xref: /openbmc/linux/fs/btrfs/ioctl.c (revision 37744fee)
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, &key, 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 	bool snapshot_force_cow = false;
752 
753 	if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
754 		return -EINVAL;
755 
756 	if (atomic_read(&root->nr_swapfiles)) {
757 		btrfs_warn(fs_info,
758 			   "cannot snapshot subvolume with active swapfile");
759 		return -ETXTBSY;
760 	}
761 
762 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
763 	if (!pending_snapshot)
764 		return -ENOMEM;
765 
766 	pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
767 			GFP_KERNEL);
768 	pending_snapshot->path = btrfs_alloc_path();
769 	if (!pending_snapshot->root_item || !pending_snapshot->path) {
770 		ret = -ENOMEM;
771 		goto free_pending;
772 	}
773 
774 	/*
775 	 * Force new buffered writes to reserve space even when NOCOW is
776 	 * possible. This is to avoid later writeback (running dealloc) to
777 	 * fallback to COW mode and unexpectedly fail with ENOSPC.
778 	 */
779 	btrfs_drew_read_lock(&root->snapshot_lock);
780 
781 	ret = btrfs_start_delalloc_snapshot(root);
782 	if (ret)
783 		goto dec_and_free;
784 
785 	/*
786 	 * All previous writes have started writeback in NOCOW mode, so now
787 	 * we force future writes to fallback to COW mode during snapshot
788 	 * creation.
789 	 */
790 	atomic_inc(&root->snapshot_force_cow);
791 	snapshot_force_cow = true;
792 
793 	btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
794 
795 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
796 			     BTRFS_BLOCK_RSV_TEMP);
797 	/*
798 	 * 1 - parent dir inode
799 	 * 2 - dir entries
800 	 * 1 - root item
801 	 * 2 - root ref/backref
802 	 * 1 - root of snapshot
803 	 * 1 - UUID item
804 	 */
805 	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
806 					&pending_snapshot->block_rsv, 8,
807 					false);
808 	if (ret)
809 		goto dec_and_free;
810 
811 	pending_snapshot->dentry = dentry;
812 	pending_snapshot->root = root;
813 	pending_snapshot->readonly = readonly;
814 	pending_snapshot->dir = dir;
815 	pending_snapshot->inherit = inherit;
816 
817 	trans = btrfs_start_transaction(root, 0);
818 	if (IS_ERR(trans)) {
819 		ret = PTR_ERR(trans);
820 		goto fail;
821 	}
822 
823 	spin_lock(&fs_info->trans_lock);
824 	list_add(&pending_snapshot->list,
825 		 &trans->transaction->pending_snapshots);
826 	spin_unlock(&fs_info->trans_lock);
827 
828 	ret = btrfs_commit_transaction(trans);
829 	if (ret)
830 		goto fail;
831 
832 	ret = pending_snapshot->error;
833 	if (ret)
834 		goto fail;
835 
836 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
837 	if (ret)
838 		goto fail;
839 
840 	inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
841 	if (IS_ERR(inode)) {
842 		ret = PTR_ERR(inode);
843 		goto fail;
844 	}
845 
846 	d_instantiate(dentry, inode);
847 	ret = 0;
848 fail:
849 	btrfs_put_root(pending_snapshot->snap);
850 	btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
851 dec_and_free:
852 	if (snapshot_force_cow)
853 		atomic_dec(&root->snapshot_force_cow);
854 	btrfs_drew_read_unlock(&root->snapshot_lock);
855 
856 free_pending:
857 	kfree(pending_snapshot->root_item);
858 	btrfs_free_path(pending_snapshot->path);
859 	kfree(pending_snapshot);
860 
861 	return ret;
862 }
863 
864 /*  copy of may_delete in fs/namei.c()
865  *	Check whether we can remove a link victim from directory dir, check
866  *  whether the type of victim is right.
867  *  1. We can't do it if dir is read-only (done in permission())
868  *  2. We should have write and exec permissions on dir
869  *  3. We can't remove anything from append-only dir
870  *  4. We can't do anything with immutable dir (done in permission())
871  *  5. If the sticky bit on dir is set we should either
872  *	a. be owner of dir, or
873  *	b. be owner of victim, or
874  *	c. have CAP_FOWNER capability
875  *  6. If the victim is append-only or immutable we can't do anything with
876  *     links pointing to it.
877  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
878  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
879  *  9. We can't remove a root or mountpoint.
880  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
881  *     nfs_async_unlink().
882  */
883 
884 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
885 {
886 	int error;
887 
888 	if (d_really_is_negative(victim))
889 		return -ENOENT;
890 
891 	BUG_ON(d_inode(victim->d_parent) != dir);
892 	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
893 
894 	error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
895 	if (error)
896 		return error;
897 	if (IS_APPEND(dir))
898 		return -EPERM;
899 	if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
900 	    IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
901 		return -EPERM;
902 	if (isdir) {
903 		if (!d_is_dir(victim))
904 			return -ENOTDIR;
905 		if (IS_ROOT(victim))
906 			return -EBUSY;
907 	} else if (d_is_dir(victim))
908 		return -EISDIR;
909 	if (IS_DEADDIR(dir))
910 		return -ENOENT;
911 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
912 		return -EBUSY;
913 	return 0;
914 }
915 
916 /* copy of may_create in fs/namei.c() */
917 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
918 {
919 	if (d_really_is_positive(child))
920 		return -EEXIST;
921 	if (IS_DEADDIR(dir))
922 		return -ENOENT;
923 	return inode_permission(dir, MAY_WRITE | MAY_EXEC);
924 }
925 
926 /*
927  * Create a new subvolume below @parent.  This is largely modeled after
928  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
929  * inside this filesystem so it's quite a bit simpler.
930  */
931 static noinline int btrfs_mksubvol(const struct path *parent,
932 				   const char *name, int namelen,
933 				   struct btrfs_root *snap_src,
934 				   bool readonly,
935 				   struct btrfs_qgroup_inherit *inherit)
936 {
937 	struct inode *dir = d_inode(parent->dentry);
938 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
939 	struct dentry *dentry;
940 	int error;
941 
942 	error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
943 	if (error == -EINTR)
944 		return error;
945 
946 	dentry = lookup_one_len(name, parent->dentry, namelen);
947 	error = PTR_ERR(dentry);
948 	if (IS_ERR(dentry))
949 		goto out_unlock;
950 
951 	error = btrfs_may_create(dir, dentry);
952 	if (error)
953 		goto out_dput;
954 
955 	/*
956 	 * even if this name doesn't exist, we may get hash collisions.
957 	 * check for them now when we can safely fail
958 	 */
959 	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
960 					       dir->i_ino, name,
961 					       namelen);
962 	if (error)
963 		goto out_dput;
964 
965 	down_read(&fs_info->subvol_sem);
966 
967 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
968 		goto out_up_read;
969 
970 	if (snap_src)
971 		error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
972 	else
973 		error = create_subvol(dir, dentry, name, namelen, inherit);
974 
975 	if (!error)
976 		fsnotify_mkdir(dir, dentry);
977 out_up_read:
978 	up_read(&fs_info->subvol_sem);
979 out_dput:
980 	dput(dentry);
981 out_unlock:
982 	inode_unlock(dir);
983 	return error;
984 }
985 
986 /*
987  * When we're defragging a range, we don't want to kick it off again
988  * if it is really just waiting for delalloc to send it down.
989  * If we find a nice big extent or delalloc range for the bytes in the
990  * file you want to defrag, we return 0 to let you know to skip this
991  * part of the file
992  */
993 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
994 {
995 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
996 	struct extent_map *em = NULL;
997 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
998 	u64 end;
999 
1000 	read_lock(&em_tree->lock);
1001 	em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1002 	read_unlock(&em_tree->lock);
1003 
1004 	if (em) {
1005 		end = extent_map_end(em);
1006 		free_extent_map(em);
1007 		if (end - offset > thresh)
1008 			return 0;
1009 	}
1010 	/* if we already have a nice delalloc here, just stop */
1011 	thresh /= 2;
1012 	end = count_range_bits(io_tree, &offset, offset + thresh,
1013 			       thresh, EXTENT_DELALLOC, 1);
1014 	if (end >= thresh)
1015 		return 0;
1016 	return 1;
1017 }
1018 
1019 /*
1020  * helper function to walk through a file and find extents
1021  * newer than a specific transid, and smaller than thresh.
1022  *
1023  * This is used by the defragging code to find new and small
1024  * extents
1025  */
1026 static int find_new_extents(struct btrfs_root *root,
1027 			    struct inode *inode, u64 newer_than,
1028 			    u64 *off, u32 thresh)
1029 {
1030 	struct btrfs_path *path;
1031 	struct btrfs_key min_key;
1032 	struct extent_buffer *leaf;
1033 	struct btrfs_file_extent_item *extent;
1034 	int type;
1035 	int ret;
1036 	u64 ino = btrfs_ino(BTRFS_I(inode));
1037 
1038 	path = btrfs_alloc_path();
1039 	if (!path)
1040 		return -ENOMEM;
1041 
1042 	min_key.objectid = ino;
1043 	min_key.type = BTRFS_EXTENT_DATA_KEY;
1044 	min_key.offset = *off;
1045 
1046 	while (1) {
1047 		ret = btrfs_search_forward(root, &min_key, path, newer_than);
1048 		if (ret != 0)
1049 			goto none;
1050 process_slot:
1051 		if (min_key.objectid != ino)
1052 			goto none;
1053 		if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1054 			goto none;
1055 
1056 		leaf = path->nodes[0];
1057 		extent = btrfs_item_ptr(leaf, path->slots[0],
1058 					struct btrfs_file_extent_item);
1059 
1060 		type = btrfs_file_extent_type(leaf, extent);
1061 		if (type == BTRFS_FILE_EXTENT_REG &&
1062 		    btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1063 		    check_defrag_in_cache(inode, min_key.offset, thresh)) {
1064 			*off = min_key.offset;
1065 			btrfs_free_path(path);
1066 			return 0;
1067 		}
1068 
1069 		path->slots[0]++;
1070 		if (path->slots[0] < btrfs_header_nritems(leaf)) {
1071 			btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1072 			goto process_slot;
1073 		}
1074 
1075 		if (min_key.offset == (u64)-1)
1076 			goto none;
1077 
1078 		min_key.offset++;
1079 		btrfs_release_path(path);
1080 	}
1081 none:
1082 	btrfs_free_path(path);
1083 	return -ENOENT;
1084 }
1085 
1086 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1087 {
1088 	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1089 	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1090 	struct extent_map *em;
1091 	u64 len = PAGE_SIZE;
1092 
1093 	/*
1094 	 * hopefully we have this extent in the tree already, try without
1095 	 * the full extent lock
1096 	 */
1097 	read_lock(&em_tree->lock);
1098 	em = lookup_extent_mapping(em_tree, start, len);
1099 	read_unlock(&em_tree->lock);
1100 
1101 	if (!em) {
1102 		struct extent_state *cached = NULL;
1103 		u64 end = start + len - 1;
1104 
1105 		/* get the big lock and read metadata off disk */
1106 		lock_extent_bits(io_tree, start, end, &cached);
1107 		em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1108 		unlock_extent_cached(io_tree, start, end, &cached);
1109 
1110 		if (IS_ERR(em))
1111 			return NULL;
1112 	}
1113 
1114 	return em;
1115 }
1116 
1117 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1118 {
1119 	struct extent_map *next;
1120 	bool ret = true;
1121 
1122 	/* this is the last extent */
1123 	if (em->start + em->len >= i_size_read(inode))
1124 		return false;
1125 
1126 	next = defrag_lookup_extent(inode, em->start + em->len);
1127 	if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1128 		ret = false;
1129 	else if ((em->block_start + em->block_len == next->block_start) &&
1130 		 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1131 		ret = false;
1132 
1133 	free_extent_map(next);
1134 	return ret;
1135 }
1136 
1137 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1138 			       u64 *last_len, u64 *skip, u64 *defrag_end,
1139 			       int compress)
1140 {
1141 	struct extent_map *em;
1142 	int ret = 1;
1143 	bool next_mergeable = true;
1144 	bool prev_mergeable = true;
1145 
1146 	/*
1147 	 * make sure that once we start defragging an extent, we keep on
1148 	 * defragging it
1149 	 */
1150 	if (start < *defrag_end)
1151 		return 1;
1152 
1153 	*skip = 0;
1154 
1155 	em = defrag_lookup_extent(inode, start);
1156 	if (!em)
1157 		return 0;
1158 
1159 	/* this will cover holes, and inline extents */
1160 	if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1161 		ret = 0;
1162 		goto out;
1163 	}
1164 
1165 	if (!*defrag_end)
1166 		prev_mergeable = false;
1167 
1168 	next_mergeable = defrag_check_next_extent(inode, em);
1169 	/*
1170 	 * we hit a real extent, if it is big or the next extent is not a
1171 	 * real extent, don't bother defragging it
1172 	 */
1173 	if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1174 	    (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1175 		ret = 0;
1176 out:
1177 	/*
1178 	 * last_len ends up being a counter of how many bytes we've defragged.
1179 	 * every time we choose not to defrag an extent, we reset *last_len
1180 	 * so that the next tiny extent will force a defrag.
1181 	 *
1182 	 * The end result of this is that tiny extents before a single big
1183 	 * extent will force at least part of that big extent to be defragged.
1184 	 */
1185 	if (ret) {
1186 		*defrag_end = extent_map_end(em);
1187 	} else {
1188 		*last_len = 0;
1189 		*skip = extent_map_end(em);
1190 		*defrag_end = 0;
1191 	}
1192 
1193 	free_extent_map(em);
1194 	return ret;
1195 }
1196 
1197 /*
1198  * it doesn't do much good to defrag one or two pages
1199  * at a time.  This pulls in a nice chunk of pages
1200  * to COW and defrag.
1201  *
1202  * It also makes sure the delalloc code has enough
1203  * dirty data to avoid making new small extents as part
1204  * of the defrag
1205  *
1206  * It's a good idea to start RA on this range
1207  * before calling this.
1208  */
1209 static int cluster_pages_for_defrag(struct inode *inode,
1210 				    struct page **pages,
1211 				    unsigned long start_index,
1212 				    unsigned long num_pages)
1213 {
1214 	unsigned long file_end;
1215 	u64 isize = i_size_read(inode);
1216 	u64 page_start;
1217 	u64 page_end;
1218 	u64 page_cnt;
1219 	int ret;
1220 	int i;
1221 	int i_done;
1222 	struct btrfs_ordered_extent *ordered;
1223 	struct extent_state *cached_state = NULL;
1224 	struct extent_io_tree *tree;
1225 	struct extent_changeset *data_reserved = NULL;
1226 	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1227 
1228 	file_end = (isize - 1) >> PAGE_SHIFT;
1229 	if (!isize || start_index > file_end)
1230 		return 0;
1231 
1232 	page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1233 
1234 	ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1235 			start_index << PAGE_SHIFT,
1236 			page_cnt << PAGE_SHIFT);
1237 	if (ret)
1238 		return ret;
1239 	i_done = 0;
1240 	tree = &BTRFS_I(inode)->io_tree;
1241 
1242 	/* step one, lock all the pages */
1243 	for (i = 0; i < page_cnt; i++) {
1244 		struct page *page;
1245 again:
1246 		page = find_or_create_page(inode->i_mapping,
1247 					   start_index + i, mask);
1248 		if (!page)
1249 			break;
1250 
1251 		page_start = page_offset(page);
1252 		page_end = page_start + PAGE_SIZE - 1;
1253 		while (1) {
1254 			lock_extent_bits(tree, page_start, page_end,
1255 					 &cached_state);
1256 			ordered = btrfs_lookup_ordered_extent(inode,
1257 							      page_start);
1258 			unlock_extent_cached(tree, page_start, page_end,
1259 					     &cached_state);
1260 			if (!ordered)
1261 				break;
1262 
1263 			unlock_page(page);
1264 			btrfs_start_ordered_extent(inode, ordered, 1);
1265 			btrfs_put_ordered_extent(ordered);
1266 			lock_page(page);
1267 			/*
1268 			 * we unlocked the page above, so we need check if
1269 			 * it was released or not.
