xref: /linux/fs/btrfs/ioctl.c (revision 702648721db590b3425c31ade294000e18808345)
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 <linux/fileattr.h>
30 #include <linux/fsverity.h>
31 #include <linux/sched/xacct.h>
32 #include "ctree.h"
33 #include "disk-io.h"
34 #include "export.h"
35 #include "transaction.h"
36 #include "btrfs_inode.h"
37 #include "print-tree.h"
38 #include "volumes.h"
39 #include "locking.h"
40 #include "backref.h"
41 #include "rcu-string.h"
42 #include "send.h"
43 #include "dev-replace.h"
44 #include "props.h"
45 #include "sysfs.h"
46 #include "qgroup.h"
47 #include "tree-log.h"
48 #include "compression.h"
49 #include "space-info.h"
50 #include "delalloc-space.h"
51 #include "block-group.h"
52 #include "subpage.h"
53 #include "fs.h"
54 #include "accessors.h"
55 #include "extent-tree.h"
56 #include "root-tree.h"
57 #include "defrag.h"
58 #include "dir-item.h"
59 #include "uuid-tree.h"
60 #include "ioctl.h"
61 #include "file.h"
62 #include "scrub.h"
63 #include "super.h"
64 
65 #ifdef CONFIG_64BIT
66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67  * structures are incorrect, as the timespec structure from userspace
68  * is 4 bytes too small. We define these alternatives here to teach
69  * the kernel about the 32-bit struct packing.
70  */
71 struct btrfs_ioctl_timespec_32 {
72 	__u64 sec;
73 	__u32 nsec;
74 } __attribute__ ((__packed__));
75 
76 struct btrfs_ioctl_received_subvol_args_32 {
77 	char	uuid[BTRFS_UUID_SIZE];	/* in */
78 	__u64	stransid;		/* in */
79 	__u64	rtransid;		/* out */
80 	struct btrfs_ioctl_timespec_32 stime; /* in */
81 	struct btrfs_ioctl_timespec_32 rtime; /* out */
82 	__u64	flags;			/* in */
83 	__u64	reserved[16];		/* in */
84 } __attribute__ ((__packed__));
85 
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 				struct btrfs_ioctl_received_subvol_args_32)
88 #endif
89 
90 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
91 struct btrfs_ioctl_send_args_32 {
92 	__s64 send_fd;			/* in */
93 	__u64 clone_sources_count;	/* in */
94 	compat_uptr_t clone_sources;	/* in */
95 	__u64 parent_root;		/* in */
96 	__u64 flags;			/* in */
97 	__u32 version;			/* in */
98 	__u8  reserved[28];		/* in */
99 } __attribute__ ((__packed__));
100 
101 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
102 			       struct btrfs_ioctl_send_args_32)
103 
104 struct btrfs_ioctl_encoded_io_args_32 {
105 	compat_uptr_t iov;
106 	compat_ulong_t iovcnt;
107 	__s64 offset;
108 	__u64 flags;
109 	__u64 len;
110 	__u64 unencoded_len;
111 	__u64 unencoded_offset;
112 	__u32 compression;
113 	__u32 encryption;
114 	__u8 reserved[64];
115 };
116 
117 #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
118 				       struct btrfs_ioctl_encoded_io_args_32)
119 #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
120 					struct btrfs_ioctl_encoded_io_args_32)
121 #endif
122 
123 /* Mask out flags that are inappropriate for the given type of inode. */
124 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
125 		unsigned int flags)
126 {
127 	if (S_ISDIR(inode->i_mode))
128 		return flags;
129 	else if (S_ISREG(inode->i_mode))
130 		return flags & ~FS_DIRSYNC_FL;
131 	else
132 		return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
133 }
134 
135 /*
136  * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
137  * ioctl.
138  */
139 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
140 {
141 	unsigned int iflags = 0;
142 	u32 flags = binode->flags;
143 	u32 ro_flags = binode->ro_flags;
144 
145 	if (flags & BTRFS_INODE_SYNC)
146 		iflags |= FS_SYNC_FL;
147 	if (flags & BTRFS_INODE_IMMUTABLE)
148 		iflags |= FS_IMMUTABLE_FL;
149 	if (flags & BTRFS_INODE_APPEND)
150 		iflags |= FS_APPEND_FL;
151 	if (flags & BTRFS_INODE_NODUMP)
152 		iflags |= FS_NODUMP_FL;
153 	if (flags & BTRFS_INODE_NOATIME)
154 		iflags |= FS_NOATIME_FL;
155 	if (flags & BTRFS_INODE_DIRSYNC)
156 		iflags |= FS_DIRSYNC_FL;
157 	if (flags & BTRFS_INODE_NODATACOW)
158 		iflags |= FS_NOCOW_FL;
159 	if (ro_flags & BTRFS_INODE_RO_VERITY)
160 		iflags |= FS_VERITY_FL;
161 
162 	if (flags & BTRFS_INODE_NOCOMPRESS)
163 		iflags |= FS_NOCOMP_FL;
164 	else if (flags & BTRFS_INODE_COMPRESS)
165 		iflags |= FS_COMPR_FL;
166 
167 	return iflags;
168 }
169 
170 /*
171  * Update inode->i_flags based on the btrfs internal flags.
172  */
173 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
174 {
175 	struct btrfs_inode *binode = BTRFS_I(inode);
176 	unsigned int new_fl = 0;
177 
178 	if (binode->flags & BTRFS_INODE_SYNC)
179 		new_fl |= S_SYNC;
180 	if (binode->flags & BTRFS_INODE_IMMUTABLE)
181 		new_fl |= S_IMMUTABLE;
182 	if (binode->flags & BTRFS_INODE_APPEND)
183 		new_fl |= S_APPEND;
184 	if (binode->flags & BTRFS_INODE_NOATIME)
185 		new_fl |= S_NOATIME;
186 	if (binode->flags & BTRFS_INODE_DIRSYNC)
187 		new_fl |= S_DIRSYNC;
188 	if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
189 		new_fl |= S_VERITY;
190 
191 	set_mask_bits(&inode->i_flags,
192 		      S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
193 		      S_VERITY, new_fl);
194 }
195 
196 /*
197  * Check if @flags are a supported and valid set of FS_*_FL flags and that
198  * the old and new flags are not conflicting
199  */
200 static int check_fsflags(unsigned int old_flags, unsigned int flags)
201 {
202 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
203 		      FS_NOATIME_FL | FS_NODUMP_FL | \
204 		      FS_SYNC_FL | FS_DIRSYNC_FL | \
205 		      FS_NOCOMP_FL | FS_COMPR_FL |
206 		      FS_NOCOW_FL))
207 		return -EOPNOTSUPP;
208 
209 	/* COMPR and NOCOMP on new/old are valid */
210 	if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
211 		return -EINVAL;
212 
213 	if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
214 		return -EINVAL;
215 
216 	/* NOCOW and compression options are mutually exclusive */
217 	if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
218 		return -EINVAL;
219 	if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
220 		return -EINVAL;
221 
222 	return 0;
223 }
224 
225 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
226 				    unsigned int flags)
227 {
228 	if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
229 		return -EPERM;
230 
231 	return 0;
232 }
233 
234 /*
235  * Set flags/xflags from the internal inode flags. The remaining items of
236  * fsxattr are zeroed.
237  */
238 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
239 {
240 	struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
241 
242 	fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
243 	return 0;
244 }
245 
246 int btrfs_fileattr_set(struct mnt_idmap *idmap,
247 		       struct dentry *dentry, struct fileattr *fa)
248 {
249 	struct inode *inode = d_inode(dentry);
250 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
251 	struct btrfs_inode *binode = BTRFS_I(inode);
252 	struct btrfs_root *root = binode->root;
253 	struct btrfs_trans_handle *trans;
254 	unsigned int fsflags, old_fsflags;
255 	int ret;
256 	const char *comp = NULL;
257 	u32 binode_flags;
258 
259 	if (btrfs_root_readonly(root))
260 		return -EROFS;
261 
262 	if (fileattr_has_fsx(fa))
263 		return -EOPNOTSUPP;
264 
265 	fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
266 	old_fsflags = btrfs_inode_flags_to_fsflags(binode);
267 	ret = check_fsflags(old_fsflags, fsflags);
268 	if (ret)
269 		return ret;
270 
271 	ret = check_fsflags_compatible(fs_info, fsflags);
272 	if (ret)
273 		return ret;
274 
275 	binode_flags = binode->flags;
276 	if (fsflags & FS_SYNC_FL)
277 		binode_flags |= BTRFS_INODE_SYNC;
278 	else
279 		binode_flags &= ~BTRFS_INODE_SYNC;
280 	if (fsflags & FS_IMMUTABLE_FL)
281 		binode_flags |= BTRFS_INODE_IMMUTABLE;
282 	else
283 		binode_flags &= ~BTRFS_INODE_IMMUTABLE;
284 	if (fsflags & FS_APPEND_FL)
285 		binode_flags |= BTRFS_INODE_APPEND;
286 	else
287 		binode_flags &= ~BTRFS_INODE_APPEND;
288 	if (fsflags & FS_NODUMP_FL)
289 		binode_flags |= BTRFS_INODE_NODUMP;
290 	else
291 		binode_flags &= ~BTRFS_INODE_NODUMP;
292 	if (fsflags & FS_NOATIME_FL)
293 		binode_flags |= BTRFS_INODE_NOATIME;
294 	else
295 		binode_flags &= ~BTRFS_INODE_NOATIME;
296 
297 	/* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
298 	if (!fa->flags_valid) {
299 		/* 1 item for the inode */
300 		trans = btrfs_start_transaction(root, 1);
301 		if (IS_ERR(trans))
302 			return PTR_ERR(trans);
303 		goto update_flags;
304 	}
305 
306 	if (fsflags & FS_DIRSYNC_FL)
307 		binode_flags |= BTRFS_INODE_DIRSYNC;
308 	else
309 		binode_flags &= ~BTRFS_INODE_DIRSYNC;
310 	if (fsflags & FS_NOCOW_FL) {
311 		if (S_ISREG(inode->i_mode)) {
312 			/*
313 			 * It's safe to turn csums off here, no extents exist.
314 			 * Otherwise we want the flag to reflect the real COW
315 			 * status of the file and will not set it.
316 			 */
317 			if (inode->i_size == 0)
318 				binode_flags |= BTRFS_INODE_NODATACOW |
319 						BTRFS_INODE_NODATASUM;
320 		} else {
321 			binode_flags |= BTRFS_INODE_NODATACOW;
322 		}
323 	} else {
324 		/*
325 		 * Revert back under same assumptions as above
326 		 */
327 		if (S_ISREG(inode->i_mode)) {
328 			if (inode->i_size == 0)
329 				binode_flags &= ~(BTRFS_INODE_NODATACOW |
330 						  BTRFS_INODE_NODATASUM);
331 		} else {
332 			binode_flags &= ~BTRFS_INODE_NODATACOW;
333 		}
334 	}
335 
336 	/*
337 	 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
338 	 * flag may be changed automatically if compression code won't make
339 	 * things smaller.
340 	 */
341 	if (fsflags & FS_NOCOMP_FL) {
342 		binode_flags &= ~BTRFS_INODE_COMPRESS;
343 		binode_flags |= BTRFS_INODE_NOCOMPRESS;
344 	} else if (fsflags & FS_COMPR_FL) {
345 
346 		if (IS_SWAPFILE(inode))
347 			return -ETXTBSY;
348 
349 		binode_flags |= BTRFS_INODE_COMPRESS;
350 		binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
351 
352 		comp = btrfs_compress_type2str(fs_info->compress_type);
353 		if (!comp || comp[0] == 0)
354 			comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
355 	} else {
356 		binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
357 	}
358 
359 	/*
360 	 * 1 for inode item
361 	 * 2 for properties
362 	 */
363 	trans = btrfs_start_transaction(root, 3);
364 	if (IS_ERR(trans))
365 		return PTR_ERR(trans);
366 
367 	if (comp) {
368 		ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
369 				     strlen(comp), 0);
370 		if (ret) {
371 			btrfs_abort_transaction(trans, ret);
372 			goto out_end_trans;
373 		}
374 	} else {
375 		ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
376 				     0, 0);
377 		if (ret && ret != -ENODATA) {
378 			btrfs_abort_transaction(trans, ret);
379 			goto out_end_trans;
380 		}
381 	}
382 
383 update_flags:
384 	binode->flags = binode_flags;
385 	btrfs_sync_inode_flags_to_i_flags(inode);
386 	inode_inc_iversion(inode);
387 	inode->i_ctime = current_time(inode);
388 	ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
389 
390  out_end_trans:
391 	btrfs_end_transaction(trans);
392 	return ret;
393 }
394 
395 /*
396  * Start exclusive operation @type, return true on success
397  */
398 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
399 			enum btrfs_exclusive_operation type)
400 {
401 	bool ret = false;
402 
403 	spin_lock(&fs_info->super_lock);
404 	if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
405 		fs_info->exclusive_operation = type;
406 		ret = true;
407 	}
408 	spin_unlock(&fs_info->super_lock);
409 
410 	return ret;
411 }
412 
413 /*
414  * Conditionally allow to enter the exclusive operation in case it's compatible
415  * with the running one.  This must be paired with btrfs_exclop_start_unlock and
416  * btrfs_exclop_finish.
417  *
418  * Compatibility:
419  * - the same type is already running
420  * - when trying to add a device and balance has been paused
421  * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
422  *   must check the condition first that would allow none -> @type
423  */
424 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
425 				 enum btrfs_exclusive_operation type)
426 {
427 	spin_lock(&fs_info->super_lock);
428 	if (fs_info->exclusive_operation == type ||
429 	    (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
430 	     type == BTRFS_EXCLOP_DEV_ADD))
431 		return true;
432 
433 	spin_unlock(&fs_info->super_lock);
434 	return false;
435 }
436 
437 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
438 {
439 	spin_unlock(&fs_info->super_lock);
440 }
441 
442 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
443 {
444 	spin_lock(&fs_info->super_lock);
445 	WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
446 	spin_unlock(&fs_info->super_lock);
447 	sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
448 }
449 
450 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
451 			  enum btrfs_exclusive_operation op)
452 {
453 	switch (op) {
454 	case BTRFS_EXCLOP_BALANCE_PAUSED:
455 		spin_lock(&fs_info->super_lock);
456 		ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
457 		       fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD);
458 		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
459 		spin_unlock(&fs_info->super_lock);
460 		break;
461 	case BTRFS_EXCLOP_BALANCE:
462 		spin_lock(&fs_info->super_lock);
463 		ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
464 		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
465 		spin_unlock(&fs_info->super_lock);
466 		break;
467 	default:
468 		btrfs_warn(fs_info,
469 			"invalid exclop balance operation %d requested", op);
470 	}
471 }
472 
473 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
474 {
475 	return put_user(inode->i_generation, arg);
476 }
477 
478 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
479 					void __user *arg)
480 {
481 	struct btrfs_device *device;
482 	struct fstrim_range range;
483 	u64 minlen = ULLONG_MAX;
484 	u64 num_devices = 0;
485 	int ret;
486 
487 	if (!capable(CAP_SYS_ADMIN))
488 		return -EPERM;
489 
490 	/*
491 	 * btrfs_trim_block_group() depends on space cache, which is not
492 	 * available in zoned filesystem. So, disallow fitrim on a zoned
493 	 * filesystem for now.
494 	 */
495 	if (btrfs_is_zoned(fs_info))
496 		return -EOPNOTSUPP;
497 
498 	/*
499 	 * If the fs is mounted with nologreplay, which requires it to be
500 	 * mounted in RO mode as well, we can not allow discard on free space
501 	 * inside block groups, because log trees refer to extents that are not
502 	 * pinned in a block group's free space cache (pinning the extents is
503 	 * precisely the first phase of replaying a log tree).
504 	 */
505 	if (btrfs_test_opt(fs_info, NOLOGREPLAY))
506 		return -EROFS;
507 
508 	rcu_read_lock();
509 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
510 				dev_list) {
511 		if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
512 			continue;
513 		num_devices++;
514 		minlen = min_t(u64, bdev_discard_granularity(device->bdev),
515 				    minlen);
516 	}
517 	rcu_read_unlock();
518 
519 	if (!num_devices)
520 		return -EOPNOTSUPP;
521 	if (copy_from_user(&range, arg, sizeof(range)))
522 		return -EFAULT;
523 
524 	/*
525 	 * NOTE: Don't truncate the range using super->total_bytes.  Bytenr of
526 	 * block group is in the logical address space, which can be any
527 	 * sectorsize aligned bytenr in  the range [0, U64_MAX].
528 	 */
529 	if (range.len < fs_info->sb->s_blocksize)
530 		return -EINVAL;
531 
532 	range.minlen = max(range.minlen, minlen);
533 	ret = btrfs_trim_fs(fs_info, &range);
534 	if (ret < 0)
535 		return ret;
536 
537 	if (copy_to_user(arg, &range, sizeof(range)))
538 		return -EFAULT;
539 
540 	return 0;
541 }
542 
543 int __pure btrfs_is_empty_uuid(u8 *uuid)
544 {
545 	int i;
546 
547 	for (i = 0; i < BTRFS_UUID_SIZE; i++) {
548 		if (uuid[i])
549 			return 0;
550 	}
551 	return 1;
552 }
553 
554 /*
555  * Calculate the number of transaction items to reserve for creating a subvolume
556  * or snapshot, not including the inode, directory entries, or parent directory.
557  */
558 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
559 {
560 	/*
561 	 * 1 to add root block
562 	 * 1 to add root item
563 	 * 1 to add root ref
564 	 * 1 to add root backref
565 	 * 1 to add UUID item
566 	 * 1 to add qgroup info
567 	 * 1 to add qgroup limit
568 	 *
569 	 * Ideally the last two would only be accounted if qgroups are enabled,
570 	 * but that can change between now and the time we would insert them.
571 	 */
572 	unsigned int num_items = 7;
573 
574 	if (inherit) {
575 		/* 2 to add qgroup relations for each inherited qgroup */
576 		num_items += 2 * inherit->num_qgroups;
577 	}
578 	return num_items;
579 }
580 
581 static noinline int create_subvol(struct mnt_idmap *idmap,
582 				  struct inode *dir, struct dentry *dentry,
583 				  struct btrfs_qgroup_inherit *inherit)
584 {
585 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
586 	struct btrfs_trans_handle *trans;
587 	struct btrfs_key key;
588 	struct btrfs_root_item *root_item;
589 	struct btrfs_inode_item *inode_item;
590 	struct extent_buffer *leaf;
591 	struct btrfs_root *root = BTRFS_I(dir)->root;
592 	struct btrfs_root *new_root;
593 	struct btrfs_block_rsv block_rsv;
594 	struct timespec64 cur_time = current_time(dir);
595 	struct btrfs_new_inode_args new_inode_args = {
596 		.dir = dir,
597 		.dentry = dentry,
598 		.subvol = true,
599 	};
600 	unsigned int trans_num_items;
601 	int ret;
602 	dev_t anon_dev;
603 	u64 objectid;
604 
605 	root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
606 	if (!root_item)
607 		return -ENOMEM;
608 
609 	ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
610 	if (ret)
611 		goto out_root_item;
612 
613 	/*
614 	 * Don't create subvolume whose level is not zero. Or qgroup will be
615 	 * screwed up since it assumes subvolume qgroup's level to be 0.
