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>
34 #include "transaction.h"
35 #include "btrfs_inode.h"
36 #include "print-tree.h"
40 #include "rcu-string.h"
42 #include "dev-replace.h"
47 #include "compression.h"
48 #include "space-info.h"
49 #include "delalloc-space.h"
50 #include "block-group.h"
53 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
54 * structures are incorrect, as the timespec structure from userspace
55 * is 4 bytes too small. We define these alternatives here to teach
56 * the kernel about the 32-bit struct packing.
58 struct btrfs_ioctl_timespec_32 {
61 } __attribute__ ((__packed__));
63 struct btrfs_ioctl_received_subvol_args_32 {
64 char uuid[BTRFS_UUID_SIZE]; /* in */
65 __u64 stransid; /* in */
66 __u64 rtransid; /* out */
67 struct btrfs_ioctl_timespec_32 stime; /* in */
68 struct btrfs_ioctl_timespec_32 rtime; /* out */
70 __u64 reserved[16]; /* in */
71 } __attribute__ ((__packed__));
73 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
74 struct btrfs_ioctl_received_subvol_args_32)
77 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
78 struct btrfs_ioctl_send_args_32 {
79 __s64 send_fd; /* in */
80 __u64 clone_sources_count; /* in */
81 compat_uptr_t clone_sources; /* in */
82 __u64 parent_root; /* in */
84 __u64 reserved[4]; /* in */
85 } __attribute__ ((__packed__));
87 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
88 struct btrfs_ioctl_send_args_32)
91 /* Mask out flags that are inappropriate for the given type of inode. */
92 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
95 if (S_ISDIR(inode->i_mode))
97 else if (S_ISREG(inode->i_mode))
98 return flags & ~FS_DIRSYNC_FL;
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
107 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
109 unsigned int iflags = 0;
110 u32 flags = binode->flags;
111 u32 ro_flags = binode->ro_flags;
113 if (flags & BTRFS_INODE_SYNC)
114 iflags |= FS_SYNC_FL;
115 if (flags & BTRFS_INODE_IMMUTABLE)
116 iflags |= FS_IMMUTABLE_FL;
117 if (flags & BTRFS_INODE_APPEND)
118 iflags |= FS_APPEND_FL;
119 if (flags & BTRFS_INODE_NODUMP)
120 iflags |= FS_NODUMP_FL;
121 if (flags & BTRFS_INODE_NOATIME)
122 iflags |= FS_NOATIME_FL;
123 if (flags & BTRFS_INODE_DIRSYNC)
124 iflags |= FS_DIRSYNC_FL;
125 if (flags & BTRFS_INODE_NODATACOW)
126 iflags |= FS_NOCOW_FL;
127 if (ro_flags & BTRFS_INODE_RO_VERITY)
128 iflags |= FS_VERITY_FL;
130 if (flags & BTRFS_INODE_NOCOMPRESS)
131 iflags |= FS_NOCOMP_FL;
132 else if (flags & BTRFS_INODE_COMPRESS)
133 iflags |= FS_COMPR_FL;
139 * Update inode->i_flags based on the btrfs internal flags.
141 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
143 struct btrfs_inode *binode = BTRFS_I(inode);
144 unsigned int new_fl = 0;
146 if (binode->flags & BTRFS_INODE_SYNC)
148 if (binode->flags & BTRFS_INODE_IMMUTABLE)
149 new_fl |= S_IMMUTABLE;
150 if (binode->flags & BTRFS_INODE_APPEND)
152 if (binode->flags & BTRFS_INODE_NOATIME)
154 if (binode->flags & BTRFS_INODE_DIRSYNC)
156 if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
159 set_mask_bits(&inode->i_flags,
160 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
165 * Check if @flags are a supported and valid set of FS_*_FL flags and that
166 * the old and new flags are not conflicting
168 static int check_fsflags(unsigned int old_flags, unsigned int flags)
170 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
171 FS_NOATIME_FL | FS_NODUMP_FL | \
172 FS_SYNC_FL | FS_DIRSYNC_FL | \
173 FS_NOCOMP_FL | FS_COMPR_FL |
177 /* COMPR and NOCOMP on new/old are valid */
178 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
181 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
184 /* NOCOW and compression options are mutually exclusive */
185 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
187 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
193 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
196 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
203 * Set flags/xflags from the internal inode flags. The remaining items of
204 * fsxattr are zeroed.
206 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
208 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
210 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
214 int btrfs_fileattr_set(struct user_namespace *mnt_userns,
215 struct dentry *dentry, struct fileattr *fa)
217 struct inode *inode = d_inode(dentry);
218 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
219 struct btrfs_inode *binode = BTRFS_I(inode);
220 struct btrfs_root *root = binode->root;
221 struct btrfs_trans_handle *trans;
222 unsigned int fsflags, old_fsflags;
224 const char *comp = NULL;
227 if (btrfs_root_readonly(root))
230 if (fileattr_has_fsx(fa))
233 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
234 old_fsflags = btrfs_inode_flags_to_fsflags(binode);
235 ret = check_fsflags(old_fsflags, fsflags);
239 ret = check_fsflags_compatible(fs_info, fsflags);
243 binode_flags = binode->flags;
244 if (fsflags & FS_SYNC_FL)
245 binode_flags |= BTRFS_INODE_SYNC;
247 binode_flags &= ~BTRFS_INODE_SYNC;
248 if (fsflags & FS_IMMUTABLE_FL)
249 binode_flags |= BTRFS_INODE_IMMUTABLE;
251 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
252 if (fsflags & FS_APPEND_FL)
253 binode_flags |= BTRFS_INODE_APPEND;
255 binode_flags &= ~BTRFS_INODE_APPEND;
256 if (fsflags & FS_NODUMP_FL)
257 binode_flags |= BTRFS_INODE_NODUMP;
259 binode_flags &= ~BTRFS_INODE_NODUMP;
260 if (fsflags & FS_NOATIME_FL)
261 binode_flags |= BTRFS_INODE_NOATIME;
263 binode_flags &= ~BTRFS_INODE_NOATIME;
265 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
266 if (!fa->flags_valid) {
267 /* 1 item for the inode */
268 trans = btrfs_start_transaction(root, 1);
270 return PTR_ERR(trans);
274 if (fsflags & FS_DIRSYNC_FL)
275 binode_flags |= BTRFS_INODE_DIRSYNC;
277 binode_flags &= ~BTRFS_INODE_DIRSYNC;
278 if (fsflags & FS_NOCOW_FL) {
279 if (S_ISREG(inode->i_mode)) {
281 * It's safe to turn csums off here, no extents exist.
282 * Otherwise we want the flag to reflect the real COW
283 * status of the file and will not set it.
285 if (inode->i_size == 0)
286 binode_flags |= BTRFS_INODE_NODATACOW |
287 BTRFS_INODE_NODATASUM;
289 binode_flags |= BTRFS_INODE_NODATACOW;
293 * Revert back under same assumptions as above
295 if (S_ISREG(inode->i_mode)) {
296 if (inode->i_size == 0)
297 binode_flags &= ~(BTRFS_INODE_NODATACOW |
298 BTRFS_INODE_NODATASUM);
300 binode_flags &= ~BTRFS_INODE_NODATACOW;
305 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
306 * flag may be changed automatically if compression code won't make
309 if (fsflags & FS_NOCOMP_FL) {
310 binode_flags &= ~BTRFS_INODE_COMPRESS;
311 binode_flags |= BTRFS_INODE_NOCOMPRESS;
312 } else if (fsflags & FS_COMPR_FL) {
314 if (IS_SWAPFILE(inode))
317 binode_flags |= BTRFS_INODE_COMPRESS;
318 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
320 comp = btrfs_compress_type2str(fs_info->compress_type);
321 if (!comp || comp[0] == 0)
322 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
324 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
331 trans = btrfs_start_transaction(root, 3);
333 return PTR_ERR(trans);
336 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
339 btrfs_abort_transaction(trans, ret);
343 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
345 if (ret && ret != -ENODATA) {
346 btrfs_abort_transaction(trans, ret);
352 binode->flags = binode_flags;
353 btrfs_sync_inode_flags_to_i_flags(inode);
354 inode_inc_iversion(inode);
355 inode->i_ctime = current_time(inode);
356 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
359 btrfs_end_transaction(trans);
364 * Start exclusive operation @type, return true on success
366 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
367 enum btrfs_exclusive_operation type)
371 spin_lock(&fs_info->super_lock);
372 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
373 fs_info->exclusive_operation = type;
376 spin_unlock(&fs_info->super_lock);
382 * Conditionally allow to enter the exclusive operation in case it's compatible
383 * with the running one. This must be paired with btrfs_exclop_start_unlock and
384 * btrfs_exclop_finish.
387 * - the same type is already running
388 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
389 * must check the condition first that would allow none -> @type
391 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
392 enum btrfs_exclusive_operation type)
394 spin_lock(&fs_info->super_lock);
395 if (fs_info->exclusive_operation == type)
398 spin_unlock(&fs_info->super_lock);
402 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
404 spin_unlock(&fs_info->super_lock);
407 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
409 spin_lock(&fs_info->super_lock);
410 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
411 spin_unlock(&fs_info->super_lock);
412 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
415 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
417 struct inode *inode = file_inode(file);
419 return put_user(inode->i_generation, arg);
422 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
425 struct btrfs_device *device;
426 struct request_queue *q;
427 struct fstrim_range range;
428 u64 minlen = ULLONG_MAX;
432 if (!capable(CAP_SYS_ADMIN))
436 * btrfs_trim_block_group() depends on space cache, which is not
437 * available in zoned filesystem. So, disallow fitrim on a zoned
438 * filesystem for now.
440 if (btrfs_is_zoned(fs_info))
444 * If the fs is mounted with nologreplay, which requires it to be
445 * mounted in RO mode as well, we can not allow discard on free space
446 * inside block groups, because log trees refer to extents that are not
447 * pinned in a block group's free space cache (pinning the extents is
448 * precisely the first phase of replaying a log tree).
450 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
454 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
458 q = bdev_get_queue(device->bdev);
459 if (blk_queue_discard(q)) {
461 minlen = min_t(u64, q->limits.discard_granularity,
469 if (copy_from_user(&range, arg, sizeof(range)))
473 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
474 * block group is in the logical address space, which can be any
475 * sectorsize aligned bytenr in the range [0, U64_MAX].
477 if (range.len < fs_info->sb->s_blocksize)
480 range.minlen = max(range.minlen, minlen);
481 ret = btrfs_trim_fs(fs_info, &range);
485 if (copy_to_user(arg, &range, sizeof(range)))
491 int __pure btrfs_is_empty_uuid(u8 *uuid)
495 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
502 static noinline int create_subvol(struct user_namespace *mnt_userns,
503 struct inode *dir, struct dentry *dentry,
504 const char *name, int namelen,
505 struct btrfs_qgroup_inherit *inherit)
507 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
508 struct btrfs_trans_handle *trans;
509 struct btrfs_key key;
510 struct btrfs_root_item *root_item;
511 struct btrfs_inode_item *inode_item;
512 struct extent_buffer *leaf;
513 struct btrfs_root *root = BTRFS_I(dir)->root;
514 struct btrfs_root *new_root;
515 struct btrfs_block_rsv block_rsv;
516 struct timespec64 cur_time = current_time(dir);
524 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
528 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
532 ret = get_anon_bdev(&anon_dev);
537 * Don't create subvolume whose level is not zero. Or qgroup will be
538 * screwed up since it assumes subvolume qgroup's level to be 0.
540 if (btrfs_qgroup_level(objectid)) {
545 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
547 * The same as the snapshot creation, please see the comment
548 * of create_snapshot().