1270 			 */
1271 			if (page->mapping != inode->i_mapping) {
1272 				unlock_page(page);
1273 				put_page(page);
1274 				goto again;
1275 			}
1276 		}
1277 
1278 		if (!PageUptodate(page)) {
1279 			btrfs_readpage(NULL, page);
1280 			lock_page(page);
1281 			if (!PageUptodate(page)) {
1282 				unlock_page(page);
1283 				put_page(page);
1284 				ret = -EIO;
1285 				break;
1286 			}
1287 		}
1288 
1289 		if (page->mapping != inode->i_mapping) {
1290 			unlock_page(page);
1291 			put_page(page);
1292 			goto again;
1293 		}
1294 
1295 		pages[i] = page;
1296 		i_done++;
1297 	}
1298 	if (!i_done || ret)
1299 		goto out;
1300 
1301 	if (!(inode->i_sb->s_flags & SB_ACTIVE))
1302 		goto out;
1303 
1304 	/*
1305 	 * so now we have a nice long stream of locked
1306 	 * and up to date pages, lets wait on them
1307 	 */
1308 	for (i = 0; i < i_done; i++)
1309 		wait_on_page_writeback(pages[i]);
1310 
1311 	page_start = page_offset(pages[0]);
1312 	page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1313 
1314 	lock_extent_bits(&BTRFS_I(inode)->io_tree,
1315 			 page_start, page_end - 1, &cached_state);
1316 	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1317 			  page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1318 			  EXTENT_DEFRAG, 0, 0, &cached_state);
1319 
1320 	if (i_done != page_cnt) {
1321 		spin_lock(&BTRFS_I(inode)->lock);
1322 		btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1323 		spin_unlock(&BTRFS_I(inode)->lock);
1324 		btrfs_delalloc_release_space(inode, data_reserved,
1325 				start_index << PAGE_SHIFT,
1326 				(page_cnt - i_done) << PAGE_SHIFT, true);
1327 	}
1328 
1329 
1330 	set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1331 			  &cached_state);
1332 
1333 	unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1334 			     page_start, page_end - 1, &cached_state);
1335 
1336 	for (i = 0; i < i_done; i++) {
1337 		clear_page_dirty_for_io(pages[i]);
1338 		ClearPageChecked(pages[i]);
1339 		set_page_extent_mapped(pages[i]);
1340 		set_page_dirty(pages[i]);
1341 		unlock_page(pages[i]);
1342 		put_page(pages[i]);
1343 	}
1344 	btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1345 	extent_changeset_free(data_reserved);
1346 	return i_done;
1347 out:
1348 	for (i = 0; i < i_done; i++) {
1349 		unlock_page(pages[i]);
1350 		put_page(pages[i]);
1351 	}
1352 	btrfs_delalloc_release_space(inode, data_reserved,
1353 			start_index << PAGE_SHIFT,
1354 			page_cnt << PAGE_SHIFT, true);
1355 	btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1356 	extent_changeset_free(data_reserved);
1357 	return ret;
1358 
1359 }
1360 
1361 int btrfs_defrag_file(struct inode *inode, struct file *file,
1362 		      struct btrfs_ioctl_defrag_range_args *range,
1363 		      u64 newer_than, unsigned long max_to_defrag)
1364 {
1365 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1366 	struct btrfs_root *root = BTRFS_I(inode)->root;
1367 	struct file_ra_state *ra = NULL;
1368 	unsigned long last_index;
1369 	u64 isize = i_size_read(inode);
1370 	u64 last_len = 0;
1371 	u64 skip = 0;
1372 	u64 defrag_end = 0;
1373 	u64 newer_off = range->start;
1374 	unsigned long i;
1375 	unsigned long ra_index = 0;
1376 	int ret;
1377 	int defrag_count = 0;
1378 	int compress_type = BTRFS_COMPRESS_ZLIB;
1379 	u32 extent_thresh = range->extent_thresh;
1380 	unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1381 	unsigned long cluster = max_cluster;
1382 	u64 new_align = ~((u64)SZ_128K - 1);
1383 	struct page **pages = NULL;
1384 	bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1385 
1386 	if (isize == 0)
1387 		return 0;
1388 
1389 	if (range->start >= isize)
1390 		return -EINVAL;
1391 
1392 	if (do_compress) {
1393 		if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1394 			return -EINVAL;
1395 		if (range->compress_type)
1396 			compress_type = range->compress_type;
1397 	}
1398 
1399 	if (extent_thresh == 0)
1400 		extent_thresh = SZ_256K;
1401 
1402 	/*
1403 	 * If we were not given a file, allocate a readahead context. As
1404 	 * readahead is just an optimization, defrag will work without it so
1405 	 * we don't error out.
1406 	 */
1407 	if (!file) {
1408 		ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1409 		if (ra)
1410 			file_ra_state_init(ra, inode->i_mapping);
1411 	} else {
1412 		ra = &file->f_ra;
1413 	}
1414 
1415 	pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1416 	if (!pages) {
1417 		ret = -ENOMEM;
1418 		goto out_ra;
1419 	}
1420 
1421 	/* find the last page to defrag */
1422 	if (range->start + range->len > range->start) {
1423 		last_index = min_t(u64, isize - 1,
1424 			 range->start + range->len - 1) >> PAGE_SHIFT;
1425 	} else {
1426 		last_index = (isize - 1) >> PAGE_SHIFT;
1427 	}
1428 
1429 	if (newer_than) {
1430 		ret = find_new_extents(root, inode, newer_than,
1431 				       &newer_off, SZ_64K);
1432 		if (!ret) {
1433 			range->start = newer_off;
1434 			/*
1435 			 * we always align our defrag to help keep
1436 			 * the extents in the file evenly spaced
1437 			 */
1438 			i = (newer_off & new_align) >> PAGE_SHIFT;
1439 		} else
1440 			goto out_ra;
1441 	} else {
1442 		i = range->start >> PAGE_SHIFT;
1443 	}
1444 	if (!max_to_defrag)
1445 		max_to_defrag = last_index - i + 1;
1446 
1447 	/*
1448 	 * make writeback starts from i, so the defrag range can be
1449 	 * written sequentially.
1450 	 */
1451 	if (i < inode->i_mapping->writeback_index)
1452 		inode->i_mapping->writeback_index = i;
1453 
1454 	while (i <= last_index && defrag_count < max_to_defrag &&
1455 	       (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1456 		/*
1457 		 * make sure we stop running if someone unmounts
1458 		 * the FS
1459 		 */
1460 		if (!(inode->i_sb->s_flags & SB_ACTIVE))
1461 			break;
1462 
1463 		if (btrfs_defrag_cancelled(fs_info)) {
1464 			btrfs_debug(fs_info, "defrag_file cancelled");
1465 			ret = -EAGAIN;
1466 			break;
1467 		}
1468 
1469 		if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1470 					 extent_thresh, &last_len, &skip,
1471 					 &defrag_end, do_compress)){
1472 			unsigned long next;
1473 			/*
1474 			 * the should_defrag function tells us how much to skip
1475 			 * bump our counter by the suggested amount
1476 			 */
1477 			next = DIV_ROUND_UP(skip, PAGE_SIZE);
1478 			i = max(i + 1, next);
1479 			continue;
1480 		}
1481 
1482 		if (!newer_than) {
1483 			cluster = (PAGE_ALIGN(defrag_end) >>
1484 				   PAGE_SHIFT) - i;
1485 			cluster = min(cluster, max_cluster);
1486 		} else {
1487 			cluster = max_cluster;
1488 		}
1489 
1490 		if (i + cluster > ra_index) {
1491 			ra_index = max(i, ra_index);
1492 			if (ra)
1493 				page_cache_sync_readahead(inode->i_mapping, ra,
1494 						file, ra_index, cluster);
1495 			ra_index += cluster;
1496 		}
1497 
1498 		inode_lock(inode);
1499 		if (IS_SWAPFILE(inode)) {
1500 			ret = -ETXTBSY;
1501 		} else {
1502 			if (do_compress)
1503 				BTRFS_I(inode)->defrag_compress = compress_type;
1504 			ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1505 		}
1506 		if (ret < 0) {
1507 			inode_unlock(inode);
1508 			goto out_ra;
1509 		}
1510 
1511 		defrag_count += ret;
1512 		balance_dirty_pages_ratelimited(inode->i_mapping);
1513 		inode_unlock(inode);
1514 
1515 		if (newer_than) {
1516 			if (newer_off == (u64)-1)
1517 				break;
1518 
1519 			if (ret > 0)
1520 				i += ret;
1521 
1522 			newer_off = max(newer_off + 1,
1523 					(u64)i << PAGE_SHIFT);
1524 
1525 			ret = find_new_extents(root, inode, newer_than,
1526 					       &newer_off, SZ_64K);
1527 			if (!ret) {
1528 				range->start = newer_off;
1529 				i = (newer_off & new_align) >> PAGE_SHIFT;
1530 			} else {
1531 				break;
1532 			}
1533 		} else {
1534 			if (ret > 0) {
1535 				i += ret;
1536 				last_len += ret << PAGE_SHIFT;
1537 			} else {
1538 				i++;
1539 				last_len = 0;
1540 			}
1541 		}
1542 	}
1543 
1544 	if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1545 		filemap_flush(inode->i_mapping);
1546 		if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1547 			     &BTRFS_I(inode)->runtime_flags))
1548 			filemap_flush(inode->i_mapping);
1549 	}
1550 
1551 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
1552 		btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1553 	} else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1554 		btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1555 	}
1556 
1557 	ret = defrag_count;
1558 
1559 out_ra:
1560 	if (do_compress) {
1561 		inode_lock(inode);
1562 		BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1563 		inode_unlock(inode);
1564 	}
1565 	if (!file)
1566 		kfree(ra);
1567 	kfree(pages);
1568 	return ret;
1569 }
1570 
1571 static noinline int btrfs_ioctl_resize(struct file *file,
1572 					void __user *arg)
1573 {
1574 	struct inode *inode = file_inode(file);
1575 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1576 	u64 new_size;
1577 	u64 old_size;
1578 	u64 devid = 1;
1579 	struct btrfs_root *root = BTRFS_I(inode)->root;
1580 	struct btrfs_ioctl_vol_args *vol_args;
1581 	struct btrfs_trans_handle *trans;
1582 	struct btrfs_device *device = NULL;
1583 	char *sizestr;
1584 	char *retptr;
1585 	char *devstr = NULL;
1586 	int ret = 0;
1587 	int mod = 0;
1588 
1589 	if (!capable(CAP_SYS_ADMIN))
1590 		return -EPERM;
1591 
1592 	ret = mnt_want_write_file(file);
1593 	if (ret)
1594 		return ret;
1595 
1596 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1597 		mnt_drop_write_file(file);
1598 		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1599 	}
1600 
1601 	vol_args = memdup_user(arg, sizeof(*vol_args));
1602 	if (IS_ERR(vol_args)) {
1603 		ret = PTR_ERR(vol_args);
1604 		goto out;
1605 	}
1606 
1607 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1608 
1609 	sizestr = vol_args->name;
1610 	devstr = strchr(sizestr, ':');
1611 	if (devstr) {
1612 		sizestr = devstr + 1;
1613 		*devstr = '\0';
1614 		devstr = vol_args->name;
1615 		ret = kstrtoull(devstr, 10, &devid);
1616 		if (ret)
1617 			goto out_free;
1618 		if (!devid) {
1619 			ret = -EINVAL;
1620 			goto out_free;
1621 		}
1622 		btrfs_info(fs_info, "resizing devid %llu", devid);
1623 	}
1624 
1625 	device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL, true);
1626 	if (!device) {
1627 		btrfs_info(fs_info, "resizer unable to find device %llu",
1628 			   devid);
1629 		ret = -ENODEV;
1630 		goto out_free;
1631 	}
1632 
1633 	if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1634 		btrfs_info(fs_info,
1635 			   "resizer unable to apply on readonly device %llu",
1636 		       devid);
1637 		ret = -EPERM;
1638 		goto out_free;
1639 	}
1640 
1641 	if (!strcmp(sizestr, "max"))
1642 		new_size = device->bdev->bd_inode->i_size;
1643 	else {
1644 		if (sizestr[0] == '-') {
1645 			mod = -1;
1646 			sizestr++;
1647 		} else if (sizestr[0] == '+') {
1648 			mod = 1;
1649 			sizestr++;
1650 		}
1651 		new_size = memparse(sizestr, &retptr);
1652 		if (*retptr != '\0' || new_size == 0) {
1653 			ret = -EINVAL;
1654 			goto out_free;
1655 		}
1656 	}
1657 
1658 	if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1659 		ret = -EPERM;
1660 		goto out_free;
1661 	}
1662 
1663 	old_size = btrfs_device_get_total_bytes(device);
1664 
1665 	if (mod < 0) {
1666 		if (new_size > old_size) {
1667 			ret = -EINVAL;
1668 			goto out_free;
1669 		}
1670 		new_size = old_size - new_size;
1671 	} else if (mod > 0) {
1672 		if (new_size > ULLONG_MAX - old_size) {
1673 			ret = -ERANGE;
1674 			goto out_free;
1675 		}
1676 		new_size = old_size + new_size;
1677 	}
1678 
1679 	if (new_size < SZ_256M) {
1680 		ret = -EINVAL;
1681 		goto out_free;
1682 	}
1683 	if (new_size > device->bdev->bd_inode->i_size) {
1684 		ret = -EFBIG;
1685 		goto out_free;
1686 	}
1687 
1688 	new_size = round_down(new_size, fs_info->sectorsize);
1689 
1690 	if (new_size > old_size) {
1691 		trans = btrfs_start_transaction(root, 0);
1692 		if (IS_ERR(trans)) {
1693 			ret = PTR_ERR(trans);
1694 			goto out_free;
1695 		}
1696 		ret = btrfs_grow_device(trans, device, new_size);
1697 		btrfs_commit_transaction(trans);
1698 	} else if (new_size < old_size) {
1699 		ret = btrfs_shrink_device(device, new_size);
1700 	} /* equal, nothing need to do */
1701 
1702 	if (ret == 0 && new_size != old_size)
1703 		btrfs_info_in_rcu(fs_info,
1704 			"resize device %s (devid %llu) from %llu to %llu",
1705 			rcu_str_deref(device->name), device->devid,
1706 			old_size, new_size);
1707 out_free:
1708 	kfree(vol_args);
1709 out:
1710 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1711 	mnt_drop_write_file(file);
1712 	return ret;
1713 }
1714 
1715 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1716 				const char *name, unsigned long fd, int subvol,
1717 				bool readonly,
1718 				struct btrfs_qgroup_inherit *inherit)
1719 {
1720 	int namelen;
1721 	int ret = 0;
1722 
1723 	if (!S_ISDIR(file_inode(file)->i_mode))
1724 		return -ENOTDIR;
1725 
1726 	ret = mnt_want_write_file(file);
1727 	if (ret)
1728 		goto out;
1729 
1730 	namelen = strlen(name);
1731 	if (strchr(name, '/')) {
1732 		ret = -EINVAL;
1733 		goto out_drop_write;
1734 	}
1735 
1736 	if (name[0] == '.' &&
1737 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1738 		ret = -EEXIST;
1739 		goto out_drop_write;
1740 	}
1741 
1742 	if (subvol) {
1743 		ret = btrfs_mksubvol(&file->f_path, name, namelen,
1744 				     NULL, readonly, inherit);
1745 	} else {
1746 		struct fd src = fdget(fd);
1747 		struct inode *src_inode;
1748 		if (!src.file) {
1749 			ret = -EINVAL;
1750 			goto out_drop_write;
1751 		}
1752 
1753 		src_inode = file_inode(src.file);
1754 		if (src_inode->i_sb != file_inode(file)->i_sb) {
1755 			btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1756 				   "Snapshot src from another FS");
1757 			ret = -EXDEV;
1758 		} else if (!inode_owner_or_capable(src_inode)) {
1759 			/*
1760 			 * Subvolume creation is not restricted, but snapshots
1761 			 * are limited to own subvolumes only
1762 			 */
1763 			ret = -EPERM;
1764 		} else {
1765 			ret = btrfs_mksubvol(&file->f_path, name, namelen,
1766 					     BTRFS_I(src_inode)->root,
1767 					     readonly, inherit);
1768 		}
1769 		fdput(src);
1770 	}
1771 out_drop_write:
1772 	mnt_drop_write_file(file);
1773 out:
1774 	return ret;
1775 }
1776 
1777 static noinline int btrfs_ioctl_snap_create(struct file *file,
1778 					    void __user *arg, int subvol)
1779 {
1780 	struct btrfs_ioctl_vol_args *vol_args;
1781 	int ret;
1782 
1783 	if (!S_ISDIR(file_inode(file)->i_mode))
1784 		return -ENOTDIR;
1785 
1786 	vol_args = memdup_user(arg, sizeof(*vol_args));
1787 	if (IS_ERR(vol_args))
1788 		return PTR_ERR(vol_args);
1789 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1790 
1791 	ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1792 					subvol, false, NULL);
1793 
1794 	kfree(vol_args);
1795 	return ret;
1796 }
1797 
1798 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1799 					       void __user *arg, int subvol)
1800 {
1801 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1802 	int ret;
1803 	bool readonly = false;
