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"
41 #include "dev-replace.h"
46 #include "compression.h"
47 #include "space-info.h"
48 #include "block-group.h"
50 #include "accessors.h"
51 #include "extent-tree.h"
52 #include "root-tree.h"
55 #include "uuid-tree.h"
62 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
63 * structures are incorrect, as the timespec structure from userspace
64 * is 4 bytes too small. We define these alternatives here to teach
65 * the kernel about the 32-bit struct packing.
67 struct btrfs_ioctl_timespec_32 {
70 } __attribute__ ((__packed__));
72 struct btrfs_ioctl_received_subvol_args_32 {
73 char uuid[BTRFS_UUID_SIZE]; /* in */
74 __u64 stransid; /* in */
75 __u64 rtransid; /* out */
76 struct btrfs_ioctl_timespec_32 stime; /* in */
77 struct btrfs_ioctl_timespec_32 rtime; /* out */
79 __u64 reserved[16]; /* in */
80 } __attribute__ ((__packed__));
82 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
83 struct btrfs_ioctl_received_subvol_args_32)
86 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
87 struct btrfs_ioctl_send_args_32 {
88 __s64 send_fd; /* in */
89 __u64 clone_sources_count; /* in */
90 compat_uptr_t clone_sources; /* in */
91 __u64 parent_root; /* in */
93 __u32 version; /* in */
94 __u8 reserved[28]; /* in */
95 } __attribute__ ((__packed__));
97 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
98 struct btrfs_ioctl_send_args_32)
100 struct btrfs_ioctl_encoded_io_args_32 {
102 compat_ulong_t iovcnt;
107 __u64 unencoded_offset;
113 #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
114 struct btrfs_ioctl_encoded_io_args_32)
115 #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
116 struct btrfs_ioctl_encoded_io_args_32)
119 /* Mask out flags that are inappropriate for the given type of inode. */
120 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
123 if (S_ISDIR(inode->i_mode))
125 else if (S_ISREG(inode->i_mode))
126 return flags & ~FS_DIRSYNC_FL;
128 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
132 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
135 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
137 unsigned int iflags = 0;
138 u32 flags = binode->flags;
139 u32 ro_flags = binode->ro_flags;
141 if (flags & BTRFS_INODE_SYNC)
142 iflags |= FS_SYNC_FL;
143 if (flags & BTRFS_INODE_IMMUTABLE)
144 iflags |= FS_IMMUTABLE_FL;
145 if (flags & BTRFS_INODE_APPEND)
146 iflags |= FS_APPEND_FL;
147 if (flags & BTRFS_INODE_NODUMP)
148 iflags |= FS_NODUMP_FL;
149 if (flags & BTRFS_INODE_NOATIME)
150 iflags |= FS_NOATIME_FL;
151 if (flags & BTRFS_INODE_DIRSYNC)
152 iflags |= FS_DIRSYNC_FL;
153 if (flags & BTRFS_INODE_NODATACOW)
154 iflags |= FS_NOCOW_FL;
155 if (ro_flags & BTRFS_INODE_RO_VERITY)
156 iflags |= FS_VERITY_FL;
158 if (flags & BTRFS_INODE_NOCOMPRESS)
159 iflags |= FS_NOCOMP_FL;
160 else if (flags & BTRFS_INODE_COMPRESS)
161 iflags |= FS_COMPR_FL;
167 * Update inode->i_flags based on the btrfs internal flags.
169 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
171 struct btrfs_inode *binode = BTRFS_I(inode);
172 unsigned int new_fl = 0;
174 if (binode->flags & BTRFS_INODE_SYNC)
176 if (binode->flags & BTRFS_INODE_IMMUTABLE)
177 new_fl |= S_IMMUTABLE;
178 if (binode->flags & BTRFS_INODE_APPEND)
180 if (binode->flags & BTRFS_INODE_NOATIME)
182 if (binode->flags & BTRFS_INODE_DIRSYNC)
184 if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
187 set_mask_bits(&inode->i_flags,
188 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
193 * Check if @flags are a supported and valid set of FS_*_FL flags and that
194 * the old and new flags are not conflicting
196 static int check_fsflags(unsigned int old_flags, unsigned int flags)
198 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
199 FS_NOATIME_FL | FS_NODUMP_FL | \
200 FS_SYNC_FL | FS_DIRSYNC_FL | \
201 FS_NOCOMP_FL | FS_COMPR_FL |
205 /* COMPR and NOCOMP on new/old are valid */
206 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
209 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
212 /* NOCOW and compression options are mutually exclusive */
213 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
215 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
221 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
224 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
230 int btrfs_check_ioctl_vol_args_path(const struct btrfs_ioctl_vol_args *vol_args)
232 if (memchr(vol_args->name, 0, sizeof(vol_args->name)) == NULL)
233 return -ENAMETOOLONG;
237 static int btrfs_check_ioctl_vol_args2_subvol_name(const struct btrfs_ioctl_vol_args_v2 *vol_args2)
239 if (memchr(vol_args2->name, 0, sizeof(vol_args2->name)) == NULL)
240 return -ENAMETOOLONG;
245 * Set flags/xflags from the internal inode flags. The remaining items of
246 * fsxattr are zeroed.
248 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
250 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
252 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
256 int btrfs_fileattr_set(struct mnt_idmap *idmap,
257 struct dentry *dentry, struct fileattr *fa)
259 struct inode *inode = d_inode(dentry);
260 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
261 struct btrfs_inode *binode = BTRFS_I(inode);
262 struct btrfs_root *root = binode->root;
263 struct btrfs_trans_handle *trans;
264 unsigned int fsflags, old_fsflags;
266 const char *comp = NULL;
269 if (btrfs_root_readonly(root))
272 if (fileattr_has_fsx(fa))
275 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
276 old_fsflags = btrfs_inode_flags_to_fsflags(binode);
277 ret = check_fsflags(old_fsflags, fsflags);
281 ret = check_fsflags_compatible(fs_info, fsflags);
285 binode_flags = binode->flags;
286 if (fsflags & FS_SYNC_FL)
287 binode_flags |= BTRFS_INODE_SYNC;
289 binode_flags &= ~BTRFS_INODE_SYNC;
290 if (fsflags & FS_IMMUTABLE_FL)
291 binode_flags |= BTRFS_INODE_IMMUTABLE;
293 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
294 if (fsflags & FS_APPEND_FL)
295 binode_flags |= BTRFS_INODE_APPEND;
297 binode_flags &= ~BTRFS_INODE_APPEND;
298 if (fsflags & FS_NODUMP_FL)
299 binode_flags |= BTRFS_INODE_NODUMP;
301 binode_flags &= ~BTRFS_INODE_NODUMP;
302 if (fsflags & FS_NOATIME_FL)
303 binode_flags |= BTRFS_INODE_NOATIME;
305 binode_flags &= ~BTRFS_INODE_NOATIME;
307 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
308 if (!fa->flags_valid) {
309 /* 1 item for the inode */
310 trans = btrfs_start_transaction(root, 1);
312 return PTR_ERR(trans);
316 if (fsflags & FS_DIRSYNC_FL)
317 binode_flags |= BTRFS_INODE_DIRSYNC;
319 binode_flags &= ~BTRFS_INODE_DIRSYNC;
320 if (fsflags & FS_NOCOW_FL) {
321 if (S_ISREG(inode->i_mode)) {
323 * It's safe to turn csums off here, no extents exist.
324 * Otherwise we want the flag to reflect the real COW
325 * status of the file and will not set it.
327 if (inode->i_size == 0)
328 binode_flags |= BTRFS_INODE_NODATACOW |
329 BTRFS_INODE_NODATASUM;
331 binode_flags |= BTRFS_INODE_NODATACOW;
335 * Revert back under same assumptions as above
337 if (S_ISREG(inode->i_mode)) {
338 if (inode->i_size == 0)
339 binode_flags &= ~(BTRFS_INODE_NODATACOW |
340 BTRFS_INODE_NODATASUM);
342 binode_flags &= ~BTRFS_INODE_NODATACOW;
347 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
348 * flag may be changed automatically if compression code won't make
351 if (fsflags & FS_NOCOMP_FL) {
352 binode_flags &= ~BTRFS_INODE_COMPRESS;
353 binode_flags |= BTRFS_INODE_NOCOMPRESS;
354 } else if (fsflags & FS_COMPR_FL) {
356 if (IS_SWAPFILE(inode))
359 binode_flags |= BTRFS_INODE_COMPRESS;
360 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
362 comp = btrfs_compress_type2str(fs_info->compress_type);
363 if (!comp || comp[0] == 0)
364 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
366 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
373 trans = btrfs_start_transaction(root, 3);
375 return PTR_ERR(trans);
378 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
381 btrfs_abort_transaction(trans, ret);
385 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
387 if (ret && ret != -ENODATA) {
388 btrfs_abort_transaction(trans, ret);
394 binode->flags = binode_flags;
395 btrfs_sync_inode_flags_to_i_flags(inode);
396 inode_inc_iversion(inode);
397 inode_set_ctime_current(inode);
398 ret = btrfs_update_inode(trans, BTRFS_I(inode));
401 btrfs_end_transaction(trans);
406 * Start exclusive operation @type, return true on success
408 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
409 enum btrfs_exclusive_operation type)
413 spin_lock(&fs_info->super_lock);
414 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
415 fs_info->exclusive_operation = type;
418 spin_unlock(&fs_info->super_lock);
424 * Conditionally allow to enter the exclusive operation in case it's compatible
425 * with the running one. This must be paired with btrfs_exclop_start_unlock and
426 * btrfs_exclop_finish.
429 * - the same type is already running
430 * - when trying to add a device and balance has been paused
431 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
432 * must check the condition first that would allow none -> @type
434 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
435 enum btrfs_exclusive_operation type)
437 spin_lock(&fs_info->super_lock);
438 if (fs_info->exclusive_operation == type ||
439 (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
440 type == BTRFS_EXCLOP_DEV_ADD))
443 spin_unlock(&fs_info->super_lock);
447 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
449 spin_unlock(&fs_info->super_lock);
452 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
454 spin_lock(&fs_info->super_lock);
455 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
456 spin_unlock(&fs_info->super_lock);
457 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
460 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
461 enum btrfs_exclusive_operation op)
464 case BTRFS_EXCLOP_BALANCE_PAUSED:
465 spin_lock(&fs_info->super_lock);
466 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
467 fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD ||
468 fs_info->exclusive_operation == BTRFS_EXCLOP_NONE ||
469 fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
470 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
471 spin_unlock(&fs_info->super_lock);
473 case BTRFS_EXCLOP_BALANCE:
474 spin_lock(&fs_info->super_lock);
475 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
476 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
477 spin_unlock(&fs_info->super_lock);
481 "invalid exclop balance operation %d requested", op);
485 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
487 return put_user(inode->i_generation, arg);
490 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
493 struct btrfs_device *device;
494 struct fstrim_range range;
495 u64 minlen = ULLONG_MAX;
499 if (!capable(CAP_SYS_ADMIN))
503 * btrfs_trim_block_group() depends on space cache, which is not
504 * available in zoned filesystem. So, disallow fitrim on a zoned
505 * filesystem for now.
507 if (btrfs_is_zoned(fs_info))
511 * If the fs is mounted with nologreplay, which requires it to be
512 * mounted in RO mode as well, we can not allow discard on free space
513 * inside block groups, because log trees refer to extents that are not
514 * pinned in a block group's free space cache (pinning the extents is
515 * precisely the first phase of replaying a log tree).
517 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
521 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
523 if (!device->bdev || !bdev_max_discard_sectors(device->bdev))
526 minlen = min_t(u64, bdev_discard_granularity(device->bdev),
533 if (copy_from_user(&range, arg, sizeof(range)))
537 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
538 * block group is in the logical address space, which can be any
539 * sectorsize aligned bytenr in the range [0, U64_MAX].
541 if (range.len < fs_info->sectorsize)
544 range.minlen = max(range.minlen, minlen);
545 ret = btrfs_trim_fs(fs_info, &range);
549 if (copy_to_user(arg, &range, sizeof(range)))
555 int __pure btrfs_is_empty_uuid(u8 *uuid)
559 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
567 * Calculate the number of transaction items to reserve for creating a subvolume
568 * or snapshot, not including the inode, directory entries, or parent directory.