550 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
554 trans = btrfs_start_transaction(root, 0);
556 ret = PTR_ERR(trans);
557 btrfs_subvolume_release_metadata(root, &block_rsv);
560 trans->block_rsv = &block_rsv;
561 trans->bytes_reserved = block_rsv.size;
563 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
567 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
568 BTRFS_NESTING_NORMAL);
574 btrfs_mark_buffer_dirty(leaf);
576 inode_item = &root_item->inode;
577 btrfs_set_stack_inode_generation(inode_item, 1);
578 btrfs_set_stack_inode_size(inode_item, 3);
579 btrfs_set_stack_inode_nlink(inode_item, 1);
580 btrfs_set_stack_inode_nbytes(inode_item,
582 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
584 btrfs_set_root_flags(root_item, 0);
585 btrfs_set_root_limit(root_item, 0);
586 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
588 btrfs_set_root_bytenr(root_item, leaf->start);
589 btrfs_set_root_generation(root_item, trans->transid);
590 btrfs_set_root_level(root_item, 0);
591 btrfs_set_root_refs(root_item, 1);
592 btrfs_set_root_used(root_item, leaf->len);
593 btrfs_set_root_last_snapshot(root_item, 0);
595 btrfs_set_root_generation_v2(root_item,
596 btrfs_root_generation(root_item));
597 generate_random_guid(root_item->uuid);
598 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
599 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
600 root_item->ctime = root_item->otime;
601 btrfs_set_root_ctransid(root_item, trans->transid);
602 btrfs_set_root_otransid(root_item, trans->transid);
604 btrfs_tree_unlock(leaf);
606 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
608 key.objectid = objectid;
610 key.type = BTRFS_ROOT_ITEM_KEY;
611 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
615 * Since we don't abort the transaction in this case, free the
616 * tree block so that we don't leak space and leave the
617 * filesystem in an inconsistent state (an extent item in the
618 * extent tree with a backreference for a root that does not
621 btrfs_tree_lock(leaf);
622 btrfs_clean_tree_block(leaf);
623 btrfs_tree_unlock(leaf);
624 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
625 free_extent_buffer(leaf);
629 free_extent_buffer(leaf);
632 key.offset = (u64)-1;
633 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
634 if (IS_ERR(new_root)) {
635 free_anon_bdev(anon_dev);
636 ret = PTR_ERR(new_root);
637 btrfs_abort_transaction(trans, ret);
640 /* Freeing will be done in btrfs_put_root() of new_root */
643 ret = btrfs_record_root_in_trans(trans, new_root);
645 btrfs_put_root(new_root);
646 btrfs_abort_transaction(trans, ret);
650 ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns);
651 btrfs_put_root(new_root);
653 /* We potentially lose an unused inode item here */
654 btrfs_abort_transaction(trans, ret);
659 * insert the directory item
661 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
663 btrfs_abort_transaction(trans, ret);
667 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
668 BTRFS_FT_DIR, index);
670 btrfs_abort_transaction(trans, ret);
674 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
675 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
677 btrfs_abort_transaction(trans, ret);
681 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
682 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
684 btrfs_abort_transaction(trans, ret);
688 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
689 BTRFS_UUID_KEY_SUBVOL, objectid);
691 btrfs_abort_transaction(trans, ret);
695 trans->block_rsv = NULL;
696 trans->bytes_reserved = 0;
697 btrfs_subvolume_release_metadata(root, &block_rsv);
699 err = btrfs_commit_transaction(trans);
704 inode = btrfs_lookup_dentry(dir, dentry);
706 return PTR_ERR(inode);
707 d_instantiate(dentry, inode);
713 free_anon_bdev(anon_dev);
718 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
719 struct dentry *dentry, bool readonly,
720 struct btrfs_qgroup_inherit *inherit)
722 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
724 struct btrfs_pending_snapshot *pending_snapshot;
725 struct btrfs_trans_handle *trans;
728 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
731 if (atomic_read(&root->nr_swapfiles)) {
733 "cannot snapshot subvolume with active swapfile");
737 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
738 if (!pending_snapshot)
741 ret = get_anon_bdev(&pending_snapshot->anon_dev);
744 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
746 pending_snapshot->path = btrfs_alloc_path();
747 if (!pending_snapshot->root_item || !pending_snapshot->path) {
752 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
753 BTRFS_BLOCK_RSV_TEMP);
755 * 1 - parent dir inode
758 * 2 - root ref/backref
759 * 1 - root of snapshot
762 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
763 &pending_snapshot->block_rsv, 8,
768 pending_snapshot->dentry = dentry;
769 pending_snapshot->root = root;
770 pending_snapshot->readonly = readonly;
771 pending_snapshot->dir = dir;
772 pending_snapshot->inherit = inherit;
774 trans = btrfs_start_transaction(root, 0);
776 ret = PTR_ERR(trans);
780 trans->pending_snapshot = pending_snapshot;
782 ret = btrfs_commit_transaction(trans);
786 ret = pending_snapshot->error;
790 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
794 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
796 ret = PTR_ERR(inode);
800 d_instantiate(dentry, inode);
802 pending_snapshot->anon_dev = 0;
804 /* Prevent double freeing of anon_dev */
805 if (ret && pending_snapshot->snap)
806 pending_snapshot->snap->anon_dev = 0;
807 btrfs_put_root(pending_snapshot->snap);
808 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
810 if (pending_snapshot->anon_dev)
811 free_anon_bdev(pending_snapshot->anon_dev);
812 kfree(pending_snapshot->root_item);
813 btrfs_free_path(pending_snapshot->path);
814 kfree(pending_snapshot);
819 /* copy of may_delete in fs/namei.c()
820 * Check whether we can remove a link victim from directory dir, check
821 * whether the type of victim is right.
822 * 1. We can't do it if dir is read-only (done in permission())
823 * 2. We should have write and exec permissions on dir
824 * 3. We can't remove anything from append-only dir
825 * 4. We can't do anything with immutable dir (done in permission())
826 * 5. If the sticky bit on dir is set we should either
827 * a. be owner of dir, or
828 * b. be owner of victim, or
829 * c. have CAP_FOWNER capability
830 * 6. If the victim is append-only or immutable we can't do anything with
831 * links pointing to it.
832 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
833 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
834 * 9. We can't remove a root or mountpoint.
835 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
836 * nfs_async_unlink().
839 static int btrfs_may_delete(struct user_namespace *mnt_userns,
840 struct inode *dir, struct dentry *victim, int isdir)
844 if (d_really_is_negative(victim))
847 BUG_ON(d_inode(victim->d_parent) != dir);
848 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
850 error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
855 if (check_sticky(mnt_userns, dir, d_inode(victim)) ||
856 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
857 IS_SWAPFILE(d_inode(victim)))
860 if (!d_is_dir(victim))
864 } else if (d_is_dir(victim))
868 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
873 /* copy of may_create in fs/namei.c() */
874 static inline int btrfs_may_create(struct user_namespace *mnt_userns,
875 struct inode *dir, struct dentry *child)
877 if (d_really_is_positive(child))
881 if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns))
883 return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
887 * Create a new subvolume below @parent. This is largely modeled after
888 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
889 * inside this filesystem so it's quite a bit simpler.
891 static noinline int btrfs_mksubvol(const struct path *parent,
892 struct user_namespace *mnt_userns,
893 const char *name, int namelen,
894 struct btrfs_root *snap_src,
896 struct btrfs_qgroup_inherit *inherit)
898 struct inode *dir = d_inode(parent->dentry);
899 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
900 struct dentry *dentry;
903 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
907 dentry = lookup_one(mnt_userns, name, parent->dentry, namelen);
908 error = PTR_ERR(dentry);
912 error = btrfs_may_create(mnt_userns, dir, dentry);
917 * even if this name doesn't exist, we may get hash collisions.
918 * check for them now when we can safely fail
920 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
926 down_read(&fs_info->subvol_sem);
928 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
932 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
934 error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit);
937 fsnotify_mkdir(dir, dentry);
939 up_read(&fs_info->subvol_sem);
943 btrfs_inode_unlock(dir, 0);
947 static noinline int btrfs_mksnapshot(const struct path *parent,
948 struct user_namespace *mnt_userns,
949 const char *name, int namelen,
950 struct btrfs_root *root,
952 struct btrfs_qgroup_inherit *inherit)
955 bool snapshot_force_cow = false;
958 * Force new buffered writes to reserve space even when NOCOW is
959 * possible. This is to avoid later writeback (running dealloc) to
960 * fallback to COW mode and unexpectedly fail with ENOSPC.
962 btrfs_drew_read_lock(&root->snapshot_lock);
964 ret = btrfs_start_delalloc_snapshot(root, false);
969 * All previous writes have started writeback in NOCOW mode, so now
970 * we force future writes to fallback to COW mode during snapshot
973 atomic_inc(&root->snapshot_force_cow);
974 snapshot_force_cow = true;
976 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
978 ret = btrfs_mksubvol(parent, mnt_userns, name, namelen,
979 root, readonly, inherit);
981 if (snapshot_force_cow)
982 atomic_dec(&root->snapshot_force_cow);
983 btrfs_drew_read_unlock(&root->snapshot_lock);
988 * When we're defragging a range, we don't want to kick it off again
989 * if it is really just waiting for delalloc to send it down.
990 * If we find a nice big extent or delalloc range for the bytes in the
991 * file you want to defrag, we return 0 to let you know to skip this
994 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
996 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
997 struct extent_map *em = NULL;
998 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1001 read_lock(&em_tree->lock);
1002 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1003 read_unlock(&em_tree->lock);
1006 end = extent_map_end(em);
1007 free_extent_map(em);
1008 if (end - offset > thresh)
1011 /* if we already have a nice delalloc here, just stop */
1013 end = count_range_bits(io_tree, &offset, offset + thresh,
1014 thresh, EXTENT_DELALLOC, 1);
1021 * helper function to walk through a file and find extents
1022 * newer than a specific transid, and smaller than thresh.
1024 * This is used by the defragging code to find new and small
1027 static int find_new_extents(struct btrfs_root *root,
1028 struct inode *inode, u64 newer_than,
1029 u64 *off, u32 thresh)
1031 struct btrfs_path *path;
1032 struct btrfs_key min_key;
1033 struct extent_buffer *leaf;
1034 struct btrfs_file_extent_item *extent;
1037 u64 ino = btrfs_ino(BTRFS_I(inode));
1039 path = btrfs_alloc_path();
1043 min_key.objectid = ino;
1044 min_key.type = BTRFS_EXTENT_DATA_KEY;
1045 min_key.offset = *off;
1048 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1052 if (min_key.objectid != ino)
1054 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1057 leaf = path->nodes[0];
1058 extent = btrfs_item_ptr(leaf, path->slots[0],
1059 struct btrfs_file_extent_item);
1061 type = btrfs_file_extent_type(leaf, extent);
1062 if (type == BTRFS_FILE_EXTENT_REG &&
1063 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1064 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1065 *off = min_key.offset;
1066 btrfs_free_path(path);
1071 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1072 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1076 if (min_key.offset == (u64)-1)
1080 btrfs_release_path(path);
1083 btrfs_free_path(path);
1087 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1089 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1090 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1091 struct extent_map *em;
1092 u64 len = PAGE_SIZE;
1095 * hopefully we have this extent in the tree already, try without
1096 * the full extent lock
1098 read_lock(&em_tree->lock);
1099 em = lookup_extent_mapping(em_tree, start, len);
1100 read_unlock(&em_tree->lock);
1103 struct extent_state *cached = NULL;
1104 u64 end = start + len - 1;
1106 /* get the big lock and read metadata off disk */
1107 lock_extent_bits(io_tree, start, end, &cached);
1108 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1109 unlock_extent_cached(io_tree, start, end, &cached);
1118 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1120 struct extent_map *next;
1123 /* this is the last extent */
1124 if (em->start + em->len >= i_size_read(inode))
1127 next = defrag_lookup_extent(inode, em->start + em->len);
1128 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1130 else if ((em->block_start + em->block_len == next->block_start) &&
1131 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1134 free_extent_map(next);
1138 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1139 u64 *last_len, u64 *skip, u64 *defrag_end,
1142 struct extent_map *em;
1144 bool next_mergeable = true;
1145 bool prev_mergeable = true;
1148 * make sure that once we start defragging an extent, we keep on
1151 if (start < *defrag_end)
1156 em = defrag_lookup_extent(inode, start);
1160 /* this will cover holes, and inline extents */
1161 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1167 prev_mergeable = false;
1169 next_mergeable = defrag_check_next_extent(inode, em);
1171 * we hit a real extent, if it is big or the next extent is not a
1172 * real extent, don't bother defragging it
1174 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1175 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1179 * last_len ends up being a counter of how many bytes we've defragged.
1180 * every time we choose not to defrag an extent, we reset *last_len
1181 * so that the next tiny extent will force a defrag.
1183 * The end result of this is that tiny extents before a single big
1184 * extent will force at least part of that big extent to be defragged.
1187 *defrag_end = extent_map_end(em);
1190 *skip = extent_map_end(em);
1194 free_extent_map(em);
1199 * it doesn't do much good to defrag one or two pages
1200 * at a time. This pulls in a nice chunk of pages
1201 * to COW and defrag.
1203 * It also makes sure the delalloc code has enough
1204 * dirty data to avoid making new small extents as part
1207 * It's a good idea to start RA on this range
1208 * before calling this.