1804 	struct btrfs_qgroup_inherit *inherit = NULL;
1805 
1806 	if (!S_ISDIR(file_inode(file)->i_mode))
1807 		return -ENOTDIR;
1808 
1809 	vol_args = memdup_user(arg, sizeof(*vol_args));
1810 	if (IS_ERR(vol_args))
1811 		return PTR_ERR(vol_args);
1812 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1813 
1814 	if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1815 		ret = -EOPNOTSUPP;
1816 		goto free_args;
1817 	}
1818 
1819 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1820 		readonly = true;
1821 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1822 		if (vol_args->size > PAGE_SIZE) {
1823 			ret = -EINVAL;
1824 			goto free_args;
1825 		}
1826 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1827 		if (IS_ERR(inherit)) {
1828 			ret = PTR_ERR(inherit);
1829 			goto free_args;
1830 		}
1831 	}
1832 
1833 	ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1834 					subvol, readonly, inherit);
1835 	if (ret)
1836 		goto free_inherit;
1837 free_inherit:
1838 	kfree(inherit);
1839 free_args:
1840 	kfree(vol_args);
1841 	return ret;
1842 }
1843 
1844 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1845 						void __user *arg)
1846 {
1847 	struct inode *inode = file_inode(file);
1848 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1849 	struct btrfs_root *root = BTRFS_I(inode)->root;
1850 	int ret = 0;
1851 	u64 flags = 0;
1852 
1853 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1854 		return -EINVAL;
1855 
1856 	down_read(&fs_info->subvol_sem);
1857 	if (btrfs_root_readonly(root))
1858 		flags |= BTRFS_SUBVOL_RDONLY;
1859 	up_read(&fs_info->subvol_sem);
1860 
1861 	if (copy_to_user(arg, &flags, sizeof(flags)))
1862 		ret = -EFAULT;
1863 
1864 	return ret;
1865 }
1866 
1867 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1868 					      void __user *arg)
1869 {
1870 	struct inode *inode = file_inode(file);
1871 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1872 	struct btrfs_root *root = BTRFS_I(inode)->root;
1873 	struct btrfs_trans_handle *trans;
1874 	u64 root_flags;
1875 	u64 flags;
1876 	int ret = 0;
1877 
1878 	if (!inode_owner_or_capable(inode))
1879 		return -EPERM;
1880 
1881 	ret = mnt_want_write_file(file);
1882 	if (ret)
1883 		goto out;
1884 
1885 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1886 		ret = -EINVAL;
1887 		goto out_drop_write;
1888 	}
1889 
1890 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1891 		ret = -EFAULT;
1892 		goto out_drop_write;
1893 	}
1894 
1895 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1896 		ret = -EOPNOTSUPP;
1897 		goto out_drop_write;
1898 	}
1899 
1900 	down_write(&fs_info->subvol_sem);
1901 
1902 	/* nothing to do */
1903 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1904 		goto out_drop_sem;
1905 
1906 	root_flags = btrfs_root_flags(&root->root_item);
1907 	if (flags & BTRFS_SUBVOL_RDONLY) {
1908 		btrfs_set_root_flags(&root->root_item,
1909 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1910 	} else {
1911 		/*
1912 		 * Block RO -> RW transition if this subvolume is involved in
1913 		 * send
1914 		 */
1915 		spin_lock(&root->root_item_lock);
1916 		if (root->send_in_progress == 0) {
1917 			btrfs_set_root_flags(&root->root_item,
1918 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1919 			spin_unlock(&root->root_item_lock);
1920 		} else {
1921 			spin_unlock(&root->root_item_lock);
1922 			btrfs_warn(fs_info,
1923 				   "Attempt to set subvolume %llu read-write during send",
1924 				   root->root_key.objectid);
1925 			ret = -EPERM;
1926 			goto out_drop_sem;
1927 		}
1928 	}
1929 
1930 	trans = btrfs_start_transaction(root, 1);
1931 	if (IS_ERR(trans)) {
1932 		ret = PTR_ERR(trans);
1933 		goto out_reset;
1934 	}
1935 
1936 	ret = btrfs_update_root(trans, fs_info->tree_root,
1937 				&root->root_key, &root->root_item);
1938 	if (ret < 0) {
1939 		btrfs_end_transaction(trans);
1940 		goto out_reset;
1941 	}
1942 
1943 	ret = btrfs_commit_transaction(trans);
1944 
1945 out_reset:
1946 	if (ret)
1947 		btrfs_set_root_flags(&root->root_item, root_flags);
1948 out_drop_sem:
1949 	up_write(&fs_info->subvol_sem);
1950 out_drop_write:
1951 	mnt_drop_write_file(file);
1952 out:
1953 	return ret;
1954 }
1955 
1956 static noinline int key_in_sk(struct btrfs_key *key,
1957 			      struct btrfs_ioctl_search_key *sk)
1958 {
1959 	struct btrfs_key test;
1960 	int ret;
1961 
1962 	test.objectid = sk->min_objectid;
1963 	test.type = sk->min_type;
1964 	test.offset = sk->min_offset;
1965 
1966 	ret = btrfs_comp_cpu_keys(key, &test);
1967 	if (ret < 0)
1968 		return 0;
1969 
1970 	test.objectid = sk->max_objectid;
1971 	test.type = sk->max_type;
1972 	test.offset = sk->max_offset;
1973 
1974 	ret = btrfs_comp_cpu_keys(key, &test);
1975 	if (ret > 0)
1976 		return 0;
1977 	return 1;
1978 }
1979 
1980 static noinline int copy_to_sk(struct btrfs_path *path,
1981 			       struct btrfs_key *key,
1982 			       struct btrfs_ioctl_search_key *sk,
1983 			       size_t *buf_size,
1984 			       char __user *ubuf,
1985 			       unsigned long *sk_offset,
1986 			       int *num_found)
1987 {
1988 	u64 found_transid;
1989 	struct extent_buffer *leaf;
1990 	struct btrfs_ioctl_search_header sh;
1991 	struct btrfs_key test;
1992 	unsigned long item_off;
1993 	unsigned long item_len;
1994 	int nritems;
1995 	int i;
1996 	int slot;
1997 	int ret = 0;
1998 
1999 	leaf = path->nodes[0];
2000 	slot = path->slots[0];
2001 	nritems = btrfs_header_nritems(leaf);
2002 
2003 	if (btrfs_header_generation(leaf) > sk->max_transid) {
2004 		i = nritems;
2005 		goto advance_key;
2006 	}
2007 	found_transid = btrfs_header_generation(leaf);
2008 
2009 	for (i = slot; i < nritems; i++) {
2010 		item_off = btrfs_item_ptr_offset(leaf, i);
2011 		item_len = btrfs_item_size_nr(leaf, i);
2012 
2013 		btrfs_item_key_to_cpu(leaf, key, i);
2014 		if (!key_in_sk(key, sk))
2015 			continue;
2016 
2017 		if (sizeof(sh) + item_len > *buf_size) {
2018 			if (*num_found) {
2019 				ret = 1;
2020 				goto out;
2021 			}
2022 
2023 			/*
2024 			 * return one empty item back for v1, which does not
2025 			 * handle -EOVERFLOW
2026 			 */
2027 
2028 			*buf_size = sizeof(sh) + item_len;
2029 			item_len = 0;
2030 			ret = -EOVERFLOW;
2031 		}
2032 
2033 		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2034 			ret = 1;
2035 			goto out;
2036 		}
2037 
2038 		sh.objectid = key->objectid;
2039 		sh.offset = key->offset;
2040 		sh.type = key->type;
2041 		sh.len = item_len;
2042 		sh.transid = found_transid;
2043 
2044 		/* copy search result header */
2045 		if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2046 			ret = -EFAULT;
2047 			goto out;
2048 		}
2049 
2050 		*sk_offset += sizeof(sh);
2051 
2052 		if (item_len) {
2053 			char __user *up = ubuf + *sk_offset;
2054 			/* copy the item */
2055 			if (read_extent_buffer_to_user(leaf, up,
2056 						       item_off, item_len)) {
2057 				ret = -EFAULT;
2058 				goto out;
2059 			}
2060 
2061 			*sk_offset += item_len;
2062 		}
2063 		(*num_found)++;
2064 
2065 		if (ret) /* -EOVERFLOW from above */
2066 			goto out;
2067 
2068 		if (*num_found >= sk->nr_items) {
2069 			ret = 1;
2070 			goto out;
2071 		}
2072 	}
2073 advance_key:
2074 	ret = 0;
2075 	test.objectid = sk->max_objectid;
2076 	test.type = sk->max_type;
2077 	test.offset = sk->max_offset;
2078 	if (btrfs_comp_cpu_keys(key, &test) >= 0)
2079 		ret = 1;
2080 	else if (key->offset < (u64)-1)
2081 		key->offset++;
2082 	else if (key->type < (u8)-1) {
2083 		key->offset = 0;
2084 		key->type++;
2085 	} else if (key->objectid < (u64)-1) {
2086 		key->offset = 0;
2087 		key->type = 0;
2088 		key->objectid++;
2089 	} else
2090 		ret = 1;
2091 out:
2092 	/*
2093 	 *  0: all items from this leaf copied, continue with next
2094 	 *  1: * more items can be copied, but unused buffer is too small
2095 	 *     * all items were found
2096 	 *     Either way, it will stops the loop which iterates to the next
2097 	 *     leaf
2098 	 *  -EOVERFLOW: item was to large for buffer
2099 	 *  -EFAULT: could not copy extent buffer back to userspace
2100 	 */
2101 	return ret;
2102 }
2103 
2104 static noinline int search_ioctl(struct inode *inode,
2105 				 struct btrfs_ioctl_search_key *sk,
2106 				 size_t *buf_size,
2107 				 char __user *ubuf)
2108 {
2109 	struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2110 	struct btrfs_root *root;
2111 	struct btrfs_key key;
2112 	struct btrfs_path *path;
2113 	int ret;
2114 	int num_found = 0;
2115 	unsigned long sk_offset = 0;
2116 
2117 	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2118 		*buf_size = sizeof(struct btrfs_ioctl_search_header);
2119 		return -EOVERFLOW;
2120 	}
2121 
2122 	path = btrfs_alloc_path();
2123 	if (!path)
2124 		return -ENOMEM;
2125 
2126 	if (sk->tree_id == 0) {
2127 		/* search the root of the inode that was passed */
2128 		root = btrfs_grab_root(BTRFS_I(inode)->root);
2129 	} else {
2130 		key.objectid = sk->tree_id;
2131 		key.type = BTRFS_ROOT_ITEM_KEY;
2132 		key.offset = (u64)-1;
2133 		root = btrfs_get_fs_root(info, &key, true);
2134 		if (IS_ERR(root)) {
2135 			btrfs_free_path(path);
2136 			return PTR_ERR(root);
2137 		}
2138 	}
2139 
2140 	key.objectid = sk->min_objectid;
2141 	key.type = sk->min_type;
2142 	key.offset = sk->min_offset;
2143 
2144 	while (1) {
2145 		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2146 		if (ret != 0) {
2147 			if (ret > 0)
2148 				ret = 0;
2149 			goto err;
2150 		}
2151 		ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2152 				 &sk_offset, &num_found);
2153 		btrfs_release_path(path);
2154 		if (ret)
2155 			break;
2156 
2157 	}
2158 	if (ret > 0)
2159 		ret = 0;
2160 err:
2161 	sk->nr_items = num_found;
2162 	btrfs_put_root(root);
2163 	btrfs_free_path(path);
2164 	return ret;
2165 }
2166 
2167 static noinline int btrfs_ioctl_tree_search(struct file *file,
2168 					   void __user *argp)
2169 {
2170 	struct btrfs_ioctl_search_args __user *uargs;
2171 	struct btrfs_ioctl_search_key sk;
2172 	struct inode *inode;
2173 	int ret;
2174 	size_t buf_size;
2175 
2176 	if (!capable(CAP_SYS_ADMIN))
2177 		return -EPERM;
2178 
2179 	uargs = (struct btrfs_ioctl_search_args __user *)argp;
2180 
2181 	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2182 		return -EFAULT;
2183 
2184 	buf_size = sizeof(uargs->buf);
2185 
2186 	inode = file_inode(file);
2187 	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2188 
2189 	/*
2190 	 * In the origin implementation an overflow is handled by returning a
2191 	 * search header with a len of zero, so reset ret.
2192 	 */
2193 	if (ret == -EOVERFLOW)
2194 		ret = 0;
2195 
2196 	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2197 		ret = -EFAULT;
2198 	return ret;
2199 }
2200 
2201 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2202 					       void __user *argp)
2203 {
2204 	struct btrfs_ioctl_search_args_v2 __user *uarg;
2205 	struct btrfs_ioctl_search_args_v2 args;
2206 	struct inode *inode;
2207 	int ret;
2208 	size_t buf_size;
2209 	const size_t buf_limit = SZ_16M;
2210 
2211 	if (!capable(CAP_SYS_ADMIN))
2212 		return -EPERM;
2213 
2214 	/* copy search header and buffer size */
2215 	uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2216 	if (copy_from_user(&args, uarg, sizeof(args)))
2217 		return -EFAULT;
2218 
2219 	buf_size = args.buf_size;
2220 
2221 	/* limit result size to 16MB */
2222 	if (buf_size > buf_limit)
2223 		buf_size = buf_limit;
2224 
2225 	inode = file_inode(file);
2226 	ret = search_ioctl(inode, &args.key, &buf_size,
2227 			   (char __user *)(&uarg->buf[0]));
2228 	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2229 		ret = -EFAULT;
2230 	else if (ret == -EOVERFLOW &&
2231 		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2232 		ret = -EFAULT;
2233 
2234 	return ret;
2235 }
2236 
2237 /*
2238  * Search INODE_REFs to identify path name of 'dirid' directory
2239  * in a 'tree_id' tree. and sets path name to 'name'.
2240  */
2241 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2242 				u64 tree_id, u64 dirid, char *name)
2243 {
2244 	struct btrfs_root *root;
2245 	struct btrfs_key key;
2246 	char *ptr;
2247 	int ret = -1;
2248 	int slot;
2249 	int len;
2250 	int total_len = 0;
2251 	struct btrfs_inode_ref *iref;
2252 	struct extent_buffer *l;
2253 	struct btrfs_path *path;
2254 
2255 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2256 		name[0]='\0';
2257 		return 0;
2258 	}
2259 
2260 	path = btrfs_alloc_path();
2261 	if (!path)
2262 		return -ENOMEM;
2263 
2264 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2265 
2266 	key.objectid = tree_id;
2267 	key.type = BTRFS_ROOT_ITEM_KEY;
2268 	key.offset = (u64)-1;
2269 	root = btrfs_get_fs_root(info, &key, true);
2270 	if (IS_ERR(root)) {
2271 		ret = PTR_ERR(root);
2272 		root = NULL;
2273 		goto out;
2274 	}
2275 
2276 	key.objectid = dirid;
2277 	key.type = BTRFS_INODE_REF_KEY;
2278 	key.offset = (u64)-1;
2279 
2280 	while (1) {
2281 		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2282 		if (ret < 0)
2283 			goto out;
2284 		else if (ret > 0) {
2285 			ret = btrfs_previous_item(root, path, dirid,
2286 						  BTRFS_INODE_REF_KEY);
2287 			if (ret < 0)
2288 				goto out;
2289 			else if (ret > 0) {
2290 				ret = -ENOENT;
2291 				goto out;
2292 			}
2293 		}
2294 
2295 		l = path->nodes[0];
2296 		slot = path->slots[0];
2297 		btrfs_item_key_to_cpu(l, &key, slot);
2298 
2299 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2300 		len = btrfs_inode_ref_name_len(l, iref);
2301 		ptr -= len + 1;
2302 		total_len += len + 1;
2303 		if (ptr < name) {
2304 			ret = -ENAMETOOLONG;
2305 			goto out;
2306 		}
2307 
2308 		*(ptr + len) = '/';
2309 		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2310 
2311 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2312 			break;
2313 
2314 		btrfs_release_path(path);
2315 		key.objectid = key.offset;
2316 		key.offset = (u64)-1;
2317 		dirid = key.objectid;
2318 	}
2319 	memmove(name, ptr, total_len);
2320 	name[total_len] = '\0';
2321 	ret = 0;
2322 out:
2323 	btrfs_put_root(root);
2324 	btrfs_free_path(path);
2325 	return ret;
2326 }
2327 
2328 static int btrfs_search_path_in_tree_user(struct inode *inode,
2329 				struct btrfs_ioctl_ino_lookup_user_args *args)
2330 {
2331 	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2332 	struct super_block *sb = inode->i_sb;
2333 	struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2334 	u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2335 	u64 dirid = args->dirid;
2336 	unsigned long item_off;
2337 	unsigned long item_len;
2338 	struct btrfs_inode_ref *iref;
2339 	struct btrfs_root_ref *rref;
2340 	struct btrfs_root *root = NULL;
2341 	struct btrfs_path *path;
2342 	struct btrfs_key key, key2;
2343 	struct extent_buffer *leaf;
2344 	struct inode *temp_inode;
2345 	char *ptr;
2346 	int slot;
2347 	int len;
2348 	int total_len = 0;
2349 	int ret;
2350 
2351 	path = btrfs_alloc_path();
2352 	if (!path)
2353 		return -ENOMEM;
2354 
2355 	/*
2356 	 * If the bottom subvolume does not exist directly under upper_limit,
2357 	 * construct the path in from the bottom up.
