2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
62 static int btrfs_clone(struct inode *src, struct inode *inode,
63 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
65 /* Mask out flags that are inappropriate for the given type of inode. */
66 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
70 else if (S_ISREG(mode))
71 return flags & ~FS_DIRSYNC_FL;
73 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
77 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
79 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
81 unsigned int iflags = 0;
83 if (flags & BTRFS_INODE_SYNC)
85 if (flags & BTRFS_INODE_IMMUTABLE)
86 iflags |= FS_IMMUTABLE_FL;
87 if (flags & BTRFS_INODE_APPEND)
88 iflags |= FS_APPEND_FL;
89 if (flags & BTRFS_INODE_NODUMP)
90 iflags |= FS_NODUMP_FL;
91 if (flags & BTRFS_INODE_NOATIME)
92 iflags |= FS_NOATIME_FL;
93 if (flags & BTRFS_INODE_DIRSYNC)
94 iflags |= FS_DIRSYNC_FL;
95 if (flags & BTRFS_INODE_NODATACOW)
96 iflags |= FS_NOCOW_FL;
98 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
99 iflags |= FS_COMPR_FL;
100 else if (flags & BTRFS_INODE_NOCOMPRESS)
101 iflags |= FS_NOCOMP_FL;
107 * Update inode->i_flags based on the btrfs internal flags.
109 void btrfs_update_iflags(struct inode *inode)
111 struct btrfs_inode *ip = BTRFS_I(inode);
113 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
115 if (ip->flags & BTRFS_INODE_SYNC)
116 inode->i_flags |= S_SYNC;
117 if (ip->flags & BTRFS_INODE_IMMUTABLE)
118 inode->i_flags |= S_IMMUTABLE;
119 if (ip->flags & BTRFS_INODE_APPEND)
120 inode->i_flags |= S_APPEND;
121 if (ip->flags & BTRFS_INODE_NOATIME)
122 inode->i_flags |= S_NOATIME;
123 if (ip->flags & BTRFS_INODE_DIRSYNC)
124 inode->i_flags |= S_DIRSYNC;
128 * Inherit flags from the parent inode.
130 * Currently only the compression flags and the cow flags are inherited.
132 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
139 flags = BTRFS_I(dir)->flags;
141 if (flags & BTRFS_INODE_NOCOMPRESS) {
142 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
143 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
144 } else if (flags & BTRFS_INODE_COMPRESS) {
145 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
146 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
149 if (flags & BTRFS_INODE_NODATACOW) {
150 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
151 if (S_ISREG(inode->i_mode))
152 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
155 btrfs_update_iflags(inode);
158 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
160 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
161 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
163 if (copy_to_user(arg, &flags, sizeof(flags)))
168 static int check_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 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
183 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
185 struct inode *inode = file_inode(file);
186 struct btrfs_inode *ip = BTRFS_I(inode);
187 struct btrfs_root *root = ip->root;
188 struct btrfs_trans_handle *trans;
189 unsigned int flags, oldflags;
192 unsigned int i_oldflags;
195 if (!inode_owner_or_capable(inode))
198 if (btrfs_root_readonly(root))
201 if (copy_from_user(&flags, arg, sizeof(flags)))
204 ret = check_flags(flags);
208 ret = mnt_want_write_file(file);
212 mutex_lock(&inode->i_mutex);
214 ip_oldflags = ip->flags;
215 i_oldflags = inode->i_flags;
216 mode = inode->i_mode;
218 flags = btrfs_mask_flags(inode->i_mode, flags);
219 oldflags = btrfs_flags_to_ioctl(ip->flags);
220 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
221 if (!capable(CAP_LINUX_IMMUTABLE)) {
227 if (flags & FS_SYNC_FL)
228 ip->flags |= BTRFS_INODE_SYNC;
230 ip->flags &= ~BTRFS_INODE_SYNC;
231 if (flags & FS_IMMUTABLE_FL)
232 ip->flags |= BTRFS_INODE_IMMUTABLE;
234 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
235 if (flags & FS_APPEND_FL)
236 ip->flags |= BTRFS_INODE_APPEND;
238 ip->flags &= ~BTRFS_INODE_APPEND;
239 if (flags & FS_NODUMP_FL)
240 ip->flags |= BTRFS_INODE_NODUMP;
242 ip->flags &= ~BTRFS_INODE_NODUMP;
243 if (flags & FS_NOATIME_FL)
244 ip->flags |= BTRFS_INODE_NOATIME;
246 ip->flags &= ~BTRFS_INODE_NOATIME;
247 if (flags & FS_DIRSYNC_FL)
248 ip->flags |= BTRFS_INODE_DIRSYNC;
250 ip->flags &= ~BTRFS_INODE_DIRSYNC;
251 if (flags & FS_NOCOW_FL) {
254 * It's safe to turn csums off here, no extents exist.
255 * Otherwise we want the flag to reflect the real COW
256 * status of the file and will not set it.
258 if (inode->i_size == 0)
259 ip->flags |= BTRFS_INODE_NODATACOW
260 | BTRFS_INODE_NODATASUM;
262 ip->flags |= BTRFS_INODE_NODATACOW;
266 * Revert back under same assuptions as above
269 if (inode->i_size == 0)
270 ip->flags &= ~(BTRFS_INODE_NODATACOW
271 | BTRFS_INODE_NODATASUM);
273 ip->flags &= ~BTRFS_INODE_NODATACOW;
278 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
279 * flag may be changed automatically if compression code won't make
282 if (flags & FS_NOCOMP_FL) {
283 ip->flags &= ~BTRFS_INODE_COMPRESS;
284 ip->flags |= BTRFS_INODE_NOCOMPRESS;
286 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
287 if (ret && ret != -ENODATA)
289 } else if (flags & FS_COMPR_FL) {
292 ip->flags |= BTRFS_INODE_COMPRESS;
293 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
295 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
299 ret = btrfs_set_prop(inode, "btrfs.compression",
300 comp, strlen(comp), 0);
305 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
308 trans = btrfs_start_transaction(root, 1);
310 ret = PTR_ERR(trans);
314 btrfs_update_iflags(inode);
315 inode_inc_iversion(inode);
316 inode->i_ctime = CURRENT_TIME;
317 ret = btrfs_update_inode(trans, root, inode);
319 btrfs_end_transaction(trans, root);
322 ip->flags = ip_oldflags;
323 inode->i_flags = i_oldflags;
327 mutex_unlock(&inode->i_mutex);
328 mnt_drop_write_file(file);
332 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
334 struct inode *inode = file_inode(file);
336 return put_user(inode->i_generation, arg);
339 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
341 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
342 struct btrfs_device *device;
343 struct request_queue *q;
344 struct fstrim_range range;
345 u64 minlen = ULLONG_MAX;
347 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
350 if (!capable(CAP_SYS_ADMIN))
354 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
358 q = bdev_get_queue(device->bdev);
359 if (blk_queue_discard(q)) {
361 minlen = min((u64)q->limits.discard_granularity,
369 if (copy_from_user(&range, arg, sizeof(range)))
371 if (range.start > total_bytes ||
372 range.len < fs_info->sb->s_blocksize)
375 range.len = min(range.len, total_bytes - range.start);
376 range.minlen = max(range.minlen, minlen);
377 ret = btrfs_trim_fs(fs_info->tree_root, &range);
381 if (copy_to_user(arg, &range, sizeof(range)))
387 int btrfs_is_empty_uuid(u8 *uuid)
391 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
398 static noinline int create_subvol(struct inode *dir,
399 struct dentry *dentry,
400 char *name, int namelen,
402 struct btrfs_qgroup_inherit *inherit)
404 struct btrfs_trans_handle *trans;
405 struct btrfs_key key;
406 struct btrfs_root_item root_item;
407 struct btrfs_inode_item *inode_item;
408 struct extent_buffer *leaf;
409 struct btrfs_root *root = BTRFS_I(dir)->root;
410 struct btrfs_root *new_root;
411 struct btrfs_block_rsv block_rsv;
412 struct timespec cur_time = CURRENT_TIME;
417 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
422 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
426 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
428 * The same as the snapshot creation, please see the comment
429 * of create_snapshot().
431 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
432 8, &qgroup_reserved, false);
436 trans = btrfs_start_transaction(root, 0);
438 ret = PTR_ERR(trans);
439 btrfs_subvolume_release_metadata(root, &block_rsv,
443 trans->block_rsv = &block_rsv;
444 trans->bytes_reserved = block_rsv.size;
446 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
450 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
451 0, objectid, NULL, 0, 0, 0);
457 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
458 btrfs_set_header_bytenr(leaf, leaf->start);
459 btrfs_set_header_generation(leaf, trans->transid);
460 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
461 btrfs_set_header_owner(leaf, objectid);
463 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
465 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
466 btrfs_header_chunk_tree_uuid(leaf),
468 btrfs_mark_buffer_dirty(leaf);
470 memset(&root_item, 0, sizeof(root_item));
472 inode_item = &root_item.inode;
473 btrfs_set_stack_inode_generation(inode_item, 1);
474 btrfs_set_stack_inode_size(inode_item, 3);
475 btrfs_set_stack_inode_nlink(inode_item, 1);
476 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
477 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
479 btrfs_set_root_flags(&root_item, 0);
480 btrfs_set_root_limit(&root_item, 0);
481 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
483 btrfs_set_root_bytenr(&root_item, leaf->start);
484 btrfs_set_root_generation(&root_item, trans->transid);
485 btrfs_set_root_level(&root_item, 0);
486 btrfs_set_root_refs(&root_item, 1);
487 btrfs_set_root_used(&root_item, leaf->len);
488 btrfs_set_root_last_snapshot(&root_item, 0);
490 btrfs_set_root_generation_v2(&root_item,
491 btrfs_root_generation(&root_item));
492 uuid_le_gen(&new_uuid);
493 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
494 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
495 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
496 root_item.ctime = root_item.otime;
497 btrfs_set_root_ctransid(&root_item, trans->transid);
498 btrfs_set_root_otransid(&root_item, trans->transid);
500 btrfs_tree_unlock(leaf);
501 free_extent_buffer(leaf);
504 btrfs_set_root_dirid(&root_item, new_dirid);
506 key.objectid = objectid;
508 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
509 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
514 key.offset = (u64)-1;
515 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
516 if (IS_ERR(new_root)) {
517 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
518 ret = PTR_ERR(new_root);
522 btrfs_record_root_in_trans(trans, new_root);
524 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
526 /* We potentially lose an unused inode item here */
527 btrfs_abort_transaction(trans, root, ret);
532 * insert the directory item
534 ret = btrfs_set_inode_index(dir, &index);
536 btrfs_abort_transaction(trans, root, ret);
540 ret = btrfs_insert_dir_item(trans, root,
541 name, namelen, dir, &key,
542 BTRFS_FT_DIR, index);
544 btrfs_abort_transaction(trans, root, ret);
548 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
549 ret = btrfs_update_inode(trans, root, dir);
552 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
553 objectid, root->root_key.objectid,
554 btrfs_ino(dir), index, name, namelen);
557 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
558 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
561 btrfs_abort_transaction(trans, root, ret);
564 trans->block_rsv = NULL;
565 trans->bytes_reserved = 0;
566 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
569 *async_transid = trans->transid;
570 err = btrfs_commit_transaction_async(trans, root, 1);
572 err = btrfs_commit_transaction(trans, root);
574 err = btrfs_commit_transaction(trans, root);
580 inode = btrfs_lookup_dentry(dir, dentry);
582 return PTR_ERR(inode);
583 d_instantiate(dentry, inode);
588 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
589 struct dentry *dentry, char *name, int namelen,
590 u64 *async_transid, bool readonly,
591 struct btrfs_qgroup_inherit *inherit)
594 struct btrfs_pending_snapshot *pending_snapshot;
595 struct btrfs_trans_handle *trans;
601 ret = btrfs_start_delalloc_inodes(root, 0);
605 btrfs_wait_ordered_extents(root, -1);
607 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
608 if (!pending_snapshot)
611 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
612 BTRFS_BLOCK_RSV_TEMP);
614 * 1 - parent dir inode
617 * 2 - root ref/backref
618 * 1 - root of snapshot
621 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
622 &pending_snapshot->block_rsv, 8,
623 &pending_snapshot->qgroup_reserved,
628 pending_snapshot->dentry = dentry;
629 pending_snapshot->root = root;
630 pending_snapshot->readonly = readonly;
631 pending_snapshot->dir = dir;
632 pending_snapshot->inherit = inherit;
634 trans = btrfs_start_transaction(root, 0);
636 ret = PTR_ERR(trans);
640 spin_lock(&root->fs_info->trans_lock);
641 list_add(&pending_snapshot->list,
642 &trans->transaction->pending_snapshots);
643 spin_unlock(&root->fs_info->trans_lock);
645 *async_transid = trans->transid;
646 ret = btrfs_commit_transaction_async(trans,
647 root->fs_info->extent_root, 1);
649 ret = btrfs_commit_transaction(trans, root);
651 ret = btrfs_commit_transaction(trans,
652 root->fs_info->extent_root);
657 ret = pending_snapshot->error;
661 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
665 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
667 ret = PTR_ERR(inode);
671 d_instantiate(dentry, inode);
674 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
675 &pending_snapshot->block_rsv,
676 pending_snapshot->qgroup_reserved);
678 kfree(pending_snapshot);
682 /* copy of check_sticky in fs/namei.c()
683 * It's inline, so penalty for filesystems that don't use sticky bit is
686 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
688 kuid_t fsuid = current_fsuid();
690 if (!(dir->i_mode & S_ISVTX))
692 if (uid_eq(inode->i_uid, fsuid))
694 if (uid_eq(dir->i_uid, fsuid))
696 return !capable(CAP_FOWNER);
699 /* copy of may_delete in fs/namei.c()
700 * Check whether we can remove a link victim from directory dir, check
701 * whether the type of victim is right.
