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"
63 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
64 * structures are incorrect, as the timespec structure from userspace
65 * is 4 bytes too small. We define these alternatives here to teach
66 * the kernel about the 32-bit struct packing.
68 struct btrfs_ioctl_timespec_32
{
71 } __attribute__ ((__packed__
));
73 struct btrfs_ioctl_received_subvol_args_32
{
74 char uuid
[BTRFS_UUID_SIZE
]; /* in */
75 __u64 stransid
; /* in */
76 __u64 rtransid
; /* out */
77 struct btrfs_ioctl_timespec_32 stime
; /* in */
78 struct btrfs_ioctl_timespec_32 rtime
; /* out */
80 __u64 reserved
[16]; /* in */
81 } __attribute__ ((__packed__
));
83 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
84 struct btrfs_ioctl_received_subvol_args_32)
88 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
89 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
);
91 /* Mask out flags that are inappropriate for the given type of inode. */
92 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
96 else if (S_ISREG(mode
))
97 return flags
& ~FS_DIRSYNC_FL
;
99 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
103 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
107 unsigned int iflags
= 0;
109 if (flags
& BTRFS_INODE_SYNC
)
110 iflags
|= FS_SYNC_FL
;
111 if (flags
& BTRFS_INODE_IMMUTABLE
)
112 iflags
|= FS_IMMUTABLE_FL
;
113 if (flags
& BTRFS_INODE_APPEND
)
114 iflags
|= FS_APPEND_FL
;
115 if (flags
& BTRFS_INODE_NODUMP
)
116 iflags
|= FS_NODUMP_FL
;
117 if (flags
& BTRFS_INODE_NOATIME
)
118 iflags
|= FS_NOATIME_FL
;
119 if (flags
& BTRFS_INODE_DIRSYNC
)
120 iflags
|= FS_DIRSYNC_FL
;
121 if (flags
& BTRFS_INODE_NODATACOW
)
122 iflags
|= FS_NOCOW_FL
;
124 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
125 iflags
|= FS_COMPR_FL
;
126 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
127 iflags
|= FS_NOCOMP_FL
;
133 * Update inode->i_flags based on the btrfs internal flags.
135 void btrfs_update_iflags(struct inode
*inode
)
137 struct btrfs_inode
*ip
= BTRFS_I(inode
);
139 inode
->i_flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
141 if (ip
->flags
& BTRFS_INODE_SYNC
)
142 inode
->i_flags
|= S_SYNC
;
143 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
144 inode
->i_flags
|= S_IMMUTABLE
;
145 if (ip
->flags
& BTRFS_INODE_APPEND
)
146 inode
->i_flags
|= S_APPEND
;
147 if (ip
->flags
& BTRFS_INODE_NOATIME
)
148 inode
->i_flags
|= S_NOATIME
;
149 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
150 inode
->i_flags
|= S_DIRSYNC
;
154 * Inherit flags from the parent inode.
156 * Currently only the compression flags and the cow flags are inherited.
158 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
165 flags
= BTRFS_I(dir
)->flags
;
167 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
168 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
169 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
170 } else if (flags
& BTRFS_INODE_COMPRESS
) {
171 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
172 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
175 if (flags
& BTRFS_INODE_NODATACOW
) {
176 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
177 if (S_ISREG(inode
->i_mode
))
178 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
181 btrfs_update_iflags(inode
);
184 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
186 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
187 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
189 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
194 static int check_flags(unsigned int flags
)
196 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
197 FS_NOATIME_FL
| FS_NODUMP_FL
| \
198 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
199 FS_NOCOMP_FL
| FS_COMPR_FL
|
203 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
209 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
211 struct inode
*inode
= file_inode(file
);
212 struct btrfs_inode
*ip
= BTRFS_I(inode
);
213 struct btrfs_root
*root
= ip
->root
;
214 struct btrfs_trans_handle
*trans
;
215 unsigned int flags
, oldflags
;
218 unsigned int i_oldflags
;
221 if (!inode_owner_or_capable(inode
))
224 if (btrfs_root_readonly(root
))
227 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
230 ret
= check_flags(flags
);
234 ret
= mnt_want_write_file(file
);
238 mutex_lock(&inode
->i_mutex
);
240 ip_oldflags
= ip
->flags
;
241 i_oldflags
= inode
->i_flags
;
242 mode
= inode
->i_mode
;
244 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
245 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
246 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
247 if (!capable(CAP_LINUX_IMMUTABLE
)) {
253 if (flags
& FS_SYNC_FL
)
254 ip
->flags
|= BTRFS_INODE_SYNC
;
256 ip
->flags
&= ~BTRFS_INODE_SYNC
;
257 if (flags
& FS_IMMUTABLE_FL
)
258 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
260 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
261 if (flags
& FS_APPEND_FL
)
262 ip
->flags
|= BTRFS_INODE_APPEND
;
264 ip
->flags
&= ~BTRFS_INODE_APPEND
;
265 if (flags
& FS_NODUMP_FL
)
266 ip
->flags
|= BTRFS_INODE_NODUMP
;
268 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
269 if (flags
& FS_NOATIME_FL
)
270 ip
->flags
|= BTRFS_INODE_NOATIME
;
272 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
273 if (flags
& FS_DIRSYNC_FL
)
274 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
276 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
277 if (flags
& FS_NOCOW_FL
) {
280 * It's safe to turn csums off here, no extents exist.
281 * Otherwise we want the flag to reflect the real COW
282 * status of the file and will not set it.
284 if (inode
->i_size
== 0)
285 ip
->flags
|= BTRFS_INODE_NODATACOW
286 | BTRFS_INODE_NODATASUM
;
288 ip
->flags
|= BTRFS_INODE_NODATACOW
;
292 * Revert back under same assuptions as above
295 if (inode
->i_size
== 0)
296 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
297 | BTRFS_INODE_NODATASUM
);
299 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
304 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
305 * flag may be changed automatically if compression code won't make
308 if (flags
& FS_NOCOMP_FL
) {
309 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
310 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
312 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
313 if (ret
&& ret
!= -ENODATA
)
315 } else if (flags
& FS_COMPR_FL
) {
318 ip
->flags
|= BTRFS_INODE_COMPRESS
;
319 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
321 if (root
->fs_info
->compress_type
== BTRFS_COMPRESS_LZO
)
325 ret
= btrfs_set_prop(inode
, "btrfs.compression",
326 comp
, strlen(comp
), 0);
331 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
334 trans
= btrfs_start_transaction(root
, 1);
336 ret
= PTR_ERR(trans
);
340 btrfs_update_iflags(inode
);
341 inode_inc_iversion(inode
);
342 inode
->i_ctime
= CURRENT_TIME
;
343 ret
= btrfs_update_inode(trans
, root
, inode
);
345 btrfs_end_transaction(trans
, root
);
348 ip
->flags
= ip_oldflags
;
349 inode
->i_flags
= i_oldflags
;
353 mutex_unlock(&inode
->i_mutex
);
354 mnt_drop_write_file(file
);
358 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
360 struct inode
*inode
= file_inode(file
);
362 return put_user(inode
->i_generation
, arg
);
365 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
367 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
368 struct btrfs_device
*device
;
369 struct request_queue
*q
;
370 struct fstrim_range range
;
371 u64 minlen
= ULLONG_MAX
;
373 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
376 if (!capable(CAP_SYS_ADMIN
))
380 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
384 q
= bdev_get_queue(device
->bdev
);
385 if (blk_queue_discard(q
)) {
387 minlen
= min((u64
)q
->limits
.discard_granularity
,
395 if (copy_from_user(&range
, arg
, sizeof(range
)))
397 if (range
.start
> total_bytes
||
398 range
.len
< fs_info
->sb
->s_blocksize
)
401 range
.len
= min(range
.len
, total_bytes
- range
.start
);
402 range
.minlen
= max(range
.minlen
, minlen
);
403 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
407 if (copy_to_user(arg
, &range
, sizeof(range
)))
413 int btrfs_is_empty_uuid(u8
*uuid
)
417 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
424 static noinline
int create_subvol(struct inode
*dir
,
425 struct dentry
*dentry
,
426 char *name
, int namelen
,
428 struct btrfs_qgroup_inherit
*inherit
)
430 struct btrfs_trans_handle
*trans
;
431 struct btrfs_key key
;
432 struct btrfs_root_item root_item
;
433 struct btrfs_inode_item
*inode_item
;
434 struct extent_buffer
*leaf
;
435 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
436 struct btrfs_root
*new_root
;
437 struct btrfs_block_rsv block_rsv
;
438 struct timespec cur_time
= CURRENT_TIME
;
443 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
448 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
452 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
454 * The same as the snapshot creation, please see the comment
455 * of create_snapshot().
457 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
458 8, &qgroup_reserved
, false);
462 trans
= btrfs_start_transaction(root
, 0);
464 ret
= PTR_ERR(trans
);
465 btrfs_subvolume_release_metadata(root
, &block_rsv
,
469 trans
->block_rsv
= &block_rsv
;
470 trans
->bytes_reserved
= block_rsv
.size
;
472 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
476 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
477 0, objectid
, NULL
, 0, 0, 0);
483 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
484 btrfs_set_header_bytenr(leaf
, leaf
->start
);
485 btrfs_set_header_generation(leaf
, trans
->transid
);
486 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
487 btrfs_set_header_owner(leaf
, objectid
);
489 write_extent_buffer(leaf
, root
->fs_info
->fsid
, btrfs_header_fsid(),
491 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
492 btrfs_header_chunk_tree_uuid(leaf
),
494 btrfs_mark_buffer_dirty(leaf
);
496 memset(&root_item
, 0, sizeof(root_item
));
498 inode_item
= &root_item
.inode
;
499 btrfs_set_stack_inode_generation(inode_item
, 1);
500 btrfs_set_stack_inode_size(inode_item
, 3);
501 btrfs_set_stack_inode_nlink(inode_item
, 1);
502 btrfs_set_stack_inode_nbytes(inode_item
, root
->leafsize
);
503 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
505 btrfs_set_root_flags(&root_item
, 0);
506 btrfs_set_root_limit(&root_item
, 0);
507 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
509 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
510 btrfs_set_root_generation(&root_item
, trans
->transid
);
511 btrfs_set_root_level(&root_item
, 0);
512 btrfs_set_root_refs(&root_item
, 1);
513 btrfs_set_root_used(&root_item
, leaf
->len
);
514 btrfs_set_root_last_snapshot(&root_item
, 0);
516 btrfs_set_root_generation_v2(&root_item
,
517 btrfs_root_generation(&root_item
));
518 uuid_le_gen(&new_uuid
);
519 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
520 btrfs_set_stack_timespec_sec(&root_item
.otime
, cur_time
.tv_sec
);
521 btrfs_set_stack_timespec_nsec(&root_item
.otime
, cur_time
.tv_nsec
);
522 root_item
.ctime
= root_item
.otime
;
523 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
524 btrfs_set_root_otransid(&root_item
, trans
->transid
);
526 btrfs_tree_unlock(leaf
);
527 free_extent_buffer(leaf
);
530 btrfs_set_root_dirid(&root_item
, new_dirid
);
532 key
.objectid
= objectid
;
534 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
535 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
540 key
.offset
= (u64
)-1;
541 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
542 if (IS_ERR(new_root
)) {
543 btrfs_abort_transaction(trans
, root
, PTR_ERR(new_root
));
544 ret
= PTR_ERR(new_root
);
548 btrfs_record_root_in_trans(trans
, new_root
);
550 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
552 /* We potentially lose an unused inode item here */
553 btrfs_abort_transaction(trans
, root
, ret
);
558 * insert the directory item
560 ret
= btrfs_set_inode_index(dir
, &index
);
562 btrfs_abort_transaction(trans
, root
, ret
);
566 ret
= btrfs_insert_dir_item(trans
, root
,
567 name
, namelen
, dir
, &key
,
568 BTRFS_FT_DIR
, index
);
570 btrfs_abort_transaction(trans
, root
, ret
);
574 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
575 ret
= btrfs_update_inode(trans
, root
, dir
);
578 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
579 objectid
, root
->root_key
.objectid
,
580 btrfs_ino(dir
), index
, name
, namelen
);
583 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
584 root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
587 btrfs_abort_transaction(trans
, root
, ret
);
590 trans
->block_rsv
= NULL
;
591 trans
->bytes_reserved
= 0;
592 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
595 *async_transid
= trans
->transid
;
596 err
= btrfs_commit_transaction_async(trans
, root
, 1);
598 err
= btrfs_commit_transaction(trans
, root
);
600 err
= btrfs_commit_transaction(trans
, root
);
606 inode
= btrfs_lookup_dentry(dir
, dentry
);
608 return PTR_ERR(inode
);
609 d_instantiate(dentry
, inode
);
614 static void btrfs_wait_nocow_write(struct btrfs_root
*root
)
620 prepare_to_wait(&root
->subv_writers
->wait
, &wait
,
621 TASK_UNINTERRUPTIBLE
);
623 writers
= percpu_counter_sum(&root
->subv_writers
->counter
);
627 finish_wait(&root
->subv_writers
->wait
, &wait
);
631 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
632 struct dentry
*dentry
, char *name
, int namelen
,
633 u64
*async_transid
, bool readonly
,
634 struct btrfs_qgroup_inherit
*inherit
)
637 struct btrfs_pending_snapshot
*pending_snapshot
;
638 struct btrfs_trans_handle
*trans
;
644 atomic_inc(&root
->will_be_snapshoted
);
645 smp_mb__after_atomic_inc();
646 btrfs_wait_nocow_write(root
);
648 ret
= btrfs_start_delalloc_inodes(root
, 0);
652 btrfs_wait_ordered_extents(root
, -1);
654 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
655 if (!pending_snapshot
) {
660 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
661 BTRFS_BLOCK_RSV_TEMP
);
663 * 1 - parent dir inode
666 * 2 - root ref/backref
667 * 1 - root of snapshot
670 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
671 &pending_snapshot
->block_rsv
, 8,
672 &pending_snapshot
->qgroup_reserved
,
677 pending_snapshot
->dentry
= dentry
;
678 pending_snapshot
->root
= root
;
679 pending_snapshot
->readonly
= readonly
;
680 pending_snapshot
->dir
= dir
;
681 pending_snapshot
->inherit
= inherit
;
683 trans
= btrfs_start_transaction(root
, 0);
685 ret
= PTR_ERR(trans
);
689 spin_lock(&root
->fs_info
->trans_lock
);
690 list_add(&pending_snapshot
->list
,
691 &trans
->transaction
->pending_snapshots
);
692 spin_unlock(&root
->fs_info
->trans_lock
);
694 *async_transid
= trans
->transid
;
695 ret
= btrfs_commit_transaction_async(trans
,
696 root
->fs_info
->extent_root
, 1);
698 ret
= btrfs_commit_transaction(trans
, root
);
700 ret
= btrfs_commit_transaction(trans
,
701 root
->fs_info
->extent_root
);
706 ret
= pending_snapshot
->error
;
710 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
714 inode
= btrfs_lookup_dentry(dentry
->d_parent
->d_inode
, dentry
);
716 ret
= PTR_ERR(inode
);
720 d_instantiate(dentry
, inode
);
723 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
724 &pending_snapshot
->block_rsv
,
725 pending_snapshot
->qgroup_reserved
);
727 kfree(pending_snapshot
);
729 atomic_dec(&root
->will_be_snapshoted
);
733 /* copy of check_sticky in fs/namei.c()
734 * It's inline, so penalty for filesystems that don't use sticky bit is
737 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
739 kuid_t fsuid
= current_fsuid();
741 if (!(dir
->i_mode
& S_ISVTX
))
743 if (uid_eq(inode
->i_uid
, fsuid
))
745 if (uid_eq(dir
->i_uid
, fsuid
))
747 return !capable(CAP_FOWNER
);
750 /* copy of may_delete in fs/namei.c()
751 * Check whether we can remove a link victim from directory dir, check
752 * whether the type of victim is right.