616 	 */
617 	if (btrfs_qgroup_level(objectid)) {
618 		ret = -ENOSPC;
619 		goto out_root_item;
620 	}
621 
622 	ret = get_anon_bdev(&anon_dev);
623 	if (ret < 0)
624 		goto out_root_item;
625 
626 	new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
627 	if (!new_inode_args.inode) {
628 		ret = -ENOMEM;
629 		goto out_anon_dev;
630 	}
631 	ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
632 	if (ret)
633 		goto out_inode;
634 	trans_num_items += create_subvol_num_items(inherit);
635 
636 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
637 	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
638 					       trans_num_items, false);
639 	if (ret)
640 		goto out_new_inode_args;
641 
642 	trans = btrfs_start_transaction(root, 0);
643 	if (IS_ERR(trans)) {
644 		ret = PTR_ERR(trans);
645 		btrfs_subvolume_release_metadata(root, &block_rsv);
646 		goto out_new_inode_args;
647 	}
648 	trans->block_rsv = &block_rsv;
649 	trans->bytes_reserved = block_rsv.size;
650 
651 	ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
652 	if (ret)
653 		goto out;
654 
655 	leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
656 				      BTRFS_NESTING_NORMAL);
657 	if (IS_ERR(leaf)) {
658 		ret = PTR_ERR(leaf);
659 		goto out;
660 	}
661 
662 	btrfs_mark_buffer_dirty(leaf);
663 
664 	inode_item = &root_item->inode;
665 	btrfs_set_stack_inode_generation(inode_item, 1);
666 	btrfs_set_stack_inode_size(inode_item, 3);
667 	btrfs_set_stack_inode_nlink(inode_item, 1);
668 	btrfs_set_stack_inode_nbytes(inode_item,
669 				     fs_info->nodesize);
670 	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
671 
672 	btrfs_set_root_flags(root_item, 0);
673 	btrfs_set_root_limit(root_item, 0);
674 	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
675 
676 	btrfs_set_root_bytenr(root_item, leaf->start);
677 	btrfs_set_root_generation(root_item, trans->transid);
678 	btrfs_set_root_level(root_item, 0);
679 	btrfs_set_root_refs(root_item, 1);
680 	btrfs_set_root_used(root_item, leaf->len);
681 	btrfs_set_root_last_snapshot(root_item, 0);
682 
683 	btrfs_set_root_generation_v2(root_item,
684 			btrfs_root_generation(root_item));
685 	generate_random_guid(root_item->uuid);
686 	btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
687 	btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
688 	root_item->ctime = root_item->otime;
689 	btrfs_set_root_ctransid(root_item, trans->transid);
690 	btrfs_set_root_otransid(root_item, trans->transid);
691 
692 	btrfs_tree_unlock(leaf);
693 
694 	btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
695 
696 	key.objectid = objectid;
697 	key.offset = 0;
698 	key.type = BTRFS_ROOT_ITEM_KEY;
699 	ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
700 				root_item);
701 	if (ret) {
702 		/*
703 		 * Since we don't abort the transaction in this case, free the
704 		 * tree block so that we don't leak space and leave the
705 		 * filesystem in an inconsistent state (an extent item in the
706 		 * extent tree with a backreference for a root that does not
707 		 * exists).
708 		 */
709 		btrfs_tree_lock(leaf);
710 		btrfs_clear_buffer_dirty(trans, leaf);
711 		btrfs_tree_unlock(leaf);
712 		btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
713 		free_extent_buffer(leaf);
714 		goto out;
715 	}
716 
717 	free_extent_buffer(leaf);
718 	leaf = NULL;
719 
720 	new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
721 	if (IS_ERR(new_root)) {
722 		ret = PTR_ERR(new_root);
723 		btrfs_abort_transaction(trans, ret);
724 		goto out;
725 	}
726 	/* anon_dev is owned by new_root now. */
727 	anon_dev = 0;
728 	BTRFS_I(new_inode_args.inode)->root = new_root;
729 	/* ... and new_root is owned by new_inode_args.inode now. */
730 
731 	ret = btrfs_record_root_in_trans(trans, new_root);
732 	if (ret) {
733 		btrfs_abort_transaction(trans, ret);
734 		goto out;
735 	}
736 
737 	ret = btrfs_uuid_tree_add(trans, root_item->uuid,
738 				  BTRFS_UUID_KEY_SUBVOL, objectid);
739 	if (ret) {
740 		btrfs_abort_transaction(trans, ret);
741 		goto out;
742 	}
743 
744 	ret = btrfs_create_new_inode(trans, &new_inode_args);
745 	if (ret) {
746 		btrfs_abort_transaction(trans, ret);
747 		goto out;
748 	}
749 
750 	d_instantiate_new(dentry, new_inode_args.inode);
751 	new_inode_args.inode = NULL;
752 
753 out:
754 	trans->block_rsv = NULL;
755 	trans->bytes_reserved = 0;
756 	btrfs_subvolume_release_metadata(root, &block_rsv);
757 
758 	if (ret)
759 		btrfs_end_transaction(trans);
760 	else
761 		ret = btrfs_commit_transaction(trans);
762 out_new_inode_args:
763 	btrfs_new_inode_args_destroy(&new_inode_args);
764 out_inode:
765 	iput(new_inode_args.inode);
766 out_anon_dev:
767 	if (anon_dev)
768 		free_anon_bdev(anon_dev);
769 out_root_item:
770 	kfree(root_item);
771 	return ret;
772 }
773 
774 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
775 			   struct dentry *dentry, bool readonly,
776 			   struct btrfs_qgroup_inherit *inherit)
777 {
778 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
779 	struct inode *inode;
780 	struct btrfs_pending_snapshot *pending_snapshot;
781 	unsigned int trans_num_items;
782 	struct btrfs_trans_handle *trans;
783 	int ret;
784 
785 	/* We do not support snapshotting right now. */
786 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
787 		btrfs_warn(fs_info,
788 			   "extent tree v2 doesn't support snapshotting yet");
789 		return -EOPNOTSUPP;
790 	}
791 
792 	if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
793 		return -EINVAL;
794 
795 	if (atomic_read(&root->nr_swapfiles)) {
796 		btrfs_warn(fs_info,
797 			   "cannot snapshot subvolume with active swapfile");
798 		return -ETXTBSY;
799 	}
800 
801 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
802 	if (!pending_snapshot)
803 		return -ENOMEM;
804 
805 	ret = get_anon_bdev(&pending_snapshot->anon_dev);
806 	if (ret < 0)
807 		goto free_pending;
808 	pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
809 			GFP_KERNEL);
810 	pending_snapshot->path = btrfs_alloc_path();
811 	if (!pending_snapshot->root_item || !pending_snapshot->path) {
812 		ret = -ENOMEM;
813 		goto free_pending;
814 	}
815 
816 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
817 			     BTRFS_BLOCK_RSV_TEMP);
818 	/*
819 	 * 1 to add dir item
820 	 * 1 to add dir index
821 	 * 1 to update parent inode item
822 	 */
823 	trans_num_items = create_subvol_num_items(inherit) + 3;
824 	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
825 					       &pending_snapshot->block_rsv,
826 					       trans_num_items, false);
827 	if (ret)
828 		goto free_pending;
829 
830 	pending_snapshot->dentry = dentry;
831 	pending_snapshot->root = root;
832 	pending_snapshot->readonly = readonly;
833 	pending_snapshot->dir = dir;
834 	pending_snapshot->inherit = inherit;
835 
836 	trans = btrfs_start_transaction(root, 0);
837 	if (IS_ERR(trans)) {
838 		ret = PTR_ERR(trans);
839 		goto fail;
840 	}
841 
842 	trans->pending_snapshot = pending_snapshot;
843 
844 	ret = btrfs_commit_transaction(trans);
845 	if (ret)
846 		goto fail;
847 
848 	ret = pending_snapshot->error;
849 	if (ret)
850 		goto fail;
851 
852 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
853 	if (ret)
854 		goto fail;
855 
856 	inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
857 	if (IS_ERR(inode)) {
858 		ret = PTR_ERR(inode);
859 		goto fail;
860 	}
861 
862 	d_instantiate(dentry, inode);
863 	ret = 0;
864 	pending_snapshot->anon_dev = 0;
865 fail:
866 	/* Prevent double freeing of anon_dev */
867 	if (ret && pending_snapshot->snap)
868 		pending_snapshot->snap->anon_dev = 0;
869 	btrfs_put_root(pending_snapshot->snap);
870 	btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
871 free_pending:
872 	if (pending_snapshot->anon_dev)
873 		free_anon_bdev(pending_snapshot->anon_dev);
874 	kfree(pending_snapshot->root_item);
875 	btrfs_free_path(pending_snapshot->path);
876 	kfree(pending_snapshot);
877 
878 	return ret;
879 }
880 
881 /*  copy of may_delete in fs/namei.c()
882  *	Check whether we can remove a link victim from directory dir, check
883  *  whether the type of victim is right.
884  *  1. We can't do it if dir is read-only (done in permission())
885  *  2. We should have write and exec permissions on dir
886  *  3. We can't remove anything from append-only dir
887  *  4. We can't do anything with immutable dir (done in permission())
888  *  5. If the sticky bit on dir is set we should either
889  *	a. be owner of dir, or
890  *	b. be owner of victim, or
891  *	c. have CAP_FOWNER capability
892  *  6. If the victim is append-only or immutable we can't do anything with
893  *     links pointing to it.
894  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
895  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
896  *  9. We can't remove a root or mountpoint.
897  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
898  *     nfs_async_unlink().
899  */
900 
901 static int btrfs_may_delete(struct mnt_idmap *idmap,
902 			    struct inode *dir, struct dentry *victim, int isdir)
903 {
904 	int error;
905 
906 	if (d_really_is_negative(victim))
907 		return -ENOENT;
908 
909 	BUG_ON(d_inode(victim->d_parent) != dir);
910 	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
911 
912 	error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
913 	if (error)
914 		return error;
915 	if (IS_APPEND(dir))
916 		return -EPERM;
917 	if (check_sticky(idmap, dir, d_inode(victim)) ||
918 	    IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
919 	    IS_SWAPFILE(d_inode(victim)))
920 		return -EPERM;
921 	if (isdir) {
922 		if (!d_is_dir(victim))
923 			return -ENOTDIR;
924 		if (IS_ROOT(victim))
925 			return -EBUSY;
926 	} else if (d_is_dir(victim))
927 		return -EISDIR;
928 	if (IS_DEADDIR(dir))
929 		return -ENOENT;
930 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
931 		return -EBUSY;
932 	return 0;
933 }
934 
935 /* copy of may_create in fs/namei.c() */
936 static inline int btrfs_may_create(struct mnt_idmap *idmap,
937 				   struct inode *dir, struct dentry *child)
938 {
939 	if (d_really_is_positive(child))
940 		return -EEXIST;
941 	if (IS_DEADDIR(dir))
942 		return -ENOENT;
943 	if (!fsuidgid_has_mapping(dir->i_sb, idmap))
944 		return -EOVERFLOW;
945 	return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
946 }
947 
948 /*
949  * Create a new subvolume below @parent.  This is largely modeled after
950  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
951  * inside this filesystem so it's quite a bit simpler.
952  */
953 static noinline int btrfs_mksubvol(const struct path *parent,
954 				   struct mnt_idmap *idmap,
955 				   const char *name, int namelen,
956 				   struct btrfs_root *snap_src,
957 				   bool readonly,
958 				   struct btrfs_qgroup_inherit *inherit)
959 {
960 	struct inode *dir = d_inode(parent->dentry);
961 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
962 	struct dentry *dentry;
963 	struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
964 	int error;
965 
966 	error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
967 	if (error == -EINTR)
968 		return error;
969 
970 	dentry = lookup_one(idmap, name, parent->dentry, namelen);
971 	error = PTR_ERR(dentry);
972 	if (IS_ERR(dentry))
973 		goto out_unlock;
974 
975 	error = btrfs_may_create(idmap, dir, dentry);
976 	if (error)
977 		goto out_dput;
978 
979 	/*
980 	 * even if this name doesn't exist, we may get hash collisions.
981 	 * check for them now when we can safely fail
982 	 */
983 	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
984 					       dir->i_ino, &name_str);
985 	if (error)
986 		goto out_dput;
987 
988 	down_read(&fs_info->subvol_sem);
989 
990 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
991 		goto out_up_read;
992 
993 	if (snap_src)
994 		error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
995 	else
996 		error = create_subvol(idmap, dir, dentry, inherit);
997 
998 	if (!error)
999 		fsnotify_mkdir(dir, dentry);
1000 out_up_read:
1001 	up_read(&fs_info->subvol_sem);
1002 out_dput:
1003 	dput(dentry);
1004 out_unlock:
1005 	btrfs_inode_unlock(BTRFS_I(dir), 0);
1006 	return error;
1007 }
1008 
1009 static noinline int btrfs_mksnapshot(const struct path *parent,
1010 				   struct mnt_idmap *idmap,
1011 				   const char *name, int namelen,
1012 				   struct btrfs_root *root,
1013 				   bool readonly,
1014 				   struct btrfs_qgroup_inherit *inherit)
1015 {
1016 	int ret;
1017 	bool snapshot_force_cow = false;
1018 
1019 	/*
1020 	 * Force new buffered writes to reserve space even when NOCOW is
1021 	 * possible. This is to avoid later writeback (running dealloc) to
1022 	 * fallback to COW mode and unexpectedly fail with ENOSPC.
1023 	 */
1024 	btrfs_drew_read_lock(&root->snapshot_lock);
1025 
1026 	ret = btrfs_start_delalloc_snapshot(root, false);
1027 	if (ret)
1028 		goto out;
1029 
1030 	/*
1031 	 * All previous writes have started writeback in NOCOW mode, so now
1032 	 * we force future writes to fallback to COW mode during snapshot
1033 	 * creation.
1034 	 */
1035 	atomic_inc(&root->snapshot_force_cow);
1036 	snapshot_force_cow = true;
1037 
1038 	btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1039 
1040 	ret = btrfs_mksubvol(parent, idmap, name, namelen,
1041 			     root, readonly, inherit);
1042 out:
1043 	if (snapshot_force_cow)
1044 		atomic_dec(&root->snapshot_force_cow);
1045 	btrfs_drew_read_unlock(&root->snapshot_lock);
1046 	return ret;
1047 }
1048 
1049 /*
1050  * Try to start exclusive operation @type or cancel it if it's running.
1051  *
1052  * Return:
1053  *   0        - normal mode, newly claimed op started
1054  *  >0        - normal mode, something else is running,
1055  *              return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1056  * ECANCELED  - cancel mode, successful cancel
1057  * ENOTCONN   - cancel mode, operation not running anymore
1058  */
1059 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1060 			enum btrfs_exclusive_operation type, bool cancel)
1061 {
1062 	if (!cancel) {
1063 		/* Start normal op */
1064 		if (!btrfs_exclop_start(fs_info, type))
1065 			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1066 		/* Exclusive operation is now claimed */
1067 		return 0;
1068 	}
1069 
1070 	/* Cancel running op */
1071 	if (btrfs_exclop_start_try_lock(fs_info, type)) {
1072 		/*
1073 		 * This blocks any exclop finish from setting it to NONE, so we
1074 		 * request cancellation. Either it runs and we will wait for it,
1075 		 * or it has finished and no waiting will happen.
