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"
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
69 struct btrfs_ioctl_timespec_32
{
72 } __attribute__ ((__packed__
));
74 struct btrfs_ioctl_received_subvol_args_32
{
75 char uuid
[BTRFS_UUID_SIZE
]; /* in */
76 __u64 stransid
; /* in */
77 __u64 rtransid
; /* out */
78 struct btrfs_ioctl_timespec_32 stime
; /* in */
79 struct btrfs_ioctl_timespec_32 rtime
; /* out */
81 __u64 reserved
[16]; /* in */
82 } __attribute__ ((__packed__
));
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
89 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
90 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
,
93 /* Mask out flags that are inappropriate for the given type of inode. */
94 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
98 else if (S_ISREG(mode
))
99 return flags
& ~FS_DIRSYNC_FL
;
101 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
105 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
107 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
109 unsigned int iflags
= 0;
111 if (flags
& BTRFS_INODE_SYNC
)
112 iflags
|= FS_SYNC_FL
;
113 if (flags
& BTRFS_INODE_IMMUTABLE
)
114 iflags
|= FS_IMMUTABLE_FL
;
115 if (flags
& BTRFS_INODE_APPEND
)
116 iflags
|= FS_APPEND_FL
;
117 if (flags
& BTRFS_INODE_NODUMP
)
118 iflags
|= FS_NODUMP_FL
;
119 if (flags
& BTRFS_INODE_NOATIME
)
120 iflags
|= FS_NOATIME_FL
;
121 if (flags
& BTRFS_INODE_DIRSYNC
)
122 iflags
|= FS_DIRSYNC_FL
;
123 if (flags
& BTRFS_INODE_NODATACOW
)
124 iflags
|= FS_NOCOW_FL
;
126 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
127 iflags
|= FS_COMPR_FL
;
128 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
129 iflags
|= FS_NOCOMP_FL
;
135 * Update inode->i_flags based on the btrfs internal flags.
137 void btrfs_update_iflags(struct inode
*inode
)
139 struct btrfs_inode
*ip
= BTRFS_I(inode
);
140 unsigned int new_fl
= 0;
142 if (ip
->flags
& BTRFS_INODE_SYNC
)
144 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
145 new_fl
|= S_IMMUTABLE
;
146 if (ip
->flags
& BTRFS_INODE_APPEND
)
148 if (ip
->flags
& BTRFS_INODE_NOATIME
)
150 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
153 set_mask_bits(&inode
->i_flags
,
154 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
,
159 * Inherit flags from the parent inode.
161 * Currently only the compression flags and the cow flags are inherited.
163 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
170 flags
= BTRFS_I(dir
)->flags
;
172 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
173 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
174 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
175 } else if (flags
& BTRFS_INODE_COMPRESS
) {
176 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
177 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
180 if (flags
& BTRFS_INODE_NODATACOW
) {
181 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
182 if (S_ISREG(inode
->i_mode
))
183 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
186 btrfs_update_iflags(inode
);
189 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
191 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
192 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
194 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
199 static int check_flags(unsigned int flags
)
201 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
202 FS_NOATIME_FL
| FS_NODUMP_FL
| \
203 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
204 FS_NOCOMP_FL
| FS_COMPR_FL
|
208 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
214 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
216 struct inode
*inode
= file_inode(file
);
217 struct btrfs_inode
*ip
= BTRFS_I(inode
);
218 struct btrfs_root
*root
= ip
->root
;
219 struct btrfs_trans_handle
*trans
;
220 unsigned int flags
, oldflags
;
223 unsigned int i_oldflags
;
226 if (!inode_owner_or_capable(inode
))
229 if (btrfs_root_readonly(root
))
232 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
235 ret
= check_flags(flags
);
239 ret
= mnt_want_write_file(file
);
245 ip_oldflags
= ip
->flags
;
246 i_oldflags
= inode
->i_flags
;
247 mode
= inode
->i_mode
;
249 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
250 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
251 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
252 if (!capable(CAP_LINUX_IMMUTABLE
)) {
258 if (flags
& FS_SYNC_FL
)
259 ip
->flags
|= BTRFS_INODE_SYNC
;
261 ip
->flags
&= ~BTRFS_INODE_SYNC
;
262 if (flags
& FS_IMMUTABLE_FL
)
263 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
265 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
266 if (flags
& FS_APPEND_FL
)
267 ip
->flags
|= BTRFS_INODE_APPEND
;
269 ip
->flags
&= ~BTRFS_INODE_APPEND
;
270 if (flags
& FS_NODUMP_FL
)
271 ip
->flags
|= BTRFS_INODE_NODUMP
;
273 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
274 if (flags
& FS_NOATIME_FL
)
275 ip
->flags
|= BTRFS_INODE_NOATIME
;
277 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
278 if (flags
& FS_DIRSYNC_FL
)
279 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
281 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
282 if (flags
& FS_NOCOW_FL
) {
285 * It's safe to turn csums off here, no extents exist.
286 * Otherwise we want the flag to reflect the real COW
287 * status of the file and will not set it.
289 if (inode
->i_size
== 0)
290 ip
->flags
|= BTRFS_INODE_NODATACOW
291 | BTRFS_INODE_NODATASUM
;
293 ip
->flags
|= BTRFS_INODE_NODATACOW
;
297 * Revert back under same assuptions as above
300 if (inode
->i_size
== 0)
301 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
302 | BTRFS_INODE_NODATASUM
);
304 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
309 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
310 * flag may be changed automatically if compression code won't make
313 if (flags
& FS_NOCOMP_FL
) {
314 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
315 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
317 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
318 if (ret
&& ret
!= -ENODATA
)
320 } else if (flags
& FS_COMPR_FL
) {
323 ip
->flags
|= BTRFS_INODE_COMPRESS
;
324 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
326 if (root
->fs_info
->compress_type
== BTRFS_COMPRESS_LZO
)
330 ret
= btrfs_set_prop(inode
, "btrfs.compression",
331 comp
, strlen(comp
), 0);
336 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
337 if (ret
&& ret
!= -ENODATA
)
339 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
342 trans
= btrfs_start_transaction(root
, 1);
344 ret
= PTR_ERR(trans
);
348 btrfs_update_iflags(inode
);
349 inode_inc_iversion(inode
);
350 inode
->i_ctime
= current_fs_time(inode
->i_sb
);
351 ret
= btrfs_update_inode(trans
, root
, inode
);
353 btrfs_end_transaction(trans
, root
);
356 ip
->flags
= ip_oldflags
;
357 inode
->i_flags
= i_oldflags
;
362 mnt_drop_write_file(file
);
366 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
368 struct inode
*inode
= file_inode(file
);
370 return put_user(inode
->i_generation
, arg
);
373 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
375 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
376 struct btrfs_device
*device
;
377 struct request_queue
*q
;
378 struct fstrim_range range
;
379 u64 minlen
= ULLONG_MAX
;
381 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
384 if (!capable(CAP_SYS_ADMIN
))
388 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
392 q
= bdev_get_queue(device
->bdev
);
393 if (blk_queue_discard(q
)) {
395 minlen
= min((u64
)q
->limits
.discard_granularity
,
403 if (copy_from_user(&range
, arg
, sizeof(range
)))
405 if (range
.start
> total_bytes
||
406 range
.len
< fs_info
->sb
->s_blocksize
)
409 range
.len
= min(range
.len
, total_bytes
- range
.start
);
410 range
.minlen
= max(range
.minlen
, minlen
);
411 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
415 if (copy_to_user(arg
, &range
, sizeof(range
)))
421 int btrfs_is_empty_uuid(u8
*uuid
)
425 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
432 static noinline
int create_subvol(struct inode
*dir
,
433 struct dentry
*dentry
,
434 char *name
, int namelen
,
436 struct btrfs_qgroup_inherit
*inherit
)
438 struct btrfs_trans_handle
*trans
;
439 struct btrfs_key key
;
440 struct btrfs_root_item root_item
;
441 struct btrfs_inode_item
*inode_item
;
442 struct extent_buffer
*leaf
;
443 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
444 struct btrfs_root
*new_root
;
445 struct btrfs_block_rsv block_rsv
;
446 struct timespec cur_time
= current_fs_time(dir
->i_sb
);
451 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
456 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
461 * Don't create subvolume whose level is not zero. Or qgroup will be
462 * screwed up since it assume subvolme qgroup's level to be 0.
464 if (btrfs_qgroup_level(objectid
))
467 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
469 * The same as the snapshot creation, please see the comment
470 * of create_snapshot().
472 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
473 8, &qgroup_reserved
, false);
477 trans
= btrfs_start_transaction(root
, 0);
479 ret
= PTR_ERR(trans
);
480 btrfs_subvolume_release_metadata(root
, &block_rsv
,
484 trans
->block_rsv
= &block_rsv
;
485 trans
->bytes_reserved
= block_rsv
.size
;
487 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
491 leaf
= btrfs_alloc_tree_block(trans
, root
, 0, objectid
, NULL
, 0, 0, 0);
497 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
498 btrfs_set_header_bytenr(leaf
, leaf
->start
);
499 btrfs_set_header_generation(leaf
, trans
->transid
);
500 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
501 btrfs_set_header_owner(leaf
, objectid
);
503 write_extent_buffer(leaf
, root
->fs_info
->fsid
, btrfs_header_fsid(),
505 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
506 btrfs_header_chunk_tree_uuid(leaf
),
508 btrfs_mark_buffer_dirty(leaf
);
510 memset(&root_item
, 0, sizeof(root_item
));
512 inode_item
= &root_item
.inode
;
513 btrfs_set_stack_inode_generation(inode_item
, 1);
514 btrfs_set_stack_inode_size(inode_item
, 3);
515 btrfs_set_stack_inode_nlink(inode_item
, 1);
516 btrfs_set_stack_inode_nbytes(inode_item
, root
->nodesize
);
517 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
519 btrfs_set_root_flags(&root_item
, 0);
520 btrfs_set_root_limit(&root_item
, 0);
521 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
523 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
524 btrfs_set_root_generation(&root_item
, trans
->transid
);
525 btrfs_set_root_level(&root_item
, 0);
526 btrfs_set_root_refs(&root_item
, 1);
527 btrfs_set_root_used(&root_item
, leaf
->len
);
528 btrfs_set_root_last_snapshot(&root_item
, 0);
530 btrfs_set_root_generation_v2(&root_item
,
531 btrfs_root_generation(&root_item
));
532 uuid_le_gen(&new_uuid
);
533 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
534 btrfs_set_stack_timespec_sec(&root_item
.otime
, cur_time
.tv_sec
);
535 btrfs_set_stack_timespec_nsec(&root_item
.otime
, cur_time
.tv_nsec
);
536 root_item
.ctime
= root_item
.otime
;
537 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
538 btrfs_set_root_otransid(&root_item
, trans
->transid
);
540 btrfs_tree_unlock(leaf
);
541 free_extent_buffer(leaf
);
544 btrfs_set_root_dirid(&root_item
, new_dirid
);
546 key
.objectid
= objectid
;
548 key
.type
= BTRFS_ROOT_ITEM_KEY
;
549 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
554 key
.offset
= (u64
)-1;
555 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
556 if (IS_ERR(new_root
)) {
557 ret
= PTR_ERR(new_root
);
558 btrfs_abort_transaction(trans
, root
, ret
);
562 btrfs_record_root_in_trans(trans
, new_root
);
564 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
566 /* We potentially lose an unused inode item here */
567 btrfs_abort_transaction(trans
, root
, ret
);
571 mutex_lock(&new_root
->objectid_mutex
);
572 new_root
->highest_objectid
= new_dirid
;
573 mutex_unlock(&new_root
->objectid_mutex
);
576 * insert the directory item
578 ret
= btrfs_set_inode_index(dir
, &index
);
580 btrfs_abort_transaction(trans
, root
, ret
);
584 ret
= btrfs_insert_dir_item(trans
, root
,
585 name
, namelen
, dir
, &key
,
586 BTRFS_FT_DIR
, index
);
588 btrfs_abort_transaction(trans
, root
, ret
);
592 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
593 ret
= btrfs_update_inode(trans
, root
, dir
);
596 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
597 objectid
, root
->root_key
.objectid
,
598 btrfs_ino(dir
), index
, name
, namelen
);
601 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
602 root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
605 btrfs_abort_transaction(trans
, root
, ret
);
608 trans
->block_rsv
= NULL
;
609 trans
->bytes_reserved
= 0;
610 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
613 *async_transid
= trans
->transid
;
614 err
= btrfs_commit_transaction_async(trans
, root
, 1);
616 err
= btrfs_commit_transaction(trans
, root
);
618 err
= btrfs_commit_transaction(trans
, root
);
624 inode
= btrfs_lookup_dentry(dir
, dentry
);
626 return PTR_ERR(inode
);
627 d_instantiate(dentry
, inode
);
632 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root
*root
)
638 prepare_to_wait(&root
->subv_writers
->wait
, &wait
,
639 TASK_UNINTERRUPTIBLE
);
641 writers
= percpu_counter_sum(&root
->subv_writers
->counter
);
645 finish_wait(&root
->subv_writers
->wait
, &wait
);
649 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
650 struct dentry
*dentry
, char *name
, int namelen
,
651 u64
*async_transid
, bool readonly
,
652 struct btrfs_qgroup_inherit
*inherit
)
655 struct btrfs_pending_snapshot
*pending_snapshot
;
656 struct btrfs_trans_handle
*trans
;
659 if (!test_bit(BTRFS_ROOT_REF_COWS
, &root
->state
))
662 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
663 if (!pending_snapshot
)
666 pending_snapshot
->root_item
= kzalloc(sizeof(struct btrfs_root_item
),
668 pending_snapshot
->path
= btrfs_alloc_path();
669 if (!pending_snapshot
->root_item
|| !pending_snapshot
->path
) {
674 atomic_inc(&root
->will_be_snapshoted
);
675 smp_mb__after_atomic();
676 btrfs_wait_for_no_snapshoting_writes(root
);
678 ret
= btrfs_start_delalloc_inodes(root
, 0);
682 btrfs_wait_ordered_extents(root
, -1);
684 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
685 BTRFS_BLOCK_RSV_TEMP
);
687 * 1 - parent dir inode
690 * 2 - root ref/backref
691 * 1 - root of snapshot
694 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
695 &pending_snapshot
->block_rsv
, 8,
696 &pending_snapshot
->qgroup_reserved
,
701 pending_snapshot
->dentry
= dentry
;
702 pending_snapshot
->root
= root
;
703 pending_snapshot
->readonly
= readonly
;
704 pending_snapshot
->dir
= dir
;
705 pending_snapshot
->inherit
= inherit
;
707 trans
= btrfs_start_transaction(root
, 0);
709 ret
= PTR_ERR(trans
);
713 spin_lock(&root
->fs_info
->trans_lock
);
714 list_add(&pending_snapshot
->list
,
715 &trans
->transaction
->pending_snapshots
);
716 spin_unlock(&root
->fs_info
->trans_lock
);
718 *async_transid
= trans
->transid
;
719 ret
= btrfs_commit_transaction_async(trans
,
720 root
->fs_info
->extent_root
, 1);
722 ret
= btrfs_commit_transaction(trans
, root
);
724 ret
= btrfs_commit_transaction(trans
,
725 root
->fs_info
->extent_root
);
730 ret
= pending_snapshot
->error
;
734 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
738 inode
= btrfs_lookup_dentry(d_inode(dentry
->d_parent
), dentry
);
740 ret
= PTR_ERR(inode
);
744 d_instantiate(dentry
, inode
);
747 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
748 &pending_snapshot
->block_rsv
,
749 pending_snapshot
->qgroup_reserved
);
751 if (atomic_dec_and_test(&root
->will_be_snapshoted
))
752 wake_up_atomic_t(&root
->will_be_snapshoted
);
754 kfree(pending_snapshot
->root_item
);
755 btrfs_free_path(pending_snapshot
->path
);
756 kfree(pending_snapshot
);
761 /* copy of may_delete in fs/namei.c()
762 * Check whether we can remove a link victim from directory dir, check
763 * whether the type of victim is right.