753 * 1. We can't do it if dir is read-only (done in permission())
754 * 2. We should have write and exec permissions on dir
755 * 3. We can't remove anything from append-only dir
756 * 4. We can't do anything with immutable dir (done in permission())
757 * 5. If the sticky bit on dir is set we should either
758 * a. be owner of dir, or
759 * b. be owner of victim, or
760 * c. have CAP_FOWNER capability
761 * 6. If the victim is append-only or immutable we can't do antyhing with
762 * links pointing to it.
763 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
764 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
765 * 9. We can't remove a root or mountpoint.
766 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
767 * nfs_async_unlink().
770 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
774 if (!victim
->d_inode
)
777 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
778 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
780 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
785 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
786 IS_APPEND(victim
->d_inode
)||
787 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
790 if (!S_ISDIR(victim
->d_inode
->i_mode
))
794 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
798 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
803 /* copy of may_create in fs/namei.c() */
804 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
810 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
814 * Create a new subvolume below @parent. This is largely modeled after
815 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816 * inside this filesystem so it's quite a bit simpler.
818 static noinline
int btrfs_mksubvol(struct path
*parent
,
819 char *name
, int namelen
,
820 struct btrfs_root
*snap_src
,
821 u64
*async_transid
, bool readonly
,
822 struct btrfs_qgroup_inherit
*inherit
)
824 struct inode
*dir
= parent
->dentry
->d_inode
;
825 struct dentry
*dentry
;
828 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
832 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
833 error
= PTR_ERR(dentry
);
841 error
= btrfs_may_create(dir
, dentry
);
846 * even if this name doesn't exist, we may get hash collisions.
847 * check for them now when we can safely fail
849 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
855 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
857 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
861 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
862 async_transid
, readonly
, inherit
);
864 error
= create_subvol(dir
, dentry
, name
, namelen
,
865 async_transid
, inherit
);
868 fsnotify_mkdir(dir
, dentry
);
870 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
874 mutex_unlock(&dir
->i_mutex
);
879 * When we're defragging a range, we don't want to kick it off again
880 * if it is really just waiting for delalloc to send it down.
881 * If we find a nice big extent or delalloc range for the bytes in the
882 * file you want to defrag, we return 0 to let you know to skip this
885 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, int thresh
)
887 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
888 struct extent_map
*em
= NULL
;
889 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
892 read_lock(&em_tree
->lock
);
893 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
894 read_unlock(&em_tree
->lock
);
897 end
= extent_map_end(em
);
899 if (end
- offset
> thresh
)
902 /* if we already have a nice delalloc here, just stop */
904 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
905 thresh
, EXTENT_DELALLOC
, 1);
912 * helper function to walk through a file and find extents
913 * newer than a specific transid, and smaller than thresh.
915 * This is used by the defragging code to find new and small
918 static int find_new_extents(struct btrfs_root
*root
,
919 struct inode
*inode
, u64 newer_than
,
920 u64
*off
, int thresh
)
922 struct btrfs_path
*path
;
923 struct btrfs_key min_key
;
924 struct extent_buffer
*leaf
;
925 struct btrfs_file_extent_item
*extent
;
928 u64 ino
= btrfs_ino(inode
);
930 path
= btrfs_alloc_path();
934 min_key
.objectid
= ino
;
935 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
936 min_key
.offset
= *off
;
939 path
->keep_locks
= 1;
940 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
943 path
->keep_locks
= 0;
944 btrfs_unlock_up_safe(path
, 1);
946 if (min_key
.objectid
!= ino
)
948 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
951 leaf
= path
->nodes
[0];
952 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
953 struct btrfs_file_extent_item
);
955 type
= btrfs_file_extent_type(leaf
, extent
);
956 if (type
== BTRFS_FILE_EXTENT_REG
&&
957 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
958 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
959 *off
= min_key
.offset
;
960 btrfs_free_path(path
);
965 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
966 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
970 if (min_key
.offset
== (u64
)-1)
974 btrfs_release_path(path
);
977 btrfs_free_path(path
);
981 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
983 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
984 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
985 struct extent_map
*em
;
986 u64 len
= PAGE_CACHE_SIZE
;
989 * hopefully we have this extent in the tree already, try without
990 * the full extent lock
992 read_lock(&em_tree
->lock
);
993 em
= lookup_extent_mapping(em_tree
, start
, len
);
994 read_unlock(&em_tree
->lock
);
997 struct extent_state
*cached
= NULL
;
998 u64 end
= start
+ len
- 1;
1000 /* get the big lock and read metadata off disk */
1001 lock_extent_bits(io_tree
, start
, end
, 0, &cached
);
1002 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
1003 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1012 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1014 struct extent_map
*next
;
1017 /* this is the last extent */
1018 if (em
->start
+ em
->len
>= i_size_read(inode
))
1021 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1022 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
||
1023 (em
->block_start
+ em
->block_len
== next
->block_start
))
1026 free_extent_map(next
);
1030 static int should_defrag_range(struct inode
*inode
, u64 start
, int thresh
,
1031 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1034 struct extent_map
*em
;
1036 bool next_mergeable
= true;
1039 * make sure that once we start defragging an extent, we keep on
1042 if (start
< *defrag_end
)
1047 em
= defrag_lookup_extent(inode
, start
);
1051 /* this will cover holes, and inline extents */
1052 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1057 next_mergeable
= defrag_check_next_extent(inode
, em
);
1060 * we hit a real extent, if it is big or the next extent is not a
1061 * real extent, don't bother defragging it
1063 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1064 (em
->len
>= thresh
|| !next_mergeable
))
1068 * last_len ends up being a counter of how many bytes we've defragged.
1069 * every time we choose not to defrag an extent, we reset *last_len
1070 * so that the next tiny extent will force a defrag.
1072 * The end result of this is that tiny extents before a single big
1073 * extent will force at least part of that big extent to be defragged.
1076 *defrag_end
= extent_map_end(em
);
1079 *skip
= extent_map_end(em
);
1083 free_extent_map(em
);
1088 * it doesn't do much good to defrag one or two pages
1089 * at a time. This pulls in a nice chunk of pages
1090 * to COW and defrag.
1092 * It also makes sure the delalloc code has enough
1093 * dirty data to avoid making new small extents as part
1096 * It's a good idea to start RA on this range
1097 * before calling this.
1099 static int cluster_pages_for_defrag(struct inode
*inode
,
1100 struct page
**pages
,
1101 unsigned long start_index
,
1102 unsigned long num_pages
)
1104 unsigned long file_end
;
1105 u64 isize
= i_size_read(inode
);
1112 struct btrfs_ordered_extent
*ordered
;
1113 struct extent_state
*cached_state
= NULL
;
1114 struct extent_io_tree
*tree
;
1115 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1117 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1118 if (!isize
|| start_index
> file_end
)
1121 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1123 ret
= btrfs_delalloc_reserve_space(inode
,
1124 page_cnt
<< PAGE_CACHE_SHIFT
);
1128 tree
= &BTRFS_I(inode
)->io_tree
;
1130 /* step one, lock all the pages */
1131 for (i
= 0; i
< page_cnt
; i
++) {
1134 page
= find_or_create_page(inode
->i_mapping
,
1135 start_index
+ i
, mask
);
1139 page_start
= page_offset(page
);
1140 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1142 lock_extent_bits(tree
, page_start
, page_end
,
1144 ordered
= btrfs_lookup_ordered_extent(inode
,
1146 unlock_extent_cached(tree
, page_start
, page_end
,
1147 &cached_state
, GFP_NOFS
);
1152 btrfs_start_ordered_extent(inode
, ordered
, 1);
1153 btrfs_put_ordered_extent(ordered
);
1156 * we unlocked the page above, so we need check if
1157 * it was released or not.
