvol, not a random directory contained
1626 		 * within it.
1627 		 */
1628 		err = -EINVAL;
1629 		if (root == dest)
1630 			goto out_dput;
1631 
1632 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1633 		if (err)
1634 			goto out_dput;
1635 
1636 		/* check if subvolume may be deleted by a non-root user */
1637 		err = btrfs_may_delete(dir, dentry, 1);
1638 		if (err)
1639 			goto out_dput;
1640 	}
1641 
1642 	if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
1643 		err = -EINVAL;
1644 		goto out_dput;
1645 	}
1646 
1647 	mutex_lock(&inode->i_mutex);
1648 	err = d_invalidate(dentry);
1649 	if (err)
1650 		goto out_unlock;
1651 
1652 	down_write(&root->fs_info->subvol_sem);
1653 
1654 	err = may_destroy_subvol(dest);
1655 	if (err)
1656 		goto out_up_write;
1657 
1658 	trans = btrfs_start_transaction(root, 0);
1659 	if (IS_ERR(trans)) {
1660 		err = PTR_ERR(trans);
1661 		goto out_up_write;
1662 	}
1663 	trans->block_rsv = &root->fs_info->global_block_rsv;
1664 
1665 	ret = btrfs_unlink_subvol(trans, root, dir,
1666 				dest->root_key.objectid,
1667 				dentry->d_name.name,
1668 				dentry->d_name.len);
1669 	BUG_ON(ret);
1670 
1671 	btrfs_record_root_in_trans(trans, dest);
1672 
1673 	memset(&dest->root_item.drop_progress, 0,
1674 		sizeof(dest->root_item.drop_progress));
1675 	dest->root_item.drop_level = 0;
1676 	btrfs_set_root_refs(&dest->root_item, 0);
1677 
1678 	if (!xchg(&dest->orphan_item_inserted, 1)) {
1679 		ret = btrfs_insert_orphan_item(trans,
1680 					root->fs_info->tree_root,
1681 					dest->root_key.objectid);
1682 		BUG_ON(ret);
1683 	}
1684 
1685 	ret = btrfs_end_transaction(trans, root);
1686 	BUG_ON(ret);
1687 	inode->i_flags |= S_DEAD;
1688 out_up_write:
1689 	up_write(&root->fs_info->subvol_sem);
1690 out_unlock:
1691 	mutex_unlock(&inode->i_mutex);
1692 	if (!err) {
1693 		shrink_dcache_sb(root->fs_info->sb);
1694 		btrfs_invalidate_inodes(dest);
1695 		d_delete(dentry);
1696 	}
1697 out_dput:
1698 	dput(dentry);
1699 out_unlock_dir:
1700 	mutex_unlock(&dir->i_mutex);
1701 	mnt_drop_write(file->f_path.mnt);
1702 out:
1703 	kfree(vol_args);
1704 	return err;
1705 }
1706 
1707 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1708 {
1709 	struct inode *inode = fdentry(file)->d_inode;
1710 	struct btrfs_root *root = BTRFS_I(inode)->root;
1711 	struct btrfs_ioctl_defrag_range_args *range;
1712 	int ret;
1713 
1714 	if (btrfs_root_readonly(root))
1715 		return -EROFS;
1716 
1717 	ret = mnt_want_write(file->f_path.mnt);
1718 	if (ret)
1719 		return ret;
1720 
1721 	switch (inode->i_mode & S_IFMT) {
1722 	case S_IFDIR:
1723 		if (!capable(CAP_SYS_ADMIN)) {
1724 			ret = -EPERM;
1725 			goto out;
1726 		}
1727 		ret = btrfs_defrag_root(root, 0);
1728 		if (ret)
1729 			goto out;
1730 		ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1731 		break;
1732 	case S_IFREG:
1733 		if (!(file->f_mode & FMODE_WRITE)) {