570 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit *inherit)
573 * 1 to add root block
576 * 1 to add root backref
578 * 1 to add qgroup info
579 * 1 to add qgroup limit
581 * Ideally the last two would only be accounted if qgroups are enabled,
582 * but that can change between now and the time we would insert them.
584 unsigned int num_items = 7;
587 /* 2 to add qgroup relations for each inherited qgroup */
588 num_items += 2 * inherit->num_qgroups;
593 static noinline int create_subvol(struct mnt_idmap *idmap,
594 struct inode *dir, struct dentry *dentry,
595 struct btrfs_qgroup_inherit *inherit)
597 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
598 struct btrfs_trans_handle *trans;
599 struct btrfs_key key;
600 struct btrfs_root_item *root_item;
601 struct btrfs_inode_item *inode_item;
602 struct extent_buffer *leaf;
603 struct btrfs_root *root = BTRFS_I(dir)->root;
604 struct btrfs_root *new_root;
605 struct btrfs_block_rsv block_rsv;
606 struct timespec64 cur_time = current_time(dir);
607 struct btrfs_new_inode_args new_inode_args = {
612 unsigned int trans_num_items;
617 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
621 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
626 * Don't create subvolume whose level is not zero. Or qgroup will be
627 * screwed up since it assumes subvolume qgroup's level to be 0.
629 if (btrfs_qgroup_level(objectid)) {
634 ret = get_anon_bdev(&anon_dev);
638 new_inode_args.inode = btrfs_new_subvol_inode(idmap, dir);
639 if (!new_inode_args.inode) {
643 ret = btrfs_new_inode_prepare(&new_inode_args, &trans_num_items);
646 trans_num_items += create_subvol_num_items(inherit);
648 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
649 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
650 trans_num_items, false);
652 goto out_new_inode_args;
654 trans = btrfs_start_transaction(root, 0);
656 ret = PTR_ERR(trans);
657 btrfs_subvolume_release_metadata(root, &block_rsv);
658 goto out_new_inode_args;
660 trans->block_rsv = &block_rsv;
661 trans->bytes_reserved = block_rsv.size;
662 /* Tree log can't currently deal with an inode which is a new root. */
663 btrfs_set_log_full_commit(trans);
665 ret = btrfs_qgroup_inherit(trans, 0, objectid, root->root_key.objectid, inherit);
669 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
670 0, BTRFS_NESTING_NORMAL);
676 btrfs_mark_buffer_dirty(trans, leaf);
678 inode_item = &root_item->inode;
679 btrfs_set_stack_inode_generation(inode_item, 1);
680 btrfs_set_stack_inode_size(inode_item, 3);
681 btrfs_set_stack_inode_nlink(inode_item, 1);
682 btrfs_set_stack_inode_nbytes(inode_item,
684 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
686 btrfs_set_root_flags(root_item, 0);
687 btrfs_set_root_limit(root_item, 0);
688 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
690 btrfs_set_root_bytenr(root_item, leaf->start);
691 btrfs_set_root_generation(root_item, trans->transid);
692 btrfs_set_root_level(root_item, 0);
693 btrfs_set_root_refs(root_item, 1);
694 btrfs_set_root_used(root_item, leaf->len);
695 btrfs_set_root_last_snapshot(root_item, 0);
697 btrfs_set_root_generation_v2(root_item,
698 btrfs_root_generation(root_item));
699 generate_random_guid(root_item->uuid);
700 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
701 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
702 root_item->ctime = root_item->otime;
703 btrfs_set_root_ctransid(root_item, trans->transid);
704 btrfs_set_root_otransid(root_item, trans->transid);
706 btrfs_tree_unlock(leaf);
708 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
710 key.objectid = objectid;
712 key.type = BTRFS_ROOT_ITEM_KEY;
713 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
717 * Since we don't abort the transaction in this case, free the
718 * tree block so that we don't leak space and leave the
719 * filesystem in an inconsistent state (an extent item in the
720 * extent tree with a backreference for a root that does not
723 btrfs_tree_lock(leaf);
724 btrfs_clear_buffer_dirty(trans, leaf);
725 btrfs_tree_unlock(leaf);
726 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
727 free_extent_buffer(leaf);
731 free_extent_buffer(leaf);
734 new_root = btrfs_get_new_fs_root(fs_info, objectid, &anon_dev);
735 if (IS_ERR(new_root)) {
736 ret = PTR_ERR(new_root);
737 btrfs_abort_transaction(trans, ret);
740 /* anon_dev is owned by new_root now. */
742 BTRFS_I(new_inode_args.inode)->root = new_root;
743 /* ... and new_root is owned by new_inode_args.inode now. */
745 ret = btrfs_record_root_in_trans(trans, new_root);
747 btrfs_abort_transaction(trans, ret);
751 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
752 BTRFS_UUID_KEY_SUBVOL, objectid);
754 btrfs_abort_transaction(trans, ret);
758 ret = btrfs_create_new_inode(trans, &new_inode_args);
760 btrfs_abort_transaction(trans, ret);
764 d_instantiate_new(dentry, new_inode_args.inode);
765 new_inode_args.inode = NULL;
768 trans->block_rsv = NULL;
769 trans->bytes_reserved = 0;
770 btrfs_subvolume_release_metadata(root, &block_rsv);
772 btrfs_end_transaction(trans);
774 btrfs_new_inode_args_destroy(&new_inode_args);
776 iput(new_inode_args.inode);
779 free_anon_bdev(anon_dev);
785 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
786 struct dentry *dentry, bool readonly,
787 struct btrfs_qgroup_inherit *inherit)
789 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
791 struct btrfs_pending_snapshot *pending_snapshot;
792 unsigned int trans_num_items;
793 struct btrfs_trans_handle *trans;
796 /* We do not support snapshotting right now. */
797 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
799 "extent tree v2 doesn't support snapshotting yet");
803 if (btrfs_root_refs(&root->root_item) == 0)
806 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
809 if (atomic_read(&root->nr_swapfiles)) {
811 "cannot snapshot subvolume with active swapfile");
815 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
816 if (!pending_snapshot)
819 ret = get_anon_bdev(&pending_snapshot->anon_dev);
822 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
824 pending_snapshot->path = btrfs_alloc_path();
825 if (!pending_snapshot->root_item || !pending_snapshot->path) {
830 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
831 BTRFS_BLOCK_RSV_TEMP);
835 * 1 to update parent inode item
837 trans_num_items = create_subvol_num_items(inherit) + 3;
838 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
839 &pending_snapshot->block_rsv,
840 trans_num_items, false);
844 pending_snapshot->dentry = dentry;
845 pending_snapshot->root = root;
846 pending_snapshot->readonly = readonly;
847 pending_snapshot->dir = dir;
848 pending_snapshot->inherit = inherit;
850 trans = btrfs_start_transaction(root, 0);
852 ret = PTR_ERR(trans);
856 trans->pending_snapshot = pending_snapshot;
858 ret = btrfs_commit_transaction(trans);
862 ret = pending_snapshot->error;
866 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
870 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
872 ret = PTR_ERR(inode);
876 d_instantiate(dentry, inode);
878 pending_snapshot->anon_dev = 0;
880 /* Prevent double freeing of anon_dev */
881 if (ret && pending_snapshot->snap)
882 pending_snapshot->snap->anon_dev = 0;
883 btrfs_put_root(pending_snapshot->snap);
884 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
886 if (pending_snapshot->anon_dev)
887 free_anon_bdev(pending_snapshot->anon_dev);
888 kfree(pending_snapshot->root_item);
889 btrfs_free_path(pending_snapshot->path);
890 kfree(pending_snapshot);
895 /* copy of may_delete in fs/namei.c()
896 * Check whether we can remove a link victim from directory dir, check
897 * whether the type of victim is right.
898 * 1. We can't do it if dir is read-only (done in permission())
899 * 2. We should have write and exec permissions on dir
900 * 3. We can't remove anything from append-only dir
901 * 4. We can't do anything with immutable dir (done in permission())
902 * 5. If the sticky bit on dir is set we should either
903 * a. be owner of dir, or
904 * b. be owner of victim, or
905 * c. have CAP_FOWNER capability
906 * 6. If the victim is append-only or immutable we can't do anything with
907 * links pointing to it.
908 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
909 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
910 * 9. We can't remove a root or mountpoint.
911 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
912 * nfs_async_unlink().
915 static int btrfs_may_delete(struct mnt_idmap *idmap,
916 struct inode *dir, struct dentry *victim, int isdir)
920 if (d_really_is_negative(victim))
923 /* The @victim is not inside @dir. */
924 if (d_inode(victim->d_parent) != dir)
926 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
928 error = inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
933 if (check_sticky(idmap, dir, d_inode(victim)) ||
934 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
935 IS_SWAPFILE(d_inode(victim)))
938 if (!d_is_dir(victim))
942 } else if (d_is_dir(victim))
946 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
951 /* copy of may_create in fs/namei.c() */
952 static inline int btrfs_may_create(struct mnt_idmap *idmap,
953 struct inode *dir, struct dentry *child)
955 if (d_really_is_positive(child))
959 if (!fsuidgid_has_mapping(dir->i_sb, idmap))
961 return inode_permission(idmap, dir, MAY_WRITE | MAY_EXEC);
965 * Create a new subvolume below @parent. This is largely modeled after
966 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
967 * inside this filesystem so it's quite a bit simpler.
969 static noinline int btrfs_mksubvol(const struct path *parent,
970 struct mnt_idmap *idmap,
971 const char *name, int namelen,
972 struct btrfs_root *snap_src,
974 struct btrfs_qgroup_inherit *inherit)
976 struct inode *dir = d_inode(parent->dentry);
977 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
978 struct dentry *dentry;
979 struct fscrypt_str name_str = FSTR_INIT((char *)name, namelen);
982 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
986 dentry = lookup_one(idmap, name, parent->dentry, namelen);
987 error = PTR_ERR(dentry);
991 error = btrfs_may_create(idmap, dir, dentry);
996 * even if this name doesn't exist, we may get hash collisions.
997 * check for them now when we can safely fail
999 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
1000 dir->i_ino, &name_str);
1004 down_read(&fs_info->subvol_sem);
1006 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
1010 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
1012 error = create_subvol(idmap, dir, dentry, inherit);
1015 fsnotify_mkdir(dir, dentry);
1017 up_read(&fs_info->subvol_sem);
1021 btrfs_inode_unlock(BTRFS_I(dir), 0);
1025 static noinline int btrfs_mksnapshot(const struct path *parent,
1026 struct mnt_idmap *idmap,
1027 const char *name, int namelen,
1028 struct btrfs_root *root,
1030 struct btrfs_qgroup_inherit *inherit)
1033 bool snapshot_force_cow = false;
1036 * Force new buffered writes to reserve space even when NOCOW is
1037 * possible. This is to avoid later writeback (running dealloc) to
1038 * fallback to COW mode and unexpectedly fail with ENOSPC.
1040 btrfs_drew_read_lock(&root->snapshot_lock);
1042 ret = btrfs_start_delalloc_snapshot(root, false);
1047 * All previous writes have started writeback in NOCOW mode, so now
1048 * we force future writes to fallback to COW mode during snapshot
1051 atomic_inc(&root->snapshot_force_cow);
1052 snapshot_force_cow = true;
1054 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1056 ret = btrfs_mksubvol(parent, idmap, name, namelen,
1057 root, readonly, inherit);
1059 if (snapshot_force_cow)
1060 atomic_dec(&root->snapshot_force_cow);
1061 btrfs_drew_read_unlock(&root->snapshot_lock);
1066 * Try to start exclusive operation @type or cancel it if it's running.
1069 * 0 - normal mode, newly claimed op started
1070 * >0 - normal mode, something else is running,
1071 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1072 * ECANCELED - cancel mode, successful cancel
1073 * ENOTCONN - cancel mode, operation not running anymore
1075 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1076 enum btrfs_exclusive_operation type, bool cancel)
1079 /* Start normal op */
1080 if (!btrfs_exclop_start(fs_info, type))
1081 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1082 /* Exclusive operation is now claimed */
1086 /* Cancel running op */
1087 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1089 * This blocks any exclop finish from setting it to NONE, so we
1090 * request cancellation. Either it runs and we will wait for it,
1091 * or it has finished and no waiting will happen.