1159 if (page
->mapping
!= inode
->i_mapping
) {
1161 page_cache_release(page
);
1166 if (!PageUptodate(page
)) {
1167 btrfs_readpage(NULL
, page
);
1169 if (!PageUptodate(page
)) {
1171 page_cache_release(page
);
1177 if (page
->mapping
!= inode
->i_mapping
) {
1179 page_cache_release(page
);
1189 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1193 * so now we have a nice long stream of locked
1194 * and up to date pages, lets wait on them
1196 for (i
= 0; i
< i_done
; i
++)
1197 wait_on_page_writeback(pages
[i
]);
1199 page_start
= page_offset(pages
[0]);
1200 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1202 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1203 page_start
, page_end
- 1, 0, &cached_state
);
1204 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1205 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1206 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1207 &cached_state
, GFP_NOFS
);
1209 if (i_done
!= page_cnt
) {
1210 spin_lock(&BTRFS_I(inode
)->lock
);
1211 BTRFS_I(inode
)->outstanding_extents
++;
1212 spin_unlock(&BTRFS_I(inode
)->lock
);
1213 btrfs_delalloc_release_space(inode
,
1214 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1218 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1219 &cached_state
, GFP_NOFS
);
1221 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1222 page_start
, page_end
- 1, &cached_state
,
1225 for (i
= 0; i
< i_done
; i
++) {
1226 clear_page_dirty_for_io(pages
[i
]);
1227 ClearPageChecked(pages
[i
]);
1228 set_page_extent_mapped(pages
[i
]);
1229 set_page_dirty(pages
[i
]);
1230 unlock_page(pages
[i
]);
1231 page_cache_release(pages
[i
]);
1235 for (i
= 0; i
< i_done
; i
++) {
1236 unlock_page(pages
[i
]);
1237 page_cache_release(pages
[i
]);
1239 btrfs_delalloc_release_space(inode
, page_cnt
<< PAGE_CACHE_SHIFT
);
1244 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1245 struct btrfs_ioctl_defrag_range_args
*range
,
1246 u64 newer_than
, unsigned long max_to_defrag
)
1248 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1249 struct file_ra_state
*ra
= NULL
;
1250 unsigned long last_index
;
1251 u64 isize
= i_size_read(inode
);
1255 u64 newer_off
= range
->start
;
1257 unsigned long ra_index
= 0;
1259 int defrag_count
= 0;
1260 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1261 int extent_thresh
= range
->extent_thresh
;
1262 unsigned long max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
1263 unsigned long cluster
= max_cluster
;
1264 u64 new_align
= ~((u64
)128 * 1024 - 1);
1265 struct page
**pages
= NULL
;
1270 if (range
->start
>= isize
)
1273 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1274 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1276 if (range
->compress_type
)
1277 compress_type
= range
->compress_type
;
1280 if (extent_thresh
== 0)
1281 extent_thresh
= 256 * 1024;
1284 * if we were not given a file, allocate a readahead
1288 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1291 file_ra_state_init(ra
, inode
->i_mapping
);
1296 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1303 /* find the last page to defrag */
1304 if (range
->start
+ range
->len
> range
->start
) {
1305 last_index
= min_t(u64
, isize
- 1,
1306 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1308 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1312 ret
= find_new_extents(root
, inode
, newer_than
,
1313 &newer_off
, 64 * 1024);
1315 range
->start
= newer_off
;
1317 * we always align our defrag to help keep
1318 * the extents in the file evenly spaced
1320 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1324 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1327 max_to_defrag
= last_index
+ 1;
1330 * make writeback starts from i, so the defrag range can be
1331 * written sequentially.
1333 if (i
< inode
->i_mapping
->writeback_index
)
1334 inode
->i_mapping
->writeback_index
= i
;
1336 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1337 (i
< (i_size_read(inode
) + PAGE_CACHE_SIZE
- 1) >>
1338 PAGE_CACHE_SHIFT
)) {
1340 * make sure we stop running if someone unmounts
1343 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1346 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1347 printk(KERN_DEBUG
"BTRFS: defrag_file cancelled\n");
1352 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1353 extent_thresh
, &last_len
, &skip
,
1354 &defrag_end
, range
->flags
&
1355 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1358 * the should_defrag function tells us how much to skip
1359 * bump our counter by the suggested amount
1361 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
1362 i
= max(i
+ 1, next
);
1367 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1368 PAGE_CACHE_SHIFT
) - i
;
1369 cluster
= min(cluster
, max_cluster
);
1371 cluster
= max_cluster
;
1374 if (i
+ cluster
> ra_index
) {
1375 ra_index
= max(i
, ra_index
);
1376 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1378 ra_index
+= max_cluster
;
1381 mutex_lock(&inode
->i_mutex
);
1382 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1383 BTRFS_I(inode
)->force_compress
= compress_type
;
1384 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1386 mutex_unlock(&inode
->i_mutex
);
1390 defrag_count
+= ret
;
1391 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1392 mutex_unlock(&inode
->i_mutex
);
1395 if (newer_off
== (u64
)-1)
1401 newer_off
= max(newer_off
+ 1,
1402 (u64
)i
<< PAGE_CACHE_SHIFT
);
1404 ret
= find_new_extents(root
, inode
,
1405 newer_than
, &newer_off
,
1408 range
->start
= newer_off
;
1409 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1416 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1424 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1425 filemap_flush(inode
->i_mapping
);
1426 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1427 &BTRFS_I(inode
)->runtime_flags
))
1428 filemap_flush(inode
->i_mapping
);
1431 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1432 /* the filemap_flush will queue IO into the worker threads, but
1433 * we have to make sure the IO is actually started and that
1434 * ordered extents get created before we return
1436 atomic_inc(&root
->fs_info
->async_submit_draining
);
1437 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1438 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1439 wait_event(root
->fs_info
->async_submit_wait
,
1440 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1441 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1443 atomic_dec(&root
->fs_info
->async_submit_draining
);
1446 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1447 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1453 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1454 mutex_lock(&inode
->i_mutex
);
1455 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1456 mutex_unlock(&inode
->i_mutex
);
1464 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1470 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1471 struct btrfs_ioctl_vol_args
*vol_args
;
1472 struct btrfs_trans_handle
*trans
;
1473 struct btrfs_device
*device
= NULL
;
1476 char *devstr
= NULL
;
1480 if (!capable(CAP_SYS_ADMIN
))
1483 ret
= mnt_want_write_file(file
);
1487 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1489 mnt_drop_write_file(file
);
1490 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1493 mutex_lock(&root
->fs_info
->volume_mutex
);
1494 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1495 if (IS_ERR(vol_args
)) {
1496 ret
= PTR_ERR(vol_args
);
1500 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1502 sizestr
= vol_args
->name
;
1503 devstr
= strchr(sizestr
, ':');
1506 sizestr
= devstr
+ 1;
1508 devstr
= vol_args
->name
;
1509 devid
= simple_strtoull(devstr
, &end
, 10);
1514 btrfs_info(root
->fs_info
, "resizing devid %llu", devid
);
1517 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1519 btrfs_info(root
->fs_info
, "resizer unable to find device %llu",
1525 if (!device
->writeable
) {
1526 btrfs_info(root
->fs_info
,
1527 "resizer unable to apply on readonly device %llu",
1533 if (!strcmp(sizestr
, "max"))
1534 new_size
= device
->bdev
->bd_inode
->i_size
;
1536 if (sizestr
[0] == '-') {
1539 } else if (sizestr
[0] == '+') {
1543 new_size
= memparse(sizestr
, &retptr
);
1544 if (*retptr
!= '\0' || new_size
== 0) {
1550 if (device
->is_tgtdev_for_dev_replace
) {
1555 old_size
= device
->total_bytes
;
1558 if (new_size
> old_size
) {
1562 new_size
= old_size
- new_size
;
1563 } else if (mod
> 0) {
1564 if (new_size
> ULLONG_MAX
- old_size
) {
1568 new_size
= old_size
+ new_size
;
1571 if (new_size
< 256 * 1024 * 1024) {
1575 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1580 do_div(new_size
, root
->sectorsize
);
1581 new_size
*= root
->sectorsize
;
1583 printk_in_rcu(KERN_INFO
"BTRFS: new size for %s is %llu\n",
1584 rcu_str_deref(device
->name
), new_size
);
1586 if (new_size
> old_size
) {
1587 trans
= btrfs_start_transaction(root
, 0);
1588 if (IS_ERR(trans
)) {
1589 ret
= PTR_ERR(trans
);
1592 ret
= btrfs_grow_device(trans
, device
, new_size
);
1593 btrfs_commit_transaction(trans
, root
);
1594 } else if (new_size
< old_size
) {
1595 ret
= btrfs_shrink_device(device
, new_size
);
1596 } /* equal, nothing need to do */
1601 mutex_unlock(&root
->fs_info
->volume_mutex
);
1602 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1603 mnt_drop_write_file(file
);
1607 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1608 char *name
, unsigned long fd
, int subvol
,
1609 u64
*transid
, bool readonly
,
1610 struct btrfs_qgroup_inherit
*inherit
)
1615 ret
= mnt_want_write_file(file
);
1619 namelen
= strlen(name
);
1620 if (strchr(name
, '/')) {
1622 goto out_drop_write
;
1625 if (name
[0] == '.' &&
1626 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1628 goto out_drop_write
;
1632 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1633 NULL
, transid
, readonly
, inherit
);
1635 struct fd src
= fdget(fd
);
1636 struct inode
*src_inode
;
1639 goto out_drop_write
;
1642 src_inode
= file_inode(src
.file
);
1643 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1644 btrfs_info(BTRFS_I(src_inode
)->root
->fs_info
,
1645 "Snapshot src from another FS");
1647 } else if (!inode_owner_or_capable(src_inode
)) {
1649 * Subvolume creation is not restricted, but snapshots
1650 * are limited to own subvolumes only
1654 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1655 BTRFS_I(src_inode
)->root
,
1656 transid
, readonly
, inherit
);
1661 mnt_drop_write_file(file
);
1666 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1667 void __user
*arg
, int subvol
)
1669 struct btrfs_ioctl_vol_args
*vol_args
;
1672 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1673 if (IS_ERR(vol_args
))
1674 return PTR_ERR(vol_args
);
1675 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1677 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1678 vol_args
->fd
, subvol
,
1685 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1686 void __user
*arg
, int subvol
)
1688 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1692 bool readonly
= false;
1693 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1695 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1696 if (IS_ERR(vol_args
))
1697 return PTR_ERR(vol_args
);
1698 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1700 if (vol_args
->flags
&
1701 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1702 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1707 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1709 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1711 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1712 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1716 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1717 if (IS_ERR(inherit
)) {
1718 ret
= PTR_ERR(inherit
);
1723 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1724 vol_args
->fd
, subvol
, ptr
,
1727 if (ret
== 0 && ptr
&&
1729 offsetof(struct btrfs_ioctl_vol_args_v2
,
1730 transid
), ptr
, sizeof(*ptr
)))
1738 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1741 struct inode
*inode
= file_inode(file
);
1742 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1746 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1749 down_read(&root
->fs_info
->subvol_sem
);
1750 if (btrfs_root_readonly(root
))
1751 flags
|= BTRFS_SUBVOL_RDONLY
;
1752 up_read(&root
->fs_info
->subvol_sem
);
1754 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1760 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1763 struct inode
*inode
= file_inode(file
);
1764 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1765 struct btrfs_trans_handle
*trans
;
1770 if (!inode_owner_or_capable(inode
))
1773 ret
= mnt_want_write_file(file
);
1777 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1779 goto out_drop_write
;
1782 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1784 goto out_drop_write
;
1787 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1789 goto out_drop_write
;
1792 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1794 goto out_drop_write
;
1797 down_write(&root
->fs_info
->subvol_sem
);
1800 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1803 root_flags
= btrfs_root_flags(&root
->root_item
);
1804 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1805 btrfs_set_root_flags(&root
->root_item
,
1806 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1809 * Block RO -> RW transition if this subvolume is involved in
1812 spin_lock(&root
->root_item_lock
);
1813 if (root
->send_in_progress
== 0) {
1814 btrfs_set_root_flags(&root
->root_item
,
1815 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1816 spin_unlock(&root
->root_item_lock
);
1818 spin_unlock(&root
->root_item_lock
);
1819 btrfs_warn(root
->fs_info
,
1820 "Attempt to set subvolume %llu read-write during send",
1821 root
->root_key
.objectid
);
1827 trans
= btrfs_start_transaction(root
, 1);
1828 if (IS_ERR(trans
)) {
1829 ret
= PTR_ERR(trans
);
1833 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1834 &root
->root_key
, &root
->root_item
);
1836 btrfs_commit_transaction(trans
, root
);
1839 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1841 up_write(&root
->fs_info
->subvol_sem
);
1843 mnt_drop_write_file(file
);
1849 * helper to check if the subvolume references other subvolumes
1851 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1853 struct btrfs_path
*path
;
1854 struct btrfs_dir_item
*di
;
1855 struct btrfs_key key
;
1859 path
= btrfs_alloc_path();
1863 /* Make sure this root isn't set as the default subvol */
1864 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
1865 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
, path
,
1866 dir_id
, "default", 7, 0);
1867 if (di
&& !IS_ERR(di
)) {
1868 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1869 if (key
.objectid
== root
->root_key
.objectid
) {
1871 btrfs_err(root
->fs_info
, "deleting default subvolume "
1872 "%llu is not allowed", key
.objectid
);
1875 btrfs_release_path(path
);
1878 key
.objectid
= root
->root_key
.objectid
;
1879 key
.type
= BTRFS_ROOT_REF_KEY
;
1880 key
.offset
= (u64
)-1;
1882 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1889 if (path
->slots
[0] > 0) {
1891 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1892 if (key
.objectid
== root
->root_key
.objectid
&&
1893 key
.type
== BTRFS_ROOT_REF_KEY
)
1897 btrfs_free_path(path
);
1901 static noinline
int key_in_sk(struct btrfs_key
*key
,
1902 struct btrfs_ioctl_search_key
*sk
)
1904 struct btrfs_key test
;
1907 test
.objectid
= sk
->min_objectid
;
1908 test
.type
= sk
->min_type
;
1909 test
.offset
= sk
->min_offset
;
1911 ret
= btrfs_comp_cpu_keys(key
, &test
);
1915 test
.objectid
= sk
->max_objectid
;
1916 test
.type
= sk
->max_type
;
1917 test
.offset
= sk
->max_offset
;
1919 ret
= btrfs_comp_cpu_keys(key
, &test
);
1925 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1926 struct btrfs_path
*path
,
1927 struct btrfs_key
*key
,
1928 struct btrfs_ioctl_search_key
*sk
,
1930 unsigned long *sk_offset
,
1934 struct extent_buffer
*leaf
;
1935 struct btrfs_ioctl_search_header sh
;
1936 unsigned long item_off
;
1937 unsigned long item_len
;
1943 leaf
= path
->nodes
[0];
1944 slot
= path
->slots
[0];
1945 nritems
= btrfs_header_nritems(leaf
);
1947 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1951 found_transid
= btrfs_header_generation(leaf
);
1953 for (i
= slot
; i
< nritems
; i
++) {
1954 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1955 item_len
= btrfs_item_size_nr(leaf
, i
);
1957 btrfs_item_key_to_cpu(leaf
, key
, i
);
1958 if (!key_in_sk(key
, sk
))
1961 if (sizeof(sh
) + item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1964 if (sizeof(sh
) + item_len
+ *sk_offset
>
1965 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1970 sh
.objectid
= key
->objectid
;
1971 sh
.offset
= key
->offset
;
1972 sh
.type
= key
->type
;
1974 sh
.transid
= found_transid
;
1976 /* copy search result header */
1977 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1978 *sk_offset
+= sizeof(sh
);
1981 char *p
= buf
+ *sk_offset
;
1983 read_extent_buffer(leaf
, p
,
1984 item_off
, item_len
);
1985 *sk_offset
+= item_len
;
1989 if (*num_found
>= sk
->nr_items
)
1994 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1996 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1999 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
2009 static noinline
int search_ioctl(struct inode
*inode
,
2010 struct btrfs_ioctl_search_args
*args
)
2012 struct btrfs_root
*root
;
2013 struct btrfs_key key
;
2014 struct btrfs_path
*path
;
2015 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
2016 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
2019 unsigned long sk_offset
= 0;
2021 path
= btrfs_alloc_path();
2025 if (sk
->tree_id
== 0) {
2026 /* search the root of the inode that was passed */
2027 root
= BTRFS_I(inode
)->root
;
2029 key
.objectid
= sk
->tree_id
;
2030 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2031 key
.offset
= (u64
)-1;
2032 root
= btrfs_read_fs_root_no_name(info
, &key
);
2034 printk(KERN_ERR
"BTRFS: could not find root %llu\n",
2036 btrfs_free_path(path
);
2041 key
.objectid
= sk
->min_objectid
;
2042 key
.type
= sk
->min_type
;
2043 key
.offset
= sk
->min_offset
;
2045 path
->keep_locks
= 1;
2048 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2054 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
2055 &sk_offset
, &num_found
);
2056 btrfs_release_path(path
);
2057 if (ret
|| num_found
>= sk
->nr_items
)
2063 sk
->nr_items
= num_found
;
2064 btrfs_free_path(path
);
2068 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2071 struct btrfs_ioctl_search_args
*args
;
2072 struct inode
*inode
;
2075 if (!capable(CAP_SYS_ADMIN
))
2078 args
= memdup_user(argp
, sizeof(*args
));
2080 return PTR_ERR(args
);
2082 inode
= file_inode(file
);
2083 ret
= search_ioctl(inode
, args
);
2084 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2091 * Search INODE_REFs to identify path name of 'dirid' directory
2092 * in a 'tree_id' tree. and sets path name to 'name'.
