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 inode *dir,
503 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 without backreferences). Also no need to have
619 * the tree block locked since it is not in any tree at this
620 * point, so no other task can find it and use it.
622 btrfs_free_tree_block(trans, root, leaf, 0, 1);
623 free_extent_buffer(leaf);
627 free_extent_buffer(leaf);
630 key.offset = (u64)-1;
631 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
632 if (IS_ERR(new_root)) {
633 free_anon_bdev(anon_dev);
634 ret = PTR_ERR(new_root);
635 btrfs_abort_transaction(trans, ret);
638 /* Freeing will be done in btrfs_put_root() of new_root */
641 ret = btrfs_record_root_in_trans(trans, new_root);
643 btrfs_put_root(new_root);
644 btrfs_abort_transaction(trans, ret);
648 ret = btrfs_create_subvol_root(trans, new_root, root);
649 btrfs_put_root(new_root);
651 /* We potentially lose an unused inode item here */
652 btrfs_abort_transaction(trans, ret);
657 * insert the directory item
659 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
661 btrfs_abort_transaction(trans, ret);
665 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
666 BTRFS_FT_DIR, index);
668 btrfs_abort_transaction(trans, ret);
672 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
673 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
675 btrfs_abort_transaction(trans, ret);
679 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
680 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
682 btrfs_abort_transaction(trans, ret);
686 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
687 BTRFS_UUID_KEY_SUBVOL, objectid);
689 btrfs_abort_transaction(trans, ret);
693 trans->block_rsv = NULL;
694 trans->bytes_reserved = 0;
695 btrfs_subvolume_release_metadata(root, &block_rsv);
697 err = btrfs_commit_transaction(trans);
702 inode = btrfs_lookup_dentry(dir, dentry);
704 return PTR_ERR(inode);
705 d_instantiate(dentry, inode);
711 free_anon_bdev(anon_dev);
716 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
717 struct dentry *dentry, bool readonly,
718 struct btrfs_qgroup_inherit *inherit)
720 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
722 struct btrfs_pending_snapshot *pending_snapshot;
723 struct btrfs_trans_handle *trans;
726 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
729 if (atomic_read(&root->nr_swapfiles)) {
731 "cannot snapshot subvolume with active swapfile");
735 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
736 if (!pending_snapshot)
739 ret = get_anon_bdev(&pending_snapshot->anon_dev);
742 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
744 pending_snapshot->path = btrfs_alloc_path();
745 if (!pending_snapshot->root_item || !pending_snapshot->path) {
750 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
751 BTRFS_BLOCK_RSV_TEMP);
753 * 1 - parent dir inode
756 * 2 - root ref/backref
757 * 1 - root of snapshot
760 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
761 &pending_snapshot->block_rsv, 8,
766 pending_snapshot->dentry = dentry;
767 pending_snapshot->root = root;
768 pending_snapshot->readonly = readonly;
769 pending_snapshot->dir = dir;
770 pending_snapshot->inherit = inherit;
772 trans = btrfs_start_transaction(root, 0);
774 ret = PTR_ERR(trans);
778 spin_lock(&fs_info->trans_lock);
779 list_add(&pending_snapshot->list,
780 &trans->transaction->pending_snapshots);
781 spin_unlock(&fs_info->trans_lock);
783 ret = btrfs_commit_transaction(trans);
787 ret = pending_snapshot->error;
791 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
795 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
797 ret = PTR_ERR(inode);
801 d_instantiate(dentry, inode);
803 pending_snapshot->anon_dev = 0;
805 /* Prevent double freeing of anon_dev */
806 if (ret && pending_snapshot->snap)
807 pending_snapshot->snap->anon_dev = 0;
808 btrfs_put_root(pending_snapshot->snap);
809 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
811 if (pending_snapshot->anon_dev)
812 free_anon_bdev(pending_snapshot->anon_dev);
813 kfree(pending_snapshot->root_item);
814 btrfs_free_path(pending_snapshot->path);
815 kfree(pending_snapshot);
820 /* copy of may_delete in fs/namei.c()
821 * Check whether we can remove a link victim from directory dir, check
822 * whether the type of victim is right.
823 * 1. We can't do it if dir is read-only (done in permission())
824 * 2. We should have write and exec permissions on dir
825 * 3. We can't remove anything from append-only dir
826 * 4. We can't do anything with immutable dir (done in permission())
827 * 5. If the sticky bit on dir is set we should either
828 * a. be owner of dir, or
829 * b. be owner of victim, or
830 * c. have CAP_FOWNER capability
831 * 6. If the victim is append-only or immutable we can't do anything with
832 * links pointing to it.
833 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
834 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
835 * 9. We can't remove a root or mountpoint.
836 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
837 * nfs_async_unlink().
840 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
844 if (d_really_is_negative(victim))
847 BUG_ON(d_inode(victim->d_parent) != dir);
848 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
850 error = inode_permission(&init_user_ns, dir, MAY_WRITE | MAY_EXEC);
855 if (check_sticky(&init_user_ns, dir, d_inode(victim)) ||
856 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
857 IS_SWAPFILE(d_inode(victim)))
860 if (!d_is_dir(victim))
864 } else if (d_is_dir(victim))
868 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
873 /* copy of may_create in fs/namei.c() */
874 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
876 if (d_really_is_positive(child))
880 return inode_permission(&init_user_ns, dir, MAY_WRITE | MAY_EXEC);
884 * Create a new subvolume below @parent. This is largely modeled after
885 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
886 * inside this filesystem so it's quite a bit simpler.
888 static noinline int btrfs_mksubvol(const struct path *parent,
889 const char *name, int namelen,
890 struct btrfs_root *snap_src,
892 struct btrfs_qgroup_inherit *inherit)
894 struct inode *dir = d_inode(parent->dentry);
895 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
896 struct dentry *dentry;
899 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
903 dentry = lookup_one_len(name, parent->dentry, namelen);
904 error = PTR_ERR(dentry);
908 error = btrfs_may_create(dir, dentry);
913 * even if this name doesn't exist, we may get hash collisions.
914 * check for them now when we can safely fail
916 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
922 down_read(&fs_info->subvol_sem);
924 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
928 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
930 error = create_subvol(dir, dentry, name, namelen, inherit);
933 fsnotify_mkdir(dir, dentry);
935 up_read(&fs_info->subvol_sem);
939 btrfs_inode_unlock(dir, 0);
943 static noinline int btrfs_mksnapshot(const struct path *parent,
944 const char *name, int namelen,
945 struct btrfs_root *root,
947 struct btrfs_qgroup_inherit *inherit)
950 bool snapshot_force_cow = false;
953 * Force new buffered writes to reserve space even when NOCOW is
954 * possible. This is to avoid later writeback (running dealloc) to
955 * fallback to COW mode and unexpectedly fail with ENOSPC.
957 btrfs_drew_read_lock(&root->snapshot_lock);
959 ret = btrfs_start_delalloc_snapshot(root, false);
964 * All previous writes have started writeback in NOCOW mode, so now
965 * we force future writes to fallback to COW mode during snapshot
968 atomic_inc(&root->snapshot_force_cow);
969 snapshot_force_cow = true;
971 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
973 ret = btrfs_mksubvol(parent, name, namelen,
974 root, readonly, inherit);
976 if (snapshot_force_cow)
977 atomic_dec(&root->snapshot_force_cow);
978 btrfs_drew_read_unlock(&root->snapshot_lock);
983 * When we're defragging a range, we don't want to kick it off again
984 * if it is really just waiting for delalloc to send it down.
985 * If we find a nice big extent or delalloc range for the bytes in the
986 * file you want to defrag, we return 0 to let you know to skip this
989 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
991 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
992 struct extent_map *em = NULL;
993 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
996 read_lock(&em_tree->lock);
997 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
998 read_unlock(&em_tree->lock);
1001 end = extent_map_end(em);
1002 free_extent_map(em);
1003 if (end - offset > thresh)
1006 /* if we already have a nice delalloc here, just stop */
1008 end = count_range_bits(io_tree, &offset, offset + thresh,
1009 thresh, EXTENT_DELALLOC, 1);
1016 * helper function to walk through a file and find extents
1017 * newer than a specific transid, and smaller than thresh.
1019 * This is used by the defragging code to find new and small
1022 static int find_new_extents(struct btrfs_root *root,
1023 struct inode *inode, u64 newer_than,
1024 u64 *off, u32 thresh)
1026 struct btrfs_path *path;
1027 struct btrfs_key min_key;
1028 struct extent_buffer *leaf;
1029 struct btrfs_file_extent_item *extent;
1032 u64 ino = btrfs_ino(BTRFS_I(inode));
1034 path = btrfs_alloc_path();
1038 min_key.objectid = ino;
1039 min_key.type = BTRFS_EXTENT_DATA_KEY;
1040 min_key.offset = *off;
1043 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1047 if (min_key.objectid != ino)
1049 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1052 leaf = path->nodes[0];
1053 extent = btrfs_item_ptr(leaf, path->slots[0],
1054 struct btrfs_file_extent_item);
1056 type = btrfs_file_extent_type(leaf, extent);
1057 if (type == BTRFS_FILE_EXTENT_REG &&
1058 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1059 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1060 *off = min_key.offset;
1061 btrfs_free_path(path);
1066 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1067 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1071 if (min_key.offset == (u64)-1)
1075 btrfs_release_path(path);
1078 btrfs_free_path(path);
1082 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1084 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1085 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1086 struct extent_map *em;
1087 u64 len = PAGE_SIZE;
1090 * hopefully we have this extent in the tree already, try without
1091 * the full extent lock
1093 read_lock(&em_tree->lock);
1094 em = lookup_extent_mapping(em_tree, start, len);
1095 read_unlock(&em_tree->lock);
1098 struct extent_state *cached = NULL;
1099 u64 end = start + len - 1;
1101 /* get the big lock and read metadata off disk */
1102 lock_extent_bits(io_tree, start, end, &cached);
1103 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1104 unlock_extent_cached(io_tree, start, end, &cached);
1113 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1115 struct extent_map *next;
1118 /* this is the last extent */
1119 if (em->start + em->len >= i_size_read(inode))
1122 next = defrag_lookup_extent(inode, em->start + em->len);
1123 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1125 else if ((em->block_start + em->block_len == next->block_start) &&
1126 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1129 free_extent_map(next);
1133 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1134 u64 *last_len, u64 *skip, u64 *defrag_end,
1137 struct extent_map *em;
1139 bool next_mergeable = true;
1140 bool prev_mergeable = true;
1143 * make sure that once we start defragging an extent, we keep on
1146 if (start < *defrag_end)
1151 em = defrag_lookup_extent(inode, start);
1155 /* this will cover holes, and inline extents */
1156 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1162 prev_mergeable = false;
1164 next_mergeable = defrag_check_next_extent(inode, em);
1166 * we hit a real extent, if it is big or the next extent is not a
1167 * real extent, don't bother defragging it
1169 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1170 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1174 * last_len ends up being a counter of how many bytes we've defragged.