1210 static int cluster_pages_for_defrag(struct inode *inode,
1211 struct page **pages,
1212 unsigned long start_index,
1213 unsigned long num_pages)
1215 unsigned long file_end;
1216 u64 isize = i_size_read(inode);
1220 u64 start = (u64)start_index << PAGE_SHIFT;
1225 struct btrfs_ordered_extent *ordered;
1226 struct extent_state *cached_state = NULL;
1227 struct extent_io_tree *tree;
1228 struct extent_changeset *data_reserved = NULL;
1229 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1231 file_end = (isize - 1) >> PAGE_SHIFT;
1232 if (!isize || start_index > file_end)
1235 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1237 ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
1238 start, page_cnt << PAGE_SHIFT);
1242 tree = &BTRFS_I(inode)->io_tree;
1244 /* step one, lock all the pages */
1245 for (i = 0; i < page_cnt; i++) {
1248 page = find_or_create_page(inode->i_mapping,
1249 start_index + i, mask);
1253 ret = set_page_extent_mapped(page);
1260 page_start = page_offset(page);
1261 page_end = page_start + PAGE_SIZE - 1;
1263 lock_extent_bits(tree, page_start, page_end,
1265 ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode),
1267 unlock_extent_cached(tree, page_start, page_end,
1273 btrfs_start_ordered_extent(ordered, 1);
1274 btrfs_put_ordered_extent(ordered);
1277 * we unlocked the page above, so we need check if
1278 * it was released or not.
1280 if (page->mapping != inode->i_mapping) {
1287 if (!PageUptodate(page)) {
1288 btrfs_readpage(NULL, page);
1290 if (!PageUptodate(page)) {
1298 if (page->mapping != inode->i_mapping) {
1310 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1314 * so now we have a nice long stream of locked
1315 * and up to date pages, lets wait on them
1317 for (i = 0; i < i_done; i++)
1318 wait_on_page_writeback(pages[i]);
1320 page_start = page_offset(pages[0]);
1321 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1323 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1324 page_start, page_end - 1, &cached_state);
1327 * When defragmenting we skip ranges that have holes or inline extents,
1328 * (check should_defrag_range()), to avoid unnecessary IO and wasting
1329 * space. At btrfs_defrag_file(), we check if a range should be defragged
1330 * before locking the inode and then, if it should, we trigger a sync
1331 * page cache readahead - we lock the inode only after that to avoid
1332 * blocking for too long other tasks that possibly want to operate on
1333 * other file ranges. But before we were able to get the inode lock,
1334 * some other task may have punched a hole in the range, or we may have
1335 * now an inline extent, in which case we should not defrag. So check
1336 * for that here, where we have the inode and the range locked, and bail
1337 * out if that happened.
1339 search_start = page_start;
1340 while (search_start < page_end) {
1341 struct extent_map *em;
1343 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, search_start,
1344 page_end - search_start);
1347 goto out_unlock_range;
1349 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1350 free_extent_map(em);
1351 /* Ok, 0 means we did not defrag anything */
1353 goto out_unlock_range;
1355 search_start = extent_map_end(em);
1356 free_extent_map(em);
1359 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1360 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1361 EXTENT_DEFRAG, 0, 0, &cached_state);
1363 if (i_done != page_cnt) {
1364 spin_lock(&BTRFS_I(inode)->lock);
1365 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1366 spin_unlock(&BTRFS_I(inode)->lock);
1367 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1368 start, (page_cnt - i_done) << PAGE_SHIFT, true);
1372 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1375 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1376 page_start, page_end - 1, &cached_state);
1378 for (i = 0; i < i_done; i++) {
1379 clear_page_dirty_for_io(pages[i]);
1380 ClearPageChecked(pages[i]);
1381 set_page_dirty(pages[i]);
1382 unlock_page(pages[i]);
1385 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1386 extent_changeset_free(data_reserved);
1390 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1391 page_start, page_end - 1, &cached_state);
1393 for (i = 0; i < i_done; i++) {
1394 unlock_page(pages[i]);
1397 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1398 start, page_cnt << PAGE_SHIFT, true);
1399 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1400 extent_changeset_free(data_reserved);
1405 int btrfs_defrag_file(struct inode *inode, struct file *file,
1406 struct btrfs_ioctl_defrag_range_args *range,
1407 u64 newer_than, unsigned long max_to_defrag)
1409 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1410 struct btrfs_root *root = BTRFS_I(inode)->root;
1411 struct file_ra_state *ra = NULL;
1412 unsigned long last_index;
1413 u64 isize = i_size_read(inode);
1417 u64 newer_off = range->start;
1419 unsigned long ra_index = 0;
1421 int defrag_count = 0;
1422 int compress_type = BTRFS_COMPRESS_ZLIB;
1423 u32 extent_thresh = range->extent_thresh;
1424 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1425 unsigned long cluster = max_cluster;
1426 u64 new_align = ~((u64)SZ_128K - 1);
1427 struct page **pages = NULL;
1428 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1433 if (range->start >= isize)
1437 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1439 if (range->compress_type)
1440 compress_type = range->compress_type;
1443 if (extent_thresh == 0)
1444 extent_thresh = SZ_256K;
1447 * If we were not given a file, allocate a readahead context. As
1448 * readahead is just an optimization, defrag will work without it so
1449 * we don't error out.
1452 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1454 file_ra_state_init(ra, inode->i_mapping);
1459 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1465 /* find the last page to defrag */
1466 if (range->start + range->len > range->start) {
1467 last_index = min_t(u64, isize - 1,
1468 range->start + range->len - 1) >> PAGE_SHIFT;
1470 last_index = (isize - 1) >> PAGE_SHIFT;
1474 ret = find_new_extents(root, inode, newer_than,
1475 &newer_off, SZ_64K);
1477 range->start = newer_off;
1479 * we always align our defrag to help keep
1480 * the extents in the file evenly spaced
1482 i = (newer_off & new_align) >> PAGE_SHIFT;
1486 i = range->start >> PAGE_SHIFT;
1489 max_to_defrag = last_index - i + 1;
1492 * make writeback starts from i, so the defrag range can be
1493 * written sequentially.
1495 if (i < inode->i_mapping->writeback_index)
1496 inode->i_mapping->writeback_index = i;
1498 while (i <= last_index && defrag_count < max_to_defrag &&
1499 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1501 * make sure we stop running if someone unmounts
1504 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1507 if (btrfs_defrag_cancelled(fs_info)) {
1508 btrfs_debug(fs_info, "defrag_file cancelled");
1513 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1514 extent_thresh, &last_len, &skip,
1515 &defrag_end, do_compress)){
1518 * the should_defrag function tells us how much to skip
1519 * bump our counter by the suggested amount
1521 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1522 i = max(i + 1, next);
1527 cluster = (PAGE_ALIGN(defrag_end) >>
1529 cluster = min(cluster, max_cluster);
1531 cluster = max_cluster;
1534 if (i + cluster > ra_index) {
1535 ra_index = max(i, ra_index);
1537 page_cache_sync_readahead(inode->i_mapping, ra,
1538 file, ra_index, cluster);
1539 ra_index += cluster;
1542 btrfs_inode_lock(inode, 0);
1543 if (IS_SWAPFILE(inode)) {
1547 BTRFS_I(inode)->defrag_compress = compress_type;
1548 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1551 btrfs_inode_unlock(inode, 0);
1555 defrag_count += ret;
1556 balance_dirty_pages_ratelimited(inode->i_mapping);
1557 btrfs_inode_unlock(inode, 0);
1560 if (newer_off == (u64)-1)
1566 newer_off = max(newer_off + 1,
1567 (u64)i << PAGE_SHIFT);
1569 ret = find_new_extents(root, inode, newer_than,
1570 &newer_off, SZ_64K);
1572 range->start = newer_off;
1573 i = (newer_off & new_align) >> PAGE_SHIFT;
1580 last_len += ret << PAGE_SHIFT;
1590 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1591 filemap_flush(inode->i_mapping);
1592 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1593 &BTRFS_I(inode)->runtime_flags))
1594 filemap_flush(inode->i_mapping);
1597 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1598 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1599 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1600 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1605 btrfs_inode_lock(inode, 0);
1606 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1607 btrfs_inode_unlock(inode, 0);
1616 * Try to start exclusive operation @type or cancel it if it's running.
1619 * 0 - normal mode, newly claimed op started
1620 * >0 - normal mode, something else is running,
1621 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1622 * ECANCELED - cancel mode, successful cancel
1623 * ENOTCONN - cancel mode, operation not running anymore
1625 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1626 enum btrfs_exclusive_operation type, bool cancel)
1629 /* Start normal op */
1630 if (!btrfs_exclop_start(fs_info, type))
1631 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1632 /* Exclusive operation is now claimed */
1636 /* Cancel running op */
1637 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1639 * This blocks any exclop finish from setting it to NONE, so we
1640 * request cancellation. Either it runs and we will wait for it,
1641 * or it has finished and no waiting will happen.
1643 atomic_inc(&fs_info->reloc_cancel_req);
1644 btrfs_exclop_start_unlock(fs_info);
1646 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1647 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1648 TASK_INTERRUPTIBLE);
1653 /* Something else is running or none */
1657 static noinline int btrfs_ioctl_resize(struct file *file,
1660 BTRFS_DEV_LOOKUP_ARGS(args);
1661 struct inode *inode = file_inode(file);
1662 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1666 struct btrfs_root *root = BTRFS_I(inode)->root;
1667 struct btrfs_ioctl_vol_args *vol_args;
1668 struct btrfs_trans_handle *trans;
1669 struct btrfs_device *device = NULL;
1672 char *devstr = NULL;
1677 if (!capable(CAP_SYS_ADMIN))
1680 ret = mnt_want_write_file(file);
1685 * Read the arguments before checking exclusivity to be able to
1686 * distinguish regular resize and cancel
1688 vol_args = memdup_user(arg, sizeof(*vol_args));
1689 if (IS_ERR(vol_args)) {
1690 ret = PTR_ERR(vol_args);
1693 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1694 sizestr = vol_args->name;
1695 cancel = (strcmp("cancel", sizestr) == 0);
1696 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1699 /* Exclusive operation is now claimed */
1701 devstr = strchr(sizestr, ':');
1703 sizestr = devstr + 1;
1705 devstr = vol_args->name;
1706 ret = kstrtoull(devstr, 10, &devid);
1713 btrfs_info(fs_info, "resizing devid %llu", devid);
1717 device = btrfs_find_device(fs_info->fs_devices, &args);
1719 btrfs_info(fs_info, "resizer unable to find device %llu",
1725 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1727 "resizer unable to apply on readonly device %llu",
1733 if (!strcmp(sizestr, "max"))
1734 new_size = device->bdev->bd_inode->i_size;
1736 if (sizestr[0] == '-') {
1739 } else if (sizestr[0] == '+') {
1743 new_size = memparse(sizestr, &retptr);
1744 if (*retptr != '\0' || new_size == 0) {
1750 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1755 old_size = btrfs_device_get_total_bytes(device);
1758 if (new_size > old_size) {
1762 new_size = old_size - new_size;
1763 } else if (mod > 0) {
1764 if (new_size > ULLONG_MAX - old_size) {
1768 new_size = old_size + new_size;
1771 if (new_size < SZ_256M) {
1775 if (new_size > device->bdev->bd_inode->i_size) {
1780 new_size = round_down(new_size, fs_info->sectorsize);
1782 if (new_size > old_size) {
1783 trans = btrfs_start_transaction(root, 0);
1784 if (IS_ERR(trans)) {
1785 ret = PTR_ERR(trans);
1788 ret = btrfs_grow_device(trans, device, new_size);
1789 btrfs_commit_transaction(trans);
1790 } else if (new_size < old_size) {
1791 ret = btrfs_shrink_device(device, new_size);
1792 } /* equal, nothing need to do */
1794 if (ret == 0 && new_size != old_size)
1795 btrfs_info_in_rcu(fs_info,
1796 "resize device %s (devid %llu) from %llu to %llu",
1797 rcu_str_deref(device->name), device->devid,
1798 old_size, new_size);