1093 atomic_inc(&fs_info->reloc_cancel_req);
1094 btrfs_exclop_start_unlock(fs_info);
1096 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1097 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1098 TASK_INTERRUPTIBLE);
1103 /* Something else is running or none */
1107 static noinline int btrfs_ioctl_resize(struct file *file,
1110 BTRFS_DEV_LOOKUP_ARGS(args);
1111 struct inode *inode = file_inode(file);
1112 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1116 struct btrfs_root *root = BTRFS_I(inode)->root;
1117 struct btrfs_ioctl_vol_args *vol_args;
1118 struct btrfs_trans_handle *trans;
1119 struct btrfs_device *device = NULL;
1122 char *devstr = NULL;
1127 if (!capable(CAP_SYS_ADMIN))
1130 ret = mnt_want_write_file(file);
1135 * Read the arguments before checking exclusivity to be able to
1136 * distinguish regular resize and cancel
1138 vol_args = memdup_user(arg, sizeof(*vol_args));
1139 if (IS_ERR(vol_args)) {
1140 ret = PTR_ERR(vol_args);
1143 ret = btrfs_check_ioctl_vol_args_path(vol_args);
1147 sizestr = vol_args->name;
1148 cancel = (strcmp("cancel", sizestr) == 0);
1149 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1152 /* Exclusive operation is now claimed */
1154 devstr = strchr(sizestr, ':');
1156 sizestr = devstr + 1;
1158 devstr = vol_args->name;
1159 ret = kstrtoull(devstr, 10, &devid);
1166 btrfs_info(fs_info, "resizing devid %llu", devid);
1170 device = btrfs_find_device(fs_info->fs_devices, &args);
1172 btrfs_info(fs_info, "resizer unable to find device %llu",
1178 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1180 "resizer unable to apply on readonly device %llu",
1186 if (!strcmp(sizestr, "max"))
1187 new_size = bdev_nr_bytes(device->bdev);
1189 if (sizestr[0] == '-') {
1192 } else if (sizestr[0] == '+') {
1196 new_size = memparse(sizestr, &retptr);
1197 if (*retptr != '\0' || new_size == 0) {
1203 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1208 old_size = btrfs_device_get_total_bytes(device);
1211 if (new_size > old_size) {
1215 new_size = old_size - new_size;
1216 } else if (mod > 0) {
1217 if (new_size > ULLONG_MAX - old_size) {
1221 new_size = old_size + new_size;
1224 if (new_size < SZ_256M) {
1228 if (new_size > bdev_nr_bytes(device->bdev)) {
1233 new_size = round_down(new_size, fs_info->sectorsize);
1235 if (new_size > old_size) {
1236 trans = btrfs_start_transaction(root, 0);
1237 if (IS_ERR(trans)) {
1238 ret = PTR_ERR(trans);
1241 ret = btrfs_grow_device(trans, device, new_size);
1242 btrfs_commit_transaction(trans);
1243 } else if (new_size < old_size) {
1244 ret = btrfs_shrink_device(device, new_size);
1245 } /* equal, nothing need to do */
1247 if (ret == 0 && new_size != old_size)
1248 btrfs_info_in_rcu(fs_info,
1249 "resize device %s (devid %llu) from %llu to %llu",
1250 btrfs_dev_name(device), device->devid,
1251 old_size, new_size);
1253 btrfs_exclop_finish(fs_info);
1257 mnt_drop_write_file(file);
1261 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1262 struct mnt_idmap *idmap,
1263 const char *name, unsigned long fd, int subvol,
1265 struct btrfs_qgroup_inherit *inherit)
1270 if (!S_ISDIR(file_inode(file)->i_mode))
1273 ret = mnt_want_write_file(file);
1277 namelen = strlen(name);
1278 if (strchr(name, '/')) {
1280 goto out_drop_write;
1283 if (name[0] == '.' &&
1284 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1286 goto out_drop_write;
1290 ret = btrfs_mksubvol(&file->f_path, idmap, name,
1291 namelen, NULL, readonly, inherit);
1293 struct fd src = fdget(fd);
1294 struct inode *src_inode;
1297 goto out_drop_write;
1300 src_inode = file_inode(src.file);
1301 if (src_inode->i_sb != file_inode(file)->i_sb) {
1302 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1303 "Snapshot src from another FS");
1305 } else if (!inode_owner_or_capable(idmap, src_inode)) {
1307 * Subvolume creation is not restricted, but snapshots
1308 * are limited to own subvolumes only
1311 } else if (btrfs_ino(BTRFS_I(src_inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1313 * Snapshots must be made with the src_inode referring
1314 * to the subvolume inode, otherwise the permission
1315 * checking above is useless because we may have
1316 * permission on a lower directory but not the subvol
1321 ret = btrfs_mksnapshot(&file->f_path, idmap,
1323 BTRFS_I(src_inode)->root,
1329 mnt_drop_write_file(file);
1334 static noinline int btrfs_ioctl_snap_create(struct file *file,
1335 void __user *arg, int subvol)
1337 struct btrfs_ioctl_vol_args *vol_args;
1340 if (!S_ISDIR(file_inode(file)->i_mode))
1343 vol_args = memdup_user(arg, sizeof(*vol_args));
1344 if (IS_ERR(vol_args))
1345 return PTR_ERR(vol_args);
1346 ret = btrfs_check_ioctl_vol_args_path(vol_args);
1350 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1351 vol_args->name, vol_args->fd, subvol,
1359 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1360 void __user *arg, int subvol)
1362 struct btrfs_ioctl_vol_args_v2 *vol_args;
1364 bool readonly = false;
1365 struct btrfs_qgroup_inherit *inherit = NULL;
1367 if (!S_ISDIR(file_inode(file)->i_mode))
1370 vol_args = memdup_user(arg, sizeof(*vol_args));
1371 if (IS_ERR(vol_args))
1372 return PTR_ERR(vol_args);
1373 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
1377 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1382 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1384 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1385 struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
1387 if (vol_args->size < sizeof(*inherit) ||
1388 vol_args->size > PAGE_SIZE) {
1392 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1393 if (IS_ERR(inherit)) {
1394 ret = PTR_ERR(inherit);
1398 ret = btrfs_qgroup_check_inherit(fs_info, inherit, vol_args->size);
1403 ret = __btrfs_ioctl_snap_create(file, file_mnt_idmap(file),
1404 vol_args->name, vol_args->fd, subvol,
1415 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
1418 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1419 struct btrfs_root *root = BTRFS_I(inode)->root;
1423 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1426 down_read(&fs_info->subvol_sem);
1427 if (btrfs_root_readonly(root))
1428 flags |= BTRFS_SUBVOL_RDONLY;
1429 up_read(&fs_info->subvol_sem);
1431 if (copy_to_user(arg, &flags, sizeof(flags)))
1437 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1440 struct inode *inode = file_inode(file);
1441 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
1442 struct btrfs_root *root = BTRFS_I(inode)->root;
1443 struct btrfs_trans_handle *trans;
1448 if (!inode_owner_or_capable(file_mnt_idmap(file), inode))
1451 ret = mnt_want_write_file(file);
1455 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1457 goto out_drop_write;
1460 if (copy_from_user(&flags, arg, sizeof(flags))) {
1462 goto out_drop_write;
1465 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1467 goto out_drop_write;
1470 down_write(&fs_info->subvol_sem);
1473 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1476 root_flags = btrfs_root_flags(&root->root_item);
1477 if (flags & BTRFS_SUBVOL_RDONLY) {
1478 btrfs_set_root_flags(&root->root_item,
1479 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1482 * Block RO -> RW transition if this subvolume is involved in
1485 spin_lock(&root->root_item_lock);
1486 if (root->send_in_progress == 0) {
1487 btrfs_set_root_flags(&root->root_item,
1488 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1489 spin_unlock(&root->root_item_lock);
1491 spin_unlock(&root->root_item_lock);
1493 "Attempt to set subvolume %llu read-write during send",
1494 root->root_key.objectid);
1500 trans = btrfs_start_transaction(root, 1);
1501 if (IS_ERR(trans)) {
1502 ret = PTR_ERR(trans);
1506 ret = btrfs_update_root(trans, fs_info->tree_root,
1507 &root->root_key, &root->root_item);
1509 btrfs_end_transaction(trans);
1513 ret = btrfs_commit_transaction(trans);
1517 btrfs_set_root_flags(&root->root_item, root_flags);
1519 up_write(&fs_info->subvol_sem);
1521 mnt_drop_write_file(file);
1526 static noinline int key_in_sk(struct btrfs_key *key,
1527 struct btrfs_ioctl_search_key *sk)
1529 struct btrfs_key test;
1532 test.objectid = sk->min_objectid;
1533 test.type = sk->min_type;
1534 test.offset = sk->min_offset;
1536 ret = btrfs_comp_cpu_keys(key, &test);
1540 test.objectid = sk->max_objectid;
1541 test.type = sk->max_type;
1542 test.offset = sk->max_offset;
1544 ret = btrfs_comp_cpu_keys(key, &test);
1550 static noinline int copy_to_sk(struct btrfs_path *path,
1551 struct btrfs_key *key,
1552 struct btrfs_ioctl_search_key *sk,
1555 unsigned long *sk_offset,
1559 struct extent_buffer *leaf;
1560 struct btrfs_ioctl_search_header sh;
1561 struct btrfs_key test;
1562 unsigned long item_off;
1563 unsigned long item_len;
1569 leaf = path->nodes[0];
1570 slot = path->slots[0];
1571 nritems = btrfs_header_nritems(leaf);
1573 if (btrfs_header_generation(leaf) > sk->max_transid) {
1577 found_transid = btrfs_header_generation(leaf);
1579 for (i = slot; i < nritems; i++) {
1580 item_off = btrfs_item_ptr_offset(leaf, i);
1581 item_len = btrfs_item_size(leaf, i);
1583 btrfs_item_key_to_cpu(leaf, key, i);
1584 if (!key_in_sk(key, sk))
1587 if (sizeof(sh) + item_len > *buf_size) {
1594 * return one empty item back for v1, which does not
1598 *buf_size = sizeof(sh) + item_len;
1603 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1608 sh.objectid = key->objectid;
1609 sh.offset = key->offset;
1610 sh.type = key->type;
1612 sh.transid = found_transid;
1615 * Copy search result header. If we fault then loop again so we
1616 * can fault in the pages and -EFAULT there if there's a
1617 * problem. Otherwise we'll fault and then copy the buffer in
1618 * properly this next time through
1620 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
1625 *sk_offset += sizeof(sh);
1628 char __user *up = ubuf + *sk_offset;
1630 * Copy the item, same behavior as above, but reset the
1631 * * sk_offset so we copy the full thing again.
1633 if (read_extent_buffer_to_user_nofault(leaf, up,
1634 item_off, item_len)) {
1636 *sk_offset -= sizeof(sh);
1640 *sk_offset += item_len;
1644 if (ret) /* -EOVERFLOW from above */
1647 if (*num_found >= sk->nr_items) {
1654 test.objectid = sk->max_objectid;
1655 test.type = sk->max_type;
1656 test.offset = sk->max_offset;
1657 if (btrfs_comp_cpu_keys(key, &test) >= 0)
1659 else if (key->offset < (u64)-1)
1661 else if (key->type < (u8)-1) {
1664 } else if (key->objectid < (u64)-1) {
1672 * 0: all items from this leaf copied, continue with next
1673 * 1: * more items can be copied, but unused buffer is too small
1674 * * all items were found
1675 * Either way, it will stops the loop which iterates to the next
1677 * -EOVERFLOW: item was to large for buffer
1678 * -EFAULT: could not copy extent buffer back to userspace
1683 static noinline int search_ioctl(struct inode *inode,
1684 struct btrfs_ioctl_search_key *sk,
1688 struct btrfs_fs_info *info = inode_to_fs_info(inode);
1689 struct btrfs_root *root;
1690 struct btrfs_key key;
1691 struct btrfs_path *path;
1694 unsigned long sk_offset = 0;
1696 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
1697 *buf_size = sizeof(struct btrfs_ioctl_search_header);
1701 path = btrfs_alloc_path();
1705 if (sk->tree_id == 0) {
1706 /* search the root of the inode that was passed */
1707 root = btrfs_grab_root(BTRFS_I(inode)->root);
1709 root = btrfs_get_fs_root(info, sk->tree_id, true);
1711 btrfs_free_path(path);
1712 return PTR_ERR(root);
1716 key.objectid = sk->min_objectid;
1717 key.type = sk->min_type;
1718 key.offset = sk->min_offset;
1723 * Ensure that the whole user buffer is faulted in at sub-page
1724 * granularity, otherwise the loop may live-lock.
