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_TIME
;
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_TIME
;
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
);
848 if (d_really_is_positive(dentry
))
851 error
= btrfs_may_create(dir
, dentry
);
856 * even if this name doesn't exist, we may get hash collisions.
857 * check for them now when we can safely fail
859 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
865 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
867 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
871 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
872 async_transid
, readonly
, inherit
);
874 error
= create_subvol(dir
, dentry
, name
, namelen
,
875 async_transid
, inherit
);
878 fsnotify_mkdir(dir
, dentry
);
880 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
889 * When we're defragging a range, we don't want to kick it off again
890 * if it is really just waiting for delalloc to send it down.
891 * If we find a nice big extent or delalloc range for the bytes in the
892 * file you want to defrag, we return 0 to let you know to skip this
895 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, u32 thresh
)
897 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
898 struct extent_map
*em
= NULL
;
899 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
902 read_lock(&em_tree
->lock
);
903 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
904 read_unlock(&em_tree
->lock
);
907 end
= extent_map_end(em
);
909 if (end
- offset
> thresh
)
912 /* if we already have a nice delalloc here, just stop */
914 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
915 thresh
, EXTENT_DELALLOC
, 1);
922 * helper function to walk through a file and find extents
923 * newer than a specific transid, and smaller than thresh.
925 * This is used by the defragging code to find new and small
928 static int find_new_extents(struct btrfs_root
*root
,
929 struct inode
*inode
, u64 newer_than
,
930 u64
*off
, u32 thresh
)
932 struct btrfs_path
*path
;
933 struct btrfs_key min_key
;
934 struct extent_buffer
*leaf
;
935 struct btrfs_file_extent_item
*extent
;
938 u64 ino
= btrfs_ino(inode
);
940 path
= btrfs_alloc_path();
944 min_key
.objectid
= ino
;
945 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
946 min_key
.offset
= *off
;
949 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
953 if (min_key
.objectid
!= ino
)
955 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
958 leaf
= path
->nodes
[0];
959 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
960 struct btrfs_file_extent_item
);
962 type
= btrfs_file_extent_type(leaf
, extent
);
963 if (type
== BTRFS_FILE_EXTENT_REG
&&
964 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
965 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
966 *off
= min_key
.offset
;
967 btrfs_free_path(path
);
972 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
973 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
977 if (min_key
.offset
== (u64
)-1)
981 btrfs_release_path(path
);
984 btrfs_free_path(path
);
988 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
990 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
991 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
992 struct extent_map
*em
;
993 u64 len
= PAGE_CACHE_SIZE
;
996 * hopefully we have this extent in the tree already, try without
997 * the full extent lock
999 read_lock(&em_tree
->lock
);
1000 em
= lookup_extent_mapping(em_tree
, start
, len
);
1001 read_unlock(&em_tree
->lock
);
1004 struct extent_state
*cached
= NULL
;
1005 u64 end
= start
+ len
- 1;
1007 /* get the big lock and read metadata off disk */
1008 lock_extent_bits(io_tree
, start
, end
, &cached
);
1009 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
1010 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1019 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1021 struct extent_map
*next
;
1024 /* this is the last extent */
1025 if (em
->start
+ em
->len
>= i_size_read(inode
))
1028 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1029 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
1031 else if ((em
->block_start
+ em
->block_len
== next
->block_start
) &&
1032 (em
->block_len
> SZ_128K
&& next
->block_len
> SZ_128K
))
1035 free_extent_map(next
);
1039 static int should_defrag_range(struct inode
*inode
, u64 start
, u32 thresh
,
1040 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1043 struct extent_map
*em
;
1045 bool next_mergeable
= true;
1046 bool prev_mergeable
= true;
1049 * make sure that once we start defragging an extent, we keep on
1052 if (start
< *defrag_end
)
1057 em
= defrag_lookup_extent(inode
, start
);
1061 /* this will cover holes, and inline extents */
1062 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1068 prev_mergeable
= false;
1070 next_mergeable
= defrag_check_next_extent(inode
, em
);
1072 * we hit a real extent, if it is big or the next extent is not a
1073 * real extent, don't bother defragging it
1075 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1076 (em
->len
>= thresh
|| (!next_mergeable
&& !prev_mergeable
)))
1080 * last_len ends up being a counter of how many bytes we've defragged.
1081 * every time we choose not to defrag an extent, we reset *last_len
1082 * so that the next tiny extent will force a defrag.
1084 * The end result of this is that tiny extents before a single big
1085 * extent will force at least part of that big extent to be defragged.
1088 *defrag_end
= extent_map_end(em
);
1091 *skip
= extent_map_end(em
);
1095 free_extent_map(em
);
1100 * it doesn't do much good to defrag one or two pages
1101 * at a time. This pulls in a nice chunk of pages
1102 * to COW and defrag.
1104 * It also makes sure the delalloc code has enough
1105 * dirty data to avoid making new small extents as part
1108 * It's a good idea to start RA on this range
1109 * before calling this.
