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 mnt_idmap *idmap,
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_set_ctime_current(inode);
388 ret = btrfs_update_inode(trans, 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_NONE ||
459 fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
460 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
461 spin_unlock(&fs_info->super_lock);
463 case BTRFS_EXCLOP_BALANCE:
464 spin_lock(&fs_info->super_lock);
465 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
466 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
467 spin_unlock(&fs_info->super_lock);
471 "invalid exclop balance operation %d requested", op);
475 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
477 return put_user(inode->i_generation, arg);
480 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
483 struct btrfs_device *device;
484 struct fstrim_range range;
485 u64 minlen = ULLONG_MAX;
489 if (!capable(CAP_SYS_ADMIN))
493 * btrfs_trim_block_group() depends on space cache, which is not
494 * available in zoned filesystem. So, disallow fitrim on a zoned
495 * filesystem for now.
497 if (btrfs_is_zoned(fs_info))
501 * If the fs is mounted with nologreplay, which requires it to be
502 * mounted in RO mode as well, we can not allow discard on free space
503 * inside block groups, because log trees refer to extents that are not
504 * pinned in a block group's free space cache (pinning the extents is
505 * precisely the first phase of replaying a log tree).
507 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
511 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
513 if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
516 minlen = min_t(u64, bdev_discard_granularity(device->bdev),
523 if (copy_from_user(&range, arg, sizeof(range)))
527 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
528 * block group is in the logical address space, which can be any
529 * sectorsize aligned bytenr in the range [0, U64_MAX].
531 if (range.len < fs_info->sb->s_blocksize)
534 range.minlen = max(range.minlen, minlen);
535 ret = btrfs_trim_fs(fs_info, &range);
539 if (copy_to_user(arg, &range, sizeof(range)))
545 int __pure btrfs_is_empty_uuid(u8 *uuid)
549 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
557 * Calculate the number of transaction items to reserve for creating a subvolume
558 * or snapshot, not including the inode, directory entries, or parent directory.
560 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
563 * 1 to add root block
566 * 1 to add root backref
568 * 1 to add qgroup info
569 * 1 to add qgroup limit
571 * Ideally the last two would only be accounted if qgroups are enabled,
572 * but that can change between now and the time we would insert them.
574 unsigned int num_items = 7;
577 /* 2 to add qgroup relations for each inherited qgroup */
578 num_items += 2 * inherit->num_qgroups;
583 static noinline int create_subvol(struct mnt_idmap *idmap,
584 struct inode *dir, struct dentry *dentry,
585 struct btrfs_qgroup_inherit *inherit)
587 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
588 struct btrfs_trans_handle *trans;
589 struct btrfs_key key;
590 struct btrfs_root_item *root_item;
591 struct btrfs_inode_item *inode_item;
592 struct extent_buffer *leaf;
593 struct btrfs_root *root = BTRFS_I(dir)->root;
594 struct btrfs_root *new_root;
595 struct btrfs_block_rsv block_rsv;
596 struct timespec64 cur_time = current_time(dir);
597 struct btrfs_new_inode_args new_inode_args = {
602 unsigned int trans_num_items;
607 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
611 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
616 * Don't create subvolume whose level is not zero. Or qgroup will be
617 * screwed up since it assumes subvolume qgroup's level to be 0.
619 if (btrfs_qgroup_level(objectid)) {
624 ret = get_anon_bdev(&anon_dev);
628 new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
629 if (!new_inode_args.inode) {
633 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
636 trans_num_items += create_subvol_num_items(inherit);
638 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
639 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
640 trans_num_items, false);
642 goto out_new_inode_args;
644 trans = btrfs_start_transaction(root, 0);
646 ret = PTR_ERR(trans);
647 btrfs_subvolume_release_metadata(root, &block_rsv);
648 goto out_new_inode_args;
650 trans->block_rsv = &block_rsv;
651 trans->bytes_reserved = block_rsv.size;
652 /* Tree log can't currently deal with an inode which is a new root. */
653 btrfs_set_log_full_commit(trans);
655 ret = btrfs_qgroup_inherit(trans, 0, objectid, root->root_key.objectid, inherit);
659 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
660 0, BTRFS_NESTING_NORMAL);
666 btrfs_mark_buffer_dirty(trans, leaf);
668 inode_item = &root_item->inode;
669 btrfs_set_stack_inode_generation(inode_item, 1);
670 btrfs_set_stack_inode_size(inode_item, 3);
671 btrfs_set_stack_inode_nlink(inode_item, 1);
672 btrfs_set_stack_inode_nbytes(inode_item,
674 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
676 btrfs_set_root_flags(root_item, 0);
677 btrfs_set_root_limit(root_item, 0);
678 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
680 btrfs_set_root_bytenr(root_item, leaf->start);
681 btrfs_set_root_generation(root_item, trans->transid);
682 btrfs_set_root_level(root_item, 0);
683 btrfs_set_root_refs(root_item, 1);
684 btrfs_set_root_used(root_item, leaf->len);
685 btrfs_set_root_last_snapshot(root_item, 0);
687 btrfs_set_root_generation_v2(root_item,
688 btrfs_root_generation(root_item));
689 generate_random_guid(root_item->uuid);
690 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
691 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
692 root_item->ctime = root_item->otime;
693 btrfs_set_root_ctransid(root_item, trans->transid);
694 btrfs_set_root_otransid(root_item, trans->transid);
696 btrfs_tree_unlock(leaf);
698 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
700 key.objectid = objectid;
702 key.type = BTRFS_ROOT_ITEM_KEY;
703 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
707 * Since we don't abort the transaction in this case, free the
708 * tree block so that we don't leak space and leave the
709 * filesystem in an inconsistent state (an extent item in the
710 * extent tree with a backreference for a root that does not
713 btrfs_tree_lock(leaf);
714 btrfs_clear_buffer_dirty(trans, leaf);
715 btrfs_tree_unlock(leaf);
716 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
717 free_extent_buffer(leaf);
721 free_extent_buffer(leaf);
724 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
725 if (IS_ERR(new_root)) {
726 ret = PTR_ERR(new_root);
727 btrfs_abort_transaction(trans, ret);
730 /* anon_dev is owned by new_root now. */
732 BTRFS_I(new_inode_args.inode)->root = new_root;
733 /* ... and new_root is owned by new_inode_args.inode now. */
735 ret = btrfs_record_root_in_trans(trans, new_root);
737 btrfs_abort_transaction(trans, ret);
741 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
742 BTRFS_UUID_KEY_SUBVOL, objectid);
744 btrfs_abort_transaction(trans, ret);
748 ret = btrfs_create_new_inode(trans, &new_inode_args);
750 btrfs_abort_transaction(trans, ret);
754 d_instantiate_new(dentry, new_inode_args.inode);
755 new_inode_args.inode = NULL;
758 trans->block_rsv = NULL;
759 trans->bytes_reserved = 0;
760 btrfs_subvolume_release_metadata(root, &block_rsv);
762 btrfs_end_transaction(trans);
764 btrfs_new_inode_args_destroy(&new_inode_args);
766 iput(new_inode_args.inode);
769 free_anon_bdev(anon_dev);
775 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
776 struct dentry *dentry, bool readonly,
777 struct btrfs_qgroup_inherit *inherit)
779 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
781 struct btrfs_pending_snapshot *pending_snapshot;
782 unsigned int trans_num_items;
783 struct btrfs_trans_handle *trans;
786 /* We do not support snapshotting right now. */
787 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
789 "extent tree v2 doesn't support snapshotting yet");
793 if (btrfs_root_refs(&root->root_item) == 0)
796 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
799 if (atomic_read(&root->nr_swapfiles)) {
801 "cannot snapshot subvolume with active swapfile");
805 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
806 if (!pending_snapshot)
809 ret = get_anon_bdev(&pending_snapshot->anon_dev);
812 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
814 pending_snapshot->path = btrfs_alloc_path();
815 if (!pending_snapshot->root_item || !pending_snapshot->path) {
820 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
821 BTRFS_BLOCK_RSV_TEMP);
825 * 1 to update parent inode item
827 trans_num_items = create_subvol_num_items(inherit) + 3;
828 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
829 &pending_snapshot->block_rsv,
830 trans_num_items, false);
834 pending_snapshot->dentry = dentry;
835 pending_snapshot->root = root;
836 pending_snapshot->readonly = readonly;
837 pending_snapshot->dir = dir;
838 pending_snapshot->inherit = inherit;
840 trans = btrfs_start_transaction(root, 0);
842 ret = PTR_ERR(trans);
846 trans->pending_snapshot = pending_snapshot;
848 ret = btrfs_commit_transaction(trans);
852 ret = pending_snapshot->error;
856 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
860 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
862 ret = PTR_ERR(inode);
866 d_instantiate(dentry, inode);
868 pending_snapshot->anon_dev = 0;
870 /* Prevent double freeing of anon_dev */
871 if (ret && pending_snapshot->snap)
872 pending_snapshot->snap->anon_dev = 0;
873 btrfs_put_root(pending_snapshot->snap);
874 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
876 if (pending_snapshot->anon_dev)
877 free_anon_bdev(pending_snapshot->anon_dev);
878 kfree(pending_snapshot->root_item);
879 btrfs_free_path(pending_snapshot->path);
880 kfree(pending_snapshot);
885 /* copy of may_delete in fs/namei.c()
886 * Check whether we can remove a link victim from directory dir, check
887 * whether the type of victim is right.
888 * 1. We can't do it if dir is read-only (done in permission())
889 * 2. We should have write and exec permissions on dir
890 * 3. We can't remove anything from append-only dir
891 * 4. We can't do anything with immutable dir (done in permission())
892 * 5. If the sticky bit on dir is set we should either
893 * a. be owner of dir, or
894 * b. be owner of victim, or
895 * c. have CAP_FOWNER capability
896 * 6. If the victim is append-only or immutable we can't do anything with
897 * links pointing to it.
898 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
899 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
900 * 9. We can't remove a root or mountpoint.
901 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
902 * nfs_async_unlink().
905 static int btrfs_may_delete(struct mnt_idmap *idmap,
906 struct inode *dir, struct dentry *victim, int isdir)
910 if (d_really_is_negative(victim))
913 BUG_ON(d_inode(victim->d_parent) != dir);
914 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
916 error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
921 if (check_sticky(idmap, dir, d_inode(victim)) ||
922 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
923 IS_SWAPFILE(d_inode(victim)))
926 if (!d_is_dir(victim))
930 } else if (d_is_dir(victim))
934 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
939 /* copy of may_create in fs/namei.c() */
940 static inline int btrfs_may_create(struct mnt_idmap *idmap,
941 struct inode *dir, struct dentry *child)
943 if (d_really_is_positive(child))
947 if (!fsuidgid_has_mapping(dir->i_sb, idmap))
949 return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
953 * Create a new subvolume below @parent. This is largely modeled after
954 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
955 * inside this filesystem so it's quite a bit simpler.
