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
619 * exists). Also no need to have the tree block locked since it
620 * is not in any tree at this point, so no other task can find
623 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
624 free_extent_buffer(leaf);
628 free_extent_buffer(leaf);
631 key.offset = (u64)-1;
632 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
633 if (IS_ERR(new_root)) {
634 free_anon_bdev(anon_dev);
635 ret = PTR_ERR(new_root);
636 btrfs_abort_transaction(trans, ret);
639 /* Freeing will be done in btrfs_put_root() of new_root */
642 ret = btrfs_record_root_in_trans(trans, new_root);
644 btrfs_put_root(new_root);
645 btrfs_abort_transaction(trans, ret);
649 ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns);
650 btrfs_put_root(new_root);
652 /* We potentially lose an unused inode item here */
653 btrfs_abort_transaction(trans, ret);
658 * insert the directory item
660 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
662 btrfs_abort_transaction(trans, ret);
666 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
667 BTRFS_FT_DIR, index);
669 btrfs_abort_transaction(trans, ret);
673 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
674 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
676 btrfs_abort_transaction(trans, ret);
680 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
681 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
683 btrfs_abort_transaction(trans, ret);
687 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
688 BTRFS_UUID_KEY_SUBVOL, objectid);
690 btrfs_abort_transaction(trans, ret);
694 trans->block_rsv = NULL;
695 trans->bytes_reserved = 0;
696 btrfs_subvolume_release_metadata(root, &block_rsv);
698 err = btrfs_commit_transaction(trans);
703 inode = btrfs_lookup_dentry(dir, dentry);
705 return PTR_ERR(inode);
706 d_instantiate(dentry, inode);
712 free_anon_bdev(anon_dev);
717 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
718 struct dentry *dentry, bool readonly,
719 struct btrfs_qgroup_inherit *inherit)
721 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
723 struct btrfs_pending_snapshot *pending_snapshot;
724 struct btrfs_trans_handle *trans;
727 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
730 if (atomic_read(&root->nr_swapfiles)) {
732 "cannot snapshot subvolume with active swapfile");
736 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
737 if (!pending_snapshot)
740 ret = get_anon_bdev(&pending_snapshot->anon_dev);
743 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
745 pending_snapshot->path = btrfs_alloc_path();
746 if (!pending_snapshot->root_item || !pending_snapshot->path) {
751 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
752 BTRFS_BLOCK_RSV_TEMP);
754 * 1 - parent dir inode
757 * 2 - root ref/backref
758 * 1 - root of snapshot
761 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
762 &pending_snapshot->block_rsv, 8,
767 pending_snapshot->dentry = dentry;
768 pending_snapshot->root = root;
769 pending_snapshot->readonly = readonly;
770 pending_snapshot->dir = dir;
771 pending_snapshot->inherit = inherit;
773 trans = btrfs_start_transaction(root, 0);
775 ret = PTR_ERR(trans);
779 trans->pending_snapshot = pending_snapshot;
781 ret = btrfs_commit_transaction(trans);
785 ret = pending_snapshot->error;
789 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
793 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
795 ret = PTR_ERR(inode);
799 d_instantiate(dentry, inode);
801 pending_snapshot->anon_dev = 0;
803 /* Prevent double freeing of anon_dev */
804 if (ret && pending_snapshot->snap)
805 pending_snapshot->snap->anon_dev = 0;
806 btrfs_put_root(pending_snapshot->snap);
807 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
809 if (pending_snapshot->anon_dev)
810 free_anon_bdev(pending_snapshot->anon_dev);
811 kfree(pending_snapshot->root_item);
812 btrfs_free_path(pending_snapshot->path);
813 kfree(pending_snapshot);
818 /* copy of may_delete in fs/namei.c()
819 * Check whether we can remove a link victim from directory dir, check
820 * whether the type of victim is right.
821 * 1. We can't do it if dir is read-only (done in permission())
822 * 2. We should have write and exec permissions on dir
823 * 3. We can't remove anything from append-only dir
824 * 4. We can't do anything with immutable dir (done in permission())
825 * 5. If the sticky bit on dir is set we should either
826 * a. be owner of dir, or
827 * b. be owner of victim, or
828 * c. have CAP_FOWNER capability
829 * 6. If the victim is append-only or immutable we can't do anything with
830 * links pointing to it.
831 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
832 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
833 * 9. We can't remove a root or mountpoint.
834 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
835 * nfs_async_unlink().
838 static int btrfs_may_delete(struct user_namespace *mnt_userns,
839 struct inode *dir, struct dentry *victim, int isdir)
843 if (d_really_is_negative(victim))
846 BUG_ON(d_inode(victim->d_parent) != dir);
847 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
849 error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
854 if (check_sticky(mnt_userns, dir, d_inode(victim)) ||
855 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
856 IS_SWAPFILE(d_inode(victim)))
859 if (!d_is_dir(victim))
863 } else if (d_is_dir(victim))
867 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
872 /* copy of may_create in fs/namei.c() */
873 static inline int btrfs_may_create(struct user_namespace *mnt_userns,
874 struct inode *dir, struct dentry *child)
876 if (d_really_is_positive(child))
880 if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns))
882 return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
886 * Create a new subvolume below @parent. This is largely modeled after
887 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
888 * inside this filesystem so it's quite a bit simpler.
890 static noinline int btrfs_mksubvol(const struct path *parent,
891 struct user_namespace *mnt_userns,
892 const char *name, int namelen,
893 struct btrfs_root *snap_src,
895 struct btrfs_qgroup_inherit *inherit)
897 struct inode *dir = d_inode(parent->dentry);
898 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
899 struct dentry *dentry;
902 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
906 dentry = lookup_one(mnt_userns, name, parent->dentry, namelen);
907 error = PTR_ERR(dentry);
911 error = btrfs_may_create(mnt_userns, dir, dentry);
916 * even if this name doesn't exist, we may get hash collisions.
917 * check for them now when we can safely fail
919 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
925 down_read(&fs_info->subvol_sem);
927 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
931 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
933 error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit);
936 fsnotify_mkdir(dir, dentry);
938 up_read(&fs_info->subvol_sem);
942 btrfs_inode_unlock(dir, 0);
946 static noinline int btrfs_mksnapshot(const struct path *parent,
947 struct user_namespace *mnt_userns,
948 const char *name, int namelen,
949 struct btrfs_root *root,
951 struct btrfs_qgroup_inherit *inherit)
954 bool snapshot_force_cow = false;
957 * Force new buffered writes to reserve space even when NOCOW is
958 * possible. This is to avoid later writeback (running dealloc) to
959 * fallback to COW mode and unexpectedly fail with ENOSPC.
961 btrfs_drew_read_lock(&root->snapshot_lock);
963 ret = btrfs_start_delalloc_snapshot(root, false);
968 * All previous writes have started writeback in NOCOW mode, so now
969 * we force future writes to fallback to COW mode during snapshot
972 atomic_inc(&root->snapshot_force_cow);
973 snapshot_force_cow = true;
975 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
977 ret = btrfs_mksubvol(parent, mnt_userns, name, namelen,
978 root, readonly, inherit);
980 if (snapshot_force_cow)
981 atomic_dec(&root->snapshot_force_cow);
982 btrfs_drew_read_unlock(&root->snapshot_lock);
987 * When we're defragging a range, we don't want to kick it off again
988 * if it is really just waiting for delalloc to send it down.
989 * If we find a nice big extent or delalloc range for the bytes in the
990 * file you want to defrag, we return 0 to let you know to skip this
993 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
995 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
996 struct extent_map *em = NULL;
997 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1000 read_lock(&em_tree->lock);
1001 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1002 read_unlock(&em_tree->lock);
1005 end = extent_map_end(em);
1006 free_extent_map(em);
1007 if (end - offset > thresh)
1010 /* if we already have a nice delalloc here, just stop */
1012 end = count_range_bits(io_tree, &offset, offset + thresh,
1013 thresh, EXTENT_DELALLOC, 1);
1020 * helper function to walk through a file and find extents
1021 * newer than a specific transid, and smaller than thresh.
1023 * This is used by the defragging code to find new and small
1026 static int find_new_extents(struct btrfs_root *root,
1027 struct inode *inode, u64 newer_than,
1028 u64 *off, u32 thresh)
1030 struct btrfs_path *path;
1031 struct btrfs_key min_key;
1032 struct extent_buffer *leaf;
1033 struct btrfs_file_extent_item *extent;
1036 u64 ino = btrfs_ino(BTRFS_I(inode));
1038 path = btrfs_alloc_path();
1042 min_key.objectid = ino;
1043 min_key.type = BTRFS_EXTENT_DATA_KEY;
1044 min_key.offset = *off;
1047 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1051 if (min_key.objectid != ino)
1053 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1056 leaf = path->nodes[0];
1057 extent = btrfs_item_ptr(leaf, path->slots[0],
1058 struct btrfs_file_extent_item);
1060 type = btrfs_file_extent_type(leaf, extent);
1061 if (type == BTRFS_FILE_EXTENT_REG &&
1062 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1063 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1064 *off = min_key.offset;
1065 btrfs_free_path(path);
1070 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1071 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1075 if (min_key.offset == (u64)-1)
1079 btrfs_release_path(path);
1082 btrfs_free_path(path);
1086 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1088 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1089 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1090 struct extent_map *em;
1091 u64 len = PAGE_SIZE;
1094 * hopefully we have this extent in the tree already, try without
1095 * the full extent lock
1097 read_lock(&em_tree->lock);
1098 em = lookup_extent_mapping(em_tree, start, len);
1099 read_unlock(&em_tree->lock);
1102 struct extent_state *cached = NULL;
1103 u64 end = start + len - 1;
1105 /* get the big lock and read metadata off disk */
1106 lock_extent_bits(io_tree, start, end, &cached);
1107 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1108 unlock_extent_cached(io_tree, start, end, &cached);
1117 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1119 struct extent_map *next;
1122 /* this is the last extent */
1123 if (em->start + em->len >= i_size_read(inode))
1126 next = defrag_lookup_extent(inode, em->start + em->len);
1127 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1129 else if ((em->block_start + em->block_len == next->block_start) &&
1130 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1133 free_extent_map(next);
1137 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1138 u64 *last_len, u64 *skip, u64 *defrag_end,
1141 struct extent_map *em;
1143 bool next_mergeable = true;
1144 bool prev_mergeable = true;
1147 * make sure that once we start defragging an extent, we keep on
1150 if (start < *defrag_end)
1155 em = defrag_lookup_extent(inode, start);
1159 /* this will cover holes, and inline extents */
1160 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1166 prev_mergeable = false;
1168 next_mergeable = defrag_check_next_extent(inode, em);
1170 * we hit a real extent, if it is big or the next extent is not a
1171 * real extent, don't bother defragging it
1173 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1174 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1178 * last_len ends up being a counter of how many bytes we've defragged.
1179 * every time we choose not to defrag an extent, we reset *last_len
1180 * so that the next tiny extent will force a defrag.
1182 * The end result of this is that tiny extents before a single big
1183 * extent will force at least part of that big extent to be defragged.