1111 static int cluster_pages_for_defrag(struct inode
*inode
,
1112 struct page
**pages
,
1113 unsigned long start_index
,
1114 unsigned long num_pages
)
1116 unsigned long file_end
;
1117 u64 isize
= i_size_read(inode
);
1124 struct btrfs_ordered_extent
*ordered
;
1125 struct extent_state
*cached_state
= NULL
;
1126 struct extent_io_tree
*tree
;
1127 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1129 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1130 if (!isize
|| start_index
> file_end
)
1133 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1135 ret
= btrfs_delalloc_reserve_space(inode
,
1136 start_index
<< PAGE_CACHE_SHIFT
,
1137 page_cnt
<< PAGE_CACHE_SHIFT
);
1141 tree
= &BTRFS_I(inode
)->io_tree
;
1143 /* step one, lock all the pages */
1144 for (i
= 0; i
< page_cnt
; i
++) {
1147 page
= find_or_create_page(inode
->i_mapping
,
1148 start_index
+ i
, mask
);
1152 page_start
= page_offset(page
);
1153 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1155 lock_extent_bits(tree
, page_start
, page_end
,
1157 ordered
= btrfs_lookup_ordered_extent(inode
,
1159 unlock_extent_cached(tree
, page_start
, page_end
,
1160 &cached_state
, GFP_NOFS
);
1165 btrfs_start_ordered_extent(inode
, ordered
, 1);
1166 btrfs_put_ordered_extent(ordered
);
1169 * we unlocked the page above, so we need check if
1170 * it was released or not.
1172 if (page
->mapping
!= inode
->i_mapping
) {
1174 page_cache_release(page
);
1179 if (!PageUptodate(page
)) {
1180 btrfs_readpage(NULL
, page
);
1182 if (!PageUptodate(page
)) {
1184 page_cache_release(page
);
1190 if (page
->mapping
!= inode
->i_mapping
) {
1192 page_cache_release(page
);
1202 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1206 * so now we have a nice long stream of locked
1207 * and up to date pages, lets wait on them
1209 for (i
= 0; i
< i_done
; i
++)
1210 wait_on_page_writeback(pages
[i
]);
1212 page_start
= page_offset(pages
[0]);
1213 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1215 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1216 page_start
, page_end
- 1, &cached_state
);
1217 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1218 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1219 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1220 &cached_state
, GFP_NOFS
);
1222 if (i_done
!= page_cnt
) {
1223 spin_lock(&BTRFS_I(inode
)->lock
);
1224 BTRFS_I(inode
)->outstanding_extents
++;
1225 spin_unlock(&BTRFS_I(inode
)->lock
);
1226 btrfs_delalloc_release_space(inode
,
1227 start_index
<< PAGE_CACHE_SHIFT
,
1228 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1232 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1233 &cached_state
, GFP_NOFS
);
1235 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1236 page_start
, page_end
- 1, &cached_state
,
1239 for (i
= 0; i
< i_done
; i
++) {
1240 clear_page_dirty_for_io(pages
[i
]);
1241 ClearPageChecked(pages
[i
]);
1242 set_page_extent_mapped(pages
[i
]);
1243 set_page_dirty(pages
[i
]);
1244 unlock_page(pages
[i
]);
1245 page_cache_release(pages
[i
]);
1249 for (i
= 0; i
< i_done
; i
++) {
1250 unlock_page(pages
[i
]);
1251 page_cache_release(pages
[i
]);
1253 btrfs_delalloc_release_space(inode
,
1254 start_index
<< PAGE_CACHE_SHIFT
,
1255 page_cnt
<< PAGE_CACHE_SHIFT
);
1260 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1261 struct btrfs_ioctl_defrag_range_args
*range
,
1262 u64 newer_than
, unsigned long max_to_defrag
)
1264 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1265 struct file_ra_state
*ra
= NULL
;
1266 unsigned long last_index
;
1267 u64 isize
= i_size_read(inode
);
1271 u64 newer_off
= range
->start
;
1273 unsigned long ra_index
= 0;
1275 int defrag_count
= 0;
1276 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1277 u32 extent_thresh
= range
->extent_thresh
;
1278 unsigned long max_cluster
= SZ_256K
>> PAGE_CACHE_SHIFT
;
1279 unsigned long cluster
= max_cluster
;
1280 u64 new_align
= ~((u64
)SZ_128K
- 1);
1281 struct page
**pages
= NULL
;
1286 if (range
->start
>= isize
)
1289 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1290 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1292 if (range
->compress_type
)
1293 compress_type
= range
->compress_type
;
1296 if (extent_thresh
== 0)
1297 extent_thresh
= SZ_256K
;
1300 * if we were not given a file, allocate a readahead
1304 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1307 file_ra_state_init(ra
, inode
->i_mapping
);
1312 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1319 /* find the last page to defrag */
1320 if (range
->start
+ range
->len
> range
->start
) {
1321 last_index
= min_t(u64
, isize
- 1,
1322 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1324 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1328 ret
= find_new_extents(root
, inode
, newer_than
,
1329 &newer_off
, SZ_64K
);
1331 range
->start
= newer_off
;
1333 * we always align our defrag to help keep
1334 * the extents in the file evenly spaced
1336 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1340 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1343 max_to_defrag
= last_index
- i
+ 1;
1346 * make writeback starts from i, so the defrag range can be
1347 * written sequentially.