1726 if (fault_in_subpage_writeable(ubuf + sk_offset,
1727 *buf_size - sk_offset))
1730 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1736 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
1737 &sk_offset, &num_found);
1738 btrfs_release_path(path);
1746 sk->nr_items = num_found;
1747 btrfs_put_root(root);
1748 btrfs_free_path(path);
1752 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
1755 struct btrfs_ioctl_search_args __user *uargs = argp;
1756 struct btrfs_ioctl_search_key sk;
1760 if (!capable(CAP_SYS_ADMIN))
1763 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
1766 buf_size = sizeof(uargs->buf);
1768 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
1771 * In the origin implementation an overflow is handled by returning a
1772 * search header with a len of zero, so reset ret.
1774 if (ret == -EOVERFLOW)
1777 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
1782 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
1785 struct btrfs_ioctl_search_args_v2 __user *uarg = argp;
1786 struct btrfs_ioctl_search_args_v2 args;
1789 const u64 buf_limit = SZ_16M;
1791 if (!capable(CAP_SYS_ADMIN))
1794 /* copy search header and buffer size */
1795 if (copy_from_user(&args, uarg, sizeof(args)))
1798 buf_size = args.buf_size;
1800 /* limit result size to 16MB */
1801 if (buf_size > buf_limit)
1802 buf_size = buf_limit;
1804 ret = search_ioctl(inode, &args.key, &buf_size,
1805 (char __user *)(&uarg->buf[0]));
1806 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
1808 else if (ret == -EOVERFLOW &&
1809 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
1816 * Search INODE_REFs to identify path name of 'dirid' directory
1817 * in a 'tree_id' tree. and sets path name to 'name'.
1819 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1820 u64 tree_id, u64 dirid, char *name)
1822 struct btrfs_root *root;
1823 struct btrfs_key key;
1829 struct btrfs_inode_ref *iref;
1830 struct extent_buffer *l;
1831 struct btrfs_path *path;
1833 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1838 path = btrfs_alloc_path();
1842 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
1844 root = btrfs_get_fs_root(info, tree_id, true);
1846 ret = PTR_ERR(root);
1851 key.objectid = dirid;
1852 key.type = BTRFS_INODE_REF_KEY;
1853 key.offset = (u64)-1;
1856 ret = btrfs_search_backwards(root, &key, path);
1865 slot = path->slots[0];
1867 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1868 len = btrfs_inode_ref_name_len(l, iref);
1870 total_len += len + 1;
1872 ret = -ENAMETOOLONG;
1877 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
1879 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1882 btrfs_release_path(path);
1883 key.objectid = key.offset;
1884 key.offset = (u64)-1;
1885 dirid = key.objectid;
1887 memmove(name, ptr, total_len);
1888 name[total_len] = '\0';
1891 btrfs_put_root(root);
1892 btrfs_free_path(path);
1896 static int btrfs_search_path_in_tree_user(struct mnt_idmap *idmap,
1897 struct inode *inode,
1898 struct btrfs_ioctl_ino_lookup_user_args *args)
1900 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1901 struct super_block *sb = inode->i_sb;
1902 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
1903 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
1904 u64 dirid = args->dirid;
1905 unsigned long item_off;
1906 unsigned long item_len;
1907 struct btrfs_inode_ref *iref;
1908 struct btrfs_root_ref *rref;
1909 struct btrfs_root *root = NULL;
1910 struct btrfs_path *path;
1911 struct btrfs_key key, key2;
1912 struct extent_buffer *leaf;
1913 struct inode *temp_inode;
1920 path = btrfs_alloc_path();
1925 * If the bottom subvolume does not exist directly under upper_limit,
1926 * construct the path in from the bottom up.
1928 if (dirid != upper_limit.objectid) {
1929 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
1931 root = btrfs_get_fs_root(fs_info, treeid, true);
1933 ret = PTR_ERR(root);
1937 key.objectid = dirid;
1938 key.type = BTRFS_INODE_REF_KEY;
1939 key.offset = (u64)-1;
1941 ret = btrfs_search_backwards(root, &key, path);
1949 leaf = path->nodes[0];
1950 slot = path->slots[0];
1952 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
1953 len = btrfs_inode_ref_name_len(leaf, iref);
1955 total_len += len + 1;
1956 if (ptr < args->path) {
1957 ret = -ENAMETOOLONG;
1962 read_extent_buffer(leaf, ptr,
1963 (unsigned long)(iref + 1), len);
1965 /* Check the read+exec permission of this directory */
1966 ret = btrfs_previous_item(root, path, dirid,
1967 BTRFS_INODE_ITEM_KEY);
1970 } else if (ret > 0) {
1975 leaf = path->nodes[0];
1976 slot = path->slots[0];
1977 btrfs_item_key_to_cpu(leaf, &key2, slot);
1978 if (key2.objectid != dirid) {
1984 * We don't need the path anymore, so release it and
1985 * avoid deadlocks and lockdep warnings in case
1986 * btrfs_iget() needs to lookup the inode from its root
1987 * btree and lock the same leaf.
1989 btrfs_release_path(path);
1990 temp_inode = btrfs_iget(sb, key2.objectid, root);
1991 if (IS_ERR(temp_inode)) {
1992 ret = PTR_ERR(temp_inode);
1995 ret = inode_permission(idmap, temp_inode,
1996 MAY_READ | MAY_EXEC);
2003 if (key.offset == upper_limit.objectid)
2005 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2010 key.objectid = key.offset;
2011 key.offset = (u64)-1;
2012 dirid = key.objectid;
2015 memmove(args->path, ptr, total_len);
2016 args->path[total_len] = '\0';
2017 btrfs_put_root(root);
2019 btrfs_release_path(path);
2022 /* Get the bottom subvolume's name from ROOT_REF */
2023 key.objectid = treeid;
2024 key.type = BTRFS_ROOT_REF_KEY;
2025 key.offset = args->treeid;
2026 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2029 } else if (ret > 0) {
2034 leaf = path->nodes[0];
2035 slot = path->slots[0];
2036 btrfs_item_key_to_cpu(leaf, &key, slot);
2038 item_off = btrfs_item_ptr_offset(leaf, slot);
2039 item_len = btrfs_item_size(leaf, slot);
2040 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2041 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2042 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2047 /* Copy subvolume's name */
2048 item_off += sizeof(struct btrfs_root_ref);
2049 item_len -= sizeof(struct btrfs_root_ref);
2050 read_extent_buffer(leaf, args->name, item_off, item_len);
2051 args->name[item_len] = 0;
2054 btrfs_put_root(root);
2056 btrfs_free_path(path);
2060 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2063 struct btrfs_ioctl_ino_lookup_args *args;
2066 args = memdup_user(argp, sizeof(*args));
2068 return PTR_ERR(args);
2071 * Unprivileged query to obtain the containing subvolume root id. The
2072 * path is reset so it's consistent with btrfs_search_path_in_tree.
2074 if (args->treeid == 0)
2075 args->treeid = root->root_key.objectid;
2077 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2082 if (!capable(CAP_SYS_ADMIN)) {
2087 ret = btrfs_search_path_in_tree(root->fs_info,
2088 args->treeid, args->objectid,
2092 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2100 * Version of ino_lookup ioctl (unprivileged)
2102 * The main differences from ino_lookup ioctl are:
2104 * 1. Read + Exec permission will be checked using inode_permission() during
2105 * path construction. -EACCES will be returned in case of failure.
2106 * 2. Path construction will be stopped at the inode number which corresponds
2107 * to the fd with which this ioctl is called. If constructed path does not
2108 * exist under fd's inode, -EACCES will be returned.
2109 * 3. The name of bottom subvolume is also searched and filled.
2111 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2113 struct btrfs_ioctl_ino_lookup_user_args *args;
2114 struct inode *inode;
2117 args = memdup_user(argp, sizeof(*args));
2119 return PTR_ERR(args);
2121 inode = file_inode(file);
2123 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2124 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2126 * The subvolume does not exist under fd with which this is
2133 ret = btrfs_search_path_in_tree_user(file_mnt_idmap(file), inode, args);
2135 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2142 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2143 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2145 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2146 struct btrfs_fs_info *fs_info;
2147 struct btrfs_root *root;
2148 struct btrfs_path *path;
2149 struct btrfs_key key;
2150 struct btrfs_root_item *root_item;
2151 struct btrfs_root_ref *rref;
2152 struct extent_buffer *leaf;
2153 unsigned long item_off;
2154 unsigned long item_len;
2158 path = btrfs_alloc_path();
2162 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2164 btrfs_free_path(path);
2168 fs_info = BTRFS_I(inode)->root->fs_info;
2170 /* Get root_item of inode's subvolume */
2171 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2172 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2174 ret = PTR_ERR(root);
2177 root_item = &root->root_item;
2179 subvol_info->treeid = key.objectid;
2181 subvol_info->generation = btrfs_root_generation(root_item);
2182 subvol_info->flags = btrfs_root_flags(root_item);
2184 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2185 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2187 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2190 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2191 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2192 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2194 subvol_info->otransid = btrfs_root_otransid(root_item);
2195 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2196 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2198 subvol_info->stransid = btrfs_root_stransid(root_item);
2199 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2200 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2202 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2203 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2204 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2206 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2207 /* Search root tree for ROOT_BACKREF of this subvolume */
2208 key.type = BTRFS_ROOT_BACKREF_KEY;
2210 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2213 } else if (path->slots[0] >=
2214 btrfs_header_nritems(path->nodes[0])) {
2215 ret = btrfs_next_leaf(fs_info->tree_root, path);
2218 } else if (ret > 0) {
2224 leaf = path->nodes[0];
2225 slot = path->slots[0];
2226 btrfs_item_key_to_cpu(leaf, &key, slot);
2227 if (key.objectid == subvol_info->treeid &&
2228 key.type == BTRFS_ROOT_BACKREF_KEY) {
2229 subvol_info->parent_id = key.offset;
2231 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2232 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2234 item_off = btrfs_item_ptr_offset(leaf, slot)
2235 + sizeof(struct btrfs_root_ref);
2236 item_len = btrfs_item_size(leaf, slot)
2237 - sizeof(struct btrfs_root_ref);
2238 read_extent_buffer(leaf, subvol_info->name,
2239 item_off, item_len);
2246 btrfs_free_path(path);
2248 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2252 btrfs_put_root(root);
2254 btrfs_free_path(path);
2260 * Return ROOT_REF information of the subvolume containing this inode
2261 * except the subvolume name.
2263 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
2266 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2267 struct btrfs_root_ref *rref;
2268 struct btrfs_path *path;
2269 struct btrfs_key key;
2270 struct extent_buffer *leaf;
2276 path = btrfs_alloc_path();
2280 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2281 if (IS_ERR(rootrefs)) {
2282 btrfs_free_path(path);
2283 return PTR_ERR(rootrefs);
2286 objectid = root->root_key.objectid;
2287 key.objectid = objectid;
2288 key.type = BTRFS_ROOT_REF_KEY;
2289 key.offset = rootrefs->min_treeid;
2292 root = root->fs_info->tree_root;
2293 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2296 } else if (path->slots[0] >=
2297 btrfs_header_nritems(path->nodes[0])) {
2298 ret = btrfs_next_leaf(root, path);
2301 } else if (ret > 0) {
2307 leaf = path->nodes[0];
2308 slot = path->slots[0];
2310 btrfs_item_key_to_cpu(leaf, &key, slot);
2311 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2316 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2321 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2322 rootrefs->rootref[found].treeid = key.offset;
2323 rootrefs->rootref[found].dirid =
2324 btrfs_root_ref_dirid(leaf, rref);
2327 ret = btrfs_next_item(root, path);
2330 } else if (ret > 0) {
2337 btrfs_free_path(path);
2339 if (!ret || ret == -EOVERFLOW) {
2340 rootrefs->num_items = found;
2341 /* update min_treeid for next search */
2343 rootrefs->min_treeid =
2344 rootrefs->rootref[found - 1].treeid + 1;
2345 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2354 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2358 struct dentry *parent = file->f_path.dentry;
2359 struct dentry *dentry;
2360 struct inode *dir = d_inode(parent);
2361 struct btrfs_fs_info *fs_info = inode_to_fs_info(dir);
2362 struct inode *inode;
2363 struct btrfs_root *root = BTRFS_I(dir)->root;
2364 struct btrfs_root *dest = NULL;
2365 struct btrfs_ioctl_vol_args *vol_args = NULL;
2366 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2367 struct mnt_idmap *idmap = file_mnt_idmap(file);
2368 char *subvol_name, *subvol_name_ptr = NULL;
2371 bool destroy_parent = false;
2373 /* We don't support snapshots with extent tree v2 yet. */
2374 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2376 "extent tree v2 doesn't support snapshot deletion yet");
2381 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2382 if (IS_ERR(vol_args2))
2383 return PTR_ERR(vol_args2);
2385 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2391 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2392 * name, same as v1 currently does.
