xref: /linux/fs/btrfs/ioctl.c (revision 2b0cfa6e49566c8fa6759734cf821aa6e8271a9e)
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_set_ctime_current(inode);
388 	ret = btrfs_update_inode(trans, 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_NONE ||
459 		       fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
460 		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
461 		spin_unlock(&fs_info->super_lock);
462 		break;
463 	case BTRFS_EXCLOP_BALANCE:
464 		spin_lock(&fs_info->super_lock);
465 		ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
466 		fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
467 		spin_unlock(&fs_info->super_lock);
468 		break;
469 	default:
470 		btrfs_warn(fs_info,
471 			"invalid exclop balance operation %d requested", op);
472 	}
473 }
474 
475 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
476 {
477 	return put_user(inode->i_generation, arg);
478 }
479 
480 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
481 					void __user *arg)
482 {
483 	struct btrfs_device *device;
484 	struct fstrim_range range;
485 	u64 minlen = ULLONG_MAX;
486 	u64 num_devices = 0;
487 	int ret;
488 
489 	if (!capable(CAP_SYS_ADMIN))
490 		return -EPERM;
491 
492 	/*
493 	 * btrfs_trim_block_group() depends on space cache, which is not
494 	 * available in zoned filesystem. So, disallow fitrim on a zoned
495 	 * filesystem for now.
496 	 */
497 	if (btrfs_is_zoned(fs_info))
498 		return -EOPNOTSUPP;
499 
500 	/*
501 	 * If the fs is mounted with nologreplay, which requires it to be
502 	 * mounted in RO mode as well, we can not allow discard on free space
503 	 * inside block groups, because log trees refer to extents that are not
504 	 * pinned in a block group's free space cache (pinning the extents is
505 	 * precisely the first phase of replaying a log tree).
506 	 */
507 	if (btrfs_test_opt(fs_info, NOLOGREPLAY))
508 		return -EROFS;
509 
510 	rcu_read_lock();
511 	list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
512 				dev_list) {
513 		if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
514 			continue;
515 		num_devices++;
516 		minlen = min_t(u64, bdev_discard_granularity(device->bdev),
517 				    minlen);
518 	}
519 	rcu_read_unlock();
520 
521 	if (!num_devices)
522 		return -EOPNOTSUPP;
523 	if (copy_from_user(&range, arg, sizeof(range)))
524 		return -EFAULT;
525 
526 	/*
527 	 * NOTE: Don't truncate the range using super->total_bytes.  Bytenr of
528 	 * block group is in the logical address space, which can be any
529 	 * sectorsize aligned bytenr in  the range [0, U64_MAX].
530 	 */
531 	if (range.len < fs_info->sb->s_blocksize)
532 		return -EINVAL;
533 
534 	range.minlen = max(range.minlen, minlen);
535 	ret = btrfs_trim_fs(fs_info, &range);
536 	if (ret < 0)
537 		return ret;
538 
539 	if (copy_to_user(arg, &range, sizeof(range)))
540 		return -EFAULT;
541 
542 	return 0;
543 }
544 
545 int __pure btrfs_is_empty_uuid(u8 *uuid)
546 {
547 	int i;
548 
549 	for (i = 0; i < BTRFS_UUID_SIZE; i++) {
550 		if (uuid[i])
551 			return 0;
552 	}
553 	return 1;
554 }
555 
556 /*
557  * Calculate the number of transaction items to reserve for creating a subvolume
558  * or snapshot, not including the inode, directory entries, or parent directory.
559  */
560 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
561 {
562 	/*
563 	 * 1 to add root block
564 	 * 1 to add root item
565 	 * 1 to add root ref
566 	 * 1 to add root backref
567 	 * 1 to add UUID item
568 	 * 1 to add qgroup info
569 	 * 1 to add qgroup limit
570 	 *
571 	 * Ideally the last two would only be accounted if qgroups are enabled,
572 	 * but that can change between now and the time we would insert them.
573 	 */
574 	unsigned int num_items = 7;
575 
576 	if (inherit) {
577 		/* 2 to add qgroup relations for each inherited qgroup */
578 		num_items += 2 * inherit->num_qgroups;
579 	}
580 	return num_items;
581 }
582 
583 static noinline int create_subvol(struct mnt_idmap *idmap,
584 				  struct inode *dir, struct dentry *dentry,
585 				  struct btrfs_qgroup_inherit *inherit)
586 {
587 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
588 	struct btrfs_trans_handle *trans;
589 	struct btrfs_key key;
590 	struct btrfs_root_item *root_item;
591 	struct btrfs_inode_item *inode_item;
592 	struct extent_buffer *leaf;
593 	struct btrfs_root *root = BTRFS_I(dir)->root;
594 	struct btrfs_root *new_root;
595 	struct btrfs_block_rsv block_rsv;
596 	struct timespec64 cur_time = current_time(dir);
597 	struct btrfs_new_inode_args new_inode_args = {
598 		.dir = dir,
599 		.dentry = dentry,
600 		.subvol = true,
601 	};
602 	unsigned int trans_num_items;
603 	int ret;
604 	dev_t anon_dev;
605 	u64 objectid;
606 
607 	root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
608 	if (!root_item)
609 		return -ENOMEM;
610 
611 	ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
612 	if (ret)
613 		goto out_root_item;
614 
615 	/*
616 	 * Don't create subvolume whose level is not zero. Or qgroup will be
617 	 * screwed up since it assumes subvolume qgroup's level to be 0.
618 	 */
619 	if (btrfs_qgroup_level(objectid)) {
620 		ret = -ENOSPC;
621 		goto out_root_item;
622 	}
623 
624 	ret = get_anon_bdev(&anon_dev);
625 	if (ret < 0)
626 		goto out_root_item;
627 
628 	new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
629 	if (!new_inode_args.inode) {
630 		ret = -ENOMEM;
631 		goto out_anon_dev;
632 	}
633 	ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
634 	if (ret)
635 		goto out_inode;
636 	trans_num_items += create_subvol_num_items(inherit);
637 
638 	btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
639 	ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
640 					       trans_num_items, false);
641 	if (ret)
642 		goto out_new_inode_args;
643 
644 	trans = btrfs_start_transaction(root, 0);
645 	if (IS_ERR(trans)) {
646 		ret = PTR_ERR(trans);
647 		btrfs_subvolume_release_metadata(root, &block_rsv);
648 		goto out_new_inode_args;
649 	}
650 	trans->block_rsv = &block_rsv;
651 	trans->bytes_reserved = block_rsv.size;
652 	/* Tree log can't currently deal with an inode which is a new root. */
653 	btrfs_set_log_full_commit(trans);
654 
655 	ret = btrfs_qgroup_inherit(trans, 0, objectid, root->root_key.objectid, inherit);
656 	if (ret)
657 		goto out;
658 
659 	leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
660 				      0, BTRFS_NESTING_NORMAL);
661 	if (IS_ERR(leaf)) {
662 		ret = PTR_ERR(leaf);
663 		goto out;
664 	}
665 
666 	btrfs_mark_buffer_dirty(trans, leaf);
667 
668 	inode_item = &root_item->inode;
669 	btrfs_set_stack_inode_generation(inode_item, 1);
670 	btrfs_set_stack_inode_size(inode_item, 3);
671 	btrfs_set_stack_inode_nlink(inode_item, 1);
672 	btrfs_set_stack_inode_nbytes(inode_item,
673 				     fs_info->nodesize);
674 	btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
675 
676 	btrfs_set_root_flags(root_item, 0);
677 	btrfs_set_root_limit(root_item, 0);
678 	btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
679 
680 	btrfs_set_root_bytenr(root_item, leaf->start);
681 	btrfs_set_root_generation(root_item, trans->transid);
682 	btrfs_set_root_level(root_item, 0);
683 	btrfs_set_root_refs(root_item, 1);
684 	btrfs_set_root_used(root_item, leaf->len);
685 	btrfs_set_root_last_snapshot(root_item, 0);
686 
687 	btrfs_set_root_generation_v2(root_item,
688 			btrfs_root_generation(root_item));
689 	generate_random_guid(root_item->uuid);
690 	btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
691 	btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
692 	root_item->ctime = root_item->otime;
693 	btrfs_set_root_ctransid(root_item, trans->transid);
694 	btrfs_set_root_otransid(root_item, trans->transid);
695 
696 	btrfs_tree_unlock(leaf);
697 
698 	btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
699 
700 	key.objectid = objectid;
701 	key.offset = 0;
702 	key.type = BTRFS_ROOT_ITEM_KEY;
703 	ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
704 				root_item);
705 	if (ret) {
706 		/*
707 		 * Since we don't abort the transaction in this case, free the
708 		 * tree block so that we don't leak space and leave the
709 		 * filesystem in an inconsistent state (an extent item in the
710 		 * extent tree with a backreference for a root that does not
711 		 * exists).
712 		 */
713 		btrfs_tree_lock(leaf);
714 		btrfs_clear_buffer_dirty(trans, leaf);
715 		btrfs_tree_unlock(leaf);
716 		btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
717 		free_extent_buffer(leaf);
718 		goto out;
719 	}
720 
721 	free_extent_buffer(leaf);
722 	leaf = NULL;
723 
724 	new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
725 	if (IS_ERR(new_root)) {
726 		ret = PTR_ERR(new_root);
727 		btrfs_abort_transaction(trans, ret);
728 		goto out;
729 	}
730 	/* anon_dev is owned by new_root now. */
731 	anon_dev = 0;
732 	BTRFS_I(new_inode_args.inode)->root = new_root;
733 	/* ... and new_root is owned by new_inode_args.inode now. */
734 
735 	ret = btrfs_record_root_in_trans(trans, new_root);
736 	if (ret) {
737 		btrfs_abort_transaction(trans, ret);
738 		goto out;
739 	}
740 
741 	ret = btrfs_uuid_tree_add(trans, root_item->uuid,
742 				  BTRFS_UUID_KEY_SUBVOL, objectid);
743 	if (ret) {
744 		btrfs_abort_transaction(trans, ret);
745 		goto out;
746 	}
747 
748 	ret = btrfs_create_new_inode(trans, &new_inode_args);
749 	if (ret) {
750 		btrfs_abort_transaction(trans, ret);
751 		goto out;
752 	}
753 
754 	d_instantiate_new(dentry, new_inode_args.inode);
755 	new_inode_args.inode = NULL;
756 
757 out:
758 	trans->block_rsv = NULL;
759 	trans->bytes_reserved = 0;
760 	btrfs_subvolume_release_metadata(root, &block_rsv);
761 
762 	btrfs_end_transaction(trans);
763 out_new_inode_args:
764 	btrfs_new_inode_args_destroy(&new_inode_args);
765 out_inode:
766 	iput(new_inode_args.inode);
767 out_anon_dev:
768 	if (anon_dev)
769 		free_anon_bdev(anon_dev);
770 out_root_item:
771 	kfree(root_item);
772 	return ret;
773 }
774 
775 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
776 			   struct dentry *dentry, bool readonly,
777 			   struct btrfs_qgroup_inherit *inherit)
778 {
779 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
780 	struct inode *inode;
781 	struct btrfs_pending_snapshot *pending_snapshot;
782 	unsigned int trans_num_items;
783 	struct btrfs_trans_handle *trans;
784 	int ret;
785 
786 	/* We do not support snapshotting right now. */
787 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
788 		btrfs_warn(fs_info,
789 			   "extent tree v2 doesn't support snapshotting yet");
790 		return -EOPNOTSUPP;
791 	}
792 
793 	if (btrfs_root_refs(&root->root_item) == 0)
794 		return -ENOENT;
795 
796 	if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
797 		return -EINVAL;
798 
799 	if (atomic_read(&root->nr_swapfiles)) {
800 		btrfs_warn(fs_info,
801 			   "cannot snapshot subvolume with active swapfile");
802 		return -ETXTBSY;
803 	}
804 
805 	pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
806 	if (!pending_snapshot)
807 		return -ENOMEM;
808 
809 	ret = get_anon_bdev(&pending_snapshot->anon_dev);
810 	if (ret < 0)
811 		goto free_pending;
812 	pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
813 			GFP_KERNEL);
814 	pending_snapshot->path = btrfs_alloc_path();
815 	if (!pending_snapshot->root_item || !pending_snapshot->path) {
816 		ret = -ENOMEM;
817 		goto free_pending;
818 	}
819 
820 	btrfs_init_block_rsv(&pending_snapshot->block_rsv,
821 			     BTRFS_BLOCK_RSV_TEMP);
822 	/*
823 	 * 1 to add dir item
824 	 * 1 to add dir index
825 	 * 1 to update parent inode item
826 	 */
827 	trans_num_items = create_subvol_num_items(inherit) + 3;
828 	ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
829 					       &pending_snapshot->block_rsv,
830 					       trans_num_items, false);
831 	if (ret)
832 		goto free_pending;
833 
834 	pending_snapshot->dentry = dentry;
835 	pending_snapshot->root = root;
836 	pending_snapshot->readonly = readonly;
837 	pending_snapshot->dir = dir;
838 	pending_snapshot->inherit = inherit;
839 
840 	trans = btrfs_start_transaction(root, 0);
841 	if (IS_ERR(trans)) {
842 		ret = PTR_ERR(trans);
843 		goto fail;
844 	}
845 
846 	trans->pending_snapshot = pending_snapshot;
847 
848 	ret = btrfs_commit_transaction(trans);
849 	if (ret)
850 		goto fail;
851 
852 	ret = pending_snapshot->error;
853 	if (ret)
854 		goto fail;
855 
856 	ret = btrfs_orphan_cleanup(pending_snapshot->snap);
857 	if (ret)
858 		goto fail;
859 
860 	inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
861 	if (IS_ERR(inode)) {
862 		ret = PTR_ERR(inode);
863 		goto fail;
864 	}
865 
866 	d_instantiate(dentry, inode);
867 	ret = 0;
868 	pending_snapshot->anon_dev = 0;
869 fail:
870 	/* Prevent double freeing of anon_dev */
871 	if (ret && pending_snapshot->snap)
872 		pending_snapshot->snap->anon_dev = 0;
873 	btrfs_put_root(pending_snapshot->snap);
874 	btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
875 free_pending:
876 	if (pending_snapshot->anon_dev)
877 		free_anon_bdev(pending_snapshot->anon_dev);
878 	kfree(pending_snapshot->root_item);
879 	btrfs_free_path(pending_snapshot->path);
880 	kfree(pending_snapshot);
881 
882 	return ret;
883 }
884 
885 /*  copy of may_delete in fs/namei.c()
886  *	Check whether we can remove a link victim from directory dir, check
887  *  whether the type of victim is right.
888  *  1. We can't do it if dir is read-only (done in permission())
889  *  2. We should have write and exec permissions on dir
890  *  3. We can't remove anything from append-only dir
891  *  4. We can't do anything with immutable dir (done in permission())
892  *  5. If the sticky bit on dir is set we should either
893  *	a. be owner of dir, or
894  *	b. be owner of victim, or
895  *	c. have CAP_FOWNER capability
896  *  6. If the victim is append-only or immutable we can't do anything with
897  *     links pointing to it.
898  *  7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
899  *  8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
900  *  9. We can't remove a root or mountpoint.
901  * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
902  *     nfs_async_unlink().
903  */
904 
905 static int btrfs_may_delete(struct mnt_idmap *idmap,
906 			    struct inode *dir, struct dentry *victim, int isdir)
907 {
908 	int error;
909 
910 	if (d_really_is_negative(victim))
911 		return -ENOENT;
912 
913 	BUG_ON(d_inode(victim->d_parent) != dir);
914 	audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
915 
916 	error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
917 	if (error)
918 		return error;
919 	if (IS_APPEND(dir))
920 		return -EPERM;
921 	if (check_sticky(idmap, dir, d_inode(victim)) ||
922 	    IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
923 	    IS_SWAPFILE(d_inode(victim)))
924 		return -EPERM;
925 	if (isdir) {
926 		if (!d_is_dir(victim))
927 			return -ENOTDIR;
928 		if (IS_ROOT(victim))
929 			return -EBUSY;
930 	} else if (d_is_dir(victim))
931 		return -EISDIR;
932 	if (IS_DEADDIR(dir))
933 		return -ENOENT;
934 	if (victim->d_flags & DCACHE_NFSFS_RENAMED)
935 		return -EBUSY;
936 	return 0;
937 }
938 
939 /* copy of may_create in fs/namei.c() */
940 static inline int btrfs_may_create(struct mnt_idmap *idmap,
941 				   struct inode *dir, struct dentry *child)
942 {
943 	if (d_really_is_positive(child))
944 		return -EEXIST;
945 	if (IS_DEADDIR(dir))
946 		return -ENOENT;
947 	if (!fsuidgid_has_mapping(dir->i_sb, idmap))
948 		return -EOVERFLOW;
949 	return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
950 }
951 
952 /*
953  * Create a new subvolume below @parent.  This is largely modeled after
954  * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
955  * inside this filesystem so it's quite a bit simpler.
956  */
957 static noinline int btrfs_mksubvol(const struct path *parent,
958 				   struct mnt_idmap *idmap,
959 				   const char *name, int namelen,
960 				   struct btrfs_root *snap_src,
961 				   bool readonly,
962 				   struct btrfs_qgroup_inherit *inherit)
963 {
964 	struct inode *dir = d_inode(parent->dentry);
965 	struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
966 	struct dentry *dentry;
967 	struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
968 	int error;
969 
970 	error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
971 	if (error == -EINTR)
972 		return error;
973 
974 	dentry = lookup_one(idmap, name, parent->dentry, namelen);
975 	error = PTR_ERR(dentry);
976 	if (IS_ERR(dentry))
977 		goto out_unlock;
978 
979 	error = btrfs_may_create(idmap, dir, dentry);
980 	if (error)
981 		goto out_dput;
982 
983 	/*
984 	 * even if this name doesn't exist, we may get hash collisions.