1076 		 */
1077 		atomic_inc(&fs_info->reloc_cancel_req);
1078 		btrfs_exclop_start_unlock(fs_info);
1079 
1080 		if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1081 			wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1082 				    TASK_INTERRUPTIBLE);
1083 
1084 		return -ECANCELED;
1085 	}
1086 
1087 	/* Something else is running or none */
1088 	return -ENOTCONN;
1089 }
1090 
1091 static noinline int btrfs_ioctl_resize(struct file *file,
1092 					void __user *arg)
1093 {
1094 	BTRFS_DEV_LOOKUP_ARGS(args);
1095 	struct inode *inode = file_inode(file);
1096 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1097 	u64 new_size;
1098 	u64 old_size;
1099 	u64 devid = 1;
1100 	struct btrfs_root *root = BTRFS_I(inode)->root;
1101 	struct btrfs_ioctl_vol_args *vol_args;
1102 	struct btrfs_trans_handle *trans;
1103 	struct btrfs_device *device = NULL;
1104 	char *sizestr;
1105 	char *retptr;
1106 	char *devstr = NULL;
1107 	int ret = 0;
1108 	int mod = 0;
1109 	bool cancel;
1110 
1111 	if (!capable(CAP_SYS_ADMIN))
1112 		return -EPERM;
1113 
1114 	ret = mnt_want_write_file(file);
1115 	if (ret)
1116 		return ret;
1117 
1118 	/*
1119 	 * Read the arguments before checking exclusivity to be able to
1120 	 * distinguish regular resize and cancel
1121 	 */
1122 	vol_args = memdup_user(arg, sizeof(*vol_args));
1123 	if (IS_ERR(vol_args)) {
1124 		ret = PTR_ERR(vol_args);
1125 		goto out_drop;
1126 	}
1127 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1128 	sizestr = vol_args->name;
1129 	cancel = (strcmp("cancel", sizestr) == 0);
1130 	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1131 	if (ret)
1132 		goto out_free;
1133 	/* Exclusive operation is now claimed */
1134 
1135 	devstr = strchr(sizestr, ':');
1136 	if (devstr) {
1137 		sizestr = devstr + 1;
1138 		*devstr = '\0';
1139 		devstr = vol_args->name;
1140 		ret = kstrtoull(devstr, 10, &devid);
1141 		if (ret)
1142 			goto out_finish;
1143 		if (!devid) {
1144 			ret = -EINVAL;
1145 			goto out_finish;
1146 		}
1147 		btrfs_info(fs_info, "resizing devid %llu", devid);
1148 	}
1149 
1150 	args.devid = devid;
1151 	device = btrfs_find_device(fs_info->fs_devices, &args);
1152 	if (!device) {
1153 		btrfs_info(fs_info, "resizer unable to find device %llu",
1154 			   devid);
1155 		ret = -ENODEV;
1156 		goto out_finish;
1157 	}
1158 
1159 	if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1160 		btrfs_info(fs_info,
1161 			   "resizer unable to apply on readonly device %llu",
1162 		       devid);
1163 		ret = -EPERM;
1164 		goto out_finish;
1165 	}
1166 
1167 	if (!strcmp(sizestr, "max"))
1168 		new_size = bdev_nr_bytes(device->bdev);
1169 	else {
1170 		if (sizestr[0] == '-') {
1171 			mod = -1;
1172 			sizestr++;
1173 		} else if (sizestr[0] == '+') {
1174 			mod = 1;
1175 			sizestr++;
1176 		}
1177 		new_size = memparse(sizestr, &retptr);
1178 		if (*retptr != '\0' || new_size == 0) {
1179 			ret = -EINVAL;
1180 			goto out_finish;
1181 		}
1182 	}
1183 
1184 	if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1185 		ret = -EPERM;
1186 		goto out_finish;
1187 	}
1188 
1189 	old_size = btrfs_device_get_total_bytes(device);
1190 
1191 	if (mod < 0) {
1192 		if (new_size > old_size) {
1193 			ret = -EINVAL;
1194 			goto out_finish;
1195 		}
1196 		new_size = old_size - new_size;
1197 	} else if (mod > 0) {
1198 		if (new_size > ULLONG_MAX - old_size) {
1199 			ret = -ERANGE;
1200 			goto out_finish;
1201 		}
1202 		new_size = old_size + new_size;
1203 	}
1204 
1205 	if (new_size < SZ_256M) {
1206 		ret = -EINVAL;
1207 		goto out_finish;
1208 	}
1209 	if (new_size > bdev_nr_bytes(device->bdev)) {
1210 		ret = -EFBIG;
1211 		goto out_finish;
1212 	}
1213 
1214 	new_size = round_down(new_size, fs_info->sectorsize);
1215 
1216 	if (new_size > old_size) {
1217 		trans = btrfs_start_transaction(root, 0);
1218 		if (IS_ERR(trans)) {
1219 			ret = PTR_ERR(trans);
1220 			goto out_finish;
1221 		}
1222 		ret = btrfs_grow_device(trans, device, new_size);
1223 		btrfs_commit_transaction(trans);
1224 	} else if (new_size < old_size) {
1225 		ret = btrfs_shrink_device(device, new_size);
1226 	} /* equal, nothing need to do */
1227 
1228 	if (ret == 0 && new_size != old_size)
1229 		btrfs_info_in_rcu(fs_info,
1230 			"resize device %s (devid %llu) from %llu to %llu",
1231 			btrfs_dev_name(device), device->devid,
1232 			old_size, new_size);
1233 out_finish:
1234 	btrfs_exclop_finish(fs_info);
1235 out_free:
1236 	kfree(vol_args);
1237 out_drop:
1238 	mnt_drop_write_file(file);
1239 	return ret;
1240 }
1241 
1242 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1243 				struct mnt_idmap *idmap,
1244 				const char *name, unsigned long fd, int subvol,
1245 				bool readonly,
1246 				struct btrfs_qgroup_inherit *inherit)
1247 {
1248 	int namelen;
1249 	int ret = 0;
1250 
1251 	if (!S_ISDIR(file_inode(file)->i_mode))
1252 		return -ENOTDIR;
1253 
1254 	ret = mnt_want_write_file(file);
1255 	if (ret)
1256 		goto out;
1257 
1258 	namelen = strlen(name);
1259 	if (strchr(name, '/')) {
1260 		ret = -EINVAL;
1261 		goto out_drop_write;
1262 	}
1263 
1264 	if (name[0] == '.' &&
1265 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1266 		ret = -EEXIST;
1267 		goto out_drop_write;
1268 	}
1269 
1270 	if (subvol) {
1271 		ret = btrfs_mksubvol(&file->f_path, idmap, name,
1272 				     namelen, NULL, readonly, inherit);
1273 	} else {
1274 		struct fd src = fdget(fd);
1275 		struct inode *src_inode;
1276 		if (!src.file) {
1277 			ret = -EINVAL;
1278 			goto out_drop_write;
1279 		}
1280 
1281 		src_inode = file_inode(src.file);
1282 		if (src_inode->i_sb != file_inode(file)->i_sb) {
1283 			btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1284 				   "Snapshot src from another FS");
1285 			ret = -EXDEV;
1286 		} else if (!inode_owner_or_capable(idmap, src_inode)) {
1287 			/*
1288 			 * Subvolume creation is not restricted, but snapshots
1289 			 * are limited to own subvolumes only
1290 			 */
1291 			ret = -EPERM;
1292 		} else {
1293 			ret = btrfs_mksnapshot(&file->f_path, idmap,
1294 					       name, namelen,
1295 					       BTRFS_I(src_inode)->root,
1296 					       readonly, inherit);
1297 		}
1298 		fdput(src);
1299 	}
1300 out_drop_write:
1301 	mnt_drop_write_file(file);
1302 out:
1303 	return ret;
1304 }
1305 
1306 static noinline int btrfs_ioctl_snap_create(struct file *file,
1307 					    void __user *arg, int subvol)
1308 {
1309 	struct btrfs_ioctl_vol_args *vol_args;
1310 	int ret;
1311 
1312 	if (!S_ISDIR(file_inode(file)->i_mode))
1313 		return -ENOTDIR;
1314 
1315 	vol_args = memdup_user(arg, sizeof(*vol_args));
1316 	if (IS_ERR(vol_args))
1317 		return PTR_ERR(vol_args);
1318 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1319 
1320 	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1321 					vol_args->name, vol_args->fd, subvol,
1322 					false, NULL);
1323 
1324 	kfree(vol_args);
1325 	return ret;
1326 }
1327 
1328 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1329 					       void __user *arg, int subvol)
1330 {
1331 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1332 	int ret;
1333 	bool readonly = false;
1334 	struct btrfs_qgroup_inherit *inherit = NULL;
1335 
1336 	if (!S_ISDIR(file_inode(file)->i_mode))
1337 		return -ENOTDIR;
1338 
1339 	vol_args = memdup_user(arg, sizeof(*vol_args));
1340 	if (IS_ERR(vol_args))
1341 		return PTR_ERR(vol_args);
1342 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1343 
1344 	if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1345 		ret = -EOPNOTSUPP;
1346 		goto free_args;
1347 	}
1348 
1349 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1350 		readonly = true;
1351 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1352 		u64 nums;
1353 
1354 		if (vol_args->size < sizeof(*inherit) ||
1355 		    vol_args->size > PAGE_SIZE) {
1356 			ret = -EINVAL;
1357 			goto free_args;
1358 		}
1359 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1360 		if (IS_ERR(inherit)) {
1361 			ret = PTR_ERR(inherit);
1362 			goto free_args;
1363 		}
1364 
1365 		if (inherit->num_qgroups > PAGE_SIZE ||
1366 		    inherit->num_ref_copies > PAGE_SIZE ||
1367 		    inherit->num_excl_copies > PAGE_SIZE) {
1368 			ret = -EINVAL;
1369 			goto free_inherit;
1370 		}
1371 
1372 		nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1373 		       2 * inherit->num_excl_copies;
1374 		if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1375 			ret = -EINVAL;
1376 			goto free_inherit;
1377 		}
1378 	}
1379 
1380 	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1381 					vol_args->name, vol_args->fd, subvol,
1382 					readonly, inherit);
1383 	if (ret)
1384 		goto free_inherit;
1385 free_inherit:
1386 	kfree(inherit);
1387 free_args:
1388 	kfree(vol_args);
1389 	return ret;
1390 }
1391 
1392 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1393 						void __user *arg)
1394 {
1395 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1396 	struct btrfs_root *root = BTRFS_I(inode)->root;
1397 	int ret = 0;
1398 	u64 flags = 0;
1399 
1400 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1401 		return -EINVAL;
1402 
1403 	down_read(&fs_info->subvol_sem);
1404 	if (btrfs_root_readonly(root))
1405 		flags |= BTRFS_SUBVOL_RDONLY;
1406 	up_read(&fs_info->subvol_sem);
1407 
1408 	if (copy_to_user(arg, &flags, sizeof(flags)))
1409 		ret = -EFAULT;
1410 
1411 	return ret;
1412 }
1413 
1414 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1415 					      void __user *arg)
1416 {
1417 	struct inode *inode = file_inode(file);
1418 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1419 	struct btrfs_root *root = BTRFS_I(inode)->root;
1420 	struct btrfs_trans_handle *trans;
1421 	u64 root_flags;
1422 	u64 flags;
1423 	int ret = 0;
1424 
1425 	if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1426 		return -EPERM;
1427 
1428 	ret = mnt_want_write_file(file);
1429 	if (ret)
1430 		goto out;
1431 
1432 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1433 		ret = -EINVAL;
1434 		goto out_drop_write;
1435 	}
1436 
1437 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1438 		ret = -EFAULT;
1439 		goto out_drop_write;
1440 	}
1441 
1442 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1443 		ret = -EOPNOTSUPP;
1444 		goto out_drop_write;
1445 	}
1446 
1447 	down_write(&fs_info->subvol_sem);
1448 
1449 	/* nothing to do */
1450 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1451 		goto out_drop_sem;
1452 
1453 	root_flags = btrfs_root_flags(&root->root_item);
1454 	if (flags & BTRFS_SUBVOL_RDONLY) {
1455 		btrfs_set_root_flags(&root->root_item,
1456 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1457 	} else {
1458 		/*
1459 		 * Block RO -> RW transition if this subvolume is involved in
1460 		 * send
1461 		 */
1462 		spin_lock(&root->root_item_lock);
1463 		if (root->send_in_progress == 0) {
1464 			btrfs_set_root_flags(&root->root_item,
1465 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1466 			spin_unlock(&root->root_item_lock);
1467 		} else {
1468 			spin_unlock(&root->root_item_lock);
1469 			btrfs_warn(fs_info,
1470 				   "Attempt to set subvolume %llu read-write during send",
1471 				   root->root_key.objectid);
1472 			ret = -EPERM;
1473 			goto out_drop_sem;
1474 		}
1475 	}
1476 
1477 	trans = btrfs_start_transaction(root, 1);
1478 	if (IS_ERR(trans)) {
1479 		ret = PTR_ERR(trans);
1480 		goto out_reset;
1481 	}
1482 
1483 	ret = btrfs_update_root(trans, fs_info->tree_root,
1484 				&root->root_key, &root->root_item);
1485 	if (ret < 0) {
1486 		btrfs_end_transaction(trans);
1487 		goto out_reset;
1488 	}
1489 
1490 	ret = btrfs_commit_transaction(trans);
1491 
1492 out_reset:
1493 	if (ret)
1494 		btrfs_set_root_flags(&root->root_item, root_flags);
1495 out_drop_sem:
1496 	up_write(&fs_info->subvol_sem);
1497 out_drop_write:
1498 	mnt_drop_write_file(file);
1499 out:
1500 	return ret;
1501 }
1502 
1503 static noinline int key_in_sk(struct btrfs_key *key,
1504 			      struct btrfs_ioctl_search_key *sk)
1505 {
1506 	struct btrfs_key test;
1507 	int ret;
1508 
1509 	test.objectid = sk->min_objectid;
1510 	test.type = sk->min_type;
1511 	test.offset = sk->min_offset;
1512 
1513 	ret = btrfs_comp_cpu_keys(key, &test);
1514 	if (ret < 0)
1515 		return 0;
1516 
1517 	test.objectid = sk->max_objectid;
1518 	test.type = sk->max_type;
1519 	test.offset = sk->max_offset;
1520 
1521 	ret = btrfs_comp_cpu_keys(key, &test);
1522 	if (ret > 0)
1523 		return 0;
1524 	return 1;
1525 }
1526 
1527 static noinline int copy_to_sk(struct btrfs_path *path,
1528 			       struct btrfs_key *key,
1529 			       struct btrfs_ioctl_search_key *sk,
1530 			       size_t *buf_size,
1531 			       char __user *ubuf,
1532 			       unsigned long *sk_offset,
1533 			       int *num_found)
1534 {
1535 	u64 found_transid;
1536 	struct extent_buffer *leaf;
1537 	struct btrfs_ioctl_search_header sh;
1538 	struct btrfs_key test;
1539 	unsigned long item_off;
1540 	unsigned long item_len;
1541 	int nritems;
1542 	int i;
1543 	int slot;
1544 	int ret = 0;
1545 
1546 	leaf = path->nodes[0];
1547 	slot = path->slots[0];
1548 	nritems = btrfs_header_nritems(leaf);
1549 
1550 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1551 		i = nritems;
1552 		goto advance_key;
1553 	}
1554 	found_transid = btrfs_header_generation(leaf);
1555 
1556 	for (i = slot; i < nritems; i++) {
1557 		item_off = btrfs_item_ptr_offset(leaf, i);
1558 		item_len = btrfs_item_size(leaf, i);
1559 
1560 		btrfs_item_key_to_cpu(leaf, key, i);
1561 		if (!key_in_sk(key, sk))
1562 			continue;
1563 
1564 		if (sizeof(sh) + item_len > *buf_size) {
1565 			if (*num_found) {
1566 				ret = 1;
1567 				goto out;
1568 			}
1569 
1570 			/*
1571 			 * return one empty item back for v1, which does not
1572 			 * handle -EOVERFLOW
1573 			 */
1574 
1575 			*buf_size = sizeof(sh) + item_len;
1576 			item_len = 0;
1577 			ret = -EOVERFLOW;
1578 		}
1579 
1580 		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1581 			ret = 1;
1582 			goto out;
1583 		}
1584 
1585 		sh.objectid = key->objectid;
1586 		sh.offset = key->offset;
1587 		sh.type = key->type;
1588 		sh.len = item_len;
1589 		sh.transid = found_transid;
1590 
1591 		/*
1592 		 * Copy search result header. If we fault then loop again so we
1593 		 * can fault in the pages and -EFAULT there if there's a
1594 		 * problem. Otherwise we'll fault and then copy the buffer in
1595 		 * properly this next time through
1596 		 */
1597 		if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1598 			ret = 0;
1599 			goto out;
1600 		}
1601 
1602 		*sk_offset += sizeof(sh);
1603 
1604 		if (item_len) {
1605 			char __user *up = ubuf + *sk_offset;
1606 			/*
1607 			 * Copy the item, same behavior as above, but reset the
1608 			 * * sk_offset so we copy the full thing again.
1609 			 */
1610 			if (read_extent_buffer_to_user_nofault(leaf, up,
1611 						item_off, item_len)) {
1612 				ret = 0;
1613 				*sk_offset -= sizeof(sh);
1614 				goto out;
1615 			}
1616 
1617 			*sk_offset += item_len;
1618 		}
1619 		(*num_found)++;
1620 
1621 		if (ret) /* -EOVERFLOW from above */
1622 			goto out;
1623 
1624 		if (*num_found >= sk->nr_items) {
1625 			ret = 1;
1626 			goto out;
1627 		}
1628 	}
1629 advance_key:
1630 	ret = 0;
1631 	test.objectid = sk->max_objectid;
1632 	test.type = sk->max_type;
1633 	test.offset = sk->max_offset;
1634 	if (btrfs_comp_cpu_keys(key, &test) >= 0)
1635 		ret = 1;
1636 	else if (key->offset < (u64)-1)
1637 		key->offset++;
1638 	else if (key->type < (u8)-1) {
1639 		key->offset = 0;
1640 		key->type++;
1641 	} else if (key->objectid < (u64)-1) {
1642 		key->offset = 0;
1643 		key->type = 0;
1644 		key->objectid++;
1645 	} else
1646 		ret = 1;
1647 out:
1648 	/*
1649 	 *  0: all items from this leaf copied, continue with next
1650 	 *  1: * more items can be copied, but unused buffer is too small
1651 	 *     * all items were found
1652 	 *     Either way, it will stops the loop which iterates to the next
1653 	 *     leaf
1654 	 *  -EOVERFLOW: item was to large for buffer
1655 	 *  -EFAULT: could not copy extent buffer back to userspace
1656 	 */
1657 	return ret;
1658 }
1659 
1660 static noinline int search_ioctl(struct inode *inode,
1661 				 struct btrfs_ioctl_search_key *sk,
1662 				 size_t *buf_size,
1663 				 char __user *ubuf)
1664 {
1665 	struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1666 	struct btrfs_root *root;
1667 	struct btrfs_key key;
1668 	struct btrfs_path *path;
1669 	int ret;
1670 	int num_found = 0;
1671 	unsigned long sk_offset = 0;
1672 
1673 	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1674 		*buf_size = sizeof(struct btrfs_ioctl_search_header);
1675 		return -EOVERFLOW;
1676 	}
1677 
1678 	path = btrfs_alloc_path();
1679 	if (!path)
1680 		return -ENOMEM;
1681 
1682 	if (sk->tree_id == 0) {
1683 		/* search the root of the inode that was passed */
1684 		root = btrfs_grab_root(BTRFS_I(inode)->root);
1685 	} else {
1686 		root = btrfs_get_fs_root(info, sk->tree_id, true);
1687 		if (IS_ERR(root)) {
1688 			btrfs_free_path(path);
1689 			return PTR_ERR(root);
1690 		}
1691 	}
1692 
1693 	key.objectid = sk->min_objectid;
1694 	key.type = sk->min_type;
1695 	key.offset = sk->min_offset;
1696 
1697 	while (1) {
1698 		ret = -EFAULT;
1699 		/*
1700 		 * Ensure that the whole user buffer is faulted in at sub-page
1701 		 * granularity, otherwise the loop may live-lock.
1702 		 */
1703 		if (fault_in_subpage_writeable(ubuf + sk_offset,
1704 					       *buf_size - sk_offset))
1705 			break;
1706 
1707 		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1708 		if (ret != 0) {
1709 			if (ret > 0)
1710 				ret = 0;
1711 			goto err;
1712 		}
1713 		ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1714 				 &sk_offset, &num_found);
1715 		btrfs_release_path(path);
1716 		if (ret)
1717 			break;
1718 
1719 	}
1720 	if (ret > 0)
1721 		ret = 0;
1722 err:
1723 	sk->nr_items = num_found;
1724 	btrfs_put_root(root);
1725 	btrfs_free_path(path);
1726 	return ret;
1727 }
1728 
1729 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1730 					    void __user *argp)
1731 {
1732 	struct btrfs_ioctl_search_args __user *uargs = argp;
1733 	struct btrfs_ioctl_search_key sk;
1734 	int ret;
1735 	size_t buf_size;
1736 
1737 	if (!capable(CAP_SYS_ADMIN))
1738 		return -EPERM;
1739 
1740 	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1741 		return -EFAULT;
1742 
1743 	buf_size = sizeof(uargs->buf);
1744 
1745 	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1746 
1747 	/*
1748 	 * In the origin implementation an overflow is handled by returning a
1749 	 * search header with a len of zero, so reset ret.