957 static noinline int btrfs_mksubvol(const struct path *parent,
958 struct mnt_idmap *idmap,
959 const char *name, int namelen,
960 struct btrfs_root *snap_src,
962 struct btrfs_qgroup_inherit *inherit)
964 struct inode *dir = d_inode(parent->dentry);
965 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
966 struct dentry *dentry;
967 struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
970 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
974 dentry = lookup_one(idmap, name, parent->dentry, namelen);
975 error = PTR_ERR(dentry);
979 error = btrfs_may_create(idmap, dir, dentry);
984 * even if this name doesn't exist, we may get hash collisions.
985 * check for them now when we can safely fail
987 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
988 dir->i_ino, &name_str);
992 down_read(&fs_info->subvol_sem);
994 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
998 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
1000 error = create_subvol(idmap, dir, dentry, inherit);
1003 fsnotify_mkdir(dir, dentry);
1005 up_read(&fs_info->subvol_sem);
1009 btrfs_inode_unlock(BTRFS_I(dir), 0);
1013 static noinline int btrfs_mksnapshot(const struct path *parent,
1014 struct mnt_idmap *idmap,
1015 const char *name, int namelen,
1016 struct btrfs_root *root,
1018 struct btrfs_qgroup_inherit *inherit)
1021 bool snapshot_force_cow = false;
1024 * Force new buffered writes to reserve space even when NOCOW is
1025 * possible. This is to avoid later writeback (running dealloc) to
1026 * fallback to COW mode and unexpectedly fail with ENOSPC.
1028 btrfs_drew_read_lock(&root->snapshot_lock);
1030 ret = btrfs_start_delalloc_snapshot(root, false);
1035 * All previous writes have started writeback in NOCOW mode, so now
1036 * we force future writes to fallback to COW mode during snapshot
1039 atomic_inc(&root->snapshot_force_cow);
1040 snapshot_force_cow = true;
1042 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1044 ret = btrfs_mksubvol(parent, idmap, name, namelen,
1045 root, readonly, inherit);
1047 if (snapshot_force_cow)
1048 atomic_dec(&root->snapshot_force_cow);
1049 btrfs_drew_read_unlock(&root->snapshot_lock);
1054 * Try to start exclusive operation @type or cancel it if it's running.
1057 * 0 - normal mode, newly claimed op started
1058 * >0 - normal mode, something else is running,
1059 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1060 * ECANCELED - cancel mode, successful cancel
1061 * ENOTCONN - cancel mode, operation not running anymore
1063 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1064 enum btrfs_exclusive_operation type, bool cancel)
1067 /* Start normal op */
1068 if (!btrfs_exclop_start(fs_info, type))
1069 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1070 /* Exclusive operation is now claimed */
1074 /* Cancel running op */
1075 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1077 * This blocks any exclop finish from setting it to NONE, so we
1078 * request cancellation. Either it runs and we will wait for it,
1079 * or it has finished and no waiting will happen.
1081 atomic_inc(&fs_info->reloc_cancel_req);
1082 btrfs_exclop_start_unlock(fs_info);
1084 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1085 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1086 TASK_INTERRUPTIBLE);
1091 /* Something else is running or none */
1095 static noinline int btrfs_ioctl_resize(struct file *file,
1098 BTRFS_DEV_LOOKUP_ARGS(args);
1099 struct inode *inode = file_inode(file);
1100 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1104 struct btrfs_root *root = BTRFS_I(inode)->root;
1105 struct btrfs_ioctl_vol_args *vol_args;
1106 struct btrfs_trans_handle *trans;
1107 struct btrfs_device *device = NULL;
1110 char *devstr = NULL;
1115 if (!capable(CAP_SYS_ADMIN))
1118 ret = mnt_want_write_file(file);
1123 * Read the arguments before checking exclusivity to be able to
1124 * distinguish regular resize and cancel
1126 vol_args = memdup_user(arg, sizeof(*vol_args));
1127 if (IS_ERR(vol_args)) {
1128 ret = PTR_ERR(vol_args);
1131 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1132 sizestr = vol_args->name;
1133 cancel = (strcmp("cancel", sizestr) == 0);
1134 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1137 /* Exclusive operation is now claimed */
1139 devstr = strchr(sizestr, ':');
1141 sizestr = devstr + 1;
1143 devstr = vol_args->name;
1144 ret = kstrtoull(devstr, 10, &devid);
1151 btrfs_info(fs_info, "resizing devid %llu", devid);
1155 device = btrfs_find_device(fs_info->fs_devices, &args);
1157 btrfs_info(fs_info, "resizer unable to find device %llu",
1163 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1165 "resizer unable to apply on readonly device %llu",
1171 if (!strcmp(sizestr, "max"))
1172 new_size = bdev_nr_bytes(device->bdev);
1174 if (sizestr[0] == '-') {
1177 } else if (sizestr[0] == '+') {
1181 new_size = memparse(sizestr, &retptr);
1182 if (*retptr != '\0' || new_size == 0) {
1188 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1193 old_size = btrfs_device_get_total_bytes(device);
1196 if (new_size > old_size) {
1200 new_size = old_size - new_size;
1201 } else if (mod > 0) {
1202 if (new_size > ULLONG_MAX - old_size) {
1206 new_size = old_size + new_size;
1209 if (new_size < SZ_256M) {
1213 if (new_size > bdev_nr_bytes(device->bdev)) {
1218 new_size = round_down(new_size, fs_info->sectorsize);
1220 if (new_size > old_size) {
1221 trans = btrfs_start_transaction(root, 0);
1222 if (IS_ERR(trans)) {
1223 ret = PTR_ERR(trans);
1226 ret = btrfs_grow_device(trans, device, new_size);
1227 btrfs_commit_transaction(trans);
1228 } else if (new_size < old_size) {
1229 ret = btrfs_shrink_device(device, new_size);
1230 } /* equal, nothing need to do */
1232 if (ret == 0 && new_size != old_size)
1233 btrfs_info_in_rcu(fs_info,
1234 "resize device %s (devid %llu) from %llu to %llu",
1235 btrfs_dev_name(device), device->devid,
1236 old_size, new_size);
1238 btrfs_exclop_finish(fs_info);
1242 mnt_drop_write_file(file);
1246 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1247 struct mnt_idmap *idmap,
1248 const char *name, unsigned long fd, int subvol,
1250 struct btrfs_qgroup_inherit *inherit)
1255 if (!S_ISDIR(file_inode(file)->i_mode))
1258 ret = mnt_want_write_file(file);
1262 namelen = strlen(name);
1263 if (strchr(name, '/')) {
1265 goto out_drop_write;
1268 if (name[0] == '.' &&
1269 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1271 goto out_drop_write;
1275 ret = btrfs_mksubvol(&file->f_path, idmap, name,
1276 namelen, NULL, readonly, inherit);
1278 struct fd src = fdget(fd);
1279 struct inode *src_inode;
1282 goto out_drop_write;
1285 src_inode = file_inode(src.file);
1286 if (src_inode->i_sb != file_inode(file)->i_sb) {
1287 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1288 "Snapshot src from another FS");
1290 } else if (!inode_owner_or_capable(idmap, src_inode)) {
1292 * Subvolume creation is not restricted, but snapshots
1293 * are limited to own subvolumes only
1296 } else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1298 * Snapshots must be made with the src_inode referring
1299 * to the subvolume inode, otherwise the permission
1300 * checking above is useless because we may have
1301 * permission on a lower directory but not the subvol
1306 ret = btrfs_mksnapshot(&file->f_path, idmap,
1308 BTRFS_I(src_inode)->root,
1314 mnt_drop_write_file(file);
1319 static noinline int btrfs_ioctl_snap_create(struct file *file,
1320 void __user *arg, int subvol)
1322 struct btrfs_ioctl_vol_args *vol_args;
1325 if (!S_ISDIR(file_inode(file)->i_mode))
1328 vol_args = memdup_user(arg, sizeof(*vol_args));
1329 if (IS_ERR(vol_args))
1330 return PTR_ERR(vol_args);
1331 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1333 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1334 vol_args->name, vol_args->fd, subvol,
1341 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1342 void __user *arg, int subvol)
1344 struct btrfs_ioctl_vol_args_v2 *vol_args;
1346 bool readonly = false;
1347 struct btrfs_qgroup_inherit *inherit = NULL;
1349 if (!S_ISDIR(file_inode(file)->i_mode))
1352 vol_args = memdup_user(arg, sizeof(*vol_args));
1353 if (IS_ERR(vol_args))
1354 return PTR_ERR(vol_args);
1355 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1357 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1362 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1364 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1367 if (vol_args->size < sizeof(*inherit) ||
1368 vol_args->size > PAGE_SIZE) {
1372 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1373 if (IS_ERR(inherit)) {
1374 ret = PTR_ERR(inherit);
1378 if (inherit->num_qgroups > PAGE_SIZE ||
1379 inherit->num_ref_copies > PAGE_SIZE ||
1380 inherit->num_excl_copies > PAGE_SIZE) {
1385 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1386 2 * inherit->num_excl_copies;
1387 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1393 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1394 vol_args->name, vol_args->fd, subvol,
1405 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1408 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1409 struct btrfs_root *root = BTRFS_I(inode)->root;
1413 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1416 down_read(&fs_info->subvol_sem);
1417 if (btrfs_root_readonly(root))
1418 flags |= BTRFS_SUBVOL_RDONLY;
1419 up_read(&fs_info->subvol_sem);
1421 if (copy_to_user(arg, &flags, sizeof(flags)))
1427 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1430 struct inode *inode = file_inode(file);
1431 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1432 struct btrfs_root *root = BTRFS_I(inode)->root;
1433 struct btrfs_trans_handle *trans;
1438 if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1441 ret = mnt_want_write_file(file);
1445 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1447 goto out_drop_write;
1450 if (copy_from_user(&flags, arg, sizeof(flags))) {
1452 goto out_drop_write;
1455 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1457 goto out_drop_write;
1460 down_write(&fs_info->subvol_sem);
1463 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1466 root_flags = btrfs_root_flags(&root->root_item);
1467 if (flags & BTRFS_SUBVOL_RDONLY) {
1468 btrfs_set_root_flags(&root->root_item,
1469 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1472 * Block RO -> RW transition if this subvolume is involved in
1475 spin_lock(&root->root_item_lock);
1476 if (root->send_in_progress == 0) {
1477 btrfs_set_root_flags(&root->root_item,
1478 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1479 spin_unlock(&root->root_item_lock);
1481 spin_unlock(&root->root_item_lock);
1483 "Attempt to set subvolume %llu read-write during send",
1484 root->root_key.objectid);
1490 trans = btrfs_start_transaction(root, 1);
1491 if (IS_ERR(trans)) {
1492 ret = PTR_ERR(trans);
1496 ret = btrfs_update_root(trans, fs_info->tree_root,
1497 &root->root_key, &root->root_item);
1499 btrfs_end_transaction(trans);
1503 ret = btrfs_commit_transaction(trans);
1507 btrfs_set_root_flags(&root->root_item, root_flags);
1509 up_write(&fs_info->subvol_sem);
1511 mnt_drop_write_file(file);
1516 static noinline int key_in_sk(struct btrfs_key *key,
1517 struct btrfs_ioctl_search_key *sk)
1519 struct btrfs_key test;
1522 test.objectid = sk->min_objectid;
1523 test.type = sk->min_type;
1524 test.offset = sk->min_offset;
1526 ret = btrfs_comp_cpu_keys(key, &test);
1530 test.objectid = sk->max_objectid;
1531 test.type = sk->max_type;
1532 test.offset = sk->max_offset;
1534 ret = btrfs_comp_cpu_keys(key, &test);
1540 static noinline int copy_to_sk(struct btrfs_path *path,
1541 struct btrfs_key *key,
1542 struct btrfs_ioctl_search_key *sk,
1545 unsigned long *sk_offset,
1549 struct extent_buffer *leaf;
1550 struct btrfs_ioctl_search_header sh;
1551 struct btrfs_key test;
1552 unsigned long item_off;
1553 unsigned long item_len;
1559 leaf = path->nodes[0];
1560 slot = path->slots[0];
1561 nritems = btrfs_header_nritems(leaf);
1563 if (btrfs_header_generation(leaf) > sk->max_transid) {
1567 found_transid = btrfs_header_generation(leaf);
1569 for (i = slot; i < nritems; i++) {
1570 item_off = btrfs_item_ptr_offset(leaf, i);
1571 item_len = btrfs_item_size(leaf, i);
1573 btrfs_item_key_to_cpu(leaf, key, i);
1574 if (!key_in_sk(key, sk))
1577 if (sizeof(sh) + item_len > *buf_size) {
1584 * return one empty item back for v1, which does not
1588 *buf_size = sizeof(sh) + item_len;
1593 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1598 sh.objectid = key->objectid;
1599 sh.offset = key->offset;
1600 sh.type = key->type;
1602 sh.transid = found_transid;
1605 * Copy search result header. If we fault then loop again so we
1606 * can fault in the pages and -EFAULT there if there's a
1607 * problem. Otherwise we'll fault and then copy the buffer in
1608 * properly this next time through
1610 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1615 *sk_offset += sizeof(sh);
1618 char __user *up = ubuf + *sk_offset;
1620 * Copy the item, same behavior as above, but reset the
1621 * * sk_offset so we copy the full thing again.