764 * 1. We can't do it if dir is read-only (done in permission())
765 * 2. We should have write and exec permissions on dir
766 * 3. We can't remove anything from append-only dir
767 * 4. We can't do anything with immutable dir (done in permission())
768 * 5. If the sticky bit on dir is set we should either
769 * a. be owner of dir, or
770 * b. be owner of victim, or
771 * c. have CAP_FOWNER capability
772 * 6. If the victim is append-only or immutable we can't do antyhing with
773 * links pointing to it.
774 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
775 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
776 * 9. We can't remove a root or mountpoint.
777 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
778 * nfs_async_unlink().
781 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
785 if (d_really_is_negative(victim
))
788 BUG_ON(d_inode(victim
->d_parent
) != dir
);
789 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
791 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
796 if (check_sticky(dir
, d_inode(victim
)) || IS_APPEND(d_inode(victim
)) ||
797 IS_IMMUTABLE(d_inode(victim
)) || IS_SWAPFILE(d_inode(victim
)))
800 if (!d_is_dir(victim
))
804 } else if (d_is_dir(victim
))
808 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
813 /* copy of may_create in fs/namei.c() */
814 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
816 if (d_really_is_positive(child
))
820 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
824 * Create a new subvolume below @parent. This is largely modeled after
825 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
826 * inside this filesystem so it's quite a bit simpler.
828 static noinline
int btrfs_mksubvol(struct path
*parent
,
829 char *name
, int namelen
,
830 struct btrfs_root
*snap_src
,
831 u64
*async_transid
, bool readonly
,
832 struct btrfs_qgroup_inherit
*inherit
)
834 struct inode
*dir
= d_inode(parent
->dentry
);
835 struct dentry
*dentry
;
838 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
842 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
843 error
= PTR_ERR(dentry
);
847 error
= btrfs_may_create(dir
, dentry
);
852 * even if this name doesn't exist, we may get hash collisions.
853 * check for them now when we can safely fail
855 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
861 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
863 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
867 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
868 async_transid
, readonly
, inherit
);
870 error
= create_subvol(dir
, dentry
, name
, namelen
,
871 async_transid
, inherit
);
874 fsnotify_mkdir(dir
, dentry
);
876 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
885 * When we're defragging a range, we don't want to kick it off again
886 * if it is really just waiting for delalloc to send it down.
887 * If we find a nice big extent or delalloc range for the bytes in the
888 * file you want to defrag, we return 0 to let you know to skip this
891 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, u32 thresh
)
893 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
894 struct extent_map
*em
= NULL
;
895 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
898 read_lock(&em_tree
->lock
);
899 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
900 read_unlock(&em_tree
->lock
);
903 end
= extent_map_end(em
);
905 if (end
- offset
> thresh
)
908 /* if we already have a nice delalloc here, just stop */
910 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
911 thresh
, EXTENT_DELALLOC
, 1);
918 * helper function to walk through a file and find extents
919 * newer than a specific transid, and smaller than thresh.
921 * This is used by the defragging code to find new and small
924 static int find_new_extents(struct btrfs_root
*root
,
925 struct inode
*inode
, u64 newer_than
,
926 u64
*off
, u32 thresh
)
928 struct btrfs_path
*path
;
929 struct btrfs_key min_key
;
930 struct extent_buffer
*leaf
;
931 struct btrfs_file_extent_item
*extent
;
934 u64 ino
= btrfs_ino(inode
);
936 path
= btrfs_alloc_path();
940 min_key
.objectid
= ino
;
941 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
942 min_key
.offset
= *off
;
945 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
949 if (min_key
.objectid
!= ino
)
951 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
954 leaf
= path
->nodes
[0];
955 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
956 struct btrfs_file_extent_item
);
958 type
= btrfs_file_extent_type(leaf
, extent
);
959 if (type
== BTRFS_FILE_EXTENT_REG
&&
960 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
961 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
962 *off
= min_key
.offset
;
963 btrfs_free_path(path
);
968 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
969 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
973 if (min_key
.offset
== (u64
)-1)
977 btrfs_release_path(path
);
980 btrfs_free_path(path
);
984 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
986 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
987 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
988 struct extent_map
*em
;
989 u64 len
= PAGE_CACHE_SIZE
;
992 * hopefully we have this extent in the tree already, try without
993 * the full extent lock
995 read_lock(&em_tree
->lock
);
996 em
= lookup_extent_mapping(em_tree
, start
, len
);
997 read_unlock(&em_tree
->lock
);
1000 struct extent_state
*cached
= NULL
;
1001 u64 end
= start
+ len
- 1;
1003 /* get the big lock and read metadata off disk */
1004 lock_extent_bits(io_tree
, start
, end
, &cached
);
1005 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
1006 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1015 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1017 struct extent_map
*next
;
1020 /* this is the last extent */
1021 if (em
->start
+ em
->len
>= i_size_read(inode
))
1024 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1025 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
1027 else if ((em
->block_start
+ em
->block_len
== next
->block_start
) &&
1028 (em
->block_len
> SZ_128K
&& next
->block_len
> SZ_128K
))
1031 free_extent_map(next
);
1035 static int should_defrag_range(struct inode
*inode
, u64 start
, u32 thresh
,
1036 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1039 struct extent_map
*em
;
1041 bool next_mergeable
= true;
1042 bool prev_mergeable
= true;
1045 * make sure that once we start defragging an extent, we keep on
1048 if (start
< *defrag_end
)
1053 em
= defrag_lookup_extent(inode
, start
);
1057 /* this will cover holes, and inline extents */
1058 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1064 prev_mergeable
= false;
1066 next_mergeable
= defrag_check_next_extent(inode
, em
);
1068 * we hit a real extent, if it is big or the next extent is not a
1069 * real extent, don't bother defragging it
1071 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1072 (em
->len
>= thresh
|| (!next_mergeable
&& !prev_mergeable
)))
1076 * last_len ends up being a counter of how many bytes we've defragged.
1077 * every time we choose not to defrag an extent, we reset *last_len
1078 * so that the next tiny extent will force a defrag.
1080 * The end result of this is that tiny extents before a single big
1081 * extent will force at least part of that big extent to be defragged.
1084 *defrag_end
= extent_map_end(em
);
1087 *skip
= extent_map_end(em
);
1091 free_extent_map(em
);
1096 * it doesn't do much good to defrag one or two pages
1097 * at a time. This pulls in a nice chunk of pages
1098 * to COW and defrag.
1100 * It also makes sure the delalloc code has enough
1101 * dirty data to avoid making new small extents as part
1104 * It's a good idea to start RA on this range
1105 * before calling this.
1107 static int cluster_pages_for_defrag(struct inode
*inode
,
1108 struct page
**pages
,
1109 unsigned long start_index
,
1110 unsigned long num_pages
)
1112 unsigned long file_end
;
1113 u64 isize
= i_size_read(inode
);
1120 struct btrfs_ordered_extent
*ordered
;
1121 struct extent_state
*cached_state
= NULL
;
1122 struct extent_io_tree
*tree
;
1123 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1125 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1126 if (!isize
|| start_index
> file_end
)
1129 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1131 ret
= btrfs_delalloc_reserve_space(inode
,
1132 start_index
<< PAGE_CACHE_SHIFT
,
1133 page_cnt
<< PAGE_CACHE_SHIFT
);
1137 tree
= &BTRFS_I(inode
)->io_tree
;
1139 /* step one, lock all the pages */
1140 for (i
= 0; i
< page_cnt
; i
++) {
1143 page
= find_or_create_page(inode
->i_mapping
,
1144 start_index
+ i
, mask
);
1148 page_start
= page_offset(page
);
1149 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1151 lock_extent_bits(tree
, page_start
, page_end
,
1153 ordered
= btrfs_lookup_ordered_extent(inode
,
1155 unlock_extent_cached(tree
, page_start
, page_end
,
1156 &cached_state
, GFP_NOFS
);
1161 btrfs_start_ordered_extent(inode
, ordered
, 1);
1162 btrfs_put_ordered_extent(ordered
);
1165 * we unlocked the page above, so we need check if
1166 * it was released or not.
1168 if (page
->mapping
!= inode
->i_mapping
) {
1170 page_cache_release(page
);
1175 if (!PageUptodate(page
)) {
1176 btrfs_readpage(NULL
, page
);
1178 if (!PageUptodate(page
)) {
1180 page_cache_release(page
);
1186 if (page
->mapping
!= inode
->i_mapping
) {
1188 page_cache_release(page
);
1198 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1202 * so now we have a nice long stream of locked
1203 * and up to date pages, lets wait on them
1205 for (i
= 0; i
< i_done
; i
++)
1206 wait_on_page_writeback(pages
[i
]);
1208 page_start
= page_offset(pages
[0]);
1209 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1211 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1212 page_start
, page_end
- 1, &cached_state
);
1213 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1214 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1215 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1216 &cached_state
, GFP_NOFS
);
1218 if (i_done
!= page_cnt
) {
1219 spin_lock(&BTRFS_I(inode
)->lock
);
1220 BTRFS_I(inode
)->outstanding_extents
++;
1221 spin_unlock(&BTRFS_I(inode
)->lock
);
1222 btrfs_delalloc_release_space(inode
,
1223 start_index
<< PAGE_CACHE_SHIFT
,
1224 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1228 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1229 &cached_state
, GFP_NOFS
);
1231 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1232 page_start
, page_end
- 1, &cached_state
,
1235 for (i
= 0; i
< i_done
; i
++) {
1236 clear_page_dirty_for_io(pages
[i
]);
1237 ClearPageChecked(pages
[i
]);
1238 set_page_extent_mapped(pages
[i
]);
1239 set_page_dirty(pages
[i
]);
1240 unlock_page(pages
[i
]);
1241 page_cache_release(pages
[i
]);
1245 for (i
= 0; i
< i_done
; i
++) {
1246 unlock_page(pages
[i
]);
1247 page_cache_release(pages
[i
]);
1249 btrfs_delalloc_release_space(inode
,
1250 start_index
<< PAGE_CACHE_SHIFT
,
1251 page_cnt
<< PAGE_CACHE_SHIFT
);
1256 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1257 struct btrfs_ioctl_defrag_range_args
*range
,
1258 u64 newer_than
, unsigned long max_to_defrag
)
1260 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1261 struct file_ra_state
*ra
= NULL
;
1262 unsigned long last_index
;
1263 u64 isize
= i_size_read(inode
);
1267 u64 newer_off
= range
->start
;
1269 unsigned long ra_index
= 0;
1271 int defrag_count
= 0;
1272 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1273 u32 extent_thresh
= range
->extent_thresh
;
1274 unsigned long max_cluster
= SZ_256K
>> PAGE_CACHE_SHIFT
;
1275 unsigned long cluster
= max_cluster
;
1276 u64 new_align
= ~((u64
)SZ_128K
- 1);
1277 struct page
**pages
= NULL
;
1282 if (range
->start
>= isize
)
1285 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1286 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1288 if (range
->compress_type
)
1289 compress_type
= range
->compress_type
;
1292 if (extent_thresh
== 0)
1293 extent_thresh
= SZ_256K
;
1296 * if we were not given a file, allocate a readahead
1300 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1303 file_ra_state_init(ra
, inode
->i_mapping
);
1308 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1315 /* find the last page to defrag */
1316 if (range
->start
+ range
->len
> range
->start
) {
1317 last_index
= min_t(u64
, isize
- 1,
1318 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1320 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1324 ret
= find_new_extents(root
, inode
, newer_than
,
1325 &newer_off
, SZ_64K
);
1327 range
->start
= newer_off
;
1329 * we always align our defrag to help keep
1330 * the extents in the file evenly spaced
1332 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1336 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1339 max_to_defrag
= last_index
- i
+ 1;
1342 * make writeback starts from i, so the defrag range can be
1343 * written sequentially.