1750 	 */
1751 	if (ret == -EOVERFLOW)
1752 		ret = 0;
1753 
1754 	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1755 		ret = -EFAULT;
1756 	return ret;
1757 }
1758 
1759 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1760 					       void __user *argp)
1761 {
1762 	struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1763 	struct btrfs_ioctl_search_args_v2 args;
1764 	int ret;
1765 	size_t buf_size;
1766 	const size_t buf_limit = SZ_16M;
1767 
1768 	if (!capable(CAP_SYS_ADMIN))
1769 		return -EPERM;
1770 
1771 	/* copy search header and buffer size */
1772 	if (copy_from_user(&args, uarg, sizeof(args)))
1773 		return -EFAULT;
1774 
1775 	buf_size = args.buf_size;
1776 
1777 	/* limit result size to 16MB */
1778 	if (buf_size > buf_limit)
1779 		buf_size = buf_limit;
1780 
1781 	ret = search_ioctl(inode, &args.key, &buf_size,
1782 			   (char __user *)(&uarg->buf[0]));
1783 	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1784 		ret = -EFAULT;
1785 	else if (ret == -EOVERFLOW &&
1786 		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1787 		ret = -EFAULT;
1788 
1789 	return ret;
1790 }
1791 
1792 /*
1793  * Search INODE_REFs to identify path name of 'dirid' directory
1794  * in a 'tree_id' tree. and sets path name to 'name'.
1795  */
1796 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1797 				u64 tree_id, u64 dirid, char *name)
1798 {
1799 	struct btrfs_root *root;
1800 	struct btrfs_key key;
1801 	char *ptr;
1802 	int ret = -1;
1803 	int slot;
1804 	int len;
1805 	int total_len = 0;
1806 	struct btrfs_inode_ref *iref;
1807 	struct extent_buffer *l;
1808 	struct btrfs_path *path;
1809 
1810 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1811 		name[0]='\0';
1812 		return 0;
1813 	}
1814 
1815 	path = btrfs_alloc_path();
1816 	if (!path)
1817 		return -ENOMEM;
1818 
1819 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1820 
1821 	root = btrfs_get_fs_root(info, tree_id, true);
1822 	if (IS_ERR(root)) {
1823 		ret = PTR_ERR(root);
1824 		root = NULL;
1825 		goto out;
1826 	}
1827 
1828 	key.objectid = dirid;
1829 	key.type = BTRFS_INODE_REF_KEY;
1830 	key.offset = (u64)-1;
1831 
1832 	while (1) {
1833 		ret = btrfs_search_backwards(root, &key, path);
1834 		if (ret < 0)
1835 			goto out;
1836 		else if (ret > 0) {
1837 			ret = -ENOENT;
1838 			goto out;
1839 		}
1840 
1841 		l = path->nodes[0];
1842 		slot = path->slots[0];
1843 
1844 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1845 		len = btrfs_inode_ref_name_len(l, iref);
1846 		ptr -= len + 1;
1847 		total_len += len + 1;
1848 		if (ptr < name) {
1849 			ret = -ENAMETOOLONG;
1850 			goto out;
1851 		}
1852 
1853 		*(ptr + len) = '/';
1854 		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1855 
1856 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1857 			break;
1858 
1859 		btrfs_release_path(path);
1860 		key.objectid = key.offset;
1861 		key.offset = (u64)-1;
1862 		dirid = key.objectid;
1863 	}
1864 	memmove(name, ptr, total_len);
1865 	name[total_len] = '\0';
1866 	ret = 0;
1867 out:
1868 	btrfs_put_root(root);
1869 	btrfs_free_path(path);
1870 	return ret;
1871 }
1872 
1873 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1874 				struct inode *inode,
1875 				struct btrfs_ioctl_ino_lookup_user_args *args)
1876 {
1877 	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1878 	struct super_block *sb = inode->i_sb;
1879 	struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1880 	u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1881 	u64 dirid = args->dirid;
1882 	unsigned long item_off;
1883 	unsigned long item_len;
1884 	struct btrfs_inode_ref *iref;
1885 	struct btrfs_root_ref *rref;
1886 	struct btrfs_root *root = NULL;
1887 	struct btrfs_path *path;
1888 	struct btrfs_key key, key2;
1889 	struct extent_buffer *leaf;
1890 	struct inode *temp_inode;
1891 	char *ptr;
1892 	int slot;
1893 	int len;
1894 	int total_len = 0;
1895 	int ret;
1896 
1897 	path = btrfs_alloc_path();
1898 	if (!path)
1899 		return -ENOMEM;
1900 
1901 	/*
1902 	 * If the bottom subvolume does not exist directly under upper_limit,
1903 	 * construct the path in from the bottom up.
1904 	 */
1905 	if (dirid != upper_limit.objectid) {
1906 		ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1907 
1908 		root = btrfs_get_fs_root(fs_info, treeid, true);
1909 		if (IS_ERR(root)) {
1910 			ret = PTR_ERR(root);
1911 			goto out;
1912 		}
1913 
1914 		key.objectid = dirid;
1915 		key.type = BTRFS_INODE_REF_KEY;
1916 		key.offset = (u64)-1;
1917 		while (1) {
1918 			ret = btrfs_search_backwards(root, &key, path);
1919 			if (ret < 0)
1920 				goto out_put;
1921 			else if (ret > 0) {
1922 				ret = -ENOENT;
1923 				goto out_put;
1924 			}
1925 
1926 			leaf = path->nodes[0];
1927 			slot = path->slots[0];
1928 
1929 			iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1930 			len = btrfs_inode_ref_name_len(leaf, iref);
1931 			ptr -= len + 1;
1932 			total_len += len + 1;
1933 			if (ptr < args->path) {
1934 				ret = -ENAMETOOLONG;
1935 				goto out_put;
1936 			}
1937 
1938 			*(ptr + len) = '/';
1939 			read_extent_buffer(leaf, ptr,
1940 					(unsigned long)(iref + 1), len);
1941 
1942 			/* Check the read+exec permission of this directory */
1943 			ret = btrfs_previous_item(root, path, dirid,
1944 						  BTRFS_INODE_ITEM_KEY);
1945 			if (ret < 0) {
1946 				goto out_put;
1947 			} else if (ret > 0) {
1948 				ret = -ENOENT;
1949 				goto out_put;
1950 			}
1951 
1952 			leaf = path->nodes[0];
1953 			slot = path->slots[0];
1954 			btrfs_item_key_to_cpu(leaf, &key2, slot);
1955 			if (key2.objectid != dirid) {
1956 				ret = -ENOENT;
1957 				goto out_put;
1958 			}
1959 
1960 			temp_inode = btrfs_iget(sb, key2.objectid, root);
1961 			if (IS_ERR(temp_inode)) {
1962 				ret = PTR_ERR(temp_inode);
1963 				goto out_put;
1964 			}
1965 			ret = inode_permission(idmap, temp_inode,
1966 					       MAY_READ | MAY_EXEC);
1967 			iput(temp_inode);
1968 			if (ret) {
1969 				ret = -EACCES;
1970 				goto out_put;
1971 			}
1972 
1973 			if (key.offset == upper_limit.objectid)
1974 				break;
1975 			if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
1976 				ret = -EACCES;
1977 				goto out_put;
1978 			}
1979 
1980 			btrfs_release_path(path);
1981 			key.objectid = key.offset;
1982 			key.offset = (u64)-1;
1983 			dirid = key.objectid;
1984 		}
1985 
1986 		memmove(args->path, ptr, total_len);
1987 		args->path[total_len] = '\0';
1988 		btrfs_put_root(root);
1989 		root = NULL;
1990 		btrfs_release_path(path);
1991 	}
1992 
1993 	/* Get the bottom subvolume's name from ROOT_REF */
1994 	key.objectid = treeid;
1995 	key.type = BTRFS_ROOT_REF_KEY;
1996 	key.offset = args->treeid;
1997 	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1998 	if (ret < 0) {
1999 		goto out;
2000 	} else if (ret > 0) {
2001 		ret = -ENOENT;
2002 		goto out;
2003 	}
2004 
2005 	leaf = path->nodes[0];
2006 	slot = path->slots[0];
2007 	btrfs_item_key_to_cpu(leaf, &key, slot);
2008 
2009 	item_off = btrfs_item_ptr_offset(leaf, slot);
2010 	item_len = btrfs_item_size(leaf, slot);
2011 	/* Check if dirid in ROOT_REF corresponds to passed dirid */
2012 	rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2013 	if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2014 		ret = -EINVAL;
2015 		goto out;
2016 	}
2017 
2018 	/* Copy subvolume's name */
2019 	item_off += sizeof(struct btrfs_root_ref);
2020 	item_len -= sizeof(struct btrfs_root_ref);
2021 	read_extent_buffer(leaf, args->name, item_off, item_len);
2022 	args->name[item_len] = 0;
2023 
2024 out_put:
2025 	btrfs_put_root(root);
2026 out:
2027 	btrfs_free_path(path);
2028 	return ret;
2029 }
2030 
2031 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2032 					   void __user *argp)
2033 {
2034 	struct btrfs_ioctl_ino_lookup_args *args;
2035 	int ret = 0;
2036 
2037 	args = memdup_user(argp, sizeof(*args));
2038 	if (IS_ERR(args))
2039 		return PTR_ERR(args);
2040 
2041 	/*
2042 	 * Unprivileged query to obtain the containing subvolume root id. The
2043 	 * path is reset so it's consistent with btrfs_search_path_in_tree.
2044 	 */
2045 	if (args->treeid == 0)
2046 		args->treeid = root->root_key.objectid;
2047 
2048 	if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2049 		args->name[0] = 0;
2050 		goto out;
2051 	}
2052 
2053 	if (!capable(CAP_SYS_ADMIN)) {
2054 		ret = -EPERM;
2055 		goto out;
2056 	}
2057 
2058 	ret = btrfs_search_path_in_tree(root->fs_info,
2059 					args->treeid, args->objectid,
2060 					args->name);
2061 
2062 out:
2063 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2064 		ret = -EFAULT;
2065 
2066 	kfree(args);
2067 	return ret;
2068 }
2069 
2070 /*
2071  * Version of ino_lookup ioctl (unprivileged)
2072  *
2073  * The main differences from ino_lookup ioctl are:
2074  *
2075  *   1. Read + Exec permission will be checked using inode_permission() during
2076  *      path construction. -EACCES will be returned in case of failure.
2077  *   2. Path construction will be stopped at the inode number which corresponds
2078  *      to the fd with which this ioctl is called. If constructed path does not
2079  *      exist under fd's inode, -EACCES will be returned.
2080  *   3. The name of bottom subvolume is also searched and filled.
2081  */
2082 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2083 {
2084 	struct btrfs_ioctl_ino_lookup_user_args *args;
2085 	struct inode *inode;
2086 	int ret;
2087 
2088 	args = memdup_user(argp, sizeof(*args));
2089 	if (IS_ERR(args))
2090 		return PTR_ERR(args);
2091 
2092 	inode = file_inode(file);
2093 
2094 	if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2095 	    BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2096 		/*
2097 		 * The subvolume does not exist under fd with which this is
2098 		 * called
2099 		 */
2100 		kfree(args);
2101 		return -EACCES;
2102 	}
2103 
2104 	ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2105 
2106 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2107 		ret = -EFAULT;
2108 
2109 	kfree(args);
2110 	return ret;
2111 }
2112 
2113 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2114 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2115 {
2116 	struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2117 	struct btrfs_fs_info *fs_info;
2118 	struct btrfs_root *root;
2119 	struct btrfs_path *path;
2120 	struct btrfs_key key;
2121 	struct btrfs_root_item *root_item;
2122 	struct btrfs_root_ref *rref;
2123 	struct extent_buffer *leaf;
2124 	unsigned long item_off;
2125 	unsigned long item_len;
2126 	int slot;
2127 	int ret = 0;
2128 
2129 	path = btrfs_alloc_path();
2130 	if (!path)
2131 		return -ENOMEM;
2132 
2133 	subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2134 	if (!subvol_info) {
2135 		btrfs_free_path(path);
2136 		return -ENOMEM;
2137 	}
2138 
2139 	fs_info = BTRFS_I(inode)->root->fs_info;
2140 
2141 	/* Get root_item of inode's subvolume */
2142 	key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2143 	root = btrfs_get_fs_root(fs_info, key.objectid, true);
2144 	if (IS_ERR(root)) {
2145 		ret = PTR_ERR(root);
2146 		goto out_free;
2147 	}
2148 	root_item = &root->root_item;
2149 
2150 	subvol_info->treeid = key.objectid;
2151 
2152 	subvol_info->generation = btrfs_root_generation(root_item);
2153 	subvol_info->flags = btrfs_root_flags(root_item);
2154 
2155 	memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2156 	memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2157 						    BTRFS_UUID_SIZE);
2158 	memcpy(subvol_info->received_uuid, root_item->received_uuid,
2159 						    BTRFS_UUID_SIZE);
2160 
2161 	subvol_info->ctransid = btrfs_root_ctransid(root_item);
2162 	subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2163 	subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2164 
2165 	subvol_info->otransid = btrfs_root_otransid(root_item);
2166 	subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2167 	subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2168 
2169 	subvol_info->stransid = btrfs_root_stransid(root_item);
2170 	subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2171 	subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2172 
2173 	subvol_info->rtransid = btrfs_root_rtransid(root_item);
2174 	subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2175 	subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2176 
2177 	if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2178 		/* Search root tree for ROOT_BACKREF of this subvolume */
2179 		key.type = BTRFS_ROOT_BACKREF_KEY;
2180 		key.offset = 0;
2181 		ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2182 		if (ret < 0) {
2183 			goto out;
2184 		} else if (path->slots[0] >=
2185 			   btrfs_header_nritems(path->nodes[0])) {
2186 			ret = btrfs_next_leaf(fs_info->tree_root, path);
2187 			if (ret < 0) {
2188 				goto out;
2189 			} else if (ret > 0) {
2190 				ret = -EUCLEAN;
2191 				goto out;
2192 			}
2193 		}
2194 
2195 		leaf = path->nodes[0];
2196 		slot = path->slots[0];
2197 		btrfs_item_key_to_cpu(leaf, &key, slot);
2198 		if (key.objectid == subvol_info->treeid &&
2199 		    key.type == BTRFS_ROOT_BACKREF_KEY) {
2200 			subvol_info->parent_id = key.offset;
2201 
2202 			rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2203 			subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2204 
2205 			item_off = btrfs_item_ptr_offset(leaf, slot)
2206 					+ sizeof(struct btrfs_root_ref);
2207 			item_len = btrfs_item_size(leaf, slot)
2208 					- sizeof(struct btrfs_root_ref);
2209 			read_extent_buffer(leaf, subvol_info->name,
2210 					   item_off, item_len);
2211 		} else {
2212 			ret = -ENOENT;
2213 			goto out;
2214 		}
2215 	}
2216 
2217 	btrfs_free_path(path);
2218 	path = NULL;
2219 	if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2220 		ret = -EFAULT;
2221 
2222 out:
2223 	btrfs_put_root(root);
2224 out_free:
2225 	btrfs_free_path(path);
2226 	kfree(subvol_info);
2227 	return ret;
2228 }
2229 
2230 /*
2231  * Return ROOT_REF information of the subvolume containing this inode
2232  * except the subvolume name.
2233  */
2234 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2235 					  void __user *argp)
2236 {
2237 	struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2238 	struct btrfs_root_ref *rref;
2239 	struct btrfs_path *path;
2240 	struct btrfs_key key;
2241 	struct extent_buffer *leaf;
2242 	u64 objectid;
2243 	int slot;
2244 	int ret;
2245 	u8 found;
2246 
2247 	path = btrfs_alloc_path();
2248 	if (!path)
2249 		return -ENOMEM;
2250 
2251 	rootrefs = memdup_user(argp, sizeof(*rootrefs));
2252 	if (IS_ERR(rootrefs)) {
2253 		btrfs_free_path(path);
2254 		return PTR_ERR(rootrefs);
2255 	}
2256 
2257 	objectid = root->root_key.objectid;
2258 	key.objectid = objectid;
2259 	key.type = BTRFS_ROOT_REF_KEY;
2260 	key.offset = rootrefs->min_treeid;
2261 	found = 0;
2262 
2263 	root = root->fs_info->tree_root;
2264 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2265 	if (ret < 0) {
2266 		goto out;
2267 	} else if (path->slots[0] >=
2268 		   btrfs_header_nritems(path->nodes[0])) {
2269 		ret = btrfs_next_leaf(root, path);
2270 		if (ret < 0) {
2271 			goto out;
2272 		} else if (ret > 0) {
2273 			ret = -EUCLEAN;
2274 			goto out;
2275 		}
2276 	}
2277 	while (1) {
2278 		leaf = path->nodes[0];
2279 		slot = path->slots[0];
2280 
2281 		btrfs_item_key_to_cpu(leaf, &key, slot);
2282 		if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2283 			ret = 0;
2284 			goto out;
2285 		}
2286 
2287 		if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2288 			ret = -EOVERFLOW;
2289 			goto out;
2290 		}
2291 
2292 		rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2293 		rootrefs->rootref[found].treeid = key.offset;
2294 		rootrefs->rootref[found].dirid =
2295 				  btrfs_root_ref_dirid(leaf, rref);
2296 		found++;
2297 
2298 		ret = btrfs_next_item(root, path);
2299 		if (ret < 0) {
2300 			goto out;
2301 		} else if (ret > 0) {
2302 			ret = -EUCLEAN;
2303 			goto out;
2304 		}
2305 	}
2306 
2307 out:
2308 	btrfs_free_path(path);
2309 
2310 	if (!ret || ret == -EOVERFLOW) {
2311 		rootrefs->num_items = found;
2312 		/* update min_treeid for next search */
2313 		if (found)
2314 			rootrefs->min_treeid =
2315 				rootrefs->rootref[found - 1].treeid + 1;
2316 		if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2317 			ret = -EFAULT;
2318 	}
2319 
2320 	kfree(rootrefs);
2321 
2322 	return ret;
2323 }
2324 
2325 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2326 					     void __user *arg,
2327 					     bool destroy_v2)
2328 {
2329 	struct dentry *parent = file->f_path.dentry;
2330 	struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2331 	struct dentry *dentry;
2332 	struct inode *dir = d_inode(parent);
2333 	struct inode *inode;
2334 	struct btrfs_root *root = BTRFS_I(dir)->root;
2335 	struct btrfs_root *dest = NULL;
2336 	struct btrfs_ioctl_vol_args *vol_args = NULL;
2337 	struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2338 	struct mnt_idmap *idmap = file_mnt_idmap(file);
2339 	char *subvol_name, *subvol_name_ptr = NULL;
2340 	int subvol_namelen;
2341 	int err = 0;
2342 	bool destroy_parent = false;
2343 
2344 	/* We don't support snapshots with extent tree v2 yet. */
2345 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2346 		btrfs_err(fs_info,
2347 			  "extent tree v2 doesn't support snapshot deletion yet");
2348 		return -EOPNOTSUPP;
2349 	}
2350 
2351 	if (destroy_v2) {
2352 		vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2353 		if (IS_ERR(vol_args2))
2354 			return PTR_ERR(vol_args2);
2355 
2356 		if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2357 			err = -EOPNOTSUPP;
2358 			goto out;
2359 		}
2360 
2361 		/*
2362 		 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2363 		 * name, same as v1 currently does.