1175 * every time we choose not to defrag an extent, we reset *last_len
1176 * so that the next tiny extent will force a defrag.
1178 * The end result of this is that tiny extents before a single big
1179 * extent will force at least part of that big extent to be defragged.
1182 *defrag_end = extent_map_end(em);
1185 *skip = extent_map_end(em);
1189 free_extent_map(em);
1194 * it doesn't do much good to defrag one or two pages
1195 * at a time. This pulls in a nice chunk of pages
1196 * to COW and defrag.
1198 * It also makes sure the delalloc code has enough
1199 * dirty data to avoid making new small extents as part
1202 * It's a good idea to start RA on this range
1203 * before calling this.
1205 static int cluster_pages_for_defrag(struct inode *inode,
1206 struct page **pages,
1207 unsigned long start_index,
1208 unsigned long num_pages)
1210 unsigned long file_end;
1211 u64 isize = i_size_read(inode);
1215 u64 start = (u64)start_index << PAGE_SHIFT;
1220 struct btrfs_ordered_extent *ordered;
1221 struct extent_state *cached_state = NULL;
1222 struct extent_io_tree *tree;
1223 struct extent_changeset *data_reserved = NULL;
1224 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1226 file_end = (isize - 1) >> PAGE_SHIFT;
1227 if (!isize || start_index > file_end)
1230 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1232 ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
1233 start, page_cnt << PAGE_SHIFT);
1237 tree = &BTRFS_I(inode)->io_tree;
1239 /* step one, lock all the pages */
1240 for (i = 0; i < page_cnt; i++) {
1243 page = find_or_create_page(inode->i_mapping,
1244 start_index + i, mask);
1248 ret = set_page_extent_mapped(page);
1255 page_start = page_offset(page);
1256 page_end = page_start + PAGE_SIZE - 1;
1258 lock_extent_bits(tree, page_start, page_end,
1260 ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode),
1262 unlock_extent_cached(tree, page_start, page_end,
1268 btrfs_start_ordered_extent(ordered, 1);
1269 btrfs_put_ordered_extent(ordered);
1272 * we unlocked the page above, so we need check if
1273 * it was released or not.
1275 if (page->mapping != inode->i_mapping) {
1282 if (!PageUptodate(page)) {
1283 btrfs_readpage(NULL, page);
1285 if (!PageUptodate(page)) {
1293 if (page->mapping != inode->i_mapping) {
1305 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1309 * so now we have a nice long stream of locked
1310 * and up to date pages, lets wait on them
1312 for (i = 0; i < i_done; i++)
1313 wait_on_page_writeback(pages[i]);
1315 page_start = page_offset(pages[0]);
1316 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1318 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1319 page_start, page_end - 1, &cached_state);
1322 * When defragmenting we skip ranges that have holes or inline extents,
1323 * (check should_defrag_range()), to avoid unnecessary IO and wasting
1324 * space. At btrfs_defrag_file(), we check if a range should be defragged
1325 * before locking the inode and then, if it should, we trigger a sync
1326 * page cache readahead - we lock the inode only after that to avoid
1327 * blocking for too long other tasks that possibly want to operate on
1328 * other file ranges. But before we were able to get the inode lock,
1329 * some other task may have punched a hole in the range, or we may have
1330 * now an inline extent, in which case we should not defrag. So check
1331 * for that here, where we have the inode and the range locked, and bail
1332 * out if that happened.
1334 search_start = page_start;
1335 while (search_start < page_end) {
1336 struct extent_map *em;
1338 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, search_start,
1339 page_end - search_start);
1342 goto out_unlock_range;
1344 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1345 free_extent_map(em);
1346 /* Ok, 0 means we did not defrag anything */
1348 goto out_unlock_range;
1350 search_start = extent_map_end(em);
1351 free_extent_map(em);
1354 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1355 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1356 EXTENT_DEFRAG, 0, 0, &cached_state);
1358 if (i_done != page_cnt) {
1359 spin_lock(&BTRFS_I(inode)->lock);
1360 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1361 spin_unlock(&BTRFS_I(inode)->lock);
1362 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1363 start, (page_cnt - i_done) << PAGE_SHIFT, true);
1367 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1370 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1371 page_start, page_end - 1, &cached_state);
1373 for (i = 0; i < i_done; i++) {
1374 clear_page_dirty_for_io(pages[i]);
1375 ClearPageChecked(pages[i]);
1376 set_page_dirty(pages[i]);
1377 unlock_page(pages[i]);
1380 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1381 extent_changeset_free(data_reserved);
1385 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1386 page_start, page_end - 1, &cached_state);
1388 for (i = 0; i < i_done; i++) {
1389 unlock_page(pages[i]);
1392 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1393 start, page_cnt << PAGE_SHIFT, true);
1394 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1395 extent_changeset_free(data_reserved);
1400 int btrfs_defrag_file(struct inode *inode, struct file *file,
1401 struct btrfs_ioctl_defrag_range_args *range,
1402 u64 newer_than, unsigned long max_to_defrag)
1404 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1405 struct btrfs_root *root = BTRFS_I(inode)->root;
1406 struct file_ra_state *ra = NULL;
1407 unsigned long last_index;
1408 u64 isize = i_size_read(inode);
1412 u64 newer_off = range->start;
1414 unsigned long ra_index = 0;
1416 int defrag_count = 0;
1417 int compress_type = BTRFS_COMPRESS_ZLIB;
1418 u32 extent_thresh = range->extent_thresh;
1419 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1420 unsigned long cluster = max_cluster;
1421 u64 new_align = ~((u64)SZ_128K - 1);
1422 struct page **pages = NULL;
1423 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1428 if (range->start >= isize)
1432 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1434 if (range->compress_type)
1435 compress_type = range->compress_type;
1438 if (extent_thresh == 0)
1439 extent_thresh = SZ_256K;
1442 * If we were not given a file, allocate a readahead context. As
1443 * readahead is just an optimization, defrag will work without it so
1444 * we don't error out.
1447 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1449 file_ra_state_init(ra, inode->i_mapping);
1454 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1460 /* find the last page to defrag */
1461 if (range->start + range->len > range->start) {
1462 last_index = min_t(u64, isize - 1,
1463 range->start + range->len - 1) >> PAGE_SHIFT;
1465 last_index = (isize - 1) >> PAGE_SHIFT;
1469 ret = find_new_extents(root, inode, newer_than,
1470 &newer_off, SZ_64K);
1472 range->start = newer_off;
1474 * we always align our defrag to help keep
1475 * the extents in the file evenly spaced
1477 i = (newer_off & new_align) >> PAGE_SHIFT;
1481 i = range->start >> PAGE_SHIFT;
1484 max_to_defrag = last_index - i + 1;
1487 * make writeback starts from i, so the defrag range can be
1488 * written sequentially.
1490 if (i < inode->i_mapping->writeback_index)
1491 inode->i_mapping->writeback_index = i;
1493 while (i <= last_index && defrag_count < max_to_defrag &&
1494 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1496 * make sure we stop running if someone unmounts
1499 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1502 if (btrfs_defrag_cancelled(fs_info)) {
1503 btrfs_debug(fs_info, "defrag_file cancelled");
1508 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1509 extent_thresh, &last_len, &skip,
1510 &defrag_end, do_compress)){
1513 * the should_defrag function tells us how much to skip
1514 * bump our counter by the suggested amount
1516 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1517 i = max(i + 1, next);
1522 cluster = (PAGE_ALIGN(defrag_end) >>
1524 cluster = min(cluster, max_cluster);
1526 cluster = max_cluster;
1529 if (i + cluster > ra_index) {
1530 ra_index = max(i, ra_index);
1532 page_cache_sync_readahead(inode->i_mapping, ra,
1533 file, ra_index, cluster);
1534 ra_index += cluster;
1537 btrfs_inode_lock(inode, 0);
1538 if (IS_SWAPFILE(inode)) {
1542 BTRFS_I(inode)->defrag_compress = compress_type;
1543 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1546 btrfs_inode_unlock(inode, 0);
1550 defrag_count += ret;
1551 balance_dirty_pages_ratelimited(inode->i_mapping);
1552 btrfs_inode_unlock(inode, 0);
1555 if (newer_off == (u64)-1)
1561 newer_off = max(newer_off + 1,
1562 (u64)i << PAGE_SHIFT);
1564 ret = find_new_extents(root, inode, newer_than,
1565 &newer_off, SZ_64K);
1567 range->start = newer_off;
1568 i = (newer_off & new_align) >> PAGE_SHIFT;
1575 last_len += ret << PAGE_SHIFT;
1585 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1586 filemap_flush(inode->i_mapping);
1587 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1588 &BTRFS_I(inode)->runtime_flags))
1589 filemap_flush(inode->i_mapping);
1592 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1593 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1594 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1595 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1600 btrfs_inode_lock(inode, 0);
1601 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1602 btrfs_inode_unlock(inode, 0);
1611 * Try to start exclusive operation @type or cancel it if it's running.
1614 * 0 - normal mode, newly claimed op started
1615 * >0 - normal mode, something else is running,
1616 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1617 * ECANCELED - cancel mode, successful cancel
1618 * ENOTCONN - cancel mode, operation not running anymore
1620 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1621 enum btrfs_exclusive_operation type, bool cancel)
1624 /* Start normal op */
1625 if (!btrfs_exclop_start(fs_info, type))
1626 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1627 /* Exclusive operation is now claimed */
1631 /* Cancel running op */
1632 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1634 * This blocks any exclop finish from setting it to NONE, so we
1635 * request cancellation. Either it runs and we will wait for it,
1636 * or it has finished and no waiting will happen.