1186 *defrag_end = extent_map_end(em);
1189 *skip = extent_map_end(em);
1193 free_extent_map(em);
1198 * it doesn't do much good to defrag one or two pages
1199 * at a time. This pulls in a nice chunk of pages
1200 * to COW and defrag.
1202 * It also makes sure the delalloc code has enough
1203 * dirty data to avoid making new small extents as part
1206 * It's a good idea to start RA on this range
1207 * before calling this.
1209 static int cluster_pages_for_defrag(struct inode *inode,
1210 struct page **pages,
1211 unsigned long start_index,
1212 unsigned long num_pages)
1214 unsigned long file_end;
1215 u64 isize = i_size_read(inode);
1219 u64 start = (u64)start_index << PAGE_SHIFT;
1224 struct btrfs_ordered_extent *ordered;
1225 struct extent_state *cached_state = NULL;
1226 struct extent_io_tree *tree;
1227 struct extent_changeset *data_reserved = NULL;
1228 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1230 file_end = (isize - 1) >> PAGE_SHIFT;
1231 if (!isize || start_index > file_end)
1234 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1236 ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
1237 start, page_cnt << PAGE_SHIFT);
1241 tree = &BTRFS_I(inode)->io_tree;
1243 /* step one, lock all the pages */
1244 for (i = 0; i < page_cnt; i++) {
1247 page = find_or_create_page(inode->i_mapping,
1248 start_index + i, mask);
1252 ret = set_page_extent_mapped(page);
1259 page_start = page_offset(page);
1260 page_end = page_start + PAGE_SIZE - 1;
1262 lock_extent_bits(tree, page_start, page_end,
1264 ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode),
1266 unlock_extent_cached(tree, page_start, page_end,
1272 btrfs_start_ordered_extent(ordered, 1);
1273 btrfs_put_ordered_extent(ordered);
1276 * we unlocked the page above, so we need check if
1277 * it was released or not.
1279 if (page->mapping != inode->i_mapping) {
1286 if (!PageUptodate(page)) {
1287 btrfs_readpage(NULL, page);
1289 if (!PageUptodate(page)) {
1297 if (page->mapping != inode->i_mapping) {
1309 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1313 * so now we have a nice long stream of locked
1314 * and up to date pages, lets wait on them
1316 for (i = 0; i < i_done; i++)
1317 wait_on_page_writeback(pages[i]);
1319 page_start = page_offset(pages[0]);
1320 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1322 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1323 page_start, page_end - 1, &cached_state);
1326 * When defragmenting we skip ranges that have holes or inline extents,
1327 * (check should_defrag_range()), to avoid unnecessary IO and wasting
1328 * space. At btrfs_defrag_file(), we check if a range should be defragged
1329 * before locking the inode and then, if it should, we trigger a sync
1330 * page cache readahead - we lock the inode only after that to avoid
1331 * blocking for too long other tasks that possibly want to operate on
1332 * other file ranges. But before we were able to get the inode lock,
1333 * some other task may have punched a hole in the range, or we may have
1334 * now an inline extent, in which case we should not defrag. So check
1335 * for that here, where we have the inode and the range locked, and bail
1336 * out if that happened.
1338 search_start = page_start;
1339 while (search_start < page_end) {
1340 struct extent_map *em;
1342 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, search_start,
1343 page_end - search_start);
1346 goto out_unlock_range;
1348 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1349 free_extent_map(em);
1350 /* Ok, 0 means we did not defrag anything */
1352 goto out_unlock_range;
1354 search_start = extent_map_end(em);
1355 free_extent_map(em);
1358 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1359 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1360 EXTENT_DEFRAG, 0, 0, &cached_state);
1362 if (i_done != page_cnt) {
1363 spin_lock(&BTRFS_I(inode)->lock);
1364 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1365 spin_unlock(&BTRFS_I(inode)->lock);
1366 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1367 start, (page_cnt - i_done) << PAGE_SHIFT, true);
1371 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1374 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1375 page_start, page_end - 1, &cached_state);
1377 for (i = 0; i < i_done; i++) {
1378 clear_page_dirty_for_io(pages[i]);
1379 ClearPageChecked(pages[i]);
1380 set_page_dirty(pages[i]);
1381 unlock_page(pages[i]);
1384 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1385 extent_changeset_free(data_reserved);
1389 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1390 page_start, page_end - 1, &cached_state);
1392 for (i = 0; i < i_done; i++) {
1393 unlock_page(pages[i]);
1396 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1397 start, page_cnt << PAGE_SHIFT, true);
1398 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1399 extent_changeset_free(data_reserved);
1404 int btrfs_defrag_file(struct inode *inode, struct file *file,
1405 struct btrfs_ioctl_defrag_range_args *range,
1406 u64 newer_than, unsigned long max_to_defrag)
1408 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1409 struct btrfs_root *root = BTRFS_I(inode)->root;
1410 struct file_ra_state *ra = NULL;
1411 unsigned long last_index;
1412 u64 isize = i_size_read(inode);
1416 u64 newer_off = range->start;
1418 unsigned long ra_index = 0;
1420 int defrag_count = 0;
1421 int compress_type = BTRFS_COMPRESS_ZLIB;
1422 u32 extent_thresh = range->extent_thresh;
1423 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1424 unsigned long cluster = max_cluster;
1425 u64 new_align = ~((u64)SZ_128K - 1);
1426 struct page **pages = NULL;
1427 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1432 if (range->start >= isize)
1436 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1438 if (range->compress_type)
1439 compress_type = range->compress_type;
1442 if (extent_thresh == 0)
1443 extent_thresh = SZ_256K;
1446 * If we were not given a file, allocate a readahead context. As
1447 * readahead is just an optimization, defrag will work without it so
1448 * we don't error out.
1451 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1453 file_ra_state_init(ra, inode->i_mapping);
1458 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1464 /* find the last page to defrag */
1465 if (range->start + range->len > range->start) {
1466 last_index = min_t(u64, isize - 1,
1467 range->start + range->len - 1) >> PAGE_SHIFT;
1469 last_index = (isize - 1) >> PAGE_SHIFT;
1473 ret = find_new_extents(root, inode, newer_than,
1474 &newer_off, SZ_64K);
1476 range->start = newer_off;
1478 * we always align our defrag to help keep
1479 * the extents in the file evenly spaced
1481 i = (newer_off & new_align) >> PAGE_SHIFT;
1485 i = range->start >> PAGE_SHIFT;
1488 max_to_defrag = last_index - i + 1;
1491 * make writeback starts from i, so the defrag range can be
1492 * written sequentially.
1494 if (i < inode->i_mapping->writeback_index)
1495 inode->i_mapping->writeback_index = i;
1497 while (i <= last_index && defrag_count < max_to_defrag &&
1498 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1500 * make sure we stop running if someone unmounts
1503 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1506 if (btrfs_defrag_cancelled(fs_info)) {
1507 btrfs_debug(fs_info, "defrag_file cancelled");
1512 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1513 extent_thresh, &last_len, &skip,
1514 &defrag_end, do_compress)){
1517 * the should_defrag function tells us how much to skip
1518 * bump our counter by the suggested amount
1520 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1521 i = max(i + 1, next);
1526 cluster = (PAGE_ALIGN(defrag_end) >>
1528 cluster = min(cluster, max_cluster);
1530 cluster = max_cluster;
1533 if (i + cluster > ra_index) {
1534 ra_index = max(i, ra_index);
1536 page_cache_sync_readahead(inode->i_mapping, ra,
1537 file, ra_index, cluster);
1538 ra_index += cluster;
1541 btrfs_inode_lock(inode, 0);
1542 if (IS_SWAPFILE(inode)) {
1546 BTRFS_I(inode)->defrag_compress = compress_type;
1547 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1550 btrfs_inode_unlock(inode, 0);
1554 defrag_count += ret;
1555 balance_dirty_pages_ratelimited(inode->i_mapping);
1556 btrfs_inode_unlock(inode, 0);
1559 if (newer_off == (u64)-1)
1565 newer_off = max(newer_off + 1,
1566 (u64)i << PAGE_SHIFT);
1568 ret = find_new_extents(root, inode, newer_than,
1569 &newer_off, SZ_64K);
1571 range->start = newer_off;
1572 i = (newer_off & new_align) >> PAGE_SHIFT;
1579 last_len += ret << PAGE_SHIFT;
1589 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1590 filemap_flush(inode->i_mapping);
1591 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1592 &BTRFS_I(inode)->runtime_flags))
1593 filemap_flush(inode->i_mapping);
1596 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1597 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1598 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1599 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1604 btrfs_inode_lock(inode, 0);
1605 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1606 btrfs_inode_unlock(inode, 0);
1615 * Try to start exclusive operation @type or cancel it if it's running.
1618 * 0 - normal mode, newly claimed op started
1619 * >0 - normal mode, something else is running,
1620 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1621 * ECANCELED - cancel mode, successful cancel
1622 * ENOTCONN - cancel mode, operation not running anymore
1624 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1625 enum btrfs_exclusive_operation type, bool cancel)
1628 /* Start normal op */
1629 if (!btrfs_exclop_start(fs_info, type))
1630 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1631 /* Exclusive operation is now claimed */
1635 /* Cancel running op */
1636 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1638 * This blocks any exclop finish from setting it to NONE, so we
1639 * request cancellation. Either it runs and we will wait for it,
1640 * or it has finished and no waiting will happen.