1734 			ret = -EINVAL;
1735 			goto out;
1736 		}
1737 
1738 		range = kzalloc(sizeof(*range), GFP_KERNEL);
1739 		if (!range) {
1740 			ret = -ENOMEM;
1741 			goto out;
1742 		}
1743 
1744 		if (argp) {
1745 			if (copy_from_user(range, argp,
1746 					   sizeof(*range))) {
1747 				ret = -EFAULT;
1748 				kfree(range);
1749 				goto out;
1750 			}
1751 			/* compression requires us to start the IO */
1752 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1753 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1754 				range->extent_thresh = (u32)-1;
1755 			}
1756 		} else {
1757 			/* the rest are all set to zero by kzalloc */
1758 			range->len = (u64)-1;
1759 		}
1760 		ret = btrfs_defrag_file(file, range);
1761 		kfree(range);
1762 		break;
1763 	default:
1764 		ret = -EINVAL;
1765 	}
1766 out:
1767 	mnt_drop_write(file->f_path.mnt);
1768 	return ret;
1769 }
1770 
1771 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
1772 {
1773 	struct btrfs_ioctl_vol_args *vol_args;
1774 	int ret;
1775 
1776 	if (!capable(CAP_SYS_ADMIN))
1777 		return -EPERM;
1778 
1779 	vol_args = memdup_user(arg, sizeof(*vol_args));
1780 	if (IS_ERR(vol_args))
1781 		return PTR_ERR(vol_args);
1782 
1783 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1784 	ret = btrfs_init_new_device(root, vol_args->name);
1785 
1786 	kfree(vol_args);
1787 	return ret;
1788 }
1789 
1790 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
1791 {
1792 	struct btrfs_ioctl_vol_args *vol_args;
1793 	int ret;
1794 
1795 	if (!capable(CAP_SYS_ADMIN))
1796 		return -EPERM;
1797 
1798 	if (root->fs_info->sb->s_flags & MS_RDONLY)
1799 		return -EROFS;
1800 
1801 	vol_args = memdup_user(arg, sizeof(*vol_args));
1802 	if (IS_ERR(vol_args))
1803 		return PTR_ERR(vol_args);
1804 
1805 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1806 	ret = btrfs_rm_device(root, vol_args->name);
1807 
1808 	kfree(vol_args);
1809 	return ret;
1810 }
1811 
1812 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1813 				       u64 off, u64 olen, u64 destoff)
1814 {
1815 	struct inode *inode = fdentry(file)->d_inode;
1816 	struct btrfs_root *root = BTRFS_I(inode)->root;
1817 	struct file *src_file;
1818 	struct inode *src;
1819 	struct btrfs_trans_handle *trans;
1820 	struct btrfs_path *path;
1821 	struct extent_buffer *leaf;
1822 	char *buf;
1823 	struct btrfs_key key;
1824 	u32 nritems;
1825 	int slot;
1826 	int ret;
1827 	u64 len = olen;
1828 	u64 bs = root->fs_info->sb->s_blocksize;
1829 	u64 hint_byte;
1830 
1831 	/*
1832 	 * TODO:
1833 	 * - split compressed inline extents.  annoying: we need to
1834 	 *   decompress into destination's address_space (the file offset
1835 	 *   may change, so source mapping won't do), then recompress (or
1836 	 *   otherwise reinsert) a subrange.
1837 	 * - allow ranges within the same file to be cloned (provided
1838 	 *   they don't overlap)?
