xref: /openbmc/linux/fs/btrfs/ioctl.c (revision af958a38)
1 /*
2  * Copyright (C) 2007 Oracle.  All rights reserved.
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public
6  * License v2 as published by the Free Software Foundation.
7  *
8  * This program is distributed in the hope that it will be useful,
9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
11  * General Public License for more details.
12  *
13  * You should have received a copy of the GNU General Public
14  * License along with this program; if not, write to the
15  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16  * Boston, MA 021110-1307, USA.
17  */
18 
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
23 #include <linux/fs.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
47 #include "ctree.h"
48 #include "disk-io.h"
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
52 #include "volumes.h"
53 #include "locking.h"
54 #include "inode-map.h"
55 #include "backref.h"
56 #include "rcu-string.h"
57 #include "send.h"
58 #include "dev-replace.h"
59 #include "props.h"
60 #include "sysfs.h"
61 #include "qgroup.h"
62 
63 #ifdef CONFIG_64BIT
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65  * structures are incorrect, as the timespec structure from userspace
66  * is 4 bytes too small. We define these alternatives here to teach
67  * the kernel about the 32-bit struct packing.
68  */
69 struct btrfs_ioctl_timespec_32 {
70 	__u64 sec;
71 	__u32 nsec;
72 } __attribute__ ((__packed__));
73 
74 struct btrfs_ioctl_received_subvol_args_32 {
75 	char	uuid[BTRFS_UUID_SIZE];	/* in */
76 	__u64	stransid;		/* in */
77 	__u64	rtransid;		/* out */
78 	struct btrfs_ioctl_timespec_32 stime; /* in */
79 	struct btrfs_ioctl_timespec_32 rtime; /* out */
80 	__u64	flags;			/* in */
81 	__u64	reserved[16];		/* in */
82 } __attribute__ ((__packed__));
83 
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 				struct btrfs_ioctl_received_subvol_args_32)
86 #endif
87 
88 
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 		       u64 off, u64 olen, u64 olen_aligned, u64 destoff);
91 
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
94 {
95 	if (S_ISDIR(mode))
96 		return flags;
97 	else if (S_ISREG(mode))
98 		return flags & ~FS_DIRSYNC_FL;
99 	else
100 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
101 }
102 
103 /*
104  * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105  */
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107 {
108 	unsigned int iflags = 0;
109 
110 	if (flags & BTRFS_INODE_SYNC)
111 		iflags |= FS_SYNC_FL;
112 	if (flags & BTRFS_INODE_IMMUTABLE)
113 		iflags |= FS_IMMUTABLE_FL;
114 	if (flags & BTRFS_INODE_APPEND)
115 		iflags |= FS_APPEND_FL;
116 	if (flags & BTRFS_INODE_NODUMP)
117 		iflags |= FS_NODUMP_FL;
118 	if (flags & BTRFS_INODE_NOATIME)
119 		iflags |= FS_NOATIME_FL;
120 	if (flags & BTRFS_INODE_DIRSYNC)
121 		iflags |= FS_DIRSYNC_FL;
122 	if (flags & BTRFS_INODE_NODATACOW)
123 		iflags |= FS_NOCOW_FL;
124 
125 	if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 		iflags |= FS_COMPR_FL;
127 	else if (flags & BTRFS_INODE_NOCOMPRESS)
128 		iflags |= FS_NOCOMP_FL;
129 
130 	return iflags;
131 }
132 
133 /*
134  * Update inode->i_flags based on the btrfs internal flags.
135  */
136 void btrfs_update_iflags(struct inode *inode)
137 {
138 	struct btrfs_inode *ip = BTRFS_I(inode);
139 	unsigned int new_fl = 0;
140 
141 	if (ip->flags & BTRFS_INODE_SYNC)
142 		new_fl |= S_SYNC;
143 	if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 		new_fl |= S_IMMUTABLE;
145 	if (ip->flags & BTRFS_INODE_APPEND)
146 		new_fl |= S_APPEND;
147 	if (ip->flags & BTRFS_INODE_NOATIME)
148 		new_fl |= S_NOATIME;
149 	if (ip->flags & BTRFS_INODE_DIRSYNC)
150 		new_fl |= S_DIRSYNC;
151 
152 	set_mask_bits(&inode->i_flags,
153 		      S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
154 		      new_fl);
155 }
156 
157 /*
158  * Inherit flags from the parent inode.
159  *
160  * Currently only the compression flags and the cow flags are inherited.
161  */
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
163 {
164 	unsigned int flags;
165 
166 	if (!dir)
167 		return;
168 
169 	flags = BTRFS_I(dir)->flags;
170 
171 	if (flags & BTRFS_INODE_NOCOMPRESS) {
172 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 	} else if (flags & BTRFS_INODE_COMPRESS) {
175 		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
177 	}
178 
179 	if (flags & BTRFS_INODE_NODATACOW) {
180 		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181 		if (S_ISREG(inode->i_mode))
182 			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
183 	}
184 
185 	btrfs_update_iflags(inode);
186 }
187 
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
189 {
190 	struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191 	unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
192 
193 	if (copy_to_user(arg, &flags, sizeof(flags)))
194 		return -EFAULT;
195 	return 0;
196 }
197 
198 static int check_flags(unsigned int flags)
199 {
200 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 		      FS_NOATIME_FL | FS_NODUMP_FL | \
202 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
203 		      FS_NOCOMP_FL | FS_COMPR_FL |
204 		      FS_NOCOW_FL))
205 		return -EOPNOTSUPP;
206 
207 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
208 		return -EINVAL;
209 
210 	return 0;
211 }
212 
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
214 {
215 	struct inode *inode = file_inode(file);
216 	struct btrfs_inode *ip = BTRFS_I(inode);
217 	struct btrfs_root *root = ip->root;
218 	struct btrfs_trans_handle *trans;
219 	unsigned int flags, oldflags;
220 	int ret;
221 	u64 ip_oldflags;
222 	unsigned int i_oldflags;
223 	umode_t mode;
224 
225 	if (!inode_owner_or_capable(inode))
226 		return -EPERM;
227 
228 	if (btrfs_root_readonly(root))
229 		return -EROFS;
230 
231 	if (copy_from_user(&flags, arg, sizeof(flags)))
232 		return -EFAULT;
233 
234 	ret = check_flags(flags);
235 	if (ret)
236 		return ret;
237 
238 	ret = mnt_want_write_file(file);
239 	if (ret)
240 		return ret;
241 
242 	mutex_lock(&inode->i_mutex);
243 
244 	ip_oldflags = ip->flags;
245 	i_oldflags = inode->i_flags;
246 	mode = inode->i_mode;
247 
248 	flags = btrfs_mask_flags(inode->i_mode, flags);
249 	oldflags = btrfs_flags_to_ioctl(ip->flags);
250 	if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 		if (!capable(CAP_LINUX_IMMUTABLE)) {
252 			ret = -EPERM;
253 			goto out_unlock;
254 		}
255 	}
256 
257 	if (flags & FS_SYNC_FL)
258 		ip->flags |= BTRFS_INODE_SYNC;
259 	else
260 		ip->flags &= ~BTRFS_INODE_SYNC;
261 	if (flags & FS_IMMUTABLE_FL)
262 		ip->flags |= BTRFS_INODE_IMMUTABLE;
263 	else
264 		ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 	if (flags & FS_APPEND_FL)
266 		ip->flags |= BTRFS_INODE_APPEND;
267 	else
268 		ip->flags &= ~BTRFS_INODE_APPEND;
269 	if (flags & FS_NODUMP_FL)
270 		ip->flags |= BTRFS_INODE_NODUMP;
271 	else
272 		ip->flags &= ~BTRFS_INODE_NODUMP;
273 	if (flags & FS_NOATIME_FL)
274 		ip->flags |= BTRFS_INODE_NOATIME;
275 	else
276 		ip->flags &= ~BTRFS_INODE_NOATIME;
277 	if (flags & FS_DIRSYNC_FL)
278 		ip->flags |= BTRFS_INODE_DIRSYNC;
279 	else
280 		ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 	if (flags & FS_NOCOW_FL) {
282 		if (S_ISREG(mode)) {
283 			/*
284 			 * It's safe to turn csums off here, no extents exist.
285 			 * Otherwise we want the flag to reflect the real COW
286 			 * status of the file and will not set it.
287 			 */
288 			if (inode->i_size == 0)
289 				ip->flags |= BTRFS_INODE_NODATACOW
290 					   | BTRFS_INODE_NODATASUM;
291 		} else {
292 			ip->flags |= BTRFS_INODE_NODATACOW;
293 		}
294 	} else {
295 		/*
296 		 * Revert back under same assuptions as above
297 		 */
298 		if (S_ISREG(mode)) {
299 			if (inode->i_size == 0)
300 				ip->flags &= ~(BTRFS_INODE_NODATACOW
301 				             | BTRFS_INODE_NODATASUM);
302 		} else {
303 			ip->flags &= ~BTRFS_INODE_NODATACOW;
304 		}
305 	}
306 
307 	/*
308 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 	 * flag may be changed automatically if compression code won't make
310 	 * things smaller.
311 	 */
312 	if (flags & FS_NOCOMP_FL) {
313 		ip->flags &= ~BTRFS_INODE_COMPRESS;
314 		ip->flags |= BTRFS_INODE_NOCOMPRESS;
315 
316 		ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 		if (ret && ret != -ENODATA)
318 			goto out_drop;
319 	} else if (flags & FS_COMPR_FL) {
320 		const char *comp;
321 
322 		ip->flags |= BTRFS_INODE_COMPRESS;
323 		ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
324 
325 		if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326 			comp = "lzo";
327 		else
328 			comp = "zlib";
329 		ret = btrfs_set_prop(inode, "btrfs.compression",
330 				     comp, strlen(comp), 0);
331 		if (ret)
332 			goto out_drop;
333 
334 	} else {
335 		ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
336 	}
337 
338 	trans = btrfs_start_transaction(root, 1);
339 	if (IS_ERR(trans)) {
340 		ret = PTR_ERR(trans);
341 		goto out_drop;
342 	}
343 
344 	btrfs_update_iflags(inode);
345 	inode_inc_iversion(inode);
346 	inode->i_ctime = CURRENT_TIME;
347 	ret = btrfs_update_inode(trans, root, inode);
348 
349 	btrfs_end_transaction(trans, root);
350  out_drop:
351 	if (ret) {
352 		ip->flags = ip_oldflags;
353 		inode->i_flags = i_oldflags;
354 	}
355 
356  out_unlock:
357 	mutex_unlock(&inode->i_mutex);
358 	mnt_drop_write_file(file);
359 	return ret;
360 }
361 
362 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
363 {
364 	struct inode *inode = file_inode(file);
365 
366 	return put_user(inode->i_generation, arg);
367 }
368 
369 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
370 {
371 	struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
372 	struct btrfs_device *device;
373 	struct request_queue *q;
374 	struct fstrim_range range;
375 	u64 minlen = ULLONG_MAX;
376 	u64 num_devices = 0;
377 	u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
378 	int ret;
379 
380 	if (!capable(CAP_SYS_ADMIN))
381 		return -EPERM;
382 
383 	rcu_read_lock();
384 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
385 				dev_list) {
386 		if (!device->bdev)
387 			continue;
388 		q = bdev_get_queue(device->bdev);
389 		if (blk_queue_discard(q)) {
390 			num_devices++;
391 			minlen = min((u64)q->limits.discard_granularity,
392 				     minlen);
393 		}
394 	}
395 	rcu_read_unlock();
396 
397 	if (!num_devices)
398 		return -EOPNOTSUPP;
399 	if (copy_from_user(&range, arg, sizeof(range)))
400 		return -EFAULT;
401 	if (range.start > total_bytes ||
402 	    range.len < fs_info->sb->s_blocksize)
403 		return -EINVAL;
404 
405 	range.len = min(range.len, total_bytes - range.start);
406 	range.minlen = max(range.minlen, minlen);
407 	ret = btrfs_trim_fs(fs_info->tree_root, &range);
408 	if (ret < 0)
409 		return ret;
410 
411 	if (copy_to_user(arg, &range, sizeof(range)))
412 		return -EFAULT;
413 
414 	return 0;
415 }
416 
417 int btrfs_is_empty_uuid(u8 *uuid)
418 {
419 	int i;
420 
421 	for (i = 0; i < BTRFS_UUID_SIZE; i++) {
422 		if (uuid[i])
423 			return 0;
424 	}
425 	return 1;
426 }
427 
428 static noinline int create_subvol(struct inode *dir,
429 				  struct dentry *dentry,
430 				  char *name, int namelen,
431 				  u64 *async_transid,
432 				  struct btrfs_qgroup_inherit *inherit)
433 {
434 	struct btrfs_trans_handle *trans;
435 	struct btrfs_key key;
436 	struct btrfs_root_item root_item;
437 	struct btrfs_inode_item *inode_item;
438 	struct extent_buffer *leaf;
439 	struct btrfs_root *root = BTRFS_I(dir)->root;
440 	struct btrfs_root *new_root;
441 	struct btrfs_block_rsv block_rsv;
442 	struct timespec cur_time = CURRENT_TIME;
443 	struct inode *inode;
444 	int ret;
445 	int err;
446 	u64 objectid;
447 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
448 	u64 index = 0;
449 	u64 qgroup_reserved;
450 	uuid_le new_uuid;
451 
452 	ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
453 	if (ret)
454 		return ret;
455 
456 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
457 	/*
458 	 * The same as the snapshot creation, please see the comment
459 	 * of create_snapshot().
460 	 */
461 	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
462 					       8, &qgroup_reserved, false);
463 	if (ret)
464 		return ret;
465 
466 	trans = btrfs_start_transaction(root, 0);
467 	if (IS_ERR(trans)) {
468 		ret = PTR_ERR(trans);
469 		btrfs_subvolume_release_metadata(root, &block_rsv,
470 						 qgroup_reserved);
471 		return ret;
472 	}
473 	trans->block_rsv = &block_rsv;
474 	trans->bytes_reserved = block_rsv.size;
475 
476 	ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
477 	if (ret)
478 		goto fail;
479 
480 	leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
481 				      0, objectid, NULL, 0, 0, 0);
482 	if (IS_ERR(leaf)) {
483 		ret = PTR_ERR(leaf);
484 		goto fail;
485 	}
486 
487 	memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
488 	btrfs_set_header_bytenr(leaf, leaf->start);
489 	btrfs_set_header_generation(leaf, trans->transid);
490 	btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
491 	btrfs_set_header_owner(leaf, objectid);
492 
493 	write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
494 			    BTRFS_FSID_SIZE);
495 	write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
496 			    btrfs_header_chunk_tree_uuid(leaf),
497 			    BTRFS_UUID_SIZE);
498 	btrfs_mark_buffer_dirty(leaf);
499 
500 	memset(&root_item, 0, sizeof(root_item));
501 
502 	inode_item = &root_item.inode;
503 	btrfs_set_stack_inode_generation(inode_item, 1);
504 	btrfs_set_stack_inode_size(inode_item, 3);
505 	btrfs_set_stack_inode_nlink(inode_item, 1);
506 	btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
507 	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
508 
509 	btrfs_set_root_flags(&root_item, 0);
510 	btrfs_set_root_limit(&root_item, 0);
511 	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
512 
513 	btrfs_set_root_bytenr(&root_item, leaf->start);
514 	btrfs_set_root_generation(&root_item, trans->transid);
515 	btrfs_set_root_level(&root_item, 0);
516 	btrfs_set_root_refs(&root_item, 1);
517 	btrfs_set_root_used(&root_item, leaf->len);
518 	btrfs_set_root_last_snapshot(&root_item, 0);
519 
520 	btrfs_set_root_generation_v2(&root_item,
521 			btrfs_root_generation(&root_item));
522 	uuid_le_gen(&new_uuid);
523 	memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
524 	btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
525 	btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
526 	root_item.ctime = root_item.otime;
527 	btrfs_set_root_ctransid(&root_item, trans->transid);
528 	btrfs_set_root_otransid(&root_item, trans->transid);
529 
530 	btrfs_tree_unlock(leaf);
531 	free_extent_buffer(leaf);
532 	leaf = NULL;
533 
534 	btrfs_set_root_dirid(&root_item, new_dirid);
535 
536 	key.objectid = objectid;
537 	key.offset = 0;
538 	btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
539 	ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
540 				&root_item);
541 	if (ret)
542 		goto fail;
543 
544 	key.offset = (u64)-1;
545 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
546 	if (IS_ERR(new_root)) {
547 		btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
548 		ret = PTR_ERR(new_root);
549 		goto fail;
550 	}
551 
552 	btrfs_record_root_in_trans(trans, new_root);
553 
554 	ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
555 	if (ret) {
556 		/* We potentially lose an unused inode item here */
557 		btrfs_abort_transaction(trans, root, ret);
558 		goto fail;
559 	}
560 
561 	/*
562 	 * insert the directory item
563 	 */
564 	ret = btrfs_set_inode_index(dir, &index);
565 	if (ret) {
566 		btrfs_abort_transaction(trans, root, ret);
567 		goto fail;
568 	}
569 
570 	ret = btrfs_insert_dir_item(trans, root,
571 				    name, namelen, dir, &key,
572 				    BTRFS_FT_DIR, index);
573 	if (ret) {
574 		btrfs_abort_transaction(trans, root, ret);
575 		goto fail;
576 	}
577 
578 	btrfs_i_size_write(dir, dir->i_size + namelen * 2);
579 	ret = btrfs_update_inode(trans, root, dir);
580 	BUG_ON(ret);
581 
582 	ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
583 				 objectid, root->root_key.objectid,
584 				 btrfs_ino(dir), index, name, namelen);
585 	BUG_ON(ret);
586 
587 	ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
588 				  root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
589 				  objectid);
590 	if (ret)
591 		btrfs_abort_transaction(trans, root, ret);
592 
593 fail:
594 	trans->block_rsv = NULL;
595 	trans->bytes_reserved = 0;
596 	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
597 
598 	if (async_transid) {
599 		*async_transid = trans->transid;
600 		err = btrfs_commit_transaction_async(trans, root, 1);
601 		if (err)
602 			err = btrfs_commit_transaction(trans, root);
603 	} else {
604 		err = btrfs_commit_transaction(trans, root);
605 	}
606 	if (err && !ret)
607 		ret = err;
608 
609 	if (!ret) {
610 		inode = btrfs_lookup_dentry(dir, dentry);
611 		if (IS_ERR(inode))
612 			return PTR_ERR(inode);
613 		d_instantiate(dentry, inode);
614 	}
615 	return ret;
616 }
617 
618 static void btrfs_wait_nocow_write(struct btrfs_root *root)
619 {
620 	s64 writers;
621 	DEFINE_WAIT(wait);
622 
623 	do {
624 		prepare_to_wait(&root->subv_writers->wait, &wait,
625 				TASK_UNINTERRUPTIBLE);
626 
627 		writers = percpu_counter_sum(&root->subv_writers->counter);
628 		if (writers)
629 			schedule();
630 
631 		finish_wait(&root->subv_writers->wait, &wait);
632 	} while (writers);
633 }
634 
635 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
636 			   struct dentry *dentry, char *name, int namelen,
637 			   u64 *async_transid, bool readonly,
638 			   struct btrfs_qgroup_inherit *inherit)
639 {
640 	struct inode *inode;
641 	struct btrfs_pending_snapshot *pending_snapshot;
642 	struct btrfs_trans_handle *trans;
643 	int ret;
644 
645 	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
646 		return -EINVAL;
647 
648 	atomic_inc(&root->will_be_snapshoted);
649 	smp_mb__after_atomic();
650 	btrfs_wait_nocow_write(root);
651 
652 	ret = btrfs_start_delalloc_inodes(root, 0);
653 	if (ret)
654 		goto out;
655 
656 	btrfs_wait_ordered_extents(root, -1);
657 
658 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
659 	if (!pending_snapshot) {
660 		ret = -ENOMEM;
661 		goto out;
662 	}
663 
664 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
665 			     BTRFS_BLOCK_RSV_TEMP);
666 	/*
667 	 * 1 - parent dir inode
668 	 * 2 - dir entries
669 	 * 1 - root item
670 	 * 2 - root ref/backref
671 	 * 1 - root of snapshot
672 	 * 1 - UUID item
673 	 */
674 	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
675 					&pending_snapshot->block_rsv, 8,
676 					&pending_snapshot->qgroup_reserved,
677 					false);
678 	if (ret)
679 		goto free;
680 
681 	pending_snapshot->dentry = dentry;
682 	pending_snapshot->root = root;
683 	pending_snapshot->readonly = readonly;
684 	pending_snapshot->dir = dir;
685 	pending_snapshot->inherit = inherit;
686 
687 	trans = btrfs_start_transaction(root, 0);
688 	if (IS_ERR(trans)) {
689 		ret = PTR_ERR(trans);
690 		goto fail;
691 	}
692 
693 	spin_lock(&root->fs_info->trans_lock);
694 	list_add(&pending_snapshot->list,
695 		 &trans->transaction->pending_snapshots);
696 	spin_unlock(&root->fs_info->trans_lock);
697 	if (async_transid) {
698 		*async_transid = trans->transid;
699 		ret = btrfs_commit_transaction_async(trans,
700 				     root->fs_info->extent_root, 1);
701 		if (ret)
702 			ret = btrfs_commit_transaction(trans, root);
703 	} else {
704 		ret = btrfs_commit_transaction(trans,
705 					       root->fs_info->extent_root);
706 	}
707 	if (ret)
708 		goto fail;
709 
710 	ret = pending_snapshot->error;
711 	if (ret)
712 		goto fail;
713 
714 	inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
715 	if (IS_ERR(inode)) {
716 		ret = PTR_ERR(inode);
717 		goto fail;
718 	}
719 
720 	d_instantiate(dentry, inode);
721 	ret = 0;
722 fail:
723 	btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
724 					 &pending_snapshot->block_rsv,
725 					 pending_snapshot->qgroup_reserved);
726 free:
727 	kfree(pending_snapshot);
728 out:
729 	atomic_dec(&root->will_be_snapshoted);
730 	return ret;
731 }
732 
733 /*  copy of check_sticky in fs/namei.c()
734 * It's inline, so penalty for filesystems that don't use sticky bit is
735 * minimal.