2394 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2395 err = btrfs_check_ioctl_vol_args2_subvol_name(vol_args2);
2398 subvol_name = vol_args2->name;
2400 err = mnt_want_write_file(file);
2404 struct inode *old_dir;
2406 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2411 err = mnt_want_write_file(file);
2415 dentry = btrfs_get_dentry(fs_info->sb,
2416 BTRFS_FIRST_FREE_OBJECTID,
2417 vol_args2->subvolid, 0);
2418 if (IS_ERR(dentry)) {
2419 err = PTR_ERR(dentry);
2420 goto out_drop_write;
2424 * Change the default parent since the subvolume being
2425 * deleted can be outside of the current mount point.
2427 parent = btrfs_get_parent(dentry);
2430 * At this point dentry->d_name can point to '/' if the
2431 * subvolume we want to destroy is outsite of the
2432 * current mount point, so we need to release the
2433 * current dentry and execute the lookup to return a new
2434 * one with ->d_name pointing to the
2435 * <mount point>/subvol_name.
2438 if (IS_ERR(parent)) {
2439 err = PTR_ERR(parent);
2440 goto out_drop_write;
2443 dir = d_inode(parent);
2446 * If v2 was used with SPEC_BY_ID, a new parent was
2447 * allocated since the subvolume can be outside of the
2448 * current mount point. Later on we need to release this
2449 * new parent dentry.
2451 destroy_parent = true;
2454 * On idmapped mounts, deletion via subvolid is
2455 * restricted to subvolumes that are immediate
2456 * ancestors of the inode referenced by the file
2457 * descriptor in the ioctl. Otherwise the idmapping
2458 * could potentially be abused to delete subvolumes
2459 * anywhere in the filesystem the user wouldn't be able
2460 * to delete without an idmapped mount.
2462 if (old_dir != dir && idmap != &nop_mnt_idmap) {
2467 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2468 fs_info, vol_args2->subvolid);
2469 if (IS_ERR(subvol_name_ptr)) {
2470 err = PTR_ERR(subvol_name_ptr);
2473 /* subvol_name_ptr is already nul terminated */
2474 subvol_name = (char *)kbasename(subvol_name_ptr);
2477 vol_args = memdup_user(arg, sizeof(*vol_args));
2478 if (IS_ERR(vol_args))
2479 return PTR_ERR(vol_args);
2481 err = btrfs_check_ioctl_vol_args_path(vol_args);
2485 subvol_name = vol_args->name;
2487 err = mnt_want_write_file(file);
2492 subvol_namelen = strlen(subvol_name);
2494 if (strchr(subvol_name, '/') ||
2495 strncmp(subvol_name, "..", subvol_namelen) == 0) {
2497 goto free_subvol_name;
2500 if (!S_ISDIR(dir->i_mode)) {
2502 goto free_subvol_name;
2505 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2507 goto free_subvol_name;
2508 dentry = lookup_one(idmap, subvol_name, parent, subvol_namelen);
2509 if (IS_ERR(dentry)) {
2510 err = PTR_ERR(dentry);
2511 goto out_unlock_dir;
2514 if (d_really_is_negative(dentry)) {
2519 inode = d_inode(dentry);
2520 dest = BTRFS_I(inode)->root;
2521 if (!capable(CAP_SYS_ADMIN)) {
2523 * Regular user. Only allow this with a special mount
2524 * option, when the user has write+exec access to the
2525 * subvol root, and when rmdir(2) would have been
2528 * Note that this is _not_ check that the subvol is
2529 * empty or doesn't contain data that we wouldn't
2530 * otherwise be able to delete.
2532 * Users who want to delete empty subvols should try
2536 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2540 * Do not allow deletion if the parent dir is the same
2541 * as the dir to be deleted. That means the ioctl
2542 * must be called on the dentry referencing the root
2543 * of the subvol, not a random directory contained
2550 err = inode_permission(idmap, inode, MAY_WRITE | MAY_EXEC);
2555 /* check if subvolume may be deleted by a user */
2556 err = btrfs_may_delete(idmap, dir, dentry, 1);
2560 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2565 btrfs_inode_lock(BTRFS_I(inode), 0);
2566 err = btrfs_delete_subvolume(BTRFS_I(dir), dentry);
2567 btrfs_inode_unlock(BTRFS_I(inode), 0);
2569 d_delete_notify(dir, dentry);
2574 btrfs_inode_unlock(BTRFS_I(dir), 0);
2576 kfree(subvol_name_ptr);
2581 mnt_drop_write_file(file);
2588 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2590 struct inode *inode = file_inode(file);
2591 struct btrfs_root *root = BTRFS_I(inode)->root;
2592 struct btrfs_ioctl_defrag_range_args range = {0};
2595 ret = mnt_want_write_file(file);
2599 if (btrfs_root_readonly(root)) {
2604 switch (inode->i_mode & S_IFMT) {
2606 if (!capable(CAP_SYS_ADMIN)) {
2610 ret = btrfs_defrag_root(root);
2614 * Note that this does not check the file descriptor for write
2615 * access. This prevents defragmenting executables that are
2616 * running and allows defrag on files open in read-only mode.
2618 if (!capable(CAP_SYS_ADMIN) &&
2619 inode_permission(&nop_mnt_idmap, inode, MAY_WRITE)) {
2625 if (copy_from_user(&range, argp, sizeof(range))) {
2629 if (range.flags & ~BTRFS_DEFRAG_RANGE_FLAGS_SUPP) {
2633 /* compression requires us to start the IO */
2634 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2635 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
2636 range.extent_thresh = (u32)-1;
2639 /* the rest are all set to zero by kzalloc */
2640 range.len = (u64)-1;
2642 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
2643 &range, BTRFS_OLDEST_GENERATION, 0);
2651 mnt_drop_write_file(file);
2655 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2657 struct btrfs_ioctl_vol_args *vol_args;
2658 bool restore_op = false;
2661 if (!capable(CAP_SYS_ADMIN))
2664 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
2665 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
2669 if (fs_info->fs_devices->temp_fsid) {
2671 "device add not supported on cloned temp-fsid mount");
2675 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
2676 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
2677 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2680 * We can do the device add because we have a paused balanced,
2681 * change the exclusive op type and remember we should bring
2682 * back the paused balance
2684 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
2685 btrfs_exclop_start_unlock(fs_info);
2689 vol_args = memdup_user(arg, sizeof(*vol_args));
2690 if (IS_ERR(vol_args)) {
2691 ret = PTR_ERR(vol_args);
2695 ret = btrfs_check_ioctl_vol_args_path(vol_args);
2699 ret = btrfs_init_new_device(fs_info, vol_args->name);
2702 btrfs_info(fs_info, "disk added %s", vol_args->name);
2708 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
2710 btrfs_exclop_finish(fs_info);
2714 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2716 BTRFS_DEV_LOOKUP_ARGS(args);
2717 struct inode *inode = file_inode(file);
2718 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2719 struct btrfs_ioctl_vol_args_v2 *vol_args;
2720 struct file *bdev_file = NULL;
2722 bool cancel = false;
2724 if (!capable(CAP_SYS_ADMIN))
2727 vol_args = memdup_user(arg, sizeof(*vol_args));
2728 if (IS_ERR(vol_args))
2729 return PTR_ERR(vol_args);
2731 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
2736 ret = btrfs_check_ioctl_vol_args2_subvol_name(vol_args);
2740 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2741 args.devid = vol_args->devid;
2742 } else if (!strcmp("cancel", vol_args->name)) {
2745 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2750 ret = mnt_want_write_file(file);
2754 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2759 /* Exclusive operation is now claimed */
2760 ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2762 btrfs_exclop_finish(fs_info);
2765 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2766 btrfs_info(fs_info, "device deleted: id %llu",
2769 btrfs_info(fs_info, "device deleted: %s",
2773 mnt_drop_write_file(file);
2777 btrfs_put_dev_args_from_path(&args);
2782 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2784 BTRFS_DEV_LOOKUP_ARGS(args);
2785 struct inode *inode = file_inode(file);
2786 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2787 struct btrfs_ioctl_vol_args *vol_args;
2788 struct file *bdev_file = NULL;
2790 bool cancel = false;
2792 if (!capable(CAP_SYS_ADMIN))
2795 vol_args = memdup_user(arg, sizeof(*vol_args));
2796 if (IS_ERR(vol_args))
2797 return PTR_ERR(vol_args);
2799 ret = btrfs_check_ioctl_vol_args_path(vol_args);
2803 if (!strcmp("cancel", vol_args->name)) {
2806 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
2811 ret = mnt_want_write_file(file);
2815 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
2818 ret = btrfs_rm_device(fs_info, &args, &bdev_file);
2820 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2821 btrfs_exclop_finish(fs_info);
2824 mnt_drop_write_file(file);
2828 btrfs_put_dev_args_from_path(&args);
2834 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2837 struct btrfs_ioctl_fs_info_args *fi_args;
2838 struct btrfs_device *device;
2839 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2843 fi_args = memdup_user(arg, sizeof(*fi_args));
2844 if (IS_ERR(fi_args))
2845 return PTR_ERR(fi_args);
2847 flags_in = fi_args->flags;
2848 memset(fi_args, 0, sizeof(*fi_args));
2851 fi_args->num_devices = fs_devices->num_devices;
2853 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
2854 if (device->devid > fi_args->max_id)
2855 fi_args->max_id = device->devid;
2859 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
2860 fi_args->nodesize = fs_info->nodesize;
2861 fi_args->sectorsize = fs_info->sectorsize;
2862 fi_args->clone_alignment = fs_info->sectorsize;
2864 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
2865 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
2866 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
2867 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
2870 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
2871 fi_args->generation = btrfs_get_fs_generation(fs_info);
2872 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
2875 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
2876 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
2877 sizeof(fi_args->metadata_uuid));
2878 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
2881 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2888 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2891 BTRFS_DEV_LOOKUP_ARGS(args);
2892 struct btrfs_ioctl_dev_info_args *di_args;
2893 struct btrfs_device *dev;
2896 di_args = memdup_user(arg, sizeof(*di_args));
2897 if (IS_ERR(di_args))
2898 return PTR_ERR(di_args);
2900 args.devid = di_args->devid;
2901 if (!btrfs_is_empty_uuid(di_args->uuid))
2902 args.uuid = di_args->uuid;
2905 dev = btrfs_find_device(fs_info->fs_devices, &args);
2911 di_args->devid = dev->devid;
2912 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2913 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2914 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2915 memcpy(di_args->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
2917 strscpy(di_args->path, btrfs_dev_name(dev), sizeof(di_args->path));
2919 di_args->path[0] = '\0';
2923 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2930 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2932 struct inode *inode = file_inode(file);
2933 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
2934 struct btrfs_root *root = BTRFS_I(inode)->root;
2935 struct btrfs_root *new_root;
2936 struct btrfs_dir_item *di;
2937 struct btrfs_trans_handle *trans;
2938 struct btrfs_path *path = NULL;
2939 struct btrfs_disk_key disk_key;
2940 struct fscrypt_str name = FSTR_INIT("default", 7);
2945 if (!capable(CAP_SYS_ADMIN))
2948 ret = mnt_want_write_file(file);
2952 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2958 objectid = BTRFS_FS_TREE_OBJECTID;
2960 new_root = btrfs_get_fs_root(fs_info, objectid, true);
2961 if (IS_ERR(new_root)) {
2962 ret = PTR_ERR(new_root);
2965 if (!is_fstree(new_root->root_key.objectid)) {
2970 path = btrfs_alloc_path();
2976 trans = btrfs_start_transaction(root, 1);
2977 if (IS_ERR(trans)) {
2978 ret = PTR_ERR(trans);
2982 dir_id = btrfs_super_root_dir(fs_info->super_copy);
2983 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
2985 if (IS_ERR_OR_NULL(di)) {
2986 btrfs_release_path(path);
2987 btrfs_end_transaction(trans);
2989 "Umm, you don't have the default diritem, this isn't going to work");
2994 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2995 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2996 btrfs_mark_buffer_dirty(trans, path->nodes[0]);
2997 btrfs_release_path(path);
2999 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3000 btrfs_end_transaction(trans);
3002 btrfs_put_root(new_root);
3003 btrfs_free_path(path);
3005 mnt_drop_write_file(file);