702 * 1. We can't do it if dir is read-only (done in permission())
703 * 2. We should have write and exec permissions on dir
704 * 3. We can't remove anything from append-only dir
705 * 4. We can't do anything with immutable dir (done in permission())
706 * 5. If the sticky bit on dir is set we should either
707 * a. be owner of dir, or
708 * b. be owner of victim, or
709 * c. have CAP_FOWNER capability
710 * 6. If the victim is append-only or immutable we can't do antyhing with
711 * links pointing to it.
712 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
713 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
714 * 9. We can't remove a root or mountpoint.
715 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
716 * nfs_async_unlink().
719 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
723 if (!victim->d_inode)
726 BUG_ON(victim->d_parent->d_inode != dir);
727 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
729 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
734 if (btrfs_check_sticky(dir, victim->d_inode)||
735 IS_APPEND(victim->d_inode)||
736 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
739 if (!S_ISDIR(victim->d_inode->i_mode))
743 } else if (S_ISDIR(victim->d_inode->i_mode))
747 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
752 /* copy of may_create in fs/namei.c() */
753 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
759 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
763 * Create a new subvolume below @parent. This is largely modeled after
764 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
765 * inside this filesystem so it's quite a bit simpler.
767 static noinline int btrfs_mksubvol(struct path *parent,
768 char *name, int namelen,
769 struct btrfs_root *snap_src,
770 u64 *async_transid, bool readonly,
771 struct btrfs_qgroup_inherit *inherit)
773 struct inode *dir = parent->dentry->d_inode;
774 struct dentry *dentry;
777 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
781 dentry = lookup_one_len(name, parent->dentry, namelen);
782 error = PTR_ERR(dentry);
790 error = btrfs_may_create(dir, dentry);
795 * even if this name doesn't exist, we may get hash collisions.
796 * check for them now when we can safely fail
798 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
804 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
806 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
810 error = create_snapshot(snap_src, dir, dentry, name, namelen,
811 async_transid, readonly, inherit);
813 error = create_subvol(dir, dentry, name, namelen,
814 async_transid, inherit);
817 fsnotify_mkdir(dir, dentry);
819 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
823 mutex_unlock(&dir->i_mutex);
828 * When we're defragging a range, we don't want to kick it off again
829 * if it is really just waiting for delalloc to send it down.
830 * If we find a nice big extent or delalloc range for the bytes in the
831 * file you want to defrag, we return 0 to let you know to skip this
834 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
836 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
837 struct extent_map *em = NULL;
838 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
841 read_lock(&em_tree->lock);
842 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
843 read_unlock(&em_tree->lock);
846 end = extent_map_end(em);
848 if (end - offset > thresh)
851 /* if we already have a nice delalloc here, just stop */
853 end = count_range_bits(io_tree, &offset, offset + thresh,
854 thresh, EXTENT_DELALLOC, 1);
861 * helper function to walk through a file and find extents
862 * newer than a specific transid, and smaller than thresh.
864 * This is used by the defragging code to find new and small
867 static int find_new_extents(struct btrfs_root *root,
868 struct inode *inode, u64 newer_than,
869 u64 *off, int thresh)
871 struct btrfs_path *path;
872 struct btrfs_key min_key;
873 struct extent_buffer *leaf;
874 struct btrfs_file_extent_item *extent;
877 u64 ino = btrfs_ino(inode);
879 path = btrfs_alloc_path();
883 min_key.objectid = ino;
884 min_key.type = BTRFS_EXTENT_DATA_KEY;
885 min_key.offset = *off;
887 path->keep_locks = 1;
890 ret = btrfs_search_forward(root, &min_key, path, newer_than);
893 if (min_key.objectid != ino)
895 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
898 leaf = path->nodes[0];
899 extent = btrfs_item_ptr(leaf, path->slots[0],
900 struct btrfs_file_extent_item);
902 type = btrfs_file_extent_type(leaf, extent);
903 if (type == BTRFS_FILE_EXTENT_REG &&
904 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
905 check_defrag_in_cache(inode, min_key.offset, thresh)) {
906 *off = min_key.offset;
907 btrfs_free_path(path);
911 if (min_key.offset == (u64)-1)
915 btrfs_release_path(path);
918 btrfs_free_path(path);
922 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
924 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
925 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
926 struct extent_map *em;
927 u64 len = PAGE_CACHE_SIZE;
930 * hopefully we have this extent in the tree already, try without
931 * the full extent lock
933 read_lock(&em_tree->lock);
934 em = lookup_extent_mapping(em_tree, start, len);
935 read_unlock(&em_tree->lock);
938 /* get the big lock and read metadata off disk */
939 lock_extent(io_tree, start, start + len - 1);
940 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
941 unlock_extent(io_tree, start, start + len - 1);
950 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
952 struct extent_map *next;
955 /* this is the last extent */
956 if (em->start + em->len >= i_size_read(inode))
959 next = defrag_lookup_extent(inode, em->start + em->len);
960 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
963 free_extent_map(next);
967 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
968 u64 *last_len, u64 *skip, u64 *defrag_end,
971 struct extent_map *em;
973 bool next_mergeable = true;
976 * make sure that once we start defragging an extent, we keep on
979 if (start < *defrag_end)
984 em = defrag_lookup_extent(inode, start);
988 /* this will cover holes, and inline extents */
989 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
994 next_mergeable = defrag_check_next_extent(inode, em);
997 * we hit a real extent, if it is big or the next extent is not a
998 * real extent, don't bother defragging it
1000 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1001 (em->len >= thresh || !next_mergeable))
1005 * last_len ends up being a counter of how many bytes we've defragged.
1006 * every time we choose not to defrag an extent, we reset *last_len
1007 * so that the next tiny extent will force a defrag.
1009 * The end result of this is that tiny extents before a single big
1010 * extent will force at least part of that big extent to be defragged.
1013 *defrag_end = extent_map_end(em);
1016 *skip = extent_map_end(em);
1020 free_extent_map(em);
1025 * it doesn't do much good to defrag one or two pages
1026 * at a time. This pulls in a nice chunk of pages
1027 * to COW and defrag.
1029 * It also makes sure the delalloc code has enough
1030 * dirty data to avoid making new small extents as part
1033 * It's a good idea to start RA on this range
1034 * before calling this.
1036 static int cluster_pages_for_defrag(struct inode *inode,
1037 struct page **pages,
1038 unsigned long start_index,
1039 unsigned long num_pages)
1041 unsigned long file_end;
1042 u64 isize = i_size_read(inode);
1049 struct btrfs_ordered_extent *ordered;
1050 struct extent_state *cached_state = NULL;
1051 struct extent_io_tree *tree;
1052 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1054 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1055 if (!isize || start_index > file_end)
1058 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1060 ret = btrfs_delalloc_reserve_space(inode,
1061 page_cnt << PAGE_CACHE_SHIFT);
1065 tree = &BTRFS_I(inode)->io_tree;
1067 /* step one, lock all the pages */
1068 for (i = 0; i < page_cnt; i++) {
1071 page = find_or_create_page(inode->i_mapping,
1072 start_index + i, mask);
1076 page_start = page_offset(page);
1077 page_end = page_start + PAGE_CACHE_SIZE - 1;
1079 lock_extent(tree, page_start, page_end);
1080 ordered = btrfs_lookup_ordered_extent(inode,
1082 unlock_extent(tree, page_start, page_end);
1087 btrfs_start_ordered_extent(inode, ordered, 1);
1088 btrfs_put_ordered_extent(ordered);
1091 * we unlocked the page above, so we need check if
1092 * it was released or not.
1094 if (page->mapping != inode->i_mapping) {
1096 page_cache_release(page);
1101 if (!PageUptodate(page)) {
1102 btrfs_readpage(NULL, page);
1104 if (!PageUptodate(page)) {
1106 page_cache_release(page);
1112 if (page->mapping != inode->i_mapping) {
1114 page_cache_release(page);
1124 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1128 * so now we have a nice long stream of locked
1129 * and up to date pages, lets wait on them
1131 for (i = 0; i < i_done; i++)
1132 wait_on_page_writeback(pages[i]);
1134 page_start = page_offset(pages[0]);
1135 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1137 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1138 page_start, page_end - 1, 0, &cached_state);
1139 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1140 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1141 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1142 &cached_state, GFP_NOFS);
1144 if (i_done != page_cnt) {
1145 spin_lock(&BTRFS_I(inode)->lock);
1146 BTRFS_I(inode)->outstanding_extents++;
1147 spin_unlock(&BTRFS_I(inode)->lock);
1148 btrfs_delalloc_release_space(inode,
1149 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1153 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1154 &cached_state, GFP_NOFS);
1156 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1157 page_start, page_end - 1, &cached_state,
1160 for (i = 0; i < i_done; i++) {
1161 clear_page_dirty_for_io(pages[i]);
1162 ClearPageChecked(pages[i]);
1163 set_page_extent_mapped(pages[i]);
1164 set_page_dirty(pages[i]);
1165 unlock_page(pages[i]);
1166 page_cache_release(pages[i]);
1170 for (i = 0; i < i_done; i++) {
1171 unlock_page(pages[i]);
1172 page_cache_release(pages[i]);
1174 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1179 int btrfs_defrag_file(struct inode *inode, struct file *file,
1180 struct btrfs_ioctl_defrag_range_args *range,
1181 u64 newer_than, unsigned long max_to_defrag)
1183 struct btrfs_root *root = BTRFS_I(inode)->root;
1184 struct file_ra_state *ra = NULL;
1185 unsigned long last_index;
1186 u64 isize = i_size_read(inode);
1190 u64 newer_off = range->start;
1192 unsigned long ra_index = 0;
1194 int defrag_count = 0;
1195 int compress_type = BTRFS_COMPRESS_ZLIB;
1196 int extent_thresh = range->extent_thresh;
1197 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1198 unsigned long cluster = max_cluster;
1199 u64 new_align = ~((u64)128 * 1024 - 1);
1200 struct page **pages = NULL;
1205 if (range->start >= isize)
1208 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1209 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1211 if (range->compress_type)
1212 compress_type = range->compress_type;
1215 if (extent_thresh == 0)
1216 extent_thresh = 256 * 1024;
1219 * if we were not given a file, allocate a readahead
1223 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1226 file_ra_state_init(ra, inode->i_mapping);
1231 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1238 /* find the last page to defrag */
1239 if (range->start + range->len > range->start) {
1240 last_index = min_t(u64, isize - 1,
1241 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1243 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1247 ret = find_new_extents(root, inode, newer_than,
1248 &newer_off, 64 * 1024);
1250 range->start = newer_off;
1252 * we always align our defrag to help keep
1253 * the extents in the file evenly spaced
1255 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1259 i = range->start >> PAGE_CACHE_SHIFT;
1262 max_to_defrag = last_index + 1;
1265 * make writeback starts from i, so the defrag range can be
1266 * written sequentially.