1345 if (i
< inode
->i_mapping
->writeback_index
)
1346 inode
->i_mapping
->writeback_index
= i
;
1348 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1349 (i
< DIV_ROUND_UP(i_size_read(inode
), PAGE_CACHE_SIZE
))) {
1351 * make sure we stop running if someone unmounts
1354 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1357 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1358 btrfs_debug(root
->fs_info
, "defrag_file cancelled");
1363 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1364 extent_thresh
, &last_len
, &skip
,
1365 &defrag_end
, range
->flags
&
1366 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1369 * the should_defrag function tells us how much to skip
1370 * bump our counter by the suggested amount
1372 next
= DIV_ROUND_UP(skip
, PAGE_CACHE_SIZE
);
1373 i
= max(i
+ 1, next
);
1378 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1379 PAGE_CACHE_SHIFT
) - i
;
1380 cluster
= min(cluster
, max_cluster
);
1382 cluster
= max_cluster
;
1385 if (i
+ cluster
> ra_index
) {
1386 ra_index
= max(i
, ra_index
);
1387 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1389 ra_index
+= cluster
;
1393 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1394 BTRFS_I(inode
)->force_compress
= compress_type
;
1395 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1397 inode_unlock(inode
);
1401 defrag_count
+= ret
;
1402 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1403 inode_unlock(inode
);
1406 if (newer_off
== (u64
)-1)
1412 newer_off
= max(newer_off
+ 1,
1413 (u64
)i
<< PAGE_CACHE_SHIFT
);
1415 ret
= find_new_extents(root
, inode
, newer_than
,
1416 &newer_off
, SZ_64K
);
1418 range
->start
= newer_off
;
1419 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1426 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1434 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1435 filemap_flush(inode
->i_mapping
);
1436 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1437 &BTRFS_I(inode
)->runtime_flags
))
1438 filemap_flush(inode
->i_mapping
);
1441 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1442 /* the filemap_flush will queue IO into the worker threads, but
1443 * we have to make sure the IO is actually started and that
1444 * ordered extents get created before we return
1446 atomic_inc(&root
->fs_info
->async_submit_draining
);
1447 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1448 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1449 wait_event(root
->fs_info
->async_submit_wait
,
1450 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1451 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1453 atomic_dec(&root
->fs_info
->async_submit_draining
);
1456 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1457 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1463 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1465 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1466 inode_unlock(inode
);
1474 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1480 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1481 struct btrfs_ioctl_vol_args
*vol_args
;
1482 struct btrfs_trans_handle
*trans
;
1483 struct btrfs_device
*device
= NULL
;
1486 char *devstr
= NULL
;
1490 if (!capable(CAP_SYS_ADMIN
))
1493 ret
= mnt_want_write_file(file
);
1497 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1499 mnt_drop_write_file(file
);
1500 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1503 mutex_lock(&root
->fs_info
->volume_mutex
);
1504 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1505 if (IS_ERR(vol_args
)) {
1506 ret
= PTR_ERR(vol_args
);
1510 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1512 sizestr
= vol_args
->name
;
1513 devstr
= strchr(sizestr
, ':');
1515 sizestr
= devstr
+ 1;
1517 devstr
= vol_args
->name
;
1518 ret
= kstrtoull(devstr
, 10, &devid
);
1525 btrfs_info(root
->fs_info
, "resizing devid %llu", devid
);
1528 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1530 btrfs_info(root
->fs_info
, "resizer unable to find device %llu",
1536 if (!device
->writeable
) {
1537 btrfs_info(root
->fs_info
,
1538 "resizer unable to apply on readonly device %llu",
1544 if (!strcmp(sizestr
, "max"))
1545 new_size
= device
->bdev
->bd_inode
->i_size
;
1547 if (sizestr
[0] == '-') {
1550 } else if (sizestr
[0] == '+') {
1554 new_size
= memparse(sizestr
, &retptr
);
1555 if (*retptr
!= '\0' || new_size
== 0) {
1561 if (device
->is_tgtdev_for_dev_replace
) {
1566 old_size
= btrfs_device_get_total_bytes(device
);
1569 if (new_size
> old_size
) {
1573 new_size
= old_size
- new_size
;
1574 } else if (mod
> 0) {
1575 if (new_size
> ULLONG_MAX
- old_size
) {
1579 new_size
= old_size
+ new_size
;
1582 if (new_size
< SZ_256M
) {
1586 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1591 new_size
= div_u64(new_size
, root
->sectorsize
);
1592 new_size
*= root
->sectorsize
;
1594 btrfs_info_in_rcu(root
->fs_info
, "new size for %s is %llu",
1595 rcu_str_deref(device
->name
), new_size
);
1597 if (new_size
> old_size
) {
1598 trans
= btrfs_start_transaction(root
, 0);
1599 if (IS_ERR(trans
)) {
1600 ret
= PTR_ERR(trans
);
1603 ret
= btrfs_grow_device(trans
, device
, new_size
);
1604 btrfs_commit_transaction(trans
, root
);
1605 } else if (new_size
< old_size
) {
1606 ret
= btrfs_shrink_device(device
, new_size
);
1607 } /* equal, nothing need to do */
1612 mutex_unlock(&root
->fs_info
->volume_mutex
);
1613 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1614 mnt_drop_write_file(file
);
1618 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1619 char *name
, unsigned long fd
, int subvol
,
1620 u64
*transid
, bool readonly
,
1621 struct btrfs_qgroup_inherit
*inherit
)
1626 ret
= mnt_want_write_file(file
);
1630 namelen
= strlen(name
);
1631 if (strchr(name
, '/')) {
1633 goto out_drop_write
;
1636 if (name
[0] == '.' &&
1637 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1639 goto out_drop_write
;
1643 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1644 NULL
, transid
, readonly
, inherit
);
1646 struct fd src
= fdget(fd
);
1647 struct inode
*src_inode
;
1650 goto out_drop_write
;
1653 src_inode
= file_inode(src
.file
);
1654 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1655 btrfs_info(BTRFS_I(src_inode
)->root
->fs_info
,
1656 "Snapshot src from another FS");
1658 } else if (!inode_owner_or_capable(src_inode
)) {
1660 * Subvolume creation is not restricted, but snapshots
1661 * are limited to own subvolumes only
1665 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1666 BTRFS_I(src_inode
)->root
,
1667 transid
, readonly
, inherit
);
1672 mnt_drop_write_file(file
);
1677 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1678 void __user
*arg
, int subvol
)
1680 struct btrfs_ioctl_vol_args
*vol_args
;
1683 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1684 if (IS_ERR(vol_args
))
1685 return PTR_ERR(vol_args
);
1686 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1688 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1689 vol_args
->fd
, subvol
,
1696 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1697 void __user
*arg
, int subvol
)
1699 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1703 bool readonly
= false;
1704 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1706 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1707 if (IS_ERR(vol_args
))
1708 return PTR_ERR(vol_args
);
1709 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1711 if (vol_args
->flags
&
1712 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1713 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1718 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1720 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1722 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1723 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1727 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1728 if (IS_ERR(inherit
)) {
1729 ret
= PTR_ERR(inherit
);
1734 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1735 vol_args
->fd
, subvol
, ptr
,
1740 if (ptr
&& copy_to_user(arg
+
1741 offsetof(struct btrfs_ioctl_vol_args_v2
,
1753 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1756 struct inode
*inode
= file_inode(file
);
1757 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1761 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1764 down_read(&root
->fs_info
->subvol_sem
);
1765 if (btrfs_root_readonly(root
))
1766 flags
|= BTRFS_SUBVOL_RDONLY
;
1767 up_read(&root
->fs_info
->subvol_sem
);
1769 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1775 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1778 struct inode
*inode
= file_inode(file
);
1779 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1780 struct btrfs_trans_handle
*trans
;
1785 if (!inode_owner_or_capable(inode
))
1788 ret
= mnt_want_write_file(file
);
1792 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1794 goto out_drop_write
;
1797 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1799 goto out_drop_write
;
1802 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1804 goto out_drop_write
;
1807 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1809 goto out_drop_write
;
1812 down_write(&root
->fs_info
->subvol_sem
);
1815 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1818 root_flags
= btrfs_root_flags(&root
->root_item
);
1819 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1820 btrfs_set_root_flags(&root
->root_item
,
1821 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1824 * Block RO -> RW transition if this subvolume is involved in
1827 spin_lock(&root
->root_item_lock
);
1828 if (root
->send_in_progress
== 0) {
1829 btrfs_set_root_flags(&root
->root_item
,
1830 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1831 spin_unlock(&root
->root_item_lock
);
1833 spin_unlock(&root
->root_item_lock
);
1834 btrfs_warn(root
->fs_info
,
1835 "Attempt to set subvolume %llu read-write during send",
1836 root
->root_key
.objectid
);
1842 trans
= btrfs_start_transaction(root
, 1);
1843 if (IS_ERR(trans
)) {
1844 ret
= PTR_ERR(trans
);
1848 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1849 &root
->root_key
, &root
->root_item
);
1851 btrfs_commit_transaction(trans
, root
);
1854 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1856 up_write(&root
->fs_info
->subvol_sem
);
1858 mnt_drop_write_file(file
);
1864 * helper to check if the subvolume references other subvolumes
1866 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1868 struct btrfs_path
*path
;
1869 struct btrfs_dir_item
*di
;
1870 struct btrfs_key key
;
1874 path
= btrfs_alloc_path();
1878 /* Make sure this root isn't set as the default subvol */
1879 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
1880 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
, path
,
1881 dir_id
, "default", 7, 0);
1882 if (di
&& !IS_ERR(di
)) {
1883 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1884 if (key
.objectid
== root
->root_key
.objectid
) {
1886 btrfs_err(root
->fs_info
, "deleting default subvolume "
1887 "%llu is not allowed", key
.objectid
);
1890 btrfs_release_path(path
);
1893 key
.objectid
= root
->root_key
.objectid
;
1894 key
.type
= BTRFS_ROOT_REF_KEY
;
1895 key
.offset
= (u64
)-1;
1897 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1904 if (path
->slots
[0] > 0) {
1906 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1907 if (key
.objectid
== root
->root_key
.objectid
&&
1908 key
.type
== BTRFS_ROOT_REF_KEY
)
1912 btrfs_free_path(path
);
1916 static noinline
int key_in_sk(struct btrfs_key
*key
,
1917 struct btrfs_ioctl_search_key
*sk
)
1919 struct btrfs_key test
;
1922 test
.objectid
= sk
->min_objectid
;
1923 test
.type
= sk
->min_type
;
1924 test
.offset
= sk
->min_offset
;
1926 ret
= btrfs_comp_cpu_keys(key
, &test
);
1930 test
.objectid
= sk
->max_objectid
;
1931 test
.type
= sk
->max_type
;
1932 test
.offset
= sk
->max_offset
;
1934 ret
= btrfs_comp_cpu_keys(key
, &test
);
1940 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1941 struct btrfs_path
*path
,
1942 struct btrfs_key
*key
,
1943 struct btrfs_ioctl_search_key
*sk
,
1946 unsigned long *sk_offset
,
1950 struct extent_buffer
*leaf
;
1951 struct btrfs_ioctl_search_header sh
;
1952 struct btrfs_key test
;
1953 unsigned long item_off
;
1954 unsigned long item_len
;
1960 leaf
= path
->nodes
[0];
1961 slot
= path
->slots
[0];
1962 nritems
= btrfs_header_nritems(leaf
);
1964 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1968 found_transid
= btrfs_header_generation(leaf
);
1970 for (i
= slot
; i
< nritems
; i
++) {
1971 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1972 item_len
= btrfs_item_size_nr(leaf
, i
);
1974 btrfs_item_key_to_cpu(leaf
, key
, i
);
1975 if (!key_in_sk(key
, sk
))
1978 if (sizeof(sh
) + item_len
> *buf_size
) {
1985 * return one empty item back for v1, which does not
1989 *buf_size
= sizeof(sh
) + item_len
;
1994 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
1999 sh
.objectid
= key
->objectid
;
2000 sh
.offset
= key
->offset
;
2001 sh
.type
= key
->type
;
2003 sh
.transid
= found_transid
;
2005 /* copy search result header */
2006 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2011 *sk_offset
+= sizeof(sh
);
2014 char __user
*up
= ubuf
+ *sk_offset
;
2016 if (read_extent_buffer_to_user(leaf
, up
,
2017 item_off
, item_len
)) {
2022 *sk_offset
+= item_len
;
2026 if (ret
) /* -EOVERFLOW from above */
2029 if (*num_found
>= sk
->nr_items
) {
2036 test
.objectid
= sk
->max_objectid
;
2037 test
.type
= sk
->max_type
;
2038 test
.offset
= sk
->max_offset
;
2039 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
2041 else if (key
->offset
< (u64
)-1)
2043 else if (key
->type
< (u8
)-1) {
2046 } else if (key
->objectid
< (u64
)-1) {
2054 * 0: all items from this leaf copied, continue with next
2055 * 1: * more items can be copied, but unused buffer is too small
2056 * * all items were found
2057 * Either way, it will stops the loop which iterates to the next
2059 * -EOVERFLOW: item was to large for buffer
2060 * -EFAULT: could not copy extent buffer back to userspace
2065 static noinline
int search_ioctl(struct inode
*inode
,
2066 struct btrfs_ioctl_search_key
*sk
,
2070 struct btrfs_root
*root
;
2071 struct btrfs_key key
;
2072 struct btrfs_path
*path
;
2073 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
2076 unsigned long sk_offset
= 0;
2078 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2079 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2083 path
= btrfs_alloc_path();
2087 if (sk
->tree_id
== 0) {
2088 /* search the root of the inode that was passed */
2089 root
= BTRFS_I(inode
)->root
;
2091 key
.objectid
= sk
->tree_id
;
2092 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2093 key
.offset
= (u64
)-1;
2094 root
= btrfs_read_fs_root_no_name(info
, &key
);
2096 btrfs_free_path(path
);
2101 key
.objectid
= sk
->min_objectid
;
2102 key
.type
= sk
->min_type
;
2103 key
.offset
= sk
->min_offset
;
2106 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2112 ret
= copy_to_sk(root
, path
, &key
, sk
, buf_size
, ubuf
,
2113 &sk_offset
, &num_found
);
2114 btrfs_release_path(path
);
2122 sk
->nr_items
= num_found
;
2123 btrfs_free_path(path
);
2127 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2130 struct btrfs_ioctl_search_args __user
*uargs
;
2131 struct btrfs_ioctl_search_key sk
;
2132 struct inode
*inode
;
2136 if (!capable(CAP_SYS_ADMIN
))
2139 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2141 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2144 buf_size
= sizeof(uargs
->buf
);
2146 inode
= file_inode(file
);
2147 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2150 * In the origin implementation an overflow is handled by returning a
2151 * search header with a len of zero, so reset ret.
2153 if (ret
== -EOVERFLOW
)
2156 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2161 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2164 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2165 struct btrfs_ioctl_search_args_v2 args
;
2166 struct inode
*inode
;
2169 const size_t buf_limit
= SZ_16M
;
2171 if (!capable(CAP_SYS_ADMIN
))
2174 /* copy search header and buffer size */
2175 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2176 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2179 buf_size
= args
.buf_size
;
2181 if (buf_size
< sizeof(struct btrfs_ioctl_search_header
))
2184 /* limit result size to 16MB */
2185 if (buf_size
> buf_limit
)
2186 buf_size
= buf_limit
;
2188 inode
= file_inode(file
);
2189 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2190 (char *)(&uarg
->buf
[0]));
2191 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2193 else if (ret
== -EOVERFLOW
&&
2194 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2201 * Search INODE_REFs to identify path name of 'dirid' directory
2202 * in a 'tree_id' tree. and sets path name to 'name'.