3009 static void get_block_group_info(struct list_head *groups_list,
3010 struct btrfs_ioctl_space_info *space)
3012 struct btrfs_block_group *block_group;
3014 space->total_bytes = 0;
3015 space->used_bytes = 0;
3017 list_for_each_entry(block_group, groups_list, list) {
3018 space->flags = block_group->flags;
3019 space->total_bytes += block_group->length;
3020 space->used_bytes += block_group->used;
3024 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3027 struct btrfs_ioctl_space_args space_args = { 0 };
3028 struct btrfs_ioctl_space_info space;
3029 struct btrfs_ioctl_space_info *dest;
3030 struct btrfs_ioctl_space_info *dest_orig;
3031 struct btrfs_ioctl_space_info __user *user_dest;
3032 struct btrfs_space_info *info;
3033 static const u64 types[] = {
3034 BTRFS_BLOCK_GROUP_DATA,
3035 BTRFS_BLOCK_GROUP_SYSTEM,
3036 BTRFS_BLOCK_GROUP_METADATA,
3037 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3045 if (copy_from_user(&space_args,
3046 (struct btrfs_ioctl_space_args __user *)arg,
3047 sizeof(space_args)))
3050 for (i = 0; i < num_types; i++) {
3051 struct btrfs_space_info *tmp;
3054 list_for_each_entry(tmp, &fs_info->space_info, list) {
3055 if (tmp->flags == types[i]) {
3064 down_read(&info->groups_sem);
3065 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3066 if (!list_empty(&info->block_groups[c]))
3069 up_read(&info->groups_sem);
3073 * Global block reserve, exported as a space_info
3077 /* space_slots == 0 means they are asking for a count */
3078 if (space_args.space_slots == 0) {
3079 space_args.total_spaces = slot_count;
3083 slot_count = min_t(u64, space_args.space_slots, slot_count);
3085 alloc_size = sizeof(*dest) * slot_count;
3087 /* we generally have at most 6 or so space infos, one for each raid
3088 * level. So, a whole page should be more than enough for everyone
3090 if (alloc_size > PAGE_SIZE)
3093 space_args.total_spaces = 0;
3094 dest = kmalloc(alloc_size, GFP_KERNEL);
3099 /* now we have a buffer to copy into */
3100 for (i = 0; i < num_types; i++) {
3101 struct btrfs_space_info *tmp;
3107 list_for_each_entry(tmp, &fs_info->space_info, list) {
3108 if (tmp->flags == types[i]) {
3116 down_read(&info->groups_sem);
3117 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3118 if (!list_empty(&info->block_groups[c])) {
3119 get_block_group_info(&info->block_groups[c],
3121 memcpy(dest, &space, sizeof(space));
3123 space_args.total_spaces++;
3129 up_read(&info->groups_sem);
3133 * Add global block reserve
3136 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3138 spin_lock(&block_rsv->lock);
3139 space.total_bytes = block_rsv->size;
3140 space.used_bytes = block_rsv->size - block_rsv->reserved;
3141 spin_unlock(&block_rsv->lock);
3142 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3143 memcpy(dest, &space, sizeof(space));
3144 space_args.total_spaces++;
3147 user_dest = (struct btrfs_ioctl_space_info __user *)
3148 (arg + sizeof(struct btrfs_ioctl_space_args));
3150 if (copy_to_user(user_dest, dest_orig, alloc_size))
3155 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3161 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3164 struct btrfs_trans_handle *trans;
3168 * Start orphan cleanup here for the given root in case it hasn't been
3169 * started already by other means. Errors are handled in the other
3170 * functions during transaction commit.
3172 btrfs_orphan_cleanup(root);
3174 trans = btrfs_attach_transaction_barrier(root);
3175 if (IS_ERR(trans)) {
3176 if (PTR_ERR(trans) != -ENOENT)
3177 return PTR_ERR(trans);
3179 /* No running transaction, don't bother */
3180 transid = btrfs_get_last_trans_committed(root->fs_info);
3183 transid = trans->transid;
3184 btrfs_commit_transaction_async(trans);
3187 if (copy_to_user(argp, &transid, sizeof(transid)))
3192 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3195 /* By default wait for the current transaction. */
3199 if (copy_from_user(&transid, argp, sizeof(transid)))
3202 return btrfs_wait_for_commit(fs_info, transid);
3205 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3207 struct btrfs_fs_info *fs_info = inode_to_fs_info(file_inode(file));
3208 struct btrfs_ioctl_scrub_args *sa;
3211 if (!capable(CAP_SYS_ADMIN))
3214 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3215 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
3219 sa = memdup_user(arg, sizeof(*sa));
3223 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
3228 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3229 ret = mnt_want_write_file(file);
3234 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3235 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3239 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3240 * error. This is important as it allows user space to know how much
3241 * progress scrub has done. For example, if scrub is canceled we get
3242 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3243 * space. Later user space can inspect the progress from the structure
3244 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3245 * previously (btrfs-progs does this).
3246 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3247 * then return -EFAULT to signal the structure was not copied or it may
3248 * be corrupt and unreliable due to a partial copy.
3250 if (copy_to_user(arg, sa, sizeof(*sa)))
3253 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3254 mnt_drop_write_file(file);
3260 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3262 if (!capable(CAP_SYS_ADMIN))
3265 return btrfs_scrub_cancel(fs_info);
3268 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3271 struct btrfs_ioctl_scrub_args *sa;
3274 if (!capable(CAP_SYS_ADMIN))
3277 sa = memdup_user(arg, sizeof(*sa));
3281 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3283 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3290 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3293 struct btrfs_ioctl_get_dev_stats *sa;
3296 sa = memdup_user(arg, sizeof(*sa));
3300 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3305 ret = btrfs_get_dev_stats(fs_info, sa);
3307 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3314 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3317 struct btrfs_ioctl_dev_replace_args *p;
3320 if (!capable(CAP_SYS_ADMIN))
3323 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3324 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
3328 p = memdup_user(arg, sizeof(*p));
3333 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3334 if (sb_rdonly(fs_info->sb)) {
3338 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3339 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3341 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3342 btrfs_exclop_finish(fs_info);
3345 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3346 btrfs_dev_replace_status(fs_info, p);
3349 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3350 p->result = btrfs_dev_replace_cancel(fs_info);
3358 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3365 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3371 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3372 struct inode_fs_paths *ipath = NULL;
3373 struct btrfs_path *path;
3375 if (!capable(CAP_DAC_READ_SEARCH))
3378 path = btrfs_alloc_path();
3384 ipa = memdup_user(arg, sizeof(*ipa));
3391 size = min_t(u32, ipa->size, 4096);
3392 ipath = init_ipath(size, root, path);
3393 if (IS_ERR(ipath)) {
3394 ret = PTR_ERR(ipath);
3399 ret = paths_from_inode(ipa->inum, ipath);
3403 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3404 rel_ptr = ipath->fspath->val[i] -
3405 (u64)(unsigned long)ipath->fspath->val;
3406 ipath->fspath->val[i] = rel_ptr;
3409 btrfs_free_path(path);
3411 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3412 ipath->fspath, size);
3419 btrfs_free_path(path);
3426 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3427 void __user *arg, int version)
3431 struct btrfs_ioctl_logical_ino_args *loi;
3432 struct btrfs_data_container *inodes = NULL;
3433 struct btrfs_path *path = NULL;
3436 if (!capable(CAP_SYS_ADMIN))
3439 loi = memdup_user(arg, sizeof(*loi));
3441 return PTR_ERR(loi);
3444 ignore_offset = false;
3445 size = min_t(u32, loi->size, SZ_64K);
3447 /* All reserved bits must be 0 for now */
3448 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3452 /* Only accept flags we have defined so far */
3453 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3457 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3458 size = min_t(u32, loi->size, SZ_16M);
3461 inodes = init_data_container(size);
3462 if (IS_ERR(inodes)) {
3463 ret = PTR_ERR(inodes);
3467 path = btrfs_alloc_path();
3472 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3473 inodes, ignore_offset);
3474 btrfs_free_path(path);
3480 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3493 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3494 struct btrfs_ioctl_balance_args *bargs)
3496 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3498 bargs->flags = bctl->flags;
3500 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3501 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3502 if (atomic_read(&fs_info->balance_pause_req))
3503 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3504 if (atomic_read(&fs_info->balance_cancel_req))
3505 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3507 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3508 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3509 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3511 spin_lock(&fs_info->balance_lock);
3512 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3513 spin_unlock(&fs_info->balance_lock);
3517 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3520 * @fs_info: the filesystem
3521 * @excl_acquired: ptr to boolean value which is set to false in case balance
3524 * Return 0 on success in which case both fs_info::balance is acquired as well
3525 * as exclusive ops are blocked. In case of failure return an error code.
3527 static int btrfs_try_lock_balance(struct btrfs_fs_info *fs_info, bool *excl_acquired)
3532 * Exclusive operation is locked. Three possibilities:
3533 * (1) some other op is running
3534 * (2) balance is running
3535 * (3) balance is paused -- special case (think resume)
3538 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
3539 *excl_acquired = true;
3540 mutex_lock(&fs_info->balance_mutex);
3544 mutex_lock(&fs_info->balance_mutex);
3545 if (fs_info->balance_ctl) {
3546 /* This is either (2) or (3) */
3547 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3553 mutex_unlock(&fs_info->balance_mutex);
3555 * Lock released to allow other waiters to
3556 * continue, we'll reexamine the status again.
3558 mutex_lock(&fs_info->balance_mutex);
3560 if (fs_info->balance_ctl &&
3561 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
3563 *excl_acquired = false;
3569 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3573 mutex_unlock(&fs_info->balance_mutex);
3577 mutex_unlock(&fs_info->balance_mutex);
3578 *excl_acquired = false;
3582 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3584 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3585 struct btrfs_fs_info *fs_info = root->fs_info;
3586 struct btrfs_ioctl_balance_args *bargs;
3587 struct btrfs_balance_control *bctl;
3588 bool need_unlock = true;
3591 if (!capable(CAP_SYS_ADMIN))
3594 ret = mnt_want_write_file(file);
3598 bargs = memdup_user(arg, sizeof(*bargs));
3599 if (IS_ERR(bargs)) {
3600 ret = PTR_ERR(bargs);
3605 ret = btrfs_try_lock_balance(fs_info, &need_unlock);
3609 lockdep_assert_held(&fs_info->balance_mutex);
3611 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3612 if (!fs_info->balance_ctl) {
3617 bctl = fs_info->balance_ctl;
3618 spin_lock(&fs_info->balance_lock);
3619 bctl->flags |= BTRFS_BALANCE_RESUME;
3620 spin_unlock(&fs_info->balance_lock);
3621 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
3626 if (bargs->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
3631 if (fs_info->balance_ctl) {
3636 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
3642 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3643 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3644 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3646 bctl->flags = bargs->flags;
3649 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3650 * bctl is freed in reset_balance_state, or, if restriper was paused
3651 * all the way until unmount, in free_fs_info. The flag should be
3652 * cleared after reset_balance_state.