2094 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2095 u64 tree_id
, u64 dirid
, char *name
)
2097 struct btrfs_root
*root
;
2098 struct btrfs_key key
;
2104 struct btrfs_inode_ref
*iref
;
2105 struct extent_buffer
*l
;
2106 struct btrfs_path
*path
;
2108 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2113 path
= btrfs_alloc_path();
2117 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2119 key
.objectid
= tree_id
;
2120 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2121 key
.offset
= (u64
)-1;
2122 root
= btrfs_read_fs_root_no_name(info
, &key
);
2124 printk(KERN_ERR
"BTRFS: could not find root %llu\n", tree_id
);
2129 key
.objectid
= dirid
;
2130 key
.type
= BTRFS_INODE_REF_KEY
;
2131 key
.offset
= (u64
)-1;
2134 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2138 ret
= btrfs_previous_item(root
, path
, dirid
,
2139 BTRFS_INODE_REF_KEY
);
2149 slot
= path
->slots
[0];
2150 btrfs_item_key_to_cpu(l
, &key
, slot
);
2152 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2153 len
= btrfs_inode_ref_name_len(l
, iref
);
2155 total_len
+= len
+ 1;
2157 ret
= -ENAMETOOLONG
;
2162 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2164 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2167 btrfs_release_path(path
);
2168 key
.objectid
= key
.offset
;
2169 key
.offset
= (u64
)-1;
2170 dirid
= key
.objectid
;
2172 memmove(name
, ptr
, total_len
);
2173 name
[total_len
] = '\0';
2176 btrfs_free_path(path
);
2180 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2183 struct btrfs_ioctl_ino_lookup_args
*args
;
2184 struct inode
*inode
;
2187 if (!capable(CAP_SYS_ADMIN
))
2190 args
= memdup_user(argp
, sizeof(*args
));
2192 return PTR_ERR(args
);
2194 inode
= file_inode(file
);
2196 if (args
->treeid
== 0)
2197 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2199 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2200 args
->treeid
, args
->objectid
,
2203 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2210 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2213 struct dentry
*parent
= file
->f_path
.dentry
;
2214 struct dentry
*dentry
;
2215 struct inode
*dir
= parent
->d_inode
;
2216 struct inode
*inode
;
2217 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2218 struct btrfs_root
*dest
= NULL
;
2219 struct btrfs_ioctl_vol_args
*vol_args
;
2220 struct btrfs_trans_handle
*trans
;
2221 struct btrfs_block_rsv block_rsv
;
2222 u64 qgroup_reserved
;
2227 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2228 if (IS_ERR(vol_args
))
2229 return PTR_ERR(vol_args
);
2231 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2232 namelen
= strlen(vol_args
->name
);
2233 if (strchr(vol_args
->name
, '/') ||
2234 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2239 err
= mnt_want_write_file(file
);
2243 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2245 goto out_drop_write
;
2246 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2247 if (IS_ERR(dentry
)) {
2248 err
= PTR_ERR(dentry
);
2249 goto out_unlock_dir
;
2252 if (!dentry
->d_inode
) {
2257 inode
= dentry
->d_inode
;
2258 dest
= BTRFS_I(inode
)->root
;
2259 if (!capable(CAP_SYS_ADMIN
)) {
2261 * Regular user. Only allow this with a special mount
2262 * option, when the user has write+exec access to the
2263 * subvol root, and when rmdir(2) would have been
2266 * Note that this is _not_ check that the subvol is
2267 * empty or doesn't contain data that we wouldn't
2268 * otherwise be able to delete.
2270 * Users who want to delete empty subvols should try
2274 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2278 * Do not allow deletion if the parent dir is the same
2279 * as the dir to be deleted. That means the ioctl
2280 * must be called on the dentry referencing the root
2281 * of the subvol, not a random directory contained
2288 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2293 /* check if subvolume may be deleted by a user */
2294 err
= btrfs_may_delete(dir
, dentry
, 1);
2298 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2303 mutex_lock(&inode
->i_mutex
);
2304 err
= d_invalidate(dentry
);
2308 down_write(&root
->fs_info
->subvol_sem
);
2310 err
= may_destroy_subvol(dest
);
2314 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2316 * One for dir inode, two for dir entries, two for root
2319 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2320 5, &qgroup_reserved
, true);
2324 trans
= btrfs_start_transaction(root
, 0);
2325 if (IS_ERR(trans
)) {
2326 err
= PTR_ERR(trans
);
2329 trans
->block_rsv
= &block_rsv
;
2330 trans
->bytes_reserved
= block_rsv
.size
;
2332 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2333 dest
->root_key
.objectid
,
2334 dentry
->d_name
.name
,
2335 dentry
->d_name
.len
);
2338 btrfs_abort_transaction(trans
, root
, ret
);
2342 btrfs_record_root_in_trans(trans
, dest
);
2344 memset(&dest
->root_item
.drop_progress
, 0,
2345 sizeof(dest
->root_item
.drop_progress
));
2346 dest
->root_item
.drop_level
= 0;
2347 btrfs_set_root_refs(&dest
->root_item
, 0);
2349 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
2350 ret
= btrfs_insert_orphan_item(trans
,
2351 root
->fs_info
->tree_root
,
2352 dest
->root_key
.objectid
);
2354 btrfs_abort_transaction(trans
, root
, ret
);
2360 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2361 dest
->root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
2362 dest
->root_key
.objectid
);
2363 if (ret
&& ret
!= -ENOENT
) {
2364 btrfs_abort_transaction(trans
, root
, ret
);
2368 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2369 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2370 dest
->root_item
.received_uuid
,
2371 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2372 dest
->root_key
.objectid
);
2373 if (ret
&& ret
!= -ENOENT
) {
2374 btrfs_abort_transaction(trans
, root
, ret
);
2381 trans
->block_rsv
= NULL
;
2382 trans
->bytes_reserved
= 0;
2383 ret
= btrfs_end_transaction(trans
, root
);
2386 inode
->i_flags
|= S_DEAD
;
2388 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2390 up_write(&root
->fs_info
->subvol_sem
);
2392 mutex_unlock(&inode
->i_mutex
);
2394 shrink_dcache_sb(root
->fs_info
->sb
);
2395 btrfs_invalidate_inodes(dest
);
2399 if (dest
->cache_inode
) {
2400 iput(dest
->cache_inode
);
2401 dest
->cache_inode
= NULL
;
2407 mutex_unlock(&dir
->i_mutex
);
2409 mnt_drop_write_file(file
);
2415 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2417 struct inode
*inode
= file_inode(file
);
2418 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2419 struct btrfs_ioctl_defrag_range_args
*range
;
2422 ret
= mnt_want_write_file(file
);
2426 if (btrfs_root_readonly(root
)) {
2431 switch (inode
->i_mode
& S_IFMT
) {
2433 if (!capable(CAP_SYS_ADMIN
)) {
2437 ret
= btrfs_defrag_root(root
);
2440 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2443 if (!(file
->f_mode
& FMODE_WRITE
)) {
2448 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2455 if (copy_from_user(range
, argp
,
2461 /* compression requires us to start the IO */
2462 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2463 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2464 range
->extent_thresh
= (u32
)-1;
2467 /* the rest are all set to zero by kzalloc */
2468 range
->len
= (u64
)-1;
2470 ret
= btrfs_defrag_file(file_inode(file
), file
,
2480 mnt_drop_write_file(file
);
2484 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2486 struct btrfs_ioctl_vol_args
*vol_args
;
2489 if (!capable(CAP_SYS_ADMIN
))
2492 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2494 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2497 mutex_lock(&root
->fs_info
->volume_mutex
);
2498 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2499 if (IS_ERR(vol_args
)) {
2500 ret
= PTR_ERR(vol_args
);
2504 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2505 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2509 mutex_unlock(&root
->fs_info
->volume_mutex
);
2510 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2514 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2516 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2517 struct btrfs_ioctl_vol_args
*vol_args
;
2520 if (!capable(CAP_SYS_ADMIN
))
2523 ret
= mnt_want_write_file(file
);
2527 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2528 if (IS_ERR(vol_args
)) {
2529 ret
= PTR_ERR(vol_args
);
2533 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2535 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2537 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2541 mutex_lock(&root
->fs_info
->volume_mutex
);
2542 ret
= btrfs_rm_device(root
, vol_args
->name
);
2543 mutex_unlock(&root
->fs_info
->volume_mutex
);
2544 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2548 mnt_drop_write_file(file
);
2552 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2554 struct btrfs_ioctl_fs_info_args
*fi_args
;
2555 struct btrfs_device
*device
;
2556 struct btrfs_device
*next
;
2557 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2560 if (!capable(CAP_SYS_ADMIN
))
2563 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2567 mutex_lock(&fs_devices
->device_list_mutex
);
2568 fi_args
->num_devices
= fs_devices
->num_devices
;
2569 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2571 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2572 if (device
->devid
> fi_args
->max_id
)
2573 fi_args
->max_id
= device
->devid
;
2575 mutex_unlock(&fs_devices
->device_list_mutex
);
2577 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2584 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2586 struct btrfs_ioctl_dev_info_args
*di_args
;
2587 struct btrfs_device
*dev
;
2588 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2590 char *s_uuid
= NULL
;
2592 if (!capable(CAP_SYS_ADMIN
))
2595 di_args
= memdup_user(arg
, sizeof(*di_args
));
2596 if (IS_ERR(di_args
))
2597 return PTR_ERR(di_args
);
2599 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2600 s_uuid
= di_args
->uuid
;
2602 mutex_lock(&fs_devices
->device_list_mutex
);
2603 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2610 di_args
->devid
= dev
->devid
;
2611 di_args
->bytes_used
= dev
->bytes_used
;
2612 di_args
->total_bytes
= dev
->total_bytes
;
2613 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2615 struct rcu_string
*name
;
2618 name
= rcu_dereference(dev
->name
);
2619 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2621 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2623 di_args
->path
[0] = '\0';
2627 mutex_unlock(&fs_devices
->device_list_mutex
);
2628 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2635 static struct page
*extent_same_get_page(struct inode
*inode
, u64 off
)
2639 struct extent_io_tree
*tree
= &BTRFS_I(inode
)->io_tree
;
2641 index
= off
>> PAGE_CACHE_SHIFT
;
2643 page
= grab_cache_page(inode
->i_mapping
, index
);
2647 if (!PageUptodate(page
)) {
2648 if (extent_read_full_page_nolock(tree
, page
, btrfs_get_extent
,
2652 if (!PageUptodate(page
)) {
2654 page_cache_release(page
);
2663 static inline void lock_extent_range(struct inode
*inode
, u64 off
, u64 len
)
2665 /* do any pending delalloc/csum calc on src, one way or
2666 another, and lock file content */
2668 struct btrfs_ordered_extent
*ordered
;
2669 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2670 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2673 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2674 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
))
2676 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2678 btrfs_put_ordered_extent(ordered