2204 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2205 u64 tree_id
, u64 dirid
, char *name
)
2207 struct btrfs_root
*root
;
2208 struct btrfs_key key
;
2214 struct btrfs_inode_ref
*iref
;
2215 struct extent_buffer
*l
;
2216 struct btrfs_path
*path
;
2218 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2223 path
= btrfs_alloc_path();
2227 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2229 key
.objectid
= tree_id
;
2230 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2231 key
.offset
= (u64
)-1;
2232 root
= btrfs_read_fs_root_no_name(info
, &key
);
2234 btrfs_err(info
, "could not find root %llu", tree_id
);
2239 key
.objectid
= dirid
;
2240 key
.type
= BTRFS_INODE_REF_KEY
;
2241 key
.offset
= (u64
)-1;
2244 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2248 ret
= btrfs_previous_item(root
, path
, dirid
,
2249 BTRFS_INODE_REF_KEY
);
2259 slot
= path
->slots
[0];
2260 btrfs_item_key_to_cpu(l
, &key
, slot
);
2262 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2263 len
= btrfs_inode_ref_name_len(l
, iref
);
2265 total_len
+= len
+ 1;
2267 ret
= -ENAMETOOLONG
;
2272 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2274 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2277 btrfs_release_path(path
);
2278 key
.objectid
= key
.offset
;
2279 key
.offset
= (u64
)-1;
2280 dirid
= key
.objectid
;
2282 memmove(name
, ptr
, total_len
);
2283 name
[total_len
] = '\0';
2286 btrfs_free_path(path
);
2290 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2293 struct btrfs_ioctl_ino_lookup_args
*args
;
2294 struct inode
*inode
;
2297 args
= memdup_user(argp
, sizeof(*args
));
2299 return PTR_ERR(args
);
2301 inode
= file_inode(file
);
2304 * Unprivileged query to obtain the containing subvolume root id. The
2305 * path is reset so it's consistent with btrfs_search_path_in_tree.
2307 if (args
->treeid
== 0)
2308 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2310 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2315 if (!capable(CAP_SYS_ADMIN
)) {
2320 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2321 args
->treeid
, args
->objectid
,
2325 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2332 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2335 struct dentry
*parent
= file
->f_path
.dentry
;
2336 struct dentry
*dentry
;
2337 struct inode
*dir
= d_inode(parent
);
2338 struct inode
*inode
;
2339 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2340 struct btrfs_root
*dest
= NULL
;
2341 struct btrfs_ioctl_vol_args
*vol_args
;
2342 struct btrfs_trans_handle
*trans
;
2343 struct btrfs_block_rsv block_rsv
;
2345 u64 qgroup_reserved
;
2350 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2351 if (IS_ERR(vol_args
))
2352 return PTR_ERR(vol_args
);
2354 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2355 namelen
= strlen(vol_args
->name
);
2356 if (strchr(vol_args
->name
, '/') ||
2357 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2362 err
= mnt_want_write_file(file
);
2367 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2369 goto out_drop_write
;
2370 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2371 if (IS_ERR(dentry
)) {
2372 err
= PTR_ERR(dentry
);
2373 goto out_unlock_dir
;
2376 if (d_really_is_negative(dentry
)) {
2381 inode
= d_inode(dentry
);
2382 dest
= BTRFS_I(inode
)->root
;
2383 if (!capable(CAP_SYS_ADMIN
)) {
2385 * Regular user. Only allow this with a special mount
2386 * option, when the user has write+exec access to the
2387 * subvol root, and when rmdir(2) would have been
2390 * Note that this is _not_ check that the subvol is
2391 * empty or doesn't contain data that we wouldn't
2392 * otherwise be able to delete.
2394 * Users who want to delete empty subvols should try
2398 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2402 * Do not allow deletion if the parent dir is the same
2403 * as the dir to be deleted. That means the ioctl
2404 * must be called on the dentry referencing the root
2405 * of the subvol, not a random directory contained
2412 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2417 /* check if subvolume may be deleted by a user */
2418 err
= btrfs_may_delete(dir
, dentry
, 1);
2422 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2430 * Don't allow to delete a subvolume with send in progress. This is
2431 * inside the i_mutex so the error handling that has to drop the bit
2432 * again is not run concurrently.
2434 spin_lock(&dest
->root_item_lock
);
2435 root_flags
= btrfs_root_flags(&dest
->root_item
);
2436 if (dest
->send_in_progress
== 0) {
2437 btrfs_set_root_flags(&dest
->root_item
,
2438 root_flags
| BTRFS_ROOT_SUBVOL_DEAD
);
2439 spin_unlock(&dest
->root_item_lock
);
2441 spin_unlock(&dest
->root_item_lock
);
2442 btrfs_warn(root
->fs_info
,
2443 "Attempt to delete subvolume %llu during send",
2444 dest
->root_key
.objectid
);
2446 goto out_unlock_inode
;
2449 down_write(&root
->fs_info
->subvol_sem
);
2451 err
= may_destroy_subvol(dest
);
2455 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2457 * One for dir inode, two for dir entries, two for root
2460 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2461 5, &qgroup_reserved
, true);
2465 trans
= btrfs_start_transaction(root
, 0);
2466 if (IS_ERR(trans
)) {
2467 err
= PTR_ERR(trans
);
2470 trans
->block_rsv
= &block_rsv
;
2471 trans
->bytes_reserved
= block_rsv
.size
;
2473 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2474 dest
->root_key
.objectid
,
2475 dentry
->d_name
.name
,
2476 dentry
->d_name
.len
);
2479 btrfs_abort_transaction(trans
, root
, ret
);
2483 btrfs_record_root_in_trans(trans
, dest
);
2485 memset(&dest
->root_item
.drop_progress
, 0,
2486 sizeof(dest
->root_item
.drop_progress
));
2487 dest
->root_item
.drop_level
= 0;
2488 btrfs_set_root_refs(&dest
->root_item
, 0);
2490 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED
, &dest
->state
)) {
2491 ret
= btrfs_insert_orphan_item(trans
,
2492 root
->fs_info
->tree_root
,
2493 dest
->root_key
.objectid
);
2495 btrfs_abort_transaction(trans
, root
, ret
);
2501 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2502 dest
->root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
2503 dest
->root_key
.objectid
);
2504 if (ret
&& ret
!= -ENOENT
) {
2505 btrfs_abort_transaction(trans
, root
, ret
);
2509 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2510 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2511 dest
->root_item
.received_uuid
,
2512 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2513 dest
->root_key
.objectid
);
2514 if (ret
&& ret
!= -ENOENT
) {
2515 btrfs_abort_transaction(trans
, root
, ret
);
2522 trans
->block_rsv
= NULL
;
2523 trans
->bytes_reserved
= 0;
2524 ret
= btrfs_end_transaction(trans
, root
);
2527 inode
->i_flags
|= S_DEAD
;
2529 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2531 up_write(&root
->fs_info
->subvol_sem
);
2533 spin_lock(&dest
->root_item_lock
);
2534 root_flags
= btrfs_root_flags(&dest
->root_item
);
2535 btrfs_set_root_flags(&dest
->root_item
,
2536 root_flags
& ~BTRFS_ROOT_SUBVOL_DEAD
);
2537 spin_unlock(&dest
->root_item_lock
);
2540 inode_unlock(inode
);
2542 d_invalidate(dentry
);
2543 btrfs_invalidate_inodes(dest
);
2545 ASSERT(dest
->send_in_progress
== 0);
2548 if (dest
->ino_cache_inode
) {
2549 iput(dest
->ino_cache_inode
);
2550 dest
->ino_cache_inode
= NULL
;
2558 mnt_drop_write_file(file
);
2564 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2566 struct inode
*inode
= file_inode(file
);
2567 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2568 struct btrfs_ioctl_defrag_range_args
*range
;
2571 ret
= mnt_want_write_file(file
);
2575 if (btrfs_root_readonly(root
)) {
2580 switch (inode
->i_mode
& S_IFMT
) {
2582 if (!capable(CAP_SYS_ADMIN
)) {
2586 ret
= btrfs_defrag_root(root
);
2589 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2592 if (!(file
->f_mode
& FMODE_WRITE
)) {
2597 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2604 if (copy_from_user(range
, argp
,
2610 /* compression requires us to start the IO */
2611 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2612 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2613 range
->extent_thresh
= (u32
)-1;
2616 /* the rest are all set to zero by kzalloc */
2617 range
->len
= (u64
)-1;
2619 ret
= btrfs_defrag_file(file_inode(file
), file
,
2629 mnt_drop_write_file(file
);
2633 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2635 struct btrfs_ioctl_vol_args
*vol_args
;
2638 if (!capable(CAP_SYS_ADMIN
))
2641 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2643 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2646 mutex_lock(&root
->fs_info
->volume_mutex
);
2647 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2648 if (IS_ERR(vol_args
)) {
2649 ret
= PTR_ERR(vol_args
);
2653 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2654 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2657 btrfs_info(root
->fs_info
, "disk added %s",vol_args
->name
);
2661 mutex_unlock(&root
->fs_info
->volume_mutex
);
2662 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2666 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2668 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2669 struct btrfs_ioctl_vol_args
*vol_args
;
2672 if (!capable(CAP_SYS_ADMIN
))
2675 ret
= mnt_want_write_file(file
);
2679 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2680 if (IS_ERR(vol_args
)) {
2681 ret
= PTR_ERR(vol_args
);
2685 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2687 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2689 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2693 mutex_lock(&root
->fs_info
->volume_mutex
);
2694 ret
= btrfs_rm_device(root
, vol_args
->name
);
2695 mutex_unlock(&root
->fs_info
->volume_mutex
);
2696 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2699 btrfs_info(root
->fs_info
, "disk deleted %s",vol_args
->name
);
2704 mnt_drop_write_file(file
);
2708 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2710 struct btrfs_ioctl_fs_info_args
*fi_args
;
2711 struct btrfs_device
*device
;
2712 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2715 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2719 mutex_lock(&fs_devices
->device_list_mutex
);
2720 fi_args
->num_devices
= fs_devices
->num_devices
;
2721 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2723 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
2724 if (device
->devid
> fi_args
->max_id
)
2725 fi_args
->max_id
= device
->devid
;
2727 mutex_unlock(&fs_devices
->device_list_mutex
);
2729 fi_args
->nodesize
= root
->fs_info
->super_copy
->nodesize
;
2730 fi_args
->sectorsize
= root
->fs_info
->super_copy
->sectorsize
;
2731 fi_args
->clone_alignment
= root
->fs_info
->super_copy
->sectorsize
;
2733 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2740 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2742 struct btrfs_ioctl_dev_info_args
*di_args
;
2743 struct btrfs_device
*dev
;
2744 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2746 char *s_uuid
= NULL
;
2748 di_args
= memdup_user(arg
, sizeof(*di_args
));
2749 if (IS_ERR(di_args
))
2750 return PTR_ERR(di_args
);
2752 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2753 s_uuid
= di_args
->uuid
;
2755 mutex_lock(&fs_devices
->device_list_mutex
);
2756 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2763 di_args
->devid
= dev
->devid
;
2764 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
2765 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
2766 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2768 struct rcu_string
*name
;
2771 name
= rcu_dereference(dev
->name
);
2772 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2774 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2776 di_args
->path
[0] = '\0';
2780 mutex_unlock(&fs_devices
->device_list_mutex
);
2781 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2788 static struct page
*extent_same_get_page(struct inode
*inode
, pgoff_t index
)
2792 page
= grab_cache_page(inode
->i_mapping
, index
);
2794 return ERR_PTR(-ENOMEM
);
2796 if (!PageUptodate(page
)) {
2799 ret
= btrfs_readpage(NULL
, page
);
2801 return ERR_PTR(ret
);
2803 if (!PageUptodate(page
)) {
2805 page_cache_release(page
);
2806 return ERR_PTR(-EIO
);
2808 if (page
->mapping
!= inode
->i_mapping
) {
2810 page_cache_release(page
);
2811 return ERR_PTR(-EAGAIN
);
2818 static int gather_extent_pages(struct inode
*inode
, struct page
**pages
,
2819 int num_pages
, u64 off
)
2822 pgoff_t index
= off
>> PAGE_CACHE_SHIFT
;
2824 for (i
= 0; i
< num_pages
; i
++) {
2826 pages
[i
] = extent_same_get_page(inode
, index
+ i
);
2827 if (IS_ERR(pages
[i
])) {
2828 int err
= PTR_ERR(pages
[i
]);
2839 static int lock_extent_range(struct inode
*inode
, u64 off
, u64 len
,
2840 bool retry_range_locking
)
2843 * Do any pending delalloc/csum calculations on inode, one way or
2844 * another, and lock file content.
2845 * The locking order is:
2848 * 2) range in the inode's io tree
2851 struct btrfs_ordered_extent
*ordered
;
2852 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2853 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2856 ordered
->file_offset
+ ordered
->len
<= off
||
2857 ordered
->file_offset
>= off
+ len
) &&
2858 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2859 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
)) {
2861 btrfs_put_ordered_extent(ordered
);
2864 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2866 btrfs_put_ordered_extent(ordered
);
2867 if (!retry_range_locking
)
2869 btrfs_wait_ordered_range(inode
, off
, len
);
2874 static void btrfs_double_inode_unlock(struct inode
*inode1
, struct inode
*inode2
)
2876 inode_unlock(inode1
);
2877 inode_unlock(inode2
);
2880 static void btrfs_double_inode_lock(struct inode
*inode1
, struct inode
*inode2
)
2882 if (inode1
< inode2
)
2883 swap(inode1
, inode2
);
2885 inode_lock_nested(inode1
, I_MUTEX_PARENT
);
2886 inode_lock_nested(inode2
, I_MUTEX_CHILD
);
2889 static void btrfs_double_extent_unlock(struct inode
*inode1
, u64 loff1
,
2890 struct inode
*inode2
, u64 loff2
, u64 len
)
2892 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2893 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2896 static int btrfs_double_extent_lock(struct inode
*inode1
, u64 loff1
,
2897 struct inode
*inode2
, u64 loff2
, u64 len
,
2898 bool retry_range_locking
)
2902 if (inode1
< inode2
) {
2903 swap(inode1
, inode2
);
2906 ret
= lock_extent_range(inode1
, loff1
, len
, retry_range_locking
);
2909 ret
= lock_extent_range(inode2
, loff2
, len
, retry_range_locking
);
2911 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
,
2918 struct page
**src_pages
;
2919 struct page
**dst_pages
;
2922 static void btrfs_cmp_data_free(struct cmp_pages
*cmp
)
2927 for (i
= 0; i
< cmp
->num_pages
; i
++) {
2928 pg
= cmp
->src_pages
[i
];
2931 page_cache_release(pg
);
2933 pg
= cmp
->dst_pages
[i
];
2936 page_cache_release(pg
);
2939 kfree(cmp
->src_pages
);
2940 kfree(cmp
->dst_pages
);
2943 static int btrfs_cmp_data_prepare(struct inode
*src
, u64 loff
,
2944 struct inode
*dst
, u64 dst_loff
,
2945 u64 len
, struct cmp_pages
*cmp
)
2948 int num_pages
= PAGE_CACHE_ALIGN(len
) >> PAGE_CACHE_SHIFT
;
2949 struct page
**src_pgarr
, **dst_pgarr
;
2952 * We must gather up all the pages before we initiate our
2953 * extent locking. We use an array for the page pointers. Size
2954 * of the array is bounded by len, which is in turn bounded by
2955 * BTRFS_MAX_DEDUPE_LEN.