3654 need_unlock = false;
3656 ret = btrfs_balance(fs_info, bctl, bargs);
3659 if (ret == 0 || ret == -ECANCELED) {
3660 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3666 mutex_unlock(&fs_info->balance_mutex);
3668 btrfs_exclop_finish(fs_info);
3670 mnt_drop_write_file(file);
3675 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
3677 if (!capable(CAP_SYS_ADMIN))
3681 case BTRFS_BALANCE_CTL_PAUSE:
3682 return btrfs_pause_balance(fs_info);
3683 case BTRFS_BALANCE_CTL_CANCEL:
3684 return btrfs_cancel_balance(fs_info);
3690 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
3693 struct btrfs_ioctl_balance_args *bargs;
3696 if (!capable(CAP_SYS_ADMIN))
3699 mutex_lock(&fs_info->balance_mutex);
3700 if (!fs_info->balance_ctl) {
3705 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
3711 btrfs_update_ioctl_balance_args(fs_info, bargs);
3713 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3718 mutex_unlock(&fs_info->balance_mutex);
3722 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
3724 struct inode *inode = file_inode(file);
3725 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3726 struct btrfs_ioctl_quota_ctl_args *sa;
3729 if (!capable(CAP_SYS_ADMIN))
3732 ret = mnt_want_write_file(file);
3736 sa = memdup_user(arg, sizeof(*sa));
3742 down_write(&fs_info->subvol_sem);
3745 case BTRFS_QUOTA_CTL_ENABLE:
3746 case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA:
3747 ret = btrfs_quota_enable(fs_info, sa);
3749 case BTRFS_QUOTA_CTL_DISABLE:
3750 ret = btrfs_quota_disable(fs_info);
3758 up_write(&fs_info->subvol_sem);
3760 mnt_drop_write_file(file);
3764 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
3766 struct inode *inode = file_inode(file);
3767 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3768 struct btrfs_root *root = BTRFS_I(inode)->root;
3769 struct btrfs_ioctl_qgroup_assign_args *sa;
3770 struct btrfs_trans_handle *trans;
3774 if (!capable(CAP_SYS_ADMIN))
3777 ret = mnt_want_write_file(file);
3781 sa = memdup_user(arg, sizeof(*sa));
3787 trans = btrfs_join_transaction(root);
3788 if (IS_ERR(trans)) {
3789 ret = PTR_ERR(trans);
3794 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
3796 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
3799 /* update qgroup status and info */
3800 mutex_lock(&fs_info->qgroup_ioctl_lock);
3801 err = btrfs_run_qgroups(trans);
3802 mutex_unlock(&fs_info->qgroup_ioctl_lock);
3804 btrfs_handle_fs_error(fs_info, err,
3805 "failed to update qgroup status and info");
3806 err = btrfs_end_transaction(trans);
3813 mnt_drop_write_file(file);
3817 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3819 struct inode *inode = file_inode(file);
3820 struct btrfs_root *root = BTRFS_I(inode)->root;
3821 struct btrfs_ioctl_qgroup_create_args *sa;
3822 struct btrfs_trans_handle *trans;
3826 if (!capable(CAP_SYS_ADMIN))
3829 ret = mnt_want_write_file(file);
3833 sa = memdup_user(arg, sizeof(*sa));
3839 if (!sa->qgroupid) {
3844 if (sa->create && is_fstree(sa->qgroupid)) {
3849 trans = btrfs_join_transaction(root);
3850 if (IS_ERR(trans)) {
3851 ret = PTR_ERR(trans);
3856 ret = btrfs_create_qgroup(trans, sa->qgroupid);
3858 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
3861 err = btrfs_end_transaction(trans);
3868 mnt_drop_write_file(file);
3872 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
3874 struct inode *inode = file_inode(file);
3875 struct btrfs_root *root = BTRFS_I(inode)->root;
3876 struct btrfs_ioctl_qgroup_limit_args *sa;
3877 struct btrfs_trans_handle *trans;
3882 if (!capable(CAP_SYS_ADMIN))
3885 ret = mnt_want_write_file(file);
3889 sa = memdup_user(arg, sizeof(*sa));
3895 trans = btrfs_join_transaction(root);
3896 if (IS_ERR(trans)) {
3897 ret = PTR_ERR(trans);
3901 qgroupid = sa->qgroupid;
3903 /* take the current subvol as qgroup */
3904 qgroupid = root->root_key.objectid;
3907 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
3909 err = btrfs_end_transaction(trans);
3916 mnt_drop_write_file(file);
3920 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
3922 struct inode *inode = file_inode(file);
3923 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3924 struct btrfs_ioctl_quota_rescan_args *qsa;
3927 if (!capable(CAP_SYS_ADMIN))
3930 ret = mnt_want_write_file(file);
3934 qsa = memdup_user(arg, sizeof(*qsa));
3945 ret = btrfs_qgroup_rescan(fs_info);
3950 mnt_drop_write_file(file);
3954 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
3957 struct btrfs_ioctl_quota_rescan_args qsa = {0};
3959 if (!capable(CAP_SYS_ADMIN))
3962 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
3964 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
3967 if (copy_to_user(arg, &qsa, sizeof(qsa)))
3973 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
3976 if (!capable(CAP_SYS_ADMIN))
3979 return btrfs_qgroup_wait_for_completion(fs_info, true);
3982 static long _btrfs_ioctl_set_received_subvol(struct file *file,
3983 struct mnt_idmap *idmap,
3984 struct btrfs_ioctl_received_subvol_args *sa)
3986 struct inode *inode = file_inode(file);
3987 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
3988 struct btrfs_root *root = BTRFS_I(inode)->root;
3989 struct btrfs_root_item *root_item = &root->root_item;
3990 struct btrfs_trans_handle *trans;
3991 struct timespec64 ct = current_time(inode);
3993 int received_uuid_changed;
3995 if (!inode_owner_or_capable(idmap, inode))
3998 ret = mnt_want_write_file(file);
4002 down_write(&fs_info->subvol_sem);
4004 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4009 if (btrfs_root_readonly(root)) {
4016 * 2 - uuid items (received uuid + subvol uuid)
4018 trans = btrfs_start_transaction(root, 3);
4019 if (IS_ERR(trans)) {
4020 ret = PTR_ERR(trans);
4025 sa->rtransid = trans->transid;
4026 sa->rtime.sec = ct.tv_sec;
4027 sa->rtime.nsec = ct.tv_nsec;
4029 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4031 if (received_uuid_changed &&
4032 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4033 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4034 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4035 root->root_key.objectid);
4036 if (ret && ret != -ENOENT) {
4037 btrfs_abort_transaction(trans, ret);
4038 btrfs_end_transaction(trans);
4042 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4043 btrfs_set_root_stransid(root_item, sa->stransid);
4044 btrfs_set_root_rtransid(root_item, sa->rtransid);
4045 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4046 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4047 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4048 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4050 ret = btrfs_update_root(trans, fs_info->tree_root,
4051 &root->root_key, &root->root_item);
4053 btrfs_end_transaction(trans);
4056 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4057 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4058 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4059 root->root_key.objectid);
4060 if (ret < 0 && ret != -EEXIST) {
4061 btrfs_abort_transaction(trans, ret);
4062 btrfs_end_transaction(trans);
4066 ret = btrfs_commit_transaction(trans);
4068 up_write(&fs_info->subvol_sem);
4069 mnt_drop_write_file(file);
4074 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4077 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4078 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4081 args32 = memdup_user(arg, sizeof(*args32));
4083 return PTR_ERR(args32);
4085 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4091 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4092 args64->stransid = args32->stransid;
4093 args64->rtransid = args32->rtransid;
4094 args64->stime.sec = args32->stime.sec;
4095 args64->stime.nsec = args32->stime.nsec;
4096 args64->rtime.sec = args32->rtime.sec;
4097 args64->rtime.nsec = args32->rtime.nsec;
4098 args64->flags = args32->flags;
4100 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), args64);
4104 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4105 args32->stransid = args64->stransid;
4106 args32->rtransid = args64->rtransid;
4107 args32->stime.sec = args64->stime.sec;
4108 args32->stime.nsec = args64->stime.nsec;
4109 args32->rtime.sec = args64->rtime.sec;
4110 args32->rtime.nsec = args64->rtime.nsec;
4111 args32->flags = args64->flags;
4113 ret = copy_to_user(arg, args32, sizeof(*args32));
4124 static long btrfs_ioctl_set_received_subvol(struct file *file,
4127 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4130 sa = memdup_user(arg, sizeof(*sa));
4134 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_idmap(file), sa);
4139 ret = copy_to_user(arg, sa, sizeof(*sa));
4148 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4153 char label[BTRFS_LABEL_SIZE];
4155 spin_lock(&fs_info->super_lock);
4156 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4157 spin_unlock(&fs_info->super_lock);
4159 len = strnlen(label, BTRFS_LABEL_SIZE);
4161 if (len == BTRFS_LABEL_SIZE) {
4163 "label is too long, return the first %zu bytes",
4167 ret = copy_to_user(arg, label, len);
4169 return ret ? -EFAULT : 0;
4172 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4174 struct inode *inode = file_inode(file);
4175 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4176 struct btrfs_root *root = BTRFS_I(inode)->root;
4177 struct btrfs_super_block *super_block = fs_info->super_copy;
4178 struct btrfs_trans_handle *trans;
4179 char label[BTRFS_LABEL_SIZE];
4182 if (!capable(CAP_SYS_ADMIN))
4185 if (copy_from_user(label, arg, sizeof(label)))
4188 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4190 "unable to set label with more than %d bytes",
4191 BTRFS_LABEL_SIZE - 1);
4195 ret = mnt_want_write_file(file);
4199 trans = btrfs_start_transaction(root, 0);
4200 if (IS_ERR(trans)) {
4201 ret = PTR_ERR(trans);
4205 spin_lock(&fs_info->super_lock);
4206 strcpy(super_block->label, label);
4207 spin_unlock(&fs_info->super_lock);
4208 ret = btrfs_commit_transaction(trans);
4211 mnt_drop_write_file(file);
4215 #define INIT_FEATURE_FLAGS(suffix) \
4216 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4217 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4218 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4220 int btrfs_ioctl_get_supported_features(void __user *arg)
4222 static const struct btrfs_ioctl_feature_flags features[3] = {
4223 INIT_FEATURE_FLAGS(SUPP),
4224 INIT_FEATURE_FLAGS(SAFE_SET),
4225 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4228 if (copy_to_user(arg, &features, sizeof(features)))
4234 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4237 struct btrfs_super_block *super_block = fs_info->super_copy;
4238 struct btrfs_ioctl_feature_flags features;
4240 features.compat_flags = btrfs_super_compat_flags(super_block);
4241 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4242 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4244 if (copy_to_user(arg, &features, sizeof(features)))
4250 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4251 enum btrfs_feature_set set,
4252 u64 change_mask, u64 flags, u64 supported_flags,
4253 u64 safe_set, u64 safe_clear)
4255 const char *type = btrfs_feature_set_name(set);
4257 u64 disallowed, unsupported;
4258 u64 set_mask = flags & change_mask;
4259 u64 clear_mask = ~flags & change_mask;
4261 unsupported = set_mask & ~supported_flags;
4263 names = btrfs_printable_features(set, unsupported);
4266 "this kernel does not support the %s feature bit%s",
4267 names, strchr(names, ',') ? "s" : "");
4271 "this kernel does not support %s bits 0x%llx",
4276 disallowed = set_mask & ~safe_set;
4278 names = btrfs_printable_features(set, disallowed);
4281 "can't set the %s feature bit%s while mounted",
4282 names, strchr(names, ',') ? "s" : "");
4286 "can't set %s bits 0x%llx while mounted",
4291 disallowed = clear_mask & ~safe_clear;
4293 names = btrfs_printable_features(set, disallowed);
4296 "can't clear the %s feature bit%s while mounted",
4297 names, strchr(names, ',') ? "s" : "");
4301 "can't clear %s bits 0x%llx while mounted",
4309 #define check_feature(fs_info, change_mask, flags, mask_base) \