2358 	 */
2359 	if (dirid != upper_limit.objectid) {
2360 		ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2361 
2362 		key.objectid = treeid;
2363 		key.type = BTRFS_ROOT_ITEM_KEY;
2364 		key.offset = (u64)-1;
2365 		root = btrfs_get_fs_root(fs_info, &key, true);
2366 		if (IS_ERR(root)) {
2367 			ret = PTR_ERR(root);
2368 			goto out;
2369 		}
2370 
2371 		key.objectid = dirid;
2372 		key.type = BTRFS_INODE_REF_KEY;
2373 		key.offset = (u64)-1;
2374 		while (1) {
2375 			ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2376 			if (ret < 0) {
2377 				goto out_put;
2378 			} else if (ret > 0) {
2379 				ret = btrfs_previous_item(root, path, dirid,
2380 							  BTRFS_INODE_REF_KEY);
2381 				if (ret < 0) {
2382 					goto out_put;
2383 				} else if (ret > 0) {
2384 					ret = -ENOENT;
2385 					goto out_put;
2386 				}
2387 			}
2388 
2389 			leaf = path->nodes[0];
2390 			slot = path->slots[0];
2391 			btrfs_item_key_to_cpu(leaf, &key, slot);
2392 
2393 			iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2394 			len = btrfs_inode_ref_name_len(leaf, iref);
2395 			ptr -= len + 1;
2396 			total_len += len + 1;
2397 			if (ptr < args->path) {
2398 				ret = -ENAMETOOLONG;
2399 				goto out_put;
2400 			}
2401 
2402 			*(ptr + len) = '/';
2403 			read_extent_buffer(leaf, ptr,
2404 					(unsigned long)(iref + 1), len);
2405 
2406 			/* Check the read+exec permission of this directory */
2407 			ret = btrfs_previous_item(root, path, dirid,
2408 						  BTRFS_INODE_ITEM_KEY);
2409 			if (ret < 0) {
2410 				goto out_put;
2411 			} else if (ret > 0) {
2412 				ret = -ENOENT;
2413 				goto out_put;
2414 			}
2415 
2416 			leaf = path->nodes[0];
2417 			slot = path->slots[0];
2418 			btrfs_item_key_to_cpu(leaf, &key2, slot);
2419 			if (key2.objectid != dirid) {
2420 				ret = -ENOENT;
2421 				goto out_put;
2422 			}
2423 
2424 			temp_inode = btrfs_iget(sb, &key2, root);
2425 			if (IS_ERR(temp_inode)) {
2426 				ret = PTR_ERR(temp_inode);
2427 				goto out_put;
2428 			}
2429 			ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2430 			iput(temp_inode);
2431 			if (ret) {
2432 				ret = -EACCES;
2433 				goto out_put;
2434 			}
2435 
2436 			if (key.offset == upper_limit.objectid)
2437 				break;
2438 			if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2439 				ret = -EACCES;
2440 				goto out_put;
2441 			}
2442 
2443 			btrfs_release_path(path);
2444 			key.objectid = key.offset;
2445 			key.offset = (u64)-1;
2446 			dirid = key.objectid;
2447 		}
2448 
2449 		memmove(args->path, ptr, total_len);
2450 		args->path[total_len] = '\0';
2451 		btrfs_put_root(root);
2452 		root = NULL;
2453 		btrfs_release_path(path);
2454 	}
2455 
2456 	/* Get the bottom subvolume's name from ROOT_REF */
2457 	key.objectid = treeid;
2458 	key.type = BTRFS_ROOT_REF_KEY;
2459 	key.offset = args->treeid;
2460 	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2461 	if (ret < 0) {
2462 		goto out;
2463 	} else if (ret > 0) {
2464 		ret = -ENOENT;
2465 		goto out;
2466 	}
2467 
2468 	leaf = path->nodes[0];
2469 	slot = path->slots[0];
2470 	btrfs_item_key_to_cpu(leaf, &key, slot);
2471 
2472 	item_off = btrfs_item_ptr_offset(leaf, slot);
2473 	item_len = btrfs_item_size_nr(leaf, slot);
2474 	/* Check if dirid in ROOT_REF corresponds to passed dirid */
2475 	rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2476 	if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2477 		ret = -EINVAL;
2478 		goto out;
2479 	}
2480 
2481 	/* Copy subvolume's name */
2482 	item_off += sizeof(struct btrfs_root_ref);
2483 	item_len -= sizeof(struct btrfs_root_ref);
2484 	read_extent_buffer(leaf, args->name, item_off, item_len);
2485 	args->name[item_len] = 0;
2486 
2487 out_put:
2488 	btrfs_put_root(root);
2489 out:
2490 	btrfs_free_path(path);
2491 	return ret;
2492 }
2493 
2494 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2495 					   void __user *argp)
2496 {
2497 	struct btrfs_ioctl_ino_lookup_args *args;
2498 	struct inode *inode;
2499 	int ret = 0;
2500 
2501 	args = memdup_user(argp, sizeof(*args));
2502 	if (IS_ERR(args))
2503 		return PTR_ERR(args);
2504 
2505 	inode = file_inode(file);
2506 
2507 	/*
2508 	 * Unprivileged query to obtain the containing subvolume root id. The
2509 	 * path is reset so it's consistent with btrfs_search_path_in_tree.
2510 	 */
2511 	if (args->treeid == 0)
2512 		args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2513 
2514 	if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2515 		args->name[0] = 0;
2516 		goto out;
2517 	}
2518 
2519 	if (!capable(CAP_SYS_ADMIN)) {
2520 		ret = -EPERM;
2521 		goto out;
2522 	}
2523 
2524 	ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2525 					args->treeid, args->objectid,
2526 					args->name);
2527 
2528 out:
2529 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2530 		ret = -EFAULT;
2531 
2532 	kfree(args);
2533 	return ret;
2534 }
2535 
2536 /*
2537  * Version of ino_lookup ioctl (unprivileged)
2538  *
2539  * The main differences from ino_lookup ioctl are:
2540  *
2541  *   1. Read + Exec permission will be checked using inode_permission() during
2542  *      path construction. -EACCES will be returned in case of failure.
2543  *   2. Path construction will be stopped at the inode number which corresponds
2544  *      to the fd with which this ioctl is called. If constructed path does not
2545  *      exist under fd's inode, -EACCES will be returned.
2546  *   3. The name of bottom subvolume is also searched and filled.
2547  */
2548 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2549 {
2550 	struct btrfs_ioctl_ino_lookup_user_args *args;
2551 	struct inode *inode;
2552 	int ret;
2553 
2554 	args = memdup_user(argp, sizeof(*args));
2555 	if (IS_ERR(args))
2556 		return PTR_ERR(args);
2557 
2558 	inode = file_inode(file);
2559 
2560 	if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2561 	    BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2562 		/*
2563 		 * The subvolume does not exist under fd with which this is
2564 		 * called
2565 		 */
2566 		kfree(args);
2567 		return -EACCES;
2568 	}
2569 
2570 	ret = btrfs_search_path_in_tree_user(inode, args);
2571 
2572 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2573 		ret = -EFAULT;
2574 
2575 	kfree(args);
2576 	return ret;
2577 }
2578 
2579 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2580 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2581 {
2582 	struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2583 	struct btrfs_fs_info *fs_info;
2584 	struct btrfs_root *root;
2585 	struct btrfs_path *path;
2586 	struct btrfs_key key;
2587 	struct btrfs_root_item *root_item;
2588 	struct btrfs_root_ref *rref;
2589 	struct extent_buffer *leaf;
2590 	unsigned long item_off;
2591 	unsigned long item_len;
2592 	struct inode *inode;
2593 	int slot;
2594 	int ret = 0;
2595 
2596 	path = btrfs_alloc_path();
2597 	if (!path)
2598 		return -ENOMEM;
2599 
2600 	subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2601 	if (!subvol_info) {
2602 		btrfs_free_path(path);
2603 		return -ENOMEM;
2604 	}
2605 
2606 	inode = file_inode(file);
2607 	fs_info = BTRFS_I(inode)->root->fs_info;
2608 
2609 	/* Get root_item of inode's subvolume */
2610 	key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2611 	key.type = BTRFS_ROOT_ITEM_KEY;
2612 	key.offset = (u64)-1;
2613 	root = btrfs_get_fs_root(fs_info, &key, true);
2614 	if (IS_ERR(root)) {
2615 		ret = PTR_ERR(root);
2616 		goto out_free;
2617 	}
2618 	root_item = &root->root_item;
2619 
2620 	subvol_info->treeid = key.objectid;
2621 
2622 	subvol_info->generation = btrfs_root_generation(root_item);
2623 	subvol_info->flags = btrfs_root_flags(root_item);
2624 
2625 	memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2626 	memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2627 						    BTRFS_UUID_SIZE);
2628 	memcpy(subvol_info->received_uuid, root_item->received_uuid,
2629 						    BTRFS_UUID_SIZE);
2630 
2631 	subvol_info->ctransid = btrfs_root_ctransid(root_item);
2632 	subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2633 	subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2634 
2635 	subvol_info->otransid = btrfs_root_otransid(root_item);
2636 	subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2637 	subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2638 
2639 	subvol_info->stransid = btrfs_root_stransid(root_item);
2640 	subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2641 	subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2642 
2643 	subvol_info->rtransid = btrfs_root_rtransid(root_item);
2644 	subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2645 	subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2646 
2647 	if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2648 		/* Search root tree for ROOT_BACKREF of this subvolume */
2649 		key.type = BTRFS_ROOT_BACKREF_KEY;
2650 		key.offset = 0;
2651 		ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2652 		if (ret < 0) {
2653 			goto out;
2654 		} else if (path->slots[0] >=
2655 			   btrfs_header_nritems(path->nodes[0])) {
2656 			ret = btrfs_next_leaf(fs_info->tree_root, path);
2657 			if (ret < 0) {
2658 				goto out;
2659 			} else if (ret > 0) {
2660 				ret = -EUCLEAN;
2661 				goto out;
2662 			}
2663 		}
2664 
2665 		leaf = path->nodes[0];
2666 		slot = path->slots[0];
2667 		btrfs_item_key_to_cpu(leaf, &key, slot);
2668 		if (key.objectid == subvol_info->treeid &&
2669 		    key.type == BTRFS_ROOT_BACKREF_KEY) {
2670 			subvol_info->parent_id = key.offset;
2671 
2672 			rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2673 			subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2674 
2675 			item_off = btrfs_item_ptr_offset(leaf, slot)
2676 					+ sizeof(struct btrfs_root_ref);
2677 			item_len = btrfs_item_size_nr(leaf, slot)
2678 					- sizeof(struct btrfs_root_ref);
2679 			read_extent_buffer(leaf, subvol_info->name,
2680 					   item_off, item_len);
2681 		} else {
2682 			ret = -ENOENT;
2683 			goto out;
2684 		}
2685 	}
2686 
2687 	if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2688 		ret = -EFAULT;
2689 
2690 out:
2691 	btrfs_put_root(root);
2692 out_free:
2693 	btrfs_free_path(path);
2694 	kzfree(subvol_info);
2695 	return ret;
2696 }
2697 
2698 /*
2699  * Return ROOT_REF information of the subvolume containing this inode
2700  * except the subvolume name.
2701  */
2702 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2703 {
2704 	struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2705 	struct btrfs_root_ref *rref;
2706 	struct btrfs_root *root;
2707 	struct btrfs_path *path;
2708 	struct btrfs_key key;
2709 	struct extent_buffer *leaf;
2710 	struct inode *inode;
2711 	u64 objectid;
2712 	int slot;
2713 	int ret;
2714 	u8 found;
2715 
2716 	path = btrfs_alloc_path();
2717 	if (!path)
2718 		return -ENOMEM;
2719 
2720 	rootrefs = memdup_user(argp, sizeof(*rootrefs));
2721 	if (IS_ERR(rootrefs)) {
2722 		btrfs_free_path(path);
2723 		return PTR_ERR(rootrefs);
2724 	}
2725 
2726 	inode = file_inode(file);
2727 	root = BTRFS_I(inode)->root->fs_info->tree_root;
2728 	objectid = BTRFS_I(inode)->root->root_key.objectid;
2729 
2730 	key.objectid = objectid;
2731 	key.type = BTRFS_ROOT_REF_KEY;
2732 	key.offset = rootrefs->min_treeid;
2733 	found = 0;
2734 
2735 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2736 	if (ret < 0) {
2737 		goto out;
2738 	} else if (path->slots[0] >=
2739 		   btrfs_header_nritems(path->nodes[0])) {
2740 		ret = btrfs_next_leaf(root, path);
2741 		if (ret < 0) {
2742 			goto out;
2743 		} else if (ret > 0) {
2744 			ret = -EUCLEAN;
2745 			goto out;
2746 		}
2747 	}
2748 	while (1) {
2749 		leaf = path->nodes[0];
2750 		slot = path->slots[0];
2751 
2752 		btrfs_item_key_to_cpu(leaf, &key, slot);
2753 		if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2754 			ret = 0;
2755 			goto out;
2756 		}
2757 
2758 		if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2759 			ret = -EOVERFLOW;
2760 			goto out;
2761 		}
2762 
2763 		rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2764 		rootrefs->rootref[found].treeid = key.offset;
2765 		rootrefs->rootref[found].dirid =
2766 				  btrfs_root_ref_dirid(leaf, rref);
2767 		found++;
2768 
2769 		ret = btrfs_next_item(root, path);
2770 		if (ret < 0) {
2771 			goto out;
2772 		} else if (ret > 0) {
2773 			ret = -EUCLEAN;
2774 			goto out;
2775 		}
2776 	}
2777 
2778 out:
2779 	if (!ret || ret == -EOVERFLOW) {
2780 		rootrefs->num_items = found;
2781 		/* update min_treeid for next search */
2782 		if (found)
2783 			rootrefs->min_treeid =
2784 				rootrefs->rootref[found - 1].treeid + 1;
2785 		if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2786 			ret = -EFAULT;
2787 	}
2788 
2789 	kfree(rootrefs);
2790 	btrfs_free_path(path);
2791 
2792 	return ret;
2793 }
2794 
2795 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2796 					     void __user *arg,
2797 					     bool destroy_v2)
2798 {
2799 	struct dentry *parent = file->f_path.dentry;
2800 	struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2801 	struct dentry *dentry;
2802 	struct inode *dir = d_inode(parent);
2803 	struct inode *inode;
2804 	struct btrfs_root *root = BTRFS_I(dir)->root;
2805 	struct btrfs_root *dest = NULL;
2806 	struct btrfs_ioctl_vol_args *vol_args = NULL;
2807 	struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2808 	char *subvol_name, *subvol_name_ptr = NULL;
2809 	int subvol_namelen;
2810 	int err = 0;
2811 	bool destroy_parent = false;
2812 
2813 	if (destroy_v2) {
2814 		vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2815 		if (IS_ERR(vol_args2))
2816 			return PTR_ERR(vol_args2);
2817 
2818 		if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2819 			err = -EOPNOTSUPP;
2820 			goto out;
2821 		}
2822 
2823 		/*
2824 		 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2825 		 * name, same as v1 currently does.
2826 		 */
2827 		if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2828 			vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2829 			subvol_name = vol_args2->name;
2830 
2831 			err = mnt_want_write_file(file);
2832 			if (err)
2833 				goto out;
2834 		} else {
2835 			if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2836 				err = -EINVAL;
2837 				goto out;
2838 			}
2839 
2840 			err = mnt_want_write_file(file);
2841 			if (err)
2842 				goto out;
2843 
2844 			dentry = btrfs_get_dentry(fs_info->sb,
2845 					BTRFS_FIRST_FREE_OBJECTID,
2846 					vol_args2->subvolid, 0, 0);
2847 			if (IS_ERR(dentry)) {
2848 				err = PTR_ERR(dentry);
2849 				goto out_drop_write;
2850 			}
2851 
2852 			/*
2853 			 * Change the default parent since the subvolume being
2854 			 * deleted can be outside of the current mount point.