1642 atomic_inc(&fs_info->reloc_cancel_req);
1643 btrfs_exclop_start_unlock(fs_info);
1645 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1646 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1647 TASK_INTERRUPTIBLE);
1652 /* Something else is running or none */
1656 static noinline int btrfs_ioctl_resize(struct file *file,
1659 struct inode *inode = file_inode(file);
1660 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1664 struct btrfs_root *root = BTRFS_I(inode)->root;
1665 struct btrfs_ioctl_vol_args *vol_args;
1666 struct btrfs_trans_handle *trans;
1667 struct btrfs_device *device = NULL;
1670 char *devstr = NULL;
1675 if (!capable(CAP_SYS_ADMIN))
1678 ret = mnt_want_write_file(file);
1683 * Read the arguments before checking exclusivity to be able to
1684 * distinguish regular resize and cancel
1686 vol_args = memdup_user(arg, sizeof(*vol_args));
1687 if (IS_ERR(vol_args)) {
1688 ret = PTR_ERR(vol_args);
1691 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1692 sizestr = vol_args->name;
1693 cancel = (strcmp("cancel", sizestr) == 0);
1694 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1697 /* Exclusive operation is now claimed */
1699 devstr = strchr(sizestr, ':');
1701 sizestr = devstr + 1;
1703 devstr = vol_args->name;
1704 ret = kstrtoull(devstr, 10, &devid);
1711 btrfs_info(fs_info, "resizing devid %llu", devid);
1714 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL);
1716 btrfs_info(fs_info, "resizer unable to find device %llu",
1722 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1724 "resizer unable to apply on readonly device %llu",
1730 if (!strcmp(sizestr, "max"))
1731 new_size = device->bdev->bd_inode->i_size;
1733 if (sizestr[0] == '-') {
1736 } else if (sizestr[0] == '+') {
1740 new_size = memparse(sizestr, &retptr);
1741 if (*retptr != '\0' || new_size == 0) {
1747 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1752 old_size = btrfs_device_get_total_bytes(device);
1755 if (new_size > old_size) {
1759 new_size = old_size - new_size;
1760 } else if (mod > 0) {
1761 if (new_size > ULLONG_MAX - old_size) {
1765 new_size = old_size + new_size;
1768 if (new_size < SZ_256M) {
1772 if (new_size > device->bdev->bd_inode->i_size) {
1777 new_size = round_down(new_size, fs_info->sectorsize);
1779 if (new_size > old_size) {
1780 trans = btrfs_start_transaction(root, 0);
1781 if (IS_ERR(trans)) {
1782 ret = PTR_ERR(trans);
1785 ret = btrfs_grow_device(trans, device, new_size);
1786 btrfs_commit_transaction(trans);
1787 } else if (new_size < old_size) {
1788 ret = btrfs_shrink_device(device, new_size);
1789 } /* equal, nothing need to do */
1791 if (ret == 0 && new_size != old_size)
1792 btrfs_info_in_rcu(fs_info,
1793 "resize device %s (devid %llu) from %llu to %llu",
1794 rcu_str_deref(device->name), device->devid,
1795 old_size, new_size);
1797 btrfs_exclop_finish(fs_info);
1801 mnt_drop_write_file(file);
1805 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1806 struct user_namespace *mnt_userns,
1807 const char *name, unsigned long fd, int subvol,
1809 struct btrfs_qgroup_inherit *inherit)
1814 if (!S_ISDIR(file_inode(file)->i_mode))
1817 ret = mnt_want_write_file(file);
1821 namelen = strlen(name);
1822 if (strchr(name, '/')) {
1824 goto out_drop_write;
1827 if (name[0] == '.' &&
1828 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1830 goto out_drop_write;
1834 ret = btrfs_mksubvol(&file->f_path, mnt_userns, name,
1835 namelen, NULL, readonly, inherit);
1837 struct fd src = fdget(fd);
1838 struct inode *src_inode;
1841 goto out_drop_write;
1844 src_inode = file_inode(src.file);
1845 if (src_inode->i_sb != file_inode(file)->i_sb) {
1846 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1847 "Snapshot src from another FS");
1849 } else if (!inode_owner_or_capable(mnt_userns, src_inode)) {
1851 * Subvolume creation is not restricted, but snapshots
1852 * are limited to own subvolumes only
1856 ret = btrfs_mksnapshot(&file->f_path, mnt_userns,
1858 BTRFS_I(src_inode)->root,
1864 mnt_drop_write_file(file);
1869 static noinline int btrfs_ioctl_snap_create(struct file *file,
1870 void __user *arg, int subvol)
1872 struct btrfs_ioctl_vol_args *vol_args;
1875 if (!S_ISDIR(file_inode(file)->i_mode))
1878 vol_args = memdup_user(arg, sizeof(*vol_args));
1879 if (IS_ERR(vol_args))
1880 return PTR_ERR(vol_args);
1881 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1883 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
1884 vol_args->name, vol_args->fd, subvol,
1891 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1892 void __user *arg, int subvol)
1894 struct btrfs_ioctl_vol_args_v2 *vol_args;
1896 bool readonly = false;
1897 struct btrfs_qgroup_inherit *inherit = NULL;
1899 if (!S_ISDIR(file_inode(file)->i_mode))
1902 vol_args = memdup_user(arg, sizeof(*vol_args));
1903 if (IS_ERR(vol_args))
1904 return PTR_ERR(vol_args);
1905 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1907 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1912 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1914 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1917 if (vol_args->size < sizeof(*inherit) ||
1918 vol_args->size > PAGE_SIZE) {
1922 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1923 if (IS_ERR(inherit)) {
1924 ret = PTR_ERR(inherit);
1928 if (inherit->num_qgroups > PAGE_SIZE ||
1929 inherit->num_ref_copies > PAGE_SIZE ||
1930 inherit->num_excl_copies > PAGE_SIZE) {
1935 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1936 2 * inherit->num_excl_copies;
1937 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1943 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
1944 vol_args->name, vol_args->fd, subvol,
1955 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1958 struct inode *inode = file_inode(file);
1959 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1960 struct btrfs_root *root = BTRFS_I(inode)->root;
1964 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1967 down_read(&fs_info->subvol_sem);
1968 if (btrfs_root_readonly(root))
1969 flags |= BTRFS_SUBVOL_RDONLY;
1970 up_read(&fs_info->subvol_sem);
1972 if (copy_to_user(arg, &flags, sizeof(flags)))
1978 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1981 struct inode *inode = file_inode(file);
1982 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1983 struct btrfs_root *root = BTRFS_I(inode)->root;
1984 struct btrfs_trans_handle *trans;
1989 if (!inode_owner_or_capable(file_mnt_user_ns(file), inode))
1992 ret = mnt_want_write_file(file);
1996 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1998 goto out_drop_write;
2001 if (copy_from_user(&flags, arg, sizeof(flags))) {
2003 goto out_drop_write;
2006 if (flags & ~BTRFS_SUBVOL_RDONLY) {
2008 goto out_drop_write;
2011 down_write(&fs_info->subvol_sem);
2014 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2017 root_flags = btrfs_root_flags(&root->root_item);
2018 if (flags & BTRFS_SUBVOL_RDONLY) {
2019 btrfs_set_root_flags(&root->root_item,
2020 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2023 * Block RO -> RW transition if this subvolume is involved in
2026 spin_lock(&root->root_item_lock);
2027 if (root->send_in_progress == 0) {
2028 btrfs_set_root_flags(&root->root_item,
2029 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2030 spin_unlock(&root->root_item_lock);
2032 spin_unlock(&root->root_item_lock);
2034 "Attempt to set subvolume %llu read-write during send",
2035 root->root_key.objectid);
2041 trans = btrfs_start_transaction(root, 1);
2042 if (IS_ERR(trans)) {
2043 ret = PTR_ERR(trans);
2047 ret = btrfs_update_root(trans, fs_info->tree_root,
2048 &root->root_key, &root->root_item);
2050 btrfs_end_transaction(trans);
2054 ret = btrfs_commit_transaction(trans);
2058 btrfs_set_root_flags(&root->root_item, root_flags);
2060 up_write(&fs_info->subvol_sem);
2062 mnt_drop_write_file(file);
2067 static noinline int key_in_sk(struct btrfs_key *key,
2068 struct btrfs_ioctl_search_key *sk)
2070 struct btrfs_key test;
2073 test.objectid = sk->min_objectid;
2074 test.type = sk->min_type;
2075 test.offset = sk->min_offset;
2077 ret = btrfs_comp_cpu_keys(key, &test);
2081 test.objectid = sk->max_objectid;
2082 test.type = sk->max_type;
2083 test.offset = sk->max_offset;
2085 ret = btrfs_comp_cpu_keys(key, &test);
2091 static noinline int copy_to_sk(struct btrfs_path *path,
2092 struct btrfs_key *key,
2093 struct btrfs_ioctl_search_key *sk,
2096 unsigned long *sk_offset,
2100 struct extent_buffer *leaf;
2101 struct btrfs_ioctl_search_header sh;
2102 struct btrfs_key test;
2103 unsigned long item_off;
2104 unsigned long item_len;
2110 leaf = path->nodes[0];
2111 slot = path->slots[0];
2112 nritems = btrfs_header_nritems(leaf);
2114 if (btrfs_header_generation(leaf) > sk->max_transid) {
2118 found_transid = btrfs_header_generation(leaf);
2120 for (i = slot; i < nritems; i++) {
2121 item_off = btrfs_item_ptr_offset(leaf, i);
2122 item_len = btrfs_item_size_nr(leaf, i);
2124 btrfs_item_key_to_cpu(leaf, key, i);
2125 if (!key_in_sk(key, sk))
2128 if (sizeof(sh) + item_len > *buf_size) {
2135 * return one empty item back for v1, which does not
2139 *buf_size = sizeof(sh) + item_len;
2144 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2149 sh.objectid = key->objectid;
2150 sh.offset = key->offset;
2151 sh.type = key->type;
2153 sh.transid = found_transid;
2156 * Copy search result header. If we fault then loop again so we
2157 * can fault in the pages and -EFAULT there if there's a
2158 * problem. Otherwise we'll fault and then copy the buffer in
2159 * properly this next time through
2161 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2166 *sk_offset += sizeof(sh);
2169 char __user *up = ubuf + *sk_offset;
2171 * Copy the item, same behavior as above, but reset the
2172 * * sk_offset so we copy the full thing again.
2174 if (read_extent_buffer_to_user_nofault(leaf, up,
2175 item_off, item_len)) {
2177 *sk_offset -= sizeof(sh);
2181 *sk_offset += item_len;
2185 if (ret) /* -EOVERFLOW from above */
2188 if (*num_found >= sk->nr_items) {
2195 test.objectid = sk->max_objectid;
2196 test.type = sk->max_type;
2197 test.offset = sk->max_offset;
2198 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2200 else if (key->offset < (u64)-1)
2202 else if (key->type < (u8)-1) {
2205 } else if (key->objectid < (u64)-1) {
2213 * 0: all items from this leaf copied, continue with next
2214 * 1: * more items can be copied, but unused buffer is too small
2215 * * all items were found
2216 * Either way, it will stops the loop which iterates to the next
2218 * -EOVERFLOW: item was to large for buffer
2219 * -EFAULT: could not copy extent buffer back to userspace
2224 static noinline int search_ioctl(struct inode *inode,
2225 struct btrfs_ioctl_search_key *sk,
2229 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2230 struct btrfs_root *root;
2231 struct btrfs_key key;
2232 struct btrfs_path *path;
2235 unsigned long sk_offset = 0;
2237 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2238 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2242 path = btrfs_alloc_path();
2246 if (sk->tree_id == 0) {
2247 /* search the root of the inode that was passed */
2248 root = btrfs_grab_root(BTRFS_I(inode)->root);
2250 root = btrfs_get_fs_root(info, sk->tree_id, true);
2252 btrfs_free_path(path);
2253 return PTR_ERR(root);
2257 key.objectid = sk->min_objectid;
2258 key.type = sk->min_type;
2259 key.offset = sk->min_offset;
2263 if (fault_in_writeable(ubuf + sk_offset, *buf_size - sk_offset))
2266 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2272 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2273 &sk_offset, &num_found);
2274 btrfs_release_path(path);
2282 sk->nr_items = num_found;
2283 btrfs_put_root(root);
2284 btrfs_free_path(path);
2288 static noinline int btrfs_ioctl_tree_search(struct file *file,
2291 struct btrfs_ioctl_search_args __user *uargs;
2292 struct btrfs_ioctl_search_key sk;
2293 struct inode *inode;
2297 if (!capable(CAP_SYS_ADMIN))
2300 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2302 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2305 buf_size = sizeof(uargs->buf);
2307 inode = file_inode(file);
2308 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2311 * In the origin implementation an overflow is handled by returning a
2312 * search header with a len of zero, so reset ret.