4310 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4311 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4312 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4313 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4315 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4317 struct inode *inode = file_inode(file);
4318 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4319 struct btrfs_root *root = BTRFS_I(inode)->root;
4320 struct btrfs_super_block *super_block = fs_info->super_copy;
4321 struct btrfs_ioctl_feature_flags flags[2];
4322 struct btrfs_trans_handle *trans;
4326 if (!capable(CAP_SYS_ADMIN))
4329 if (copy_from_user(flags, arg, sizeof(flags)))
4333 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4334 !flags[0].incompat_flags)
4337 ret = check_feature(fs_info, flags[0].compat_flags,
4338 flags[1].compat_flags, COMPAT);
4342 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4343 flags[1].compat_ro_flags, COMPAT_RO);
4347 ret = check_feature(fs_info, flags[0].incompat_flags,
4348 flags[1].incompat_flags, INCOMPAT);
4352 ret = mnt_want_write_file(file);
4356 trans = btrfs_start_transaction(root, 0);
4357 if (IS_ERR(trans)) {
4358 ret = PTR_ERR(trans);
4359 goto out_drop_write;
4362 spin_lock(&fs_info->super_lock);
4363 newflags = btrfs_super_compat_flags(super_block);
4364 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4365 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4366 btrfs_set_super_compat_flags(super_block, newflags);
4368 newflags = btrfs_super_compat_ro_flags(super_block);
4369 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4370 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4371 btrfs_set_super_compat_ro_flags(super_block, newflags);
4373 newflags = btrfs_super_incompat_flags(super_block);
4374 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4375 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4376 btrfs_set_super_incompat_flags(super_block, newflags);
4377 spin_unlock(&fs_info->super_lock);
4379 ret = btrfs_commit_transaction(trans);
4381 mnt_drop_write_file(file);
4386 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
4388 struct btrfs_ioctl_send_args *arg;
4392 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4393 struct btrfs_ioctl_send_args_32 args32 = { 0 };
4395 ret = copy_from_user(&args32, argp, sizeof(args32));
4398 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4401 arg->send_fd = args32.send_fd;
4402 arg->clone_sources_count = args32.clone_sources_count;
4403 arg->clone_sources = compat_ptr(args32.clone_sources);
4404 arg->parent_root = args32.parent_root;
4405 arg->flags = args32.flags;
4406 arg->version = args32.version;
4407 memcpy(arg->reserved, args32.reserved,
4408 sizeof(args32.reserved));
4413 arg = memdup_user(argp, sizeof(*arg));
4415 return PTR_ERR(arg);
4417 ret = btrfs_ioctl_send(inode, arg);
4422 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
4425 struct btrfs_ioctl_encoded_io_args args = { 0 };
4426 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
4429 struct iovec iovstack[UIO_FASTIOV];
4430 struct iovec *iov = iovstack;
4431 struct iov_iter iter;
4436 if (!capable(CAP_SYS_ADMIN)) {
4442 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4443 struct btrfs_ioctl_encoded_io_args_32 args32;
4445 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
4447 if (copy_from_user(&args32, argp, copy_end)) {
4451 args.iov = compat_ptr(args32.iov);
4452 args.iovcnt = args32.iovcnt;
4453 args.offset = args32.offset;
4454 args.flags = args32.flags;
4459 copy_end = copy_end_kernel;
4460 if (copy_from_user(&args, argp, copy_end)) {
4465 if (args.flags != 0) {
4470 ret = import_iovec(ITER_DEST, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4475 if (iov_iter_count(&iter) == 0) {
4480 ret = rw_verify_area(READ, file, &pos, args.len);
4484 init_sync_kiocb(&kiocb, file);
4487 ret = btrfs_encoded_read(&kiocb, &iter, &args);
4489 fsnotify_access(file);
4490 if (copy_to_user(argp + copy_end,
4491 (char *)&args + copy_end_kernel,
4492 sizeof(args) - copy_end_kernel))
4500 add_rchar(current, ret);
4505 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
4507 struct btrfs_ioctl_encoded_io_args args;
4508 struct iovec iovstack[UIO_FASTIOV];
4509 struct iovec *iov = iovstack;
4510 struct iov_iter iter;
4515 if (!capable(CAP_SYS_ADMIN)) {
4520 if (!(file->f_mode & FMODE_WRITE)) {
4526 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4527 struct btrfs_ioctl_encoded_io_args_32 args32;
4529 if (copy_from_user(&args32, argp, sizeof(args32))) {
4533 args.iov = compat_ptr(args32.iov);
4534 args.iovcnt = args32.iovcnt;
4535 args.offset = args32.offset;
4536 args.flags = args32.flags;
4537 args.len = args32.len;
4538 args.unencoded_len = args32.unencoded_len;
4539 args.unencoded_offset = args32.unencoded_offset;
4540 args.compression = args32.compression;
4541 args.encryption = args32.encryption;
4542 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
4547 if (copy_from_user(&args, argp, sizeof(args))) {
4554 if (args.flags != 0)
4556 if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
4558 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
4559 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
4561 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
4562 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
4564 if (args.unencoded_offset > args.unencoded_len)
4566 if (args.len > args.unencoded_len - args.unencoded_offset)
4569 ret = import_iovec(ITER_SOURCE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
4574 if (iov_iter_count(&iter) == 0) {
4579 ret = rw_verify_area(WRITE, file, &pos, args.len);
4583 init_sync_kiocb(&kiocb, file);
4584 ret = kiocb_set_rw_flags(&kiocb, 0);
4589 file_start_write(file);
4591 ret = btrfs_do_write_iter(&kiocb, &iter, &args);
4593 fsnotify_modify(file);
4595 file_end_write(file);
4600 add_wchar(current, ret);
4605 long btrfs_ioctl(struct file *file, unsigned int
4606 cmd, unsigned long arg)
4608 struct inode *inode = file_inode(file);
4609 struct btrfs_fs_info *fs_info = inode_to_fs_info(inode);
4610 struct btrfs_root *root = BTRFS_I(inode)->root;
4611 void __user *argp = (void __user *)arg;
4614 case FS_IOC_GETVERSION:
4615 return btrfs_ioctl_getversion(inode, argp);
4616 case FS_IOC_GETFSLABEL:
4617 return btrfs_ioctl_get_fslabel(fs_info, argp);
4618 case FS_IOC_SETFSLABEL:
4619 return btrfs_ioctl_set_fslabel(file, argp);
4621 return btrfs_ioctl_fitrim(fs_info, argp);
4622 case BTRFS_IOC_SNAP_CREATE:
4623 return btrfs_ioctl_snap_create(file, argp, 0);
4624 case BTRFS_IOC_SNAP_CREATE_V2:
4625 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4626 case BTRFS_IOC_SUBVOL_CREATE:
4627 return btrfs_ioctl_snap_create(file, argp, 1);
4628 case BTRFS_IOC_SUBVOL_CREATE_V2:
4629 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4630 case BTRFS_IOC_SNAP_DESTROY:
4631 return btrfs_ioctl_snap_destroy(file, argp, false);
4632 case BTRFS_IOC_SNAP_DESTROY_V2:
4633 return btrfs_ioctl_snap_destroy(file, argp, true);
4634 case BTRFS_IOC_SUBVOL_GETFLAGS:
4635 return btrfs_ioctl_subvol_getflags(inode, argp);
4636 case BTRFS_IOC_SUBVOL_SETFLAGS:
4637 return btrfs_ioctl_subvol_setflags(file, argp);
4638 case BTRFS_IOC_DEFAULT_SUBVOL:
4639 return btrfs_ioctl_default_subvol(file, argp);
4640 case BTRFS_IOC_DEFRAG:
4641 return btrfs_ioctl_defrag(file, NULL);
4642 case BTRFS_IOC_DEFRAG_RANGE:
4643 return btrfs_ioctl_defrag(file, argp);
4644 case BTRFS_IOC_RESIZE:
4645 return btrfs_ioctl_resize(file, argp);
4646 case BTRFS_IOC_ADD_DEV:
4647 return btrfs_ioctl_add_dev(fs_info, argp);
4648 case BTRFS_IOC_RM_DEV:
4649 return btrfs_ioctl_rm_dev(file, argp);
4650 case BTRFS_IOC_RM_DEV_V2:
4651 return btrfs_ioctl_rm_dev_v2(file, argp);
4652 case BTRFS_IOC_FS_INFO:
4653 return btrfs_ioctl_fs_info(fs_info, argp);
4654 case BTRFS_IOC_DEV_INFO:
4655 return btrfs_ioctl_dev_info(fs_info, argp);
4656 case BTRFS_IOC_TREE_SEARCH:
4657 return btrfs_ioctl_tree_search(inode, argp);
4658 case BTRFS_IOC_TREE_SEARCH_V2:
4659 return btrfs_ioctl_tree_search_v2(inode, argp);
4660 case BTRFS_IOC_INO_LOOKUP:
4661 return btrfs_ioctl_ino_lookup(root, argp);
4662 case BTRFS_IOC_INO_PATHS:
4663 return btrfs_ioctl_ino_to_path(root, argp);
4664 case BTRFS_IOC_LOGICAL_INO:
4665 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4666 case BTRFS_IOC_LOGICAL_INO_V2:
4667 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4668 case BTRFS_IOC_SPACE_INFO:
4669 return btrfs_ioctl_space_info(fs_info, argp);
4670 case BTRFS_IOC_SYNC: {
4673 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4676 ret = btrfs_sync_fs(inode->i_sb, 1);
4678 * The transaction thread may want to do more work,
4679 * namely it pokes the cleaner kthread that will start
4680 * processing uncleaned subvols.
4682 wake_up_process(fs_info->transaction_kthread);
4685 case BTRFS_IOC_START_SYNC:
4686 return btrfs_ioctl_start_sync(root, argp);
4687 case BTRFS_IOC_WAIT_SYNC:
4688 return btrfs_ioctl_wait_sync(fs_info, argp);
4689 case BTRFS_IOC_SCRUB:
4690 return btrfs_ioctl_scrub(file, argp);
4691 case BTRFS_IOC_SCRUB_CANCEL:
4692 return btrfs_ioctl_scrub_cancel(fs_info);
4693 case BTRFS_IOC_SCRUB_PROGRESS:
4694 return btrfs_ioctl_scrub_progress(fs_info, argp);
4695 case BTRFS_IOC_BALANCE_V2:
4696 return btrfs_ioctl_balance(file, argp);
4697 case BTRFS_IOC_BALANCE_CTL:
4698 return btrfs_ioctl_balance_ctl(fs_info, arg);
4699 case BTRFS_IOC_BALANCE_PROGRESS:
4700 return btrfs_ioctl_balance_progress(fs_info, argp);
4701 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4702 return btrfs_ioctl_set_received_subvol(file, argp);
4704 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4705 return btrfs_ioctl_set_received_subvol_32(file, argp);
4707 case BTRFS_IOC_SEND:
4708 return _btrfs_ioctl_send(inode, argp, false);
4709 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4710 case BTRFS_IOC_SEND_32:
4711 return _btrfs_ioctl_send(inode, argp, true);
4713 case BTRFS_IOC_GET_DEV_STATS:
4714 return btrfs_ioctl_get_dev_stats(fs_info, argp);
4715 case BTRFS_IOC_QUOTA_CTL:
4716 return btrfs_ioctl_quota_ctl(file, argp);
4717 case BTRFS_IOC_QGROUP_ASSIGN:
4718 return btrfs_ioctl_qgroup_assign(file, argp);
4719 case BTRFS_IOC_QGROUP_CREATE:
4720 return btrfs_ioctl_qgroup_create(file, argp);
4721 case BTRFS_IOC_QGROUP_LIMIT:
4722 return btrfs_ioctl_qgroup_limit(file, argp);
4723 case BTRFS_IOC_QUOTA_RESCAN:
4724 return btrfs_ioctl_quota_rescan(file, argp);
4725 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4726 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
4727 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4728 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
4729 case BTRFS_IOC_DEV_REPLACE:
4730 return btrfs_ioctl_dev_replace(fs_info, argp);
4731 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4732 return btrfs_ioctl_get_supported_features(argp);
4733 case BTRFS_IOC_GET_FEATURES:
4734 return btrfs_ioctl_get_features(fs_info, argp);
4735 case BTRFS_IOC_SET_FEATURES:
4736 return btrfs_ioctl_set_features(file, argp);
4737 case BTRFS_IOC_GET_SUBVOL_INFO:
4738 return btrfs_ioctl_get_subvol_info(inode, argp);
4739 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
4740 return btrfs_ioctl_get_subvol_rootref(root, argp);
4741 case BTRFS_IOC_INO_LOOKUP_USER:
4742 return btrfs_ioctl_ino_lookup_user(file, argp);
4743 case FS_IOC_ENABLE_VERITY:
4744 return fsverity_ioctl_enable(file, (const void __user *)argp);
4745 case FS_IOC_MEASURE_VERITY:
4746 return fsverity_ioctl_measure(file, argp);
4747 case BTRFS_IOC_ENCODED_READ:
4748 return btrfs_ioctl_encoded_read(file, argp, false);
4749 case BTRFS_IOC_ENCODED_WRITE:
4750 return btrfs_ioctl_encoded_write(file, argp, false);
4751 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4752 case BTRFS_IOC_ENCODED_READ_32:
4753 return btrfs_ioctl_encoded_read(file, argp, true);
4754 case BTRFS_IOC_ENCODED_WRITE_32:
4755 return btrfs_ioctl_encoded_write(file, argp, true);
4762 #ifdef CONFIG_COMPAT
4763 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
4766 * These all access 32-bit values anyway so no further
4767 * handling is necessary.
4770 case FS_IOC32_GETVERSION:
4771 cmd = FS_IOC_GETVERSION;
4775 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));