1349 if (i
< inode
->i_mapping
->writeback_index
)
1350 inode
->i_mapping
->writeback_index
= i
;
1352 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1353 (i
< DIV_ROUND_UP(i_size_read(inode
), PAGE_CACHE_SIZE
))) {
1355 * make sure we stop running if someone unmounts
1358 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1361 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1362 btrfs_debug(root
->fs_info
, "defrag_file cancelled");
1367 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1368 extent_thresh
, &last_len
, &skip
,
1369 &defrag_end
, range
->flags
&
1370 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1373 * the should_defrag function tells us how much to skip
1374 * bump our counter by the suggested amount
1376 next
= DIV_ROUND_UP(skip
, PAGE_CACHE_SIZE
);
1377 i
= max(i
+ 1, next
);
1382 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1383 PAGE_CACHE_SHIFT
) - i
;
1384 cluster
= min(cluster
, max_cluster
);
1386 cluster
= max_cluster
;
1389 if (i
+ cluster
> ra_index
) {
1390 ra_index
= max(i
, ra_index
);
1391 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1393 ra_index
+= cluster
;
1397 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1398 BTRFS_I(inode
)->force_compress
= compress_type
;
1399 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1401 inode_unlock(inode
);
1405 defrag_count
+= ret
;
1406 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1407 inode_unlock(inode
);
1410 if (newer_off
== (u64
)-1)
1416 newer_off
= max(newer_off
+ 1,
1417 (u64
)i
<< PAGE_CACHE_SHIFT
);
1419 ret
= find_new_extents(root
, inode
, newer_than
,
1420 &newer_off
, SZ_64K
);
1422 range
->start
= newer_off
;
1423 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1430 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1438 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1439 filemap_flush(inode
->i_mapping
);
1440 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1441 &BTRFS_I(inode
)->runtime_flags
))
1442 filemap_flush(inode
->i_mapping
);
1445 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1446 /* the filemap_flush will queue IO into the worker threads, but
1447 * we have to make sure the IO is actually started and that
1448 * ordered extents get created before we return
1450 atomic_inc(&root
->fs_info
->async_submit_draining
);
1451 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1452 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1453 wait_event(root
->fs_info
->async_submit_wait
,
1454 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1455 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1457 atomic_dec(&root
->fs_info
->async_submit_draining
);
1460 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1461 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1467 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1469 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1470 inode_unlock(inode
);
1478 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1484 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1485 struct btrfs_ioctl_vol_args
*vol_args
;
1486 struct btrfs_trans_handle
*trans
;
1487 struct btrfs_device
*device
= NULL
;
1490 char *devstr
= NULL
;
1494 if (!capable(CAP_SYS_ADMIN
))
1497 ret
= mnt_want_write_file(file
);
1501 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1503 mnt_drop_write_file(file
);
1504 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1507 mutex_lock(&root
->fs_info
->volume_mutex
);
1508 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1509 if (IS_ERR(vol_args
)) {
1510 ret
= PTR_ERR(vol_args
);
1514 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1516 sizestr
= vol_args
->name
;
1517 devstr
= strchr(sizestr
, ':');
1519 sizestr
= devstr
+ 1;
1521 devstr
= vol_args
->name
;
1522 ret
= kstrtoull(devstr
, 10, &devid
);
1529 btrfs_info(root
->fs_info
, "resizing devid %llu", devid
);
1532 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1534 btrfs_info(root
->fs_info
, "resizer unable to find device %llu",
1540 if (!device
->writeable
) {
1541 btrfs_info(root
->fs_info
,
1542 "resizer unable to apply on readonly device %llu",
1548 if (!strcmp(sizestr
, "max"))
1549 new_size
= device
->bdev
->bd_inode
->i_size
;
1551 if (sizestr
[0] == '-') {
1554 } else if (sizestr
[0] == '+') {
1558 new_size
= memparse(sizestr
, &retptr
);
1559 if (*retptr
!= '\0' || new_size
== 0) {
1565 if (device
->is_tgtdev_for_dev_replace
) {
1570 old_size
= btrfs_device_get_total_bytes(device
);
1573 if (new_size
> old_size
) {
1577 new_size
= old_size
- new_size
;
1578 } else if (mod
> 0) {
1579 if (new_size
> ULLONG_MAX
- old_size
) {
1583 new_size
= old_size
+ new_size
;
1586 if (new_size
< SZ_256M
) {
1590 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1595 new_size
= div_u64(new_size
, root
->sectorsize
);
1596 new_size
*= root
->sectorsize
;
1598 btrfs_info_in_rcu(root
->fs_info
, "new size for %s is %llu",
1599 rcu_str_deref(device
->name
), new_size
);
1601 if (new_size
> old_size
) {
1602 trans
= btrfs_start_transaction(root
, 0);
1603 if (IS_ERR(trans
)) {
1604 ret
= PTR_ERR(trans
);
1607 ret
= btrfs_grow_device(trans
, device
, new_size
);
1608 btrfs_commit_transaction(trans
, root
);
1609 } else if (new_size
< old_size
) {
1610 ret
