xref: /openbmc/linux/fs/btrfs/ioctl.c (revision 62975d27)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (C) 2007 Oracle.  All rights reserved.
4  */
5 
6 #include <linux/kernel.h>
7 #include <linux/bio.h>
8 #include <linux/file.h>
9 #include <linux/fs.h>
10 #include <linux/fsnotify.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/time.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/namei.h>
18 #include <linux/writeback.h>
19 #include <linux/compat.h>
20 #include <linux/security.h>
21 #include <linux/xattr.h>
22 #include <linux/mm.h>
23 #include <linux/slab.h>
24 #include <linux/blkdev.h>
25 #include <linux/uuid.h>
26 #include <linux/btrfs.h>
27 #include <linux/uaccess.h>
28 #include <linux/iversion.h>
29 #include "ctree.h"
30 #include "disk-io.h"
31 #include "export.h"
32 #include "transaction.h"
33 #include "btrfs_inode.h"
34 #include "print-tree.h"
35 #include "volumes.h"
36 #include "locking.h"
37 #include "inode-map.h"
38 #include "backref.h"
39 #include "rcu-string.h"
40 #include "send.h"
41 #include "dev-replace.h"
42 #include "props.h"
43 #include "sysfs.h"
44 #include "qgroup.h"
45 #include "tree-log.h"
46 #include "compression.h"
47 #include "space-info.h"
48 #include "delalloc-space.h"
49 #include "block-group.h"
50 
51 #ifdef CONFIG_64BIT
52 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
53  * structures are incorrect, as the timespec structure from userspace
54  * is 4 bytes too small. We define these alternatives here to teach
55  * the kernel about the 32-bit struct packing.
56  */
57 struct btrfs_ioctl_timespec_32 {
58 	__u64 sec;
59 	__u32 nsec;
60 } __attribute__ ((__packed__));
61 
62 struct btrfs_ioctl_received_subvol_args_32 {
63 	char	uuid[BTRFS_UUID_SIZE];	/* in */
64 	__u64	stransid;		/* in */
65 	__u64	rtransid;		/* out */
66 	struct btrfs_ioctl_timespec_32 stime; /* in */
67 	struct btrfs_ioctl_timespec_32 rtime; /* out */
68 	__u64	flags;			/* in */
69 	__u64	reserved[16];		/* in */
70 } __attribute__ ((__packed__));
71 
72 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
73 				struct btrfs_ioctl_received_subvol_args_32)
74 #endif
75 
76 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
77 struct btrfs_ioctl_send_args_32 {
78 	__s64 send_fd;			/* in */
79 	__u64 clone_sources_count;	/* in */
80 	compat_uptr_t clone_sources;	/* in */
81 	__u64 parent_root;		/* in */
82 	__u64 flags;			/* in */
83 	__u64 reserved[4];		/* in */
84 } __attribute__ ((__packed__));
85 
86 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
87 			       struct btrfs_ioctl_send_args_32)
88 #endif
89 
90 /* Mask out flags that are inappropriate for the given type of inode. */
91 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
92 		unsigned int flags)
93 {
94 	if (S_ISDIR(inode->i_mode))
95 		return flags;
96 	else if (S_ISREG(inode->i_mode))
97 		return flags & ~FS_DIRSYNC_FL;
98 	else
99 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
100 }
101 
102 /*
103  * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
104  * ioctl.
105  */
106 static unsigned int btrfs_inode_flags_to_fsflags(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_NOCOMPRESS)
126 		iflags |= FS_NOCOMP_FL;
127 	else if (flags & BTRFS_INODE_COMPRESS)
128 		iflags |= FS_COMPR_FL;
129 
130 	return iflags;
131 }
132 
133 /*
134  * Update inode->i_flags based on the btrfs internal flags.
135  */
136 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
137 {
138 	struct btrfs_inode *binode = BTRFS_I(inode);
139 	unsigned int new_fl = 0;
140 
141 	if (binode->flags & BTRFS_INODE_SYNC)
142 		new_fl |= S_SYNC;
143 	if (binode->flags & BTRFS_INODE_IMMUTABLE)
144 		new_fl |= S_IMMUTABLE;
145 	if (binode->flags & BTRFS_INODE_APPEND)
146 		new_fl |= S_APPEND;
147 	if (binode->flags & BTRFS_INODE_NOATIME)
148 		new_fl |= S_NOATIME;
149 	if (binode->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 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
158 {
159 	struct btrfs_inode *binode = BTRFS_I(file_inode(file));
160 	unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
161 
162 	if (copy_to_user(arg, &flags, sizeof(flags)))
163 		return -EFAULT;
164 	return 0;
165 }
166 
167 /* Check if @flags are a supported and valid set of FS_*_FL flags */
168 static int check_fsflags(unsigned int flags)
169 {
170 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
171 		      FS_NOATIME_FL | FS_NODUMP_FL | \
172 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
173 		      FS_NOCOMP_FL | FS_COMPR_FL |
174 		      FS_NOCOW_FL))
175 		return -EOPNOTSUPP;
176 
177 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
178 		return -EINVAL;
179 
180 	return 0;
181 }
182 
183 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
184 {
185 	struct inode *inode = file_inode(file);
186 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
187 	struct btrfs_inode *binode = BTRFS_I(inode);
188 	struct btrfs_root *root = binode->root;
189 	struct btrfs_trans_handle *trans;
190 	unsigned int fsflags, old_fsflags;
191 	int ret;
192 	const char *comp = NULL;
193 	u32 binode_flags = binode->flags;
194 
195 	if (!inode_owner_or_capable(inode))
196 		return -EPERM;
197 
198 	if (btrfs_root_readonly(root))
199 		return -EROFS;
200 
201 	if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
202 		return -EFAULT;
203 
204 	ret = check_fsflags(fsflags);
205 	if (ret)
206 		return ret;
207 
208 	ret = mnt_want_write_file(file);
209 	if (ret)
210 		return ret;
211 
212 	inode_lock(inode);
213 
214 	fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
215 	old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
216 	ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
217 	if (ret)
218 		goto out_unlock;
219 
220 	if (fsflags & FS_SYNC_FL)
221 		binode_flags |= BTRFS_INODE_SYNC;
222 	else
223 		binode_flags &= ~BTRFS_INODE_SYNC;
224 	if (fsflags & FS_IMMUTABLE_FL)
225 		binode_flags |= BTRFS_INODE_IMMUTABLE;
226 	else
227 		binode_flags &= ~BTRFS_INODE_IMMUTABLE;
228 	if (fsflags & FS_APPEND_FL)
229 		binode_flags |= BTRFS_INODE_APPEND;
230 	else
231 		binode_flags &= ~BTRFS_INODE_APPEND;
232 	if (fsflags & FS_NODUMP_FL)
233 		binode_flags |= BTRFS_INODE_NODUMP;
234 	else
235 		binode_flags &= ~BTRFS_INODE_NODUMP;
236 	if (fsflags & FS_NOATIME_FL)
237 		binode_flags |= BTRFS_INODE_NOATIME;
238 	else
239 		binode_flags &= ~BTRFS_INODE_NOATIME;
240 	if (fsflags & FS_DIRSYNC_FL)
241 		binode_flags |= BTRFS_INODE_DIRSYNC;
242 	else
243 		binode_flags &= ~BTRFS_INODE_DIRSYNC;
244 	if (fsflags & FS_NOCOW_FL) {
245 		if (S_ISREG(inode->i_mode)) {
246 			/*
247 			 * It's safe to turn csums off here, no extents exist.
248 			 * Otherwise we want the flag to reflect the real COW
249 			 * status of the file and will not set it.
250 			 */
251 			if (inode->i_size == 0)
252 				binode_flags |= BTRFS_INODE_NODATACOW |
253 						BTRFS_INODE_NODATASUM;
254 		} else {
255 			binode_flags |= BTRFS_INODE_NODATACOW;
256 		}
257 	} else {
258 		/*
259 		 * Revert back under same assumptions as above
260 		 */
261 		if (S_ISREG(inode->i_mode)) {
262 			if (inode->i_size == 0)
263 				binode_flags &= ~(BTRFS_INODE_NODATACOW |
264 						  BTRFS_INODE_NODATASUM);
265 		} else {
266 			binode_flags &= ~BTRFS_INODE_NODATACOW;
267 		}
268 	}
269 
270 	/*
271 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
272 	 * flag may be changed automatically if compression code won't make
273 	 * things smaller.
274 	 */
275 	if (fsflags & FS_NOCOMP_FL) {
276 		binode_flags &= ~BTRFS_INODE_COMPRESS;
277 		binode_flags |= BTRFS_INODE_NOCOMPRESS;
278 	} else if (fsflags & FS_COMPR_FL) {
279 
280 		if (IS_SWAPFILE(inode)) {
281 			ret = -ETXTBSY;
282 			goto out_unlock;
283 		}
284 
285 		binode_flags |= BTRFS_INODE_COMPRESS;
286 		binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
287 
288 		comp = btrfs_compress_type2str(fs_info->compress_type);
289 		if (!comp || comp[0] == 0)
290 			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
291 	} else {
292 		binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
293 	}
294 
295 	/*
296 	 * 1 for inode item
297 	 * 2 for properties
298 	 */
299 	trans = btrfs_start_transaction(root, 3);
300 	if (IS_ERR(trans)) {
301 		ret = PTR_ERR(trans);
302 		goto out_unlock;
303 	}
304 
305 	if (comp) {
306 		ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
307 				     strlen(comp), 0);
308 		if (ret) {
309 			btrfs_abort_transaction(trans, ret);
310 			goto out_end_trans;
311 		}
312 	} else {
313 		ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
314 				     0, 0);
315 		if (ret && ret != -ENODATA) {
316 			btrfs_abort_transaction(trans, ret);
317 			goto out_end_trans;
318 		}
319 	}
320 
321 	binode->flags = binode_flags;
322 	btrfs_sync_inode_flags_to_i_flags(inode);
323 	inode_inc_iversion(inode);
324 	inode->i_ctime = current_time(inode);
325 	ret = btrfs_update_inode(trans, root, inode);
326 
327  out_end_trans:
328 	btrfs_end_transaction(trans);
329  out_unlock:
330 	inode_unlock(inode);
331 	mnt_drop_write_file(file);
332 	return ret;
333 }
334 
335 /*
336  * Translate btrfs internal inode flags to xflags as expected by the
337  * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
338  * silently dropped.
339  */
340 static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
341 {
342 	unsigned int xflags = 0;
343 
344 	if (flags & BTRFS_INODE_APPEND)
345 		xflags |= FS_XFLAG_APPEND;
346 	if (flags & BTRFS_INODE_IMMUTABLE)
347 		xflags |= FS_XFLAG_IMMUTABLE;
348 	if (flags & BTRFS_INODE_NOATIME)
349 		xflags |= FS_XFLAG_NOATIME;
350 	if (flags & BTRFS_INODE_NODUMP)
351 		xflags |= FS_XFLAG_NODUMP;
352 	if (flags & BTRFS_INODE_SYNC)
353 		xflags |= FS_XFLAG_SYNC;
354 
355 	return xflags;
356 }
357 
358 /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
359 static int check_xflags(unsigned int flags)
360 {
361 	if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
362 		      FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
363 		return -EOPNOTSUPP;
364 	return 0;
365 }
366 
367 /*
368  * Set the xflags from the internal inode flags. The remaining items of fsxattr
369  * are zeroed.
370  */
371 static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
372 {
373 	struct btrfs_inode *binode = BTRFS_I(file_inode(file));
374 	struct fsxattr fa;
375 
376 	simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags));
377 	if (copy_to_user(arg, &fa, sizeof(fa)))
378 		return -EFAULT;
379 
380 	return 0;
381 }
382 
383 static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
384 {
385 	struct inode *inode = file_inode(file);
386 	struct btrfs_inode *binode = BTRFS_I(inode);
387 	struct btrfs_root *root = binode->root;
388 	struct btrfs_trans_handle *trans;
389 	struct fsxattr fa, old_fa;
390 	unsigned old_flags;
391 	unsigned old_i_flags;
392 	int ret = 0;
393 
394 	if (!inode_owner_or_capable(inode))
395 		return -EPERM;
396 
397 	if (btrfs_root_readonly(root))
398 		return -EROFS;
399 
400 	if (copy_from_user(&fa, arg, sizeof(fa)))
401 		return -EFAULT;
402 
403 	ret = check_xflags(fa.fsx_xflags);
404 	if (ret)
405 		return ret;
406 
407 	if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
408 		return -EOPNOTSUPP;
409 
410 	ret = mnt_want_write_file(file);
411 	if (ret)
412 		return ret;
413 
414 	inode_lock(inode);
415 
416 	old_flags = binode->flags;
417 	old_i_flags = inode->i_flags;
418 
419 	simple_fill_fsxattr(&old_fa,
420 			    btrfs_inode_flags_to_xflags(binode->flags));
421 	ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
422 	if (ret)
423 		goto out_unlock;
424 
425 	if (fa.fsx_xflags & FS_XFLAG_SYNC)
426 		binode->flags |= BTRFS_INODE_SYNC;
427 	else
428 		binode->flags &= ~BTRFS_INODE_SYNC;
429 	if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
430 		binode->flags |= BTRFS_INODE_IMMUTABLE;
431 	else
432 		binode->flags &= ~BTRFS_INODE_IMMUTABLE;
433 	if (fa.fsx_xflags & FS_XFLAG_APPEND)
434 		binode->flags |= BTRFS_INODE_APPEND;
435 	else
436 		binode->flags &= ~BTRFS_INODE_APPEND;
437 	if (fa.fsx_xflags & FS_XFLAG_NODUMP)
438 		binode->flags |= BTRFS_INODE_NODUMP;
439 	else
440 		binode->flags &= ~BTRFS_INODE_NODUMP;
441 	if (fa.fsx_xflags & FS_XFLAG_NOATIME)
442 		binode->flags |= BTRFS_INODE_NOATIME;
443 	else
444 		binode->flags &= ~BTRFS_INODE_NOATIME;
445 
446 	/* 1 item for the inode */
447 	trans = btrfs_start_transaction(root, 1);
448 	if (IS_ERR(trans)) {
449 		ret = PTR_ERR(trans);
450 		goto out_unlock;
451 	}
452 
453 	btrfs_sync_inode_flags_to_i_flags(inode);
454 	inode_inc_iversion(inode);
455 	inode->i_ctime = current_time(inode);
456 	ret = btrfs_update_inode(trans, root, inode);
457 
458 	btrfs_end_transaction(trans);
459 
460 out_unlock:
461 	if (ret) {
462 		binode->flags = old_flags;
463 		inode->i_flags = old_i_flags;
464 	}
465 
466 	inode_unlock(inode);
467 	mnt_drop_write_file(file);
468 
469 	return ret;
470 }
471 
472 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
473 {
474 	struct inode *inode = file_inode(file);
475 
476 	return put_user(inode->i_generation, arg);
477 }
478 
479 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
480 					void __user *arg)
481 {
482 	struct btrfs_device *device;
483 	struct request_queue *q;
484 	struct fstrim_range range;
485 	u64 minlen = ULLONG_MAX;
486 	u64 num_devices = 0;
487 	int ret;
488 
489 	if (!capable(CAP_SYS_ADMIN))
490 		return -EPERM;
491 
492 	/*
493 	 * If the fs is mounted with nologreplay, which requires it to be
494 	 * mounted in RO mode as well, we can not allow discard on free space
495 	 * inside block groups, because log trees refer to extents that are not
496 	 * pinned in a block group's free space cache (pinning the extents is
497 	 * precisely the first phase of replaying a log tree).
498 	 */
499 	if (btrfs_test_opt(fs_info, NOLOGREPLAY))
500 		return -EROFS;
501 
502 	rcu_read_lock();
503 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
504 				dev_list) {
505 		if (!device->bdev)
506 			continue;
507 		q = bdev_get_queue(device->bdev);
508 		if (blk_queue_discard(q)) {
509 			num_devices++;
510 			minlen = min_t(u64, q->limits.discard_granularity,
511 				     minlen);
512 		}
513 	}
514 	rcu_read_unlock();
515 
516 	if (!num_devices)
517 		return -EOPNOTSUPP;
518 	if (copy_from_user(&range, arg, sizeof(range)))
519 		return -EFAULT;
520 
521 	/*
522 	 * NOTE: Don't truncate the range using super->total_bytes.  Bytenr of
523 	 * block group is in the logical address space, which can be any
524 	 * sectorsize aligned bytenr in  the range [0, U64_MAX].
525 	 */
526 	if (range.len < fs_info->sb->s_blocksize)
527 		return -EINVAL;
528 
529 	range.minlen = max(range.minlen, minlen);
530 	ret = btrfs_trim_fs(fs_info, &range);
531 	if (ret < 0)
532 		return ret;
533 
534 	if (copy_to_user(arg, &range, sizeof(range)))
535 		return -EFAULT;
536 
537 	return 0;
538 }
539 
540 int __pure btrfs_is_empty_uuid(u8 *uuid)
541 {
542 	int i;
543 
544 	for (i = 0; i < BTRFS_UUID_SIZE; i++) {
545 		if (uuid[i])
546 			return 0;
547 	}
548 	return 1;
549 }
550 
551 static noinline int create_subvol(struct inode *dir,
552 				  struct dentry *dentry,
553 				  const char *name, int namelen,
554 				  struct btrfs_qgroup_inherit *inherit)
555 {
556 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
557 	struct btrfs_trans_handle *trans;
558 	struct btrfs_key key;
559 	struct btrfs_root_item *root_item;
560 	struct btrfs_inode_item *inode_item;
561 	struct extent_buffer *leaf;
562 	struct btrfs_root *root = BTRFS_I(dir)->root;
563 	struct btrfs_root *new_root;
564 	struct btrfs_block_rsv block_rsv;
565 	struct timespec64 cur_time = current_time(dir);
566 	struct inode *inode;
567 	int ret;
568 	int err;
569 	u64 objectid;
570 	u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
571 	u64 index = 0;
572 
573 	root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
574 	if (!root_item)
575 		return -ENOMEM;
576 
577 	ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
578 	if (ret)
579 		goto fail_free;
580 
581 	/*
582 	 * Don't create subvolume whose level is not zero. Or qgroup will be
583 	 * screwed up since it assumes subvolume qgroup's level to be 0.
584 	 */
585 	if (btrfs_qgroup_level(objectid)) {
586 		ret = -ENOSPC;
587 		goto fail_free;
588 	}
589 
590 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
591 	/*
592 	 * The same as the snapshot creation, please see the comment
593 	 * of create_snapshot().