2314 if (ret == -EOVERFLOW)
2317 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2322 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2325 struct btrfs_ioctl_search_args_v2 __user *uarg;
2326 struct btrfs_ioctl_search_args_v2 args;
2327 struct inode *inode;
2330 const size_t buf_limit = SZ_16M;
2332 if (!capable(CAP_SYS_ADMIN))
2335 /* copy search header and buffer size */
2336 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2337 if (copy_from_user(&args, uarg, sizeof(args)))
2340 buf_size = args.buf_size;
2342 /* limit result size to 16MB */
2343 if (buf_size > buf_limit)
2344 buf_size = buf_limit;
2346 inode = file_inode(file);
2347 ret = search_ioctl(inode, &args.key, &buf_size,
2348 (char __user *)(&uarg->buf[0]));
2349 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2351 else if (ret == -EOVERFLOW &&
2352 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2359 * Search INODE_REFs to identify path name of 'dirid' directory
2360 * in a 'tree_id' tree. and sets path name to 'name'.
2362 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2363 u64 tree_id, u64 dirid, char *name)
2365 struct btrfs_root *root;
2366 struct btrfs_key key;
2372 struct btrfs_inode_ref *iref;
2373 struct extent_buffer *l;
2374 struct btrfs_path *path;
2376 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2381 path = btrfs_alloc_path();
2385 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2387 root = btrfs_get_fs_root(info, tree_id, true);
2389 ret = PTR_ERR(root);
2394 key.objectid = dirid;
2395 key.type = BTRFS_INODE_REF_KEY;
2396 key.offset = (u64)-1;
2399 ret = btrfs_search_backwards(root, &key, path);
2408 slot = path->slots[0];
2410 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2411 len = btrfs_inode_ref_name_len(l, iref);
2413 total_len += len + 1;
2415 ret = -ENAMETOOLONG;
2420 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2422 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2425 btrfs_release_path(path);
2426 key.objectid = key.offset;
2427 key.offset = (u64)-1;
2428 dirid = key.objectid;
2430 memmove(name, ptr, total_len);
2431 name[total_len] = '\0';
2434 btrfs_put_root(root);
2435 btrfs_free_path(path);
2439 static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns,
2440 struct inode *inode,
2441 struct btrfs_ioctl_ino_lookup_user_args *args)
2443 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2444 struct super_block *sb = inode->i_sb;
2445 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2446 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2447 u64 dirid = args->dirid;
2448 unsigned long item_off;
2449 unsigned long item_len;
2450 struct btrfs_inode_ref *iref;
2451 struct btrfs_root_ref *rref;
2452 struct btrfs_root *root = NULL;
2453 struct btrfs_path *path;
2454 struct btrfs_key key, key2;
2455 struct extent_buffer *leaf;
2456 struct inode *temp_inode;
2463 path = btrfs_alloc_path();
2468 * If the bottom subvolume does not exist directly under upper_limit,
2469 * construct the path in from the bottom up.
2471 if (dirid != upper_limit.objectid) {
2472 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2474 root = btrfs_get_fs_root(fs_info, treeid, true);
2476 ret = PTR_ERR(root);
2480 key.objectid = dirid;
2481 key.type = BTRFS_INODE_REF_KEY;
2482 key.offset = (u64)-1;
2484 ret = btrfs_search_backwards(root, &key, path);
2492 leaf = path->nodes[0];
2493 slot = path->slots[0];
2495 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2496 len = btrfs_inode_ref_name_len(leaf, iref);
2498 total_len += len + 1;
2499 if (ptr < args->path) {
2500 ret = -ENAMETOOLONG;
2505 read_extent_buffer(leaf, ptr,
2506 (unsigned long)(iref + 1), len);
2508 /* Check the read+exec permission of this directory */
2509 ret = btrfs_previous_item(root, path, dirid,
2510 BTRFS_INODE_ITEM_KEY);
2513 } else if (ret > 0) {
2518 leaf = path->nodes[0];
2519 slot = path->slots[0];
2520 btrfs_item_key_to_cpu(leaf, &key2, slot);
2521 if (key2.objectid != dirid) {
2526 temp_inode = btrfs_iget(sb, key2.objectid, root);
2527 if (IS_ERR(temp_inode)) {
2528 ret = PTR_ERR(temp_inode);
2531 ret = inode_permission(mnt_userns, temp_inode,
2532 MAY_READ | MAY_EXEC);
2539 if (key.offset == upper_limit.objectid)
2541 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2546 btrfs_release_path(path);
2547 key.objectid = key.offset;
2548 key.offset = (u64)-1;
2549 dirid = key.objectid;
2552 memmove(args->path, ptr, total_len);
2553 args->path[total_len] = '\0';
2554 btrfs_put_root(root);
2556 btrfs_release_path(path);
2559 /* Get the bottom subvolume's name from ROOT_REF */
2560 key.objectid = treeid;
2561 key.type = BTRFS_ROOT_REF_KEY;
2562 key.offset = args->treeid;
2563 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2566 } else if (ret > 0) {
2571 leaf = path->nodes[0];
2572 slot = path->slots[0];
2573 btrfs_item_key_to_cpu(leaf, &key, slot);
2575 item_off = btrfs_item_ptr_offset(leaf, slot);
2576 item_len = btrfs_item_size_nr(leaf, slot);
2577 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2578 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2579 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2584 /* Copy subvolume's name */
2585 item_off += sizeof(struct btrfs_root_ref);
2586 item_len -= sizeof(struct btrfs_root_ref);
2587 read_extent_buffer(leaf, args->name, item_off, item_len);
2588 args->name[item_len] = 0;
2591 btrfs_put_root(root);
2593 btrfs_free_path(path);
2597 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2600 struct btrfs_ioctl_ino_lookup_args *args;
2601 struct inode *inode;
2604 args = memdup_user(argp, sizeof(*args));
2606 return PTR_ERR(args);
2608 inode = file_inode(file);
2611 * Unprivileged query to obtain the containing subvolume root id. The
2612 * path is reset so it's consistent with btrfs_search_path_in_tree.
2614 if (args->treeid == 0)
2615 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2617 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2622 if (!capable(CAP_SYS_ADMIN)) {
2627 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2628 args->treeid, args->objectid,
2632 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2640 * Version of ino_lookup ioctl (unprivileged)
2642 * The main differences from ino_lookup ioctl are:
2644 * 1. Read + Exec permission will be checked using inode_permission() during
2645 * path construction. -EACCES will be returned in case of failure.
2646 * 2. Path construction will be stopped at the inode number which corresponds
2647 * to the fd with which this ioctl is called. If constructed path does not
2648 * exist under fd's inode, -EACCES will be returned.
2649 * 3. The name of bottom subvolume is also searched and filled.
2651 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2653 struct btrfs_ioctl_ino_lookup_user_args *args;
2654 struct inode *inode;
2657 args = memdup_user(argp, sizeof(*args));
2659 return PTR_ERR(args);
2661 inode = file_inode(file);
2663 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2664 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2666 * The subvolume does not exist under fd with which this is
2673 ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args);
2675 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2682 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2683 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2685 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2686 struct btrfs_fs_info *fs_info;
2687 struct btrfs_root *root;
2688 struct btrfs_path *path;
2689 struct btrfs_key key;
2690 struct btrfs_root_item *root_item;
2691 struct btrfs_root_ref *rref;
2692 struct extent_buffer *leaf;
2693 unsigned long item_off;
2694 unsigned long item_len;
2695 struct inode *inode;
2699 path = btrfs_alloc_path();
2703 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2705 btrfs_free_path(path);
2709 inode = file_inode(file);
2710 fs_info = BTRFS_I(inode)->root->fs_info;
2712 /* Get root_item of inode's subvolume */
2713 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2714 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2716 ret = PTR_ERR(root);
2719 root_item = &root->root_item;
2721 subvol_info->treeid = key.objectid;
2723 subvol_info->generation = btrfs_root_generation(root_item);
2724 subvol_info->flags = btrfs_root_flags(root_item);
2726 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2727 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2729 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2732 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2733 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2734 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2736 subvol_info->otransid = btrfs_root_otransid(root_item);
2737 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2738 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2740 subvol_info->stransid = btrfs_root_stransid(root_item);
2741 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2742 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2744 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2745 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2746 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2748 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2749 /* Search root tree for ROOT_BACKREF of this subvolume */
2750 key.type = BTRFS_ROOT_BACKREF_KEY;
2752 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2755 } else if (path->slots[0] >=
2756 btrfs_header_nritems(path->nodes[0])) {
2757 ret = btrfs_next_leaf(fs_info->tree_root, path);
2760 } else if (ret > 0) {
2766 leaf = path->nodes[0];
2767 slot = path->slots[0];
2768 btrfs_item_key_to_cpu(leaf, &key, slot);
2769 if (key.objectid == subvol_info->treeid &&
2770 key.type == BTRFS_ROOT_BACKREF_KEY) {
2771 subvol_info->parent_id = key.offset;
2773 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2774 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2776 item_off = btrfs_item_ptr_offset(leaf, slot)
2777 + sizeof(struct btrfs_root_ref);
2778 item_len = btrfs_item_size_nr(leaf, slot)
2779 - sizeof(struct btrfs_root_ref);
2780 read_extent_buffer(leaf, subvol_info->name,
2781 item_off, item_len);
2788 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2792 btrfs_put_root(root);
2794 btrfs_free_path(path);
2800 * Return ROOT_REF information of the subvolume containing this inode
2801 * except the subvolume name.
2803 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2805 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2806 struct btrfs_root_ref *rref;
2807 struct btrfs_root *root;
2808 struct btrfs_path *path;
2809 struct btrfs_key key;
2810 struct extent_buffer *leaf;
2811 struct inode *inode;
2817 path = btrfs_alloc_path();
2821 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2822 if (IS_ERR(rootrefs)) {
2823 btrfs_free_path(path);
2824 return PTR_ERR(rootrefs);
2827 inode = file_inode(file);
2828 root = BTRFS_I(inode)->root->fs_info->tree_root;
2829 objectid = BTRFS_I(inode)->root->root_key.objectid;
2831 key.objectid = objectid;
2832 key.type = BTRFS_ROOT_REF_KEY;
2833 key.offset = rootrefs->min_treeid;
2836 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2839 } else if (path->slots[0] >=
2840 btrfs_header_nritems(path->nodes[0])) {
2841 ret = btrfs_next_leaf(root, path);
2844 } else if (ret > 0) {
2850 leaf = path->nodes[0];
2851 slot = path->slots[0];
2853 btrfs_item_key_to_cpu(leaf, &key, slot);
2854 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2859 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2864 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2865 rootrefs->rootref[found].treeid = key.offset;
2866 rootrefs->rootref[found].dirid =
2867 btrfs_root_ref_dirid(leaf, rref);
2870 ret = btrfs_next_item(root, path);
2873 } else if (ret > 0) {
2880 if (!ret || ret == -EOVERFLOW) {
2881 rootrefs->num_items = found;
2882 /* update min_treeid for next search */
2884 rootrefs->min_treeid =
2885 rootrefs->rootref[found - 1].treeid + 1;
2886 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2891 btrfs_free_path(path);
2896 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2900 struct dentry *parent = file->f_path.dentry;
2901 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2902 struct dentry *dentry;
2903 struct inode *dir = d_inode(parent);
2904 struct inode *inode;
2905 struct btrfs_root *root = BTRFS_I(dir)->root;
2906 struct btrfs_root *dest = NULL;
2907 struct btrfs_ioctl_vol_args *vol_args = NULL;
2908 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2909 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
2910 char *subvol_name, *subvol_name_ptr = NULL;
2913 bool destroy_parent = false;
2916 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2917 if (IS_ERR(vol_args2))
2918 return PTR_ERR(vol_args2);
2920 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2926 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2927 * name, same as v1 currently does.