1839 	 */
1840 
1841 	/* the destination must be opened for writing */
1842 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
1843 		return -EINVAL;
1844 
1845 	if (btrfs_root_readonly(root))
1846 		return -EROFS;
1847 
1848 	ret = mnt_want_write(file->f_path.mnt);
1849 	if (ret)
1850 		return ret;
1851 
1852 	src_file = fget(srcfd);
1853 	if (!src_file) {
1854 		ret = -EBADF;
1855 		goto out_drop_write;
1856 	}
1857 
1858 	src = src_file->f_dentry->d_inode;
1859 
1860 	ret = -EINVAL;
1861 	if (src == inode)
1862 		goto out_fput;
1863 
1864 	/* the src must be open for reading */
1865 	if (!(src_file->f_mode & FMODE_READ))
1866 		goto out_fput;
1867 
1868 	ret = -EISDIR;
1869 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
1870 		goto out_fput;
1871 
1872 	ret = -EXDEV;
1873 	if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
1874 		goto out_fput;
1875 
1876 	ret = -ENOMEM;
1877 	buf = vmalloc(btrfs_level_size(root, 0));
1878 	if (!buf)
1879 		goto out_fput;
1880 
1881 	path = btrfs_alloc_path();
1882 	if (!path) {
1883 		vfree(buf);
1884 		goto out_fput;
1885 	}
1886 	path->reada = 2;
1887 
1888 	if (inode < src) {
1889 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
1890 		mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
1891 	} else {
1892 		mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
1893 		mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
1894 	}
1895 
1896 	/* determine range to clone */
1897 	ret = -EINVAL;
1898 	if (off + len > src->i_size || off + len < off)
1899 		goto out_unlock;
1900 	if (len == 0)
1901 		olen = len = src->i_size - off;
1902 	/* if we extend to eof, continue to block boundary */
1903 	if (off + len == src->i_size)
1904 		len = ALIGN(src->i_size, bs) - off;
1905 
1906 	/* verify the end result is block aligned */
1907 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
1908 	    !IS_ALIGNED(destoff, bs))
1909 		goto out_unlock;
1910 
1911 	/* do any pending delalloc/csum calc on src, one way or
1912 	   another, and lock file content */
1913 	while (1) {
1914 		struct btrfs_ordered_extent *ordered;
1915 		lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1916 		ordered = btrfs_lookup_first_ordered_extent(src, off+len);
1917 		if (!ordered &&
1918 		    !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
1919 				   EXTENT_DELALLOC, 0, NULL))
1920 			break;
1921 		unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1922 		if (ordered)
1923 			btrfs_put_ordered_extent(ordered);
1924 		btrfs_wait_ordered_range(src, off, len);
1925 	}
1926 
1927 	/* clone data */
1928 	key.objectid = src->i_ino;
1929 	key.type = BTRFS_EXTENT_DATA_KEY;
1930 	key.offset = 0;
1931 
1932 	while (1) {
1933 		/*
1934 		 * note the key will change type as we walk through the
1935 		 * tree.
1936 		 */
1937 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1938 		if (ret < 0)
1939 			goto out;
1940 
1941 		nritems = btrfs_header_nritems(path->nodes[0]);
1942 		if (path->slots[0] >= nritems) {
1943 			ret = btrfs_next_leaf(root, path);
1944 			if (ret < 0)
1945 				goto out;
1946 			if (ret > 0)
1947 				break;
1948 			nritems = btrfs_header_nritems(path->nodes[0]);
1949 		}
1950 		leaf = path->nodes[0];
1951 		slot = path->slots[0];
1952 
1953 		btrfs_item_key_to_cpu(leaf, &key, slot);
1954 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1955 		    key.objectid != src->i_ino)
1956 			break;
1957 
1958 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1959 			struct btrfs_file_extent_item *extent;
1960 			int type;
