1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/kernel.h>
8 #include <linux/file.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>
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>
35 #include "transaction.h"
36 #include "btrfs_inode.h"
37 #include "print-tree.h"
41 #include "rcu-string.h"
43 #include "dev-replace.h"
48 #include "compression.h"
49 #include "space-info.h"
50 #include "delalloc-space.h"
51 #include "block-group.h"
54 #include "accessors.h"
55 #include "extent-tree.h"
56 #include "root-tree.h"
59 #include "uuid-tree.h"
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.
71 struct btrfs_ioctl_timespec_32 {
74 } __attribute__ ((__packed__));
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 */
83 __u64 reserved[16]; /* in */
84 } __attribute__ ((__packed__));
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
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 */
97 __u32 version; /* in */
98 __u8 reserved[28]; /* in */
99 } __attribute__ ((__packed__));
101 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
102 struct btrfs_ioctl_send_args_32)
104 struct btrfs_ioctl_encoded_io_args_32 {
106 compat_ulong_t iovcnt;
111 __u64 unencoded_offset;
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)
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,
127 if (S_ISDIR(inode->i_mode))
129 else if (S_ISREG(inode->i_mode))
130 return flags & ~FS_DIRSYNC_FL;
132 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
136 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
139 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
141 unsigned int iflags = 0;
142 u32 flags = binode->flags;
143 u32 ro_flags = binode->ro_flags;
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;
162 if (flags & BTRFS_INODE_NOCOMPRESS)
163 iflags |= FS_NOCOMP_FL;
164 else if (flags & BTRFS_INODE_COMPRESS)
165 iflags |= FS_COMPR_FL;
171 * Update inode->i_flags based on the btrfs internal flags.
173 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
175 struct btrfs_inode *binode = BTRFS_I(inode);
176 unsigned int new_fl = 0;
178 if (binode->flags & BTRFS_INODE_SYNC)
180 if (binode->flags & BTRFS_INODE_IMMUTABLE)
181 new_fl |= S_IMMUTABLE;
182 if (binode->flags & BTRFS_INODE_APPEND)
184 if (binode->flags & BTRFS_INODE_NOATIME)
186 if (binode->flags & BTRFS_INODE_DIRSYNC)
188 if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
191 set_mask_bits(&inode->i_flags,
192 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
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
200 static int check_fsflags(unsigned int old_flags, unsigned int flags)
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 |
209 /* COMPR and NOCOMP on new/old are valid */
210 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
213 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
216 /* NOCOW and compression options are mutually exclusive */
217 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
219 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
225 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
228 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
235 * Set flags/xflags from the internal inode flags. The remaining items of
236 * fsxattr are zeroed.
238 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
240 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
242 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
246 int btrfs_fileattr_set(struct user_namespace *mnt_userns,
247 struct dentry *dentry, struct fileattr *fa)
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;
256 const char *comp = NULL;
259 if (btrfs_root_readonly(root))
262 if (fileattr_has_fsx(fa))
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);
271 ret = check_fsflags_compatible(fs_info, fsflags);
275 binode_flags = binode->flags;
276 if (fsflags & FS_SYNC_FL)
277 binode_flags |= BTRFS_INODE_SYNC;
279 binode_flags &= ~BTRFS_INODE_SYNC;
280 if (fsflags & FS_IMMUTABLE_FL)
281 binode_flags |= BTRFS_INODE_IMMUTABLE;
283 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
284 if (fsflags & FS_APPEND_FL)
285 binode_flags |= BTRFS_INODE_APPEND;
287 binode_flags &= ~BTRFS_INODE_APPEND;
288 if (fsflags & FS_NODUMP_FL)
289 binode_flags |= BTRFS_INODE_NODUMP;
291 binode_flags &= ~BTRFS_INODE_NODUMP;
292 if (fsflags & FS_NOATIME_FL)
293 binode_flags |= BTRFS_INODE_NOATIME;
295 binode_flags &= ~BTRFS_INODE_NOATIME;
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);
302 return PTR_ERR(trans);
306 if (fsflags & FS_DIRSYNC_FL)
307 binode_flags |= BTRFS_INODE_DIRSYNC;
309 binode_flags &= ~BTRFS_INODE_DIRSYNC;
310 if (fsflags & FS_NOCOW_FL) {
311 if (S_ISREG(inode->i_mode)) {
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.
317 if (inode->i_size == 0)
318 binode_flags |= BTRFS_INODE_NODATACOW |
319 BTRFS_INODE_NODATASUM;
321 binode_flags |= BTRFS_INODE_NODATACOW;
325 * Revert back under same assumptions as above
327 if (S_ISREG(inode->i_mode)) {
328 if (inode->i_size == 0)
329 binode_flags &= ~(BTRFS_INODE_NODATACOW |
330 BTRFS_INODE_NODATASUM);
332 binode_flags &= ~BTRFS_INODE_NODATACOW;
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
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) {
346 if (IS_SWAPFILE(inode))
349 binode_flags |= BTRFS_INODE_COMPRESS;
350 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
352 comp = btrfs_compress_type2str(fs_info->compress_type);
353 if (!comp || comp[0] == 0)
354 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
356 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
363 trans = btrfs_start_transaction(root, 3);
365 return PTR_ERR(trans);
368 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
371 btrfs_abort_transaction(trans, ret);
375 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
377 if (ret && ret != -ENODATA) {
378 btrfs_abort_transaction(trans, ret);
384 binode->flags = binode_flags;
385 btrfs_sync_inode_flags_to_i_flags(inode);
386 inode_inc_iversion(inode);
387 inode->i_ctime = current_time(inode);
388 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
391 btrfs_end_transaction(trans);
396 * Start exclusive operation @type, return true on success
398 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
399 enum btrfs_exclusive_operation type)
403 spin_lock(&fs_info->super_lock);
404 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
405 fs_info->exclusive_operation = type;
408 spin_unlock(&fs_info->super_lock);
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.
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
424 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
425 enum btrfs_exclusive_operation type)
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))
433 spin_unlock(&fs_info->super_lock);
437 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
439 spin_unlock(&fs_info->super_lock);
442 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
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");
450 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
451 enum btrfs_exclusive_operation op)
454 case BTRFS_EXCLOP_BALANCE_PAUSED:
455 spin_lock(&fs_info->super_lock);
456 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
457 fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD);
458 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
459 spin_unlock(&fs_info->super_lock);
461 case BTRFS_EXCLOP_BALANCE:
462 spin_lock(&fs_info->super_lock);
463 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
464 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
465 spin_unlock(&fs_info->super_lock);
469 "invalid exclop balance operation %d requested", op);
473 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
475 return put_user(inode->i_generation, arg);
478 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
481 struct btrfs_device *device;
482 struct fstrim_range range;
483 u64 minlen = ULLONG_MAX;
487 if (!capable(CAP_SYS_ADMIN))
491 * btrfs_trim_block_group() depends on space cache, which is not
492 * available in zoned filesystem. So, disallow fitrim on a zoned
493 * filesystem for now.
495 if (btrfs_is_zoned(fs_info))
499 * If the fs is mounted with nologreplay, which requires it to be
500 * mounted in RO mode as well, we can not allow discard on free space
501 * inside block groups, because log trees refer to extents that are not
502 * pinned in a block group's free space cache (pinning the extents is
503 * precisely the first phase of replaying a log tree).
505 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
509 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
511 if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
514 minlen = min_t(u64, bdev_discard_granularity(device->bdev),
521 if (copy_from_user(&range, arg, sizeof(range)))
525 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
526 * block group is in the logical address space, which can be any
527 * sectorsize aligned bytenr in the range [0, U64_MAX].
529 if (range.len < fs_info->sb->s_blocksize)
532 range.minlen = max(range.minlen, minlen);
533 ret = btrfs_trim_fs(fs_info, &range);
537 if (copy_to_user(arg, &range, sizeof(range)))
543 int __pure btrfs_is_empty_uuid(u8 *uuid)
547 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
555 * Calculate the number of transaction items to reserve for creating a subvolume
556 * or snapshot, not including the inode, directory entries, or parent directory.
558 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
561 * 1 to add root block
564 * 1 to add root backref
566 * 1 to add qgroup info
567 * 1 to add qgroup limit
569 * Ideally the last two would only be accounted if qgroups are enabled,
570 * but that can change between now and the time we would insert them.
572 unsigned int num_items = 7;
575 /* 2 to add qgroup relations for each inherited qgroup */
576 num_items += 2 * inherit->num_qgroups;
581 static noinline int create_subvol(struct user_namespace *mnt_userns,
582 struct inode *dir, struct dentry *dentry,
583 struct btrfs_qgroup_inherit *inherit)
585 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
586 struct btrfs_trans_handle *trans;
587 struct btrfs_key key;
588 struct btrfs_root_item *root_item;
589 struct btrfs_inode_item *inode_item;
590 struct extent_buffer *leaf;
591 struct btrfs_root *root = BTRFS_I(dir)->root;
592 struct btrfs_root *new_root;
593 struct btrfs_block_rsv block_rsv;
594 struct timespec64 cur_time = current_time(dir);
595 struct btrfs_new_inode_args new_inode_args = {
600 unsigned int trans_num_items;
605 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
609 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
614 * Don't create subvolume whose level is not zero. Or qgroup will be
615 * screwed up since it assumes subvolume qgroup's level to be 0.