985 	 * check for them now when we can safely fail
986 	 */
987 	error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
988 					       dir->i_ino, &name_str);
989 	if (error)
990 		goto out_dput;
991 
992 	down_read(&fs_info->subvol_sem);
993 
994 	if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
995 		goto out_up_read;
996 
997 	if (snap_src)
998 		error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
999 	else
1000 		error = create_subvol(idmap, dir, dentry, inherit);
1001 
1002 	if (!error)
1003 		fsnotify_mkdir(dir, dentry);
1004 out_up_read:
1005 	up_read(&fs_info->subvol_sem);
1006 out_dput:
1007 	dput(dentry);
1008 out_unlock:
1009 	btrfs_inode_unlock(BTRFS_I(dir), 0);
1010 	return error;
1011 }
1012 
1013 static noinline int btrfs_mksnapshot(const struct path *parent,
1014 				   struct mnt_idmap *idmap,
1015 				   const char *name, int namelen,
1016 				   struct btrfs_root *root,
1017 				   bool readonly,
1018 				   struct btrfs_qgroup_inherit *inherit)
1019 {
1020 	int ret;
1021 	bool snapshot_force_cow = false;
1022 
1023 	/*
1024 	 * Force new buffered writes to reserve space even when NOCOW is
1025 	 * possible. This is to avoid later writeback (running dealloc) to
1026 	 * fallback to COW mode and unexpectedly fail with ENOSPC.
1027 	 */
1028 	btrfs_drew_read_lock(&root->snapshot_lock);
1029 
1030 	ret = btrfs_start_delalloc_snapshot(root, false);
1031 	if (ret)
1032 		goto out;
1033 
1034 	/*
1035 	 * All previous writes have started writeback in NOCOW mode, so now
1036 	 * we force future writes to fallback to COW mode during snapshot
1037 	 * creation.
1038 	 */
1039 	atomic_inc(&root->snapshot_force_cow);
1040 	snapshot_force_cow = true;
1041 
1042 	btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1043 
1044 	ret = btrfs_mksubvol(parent, idmap, name, namelen,
1045 			     root, readonly, inherit);
1046 out:
1047 	if (snapshot_force_cow)
1048 		atomic_dec(&root->snapshot_force_cow);
1049 	btrfs_drew_read_unlock(&root->snapshot_lock);
1050 	return ret;
1051 }
1052 
1053 /*
1054  * Try to start exclusive operation @type or cancel it if it's running.
1055  *
1056  * Return:
1057  *   0        - normal mode, newly claimed op started
1058  *  >0        - normal mode, something else is running,
1059  *              return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1060  * ECANCELED  - cancel mode, successful cancel
1061  * ENOTCONN   - cancel mode, operation not running anymore
1062  */
1063 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1064 			enum btrfs_exclusive_operation type, bool cancel)
1065 {
1066 	if (!cancel) {
1067 		/* Start normal op */
1068 		if (!btrfs_exclop_start(fs_info, type))
1069 			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1070 		/* Exclusive operation is now claimed */
1071 		return 0;
1072 	}
1073 
1074 	/* Cancel running op */
1075 	if (btrfs_exclop_start_try_lock(fs_info, type)) {
1076 		/*
1077 		 * This blocks any exclop finish from setting it to NONE, so we
1078 		 * request cancellation. Either it runs and we will wait for it,
1079 		 * or it has finished and no waiting will happen.
1080 		 */
1081 		atomic_inc(&fs_info->reloc_cancel_req);
1082 		btrfs_exclop_start_unlock(fs_info);
1083 
1084 		if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1085 			wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1086 				    TASK_INTERRUPTIBLE);
1087 
1088 		return -ECANCELED;
1089 	}
1090 
1091 	/* Something else is running or none */
1092 	return -ENOTCONN;
1093 }
1094 
1095 static noinline int btrfs_ioctl_resize(struct file *file,
1096 					void __user *arg)
1097 {
1098 	BTRFS_DEV_LOOKUP_ARGS(args);
1099 	struct inode *inode = file_inode(file);
1100 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1101 	u64 new_size;
1102 	u64 old_size;
1103 	u64 devid = 1;
1104 	struct btrfs_root *root = BTRFS_I(inode)->root;
1105 	struct btrfs_ioctl_vol_args *vol_args;
1106 	struct btrfs_trans_handle *trans;
1107 	struct btrfs_device *device = NULL;
1108 	char *sizestr;
1109 	char *retptr;
1110 	char *devstr = NULL;
1111 	int ret = 0;
1112 	int mod = 0;
1113 	bool cancel;
1114 
1115 	if (!capable(CAP_SYS_ADMIN))
1116 		return -EPERM;
1117 
1118 	ret = mnt_want_write_file(file);
1119 	if (ret)
1120 		return ret;
1121 
1122 	/*
1123 	 * Read the arguments before checking exclusivity to be able to
1124 	 * distinguish regular resize and cancel
1125 	 */
1126 	vol_args = memdup_user(arg, sizeof(*vol_args));
1127 	if (IS_ERR(vol_args)) {
1128 		ret = PTR_ERR(vol_args);
1129 		goto out_drop;
1130 	}
1131 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1132 	sizestr = vol_args->name;
1133 	cancel = (strcmp("cancel", sizestr) == 0);
1134 	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1135 	if (ret)
1136 		goto out_free;
1137 	/* Exclusive operation is now claimed */
1138 
1139 	devstr = strchr(sizestr, ':');
1140 	if (devstr) {
1141 		sizestr = devstr + 1;
1142 		*devstr = '\0';
1143 		devstr = vol_args->name;
1144 		ret = kstrtoull(devstr, 10, &devid);
1145 		if (ret)
1146 			goto out_finish;
1147 		if (!devid) {
1148 			ret = -EINVAL;
1149 			goto out_finish;
1150 		}
1151 		btrfs_info(fs_info, "resizing devid %llu", devid);
1152 	}
1153 
1154 	args.devid = devid;
1155 	device = btrfs_find_device(fs_info->fs_devices, &args);
1156 	if (!device) {
1157 		btrfs_info(fs_info, "resizer unable to find device %llu",
1158 			   devid);
1159 		ret = -ENODEV;
1160 		goto out_finish;
1161 	}
1162 
1163 	if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1164 		btrfs_info(fs_info,
1165 			   "resizer unable to apply on readonly device %llu",
1166 		       devid);
1167 		ret = -EPERM;
1168 		goto out_finish;
1169 	}
1170 
1171 	if (!strcmp(sizestr, "max"))
1172 		new_size = bdev_nr_bytes(device->bdev);
1173 	else {
1174 		if (sizestr[0] == '-') {
1175 			mod = -1;
1176 			sizestr++;
1177 		} else if (sizestr[0] == '+') {
1178 			mod = 1;
1179 			sizestr++;
1180 		}
1181 		new_size = memparse(sizestr, &retptr);
1182 		if (*retptr != '\0' || new_size == 0) {
1183 			ret = -EINVAL;
1184 			goto out_finish;
1185 		}
1186 	}
1187 
1188 	if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1189 		ret = -EPERM;
1190 		goto out_finish;
1191 	}
1192 
1193 	old_size = btrfs_device_get_total_bytes(device);
1194 
1195 	if (mod < 0) {
1196 		if (new_size > old_size) {
1197 			ret = -EINVAL;
1198 			goto out_finish;
1199 		}
1200 		new_size = old_size - new_size;
1201 	} else if (mod > 0) {
1202 		if (new_size > ULLONG_MAX - old_size) {
1203 			ret = -ERANGE;
1204 			goto out_finish;
1205 		}
1206 		new_size = old_size + new_size;
1207 	}
1208 
1209 	if (new_size < SZ_256M) {
1210 		ret = -EINVAL;
1211 		goto out_finish;
1212 	}
1213 	if (new_size > bdev_nr_bytes(device->bdev)) {
1214 		ret = -EFBIG;
1215 		goto out_finish;
1216 	}
1217 
1218 	new_size = round_down(new_size, fs_info->sectorsize);
1219 
1220 	if (new_size > old_size) {
1221 		trans = btrfs_start_transaction(root, 0);
1222 		if (IS_ERR(trans)) {
1223 			ret = PTR_ERR(trans);
1224 			goto out_finish;
1225 		}
1226 		ret = btrfs_grow_device(trans, device, new_size);
1227 		btrfs_commit_transaction(trans);
1228 	} else if (new_size < old_size) {
1229 		ret = btrfs_shrink_device(device, new_size);
1230 	} /* equal, nothing need to do */
1231 
1232 	if (ret == 0 && new_size != old_size)
1233 		btrfs_info_in_rcu(fs_info,
1234 			"resize device %s (devid %llu) from %llu to %llu",
1235 			btrfs_dev_name(device), device->devid,
1236 			old_size, new_size);
1237 out_finish:
1238 	btrfs_exclop_finish(fs_info);
1239 out_free:
1240 	kfree(vol_args);
1241 out_drop:
1242 	mnt_drop_write_file(file);
1243 	return ret;
1244 }
1245 
1246 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1247 				struct mnt_idmap *idmap,
1248 				const char *name, unsigned long fd, int subvol,
1249 				bool readonly,
1250 				struct btrfs_qgroup_inherit *inherit)
1251 {
1252 	int namelen;
1253 	int ret = 0;
1254 
1255 	if (!S_ISDIR(file_inode(file)->i_mode))
1256 		return -ENOTDIR;
1257 
1258 	ret = mnt_want_write_file(file);
1259 	if (ret)
1260 		goto out;
1261 
1262 	namelen = strlen(name);
1263 	if (strchr(name, '/')) {
1264 		ret = -EINVAL;
1265 		goto out_drop_write;
1266 	}
1267 
1268 	if (name[0] == '.' &&
1269 	   (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1270 		ret = -EEXIST;
1271 		goto out_drop_write;
1272 	}
1273 
1274 	if (subvol) {
1275 		ret = btrfs_mksubvol(&file->f_path, idmap, name,
1276 				     namelen, NULL, readonly, inherit);
1277 	} else {
1278 		struct fd src = fdget(fd);
1279 		struct inode *src_inode;
1280 		if (!src.file) {
1281 			ret = -EINVAL;
1282 			goto out_drop_write;
1283 		}
1284 
1285 		src_inode = file_inode(src.file);
1286 		if (src_inode->i_sb != file_inode(file)->i_sb) {
1287 			btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1288 				   "Snapshot src from another FS");
1289 			ret = -EXDEV;
1290 		} else if (!inode_owner_or_capable(idmap, src_inode)) {
1291 			/*
1292 			 * Subvolume creation is not restricted, but snapshots
1293 			 * are limited to own subvolumes only
1294 			 */
1295 			ret = -EPERM;
1296 		} else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1297 			/*
1298 			 * Snapshots must be made with the src_inode referring
1299 			 * to the subvolume inode, otherwise the permission
1300 			 * checking above is useless because we may have
1301 			 * permission on a lower directory but not the subvol
1302 			 * itself.
1303 			 */
1304 			ret = -EINVAL;
1305 		} else {
1306 			ret = btrfs_mksnapshot(&file->f_path, idmap,
1307 					       name, namelen,
1308 					       BTRFS_I(src_inode)->root,
1309 					       readonly, inherit);
1310 		}
1311 		fdput(src);
1312 	}
1313 out_drop_write:
1314 	mnt_drop_write_file(file);
1315 out:
1316 	return ret;
1317 }
1318 
1319 static noinline int btrfs_ioctl_snap_create(struct file *file,
1320 					    void __user *arg, int subvol)
1321 {
1322 	struct btrfs_ioctl_vol_args *vol_args;
1323 	int ret;
1324 
1325 	if (!S_ISDIR(file_inode(file)->i_mode))
1326 		return -ENOTDIR;
1327 
1328 	vol_args = memdup_user(arg, sizeof(*vol_args));
1329 	if (IS_ERR(vol_args))
1330 		return PTR_ERR(vol_args);
1331 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1332 
1333 	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1334 					vol_args->name, vol_args->fd, subvol,
1335 					false, NULL);
1336 
1337 	kfree(vol_args);
1338 	return ret;
1339 }
1340 
1341 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1342 					       void __user *arg, int subvol)
1343 {
1344 	struct btrfs_ioctl_vol_args_v2 *vol_args;
1345 	int ret;
1346 	bool readonly = false;
1347 	struct btrfs_qgroup_inherit *inherit = NULL;
1348 
1349 	if (!S_ISDIR(file_inode(file)->i_mode))
1350 		return -ENOTDIR;
1351 
1352 	vol_args = memdup_user(arg, sizeof(*vol_args));
1353 	if (IS_ERR(vol_args))
1354 		return PTR_ERR(vol_args);
1355 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1356 
1357 	if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1358 		ret = -EOPNOTSUPP;
1359 		goto free_args;
1360 	}
1361 
1362 	if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1363 		readonly = true;
1364 	if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1365 		u64 nums;
1366 
1367 		if (vol_args->size < sizeof(*inherit) ||
1368 		    vol_args->size > PAGE_SIZE) {
1369 			ret = -EINVAL;
1370 			goto free_args;
1371 		}
1372 		inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1373 		if (IS_ERR(inherit)) {
1374 			ret = PTR_ERR(inherit);
1375 			goto free_args;
1376 		}
1377 
1378 		if (inherit->num_qgroups > PAGE_SIZE ||
1379 		    inherit->num_ref_copies > PAGE_SIZE ||
1380 		    inherit->num_excl_copies > PAGE_SIZE) {
1381 			ret = -EINVAL;
1382 			goto free_inherit;
1383 		}
1384 
1385 		nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1386 		       2 * inherit->num_excl_copies;
1387 		if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1388 			ret = -EINVAL;
1389 			goto free_inherit;
1390 		}
1391 	}
1392 
1393 	ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1394 					vol_args->name, vol_args->fd, subvol,
1395 					readonly, inherit);
1396 	if (ret)
1397 		goto free_inherit;
1398 free_inherit:
1399 	kfree(inherit);
1400 free_args:
1401 	kfree(vol_args);
1402 	return ret;
1403 }
1404 
1405 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1406 						void __user *arg)
1407 {
1408 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1409 	struct btrfs_root *root = BTRFS_I(inode)->root;
1410 	int ret = 0;
1411 	u64 flags = 0;
1412 
1413 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1414 		return -EINVAL;
1415 
1416 	down_read(&fs_info->subvol_sem);
1417 	if (btrfs_root_readonly(root))
1418 		flags |= BTRFS_SUBVOL_RDONLY;
1419 	up_read(&fs_info->subvol_sem);
1420 
1421 	if (copy_to_user(arg, &flags, sizeof(flags)))
1422 		ret = -EFAULT;
1423 
1424 	return ret;
1425 }
1426 
1427 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1428 					      void __user *arg)
1429 {
1430 	struct inode *inode = file_inode(file);
1431 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1432 	struct btrfs_root *root = BTRFS_I(inode)->root;
1433 	struct btrfs_trans_handle *trans;
1434 	u64 root_flags;
1435 	u64 flags;
1436 	int ret = 0;
1437 
1438 	if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1439 		return -EPERM;
1440 
1441 	ret = mnt_want_write_file(file);
1442 	if (ret)
1443 		goto out;
1444 
1445 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1446 		ret = -EINVAL;
1447 		goto out_drop_write;
1448 	}
1449 
1450 	if (copy_from_user(&flags, arg, sizeof(flags))) {
1451 		ret = -EFAULT;
1452 		goto out_drop_write;
1453 	}
1454 
1455 	if (flags & ~BTRFS_SUBVOL_RDONLY) {
1456 		ret = -EOPNOTSUPP;
1457 		goto out_drop_write;
1458 	}
1459 
1460 	down_write(&fs_info->subvol_sem);
1461 
1462 	/* nothing to do */
1463 	if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1464 		goto out_drop_sem;
1465 
1466 	root_flags = btrfs_root_flags(&root->root_item);
1467 	if (flags & BTRFS_SUBVOL_RDONLY) {
1468 		btrfs_set_root_flags(&root->root_item,
1469 				     root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1470 	} else {
1471 		/*
1472 		 * Block RO -> RW transition if this subvolume is involved in
1473 		 * send
1474 		 */
1475 		spin_lock(&root->root_item_lock);
1476 		if (root->send_in_progress == 0) {
1477 			btrfs_set_root_flags(&root->root_item,
1478 				     root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1479 			spin_unlock(&root->root_item_lock);
1480 		} else {
1481 			spin_unlock(&root->root_item_lock);
1482 			btrfs_warn(fs_info,
1483 				   "Attempt to set subvolume %llu read-write during send",
1484 				   root->root_key.objectid);
1485 			ret = -EPERM;
1486 			goto out_drop_sem;
1487 		}
1488 	}
1489 
1490 	trans = btrfs_start_transaction(root, 1);
1491 	if (IS_ERR(trans)) {
1492 		ret = PTR_ERR(trans);
1493 		goto out_reset;
1494 	}
1495 
1496 	ret = btrfs_update_root(trans, fs_info->tree_root,
1497 				&root->root_key, &root->root_item);
1498 	if (ret < 0) {
1499 		btrfs_end_transaction(trans);
1500 		goto out_reset;
1501 	}
1502 
1503 	ret = btrfs_commit_transaction(trans);
1504 
1505 out_reset:
1506 	if (ret)
1507 		btrfs_set_root_flags(&root->root_item, root_flags);
1508 out_drop_sem:
1509 	up_write(&fs_info->subvol_sem);
1510 out_drop_write:
1511 	mnt_drop_write_file(file);
1512 out:
1513 	return ret;
1514 }
1515 
1516 static noinline int key_in_sk(struct btrfs_key *key,
1517 			      struct btrfs_ioctl_search_key *sk)
1518 {
1519 	struct btrfs_key test;
1520 	int ret;
1521 
1522 	test.objectid = sk->min_objectid;
1523 	test.type = sk->min_type;
1524 	test.offset = sk->min_offset;
1525 
1526 	ret = btrfs_comp_cpu_keys(key, &test);
1527 	if (ret < 0)
1528 		return 0;
1529 
1530 	test.objectid = sk->max_objectid;
1531 	test.type = sk->max_type;
1532 	test.offset = sk->max_offset;
1533 
1534 	ret = btrfs_comp_cpu_keys(key, &test);
1535 	if (ret > 0)
1536 		return 0;
1537 	return 1;
1538 }
1539 
1540 static noinline int copy_to_sk(struct btrfs_path *path,
1541 			       struct btrfs_key *key,
1542 			       struct btrfs_ioctl_search_key *sk,
1543 			       u64 *buf_size,
1544 			       char __user *ubuf,
1545 			       unsigned long *sk_offset,
1546 			       int *num_found)
1547 {
1548 	u64 found_transid;
1549 	struct extent_buffer *leaf;
1550 	struct btrfs_ioctl_search_header sh;
1551 	struct btrfs_key test;
1552 	unsigned long item_off;
1553 	unsigned long item_len;
1554 	int nritems;
1555 	int i;
1556 	int slot;
1557 	int ret = 0;
1558 
1559 	leaf = path->nodes[0];
1560 	slot = path->slots[0];
1561 	nritems = btrfs_header_nritems(leaf);
1562 
1563 	if (btrfs_header_generation(leaf) > sk->max_transid) {
1564 		i = nritems;
1565 		goto advance_key;
1566 	}
1567 	found_transid = btrfs_header_generation(leaf);
1568 
1569 	for (i = slot; i < nritems; i++) {
1570 		item_off = btrfs_item_ptr_offset(leaf, i);
1571 		item_len = btrfs_item_size(leaf, i);
1572 
1573 		btrfs_item_key_to_cpu(leaf, key, i);
1574 		if (!key_in_sk(key, sk))
1575 			continue;
1576 
1577 		if (sizeof(sh) + item_len > *buf_size) {
1578 			if (*num_found) {
1579 				ret = 1;
1580 				goto out;
1581 			}
1582 
1583 			/*
1584 			 * return one empty item back for v1, which does not
1585 			 * handle -EOVERFLOW
1586 			 */
1587 
1588 			*buf_size = sizeof(sh) + item_len;
1589 			item_len = 0;
1590 			ret = -EOVERFLOW;
1591 		}
1592 
1593 		if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1594 			ret = 1;
1595 			goto out;
1596 		}
1597 
1598 		sh.objectid = key->objectid;
1599 		sh.offset = key->offset;
1600 		sh.type = key->type;
1601 		sh.len = item_len;
1602 		sh.transid = found_transid;
1603 
1604 		/*
1605 		 * Copy search result header. If we fault then loop again so we
1606 		 * can fault in the pages and -EFAULT there if there's a
1607 		 * problem. Otherwise we'll fault and then copy the buffer in
1608 		 * properly this next time through
1609 		 */
1610 		if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1611 			ret = 0;
1612 			goto out;
1613 		}
1614 
1615 		*sk_offset += sizeof(sh);
1616 
1617 		if (item_len) {
1618 			char __user *up = ubuf + *sk_offset;
1619 			/*
1620 			 * Copy the item, same behavior as above, but reset the
1621 			 * * sk_offset so we copy the full thing again.
