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>
47 #include "transaction.h"
48 #include "btrfs_inode.h"
50 #include "print-tree.h"
53 #include "inode-map.h"
55 /* Mask out flags that are inappropriate for the given type of inode. */
56 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
60 else if (S_ISREG(mode
))
61 return flags
& ~FS_DIRSYNC_FL
;
63 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
67 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
69 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
71 unsigned int iflags
= 0;
73 if (flags
& BTRFS_INODE_SYNC
)
75 if (flags
& BTRFS_INODE_IMMUTABLE
)
76 iflags
|= FS_IMMUTABLE_FL
;
77 if (flags
& BTRFS_INODE_APPEND
)
78 iflags
|= FS_APPEND_FL
;
79 if (flags
& BTRFS_INODE_NODUMP
)
80 iflags
|= FS_NODUMP_FL
;
81 if (flags
& BTRFS_INODE_NOATIME
)
82 iflags
|= FS_NOATIME_FL
;
83 if (flags
& BTRFS_INODE_DIRSYNC
)
84 iflags
|= FS_DIRSYNC_FL
;
85 if (flags
& BTRFS_INODE_NODATACOW
)
86 iflags
|= FS_NOCOW_FL
;
88 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
89 iflags
|= FS_COMPR_FL
;
90 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
91 iflags
|= FS_NOCOMP_FL
;
97 * Update inode->i_flags based on the btrfs internal flags.
99 void btrfs_update_iflags(struct inode
*inode
)
101 struct btrfs_inode
*ip
= BTRFS_I(inode
);
103 inode
->i_flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
105 if (ip
->flags
& BTRFS_INODE_SYNC
)
106 inode
->i_flags
|= S_SYNC
;
107 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
108 inode
->i_flags
|= S_IMMUTABLE
;
109 if (ip
->flags
& BTRFS_INODE_APPEND
)
110 inode
->i_flags
|= S_APPEND
;
111 if (ip
->flags
& BTRFS_INODE_NOATIME
)
112 inode
->i_flags
|= S_NOATIME
;
113 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
114 inode
->i_flags
|= S_DIRSYNC
;
118 * Inherit flags from the parent inode.
120 * Currently only the compression flags and the cow flags are inherited.
122 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
129 flags
= BTRFS_I(dir
)->flags
;
131 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
132 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
133 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
134 } else if (flags
& BTRFS_INODE_COMPRESS
) {
135 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
136 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
139 if (flags
& BTRFS_INODE_NODATACOW
)
140 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
142 btrfs_update_iflags(inode
);
145 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
147 struct btrfs_inode
*ip
= BTRFS_I(file
->f_path
.dentry
->d_inode
);
148 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
150 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
155 static int check_flags(unsigned int flags
)
157 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
158 FS_NOATIME_FL
| FS_NODUMP_FL
| \
159 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
160 FS_NOCOMP_FL
| FS_COMPR_FL
|
164 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
170 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
172 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
173 struct btrfs_inode
*ip
= BTRFS_I(inode
);
174 struct btrfs_root
*root
= ip
->root
;
175 struct btrfs_trans_handle
*trans
;
176 unsigned int flags
, oldflags
;
179 if (btrfs_root_readonly(root
))
182 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
185 ret
= check_flags(flags
);
189 if (!inode_owner_or_capable(inode
))
192 mutex_lock(&inode
->i_mutex
);
194 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
195 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
196 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
197 if (!capable(CAP_LINUX_IMMUTABLE
)) {
203 ret
= mnt_want_write(file
->f_path
.mnt
);
207 if (flags
& FS_SYNC_FL
)
208 ip
->flags
|= BTRFS_INODE_SYNC
;
210 ip
->flags
&= ~BTRFS_INODE_SYNC
;
211 if (flags
& FS_IMMUTABLE_FL
)
212 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
214 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
215 if (flags
& FS_APPEND_FL
)
216 ip
->flags
|= BTRFS_INODE_APPEND
;
218 ip
->flags
&= ~BTRFS_INODE_APPEND
;
219 if (flags
& FS_NODUMP_FL
)
220 ip
->flags
|= BTRFS_INODE_NODUMP
;
222 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
223 if (flags
& FS_NOATIME_FL
)
224 ip
->flags
|= BTRFS_INODE_NOATIME
;
226 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
227 if (flags
& FS_DIRSYNC_FL
)
228 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
230 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
231 if (flags
& FS_NOCOW_FL
)
232 ip
->flags
|= BTRFS_INODE_NODATACOW
;
234 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
237 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
238 * flag may be changed automatically if compression code won't make
241 if (flags
& FS_NOCOMP_FL
) {
242 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
243 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
244 } else if (flags
& FS_COMPR_FL
) {
245 ip
->flags
|= BTRFS_INODE_COMPRESS
;
246 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
248 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
251 trans
= btrfs_join_transaction(root
);
252 BUG_ON(IS_ERR(trans
));
254 ret
= btrfs_update_inode(trans
, root
, inode
);
257 btrfs_update_iflags(inode
);
258 inode
->i_ctime
= CURRENT_TIME
;
259 btrfs_end_transaction(trans
, root
);
261 mnt_drop_write(file
->f_path
.mnt
);
265 mutex_unlock(&inode
->i_mutex
);
269 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
271 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
273 return put_user(inode
->i_generation
, arg
);
276 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
278 struct btrfs_root
*root
= fdentry(file
)->d_sb
->s_fs_info
;
279 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
280 struct btrfs_device
*device
;
281 struct request_queue
*q
;
282 struct fstrim_range range
;
283 u64 minlen
= ULLONG_MAX
;
285 u64 total_bytes
= btrfs_super_total_bytes(root
->fs_info
->super_copy
);
288 if (!capable(CAP_SYS_ADMIN
))
292 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
296 q
= bdev_get_queue(device
->bdev
);
297 if (blk_queue_discard(q
)) {
299 minlen
= min((u64
)q
->limits
.discard_granularity
,
307 if (copy_from_user(&range
, arg
, sizeof(range
)))
309 if (range
.start
> total_bytes
)
312 range
.len
= min(range
.len
, total_bytes
- range
.start
);
313 range
.minlen
= max(range
.minlen
, minlen
);
314 ret
= btrfs_trim_fs(root
, &range
);
318 if (copy_to_user(arg
, &range
, sizeof(range
)))
324 static noinline
int create_subvol(struct btrfs_root
*root
,
325 struct dentry
*dentry
,
326 char *name
, int namelen
,
329 struct btrfs_trans_handle
*trans
;
330 struct btrfs_key key
;
331 struct btrfs_root_item root_item
;
332 struct btrfs_inode_item
*inode_item
;
333 struct extent_buffer
*leaf
;
334 struct btrfs_root
*new_root
;
335 struct dentry
*parent
= dentry
->d_parent
;
340 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
343 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
347 dir
= parent
->d_inode
;
355 trans
= btrfs_start_transaction(root
, 6);
357 return PTR_ERR(trans
);
359 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
360 0, objectid
, NULL
, 0, 0, 0);
366 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
367 btrfs_set_header_bytenr(leaf
, leaf
->start
);
368 btrfs_set_header_generation(leaf
, trans
->transid
);
369 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
370 btrfs_set_header_owner(leaf
, objectid
);
372 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
373 (unsigned long)btrfs_header_fsid(leaf
),
375 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
376 (unsigned long)btrfs_header_chunk_tree_uuid(leaf
),
378 btrfs_mark_buffer_dirty(leaf
);
380 inode_item
= &root_item
.inode
;
381 memset(inode_item
, 0, sizeof(*inode_item
));
382 inode_item
->generation
= cpu_to_le64(1);
383 inode_item
->size
= cpu_to_le64(3);
384 inode_item
->nlink
= cpu_to_le32(1);
385 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
386 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
389 root_item
.byte_limit
= 0;
390 inode_item
->flags
= cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT
);
392 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
393 btrfs_set_root_generation(&root_item
, trans
->transid
);
394 btrfs_set_root_level(&root_item
, 0);
395 btrfs_set_root_refs(&root_item
, 1);
396 btrfs_set_root_used(&root_item
, leaf
->len
);
397 btrfs_set_root_last_snapshot(&root_item
, 0);
399 memset(&root_item
.drop_progress
, 0, sizeof(root_item
.drop_progress
));
400 root_item
.drop_level
= 0;
402 btrfs_tree_unlock(leaf
);
403 free_extent_buffer(leaf
);
406 btrfs_set_root_dirid(&root_item
, new_dirid
);
408 key
.objectid
= objectid
;
410 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
411 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