1268 if (i < inode->i_mapping->writeback_index)
1269 inode->i_mapping->writeback_index = i;
1271 while (i <= last_index && defrag_count < max_to_defrag &&
1272 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1273 PAGE_CACHE_SHIFT)) {
1275 * make sure we stop running if someone unmounts
1278 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1281 if (btrfs_defrag_cancelled(root->fs_info)) {
1282 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1287 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1288 extent_thresh, &last_len, &skip,
1289 &defrag_end, range->flags &
1290 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1293 * the should_defrag function tells us how much to skip
1294 * bump our counter by the suggested amount
1296 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1297 i = max(i + 1, next);
1302 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1303 PAGE_CACHE_SHIFT) - i;
1304 cluster = min(cluster, max_cluster);
1306 cluster = max_cluster;
1309 if (i + cluster > ra_index) {
1310 ra_index = max(i, ra_index);
1311 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1313 ra_index += max_cluster;
1316 mutex_lock(&inode->i_mutex);
1317 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1318 BTRFS_I(inode)->force_compress = compress_type;
1319 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1321 mutex_unlock(&inode->i_mutex);
1325 defrag_count += ret;
1326 balance_dirty_pages_ratelimited(inode->i_mapping);
1327 mutex_unlock(&inode->i_mutex);
1330 if (newer_off == (u64)-1)
1336 newer_off = max(newer_off + 1,
1337 (u64)i << PAGE_CACHE_SHIFT);
1339 ret = find_new_extents(root, inode,
1340 newer_than, &newer_off,
1343 range->start = newer_off;
1344 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1351 last_len += ret << PAGE_CACHE_SHIFT;
1359 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1360 filemap_flush(inode->i_mapping);
1362 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1363 /* the filemap_flush will queue IO into the worker threads, but
1364 * we have to make sure the IO is actually started and that
1365 * ordered extents get created before we return
1367 atomic_inc(&root->fs_info->async_submit_draining);
1368 while (atomic_read(&root->fs_info->nr_async_submits) ||
1369 atomic_read(&root->fs_info->async_delalloc_pages)) {
1370 wait_event(root->fs_info->async_submit_wait,
1371 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1372 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1374 atomic_dec(&root->fs_info->async_submit_draining);
1377 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1378 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1384 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1385 mutex_lock(&inode->i_mutex);
1386 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1387 mutex_unlock(&inode->i_mutex);
1395 static noinline int btrfs_ioctl_resize(struct file *file,
1401 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1402 struct btrfs_ioctl_vol_args *vol_args;
1403 struct btrfs_trans_handle *trans;
1404 struct btrfs_device *device = NULL;
1406 char *devstr = NULL;
1410 if (!capable(CAP_SYS_ADMIN))
1413 ret = mnt_want_write_file(file);
1417 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1419 mnt_drop_write_file(file);
1420 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1423 mutex_lock(&root->fs_info->volume_mutex);
1424 vol_args = memdup_user(arg, sizeof(*vol_args));
1425 if (IS_ERR(vol_args)) {
1426 ret = PTR_ERR(vol_args);
1430 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1432 sizestr = vol_args->name;
1433 devstr = strchr(sizestr, ':');
1436 sizestr = devstr + 1;
1438 devstr = vol_args->name;
1439 devid = simple_strtoull(devstr, &end, 10);
1444 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1447 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1449 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1455 if (!device->writeable) {
1456 btrfs_info(root->fs_info,
1457 "resizer unable to apply on readonly device %llu",
1463 if (!strcmp(sizestr, "max"))
1464 new_size = device->bdev->bd_inode->i_size;
1466 if (sizestr[0] == '-') {
1469 } else if (sizestr[0] == '+') {
1473 new_size = memparse(sizestr, NULL);
1474 if (new_size == 0) {
1480 if (device->is_tgtdev_for_dev_replace) {
1485 old_size = device->total_bytes;
1488 if (new_size > old_size) {
1492 new_size = old_size - new_size;
1493 } else if (mod > 0) {
1494 if (new_size > ULLONG_MAX - old_size) {
1498 new_size = old_size + new_size;
1501 if (new_size < 256 * 1024 * 1024) {
1505 if (new_size > device->bdev->bd_inode->i_size) {
1510 do_div(new_size, root->sectorsize);
1511 new_size *= root->sectorsize;
1513 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1514 rcu_str_deref(device->name), new_size);
1516 if (new_size > old_size) {
1517 trans = btrfs_start_transaction(root, 0);
1518 if (IS_ERR(trans)) {
1519 ret = PTR_ERR(trans);
1522 ret = btrfs_grow_device(trans, device, new_size);
1523 btrfs_commit_transaction(trans, root);
1524 } else if (new_size < old_size) {
1525 ret = btrfs_shrink_device(device, new_size);
1526 } /* equal, nothing need to do */
1531 mutex_unlock(&root->fs_info->volume_mutex);
1532 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1533 mnt_drop_write_file(file);
1537 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1538 char *name, unsigned long fd, int subvol,
1539 u64 *transid, bool readonly,
1540 struct btrfs_qgroup_inherit *inherit)
1545 ret = mnt_want_write_file(file);
1549 namelen = strlen(name);
1550 if (strchr(name, '/')) {
1552 goto out_drop_write;
1555 if (name[0] == '.' &&
1556 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1558 goto out_drop_write;
1562 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1563 NULL, transid, readonly, inherit);
1565 struct fd src = fdget(fd);
1566 struct inode *src_inode;
1569 goto out_drop_write;
1572 src_inode = file_inode(src.file);
1573 if (src_inode->i_sb != file_inode(file)->i_sb) {
1574 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1575 "Snapshot src from another FS");
1577 } else if (!inode_owner_or_capable(src_inode)) {
1579 * Subvolume creation is not restricted, but snapshots
1580 * are limited to own subvolumes only
1584 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1585 BTRFS_I(src_inode)->root,
1586 transid, readonly, inherit);
1591 mnt_drop_write_file(file);
1596 static noinline int btrfs_ioctl_snap_create(struct file *file,
1597 void __user *arg, int subvol)
1599 struct btrfs_ioctl_vol_args *vol_args;
1602 vol_args = memdup_user(arg, sizeof(*vol_args));
1603 if (IS_ERR(vol_args))
1604 return PTR_ERR(vol_args);
1605 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1607 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1608 vol_args->fd, subvol,
1615 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1616 void __user *arg, int subvol)
1618 struct btrfs_ioctl_vol_args_v2 *vol_args;
1622 bool readonly = false;
1623 struct btrfs_qgroup_inherit *inherit = NULL;
1625 vol_args = memdup_user(arg, sizeof(*vol_args));
1626 if (IS_ERR(vol_args))
1627 return PTR_ERR(vol_args);
1628 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1630 if (vol_args->flags &
1631 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1632 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1637 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1639 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1641 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1642 if (vol_args->size > PAGE_CACHE_SIZE) {
1646 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1647 if (IS_ERR(inherit)) {
1648 ret = PTR_ERR(inherit);
1653 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1654 vol_args->fd, subvol, ptr,
1657 if (ret == 0 && ptr &&
1659 offsetof(struct btrfs_ioctl_vol_args_v2,
1660 transid), ptr, sizeof(*ptr)))
1668 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1671 struct inode *inode = file_inode(file);
1672 struct btrfs_root *root = BTRFS_I(inode)->root;
1676 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1679 down_read(&root->fs_info->subvol_sem);
1680 if (btrfs_root_readonly(root))
1681 flags |= BTRFS_SUBVOL_RDONLY;
1682 up_read(&root->fs_info->subvol_sem);
1684 if (copy_to_user(arg, &flags, sizeof(flags)))
1690 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1693 struct inode *inode = file_inode(file);
1694 struct btrfs_root *root = BTRFS_I(inode)->root;
1695 struct btrfs_trans_handle *trans;
1700 if (!inode_owner_or_capable(inode))
1703 ret = mnt_want_write_file(file);
1707 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1709 goto out_drop_write;
1712 if (copy_from_user(&flags, arg, sizeof(flags))) {
1714 goto out_drop_write;
1717 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1719 goto out_drop_write;
1722 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1724 goto out_drop_write;
1727 down_write(&root->fs_info->subvol_sem);
1730 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1733 root_flags = btrfs_root_flags(&root->root_item);
1734 if (flags & BTRFS_SUBVOL_RDONLY) {
1735 btrfs_set_root_flags(&root->root_item,
1736 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1739 * Block RO -> RW transition if this subvolume is involved in
1742 spin_lock(&root->root_item_lock);
1743 if (root->send_in_progress == 0) {
1744 btrfs_set_root_flags(&root->root_item,
1745 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1746 spin_unlock(&root->root_item_lock);
1748 spin_unlock(&root->root_item_lock);
1749 btrfs_warn(root->fs_info,
1750 "Attempt to set subvolume %llu read-write during send",
1751 root->root_key.objectid);
1757 trans = btrfs_start_transaction(root, 1);
1758 if (IS_ERR(trans)) {
1759 ret = PTR_ERR(trans);
1763 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1764 &root->root_key, &root->root_item);
1766 btrfs_commit_transaction(trans, root);
1769 btrfs_set_root_flags(&root->root_item, root_flags);
1771 up_write(&root->fs_info->subvol_sem);
1773 mnt_drop_write_file(file);
1779 * helper to check if the subvolume references other subvolumes
1781 static noinline int may_destroy_subvol(struct btrfs_root *root)
1783 struct btrfs_path *path;
1784 struct btrfs_dir_item *di;
1785 struct btrfs_key key;
1789 path = btrfs_alloc_path();
1793 /* Make sure this root isn't set as the default subvol */
1794 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1795 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1796 dir_id, "default", 7, 0);
1797 if (di && !IS_ERR(di)) {
1798 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1799 if (key.objectid == root->root_key.objectid) {
1803 btrfs_release_path(path);
1806 key.objectid = root->root_key.objectid;
1807 key.type = BTRFS_ROOT_REF_KEY;
1808 key.offset = (u64)-1;
1810 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1817 if (path->slots[0] > 0) {
1819 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1820 if (key.objectid == root->root_key.objectid &&
1821 key.type == BTRFS_ROOT_REF_KEY)
1825 btrfs_free_path(path);
1829 static noinline int key_in_sk(struct btrfs_key *key,
1830 struct btrfs_ioctl_search_key *sk)
1832 struct btrfs_key test;
1835 test.objectid = sk->min_objectid;
1836 test.type = sk->min_type;
1837 test.offset = sk->min_offset;
1839 ret = btrfs_comp_cpu_keys(key, &test);
1843 test.objectid = sk->max_objectid;
1844 test.type = sk->max_type;
1845 test.offset = sk->max_offset;
1847 ret = btrfs_comp_cpu_keys(key, &test);
1853 static noinline int copy_to_sk(struct btrfs_root *root,
1854 struct btrfs_path *path,
1855 struct btrfs_key *key,
1856 struct btrfs_ioctl_search_key *sk,
1858 unsigned long *sk_offset,
1862 struct extent_buffer *leaf;
1863 struct btrfs_ioctl_search_header sh;
1864 unsigned long item_off;
1865 unsigned long item_len;
1871 leaf = path->nodes[0];
1872 slot = path->slots[0];
1873 nritems = btrfs_header_nritems(leaf);
1875 if (btrfs_header_generation(leaf) > sk->max_transid) {
1879 found_transid = btrfs_header_generation(leaf);
1881 for (i = slot; i < nritems; i++) {
1882 item_off = btrfs_item_ptr_offset(leaf, i);
1883 item_len = btrfs_item_size_nr(leaf, i);
1885 btrfs_item_key_to_cpu(leaf, key, i);
1886 if (!key_in_sk(key, sk))
1889 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1892 if (sizeof(sh) + item_len + *sk_offset >
1893 BTRFS_SEARCH_ARGS_BUFSIZE) {
1898 sh.objectid = key->objectid;
1899 sh.offset = key->offset;
1900 sh.type = key->type;
1902 sh.transid = found_transid;
1904 /* copy search result header */
1905 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1906 *sk_offset += sizeof(sh);
1909 char *p = buf + *sk_offset;
1911 read_extent_buffer(leaf, p,
1912 item_off, item_len);
1913 *sk_offset += item_len;
1917 if (*num_found >= sk->nr_items)
1922 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1924 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1927 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1937 static noinline int search_ioctl(struct inode *inode,
1938 struct btrfs_ioctl_search_args *args)
1940 struct btrfs_root *root;
1941 struct btrfs_key key;
1942 struct btrfs_path *path;
1943 struct btrfs_ioctl_search_key *sk = &args->key;
1944 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1947 unsigned long sk_offset = 0;
1949 path = btrfs_alloc_path();
1953 if (sk->tree_id == 0) {
1954 /* search the root of the inode that was passed */
1955 root = BTRFS_I(inode)->root;
1957 key.objectid = sk->tree_id;
1958 key.type = BTRFS_ROOT_ITEM_KEY;
1959 key.offset = (u64)-1;
1960 root = btrfs_read_fs_root_no_name(info, &key);
1962 printk(KERN_ERR "BTRFS: could not find root %llu\n",
1964 btrfs_free_path(path);
1969 key.objectid = sk->min_objectid;
1970 key.type = sk->min_type;
1971 key.offset = sk->min_offset;
1973 path->keep_locks = 1;
1976 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
1982 ret = copy_to_sk(root, path, &key, sk, args->buf,
1983 &sk_offset, &num_found);
1984 btrfs_release_path(path);
1985 if (ret || num_found >= sk->nr_items)
1991 sk->nr_items = num_found;
1992 btrfs_free_path(path);
1996 static noinline int btrfs_ioctl_tree_search(struct file *file,
1999 struct btrfs_ioctl_search_args *args;
2000 struct inode *inode;
2003 if (!capable(CAP_SYS_ADMIN))
2006 args = memdup_user(argp, sizeof(*args));
2008 return PTR_ERR(args);
2010 inode = file_inode(file);
2011 ret = search_ioctl(inode, args);
2012 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2019 * Search INODE_REFs to identify path name of 'dirid' directory
2020 * in a 'tree_id' tree. and sets path name to 'name'.