594 	 */
595 	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
596 	if (ret)
597 		goto fail_free;
598 
599 	trans = btrfs_start_transaction(root, 0);
600 	if (IS_ERR(trans)) {
601 		ret = PTR_ERR(trans);
602 		btrfs_subvolume_release_metadata(fs_info, &block_rsv);
603 		goto fail_free;
604 	}
605 	trans->block_rsv = &block_rsv;
606 	trans->bytes_reserved = block_rsv.size;
607 
608 	ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
609 	if (ret)
610 		goto fail;
611 
612 	leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
613 	if (IS_ERR(leaf)) {
614 		ret = PTR_ERR(leaf);
615 		goto fail;
616 	}
617 
618 	btrfs_mark_buffer_dirty(leaf);
619 
620 	inode_item = &root_item->inode;
621 	btrfs_set_stack_inode_generation(inode_item, 1);
622 	btrfs_set_stack_inode_size(inode_item, 3);
623 	btrfs_set_stack_inode_nlink(inode_item, 1);
624 	btrfs_set_stack_inode_nbytes(inode_item,
625 				     fs_info->nodesize);
626 	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
627 
628 	btrfs_set_root_flags(root_item, 0);
629 	btrfs_set_root_limit(root_item, 0);
630 	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
631 
632 	btrfs_set_root_bytenr(root_item, leaf->start);
633 	btrfs_set_root_generation(root_item, trans->transid);
634 	btrfs_set_root_level(root_item, 0);
635 	btrfs_set_root_refs(root_item, 1);
636 	btrfs_set_root_used(root_item, leaf->len);
637 	btrfs_set_root_last_snapshot(root_item, 0);
638 
639 	btrfs_set_root_generation_v2(root_item,
640 			btrfs_root_generation(root_item));
641 	generate_random_guid(root_item->uuid);
642 	btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
643 	btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
644 	root_item->ctime = root_item->otime;
645 	btrfs_set_root_ctransid(root_item, trans->transid);
646 	btrfs_set_root_otransid(root_item, trans->transid);
647 
648 	btrfs_tree_unlock(leaf);
649 	free_extent_buffer(leaf);
650 	leaf = NULL;
651 
652 	btrfs_set_root_dirid(root_item, new_dirid);
653 
654 	key.objectid = objectid;
655 	key.offset = 0;
656 	key.type = BTRFS_ROOT_ITEM_KEY;
657 	ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
658 				root_item);
659 	if (ret)
660 		goto fail;
661 
662 	key.offset = (u64)-1;
663 	new_root = btrfs_get_fs_root(fs_info, objectid, true);
664 	if (IS_ERR(new_root)) {
665 		ret = PTR_ERR(new_root);
666 		btrfs_abort_transaction(trans, ret);
667 		goto fail;
668 	}
669 
670 	btrfs_record_root_in_trans(trans, new_root);
671 
672 	ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
673 	btrfs_put_root(new_root);
674 	if (ret) {
675 		/* We potentially lose an unused inode item here */
676 		btrfs_abort_transaction(trans, ret);
677 		goto fail;
678 	}
679 
680 	mutex_lock(&new_root->objectid_mutex);
681 	new_root->highest_objectid = new_dirid;
682 	mutex_unlock(&new_root->objectid_mutex);
683 
684 	/*
685 	 * insert the directory item
686 	 */
687 	ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
688 	if (ret) {
689 		btrfs_abort_transaction(trans, ret);
690 		goto fail;
691 	}
692 
693 	ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
694 				    BTRFS_FT_DIR, index);
695 	if (ret) {
696 		btrfs_abort_transaction(trans, ret);
697 		goto fail;
698 	}
699 
700 	btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
701 	ret = btrfs_update_inode(trans, root, dir);
702 	if (ret) {
703 		btrfs_abort_transaction(trans, ret);
704 		goto fail;
705 	}
706 
707 	ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
708 				 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
709 	if (ret) {
710 		btrfs_abort_transaction(trans, ret);
711 		goto fail;
712 	}
713 
714 	ret = btrfs_uuid_tree_add(trans, root_item->uuid,
715 				  BTRFS_UUID_KEY_SUBVOL, objectid);
716 	if (ret)
717 		btrfs_abort_transaction(trans, ret);
718 
719 fail:
720 	kfree(root_item);
721 	trans->block_rsv = NULL;
722 	trans->bytes_reserved = 0;
723 	btrfs_subvolume_release_metadata(fs_info, &block_rsv);
724 
725 	err = btrfs_commit_transaction(trans);
726 	if (err && !ret)
727 		ret = err;
728 
729 	if (!ret) {
730 		inode = btrfs_lookup_dentry(dir, dentry);
731 		if (IS_ERR(inode))
732 			return PTR_ERR(inode);
733 		d_instantiate(dentry, inode);
734 	}
735 	return ret;
736 
737 fail_free:
738 	kfree(root_item);
739 	return ret;
740 }
741 
742 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
743 			   struct dentry *dentry, bool readonly,
744 			   struct btrfs_qgroup_inherit *inherit)
745 {
746 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
747 	struct inode *inode;
748 	struct btrfs_pending_snapshot *pending_snapshot;
749 	struct btrfs_trans_handle *trans;
750 	int ret;
751 
752 	if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
753 		return -EINVAL;
754 
755 	if (atomic_read(&root->nr_swapfiles)) {
756 		btrfs_warn(fs_info,
757 			   "cannot snapshot subvolume with active swapfile");
758 		return -ETXTBSY;
759 	}
760 
761 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
762 	if (!pending_snapshot)
763 		return -ENOMEM;
764 
765 	pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
766 			GFP_KERNEL);
767 	pending_snapshot->path = btrfs_alloc_path();
768 	if (!pending_snapshot->root_item || !pending_snapshot->path) {
769 		ret = -ENOMEM;
770 		goto free_pending;
771 	}
772 
773 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
774 			     BTRFS_BLOCK_RSV_TEMP);
775 	/*
776 	 * 1 - parent dir inode
777 	 * 2 - dir entries
778 	 * 1 - root item
779 	 * 2 - root ref/backref
780 	 * 1 - root of snapshot
781 	 * 1 - UUID item
782 	 */
783 	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
784 					&pending_snapshot->block_rsv, 8,
785 					false);
786 	if (ret)
787 		goto free_pending;
788 
789 	pending_snapshot->dentry = dentry;
790 	pending_snapshot->root = root;
791 	pending_snapshot->readonly = readonly;
792 	pending_snapshot->dir = dir;
793 	pending_snapshot->inherit = inherit;
794 
795 	trans = btrfs_start_transaction(root, 0);
796 	if (IS_ERR(trans)) {
797 		ret = PTR_ERR(trans);
798 		goto fail;
799 	}
800 
801 	spin_lock(&fs_info->trans_lock);
802 	list_add(&pending_snapshot->list,
803 		 &trans->transaction->pending_snapshots);
804 	spin_unlock(&fs_info->trans_lock);
805 
806 	ret = btrfs_commit_transaction(trans);
807 	if (ret)
808 		goto fail;
809 
810 	ret = pending_snapshot->error;
811 	if (ret)
812 		goto fail;
813 
814 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
815 	if (ret)
816 		goto fail;
817 
818 	inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
819 	if (IS_ERR(inode)) {
820 		ret = PTR_ERR(inode);
821 		goto fail;
822 	}
823 
824 	d_instantiate(dentry, inode);
825 	ret = 0;
826 fail:
827 	btrfs_put_root(pending_snapshot->snap);
828 	btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
829 free_pending:
830 	kfree(pending_snapshot->root_item);
831 	btrfs_free_path(pending_snapshot->path);
832 	kfree(pending_snapshot);
833 
834 	return ret;
835 }
836 
837 /*  copy of may_delete in fs/namei.c()
838  *	Check whether we can remove a link victim from directory dir, check
839  *  whether the type of victim is right.
840  *  1. We can't do it if dir is read-only (done in permission())
841  *  2. We should have write and exec permissions on dir
842  *  3. We can't remove anything from append-only dir
843  *  4. We can't do anything with immutable dir (done in permission())
844  *  5. If the sticky bit on dir is set we should either
845  *	a. be owner of dir, or
846  *	b. be owner of victim, or
847  *	c. have CAP_FOWNER capability
848  *  6. If the victim is append-only or immutable we can't do anything with
849  *     links pointing to it.
850  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
851  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
852  *  9. We can't remove a root or mountpoint.
853  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
854  *     nfs_async_unlink().
855  */
856 
857 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
858 {
859 	int error;
860 
861 	if (d_really_is_negative(victim))
862 		return -ENOENT;
863 
864 	BUG_ON(d_inode(victim->d_parent) != dir);
865 	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
866 
867 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
868 	if (error)
869 		return error;
870 	if (IS_APPEND(dir))
871 		return -EPERM;
872 	if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
873 	    IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
874 		return -EPERM;
875 	if (isdir) {
876 		if (!d_is_dir(victim))
877 			return -ENOTDIR;
878 		if (IS_ROOT(victim))
879 			return -EBUSY;
880 	} else if (d_is_dir(victim))
881 		return -EISDIR;
882 	if (IS_DEADDIR(dir))
883 		return -ENOENT;
884 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
885 		return -EBUSY;
886 	return 0;
887 }
888 
889 /* copy of may_create in fs/namei.c() */
890 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
891 {
892 	if (d_really_is_positive(child))
893 		return -EEXIST;
894 	if (IS_DEADDIR(dir))
895 		return -ENOENT;
896 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
897 }
898 
899 /*
900  * Create a new subvolume below @parent.  This is largely modeled after
901  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
902  * inside this filesystem so it's quite a bit simpler.
903  */
904 static noinline int btrfs_mksubvol(const struct path *parent,
905 				   const char *name, int namelen,
906 				   struct btrfs_root *snap_src,
907 				   bool readonly,
908 				   struct btrfs_qgroup_inherit *inherit)
909 {
910 	struct inode *dir = d_inode(parent->dentry);
911 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
912 	struct dentry *dentry;
913 	int error;
914 
915 	error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
916 	if (error == -EINTR)
917 		return error;
918 
919 	dentry = lookup_one_len(name, parent->dentry, namelen);
920 	error = PTR_ERR(dentry);
921 	if (IS_ERR(dentry))
922 		goto out_unlock;
923 
924 	error = btrfs_may_create(dir, dentry);
925 	if (error)
926 		goto out_dput;
927 
928 	/*
929 	 * even if this name doesn't exist, we may get hash collisions.
930 	 * check for them now when we can safely fail
931 	 */
932 	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
933 					       dir->i_ino, name,
934 					       namelen);
935 	if (error)
936 		goto out_dput;
937 
938 	down_read(&fs_info->subvol_sem);
939 
940 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
941 		goto out_up_read;
942 
943 	if (snap_src)
944 		error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
945 	else
946 		error = create_subvol(dir, dentry, name, namelen, inherit);
947 
948 	if (!error)
949 		fsnotify_mkdir(dir, dentry);
950 out_up_read:
951 	up_read(&fs_info->subvol_sem);
952 out_dput:
953 	dput(dentry);
954 out_unlock:
955 	inode_unlock(dir);
956 	return error;
957 }
958 
959 static noinline int btrfs_mksnapshot(const struct path *parent,
960 				   const char *name, int namelen,
961 				   struct btrfs_root *root,
962 				   bool readonly,
963 				   struct btrfs_qgroup_inherit *inherit)
964 {
965 	int ret;
966 	bool snapshot_force_cow = false;
967 
968 	/*
969 	 * Force new buffered writes to reserve space even when NOCOW is
970 	 * possible. This is to avoid later writeback (running dealloc) to
971 	 * fallback to COW mode and unexpectedly fail with ENOSPC.
972 	 */
973 	btrfs_drew_read_lock(&root->snapshot_lock);
974 
975 	ret = btrfs_start_delalloc_snapshot(root);
976 	if (ret)
977 		goto out;
978 
979 	/*
980 	 * All previous writes have started writeback in NOCOW mode, so now
981 	 * we force future writes to fallback to COW mode during snapshot
982 	 * creation.
983 	 */
984 	atomic_inc(&root->snapshot_force_cow);
985 	snapshot_force_cow = true;
986 
987 	btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
988 
989 	ret = btrfs_mksubvol(parent, name, namelen,
990 			     root, readonly, inherit);
991 out:
992 	if (snapshot_force_cow)
993 		atomic_dec(&root->snapshot_force_cow);
994 	btrfs_drew_read_unlock(&root->snapshot_lock);
995 	return ret;
996 }
997 
998 /*
999  * When we're defragging a range, we don't want to kick it off again
1000  * if it is really just waiting for delalloc to send it down.
1001  * If we find a nice big extent or delalloc range for the bytes in the
1002  * file you want to defrag, we return 0 to let you know to skip this
1003  * part of the file
1004  */
1005 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
1006 {
1007 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1008 	struct extent_map *em = NULL;
1009 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1010 	u64 end;
1011 
1012 	read_lock(&em_tree->lock);
1013 	em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1014 	read_unlock(&em_tree->lock);
1015 
1016 	if (em) {
1017 		end = extent_map_end(em);
1018 		free_extent_map(em);
1019 		if (end - offset > thresh)
1020 			return 0;
1021 	}
1022 	/* if we already have a nice delalloc here, just stop */
1023 	thresh /= 2;
1024 	end = count_range_bits(io_tree, &offset, offset + thresh,
1025 			       thresh, EXTENT_DELALLOC, 1);
1026 	if (end >= thresh)
1027 		return 0;
1028 	return 1;
1029 }
1030 
1031 /*
1032  * helper function to walk through a file and find extents
1033  * newer than a specific transid, and smaller than thresh.
1034  *
1035  * This is used by the defragging code to find new and small
1036  * extents
1037  */
1038 static int find_new_extents(struct btrfs_root *root,
1039 			    struct inode *inode, u64 newer_than,
1040 			    u64 *off, u32 thresh)
1041 {
1042 	struct btrfs_path *path;
1043 	struct btrfs_key min_key;
1044 	struct extent_buffer *leaf;
1045 	struct btrfs_file_extent_item *extent;
1046 	int type;
1047 	int ret;
1048 	u64 ino = btrfs_ino(BTRFS_I(inode));
1049 
1050 	path = btrfs_alloc_path();
1051 	if (!path)
1052 		return -ENOMEM;
1053 
1054 	min_key.objectid = ino;
1055 	min_key.type = BTRFS_EXTENT_DATA_KEY;
1056 	min_key.offset = *off;
1057 
1058 	while (1) {
1059 		ret = btrfs_search_forward(root, &min_key, path, newer_than);
1060 		if (ret != 0)
1061 			goto none;
1062 process_slot:
1063 		if (min_key.objectid != ino)
1064 			goto none;
1065 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1066 			goto none;
1067 
1068 		leaf = path->nodes[0];
1069 		extent = btrfs_item_ptr(leaf, path->slots[0],
1070 					struct btrfs_file_extent_item);
1071 
1072 		type = btrfs_file_extent_type(leaf, extent);
1073 		if (type == BTRFS_FILE_EXTENT_REG &&
1074 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1075 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
1076 			*off = min_key.offset;
1077 			btrfs_free_path(path);
1078 			return 0;
1079 		}
1080 
1081 		path->slots[0]++;
1082 		if (path->slots[0] < btrfs_header_nritems(leaf)) {
1083 			btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1084 			goto process_slot;
1085 		}
1086 
1087 		if (min_key.offset == (u64)-1)
1088 			goto none;
1089 
1090 		min_key.offset++;
1091 		btrfs_release_path(path);
1092 	}
1093 none:
1094 	btrfs_free_path(path);
1095 	return -ENOENT;
1096 }
1097 
1098 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1099 {
1100 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1101 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1102 	struct extent_map *em;
1103 	u64 len = PAGE_SIZE;
1104 
1105 	/*
1106 	 * hopefully we have this extent in the tree already, try without
1107 	 * the full extent lock
1108 	 */
1109 	read_lock(&em_tree->lock);
1110 	em = lookup_extent_mapping(em_tree, start, len);
1111 	read_unlock(&em_tree->lock);
1112 
1113 	if (!em) {
1114 		struct extent_state *cached = NULL;
1115 		u64 end = start + len - 1;
1116 
1117 		/* get the big lock and read metadata off disk */
1118 		lock_extent_bits(io_tree, start, end, &cached);
1119 		em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1120 		unlock_extent_cached(io_tree, start, end, &cached);
1121 
1122 		if (IS_ERR(em))
1123 			return NULL;
1124 	}
1125 
1126 	return em;
1127 }
1128 
1129 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1130 {
1131 	struct extent_map *next;
1132 	bool ret = true;
1133 
1134 	/* this is the last extent */
1135 	if (em->start + em->len >= i_size_read(inode))
1136 		return false;
1137 
1138 	next = defrag_lookup_extent(inode, em->start + em->len);
1139 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1140 		ret = false;
1141 	else if ((em->block_start + em->block_len == next->block_start) &&
1142 		 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1143 		ret = false;
1144 
1145 	free_extent_map(next);
1146 	return ret;
1147 }
1148 
1149 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1150 			       u64 *last_len, u64 *skip, u64 *defrag_end,
1151 			       int compress)
1152 {
1153 	struct extent_map *em;
1154 	int ret = 1;
1155 	bool next_mergeable = true;
1156 	bool prev_mergeable = true;
1157 
1158 	/*
1159 	 * make sure that once we start defragging an extent, we keep on
1160 	 * defragging it
1161 	 */
1162 	if (start < *defrag_end)
1163 		return 1;
1164 
1165 	*skip = 0;
1166 
1167 	em = defrag_lookup_extent(inode, start);
1168 	if (!em)
1169 		return 0;
1170 
1171 	/* this will cover holes, and inline extents */
1172 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1173 		ret = 0;
1174 		goto out;
1175 	}
1176 
1177 	if (!*defrag_end)
1178 		prev_mergeable = false;
1179 
1180 	next_mergeable = defrag_check_next_extent(inode, em);
1181 	/*
1182 	 * we hit a real extent, if it is big or the next extent is not a
1183 	 * real extent, don't bother defragging it
1184 	 */
1185 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1186 	    (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1187 		ret = 0;
1188 out:
1189 	/*
1190 	 * last_len ends up being a counter of how many bytes we've defragged.
1191 	 * every time we choose not to defrag an extent, we reset *last_len
1192 	 * so that the next tiny extent will force a defrag.
1193 	 *
1194 	 * The end result of this is that tiny extents before a single big
1195 	 * extent will force at least part of that big extent to be defragged.
1196 	 */
1197 	if (ret) {
1198 		*defrag_end = extent_map_end(em);
1199 	} else {
1200 		*last_len = 0;
1201 		*skip = extent_map_end(em);
1202 		*defrag_end = 0;
1203 	}
1204 
1205 	free_extent_map(em);
1206 	return ret;
1207 }
1208 
1209 /*
1210  * it doesn't do much good to defrag one or two pages
1211  * at a time.  This pulls in a nice chunk of pages
1212  * to COW and defrag.
1213  *
1214  * It also makes sure the delalloc code has enough
1215  * dirty data to avoid making new small extents as part
1216  * of the defrag
1217  *
1218  * It's a good idea to start RA on this range
1219  * before calling this.
1220  */
1221 static int cluster_pages_for_defrag(struct inode *inode,
1222 				    struct page **pages,
1223 				    unsigned long start_index,
1224 				    unsigned long num_pages)
1225 {
1226 	unsigned long file_end;
1227 	u64 isize = i_size_read(inode);
1228 	u64 page_start;
1229 	u64 page_end;
1230 	u64 page_cnt;
1231 	int ret;
1232 	int i;
1233 	int i_done;
1234 	struct btrfs_ordered_extent *ordered;
1235 	struct extent_state *cached_state = NULL;
1236 	struct extent_io_tree *tree;
1237 	struct extent_changeset *data_reserved = NULL;
1238 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1239 
1240 	file_end = (isize - 1) >> PAGE_SHIFT;
1241 	if (!isize || start_index > file_end)
1242 		return 0;
1243 
1244 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1245 
1246 	ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1247 			start_index << PAGE_SHIFT,
1248 			page_cnt << PAGE_SHIFT);
1249 	if (ret)
1250 		return ret;
1251 	i_done = 0;
1252 	tree = &BTRFS_I(inode)->io_tree;
1253 
1254 	/* step one, lock all the pages */
1255 	for (i = 0; i < page_cnt; i++) {
1256 		struct page *page;
1257 again:
1258 		page = find_or_create_page(inode->i_mapping,
1259 					   start_index + i, mask);
1260 		if (!page)
1261 			break;
1262 
1263 		page_start = page_offset(page);
1264 		page_end = page_start + PAGE_SIZE - 1;
1265 		while (1) {
1266 			lock_extent_bits(tree, page_start, page_end,
1267 					 &cached_state);
1268 			ordered = btrfs_lookup_ordered_extent(inode,
1269 							      page_start);
1270 			unlock_extent_cached(tree, page_start, page_end,
1271 					     &cached_state);
1272 			if (!ordered)
1273 				break;
1274 
1275 			unlock_page(page);
1276 			btrfs_start_ordered_extent(inode, ordered, 1);
1277 			btrfs_put_ordered_extent(ordered);
1278 			lock_page(page);
1279 			/*
1280 			 * we unlocked the page above, so we need check if
1281 			 * it was released or not.