1623 if (read_extent_buffer_to_user_nofault(leaf, up,
1624 item_off, item_len)) {
1626 *sk_offset -= sizeof(sh);
1630 *sk_offset += item_len;
1634 if (ret) /* -EOVERFLOW from above */
1637 if (*num_found >= sk->nr_items) {
1644 test.objectid = sk->max_objectid;
1645 test.type = sk->max_type;
1646 test.offset = sk->max_offset;
1647 if (btrfs_comp_cpu_keys(key, &test) >= 0)
1649 else if (key->offset < (u64)-1)
1651 else if (key->type < (u8)-1) {
1654 } else if (key->objectid < (u64)-1) {
1662 * 0: all items from this leaf copied, continue with next
1663 * 1: * more items can be copied, but unused buffer is too small
1664 * * all items were found
1665 * Either way, it will stops the loop which iterates to the next
1667 * -EOVERFLOW: item was to large for buffer
1668 * -EFAULT: could not copy extent buffer back to userspace
1673 static noinline int search_ioctl(struct inode *inode,
1674 struct btrfs_ioctl_search_key *sk,
1678 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
1679 struct btrfs_root *root;
1680 struct btrfs_key key;
1681 struct btrfs_path *path;
1684 unsigned long sk_offset = 0;
1686 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1687 *buf_size = sizeof(struct btrfs_ioctl_search_header);
1691 path = btrfs_alloc_path();
1695 if (sk->tree_id == 0) {
1696 /* search the root of the inode that was passed */
1697 root = btrfs_grab_root(BTRFS_I(inode)->root);
1699 root = btrfs_get_fs_root(info, sk->tree_id, true);
1701 btrfs_free_path(path);
1702 return PTR_ERR(root);
1706 key.objectid = sk->min_objectid;
1707 key.type = sk->min_type;
1708 key.offset = sk->min_offset;
1713 * Ensure that the whole user buffer is faulted in at sub-page
1714 * granularity, otherwise the loop may live-lock.
1716 if (fault_in_subpage_writeable(ubuf + sk_offset,
1717 *buf_size - sk_offset))
1720 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1726 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1727 &sk_offset, &num_found);
1728 btrfs_release_path(path);
1736 sk->nr_items = num_found;
1737 btrfs_put_root(root);
1738 btrfs_free_path(path);
1742 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1745 struct btrfs_ioctl_search_args __user *uargs = argp;
1746 struct btrfs_ioctl_search_key sk;
1750 if (!capable(CAP_SYS_ADMIN))
1753 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1756 buf_size = sizeof(uargs->buf);
1758 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1761 * In the origin implementation an overflow is handled by returning a
1762 * search header with a len of zero, so reset ret.
1764 if (ret == -EOVERFLOW)
1767 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1772 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1775 struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1776 struct btrfs_ioctl_search_args_v2 args;
1779 const u64 buf_limit = SZ_16M;
1781 if (!capable(CAP_SYS_ADMIN))
1784 /* copy search header and buffer size */
1785 if (copy_from_user(&args, uarg, sizeof(args)))
1788 buf_size = args.buf_size;
1790 /* limit result size to 16MB */
1791 if (buf_size > buf_limit)
1792 buf_size = buf_limit;
1794 ret = search_ioctl(inode, &args.key, &buf_size,
1795 (char __user *)(&uarg->buf[0]));
1796 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1798 else if (ret == -EOVERFLOW &&
1799 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1806 * Search INODE_REFs to identify path name of 'dirid' directory
1807 * in a 'tree_id' tree. and sets path name to 'name'.
1809 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1810 u64 tree_id, u64 dirid, char *name)
1812 struct btrfs_root *root;
1813 struct btrfs_key key;
1819 struct btrfs_inode_ref *iref;
1820 struct extent_buffer *l;
1821 struct btrfs_path *path;
1823 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1828 path = btrfs_alloc_path();
1832 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1834 root = btrfs_get_fs_root(info, tree_id, true);
1836 ret = PTR_ERR(root);
1841 key.objectid = dirid;
1842 key.type = BTRFS_INODE_REF_KEY;
1843 key.offset = (u64)-1;
1846 ret = btrfs_search_backwards(root, &key, path);
1855 slot = path->slots[0];
1857 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1858 len = btrfs_inode_ref_name_len(l, iref);
1860 total_len += len + 1;
1862 ret = -ENAMETOOLONG;
1867 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1869 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1872 btrfs_release_path(path);
1873 key.objectid = key.offset;
1874 key.offset = (u64)-1;
1875 dirid = key.objectid;
1877 memmove(name, ptr, total_len);
1878 name[total_len] = '\0';
1881 btrfs_put_root(root);
1882 btrfs_free_path(path);
1886 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1887 struct inode *inode,
1888 struct btrfs_ioctl_ino_lookup_user_args *args)
1890 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1891 struct super_block *sb = inode->i_sb;
1892 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1893 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1894 u64 dirid = args->dirid;
1895 unsigned long item_off;
1896 unsigned long item_len;
1897 struct btrfs_inode_ref *iref;
1898 struct btrfs_root_ref *rref;
1899 struct btrfs_root *root = NULL;
1900 struct btrfs_path *path;
1901 struct btrfs_key key, key2;
1902 struct extent_buffer *leaf;
1903 struct inode *temp_inode;
1910 path = btrfs_alloc_path();
1915 * If the bottom subvolume does not exist directly under upper_limit,
1916 * construct the path in from the bottom up.
1918 if (dirid != upper_limit.objectid) {
1919 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1921 root = btrfs_get_fs_root(fs_info, treeid, true);
1923 ret = PTR_ERR(root);
1927 key.objectid = dirid;
1928 key.type = BTRFS_INODE_REF_KEY;
1929 key.offset = (u64)-1;
1931 ret = btrfs_search_backwards(root, &key, path);
1939 leaf = path->nodes[0];
1940 slot = path->slots[0];
1942 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1943 len = btrfs_inode_ref_name_len(leaf, iref);
1945 total_len += len + 1;
1946 if (ptr < args->path) {
1947 ret = -ENAMETOOLONG;
1952 read_extent_buffer(leaf, ptr,
1953 (unsigned long)(iref + 1), len);
1955 /* Check the read+exec permission of this directory */
1956 ret = btrfs_previous_item(root, path, dirid,
1957 BTRFS_INODE_ITEM_KEY);
1960 } else if (ret > 0) {
1965 leaf = path->nodes[0];
1966 slot = path->slots[0];
1967 btrfs_item_key_to_cpu(leaf, &key2, slot);
1968 if (key2.objectid != dirid) {
1974 * We don't need the path anymore, so release it and
1975 * avoid deadlocks and lockdep warnings in case
1976 * btrfs_iget() needs to lookup the inode from its root
1977 * btree and lock the same leaf.
1979 btrfs_release_path(path);
1980 temp_inode = btrfs_iget(sb, key2.objectid, root);
1981 if (IS_ERR(temp_inode)) {
1982 ret = PTR_ERR(temp_inode);
1985 ret = inode_permission(idmap, temp_inode,
1986 MAY_READ | MAY_EXEC);
1993 if (key.offset == upper_limit.objectid)
1995 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2000 key.objectid = key.offset;
2001 key.offset = (u64)-1;
2002 dirid = key.objectid;
2005 memmove(args->path, ptr, total_len);
2006 args->path[total_len] = '\0';
2007 btrfs_put_root(root);
2009 btrfs_release_path(path);
2012 /* Get the bottom subvolume's name from ROOT_REF */
2013 key.objectid = treeid;
2014 key.type = BTRFS_ROOT_REF_KEY;
2015 key.offset = args->treeid;
2016 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2019 } else if (ret > 0) {
2024 leaf = path->nodes[0];
2025 slot = path->slots[0];
2026 btrfs_item_key_to_cpu(leaf, &key, slot);
2028 item_off = btrfs_item_ptr_offset(leaf, slot);
2029 item_len = btrfs_item_size(leaf, slot);
2030 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2031 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2032 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2037 /* Copy subvolume's name */
2038 item_off += sizeof(struct btrfs_root_ref);
2039 item_len -= sizeof(struct btrfs_root_ref);
2040 read_extent_buffer(leaf, args->name, item_off, item_len);
2041 args->name[item_len] = 0;
2044 btrfs_put_root(root);
2046 btrfs_free_path(path);
2050 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2053 struct btrfs_ioctl_ino_lookup_args *args;
2056 args = memdup_user(argp, sizeof(*args));
2058 return PTR_ERR(args);
2061 * Unprivileged query to obtain the containing subvolume root id. The
2062 * path is reset so it's consistent with btrfs_search_path_in_tree.