2022 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2023 u64 tree_id, u64 dirid, char *name)
2025 struct btrfs_root *root;
2026 struct btrfs_key key;
2032 struct btrfs_inode_ref *iref;
2033 struct extent_buffer *l;
2034 struct btrfs_path *path;
2036 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2041 path = btrfs_alloc_path();
2045 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2047 key.objectid = tree_id;
2048 key.type = BTRFS_ROOT_ITEM_KEY;
2049 key.offset = (u64)-1;
2050 root = btrfs_read_fs_root_no_name(info, &key);
2052 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2057 key.objectid = dirid;
2058 key.type = BTRFS_INODE_REF_KEY;
2059 key.offset = (u64)-1;
2062 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2066 ret = btrfs_previous_item(root, path, dirid,
2067 BTRFS_INODE_REF_KEY);
2077 slot = path->slots[0];
2078 btrfs_item_key_to_cpu(l, &key, slot);
2080 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2081 len = btrfs_inode_ref_name_len(l, iref);
2083 total_len += len + 1;
2085 ret = -ENAMETOOLONG;
2090 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2092 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2095 btrfs_release_path(path);
2096 key.objectid = key.offset;
2097 key.offset = (u64)-1;
2098 dirid = key.objectid;
2100 memmove(name, ptr, total_len);
2101 name[total_len] = '\0';
2104 btrfs_free_path(path);
2108 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2111 struct btrfs_ioctl_ino_lookup_args *args;
2112 struct inode *inode;
2115 if (!capable(CAP_SYS_ADMIN))
2118 args = memdup_user(argp, sizeof(*args));
2120 return PTR_ERR(args);
2122 inode = file_inode(file);
2124 if (args->treeid == 0)
2125 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2127 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2128 args->treeid, args->objectid,
2131 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2138 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2141 struct dentry *parent = file->f_path.dentry;
2142 struct dentry *dentry;
2143 struct inode *dir = parent->d_inode;
2144 struct inode *inode;
2145 struct btrfs_root *root = BTRFS_I(dir)->root;
2146 struct btrfs_root *dest = NULL;
2147 struct btrfs_ioctl_vol_args *vol_args;
2148 struct btrfs_trans_handle *trans;
2149 struct btrfs_block_rsv block_rsv;
2150 u64 qgroup_reserved;
2155 vol_args = memdup_user(arg, sizeof(*vol_args));
2156 if (IS_ERR(vol_args))
2157 return PTR_ERR(vol_args);
2159 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2160 namelen = strlen(vol_args->name);
2161 if (strchr(vol_args->name, '/') ||
2162 strncmp(vol_args->name, "..", namelen) == 0) {
2167 err = mnt_want_write_file(file);
2171 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2173 goto out_drop_write;
2174 dentry = lookup_one_len(vol_args->name, parent, namelen);
2175 if (IS_ERR(dentry)) {
2176 err = PTR_ERR(dentry);
2177 goto out_unlock_dir;
2180 if (!dentry->d_inode) {
2185 inode = dentry->d_inode;
2186 dest = BTRFS_I(inode)->root;
2187 if (!capable(CAP_SYS_ADMIN)) {
2189 * Regular user. Only allow this with a special mount
2190 * option, when the user has write+exec access to the
2191 * subvol root, and when rmdir(2) would have been
2194 * Note that this is _not_ check that the subvol is
2195 * empty or doesn't contain data that we wouldn't
2196 * otherwise be able to delete.
2198 * Users who want to delete empty subvols should try
2202 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2206 * Do not allow deletion if the parent dir is the same
2207 * as the dir to be deleted. That means the ioctl
2208 * must be called on the dentry referencing the root
2209 * of the subvol, not a random directory contained
2216 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2221 /* check if subvolume may be deleted by a user */
2222 err = btrfs_may_delete(dir, dentry, 1);
2226 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2231 mutex_lock(&inode->i_mutex);
2232 err = d_invalidate(dentry);
2236 down_write(&root->fs_info->subvol_sem);
2238 err = may_destroy_subvol(dest);
2242 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2244 * One for dir inode, two for dir entries, two for root
2247 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2248 5, &qgroup_reserved, true);
2252 trans = btrfs_start_transaction(root, 0);
2253 if (IS_ERR(trans)) {
2254 err = PTR_ERR(trans);
2257 trans->block_rsv = &block_rsv;
2258 trans->bytes_reserved = block_rsv.size;
2260 ret = btrfs_unlink_subvol(trans, root, dir,
2261 dest->root_key.objectid,
2262 dentry->d_name.name,
2263 dentry->d_name.len);
2266 btrfs_abort_transaction(trans, root, ret);
2270 btrfs_record_root_in_trans(trans, dest);
2272 memset(&dest->root_item.drop_progress, 0,
2273 sizeof(dest->root_item.drop_progress));
2274 dest->root_item.drop_level = 0;
2275 btrfs_set_root_refs(&dest->root_item, 0);
2277 if (!xchg(&dest->orphan_item_inserted, 1)) {
2278 ret = btrfs_insert_orphan_item(trans,
2279 root->fs_info->tree_root,
2280 dest->root_key.objectid);
2282 btrfs_abort_transaction(trans, root, ret);
2288 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2289 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2290 dest->root_key.objectid);
2291 if (ret && ret != -ENOENT) {
2292 btrfs_abort_transaction(trans, root, ret);
2296 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2297 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2298 dest->root_item.received_uuid,
2299 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2300 dest->root_key.objectid);
2301 if (ret && ret != -ENOENT) {
2302 btrfs_abort_transaction(trans, root, ret);
2309 trans->block_rsv = NULL;
2310 trans->bytes_reserved = 0;
2311 ret = btrfs_end_transaction(trans, root);
2314 inode->i_flags |= S_DEAD;
2316 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2318 up_write(&root->fs_info->subvol_sem);
2320 mutex_unlock(&inode->i_mutex);
2322 shrink_dcache_sb(root->fs_info->sb);
2323 btrfs_invalidate_inodes(dest);
2327 if (dest->cache_inode) {
2328 iput(dest->cache_inode);
2329 dest->cache_inode = NULL;
2335 mutex_unlock(&dir->i_mutex);
2337 mnt_drop_write_file(file);
2343 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2345 struct inode *inode = file_inode(file);
2346 struct btrfs_root *root = BTRFS_I(inode)->root;
2347 struct btrfs_ioctl_defrag_range_args *range;
2350 ret = mnt_want_write_file(file);
2354 if (btrfs_root_readonly(root)) {
2359 switch (inode->i_mode & S_IFMT) {
2361 if (!capable(CAP_SYS_ADMIN)) {
2365 ret = btrfs_defrag_root(root);
2368 ret = btrfs_defrag_root(root->fs_info->extent_root);
2371 if (!(file->f_mode & FMODE_WRITE)) {
2376 range = kzalloc(sizeof(*range), GFP_KERNEL);
2383 if (copy_from_user(range, argp,
2389 /* compression requires us to start the IO */
2390 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2391 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2392 range->extent_thresh = (u32)-1;
2395 /* the rest are all set to zero by kzalloc */
2396 range->len = (u64)-1;
2398 ret = btrfs_defrag_file(file_inode(file), file,
2408 mnt_drop_write_file(file);
2412 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2414 struct btrfs_ioctl_vol_args *vol_args;
2417 if (!capable(CAP_SYS_ADMIN))
2420 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2422 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2425 mutex_lock(&root->fs_info->volume_mutex);
2426 vol_args = memdup_user(arg, sizeof(*vol_args));
2427 if (IS_ERR(vol_args)) {
2428 ret = PTR_ERR(vol_args);
2432 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2433 ret = btrfs_init_new_device(root, vol_args->name);
2437 mutex_unlock(&root->fs_info->volume_mutex);
2438 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2442 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2444 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2445 struct btrfs_ioctl_vol_args *vol_args;
2448 if (!capable(CAP_SYS_ADMIN))
2451 ret = mnt_want_write_file(file);
2455 vol_args = memdup_user(arg, sizeof(*vol_args));
2456 if (IS_ERR(vol_args)) {
2457 ret = PTR_ERR(vol_args);
2461 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2463 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2465 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2469 mutex_lock(&root->fs_info->volume_mutex);
2470 ret = btrfs_rm_device(root, vol_args->name);
2471 mutex_unlock(&root->fs_info->volume_mutex);
2472 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2476 mnt_drop_write_file(file);
2480 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2482 struct btrfs_ioctl_fs_info_args *fi_args;
2483 struct btrfs_device *device;
2484 struct btrfs_device *next;
2485 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2488 if (!capable(CAP_SYS_ADMIN))
2491 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2495 mutex_lock(&fs_devices->device_list_mutex);
2496 fi_args->num_devices = fs_devices->num_devices;
2497 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2499 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2500 if (device->devid > fi_args->max_id)
2501 fi_args->max_id = device->devid;
2503 mutex_unlock(&fs_devices->device_list_mutex);
2505 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2512 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2514 struct btrfs_ioctl_dev_info_args *di_args;
2515 struct btrfs_device *dev;
2516 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2518 char *s_uuid = NULL;
2520 if (!capable(CAP_SYS_ADMIN))
2523 di_args = memdup_user(arg, sizeof(*di_args));
2524 if (IS_ERR(di_args))
2525 return PTR_ERR(di_args);
2527 if (!btrfs_is_empty_uuid(di_args->uuid))
2528 s_uuid = di_args->uuid;
2530 mutex_lock(&fs_devices->device_list_mutex);
2531 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2538 di_args->devid = dev->devid;
2539 di_args->bytes_used = dev->bytes_used;
2540 di_args->total_bytes = dev->total_bytes;
2541 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2543 struct rcu_string *name;
2546 name = rcu_dereference(dev->name);
2547 strncpy(di_args->path, name->str, sizeof(di_args->path));
2549 di_args->path[sizeof(di_args->path) - 1] = 0;
2551 di_args->path[0] = '\0';
2555 mutex_unlock(&fs_devices->device_list_mutex);
2556 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2563 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2567 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2569 index = off >> PAGE_CACHE_SHIFT;
2571 page = grab_cache_page(inode->i_mapping, index);
2575 if (!PageUptodate(page)) {
2576 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2580 if (!PageUptodate(page)) {
2582 page_cache_release(page);
2591 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2593 /* do any pending delalloc/csum calc on src, one way or
2594 another, and lock file content */
2596 struct btrfs_ordered_extent *ordered;
2597 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2598 ordered = btrfs_lookup_first_ordered_extent(inode,
2601 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2602 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2604 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2606 btrfs_put_ordered_extent(ordered);
2607 btrfs_wait_ordered_range(inode, off, len);
2611 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2612 struct inode *inode2, u64 loff2, u64 len)
2614 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2615 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2617 mutex_unlock(&inode1->i_mutex);
2618 mutex_unlock(&inode2->i_mutex);
2621 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2622 struct inode *inode2, u64 loff2, u64 len)
2624 if (inode1 < inode2) {
2625 swap(inode1, inode2);
2629 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2630 lock_extent_range(inode1, loff1, len);
2631 if (inode1 != inode2) {
2632 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2633 lock_extent_range(inode2, loff2, len);
2637 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2638 u64 dst_loff, u64 len)
2641 struct page *src_page, *dst_page;
2642 unsigned int cmp_len = PAGE_CACHE_SIZE;
2643 void *addr, *dst_addr;
2646 if (len < PAGE_CACHE_SIZE)
2649 src_page = extent_same_get_page(src, loff);
2652 dst_page = extent_same_get_page(dst, dst_loff);
2654 page_cache_release(src_page);
2657 addr = kmap_atomic(src_page);
2658 dst_addr = kmap_atomic(dst_page);
2660 flush_dcache_page(src_page);
2661 flush_dcache_page(dst_page);
2663 if (memcmp(addr, dst_addr, cmp_len))
2664 ret = BTRFS_SAME_DATA_DIFFERS;
2666 kunmap_atomic(addr);
2667 kunmap_atomic(dst_addr);
2668 page_cache_release(src_page);
2669 page_cache_release(dst_page);
2675 dst_loff += cmp_len;
2682 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2684 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2686 if (off + len > inode->i_size || off + len < off)
2688 /* Check that we are block aligned - btrfs_clone() requires this */
2689 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2695 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2696 struct inode *dst, u64 dst_loff)
2701 * btrfs_clone() can't handle extents in the same file
2702 * yet. Once that works, we can drop this check and replace it
2703 * with a check for the same inode, but overlapping extents.