1282 			 */
1283 			if (page->mapping != inode->i_mapping) {
1284 				unlock_page(page);
1285 				put_page(page);
1286 				goto again;
1287 			}
1288 		}
1289 
1290 		if (!PageUptodate(page)) {
1291 			btrfs_readpage(NULL, page);
1292 			lock_page(page);
1293 			if (!PageUptodate(page)) {
1294 				unlock_page(page);
1295 				put_page(page);
1296 				ret = -EIO;
1297 				break;
1298 			}
1299 		}
1300 
1301 		if (page->mapping != inode->i_mapping) {
1302 			unlock_page(page);
1303 			put_page(page);
1304 			goto again;
1305 		}
1306 
1307 		pages[i] = page;
1308 		i_done++;
1309 	}
1310 	if (!i_done || ret)
1311 		goto out;
1312 
1313 	if (!(inode->i_sb->s_flags & SB_ACTIVE))
1314 		goto out;
1315 
1316 	/*
1317 	 * so now we have a nice long stream of locked
1318 	 * and up to date pages, lets wait on them
1319 	 */
1320 	for (i = 0; i < i_done; i++)
1321 		wait_on_page_writeback(pages[i]);
1322 
1323 	page_start = page_offset(pages[0]);
1324 	page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1325 
1326 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1327 			 page_start, page_end - 1, &cached_state);
1328 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1329 			  page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1330 			  EXTENT_DEFRAG, 0, 0, &cached_state);
1331 
1332 	if (i_done != page_cnt) {
1333 		spin_lock(&BTRFS_I(inode)->lock);
1334 		btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1335 		spin_unlock(&BTRFS_I(inode)->lock);
1336 		btrfs_delalloc_release_space(inode, data_reserved,
1337 				start_index << PAGE_SHIFT,
1338 				(page_cnt - i_done) << PAGE_SHIFT, true);
1339 	}
1340 
1341 
1342 	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1343 			  &cached_state);
1344 
1345 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1346 			     page_start, page_end - 1, &cached_state);
1347 
1348 	for (i = 0; i < i_done; i++) {
1349 		clear_page_dirty_for_io(pages[i]);
1350 		ClearPageChecked(pages[i]);
1351 		set_page_extent_mapped(pages[i]);
1352 		set_page_dirty(pages[i]);
1353 		unlock_page(pages[i]);
1354 		put_page(pages[i]);
1355 	}
1356 	btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1357 	extent_changeset_free(data_reserved);
1358 	return i_done;
1359 out:
1360 	for (i = 0; i < i_done; i++) {
1361 		unlock_page(pages[i]);
1362 		put_page(pages[i]);
1363 	}
1364 	btrfs_delalloc_release_space(inode, data_reserved,
1365 			start_index << PAGE_SHIFT,
1366 			page_cnt << PAGE_SHIFT, true);
1367 	btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1368 	extent_changeset_free(data_reserved);
1369 	return ret;
1370 
1371 }
1372 
1373 int btrfs_defrag_file(struct inode *inode, struct file *file,
1374 		      struct btrfs_ioctl_defrag_range_args *range,
1375 		      u64 newer_than, unsigned long max_to_defrag)
1376 {
1377 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1378 	struct btrfs_root *root = BTRFS_I(inode)->root;
1379 	struct file_ra_state *ra = NULL;
1380 	unsigned long last_index;
1381 	u64 isize = i_size_read(inode);
1382 	u64 last_len = 0;
1383 	u64 skip = 0;
1384 	u64 defrag_end = 0;
1385 	u64 newer_off = range->start;
1386 	unsigned long i;
1387 	unsigned long ra_index = 0;
1388 	int ret;
1389 	int defrag_count = 0;
1390 	int compress_type = BTRFS_COMPRESS_ZLIB;
1391 	u32 extent_thresh = range->extent_thresh;
1392 	unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1393 	unsigned long cluster = max_cluster;
1394 	u64 new_align = ~((u64)SZ_128K - 1);
1395 	struct page **pages = NULL;
1396 	bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1397 
1398 	if (isize == 0)
1399 		return 0;
1400 
1401 	if (range->start >= isize)
1402 		return -EINVAL;
1403 
1404 	if (do_compress) {
1405 		if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1406 			return -EINVAL;
1407 		if (range->compress_type)
1408 			compress_type = range->compress_type;
1409 	}
1410 
1411 	if (extent_thresh == 0)
1412 		extent_thresh = SZ_256K;
1413 
1414 	/*
1415 	 * If we were not given a file, allocate a readahead context. As
1416 	 * readahead is just an optimization, defrag will work without it so
1417 	 * we don't error out.
1418 	 */
1419 	if (!file) {
1420 		ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1421 		if (ra)
1422 			file_ra_state_init(ra, inode->i_mapping);
1423 	} else {
1424 		ra = &file->f_ra;
1425 	}
1426 
1427 	pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1428 	if (!pages) {
1429 		ret = -ENOMEM;
1430 		goto out_ra;
1431 	}
1432 
1433 	/* find the last page to defrag */
1434 	if (range->start + range->len > range->start) {
1435 		last_index = min_t(u64, isize - 1,
1436 			 range->start + range->len - 1) >> PAGE_SHIFT;
1437 	} else {
1438 		last_index = (isize - 1) >> PAGE_SHIFT;
1439 	}
1440 
1441 	if (newer_than) {
1442 		ret = find_new_extents(root, inode, newer_than,
1443 				       &newer_off, SZ_64K);
1444 		if (!ret) {
1445 			range->start = newer_off;
1446 			/*
1447 			 * we always align our defrag to help keep
1448 			 * the extents in the file evenly spaced
1449 			 */
1450 			i = (newer_off & new_align) >> PAGE_SHIFT;
1451 		} else
1452 			goto out_ra;
1453 	} else {
1454 		i = range->start >> PAGE_SHIFT;
1455 	}
1456 	if (!max_to_defrag)
1457 		max_to_defrag = last_index - i + 1;
1458 
1459 	/*
1460 	 * make writeback starts from i, so the defrag range can be
1461 	 * written sequentially.
1462 	 */
1463 	if (i < inode->i_mapping->writeback_index)
1464 		inode->i_mapping->writeback_index = i;
1465 
1466 	while (i <= last_index && defrag_count < max_to_defrag &&
1467 	       (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1468 		/*
1469 		 * make sure we stop running if someone unmounts
1470 		 * the FS
1471 		 */
1472 		if (!(inode->i_sb->s_flags & SB_ACTIVE))
1473 			break;
1474 
1475 		if (btrfs_defrag_cancelled(fs_info)) {
1476 			btrfs_debug(fs_info, "defrag_file cancelled");
1477 			ret = -EAGAIN;
1478 			break;
1479 		}
1480 
1481 		if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1482 					 extent_thresh, &last_len, &skip,
1483 					 &defrag_end, do_compress)){
1484 			unsigned long next;
1485 			/*
1486 			 * the should_defrag function tells us how much to skip
1487 			 * bump our counter by the suggested amount
1488 			 */
1489 			next = DIV_ROUND_UP(skip, PAGE_SIZE);
1490 			i = max(i + 1, next);
1491 			continue;
1492 		}
1493 
1494 		if (!newer_than) {
1495 			cluster = (PAGE_ALIGN(defrag_end) >>
1496 				   PAGE_SHIFT) - i;
1497 			cluster = min(cluster, max_cluster);
1498 		} else {
1499 			cluster = max_cluster;
1500 		}
1501 
1502 		if (i + cluster > ra_index) {
1503 			ra_index = max(i, ra_index);
1504 			if (ra)
1505 				page_cache_sync_readahead(inode->i_mapping, ra,
1506 						file, ra_index, cluster);
1507 			ra_index += cluster;
1508 		}
1509 
1510 		inode_lock(inode);
1511 		if (IS_SWAPFILE(inode)) {
1512 			ret = -ETXTBSY;
1513 		} else {
1514 			if (do_compress)
1515 				BTRFS_I(inode)->defrag_compress = compress_type;
1516 			ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1517 		}
1518 		if (ret < 0) {
1519 			inode_unlock(inode);
1520 			goto out_ra;
1521 		}
1522 
1523 		defrag_count += ret;
1524 		balance_dirty_pages_ratelimited(inode->i_mapping);
1525 		inode_unlock(inode);
1526 
1527 		if (newer_than) {
1528 			if (newer_off == (u64)-1)
1529 				break;
1530 
1531 			if (ret > 0)
1532 				i += ret;
1533 
1534 			newer_off = max(newer_off + 1,
1535 					(u64)i << PAGE_SHIFT);
1536 
1537 			ret = find_new_extents(root, inode, newer_than,
1538 					       &newer_off, SZ_64K);
1539 			if (!ret) {
1540 				range->start = newer_off;
1541 				i = (newer_off & new_align) >> PAGE_SHIFT;
1542 			} else {
1543 				break;
1544 			}
1545 		} else {
1546 			if (ret > 0) {
1547 				i += ret;
1548 				last_len += ret << PAGE_SHIFT;
1549 			} else {
1550 				i++;
1551 				last_len = 0;
1552 			}
1553 		}
1554 	}
1555 
1556 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1557 		filemap_flush(inode->i_mapping);
1558 		if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1559 			     &BTRFS_I(inode)->runtime_flags))
1560 			filemap_flush(inode->i_mapping);
1561 	}
1562 
1563 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1564 		btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1565 	} else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1566 		btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1567 	}
1568 
1569 	ret = defrag_count;
1570 
1571 out_ra:
1572 	if (do_compress) {
1573 		inode_lock(inode);
1574 		BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1575 		inode_unlock(inode);
1576 	}
1577 	if (!file)
1578 		kfree(ra);
1579 	kfree(pages);
1580 	return ret;
1581 }
1582 
1583 static noinline int btrfs_ioctl_resize(struct file *file,
1584 					void __user *arg)
1585 {
1586 	struct inode *inode = file_inode(file);
1587 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1588 	u64 new_size;
1589 	u64 old_size;
1590 	u64 devid = 1;
1591 	struct btrfs_root *root = BTRFS_I(inode)->root;
1592 	struct btrfs_ioctl_vol_args *vol_args;
1593 	struct btrfs_trans_handle *trans;
1594 	struct btrfs_device *device = NULL;
1595 	char *sizestr;
1596 	char *retptr;
1597 	char *devstr = NULL;
1598 	int ret = 0;
1599 	int mod = 0;
1600 
1601 	if (!capable(CAP_SYS_ADMIN))
1602 		return -EPERM;
1603 
1604 	ret = mnt_want_write_file(file);
1605 	if (ret)
1606 		return ret;
1607 
1608 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1609 		mnt_drop_write_file(file);
1610 		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1611 	}
1612 
1613 	vol_args = memdup_user(arg, sizeof(*vol_args));
1614 	if (IS_ERR(vol_args)) {
1615 		ret = PTR_ERR(vol_args);
1616 		goto out;
1617 	}
1618 
1619 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1620 
1621 	sizestr = vol_args->name;
1622 	devstr = strchr(sizestr, ':');
1623 	if (devstr) {
1624 		sizestr = devstr + 1;
1625 		*devstr = '\0';
1626 		devstr = vol_args->name;
1627 		ret = kstrtoull(devstr, 10, &devid);
1628 		if (ret)
1629 			goto out_free;
1630 		if (!devid) {
1631 			ret = -EINVAL;
1632 			goto out_free;
1633 		}
1634 		btrfs_info(fs_info, "resizing devid %llu", devid);
1635 	}
1636 
1637 	device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true);
1638 	if (!device) {
1639 		btrfs_info(fs_info, "resizer unable to find device %llu",
1640 			   devid);
1641 		ret = -ENODEV;
1642 		goto out_free;
1643 	}
1644 
1645 	if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1646 		btrfs_info(fs_info,
1647 			   "resizer unable to apply on readonly device %llu",
1648 		       devid);
1649 		ret = -EPERM;
1650 		goto out_free;
1651 	}
1652 
1653 	if (!strcmp(sizestr, "max"))
1654 		new_size = device->bdev->bd_inode->i_size;
1655 	else {
1656 		if (sizestr[0] == '-') {
1657 			mod = -1;
1658 			sizestr++;
1659 		} else if (sizestr[0] == '+') {
1660 			mod = 1;
1661 			sizestr++;
1662 		}
1663 		new_size = memparse(sizestr, &retptr);
1664 		if (*retptr != '\0' || new_size == 0) {
1665 			ret = -EINVAL;
1666 			goto out_free;
1667 		}
1668 	}
1669 
1670 	if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1671 		ret = -EPERM;
1672 		goto out_free;
1673 	}
1674 
1675 	old_size = btrfs_device_get_total_bytes(device);
1676 
1677 	if (mod < 0) {
1678 		if (new_size > old_size) {
1679 			ret = -EINVAL;
1680 			goto out_free;
1681 		}
1682 		new_size = old_size - new_size;
1683 	} else if (mod > 0) {
1684 		if (new_size > ULLONG_MAX - old_size) {
1685 			ret = -ERANGE;
1686 			goto out_free;
1687 		}
1688 		new_size = old_size + new_size;
1689 	}
1690 
1691 	if (new_size < SZ_256M) {
1692 		ret = -EINVAL;
1693 		goto out_free;
1694 	}
1695 	if (new_size > device->bdev->bd_inode->i_size) {
1696 		ret = -EFBIG;
1697 		goto out_free;
1698 	}
1699 
1700 	new_size = round_down(new_size, fs_info->sectorsize);
1701 
1702 	if (new_size > old_size) {
1703 		trans = btrfs_start_transaction(root, 0);
1704 		if (IS_ERR(trans)) {
1705 			ret = PTR_ERR(trans);
1706 			goto out_free;
1707 		}
1708 		ret = btrfs_grow_device(trans, device, new_size);
1709 		btrfs_commit_transaction(trans);
1710 	} else if (new_size < old_size) {
1711 		ret = btrfs_shrink_device(device, new_size);
1712 	} /* equal, nothing need to do */
1713 
1714 	if (ret == 0 && new_size != old_size)
1715 		btrfs_info_in_rcu(fs_info,
1716 			"resize device %s (devid %llu) from %llu to %llu",
1717 			rcu_str_deref(device->name), device->devid,
1718 			old_size, new_size);
1719 out_free:
1720 	kfree(vol_args);
1721 out:
1722 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1723 	mnt_drop_write_file(file);
1724 	return ret;
1725 }
1726 
1727 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1728 				const char *name, unsigned long fd, int subvol,
1729 				bool readonly,
1730 				struct btrfs_qgroup_inherit *inherit)
1731 {
1732 	int namelen;
1733 	int ret = 0;
1734 
1735 	if (!S_ISDIR(file_inode(file)->i_mode))
1736 		return -ENOTDIR;
1737 
1738 	ret = mnt_want_write_file(file);
1739 	if (ret)
1740 		goto out;
1741 
1742 	namelen = strlen(name);
1743 	if (strchr(name, '/')) {
1744 		ret = -EINVAL;
1745 		goto out_drop_write;
1746 	}
1747 
1748 	if (name[0] == '.' &&
1749 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1750 		ret = -EEXIST;
1751 		goto out_drop_write;
1752 	}
1753 
1754 	if (subvol) {
1755 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1756 				     NULL, readonly, inherit);
1757 	} else {
1758 		struct fd src = fdget(fd);
1759 		struct inode *src_inode;
1760 		if (!src.file) {
1761 			ret = -EINVAL;
1762 			goto out_drop_write;
1763 		}
1764 
1765 		src_inode = file_inode(src.file);
1766 		if (src_inode->i_sb != file_inode(file)->i_sb) {
1767 			btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1768 				   "Snapshot src from another FS");
1769 			ret = -EXDEV;
1770 		} else if (!inode_owner_or_capable(src_inode)) {
1771 			/*
1772 			 * Subvolume creation is not restricted, but snapshots
1773 			 * are limited to own subvolumes only
1774 			 */
1775 			ret = -EPERM;
1776 		} else {
1777 			ret = btrfs_mksnapshot(&file->f_path, name, namelen,
1778 					     BTRFS_I(src_inode)->root,
1779 					     readonly, inherit);
1780 		}
1781 		fdput(src);
1782 	}
1783 out_drop_write:
1784 	mnt_drop_write_file(file);
1785 out:
1786 	return ret;
1787 }
1788 
1789 static noinline int btrfs_ioctl_snap_create(struct file *file,
1790 					    void __user *arg, int subvol)
1791 {
1792 	struct btrfs_ioctl_vol_args *vol_args;
1793 	int ret;
1794 
1795 	if (!S_ISDIR(file_inode(file)->i_mode))
1796 		return -ENOTDIR;
1797 
1798 	vol_args = memdup_user(arg, sizeof(*vol_args));
1799 	if (IS_ERR(vol_args))
1800 		return PTR_ERR(vol_args);
1801 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1802 
1803 	ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1804 					subvol, false, NULL);
1805 
1806 	kfree(vol_args);
1807 	return ret;
1808 }
1809 
1810 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1811 					       void __user *arg, int subvol)
1812 {
1813 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1814 	int ret;
1815 	bool readonly = false;
1816 	struct btrfs_qgroup_inherit *inherit = NULL;
1817 
1818 	if (!S_ISDIR(file_inode(file)->i_mode))
1819 		return -ENOTDIR;
1820 
1821 	vol_args = memdup_user(arg, sizeof(*vol_args));
1822 	if (IS_ERR(vol_args))
1823 		return PTR_ERR(vol_args);
1824 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1825 
1826 	if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1827 		ret = -EOPNOTSUPP;
1828 		goto free_args;
1829 	}
1830 
1831 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1832 		readonly = true;
1833 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1834 		if (vol_args->size > PAGE_SIZE) {
1835 			ret = -EINVAL;
1836 			goto free_args;
1837 		}
1838 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1839 		if (IS_ERR(inherit)) {
1840 			ret = PTR_ERR(inherit);
1841 			goto free_args;
1842 		}
1843 	}
1844 
1845 	ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1846 					subvol, readonly, inherit);
1847 	if (ret)
1848 		goto free_inherit;
1849 free_inherit:
1850 	kfree(inherit);
1851 free_args:
1852 	kfree(vol_args);
1853 	return ret;
1854 }
1855 
1856 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1857 						void __user *arg)
1858 {
1859 	struct inode *inode = file_inode(file);
1860 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1861 	struct btrfs_root *root = BTRFS_I(inode)->root;
1862 	int ret = 0;
1863 	u64 flags = 0;
1864 
1865 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1866 		return -EINVAL;
1867 
1868 	down_read(&fs_info->subvol_sem);
1869 	if (btrfs_root_readonly(root))
1870 		flags |= BTRFS_SUBVOL_RDONLY;
1871 	up_read(&fs_info->subvol_sem);
1872 
1873 	if (copy_to_user(arg, &flags, sizeof(flags)))
1874 		ret = -EFAULT;
1875 
1876 	return ret;
1877 }
1878 
1879 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1880 					      void __user *arg)
1881 {
1882 	struct inode *inode = file_inode(file);
1883 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1884 	struct btrfs_root *root = BTRFS_I(inode)->root;
1885 	struct btrfs_trans_handle *trans;
1886 	u64 root_flags;