617 if (btrfs_qgroup_level(objectid)) {
622 ret = get_anon_bdev(&anon_dev);
626 new_inode_args.inode = btrfs_new_subvol_inode(mnt_userns, dir);
627 if (!new_inode_args.inode) {
631 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
634 trans_num_items += create_subvol_num_items(inherit);
636 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
637 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
638 trans_num_items, false);
640 goto out_new_inode_args;
642 trans = btrfs_start_transaction(root, 0);
644 ret = PTR_ERR(trans);
645 btrfs_subvolume_release_metadata(root, &block_rsv);
646 goto out_new_inode_args;
648 trans->block_rsv = &block_rsv;
649 trans->bytes_reserved = block_rsv.size;
651 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
655 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
656 BTRFS_NESTING_NORMAL);
662 btrfs_mark_buffer_dirty(leaf);
664 inode_item = &root_item->inode;
665 btrfs_set_stack_inode_generation(inode_item, 1);
666 btrfs_set_stack_inode_size(inode_item, 3);
667 btrfs_set_stack_inode_nlink(inode_item, 1);
668 btrfs_set_stack_inode_nbytes(inode_item,
670 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
672 btrfs_set_root_flags(root_item, 0);
673 btrfs_set_root_limit(root_item, 0);
674 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
676 btrfs_set_root_bytenr(root_item, leaf->start);
677 btrfs_set_root_generation(root_item, trans->transid);
678 btrfs_set_root_level(root_item, 0);
679 btrfs_set_root_refs(root_item, 1);
680 btrfs_set_root_used(root_item, leaf->len);
681 btrfs_set_root_last_snapshot(root_item, 0);
683 btrfs_set_root_generation_v2(root_item,
684 btrfs_root_generation(root_item));
685 generate_random_guid(root_item->uuid);
686 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
687 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
688 root_item->ctime = root_item->otime;
689 btrfs_set_root_ctransid(root_item, trans->transid);
690 btrfs_set_root_otransid(root_item, trans->transid);
692 btrfs_tree_unlock(leaf);
694 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
696 key.objectid = objectid;
698 key.type = BTRFS_ROOT_ITEM_KEY;
699 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
703 * Since we don't abort the transaction in this case, free the
704 * tree block so that we don't leak space and leave the
705 * filesystem in an inconsistent state (an extent item in the
706 * extent tree with a backreference for a root that does not
709 btrfs_tree_lock(leaf);
710 btrfs_clean_tree_block(leaf);
711 btrfs_tree_unlock(leaf);
712 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
713 free_extent_buffer(leaf);
717 free_extent_buffer(leaf);
720 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
721 if (IS_ERR(new_root)) {
722 ret = PTR_ERR(new_root);
723 btrfs_abort_transaction(trans, ret);
726 /* anon_dev is owned by new_root now. */
728 BTRFS_I(new_inode_args.inode)->root = new_root;
729 /* ... and new_root is owned by new_inode_args.inode now. */
731 ret = btrfs_record_root_in_trans(trans, new_root);
733 btrfs_abort_transaction(trans, ret);
737 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
738 BTRFS_UUID_KEY_SUBVOL, objectid);
740 btrfs_abort_transaction(trans, ret);
744 ret = btrfs_create_new_inode(trans, &new_inode_args);
746 btrfs_abort_transaction(trans, ret);
750 d_instantiate_new(dentry, new_inode_args.inode);
751 new_inode_args.inode = NULL;
754 trans->block_rsv = NULL;
755 trans->bytes_reserved = 0;
756 btrfs_subvolume_release_metadata(root, &block_rsv);
759 btrfs_end_transaction(trans);
761 ret = btrfs_commit_transaction(trans);
763 btrfs_new_inode_args_destroy(&new_inode_args);
765 iput(new_inode_args.inode);
768 free_anon_bdev(anon_dev);
774 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
775 struct dentry *dentry, bool readonly,
776 struct btrfs_qgroup_inherit *inherit)
778 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
780 struct btrfs_pending_snapshot *pending_snapshot;
781 unsigned int trans_num_items;
782 struct btrfs_trans_handle *trans;
785 /* We do not support snapshotting right now. */
786 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
788 "extent tree v2 doesn't support snapshotting yet");
792 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
795 if (atomic_read(&root->nr_swapfiles)) {
797 "cannot snapshot subvolume with active swapfile");
801 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
802 if (!pending_snapshot)
805 ret = get_anon_bdev(&pending_snapshot->anon_dev);
808 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
810 pending_snapshot->path = btrfs_alloc_path();
811 if (!pending_snapshot->root_item || !pending_snapshot->path) {
816 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
817 BTRFS_BLOCK_RSV_TEMP);
821 * 1 to update parent inode item
823 trans_num_items = create_subvol_num_items(inherit) + 3;
824 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
825 &pending_snapshot->block_rsv,
826 trans_num_items, false);
830 pending_snapshot->dentry = dentry;
831 pending_snapshot->root = root;
832 pending_snapshot->readonly = readonly;
833 pending_snapshot->dir = dir;
834 pending_snapshot->inherit = inherit;
836 trans = btrfs_start_transaction(root, 0);
838 ret = PTR_ERR(trans);
842 trans->pending_snapshot = pending_snapshot;
844 ret = btrfs_commit_transaction(trans);
848 ret = pending_snapshot->error;
852 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
856 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
858 ret = PTR_ERR(inode);
862 d_instantiate(dentry, inode);
864 pending_snapshot->anon_dev = 0;
866 /* Prevent double freeing of anon_dev */
867 if (ret && pending_snapshot->snap)
868 pending_snapshot->snap->anon_dev = 0;
869 btrfs_put_root(pending_snapshot->snap);
870 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
872 if (pending_snapshot->anon_dev)
873 free_anon_bdev(pending_snapshot->anon_dev);
874 kfree(pending_snapshot->root_item);
875 btrfs_free_path(pending_snapshot->path);
876 kfree(pending_snapshot);
881 /* copy of may_delete in fs/namei.c()
882 * Check whether we can remove a link victim from directory dir, check
883 * whether the type of victim is right.
884 * 1. We can't do it if dir is read-only (done in permission())
885 * 2. We should have write and exec permissions on dir
886 * 3. We can't remove anything from append-only dir
887 * 4. We can't do anything with immutable dir (done in permission())
888 * 5. If the sticky bit on dir is set we should either
889 * a. be owner of dir, or
890 * b. be owner of victim, or
891 * c. have CAP_FOWNER capability
892 * 6. If the victim is append-only or immutable we can't do anything with
893 * links pointing to it.
894 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
895 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
896 * 9. We can't remove a root or mountpoint.
897 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
898 * nfs_async_unlink().
901 static int btrfs_may_delete(struct user_namespace *mnt_userns,
902 struct inode *dir, struct dentry *victim, int isdir)
906 if (d_really_is_negative(victim))
909 BUG_ON(d_inode(victim->d_parent) != dir);
910 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
912 error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
917 if (check_sticky(mnt_userns, dir, d_inode(victim)) ||
918 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
919 IS_SWAPFILE(d_inode(victim)))
922 if (!d_is_dir(victim))
926 } else if (d_is_dir(victim))
930 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
935 /* copy of may_create in fs/namei.c() */
936 static inline int btrfs_may_create(struct user_namespace *mnt_userns,
937 struct inode *dir, struct dentry *child)
939 if (d_really_is_positive(child))
943 if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns))
945 return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
949 * Create a new subvolume below @parent. This is largely modeled after
950 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
951 * inside this filesystem so it's quite a bit simpler.
953 static noinline int btrfs_mksubvol(const struct path *parent,
954 struct user_namespace *mnt_userns,
955 const char *name, int namelen,
956 struct btrfs_root *snap_src,
958 struct btrfs_qgroup_inherit *inherit)
960 struct inode *dir = d_inode(parent->dentry);
961 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
962 struct dentry *dentry;
963 struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
966 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
970 dentry = lookup_one(mnt_userns, name, parent->dentry, namelen);
971 error = PTR_ERR(dentry);
975 error = btrfs_may_create(mnt_userns, dir, dentry);
980 * even if this name doesn't exist, we may get hash collisions.
981 * check for them now when we can safely fail
983 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
984 dir->i_ino, &name_str);
988 down_read(&fs_info->subvol_sem);
990 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
994 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
996 error = create_subvol(mnt_userns, dir, dentry, inherit);
999 fsnotify_mkdir(dir, dentry);
1001 up_read(&fs_info->subvol_sem);
1005 btrfs_inode_unlock(BTRFS_I(dir), 0);
1009 static noinline int btrfs_mksnapshot(const struct path *parent,
1010 struct user_namespace *mnt_userns,
1011 const char *name, int namelen,
1012 struct btrfs_root *root,
1014 struct btrfs_qgroup_inherit *inherit)
1017 bool snapshot_force_cow = false;
1020 * Force new buffered writes to reserve space even when NOCOW is
1021 * possible. This is to avoid later writeback (running dealloc) to
1022 * fallback to COW mode and unexpectedly fail with ENOSPC.
1024 btrfs_drew_read_lock(&root->snapshot_lock);
1026 ret = btrfs_start_delalloc_snapshot(root, false);
1031 * All previous writes have started writeback in NOCOW mode, so now
1032 * we force future writes to fallback to COW mode during snapshot
1035 atomic_inc(&root->snapshot_force_cow);
1036 snapshot_force_cow = true;
1038 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1040 ret = btrfs_mksubvol(parent, mnt_userns, name, namelen,
1041 root, readonly, inherit);
1043 if (snapshot_force_cow)
1044 atomic_dec(&root->snapshot_force_cow);
1045 btrfs_drew_read_unlock(&root->snapshot_lock);
1050 * Try to start exclusive operation @type or cancel it if it's running.
1053 * 0 - normal mode, newly claimed op started
1054 * >0 - normal mode, something else is running,
1055 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1056 * ECANCELED - cancel mode, successful cancel
1057 * ENOTCONN - cancel mode, operation not running anymore
1059 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1060 enum btrfs_exclusive_operation type, bool cancel)
1063 /* Start normal op */
1064 if (!btrfs_exclop_start(fs_info, type))
1065 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1066 /* Exclusive operation is now claimed */
1070 /* Cancel running op */
1071 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1073 * This blocks any exclop finish from setting it to NONE, so we
1074 * request cancellation. Either it runs and we will wait for it,
1075 * or it has finished and no waiting will happen.