416 key
.offset
= (u64
)-1;
417 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
418 BUG_ON(IS_ERR(new_root
));
420 btrfs_record_root_in_trans(trans
, new_root
);
422 ret
= btrfs_create_subvol_root(trans
, new_root
, new_dirid
);
424 * insert the directory item
426 ret
= btrfs_set_inode_index(dir
, &index
);
429 ret
= btrfs_insert_dir_item(trans
, root
,
430 name
, namelen
, dir
, &key
,
431 BTRFS_FT_DIR
, index
);
435 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
436 ret
= btrfs_update_inode(trans
, root
, dir
);
439 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
440 objectid
, root
->root_key
.objectid
,
441 btrfs_ino(dir
), index
, name
, namelen
);
445 d_instantiate(dentry
, btrfs_lookup_dentry(dir
, dentry
));
448 *async_transid
= trans
->transid
;
449 err
= btrfs_commit_transaction_async(trans
, root
, 1);
451 err
= btrfs_commit_transaction(trans
, root
);
458 static int create_snapshot(struct btrfs_root
*root
, struct dentry
*dentry
,
459 char *name
, int namelen
, u64
*async_transid
,
463 struct btrfs_pending_snapshot
*pending_snapshot
;
464 struct btrfs_trans_handle
*trans
;
470 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
471 if (!pending_snapshot
)
474 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
);
475 pending_snapshot
->dentry
= dentry
;
476 pending_snapshot
->root
= root
;
477 pending_snapshot
->readonly
= readonly
;
479 trans
= btrfs_start_transaction(root
->fs_info
->extent_root
, 5);
481 ret
= PTR_ERR(trans
);
485 ret
= btrfs_snap_reserve_metadata(trans
, pending_snapshot
);
488 spin_lock(&root
->fs_info
->trans_lock
);
489 list_add(&pending_snapshot
->list
,
490 &trans
->transaction
->pending_snapshots
);
491 spin_unlock(&root
->fs_info
->trans_lock
);
493 *async_transid
= trans
->transid
;
494 ret
= btrfs_commit_transaction_async(trans
,
495 root
->fs_info
->extent_root
, 1);
497 ret
= btrfs_commit_transaction(trans
,
498 root
->fs_info
->extent_root
);
502 ret
= pending_snapshot
->error
;
506 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
510 inode
= btrfs_lookup_dentry(dentry
->d_parent
->d_inode
, dentry
);
512 ret
= PTR_ERR(inode
);
516 d_instantiate(dentry
, inode
);
519 kfree(pending_snapshot
);
523 /* copy of check_sticky in fs/namei.c()
524 * It's inline, so penalty for filesystems that don't use sticky bit is
527 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
529 uid_t fsuid
= current_fsuid();
531 if (!(dir
->i_mode
& S_ISVTX
))
533 if (inode
->i_uid
== fsuid
)
535 if (dir
->i_uid
== fsuid
)
537 return !capable(CAP_FOWNER
);
540 /* copy of may_delete in fs/namei.c()
541 * Check whether we can remove a link victim from directory dir, check
542 * whether the type of victim is right.
543 * 1. We can't do it if dir is read-only (done in permission())
544 * 2. We should have write and exec permissions on dir
545 * 3. We can't remove anything from append-only dir
546 * 4. We can't do anything with immutable dir (done in permission())
547 * 5. If the sticky bit on dir is set we should either
548 * a. be owner of dir, or
549 * b. be owner of victim, or
550 * c. have CAP_FOWNER capability
551 * 6. If the victim is append-only or immutable we can't do antyhing with
552 * links pointing to it.
553 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
554 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
555 * 9. We can't remove a root or mountpoint.
556 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
557 * nfs_async_unlink().
560 static int btrfs_may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
564 if (!victim
->d_inode
)
567 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
568 audit_inode_child(victim
, dir
);
570 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
575 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
576 IS_APPEND(victim
->d_inode
)||
577 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
580 if (!S_ISDIR(victim
->d_inode
->i_mode
))
584 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
588 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
593 /* copy of may_create in fs/namei.c() */
594 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
600 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
604 * Create a new subvolume below @parent. This is largely modeled after
605 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
606 * inside this filesystem so it's quite a bit simpler.
608 static noinline
int btrfs_mksubvol(struct path
*parent
,
609 char *name
, int namelen
,
610 struct btrfs_root
*snap_src
,
611 u64
*async_transid
, bool readonly
)
613 struct inode
*dir
= parent
->dentry
->d_inode
;
614 struct dentry
*dentry
;
617 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
619 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
620 error
= PTR_ERR(dentry
);
628 error
= mnt_want_write(parent
->mnt
);
632 error
= btrfs_may_create(dir
, dentry
);
636 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
638 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
642 error
= create_snapshot(snap_src
, dentry
,
643 name
, namelen
, async_transid
, readonly
);
645 error
= create_subvol(BTRFS_I(dir
)->root
, dentry
,
646 name
, namelen
, async_transid
);
649 fsnotify_mkdir(dir
, dentry
);
651 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
653 mnt_drop_write(parent
->mnt
);
657 mutex_unlock(&dir
->i_mutex
);
662 * When we're defragging a range, we don't want to kick it off again
663 * if it is really just waiting for delalloc to send it down.
664 * If we find a nice big extent or delalloc range for the bytes in the
665 * file you want to defrag, we return 0 to let you know to skip this
668 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, int thresh
)
670 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
671 struct extent_map
*em
= NULL
;
672 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
675 read_lock(&em_tree
->lock
);
676 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
677 read_unlock(&em_tree
->lock
);
680 end
= extent_map_end(em
);
682 if (end
- offset
> thresh
)
685 /* if we already have a nice delalloc here, just stop */
687 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
688 thresh
, EXTENT_DELALLOC
, 1);
695 * helper function to walk through a file and find extents
696 * newer than a specific transid, and smaller than thresh.
698 * This is used by the defragging code to find new and small
701 static int find_new_extents(struct btrfs_root
*root
,
702 struct inode
*inode
, u64 newer_than
,
703 u64
*off
, int thresh
)
705 struct btrfs_path
*path
;
706 struct btrfs_key min_key
;
707 struct btrfs_key max_key
;
708 struct extent_buffer
*leaf
;
709 struct btrfs_file_extent_item
*extent
;
712 u64 ino
= btrfs_ino(inode
);
714 path
= btrfs_alloc_path();
718 min_key
.objectid
= ino
;
719 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
720 min_key
.offset
= *off
;
722 max_key
.objectid
= ino
;
723 max_key
.type
= (u8
)-1;
724 max_key
.offset
= (u64
)-1;
726 path
->keep_locks
= 1;
729 ret
= btrfs_search_forward(root
, &min_key
, &max_key
,
730 path
, 0, newer_than
);
733 if (min_key
.objectid
!= ino
)
735 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
738 leaf
= path
->nodes
[0];
739 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
740 struct btrfs_file_extent_item
);
742 type
= btrfs_file_extent_type(leaf
, extent
);
743 if (type
== BTRFS_FILE_EXTENT_REG
&&
744 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
745 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
746 *off
= min_key
.offset
;
747 btrfs_free_path(path
);
751 if (min_key
.offset
== (u64
)-1)
755 btrfs_release_path(path
);
758 btrfs_free_path(path
);
762 static int should_defrag_range(struct inode
*inode
, u64 start
, u64 len
,
763 int thresh
, u64
*last_len
, u64
*skip
,
766 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
767 struct extent_map
*em
= NULL
;
768 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
772 * make sure that once we start defragging an extent, we keep on
775 if (start
< *defrag_end
)
781 * hopefully we have this extent in the tree already, try without
782 * the full extent lock
784 read_lock(&em_tree
->lock
);
785 em
= lookup_extent_mapping(em_tree
, start
, len
);
786 read_unlock(&em_tree
->lock
);
789 /* get the big lock and read metadata off disk */
790 lock_extent(io_tree
, start
, start
+ len
- 1, GFP_NOFS
);
791 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
792 unlock_extent(io_tree
, start
, start
+ len
- 1, GFP_NOFS
);
798 /* this will cover holes, and inline extents */
799 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
)
803 * we hit a real extent, if it is big don't bother defragging it again
805 if ((*last_len
== 0 || *last_len
>= thresh
) && em
->len
>= thresh
)
809 * last_len ends up being a counter of how many bytes we've defragged.