736 */
737 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
738 {
739 	kuid_t fsuid = current_fsuid();
740 
741 	if (!(dir->i_mode & S_ISVTX))
742 		return 0;
743 	if (uid_eq(inode->i_uid, fsuid))
744 		return 0;
745 	if (uid_eq(dir->i_uid, fsuid))
746 		return 0;
747 	return !capable(CAP_FOWNER);
748 }
749 
750 /*  copy of may_delete in fs/namei.c()
751  *	Check whether we can remove a link victim from directory dir, check
752  *  whether the type of victim is right.
753  *  1. We can't do it if dir is read-only (done in permission())
754  *  2. We should have write and exec permissions on dir
755  *  3. We can't remove anything from append-only dir
756  *  4. We can't do anything with immutable dir (done in permission())
757  *  5. If the sticky bit on dir is set we should either
758  *	a. be owner of dir, or
759  *	b. be owner of victim, or
760  *	c. have CAP_FOWNER capability
761  *  6. If the victim is append-only or immutable we can't do antyhing with
762  *     links pointing to it.
763  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
764  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
765  *  9. We can't remove a root or mountpoint.
766  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
767  *     nfs_async_unlink().
768  */
769 
770 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
771 {
772 	int error;
773 
774 	if (!victim->d_inode)
775 		return -ENOENT;
776 
777 	BUG_ON(victim->d_parent->d_inode != dir);
778 	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
779 
780 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
781 	if (error)
782 		return error;
783 	if (IS_APPEND(dir))
784 		return -EPERM;
785 	if (btrfs_check_sticky(dir, victim->d_inode)||
786 		IS_APPEND(victim->d_inode)||
787 	    IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
788 		return -EPERM;
789 	if (isdir) {
790 		if (!S_ISDIR(victim->d_inode->i_mode))
791 			return -ENOTDIR;
792 		if (IS_ROOT(victim))
793 			return -EBUSY;
794 	} else if (S_ISDIR(victim->d_inode->i_mode))
795 		return -EISDIR;
796 	if (IS_DEADDIR(dir))
797 		return -ENOENT;
798 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
799 		return -EBUSY;
800 	return 0;
801 }
802 
803 /* copy of may_create in fs/namei.c() */
804 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
805 {
806 	if (child->d_inode)
807 		return -EEXIST;
808 	if (IS_DEADDIR(dir))
809 		return -ENOENT;
810 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
811 }
812 
813 /*
814  * Create a new subvolume below @parent.  This is largely modeled after
815  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816  * inside this filesystem so it's quite a bit simpler.
817  */
818 static noinline int btrfs_mksubvol(struct path *parent,
819 				   char *name, int namelen,
820 				   struct btrfs_root *snap_src,
821 				   u64 *async_transid, bool readonly,
822 				   struct btrfs_qgroup_inherit *inherit)
823 {
824 	struct inode *dir  = parent->dentry->d_inode;
825 	struct dentry *dentry;
826 	int error;
827 
828 	error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
829 	if (error == -EINTR)
830 		return error;
831 
832 	dentry = lookup_one_len(name, parent->dentry, namelen);
833 	error = PTR_ERR(dentry);
834 	if (IS_ERR(dentry))
835 		goto out_unlock;
836 
837 	error = -EEXIST;
838 	if (dentry->d_inode)
839 		goto out_dput;
840 
841 	error = btrfs_may_create(dir, dentry);
842 	if (error)
843 		goto out_dput;
844 
845 	/*
846 	 * even if this name doesn't exist, we may get hash collisions.
847 	 * check for them now when we can safely fail
848 	 */
849 	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
850 					       dir->i_ino, name,
851 					       namelen);
852 	if (error)
853 		goto out_dput;
854 
855 	down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
856 
857 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
858 		goto out_up_read;
859 
860 	if (snap_src) {
861 		error = create_snapshot(snap_src, dir, dentry, name, namelen,
862 					async_transid, readonly, inherit);
863 	} else {
864 		error = create_subvol(dir, dentry, name, namelen,
865 				      async_transid, inherit);
866 	}
867 	if (!error)
868 		fsnotify_mkdir(dir, dentry);
869 out_up_read:
870 	up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
871 out_dput:
872 	dput(dentry);
873 out_unlock:
874 	mutex_unlock(&dir->i_mutex);
875 	return error;
876 }
877 
878 /*
879  * When we're defragging a range, we don't want to kick it off again
880  * if it is really just waiting for delalloc to send it down.
881  * If we find a nice big extent or delalloc range for the bytes in the
882  * file you want to defrag, we return 0 to let you know to skip this
883  * part of the file
884  */
885 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
886 {
887 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
888 	struct extent_map *em = NULL;
889 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
890 	u64 end;
891 
892 	read_lock(&em_tree->lock);
893 	em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
894 	read_unlock(&em_tree->lock);
895 
896 	if (em) {
897 		end = extent_map_end(em);
898 		free_extent_map(em);
899 		if (end - offset > thresh)
900 			return 0;
901 	}
902 	/* if we already have a nice delalloc here, just stop */
903 	thresh /= 2;
904 	end = count_range_bits(io_tree, &offset, offset + thresh,
905 			       thresh, EXTENT_DELALLOC, 1);
906 	if (end >= thresh)
907 		return 0;
908 	return 1;
909 }
910 
911 /*
912  * helper function to walk through a file and find extents
913  * newer than a specific transid, and smaller than thresh.
914  *
915  * This is used by the defragging code to find new and small
916  * extents
917  */
918 static int find_new_extents(struct btrfs_root *root,
919 			    struct inode *inode, u64 newer_than,
920 			    u64 *off, int thresh)
921 {
922 	struct btrfs_path *path;
923 	struct btrfs_key min_key;
924 	struct extent_buffer *leaf;
925 	struct btrfs_file_extent_item *extent;
926 	int type;
927 	int ret;
928 	u64 ino = btrfs_ino(inode);
929 
930 	path = btrfs_alloc_path();
931 	if (!path)
932 		return -ENOMEM;
933 
934 	min_key.objectid = ino;
935 	min_key.type = BTRFS_EXTENT_DATA_KEY;
936 	min_key.offset = *off;
937 
938 	while (1) {
939 		path->keep_locks = 1;
940 		ret = btrfs_search_forward(root, &min_key, path, newer_than);
941 		if (ret != 0)
942 			goto none;
943 		path->keep_locks = 0;
944 		btrfs_unlock_up_safe(path, 1);
945 process_slot:
946 		if (min_key.objectid != ino)
947 			goto none;
948 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
949 			goto none;
950 
951 		leaf = path->nodes[0];
952 		extent = btrfs_item_ptr(leaf, path->slots[0],
953 					struct btrfs_file_extent_item);
954 
955 		type = btrfs_file_extent_type(leaf, extent);
956 		if (type == BTRFS_FILE_EXTENT_REG &&
957 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
958 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
959 			*off = min_key.offset;
960 			btrfs_free_path(path);
961 			return 0;
962 		}
963 
964 		path->slots[0]++;
965 		if (path->slots[0] < btrfs_header_nritems(leaf)) {
966 			btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
967 			goto process_slot;
968 		}
969 
970 		if (min_key.offset == (u64)-1)
971 			goto none;
972 
973 		min_key.offset++;
974 		btrfs_release_path(path);
975 	}
976 none:
977 	btrfs_free_path(path);
978 	return -ENOENT;
979 }
980 
981 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
982 {
983 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
984 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
985 	struct extent_map *em;
986 	u64 len = PAGE_CACHE_SIZE;
987 
988 	/*
989 	 * hopefully we have this extent in the tree already, try without
990 	 * the full extent lock
991 	 */
992 	read_lock(&em_tree->lock);
993 	em = lookup_extent_mapping(em_tree, start, len);
994 	read_unlock(&em_tree->lock);
995 
996 	if (!em) {
997 		struct extent_state *cached = NULL;
998 		u64 end = start + len - 1;
999 
1000 		/* get the big lock and read metadata off disk */
1001 		lock_extent_bits(io_tree, start, end, 0, &cached);
1002 		em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1003 		unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1004 
1005 		if (IS_ERR(em))
1006 			return NULL;
1007 	}
1008 
1009 	return em;
1010 }
1011 
1012 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1013 {
1014 	struct extent_map *next;
1015 	bool ret = true;
1016 
1017 	/* this is the last extent */
1018 	if (em->start + em->len >= i_size_read(inode))
1019 		return false;
1020 
1021 	next = defrag_lookup_extent(inode, em->start + em->len);
1022 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1023 		ret = false;
1024 	else if ((em->block_start + em->block_len == next->block_start) &&
1025 		 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1026 		ret = false;
1027 
1028 	free_extent_map(next);
1029 	return ret;
1030 }
1031 
1032 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
1033 			       u64 *last_len, u64 *skip, u64 *defrag_end,
1034 			       int compress)
1035 {
1036 	struct extent_map *em;
1037 	int ret = 1;
1038 	bool next_mergeable = true;
1039 
1040 	/*
1041 	 * make sure that once we start defragging an extent, we keep on
1042 	 * defragging it
1043 	 */
1044 	if (start < *defrag_end)
1045 		return 1;
1046 
1047 	*skip = 0;
1048 
1049 	em = defrag_lookup_extent(inode, start);
1050 	if (!em)
1051 		return 0;
1052 
1053 	/* this will cover holes, and inline extents */
1054 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1055 		ret = 0;
1056 		goto out;
1057 	}
1058 
1059 	next_mergeable = defrag_check_next_extent(inode, em);
1060 	/*
1061 	 * we hit a real extent, if it is big or the next extent is not a
1062 	 * real extent, don't bother defragging it
1063 	 */
1064 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1065 	    (em->len >= thresh || !next_mergeable))
1066 		ret = 0;
1067 out:
1068 	/*
1069 	 * last_len ends up being a counter of how many bytes we've defragged.
1070 	 * every time we choose not to defrag an extent, we reset *last_len
1071 	 * so that the next tiny extent will force a defrag.
1072 	 *
1073 	 * The end result of this is that tiny extents before a single big
1074 	 * extent will force at least part of that big extent to be defragged.
1075 	 */
1076 	if (ret) {
1077 		*defrag_end = extent_map_end(em);
1078 	} else {
1079 		*last_len = 0;
1080 		*skip = extent_map_end(em);
1081 		*defrag_end = 0;
1082 	}
1083 
1084 	free_extent_map(em);
1085 	return ret;
1086 }
1087 
1088 /*
1089  * it doesn't do much good to defrag one or two pages
1090  * at a time.  This pulls in a nice chunk of pages
1091  * to COW and defrag.
1092  *
1093  * It also makes sure the delalloc code has enough
1094  * dirty data to avoid making new small extents as part
1095  * of the defrag
1096  *
1097  * It's a good idea to start RA on this range
1098  * before calling this.
1099  */
1100 static int cluster_pages_for_defrag(struct inode *inode,
1101 				    struct page **pages,
1102 				    unsigned long start_index,
1103 				    unsigned long num_pages)
1104 {
1105 	unsigned long file_end;
1106 	u64 isize = i_size_read(inode);
1107 	u64 page_start;
1108 	u64 page_end;
1109 	u64 page_cnt;
1110 	int ret;
1111 	int i;
1112 	int i_done;
1113 	struct btrfs_ordered_extent *ordered;
1114 	struct extent_state *cached_state = NULL;
1115 	struct extent_io_tree *tree;
1116 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1117 
1118 	file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1119 	if (!isize || start_index > file_end)
1120 		return 0;
1121 
1122 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1123 
1124 	ret = btrfs_delalloc_reserve_space(inode,
1125 					   page_cnt << PAGE_CACHE_SHIFT);
1126 	if (ret)
1127 		return ret;
1128 	i_done = 0;
1129 	tree = &BTRFS_I(inode)->io_tree;
1130 
1131 	/* step one, lock all the pages */
1132 	for (i = 0; i < page_cnt; i++) {
1133 		struct page *page;
1134 again:
1135 		page = find_or_create_page(inode->i_mapping,
1136 					   start_index + i, mask);
1137 		if (!page)
1138 			break;
1139 
1140 		page_start = page_offset(page);
1141 		page_end = page_start + PAGE_CACHE_SIZE - 1;
1142 		while (1) {
1143 			lock_extent_bits(tree, page_start, page_end,
1144 					 0, &cached_state);
1145 			ordered = btrfs_lookup_ordered_extent(inode,
1146 							      page_start);
1147 			unlock_extent_cached(tree, page_start, page_end,
1148 					     &cached_state, GFP_NOFS);
1149 			if (!ordered)
1150 				break;
1151 
1152 			unlock_page(page);
1153 			btrfs_start_ordered_extent(inode, ordered, 1);
1154 			btrfs_put_ordered_extent(ordered);
1155 			lock_page(page);
1156 			/*
1157 			 * we unlocked the page above, so we need check if
1158 			 * it was released or not.
1159 			 */
1160 			if (page->mapping != inode->i_mapping) {
1161 				unlock_page(page);
1162 				page_cache_release(page);
1163 				goto again;
1164 			}
1165 		}
1166 
1167 		if (!PageUptodate(page)) {
1168 			btrfs_readpage(NULL, page);
1169 			lock_page(page);
1170 			if (!PageUptodate(page)) {
1171 				unlock_page(page);
1172 				page_cache_release(page);
1173 				ret = -EIO;
1174 				break;
1175 			}
1176 		}
1177 
1178 		if (page->mapping != inode->i_mapping) {
1179 			unlock_page(page);
1180 			page_cache_release(page);
1181 			goto again;
1182 		}
1183 
1184 		pages[i] = page;
1185 		i_done++;
1186 	}
1187 	if (!i_done || ret)
1188 		goto out;
1189 
1190 	if (!(inode->i_sb->s_flags & MS_ACTIVE))
1191 		goto out;
1192 
1193 	/*
1194 	 * so now we have a nice long stream of locked
1195 	 * and up to date pages, lets wait on them
1196 	 */
1197 	for (i = 0; i < i_done; i++)
1198 		wait_on_page_writeback(pages[i]);
1199 
1200 	page_start = page_offset(pages[0]);
1201 	page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1202 
1203 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1204 			 page_start, page_end - 1, 0, &cached_state);
1205 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1206 			  page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1207 			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1208 			  &cached_state, GFP_NOFS);
1209 
1210 	if (i_done != page_cnt) {
1211 		spin_lock(&BTRFS_I(inode)->lock);
1212 		BTRFS_I(inode)->outstanding_extents++;
1213 		spin_unlock(&BTRFS_I(inode)->lock);
1214 		btrfs_delalloc_release_space(inode,
1215 				     (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1216 	}
1217 
1218 
1219 	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1220 			  &cached_state, GFP_NOFS);
1221 
1222 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1223 			     page_start, page_end - 1, &cached_state,
1224 			     GFP_NOFS);
1225 
1226 	for (i = 0; i < i_done; i++) {
1227 		clear_page_dirty_for_io(pages[i]);
1228 		ClearPageChecked(pages[i]);
1229 		set_page_extent_mapped(pages[i]);
1230 		set_page_dirty(pages[i]);
1231 		unlock_page(pages[i]);
1232 		page_cache_release(pages[i]);
1233 	}
1234 	return i_done;
1235 out:
1236 	for (i = 0; i < i_done; i++) {
1237 		unlock_page(pages[i]);
1238 		page_cache_release(pages[i]);
1239 	}
1240 	btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1241 	return ret;
1242 
1243 }
1244 
1245 int btrfs_defrag_file(struct inode *inode, struct file *file,
1246 		      struct btrfs_ioctl_defrag_range_args *range,
1247 		      u64 newer_than, unsigned long max_to_defrag)
1248 {
1249 	struct btrfs_root *root = BTRFS_I(inode)->root;
1250 	struct file_ra_state *ra = NULL;
1251 	unsigned long last_index;
1252 	u64 isize = i_size_read(inode);
1253 	u64 last_len = 0;
1254 	u64 skip = 0;
1255 	u64 defrag_end = 0;
1256 	u64 newer_off = range->start;
1257 	unsigned long i;
1258 	unsigned long ra_index = 0;
1259 	int ret;
1260 	int defrag_count = 0;
1261 	int compress_type = BTRFS_COMPRESS_ZLIB;
1262 	int extent_thresh = range->extent_thresh;
1263 	unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1264 	unsigned long cluster = max_cluster;
1265 	u64 new_align = ~((u64)128 * 1024 - 1);
1266 	struct page **pages = NULL;
1267 
1268 	if (isize == 0)
1269 		return 0;
1270 
1271 	if (range->start >= isize)
1272 		return -EINVAL;
1273 
1274 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1275 		if (range->compress_type > BTRFS_COMPRESS_TYPES)
1276 			return -EINVAL;
1277 		if (range->compress_type)
1278 			compress_type = range->compress_type;
1279 	}
1280 
1281 	if (extent_thresh == 0)
1282 		extent_thresh = 256 * 1024;
1283 
1284 	/*
1285 	 * if we were not given a file, allocate a readahead
1286 	 * context
1287 	 */
1288 	if (!file) {
1289 		ra = kzalloc(sizeof(*ra), GFP_NOFS);
1290 		if (!ra)
1291 			return -ENOMEM;
1292 		file_ra_state_init(ra, inode->i_mapping);
1293 	} else {
1294 		ra = &file->f_ra;
1295 	}
1296 
1297 	pages = kmalloc_array(max_cluster, sizeof(struct page *),
1298 			GFP_NOFS);
1299 	if (!pages) {
1300 		ret = -ENOMEM;
1301 		goto out_ra;
1302 	}
1303 
1304 	/* find the last page to defrag */
1305 	if (range->start + range->len > range->start) {
1306 		last_index = min_t(u64, isize - 1,
1307 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1308 	} else {
1309 		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1310 	}
1311 
1312 	if (newer_than) {
1313 		ret = find_new_extents(root, inode, newer_than,
1314 				       &newer_off, 64 * 1024);
1315 		if (!ret) {
1316 			range->start = newer_off;
1317 			/*
1318 			 * we always align our defrag to help keep
1319 			 * the extents in the file evenly spaced
1320 			 */
1321 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1322 		} else
1323 			goto out_ra;
1324 	} else {
1325 		i = range->start >> PAGE_CACHE_SHIFT;
1326 	}
1327 	if (!max_to_defrag)
1328 		max_to_defrag = last_index + 1;
1329 
1330 	/*
1331 	 * make writeback starts from i, so the defrag range can be
1332 	 * written sequentially.