2929 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2930 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2931 subvol_name = vol_args2->name;
2933 err = mnt_want_write_file(file);
2937 struct inode *old_dir;
2939 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2944 err = mnt_want_write_file(file);
2948 dentry = btrfs_get_dentry(fs_info->sb,
2949 BTRFS_FIRST_FREE_OBJECTID,
2950 vol_args2->subvolid, 0, 0);
2951 if (IS_ERR(dentry)) {
2952 err = PTR_ERR(dentry);
2953 goto out_drop_write;
2957 * Change the default parent since the subvolume being
2958 * deleted can be outside of the current mount point.
2960 parent = btrfs_get_parent(dentry);
2963 * At this point dentry->d_name can point to '/' if the
2964 * subvolume we want to destroy is outsite of the
2965 * current mount point, so we need to release the
2966 * current dentry and execute the lookup to return a new
2967 * one with ->d_name pointing to the
2968 * <mount point>/subvol_name.
2971 if (IS_ERR(parent)) {
2972 err = PTR_ERR(parent);
2973 goto out_drop_write;
2976 dir = d_inode(parent);
2979 * If v2 was used with SPEC_BY_ID, a new parent was
2980 * allocated since the subvolume can be outside of the
2981 * current mount point. Later on we need to release this
2982 * new parent dentry.
2984 destroy_parent = true;
2987 * On idmapped mounts, deletion via subvolid is
2988 * restricted to subvolumes that are immediate
2989 * ancestors of the inode referenced by the file
2990 * descriptor in the ioctl. Otherwise the idmapping
2991 * could potentially be abused to delete subvolumes
2992 * anywhere in the filesystem the user wouldn't be able
2993 * to delete without an idmapped mount.
2995 if (old_dir != dir && mnt_userns != &init_user_ns) {
3000 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
3001 fs_info, vol_args2->subvolid);
3002 if (IS_ERR(subvol_name_ptr)) {
3003 err = PTR_ERR(subvol_name_ptr);
3006 /* subvol_name_ptr is already nul terminated */
3007 subvol_name = (char *)kbasename(subvol_name_ptr);
3010 vol_args = memdup_user(arg, sizeof(*vol_args));
3011 if (IS_ERR(vol_args))
3012 return PTR_ERR(vol_args);
3014 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
3015 subvol_name = vol_args->name;
3017 err = mnt_want_write_file(file);
3022 subvol_namelen = strlen(subvol_name);
3024 if (strchr(subvol_name, '/') ||
3025 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3027 goto free_subvol_name;
3030 if (!S_ISDIR(dir->i_mode)) {
3032 goto free_subvol_name;
3035 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3037 goto free_subvol_name;
3038 dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen);
3039 if (IS_ERR(dentry)) {
3040 err = PTR_ERR(dentry);
3041 goto out_unlock_dir;
3044 if (d_really_is_negative(dentry)) {
3049 inode = d_inode(dentry);
3050 dest = BTRFS_I(inode)->root;
3051 if (!capable(CAP_SYS_ADMIN)) {
3053 * Regular user. Only allow this with a special mount
3054 * option, when the user has write+exec access to the
3055 * subvol root, and when rmdir(2) would have been
3058 * Note that this is _not_ check that the subvol is
3059 * empty or doesn't contain data that we wouldn't
3060 * otherwise be able to delete.
3062 * Users who want to delete empty subvols should try
3066 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3070 * Do not allow deletion if the parent dir is the same
3071 * as the dir to be deleted. That means the ioctl
3072 * must be called on the dentry referencing the root
3073 * of the subvol, not a random directory contained
3080 err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC);
3085 /* check if subvolume may be deleted by a user */
3086 err = btrfs_may_delete(mnt_userns, dir, dentry, 1);
3090 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3095 btrfs_inode_lock(inode, 0);
3096 err = btrfs_delete_subvolume(dir, dentry);
3097 btrfs_inode_unlock(inode, 0);
3099 d_delete_notify(dir, dentry);
3104 btrfs_inode_unlock(dir, 0);
3106 kfree(subvol_name_ptr);
3111 mnt_drop_write_file(file);
3118 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3120 struct inode *inode = file_inode(file);
3121 struct btrfs_root *root = BTRFS_I(inode)->root;
3122 struct btrfs_ioctl_defrag_range_args range = {0};
3125 ret = mnt_want_write_file(file);
3129 if (btrfs_root_readonly(root)) {
3134 /* Subpage defrag will be supported in later commits */
3135 if (root->fs_info->sectorsize < PAGE_SIZE) {
3140 switch (inode->i_mode & S_IFMT) {
3142 if (!capable(CAP_SYS_ADMIN)) {
3146 ret = btrfs_defrag_root(root);
3150 * Note that this does not check the file descriptor for write
3151 * access. This prevents defragmenting executables that are
3152 * running and allows defrag on files open in read-only mode.
3154 if (!capable(CAP_SYS_ADMIN) &&
3155 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
3161 if (copy_from_user(&range, argp, sizeof(range))) {
3165 /* compression requires us to start the IO */
3166 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3167 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
3168 range.extent_thresh = (u32)-1;
3171 /* the rest are all set to zero by kzalloc */
3172 range.len = (u64)-1;
3174 ret = btrfs_defrag_file(file_inode(file), file,
3175 &range, BTRFS_OLDEST_GENERATION, 0);
3183 mnt_drop_write_file(file);
3187 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3189 struct btrfs_ioctl_vol_args *vol_args;
3192 if (!capable(CAP_SYS_ADMIN))
3195 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD))
3196 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3198 vol_args = memdup_user(arg, sizeof(*vol_args));
3199 if (IS_ERR(vol_args)) {
3200 ret = PTR_ERR(vol_args);
3204 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3205 ret = btrfs_init_new_device(fs_info, vol_args->name);
3208 btrfs_info(fs_info, "disk added %s", vol_args->name);
3212 btrfs_exclop_finish(fs_info);
3216 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3218 struct inode *inode = file_inode(file);
3219 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3220 struct btrfs_ioctl_vol_args_v2 *vol_args;
3221 struct block_device *bdev = NULL;
3224 bool cancel = false;
3226 if (!capable(CAP_SYS_ADMIN))
3229 ret = mnt_want_write_file(file);
3233 vol_args = memdup_user(arg, sizeof(*vol_args));
3234 if (IS_ERR(vol_args)) {
3235 ret = PTR_ERR(vol_args);
3239 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3243 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3244 if (!(vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) &&
3245 strcmp("cancel", vol_args->name) == 0)
3248 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3252 /* Exclusive operation is now claimed */
3254 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3255 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid, &bdev, &mode);
3257 ret = btrfs_rm_device(fs_info, vol_args->name, 0, &bdev, &mode);
3259 btrfs_exclop_finish(fs_info);
3262 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3263 btrfs_info(fs_info, "device deleted: id %llu",
3266 btrfs_info(fs_info, "device deleted: %s",
3272 mnt_drop_write_file(file);
3274 blkdev_put(bdev, mode);
3278 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3280 struct inode *inode = file_inode(file);
3281 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3282 struct btrfs_ioctl_vol_args *vol_args;
3283 struct block_device *bdev = NULL;
3288 if (!capable(CAP_SYS_ADMIN))
3291 ret = mnt_want_write_file(file);
3295 vol_args = memdup_user(arg, sizeof(*vol_args));
3296 if (IS_ERR(vol_args)) {
3297 ret = PTR_ERR(vol_args);
3298 goto out_drop_write;
3300 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3301 cancel = (strcmp("cancel", vol_args->name) == 0);
3303 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3306 ret = btrfs_rm_device(fs_info, vol_args->name, 0, &bdev, &mode);
3308 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3309 btrfs_exclop_finish(fs_info);
3314 mnt_drop_write_file(file);
3316 blkdev_put(bdev, mode);
3320 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3323 struct btrfs_ioctl_fs_info_args *fi_args;
3324 struct btrfs_device *device;
3325 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3329 fi_args = memdup_user(arg, sizeof(*fi_args));
3330 if (IS_ERR(fi_args))
3331 return PTR_ERR(fi_args);
3333 flags_in = fi_args->flags;
3334 memset(fi_args, 0, sizeof(*fi_args));
3337 fi_args->num_devices = fs_devices->num_devices;
3339 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3340 if (device->devid > fi_args->max_id)
3341 fi_args->max_id = device->devid;
3345 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3346 fi_args->nodesize = fs_info->nodesize;
3347 fi_args->sectorsize = fs_info->sectorsize;
3348 fi_args->clone_alignment = fs_info->sectorsize;
3350 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3351 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3352 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3353 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3356 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3357 fi_args->generation = fs_info->generation;
3358 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3361 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3362 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3363 sizeof(fi_args->metadata_uuid));
3364 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3367 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3374 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3377 struct btrfs_ioctl_dev_info_args *di_args;
3378 struct btrfs_device *dev;
3380 char *s_uuid = NULL;
3382 di_args = memdup_user(arg, sizeof(*di_args));
3383 if (IS_ERR(di_args))
3384 return PTR_ERR(di_args);
3386 if (!btrfs_is_empty_uuid(di_args->uuid))
3387 s_uuid = di_args->uuid;
3390 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3398 di_args->devid = dev->devid;
3399 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3400 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3401 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3403 strncpy(di_args->path, rcu_str_deref(dev->name),
3404 sizeof(di_args->path) - 1);
3405 di_args->path[sizeof(di_args->path) - 1] = 0;
3407 di_args->path[0] = '\0';
3412 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3419 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3421 struct inode *inode = file_inode(file);
3422 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3423 struct btrfs_root *root = BTRFS_I(inode)->root;
3424 struct btrfs_root *new_root;
3425 struct btrfs_dir_item *di;
3426 struct btrfs_trans_handle *trans;
3427 struct btrfs_path *path = NULL;
3428 struct btrfs_disk_key disk_key;
3433 if (!capable(CAP_SYS_ADMIN))
3436 ret = mnt_want_write_file(file);
3440 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3446 objectid = BTRFS_FS_TREE_OBJECTID;
3448 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3449 if (IS_ERR(new_root)) {
3450 ret = PTR_ERR(new_root);
3453 if (!is_fstree(new_root->root_key.objectid)) {
3458 path = btrfs_alloc_path();
3464 trans = btrfs_start_transaction(root, 1);
3465 if (IS_ERR(trans)) {
3466 ret = PTR_ERR(trans);
3470 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3471 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3472 dir_id, "default", 7, 1);
3473 if (IS_ERR_OR_NULL(di)) {
3474 btrfs_release_path(path);
3475 btrfs_end_transaction(trans);
3477 "Umm, you don't have the default diritem, this isn't going to work");
3482 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3483 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3484 btrfs_mark_buffer_dirty(path->nodes[0]);
3485 btrfs_release_path(path);
3487 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3488 btrfs_end_transaction(trans);
3490 btrfs_put_root(new_root);
3491 btrfs_free_path(path);
3493 mnt_drop_write_file(file);
3497 static void get_block_group_info(struct list_head *groups_list,
3498 struct btrfs_ioctl_space_info *space)