1638 atomic_inc(&fs_info->reloc_cancel_req);
1639 btrfs_exclop_start_unlock(fs_info);
1641 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1642 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1643 TASK_INTERRUPTIBLE);
1648 /* Something else is running or none */
1652 static noinline int btrfs_ioctl_resize(struct file *file,
1655 struct inode *inode = file_inode(file);
1656 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1660 struct btrfs_root *root = BTRFS_I(inode)->root;
1661 struct btrfs_ioctl_vol_args *vol_args;
1662 struct btrfs_trans_handle *trans;
1663 struct btrfs_device *device = NULL;
1666 char *devstr = NULL;
1671 if (!capable(CAP_SYS_ADMIN))
1674 ret = mnt_want_write_file(file);
1679 * Read the arguments before checking exclusivity to be able to
1680 * distinguish regular resize and cancel
1682 vol_args = memdup_user(arg, sizeof(*vol_args));
1683 if (IS_ERR(vol_args)) {
1684 ret = PTR_ERR(vol_args);
1687 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1688 sizestr = vol_args->name;
1689 cancel = (strcmp("cancel", sizestr) == 0);
1690 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1693 /* Exclusive operation is now claimed */
1695 devstr = strchr(sizestr, ':');
1697 sizestr = devstr + 1;
1699 devstr = vol_args->name;
1700 ret = kstrtoull(devstr, 10, &devid);
1707 btrfs_info(fs_info, "resizing devid %llu", devid);
1710 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL);
1712 btrfs_info(fs_info, "resizer unable to find device %llu",
1718 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1720 "resizer unable to apply on readonly device %llu",
1726 if (!strcmp(sizestr, "max"))
1727 new_size = device->bdev->bd_inode->i_size;
1729 if (sizestr[0] == '-') {
1732 } else if (sizestr[0] == '+') {
1736 new_size = memparse(sizestr, &retptr);
1737 if (*retptr != '\0' || new_size == 0) {
1743 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1748 old_size = btrfs_device_get_total_bytes(device);
1751 if (new_size > old_size) {
1755 new_size = old_size - new_size;
1756 } else if (mod > 0) {
1757 if (new_size > ULLONG_MAX - old_size) {
1761 new_size = old_size + new_size;
1764 if (new_size < SZ_256M) {
1768 if (new_size > device->bdev->bd_inode->i_size) {
1773 new_size = round_down(new_size, fs_info->sectorsize);
1775 if (new_size > old_size) {
1776 trans = btrfs_start_transaction(root, 0);
1777 if (IS_ERR(trans)) {
1778 ret = PTR_ERR(trans);
1781 ret = btrfs_grow_device(trans, device, new_size);
1782 btrfs_commit_transaction(trans);
1783 } else if (new_size < old_size) {
1784 ret = btrfs_shrink_device(device, new_size);
1785 } /* equal, nothing need to do */
1787 if (ret == 0 && new_size != old_size)
1788 btrfs_info_in_rcu(fs_info,
1789 "resize device %s (devid %llu) from %llu to %llu",
1790 rcu_str_deref(device->name), device->devid,
1791 old_size, new_size);
1793 btrfs_exclop_finish(fs_info);
1797 mnt_drop_write_file(file);
1801 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1802 const char *name, unsigned long fd, int subvol,
1804 struct btrfs_qgroup_inherit *inherit)
1809 if (!S_ISDIR(file_inode(file)->i_mode))
1812 ret = mnt_want_write_file(file);
1816 namelen = strlen(name);
1817 if (strchr(name, '/')) {
1819 goto out_drop_write;
1822 if (name[0] == '.' &&
1823 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1825 goto out_drop_write;
1829 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1830 NULL, readonly, inherit);
1832 struct fd src = fdget(fd);
1833 struct inode *src_inode;
1836 goto out_drop_write;
1839 src_inode = file_inode(src.file);
1840 if (src_inode->i_sb != file_inode(file)->i_sb) {
1841 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1842 "Snapshot src from another FS");
1844 } else if (!inode_owner_or_capable(&init_user_ns, src_inode)) {
1846 * Subvolume creation is not restricted, but snapshots
1847 * are limited to own subvolumes only
1851 ret = btrfs_mksnapshot(&file->f_path, name, namelen,
1852 BTRFS_I(src_inode)->root,
1858 mnt_drop_write_file(file);
1863 static noinline int btrfs_ioctl_snap_create(struct file *file,
1864 void __user *arg, int subvol)
1866 struct btrfs_ioctl_vol_args *vol_args;
1869 if (!S_ISDIR(file_inode(file)->i_mode))
1872 vol_args = memdup_user(arg, sizeof(*vol_args));
1873 if (IS_ERR(vol_args))
1874 return PTR_ERR(vol_args);
1875 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1877 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1878 subvol, false, NULL);
1884 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1885 void __user *arg, int subvol)
1887 struct btrfs_ioctl_vol_args_v2 *vol_args;
1889 bool readonly = false;
1890 struct btrfs_qgroup_inherit *inherit = NULL;
1892 if (!S_ISDIR(file_inode(file)->i_mode))
1895 vol_args = memdup_user(arg, sizeof(*vol_args));
1896 if (IS_ERR(vol_args))
1897 return PTR_ERR(vol_args);
1898 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1900 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1905 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1907 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1910 if (vol_args->size < sizeof(*inherit) ||
1911 vol_args->size > PAGE_SIZE) {
1915 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1916 if (IS_ERR(inherit)) {
1917 ret = PTR_ERR(inherit);
1921 if (inherit->num_qgroups > PAGE_SIZE ||
1922 inherit->num_ref_copies > PAGE_SIZE ||
1923 inherit->num_excl_copies > PAGE_SIZE) {
1928 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1929 2 * inherit->num_excl_copies;
1930 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1936 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1937 subvol, readonly, inherit);
1947 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1950 struct inode *inode = file_inode(file);
1951 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1952 struct btrfs_root *root = BTRFS_I(inode)->root;
1956 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1959 down_read(&fs_info->subvol_sem);
1960 if (btrfs_root_readonly(root))
1961 flags |= BTRFS_SUBVOL_RDONLY;
1962 up_read(&fs_info->subvol_sem);
1964 if (copy_to_user(arg, &flags, sizeof(flags)))
1970 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1973 struct inode *inode = file_inode(file);
1974 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1975 struct btrfs_root *root = BTRFS_I(inode)->root;
1976 struct btrfs_trans_handle *trans;
1981 if (!inode_owner_or_capable(&init_user_ns, inode))
1984 ret = mnt_want_write_file(file);
1988 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1990 goto out_drop_write;
1993 if (copy_from_user(&flags, arg, sizeof(flags))) {
1995 goto out_drop_write;
1998 if (flags & ~BTRFS_SUBVOL_RDONLY) {
2000 goto out_drop_write;
2003 down_write(&fs_info->subvol_sem);
2006 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2009 root_flags = btrfs_root_flags(&root->root_item);
2010 if (flags & BTRFS_SUBVOL_RDONLY) {
2011 btrfs_set_root_flags(&root->root_item,
2012 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2015 * Block RO -> RW transition if this subvolume is involved in
2018 spin_lock(&root->root_item_lock);
2019 if (root->send_in_progress == 0) {
2020 btrfs_set_root_flags(&root->root_item,
2021 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2022 spin_unlock(&root->root_item_lock);
2024 spin_unlock(&root->root_item_lock);
2026 "Attempt to set subvolume %llu read-write during send",
2027 root->root_key.objectid);
2033 trans = btrfs_start_transaction(root, 1);
2034 if (IS_ERR(trans)) {
2035 ret = PTR_ERR(trans);
2039 ret = btrfs_update_root(trans, fs_info->tree_root,
2040 &root->root_key, &root->root_item);
2042 btrfs_end_transaction(trans);
2046 ret = btrfs_commit_transaction(trans);
2050 btrfs_set_root_flags(&root->root_item, root_flags);
2052 up_write(&fs_info->subvol_sem);
2054 mnt_drop_write_file(file);
2059 static noinline int key_in_sk(struct btrfs_key *key,
2060 struct btrfs_ioctl_search_key *sk)
2062 struct btrfs_key test;
2065 test.objectid = sk->min_objectid;
2066 test.type = sk->min_type;
2067 test.offset = sk->min_offset;
2069 ret = btrfs_comp_cpu_keys(key, &test);
2073 test.objectid = sk->max_objectid;
2074 test.type = sk->max_type;
2075 test.offset = sk->max_offset;
2077 ret = btrfs_comp_cpu_keys(key, &test);
2083 static noinline int copy_to_sk(struct btrfs_path *path,
2084 struct btrfs_key *key,
2085 struct btrfs_ioctl_search_key *sk,
2088 unsigned long *sk_offset,
2092 struct extent_buffer *leaf;
2093 struct btrfs_ioctl_search_header sh;
2094 struct btrfs_key test;
2095 unsigned long item_off;
2096 unsigned long item_len;
2102 leaf = path->nodes[0];
2103 slot = path->slots[0];
2104 nritems = btrfs_header_nritems(leaf);
2106 if (btrfs_header_generation(leaf) > sk->max_transid) {
2110 found_transid = btrfs_header_generation(leaf);
2112 for (i = slot; i < nritems; i++) {
2113 item_off = btrfs_item_ptr_offset(leaf, i);
2114 item_len = btrfs_item_size_nr(leaf, i);
2116 btrfs_item_key_to_cpu(leaf, key, i);
2117 if (!key_in_sk(key, sk))
2120 if (sizeof(sh) + item_len > *buf_size) {
2127 * return one empty item back for v1, which does not
2131 *buf_size = sizeof(sh) + item_len;
2136 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2141 sh.objectid = key->objectid;
2142 sh.offset = key->offset;
2143 sh.type = key->type;
2145 sh.transid = found_transid;
2148 * Copy search result header. If we fault then loop again so we
2149 * can fault in the pages and -EFAULT there if there's a
2150 * problem. Otherwise we'll fault and then copy the buffer in
2151 * properly this next time through
2153 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2158 *sk_offset += sizeof(sh);
2161 char __user *up = ubuf + *sk_offset;
2163 * Copy the item, same behavior as above, but reset the
2164 * * sk_offset so we copy the full thing again.
2166 if (read_extent_buffer_to_user_nofault(leaf, up,
2167 item_off, item_len)) {
2169 *sk_offset -= sizeof(sh);
2173 *sk_offset += item_len;
2177 if (ret) /* -EOVERFLOW from above */
2180 if (*num_found >= sk->nr_items) {
2187 test.objectid = sk->max_objectid;
2188 test.type = sk->max_type;
2189 test.offset = sk->max_offset;
2190 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2192 else if (key->offset < (u64)-1)
2194 else if (key->type < (u8)-1) {
2197 } else if (key->objectid < (u64)-1) {
2205 * 0: all items from this leaf copied, continue with next
2206 * 1: * more items can be copied, but unused buffer is too small
2207 * * all items were found
2208 * Either way, it will stops the loop which iterates to the next
2210 * -EOVERFLOW: item was to large for buffer
2211 * -EFAULT: could not copy extent buffer back to userspace
2216 static noinline int search_ioctl(struct inode *inode,
2217 struct btrfs_ioctl_search_key *sk,
2221 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2222 struct btrfs_root *root;
2223 struct btrfs_key key;
2224 struct btrfs_path *path;
2227 unsigned long sk_offset = 0;
2229 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2230 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2234 path = btrfs_alloc_path();
2238 if (sk->tree_id == 0) {
2239 /* search the root of the inode that was passed */
2240 root = btrfs_grab_root(BTRFS_I(inode)->root);
2242 root = btrfs_get_fs_root(info, sk->tree_id, true);
2244 btrfs_free_path(path);
2245 return PTR_ERR(root);
2249 key.objectid = sk->min_objectid;
2250 key.type = sk->min_type;
2251 key.offset = sk->min_offset;
2254 ret = fault_in_pages_writeable(ubuf + sk_offset,
2255 *buf_size - sk_offset);
2259 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2265 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2266 &sk_offset, &num_found);
2267 btrfs_release_path(path);
2275 sk->nr_items = num_found;
2276 btrfs_put_root(root);
2277 btrfs_free_path(path);
2281 static noinline int btrfs_ioctl_tree_search(struct file *file,
2284 struct btrfs_ioctl_search_args __user *uargs;
2285 struct btrfs_ioctl_search_key sk;
2286 struct inode *inode;
2290 if (!capable(CAP_SYS_ADMIN))
2293 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2295 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2298 buf_size = sizeof(uargs->buf);
2300 inode = file_inode(file);
2301 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2304 * In the origin implementation an overflow is handled by returning a
2305 * search header with a len of zero, so reset ret.