= btrfs_shrink_device(device
, new_size
);
1611 } /* equal, nothing need to do */
1616 mutex_unlock(&root
->fs_info
->volume_mutex
);
1617 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1618 mnt_drop_write_file(file
);
1622 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1623 char *name
, unsigned long fd
, int subvol
,
1624 u64
*transid
, bool readonly
,
1625 struct btrfs_qgroup_inherit
*inherit
)
1630 ret
= mnt_want_write_file(file
);
1634 namelen
= strlen(name
);
1635 if (strchr(name
, '/')) {
1637 goto out_drop_write
;
1640 if (name
[0] == '.' &&
1641 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1643 goto out_drop_write
;
1647 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1648 NULL
, transid
, readonly
, inherit
);
1650 struct fd src
= fdget(fd
);
1651 struct inode
*src_inode
;
1654 goto out_drop_write
;
1657 src_inode
= file_inode(src
.file
);
1658 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1659 btrfs_info(BTRFS_I(src_inode
)->root
->fs_info
,
1660 "Snapshot src from another FS");
1662 } else if (!inode_owner_or_capable(src_inode
)) {
1664 * Subvolume creation is not restricted, but snapshots
1665 * are limited to own subvolumes only
1669 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1670 BTRFS_I(src_inode
)->root
,
1671 transid
, readonly
, inherit
);
1676 mnt_drop_write_file(file
);
1681 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1682 void __user
*arg
, int subvol
)
1684 struct btrfs_ioctl_vol_args
*vol_args
;
1687 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1688 if (IS_ERR(vol_args
))
1689 return PTR_ERR(vol_args
);
1690 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1692 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1693 vol_args
->fd
, subvol
,
1700 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1701 void __user
*arg
, int subvol
)
1703 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1707 bool readonly
= false;
1708 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1710 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1711 if (IS_ERR(vol_args
))
1712 return PTR_ERR(vol_args
);
1713 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1715 if (vol_args
->flags
&
1716 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1717 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1722 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1724 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1726 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1727 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1731 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1732 if (IS_ERR(inherit
)) {
1733 ret
= PTR_ERR(inherit
);
1738 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1739 vol_args
->fd
, subvol
, ptr
,
1744 if (ptr
&& copy_to_user(arg
+
1745 offsetof(struct btrfs_ioctl_vol_args_v2
,
1757 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1760 struct inode
*inode
= file_inode(file
);
1761 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1765 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1768 down_read(&root
->fs_info
->subvol_sem
);
1769 if (btrfs_root_readonly(root
))
1770 flags
|= BTRFS_SUBVOL_RDONLY
;
1771 up_read(&root
->fs_info
->subvol_sem
);
1773 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1779 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1782 struct inode
*inode
= file_inode(file
);
1783 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1784 struct btrfs_trans_handle
*trans
;
1789 if (!inode_owner_or_capable(inode
))
1792 ret
= mnt_want_write_file(file
);
1796 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1798 goto out_drop_write
;
1801 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1803 goto out_drop_write
;
1806 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1808 goto out_drop_write
;
1811 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1813 goto out_drop_write
;
1816 down_write(&root
->fs_info
->subvol_sem
);
1819 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1822 root_flags
= btrfs_root_flags(&root
->root_item
);
1823 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1824 btrfs_set_root_flags(&root
->root_item
,
1825 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1828 * Block RO -> RW transition if this subvolume is involved in
1831 spin_lock(&root
->root_item_lock
);
1832 if (root
->send_in_progress
== 0) {
1833 btrfs_set_root_flags(&root
->root_item
,
1834 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1835 spin_unlock(&root
->root_item_lock
);
1837 spin_unlock(&root
->root_item_lock
);
1838 btrfs_warn(root
->fs_info
,
1839 "Attempt to set subvolume %llu read-write during send",
1840 root
->root_key
.objectid
);
1846 trans
= btrfs_start_transaction(root
, 1);
1847 if (IS_ERR(trans
)) {
1848 ret
= PTR_ERR(trans
);
1852 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1853 &root
->root_key
, &root
->root_item
);
1855 btrfs_commit_transaction(trans
, root
);
1858 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1860 up_write(&root
->fs_info
->subvol_sem
);
1862 mnt_drop_write_file(file
);
1868 * helper to check if the subvolume references other subvolumes
1870 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1872 struct btrfs_path
*path
;
1873 struct btrfs_dir_item
*di
;
1874 struct btrfs_key key