1622 			 */
1623 			if (read_extent_buffer_to_user_nofault(leaf, up,
1624 						item_off, item_len)) {
1625 				ret = 0;
1626 				*sk_offset -= sizeof(sh);
1627 				goto out;
1628 			}
1629 
1630 			*sk_offset += item_len;
1631 		}
1632 		(*num_found)++;
1633 
1634 		if (ret) /* -EOVERFLOW from above */
1635 			goto out;
1636 
1637 		if (*num_found >= sk->nr_items) {
1638 			ret = 1;
1639 			goto out;
1640 		}
1641 	}
1642 advance_key:
1643 	ret = 0;
1644 	test.objectid = sk->max_objectid;
1645 	test.type = sk->max_type;
1646 	test.offset = sk->max_offset;
1647 	if (btrfs_comp_cpu_keys(key, &test) >= 0)
1648 		ret = 1;
1649 	else if (key->offset < (u64)-1)
1650 		key->offset++;
1651 	else if (key->type < (u8)-1) {
1652 		key->offset = 0;
1653 		key->type++;
1654 	} else if (key->objectid < (u64)-1) {
1655 		key->offset = 0;
1656 		key->type = 0;
1657 		key->objectid++;
1658 	} else
1659 		ret = 1;
1660 out:
1661 	/*
1662 	 *  0: all items from this leaf copied, continue with next
1663 	 *  1: * more items can be copied, but unused buffer is too small
1664 	 *     * all items were found
1665 	 *     Either way, it will stops the loop which iterates to the next
1666 	 *     leaf
1667 	 *  -EOVERFLOW: item was to large for buffer
1668 	 *  -EFAULT: could not copy extent buffer back to userspace
1669 	 */
1670 	return ret;
1671 }
1672 
1673 static noinline int search_ioctl(struct inode *inode,
1674 				 struct btrfs_ioctl_search_key *sk,
1675 				 u64 *buf_size,
1676 				 char __user *ubuf)
1677 {
1678 	struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1679 	struct btrfs_root *root;
1680 	struct btrfs_key key;
1681 	struct btrfs_path *path;
1682 	int ret;
1683 	int num_found = 0;
1684 	unsigned long sk_offset = 0;
1685 
1686 	if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1687 		*buf_size = sizeof(struct btrfs_ioctl_search_header);
1688 		return -EOVERFLOW;
1689 	}
1690 
1691 	path = btrfs_alloc_path();
1692 	if (!path)
1693 		return -ENOMEM;
1694 
1695 	if (sk->tree_id == 0) {
1696 		/* search the root of the inode that was passed */
1697 		root = btrfs_grab_root(BTRFS_I(inode)->root);
1698 	} else {
1699 		root = btrfs_get_fs_root(info, sk->tree_id, true);
1700 		if (IS_ERR(root)) {
1701 			btrfs_free_path(path);
1702 			return PTR_ERR(root);
1703 		}
1704 	}
1705 
1706 	key.objectid = sk->min_objectid;
1707 	key.type = sk->min_type;
1708 	key.offset = sk->min_offset;
1709 
1710 	while (1) {
1711 		ret = -EFAULT;
1712 		/*
1713 		 * Ensure that the whole user buffer is faulted in at sub-page
1714 		 * granularity, otherwise the loop may live-lock.
1715 		 */
1716 		if (fault_in_subpage_writeable(ubuf + sk_offset,
1717 					       *buf_size - sk_offset))
1718 			break;
1719 
1720 		ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1721 		if (ret != 0) {
1722 			if (ret > 0)
1723 				ret = 0;
1724 			goto err;
1725 		}
1726 		ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1727 				 &sk_offset, &num_found);
1728 		btrfs_release_path(path);
1729 		if (ret)
1730 			break;
1731 
1732 	}
1733 	if (ret > 0)
1734 		ret = 0;
1735 err:
1736 	sk->nr_items = num_found;
1737 	btrfs_put_root(root);
1738 	btrfs_free_path(path);
1739 	return ret;
1740 }
1741 
1742 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1743 					    void __user *argp)
1744 {
1745 	struct btrfs_ioctl_search_args __user *uargs = argp;
1746 	struct btrfs_ioctl_search_key sk;
1747 	int ret;
1748 	u64 buf_size;
1749 
1750 	if (!capable(CAP_SYS_ADMIN))
1751 		return -EPERM;
1752 
1753 	if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1754 		return -EFAULT;
1755 
1756 	buf_size = sizeof(uargs->buf);
1757 
1758 	ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1759 
1760 	/*
1761 	 * In the origin implementation an overflow is handled by returning a
1762 	 * search header with a len of zero, so reset ret.
1763 	 */
1764 	if (ret == -EOVERFLOW)
1765 		ret = 0;
1766 
1767 	if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1768 		ret = -EFAULT;
1769 	return ret;
1770 }
1771 
1772 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1773 					       void __user *argp)
1774 {
1775 	struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1776 	struct btrfs_ioctl_search_args_v2 args;
1777 	int ret;
1778 	u64 buf_size;
1779 	const u64 buf_limit = SZ_16M;
1780 
1781 	if (!capable(CAP_SYS_ADMIN))
1782 		return -EPERM;
1783 
1784 	/* copy search header and buffer size */
1785 	if (copy_from_user(&args, uarg, sizeof(args)))
1786 		return -EFAULT;
1787 
1788 	buf_size = args.buf_size;
1789 
1790 	/* limit result size to 16MB */
1791 	if (buf_size > buf_limit)
1792 		buf_size = buf_limit;
1793 
1794 	ret = search_ioctl(inode, &args.key, &buf_size,
1795 			   (char __user *)(&uarg->buf[0]));
1796 	if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1797 		ret = -EFAULT;
1798 	else if (ret == -EOVERFLOW &&
1799 		copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1800 		ret = -EFAULT;
1801 
1802 	return ret;
1803 }
1804 
1805 /*
1806  * Search INODE_REFs to identify path name of 'dirid' directory
1807  * in a 'tree_id' tree. and sets path name to 'name'.
1808  */
1809 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1810 				u64 tree_id, u64 dirid, char *name)
1811 {
1812 	struct btrfs_root *root;
1813 	struct btrfs_key key;
1814 	char *ptr;
1815 	int ret = -1;
1816 	int slot;
1817 	int len;
1818 	int total_len = 0;
1819 	struct btrfs_inode_ref *iref;
1820 	struct extent_buffer *l;
1821 	struct btrfs_path *path;
1822 
1823 	if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1824 		name[0]='\0';
1825 		return 0;
1826 	}
1827 
1828 	path = btrfs_alloc_path();
1829 	if (!path)
1830 		return -ENOMEM;
1831 
1832 	ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1833 
1834 	root = btrfs_get_fs_root(info, tree_id, true);
1835 	if (IS_ERR(root)) {
1836 		ret = PTR_ERR(root);
1837 		root = NULL;
1838 		goto out;
1839 	}
1840 
1841 	key.objectid = dirid;
1842 	key.type = BTRFS_INODE_REF_KEY;
1843 	key.offset = (u64)-1;
1844 
1845 	while (1) {
1846 		ret = btrfs_search_backwards(root, &key, path);
1847 		if (ret < 0)
1848 			goto out;
1849 		else if (ret > 0) {
1850 			ret = -ENOENT;
1851 			goto out;
1852 		}
1853 
1854 		l = path->nodes[0];
1855 		slot = path->slots[0];
1856 
1857 		iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1858 		len = btrfs_inode_ref_name_len(l, iref);
1859 		ptr -= len + 1;
1860 		total_len += len + 1;
1861 		if (ptr < name) {
1862 			ret = -ENAMETOOLONG;
1863 			goto out;
1864 		}
1865 
1866 		*(ptr + len) = '/';
1867 		read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1868 
1869 		if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1870 			break;
1871 
1872 		btrfs_release_path(path);
1873 		key.objectid = key.offset;
1874 		key.offset = (u64)-1;
1875 		dirid = key.objectid;
1876 	}
1877 	memmove(name, ptr, total_len);
1878 	name[total_len] = '\0';
1879 	ret = 0;
1880 out:
1881 	btrfs_put_root(root);
1882 	btrfs_free_path(path);
1883 	return ret;
1884 }
1885 
1886 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1887 				struct inode *inode,
1888 				struct btrfs_ioctl_ino_lookup_user_args *args)
1889 {
1890 	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1891 	struct super_block *sb = inode->i_sb;
1892 	struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1893 	u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1894 	u64 dirid = args->dirid;
1895 	unsigned long item_off;
1896 	unsigned long item_len;
1897 	struct btrfs_inode_ref *iref;
1898 	struct btrfs_root_ref *rref;
1899 	struct btrfs_root *root = NULL;
1900 	struct btrfs_path *path;
1901 	struct btrfs_key key, key2;
1902 	struct extent_buffer *leaf;
1903 	struct inode *temp_inode;
1904 	char *ptr;
1905 	int slot;
1906 	int len;
1907 	int total_len = 0;
1908 	int ret;
1909 
1910 	path = btrfs_alloc_path();
1911 	if (!path)
1912 		return -ENOMEM;
1913 
1914 	/*
1915 	 * If the bottom subvolume does not exist directly under upper_limit,
1916 	 * construct the path in from the bottom up.
1917 	 */
1918 	if (dirid != upper_limit.objectid) {
1919 		ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1920 
1921 		root = btrfs_get_fs_root(fs_info, treeid, true);
1922 		if (IS_ERR(root)) {
1923 			ret = PTR_ERR(root);
1924 			goto out;
1925 		}
1926 
1927 		key.objectid = dirid;
1928 		key.type = BTRFS_INODE_REF_KEY;
1929 		key.offset = (u64)-1;
1930 		while (1) {
1931 			ret = btrfs_search_backwards(root, &key, path);
1932 			if (ret < 0)
1933 				goto out_put;
1934 			else if (ret > 0) {
1935 				ret = -ENOENT;
1936 				goto out_put;
1937 			}
1938 
1939 			leaf = path->nodes[0];
1940 			slot = path->slots[0];
1941 
1942 			iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1943 			len = btrfs_inode_ref_name_len(leaf, iref);
1944 			ptr -= len + 1;
1945 			total_len += len + 1;
1946 			if (ptr < args->path) {
1947 				ret = -ENAMETOOLONG;
1948 				goto out_put;
1949 			}
1950 
1951 			*(ptr + len) = '/';
1952 			read_extent_buffer(leaf, ptr,
1953 					(unsigned long)(iref + 1), len);
1954 
1955 			/* Check the read+exec permission of this directory */
1956 			ret = btrfs_previous_item(root, path, dirid,
1957 						  BTRFS_INODE_ITEM_KEY);
1958 			if (ret < 0) {
1959 				goto out_put;
1960 			} else if (ret > 0) {
1961 				ret = -ENOENT;
1962 				goto out_put;
1963 			}
1964 
1965 			leaf = path->nodes[0];
1966 			slot = path->slots[0];
1967 			btrfs_item_key_to_cpu(leaf, &key2, slot);
1968 			if (key2.objectid != dirid) {
1969 				ret = -ENOENT;
1970 				goto out_put;
1971 			}
1972 
1973 			/*
1974 			 * We don't need the path anymore, so release it and
1975 			 * avoid deadlocks and lockdep warnings in case
1976 			 * btrfs_iget() needs to lookup the inode from its root
1977 			 * btree and lock the same leaf.
1978 			 */
1979 			btrfs_release_path(path);
1980 			temp_inode = btrfs_iget(sb, key2.objectid, root);
1981 			if (IS_ERR(temp_inode)) {
1982 				ret = PTR_ERR(temp_inode);
1983 				goto out_put;
1984 			}
1985 			ret = inode_permission(idmap, temp_inode,
1986 					       MAY_READ | MAY_EXEC);
1987 			iput(temp_inode);
1988 			if (ret) {
1989 				ret = -EACCES;
1990 				goto out_put;
1991 			}
1992 
1993 			if (key.offset == upper_limit.objectid)
1994 				break;
1995 			if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
1996 				ret = -EACCES;
1997 				goto out_put;
1998 			}
1999 
2000 			key.objectid = key.offset;
2001 			key.offset = (u64)-1;
2002 			dirid = key.objectid;
2003 		}
2004 
2005 		memmove(args->path, ptr, total_len);
2006 		args->path[total_len] = '\0';
2007 		btrfs_put_root(root);
2008 		root = NULL;
2009 		btrfs_release_path(path);
2010 	}
2011 
2012 	/* Get the bottom subvolume's name from ROOT_REF */
2013 	key.objectid = treeid;
2014 	key.type = BTRFS_ROOT_REF_KEY;
2015 	key.offset = args->treeid;
2016 	ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2017 	if (ret < 0) {
2018 		goto out;
2019 	} else if (ret > 0) {
2020 		ret = -ENOENT;
2021 		goto out;
2022 	}
2023 
2024 	leaf = path->nodes[0];
2025 	slot = path->slots[0];
2026 	btrfs_item_key_to_cpu(leaf, &key, slot);
2027 
2028 	item_off = btrfs_item_ptr_offset(leaf, slot);
2029 	item_len = btrfs_item_size(leaf, slot);
2030 	/* Check if dirid in ROOT_REF corresponds to passed dirid */
2031 	rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2032 	if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2033 		ret = -EINVAL;
2034 		goto out;
2035 	}
2036 
2037 	/* Copy subvolume's name */
2038 	item_off += sizeof(struct btrfs_root_ref);
2039 	item_len -= sizeof(struct btrfs_root_ref);
2040 	read_extent_buffer(leaf, args->name, item_off, item_len);
2041 	args->name[item_len] = 0;
2042 
2043 out_put:
2044 	btrfs_put_root(root);
2045 out:
2046 	btrfs_free_path(path);
2047 	return ret;
2048 }
2049 
2050 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2051 					   void __user *argp)
2052 {
2053 	struct btrfs_ioctl_ino_lookup_args *args;
2054 	int ret = 0;
2055 
2056 	args = memdup_user(argp, sizeof(*args));
2057 	if (IS_ERR(args))
2058 		return PTR_ERR(args);
2059 
2060 	/*
2061 	 * Unprivileged query to obtain the containing subvolume root id. The
2062 	 * path is reset so it's consistent with btrfs_search_path_in_tree.