2364 		 */
2365 		if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2366 			vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2367 			subvol_name = vol_args2->name;
2368 
2369 			err = mnt_want_write_file(file);
2370 			if (err)
2371 				goto out;
2372 		} else {
2373 			struct inode *old_dir;
2374 
2375 			if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2376 				err = -EINVAL;
2377 				goto out;
2378 			}
2379 
2380 			err = mnt_want_write_file(file);
2381 			if (err)
2382 				goto out;
2383 
2384 			dentry = btrfs_get_dentry(fs_info->sb,
2385 					BTRFS_FIRST_FREE_OBJECTID,
2386 					vol_args2->subvolid, 0);
2387 			if (IS_ERR(dentry)) {
2388 				err = PTR_ERR(dentry);
2389 				goto out_drop_write;
2390 			}
2391 
2392 			/*
2393 			 * Change the default parent since the subvolume being
2394 			 * deleted can be outside of the current mount point.
2395 			 */
2396 			parent = btrfs_get_parent(dentry);
2397 
2398 			/*
2399 			 * At this point dentry->d_name can point to '/' if the
2400 			 * subvolume we want to destroy is outsite of the
2401 			 * current mount point, so we need to release the
2402 			 * current dentry and execute the lookup to return a new
2403 			 * one with ->d_name pointing to the
2404 			 * <mount point>/subvol_name.
2405 			 */
2406 			dput(dentry);
2407 			if (IS_ERR(parent)) {
2408 				err = PTR_ERR(parent);
2409 				goto out_drop_write;
2410 			}
2411 			old_dir = dir;
2412 			dir = d_inode(parent);
2413 
2414 			/*
2415 			 * If v2 was used with SPEC_BY_ID, a new parent was
2416 			 * allocated since the subvolume can be outside of the
2417 			 * current mount point. Later on we need to release this
2418 			 * new parent dentry.
2419 			 */
2420 			destroy_parent = true;
2421 
2422 			/*
2423 			 * On idmapped mounts, deletion via subvolid is
2424 			 * restricted to subvolumes that are immediate
2425 			 * ancestors of the inode referenced by the file
2426 			 * descriptor in the ioctl. Otherwise the idmapping
2427 			 * could potentially be abused to delete subvolumes
2428 			 * anywhere in the filesystem the user wouldn't be able
2429 			 * to delete without an idmapped mount.
2430 			 */
2431 			if (old_dir != dir && idmap != &nop_mnt_idmap) {
2432 				err = -EOPNOTSUPP;
2433 				goto free_parent;
2434 			}
2435 
2436 			subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2437 						fs_info, vol_args2->subvolid);
2438 			if (IS_ERR(subvol_name_ptr)) {
2439 				err = PTR_ERR(subvol_name_ptr);
2440 				goto free_parent;
2441 			}
2442 			/* subvol_name_ptr is already nul terminated */
2443 			subvol_name = (char *)kbasename(subvol_name_ptr);
2444 		}
2445 	} else {
2446 		vol_args = memdup_user(arg, sizeof(*vol_args));
2447 		if (IS_ERR(vol_args))
2448 			return PTR_ERR(vol_args);
2449 
2450 		vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2451 		subvol_name = vol_args->name;
2452 
2453 		err = mnt_want_write_file(file);
2454 		if (err)
2455 			goto out;
2456 	}
2457 
2458 	subvol_namelen = strlen(subvol_name);
2459 
2460 	if (strchr(subvol_name, '/') ||
2461 	    strncmp(subvol_name, "..", subvol_namelen) == 0) {
2462 		err = -EINVAL;
2463 		goto free_subvol_name;
2464 	}
2465 
2466 	if (!S_ISDIR(dir->i_mode)) {
2467 		err = -ENOTDIR;
2468 		goto free_subvol_name;
2469 	}
2470 
2471 	err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2472 	if (err == -EINTR)
2473 		goto free_subvol_name;
2474 	dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2475 	if (IS_ERR(dentry)) {
2476 		err = PTR_ERR(dentry);
2477 		goto out_unlock_dir;
2478 	}
2479 
2480 	if (d_really_is_negative(dentry)) {
2481 		err = -ENOENT;
2482 		goto out_dput;
2483 	}
2484 
2485 	inode = d_inode(dentry);
2486 	dest = BTRFS_I(inode)->root;
2487 	if (!capable(CAP_SYS_ADMIN)) {
2488 		/*
2489 		 * Regular user.  Only allow this with a special mount
2490 		 * option, when the user has write+exec access to the
2491 		 * subvol root, and when rmdir(2) would have been
2492 		 * allowed.
2493 		 *
2494 		 * Note that this is _not_ check that the subvol is
2495 		 * empty or doesn't contain data that we wouldn't
2496 		 * otherwise be able to delete.
2497 		 *
2498 		 * Users who want to delete empty subvols should try
2499 		 * rmdir(2).
2500 		 */
2501 		err = -EPERM;
2502 		if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2503 			goto out_dput;
2504 
2505 		/*
2506 		 * Do not allow deletion if the parent dir is the same
2507 		 * as the dir to be deleted.  That means the ioctl
2508 		 * must be called on the dentry referencing the root
2509 		 * of the subvol, not a random directory contained
2510 		 * within it.
2511 		 */
2512 		err = -EINVAL;
2513 		if (root == dest)
2514 			goto out_dput;
2515 
2516 		err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2517 		if (err)
2518 			goto out_dput;
2519 	}
2520 
2521 	/* check if subvolume may be deleted by a user */
2522 	err = btrfs_may_delete(idmap, dir, dentry, 1);
2523 	if (err)
2524 		goto out_dput;
2525 
2526 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2527 		err = -EINVAL;
2528 		goto out_dput;
2529 	}
2530 
2531 	btrfs_inode_lock(BTRFS_I(inode), 0);
2532 	err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2533 	btrfs_inode_unlock(BTRFS_I(inode), 0);
2534 	if (!err)
2535 		d_delete_notify(dir, dentry);
2536 
2537 out_dput:
2538 	dput(dentry);
2539 out_unlock_dir:
2540 	btrfs_inode_unlock(BTRFS_I(dir), 0);
2541 free_subvol_name:
2542 	kfree(subvol_name_ptr);
2543 free_parent:
2544 	if (destroy_parent)
2545 		dput(parent);
2546 out_drop_write:
2547 	mnt_drop_write_file(file);
2548 out:
2549 	kfree(vol_args2);
2550 	kfree(vol_args);
2551 	return err;
2552 }
2553 
2554 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2555 {
2556 	struct inode *inode = file_inode(file);
2557 	struct btrfs_root *root = BTRFS_I(inode)->root;
2558 	struct btrfs_ioctl_defrag_range_args range = {0};
2559 	int ret;
2560 
2561 	ret = mnt_want_write_file(file);
2562 	if (ret)
2563 		return ret;
2564 
2565 	if (btrfs_root_readonly(root)) {
2566 		ret = -EROFS;
2567 		goto out;
2568 	}
2569 
2570 	switch (inode->i_mode & S_IFMT) {
2571 	case S_IFDIR:
2572 		if (!capable(CAP_SYS_ADMIN)) {
2573 			ret = -EPERM;
2574 			goto out;
2575 		}
2576 		ret = btrfs_defrag_root(root);
2577 		break;
2578 	case S_IFREG:
2579 		/*
2580 		 * Note that this does not check the file descriptor for write
2581 		 * access. This prevents defragmenting executables that are
2582 		 * running and allows defrag on files open in read-only mode.
2583 		 */
2584 		if (!capable(CAP_SYS_ADMIN) &&
2585 		    inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2586 			ret = -EPERM;
2587 			goto out;
2588 		}
2589 
2590 		if (argp) {
2591 			if (copy_from_user(&range, argp, sizeof(range))) {
2592 				ret = -EFAULT;
2593 				goto out;
2594 			}
2595 			/* compression requires us to start the IO */
2596 			if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2597 				range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2598 				range.extent_thresh = (u32)-1;
2599 			}
2600 		} else {
2601 			/* the rest are all set to zero by kzalloc */
2602 			range.len = (u64)-1;
2603 		}
2604 		ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2605 					&range, BTRFS_OLDEST_GENERATION, 0);
2606 		if (ret > 0)
2607 			ret = 0;
2608 		break;
2609 	default:
2610 		ret = -EINVAL;
2611 	}
2612 out:
2613 	mnt_drop_write_file(file);
2614 	return ret;
2615 }
2616 
2617 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2618 {
2619 	struct btrfs_ioctl_vol_args *vol_args;
2620 	bool restore_op = false;
2621 	int ret;
2622 
2623 	if (!capable(CAP_SYS_ADMIN))
2624 		return -EPERM;
2625 
2626 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2627 		btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2628 		return -EINVAL;
2629 	}
2630 
2631 	if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2632 		if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2633 			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2634 
2635 		/*
2636 		 * We can do the device add because we have a paused balanced,
2637 		 * change the exclusive op type and remember we should bring
2638 		 * back the paused balance
2639 		 */
2640 		fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2641 		btrfs_exclop_start_unlock(fs_info);
2642 		restore_op = true;
2643 	}
2644 
2645 	vol_args = memdup_user(arg, sizeof(*vol_args));
2646 	if (IS_ERR(vol_args)) {
2647 		ret = PTR_ERR(vol_args);
2648 		goto out;
2649 	}
2650 
2651 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2652 	ret = btrfs_init_new_device(fs_info, vol_args->name);
2653 
2654 	if (!ret)
2655 		btrfs_info(fs_info, "disk added %s", vol_args->name);
2656 
2657 	kfree(vol_args);
2658 out:
2659 	if (restore_op)
2660 		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2661 	else
2662 		btrfs_exclop_finish(fs_info);
2663 	return ret;
2664 }
2665 
2666 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2667 {
2668 	BTRFS_DEV_LOOKUP_ARGS(args);
2669 	struct inode *inode = file_inode(file);
2670 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2671 	struct btrfs_ioctl_vol_args_v2 *vol_args;
2672 	struct block_device *bdev = NULL;
2673 	fmode_t mode;
2674 	int ret;
2675 	bool cancel = false;
2676 
2677 	if (!capable(CAP_SYS_ADMIN))
2678 		return -EPERM;
2679 
2680 	vol_args = memdup_user(arg, sizeof(*vol_args));
2681 	if (IS_ERR(vol_args))
2682 		return PTR_ERR(vol_args);
2683 
2684 	if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2685 		ret = -EOPNOTSUPP;
2686 		goto out;
2687 	}
2688 
2689 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2690 	if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2691 		args.devid = vol_args->devid;
2692 	} else if (!strcmp("cancel", vol_args->name)) {
2693 		cancel = true;
2694 	} else {
2695 		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2696 		if (ret)
2697 			goto out;
2698 	}
2699 
2700 	ret = mnt_want_write_file(file);
2701 	if (ret)
2702 		goto out;
2703 
2704 	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2705 					   cancel);
2706 	if (ret)
2707 		goto err_drop;
2708 
2709 	/* Exclusive operation is now claimed */
2710 	ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
2711 
2712 	btrfs_exclop_finish(fs_info);
2713 
2714 	if (!ret) {
2715 		if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2716 			btrfs_info(fs_info, "device deleted: id %llu",
2717 					vol_args->devid);
2718 		else
2719 			btrfs_info(fs_info, "device deleted: %s",
2720 					vol_args->name);
2721 	}
2722 err_drop:
2723 	mnt_drop_write_file(file);
2724 	if (bdev)
2725 		blkdev_put(bdev, mode);
2726 out:
2727 	btrfs_put_dev_args_from_path(&args);
2728 	kfree(vol_args);
2729 	return ret;
2730 }
2731 
2732 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2733 {
2734 	BTRFS_DEV_LOOKUP_ARGS(args);
2735 	struct inode *inode = file_inode(file);
2736 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2737 	struct btrfs_ioctl_vol_args *vol_args;
2738 	struct block_device *bdev = NULL;
2739 	fmode_t mode;
2740 	int ret;
2741 	bool cancel = false;
2742 
2743 	if (!capable(CAP_SYS_ADMIN))
2744 		return -EPERM;
2745 
2746 	vol_args = memdup_user(arg, sizeof(*vol_args));
2747 	if (IS_ERR(vol_args))
2748 		return PTR_ERR(vol_args);
2749 
2750 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2751 	if (!strcmp("cancel", vol_args->name)) {
2752 		cancel = true;
2753 	} else {
2754 		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2755 		if (ret)
2756 			goto out;
2757 	}
2758 
2759 	ret = mnt_want_write_file(file);
2760 	if (ret)
2761 		goto out;
2762 
2763 	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2764 					   cancel);
2765 	if (ret == 0) {
2766 		ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
2767 		if (!ret)
2768 			btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2769 		btrfs_exclop_finish(fs_info);
2770 	}
2771 
2772 	mnt_drop_write_file(file);
2773 	if (bdev)
2774 		blkdev_put(bdev, mode);
2775 out:
2776 	btrfs_put_dev_args_from_path(&args);
2777 	kfree(vol_args);
2778 	return ret;
2779 }
2780 
2781 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2782 				void __user *arg)
2783 {
2784 	struct btrfs_ioctl_fs_info_args *fi_args;
2785 	struct btrfs_device *device;
2786 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2787 	u64 flags_in;
2788 	int ret = 0;
2789 
2790 	fi_args = memdup_user(arg, sizeof(*fi_args));
2791 	if (IS_ERR(fi_args))
2792 		return PTR_ERR(fi_args);
2793 
2794 	flags_in = fi_args->flags;
2795 	memset(fi_args, 0, sizeof(*fi_args));
2796 
2797 	rcu_read_lock();
2798 	fi_args->num_devices = fs_devices->num_devices;
2799 
2800 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2801 		if (device->devid > fi_args->max_id)
2802 			fi_args->max_id = device->devid;
2803 	}
2804 	rcu_read_unlock();
2805 
2806 	memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2807 	fi_args->nodesize = fs_info->nodesize;
2808 	fi_args->sectorsize = fs_info->sectorsize;
2809 	fi_args->clone_alignment = fs_info->sectorsize;
2810 
2811 	if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2812 		fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2813 		fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2814 		fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2815 	}
2816 
2817 	if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2818 		fi_args->generation = fs_info->generation;
2819 		fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2820 	}
2821 
2822 	if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2823 		memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2824 		       sizeof(fi_args->metadata_uuid));
2825 		fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2826 	}
2827 
2828 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2829 		ret = -EFAULT;
2830 
2831 	kfree(fi_args);
2832 	return ret;
2833 }
2834 
2835 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2836 				 void __user *arg)
2837 {
2838 	BTRFS_DEV_LOOKUP_ARGS(args);
2839 	struct btrfs_ioctl_dev_info_args *di_args;
2840 	struct btrfs_device *dev;
2841 	int ret = 0;
2842 
2843 	di_args = memdup_user(arg, sizeof(*di_args));
2844 	if (IS_ERR(di_args))
2845 		return PTR_ERR(di_args);
2846 
2847 	args.devid = di_args->devid;
2848 	if (!btrfs_is_empty_uuid(di_args->uuid))
2849 		args.uuid = di_args->uuid;
2850 
2851 	rcu_read_lock();
2852 	dev = btrfs_find_device(fs_info->fs_devices, &args);
2853 	if (!dev) {
2854 		ret = -ENODEV;
2855 		goto out;
2856 	}
2857 
2858 	di_args->devid = dev->devid;
2859 	di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2860 	di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2861 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2862 	memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2863 	if (dev->name)
2864 		strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2865 	else
2866 		di_args->path[0] = '\0';
2867 
2868 out:
2869 	rcu_read_unlock();
2870 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2871 		ret = -EFAULT;
2872 
2873 	kfree(di_args);
2874 	return ret;
2875 }
2876 
2877 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2878 {
2879 	struct inode *inode = file_inode(file);
2880 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2881 	struct btrfs_root *root = BTRFS_I(inode)->root;
2882 	struct btrfs_root *new_root;
2883 	struct btrfs_dir_item *di;
2884 	struct btrfs_trans_handle *trans;
2885 	struct btrfs_path *path = NULL;
2886 	struct btrfs_disk_key disk_key;
2887 	struct fscrypt_str name = FSTR_INIT("default", 7);
2888 	u64 objectid = 0;
2889 	u64 dir_id;
2890 	int ret;
2891 
2892 	if (!capable(CAP_SYS_ADMIN))
2893 		return -EPERM;
2894 
2895 	ret = mnt_want_write_file(file);
2896 	if (ret)
2897 		return ret;
2898 
2899 	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2900 		ret = -EFAULT;
2901 		goto out;
2902 	}
2903 
2904 	if (!objectid)
2905 		objectid = BTRFS_FS_TREE_OBJECTID;
2906 
2907 	new_root = btrfs_get_fs_root(fs_info, objectid, true);
2908 	if (IS_ERR(new_root)) {
2909 		ret = PTR_ERR(new_root);
2910 		goto out;
2911 	}
2912 	if (!is_fstree(new_root->root_key.objectid)) {
2913 		ret = -ENOENT;
2914 		goto out_free;
2915 	}
2916 
2917 	path = btrfs_alloc_path();
2918 	if (!path) {
2919 		ret = -ENOMEM;
2920 		goto out_free;
2921 	}
2922 
2923 	trans = btrfs_start_transaction(root, 1);
2924 	if (IS_ERR(trans)) {
2925 		ret = PTR_ERR(trans);
2926 		goto out_free;
2927 	}
2928 
2929 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
2930 	di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2931 				   dir_id, &name, 1);
2932 	if (IS_ERR_OR_NULL(di)) {
2933 		btrfs_release_path(path);
2934 		btrfs_end_transaction(trans);
2935 		btrfs_err(fs_info,
2936 			  "Umm, you don't have the default diritem, this isn't going to work");
2937 		ret = -ENOENT;
2938 		goto out_free;
2939 	}
2940 
2941 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2942 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2943 	btrfs_mark_buffer_dirty(path->nodes[0]);
2944 	btrfs_release_path(path);
2945 
2946 	btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2947 	btrfs_end_transaction(trans);
2948 out_free:
2949 	btrfs_put_root(new_root);
2950 	btrfs_free_path(path);
2951 out:
2952 	mnt_drop_write_file(file);
2953 	return ret;
2954 }
2955 
2956 static void get_block_group_info(struct list_head *groups_list,
2957 				 struct btrfs_ioctl_space_info *space)
2958 {
2959 	struct btrfs_block_group *block_group;
2960 
2961 	space->total_bytes = 0;
2962 	space->used_bytes = 0;
2963 	space->flags = 0;