;
1878 path
= btrfs_alloc_path();
1882 /* Make sure this root isn't set as the default subvol */
1883 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
1884 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
, path
,
1885 dir_id
, "default", 7, 0);
1886 if (di
&& !IS_ERR(di
)) {
1887 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1888 if (key
.objectid
== root
->root_key
.objectid
) {
1890 btrfs_err(root
->fs_info
, "deleting default subvolume "
1891 "%llu is not allowed", key
.objectid
);
1894 btrfs_release_path(path
);
1897 key
.objectid
= root
->root_key
.objectid
;
1898 key
.type
= BTRFS_ROOT_REF_KEY
;
1899 key
.offset
= (u64
)-1;
1901 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1908 if (path
->slots
[0] > 0) {
1910 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1911 if (key
.objectid
== root
->root_key
.objectid
&&
1912 key
.type
== BTRFS_ROOT_REF_KEY
)
1916 btrfs_free_path(path
);
1920 static noinline
int key_in_sk(struct btrfs_key
*key
,
1921 struct btrfs_ioctl_search_key
*sk
)
1923 struct btrfs_key test
;
1926 test
.objectid
= sk
->min_objectid
;
1927 test
.type
= sk
->min_type
;
1928 test
.offset
= sk
->min_offset
;
1930 ret
= btrfs_comp_cpu_keys(key
, &test
);
1934 test
.objectid
= sk
->max_objectid
;
1935 test
.type
= sk
->max_type
;
1936 test
.offset
= sk
->max_offset
;
1938 ret
= btrfs_comp_cpu_keys(key
, &test
);
1944 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1945 struct btrfs_path
*path
,
1946 struct btrfs_key
*key
,
1947 struct btrfs_ioctl_search_key
*sk
,
1950 unsigned long *sk_offset
,
1954 struct extent_buffer
*leaf
;
1955 struct btrfs_ioctl_search_header sh
;
1956 struct btrfs_key test
;
1957 unsigned long item_off
;
1958 unsigned long item_len
;
1964 leaf
= path
->nodes
[0];
1965 slot
= path
->slots
[0];
1966 nritems
= btrfs_header_nritems(leaf
);
1968 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1972 found_transid
= btrfs_header_generation(leaf
);
1974 for (i
= slot
; i
< nritems
; i
++) {
1975 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1976 item_len
= btrfs_item_size_nr(leaf
, i
);
1978 btrfs_item_key_to_cpu(leaf
, key
, i
);
1979 if (!key_in_sk(key
, sk
))
1982 if (sizeof(sh
) + item_len
> *buf_size
) {
1989 * return one empty item back for v1, which does not
1993 *buf_size
= sizeof(sh
) + item_len
;
1998 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
2003 sh
.objectid
= key
->objectid
;
2004 sh
.offset
= key
->offset
;
2005 sh
.type
= key
->type
;
2007 sh
.transid
= found_transid
;
2009 /* copy search result header */
2010 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2015 *sk_offset
+= sizeof(sh
);
2018 char __user
*up
= ubuf
+ *sk_offset
;
2020 if (read_extent_buffer_to_user(leaf
, up
,
2021 item_off
, item_len
)) {
2026 *sk_offset
+= item_len
;
2030 if (ret
) /* -EOVERFLOW from above */
2033 if (*num_found
>= sk
->nr_items
) {
2040 test
.objectid
= sk
->max_objectid
;
2041 test
.type
= sk
->max_type
;
2042 test
.offset
= sk
->max_offset
;
2043 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
2045 else if (key
->offset
< (u64
)-1)
2047 else if (key
->type
< (u8
)-1) {
2050 } else if (key
->objectid
< (u64
)-1) {
2058 * 0: all items from this leaf copied, continue with next
2059 * 1: * more items can be copied, but unused buffer is too small
2060 * * all items were found
2061 * Either way, it will stops the loop which iterates to the next
2063 * -EOVERFLOW: item was to large for buffer
2064 * -EFAULT: could not copy extent buffer back to userspace
2069 static noinline
int search_ioctl(struct inode
*inode
,
2070 struct btrfs_ioctl_search_key
*sk
,
2074 struct btrfs_root
*root
;
2075 struct btrfs_key key
;
2076 struct btrfs_path
*path
;
2077 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
2080 unsigned long sk_offset
= 0;
2082 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2083 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2087 path
= btrfs_alloc_path();
2091 if (sk
->tree_id
== 0) {
2092 /* search the root of the inode that was passed */
2093 root
= BTRFS_I(inode
)->root
;
2095 key
.objectid
= sk
->tree_id
;
2096 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2097 key
.offset
= (u64
)-1;
2098 root
= btrfs_read_fs_root_no_name(info
, &key
);
2100 btrfs_err(info
, "could not find root %llu",
2102 btrfs_free_path(path
);
2107 key
.objectid
= sk
->min_objectid
;
2108 key
.type
= sk
->min_type
;
2109 key
.offset
= sk
->min_offset
;
2112 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2118 ret
= copy_to_sk(root
, path
, &key
, sk
, buf_size
, ubuf
,
2119 &sk_offset
, &num_found
);
2120 btrfs_release_path(path
);
2128 sk
->nr_items
= num_found
;
2129 btrfs_free_path(path
);
2133 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2136 struct btrfs_ioctl_search_args __user
*uargs
;
2137 struct btrfs_ioctl_search_key sk
;
2138 struct inode
*inode
;
2142 if (!capable(CAP_SYS_ADMIN
))
2145 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2147 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2150 buf_size
= sizeof(uargs
->buf
);
2152 inode
= file_inode(file
);
2153 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2156 * In the origin implementation an overflow is handled by returning a
2157 * search header with a len of zero, so reset ret.