1887 	u64 flags;
1888 	int ret = 0;
1889 
1890 	if (!inode_owner_or_capable(inode))
1891 		return -EPERM;
1892 
1893 	ret = mnt_want_write_file(file);
1894 	if (ret)
1895 		goto out;
1896 
1897 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1898 		ret = -EINVAL;
1899 		goto out_drop_write;
1900 	}
1901 
1902 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1903 		ret = -EFAULT;
1904 		goto out_drop_write;
1905 	}
1906 
1907 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1908 		ret = -EOPNOTSUPP;
1909 		goto out_drop_write;
1910 	}
1911 
1912 	down_write(&fs_info->subvol_sem);
1913 
1914 	/* nothing to do */
1915 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1916 		goto out_drop_sem;
1917 
1918 	root_flags = btrfs_root_flags(&root->root_item);
1919 	if (flags & BTRFS_SUBVOL_RDONLY) {
1920 		btrfs_set_root_flags(&root->root_item,
1921 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1922 	} else {
1923 		/*
1924 		 * Block RO -> RW transition if this subvolume is involved in
1925 		 * send
1926 		 */
1927 		spin_lock(&root->root_item_lock);
1928 		if (root->send_in_progress == 0) {
1929 			btrfs_set_root_flags(&root->root_item,
1930 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1931 			spin_unlock(&root->root_item_lock);
1932 		} else {
1933 			spin_unlock(&root->root_item_lock);
1934 			btrfs_warn(fs_info,
1935 				   "Attempt to set subvolume %llu read-write during send",
1936 				   root->root_key.objectid);
1937 			ret = -EPERM;
1938 			goto out_drop_sem;
1939 		}
1940 	}
1941 
1942 	trans = btrfs_start_transaction(root, 1);
1943 	if (IS_ERR(trans)) {
1944 		ret = PTR_ERR(trans);
1945 		goto out_reset;
1946 	}
1947 
1948 	ret = btrfs_update_root(trans, fs_info->tree_root,
1949 				&root->root_key, &root->root_item);
1950 	if (ret < 0) {
1951 		btrfs_end_transaction(trans);
1952 		goto out_reset;
1953 	}
1954 
1955 	ret = btrfs_commit_transaction(trans);
1956 
1957 out_reset:
1958 	if (ret)
1959 		btrfs_set_root_flags(&root->root_item, root_flags);
1960 out_drop_sem:
1961 	up_write(&fs_info->subvol_sem);
1962 out_drop_write:
1963 	mnt_drop_write_file(file);
1964 out:
1965 	return ret;
1966 }
1967 
1968 static noinline int key_in_sk(struct btrfs_key *key,
1969 			      struct btrfs_ioctl_search_key *sk)
1970 {
1971 	struct btrfs_key test;
1972 	int ret;
1973 
1974 	test.objectid = sk->min_objectid;
1975 	test.type = sk->min_type;
1976 	test.offset = sk->min_offset;
1977 
1978 	ret = btrfs_comp_cpu_keys(key, &test);
1979 	if (ret < 0)
1980 		return 0;
1981 
1982 	test.objectid = sk->max_objectid;
1983 	test.type = sk->max_type;
1984 	test.offset = sk->max_offset;
1985 
1986 	ret = btrfs_comp_cpu_keys(key, &test);
1987 	if (ret > 0)
1988 		return 0;
1989 	return 1;
1990 }
1991 
1992 static noinline int copy_to_sk(struct btrfs_path *path,
1993 			       struct btrfs_key *key,
1994 			       struct btrfs_ioctl_search_key *sk,
1995 			       size_t *buf_size,
1996 			       char __user *ubuf,
1997 			       unsigned long *sk_offset,
1998 			       int *num_found)
1999 {
2000 	u64 found_transid;
2001 	struct extent_buffer *leaf;
2002 	struct btrfs_ioctl_search_header sh;
2003 	struct btrfs_key test;
2004 	unsigned long item_off;
2005 	unsigned long item_len;
2006 	int nritems;
2007 	int i;
2008 	int slot;
2009 	int ret = 0;
2010 
2011 	leaf = path->nodes[0];
2012 	slot = path->slots[0];
2013 	nritems = btrfs_header_nritems(leaf);
2014 
2015 	if (btrfs_header_generation(leaf) > sk->max_transid) {
2016 		i = nritems;
2017 		goto advance_key;
2018 	}
2019 	found_transid = btrfs_header_generation(leaf);
2020 
2021 	for (i = slot; i < nritems; i++) {
2022 		item_off = btrfs_item_ptr_offset(leaf, i);
2023 		item_len = btrfs_item_size_nr(leaf, i);
2024 
2025 		btrfs_item_key_to_cpu(leaf, key, i);
2026 		if (!key_in_sk(key, sk))
2027 			continue;
2028 
2029 		if (sizeof(sh) + item_len > *buf_size) {
2030 			if (*num_found) {
2031 				ret = 1;
2032 				goto out;
2033 			}
2034 
2035 			/*
2036 			 * return one empty item back for v1, which does not
2037 			 * handle -EOVERFLOW
2038 			 */
2039 
2040 			*buf_size = sizeof(sh) + item_len;
2041 			item_len = 0;
2042 			ret = -EOVERFLOW;
2043 		}
2044 
2045 		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2046 			ret = 1;
2047 			goto out;
2048 		}
2049 
2050 		sh.objectid = key->objectid;
2051 		sh.offset = key->offset;
2052 		sh.type = key->type;
2053 		sh.len = item_len;
2054 		sh.transid = found_transid;
2055 
2056 		/* copy search result header */
2057 		if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2058 			ret = -EFAULT;
2059 			goto out;
2060 		}
2061 
2062 		*sk_offset += sizeof(sh);
2063 
2064 		if (item_len) {
2065 			char __user *up = ubuf + *sk_offset;
2066 			/* copy the item */
2067 			if (read_extent_buffer_to_user(leaf, up,
2068 						       item_off, item_len)) {
2069 				ret = -EFAULT;
2070 				goto out;
2071 			}
2072 
2073 			*sk_offset += item_len;
2074 		}
2075 		(*num_found)++;
2076 
2077 		if (ret) /* -EOVERFLOW from above */
2078 			goto out;
2079 
2080 		if (*num_found >= sk->nr_items) {
2081 			ret = 1;
2082 			goto out;
2083 		}
2084 	}
2085 advance_key:
2086 	ret = 0;
2087 	test.objectid = sk->max_objectid;
2088 	test.type = sk->max_type;
2089 	test.offset = sk->max_offset;
2090 	if (btrfs_comp_cpu_keys(key, &test) >= 0)
2091 		ret = 1;
2092 	else if (key->offset < (u64)-1)
2093 		key->offset++;
2094 	else if (key->type < (u8)-1) {
2095 		key->offset = 0;
2096 		key->type++;
2097 	} else if (key->objectid < (u64)-1) {
2098 		key->offset = 0;
2099 		key->type = 0;
2100 		key->objectid++;
2101 	} else
2102 		ret = 1;
2103 out:
2104 	/*
2105 	 *  0: all items from this leaf copied, continue with next
2106 	 *  1: * more items can be copied, but unused buffer is too small
2107 	 *     * all items were found
2108 	 *     Either way, it will stops the loop which iterates to the next
2109 	 *     leaf
2110 	 *  -EOVERFLOW: item was to large for buffer
2111 	 *  -EFAULT: could not copy extent buffer back to userspace
2112 	 */
2113 	return ret;
2114 }
2115 
2116 static noinline int search_ioctl(struct inode *inode,
2117 				 struct btrfs_ioctl_search_key *sk,
2118 				 size_t *buf_size,
2119 				 char __user *ubuf)
2120 {
2121 	struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2122 	struct btrfs_root *root;
2123 	struct btrfs_key key;
2124 	struct btrfs_path *path;
2125 	int ret;
2126 	int num_found = 0;
2127 	unsigned long sk_offset = 0;
2128 
2129 	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2130 		*buf_size = sizeof(struct btrfs_ioctl_search_header);
2131 		return -EOVERFLOW;
2132 	}
2133 
2134 	path = btrfs_alloc_path();
2135 	if (!path)
2136 		return -ENOMEM;
2137 
2138 	if (sk->tree_id == 0) {
2139 		/* search the root of the inode that was passed */
2140 		root = btrfs_grab_root(BTRFS_I(inode)->root);
2141 	} else {
2142 		root = btrfs_get_fs_root(info, sk->tree_id, true);
2143 		if (IS_ERR(root)) {
2144 			btrfs_free_path(path);
2145 			return PTR_ERR(root);
2146 		}
2147 	}
2148 
2149 	key.objectid = sk->min_objectid;
2150 	key.type = sk->min_type;
2151 	key.offset = sk->min_offset;
2152 
2153 	while (1) {
2154 		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2155 		if (ret != 0) {
2156 			if (ret > 0)
2157 				ret = 0;
2158 			goto err;
2159 		}
2160 		ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2161 				 &sk_offset, &num_found);
2162 		btrfs_release_path(path);
2163 		if (ret)
2164 			break;
2165 
2166 	}
2167 	if (ret > 0)
2168 		ret = 0;
2169 err:
2170 	sk->nr_items = num_found;
2171 	btrfs_put_root(root);
2172 	btrfs_free_path(path);
2173 	return ret;
2174 }
2175 
2176 static noinline int btrfs_ioctl_tree_search(struct file *file,
2177 					   void __user *argp)
2178 {
2179 	struct btrfs_ioctl_search_args __user *uargs;
2180 	struct btrfs_ioctl_search_key sk;
2181 	struct inode *inode;
2182 	int ret;
2183 	size_t buf_size;
2184 
2185 	if (!capable(CAP_SYS_ADMIN))
2186 		return -EPERM;
2187 
2188 	uargs = (struct btrfs_ioctl_search_args __user *)argp;
2189 
2190 	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2191 		return -EFAULT;
2192 
2193 	buf_size = sizeof(uargs->buf);
2194 
2195 	inode = file_inode(file);
2196 	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2197 
2198 	/*
2199 	 * In the origin implementation an overflow is handled by returning a
2200 	 * search header with a len of zero, so reset ret.
2201 	 */
2202 	if (ret == -EOVERFLOW)
2203 		ret = 0;
2204 
2205 	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2206 		ret = -EFAULT;
2207 	return ret;
2208 }
2209 
2210 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2211 					       void __user *argp)
2212 {
2213 	struct btrfs_ioctl_search_args_v2 __user *uarg;
2214 	struct btrfs_ioctl_search_args_v2 args;
2215 	struct inode *inode;
2216 	int ret;
2217 	size_t buf_size;
2218 	const size_t buf_limit = SZ_16M;
2219 
2220 	if (!capable(CAP_SYS_ADMIN))
2221 		return -EPERM;
2222 
2223 	/* copy search header and buffer size */
2224 	uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2225 	if (copy_from_user(&args, uarg, sizeof(args)))
2226 		return -EFAULT;
2227 
2228 	buf_size = args.buf_size;
2229 
2230 	/* limit result size to 16MB */
2231 	if (buf_size > buf_limit)
2232 		buf_size = buf_limit;
2233 
2234 	inode = file_inode(file);
2235 	ret = search_ioctl(inode, &args.key, &buf_size,
2236 			   (char __user *)(&uarg->buf[0]));
2237 	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2238 		ret = -EFAULT;
2239 	else if (ret == -EOVERFLOW &&
2240 		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2241 		ret = -EFAULT;
2242 
2243 	return ret;
2244 }
2245 
2246 /*
2247  * Search INODE_REFs to identify path name of 'dirid' directory
2248  * in a 'tree_id' tree. and sets path name to 'name'.
2249  */
2250 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2251 				u64 tree_id, u64 dirid, char *name)
2252 {
2253 	struct btrfs_root *root;
2254 	struct btrfs_key key;
2255 	char *ptr;
2256 	int ret = -1;
2257 	int slot;
2258 	int len;
2259 	int total_len = 0;
2260 	struct btrfs_inode_ref *iref;
2261 	struct extent_buffer *l;
2262 	struct btrfs_path *path;
2263 
2264 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2265 		name[0]='\0';
2266 		return 0;
2267 	}
2268 
2269 	path = btrfs_alloc_path();
2270 	if (!path)
2271 		return -ENOMEM;
2272 
2273 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2274 
2275 	root = btrfs_get_fs_root(info, tree_id, true);
2276 	if (IS_ERR(root)) {
2277 		ret = PTR_ERR(root);
2278 		root = NULL;
2279 		goto out;
2280 	}
2281 
2282 	key.objectid = dirid;
2283 	key.type = BTRFS_INODE_REF_KEY;
2284 	key.offset = (u64)-1;
2285 
2286 	while (1) {
2287 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2288 		if (ret < 0)
2289 			goto out;
2290 		else if (ret > 0) {
2291 			ret = btrfs_previous_item(root, path, dirid,
2292 						  BTRFS_INODE_REF_KEY);
2293 			if (ret < 0)
2294 				goto out;
2295 			else if (ret > 0) {
2296 				ret = -ENOENT;
2297 				goto out;
2298 			}
2299 		}
2300 
2301 		l = path->nodes[0];
2302 		slot = path->slots[0];
2303 		btrfs_item_key_to_cpu(l, &key, slot);
2304 
2305 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2306 		len = btrfs_inode_ref_name_len(l, iref);
2307 		ptr -= len + 1;
2308 		total_len += len + 1;
2309 		if (ptr < name) {
2310 			ret = -ENAMETOOLONG;
2311 			goto out;
2312 		}
2313 
2314 		*(ptr + len) = '/';
2315 		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2316 
2317 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2318 			break;
2319 
2320 		btrfs_release_path(path);
2321 		key.objectid = key.offset;
2322 		key.offset = (u64)-1;
2323 		dirid = key.objectid;
2324 	}
2325 	memmove(name, ptr, total_len);
2326 	name[total_len] = '\0';
2327 	ret = 0;
2328 out:
2329 	btrfs_put_root(root);
2330 	btrfs_free_path(path);
2331 	return ret;
2332 }
2333 
2334 static int btrfs_search_path_in_tree_user(struct inode *inode,
2335 				struct btrfs_ioctl_ino_lookup_user_args *args)
2336 {
2337 	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2338 	struct super_block *sb = inode->i_sb;
2339 	struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2340 	u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2341 	u64 dirid = args->dirid;
2342 	unsigned long item_off;
2343 	unsigned long item_len;
2344 	struct btrfs_inode_ref *iref;
2345 	struct btrfs_root_ref *rref;
2346 	struct btrfs_root *root = NULL;
2347 	struct btrfs_path *path;
2348 	struct btrfs_key key, key2;
2349 	struct extent_buffer *leaf;
2350 	struct inode *temp_inode;
2351 	char *ptr;
2352 	int slot;
2353 	int len;
2354 	int total_len = 0;
2355 	int ret;
2356 
2357 	path = btrfs_alloc_path();
2358 	if (!path)
2359 		return -ENOMEM;
2360 
2361 	/*
2362 	 * If the bottom subvolume does not exist directly under upper_limit,
2363 	 * construct the path in from the bottom up.
2364 	 */
2365 	if (dirid != upper_limit.objectid) {
2366 		ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2367 
2368 		root = btrfs_get_fs_root(fs_info, treeid, true);
2369 		if (IS_ERR(root)) {
2370 			ret = PTR_ERR(root);
2371 			goto out;
2372 		}
2373 
2374 		key.objectid = dirid;
2375 		key.type = BTRFS_INODE_REF_KEY;
2376 		key.offset = (u64)-1;
2377 		while (1) {
2378 			ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2379 			if (ret < 0) {
2380 				goto out_put;
2381 			} else if (ret > 0) {
2382 				ret = btrfs_previous_item(root, path, dirid,
2383 							  BTRFS_INODE_REF_KEY);
2384 				if (ret < 0) {
2385 					goto out_put;
2386 				} else if (ret > 0) {
2387 					ret = -ENOENT;
2388 					goto out_put;
2389 				}
2390 			}
2391 
2392 			leaf = path->nodes[0];
2393 			slot = path->slots[0];
2394 			btrfs_item_key_to_cpu(leaf, &key, slot);
2395 
2396 			iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2397 			len = btrfs_inode_ref_name_len(leaf, iref);
2398 			ptr -= len + 1;
2399 			total_len += len + 1;
2400 			if (ptr < args->path) {
2401 				ret = -ENAMETOOLONG;
2402 				goto out_put;
2403 			}
2404 
2405 			*(ptr + len) = '/';
2406 			read_extent_buffer(leaf, ptr,
2407 					(unsigned long)(iref + 1), len);
2408 
2409 			/* Check the read+exec permission of this directory */
2410 			ret = btrfs_previous_item(root, path, dirid,
2411 						  BTRFS_INODE_ITEM_KEY);
2412 			if (ret < 0) {
2413 				goto out_put;
2414 			} else if (ret > 0) {
2415 				ret = -ENOENT;
2416 				goto out_put;
2417 			}
2418 
2419 			leaf = path->nodes[0];
2420 			slot = path->slots[0];
2421 			btrfs_item_key_to_cpu(leaf, &key2, slot);
2422 			if (key2.objectid != dirid) {
2423 				ret = -ENOENT;
2424 				goto out_put;
2425 			}
2426 
2427 			temp_inode = btrfs_iget(sb, key2.objectid, root);
2428 			if (IS_ERR(temp_inode)) {
2429 				ret = PTR_ERR(temp_inode);
2430 				goto out_put;
2431 			}
2432 			ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2433 			iput(temp_inode);
2434 			if (ret) {
2435 				ret = -EACCES;
2436 				goto out_put;
2437 			}
2438 
2439 			if (key.offset == upper_limit.objectid)
2440 				break;
2441 			if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2442 				ret = -EACCES;
2443 				goto out_put;
2444 			}
2445 
2446 			btrfs_release_path(path);
2447 			key.objectid = key.offset;
2448 			key.offset = (u64)-1;
2449 			dirid = key.objectid;
2450 		}
2451 
2452 		memmove(args->path, ptr, total_len);
2453 		args->path[total_len] = '\0';
2454 		btrfs_put_root(root);
2455 		root = NULL;
2456 		btrfs_release_path(path);
2457 	}
2458 
2459 	/* Get the bottom subvolume's name from ROOT_REF */
2460 	key.objectid = treeid;
2461 	key.type = BTRFS_ROOT_REF_KEY;
2462 	key.offset = args->treeid;
2463 	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2464 	if (ret < 0) {
2465 		goto out;
2466 	} else if (ret > 0) {
2467 		ret = -ENOENT;
2468 		goto out;
2469 	}
2470 
2471 	leaf = path->nodes[0];
2472 	slot = path->slots[0];
2473 	btrfs_item_key_to_cpu(leaf, &key, slot);
2474 
2475 	item_off = btrfs_item_ptr_offset(leaf, slot);
2476 	item_len = btrfs_item_size_nr(leaf, slot);
2477 	/* Check if dirid in ROOT_REF corresponds to passed dirid */
2478 	rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2479 	if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2480 		ret = -EINVAL;
2481 		goto out;
2482 	}
2483 
2484 	/* Copy subvolume's name */
2485 	item_off += sizeof(struct btrfs_root_ref);
2486 	item_len -= sizeof(struct btrfs_root_ref);
2487 	read_extent_buffer(leaf, args->name, item_off, item_len);
2488 	args->name[item_len] = 0;
2489 
2490 out_put:
2491 	btrfs_put_root(root);
2492 out:
2493 	btrfs_free_path(path);
2494 	return ret;
2495 }
2496 
2497 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2498 					   void __user *argp)
2499 {
2500 	struct btrfs_ioctl_ino_lookup_args *args;
2501 	struct inode *inode;
2502 	int ret = 0;
2503 
2504 	args = memdup_user(argp, sizeof(*args));
2505 	if (IS_ERR(args))
2506 		return PTR_ERR(args);
2507 
2508 	inode = file_inode(file);
2509 
2510 	/*
2511 	 * Unprivileged query to obtain the containing subvolume root id. The
2512 	 * path is reset so it's consistent with btrfs_search_path_in_tree.