2064 if (args->treeid == 0)
2065 args->treeid = root->root_key.objectid;
2067 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2072 if (!capable(CAP_SYS_ADMIN)) {
2077 ret = btrfs_search_path_in_tree(root->fs_info,
2078 args->treeid, args->objectid,
2082 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2090 * Version of ino_lookup ioctl (unprivileged)
2092 * The main differences from ino_lookup ioctl are:
2094 * 1. Read + Exec permission will be checked using inode_permission() during
2095 * path construction. -EACCES will be returned in case of failure.
2096 * 2. Path construction will be stopped at the inode number which corresponds
2097 * to the fd with which this ioctl is called. If constructed path does not
2098 * exist under fd's inode, -EACCES will be returned.
2099 * 3. The name of bottom subvolume is also searched and filled.
2101 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2103 struct btrfs_ioctl_ino_lookup_user_args *args;
2104 struct inode *inode;
2107 args = memdup_user(argp, sizeof(*args));
2109 return PTR_ERR(args);
2111 inode = file_inode(file);
2113 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2114 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2116 * The subvolume does not exist under fd with which this is
2123 ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2125 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2132 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2133 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2135 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2136 struct btrfs_fs_info *fs_info;
2137 struct btrfs_root *root;
2138 struct btrfs_path *path;
2139 struct btrfs_key key;
2140 struct btrfs_root_item *root_item;
2141 struct btrfs_root_ref *rref;
2142 struct extent_buffer *leaf;
2143 unsigned long item_off;
2144 unsigned long item_len;
2148 path = btrfs_alloc_path();
2152 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2154 btrfs_free_path(path);
2158 fs_info = BTRFS_I(inode)->root->fs_info;
2160 /* Get root_item of inode's subvolume */
2161 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2162 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2164 ret = PTR_ERR(root);
2167 root_item = &root->root_item;
2169 subvol_info->treeid = key.objectid;
2171 subvol_info->generation = btrfs_root_generation(root_item);
2172 subvol_info->flags = btrfs_root_flags(root_item);
2174 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2175 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2177 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2180 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2181 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2182 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2184 subvol_info->otransid = btrfs_root_otransid(root_item);
2185 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2186 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2188 subvol_info->stransid = btrfs_root_stransid(root_item);
2189 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2190 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2192 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2193 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2194 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2196 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2197 /* Search root tree for ROOT_BACKREF of this subvolume */
2198 key.type = BTRFS_ROOT_BACKREF_KEY;
2200 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2203 } else if (path->slots[0] >=
2204 btrfs_header_nritems(path->nodes[0])) {
2205 ret = btrfs_next_leaf(fs_info->tree_root, path);
2208 } else if (ret > 0) {
2214 leaf = path->nodes[0];
2215 slot = path->slots[0];
2216 btrfs_item_key_to_cpu(leaf, &key, slot);
2217 if (key.objectid == subvol_info->treeid &&
2218 key.type == BTRFS_ROOT_BACKREF_KEY) {
2219 subvol_info->parent_id = key.offset;
2221 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2222 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2224 item_off = btrfs_item_ptr_offset(leaf, slot)
2225 + sizeof(struct btrfs_root_ref);
2226 item_len = btrfs_item_size(leaf, slot)
2227 - sizeof(struct btrfs_root_ref);
2228 read_extent_buffer(leaf, subvol_info->name,
2229 item_off, item_len);
2236 btrfs_free_path(path);
2238 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2242 btrfs_put_root(root);
2244 btrfs_free_path(path);
2250 * Return ROOT_REF information of the subvolume containing this inode
2251 * except the subvolume name.
2253 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2256 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2257 struct btrfs_root_ref *rref;
2258 struct btrfs_path *path;
2259 struct btrfs_key key;
2260 struct extent_buffer *leaf;
2266 path = btrfs_alloc_path();
2270 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2271 if (IS_ERR(rootrefs)) {
2272 btrfs_free_path(path);
2273 return PTR_ERR(rootrefs);
2276 objectid = root->root_key.objectid;
2277 key.objectid = objectid;
2278 key.type = BTRFS_ROOT_REF_KEY;
2279 key.offset = rootrefs->min_treeid;
2282 root = root->fs_info->tree_root;
2283 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2286 } else if (path->slots[0] >=
2287 btrfs_header_nritems(path->nodes[0])) {
2288 ret = btrfs_next_leaf(root, path);
2291 } else if (ret > 0) {
2297 leaf = path->nodes[0];
2298 slot = path->slots[0];
2300 btrfs_item_key_to_cpu(leaf, &key, slot);
2301 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2306 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2311 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2312 rootrefs->rootref[found].treeid = key.offset;
2313 rootrefs->rootref[found].dirid =
2314 btrfs_root_ref_dirid(leaf, rref);
2317 ret = btrfs_next_item(root, path);
2320 } else if (ret > 0) {
2327 btrfs_free_path(path);
2329 if (!ret || ret == -EOVERFLOW) {
2330 rootrefs->num_items = found;
2331 /* update min_treeid for next search */
2333 rootrefs->min_treeid =
2334 rootrefs->rootref[found - 1].treeid + 1;
2335 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2344 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2348 struct dentry *parent = file->f_path.dentry;
2349 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2350 struct dentry *dentry;
2351 struct inode *dir = d_inode(parent);
2352 struct inode *inode;
2353 struct btrfs_root *root = BTRFS_I(dir)->root;
2354 struct btrfs_root *dest = NULL;
2355 struct btrfs_ioctl_vol_args *vol_args = NULL;
2356 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2357 struct mnt_idmap *idmap = file_mnt_idmap(file);
2358 char *subvol_name, *subvol_name_ptr = NULL;
2361 bool destroy_parent = false;
2363 /* We don't support snapshots with extent tree v2 yet. */
2364 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2366 "extent tree v2 doesn't support snapshot deletion yet");
2371 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2372 if (IS_ERR(vol_args2))
2373 return PTR_ERR(vol_args2);
2375 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2381 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2382 * name, same as v1 currently does.
2384 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2385 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2386 subvol_name = vol_args2->name;
2388 err = mnt_want_write_file(file);
2392 struct inode *old_dir;
2394 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2399 err = mnt_want_write_file(file);
2403 dentry = btrfs_get_dentry(fs_info->sb,
2404 BTRFS_FIRST_FREE_OBJECTID,
2405 vol_args2->subvolid, 0);
2406 if (IS_ERR(dentry)) {
2407 err = PTR_ERR(dentry);
2408 goto out_drop_write;
2412 * Change the default parent since the subvolume being
2413 * deleted can be outside of the current mount point.
2415 parent = btrfs_get_parent(dentry);
2418 * At this point dentry->d_name can point to '/' if the
2419 * subvolume we want to destroy is outsite of the
2420 * current mount point, so we need to release the
2421 * current dentry and execute the lookup to return a new
2422 * one with ->d_name pointing to the
2423 * <mount point>/subvol_name.
2426 if (IS_ERR(parent)) {
2427 err = PTR_ERR(parent);
2428 goto out_drop_write;
2431 dir = d_inode(parent);
2434 * If v2 was used with SPEC_BY_ID, a new parent was
2435 * allocated since the subvolume can be outside of the
2436 * current mount point. Later on we need to release this
2437 * new parent dentry.
2439 destroy_parent = true;
2442 * On idmapped mounts, deletion via subvolid is
2443 * restricted to subvolumes that are immediate
2444 * ancestors of the inode referenced by the file
2445 * descriptor in the ioctl. Otherwise the idmapping
2446 * could potentially be abused to delete subvolumes
2447 * anywhere in the filesystem the user wouldn't be able
2448 * to delete without an idmapped mount.
2450 if (old_dir != dir && idmap != &nop_mnt_idmap) {
2455 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2456 fs_info, vol_args2->subvolid);
2457 if (IS_ERR(subvol_name_ptr)) {
2458 err = PTR_ERR(subvol_name_ptr);
2461 /* subvol_name_ptr is already nul terminated */
2462 subvol_name = (char *)kbasename(subvol_name_ptr);
2465 vol_args = memdup_user(arg, sizeof(*vol_args));
2466 if (IS_ERR(vol_args))
2467 return PTR_ERR(vol_args);
2469 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
2470 subvol_name = vol_args->name;
2472 err = mnt_want_write_file(file);
2477 subvol_namelen = strlen(subvol_name);
2479 if (strchr(subvol_name, '/') ||
2480 strncmp(subvol_name, "..", subvol_namelen) == 0) {
2482 goto free_subvol_name;
2485 if (!S_ISDIR(dir->i_mode)) {
2487 goto free_subvol_name;
2490 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2492 goto free_subvol_name;
2493 dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2494 if (IS_ERR(dentry)) {
2495 err = PTR_ERR(dentry);
2496 goto out_unlock_dir;
2499 if (d_really_is_negative(dentry)) {
2504 inode = d_inode(dentry);
2505 dest = BTRFS_I(inode)->root;
2506 if (!capable(CAP_SYS_ADMIN)) {
2508 * Regular user. Only allow this with a special mount
2509 * option, when the user has write+exec access to the
2510 * subvol root, and when rmdir(2) would have been
2513 * Note that this is _not_ check that the subvol is
2514 * empty or doesn't contain data that we wouldn't
2515 * otherwise be able to delete.
2517 * Users who want to delete empty subvols should try
2521 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2525 * Do not allow deletion if the parent dir is the same
2526 * as the dir to be deleted. That means the ioctl
2527 * must be called on the dentry referencing the root
2528 * of the subvol, not a random directory contained
2535 err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2540 /* check if subvolume may be deleted by a user */
2541 err = btrfs_may_delete(idmap, dir, dentry, 1);
2545 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2550 btrfs_inode_lock(BTRFS_I(inode), 0);
2551 err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2552 btrfs_inode_unlock(BTRFS_I(inode), 0);
2554 d_delete_notify(dir, dentry);
2559 btrfs_inode_unlock(BTRFS_I(dir), 0);
2561 kfree(subvol_name_ptr);
2566 mnt_drop_write_file(file);
2573 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2575 struct inode *inode = file_inode(file);
2576 struct btrfs_root *root = BTRFS_I(inode)->root;
2577 struct btrfs_ioctl_defrag_range_args range = {0};
2580 ret = mnt_want_write_file(file);
2584 if (btrfs_root_readonly(root)) {
2589 switch (inode->i_mode & S_IFMT) {
2591 if (!capable(CAP_SYS_ADMIN)) {
2595 ret = btrfs_defrag_root(root);
2599 * Note that this does not check the file descriptor for write
2600 * access. This prevents defragmenting executables that are
2601 * running and allows defrag on files open in read-only mode.