810 * every time we choose not to defrag an extent, we reset *last_len
811 * so that the next tiny extent will force a defrag.
813 * The end result of this is that tiny extents before a single big
814 * extent will force at least part of that big extent to be defragged.
817 *defrag_end
= extent_map_end(em
);
820 *skip
= extent_map_end(em
);
829 * it doesn't do much good to defrag one or two pages
830 * at a time. This pulls in a nice chunk of pages
833 * It also makes sure the delalloc code has enough
834 * dirty data to avoid making new small extents as part
837 * It's a good idea to start RA on this range
838 * before calling this.
840 static int cluster_pages_for_defrag(struct inode
*inode
,
842 unsigned long start_index
,
845 unsigned long file_end
;
846 u64 isize
= i_size_read(inode
);
852 struct btrfs_ordered_extent
*ordered
;
853 struct extent_state
*cached_state
= NULL
;
854 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
858 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
860 ret
= btrfs_delalloc_reserve_space(inode
,
861 num_pages
<< PAGE_CACHE_SHIFT
);
868 /* step one, lock all the pages */
869 for (i
= 0; i
< num_pages
; i
++) {
871 page
= find_or_create_page(inode
->i_mapping
,
872 start_index
+ i
, mask
);
876 if (!PageUptodate(page
)) {
877 btrfs_readpage(NULL
, page
);
879 if (!PageUptodate(page
)) {
881 page_cache_release(page
);
886 isize
= i_size_read(inode
);
887 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
888 if (!isize
|| page
->index
> file_end
||
889 page
->mapping
!= inode
->i_mapping
) {
890 /* whoops, we blew past eof, skip this page */
892 page_cache_release(page
);
901 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
905 * so now we have a nice long stream of locked
906 * and up to date pages, lets wait on them
908 for (i
= 0; i
< i_done
; i
++)
909 wait_on_page_writeback(pages
[i
]);
911 page_start
= page_offset(pages
[0]);
912 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
914 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
915 page_start
, page_end
- 1, 0, &cached_state
,
917 ordered
= btrfs_lookup_first_ordered_extent(inode
, page_end
- 1);
919 ordered
->file_offset
+ ordered
->len
> page_start
&&
920 ordered
->file_offset
< page_end
) {
921 btrfs_put_ordered_extent(ordered
);
922 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
923 page_start
, page_end
- 1,
924 &cached_state
, GFP_NOFS
);
925 for (i
= 0; i
< i_done
; i
++) {
926 unlock_page(pages
[i
]);
927 page_cache_release(pages
[i
]);
929 btrfs_wait_ordered_range(inode
, page_start
,
930 page_end
- page_start
);
934 btrfs_put_ordered_extent(ordered
);
936 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
937 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
938 EXTENT_DO_ACCOUNTING
, 0, 0, &cached_state
,
941 if (i_done
!= num_pages
) {
942 spin_lock(&BTRFS_I(inode
)->lock
);
943 BTRFS_I(inode
)->outstanding_extents
++;
944 spin_unlock(&BTRFS_I(inode
)->lock
);
945 btrfs_delalloc_release_space(inode
,
946 (num_pages
- i_done
) << PAGE_CACHE_SHIFT
);
950 btrfs_set_extent_delalloc(inode
, page_start
, page_end
- 1,
953 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
954 page_start
, page_end
- 1, &cached_state
,
957 for (i
= 0; i
< i_done
; i
++) {
958 clear_page_dirty_for_io(pages
[i
]);
959 ClearPageChecked(pages
[i
]);
960 set_page_extent_mapped(pages
[i
]);
961 set_page_dirty(pages
[i
]);
962 unlock_page(pages
[i
]);
963 page_cache_release(pages
[i
]);
967 for (i
= 0; i
< i_done
; i
++) {
968 unlock_page(pages
[i
]);
969 page_cache_release(pages
[i
]);
971 btrfs_delalloc_release_space(inode
, num_pages
<< PAGE_CACHE_SHIFT
);
976 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
977 struct btrfs_ioctl_defrag_range_args
*range
,
978 u64 newer_than
, unsigned long max_to_defrag
)
980 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
981 struct btrfs_super_block
*disk_super
;
982 struct file_ra_state
*ra
= NULL
;
983 unsigned long last_index
;
984 u64 isize
= i_size_read(inode
);
989 u64 newer_off
= range
->start
;
991 unsigned long ra_index
= 0;
993 int defrag_count
= 0;
994 int compress_type
= BTRFS_COMPRESS_ZLIB
;
995 int extent_thresh
= range
->extent_thresh
;
996 int max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
997 int cluster
= max_cluster
;
998 u64 new_align
= ~((u64
)128 * 1024 - 1);
999 struct page
**pages
= NULL
;
1001 if (extent_thresh
== 0)
1002 extent_thresh
= 256 * 1024;
1004 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1005 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1007 if (range
->compress_type
)
1008 compress_type
= range
->compress_type
;
1015 * if we were not given a file, allocate a readahead
1019 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1022 file_ra_state_init(ra
, inode
->i_mapping
);
1027 pages
= kmalloc(sizeof(struct page
*) * max_cluster
,
1034 /* find the last page to defrag */
1035 if (range
->start
+ range
->len
> range
->start
) {
1036 last_index
= min_t(u64
, isize
- 1,
1037 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1039 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1043 ret
= find_new_extents(root
, inode
, newer_than
,
1044 &newer_off
, 64 * 1024);
1046 range
->start
= newer_off
;
1048 * we always align our defrag to help keep
1049 * the extents in the file evenly spaced
1051 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1055 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1058 max_to_defrag
= last_index
;
1061 * make writeback starts from i, so the defrag range can be
1062 * written sequentially.