2307 if (ret == -EOVERFLOW)
2310 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2315 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2318 struct btrfs_ioctl_search_args_v2 __user *uarg;
2319 struct btrfs_ioctl_search_args_v2 args;
2320 struct inode *inode;
2323 const size_t buf_limit = SZ_16M;
2325 if (!capable(CAP_SYS_ADMIN))
2328 /* copy search header and buffer size */
2329 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2330 if (copy_from_user(&args, uarg, sizeof(args)))
2333 buf_size = args.buf_size;
2335 /* limit result size to 16MB */
2336 if (buf_size > buf_limit)
2337 buf_size = buf_limit;
2339 inode = file_inode(file);
2340 ret = search_ioctl(inode, &args.key, &buf_size,
2341 (char __user *)(&uarg->buf[0]));
2342 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2344 else if (ret == -EOVERFLOW &&
2345 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2352 * Search INODE_REFs to identify path name of 'dirid' directory
2353 * in a 'tree_id' tree. and sets path name to 'name'.
2355 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2356 u64 tree_id, u64 dirid, char *name)
2358 struct btrfs_root *root;
2359 struct btrfs_key key;
2365 struct btrfs_inode_ref *iref;
2366 struct extent_buffer *l;
2367 struct btrfs_path *path;
2369 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2374 path = btrfs_alloc_path();
2378 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2380 root = btrfs_get_fs_root(info, tree_id, true);
2382 ret = PTR_ERR(root);
2387 key.objectid = dirid;
2388 key.type = BTRFS_INODE_REF_KEY;
2389 key.offset = (u64)-1;
2392 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2396 ret = btrfs_previous_item(root, path, dirid,
2397 BTRFS_INODE_REF_KEY);
2407 slot = path->slots[0];
2408 btrfs_item_key_to_cpu(l, &key, slot);
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 inode *inode,
2440 struct btrfs_ioctl_ino_lookup_user_args *args)
2442 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2443 struct super_block *sb = inode->i_sb;
2444 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2445 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2446 u64 dirid = args->dirid;
2447 unsigned long item_off;
2448 unsigned long item_len;
2449 struct btrfs_inode_ref *iref;
2450 struct btrfs_root_ref *rref;
2451 struct btrfs_root *root = NULL;
2452 struct btrfs_path *path;
2453 struct btrfs_key key, key2;
2454 struct extent_buffer *leaf;
2455 struct inode *temp_inode;
2462 path = btrfs_alloc_path();
2467 * If the bottom subvolume does not exist directly under upper_limit,
2468 * construct the path in from the bottom up.
2470 if (dirid != upper_limit.objectid) {
2471 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2473 root = btrfs_get_fs_root(fs_info, treeid, true);
2475 ret = PTR_ERR(root);
2479 key.objectid = dirid;
2480 key.type = BTRFS_INODE_REF_KEY;
2481 key.offset = (u64)-1;
2483 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2486 } else if (ret > 0) {
2487 ret = btrfs_previous_item(root, path, dirid,
2488 BTRFS_INODE_REF_KEY);
2491 } else if (ret > 0) {
2497 leaf = path->nodes[0];
2498 slot = path->slots[0];
2499 btrfs_item_key_to_cpu(leaf, &key, slot);
2501 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2502 len = btrfs_inode_ref_name_len(leaf, iref);
2504 total_len += len + 1;
2505 if (ptr < args->path) {
2506 ret = -ENAMETOOLONG;
2511 read_extent_buffer(leaf, ptr,
2512 (unsigned long)(iref + 1), len);
2514 /* Check the read+exec permission of this directory */
2515 ret = btrfs_previous_item(root, path, dirid,
2516 BTRFS_INODE_ITEM_KEY);
2519 } else if (ret > 0) {
2524 leaf = path->nodes[0];
2525 slot = path->slots[0];
2526 btrfs_item_key_to_cpu(leaf, &key2, slot);
2527 if (key2.objectid != dirid) {
2532 temp_inode = btrfs_iget(sb, key2.objectid, root);
2533 if (IS_ERR(temp_inode)) {
2534 ret = PTR_ERR(temp_inode);
2537 ret = inode_permission(&init_user_ns, temp_inode,
2538 MAY_READ | MAY_EXEC);
2545 if (key.offset == upper_limit.objectid)
2547 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2552 btrfs_release_path(path);
2553 key.objectid = key.offset;
2554 key.offset = (u64)-1;
2555 dirid = key.objectid;
2558 memmove(args->path, ptr, total_len);
2559 args->path[total_len] = '\0';
2560 btrfs_put_root(root);
2562 btrfs_release_path(path);
2565 /* Get the bottom subvolume's name from ROOT_REF */
2566 key.objectid = treeid;
2567 key.type = BTRFS_ROOT_REF_KEY;
2568 key.offset = args->treeid;
2569 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2572 } else if (ret > 0) {
2577 leaf = path->nodes[0];
2578 slot = path->slots[0];
2579 btrfs_item_key_to_cpu(leaf, &key, slot);
2581 item_off = btrfs_item_ptr_offset(leaf, slot);
2582 item_len = btrfs_item_size_nr(leaf, slot);
2583 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2584 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2585 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2590 /* Copy subvolume's name */
2591 item_off += sizeof(struct btrfs_root_ref);
2592 item_len -= sizeof(struct btrfs_root_ref);
2593 read_extent_buffer(leaf, args->name, item_off, item_len);
2594 args->name[item_len] = 0;
2597 btrfs_put_root(root);
2599 btrfs_free_path(path);
2603 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2606 struct btrfs_ioctl_ino_lookup_args *args;
2607 struct inode *inode;
2610 args = memdup_user(argp, sizeof(*args));
2612 return PTR_ERR(args);
2614 inode = file_inode(file);
2617 * Unprivileged query to obtain the containing subvolume root id. The
2618 * path is reset so it's consistent with btrfs_search_path_in_tree.
2620 if (args->treeid == 0)
2621 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2623 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2628 if (!capable(CAP_SYS_ADMIN)) {
2633 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2634 args->treeid, args->objectid,
2638 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2646 * Version of ino_lookup ioctl (unprivileged)
2648 * The main differences from ino_lookup ioctl are:
2650 * 1. Read + Exec permission will be checked using inode_permission() during
2651 * path construction. -EACCES will be returned in case of failure.
2652 * 2. Path construction will be stopped at the inode number which corresponds
2653 * to the fd with which this ioctl is called. If constructed path does not
2654 * exist under fd's inode, -EACCES will be returned.
2655 * 3. The name of bottom subvolume is also searched and filled.
2657 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2659 struct btrfs_ioctl_ino_lookup_user_args *args;
2660 struct inode *inode;
2663 args = memdup_user(argp, sizeof(*args));
2665 return PTR_ERR(args);
2667 inode = file_inode(file);
2669 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2670 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2672 * The subvolume does not exist under fd with which this is
2679 ret = btrfs_search_path_in_tree_user(inode, args);
2681 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2688 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2689 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2691 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2692 struct btrfs_fs_info *fs_info;
2693 struct btrfs_root *root;
2694 struct btrfs_path *path;
2695 struct btrfs_key key;
2696 struct btrfs_root_item *root_item;
2697 struct btrfs_root_ref *rref;
2698 struct extent_buffer *leaf;
2699 unsigned long item_off;
2700 unsigned long item_len;
2701 struct inode *inode;
2705 path = btrfs_alloc_path();
2709 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2711 btrfs_free_path(path);
2715 inode = file_inode(file);
2716 fs_info = BTRFS_I(inode)->root->fs_info;
2718 /* Get root_item of inode's subvolume */
2719 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2720 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2722 ret = PTR_ERR(root);
2725 root_item = &root->root_item;
2727 subvol_info->treeid = key.objectid;
2729 subvol_info->generation = btrfs_root_generation(root_item);
2730 subvol_info->flags = btrfs_root_flags(root_item);
2732 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2733 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2735 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2738 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2739 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2740 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2742 subvol_info->otransid = btrfs_root_otransid(root_item);
2743 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2744 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2746 subvol_info->stransid = btrfs_root_stransid(root_item);
2747 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2748 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2750 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2751 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2752 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2754 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2755 /* Search root tree for ROOT_BACKREF of this subvolume */
2756 key.type = BTRFS_ROOT_BACKREF_KEY;
2758 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2761 } else if (path->slots[0] >=
2762 btrfs_header_nritems(path->nodes[0])) {
2763 ret = btrfs_next_leaf(fs_info->tree_root, path);
2766 } else if (ret > 0) {
2772 leaf = path->nodes[0];
2773 slot = path->slots[0];
2774 btrfs_item_key_to_cpu(leaf, &key, slot);
2775 if (key.objectid == subvol_info->treeid &&
2776 key.type == BTRFS_ROOT_BACKREF_KEY) {
2777 subvol_info->parent_id = key.offset;
2779 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2780 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2782 item_off = btrfs_item_ptr_offset(leaf, slot)
2783 + sizeof(struct btrfs_root_ref);
2784 item_len = btrfs_item_size_nr(leaf, slot)
2785 - sizeof(struct btrfs_root_ref);
2786 read_extent_buffer(leaf, subvol_info->name,
2787 item_off, item_len);
2794 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2798 btrfs_put_root(root);
2800 btrfs_free_path(path);
2806 * Return ROOT_REF information of the subvolume containing this inode
2807 * except the subvolume name.