1800 btrfs_exclop_finish(fs_info);
1804 mnt_drop_write_file(file);
1808 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1809 struct user_namespace *mnt_userns,
1810 const char *name, unsigned long fd, int subvol,
1812 struct btrfs_qgroup_inherit *inherit)
1817 if (!S_ISDIR(file_inode(file)->i_mode))
1820 ret = mnt_want_write_file(file);
1824 namelen = strlen(name);
1825 if (strchr(name, '/')) {
1827 goto out_drop_write;
1830 if (name[0] == '.' &&
1831 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1833 goto out_drop_write;
1837 ret = btrfs_mksubvol(&file->f_path, mnt_userns, name,
1838 namelen, NULL, readonly, inherit);
1840 struct fd src = fdget(fd);
1841 struct inode *src_inode;
1844 goto out_drop_write;
1847 src_inode = file_inode(src.file);
1848 if (src_inode->i_sb != file_inode(file)->i_sb) {
1849 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1850 "Snapshot src from another FS");
1852 } else if (!inode_owner_or_capable(mnt_userns, src_inode)) {
1854 * Subvolume creation is not restricted, but snapshots
1855 * are limited to own subvolumes only
1859 ret = btrfs_mksnapshot(&file->f_path, mnt_userns,
1861 BTRFS_I(src_inode)->root,
1867 mnt_drop_write_file(file);
1872 static noinline int btrfs_ioctl_snap_create(struct file *file,
1873 void __user *arg, int subvol)
1875 struct btrfs_ioctl_vol_args *vol_args;
1878 if (!S_ISDIR(file_inode(file)->i_mode))
1881 vol_args = memdup_user(arg, sizeof(*vol_args));
1882 if (IS_ERR(vol_args))
1883 return PTR_ERR(vol_args);
1884 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1886 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
1887 vol_args->name, vol_args->fd, subvol,
1894 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1895 void __user *arg, int subvol)
1897 struct btrfs_ioctl_vol_args_v2 *vol_args;
1899 bool readonly = false;
1900 struct btrfs_qgroup_inherit *inherit = NULL;
1902 if (!S_ISDIR(file_inode(file)->i_mode))
1905 vol_args = memdup_user(arg, sizeof(*vol_args));
1906 if (IS_ERR(vol_args))
1907 return PTR_ERR(vol_args);
1908 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1910 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1915 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1917 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1920 if (vol_args->size < sizeof(*inherit) ||
1921 vol_args->size > PAGE_SIZE) {
1925 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1926 if (IS_ERR(inherit)) {
1927 ret = PTR_ERR(inherit);
1931 if (inherit->num_qgroups > PAGE_SIZE ||
1932 inherit->num_ref_copies > PAGE_SIZE ||
1933 inherit->num_excl_copies > PAGE_SIZE) {
1938 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1939 2 * inherit->num_excl_copies;
1940 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1946 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
1947 vol_args->name, vol_args->fd, subvol,
1958 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1961 struct inode *inode = file_inode(file);
1962 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1963 struct btrfs_root *root = BTRFS_I(inode)->root;
1967 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1970 down_read(&fs_info->subvol_sem);
1971 if (btrfs_root_readonly(root))
1972 flags |= BTRFS_SUBVOL_RDONLY;
1973 up_read(&fs_info->subvol_sem);
1975 if (copy_to_user(arg, &flags, sizeof(flags)))
1981 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1984 struct inode *inode = file_inode(file);
1985 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1986 struct btrfs_root *root = BTRFS_I(inode)->root;
1987 struct btrfs_trans_handle *trans;
1992 if (!inode_owner_or_capable(file_mnt_user_ns(file), inode))
1995 ret = mnt_want_write_file(file);
1999 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2001 goto out_drop_write;
2004 if (copy_from_user(&flags, arg, sizeof(flags))) {
2006 goto out_drop_write;
2009 if (flags & ~BTRFS_SUBVOL_RDONLY) {
2011 goto out_drop_write;
2014 down_write(&fs_info->subvol_sem);
2017 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2020 root_flags = btrfs_root_flags(&root->root_item);
2021 if (flags & BTRFS_SUBVOL_RDONLY) {
2022 btrfs_set_root_flags(&root->root_item,
2023 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2026 * Block RO -> RW transition if this subvolume is involved in
2029 spin_lock(&root->root_item_lock);
2030 if (root->send_in_progress == 0) {
2031 btrfs_set_root_flags(&root->root_item,
2032 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2033 spin_unlock(&root->root_item_lock);
2035 spin_unlock(&root->root_item_lock);
2037 "Attempt to set subvolume %llu read-write during send",
2038 root->root_key.objectid);
2044 trans = btrfs_start_transaction(root, 1);
2045 if (IS_ERR(trans)) {
2046 ret = PTR_ERR(trans);
2050 ret = btrfs_update_root(trans, fs_info->tree_root,
2051 &root->root_key, &root->root_item);
2053 btrfs_end_transaction(trans);
2057 ret = btrfs_commit_transaction(trans);
2061 btrfs_set_root_flags(&root->root_item, root_flags);
2063 up_write(&fs_info->subvol_sem);
2065 mnt_drop_write_file(file);
2070 static noinline int key_in_sk(struct btrfs_key *key,
2071 struct btrfs_ioctl_search_key *sk)
2073 struct btrfs_key test;
2076 test.objectid = sk->min_objectid;
2077 test.type = sk->min_type;
2078 test.offset = sk->min_offset;
2080 ret = btrfs_comp_cpu_keys(key, &test);
2084 test.objectid = sk->max_objectid;
2085 test.type = sk->max_type;
2086 test.offset = sk->max_offset;
2088 ret = btrfs_comp_cpu_keys(key, &test);
2094 static noinline int copy_to_sk(struct btrfs_path *path,
2095 struct btrfs_key *key,
2096 struct btrfs_ioctl_search_key *sk,
2099 unsigned long *sk_offset,
2103 struct extent_buffer *leaf;
2104 struct btrfs_ioctl_search_header sh;
2105 struct btrfs_key test;
2106 unsigned long item_off;
2107 unsigned long item_len;
2113 leaf = path->nodes[0];
2114 slot = path->slots[0];
2115 nritems = btrfs_header_nritems(leaf);
2117 if (btrfs_header_generation(leaf) > sk->max_transid) {
2121 found_transid = btrfs_header_generation(leaf);
2123 for (i = slot; i < nritems; i++) {
2124 item_off = btrfs_item_ptr_offset(leaf, i);
2125 item_len = btrfs_item_size_nr(leaf, i);
2127 btrfs_item_key_to_cpu(leaf, key, i);
2128 if (!key_in_sk(key, sk))
2131 if (sizeof(sh) + item_len > *buf_size) {
2138 * return one empty item back for v1, which does not
2142 *buf_size = sizeof(sh) + item_len;
2147 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2152 sh.objectid = key->objectid;
2153 sh.offset = key->offset;
2154 sh.type = key->type;
2156 sh.transid = found_transid;
2159 * Copy search result header. If we fault then loop again so we
2160 * can fault in the pages and -EFAULT there if there's a
2161 * problem. Otherwise we'll fault and then copy the buffer in
2162 * properly this next time through
2164 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2169 *sk_offset += sizeof(sh);
2172 char __user *up = ubuf + *sk_offset;
2174 * Copy the item, same behavior as above, but reset the
2175 * * sk_offset so we copy the full thing again.
2177 if (read_extent_buffer_to_user_nofault(leaf, up,
2178 item_off, item_len)) {
2180 *sk_offset -= sizeof(sh);
2184 *sk_offset += item_len;
2188 if (ret) /* -EOVERFLOW from above */
2191 if (*num_found >= sk->nr_items) {
2198 test.objectid = sk->max_objectid;
2199 test.type = sk->max_type;
2200 test.offset = sk->max_offset;
2201 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2203 else if (key->offset < (u64)-1)
2205 else if (key->type < (u8)-1) {
2208 } else if (key->objectid < (u64)-1) {
2216 * 0: all items from this leaf copied, continue with next
2217 * 1: * more items can be copied, but unused buffer is too small
2218 * * all items were found
2219 * Either way, it will stops the loop which iterates to the next
2221 * -EOVERFLOW: item was to large for buffer
2222 * -EFAULT: could not copy extent buffer back to userspace
2227 static noinline int search_ioctl(struct inode *inode,
2228 struct btrfs_ioctl_search_key *sk,
2232 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2233 struct btrfs_root *root;
2234 struct btrfs_key key;
2235 struct btrfs_path *path;
2238 unsigned long sk_offset = 0;
2240 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2241 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2245 path = btrfs_alloc_path();
2249 if (sk->tree_id == 0) {
2250 /* search the root of the inode that was passed */
2251 root = btrfs_grab_root(BTRFS_I(inode)->root);
2253 root = btrfs_get_fs_root(info, sk->tree_id, true);
2255 btrfs_free_path(path);
2256 return PTR_ERR(root);
2260 key.objectid = sk->min_objectid;
2261 key.type = sk->min_type;
2262 key.offset = sk->min_offset;
2266 if (fault_in_writeable(ubuf + sk_offset, *buf_size - sk_offset))
2269 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2275 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2276 &sk_offset, &num_found);
2277 btrfs_release_path(path);
2285 sk->nr_items = num_found;
2286 btrfs_put_root(root);
2287 btrfs_free_path(path);
2291 static noinline int btrfs_ioctl_tree_search(struct file *file,
2294 struct btrfs_ioctl_search_args __user *uargs;
2295 struct btrfs_ioctl_search_key sk;
2296 struct inode *inode;
2300 if (!capable(CAP_SYS_ADMIN))
2303 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2305 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2308 buf_size = sizeof(uargs->buf);
2310 inode = file_inode(file);
2311 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2314 * In the origin implementation an overflow is handled by returning a
2315 * search header with a len of zero, so reset ret.
2317 if (ret == -EOVERFLOW)
2320 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2325 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2328 struct btrfs_ioctl_search_args_v2 __user *uarg;
2329 struct btrfs_ioctl_search_args_v2 args;
2330 struct inode *inode;
2333 const size_t buf_limit = SZ_16M;
2335 if (!capable(CAP_SYS_ADMIN))
2338 /* copy search header and buffer size */
2339 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2340 if (copy_from_user(&args, uarg, sizeof(args)))
2343 buf_size = args.buf_size;
2345 /* limit result size to 16MB */
2346 if (buf_size > buf_limit)
2347 buf_size = buf_limit;
2349 inode = file_inode(file);
2350 ret = search_ioctl(inode, &args.key, &buf_size,
2351 (char __user *)(&uarg->buf[0]));
2352 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2354 else if (ret == -EOVERFLOW &&
2355 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2362 * Search INODE_REFs to identify path name of 'dirid' directory
2363 * in a 'tree_id' tree. and sets path name to 'name'.
2365 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2366 u64 tree_id, u64 dirid, char *name)
2368 struct btrfs_root *root;
2369 struct btrfs_key key;
2375 struct btrfs_inode_ref *iref;
2376 struct extent_buffer *l;
2377 struct btrfs_path *path;
2379 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2384 path = btrfs_alloc_path();
2388 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2390 root = btrfs_get_fs_root(info, tree_id, true);
2392 ret = PTR_ERR(root);
2397 key.objectid = dirid;
2398 key.type = BTRFS_INODE_REF_KEY;
2399 key.offset = (u64)-1;
2402 ret = btrfs_search_backwards(root, &key, path);
2411 slot = path->slots[0];
2413 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2414 len = btrfs_inode_ref_name_len(l, iref);
2416 total_len += len + 1;
2418 ret = -ENAMETOOLONG;
2423 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2425 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2428 btrfs_release_path(path);
2429 key.objectid = key.offset;
2430 key.offset = (u64)-1;
2431 dirid = key.objectid;
2433 memmove(name, ptr, total_len);
2434 name[total_len] = '\0';
2437 btrfs_put_root(root);
2438 btrfs_free_path(path);
2442 static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns,
2443 struct inode *inode,
2444 struct btrfs_ioctl_ino_lookup_user_args *args)
2446 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2447 struct super_block *sb = inode->i_sb;
2448 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2449 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2450 u64 dirid = args->dirid;
2451 unsigned long item_off;
2452 unsigned long item_len;
2453 struct btrfs_inode_ref *iref;
2454 struct btrfs_root_ref *rref;
2455 struct btrfs_root *root = NULL;
2456 struct btrfs_path *path;
2457 struct btrfs_key key, key2;
2458 struct extent_buffer *leaf;
2459 struct inode *temp_inode;
2466 path = btrfs_alloc_path();
2471 * If the bottom subvolume does not exist directly under upper_limit,
2472 * construct the path in from the bottom up.