2964 	list_for_each_entry(block_group, groups_list, list) {
2965 		space->flags = block_group->flags;
2966 		space->total_bytes += block_group->length;
2967 		space->used_bytes += block_group->used;
2968 	}
2969 }
2970 
2971 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
2972 				   void __user *arg)
2973 {
2974 	struct btrfs_ioctl_space_args space_args;
2975 	struct btrfs_ioctl_space_info space;
2976 	struct btrfs_ioctl_space_info *dest;
2977 	struct btrfs_ioctl_space_info *dest_orig;
2978 	struct btrfs_ioctl_space_info __user *user_dest;
2979 	struct btrfs_space_info *info;
2980 	static const u64 types[] = {
2981 		BTRFS_BLOCK_GROUP_DATA,
2982 		BTRFS_BLOCK_GROUP_SYSTEM,
2983 		BTRFS_BLOCK_GROUP_METADATA,
2984 		BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
2985 	};
2986 	int num_types = 4;
2987 	int alloc_size;
2988 	int ret = 0;
2989 	u64 slot_count = 0;
2990 	int i, c;
2991 
2992 	if (copy_from_user(&space_args,
2993 			   (struct btrfs_ioctl_space_args __user *)arg,
2994 			   sizeof(space_args)))
2995 		return -EFAULT;
2996 
2997 	for (i = 0; i < num_types; i++) {
2998 		struct btrfs_space_info *tmp;
2999 
3000 		info = NULL;
3001 		list_for_each_entry(tmp, &fs_info->space_info, list) {
3002 			if (tmp->flags == types[i]) {
3003 				info = tmp;
3004 				break;
3005 			}
3006 		}
3007 
3008 		if (!info)
3009 			continue;
3010 
3011 		down_read(&info->groups_sem);
3012 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3013 			if (!list_empty(&info->block_groups[c]))
3014 				slot_count++;
3015 		}
3016 		up_read(&info->groups_sem);
3017 	}
3018 
3019 	/*
3020 	 * Global block reserve, exported as a space_info
3021 	 */
3022 	slot_count++;
3023 
3024 	/* space_slots == 0 means they are asking for a count */
3025 	if (space_args.space_slots == 0) {
3026 		space_args.total_spaces = slot_count;
3027 		goto out;
3028 	}
3029 
3030 	slot_count = min_t(u64, space_args.space_slots, slot_count);
3031 
3032 	alloc_size = sizeof(*dest) * slot_count;
3033 
3034 	/* we generally have at most 6 or so space infos, one for each raid
3035 	 * level.  So, a whole page should be more than enough for everyone
3036 	 */
3037 	if (alloc_size > PAGE_SIZE)
3038 		return -ENOMEM;
3039 
3040 	space_args.total_spaces = 0;
3041 	dest = kmalloc(alloc_size, GFP_KERNEL);
3042 	if (!dest)
3043 		return -ENOMEM;
3044 	dest_orig = dest;
3045 
3046 	/* now we have a buffer to copy into */
3047 	for (i = 0; i < num_types; i++) {
3048 		struct btrfs_space_info *tmp;
3049 
3050 		if (!slot_count)
3051 			break;
3052 
3053 		info = NULL;
3054 		list_for_each_entry(tmp, &fs_info->space_info, list) {
3055 			if (tmp->flags == types[i]) {
3056 				info = tmp;
3057 				break;
3058 			}
3059 		}
3060 
3061 		if (!info)
3062 			continue;
3063 		down_read(&info->groups_sem);
3064 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3065 			if (!list_empty(&info->block_groups[c])) {
3066 				get_block_group_info(&info->block_groups[c],
3067 						     &space);
3068 				memcpy(dest, &space, sizeof(space));
3069 				dest++;
3070 				space_args.total_spaces++;
3071 				slot_count--;
3072 			}
3073 			if (!slot_count)
3074 				break;
3075 		}
3076 		up_read(&info->groups_sem);
3077 	}
3078 
3079 	/*
3080 	 * Add global block reserve
3081 	 */
3082 	if (slot_count) {
3083 		struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3084 
3085 		spin_lock(&block_rsv->lock);
3086 		space.total_bytes = block_rsv->size;
3087 		space.used_bytes = block_rsv->size - block_rsv->reserved;
3088 		spin_unlock(&block_rsv->lock);
3089 		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3090 		memcpy(dest, &space, sizeof(space));
3091 		space_args.total_spaces++;
3092 	}
3093 
3094 	user_dest = (struct btrfs_ioctl_space_info __user *)
3095 		(arg + sizeof(struct btrfs_ioctl_space_args));
3096 
3097 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3098 		ret = -EFAULT;
3099 
3100 	kfree(dest_orig);
3101 out:
3102 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3103 		ret = -EFAULT;
3104 
3105 	return ret;
3106 }
3107 
3108 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3109 					    void __user *argp)
3110 {
3111 	struct btrfs_trans_handle *trans;
3112 	u64 transid;
3113 
3114 	trans = btrfs_attach_transaction_barrier(root);
3115 	if (IS_ERR(trans)) {
3116 		if (PTR_ERR(trans) != -ENOENT)
3117 			return PTR_ERR(trans);
3118 
3119 		/* No running transaction, don't bother */
3120 		transid = root->fs_info->last_trans_committed;
3121 		goto out;
3122 	}
3123 	transid = trans->transid;
3124 	btrfs_commit_transaction_async(trans);
3125 out:
3126 	if (argp)
3127 		if (copy_to_user(argp, &transid, sizeof(transid)))
3128 			return -EFAULT;
3129 	return 0;
3130 }
3131 
3132 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3133 					   void __user *argp)
3134 {
3135 	u64 transid;
3136 
3137 	if (argp) {
3138 		if (copy_from_user(&transid, argp, sizeof(transid)))
3139 			return -EFAULT;
3140 	} else {
3141 		transid = 0;  /* current trans */
3142 	}
3143 	return btrfs_wait_for_commit(fs_info, transid);
3144 }
3145 
3146 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3147 {
3148 	struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3149 	struct btrfs_ioctl_scrub_args *sa;
3150 	int ret;
3151 
3152 	if (!capable(CAP_SYS_ADMIN))
3153 		return -EPERM;
3154 
3155 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3156 		btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3157 		return -EINVAL;
3158 	}
3159 
3160 	sa = memdup_user(arg, sizeof(*sa));
3161 	if (IS_ERR(sa))
3162 		return PTR_ERR(sa);
3163 
3164 	if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3165 		ret = -EOPNOTSUPP;
3166 		goto out;
3167 	}
3168 
3169 	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3170 		ret = mnt_want_write_file(file);
3171 		if (ret)
3172 			goto out;
3173 	}
3174 
3175 	ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3176 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3177 			      0);
3178 
3179 	/*
3180 	 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3181 	 * error. This is important as it allows user space to know how much
3182 	 * progress scrub has done. For example, if scrub is canceled we get
3183 	 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3184 	 * space. Later user space can inspect the progress from the structure
3185 	 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3186 	 * previously (btrfs-progs does this).
3187 	 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3188 	 * then return -EFAULT to signal the structure was not copied or it may
3189 	 * be corrupt and unreliable due to a partial copy.
3190 	 */
3191 	if (copy_to_user(arg, sa, sizeof(*sa)))
3192 		ret = -EFAULT;
3193 
3194 	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3195 		mnt_drop_write_file(file);
3196 out:
3197 	kfree(sa);
3198 	return ret;
3199 }
3200 
3201 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3202 {
3203 	if (!capable(CAP_SYS_ADMIN))
3204 		return -EPERM;
3205 
3206 	return btrfs_scrub_cancel(fs_info);
3207 }
3208 
3209 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3210 				       void __user *arg)
3211 {
3212 	struct btrfs_ioctl_scrub_args *sa;
3213 	int ret;
3214 
3215 	if (!capable(CAP_SYS_ADMIN))
3216 		return -EPERM;
3217 
3218 	sa = memdup_user(arg, sizeof(*sa));
3219 	if (IS_ERR(sa))
3220 		return PTR_ERR(sa);
3221 
3222 	ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3223 
3224 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3225 		ret = -EFAULT;
3226 
3227 	kfree(sa);
3228 	return ret;
3229 }
3230 
3231 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3232 				      void __user *arg)
3233 {
3234 	struct btrfs_ioctl_get_dev_stats *sa;
3235 	int ret;
3236 
3237 	sa = memdup_user(arg, sizeof(*sa));
3238 	if (IS_ERR(sa))
3239 		return PTR_ERR(sa);
3240 
3241 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3242 		kfree(sa);
3243 		return -EPERM;
3244 	}
3245 
3246 	ret = btrfs_get_dev_stats(fs_info, sa);
3247 
3248 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3249 		ret = -EFAULT;
3250 
3251 	kfree(sa);
3252 	return ret;
3253 }
3254 
3255 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3256 				    void __user *arg)
3257 {
3258 	struct btrfs_ioctl_dev_replace_args *p;
3259 	int ret;
3260 
3261 	if (!capable(CAP_SYS_ADMIN))
3262 		return -EPERM;
3263 
3264 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3265 		btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3266 		return -EINVAL;
3267 	}
3268 
3269 	p = memdup_user(arg, sizeof(*p));
3270 	if (IS_ERR(p))
3271 		return PTR_ERR(p);
3272 
3273 	switch (p->cmd) {
3274 	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3275 		if (sb_rdonly(fs_info->sb)) {
3276 			ret = -EROFS;
3277 			goto out;
3278 		}
3279 		if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3280 			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3281 		} else {
3282 			ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3283 			btrfs_exclop_finish(fs_info);
3284 		}
3285 		break;
3286 	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3287 		btrfs_dev_replace_status(fs_info, p);
3288 		ret = 0;
3289 		break;
3290 	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3291 		p->result = btrfs_dev_replace_cancel(fs_info);
3292 		ret = 0;
3293 		break;
3294 	default:
3295 		ret = -EINVAL;
3296 		break;
3297 	}
3298 
3299 	if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3300 		ret = -EFAULT;
3301 out:
3302 	kfree(p);
3303 	return ret;
3304 }
3305 
3306 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3307 {
3308 	int ret = 0;
3309 	int i;
3310 	u64 rel_ptr;
3311 	int size;
3312 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3313 	struct inode_fs_paths *ipath = NULL;
3314 	struct btrfs_path *path;
3315 
3316 	if (!capable(CAP_DAC_READ_SEARCH))
3317 		return -EPERM;
3318 
3319 	path = btrfs_alloc_path();
3320 	if (!path) {
3321 		ret = -ENOMEM;
3322 		goto out;
3323 	}
3324 
3325 	ipa = memdup_user(arg, sizeof(*ipa));
3326 	if (IS_ERR(ipa)) {
3327 		ret = PTR_ERR(ipa);
3328 		ipa = NULL;
3329 		goto out;
3330 	}
3331 
3332 	size = min_t(u32, ipa->size, 4096);
3333 	ipath = init_ipath(size, root, path);
3334 	if (IS_ERR(ipath)) {
3335 		ret = PTR_ERR(ipath);
3336 		ipath = NULL;
3337 		goto out;
3338 	}
3339 
3340 	ret = paths_from_inode(ipa->inum, ipath);
3341 	if (ret < 0)
3342 		goto out;
3343 
3344 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3345 		rel_ptr = ipath->fspath->val[i] -
3346 			  (u64)(unsigned long)ipath->fspath->val;
3347 		ipath->fspath->val[i] = rel_ptr;
3348 	}
3349 
3350 	btrfs_free_path(path);
3351 	path = NULL;
3352 	ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3353 			   ipath->fspath, size);
3354 	if (ret) {
3355 		ret = -EFAULT;
3356 		goto out;
3357 	}
3358 
3359 out:
3360 	btrfs_free_path(path);
3361 	free_ipath(ipath);
3362 	kfree(ipa);
3363 
3364 	return ret;
3365 }
3366 
3367 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3368 					void __user *arg, int version)
3369 {
3370 	int ret = 0;
3371 	int size;
3372 	struct btrfs_ioctl_logical_ino_args *loi;
3373 	struct btrfs_data_container *inodes = NULL;
3374 	struct btrfs_path *path = NULL;
3375 	bool ignore_offset;
3376 
3377 	if (!capable(CAP_SYS_ADMIN))
3378 		return -EPERM;
3379 
3380 	loi = memdup_user(arg, sizeof(*loi));
3381 	if (IS_ERR(loi))
3382 		return PTR_ERR(loi);
3383 
3384 	if (version == 1) {
3385 		ignore_offset = false;
3386 		size = min_t(u32, loi->size, SZ_64K);
3387 	} else {
3388 		/* All reserved bits must be 0 for now */
3389 		if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3390 			ret = -EINVAL;
3391 			goto out_loi;
3392 		}
3393 		/* Only accept flags we have defined so far */
3394 		if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3395 			ret = -EINVAL;
3396 			goto out_loi;
3397 		}
3398 		ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3399 		size = min_t(u32, loi->size, SZ_16M);
3400 	}
3401 
3402 	inodes = init_data_container(size);
3403 	if (IS_ERR(inodes)) {
3404 		ret = PTR_ERR(inodes);
3405 		goto out_loi;
3406 	}
3407 
3408 	path = btrfs_alloc_path();
3409 	if (!path) {
3410 		ret = -ENOMEM;
3411 		goto out;
3412 	}
3413 	ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3414 					  inodes, ignore_offset);
3415 	btrfs_free_path(path);
3416 	if (ret == -EINVAL)
3417 		ret = -ENOENT;
3418 	if (ret < 0)
3419 		goto out;
3420 
3421 	ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3422 			   size);
3423 	if (ret)
3424 		ret = -EFAULT;
3425 
3426 out:
3427 	kvfree(inodes);
3428 out_loi:
3429 	kfree(loi);
3430 
3431 	return ret;
3432 }
3433 
3434 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3435 			       struct btrfs_ioctl_balance_args *bargs)
3436 {
3437 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3438 
3439 	bargs->flags = bctl->flags;
3440 
3441 	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3442 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3443 	if (atomic_read(&fs_info->balance_pause_req))
3444 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3445 	if (atomic_read(&fs_info->balance_cancel_req))
3446 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3447 
3448 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3449 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3450 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3451 
3452 	spin_lock(&fs_info->balance_lock);
3453 	memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3454 	spin_unlock(&fs_info->balance_lock);
3455 }
3456 
3457 /*
3458  * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3459  * required.
3460  *
3461  * @fs_info:       the filesystem
3462  * @excl_acquired: ptr to boolean value which is set to false in case balance
3463  *                 is being resumed
3464  *
3465  * Return 0 on success in which case both fs_info::balance is acquired as well
3466  * as exclusive ops are blocked. In case of failure return an error code.
3467  */
3468 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3469 {
3470 	int ret;
3471 
3472 	/*
3473 	 * Exclusive operation is locked. Three possibilities:
3474 	 *   (1) some other op is running
3475 	 *   (2) balance is running
3476 	 *   (3) balance is paused -- special case (think resume)
3477 	 */
3478 	while (1) {
3479 		if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3480 			*excl_acquired = true;
3481 			mutex_lock(&fs_info->balance_mutex);
3482 			return 0;
3483 		}
3484 
3485 		mutex_lock(&fs_info->balance_mutex);
3486 		if (fs_info->balance_ctl) {
3487 			/* This is either (2) or (3) */
3488 			if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3489 				/* This is (2) */
3490 				ret = -EINPROGRESS;
3491 				goto out_failure;
3492 
3493 			} else {
3494 				mutex_unlock(&fs_info->balance_mutex);
3495 				/*
3496 				 * Lock released to allow other waiters to
3497 				 * continue, we'll reexamine the status again.
3498 				 */
3499 				mutex_lock(&fs_info->balance_mutex);
3500 
3501 				if (fs_info->balance_ctl &&
3502 				    !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3503 					/* This is (3) */
3504 					*excl_acquired = false;
3505 					return 0;
3506 				}
3507 			}
3508 		} else {
3509 			/* This is (1) */
3510 			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3511 			goto out_failure;
3512 		}
3513 
3514 		mutex_unlock(&fs_info->balance_mutex);
3515 	}
3516 
3517 out_failure:
3518 	mutex_unlock(&fs_info->balance_mutex);
3519 	*excl_acquired = false;
3520 	return ret;
3521 }
3522 
3523 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3524 {
3525 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3526 	struct btrfs_fs_info *fs_info = root->fs_info;
3527 	struct btrfs_ioctl_balance_args *bargs;
3528 	struct btrfs_balance_control *bctl;
3529 	bool need_unlock = true;
3530 	int ret;
3531 
3532 	if (!capable(CAP_SYS_ADMIN))
3533 		return -EPERM;
3534 
3535 	ret = mnt_want_write_file(file);
3536 	if (ret)
3537 		return ret;
3538 
3539 	bargs = memdup_user(arg, sizeof(*bargs));
3540 	if (IS_ERR(bargs)) {
3541 		ret = PTR_ERR(bargs);
3542 		bargs = NULL;
3543 		goto out;
3544 	}
3545 
3546 	ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3547 	if (ret)
3548 		goto out;
3549 
3550 	lockdep_assert_held(&fs_info->balance_mutex);
3551 
3552 	if (bargs->flags & BTRFS_BALANCE_RESUME) {
3553 		if (!fs_info->balance_ctl) {
3554 			ret = -ENOTCONN;
3555 			goto out_unlock;
3556 		}
3557 
3558 		bctl = fs_info->balance_ctl;
3559 		spin_lock(&fs_info->balance_lock);
3560 		bctl->flags |= BTRFS_BALANCE_RESUME;
3561 		spin_unlock(&fs_info->balance_lock);
3562 		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3563 
3564 		goto do_balance;
3565 	}
3566 
3567 	if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3568 		ret = -EINVAL;
3569 		goto out_unlock;
3570 	}
3571 
3572 	if (fs_info->balance_ctl) {
3573 		ret = -EINPROGRESS;
3574 		goto out_unlock;
3575 	}
3576 
3577 	bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3578 	if (!bctl) {
3579 		ret = -ENOMEM;
3580 		goto out_unlock;
3581 	}
3582 
3583 	memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3584 	memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3585 	memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3586 
3587 	bctl->flags = bargs->flags;
3588 do_balance:
3589 	/*
3590 	 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3591 	 * bctl is freed in reset_balance_state, or, if restriper was paused
3592 	 * all the way until unmount, in free_fs_info.  The flag should be
3593 	 * cleared after reset_balance_state.