1064 if (i
< inode
->i_mapping
->writeback_index
)
1065 inode
->i_mapping
->writeback_index
= i
;
1067 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1068 (i
< (i_size_read(inode
) + PAGE_CACHE_SIZE
- 1) >>
1069 PAGE_CACHE_SHIFT
)) {
1071 * make sure we stop running if someone unmounts
1074 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1078 !should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1085 * the should_defrag function tells us how much to skip
1086 * bump our counter by the suggested amount
1088 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
1089 i
= max(i
+ 1, next
);
1094 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1095 PAGE_CACHE_SHIFT
) - i
;
1096 cluster
= min(cluster
, max_cluster
);
1098 cluster
= max_cluster
;
1101 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1102 BTRFS_I(inode
)->force_compress
= compress_type
;
1104 if (i
+ cluster
> ra_index
) {
1105 ra_index
= max(i
, ra_index
);
1106 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1108 ra_index
+= max_cluster
;
1111 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1115 defrag_count
+= ret
;
1116 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, ret
);
1119 if (newer_off
== (u64
)-1)
1122 newer_off
= max(newer_off
+ 1,
1123 (u64
)i
<< PAGE_CACHE_SHIFT
);
1125 ret
= find_new_extents(root
, inode
,
1126 newer_than
, &newer_off
,
1129 range
->start
= newer_off
;
1130 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1137 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1145 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
))
1146 filemap_flush(inode
->i_mapping
);
1148 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1149 /* the filemap_flush will queue IO into the worker threads, but
1150 * we have to make sure the IO is actually started and that
1151 * ordered extents get created before we return
1153 atomic_inc(&root
->fs_info
->async_submit_draining
);
1154 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1155 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1156 wait_event(root
->fs_info
->async_submit_wait
,
1157 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1158 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1160 atomic_dec(&root
->fs_info
->async_submit_draining
);
1162 mutex_lock(&inode
->i_mutex
);
1163 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1164 mutex_unlock(&inode
->i_mutex
);
1167 disk_super
= root
->fs_info
->super_copy
;
1168 features
= btrfs_super_incompat_flags(disk_super
);
1169 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1170 features
|= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO
;
1171 btrfs_set_super_incompat_flags(disk_super
, features
);
1183 static noinline
int btrfs_ioctl_resize(struct btrfs_root
*root
,
1189 struct btrfs_ioctl_vol_args
*vol_args
;
1190 struct btrfs_trans_handle
*trans
;
1191 struct btrfs_device
*device
= NULL
;
1193 char *devstr
= NULL
;
1197 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
1200 if (!capable(CAP_SYS_ADMIN
))
1203 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1204 if (IS_ERR(vol_args
))
1205 return PTR_ERR(vol_args
);
1207 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1209 mutex_lock(&root
->fs_info
->volume_mutex
);
1210 sizestr
= vol_args
->name
;
1211 devstr
= strchr(sizestr
, ':');
1214 sizestr
= devstr
+ 1;
1216 devstr
= vol_args
->name
;
1217 devid
= simple_strtoull(devstr
, &end
, 10);
1218 printk(KERN_INFO
"resizing devid %llu\n",
1219 (unsigned long long)devid
);
1221 device
= btrfs_find_device(root
, devid
, NULL
, NULL
);
1223 printk(KERN_INFO
"resizer unable to find device %llu\n",
1224 (unsigned long long)devid
);
1228 if (!strcmp(sizestr
, "max"))
1229 new_size
= device
->bdev
->bd_inode
->i_size
;
1231 if (sizestr
[0] == '-') {
1234 } else if (sizestr
[0] == '+') {
1238 new_size
= memparse(sizestr
, NULL
);
1239 if (new_size
== 0) {
1245 old_size
= device
->total_bytes
;
1248 if (new_size
> old_size
) {
1252 new_size
= old_size
- new_size
;
1253 } else if (mod
> 0) {
1254 new_size
= old_size
+ new_size
;
1257 if (new_size
< 256 * 1024 * 1024) {
1261 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1266 do_div(new_size
, root
->sectorsize
);
1267 new_size
*= root
->sectorsize
;
1269 printk(KERN_INFO
"new size for %s is %llu\n",
1270 device
->name
, (unsigned long long)new_size
);
1272 if (new_size
> old_size
) {
1273 trans
= btrfs_start_transaction(root
, 0);
1274 if (IS_ERR(trans
)) {
1275 ret
= PTR_ERR(trans
);
1278 ret
= btrfs_grow_device(trans
, device
, new_size
);
1279 btrfs_commit_transaction(trans
, root
);
1281 ret
= btrfs_shrink_device(device
, new_size
);
1285 mutex_unlock(&root
->fs_info
->volume_mutex
);
1290 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1297 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
1298 struct file
*src_file
;
1302 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
1305 namelen
= strlen(name
);
1306 if (strchr(name
, '/')) {
1312 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1313 NULL
, transid
, readonly
);
1315 struct inode
*src_inode
;
1316 src_file
= fget(fd
);
1322 src_inode
= src_file
->f_path
.dentry
->d_inode
;
1323 if (src_inode
->i_sb
!= file
->f_path
.dentry
->d_inode
->i_sb
) {
1324 printk(KERN_INFO
"btrfs: Snapshot src from "
1330 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1331 BTRFS_I(src_inode
)->root
,
1339 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1340 void __user
*arg
, int subvol
)
1342 struct btrfs_ioctl_vol_args
*vol_args
;
1345 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1346 if (IS_ERR(vol_args
))
1347 return PTR_ERR(vol_args
);
1348 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1350 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1351 vol_args
->fd
, subvol
,
1358 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1359 void __user
*arg
, int subvol
)
1361 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1365 bool readonly
= false;
1367 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1368 if (IS_ERR(vol_args
))
1369 return PTR_ERR(vol_args
);
1370 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1372 if (vol_args
->flags
&
1373 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
)) {
1378 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1380 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1383 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1384 vol_args
->fd
, subvol
,
1387 if (ret
== 0 && ptr
&&
1389 offsetof(struct btrfs_ioctl_vol_args_v2
,
1390 transid
), ptr
, sizeof(*ptr
)))
1397 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1400 struct inode
*inode
= fdentry(file
)->d_inode
;
1401 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1405 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1408 down_read(&root
->fs_info
->subvol_sem
);
1409 if (btrfs_root_readonly(root
))
1410 flags
|= BTRFS_SUBVOL_RDONLY
;
1411 up_read(&root
->fs_info
->subvol_sem
);
1413 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1419 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1422 struct inode
*inode
= fdentry(file
)->d_inode
;
1423 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1424 struct btrfs_trans_handle
*trans
;
1429 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
1432 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1435 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
1438 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1441 if (flags
& ~BTRFS_SUBVOL_RDONLY
)
1444 if (!inode_owner_or_capable(inode
))
1447 down_write(&root
->fs_info
->subvol_sem
);
1450 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1453 root_flags
= btrfs_root_flags(&root
->root_item
);
1454 if (flags
& BTRFS_SUBVOL_RDONLY
)
1455 btrfs_set_root_flags(&root
->root_item
,
1456 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1458 btrfs_set_root_flags(&root
->root_item
,
1459 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1461 trans
= btrfs_start_transaction(root
, 1);
1462 if (IS_ERR(trans
)) {
1463 ret
= PTR_ERR(trans
);
1467 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1468 &root