2513 	 */
2514 	if (args->treeid == 0)
2515 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2516 
2517 	if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2518 		args->name[0] = 0;
2519 		goto out;
2520 	}
2521 
2522 	if (!capable(CAP_SYS_ADMIN)) {
2523 		ret = -EPERM;
2524 		goto out;
2525 	}
2526 
2527 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2528 					args->treeid, args->objectid,
2529 					args->name);
2530 
2531 out:
2532 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2533 		ret = -EFAULT;
2534 
2535 	kfree(args);
2536 	return ret;
2537 }
2538 
2539 /*
2540  * Version of ino_lookup ioctl (unprivileged)
2541  *
2542  * The main differences from ino_lookup ioctl are:
2543  *
2544  *   1. Read + Exec permission will be checked using inode_permission() during
2545  *      path construction. -EACCES will be returned in case of failure.
2546  *   2. Path construction will be stopped at the inode number which corresponds
2547  *      to the fd with which this ioctl is called. If constructed path does not
2548  *      exist under fd's inode, -EACCES will be returned.
2549  *   3. The name of bottom subvolume is also searched and filled.
2550  */
2551 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2552 {
2553 	struct btrfs_ioctl_ino_lookup_user_args *args;
2554 	struct inode *inode;
2555 	int ret;
2556 
2557 	args = memdup_user(argp, sizeof(*args));
2558 	if (IS_ERR(args))
2559 		return PTR_ERR(args);
2560 
2561 	inode = file_inode(file);
2562 
2563 	if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2564 	    BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2565 		/*
2566 		 * The subvolume does not exist under fd with which this is
2567 		 * called
2568 		 */
2569 		kfree(args);
2570 		return -EACCES;
2571 	}
2572 
2573 	ret = btrfs_search_path_in_tree_user(inode, args);
2574 
2575 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2576 		ret = -EFAULT;
2577 
2578 	kfree(args);
2579 	return ret;
2580 }
2581 
2582 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2583 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2584 {
2585 	struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2586 	struct btrfs_fs_info *fs_info;
2587 	struct btrfs_root *root;
2588 	struct btrfs_path *path;
2589 	struct btrfs_key key;
2590 	struct btrfs_root_item *root_item;
2591 	struct btrfs_root_ref *rref;
2592 	struct extent_buffer *leaf;
2593 	unsigned long item_off;
2594 	unsigned long item_len;
2595 	struct inode *inode;
2596 	int slot;
2597 	int ret = 0;
2598 
2599 	path = btrfs_alloc_path();
2600 	if (!path)
2601 		return -ENOMEM;
2602 
2603 	subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2604 	if (!subvol_info) {
2605 		btrfs_free_path(path);
2606 		return -ENOMEM;
2607 	}
2608 
2609 	inode = file_inode(file);
2610 	fs_info = BTRFS_I(inode)->root->fs_info;
2611 
2612 	/* Get root_item of inode's subvolume */
2613 	key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2614 	root = btrfs_get_fs_root(fs_info, key.objectid, true);
2615 	if (IS_ERR(root)) {
2616 		ret = PTR_ERR(root);
2617 		goto out_free;
2618 	}
2619 	root_item = &root->root_item;
2620 
2621 	subvol_info->treeid = key.objectid;
2622 
2623 	subvol_info->generation = btrfs_root_generation(root_item);
2624 	subvol_info->flags = btrfs_root_flags(root_item);
2625 
2626 	memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2627 	memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2628 						    BTRFS_UUID_SIZE);
2629 	memcpy(subvol_info->received_uuid, root_item->received_uuid,
2630 						    BTRFS_UUID_SIZE);
2631 
2632 	subvol_info->ctransid = btrfs_root_ctransid(root_item);
2633 	subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2634 	subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2635 
2636 	subvol_info->otransid = btrfs_root_otransid(root_item);
2637 	subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2638 	subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2639 
2640 	subvol_info->stransid = btrfs_root_stransid(root_item);
2641 	subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2642 	subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2643 
2644 	subvol_info->rtransid = btrfs_root_rtransid(root_item);
2645 	subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2646 	subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2647 
2648 	if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2649 		/* Search root tree for ROOT_BACKREF of this subvolume */
2650 		key.type = BTRFS_ROOT_BACKREF_KEY;
2651 		key.offset = 0;
2652 		ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2653 		if (ret < 0) {
2654 			goto out;
2655 		} else if (path->slots[0] >=
2656 			   btrfs_header_nritems(path->nodes[0])) {
2657 			ret = btrfs_next_leaf(fs_info->tree_root, path);
2658 			if (ret < 0) {
2659 				goto out;
2660 			} else if (ret > 0) {
2661 				ret = -EUCLEAN;
2662 				goto out;
2663 			}
2664 		}
2665 
2666 		leaf = path->nodes[0];
2667 		slot = path->slots[0];
2668 		btrfs_item_key_to_cpu(leaf, &key, slot);
2669 		if (key.objectid == subvol_info->treeid &&
2670 		    key.type == BTRFS_ROOT_BACKREF_KEY) {
2671 			subvol_info->parent_id = key.offset;
2672 
2673 			rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2674 			subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2675 
2676 			item_off = btrfs_item_ptr_offset(leaf, slot)
2677 					+ sizeof(struct btrfs_root_ref);
2678 			item_len = btrfs_item_size_nr(leaf, slot)
2679 					- sizeof(struct btrfs_root_ref);
2680 			read_extent_buffer(leaf, subvol_info->name,
2681 					   item_off, item_len);
2682 		} else {
2683 			ret = -ENOENT;
2684 			goto out;
2685 		}
2686 	}
2687 
2688 	if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2689 		ret = -EFAULT;
2690 
2691 out:
2692 	btrfs_put_root(root);
2693 out_free:
2694 	btrfs_free_path(path);
2695 	kfree(subvol_info);
2696 	return ret;
2697 }
2698 
2699 /*
2700  * Return ROOT_REF information of the subvolume containing this inode
2701  * except the subvolume name.
2702  */
2703 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2704 {
2705 	struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2706 	struct btrfs_root_ref *rref;
2707 	struct btrfs_root *root;
2708 	struct btrfs_path *path;
2709 	struct btrfs_key key;
2710 	struct extent_buffer *leaf;
2711 	struct inode *inode;
2712 	u64 objectid;
2713 	int slot;
2714 	int ret;
2715 	u8 found;
2716 
2717 	path = btrfs_alloc_path();
2718 	if (!path)
2719 		return -ENOMEM;
2720 
2721 	rootrefs = memdup_user(argp, sizeof(*rootrefs));
2722 	if (IS_ERR(rootrefs)) {
2723 		btrfs_free_path(path);
2724 		return PTR_ERR(rootrefs);
2725 	}
2726 
2727 	inode = file_inode(file);
2728 	root = BTRFS_I(inode)->root->fs_info->tree_root;
2729 	objectid = BTRFS_I(inode)->root->root_key.objectid;
2730 
2731 	key.objectid = objectid;
2732 	key.type = BTRFS_ROOT_REF_KEY;
2733 	key.offset = rootrefs->min_treeid;
2734 	found = 0;
2735 
2736 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2737 	if (ret < 0) {
2738 		goto out;
2739 	} else if (path->slots[0] >=
2740 		   btrfs_header_nritems(path->nodes[0])) {
2741 		ret = btrfs_next_leaf(root, path);
2742 		if (ret < 0) {
2743 			goto out;
2744 		} else if (ret > 0) {
2745 			ret = -EUCLEAN;
2746 			goto out;
2747 		}
2748 	}
2749 	while (1) {
2750 		leaf = path->nodes[0];
2751 		slot = path->slots[0];
2752 
2753 		btrfs_item_key_to_cpu(leaf, &key, slot);
2754 		if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2755 			ret = 0;
2756 			goto out;
2757 		}
2758 
2759 		if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2760 			ret = -EOVERFLOW;
2761 			goto out;
2762 		}
2763 
2764 		rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2765 		rootrefs->rootref[found].treeid = key.offset;
2766 		rootrefs->rootref[found].dirid =
2767 				  btrfs_root_ref_dirid(leaf, rref);
2768 		found++;
2769 
2770 		ret = btrfs_next_item(root, path);
2771 		if (ret < 0) {
2772 			goto out;
2773 		} else if (ret > 0) {
2774 			ret = -EUCLEAN;
2775 			goto out;
2776 		}
2777 	}
2778 
2779 out:
2780 	if (!ret || ret == -EOVERFLOW) {
2781 		rootrefs->num_items = found;
2782 		/* update min_treeid for next search */
2783 		if (found)
2784 			rootrefs->min_treeid =
2785 				rootrefs->rootref[found - 1].treeid + 1;
2786 		if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2787 			ret = -EFAULT;
2788 	}
2789 
2790 	kfree(rootrefs);
2791 	btrfs_free_path(path);
2792 
2793 	return ret;
2794 }
2795 
2796 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2797 					     void __user *arg,
2798 					     bool destroy_v2)
2799 {
2800 	struct dentry *parent = file->f_path.dentry;
2801 	struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2802 	struct dentry *dentry;
2803 	struct inode *dir = d_inode(parent);
2804 	struct inode *inode;
2805 	struct btrfs_root *root = BTRFS_I(dir)->root;
2806 	struct btrfs_root *dest = NULL;
2807 	struct btrfs_ioctl_vol_args *vol_args = NULL;
2808 	struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2809 	char *subvol_name, *subvol_name_ptr = NULL;
2810 	int subvol_namelen;
2811 	int err = 0;
2812 	bool destroy_parent = false;
2813 
2814 	if (destroy_v2) {
2815 		vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2816 		if (IS_ERR(vol_args2))
2817 			return PTR_ERR(vol_args2);
2818 
2819 		if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2820 			err = -EOPNOTSUPP;
2821 			goto out;
2822 		}
2823 
2824 		/*
2825 		 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2826 		 * name, same as v1 currently does.
2827 		 */
2828 		if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2829 			vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2830 			subvol_name = vol_args2->name;
2831 
2832 			err = mnt_want_write_file(file);
2833 			if (err)
2834 				goto out;
2835 		} else {
2836 			if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2837 				err = -EINVAL;
2838 				goto out;
2839 			}
2840 
2841 			err = mnt_want_write_file(file);
2842 			if (err)
2843 				goto out;
2844 
2845 			dentry = btrfs_get_dentry(fs_info->sb,
2846 					BTRFS_FIRST_FREE_OBJECTID,
2847 					vol_args2->subvolid, 0, 0);
2848 			if (IS_ERR(dentry)) {
2849 				err = PTR_ERR(dentry);
2850 				goto out_drop_write;
2851 			}
2852 
2853 			/*
2854 			 * Change the default parent since the subvolume being
2855 			 * deleted can be outside of the current mount point.
2856 			 */
2857 			parent = btrfs_get_parent(dentry);
2858 
2859 			/*
2860 			 * At this point dentry->d_name can point to '/' if the
2861 			 * subvolume we want to destroy is outsite of the
2862 			 * current mount point, so we need to release the
2863 			 * current dentry and execute the lookup to return a new
2864 			 * one with ->d_name pointing to the
2865 			 * <mount point>/subvol_name.
2866 			 */
2867 			dput(dentry);
2868 			if (IS_ERR(parent)) {
2869 				err = PTR_ERR(parent);
2870 				goto out_drop_write;
2871 			}
2872 			dir = d_inode(parent);
2873 
2874 			/*
2875 			 * If v2 was used with SPEC_BY_ID, a new parent was
2876 			 * allocated since the subvolume can be outside of the
2877 			 * current mount point. Later on we need to release this
2878 			 * new parent dentry.
2879 			 */
2880 			destroy_parent = true;
2881 
2882 			subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2883 						fs_info, vol_args2->subvolid);
2884 			if (IS_ERR(subvol_name_ptr)) {
2885 				err = PTR_ERR(subvol_name_ptr);
2886 				goto free_parent;
2887 			}
2888 			/* subvol_name_ptr is already NULL termined */
2889 			subvol_name = (char *)kbasename(subvol_name_ptr);
2890 		}
2891 	} else {
2892 		vol_args = memdup_user(arg, sizeof(*vol_args));
2893 		if (IS_ERR(vol_args))
2894 			return PTR_ERR(vol_args);
2895 
2896 		vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2897 		subvol_name = vol_args->name;
2898 
2899 		err = mnt_want_write_file(file);
2900 		if (err)
2901 			goto out;
2902 	}
2903 
2904 	subvol_namelen = strlen(subvol_name);
2905 
2906 	if (strchr(subvol_name, '/') ||
2907 	    strncmp(subvol_name, "..", subvol_namelen) == 0) {
2908 		err = -EINVAL;
2909 		goto free_subvol_name;
2910 	}
2911 
2912 	if (!S_ISDIR(dir->i_mode)) {
2913 		err = -ENOTDIR;
2914 		goto free_subvol_name;
2915 	}
2916 
2917 	err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2918 	if (err == -EINTR)
2919 		goto free_subvol_name;
2920 	dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
2921 	if (IS_ERR(dentry)) {
2922 		err = PTR_ERR(dentry);
2923 		goto out_unlock_dir;
2924 	}
2925 
2926 	if (d_really_is_negative(dentry)) {
2927 		err = -ENOENT;
2928 		goto out_dput;
2929 	}
2930 
2931 	inode = d_inode(dentry);
2932 	dest = BTRFS_I(inode)->root;
2933 	if (!capable(CAP_SYS_ADMIN)) {
2934 		/*
2935 		 * Regular user.  Only allow this with a special mount
2936 		 * option, when the user has write+exec access to the
2937 		 * subvol root, and when rmdir(2) would have been
2938 		 * allowed.
2939 		 *
2940 		 * Note that this is _not_ check that the subvol is
2941 		 * empty or doesn't contain data that we wouldn't
2942 		 * otherwise be able to delete.
2943 		 *
2944 		 * Users who want to delete empty subvols should try
2945 		 * rmdir(2).
2946 		 */
2947 		err = -EPERM;
2948 		if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2949 			goto out_dput;
2950 
2951 		/*
2952 		 * Do not allow deletion if the parent dir is the same
2953 		 * as the dir to be deleted.  That means the ioctl
2954 		 * must be called on the dentry referencing the root
2955 		 * of the subvol, not a random directory contained
2956 		 * within it.
2957 		 */
2958 		err = -EINVAL;
2959 		if (root == dest)
2960 			goto out_dput;
2961 
2962 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2963 		if (err)
2964 			goto out_dput;
2965 	}
2966 
2967 	/* check if subvolume may be deleted by a user */
2968 	err = btrfs_may_delete(dir, dentry, 1);
2969 	if (err)
2970 		goto out_dput;
2971 
2972 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2973 		err = -EINVAL;
2974 		goto out_dput;
2975 	}
2976 
2977 	inode_lock(inode);
2978 	err = btrfs_delete_subvolume(dir, dentry);
2979 	inode_unlock(inode);
2980 	if (!err) {
2981 		fsnotify_rmdir(dir, dentry);
2982 		d_delete(dentry);
2983 	}
2984 
2985 out_dput:
2986 	dput(dentry);
2987 out_unlock_dir:
2988 	inode_unlock(dir);
2989 free_subvol_name:
2990 	kfree(subvol_name_ptr);
2991 free_parent:
2992 	if (destroy_parent)
2993 		dput(parent);
2994 out_drop_write:
2995 	mnt_drop_write_file(file);
2996 out:
2997 	kfree(vol_args2);
2998 	kfree(vol_args);
2999 	return err;
3000 }
3001 
3002 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3003 {
3004 	struct inode *inode = file_inode(file);
3005 	struct btrfs_root *root = BTRFS_I(inode)->root;
3006 	struct btrfs_ioctl_defrag_range_args *range;
3007 	int ret;
3008 
3009 	ret = mnt_want_write_file(file);
3010 	if (ret)
3011 		return ret;
3012 
3013 	if (btrfs_root_readonly(root)) {
3014 		ret = -EROFS;
3015 		goto out;
3016 	}
3017 
3018 	switch (inode->i_mode & S_IFMT) {
3019 	case S_IFDIR:
3020 		if (!capable(CAP_SYS_ADMIN)) {
3021 			ret = -EPERM;
3022 			goto out;
3023 		}
3024 		ret = btrfs_defrag_root(root);
3025 		break;
3026 	case S_IFREG:
3027 		/*
3028 		 * Note that this does not check the file descriptor for write
3029 		 * access. This prevents defragmenting executables that are
3030 		 * running and allows defrag on files open in read-only mode.