2474 if (dirid != upper_limit.objectid) {
2475 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2477 root = btrfs_get_fs_root(fs_info, treeid, true);
2479 ret = PTR_ERR(root);
2483 key.objectid = dirid;
2484 key.type = BTRFS_INODE_REF_KEY;
2485 key.offset = (u64)-1;
2487 ret = btrfs_search_backwards(root, &key, path);
2495 leaf = path->nodes[0];
2496 slot = path->slots[0];
2498 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2499 len = btrfs_inode_ref_name_len(leaf, iref);
2501 total_len += len + 1;
2502 if (ptr < args->path) {
2503 ret = -ENAMETOOLONG;
2508 read_extent_buffer(leaf, ptr,
2509 (unsigned long)(iref + 1), len);
2511 /* Check the read+exec permission of this directory */
2512 ret = btrfs_previous_item(root, path, dirid,
2513 BTRFS_INODE_ITEM_KEY);
2516 } else if (ret > 0) {
2521 leaf = path->nodes[0];
2522 slot = path->slots[0];
2523 btrfs_item_key_to_cpu(leaf, &key2, slot);
2524 if (key2.objectid != dirid) {
2529 temp_inode = btrfs_iget(sb, key2.objectid, root);
2530 if (IS_ERR(temp_inode)) {
2531 ret = PTR_ERR(temp_inode);
2534 ret = inode_permission(mnt_userns, temp_inode,
2535 MAY_READ | MAY_EXEC);
2542 if (key.offset == upper_limit.objectid)
2544 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2549 btrfs_release_path(path);
2550 key.objectid = key.offset;
2551 key.offset = (u64)-1;
2552 dirid = key.objectid;
2555 memmove(args->path, ptr, total_len);
2556 args->path[total_len] = '\0';
2557 btrfs_put_root(root);
2559 btrfs_release_path(path);
2562 /* Get the bottom subvolume's name from ROOT_REF */
2563 key.objectid = treeid;
2564 key.type = BTRFS_ROOT_REF_KEY;
2565 key.offset = args->treeid;
2566 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2569 } else if (ret > 0) {
2574 leaf = path->nodes[0];
2575 slot = path->slots[0];
2576 btrfs_item_key_to_cpu(leaf, &key, slot);
2578 item_off = btrfs_item_ptr_offset(leaf, slot);
2579 item_len = btrfs_item_size_nr(leaf, slot);
2580 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2581 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2582 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2587 /* Copy subvolume's name */
2588 item_off += sizeof(struct btrfs_root_ref);
2589 item_len -= sizeof(struct btrfs_root_ref);
2590 read_extent_buffer(leaf, args->name, item_off, item_len);
2591 args->name[item_len] = 0;
2594 btrfs_put_root(root);
2596 btrfs_free_path(path);
2600 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2603 struct btrfs_ioctl_ino_lookup_args *args;
2604 struct inode *inode;
2607 args = memdup_user(argp, sizeof(*args));
2609 return PTR_ERR(args);
2611 inode = file_inode(file);
2614 * Unprivileged query to obtain the containing subvolume root id. The
2615 * path is reset so it's consistent with btrfs_search_path_in_tree.
2617 if (args->treeid == 0)
2618 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2620 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2625 if (!capable(CAP_SYS_ADMIN)) {
2630 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2631 args->treeid, args->objectid,
2635 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2643 * Version of ino_lookup ioctl (unprivileged)
2645 * The main differences from ino_lookup ioctl are:
2647 * 1. Read + Exec permission will be checked using inode_permission() during
2648 * path construction. -EACCES will be returned in case of failure.
2649 * 2. Path construction will be stopped at the inode number which corresponds
2650 * to the fd with which this ioctl is called. If constructed path does not
2651 * exist under fd's inode, -EACCES will be returned.
2652 * 3. The name of bottom subvolume is also searched and filled.
2654 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2656 struct btrfs_ioctl_ino_lookup_user_args *args;
2657 struct inode *inode;
2660 args = memdup_user(argp, sizeof(*args));
2662 return PTR_ERR(args);
2664 inode = file_inode(file);
2666 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2667 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2669 * The subvolume does not exist under fd with which this is
2676 ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args);
2678 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2685 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2686 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2688 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2689 struct btrfs_fs_info *fs_info;
2690 struct btrfs_root *root;
2691 struct btrfs_path *path;
2692 struct btrfs_key key;
2693 struct btrfs_root_item *root_item;
2694 struct btrfs_root_ref *rref;
2695 struct extent_buffer *leaf;
2696 unsigned long item_off;
2697 unsigned long item_len;
2698 struct inode *inode;
2702 path = btrfs_alloc_path();
2706 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2708 btrfs_free_path(path);
2712 inode = file_inode(file);
2713 fs_info = BTRFS_I(inode)->root->fs_info;
2715 /* Get root_item of inode's subvolume */
2716 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2717 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2719 ret = PTR_ERR(root);
2722 root_item = &root->root_item;
2724 subvol_info->treeid = key.objectid;
2726 subvol_info->generation = btrfs_root_generation(root_item);
2727 subvol_info->flags = btrfs_root_flags(root_item);
2729 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2730 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2732 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2735 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2736 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2737 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2739 subvol_info->otransid = btrfs_root_otransid(root_item);
2740 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2741 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2743 subvol_info->stransid = btrfs_root_stransid(root_item);
2744 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2745 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2747 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2748 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2749 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2751 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2752 /* Search root tree for ROOT_BACKREF of this subvolume */
2753 key.type = BTRFS_ROOT_BACKREF_KEY;
2755 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2758 } else if (path->slots[0] >=
2759 btrfs_header_nritems(path->nodes[0])) {
2760 ret = btrfs_next_leaf(fs_info->tree_root, path);
2763 } else if (ret > 0) {
2769 leaf = path->nodes[0];
2770 slot = path->slots[0];
2771 btrfs_item_key_to_cpu(leaf, &key, slot);
2772 if (key.objectid == subvol_info->treeid &&
2773 key.type == BTRFS_ROOT_BACKREF_KEY) {
2774 subvol_info->parent_id = key.offset;
2776 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2777 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2779 item_off = btrfs_item_ptr_offset(leaf, slot)
2780 + sizeof(struct btrfs_root_ref);
2781 item_len = btrfs_item_size_nr(leaf, slot)
2782 - sizeof(struct btrfs_root_ref);
2783 read_extent_buffer(leaf, subvol_info->name,
2784 item_off, item_len);
2791 btrfs_free_path(path);
2793 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2797 btrfs_put_root(root);
2799 btrfs_free_path(path);
2805 * Return ROOT_REF information of the subvolume containing this inode
2806 * except the subvolume name.
2808 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2810 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2811 struct btrfs_root_ref *rref;
2812 struct btrfs_root *root;
2813 struct btrfs_path *path;
2814 struct btrfs_key key;
2815 struct extent_buffer *leaf;
2816 struct inode *inode;
2822 path = btrfs_alloc_path();
2826 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2827 if (IS_ERR(rootrefs)) {
2828 btrfs_free_path(path);
2829 return PTR_ERR(rootrefs);
2832 inode = file_inode(file);
2833 root = BTRFS_I(inode)->root->fs_info->tree_root;
2834 objectid = BTRFS_I(inode)->root->root_key.objectid;
2836 key.objectid = objectid;
2837 key.type = BTRFS_ROOT_REF_KEY;
2838 key.offset = rootrefs->min_treeid;
2841 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2844 } else if (path->slots[0] >=
2845 btrfs_header_nritems(path->nodes[0])) {
2846 ret = btrfs_next_leaf(root, path);
2849 } else if (ret > 0) {
2855 leaf = path->nodes[0];
2856 slot = path->slots[0];
2858 btrfs_item_key_to_cpu(leaf, &key, slot);
2859 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2864 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2869 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2870 rootrefs->rootref[found].treeid = key.offset;
2871 rootrefs->rootref[found].dirid =
2872 btrfs_root_ref_dirid(leaf, rref);
2875 ret = btrfs_next_item(root, path);
2878 } else if (ret > 0) {
2885 btrfs_free_path(path);
2887 if (!ret || ret == -EOVERFLOW) {
2888 rootrefs->num_items = found;
2889 /* update min_treeid for next search */
2891 rootrefs->min_treeid =
2892 rootrefs->rootref[found - 1].treeid + 1;
2893 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2902 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2906 struct dentry *parent = file->f_path.dentry;
2907 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2908 struct dentry *dentry;
2909 struct inode *dir = d_inode(parent);
2910 struct inode *inode;
2911 struct btrfs_root *root = BTRFS_I(dir)->root;
2912 struct btrfs_root *dest = NULL;
2913 struct btrfs_ioctl_vol_args *vol_args = NULL;
2914 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2915 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
2916 char *subvol_name, *subvol_name_ptr = NULL;
2919 bool destroy_parent = false;
2922 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2923 if (IS_ERR(vol_args2))
2924 return PTR_ERR(vol_args2);
2926 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2932 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2933 * name, same as v1 currently does.
2935 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2936 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2937 subvol_name = vol_args2->name;
2939 err = mnt_want_write_file(file);
2943 struct inode *old_dir;
2945 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2950 err = mnt_want_write_file(file);
2954 dentry = btrfs_get_dentry(fs_info->sb,
2955 BTRFS_FIRST_FREE_OBJECTID,
2956 vol_args2->subvolid, 0, 0);
2957 if (IS_ERR(dentry)) {
2958 err = PTR_ERR(dentry);
2959 goto out_drop_write;
2963 * Change the default parent since the subvolume being
2964 * deleted can be outside of the current mount point.
2966 parent = btrfs_get_parent(dentry);
2969 * At this point dentry->d_name can point to '/' if the
2970 * subvolume we want to destroy is outsite of the
2971 * current mount point, so we need to release the
2972 * current dentry and execute the lookup to return a new
2973 * one with ->d_name pointing to the
2974 * <mount point>/subvol_name.
2977 if (IS_ERR(parent)) {
2978 err = PTR_ERR(parent);
2979 goto out_drop_write;
2982 dir = d_inode(parent);
2985 * If v2 was used with SPEC_BY_ID, a new parent was
2986 * allocated since the subvolume can be outside of the
2987 * current mount point. Later on we need to release this
2988 * new parent dentry.
2990 destroy_parent = true;
2993 * On idmapped mounts, deletion via subvolid is
2994 * restricted to subvolumes that are immediate
2995 * ancestors of the inode referenced by the file
2996 * descriptor in the ioctl. Otherwise the idmapping
2997 * could potentially be abused to delete subvolumes
2998 * anywhere in the filesystem the user wouldn't be able
2999 * to delete without an idmapped mount.
3001 if (old_dir != dir && mnt_userns != &init_user_ns) {
3006 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
3007 fs_info, vol_args2->subvolid);
3008 if (IS_ERR(subvol_name_ptr)) {
3009 err = PTR_ERR(subvol_name_ptr);
3012 /* subvol_name_ptr is already nul terminated */
3013 subvol_name = (char *)kbasename(subvol_name_ptr);
3016 vol_args = memdup_user(arg, sizeof(*vol_args));
3017 if (IS_ERR(vol_args))
3018 return PTR_ERR(vol_args);
3020 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
3021 subvol_name = vol_args->name;
3023 err = mnt_want_write_file(file);
3028 subvol_namelen = strlen(subvol_name);
3030 if (strchr(subvol_name, '/') ||
3031 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3033 goto free_subvol_name;
3036 if (!S_ISDIR(dir->i_mode)) {
3038 goto free_subvol_name;
3041 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3043 goto free_subvol_name;
3044 dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen);
3045 if (IS_ERR(dentry)) {
3046 err = PTR_ERR(dentry);
3047 goto out_unlock_dir;
3050 if (d_really_is_negative(dentry)) {
3055 inode = d_inode(dentry);
3056 dest = BTRFS_I(inode)->root;
3057 if (!capable(CAP_SYS_ADMIN)) {
3059 * Regular user. Only allow this with a special mount
3060 * option, when the user has write+exec access to the
3061 * subvol root, and when rmdir(2) would have been
3064 * Note that this is _not_ check that the subvol is
3065 * empty or doesn't contain data that we wouldn't
3066 * otherwise be able to delete.
3068 * Users who want to delete empty subvols should try
3072 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3076 * Do not allow deletion if the parent dir is the same
3077 * as the dir to be deleted. That means the ioctl
3078 * must be called on the dentry referencing the root
3079 * of the subvol, not a random directory contained
3086 err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC);
3091 /* check if subvolume may be deleted by a user */
3092 err = btrfs_may_delete(mnt_userns, dir, dentry, 1);
3096 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3101 btrfs_inode_lock(inode, 0);
3102 err = btrfs_delete_subvolume(dir, dentry);
3103 btrfs_inode_unlock(inode, 0);
3105 d_delete_notify(dir, dentry);
3110 btrfs_inode_unlock(dir, 0);
3112 kfree(subvol_name_ptr);
3117 mnt_drop_write_file(file);
3124 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3126 struct inode *inode = file_inode(file);
3127 struct btrfs_root *root = BTRFS_I(inode)->root;
3128 struct btrfs_ioctl_defrag_range_args range = {0};
3131 ret = mnt_want_write_file(file);
3135 if (btrfs_root_readonly(root)) {
3140 /* Subpage defrag will be supported in later commits */
3141 if (root->fs_info->sectorsize < PAGE_SIZE) {
3146 switch (inode->i_mode & S_IFMT) {
3148 if (!capable(CAP_SYS_ADMIN)) {
3152 ret = btrfs_defrag_root(root);
3156 * Note that this does not check the file descriptor for write
3157 * access. This prevents defragmenting executables that are
3158 * running and allows defrag on files open in read-only mode.