1333 	 */
1334 	if (i < inode->i_mapping->writeback_index)
1335 		inode->i_mapping->writeback_index = i;
1336 
1337 	while (i <= last_index && defrag_count < max_to_defrag &&
1338 	       (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1339 		PAGE_CACHE_SHIFT)) {
1340 		/*
1341 		 * make sure we stop running if someone unmounts
1342 		 * the FS
1343 		 */
1344 		if (!(inode->i_sb->s_flags & MS_ACTIVE))
1345 			break;
1346 
1347 		if (btrfs_defrag_cancelled(root->fs_info)) {
1348 			printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1349 			ret = -EAGAIN;
1350 			break;
1351 		}
1352 
1353 		if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1354 					 extent_thresh, &last_len, &skip,
1355 					 &defrag_end, range->flags &
1356 					 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1357 			unsigned long next;
1358 			/*
1359 			 * the should_defrag function tells us how much to skip
1360 			 * bump our counter by the suggested amount
1361 			 */
1362 			next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1363 			i = max(i + 1, next);
1364 			continue;
1365 		}
1366 
1367 		if (!newer_than) {
1368 			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1369 				   PAGE_CACHE_SHIFT) - i;
1370 			cluster = min(cluster, max_cluster);
1371 		} else {
1372 			cluster = max_cluster;
1373 		}
1374 
1375 		if (i + cluster > ra_index) {
1376 			ra_index = max(i, ra_index);
1377 			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1378 				       cluster);
1379 			ra_index += max_cluster;
1380 		}
1381 
1382 		mutex_lock(&inode->i_mutex);
1383 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1384 			BTRFS_I(inode)->force_compress = compress_type;
1385 		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1386 		if (ret < 0) {
1387 			mutex_unlock(&inode->i_mutex);
1388 			goto out_ra;
1389 		}
1390 
1391 		defrag_count += ret;
1392 		balance_dirty_pages_ratelimited(inode->i_mapping);
1393 		mutex_unlock(&inode->i_mutex);
1394 
1395 		if (newer_than) {
1396 			if (newer_off == (u64)-1)
1397 				break;
1398 
1399 			if (ret > 0)
1400 				i += ret;
1401 
1402 			newer_off = max(newer_off + 1,
1403 					(u64)i << PAGE_CACHE_SHIFT);
1404 
1405 			ret = find_new_extents(root, inode,
1406 					       newer_than, &newer_off,
1407 					       64 * 1024);
1408 			if (!ret) {
1409 				range->start = newer_off;
1410 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1411 			} else {
1412 				break;
1413 			}
1414 		} else {
1415 			if (ret > 0) {
1416 				i += ret;
1417 				last_len += ret << PAGE_CACHE_SHIFT;
1418 			} else {
1419 				i++;
1420 				last_len = 0;
1421 			}
1422 		}
1423 	}
1424 
1425 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1426 		filemap_flush(inode->i_mapping);
1427 		if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1428 			     &BTRFS_I(inode)->runtime_flags))
1429 			filemap_flush(inode->i_mapping);
1430 	}
1431 
1432 	if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1433 		/* the filemap_flush will queue IO into the worker threads, but
1434 		 * we have to make sure the IO is actually started and that
1435 		 * ordered extents get created before we return
1436 		 */
1437 		atomic_inc(&root->fs_info->async_submit_draining);
1438 		while (atomic_read(&root->fs_info->nr_async_submits) ||
1439 		      atomic_read(&root->fs_info->async_delalloc_pages)) {
1440 			wait_event(root->fs_info->async_submit_wait,
1441 			   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1442 			    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1443 		}
1444 		atomic_dec(&root->fs_info->async_submit_draining);
1445 	}
1446 
1447 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1448 		btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1449 	}
1450 
1451 	ret = defrag_count;
1452 
1453 out_ra:
1454 	if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1455 		mutex_lock(&inode->i_mutex);
1456 		BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1457 		mutex_unlock(&inode->i_mutex);
1458 	}
1459 	if (!file)
1460 		kfree(ra);
1461 	kfree(pages);
1462 	return ret;
1463 }
1464 
1465 static noinline int btrfs_ioctl_resize(struct file *file,
1466 					void __user *arg)
1467 {
1468 	u64 new_size;
1469 	u64 old_size;
1470 	u64 devid = 1;
1471 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1472 	struct btrfs_ioctl_vol_args *vol_args;
1473 	struct btrfs_trans_handle *trans;
1474 	struct btrfs_device *device = NULL;
1475 	char *sizestr;
1476 	char *retptr;
1477 	char *devstr = NULL;
1478 	int ret = 0;
1479 	int mod = 0;
1480 
1481 	if (!capable(CAP_SYS_ADMIN))
1482 		return -EPERM;
1483 
1484 	ret = mnt_want_write_file(file);
1485 	if (ret)
1486 		return ret;
1487 
1488 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1489 			1)) {
1490 		mnt_drop_write_file(file);
1491 		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1492 	}
1493 
1494 	mutex_lock(&root->fs_info->volume_mutex);
1495 	vol_args = memdup_user(arg, sizeof(*vol_args));
1496 	if (IS_ERR(vol_args)) {
1497 		ret = PTR_ERR(vol_args);
1498 		goto out;
1499 	}
1500 
1501 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1502 
1503 	sizestr = vol_args->name;
1504 	devstr = strchr(sizestr, ':');
1505 	if (devstr) {
1506 		sizestr = devstr + 1;
1507 		*devstr = '\0';
1508 		devstr = vol_args->name;
1509 		ret = kstrtoull(devstr, 10, &devid);
1510 		if (ret)
1511 			goto out_free;
1512 		if (!devid) {
1513 			ret = -EINVAL;
1514 			goto out_free;
1515 		}
1516 		btrfs_info(root->fs_info, "resizing devid %llu", devid);
1517 	}
1518 
1519 	device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1520 	if (!device) {
1521 		btrfs_info(root->fs_info, "resizer unable to find device %llu",
1522 		       devid);
1523 		ret = -ENODEV;
1524 		goto out_free;
1525 	}
1526 
1527 	if (!device->writeable) {
1528 		btrfs_info(root->fs_info,
1529 			   "resizer unable to apply on readonly device %llu",
1530 		       devid);
1531 		ret = -EPERM;
1532 		goto out_free;
1533 	}
1534 
1535 	if (!strcmp(sizestr, "max"))
1536 		new_size = device->bdev->bd_inode->i_size;
1537 	else {
1538 		if (sizestr[0] == '-') {
1539 			mod = -1;
1540 			sizestr++;
1541 		} else if (sizestr[0] == '+') {
1542 			mod = 1;
1543 			sizestr++;
1544 		}
1545 		new_size = memparse(sizestr, &retptr);
1546 		if (*retptr != '\0' || new_size == 0) {
1547 			ret = -EINVAL;
1548 			goto out_free;
1549 		}
1550 	}
1551 
1552 	if (device->is_tgtdev_for_dev_replace) {
1553 		ret = -EPERM;
1554 		goto out_free;
1555 	}
1556 
1557 	old_size = device->total_bytes;
1558 
1559 	if (mod < 0) {
1560 		if (new_size > old_size) {
1561 			ret = -EINVAL;
1562 			goto out_free;
1563 		}
1564 		new_size = old_size - new_size;
1565 	} else if (mod > 0) {
1566 		if (new_size > ULLONG_MAX - old_size) {
1567 			ret = -ERANGE;
1568 			goto out_free;
1569 		}
1570 		new_size = old_size + new_size;
1571 	}
1572 
1573 	if (new_size < 256 * 1024 * 1024) {
1574 		ret = -EINVAL;
1575 		goto out_free;
1576 	}
1577 	if (new_size > device->bdev->bd_inode->i_size) {
1578 		ret = -EFBIG;
1579 		goto out_free;
1580 	}
1581 
1582 	do_div(new_size, root->sectorsize);
1583 	new_size *= root->sectorsize;
1584 
1585 	printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1586 		      rcu_str_deref(device->name), new_size);
1587 
1588 	if (new_size > old_size) {
1589 		trans = btrfs_start_transaction(root, 0);
1590 		if (IS_ERR(trans)) {
1591 			ret = PTR_ERR(trans);
1592 			goto out_free;
1593 		}
1594 		ret = btrfs_grow_device(trans, device, new_size);
1595 		btrfs_commit_transaction(trans, root);
1596 	} else if (new_size < old_size) {
1597 		ret = btrfs_shrink_device(device, new_size);
1598 	} /* equal, nothing need to do */
1599 
1600 out_free:
1601 	kfree(vol_args);
1602 out:
1603 	mutex_unlock(&root->fs_info->volume_mutex);
1604 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1605 	mnt_drop_write_file(file);
1606 	return ret;
1607 }
1608 
1609 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1610 				char *name, unsigned long fd, int subvol,
1611 				u64 *transid, bool readonly,
1612 				struct btrfs_qgroup_inherit *inherit)
1613 {
1614 	int namelen;
1615 	int ret = 0;
1616 
1617 	ret = mnt_want_write_file(file);
1618 	if (ret)
1619 		goto out;
1620 
1621 	namelen = strlen(name);
1622 	if (strchr(name, '/')) {
1623 		ret = -EINVAL;
1624 		goto out_drop_write;
1625 	}
1626 
1627 	if (name[0] == '.' &&
1628 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1629 		ret = -EEXIST;
1630 		goto out_drop_write;
1631 	}
1632 
1633 	if (subvol) {
1634 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1635 				     NULL, transid, readonly, inherit);
1636 	} else {
1637 		struct fd src = fdget(fd);
1638 		struct inode *src_inode;
1639 		if (!src.file) {
1640 			ret = -EINVAL;
1641 			goto out_drop_write;
1642 		}
1643 
1644 		src_inode = file_inode(src.file);
1645 		if (src_inode->i_sb != file_inode(file)->i_sb) {
1646 			btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1647 				   "Snapshot src from another FS");
1648 			ret = -EXDEV;
1649 		} else if (!inode_owner_or_capable(src_inode)) {
1650 			/*
1651 			 * Subvolume creation is not restricted, but snapshots
1652 			 * are limited to own subvolumes only
1653 			 */
1654 			ret = -EPERM;
1655 		} else {
1656 			ret = btrfs_mksubvol(&file->f_path, name, namelen,
1657 					     BTRFS_I(src_inode)->root,
1658 					     transid, readonly, inherit);
1659 		}
1660 		fdput(src);
1661 	}
1662 out_drop_write:
1663 	mnt_drop_write_file(file);
1664 out:
1665 	return ret;
1666 }
1667 
1668 static noinline int btrfs_ioctl_snap_create(struct file *file,
1669 					    void __user *arg, int subvol)
1670 {
1671 	struct btrfs_ioctl_vol_args *vol_args;
1672 	int ret;
1673 
1674 	vol_args = memdup_user(arg, sizeof(*vol_args));
1675 	if (IS_ERR(vol_args))
1676 		return PTR_ERR(vol_args);
1677 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1678 
1679 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1680 					      vol_args->fd, subvol,
1681 					      NULL, false, NULL);
1682 
1683 	kfree(vol_args);
1684 	return ret;
1685 }
1686 
1687 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1688 					       void __user *arg, int subvol)
1689 {
1690 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1691 	int ret;
1692 	u64 transid = 0;
1693 	u64 *ptr = NULL;
1694 	bool readonly = false;
1695 	struct btrfs_qgroup_inherit *inherit = NULL;
1696 
1697 	vol_args = memdup_user(arg, sizeof(*vol_args));
1698 	if (IS_ERR(vol_args))
1699 		return PTR_ERR(vol_args);
1700 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1701 
1702 	if (vol_args->flags &
1703 	    ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1704 	      BTRFS_SUBVOL_QGROUP_INHERIT)) {
1705 		ret = -EOPNOTSUPP;
1706 		goto free_args;
1707 	}
1708 
1709 	if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1710 		ptr = &transid;
1711 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1712 		readonly = true;
1713 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1714 		if (vol_args->size > PAGE_CACHE_SIZE) {
1715 			ret = -EINVAL;
1716 			goto free_args;
1717 		}
1718 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1719 		if (IS_ERR(inherit)) {
1720 			ret = PTR_ERR(inherit);
1721 			goto free_args;
1722 		}
1723 	}
1724 
1725 	ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1726 					      vol_args->fd, subvol, ptr,
1727 					      readonly, inherit);
1728 	if (ret)
1729 		goto free_inherit;
1730 
1731 	if (ptr && copy_to_user(arg +
1732 				offsetof(struct btrfs_ioctl_vol_args_v2,
1733 					transid),
1734 				ptr, sizeof(*ptr)))
1735 		ret = -EFAULT;
1736 
1737 free_inherit:
1738 	kfree(inherit);
1739 free_args:
1740 	kfree(vol_args);
1741 	return ret;
1742 }
1743 
1744 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1745 						void __user *arg)
1746 {
1747 	struct inode *inode = file_inode(file);
1748 	struct btrfs_root *root = BTRFS_I(inode)->root;
1749 	int ret = 0;
1750 	u64 flags = 0;
1751 
1752 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1753 		return -EINVAL;
1754 
1755 	down_read(&root->fs_info->subvol_sem);
1756 	if (btrfs_root_readonly(root))
1757 		flags |= BTRFS_SUBVOL_RDONLY;
1758 	up_read(&root->fs_info->subvol_sem);
1759 
1760 	if (copy_to_user(arg, &flags, sizeof(flags)))
1761 		ret = -EFAULT;
1762 
1763 	return ret;
1764 }
1765 
1766 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1767 					      void __user *arg)
1768 {
1769 	struct inode *inode = file_inode(file);
1770 	struct btrfs_root *root = BTRFS_I(inode)->root;
1771 	struct btrfs_trans_handle *trans;
1772 	u64 root_flags;
1773 	u64 flags;
1774 	int ret = 0;
1775 
1776 	if (!inode_owner_or_capable(inode))
1777 		return -EPERM;
1778 
1779 	ret = mnt_want_write_file(file);
1780 	if (ret)
1781 		goto out;
1782 
1783 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1784 		ret = -EINVAL;
1785 		goto out_drop_write;
1786 	}
1787 
1788 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1789 		ret = -EFAULT;
1790 		goto out_drop_write;
1791 	}
1792 
1793 	if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1794 		ret = -EINVAL;
1795 		goto out_drop_write;
1796 	}
1797 
1798 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1799 		ret = -EOPNOTSUPP;
1800 		goto out_drop_write;
1801 	}
1802 
1803 	down_write(&root->fs_info->subvol_sem);
1804 
1805 	/* nothing to do */
1806 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1807 		goto out_drop_sem;
1808 
1809 	root_flags = btrfs_root_flags(&root->root_item);
1810 	if (flags & BTRFS_SUBVOL_RDONLY) {
1811 		btrfs_set_root_flags(&root->root_item,
1812 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1813 	} else {
1814 		/*
1815 		 * Block RO -> RW transition if this subvolume is involved in
1816 		 * send
1817 		 */
1818 		spin_lock(&root->root_item_lock);
1819 		if (root->send_in_progress == 0) {
1820 			btrfs_set_root_flags(&root->root_item,
1821 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1822 			spin_unlock(&root->root_item_lock);
1823 		} else {
1824 			spin_unlock(&root->root_item_lock);
1825 			btrfs_warn(root->fs_info,
1826 			"Attempt to set subvolume %llu read-write during send",
1827 					root->root_key.objectid);
1828 			ret = -EPERM;
1829 			goto out_drop_sem;
1830 		}
1831 	}
1832 
1833 	trans = btrfs_start_transaction(root, 1);
1834 	if (IS_ERR(trans)) {
1835 		ret = PTR_ERR(trans);
1836 		goto out_reset;
1837 	}
1838 
1839 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
1840 				&root->root_key, &root->root_item);
1841 
1842 	btrfs_commit_transaction(trans, root);
1843 out_reset:
1844 	if (ret)
1845 		btrfs_set_root_flags(&root->root_item, root_flags);
1846 out_drop_sem:
1847 	up_write(&root->fs_info->subvol_sem);
1848 out_drop_write:
1849 	mnt_drop_write_file(file);
1850 out:
1851 	return ret;
1852 }
1853 
1854 /*
1855  * helper to check if the subvolume references other subvolumes
1856  */
1857 static noinline int may_destroy_subvol(struct btrfs_root *root)
1858 {
1859 	struct btrfs_path *path;
1860 	struct btrfs_dir_item *di;
1861 	struct btrfs_key key;
1862 	u64 dir_id;
1863 	int ret;
1864 
1865 	path = btrfs_alloc_path();
1866 	if (!path)
1867 		return -ENOMEM;
1868 
1869 	/* Make sure this root isn't set as the default subvol */
1870 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1871 	di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1872 				   dir_id, "default", 7, 0);
1873 	if (di && !IS_ERR(di)) {
1874 		btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1875 		if (key.objectid == root->root_key.objectid) {
1876 			ret = -EPERM;
1877 			btrfs_err(root->fs_info, "deleting default subvolume "
1878 				  "%llu is not allowed", key.objectid);
1879 			goto out;
1880 		}
1881 		btrfs_release_path(path);
1882 	}
1883 
1884 	key.objectid = root->root_key.objectid;
1885 	key.type = BTRFS_ROOT_REF_KEY;
1886 	key.offset = (u64)-1;
1887 
1888 	ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1889 				&key, path, 0, 0);
1890 	if (ret < 0)
1891 		goto out;
1892 	BUG_ON(ret == 0);
1893 
1894 	ret = 0;
1895 	if (path->slots[0] > 0) {
1896 		path->slots[0]--;
1897 		btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1898 		if (key.objectid == root->root_key.objectid &&
1899 		    key.type == BTRFS_ROOT_REF_KEY)
1900 			ret = -ENOTEMPTY;
1901 	}
1902 out:
1903 	btrfs_free_path(path);
1904 	return ret;
1905 }
1906 
1907 static noinline int key_in_sk(struct btrfs_key *key,
1908 			      struct btrfs_ioctl_search_key *sk)
1909 {
1910 	struct btrfs_key test;
1911 	int ret;
1912 
1913 	test.objectid = sk->min_objectid;
1914 	test.type = sk->min_type;
1915 	test.offset = sk->min_offset;
1916 
1917 	ret = btrfs_comp_cpu_keys(key, &test);
1918 	if (ret < 0)
1919 		return 0;
1920 
1921 	test.objectid = sk->max_objectid;
1922 	test.type = sk->max_type;
1923 	test.offset = sk->max_offset;
1924 
1925 	ret = btrfs_comp_cpu_keys(key, &test);
1926 	if (ret > 0)
1927 		return 0;
1928 	return 1;
1929 }
1930 
1931 static noinline int copy_to_sk(struct btrfs_root *root,
1932 			       struct btrfs_path *path,
1933 			       struct btrfs_key *key,
1934 			       struct btrfs_ioctl_search_key *sk,
1935 			       size_t *buf_size,
1936 			       char __user *ubuf,
1937 			       unsigned long *sk_offset,
1938 			       int *num_found)
1939 {
1940 	u64 found_transid;
1941 	struct extent_buffer *leaf;
1942 	struct btrfs_ioctl_search_header sh;
1943 	unsigned long item_off;
1944 	unsigned long item_len;
1945 	int nritems;
1946 	int i;
1947 	int slot;
1948 	int ret = 0;
1949 
1950 	leaf = path->nodes[0];
1951 	slot = path->slots[0];
1952 	nritems = btrfs_header_nritems(leaf);
1953 
1954 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1955 		i = nritems;
1956 		goto advance_key;
1957 	}
1958 	found_transid = btrfs_header_generation(leaf);
1959 
1960 	for (i = slot; i < nritems; i++) {
1961 		item_off = btrfs_item_ptr_offset(leaf, i);
1962 		item_len = btrfs_item_size_nr(leaf, i);
1963 
1964 		btrfs_item_key_to_cpu(leaf, key, i);
1965 		if (!key_in_sk(key, sk))
1966 			continue;
1967 
1968 		if (sizeof(sh) + item_len > *buf_size) {
1969 			if (*num_found) {
1970 				ret = 1;
1971 				goto out;
1972 			}
1973 
1974 			/*
1975 			 * return one empty item back for v1, which does not
1976 			 * handle -EOVERFLOW
1977 			 */
1978 
1979 			*buf_size = sizeof(sh) + item_len;
1980 			item_len = 0;
1981 			ret = -EOVERFLOW;
1982 		}
1983 
1984 		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1985 			ret = 1;
1986 			goto out;
1987 		}
1988 
1989 		sh.objectid = key->objectid;
1990 		sh.offset = key->offset;
1991 		sh.type = key->type;
1992 		sh.len = item_len;
1993 		sh.transid = found_transid;
1994 
1995 		/* copy search result header */
1996 		if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1997 			ret = -EFAULT;
1998 			goto out;
1999 		}
2000 
2001 		*sk_offset += sizeof(sh);
2002 
2003 		if (item_len) {
2004 			char __user *up = ubuf + *sk_offset;
2005 			/* copy the item */
2006 			if (read_extent_buffer_to_user(leaf, up,
2007 						       item_off, item_len)) {
2008 				ret = -EFAULT;
2009 				goto out;
2010 			}
2011 
2012 			*sk_offset += item_len;
2013 		}
2014 		(*num_found)++;
2015 
2016 		if (ret) /* -EOVERFLOW from above */
2017 			goto out;
2018 
2019 		if (*num_found >= sk->nr_items) {
2020 			ret = 1;
2021 			goto out;
2022 		}
2023 	}
2024 advance_key:
2025 	ret = 0;
2026 	if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2027 		key->offset++;
2028 	else if (key->type < (u8)-1 && key->type < sk->max_type) {
2029 		key->offset = 0;
2030 		key->type++;
2031 	} else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2032 		key->offset = 0;
2033 		key->type = 0;
2034 		key->objectid++;
2035 	} else
2036 		ret = 1;
2037 out:
2038 	/*
2039 	 *  0: all items from this leaf copied, continue with next
2040 	 *  1: * more items can be copied, but unused buffer is too small
2041 	 *     * all items were found
2042 	 *     Either way, it will stops the loop which iterates to the next
2043 	 *     leaf
2044 	 *  -EOVERFLOW: item was to large for buffer
2045 	 *  -EFAULT: could not copy extent buffer back to userspace
2046 	 */
2047 	return ret;
2048 }
2049 
2050 static noinline int search_ioctl(struct inode *inode,
2051 				 struct btrfs_ioctl_search_key *sk,
2052 				 size_t *buf_size,
2053 				 char __user *ubuf)
2054 {
2055 	struct btrfs_root *root;
2056 	struct btrfs_key key;
2057 	struct btrfs_path *path;
2058 	struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2059 	int ret;
2060 	int num_found = 0;
2061 	unsigned long sk_offset = 0;
2062 
2063 	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2064 		*buf_size = sizeof(struct btrfs_ioctl_search_header);
2065 		return -EOVERFLOW;
2066 	}
2067 
2068 	path = btrfs_alloc_path();
2069 	if (!path)
2070 		return -ENOMEM;
2071 
2072 	if (sk->tree_id == 0) {
2073 		/* search the root of the inode that was passed */
2074 		root = BTRFS_I(inode)->root;
2075 	} else {
2076 		key.objectid = sk->tree_id;
2077 		key.type = BTRFS_ROOT_ITEM_KEY;
2078 		key.offset = (u64)-1;
2079 		root = btrfs_read_fs_root_no_name(info, &key);
2080 		if (IS_ERR(root)) {
2081 			printk(KERN_ERR "BTRFS: could not find root %llu\n",
2082 			       sk->tree_id);
2083 			btrfs_free_path(path);
2084 			return -ENOENT;
2085 		}
2086 	}
2087 
2088 	key.objectid = sk->min_objectid;
2089 	key.type = sk->min_type;
2090 	key.offset = sk->min_offset;
2091 
2092 	path->keep_locks = 1;
2093 
2094 	while (1) {
2095 		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2096 		if (ret != 0) {
2097 			if (ret > 0)
2098 				ret = 0;
2099 			goto err;
2100 		}
2101 		ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2102 				 &sk_offset, &num_found);
2103 		btrfs_release_path(path);
2104 		if (ret)
2105 			break;
2106 
2107 	}
2108 	if (ret > 0)
2109 		ret = 0;
2110 err:
2111 	sk->nr_items = num_found;
2112 	btrfs_free_path(path);
2113 	return ret;
2114 }
2115 
2116 static noinline int btrfs_ioctl_tree_search(struct file *file,
2117 					   void __user *argp)
2118 {
2119 	struct btrfs_ioctl_search_args __user *uargs;
2120 	struct btrfs_ioctl_search_key sk;
2121 	struct inode *inode;
2122 	int ret;
2123 	size_t buf_size;
2124 
2125 	if (!capable(CAP_SYS_ADMIN))
2126 		return -EPERM;
2127 
2128 	uargs = (struct btrfs_ioctl_search_args __user *)argp;
2129 
2130 	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2131 		return -EFAULT;
2132 
2133 	buf_size = sizeof(uargs->buf);
2134 
2135 	inode = file_inode(file);
2136 	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2137 
2138 	/*
2139 	 * In the origin implementation an overflow is handled by returning a
2140 	 * search header with a len of zero, so reset ret.