);
2679 btrfs_wait_ordered_range(inode
, off
, len
);
2683 static void btrfs_double_unlock(struct inode
*inode1
, u64 loff1
,
2684 struct inode
*inode2
, u64 loff2
, u64 len
)
2686 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2687 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2689 mutex_unlock(&inode1
->i_mutex
);
2690 mutex_unlock(&inode2
->i_mutex
);
2693 static void btrfs_double_lock(struct inode
*inode1
, u64 loff1
,
2694 struct inode
*inode2
, u64 loff2
, u64 len
)
2696 if (inode1
< inode2
) {
2697 swap(inode1
, inode2
);
2701 mutex_lock_nested(&inode1
->i_mutex
, I_MUTEX_PARENT
);
2702 lock_extent_range(inode1
, loff1
, len
);
2703 if (inode1
!= inode2
) {
2704 mutex_lock_nested(&inode2
->i_mutex
, I_MUTEX_CHILD
);
2705 lock_extent_range(inode2
, loff2
, len
);
2709 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
2710 u64 dst_loff
, u64 len
)
2713 struct page
*src_page
, *dst_page
;
2714 unsigned int cmp_len
= PAGE_CACHE_SIZE
;
2715 void *addr
, *dst_addr
;
2718 if (len
< PAGE_CACHE_SIZE
)
2721 src_page
= extent_same_get_page(src
, loff
);
2724 dst_page
= extent_same_get_page(dst
, dst_loff
);
2726 page_cache_release(src_page
);
2729 addr
= kmap_atomic(src_page
);
2730 dst_addr
= kmap_atomic(dst_page
);
2732 flush_dcache_page(src_page
);
2733 flush_dcache_page(dst_page
);
2735 if (memcmp(addr
, dst_addr
, cmp_len
))
2736 ret
= BTRFS_SAME_DATA_DIFFERS
;
2738 kunmap_atomic(addr
);
2739 kunmap_atomic(dst_addr
);
2740 page_cache_release(src_page
);
2741 page_cache_release(dst_page
);
2747 dst_loff
+= cmp_len
;
2754 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64 len
)
2756 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
2758 if (off
+ len
> inode
->i_size
|| off
+ len
< off
)
2760 /* Check that we are block aligned - btrfs_clone() requires this */
2761 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
2767 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 len
,
2768 struct inode
*dst
, u64 dst_loff
)
2773 * btrfs_clone() can't handle extents in the same file
2774 * yet. Once that works, we can drop this check and replace it
2775 * with a check for the same inode, but overlapping extents.
2780 btrfs_double_lock(src
, loff
, dst
, dst_loff
, len
);
2782 ret
= extent_same_check_offsets(src
, loff
, len
);
2786 ret
= extent_same_check_offsets(dst
, dst_loff
, len
);
2790 /* don't make the dst file partly checksummed */
2791 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2792 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
2797 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, len
);
2799 ret
= btrfs_clone(src
, dst
, loff
, len
, len
, dst_loff
);
2802 btrfs_double_unlock(src
, loff
, dst
, dst_loff
, len
);
2807 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2809 static long btrfs_ioctl_file_extent_same(struct file
*file
,
2810 struct btrfs_ioctl_same_args __user
*argp
)
2812 struct btrfs_ioctl_same_args
*same
;
2813 struct btrfs_ioctl_same_extent_info
*info
;
2814 struct inode
*src
= file_inode(file
);
2820 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
2821 bool is_admin
= capable(CAP_SYS_ADMIN
);
2824 if (!(file
->f_mode
& FMODE_READ
))
2827 ret
= mnt_want_write_file(file
);
2831 if (get_user(count
, &argp
->dest_count
)) {
2836 size
= offsetof(struct btrfs_ioctl_same_args __user
, info
[count
]);
2838 same
= memdup_user(argp
, size
);
2841 ret
= PTR_ERR(same
);
2845 off
= same
->logical_offset
;
2849 * Limit the total length we will dedupe for each operation.
2850 * This is intended to bound the total time spent in this
2851 * ioctl to something sane.
2853 if (len
> BTRFS_MAX_DEDUPE_LEN
)
2854 len
= BTRFS_MAX_DEDUPE_LEN
;
2856 if (WARN_ON_ONCE(bs
< PAGE_CACHE_SIZE
)) {
2858 * Btrfs does not support blocksize < page_size. As a
2859 * result, btrfs_cmp_data() won't correctly handle
2860 * this situation without an update.
2867 if (S_ISDIR(src
->i_mode
))
2871 if (!S_ISREG(src
->i_mode
))
2874 /* pre-format output fields to sane values */
2875 for (i
= 0; i
< count
; i
++) {
2876 same
->info
[i
].bytes_deduped
= 0ULL;
2877 same
->info
[i
].status
= 0;
2880 for (i
= 0, info
= same
->info
; i
< count
; i
++, info
++) {
2882 struct fd dst_file
= fdget(info
->fd
);
2883 if (!dst_file
.file
) {
2884 info
->status
= -EBADF
;
2887 dst
= file_inode(dst_file
.file
);
2889 if (!(is_admin
|| (dst_file
.file
->f_mode
& FMODE_WRITE
))) {
2890 info
->status
= -EINVAL
;
2891 } else if (file
->f_path
.mnt
!= dst_file
.file
->f_path
.mnt
) {
2892 info
->status
= -EXDEV
;
2893 } else if (S_ISDIR(dst
->i_mode
)) {
2894 info
->status
= -EISDIR
;
2895 } else if (!S_ISREG(dst
->i_mode
)) {
2896 info
->status
= -EACCES
;
2898 info
->status
= btrfs_extent_same(src
, off
, len
, dst
,
2899 info
->logical_offset
);
2900 if (info
->status
== 0)
2901 info
->bytes_deduped
+= len
;
2906 ret
= copy_to_user(argp
, same
, size
);
2911 mnt_drop_write_file(file
);
2916 * btrfs_clone() - clone a range from inode file to another
2918 * @src: Inode to clone from
2919 * @inode: Inode to clone to
2920 * @off: Offset within source to start clone from
2921 * @olen: Original length, passed by user, of range to clone
2922 * @olen_aligned: Block-aligned value of olen, extent_same uses
2923 * identical values here
2924 * @destoff: Offset within @inode to start clone
2926 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
2927 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
)
2929 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2930 struct btrfs_path
*path
= NULL
;
2931 struct extent_buffer
*leaf
;
2932 struct btrfs_trans_handle
*trans
;
2934 struct btrfs_key key
;
2938 u64 len
= olen_aligned
;
2941 buf
= vmalloc(btrfs_level_size(root
, 0));
2945 path
= btrfs_alloc_path();
2953 key
.objectid
= btrfs_ino(src
);
2954 key
.type
= BTRFS_EXTENT_DATA_KEY
;
2959 * note the key will change type as we walk through the
2962 path
->leave_spinning
= 1;
2963 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
2968 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2970 if (path
->slots
[0] >= nritems
) {
2971 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
2976 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2978 leaf
= path
->nodes
[0];
2979 slot
= path
->slots
[0];
2981 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2982 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
2983 key
.objectid
!= btrfs_ino(src
))
2986 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
2987 struct btrfs_file_extent_item
*extent
;
2990 struct btrfs_key new_key
;
2991 u64 disko
= 0, diskl
= 0;
2992 u64 datao
= 0, datal
= 0;
2996 extent
= btrfs_item_ptr(leaf
, slot
,
2997 struct btrfs_file_extent_item
);
2998 comp
= btrfs_file_extent_compression(leaf
, extent
);
2999 type
= btrfs_file_extent_type(leaf
, extent
);
3000 if (type
== BTRFS_FILE_EXTENT_REG
||
3001 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3002 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3004 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3006 datao
= btrfs_file_extent_offset(leaf
, extent
);
3007 datal
= btrfs_file_extent_num_bytes(leaf
,
3009 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3010 /* take upper bound, may be compressed */
3011 datal
= btrfs_file_extent_ram_bytes(leaf
,
3015 if (key
.offset
+ datal
<= off
||
3016 key
.offset
>= off
+ len
- 1) {
3021 size
= btrfs_item_size_nr(leaf
, slot
);
3022 read_extent_buffer(leaf
, buf
,
3023 btrfs_item_ptr_offset(leaf
, slot
),
3026 btrfs_release_path(path
);
3027 path
->leave_spinning
= 0;
3029 memcpy(&new_key
, &key
, sizeof(new_key
));
3030 new_key
.objectid
= btrfs_ino(inode
);
3031 if (off
<= key
.offset
)
3032 new_key
.offset
= key
.offset
+ destoff
- off
;
3034 new_key
.offset
= destoff
;
3037 * 1 - adjusting old extent (we may have to split it)
3038 * 1 - add new extent
3041 trans
= btrfs_start_transaction(root
, 3);
3042 if (IS_ERR(trans
)) {
3043 ret
= PTR_ERR(trans
);
3047 if (type
== BTRFS_FILE_EXTENT_REG
||
3048 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3050 * a | --- range to clone ---| b
3051 * | ------------- extent ------------- |
3054 /* substract range b */
3055 if (key
.offset
+ datal
> off
+ len
)
3056 datal
= off
+ len
- key
.offset
;
3058 /* substract range a */
3059 if (off
> key
.offset
) {
3060 datao
+= off
- key
.offset
;
3061 datal
-= off
- key
.offset
;
3064 ret
= btrfs_drop_extents(trans
, root
, inode
,
3066 new_key
.offset
+ datal
,
3069 if (ret
!= -EOPNOTSUPP
)
3070 btrfs_abort_transaction(trans
,
3072 btrfs_end_transaction(trans
, root
);
3076 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3079 btrfs_abort_transaction(trans
, root
,
3081 btrfs_end_transaction(trans
, root
);
3085 leaf
= path
->nodes
[0];
3086 slot
= path
->slots
[0];
3087 write_extent_buffer(leaf
, buf
,
3088 btrfs_item_ptr_offset(leaf
, slot
),
3091 extent
= btrfs_item_ptr(leaf
, slot
,
3092 struct btrfs_file_extent_item
);
3094 /* disko == 0 means it's a hole */
3098 btrfs_set_file_extent_offset(leaf
, extent
,
3100 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3103 inode_add_bytes(inode
, datal
);
3104 ret
= btrfs_inc_extent_ref(trans
, root
,
3106 root
->root_key
.objectid
,
3108 new_key
.offset
- datao
,
3111 btrfs_abort_transaction(trans
,
3114 btrfs_end_transaction(trans
,
3120 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3123 if (off
> key
.offset
) {
3124 skip
= off
- key
.offset
;
3125 new_key
.offset
+= skip
;
3128 if (key
.offset
+ datal
> off
+ len
)
3129 trim
= key
.offset
+ datal
- (off
+ len
);
3131 if (comp
&& (skip
|| trim
)) {
3133 btrfs_end_transaction(trans
, root
);
3136 size
-= skip
+ trim
;
3137 datal
-= skip
+ trim
;
3139 ret
= btrfs_drop_extents(trans
, root
, inode
,
3141 new_key
.offset
+ datal
,
3144 if (ret
!= -EOPNOTSUPP
)
3145 btrfs_abort_transaction(trans
,
3147 btrfs_end_transaction(trans
, root
);
3151 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3154 btrfs_abort_transaction(trans
, root
,
3156 btrfs_end_transaction(trans
, root
);
3162 btrfs_file_extent_calc_inline_size(0);
3163 memmove(buf
+start
, buf
+start
+skip
,
3167 leaf
= path
->nodes
[0];
3168 slot
= path
->slots
[0];
3169 write_extent_buffer(leaf
, buf
,
3170 btrfs_item_ptr_offset(leaf
, slot
),
3172 inode_add_bytes(inode
, datal
);
3175 btrfs_mark_buffer_dirty(leaf
);
3176 btrfs_release_path(path
);
3178 inode_inc_iversion(inode
);
3179 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3182 * we round up to the block size at eof when
3183 * determining which extents to clone above,
3184 * but shouldn't round up the file size
3186 endoff
= new_key
.offset
+ datal
;
3187 if (endoff
> destoff
+olen
)
3188 endoff
= destoff
+olen
;
3189 if (endoff
> inode
->i_size
)
3190 btrfs_i_size_write(inode
, endoff
);
3192 ret
= btrfs_update_inode(trans
, root
, inode
);
3194 btrfs_abort_transaction(trans
, root
, ret
);
3195 btrfs_end_transaction(trans
, root
);
3198 ret
= btrfs_end_transaction(trans
, root
);
3200 btrfs_release_path(path
);
3206 btrfs_release_path(path
);
3207 btrfs_free_path(path
);
3212 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
3213 u64 off
, u64 olen
, u64 destoff
)
3215 struct inode
*inode
= file_inode(file
);
3216 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3221 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
3226 * - split compressed inline extents. annoying: we need to
3227 * decompress into destination's address_space (the file offset
3228 * may change, so source mapping won't do), then recompress (or
3229 * otherwise reinsert) a subrange.