2063 	 */
2064 	if (args->treeid == 0)
2065 		args->treeid = root->root_key.objectid;
2066 
2067 	if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2068 		args->name[0] = 0;
2069 		goto out;
2070 	}
2071 
2072 	if (!capable(CAP_SYS_ADMIN)) {
2073 		ret = -EPERM;
2074 		goto out;
2075 	}
2076 
2077 	ret = btrfs_search_path_in_tree(root->fs_info,
2078 					args->treeid, args->objectid,
2079 					args->name);
2080 
2081 out:
2082 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2083 		ret = -EFAULT;
2084 
2085 	kfree(args);
2086 	return ret;
2087 }
2088 
2089 /*
2090  * Version of ino_lookup ioctl (unprivileged)
2091  *
2092  * The main differences from ino_lookup ioctl are:
2093  *
2094  *   1. Read + Exec permission will be checked using inode_permission() during
2095  *      path construction. -EACCES will be returned in case of failure.
2096  *   2. Path construction will be stopped at the inode number which corresponds
2097  *      to the fd with which this ioctl is called. If constructed path does not
2098  *      exist under fd's inode, -EACCES will be returned.
2099  *   3. The name of bottom subvolume is also searched and filled.
2100  */
2101 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2102 {
2103 	struct btrfs_ioctl_ino_lookup_user_args *args;
2104 	struct inode *inode;
2105 	int ret;
2106 
2107 	args = memdup_user(argp, sizeof(*args));
2108 	if (IS_ERR(args))
2109 		return PTR_ERR(args);
2110 
2111 	inode = file_inode(file);
2112 
2113 	if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2114 	    BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2115 		/*
2116 		 * The subvolume does not exist under fd with which this is
2117 		 * called
2118 		 */
2119 		kfree(args);
2120 		return -EACCES;
2121 	}
2122 
2123 	ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2124 
2125 	if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2126 		ret = -EFAULT;
2127 
2128 	kfree(args);
2129 	return ret;
2130 }
2131 
2132 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2133 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2134 {
2135 	struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2136 	struct btrfs_fs_info *fs_info;
2137 	struct btrfs_root *root;
2138 	struct btrfs_path *path;
2139 	struct btrfs_key key;
2140 	struct btrfs_root_item *root_item;
2141 	struct btrfs_root_ref *rref;
2142 	struct extent_buffer *leaf;
2143 	unsigned long item_off;
2144 	unsigned long item_len;
2145 	int slot;
2146 	int ret = 0;
2147 
2148 	path = btrfs_alloc_path();
2149 	if (!path)
2150 		return -ENOMEM;
2151 
2152 	subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2153 	if (!subvol_info) {
2154 		btrfs_free_path(path);
2155 		return -ENOMEM;
2156 	}
2157 
2158 	fs_info = BTRFS_I(inode)->root->fs_info;
2159 
2160 	/* Get root_item of inode's subvolume */
2161 	key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2162 	root = btrfs_get_fs_root(fs_info, key.objectid, true);
2163 	if (IS_ERR(root)) {
2164 		ret = PTR_ERR(root);
2165 		goto out_free;
2166 	}
2167 	root_item = &root->root_item;
2168 
2169 	subvol_info->treeid = key.objectid;
2170 
2171 	subvol_info->generation = btrfs_root_generation(root_item);
2172 	subvol_info->flags = btrfs_root_flags(root_item);
2173 
2174 	memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2175 	memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2176 						    BTRFS_UUID_SIZE);
2177 	memcpy(subvol_info->received_uuid, root_item->received_uuid,
2178 						    BTRFS_UUID_SIZE);
2179 
2180 	subvol_info->ctransid = btrfs_root_ctransid(root_item);
2181 	subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2182 	subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2183 
2184 	subvol_info->otransid = btrfs_root_otransid(root_item);
2185 	subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2186 	subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2187 
2188 	subvol_info->stransid = btrfs_root_stransid(root_item);
2189 	subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2190 	subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2191 
2192 	subvol_info->rtransid = btrfs_root_rtransid(root_item);
2193 	subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2194 	subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2195 
2196 	if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2197 		/* Search root tree for ROOT_BACKREF of this subvolume */
2198 		key.type = BTRFS_ROOT_BACKREF_KEY;
2199 		key.offset = 0;
2200 		ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2201 		if (ret < 0) {
2202 			goto out;
2203 		} else if (path->slots[0] >=
2204 			   btrfs_header_nritems(path->nodes[0])) {
2205 			ret = btrfs_next_leaf(fs_info->tree_root, path);
2206 			if (ret < 0) {
2207 				goto out;
2208 			} else if (ret > 0) {
2209 				ret = -EUCLEAN;
2210 				goto out;
2211 			}
2212 		}
2213 
2214 		leaf = path->nodes[0];
2215 		slot = path->slots[0];
2216 		btrfs_item_key_to_cpu(leaf, &key, slot);
2217 		if (key.objectid == subvol_info->treeid &&
2218 		    key.type == BTRFS_ROOT_BACKREF_KEY) {
2219 			subvol_info->parent_id = key.offset;
2220 
2221 			rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2222 			subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2223 
2224 			item_off = btrfs_item_ptr_offset(leaf, slot)
2225 					+ sizeof(struct btrfs_root_ref);
2226 			item_len = btrfs_item_size(leaf, slot)
2227 					- sizeof(struct btrfs_root_ref);
2228 			read_extent_buffer(leaf, subvol_info->name,
2229 					   item_off, item_len);
2230 		} else {
2231 			ret = -ENOENT;
2232 			goto out;
2233 		}
2234 	}
2235 
2236 	btrfs_free_path(path);
2237 	path = NULL;
2238 	if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2239 		ret = -EFAULT;
2240 
2241 out:
2242 	btrfs_put_root(root);
2243 out_free:
2244 	btrfs_free_path(path);
2245 	kfree(subvol_info);
2246 	return ret;
2247 }
2248 
2249 /*
2250  * Return ROOT_REF information of the subvolume containing this inode
2251  * except the subvolume name.
2252  */
2253 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2254 					  void __user *argp)
2255 {
2256 	struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2257 	struct btrfs_root_ref *rref;
2258 	struct btrfs_path *path;
2259 	struct btrfs_key key;
2260 	struct extent_buffer *leaf;
2261 	u64 objectid;
2262 	int slot;
2263 	int ret;
2264 	u8 found;
2265 
2266 	path = btrfs_alloc_path();
2267 	if (!path)
2268 		return -ENOMEM;
2269 
2270 	rootrefs = memdup_user(argp, sizeof(*rootrefs));
2271 	if (IS_ERR(rootrefs)) {
2272 		btrfs_free_path(path);
2273 		return PTR_ERR(rootrefs);
2274 	}
2275 
2276 	objectid = root->root_key.objectid;
2277 	key.objectid = objectid;
2278 	key.type = BTRFS_ROOT_REF_KEY;
2279 	key.offset = rootrefs->min_treeid;
2280 	found = 0;
2281 
2282 	root = root->fs_info->tree_root;
2283 	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2284 	if (ret < 0) {
2285 		goto out;
2286 	} else if (path->slots[0] >=
2287 		   btrfs_header_nritems(path->nodes[0])) {
2288 		ret = btrfs_next_leaf(root, path);
2289 		if (ret < 0) {
2290 			goto out;
2291 		} else if (ret > 0) {
2292 			ret = -EUCLEAN;
2293 			goto out;
2294 		}
2295 	}
2296 	while (1) {
2297 		leaf = path->nodes[0];
2298 		slot = path->slots[0];
2299 
2300 		btrfs_item_key_to_cpu(leaf, &key, slot);
2301 		if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2302 			ret = 0;
2303 			goto out;
2304 		}
2305 
2306 		if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2307 			ret = -EOVERFLOW;
2308 			goto out;
2309 		}
2310 
2311 		rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2312 		rootrefs->rootref[found].treeid = key.offset;
2313 		rootrefs->rootref[found].dirid =
2314 				  btrfs_root_ref_dirid(leaf, rref);
2315 		found++;
2316 
2317 		ret = btrfs_next_item(root, path);
2318 		if (ret < 0) {
2319 			goto out;
2320 		} else if (ret > 0) {
2321 			ret = -EUCLEAN;
2322 			goto out;
2323 		}
2324 	}
2325 
2326 out:
2327 	btrfs_free_path(path);
2328 
2329 	if (!ret || ret == -EOVERFLOW) {
2330 		rootrefs->num_items = found;
2331 		/* update min_treeid for next search */
2332 		if (found)
2333 			rootrefs->min_treeid =
2334 				rootrefs->rootref[found - 1].treeid + 1;
2335 		if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2336 			ret = -EFAULT;
2337 	}
2338 
2339 	kfree(rootrefs);
2340 
2341 	return ret;
2342 }
2343 
2344 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2345 					     void __user *arg,
2346 					     bool destroy_v2)
2347 {
2348 	struct dentry *parent = file->f_path.dentry;
2349 	struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2350 	struct dentry *dentry;
2351 	struct inode *dir = d_inode(parent);
2352 	struct inode *inode;
2353 	struct btrfs_root *root = BTRFS_I(dir)->root;
2354 	struct btrfs_root *dest = NULL;
2355 	struct btrfs_ioctl_vol_args *vol_args = NULL;
2356 	struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2357 	struct mnt_idmap *idmap = file_mnt_idmap(file);
2358 	char *subvol_name, *subvol_name_ptr = NULL;
2359 	int subvol_namelen;
2360 	int err = 0;
2361 	bool destroy_parent = false;
2362 
2363 	/* We don't support snapshots with extent tree v2 yet. */
2364 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2365 		btrfs_err(fs_info,
2366 			  "extent tree v2 doesn't support snapshot deletion yet");
2367 		return -EOPNOTSUPP;
2368 	}
2369 
2370 	if (destroy_v2) {
2371 		vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2372 		if (IS_ERR(vol_args2))
2373 			return PTR_ERR(vol_args2);
2374 
2375 		if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2376 			err = -EOPNOTSUPP;
2377 			goto out;
2378 		}
2379 
2380 		/*
2381 		 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2382 		 * name, same as v1 currently does.
2383 		 */
2384 		if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2385 			vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2386 			subvol_name = vol_args2->name;
2387 
2388 			err = mnt_want_write_file(file);
2389 			if (err)
2390 				goto out;
2391 		} else {
2392 			struct inode *old_dir;
2393 
2394 			if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2395 				err = -EINVAL;
2396 				goto out;
2397 			}
2398 
2399 			err = mnt_want_write_file(file);
2400 			if (err)
2401 				goto out;
2402 
2403 			dentry = btrfs_get_dentry(fs_info->sb,
2404 					BTRFS_FIRST_FREE_OBJECTID,
2405 					vol_args2->subvolid, 0);
2406 			if (IS_ERR(dentry)) {
2407 				err = PTR_ERR(dentry);
2408 				goto out_drop_write;
2409 			}
2410 
2411 			/*
2412 			 * Change the default parent since the subvolume being
2413 			 * deleted can be outside of the current mount point.
2414 			 */
2415 			parent = btrfs_get_parent(dentry);
2416 
2417 			/*
2418 			 * At this point dentry->d_name can point to '/' if the
2419 			 * subvolume we want to destroy is outsite of the
2420 			 * current mount point, so we need to release the
2421 			 * current dentry and execute the lookup to return a new
2422 			 * one with ->d_name pointing to the
2423 			 * <mount point>/subvol_name.
2424 			 */
2425 			dput(dentry);
2426 			if (IS_ERR(parent)) {
2427 				err = PTR_ERR(parent);
2428 				goto out_drop_write;
2429 			}
2430 			old_dir = dir;
2431 			dir = d_inode(parent);
2432 
2433 			/*
2434 			 * If v2 was used with SPEC_BY_ID, a new parent was
2435 			 * allocated since the subvolume can be outside of the
2436 			 * current mount point. Later on we need to release this
2437 			 * new parent dentry.
2438 			 */
2439 			destroy_parent = true;
2440 
2441 			/*
2442 			 * On idmapped mounts, deletion via subvolid is
2443 			 * restricted to subvolumes that are immediate
2444 			 * ancestors of the inode referenced by the file
2445 			 * descriptor in the ioctl. Otherwise the idmapping
2446 			 * could potentially be abused to delete subvolumes
2447 			 * anywhere in the filesystem the user wouldn't be able
2448 			 * to delete without an idmapped mount.
2449 			 */
2450 			if (old_dir != dir && idmap != &nop_mnt_idmap) {
2451 				err = -EOPNOTSUPP;
2452 				goto free_parent;
2453 			}
2454 
2455 			subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2456 						fs_info, vol_args2->subvolid);
2457 			if (IS_ERR(subvol_name_ptr)) {
2458 				err = PTR_ERR(subvol_name_ptr);
2459 				goto free_parent;
2460 			}
2461 			/* subvol_name_ptr is already nul terminated */
2462 			subvol_name = (char *)kbasename(subvol_name_ptr);
2463 		}
2464 	} else {
2465 		vol_args = memdup_user(arg, sizeof(*vol_args));
2466 		if (IS_ERR(vol_args))
2467 			return PTR_ERR(vol_args);
2468 
2469 		vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2470 		subvol_name = vol_args->name;
2471 
2472 		err = mnt_want_write_file(file);
2473 		if (err)
2474 			goto out;
2475 	}
2476 
2477 	subvol_namelen = strlen(subvol_name);
2478 
2479 	if (strchr(subvol_name, '/') ||
2480 	    strncmp(subvol_name, "..", subvol_namelen) == 0) {
2481 		err = -EINVAL;
2482 		goto free_subvol_name;
2483 	}
2484 
2485 	if (!S_ISDIR(dir->i_mode)) {
2486 		err = -ENOTDIR;
2487 		goto free_subvol_name;
2488 	}
2489 
2490 	err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2491 	if (err == -EINTR)
2492 		goto free_subvol_name;
2493 	dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2494 	if (IS_ERR(dentry)) {
2495 		err = PTR_ERR(dentry);
2496 		goto out_unlock_dir;
2497 	}
2498 
2499 	if (d_really_is_negative(dentry)) {
2500 		err = -ENOENT;
2501 		goto out_dput;
2502 	}
2503 
2504 	inode = d_inode(dentry);
2505 	dest = BTRFS_I(inode)->root;
2506 	if (!capable(CAP_SYS_ADMIN)) {
2507 		/*
2508 		 * Regular user.  Only allow this with a special mount
2509 		 * option, when the user has write+exec access to the
2510 		 * subvol root, and when rmdir(2) would have been
2511 		 * allowed.
2512 		 *
2513 		 * Note that this is _not_ check that the subvol is
2514 		 * empty or doesn't contain data that we wouldn't
2515 		 * otherwise be able to delete.
2516 		 *
2517 		 * Users who want to delete empty subvols should try
2518 		 * rmdir(2).
2519 		 */
2520 		err = -EPERM;
2521 		if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2522 			goto out_dput;
2523 
2524 		/*
2525 		 * Do not allow deletion if the parent dir is the same
2526 		 * as the dir to be deleted.  That means the ioctl
2527 		 * must be called on the dentry referencing the root
2528 		 * of the subvol, not a random directory contained
2529 		 * within it.
2530 		 */
2531 		err = -EINVAL;
2532 		if (root == dest)
2533 			goto out_dput;
2534 
2535 		err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2536 		if (err)
2537 			goto out_dput;
2538 	}
2539 
2540 	/* check if subvolume may be deleted by a user */
2541 	err = btrfs_may_delete(idmap, dir, dentry, 1);
2542 	if (err)
2543 		goto out_dput;
2544 
2545 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2546 		err = -EINVAL;
2547 		goto out_dput;
2548 	}
2549 
2550 	btrfs_inode_lock(BTRFS_I(inode), 0);
2551 	err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2552 	btrfs_inode_unlock(BTRFS_I(inode), 0);
2553 	if (!err)
2554 		d_delete_notify(dir, dentry);
2555 
2556 out_dput:
2557 	dput(dentry);
2558 out_unlock_dir:
2559 	btrfs_inode_unlock(BTRFS_I(dir), 0);
2560 free_subvol_name:
2561 	kfree(subvol_name_ptr);
2562 free_parent:
2563 	if (destroy_parent)
2564 		dput(parent);
2565 out_drop_write:
2566 	mnt_drop_write_file(file);
2567 out:
2568 	kfree(vol_args2);
2569 	kfree(vol_args);
2570 	return err;
2571 }
2572 
2573 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2574 {
2575 	struct inode *inode = file_inode(file);
2576 	struct btrfs_root *root = BTRFS_I(inode)->root;
2577 	struct btrfs_ioctl_defrag_range_args range = {0};
2578 	int ret;
2579 
2580 	ret = mnt_want_write_file(file);
2581 	if (ret)
2582 		return ret;
2583 
2584 	if (btrfs_root_readonly(root)) {
2585 		ret = -EROFS;
2586 		goto out;
2587 	}
2588 
2589 	switch (inode->i_mode & S_IFMT) {
2590 	case S_IFDIR:
2591 		if (!capable(CAP_SYS_ADMIN)) {
2592 			ret = -EPERM;
2593 			goto out;
2594 		}
2595 		ret = btrfs_defrag_root(root);
2596 		break;
2597 	case S_IFREG:
2598 		/*
2599 		 * Note that this does not check the file descriptor for write
2600 		 * access. This prevents defragmenting executables that are
2601 		 * running and allows defrag on files open in read-only mode.