->root_key
, &root
->root_item
);
1470 btrfs_commit_transaction(trans
, root
);
1473 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1475 up_write(&root
->fs_info
->subvol_sem
);
1480 * helper to check if the subvolume references other subvolumes
1482 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1484 struct btrfs_path
*path
;
1485 struct btrfs_key key
;
1488 path
= btrfs_alloc_path();
1492 key
.objectid
= root
->root_key
.objectid
;
1493 key
.type
= BTRFS_ROOT_REF_KEY
;
1494 key
.offset
= (u64
)-1;
1496 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1503 if (path
->slots
[0] > 0) {
1505 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1506 if (key
.objectid
== root
->root_key
.objectid
&&
1507 key
.type
== BTRFS_ROOT_REF_KEY
)
1511 btrfs_free_path(path
);
1515 static noinline
int key_in_sk(struct btrfs_key
*key
,
1516 struct btrfs_ioctl_search_key
*sk
)
1518 struct btrfs_key test
;
1521 test
.objectid
= sk
->min_objectid
;
1522 test
.type
= sk
->min_type
;
1523 test
.offset
= sk
->min_offset
;
1525 ret
= btrfs_comp_cpu_keys(key
, &test
);
1529 test
.objectid
= sk
->max_objectid
;
1530 test
.type
= sk
->max_type
;
1531 test
.offset
= sk
->max_offset
;
1533 ret
= btrfs_comp_cpu_keys(key
, &test
);
1539 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1540 struct btrfs_path
*path
,
1541 struct btrfs_key
*key
,
1542 struct btrfs_ioctl_search_key
*sk
,
1544 unsigned long *sk_offset
,
1548 struct extent_buffer
*leaf
;
1549 struct btrfs_ioctl_search_header sh
;
1550 unsigned long item_off
;
1551 unsigned long item_len
;
1557 leaf
= path
->nodes
[0];
1558 slot
= path
->slots
[0];
1559 nritems
= btrfs_header_nritems(leaf
);
1561 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1565 found_transid
= btrfs_header_generation(leaf
);
1567 for (i
= slot
; i
< nritems
; i
++) {
1568 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1569 item_len
= btrfs_item_size_nr(leaf
, i
);
1571 if (item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1574 if (sizeof(sh
) + item_len
+ *sk_offset
>
1575 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1580 btrfs_item_key_to_cpu(leaf
, key
, i
);
1581 if (!key_in_sk(key
, sk
))
1584 sh
.objectid
= key
->objectid
;
1585 sh
.offset
= key
->offset
;
1586 sh
.type
= key
->type
;
1588 sh
.transid
= found_transid
;
1590 /* copy search result header */
1591 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1592 *sk_offset
+= sizeof(sh
);
1595 char *p
= buf
+ *sk_offset
;
1597 read_extent_buffer(leaf
, p
,
1598 item_off
, item_len
);
1599 *sk_offset
+= item_len
;
1603 if (*num_found
>= sk
->nr_items
)
1608 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1610 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1613 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
1623 static noinline
int search_ioctl(struct inode
*inode
,
1624 struct btrfs_ioctl_search_args
*args
)
1626 struct btrfs_root
*root
;
1627 struct btrfs_key key
;
1628 struct btrfs_key max_key
;
1629 struct btrfs_path
*path
;
1630 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
1631 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
1634 unsigned long sk_offset
= 0;
1636 path
= btrfs_alloc_path();
1640 if (sk
->tree_id
== 0) {
1641 /* search the root of the inode that was passed */
1642 root
= BTRFS_I(inode
)->root
;
1644 key
.objectid
= sk
->tree_id
;
1645 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1646 key
.offset
= (u64
)-1;
1647 root
= btrfs_read_fs_root_no_name(info
, &key
);
1649 printk(KERN_ERR
"could not find root %llu\n",
1651 btrfs_free_path(path
);
1656 key
.objectid
= sk
->min_objectid
;
1657 key
.type
= sk
->min_type
;
1658 key
.offset
= sk
->min_offset
;
1660 max_key
.objectid
= sk
->max_objectid
;
1661 max_key
.type
= sk
->max_type
;
1662 max_key
.offset
= sk
->max_offset
;
1664 path
->keep_locks
= 1;
1667 ret
= btrfs_search_forward(root
, &key
, &max_key
, path
, 0,
1674 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
1675 &sk_offset
, &num_found
);
1676 btrfs_release_path(path
);
1677 if (ret
|| num_found
>= sk
->nr_items
)
1683 sk
->nr_items
= num_found
;
1684 btrfs_free_path(path
);
1688 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
1691 struct btrfs_ioctl_search_args
*args
;
1692 struct inode
*inode
;
1695 if (!capable(CAP_SYS_ADMIN
))
1698 args
= memdup_user(argp
, sizeof(*args
));
1700 return PTR_ERR(args
);
1702 inode
= fdentry(file
)->d_inode
;
1703 ret
= search_ioctl(inode
, args
);
1704 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1711 * Search INODE_REFs to identify path name of 'dirid' directory
1712 * in a 'tree_id' tree. and sets path name to 'name'.
1714 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1715 u64 tree_id
, u64 dirid
, char *name
)
1717 struct btrfs_root
*root
;
1718 struct btrfs_key key
;
1724 struct btrfs_inode_ref
*iref
;
1725 struct extent_buffer
*l
;
1726 struct btrfs_path
*path
;
1728 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1733 path
= btrfs_alloc_path();
1737 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
1739 key
.objectid
= tree_id
;
1740 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1741 key
.offset
= (u64
)-1;
1742 root
= btrfs_read_fs_root_no_name(info
, &key
);
1744 printk(KERN_ERR
"could not find root %llu\n", tree_id
);
1749 key
.objectid
= dirid
;
1750 key
.type
= BTRFS_INODE_REF_KEY
;
1751 key
.offset
= (u64
)-1;
1754 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1759 slot
= path
->slots
[0];
1760 if (ret
> 0 && slot
> 0)
1762 btrfs_item_key_to_cpu(l
, &key
, slot
);
1764 if (ret
> 0 && (key
.objectid
!= dirid
||
1765 key
.type
!= BTRFS_INODE_REF_KEY
)) {
1770 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
1771 len
= btrfs_inode_ref_name_len(l
, iref
);
1773 total_len
+= len
+ 1;
1778 read_extent_buffer(l
, ptr
,(unsigned long)(iref
+ 1), len
);
1780 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
1783 btrfs_release_path(path
);
1784 key
.objectid
= key
.offset
;
1785 key
.offset
= (u64
)-1;
1786 dirid
= key
.objectid
;
1790 memmove(name
, ptr
, total_len
);
1791 name
[total_len
]='\0';
1794 btrfs_free_path(path
);
1798 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
1801 struct btrfs_ioctl_ino_lookup_args
*args
;
1802 struct inode
*inode
;
1805 if (!capable(CAP_SYS_ADMIN
))
1808 args
= memdup_user(argp
, sizeof(*args
));
1810 return PTR_ERR(args
);
1812 inode
= fdentry(file
)->d_inode
;
1814 if (args
->treeid
== 0)
1815 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
1817 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
1818 args
->treeid
, args
->objectid
,
1821 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1828 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
1831 struct dentry
*parent
= fdentry(file
);
1832 struct dentry
*dentry
;
1833 struct inode
*dir
= parent
->d_inode
;
1834 struct inode
*inode
;
1835 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1836 struct btrfs_root
*dest
= NULL
;
1837 struct btrfs_ioctl_vol_args
*vol_args
;
1838 struct btrfs_trans_handle
*trans
;
1843 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1844 if (IS_ERR(vol_args
))
1845 return PTR_ERR(vol_args
);
1847 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1848 namelen
= strlen(vol_args
->name
);
1849 if (strchr(vol_args
->name
, '/') ||
1850 strncmp(vol_args
->name
, "..", namelen
) == 0) {
1855 err
= mnt_want_write(file
->f_path
.mnt
);
1859 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
1860 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
1861 if (IS_ERR(dentry
)) {
1862 err
= PTR_ERR(dentry
);
1863 goto out_unlock_dir
;
1866 if (!dentry
->d_inode
) {
1871 inode
= dentry
->d_inode
;
1872 dest
= BTRFS_I(inode
)->root
;
1873 if (!capable(CAP_SYS_ADMIN
)){
1875 * Regular user. Only allow this with a special mount
1876 * option, when the user has write+exec access to the
1877 * subvol root, and when rmdir(2) would have been
1880 * Note that this is _not_ check that the subvol is
1881 * empty or doesn't contain data that we wouldn't
1882 * otherwise be able to delete.