2809 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2811 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2812 struct btrfs_root_ref *rref;
2813 struct btrfs_root *root;
2814 struct btrfs_path *path;
2815 struct btrfs_key key;
2816 struct extent_buffer *leaf;
2817 struct inode *inode;
2823 path = btrfs_alloc_path();
2827 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2828 if (IS_ERR(rootrefs)) {
2829 btrfs_free_path(path);
2830 return PTR_ERR(rootrefs);
2833 inode = file_inode(file);
2834 root = BTRFS_I(inode)->root->fs_info->tree_root;
2835 objectid = BTRFS_I(inode)->root->root_key.objectid;
2837 key.objectid = objectid;
2838 key.type = BTRFS_ROOT_REF_KEY;
2839 key.offset = rootrefs->min_treeid;
2842 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2845 } else if (path->slots[0] >=
2846 btrfs_header_nritems(path->nodes[0])) {
2847 ret = btrfs_next_leaf(root, path);
2850 } else if (ret > 0) {
2856 leaf = path->nodes[0];
2857 slot = path->slots[0];
2859 btrfs_item_key_to_cpu(leaf, &key, slot);
2860 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2865 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2870 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2871 rootrefs->rootref[found].treeid = key.offset;
2872 rootrefs->rootref[found].dirid =
2873 btrfs_root_ref_dirid(leaf, rref);
2876 ret = btrfs_next_item(root, path);
2879 } else if (ret > 0) {
2886 if (!ret || ret == -EOVERFLOW) {
2887 rootrefs->num_items = found;
2888 /* update min_treeid for next search */
2890 rootrefs->min_treeid =
2891 rootrefs->rootref[found - 1].treeid + 1;
2892 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2897 btrfs_free_path(path);
2902 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2906 struct dentry *parent = file->f_path.dentry;
2907 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2908 struct dentry *dentry;
2909 struct inode *dir = d_inode(parent);
2910 struct inode *inode;
2911 struct btrfs_root *root = BTRFS_I(dir)->root;
2912 struct btrfs_root *dest = NULL;
2913 struct btrfs_ioctl_vol_args *vol_args = NULL;
2914 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2915 char *subvol_name, *subvol_name_ptr = NULL;
2918 bool destroy_parent = false;
2921 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2922 if (IS_ERR(vol_args2))
2923 return PTR_ERR(vol_args2);
2925 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2931 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2932 * name, same as v1 currently does.
2934 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2935 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2936 subvol_name = vol_args2->name;
2938 err = mnt_want_write_file(file);
2942 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2947 err = mnt_want_write_file(file);
2951 dentry = btrfs_get_dentry(fs_info->sb,
2952 BTRFS_FIRST_FREE_OBJECTID,
2953 vol_args2->subvolid, 0, 0);
2954 if (IS_ERR(dentry)) {
2955 err = PTR_ERR(dentry);
2956 goto out_drop_write;
2960 * Change the default parent since the subvolume being
2961 * deleted can be outside of the current mount point.
2963 parent = btrfs_get_parent(dentry);
2966 * At this point dentry->d_name can point to '/' if the
2967 * subvolume we want to destroy is outsite of the
2968 * current mount point, so we need to release the
2969 * current dentry and execute the lookup to return a new
2970 * one with ->d_name pointing to the
2971 * <mount point>/subvol_name.
2974 if (IS_ERR(parent)) {
2975 err = PTR_ERR(parent);
2976 goto out_drop_write;
2978 dir = d_inode(parent);
2981 * If v2 was used with SPEC_BY_ID, a new parent was
2982 * allocated since the subvolume can be outside of the
2983 * current mount point. Later on we need to release this
2984 * new parent dentry.
2986 destroy_parent = true;
2988 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
2989 fs_info, vol_args2->subvolid);
2990 if (IS_ERR(subvol_name_ptr)) {
2991 err = PTR_ERR(subvol_name_ptr);
2994 /* subvol_name_ptr is already nul terminated */
2995 subvol_name = (char *)kbasename(subvol_name_ptr);
2998 vol_args = memdup_user(arg, sizeof(*vol_args));
2999 if (IS_ERR(vol_args))
3000 return PTR_ERR(vol_args);
3002 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
3003 subvol_name = vol_args->name;
3005 err = mnt_want_write_file(file);
3010 subvol_namelen = strlen(subvol_name);
3012 if (strchr(subvol_name, '/') ||
3013 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3015 goto free_subvol_name;
3018 if (!S_ISDIR(dir->i_mode)) {
3020 goto free_subvol_name;
3023 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3025 goto free_subvol_name;
3026 dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
3027 if (IS_ERR(dentry)) {
3028 err = PTR_ERR(dentry);
3029 goto out_unlock_dir;
3032 if (d_really_is_negative(dentry)) {
3037 inode = d_inode(dentry);
3038 dest = BTRFS_I(inode)->root;
3039 if (!capable(CAP_SYS_ADMIN)) {
3041 * Regular user. Only allow this with a special mount
3042 * option, when the user has write+exec access to the
3043 * subvol root, and when rmdir(2) would have been
3046 * Note that this is _not_ check that the subvol is
3047 * empty or doesn't contain data that we wouldn't
3048 * otherwise be able to delete.
3050 * Users who want to delete empty subvols should try
3054 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3058 * Do not allow deletion if the parent dir is the same
3059 * as the dir to be deleted. That means the ioctl
3060 * must be called on the dentry referencing the root
3061 * of the subvol, not a random directory contained
3068 err = inode_permission(&init_user_ns, inode,
3069 MAY_WRITE | MAY_EXEC);
3074 /* check if subvolume may be deleted by a user */
3075 err = btrfs_may_delete(dir, dentry, 1);
3079 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3084 btrfs_inode_lock(inode, 0);
3085 err = btrfs_delete_subvolume(dir, dentry);
3086 btrfs_inode_unlock(inode, 0);
3088 fsnotify_rmdir(dir, dentry);
3095 btrfs_inode_unlock(dir, 0);
3097 kfree(subvol_name_ptr);
3102 mnt_drop_write_file(file);
3109 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3111 struct inode *inode = file_inode(file);
3112 struct btrfs_root *root = BTRFS_I(inode)->root;
3113 struct btrfs_ioctl_defrag_range_args *range;
3116 ret = mnt_want_write_file(file);
3120 if (btrfs_root_readonly(root)) {
3125 /* Subpage defrag will be supported in later commits */
3126 if (root->fs_info->sectorsize < PAGE_SIZE) {
3131 switch (inode->i_mode & S_IFMT) {
3133 if (!capable(CAP_SYS_ADMIN)) {
3137 ret = btrfs_defrag_root(root);
3141 * Note that this does not check the file descriptor for write
3142 * access. This prevents defragmenting executables that are
3143 * running and allows defrag on files open in read-only mode.
3145 if (!capable(CAP_SYS_ADMIN) &&
3146 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
3151 range = kzalloc(sizeof(*range), GFP_KERNEL);
3158 if (copy_from_user(range, argp,
3164 /* compression requires us to start the IO */
3165 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3166 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3167 range->extent_thresh = (u32)-1;
3170 /* the rest are all set to zero by kzalloc */
3171 range->len = (u64)-1;
3173 ret = btrfs_defrag_file(file_inode(file), file,
3174 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;
3222 bool cancel = false;
3224 if (!capable(CAP_SYS_ADMIN))
3227 ret = mnt_want_write_file(file);
3231 vol_args = memdup_user(arg, sizeof(*vol_args));
3232 if (IS_ERR(vol_args)) {
3233 ret = PTR_ERR(vol_args);
3237 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3241 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3242 if (!(vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) &&
3243 strcmp("cancel", vol_args->name) == 0)
3246 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3250 /* Exclusive operation is now claimed */
3252 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3253 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3255 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3257 btrfs_exclop_finish(fs_info);
3260 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3261 btrfs_info(fs_info, "device deleted: id %llu",
3264 btrfs_info(fs_info, "device deleted: %s",
3270 mnt_drop_write_file(file);
3274 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3276 struct inode *inode = file_inode(file);
3277 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3278 struct btrfs_ioctl_vol_args *vol_args;
3282 if (!capable(CAP_SYS_ADMIN))
3285 ret = mnt_want_write_file(file);
3289 vol_args = memdup_user(arg, sizeof(*vol_args));
3290 if (IS_ERR(vol_args)) {
3291 ret = PTR_ERR(vol_args);
3292 goto out_drop_write;
3294 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3295 cancel = (strcmp("cancel", vol_args->name) == 0);
3297 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3300 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3302 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3303 btrfs_exclop_finish(fs_info);
3308 mnt_drop_write_file(file);
3313 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3316 struct btrfs_ioctl_fs_info_args *fi_args;
3317 struct btrfs_device *device;
3318 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3322 fi_args = memdup_user(arg, sizeof(*fi_args));
3323 if (IS_ERR(fi_args))
3324 return PTR_ERR(fi_args);
3326 flags_in = fi_args->flags;
3327 memset(fi_args, 0, sizeof(*fi_args));
3330 fi_args->num_devices = fs_devices->num_devices;
3332 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3333 if (device->devid > fi_args->max_id)
3334 fi_args->max_id = device->devid;
3338 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3339 fi_args->nodesize = fs_info->nodesize;
3340 fi_args->sectorsize = fs_info->sectorsize;
3341 fi_args->clone_alignment = fs_info->sectorsize;
3343 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3344 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3345 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3346 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3349 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3350 fi_args->generation = fs_info->generation;
3351 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3354 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3355 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3356 sizeof(fi_args->metadata_uuid));
3357 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3360 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3367 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3370 struct btrfs_ioctl_dev_info_args *di_args;
3371 struct btrfs_device *dev;
3373 char *s_uuid = NULL;
3375 di_args = memdup_user(arg, sizeof(*di_args));
3376 if (IS_ERR(di_args))
3377 return PTR_ERR(di_args);
3379 if (!btrfs_is_empty_uuid(di_args->uuid))
3380 s_uuid = di_args->uuid;
3383 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3391 di_args->devid = dev->devid;
3392 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3393 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3394 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3396 strncpy(di_args->path, rcu_str_deref(dev->name),
3397 sizeof(di_args->path) - 1);
3398 di_args->path[sizeof(di_args->path) - 1] = 0;
3400 di_args->path[0] = '\0';
3405 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3412 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3414 struct inode *inode = file_inode(file);
3415 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3416 struct btrfs_root *root = BTRFS_I(inode)->root;
3417 struct btrfs_root *new_root;
3418 struct btrfs_dir_item *di;
3419 struct btrfs_trans_handle *trans;
3420 struct btrfs_path *path = NULL;
3421 struct btrfs_disk_key disk_key;
3426 if (!capable(CAP_SYS_ADMIN))
3429 ret = mnt_want_write_file(file);
3433 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3439 objectid = BTRFS_FS_TREE_OBJECTID;
3441 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3442 if (IS_ERR(new_root)) {
3443 ret = PTR_ERR(new_root);
3446 if (!is_fstree(new_root->root_key.objectid)) {
3451 path = btrfs_alloc_path();
3457 trans = btrfs_start_transaction(root, 1);
3458 if (IS_ERR(trans)) {
3459 ret = PTR_ERR(trans);
3463 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3464 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3465 dir_id, "default", 7, 1);
3466 if (IS_ERR_OR_NULL(di)) {
3467 btrfs_release_path(path);
3468 btrfs_end_transaction(trans);