2603 if (!capable(CAP_SYS_ADMIN) &&
2604 inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2610 if (copy_from_user(&range, argp, sizeof(range))) {
2614 if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2618 /* compression requires us to start the IO */
2619 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2620 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2621 range.extent_thresh = (u32)-1;
2624 /* the rest are all set to zero by kzalloc */
2625 range.len = (u64)-1;
2627 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2628 &range, BTRFS_OLDEST_GENERATION, 0);
2636 mnt_drop_write_file(file);
2640 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2642 struct btrfs_ioctl_vol_args *vol_args;
2643 bool restore_op = false;
2646 if (!capable(CAP_SYS_ADMIN))
2649 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2650 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2654 if (fs_info->fs_devices->temp_fsid) {
2656 "device add not supported on cloned temp-fsid mount");
2660 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2661 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2662 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2665 * We can do the device add because we have a paused balanced,
2666 * change the exclusive op type and remember we should bring
2667 * back the paused balance
2669 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2670 btrfs_exclop_start_unlock(fs_info);
2674 vol_args = memdup_user(arg, sizeof(*vol_args));
2675 if (IS_ERR(vol_args)) {
2676 ret = PTR_ERR(vol_args);
2680 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2681 ret = btrfs_init_new_device(fs_info, vol_args->name);
2684 btrfs_info(fs_info, "disk added %s", vol_args->name);
2689 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2691 btrfs_exclop_finish(fs_info);
2695 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2697 BTRFS_DEV_LOOKUP_ARGS(args);
2698 struct inode *inode = file_inode(file);
2699 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2700 struct btrfs_ioctl_vol_args_v2 *vol_args;
2701 struct bdev_handle *bdev_handle = NULL;
2703 bool cancel = false;
2705 if (!capable(CAP_SYS_ADMIN))
2708 vol_args = memdup_user(arg, sizeof(*vol_args));
2709 if (IS_ERR(vol_args))
2710 return PTR_ERR(vol_args);
2712 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2717 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2718 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2719 args.devid = vol_args->devid;
2720 } else if (!strcmp("cancel", vol_args->name)) {
2723 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2728 ret = mnt_want_write_file(file);
2732 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2737 /* Exclusive operation is now claimed */
2738 ret = btrfs_rm_device(fs_info, &args, &bdev_handle);
2740 btrfs_exclop_finish(fs_info);
2743 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2744 btrfs_info(fs_info, "device deleted: id %llu",
2747 btrfs_info(fs_info, "device deleted: %s",
2751 mnt_drop_write_file(file);
2753 bdev_release(bdev_handle);
2755 btrfs_put_dev_args_from_path(&args);
2760 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2762 BTRFS_DEV_LOOKUP_ARGS(args);
2763 struct inode *inode = file_inode(file);
2764 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2765 struct btrfs_ioctl_vol_args *vol_args;
2766 struct bdev_handle *bdev_handle = NULL;
2768 bool cancel = false;
2770 if (!capable(CAP_SYS_ADMIN))
2773 vol_args = memdup_user(arg, sizeof(*vol_args));
2774 if (IS_ERR(vol_args))
2775 return PTR_ERR(vol_args);
2777 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2778 if (!strcmp("cancel", vol_args->name)) {
2781 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2786 ret = mnt_want_write_file(file);
2790 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2793 ret = btrfs_rm_device(fs_info, &args, &bdev_handle);
2795 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2796 btrfs_exclop_finish(fs_info);
2799 mnt_drop_write_file(file);
2801 bdev_release(bdev_handle);
2803 btrfs_put_dev_args_from_path(&args);
2808 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2811 struct btrfs_ioctl_fs_info_args *fi_args;
2812 struct btrfs_device *device;
2813 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2817 fi_args = memdup_user(arg, sizeof(*fi_args));
2818 if (IS_ERR(fi_args))
2819 return PTR_ERR(fi_args);
2821 flags_in = fi_args->flags;
2822 memset(fi_args, 0, sizeof(*fi_args));
2825 fi_args->num_devices = fs_devices->num_devices;
2827 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2828 if (device->devid > fi_args->max_id)
2829 fi_args->max_id = device->devid;
2833 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2834 fi_args->nodesize = fs_info->nodesize;
2835 fi_args->sectorsize = fs_info->sectorsize;
2836 fi_args->clone_alignment = fs_info->sectorsize;
2838 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2839 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2840 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2841 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2844 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2845 fi_args->generation = btrfs_get_fs_generation(fs_info);
2846 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2849 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2850 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2851 sizeof(fi_args->metadata_uuid));
2852 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2855 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2862 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2865 BTRFS_DEV_LOOKUP_ARGS(args);
2866 struct btrfs_ioctl_dev_info_args *di_args;
2867 struct btrfs_device *dev;
2870 di_args = memdup_user(arg, sizeof(*di_args));
2871 if (IS_ERR(di_args))
2872 return PTR_ERR(di_args);
2874 args.devid = di_args->devid;
2875 if (!btrfs_is_empty_uuid(di_args->uuid))
2876 args.uuid = di_args->uuid;
2879 dev = btrfs_find_device(fs_info->fs_devices, &args);
2885 di_args->devid = dev->devid;
2886 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2887 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2888 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2889 memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2891 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2893 di_args->path[0] = '\0';
2897 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2904 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2906 struct inode *inode = file_inode(file);
2907 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2908 struct btrfs_root *root = BTRFS_I(inode)->root;
2909 struct btrfs_root *new_root;
2910 struct btrfs_dir_item *di;
2911 struct btrfs_trans_handle *trans;
2912 struct btrfs_path *path = NULL;
2913 struct btrfs_disk_key disk_key;
2914 struct fscrypt_str name = FSTR_INIT("default", 7);
2919 if (!capable(CAP_SYS_ADMIN))
2922 ret = mnt_want_write_file(file);
2926 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2932 objectid = BTRFS_FS_TREE_OBJECTID;
2934 new_root = btrfs_get_fs_root(fs_info, objectid, true);
2935 if (IS_ERR(new_root)) {
2936 ret = PTR_ERR(new_root);
2939 if (!is_fstree(new_root->root_key.objectid)) {
2944 path = btrfs_alloc_path();
2950 trans = btrfs_start_transaction(root, 1);
2951 if (IS_ERR(trans)) {
2952 ret = PTR_ERR(trans);
2956 dir_id = btrfs_super_root_dir(fs_info->super_copy);
2957 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2959 if (IS_ERR_OR_NULL(di)) {
2960 btrfs_release_path(path);
2961 btrfs_end_transaction(trans);
2963 "Umm, you don't have the default diritem, this isn't going to work");
2968 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2969 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2970 btrfs_mark_buffer_dirty(trans, path->nodes[0]);
2971 btrfs_release_path(path);
2973 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
2974 btrfs_end_transaction(trans);
2976 btrfs_put_root(new_root);
2977 btrfs_free_path(path);
2979 mnt_drop_write_file(file);
2983 static void get_block_group_info(struct list_head *groups_list,
2984 struct btrfs_ioctl_space_info *space)
2986 struct btrfs_block_group *block_group;
2988 space->total_bytes = 0;
2989 space->used_bytes = 0;
2991 list_for_each_entry(block_group, groups_list, list) {
2992 space->flags = block_group->flags;
2993 space->total_bytes += block_group->length;
2994 space->used_bytes += block_group->used;
2998 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3001 struct btrfs_ioctl_space_args space_args = { 0 };
3002 struct btrfs_ioctl_space_info space;
3003 struct btrfs_ioctl_space_info *dest;
3004 struct btrfs_ioctl_space_info *dest_orig;
3005 struct btrfs_ioctl_space_info __user *user_dest;
3006 struct btrfs_space_info *info;
3007 static const u64 types[] = {
3008 BTRFS_BLOCK_GROUP_DATA,
3009 BTRFS_BLOCK_GROUP_SYSTEM,
3010 BTRFS_BLOCK_GROUP_METADATA,
3011 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3019 if (copy_from_user(&space_args,
3020 (struct btrfs_ioctl_space_args __user *)arg,
3021 sizeof(space_args)))
3024 for (i = 0; i < num_types; i++) {
3025 struct btrfs_space_info *tmp;
3028 list_for_each_entry(tmp, &fs_info->space_info, list) {
3029 if (tmp->flags == types[i]) {
3038 down_read(&info->groups_sem);
3039 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3040 if (!list_empty(&info->block_groups[c]))
3043 up_read(&info->groups_sem);
3047 * Global block reserve, exported as a space_info
3051 /* space_slots == 0 means they are asking for a count */
3052 if (space_args.space_slots == 0) {
3053 space_args.total_spaces = slot_count;
3057 slot_count = min_t(u64, space_args.space_slots, slot_count);
3059 alloc_size = sizeof(*dest) * slot_count;
3061 /* we generally have at most 6 or so space infos, one for each raid
3062 * level. So, a whole page should be more than enough for everyone
3064 if (alloc_size > PAGE_SIZE)
3067 space_args.total_spaces = 0;
3068 dest = kmalloc(alloc_size, GFP_KERNEL);
3073 /* now we have a buffer to copy into */
3074 for (i = 0; i < num_types; i++) {
3075 struct btrfs_space_info *tmp;
3081 list_for_each_entry(tmp, &fs_info->space_info, list) {
3082 if (tmp->flags == types[i]) {
3090 down_read(&info->groups_sem);
3091 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3092 if (!list_empty(&info->block_groups[c])) {
3093 get_block_group_info(&info->block_groups[c],
3095 memcpy(dest, &space, sizeof(space));
3097 space_args.total_spaces++;
3103 up_read(&info->groups_sem);
3107 * Add global block reserve
3110 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3112 spin_lock(&block_rsv->lock);
3113 space.total_bytes = block_rsv->size;
3114 space.used_bytes = block_rsv->size - block_rsv->reserved;
3115 spin_unlock(&block_rsv->lock);
3116 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3117 memcpy(dest, &space, sizeof(space));
3118 space_args.total_spaces++;
3121 user_dest = (struct btrfs_ioctl_space_info __user *)
3122 (arg + sizeof(struct btrfs_ioctl_space_args));
3124 if (copy_to_user(user_dest, dest_orig, alloc_size))
3129 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3135 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3138 struct btrfs_trans_handle *trans;
3142 * Start orphan cleanup here for the given root in case it hasn't been
3143 * started already by other means. Errors are handled in the other
3144 * functions during transaction commit.