3160 if (!capable(CAP_SYS_ADMIN) &&
3161 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
3167 if (copy_from_user(&range, argp, sizeof(range))) {
3171 /* compression requires us to start the IO */
3172 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3173 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
3174 range.extent_thresh = (u32)-1;
3177 /* the rest are all set to zero by kzalloc */
3178 range.len = (u64)-1;
3180 ret = btrfs_defrag_file(file_inode(file), file,
3181 &range, BTRFS_OLDEST_GENERATION, 0);
3189 mnt_drop_write_file(file);
3193 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3195 struct btrfs_ioctl_vol_args *vol_args;
3198 if (!capable(CAP_SYS_ADMIN))
3201 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD))
3202 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3204 vol_args = memdup_user(arg, sizeof(*vol_args));
3205 if (IS_ERR(vol_args)) {
3206 ret = PTR_ERR(vol_args);
3210 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3211 ret = btrfs_init_new_device(fs_info, vol_args->name);
3214 btrfs_info(fs_info, "disk added %s", vol_args->name);
3218 btrfs_exclop_finish(fs_info);
3222 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3224 BTRFS_DEV_LOOKUP_ARGS(args);
3225 struct inode *inode = file_inode(file);
3226 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3227 struct btrfs_ioctl_vol_args_v2 *vol_args;
3228 struct block_device *bdev = NULL;
3231 bool cancel = false;
3233 if (!capable(CAP_SYS_ADMIN))
3236 vol_args = memdup_user(arg, sizeof(*vol_args));
3237 if (IS_ERR(vol_args))
3238 return PTR_ERR(vol_args);
3240 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3245 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3246 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3247 args.devid = vol_args->devid;
3248 } else if (!strcmp("cancel", vol_args->name)) {
3251 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
3256 ret = mnt_want_write_file(file);
3260 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3265 /* Exclusive operation is now claimed */
3266 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
3268 btrfs_exclop_finish(fs_info);
3271 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3272 btrfs_info(fs_info, "device deleted: id %llu",
3275 btrfs_info(fs_info, "device deleted: %s",
3279 mnt_drop_write_file(file);
3281 blkdev_put(bdev, mode);
3283 btrfs_put_dev_args_from_path(&args);
3288 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3290 BTRFS_DEV_LOOKUP_ARGS(args);
3291 struct inode *inode = file_inode(file);
3292 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3293 struct btrfs_ioctl_vol_args *vol_args;
3294 struct block_device *bdev = NULL;
3297 bool cancel = false;
3299 if (!capable(CAP_SYS_ADMIN))
3302 vol_args = memdup_user(arg, sizeof(*vol_args));
3303 if (IS_ERR(vol_args))
3304 return PTR_ERR(vol_args);
3306 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3307 if (!strcmp("cancel", vol_args->name)) {
3310 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
3315 ret = mnt_want_write_file(file);
3319 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3322 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
3324 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3325 btrfs_exclop_finish(fs_info);
3328 mnt_drop_write_file(file);
3330 blkdev_put(bdev, mode);
3332 btrfs_put_dev_args_from_path(&args);
3337 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3340 struct btrfs_ioctl_fs_info_args *fi_args;
3341 struct btrfs_device *device;
3342 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3346 fi_args = memdup_user(arg, sizeof(*fi_args));
3347 if (IS_ERR(fi_args))
3348 return PTR_ERR(fi_args);
3350 flags_in = fi_args->flags;
3351 memset(fi_args, 0, sizeof(*fi_args));
3354 fi_args->num_devices = fs_devices->num_devices;
3356 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3357 if (device->devid > fi_args->max_id)
3358 fi_args->max_id = device->devid;
3362 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3363 fi_args->nodesize = fs_info->nodesize;
3364 fi_args->sectorsize = fs_info->sectorsize;
3365 fi_args->clone_alignment = fs_info->sectorsize;
3367 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3368 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3369 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3370 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3373 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3374 fi_args->generation = fs_info->generation;
3375 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3378 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3379 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3380 sizeof(fi_args->metadata_uuid));
3381 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3384 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3391 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3394 BTRFS_DEV_LOOKUP_ARGS(args);
3395 struct btrfs_ioctl_dev_info_args *di_args;
3396 struct btrfs_device *dev;
3399 di_args = memdup_user(arg, sizeof(*di_args));
3400 if (IS_ERR(di_args))
3401 return PTR_ERR(di_args);
3403 args.devid = di_args->devid;
3404 if (!btrfs_is_empty_uuid(di_args->uuid))
3405 args.uuid = di_args->uuid;
3408 dev = btrfs_find_device(fs_info->fs_devices, &args);
3414 di_args->devid = dev->devid;
3415 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3416 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3417 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3419 strncpy(di_args->path, rcu_str_deref(dev->name),
3420 sizeof(di_args->path) - 1);
3421 di_args->path[sizeof(di_args->path) - 1] = 0;
3423 di_args->path[0] = '\0';
3428 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3435 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3437 struct inode *inode = file_inode(file);
3438 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3439 struct btrfs_root *root = BTRFS_I(inode)->root;
3440 struct btrfs_root *new_root;
3441 struct btrfs_dir_item *di;
3442 struct btrfs_trans_handle *trans;
3443 struct btrfs_path *path = NULL;
3444 struct btrfs_disk_key disk_key;
3449 if (!capable(CAP_SYS_ADMIN))
3452 ret = mnt_want_write_file(file);
3456 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3462 objectid = BTRFS_FS_TREE_OBJECTID;
3464 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3465 if (IS_ERR(new_root)) {
3466 ret = PTR_ERR(new_root);
3469 if (!is_fstree(new_root->root_key.objectid)) {
3474 path = btrfs_alloc_path();
3480 trans = btrfs_start_transaction(root, 1);
3481 if (IS_ERR(trans)) {
3482 ret = PTR_ERR(trans);
3486 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3487 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3488 dir_id, "default", 7, 1);
3489 if (IS_ERR_OR_NULL(di)) {
3490 btrfs_release_path(path);
3491 btrfs_end_transaction(trans);
3493 "Umm, you don't have the default diritem, this isn't going to work");
3498 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3499 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3500 btrfs_mark_buffer_dirty(path->nodes[0]);
3501 btrfs_release_path(path);
3503 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3504 btrfs_end_transaction(trans);
3506 btrfs_put_root(new_root);
3507 btrfs_free_path(path);
3509 mnt_drop_write_file(file);
3513 static void get_block_group_info(struct list_head *groups_list,
3514 struct btrfs_ioctl_space_info *space)
3516 struct btrfs_block_group *block_group;
3518 space->total_bytes = 0;
3519 space->used_bytes = 0;
3521 list_for_each_entry(block_group, groups_list, list) {
3522 space->flags = block_group->flags;
3523 space->total_bytes += block_group->length;
3524 space->used_bytes += block_group->used;
3528 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3531 struct btrfs_ioctl_space_args space_args;
3532 struct btrfs_ioctl_space_info space;
3533 struct btrfs_ioctl_space_info *dest;
3534 struct btrfs_ioctl_space_info *dest_orig;
3535 struct btrfs_ioctl_space_info __user *user_dest;
3536 struct btrfs_space_info *info;
3537 static const u64 types[] = {
3538 BTRFS_BLOCK_GROUP_DATA,
3539 BTRFS_BLOCK_GROUP_SYSTEM,
3540 BTRFS_BLOCK_GROUP_METADATA,
3541 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3549 if (copy_from_user(&space_args,
3550 (struct btrfs_ioctl_space_args __user *)arg,
3551 sizeof(space_args)))
3554 for (i = 0; i < num_types; i++) {
3555 struct btrfs_space_info *tmp;
3558 list_for_each_entry(tmp, &fs_info->space_info, list) {
3559 if (tmp->flags == types[i]) {
3568 down_read(&info->groups_sem);
3569 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3570 if (!list_empty(&info->block_groups[c]))
3573 up_read(&info->groups_sem);
3577 * Global block reserve, exported as a space_info
3581 /* space_slots == 0 means they are asking for a count */
3582 if (space_args.space_slots == 0) {
3583 space_args.total_spaces = slot_count;
3587 slot_count = min_t(u64, space_args.space_slots, slot_count);
3589 alloc_size = sizeof(*dest) * slot_count;
3591 /* we generally have at most 6 or so space infos, one for each raid
3592 * level. So, a whole page should be more than enough for everyone
3594 if (alloc_size > PAGE_SIZE)
3597 space_args.total_spaces = 0;
3598 dest = kmalloc(alloc_size, GFP_KERNEL);
3603 /* now we have a buffer to copy into */
3604 for (i = 0; i < num_types; i++) {
3605 struct btrfs_space_info *tmp;
3611 list_for_each_entry(tmp, &fs_info->space_info, list) {
3612 if (tmp->flags == types[i]) {
3620 down_read(&info->groups_sem);
3621 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3622 if (!list_empty(&info->block_groups[c])) {
3623 get_block_group_info(&info->block_groups[c],
3625 memcpy(dest, &space, sizeof(space));
3627 space_args.total_spaces++;
3633 up_read(&info->groups_sem);
3637 * Add global block reserve
3640 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3642 spin_lock(&block_rsv->lock);
3643 space.total_bytes = block_rsv->size;
3644 space.used_bytes = block_rsv->size - block_rsv->reserved;
3645 spin_unlock(&block_rsv->lock);
3646 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3647 memcpy(dest, &space, sizeof(space));
3648 space_args.total_spaces++;
3651 user_dest = (struct btrfs_ioctl_space_info __user *)
3652 (arg + sizeof(struct btrfs_ioctl_space_args));
3654 if (copy_to_user(user_dest, dest_orig, alloc_size))
3659 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3665 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3668 struct btrfs_trans_handle *trans;
3672 trans = btrfs_attach_transaction_barrier(root);
3673 if (IS_ERR(trans)) {
3674 if (PTR_ERR(trans) != -ENOENT)
3675 return PTR_ERR(trans);
3677 /* No running transaction, don't bother */
3678 transid = root->fs_info->last_trans_committed;
3681 transid = trans->transid;
3682 ret = btrfs_commit_transaction_async(trans);
3684 btrfs_end_transaction(trans);
3689 if (copy_to_user(argp, &transid, sizeof(transid)))
3694 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3700 if (copy_from_user(&transid, argp, sizeof(transid)))
3703 transid = 0; /* current trans */
3705 return btrfs_wait_for_commit(fs_info, transid);
3708 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3710 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3711 struct btrfs_ioctl_scrub_args *sa;
3714 if (!capable(CAP_SYS_ADMIN))
3717 sa = memdup_user(arg, sizeof(*sa));
3721 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3722 ret = mnt_want_write_file(file);
3727 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3728 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3732 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3733 * error. This is important as it allows user space to know how much
3734 * progress scrub has done. For example, if scrub is canceled we get
3735 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3736 * space. Later user space can inspect the progress from the structure
3737 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3738 * previously (btrfs-progs does this).
3739 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3740 * then return -EFAULT to signal the structure was not copied or it may
3741 * be corrupt and unreliable due to a partial copy.