1077 atomic_inc(&fs_info->reloc_cancel_req);
1078 btrfs_exclop_start_unlock(fs_info);
1080 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1081 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1082 TASK_INTERRUPTIBLE);
1087 /* Something else is running or none */
1091 static noinline int btrfs_ioctl_resize(struct file *file,
1094 BTRFS_DEV_LOOKUP_ARGS(args);
1095 struct inode *inode = file_inode(file);
1096 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1100 struct btrfs_root *root = BTRFS_I(inode)->root;
1101 struct btrfs_ioctl_vol_args *vol_args;
1102 struct btrfs_trans_handle *trans;
1103 struct btrfs_device *device = NULL;
1106 char *devstr = NULL;
1111 if (!capable(CAP_SYS_ADMIN))
1114 ret = mnt_want_write_file(file);
1119 * Read the arguments before checking exclusivity to be able to
1120 * distinguish regular resize and cancel
1122 vol_args = memdup_user(arg, sizeof(*vol_args));
1123 if (IS_ERR(vol_args)) {
1124 ret = PTR_ERR(vol_args);
1127 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1128 sizestr = vol_args->name;
1129 cancel = (strcmp("cancel", sizestr) == 0);
1130 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1133 /* Exclusive operation is now claimed */
1135 devstr = strchr(sizestr, ':');
1137 sizestr = devstr + 1;
1139 devstr = vol_args->name;
1140 ret = kstrtoull(devstr, 10, &devid);
1147 btrfs_info(fs_info, "resizing devid %llu", devid);
1151 device = btrfs_find_device(fs_info->fs_devices, &args);
1153 btrfs_info(fs_info, "resizer unable to find device %llu",
1159 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1161 "resizer unable to apply on readonly device %llu",
1167 if (!strcmp(sizestr, "max"))
1168 new_size = bdev_nr_bytes(device->bdev);
1170 if (sizestr[0] == '-') {
1173 } else if (sizestr[0] == '+') {
1177 new_size = memparse(sizestr, &retptr);
1178 if (*retptr != '\0' || new_size == 0) {
1184 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1189 old_size = btrfs_device_get_total_bytes(device);
1192 if (new_size > old_size) {
1196 new_size = old_size - new_size;
1197 } else if (mod > 0) {
1198 if (new_size > ULLONG_MAX - old_size) {
1202 new_size = old_size + new_size;
1205 if (new_size < SZ_256M) {
1209 if (new_size > bdev_nr_bytes(device->bdev)) {
1214 new_size = round_down(new_size, fs_info->sectorsize);
1216 if (new_size > old_size) {
1217 trans = btrfs_start_transaction(root, 0);
1218 if (IS_ERR(trans)) {
1219 ret = PTR_ERR(trans);
1222 ret = btrfs_grow_device(trans, device, new_size);
1223 btrfs_commit_transaction(trans);
1224 } else if (new_size < old_size) {
1225 ret = btrfs_shrink_device(device, new_size);
1226 } /* equal, nothing need to do */
1228 if (ret == 0 && new_size != old_size)
1229 btrfs_info_in_rcu(fs_info,
1230 "resize device %s (devid %llu) from %llu to %llu",
1231 btrfs_dev_name(device), device->devid,
1232 old_size, new_size);
1234 btrfs_exclop_finish(fs_info);
1238 mnt_drop_write_file(file);
1242 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1243 struct user_namespace *mnt_userns,
1244 const char *name, unsigned long fd, int subvol,
1246 struct btrfs_qgroup_inherit *inherit)
1251 if (!S_ISDIR(file_inode(file)->i_mode))
1254 ret = mnt_want_write_file(file);
1258 namelen = strlen(name);
1259 if (strchr(name, '/')) {
1261 goto out_drop_write;
1264 if (name[0] == '.' &&
1265 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1267 goto out_drop_write;
1271 ret = btrfs_mksubvol(&file->f_path, mnt_userns, name,
1272 namelen, NULL, readonly, inherit);
1274 struct fd src = fdget(fd);
1275 struct inode *src_inode;
1278 goto out_drop_write;
1281 src_inode = file_inode(src.file);
1282 if (src_inode->i_sb != file_inode(file)->i_sb) {
1283 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1284 "Snapshot src from another FS");
1286 } else if (!inode_owner_or_capable(mnt_userns, src_inode)) {
1288 * Subvolume creation is not restricted, but snapshots
1289 * are limited to own subvolumes only
1293 ret = btrfs_mksnapshot(&file->f_path, mnt_userns,
1295 BTRFS_I(src_inode)->root,
1301 mnt_drop_write_file(file);
1306 static noinline int btrfs_ioctl_snap_create(struct file *file,
1307 void __user *arg, int subvol)
1309 struct btrfs_ioctl_vol_args *vol_args;
1312 if (!S_ISDIR(file_inode(file)->i_mode))
1315 vol_args = memdup_user(arg, sizeof(*vol_args));
1316 if (IS_ERR(vol_args))
1317 return PTR_ERR(vol_args);
1318 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1320 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
1321 vol_args->name, vol_args->fd, subvol,
1328 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1329 void __user *arg, int subvol)
1331 struct btrfs_ioctl_vol_args_v2 *vol_args;
1333 bool readonly = false;
1334 struct btrfs_qgroup_inherit *inherit = NULL;
1336 if (!S_ISDIR(file_inode(file)->i_mode))
1339 vol_args = memdup_user(arg, sizeof(*vol_args));
1340 if (IS_ERR(vol_args))
1341 return PTR_ERR(vol_args);
1342 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1344 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1349 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1351 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1354 if (vol_args->size < sizeof(*inherit) ||
1355 vol_args->size > PAGE_SIZE) {
1359 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1360 if (IS_ERR(inherit)) {
1361 ret = PTR_ERR(inherit);
1365 if (inherit->num_qgroups > PAGE_SIZE ||
1366 inherit->num_ref_copies > PAGE_SIZE ||
1367 inherit->num_excl_copies > PAGE_SIZE) {
1372 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1373 2 * inherit->num_excl_copies;
1374 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1380 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
1381 vol_args->name, vol_args->fd, subvol,
1392 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1395 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1396 struct btrfs_root *root = BTRFS_I(inode)->root;
1400 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1403 down_read(&fs_info->subvol_sem);
1404 if (btrfs_root_readonly(root))
1405 flags |= BTRFS_SUBVOL_RDONLY;
1406 up_read(&fs_info->subvol_sem);
1408 if (copy_to_user(arg, &flags, sizeof(flags)))
1414 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1417 struct inode *inode = file_inode(file);
1418 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1419 struct btrfs_root *root = BTRFS_I(inode)->root;
1420 struct btrfs_trans_handle *trans;
1425 if (!inode_owner_or_capable(file_mnt_user_ns(file), inode))
1428 ret = mnt_want_write_file(file);
1432 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1434 goto out_drop_write;
1437 if (copy_from_user(&flags, arg, sizeof(flags))) {
1439 goto out_drop_write;
1442 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1444 goto out_drop_write;
1447 down_write(&fs_info->subvol_sem);
1450 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1453 root_flags = btrfs_root_flags(&root->root_item);
1454 if (flags & BTRFS_SUBVOL_RDONLY) {
1455 btrfs_set_root_flags(&root->root_item,
1456 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1459 * Block RO -> RW transition if this subvolume is involved in
1462 spin_lock(&root->root_item_lock);
1463 if (root->send_in_progress == 0) {
1464 btrfs_set_root_flags(&root->root_item,
1465 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1466 spin_unlock(&root->root_item_lock);
1468 spin_unlock(&root->root_item_lock);
1470 "Attempt to set subvolume %llu read-write during send",
1471 root->root_key.objectid);
1477 trans = btrfs_start_transaction(root, 1);
1478 if (IS_ERR(trans)) {
1479 ret = PTR_ERR(trans);
1483 ret = btrfs_update_root(trans, fs_info->tree_root,
1484 &root->root_key, &root->root_item);
1486 btrfs_end_transaction(trans);
1490 ret = btrfs_commit_transaction(trans);
1494 btrfs_set_root_flags(&root->root_item, root_flags);
1496 up_write(&fs_info->subvol_sem);
1498 mnt_drop_write_file(file);
1503 static noinline int key_in_sk(struct btrfs_key *key,
1504 struct btrfs_ioctl_search_key *sk)
1506 struct btrfs_key test;
1509 test.objectid = sk->min_objectid;
1510 test.type = sk->min_type;
1511 test.offset = sk->min_offset;
1513 ret = btrfs_comp_cpu_keys(key, &test);
1517 test.objectid = sk->max_objectid;
1518 test.type = sk->max_type;
1519 test.offset = sk->max_offset;
1521 ret = btrfs_comp_cpu_keys(key, &test);
1527 static noinline int copy_to_sk(struct btrfs_path *path,
1528 struct btrfs_key *key,
1529 struct btrfs_ioctl_search_key *sk,
1532 unsigned long *sk_offset,
1536 struct extent_buffer *leaf;
1537 struct btrfs_ioctl_search_header sh;
1538 struct btrfs_key test;
1539 unsigned long item_off;
1540 unsigned long item_len;
1546 leaf = path->nodes[0];
1547 slot = path->slots[0];
1548 nritems = btrfs_header_nritems(leaf);
1550 if (btrfs_header_generation(leaf) > sk->max_transid) {
1554 found_transid = btrfs_header_generation(leaf);
1556 for (i = slot; i < nritems; i++) {
1557 item_off = btrfs_item_ptr_offset(leaf, i);
1558 item_len = btrfs_item_size(leaf, i);
1560 btrfs_item_key_to_cpu(leaf, key, i);
1561 if (!key_in_sk(key, sk))
1564 if (sizeof(sh) + item_len > *buf_size) {
1571 * return one empty item back for v1, which does not
1575 *buf_size = sizeof(sh) + item_len;
1580 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1585 sh.objectid = key->objectid;
1586 sh.offset = key->offset;
1587 sh.type = key->type;
1589 sh.transid = found_transid;
1592 * Copy search result header. If we fault then loop again so we
1593 * can fault in the pages and -EFAULT there if there's a
1594 * problem. Otherwise we'll fault and then copy the buffer in
1595 * properly this next time through
1597 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1602 *sk_offset += sizeof(sh);
1605 char __user *up = ubuf + *sk_offset;
1607 * Copy the item, same behavior as above, but reset the
1608 * * sk_offset so we copy the full thing again.
1610 if (read_extent_buffer_to_user_nofault(leaf, up,
1611 item_off, item_len)) {
1613 *sk_offset -= sizeof(sh);
1617 *sk_offset += item_len;
1621 if (ret) /* -EOVERFLOW from above */
1624 if (*num_found >= sk->nr_items) {
1631 test.objectid = sk->max_objectid;
1632 test.type = sk->max_type;
1633 test.offset = sk->max_offset;
1634 if (btrfs_comp_cpu_keys(key, &test) >= 0)
1636 else if (key->offset < (u64)-1)
1638 else if (key->type < (u8)-1) {
1641 } else if (key->objectid < (u64)-1) {
1649 * 0: all items from this leaf copied, continue with next
1650 * 1: * more items can be copied, but unused buffer is too small
1651 * * all items were found
1652 * Either way, it will stops the loop which iterates to the next
1654 * -EOVERFLOW: item was to large for buffer
1655 * -EFAULT: could not copy extent buffer back to userspace
1660 static noinline int search_ioctl(struct inode *inode,
1661 struct btrfs_ioctl_search_key *sk,
1665 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1666 struct btrfs_root *root;
1667 struct btrfs_key key;
1668 struct btrfs_path *path;
1671 unsigned long sk_offset = 0;
1673 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1674 *buf_size = sizeof(struct btrfs_ioctl_search_header);
1678 path = btrfs_alloc_path();
1682 if (sk->tree_id == 0) {
1683 /* search the root of the inode that was passed */
1684 root = btrfs_grab_root(BTRFS_I(inode)->root);
1686 root = btrfs_get_fs_root(info, sk->tree_id, true);
1688 btrfs_free_path(path);
1689 return PTR_ERR(root);
1693 key.objectid = sk->min_objectid;
1694 key.type = sk->min_type;
1695 key.offset = sk->min_offset;
1700 * Ensure that the whole user buffer is faulted in at sub-page
1701 * granularity, otherwise the loop may live-lock.
1703 if (fault_in_subpage_writeable(ubuf + sk_offset,
1704 *buf_size - sk_offset))
1707 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1713 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1714 &sk_offset, &num_found);
1715 btrfs_release_path(path);
1723 sk->nr_items = num_found;
1724 btrfs_put_root(root);
1725 btrfs_free_path(path);
1729 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1732 struct btrfs_ioctl_search_args __user *uargs = argp;
1733 struct btrfs_ioctl_search_key sk;
1737 if (!capable(CAP_SYS_ADMIN))
1740 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1743 buf_size = sizeof(uargs->buf);
1745 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1748 * In the origin implementation an overflow is handled by returning a
1749 * search header with a len of zero, so reset ret.
1751 if (ret == -EOVERFLOW)
1754 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1759 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1762 struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1763 struct btrfs_ioctl_search_args_v2 args;
1766 const size_t buf_limit = SZ_16M;
1768 if (!capable(CAP_SYS_ADMIN))
1771 /* copy search header and buffer size */
1772 if (copy_from_user(&args, uarg, sizeof(args)))
1775 buf_size = args.buf_size;
1777 /* limit result size to 16MB */
1778 if (buf_size > buf_limit)
1779 buf_size = buf_limit;
1781 ret = search_ioctl(inode, &args.key, &buf_size,
1782 (char __user *)(&uarg->buf[0]));
1783 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1785 else if (ret == -EOVERFLOW &&
1786 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1793 * Search INODE_REFs to identify path name of 'dirid' directory
1794 * in a 'tree_id' tree. and sets path name to 'name'.