3146 btrfs_orphan_cleanup(root);
3148 trans = btrfs_attach_transaction_barrier(root);
3149 if (IS_ERR(trans)) {
3150 if (PTR_ERR(trans) != -ENOENT)
3151 return PTR_ERR(trans);
3153 /* No running transaction, don't bother */
3154 transid = btrfs_get_last_trans_committed(root->fs_info);
3157 transid = trans->transid;
3158 btrfs_commit_transaction_async(trans);
3161 if (copy_to_user(argp, &transid, sizeof(transid)))
3166 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3169 /* By default wait for the current transaction. */
3173 if (copy_from_user(&transid, argp, sizeof(transid)))
3176 return btrfs_wait_for_commit(fs_info, transid);
3179 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3181 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3182 struct btrfs_ioctl_scrub_args *sa;
3185 if (!capable(CAP_SYS_ADMIN))
3188 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3189 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3193 sa = memdup_user(arg, sizeof(*sa));
3197 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3202 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3203 ret = mnt_want_write_file(file);
3208 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3209 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3213 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3214 * error. This is important as it allows user space to know how much
3215 * progress scrub has done. For example, if scrub is canceled we get
3216 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3217 * space. Later user space can inspect the progress from the structure
3218 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3219 * previously (btrfs-progs does this).
3220 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3221 * then return -EFAULT to signal the structure was not copied or it may
3222 * be corrupt and unreliable due to a partial copy.
3224 if (copy_to_user(arg, sa, sizeof(*sa)))
3227 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3228 mnt_drop_write_file(file);
3234 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3236 if (!capable(CAP_SYS_ADMIN))
3239 return btrfs_scrub_cancel(fs_info);
3242 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3245 struct btrfs_ioctl_scrub_args *sa;
3248 if (!capable(CAP_SYS_ADMIN))
3251 sa = memdup_user(arg, sizeof(*sa));
3255 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3257 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3264 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3267 struct btrfs_ioctl_get_dev_stats *sa;
3270 sa = memdup_user(arg, sizeof(*sa));
3274 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3279 ret = btrfs_get_dev_stats(fs_info, sa);
3281 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3288 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3291 struct btrfs_ioctl_dev_replace_args *p;
3294 if (!capable(CAP_SYS_ADMIN))
3297 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3298 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3302 p = memdup_user(arg, sizeof(*p));
3307 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3308 if (sb_rdonly(fs_info->sb)) {
3312 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3313 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3315 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3316 btrfs_exclop_finish(fs_info);
3319 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3320 btrfs_dev_replace_status(fs_info, p);
3323 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3324 p->result = btrfs_dev_replace_cancel(fs_info);
3332 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3339 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3345 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3346 struct inode_fs_paths *ipath = NULL;
3347 struct btrfs_path *path;
3349 if (!capable(CAP_DAC_READ_SEARCH))
3352 path = btrfs_alloc_path();
3358 ipa = memdup_user(arg, sizeof(*ipa));
3365 size = min_t(u32, ipa->size, 4096);
3366 ipath = init_ipath(size, root, path);
3367 if (IS_ERR(ipath)) {
3368 ret = PTR_ERR(ipath);
3373 ret = paths_from_inode(ipa->inum, ipath);
3377 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3378 rel_ptr = ipath->fspath->val[i] -
3379 (u64)(unsigned long)ipath->fspath->val;
3380 ipath->fspath->val[i] = rel_ptr;
3383 btrfs_free_path(path);
3385 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3386 ipath->fspath, size);
3393 btrfs_free_path(path);
3400 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3401 void __user *arg, int version)
3405 struct btrfs_ioctl_logical_ino_args *loi;
3406 struct btrfs_data_container *inodes = NULL;
3407 struct btrfs_path *path = NULL;
3410 if (!capable(CAP_SYS_ADMIN))
3413 loi = memdup_user(arg, sizeof(*loi));
3415 return PTR_ERR(loi);
3418 ignore_offset = false;
3419 size = min_t(u32, loi->size, SZ_64K);
3421 /* All reserved bits must be 0 for now */
3422 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3426 /* Only accept flags we have defined so far */
3427 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3431 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3432 size = min_t(u32, loi->size, SZ_16M);
3435 inodes = init_data_container(size);
3436 if (IS_ERR(inodes)) {
3437 ret = PTR_ERR(inodes);
3441 path = btrfs_alloc_path();
3446 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3447 inodes, ignore_offset);
3448 btrfs_free_path(path);
3454 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3467 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3468 struct btrfs_ioctl_balance_args *bargs)
3470 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3472 bargs->flags = bctl->flags;
3474 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3475 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3476 if (atomic_read(&fs_info->balance_pause_req))
3477 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3478 if (atomic_read(&fs_info->balance_cancel_req))
3479 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3481 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3482 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3483 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3485 spin_lock(&fs_info->balance_lock);
3486 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3487 spin_unlock(&fs_info->balance_lock);
3491 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3494 * @fs_info: the filesystem
3495 * @excl_acquired: ptr to boolean value which is set to false in case balance
3498 * Return 0 on success in which case both fs_info::balance is acquired as well
3499 * as exclusive ops are blocked. In case of failure return an error code.
3501 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3506 * Exclusive operation is locked. Three possibilities:
3507 * (1) some other op is running
3508 * (2) balance is running
3509 * (3) balance is paused -- special case (think resume)
3512 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3513 *excl_acquired = true;
3514 mutex_lock(&fs_info->balance_mutex);
3518 mutex_lock(&fs_info->balance_mutex);
3519 if (fs_info->balance_ctl) {
3520 /* This is either (2) or (3) */
3521 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3527 mutex_unlock(&fs_info->balance_mutex);
3529 * Lock released to allow other waiters to
3530 * continue, we'll reexamine the status again.
3532 mutex_lock(&fs_info->balance_mutex);
3534 if (fs_info->balance_ctl &&
3535 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3537 *excl_acquired = false;
3543 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3547 mutex_unlock(&fs_info->balance_mutex);
3551 mutex_unlock(&fs_info->balance_mutex);
3552 *excl_acquired = false;
3556 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3558 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3559 struct btrfs_fs_info *fs_info = root->fs_info;
3560 struct btrfs_ioctl_balance_args *bargs;
3561 struct btrfs_balance_control *bctl;
3562 bool need_unlock = true;
3565 if (!capable(CAP_SYS_ADMIN))
3568 ret = mnt_want_write_file(file);
3572 bargs = memdup_user(arg, sizeof(*bargs));
3573 if (IS_ERR(bargs)) {
3574 ret = PTR_ERR(bargs);
3579 ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3583 lockdep_assert_held(&fs_info->balance_mutex);
3585 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3586 if (!fs_info->balance_ctl) {
3591 bctl = fs_info->balance_ctl;
3592 spin_lock(&fs_info->balance_lock);
3593 bctl->flags |= BTRFS_BALANCE_RESUME;
3594 spin_unlock(&fs_info->balance_lock);
3595 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3600 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3605 if (fs_info->balance_ctl) {
3610 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3616 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3617 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3618 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3620 bctl->flags = bargs->flags;
3623 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3624 * bctl is freed in reset_balance_state, or, if restriper was paused
3625 * all the way until unmount, in free_fs_info. The flag should be
3626 * cleared after reset_balance_state.
3628 need_unlock = false;
3630 ret = btrfs_balance(fs_info, bctl, bargs);
3633 if (ret == 0 || ret == -ECANCELED) {
3634 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3640 mutex_unlock(&fs_info->balance_mutex);
3642 btrfs_exclop_finish(fs_info);
3644 mnt_drop_write_file(file);
3649 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3651 if (!capable(CAP_SYS_ADMIN))
3655 case BTRFS_BALANCE_CTL_PAUSE:
3656 return btrfs_pause_balance(fs_info);
3657 case BTRFS_BALANCE_CTL_CANCEL:
3658 return btrfs_cancel_balance(fs_info);
3664 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3667 struct btrfs_ioctl_balance_args *bargs;
3670 if (!capable(CAP_SYS_ADMIN))
3673 mutex_lock(&fs_info->balance_mutex);
3674 if (!fs_info->balance_ctl) {
3679 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3685 btrfs_update_ioctl_balance_args(fs_info, bargs);
3687 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3692 mutex_unlock(&fs_info->balance_mutex);
3696 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3698 struct inode *inode = file_inode(file);
3699 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3700 struct btrfs_ioctl_quota_ctl_args *sa;
3703 if (!capable(CAP_SYS_ADMIN))
3706 ret = mnt_want_write_file(file);
3710 sa = memdup_user(arg, sizeof(*sa));
3716 down_write(&fs_info->subvol_sem);
3719 case BTRFS_QUOTA_CTL_ENABLE:
3720 case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3721 ret = btrfs_quota_enable(fs_info, sa);
3723 case BTRFS_QUOTA_CTL_DISABLE:
3724 ret = btrfs_quota_disable(fs_info);
3732 up_write(&fs_info->subvol_sem);
3734 mnt_drop_write_file(file);
3738 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3740 struct inode *inode = file_inode(file);
3741 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3742 struct btrfs_root *root = BTRFS_I(inode)->root;
3743 struct btrfs_ioctl_qgroup_assign_args *sa;
3744 struct btrfs_trans_handle *trans;
3748 if (!capable(CAP_SYS_ADMIN))
3751 ret = mnt_want_write_file(file);