1884 * Users who want to delete empty subvols should try
1888 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
1892 * Do not allow deletion if the parent dir is the same
1893 * as the dir to be deleted. That means the ioctl
1894 * must be called on the dentry referencing the root
1895 * of the subvol, not a random directory contained
1902 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
1906 /* check if subvolume may be deleted by a non-root user */
1907 err
= btrfs_may_delete(dir
, dentry
, 1);
1912 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1917 mutex_lock(&inode
->i_mutex
);
1918 err
= d_invalidate(dentry
);
1922 down_write(&root
->fs_info
->subvol_sem
);
1924 err
= may_destroy_subvol(dest
);
1928 trans
= btrfs_start_transaction(root
, 0);
1929 if (IS_ERR(trans
)) {
1930 err
= PTR_ERR(trans
);
1933 trans
->block_rsv
= &root
->fs_info
->global_block_rsv
;
1935 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
1936 dest
->root_key
.objectid
,
1937 dentry
->d_name
.name
,
1938 dentry
->d_name
.len
);
1941 btrfs_record_root_in_trans(trans
, dest
);
1943 memset(&dest
->root_item
.drop_progress
, 0,
1944 sizeof(dest
->root_item
.drop_progress
));
1945 dest
->root_item
.drop_level
= 0;
1946 btrfs_set_root_refs(&dest
->root_item
, 0);
1948 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
1949 ret
= btrfs_insert_orphan_item(trans
,
1950 root
->fs_info
->tree_root
,
1951 dest
->root_key
.objectid
);
1955 ret
= btrfs_end_transaction(trans
, root
);
1957 inode
->i_flags
|= S_DEAD
;
1959 up_write(&root
->fs_info
->subvol_sem
);
1961 mutex_unlock(&inode
->i_mutex
);
1963 shrink_dcache_sb(root
->fs_info
->sb
);
1964 btrfs_invalidate_inodes(dest
);
1970 mutex_unlock(&dir
->i_mutex
);
1971 mnt_drop_write(file
->f_path
.mnt
);
1977 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
1979 struct inode
*inode
= fdentry(file
)->d_inode
;
1980 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1981 struct btrfs_ioctl_defrag_range_args
*range
;
1984 if (btrfs_root_readonly(root
))
1987 ret
= mnt_want_write(file
->f_path
.mnt
);
1991 switch (inode
->i_mode
& S_IFMT
) {
1993 if (!capable(CAP_SYS_ADMIN
)) {
1997 ret
= btrfs_defrag_root(root
, 0);
2000 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
2003 if (!(file
->f_mode
& FMODE_WRITE
)) {
2008 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2015 if (copy_from_user(range
, argp
,
2021 /* compression requires us to start the IO */
2022 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2023 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2024 range
->extent_thresh
= (u32
)-1;
2027 /* the rest are all set to zero by kzalloc */
2028 range
->len
= (u64
)-1;
2030 ret
= btrfs_defrag_file(fdentry(file
)->d_inode
, file
,
2040 mnt_drop_write(file
->f_path
.mnt
);
2044 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2046 struct btrfs_ioctl_vol_args
*vol_args
;
2049 if (!capable(CAP_SYS_ADMIN
))
2052 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2053 if (IS_ERR(vol_args
))
2054 return PTR_ERR(vol_args
);
2056 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2057 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2063 static long btrfs_ioctl_rm_dev(struct btrfs_root
*root
, void __user
*arg
)
2065 struct btrfs_ioctl_vol_args
*vol_args
;
2068 if (!capable(CAP_SYS_ADMIN
))
2071 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
2074 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2075 if (IS_ERR(vol_args
))
2076 return PTR_ERR(vol_args
);
2078 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2079 ret
= btrfs_rm_device(root
, vol_args
->name
);
2085 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2087 struct btrfs_ioctl_fs_info_args
*fi_args
;
2088 struct btrfs_device
*device
;
2089 struct btrfs_device
*next
;
2090 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2093 if (!capable(CAP_SYS_ADMIN
))
2096 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2100 fi_args
->num_devices
= fs_devices
->num_devices
;
2101 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2103 mutex_lock(&fs_devices
->device_list_mutex
);
2104 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2105 if (device
->devid
> fi_args
->max_id
)
2106 fi_args
->max_id
= device
->devid
;
2108 mutex_unlock(&fs_devices
->device_list_mutex
);
2110 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2117 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2119 struct btrfs_ioctl_dev_info_args
*di_args
;
2120 struct btrfs_device
*dev
;
2121 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2123 char *s_uuid
= NULL
;
2124 char empty_uuid
[BTRFS_UUID_SIZE
] = {0};
2126 if (!capable(CAP_SYS_ADMIN
))
2129 di_args
= memdup_user(arg
, sizeof(*di_args
));
2130 if (IS_ERR(di_args
))
2131 return PTR_ERR(di_args
);
2133 if (memcmp(empty_uuid
, di_args
->uuid
, BTRFS_UUID_SIZE
) != 0)
2134 s_uuid
= di_args
->uuid
;
2136 mutex_lock(&fs_devices
->device_list_mutex
);
2137 dev
= btrfs_find_device(root
, di_args
->devid
, s_uuid
, NULL
);
2138 mutex_unlock(&fs_devices
->device_list_mutex
);
2145 di_args
->devid
= dev
->devid
;
2146 di_args
->bytes_used
= dev
->bytes_used
;
2147 di_args
->total_bytes
= dev
->total_bytes
;
2148 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2149 strncpy(di_args
->path
, dev
->name
, sizeof(di_args
->path
));
2152 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2159 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
2160 u64 off
, u64 olen
, u64 destoff
)
2162 struct inode
*inode
= fdentry(file
)->d_inode
;
2163 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2164 struct file
*src_file
;
2166 struct btrfs_trans_handle
*trans
;
2167 struct btrfs_path
*path
;
2168 struct extent_buffer
*leaf
;
2170 struct btrfs_key key
;
2175 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
2180 * - split compressed inline extents. annoying: we need to
2181 * decompress into destination's address_space (the file offset
2182 * may change, so source mapping won't do), then recompress (or
2183 * otherwise reinsert) a subrange.
2184 * - allow ranges within the same file to be cloned (provided
2185 * they don't overlap)?
2188 /* the destination must be opened for writing */
2189 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
2192 if (btrfs_root_readonly(root
))
2195 ret
= mnt_want_write(file
->f_path
.mnt
);
2199 src_file
= fget(srcfd
);
2202 goto out_drop_write
;
2205 src
= src_file
->f_dentry
->d_inode
;
2211 /* the src must be open for reading */
2212 if (!(src_file
->f_mode
& FMODE_READ
))
2215 /* don't make the dst file partly checksummed */
2216 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2217 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
2221 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
2225 if (src
->i_sb
!= inode
->i_sb
|| BTRFS_I(src
)->root
!= root
)
2229 buf
= vmalloc(btrfs_level_size(root
, 0));
2233 path
= btrfs_alloc_path();
2241 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
2242 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
2244 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
2245 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
2248 /* determine range to clone */
2250 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
2253 olen
= len
= src
->i_size
- off
;
2254 /* if we extend to eof, continue to block boundary */
2255 if (off
+ len
== src
->i_size
)
2256 len
= ALIGN(src
->i_size
, bs
) - off
;
2258 /* verify the end result is block aligned */
2259 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
2260 !IS_ALIGNED(destoff
, bs
))
2263 if (destoff
> inode
->i_size
) {
2264 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
2269 /* truncate page cache pages from target inode range */
2270 truncate_inode_pages_range(&inode
->i_data
, destoff
,
2271 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
2273 /* do any pending delalloc/csum calc on src, one way or
2274 another, and lock file content */
2276 struct btrfs_ordered_extent
*ordered
;
2277 lock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+len
, GFP_NOFS
);
2278 ordered
= btrfs_lookup_first_ordered_extent(src
, off
+len
);
2280 !test_range_bit(&BTRFS_I(src
)->io_tree
, off
, off
+len
,
2281 EXTENT_DELALLOC
, 0, NULL
))
2283 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+len
, GFP_NOFS
);
2285 btrfs_put_ordered_extent(ordered
);
2286 btrfs_wait_ordered_range(src
, off
, len
);
2290 key
.objectid
= btrfs_ino(src
);
2291 key
.type
= BTRFS_EXTENT_DATA_KEY
;
2296 * note the key will change type as we walk through the
2299 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2303 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2304 if (path
->slots
[0] >= nritems
) {
2305 ret
= btrfs_next_leaf(root
, path
);
2310 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2312 leaf
= path
->nodes
[0];
2313 slot
= path
->slots
[0];
2315 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2316 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
2317 key
.objectid
!= btrfs_ino(src
))
2320 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
2321 struct btrfs_file_extent_item
*extent
;
2324 struct btrfs_key new_key
;
2325 u64 disko
= 0, diskl
= 0;
2326 u64 datao
= 0, datal
= 0;
2330 size
= btrfs_item_size_nr(leaf
, slot
);
2331 read_extent_buffer(leaf
, buf
,
2332 btrfs_item_ptr_offset(leaf
, slot
),
2335 extent
= btrfs_item_ptr(leaf
, slot
,
2336 struct btrfs_file_extent_item
);
2337 comp
= btrfs_file_extent_compression(leaf
, extent
);
2338 type
= btrfs_file_extent_type(leaf
, extent
);
2339 if (type
== BTRFS_FILE_EXTENT_REG
||
2340 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2341 disko
= btrfs_file_extent_disk_bytenr(leaf