2159 if (ret
== -EOVERFLOW
)
2162 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2167 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2170 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2171 struct btrfs_ioctl_search_args_v2 args
;
2172 struct inode
*inode
;
2175 const size_t buf_limit
= SZ_16M
;
2177 if (!capable(CAP_SYS_ADMIN
))
2180 /* copy search header and buffer size */
2181 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2182 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2185 buf_size
= args
.buf_size
;
2187 if (buf_size
< sizeof(struct btrfs_ioctl_search_header
))
2190 /* limit result size to 16MB */
2191 if (buf_size
> buf_limit
)
2192 buf_size
= buf_limit
;
2194 inode
= file_inode(file
);
2195 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2196 (char *)(&uarg
->buf
[0]));
2197 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2199 else if (ret
== -EOVERFLOW
&&
2200 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2207 * Search INODE_REFs to identify path name of 'dirid' directory
2208 * in a 'tree_id' tree. and sets path name to 'name'.
2210 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2211 u64 tree_id
, u64 dirid
, char *name
)
2213 struct btrfs_root
*root
;
2214 struct btrfs_key key
;
2220 struct btrfs_inode_ref
*iref
;
2221 struct extent_buffer
*l
;
2222 struct btrfs_path
*path
;
2224 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2229 path
= btrfs_alloc_path();
2233 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2235 key
.objectid
= tree_id
;
2236 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2237 key
.offset
= (u64
)-1;
2238 root
= btrfs_read_fs_root_no_name(info
, &key
);
2240 btrfs_err(info
, "could not find root %llu", tree_id
);
2245 key
.objectid
= dirid
;
2246 key
.type
= BTRFS_INODE_REF_KEY
;
2247 key
.offset
= (u64
)-1;
2250 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2254 ret
= btrfs_previous_item(root
, path
, dirid
,
2255 BTRFS_INODE_REF_KEY
);
2265 slot
= path
->slots
[0];
2266 btrfs_item_key_to_cpu(l
, &key
, slot
);
2268 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2269 len
= btrfs_inode_ref_name_len(l
, iref
);
2271 total_len
+= len
+ 1;
2273 ret
= -ENAMETOOLONG
;
2278 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2280 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2283 btrfs_release_path(path
);
2284 key
.objectid
= key
.offset
;
2285 key
.offset
= (u64
)-1;
2286 dirid
= key
.objectid
;
2288 memmove(name
, ptr
, total_len
);
2289 name
[total_len
] = '\0';
2292 btrfs_free_path(path
);
2296 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2299 struct btrfs_ioctl_ino_lookup_args
*args
;
2300 struct inode
*inode
;
2303 args
= memdup_user(argp
, sizeof(*args
));
2305 return PTR_ERR(args
);
2307 inode
= file_inode(file
);
2310 * Unprivileged query to obtain the containing subvolume root id. The
2311 * path is reset so it's consistent with btrfs_search_path_in_tree.
2313 if (args
->treeid
== 0)
2314 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2316 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2321 if (!capable(CAP_SYS_ADMIN
)) {
2326 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2327 args
->treeid
, args
->objectid
,
2331 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2338 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2341 struct dentry
*parent
= file
->f_path
.dentry
;
2342 struct dentry
*dentry
;
2343 struct inode
*dir
= d_inode(parent
);
2344 struct inode
*inode
;
2345 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2346 struct btrfs_root
*dest
= NULL
;
2347 struct btrfs_ioctl_vol_args
*vol_args
;
2348 struct btrfs_trans_handle
*trans
;
2349 struct btrfs_block_rsv block_rsv
;
2351 u64 qgroup_reserved
;
2356 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2357 if (IS_ERR(vol_args
))
2358 return PTR_ERR(vol_args
);
2360 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2361 namelen
= strlen(vol_args
->name
);
2362 if (strchr(vol_args
->name
, '/') ||
2363 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2368 err
= mnt_want_write_file(file
);
2373 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2375 goto out_drop_write
;
2376 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2377 if (IS_ERR(dentry
)) {
2378 err
= PTR_ERR(dentry
);
2379 goto out_unlock_dir
;
2382 if (d_really_is_negative(dentry
)) {
2387 inode
= d_inode(dentry
);
2388 dest
= BTRFS_I(inode
)->root
;
2389 if (!capable(CAP_SYS_ADMIN
)) {
2391 * Regular user. Only allow this with a special mount
2392 * option, when the user has write+exec access to the
2393 * subvol root, and when rmdir(2) would have been
2396 * Note that this is _not_ check that the subvol is
2397 * empty or doesn't contain data that we wouldn't
2398 * otherwise be able to delete.