3470 "Umm, you don't have the default diritem, this isn't going to work");
3475 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3476 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3477 btrfs_mark_buffer_dirty(path->nodes[0]);
3478 btrfs_release_path(path);
3480 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3481 btrfs_end_transaction(trans);
3483 btrfs_put_root(new_root);
3484 btrfs_free_path(path);
3486 mnt_drop_write_file(file);
3490 static void get_block_group_info(struct list_head *groups_list,
3491 struct btrfs_ioctl_space_info *space)
3493 struct btrfs_block_group *block_group;
3495 space->total_bytes = 0;
3496 space->used_bytes = 0;
3498 list_for_each_entry(block_group, groups_list, list) {
3499 space->flags = block_group->flags;
3500 space->total_bytes += block_group->length;
3501 space->used_bytes += block_group->used;
3505 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3508 struct btrfs_ioctl_space_args space_args;
3509 struct btrfs_ioctl_space_info space;
3510 struct btrfs_ioctl_space_info *dest;
3511 struct btrfs_ioctl_space_info *dest_orig;
3512 struct btrfs_ioctl_space_info __user *user_dest;
3513 struct btrfs_space_info *info;
3514 static const u64 types[] = {
3515 BTRFS_BLOCK_GROUP_DATA,
3516 BTRFS_BLOCK_GROUP_SYSTEM,
3517 BTRFS_BLOCK_GROUP_METADATA,
3518 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3526 if (copy_from_user(&space_args,
3527 (struct btrfs_ioctl_space_args __user *)arg,
3528 sizeof(space_args)))
3531 for (i = 0; i < num_types; i++) {
3532 struct btrfs_space_info *tmp;
3535 list_for_each_entry(tmp, &fs_info->space_info, list) {
3536 if (tmp->flags == types[i]) {
3545 down_read(&info->groups_sem);
3546 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3547 if (!list_empty(&info->block_groups[c]))
3550 up_read(&info->groups_sem);
3554 * Global block reserve, exported as a space_info
3558 /* space_slots == 0 means they are asking for a count */
3559 if (space_args.space_slots == 0) {
3560 space_args.total_spaces = slot_count;
3564 slot_count = min_t(u64, space_args.space_slots, slot_count);
3566 alloc_size = sizeof(*dest) * slot_count;
3568 /* we generally have at most 6 or so space infos, one for each raid
3569 * level. So, a whole page should be more than enough for everyone
3571 if (alloc_size > PAGE_SIZE)
3574 space_args.total_spaces = 0;
3575 dest = kmalloc(alloc_size, GFP_KERNEL);
3580 /* now we have a buffer to copy into */
3581 for (i = 0; i < num_types; i++) {
3582 struct btrfs_space_info *tmp;
3588 list_for_each_entry(tmp, &fs_info->space_info, list) {
3589 if (tmp->flags == types[i]) {
3597 down_read(&info->groups_sem);
3598 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3599 if (!list_empty(&info->block_groups[c])) {
3600 get_block_group_info(&info->block_groups[c],
3602 memcpy(dest, &space, sizeof(space));
3604 space_args.total_spaces++;
3610 up_read(&info->groups_sem);
3614 * Add global block reserve
3617 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3619 spin_lock(&block_rsv->lock);
3620 space.total_bytes = block_rsv->size;
3621 space.used_bytes = block_rsv->size - block_rsv->reserved;
3622 spin_unlock(&block_rsv->lock);
3623 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3624 memcpy(dest, &space, sizeof(space));
3625 space_args.total_spaces++;
3628 user_dest = (struct btrfs_ioctl_space_info __user *)
3629 (arg + sizeof(struct btrfs_ioctl_space_args));
3631 if (copy_to_user(user_dest, dest_orig, alloc_size))
3636 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3642 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3645 struct btrfs_trans_handle *trans;
3649 trans = btrfs_attach_transaction_barrier(root);
3650 if (IS_ERR(trans)) {
3651 if (PTR_ERR(trans) != -ENOENT)
3652 return PTR_ERR(trans);
3654 /* No running transaction, don't bother */
3655 transid = root->fs_info->last_trans_committed;
3658 transid = trans->transid;
3659 ret = btrfs_commit_transaction_async(trans);
3661 btrfs_end_transaction(trans);
3666 if (copy_to_user(argp, &transid, sizeof(transid)))
3671 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3677 if (copy_from_user(&transid, argp, sizeof(transid)))
3680 transid = 0; /* current trans */
3682 return btrfs_wait_for_commit(fs_info, transid);
3685 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3687 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3688 struct btrfs_ioctl_scrub_args *sa;
3691 if (!capable(CAP_SYS_ADMIN))
3694 sa = memdup_user(arg, sizeof(*sa));
3698 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3699 ret = mnt_want_write_file(file);
3704 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3705 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3709 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3710 * error. This is important as it allows user space to know how much
3711 * progress scrub has done. For example, if scrub is canceled we get
3712 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3713 * space. Later user space can inspect the progress from the structure
3714 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3715 * previously (btrfs-progs does this).
3716 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3717 * then return -EFAULT to signal the structure was not copied or it may
3718 * be corrupt and unreliable due to a partial copy.
3720 if (copy_to_user(arg, sa, sizeof(*sa)))
3723 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3724 mnt_drop_write_file(file);
3730 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3732 if (!capable(CAP_SYS_ADMIN))
3735 return btrfs_scrub_cancel(fs_info);
3738 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3741 struct btrfs_ioctl_scrub_args *sa;
3744 if (!capable(CAP_SYS_ADMIN))
3747 sa = memdup_user(arg, sizeof(*sa));
3751 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3753 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3760 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3763 struct btrfs_ioctl_get_dev_stats *sa;
3766 sa = memdup_user(arg, sizeof(*sa));
3770 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3775 ret = btrfs_get_dev_stats(fs_info, sa);
3777 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3784 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3787 struct btrfs_ioctl_dev_replace_args *p;
3790 if (!capable(CAP_SYS_ADMIN))
3793 p = memdup_user(arg, sizeof(*p));
3798 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3799 if (sb_rdonly(fs_info->sb)) {
3803 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3804 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3806 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3807 btrfs_exclop_finish(fs_info);
3810 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3811 btrfs_dev_replace_status(fs_info, p);
3814 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3815 p->result = btrfs_dev_replace_cancel(fs_info);
3823 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3830 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3836 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3837 struct inode_fs_paths *ipath = NULL;
3838 struct btrfs_path *path;
3840 if (!capable(CAP_DAC_READ_SEARCH))
3843 path = btrfs_alloc_path();
3849 ipa = memdup_user(arg, sizeof(*ipa));
3856 size = min_t(u32, ipa->size, 4096);
3857 ipath = init_ipath(size, root, path);
3858 if (IS_ERR(ipath)) {
3859 ret = PTR_ERR(ipath);
3864 ret = paths_from_inode(ipa->inum, ipath);
3868 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3869 rel_ptr = ipath->fspath->val[i] -
3870 (u64)(unsigned long)ipath->fspath->val;
3871 ipath->fspath->val[i] = rel_ptr;
3874 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3875 ipath->fspath, size);
3882 btrfs_free_path(path);
3889 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3891 struct btrfs_data_container *inodes = ctx;
3892 const size_t c = 3 * sizeof(u64);
3894 if (inodes->bytes_left >= c) {
3895 inodes->bytes_left -= c;
3896 inodes->val[inodes->elem_cnt] = inum;
3897 inodes->val[inodes->elem_cnt + 1] = offset;
3898 inodes->val[inodes->elem_cnt + 2] = root;
3899 inodes->elem_cnt += 3;
3901 inodes->bytes_missing += c - inodes->bytes_left;
3902 inodes->bytes_left = 0;
3903 inodes->elem_missed += 3;
3909 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3910 void __user *arg, int version)
3914 struct btrfs_ioctl_logical_ino_args *loi;
3915 struct btrfs_data_container *inodes = NULL;
3916 struct btrfs_path *path = NULL;
3919 if (!capable(CAP_SYS_ADMIN))
3922 loi = memdup_user(arg, sizeof(*loi));
3924 return PTR_ERR(loi);
3927 ignore_offset = false;
3928 size = min_t(u32, loi->size, SZ_64K);
3930 /* All reserved bits must be 0 for now */
3931 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3935 /* Only accept flags we have defined so far */
3936 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3940 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3941 size = min_t(u32, loi->size, SZ_16M);
3944 path = btrfs_alloc_path();
3950 inodes = init_data_container(size);
3951 if (IS_ERR(inodes)) {
3952 ret = PTR_ERR(inodes);
3957 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3958 build_ino_list, inodes, ignore_offset);
3964 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3970 btrfs_free_path(path);
3978 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3979 struct btrfs_ioctl_balance_args *bargs)
3981 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3983 bargs->flags = bctl->flags;
3985 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
3986 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3987 if (atomic_read(&fs_info->balance_pause_req))
3988 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3989 if (atomic_read(&fs_info->balance_cancel_req))
3990 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3992 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3993 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3994 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3996 spin_lock(&fs_info->balance_lock);
3997 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3998 spin_unlock(&fs_info->balance_lock);
4001 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4003 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4004 struct btrfs_fs_info *fs_info = root->fs_info;
4005 struct btrfs_ioctl_balance_args *bargs;
4006 struct btrfs_balance_control *bctl;
4007 bool need_unlock; /* for mut. excl. ops lock */
4010 if (!capable(CAP_SYS_ADMIN))
4013 ret = mnt_want_write_file(file);
4018 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4019 mutex_lock(&fs_info->balance_mutex);
4025 * mut. excl. ops lock is locked. Three possibilities:
4026 * (1) some other op is running
4027 * (2) balance is running
4028 * (3) balance is paused -- special case (think resume)
4030 mutex_lock(&fs_info->balance_mutex);
4031 if (fs_info->balance_ctl) {
4032 /* this is either (2) or (3) */
4033 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4034 mutex_unlock(&fs_info->balance_mutex);
4036 * Lock released to allow other waiters to continue,
4037 * we'll reexamine the status again.
4039 mutex_lock(&fs_info->balance_mutex);
4041 if (fs_info->balance_ctl &&
4042 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4044 need_unlock = false;
4048 mutex_unlock(&fs_info->balance_mutex);
4052 mutex_unlock(&fs_info->balance_mutex);
4058 mutex_unlock(&fs_info->balance_mutex);
4059 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4066 bargs = memdup_user(arg, sizeof(*bargs));
4067 if (IS_ERR(bargs)) {
4068 ret = PTR_ERR(bargs);
4072 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4073 if (!fs_info->balance_ctl) {
4078 bctl = fs_info->balance_ctl;
4079 spin_lock(&fs_info->balance_lock);
4080 bctl->flags |= BTRFS_BALANCE_RESUME;
4081 spin_unlock(&fs_info->balance_lock);
4089 if (fs_info->balance_ctl) {
4094 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4101 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4102 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4103 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4105 bctl->flags = bargs->flags;
4107 /* balance everything - no filters */
4108 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4111 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4118 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4119 * bctl is freed in reset_balance_state, or, if restriper was paused
4120 * all the way until unmount, in free_fs_info. The flag should be
4121 * cleared after reset_balance_state.