1961 			u32 size;
1962 			struct btrfs_key new_key;
1963 			u64 disko = 0, diskl = 0;
1964 			u64 datao = 0, datal = 0;
1965 			u8 comp;
1966 			u64 endoff;
1967 
1968 			size = btrfs_item_size_nr(leaf, slot);
1969 			read_extent_buffer(leaf, buf,
1970 					   btrfs_item_ptr_offset(leaf, slot),
1971 					   size);
1972 
1973 			extent = btrfs_item_ptr(leaf, slot,
1974 						struct btrfs_file_extent_item);
1975 			comp = btrfs_file_extent_compression(leaf, extent);
1976 			type = btrfs_file_extent_type(leaf, extent);
1977 			if (type == BTRFS_FILE_EXTENT_REG ||
1978 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
1979 				disko = btrfs_file_extent_disk_bytenr(leaf,
1980 								      extent);
1981 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
1982 								 extent);
1983 				datao = btrfs_file_extent_offset(leaf, extent);
1984 				datal = btrfs_file_extent_num_bytes(leaf,
1985 								    extent);
1986 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
1987 				/* take upper bound, may be compressed */
1988 				datal = btrfs_file_extent_ram_bytes(leaf,
1989 								    extent);
1990 			}
1991 			btrfs_release_path(root, path);
1992 
1993 			if (key.offset + datal <= off ||
1994 			    key.offset >= off+len)
1995 				goto next;
1996 
1997 			memcpy(&new_key, &key, sizeof(new_key));
1998 			new_key.objectid = inode->i_ino;
1999 			if (off <= key.offset)
2000 				new_key.offset = key.offset + destoff - off;
2001 			else
2002 				new_key.offset = destoff;
2003 
2004 			trans = btrfs_start_transaction(root, 1);
2005 			if (IS_ERR(trans)) {
2006 				ret = PTR_ERR(trans);
2007 				goto out;
2008 			}
2009 
2010 			if (type == BTRFS_FILE_EXTENT_REG ||
2011 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
2012 				if (off > key.offset) {
2013 					datao += off - key.offset;
2014 					datal -= off - key.offset;
2015 				}
2016 
2017 				if (key.offset + datal > off + len)
2018 					datal = off + len - key.offset;
2019 
2020 				ret = btrfs_drop_extents(trans, inode,
2021 							 new_key.offset,
2022 							 new_key.offset + datal,
2023 							 &hint_byte, 1);
2024 				BUG_ON(ret);
2025 
2026 				ret = btrfs_insert_empty_item(trans, root, path,
2027 							      &new_key, size);
2028 				BUG_ON(ret);
2029 
2030 				leaf = path->nodes[0];
2031 				slot = path->slots[0];
2032 				write_extent_buffer(leaf, buf,
2033 					    btrfs_item_ptr_offset(leaf, slot),
2034 					    size);
2035 
2036 				extent = btrfs_item_ptr(leaf, slot,
2037 						struct btrfs_file_extent_item);
2038 
2039 				/* disko == 0 means it's a hole */
2040 				if (!disko)
2041 					datao = 0;
2042 
2043 				btrfs_set_file_extent_offset(leaf, extent,
2044 							     datao);
2045 				btrfs_set_file_extent_num_bytes(leaf, extent,
2046 								datal);
2047 				if (disko) {
2048 					inode_add_bytes(inode, datal);
2049 					ret = btrfs_inc_extent_ref(trans, root,
2050 							disko, diskl, 0,
2051 							root->root_key.objectid,
2052 							inode->i_ino,
2053 							new_key.offset - datao);
2054 					BUG_ON(ret);
2055 				}
2056 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
2057 				u64 skip = 0;
2058 				u64 trim = 0;
2059 				if (off > key.offset) {
2060 					skip = off - key.offset;
2061 					new_key.offset += skip;
2062 				}
2063 
2064 				if (key.offset + datal > off+len)
2065 					trim = key.offset + datal - (off+len);
2066 
2067 				if (comp && (skip || trim)) {
2068 					ret = -EINVAL;
2069 					btrfs_end_transaction(trans, root);
2070 					goto out;
2071 				}
2072 				size -= skip + trim;