2957 src_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
2958 dst_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
2959 if (!src_pgarr
|| !dst_pgarr
) {
2964 cmp
->num_pages
= num_pages
;
2965 cmp
->src_pages
= src_pgarr
;
2966 cmp
->dst_pages
= dst_pgarr
;
2968 ret
= gather_extent_pages(src
, cmp
->src_pages
, cmp
->num_pages
, loff
);
2972 ret
= gather_extent_pages(dst
, cmp
->dst_pages
, cmp
->num_pages
, dst_loff
);
2976 btrfs_cmp_data_free(cmp
);
2980 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
2981 u64 dst_loff
, u64 len
, struct cmp_pages
*cmp
)
2985 struct page
*src_page
, *dst_page
;
2986 unsigned int cmp_len
= PAGE_CACHE_SIZE
;
2987 void *addr
, *dst_addr
;
2991 if (len
< PAGE_CACHE_SIZE
)
2994 BUG_ON(i
>= cmp
->num_pages
);
2996 src_page
= cmp
->src_pages
[i
];
2997 dst_page
= cmp
->dst_pages
[i
];
2998 ASSERT(PageLocked(src_page
));
2999 ASSERT(PageLocked(dst_page
));
3001 addr
= kmap_atomic(src_page
);
3002 dst_addr
= kmap_atomic(dst_page
);
3004 flush_dcache_page(src_page
);
3005 flush_dcache_page(dst_page
);
3007 if (memcmp(addr
, dst_addr
, cmp_len
))
3010 kunmap_atomic(addr
);
3011 kunmap_atomic(dst_addr
);
3023 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64
*plen
,
3027 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
3029 if (off
+ olen
> inode
->i_size
|| off
+ olen
< off
)
3032 /* if we extend to eof, continue to block boundary */
3033 if (off
+ len
== inode
->i_size
)
3034 *plen
= len
= ALIGN(inode
->i_size
, bs
) - off
;
3036 /* Check that we are block aligned - btrfs_clone() requires this */
3037 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
3043 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 olen
,
3044 struct inode
*dst
, u64 dst_loff
)
3048 struct cmp_pages cmp
;
3050 u64 same_lock_start
= 0;
3051 u64 same_lock_len
= 0;
3062 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3067 * Single inode case wants the same checks, except we
3068 * don't want our length pushed out past i_size as
3069 * comparing that data range makes no sense.
3071 * extent_same_check_offsets() will do this for an
3072 * unaligned length at i_size, so catch it here and
3073 * reject the request.
3075 * This effectively means we require aligned extents
3076 * for the single-inode case, whereas the other cases
3077 * allow an unaligned length so long as it ends at
3085 /* Check for overlapping ranges */
3086 if (dst_loff
+ len
> loff
&& dst_loff
< loff
+ len
) {
3091 same_lock_start
= min_t(u64
, loff
, dst_loff
);
3092 same_lock_len
= max_t(u64
, loff
, dst_loff
) + len
- same_lock_start
;
3094 btrfs_double_inode_lock(src
, dst
);
3096 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3100 ret
= extent_same_check_offsets(dst
, dst_loff
, &len
, olen
);
3105 /* don't make the dst file partly checksummed */
3106 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3107 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
3113 ret
= btrfs_cmp_data_prepare(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3118 ret
= lock_extent_range(src
, same_lock_start
, same_lock_len
,
3121 ret
= btrfs_double_extent_lock(src
, loff
, dst
, dst_loff
, len
,
3124 * If one of the inodes has dirty pages in the respective range or
3125 * ordered extents, we need to flush dellaloc and wait for all ordered
3126 * extents in the range. We must unlock the pages and the ranges in the
3127 * io trees to avoid deadlocks when flushing delalloc (requires locking
3128 * pages) and when waiting for ordered extents to complete (they require
3131 if (ret
== -EAGAIN
) {
3133 * Ranges in the io trees already unlocked. Now unlock all
3134 * pages before waiting for all IO to complete.
3136 btrfs_cmp_data_free(&cmp
);
3138 btrfs_wait_ordered_range(src
, same_lock_start
,
3141 btrfs_wait_ordered_range(src
, loff
, len
);
3142 btrfs_wait_ordered_range(dst
, dst_loff
, len
);
3148 /* ranges in the io trees already unlocked */
3149 btrfs_cmp_data_free(&cmp
);
3153 /* pass original length for comparison so we stay within i_size */
3154 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3156 ret
= btrfs_clone(src
, dst
, loff
, olen
, len
, dst_loff
, 1);
3159 unlock_extent(&BTRFS_I(src
)->io_tree
, same_lock_start
,
3160 same_lock_start
+ same_lock_len
- 1);
3162 btrfs_double_extent_unlock(src
, loff
, dst
, dst_loff
, len
);
3164 btrfs_cmp_data_free(&cmp
);
3169 btrfs_double_inode_unlock(src
, dst
);
3174 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3176 ssize_t
btrfs_dedupe_file_range(struct file
*src_file
, u64 loff
, u64 olen
,
3177 struct file
*dst_file
, u64 dst_loff
)
3179 struct inode
*src
= file_inode(src_file
);
3180 struct inode
*dst
= file_inode(dst_file
);
3181 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
3184 if (olen
> BTRFS_MAX_DEDUPE_LEN
)
3185 olen
= BTRFS_MAX_DEDUPE_LEN
;
3187 if (WARN_ON_ONCE(bs
< PAGE_CACHE_SIZE
)) {
3189 * Btrfs does not support blocksize < page_size. As a
3190 * result, btrfs_cmp_data() won't correctly handle
3191 * this situation without an update.
3196 res
= btrfs_extent_same(src
, loff
, olen
, dst
, dst_loff
);
3202 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3203 struct inode
*inode
,
3209 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3212 inode_inc_iversion(inode
);
3213 if (!no_time_update
)
3214 inode
->i_mtime
= inode
->i_ctime
= current_fs_time(inode
->i_sb
);
3216 * We round up to the block size at eof when determining which
3217 * extents to clone above, but shouldn't round up the file size.
3219 if (endoff
> destoff
+ olen
)
3220 endoff
= destoff
+ olen
;
3221 if (endoff
> inode
->i_size
)
3222 btrfs_i_size_write(inode
, endoff
);
3224 ret
= btrfs_update_inode(trans
, root
, inode
);
3226 btrfs_abort_transaction(trans
, root
, ret
);
3227 btrfs_end_transaction(trans
, root
);
3230 ret
= btrfs_end_transaction(trans
, root
);
3235 static void clone_update_extent_map(struct inode
*inode
,
3236 const struct btrfs_trans_handle
*trans
,
3237 const struct btrfs_path
*path
,
3238 const u64 hole_offset
,
3241 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
3242 struct extent_map
*em
;
3245 em
= alloc_extent_map();
3247 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3248 &BTRFS_I(inode
)->runtime_flags
);
3253 struct btrfs_file_extent_item
*fi
;
3255 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3256 struct btrfs_file_extent_item
);
3257 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3258 em
->generation
= -1;
3259 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3260 BTRFS_FILE_EXTENT_INLINE
)
3261 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3262 &BTRFS_I(inode
)->runtime_flags
);
3264 em
->start
= hole_offset
;
3266 em
->ram_bytes
= em
->len
;
3267 em
->orig_start
= hole_offset
;
3268 em
->block_start
= EXTENT_MAP_HOLE
;
3270 em
->orig_block_len
= 0;
3271 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3272 em
->generation
= trans
->transid
;
3276 write_lock(&em_tree
->lock
);
3277 ret
= add_extent_mapping(em_tree
, em
, 1);
3278 write_unlock(&em_tree
->lock
);
3279 if (ret
!= -EEXIST
) {
3280 free_extent_map(em
);
3283 btrfs_drop_extent_cache(inode
, em
->start
,
3284 em
->start
+ em
->len
- 1, 0);
3288 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3289 &BTRFS_I(inode
)->runtime_flags
);
3293 * Make sure we do not end up inserting an inline extent into a file that has
3294 * already other (non-inline) extents. If a file has an inline extent it can
3295 * not have any other extents and the (single) inline extent must start at the
3296 * file offset 0. Failing to respect these rules will lead to file corruption,
3297 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3299 * We can have extents that have been already written to disk or we can have
3300 * dirty ranges still in delalloc, in which case the extent maps and items are
3301 * created only when we run delalloc, and the delalloc ranges might fall outside
3302 * the range we are currently locking in the inode's io tree. So we check the
3303 * inode's i_size because of that (i_size updates are done while holding the
3304 * i_mutex, which we are holding here).
3305 * We also check to see if the inode has a size not greater than "datal" but has
3306 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3307 * protected against such concurrent fallocate calls by the i_mutex).
3309 * If the file has no extents but a size greater than datal, do not allow the
3310 * copy because we would need turn the inline extent into a non-inline one (even
3311 * with NO_HOLES enabled). If we find our destination inode only has one inline
3312 * extent, just overwrite it with the source inline extent if its size is less
3313 * than the source extent's size, or we could copy the source inline extent's
3314 * data into the destination inode's inline extent if the later is greater then
3317 static int clone_copy_inline_extent(struct inode
*src
,
3319 struct btrfs_trans_handle
*trans
,
3320 struct btrfs_path
*path
,
3321 struct btrfs_key
*new_key
,
3322 const u64 drop_start
,
3328 struct btrfs_root
*root
= BTRFS_I(dst
)->root
;
3329 const u64 aligned_end
= ALIGN(new_key
->offset
+ datal
,
3332 struct btrfs_key key
;
3334 if (new_key
->offset
> 0)
3337 key
.objectid
= btrfs_ino(dst
);
3338 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3340 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
3343 } else if (ret
> 0) {
3344 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3345 ret
= btrfs_next_leaf(root
, path
);
3349 goto copy_inline_extent
;
3351 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
3352 if (key
.objectid
== btrfs_ino(dst
) &&
3353 key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3354 ASSERT(key
.offset
> 0);
3357 } else if (i_size_read(dst
) <= datal
) {
3358 struct btrfs_file_extent_item
*ei
;
3362 * If the file size is <= datal, make sure there are no other
3363 * extents following (can happen do to an fallocate call with
3364 * the flag FALLOC_FL_KEEP_SIZE).
3366 ei
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3367 struct btrfs_file_extent_item
);
3369 * If it's an inline extent, it can not have other extents
3372 if (btrfs_file_extent_type(path
->nodes
[0], ei
) ==
3373 BTRFS_FILE_EXTENT_INLINE
)
3374 goto copy_inline_extent
;
3376 ext_len
= btrfs_file_extent_num_bytes(path
->nodes
[0], ei
);
3377 if (ext_len
> aligned_end
)
3380 ret
= btrfs_next_item(root
, path
);
3383 } else if (ret
== 0) {
3384 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3386 if (key
.objectid
== btrfs_ino(dst
) &&
3387 key
.type
== BTRFS_EXTENT_DATA_KEY
)
3394 * We have no extent items, or we have an extent at offset 0 which may
3395 * or may not be inlined. All these cases are dealt the same way.
3397 if (i_size_read(dst
) > datal
) {
3399 * If the destination inode has an inline extent...
3400 * This would require copying the data from the source inline
3401 * extent into the beginning of the destination's inline extent.
3402 * But this is really complex, both extents can be compressed
3403 * or just one of them, which would require decompressing and
3404 * re-compressing data (which could increase the new compressed
3405 * size, not allowing the compressed data to fit anymore in an
3407 * So just don't support this case for now (it should be rare,
3408 * we are not really saving space when cloning inline extents).
3413 btrfs_release_path(path
);
3414 ret
= btrfs_drop_extents(trans
, root
, dst
, drop_start
, aligned_end
, 1);
3417 ret
= btrfs_insert_empty_item(trans
, root
, path
, new_key
, size
);
3422 const u32 start
= btrfs_file_extent_calc_inline_size(0);
3424 memmove(inline_data
+ start
, inline_data
+ start
+ skip
, datal
);
3427 write_extent_buffer(path
->nodes
[0], inline_data
,
3428 btrfs_item_ptr_offset(path
->nodes
[0],
3431 inode_add_bytes(dst
, datal
);
3437 * btrfs_clone() - clone a range from inode file to another
3439 * @src: Inode to clone from
3440 * @inode: Inode to clone to
3441 * @off: Offset within source to start clone from
3442 * @olen: Original length, passed by user, of range to clone
3443 * @olen_aligned: Block-aligned value of olen
3444 * @destoff: Offset within @inode to start clone
3445 * @no_time_update: Whether to update mtime/ctime on the target inode
3447 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3448 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3449 const u64 destoff
, int no_time_update
)
3451 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3452 struct btrfs_path
*path
= NULL
;
3453 struct extent_buffer
*leaf
;
3454 struct btrfs_trans_handle
*trans
;
3456 struct btrfs_key key
;
3460 const u64 len
= olen_aligned
;
3461 u64 last_dest_end
= destoff
;
3464 buf
= vmalloc(root
->nodesize
);
3468 path
= btrfs_alloc_path();
3474 path
->reada
= READA_FORWARD
;
3476 key
.objectid
= btrfs_ino(src
);
3477 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3481 u64 next_key_min_offset
= key
.offset
+ 1;
3484 * note the key will change type as we walk through the
3487 path
->leave_spinning
= 1;
3488 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3493 * First search, if no extent item that starts at offset off was
3494 * found but the previous item is an extent item, it's possible
3495 * it might overlap our target range, therefore process it.