3231 * - split destination inode's inline extents. The inline extents can
3232 * be either compressed or non-compressed.
3235 /* the destination must be opened for writing */
3236 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
3239 if (btrfs_root_readonly(root
))
3242 ret
= mnt_want_write_file(file
);
3246 src_file
= fdget(srcfd
);
3247 if (!src_file
.file
) {
3249 goto out_drop_write
;
3253 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
3256 src
= file_inode(src_file
.file
);
3262 /* the src must be open for reading */
3263 if (!(src_file
.file
->f_mode
& FMODE_READ
))
3266 /* don't make the dst file partly checksummed */
3267 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3268 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3272 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3276 if (src
->i_sb
!= inode
->i_sb
)
3281 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
3282 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
3284 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
3285 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
3288 mutex_lock(&src
->i_mutex
);
3291 /* determine range to clone */
3293 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3296 olen
= len
= src
->i_size
- off
;
3297 /* if we extend to eof, continue to block boundary */
3298 if (off
+ len
== src
->i_size
)
3299 len
= ALIGN(src
->i_size
, bs
) - off
;
3301 /* verify the end result is block aligned */
3302 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3303 !IS_ALIGNED(destoff
, bs
))
3306 /* verify if ranges are overlapped within the same file */
3308 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3312 if (destoff
> inode
->i_size
) {
3313 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3318 /* truncate page cache pages from target inode range */
3319 truncate_inode_pages_range(&inode
->i_data
, destoff
,
3320 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
3322 lock_extent_range(src
, off
, len
);
3324 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
);
3326 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
3330 mutex_unlock(&src
->i_mutex
);
3331 mutex_unlock(&inode
->i_mutex
);
3333 mutex_unlock(&inode
->i_mutex
);
3334 mutex_unlock(&src
->i_mutex
);
3337 mutex_unlock(&src
->i_mutex
);
3342 mnt_drop_write_file(file
);
3346 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
3348 struct btrfs_ioctl_clone_range_args args
;
3350 if (copy_from_user(&args
, argp
, sizeof(args
)))
3352 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
3353 args
.src_length
, args
.dest_offset
);
3357 * there are many ways the trans_start and trans_end ioctls can lead
3358 * to deadlocks. They should only be used by applications that
3359 * basically own the machine, and have a very in depth understanding
3360 * of all the possible deadlocks and enospc problems.
3362 static long btrfs_ioctl_trans_start(struct file
*file
)
3364 struct inode
*inode
= file_inode(file
);
3365 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3366 struct btrfs_trans_handle
*trans
;
3370 if (!capable(CAP_SYS_ADMIN
))
3374 if (file
->private_data
)
3378 if (btrfs_root_readonly(root
))
3381 ret
= mnt_want_write_file(file
);
3385 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
3388 trans
= btrfs_start_ioctl_transaction(root
);
3392 file
->private_data
= trans
;
3396 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3397 mnt_drop_write_file(file
);
3402 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
3404 struct inode
*inode
= file_inode(file
);
3405 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3406 struct btrfs_root
*new_root
;
3407 struct btrfs_dir_item
*di
;
3408 struct btrfs_trans_handle
*trans
;
3409 struct btrfs_path
*path
;
3410 struct btrfs_key location
;
3411 struct btrfs_disk_key disk_key
;
3416 if (!capable(CAP_SYS_ADMIN
))
3419 ret
= mnt_want_write_file(file
);
3423 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
3429 objectid
= BTRFS_FS_TREE_OBJECTID
;
3431 location
.objectid
= objectid
;
3432 location
.type
= BTRFS_ROOT_ITEM_KEY
;
3433 location
.offset
= (u64
)-1;
3435 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
3436 if (IS_ERR(new_root
)) {
3437 ret
= PTR_ERR(new_root
);
3441 path
= btrfs_alloc_path();
3446 path
->leave_spinning
= 1;
3448 trans
= btrfs_start_transaction(root
, 1);
3449 if (IS_ERR(trans
)) {
3450 btrfs_free_path(path
);
3451 ret
= PTR_ERR(trans
);
3455 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
3456 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
3457 dir_id
, "default", 7, 1);
3458 if (IS_ERR_OR_NULL(di
)) {
3459 btrfs_free_path(path
);
3460 btrfs_end_transaction(trans
, root
);
3461 btrfs_err(new_root
->fs_info
, "Umm, you don't have the default dir"
3462 "item, this isn't going to work");
3467 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
3468 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
3469 btrfs_mark_buffer_dirty(path
->nodes
[0]);
3470 btrfs_free_path(path
);
3472 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
3473 btrfs_end_transaction(trans
, root
);
3475 mnt_drop_write_file(file
);
3479 void btrfs_get_block_group_info(struct list_head
*groups_list
,
3480 struct btrfs_ioctl_space_info
*space
)
3482 struct btrfs_block_group_cache
*block_group
;
3484 space
->total_bytes
= 0;
3485 space
->used_bytes
= 0;
3487 list_for_each_entry(block_group
, groups_list
, list
) {
3488 space
->flags
= block_group
->flags
;
3489 space
->total_bytes
+= block_group
->key
.offset
;
3490 space
->used_bytes
+=
3491 btrfs_block_group_used(&block_group
->item
);
3495 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
3497 struct btrfs_ioctl_space_args space_args
;
3498 struct btrfs_ioctl_space_info space
;
3499 struct btrfs_ioctl_space_info
*dest
;
3500 struct btrfs_ioctl_space_info
*dest_orig
;
3501 struct btrfs_ioctl_space_info __user
*user_dest
;
3502 struct btrfs_space_info
*info
;
3503 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
3504 BTRFS_BLOCK_GROUP_SYSTEM
,
3505 BTRFS_BLOCK_GROUP_METADATA
,
3506 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
3513 if (copy_from_user(&space_args
,
3514 (struct btrfs_ioctl_space_args __user
*)arg
,
3515 sizeof(space_args
)))
3518 for (i
= 0; i
< num_types
; i
++) {
3519 struct btrfs_space_info
*tmp
;
3523 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3525 if (tmp
->flags
== types
[i
]) {
3535 down_read(&info
->groups_sem
);
3536 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3537 if (!list_empty(&info
->block_groups
[c
]))
3540 up_read(&info
->groups_sem
);
3544 * Global block reserve, exported as a space_info
3548 /* space_slots == 0 means they are asking for a count */
3549 if (space_args
.space_slots
== 0) {
3550 space_args
.total_spaces
= slot_count
;
3554 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
3556 alloc_size
= sizeof(*dest
) * slot_count
;
3558 /* we generally have at most 6 or so space infos, one for each raid
3559 * level. So, a whole page should be more than enough for everyone
3561 if (alloc_size
> PAGE_CACHE_SIZE
)
3564 space_args
.total_spaces
= 0;
3565 dest
= kmalloc(alloc_size
, GFP_NOFS
);
3570 /* now we have a buffer to copy into */
3571 for (i
= 0; i
< num_types
; i
++) {
3572 struct btrfs_space_info
*tmp
;
3579 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3581 if (tmp
->flags
== types
[i
]) {
3590 down_read(&info
->groups_sem
);
3591 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3592 if (!list_empty(&info
->block_groups
[c
])) {
3593 btrfs_get_block_group_info(
3594 &info
->block_groups
[c
], &space
);
3595 memcpy(dest
, &space
, sizeof(space
));
3597 space_args
.total_spaces
++;
3603 up_read(&info
->groups_sem
);
3607 * Add global block reserve
3610 struct btrfs_block_rsv
*block_rsv
= &root
->fs_info
->global_block_rsv
;
3612 spin_lock(&block_rsv
->lock
);
3613 space
.total_bytes
= block_rsv
->size
;
3614 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
3615 spin_unlock(&block_rsv
->lock
);
3616 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
3617 memcpy(dest
, &space
, sizeof(space
));
3618 space_args
.total_spaces
++;
3621 user_dest
= (struct btrfs_ioctl_space_info __user
*)
3622 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
3624 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
3629 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
3636 * there are many ways the trans_start and trans_end ioctls can lead
3637 * to deadlocks. They should only be used by applications that
3638 * basically own the machine, and have a very in depth understanding
3639 * of all the possible deadlocks and enospc problems.