2855 			 */
2856 			parent = btrfs_get_parent(dentry);
2857 
2858 			/*
2859 			 * At this point dentry->d_name can point to '/' if the
2860 			 * subvolume we want to destroy is outsite of the
2861 			 * current mount point, so we need to release the
2862 			 * current dentry and execute the lookup to return a new
2863 			 * one with ->d_name pointing to the
2864 			 * <mount point>/subvol_name.
2865 			 */
2866 			dput(dentry);
2867 			if (IS_ERR(parent)) {
2868 				err = PTR_ERR(parent);
2869 				goto out_drop_write;
2870 			}
2871 			dir = d_inode(parent);
2872 
2873 			/*
2874 			 * If v2 was used with SPEC_BY_ID, a new parent was
2875 			 * allocated since the subvolume can be outside of the
2876 			 * current mount point. Later on we need to release this
2877 			 * new parent dentry.
2878 			 */
2879 			destroy_parent = true;
2880 
2881 			subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2882 						fs_info, vol_args2->subvolid);
2883 			if (IS_ERR(subvol_name_ptr)) {
2884 				err = PTR_ERR(subvol_name_ptr);
2885 				goto free_parent;
2886 			}
2887 			/* subvol_name_ptr is already NULL termined */
2888 			subvol_name = (char *)kbasename(subvol_name_ptr);
2889 		}
2890 	} else {
2891 		vol_args = memdup_user(arg, sizeof(*vol_args));
2892 		if (IS_ERR(vol_args))
2893 			return PTR_ERR(vol_args);
2894 
2895 		vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2896 		subvol_name = vol_args->name;
2897 
2898 		err = mnt_want_write_file(file);
2899 		if (err)
2900 			goto out;
2901 	}
2902 
2903 	subvol_namelen = strlen(subvol_name);
2904 
2905 	if (strchr(subvol_name, '/') ||
2906 	    strncmp(subvol_name, "..", subvol_namelen) == 0) {
2907 		err = -EINVAL;
2908 		goto free_subvol_name;
2909 	}
2910 
2911 	if (!S_ISDIR(dir->i_mode)) {
2912 		err = -ENOTDIR;
2913 		goto free_subvol_name;
2914 	}
2915 
2916 	err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2917 	if (err == -EINTR)
2918 		goto free_subvol_name;
2919 	dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
2920 	if (IS_ERR(dentry)) {
2921 		err = PTR_ERR(dentry);
2922 		goto out_unlock_dir;
2923 	}
2924 
2925 	if (d_really_is_negative(dentry)) {
2926 		err = -ENOENT;
2927 		goto out_dput;
2928 	}
2929 
2930 	inode = d_inode(dentry);
2931 	dest = BTRFS_I(inode)->root;
2932 	if (!capable(CAP_SYS_ADMIN)) {
2933 		/*
2934 		 * Regular user.  Only allow this with a special mount
2935 		 * option, when the user has write+exec access to the
2936 		 * subvol root, and when rmdir(2) would have been
2937 		 * allowed.
2938 		 *
2939 		 * Note that this is _not_ check that the subvol is
2940 		 * empty or doesn't contain data that we wouldn't
2941 		 * otherwise be able to delete.
2942 		 *
2943 		 * Users who want to delete empty subvols should try
2944 		 * rmdir(2).
2945 		 */
2946 		err = -EPERM;
2947 		if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2948 			goto out_dput;
2949 
2950 		/*
2951 		 * Do not allow deletion if the parent dir is the same
2952 		 * as the dir to be deleted.  That means the ioctl
2953 		 * must be called on the dentry referencing the root
2954 		 * of the subvol, not a random directory contained
2955 		 * within it.
2956 		 */
2957 		err = -EINVAL;
2958 		if (root == dest)
2959 			goto out_dput;
2960 
2961 		err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2962 		if (err)
2963 			goto out_dput;
2964 	}
2965 
2966 	/* check if subvolume may be deleted by a user */
2967 	err = btrfs_may_delete(dir, dentry, 1);
2968 	if (err)
2969 		goto out_dput;
2970 
2971 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2972 		err = -EINVAL;
2973 		goto out_dput;
2974 	}
2975 
2976 	inode_lock(inode);
2977 	err = btrfs_delete_subvolume(dir, dentry);
2978 	inode_unlock(inode);
2979 	if (!err) {
2980 		fsnotify_rmdir(dir, dentry);
2981 		d_delete(dentry);
2982 	}
2983 
2984 out_dput:
2985 	dput(dentry);
2986 out_unlock_dir:
2987 	inode_unlock(dir);
2988 free_subvol_name:
2989 	kfree(subvol_name_ptr);
2990 free_parent:
2991 	if (destroy_parent)
2992 		dput(parent);
2993 out_drop_write:
2994 	mnt_drop_write_file(file);
2995 out:
2996 	kfree(vol_args2);
2997 	kfree(vol_args);
2998 	return err;
2999 }
3000 
3001 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3002 {
3003 	struct inode *inode = file_inode(file);
3004 	struct btrfs_root *root = BTRFS_I(inode)->root;
3005 	struct btrfs_ioctl_defrag_range_args *range;
3006 	int ret;
3007 
3008 	ret = mnt_want_write_file(file);
3009 	if (ret)
3010 		return ret;
3011 
3012 	if (btrfs_root_readonly(root)) {
3013 		ret = -EROFS;
3014 		goto out;
3015 	}
3016 
3017 	switch (inode->i_mode & S_IFMT) {
3018 	case S_IFDIR:
3019 		if (!capable(CAP_SYS_ADMIN)) {
3020 			ret = -EPERM;
3021 			goto out;
3022 		}
3023 		ret = btrfs_defrag_root(root);
3024 		break;
3025 	case S_IFREG:
3026 		/*
3027 		 * Note that this does not check the file descriptor for write
3028 		 * access. This prevents defragmenting executables that are
3029 		 * running and allows defrag on files open in read-only mode.
3030 		 */
3031 		if (!capable(CAP_SYS_ADMIN) &&
3032 		    inode_permission(inode, MAY_WRITE)) {
3033 			ret = -EPERM;
3034 			goto out;
3035 		}
3036 
3037 		range = kzalloc(sizeof(*range), GFP_KERNEL);
3038 		if (!range) {
3039 			ret = -ENOMEM;
3040 			goto out;
3041 		}
3042 
3043 		if (argp) {
3044 			if (copy_from_user(range, argp,
3045 					   sizeof(*range))) {
3046 				ret = -EFAULT;
3047 				kfree(range);
3048 				goto out;
3049 			}
3050 			/* compression requires us to start the IO */
3051 			if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3052 				range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3053 				range->extent_thresh = (u32)-1;
3054 			}
3055 		} else {
3056 			/* the rest are all set to zero by kzalloc */
3057 			range->len = (u64)-1;
3058 		}
3059 		ret = btrfs_defrag_file(file_inode(file), file,
3060 					range, BTRFS_OLDEST_GENERATION, 0);
3061 		if (ret > 0)
3062 			ret = 0;
3063 		kfree(range);
3064 		break;
3065 	default:
3066 		ret = -EINVAL;
3067 	}
3068 out:
3069 	mnt_drop_write_file(file);
3070 	return ret;
3071 }
3072 
3073 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3074 {
3075 	struct btrfs_ioctl_vol_args *vol_args;
3076 	int ret;
3077 
3078 	if (!capable(CAP_SYS_ADMIN))
3079 		return -EPERM;
3080 
3081 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
3082 		return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3083 
3084 	vol_args = memdup_user(arg, sizeof(*vol_args));
3085 	if (IS_ERR(vol_args)) {
3086 		ret = PTR_ERR(vol_args);
3087 		goto out;
3088 	}
3089 
3090 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3091 	ret = btrfs_init_new_device(fs_info, vol_args->name);
3092 
3093 	if (!ret)
3094 		btrfs_info(fs_info, "disk added %s", vol_args->name);
3095 
3096 	kfree(vol_args);
3097 out:
3098 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3099 	return ret;
3100 }
3101 
3102 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3103 {
3104 	struct inode *inode = file_inode(file);
3105 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3106 	struct btrfs_ioctl_vol_args_v2 *vol_args;
3107 	int ret;
3108 
3109 	if (!capable(CAP_SYS_ADMIN))
3110 		return -EPERM;
3111 
3112 	ret = mnt_want_write_file(file);
3113 	if (ret)
3114 		return ret;
3115 
3116 	vol_args = memdup_user(arg, sizeof(*vol_args));
3117 	if (IS_ERR(vol_args)) {
3118 		ret = PTR_ERR(vol_args);
3119 		goto err_drop;
3120 	}
3121 
3122 	if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3123 		ret = -EOPNOTSUPP;
3124 		goto out;
3125 	}
3126 
3127 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3128 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3129 		goto out;
3130 	}
3131 
3132 	if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3133 		ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3134 	} else {
3135 		vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3136 		ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3137 	}
3138 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3139 
3140 	if (!ret) {
3141 		if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3142 			btrfs_info(fs_info, "device deleted: id %llu",
3143 					vol_args->devid);
3144 		else
3145 			btrfs_info(fs_info, "device deleted: %s",
3146 					vol_args->name);
3147 	}
3148 out:
3149 	kfree(vol_args);
3150 err_drop:
3151 	mnt_drop_write_file(file);
3152 	return ret;
3153 }
3154 
3155 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3156 {
3157 	struct inode *inode = file_inode(file);
3158 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3159 	struct btrfs_ioctl_vol_args *vol_args;
3160 	int ret;
3161 
3162 	if (!capable(CAP_SYS_ADMIN))
3163 		return -EPERM;
3164 
3165 	ret = mnt_want_write_file(file);
3166 	if (ret)
3167 		return ret;
3168 
3169 	if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3170 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3171 		goto out_drop_write;
3172 	}
3173 
3174 	vol_args = memdup_user(arg, sizeof(*vol_args));
3175 	if (IS_ERR(vol_args)) {
3176 		ret = PTR_ERR(vol_args);
3177 		goto out;
3178 	}
3179 
3180 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3181 	ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3182 
3183 	if (!ret)
3184 		btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3185 	kfree(vol_args);
3186 out:
3187 	clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3188 out_drop_write:
3189 	mnt_drop_write_file(file);
3190 
3191 	return ret;
3192 }
3193 
3194 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3195 				void __user *arg)
3196 {
3197 	struct btrfs_ioctl_fs_info_args *fi_args;
3198 	struct btrfs_device *device;
3199 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3200 	int ret = 0;
3201 
3202 	fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
3203 	if (!fi_args)
3204 		return -ENOMEM;
3205 
3206 	rcu_read_lock();
3207 	fi_args->num_devices = fs_devices->num_devices;
3208 
3209 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3210 		if (device->devid > fi_args->max_id)
3211 			fi_args->max_id = device->devid;
3212 	}
3213 	rcu_read_unlock();
3214 
3215 	memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3216 	fi_args->nodesize = fs_info->nodesize;
3217 	fi_args->sectorsize = fs_info->sectorsize;
3218 	fi_args->clone_alignment = fs_info->sectorsize;
3219 
3220 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3221 		ret = -EFAULT;
3222 
3223 	kfree(fi_args);
3224 	return ret;
3225 }
3226 
3227 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3228 				 void __user *arg)
3229 {
3230 	struct btrfs_ioctl_dev_info_args *di_args;
3231 	struct btrfs_device *dev;
3232 	int ret = 0;
3233 	char *s_uuid = NULL;
3234 
3235 	di_args = memdup_user(arg, sizeof(*di_args));
3236 	if (IS_ERR(di_args))
3237 		return PTR_ERR(di_args);
3238 
3239 	if (!btrfs_is_empty_uuid(di_args->uuid))
3240 		s_uuid = di_args->uuid;
3241 
3242 	rcu_read_lock();
3243 	dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3244 				NULL, true);
3245 
3246 	if (!dev) {
3247 		ret = -ENODEV;
3248 		goto out;
3249 	}
3250 
3251 	di_args->devid = dev->devid;
3252 	di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3253 	di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3254 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3255 	if (dev->name) {
3256 		strncpy(di_args->path, rcu_str_deref(dev->name),
3257 				sizeof(di_args->path) - 1);
3258 		di_args->path[sizeof(di_args->path) - 1] = 0;
3259 	} else {
3260 		di_args->path[0] = '\0';
3261 	}
3262 
3263 out:
3264 	rcu_read_unlock();
3265 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3266 		ret = -EFAULT;
3267 
3268 	kfree(di_args);
3269 	return ret;
3270 }
3271 
3272 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3273 {
3274 	struct inode *inode = file_inode(file);
3275 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3276 	struct btrfs_root *root = BTRFS_I(inode)->root;
3277 	struct btrfs_root *new_root;
3278 	struct btrfs_dir_item *di;
3279 	struct btrfs_trans_handle *trans;
3280 	struct btrfs_path *path = NULL;
3281 	struct btrfs_key location;
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 	location.objectid = objectid;
3303 	location.type = BTRFS_ROOT_ITEM_KEY;
3304 	location.offset = (u64)-1;
3305 
3306 	new_root = btrfs_get_fs_root(fs_info, &location, true);
3307 	if (IS_ERR(new_root)) {
3308 		ret = PTR_ERR(new_root);
3309 		goto out;
3310 	}
3311 	if (!is_fstree(new_root->root_key.objectid)) {
3312 		ret = -ENOENT;
3313 		goto out_free;
3314 	}
3315 
3316 	path = btrfs_alloc_path();
3317 	if (!path) {
3318 		ret = -ENOMEM;
3319 		goto out_free;
3320 	}
3321 	path->leave_spinning = 1;
3322 
3323 	trans = btrfs_start_transaction(root, 1);
3324 	if (IS_ERR(trans)) {
3325 		ret = PTR_ERR(trans);
3326 		goto out_free;
3327 	}
3328 
3329 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
3330 	di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3331 				   dir_id, "default", 7, 1);
3332 	if (IS_ERR_OR_NULL(di)) {
3333 		btrfs_release_path(path);
3334 		btrfs_end_transaction(trans);
3335 		btrfs_err(fs_info,
3336 			  "Umm, you don't have the default diritem, this isn't going to work");
3337 		ret = -ENOENT;
3338 		goto out_free;
3339 	}
3340 
3341 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3342 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3343 	btrfs_mark_buffer_dirty(path->nodes[0]);
3344 	btrfs_release_path(path);
3345 
3346 	btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3347 	btrfs_end_transaction(trans);
3348 out_free:
3349 	btrfs_put_root(new_root);
3350 	btrfs_free_path(path);
3351 out:
3352 	mnt_drop_write_file(file);
3353 	return ret;
3354 }
3355 
3356 static void get_block_group_info(struct list_head *groups_list,
3357 				 struct btrfs_ioctl_space_info *space)
3358 {
3359 	struct btrfs_block_group *block_group;
3360 
3361 	space->total_bytes = 0;
3362 	space->used_bytes = 0;
3363 	space->flags = 0;
3364 	list_for_each_entry(block_group, groups_list, list) {
3365 		space->flags = block_group->flags;
3366 		space->total_bytes += block_group->length;
3367 		space->used_bytes += block_group->used;
3368 	}
3369 }
3370 
3371 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3372 				   void __user *arg)
3373 {
3374 	struct btrfs_ioctl_space_args space_args;
3375 	struct btrfs_ioctl_space_info space;
3376 	struct btrfs_ioctl_space_info *dest;
3377 	struct btrfs_ioctl_space_info *dest_orig;
3378 	struct btrfs_ioctl_space_info __user *user_dest;
3379 	struct btrfs_space_info *info;
3380 	static const u64 types[] = {
3381 		BTRFS_BLOCK_GROUP_DATA,
3382 		BTRFS_BLOCK_GROUP_SYSTEM,
3383 		BTRFS_BLOCK_GROUP_METADATA,
3384 		BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3385 	};
3386 	int num_types = 4;
3387 	int alloc_size;
3388 	int ret = 0;
3389 	u64 slot_count = 0;
3390 	int i, c;
3391 
3392 	if (copy_from_user(&space_args,
3393 			   (struct btrfs_ioctl_space_args __user *)arg,
3394 			   sizeof(space_args)))
3395 		return -EFAULT;
3396 
3397 	for (i = 0; i < num_types; i++) {
3398 		struct btrfs_space_info *tmp;
3399 
3400 		info = NULL;
3401 		rcu_read_lock();
3402 		list_for_each_entry_rcu(tmp, &fs_info->space_info,
3403 					list) {
3404 			if (tmp->flags == types[i]) {
3405 				info = tmp;
3406 				break;
3407 			}
3408 		}
3409 		rcu_read_unlock();
3410 
3411 		if (!info)
3412 			continue;
3413 
3414 		down_read(&info->groups_sem);
3415 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3416 			if (!list_empty(&info->block_groups[c]))
3417 				slot_count++;
3418 		}
3419 		up_read(&info->groups_sem);
3420 	}
3421 
3422 	/*
3423 	 * Global block reserve, exported as a space_info
3424 	 */
3425 	slot_count++;
3426 
3427 	/* space_slots == 0 means they are asking for a count */
3428 	if (space_args.space_slots == 0) {
3429 		space_args.total_spaces = slot_count;
3430 		goto out;
3431 	}
3432 
3433 	slot_count = min_t(u64, space_args.space_slots, slot_count);
3434 
3435 	alloc_size = sizeof(*dest) * slot_count;
3436 
3437 	/* we generally have at most 6 or so space infos, one for each raid
3438 	 * level.  So, a whole page should be more than enough for everyone
3439 	 */
3440 	if (alloc_size > PAGE_SIZE)
3441 		return -ENOMEM;
3442 
3443 	space_args.total_spaces = 0;
3444 	dest = kmalloc(alloc_size, GFP_KERNEL);
3445 	if (!dest)
3446 		return -ENOMEM;
3447 	dest_orig = dest;
3448 
3449 	/* now we have a buffer to copy into */
3450 	for (i = 0; i < num_types; i++) {
3451 		struct btrfs_space_info *tmp;
3452 
3453 		if (!slot_count)
3454 			break;
3455 
3456 		info = NULL;
3457 		rcu_read_lock();
3458 		list_for_each_entry_rcu(tmp, &fs_info->space_info,
3459 					list) {
3460 			if (tmp->flags == types[i]) {
3461 				info = tmp;
3462 				break;
3463 			}
3464 		}
3465 		rcu_read_unlock();
3466 
3467 		if (!info)
3468 			continue;
3469 		down_read(&info->groups_sem);
3470 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3471 			if (!list_empty(&info->block_groups[c])) {
3472 				get_block_group_info(&info->block_groups[c],
3473 						     &space);
3474 				memcpy(dest, &space, sizeof(space));
3475 				dest++;
3476 				space_args.total_spaces++;
3477 				slot_count--;
3478 			}
3479 			if (!slot_count)
3480 				break;
3481 		}
3482 		up_read(&info->groups_sem);
3483 	}
3484 
3485 	/*
3486 	 * Add global block reserve
3487 	 */
3488 	if (slot_count) {
3489 		struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3490 
3491 		spin_lock(&block_rsv->lock);
3492 		space.total_bytes = block_rsv->size;
3493 		space.used_bytes = block_rsv->size - block_rsv->reserved;
3494 		spin_unlock(&block_rsv->lock);
3495 		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3496 		memcpy(dest, &space, sizeof(space));
3497 		space_args.total_spaces++;
3498 	}
3499 
3500 	user_dest = (struct btrfs_ioctl_space_info __user *)
3501 		(arg + sizeof(struct btrfs_ioctl_space_args));
3502 
3503 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3504 		ret = -EFAULT;
3505 
3506 	kfree(dest_orig);
3507 out:
3508 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3509 		ret = -EFAULT;
3510 
3511 	return ret;
3512 }
3513 
3514 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3515 					    void __user *argp)
3516 {
3517 	struct btrfs_trans_handle *trans;
3518 	u64 transid;
3519 	int ret;
3520 
3521 	trans = btrfs_attach_transaction_barrier(root);
3522 	if (IS_ERR(trans)) {
3523 		if (PTR_ERR(trans) != -ENOENT)
3524 			return PTR_ERR(trans);
3525 
3526 		/* No running transaction, don't bother */
3527 		transid = root->fs_info->last_trans_committed;
3528 		goto out;
3529 	}
3530 	transid = trans->transid;
3531 	ret = btrfs_commit_transaction_async(trans, 0);
3532 	if (ret) {
3533 		btrfs_end_transaction(trans);
3534 		return ret;
3535 	}
3536 out:
3537 	if (argp)
3538 		if (copy_to_user(argp, &transid, sizeof(transid)))
3539 			return -EFAULT;
3540 	return 0;
3541 }
3542 
3543 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3544 					   void __user *argp)
3545 {
3546 	u64 transid;
3547 
3548 	if (argp) {
3549 		if (copy_from_user(&transid, argp, sizeof(transid)))
3550 			return -EFAULT;
3551 	} else {
3552 		transid = 0;  /* current trans */
3553 	}
3554 	return btrfs_wait_for_commit(fs_info, transid);
3555 }
3556 
3557 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3558 {
3559 	struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3560 	struct btrfs_ioctl_scrub_args *sa;
3561 	int ret;
3562 
3563 	if (!capable(CAP_SYS_ADMIN))
3564 		return -EPERM;
3565 
3566 	sa = memdup_user(arg, sizeof(*sa));
3567 	if (IS_ERR(sa))
3568 		return PTR_ERR(sa);
3569 
3570 	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3571 		ret = mnt_want_write_file(file);
3572 		if (ret)
3573 			goto out;
3574 	}
3575 
3576 	ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3577 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3578 			      0);
3579 
3580 	/*
3581 	 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3582 	 * error. This is important as it allows user space to know how much
3583 	 * progress scrub has done. For example, if scrub is canceled we get
3584 	 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3585 	 * space. Later user space can inspect the progress from the structure
3586 	 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3587 	 * previously (btrfs-progs does this).