2708 btrfs_double_lock(src, loff, dst, dst_loff, len);
2710 ret = extent_same_check_offsets(src, loff, len);
2714 ret = extent_same_check_offsets(dst, dst_loff, len);
2718 /* don't make the dst file partly checksummed */
2719 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2720 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2725 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2727 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2730 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2735 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2737 static long btrfs_ioctl_file_extent_same(struct file *file,
2740 struct btrfs_ioctl_same_args tmp;
2741 struct btrfs_ioctl_same_args *same;
2742 struct btrfs_ioctl_same_extent_info *info;
2743 struct inode *src = file->f_dentry->d_inode;
2744 struct file *dst_file = NULL;
2751 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2752 bool is_admin = capable(CAP_SYS_ADMIN);
2754 if (!(file->f_mode & FMODE_READ))
2757 ret = mnt_want_write_file(file);
2761 if (copy_from_user(&tmp,
2762 (struct btrfs_ioctl_same_args __user *)argp,
2768 size = sizeof(tmp) +
2769 tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info);
2771 same = memdup_user((struct btrfs_ioctl_same_args __user *)argp, size);
2774 ret = PTR_ERR(same);
2778 off = same->logical_offset;
2782 * Limit the total length we will dedupe for each operation.
2783 * This is intended to bound the total time spent in this
2784 * ioctl to something sane.
2786 if (len > BTRFS_MAX_DEDUPE_LEN)
2787 len = BTRFS_MAX_DEDUPE_LEN;
2789 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2791 * Btrfs does not support blocksize < page_size. As a
2792 * result, btrfs_cmp_data() won't correctly handle
2793 * this situation without an update.
2800 if (S_ISDIR(src->i_mode))
2804 if (!S_ISREG(src->i_mode))
2807 /* pre-format output fields to sane values */
2808 for (i = 0; i < same->dest_count; i++) {
2809 same->info[i].bytes_deduped = 0ULL;
2810 same->info[i].status = 0;
2814 for (i = 0; i < same->dest_count; i++) {
2815 info = &same->info[i];
2817 dst_file = fget(info->fd);
2819 info->status = -EBADF;
2823 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
2824 info->status = -EINVAL;
2828 info->status = -EXDEV;
2829 if (file->f_path.mnt != dst_file->f_path.mnt)
2832 dst = dst_file->f_dentry->d_inode;
2833 if (src->i_sb != dst->i_sb)
2836 if (S_ISDIR(dst->i_mode)) {
2837 info->status = -EISDIR;
2841 if (!S_ISREG(dst->i_mode)) {
2842 info->status = -EACCES;
2846 info->status = btrfs_extent_same(src, off, len, dst,
2847 info->logical_offset);
2848 if (info->status == 0)
2849 info->bytes_deduped += len;
2856 ret = copy_to_user(argp, same, size);
2861 mnt_drop_write_file(file);
2866 * btrfs_clone() - clone a range from inode file to another
2868 * @src: Inode to clone from
2869 * @inode: Inode to clone to
2870 * @off: Offset within source to start clone from
2871 * @olen: Original length, passed by user, of range to clone
2872 * @olen_aligned: Block-aligned value of olen, extent_same uses
2873 * identical values here
2874 * @destoff: Offset within @inode to start clone
2876 static int btrfs_clone(struct inode *src, struct inode *inode,
2877 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
2879 struct btrfs_root *root = BTRFS_I(inode)->root;
2880 struct btrfs_path *path = NULL;
2881 struct extent_buffer *leaf;
2882 struct btrfs_trans_handle *trans;
2884 struct btrfs_key key;
2888 u64 len = olen_aligned;
2891 buf = vmalloc(btrfs_level_size(root, 0));
2895 path = btrfs_alloc_path();
2903 key.objectid = btrfs_ino(src);
2904 key.type = BTRFS_EXTENT_DATA_KEY;
2909 * note the key will change type as we walk through the
2912 path->leave_spinning = 1;
2913 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2918 nritems = btrfs_header_nritems(path->nodes[0]);
2920 if (path->slots[0] >= nritems) {
2921 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2926 nritems = btrfs_header_nritems(path->nodes[0]);
2928 leaf = path->nodes[0];
2929 slot = path->slots[0];
2931 btrfs_item_key_to_cpu(leaf, &key, slot);
2932 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2933 key.objectid != btrfs_ino(src))
2936 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2937 struct btrfs_file_extent_item *extent;
2940 struct btrfs_key new_key;
2941 u64 disko = 0, diskl = 0;
2942 u64 datao = 0, datal = 0;
2946 extent = btrfs_item_ptr(leaf, slot,
2947 struct btrfs_file_extent_item);
2948 comp = btrfs_file_extent_compression(leaf, extent);
2949 type = btrfs_file_extent_type(leaf, extent);
2950 if (type == BTRFS_FILE_EXTENT_REG ||
2951 type == BTRFS_FILE_EXTENT_PREALLOC) {
2952 disko = btrfs_file_extent_disk_bytenr(leaf,
2954 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2956 datao = btrfs_file_extent_offset(leaf, extent);
2957 datal = btrfs_file_extent_num_bytes(leaf,
2959 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2960 /* take upper bound, may be compressed */
2961 datal = btrfs_file_extent_ram_bytes(leaf,
2965 if (key.offset + datal <= off ||
2966 key.offset >= off + len - 1) {
2971 size = btrfs_item_size_nr(leaf, slot);
2972 read_extent_buffer(leaf, buf,
2973 btrfs_item_ptr_offset(leaf, slot),
2976 btrfs_release_path(path);
2977 path->leave_spinning = 0;
2979 memcpy(&new_key, &key, sizeof(new_key));
2980 new_key.objectid = btrfs_ino(inode);
2981 if (off <= key.offset)
2982 new_key.offset = key.offset + destoff - off;
2984 new_key.offset = destoff;
2987 * 1 - adjusting old extent (we may have to split it)
2988 * 1 - add new extent
2991 trans = btrfs_start_transaction(root, 3);
2992 if (IS_ERR(trans)) {
2993 ret = PTR_ERR(trans);
2997 if (type == BTRFS_FILE_EXTENT_REG ||
2998 type == BTRFS_FILE_EXTENT_PREALLOC) {
3000 * a | --- range to clone ---| b
3001 * | ------------- extent ------------- |
3004 /* substract range b */
3005 if (key.offset + datal > off + len)
3006 datal = off + len - key.offset;
3008 /* substract range a */
3009 if (off > key.offset) {
3010 datao += off - key.offset;
3011 datal -= off - key.offset;
3014 ret = btrfs_drop_extents(trans, root, inode,
3016 new_key.offset + datal,
3019 btrfs_abort_transaction(trans, root,
3021 btrfs_end_transaction(trans, root);
3025 ret = btrfs_insert_empty_item(trans, root, path,
3028 btrfs_abort_transaction(trans, root,
3030 btrfs_end_transaction(trans, root);
3034 leaf = path->nodes[0];
3035 slot = path->slots[0];
3036 write_extent_buffer(leaf, buf,
3037 btrfs_item_ptr_offset(leaf, slot),
3040 extent = btrfs_item_ptr(leaf, slot,
3041 struct btrfs_file_extent_item);
3043 /* disko == 0 means it's a hole */
3047 btrfs_set_file_extent_offset(leaf, extent,
3049 btrfs_set_file_extent_num_bytes(leaf, extent,
3052 inode_add_bytes(inode, datal);
3053 ret = btrfs_inc_extent_ref(trans, root,
3055 root->root_key.objectid,
3057 new_key.offset - datao,
3060 btrfs_abort_transaction(trans,
3063 btrfs_end_transaction(trans,
3069 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3072 if (off > key.offset) {
3073 skip = off - key.offset;
3074 new_key.offset += skip;
3077 if (key.offset + datal > off + len)
3078 trim = key.offset + datal - (off + len);
3080 if (comp && (skip || trim)) {
3082 btrfs_end_transaction(trans, root);
3085 size -= skip + trim;
3086 datal -= skip + trim;
3088 ret = btrfs_drop_extents(trans, root, inode,
3090 new_key.offset + datal,
3093 btrfs_abort_transaction(trans, root,
3095 btrfs_end_transaction(trans, root);
3099 ret = btrfs_insert_empty_item(trans, root, path,
3102 btrfs_abort_transaction(trans, root,
3104 btrfs_end_transaction(trans, root);
3110 btrfs_file_extent_calc_inline_size(0);
3111 memmove(buf+start, buf+start+skip,
3115 leaf = path->nodes[0];
3116 slot = path->slots[0];
3117 write_extent_buffer(leaf, buf,
3118 btrfs_item_ptr_offset(leaf, slot),
3120 inode_add_bytes(inode, datal);
3123 btrfs_mark_buffer_dirty(leaf);
3124 btrfs_release_path(path);
3126 inode_inc_iversion(inode);
3127 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3130 * we round up to the block size at eof when
3131 * determining which extents to clone above,
3132 * but shouldn't round up the file size
3134 endoff = new_key.offset + datal;
3135 if (endoff > destoff+olen)
3136 endoff = destoff+olen;
3137 if (endoff > inode->i_size)
3138 btrfs_i_size_write(inode, endoff);
3140 ret = btrfs_update_inode(trans, root, inode);
3142 btrfs_abort_transaction(trans, root, ret);
3143 btrfs_end_transaction(trans, root);
3146 ret = btrfs_end_transaction(trans, root);
3148 btrfs_release_path(path);
3154 btrfs_release_path(path);
3155 btrfs_free_path(path);
3160 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3161 u64 off, u64 olen, u64 destoff)
3163 struct inode *inode = file_inode(file);
3164 struct btrfs_root *root = BTRFS_I(inode)->root;
3169 u64 bs = root->fs_info->sb->s_blocksize;
3174 * - split compressed inline extents. annoying: we need to
3175 * decompress into destination's address_space (the file offset
3176 * may change, so source mapping won't do), then recompress (or
3177 * otherwise reinsert) a subrange.