3743 if (copy_to_user(arg, sa, sizeof(*sa)))
3746 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3747 mnt_drop_write_file(file);
3753 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3755 if (!capable(CAP_SYS_ADMIN))
3758 return btrfs_scrub_cancel(fs_info);
3761 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3764 struct btrfs_ioctl_scrub_args *sa;
3767 if (!capable(CAP_SYS_ADMIN))
3770 sa = memdup_user(arg, sizeof(*sa));
3774 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3776 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3783 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3786 struct btrfs_ioctl_get_dev_stats *sa;
3789 sa = memdup_user(arg, sizeof(*sa));
3793 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3798 ret = btrfs_get_dev_stats(fs_info, sa);
3800 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3807 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3810 struct btrfs_ioctl_dev_replace_args *p;
3813 if (!capable(CAP_SYS_ADMIN))
3816 p = memdup_user(arg, sizeof(*p));
3821 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3822 if (sb_rdonly(fs_info->sb)) {
3826 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3827 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3829 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3830 btrfs_exclop_finish(fs_info);
3833 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3834 btrfs_dev_replace_status(fs_info, p);
3837 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3838 p->result = btrfs_dev_replace_cancel(fs_info);
3846 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3853 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3859 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3860 struct inode_fs_paths *ipath = NULL;
3861 struct btrfs_path *path;
3863 if (!capable(CAP_DAC_READ_SEARCH))
3866 path = btrfs_alloc_path();
3872 ipa = memdup_user(arg, sizeof(*ipa));
3879 size = min_t(u32, ipa->size, 4096);
3880 ipath = init_ipath(size, root, path);
3881 if (IS_ERR(ipath)) {
3882 ret = PTR_ERR(ipath);
3887 ret = paths_from_inode(ipa->inum, ipath);
3891 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3892 rel_ptr = ipath->fspath->val[i] -
3893 (u64)(unsigned long)ipath->fspath->val;
3894 ipath->fspath->val[i] = rel_ptr;
3897 btrfs_free_path(path);
3899 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3900 ipath->fspath, size);
3907 btrfs_free_path(path);
3914 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3915 void __user *arg, int version)
3919 struct btrfs_ioctl_logical_ino_args *loi;
3920 struct btrfs_data_container *inodes = NULL;
3921 struct btrfs_path *path = NULL;
3924 if (!capable(CAP_SYS_ADMIN))
3927 loi = memdup_user(arg, sizeof(*loi));
3929 return PTR_ERR(loi);
3932 ignore_offset = false;
3933 size = min_t(u32, loi->size, SZ_64K);
3935 /* All reserved bits must be 0 for now */
3936 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3940 /* Only accept flags we have defined so far */
3941 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3945 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3946 size = min_t(u32, loi->size, SZ_16M);
3949 path = btrfs_alloc_path();
3955 inodes = init_data_container(size);
3956 if (IS_ERR(inodes)) {
3957 ret = PTR_ERR(inodes);
3962 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3963 inodes, ignore_offset);
3969 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3975 btrfs_free_path(path);
3983 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3984 struct btrfs_ioctl_balance_args *bargs)
3986 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3988 bargs->flags = bctl->flags;
3990 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3991 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3992 if (atomic_read(&fs_info->balance_pause_req))
3993 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3994 if (atomic_read(&fs_info->balance_cancel_req))
3995 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3997 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3998 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3999 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4001 spin_lock(&fs_info->balance_lock);
4002 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4003 spin_unlock(&fs_info->balance_lock);
4006 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4008 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4009 struct btrfs_fs_info *fs_info = root->fs_info;
4010 struct btrfs_ioctl_balance_args *bargs;
4011 struct btrfs_balance_control *bctl;
4012 bool need_unlock; /* for mut. excl. ops lock */
4015 if (!capable(CAP_SYS_ADMIN))
4018 ret = mnt_want_write_file(file);
4023 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4024 mutex_lock(&fs_info->balance_mutex);
4030 * mut. excl. ops lock is locked. Three possibilities:
4031 * (1) some other op is running
4032 * (2) balance is running
4033 * (3) balance is paused -- special case (think resume)
4035 mutex_lock(&fs_info->balance_mutex);
4036 if (fs_info->balance_ctl) {
4037 /* this is either (2) or (3) */
4038 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4039 mutex_unlock(&fs_info->balance_mutex);
4041 * Lock released to allow other waiters to continue,
4042 * we'll reexamine the status again.
4044 mutex_lock(&fs_info->balance_mutex);
4046 if (fs_info->balance_ctl &&
4047 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4049 need_unlock = false;
4053 mutex_unlock(&fs_info->balance_mutex);
4057 mutex_unlock(&fs_info->balance_mutex);
4063 mutex_unlock(&fs_info->balance_mutex);
4064 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4071 bargs = memdup_user(arg, sizeof(*bargs));
4072 if (IS_ERR(bargs)) {
4073 ret = PTR_ERR(bargs);
4077 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4078 if (!fs_info->balance_ctl) {
4083 bctl = fs_info->balance_ctl;
4084 spin_lock(&fs_info->balance_lock);
4085 bctl->flags |= BTRFS_BALANCE_RESUME;
4086 spin_unlock(&fs_info->balance_lock);
4094 if (fs_info->balance_ctl) {
4099 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4106 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4107 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4108 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4110 bctl->flags = bargs->flags;
4112 /* balance everything - no filters */
4113 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4116 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4123 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4124 * bctl is freed in reset_balance_state, or, if restriper was paused
4125 * all the way until unmount, in free_fs_info. The flag should be
4126 * cleared after reset_balance_state.
4128 need_unlock = false;
4130 ret = btrfs_balance(fs_info, bctl, bargs);
4133 if ((ret == 0 || ret == -ECANCELED) && arg) {
4134 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4143 mutex_unlock(&fs_info->balance_mutex);
4145 btrfs_exclop_finish(fs_info);
4147 mnt_drop_write_file(file);
4151 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4153 if (!capable(CAP_SYS_ADMIN))
4157 case BTRFS_BALANCE_CTL_PAUSE:
4158 return btrfs_pause_balance(fs_info);
4159 case BTRFS_BALANCE_CTL_CANCEL:
4160 return btrfs_cancel_balance(fs_info);
4166 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4169 struct btrfs_ioctl_balance_args *bargs;
4172 if (!capable(CAP_SYS_ADMIN))
4175 mutex_lock(&fs_info->balance_mutex);
4176 if (!fs_info->balance_ctl) {
4181 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4187 btrfs_update_ioctl_balance_args(fs_info, bargs);
4189 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4194 mutex_unlock(&fs_info->balance_mutex);
4198 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4200 struct inode *inode = file_inode(file);
4201 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4202 struct btrfs_ioctl_quota_ctl_args *sa;
4205 if (!capable(CAP_SYS_ADMIN))
4208 ret = mnt_want_write_file(file);
4212 sa = memdup_user(arg, sizeof(*sa));
4218 down_write(&fs_info->subvol_sem);
4221 case BTRFS_QUOTA_CTL_ENABLE:
4222 ret = btrfs_quota_enable(fs_info);
4224 case BTRFS_QUOTA_CTL_DISABLE:
4225 ret = btrfs_quota_disable(fs_info);
4233 up_write(&fs_info->subvol_sem);
4235 mnt_drop_write_file(file);
4239 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4241 struct inode *inode = file_inode(file);
4242 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4243 struct btrfs_root *root = BTRFS_I(inode)->root;
4244 struct btrfs_ioctl_qgroup_assign_args *sa;
4245 struct btrfs_trans_handle *trans;
4249 if (!capable(CAP_SYS_ADMIN))
4252 ret = mnt_want_write_file(file);
4256 sa = memdup_user(arg, sizeof(*sa));
4262 trans = btrfs_join_transaction(root);
4263 if (IS_ERR(trans)) {
4264 ret = PTR_ERR(trans);
4269 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4271 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4274 /* update qgroup status and info */
4275 err = btrfs_run_qgroups(trans);
4277 btrfs_handle_fs_error(fs_info, err,
4278 "failed to update qgroup status and info");
4279 err = btrfs_end_transaction(trans);
4286 mnt_drop_write_file(file);
4290 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4292 struct inode *inode = file_inode(file);
4293 struct btrfs_root *root = BTRFS_I(inode)->root;
4294 struct btrfs_ioctl_qgroup_create_args *sa;
4295 struct btrfs_trans_handle *trans;
4299 if (!capable(CAP_SYS_ADMIN))
4302 ret = mnt_want_write_file(file);
4306 sa = memdup_user(arg, sizeof(*sa));
4312 if (!sa->qgroupid) {
4317 trans = btrfs_join_transaction(root);
4318 if (IS_ERR(trans)) {
4319 ret = PTR_ERR(trans);
4324 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4326 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4329 err = btrfs_end_transaction(trans);
4336 mnt_drop_write_file(file);
4340 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4342 struct inode *inode = file_inode(file);
4343 struct btrfs_root *root = BTRFS_I(inode)->root;
4344 struct btrfs_ioctl_qgroup_limit_args *sa;
4345 struct btrfs_trans_handle *trans;
4350 if (!capable(CAP_SYS_ADMIN))
4353 ret = mnt_want_write_file(file);
4357 sa = memdup_user(arg, sizeof(*sa));
4363 trans = btrfs_join_transaction(root);
4364 if (IS_ERR(trans)) {
4365 ret = PTR_ERR(trans);
4369 qgroupid = sa->qgroupid;
4371 /* take the current subvol as qgroup */
4372 qgroupid = root->root_key.objectid;
4375 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4377 err = btrfs_end_transaction(trans);
4384 mnt_drop_write_file(file);
4388 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4390 struct inode *inode = file_inode(file);
4391 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4392 struct btrfs_ioctl_quota_rescan_args *qsa;
4395 if (!capable(CAP_SYS_ADMIN))
4398 ret = mnt_want_write_file(file);
4402 qsa = memdup_user(arg, sizeof(*qsa));
4413 ret = btrfs_qgroup_rescan(fs_info);
4418 mnt_drop_write_file(file);
4422 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4425 struct btrfs_ioctl_quota_rescan_args qsa = {0};
4428 if (!capable(CAP_SYS_ADMIN))
4431 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4433 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
4436 if (copy_to_user(arg, &qsa, sizeof(qsa)))
4442 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4445 if (!capable(CAP_SYS_ADMIN))
4448 return btrfs_qgroup_wait_for_completion(fs_info, true);
4451 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4452 struct user_namespace *mnt_userns,
4453 struct btrfs_ioctl_received_subvol_args *sa)
4455 struct inode *inode = file_inode(file);
4456 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4457 struct btrfs_root *root = BTRFS_I(inode)->root;
4458 struct btrfs_root_item *root_item = &root->root_item;
4459 struct btrfs_trans_handle *trans;
4460 struct timespec64 ct = current_time(inode);
4462 int received_uuid_changed;
4464 if (!inode_owner_or_capable(mnt_userns, inode))
4467 ret = mnt_want_write_file(file);
4471 down_write(&fs_info->subvol_sem);
4473 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4478 if (btrfs_root_readonly(root)) {
4485 * 2 - uuid items (received uuid + subvol uuid)
4487 trans = btrfs_start_transaction(root, 3);
4488 if (IS_ERR(trans)) {
4489 ret = PTR_ERR(trans);
4494 sa->rtransid = trans->transid;
4495 sa->rtime.sec = ct.tv_sec;
4496 sa->rtime.nsec = ct.tv_nsec;
4498 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4500 if (received_uuid_changed &&
4501 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4502 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4503 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4504 root->root_key.objectid);
4505 if (ret && ret != -ENOENT) {
4506 btrfs_abort_transaction(trans, ret);
4507 btrfs_end_transaction(trans);
4511 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4512 btrfs_set_root_stransid(root_item, sa->stransid);
4513 btrfs_set_root_rtransid(root_item, sa->rtransid);
4514 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4515 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4516 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4517 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4519 ret = btrfs_update_root(trans, fs_info->tree_root,
4520 &root->root_key, &root->root_item);
4522 btrfs_end_transaction(trans);
4525 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4526 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4527 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4528 root->root_key.objectid);
4529 if (ret < 0 && ret != -EEXIST) {
4530 btrfs_abort_transaction(trans, ret);
4531 btrfs_end_transaction(trans);
4535 ret = btrfs_commit_transaction(trans);
4537 up_write(&fs_info->subvol_sem);
4538 mnt_drop_write_file(file);