3588 	 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3589 	 * then return -EFAULT to signal the structure was not copied or it may
3590 	 * be corrupt and unreliable due to a partial copy.
3591 	 */
3592 	if (copy_to_user(arg, sa, sizeof(*sa)))
3593 		ret = -EFAULT;
3594 
3595 	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3596 		mnt_drop_write_file(file);
3597 out:
3598 	kfree(sa);
3599 	return ret;
3600 }
3601 
3602 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3603 {
3604 	if (!capable(CAP_SYS_ADMIN))
3605 		return -EPERM;
3606 
3607 	return btrfs_scrub_cancel(fs_info);
3608 }
3609 
3610 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3611 				       void __user *arg)
3612 {
3613 	struct btrfs_ioctl_scrub_args *sa;
3614 	int ret;
3615 
3616 	if (!capable(CAP_SYS_ADMIN))
3617 		return -EPERM;
3618 
3619 	sa = memdup_user(arg, sizeof(*sa));
3620 	if (IS_ERR(sa))
3621 		return PTR_ERR(sa);
3622 
3623 	ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3624 
3625 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3626 		ret = -EFAULT;
3627 
3628 	kfree(sa);
3629 	return ret;
3630 }
3631 
3632 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3633 				      void __user *arg)
3634 {
3635 	struct btrfs_ioctl_get_dev_stats *sa;
3636 	int ret;
3637 
3638 	sa = memdup_user(arg, sizeof(*sa));
3639 	if (IS_ERR(sa))
3640 		return PTR_ERR(sa);
3641 
3642 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3643 		kfree(sa);
3644 		return -EPERM;
3645 	}
3646 
3647 	ret = btrfs_get_dev_stats(fs_info, sa);
3648 
3649 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3650 		ret = -EFAULT;
3651 
3652 	kfree(sa);
3653 	return ret;
3654 }
3655 
3656 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3657 				    void __user *arg)
3658 {
3659 	struct btrfs_ioctl_dev_replace_args *p;
3660 	int ret;
3661 
3662 	if (!capable(CAP_SYS_ADMIN))
3663 		return -EPERM;
3664 
3665 	p = memdup_user(arg, sizeof(*p));
3666 	if (IS_ERR(p))
3667 		return PTR_ERR(p);
3668 
3669 	switch (p->cmd) {
3670 	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3671 		if (sb_rdonly(fs_info->sb)) {
3672 			ret = -EROFS;
3673 			goto out;
3674 		}
3675 		if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3676 			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3677 		} else {
3678 			ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3679 			clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
3680 		}
3681 		break;
3682 	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3683 		btrfs_dev_replace_status(fs_info, p);
3684 		ret = 0;
3685 		break;
3686 	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3687 		p->result = btrfs_dev_replace_cancel(fs_info);
3688 		ret = 0;
3689 		break;
3690 	default:
3691 		ret = -EINVAL;
3692 		break;
3693 	}
3694 
3695 	if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3696 		ret = -EFAULT;
3697 out:
3698 	kfree(p);
3699 	return ret;
3700 }
3701 
3702 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3703 {
3704 	int ret = 0;
3705 	int i;
3706 	u64 rel_ptr;
3707 	int size;
3708 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3709 	struct inode_fs_paths *ipath = NULL;
3710 	struct btrfs_path *path;
3711 
3712 	if (!capable(CAP_DAC_READ_SEARCH))
3713 		return -EPERM;
3714 
3715 	path = btrfs_alloc_path();
3716 	if (!path) {
3717 		ret = -ENOMEM;
3718 		goto out;
3719 	}
3720 
3721 	ipa = memdup_user(arg, sizeof(*ipa));
3722 	if (IS_ERR(ipa)) {
3723 		ret = PTR_ERR(ipa);
3724 		ipa = NULL;
3725 		goto out;
3726 	}
3727 
3728 	size = min_t(u32, ipa->size, 4096);
3729 	ipath = init_ipath(size, root, path);
3730 	if (IS_ERR(ipath)) {
3731 		ret = PTR_ERR(ipath);
3732 		ipath = NULL;
3733 		goto out;
3734 	}
3735 
3736 	ret = paths_from_inode(ipa->inum, ipath);
3737 	if (ret < 0)
3738 		goto out;
3739 
3740 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3741 		rel_ptr = ipath->fspath->val[i] -
3742 			  (u64)(unsigned long)ipath->fspath->val;
3743 		ipath->fspath->val[i] = rel_ptr;
3744 	}
3745 
3746 	ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3747 			   ipath->fspath, size);
3748 	if (ret) {
3749 		ret = -EFAULT;
3750 		goto out;
3751 	}
3752 
3753 out:
3754 	btrfs_free_path(path);
3755 	free_ipath(ipath);
3756 	kfree(ipa);
3757 
3758 	return ret;
3759 }
3760 
3761 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3762 {
3763 	struct btrfs_data_container *inodes = ctx;
3764 	const size_t c = 3 * sizeof(u64);
3765 
3766 	if (inodes->bytes_left >= c) {
3767 		inodes->bytes_left -= c;
3768 		inodes->val[inodes->elem_cnt] = inum;
3769 		inodes->val[inodes->elem_cnt + 1] = offset;
3770 		inodes->val[inodes->elem_cnt + 2] = root;
3771 		inodes->elem_cnt += 3;
3772 	} else {
3773 		inodes->bytes_missing += c - inodes->bytes_left;
3774 		inodes->bytes_left = 0;
3775 		inodes->elem_missed += 3;
3776 	}
3777 
3778 	return 0;
3779 }
3780 
3781 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3782 					void __user *arg, int version)
3783 {
3784 	int ret = 0;
3785 	int size;
3786 	struct btrfs_ioctl_logical_ino_args *loi;
3787 	struct btrfs_data_container *inodes = NULL;
3788 	struct btrfs_path *path = NULL;
3789 	bool ignore_offset;
3790 
3791 	if (!capable(CAP_SYS_ADMIN))
3792 		return -EPERM;
3793 
3794 	loi = memdup_user(arg, sizeof(*loi));
3795 	if (IS_ERR(loi))
3796 		return PTR_ERR(loi);
3797 
3798 	if (version == 1) {
3799 		ignore_offset = false;
3800 		size = min_t(u32, loi->size, SZ_64K);
3801 	} else {
3802 		/* All reserved bits must be 0 for now */
3803 		if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3804 			ret = -EINVAL;
3805 			goto out_loi;
3806 		}
3807 		/* Only accept flags we have defined so far */
3808 		if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3809 			ret = -EINVAL;
3810 			goto out_loi;
3811 		}
3812 		ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3813 		size = min_t(u32, loi->size, SZ_16M);
3814 	}
3815 
3816 	path = btrfs_alloc_path();
3817 	if (!path) {
3818 		ret = -ENOMEM;
3819 		goto out;
3820 	}
3821 
3822 	inodes = init_data_container(size);
3823 	if (IS_ERR(inodes)) {
3824 		ret = PTR_ERR(inodes);
3825 		inodes = NULL;
3826 		goto out;
3827 	}
3828 
3829 	ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3830 					  build_ino_list, inodes, ignore_offset);
3831 	if (ret == -EINVAL)
3832 		ret = -ENOENT;
3833 	if (ret < 0)
3834 		goto out;
3835 
3836 	ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3837 			   size);
3838 	if (ret)
3839 		ret = -EFAULT;
3840 
3841 out:
3842 	btrfs_free_path(path);
3843 	kvfree(inodes);
3844 out_loi:
3845 	kfree(loi);
3846 
3847 	return ret;
3848 }
3849 
3850 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3851 			       struct btrfs_ioctl_balance_args *bargs)
3852 {
3853 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3854 
3855 	bargs->flags = bctl->flags;
3856 
3857 	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3858 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3859 	if (atomic_read(&fs_info->balance_pause_req))
3860 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3861 	if (atomic_read(&fs_info->balance_cancel_req))
3862 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3863 
3864 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3865 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3866 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3867 
3868 	spin_lock(&fs_info->balance_lock);
3869 	memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3870 	spin_unlock(&fs_info->balance_lock);
3871 }
3872 
3873 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3874 {
3875 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3876 	struct btrfs_fs_info *fs_info = root->fs_info;
3877 	struct btrfs_ioctl_balance_args *bargs;
3878 	struct btrfs_balance_control *bctl;
3879 	bool need_unlock; /* for mut. excl. ops lock */
3880 	int ret;
3881 
3882 	if (!capable(CAP_SYS_ADMIN))
3883 		return -EPERM;
3884 
3885 	ret = mnt_want_write_file(file);
3886 	if (ret)
3887 		return ret;
3888 
3889 again:
3890 	if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
3891 		mutex_lock(&fs_info->balance_mutex);
3892 		need_unlock = true;
3893 		goto locked;
3894 	}
3895 
3896 	/*
3897 	 * mut. excl. ops lock is locked.  Three possibilities:
3898 	 *   (1) some other op is running
3899 	 *   (2) balance is running
3900 	 *   (3) balance is paused -- special case (think resume)
3901 	 */
3902 	mutex_lock(&fs_info->balance_mutex);
3903 	if (fs_info->balance_ctl) {
3904 		/* this is either (2) or (3) */
3905 		if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3906 			mutex_unlock(&fs_info->balance_mutex);
3907 			/*
3908 			 * Lock released to allow other waiters to continue,
3909 			 * we'll reexamine the status again.
3910 			 */
3911 			mutex_lock(&fs_info->balance_mutex);
3912 
3913 			if (fs_info->balance_ctl &&
3914 			    !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3915 				/* this is (3) */
3916 				need_unlock = false;
3917 				goto locked;
3918 			}
3919 
3920 			mutex_unlock(&fs_info->balance_mutex);
3921 			goto again;
3922 		} else {
3923 			/* this is (2) */
3924 			mutex_unlock(&fs_info->balance_mutex);
3925 			ret = -EINPROGRESS;
3926 			goto out;
3927 		}
3928 	} else {
3929 		/* this is (1) */
3930 		mutex_unlock(&fs_info->balance_mutex);
3931 		ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3932 		goto out;
3933 	}
3934 
3935 locked:
3936 	BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
3937 
3938 	if (arg) {
3939 		bargs = memdup_user(arg, sizeof(*bargs));
3940 		if (IS_ERR(bargs)) {
3941 			ret = PTR_ERR(bargs);
3942 			goto out_unlock;
3943 		}
3944 
3945 		if (bargs->flags & BTRFS_BALANCE_RESUME) {
3946 			if (!fs_info->balance_ctl) {
3947 				ret = -ENOTCONN;
3948 				goto out_bargs;
3949 			}
3950 
3951 			bctl = fs_info->balance_ctl;
3952 			spin_lock(&fs_info->balance_lock);
3953 			bctl->flags |= BTRFS_BALANCE_RESUME;
3954 			spin_unlock(&fs_info->balance_lock);
3955 
3956 			goto do_balance;
3957 		}
3958 	} else {
3959 		bargs = NULL;
3960 	}
3961 
3962 	if (fs_info->balance_ctl) {
3963 		ret = -EINPROGRESS;
3964 		goto out_bargs;
3965 	}
3966 
3967 	bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3968 	if (!bctl) {
3969 		ret = -ENOMEM;
3970 		goto out_bargs;
3971 	}
3972 
3973 	if (arg) {
3974 		memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3975 		memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3976 		memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3977 
3978 		bctl->flags = bargs->flags;
3979 	} else {
3980 		/* balance everything - no filters */
3981 		bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3982 	}
3983 
3984 	if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3985 		ret = -EINVAL;
3986 		goto out_bctl;
3987 	}
3988 
3989 do_balance:
3990 	/*
3991 	 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
3992 	 * btrfs_balance.  bctl is freed in reset_balance_state, or, if
3993 	 * restriper was paused all the way until unmount, in free_fs_info.