3178 * - allow ranges within the same file to be cloned (provided
3179 * they don't overlap)?
3182 /* the destination must be opened for writing */
3183 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3186 if (btrfs_root_readonly(root))
3189 ret = mnt_want_write_file(file);
3193 src_file = fdget(srcfd);
3194 if (!src_file.file) {
3196 goto out_drop_write;
3200 if (src_file.file->f_path.mnt != file->f_path.mnt)
3203 src = file_inode(src_file.file);
3209 /* the src must be open for reading */
3210 if (!(src_file.file->f_mode & FMODE_READ))
3213 /* don't make the dst file partly checksummed */
3214 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3215 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3219 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3223 if (src->i_sb != inode->i_sb)
3228 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3229 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3231 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3232 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3235 mutex_lock(&src->i_mutex);
3238 /* determine range to clone */
3240 if (off + len > src->i_size || off + len < off)
3243 olen = len = src->i_size - off;
3244 /* if we extend to eof, continue to block boundary */
3245 if (off + len == src->i_size)
3246 len = ALIGN(src->i_size, bs) - off;
3248 /* verify the end result is block aligned */
3249 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3250 !IS_ALIGNED(destoff, bs))
3253 /* verify if ranges are overlapped within the same file */
3255 if (destoff + len > off && destoff < off + len)
3259 if (destoff > inode->i_size) {
3260 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3265 /* truncate page cache pages from target inode range */
3266 truncate_inode_pages_range(&inode->i_data, destoff,
3267 PAGE_CACHE_ALIGN(destoff + len) - 1);
3269 lock_extent_range(src, off, len);
3271 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3273 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3277 mutex_unlock(&src->i_mutex);
3278 mutex_unlock(&inode->i_mutex);
3280 mutex_unlock(&inode->i_mutex);
3281 mutex_unlock(&src->i_mutex);
3284 mutex_unlock(&src->i_mutex);
3289 mnt_drop_write_file(file);
3293 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3295 struct btrfs_ioctl_clone_range_args args;
3297 if (copy_from_user(&args, argp, sizeof(args)))
3299 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3300 args.src_length, args.dest_offset);
3304 * there are many ways the trans_start and trans_end ioctls can lead
3305 * to deadlocks. They should only be used by applications that
3306 * basically own the machine, and have a very in depth understanding
3307 * of all the possible deadlocks and enospc problems.
3309 static long btrfs_ioctl_trans_start(struct file *file)
3311 struct inode *inode = file_inode(file);
3312 struct btrfs_root *root = BTRFS_I(inode)->root;
3313 struct btrfs_trans_handle *trans;
3317 if (!capable(CAP_SYS_ADMIN))
3321 if (file->private_data)
3325 if (btrfs_root_readonly(root))
3328 ret = mnt_want_write_file(file);
3332 atomic_inc(&root->fs_info->open_ioctl_trans);
3335 trans = btrfs_start_ioctl_transaction(root);
3339 file->private_data = trans;
3343 atomic_dec(&root->fs_info->open_ioctl_trans);
3344 mnt_drop_write_file(file);
3349 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3351 struct inode *inode = file_inode(file);
3352 struct btrfs_root *root = BTRFS_I(inode)->root;
3353 struct btrfs_root *new_root;
3354 struct btrfs_dir_item *di;
3355 struct btrfs_trans_handle *trans;
3356 struct btrfs_path *path;
3357 struct btrfs_key location;
3358 struct btrfs_disk_key disk_key;
3363 if (!capable(CAP_SYS_ADMIN))
3366 ret = mnt_want_write_file(file);
3370 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3376 objectid = BTRFS_FS_TREE_OBJECTID;
3378 location.objectid = objectid;
3379 location.type = BTRFS_ROOT_ITEM_KEY;
3380 location.offset = (u64)-1;
3382 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3383 if (IS_ERR(new_root)) {
3384 ret = PTR_ERR(new_root);
3388 path = btrfs_alloc_path();
3393 path->leave_spinning = 1;
3395 trans = btrfs_start_transaction(root, 1);
3396 if (IS_ERR(trans)) {
3397 btrfs_free_path(path);
3398 ret = PTR_ERR(trans);
3402 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3403 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3404 dir_id, "default", 7, 1);
3405 if (IS_ERR_OR_NULL(di)) {
3406 btrfs_free_path(path);
3407 btrfs_end_transaction(trans, root);
3408 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3409 "item, this isn't going to work");
3414 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3415 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3416 btrfs_mark_buffer_dirty(path->nodes[0]);
3417 btrfs_free_path(path);
3419 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3420 btrfs_end_transaction(trans, root);
3422 mnt_drop_write_file(file);
3426 void btrfs_get_block_group_info(struct list_head *groups_list,
3427 struct btrfs_ioctl_space_info *space)
3429 struct btrfs_block_group_cache *block_group;
3431 space->total_bytes = 0;
3432 space->used_bytes = 0;
3434 list_for_each_entry(block_group, groups_list, list) {
3435 space->flags = block_group->flags;
3436 space->total_bytes += block_group->key.offset;
3437 space->used_bytes +=
3438 btrfs_block_group_used(&block_group->item);
3442 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3444 struct btrfs_ioctl_space_args space_args;
3445 struct btrfs_ioctl_space_info space;
3446 struct btrfs_ioctl_space_info *dest;
3447 struct btrfs_ioctl_space_info *dest_orig;
3448 struct btrfs_ioctl_space_info __user *user_dest;
3449 struct btrfs_space_info *info;
3450 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3451 BTRFS_BLOCK_GROUP_SYSTEM,
3452 BTRFS_BLOCK_GROUP_METADATA,
3453 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3460 if (copy_from_user(&space_args,
3461 (struct btrfs_ioctl_space_args __user *)arg,
3462 sizeof(space_args)))
3465 for (i = 0; i < num_types; i++) {
3466 struct btrfs_space_info *tmp;
3470 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3472 if (tmp->flags == types[i]) {
3482 down_read(&info->groups_sem);
3483 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3484 if (!list_empty(&info->block_groups[c]))
3487 up_read(&info->groups_sem);
3490 /* space_slots == 0 means they are asking for a count */
3491 if (space_args.space_slots == 0) {
3492 space_args.total_spaces = slot_count;
3496 slot_count = min_t(u64, space_args.space_slots, slot_count);
3498 alloc_size = sizeof(*dest) * slot_count;
3500 /* we generally have at most 6 or so space infos, one for each raid
3501 * level. So, a whole page should be more than enough for everyone
3503 if (alloc_size > PAGE_CACHE_SIZE)
3506 space_args.total_spaces = 0;
3507 dest = kmalloc(alloc_size, GFP_NOFS);
3512 /* now we have a buffer to copy into */
3513 for (i = 0; i < num_types; i++) {
3514 struct btrfs_space_info *tmp;
3521 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3523 if (tmp->flags == types[i]) {
3532 down_read(&info->groups_sem);
3533 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3534 if (!list_empty(&info->block_groups[c])) {
3535 btrfs_get_block_group_info(
3536 &info->block_groups[c], &space);
3537 memcpy(dest, &space, sizeof(space));
3539 space_args.total_spaces++;
3545 up_read(&info->groups_sem);
3548 user_dest = (struct btrfs_ioctl_space_info __user *)
3549 (arg + sizeof(struct btrfs_ioctl_space_args));
3551 if (copy_to_user(user_dest, dest_orig, alloc_size))
3556 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3562 static long btrfs_ioctl_global_rsv(struct btrfs_root *root, void __user *arg)
3564 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3567 spin_lock(&block_rsv->lock);
3568 reserved = block_rsv->reserved;
3569 spin_unlock(&block_rsv->lock);
3571 if (arg && copy_to_user(arg, &reserved, sizeof(reserved)))
3577 * there are many ways the trans_start and trans_end ioctls can lead
3578 * to deadlocks. They should only be used by applications that
3579 * basically own the machine, and have a very in depth understanding
3580 * of all the possible deadlocks and enospc problems.