2141 	 */
2142 	if (ret == -EOVERFLOW)
2143 		ret = 0;
2144 
2145 	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2146 		ret = -EFAULT;
2147 	return ret;
2148 }
2149 
2150 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2151 					       void __user *argp)
2152 {
2153 	struct btrfs_ioctl_search_args_v2 __user *uarg;
2154 	struct btrfs_ioctl_search_args_v2 args;
2155 	struct inode *inode;
2156 	int ret;
2157 	size_t buf_size;
2158 	const size_t buf_limit = 16 * 1024 * 1024;
2159 
2160 	if (!capable(CAP_SYS_ADMIN))
2161 		return -EPERM;
2162 
2163 	/* copy search header and buffer size */
2164 	uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2165 	if (copy_from_user(&args, uarg, sizeof(args)))
2166 		return -EFAULT;
2167 
2168 	buf_size = args.buf_size;
2169 
2170 	if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2171 		return -EOVERFLOW;
2172 
2173 	/* limit result size to 16MB */
2174 	if (buf_size > buf_limit)
2175 		buf_size = buf_limit;
2176 
2177 	inode = file_inode(file);
2178 	ret = search_ioctl(inode, &args.key, &buf_size,
2179 			   (char *)(&uarg->buf[0]));
2180 	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2181 		ret = -EFAULT;
2182 	else if (ret == -EOVERFLOW &&
2183 		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2184 		ret = -EFAULT;
2185 
2186 	return ret;
2187 }
2188 
2189 /*
2190  * Search INODE_REFs to identify path name of 'dirid' directory
2191  * in a 'tree_id' tree. and sets path name to 'name'.
2192  */
2193 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2194 				u64 tree_id, u64 dirid, char *name)
2195 {
2196 	struct btrfs_root *root;
2197 	struct btrfs_key key;
2198 	char *ptr;
2199 	int ret = -1;
2200 	int slot;
2201 	int len;
2202 	int total_len = 0;
2203 	struct btrfs_inode_ref *iref;
2204 	struct extent_buffer *l;
2205 	struct btrfs_path *path;
2206 
2207 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2208 		name[0]='\0';
2209 		return 0;
2210 	}
2211 
2212 	path = btrfs_alloc_path();
2213 	if (!path)
2214 		return -ENOMEM;
2215 
2216 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2217 
2218 	key.objectid = tree_id;
2219 	key.type = BTRFS_ROOT_ITEM_KEY;
2220 	key.offset = (u64)-1;
2221 	root = btrfs_read_fs_root_no_name(info, &key);
2222 	if (IS_ERR(root)) {
2223 		printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2224 		ret = -ENOENT;
2225 		goto out;
2226 	}
2227 
2228 	key.objectid = dirid;
2229 	key.type = BTRFS_INODE_REF_KEY;
2230 	key.offset = (u64)-1;
2231 
2232 	while (1) {
2233 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2234 		if (ret < 0)
2235 			goto out;
2236 		else if (ret > 0) {
2237 			ret = btrfs_previous_item(root, path, dirid,
2238 						  BTRFS_INODE_REF_KEY);
2239 			if (ret < 0)
2240 				goto out;
2241 			else if (ret > 0) {
2242 				ret = -ENOENT;
2243 				goto out;
2244 			}
2245 		}
2246 
2247 		l = path->nodes[0];
2248 		slot = path->slots[0];
2249 		btrfs_item_key_to_cpu(l, &key, slot);
2250 
2251 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2252 		len = btrfs_inode_ref_name_len(l, iref);
2253 		ptr -= len + 1;
2254 		total_len += len + 1;
2255 		if (ptr < name) {
2256 			ret = -ENAMETOOLONG;
2257 			goto out;
2258 		}
2259 
2260 		*(ptr + len) = '/';
2261 		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2262 
2263 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2264 			break;
2265 
2266 		btrfs_release_path(path);
2267 		key.objectid = key.offset;
2268 		key.offset = (u64)-1;
2269 		dirid = key.objectid;
2270 	}
2271 	memmove(name, ptr, total_len);
2272 	name[total_len] = '\0';
2273 	ret = 0;
2274 out:
2275 	btrfs_free_path(path);
2276 	return ret;
2277 }
2278 
2279 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2280 					   void __user *argp)
2281 {
2282 	 struct btrfs_ioctl_ino_lookup_args *args;
2283 	 struct inode *inode;
2284 	 int ret;
2285 
2286 	if (!capable(CAP_SYS_ADMIN))
2287 		return -EPERM;
2288 
2289 	args = memdup_user(argp, sizeof(*args));
2290 	if (IS_ERR(args))
2291 		return PTR_ERR(args);
2292 
2293 	inode = file_inode(file);
2294 
2295 	if (args->treeid == 0)
2296 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2297 
2298 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2299 					args->treeid, args->objectid,
2300 					args->name);
2301 
2302 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2303 		ret = -EFAULT;
2304 
2305 	kfree(args);
2306 	return ret;
2307 }
2308 
2309 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2310 					     void __user *arg)
2311 {
2312 	struct dentry *parent = file->f_path.dentry;
2313 	struct dentry *dentry;
2314 	struct inode *dir = parent->d_inode;
2315 	struct inode *inode;
2316 	struct btrfs_root *root = BTRFS_I(dir)->root;
2317 	struct btrfs_root *dest = NULL;
2318 	struct btrfs_ioctl_vol_args *vol_args;
2319 	struct btrfs_trans_handle *trans;
2320 	struct btrfs_block_rsv block_rsv;
2321 	u64 root_flags;
2322 	u64 qgroup_reserved;
2323 	int namelen;
2324 	int ret;
2325 	int err = 0;
2326 
2327 	vol_args = memdup_user(arg, sizeof(*vol_args));
2328 	if (IS_ERR(vol_args))
2329 		return PTR_ERR(vol_args);
2330 
2331 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2332 	namelen = strlen(vol_args->name);
2333 	if (strchr(vol_args->name, '/') ||
2334 	    strncmp(vol_args->name, "..", namelen) == 0) {
2335 		err = -EINVAL;
2336 		goto out;
2337 	}
2338 
2339 	err = mnt_want_write_file(file);
2340 	if (err)
2341 		goto out;
2342 
2343 
2344 	err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2345 	if (err == -EINTR)
2346 		goto out_drop_write;
2347 	dentry = lookup_one_len(vol_args->name, parent, namelen);
2348 	if (IS_ERR(dentry)) {
2349 		err = PTR_ERR(dentry);
2350 		goto out_unlock_dir;
2351 	}
2352 
2353 	if (!dentry->d_inode) {
2354 		err = -ENOENT;
2355 		goto out_dput;
2356 	}
2357 
2358 	inode = dentry->d_inode;
2359 	dest = BTRFS_I(inode)->root;
2360 	if (!capable(CAP_SYS_ADMIN)) {
2361 		/*
2362 		 * Regular user.  Only allow this with a special mount
2363 		 * option, when the user has write+exec access to the
2364 		 * subvol root, and when rmdir(2) would have been
2365 		 * allowed.
2366 		 *
2367 		 * Note that this is _not_ check that the subvol is
2368 		 * empty or doesn't contain data that we wouldn't
2369 		 * otherwise be able to delete.
2370 		 *
2371 		 * Users who want to delete empty subvols should try
2372 		 * rmdir(2).
2373 		 */
2374 		err = -EPERM;
2375 		if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2376 			goto out_dput;
2377 
2378 		/*
2379 		 * Do not allow deletion if the parent dir is the same
2380 		 * as the dir to be deleted.  That means the ioctl
2381 		 * must be called on the dentry referencing the root
2382 		 * of the subvol, not a random directory contained
2383 		 * within it.
2384 		 */
2385 		err = -EINVAL;
2386 		if (root == dest)
2387 			goto out_dput;
2388 
2389 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2390 		if (err)
2391 			goto out_dput;
2392 	}
2393 
2394 	/* check if subvolume may be deleted by a user */
2395 	err = btrfs_may_delete(dir, dentry, 1);
2396 	if (err)
2397 		goto out_dput;
2398 
2399 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2400 		err = -EINVAL;
2401 		goto out_dput;
2402 	}
2403 
2404 	mutex_lock(&inode->i_mutex);
2405 
2406 	/*
2407 	 * Don't allow to delete a subvolume with send in progress. This is
2408 	 * inside the i_mutex so the error handling that has to drop the bit
2409 	 * again is not run concurrently.
2410 	 */
2411 	spin_lock(&dest->root_item_lock);
2412 	root_flags = btrfs_root_flags(&dest->root_item);
2413 	if (dest->send_in_progress == 0) {
2414 		btrfs_set_root_flags(&dest->root_item,
2415 				root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2416 		spin_unlock(&dest->root_item_lock);
2417 	} else {
2418 		spin_unlock(&dest->root_item_lock);
2419 		btrfs_warn(root->fs_info,
2420 			"Attempt to delete subvolume %llu during send",
2421 			dest->root_key.objectid);
2422 		err = -EPERM;
2423 		goto out_dput;
2424 	}
2425 
2426 	err = d_invalidate(dentry);
2427 	if (err)
2428 		goto out_unlock;
2429 
2430 	down_write(&root->fs_info->subvol_sem);
2431 
2432 	err = may_destroy_subvol(dest);
2433 	if (err)
2434 		goto out_up_write;
2435 
2436 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2437 	/*
2438 	 * One for dir inode, two for dir entries, two for root
2439 	 * ref/backref.
2440 	 */
2441 	err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2442 					       5, &qgroup_reserved, true);
2443 	if (err)
2444 		goto out_up_write;
2445 
2446 	trans = btrfs_start_transaction(root, 0);
2447 	if (IS_ERR(trans)) {
2448 		err = PTR_ERR(trans);
2449 		goto out_release;
2450 	}
2451 	trans->block_rsv = &block_rsv;
2452 	trans->bytes_reserved = block_rsv.size;
2453 
2454 	ret = btrfs_unlink_subvol(trans, root, dir,
2455 				dest->root_key.objectid,
2456 				dentry->d_name.name,
2457 				dentry->d_name.len);
2458 	if (ret) {
2459 		err = ret;
2460 		btrfs_abort_transaction(trans, root, ret);
2461 		goto out_end_trans;
2462 	}
2463 
2464 	btrfs_record_root_in_trans(trans, dest);
2465 
2466 	memset(&dest->root_item.drop_progress, 0,
2467 		sizeof(dest->root_item.drop_progress));
2468 	dest->root_item.drop_level = 0;
2469 	btrfs_set_root_refs(&dest->root_item, 0);
2470 
2471 	if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2472 		ret = btrfs_insert_orphan_item(trans,
2473 					root->fs_info->tree_root,
2474 					dest->root_key.objectid);
2475 		if (ret) {
2476 			btrfs_abort_transaction(trans, root, ret);
2477 			err = ret;
2478 			goto out_end_trans;
2479 		}
2480 	}
2481 
2482 	ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2483 				  dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2484 				  dest->root_key.objectid);
2485 	if (ret && ret != -ENOENT) {
2486 		btrfs_abort_transaction(trans, root, ret);
2487 		err = ret;
2488 		goto out_end_trans;
2489 	}
2490 	if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2491 		ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2492 					  dest->root_item.received_uuid,
2493 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2494 					  dest->root_key.objectid);
2495 		if (ret && ret != -ENOENT) {
2496 			btrfs_abort_transaction(trans, root, ret);
2497 			err = ret;
2498 			goto out_end_trans;
2499 		}
2500 	}
2501 
2502 out_end_trans:
2503 	trans->block_rsv = NULL;
2504 	trans->bytes_reserved = 0;
2505 	ret = btrfs_end_transaction(trans, root);
2506 	if (ret && !err)
2507 		err = ret;
2508 	inode->i_flags |= S_DEAD;
2509 out_release:
2510 	btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2511 out_up_write:
2512 	up_write(&root->fs_info->subvol_sem);
2513 out_unlock:
2514 	if (err) {
2515 		spin_lock(&dest->root_item_lock);
2516 		root_flags = btrfs_root_flags(&dest->root_item);
2517 		btrfs_set_root_flags(&dest->root_item,
2518 				root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2519 		spin_unlock(&dest->root_item_lock);
2520 	}
2521 	mutex_unlock(&inode->i_mutex);
2522 	if (!err) {
2523 		shrink_dcache_sb(root->fs_info->sb);
2524 		btrfs_invalidate_inodes(dest);
2525 		d_delete(dentry);
2526 		ASSERT(dest->send_in_progress == 0);
2527 
2528 		/* the last ref */
2529 		if (dest->cache_inode) {
2530 			iput(dest->cache_inode);
2531 			dest->cache_inode = NULL;
2532 		}
2533 	}
2534 out_dput:
2535 	dput(dentry);
2536 out_unlock_dir:
2537 	mutex_unlock(&dir->i_mutex);
2538 out_drop_write:
2539 	mnt_drop_write_file(file);
2540 out:
2541 	kfree(vol_args);
2542 	return err;
2543 }
2544 
2545 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2546 {
2547 	struct inode *inode = file_inode(file);
2548 	struct btrfs_root *root = BTRFS_I(inode)->root;
2549 	struct btrfs_ioctl_defrag_range_args *range;
2550 	int ret;
2551 
2552 	ret = mnt_want_write_file(file);
2553 	if (ret)
2554 		return ret;
2555 
2556 	if (btrfs_root_readonly(root)) {
2557 		ret = -EROFS;
2558 		goto out;
2559 	}
2560 
2561 	switch (inode->i_mode & S_IFMT) {
2562 	case S_IFDIR:
2563 		if (!capable(CAP_SYS_ADMIN)) {
2564 			ret = -EPERM;
2565 			goto out;
2566 		}
2567 		ret = btrfs_defrag_root(root);
2568 		if (ret)
2569 			goto out;
2570 		ret = btrfs_defrag_root(root->fs_info->extent_root);
2571 		break;
2572 	case S_IFREG:
2573 		if (!(file->f_mode & FMODE_WRITE)) {
2574 			ret = -EINVAL;
2575 			goto out;
2576 		}
2577 
2578 		range = kzalloc(sizeof(*range), GFP_KERNEL);
2579 		if (!range) {
2580 			ret = -ENOMEM;
2581 			goto out;
2582 		}
2583 
2584 		if (argp) {
2585 			if (copy_from_user(range, argp,
2586 					   sizeof(*range))) {
2587 				ret = -EFAULT;
2588 				kfree(range);
2589 				goto out;
2590 			}
2591 			/* compression requires us to start the IO */
2592 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2593 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2594 				range->extent_thresh = (u32)-1;
2595 			}
2596 		} else {
2597 			/* the rest are all set to zero by kzalloc */
2598 			range->len = (u64)-1;
2599 		}
2600 		ret = btrfs_defrag_file(file_inode(file), file,
2601 					range, 0, 0);
2602 		if (ret > 0)
2603 			ret = 0;
2604 		kfree(range);
2605 		break;
2606 	default:
2607 		ret = -EINVAL;
2608 	}
2609 out:
2610 	mnt_drop_write_file(file);
2611 	return ret;
2612 }
2613 
2614 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2615 {
2616 	struct btrfs_ioctl_vol_args *vol_args;
2617 	int ret;
2618 
2619 	if (!capable(CAP_SYS_ADMIN))
2620 		return -EPERM;
2621 
2622 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2623 			1)) {
2624 		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2625 	}
2626 
2627 	mutex_lock(&root->fs_info->volume_mutex);
2628 	vol_args = memdup_user(arg, sizeof(*vol_args));
2629 	if (IS_ERR(vol_args)) {
2630 		ret = PTR_ERR(vol_args);
2631 		goto out;
2632 	}
2633 
2634 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2635 	ret = btrfs_init_new_device(root, vol_args->name);
2636 
2637 	kfree(vol_args);
2638 out:
2639 	mutex_unlock(&root->fs_info->volume_mutex);
2640 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2641 	return ret;
2642 }
2643 
2644 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2645 {
2646 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2647 	struct btrfs_ioctl_vol_args *vol_args;
2648 	int ret;
2649 
2650 	if (!capable(CAP_SYS_ADMIN))
2651 		return -EPERM;
2652 
2653 	ret = mnt_want_write_file(file);
2654 	if (ret)
2655 		return ret;
2656 
2657 	vol_args = memdup_user(arg, sizeof(*vol_args));
2658 	if (IS_ERR(vol_args)) {
2659 		ret = PTR_ERR(vol_args);
2660 		goto err_drop;
2661 	}
2662 
2663 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2664 
2665 	if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2666 			1)) {
2667 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2668 		goto out;
2669 	}
2670 
2671 	mutex_lock(&root->fs_info->volume_mutex);
2672 	ret = btrfs_rm_device(root, vol_args->name);
2673 	mutex_unlock(&root->fs_info->volume_mutex);
2674 	atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2675 
2676 out:
2677 	kfree(vol_args);
2678 err_drop:
2679 	mnt_drop_write_file(file);
2680 	return ret;
2681 }
2682 
2683 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2684 {
2685 	struct btrfs_ioctl_fs_info_args *fi_args;
2686 	struct btrfs_device *device;
2687 	struct btrfs_device *next;
2688 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2689 	int ret = 0;
2690 
2691 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2692 	if (!fi_args)
2693 		return -ENOMEM;
2694 
2695 	mutex_lock(&fs_devices->device_list_mutex);
2696 	fi_args->num_devices = fs_devices->num_devices;
2697 	memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2698 
2699 	list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2700 		if (device->devid > fi_args->max_id)
2701 			fi_args->max_id = device->devid;
2702 	}
2703 	mutex_unlock(&fs_devices->device_list_mutex);
2704 
2705 	fi_args->nodesize = root->fs_info->super_copy->nodesize;
2706 	fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2707 	fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2708 
2709 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2710 		ret = -EFAULT;
2711 
2712 	kfree(fi_args);
2713 	return ret;
2714 }
2715 
2716 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2717 {
2718 	struct btrfs_ioctl_dev_info_args *di_args;
2719 	struct btrfs_device *dev;
2720 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2721 	int ret = 0;
2722 	char *s_uuid = NULL;
2723 
2724 	di_args = memdup_user(arg, sizeof(*di_args));
2725 	if (IS_ERR(di_args))
2726 		return PTR_ERR(di_args);
2727 
2728 	if (!btrfs_is_empty_uuid(di_args->uuid))
2729 		s_uuid = di_args->uuid;
2730 
2731 	mutex_lock(&fs_devices->device_list_mutex);
2732 	dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2733 
2734 	if (!dev) {
2735 		ret = -ENODEV;
2736 		goto out;
2737 	}
2738 
2739 	di_args->devid = dev->devid;
2740 	di_args->bytes_used = dev->bytes_used;
2741 	di_args->total_bytes = dev->total_bytes;
2742 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2743 	if (dev->name) {
2744 		struct rcu_string *name;
2745 
2746 		rcu_read_lock();
2747 		name = rcu_dereference(dev->name);
2748 		strncpy(di_args->path, name->str, sizeof(di_args->path));
2749 		rcu_read_unlock();
2750 		di_args->path[sizeof(di_args->path) - 1] = 0;
2751 	} else {
2752 		di_args->path[0] = '\0';
2753 	}
2754 
2755 out:
2756 	mutex_unlock(&fs_devices->device_list_mutex);
2757 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2758 		ret = -EFAULT;
2759 
2760 	kfree(di_args);
2761 	return ret;
2762 }
2763 
2764 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2765 {
2766 	struct page *page;
2767 	pgoff_t index;
2768 	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2769 
2770 	index = off >> PAGE_CACHE_SHIFT;
2771 
2772 	page = grab_cache_page(inode->i_mapping, index);
2773 	if (!page)
2774 		return NULL;
2775 
2776 	if (!PageUptodate(page)) {
2777 		if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2778 						 0))
2779 			return NULL;
2780 		lock_page(page);
2781 		if (!PageUptodate(page)) {
2782 			unlock_page(page);
2783 			page_cache_release(page);
2784 			return NULL;
2785 		}
2786 	}
2787 	unlock_page(page);
2788 
2789 	return page;
2790 }
2791 
2792 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2793 {
2794 	/* do any pending delalloc/csum calc on src, one way or
2795 	   another, and lock file content */
2796 	while (1) {
2797 		struct btrfs_ordered_extent *ordered;
2798 		lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2799 		ordered = btrfs_lookup_first_ordered_extent(inode,
2800 							    off + len - 1);
2801 		if ((!ordered ||
2802 		     ordered->file_offset + ordered->len <= off ||
2803 		     ordered->file_offset >= off + len) &&
2804 		    !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2805 				    off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2806 			if (ordered)
2807 				btrfs_put_ordered_extent(ordered);
2808 			break;
2809 		}
2810 		unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2811 		if (ordered)
2812 			btrfs_put_ordered_extent(ordered);
2813 		btrfs_wait_ordered_range(inode, off, len);
2814 	}
2815 }
2816 
2817 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2818 				struct inode *inode2, u64 loff2, u64 len)
2819 {
2820 	unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2821 	unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2822 
2823 	mutex_unlock(&inode1->i_mutex);
2824 	mutex_unlock(&inode2->i_mutex);
2825 }
2826 
2827 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2828 			      struct inode *inode2, u64 loff2, u64 len)
2829 {
2830 	if (inode1 < inode2) {
2831 		swap(inode1, inode2);
2832 		swap(loff1, loff2);
2833 	}
2834 
2835 	mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2836 	lock_extent_range(inode1, loff1, len);
2837 	if (inode1 != inode2) {
2838 		mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2839 		lock_extent_range(inode2, loff2, len);
2840 	}
2841 }
2842 
2843 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2844 			  u64 dst_loff, u64 len)
2845 {
2846 	int ret = 0;
2847 	struct page *src_page, *dst_page;
2848 	unsigned int cmp_len = PAGE_CACHE_SIZE;
2849 	void *addr, *dst_addr;
2850 
2851 	while (len) {
2852 		if (len < PAGE_CACHE_SIZE)
2853 			cmp_len = len;
2854 
2855 		src_page = extent_same_get_page(src, loff);
2856 		if (!src_page)
2857 			return -EINVAL;
2858 		dst_page = extent_same_get_page(dst, dst_loff);
2859 		if (!dst_page) {
2860 			page_cache_release(src_page);
2861 			return -EINVAL;
2862 		}
2863 		addr = kmap_atomic(src_page);
2864 		dst_addr = kmap_atomic(dst_page);
2865 
2866 		flush_dcache_page(src_page);
2867 		flush_dcache_page(dst_page);
2868 
2869 		if (memcmp(addr, dst_addr, cmp_len))
2870 			ret = BTRFS_SAME_DATA_DIFFERS;
2871 
2872 		kunmap_atomic(addr);
2873 		kunmap_atomic(dst_addr);
2874 		page_cache_release(src_page);
2875 		page_cache_release(dst_page);
2876 
2877 		if (ret)
2878 			break;
2879 
2880 		loff += cmp_len;
2881 		dst_loff += cmp_len;
2882 		len -= cmp_len;
2883 	}
2884 
2885 	return ret;
2886 }
2887 
2888 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2889 {
2890 	u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2891 
2892 	if (off + len > inode->i_size || off + len < off)
2893 		return -EINVAL;
2894 	/* Check that we are block aligned - btrfs_clone() requires this */
2895 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2896 		return -EINVAL;
2897 
2898 	return 0;
2899 }
2900 
2901 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2902 			     struct inode *dst, u64 dst_loff)
2903 {
2904 	int ret;
2905 
2906 	/*
2907 	 * btrfs_clone() can't handle extents in the same file
2908 	 * yet. Once that works, we can drop this check and replace it
2909 	 * with a check for the same inode, but overlapping extents.