1796 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1797 u64 tree_id, u64 dirid, char *name)
1799 struct btrfs_root *root;
1800 struct btrfs_key key;
1806 struct btrfs_inode_ref *iref;
1807 struct extent_buffer *l;
1808 struct btrfs_path *path;
1810 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1815 path = btrfs_alloc_path();
1819 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1821 root = btrfs_get_fs_root(info, tree_id, true);
1823 ret = PTR_ERR(root);
1828 key.objectid = dirid;
1829 key.type = BTRFS_INODE_REF_KEY;
1830 key.offset = (u64)-1;
1833 ret = btrfs_search_backwards(root, &key, path);
1842 slot = path->slots[0];
1844 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1845 len = btrfs_inode_ref_name_len(l, iref);
1847 total_len += len + 1;
1849 ret = -ENAMETOOLONG;
1854 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1856 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1859 btrfs_release_path(path);
1860 key.objectid = key.offset;
1861 key.offset = (u64)-1;
1862 dirid = key.objectid;
1864 memmove(name, ptr, total_len);
1865 name[total_len] = '\0';
1868 btrfs_put_root(root);
1869 btrfs_free_path(path);
1873 static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns,
1874 struct inode *inode,
1875 struct btrfs_ioctl_ino_lookup_user_args *args)
1877 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1878 struct super_block *sb = inode->i_sb;
1879 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1880 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1881 u64 dirid = args->dirid;
1882 unsigned long item_off;
1883 unsigned long item_len;
1884 struct btrfs_inode_ref *iref;
1885 struct btrfs_root_ref *rref;
1886 struct btrfs_root *root = NULL;
1887 struct btrfs_path *path;
1888 struct btrfs_key key, key2;
1889 struct extent_buffer *leaf;
1890 struct inode *temp_inode;
1897 path = btrfs_alloc_path();
1902 * If the bottom subvolume does not exist directly under upper_limit,
1903 * construct the path in from the bottom up.
1905 if (dirid != upper_limit.objectid) {
1906 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1908 root = btrfs_get_fs_root(fs_info, treeid, true);
1910 ret = PTR_ERR(root);
1914 key.objectid = dirid;
1915 key.type = BTRFS_INODE_REF_KEY;
1916 key.offset = (u64)-1;
1918 ret = btrfs_search_backwards(root, &key, path);
1926 leaf = path->nodes[0];
1927 slot = path->slots[0];
1929 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1930 len = btrfs_inode_ref_name_len(leaf, iref);
1932 total_len += len + 1;
1933 if (ptr < args->path) {
1934 ret = -ENAMETOOLONG;
1939 read_extent_buffer(leaf, ptr,
1940 (unsigned long)(iref + 1), len);
1942 /* Check the read+exec permission of this directory */
1943 ret = btrfs_previous_item(root, path, dirid,
1944 BTRFS_INODE_ITEM_KEY);
1947 } else if (ret > 0) {
1952 leaf = path->nodes[0];
1953 slot = path->slots[0];
1954 btrfs_item_key_to_cpu(leaf, &key2, slot);
1955 if (key2.objectid != dirid) {
1960 temp_inode = btrfs_iget(sb, key2.objectid, root);
1961 if (IS_ERR(temp_inode)) {
1962 ret = PTR_ERR(temp_inode);
1965 ret = inode_permission(mnt_userns, temp_inode,
1966 MAY_READ | MAY_EXEC);
1973 if (key.offset == upper_limit.objectid)
1975 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
1980 btrfs_release_path(path);
1981 key.objectid = key.offset;
1982 key.offset = (u64)-1;
1983 dirid = key.objectid;
1986 memmove(args->path, ptr, total_len);
1987 args->path[total_len] = '\0';
1988 btrfs_put_root(root);
1990 btrfs_release_path(path);
1993 /* Get the bottom subvolume's name from ROOT_REF */
1994 key.objectid = treeid;
1995 key.type = BTRFS_ROOT_REF_KEY;
1996 key.offset = args->treeid;
1997 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2000 } else if (ret > 0) {
2005 leaf = path->nodes[0];
2006 slot = path->slots[0];
2007 btrfs_item_key_to_cpu(leaf, &key, slot);
2009 item_off = btrfs_item_ptr_offset(leaf, slot);
2010 item_len = btrfs_item_size(leaf, slot);
2011 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2012 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2013 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2018 /* Copy subvolume's name */
2019 item_off += sizeof(struct btrfs_root_ref);
2020 item_len -= sizeof(struct btrfs_root_ref);
2021 read_extent_buffer(leaf, args->name, item_off, item_len);
2022 args->name[item_len] = 0;
2025 btrfs_put_root(root);
2027 btrfs_free_path(path);
2031 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2034 struct btrfs_ioctl_ino_lookup_args *args;
2037 args = memdup_user(argp, sizeof(*args));
2039 return PTR_ERR(args);
2042 * Unprivileged query to obtain the containing subvolume root id. The
2043 * path is reset so it's consistent with btrfs_search_path_in_tree.
2045 if (args->treeid == 0)
2046 args->treeid = root->root_key.objectid;
2048 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2053 if (!capable(CAP_SYS_ADMIN)) {
2058 ret = btrfs_search_path_in_tree(root->fs_info,
2059 args->treeid, args->objectid,
2063 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2071 * Version of ino_lookup ioctl (unprivileged)
2073 * The main differences from ino_lookup ioctl are:
2075 * 1. Read + Exec permission will be checked using inode_permission() during
2076 * path construction. -EACCES will be returned in case of failure.
2077 * 2. Path construction will be stopped at the inode number which corresponds
2078 * to the fd with which this ioctl is called. If constructed path does not
2079 * exist under fd's inode, -EACCES will be returned.
2080 * 3. The name of bottom subvolume is also searched and filled.
2082 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2084 struct btrfs_ioctl_ino_lookup_user_args *args;
2085 struct inode *inode;
2088 args = memdup_user(argp, sizeof(*args));
2090 return PTR_ERR(args);
2092 inode = file_inode(file);
2094 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2095 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2097 * The subvolume does not exist under fd with which this is
2104 ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args);
2106 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2113 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2114 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2116 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2117 struct btrfs_fs_info *fs_info;
2118 struct btrfs_root *root;
2119 struct btrfs_path *path;
2120 struct btrfs_key key;
2121 struct btrfs_root_item *root_item;
2122 struct btrfs_root_ref *rref;
2123 struct extent_buffer *leaf;
2124 unsigned long item_off;
2125 unsigned long item_len;
2129 path = btrfs_alloc_path();
2133 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2135 btrfs_free_path(path);
2139 fs_info = BTRFS_I(inode)->root->fs_info;
2141 /* Get root_item of inode's subvolume */
2142 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2143 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2145 ret = PTR_ERR(root);
2148 root_item = &root->root_item;
2150 subvol_info->treeid = key.objectid;
2152 subvol_info->generation = btrfs_root_generation(root_item);
2153 subvol_info->flags = btrfs_root_flags(root_item);
2155 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2156 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2158 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2161 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2162 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2163 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2165 subvol_info->otransid = btrfs_root_otransid(root_item);
2166 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2167 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2169 subvol_info->stransid = btrfs_root_stransid(root_item);
2170 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2171 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2173 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2174 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2175 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2177 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2178 /* Search root tree for ROOT_BACKREF of this subvolume */
2179 key.type = BTRFS_ROOT_BACKREF_KEY;
2181 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2184 } else if (path->slots[0] >=
2185 btrfs_header_nritems(path->nodes[0])) {
2186 ret = btrfs_next_leaf(fs_info->tree_root, path);
2189 } else if (ret > 0) {
2195 leaf = path->nodes[0];
2196 slot = path->slots[0];
2197 btrfs_item_key_to_cpu(leaf, &key, slot);
2198 if (key.objectid == subvol_info->treeid &&
2199 key.type == BTRFS_ROOT_BACKREF_KEY) {
2200 subvol_info->parent_id = key.offset;
2202 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2203 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2205 item_off = btrfs_item_ptr_offset(leaf, slot)
2206 + sizeof(struct btrfs_root_ref);
2207 item_len = btrfs_item_size(leaf, slot)
2208 - sizeof(struct btrfs_root_ref);
2209 read_extent_buffer(leaf, subvol_info->name,
2210 item_off, item_len);
2217 btrfs_free_path(path);
2219 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2223 btrfs_put_root(root);
2225 btrfs_free_path(path);
2231 * Return ROOT_REF information of the subvolume containing this inode
2232 * except the subvolume name.
2234 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2237 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2238 struct btrfs_root_ref *rref;
2239 struct btrfs_path *path;
2240 struct btrfs_key key;
2241 struct extent_buffer *leaf;
2247 path = btrfs_alloc_path();
2251 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2252 if (IS_ERR(rootrefs)) {
2253 btrfs_free_path(path);
2254 return PTR_ERR(rootrefs);
2257 objectid = root->root_key.objectid;
2258 key.objectid = objectid;
2259 key.type = BTRFS_ROOT_REF_KEY;
2260 key.offset = rootrefs->min_treeid;
2263 root = root->fs_info->tree_root;
2264 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2267 } else if (path->slots[0] >=
2268 btrfs_header_nritems(path->nodes[0])) {
2269 ret = btrfs_next_leaf(root, path);
2272 } else if (ret > 0) {
2278 leaf = path->nodes[0];
2279 slot = path->slots[0];
2281 btrfs_item_key_to_cpu(leaf, &key, slot);
2282 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2287 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2292 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2293 rootrefs->rootref[found].treeid = key.offset;
2294 rootrefs->rootref[found].dirid =
2295 btrfs_root_ref_dirid(leaf, rref);
2298 ret = btrfs_next_item(root, path);
2301 } else if (ret > 0) {
2308 btrfs_free_path(path);
2310 if (!ret || ret == -EOVERFLOW) {
2311 rootrefs->num_items = found;
2312 /* update min_treeid for next search */
2314 rootrefs->min_treeid =
2315 rootrefs->rootref[found - 1].treeid + 1;
2316 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2325 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2329 struct dentry *parent = file->f_path.dentry;
2330 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2331 struct dentry *dentry;
2332 struct inode *dir = d_inode(parent);
2333 struct inode *inode;
2334 struct btrfs_root *root = BTRFS_I(dir)->root;
2335 struct btrfs_root *dest = NULL;
2336 struct btrfs_ioctl_vol_args *vol_args = NULL;
2337 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2338 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
2339 char *subvol_name, *subvol_name_ptr = NULL;
2342 bool destroy_parent = false;
2344 /* We don't support snapshots with extent tree v2 yet. */
2345 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2347 "extent tree v2 doesn't support snapshot deletion yet");
2352 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2353 if (IS_ERR(vol_args2))
2354 return PTR_ERR(vol_args2);
2356 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2362 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2363 * name, same as v1 currently does.
2365 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2366 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2367 subvol_name = vol_args2->name;
2369 err = mnt_want_write_file(file);
2373 struct inode *old_dir;
2375 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2380 err = mnt_want_write_file(file);
2384 dentry = btrfs_get_dentry(fs_info->sb,
2385 BTRFS_FIRST_FREE_OBJECTID,
2386 vol_args2->subvolid, 0);
2387 if (IS_ERR(dentry)) {
2388 err = PTR_ERR(dentry);
2389 goto out_drop_write;
2393 * Change the default parent since the subvolume being
2394 * deleted can be outside of the current mount point.
2396 parent = btrfs_get_parent(dentry);
2399 * At this point dentry->d_name can point to '/' if the
2400 * subvolume we want to destroy is outsite of the
2401 * current mount point, so we need to release the
2402 * current dentry and execute the lookup to return a new
2403 * one with ->d_name pointing to the
2404 * <mount point>/subvol_name.
2407 if (IS_ERR(parent)) {
2408 err = PTR_ERR(parent);
2409 goto out_drop_write;
2412 dir = d_inode(parent);
2415 * If v2 was used with SPEC_BY_ID, a new parent was
2416 * allocated since the subvolume can be outside of the
2417 * current mount point. Later on we need to release this
2418 * new parent dentry.