3582 long btrfs_ioctl_trans_end(struct file *file)
3584 struct inode *inode = file_inode(file);
3585 struct btrfs_root *root = BTRFS_I(inode)->root;
3586 struct btrfs_trans_handle *trans;
3588 trans = file->private_data;
3591 file->private_data = NULL;
3593 btrfs_end_transaction(trans, root);
3595 atomic_dec(&root->fs_info->open_ioctl_trans);
3597 mnt_drop_write_file(file);
3601 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3604 struct btrfs_trans_handle *trans;
3608 trans = btrfs_attach_transaction_barrier(root);
3609 if (IS_ERR(trans)) {
3610 if (PTR_ERR(trans) != -ENOENT)
3611 return PTR_ERR(trans);
3613 /* No running transaction, don't bother */
3614 transid = root->fs_info->last_trans_committed;
3617 transid = trans->transid;
3618 ret = btrfs_commit_transaction_async(trans, root, 0);
3620 btrfs_end_transaction(trans, root);
3625 if (copy_to_user(argp, &transid, sizeof(transid)))
3630 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3636 if (copy_from_user(&transid, argp, sizeof(transid)))
3639 transid = 0; /* current trans */
3641 return btrfs_wait_for_commit(root, transid);
3644 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3646 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3647 struct btrfs_ioctl_scrub_args *sa;
3650 if (!capable(CAP_SYS_ADMIN))
3653 sa = memdup_user(arg, sizeof(*sa));
3657 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3658 ret = mnt_want_write_file(file);
3663 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3664 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3667 if (copy_to_user(arg, sa, sizeof(*sa)))
3670 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3671 mnt_drop_write_file(file);
3677 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3679 if (!capable(CAP_SYS_ADMIN))
3682 return btrfs_scrub_cancel(root->fs_info);
3685 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3688 struct btrfs_ioctl_scrub_args *sa;
3691 if (!capable(CAP_SYS_ADMIN))
3694 sa = memdup_user(arg, sizeof(*sa));
3698 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3700 if (copy_to_user(arg, sa, sizeof(*sa)))
3707 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3710 struct btrfs_ioctl_get_dev_stats *sa;
3713 sa = memdup_user(arg, sizeof(*sa));
3717 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3722 ret = btrfs_get_dev_stats(root, sa);
3724 if (copy_to_user(arg, sa, sizeof(*sa)))
3731 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3733 struct btrfs_ioctl_dev_replace_args *p;
3736 if (!capable(CAP_SYS_ADMIN))
3739 p = memdup_user(arg, sizeof(*p));
3744 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3745 if (root->fs_info->sb->s_flags & MS_RDONLY) {
3750 &root->fs_info->mutually_exclusive_operation_running,
3752 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3754 ret = btrfs_dev_replace_start(root, p);
3756 &root->fs_info->mutually_exclusive_operation_running,
3760 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3761 btrfs_dev_replace_status(root->fs_info, p);
3764 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3765 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3772 if (copy_to_user(arg, p, sizeof(*p)))
3779 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3785 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3786 struct inode_fs_paths *ipath = NULL;
3787 struct btrfs_path *path;
3789 if (!capable(CAP_DAC_READ_SEARCH))
3792 path = btrfs_alloc_path();
3798 ipa = memdup_user(arg, sizeof(*ipa));
3805 size = min_t(u32, ipa->size, 4096);
3806 ipath = init_ipath(size, root, path);
3807 if (IS_ERR(ipath)) {
3808 ret = PTR_ERR(ipath);
3813 ret = paths_from_inode(ipa->inum, ipath);
3817 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3818 rel_ptr = ipath->fspath->val[i] -
3819 (u64)(unsigned long)ipath->fspath->val;
3820 ipath->fspath->val[i] = rel_ptr;
3823 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3824 (void *)(unsigned long)ipath->fspath, size);
3831 btrfs_free_path(path);
3838 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3840 struct btrfs_data_container *inodes = ctx;
3841 const size_t c = 3 * sizeof(u64);
3843 if (inodes->bytes_left >= c) {
3844 inodes->bytes_left -= c;
3845 inodes->val[inodes->elem_cnt] = inum;
3846 inodes->val[inodes->elem_cnt + 1] = offset;
3847 inodes->val[inodes->elem_cnt + 2] = root;
3848 inodes->elem_cnt += 3;
3850 inodes->bytes_missing += c - inodes->bytes_left;
3851 inodes->bytes_left = 0;
3852 inodes->elem_missed += 3;
3858 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3863 struct btrfs_ioctl_logical_ino_args *loi;
3864 struct btrfs_data_container *inodes = NULL;
3865 struct btrfs_path *path = NULL;
3867 if (!capable(CAP_SYS_ADMIN))
3870 loi = memdup_user(arg, sizeof(*loi));
3877 path = btrfs_alloc_path();
3883 size = min_t(u32, loi->size, 64 * 1024);
3884 inodes = init_data_container(size);
3885 if (IS_ERR(inodes)) {
3886 ret = PTR_ERR(inodes);
3891 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3892 build_ino_list, inodes);
3898 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3899 (void *)(unsigned long)inodes, size);
3904 btrfs_free_path(path);
3911 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3912 struct btrfs_ioctl_balance_args *bargs)
3914 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3916 bargs->flags = bctl->flags;
3918 if (atomic_read(&fs_info->balance_running))
3919 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3920 if (atomic_read(&fs_info->balance_pause_req))
3921 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3922 if (atomic_read(&fs_info->balance_cancel_req))
3923 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3925 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3926 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3927 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3930 spin_lock(&fs_info->balance_lock);
3931 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3932 spin_unlock(&fs_info->balance_lock);
3934 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3938 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3940 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3941 struct btrfs_fs_info *fs_info = root->fs_info;
3942 struct btrfs_ioctl_balance_args *bargs;
3943 struct btrfs_balance_control *bctl;
3944 bool need_unlock; /* for mut. excl. ops lock */
3947 if (!capable(CAP_SYS_ADMIN))
3950 ret = mnt_want_write_file(file);
3955 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3956 mutex_lock(&fs_info->volume_mutex);
3957 mutex_lock(&fs_info->balance_mutex);
3963 * mut. excl. ops lock is locked. Three possibilites:
3964 * (1) some other op is running
3965 * (2) balance is running
3966 * (3) balance is paused -- special case (think resume)
3968 mutex_lock(&fs_info->balance_mutex);
3969 if (fs_info->balance_ctl) {
3970 /* this is either (2) or (3) */
3971 if (!atomic_read(&fs_info->balance_running)) {
3972 mutex_unlock(&fs_info->balance_mutex);
3973 if (!mutex_trylock(&fs_info->volume_mutex))
3975 mutex_lock(&fs_info->balance_mutex);
3977 if (fs_info->balance_ctl &&
3978 !atomic_read(&fs_info->balance_running)) {
3980 need_unlock = false;
3984 mutex_unlock(&fs_info->balance_mutex);
3985 mutex_unlock(&fs_info->volume_mutex);
3989 mutex_unlock(&fs_info->balance_mutex);
3995 mutex_unlock(&fs_info->balance_mutex);
3996 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4001 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4004 bargs = memdup_user(arg, sizeof(*bargs));
4005 if (IS_ERR(bargs)) {
4006 ret = PTR_ERR(bargs);
4010 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4011 if (!fs_info->balance_ctl) {
4016 bctl = fs_info->balance_ctl;
4017 spin_lock(&fs_info->balance_lock);
4018 bctl->flags |= BTRFS_BALANCE_RESUME;
4019 spin_unlock(&fs_info->balance_lock);
4027 if (fs_info->balance_ctl) {
4032 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4038 bctl->fs_info = fs_info;
4040 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4041 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4042 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4044 bctl->flags = bargs->flags;
4046 /* balance everything - no filters */
4047 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4052 * Ownership of bctl and mutually_exclusive_operation_running
4053 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4054 * or, if restriper was paused all the way until unmount, in
4055 * free_fs_info. mutually_exclusive_operation_running is
4056 * cleared in __cancel_balance.
4058 need_unlock = false;
4060 ret = btrfs_balance(bctl, bargs);
4063 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4070 mutex_unlock(&fs_info->balance_mutex);
4071 mutex_unlock(&fs_info->volume_mutex);
4073 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4075 mnt_drop_write_file(file);
4079 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4081 if (!capable(CAP_SYS_ADMIN))
4085 case BTRFS_BALANCE_CTL_PAUSE:
4086 return btrfs_pause_balance(root->fs_info);
4087 case BTRFS_BALANCE_CTL_CANCEL:
4088 return btrfs_cancel_balance(root->fs_info);
4094 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4097 struct btrfs_fs_info *fs_info = root->fs_info;
4098 struct btrfs_ioctl_balance_args *bargs;
4101 if (!capable(CAP_SYS_ADMIN))
4104 mutex_lock(&fs_info->balance_mutex);
4105 if (!fs_info->balance_ctl) {
4110 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4116 update_ioctl_balance_args(fs_info, 1, bargs);
4118 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4123 mutex_unlock(&fs_info->balance_mutex);
4127 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4129 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4130 struct btrfs_ioctl_quota_ctl_args *sa;
4131 struct btrfs_trans_handle *trans = NULL;
4135 if (!capable(CAP_SYS_ADMIN))
4138 ret = mnt_want_write_file(file);
4142 sa = memdup_user(arg, sizeof(*sa));
4148 down_write(&root->fs_info->subvol_sem);
4149 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4150 if (IS_ERR(trans)) {
4151 ret = PTR_ERR(trans);
4156 case BTRFS_QUOTA_CTL_ENABLE:
4157 ret = btrfs_quota_enable(trans, root->fs_info);
4159 case BTRFS_QUOTA_CTL_DISABLE:
4160 ret = btrfs_quota_disable(trans, root->fs_info);
4167 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4172 up_write(&root->fs_info->subvol_sem);
4174 mnt_drop_write_file(file);
4178 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4180 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4181 struct btrfs_ioctl_qgroup_assign_args *sa;
4182 struct btrfs_trans_handle *trans;
4186 if (!capable(CAP_SYS_ADMIN))
4189 ret = mnt_want_write_file(file);
4193 sa = memdup_user(arg, sizeof(*sa));
4199 trans = btrfs_join_transaction(root);
4200 if (IS_ERR(trans)) {
4201 ret = PTR_ERR(trans);
4205 /* FIXME: check if the IDs really exist */
4207 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4210 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4214 err = btrfs_end_transaction(trans, root);
4221 mnt_drop_write_file(file);
4225 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4227 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4228 struct btrfs_ioctl_qgroup_create_args *sa;
4229 struct btrfs_trans_handle *trans;
4233 if (!capable(CAP_SYS_ADMIN))
4236 ret = mnt_want_write_file(file);
4240 sa = memdup_user(arg, sizeof(*sa));
4246 if (!sa->qgroupid) {
4251 trans = btrfs_join_transaction(root);
4252 if (IS_ERR(trans)) {
4253 ret = PTR_ERR(trans);
4257 /* FIXME: check if the IDs really exist */
4259 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4262 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4265 err = btrfs_end_transaction(trans, root);
4272 mnt_drop_write_file(file);
4276 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4278 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4279 struct btrfs_ioctl_qgroup_limit_args *sa;
4280 struct btrfs_trans_handle *trans;
4285 if (!capable(CAP_SYS_ADMIN))
4288 ret = mnt_want_write_file(file);
4292 sa = memdup_user(arg, sizeof(*sa));
4298 trans = btrfs_join_transaction(root);
4299 if (IS_ERR(trans)) {
4300 ret = PTR_ERR(trans);
4304 qgroupid = sa->qgroupid;
4306 /* take the current subvol as qgroup */
4307 qgroupid = root->root_key.objectid;
4310 /* FIXME: check if the IDs really exist */
4311 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4313 err = btrfs_end_transaction(trans, root);
4320 mnt_drop_write_file(file);
4324 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4326 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4327 struct btrfs_ioctl_quota_rescan_args *qsa;
4330 if (!capable(CAP_SYS_ADMIN))
4333 ret = mnt_want_write_file(file);
4337 qsa = memdup_user(arg, sizeof(*qsa));
4348 ret = btrfs_qgroup_rescan(root->fs_info);
4353 mnt_drop_write_file(file);
4357 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4359 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4360 struct btrfs_ioctl_quota_rescan_args *qsa;
4363 if (!capable(CAP_SYS_ADMIN))