2073 				datal -= skip + trim;
2074 
2075 				ret = btrfs_drop_extents(trans, inode,
2076 							 new_key.offset,
2077 							 new_key.offset + datal,
2078 							 &hint_byte, 1);
2079 				BUG_ON(ret);
2080 
2081 				ret = btrfs_insert_empty_item(trans, root, path,
2082 							      &new_key, size);
2083 				BUG_ON(ret);
2084 
2085 				if (skip) {
2086 					u32 start =
2087 					  btrfs_file_extent_calc_inline_size(0);
2088 					memmove(buf+start, buf+start+skip,
2089 						datal);
2090 				}
2091 
2092 				leaf = path->nodes[0];
2093 				slot = path->slots[0];
2094 				write_extent_buffer(leaf, buf,
2095 					    btrfs_item_ptr_offset(leaf, slot),
2096 					    size);
2097 				inode_add_bytes(inode, datal);
2098 			}
2099 
2100 			btrfs_mark_buffer_dirty(leaf);
2101 			btrfs_release_path(root, path);
2102 
2103 			inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2104 
2105 			/*
2106 			 * we round up to the block size at eof when
2107 			 * determining which extents to clone above,
2108 			 * but shouldn't round up the file size
2109 			 */
2110 			endoff = new_key.offset + datal;
2111 			if (endoff > destoff+olen)
2112 				endoff = destoff+olen;
2113 			if (endoff > inode->i_size)
2114 				btrfs_i_size_write(inode, endoff);
2115 
2116 			BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
2117 			ret = btrfs_update_inode(trans, root, inode);
2118 			BUG_ON(ret);
2119 			btrfs_end_transaction(trans, root);
2120 		}
2121 next:
2122 		btrfs_release_path(root, path);
2123 		key.offset++;
2124 	}
2125 	ret = 0;
2126 out:
2127 	btrfs_release_path(root, path);
2128 	unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2129 out_unlock:
2130 	mutex_unlock(&src->i_mutex);
2131 	mutex_unlock(&inode->i_mutex);
2132 	vfree(buf);
2133 	btrfs_free_path(path);
2134 out_fput:
2135 	fput(src_file);
2136 out_drop_write:
2137 	mnt_drop_write(file->f_path.mnt);
2138 	return ret;
2139 }
2140 
2141 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2142 {
2143 	struct btrfs_ioctl_clone_range_args args;
2144 
2145 	if (copy_from_user(&args, argp, sizeof(args)))
2146 		return -EFAULT;
2147 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2148 				 args.src_length, args.dest_offset);
2149 }
2150 
2151 /*
2152  * there are many ways the trans_start and trans_end ioctls can lead
2153  * to deadlocks.  They should only be used by applications that
2154  * basically own the machine, and have a very in depth understanding
2155  * of all the possible deadlocks and enospc problems.
2156  */
2157 static long btrfs_ioctl_trans_start(struct file *file)
2158 {
2159 	struct inode *inode = fdentry(file)->d_inode;
2160 	struct btrfs_root *root = BTRFS_I(inode)->root;
2161 	struct btrfs_trans_handle *trans;
2162 	int ret;
2163 
2164 	ret = -EPERM;
2165 	if (!capable(CAP_SYS_ADMIN))
2166 		goto out;
2167 
2168 	ret = -EINPROGRESS;
2169 	if (file->private_data)
2170 		goto out;
2171 
2172 	ret = -EROFS;
2173 	if (btrfs_root_readonly(root))
2174 		goto out;
2175 
2176 	ret = mnt_want_write(file->f_path.mnt);
2177 	if (ret)
2178 		goto out;
2179 
2180 	mutex_lock(&root->fs_info->trans_mutex);
2181 	root->fs_info->open_ioctl_trans++;
2182 	mutex_unlock(&root->fs_info->trans_mutex);
2183 
2184 	ret = -ENOMEM;
2185 	trans = btrfs_start_ioctl_transaction(root, 0);
2186 	if (IS_ERR(trans))
2187 		goto out_drop;
2188 
2189 	file->private_data = trans;
2190 	return 0;
2191 
2192 out_drop:
2193 	mutex_lock(&root->fs_info->trans_mutex);
2194 	root->fs_info->open_ioctl_trans--;
2195 	mutex_unlock(&root->fs_info->trans_mutex);