3755 sa = memdup_user(arg, sizeof(*sa));
3761 trans = btrfs_join_transaction(root);
3762 if (IS_ERR(trans)) {
3763 ret = PTR_ERR(trans);
3768 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3770 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3773 /* update qgroup status and info */
3774 mutex_lock(&fs_info->qgroup_ioctl_lock);
3775 err = btrfs_run_qgroups(trans);
3776 mutex_unlock(&fs_info->qgroup_ioctl_lock);
3778 btrfs_handle_fs_error(fs_info, err,
3779 "failed to update qgroup status and info");
3780 err = btrfs_end_transaction(trans);
3787 mnt_drop_write_file(file);
3791 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3793 struct inode *inode = file_inode(file);
3794 struct btrfs_root *root = BTRFS_I(inode)->root;
3795 struct btrfs_ioctl_qgroup_create_args *sa;
3796 struct btrfs_trans_handle *trans;
3800 if (!capable(CAP_SYS_ADMIN))
3803 ret = mnt_want_write_file(file);
3807 sa = memdup_user(arg, sizeof(*sa));
3813 if (!sa->qgroupid) {
3818 trans = btrfs_join_transaction(root);
3819 if (IS_ERR(trans)) {
3820 ret = PTR_ERR(trans);
3825 ret = btrfs_create_qgroup(trans, sa->qgroupid);
3827 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3830 err = btrfs_end_transaction(trans);
3837 mnt_drop_write_file(file);
3841 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3843 struct inode *inode = file_inode(file);
3844 struct btrfs_root *root = BTRFS_I(inode)->root;
3845 struct btrfs_ioctl_qgroup_limit_args *sa;
3846 struct btrfs_trans_handle *trans;
3851 if (!capable(CAP_SYS_ADMIN))
3854 ret = mnt_want_write_file(file);
3858 sa = memdup_user(arg, sizeof(*sa));
3864 trans = btrfs_join_transaction(root);
3865 if (IS_ERR(trans)) {
3866 ret = PTR_ERR(trans);
3870 qgroupid = sa->qgroupid;
3872 /* take the current subvol as qgroup */
3873 qgroupid = root->root_key.objectid;
3876 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3878 err = btrfs_end_transaction(trans);
3885 mnt_drop_write_file(file);
3889 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3891 struct inode *inode = file_inode(file);
3892 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3893 struct btrfs_ioctl_quota_rescan_args *qsa;
3896 if (!capable(CAP_SYS_ADMIN))
3899 ret = mnt_want_write_file(file);
3903 qsa = memdup_user(arg, sizeof(*qsa));
3914 ret = btrfs_qgroup_rescan(fs_info);
3919 mnt_drop_write_file(file);
3923 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3926 struct btrfs_ioctl_quota_rescan_args qsa = {0};
3928 if (!capable(CAP_SYS_ADMIN))
3931 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3933 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3936 if (copy_to_user(arg, &qsa, sizeof(qsa)))
3942 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3945 if (!capable(CAP_SYS_ADMIN))
3948 return btrfs_qgroup_wait_for_completion(fs_info, true);
3951 static long _btrfs_ioctl_set_received_subvol(struct file *file,
3952 struct mnt_idmap *idmap,
3953 struct btrfs_ioctl_received_subvol_args *sa)
3955 struct inode *inode = file_inode(file);
3956 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3957 struct btrfs_root *root = BTRFS_I(inode)->root;
3958 struct btrfs_root_item *root_item = &root->root_item;
3959 struct btrfs_trans_handle *trans;
3960 struct timespec64 ct = current_time(inode);
3962 int received_uuid_changed;
3964 if (!inode_owner_or_capable(idmap, inode))
3967 ret = mnt_want_write_file(file);
3971 down_write(&fs_info->subvol_sem);
3973 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3978 if (btrfs_root_readonly(root)) {
3985 * 2 - uuid items (received uuid + subvol uuid)
3987 trans = btrfs_start_transaction(root, 3);
3988 if (IS_ERR(trans)) {
3989 ret = PTR_ERR(trans);
3994 sa->rtransid = trans->transid;
3995 sa->rtime.sec = ct.tv_sec;
3996 sa->rtime.nsec = ct.tv_nsec;
3998 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4000 if (received_uuid_changed &&
4001 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4002 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4003 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4004 root->root_key.objectid);
4005 if (ret && ret != -ENOENT) {
4006 btrfs_abort_transaction(trans, ret);
4007 btrfs_end_transaction(trans);
4011 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4012 btrfs_set_root_stransid(root_item, sa->stransid);
4013 btrfs_set_root_rtransid(root_item, sa->rtransid);
4014 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4015 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4016 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4017 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4019 ret = btrfs_update_root(trans, fs_info->tree_root,
4020 &root->root_key, &root->root_item);
4022 btrfs_end_transaction(trans);
4025 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4026 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4027 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4028 root->root_key.objectid);
4029 if (ret < 0 && ret != -EEXIST) {
4030 btrfs_abort_transaction(trans, ret);
4031 btrfs_end_transaction(trans);
4035 ret = btrfs_commit_transaction(trans);
4037 up_write(&fs_info->subvol_sem);
4038 mnt_drop_write_file(file);
4043 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4046 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4047 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4050 args32 = memdup_user(arg, sizeof(*args32));
4052 return PTR_ERR(args32);
4054 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4060 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4061 args64->stransid = args32->stransid;
4062 args64->rtransid = args32->rtransid;
4063 args64->stime.sec = args32->stime.sec;
4064 args64->stime.nsec = args32->stime.nsec;
4065 args64->rtime.sec = args32->rtime.sec;
4066 args64->rtime.nsec = args32->rtime.nsec;
4067 args64->flags = args32->flags;
4069 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4073 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4074 args32->stransid = args64->stransid;
4075 args32->rtransid = args64->rtransid;
4076 args32->stime.sec = args64->stime.sec;
4077 args32->stime.nsec = args64->stime.nsec;
4078 args32->rtime.sec = args64->rtime.sec;
4079 args32->rtime.nsec = args64->rtime.nsec;
4080 args32->flags = args64->flags;
4082 ret = copy_to_user(arg, args32, sizeof(*args32));
4093 static long btrfs_ioctl_set_received_subvol(struct file *file,
4096 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4099 sa = memdup_user(arg, sizeof(*sa));
4103 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4108 ret = copy_to_user(arg, sa, sizeof(*sa));
4117 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4122 char label[BTRFS_LABEL_SIZE];
4124 spin_lock(&fs_info->super_lock);
4125 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4126 spin_unlock(&fs_info->super_lock);
4128 len = strnlen(label, BTRFS_LABEL_SIZE);
4130 if (len == BTRFS_LABEL_SIZE) {
4132 "label is too long, return the first %zu bytes",
4136 ret = copy_to_user(arg, label, len);
4138 return ret ? -EFAULT : 0;
4141 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4143 struct inode *inode = file_inode(file);
4144 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4145 struct btrfs_root *root = BTRFS_I(inode)->root;
4146 struct btrfs_super_block *super_block = fs_info->super_copy;
4147 struct btrfs_trans_handle *trans;
4148 char label[BTRFS_LABEL_SIZE];
4151 if (!capable(CAP_SYS_ADMIN))
4154 if (copy_from_user(label, arg, sizeof(label)))
4157 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4159 "unable to set label with more than %d bytes",
4160 BTRFS_LABEL_SIZE - 1);
4164 ret = mnt_want_write_file(file);
4168 trans = btrfs_start_transaction(root, 0);
4169 if (IS_ERR(trans)) {
4170 ret = PTR_ERR(trans);
4174 spin_lock(&fs_info->super_lock);
4175 strcpy(super_block->label, label);
4176 spin_unlock(&fs_info->super_lock);
4177 ret = btrfs_commit_transaction(trans);
4180 mnt_drop_write_file(file);
4184 #define INIT_FEATURE_FLAGS(suffix) \
4185 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4186 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4187 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4189 int btrfs_ioctl_get_supported_features(void __user *arg)
4191 static const struct btrfs_ioctl_feature_flags features[3] = {
4192 INIT_FEATURE_FLAGS(SUPP),
4193 INIT_FEATURE_FLAGS(SAFE_SET),
4194 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4197 if (copy_to_user(arg, &features, sizeof(features)))
4203 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4206 struct btrfs_super_block *super_block = fs_info->super_copy;
4207 struct btrfs_ioctl_feature_flags features;
4209 features.compat_flags = btrfs_super_compat_flags(super_block);
4210 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4211 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4213 if (copy_to_user(arg, &features, sizeof(features)))
4219 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4220 enum btrfs_feature_set set,
4221 u64 change_mask, u64 flags, u64 supported_flags,
4222 u64 safe_set, u64 safe_clear)
4224 const char *type = btrfs_feature_set_name(set);
4226 u64 disallowed, unsupported;
4227 u64 set_mask = flags & change_mask;
4228 u64 clear_mask = ~flags & change_mask;
4230 unsupported = set_mask & ~supported_flags;
4232 names = btrfs_printable_features(set, unsupported);
4235 "this kernel does not support the %s feature bit%s",
4236 names, strchr(names, ',') ? "s" : "");
4240 "this kernel does not support %s bits 0x%llx",
4245 disallowed = set_mask & ~safe_set;
4247 names = btrfs_printable_features(set, disallowed);
4250 "can't set the %s feature bit%s while mounted",
4251 names, strchr(names, ',') ? "s" : "");
4255 "can't set %s bits 0x%llx while mounted",
4260 disallowed = clear_mask & ~safe_clear;
4262 names = btrfs_printable_features(set, disallowed);
4265 "can't clear the %s feature bit%s while mounted",
4266 names, strchr(names, ',') ? "s" : "");
4270 "can't clear %s bits 0x%llx while mounted",
4278 #define check_feature(fs_info, change_mask, flags, mask_base) \
4279 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4280 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4281 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4282 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4284 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4286 struct inode *inode = file_inode(file);
4287 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4288 struct btrfs_root *root = BTRFS_I(inode)->root;
4289 struct btrfs_super_block *super_block = fs_info->super_copy;
4290 struct btrfs_ioctl_feature_flags flags[2];
4291 struct btrfs_trans_handle *trans;
4295 if (!capable(CAP_SYS_ADMIN))
4298 if (copy_from_user(flags, arg, sizeof(flags)))
4302 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4303 !flags[0].incompat_flags)
4306 ret = check_feature(fs_info, flags[0].compat_flags,
4307 flags[1].compat_flags, COMPAT);
4311 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4312 flags[1].compat_ro_flags, COMPAT_RO);
4316 ret = check_feature(fs_info, flags[0].incompat_flags,
4317 flags[1].incompat_flags, INCOMPAT);