,
2343 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
2345 datao
= btrfs_file_extent_offset(leaf
, extent
);
2346 datal
= btrfs_file_extent_num_bytes(leaf
,
2348 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2349 /* take upper bound, may be compressed */
2350 datal
= btrfs_file_extent_ram_bytes(leaf
,
2353 btrfs_release_path(path
);
2355 if (key
.offset
+ datal
<= off
||
2356 key
.offset
>= off
+len
)
2359 memcpy(&new_key
, &key
, sizeof(new_key
));
2360 new_key
.objectid
= btrfs_ino(inode
);
2361 if (off
<= key
.offset
)
2362 new_key
.offset
= key
.offset
+ destoff
- off
;
2364 new_key
.offset
= destoff
;
2367 * 1 - adjusting old extent (we may have to split it)
2368 * 1 - add new extent
2371 trans
= btrfs_start_transaction(root
, 3);
2372 if (IS_ERR(trans
)) {
2373 ret
= PTR_ERR(trans
);
2377 if (type
== BTRFS_FILE_EXTENT_REG
||
2378 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2380 * a | --- range to clone ---| b
2381 * | ------------- extent ------------- |
2384 /* substract range b */
2385 if (key
.offset
+ datal
> off
+ len
)
2386 datal
= off
+ len
- key
.offset
;
2388 /* substract range a */
2389 if (off
> key
.offset
) {
2390 datao
+= off
- key
.offset
;
2391 datal
-= off
- key
.offset
;
2394 ret
= btrfs_drop_extents(trans
, inode
,
2396 new_key
.offset
+ datal
,
2400 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2404 leaf
= path
->nodes
[0];
2405 slot
= path
->slots
[0];
2406 write_extent_buffer(leaf
, buf
,
2407 btrfs_item_ptr_offset(leaf
, slot
),
2410 extent
= btrfs_item_ptr(leaf
, slot
,
2411 struct btrfs_file_extent_item
);
2413 /* disko == 0 means it's a hole */
2417 btrfs_set_file_extent_offset(leaf
, extent
,
2419 btrfs_set_file_extent_num_bytes(leaf
, extent
,
2422 inode_add_bytes(inode
, datal
);
2423 ret
= btrfs_inc_extent_ref(trans
, root
,
2425 root
->root_key
.objectid
,
2427 new_key
.offset
- datao
);
2430 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2433 if (off
> key
.offset
) {
2434 skip
= off
- key
.offset
;
2435 new_key
.offset
+= skip
;
2438 if (key
.offset
+ datal
> off
+len
)
2439 trim
= key
.offset
+ datal
- (off
+len
);
2441 if (comp
&& (skip
|| trim
)) {
2443 btrfs_end_transaction(trans
, root
);
2446 size
-= skip
+ trim
;
2447 datal
-= skip
+ trim
;
2449 ret
= btrfs_drop_extents(trans
, inode
,
2451 new_key
.offset
+ datal
,
2455 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2461 btrfs_file_extent_calc_inline_size(0);
2462 memmove(buf
+start
, buf
+start
+skip
,
2466 leaf
= path
->nodes
[0];
2467 slot
= path
->slots
[0];
2468 write_extent_buffer(leaf
, buf
,
2469 btrfs_item_ptr_offset(leaf
, slot
),
2471 inode_add_bytes(inode
, datal
);
2474 btrfs_mark_buffer_dirty(leaf
);
2475 btrfs_release_path(path
);
2477 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2480 * we round up to the block size at eof when
2481 * determining which extents to clone above,
2482 * but shouldn't round up the file size
2484 endoff
= new_key
.offset
+ datal
;
2485 if (endoff
> destoff
+olen
)
2486 endoff
= destoff
+olen
;
2487 if (endoff
> inode
->i_size
)
2488 btrfs_i_size_write(inode
, endoff
);
2490 ret
= btrfs_update_inode(trans
, root
, inode
);
2492 btrfs_end_transaction(trans
, root
);
2495 btrfs_release_path(path
);
2500 btrfs_release_path(path
);
2501 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+len
, GFP_NOFS
);
2503 mutex_unlock(&src
->i_mutex
);
2504 mutex_unlock(&inode
->i_mutex
);
2506 btrfs_free_path(path
);
2510 mnt_drop_write(file
->f_path
.mnt
);
2514 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
2516 struct btrfs_ioctl_clone_range_args args
;
2518 if (copy_from_user(&args
, argp
, sizeof(args
)))
2520 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
2521 args
.src_length
, args
.dest_offset
);
2525 * there are many ways the trans_start and trans_end ioctls can lead
2526 * to deadlocks. They should only be used by applications that
2527 * basically own the machine, and have a very in depth understanding
2528 * of all the possible deadlocks and enospc problems.
2530 static long btrfs_ioctl_trans_start(struct file
*file
)
2532 struct inode
*inode
= fdentry(file
)->d_inode
;
2533 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2534 struct btrfs_trans_handle
*trans
;
2538 if (!capable(CAP_SYS_ADMIN
))
2542 if (file
->private_data
)
2546 if (btrfs_root_readonly(root
))
2549 ret
= mnt_want_write(file
->f_path
.mnt
);
2553 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
2556 trans
= btrfs_start_ioctl_transaction(root
);
2560 file
->private_data
= trans
;
2564 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
2565 mnt_drop_write(file
->f_path
.mnt
);
2570 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2572 struct inode
*inode
= fdentry(file
)->d_inode
;
2573 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2574 struct btrfs_root
*new_root
;
2575 struct btrfs_dir_item
*di
;
2576 struct btrfs_trans_handle
*trans
;
2577 struct btrfs_path
*path
;
2578 struct btrfs_key location
;
2579 struct btrfs_disk_key disk_key
;
2580 struct btrfs_super_block
*disk_super
;
2585 if (!capable(CAP_SYS_ADMIN
))
2588 if (copy_from_user(&objectid
, argp
, sizeof(objectid
)))
2592 objectid
= root
->root_key
.objectid
;
2594 location
.objectid
= objectid
;
2595 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2596 location
.offset
= (u64
)-1;
2598 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
2599 if (IS_ERR(new_root
))
2600 return PTR_ERR(new_root
);
2602 if (btrfs_root_refs(&new_root
->root_item
) == 0)
2605 path
= btrfs_alloc_path();
2608 path
->leave_spinning
= 1;
2610 trans
= btrfs_start_transaction(root
, 1);
2611 if (IS_ERR(trans
)) {
2612 btrfs_free_path(path
);
2613 return PTR_ERR(trans
);
2616 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
2617 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
2618 dir_id
, "default", 7, 1);
2619 if (IS_ERR_OR_NULL(di
)) {
2620 btrfs_free_path(path
);
2621 btrfs_end_transaction(trans
, root
);
2622 printk(KERN_ERR
"Umm, you don't have the default dir item, "
2623 "this isn't going to work\n");
2627 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
2628 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
2629 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2630 btrfs_free_path(path
);
2632 disk_super
= root
->fs_info
->super_copy
;
2633 features
= btrfs_super_incompat_flags(disk_super
);
2634 if (!(features
& BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL
)) {
2635 features
|= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL
;
2636 btrfs_set_super_incompat_flags(disk_super
, features
);
2638 btrfs_end_transaction(trans
, root
);
2643 static void get_block_group_info(struct list_head
*groups_list
,
2644 struct btrfs_ioctl_space_info
*space
)
2646 struct btrfs_block_group_cache
*block_group
;
2648 space
->total_bytes
= 0;
2649 space
->used_bytes
= 0;
2651 list_for_each_entry(block_group
, groups_list
, list
) {
2652 space
->flags
= block_group
->flags
;
2653 space
->total_bytes
+= block_group
->key
.offset
;
2654 space
->used_bytes
+=
2655 btrfs_block_group_used(&block_group
->item
);
2659 long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
2661 struct btrfs_ioctl_space_args space_args
;
2662 struct btrfs_ioctl_space_info space
;
2663 struct btrfs_ioctl_space_info
*dest
;
2664 struct btrfs_ioctl_space_info
*dest_orig
;
2665 struct btrfs_ioctl_space_info __user
*user_dest
;
2666 struct btrfs_space_info
*info
;
2667 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
2668 BTRFS_BLOCK_GROUP_SYSTEM
,
2669 BTRFS_BLOCK_GROUP_METADATA
,
2670 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
2677 if (copy_from_user(&space_args
,
2678 (struct btrfs_ioctl_space_args __user
*)arg
,
2679 sizeof(space_args
)))
2682 for (i
= 0; i
< num_types
; i
++) {
2683 struct btrfs_space_info
*tmp
;
2687 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
2689 if (tmp
->flags
== types
[i
]) {
2699 down_read(&info
->groups_sem
);
2700 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
2701 if (!list_empty(&info
->block_groups
[c
]))
2704 up_read(&info
->groups_sem
);
2707 /* space_slots == 0 means they are asking for a count */
2708 if (space_args
.space_slots
== 0) {
2709 space_args
.total_spaces
= slot_count
;
2713 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
2715 alloc_size
= sizeof(*dest
) * slot_count
;
2717 /* we generally have at most 6 or so space infos, one for each raid
2718 * level. So, a whole page should be more than enough for everyone
2720 if (alloc_size
> PAGE_CACHE_SIZE
)
2723 space_args
.total_spaces
= 0;
2724 dest
= kmalloc(alloc_size
, GFP_NOFS
);
2729 /* now we have a buffer to copy into */
2730 for (i
= 0; i
< num_types
; i
++) {
2731 struct btrfs_space_info
*tmp
;
2738 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
2740 if (tmp
->flags
== types
[i
]) {
2749 down_read(&info
->groups_sem
);
2750 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
2751 if (!list_empty(&info
->block_groups
[c
])) {
2752 get_block_group_info(&info
->block_groups
[c
],
2754 memcpy(dest
, &space
, sizeof(space
));
2756 space_args
.total_spaces
++;
2762 up_read(&info
->groups_sem
);
2765 user_dest
= (struct btrfs_ioctl_space_info
*)
2766 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
2768 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
2773 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
2780 * there are many ways the trans_start and trans_end ioctls can lead
2781 * to deadlocks. They should only be used by applications that
2782 * basically own the machine, and have a very in depth understanding
2783 * of all the possible deadlocks and enospc problems.