2602 		 */
2603 		if (!capable(CAP_SYS_ADMIN) &&
2604 		    inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2605 			ret = -EPERM;
2606 			goto out;
2607 		}
2608 
2609 		if (argp) {
2610 			if (copy_from_user(&range, argp, sizeof(range))) {
2611 				ret = -EFAULT;
2612 				goto out;
2613 			}
2614 			if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2615 				ret = -EOPNOTSUPP;
2616 				goto out;
2617 			}
2618 			/* compression requires us to start the IO */
2619 			if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2620 				range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2621 				range.extent_thresh = (u32)-1;
2622 			}
2623 		} else {
2624 			/* the rest are all set to zero by kzalloc */
2625 			range.len = (u64)-1;
2626 		}
2627 		ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2628 					&range, BTRFS_OLDEST_GENERATION, 0);
2629 		if (ret > 0)
2630 			ret = 0;
2631 		break;
2632 	default:
2633 		ret = -EINVAL;
2634 	}
2635 out:
2636 	mnt_drop_write_file(file);
2637 	return ret;
2638 }
2639 
2640 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2641 {
2642 	struct btrfs_ioctl_vol_args *vol_args;
2643 	bool restore_op = false;
2644 	int ret;
2645 
2646 	if (!capable(CAP_SYS_ADMIN))
2647 		return -EPERM;
2648 
2649 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2650 		btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2651 		return -EINVAL;
2652 	}
2653 
2654 	if (fs_info->fs_devices->temp_fsid) {
2655 		btrfs_err(fs_info,
2656 			  "device add not supported on cloned temp-fsid mount");
2657 		return -EINVAL;
2658 	}
2659 
2660 	if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2661 		if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2662 			return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2663 
2664 		/*
2665 		 * We can do the device add because we have a paused balanced,
2666 		 * change the exclusive op type and remember we should bring
2667 		 * back the paused balance
2668 		 */
2669 		fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2670 		btrfs_exclop_start_unlock(fs_info);
2671 		restore_op = true;
2672 	}
2673 
2674 	vol_args = memdup_user(arg, sizeof(*vol_args));
2675 	if (IS_ERR(vol_args)) {
2676 		ret = PTR_ERR(vol_args);
2677 		goto out;
2678 	}
2679 
2680 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2681 	ret = btrfs_init_new_device(fs_info, vol_args->name);
2682 
2683 	if (!ret)
2684 		btrfs_info(fs_info, "disk added %s", vol_args->name);
2685 
2686 	kfree(vol_args);
2687 out:
2688 	if (restore_op)
2689 		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2690 	else
2691 		btrfs_exclop_finish(fs_info);
2692 	return ret;
2693 }
2694 
2695 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2696 {
2697 	BTRFS_DEV_LOOKUP_ARGS(args);
2698 	struct inode *inode = file_inode(file);
2699 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2700 	struct btrfs_ioctl_vol_args_v2 *vol_args;
2701 	struct bdev_handle *bdev_handle = NULL;
2702 	int ret;
2703 	bool cancel = false;
2704 
2705 	if (!capable(CAP_SYS_ADMIN))
2706 		return -EPERM;
2707 
2708 	vol_args = memdup_user(arg, sizeof(*vol_args));
2709 	if (IS_ERR(vol_args))
2710 		return PTR_ERR(vol_args);
2711 
2712 	if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2713 		ret = -EOPNOTSUPP;
2714 		goto out;
2715 	}
2716 
2717 	vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2718 	if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2719 		args.devid = vol_args->devid;
2720 	} else if (!strcmp("cancel", vol_args->name)) {
2721 		cancel = true;
2722 	} else {
2723 		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2724 		if (ret)
2725 			goto out;
2726 	}
2727 
2728 	ret = mnt_want_write_file(file);
2729 	if (ret)
2730 		goto out;
2731 
2732 	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2733 					   cancel);
2734 	if (ret)
2735 		goto err_drop;
2736 
2737 	/* Exclusive operation is now claimed */
2738 	ret = btrfs_rm_device(fs_info, &args, &bdev_handle);
2739 
2740 	btrfs_exclop_finish(fs_info);
2741 
2742 	if (!ret) {
2743 		if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2744 			btrfs_info(fs_info, "device deleted: id %llu",
2745 					vol_args->devid);
2746 		else
2747 			btrfs_info(fs_info, "device deleted: %s",
2748 					vol_args->name);
2749 	}
2750 err_drop:
2751 	mnt_drop_write_file(file);
2752 	if (bdev_handle)
2753 		bdev_release(bdev_handle);
2754 out:
2755 	btrfs_put_dev_args_from_path(&args);
2756 	kfree(vol_args);
2757 	return ret;
2758 }
2759 
2760 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2761 {
2762 	BTRFS_DEV_LOOKUP_ARGS(args);
2763 	struct inode *inode = file_inode(file);
2764 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2765 	struct btrfs_ioctl_vol_args *vol_args;
2766 	struct bdev_handle *bdev_handle = NULL;
2767 	int ret;
2768 	bool cancel = false;
2769 
2770 	if (!capable(CAP_SYS_ADMIN))
2771 		return -EPERM;
2772 
2773 	vol_args = memdup_user(arg, sizeof(*vol_args));
2774 	if (IS_ERR(vol_args))
2775 		return PTR_ERR(vol_args);
2776 
2777 	vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2778 	if (!strcmp("cancel", vol_args->name)) {
2779 		cancel = true;
2780 	} else {
2781 		ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2782 		if (ret)
2783 			goto out;
2784 	}
2785 
2786 	ret = mnt_want_write_file(file);
2787 	if (ret)
2788 		goto out;
2789 
2790 	ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2791 					   cancel);
2792 	if (ret == 0) {
2793 		ret = btrfs_rm_device(fs_info, &args, &bdev_handle);
2794 		if (!ret)
2795 			btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2796 		btrfs_exclop_finish(fs_info);
2797 	}
2798 
2799 	mnt_drop_write_file(file);
2800 	if (bdev_handle)
2801 		bdev_release(bdev_handle);
2802 out:
2803 	btrfs_put_dev_args_from_path(&args);
2804 	kfree(vol_args);
2805 	return ret;
2806 }
2807 
2808 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2809 				void __user *arg)
2810 {
2811 	struct btrfs_ioctl_fs_info_args *fi_args;
2812 	struct btrfs_device *device;
2813 	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2814 	u64 flags_in;
2815 	int ret = 0;
2816 
2817 	fi_args = memdup_user(arg, sizeof(*fi_args));
2818 	if (IS_ERR(fi_args))
2819 		return PTR_ERR(fi_args);
2820 
2821 	flags_in = fi_args->flags;
2822 	memset(fi_args, 0, sizeof(*fi_args));
2823 
2824 	rcu_read_lock();
2825 	fi_args->num_devices = fs_devices->num_devices;
2826 
2827 	list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2828 		if (device->devid > fi_args->max_id)
2829 			fi_args->max_id = device->devid;
2830 	}
2831 	rcu_read_unlock();
2832 
2833 	memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2834 	fi_args->nodesize = fs_info->nodesize;
2835 	fi_args->sectorsize = fs_info->sectorsize;
2836 	fi_args->clone_alignment = fs_info->sectorsize;
2837 
2838 	if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2839 		fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2840 		fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2841 		fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2842 	}
2843 
2844 	if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2845 		fi_args->generation = btrfs_get_fs_generation(fs_info);
2846 		fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2847 	}
2848 
2849 	if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2850 		memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2851 		       sizeof(fi_args->metadata_uuid));
2852 		fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2853 	}
2854 
2855 	if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2856 		ret = -EFAULT;
2857 
2858 	kfree(fi_args);
2859 	return ret;
2860 }
2861 
2862 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2863 				 void __user *arg)
2864 {
2865 	BTRFS_DEV_LOOKUP_ARGS(args);
2866 	struct btrfs_ioctl_dev_info_args *di_args;
2867 	struct btrfs_device *dev;
2868 	int ret = 0;
2869 
2870 	di_args = memdup_user(arg, sizeof(*di_args));
2871 	if (IS_ERR(di_args))
2872 		return PTR_ERR(di_args);
2873 
2874 	args.devid = di_args->devid;
2875 	if (!btrfs_is_empty_uuid(di_args->uuid))
2876 		args.uuid = di_args->uuid;
2877 
2878 	rcu_read_lock();
2879 	dev = btrfs_find_device(fs_info->fs_devices, &args);
2880 	if (!dev) {
2881 		ret = -ENODEV;
2882 		goto out;
2883 	}
2884 
2885 	di_args->devid = dev->devid;
2886 	di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2887 	di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2888 	memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2889 	memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2890 	if (dev->name)
2891 		strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2892 	else
2893 		di_args->path[0] = '\0';
2894 
2895 out:
2896 	rcu_read_unlock();
2897 	if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2898 		ret = -EFAULT;
2899 
2900 	kfree(di_args);
2901 	return ret;
2902 }
2903 
2904 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2905 {
2906 	struct inode *inode = file_inode(file);
2907 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2908 	struct btrfs_root *root = BTRFS_I(inode)->root;
2909 	struct btrfs_root *new_root;
2910 	struct btrfs_dir_item *di;
2911 	struct btrfs_trans_handle *trans;
2912 	struct btrfs_path *path = NULL;
2913 	struct btrfs_disk_key disk_key;
2914 	struct fscrypt_str name = FSTR_INIT("default", 7);
2915 	u64 objectid = 0;
2916 	u64 dir_id;
2917 	int ret;
2918 
2919 	if (!capable(CAP_SYS_ADMIN))
2920 		return -EPERM;
2921 
2922 	ret = mnt_want_write_file(file);
2923 	if (ret)
2924 		return ret;
2925 
2926 	if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2927 		ret = -EFAULT;
2928 		goto out;
2929 	}
2930 
2931 	if (!objectid)
2932 		objectid = BTRFS_FS_TREE_OBJECTID;
2933 
2934 	new_root = btrfs_get_fs_root(fs_info, objectid, true);
2935 	if (IS_ERR(new_root)) {
2936 		ret = PTR_ERR(new_root);
2937 		goto out;
2938 	}
2939 	if (!is_fstree(new_root->root_key.objectid)) {
2940 		ret = -ENOENT;
2941 		goto out_free;
2942 	}
2943 
2944 	path = btrfs_alloc_path();
2945 	if (!path) {
2946 		ret = -ENOMEM;
2947 		goto out_free;
2948 	}
2949 
2950 	trans = btrfs_start_transaction(root, 1);
2951 	if (IS_ERR(trans)) {
2952 		ret = PTR_ERR(trans);
2953 		goto out_free;
2954 	}
2955 
2956 	dir_id = btrfs_super_root_dir(fs_info->super_copy);
2957 	di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2958 				   dir_id, &name, 1);
2959 	if (IS_ERR_OR_NULL(di)) {
2960 		btrfs_release_path(path);
2961 		btrfs_end_transaction(trans);
2962 		btrfs_err(fs_info,
2963 			  "Umm, you don't have the default diritem, this isn't going to work");
2964 		ret = -ENOENT;
2965 		goto out_free;
2966 	}
2967 
2968 	btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2969 	btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2970 	btrfs_mark_buffer_dirty(trans, path->nodes[0]);
2971 	btrfs_release_path(path);
2972 
2973 	btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2974 	btrfs_end_transaction(trans);
2975 out_free:
2976 	btrfs_put_root(new_root);
2977 	btrfs_free_path(path);
2978 out:
2979 	mnt_drop_write_file(file);
2980 	return ret;
2981 }
2982 
2983 static void get_block_group_info(struct list_head *groups_list,
2984 				 struct btrfs_ioctl_space_info *space)
2985 {
2986 	struct btrfs_block_group *block_group;
2987 
2988 	space->total_bytes = 0;
2989 	space->used_bytes = 0;
2990 	space->flags = 0;
2991 	list_for_each_entry(block_group, groups_list, list) {
2992 		space->flags = block_group->flags;
2993 		space->total_bytes += block_group->length;
2994 		space->used_bytes += block_group->used;
2995 	}
2996 }
2997 
2998 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
2999 				   void __user *arg)
3000 {
3001 	struct btrfs_ioctl_space_args space_args = { 0 };
3002 	struct btrfs_ioctl_space_info space;
3003 	struct btrfs_ioctl_space_info *dest;
3004 	struct btrfs_ioctl_space_info *dest_orig;
3005 	struct btrfs_ioctl_space_info __user *user_dest;
3006 	struct btrfs_space_info *info;
3007 	static const u64 types[] = {
3008 		BTRFS_BLOCK_GROUP_DATA,
3009 		BTRFS_BLOCK_GROUP_SYSTEM,
3010 		BTRFS_BLOCK_GROUP_METADATA,
3011 		BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3012 	};
3013 	int num_types = 4;
3014 	int alloc_size;
3015 	int ret = 0;
3016 	u64 slot_count = 0;
3017 	int i, c;
3018 
3019 	if (copy_from_user(&space_args,
3020 			   (struct btrfs_ioctl_space_args __user *)arg,
3021 			   sizeof(space_args)))
3022 		return -EFAULT;
3023 
3024 	for (i = 0; i < num_types; i++) {
3025 		struct btrfs_space_info *tmp;
3026 
3027 		info = NULL;
3028 		list_for_each_entry(tmp, &fs_info->space_info, list) {
3029 			if (tmp->flags == types[i]) {
3030 				info = tmp;
3031 				break;
3032 			}
3033 		}
3034 
3035 		if (!info)
3036 			continue;
3037 
3038 		down_read(&info->groups_sem);
3039 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3040 			if (!list_empty(&info->block_groups[c]))
3041 				slot_count++;
3042 		}
3043 		up_read(&info->groups_sem);
3044 	}
3045 
3046 	/*
3047 	 * Global block reserve, exported as a space_info
3048 	 */
3049 	slot_count++;
3050 
3051 	/* space_slots == 0 means they are asking for a count */
3052 	if (space_args.space_slots == 0) {
3053 		space_args.total_spaces = slot_count;
3054 		goto out;
3055 	}
3056 
3057 	slot_count = min_t(u64, space_args.space_slots, slot_count);
3058 
3059 	alloc_size = sizeof(*dest) * slot_count;
3060 
3061 	/* we generally have at most 6 or so space infos, one for each raid
3062 	 * level.  So, a whole page should be more than enough for everyone
3063 	 */
3064 	if (alloc_size > PAGE_SIZE)
3065 		return -ENOMEM;
3066 
3067 	space_args.total_spaces = 0;
3068 	dest = kmalloc(alloc_size, GFP_KERNEL);
3069 	if (!dest)
3070 		return -ENOMEM;
3071 	dest_orig = dest;
3072 
3073 	/* now we have a buffer to copy into */
3074 	for (i = 0; i < num_types; i++) {
3075 		struct btrfs_space_info *tmp;
3076 
3077 		if (!slot_count)
3078 			break;
3079 
3080 		info = NULL;
3081 		list_for_each_entry(tmp, &fs_info->space_info, list) {
3082 			if (tmp->flags == types[i]) {
3083 				info = tmp;
3084 				break;
3085 			}
3086 		}
3087 
3088 		if (!info)
3089 			continue;
3090 		down_read(&info->groups_sem);
3091 		for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3092 			if (!list_empty(&info->block_groups[c])) {
3093 				get_block_group_info(&info->block_groups[c],
3094 						     &space);
3095 				memcpy(dest, &space, sizeof(space));
3096 				dest++;
3097 				space_args.total_spaces++;
3098 				slot_count--;
3099 			}
3100 			if (!slot_count)
3101 				break;
3102 		}
3103 		up_read(&info->groups_sem);
3104 	}
3105 
3106 	/*
3107 	 * Add global block reserve
3108 	 */
3109 	if (slot_count) {
3110 		struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3111 
3112 		spin_lock(&block_rsv->lock);
3113 		space.total_bytes = block_rsv->size;
3114 		space.used_bytes = block_rsv->size - block_rsv->reserved;
3115 		spin_unlock(&block_rsv->lock);
3116 		space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3117 		memcpy(dest, &space, sizeof(space));
3118 		space_args.total_spaces++;
3119 	}
3120 
3121 	user_dest = (struct btrfs_ioctl_space_info __user *)
3122 		(arg + sizeof(struct btrfs_ioctl_space_args));
3123 
3124 	if (copy_to_user(user_dest, dest_orig, alloc_size))
3125 		ret = -EFAULT;
3126 
3127 	kfree(dest_orig);
3128 out:
3129 	if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3130 		ret = -EFAULT;
3131 
3132 	return ret;
3133 }
3134 
3135 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3136 					    void __user *argp)
3137 {
3138 	struct btrfs_trans_handle *trans;
3139 	u64 transid;
3140 
3141 	/*
3142 	 * Start orphan cleanup here for the given root in case it hasn't been
3143 	 * started already by other means. Errors are handled in the other
3144 	 * functions during transaction commit.