4321 ret = mnt_want_write_file(file);
4325 trans = btrfs_start_transaction(root, 0);
4326 if (IS_ERR(trans)) {
4327 ret = PTR_ERR(trans);
4328 goto out_drop_write;
4331 spin_lock(&fs_info->super_lock);
4332 newflags = btrfs_super_compat_flags(super_block);
4333 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4334 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4335 btrfs_set_super_compat_flags(super_block, newflags);
4337 newflags = btrfs_super_compat_ro_flags(super_block);
4338 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4339 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4340 btrfs_set_super_compat_ro_flags(super_block, newflags);
4342 newflags = btrfs_super_incompat_flags(super_block);
4343 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4344 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4345 btrfs_set_super_incompat_flags(super_block, newflags);
4346 spin_unlock(&fs_info->super_lock);
4348 ret = btrfs_commit_transaction(trans);
4350 mnt_drop_write_file(file);
4355 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4357 struct btrfs_ioctl_send_args *arg;
4361 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4362 struct btrfs_ioctl_send_args_32 args32 = { 0 };
4364 ret = copy_from_user(&args32, argp, sizeof(args32));
4367 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4370 arg->send_fd = args32.send_fd;
4371 arg->clone_sources_count = args32.clone_sources_count;
4372 arg->clone_sources = compat_ptr(args32.clone_sources);
4373 arg->parent_root = args32.parent_root;
4374 arg->flags = args32.flags;
4375 arg->version = args32.version;
4376 memcpy(arg->reserved, args32.reserved,
4377 sizeof(args32.reserved));
4382 arg = memdup_user(argp, sizeof(*arg));
4384 return PTR_ERR(arg);
4386 ret = btrfs_ioctl_send(inode, arg);
4391 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4394 struct btrfs_ioctl_encoded_io_args args = { 0 };
4395 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4398 struct iovec iovstack[UIO_FASTIOV];
4399 struct iovec *iov = iovstack;
4400 struct iov_iter iter;
4405 if (!capable(CAP_SYS_ADMIN)) {
4411 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4412 struct btrfs_ioctl_encoded_io_args_32 args32;
4414 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4416 if (copy_from_user(&args32, argp, copy_end)) {
4420 args.iov = compat_ptr(args32.iov);
4421 args.iovcnt = args32.iovcnt;
4422 args.offset = args32.offset;
4423 args.flags = args32.flags;
4428 copy_end = copy_end_kernel;
4429 if (copy_from_user(&args, argp, copy_end)) {
4434 if (args.flags != 0) {
4439 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4444 if (iov_iter_count(&iter) == 0) {
4449 ret = rw_verify_area(READ, file, &pos, args.len);
4453 init_sync_kiocb(&kiocb, file);
4456 ret = btrfs_encoded_read(&kiocb, &iter, &args);
4458 fsnotify_access(file);
4459 if (copy_to_user(argp + copy_end,
4460 (char *)&args + copy_end_kernel,
4461 sizeof(args) - copy_end_kernel))
4469 add_rchar(current, ret);
4474 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4476 struct btrfs_ioctl_encoded_io_args args;
4477 struct iovec iovstack[UIO_FASTIOV];
4478 struct iovec *iov = iovstack;
4479 struct iov_iter iter;
4484 if (!capable(CAP_SYS_ADMIN)) {
4489 if (!(file->f_mode & FMODE_WRITE)) {
4495 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4496 struct btrfs_ioctl_encoded_io_args_32 args32;
4498 if (copy_from_user(&args32, argp, sizeof(args32))) {
4502 args.iov = compat_ptr(args32.iov);
4503 args.iovcnt = args32.iovcnt;
4504 args.offset = args32.offset;
4505 args.flags = args32.flags;
4506 args.len = args32.len;
4507 args.unencoded_len = args32.unencoded_len;
4508 args.unencoded_offset = args32.unencoded_offset;
4509 args.compression = args32.compression;
4510 args.encryption = args32.encryption;
4511 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4516 if (copy_from_user(&args, argp, sizeof(args))) {
4523 if (args.flags != 0)
4525 if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4527 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4528 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4530 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4531 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4533 if (args.unencoded_offset > args.unencoded_len)
4535 if (args.len > args.unencoded_len - args.unencoded_offset)
4538 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4543 if (iov_iter_count(&iter) == 0) {
4548 ret = rw_verify_area(WRITE, file, &pos, args.len);
4552 init_sync_kiocb(&kiocb, file);
4553 ret = kiocb_set_rw_flags(&kiocb, 0);
4558 file_start_write(file);
4560 ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4562 fsnotify_modify(file);
4564 file_end_write(file);
4569 add_wchar(current, ret);
4574 long btrfs_ioctl(struct file *file, unsigned int
4575 cmd, unsigned long arg)
4577 struct inode *inode = file_inode(file);
4578 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4579 struct btrfs_root *root = BTRFS_I(inode)->root;
4580 void __user *argp = (void __user *)arg;
4583 case FS_IOC_GETVERSION:
4584 return btrfs_ioctl_getversion(inode, argp);
4585 case FS_IOC_GETFSLABEL:
4586 return btrfs_ioctl_get_fslabel(fs_info, argp);
4587 case FS_IOC_SETFSLABEL:
4588 return btrfs_ioctl_set_fslabel(file, argp);
4590 return btrfs_ioctl_fitrim(fs_info, argp);
4591 case BTRFS_IOC_SNAP_CREATE:
4592 return btrfs_ioctl_snap_create(file, argp, 0);
4593 case BTRFS_IOC_SNAP_CREATE_V2:
4594 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4595 case BTRFS_IOC_SUBVOL_CREATE:
4596 return btrfs_ioctl_snap_create(file, argp, 1);
4597 case BTRFS_IOC_SUBVOL_CREATE_V2:
4598 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4599 case BTRFS_IOC_SNAP_DESTROY:
4600 return btrfs_ioctl_snap_destroy(file, argp, false);
4601 case BTRFS_IOC_SNAP_DESTROY_V2:
4602 return btrfs_ioctl_snap_destroy(file, argp, true);
4603 case BTRFS_IOC_SUBVOL_GETFLAGS:
4604 return btrfs_ioctl_subvol_getflags(inode, argp);
4605 case BTRFS_IOC_SUBVOL_SETFLAGS:
4606 return btrfs_ioctl_subvol_setflags(file, argp);
4607 case BTRFS_IOC_DEFAULT_SUBVOL:
4608 return btrfs_ioctl_default_subvol(file, argp);
4609 case BTRFS_IOC_DEFRAG:
4610 return btrfs_ioctl_defrag(file, NULL);
4611 case BTRFS_IOC_DEFRAG_RANGE:
4612 return btrfs_ioctl_defrag(file, argp);
4613 case BTRFS_IOC_RESIZE:
4614 return btrfs_ioctl_resize(file, argp);
4615 case BTRFS_IOC_ADD_DEV:
4616 return btrfs_ioctl_add_dev(fs_info, argp);
4617 case BTRFS_IOC_RM_DEV:
4618 return btrfs_ioctl_rm_dev(file, argp);
4619 case BTRFS_IOC_RM_DEV_V2:
4620 return btrfs_ioctl_rm_dev_v2(file, argp);
4621 case BTRFS_IOC_FS_INFO:
4622 return btrfs_ioctl_fs_info(fs_info, argp);
4623 case BTRFS_IOC_DEV_INFO:
4624 return btrfs_ioctl_dev_info(fs_info, argp);
4625 case BTRFS_IOC_TREE_SEARCH:
4626 return btrfs_ioctl_tree_search(inode, argp);
4627 case BTRFS_IOC_TREE_SEARCH_V2:
4628 return btrfs_ioctl_tree_search_v2(inode, argp);
4629 case BTRFS_IOC_INO_LOOKUP:
4630 return btrfs_ioctl_ino_lookup(root, argp);
4631 case BTRFS_IOC_INO_PATHS:
4632 return btrfs_ioctl_ino_to_path(root, argp);
4633 case BTRFS_IOC_LOGICAL_INO:
4634 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4635 case BTRFS_IOC_LOGICAL_INO_V2:
4636 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4637 case BTRFS_IOC_SPACE_INFO:
4638 return btrfs_ioctl_space_info(fs_info, argp);
4639 case BTRFS_IOC_SYNC: {
4642 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4645 ret = btrfs_sync_fs(inode->i_sb, 1);
4647 * The transaction thread may want to do more work,
4648 * namely it pokes the cleaner kthread that will start
4649 * processing uncleaned subvols.
4651 wake_up_process(fs_info->transaction_kthread);
4654 case BTRFS_IOC_START_SYNC:
4655 return btrfs_ioctl_start_sync(root, argp);
4656 case BTRFS_IOC_WAIT_SYNC:
4657 return btrfs_ioctl_wait_sync(fs_info, argp);
4658 case BTRFS_IOC_SCRUB:
4659 return btrfs_ioctl_scrub(file, argp);
4660 case BTRFS_IOC_SCRUB_CANCEL:
4661 return btrfs_ioctl_scrub_cancel(fs_info);
4662 case BTRFS_IOC_SCRUB_PROGRESS:
4663 return btrfs_ioctl_scrub_progress(fs_info, argp);
4664 case BTRFS_IOC_BALANCE_V2:
4665 return btrfs_ioctl_balance(file, argp);
4666 case BTRFS_IOC_BALANCE_CTL:
4667 return btrfs_ioctl_balance_ctl(fs_info, arg);
4668 case BTRFS_IOC_BALANCE_PROGRESS:
4669 return btrfs_ioctl_balance_progress(fs_info, argp);
4670 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4671 return btrfs_ioctl_set_received_subvol(file, argp);
4673 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4674 return btrfs_ioctl_set_received_subvol_32(file, argp);
4676 case BTRFS_IOC_SEND:
4677 return _btrfs_ioctl_send(inode, argp, false);
4678 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4679 case BTRFS_IOC_SEND_32:
4680 return _btrfs_ioctl_send(inode, argp, true);
4682 case BTRFS_IOC_GET_DEV_STATS:
4683 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4684 case BTRFS_IOC_QUOTA_CTL:
4685 return btrfs_ioctl_quota_ctl(file, argp);
4686 case BTRFS_IOC_QGROUP_ASSIGN:
4687 return btrfs_ioctl_qgroup_assign(file, argp);
4688 case BTRFS_IOC_QGROUP_CREATE:
4689 return btrfs_ioctl_qgroup_create(file, argp);
4690 case BTRFS_IOC_QGROUP_LIMIT:
4691 return btrfs_ioctl_qgroup_limit(file, argp);
4692 case BTRFS_IOC_QUOTA_RESCAN:
4693 return btrfs_ioctl_quota_rescan(file, argp);
4694 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4695 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4696 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4697 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4698 case BTRFS_IOC_DEV_REPLACE:
4699 return btrfs_ioctl_dev_replace(fs_info, argp);
4700 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4701 return btrfs_ioctl_get_supported_features(argp);
4702 case BTRFS_IOC_GET_FEATURES:
4703 return btrfs_ioctl_get_features(fs_info, argp);
4704 case BTRFS_IOC_SET_FEATURES:
4705 return btrfs_ioctl_set_features(file, argp);
4706 case BTRFS_IOC_GET_SUBVOL_INFO:
4707 return btrfs_ioctl_get_subvol_info(inode, argp);
4708 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4709 return btrfs_ioctl_get_subvol_rootref(root, argp);
4710 case BTRFS_IOC_INO_LOOKUP_USER:
4711 return btrfs_ioctl_ino_lookup_user(file, argp);
4712 case FS_IOC_ENABLE_VERITY:
4713 return fsverity_ioctl_enable(file, (const void __user *)argp);
4714 case FS_IOC_MEASURE_VERITY:
4715 return fsverity_ioctl_measure(file, argp);
4716 case BTRFS_IOC_ENCODED_READ:
4717 return btrfs_ioctl_encoded_read(file, argp, false);
4718 case BTRFS_IOC_ENCODED_WRITE:
4719 return btrfs_ioctl_encoded_write(file, argp, false);
4720 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4721 case BTRFS_IOC_ENCODED_READ_32:
4722 return btrfs_ioctl_encoded_read(file, argp, true);
4723 case BTRFS_IOC_ENCODED_WRITE_32:
4724 return btrfs_ioctl_encoded_write(file, argp, true);
4731 #ifdef CONFIG_COMPAT
4732 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4735 * These all access 32-bit values anyway so no further
4736 * handling is necessary.
4739 case FS_IOC32_GETVERSION:
4740 cmd = FS_IOC_GETVERSION;
4744 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));