4543 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4546 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4547 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4550 args32 = memdup_user(arg, sizeof(*args32));
4552 return PTR_ERR(args32);
4554 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4560 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4561 args64->stransid = args32->stransid;
4562 args64->rtransid = args32->rtransid;
4563 args64->stime.sec = args32->stime.sec;
4564 args64->stime.nsec = args32->stime.nsec;
4565 args64->rtime.sec = args32->rtime.sec;
4566 args64->rtime.nsec = args32->rtime.nsec;
4567 args64->flags = args32->flags;
4569 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64);
4573 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4574 args32->stransid = args64->stransid;
4575 args32->rtransid = args64->rtransid;
4576 args32->stime.sec = args64->stime.sec;
4577 args32->stime.nsec = args64->stime.nsec;
4578 args32->rtime.sec = args64->rtime.sec;
4579 args32->rtime.nsec = args64->rtime.nsec;
4580 args32->flags = args64->flags;
4582 ret = copy_to_user(arg, args32, sizeof(*args32));
4593 static long btrfs_ioctl_set_received_subvol(struct file *file,
4596 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4599 sa = memdup_user(arg, sizeof(*sa));
4603 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa);
4608 ret = copy_to_user(arg, sa, sizeof(*sa));
4617 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4622 char label[BTRFS_LABEL_SIZE];
4624 spin_lock(&fs_info->super_lock);
4625 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4626 spin_unlock(&fs_info->super_lock);
4628 len = strnlen(label, BTRFS_LABEL_SIZE);
4630 if (len == BTRFS_LABEL_SIZE) {
4632 "label is too long, return the first %zu bytes",
4636 ret = copy_to_user(arg, label, len);
4638 return ret ? -EFAULT : 0;
4641 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4643 struct inode *inode = file_inode(file);
4644 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4645 struct btrfs_root *root = BTRFS_I(inode)->root;
4646 struct btrfs_super_block *super_block = fs_info->super_copy;
4647 struct btrfs_trans_handle *trans;
4648 char label[BTRFS_LABEL_SIZE];
4651 if (!capable(CAP_SYS_ADMIN))
4654 if (copy_from_user(label, arg, sizeof(label)))
4657 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4659 "unable to set label with more than %d bytes",
4660 BTRFS_LABEL_SIZE - 1);
4664 ret = mnt_want_write_file(file);
4668 trans = btrfs_start_transaction(root, 0);
4669 if (IS_ERR(trans)) {
4670 ret = PTR_ERR(trans);
4674 spin_lock(&fs_info->super_lock);
4675 strcpy(super_block->label, label);
4676 spin_unlock(&fs_info->super_lock);
4677 ret = btrfs_commit_transaction(trans);
4680 mnt_drop_write_file(file);
4684 #define INIT_FEATURE_FLAGS(suffix) \
4685 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4686 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4687 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4689 int btrfs_ioctl_get_supported_features(void __user *arg)
4691 static const struct btrfs_ioctl_feature_flags features[3] = {
4692 INIT_FEATURE_FLAGS(SUPP),
4693 INIT_FEATURE_FLAGS(SAFE_SET),
4694 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4697 if (copy_to_user(arg, &features, sizeof(features)))
4703 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4706 struct btrfs_super_block *super_block = fs_info->super_copy;
4707 struct btrfs_ioctl_feature_flags features;
4709 features.compat_flags = btrfs_super_compat_flags(super_block);
4710 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4711 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4713 if (copy_to_user(arg, &features, sizeof(features)))
4719 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4720 enum btrfs_feature_set set,
4721 u64 change_mask, u64 flags, u64 supported_flags,
4722 u64 safe_set, u64 safe_clear)
4724 const char *type = btrfs_feature_set_name(set);
4726 u64 disallowed, unsupported;
4727 u64 set_mask = flags & change_mask;
4728 u64 clear_mask = ~flags & change_mask;
4730 unsupported = set_mask & ~supported_flags;
4732 names = btrfs_printable_features(set, unsupported);
4735 "this kernel does not support the %s feature bit%s",
4736 names, strchr(names, ',') ? "s" : "");
4740 "this kernel does not support %s bits 0x%llx",
4745 disallowed = set_mask & ~safe_set;
4747 names = btrfs_printable_features(set, disallowed);
4750 "can't set the %s feature bit%s while mounted",
4751 names, strchr(names, ',') ? "s" : "");
4755 "can't set %s bits 0x%llx while mounted",
4760 disallowed = clear_mask & ~safe_clear;
4762 names = btrfs_printable_features(set, disallowed);
4765 "can't clear the %s feature bit%s while mounted",
4766 names, strchr(names, ',') ? "s" : "");
4770 "can't clear %s bits 0x%llx while mounted",
4778 #define check_feature(fs_info, change_mask, flags, mask_base) \
4779 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4780 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4781 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4782 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4784 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4786 struct inode *inode = file_inode(file);
4787 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4788 struct btrfs_root *root = BTRFS_I(inode)->root;
4789 struct btrfs_super_block *super_block = fs_info->super_copy;
4790 struct btrfs_ioctl_feature_flags flags[2];
4791 struct btrfs_trans_handle *trans;
4795 if (!capable(CAP_SYS_ADMIN))
4798 if (copy_from_user(flags, arg, sizeof(flags)))
4802 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4803 !flags[0].incompat_flags)
4806 ret = check_feature(fs_info, flags[0].compat_flags,
4807 flags[1].compat_flags, COMPAT);
4811 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4812 flags[1].compat_ro_flags, COMPAT_RO);
4816 ret = check_feature(fs_info, flags[0].incompat_flags,
4817 flags[1].incompat_flags, INCOMPAT);
4821 ret = mnt_want_write_file(file);
4825 trans = btrfs_start_transaction(root, 0);
4826 if (IS_ERR(trans)) {
4827 ret = PTR_ERR(trans);
4828 goto out_drop_write;
4831 spin_lock(&fs_info->super_lock);
4832 newflags = btrfs_super_compat_flags(super_block);
4833 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4834 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4835 btrfs_set_super_compat_flags(super_block, newflags);
4837 newflags = btrfs_super_compat_ro_flags(super_block);
4838 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4839 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4840 btrfs_set_super_compat_ro_flags(super_block, newflags);
4842 newflags = btrfs_super_incompat_flags(super_block);
4843 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4844 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4845 btrfs_set_super_incompat_flags(super_block, newflags);
4846 spin_unlock(&fs_info->super_lock);
4848 ret = btrfs_commit_transaction(trans);
4850 mnt_drop_write_file(file);
4855 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4857 struct btrfs_ioctl_send_args *arg;
4861 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4862 struct btrfs_ioctl_send_args_32 args32;
4864 ret = copy_from_user(&args32, argp, sizeof(args32));
4867 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4870 arg->send_fd = args32.send_fd;
4871 arg->clone_sources_count = args32.clone_sources_count;
4872 arg->clone_sources = compat_ptr(args32.clone_sources);
4873 arg->parent_root = args32.parent_root;
4874 arg->flags = args32.flags;
4875 memcpy(arg->reserved, args32.reserved,
4876 sizeof(args32.reserved));
4881 arg = memdup_user(argp, sizeof(*arg));
4883 return PTR_ERR(arg);
4885 ret = btrfs_ioctl_send(file, arg);
4890 long btrfs_ioctl(struct file *file, unsigned int
4891 cmd, unsigned long arg)
4893 struct inode *inode = file_inode(file);
4894 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4895 struct btrfs_root *root = BTRFS_I(inode)->root;
4896 void __user *argp = (void __user *)arg;
4899 case FS_IOC_GETVERSION:
4900 return btrfs_ioctl_getversion(file, argp);
4901 case FS_IOC_GETFSLABEL:
4902 return btrfs_ioctl_get_fslabel(fs_info, argp);
4903 case FS_IOC_SETFSLABEL:
4904 return btrfs_ioctl_set_fslabel(file, argp);
4906 return btrfs_ioctl_fitrim(fs_info, argp);
4907 case BTRFS_IOC_SNAP_CREATE:
4908 return btrfs_ioctl_snap_create(file, argp, 0);
4909 case BTRFS_IOC_SNAP_CREATE_V2:
4910 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4911 case BTRFS_IOC_SUBVOL_CREATE:
4912 return btrfs_ioctl_snap_create(file, argp, 1);
4913 case BTRFS_IOC_SUBVOL_CREATE_V2:
4914 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4915 case BTRFS_IOC_SNAP_DESTROY:
4916 return btrfs_ioctl_snap_destroy(file, argp, false);
4917 case BTRFS_IOC_SNAP_DESTROY_V2:
4918 return btrfs_ioctl_snap_destroy(file, argp, true);
4919 case BTRFS_IOC_SUBVOL_GETFLAGS:
4920 return btrfs_ioctl_subvol_getflags(file, argp);
4921 case BTRFS_IOC_SUBVOL_SETFLAGS:
4922 return btrfs_ioctl_subvol_setflags(file, argp);
4923 case BTRFS_IOC_DEFAULT_SUBVOL:
4924 return btrfs_ioctl_default_subvol(file, argp);
4925 case BTRFS_IOC_DEFRAG:
4926 return btrfs_ioctl_defrag(file, NULL);
4927 case BTRFS_IOC_DEFRAG_RANGE:
4928 return btrfs_ioctl_defrag(file, argp);
4929 case BTRFS_IOC_RESIZE:
4930 return btrfs_ioctl_resize(file, argp);
4931 case BTRFS_IOC_ADD_DEV:
4932 return btrfs_ioctl_add_dev(fs_info, argp);
4933 case BTRFS_IOC_RM_DEV:
4934 return btrfs_ioctl_rm_dev(file, argp);
4935 case BTRFS_IOC_RM_DEV_V2:
4936 return btrfs_ioctl_rm_dev_v2(file, argp);
4937 case BTRFS_IOC_FS_INFO:
4938 return btrfs_ioctl_fs_info(fs_info, argp);
4939 case BTRFS_IOC_DEV_INFO:
4940 return btrfs_ioctl_dev_info(fs_info, argp);
4941 case BTRFS_IOC_BALANCE:
4942 return btrfs_ioctl_balance(file, NULL);
4943 case BTRFS_IOC_TREE_SEARCH:
4944 return btrfs_ioctl_tree_search(file, argp);
4945 case BTRFS_IOC_TREE_SEARCH_V2:
4946 return btrfs_ioctl_tree_search_v2(file, argp);
4947 case BTRFS_IOC_INO_LOOKUP:
4948 return btrfs_ioctl_ino_lookup(file, argp);
4949 case BTRFS_IOC_INO_PATHS:
4950 return btrfs_ioctl_ino_to_path(root, argp);
4951 case BTRFS_IOC_LOGICAL_INO:
4952 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4953 case BTRFS_IOC_LOGICAL_INO_V2:
4954 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4955 case BTRFS_IOC_SPACE_INFO:
4956 return btrfs_ioctl_space_info(fs_info, argp);
4957 case BTRFS_IOC_SYNC: {
4960 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4963 ret = btrfs_sync_fs(inode->i_sb, 1);
4965 * The transaction thread may want to do more work,
4966 * namely it pokes the cleaner kthread that will start
4967 * processing uncleaned subvols.
4969 wake_up_process(fs_info->transaction_kthread);
4972 case BTRFS_IOC_START_SYNC:
4973 return btrfs_ioctl_start_sync(root, argp);
4974 case BTRFS_IOC_WAIT_SYNC:
4975 return btrfs_ioctl_wait_sync(fs_info, argp);
4976 case BTRFS_IOC_SCRUB:
4977 return btrfs_ioctl_scrub(file, argp);
4978 case BTRFS_IOC_SCRUB_CANCEL:
4979 return btrfs_ioctl_scrub_cancel(fs_info);
4980 case BTRFS_IOC_SCRUB_PROGRESS:
4981 return btrfs_ioctl_scrub_progress(fs_info, argp);
4982 case BTRFS_IOC_BALANCE_V2:
4983 return btrfs_ioctl_balance(file, argp);
4984 case BTRFS_IOC_BALANCE_CTL:
4985 return btrfs_ioctl_balance_ctl(fs_info, arg);
4986 case BTRFS_IOC_BALANCE_PROGRESS:
4987 return btrfs_ioctl_balance_progress(fs_info, argp);
4988 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4989 return btrfs_ioctl_set_received_subvol(file, argp);
4991 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4992 return btrfs_ioctl_set_received_subvol_32(file, argp);
4994 case BTRFS_IOC_SEND:
4995 return _btrfs_ioctl_send(file, argp, false);
4996 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4997 case BTRFS_IOC_SEND_32:
4998 return _btrfs_ioctl_send(file, argp, true);
5000 case BTRFS_IOC_GET_DEV_STATS:
5001 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5002 case BTRFS_IOC_QUOTA_CTL:
5003 return btrfs_ioctl_quota_ctl(file, argp);
5004 case BTRFS_IOC_QGROUP_ASSIGN:
5005 return btrfs_ioctl_qgroup_assign(file, argp);
5006 case BTRFS_IOC_QGROUP_CREATE:
5007 return btrfs_ioctl_qgroup_create(file, argp);
5008 case BTRFS_IOC_QGROUP_LIMIT:
5009 return btrfs_ioctl_qgroup_limit(file, argp);
5010 case BTRFS_IOC_QUOTA_RESCAN:
5011 return btrfs_ioctl_quota_rescan(file, argp);
5012 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5013 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5014 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5015 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5016 case BTRFS_IOC_DEV_REPLACE:
5017 return btrfs_ioctl_dev_replace(fs_info, argp);
5018 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5019 return btrfs_ioctl_get_supported_features(argp);
5020 case BTRFS_IOC_GET_FEATURES:
5021 return btrfs_ioctl_get_features(fs_info, argp);
5022 case BTRFS_IOC_SET_FEATURES:
5023 return btrfs_ioctl_set_features(file, argp);
5024 case BTRFS_IOC_GET_SUBVOL_INFO:
5025 return btrfs_ioctl_get_subvol_info(file, argp);
5026 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5027 return btrfs_ioctl_get_subvol_rootref(file, argp);
5028 case BTRFS_IOC_INO_LOOKUP_USER:
5029 return btrfs_ioctl_ino_lookup_user(file, argp);
5030 case FS_IOC_ENABLE_VERITY:
5031 return fsverity_ioctl_enable(file, (const void __user *)argp);
5032 case FS_IOC_MEASURE_VERITY:
5033 return fsverity_ioctl_measure(file, argp);
5039 #ifdef CONFIG_COMPAT
5040 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5043 * These all access 32-bit values anyway so no further
5044 * handling is necessary.
5047 case FS_IOC32_GETVERSION:
5048 cmd = FS_IOC_GETVERSION;
5052 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));