2785 long btrfs_ioctl_trans_end(struct file
*file
)
2787 struct inode
*inode
= fdentry(file
)->d_inode
;
2788 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2789 struct btrfs_trans_handle
*trans
;
2791 trans
= file
->private_data
;
2794 file
->private_data
= NULL
;
2796 btrfs_end_transaction(trans
, root
);
2798 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
2800 mnt_drop_write(file
->f_path
.mnt
);
2804 static noinline
long btrfs_ioctl_start_sync(struct file
*file
, void __user
*argp
)
2806 struct btrfs_root
*root
= BTRFS_I(file
->f_dentry
->d_inode
)->root
;
2807 struct btrfs_trans_handle
*trans
;
2811 trans
= btrfs_start_transaction(root
, 0);
2813 return PTR_ERR(trans
);
2814 transid
= trans
->transid
;
2815 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
2817 btrfs_end_transaction(trans
, root
);
2822 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
2827 static noinline
long btrfs_ioctl_wait_sync(struct file
*file
, void __user
*argp
)
2829 struct btrfs_root
*root
= BTRFS_I(file
->f_dentry
->d_inode
)->root
;
2833 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
2836 transid
= 0; /* current trans */
2838 return btrfs_wait_for_commit(root
, transid
);
2841 static long btrfs_ioctl_scrub(struct btrfs_root
*root
, void __user
*arg
)
2844 struct btrfs_ioctl_scrub_args
*sa
;
2846 if (!capable(CAP_SYS_ADMIN
))
2849 sa
= memdup_user(arg
, sizeof(*sa
));
2853 ret
= btrfs_scrub_dev(root
, sa
->devid
, sa
->start
, sa
->end
,
2854 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
);
2856 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
2863 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
2865 if (!capable(CAP_SYS_ADMIN
))
2868 return btrfs_scrub_cancel(root
);
2871 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
2874 struct btrfs_ioctl_scrub_args
*sa
;
2877 if (!capable(CAP_SYS_ADMIN
))
2880 sa
= memdup_user(arg
, sizeof(*sa
));
2884 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
2886 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
2893 long btrfs_ioctl(struct file
*file
, unsigned int
2894 cmd
, unsigned long arg
)
2896 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
2897 void __user
*argp
= (void __user
*)arg
;
2900 case FS_IOC_GETFLAGS
:
2901 return btrfs_ioctl_getflags(file
, argp
);
2902 case FS_IOC_SETFLAGS
:
2903 return btrfs_ioctl_setflags(file
, argp
);
2904 case FS_IOC_GETVERSION
:
2905 return btrfs_ioctl_getversion(file
, argp
);
2907 return btrfs_ioctl_fitrim(file
, argp
);
2908 case BTRFS_IOC_SNAP_CREATE
:
2909 return btrfs_ioctl_snap_create(file
, argp
, 0);
2910 case BTRFS_IOC_SNAP_CREATE_V2
:
2911 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
2912 case BTRFS_IOC_SUBVOL_CREATE
:
2913 return btrfs_ioctl_snap_create(file
, argp
, 1);
2914 case BTRFS_IOC_SNAP_DESTROY
:
2915 return btrfs_ioctl_snap_destroy(file
, argp
);
2916 case BTRFS_IOC_SUBVOL_GETFLAGS
:
2917 return btrfs_ioctl_subvol_getflags(file
, argp
);
2918 case BTRFS_IOC_SUBVOL_SETFLAGS
:
2919 return btrfs_ioctl_subvol_setflags(file
, argp
);
2920 case BTRFS_IOC_DEFAULT_SUBVOL
:
2921 return btrfs_ioctl_default_subvol(file
, argp
);
2922 case BTRFS_IOC_DEFRAG
:
2923 return btrfs_ioctl_defrag(file
, NULL
);
2924 case BTRFS_IOC_DEFRAG_RANGE
:
2925 return btrfs_ioctl_defrag(file
, argp
);
2926 case BTRFS_IOC_RESIZE
:
2927 return btrfs_ioctl_resize(root
, argp
);
2928 case BTRFS_IOC_ADD_DEV
:
2929 return btrfs_ioctl_add_dev(root
, argp
);
2930 case BTRFS_IOC_RM_DEV
:
2931 return btrfs_ioctl_rm_dev(root
, argp
);
2932 case BTRFS_IOC_FS_INFO
:
2933 return btrfs_ioctl_fs_info(root
, argp
);
2934 case BTRFS_IOC_DEV_INFO
:
2935 return btrfs_ioctl_dev_info(root
, argp
);
2936 case BTRFS_IOC_BALANCE
:
2937 return btrfs_balance(root
->fs_info
->dev_root
);
2938 case BTRFS_IOC_CLONE
:
2939 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
2940 case BTRFS_IOC_CLONE_RANGE
:
2941 return btrfs_ioctl_clone_range(file
, argp
);
2942 case BTRFS_IOC_TRANS_START
:
2943 return btrfs_ioctl_trans_start(file
);
2944 case BTRFS_IOC_TRANS_END
:
2945 return btrfs_ioctl_trans_end(file
);
2946 case BTRFS_IOC_TREE_SEARCH
:
2947 return btrfs_ioctl_tree_search(file
, argp
);
2948 case BTRFS_IOC_INO_LOOKUP
:
2949 return btrfs_ioctl_ino_lookup(file
, argp
);
2950 case BTRFS_IOC_SPACE_INFO
:
2951 return btrfs_ioctl_space_info(root
, argp
);
2952 case BTRFS_IOC_SYNC
:
2953 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
2955 case BTRFS_IOC_START_SYNC
:
2956 return btrfs_ioctl_start_sync(file
, argp
);
2957 case BTRFS_IOC_WAIT_SYNC
:
2958 return btrfs_ioctl_wait_sync(file
, argp
);
2959 case BTRFS_IOC_SCRUB
:
2960 return btrfs_ioctl_scrub(root
, argp
);
2961 case BTRFS_IOC_SCRUB_CANCEL
:
2962 return btrfs_ioctl_scrub_cancel(root
, argp
);
2963 case BTRFS_IOC_SCRUB_PROGRESS
:
2964 return btrfs_ioctl_scrub_progress(root
, argp
);