3497 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3498 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3499 path
->slots
[0] - 1);
3500 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3504 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3506 if (path
->slots
[0] >= nritems
) {
3507 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3512 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3514 leaf
= path
->nodes
[0];
3515 slot
= path
->slots
[0];
3517 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3518 if (key
.type
> BTRFS_EXTENT_DATA_KEY
||
3519 key
.objectid
!= btrfs_ino(src
))
3522 if (key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3523 struct btrfs_file_extent_item
*extent
;
3526 struct btrfs_key new_key
;
3527 u64 disko
= 0, diskl
= 0;
3528 u64 datao
= 0, datal
= 0;
3532 extent
= btrfs_item_ptr(leaf
, slot
,
3533 struct btrfs_file_extent_item
);
3534 comp
= btrfs_file_extent_compression(leaf
, extent
);
3535 type
= btrfs_file_extent_type(leaf
, extent
);
3536 if (type
== BTRFS_FILE_EXTENT_REG
||
3537 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3538 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3540 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3542 datao
= btrfs_file_extent_offset(leaf
, extent
);
3543 datal
= btrfs_file_extent_num_bytes(leaf
,
3545 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3546 /* take upper bound, may be compressed */
3547 datal
= btrfs_file_extent_ram_bytes(leaf
,
3552 * The first search might have left us at an extent
3553 * item that ends before our target range's start, can
3554 * happen if we have holes and NO_HOLES feature enabled.
3556 if (key
.offset
+ datal
<= off
) {
3559 } else if (key
.offset
>= off
+ len
) {
3562 next_key_min_offset
= key
.offset
+ datal
;
3563 size
= btrfs_item_size_nr(leaf
, slot
);
3564 read_extent_buffer(leaf
, buf
,
3565 btrfs_item_ptr_offset(leaf
, slot
),
3568 btrfs_release_path(path
);
3569 path
->leave_spinning
= 0;
3571 memcpy(&new_key
, &key
, sizeof(new_key
));
3572 new_key
.objectid
= btrfs_ino(inode
);
3573 if (off
<= key
.offset
)
3574 new_key
.offset
= key
.offset
+ destoff
- off
;
3576 new_key
.offset
= destoff
;
3579 * Deal with a hole that doesn't have an extent item
3580 * that represents it (NO_HOLES feature enabled).
3581 * This hole is either in the middle of the cloning
3582 * range or at the beginning (fully overlaps it or
3583 * partially overlaps it).
3585 if (new_key
.offset
!= last_dest_end
)
3586 drop_start
= last_dest_end
;
3588 drop_start
= new_key
.offset
;
3591 * 1 - adjusting old extent (we may have to split it)
3592 * 1 - add new extent
3595 trans
= btrfs_start_transaction(root
, 3);
3596 if (IS_ERR(trans
)) {
3597 ret
= PTR_ERR(trans
);
3601 if (type
== BTRFS_FILE_EXTENT_REG
||
3602 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3604 * a | --- range to clone ---| b
3605 * | ------------- extent ------------- |
3608 /* subtract range b */
3609 if (key
.offset
+ datal
> off
+ len
)
3610 datal
= off
+ len
- key
.offset
;
3612 /* subtract range a */
3613 if (off
> key
.offset
) {
3614 datao
+= off
- key
.offset
;
3615 datal
-= off
- key
.offset
;
3618 ret
= btrfs_drop_extents(trans
, root
, inode
,
3620 new_key
.offset
+ datal
,
3623 if (ret
!= -EOPNOTSUPP
)
3624 btrfs_abort_transaction(trans
,
3626 btrfs_end_transaction(trans
, root
);
3630 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3633 btrfs_abort_transaction(trans
, root
,
3635 btrfs_end_transaction(trans
, root
);
3639 leaf
= path
->nodes
[0];
3640 slot
= path
->slots
[0];
3641 write_extent_buffer(leaf
, buf
,
3642 btrfs_item_ptr_offset(leaf
, slot
),
3645 extent
= btrfs_item_ptr(leaf
, slot
,
3646 struct btrfs_file_extent_item
);
3648 /* disko == 0 means it's a hole */
3652 btrfs_set_file_extent_offset(leaf
, extent
,
3654 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3658 inode_add_bytes(inode
, datal
);
3659 ret
= btrfs_inc_extent_ref(trans
, root
,
3661 root
->root_key
.objectid
,
3663 new_key
.offset
- datao
);
3665 btrfs_abort_transaction(trans
,
3668 btrfs_end_transaction(trans
,
3674 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3678 if (off
> key
.offset
) {
3679 skip
= off
- key
.offset
;
3680 new_key
.offset
+= skip
;
3683 if (key
.offset
+ datal
> off
+ len
)
3684 trim
= key
.offset
+ datal
- (off
+ len
);
3686 if (comp
&& (skip
|| trim
)) {
3688 btrfs_end_transaction(trans
, root
);
3691 size
-= skip
+ trim
;
3692 datal
-= skip
+ trim
;
3694 ret
= clone_copy_inline_extent(src
, inode
,
3701 if (ret
!= -EOPNOTSUPP
)
3702 btrfs_abort_transaction(trans
,
3705 btrfs_end_transaction(trans
, root
);
3708 leaf
= path
->nodes
[0];
3709 slot
= path
->slots
[0];
3712 /* If we have an implicit hole (NO_HOLES feature). */
3713 if (drop_start
< new_key
.offset
)
3714 clone_update_extent_map(inode
, trans
,
3716 new_key
.offset
- drop_start
);
3718 clone_update_extent_map(inode
, trans
, path
, 0, 0);
3720 btrfs_mark_buffer_dirty(leaf
);
3721 btrfs_release_path(path
);
3723 last_dest_end
= ALIGN(new_key
.offset
+ datal
,
3725 ret
= clone_finish_inode_update(trans
, inode
,
3731 if (new_key
.offset
+ datal
>= destoff
+ len
)
3734 btrfs_release_path(path
);
3735 key
.offset
= next_key_min_offset
;
3739 if (last_dest_end
< destoff
+ len
) {
3741 * We have an implicit hole (NO_HOLES feature is enabled) that
3742 * fully or partially overlaps our cloning range at its end.
3744 btrfs_release_path(path
);
3747 * 1 - remove extent(s)
3750 trans
= btrfs_start_transaction(root
, 2);
3751 if (IS_ERR(trans
)) {
3752 ret
= PTR_ERR(trans
);
3755 ret
= btrfs_drop_extents(trans
, root
, inode
,
3756 last_dest_end
, destoff
+ len
, 1);
3758 if (ret
!= -EOPNOTSUPP
)
3759 btrfs_abort_transaction(trans
, root
, ret
);
3760 btrfs_end_transaction(trans
, root
);
3763 clone_update_extent_map(inode
, trans
, NULL
, last_dest_end
,
3764 destoff
+ len
- last_dest_end
);
3765 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3766 destoff
, olen
, no_time_update
);
3770 btrfs_free_path(path
);
3775 static noinline
int btrfs_clone_files(struct file
*file
, struct file
*file_src
,
3776 u64 off
, u64 olen
, u64 destoff
)
3778 struct inode
*inode
= file_inode(file
);
3779 struct inode
*src
= file_inode(file_src
);
3780 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3783 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
3784 int same_inode
= src
== inode
;
3788 * - split compressed inline extents. annoying: we need to
3789 * decompress into destination's address_space (the file offset
3790 * may change, so source mapping won't do), then recompress (or
3791 * otherwise reinsert) a subrange.
3793 * - split destination inode's inline extents. The inline extents can
3794 * be either compressed or non-compressed.
3797 if (btrfs_root_readonly(root
))
3800 if (file_src
->f_path
.mnt
!= file
->f_path
.mnt
||
3801 src
->i_sb
!= inode
->i_sb
)
3804 /* don't make the dst file partly checksummed */
3805 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3806 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3809 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3813 btrfs_double_inode_lock(src
, inode
);
3818 /* determine range to clone */
3820 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3823 olen
= len
= src
->i_size
- off
;
3824 /* if we extend to eof, continue to block boundary */
3825 if (off
+ len
== src
->i_size
)
3826 len
= ALIGN(src
->i_size
, bs
) - off
;
3833 /* verify the end result is block aligned */
3834 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3835 !IS_ALIGNED(destoff
, bs
))
3838 /* verify if ranges are overlapped within the same file */
3840 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3844 if (destoff
> inode
->i_size
) {
3845 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3851 * Lock the target range too. Right after we replace the file extent
3852 * items in the fs tree (which now point to the cloned data), we might
3853 * have a worker replace them with extent items relative to a write
3854 * operation that was issued before this clone operation (i.e. confront
3855 * with inode.c:btrfs_finish_ordered_io).
3858 u64 lock_start
= min_t(u64
, off
, destoff
);
3859 u64 lock_len
= max_t(u64
, off
, destoff
) + len
- lock_start
;
3861 ret
= lock_extent_range(src
, lock_start
, lock_len
, true);
3863 ret
= btrfs_double_extent_lock(src
, off
, inode
, destoff
, len
,
3868 /* ranges in the io trees already unlocked */
3872 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
, 0);
3875 u64 lock_start
= min_t(u64
, off
, destoff
);
3876 u64 lock_end
= max_t(u64
, off
, destoff
) + len
- 1;
3878 unlock_extent(&BTRFS_I(src
)->io_tree
, lock_start
, lock_end
);
3880 btrfs_double_extent_unlock(src
, off
, inode
, destoff
, len
);
3883 * Truncate page cache pages so that future reads will see the cloned
3884 * data immediately and not the previous data.
3886 truncate_inode_pages_range(&inode
->i_data
,
3887 round_down(destoff
, PAGE_CACHE_SIZE
),
3888 round_up(destoff
+ len
, PAGE_CACHE_SIZE
) - 1);
3891 btrfs_double_inode_unlock(src
, inode
);
3897 ssize_t
btrfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
3898 struct file
*file_out
, loff_t pos_out
,
3899 size_t len
, unsigned int flags
)
3903 ret
= btrfs_clone_files(file_out
, file_in
, pos_in
, len
, pos_out
);
3909 int btrfs_clone_file_range(struct file
*src_file
, loff_t off
,
3910 struct file
*dst_file
, loff_t destoff
, u64 len
)
3912 return btrfs_clone_files(dst_file
, src_file
, off
, len
, destoff
);
3916 * there are many ways the trans_start and trans_end ioctls can lead
3917 * to deadlocks. They should only be used by applications that
3918 * basically own the machine, and have a very in depth understanding
3919 * of all the possible deadlocks and enospc problems.
3921 static long btrfs_ioctl_trans_start(struct file
*file
)
3923 struct inode
*inode
= file_inode(file
);
3924 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3925 struct btrfs_trans_handle
*trans
;
3929 if (!capable(CAP_SYS_ADMIN
))
3933 if (file
->private_data
)
3937 if (btrfs_root_readonly(root
))
3940 ret
= mnt_want_write_file(file
);
3944 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
3947 trans
= btrfs_start_ioctl_transaction(root
);
3951 file
->private_data
= trans
;
3955 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3956 mnt_drop_write_file(file
);
3961 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
3963 struct inode
*inode
= file_inode(file
);
3964 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3965 struct btrfs_root
*new_root
;
3966 struct btrfs_dir_item
*di
;
3967 struct btrfs_trans_handle
*trans
;
3968 struct btrfs_path
*path
;
3969 struct btrfs_key location
;
3970 struct btrfs_disk_key disk_key
;
3975 if (!capable(CAP_SYS_ADMIN
))
3978 ret
= mnt_want_write_file(file
);
3982 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
3988 objectid
= BTRFS_FS_TREE_OBJECTID
;
3990 location
.objectid
= objectid
;
3991 location
.type
= BTRFS_ROOT_ITEM_KEY
;
3992 location
.offset
= (u64
)-1;
3994 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
3995 if (IS_ERR(new_root
)) {
3996 ret
= PTR_ERR(new_root
);
4000 path
= btrfs_alloc_path();
4005 path
->leave_spinning
= 1;
4007 trans
= btrfs_start_transaction(root
, 1);
4008 if (IS_ERR(trans
)) {
4009 btrfs_free_path(path
);
4010 ret
= PTR_ERR(trans
);
4014 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
4015 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
4016 dir_id
, "default", 7, 1);
4017 if (IS_ERR_OR_NULL(di
)) {
4018 btrfs_free_path(path
);
4019 btrfs_end_transaction(trans
, root
);
4020 btrfs_err(new_root
->fs_info
, "Umm, you don't have the default dir"
4021 "item, this isn't going to work");
4026 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
4027 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
4028 btrfs_mark_buffer_dirty(path
->nodes
[0]);
4029 btrfs_free_path(path
);
4031 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
4032 btrfs_end_transaction(trans
, root
);
4034 mnt_drop_write_file(file
);
4038 void btrfs_get_block_group_info(struct list_head
*groups_list
,
4039 struct btrfs_ioctl_space_info
*space
)
4041 struct btrfs_block_group_cache
*block_group
;
4043 space
->total_bytes
= 0;
4044 space
->used_bytes
= 0;
4046 list_for_each_entry(block_group
, groups_list
, list
) {
4047 space
->flags
= block_group
->flags
;
4048 space
->total_bytes
+= block_group
->key
.offset
;
4049 space
->used_bytes
+=
4050 btrfs_block_group_used(&block_group
->item
);
4054 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
4056 struct btrfs_ioctl_space_args space_args
;
4057 struct btrfs_ioctl_space_info space
;
4058 struct btrfs_ioctl_space_info
*dest
;
4059 struct btrfs_ioctl_space_info
*dest_orig
;
4060 struct btrfs_ioctl_space_info __user
*user_dest
;
4061 struct btrfs_space_info
*info
;
4062 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
4063 BTRFS_BLOCK_GROUP_SYSTEM
,
4064 BTRFS_BLOCK_GROUP_METADATA
,
4065 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
4072 if (copy_from_user(&space_args
,
4073 (struct btrfs_ioctl_space_args __user
*)arg
,
4074 sizeof(space_args
)))
4077 for (i
= 0; i
< num_types
; i
++) {
4078 struct btrfs_space_info
*tmp
;
4082 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4084 if (tmp
->flags
== types
[i
]) {
4094 down_read(&info
->groups_sem
);
4095 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4096 if (!list_empty(&info
->block_groups
[c
]))
4099 up_read(&info
->groups_sem
);
4103 * Global block reserve, exported as a space_info
4107 /* space_slots == 0 means they are asking for a count */
4108 if (space_args
.space_slots
== 0) {
4109 space_args
.total_spaces
= slot_count
;
4113 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
4115 alloc_size
= sizeof(*dest
) * slot_count
;
4117 /* we generally have at most 6 or so space infos, one for each raid
4118 * level. So, a whole page should be more than enough for everyone
4120 if (alloc_size
> PAGE_CACHE_SIZE
)
4123 space_args
.total_spaces
= 0;
4124 dest
= kmalloc(alloc_size
, GFP_KERNEL
);
4129 /* now we have a buffer to copy into */
4130 for (i
= 0; i
< num_types
; i
++) {
4131 struct btrfs_space_info
*tmp
;
4138 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4140 if (tmp
->flags
== types
[i
]) {
4149 down_read(&info
->groups_sem
);
4150 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4151 if (!list_empty(&info
->block_groups
[c
])) {
4152 btrfs_get_block_group_info(
4153 &info
->block_groups
[c
], &space
);
4154 memcpy(dest
, &space
, sizeof(space
));
4156 space_args
.total_spaces
++;
4162 up_read(&info
->groups_sem
);
4166 * Add global block reserve
4169 struct btrfs_block_rsv
*block_rsv
= &root
->fs_info
->global_block_rsv
;
4171 spin_lock(&block_rsv
->lock
);
4172 space
.total_bytes
= block_rsv
->size
;
4173 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4174 spin_unlock(&block_rsv
->lock
);
4175 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4176 memcpy(dest
, &space
, sizeof(space
));
4177 space_args
.total_spaces
++;
4180 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4181 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4183 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4188 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4195 * there are many ways the trans_start and trans_end ioctls can lead
4196 * to deadlocks. They should only be used by applications that
4197 * basically own the machine, and have a very in depth understanding
4198 * of all the possible deadlocks and enospc problems.