3145 	 */
3146 	btrfs_orphan_cleanup(root);
3147 
3148 	trans = btrfs_attach_transaction_barrier(root);
3149 	if (IS_ERR(trans)) {
3150 		if (PTR_ERR(trans) != -ENOENT)
3151 			return PTR_ERR(trans);
3152 
3153 		/* No running transaction, don't bother */
3154 		transid = btrfs_get_last_trans_committed(root->fs_info);
3155 		goto out;
3156 	}
3157 	transid = trans->transid;
3158 	btrfs_commit_transaction_async(trans);
3159 out:
3160 	if (argp)
3161 		if (copy_to_user(argp, &transid, sizeof(transid)))
3162 			return -EFAULT;
3163 	return 0;
3164 }
3165 
3166 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3167 					   void __user *argp)
3168 {
3169 	/* By default wait for the current transaction. */
3170 	u64 transid = 0;
3171 
3172 	if (argp)
3173 		if (copy_from_user(&transid, argp, sizeof(transid)))
3174 			return -EFAULT;
3175 
3176 	return btrfs_wait_for_commit(fs_info, transid);
3177 }
3178 
3179 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3180 {
3181 	struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3182 	struct btrfs_ioctl_scrub_args *sa;
3183 	int ret;
3184 
3185 	if (!capable(CAP_SYS_ADMIN))
3186 		return -EPERM;
3187 
3188 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3189 		btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3190 		return -EINVAL;
3191 	}
3192 
3193 	sa = memdup_user(arg, sizeof(*sa));
3194 	if (IS_ERR(sa))
3195 		return PTR_ERR(sa);
3196 
3197 	if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3198 		ret = -EOPNOTSUPP;
3199 		goto out;
3200 	}
3201 
3202 	if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3203 		ret = mnt_want_write_file(file);
3204 		if (ret)
3205 			goto out;
3206 	}
3207 
3208 	ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3209 			      &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3210 			      0);
3211 
3212 	/*
3213 	 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3214 	 * error. This is important as it allows user space to know how much
3215 	 * progress scrub has done. For example, if scrub is canceled we get
3216 	 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3217 	 * space. Later user space can inspect the progress from the structure
3218 	 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3219 	 * previously (btrfs-progs does this).
3220 	 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3221 	 * then return -EFAULT to signal the structure was not copied or it may
3222 	 * be corrupt and unreliable due to a partial copy.
3223 	 */
3224 	if (copy_to_user(arg, sa, sizeof(*sa)))
3225 		ret = -EFAULT;
3226 
3227 	if (!(sa->flags & BTRFS_SCRUB_READONLY))
3228 		mnt_drop_write_file(file);
3229 out:
3230 	kfree(sa);
3231 	return ret;
3232 }
3233 
3234 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3235 {
3236 	if (!capable(CAP_SYS_ADMIN))
3237 		return -EPERM;
3238 
3239 	return btrfs_scrub_cancel(fs_info);
3240 }
3241 
3242 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3243 				       void __user *arg)
3244 {
3245 	struct btrfs_ioctl_scrub_args *sa;
3246 	int ret;
3247 
3248 	if (!capable(CAP_SYS_ADMIN))
3249 		return -EPERM;
3250 
3251 	sa = memdup_user(arg, sizeof(*sa));
3252 	if (IS_ERR(sa))
3253 		return PTR_ERR(sa);
3254 
3255 	ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3256 
3257 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3258 		ret = -EFAULT;
3259 
3260 	kfree(sa);
3261 	return ret;
3262 }
3263 
3264 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3265 				      void __user *arg)
3266 {
3267 	struct btrfs_ioctl_get_dev_stats *sa;
3268 	int ret;
3269 
3270 	sa = memdup_user(arg, sizeof(*sa));
3271 	if (IS_ERR(sa))
3272 		return PTR_ERR(sa);
3273 
3274 	if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3275 		kfree(sa);
3276 		return -EPERM;
3277 	}
3278 
3279 	ret = btrfs_get_dev_stats(fs_info, sa);
3280 
3281 	if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3282 		ret = -EFAULT;
3283 
3284 	kfree(sa);
3285 	return ret;
3286 }
3287 
3288 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3289 				    void __user *arg)
3290 {
3291 	struct btrfs_ioctl_dev_replace_args *p;
3292 	int ret;
3293 
3294 	if (!capable(CAP_SYS_ADMIN))
3295 		return -EPERM;
3296 
3297 	if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3298 		btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3299 		return -EINVAL;
3300 	}
3301 
3302 	p = memdup_user(arg, sizeof(*p));
3303 	if (IS_ERR(p))
3304 		return PTR_ERR(p);
3305 
3306 	switch (p->cmd) {
3307 	case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3308 		if (sb_rdonly(fs_info->sb)) {
3309 			ret = -EROFS;
3310 			goto out;
3311 		}
3312 		if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3313 			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3314 		} else {
3315 			ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3316 			btrfs_exclop_finish(fs_info);
3317 		}
3318 		break;
3319 	case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3320 		btrfs_dev_replace_status(fs_info, p);
3321 		ret = 0;
3322 		break;
3323 	case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3324 		p->result = btrfs_dev_replace_cancel(fs_info);
3325 		ret = 0;
3326 		break;
3327 	default:
3328 		ret = -EINVAL;
3329 		break;
3330 	}
3331 
3332 	if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3333 		ret = -EFAULT;
3334 out:
3335 	kfree(p);
3336 	return ret;
3337 }
3338 
3339 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3340 {
3341 	int ret = 0;
3342 	int i;
3343 	u64 rel_ptr;
3344 	int size;
3345 	struct btrfs_ioctl_ino_path_args *ipa = NULL;
3346 	struct inode_fs_paths *ipath = NULL;
3347 	struct btrfs_path *path;
3348 
3349 	if (!capable(CAP_DAC_READ_SEARCH))
3350 		return -EPERM;
3351 
3352 	path = btrfs_alloc_path();
3353 	if (!path) {
3354 		ret = -ENOMEM;
3355 		goto out;
3356 	}
3357 
3358 	ipa = memdup_user(arg, sizeof(*ipa));
3359 	if (IS_ERR(ipa)) {
3360 		ret = PTR_ERR(ipa);
3361 		ipa = NULL;
3362 		goto out;
3363 	}
3364 
3365 	size = min_t(u32, ipa->size, 4096);
3366 	ipath = init_ipath(size, root, path);
3367 	if (IS_ERR(ipath)) {
3368 		ret = PTR_ERR(ipath);
3369 		ipath = NULL;
3370 		goto out;
3371 	}
3372 
3373 	ret = paths_from_inode(ipa->inum, ipath);
3374 	if (ret < 0)
3375 		goto out;
3376 
3377 	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3378 		rel_ptr = ipath->fspath->val[i] -
3379 			  (u64)(unsigned long)ipath->fspath->val;
3380 		ipath->fspath->val[i] = rel_ptr;
3381 	}
3382 
3383 	btrfs_free_path(path);
3384 	path = NULL;
3385 	ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3386 			   ipath->fspath, size);
3387 	if (ret) {
3388 		ret = -EFAULT;
3389 		goto out;
3390 	}
3391 
3392 out:
3393 	btrfs_free_path(path);
3394 	free_ipath(ipath);
3395 	kfree(ipa);
3396 
3397 	return ret;
3398 }
3399 
3400 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3401 					void __user *arg, int version)
3402 {
3403 	int ret = 0;
3404 	int size;
3405 	struct btrfs_ioctl_logical_ino_args *loi;
3406 	struct btrfs_data_container *inodes = NULL;
3407 	struct btrfs_path *path = NULL;
3408 	bool ignore_offset;
3409 
3410 	if (!capable(CAP_SYS_ADMIN))
3411 		return -EPERM;
3412 
3413 	loi = memdup_user(arg, sizeof(*loi));
3414 	if (IS_ERR(loi))
3415 		return PTR_ERR(loi);
3416 
3417 	if (version == 1) {
3418 		ignore_offset = false;
3419 		size = min_t(u32, loi->size, SZ_64K);
3420 	} else {
3421 		/* All reserved bits must be 0 for now */
3422 		if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3423 			ret = -EINVAL;
3424 			goto out_loi;
3425 		}
3426 		/* Only accept flags we have defined so far */
3427 		if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3428 			ret = -EINVAL;
3429 			goto out_loi;
3430 		}
3431 		ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3432 		size = min_t(u32, loi->size, SZ_16M);
3433 	}
3434 
3435 	inodes = init_data_container(size);
3436 	if (IS_ERR(inodes)) {
3437 		ret = PTR_ERR(inodes);
3438 		goto out_loi;
3439 	}
3440 
3441 	path = btrfs_alloc_path();
3442 	if (!path) {
3443 		ret = -ENOMEM;
3444 		goto out;
3445 	}
3446 	ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3447 					  inodes, ignore_offset);
3448 	btrfs_free_path(path);
3449 	if (ret == -EINVAL)
3450 		ret = -ENOENT;
3451 	if (ret < 0)
3452 		goto out;
3453 
3454 	ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3455 			   size);
3456 	if (ret)
3457 		ret = -EFAULT;
3458 
3459 out:
3460 	kvfree(inodes);
3461 out_loi:
3462 	kfree(loi);
3463 
3464 	return ret;
3465 }
3466 
3467 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3468 			       struct btrfs_ioctl_balance_args *bargs)
3469 {
3470 	struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3471 
3472 	bargs->flags = bctl->flags;
3473 
3474 	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3475 		bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3476 	if (atomic_read(&fs_info->balance_pause_req))
3477 		bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3478 	if (atomic_read(&fs_info->balance_cancel_req))
3479 		bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3480 
3481 	memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3482 	memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3483 	memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3484 
3485 	spin_lock(&fs_info->balance_lock);
3486 	memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3487 	spin_unlock(&fs_info->balance_lock);
3488 }
3489 
3490 /*
3491  * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3492  * required.
3493  *
3494  * @fs_info:       the filesystem
3495  * @excl_acquired: ptr to boolean value which is set to false in case balance
3496  *                 is being resumed
3497  *
3498  * Return 0 on success in which case both fs_info::balance is acquired as well
3499  * as exclusive ops are blocked. In case of failure return an error code.
3500  */
3501 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3502 {
3503 	int ret;
3504 
3505 	/*
3506 	 * Exclusive operation is locked. Three possibilities:
3507 	 *   (1) some other op is running
3508 	 *   (2) balance is running
3509 	 *   (3) balance is paused -- special case (think resume)
3510 	 */
3511 	while (1) {
3512 		if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3513 			*excl_acquired = true;
3514 			mutex_lock(&fs_info->balance_mutex);
3515 			return 0;
3516 		}
3517 
3518 		mutex_lock(&fs_info->balance_mutex);
3519 		if (fs_info->balance_ctl) {
3520 			/* This is either (2) or (3) */
3521 			if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3522 				/* This is (2) */
3523 				ret = -EINPROGRESS;
3524 				goto out_failure;
3525 
3526 			} else {
3527 				mutex_unlock(&fs_info->balance_mutex);
3528 				/*
3529 				 * Lock released to allow other waiters to
3530 				 * continue, we'll reexamine the status again.
3531 				 */
3532 				mutex_lock(&fs_info->balance_mutex);
3533 
3534 				if (fs_info->balance_ctl &&
3535 				    !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3536 					/* This is (3) */
3537 					*excl_acquired = false;
3538 					return 0;
3539 				}
3540 			}
3541 		} else {
3542 			/* This is (1) */
3543 			ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3544 			goto out_failure;
3545 		}
3546 
3547 		mutex_unlock(&fs_info->balance_mutex);
3548 	}
3549 
3550 out_failure:
3551 	mutex_unlock(&fs_info->balance_mutex);
3552 	*excl_acquired = false;
3553 	return ret;
3554 }
3555 
3556 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3557 {
3558 	struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3559 	struct btrfs_fs_info *fs_info = root->fs_info;
3560 	struct btrfs_ioctl_balance_args *bargs;
3561 	struct btrfs_balance_control *bctl;
3562 	bool need_unlock = true;
3563 	int ret;
3564 
3565 	if (!capable(CAP_SYS_ADMIN))
3566 		return -EPERM;
3567 
3568 	ret = mnt_want_write_file(file);
3569 	if (ret)
3570 		return ret;
3571 
3572 	bargs = memdup_user(arg, sizeof(*bargs));
3573 	if (IS_ERR(bargs)) {
3574 		ret = PTR_ERR(bargs);
3575 		bargs = NULL;
3576 		goto out;
3577 	}
3578 
3579 	ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3580 	if (ret)
3581 		goto out;
3582 
3583 	lockdep_assert_held(&fs_info->balance_mutex);
3584 
3585 	if (bargs->flags & BTRFS_BALANCE_RESUME) {
3586 		if (!fs_info->balance_ctl) {
3587 			ret = -ENOTCONN;
3588 			goto out_unlock;
3589 		}
3590 
3591 		bctl = fs_info->balance_ctl;
3592 		spin_lock(&fs_info->balance_lock);
3593 		bctl->flags |= BTRFS_BALANCE_RESUME;
3594 		spin_unlock(&fs_info->balance_lock);
3595 		btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3596 
3597 		goto do_balance;
3598 	}
3599 
3600 	if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3601 		ret = -EINVAL;
3602 		goto out_unlock;
3603 	}
3604 
3605 	if (fs_info->balance_ctl) {
3606 		ret = -EINPROGRESS;
3607 		goto out_unlock;
3608 	}
3609 
3610 	bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3611 	if (!bctl) {
3612 		ret = -ENOMEM;
3613 		goto out_unlock;
3614 	}
3615 
3616 	memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3617 	memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3618 	memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3619 
3620 	bctl->flags = bargs->flags;
3621 do_balance:
3622 	/*
3623 	 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3624 	 * bctl is freed in reset_balance_state, or, if restriper was paused
3625 	 * all the way until unmount, in free_fs_info.  The flag should be
3626 	 * cleared after reset_balance_state.