3994 	 * The flag should be cleared after reset_balance_state.
3995 	 */
3996 	need_unlock = false;
3997 
3998 	ret = btrfs_balance(fs_info, bctl, bargs);
3999 	bctl = NULL;
4000 
4001 	if ((ret == 0 || ret == -ECANCELED) && arg) {
4002 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
4003 			ret = -EFAULT;
4004 	}
4005 
4006 out_bctl:
4007 	kfree(bctl);
4008 out_bargs:
4009 	kfree(bargs);
4010 out_unlock:
4011 	mutex_unlock(&fs_info->balance_mutex);
4012 	if (need_unlock)
4013 		clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4014 out:
4015 	mnt_drop_write_file(file);
4016 	return ret;
4017 }
4018 
4019 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4020 {
4021 	if (!capable(CAP_SYS_ADMIN))
4022 		return -EPERM;
4023 
4024 	switch (cmd) {
4025 	case BTRFS_BALANCE_CTL_PAUSE:
4026 		return btrfs_pause_balance(fs_info);
4027 	case BTRFS_BALANCE_CTL_CANCEL:
4028 		return btrfs_cancel_balance(fs_info);
4029 	}
4030 
4031 	return -EINVAL;
4032 }
4033 
4034 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4035 					 void __user *arg)
4036 {
4037 	struct btrfs_ioctl_balance_args *bargs;
4038 	int ret = 0;
4039 
4040 	if (!capable(CAP_SYS_ADMIN))
4041 		return -EPERM;
4042 
4043 	mutex_lock(&fs_info->balance_mutex);
4044 	if (!fs_info->balance_ctl) {
4045 		ret = -ENOTCONN;
4046 		goto out;
4047 	}
4048 
4049 	bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4050 	if (!bargs) {
4051 		ret = -ENOMEM;
4052 		goto out;
4053 	}
4054 
4055 	btrfs_update_ioctl_balance_args(fs_info, bargs);
4056 
4057 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
4058 		ret = -EFAULT;
4059 
4060 	kfree(bargs);
4061 out:
4062 	mutex_unlock(&fs_info->balance_mutex);
4063 	return ret;
4064 }
4065 
4066 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4067 {
4068 	struct inode *inode = file_inode(file);
4069 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4070 	struct btrfs_ioctl_quota_ctl_args *sa;
4071 	int ret;
4072 
4073 	if (!capable(CAP_SYS_ADMIN))
4074 		return -EPERM;
4075 
4076 	ret = mnt_want_write_file(file);
4077 	if (ret)
4078 		return ret;
4079 
4080 	sa = memdup_user(arg, sizeof(*sa));
4081 	if (IS_ERR(sa)) {
4082 		ret = PTR_ERR(sa);
4083 		goto drop_write;
4084 	}
4085 
4086 	down_write(&fs_info->subvol_sem);
4087 
4088 	switch (sa->cmd) {
4089 	case BTRFS_QUOTA_CTL_ENABLE:
4090 		ret = btrfs_quota_enable(fs_info);
4091 		break;
4092 	case BTRFS_QUOTA_CTL_DISABLE:
4093 		ret = btrfs_quota_disable(fs_info);
4094 		break;
4095 	default:
4096 		ret = -EINVAL;
4097 		break;
4098 	}
4099 
4100 	kfree(sa);
4101 	up_write(&fs_info->subvol_sem);
4102 drop_write:
4103 	mnt_drop_write_file(file);
4104 	return ret;
4105 }
4106 
4107 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4108 {
4109 	struct inode *inode = file_inode(file);
4110 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4111 	struct btrfs_root *root = BTRFS_I(inode)->root;
4112 	struct btrfs_ioctl_qgroup_assign_args *sa;
4113 	struct btrfs_trans_handle *trans;
4114 	int ret;
4115 	int err;
4116 
4117 	if (!capable(CAP_SYS_ADMIN))
4118 		return -EPERM;
4119 
4120 	ret = mnt_want_write_file(file);
4121 	if (ret)
4122 		return ret;
4123 
4124 	sa = memdup_user(arg, sizeof(*sa));
4125 	if (IS_ERR(sa)) {
4126 		ret = PTR_ERR(sa);
4127 		goto drop_write;
4128 	}
4129 
4130 	trans = btrfs_join_transaction(root);
4131 	if (IS_ERR(trans)) {
4132 		ret = PTR_ERR(trans);
4133 		goto out;
4134 	}
4135 
4136 	if (sa->assign) {
4137 		ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4138 	} else {
4139 		ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4140 	}
4141 
4142 	/* update qgroup status and info */
4143 	err = btrfs_run_qgroups(trans);
4144 	if (err < 0)
4145 		btrfs_handle_fs_error(fs_info, err,
4146 				      "failed to update qgroup status and info");
4147 	err = btrfs_end_transaction(trans);
4148 	if (err && !ret)
4149 		ret = err;
4150 
4151 out:
4152 	kfree(sa);
4153 drop_write:
4154 	mnt_drop_write_file(file);
4155 	return ret;
4156 }
4157 
4158 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4159 {
4160 	struct inode *inode = file_inode(file);
4161 	struct btrfs_root *root = BTRFS_I(inode)->root;
4162 	struct btrfs_ioctl_qgroup_create_args *sa;
4163 	struct btrfs_trans_handle *trans;
4164 	int ret;
4165 	int err;
4166 
4167 	if (!capable(CAP_SYS_ADMIN))
4168 		return -EPERM;
4169 
4170 	ret = mnt_want_write_file(file);
4171 	if (ret)
4172 		return ret;
4173 
4174 	sa = memdup_user(arg, sizeof(*sa));
4175 	if (IS_ERR(sa)) {
4176 		ret = PTR_ERR(sa);
4177 		goto drop_write;
4178 	}
4179 
4180 	if (!sa->qgroupid) {
4181 		ret = -EINVAL;
4182 		goto out;
4183 	}
4184 
4185 	trans = btrfs_join_transaction(root);
4186 	if (IS_ERR(trans)) {
4187 		ret = PTR_ERR(trans);
4188 		goto out;
4189 	}
4190 
4191 	if (sa->create) {
4192 		ret = btrfs_create_qgroup(trans, sa->qgroupid);
4193 	} else {
4194 		ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4195 	}
4196 
4197 	err = btrfs_end_transaction(trans);
4198 	if (err && !ret)
4199 		ret = err;
4200 
4201 out:
4202 	kfree(sa);
4203 drop_write:
4204 	mnt_drop_write_file(file);
4205 	return ret;
4206 }
4207 
4208 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4209 {
4210 	struct inode *inode = file_inode(file);
4211 	struct btrfs_root *root = BTRFS_I(inode)->root;
4212 	struct btrfs_ioctl_qgroup_limit_args *sa;
4213 	struct btrfs_trans_handle *trans;
4214 	int ret;
4215 	int err;
4216 	u64 qgroupid;
4217 
4218 	if (!capable(CAP_SYS_ADMIN))
4219 		return -EPERM;
4220 
4221 	ret = mnt_want_write_file(file);
4222 	if (ret)
4223 		return ret;
4224 
4225 	sa = memdup_user(arg, sizeof(*sa));
4226 	if (IS_ERR(sa)) {
4227 		ret = PTR_ERR(sa);
4228 		goto drop_write;
4229 	}
4230 
4231 	trans = btrfs_join_transaction(root);
4232 	if (IS_ERR(trans)) {
4233 		ret = PTR_ERR(trans);
4234 		goto out;
4235 	}
4236 
4237 	qgroupid = sa->qgroupid;
4238 	if (!qgroupid) {
4239 		/* take the current subvol as qgroup */
4240 		qgroupid = root->root_key.objectid;
4241 	}
4242 
4243 	ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4244 
4245 	err = btrfs_end_transaction(trans);
4246 	if (err && !ret)
4247 		ret = err;
4248 
4249 out:
4250 	kfree(sa);
4251 drop_write:
4252 	mnt_drop_write_file(file);
4253 	return ret;
4254 }
4255 
4256 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4257 {
4258 	struct inode *inode = file_inode(file);
4259 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4260 	struct btrfs_ioctl_quota_rescan_args *qsa;
4261 	int ret;
4262 
4263 	if (!capable(CAP_SYS_ADMIN))
4264 		return -EPERM;
4265 
4266 	ret = mnt_want_write_file(file);
4267 	if (ret)
4268 		return ret;
4269 
4270 	qsa = memdup_user(arg, sizeof(*qsa));
4271 	if (IS_ERR(qsa)) {
4272 		ret = PTR_ERR(qsa);
4273 		goto drop_write;
4274 	}
4275 
4276 	if (qsa->flags) {
4277 		ret = -EINVAL;
4278 		goto out;
4279 	}
4280 
4281 	ret = btrfs_qgroup_rescan(fs_info);
4282 
4283 out:
4284 	kfree(qsa);
4285 drop_write:
4286 	mnt_drop_write_file(file);
4287 	return ret;
4288 }
4289 
4290 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4291 						void __user *arg)
4292 {
4293 	struct btrfs_ioctl_quota_rescan_args *qsa;
4294 	int ret = 0;
4295 
4296 	if (!capable(CAP_SYS_ADMIN))
4297 		return -EPERM;
4298 
4299 	qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4300 	if (!qsa)
4301 		return -ENOMEM;
4302 
4303 	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4304 		qsa->flags = 1;
4305 		qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4306 	}
4307 
4308 	if (copy_to_user(arg, qsa, sizeof(*qsa)))
4309 		ret = -EFAULT;
4310 
4311 	kfree(qsa);
4312 	return ret;
4313 }
4314 
4315 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4316 						void __user *arg)
4317 {
4318 	if (!capable(CAP_SYS_ADMIN))
4319 		return -EPERM;
4320 
4321 	return btrfs_qgroup_wait_for_completion(fs_info, true);
4322 }
4323 
4324 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4325 					    struct btrfs_ioctl_received_subvol_args *sa)
4326 {
4327 	struct inode *inode = file_inode(file);
4328 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4329 	struct btrfs_root *root = BTRFS_I(inode)->root;
4330 	struct btrfs_root_item *root_item = &root->root_item;
4331 	struct btrfs_trans_handle *trans;
4332 	struct timespec64 ct = current_time(inode);
4333 	int ret = 0;
4334 	int received_uuid_changed;
4335 
4336 	if (!inode_owner_or_capable(inode))
4337 		return -EPERM;
4338 
4339 	ret = mnt_want_write_file(file);
4340 	if (ret < 0)
4341 		return ret;
4342 
4343 	down_write(&fs_info->subvol_sem);
4344 
4345 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4346 		ret = -EINVAL;
4347 		goto out;
4348 	}
4349 
4350 	if (btrfs_root_readonly(root)) {
4351 		ret = -EROFS;
4352 		goto out;
4353 	}
4354 
4355 	/*
4356 	 * 1 - root item
4357 	 * 2 - uuid items (received uuid + subvol uuid)
4358 	 */
4359 	trans = btrfs_start_transaction(root, 3);
4360 	if (IS_ERR(trans)) {
4361 		ret = PTR_ERR(trans);
4362 		trans = NULL;
4363 		goto out;
4364 	}
4365 
4366 	sa->rtransid = trans->transid;
4367 	sa->rtime.sec = ct.tv_sec;
4368 	sa->rtime.nsec = ct.tv_nsec;
4369 
4370 	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4371 				       BTRFS_UUID_SIZE);
4372 	if (received_uuid_changed &&
4373 	    !btrfs_is_empty_uuid(root_item->received_uuid)) {
4374 		ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4375 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4376 					  root->root_key.objectid);
4377 		if (ret && ret != -ENOENT) {
4378 		        btrfs_abort_transaction(trans, ret);
4379 		        btrfs_end_transaction(trans);
4380 		        goto out;
4381 		}
4382 	}
4383 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4384 	btrfs_set_root_stransid(root_item, sa->stransid);
4385 	btrfs_set_root_rtransid(root_item, sa->rtransid);
4386 	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4387 	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4388 	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4389 	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4390 
4391 	ret = btrfs_update_root(trans, fs_info->tree_root,
4392 				&root->root_key, &root->root_item);
4393 	if (ret < 0) {
4394 		btrfs_end_transaction(trans);
4395 		goto out;
4396 	}
4397 	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4398 		ret = btrfs_uuid_tree_add(trans, sa->uuid,
4399 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4400 					  root->root_key.objectid);
4401 		if (ret < 0 && ret != -EEXIST) {
4402 			btrfs_abort_transaction(trans, ret);
4403 			btrfs_end_transaction(trans);
4404 			goto out;
4405 		}
4406 	}
4407 	ret = btrfs_commit_transaction(trans);
4408 out:
4409 	up_write(&fs_info->subvol_sem);
4410 	mnt_drop_write_file(file);
4411 	return ret;
4412 }
4413 
4414 #ifdef CONFIG_64BIT
4415 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4416 						void __user *arg)
4417 {
4418 	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4419 	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4420 	int ret = 0;
4421 
4422 	args32 = memdup_user(arg, sizeof(*args32));
4423 	if (IS_ERR(args32))
4424 		return PTR_ERR(args32);
4425 
4426 	args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4427 	if (!args64) {
4428 		ret = -ENOMEM;
4429 		goto out;
4430 	}
4431 
4432 	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4433 	args64->stransid = args32->stransid;
4434 	args64->rtransid = args32->rtransid;
4435 	args64->stime.sec = args32->stime.sec;
4436 	args64->stime.nsec = args32->stime.nsec;
4437 	args64->rtime.sec = args32->rtime.sec;
4438 	args64->rtime.nsec = args32->rtime.nsec;
4439 	args64->flags = args32->flags;
4440 
4441 	ret = _btrfs_ioctl_set_received_subvol(file, args64);
4442 	if (ret)
4443 		goto out;
4444 
4445 	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4446 	args32->stransid = args64->stransid;
4447 	args32->rtransid = args64->rtransid;
4448 	args32->stime.sec = args64->stime.sec;
4449 	args32->stime.nsec = args64->stime.nsec;
4450 	args32->rtime.sec = args64->rtime.sec;
4451 	args32->rtime.nsec = args64->rtime.nsec;
4452 	args32->flags = args64->flags;
4453 
4454 	ret = copy_to_user(arg, args32, sizeof(*args32));
4455 	if (ret)
4456 		ret = -EFAULT;
4457 
4458 out:
4459 	kfree(args32);
4460 	kfree(args64);
4461 	return ret;
4462 }
4463 #endif
4464 
4465 static long btrfs_ioctl_set_received_subvol(struct file *file,
4466 					    void __user *arg)
4467 {
4468 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4469 	int ret = 0;
4470 
4471 	sa = memdup_user(arg, sizeof(*sa));
4472 	if (IS_ERR(sa))
4473 		return PTR_ERR(sa);
4474 
4475 	ret = _btrfs_ioctl_set_received_subvol(file, sa);
4476 
4477 	if (ret)
4478 		goto out;
4479 
4480 	ret = copy_to_user(arg, sa, sizeof(*sa));
4481 	if (ret)
4482 		ret = -EFAULT;
4483 
4484 out:
4485 	kfree(sa);
4486 	return ret;
4487 }
4488 
4489 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4490 					void __user *arg)
4491 {
4492 	size_t len;
4493 	int ret;
4494 	char label[BTRFS_LABEL_SIZE];
4495 
4496 	spin_lock(&fs_info->super_lock);
4497 	memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4498 	spin_unlock(&fs_info->super_lock);
4499 
4500 	len = strnlen(label, BTRFS_LABEL_SIZE);
4501 
4502 	if (len == BTRFS_LABEL_SIZE) {
4503 		btrfs_warn(fs_info,
4504 			   "label is too long, return the first %zu bytes",
4505 			   --len);
4506 	}
4507 
4508 	ret = copy_to_user(arg, label, len);
4509 
4510 	return ret ? -EFAULT : 0;
4511 }
4512 
4513 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4514 {
4515 	struct inode *inode = file_inode(file);
4516 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4517 	struct btrfs_root *root = BTRFS_I(inode)->root;
4518 	struct btrfs_super_block *super_block = fs_info->super_copy;
4519 	struct btrfs_trans_handle *trans;
4520 	char label[BTRFS_LABEL_SIZE];
4521 	int ret;
4522 
4523 	if (!capable(CAP_SYS_ADMIN))
4524 		return -EPERM;
4525 
4526 	if (copy_from_user(label, arg, sizeof(label)))
4527 		return -EFAULT;
4528 
4529 	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4530 		btrfs_err(fs_info,