3500 struct btrfs_block_group *block_group;
3502 space->total_bytes = 0;
3503 space->used_bytes = 0;
3505 list_for_each_entry(block_group, groups_list, list) {
3506 space->flags = block_group->flags;
3507 space->total_bytes += block_group->length;
3508 space->used_bytes += block_group->used;
3512 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3515 struct btrfs_ioctl_space_args space_args;
3516 struct btrfs_ioctl_space_info space;
3517 struct btrfs_ioctl_space_info *dest;
3518 struct btrfs_ioctl_space_info *dest_orig;
3519 struct btrfs_ioctl_space_info __user *user_dest;
3520 struct btrfs_space_info *info;
3521 static const u64 types[] = {
3522 BTRFS_BLOCK_GROUP_DATA,
3523 BTRFS_BLOCK_GROUP_SYSTEM,
3524 BTRFS_BLOCK_GROUP_METADATA,
3525 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3533 if (copy_from_user(&space_args,
3534 (struct btrfs_ioctl_space_args __user *)arg,
3535 sizeof(space_args)))
3538 for (i = 0; i < num_types; i++) {
3539 struct btrfs_space_info *tmp;
3542 list_for_each_entry(tmp, &fs_info->space_info, list) {
3543 if (tmp->flags == types[i]) {
3552 down_read(&info->groups_sem);
3553 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3554 if (!list_empty(&info->block_groups[c]))
3557 up_read(&info->groups_sem);
3561 * Global block reserve, exported as a space_info
3565 /* space_slots == 0 means they are asking for a count */
3566 if (space_args.space_slots == 0) {
3567 space_args.total_spaces = slot_count;
3571 slot_count = min_t(u64, space_args.space_slots, slot_count);
3573 alloc_size = sizeof(*dest) * slot_count;
3575 /* we generally have at most 6 or so space infos, one for each raid
3576 * level. So, a whole page should be more than enough for everyone
3578 if (alloc_size > PAGE_SIZE)
3581 space_args.total_spaces = 0;
3582 dest = kmalloc(alloc_size, GFP_KERNEL);
3587 /* now we have a buffer to copy into */
3588 for (i = 0; i < num_types; i++) {
3589 struct btrfs_space_info *tmp;
3595 list_for_each_entry(tmp, &fs_info->space_info, list) {
3596 if (tmp->flags == types[i]) {
3604 down_read(&info->groups_sem);
3605 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3606 if (!list_empty(&info->block_groups[c])) {
3607 get_block_group_info(&info->block_groups[c],
3609 memcpy(dest, &space, sizeof(space));
3611 space_args.total_spaces++;
3617 up_read(&info->groups_sem);
3621 * Add global block reserve
3624 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3626 spin_lock(&block_rsv->lock);
3627 space.total_bytes = block_rsv->size;
3628 space.used_bytes = block_rsv->size - block_rsv->reserved;
3629 spin_unlock(&block_rsv->lock);
3630 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3631 memcpy(dest, &space, sizeof(space));
3632 space_args.total_spaces++;
3635 user_dest = (struct btrfs_ioctl_space_info __user *)
3636 (arg + sizeof(struct btrfs_ioctl_space_args));
3638 if (copy_to_user(user_dest, dest_orig, alloc_size))
3643 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3649 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3652 struct btrfs_trans_handle *trans;
3656 trans = btrfs_attach_transaction_barrier(root);
3657 if (IS_ERR(trans)) {
3658 if (PTR_ERR(trans) != -ENOENT)
3659 return PTR_ERR(trans);
3661 /* No running transaction, don't bother */
3662 transid = root->fs_info->last_trans_committed;
3665 transid = trans->transid;
3666 ret = btrfs_commit_transaction_async(trans);
3668 btrfs_end_transaction(trans);
3673 if (copy_to_user(argp, &transid, sizeof(transid)))
3678 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3684 if (copy_from_user(&transid, argp, sizeof(transid)))
3687 transid = 0; /* current trans */
3689 return btrfs_wait_for_commit(fs_info, transid);
3692 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3694 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3695 struct btrfs_ioctl_scrub_args *sa;
3698 if (!capable(CAP_SYS_ADMIN))
3701 sa = memdup_user(arg, sizeof(*sa));
3705 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3706 ret = mnt_want_write_file(file);
3711 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3712 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3716 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3717 * error. This is important as it allows user space to know how much
3718 * progress scrub has done. For example, if scrub is canceled we get
3719 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3720 * space. Later user space can inspect the progress from the structure
3721 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3722 * previously (btrfs-progs does this).
3723 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3724 * then return -EFAULT to signal the structure was not copied or it may
3725 * be corrupt and unreliable due to a partial copy.
3727 if (copy_to_user(arg, sa, sizeof(*sa)))
3730 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3731 mnt_drop_write_file(file);
3737 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3739 if (!capable(CAP_SYS_ADMIN))
3742 return btrfs_scrub_cancel(fs_info);
3745 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3748 struct btrfs_ioctl_scrub_args *sa;
3751 if (!capable(CAP_SYS_ADMIN))
3754 sa = memdup_user(arg, sizeof(*sa));
3758 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3760 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3767 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3770 struct btrfs_ioctl_get_dev_stats *sa;
3773 sa = memdup_user(arg, sizeof(*sa));
3777 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3782 ret = btrfs_get_dev_stats(fs_info, sa);
3784 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3791 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3794 struct btrfs_ioctl_dev_replace_args *p;
3797 if (!capable(CAP_SYS_ADMIN))
3800 p = memdup_user(arg, sizeof(*p));
3805 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3806 if (sb_rdonly(fs_info->sb)) {
3810 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3811 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3813 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3814 btrfs_exclop_finish(fs_info);
3817 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3818 btrfs_dev_replace_status(fs_info, p);
3821 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3822 p->result = btrfs_dev_replace_cancel(fs_info);
3830 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3837 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3843 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3844 struct inode_fs_paths *ipath = NULL;
3845 struct btrfs_path *path;
3847 if (!capable(CAP_DAC_READ_SEARCH))
3850 path = btrfs_alloc_path();
3856 ipa = memdup_user(arg, sizeof(*ipa));
3863 size = min_t(u32, ipa->size, 4096);
3864 ipath = init_ipath(size, root, path);
3865 if (IS_ERR(ipath)) {
3866 ret = PTR_ERR(ipath);
3871 ret = paths_from_inode(ipa->inum, ipath);
3875 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3876 rel_ptr = ipath->fspath->val[i] -
3877 (u64)(unsigned long)ipath->fspath->val;
3878 ipath->fspath->val[i] = rel_ptr;
3881 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3882 ipath->fspath, size);
3889 btrfs_free_path(path);
3896 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3898 struct btrfs_data_container *inodes = ctx;
3899 const size_t c = 3 * sizeof(u64);
3901 if (inodes->bytes_left >= c) {
3902 inodes->bytes_left -= c;
3903 inodes->val[inodes->elem_cnt] = inum;
3904 inodes->val[inodes->elem_cnt + 1] = offset;
3905 inodes->val[inodes->elem_cnt + 2] = root;
3906 inodes->elem_cnt += 3;
3908 inodes->bytes_missing += c - inodes->bytes_left;
3909 inodes->bytes_left = 0;
3910 inodes->elem_missed += 3;
3916 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3917 void __user *arg, int version)
3921 struct btrfs_ioctl_logical_ino_args *loi;
3922 struct btrfs_data_container *inodes = NULL;
3923 struct btrfs_path *path = NULL;
3926 if (!capable(CAP_SYS_ADMIN))
3929 loi = memdup_user(arg, sizeof(*loi));
3931 return PTR_ERR(loi);
3934 ignore_offset = false;
3935 size = min_t(u32, loi->size, SZ_64K);
3937 /* All reserved bits must be 0 for now */
3938 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3942 /* Only accept flags we have defined so far */
3943 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3947 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3948 size = min_t(u32, loi->size, SZ_16M);
3951 path = btrfs_alloc_path();
3957 inodes = init_data_container(size);
3958 if (IS_ERR(inodes)) {
3959 ret = PTR_ERR(inodes);
3964 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3965 build_ino_list, inodes, ignore_offset);
3971 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3977 btrfs_free_path(path);
3985 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3986 struct btrfs_ioctl_balance_args *bargs)
3988 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3990 bargs->flags = bctl->flags;
3992 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3993 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3994 if (atomic_read(&fs_info->balance_pause_req))
3995 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3996 if (atomic_read(&fs_info->balance_cancel_req))
3997 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3999 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4000 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4001 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4003 spin_lock(&fs_info->balance_lock);
4004 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4005 spin_unlock(&fs_info->balance_lock);
4008 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4010 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4011 struct btrfs_fs_info *fs_info = root->fs_info;
4012 struct btrfs_ioctl_balance_args *bargs;
4013 struct btrfs_balance_control *bctl;
4014 bool need_unlock; /* for mut. excl. ops lock */
4017 if (!capable(CAP_SYS_ADMIN))
4020 ret = mnt_want_write_file(file);
4025 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4026 mutex_lock(&fs_info->balance_mutex);
4032 * mut. excl. ops lock is locked. Three possibilities:
4033 * (1) some other op is running
4034 * (2) balance is running
4035 * (3) balance is paused -- special case (think resume)
4037 mutex_lock(&fs_info->balance_mutex);
4038 if (fs_info->balance_ctl) {
4039 /* this is either (2) or (3) */
4040 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4041 mutex_unlock(&fs_info->balance_mutex);
4043 * Lock released to allow other waiters to continue,
4044 * we'll reexamine the status again.
4046 mutex_lock(&fs_info->balance_mutex);
4048 if (fs_info->balance_ctl &&
4049 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4051 need_unlock = false;
4055 mutex_unlock(&fs_info->balance_mutex);
4059 mutex_unlock(&fs_info->balance_mutex);
4065 mutex_unlock(&fs_info->balance_mutex);
4066 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4073 bargs = memdup_user(arg, sizeof(*bargs));
4074 if (IS_ERR(bargs)) {
4075 ret = PTR_ERR(bargs);
4079 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4080 if (!fs_info->balance_ctl) {
4085 bctl = fs_info->balance_ctl;
4086 spin_lock(&fs_info->balance_lock);
4087 bctl->flags |= BTRFS_BALANCE_RESUME;
4088 spin_unlock(&fs_info->balance_lock);
4096 if (fs_info->balance_ctl) {
4101 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4108 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4109 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4110 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4112 bctl->flags = bargs->flags;
4114 /* balance everything - no filters */
4115 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4118 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4125 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4126 * bctl is freed in reset_balance_state, or, if restriper was paused
4127 * all the way until unmount, in free_fs_info. The flag should be
4128 * cleared after reset_balance_state.