2910 	 */
2911 	if (src == dst)
2912 		return -EINVAL;
2913 
2914 	btrfs_double_lock(src, loff, dst, dst_loff, len);
2915 
2916 	ret = extent_same_check_offsets(src, loff, len);
2917 	if (ret)
2918 		goto out_unlock;
2919 
2920 	ret = extent_same_check_offsets(dst, dst_loff, len);
2921 	if (ret)
2922 		goto out_unlock;
2923 
2924 	/* don't make the dst file partly checksummed */
2925 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2926 	    (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2927 		ret = -EINVAL;
2928 		goto out_unlock;
2929 	}
2930 
2931 	ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2932 	if (ret == 0)
2933 		ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2934 
2935 out_unlock:
2936 	btrfs_double_unlock(src, loff, dst, dst_loff, len);
2937 
2938 	return ret;
2939 }
2940 
2941 #define BTRFS_MAX_DEDUPE_LEN	(16 * 1024 * 1024)
2942 
2943 static long btrfs_ioctl_file_extent_same(struct file *file,
2944 			struct btrfs_ioctl_same_args __user *argp)
2945 {
2946 	struct btrfs_ioctl_same_args *same;
2947 	struct btrfs_ioctl_same_extent_info *info;
2948 	struct inode *src = file_inode(file);
2949 	u64 off;
2950 	u64 len;
2951 	int i;
2952 	int ret;
2953 	unsigned long size;
2954 	u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2955 	bool is_admin = capable(CAP_SYS_ADMIN);
2956 	u16 count;
2957 
2958 	if (!(file->f_mode & FMODE_READ))
2959 		return -EINVAL;
2960 
2961 	ret = mnt_want_write_file(file);
2962 	if (ret)
2963 		return ret;
2964 
2965 	if (get_user(count, &argp->dest_count)) {
2966 		ret = -EFAULT;
2967 		goto out;
2968 	}
2969 
2970 	size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2971 
2972 	same = memdup_user(argp, size);
2973 
2974 	if (IS_ERR(same)) {
2975 		ret = PTR_ERR(same);
2976 		goto out;
2977 	}
2978 
2979 	off = same->logical_offset;
2980 	len = same->length;
2981 
2982 	/*
2983 	 * Limit the total length we will dedupe for each operation.
2984 	 * This is intended to bound the total time spent in this
2985 	 * ioctl to something sane.
2986 	 */
2987 	if (len > BTRFS_MAX_DEDUPE_LEN)
2988 		len = BTRFS_MAX_DEDUPE_LEN;
2989 
2990 	if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2991 		/*
2992 		 * Btrfs does not support blocksize < page_size. As a
2993 		 * result, btrfs_cmp_data() won't correctly handle
2994 		 * this situation without an update.
2995 		 */
2996 		ret = -EINVAL;
2997 		goto out;
2998 	}
2999 
3000 	ret = -EISDIR;
3001 	if (S_ISDIR(src->i_mode))
3002 		goto out;
3003 
3004 	ret = -EACCES;
3005 	if (!S_ISREG(src->i_mode))
3006 		goto out;
3007 
3008 	/* pre-format output fields to sane values */
3009 	for (i = 0; i < count; i++) {
3010 		same->info[i].bytes_deduped = 0ULL;
3011 		same->info[i].status = 0;
3012 	}
3013 
3014 	for (i = 0, info = same->info; i < count; i++, info++) {
3015 		struct inode *dst;
3016 		struct fd dst_file = fdget(info->fd);
3017 		if (!dst_file.file) {
3018 			info->status = -EBADF;
3019 			continue;
3020 		}
3021 		dst = file_inode(dst_file.file);
3022 
3023 		if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3024 			info->status = -EINVAL;
3025 		} else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3026 			info->status = -EXDEV;
3027 		} else if (S_ISDIR(dst->i_mode)) {
3028 			info->status = -EISDIR;
3029 		} else if (!S_ISREG(dst->i_mode)) {
3030 			info->status = -EACCES;
3031 		} else {
3032 			info->status = btrfs_extent_same(src, off, len, dst,
3033 							info->logical_offset);
3034 			if (info->status == 0)
3035 				info->bytes_deduped += len;
3036 		}
3037 		fdput(dst_file);
3038 	}
3039 
3040 	ret = copy_to_user(argp, same, size);
3041 	if (ret)
3042 		ret = -EFAULT;
3043 
3044 out:
3045 	mnt_drop_write_file(file);
3046 	return ret;
3047 }
3048 
3049 /* Helper to check and see if this root currently has a ref on the given disk
3050  * bytenr.  If it does then we need to update the quota for this root.  This
3051  * doesn't do anything if quotas aren't enabled.
3052  */
3053 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3054 		     u64 disko)
3055 {
3056 	struct seq_list tree_mod_seq_elem = {};
3057 	struct ulist *roots;
3058 	struct ulist_iterator uiter;
3059 	struct ulist_node *root_node = NULL;
3060 	int ret;
3061 
3062 	if (!root->fs_info->quota_enabled)
3063 		return 1;
3064 
3065 	btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3066 	ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3067 				   tree_mod_seq_elem.seq, &roots);
3068 	if (ret < 0)
3069 		goto out;
3070 	ret = 0;
3071 	ULIST_ITER_INIT(&uiter);
3072 	while ((root_node = ulist_next(roots, &uiter))) {
3073 		if (root_node->val == root->objectid) {
3074 			ret = 1;
3075 			break;
3076 		}
3077 	}
3078 	ulist_free(roots);
3079 out:
3080 	btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3081 	return ret;
3082 }
3083 
3084 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3085 				     struct inode *inode,
3086 				     u64 endoff,
3087 				     const u64 destoff,
3088 				     const u64 olen)
3089 {
3090 	struct btrfs_root *root = BTRFS_I(inode)->root;
3091 	int ret;
3092 
3093 	inode_inc_iversion(inode);
3094 	inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3095 	/*
3096 	 * We round up to the block size at eof when determining which
3097 	 * extents to clone above, but shouldn't round up the file size.
3098 	 */
3099 	if (endoff > destoff + olen)
3100 		endoff = destoff + olen;
3101 	if (endoff > inode->i_size)
3102 		btrfs_i_size_write(inode, endoff);
3103 
3104 	ret = btrfs_update_inode(trans, root, inode);
3105 	if (ret) {
3106 		btrfs_abort_transaction(trans, root, ret);
3107 		btrfs_end_transaction(trans, root);
3108 		goto out;
3109 	}
3110 	ret = btrfs_end_transaction(trans, root);
3111 out:
3112 	return ret;
3113 }
3114 
3115 static void clone_update_extent_map(struct inode *inode,
3116 				    const struct btrfs_trans_handle *trans,
3117 				    const struct btrfs_path *path,
3118 				    const u64 hole_offset,
3119 				    const u64 hole_len)
3120 {
3121 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3122 	struct extent_map *em;
3123 	int ret;
3124 
3125 	em = alloc_extent_map();
3126 	if (!em) {
3127 		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3128 			&BTRFS_I(inode)->runtime_flags);
3129 		return;
3130 	}
3131 
3132 	if (path) {
3133 		struct btrfs_file_extent_item *fi;
3134 
3135 		fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3136 				    struct btrfs_file_extent_item);
3137 		btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3138 		em->generation = -1;
3139 		if (btrfs_file_extent_type(path->nodes[0], fi) ==
3140 		    BTRFS_FILE_EXTENT_INLINE)
3141 			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3142 				&BTRFS_I(inode)->runtime_flags);
3143 	} else {
3144 		em->start = hole_offset;
3145 		em->len = hole_len;
3146 		em->ram_bytes = em->len;
3147 		em->orig_start = hole_offset;
3148 		em->block_start = EXTENT_MAP_HOLE;
3149 		em->block_len = 0;
3150 		em->orig_block_len = 0;
3151 		em->compress_type = BTRFS_COMPRESS_NONE;
3152 		em->generation = trans->transid;
3153 	}
3154 
3155 	while (1) {
3156 		write_lock(&em_tree->lock);
3157 		ret = add_extent_mapping(em_tree, em, 1);
3158 		write_unlock(&em_tree->lock);
3159 		if (ret != -EEXIST) {
3160 			free_extent_map(em);
3161 			break;
3162 		}
3163 		btrfs_drop_extent_cache(inode, em->start,
3164 					em->start + em->len - 1, 0);
3165 	}
3166 
3167 	if (unlikely(ret))
3168 		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3169 			&BTRFS_I(inode)->runtime_flags);
3170 }
3171 
3172 /**
3173  * btrfs_clone() - clone a range from inode file to another
3174  *
3175  * @src: Inode to clone from
3176  * @inode: Inode to clone to
3177  * @off: Offset within source to start clone from
3178  * @olen: Original length, passed by user, of range to clone
3179  * @olen_aligned: Block-aligned value of olen, extent_same uses
3180  *               identical values here
3181  * @destoff: Offset within @inode to start clone
3182  */
3183 static int btrfs_clone(struct inode *src, struct inode *inode,
3184 		       const u64 off, const u64 olen, const u64 olen_aligned,
3185 		       const u64 destoff)
3186 {
3187 	struct btrfs_root *root = BTRFS_I(inode)->root;
3188 	struct btrfs_path *path = NULL;
3189 	struct extent_buffer *leaf;
3190 	struct btrfs_trans_handle *trans;
3191 	char *buf = NULL;
3192 	struct btrfs_key key;
3193 	u32 nritems;
3194 	int slot;
3195 	int ret;
3196 	int no_quota;
3197 	const u64 len = olen_aligned;
3198 	u64 last_disko = 0;
3199 	u64 last_dest_end = destoff;
3200 
3201 	ret = -ENOMEM;
3202 	buf = vmalloc(btrfs_level_size(root, 0));
3203 	if (!buf)
3204 		return ret;
3205 
3206 	path = btrfs_alloc_path();
3207 	if (!path) {
3208 		vfree(buf);
3209 		return ret;
3210 	}
3211 
3212 	path->reada = 2;
3213 	/* clone data */
3214 	key.objectid = btrfs_ino(src);
3215 	key.type = BTRFS_EXTENT_DATA_KEY;
3216 	key.offset = off;
3217 
3218 	while (1) {
3219 		/*
3220 		 * note the key will change type as we walk through the
3221 		 * tree.
3222 		 */
3223 		path->leave_spinning = 1;
3224 		ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3225 				0, 0);
3226 		if (ret < 0)
3227 			goto out;
3228 		/*
3229 		 * First search, if no extent item that starts at offset off was
3230 		 * found but the previous item is an extent item, it's possible
3231 		 * it might overlap our target range, therefore process it.
3232 		 */
3233 		if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3234 			btrfs_item_key_to_cpu(path->nodes[0], &key,
3235 					      path->slots[0] - 1);
3236 			if (key.type == BTRFS_EXTENT_DATA_KEY)
3237 				path->slots[0]--;
3238 		}
3239 
3240 		nritems = btrfs_header_nritems(path->nodes[0]);
3241 process_slot:
3242 		no_quota = 1;
3243 		if (path->slots[0] >= nritems) {
3244 			ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3245 			if (ret < 0)
3246 				goto out;
3247 			if (ret > 0)
3248 				break;
3249 			nritems = btrfs_header_nritems(path->nodes[0]);
3250 		}
3251 		leaf = path->nodes[0];
3252 		slot = path->slots[0];
3253 
3254 		btrfs_item_key_to_cpu(leaf, &key, slot);
3255 		if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
3256 		    key.objectid != btrfs_ino(src))
3257 			break;
3258 
3259 		if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
3260 			struct btrfs_file_extent_item *extent;
3261 			int type;
3262 			u32 size;
3263 			struct btrfs_key new_key;
3264 			u64 disko = 0, diskl = 0;
3265 			u64 datao = 0, datal = 0;
3266 			u8 comp;
3267 			u64 drop_start;
3268 
3269 			extent = btrfs_item_ptr(leaf, slot,
3270 						struct btrfs_file_extent_item);
3271 			comp = btrfs_file_extent_compression(leaf, extent);
3272 			type = btrfs_file_extent_type(leaf, extent);
3273 			if (type == BTRFS_FILE_EXTENT_REG ||
3274 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
3275 				disko = btrfs_file_extent_disk_bytenr(leaf,
3276 								      extent);
3277 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
3278 								 extent);
3279 				datao = btrfs_file_extent_offset(leaf, extent);
3280 				datal = btrfs_file_extent_num_bytes(leaf,
3281 								    extent);
3282 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
3283 				/* take upper bound, may be compressed */
3284 				datal = btrfs_file_extent_ram_bytes(leaf,
3285 								    extent);
3286 			}
3287 
3288 			/*
3289 			 * The first search might have left us at an extent
3290 			 * item that ends before our target range's start, can
3291 			 * happen if we have holes and NO_HOLES feature enabled.
3292 			 */
3293 			if (key.offset + datal <= off) {
3294 				path->slots[0]++;
3295 				goto process_slot;
3296 			} else if (key.offset >= off + len) {
3297 				break;
3298 			}
3299 
3300 			size = btrfs_item_size_nr(leaf, slot);
3301 			read_extent_buffer(leaf, buf,
3302 					   btrfs_item_ptr_offset(leaf, slot),
3303 					   size);
3304 
3305 			btrfs_release_path(path);
3306 			path->leave_spinning = 0;
3307 
3308 			memcpy(&new_key, &key, sizeof(new_key));
3309 			new_key.objectid = btrfs_ino(inode);
3310 			if (off <= key.offset)
3311 				new_key.offset = key.offset + destoff - off;
3312 			else
3313 				new_key.offset = destoff;
3314 
3315 			/*
3316 			 * Deal with a hole that doesn't have an extent item
3317 			 * that represents it (NO_HOLES feature enabled).
3318 			 * This hole is either in the middle of the cloning
3319 			 * range or at the beginning (fully overlaps it or
3320 			 * partially overlaps it).
3321 			 */
3322 			if (new_key.offset != last_dest_end)
3323 				drop_start = last_dest_end;
3324 			else
3325 				drop_start = new_key.offset;
3326 
3327 			/*
3328 			 * 1 - adjusting old extent (we may have to split it)
3329 			 * 1 - add new extent
3330 			 * 1 - inode update
3331 			 */
3332 			trans = btrfs_start_transaction(root, 3);
3333 			if (IS_ERR(trans)) {
3334 				ret = PTR_ERR(trans);
3335 				goto out;
3336 			}
3337 
3338 			if (type == BTRFS_FILE_EXTENT_REG ||
3339 			    type == BTRFS_FILE_EXTENT_PREALLOC) {
3340 				/*
3341 				 *    a  | --- range to clone ---|  b
3342 				 * | ------------- extent ------------- |
3343 				 */
3344 
3345 				/* subtract range b */
3346 				if (key.offset + datal > off + len)
3347 					datal = off + len - key.offset;
3348 
3349 				/* subtract range a */
3350 				if (off > key.offset) {
3351 					datao += off - key.offset;
3352 					datal -= off - key.offset;
3353 				}
3354 
3355 				ret = btrfs_drop_extents(trans, root, inode,
3356 							 drop_start,
3357 							 new_key.offset + datal,
3358 							 1);
3359 				if (ret) {
3360 					if (ret != -EOPNOTSUPP)
3361 						btrfs_abort_transaction(trans,
3362 								root, ret);
3363 					btrfs_end_transaction(trans, root);
3364 					goto out;
3365 				}
3366 
3367 				ret = btrfs_insert_empty_item(trans, root, path,
3368 							      &new_key, size);
3369 				if (ret) {
3370 					btrfs_abort_transaction(trans, root,
3371 								ret);
3372 					btrfs_end_transaction(trans, root);
3373 					goto out;
3374 				}
3375 
3376 				leaf = path->nodes[0];
3377 				slot = path->slots[0];
3378 				write_extent_buffer(leaf, buf,
3379 					    btrfs_item_ptr_offset(leaf, slot),
3380 					    size);
3381 
3382 				extent = btrfs_item_ptr(leaf, slot,
3383 						struct btrfs_file_extent_item);
3384 
3385 				/* disko == 0 means it's a hole */
3386 				if (!disko)
3387 					datao = 0;
3388 
3389 				btrfs_set_file_extent_offset(leaf, extent,
3390 							     datao);
3391 				btrfs_set_file_extent_num_bytes(leaf, extent,
3392 								datal);
3393 
3394 				/*
3395 				 * We need to look up the roots that point at
3396 				 * this bytenr and see if the new root does.  If
3397 				 * it does not we need to make sure we update
3398 				 * quotas appropriately.