4200 long btrfs_ioctl_trans_end(struct file
*file
)
4202 struct inode
*inode
= file_inode(file
);
4203 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4204 struct btrfs_trans_handle
*trans
;
4206 trans
= file
->private_data
;
4209 file
->private_data
= NULL
;
4211 btrfs_end_transaction(trans
, root
);
4213 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4215 mnt_drop_write_file(file
);
4219 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4222 struct btrfs_trans_handle
*trans
;
4226 trans
= btrfs_attach_transaction_barrier(root
);
4227 if (IS_ERR(trans
)) {
4228 if (PTR_ERR(trans
) != -ENOENT
)
4229 return PTR_ERR(trans
);
4231 /* No running transaction, don't bother */
4232 transid
= root
->fs_info
->last_trans_committed
;
4235 transid
= trans
->transid
;
4236 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
4238 btrfs_end_transaction(trans
, root
);
4243 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4248 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
4254 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4257 transid
= 0; /* current trans */
4259 return btrfs_wait_for_commit(root
, transid
);
4262 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4264 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4265 struct btrfs_ioctl_scrub_args
*sa
;
4268 if (!capable(CAP_SYS_ADMIN
))
4271 sa
= memdup_user(arg
, sizeof(*sa
));
4275 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4276 ret
= mnt_want_write_file(file
);
4281 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4282 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4285 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4288 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4289 mnt_drop_write_file(file
);
4295 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
4297 if (!capable(CAP_SYS_ADMIN
))
4300 return btrfs_scrub_cancel(root
->fs_info
);
4303 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
4306 struct btrfs_ioctl_scrub_args
*sa
;
4309 if (!capable(CAP_SYS_ADMIN
))
4312 sa
= memdup_user(arg
, sizeof(*sa
));
4316 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
4318 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4325 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
4328 struct btrfs_ioctl_get_dev_stats
*sa
;
4331 sa
= memdup_user(arg
, sizeof(*sa
));
4335 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4340 ret
= btrfs_get_dev_stats(root
, sa
);
4342 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4349 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
4351 struct btrfs_ioctl_dev_replace_args
*p
;
4354 if (!capable(CAP_SYS_ADMIN
))
4357 p
= memdup_user(arg
, sizeof(*p
));
4362 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4363 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
) {
4368 &root
->fs_info
->mutually_exclusive_operation_running
,
4370 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4372 ret
= btrfs_dev_replace_start(root
, p
);
4374 &root
->fs_info
->mutually_exclusive_operation_running
,
4378 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4379 btrfs_dev_replace_status(root
->fs_info
, p
);
4382 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4383 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
4390 if (copy_to_user(arg
, p
, sizeof(*p
)))
4397 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4403 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4404 struct inode_fs_paths
*ipath
= NULL
;
4405 struct btrfs_path
*path
;
4407 if (!capable(CAP_DAC_READ_SEARCH
))
4410 path
= btrfs_alloc_path();
4416 ipa
= memdup_user(arg
, sizeof(*ipa
));
4423 size
= min_t(u32
, ipa
->size
, 4096);
4424 ipath
= init_ipath(size
, root
, path
);
4425 if (IS_ERR(ipath
)) {
4426 ret
= PTR_ERR(ipath
);
4431 ret
= paths_from_inode(ipa
->inum
, ipath
);
4435 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4436 rel_ptr
= ipath
->fspath
->val
[i
] -
4437 (u64
)(unsigned long)ipath
->fspath
->val
;
4438 ipath
->fspath
->val
[i
] = rel_ptr
;
4441 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
4442 (void *)(unsigned long)ipath
->fspath
, size
);
4449 btrfs_free_path(path
);
4456 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4458 struct btrfs_data_container
*inodes
= ctx
;
4459 const size_t c
= 3 * sizeof(u64
);
4461 if (inodes
->bytes_left
>= c
) {
4462 inodes
->bytes_left
-= c
;
4463 inodes
->val
[inodes
->elem_cnt
] = inum
;
4464 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4465 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4466 inodes
->elem_cnt
+= 3;
4468 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4469 inodes
->bytes_left
= 0;
4470 inodes
->elem_missed
+= 3;
4476 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
4481 struct btrfs_ioctl_logical_ino_args
*loi
;
4482 struct btrfs_data_container
*inodes
= NULL
;
4483 struct btrfs_path
*path
= NULL
;
4485 if (!capable(CAP_SYS_ADMIN
))
4488 loi
= memdup_user(arg
, sizeof(*loi
));
4495 path
= btrfs_alloc_path();
4501 size
= min_t(u32
, loi
->size
, SZ_64K
);
4502 inodes
= init_data_container(size
);
4503 if (IS_ERR(inodes
)) {
4504 ret
= PTR_ERR(inodes
);
4509 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
4510 build_ino_list
, inodes
);
4516 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
4517 (void *)(unsigned long)inodes
, size
);
4522 btrfs_free_path(path
);
4529 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
4530 struct btrfs_ioctl_balance_args
*bargs
)
4532 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4534 bargs
->flags
= bctl
->flags
;
4536 if (atomic_read(&fs_info
->balance_running
))
4537 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4538 if (atomic_read(&fs_info
->balance_pause_req
))
4539 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4540 if (atomic_read(&fs_info
->balance_cancel_req
))
4541 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4543 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4544 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4545 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4548 spin_lock(&fs_info
->balance_lock
);
4549 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4550 spin_unlock(&fs_info
->balance_lock
);
4552 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4556 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4558 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4559 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4560 struct btrfs_ioctl_balance_args
*bargs
;
4561 struct btrfs_balance_control
*bctl
;
4562 bool need_unlock
; /* for mut. excl. ops lock */
4565 if (!capable(CAP_SYS_ADMIN
))
4568 ret
= mnt_want_write_file(file
);
4573 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4574 mutex_lock(&fs_info
->volume_mutex
);
4575 mutex_lock(&fs_info
->balance_mutex
);
4581 * mut. excl. ops lock is locked. Three possibilites:
4582 * (1) some other op is running
4583 * (2) balance is running
4584 * (3) balance is paused -- special case (think resume)
4586 mutex_lock(&fs_info
->balance_mutex
);
4587 if (fs_info
->balance_ctl
) {
4588 /* this is either (2) or (3) */
4589 if (!atomic_read(&fs_info
->balance_running
)) {
4590 mutex_unlock(&fs_info
->balance_mutex
);
4591 if (!mutex_trylock(&fs_info
->volume_mutex
))
4593 mutex_lock(&fs_info
->balance_mutex
);
4595 if (fs_info
->balance_ctl
&&
4596 !atomic_read(&fs_info
->balance_running
)) {
4598 need_unlock
= false;
4602 mutex_unlock(&fs_info
->balance_mutex
);
4603 mutex_unlock(&fs_info
->volume_mutex
);
4607 mutex_unlock(&fs_info
->balance_mutex
);
4613 mutex_unlock(&fs_info
->balance_mutex
);
4614 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4619 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4622 bargs
= memdup_user(arg
, sizeof(*bargs
));
4623 if (IS_ERR(bargs
)) {
4624 ret
= PTR_ERR(bargs
);
4628 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4629 if (!fs_info
->balance_ctl
) {
4634 bctl
= fs_info
->balance_ctl
;
4635 spin_lock(&fs_info
->balance_lock
);
4636 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4637 spin_unlock(&fs_info
->balance_lock
);
4645 if (fs_info
->balance_ctl
) {
4650 bctl
= kzalloc(sizeof(*bctl
), GFP_KERNEL
);
4656 bctl
->fs_info
= fs_info
;
4658 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4659 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4660 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4662 bctl
->flags
= bargs
->flags
;
4664 /* balance everything - no filters */
4665 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4668 if (bctl
->flags
& ~(BTRFS_BALANCE_ARGS_MASK
| BTRFS_BALANCE_TYPE_MASK
)) {
4675 * Ownership of bctl and mutually_exclusive_operation_running
4676 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4677 * or, if restriper was paused all the way until unmount, in
4678 * free_fs_info. mutually_exclusive_operation_running is
4679 * cleared in __cancel_balance.
4681 need_unlock
= false;
4683 ret
= btrfs_balance(bctl
, bargs
);
4687 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4696 mutex_unlock(&fs_info
->balance_mutex
);
4697 mutex_unlock(&fs_info
->volume_mutex
);
4699 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4701 mnt_drop_write_file(file
);
4705 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4707 if (!capable(CAP_SYS_ADMIN
))
4711 case BTRFS_BALANCE_CTL_PAUSE
:
4712 return btrfs_pause_balance(root
->fs_info
);
4713 case BTRFS_BALANCE_CTL_CANCEL
:
4714 return btrfs_cancel_balance(root
->fs_info
);
4720 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4723 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4724 struct btrfs_ioctl_balance_args
*bargs
;
4727 if (!capable(CAP_SYS_ADMIN
))
4730 mutex_lock(&fs_info
->balance_mutex
);
4731 if (!fs_info
->balance_ctl
) {
4736 bargs
= kzalloc(sizeof(*bargs
), GFP_KERNEL
);
4742 update_ioctl_balance_args(fs_info
, 1, bargs
);
4744 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4749 mutex_unlock(&fs_info
->balance_mutex
);
4753 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4755 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4756 struct btrfs_ioctl_quota_ctl_args
*sa
;
4757 struct btrfs_trans_handle
*trans
= NULL
;
4761 if (!capable(CAP_SYS_ADMIN
))
4764 ret
= mnt_want_write_file(file
);
4768 sa
= memdup_user(arg
, sizeof(*sa
));
4774 down_write(&root
->fs_info
->subvol_sem
);
4775 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
4776 if (IS_ERR(trans
)) {
4777 ret
= PTR_ERR(trans
);
4782 case BTRFS_QUOTA_CTL_ENABLE
:
4783 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
4785 case BTRFS_QUOTA_CTL_DISABLE
:
4786 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
4793 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
4798 up_write(&root
->fs_info
->subvol_sem
);
4800 mnt_drop_write_file(file
);
4804 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4806 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4807 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4808 struct btrfs_trans_handle
*trans
;
4812 if (!capable(CAP_SYS_ADMIN
))
4815 ret
= mnt_want_write_file(file
);
4819 sa
= memdup_user(arg
, sizeof(*sa
));
4825 trans
= btrfs_join_transaction(root
);
4826 if (IS_ERR(trans
)) {
4827 ret
= PTR_ERR(trans
);
4831 /* FIXME: check if the IDs really exist */
4833 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
4836 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
4840 /* update qgroup status and info */
4841 err
= btrfs_run_qgroups(trans
, root
->fs_info
);
4843 btrfs_std_error(root
->fs_info
, ret
,
4844 "failed to update qgroup status and info\n");
4845 err
= btrfs_end_transaction(trans
, root
);
4852 mnt_drop_write_file(file
);
4856 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4858 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4859 struct btrfs_ioctl_qgroup_create_args
*sa
;
4860 struct btrfs_trans_handle
*trans
;
4864 if (!capable(CAP_SYS_ADMIN
))
4867 ret
= mnt_want_write_file(file
);
4871 sa
= memdup_user(arg
, sizeof(*sa
));
4877 if (!sa
->qgroupid
) {
4882 trans
= btrfs_join_transaction(root
);
4883 if (IS_ERR(trans
)) {
4884 ret
= PTR_ERR(trans
);
4888 /* FIXME: check if the IDs really exist */
4890 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4892 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4895 err
= btrfs_end_transaction(trans
, root
);
4902 mnt_drop_write_file(file
);
4906 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4908 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4909 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4910 struct btrfs_trans_handle
*trans
;
4915 if (!capable(CAP_SYS_ADMIN
))
4918 ret
= mnt_want_write_file(file
);
4922 sa
= memdup_user(arg
, sizeof(*sa
));
4928 trans
= btrfs_join_transaction(root
);
4929 if (IS_ERR(trans
)) {
4930 ret
= PTR_ERR(trans
);
4934 qgroupid
= sa
->qgroupid
;
4936 /* take the current subvol as qgroup */
4937 qgroupid
= root
->root_key
.objectid
;
4940 /* FIXME: check if the IDs really exist */
4941 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
4943 err
= btrfs_end_transaction(trans
, root
);
4950 mnt_drop_write_file(file
);
4954 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
4956 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4957 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4960 if (!capable(CAP_SYS_ADMIN
))
4963 ret
= mnt_want_write_file(file
);
4967 qsa
= memdup_user(arg
, sizeof(*qsa
));
4978 ret
= btrfs_qgroup_rescan(root
->fs_info
);
4983 mnt_drop_write_file(file
);
4987 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
4989 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4990 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4993 if (!capable(CAP_SYS_ADMIN
))
4996 qsa
= kzalloc(sizeof(*qsa
), GFP_KERNEL
);
5000 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
5002 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
5005 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
5012 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
5014 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5016 if (!capable(CAP_SYS_ADMIN
))
5019 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
5022 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
5023 struct btrfs_ioctl_received_subvol_args
*sa
)
5025 struct inode
*inode
= file_inode(file
);
5026 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5027 struct btrfs_root_item
*root_item
= &root
->root_item
;
5028 struct btrfs_trans_handle
*trans
;
5029 struct timespec ct