3641 long btrfs_ioctl_trans_end(struct file
*file
)
3643 struct inode
*inode
= file_inode(file
);
3644 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3645 struct btrfs_trans_handle
*trans
;
3647 trans
= file
->private_data
;
3650 file
->private_data
= NULL
;
3652 btrfs_end_transaction(trans
, root
);
3654 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3656 mnt_drop_write_file(file
);
3660 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
3663 struct btrfs_trans_handle
*trans
;
3667 trans
= btrfs_attach_transaction_barrier(root
);
3668 if (IS_ERR(trans
)) {
3669 if (PTR_ERR(trans
) != -ENOENT
)
3670 return PTR_ERR(trans
);
3672 /* No running transaction, don't bother */
3673 transid
= root
->fs_info
->last_trans_committed
;
3676 transid
= trans
->transid
;
3677 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
3679 btrfs_end_transaction(trans
, root
);
3684 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
3689 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
3695 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
3698 transid
= 0; /* current trans */
3700 return btrfs_wait_for_commit(root
, transid
);
3703 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
3705 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3706 struct btrfs_ioctl_scrub_args
*sa
;
3709 if (!capable(CAP_SYS_ADMIN
))
3712 sa
= memdup_user(arg
, sizeof(*sa
));
3716 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
3717 ret
= mnt_want_write_file(file
);
3722 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
3723 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
3726 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3729 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
3730 mnt_drop_write_file(file
);
3736 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
3738 if (!capable(CAP_SYS_ADMIN
))
3741 return btrfs_scrub_cancel(root
->fs_info
);
3744 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
3747 struct btrfs_ioctl_scrub_args
*sa
;
3750 if (!capable(CAP_SYS_ADMIN
))
3753 sa
= memdup_user(arg
, sizeof(*sa
));
3757 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
3759 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3766 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
3769 struct btrfs_ioctl_get_dev_stats
*sa
;
3772 sa
= memdup_user(arg
, sizeof(*sa
));
3776 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
3781 ret
= btrfs_get_dev_stats(root
, sa
);
3783 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3790 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
3792 struct btrfs_ioctl_dev_replace_args
*p
;
3795 if (!capable(CAP_SYS_ADMIN
))
3798 p
= memdup_user(arg
, sizeof(*p
));
3803 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
3804 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
) {
3809 &root
->fs_info
->mutually_exclusive_operation_running
,
3811 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
3813 ret
= btrfs_dev_replace_start(root
, p
);
3815 &root
->fs_info
->mutually_exclusive_operation_running
,
3819 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
3820 btrfs_dev_replace_status(root
->fs_info
, p
);
3823 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
3824 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
3831 if (copy_to_user(arg
, p
, sizeof(*p
)))
3838 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
3844 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
3845 struct inode_fs_paths
*ipath
= NULL
;
3846 struct btrfs_path
*path
;
3848 if (!capable(CAP_DAC_READ_SEARCH
))
3851 path
= btrfs_alloc_path();
3857 ipa
= memdup_user(arg
, sizeof(*ipa
));
3864 size
= min_t(u32
, ipa
->size
, 4096);
3865 ipath
= init_ipath(size
, root
, path
);
3866 if (IS_ERR(ipath
)) {
3867 ret
= PTR_ERR(ipath
);
3872 ret
= paths_from_inode(ipa
->inum
, ipath
);
3876 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
3877 rel_ptr
= ipath
->fspath
->val
[i
] -
3878 (u64
)(unsigned long)ipath
->fspath
->val
;
3879 ipath
->fspath
->val
[i
] = rel_ptr
;
3882 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
3883 (void *)(unsigned long)ipath
->fspath
, size
);
3890 btrfs_free_path(path
);
3897 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
3899 struct btrfs_data_container
*inodes
= ctx
;
3900 const size_t c
= 3 * sizeof(u64
);
3902 if (inodes
->bytes_left
>= c
) {
3903 inodes
->bytes_left
-= c
;
3904 inodes
->val
[inodes
->elem_cnt
] = inum
;
3905 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
3906 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
3907 inodes
->elem_cnt
+= 3;
3909 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
3910 inodes
->bytes_left
= 0;
3911 inodes
->elem_missed
+= 3;
3917 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
3922 struct btrfs_ioctl_logical_ino_args
*loi
;
3923 struct btrfs_data_container
*inodes
= NULL
;
3924 struct btrfs_path
*path
= NULL
;
3926 if (!capable(CAP_SYS_ADMIN
))
3929 loi
= memdup_user(arg
, sizeof(*loi
));
3936 path
= btrfs_alloc_path();
3942 size
= min_t(u32
, loi
->size
, 64 * 1024);
3943 inodes
= init_data_container(size
);
3944 if (IS_ERR(inodes
)) {
3945 ret
= PTR_ERR(inodes
);
3950 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
3951 build_ino_list
, inodes
);
3957 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
3958 (void *)(unsigned long)inodes
, size
);
3963 btrfs_free_path(path
);
3970 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
3971 struct btrfs_ioctl_balance_args
*bargs
)
3973 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
3975 bargs
->flags
= bctl
->flags
;
3977 if (atomic_read(&fs_info
->balance_running
))
3978 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
3979 if (atomic_read(&fs_info
->balance_pause_req
))
3980 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
3981 if (atomic_read(&fs_info
->balance_cancel_req
))
3982 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
3984 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
3985 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
3986 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
3989 spin_lock(&fs_info
->balance_lock
);
3990 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3991 spin_unlock(&fs_info
->balance_lock
);
3993 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3997 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
3999 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4000 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4001 struct btrfs_ioctl_balance_args
*bargs
;
4002 struct btrfs_balance_control
*bctl
;
4003 bool need_unlock
; /* for mut. excl. ops lock */
4006 if (!capable(CAP_SYS_ADMIN
))
4009 ret
= mnt_want_write_file(file
);
4014 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4015 mutex_lock(&fs_info
->volume_mutex
);
4016 mutex_lock(&fs_info
->balance_mutex
);
4022 * mut. excl. ops lock is locked. Three possibilites:
4023 * (1) some other op is running
4024 * (2) balance is running
4025 * (3) balance is paused -- special case (think resume)
4027 mutex_lock(&fs_info
->balance_mutex
);
4028 if (fs_info
->balance_ctl
) {
4029 /* this is either (2) or (3) */
4030 if (!atomic_read(&fs_info
->balance_running
)) {
4031 mutex_unlock(&fs_info
->balance_mutex
);
4032 if (!mutex_trylock(&fs_info
->volume_mutex
))
4034 mutex_lock(&fs_info
->balance_mutex
);
4036 if (fs_info
->balance_ctl
&&
4037 !atomic_read(&fs_info
->balance_running
)) {
4039 need_unlock
= false;
4043 mutex_unlock(&fs_info
->balance_mutex
);
4044 mutex_unlock(&fs_info
->volume_mutex
);
4048 mutex_unlock(&fs_info
->balance_mutex
);
4054 mutex_unlock(&fs_info
->balance_mutex
);
4055 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4060 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4063 bargs
= memdup_user(arg
, sizeof(*bargs
));
4064 if (IS_ERR(bargs
)) {
4065 ret
= PTR_ERR(bargs
);
4069 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4070 if (!fs_info
->balance_ctl
) {
4075 bctl
= fs_info
->balance_ctl
;
4076 spin_lock(&fs_info
->balance_lock
);
4077 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4078 spin_unlock(&fs_info
->balance_lock
);
4086 if (fs_info
->balance_ctl
) {
4091 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
4097 bctl
->fs_info
= fs_info
;
4099 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4100 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4101 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4103 bctl
->flags
= bargs
->flags
;
4105 /* balance everything - no filters */
4106 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4111 * Ownership of bctl and mutually_exclusive_operation_running
4112 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4113 * or, if restriper was paused all the way until unmount, in
4114 * free_fs_info. mutually_exclusive_operation_running is
4115 * cleared in __cancel_balance.
4117 need_unlock
= false;
4119 ret
= btrfs_balance(bctl
, bargs
);
4122 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4129 mutex_unlock(&fs_info
->balance_mutex
);
4130 mutex_unlock(&fs_info
->volume_mutex
);
4132 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4134 mnt_drop_write_file(file
);
4138 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4140 if (!capable(CAP_SYS_ADMIN
))
4144 case BTRFS_BALANCE_CTL_PAUSE
:
4145 return btrfs_pause_balance(root
->fs_info
);
4146 case BTRFS_BALANCE_CTL_CANCEL
:
4147 return btrfs_cancel_balance(root
->fs_info
);
4153 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4156 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4157 struct btrfs_ioctl_balance_args
*bargs
;
4160 if (!capable(CAP_SYS_ADMIN
))
4163 mutex_lock(&fs_info
->balance_mutex
);
4164 if (!fs_info
->balance_ctl
) {
4169 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
4175 update_ioctl_balance_args(fs_info
, 1, bargs
);
4177 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4182 mutex_unlock(&fs_info
->balance_mutex
);
4186 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4188 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4189 struct btrfs_ioctl_quota_ctl_args
*sa
;
4190 struct btrfs_trans_handle
*trans
= NULL
;
4194 if (!capable(CAP_SYS_ADMIN
))
4197 ret
= mnt_want_write_file(file
);
4201 sa
= memdup_user(arg
, sizeof(*sa
));
4207 down_write(&root
->fs_info
->subvol_sem
);
4208 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
4209 if (IS_ERR(trans
)) {
4210 ret
= PTR_ERR(trans
);
4215 case BTRFS_QUOTA_CTL_ENABLE
:
4216 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
4218 case BTRFS_QUOTA_CTL_DISABLE
:
4219 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
4226 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
4231 up_write(&root
->fs_info
->subvol_sem
);
4233 mnt_drop_write_file(file
);
4237 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4239 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4240 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4241 struct btrfs_trans_handle
*trans
;
4245 if (!capable(CAP_SYS_ADMIN
))
4248 ret
= mnt_want_write_file(file
);
4252 sa
= memdup_user(arg
, sizeof(*sa
));
4258 trans
= btrfs_join_transaction(root
);
4259 if (IS_ERR(trans
)) {
4260 ret
= PTR_ERR(trans
);
4264 /* FIXME: check if the IDs really exist */
4266 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
4269 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
4273 err
= btrfs_end_transaction(trans
, root
);
4280 mnt_drop_write_file(file
);
4284 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4286 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4287 struct btrfs_ioctl_qgroup_create_args
*sa
;
4288 struct btrfs_trans_handle
*trans
;
4292 if (!capable(CAP_SYS_ADMIN
))
4295 ret
= mnt_want_write_file(file
);
4299 sa
= memdup_user(arg
, sizeof(*sa
));
4305 if (!sa
->qgroupid
) {
4310 trans
= btrfs_join_transaction(root
);
4311 if (IS_ERR(trans
)) {
4312 ret
= PTR_ERR(trans
);
4316 /* FIXME: check if the IDs really exist */
4318 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
,
4321 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4324 err
= btrfs_end_transaction(trans
, root
);
4331 mnt_drop_write_file(file
);
4335 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4337 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4338 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4339 struct btrfs_trans_handle
*trans
;
4344 if (!capable(CAP_SYS_ADMIN
))
4347 ret
= mnt_want_write_file(file
);
4351 sa
= memdup_user(arg
, sizeof(*sa
));
4357 trans
= btrfs_join_transaction(root
);
4358 if (IS_ERR(trans
)) {
4359 ret
= PTR_ERR(trans
);
4363 qgroupid
= sa
->qgroupid
;
4365 /* take the current subvol as qgroup */
4366 qgroupid
= root
->root_key
.objectid
;
4369 /* FIXME: check if the IDs really exist */
4370 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
4372 err
= btrfs_end_transaction(trans
, root
);
4379 mnt_drop_write_file(file
);
4383 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
4385 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4386 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4389 if (!capable(CAP_SYS_ADMIN
))
4392 ret
= mnt_want_write_file(file
);
4396 qsa
= memdup_user(arg
, sizeof(*qsa
));
4407 ret
= btrfs_qgroup_rescan(root
->fs_info
);
4412 mnt_drop_write_file(file
);
4416 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
4418 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4419 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4422 if (!capable(CAP_SYS_ADMIN
))
4425 qsa
= kzalloc(sizeof(*qsa
), GFP_NOFS
);
4429 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
4431 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
4434 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
4441 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
4443 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4445 if (!capable(CAP_SYS_ADMIN
))
4448 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
4451 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
4452 struct btrfs_ioctl_received_subvol_args
*sa
)
4454 struct inode
*inode
= file_inode(file
);
4455 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4456 struct btrfs_root_item
*root_item
= &root
->root_item
;
4457 struct btrfs_trans_handle
*trans
;
4458 struct timespec ct
= CURRENT_TIME
;
4460 int received_uuid_changed
;
4462 if (!inode_owner_or_capable(inode