4130 need_unlock = false;
4132 ret = btrfs_balance(fs_info, bctl, bargs);
4135 if ((ret == 0 || ret == -ECANCELED) && arg) {
4136 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4145 mutex_unlock(&fs_info->balance_mutex);
4147 btrfs_exclop_finish(fs_info);
4149 mnt_drop_write_file(file);
4153 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4155 if (!capable(CAP_SYS_ADMIN))
4159 case BTRFS_BALANCE_CTL_PAUSE:
4160 return btrfs_pause_balance(fs_info);
4161 case BTRFS_BALANCE_CTL_CANCEL:
4162 return btrfs_cancel_balance(fs_info);
4168 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4171 struct btrfs_ioctl_balance_args *bargs;
4174 if (!capable(CAP_SYS_ADMIN))
4177 mutex_lock(&fs_info->balance_mutex);
4178 if (!fs_info->balance_ctl) {
4183 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4189 btrfs_update_ioctl_balance_args(fs_info, bargs);
4191 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4196 mutex_unlock(&fs_info->balance_mutex);
4200 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4202 struct inode *inode = file_inode(file);
4203 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4204 struct btrfs_ioctl_quota_ctl_args *sa;
4207 if (!capable(CAP_SYS_ADMIN))
4210 ret = mnt_want_write_file(file);
4214 sa = memdup_user(arg, sizeof(*sa));
4220 down_write(&fs_info->subvol_sem);
4223 case BTRFS_QUOTA_CTL_ENABLE:
4224 ret = btrfs_quota_enable(fs_info);
4226 case BTRFS_QUOTA_CTL_DISABLE:
4227 ret = btrfs_quota_disable(fs_info);
4235 up_write(&fs_info->subvol_sem);
4237 mnt_drop_write_file(file);
4241 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4243 struct inode *inode = file_inode(file);
4244 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4245 struct btrfs_root *root = BTRFS_I(inode)->root;
4246 struct btrfs_ioctl_qgroup_assign_args *sa;
4247 struct btrfs_trans_handle *trans;
4251 if (!capable(CAP_SYS_ADMIN))
4254 ret = mnt_want_write_file(file);
4258 sa = memdup_user(arg, sizeof(*sa));
4264 trans = btrfs_join_transaction(root);
4265 if (IS_ERR(trans)) {
4266 ret = PTR_ERR(trans);
4271 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4273 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4276 /* update qgroup status and info */
4277 err = btrfs_run_qgroups(trans);
4279 btrfs_handle_fs_error(fs_info, err,
4280 "failed to update qgroup status and info");
4281 err = btrfs_end_transaction(trans);
4288 mnt_drop_write_file(file);
4292 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4294 struct inode *inode = file_inode(file);
4295 struct btrfs_root *root = BTRFS_I(inode)->root;
4296 struct btrfs_ioctl_qgroup_create_args *sa;
4297 struct btrfs_trans_handle *trans;
4301 if (!capable(CAP_SYS_ADMIN))
4304 ret = mnt_want_write_file(file);
4308 sa = memdup_user(arg, sizeof(*sa));
4314 if (!sa->qgroupid) {
4319 trans = btrfs_join_transaction(root);
4320 if (IS_ERR(trans)) {
4321 ret = PTR_ERR(trans);
4326 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4328 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4331 err = btrfs_end_transaction(trans);
4338 mnt_drop_write_file(file);
4342 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4344 struct inode *inode = file_inode(file);
4345 struct btrfs_root *root = BTRFS_I(inode)->root;
4346 struct btrfs_ioctl_qgroup_limit_args *sa;
4347 struct btrfs_trans_handle *trans;
4352 if (!capable(CAP_SYS_ADMIN))
4355 ret = mnt_want_write_file(file);
4359 sa = memdup_user(arg, sizeof(*sa));
4365 trans = btrfs_join_transaction(root);
4366 if (IS_ERR(trans)) {
4367 ret = PTR_ERR(trans);
4371 qgroupid = sa->qgroupid;
4373 /* take the current subvol as qgroup */
4374 qgroupid = root->root_key.objectid;
4377 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4379 err = btrfs_end_transaction(trans);
4386 mnt_drop_write_file(file);
4390 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4392 struct inode *inode = file_inode(file);
4393 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4394 struct btrfs_ioctl_quota_rescan_args *qsa;
4397 if (!capable(CAP_SYS_ADMIN))
4400 ret = mnt_want_write_file(file);
4404 qsa = memdup_user(arg, sizeof(*qsa));
4415 ret = btrfs_qgroup_rescan(fs_info);
4420 mnt_drop_write_file(file);
4424 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4427 struct btrfs_ioctl_quota_rescan_args qsa = {0};
4430 if (!capable(CAP_SYS_ADMIN))
4433 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4435 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
4438 if (copy_to_user(arg, &qsa, sizeof(qsa)))
4444 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4447 if (!capable(CAP_SYS_ADMIN))
4450 return btrfs_qgroup_wait_for_completion(fs_info, true);
4453 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4454 struct user_namespace *mnt_userns,
4455 struct btrfs_ioctl_received_subvol_args *sa)
4457 struct inode *inode = file_inode(file);
4458 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4459 struct btrfs_root *root = BTRFS_I(inode)->root;
4460 struct btrfs_root_item *root_item = &root->root_item;
4461 struct btrfs_trans_handle *trans;
4462 struct timespec64 ct = current_time(inode);
4464 int received_uuid_changed;
4466 if (!inode_owner_or_capable(mnt_userns, inode))
4469 ret = mnt_want_write_file(file);
4473 down_write(&fs_info->subvol_sem);
4475 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4480 if (btrfs_root_readonly(root)) {
4487 * 2 - uuid items (received uuid + subvol uuid)
4489 trans = btrfs_start_transaction(root, 3);
4490 if (IS_ERR(trans)) {
4491 ret = PTR_ERR(trans);
4496 sa->rtransid = trans->transid;
4497 sa->rtime.sec = ct.tv_sec;
4498 sa->rtime.nsec = ct.tv_nsec;
4500 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4502 if (received_uuid_changed &&
4503 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4504 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4505 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4506 root->root_key.objectid);
4507 if (ret && ret != -ENOENT) {
4508 btrfs_abort_transaction(trans, ret);
4509 btrfs_end_transaction(trans);
4513 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4514 btrfs_set_root_stransid(root_item, sa->stransid);
4515 btrfs_set_root_rtransid(root_item, sa->rtransid);
4516 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4517 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4518 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4519 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4521 ret = btrfs_update_root(trans, fs_info->tree_root,
4522 &root->root_key, &root->root_item);
4524 btrfs_end_transaction(trans);
4527 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4528 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4529 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4530 root->root_key.objectid);
4531 if (ret < 0 && ret != -EEXIST) {
4532 btrfs_abort_transaction(trans, ret);
4533 btrfs_end_transaction(trans);
4537 ret = btrfs_commit_transaction(trans);
4539 up_write(&fs_info->subvol_sem);
4540 mnt_drop_write_file(file);
4545 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4548 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4549 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4552 args32 = memdup_user(arg, sizeof(*args32));
4554 return PTR_ERR(args32);
4556 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4562 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4563 args64->stransid = args32->stransid;
4564 args64->rtransid = args32->rtransid;
4565 args64->stime.sec = args32->stime.sec;
4566 args64->stime.nsec = args32->stime.nsec;
4567 args64->rtime.sec = args32->rtime.sec;
4568 args64->rtime.nsec = args32->rtime.nsec;
4569 args64->flags = args32->flags;
4571 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64);
4575 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4576 args32->stransid = args64->stransid;
4577 args32->rtransid = args64->rtransid;
4578 args32->stime.sec = args64->stime.sec;
4579 args32->stime.nsec = args64->stime.nsec;
4580 args32->rtime.sec = args64->rtime.sec;
4581 args32->rtime.nsec = args64->rtime.nsec;
4582 args32->flags = args64->flags;
4584 ret = copy_to_user(arg, args32, sizeof(*args32));
4595 static long btrfs_ioctl_set_received_subvol(struct file *file,
4598 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4601 sa = memdup_user(arg, sizeof(*sa));
4605 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa);
4610 ret = copy_to_user(arg, sa, sizeof(*sa));
4619 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4624 char label[BTRFS_LABEL_SIZE];
4626 spin_lock(&fs_info->super_lock);
4627 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4628 spin_unlock(&fs_info->super_lock);
4630 len = strnlen(label, BTRFS_LABEL_SIZE);
4632 if (len == BTRFS_LABEL_SIZE) {
4634 "label is too long, return the first %zu bytes",
4638 ret = copy_to_user(arg, label, len);
4640 return ret ? -EFAULT : 0;
4643 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4645 struct inode *inode = file_inode(file);
4646 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4647 struct btrfs_root *root = BTRFS_I(inode)->root;
4648 struct btrfs_super_block *super_block = fs_info->super_copy;
4649 struct btrfs_trans_handle *trans;
4650 char label[BTRFS_LABEL_SIZE];
4653 if (!capable(CAP_SYS_ADMIN))
4656 if (copy_from_user(label, arg, sizeof(label)))
4659 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4661 "unable to set label with more than %d bytes",
4662 BTRFS_LABEL_SIZE - 1);
4666 ret = mnt_want_write_file(file);
4670 trans = btrfs_start_transaction(root, 0);
4671 if (IS_ERR(trans)) {
4672 ret = PTR_ERR(trans);
4676 spin_lock(&fs_info->super_lock);
4677 strcpy(super_block->label, label);
4678 spin_unlock(&fs_info->super_lock);
4679 ret = btrfs_commit_transaction(trans);
4682 mnt_drop_write_file(file);
4686 #define INIT_FEATURE_FLAGS(suffix) \
4687 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4688 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4689 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4691 int btrfs_ioctl_get_supported_features(void __user *arg)
4693 static const struct btrfs_ioctl_feature_flags features[3] = {
4694 INIT_FEATURE_FLAGS(SUPP),
4695 INIT_FEATURE_FLAGS(SAFE_SET),
4696 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4699 if (copy_to_user(arg, &features, sizeof(features)))