4366 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4370 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4372 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4375 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4382 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4384 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4386 if (!capable(CAP_SYS_ADMIN))
4389 return btrfs_qgroup_wait_for_completion(root->fs_info);
4392 static long btrfs_ioctl_set_received_subvol(struct file *file,
4395 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4396 struct inode *inode = file_inode(file);
4397 struct btrfs_root *root = BTRFS_I(inode)->root;
4398 struct btrfs_root_item *root_item = &root->root_item;
4399 struct btrfs_trans_handle *trans;
4400 struct timespec ct = CURRENT_TIME;
4402 int received_uuid_changed;
4404 if (!inode_owner_or_capable(inode))
4407 ret = mnt_want_write_file(file);
4411 down_write(&root->fs_info->subvol_sem);
4413 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4418 if (btrfs_root_readonly(root)) {
4423 sa = memdup_user(arg, sizeof(*sa));
4432 * 2 - uuid items (received uuid + subvol uuid)
4434 trans = btrfs_start_transaction(root, 3);
4435 if (IS_ERR(trans)) {
4436 ret = PTR_ERR(trans);
4441 sa->rtransid = trans->transid;
4442 sa->rtime.sec = ct.tv_sec;
4443 sa->rtime.nsec = ct.tv_nsec;
4445 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4447 if (received_uuid_changed &&
4448 !btrfs_is_empty_uuid(root_item->received_uuid))
4449 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4450 root_item->received_uuid,
4451 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4452 root->root_key.objectid);
4453 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4454 btrfs_set_root_stransid(root_item, sa->stransid);
4455 btrfs_set_root_rtransid(root_item, sa->rtransid);
4456 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4457 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4458 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4459 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4461 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4462 &root->root_key, &root->root_item);
4464 btrfs_end_transaction(trans, root);
4467 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4468 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4470 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4471 root->root_key.objectid);
4472 if (ret < 0 && ret != -EEXIST) {
4473 btrfs_abort_transaction(trans, root, ret);
4477 ret = btrfs_commit_transaction(trans, root);
4479 btrfs_abort_transaction(trans, root, ret);
4483 ret = copy_to_user(arg, sa, sizeof(*sa));
4489 up_write(&root->fs_info->subvol_sem);
4490 mnt_drop_write_file(file);
4494 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4496 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4499 char label[BTRFS_LABEL_SIZE];
4501 spin_lock(&root->fs_info->super_lock);
4502 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4503 spin_unlock(&root->fs_info->super_lock);
4505 len = strnlen(label, BTRFS_LABEL_SIZE);
4507 if (len == BTRFS_LABEL_SIZE) {
4508 btrfs_warn(root->fs_info,
4509 "label is too long, return the first %zu bytes", --len);
4512 ret = copy_to_user(arg, label, len);
4514 return ret ? -EFAULT : 0;
4517 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4519 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4520 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4521 struct btrfs_trans_handle *trans;
4522 char label[BTRFS_LABEL_SIZE];
4525 if (!capable(CAP_SYS_ADMIN))
4528 if (copy_from_user(label, arg, sizeof(label)))
4531 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4532 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
4533 BTRFS_LABEL_SIZE - 1);
4537 ret = mnt_want_write_file(file);
4541 trans = btrfs_start_transaction(root, 0);
4542 if (IS_ERR(trans)) {
4543 ret = PTR_ERR(trans);
4547 spin_lock(&root->fs_info->super_lock);
4548 strcpy(super_block->label, label);
4549 spin_unlock(&root->fs_info->super_lock);
4550 ret = btrfs_end_transaction(trans, root);
4553 mnt_drop_write_file(file);
4557 #define INIT_FEATURE_FLAGS(suffix) \
4558 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4559 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4560 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4562 static int btrfs_ioctl_get_supported_features(struct file *file,
4565 static struct btrfs_ioctl_feature_flags features[3] = {
4566 INIT_FEATURE_FLAGS(SUPP),
4567 INIT_FEATURE_FLAGS(SAFE_SET),
4568 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4571 if (copy_to_user(arg, &features, sizeof(features)))
4577 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
4579 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4580 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4581 struct btrfs_ioctl_feature_flags features;
4583 features.compat_flags = btrfs_super_compat_flags(super_block);
4584 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4585 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4587 if (copy_to_user(arg, &features, sizeof(features)))
4593 static int check_feature_bits(struct btrfs_root *root,
4594 enum btrfs_feature_set set,
4595 u64 change_mask, u64 flags, u64 supported_flags,
4596 u64 safe_set, u64 safe_clear)
4598 const char *type = btrfs_feature_set_names[set];
4600 u64 disallowed, unsupported;
4601 u64 set_mask = flags & change_mask;
4602 u64 clear_mask = ~flags & change_mask;
4604 unsupported = set_mask & ~supported_flags;
4606 names = btrfs_printable_features(set, unsupported);
4608 btrfs_warn(root->fs_info,
4609 "this kernel does not support the %s feature bit%s",
4610 names, strchr(names, ',') ? "s" : "");
4613 btrfs_warn(root->fs_info,
4614 "this kernel does not support %s bits 0x%llx",
4619 disallowed = set_mask & ~safe_set;
4621 names = btrfs_printable_features(set, disallowed);
4623 btrfs_warn(root->fs_info,
4624 "can't set the %s feature bit%s while mounted",
4625 names, strchr(names, ',') ? "s" : "");
4628 btrfs_warn(root->fs_info,
4629 "can't set %s bits 0x%llx while mounted",
4634 disallowed = clear_mask & ~safe_clear;
4636 names = btrfs_printable_features(set, disallowed);
4638 btrfs_warn(root->fs_info,
4639 "can't clear the %s feature bit%s while mounted",
4640 names, strchr(names, ',') ? "s" : "");
4643 btrfs_warn(root->fs_info,
4644 "can't clear %s bits 0x%llx while mounted",
4652 #define check_feature(root, change_mask, flags, mask_base) \
4653 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
4654 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4655 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4656 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4658 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4660 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4661 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4662 struct btrfs_ioctl_feature_flags flags[2];
4663 struct btrfs_trans_handle *trans;
4667 if (!capable(CAP_SYS_ADMIN))
4670 if (copy_from_user(flags, arg, sizeof(flags)))
4674 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4675 !flags[0].incompat_flags)
4678 ret = check_feature(root, flags[0].compat_flags,
4679 flags[1].compat_flags, COMPAT);
4683 ret = check_feature(root, flags[0].compat_ro_flags,
4684 flags[1].compat_ro_flags, COMPAT_RO);
4688 ret = check_feature(root, flags[0].incompat_flags,
4689 flags[1].incompat_flags, INCOMPAT);
4693 trans = btrfs_start_transaction(root, 1);
4695 return PTR_ERR(trans);
4697 spin_lock(&root->fs_info->super_lock);
4698 newflags = btrfs_super_compat_flags(super_block);
4699 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4700 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4701 btrfs_set_super_compat_flags(super_block, newflags);
4703 newflags = btrfs_super_compat_ro_flags(super_block);
4704 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4705 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4706 btrfs_set_super_compat_ro_flags(super_block, newflags);
4708 newflags = btrfs_super_incompat_flags(super_block);
4709 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4710 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4711 btrfs_set_super_incompat_flags(super_block, newflags);
4712 spin_unlock(&root->fs_info->super_lock);
4714 return btrfs_end_transaction(trans, root);
4717 long btrfs_ioctl(struct file *file, unsigned int
4718 cmd, unsigned long arg)
4720 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4721 void __user *argp = (void __user *)arg;
4724 case FS_IOC_GETFLAGS:
4725 return btrfs_ioctl_getflags(file, argp);
4726 case FS_IOC_SETFLAGS:
4727 return btrfs_ioctl_setflags(file, argp);
4728 case FS_IOC_GETVERSION:
4729 return btrfs_ioctl_getversion(file, argp);
4731 return btrfs_ioctl_fitrim(file, argp);
4732 case BTRFS_IOC_SNAP_CREATE:
4733 return btrfs_ioctl_snap_create(file, argp, 0);
4734 case BTRFS_IOC_SNAP_CREATE_V2:
4735 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4736 case BTRFS_IOC_SUBVOL_CREATE:
4737 return btrfs_ioctl_snap_create(file, argp, 1);
4738 case BTRFS_IOC_SUBVOL_CREATE_V2:
4739 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4740 case BTRFS_IOC_SNAP_DESTROY:
4741 return btrfs_ioctl_snap_destroy(file, argp);
4742 case BTRFS_IOC_SUBVOL_GETFLAGS:
4743 return btrfs_ioctl_subvol_getflags(file, argp);
4744 case BTRFS_IOC_SUBVOL_SETFLAGS:
4745 return btrfs_ioctl_subvol_setflags(file, argp);
4746 case BTRFS_IOC_DEFAULT_SUBVOL:
4747 return btrfs_ioctl_default_subvol(file, argp);
4748 case BTRFS_IOC_DEFRAG:
4749 return btrfs_ioctl_defrag(file, NULL);
4750 case BTRFS_IOC_DEFRAG_RANGE:
4751 return btrfs_ioctl_defrag(file, argp);
4752 case BTRFS_IOC_RESIZE:
4753 return btrfs_ioctl_resize(file, argp);
4754 case BTRFS_IOC_ADD_DEV:
4755 return btrfs_ioctl_add_dev(root, argp);
4756 case BTRFS_IOC_RM_DEV:
4757 return btrfs_ioctl_rm_dev(file, argp);
4758 case BTRFS_IOC_FS_INFO:
4759 return btrfs_ioctl_fs_info(root, argp);
4760 case BTRFS_IOC_DEV_INFO:
4761 return btrfs_ioctl_dev_info(root, argp);
4762 case BTRFS_IOC_BALANCE:
4763 return btrfs_ioctl_balance(file, NULL);
4764 case BTRFS_IOC_CLONE:
4765 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4766 case BTRFS_IOC_CLONE_RANGE:
4767 return btrfs_ioctl_clone_range(file, argp);
4768 case BTRFS_IOC_TRANS_START:
4769 return btrfs_ioctl_trans_start(file);
4770 case BTRFS_IOC_TRANS_END:
4771 return btrfs_ioctl_trans_end(file);
4772 case BTRFS_IOC_TREE_SEARCH:
4773 return btrfs_ioctl_tree_search(file, argp);
4774 case BTRFS_IOC_INO_LOOKUP:
4775 return btrfs_ioctl_ino_lookup(file, argp);
4776 case BTRFS_IOC_INO_PATHS:
4777 return btrfs_ioctl_ino_to_path(root, argp);
4778 case BTRFS_IOC_LOGICAL_INO:
4779 return btrfs_ioctl_logical_to_ino(root, argp);
4780 case BTRFS_IOC_SPACE_INFO:
4781 return btrfs_ioctl_space_info(root, argp);
4782 case BTRFS_IOC_GLOBAL_RSV:
4783 return btrfs_ioctl_global_rsv(root, argp);
4784 case BTRFS_IOC_SYNC: {
4787 ret = btrfs_start_delalloc_roots(root->fs_info, 0);
4790 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
4793 case BTRFS_IOC_START_SYNC:
4794 return btrfs_ioctl_start_sync(root, argp);
4795 case BTRFS_IOC_WAIT_SYNC:
4796 return btrfs_ioctl_wait_sync(root, argp);
4797 case BTRFS_IOC_SCRUB:
4798 return btrfs_ioctl_scrub(file, argp);
4799 case BTRFS_IOC_SCRUB_CANCEL:
4800 return btrfs_ioctl_scrub_cancel(root, argp);
4801 case BTRFS_IOC_SCRUB_PROGRESS:
4802 return btrfs_ioctl_scrub_progress(root, argp);
4803 case BTRFS_IOC_BALANCE_V2:
4804 return btrfs_ioctl_balance(file, argp);
4805 case BTRFS_IOC_BALANCE_CTL:
4806 return btrfs_ioctl_balance_ctl(root, arg);
4807 case BTRFS_IOC_BALANCE_PROGRESS:
4808 return btrfs_ioctl_balance_progress(root, argp);
4809 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4810 return btrfs_ioctl_set_received_subvol(file, argp);
4811 case BTRFS_IOC_SEND:
4812 return btrfs_ioctl_send(file, argp);
4813 case BTRFS_IOC_GET_DEV_STATS:
4814 return btrfs_ioctl_get_dev_stats(root, argp);
4815 case BTRFS_IOC_QUOTA_CTL:
4816 return btrfs_ioctl_quota_ctl(file, argp);
4817 case BTRFS_IOC_QGROUP_ASSIGN:
4818 return btrfs_ioctl_qgroup_assign(file, argp);
4819 case BTRFS_IOC_QGROUP_CREATE:
4820 return btrfs_ioctl_qgroup_create(file, argp);
4821 case BTRFS_IOC_QGROUP_LIMIT:
4822 return btrfs_ioctl_qgroup_limit(file, argp);
4823 case BTRFS_IOC_QUOTA_RESCAN:
4824 return btrfs_ioctl_quota_rescan(file, argp);
4825 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4826 return btrfs_ioctl_quota_rescan_status(file, argp);
4827 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4828 return btrfs_ioctl_quota_rescan_wait(file, argp);
4829 case BTRFS_IOC_DEV_REPLACE:
4830 return btrfs_ioctl_dev_replace(root, argp);
4831 case BTRFS_IOC_GET_FSLABEL:
4832 return btrfs_ioctl_get_fslabel(file, argp);
4833 case BTRFS_IOC_SET_FSLABEL:
4834 return btrfs_ioctl_set_fslabel(file, argp);
4835 case BTRFS_IOC_FILE_EXTENT_SAME:
4836 return btrfs_ioctl_file_extent_same(file, argp);
4837 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4838 return btrfs_ioctl_get_supported_features(file, argp);
4839 case BTRFS_IOC_GET_FEATURES:
4840 return btrfs_ioctl_get_features(file, argp);
4841 case BTRFS_IOC_SET_FEATURES:
4842 return btrfs_ioctl_set_features(file, argp);