3399 				 */
3400 				if (disko && root != BTRFS_I(src)->root &&
3401 				    disko != last_disko) {
3402 					no_quota = check_ref(trans, root,
3403 							     disko);
3404 					if (no_quota < 0) {
3405 						btrfs_abort_transaction(trans,
3406 									root,
3407 									ret);
3408 						btrfs_end_transaction(trans,
3409 								      root);
3410 						ret = no_quota;
3411 						goto out;
3412 					}
3413 				}
3414 
3415 				if (disko) {
3416 					inode_add_bytes(inode, datal);
3417 					ret = btrfs_inc_extent_ref(trans, root,
3418 							disko, diskl, 0,
3419 							root->root_key.objectid,
3420 							btrfs_ino(inode),
3421 							new_key.offset - datao,
3422 							no_quota);
3423 					if (ret) {
3424 						btrfs_abort_transaction(trans,
3425 									root,
3426 									ret);
3427 						btrfs_end_transaction(trans,
3428 								      root);
3429 						goto out;
3430 
3431 					}
3432 				}
3433 			} else if (type == BTRFS_FILE_EXTENT_INLINE) {
3434 				u64 skip = 0;
3435 				u64 trim = 0;
3436 				u64 aligned_end = 0;
3437 
3438 				if (off > key.offset) {
3439 					skip = off - key.offset;
3440 					new_key.offset += skip;
3441 				}
3442 
3443 				if (key.offset + datal > off + len)
3444 					trim = key.offset + datal - (off + len);
3445 
3446 				if (comp && (skip || trim)) {
3447 					ret = -EINVAL;
3448 					btrfs_end_transaction(trans, root);
3449 					goto out;
3450 				}
3451 				size -= skip + trim;
3452 				datal -= skip + trim;
3453 
3454 				aligned_end = ALIGN(new_key.offset + datal,
3455 						    root->sectorsize);
3456 				ret = btrfs_drop_extents(trans, root, inode,
3457 							 drop_start,
3458 							 aligned_end,
3459 							 1);
3460 				if (ret) {
3461 					if (ret != -EOPNOTSUPP)
3462 						btrfs_abort_transaction(trans,
3463 							root, ret);
3464 					btrfs_end_transaction(trans, root);
3465 					goto out;
3466 				}
3467 
3468 				ret = btrfs_insert_empty_item(trans, root, path,
3469 							      &new_key, size);
3470 				if (ret) {
3471 					btrfs_abort_transaction(trans, root,
3472 								ret);
3473 					btrfs_end_transaction(trans, root);
3474 					goto out;
3475 				}
3476 
3477 				if (skip) {
3478 					u32 start =
3479 					  btrfs_file_extent_calc_inline_size(0);
3480 					memmove(buf+start, buf+start+skip,
3481 						datal);
3482 				}
3483 
3484 				leaf = path->nodes[0];
3485 				slot = path->slots[0];
3486 				write_extent_buffer(leaf, buf,
3487 					    btrfs_item_ptr_offset(leaf, slot),
3488 					    size);
3489 				inode_add_bytes(inode, datal);
3490 			}
3491 
3492 			/* If we have an implicit hole (NO_HOLES feature). */
3493 			if (drop_start < new_key.offset)
3494 				clone_update_extent_map(inode, trans,
3495 						NULL, drop_start,
3496 						new_key.offset - drop_start);
3497 
3498 			clone_update_extent_map(inode, trans, path, 0, 0);
3499 
3500 			btrfs_mark_buffer_dirty(leaf);
3501 			btrfs_release_path(path);
3502 
3503 			last_dest_end = ALIGN(new_key.offset + datal,
3504 					      root->sectorsize);
3505 			ret = clone_finish_inode_update(trans, inode,
3506 							last_dest_end,
3507 							destoff, olen);
3508 			if (ret)
3509 				goto out;
3510 			if (new_key.offset + datal >= destoff + len)
3511 				break;
3512 		}
3513 		btrfs_release_path(path);
3514 		key.offset++;
3515 	}
3516 	ret = 0;
3517 
3518 	if (last_dest_end < destoff + len) {
3519 		/*
3520 		 * We have an implicit hole (NO_HOLES feature is enabled) that
3521 		 * fully or partially overlaps our cloning range at its end.
3522 		 */
3523 		btrfs_release_path(path);
3524 
3525 		/*
3526 		 * 1 - remove extent(s)
3527 		 * 1 - inode update
3528 		 */
3529 		trans = btrfs_start_transaction(root, 2);
3530 		if (IS_ERR(trans)) {
3531 			ret = PTR_ERR(trans);
3532 			goto out;
3533 		}
3534 		ret = btrfs_drop_extents(trans, root, inode,
3535 					 last_dest_end, destoff + len, 1);
3536 		if (ret) {
3537 			if (ret != -EOPNOTSUPP)
3538 				btrfs_abort_transaction(trans, root, ret);
3539 			btrfs_end_transaction(trans, root);
3540 			goto out;
3541 		}
3542 		clone_update_extent_map(inode, trans, NULL, last_dest_end,
3543 					destoff + len - last_dest_end);
3544 		ret = clone_finish_inode_update(trans, inode, destoff + len,
3545 						destoff, olen);
3546 	}
3547 
3548 out:
3549 	btrfs_free_path(path);
3550 	vfree(buf);
3551 	return ret;
3552 }
3553 
3554 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3555 				       u64 off, u64 olen, u64 destoff)
3556 {
3557 	struct inode *inode = file_inode(file);
3558 	struct btrfs_root *root = BTRFS_I(inode)->root;
3559 	struct fd src_file;
3560 	struct inode *src;
3561 	int ret;
3562 	u64 len = olen;
3563 	u64 bs = root->fs_info->sb->s_blocksize;
3564 	int same_inode = 0;
3565 
3566 	/*
3567 	 * TODO:
3568 	 * - split compressed inline extents.  annoying: we need to
3569 	 *   decompress into destination's address_space (the file offset
3570 	 *   may change, so source mapping won't do), then recompress (or
3571 	 *   otherwise reinsert) a subrange.
3572 	 *
3573 	 * - split destination inode's inline extents.  The inline extents can
3574 	 *   be either compressed or non-compressed.
3575 	 */
3576 
3577 	/* the destination must be opened for writing */
3578 	if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3579 		return -EINVAL;
3580 
3581 	if (btrfs_root_readonly(root))
3582 		return -EROFS;
3583 
3584 	ret = mnt_want_write_file(file);
3585 	if (ret)
3586 		return ret;
3587 
3588 	src_file = fdget(srcfd);
3589 	if (!src_file.file) {
3590 		ret = -EBADF;
3591 		goto out_drop_write;
3592 	}
3593 
3594 	ret = -EXDEV;
3595 	if (src_file.file->f_path.mnt != file->f_path.mnt)
3596 		goto out_fput;
3597 
3598 	src = file_inode(src_file.file);
3599 
3600 	ret = -EINVAL;
3601 	if (src == inode)
3602 		same_inode = 1;
3603 
3604 	/* the src must be open for reading */
3605 	if (!(src_file.file->f_mode & FMODE_READ))
3606 		goto out_fput;
3607 
3608 	/* don't make the dst file partly checksummed */
3609 	if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3610 	    (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3611 		goto out_fput;
3612 
3613 	ret = -EISDIR;
3614 	if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3615 		goto out_fput;
3616 
3617 	ret = -EXDEV;
3618 	if (src->i_sb != inode->i_sb)
3619 		goto out_fput;
3620 
3621 	if (!same_inode) {
3622 		if (inode < src) {
3623 			mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3624 			mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3625 		} else {
3626 			mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3627 			mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3628 		}
3629 	} else {
3630 		mutex_lock(&src->i_mutex);
3631 	}
3632 
3633 	/* determine range to clone */
3634 	ret = -EINVAL;
3635 	if (off + len > src->i_size || off + len < off)
3636 		goto out_unlock;
3637 	if (len == 0)
3638 		olen = len = src->i_size - off;
3639 	/* if we extend to eof, continue to block boundary */
3640 	if (off + len == src->i_size)
3641 		len = ALIGN(src->i_size, bs) - off;
3642 
3643 	/* verify the end result is block aligned */
3644 	if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3645 	    !IS_ALIGNED(destoff, bs))
3646 		goto out_unlock;
3647 
3648 	/* verify if ranges are overlapped within the same file */
3649 	if (same_inode) {
3650 		if (destoff + len > off && destoff < off + len)
3651 			goto out_unlock;
3652 	}
3653 
3654 	if (destoff > inode->i_size) {
3655 		ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3656 		if (ret)
3657 			goto out_unlock;
3658 	}
3659 
3660 	/*
3661 	 * Lock the target range too. Right after we replace the file extent
3662 	 * items in the fs tree (which now point to the cloned data), we might
3663 	 * have a worker replace them with extent items relative to a write
3664 	 * operation that was issued before this clone operation (i.e. confront
3665 	 * with inode.c:btrfs_finish_ordered_io).
3666 	 */
3667 	if (same_inode) {
3668 		u64 lock_start = min_t(u64, off, destoff);
3669 		u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3670 
3671 		lock_extent_range(src, lock_start, lock_len);
3672 	} else {
3673 		lock_extent_range(src, off, len);
3674 		lock_extent_range(inode, destoff, len);
3675 	}
3676 
3677 	ret = btrfs_clone(src, inode, off, olen, len, destoff);
3678 
3679 	if (same_inode) {
3680 		u64 lock_start = min_t(u64, off, destoff);
3681 		u64 lock_end = max_t(u64, off, destoff) + len - 1;
3682 
3683 		unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3684 	} else {
3685 		unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3686 		unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3687 			      destoff + len - 1);
3688 	}
3689 	/*
3690 	 * Truncate page cache pages so that future reads will see the cloned
3691 	 * data immediately and not the previous data.
3692 	 */
3693 	truncate_inode_pages_range(&inode->i_data, destoff,
3694 				   PAGE_CACHE_ALIGN(destoff + len) - 1);
3695 out_unlock:
3696 	if (!same_inode) {
3697 		if (inode < src) {
3698 			mutex_unlock(&src->i_mutex);
3699 			mutex_unlock(&inode->i_mutex);
3700 		} else {
3701 			mutex_unlock(&inode->i_mutex);
3702 			mutex_unlock(&src->i_mutex);
3703 		}
3704 	} else {
3705 		mutex_unlock(&src->i_mutex);
3706 	}
3707 out_fput:
3708 	fdput(src_file);
3709 out_drop_write:
3710 	mnt_drop_write_file(file);
3711 	return ret;
3712 }
3713 
3714 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3715 {
3716 	struct btrfs_ioctl_clone_range_args args;
3717 
3718 	if (copy_from_user(&args, argp, sizeof(args)))
3719 		return -EFAULT;
3720 	return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3721 				 args.src_length, args.dest_offset);
3722 }
3723 
3724 /*
3725  * there are many ways the trans_start and trans_end ioctls can lead
3726  * to deadlocks.  They should only be used by applications that
3727  * basically own the machine, and have a very in depth understanding
3728  * of all the possible deadlocks and enospc problems.
3729  */
3730 static long btrfs_ioctl_trans_start(struct file *file)
3731 {
3732 	struct inode *inode = file_inode(file);
3733 	struct btrfs_root *root = BTRFS_I(inode)->root;
3734 	struct btrfs_trans_handle *trans;
3735 	int ret;
3736 
3737 	ret = -EPERM;
3738 	if (!capable(CAP_SYS_ADMIN))
3739 		goto out;
3740 
3741 	ret = -EINPROGRESS;
3742 	if (file->private_data)
3743 		goto out;
3744 
3745 	ret = -EROFS;
3746 	if (btrfs_root_readonly(root))
3747 		goto out;
3748 
3749 	ret = mnt_want_write_file(file);
3750 	if (ret)
3751 		goto out;
3752 
3753 	atomic_inc(&root->fs_info->open_ioctl_trans);
3754 
3755 	ret = -ENOMEM;
3756 	trans = btrfs_start_ioctl_transaction(root);
3757 	if (IS_ERR(trans))
3758 		goto out_drop;
3759 
3760 	file->private_data = trans;
3761 	return 0;
3762 
3763 out_drop:
3764 	atomic_dec(&root->fs_info->open_ioctl_trans);
3765 	mnt_drop_write_file(file);
3766 out:
3767 	return ret;
3768 }
3769 
3770 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3771 {
3772 	struct inode *inode = file_inode(file);
3773 	struct btrfs_root *root = BTRFS_I(inode)->root;
3774 	struct btrfs_root *new_root;
3775 	struct btrfs_dir_item *di;
3776 	struct btrfs_trans_handle *trans;
3777 	struct btrfs_path *path;
3778 	struct btrfs_key location;
3779 	struct btrfs_disk_key disk_key;
3780 	u64 objectid = 0;
3781 	u64 dir_id;
3782 	int ret;
3783 
3784 	if (!capable(CAP_SYS_ADMIN))
3785 		return -EPERM;
3786 
3787 	ret = mnt_want_write_file(file);
3788 	if (ret)
3789 		return ret;
3790 
3791 	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3792 		ret = -EFAULT;
3793 		goto out;
3794 	}
3795 
3796 	if (!objectid)
3797 		objectid = BTRFS_FS_TREE_OBJECTID;
3798 
3799 	location.objectid = objectid;
3800 	location.type = BTRFS_ROOT_ITEM_KEY;
3801 	location.offset = (u64)-1;
3802 
3803 	new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3804 	if (IS_ERR(new_root)) {
3805 		ret = PTR_ERR(new_root);
3806 		goto out;
3807 	}
3808 
3809 	path = btrfs_alloc_path();
3810 	if (!path) {
3811 		ret = -ENOMEM;
3812 		goto out;
3813 	}
3814 	path->leave_spinning = 1;
3815 
3816 	trans = btrfs_start_transaction(root, 1);
3817 	if (IS_ERR(trans)) {
3818 		btrfs_free_path(path);
3819 		ret = PTR_ERR(trans);
3820 		goto out;
3821 	}
3822 
3823 	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3824 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3825 				   dir_id, "default", 7, 1);
3826 	if (IS_ERR_OR_NULL(di)) {
3827 		btrfs_free_path(path);
3828 		btrfs_end_transaction(trans, root);
3829 		btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3830 			   "item, this isn't going to work");
3831 		ret = -ENOENT;
3832 		goto out;
3833 	}
3834 
3835 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3836 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3837 	btrfs_mark_buffer_dirty(path->nodes[0]);
3838 	btrfs_free_path(path);
3839 
3840 	btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3841 	btrfs_end_transaction(trans, root);
3842 out:
3843 	mnt_drop_write_file(file);
3844 	return ret;
3845 }
3846 
3847 void btrfs_get_block_group_info(struct list_head *groups_list,
3848 				struct btrfs_ioctl_space_info *space)
3849 {
3850 	struct btrfs_block_group_cache *block_group;
3851 
3852 	space->total_bytes = 0;
3853 	space->used_bytes = 0;
3854 	space->flags = 0;
3855 	list_for_each_entry(block_group, groups_list, list) {
3856 		space->flags = block_group->flags;
3857 		space->total_bytes += block_group->key.offset;
3858 		space->used_bytes +=
3859 			btrfs_block_group_used(&block_group->item);
3860 	}
3861 }
3862 
3863 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3864 {
3865 	struct btrfs_ioctl_space_args space_args;
3866 	struct btrfs_ioctl_space_info space;
3867 	struct btrfs_ioctl_space_info *dest;
3868 	struct btrfs_ioctl_space_info *dest_orig;
3869 	struct btrfs_ioctl_space_info __user *user_dest;
3870 	struct btrfs_space_info *info;
3871 	u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3872 		       BTRFS_BLOCK_GROUP_SYSTEM,
3873 		       BTRFS_BLOCK_GROUP_METADATA,
3874 		       BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3875 	int num_types = 4;
3876 	int alloc_size;
3877 	int ret = 0;
3878 	u64 slot_count = 0;
3879 	int i, c;
3880 
3881 	if (copy_from_user(&space_args,
3882 			   (struct btrfs_ioctl_space_args __user *)arg,
3883 			   sizeof(space_args)))
3884 		return -EFAULT;
3885 
3886 	for (i = 0; i < num_types; i++) {
3887 		struct btrfs_space_info *tmp;
3888 
3889 		info = NULL;
3890 		rcu_read_lock();
3891 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3892 					list) {
3893 			if (tmp->flags == types[i]) {
3894 				info = tmp;
3895 				break;
3896 			}
3897 		}
3898 		rcu_read_unlock();
3899 
3900 		if (!info)
3901 			continue;
3902 
3903 		down_read(&info->groups_sem);
3904 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3905 			if (!list_empty(&info->block_groups[c]))
3906 				slot_count++;
3907 		}
3908 		up_read(&info->groups_sem);
3909 	}
3910 
3911 	/*
3912 	 * Global block reserve, exported as a space_info
3913 	 */
3914 	slot_count++;
3915 
3916 	/* space_slots == 0 means they are asking for a count */
3917 	if (space_args.space_slots == 0) {
3918 		space_args.total_spaces = slot_count;
3919 		goto out;
3920 	}
3921 
3922 	slot_count = min_t(u64, space_args.space_slots, slot_count);
3923 
3924 	alloc_size = sizeof(*dest) * slot_count;
3925 
3926 	/* we generally have at most 6 or so space infos, one for each raid
3927 	 * level.  So, a whole page should be more than enough for everyone
3928 	 */
3929 	if (alloc_size > PAGE_CACHE_SIZE)
3930 		return -ENOMEM;
3931 
3932 	space_args.total_spaces = 0;
3933 	dest = kmalloc(alloc_size, GFP_NOFS);
3934 	if (!dest)
3935 		return -ENOMEM;
3936 	dest_orig = dest;
3937 
3938 	/* now we have a buffer to copy into */
3939 	for (i = 0; i < num_types; i++) {
3940 		struct btrfs_space_info *tmp;
3941 
3942 		if (!slot_count)
3943 			break;
3944 
3945 		info = NULL;
3946 		rcu_read_lock();
3947 		list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3948 					list) {
3949 			if (tmp->flags == types[i]) {
3950 				info = tmp;
3951 				break;
3952 			}
3953 		}
3954 		rcu_read_unlock();
3955 
3956 		if (!info)
3957 			continue;
3958 		down_read(&info->groups_sem);
3959 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3960 			if (!list_empty(&info->block_groups[c])) {
3961 				btrfs_get_block_group_info(
3962 					&info->block_groups[c], &space);
3963 				memcpy(dest, &space, sizeof(space));
3964 				dest++;
3965 				space_args.total_spaces++;
3966 				slot_count--;
3967 			}
3968 			if (!slot_count)
3969 				break;
3970 		}
3971 		up_read(&info->groups_sem);
3972 	}
3973 
3974 	/*
3975 	 * Add global block reserve
3976 	 */
3977 	if (slot_count) {
3978 		struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3979 
3980 		spin_lock(&block_rsv->lock);
3981 		space.total_bytes = block_rsv->size;
3982 		space.used_bytes = block_rsv->size - block_rsv->reserved;
3983 		spin_unlock(&block_rsv->lock);
3984 		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3985 		memcpy(dest, &space, sizeof(space));
3986 		space_args.total_spaces++;
3987 	}
3988 
3989 	user_dest = (struct btrfs_ioctl_space_info __user *)
3990 		(arg + sizeof(struct btrfs_ioctl_space_args));
3991 
3992 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3993 		ret = -EFAULT;
3994 
3995 	kfree(dest_orig);
3996 out:
3997 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3998 		ret = -EFAULT;
3999 
4000 	return ret;
4001 }
4002 
4003 /*
4004  * there are many ways the trans_start and trans_end ioctls can lead
4005  * to deadlocks.  They should only be used by applications that
4006  * basically own the machine, and have a very in depth understanding
4007  * of all the possible deadlocks and enospc problems.