2420 destroy_parent = true;
2423 * On idmapped mounts, deletion via subvolid is
2424 * restricted to subvolumes that are immediate
2425 * ancestors of the inode referenced by the file
2426 * descriptor in the ioctl. Otherwise the idmapping
2427 * could potentially be abused to delete subvolumes
2428 * anywhere in the filesystem the user wouldn't be able
2429 * to delete without an idmapped mount.
2431 if (old_dir != dir && mnt_userns != &init_user_ns) {
2436 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2437 fs_info, vol_args2->subvolid);
2438 if (IS_ERR(subvol_name_ptr)) {
2439 err = PTR_ERR(subvol_name_ptr);
2442 /* subvol_name_ptr is already nul terminated */
2443 subvol_name = (char *)kbasename(subvol_name_ptr);
2446 vol_args = memdup_user(arg, sizeof(*vol_args));
2447 if (IS_ERR(vol_args))
2448 return PTR_ERR(vol_args);
2450 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2451 subvol_name = vol_args->name;
2453 err = mnt_want_write_file(file);
2458 subvol_namelen = strlen(subvol_name);
2460 if (strchr(subvol_name, '/') ||
2461 strncmp(subvol_name, "..", subvol_namelen) == 0) {
2463 goto free_subvol_name;
2466 if (!S_ISDIR(dir->i_mode)) {
2468 goto free_subvol_name;
2471 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2473 goto free_subvol_name;
2474 dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen);
2475 if (IS_ERR(dentry)) {
2476 err = PTR_ERR(dentry);
2477 goto out_unlock_dir;
2480 if (d_really_is_negative(dentry)) {
2485 inode = d_inode(dentry);
2486 dest = BTRFS_I(inode)->root;
2487 if (!capable(CAP_SYS_ADMIN)) {
2489 * Regular user. Only allow this with a special mount
2490 * option, when the user has write+exec access to the
2491 * subvol root, and when rmdir(2) would have been
2494 * Note that this is _not_ check that the subvol is
2495 * empty or doesn't contain data that we wouldn't
2496 * otherwise be able to delete.
2498 * Users who want to delete empty subvols should try
2502 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2506 * Do not allow deletion if the parent dir is the same
2507 * as the dir to be deleted. That means the ioctl
2508 * must be called on the dentry referencing the root
2509 * of the subvol, not a random directory contained
2516 err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC);
2521 /* check if subvolume may be deleted by a user */
2522 err = btrfs_may_delete(mnt_userns, dir, dentry, 1);
2526 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2531 btrfs_inode_lock(BTRFS_I(inode), 0);
2532 err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2533 btrfs_inode_unlock(BTRFS_I(inode), 0);
2535 d_delete_notify(dir, dentry);
2540 btrfs_inode_unlock(BTRFS_I(dir), 0);
2542 kfree(subvol_name_ptr);
2547 mnt_drop_write_file(file);
2554 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2556 struct inode *inode = file_inode(file);
2557 struct btrfs_root *root = BTRFS_I(inode)->root;
2558 struct btrfs_ioctl_defrag_range_args range = {0};
2561 ret = mnt_want_write_file(file);
2565 if (btrfs_root_readonly(root)) {
2570 switch (inode->i_mode & S_IFMT) {
2572 if (!capable(CAP_SYS_ADMIN)) {
2576 ret = btrfs_defrag_root(root);
2580 * Note that this does not check the file descriptor for write
2581 * access. This prevents defragmenting executables that are
2582 * running and allows defrag on files open in read-only mode.
2584 if (!capable(CAP_SYS_ADMIN) &&
2585 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
2591 if (copy_from_user(&range, argp, sizeof(range))) {
2595 /* compression requires us to start the IO */
2596 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2597 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2598 range.extent_thresh = (u32)-1;
2601 /* the rest are all set to zero by kzalloc */
2602 range.len = (u64)-1;
2604 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2605 &range, BTRFS_OLDEST_GENERATION, 0);
2613 mnt_drop_write_file(file);
2617 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2619 struct btrfs_ioctl_vol_args *vol_args;
2620 bool restore_op = false;
2623 if (!capable(CAP_SYS_ADMIN))
2626 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2627 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2631 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2632 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2633 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2636 * We can do the device add because we have a paused balanced,
2637 * change the exclusive op type and remember we should bring
2638 * back the paused balance
2640 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2641 btrfs_exclop_start_unlock(fs_info);
2645 vol_args = memdup_user(arg, sizeof(*vol_args));
2646 if (IS_ERR(vol_args)) {
2647 ret = PTR_ERR(vol_args);
2651 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2652 ret = btrfs_init_new_device(fs_info, vol_args->name);
2655 btrfs_info(fs_info, "disk added %s", vol_args->name);
2660 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2662 btrfs_exclop_finish(fs_info);
2666 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2668 BTRFS_DEV_LOOKUP_ARGS(args);
2669 struct inode *inode = file_inode(file);
2670 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2671 struct btrfs_ioctl_vol_args_v2 *vol_args;
2672 struct block_device *bdev = NULL;
2675 bool cancel = false;
2677 if (!capable(CAP_SYS_ADMIN))
2680 vol_args = memdup_user(arg, sizeof(*vol_args));
2681 if (IS_ERR(vol_args))
2682 return PTR_ERR(vol_args);
2684 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2689 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2690 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2691 args.devid = vol_args->devid;
2692 } else if (!strcmp("cancel", vol_args->name)) {
2695 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2700 ret = mnt_want_write_file(file);
2704 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2709 /* Exclusive operation is now claimed */
2710 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
2712 btrfs_exclop_finish(fs_info);
2715 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2716 btrfs_info(fs_info, "device deleted: id %llu",
2719 btrfs_info(fs_info, "device deleted: %s",
2723 mnt_drop_write_file(file);
2725 blkdev_put(bdev, mode);
2727 btrfs_put_dev_args_from_path(&args);
2732 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2734 BTRFS_DEV_LOOKUP_ARGS(args);
2735 struct inode *inode = file_inode(file);
2736 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2737 struct btrfs_ioctl_vol_args *vol_args;
2738 struct block_device *bdev = NULL;
2741 bool cancel = false;
2743 if (!capable(CAP_SYS_ADMIN))
2746 vol_args = memdup_user(arg, sizeof(*vol_args));
2747 if (IS_ERR(vol_args))
2748 return PTR_ERR(vol_args);
2750 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2751 if (!strcmp("cancel", vol_args->name)) {
2754 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2759 ret = mnt_want_write_file(file);
2763 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2766 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
2768 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2769 btrfs_exclop_finish(fs_info);
2772 mnt_drop_write_file(file);
2774 blkdev_put(bdev, mode);
2776 btrfs_put_dev_args_from_path(&args);
2781 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2784 struct btrfs_ioctl_fs_info_args *fi_args;
2785 struct btrfs_device *device;
2786 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2790 fi_args = memdup_user(arg, sizeof(*fi_args));
2791 if (IS_ERR(fi_args))
2792 return PTR_ERR(fi_args);
2794 flags_in = fi_args->flags;
2795 memset(fi_args, 0, sizeof(*fi_args));
2798 fi_args->num_devices = fs_devices->num_devices;
2800 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2801 if (device->devid > fi_args->max_id)
2802 fi_args->max_id = device->devid;
2806 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2807 fi_args->nodesize = fs_info->nodesize;
2808 fi_args->sectorsize = fs_info->sectorsize;
2809 fi_args->clone_alignment = fs_info->sectorsize;
2811 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2812 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2813 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2814 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2817 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2818 fi_args->generation = fs_info->generation;
2819 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2822 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2823 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2824 sizeof(fi_args->metadata_uuid));
2825 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2828 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2835 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2838 BTRFS_DEV_LOOKUP_ARGS(args);
2839 struct btrfs_ioctl_dev_info_args *di_args;
2840 struct btrfs_device *dev;
2843 di_args = memdup_user(arg, sizeof(*di_args));
2844 if (IS_ERR(di_args))
2845 return PTR_ERR(di_args);
2847 args.devid = di_args->devid;
2848 if (!btrfs_is_empty_uuid(di_args->uuid))
2849 args.uuid = di_args->uuid;
2852 dev = btrfs_find_device(fs_info->fs_devices, &args);
2858 di_args->devid = dev->devid;
2859 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2860 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2861 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2863 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2865 di_args->path[0] = '\0';
2869 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2876 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2878 struct inode *inode = file_inode(file);
2879 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2880 struct btrfs_root *root = BTRFS_I(inode)->root;
2881 struct btrfs_root *new_root;
2882 struct btrfs_dir_item *di;
2883 struct btrfs_trans_handle *trans;
2884 struct btrfs_path *path = NULL;
2885 struct btrfs_disk_key disk_key;
2886 struct fscrypt_str name = FSTR_INIT("default", 7);
2891 if (!capable(CAP_SYS_ADMIN))
2894 ret = mnt_want_write_file(file);
2898 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2904 objectid = BTRFS_FS_TREE_OBJECTID;
2906 new_root = btrfs_get_fs_root(fs_info, objectid, true);
2907 if (IS_ERR(new_root)) {
2908 ret = PTR_ERR(new_root);
2911 if (!is_fstree(new_root->root_key.objectid)) {
2916 path = btrfs_alloc_path();
2922 trans = btrfs_start_transaction(root, 1);
2923 if (IS_ERR(trans)) {
2924 ret = PTR_ERR(trans);
2928 dir_id = btrfs_super_root_dir(fs_info->super_copy);
2929 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2931 if (IS_ERR_OR_NULL(di)) {
2932 btrfs_release_path(path);
2933 btrfs_end_transaction(trans);
2935 "Umm, you don't have the default diritem, this isn't going to work");
2940 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2941 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2942 btrfs_mark_buffer_dirty(path->nodes[0]);
2943 btrfs_release_path(path);
2945 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2946 btrfs_end_transaction(trans);
2948 btrfs_put_root(new_root);
2949 btrfs_free_path(path);
2951 mnt_drop_write_file(file);
2955 static void get_block_group_info(struct list_head *groups_list,
2956 struct btrfs_ioctl_space_info *space)
2958 struct btrfs_block_group *block_group;
2960 space->total_bytes = 0;
2961 space->used_bytes = 0;
2963 list_for_each_entry(block_group, groups_list, list) {
2964 space->flags = block_group->flags;
2965 space->total_bytes += block_group->length;
2966 space->used_bytes += block_group->used;
2970 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
2973 struct btrfs_ioctl_space_args space_args;
2974 struct btrfs_ioctl_space_info space;
2975 struct btrfs_ioctl_space_info *dest;
2976 struct btrfs_ioctl_space_info *dest_orig;
2977 struct btrfs_ioctl_space_info __user *user_dest;
2978 struct btrfs_space_info *info;
2979 static const u64 types[] = {
2980 BTRFS_BLOCK_GROUP_DATA,
2981 BTRFS_BLOCK_GROUP_SYSTEM,
2982 BTRFS_BLOCK_GROUP_METADATA,
2983 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
2991 if (copy_from_user(&space_args,
2992 (struct btrfs_ioctl_space_args __user *)arg,
2993 sizeof(space_args)))
2996 for (i = 0; i < num_types; i++) {
2997 struct btrfs_space_info *tmp;
3000 list_for_each_entry(tmp, &fs_info->space_info, list) {
3001 if (tmp->flags == types[i]) {
3010 down_read(&info->groups_sem);
3011 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3012 if (!list_empty(&info->block_groups[c]))
3015 up_read(&info->groups_sem);
3019 * Global block reserve, exported as a space_info
3023 /* space_slots == 0 means they are asking for a count */
3024 if (space_args.space_slots == 0) {
3025 space_args.total_spaces = slot_count;
3029 slot_count = min_t(u64, space_args.space_slots, slot_count);
3031 alloc_size = sizeof(*dest) * slot_count;
3033 /* we generally have at most 6 or so space infos, one for each raid
3034 * level. So, a whole page should be more than enough for everyone
3036 if (alloc_size > PAGE_SIZE)
3039 space_args.total_spaces = 0;
3040 dest = kmalloc(alloc_size, GFP_KERNEL);
3045 /* now we have a buffer to copy into */
3046 for (i = 0; i < num_types; i++) {
3047 struct btrfs_space_info *tmp;
3053 list_for_each_entry(tmp, &fs_info->space_info, list) {
3054 if (tmp->flags == types[i]) {
3062 down_read(&info->groups_sem);
3063 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3064 if (!list_empty(&info->block_groups[c])) {
3065 get_block_group_info(&info->block_groups[c],
3067 memcpy(dest, &space, sizeof(space));
3069 space_args.total_spaces++;
3075 up_read(&info->groups_sem);
3079 * Add global block reserve
3082 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3084 spin_lock(&block_rsv->lock);
3085 space.total_bytes = block_rsv->size;
3086 space.used_bytes = block_rsv->size - block_rsv->reserved;
3087 spin_unlock(&block_rsv->lock);
3088 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3089 memcpy(dest, &space, sizeof(space));
3090 space_args.total_spaces++;
3093 user_dest = (struct btrfs_ioctl_space_info __user *)
3094 (arg + sizeof(struct btrfs_ioctl_space_args));
3096 if (copy_to_user(user_dest, dest_orig, alloc_size))
3101 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3107 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3110 struct btrfs_trans_handle *trans;
3113 trans = btrfs_attach_transaction_barrier(root);
3114 if (IS_ERR(trans)) {
3115 if (PTR_ERR(trans) != -ENOENT)
3116 return PTR_ERR(trans);
3118 /* No running transaction, don't bother */
3119 transid = root->fs_info->last_trans_committed;
3122 transid = trans->transid;
3123 btrfs_commit_transaction_async(trans);
3126 if (copy_to_user(argp, &transid, sizeof(transid)))
3131 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3137 if (copy_from_user(&transid, argp, sizeof(transid)))
3140 transid = 0; /* current trans */
3142 return btrfs_wait_for_commit(fs_info, transid);
3145 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3147 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3148 struct btrfs_ioctl_scrub_args *sa;
3151 if (!capable(CAP_SYS_ADMIN))
3154 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3155 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3159 sa = memdup_user(arg, sizeof(*sa));
3163 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3164 ret = mnt_want_write_file(file);
3169 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3170 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3174 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3175 * error. This is important as it allows user space to know how much
3176 * progress scrub has done. For example, if scrub is canceled we get
3177 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3178 * space. Later user space can inspect the progress from the structure
3179 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3180 * previously (btrfs-progs does this).