2196 	mnt_drop_write(file->f_path.mnt);
2197 out:
2198 	return ret;
2199 }
2200 
2201 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2202 {
2203 	struct inode *inode = fdentry(file)->d_inode;
2204 	struct btrfs_root *root = BTRFS_I(inode)->root;
2205 	struct btrfs_root *new_root;
2206 	struct btrfs_dir_item *di;
2207 	struct btrfs_trans_handle *trans;
2208 	struct btrfs_path *path;
2209 	struct btrfs_key location;
2210 	struct btrfs_disk_key disk_key;
2211 	struct btrfs_super_block *disk_super;
2212 	u64 features;
2213 	u64 objectid = 0;
2214 	u64 dir_id;
2215 
2216 	if (!capable(CAP_SYS_ADMIN))
2217 		return -EPERM;
2218 
2219 	if (copy_from_user(&objectid, argp, sizeof(objectid)))
2220 		return -EFAULT;
2221 
2222 	if (!objectid)
2223 		objectid = root->root_key.objectid;
2224 
2225 	location.objectid = objectid;
2226 	location.type = BTRFS_ROOT_ITEM_KEY;
2227 	location.offset = (u64)-1;
2228 
2229 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2230 	if (IS_ERR(new_root))
2231 		return PTR_ERR(new_root);
2232 
2233 	if (btrfs_root_refs(&new_root->root_item) == 0)
2234 		return -ENOENT;
2235 
2236 	path = btrfs_alloc_path();
2237 	if (!path)
2238 		return -ENOMEM;
2239 	path->leave_spinning = 1;
2240 
2241 	trans = btrfs_start_transaction(root, 1);
2242 	if (IS_ERR(trans)) {
2243 		btrfs_free_path(path);
2244 		return PTR_ERR(trans);
2245 	}
2246 
2247 	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2248 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2249 				   dir_id, "default", 7, 1);
2250 	if (IS_ERR_OR_NULL(di)) {
2251 		btrfs_free_path(path);
2252 		btrfs_end_transaction(trans, root);
2253 		printk(KERN_ERR "Umm, you don't have the default dir item, "
2254 		       "this isn't going to work\n");
2255 		return -ENOENT;
2256 	}
2257 
2258 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2259 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2260 	btrfs_mark_buffer_dirty(path->nodes[0]);
2261 	btrfs_free_path(path);
2262 
2263 	disk_super = &root->fs_info->super_copy;
2264 	features = btrfs_super_incompat_flags(disk_super);
2265 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2266 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2267 		btrfs_set_super_incompat_flags(disk_super, features);
2268 	}
2269 	btrfs_end_transaction(trans, root);
2270 
2271 	return 0;
2272 }
2273 
2274 static void get_block_group_info(struct list_head *groups_list,
2275 				 struct btrfs_ioctl_space_info *space)
2276 {
2277 	struct btrfs_block_group_cache *block_group;
2278 
2279 	space->total_bytes = 0;
2280 	space->used_bytes = 0;
2281 	space->flags = 0;
2282 	list_for_each_entry(block_group, groups_list, list) {
2283 		space->flags = block_group->flags;
2284 		space->total_bytes += block_group->key.offset;
2285 		space->used_bytes +=
2286 			btrfs_block_group_used(&block_group->item);
2287 	}
2288 }
2289 
2290 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2291 {
2292 	struct btrfs_ioctl_space_args space_args;
2293 	struct btrfs_ioctl_space_info space;
2294 	struct btrfs_ioctl_space_info *dest;
2295 	struct btrfs_ioctl_space_info *dest_orig;
2296 	struct btrfs_ioctl_space_info __user *user_dest;
2297 	struct btrfs_space_info *info;
2298 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2299 		       BTRFS_BLOCK_GROUP_SYSTEM,
2300 		       BTRFS_BLOCK_GROUP_METADATA,
2301 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2302 	int num_types = 4;
2303 	int alloc_size;
2304 	int ret = 0;
2305 	u64 slot_count = 0;
2306 	int i, c;