3031 		 */
3032 		if (!capable(CAP_SYS_ADMIN) &&
3033 		    inode_permission(inode, MAY_WRITE)) {
3034 			ret = -EPERM;
3035 			goto out;
3036 		}
3037 
3038 		range = kzalloc(sizeof(*range), GFP_KERNEL);
3039 		if (!range) {
3040 			ret = -ENOMEM;
3041 			goto out;
3042 		}
3043 
3044 		if (argp) {
3045 			if (copy_from_user(range, argp,
3046 					   sizeof(*range))) {
3047 				ret = -EFAULT;
3048 				kfree(range);
3049 				goto out;
3050 			}
3051 			/* compression requires us to start the IO */
3052 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3053 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3054 				range->extent_thresh = (u32)-1;
3055 			}
3056 		} else {
3057 			/* the rest are all set to zero by kzalloc */
3058 			range->len = (u64)-1;
3059 		}
3060 		ret = btrfs_defrag_file(file_inode(file), file,
3061 					range, BTRFS_OLDEST_GENERATION, 0);
3062 		if (ret > 0)
3063 			ret = 0;
3064 		kfree(range);
3065 		break;
3066 	default:
3067 		ret = -EINVAL;
3068 	}
3069 out:
3070 	mnt_drop_write_file(file);
3071 	return ret;
3072 }
3073 
3074 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3075 {
3076 	struct btrfs_ioctl_vol_args *vol_args;
3077 	int ret;
3078 
3079 	if (!capable(CAP_SYS_ADMIN))
3080 		return -EPERM;
3081 
3082 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
3083 		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3084 
3085 	vol_args = memdup_user(arg, sizeof(*vol_args));
3086 	if (IS_ERR(vol_args)) {
3087 		ret = PTR_ERR(vol_args);
3088 		goto out;
3089 	}
3090 
3091 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3092 	ret = btrfs_init_new_device(fs_info, vol_args->name);
3093 
3094 	if (!ret)
3095 		btrfs_info(fs_info, "disk added %s", vol_args->name);
3096 
3097 	kfree(vol_args);
3098 out:
3099 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3100 	return ret;
3101 }
3102 
3103 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3104 {
3105 	struct inode *inode = file_inode(file);
3106 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3107 	struct btrfs_ioctl_vol_args_v2 *vol_args;
3108 	int ret;
3109 
3110 	if (!capable(CAP_SYS_ADMIN))
3111 		return -EPERM;
3112 
3113 	ret = mnt_want_write_file(file);
3114 	if (ret)
3115 		return ret;
3116 
3117 	vol_args = memdup_user(arg, sizeof(*vol_args));
3118 	if (IS_ERR(vol_args)) {
3119 		ret = PTR_ERR(vol_args);
3120 		goto err_drop;
3121 	}
3122 
3123 	if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3124 		ret = -EOPNOTSUPP;
3125 		goto out;
3126 	}
3127 
3128 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3129 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3130 		goto out;
3131 	}
3132 
3133 	if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3134 		ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3135 	} else {
3136 		vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3137 		ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3138 	}
3139 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3140 
3141 	if (!ret) {
3142 		if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3143 			btrfs_info(fs_info, "device deleted: id %llu",
3144 					vol_args->devid);
3145 		else
3146 			btrfs_info(fs_info, "device deleted: %s",
3147 					vol_args->name);
3148 	}
3149 out:
3150 	kfree(vol_args);
3151 err_drop:
3152 	mnt_drop_write_file(file);
3153 	return ret;
3154 }
3155 
3156 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3157 {
3158 	struct inode *inode = file_inode(file);
3159 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3160 	struct btrfs_ioctl_vol_args *vol_args;
3161 	int ret;
3162 
3163 	if (!capable(CAP_SYS_ADMIN))
3164 		return -EPERM;
3165 
3166 	ret = mnt_want_write_file(file);
3167 	if (ret)
3168 		return ret;
3169 
3170 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3171 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3172 		goto out_drop_write;
3173 	}
3174 
3175 	vol_args = memdup_user(arg, sizeof(*vol_args));
3176 	if (IS_ERR(vol_args)) {
3177 		ret = PTR_ERR(vol_args);
3178 		goto out;
3179 	}
3180 
3181 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3182 	ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3183 
3184 	if (!ret)
3185 		btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3186 	kfree(vol_args);
3187 out:
3188 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3189 out_drop_write:
3190 	mnt_drop_write_file(file);
3191 
3192 	return ret;
3193 }
3194 
3195 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3196 				void __user *arg)
3197 {
3198 	struct btrfs_ioctl_fs_info_args *fi_args;
3199 	struct btrfs_device *device;
3200 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3201 	int ret = 0;
3202 
3203 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3204 	if (!fi_args)
3205 		return -ENOMEM;
3206 
3207 	rcu_read_lock();
3208 	fi_args->num_devices = fs_devices->num_devices;
3209 
3210 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3211 		if (device->devid > fi_args->max_id)
3212 			fi_args->max_id = device->devid;
3213 	}
3214 	rcu_read_unlock();
3215 
3216 	memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3217 	fi_args->nodesize = fs_info->nodesize;
3218 	fi_args->sectorsize = fs_info->sectorsize;
3219 	fi_args->clone_alignment = fs_info->sectorsize;
3220 
3221 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3222 		ret = -EFAULT;
3223 
3224 	kfree(fi_args);
3225 	return ret;
3226 }
3227 
3228 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3229 				 void __user *arg)
3230 {
3231 	struct btrfs_ioctl_dev_info_args *di_args;
3232 	struct btrfs_device *dev;
3233 	int ret = 0;
3234 	char *s_uuid = NULL;
3235 
3236 	di_args = memdup_user(arg, sizeof(*di_args));
3237 	if (IS_ERR(di_args))
3238 		return PTR_ERR(di_args);
3239 
3240 	if (!btrfs_is_empty_uuid(di_args->uuid))
3241 		s_uuid = di_args->uuid;
3242 
3243 	rcu_read_lock();
3244 	dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3245 				NULL, true);
3246 
3247 	if (!dev) {
3248 		ret = -ENODEV;
3249 		goto out;
3250 	}
3251 
3252 	di_args->devid = dev->devid;
3253 	di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3254 	di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3255 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3256 	if (dev->name) {
3257 		strncpy(di_args->path, rcu_str_deref(dev->name),
3258 				sizeof(di_args->path) - 1);
3259 		di_args->path[sizeof(di_args->path) - 1] = 0;
3260 	} else {
3261 		di_args->path[0] = '\0';
3262 	}
3263 
3264 out:
3265 	rcu_read_unlock();
3266 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3267 		ret = -EFAULT;
3268 
3269 	kfree(di_args);
3270 	return ret;
3271 }
3272 
3273 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3274 {
3275 	struct inode *inode = file_inode(file);
3276 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3277 	struct btrfs_root *root = BTRFS_I(inode)->root;
3278 	struct btrfs_root *new_root;
3279 	struct btrfs_dir_item *di;
3280 	struct btrfs_trans_handle *trans;
3281 	struct btrfs_path *path = NULL;
3282 	struct btrfs_disk_key disk_key;
3283 	u64 objectid = 0;
3284 	u64 dir_id;
3285 	int ret;
3286 
3287 	if (!capable(CAP_SYS_ADMIN))
3288 		return -EPERM;
3289 
3290 	ret = mnt_want_write_file(file);
3291 	if (ret)
3292 		return ret;
3293 
3294 	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3295 		ret = -EFAULT;
3296 		goto out;
3297 	}
3298 
3299 	if (!objectid)
3300 		objectid = BTRFS_FS_TREE_OBJECTID;
3301 
3302 	new_root = btrfs_get_fs_root(fs_info, objectid, true);
3303 	if (IS_ERR(new_root)) {
3304 		ret = PTR_ERR(new_root);
3305 		goto out;
3306 	}
3307 	if (!is_fstree(new_root->root_key.objectid)) {
3308 		ret = -ENOENT;
3309 		goto out_free;
3310 	}
3311 
3312 	path = btrfs_alloc_path();
3313 	if (!path) {
3314 		ret = -ENOMEM;
3315 		goto out_free;
3316 	}
3317 	path->leave_spinning = 1;
3318 
3319 	trans = btrfs_start_transaction(root, 1);
3320 	if (IS_ERR(trans)) {
3321 		ret = PTR_ERR(trans);
3322 		goto out_free;
3323 	}
3324 
3325 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
3326 	di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3327 				   dir_id, "default", 7, 1);
3328 	if (IS_ERR_OR_NULL(di)) {
3329 		btrfs_release_path(path);
3330 		btrfs_end_transaction(trans);
3331 		btrfs_err(fs_info,
3332 			  "Umm, you don't have the default diritem, this isn't going to work");
3333 		ret = -ENOENT;
3334 		goto out_free;
3335 	}
3336 
3337 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3338 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3339 	btrfs_mark_buffer_dirty(path->nodes[0]);
3340 	btrfs_release_path(path);
3341 
3342 	btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3343 	btrfs_end_transaction(trans);
3344 out_free:
3345 	btrfs_put_root(new_root);
3346 	btrfs_free_path(path);
3347 out:
3348 	mnt_drop_write_file(file);
3349 	return ret;
3350 }
3351 
3352 static void get_block_group_info(struct list_head *groups_list,
3353 				 struct btrfs_ioctl_space_info *space)
3354 {
3355 	struct btrfs_block_group *block_group;
3356 
3357 	space->total_bytes = 0;
3358 	space->used_bytes = 0;
3359 	space->flags = 0;
3360 	list_for_each_entry(block_group, groups_list, list) {
3361 		space->flags = block_group->flags;
3362 		space->total_bytes += block_group->length;
3363 		space->used_bytes += block_group->used;
3364 	}
3365 }
3366 
3367 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3368 				   void __user *arg)
3369 {
3370 	struct btrfs_ioctl_space_args space_args;
3371 	struct btrfs_ioctl_space_info space;
3372 	struct btrfs_ioctl_space_info *dest;
3373 	struct btrfs_ioctl_space_info *dest_orig;
3374 	struct btrfs_ioctl_space_info __user *user_dest;
3375 	struct btrfs_space_info *info;
3376 	static const u64 types[] = {
3377 		BTRFS_BLOCK_GROUP_DATA,
3378 		BTRFS_BLOCK_GROUP_SYSTEM,
3379 		BTRFS_BLOCK_GROUP_METADATA,
3380 		BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3381 	};
3382 	int num_types = 4;
3383 	int alloc_size;
3384 	int ret = 0;
3385 	u64 slot_count = 0;
3386 	int i, c;
3387 
3388 	if (copy_from_user(&space_args,
3389 			   (struct btrfs_ioctl_space_args __user *)arg,
3390 			   sizeof(space_args)))
3391 		return -EFAULT;
3392 
3393 	for (i = 0; i < num_types; i++) {
3394 		struct btrfs_space_info *tmp;
3395 
3396 		info = NULL;
3397 		rcu_read_lock();
3398 		list_for_each_entry_rcu(tmp, &fs_info->space_info,
3399 					list) {
3400 			if (tmp->flags == types[i]) {
3401 				info = tmp;
3402 				break;
3403 			}
3404 		}
3405 		rcu_read_unlock();
3406 
3407 		if (!info)
3408 			continue;
3409 
3410 		down_read(&info->groups_sem);
3411 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3412 			if (!list_empty(&info->block_groups[c]))
3413 				slot_count++;
3414 		}
3415 		up_read(&info->groups_sem);
3416 	}
3417 
3418 	/*
3419 	 * Global block reserve, exported as a space_info
3420 	 */
3421 	slot_count++;
3422 
3423 	/* space_slots == 0 means they are asking for a count */
3424 	if (space_args.space_slots == 0) {
3425 		space_args.total_spaces = slot_count;
3426 		goto out;
3427 	}
3428 
3429 	slot_count = min_t(u64, space_args.space_slots, slot_count);
3430 
3431 	alloc_size = sizeof(*dest) * slot_count;
3432 
3433 	/* we generally have at most 6 or so space infos, one for each raid
3434 	 * level.  So, a whole page should be more than enough for everyone
3435 	 */
3436 	if (alloc_size > PAGE_SIZE)
3437 		return -ENOMEM;
3438 
3439 	space_args.total_spaces = 0;
3440 	dest = kmalloc(alloc_size, GFP_KERNEL);
3441 	if (!dest)
3442 		return -ENOMEM;
3443 	dest_orig = dest;
3444 
3445 	/* now we have a buffer to copy into */
3446 	for (i = 0; i < num_types; i++) {
3447 		struct btrfs_space_info *tmp;
3448 
3449 		if (!slot_count)
3450 			break;
3451 
3452 		info = NULL;
3453 		rcu_read_lock();
3454 		list_for_each_entry_rcu(tmp, &fs_info->space_info,
3455 					list) {
3456 			if (tmp->flags == types[i]) {
3457 				info = tmp;
3458 				break;
3459 			}
3460 		}
3461 		rcu_read_unlock();
3462 
3463 		if (!info)
3464 			continue;
3465 		down_read(&info->groups_sem);
3466 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3467 			if (!list_empty(&info->block_groups[c])) {
3468 				get_block_group_info(&info->block_groups[c],
3469 						     &space);
3470 				memcpy(dest, &space, sizeof(space));
3471 				dest++;
3472 				space_args.total_spaces++;
3473 				slot_count--;
3474 			}
3475 			if (!slot_count)
3476 				break;
3477 		}
3478 		up_read(&info->groups_sem);
3479 	}
3480 
3481 	/*
3482 	 * Add global block reserve
3483 	 */
3484 	if (slot_count) {
3485 		struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3486 
3487 		spin_lock(&block_rsv->lock);
3488 		space.total_bytes = block_rsv->size;
3489 		space.used_bytes = block_rsv->size - block_rsv->reserved;
3490 		spin_unlock(&block_rsv->lock);
3491 		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3492 		memcpy(dest, &space, sizeof(space));
3493 		space_args.total_spaces++;
3494 	}
3495 
3496 	user_dest = (struct btrfs_ioctl_space_info __user *)
3497 		(arg + sizeof(struct btrfs_ioctl_space_args));
3498 
3499 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3500 		ret = -EFAULT;
3501 
3502 	kfree(dest_orig);
3503 out:
3504 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3505 		ret = -EFAULT;
3506 
3507 	return ret;
3508 }
3509 
3510 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3511 					    void __user *argp)
3512 {
3513 	struct btrfs_trans_handle *trans;
3514 	u64 transid;
3515 	int ret;
3516 
3517 	trans = btrfs_attach_transaction_barrier(root);
3518 	if (IS_ERR(trans)) {
3519 		if (PTR_ERR(trans) != -ENOENT)
3520 			return PTR_ERR(trans);
3521 
3522 		/* No running transaction, don't bother */
3523 		transid = root->fs_info->last_trans_committed;
3524 		goto out;
3525 	}
3526 	transid = trans->transid;
3527 	ret = btrfs_commit_transaction_async(trans, 0);
3528 	if (ret) {
3529 		btrfs_end_transaction(trans);
3530 		return ret;
3531 	}
3532 out:
3533 	if (argp)
3534 		if (copy_to_user(argp, &transid, sizeof(transid)))
3535 			return -EFAULT;
3536 	return 0;
3537 }
3538 
3539 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3540 					   void __user *argp)
3541 {
3542 	u64 transid;
3543 
3544 	if (argp) {
3545 		if (copy_from_user(&transid, argp, sizeof(transid)))
3546 			return -EFAULT;
3547 	} else {
3548 		transid = 0;  /* current trans */
3549 	}
3550 	return btrfs_wait_for_commit(fs_info, transid);
3551 }
3552 
3553 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3554 {
3555 	struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3556 	struct btrfs_ioctl_scrub_args *sa;
3557 	int ret;
3558 
3559 	if (!capable(CAP_SYS_ADMIN))
3560 		return -EPERM;
3561 
3562 	sa = memdup_user(arg, sizeof(*sa));
3563 	if (IS_ERR(sa))
3564 		return PTR_ERR(sa);
3565 
3566 	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3567 		ret = mnt_want_write_file(file);
3568 		if (ret)
3569 			goto out;
3570 	}
3571 
3572 	ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3573 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3574 			      0);
3575 
3576 	/*
3577 	 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3578 	 * error. This is important as it allows user space to know how much
3579 	 * progress scrub has done. For example, if scrub is canceled we get
3580 	 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3581 	 * space. Later user space can inspect the progress from the structure
3582 	 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3583 	 * previously (btrfs-progs does this).
3584 	 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3585 	 * then return -EFAULT to signal the structure was not copied or it may
3586 	 * be corrupt and unreliable due to a partial copy.
3587 	 */
3588 	if (copy_to_user(arg, sa, sizeof(*sa)))
3589 		ret = -EFAULT;
3590 
3591 	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3592 		mnt_drop_write_file(file);
3593 out:
3594 	kfree(sa);
3595 	return ret;
3596 }
3597 
3598 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3599 {
3600 	if (!capable(CAP_SYS_ADMIN))
3601 		return -EPERM;
3602 
3603 	return btrfs_scrub_cancel(fs_info);
3604 }
3605 
3606 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3607 				       void __user *arg)
3608 {
3609 	struct btrfs_ioctl_scrub_args *sa;
3610 	int ret;
3611 
3612 	if (!capable(CAP_SYS_ADMIN))
3613 		return -EPERM;
3614 
3615 	sa = memdup_user(arg, sizeof(*sa));
3616 	if (IS_ERR(sa))
3617 		return PTR_ERR(sa);
3618 
3619 	ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3620 
3621 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3622 		ret = -EFAULT;
3623 
3624 	kfree(sa);
3625 	return ret;
3626 }
3627 
3628 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3629 				      void __user *arg)
3630 {
3631 	struct btrfs_ioctl_get_dev_stats *sa;
3632 	int ret;
3633 
3634 	sa = memdup_user(arg, sizeof(*sa));
3635 	if (IS_ERR(sa))
3636 		return PTR_ERR(sa);
3637 
3638 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3639 		kfree(sa);
3640 		return -EPERM;
3641 	}
3642 
3643 	ret = btrfs_get_dev_stats(fs_info, sa);
3644 
3645 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3646 		ret = -EFAULT;
3647 
3648 	kfree(sa);
3649 	return ret;
3650 }
3651 
3652 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3653 				    void __user *arg)
3654 {
3655 	struct btrfs_ioctl_dev_replace_args *p;
3656 	int ret;
3657 
3658 	if (!capable(CAP_SYS_ADMIN))
3659 		return -EPERM;
3660 
3661 	p = memdup_user(arg, sizeof(*p));
3662 	if (IS_ERR(p))
3663 		return PTR_ERR(p);
3664 
3665 	switch (p->cmd) {
3666 	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3667 		if (sb_rdonly(fs_info->sb)) {
3668 			ret = -EROFS;
3669 			goto out;
3670 		}
3671 		if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3672 			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3673 		} else {
3674 			ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3675 			clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3676 		}
3677 		break;
3678 	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3679 		btrfs_dev_replace_status(fs_info, p);
3680 		ret = 0;
3681 		break;
3682 	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3683 		p->result = btrfs_dev_replace_cancel(fs_info);
3684 		ret = 0;
3685 		break;
3686 	default:
3687 		ret = -EINVAL;
3688 		break;
3689 	}
3690 
3691 	if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3692 		ret = -EFAULT;
3693 out:
3694 	kfree(p);
3695 	return ret;
3696 }
3697 
3698 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3699 {
3700 	int ret = 0;
3701 	int i;
3702 	u64 rel_ptr;
3703 	int size;
3704 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3705 	struct inode_fs_paths *ipath = NULL;
3706 	struct btrfs_path *path;
3707 
3708 	if (!capable(CAP_DAC_READ_SEARCH))
3709 		return -EPERM;
3710 
3711 	path = btrfs_alloc_path();
3712 	if (!path) {
3713 		ret = -ENOMEM;
3714 		goto out;
3715 	}
3716 
3717 	ipa = memdup_user(arg, sizeof(*ipa));
3718 	if (IS_ERR(ipa)) {
3719 		ret = PTR_ERR(ipa);
3720 		ipa = NULL;
3721 		goto out;
3722 	}
3723 
3724 	size = min_t(u32, ipa->size, 4096);
3725 	ipath = init_ipath(size, root, path);
3726 	if (IS_ERR(ipath)) {
3727 		ret = PTR_ERR(ipath);
3728 		ipath = NULL;
3729 		goto out;
3730 	}
3731 
3732 	ret = paths_from_inode(ipa->inum, ipath);
3733 	if (ret < 0)
3734 		goto out;
3735 
3736 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3737 		rel_ptr = ipath->fspath->val[i] -
3738 			  (u64)(unsigned long)ipath->fspath->val;
3739 		ipath->fspath->val[i] = rel_ptr;
3740 	}
3741 
3742 	ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3743 			   ipath->fspath, size);
3744 	if (ret) {
3745 		ret = -EFAULT;
3746 		goto out;
3747 	}
3748 
3749 out:
3750 	btrfs_free_path(path);
3751 	free_ipath(ipath);
3752 	kfree(ipa);
3753 
3754 	return ret;
3755 }
3756 
3757 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3758 {
3759 	struct btrfs_data_container *inodes = ctx;
3760 	const size_t c = 3 * sizeof(u64);
3761 
3762 	if (inodes->bytes_left >= c) {
3763 		inodes->bytes_left -= c;
3764 		inodes->val[inodes->elem_cnt] = inum;
3765 		inodes->val[inodes->elem_cnt + 1] = offset;
3766 		inodes->val[inodes->elem_cnt + 2] = root;
3767 		inodes->elem_cnt += 3;
3768 	} else {
3769 		inodes->bytes_missing += c - inodes->bytes_left;
3770 		inodes->bytes_left = 0;
3771 		inodes->elem_missed += 3;
3772 	}
3773 
3774 	return 0;
3775 }
3776 
3777 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3778 					void __user *arg, int version)
3779 {
3780 	int ret = 0;
3781 	int size;
3782 	struct btrfs_ioctl_logical_ino_args *loi;
3783 	struct btrfs_data_container *inodes = NULL;
3784 	struct btrfs_path *path = NULL;
3785 	bool ignore_offset;
3786 
3787 	if (!capable(CAP_SYS_ADMIN))
3788 		return -EPERM;
3789 
3790 	loi = memdup_user(arg, sizeof(*loi));
3791 	if (IS_ERR(loi))
3792 		return PTR_ERR(loi);
3793 
3794 	if (version == 1) {
3795 		ignore_offset = false;
3796 		size = min_t(u32, loi->size, SZ_64K);
3797 	} else {
3798 		/* All reserved bits must be 0 for now */
3799 		if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3800 			ret = -EINVAL;
3801 			goto out_loi;
3802 		}
3803 		/* Only accept flags we have defined so far */
3804 		if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3805 			ret = -EINVAL;
3806 			goto out_loi;
3807 		}
3808 		ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3809 		size = min_t(u32, loi->size, SZ_16M);
3810 	}
3811 
3812 	path = btrfs_alloc_path();
3813 	if (!path) {
3814 		ret = -ENOMEM;
3815 		goto out;
3816 	}
3817 
3818 	inodes = init_data_container(size);
3819 	if (IS_ERR(inodes)) {
3820 		ret = PTR_ERR(inodes);
3821 		inodes = NULL;
3822 		goto out;
3823 	}
3824 
3825 	ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3826 					  build_ino_list, inodes, ignore_offset);
3827 	if (ret == -EINVAL)
3828 		ret = -ENOENT;
3829 	if (ret < 0)
3830 		goto out;
3831 
3832 	ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3833 			   size);
3834 	if (ret)
3835 		ret = -EFAULT;
3836 
3837 out:
3838 	btrfs_free_path(path);
3839 	kvfree(inodes);
3840 out_loi:
3841 	kfree(loi);
3842 
3843 	return ret;
3844 }
3845 
3846 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3847 			       struct btrfs_ioctl_balance_args *bargs)
3848 {
3849 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3850 
3851 	bargs->flags = bctl->flags;
3852 
3853 	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3854 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3855 	if (atomic_read(&fs_info->balance_pause_req))
3856 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3857 	if (atomic_read(&fs_info->balance_cancel_req))
3858 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3859 
3860 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3861 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3862 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3863 
3864 	spin_lock(&fs_info->balance_lock);
3865 	memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3866 	spin_unlock(&fs_info->balance_lock);
3867 }
3868 
3869 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3870 {
3871 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3872 	struct btrfs_fs_info *fs_info = root->fs_info;
3873 	struct btrfs_ioctl_balance_args *bargs;
3874 	struct btrfs_balance_control *bctl;
3875 	bool need_unlock; /* for mut. excl. ops lock */
3876 	int ret;
3877 
3878 	if (!capable(CAP_SYS_ADMIN))
3879 		return -EPERM;
3880 
3881 	ret = mnt_want_write_file(file);
3882 	if (ret)
3883 		return ret;
3884 
3885 again:
3886 	if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3887 		mutex_lock(&fs_info->balance_mutex);
3888 		need_unlock = true;
3889 		goto locked;
3890 	}
3891 
3892 	/*
3893 	 * mut. excl. ops lock is locked.  Three possibilities:
3894 	 *   (1) some other op is running
3895 	 *   (2) balance is running
3896 	 *   (3) balance is paused -- special case (think resume)
3897 	 */
3898 	mutex_lock(&fs_info->balance_mutex);
3899 	if (fs_info->balance_ctl) {
3900 		/* this is either (2) or (3) */
3901 		if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3902 			mutex_unlock(&fs_info->balance_mutex);
3903 			/*
3904 			 * Lock released to allow other waiters to continue,
3905 			 * we'll reexamine the status again.