3181 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3182 * then return -EFAULT to signal the structure was not copied or it may
3183 * be corrupt and unreliable due to a partial copy.
3185 if (copy_to_user(arg, sa, sizeof(*sa)))
3188 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3189 mnt_drop_write_file(file);
3195 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3197 if (!capable(CAP_SYS_ADMIN))
3200 return btrfs_scrub_cancel(fs_info);
3203 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3206 struct btrfs_ioctl_scrub_args *sa;
3209 if (!capable(CAP_SYS_ADMIN))
3212 sa = memdup_user(arg, sizeof(*sa));
3216 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3218 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3225 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3228 struct btrfs_ioctl_get_dev_stats *sa;
3231 sa = memdup_user(arg, sizeof(*sa));
3235 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3240 ret = btrfs_get_dev_stats(fs_info, sa);
3242 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3249 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3252 struct btrfs_ioctl_dev_replace_args *p;
3255 if (!capable(CAP_SYS_ADMIN))
3258 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3259 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3263 p = memdup_user(arg, sizeof(*p));
3268 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3269 if (sb_rdonly(fs_info->sb)) {
3273 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3274 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3276 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3277 btrfs_exclop_finish(fs_info);
3280 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3281 btrfs_dev_replace_status(fs_info, p);
3284 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3285 p->result = btrfs_dev_replace_cancel(fs_info);
3293 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3300 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3306 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3307 struct inode_fs_paths *ipath = NULL;
3308 struct btrfs_path *path;
3310 if (!capable(CAP_DAC_READ_SEARCH))
3313 path = btrfs_alloc_path();
3319 ipa = memdup_user(arg, sizeof(*ipa));
3326 size = min_t(u32, ipa->size, 4096);
3327 ipath = init_ipath(size, root, path);
3328 if (IS_ERR(ipath)) {
3329 ret = PTR_ERR(ipath);
3334 ret = paths_from_inode(ipa->inum, ipath);
3338 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3339 rel_ptr = ipath->fspath->val[i] -
3340 (u64)(unsigned long)ipath->fspath->val;
3341 ipath->fspath->val[i] = rel_ptr;
3344 btrfs_free_path(path);
3346 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3347 ipath->fspath, size);
3354 btrfs_free_path(path);
3361 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3362 void __user *arg, int version)
3366 struct btrfs_ioctl_logical_ino_args *loi;
3367 struct btrfs_data_container *inodes = NULL;
3368 struct btrfs_path *path = NULL;
3371 if (!capable(CAP_SYS_ADMIN))
3374 loi = memdup_user(arg, sizeof(*loi));
3376 return PTR_ERR(loi);
3379 ignore_offset = false;
3380 size = min_t(u32, loi->size, SZ_64K);
3382 /* All reserved bits must be 0 for now */
3383 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3387 /* Only accept flags we have defined so far */
3388 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3392 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3393 size = min_t(u32, loi->size, SZ_16M);
3396 inodes = init_data_container(size);
3397 if (IS_ERR(inodes)) {
3398 ret = PTR_ERR(inodes);
3402 path = btrfs_alloc_path();
3407 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3408 inodes, ignore_offset);
3409 btrfs_free_path(path);
3415 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3428 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3429 struct btrfs_ioctl_balance_args *bargs)
3431 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3433 bargs->flags = bctl->flags;
3435 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3436 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3437 if (atomic_read(&fs_info->balance_pause_req))
3438 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3439 if (atomic_read(&fs_info->balance_cancel_req))
3440 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3442 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3443 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3444 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3446 spin_lock(&fs_info->balance_lock);
3447 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3448 spin_unlock(&fs_info->balance_lock);
3452 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3455 * @fs_info: the filesystem
3456 * @excl_acquired: ptr to boolean value which is set to false in case balance
3459 * Return 0 on success in which case both fs_info::balance is acquired as well
3460 * as exclusive ops are blocked. In case of failure return an error code.
3462 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3467 * Exclusive operation is locked. Three possibilities:
3468 * (1) some other op is running
3469 * (2) balance is running
3470 * (3) balance is paused -- special case (think resume)
3473 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3474 *excl_acquired = true;
3475 mutex_lock(&fs_info->balance_mutex);
3479 mutex_lock(&fs_info->balance_mutex);
3480 if (fs_info->balance_ctl) {
3481 /* This is either (2) or (3) */
3482 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3488 mutex_unlock(&fs_info->balance_mutex);
3490 * Lock released to allow other waiters to
3491 * continue, we'll reexamine the status again.
3493 mutex_lock(&fs_info->balance_mutex);
3495 if (fs_info->balance_ctl &&
3496 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3498 *excl_acquired = false;
3504 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3508 mutex_unlock(&fs_info->balance_mutex);
3512 mutex_unlock(&fs_info->balance_mutex);
3513 *excl_acquired = false;
3517 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3519 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3520 struct btrfs_fs_info *fs_info = root->fs_info;
3521 struct btrfs_ioctl_balance_args *bargs;
3522 struct btrfs_balance_control *bctl;
3523 bool need_unlock = true;
3526 if (!capable(CAP_SYS_ADMIN))
3529 ret = mnt_want_write_file(file);
3533 bargs = memdup_user(arg, sizeof(*bargs));
3534 if (IS_ERR(bargs)) {
3535 ret = PTR_ERR(bargs);
3540 ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3544 lockdep_assert_held(&fs_info->balance_mutex);
3546 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3547 if (!fs_info->balance_ctl) {
3552 bctl = fs_info->balance_ctl;
3553 spin_lock(&fs_info->balance_lock);
3554 bctl->flags |= BTRFS_BALANCE_RESUME;
3555 spin_unlock(&fs_info->balance_lock);
3556 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3561 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3566 if (fs_info->balance_ctl) {
3571 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3577 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3578 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3579 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3581 bctl->flags = bargs->flags;
3584 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3585 * bctl is freed in reset_balance_state, or, if restriper was paused
3586 * all the way until unmount, in free_fs_info. The flag should be
3587 * cleared after reset_balance_state.