2307 
2308 	if (copy_from_user(&space_args,
2309 			   (struct btrfs_ioctl_space_args __user *)arg,
2310 			   sizeof(space_args)))
2311 		return -EFAULT;
2312 
2313 	for (i = 0; i < num_types; i++) {
2314 		struct btrfs_space_info *tmp;
2315 
2316 		info = NULL;
2317 		rcu_read_lock();
2318 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2319 					list) {
2320 			if (tmp->flags == types[i]) {
2321 				info = tmp;
2322 				break;
2323 			}
2324 		}
2325 		rcu_read_unlock();
2326 
2327 		if (!info)
2328 			continue;
2329 
2330 		down_read(&info->groups_sem);
2331 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2332 			if (!list_empty(&info->block_groups[c]))
2333 				slot_count++;
2334 		}
2335 		up_read(&info->groups_sem);
2336 	}
2337 
2338 	/* space_slots == 0 means they are asking for a count */
2339 	if (space_args.space_slots == 0) {
2340 		space_args.total_spaces = slot_count;
2341 		goto out;
2342 	}
2343 
2344 	slot_count = min_t(u64, space_args.space_slots, slot_count);
2345 
2346 	alloc_size = sizeof(*dest) * slot_count;
2347 
2348 	/* we generally have at most 6 or so space infos, one for each raid
2349 	 * level.  So, a whole page should be more than enough for everyone
2350 	 */
2351 	if (alloc_size > PAGE_CACHE_SIZE)
2352 		return -ENOMEM;
2353 
2354 	space_args.total_spaces = 0;
2355 	dest = kmalloc(alloc_size, GFP_NOFS);
2356 	if (!dest)
2357 		return -ENOMEM;
2358 	dest_orig = dest;
2359 
2360 	/* now we have a buffer to copy into */
2361 	for (i = 0; i < num_types; i++) {
2362 		struct btrfs_space_info *tmp;
2363 
2364 		if (!slot_count)
2365 			break;
2366 
2367 		info = NULL;
2368 		rcu_read_lock();
2369 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2370 					list) {
2371 			if (tmp->flags == types[i]) {
2372 				info = tmp;
2373 				break;
2374 			}
2375 		}
2376 		rcu_read_unlock();
2377 
2378 		if (!info)
2379 			continue;
2380 		down_read(&info->groups_sem);
2381 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2382 			if (!list_empty(&info->block_groups[c])) {
2383 				get_block_group_info(&info->block_groups[c],
2384 						     &space);
2385 				memcpy(dest, &space, sizeof(space));
2386 				dest++;
2387 				space_args.total_spaces++;
2388 				slot_count--;
2389 			}
2390 			if (!slot_count)
2391 				break;
2392 		}
2393 		up_read(&info->groups_sem);
2394 	}
2395 
2396 	user_dest = (struct btrfs_ioctl_space_info *)
2397 		(arg + sizeof(struct btrfs_ioctl_space_args));
2398 
2399 	if (copy_to_user(user_dest, dest_orig, alloc_size))
2400 		ret = -EFAULT;
2401 
2402 	kfree(dest_orig);
2403 out:
2404 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2405 		ret = -EFAULT;
2406 
2407 	return ret;
2408 }
2409 
2410 /*
2411  * there are many ways the trans_start and trans_end ioctls can lead
2412  * to deadlocks.  They should only be used by applications that
2413  * basically own the machine, and have a very in depth understanding
2414  * of all the possible deadlocks and enospc problems.
2415  */
2416 long btrfs_ioctl_trans_end(struct file *file)
2417 {
2418 	struct inode *inode = fdentry(file)->d_inode;
2419 	struct btrfs_root *root = BTRFS_I(inode)->root;
2420 	struct btrfs_trans_handle *trans;
2421 
2422 	trans = file->private_data;
2423 	if (!trans)
2424 		return -EINVAL;
2425 	file->private_data = NULL;
2426 
2427 	btrfs_end_transaction(trans, root);
2428 
2429 	mutex_lock(&root->fs_info->trans_mutex);
2430 	root->fs_info->open_ioctl_trans--;
2431 	mutex_unlock(&root->fs_info->trans_mutex);
2432 
2433 	mnt_drop_write(file->f_path.mnt);