3627 	 */
3628 	need_unlock = false;
3629 
3630 	ret = btrfs_balance(fs_info, bctl, bargs);
3631 	bctl = NULL;
3632 
3633 	if (ret == 0 || ret == -ECANCELED) {
3634 		if (copy_to_user(arg, bargs, sizeof(*bargs)))
3635 			ret = -EFAULT;
3636 	}
3637 
3638 	kfree(bctl);
3639 out_unlock:
3640 	mutex_unlock(&fs_info->balance_mutex);
3641 	if (need_unlock)
3642 		btrfs_exclop_finish(fs_info);
3643 out:
3644 	mnt_drop_write_file(file);
3645 	kfree(bargs);
3646 	return ret;
3647 }
3648 
3649 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3650 {
3651 	if (!capable(CAP_SYS_ADMIN))
3652 		return -EPERM;
3653 
3654 	switch (cmd) {
3655 	case BTRFS_BALANCE_CTL_PAUSE:
3656 		return btrfs_pause_balance(fs_info);
3657 	case BTRFS_BALANCE_CTL_CANCEL:
3658 		return btrfs_cancel_balance(fs_info);
3659 	}
3660 
3661 	return -EINVAL;
3662 }
3663 
3664 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3665 					 void __user *arg)
3666 {
3667 	struct btrfs_ioctl_balance_args *bargs;
3668 	int ret = 0;
3669 
3670 	if (!capable(CAP_SYS_ADMIN))
3671 		return -EPERM;
3672 
3673 	mutex_lock(&fs_info->balance_mutex);
3674 	if (!fs_info->balance_ctl) {
3675 		ret = -ENOTCONN;
3676 		goto out;
3677 	}
3678 
3679 	bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3680 	if (!bargs) {
3681 		ret = -ENOMEM;
3682 		goto out;
3683 	}
3684 
3685 	btrfs_update_ioctl_balance_args(fs_info, bargs);
3686 
3687 	if (copy_to_user(arg, bargs, sizeof(*bargs)))
3688 		ret = -EFAULT;
3689 
3690 	kfree(bargs);
3691 out:
3692 	mutex_unlock(&fs_info->balance_mutex);
3693 	return ret;
3694 }
3695 
3696 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3697 {
3698 	struct inode *inode = file_inode(file);
3699 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3700 	struct btrfs_ioctl_quota_ctl_args *sa;
3701 	int ret;
3702 
3703 	if (!capable(CAP_SYS_ADMIN))
3704 		return -EPERM;
3705 
3706 	ret = mnt_want_write_file(file);
3707 	if (ret)
3708 		return ret;
3709 
3710 	sa = memdup_user(arg, sizeof(*sa));
3711 	if (IS_ERR(sa)) {
3712 		ret = PTR_ERR(sa);
3713 		goto drop_write;
3714 	}
3715 
3716 	down_write(&fs_info->subvol_sem);
3717 
3718 	switch (sa->cmd) {
3719 	case BTRFS_QUOTA_CTL_ENABLE:
3720 	case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3721 		ret = btrfs_quota_enable(fs_info, sa);
3722 		break;
3723 	case BTRFS_QUOTA_CTL_DISABLE:
3724 		ret = btrfs_quota_disable(fs_info);
3725 		break;
3726 	default:
3727 		ret = -EINVAL;
3728 		break;
3729 	}
3730 
3731 	kfree(sa);
3732 	up_write(&fs_info->subvol_sem);
3733 drop_write:
3734 	mnt_drop_write_file(file);
3735 	return ret;
3736 }
3737 
3738 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3739 {
3740 	struct inode *inode = file_inode(file);
3741 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3742 	struct btrfs_root *root = BTRFS_I(inode)->root;
3743 	struct btrfs_ioctl_qgroup_assign_args *sa;
3744 	struct btrfs_trans_handle *trans;
3745 	int ret;
3746 	int err;
3747 
3748 	if (!capable(CAP_SYS_ADMIN))
3749 		return -EPERM;
3750 
3751 	ret = mnt_want_write_file(file);
3752 	if (ret)
3753 		return ret;
3754 
3755 	sa = memdup_user(arg, sizeof(*sa));
3756 	if (IS_ERR(sa)) {
3757 		ret = PTR_ERR(sa);
3758 		goto drop_write;
3759 	}
3760 
3761 	trans = btrfs_join_transaction(root);
3762 	if (IS_ERR(trans)) {
3763 		ret = PTR_ERR(trans);
3764 		goto out;
3765 	}
3766 
3767 	if (sa->assign) {
3768 		ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3769 	} else {
3770 		ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3771 	}
3772 
3773 	/* update qgroup status and info */
3774 	mutex_lock(&fs_info->qgroup_ioctl_lock);
3775 	err = btrfs_run_qgroups(trans);
3776 	mutex_unlock(&fs_info->qgroup_ioctl_lock);
3777 	if (err < 0)
3778 		btrfs_handle_fs_error(fs_info, err,
3779 				      "failed to update qgroup status and info");
3780 	err = btrfs_end_transaction(trans);
3781 	if (err && !ret)
3782 		ret = err;
3783 
3784 out:
3785 	kfree(sa);
3786 drop_write:
3787 	mnt_drop_write_file(file);
3788 	return ret;
3789 }
3790 
3791 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3792 {
3793 	struct inode *inode = file_inode(file);
3794 	struct btrfs_root *root = BTRFS_I(inode)->root;
3795 	struct btrfs_ioctl_qgroup_create_args *sa;
3796 	struct btrfs_trans_handle *trans;
3797 	int ret;
3798 	int err;
3799 
3800 	if (!capable(CAP_SYS_ADMIN))
3801 		return -EPERM;
3802 
3803 	ret = mnt_want_write_file(file);
3804 	if (ret)
3805 		return ret;
3806 
3807 	sa = memdup_user(arg, sizeof(*sa));
3808 	if (IS_ERR(sa)) {
3809 		ret = PTR_ERR(sa);
3810 		goto drop_write;
3811 	}
3812 
3813 	if (!sa->qgroupid) {
3814 		ret = -EINVAL;
3815 		goto out;
3816 	}
3817 
3818 	if (sa->create && is_fstree(sa->qgroupid)) {
3819 		ret = -EINVAL;
3820 		goto out;
3821 	}
3822 
3823 	trans = btrfs_join_transaction(root);
3824 	if (IS_ERR(trans)) {
3825 		ret = PTR_ERR(trans);
3826 		goto out;
3827 	}
3828 
3829 	if (sa->create) {
3830 		ret = btrfs_create_qgroup(trans, sa->qgroupid);
3831 	} else {
3832 		ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3833 	}
3834 
3835 	err = btrfs_end_transaction(trans);
3836 	if (err && !ret)
3837 		ret = err;
3838 
3839 out:
3840 	kfree(sa);
3841 drop_write:
3842 	mnt_drop_write_file(file);
3843 	return ret;
3844 }
3845 
3846 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3847 {
3848 	struct inode *inode = file_inode(file);
3849 	struct btrfs_root *root = BTRFS_I(inode)->root;
3850 	struct btrfs_ioctl_qgroup_limit_args *sa;
3851 	struct btrfs_trans_handle *trans;
3852 	int ret;
3853 	int err;
3854 	u64 qgroupid;
3855 
3856 	if (!capable(CAP_SYS_ADMIN))
3857 		return -EPERM;
3858 
3859 	ret = mnt_want_write_file(file);
3860 	if (ret)
3861 		return ret;
3862 
3863 	sa = memdup_user(arg, sizeof(*sa));
3864 	if (IS_ERR(sa)) {
3865 		ret = PTR_ERR(sa);
3866 		goto drop_write;
3867 	}
3868 
3869 	trans = btrfs_join_transaction(root);
3870 	if (IS_ERR(trans)) {
3871 		ret = PTR_ERR(trans);
3872 		goto out;
3873 	}
3874 
3875 	qgroupid = sa->qgroupid;
3876 	if (!qgroupid) {
3877 		/* take the current subvol as qgroup */
3878 		qgroupid = root->root_key.objectid;
3879 	}
3880 
3881 	ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3882 
3883 	err = btrfs_end_transaction(trans);
3884 	if (err && !ret)
3885 		ret = err;
3886 
3887 out:
3888 	kfree(sa);
3889 drop_write:
3890 	mnt_drop_write_file(file);
3891 	return ret;
3892 }
3893 
3894 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3895 {
3896 	struct inode *inode = file_inode(file);
3897 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3898 	struct btrfs_ioctl_quota_rescan_args *qsa;
3899 	int ret;
3900 
3901 	if (!capable(CAP_SYS_ADMIN))
3902 		return -EPERM;
3903 
3904 	ret = mnt_want_write_file(file);
3905 	if (ret)
3906 		return ret;
3907 
3908 	qsa = memdup_user(arg, sizeof(*qsa));
3909 	if (IS_ERR(qsa)) {
3910 		ret = PTR_ERR(qsa);
3911 		goto drop_write;
3912 	}
3913 
3914 	if (qsa->flags) {
3915 		ret = -EINVAL;
3916 		goto out;
3917 	}
3918 
3919 	ret = btrfs_qgroup_rescan(fs_info);
3920 
3921 out:
3922 	kfree(qsa);
3923 drop_write:
3924 	mnt_drop_write_file(file);
3925 	return ret;
3926 }
3927 
3928 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3929 						void __user *arg)
3930 {
3931 	struct btrfs_ioctl_quota_rescan_args qsa = {0};
3932 
3933 	if (!capable(CAP_SYS_ADMIN))
3934 		return -EPERM;
3935 
3936 	if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3937 		qsa.flags = 1;
3938 		qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3939 	}
3940 
3941 	if (copy_to_user(arg, &qsa, sizeof(qsa)))
3942 		return -EFAULT;
3943 
3944 	return 0;
3945 }
3946 
3947 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3948 						void __user *arg)
3949 {
3950 	if (!capable(CAP_SYS_ADMIN))
3951 		return -EPERM;
3952 
3953 	return btrfs_qgroup_wait_for_completion(fs_info, true);
3954 }
3955 
3956 static long _btrfs_ioctl_set_received_subvol(struct file *file,
3957 					    struct mnt_idmap *idmap,
3958 					    struct btrfs_ioctl_received_subvol_args *sa)
3959 {
3960 	struct inode *inode = file_inode(file);
3961 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3962 	struct btrfs_root *root = BTRFS_I(inode)->root;
3963 	struct btrfs_root_item *root_item = &root->root_item;
3964 	struct btrfs_trans_handle *trans;
3965 	struct timespec64 ct = current_time(inode);
3966 	int ret = 0;
3967 	int received_uuid_changed;
3968 
3969 	if (!inode_owner_or_capable(idmap, inode))
3970 		return -EPERM;
3971 
3972 	ret = mnt_want_write_file(file);
3973 	if (ret < 0)
3974 		return ret;
3975 
3976 	down_write(&fs_info->subvol_sem);
3977 
3978 	if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3979 		ret = -EINVAL;
3980 		goto out;
3981 	}
3982 
3983 	if (btrfs_root_readonly(root)) {
3984 		ret = -EROFS;
3985 		goto out;
3986 	}
3987 
3988 	/*
3989 	 * 1 - root item
3990 	 * 2 - uuid items (received uuid + subvol uuid)
3991 	 */
3992 	trans = btrfs_start_transaction(root, 3);
3993 	if (IS_ERR(trans)) {
3994 		ret = PTR_ERR(trans);
3995 		trans = NULL;
3996 		goto out;
3997 	}
3998 
3999 	sa->rtransid = trans->transid;
4000 	sa->rtime.sec = ct.tv_sec;
4001 	sa->rtime.nsec = ct.tv_nsec;
4002 
4003 	received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4004 				       BTRFS_UUID_SIZE);
4005 	if (received_uuid_changed &&
4006 	    !btrfs_is_empty_uuid(root_item->received_uuid)) {
4007 		ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4008 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4009 					  root->root_key.objectid);
4010 		if (ret && ret != -ENOENT) {
4011 		        btrfs_abort_transaction(trans, ret);
4012 		        btrfs_end_transaction(trans);
4013 		        goto out;
4014 		}
4015 	}
4016 	memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4017 	btrfs_set_root_stransid(root_item, sa->stransid);
4018 	btrfs_set_root_rtransid(root_item, sa->rtransid);
4019 	btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4020 	btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4021 	btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4022 	btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4023 
4024 	ret = btrfs_update_root(trans, fs_info->tree_root,
4025 				&root->root_key, &root->root_item);
4026 	if (ret < 0) {
4027 		btrfs_end_transaction(trans);
4028 		goto out;
4029 	}
4030 	if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4031 		ret = btrfs_uuid_tree_add(trans, sa->uuid,
4032 					  BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4033 					  root->root_key.objectid);
4034 		if (ret < 0 && ret != -EEXIST) {
4035 			btrfs_abort_transaction(trans, ret);
4036 			btrfs_end_transaction(trans);
4037 			goto out;
4038 		}
4039 	}
4040 	ret = btrfs_commit_transaction(trans);
4041 out:
4042 	up_write(&fs_info->subvol_sem);
4043 	mnt_drop_write_file(file);
4044 	return ret;
4045 }
4046 
4047 #ifdef CONFIG_64BIT
4048 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4049 						void __user *arg)
4050 {
4051 	struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4052 	struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4053 	int ret = 0;
4054 
4055 	args32 = memdup_user(arg, sizeof(*args32));
4056 	if (IS_ERR(args32))
4057 		return PTR_ERR(args32);
4058 
4059 	args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4060 	if (!args64) {
4061 		ret = -ENOMEM;
4062 		goto out;
4063 	}
4064 
4065 	memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4066 	args64->stransid = args32->stransid;
4067 	args64->rtransid = args32->rtransid;
4068 	args64->stime.sec = args32->stime.sec;
4069 	args64->stime.nsec = args32->stime.nsec;
4070 	args64->rtime.sec = args32->rtime.sec;
4071 	args64->rtime.nsec = args32->rtime.nsec;
4072 	args64->flags = args32->flags;
4073 
4074 	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4075 	if (ret)
4076 		goto out;
4077 
4078 	memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4079 	args32->stransid = args64->stransid;
4080 	args32->rtransid = args64->rtransid;
4081 	args32->stime.sec = args64->stime.sec;
4082 	args32->stime.nsec = args64->stime.nsec;
4083 	args32->rtime.sec = args64->rtime.sec;
4084 	args32->rtime.nsec = args64->rtime.nsec;
4085 	args32->flags = args64->flags;
4086 
4087 	ret = copy_to_user(arg, args32, sizeof(*args32));
4088 	if (ret)
4089 		ret = -EFAULT;
4090 
4091 out:
4092 	kfree(args32);
4093 	kfree(args64);
4094 	return ret;
4095 }
4096 #endif
4097 
4098 static long btrfs_ioctl_set_received_subvol(struct file *file,
4099 					    void __user *arg)
4100 {
4101 	struct btrfs_ioctl_received_subvol_args *sa = NULL;
4102 	int ret = 0;
4103 
4104 	sa = memdup_user(arg, sizeof(*sa));
4105 	if (IS_ERR(sa))
4106 		return PTR_ERR(sa);
4107 
4108 	ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4109 
4110 	if (ret)
4111 		goto out;
4112 
4113 	ret = copy_to_user(arg, sa, sizeof(*sa));
4114 	if (ret)
4115 		ret = -EFAULT;
4116 
4117 out:
4118 	kfree(sa);
4119 	return ret;
4120 }
4121 
4122 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4123 					void __user *arg)
4124 {
4125 	size_t len;
4126 	int ret;
4127 	char label[BTRFS_LABEL_SIZE];
4128 
4129 	spin_lock(&fs_info->super_lock);
4130 	memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4131 	spin_unlock(&fs_info->super_lock);
4132 
4133 	len = strnlen(label, BTRFS_LABEL_SIZE);
4134 
4135 	if (len == BTRFS_LABEL_SIZE) {
4136 		btrfs_warn(fs_info,
4137 			   "label is too long, return the first %zu bytes",
4138 			   --len);
4139 	}
4140 
4141 	ret = copy_to_user(arg, label, len);
4142 
4143 	return ret ? -EFAULT : 0;
4144 }
4145 
4146 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4147 {
4148 	struct inode *inode = file_inode(file);
4149 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4150 	struct btrfs_root *root = BTRFS_I(inode)->root;
4151 	struct btrfs_super_block *super_block = fs_info->super_copy;
4152 	struct btrfs_trans_handle *trans;
4153 	char label[BTRFS_LABEL_SIZE];
4154 	int ret;
4155 
4156 	if (!capable(CAP_SYS_ADMIN))
4157 		return -EPERM;
4158 
4159 	if (copy_from_user(label, arg, sizeof(label)))
4160 		return -EFAULT;
4161 
4162 	if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4163 		btrfs_err(fs_info,
4164 			  "unable to set label with more than %d bytes",
4165 			  BTRFS_LABEL_SIZE - 1);
4166 		return -EINVAL;
4167 	}
4168 
4169 	ret = mnt_want_write_file(file);
4170 	if (ret)
4171 		return ret;
4172 
4173 	trans = btrfs_start_transaction(root, 0);
4174 	if (IS_ERR(trans)) {
4175 		ret = PTR_ERR(trans);
4176 		goto out_unlock;
4177 	}
4178 
4179 	spin_lock(&fs_info->super_lock);
4180 	strcpy(super_block->label, label);
4181 	spin_unlock(&fs_info->super_lock);
4182 	ret = btrfs_commit_transaction(trans);
4183 
4184 out_unlock:
4185 	mnt_drop_write_file(file);
4186 	return ret;
4187 }
4188 
4189 #define INIT_FEATURE_FLAGS(suffix) \
4190 	{ .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4191 	  .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4192 	  .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4193 
4194 int btrfs_ioctl_get_supported_features(void __user *arg)
4195 {
4196 	static const struct btrfs_ioctl_feature_flags features[3] = {
4197 		INIT_FEATURE_FLAGS(SUPP),
4198 		INIT_FEATURE_FLAGS(SAFE_SET),
4199 		INIT_FEATURE_FLAGS(SAFE_CLEAR)
4200 	};
4201 
4202 	if (copy_to_user(arg, &features, sizeof(features)))
4203 		return -EFAULT;
4204 
4205 	return 0;
4206 }
4207 
4208 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4209 					void __user *arg)
4210 {
4211 	struct btrfs_super_block *super_block = fs_info->super_copy;
4212 	struct btrfs_ioctl_feature_flags features;
4213 
4214 	features.compat_flags = btrfs_super_compat_flags(super_block);
4215 	features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4216 	features.incompat_flags = btrfs_super_incompat_flags(super_block);
4217 
4218 	if (copy_to_user(arg, &features, sizeof(features)))
4219 		return -EFAULT;
4220 
4221 	return 0;
4222 }
4223 
4224 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4225 			      enum btrfs_feature_set set,
4226 			      u64 change_mask, u64 flags, u64 supported_flags,
4227 			      u64 safe_set, u64 safe_clear)
4228 {
4229 	const char *type = btrfs_feature_set_name(set);
4230 	char *names;
4231 	u64 disallowed, unsupported;
4232 	u64 set_mask = flags & change_mask;
4233 	u64 clear_mask = ~flags & change_mask;
4234 
4235 	unsupported = set_mask & ~supported_flags;
4236 	if (unsupported) {
4237 		names = btrfs_printable_features(set, unsupported);
4238 		if (names) {
4239 			btrfs_warn(fs_info,
4240 				   "this kernel does not support the %s feature bit%s",
4241 				   names, strchr(names, ',') ? "s" : "");
4242 			kfree(names);
4243 		} else
4244 			btrfs_warn(fs_info,