4008  */
4009 long btrfs_ioctl_trans_end(struct file *file)
4010 {
4011 	struct inode *inode = file_inode(file);
4012 	struct btrfs_root *root = BTRFS_I(inode)->root;
4013 	struct btrfs_trans_handle *trans;
4014 
4015 	trans = file->private_data;
4016 	if (!trans)
4017 		return -EINVAL;
4018 	file->private_data = NULL;
4019 
4020 	btrfs_end_transaction(trans, root);
4021 
4022 	atomic_dec(&root->fs_info->open_ioctl_trans);
4023 
4024 	mnt_drop_write_file(file);
4025 	return 0;
4026 }
4027 
4028 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4029 					    void __user *argp)
4030 {
4031 	struct btrfs_trans_handle *trans;
4032 	u64 transid;
4033 	int ret;
4034 
4035 	trans = btrfs_attach_transaction_barrier(root);
4036 	if (IS_ERR(trans)) {
4037 		if (PTR_ERR(trans) != -ENOENT)
4038 			return PTR_ERR(trans);
4039 
4040 		/* No running transaction, don't bother */
4041 		transid = root->fs_info->last_trans_committed;
4042 		goto out;
4043 	}
4044 	transid = trans->transid;
4045 	ret = btrfs_commit_transaction_async(trans, root, 0);
4046 	if (ret) {
4047 		btrfs_end_transaction(trans, root);
4048 		return ret;
4049 	}
4050 out:
4051 	if (argp)
4052 		if (copy_to_user(argp, &transid, sizeof(transid)))
4053 			return -EFAULT;
4054 	return 0;
4055 }
4056 
4057 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4058 					   void __user *argp)
4059 {
4060 	u64 transid;
4061 
4062 	if (argp) {
4063 		if (copy_from_user(&transid, argp, sizeof(transid)))
4064 			return -EFAULT;
4065 	} else {
4066 		transid = 0;  /* current trans */
4067 	}
4068 	return btrfs_wait_for_commit(root, transid);
4069 }
4070 
4071 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4072 {
4073 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4074 	struct btrfs_ioctl_scrub_args *sa;
4075 	int ret;
4076 
4077 	if (!capable(CAP_SYS_ADMIN))
4078 		return -EPERM;
4079 
4080 	sa = memdup_user(arg, sizeof(*sa));
4081 	if (IS_ERR(sa))
4082 		return PTR_ERR(sa);
4083 
4084 	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4085 		ret = mnt_want_write_file(file);
4086 		if (ret)
4087 			goto out;
4088 	}
4089 
4090 	ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4091 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4092 			      0);
4093 
4094 	if (copy_to_user(arg, sa, sizeof(*sa)))
4095 		ret = -EFAULT;
4096 
4097 	if (!(sa->flags & BTRFS_SCRUB_READONLY))
4098 		mnt_drop_write_file(file);
4099 out:
4100 	kfree(sa);
4101 	return ret;
4102 }
4103 
4104 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4105 {
4106 	if (!capable(CAP_SYS_ADMIN))
4107 		return -EPERM;
4108 
4109 	return btrfs_scrub_cancel(root->fs_info);
4110 }
4111 
4112 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4113 				       void __user *arg)
4114 {
4115 	struct btrfs_ioctl_scrub_args *sa;
4116 	int ret;
4117 
4118 	if (!capable(CAP_SYS_ADMIN))
4119 		return -EPERM;
4120 
4121 	sa = memdup_user(arg, sizeof(*sa));
4122 	if (IS_ERR(sa))
4123 		return PTR_ERR(sa);
4124 
4125 	ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4126 
4127 	if (copy_to_user(arg, sa, sizeof(*sa)))
4128 		ret = -EFAULT;
4129 
4130 	kfree(sa);
4131 	return ret;
4132 }
4133 
4134 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4135 				      void __user *arg)
4136 {
4137 	struct btrfs_ioctl_get_dev_stats *sa;
4138 	int ret;
4139 
4140 	sa = memdup_user(arg, sizeof(*sa));
4141 	if (IS_ERR(sa))
4142 		return PTR_ERR(sa);
4143 
4144 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4145 		kfree(sa);
4146 		return -EPERM;
4147 	}
4148 
4149 	ret = btrfs_get_dev_stats(root, sa);
4150 
4151 	if (copy_to_user(arg, sa, sizeof(*sa)))
4152 		ret = -EFAULT;
4153 
4154 	kfree(sa);
4155 	return ret;
4156 }
4157 
4158 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4159 {
4160 	struct btrfs_ioctl_dev_replace_args *p;
4161 	int ret;
4162 
4163 	if (!capable(CAP_SYS_ADMIN))
4164 		return -EPERM;
4165 
4166 	p = memdup_user(arg, sizeof(*p));
4167 	if (IS_ERR(p))
4168 		return PTR_ERR(p);
4169 
4170 	switch (p->cmd) {
4171 	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4172 		if (root->fs_info->sb->s_flags & MS_RDONLY) {
4173 			ret = -EROFS;
4174 			goto out;
4175 		}
4176 		if (atomic_xchg(
4177 			&root->fs_info->mutually_exclusive_operation_running,
4178 			1)) {
4179 			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4180 		} else {
4181 			ret = btrfs_dev_replace_start(root, p);
4182 			atomic_set(
4183 			 &root->fs_info->mutually_exclusive_operation_running,
4184 			 0);
4185 		}
4186 		break;
4187 	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4188 		btrfs_dev_replace_status(root->fs_info, p);
4189 		ret = 0;
4190 		break;
4191 	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4192 		ret = btrfs_dev_replace_cancel(root->fs_info, p);
4193 		break;
4194 	default:
4195 		ret = -EINVAL;
4196 		break;
4197 	}
4198 
4199 	if (copy_to_user(arg, p, sizeof(*p)))
4200 		ret = -EFAULT;
4201 out:
4202 	kfree(p);
4203 	return ret;
4204 }
4205 
4206 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4207 {
4208 	int ret = 0;
4209 	int i;
4210 	u64 rel_ptr;
4211 	int size;
4212 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
4213 	struct inode_fs_paths *ipath = NULL;
4214 	struct btrfs_path *path;
4215 
4216 	if (!capable(CAP_DAC_READ_SEARCH))
4217 		return -EPERM;
4218 
4219 	path = btrfs_alloc_path();
4220 	if (!path) {
4221 		ret = -ENOMEM;
4222 		goto out;
4223 	}
4224 
4225 	ipa = memdup_user(arg, sizeof(*ipa));
4226 	if (IS_ERR(ipa)) {
4227 		ret = PTR_ERR(ipa);
4228 		ipa = NULL;
4229 		goto out;
4230 	}
4231 
4232 	size = min_t(u32, ipa->size, 4096);
4233 	ipath = init_ipath(size, root, path);
4234 	if (IS_ERR(ipath)) {
4235 		ret = PTR_ERR(ipath);
4236 		ipath = NULL;
4237 		goto out;
4238 	}
4239 
4240 	ret = paths_from_inode(ipa->inum, ipath);
4241 	if (ret < 0)
4242 		goto out;
4243 
4244 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4245 		rel_ptr = ipath->fspath->val[i] -
4246 			  (u64)(unsigned long)ipath->fspath->val;
4247 		ipath->fspath->val[i] = rel_ptr;
4248 	}
4249 
4250 	ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4251 			   (void *)(unsigned long)ipath->fspath, size);
4252 	if (ret) {
4253 		ret = -EFAULT;
4254 		goto out;
4255 	}
4256 
4257 out:
4258 	btrfs_free_path(path);
4259 	free_ipath(ipath);
4260 	kfree(ipa);
4261 
4262 	return ret;
4263 }
4264 
4265 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4266 {
4267 	struct btrfs_data_container *inodes = ctx;
4268 	const size_t c = 3 * sizeof(u64);
4269 
4270 	if (inodes->bytes_left >= c) {
4271 		inodes->bytes_left -= c;
4272 		inodes->val[inodes->elem_cnt] = inum;
4273 		inodes->val[inodes->elem_cnt + 1] = offset;
4274 		inodes->val[inodes->elem_cnt + 2] = root;
4275 		inodes->elem_cnt += 3;
4276 	} else {
4277 		inodes->bytes_missing += c - inodes->bytes_left;
4278 		inodes->bytes_left = 0;
4279 		inodes->elem_missed += 3;
4280 	}
4281 
4282 	return 0;
4283 }
4284 
4285 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4286 					void __user *arg)
4287 {
4288 	int ret = 0;
4289 	int size;
4290 	struct btrfs_ioctl_logical_ino_args *loi;
4291 	struct btrfs_data_container *inodes = NULL;
4292 	struct btrfs_path *path = NULL;
4293 
4294 	if (!capable(CAP_SYS_ADMIN))
4295 		return -EPERM;
4296 
4297 	loi = memdup_user(arg, sizeof(*loi));
4298 	if (IS_ERR(loi)) {
4299 		ret = PTR_ERR(loi);
4300 		loi = NULL;
4301 		goto out;
4302 	}
4303 
4304 	path = btrfs_alloc_path();
4305 	if (!path) {
4306 		ret = -ENOMEM;
4307 		goto out;
4308 	}
4309 
4310 	size = min_t(u32, loi->size, 64 * 1024);
4311 	inodes = init_data_container(size);
4312 	if (IS_ERR(inodes)) {
4313 		ret = PTR_ERR(inodes);
4314 		inodes = NULL;
4315 		goto out;
4316 	}
4317 
4318 	ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4319 					  build_ino_list, inodes);
4320 	if (ret == -EINVAL)
4321 		ret = -ENOENT;
4322 	if (ret < 0)
4323 		goto out;
4324 
4325 	ret = copy_to_user((void *)(unsigned long)loi->inodes,
4326 			   (void *)(unsigned long)inodes, size);
4327 	if (ret)
4328 		ret = -EFAULT;
4329 
4330 out:
4331 	btrfs_free_path(path);
4332 	vfree(inodes);
4333 	kfree(loi);
4334 
4335 	return ret;
4336 }
4337 
4338 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4339 			       struct btrfs_ioctl_balance_args *bargs)
4340 {
4341 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4342 
4343 	bargs->flags = bctl->flags;
4344 
4345 	if (atomic_read(&fs_info->balance_running))
4346 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4347 	if (atomic_read(&fs_info->balance_pause_req))
4348 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4349 	if (atomic_read(&fs_info->balance_cancel_req))
4350 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4351 
4352 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4353 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4354 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4355 
4356 	if (lock) {
4357 		spin_lock(&fs_info->balance_lock);
4358 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4359 		spin_unlock(&fs_info->balance_lock);
4360 	} else {
4361 		memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4362 	}
4363 }
4364 
4365 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4366 {
4367 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4368 	struct btrfs_fs_info *fs_info = root->fs_info;
4369 	struct btrfs_ioctl_balance_args *bargs;
4370 	struct btrfs_balance_control *bctl;
4371 	bool need_unlock; /* for mut. excl. ops lock */
4372 	int ret;
4373 
4374 	if (!capable(CAP_SYS_ADMIN))
4375 		return -EPERM;
4376 
4377 	ret = mnt_want_write_file(file);
4378 	if (ret)
4379 		return ret;
4380 
4381 again:
4382 	if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4383 		mutex_lock(&fs_info->volume_mutex);
4384 		mutex_lock(&fs_info->balance_mutex);
4385 		need_unlock = true;
4386 		goto locked;
4387 	}
4388 
4389 	/*
4390 	 * mut. excl. ops lock is locked.  Three possibilites:
4391 	 *   (1) some other op is running
4392 	 *   (2) balance is running
4393 	 *   (3) balance is paused -- special case (think resume)
4394 	 */
4395 	mutex_lock(&fs_info->balance_mutex);
4396 	if (fs_info->balance_ctl) {
4397 		/* this is either (2) or (3) */
4398 		if (!atomic_read(&fs_info->balance_running)) {
4399 			mutex_unlock(&fs_info->balance_mutex);
4400 			if (!mutex_trylock(&fs_info->volume_mutex))
4401 				goto again;
4402 			mutex_lock(&fs_info->balance_mutex);
4403 
4404 			if (fs_info->balance_ctl &&
4405 			    !atomic_read(&fs_info->balance_running)) {
4406 				/* this is (3) */
4407 				need_unlock = false;
4408 				goto locked;
4409 			}
4410 
4411 			mutex_unlock(&fs_info->balance_mutex);
4412 			mutex_unlock(&fs_info->volume_mutex);
4413 			goto again;
4414 		} else {
4415 			/* this is (2) */
4416 			mutex_unlock(&fs_info->balance_mutex);
4417 			ret = -EINPROGRESS;
4418 			goto out;
4419 		}
4420 	} else {
4421 		/* this is (1) */
4422 		mutex_unlock(&fs_info->balance_mutex);
4423 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4424 		goto out;
4425 	}
4426 
4427 locked:
4428 	BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4429 
4430 	if (arg) {
4431 		bargs = memdup_user(arg, sizeof(*bargs));
4432 		if (IS_ERR(bargs)) {
4433 			ret = PTR_ERR(bargs);
4434 			goto out_unlock;
4435 		}
4436 
4437 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
4438 			if (!fs_info->balance_ctl) {
4439 				ret = -ENOTCONN;
4440 				goto out_bargs;
4441 			}
4442 
4443 			bctl = fs_info->balance_ctl;
4444 			spin_lock(&fs_info->balance_lock);
4445 			bctl->flags |= BTRFS_BALANCE_RESUME;
4446 			spin_unlock(&fs_info->balance_lock);
4447 
4448 			goto do_balance;
4449 		}
4450 	} else {
4451 		bargs = NULL;
4452 	}
4453 
4454 	if (fs_info->balance_ctl) {
4455 		ret = -EINPROGRESS;
4456 		goto out_bargs;
4457 	}
4458 
4459 	bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4460 	if (!bctl) {
4461 		ret = -ENOMEM;
4462 		goto out_bargs;
4463 	}
4464 
4465 	bctl->fs_info = fs_info;
4466 	if (arg) {
4467 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4468 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4469 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4470 
4471 		bctl->flags = bargs->flags;
4472 	} else {
4473 		/* balance everything - no filters */
4474 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4475 	}
4476 
4477 do_balance:
4478 	/*
4479 	 * Ownership of bctl and mutually_exclusive_operation_running
4480 	 * goes to to btrfs_balance.  bctl is freed in __cancel_balance,
4481 	 * or, if restriper was paused all the way until unmount, in
4482 	 * free_fs_info.  mutually_exclusive_operation_running is
4483 	 * cleared in __cancel_balance.
4484 	 */
4485 	need_unlock = false;
4486 
4487 	ret = btrfs_balance(bctl, bargs);
4488 
4489 	if (arg) {
4490 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
4491 			ret = -EFAULT;
4492 	}
4493 
4494 out_bargs:
4495 	kfree(bargs);
4496 out_unlock:
4497 	mutex_unlock(&fs_info->balance_mutex);
4498 	mutex_unlock(&fs_info->volume_mutex);
4499 	if (need_unlock)
4500 		atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4501 out:
4502 	mnt_drop_write_file(file);
4503 	return ret;
4504 }
4505 
4506 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4507 {
4508 	if (!capable(CAP_SYS_ADMIN))
4509 		return -EPERM;
4510 
4511 	switch (cmd) {
4512 	case BTRFS_BALANCE_CTL_PAUSE:
4513 		return btrfs_pause_balance(root->fs_info);
4514 	case BTRFS_BALANCE_CTL_CANCEL:
4515 		return btrfs_cancel_balance(root->fs_info);
4516 	}
4517 
4518 	return -EINVAL;
4519 }
4520 
4521 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4522 					 void __user *arg)
4523 {
4524 	struct btrfs_fs_info *fs_info = root->fs_info;
4525 	struct btrfs_ioctl_balance_args *bargs;
4526 	int ret = 0;
4527 
4528 	if (!capable(CAP_SYS_ADMIN))
4529 		return -EPERM;
4530 
4531 	mutex_lock(&fs_info->balance_mutex);
4532 	if (!fs_info->balance_ctl) {
4533 		ret = -ENOTCONN;
4534 		goto out;
4535 	}
4536 
4537 	bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4538 	if (!bargs) {
4539 		ret = -ENOMEM;
4540 		goto out;
4541 	}
4542 
4543 	update_ioctl_balance_args(fs_info, 1, bargs);
4544 
4545 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
4546 		ret = -EFAULT;
4547 
4548 	kfree(bargs);
4549 out:
4550 	mutex_unlock(&fs_info->balance_mutex);
4551 	return ret;
4552 }
4553 
4554 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4555 {
4556 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4557 	struct btrfs_ioctl_quota_ctl_args *sa;
4558 	struct btrfs_trans_handle *trans = NULL;
4559 	int ret;
4560 	int err;
4561 
4562 	if (!capable(CAP_SYS_ADMIN))
4563 		return -EPERM;
4564 
4565 	ret = mnt_want_write_file(file);
4566 	if (ret)
4567 		return ret;
4568 
4569 	sa = memdup_user(arg, sizeof(*sa));
4570 	if (IS_ERR(sa)) {
4571 		ret = PTR_ERR(sa);
4572 		goto drop_write;
4573 	}
4574 
4575 	down_write(&root->fs_info->subvol_sem);
4576 	trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4577 	if (IS_ERR(trans)) {
4578 		ret = PTR_ERR(trans);
4579 		goto out;
4580 	}
4581 
4582 	switch (sa->cmd) {
4583 	case BTRFS_QUOTA_CTL_ENABLE:
4584 		ret = btrfs_quota_enable(trans, root->fs_info);
4585 		break;
4586 	case BTRFS_QUOTA_CTL_DISABLE:
4587 		ret = btrfs_quota_disable(trans, root->fs_info);
4588 		break;
4589 	default:
4590 		ret = -EINVAL;
4591 		break;
4592 	}
4593 
4594 	err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4595 	if (err && !ret)
4596 		ret = err;
4597 out:
4598 	kfree(sa);
4599 	up_write(&root->fs_info->subvol_sem);
4600 drop_write:
4601 	mnt_drop_write_file(file);
4602 	return ret;
4603 }
4604 
4605 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4606 {
4607 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4608 	struct btrfs_ioctl_qgroup_assign_args *sa;
4609 	struct btrfs_trans_handle *trans;
4610 	int ret;
4611 	int err;
4612 
4613 	if (!capable(CAP_SYS_ADMIN))
4614 		return -EPERM;
4615 
4616 	ret = mnt_want_write_file(file);
4617 	if (ret)
4618 		return ret;
4619 
4620 	sa = memdup_user(arg, sizeof(*sa));
4621 	if (IS_ERR(sa)) {
4622 		ret = PTR_ERR(sa);
4623 		goto drop_write;
4624 	}
4625 
4626 	trans = btrfs_join_transaction(root);
4627 	if (IS_ERR(trans)) {
4628 		ret = PTR_ERR(trans);
4629 		goto out;
4630 	}
4631 
4632 	/* FIXME: check if the IDs really exist */
4633 	if (sa->assign) {
4634 		ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4635 						sa->src, sa->dst);
4636 	} else {
4637 		ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4638 						sa->src, sa->dst);
4639 	}
4640 
4641 	err = btrfs_end_transaction(trans, root);
4642 	if (err && !ret)
4643 		ret = err;
4644 
4645 out:
4646 	kfree(sa);
4647 drop_write:
4648 	mnt_drop_write_file(file);
4649 	return ret;
4650 }
4651 
4652 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4653 {
4654 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4655 	struct btrfs_ioctl_qgroup_create_args *sa;
4656 	struct btrfs_trans_handle *trans;
4657 	int ret;
4658 	int err;
4659 
4660 	if (!capable(CAP_SYS_ADMIN))
4661 		return -EPERM;
4662 
4663 	ret = mnt_want_write_file(file);
4664 	if (ret)
4665 		return ret;
4666 
4667 	sa = memdup_user(arg, sizeof(*sa));
4668 	if (IS_ERR(sa)) {
4669 		ret = PTR_ERR(sa);
4670 		goto drop_write;
4671 	}
4672 
4673 	if (!sa->qgroupid) {
4674 		ret = -EINVAL;
4675 		goto out;
4676 	}
4677 
4678 	trans = btrfs_join_transaction(root);
4679 	if (IS_ERR(trans)) {
4680 		ret = PTR_ERR(trans);
4681 		goto out;
4682 	}
4683 
4684 	/* FIXME: check if the IDs really exist */
4685 	if (sa->create) {
4686 		ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4687 					  NULL);
4688 	} else {
4689 		ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4690 	}
4691 
4692 	err = btrfs_end_transaction(trans, root);
4693 	if (err && !ret)
4694 		ret = err;
4695 
4696 out:
4697 	kfree(sa);
4698 drop_write:
4699 	mnt_drop_write_file(file);
4700 	return ret;
4701 }
4702 
4703 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4704 {
4705 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4706 	struct btrfs_ioctl_qgroup_limit_args *sa;
4707 	struct btrfs_trans_handle *trans;
4708 	int ret;
4709 	int err;
4710 	u64 qgroupid;
4711 
4712 	if (!capable(CAP_SYS_ADMIN))
4713 		return -EPERM;
4714 
4715 	ret = mnt_want_write_file(file);
4716 	if (ret)
4717 		return ret;
4718 
4719 	sa = memdup_user(arg, sizeof(*sa));
4720 	if (IS_ERR(sa)) {
4721 		ret = PTR_ERR(sa);
4722 		goto drop_write;
4723 	}
4724 
4725 	trans = btrfs_join_transaction(root);
4726 	if (IS_ERR(trans)) {
4727 		ret = PTR_ERR(trans);
4728 		goto out;
4729 	}
4730 
4731 	qgroupid = sa->qgroupid;
4732 	if (!qgroupid) {
4733 		/* take the current subvol as qgroup */
4734 		qgroupid = root->root_key.objectid;
4735 	}
4736 
4737 	/* FIXME: check if the IDs really exist */
4738 	ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4739 
4740 	err = btrfs_end_transaction(trans, root);
4741 	if (err && !ret)
4742 		ret = err;
4743 
4744 out:
4745 	kfree(sa);
4746 drop_write:
4747 	mnt_drop_write_file(file);
4748 	return ret;
4749 }
4750 
4751 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4752 {