3906 			 */
3907 			mutex_lock(&fs_info->balance_mutex);
3908 
3909 			if (fs_info->balance_ctl &&
3910 			    !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3911 				/* this is (3) */
3912 				need_unlock = false;
3913 				goto locked;
3914 			}
3915 
3916 			mutex_unlock(&fs_info->balance_mutex);
3917 			goto again;
3918 		} else {
3919 			/* this is (2) */
3920 			mutex_unlock(&fs_info->balance_mutex);
3921 			ret = -EINPROGRESS;
3922 			goto out;
3923 		}
3924 	} else {
3925 		/* this is (1) */
3926 		mutex_unlock(&fs_info->balance_mutex);
3927 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3928 		goto out;
3929 	}
3930 
3931 locked:
3932 	BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
3933 
3934 	if (arg) {
3935 		bargs = memdup_user(arg, sizeof(*bargs));
3936 		if (IS_ERR(bargs)) {
3937 			ret = PTR_ERR(bargs);
3938 			goto out_unlock;
3939 		}
3940 
3941 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3942 			if (!fs_info->balance_ctl) {
3943 				ret = -ENOTCONN;
3944 				goto out_bargs;
3945 			}
3946 
3947 			bctl = fs_info->balance_ctl;
3948 			spin_lock(&fs_info->balance_lock);
3949 			bctl->flags |= BTRFS_BALANCE_RESUME;
3950 			spin_unlock(&fs_info->balance_lock);
3951 
3952 			goto do_balance;
3953 		}
3954 	} else {
3955 		bargs = NULL;
3956 	}
3957 
3958 	if (fs_info->balance_ctl) {
3959 		ret = -EINPROGRESS;
3960 		goto out_bargs;
3961 	}
3962 
3963 	bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3964 	if (!bctl) {
3965 		ret = -ENOMEM;
3966 		goto out_bargs;
3967 	}
3968 
3969 	if (arg) {
3970 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3971 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3972 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3973 
3974 		bctl->flags = bargs->flags;
3975 	} else {
3976 		/* balance everything - no filters */
3977 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3978 	}
3979 
3980 	if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3981 		ret = -EINVAL;
3982 		goto out_bctl;
3983 	}
3984 
3985 do_balance:
3986 	/*
3987 	 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
3988 	 * btrfs_balance.  bctl is freed in reset_balance_state, or, if
3989 	 * restriper was paused all the way until unmount, in free_fs_info.
3990 	 * The flag should be cleared after reset_balance_state.
3991 	 */
3992 	need_unlock = false;
3993 
3994 	ret = btrfs_balance(fs_info, bctl, bargs);
3995 	bctl = NULL;
3996 
3997 	if ((ret == 0 || ret == -ECANCELED) && arg) {
3998 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3999 			ret = -EFAULT;
4000 	}
4001 
4002 out_bctl:
4003 	kfree(bctl);
4004 out_bargs:
4005 	kfree(bargs);
4006 out_unlock:
4007 	mutex_unlock(&fs_info->balance_mutex);
4008 	if (need_unlock)
4009 		clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4010 out:
4011 	mnt_drop_write_file(file);
4012 	return ret;
4013 }
4014 
4015 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4016 {
4017 	if (!capable(CAP_SYS_ADMIN))
4018 		return -EPERM;
4019 
4020 	switch (cmd) {
4021 	case BTRFS_BALANCE_CTL_PAUSE:
4022 		return btrfs_pause_balance(fs_info);
4023 	case BTRFS_BALANCE_CTL_CANCEL:
4024 		return btrfs_cancel_balance(fs_info);
4025 	}
4026 
4027 	return -EINVAL;
4028 }
4029 
4030 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4031 					 void __user *arg)
4032 {
4033 	struct btrfs_ioctl_balance_args *bargs;
4034 	int ret = 0;
4035 
4036 	if (!capable(CAP_SYS_ADMIN))
4037 		return -EPERM;
4038 
4039 	mutex_lock(&fs_info->balance_mutex);
4040 	if (!fs_info->balance_ctl) {
4041 		ret = -ENOTCONN;
4042 		goto out;
4043 	}
4044 
4045 	bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4046 	if (!bargs) {
4047 		ret = -ENOMEM;
4048 		goto out;
4049 	}
4050 
4051 	btrfs_update_ioctl_balance_args(fs_info, bargs);
4052 
4053 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
4054 		ret = -EFAULT;
4055 
4056 	kfree(bargs);
4057 out:
4058 	mutex_unlock(&fs_info->balance_mutex);
4059 	return ret;
4060 }
4061 
4062 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4063 {
4064 	struct inode *inode = file_inode(file);
4065 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4066 	struct btrfs_ioctl_quota_ctl_args *sa;
4067 	int ret;
4068 
4069 	if (!capable(CAP_SYS_ADMIN))
4070 		return -EPERM;
4071 
4072 	ret = mnt_want_write_file(file);
4073 	if (ret)
4074 		return ret;
4075 
4076 	sa = memdup_user(arg, sizeof(*sa));
4077 	if (IS_ERR(sa)) {
4078 		ret = PTR_ERR(sa);
4079 		goto drop_write;
4080 	}
4081 
4082 	down_write(&fs_info->subvol_sem);
4083 
4084 	switch (sa->cmd) {
4085 	case BTRFS_QUOTA_CTL_ENABLE:
4086 		ret = btrfs_quota_enable(fs_info);
4087 		break;
4088 	case BTRFS_QUOTA_CTL_DISABLE:
4089 		ret = btrfs_quota_disable(fs_info);
4090 		break;
4091 	default:
4092 		ret = -EINVAL;
4093 		break;
4094 	}
4095 
4096 	kfree(sa);
4097 	up_write(&fs_info->subvol_sem);
4098 drop_write:
4099 	mnt_drop_write_file(file);
4100 	return ret;
4101 }
4102 
4103 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4104 {
4105 	struct inode *inode = file_inode(file);
4106 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4107 	struct btrfs_root *root = BTRFS_I(inode)->root;
4108 	struct btrfs_ioctl_qgroup_assign_args *sa;
4109 	struct btrfs_trans_handle *trans;
4110 	int ret;
4111 	int err;
4112 
4113 	if (!capable(CAP_SYS_ADMIN))
4114 		return -EPERM;
4115 
4116 	ret = mnt_want_write_file(file);
4117 	if (ret)
4118 		return ret;
4119 
4120 	sa = memdup_user(arg, sizeof(*sa));
4121 	if (IS_ERR(sa)) {
4122 		ret = PTR_ERR(sa);
4123 		goto drop_write;
4124 	}
4125 
4126 	trans = btrfs_join_transaction(root);
4127 	if (IS_ERR(trans)) {
4128 		ret = PTR_ERR(trans);
4129 		goto out;
4130 	}
4131 
4132 	if (sa->assign) {
4133 		ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4134 	} else {
4135 		ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4136 	}
4137 
4138 	/* update qgroup status and info */
4139 	err = btrfs_run_qgroups(trans);
4140 	if (err < 0)
4141 		btrfs_handle_fs_error(fs_info, err,
4142 				      "failed to update qgroup status and info");
4143 	err = btrfs_end_transaction(trans);
4144 	if (err && !ret)
4145 		ret = err;
4146 
4147 out:
4148 	kfree(sa);
4149 drop_write:
4150 	mnt_drop_write_file(file);
4151 	return ret;
4152 }
4153 
4154 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4155 {
4156 	struct inode *inode = file_inode(file);
4157 	struct btrfs_root *root = BTRFS_I(inode)->root;
4158 	struct btrfs_ioctl_qgroup_create_args *sa;
4159 	struct btrfs_trans_handle *trans;
4160 	int ret;
4161 	int err;
4162 
4163 	if (!capable(CAP_SYS_ADMIN))
4164 		return -EPERM;
4165 
4166 	ret = mnt_want_write_file(file);
4167 	if (ret)
4168 		return ret;
4169 
4170 	sa = memdup_user(arg, sizeof(*sa));
4171 	if (IS_ERR(sa)) {
4172 		ret = PTR_ERR(sa);
4173 		goto drop_write;
4174 	}
4175 
4176 	if (!sa->qgroupid) {
4177 		ret = -EINVAL;
4178 		goto out;
4179 	}
4180 
4181 	trans = btrfs_join_transaction(root);
4182 	if (IS_ERR(trans)) {
4183 		ret = PTR_ERR(trans);
4184 		goto out;
4185 	}
4186 
4187 	if (sa->create) {
4188 		ret = btrfs_create_qgroup(trans, sa->qgroupid);
4189 	} else {
4190 		ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4191 	}
4192 
4193 	err = btrfs_end_transaction(trans);
4194 	if (err && !ret)
4195 		ret = err;
4196 
4197 out:
4198 	kfree(sa);
4199 drop_write:
4200 	mnt_drop_write_file(file);
4201 	return ret;
4202 }
4203 
4204 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4205 {
4206 	struct inode *inode = file_inode(file);
4207 	struct btrfs_root *root = BTRFS_I(inode)->root;
4208 	struct btrfs_ioctl_qgroup_limit_args *sa;
4209 	struct btrfs_trans_handle *trans;
4210 	int ret;
4211 	int err;
4212 	u64 qgroupid;
4213 
4214 	if (!capable(CAP_SYS_ADMIN))
4215 		return -EPERM;
4216 
4217 	ret = mnt_want_write_file(file);
4218 	if (ret)
4219 		return ret;
4220 
4221 	sa = memdup_user(arg, sizeof(*sa));
4222 	if (IS_ERR(sa)) {
4223 		ret = PTR_ERR(sa);
4224 		goto drop_write;
4225 	}
4226 
4227 	trans = btrfs_join_transaction(root);
4228 	if (IS_ERR(trans)) {
4229 		ret = PTR_ERR(trans);
4230 		goto out;
4231 	}
4232 
4233 	qgroupid = sa->qgroupid;
4234 	if (!qgroupid) {
4235 		/* take the current subvol as qgroup */
4236 		qgroupid = root->root_key.objectid;
4237 	}
4238 
4239 	ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4240 
4241 	err = btrfs_end_transaction(trans);
4242 	if (err && !ret)
4243 		ret = err;
4244 
4245 out:
4246 	kfree(sa);
4247 drop_write:
4248 	mnt_drop_write_file(file);
4249 	return ret;
4250 }
4251 
4252 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4253 {
4254 	struct inode *inode = file_inode(file);
4255 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4256 	struct btrfs_ioctl_quota_rescan_args *qsa;
4257 	int ret;
4258 
4259 	if (!capable(CAP_SYS_ADMIN))
4260 		return -EPERM;
4261 
4262 	ret = mnt_want_write_file(file);
4263 	if (ret)
4264 		return ret;
4265 
4266 	qsa = memdup_user(arg, sizeof(*qsa));
4267 	if (IS_ERR(qsa)) {
4268 		ret = PTR_ERR(qsa);
4269 		goto drop_write;
4270 	}
4271 
4272 	if (qsa->flags) {
4273 		ret = -EINVAL;
4274 		goto out;
4275 	}
4276 
4277 	ret = btrfs_qgroup_rescan(fs_info);
4278 
4279 out:
4280 	kfree(qsa);
4281 drop_write:
4282 	mnt_drop_write_file(file);
4283 	return ret;
4284 }
4285 
4286 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4287 						void __user *arg)
4288 {
4289 	struct btrfs_ioctl_quota_rescan_args *qsa;
4290 	int ret = 0;
4291 
4292 	if (!capable(CAP_SYS_ADMIN))
4293 		return -EPERM;
4294 
4295 	qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4296 	if (!qsa)
4297 		return -ENOMEM;
4298 
4299 	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4300 		qsa->flags = 1;
4301 		qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4302 	}
4303 
4304 	if (copy_to_user(arg, qsa, sizeof(*qsa)))
4305 		ret = -EFAULT;
4306 
4307 	kfree(qsa);
4308 	return ret;
4309 }
4310 
4311 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4312 						void __user *arg)
4313 {
4314 	if (!capable(CAP_SYS_ADMIN))
4315 		return -EPERM;
4316 
4317 	return btrfs_qgroup_wait_for_completion(fs_info, true);
4318 }
4319 
4320 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4321 					    struct btrfs_ioctl_received_subvol_args *sa)
4322 {
4323 	struct inode *inode = file_inode(file);
4324 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4325 	struct btrfs_root *root = BTRFS_I(inode)->root;
4326 	struct btrfs_root_item *root_item = &root->root_item;
4327 	struct btrfs_trans_handle *trans;
4328 	struct timespec64 ct = current_time(inode);
4329 	int ret = 0;
4330 	int received_uuid_changed;
4331 
4332 	if (!inode_owner_or_capable(inode))
4333 		return -EPERM;
4334 
4335 	ret = mnt_want_write_file(file);
4336 	if (ret < 0)
4337 		return ret;
4338 
4339 	down_write(&fs_info->subvol_sem);
4340 
4341 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4342 		ret = -EINVAL;
4343 		goto out;
4344 	}
4345 
4346 	if (btrfs_root_readonly(root)) {
4347 		ret = -EROFS;
4348 		goto out;
4349 	}
4350 
4351 	/*
4352 	 * 1 - root item
4353 	 * 2 - uuid items (received uuid + subvol uuid)
4354 	 */
4355 	trans = btrfs_start_transaction(root, 3);
4356 	if (IS_ERR(trans)) {
4357 		ret = PTR_ERR(trans);
4358 		trans = NULL;
4359 		goto out;
4360 	}
4361 
4362 	sa->rtransid = trans->transid;
4363 	sa->rtime.sec = ct.tv_sec;
4364 	sa->rtime.nsec = ct.tv_nsec;
4365 
4366 	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4367 				       BTRFS_UUID_SIZE);
4368 	if (received_uuid_changed &&
4369 	    !btrfs_is_empty_uuid(root_item->received_uuid)) {
4370 		ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4371 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4372 					  root->root_key.objectid);
4373 		if (ret && ret != -ENOENT) {
4374 		        btrfs_abort_transaction(trans, ret);
4375 		        btrfs_end_transaction(trans);
4376 		        goto out;
4377 		}
4378 	}
4379 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4380 	btrfs_set_root_stransid(root_item, sa->stransid);
4381 	btrfs_set_root_rtransid(root_item, sa->rtransid);
4382 	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4383 	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4384 	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4385 	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4386 
4387 	ret = btrfs_update_root(trans, fs_info->tree_root,
4388 				&root->root_key, &root->root_item);
4389 	if (ret < 0) {
4390 		btrfs_end_transaction(trans);
4391 		goto out;
4392 	}
4393 	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4394 		ret = btrfs_uuid_tree_add(trans, sa->uuid,
4395 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4396 					  root->root_key.objectid);
4397 		if (ret < 0 && ret != -EEXIST) {
4398 			btrfs_abort_transaction(trans, ret);
4399 			btrfs_end_transaction(trans);
4400 			goto out;
4401 		}
4402 	}
4403 	ret = btrfs_commit_transaction(trans);
4404 out:
4405 	up_write(&fs_info->subvol_sem);
4406 	mnt_drop_write_file(file);
4407 	return ret;
4408 }
4409 
4410 #ifdef CONFIG_64BIT
4411 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4412 						void __user *arg)
4413 {
4414 	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4415 	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4416 	int ret = 0;
4417 
4418 	args32 = memdup_user(arg, sizeof(*args32));
4419 	if (IS_ERR(args32))
4420 		return PTR_ERR(args32);
4421 
4422 	args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4423 	if (!args64) {
4424 		ret = -ENOMEM;
4425 		goto out;
4426 	}
4427 
4428 	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4429 	args64->stransid = args32->stransid;
4430 	args64->rtransid = args32->rtransid;
4431 	args64->stime.sec = args32->stime.sec;
4432 	args64->stime.nsec = args32->stime.nsec;
4433 	args64->rtime.sec = args32->rtime.sec;
4434 	args64->rtime.nsec = args32->rtime.nsec;
4435 	args64->flags = args32->flags;
4436 
4437 	ret = _btrfs_ioctl_set_received_subvol(file, args64);
4438 	if (ret)
4439 		goto out;
4440 
4441 	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4442 	args32->stransid = args64->stransid;
4443 	args32->rtransid = args64->rtransid;
4444 	args32->stime.sec = args64->stime.sec;
4445 	args32->stime.nsec = args64->stime.nsec;
4446 	args32->rtime.sec = args64->rtime.sec;
4447 	args32->rtime.nsec = args64->rtime.nsec;
4448 	args32->flags = args64->flags;
4449 
4450 	ret = copy_to_user(arg, args32, sizeof(*args32));
4451 	if (ret)
4452 		ret = -EFAULT;
4453 
4454 out:
4455 	kfree(args32);
4456 	kfree(args64);
4457 	return ret;
4458 }
4459 #endif
4460 
4461 static long btrfs_ioctl_set_received_subvol(struct file *file,
4462 					    void __user *arg)
4463 {
4464 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4465 	int ret = 0;
4466 
4467 	sa = memdup_user(arg, sizeof(*sa));
4468 	if (IS_ERR(sa))
4469 		return PTR_ERR(sa);
4470 
4471 	ret = _btrfs_ioctl_set_received_subvol(file, sa);
4472 
4473 	if (ret)
4474 		goto out;
4475 
4476 	ret = copy_to_user(arg, sa, sizeof(*sa));
4477 	if (ret)
4478 		ret = -EFAULT;
4479 
4480 out:
4481 	kfree(sa);
4482 	return ret;
4483 }
4484 
4485 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4486 					void __user *arg)
4487 {
4488 	size_t len;