3589 need_unlock = false;
3591 ret = btrfs_balance(fs_info, bctl, bargs);
3594 if (ret == 0 || ret == -ECANCELED) {
3595 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3601 mutex_unlock(&fs_info->balance_mutex);
3603 btrfs_exclop_finish(fs_info);
3605 mnt_drop_write_file(file);
3610 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3612 if (!capable(CAP_SYS_ADMIN))
3616 case BTRFS_BALANCE_CTL_PAUSE:
3617 return btrfs_pause_balance(fs_info);
3618 case BTRFS_BALANCE_CTL_CANCEL:
3619 return btrfs_cancel_balance(fs_info);
3625 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3628 struct btrfs_ioctl_balance_args *bargs;
3631 if (!capable(CAP_SYS_ADMIN))
3634 mutex_lock(&fs_info->balance_mutex);
3635 if (!fs_info->balance_ctl) {
3640 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3646 btrfs_update_ioctl_balance_args(fs_info, bargs);
3648 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3653 mutex_unlock(&fs_info->balance_mutex);
3657 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3659 struct inode *inode = file_inode(file);
3660 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3661 struct btrfs_ioctl_quota_ctl_args *sa;
3664 if (!capable(CAP_SYS_ADMIN))
3667 ret = mnt_want_write_file(file);
3671 sa = memdup_user(arg, sizeof(*sa));
3677 down_write(&fs_info->subvol_sem);
3680 case BTRFS_QUOTA_CTL_ENABLE:
3681 ret = btrfs_quota_enable(fs_info);
3683 case BTRFS_QUOTA_CTL_DISABLE:
3684 ret = btrfs_quota_disable(fs_info);
3692 up_write(&fs_info->subvol_sem);
3694 mnt_drop_write_file(file);
3698 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3700 struct inode *inode = file_inode(file);
3701 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3702 struct btrfs_root *root = BTRFS_I(inode)->root;
3703 struct btrfs_ioctl_qgroup_assign_args *sa;
3704 struct btrfs_trans_handle *trans;
3708 if (!capable(CAP_SYS_ADMIN))
3711 ret = mnt_want_write_file(file);
3715 sa = memdup_user(arg, sizeof(*sa));
3721 trans = btrfs_join_transaction(root);
3722 if (IS_ERR(trans)) {
3723 ret = PTR_ERR(trans);
3728 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3730 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3733 /* update qgroup status and info */
3734 err = btrfs_run_qgroups(trans);
3736 btrfs_handle_fs_error(fs_info, err,
3737 "failed to update qgroup status and info");
3738 err = btrfs_end_transaction(trans);
3745 mnt_drop_write_file(file);
3749 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3751 struct inode *inode = file_inode(file);
3752 struct btrfs_root *root = BTRFS_I(inode)->root;
3753 struct btrfs_ioctl_qgroup_create_args *sa;
3754 struct btrfs_trans_handle *trans;
3758 if (!capable(CAP_SYS_ADMIN))
3761 ret = mnt_want_write_file(file);
3765 sa = memdup_user(arg, sizeof(*sa));
3771 if (!sa->qgroupid) {
3776 trans = btrfs_join_transaction(root);
3777 if (IS_ERR(trans)) {
3778 ret = PTR_ERR(trans);
3783 ret = btrfs_create_qgroup(trans, sa->qgroupid);
3785 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3788 err = btrfs_end_transaction(trans);
3795 mnt_drop_write_file(file);
3799 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3801 struct inode *inode = file_inode(file);
3802 struct btrfs_root *root = BTRFS_I(inode)->root;
3803 struct btrfs_ioctl_qgroup_limit_args *sa;
3804 struct btrfs_trans_handle *trans;
3809 if (!capable(CAP_SYS_ADMIN))
3812 ret = mnt_want_write_file(file);
3816 sa = memdup_user(arg, sizeof(*sa));
3822 trans = btrfs_join_transaction(root);
3823 if (IS_ERR(trans)) {
3824 ret = PTR_ERR(trans);
3828 qgroupid = sa->qgroupid;
3830 /* take the current subvol as qgroup */
3831 qgroupid = root->root_key.objectid;
3834 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3836 err = btrfs_end_transaction(trans);
3843 mnt_drop_write_file(file);
3847 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3849 struct inode *inode = file_inode(file);
3850 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3851 struct btrfs_ioctl_quota_rescan_args *qsa;
3854 if (!capable(CAP_SYS_ADMIN))
3857 ret = mnt_want_write_file(file);
3861 qsa = memdup_user(arg, sizeof(*qsa));
3872 ret = btrfs_qgroup_rescan(fs_info);
3877 mnt_drop_write_file(file);
3881 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3884 struct btrfs_ioctl_quota_rescan_args qsa = {0};
3886 if (!capable(CAP_SYS_ADMIN))
3889 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3891 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3894 if (copy_to_user(arg, &qsa, sizeof(qsa)))
3900 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3903 if (!capable(CAP_SYS_ADMIN))
3906 return btrfs_qgroup_wait_for_completion(fs_info, true);
3909 static long _btrfs_ioctl_set_received_subvol(struct file *file,
3910 struct user_namespace *mnt_userns,
3911 struct btrfs_ioctl_received_subvol_args *sa)
3913 struct inode *inode = file_inode(file);
3914 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3915 struct btrfs_root *root = BTRFS_I(inode)->root;
3916 struct btrfs_root_item *root_item = &root->root_item;
3917 struct btrfs_trans_handle *trans;
3918 struct timespec64 ct = current_time(inode);
3920 int received_uuid_changed;
3922 if (!inode_owner_or_capable(mnt_userns, inode))
3925 ret = mnt_want_write_file(file);
3929 down_write(&fs_info->subvol_sem);
3931 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3936 if (btrfs_root_readonly(root)) {
3943 * 2 - uuid items (received uuid + subvol uuid)
3945 trans = btrfs_start_transaction(root, 3);
3946 if (IS_ERR(trans)) {
3947 ret = PTR_ERR(trans);
3952 sa->rtransid = trans->transid;
3953 sa->rtime.sec = ct.tv_sec;
3954 sa->rtime.nsec = ct.tv_nsec;
3956 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
3958 if (received_uuid_changed &&
3959 !btrfs_is_empty_uuid(root_item->received_uuid)) {
3960 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
3961 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
3962 root->root_key.objectid);
3963 if (ret && ret != -ENOENT) {
3964 btrfs_abort_transaction(trans, ret);
3965 btrfs_end_transaction(trans);
3969 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3970 btrfs_set_root_stransid(root_item, sa->stransid);
3971 btrfs_set_root_rtransid(root_item, sa->rtransid);
3972 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
3973 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
3974 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
3975 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
3977 ret = btrfs_update_root(trans, fs_info->tree_root,
3978 &root->root_key, &root->root_item);
3980 btrfs_end_transaction(trans);
3983 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
3984 ret = btrfs_uuid_tree_add(trans, sa->uuid,
3985 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
3986 root->root_key.objectid);
3987 if (ret < 0 && ret != -EEXIST) {
3988 btrfs_abort_transaction(trans, ret);
3989 btrfs_end_transaction(trans);
3993 ret = btrfs_commit_transaction(trans);
3995 up_write(&fs_info->subvol_sem);
3996 mnt_drop_write_file(file);
4001 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4004 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4005 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4008 args32 = memdup_user(arg, sizeof(*args32));
4010 return PTR_ERR(args32);
4012 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4018 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4019 args64->stransid = args32->stransid;
4020 args64->rtransid = args32->rtransid;
4021 args64->stime.sec = args32->stime.sec;
4022 args64->stime.nsec = args32->stime.nsec;
4023 args64->rtime.sec = args32->rtime.sec;
4024 args64->rtime.nsec = args32->rtime.nsec;
4025 args64->flags = args32->flags;
4027 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64);
4031 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4032 args32->stransid = args64->stransid;
4033 args32->rtransid = args64->rtransid;
4034 args32->stime.sec = args64->stime.sec;
4035 args32->stime.nsec = args64->stime.nsec;
4036 args32->rtime.sec = args64->rtime.sec;
4037 args32->rtime.nsec = args64->rtime.nsec;
4038 args32->flags = args64->flags;
4040 ret = copy_to_user(arg, args32, sizeof(*args32));
4051 static long btrfs_ioctl_set_received_subvol(struct file *file,
4054 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4057 sa = memdup_user(arg, sizeof(*sa));
4061 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa);
4066 ret = copy_to_user(arg, sa, sizeof(*sa));
4075 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4080 char label[BTRFS_LABEL_SIZE];
4082 spin_lock(&fs_info->super_lock);
4083 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4084 spin_unlock(&fs_info->super_lock);
4086 len = strnlen(label, BTRFS_LABEL_SIZE);
4088 if (len == BTRFS_LABEL_SIZE) {
4090 "label is too long, return the first %zu bytes",
4094 ret = copy_to_user(arg, label, len);
4096 return ret ? -EFAULT : 0;
4099 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4101 struct inode *inode = file_inode(file);
4102 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4103 struct btrfs_root *root = BTRFS_I(inode)->root;
4104 struct btrfs_super_block *super_block = fs_info->super_copy;
4105 struct btrfs_trans_handle *trans;
4106 char label[BTRFS_LABEL_SIZE];
4109 if (!capable(CAP_SYS_ADMIN))
4112 if (copy_from_user(label, arg, sizeof(label)))
4115 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4117 "unable to set label with more than %d bytes",
4118 BTRFS_LABEL_SIZE - 1);
4122 ret = mnt_want_write_file(file);
4126 trans = btrfs_start_transaction(root, 0);
4127 if (IS_ERR(trans)) {
4128 ret = PTR_ERR(trans);
4132 spin_lock(&fs_info->super_lock);
4133 strcpy(super_block->label, label);
4134 spin_unlock(&fs_info->super_lock);
4135 ret = btrfs_commit_transaction(trans);
4138 mnt_drop_write_file(file);
4142 #define INIT_FEATURE_FLAGS(suffix) \
4143 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4144 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4145 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4147 int btrfs_ioctl_get_supported_features(void __user *arg)
4149 static const struct btrfs_ioctl_feature_flags features[3] = {
4150 INIT_FEATURE_FLAGS(SUPP),
4151 INIT_FEATURE_FLAGS(SAFE_SET),
4152 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4155 if (copy_to_user(arg, &features, sizeof(features)))
4161 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4164 struct btrfs_super_block *super_block = fs_info->super_copy;
4165 struct btrfs_ioctl_feature_flags features;
4167 features.compat_flags = btrfs_super_compat_flags(super_block);
4168 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4169 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4171 if (copy_to_user(arg, &features, sizeof(features)))
4177 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4178 enum btrfs_feature_set set,
4179 u64 change_mask, u64 flags, u64 supported_flags,
4180 u64 safe_set, u64 safe_clear)
4182 const char *type = btrfs_feature_set_name(set);
4184 u64 disallowed, unsupported;
4185 u64 set_mask = flags & change_mask;
4186 u64 clear_mask = ~flags & change_mask;
4188 unsupported = set_mask & ~supported_flags;
4190 names = btrfs_printable_features(set, unsupported);
4193 "this kernel does not support the %s feature bit%s",
4194 names, strchr(names, ',') ? "s" : "");
4198 "this kernel does not support %s bits 0x%llx",
4203 disallowed = set_mask & ~safe_set;
4205 names = btrfs_printable_features(set, disallowed);
4208 "can't set the %s feature bit%s while mounted",
4209 names, strchr(names, ',') ? "s" : "");
4213 "can't set %s bits 0x%llx while mounted",
4218 disallowed = clear_mask & ~safe_clear;
4220 names = btrfs_printable_features(set, disallowed);
4223 "can't clear the %s feature bit%s while mounted",
4224 names, strchr(names, ',') ? "s" : "");
4228 "can't clear %s bits 0x%llx while mounted",
4236 #define check_feature(fs_info, change_mask, flags, mask_base) \