))
4465 ret
= mnt_want_write_file(file
);
4469 down_write(&root
->fs_info
->subvol_sem
);
4471 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
4476 if (btrfs_root_readonly(root
)) {
4483 * 2 - uuid items (received uuid + subvol uuid)
4485 trans
= btrfs_start_transaction(root
, 3);
4486 if (IS_ERR(trans
)) {
4487 ret
= PTR_ERR(trans
);
4492 sa
->rtransid
= trans
->transid
;
4493 sa
->rtime
.sec
= ct
.tv_sec
;
4494 sa
->rtime
.nsec
= ct
.tv_nsec
;
4496 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
4498 if (received_uuid_changed
&&
4499 !btrfs_is_empty_uuid(root_item
->received_uuid
))
4500 btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
4501 root_item
->received_uuid
,
4502 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
4503 root
->root_key
.objectid
);
4504 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
4505 btrfs_set_root_stransid(root_item
, sa
->stransid
);
4506 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
4507 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
4508 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
4509 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
4510 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
4512 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
4513 &root
->root_key
, &root
->root_item
);
4515 btrfs_end_transaction(trans
, root
);
4518 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
4519 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
4521 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
4522 root
->root_key
.objectid
);
4523 if (ret
< 0 && ret
!= -EEXIST
) {
4524 btrfs_abort_transaction(trans
, root
, ret
);
4528 ret
= btrfs_commit_transaction(trans
, root
);
4530 btrfs_abort_transaction(trans
, root
, ret
);
4535 up_write(&root
->fs_info
->subvol_sem
);
4536 mnt_drop_write_file(file
);
4541 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
4544 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
4545 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
4548 args32
= memdup_user(arg
, sizeof(*args32
));
4549 if (IS_ERR(args32
)) {
4550 ret
= PTR_ERR(args32
);
4555 args64
= kmalloc(sizeof(*args64
), GFP_NOFS
);
4561 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
4562 args64
->stransid
= args32
->stransid
;
4563 args64
->rtransid
= args32
->rtransid
;
4564 args64
->stime
.sec
= args32
->stime
.sec
;
4565 args64
->stime
.nsec
= args32
->stime
.nsec
;
4566 args64
->rtime
.sec
= args32
->rtime
.sec
;
4567 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
4568 args64
->flags
= args32
->flags
;
4570 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
4574 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
4575 args32
->stransid
= args64
->stransid
;
4576 args32
->rtransid
= args64
->rtransid
;
4577 args32
->stime
.sec
= args64
->stime
.sec
;
4578 args32
->stime
.nsec
= args64
->stime
.nsec
;
4579 args32
->rtime
.sec
= args64
->rtime
.sec
;
4580 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
4581 args32
->flags
= args64
->flags
;
4583 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
4594 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
4597 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
4600 sa
= memdup_user(arg
, sizeof(*sa
));
4607 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
4612 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
4621 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
4623 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4626 char label
[BTRFS_LABEL_SIZE
];
4628 spin_lock(&root
->fs_info
->super_lock
);
4629 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
4630 spin_unlock(&root
->fs_info
->super_lock
);
4632 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
4634 if (len
== BTRFS_LABEL_SIZE
) {
4635 btrfs_warn(root
->fs_info
,
4636 "label is too long, return the first %zu bytes", --len
);
4639 ret
= copy_to_user(arg
, label
, len
);
4641 return ret
? -EFAULT
: 0;
4644 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
4646 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4647 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
4648 struct btrfs_trans_handle
*trans
;
4649 char label
[BTRFS_LABEL_SIZE
];
4652 if (!capable(CAP_SYS_ADMIN
))
4655 if (copy_from_user(label
, arg
, sizeof(label
)))
4658 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
4659 btrfs_err(root
->fs_info
, "unable to set label with more than %d bytes",
4660 BTRFS_LABEL_SIZE
- 1);
4664 ret
= mnt_want_write_file(file
);
4668 trans
= btrfs_start_transaction(root
, 0);
4669 if (IS_ERR(trans
)) {
4670 ret
= PTR_ERR(trans
);
4674 spin_lock(&root
->fs_info
->super_lock
);
4675 strcpy(super_block
->label
, label
);
4676 spin_unlock(&root
->fs_info
->super_lock
);
4677 ret
= btrfs_commit_transaction(trans
, root
);
4680 mnt_drop_write_file(file
);
4684 #define INIT_FEATURE_FLAGS(suffix) \
4685 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4686 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4687 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4689 static int btrfs_ioctl_get_supported_features(struct file
*file
,
4692 static struct btrfs_ioctl_feature_flags features
[3] = {
4693 INIT_FEATURE_FLAGS(SUPP
),
4694 INIT_FEATURE_FLAGS(SAFE_SET
),
4695 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
4698 if (copy_to_user(arg
, &features
, sizeof(features
)))
4704 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
4706 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4707 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
4708 struct btrfs_ioctl_feature_flags features
;
4710 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
4711 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
4712 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
4714 if (copy_to_user(arg
, &features
, sizeof(features
)))
4720 static int check_feature_bits(struct btrfs_root
*root
,
4721 enum btrfs_feature_set set
,
4722 u64 change_mask
, u64 flags
, u64 supported_flags
,
4723 u64 safe_set
, u64 safe_clear
)
4725 const char *type
= btrfs_feature_set_names
[set
];
4727 u64 disallowed
, unsupported
;
4728 u64 set_mask
= flags
& change_mask
;
4729 u64 clear_mask
= ~flags
& change_mask
;
4731 unsupported
= set_mask
& ~supported_flags
;
4733 names
= btrfs_printable_features(set
, unsupported
);
4735 btrfs_warn(root
->fs_info
,
4736 "this kernel does not support the %s feature bit%s",
4737 names
, strchr(names
, ',') ? "s" : "");
4740 btrfs_warn(root
->fs_info
,
4741 "this kernel does not support %s bits 0x%llx",
4746 disallowed
= set_mask
& ~safe_set
;
4748 names
= btrfs_printable_features(set
, disallowed
);
4750 btrfs_warn(root
->fs_info
,
4751 "can't set the %s feature bit%s while mounted",
4752 names
, strchr(names
, ',') ? "s" : "");
4755 btrfs_warn(root
->fs_info
,
4756 "can't set %s bits 0x%llx while mounted",
4761 disallowed
= clear_mask
& ~safe_clear
;
4763 names
= btrfs_printable_features(set
, disallowed
);
4765 btrfs_warn(root
->fs_info
,
4766 "can't clear the %s feature bit%s while mounted",
4767 names
, strchr(names
, ',') ? "s" : "");
4770 btrfs_warn(root
->fs_info
,
4771 "can't clear %s bits 0x%llx while mounted",
4779 #define check_feature(root, change_mask, flags, mask_base) \
4780 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
4781 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4782 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4783 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4785 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
4787 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4788 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
4789 struct btrfs_ioctl_feature_flags flags
[2];
4790 struct btrfs_trans_handle
*trans
;
4794 if (!capable(CAP_SYS_ADMIN
))
4797 if (copy_from_user(flags
, arg
, sizeof(flags
)))
4801 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
4802 !flags
[0].incompat_flags
)
4805 ret
= check_feature(root
, flags
[0].compat_flags
,
4806 flags
[1].compat_flags
, COMPAT
);
4810 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
4811 flags
[1].compat_ro_flags
, COMPAT_RO
);
4815 ret
= check_feature(root
, flags
[0].incompat_flags
,
4816 flags
[1].incompat_flags
, INCOMPAT
);
4820 trans
= btrfs_start_transaction(root
, 0);
4822 return PTR_ERR(trans
);
4824 spin_lock(&root
->fs_info
->super_lock
);
4825 newflags
= btrfs_super_compat_flags(super_block
);
4826 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
4827 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
4828 btrfs_set_super_compat_flags(super_block
, newflags
);
4830 newflags
= btrfs_super_compat_ro_flags(super_block
);
4831 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
4832 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
4833 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
4835 newflags
= btrfs_super_incompat_flags(super_block
);
4836 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
4837 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
4838 btrfs_set_super_incompat_flags(super_block
, newflags
);
4839 spin_unlock(&root
->fs_info
->super_lock
);
4841 return btrfs_commit_transaction(trans
, root
);
4844 long btrfs_ioctl(struct file
*file
, unsigned int
4845 cmd
, unsigned long arg
)
4847 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4848 void __user
*argp
= (void __user
*)arg
;
4851 case FS_IOC_GETFLAGS
:
4852 return btrfs_ioctl_getflags(file
, argp
);
4853 case FS_IOC_SETFLAGS
:
4854 return btrfs_ioctl_setflags(file
, argp
);
4855 case FS_IOC_GETVERSION
:
4856 return btrfs_ioctl_getversion(file
, argp
);
4858 return btrfs_ioctl_fitrim(file
, argp
);
4859 case BTRFS_IOC_SNAP_CREATE
:
4860 return btrfs_ioctl_snap_create(file
, argp
, 0);
4861 case BTRFS_IOC_SNAP_CREATE_V2
:
4862 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
4863 case BTRFS_IOC_SUBVOL_CREATE
:
4864 return btrfs_ioctl_snap_create(file
, argp
, 1);
4865 case BTRFS_IOC_SUBVOL_CREATE_V2
:
4866 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
4867 case BTRFS_IOC_SNAP_DESTROY
:
4868 return btrfs_ioctl_snap_destroy(file
, argp
);
4869 case BTRFS_IOC_SUBVOL_GETFLAGS
:
4870 return btrfs_ioctl_subvol_getflags(file
, argp
);
4871 case BTRFS_IOC_SUBVOL_SETFLAGS
:
4872 return btrfs_ioctl_subvol_setflags(file
, argp
);
4873 case BTRFS_IOC_DEFAULT_SUBVOL
:
4874 return btrfs_ioctl_default_subvol(file
, argp
);
4875 case BTRFS_IOC_DEFRAG
:
4876 return btrfs_ioctl_defrag(file
, NULL
);
4877 case BTRFS_IOC_DEFRAG_RANGE
:
4878 return btrfs_ioctl_defrag(file
, argp
);
4879 case BTRFS_IOC_RESIZE
:
4880 return btrfs_ioctl_resize(file
, argp
);
4881 case BTRFS_IOC_ADD_DEV
:
4882 return btrfs_ioctl_add_dev(root
, argp
);
4883 case BTRFS_IOC_RM_DEV
:
4884 return btrfs_ioctl_rm_dev(file
, argp
);
4885 case BTRFS_IOC_FS_INFO
:
4886 return btrfs_ioctl_fs_info(root
, argp
);
4887 case BTRFS_IOC_DEV_INFO
:
4888 return btrfs_ioctl_dev_info(root
, argp
);
4889 case BTRFS_IOC_BALANCE
:
4890 return btrfs_ioctl_balance(file
, NULL
);
4891 case BTRFS_IOC_CLONE
:
4892 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
4893 case BTRFS_IOC_CLONE_RANGE
:
4894 return btrfs_ioctl_clone_range(file
, argp
);
4895 case BTRFS_IOC_TRANS_START
:
4896 return btrfs_ioctl_trans_start(file
);
4897 case BTRFS_IOC_TRANS_END
:
4898 return btrfs_ioctl_trans_end(file
);
4899 case BTRFS_IOC_TREE_SEARCH
:
4900 return btrfs_ioctl_tree_search(file
, argp
);
4901 case BTRFS_IOC_INO_LOOKUP
:
4902 return btrfs_ioctl_ino_lookup(file
, argp
);
4903 case BTRFS_IOC_INO_PATHS
:
4904 return btrfs_ioctl_ino_to_path(root
, argp
);
4905 case BTRFS_IOC_LOGICAL_INO
:
4906 return btrfs_ioctl_logical_to_ino(root
, argp
);
4907 case BTRFS_IOC_SPACE_INFO
:
4908 return btrfs_ioctl_space_info(root
, argp
);
4909 case BTRFS_IOC_SYNC
: {
4912 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
4915 ret
= btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
4918 case BTRFS_IOC_START_SYNC
:
4919 return btrfs_ioctl_start_sync(root
, argp
);
4920 case BTRFS_IOC_WAIT_SYNC
:
4921 return btrfs_ioctl_wait_sync(root
, argp
);
4922 case BTRFS_IOC_SCRUB
:
4923 return btrfs_ioctl_scrub(file
, argp
);
4924 case BTRFS_IOC_SCRUB_CANCEL
:
4925 return btrfs_ioctl_scrub_cancel(root
, argp
);
4926 case BTRFS_IOC_SCRUB_PROGRESS
:
4927 return btrfs_ioctl_scrub_progress(root
, argp
);
4928 case BTRFS_IOC_BALANCE_V2
:
4929 return btrfs_ioctl_balance(file
, argp
);
4930 case BTRFS_IOC_BALANCE_CTL
:
4931 return btrfs_ioctl_balance_ctl(root
, arg
);
4932 case BTRFS_IOC_BALANCE_PROGRESS
:
4933 return btrfs_ioctl_balance_progress(root
, argp
);
4934 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
4935 return btrfs_ioctl_set_received_subvol(file
, argp
);
4937 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
4938 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
4940 case BTRFS_IOC_SEND
:
4941 return btrfs_ioctl_send(file
, argp
);
4942 case BTRFS_IOC_GET_DEV_STATS
:
4943 return btrfs_ioctl_get_dev_stats(root
, argp
);
4944 case BTRFS_IOC_QUOTA_CTL
:
4945 return btrfs_ioctl_quota_ctl(file
, argp
);
4946 case BTRFS_IOC_QGROUP_ASSIGN
:
4947 return btrfs_ioctl_qgroup_assign(file
, argp
);
4948 case BTRFS_IOC_QGROUP_CREATE
:
4949 return btrfs_ioctl_qgroup_create(file
, argp
);
4950 case BTRFS_IOC_QGROUP_LIMIT
:
4951 return btrfs_ioctl_qgroup_limit(file
, argp
);
4952 case BTRFS_IOC_QUOTA_RESCAN
:
4953 return btrfs_ioctl_quota_rescan(file
, argp
);
4954 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
4955 return btrfs_ioctl_quota_rescan_status(file
, argp
);
4956 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
4957 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
4958 case BTRFS_IOC_DEV_REPLACE
:
4959 return btrfs_ioctl_dev_replace(root
, argp
);
4960 case BTRFS_IOC_GET_FSLABEL
:
4961 return btrfs_ioctl_get_fslabel(file
, argp
);
4962 case BTRFS_IOC_SET_FSLABEL
:
4963 return btrfs_ioctl_set_fslabel(file
, argp
);
4964 case BTRFS_IOC_FILE_EXTENT_SAME
:
4965 return btrfs_ioctl_file_extent_same(file
, argp
);
4966 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
4967 return btrfs_ioctl_get_supported_features(file
, argp
);
4968 case BTRFS_IOC_GET_FEATURES
:
4969 return btrfs_ioctl_get_features(file
, argp
);
4970 case BTRFS_IOC_SET_FEATURES
:
4971 return btrfs_ioctl_set_features(file
, argp
);