= current_fs_time(inode
->i_sb
);
5031 int received_uuid_changed
;
5033 if (!inode_owner_or_capable(inode
))
5036 ret
= mnt_want_write_file(file
);
5040 down_write(&root
->fs_info
->subvol_sem
);
5042 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
5047 if (btrfs_root_readonly(root
)) {
5054 * 2 - uuid items (received uuid + subvol uuid)
5056 trans
= btrfs_start_transaction(root
, 3);
5057 if (IS_ERR(trans
)) {
5058 ret
= PTR_ERR(trans
);
5063 sa
->rtransid
= trans
->transid
;
5064 sa
->rtime
.sec
= ct
.tv_sec
;
5065 sa
->rtime
.nsec
= ct
.tv_nsec
;
5067 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
5069 if (received_uuid_changed
&&
5070 !btrfs_is_empty_uuid(root_item
->received_uuid
))
5071 btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
5072 root_item
->received_uuid
,
5073 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5074 root
->root_key
.objectid
);
5075 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
5076 btrfs_set_root_stransid(root_item
, sa
->stransid
);
5077 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
5078 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
5079 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
5080 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
5081 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
5083 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
5084 &root
->root_key
, &root
->root_item
);
5086 btrfs_end_transaction(trans
, root
);
5089 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
5090 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
5092 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5093 root
->root_key
.objectid
);
5094 if (ret
< 0 && ret
!= -EEXIST
) {
5095 btrfs_abort_transaction(trans
, root
, ret
);
5099 ret
= btrfs_commit_transaction(trans
, root
);
5101 btrfs_abort_transaction(trans
, root
, ret
);
5106 up_write(&root
->fs_info
->subvol_sem
);
5107 mnt_drop_write_file(file
);
5112 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
5115 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
5116 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
5119 args32
= memdup_user(arg
, sizeof(*args32
));
5120 if (IS_ERR(args32
)) {
5121 ret
= PTR_ERR(args32
);
5126 args64
= kmalloc(sizeof(*args64
), GFP_KERNEL
);
5132 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
5133 args64
->stransid
= args32
->stransid
;
5134 args64
->rtransid
= args32
->rtransid
;
5135 args64
->stime
.sec
= args32
->stime
.sec
;
5136 args64
->stime
.nsec
= args32
->stime
.nsec
;
5137 args64
->rtime
.sec
= args32
->rtime
.sec
;
5138 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
5139 args64
->flags
= args32
->flags
;
5141 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
5145 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
5146 args32
->stransid
= args64
->stransid
;
5147 args32
->rtransid
= args64
->rtransid
;
5148 args32
->stime
.sec
= args64
->stime
.sec
;
5149 args32
->stime
.nsec
= args64
->stime
.nsec
;
5150 args32
->rtime
.sec
= args64
->rtime
.sec
;
5151 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
5152 args32
->flags
= args64
->flags
;
5154 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
5165 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
5168 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
5171 sa
= memdup_user(arg
, sizeof(*sa
));
5178 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5183 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5192 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5194 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5197 char label
[BTRFS_LABEL_SIZE
];
5199 spin_lock(&root
->fs_info
->super_lock
);
5200 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5201 spin_unlock(&root
->fs_info
->super_lock
);
5203 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5205 if (len
== BTRFS_LABEL_SIZE
) {
5206 btrfs_warn(root
->fs_info
,
5207 "label is too long, return the first %zu bytes", --len
);
5210 ret
= copy_to_user(arg
, label
, len
);
5212 return ret
? -EFAULT
: 0;
5215 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5217 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5218 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5219 struct btrfs_trans_handle
*trans
;
5220 char label
[BTRFS_LABEL_SIZE
];
5223 if (!capable(CAP_SYS_ADMIN
))
5226 if (copy_from_user(label
, arg
, sizeof(label
)))
5229 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5230 btrfs_err(root
->fs_info
, "unable to set label with more than %d bytes",
5231 BTRFS_LABEL_SIZE
- 1);
5235 ret
= mnt_want_write_file(file
);
5239 trans
= btrfs_start_transaction(root
, 0);
5240 if (IS_ERR(trans
)) {
5241 ret
= PTR_ERR(trans
);
5245 spin_lock(&root
->fs_info
->super_lock
);
5246 strcpy(super_block
->label
, label
);
5247 spin_unlock(&root
->fs_info
->super_lock
);
5248 ret
= btrfs_commit_transaction(trans
, root
);
5251 mnt_drop_write_file(file
);
5255 #define INIT_FEATURE_FLAGS(suffix) \
5256 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5257 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5258 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5260 static int btrfs_ioctl_get_supported_features(struct file
*file
,
5263 static const struct btrfs_ioctl_feature_flags features
[3] = {
5264 INIT_FEATURE_FLAGS(SUPP
),
5265 INIT_FEATURE_FLAGS(SAFE_SET
),
5266 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5269 if (copy_to_user(arg
, &features
, sizeof(features
)))
5275 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5277 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5278 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5279 struct btrfs_ioctl_feature_flags features
;
5281 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5282 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5283 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5285 if (copy_to_user(arg
, &features
, sizeof(features
)))
5291 static int check_feature_bits(struct btrfs_root
*root
,
5292 enum btrfs_feature_set set
,
5293 u64 change_mask
, u64 flags
, u64 supported_flags
,
5294 u64 safe_set
, u64 safe_clear
)
5296 const char *type
= btrfs_feature_set_names
[set
];
5298 u64 disallowed
, unsupported
;
5299 u64 set_mask
= flags
& change_mask
;
5300 u64 clear_mask
= ~flags
& change_mask
;
5302 unsupported
= set_mask
& ~supported_flags
;
5304 names
= btrfs_printable_features(set
, unsupported
);
5306 btrfs_warn(root
->fs_info
,
5307 "this kernel does not support the %s feature bit%s",
5308 names
, strchr(names
, ',') ? "s" : "");
5311 btrfs_warn(root
->fs_info
,
5312 "this kernel does not support %s bits 0x%llx",
5317 disallowed
= set_mask
& ~safe_set
;
5319 names
= btrfs_printable_features(set
, disallowed
);
5321 btrfs_warn(root
->fs_info
,
5322 "can't set the %s feature bit%s while mounted",
5323 names
, strchr(names
, ',') ? "s" : "");
5326 btrfs_warn(root
->fs_info
,
5327 "can't set %s bits 0x%llx while mounted",
5332 disallowed
= clear_mask
& ~safe_clear
;
5334 names
= btrfs_printable_features(set
, disallowed
);
5336 btrfs_warn(root
->fs_info
,
5337 "can't clear the %s feature bit%s while mounted",
5338 names
, strchr(names
, ',') ? "s" : "");
5341 btrfs_warn(root
->fs_info
,
5342 "can't clear %s bits 0x%llx while mounted",
5350 #define check_feature(root, change_mask, flags, mask_base) \
5351 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5352 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5353 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5354 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5356 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5358 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5359 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5360 struct btrfs_ioctl_feature_flags flags
[2];
5361 struct btrfs_trans_handle
*trans
;
5365 if (!capable(CAP_SYS_ADMIN
))
5368 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5372 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5373 !flags
[0].incompat_flags
)
5376 ret
= check_feature(root
, flags
[0].compat_flags
,
5377 flags
[1].compat_flags
, COMPAT
);
5381 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
5382 flags
[1].compat_ro_flags
, COMPAT_RO
);
5386 ret
= check_feature(root
, flags
[0].incompat_flags
,
5387 flags
[1].incompat_flags
, INCOMPAT
);
5391 trans
= btrfs_start_transaction(root
, 0);
5393 return PTR_ERR(trans
);
5395 spin_lock(&root
->fs_info
->super_lock
);
5396 newflags
= btrfs_super_compat_flags(super_block
);
5397 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5398 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5399 btrfs_set_super_compat_flags(super_block
, newflags
);
5401 newflags
= btrfs_super_compat_ro_flags(super_block
);
5402 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5403 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5404 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5406 newflags
= btrfs_super_incompat_flags(super_block
);
5407 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5408 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5409 btrfs_set_super_incompat_flags(super_block
, newflags
);
5410 spin_unlock(&root
->fs_info
->super_lock
);
5412 return btrfs_commit_transaction(trans
, root
);
5415 long btrfs_ioctl(struct file
*file
, unsigned int
5416 cmd
, unsigned long arg
)
5418 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5419 void __user
*argp
= (void __user
*)arg
;
5422 case FS_IOC_GETFLAGS
:
5423 return btrfs_ioctl_getflags(file
, argp
);
5424 case FS_IOC_SETFLAGS
:
5425 return btrfs_ioctl_setflags(file
, argp
);
5426 case FS_IOC_GETVERSION
:
5427 return btrfs_ioctl_getversion(file
, argp
);
5429 return btrfs_ioctl_fitrim(file
, argp
);
5430 case BTRFS_IOC_SNAP_CREATE
:
5431 return btrfs_ioctl_snap_create(file
, argp
, 0);
5432 case BTRFS_IOC_SNAP_CREATE_V2
:
5433 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5434 case BTRFS_IOC_SUBVOL_CREATE
:
5435 return btrfs_ioctl_snap_create(file
, argp
, 1);
5436 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5437 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5438 case BTRFS_IOC_SNAP_DESTROY
:
5439 return btrfs_ioctl_snap_destroy(file
, argp
);
5440 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5441 return btrfs_ioctl_subvol_getflags(file
, argp
);
5442 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5443 return btrfs_ioctl_subvol_setflags(file
, argp
);
5444 case BTRFS_IOC_DEFAULT_SUBVOL
:
5445 return btrfs_ioctl_default_subvol(file
, argp
);
5446 case BTRFS_IOC_DEFRAG
:
5447 return btrfs_ioctl_defrag(file
, NULL
);
5448 case BTRFS_IOC_DEFRAG_RANGE
:
5449 return btrfs_ioctl_defrag(file
, argp
);
5450 case BTRFS_IOC_RESIZE
:
5451 return btrfs_ioctl_resize(file
, argp
);
5452 case BTRFS_IOC_ADD_DEV
:
5453 return btrfs_ioctl_add_dev(root
, argp
);
5454 case BTRFS_IOC_RM_DEV
:
5455 return btrfs_ioctl_rm_dev(file
, argp
);
5456 case BTRFS_IOC_FS_INFO
:
5457 return btrfs_ioctl_fs_info(root
, argp
);
5458 case BTRFS_IOC_DEV_INFO
:
5459 return btrfs_ioctl_dev_info(root
, argp
);
5460 case BTRFS_IOC_BALANCE
:
5461 return btrfs_ioctl_balance(file
, NULL
);
5462 case BTRFS_IOC_TRANS_START
:
5463 return btrfs_ioctl_trans_start(file
);
5464 case BTRFS_IOC_TRANS_END
:
5465 return btrfs_ioctl_trans_end(file
);
5466 case BTRFS_IOC_TREE_SEARCH
:
5467 return btrfs_ioctl_tree_search(file
, argp
);
5468 case BTRFS_IOC_TREE_SEARCH_V2
:
5469 return btrfs_ioctl_tree_search_v2(file
, argp
);
5470 case BTRFS_IOC_INO_LOOKUP
:
5471 return btrfs_ioctl_ino_lookup(file
, argp
);
5472 case BTRFS_IOC_INO_PATHS
:
5473 return btrfs_ioctl_ino_to_path(root
, argp
);
5474 case BTRFS_IOC_LOGICAL_INO
:
5475 return btrfs_ioctl_logical_to_ino(root
, argp
);
5476 case BTRFS_IOC_SPACE_INFO
:
5477 return btrfs_ioctl_space_info(root
, argp
);
5478 case BTRFS_IOC_SYNC
: {
5481 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
5484 ret
= btrfs_sync_fs(file_inode(file
)->i_sb
, 1);
5486 * The transaction thread may want to do more work,
5487 * namely it pokes the cleaner ktread that will start
5488 * processing uncleaned subvols.
5490 wake_up_process(root
->fs_info
->transaction_kthread
);
5493 case BTRFS_IOC_START_SYNC
:
5494 return btrfs_ioctl_start_sync(root
, argp
);
5495 case BTRFS_IOC_WAIT_SYNC
:
5496 return btrfs_ioctl_wait_sync(root
, argp
);
5497 case BTRFS_IOC_SCRUB
:
5498 return btrfs_ioctl_scrub(file
, argp
);
5499 case BTRFS_IOC_SCRUB_CANCEL
:
5500 return btrfs_ioctl_scrub_cancel(root
, argp
);
5501 case BTRFS_IOC_SCRUB_PROGRESS
:
5502 return btrfs_ioctl_scrub_progress(root
, argp
);
5503 case BTRFS_IOC_BALANCE_V2
:
5504 return btrfs_ioctl_balance(file
, argp
);
5505 case BTRFS_IOC_BALANCE_CTL
:
5506 return btrfs_ioctl_balance_ctl(root
, arg
);
5507 case BTRFS_IOC_BALANCE_PROGRESS
:
5508 return btrfs_ioctl_balance_progress(root
, argp
);
5509 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5510 return btrfs_ioctl_set_received_subvol(file
, argp
);
5512 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5513 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5515 case BTRFS_IOC_SEND
:
5516 return btrfs_ioctl_send(file
, argp
);
5517 case BTRFS_IOC_GET_DEV_STATS
:
5518 return btrfs_ioctl_get_dev_stats(root
, argp
);
5519 case BTRFS_IOC_QUOTA_CTL
:
5520 return btrfs_ioctl_quota_ctl(file
, argp
);
5521 case BTRFS_IOC_QGROUP_ASSIGN
:
5522 return btrfs_ioctl_qgroup_assign(file
, argp
);
5523 case BTRFS_IOC_QGROUP_CREATE
:
5524 return btrfs_ioctl_qgroup_create(file
, argp
);
5525 case BTRFS_IOC_QGROUP_LIMIT
:
5526 return btrfs_ioctl_qgroup_limit(file
, argp
);
5527 case BTRFS_IOC_QUOTA_RESCAN
:
5528 return btrfs_ioctl_quota_rescan(file
, argp
);
5529 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5530 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5531 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5532 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5533 case BTRFS_IOC_DEV_REPLACE
:
5534 return btrfs_ioctl_dev_replace(root
, argp
);
5535 case BTRFS_IOC_GET_FSLABEL
:
5536 return btrfs_ioctl_get_fslabel(file
, argp
);
5537 case BTRFS_IOC_SET_FSLABEL
:
5538 return btrfs_ioctl_set_fslabel(file
, argp
);
5539 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5540 return btrfs_ioctl_get_supported_features(file
, argp
);
5541 case BTRFS_IOC_GET_FEATURES
:
5542 return btrfs_ioctl_get_features(file
, argp
);
5543 case BTRFS_IOC_SET_FEATURES
:
5544 return btrfs_ioctl_set_features(file
, argp
);