4489 	int ret;
4490 	char label[BTRFS_LABEL_SIZE];
4491 
4492 	spin_lock(&fs_info->super_lock);
4493 	memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4494 	spin_unlock(&fs_info->super_lock);
4495 
4496 	len = strnlen(label, BTRFS_LABEL_SIZE);
4497 
4498 	if (len == BTRFS_LABEL_SIZE) {
4499 		btrfs_warn(fs_info,
4500 			   "label is too long, return the first %zu bytes",
4501 			   --len);
4502 	}
4503 
4504 	ret = copy_to_user(arg, label, len);
4505 
4506 	return ret ? -EFAULT : 0;
4507 }
4508 
4509 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4510 {
4511 	struct inode *inode = file_inode(file);
4512 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4513 	struct btrfs_root *root = BTRFS_I(inode)->root;
4514 	struct btrfs_super_block *super_block = fs_info->super_copy;
4515 	struct btrfs_trans_handle *trans;
4516 	char label[BTRFS_LABEL_SIZE];
4517 	int ret;
4518 
4519 	if (!capable(CAP_SYS_ADMIN))
4520 		return -EPERM;
4521 
4522 	if (copy_from_user(label, arg, sizeof(label)))
4523 		return -EFAULT;
4524 
4525 	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4526 		btrfs_err(fs_info,
4527 			  "unable to set label with more than %d bytes",
4528 			  BTRFS_LABEL_SIZE - 1);
4529 		return -EINVAL;
4530 	}
4531 
4532 	ret = mnt_want_write_file(file);
4533 	if (ret)
4534 		return ret;
4535 
4536 	trans = btrfs_start_transaction(root, 0);
4537 	if (IS_ERR(trans)) {
4538 		ret = PTR_ERR(trans);
4539 		goto out_unlock;
4540 	}
4541 
4542 	spin_lock(&fs_info->super_lock);
4543 	strcpy(super_block->label, label);
4544 	spin_unlock(&fs_info->super_lock);
4545 	ret = btrfs_commit_transaction(trans);
4546 
4547 out_unlock:
4548 	mnt_drop_write_file(file);
4549 	return ret;
4550 }
4551 
4552 #define INIT_FEATURE_FLAGS(suffix) \
4553 	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4554 	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4555 	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4556 
4557 int btrfs_ioctl_get_supported_features(void __user *arg)
4558 {
4559 	static const struct btrfs_ioctl_feature_flags features[3] = {
4560 		INIT_FEATURE_FLAGS(SUPP),
4561 		INIT_FEATURE_FLAGS(SAFE_SET),
4562 		INIT_FEATURE_FLAGS(SAFE_CLEAR)
4563 	};
4564 
4565 	if (copy_to_user(arg, &features, sizeof(features)))
4566 		return -EFAULT;
4567 
4568 	return 0;
4569 }
4570 
4571 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4572 					void __user *arg)
4573 {
4574 	struct btrfs_super_block *super_block = fs_info->super_copy;
4575 	struct btrfs_ioctl_feature_flags features;
4576 
4577 	features.compat_flags = btrfs_super_compat_flags(super_block);
4578 	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4579 	features.incompat_flags = btrfs_super_incompat_flags(super_block);
4580 
4581 	if (copy_to_user(arg, &features, sizeof(features)))
4582 		return -EFAULT;
4583 
4584 	return 0;
4585 }
4586 
4587 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4588 			      enum btrfs_feature_set set,
4589 			      u64 change_mask, u64 flags, u64 supported_flags,
4590 			      u64 safe_set, u64 safe_clear)
4591 {
4592 	const char *type = btrfs_feature_set_name(set);
4593 	char *names;
4594 	u64 disallowed, unsupported;
4595 	u64 set_mask = flags & change_mask;
4596 	u64 clear_mask = ~flags & change_mask;
4597 
4598 	unsupported = set_mask & ~supported_flags;
4599 	if (unsupported) {
4600 		names = btrfs_printable_features(set, unsupported);
4601 		if (names) {
4602 			btrfs_warn(fs_info,
4603 				   "this kernel does not support the %s feature bit%s",
4604 				   names, strchr(names, ',') ? "s" : "");
4605 			kfree(names);
4606 		} else
4607 			btrfs_warn(fs_info,
4608 				   "this kernel does not support %s bits 0x%llx",
4609 				   type, unsupported);
4610 		return -EOPNOTSUPP;
4611 	}
4612 
4613 	disallowed = set_mask & ~safe_set;
4614 	if (disallowed) {
4615 		names = btrfs_printable_features(set, disallowed);
4616 		if (names) {
4617 			btrfs_warn(fs_info,
4618 				   "can't set the %s feature bit%s while mounted",
4619 				   names, strchr(names, ',') ? "s" : "");
4620 			kfree(names);
4621 		} else
4622 			btrfs_warn(fs_info,
4623 				   "can't set %s bits 0x%llx while mounted",
4624 				   type, disallowed);
4625 		return -EPERM;
4626 	}
4627 
4628 	disallowed = clear_mask & ~safe_clear;
4629 	if (disallowed) {
4630 		names = btrfs_printable_features(set, disallowed);
4631 		if (names) {
4632 			btrfs_warn(fs_info,
4633 				   "can't clear the %s feature bit%s while mounted",
4634 				   names, strchr(names, ',') ? "s" : "");
4635 			kfree(names);
4636 		} else
4637 			btrfs_warn(fs_info,
4638 				   "can't clear %s bits 0x%llx while mounted",
4639 				   type, disallowed);
4640 		return -EPERM;
4641 	}
4642 
4643 	return 0;
4644 }
4645 
4646 #define check_feature(fs_info, change_mask, flags, mask_base)	\
4647 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,	\
4648 		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
4649 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
4650 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4651 
4652 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4653 {
4654 	struct inode *inode = file_inode(file);
4655 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4656 	struct btrfs_root *root = BTRFS_I(inode)->root;
4657 	struct btrfs_super_block *super_block = fs_info->super_copy;
4658 	struct btrfs_ioctl_feature_flags flags[2];
4659 	struct btrfs_trans_handle *trans;
4660 	u64 newflags;
4661 	int ret;
4662 
4663 	if (!capable(CAP_SYS_ADMIN))
4664 		return -EPERM;
4665 
4666 	if (copy_from_user(flags, arg, sizeof(flags)))
4667 		return -EFAULT;
4668 
4669 	/* Nothing to do */
4670 	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4671 	    !flags[0].incompat_flags)
4672 		return 0;
4673 
4674 	ret = check_feature(fs_info, flags[0].compat_flags,
4675 			    flags[1].compat_flags, COMPAT);
4676 	if (ret)
4677 		return ret;
4678 
4679 	ret = check_feature(fs_info, flags[0].compat_ro_flags,
4680 			    flags[1].compat_ro_flags, COMPAT_RO);
4681 	if (ret)
4682 		return ret;
4683 
4684 	ret = check_feature(fs_info, flags[0].incompat_flags,
4685 			    flags[1].incompat_flags, INCOMPAT);
4686 	if (ret)
4687 		return ret;
4688 
4689 	ret = mnt_want_write_file(file);
4690 	if (ret)
4691 		return ret;
4692 
4693 	trans = btrfs_start_transaction(root, 0);
4694 	if (IS_ERR(trans)) {
4695 		ret = PTR_ERR(trans);
4696 		goto out_drop_write;
4697 	}
4698 
4699 	spin_lock(&fs_info->super_lock);
4700 	newflags = btrfs_super_compat_flags(super_block);
4701 	newflags |= flags[0].compat_flags & flags[1].compat_flags;
4702 	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4703 	btrfs_set_super_compat_flags(super_block, newflags);
4704 
4705 	newflags = btrfs_super_compat_ro_flags(super_block);
4706 	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4707 	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4708 	btrfs_set_super_compat_ro_flags(super_block, newflags);
4709 
4710 	newflags = btrfs_super_incompat_flags(super_block);
4711 	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4712 	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4713 	btrfs_set_super_incompat_flags(super_block, newflags);
4714 	spin_unlock(&fs_info->super_lock);
4715 
4716 	ret = btrfs_commit_transaction(trans);
4717 out_drop_write:
4718 	mnt_drop_write_file(file);
4719 
4720 	return ret;
4721 }
4722 
4723 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4724 {
4725 	struct btrfs_ioctl_send_args *arg;
4726 	int ret;
4727 
4728 	if (compat) {
4729 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4730 		struct btrfs_ioctl_send_args_32 args32;
4731 
4732 		ret = copy_from_user(&args32, argp, sizeof(args32));
4733 		if (ret)
4734 			return -EFAULT;
4735 		arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4736 		if (!arg)
4737 			return -ENOMEM;
4738 		arg->send_fd = args32.send_fd;
4739 		arg->clone_sources_count = args32.clone_sources_count;
4740 		arg->clone_sources = compat_ptr(args32.clone_sources);
4741 		arg->parent_root = args32.parent_root;
4742 		arg->flags = args32.flags;
4743 		memcpy(arg->reserved, args32.reserved,
4744 		       sizeof(args32.reserved));
4745 #else
4746 		return -ENOTTY;
4747 #endif
4748 	} else {
4749 		arg = memdup_user(argp, sizeof(*arg));
4750 		if (IS_ERR(arg))
4751 			return PTR_ERR(arg);
4752 	}
4753 	ret = btrfs_ioctl_send(file, arg);
4754 	kfree(arg);
4755 	return ret;
4756 }
4757 
4758 long btrfs_ioctl(struct file *file, unsigned int
4759 		cmd, unsigned long arg)
4760 {
4761 	struct inode *inode = file_inode(file);
4762 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4763 	struct btrfs_root *root = BTRFS_I(inode)->root;
4764 	void __user *argp = (void __user *)arg;
4765 
4766 	switch (cmd) {
4767 	case FS_IOC_GETFLAGS:
4768 		return btrfs_ioctl_getflags(file, argp);
4769 	case FS_IOC_SETFLAGS:
4770 		return btrfs_ioctl_setflags(file, argp);
4771 	case FS_IOC_GETVERSION:
4772 		return btrfs_ioctl_getversion(file, argp);
4773 	case FS_IOC_GETFSLABEL:
4774 		return btrfs_ioctl_get_fslabel(fs_info, argp);
4775 	case FS_IOC_SETFSLABEL:
4776 		return btrfs_ioctl_set_fslabel(file, argp);
4777 	case FITRIM:
4778 		return btrfs_ioctl_fitrim(fs_info, argp);
4779 	case BTRFS_IOC_SNAP_CREATE:
4780 		return btrfs_ioctl_snap_create(file, argp, 0);
4781 	case BTRFS_IOC_SNAP_CREATE_V2:
4782 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
4783 	case BTRFS_IOC_SUBVOL_CREATE:
4784 		return btrfs_ioctl_snap_create(file, argp, 1);
4785 	case BTRFS_IOC_SUBVOL_CREATE_V2:
4786 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
4787 	case BTRFS_IOC_SNAP_DESTROY:
4788 		return btrfs_ioctl_snap_destroy(file, argp, false);
4789 	case BTRFS_IOC_SNAP_DESTROY_V2:
4790 		return btrfs_ioctl_snap_destroy(file, argp, true);
4791 	case BTRFS_IOC_SUBVOL_GETFLAGS:
4792 		return btrfs_ioctl_subvol_getflags(file, argp);
4793 	case BTRFS_IOC_SUBVOL_SETFLAGS:
4794 		return btrfs_ioctl_subvol_setflags(file, argp);
4795 	case BTRFS_IOC_DEFAULT_SUBVOL:
4796 		return btrfs_ioctl_default_subvol(file, argp);
4797 	case BTRFS_IOC_DEFRAG:
4798 		return btrfs_ioctl_defrag(file, NULL);
4799 	case BTRFS_IOC_DEFRAG_RANGE:
4800 		return btrfs_ioctl_defrag(file, argp);
4801 	case BTRFS_IOC_RESIZE:
4802 		return btrfs_ioctl_resize(file, argp);
4803 	case BTRFS_IOC_ADD_DEV:
4804 		return btrfs_ioctl_add_dev(fs_info, argp);
4805 	case BTRFS_IOC_RM_DEV:
4806 		return btrfs_ioctl_rm_dev(file, argp);
4807 	case BTRFS_IOC_RM_DEV_V2:
4808 		return btrfs_ioctl_rm_dev_v2(file, argp);
4809 	case BTRFS_IOC_FS_INFO:
4810 		return btrfs_ioctl_fs_info(fs_info, argp);
4811 	case BTRFS_IOC_DEV_INFO:
4812 		return btrfs_ioctl_dev_info(fs_info, argp);
4813 	case BTRFS_IOC_BALANCE:
4814 		return btrfs_ioctl_balance(file, NULL);
4815 	case BTRFS_IOC_TREE_SEARCH:
4816 		return btrfs_ioctl_tree_search(file, argp);
4817 	case BTRFS_IOC_TREE_SEARCH_V2:
4818 		return btrfs_ioctl_tree_search_v2(file, argp);
4819 	case BTRFS_IOC_INO_LOOKUP:
4820 		return btrfs_ioctl_ino_lookup(file, argp);
4821 	case BTRFS_IOC_INO_PATHS:
4822 		return btrfs_ioctl_ino_to_path(root, argp);
4823 	case BTRFS_IOC_LOGICAL_INO:
4824 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4825 	case BTRFS_IOC_LOGICAL_INO_V2:
4826 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4827 	case BTRFS_IOC_SPACE_INFO:
4828 		return btrfs_ioctl_space_info(fs_info, argp);
4829 	case BTRFS_IOC_SYNC: {
4830 		int ret;
4831 
4832 		ret = btrfs_start_delalloc_roots(fs_info, -1);
4833 		if (ret)
4834 			return ret;
4835 		ret = btrfs_sync_fs(inode->i_sb, 1);
4836 		/*
4837 		 * The transaction thread may want to do more work,
4838 		 * namely it pokes the cleaner kthread that will start
4839 		 * processing uncleaned subvols.
4840 		 */
4841 		wake_up_process(fs_info->transaction_kthread);
4842 		return ret;
4843 	}
4844 	case BTRFS_IOC_START_SYNC:
4845 		return btrfs_ioctl_start_sync(root, argp);
4846 	case BTRFS_IOC_WAIT_SYNC:
4847 		return btrfs_ioctl_wait_sync(fs_info, argp);
4848 	case BTRFS_IOC_SCRUB:
4849 		return btrfs_ioctl_scrub(file, argp);
4850 	case BTRFS_IOC_SCRUB_CANCEL:
4851 		return btrfs_ioctl_scrub_cancel(fs_info);
4852 	case BTRFS_IOC_SCRUB_PROGRESS:
4853 		return btrfs_ioctl_scrub_progress(fs_info, argp);
4854 	case BTRFS_IOC_BALANCE_V2:
4855 		return btrfs_ioctl_balance(file, argp);
4856 	case BTRFS_IOC_BALANCE_CTL:
4857 		return btrfs_ioctl_balance_ctl(fs_info, arg);
4858 	case BTRFS_IOC_BALANCE_PROGRESS:
4859 		return btrfs_ioctl_balance_progress(fs_info, argp);
4860 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4861 		return btrfs_ioctl_set_received_subvol(file, argp);
4862 #ifdef CONFIG_64BIT
4863 	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4864 		return btrfs_ioctl_set_received_subvol_32(file, argp);
4865 #endif
4866 	case BTRFS_IOC_SEND:
4867 		return _btrfs_ioctl_send(file, argp, false);
4868 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4869 	case BTRFS_IOC_SEND_32:
4870 		return _btrfs_ioctl_send(file, argp, true);
4871 #endif
4872 	case BTRFS_IOC_GET_DEV_STATS:
4873 		return btrfs_ioctl_get_dev_stats(fs_info, argp);
4874 	case BTRFS_IOC_QUOTA_CTL:
4875 		return btrfs_ioctl_quota_ctl(file, argp);
4876 	case BTRFS_IOC_QGROUP_ASSIGN:
4877 		return btrfs_ioctl_qgroup_assign(file, argp);
4878 	case BTRFS_IOC_QGROUP_CREATE:
4879 		return btrfs_ioctl_qgroup_create(file, argp);
4880 	case BTRFS_IOC_QGROUP_LIMIT:
4881 		return btrfs_ioctl_qgroup_limit(file, argp);
4882 	case BTRFS_IOC_QUOTA_RESCAN:
4883 		return btrfs_ioctl_quota_rescan(file, argp);
4884 	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4885 		return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4886 	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4887 		return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4888 	case BTRFS_IOC_DEV_REPLACE:
4889 		return btrfs_ioctl_dev_replace(fs_info, argp);
4890 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4891 		return btrfs_ioctl_get_supported_features(argp);
4892 	case BTRFS_IOC_GET_FEATURES:
4893 		return btrfs_ioctl_get_features(fs_info, argp);
4894 	case BTRFS_IOC_SET_FEATURES:
4895 		return btrfs_ioctl_set_features(file, argp);
4896 	case FS_IOC_FSGETXATTR:
4897 		return btrfs_ioctl_fsgetxattr(file, argp);
4898 	case FS_IOC_FSSETXATTR:
4899 		return btrfs_ioctl_fssetxattr(file, argp);
4900 	case BTRFS_IOC_GET_SUBVOL_INFO:
4901 		return btrfs_ioctl_get_subvol_info(file, argp);
4902 	case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4903 		return btrfs_ioctl_get_subvol_rootref(file, argp);
4904 	case BTRFS_IOC_INO_LOOKUP_USER:
4905 		return btrfs_ioctl_ino_lookup_user(file, argp);
4906 	}
4907 
4908 	return -ENOTTY;
4909 }
4910 
4911 #ifdef CONFIG_COMPAT
4912 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4913 {
4914 	/*
4915 	 * These all access 32-bit values anyway so no further
4916 	 * handling is necessary.
4917 	 */
4918 	switch (cmd) {
4919 	case FS_IOC32_GETFLAGS:
4920 		cmd = FS_IOC_GETFLAGS;
4921 		break;
4922 	case FS_IOC32_SETFLAGS:
4923 		cmd = FS_IOC_SETFLAGS;
4924 		break;
4925 	case FS_IOC32_GETVERSION:
4926 		cmd = FS_IOC_GETVERSION;
4927 		break;
4928 	}
4929 
4930 	return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4931 }
4932 #endif
4933