2434 	return 0;
2435 }
2436 
2437 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2438 {
2439 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2440 	struct btrfs_trans_handle *trans;
2441 	u64 transid;
2442 	int ret;
2443 
2444 	trans = btrfs_start_transaction(root, 0);
2445 	if (IS_ERR(trans))
2446 		return PTR_ERR(trans);
2447 	transid = trans->transid;
2448 	ret = btrfs_commit_transaction_async(trans, root, 0);
2449 	if (ret) {
2450 		btrfs_end_transaction(trans, root);
2451 		return ret;
2452 	}
2453 
2454 	if (argp)
2455 		if (copy_to_user(argp, &transid, sizeof(transid)))
2456 			return -EFAULT;
2457 	return 0;
2458 }
2459 
2460 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2461 {
2462 	struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2463 	u64 transid;
2464 
2465 	if (argp) {
2466 		if (copy_from_user(&transid, argp, sizeof(transid)))
2467 			return -EFAULT;
2468 	} else {
2469 		transid = 0;  /* current trans */
2470 	}
2471 	return btrfs_wait_for_commit(root, transid);
2472 }
2473 
2474 long btrfs_ioctl(struct file *file, unsigned int
2475 		cmd, unsigned long arg)
2476 {
2477 	struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2478 	void __user *argp = (void __user *)arg;
2479 
2480 	switch (cmd) {
2481 	case FS_IOC_GETFLAGS:
2482 		return btrfs_ioctl_getflags(file, argp);
2483 	case FS_IOC_SETFLAGS:
2484 		return btrfs_ioctl_setflags(file, argp);
2485 	case FS_IOC_GETVERSION:
2486 		return btrfs_ioctl_getversion(file, argp);
2487 	case FITRIM:
2488 		return btrfs_ioctl_fitrim(file, argp);
2489 	case BTRFS_IOC_SNAP_CREATE:
2490 		return btrfs_ioctl_snap_create(file, argp, 0);
2491 	case BTRFS_IOC_SNAP_CREATE_V2:
2492 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
2493 	case BTRFS_IOC_SUBVOL_CREATE:
2494 		return btrfs_ioctl_snap_create(file, argp, 1);
2495 	case BTRFS_IOC_SNAP_DESTROY:
2496 		return btrfs_ioctl_snap_destroy(file, argp);
2497 	case BTRFS_IOC_SUBVOL_GETFLAGS:
2498 		return btrfs_ioctl_subvol_getflags(file, argp);
2499 	case BTRFS_IOC_SUBVOL_SETFLAGS:
2500 		return btrfs_ioctl_subvol_setflags(file, argp);
2501 	case BTRFS_IOC_DEFAULT_SUBVOL:
2502 		return btrfs_ioctl_default_subvol(file, argp);
2503 	case BTRFS_IOC_DEFRAG:
2504 		return btrfs_ioctl_defrag(file, NULL);
2505 	case BTRFS_IOC_DEFRAG_RANGE:
2506 		return btrfs_ioctl_defrag(file, argp);
2507 	case BTRFS_IOC_RESIZE:
2508 		return btrfs_ioctl_resize(root, argp);
2509 	case BTRFS_IOC_ADD_DEV:
2510 		return btrfs_ioctl_add_dev(root, argp);
2511 	case BTRFS_IOC_RM_DEV:
2512 		return btrfs_ioctl_rm_dev(root, argp);
2513 	case BTRFS_IOC_BALANCE:
2514 		return btrfs_balance(root->fs_info->dev_root);
2515 	case BTRFS_IOC_CLONE:
2516 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2517 	case BTRFS_IOC_CLONE_RANGE:
2518 		return btrfs_ioctl_clone_range(file, argp);
2519 	case BTRFS_IOC_TRANS_START:
2520 		return btrfs_ioctl_trans_start(file);
2521 	case BTRFS_IOC_TRANS_END:
2522 		return btrfs_ioctl_trans_end(file);
2523 	case BTRFS_IOC_TREE_SEARCH:
2524 		return btrfs_ioctl_tree_search(file, argp);
2525 	case BTRFS_IOC_INO_LOOKUP:
2526 		return btrfs_ioctl_ino_lookup(file, argp);
2527 	case BTRFS_IOC_SPACE_INFO:
2528 		return btrfs_ioctl_space_info(root, argp);
2529 	case BTRFS_IOC_SYNC:
2530 		btrfs_sync_fs(file->f_dentry->d_sb, 1);
2531 		return 0;
2532 	case BTRFS_IOC_START_SYNC:
2533 		return btrfs_ioctl_start_sync(file, argp);
2534 	case BTRFS_IOC_WAIT_SYNC:
2535 		return btrfs_ioctl_wait_sync(file, argp);
2536 	}
2537 
2538 	return -ENOTTY;
2539 }
2540