4123 need_unlock = false;
4125 ret = btrfs_balance(fs_info, bctl, bargs);
4128 if ((ret == 0 || ret == -ECANCELED) && arg) {
4129 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4138 mutex_unlock(&fs_info->balance_mutex);
4140 btrfs_exclop_finish(fs_info);
4142 mnt_drop_write_file(file);
4146 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4148 if (!capable(CAP_SYS_ADMIN))
4152 case BTRFS_BALANCE_CTL_PAUSE:
4153 return btrfs_pause_balance(fs_info);
4154 case BTRFS_BALANCE_CTL_CANCEL:
4155 return btrfs_cancel_balance(fs_info);
4161 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4164 struct btrfs_ioctl_balance_args *bargs;
4167 if (!capable(CAP_SYS_ADMIN))
4170 mutex_lock(&fs_info->balance_mutex);
4171 if (!fs_info->balance_ctl) {
4176 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4182 btrfs_update_ioctl_balance_args(fs_info, bargs);
4184 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4189 mutex_unlock(&fs_info->balance_mutex);
4193 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4195 struct inode *inode = file_inode(file);
4196 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4197 struct btrfs_ioctl_quota_ctl_args *sa;
4200 if (!capable(CAP_SYS_ADMIN))
4203 ret = mnt_want_write_file(file);
4207 sa = memdup_user(arg, sizeof(*sa));
4213 down_write(&fs_info->subvol_sem);
4216 case BTRFS_QUOTA_CTL_ENABLE:
4217 ret = btrfs_quota_enable(fs_info);
4219 case BTRFS_QUOTA_CTL_DISABLE:
4220 ret = btrfs_quota_disable(fs_info);
4228 up_write(&fs_info->subvol_sem);
4230 mnt_drop_write_file(file);
4234 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4236 struct inode *inode = file_inode(file);
4237 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4238 struct btrfs_root *root = BTRFS_I(inode)->root;
4239 struct btrfs_ioctl_qgroup_assign_args *sa;
4240 struct btrfs_trans_handle *trans;
4244 if (!capable(CAP_SYS_ADMIN))
4247 ret = mnt_want_write_file(file);
4251 sa = memdup_user(arg, sizeof(*sa));
4257 trans = btrfs_join_transaction(root);
4258 if (IS_ERR(trans)) {
4259 ret = PTR_ERR(trans);
4264 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4266 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4269 /* update qgroup status and info */
4270 err = btrfs_run_qgroups(trans);
4272 btrfs_handle_fs_error(fs_info, err,
4273 "failed to update qgroup status and info");
4274 err = btrfs_end_transaction(trans);
4281 mnt_drop_write_file(file);
4285 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4287 struct inode *inode = file_inode(file);
4288 struct btrfs_root *root = BTRFS_I(inode)->root;
4289 struct btrfs_ioctl_qgroup_create_args *sa;
4290 struct btrfs_trans_handle *trans;
4294 if (!capable(CAP_SYS_ADMIN))
4297 ret = mnt_want_write_file(file);
4301 sa = memdup_user(arg, sizeof(*sa));
4307 if (!sa->qgroupid) {
4312 trans = btrfs_join_transaction(root);
4313 if (IS_ERR(trans)) {
4314 ret = PTR_ERR(trans);
4319 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4321 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4324 err = btrfs_end_transaction(trans);
4331 mnt_drop_write_file(file);
4335 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4337 struct inode *inode = file_inode(file);
4338 struct btrfs_root *root = BTRFS_I(inode)->root;
4339 struct btrfs_ioctl_qgroup_limit_args *sa;
4340 struct btrfs_trans_handle *trans;
4345 if (!capable(CAP_SYS_ADMIN))
4348 ret = mnt_want_write_file(file);
4352 sa = memdup_user(arg, sizeof(*sa));
4358 trans = btrfs_join_transaction(root);
4359 if (IS_ERR(trans)) {
4360 ret = PTR_ERR(trans);
4364 qgroupid = sa->qgroupid;
4366 /* take the current subvol as qgroup */
4367 qgroupid = root->root_key.objectid;
4370 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4372 err = btrfs_end_transaction(trans);
4379 mnt_drop_write_file(file);
4383 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4385 struct inode *inode = file_inode(file);
4386 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4387 struct btrfs_ioctl_quota_rescan_args *qsa;
4390 if (!capable(CAP_SYS_ADMIN))
4393 ret = mnt_want_write_file(file);
4397 qsa = memdup_user(arg, sizeof(*qsa));
4408 ret = btrfs_qgroup_rescan(fs_info);
4413 mnt_drop_write_file(file);
4417 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4420 struct btrfs_ioctl_quota_rescan_args *qsa;
4423 if (!capable(CAP_SYS_ADMIN))
4426 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4430 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4432 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4435 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4442 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4445 if (!capable(CAP_SYS_ADMIN))
4448 return btrfs_qgroup_wait_for_completion(fs_info, true);
4451 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4452 struct btrfs_ioctl_received_subvol_args *sa)
4454 struct inode *inode = file_inode(file);
4455 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4456 struct btrfs_root *root = BTRFS_I(inode)->root;
4457 struct btrfs_root_item *root_item = &root->root_item;
4458 struct btrfs_trans_handle *trans;
4459 struct timespec64 ct = current_time(inode);
4461 int received_uuid_changed;
4463 if (!inode_owner_or_capable(&init_user_ns, inode))
4466 ret = mnt_want_write_file(file);
4470 down_write(&fs_info->subvol_sem);
4472 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4477 if (btrfs_root_readonly(root)) {
4484 * 2 - uuid items (received uuid + subvol uuid)
4486 trans = btrfs_start_transaction(root, 3);
4487 if (IS_ERR(trans)) {
4488 ret = PTR_ERR(trans);
4493 sa->rtransid = trans->transid;
4494 sa->rtime.sec = ct.tv_sec;
4495 sa->rtime.nsec = ct.tv_nsec;
4497 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4499 if (received_uuid_changed &&
4500 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4501 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4502 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4503 root->root_key.objectid);
4504 if (ret && ret != -ENOENT) {
4505 btrfs_abort_transaction(trans, ret);
4506 btrfs_end_transaction(trans);
4510 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4511 btrfs_set_root_stransid(root_item, sa->stransid);
4512 btrfs_set_root_rtransid(root_item, sa->rtransid);
4513 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4514 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4515 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4516 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4518 ret = btrfs_update_root(trans, fs_info->tree_root,
4519 &root->root_key, &root->root_item);
4521 btrfs_end_transaction(trans);
4524 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4525 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4526 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4527 root->root_key.objectid);
4528 if (ret < 0 && ret != -EEXIST) {
4529 btrfs_abort_transaction(trans, ret);
4530 btrfs_end_transaction(trans);
4534 ret = btrfs_commit_transaction(trans);
4536 up_write(&fs_info->subvol_sem);
4537 mnt_drop_write_file(file);
4542 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4545 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4546 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4549 args32 = memdup_user(arg, sizeof(*args32));
4551 return PTR_ERR(args32);
4553 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4559 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4560 args64->stransid = args32->stransid;
4561 args64->rtransid = args32->rtransid;
4562 args64->stime.sec = args32->stime.sec;
4563 args64->stime.nsec = args32->stime.nsec;
4564 args64->rtime.sec = args32->rtime.sec;
4565 args64->rtime.nsec = args32->rtime.nsec;
4566 args64->flags = args32->flags;
4568 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4572 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4573 args32->stransid = args64->stransid;
4574 args32->rtransid = args64->rtransid;
4575 args32->stime.sec = args64->stime.sec;
4576 args32->stime.nsec = args64->stime.nsec;
4577 args32->rtime.sec = args64->rtime.sec;
4578 args32->rtime.nsec = args64->rtime.nsec;
4579 args32->flags = args64->flags;
4581 ret = copy_to_user(arg, args32, sizeof(*args32));
4592 static long btrfs_ioctl_set_received_subvol(struct file *file,
4595 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4598 sa = memdup_user(arg, sizeof(*sa));
4602 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4607 ret = copy_to_user(arg, sa, sizeof(*sa));
4616 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4621 char label[BTRFS_LABEL_SIZE];
4623 spin_lock(&fs_info->super_lock);
4624 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4625 spin_unlock(&fs_info->super_lock);
4627 len = strnlen(label, BTRFS_LABEL_SIZE);
4629 if (len == BTRFS_LABEL_SIZE) {
4631 "label is too long, return the first %zu bytes",
4635 ret = copy_to_user(arg, label, len);
4637 return ret ? -EFAULT : 0;
4640 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4642 struct inode *inode = file_inode(file);
4643 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4644 struct btrfs_root *root = BTRFS_I(inode)->root;
4645 struct btrfs_super_block *super_block = fs_info->super_copy;
4646 struct btrfs_trans_handle *trans;
4647 char label[BTRFS_LABEL_SIZE];
4650 if (!capable(CAP_SYS_ADMIN))
4653 if (copy_from_user(label, arg, sizeof(label)))
4656 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4658 "unable to set label with more than %d bytes",
4659 BTRFS_LABEL_SIZE - 1);
4663 ret = mnt_want_write_file(file);
4667 trans = btrfs_start_transaction(root, 0);
4668 if (IS_ERR(trans)) {
4669 ret = PTR_ERR(trans);
4673 spin_lock(&fs_info->super_lock);
4674 strcpy(super_block->label, label);
4675 spin_unlock(&fs_info->super_lock);
4676 ret = btrfs_commit_transaction(trans);
4679 mnt_drop_write_file(file);
4683 #define INIT_FEATURE_FLAGS(suffix) \
4684 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4685 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4686 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4688 int btrfs_ioctl_get_supported_features(void __user *arg)
4690 static const struct btrfs_ioctl_feature_flags features[3] = {
4691 INIT_FEATURE_FLAGS(SUPP),
4692 INIT_FEATURE_FLAGS(SAFE_SET),
4693 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4696 if (copy_to_user(arg, &features, sizeof(features)))