3594 	 */
3595 	need_unlock = false;
3596 
3597 	ret = btrfs_balance(fs_info, bctl, bargs);
3598 	bctl = NULL;
3599 
3600 	if (ret == 0 || ret == -ECANCELED) {
3601 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3602 			ret = -EFAULT;
3603 	}
3604 
3605 	kfree(bctl);
3606 out_unlock:
3607 	mutex_unlock(&fs_info->balance_mutex);
3608 	if (need_unlock)
3609 		btrfs_exclop_finish(fs_info);
3610 out:
3611 	mnt_drop_write_file(file);
3612 	kfree(bargs);
3613 	return ret;
3614 }
3615 
3616 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3617 {
3618 	if (!capable(CAP_SYS_ADMIN))
3619 		return -EPERM;
3620 
3621 	switch (cmd) {
3622 	case BTRFS_BALANCE_CTL_PAUSE:
3623 		return btrfs_pause_balance(fs_info);
3624 	case BTRFS_BALANCE_CTL_CANCEL:
3625 		return btrfs_cancel_balance(fs_info);
3626 	}
3627 
3628 	return -EINVAL;
3629 }
3630 
3631 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3632 					 void __user *arg)
3633 {
3634 	struct btrfs_ioctl_balance_args *bargs;
3635 	int ret = 0;
3636 
3637 	if (!capable(CAP_SYS_ADMIN))
3638 		return -EPERM;
3639 
3640 	mutex_lock(&fs_info->balance_mutex);
3641 	if (!fs_info->balance_ctl) {
3642 		ret = -ENOTCONN;
3643 		goto out;
3644 	}
3645 
3646 	bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3647 	if (!bargs) {
3648 		ret = -ENOMEM;
3649 		goto out;
3650 	}
3651 
3652 	btrfs_update_ioctl_balance_args(fs_info, bargs);
3653 
3654 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3655 		ret = -EFAULT;
3656 
3657 	kfree(bargs);
3658 out:
3659 	mutex_unlock(&fs_info->balance_mutex);
3660 	return ret;
3661 }
3662 
3663 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3664 {
3665 	struct inode *inode = file_inode(file);
3666 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3667 	struct btrfs_ioctl_quota_ctl_args *sa;
3668 	int ret;
3669 
3670 	if (!capable(CAP_SYS_ADMIN))
3671 		return -EPERM;
3672 
3673 	ret = mnt_want_write_file(file);
3674 	if (ret)
3675 		return ret;
3676 
3677 	sa = memdup_user(arg, sizeof(*sa));
3678 	if (IS_ERR(sa)) {
3679 		ret = PTR_ERR(sa);
3680 		goto drop_write;
3681 	}
3682 
3683 	down_write(&fs_info->subvol_sem);
3684 
3685 	switch (sa->cmd) {
3686 	case BTRFS_QUOTA_CTL_ENABLE:
3687 		ret = btrfs_quota_enable(fs_info);
3688 		break;
3689 	case BTRFS_QUOTA_CTL_DISABLE:
3690 		ret = btrfs_quota_disable(fs_info);
3691 		break;
3692 	default:
3693 		ret = -EINVAL;
3694 		break;
3695 	}
3696 
3697 	kfree(sa);
3698 	up_write(&fs_info->subvol_sem);
3699 drop_write:
3700 	mnt_drop_write_file(file);
3701 	return ret;
3702 }
3703 
3704 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3705 {
3706 	struct inode *inode = file_inode(file);
3707 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3708 	struct btrfs_root *root = BTRFS_I(inode)->root;
3709 	struct btrfs_ioctl_qgroup_assign_args *sa;
3710 	struct btrfs_trans_handle *trans;
3711 	int ret;
3712 	int err;
3713 
3714 	if (!capable(CAP_SYS_ADMIN))
3715 		return -EPERM;
3716 
3717 	ret = mnt_want_write_file(file);
3718 	if (ret)
3719 		return ret;
3720 
3721 	sa = memdup_user(arg, sizeof(*sa));
3722 	if (IS_ERR(sa)) {
3723 		ret = PTR_ERR(sa);
3724 		goto drop_write;
3725 	}
3726 
3727 	trans = btrfs_join_transaction(root);
3728 	if (IS_ERR(trans)) {
3729 		ret = PTR_ERR(trans);
3730 		goto out;
3731 	}
3732 
3733 	if (sa->assign) {
3734 		ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3735 	} else {
3736 		ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3737 	}
3738 
3739 	/* update qgroup status and info */
3740 	mutex_lock(&fs_info->qgroup_ioctl_lock);
3741 	err = btrfs_run_qgroups(trans);
3742 	mutex_unlock(&fs_info->qgroup_ioctl_lock);
3743 	if (err < 0)
3744 		btrfs_handle_fs_error(fs_info, err,
3745 				      "failed to update qgroup status and info");
3746 	err = btrfs_end_transaction(trans);
3747 	if (err && !ret)
3748 		ret = err;
3749 
3750 out:
3751 	kfree(sa);
3752 drop_write:
3753 	mnt_drop_write_file(file);
3754 	return ret;
3755 }
3756 
3757 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3758 {
3759 	struct inode *inode = file_inode(file);
3760 	struct btrfs_root *root = BTRFS_I(inode)->root;
3761 	struct btrfs_ioctl_qgroup_create_args *sa;
3762 	struct btrfs_trans_handle *trans;
3763 	int ret;
3764 	int err;
3765 
3766 	if (!capable(CAP_SYS_ADMIN))
3767 		return -EPERM;
3768 
3769 	ret = mnt_want_write_file(file);
3770 	if (ret)
3771 		return ret;
3772 
3773 	sa = memdup_user(arg, sizeof(*sa));
3774 	if (IS_ERR(sa)) {
3775 		ret = PTR_ERR(sa);
3776 		goto drop_write;
3777 	}
3778 
3779 	if (!sa->qgroupid) {
3780 		ret = -EINVAL;
3781 		goto out;
3782 	}
3783 
3784 	trans = btrfs_join_transaction(root);
3785 	if (IS_ERR(trans)) {
3786 		ret = PTR_ERR(trans);
3787 		goto out;
3788 	}
3789 
3790 	if (sa->create) {
3791 		ret = btrfs_create_qgroup(trans, sa->qgroupid);
3792 	} else {
3793 		ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3794 	}
3795 
3796 	err = btrfs_end_transaction(trans);
3797 	if (err && !ret)
3798 		ret = err;
3799 
3800 out:
3801 	kfree(sa);
3802 drop_write:
3803 	mnt_drop_write_file(file);
3804 	return ret;
3805 }
3806 
3807 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3808 {
3809 	struct inode *inode = file_inode(file);
3810 	struct btrfs_root *root = BTRFS_I(inode)->root;
3811 	struct btrfs_ioctl_qgroup_limit_args *sa;
3812 	struct btrfs_trans_handle *trans;
3813 	int ret;
3814 	int err;
3815 	u64 qgroupid;
3816 
3817 	if (!capable(CAP_SYS_ADMIN))
3818 		return -EPERM;
3819 
3820 	ret = mnt_want_write_file(file);
3821 	if (ret)
3822 		return ret;
3823 
3824 	sa = memdup_user(arg, sizeof(*sa));
3825 	if (IS_ERR(sa)) {
3826 		ret = PTR_ERR(sa);
3827 		goto drop_write;
3828 	}
3829 
3830 	trans = btrfs_join_transaction(root);
3831 	if (IS_ERR(trans)) {
3832 		ret = PTR_ERR(trans);
3833 		goto out;
3834 	}
3835 
3836 	qgroupid = sa->qgroupid;
3837 	if (!qgroupid) {
3838 		/* take the current subvol as qgroup */
3839 		qgroupid = root->root_key.objectid;
3840 	}
3841 
3842 	ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3843 
3844 	err = btrfs_end_transaction(trans);
3845 	if (err && !ret)
3846 		ret = err;
3847 
3848 out:
3849 	kfree(sa);
3850 drop_write:
3851 	mnt_drop_write_file(file);
3852 	return ret;
3853 }
3854 
3855 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3856 {
3857 	struct inode *inode = file_inode(file);
3858 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3859 	struct btrfs_ioctl_quota_rescan_args *qsa;
3860 	int ret;
3861 
3862 	if (!capable(CAP_SYS_ADMIN))
3863 		return -EPERM;
3864 
3865 	ret = mnt_want_write_file(file);
3866 	if (ret)
3867 		return ret;
3868 
3869 	qsa = memdup_user(arg, sizeof(*qsa));
3870 	if (IS_ERR(qsa)) {
3871 		ret = PTR_ERR(qsa);
3872 		goto drop_write;
3873 	}
3874 
3875 	if (qsa->flags) {
3876 		ret = -EINVAL;
3877 		goto out;
3878 	}
3879 
3880 	ret = btrfs_qgroup_rescan(fs_info);
3881 
3882 out:
3883 	kfree(qsa);
3884 drop_write:
3885 	mnt_drop_write_file(file);
3886 	return ret;
3887 }
3888 
3889 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3890 						void __user *arg)
3891 {
3892 	struct btrfs_ioctl_quota_rescan_args qsa = {0};
3893 
3894 	if (!capable(CAP_SYS_ADMIN))
3895 		return -EPERM;
3896 
3897 	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3898 		qsa.flags = 1;
3899 		qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3900 	}
3901 
3902 	if (copy_to_user(arg, &qsa, sizeof(qsa)))
3903 		return -EFAULT;
3904 
3905 	return 0;
3906 }
3907 
3908 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3909 						void __user *arg)
3910 {
3911 	if (!capable(CAP_SYS_ADMIN))
3912 		return -EPERM;
3913 
3914 	return btrfs_qgroup_wait_for_completion(fs_info, true);
3915 }
3916 
3917 static long _btrfs_ioctl_set_received_subvol(struct file *file,
3918 					    struct mnt_idmap *idmap,
3919 					    struct btrfs_ioctl_received_subvol_args *sa)
3920 {
3921 	struct inode *inode = file_inode(file);
3922 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3923 	struct btrfs_root *root = BTRFS_I(inode)->root;
3924 	struct btrfs_root_item *root_item = &root->root_item;
3925 	struct btrfs_trans_handle *trans;
3926 	struct timespec64 ct = current_time(inode);
3927 	int ret = 0;
3928 	int received_uuid_changed;
3929 
3930 	if (!inode_owner_or_capable(idmap, inode))
3931 		return -EPERM;
3932 
3933 	ret = mnt_want_write_file(file);
3934 	if (ret < 0)
3935 		return ret;
3936 
3937 	down_write(&fs_info->subvol_sem);
3938 
3939 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3940 		ret = -EINVAL;
3941 		goto out;
3942 	}
3943 
3944 	if (btrfs_root_readonly(root)) {
3945 		ret = -EROFS;
3946 		goto out;
3947 	}
3948 
3949 	/*
3950 	 * 1 - root item
3951 	 * 2 - uuid items (received uuid + subvol uuid)
3952 	 */
3953 	trans = btrfs_start_transaction(root, 3);
3954 	if (IS_ERR(trans)) {
3955 		ret = PTR_ERR(trans);
3956 		trans = NULL;
3957 		goto out;
3958 	}
3959 
3960 	sa->rtransid = trans->transid;
3961 	sa->rtime.sec = ct.tv_sec;
3962 	sa->rtime.nsec = ct.tv_nsec;
3963 
3964 	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
3965 				       BTRFS_UUID_SIZE);
3966 	if (received_uuid_changed &&
3967 	    !btrfs_is_empty_uuid(root_item->received_uuid)) {
3968 		ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
3969 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
3970 					  root->root_key.objectid);
3971 		if (ret && ret != -ENOENT) {
3972 		        btrfs_abort_transaction(trans, ret);
3973 		        btrfs_end_transaction(trans);
3974 		        goto out;
3975 		}
3976 	}
3977 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3978 	btrfs_set_root_stransid(root_item, sa->stransid);
3979 	btrfs_set_root_rtransid(root_item, sa->rtransid);
3980 	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
3981 	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
3982 	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
3983 	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
3984 
3985 	ret = btrfs_update_root(trans, fs_info->tree_root,
3986 				&root->root_key, &root->root_item);
3987 	if (ret < 0) {
3988 		btrfs_end_transaction(trans);
3989 		goto out;
3990 	}
3991 	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
3992 		ret = btrfs_uuid_tree_add(trans, sa->uuid,
3993 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
3994 					  root->root_key.objectid);
3995 		if (ret < 0 && ret != -EEXIST) {
3996 			btrfs_abort_transaction(trans, ret);
3997 			btrfs_end_transaction(trans);
3998 			goto out;
3999 		}
4000 	}
4001 	ret = btrfs_commit_transaction(trans);
4002 out:
4003 	up_write(&fs_info->subvol_sem);
4004 	mnt_drop_write_file(file);
4005 	return ret;
4006 }
4007 
4008 #ifdef CONFIG_64BIT
4009 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4010 						void __user *arg)
4011 {
4012 	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4013 	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4014 	int ret = 0;
4015 
4016 	args32 = memdup_user(arg, sizeof(*args32));
4017 	if (IS_ERR(args32))
4018 		return PTR_ERR(args32);
4019 
4020 	args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4021 	if (!args64) {
4022 		ret = -ENOMEM;
4023 		goto out;
4024 	}
4025 
4026 	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4027 	args64->stransid = args32->stransid;
4028 	args64->rtransid = args32->rtransid;
4029 	args64->stime.sec = args32->stime.sec;
4030 	args64->stime.nsec = args32->stime.nsec;
4031 	args64->rtime.sec = args32->rtime.sec;
4032 	args64->rtime.nsec = args32->rtime.nsec;
4033 	args64->flags = args32->flags;
4034 
4035 	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4036 	if (ret)
4037 		goto out;
4038 
4039 	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4040 	args32->stransid = args64->stransid;
4041 	args32->rtransid = args64->rtransid;
4042 	args32->stime.sec = args64->stime.sec;
4043 	args32->stime.nsec = args64->stime.nsec;
4044 	args32->rtime.sec = args64->rtime.sec;
4045 	args32->rtime.nsec = args64->rtime.nsec;
4046 	args32->flags = args64->flags;
4047 
4048 	ret = copy_to_user(arg, args32, sizeof(*args32));
4049 	if (ret)
4050 		ret = -EFAULT;
4051 
4052 out:
4053 	kfree(args32);
4054 	kfree(args64);
4055 	return ret;
4056 }
4057 #endif
4058 
4059 static long btrfs_ioctl_set_received_subvol(struct file *file,
4060 					    void __user *arg)
4061 {
4062 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4063 	int ret = 0;
4064 
4065 	sa = memdup_user(arg, sizeof(*sa));
4066 	if (IS_ERR(sa))
4067 		return PTR_ERR(sa);
4068 
4069 	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4070 
4071 	if (ret)
4072 		goto out;
4073 
4074 	ret = copy_to_user(arg, sa, sizeof(*sa));
4075 	if (ret)
4076 		ret = -EFAULT;
4077 
4078 out:
4079 	kfree(sa);
4080 	return ret;
4081 }
4082 
4083 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4084 					void __user *arg)
4085 {
4086 	size_t len;
4087 	int ret;
4088 	char label[BTRFS_LABEL_SIZE];
4089 
4090 	spin_lock(&fs_info->super_lock);
4091 	memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4092 	spin_unlock(&fs_info->super_lock);
4093 
4094 	len = strnlen(label, BTRFS_LABEL_SIZE);
4095 
4096 	if (len == BTRFS_LABEL_SIZE) {
4097 		btrfs_warn(fs_info,
4098 			   "label is too long, return the first %zu bytes",
4099 			   --len);
4100 	}
4101 
4102 	ret = copy_to_user(arg, label, len);
4103 
4104 	return ret ? -EFAULT : 0;
4105 }
4106 
4107 static int btrfs_ioctl_set_fslabel(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_super_block *super_block = fs_info->super_copy;
4113 	struct btrfs_trans_handle *trans;
4114 	char label[BTRFS_LABEL_SIZE];
4115 	int ret;
4116 
4117 	if (!capable(CAP_SYS_ADMIN))
4118 		return -EPERM;
4119 
4120 	if (copy_from_user(label, arg, sizeof(label)))
4121 		return -EFAULT;
4122 
4123 	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4124 		btrfs_err(fs_info,
4125 			  "unable to set label with more than %d bytes",
4126 			  BTRFS_LABEL_SIZE - 1);
4127 		return -EINVAL;
4128 	}
4129 
4130 	ret = mnt_want_write_file(file);
4131 	if (ret)
4132 		return ret;
4133 
4134 	trans = btrfs_start_transaction(root, 0);
4135 	if (IS_ERR(trans)) {
4136 		ret = PTR_ERR(trans);
4137 		goto out_unlock;
4138 	}
4139 
4140 	spin_lock(&fs_info->super_lock);
4141 	strcpy(super_block->label, label);
4142 	spin_unlock(&fs_info->super_lock);
4143 	ret = btrfs_commit_transaction(trans);
4144 
4145 out_unlock:
4146 	mnt_drop_write_file(file);
4147 	return ret;
4148 }
4149 
4150 #define INIT_FEATURE_FLAGS(suffix) \
4151 	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4152 	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4153 	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4154 
4155 int btrfs_ioctl_get_supported_features(void __user *arg)
4156 {
4157 	static const struct btrfs_ioctl_feature_flags features[3] = {
4158 		INIT_FEATURE_FLAGS(SUPP),
4159 		INIT_FEATURE_FLAGS(SAFE_SET),
4160 		INIT_FEATURE_FLAGS(SAFE_CLEAR)
4161 	};
4162 
4163 	if (copy_to_user(arg, &features, sizeof(features)))
4164 		return -EFAULT;
4165 
4166 	return 0;
4167 }
4168 
4169 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4170 					void __user *arg)
4171 {
4172 	struct btrfs_super_block *super_block = fs_info->super_copy;
4173 	struct btrfs_ioctl_feature_flags features;
4174 
4175 	features.compat_flags = btrfs_super_compat_flags(super_block);
4176 	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4177 	features.incompat_flags = btrfs_super_incompat_flags(super_block);
4178 
4179 	if (copy_to_user(arg, &features, sizeof(features)))
4180 		return -EFAULT;
4181 
4182 	return 0;
4183 }
4184 
4185 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4186 			      enum btrfs_feature_set set,
4187 			      u64 change_mask, u64 flags, u64 supported_flags,
4188 			      u64 safe_set, u64 safe_clear)
4189 {
4190 	const char *type = btrfs_feature_set_name(set);
4191 	char *names;
4192 	u64 disallowed, unsupported;
4193 	u64 set_mask = flags & change_mask;
4194 	u64 clear_mask = ~flags & change_mask;
4195 
4196 	unsupported = set_mask & ~supported_flags;
4197 	if (unsupported) {
4198 		names = btrfs_printable_features(set, unsupported);
4199 		if (names) {
4200 			btrfs_warn(fs_info,
4201 				   "this kernel does not support the %s feature bit%s",
4202 				   names, strchr(names, ',') ? "s" : "");
4203 			kfree(names);
4204 		} else
4205 			btrfs_warn(fs_info,
4206 				   "this kernel does not support %s bits 0x%llx",
4207 				   type, unsupported);
4208 		return -EOPNOTSUPP;
4209 	}
4210 
4211 	disallowed = set_mask & ~safe_set;