4531 			  "unable to set label with more than %d bytes",
4532 			  BTRFS_LABEL_SIZE - 1);
4533 		return -EINVAL;
4534 	}
4535 
4536 	ret = mnt_want_write_file(file);
4537 	if (ret)
4538 		return ret;
4539 
4540 	trans = btrfs_start_transaction(root, 0);
4541 	if (IS_ERR(trans)) {
4542 		ret = PTR_ERR(trans);
4543 		goto out_unlock;
4544 	}
4545 
4546 	spin_lock(&fs_info->super_lock);
4547 	strcpy(super_block->label, label);
4548 	spin_unlock(&fs_info->super_lock);
4549 	ret = btrfs_commit_transaction(trans);
4550 
4551 out_unlock:
4552 	mnt_drop_write_file(file);
4553 	return ret;
4554 }
4555 
4556 #define INIT_FEATURE_FLAGS(suffix) \
4557 	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4558 	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4559 	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4560 
4561 int btrfs_ioctl_get_supported_features(void __user *arg)
4562 {
4563 	static const struct btrfs_ioctl_feature_flags features[3] = {
4564 		INIT_FEATURE_FLAGS(SUPP),
4565 		INIT_FEATURE_FLAGS(SAFE_SET),
4566 		INIT_FEATURE_FLAGS(SAFE_CLEAR)
4567 	};
4568 
4569 	if (copy_to_user(arg, &features, sizeof(features)))
4570 		return -EFAULT;
4571 
4572 	return 0;
4573 }
4574 
4575 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4576 					void __user *arg)
4577 {
4578 	struct btrfs_super_block *super_block = fs_info->super_copy;
4579 	struct btrfs_ioctl_feature_flags features;
4580 
4581 	features.compat_flags = btrfs_super_compat_flags(super_block);
4582 	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4583 	features.incompat_flags = btrfs_super_incompat_flags(super_block);
4584 
4585 	if (copy_to_user(arg, &features, sizeof(features)))
4586 		return -EFAULT;
4587 
4588 	return 0;
4589 }
4590 
4591 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4592 			      enum btrfs_feature_set set,
4593 			      u64 change_mask, u64 flags, u64 supported_flags,
4594 			      u64 safe_set, u64 safe_clear)
4595 {
4596 	const char *type = btrfs_feature_set_name(set);
4597 	char *names;
4598 	u64 disallowed, unsupported;
4599 	u64 set_mask = flags & change_mask;
4600 	u64 clear_mask = ~flags & change_mask;
4601 
4602 	unsupported = set_mask & ~supported_flags;
4603 	if (unsupported) {
4604 		names = btrfs_printable_features(set, unsupported);
4605 		if (names) {
4606 			btrfs_warn(fs_info,
4607 				   "this kernel does not support the %s feature bit%s",
4608 				   names, strchr(names, ',') ? "s" : "");
4609 			kfree(names);
4610 		} else
4611 			btrfs_warn(fs_info,
4612 				   "this kernel does not support %s bits 0x%llx",
4613 				   type, unsupported);
4614 		return -EOPNOTSUPP;
4615 	}
4616 
4617 	disallowed = set_mask & ~safe_set;
4618 	if (disallowed) {
4619 		names = btrfs_printable_features(set, disallowed);
4620 		if (names) {
4621 			btrfs_warn(fs_info,
4622 				   "can't set the %s feature bit%s while mounted",
4623 				   names, strchr(names, ',') ? "s" : "");
4624 			kfree(names);
4625 		} else
4626 			btrfs_warn(fs_info,
4627 				   "can't set %s bits 0x%llx while mounted",
4628 				   type, disallowed);
4629 		return -EPERM;
4630 	}
4631 
4632 	disallowed = clear_mask & ~safe_clear;
4633 	if (disallowed) {
4634 		names = btrfs_printable_features(set, disallowed);
4635 		if (names) {
4636 			btrfs_warn(fs_info,
4637 				   "can't clear the %s feature bit%s while mounted",
4638 				   names, strchr(names, ',') ? "s" : "");
4639 			kfree(names);
4640 		} else
4641 			btrfs_warn(fs_info,
4642 				   "can't clear %s bits 0x%llx while mounted",
4643 				   type, disallowed);
4644 		return -EPERM;
4645 	}
4646 
4647 	return 0;
4648 }
4649 
4650 #define check_feature(fs_info, change_mask, flags, mask_base)	\
4651 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,	\
4652 		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
4653 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
4654 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4655 
4656 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4657 {
4658 	struct inode *inode = file_inode(file);
4659 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4660 	struct btrfs_root *root = BTRFS_I(inode)->root;
4661 	struct btrfs_super_block *super_block = fs_info->super_copy;
4662 	struct btrfs_ioctl_feature_flags flags[2];
4663 	struct btrfs_trans_handle *trans;
4664 	u64 newflags;
4665 	int ret;
4666 
4667 	if (!capable(CAP_SYS_ADMIN))
4668 		return -EPERM;
4669 
4670 	if (copy_from_user(flags, arg, sizeof(flags)))
4671 		return -EFAULT;
4672 
4673 	/* Nothing to do */
4674 	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4675 	    !flags[0].incompat_flags)
4676 		return 0;
4677 
4678 	ret = check_feature(fs_info, flags[0].compat_flags,
4679 			    flags[1].compat_flags, COMPAT);
4680 	if (ret)
4681 		return ret;
4682 
4683 	ret = check_feature(fs_info, flags[0].compat_ro_flags,
4684 			    flags[1].compat_ro_flags, COMPAT_RO);
4685 	if (ret)
4686 		return ret;
4687 
4688 	ret = check_feature(fs_info, flags[0].incompat_flags,
4689 			    flags[1].incompat_flags, INCOMPAT);
4690 	if (ret)
4691 		return ret;
4692 
4693 	ret = mnt_want_write_file(file);
4694 	if (ret)
4695 		return ret;
4696 
4697 	trans = btrfs_start_transaction(root, 0);
4698 	if (IS_ERR(trans)) {
4699 		ret = PTR_ERR(trans);
4700 		goto out_drop_write;
4701 	}
4702 
4703 	spin_lock(&fs_info->super_lock);
4704 	newflags = btrfs_super_compat_flags(super_block);
4705 	newflags |= flags[0].compat_flags & flags[1].compat_flags;
4706 	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4707 	btrfs_set_super_compat_flags(super_block, newflags);
4708 
4709 	newflags = btrfs_super_compat_ro_flags(super_block);
4710 	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4711 	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4712 	btrfs_set_super_compat_ro_flags(super_block, newflags);
4713 
4714 	newflags = btrfs_super_incompat_flags(super_block);
4715 	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4716 	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4717 	btrfs_set_super_incompat_flags(super_block, newflags);
4718 	spin_unlock(&fs_info->super_lock);
4719 
4720 	ret = btrfs_commit_transaction(trans);
4721 out_drop_write:
4722 	mnt_drop_write_file(file);
4723 
4724 	return ret;
4725 }
4726 
4727 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4728 {
4729 	struct btrfs_ioctl_send_args *arg;
4730 	int ret;
4731 
4732 	if (compat) {
4733 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4734 		struct btrfs_ioctl_send_args_32 args32;
4735 
4736 		ret = copy_from_user(&args32, argp, sizeof(args32));
4737 		if (ret)
4738 			return -EFAULT;
4739 		arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4740 		if (!arg)
4741 			return -ENOMEM;
4742 		arg->send_fd = args32.send_fd;
4743 		arg->clone_sources_count = args32.clone_sources_count;
4744 		arg->clone_sources = compat_ptr(args32.clone_sources);
4745 		arg->parent_root = args32.parent_root;
4746 		arg->flags = args32.flags;
4747 		memcpy(arg->reserved, args32.reserved,
4748 		       sizeof(args32.reserved));
4749 #else
4750 		return -ENOTTY;
4751 #endif
4752 	} else {
4753 		arg = memdup_user(argp, sizeof(*arg));
4754 		if (IS_ERR(arg))
4755 			return PTR_ERR(arg);
4756 	}
4757 	ret = btrfs_ioctl_send(file, arg);
4758 	kfree(arg);
4759 	return ret;
4760 }
4761 
4762 long btrfs_ioctl(struct file *file, unsigned int
4763 		cmd, unsigned long arg)
4764 {
4765 	struct inode *inode = file_inode(file);
4766 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4767 	struct btrfs_root *root = BTRFS_I(inode)->root;
4768 	void __user *argp = (void __user *)arg;
4769 
4770 	switch (cmd) {
4771 	case FS_IOC_GETFLAGS:
4772 		return btrfs_ioctl_getflags(file, argp);
4773 	case FS_IOC_SETFLAGS:
4774 		return btrfs_ioctl_setflags(file, argp);
4775 	case FS_IOC_GETVERSION:
4776 		return btrfs_ioctl_getversion(file, argp);
4777 	case FS_IOC_GETFSLABEL:
4778 		return btrfs_ioctl_get_fslabel(fs_info, argp);
4779 	case FS_IOC_SETFSLABEL:
4780 		return btrfs_ioctl_set_fslabel(file, argp);
4781 	case FITRIM:
4782 		return btrfs_ioctl_fitrim(fs_info, argp);
4783 	case BTRFS_IOC_SNAP_CREATE:
4784 		return btrfs_ioctl_snap_create(file, argp, 0);
4785 	case BTRFS_IOC_SNAP_CREATE_V2:
4786 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
4787 	case BTRFS_IOC_SUBVOL_CREATE:
4788 		return btrfs_ioctl_snap_create(file, argp, 1);
4789 	case BTRFS_IOC_SUBVOL_CREATE_V2:
4790 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
4791 	case BTRFS_IOC_SNAP_DESTROY:
4792 		return btrfs_ioctl_snap_destroy(file, argp, false);
4793 	case BTRFS_IOC_SNAP_DESTROY_V2:
4794 		return btrfs_ioctl_snap_destroy(file, argp, true);
4795 	case BTRFS_IOC_SUBVOL_GETFLAGS:
4796 		return btrfs_ioctl_subvol_getflags(file, argp);
4797 	case BTRFS_IOC_SUBVOL_SETFLAGS:
4798 		return btrfs_ioctl_subvol_setflags(file, argp);
4799 	case BTRFS_IOC_DEFAULT_SUBVOL:
4800 		return btrfs_ioctl_default_subvol(file, argp);
4801 	case BTRFS_IOC_DEFRAG:
4802 		return btrfs_ioctl_defrag(file, NULL);
4803 	case BTRFS_IOC_DEFRAG_RANGE:
4804 		return btrfs_ioctl_defrag(file, argp);
4805 	case BTRFS_IOC_RESIZE:
4806 		return btrfs_ioctl_resize(file, argp);
4807 	case BTRFS_IOC_ADD_DEV:
4808 		return btrfs_ioctl_add_dev(fs_info, argp);
4809 	case BTRFS_IOC_RM_DEV:
4810 		return btrfs_ioctl_rm_dev(file, argp);
4811 	case BTRFS_IOC_RM_DEV_V2:
4812 		return btrfs_ioctl_rm_dev_v2(file, argp);
4813 	case BTRFS_IOC_FS_INFO:
4814 		return btrfs_ioctl_fs_info(fs_info, argp);
4815 	case BTRFS_IOC_DEV_INFO:
4816 		return btrfs_ioctl_dev_info(fs_info, argp);
4817 	case BTRFS_IOC_BALANCE:
4818 		return btrfs_ioctl_balance(file, NULL);
4819 	case BTRFS_IOC_TREE_SEARCH:
4820 		return btrfs_ioctl_tree_search(file, argp);
4821 	case BTRFS_IOC_TREE_SEARCH_V2:
4822 		return btrfs_ioctl_tree_search_v2(file, argp);
4823 	case BTRFS_IOC_INO_LOOKUP:
4824 		return btrfs_ioctl_ino_lookup(file, argp);
4825 	case BTRFS_IOC_INO_PATHS:
4826 		return btrfs_ioctl_ino_to_path(root, argp);
4827 	case BTRFS_IOC_LOGICAL_INO:
4828 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4829 	case BTRFS_IOC_LOGICAL_INO_V2:
4830 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4831 	case BTRFS_IOC_SPACE_INFO:
4832 		return btrfs_ioctl_space_info(fs_info, argp);
4833 	case BTRFS_IOC_SYNC: {
4834 		int ret;
4835 
4836 		ret = btrfs_start_delalloc_roots(fs_info, -1);
4837 		if (ret)
4838 			return ret;
4839 		ret = btrfs_sync_fs(inode->i_sb, 1);
4840 		/*
4841 		 * The transaction thread may want to do more work,
4842 		 * namely it pokes the cleaner kthread that will start
4843 		 * processing uncleaned subvols.
4844 		 */
4845 		wake_up_process(fs_info->transaction_kthread);
4846 		return ret;
4847 	}
4848 	case BTRFS_IOC_START_SYNC:
4849 		return btrfs_ioctl_start_sync(root, argp);
4850 	case BTRFS_IOC_WAIT_SYNC:
4851 		return btrfs_ioctl_wait_sync(fs_info, argp);
4852 	case BTRFS_IOC_SCRUB:
4853 		return btrfs_ioctl_scrub(file, argp);
4854 	case BTRFS_IOC_SCRUB_CANCEL:
4855 		return btrfs_ioctl_scrub_cancel(fs_info);
4856 	case BTRFS_IOC_SCRUB_PROGRESS:
4857 		return btrfs_ioctl_scrub_progress(fs_info, argp);
4858 	case BTRFS_IOC_BALANCE_V2:
4859 		return btrfs_ioctl_balance(file, argp);
4860 	case BTRFS_IOC_BALANCE_CTL:
4861 		return btrfs_ioctl_balance_ctl(fs_info, arg);
4862 	case BTRFS_IOC_BALANCE_PROGRESS:
4863 		return btrfs_ioctl_balance_progress(fs_info, argp);
4864 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4865 		return btrfs_ioctl_set_received_subvol(file, argp);
4866 #ifdef CONFIG_64BIT
4867 	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4868 		return btrfs_ioctl_set_received_subvol_32(file, argp);
4869 #endif
4870 	case BTRFS_IOC_SEND:
4871 		return _btrfs_ioctl_send(file, argp, false);
4872 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4873 	case BTRFS_IOC_SEND_32:
4874 		return _btrfs_ioctl_send(file, argp, true);
4875 #endif
4876 	case BTRFS_IOC_GET_DEV_STATS:
4877 		return btrfs_ioctl_get_dev_stats(fs_info, argp);
4878 	case BTRFS_IOC_QUOTA_CTL:
4879 		return btrfs_ioctl_quota_ctl(file, argp);
4880 	case BTRFS_IOC_QGROUP_ASSIGN:
4881 		return btrfs_ioctl_qgroup_assign(file, argp);
4882 	case BTRFS_IOC_QGROUP_CREATE:
4883 		return btrfs_ioctl_qgroup_create(file, argp);
4884 	case BTRFS_IOC_QGROUP_LIMIT:
4885 		return btrfs_ioctl_qgroup_limit(file, argp);
4886 	case BTRFS_IOC_QUOTA_RESCAN:
4887 		return btrfs_ioctl_quota_rescan(file, argp);
4888 	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4889 		return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4890 	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4891 		return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4892 	case BTRFS_IOC_DEV_REPLACE:
4893 		return btrfs_ioctl_dev_replace(fs_info, argp);
4894 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4895 		return btrfs_ioctl_get_supported_features(argp);
4896 	case BTRFS_IOC_GET_FEATURES:
4897 		return btrfs_ioctl_get_features(fs_info, argp);
4898 	case BTRFS_IOC_SET_FEATURES:
4899 		return btrfs_ioctl_set_features(file, argp);
4900 	case FS_IOC_FSGETXATTR:
4901 		return btrfs_ioctl_fsgetxattr(file, argp);
4902 	case FS_IOC_FSSETXATTR:
4903 		return btrfs_ioctl_fssetxattr(file, argp);
4904 	case BTRFS_IOC_GET_SUBVOL_INFO:
4905 		return btrfs_ioctl_get_subvol_info(file, argp);
4906 	case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4907 		return btrfs_ioctl_get_subvol_rootref(file, argp);
4908 	case BTRFS_IOC_INO_LOOKUP_USER:
4909 		return btrfs_ioctl_ino_lookup_user(file, argp);
4910 	}
4911 
4912 	return -ENOTTY;
4913 }
4914 
4915 #ifdef CONFIG_COMPAT
4916 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4917 {
4918 	/*
4919 	 * These all access 32-bit values anyway so no further
4920 	 * handling is necessary.
4921 	 */
4922 	switch (cmd) {
4923 	case FS_IOC32_GETFLAGS:
4924 		cmd = FS_IOC_GETFLAGS;
4925 		break;
4926 	case FS_IOC32_SETFLAGS:
4927 		cmd = FS_IOC_SETFLAGS;
4928 		break;
4929 	case FS_IOC32_GETVERSION:
4930 		cmd = FS_IOC_GETVERSION;
4931 		break;
4932 	}
4933 
4934 	return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4935 }
4936 #endif
4937