2400 * Users who want to delete empty subvols should try
2404 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2408 * Do not allow deletion if the parent dir is the same
2409 * as the dir to be deleted. That means the ioctl
2410 * must be called on the dentry referencing the root
2411 * of the subvol, not a random directory contained
2418 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2423 /* check if subvolume may be deleted by a user */
2424 err
= btrfs_may_delete(dir
, dentry
, 1);
2428 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2436 * Don't allow to delete a subvolume with send in progress. This is
2437 * inside the i_mutex so the error handling that has to drop the bit
2438 * again is not run concurrently.
2440 spin_lock(&dest
->root_item_lock
);
2441 root_flags
= btrfs_root_flags(&dest
->root_item
);
2442 if (dest
->send_in_progress
== 0) {
2443 btrfs_set_root_flags(&dest
->root_item
,
2444 root_flags
| BTRFS_ROOT_SUBVOL_DEAD
);
2445 spin_unlock(&dest
->root_item_lock
);
2447 spin_unlock(&dest
->root_item_lock
);
2448 btrfs_warn(root
->fs_info
,
2449 "Attempt to delete subvolume %llu during send",
2450 dest
->root_key
.objectid
);
2452 goto out_unlock_inode
;
2455 down_write(&root
->fs_info
->subvol_sem
);
2457 err
= may_destroy_subvol(dest
);
2461 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2463 * One for dir inode, two for dir entries, two for root
2466 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2467 5, &qgroup_reserved
, true);
2471 trans
= btrfs_start_transaction(root
, 0);
2472 if (IS_ERR(trans
)) {
2473 err
= PTR_ERR(trans
);
2476 trans
->block_rsv
= &block_rsv
;
2477 trans
->bytes_reserved
= block_rsv
.size
;
2479 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2480 dest
->root_key
.objectid
,
2481 dentry
->d_name
.name
,
2482 dentry
->d_name
.len
);
2485 btrfs_abort_transaction(trans
, root
, ret
);
2489 btrfs_record_root_in_trans(trans
, dest
);
2491 memset(&dest
->root_item
.drop_progress
, 0,
2492 sizeof(dest
->root_item
.drop_progress
));
2493 dest
->root_item
.drop_level
= 0;
2494 btrfs_set_root_refs(&dest
->root_item
, 0);
2496 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED
, &dest
->state
)) {
2497 ret
= btrfs_insert_orphan_item(trans
,
2498 root
->fs_info
->tree_root
,
2499 dest
->root_key
.objectid
);
2501 btrfs_abort_transaction(trans
, root
, ret
);
2507 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2508 dest
->root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
2509 dest
->root_key
.objectid
);
2510 if (ret
&& ret
!= -ENOENT
) {
2511 btrfs_abort_transaction(trans
, root
, ret
);
2515 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2516 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2517 dest
->root_item
.received_uuid
,
2518 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2519 dest
->root_key
.objectid
);
2520 if (ret
&& ret
!= -ENOENT
) {
2521 btrfs_abort_transaction(trans
, root
, ret
);
2528 trans
->block_rsv
= NULL
;
2529 trans
->bytes_reserved
= 0;
2530 ret
= btrfs_end_transaction(trans
, root
);
2533 inode
->i_flags
|= S_DEAD
;
2535 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2537 up_write(&root
->fs_info
->subvol_sem
);
2539 spin_lock(&dest
->root_item_lock
);
2540 root_flags
= btrfs_root_flags(&dest
->root_item
);
2541 btrfs_set_root_flags(&dest
->root_item
,
2542 root_flags
& ~BTRFS_ROOT_SUBVOL_DEAD
);
2543 spin_unlock(&dest
->root_item_lock
);
2546 inode_unlock(inode
);
2548 d_invalidate(dentry
);
2549 btrfs_invalidate_inodes(dest
);
2551 ASSERT(dest
->send_in_progress
== 0);
2554 if (dest
->ino_cache_inode
) {
2555 iput(dest
->ino_cache_inode
);
2556 dest
->ino_cache_inode
= NULL
;
2564 mnt_drop_write_file(file
);
2570 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2572 struct inode
*inode
= file_inode(file
);
2573 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2574 struct btrfs_ioctl_defrag_range_args
*range
;
2577 ret
= mnt_want_write_file(file
);
2581 if (btrfs_root_readonly(root
)) {
2586 switch (inode
->i_mode
& S_IFMT
) {
2588 if (!capable(CAP_SYS_ADMIN
)) {
2592 ret
= btrfs_defrag_root(root
);
2595 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2598 if (!(file
->f_mode
& FMODE_WRITE
)) {
2603 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2610 if (copy_from_user(range
, argp
,
2616 /* compression requires us to start the IO */
2617 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2618 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2619 range