4753 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4754 	struct btrfs_ioctl_quota_rescan_args *qsa;
4755 	int ret;
4756 
4757 	if (!capable(CAP_SYS_ADMIN))
4758 		return -EPERM;
4759 
4760 	ret = mnt_want_write_file(file);
4761 	if (ret)
4762 		return ret;
4763 
4764 	qsa = memdup_user(arg, sizeof(*qsa));
4765 	if (IS_ERR(qsa)) {
4766 		ret = PTR_ERR(qsa);
4767 		goto drop_write;
4768 	}
4769 
4770 	if (qsa->flags) {
4771 		ret = -EINVAL;
4772 		goto out;
4773 	}
4774 
4775 	ret = btrfs_qgroup_rescan(root->fs_info);
4776 
4777 out:
4778 	kfree(qsa);
4779 drop_write:
4780 	mnt_drop_write_file(file);
4781 	return ret;
4782 }
4783 
4784 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4785 {
4786 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4787 	struct btrfs_ioctl_quota_rescan_args *qsa;
4788 	int ret = 0;
4789 
4790 	if (!capable(CAP_SYS_ADMIN))
4791 		return -EPERM;
4792 
4793 	qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4794 	if (!qsa)
4795 		return -ENOMEM;
4796 
4797 	if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4798 		qsa->flags = 1;
4799 		qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4800 	}
4801 
4802 	if (copy_to_user(arg, qsa, sizeof(*qsa)))
4803 		ret = -EFAULT;
4804 
4805 	kfree(qsa);
4806 	return ret;
4807 }
4808 
4809 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4810 {
4811 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4812 
4813 	if (!capable(CAP_SYS_ADMIN))
4814 		return -EPERM;
4815 
4816 	return btrfs_qgroup_wait_for_completion(root->fs_info);
4817 }
4818 
4819 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4820 					    struct btrfs_ioctl_received_subvol_args *sa)
4821 {
4822 	struct inode *inode = file_inode(file);
4823 	struct btrfs_root *root = BTRFS_I(inode)->root;
4824 	struct btrfs_root_item *root_item = &root->root_item;
4825 	struct btrfs_trans_handle *trans;
4826 	struct timespec ct = CURRENT_TIME;
4827 	int ret = 0;
4828 	int received_uuid_changed;
4829 
4830 	if (!inode_owner_or_capable(inode))
4831 		return -EPERM;
4832 
4833 	ret = mnt_want_write_file(file);
4834 	if (ret < 0)
4835 		return ret;
4836 
4837 	down_write(&root->fs_info->subvol_sem);
4838 
4839 	if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4840 		ret = -EINVAL;
4841 		goto out;
4842 	}
4843 
4844 	if (btrfs_root_readonly(root)) {
4845 		ret = -EROFS;
4846 		goto out;
4847 	}
4848 
4849 	/*
4850 	 * 1 - root item
4851 	 * 2 - uuid items (received uuid + subvol uuid)
4852 	 */
4853 	trans = btrfs_start_transaction(root, 3);
4854 	if (IS_ERR(trans)) {
4855 		ret = PTR_ERR(trans);
4856 		trans = NULL;
4857 		goto out;
4858 	}
4859 
4860 	sa->rtransid = trans->transid;
4861 	sa->rtime.sec = ct.tv_sec;
4862 	sa->rtime.nsec = ct.tv_nsec;
4863 
4864 	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4865 				       BTRFS_UUID_SIZE);
4866 	if (received_uuid_changed &&
4867 	    !btrfs_is_empty_uuid(root_item->received_uuid))
4868 		btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4869 				    root_item->received_uuid,
4870 				    BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4871 				    root->root_key.objectid);
4872 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4873 	btrfs_set_root_stransid(root_item, sa->stransid);
4874 	btrfs_set_root_rtransid(root_item, sa->rtransid);
4875 	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4876 	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4877 	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4878 	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4879 
4880 	ret = btrfs_update_root(trans, root->fs_info->tree_root,
4881 				&root->root_key, &root->root_item);
4882 	if (ret < 0) {
4883 		btrfs_end_transaction(trans, root);
4884 		goto out;
4885 	}
4886 	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4887 		ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4888 					  sa->uuid,
4889 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4890 					  root->root_key.objectid);
4891 		if (ret < 0 && ret != -EEXIST) {
4892 			btrfs_abort_transaction(trans, root, ret);
4893 			goto out;
4894 		}
4895 	}
4896 	ret = btrfs_commit_transaction(trans, root);
4897 	if (ret < 0) {
4898 		btrfs_abort_transaction(trans, root, ret);
4899 		goto out;
4900 	}
4901 
4902 out:
4903 	up_write(&root->fs_info->subvol_sem);
4904 	mnt_drop_write_file(file);
4905 	return ret;
4906 }
4907 
4908 #ifdef CONFIG_64BIT
4909 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4910 						void __user *arg)
4911 {
4912 	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4913 	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4914 	int ret = 0;
4915 
4916 	args32 = memdup_user(arg, sizeof(*args32));
4917 	if (IS_ERR(args32)) {
4918 		ret = PTR_ERR(args32);
4919 		args32 = NULL;
4920 		goto out;
4921 	}
4922 
4923 	args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4924 	if (!args64) {
4925 		ret = -ENOMEM;
4926 		goto out;
4927 	}
4928 
4929 	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4930 	args64->stransid = args32->stransid;
4931 	args64->rtransid = args32->rtransid;
4932 	args64->stime.sec = args32->stime.sec;
4933 	args64->stime.nsec = args32->stime.nsec;
4934 	args64->rtime.sec = args32->rtime.sec;
4935 	args64->rtime.nsec = args32->rtime.nsec;
4936 	args64->flags = args32->flags;
4937 
4938 	ret = _btrfs_ioctl_set_received_subvol(file, args64);
4939 	if (ret)
4940 		goto out;
4941 
4942 	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4943 	args32->stransid = args64->stransid;
4944 	args32->rtransid = args64->rtransid;
4945 	args32->stime.sec = args64->stime.sec;
4946 	args32->stime.nsec = args64->stime.nsec;
4947 	args32->rtime.sec = args64->rtime.sec;
4948 	args32->rtime.nsec = args64->rtime.nsec;
4949 	args32->flags = args64->flags;
4950 
4951 	ret = copy_to_user(arg, args32, sizeof(*args32));
4952 	if (ret)
4953 		ret = -EFAULT;
4954 
4955 out:
4956 	kfree(args32);
4957 	kfree(args64);
4958 	return ret;
4959 }
4960 #endif
4961 
4962 static long btrfs_ioctl_set_received_subvol(struct file *file,
4963 					    void __user *arg)
4964 {
4965 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4966 	int ret = 0;
4967 
4968 	sa = memdup_user(arg, sizeof(*sa));
4969 	if (IS_ERR(sa)) {
4970 		ret = PTR_ERR(sa);
4971 		sa = NULL;
4972 		goto out;
4973 	}
4974 
4975 	ret = _btrfs_ioctl_set_received_subvol(file, sa);
4976 
4977 	if (ret)
4978 		goto out;
4979 
4980 	ret = copy_to_user(arg, sa, sizeof(*sa));
4981 	if (ret)
4982 		ret = -EFAULT;
4983 
4984 out:
4985 	kfree(sa);
4986 	return ret;
4987 }
4988 
4989 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4990 {
4991 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4992 	size_t len;
4993 	int ret;
4994 	char label[BTRFS_LABEL_SIZE];
4995 
4996 	spin_lock(&root->fs_info->super_lock);
4997 	memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4998 	spin_unlock(&root->fs_info->super_lock);
4999 
5000 	len = strnlen(label, BTRFS_LABEL_SIZE);
5001 
5002 	if (len == BTRFS_LABEL_SIZE) {
5003 		btrfs_warn(root->fs_info,
5004 			"label is too long, return the first %zu bytes", --len);
5005 	}
5006 
5007 	ret = copy_to_user(arg, label, len);
5008 
5009 	return ret ? -EFAULT : 0;
5010 }
5011 
5012 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5013 {
5014 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5015 	struct btrfs_super_block *super_block = root->fs_info->super_copy;
5016 	struct btrfs_trans_handle *trans;
5017 	char label[BTRFS_LABEL_SIZE];
5018 	int ret;
5019 
5020 	if (!capable(CAP_SYS_ADMIN))
5021 		return -EPERM;
5022 
5023 	if (copy_from_user(label, arg, sizeof(label)))
5024 		return -EFAULT;
5025 
5026 	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5027 		btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5028 		       BTRFS_LABEL_SIZE - 1);
5029 		return -EINVAL;
5030 	}
5031 
5032 	ret = mnt_want_write_file(file);
5033 	if (ret)
5034 		return ret;
5035 
5036 	trans = btrfs_start_transaction(root, 0);
5037 	if (IS_ERR(trans)) {
5038 		ret = PTR_ERR(trans);
5039 		goto out_unlock;
5040 	}
5041 
5042 	spin_lock(&root->fs_info->super_lock);
5043 	strcpy(super_block->label, label);
5044 	spin_unlock(&root->fs_info->super_lock);
5045 	ret = btrfs_commit_transaction(trans, root);
5046 
5047 out_unlock:
5048 	mnt_drop_write_file(file);
5049 	return ret;
5050 }
5051 
5052 #define INIT_FEATURE_FLAGS(suffix) \
5053 	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5054 	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5055 	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5056 
5057 static int btrfs_ioctl_get_supported_features(struct file *file,
5058 					      void __user *arg)
5059 {
5060 	static struct btrfs_ioctl_feature_flags features[3] = {
5061 		INIT_FEATURE_FLAGS(SUPP),
5062 		INIT_FEATURE_FLAGS(SAFE_SET),
5063 		INIT_FEATURE_FLAGS(SAFE_CLEAR)
5064 	};
5065 
5066 	if (copy_to_user(arg, &features, sizeof(features)))
5067 		return -EFAULT;
5068 
5069 	return 0;
5070 }
5071 
5072 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5073 {
5074 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5075 	struct btrfs_super_block *super_block = root->fs_info->super_copy;
5076 	struct btrfs_ioctl_feature_flags features;
5077 
5078 	features.compat_flags = btrfs_super_compat_flags(super_block);
5079 	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5080 	features.incompat_flags = btrfs_super_incompat_flags(super_block);
5081 
5082 	if (copy_to_user(arg, &features, sizeof(features)))
5083 		return -EFAULT;
5084 
5085 	return 0;
5086 }
5087 
5088 static int check_feature_bits(struct btrfs_root *root,
5089 			      enum btrfs_feature_set set,
5090 			      u64 change_mask, u64 flags, u64 supported_flags,
5091 			      u64 safe_set, u64 safe_clear)
5092 {
5093 	const char *type = btrfs_feature_set_names[set];
5094 	char *names;
5095 	u64 disallowed, unsupported;
5096 	u64 set_mask = flags & change_mask;
5097 	u64 clear_mask = ~flags & change_mask;
5098 
5099 	unsupported = set_mask & ~supported_flags;
5100 	if (unsupported) {
5101 		names = btrfs_printable_features(set, unsupported);
5102 		if (names) {
5103 			btrfs_warn(root->fs_info,
5104 			   "this kernel does not support the %s feature bit%s",
5105 			   names, strchr(names, ',') ? "s" : "");
5106 			kfree(names);
5107 		} else
5108 			btrfs_warn(root->fs_info,
5109 			   "this kernel does not support %s bits 0x%llx",
5110 			   type, unsupported);
5111 		return -EOPNOTSUPP;
5112 	}
5113 
5114 	disallowed = set_mask & ~safe_set;
5115 	if (disallowed) {
5116 		names = btrfs_printable_features(set, disallowed);
5117 		if (names) {
5118 			btrfs_warn(root->fs_info,
5119 			   "can't set the %s feature bit%s while mounted",
5120 			   names, strchr(names, ',') ? "s" : "");
5121 			kfree(names);
5122 		} else
5123 			btrfs_warn(root->fs_info,
5124 			   "can't set %s bits 0x%llx while mounted",
5125 			   type, disallowed);
5126 		return -EPERM;
5127 	}
5128 
5129 	disallowed = clear_mask & ~safe_clear;
5130 	if (disallowed) {
5131 		names = btrfs_printable_features(set, disallowed);
5132 		if (names) {
5133 			btrfs_warn(root->fs_info,
5134 			   "can't clear the %s feature bit%s while mounted",
5135 			   names, strchr(names, ',') ? "s" : "");
5136 			kfree(names);
5137 		} else
5138 			btrfs_warn(root->fs_info,
5139 			   "can't clear %s bits 0x%llx while mounted",
5140 			   type, disallowed);
5141 		return -EPERM;
5142 	}
5143 
5144 	return 0;
5145 }
5146 
5147 #define check_feature(root, change_mask, flags, mask_base)	\
5148 check_feature_bits(root, FEAT_##mask_base, change_mask, flags,	\
5149 		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
5150 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
5151 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5152 
5153 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5154 {
5155 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5156 	struct btrfs_super_block *super_block = root->fs_info->super_copy;
5157 	struct btrfs_ioctl_feature_flags flags[2];
5158 	struct btrfs_trans_handle *trans;
5159 	u64 newflags;
5160 	int ret;
5161 
5162 	if (!capable(CAP_SYS_ADMIN))
5163 		return -EPERM;
5164 
5165 	if (copy_from_user(flags, arg, sizeof(flags)))
5166 		return -EFAULT;
5167 
5168 	/* Nothing to do */
5169 	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5170 	    !flags[0].incompat_flags)
5171 		return 0;
5172 
5173 	ret = check_feature(root, flags[0].compat_flags,
5174 			    flags[1].compat_flags, COMPAT);
5175 	if (ret)
5176 		return ret;
5177 
5178 	ret = check_feature(root, flags[0].compat_ro_flags,
5179 			    flags[1].compat_ro_flags, COMPAT_RO);
5180 	if (ret)
5181 		return ret;
5182 
5183 	ret = check_feature(root, flags[0].incompat_flags,
5184 			    flags[1].incompat_flags, INCOMPAT);
5185 	if (ret)
5186 		return ret;
5187 
5188 	trans = btrfs_start_transaction(root, 0);
5189 	if (IS_ERR(trans))
5190 		return PTR_ERR(trans);
5191 
5192 	spin_lock(&root->fs_info->super_lock);
5193 	newflags = btrfs_super_compat_flags(super_block);
5194 	newflags |= flags[0].compat_flags & flags[1].compat_flags;
5195 	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5196 	btrfs_set_super_compat_flags(super_block, newflags);
5197 
5198 	newflags = btrfs_super_compat_ro_flags(super_block);
5199 	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5200 	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5201 	btrfs_set_super_compat_ro_flags(super_block, newflags);
5202 
5203 	newflags = btrfs_super_incompat_flags(super_block);
5204 	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5205 	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5206 	btrfs_set_super_incompat_flags(super_block, newflags);
5207 	spin_unlock(&root->fs_info->super_lock);
5208 
5209 	return btrfs_commit_transaction(trans, root);
5210 }
5211 
5212 long btrfs_ioctl(struct file *file, unsigned int
5213 		cmd, unsigned long arg)
5214 {
5215 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5216 	void __user *argp = (void __user *)arg;
5217 
5218 	switch (cmd) {
5219 	case FS_IOC_GETFLAGS:
5220 		return btrfs_ioctl_getflags(file, argp);
5221 	case FS_IOC_SETFLAGS:
5222 		return btrfs_ioctl_setflags(file, argp);
5223 	case FS_IOC_GETVERSION:
5224 		return btrfs_ioctl_getversion(file, argp);
5225 	case FITRIM:
5226 		return btrfs_ioctl_fitrim(file, argp);
5227 	case BTRFS_IOC_SNAP_CREATE:
5228 		return btrfs_ioctl_snap_create(file, argp, 0);
5229 	case BTRFS_IOC_SNAP_CREATE_V2:
5230 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
5231 	case BTRFS_IOC_SUBVOL_CREATE:
5232 		return btrfs_ioctl_snap_create(file, argp, 1);
5233 	case BTRFS_IOC_SUBVOL_CREATE_V2:
5234 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
5235 	case BTRFS_IOC_SNAP_DESTROY:
5236 		return btrfs_ioctl_snap_destroy(file, argp);
5237 	case BTRFS_IOC_SUBVOL_GETFLAGS:
5238 		return btrfs_ioctl_subvol_getflags(file, argp);
5239 	case BTRFS_IOC_SUBVOL_SETFLAGS:
5240 		return btrfs_ioctl_subvol_setflags(file, argp);
5241 	case BTRFS_IOC_DEFAULT_SUBVOL:
5242 		return btrfs_ioctl_default_subvol(file, argp);
5243 	case BTRFS_IOC_DEFRAG:
5244 		return btrfs_ioctl_defrag(file, NULL);
5245 	case BTRFS_IOC_DEFRAG_RANGE:
5246 		return btrfs_ioctl_defrag(file, argp);
5247 	case BTRFS_IOC_RESIZE:
5248 		return btrfs_ioctl_resize(file, argp);
5249 	case BTRFS_IOC_ADD_DEV:
5250 		return btrfs_ioctl_add_dev(root, argp);
5251 	case BTRFS_IOC_RM_DEV:
5252 		return btrfs_ioctl_rm_dev(file, argp);
5253 	case BTRFS_IOC_FS_INFO:
5254 		return btrfs_ioctl_fs_info(root, argp);
5255 	case BTRFS_IOC_DEV_INFO:
5256 		return btrfs_ioctl_dev_info(root, argp);
5257 	case BTRFS_IOC_BALANCE:
5258 		return btrfs_ioctl_balance(file, NULL);
5259 	case BTRFS_IOC_CLONE:
5260 		return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5261 	case BTRFS_IOC_CLONE_RANGE:
5262 		return btrfs_ioctl_clone_range(file, argp);
5263 	case BTRFS_IOC_TRANS_START:
5264 		return btrfs_ioctl_trans_start(file);
5265 	case BTRFS_IOC_TRANS_END:
5266 		return btrfs_ioctl_trans_end(file);
5267 	case BTRFS_IOC_TREE_SEARCH:
5268 		return btrfs_ioctl_tree_search(file, argp);
5269 	case BTRFS_IOC_TREE_SEARCH_V2:
5270 		return btrfs_ioctl_tree_search_v2(file, argp);
5271 	case BTRFS_IOC_INO_LOOKUP:
5272 		return btrfs_ioctl_ino_lookup(file, argp);
5273 	case BTRFS_IOC_INO_PATHS:
5274 		return btrfs_ioctl_ino_to_path(root, argp);
5275 	case BTRFS_IOC_LOGICAL_INO:
5276 		return btrfs_ioctl_logical_to_ino(root, argp);
5277 	case BTRFS_IOC_SPACE_INFO:
5278 		return btrfs_ioctl_space_info(root, argp);
5279 	case BTRFS_IOC_SYNC: {
5280 		int ret;
5281 
5282 		ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5283 		if (ret)
5284 			return ret;
5285 		ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5286 		return ret;
5287 	}
5288 	case BTRFS_IOC_START_SYNC:
5289 		return btrfs_ioctl_start_sync(root, argp);
5290 	case BTRFS_IOC_WAIT_SYNC:
5291 		return btrfs_ioctl_wait_sync(root, argp);
5292 	case BTRFS_IOC_SCRUB:
5293 		return btrfs_ioctl_scrub(file, argp);
5294 	case BTRFS_IOC_SCRUB_CANCEL:
5295 		return btrfs_ioctl_scrub_cancel(root, argp);
5296 	case BTRFS_IOC_SCRUB_PROGRESS:
5297 		return btrfs_ioctl_scrub_progress(root, argp);
5298 	case BTRFS_IOC_BALANCE_V2:
5299 		return btrfs_ioctl_balance(file, argp);
5300 	case BTRFS_IOC_BALANCE_CTL:
5301 		return btrfs_ioctl_balance_ctl(root, arg);
5302 	case BTRFS_IOC_BALANCE_PROGRESS:
5303 		return btrfs_ioctl_balance_progress(root, argp);
5304 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5305 		return btrfs_ioctl_set_received_subvol(file, argp);
5306 #ifdef CONFIG_64BIT
5307 	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5308 		return btrfs_ioctl_set_received_subvol_32(file, argp);
5309 #endif
5310 	case BTRFS_IOC_SEND:
5311 		return btrfs_ioctl_send(file, argp);
5312 	case BTRFS_IOC_GET_DEV_STATS:
5313 		return btrfs_ioctl_get_dev_stats(root, argp);
5314 	case BTRFS_IOC_QUOTA_CTL:
5315 		return btrfs_ioctl_quota_ctl(file, argp);
5316 	case BTRFS_IOC_QGROUP_ASSIGN:
5317 		return btrfs_ioctl_qgroup_assign(file, argp);
5318 	case BTRFS_IOC_QGROUP_CREATE:
5319 		return btrfs_ioctl_qgroup_create(file, argp);
5320 	case BTRFS_IOC_QGROUP_LIMIT:
5321 		return btrfs_ioctl_qgroup_limit(file, argp);
5322 	case BTRFS_IOC_QUOTA_RESCAN:
5323 		return btrfs_ioctl_quota_rescan(file, argp);
5324 	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5325 		return btrfs_ioctl_quota_rescan_status(file, argp);
5326 	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5327 		return btrfs_ioctl_quota_rescan_wait(file, argp);
5328 	case BTRFS_IOC_DEV_REPLACE:
5329 		return btrfs_ioctl_dev_replace(root, argp);
5330 	case BTRFS_IOC_GET_FSLABEL:
5331 		return btrfs_ioctl_get_fslabel(file, argp);
5332 	case BTRFS_IOC_SET_FSLABEL:
5333 		return btrfs_ioctl_set_fslabel(file, argp);
5334 	case BTRFS_IOC_FILE_EXTENT_SAME:
5335 		return btrfs_ioctl_file_extent_same(file, argp);
5336 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5337 		return btrfs_ioctl_get_supported_features(file, argp);
5338 	case BTRFS_IOC_GET_FEATURES:
5339 		return btrfs_ioctl_get_features(file, argp);
5340 	case BTRFS_IOC_SET_FEATURES:
5341 		return btrfs_ioctl_set_features(file, argp);
5342 	}
5343 
5344 	return -ENOTTY;
5345 }
5346