4702 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4705 struct btrfs_super_block *super_block = fs_info->super_copy;
4706 struct btrfs_ioctl_feature_flags features;
4708 features.compat_flags = btrfs_super_compat_flags(super_block);
4709 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4710 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4712 if (copy_to_user(arg, &features, sizeof(features)))
4718 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4719 enum btrfs_feature_set set,
4720 u64 change_mask, u64 flags, u64 supported_flags,
4721 u64 safe_set, u64 safe_clear)
4723 const char *type = btrfs_feature_set_name(set);
4725 u64 disallowed, unsupported;
4726 u64 set_mask = flags & change_mask;
4727 u64 clear_mask = ~flags & change_mask;
4729 unsupported = set_mask & ~supported_flags;
4731 names = btrfs_printable_features(set, unsupported);
4734 "this kernel does not support the %s feature bit%s",
4735 names, strchr(names, ',') ? "s" : "");
4739 "this kernel does not support %s bits 0x%llx",
4744 disallowed = set_mask & ~safe_set;
4746 names = btrfs_printable_features(set, disallowed);
4749 "can't set the %s feature bit%s while mounted",
4750 names, strchr(names, ',') ? "s" : "");
4754 "can't set %s bits 0x%llx while mounted",
4759 disallowed = clear_mask & ~safe_clear;
4761 names = btrfs_printable_features(set, disallowed);
4764 "can't clear the %s feature bit%s while mounted",
4765 names, strchr(names, ',') ? "s" : "");
4769 "can't clear %s bits 0x%llx while mounted",
4777 #define check_feature(fs_info, change_mask, flags, mask_base) \
4778 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4779 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4780 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4781 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4783 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4785 struct inode *inode = file_inode(file);
4786 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4787 struct btrfs_root *root = BTRFS_I(inode)->root;
4788 struct btrfs_super_block *super_block = fs_info->super_copy;
4789 struct btrfs_ioctl_feature_flags flags[2];
4790 struct btrfs_trans_handle *trans;
4794 if (!capable(CAP_SYS_ADMIN))
4797 if (copy_from_user(flags, arg, sizeof(flags)))
4801 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4802 !flags[0].incompat_flags)
4805 ret = check_feature(fs_info, flags[0].compat_flags,
4806 flags[1].compat_flags, COMPAT);
4810 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4811 flags[1].compat_ro_flags, COMPAT_RO);
4815 ret = check_feature(fs_info, flags[0].incompat_flags,
4816 flags[1].incompat_flags, INCOMPAT);
4820 ret = mnt_want_write_file(file);
4824 trans = btrfs_start_transaction(root, 0);
4825 if (IS_ERR(trans)) {
4826 ret = PTR_ERR(trans);
4827 goto out_drop_write;
4830 spin_lock(&fs_info->super_lock);
4831 newflags = btrfs_super_compat_flags(super_block);
4832 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4833 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4834 btrfs_set_super_compat_flags(super_block, newflags);
4836 newflags = btrfs_super_compat_ro_flags(super_block);
4837 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4838 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4839 btrfs_set_super_compat_ro_flags(super_block, newflags);
4841 newflags = btrfs_super_incompat_flags(super_block);
4842 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4843 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4844 btrfs_set_super_incompat_flags(super_block, newflags);
4845 spin_unlock(&fs_info->super_lock);
4847 ret = btrfs_commit_transaction(trans);
4849 mnt_drop_write_file(file);
4854 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4856 struct btrfs_ioctl_send_args *arg;
4860 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4861 struct btrfs_ioctl_send_args_32 args32;
4863 ret = copy_from_user(&args32, argp, sizeof(args32));
4866 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4869 arg->send_fd = args32.send_fd;
4870 arg->clone_sources_count = args32.clone_sources_count;
4871 arg->clone_sources = compat_ptr(args32.clone_sources);
4872 arg->parent_root = args32.parent_root;
4873 arg->flags = args32.flags;
4874 memcpy(arg->reserved, args32.reserved,
4875 sizeof(args32.reserved));
4880 arg = memdup_user(argp, sizeof(*arg));
4882 return PTR_ERR(arg);
4884 ret = btrfs_ioctl_send(file, arg);
4889 long btrfs_ioctl(struct file *file, unsigned int
4890 cmd, unsigned long arg)
4892 struct inode *inode = file_inode(file);
4893 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4894 struct btrfs_root *root = BTRFS_I(inode)->root;
4895 void __user *argp = (void __user *)arg;
4898 case FS_IOC_GETVERSION:
4899 return btrfs_ioctl_getversion(file, argp);
4900 case FS_IOC_GETFSLABEL:
4901 return btrfs_ioctl_get_fslabel(fs_info, argp);
4902 case FS_IOC_SETFSLABEL:
4903 return btrfs_ioctl_set_fslabel(file, argp);
4905 return btrfs_ioctl_fitrim(fs_info, argp);
4906 case BTRFS_IOC_SNAP_CREATE:
4907 return btrfs_ioctl_snap_create(file, argp, 0);
4908 case BTRFS_IOC_SNAP_CREATE_V2:
4909 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4910 case BTRFS_IOC_SUBVOL_CREATE:
4911 return btrfs_ioctl_snap_create(file, argp, 1);
4912 case BTRFS_IOC_SUBVOL_CREATE_V2:
4913 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4914 case BTRFS_IOC_SNAP_DESTROY:
4915 return btrfs_ioctl_snap_destroy(file, argp, false);
4916 case BTRFS_IOC_SNAP_DESTROY_V2:
4917 return btrfs_ioctl_snap_destroy(file, argp, true);
4918 case BTRFS_IOC_SUBVOL_GETFLAGS:
4919 return btrfs_ioctl_subvol_getflags(file, argp);
4920 case BTRFS_IOC_SUBVOL_SETFLAGS:
4921 return btrfs_ioctl_subvol_setflags(file, argp);
4922 case BTRFS_IOC_DEFAULT_SUBVOL:
4923 return btrfs_ioctl_default_subvol(file, argp);
4924 case BTRFS_IOC_DEFRAG:
4925 return btrfs_ioctl_defrag(file, NULL);
4926 case BTRFS_IOC_DEFRAG_RANGE:
4927 return btrfs_ioctl_defrag(file, argp);
4928 case BTRFS_IOC_RESIZE:
4929 return btrfs_ioctl_resize(file, argp);
4930 case BTRFS_IOC_ADD_DEV:
4931 return btrfs_ioctl_add_dev(fs_info, argp);
4932 case BTRFS_IOC_RM_DEV:
4933 return btrfs_ioctl_rm_dev(file, argp);
4934 case BTRFS_IOC_RM_DEV_V2:
4935 return btrfs_ioctl_rm_dev_v2(file, argp);
4936 case BTRFS_IOC_FS_INFO:
4937 return btrfs_ioctl_fs_info(fs_info, argp);
4938 case BTRFS_IOC_DEV_INFO:
4939 return btrfs_ioctl_dev_info(fs_info, argp);
4940 case BTRFS_IOC_BALANCE:
4941 return btrfs_ioctl_balance(file, NULL);
4942 case BTRFS_IOC_TREE_SEARCH:
4943 return btrfs_ioctl_tree_search(file, argp);
4944 case BTRFS_IOC_TREE_SEARCH_V2:
4945 return btrfs_ioctl_tree_search_v2(file, argp);
4946 case BTRFS_IOC_INO_LOOKUP:
4947 return btrfs_ioctl_ino_lookup(file, argp);
4948 case BTRFS_IOC_INO_PATHS:
4949 return btrfs_ioctl_ino_to_path(root, argp);
4950 case BTRFS_IOC_LOGICAL_INO:
4951 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4952 case BTRFS_IOC_LOGICAL_INO_V2:
4953 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4954 case BTRFS_IOC_SPACE_INFO:
4955 return btrfs_ioctl_space_info(fs_info, argp);
4956 case BTRFS_IOC_SYNC: {
4959 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4962 ret = btrfs_sync_fs(inode->i_sb, 1);
4964 * The transaction thread may want to do more work,
4965 * namely it pokes the cleaner kthread that will start
4966 * processing uncleaned subvols.
4968 wake_up_process(fs_info->transaction_kthread);
4971 case BTRFS_IOC_START_SYNC:
4972 return btrfs_ioctl_start_sync(root, argp);
4973 case BTRFS_IOC_WAIT_SYNC:
4974 return btrfs_ioctl_wait_sync(fs_info, argp);
4975 case BTRFS_IOC_SCRUB:
4976 return btrfs_ioctl_scrub(file, argp);
4977 case BTRFS_IOC_SCRUB_CANCEL:
4978 return btrfs_ioctl_scrub_cancel(fs_info);
4979 case BTRFS_IOC_SCRUB_PROGRESS:
4980 return btrfs_ioctl_scrub_progress(fs_info, argp);
4981 case BTRFS_IOC_BALANCE_V2:
4982 return btrfs_ioctl_balance(file, argp);
4983 case BTRFS_IOC_BALANCE_CTL:
4984 return btrfs_ioctl_balance_ctl(fs_info, arg);
4985 case BTRFS_IOC_BALANCE_PROGRESS:
4986 return btrfs_ioctl_balance_progress(fs_info, argp);
4987 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4988 return btrfs_ioctl_set_received_subvol(file, argp);
4990 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4991 return btrfs_ioctl_set_received_subvol_32(file, argp);
4993 case BTRFS_IOC_SEND:
4994 return _btrfs_ioctl_send(file, argp, false);
4995 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4996 case BTRFS_IOC_SEND_32:
4997 return _btrfs_ioctl_send(file, argp, true);
4999 case BTRFS_IOC_GET_DEV_STATS:
5000 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5001 case BTRFS_IOC_QUOTA_CTL:
5002 return btrfs_ioctl_quota_ctl(file, argp);
5003 case BTRFS_IOC_QGROUP_ASSIGN:
5004 return btrfs_ioctl_qgroup_assign(file, argp);
5005 case BTRFS_IOC_QGROUP_CREATE:
5006 return btrfs_ioctl_qgroup_create(file, argp);
5007 case BTRFS_IOC_QGROUP_LIMIT:
5008 return btrfs_ioctl_qgroup_limit(file, argp);
5009 case BTRFS_IOC_QUOTA_RESCAN:
5010 return btrfs_ioctl_quota_rescan(file, argp);
5011 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5012 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5013 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5014 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5015 case BTRFS_IOC_DEV_REPLACE:
5016 return btrfs_ioctl_dev_replace(fs_info, argp);
5017 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5018 return btrfs_ioctl_get_supported_features(argp);
5019 case BTRFS_IOC_GET_FEATURES:
5020 return btrfs_ioctl_get_features(fs_info, argp);
5021 case BTRFS_IOC_SET_FEATURES:
5022 return btrfs_ioctl_set_features(file, argp);
5023 case BTRFS_IOC_GET_SUBVOL_INFO:
5024 return btrfs_ioctl_get_subvol_info(file, argp);
5025 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5026 return btrfs_ioctl_get_subvol_rootref(file, argp);
5027 case BTRFS_IOC_INO_LOOKUP_USER:
5028 return btrfs_ioctl_ino_lookup_user(file, argp);
5029 case FS_IOC_ENABLE_VERITY:
5030 return fsverity_ioctl_enable(file, (const void __user *)argp);
5031 case FS_IOC_MEASURE_VERITY:
5032 return fsverity_ioctl_measure(file, argp);
5038 #ifdef CONFIG_COMPAT
5039 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5042 * These all access 32-bit values anyway so no further
5043 * handling is necessary.
5046 case FS_IOC32_GETVERSION:
5047 cmd = FS_IOC_GETVERSION;
5051 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));