4705 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4708 struct btrfs_super_block *super_block = fs_info->super_copy;
4709 struct btrfs_ioctl_feature_flags features;
4711 features.compat_flags = btrfs_super_compat_flags(super_block);
4712 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4713 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4715 if (copy_to_user(arg, &features, sizeof(features)))
4721 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4722 enum btrfs_feature_set set,
4723 u64 change_mask, u64 flags, u64 supported_flags,
4724 u64 safe_set, u64 safe_clear)
4726 const char *type = btrfs_feature_set_name(set);
4728 u64 disallowed, unsupported;
4729 u64 set_mask = flags & change_mask;
4730 u64 clear_mask = ~flags & change_mask;
4732 unsupported = set_mask & ~supported_flags;
4734 names = btrfs_printable_features(set, unsupported);
4737 "this kernel does not support the %s feature bit%s",
4738 names, strchr(names, ',') ? "s" : "");
4742 "this kernel does not support %s bits 0x%llx",
4747 disallowed = set_mask & ~safe_set;
4749 names = btrfs_printable_features(set, disallowed);
4752 "can't set the %s feature bit%s while mounted",
4753 names, strchr(names, ',') ? "s" : "");
4757 "can't set %s bits 0x%llx while mounted",
4762 disallowed = clear_mask & ~safe_clear;
4764 names = btrfs_printable_features(set, disallowed);
4767 "can't clear the %s feature bit%s while mounted",
4768 names, strchr(names, ',') ? "s" : "");
4772 "can't clear %s bits 0x%llx while mounted",
4780 #define check_feature(fs_info, change_mask, flags, mask_base) \
4781 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4782 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4783 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4784 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4786 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4788 struct inode *inode = file_inode(file);
4789 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4790 struct btrfs_root *root = BTRFS_I(inode)->root;
4791 struct btrfs_super_block *super_block = fs_info->super_copy;
4792 struct btrfs_ioctl_feature_flags flags[2];
4793 struct btrfs_trans_handle *trans;
4797 if (!capable(CAP_SYS_ADMIN))
4800 if (copy_from_user(flags, arg, sizeof(flags)))
4804 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4805 !flags[0].incompat_flags)
4808 ret = check_feature(fs_info, flags[0].compat_flags,
4809 flags[1].compat_flags, COMPAT);
4813 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4814 flags[1].compat_ro_flags, COMPAT_RO);
4818 ret = check_feature(fs_info, flags[0].incompat_flags,
4819 flags[1].incompat_flags, INCOMPAT);
4823 ret = mnt_want_write_file(file);
4827 trans = btrfs_start_transaction(root, 0);
4828 if (IS_ERR(trans)) {
4829 ret = PTR_ERR(trans);
4830 goto out_drop_write;
4833 spin_lock(&fs_info->super_lock);
4834 newflags = btrfs_super_compat_flags(super_block);
4835 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4836 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4837 btrfs_set_super_compat_flags(super_block, newflags);
4839 newflags = btrfs_super_compat_ro_flags(super_block);
4840 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4841 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4842 btrfs_set_super_compat_ro_flags(super_block, newflags);
4844 newflags = btrfs_super_incompat_flags(super_block);
4845 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4846 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4847 btrfs_set_super_incompat_flags(super_block, newflags);
4848 spin_unlock(&fs_info->super_lock);
4850 ret = btrfs_commit_transaction(trans);
4852 mnt_drop_write_file(file);
4857 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4859 struct btrfs_ioctl_send_args *arg;
4863 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4864 struct btrfs_ioctl_send_args_32 args32;
4866 ret = copy_from_user(&args32, argp, sizeof(args32));
4869 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4872 arg->send_fd = args32.send_fd;
4873 arg->clone_sources_count = args32.clone_sources_count;
4874 arg->clone_sources = compat_ptr(args32.clone_sources);
4875 arg->parent_root = args32.parent_root;
4876 arg->flags = args32.flags;
4877 memcpy(arg->reserved, args32.reserved,
4878 sizeof(args32.reserved));
4883 arg = memdup_user(argp, sizeof(*arg));
4885 return PTR_ERR(arg);
4887 ret = btrfs_ioctl_send(file, arg);
4892 long btrfs_ioctl(struct file *file, unsigned int
4893 cmd, unsigned long arg)
4895 struct inode *inode = file_inode(file);
4896 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4897 struct btrfs_root *root = BTRFS_I(inode)->root;
4898 void __user *argp = (void __user *)arg;
4901 case FS_IOC_GETVERSION:
4902 return btrfs_ioctl_getversion(file, argp);
4903 case FS_IOC_GETFSLABEL:
4904 return btrfs_ioctl_get_fslabel(fs_info, argp);
4905 case FS_IOC_SETFSLABEL:
4906 return btrfs_ioctl_set_fslabel(file, argp);
4908 return btrfs_ioctl_fitrim(fs_info, argp);
4909 case BTRFS_IOC_SNAP_CREATE:
4910 return btrfs_ioctl_snap_create(file, argp, 0);
4911 case BTRFS_IOC_SNAP_CREATE_V2:
4912 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4913 case BTRFS_IOC_SUBVOL_CREATE:
4914 return btrfs_ioctl_snap_create(file, argp, 1);
4915 case BTRFS_IOC_SUBVOL_CREATE_V2:
4916 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4917 case BTRFS_IOC_SNAP_DESTROY:
4918 return btrfs_ioctl_snap_destroy(file, argp, false);
4919 case BTRFS_IOC_SNAP_DESTROY_V2:
4920 return btrfs_ioctl_snap_destroy(file, argp, true);
4921 case BTRFS_IOC_SUBVOL_GETFLAGS:
4922 return btrfs_ioctl_subvol_getflags(file, argp);
4923 case BTRFS_IOC_SUBVOL_SETFLAGS:
4924 return btrfs_ioctl_subvol_setflags(file, argp);
4925 case BTRFS_IOC_DEFAULT_SUBVOL:
4926 return btrfs_ioctl_default_subvol(file, argp);
4927 case BTRFS_IOC_DEFRAG:
4928 return btrfs_ioctl_defrag(file, NULL);
4929 case BTRFS_IOC_DEFRAG_RANGE:
4930 return btrfs_ioctl_defrag(file, argp);
4931 case BTRFS_IOC_RESIZE:
4932 return btrfs_ioctl_resize(file, argp);
4933 case BTRFS_IOC_ADD_DEV:
4934 return btrfs_ioctl_add_dev(fs_info, argp);
4935 case BTRFS_IOC_RM_DEV:
4936 return btrfs_ioctl_rm_dev(file, argp);
4937 case BTRFS_IOC_RM_DEV_V2:
4938 return btrfs_ioctl_rm_dev_v2(file, argp);
4939 case BTRFS_IOC_FS_INFO:
4940 return btrfs_ioctl_fs_info(fs_info, argp);
4941 case BTRFS_IOC_DEV_INFO:
4942 return btrfs_ioctl_dev_info(fs_info, argp);
4943 case BTRFS_IOC_BALANCE:
4944 return btrfs_ioctl_balance(file, NULL);
4945 case BTRFS_IOC_TREE_SEARCH:
4946 return btrfs_ioctl_tree_search(file, argp);
4947 case BTRFS_IOC_TREE_SEARCH_V2:
4948 return btrfs_ioctl_tree_search_v2(file, argp);
4949 case BTRFS_IOC_INO_LOOKUP:
4950 return btrfs_ioctl_ino_lookup(file, argp);
4951 case BTRFS_IOC_INO_PATHS:
4952 return btrfs_ioctl_ino_to_path(root, argp);
4953 case BTRFS_IOC_LOGICAL_INO:
4954 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4955 case BTRFS_IOC_LOGICAL_INO_V2:
4956 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4957 case BTRFS_IOC_SPACE_INFO:
4958 return btrfs_ioctl_space_info(fs_info, argp);
4959 case BTRFS_IOC_SYNC: {
4962 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4965 ret = btrfs_sync_fs(inode->i_sb, 1);
4967 * The transaction thread may want to do more work,
4968 * namely it pokes the cleaner kthread that will start
4969 * processing uncleaned subvols.
4971 wake_up_process(fs_info->transaction_kthread);
4974 case BTRFS_IOC_START_SYNC:
4975 return btrfs_ioctl_start_sync(root, argp);
4976 case BTRFS_IOC_WAIT_SYNC:
4977 return btrfs_ioctl_wait_sync(fs_info, argp);
4978 case BTRFS_IOC_SCRUB:
4979 return btrfs_ioctl_scrub(file, argp);
4980 case BTRFS_IOC_SCRUB_CANCEL:
4981 return btrfs_ioctl_scrub_cancel(fs_info);
4982 case BTRFS_IOC_SCRUB_PROGRESS:
4983 return btrfs_ioctl_scrub_progress(fs_info, argp);
4984 case BTRFS_IOC_BALANCE_V2:
4985 return btrfs_ioctl_balance(file, argp);
4986 case BTRFS_IOC_BALANCE_CTL:
4987 return btrfs_ioctl_balance_ctl(fs_info, arg);
4988 case BTRFS_IOC_BALANCE_PROGRESS:
4989 return btrfs_ioctl_balance_progress(fs_info, argp);
4990 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4991 return btrfs_ioctl_set_received_subvol(file, argp);
4993 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4994 return btrfs_ioctl_set_received_subvol_32(file, argp);
4996 case BTRFS_IOC_SEND:
4997 return _btrfs_ioctl_send(file, argp, false);
4998 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4999 case BTRFS_IOC_SEND_32:
5000 return _btrfs_ioctl_send(file, argp, true);
5002 case BTRFS_IOC_GET_DEV_STATS:
5003 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5004 case BTRFS_IOC_QUOTA_CTL:
5005 return btrfs_ioctl_quota_ctl(file, argp);
5006 case BTRFS_IOC_QGROUP_ASSIGN:
5007 return btrfs_ioctl_qgroup_assign(file, argp);
5008 case BTRFS_IOC_QGROUP_CREATE:
5009 return btrfs_ioctl_qgroup_create(file, argp);
5010 case BTRFS_IOC_QGROUP_LIMIT:
5011 return btrfs_ioctl_qgroup_limit(file, argp);
5012 case BTRFS_IOC_QUOTA_RESCAN:
5013 return btrfs_ioctl_quota_rescan(file, argp);
5014 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5015 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5016 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5017 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5018 case BTRFS_IOC_DEV_REPLACE:
5019 return btrfs_ioctl_dev_replace(fs_info, argp);
5020 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5021 return btrfs_ioctl_get_supported_features(argp);
5022 case BTRFS_IOC_GET_FEATURES:
5023 return btrfs_ioctl_get_features(fs_info, argp);
5024 case BTRFS_IOC_SET_FEATURES:
5025 return btrfs_ioctl_set_features(file, argp);
5026 case BTRFS_IOC_GET_SUBVOL_INFO:
5027 return btrfs_ioctl_get_subvol_info(file, argp);
5028 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5029 return btrfs_ioctl_get_subvol_rootref(file, argp);
5030 case BTRFS_IOC_INO_LOOKUP_USER:
5031 return btrfs_ioctl_ino_lookup_user(file, argp);
5032 case FS_IOC_ENABLE_VERITY:
5033 return fsverity_ioctl_enable(file, (const void __user *)argp);
5034 case FS_IOC_MEASURE_VERITY:
5035 return fsverity_ioctl_measure(file, argp);
5041 #ifdef CONFIG_COMPAT
5042 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5045 * These all access 32-bit values anyway so no further
5046 * handling is necessary.
5049 case FS_IOC32_GETVERSION:
5050 cmd = FS_IOC_GETVERSION;
5054 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));