4212 	if (disallowed) {
4213 		names = btrfs_printable_features(set, disallowed);
4214 		if (names) {
4215 			btrfs_warn(fs_info,
4216 				   "can't set the %s feature bit%s while mounted",
4217 				   names, strchr(names, ',') ? "s" : "");
4218 			kfree(names);
4219 		} else
4220 			btrfs_warn(fs_info,
4221 				   "can't set %s bits 0x%llx while mounted",
4222 				   type, disallowed);
4223 		return -EPERM;
4224 	}
4225 
4226 	disallowed = clear_mask & ~safe_clear;
4227 	if (disallowed) {
4228 		names = btrfs_printable_features(set, disallowed);
4229 		if (names) {
4230 			btrfs_warn(fs_info,
4231 				   "can't clear the %s feature bit%s while mounted",
4232 				   names, strchr(names, ',') ? "s" : "");
4233 			kfree(names);
4234 		} else
4235 			btrfs_warn(fs_info,
4236 				   "can't clear %s bits 0x%llx while mounted",
4237 				   type, disallowed);
4238 		return -EPERM;
4239 	}
4240 
4241 	return 0;
4242 }
4243 
4244 #define check_feature(fs_info, change_mask, flags, mask_base)	\
4245 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,	\
4246 		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
4247 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
4248 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4249 
4250 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4251 {
4252 	struct inode *inode = file_inode(file);
4253 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4254 	struct btrfs_root *root = BTRFS_I(inode)->root;
4255 	struct btrfs_super_block *super_block = fs_info->super_copy;
4256 	struct btrfs_ioctl_feature_flags flags[2];
4257 	struct btrfs_trans_handle *trans;
4258 	u64 newflags;
4259 	int ret;
4260 
4261 	if (!capable(CAP_SYS_ADMIN))
4262 		return -EPERM;
4263 
4264 	if (copy_from_user(flags, arg, sizeof(flags)))
4265 		return -EFAULT;
4266 
4267 	/* Nothing to do */
4268 	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4269 	    !flags[0].incompat_flags)
4270 		return 0;
4271 
4272 	ret = check_feature(fs_info, flags[0].compat_flags,
4273 			    flags[1].compat_flags, COMPAT);
4274 	if (ret)
4275 		return ret;
4276 
4277 	ret = check_feature(fs_info, flags[0].compat_ro_flags,
4278 			    flags[1].compat_ro_flags, COMPAT_RO);
4279 	if (ret)
4280 		return ret;
4281 
4282 	ret = check_feature(fs_info, flags[0].incompat_flags,
4283 			    flags[1].incompat_flags, INCOMPAT);
4284 	if (ret)
4285 		return ret;
4286 
4287 	ret = mnt_want_write_file(file);
4288 	if (ret)
4289 		return ret;
4290 
4291 	trans = btrfs_start_transaction(root, 0);
4292 	if (IS_ERR(trans)) {
4293 		ret = PTR_ERR(trans);
4294 		goto out_drop_write;
4295 	}
4296 
4297 	spin_lock(&fs_info->super_lock);
4298 	newflags = btrfs_super_compat_flags(super_block);
4299 	newflags |= flags[0].compat_flags & flags[1].compat_flags;
4300 	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4301 	btrfs_set_super_compat_flags(super_block, newflags);
4302 
4303 	newflags = btrfs_super_compat_ro_flags(super_block);
4304 	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4305 	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4306 	btrfs_set_super_compat_ro_flags(super_block, newflags);
4307 
4308 	newflags = btrfs_super_incompat_flags(super_block);
4309 	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4310 	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4311 	btrfs_set_super_incompat_flags(super_block, newflags);
4312 	spin_unlock(&fs_info->super_lock);
4313 
4314 	ret = btrfs_commit_transaction(trans);
4315 out_drop_write:
4316 	mnt_drop_write_file(file);
4317 
4318 	return ret;
4319 }
4320 
4321 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4322 {
4323 	struct btrfs_ioctl_send_args *arg;
4324 	int ret;
4325 
4326 	if (compat) {
4327 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4328 		struct btrfs_ioctl_send_args_32 args32;
4329 
4330 		ret = copy_from_user(&args32, argp, sizeof(args32));
4331 		if (ret)
4332 			return -EFAULT;
4333 		arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4334 		if (!arg)
4335 			return -ENOMEM;
4336 		arg->send_fd = args32.send_fd;
4337 		arg->clone_sources_count = args32.clone_sources_count;
4338 		arg->clone_sources = compat_ptr(args32.clone_sources);
4339 		arg->parent_root = args32.parent_root;
4340 		arg->flags = args32.flags;
4341 		memcpy(arg->reserved, args32.reserved,
4342 		       sizeof(args32.reserved));
4343 #else
4344 		return -ENOTTY;
4345 #endif
4346 	} else {
4347 		arg = memdup_user(argp, sizeof(*arg));
4348 		if (IS_ERR(arg))
4349 			return PTR_ERR(arg);
4350 	}
4351 	ret = btrfs_ioctl_send(inode, arg);
4352 	kfree(arg);
4353 	return ret;
4354 }
4355 
4356 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4357 				    bool compat)
4358 {
4359 	struct btrfs_ioctl_encoded_io_args args = { 0 };
4360 	size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4361 					     flags);
4362 	size_t copy_end;
4363 	struct iovec iovstack[UIO_FASTIOV];
4364 	struct iovec *iov = iovstack;
4365 	struct iov_iter iter;
4366 	loff_t pos;
4367 	struct kiocb kiocb;
4368 	ssize_t ret;
4369 
4370 	if (!capable(CAP_SYS_ADMIN)) {
4371 		ret = -EPERM;
4372 		goto out_acct;
4373 	}
4374 
4375 	if (compat) {
4376 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4377 		struct btrfs_ioctl_encoded_io_args_32 args32;
4378 
4379 		copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4380 				       flags);
4381 		if (copy_from_user(&args32, argp, copy_end)) {
4382 			ret = -EFAULT;
4383 			goto out_acct;
4384 		}
4385 		args.iov = compat_ptr(args32.iov);
4386 		args.iovcnt = args32.iovcnt;
4387 		args.offset = args32.offset;
4388 		args.flags = args32.flags;
4389 #else
4390 		return -ENOTTY;
4391 #endif
4392 	} else {
4393 		copy_end = copy_end_kernel;
4394 		if (copy_from_user(&args, argp, copy_end)) {
4395 			ret = -EFAULT;
4396 			goto out_acct;
4397 		}
4398 	}
4399 	if (args.flags != 0) {
4400 		ret = -EINVAL;
4401 		goto out_acct;
4402 	}
4403 
4404 	ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4405 			   &iov, &iter);
4406 	if (ret < 0)
4407 		goto out_acct;
4408 
4409 	if (iov_iter_count(&iter) == 0) {
4410 		ret = 0;
4411 		goto out_iov;
4412 	}
4413 	pos = args.offset;
4414 	ret = rw_verify_area(READ, file, &pos, args.len);
4415 	if (ret < 0)
4416 		goto out_iov;
4417 
4418 	init_sync_kiocb(&kiocb, file);
4419 	kiocb.ki_pos = pos;
4420 
4421 	ret = btrfs_encoded_read(&kiocb, &iter, &args);
4422 	if (ret >= 0) {
4423 		fsnotify_access(file);
4424 		if (copy_to_user(argp + copy_end,
4425 				 (char *)&args + copy_end_kernel,
4426 				 sizeof(args) - copy_end_kernel))
4427 			ret = -EFAULT;
4428 	}
4429 
4430 out_iov:
4431 	kfree(iov);
4432 out_acct:
4433 	if (ret > 0)
4434 		add_rchar(current, ret);
4435 	inc_syscr(current);
4436 	return ret;
4437 }
4438 
4439 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4440 {
4441 	struct btrfs_ioctl_encoded_io_args args;
4442 	struct iovec iovstack[UIO_FASTIOV];
4443 	struct iovec *iov = iovstack;
4444 	struct iov_iter iter;
4445 	loff_t pos;
4446 	struct kiocb kiocb;
4447 	ssize_t ret;
4448 
4449 	if (!capable(CAP_SYS_ADMIN)) {
4450 		ret = -EPERM;
4451 		goto out_acct;
4452 	}
4453 
4454 	if (!(file->f_mode & FMODE_WRITE)) {
4455 		ret = -EBADF;
4456 		goto out_acct;
4457 	}
4458 
4459 	if (compat) {
4460 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4461 		struct btrfs_ioctl_encoded_io_args_32 args32;
4462 
4463 		if (copy_from_user(&args32, argp, sizeof(args32))) {
4464 			ret = -EFAULT;
4465 			goto out_acct;
4466 		}
4467 		args.iov = compat_ptr(args32.iov);
4468 		args.iovcnt = args32.iovcnt;
4469 		args.offset = args32.offset;
4470 		args.flags = args32.flags;
4471 		args.len = args32.len;
4472 		args.unencoded_len = args32.unencoded_len;
4473 		args.unencoded_offset = args32.unencoded_offset;
4474 		args.compression = args32.compression;
4475 		args.encryption = args32.encryption;
4476 		memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4477 #else
4478 		return -ENOTTY;
4479 #endif
4480 	} else {
4481 		if (copy_from_user(&args, argp, sizeof(args))) {
4482 			ret = -EFAULT;
4483 			goto out_acct;
4484 		}
4485 	}
4486 
4487 	ret = -EINVAL;
4488 	if (args.flags != 0)
4489 		goto out_acct;
4490 	if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4491 		goto out_acct;
4492 	if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4493 	    args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4494 		goto out_acct;
4495 	if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4496 	    args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4497 		goto out_acct;
4498 	if (args.unencoded_offset > args.unencoded_len)
4499 		goto out_acct;
4500 	if (args.len > args.unencoded_len - args.unencoded_offset)
4501 		goto out_acct;
4502 
4503 	ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4504 			   &iov, &iter);
4505 	if (ret < 0)
4506 		goto out_acct;
4507 
4508 	file_start_write(file);
4509 
4510 	if (iov_iter_count(&iter) == 0) {
4511 		ret = 0;
4512 		goto out_end_write;
4513 	}
4514 	pos = args.offset;
4515 	ret = rw_verify_area(WRITE, file, &pos, args.len);
4516 	if (ret < 0)
4517 		goto out_end_write;
4518 
4519 	init_sync_kiocb(&kiocb, file);
4520 	ret = kiocb_set_rw_flags(&kiocb, 0);
4521 	if (ret)
4522 		goto out_end_write;
4523 	kiocb.ki_pos = pos;
4524 
4525 	ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4526 	if (ret > 0)
4527 		fsnotify_modify(file);
4528 
4529 out_end_write:
4530 	file_end_write(file);
4531 	kfree(iov);
4532 out_acct:
4533 	if (ret > 0)
4534 		add_wchar(current, ret);
4535 	inc_syscw(current);
4536 	return ret;
4537 }
4538 
4539 long btrfs_ioctl(struct file *file, unsigned int
4540 		cmd, unsigned long arg)
4541 {
4542 	struct inode *inode = file_inode(file);
4543 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4544 	struct btrfs_root *root = BTRFS_I(inode)->root;
4545 	void __user *argp = (void __user *)arg;
4546 
4547 	switch (cmd) {
4548 	case FS_IOC_GETVERSION:
4549 		return btrfs_ioctl_getversion(inode, argp);
4550 	case FS_IOC_GETFSLABEL:
4551 		return btrfs_ioctl_get_fslabel(fs_info, argp);
4552 	case FS_IOC_SETFSLABEL:
4553 		return btrfs_ioctl_set_fslabel(file, argp);
4554 	case FITRIM:
4555 		return btrfs_ioctl_fitrim(fs_info, argp);
4556 	case BTRFS_IOC_SNAP_CREATE:
4557 		return btrfs_ioctl_snap_create(file, argp, 0);
4558 	case BTRFS_IOC_SNAP_CREATE_V2:
4559 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
4560 	case BTRFS_IOC_SUBVOL_CREATE:
4561 		return btrfs_ioctl_snap_create(file, argp, 1);
4562 	case BTRFS_IOC_SUBVOL_CREATE_V2:
4563 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
4564 	case BTRFS_IOC_SNAP_DESTROY:
4565 		return btrfs_ioctl_snap_destroy(file, argp, false);
4566 	case BTRFS_IOC_SNAP_DESTROY_V2:
4567 		return btrfs_ioctl_snap_destroy(file, argp, true);
4568 	case BTRFS_IOC_SUBVOL_GETFLAGS:
4569 		return btrfs_ioctl_subvol_getflags(inode, argp);
4570 	case BTRFS_IOC_SUBVOL_SETFLAGS:
4571 		return btrfs_ioctl_subvol_setflags(file, argp);
4572 	case BTRFS_IOC_DEFAULT_SUBVOL:
4573 		return btrfs_ioctl_default_subvol(file, argp);
4574 	case BTRFS_IOC_DEFRAG:
4575 		return btrfs_ioctl_defrag(file, NULL);
4576 	case BTRFS_IOC_DEFRAG_RANGE:
4577 		return btrfs_ioctl_defrag(file, argp);
4578 	case BTRFS_IOC_RESIZE:
4579 		return btrfs_ioctl_resize(file, argp);
4580 	case BTRFS_IOC_ADD_DEV:
4581 		return btrfs_ioctl_add_dev(fs_info, argp);
4582 	case BTRFS_IOC_RM_DEV:
4583 		return btrfs_ioctl_rm_dev(file, argp);
4584 	case BTRFS_IOC_RM_DEV_V2:
4585 		return btrfs_ioctl_rm_dev_v2(file, argp);
4586 	case BTRFS_IOC_FS_INFO:
4587 		return btrfs_ioctl_fs_info(fs_info, argp);
4588 	case BTRFS_IOC_DEV_INFO:
4589 		return btrfs_ioctl_dev_info(fs_info, argp);
4590 	case BTRFS_IOC_TREE_SEARCH:
4591 		return btrfs_ioctl_tree_search(inode, argp);
4592 	case BTRFS_IOC_TREE_SEARCH_V2:
4593 		return btrfs_ioctl_tree_search_v2(inode, argp);
4594 	case BTRFS_IOC_INO_LOOKUP:
4595 		return btrfs_ioctl_ino_lookup(root, argp);
4596 	case BTRFS_IOC_INO_PATHS:
4597 		return btrfs_ioctl_ino_to_path(root, argp);
4598 	case BTRFS_IOC_LOGICAL_INO:
4599 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4600 	case BTRFS_IOC_LOGICAL_INO_V2:
4601 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4602 	case BTRFS_IOC_SPACE_INFO:
4603 		return btrfs_ioctl_space_info(fs_info, argp);
4604 	case BTRFS_IOC_SYNC: {
4605 		int ret;
4606 
4607 		ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4608 		if (ret)
4609 			return ret;
4610 		ret = btrfs_sync_fs(inode->i_sb, 1);
4611 		/*
4612 		 * The transaction thread may want to do more work,
4613 		 * namely it pokes the cleaner kthread that will start
4614 		 * processing uncleaned subvols.
4615 		 */
4616 		wake_up_process(fs_info->transaction_kthread);
4617 		return ret;
4618 	}
4619 	case BTRFS_IOC_START_SYNC:
4620 		return btrfs_ioctl_start_sync(root, argp);
4621 	case BTRFS_IOC_WAIT_SYNC:
4622 		return btrfs_ioctl_wait_sync(fs_info, argp);
4623 	case BTRFS_IOC_SCRUB:
4624 		return btrfs_ioctl_scrub(file, argp);
4625 	case BTRFS_IOC_SCRUB_CANCEL:
4626 		return btrfs_ioctl_scrub_cancel(fs_info);
4627 	case BTRFS_IOC_SCRUB_PROGRESS:
4628 		return btrfs_ioctl_scrub_progress(fs_info, argp);
4629 	case BTRFS_IOC_BALANCE_V2:
4630 		return btrfs_ioctl_balance(file, argp);
4631 	case BTRFS_IOC_BALANCE_CTL:
4632 		return btrfs_ioctl_balance_ctl(fs_info, arg);
4633 	case BTRFS_IOC_BALANCE_PROGRESS:
4634 		return btrfs_ioctl_balance_progress(fs_info, argp);
4635 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4636 		return btrfs_ioctl_set_received_subvol(file, argp);
4637 #ifdef CONFIG_64BIT
4638 	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4639 		return btrfs_ioctl_set_received_subvol_32(file, argp);
4640 #endif
4641 	case BTRFS_IOC_SEND:
4642 		return _btrfs_ioctl_send(inode, argp, false);
4643 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4644 	case BTRFS_IOC_SEND_32:
4645 		return _btrfs_ioctl_send(inode, argp, true);
4646 #endif
4647 	case BTRFS_IOC_GET_DEV_STATS:
4648 		return btrfs_ioctl_get_dev_stats(fs_info, argp);
4649 	case BTRFS_IOC_QUOTA_CTL:
4650 		return btrfs_ioctl_quota_ctl(file, argp);
4651 	case BTRFS_IOC_QGROUP_ASSIGN:
4652 		return btrfs_ioctl_qgroup_assign(file, argp);
4653 	case BTRFS_IOC_QGROUP_CREATE:
4654 		return btrfs_ioctl_qgroup_create(file, argp);
4655 	case BTRFS_IOC_QGROUP_LIMIT:
4656 		return btrfs_ioctl_qgroup_limit(file, argp);
4657 	case BTRFS_IOC_QUOTA_RESCAN:
4658 		return btrfs_ioctl_quota_rescan(file, argp);
4659 	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4660 		return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4661 	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4662 		return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4663 	case BTRFS_IOC_DEV_REPLACE:
4664 		return btrfs_ioctl_dev_replace(fs_info, argp);
4665 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4666 		return btrfs_ioctl_get_supported_features(argp);
4667 	case BTRFS_IOC_GET_FEATURES:
4668 		return btrfs_ioctl_get_features(fs_info, argp);
4669 	case BTRFS_IOC_SET_FEATURES:
4670 		return btrfs_ioctl_set_features(file, argp);
4671 	case BTRFS_IOC_GET_SUBVOL_INFO:
4672 		return btrfs_ioctl_get_subvol_info(inode, argp);
4673 	case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4674 		return btrfs_ioctl_get_subvol_rootref(root, argp);
4675 	case BTRFS_IOC_INO_LOOKUP_USER:
4676 		return btrfs_ioctl_ino_lookup_user(file, argp);
4677 	case FS_IOC_ENABLE_VERITY:
4678 		return fsverity_ioctl_enable(file, (const void __user *)argp);
4679 	case FS_IOC_MEASURE_VERITY:
4680 		return fsverity_ioctl_measure(file, argp);
4681 	case BTRFS_IOC_ENCODED_READ:
4682 		return btrfs_ioctl_encoded_read(file, argp, false);
4683 	case BTRFS_IOC_ENCODED_WRITE:
4684 		return btrfs_ioctl_encoded_write(file, argp, false);
4685 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4686 	case BTRFS_IOC_ENCODED_READ_32:
4687 		return btrfs_ioctl_encoded_read(file, argp, true);
4688 	case BTRFS_IOC_ENCODED_WRITE_32:
4689 		return btrfs_ioctl_encoded_write(file, argp, true);
4690 #endif
4691 	}
4692 
4693 	return -ENOTTY;
4694 }
4695 
4696 #ifdef CONFIG_COMPAT
4697 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4698 {
4699 	/*
4700 	 * These all access 32-bit values anyway so no further
4701 	 * handling is necessary.
4702 	 */
4703 	switch (cmd) {
4704 	case FS_IOC32_GETVERSION:
4705 		cmd = FS_IOC_GETVERSION;
4706 		break;
4707 	}
4708 
4709 	return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4710 }
4711 #endif
4712