4245 				   "this kernel does not support %s bits 0x%llx",
4246 				   type, unsupported);
4247 		return -EOPNOTSUPP;
4248 	}
4249 
4250 	disallowed = set_mask & ~safe_set;
4251 	if (disallowed) {
4252 		names = btrfs_printable_features(set, disallowed);
4253 		if (names) {
4254 			btrfs_warn(fs_info,
4255 				   "can't set the %s feature bit%s while mounted",
4256 				   names, strchr(names, ',') ? "s" : "");
4257 			kfree(names);
4258 		} else
4259 			btrfs_warn(fs_info,
4260 				   "can't set %s bits 0x%llx while mounted",
4261 				   type, disallowed);
4262 		return -EPERM;
4263 	}
4264 
4265 	disallowed = clear_mask & ~safe_clear;
4266 	if (disallowed) {
4267 		names = btrfs_printable_features(set, disallowed);
4268 		if (names) {
4269 			btrfs_warn(fs_info,
4270 				   "can't clear the %s feature bit%s while mounted",
4271 				   names, strchr(names, ',') ? "s" : "");
4272 			kfree(names);
4273 		} else
4274 			btrfs_warn(fs_info,
4275 				   "can't clear %s bits 0x%llx while mounted",
4276 				   type, disallowed);
4277 		return -EPERM;
4278 	}
4279 
4280 	return 0;
4281 }
4282 
4283 #define check_feature(fs_info, change_mask, flags, mask_base)	\
4284 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags,	\
4285 		   BTRFS_FEATURE_ ## mask_base ## _SUPP,	\
4286 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_SET,	\
4287 		   BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4288 
4289 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4290 {
4291 	struct inode *inode = file_inode(file);
4292 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4293 	struct btrfs_root *root = BTRFS_I(inode)->root;
4294 	struct btrfs_super_block *super_block = fs_info->super_copy;
4295 	struct btrfs_ioctl_feature_flags flags[2];
4296 	struct btrfs_trans_handle *trans;
4297 	u64 newflags;
4298 	int ret;
4299 
4300 	if (!capable(CAP_SYS_ADMIN))
4301 		return -EPERM;
4302 
4303 	if (copy_from_user(flags, arg, sizeof(flags)))
4304 		return -EFAULT;
4305 
4306 	/* Nothing to do */
4307 	if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4308 	    !flags[0].incompat_flags)
4309 		return 0;
4310 
4311 	ret = check_feature(fs_info, flags[0].compat_flags,
4312 			    flags[1].compat_flags, COMPAT);
4313 	if (ret)
4314 		return ret;
4315 
4316 	ret = check_feature(fs_info, flags[0].compat_ro_flags,
4317 			    flags[1].compat_ro_flags, COMPAT_RO);
4318 	if (ret)
4319 		return ret;
4320 
4321 	ret = check_feature(fs_info, flags[0].incompat_flags,
4322 			    flags[1].incompat_flags, INCOMPAT);
4323 	if (ret)
4324 		return ret;
4325 
4326 	ret = mnt_want_write_file(file);
4327 	if (ret)
4328 		return ret;
4329 
4330 	trans = btrfs_start_transaction(root, 0);
4331 	if (IS_ERR(trans)) {
4332 		ret = PTR_ERR(trans);
4333 		goto out_drop_write;
4334 	}
4335 
4336 	spin_lock(&fs_info->super_lock);
4337 	newflags = btrfs_super_compat_flags(super_block);
4338 	newflags |= flags[0].compat_flags & flags[1].compat_flags;
4339 	newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4340 	btrfs_set_super_compat_flags(super_block, newflags);
4341 
4342 	newflags = btrfs_super_compat_ro_flags(super_block);
4343 	newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4344 	newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4345 	btrfs_set_super_compat_ro_flags(super_block, newflags);
4346 
4347 	newflags = btrfs_super_incompat_flags(super_block);
4348 	newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4349 	newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4350 	btrfs_set_super_incompat_flags(super_block, newflags);
4351 	spin_unlock(&fs_info->super_lock);
4352 
4353 	ret = btrfs_commit_transaction(trans);
4354 out_drop_write:
4355 	mnt_drop_write_file(file);
4356 
4357 	return ret;
4358 }
4359 
4360 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4361 {
4362 	struct btrfs_ioctl_send_args *arg;
4363 	int ret;
4364 
4365 	if (compat) {
4366 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4367 		struct btrfs_ioctl_send_args_32 args32 = { 0 };
4368 
4369 		ret = copy_from_user(&args32, argp, sizeof(args32));
4370 		if (ret)
4371 			return -EFAULT;
4372 		arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4373 		if (!arg)
4374 			return -ENOMEM;
4375 		arg->send_fd = args32.send_fd;
4376 		arg->clone_sources_count = args32.clone_sources_count;
4377 		arg->clone_sources = compat_ptr(args32.clone_sources);
4378 		arg->parent_root = args32.parent_root;
4379 		arg->flags = args32.flags;
4380 		arg->version = args32.version;
4381 		memcpy(arg->reserved, args32.reserved,
4382 		       sizeof(args32.reserved));
4383 #else
4384 		return -ENOTTY;
4385 #endif
4386 	} else {
4387 		arg = memdup_user(argp, sizeof(*arg));
4388 		if (IS_ERR(arg))
4389 			return PTR_ERR(arg);
4390 	}
4391 	ret = btrfs_ioctl_send(inode, arg);
4392 	kfree(arg);
4393 	return ret;
4394 }
4395 
4396 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4397 				    bool compat)
4398 {
4399 	struct btrfs_ioctl_encoded_io_args args = { 0 };
4400 	size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4401 					     flags);
4402 	size_t copy_end;
4403 	struct iovec iovstack[UIO_FASTIOV];
4404 	struct iovec *iov = iovstack;
4405 	struct iov_iter iter;
4406 	loff_t pos;
4407 	struct kiocb kiocb;
4408 	ssize_t ret;
4409 
4410 	if (!capable(CAP_SYS_ADMIN)) {
4411 		ret = -EPERM;
4412 		goto out_acct;
4413 	}
4414 
4415 	if (compat) {
4416 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4417 		struct btrfs_ioctl_encoded_io_args_32 args32;
4418 
4419 		copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4420 				       flags);
4421 		if (copy_from_user(&args32, argp, copy_end)) {
4422 			ret = -EFAULT;
4423 			goto out_acct;
4424 		}
4425 		args.iov = compat_ptr(args32.iov);
4426 		args.iovcnt = args32.iovcnt;
4427 		args.offset = args32.offset;
4428 		args.flags = args32.flags;
4429 #else
4430 		return -ENOTTY;
4431 #endif
4432 	} else {
4433 		copy_end = copy_end_kernel;
4434 		if (copy_from_user(&args, argp, copy_end)) {
4435 			ret = -EFAULT;
4436 			goto out_acct;
4437 		}
4438 	}
4439 	if (args.flags != 0) {
4440 		ret = -EINVAL;
4441 		goto out_acct;
4442 	}
4443 
4444 	ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4445 			   &iov, &iter);
4446 	if (ret < 0)
4447 		goto out_acct;
4448 
4449 	if (iov_iter_count(&iter) == 0) {
4450 		ret = 0;
4451 		goto out_iov;
4452 	}
4453 	pos = args.offset;
4454 	ret = rw_verify_area(READ, file, &pos, args.len);
4455 	if (ret < 0)
4456 		goto out_iov;
4457 
4458 	init_sync_kiocb(&kiocb, file);
4459 	kiocb.ki_pos = pos;
4460 
4461 	ret = btrfs_encoded_read(&kiocb, &iter, &args);
4462 	if (ret >= 0) {
4463 		fsnotify_access(file);
4464 		if (copy_to_user(argp + copy_end,
4465 				 (char *)&args + copy_end_kernel,
4466 				 sizeof(args) - copy_end_kernel))
4467 			ret = -EFAULT;
4468 	}
4469 
4470 out_iov:
4471 	kfree(iov);
4472 out_acct:
4473 	if (ret > 0)
4474 		add_rchar(current, ret);
4475 	inc_syscr(current);
4476 	return ret;
4477 }
4478 
4479 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4480 {
4481 	struct btrfs_ioctl_encoded_io_args args;
4482 	struct iovec iovstack[UIO_FASTIOV];
4483 	struct iovec *iov = iovstack;
4484 	struct iov_iter iter;
4485 	loff_t pos;
4486 	struct kiocb kiocb;
4487 	ssize_t ret;
4488 
4489 	if (!capable(CAP_SYS_ADMIN)) {
4490 		ret = -EPERM;
4491 		goto out_acct;
4492 	}
4493 
4494 	if (!(file->f_mode & FMODE_WRITE)) {
4495 		ret = -EBADF;
4496 		goto out_acct;
4497 	}
4498 
4499 	if (compat) {
4500 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4501 		struct btrfs_ioctl_encoded_io_args_32 args32;
4502 
4503 		if (copy_from_user(&args32, argp, sizeof(args32))) {
4504 			ret = -EFAULT;
4505 			goto out_acct;
4506 		}
4507 		args.iov = compat_ptr(args32.iov);
4508 		args.iovcnt = args32.iovcnt;
4509 		args.offset = args32.offset;
4510 		args.flags = args32.flags;
4511 		args.len = args32.len;
4512 		args.unencoded_len = args32.unencoded_len;
4513 		args.unencoded_offset = args32.unencoded_offset;
4514 		args.compression = args32.compression;
4515 		args.encryption = args32.encryption;
4516 		memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4517 #else
4518 		return -ENOTTY;
4519 #endif
4520 	} else {
4521 		if (copy_from_user(&args, argp, sizeof(args))) {
4522 			ret = -EFAULT;
4523 			goto out_acct;
4524 		}
4525 	}
4526 
4527 	ret = -EINVAL;
4528 	if (args.flags != 0)
4529 		goto out_acct;
4530 	if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4531 		goto out_acct;
4532 	if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4533 	    args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4534 		goto out_acct;
4535 	if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4536 	    args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4537 		goto out_acct;
4538 	if (args.unencoded_offset > args.unencoded_len)
4539 		goto out_acct;
4540 	if (args.len > args.unencoded_len - args.unencoded_offset)
4541 		goto out_acct;
4542 
4543 	ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4544 			   &iov, &iter);
4545 	if (ret < 0)
4546 		goto out_acct;
4547 
4548 	if (iov_iter_count(&iter) == 0) {
4549 		ret = 0;
4550 		goto out_iov;
4551 	}
4552 	pos = args.offset;
4553 	ret = rw_verify_area(WRITE, file, &pos, args.len);
4554 	if (ret < 0)
4555 		goto out_iov;
4556 
4557 	init_sync_kiocb(&kiocb, file);
4558 	ret = kiocb_set_rw_flags(&kiocb, 0);
4559 	if (ret)
4560 		goto out_iov;
4561 	kiocb.ki_pos = pos;
4562 
4563 	file_start_write(file);
4564 
4565 	ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4566 	if (ret > 0)
4567 		fsnotify_modify(file);
4568 
4569 	file_end_write(file);
4570 out_iov:
4571 	kfree(iov);
4572 out_acct:
4573 	if (ret > 0)
4574 		add_wchar(current, ret);
4575 	inc_syscw(current);
4576 	return ret;
4577 }
4578 
4579 long btrfs_ioctl(struct file *file, unsigned int
4580 		cmd, unsigned long arg)
4581 {
4582 	struct inode *inode = file_inode(file);
4583 	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4584 	struct btrfs_root *root = BTRFS_I(inode)->root;
4585 	void __user *argp = (void __user *)arg;
4586 
4587 	switch (cmd) {
4588 	case FS_IOC_GETVERSION:
4589 		return btrfs_ioctl_getversion(inode, argp);
4590 	case FS_IOC_GETFSLABEL:
4591 		return btrfs_ioctl_get_fslabel(fs_info, argp);
4592 	case FS_IOC_SETFSLABEL:
4593 		return btrfs_ioctl_set_fslabel(file, argp);
4594 	case FITRIM:
4595 		return btrfs_ioctl_fitrim(fs_info, argp);
4596 	case BTRFS_IOC_SNAP_CREATE:
4597 		return btrfs_ioctl_snap_create(file, argp, 0);
4598 	case BTRFS_IOC_SNAP_CREATE_V2:
4599 		return btrfs_ioctl_snap_create_v2(file, argp, 0);
4600 	case BTRFS_IOC_SUBVOL_CREATE:
4601 		return btrfs_ioctl_snap_create(file, argp, 1);
4602 	case BTRFS_IOC_SUBVOL_CREATE_V2:
4603 		return btrfs_ioctl_snap_create_v2(file, argp, 1);
4604 	case BTRFS_IOC_SNAP_DESTROY:
4605 		return btrfs_ioctl_snap_destroy(file, argp, false);
4606 	case BTRFS_IOC_SNAP_DESTROY_V2:
4607 		return btrfs_ioctl_snap_destroy(file, argp, true);
4608 	case BTRFS_IOC_SUBVOL_GETFLAGS:
4609 		return btrfs_ioctl_subvol_getflags(inode, argp);
4610 	case BTRFS_IOC_SUBVOL_SETFLAGS:
4611 		return btrfs_ioctl_subvol_setflags(file, argp);
4612 	case BTRFS_IOC_DEFAULT_SUBVOL:
4613 		return btrfs_ioctl_default_subvol(file, argp);
4614 	case BTRFS_IOC_DEFRAG:
4615 		return btrfs_ioctl_defrag(file, NULL);
4616 	case BTRFS_IOC_DEFRAG_RANGE:
4617 		return btrfs_ioctl_defrag(file, argp);
4618 	case BTRFS_IOC_RESIZE:
4619 		return btrfs_ioctl_resize(file, argp);
4620 	case BTRFS_IOC_ADD_DEV:
4621 		return btrfs_ioctl_add_dev(fs_info, argp);
4622 	case BTRFS_IOC_RM_DEV:
4623 		return btrfs_ioctl_rm_dev(file, argp);
4624 	case BTRFS_IOC_RM_DEV_V2:
4625 		return btrfs_ioctl_rm_dev_v2(file, argp);
4626 	case BTRFS_IOC_FS_INFO:
4627 		return btrfs_ioctl_fs_info(fs_info, argp);
4628 	case BTRFS_IOC_DEV_INFO:
4629 		return btrfs_ioctl_dev_info(fs_info, argp);
4630 	case BTRFS_IOC_TREE_SEARCH:
4631 		return btrfs_ioctl_tree_search(inode, argp);
4632 	case BTRFS_IOC_TREE_SEARCH_V2:
4633 		return btrfs_ioctl_tree_search_v2(inode, argp);
4634 	case BTRFS_IOC_INO_LOOKUP:
4635 		return btrfs_ioctl_ino_lookup(root, argp);
4636 	case BTRFS_IOC_INO_PATHS:
4637 		return btrfs_ioctl_ino_to_path(root, argp);
4638 	case BTRFS_IOC_LOGICAL_INO:
4639 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4640 	case BTRFS_IOC_LOGICAL_INO_V2:
4641 		return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4642 	case BTRFS_IOC_SPACE_INFO:
4643 		return btrfs_ioctl_space_info(fs_info, argp);
4644 	case BTRFS_IOC_SYNC: {
4645 		int ret;
4646 
4647 		ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4648 		if (ret)
4649 			return ret;
4650 		ret = btrfs_sync_fs(inode->i_sb, 1);
4651 		/*
4652 		 * The transaction thread may want to do more work,
4653 		 * namely it pokes the cleaner kthread that will start
4654 		 * processing uncleaned subvols.
4655 		 */
4656 		wake_up_process(fs_info->transaction_kthread);
4657 		return ret;
4658 	}
4659 	case BTRFS_IOC_START_SYNC:
4660 		return btrfs_ioctl_start_sync(root, argp);
4661 	case BTRFS_IOC_WAIT_SYNC:
4662 		return btrfs_ioctl_wait_sync(fs_info, argp);
4663 	case BTRFS_IOC_SCRUB:
4664 		return btrfs_ioctl_scrub(file, argp);
4665 	case BTRFS_IOC_SCRUB_CANCEL:
4666 		return btrfs_ioctl_scrub_cancel(fs_info);
4667 	case BTRFS_IOC_SCRUB_PROGRESS:
4668 		return btrfs_ioctl_scrub_progress(fs_info, argp);
4669 	case BTRFS_IOC_BALANCE_V2:
4670 		return btrfs_ioctl_balance(file, argp);
4671 	case BTRFS_IOC_BALANCE_CTL:
4672 		return btrfs_ioctl_balance_ctl(fs_info, arg);
4673 	case BTRFS_IOC_BALANCE_PROGRESS:
4674 		return btrfs_ioctl_balance_progress(fs_info, argp);
4675 	case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4676 		return btrfs_ioctl_set_received_subvol(file, argp);
4677 #ifdef CONFIG_64BIT
4678 	case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4679 		return btrfs_ioctl_set_received_subvol_32(file, argp);
4680 #endif
4681 	case BTRFS_IOC_SEND:
4682 		return _btrfs_ioctl_send(inode, argp, false);
4683 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4684 	case BTRFS_IOC_SEND_32:
4685 		return _btrfs_ioctl_send(inode, argp, true);
4686 #endif
4687 	case BTRFS_IOC_GET_DEV_STATS:
4688 		return btrfs_ioctl_get_dev_stats(fs_info, argp);
4689 	case BTRFS_IOC_QUOTA_CTL:
4690 		return btrfs_ioctl_quota_ctl(file, argp);
4691 	case BTRFS_IOC_QGROUP_ASSIGN:
4692 		return btrfs_ioctl_qgroup_assign(file, argp);
4693 	case BTRFS_IOC_QGROUP_CREATE:
4694 		return btrfs_ioctl_qgroup_create(file, argp);
4695 	case BTRFS_IOC_QGROUP_LIMIT:
4696 		return btrfs_ioctl_qgroup_limit(file, argp);
4697 	case BTRFS_IOC_QUOTA_RESCAN:
4698 		return btrfs_ioctl_quota_rescan(file, argp);
4699 	case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4700 		return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4701 	case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4702 		return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4703 	case BTRFS_IOC_DEV_REPLACE:
4704 		return btrfs_ioctl_dev_replace(fs_info, argp);
4705 	case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4706 		return btrfs_ioctl_get_supported_features(argp);
4707 	case BTRFS_IOC_GET_FEATURES:
4708 		return btrfs_ioctl_get_features(fs_info, argp);
4709 	case BTRFS_IOC_SET_FEATURES:
4710 		return btrfs_ioctl_set_features(file, argp);
4711 	case BTRFS_IOC_GET_SUBVOL_INFO:
4712 		return btrfs_ioctl_get_subvol_info(inode, argp);
4713 	case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4714 		return btrfs_ioctl_get_subvol_rootref(root, argp);
4715 	case BTRFS_IOC_INO_LOOKUP_USER:
4716 		return btrfs_ioctl_ino_lookup_user(file, argp);
4717 	case FS_IOC_ENABLE_VERITY:
4718 		return fsverity_ioctl_enable(file, (const void __user *)argp);
4719 	case FS_IOC_MEASURE_VERITY:
4720 		return fsverity_ioctl_measure(file, argp);
4721 	case BTRFS_IOC_ENCODED_READ:
4722 		return btrfs_ioctl_encoded_read(file, argp, false);
4723 	case BTRFS_IOC_ENCODED_WRITE:
4724 		return btrfs_ioctl_encoded_write(file, argp, false);
4725 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4726 	case BTRFS_IOC_ENCODED_READ_32:
4727 		return btrfs_ioctl_encoded_read(file, argp, true);
4728 	case BTRFS_IOC_ENCODED_WRITE_32:
4729 		return btrfs_ioctl_encoded_write(file, argp, true);
4730 #endif
4731 	}
4732 
4733 	return -ENOTTY;
4734 }
4735 
4736 #ifdef CONFIG_COMPAT
4737 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4738 {
4739 	/*
4740 	 * These all access 32-bit values anyway so no further
4741 	 * handling is necessary.
4742 	 */
4743 	switch (cmd) {
4744 	case FS_IOC32_GETVERSION:
4745 		cmd = FS_IOC_GETVERSION;
4746 		break;
4747 	}
4748 
4749 	return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
4750 }
4751 #endif
4752