4237 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4238 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4239 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4240 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4242 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4244 struct inode *inode = file_inode(file);
4245 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4246 struct btrfs_root *root = BTRFS_I(inode)->root;
4247 struct btrfs_super_block *super_block = fs_info->super_copy;
4248 struct btrfs_ioctl_feature_flags flags[2];
4249 struct btrfs_trans_handle *trans;
4253 if (!capable(CAP_SYS_ADMIN))
4256 if (copy_from_user(flags, arg, sizeof(flags)))
4260 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4261 !flags[0].incompat_flags)
4264 ret = check_feature(fs_info, flags[0].compat_flags,
4265 flags[1].compat_flags, COMPAT);
4269 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4270 flags[1].compat_ro_flags, COMPAT_RO);
4274 ret = check_feature(fs_info, flags[0].incompat_flags,
4275 flags[1].incompat_flags, INCOMPAT);
4279 ret = mnt_want_write_file(file);
4283 trans = btrfs_start_transaction(root, 0);
4284 if (IS_ERR(trans)) {
4285 ret = PTR_ERR(trans);
4286 goto out_drop_write;
4289 spin_lock(&fs_info->super_lock);
4290 newflags = btrfs_super_compat_flags(super_block);
4291 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4292 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4293 btrfs_set_super_compat_flags(super_block, newflags);
4295 newflags = btrfs_super_compat_ro_flags(super_block);
4296 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4297 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4298 btrfs_set_super_compat_ro_flags(super_block, newflags);
4300 newflags = btrfs_super_incompat_flags(super_block);
4301 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4302 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4303 btrfs_set_super_incompat_flags(super_block, newflags);
4304 spin_unlock(&fs_info->super_lock);
4306 ret = btrfs_commit_transaction(trans);
4308 mnt_drop_write_file(file);
4313 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4315 struct btrfs_ioctl_send_args *arg;
4319 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4320 struct btrfs_ioctl_send_args_32 args32;
4322 ret = copy_from_user(&args32, argp, sizeof(args32));
4325 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4328 arg->send_fd = args32.send_fd;
4329 arg->clone_sources_count = args32.clone_sources_count;
4330 arg->clone_sources = compat_ptr(args32.clone_sources);
4331 arg->parent_root = args32.parent_root;
4332 arg->flags = args32.flags;
4333 memcpy(arg->reserved, args32.reserved,
4334 sizeof(args32.reserved));
4339 arg = memdup_user(argp, sizeof(*arg));
4341 return PTR_ERR(arg);
4343 ret = btrfs_ioctl_send(inode, arg);
4348 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4351 struct btrfs_ioctl_encoded_io_args args = { 0 };
4352 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4355 struct iovec iovstack[UIO_FASTIOV];
4356 struct iovec *iov = iovstack;
4357 struct iov_iter iter;
4362 if (!capable(CAP_SYS_ADMIN)) {
4368 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4369 struct btrfs_ioctl_encoded_io_args_32 args32;
4371 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4373 if (copy_from_user(&args32, argp, copy_end)) {
4377 args.iov = compat_ptr(args32.iov);
4378 args.iovcnt = args32.iovcnt;
4379 args.offset = args32.offset;
4380 args.flags = args32.flags;
4385 copy_end = copy_end_kernel;
4386 if (copy_from_user(&args, argp, copy_end)) {
4391 if (args.flags != 0) {
4396 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4401 if (iov_iter_count(&iter) == 0) {
4406 ret = rw_verify_area(READ, file, &pos, args.len);
4410 init_sync_kiocb(&kiocb, file);
4413 ret = btrfs_encoded_read(&kiocb, &iter, &args);
4415 fsnotify_access(file);
4416 if (copy_to_user(argp + copy_end,
4417 (char *)&args + copy_end_kernel,
4418 sizeof(args) - copy_end_kernel))
4426 add_rchar(current, ret);
4431 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4433 struct btrfs_ioctl_encoded_io_args args;
4434 struct iovec iovstack[UIO_FASTIOV];
4435 struct iovec *iov = iovstack;
4436 struct iov_iter iter;
4441 if (!capable(CAP_SYS_ADMIN)) {
4446 if (!(file->f_mode & FMODE_WRITE)) {
4452 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4453 struct btrfs_ioctl_encoded_io_args_32 args32;
4455 if (copy_from_user(&args32, argp, sizeof(args32))) {
4459 args.iov = compat_ptr(args32.iov);
4460 args.iovcnt = args32.iovcnt;
4461 args.offset = args32.offset;
4462 args.flags = args32.flags;
4463 args.len = args32.len;
4464 args.unencoded_len = args32.unencoded_len;
4465 args.unencoded_offset = args32.unencoded_offset;
4466 args.compression = args32.compression;
4467 args.encryption = args32.encryption;
4468 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4473 if (copy_from_user(&args, argp, sizeof(args))) {
4480 if (args.flags != 0)
4482 if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4484 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4485 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4487 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4488 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4490 if (args.unencoded_offset > args.unencoded_len)
4492 if (args.len > args.unencoded_len - args.unencoded_offset)
4495 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4500 file_start_write(file);
4502 if (iov_iter_count(&iter) == 0) {
4507 ret = rw_verify_area(WRITE, file, &pos, args.len);
4511 init_sync_kiocb(&kiocb, file);
4512 ret = kiocb_set_rw_flags(&kiocb, 0);
4517 ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4519 fsnotify_modify(file);
4522 file_end_write(file);
4526 add_wchar(current, ret);
4531 long btrfs_ioctl(struct file *file, unsigned int
4532 cmd, unsigned long arg)
4534 struct inode *inode = file_inode(file);
4535 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4536 struct btrfs_root *root = BTRFS_I(inode)->root;
4537 void __user *argp = (void __user *)arg;
4540 case FS_IOC_GETVERSION:
4541 return btrfs_ioctl_getversion(inode, argp);
4542 case FS_IOC_GETFSLABEL:
4543 return btrfs_ioctl_get_fslabel(fs_info, argp);
4544 case FS_IOC_SETFSLABEL:
4545 return btrfs_ioctl_set_fslabel(file, argp);
4547 return btrfs_ioctl_fitrim(fs_info, argp);
4548 case BTRFS_IOC_SNAP_CREATE:
4549 return btrfs_ioctl_snap_create(file, argp, 0);
4550 case BTRFS_IOC_SNAP_CREATE_V2:
4551 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4552 case BTRFS_IOC_SUBVOL_CREATE:
4553 return btrfs_ioctl_snap_create(file, argp, 1);
4554 case BTRFS_IOC_SUBVOL_CREATE_V2:
4555 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4556 case BTRFS_IOC_SNAP_DESTROY:
4557 return btrfs_ioctl_snap_destroy(file, argp, false);
4558 case BTRFS_IOC_SNAP_DESTROY_V2:
4559 return btrfs_ioctl_snap_destroy(file, argp, true);
4560 case BTRFS_IOC_SUBVOL_GETFLAGS:
4561 return btrfs_ioctl_subvol_getflags(inode, argp);
4562 case BTRFS_IOC_SUBVOL_SETFLAGS:
4563 return btrfs_ioctl_subvol_setflags(file, argp);
4564 case BTRFS_IOC_DEFAULT_SUBVOL:
4565 return btrfs_ioctl_default_subvol(file, argp);
4566 case BTRFS_IOC_DEFRAG:
4567 return btrfs_ioctl_defrag(file, NULL);
4568 case BTRFS_IOC_DEFRAG_RANGE:
4569 return btrfs_ioctl_defrag(file, argp);
4570 case BTRFS_IOC_RESIZE:
4571 return btrfs_ioctl_resize(file, argp);
4572 case BTRFS_IOC_ADD_DEV:
4573 return btrfs_ioctl_add_dev(fs_info, argp);
4574 case BTRFS_IOC_RM_DEV:
4575 return btrfs_ioctl_rm_dev(file, argp);
4576 case BTRFS_IOC_RM_DEV_V2:
4577 return btrfs_ioctl_rm_dev_v2(file, argp);
4578 case BTRFS_IOC_FS_INFO:
4579 return btrfs_ioctl_fs_info(fs_info, argp);
4580 case BTRFS_IOC_DEV_INFO:
4581 return btrfs_ioctl_dev_info(fs_info, argp);
4582 case BTRFS_IOC_TREE_SEARCH:
4583 return btrfs_ioctl_tree_search(inode, argp);
4584 case BTRFS_IOC_TREE_SEARCH_V2:
4585 return btrfs_ioctl_tree_search_v2(inode, argp);
4586 case BTRFS_IOC_INO_LOOKUP:
4587 return btrfs_ioctl_ino_lookup(root, argp);
4588 case BTRFS_IOC_INO_PATHS:
4589 return btrfs_ioctl_ino_to_path(root, argp);
4590 case BTRFS_IOC_LOGICAL_INO:
4591 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4592 case BTRFS_IOC_LOGICAL_INO_V2:
4593 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4594 case BTRFS_IOC_SPACE_INFO:
4595 return btrfs_ioctl_space_info(fs_info, argp);
4596 case BTRFS_IOC_SYNC: {
4599 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4602 ret = btrfs_sync_fs(inode->i_sb, 1);
4604 * The transaction thread may want to do more work,
4605 * namely it pokes the cleaner kthread that will start
4606 * processing uncleaned subvols.
4608 wake_up_process(fs_info->transaction_kthread);
4611 case BTRFS_IOC_START_SYNC:
4612 return btrfs_ioctl_start_sync(root, argp);
4613 case BTRFS_IOC_WAIT_SYNC:
4614 return btrfs_ioctl_wait_sync(fs_info, argp);
4615 case BTRFS_IOC_SCRUB:
4616 return btrfs_ioctl_scrub(file, argp);
4617 case BTRFS_IOC_SCRUB_CANCEL:
4618 return btrfs_ioctl_scrub_cancel(fs_info);
4619 case BTRFS_IOC_SCRUB_PROGRESS:
4620 return btrfs_ioctl_scrub_progress(fs_info, argp);
4621 case BTRFS_IOC_BALANCE_V2:
4622 return btrfs_ioctl_balance(file, argp);
4623 case BTRFS_IOC_BALANCE_CTL:
4624 return btrfs_ioctl_balance_ctl(fs_info, arg);
4625 case BTRFS_IOC_BALANCE_PROGRESS:
4626 return btrfs_ioctl_balance_progress(fs_info, argp);
4627 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4628 return btrfs_ioctl_set_received_subvol(file, argp);
4630 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4631 return btrfs_ioctl_set_received_subvol_32(file, argp);
4633 case BTRFS_IOC_SEND:
4634 return _btrfs_ioctl_send(inode, argp, false);
4635 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4636 case BTRFS_IOC_SEND_32:
4637 return _btrfs_ioctl_send(inode, argp, true);
4639 case BTRFS_IOC_GET_DEV_STATS:
4640 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4641 case BTRFS_IOC_QUOTA_CTL:
4642 return btrfs_ioctl_quota_ctl(file, argp);
4643 case BTRFS_IOC_QGROUP_ASSIGN:
4644 return btrfs_ioctl_qgroup_assign(file, argp);
4645 case BTRFS_IOC_QGROUP_CREATE:
4646 return btrfs_ioctl_qgroup_create(file, argp);
4647 case BTRFS_IOC_QGROUP_LIMIT:
4648 return btrfs_ioctl_qgroup_limit(file, argp);
4649 case BTRFS_IOC_QUOTA_RESCAN:
4650 return btrfs_ioctl_quota_rescan(file, argp);
4651 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4652 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4653 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4654 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4655 case BTRFS_IOC_DEV_REPLACE:
4656 return btrfs_ioctl_dev_replace(fs_info, argp);
4657 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4658 return btrfs_ioctl_get_supported_features(argp);
4659 case BTRFS_IOC_GET_FEATURES:
4660 return btrfs_ioctl_get_features(fs_info, argp);
4661 case BTRFS_IOC_SET_FEATURES:
4662 return btrfs_ioctl_set_features(file, argp);
4663 case BTRFS_IOC_GET_SUBVOL_INFO:
4664 return btrfs_ioctl_get_subvol_info(inode, argp);
4665 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4666 return btrfs_ioctl_get_subvol_rootref(root, argp);
4667 case BTRFS_IOC_INO_LOOKUP_USER:
4668 return btrfs_ioctl_ino_lookup_user(file, argp);
4669 case FS_IOC_ENABLE_VERITY:
4670 return fsverity_ioctl_enable(file, (const void __user *)argp);
4671 case FS_IOC_MEASURE_VERITY:
4672 return fsverity_ioctl_measure(file, argp);
4673 case BTRFS_IOC_ENCODED_READ:
4674 return btrfs_ioctl_encoded_read(file, argp, false);
4675 case BTRFS_IOC_ENCODED_WRITE:
4676 return btrfs_ioctl_encoded_write(file, argp, false);
4677 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4678 case BTRFS_IOC_ENCODED_READ_32:
4679 return btrfs_ioctl_encoded_read(file, argp, true);
4680 case BTRFS_IOC_ENCODED_WRITE_32:
4681 return btrfs_ioctl_encoded_write(file, argp, true);
4688 #ifdef CONFIG_COMPAT
4689 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4692 * These all access 32-bit values anyway so no further
4693 * handling is necessary.
4696 case FS_IOC32_GETVERSION:
4697 cmd = FS_IOC_GETVERSION;
4701 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));