->extent_thresh
= (u32
)-1;
2622 /* the rest are all set to zero by kzalloc */
2623 range
->len
= (u64
)-1;
2625 ret
= btrfs_defrag_file(file_inode(file
), file
,
2635 mnt_drop_write_file(file
);
2639 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2641 struct btrfs_ioctl_vol_args
*vol_args
;
2644 if (!capable(CAP_SYS_ADMIN
))
2647 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2649 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2652 mutex_lock(&root
->fs_info
->volume_mutex
);
2653 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2654 if (IS_ERR(vol_args
)) {
2655 ret
= PTR_ERR(vol_args
);
2659 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2660 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2663 btrfs_info(root
->fs_info
, "disk added %s",vol_args
->name
);
2667 mutex_unlock(&root
->fs_info
->volume_mutex
);
2668 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2672 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2674 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2675 struct btrfs_ioctl_vol_args
*vol_args
;
2678 if (!capable(CAP_SYS_ADMIN
))
2681 ret
= mnt_want_write_file(file
);
2685 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2686 if (IS_ERR(vol_args
)) {
2687 ret
= PTR_ERR(vol_args
);
2691 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2693 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2695 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2699 mutex_lock(&root
->fs_info
->volume_mutex
);
2700 ret
= btrfs_rm_device(root
, vol_args
->name
);
2701 mutex_unlock(&root
->fs_info
->volume_mutex
);
2702 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2705 btrfs_info(root
->fs_info
, "disk deleted %s",vol_args
->name
);
2710 mnt_drop_write_file(file
);
2714 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2716 struct btrfs_ioctl_fs_info_args
*fi_args
;
2717 struct btrfs_device
*device
;
2718 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2721 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2725 mutex_lock(&fs_devices
->device_list_mutex
);
2726 fi_args
->num_devices
= fs_devices
->num_devices
;
2727 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2729 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
2730 if (device
->devid
> fi_args
->max_id
)
2731 fi_args
->max_id
= device
->devid
;
2733 mutex_unlock(&fs_devices
->device_list_mutex
);
2735 fi_args
->nodesize
= root
->fs_info
->super_copy
->nodesize
;
2736 fi_args
->sectorsize
= root
->fs_info
->super_copy
->sectorsize
;
2737 fi_args
->clone_alignment
= root
->fs_info
->super_copy
->sectorsize
;
2739 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2746 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2748 struct btrfs_ioctl_dev_info_args
*di_args
;
2749 struct btrfs_device
*dev
;
2750 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2752 char *s_uuid
= NULL
;
2754 di_args
= memdup_user(arg
, sizeof(*di_args
));
2755 if (IS_ERR(di_args
))
2756 return PTR_ERR(di_args
);
2758 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2759 s_uuid
= di_args
->uuid
;
2761 mutex_lock(&fs_devices
->device_list_mutex
);
2762 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2769 di_args
->devid
= dev
->devid
;
2770 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
2771 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
2772 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2774 struct rcu_string
*name
;
2777 name
= rcu_dereference(dev
->name
);
2778 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2780 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2782 di_args
->path
[0] = '\0';
2786 mutex_unlock(&fs_devices
->device_list_mutex
);
2787 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2794 static struct page
*extent_same_get_page(struct inode
*inode
, pgoff_t index
)
2798 page
= grab_cache_page(inode
->i_mapping
, index
);
2800 return ERR_PTR(-ENOMEM
);
2802 if (!PageUptodate(page
)) {
2805 ret
= btrfs_readpage(NULL
, page
);
2807 return ERR_PTR(ret
);
2809 if (!PageUptodate(page
)) {
2811 page_cache_release(page
);
2812 return ERR_PTR(-EIO
);
2814 if (page
->mapping
!= inode
->i_mapping
) {
2816 page_cache_release(page
);
2817 return ERR_PTR(-EAGAIN
);
2824 static int gather_extent_pages(struct inode
*inode
, struct page
**pages
,
2825 int num_pages
, u64 off
)
2828 pgoff_t index
= off
>> PAGE_CACHE_SHIFT
;
2830 for (i
= 0; i
< num_pages
; i
++) {
2832 pages
[i
] = extent_same_get_page(inode
, index
+ i
);
2833 if (IS_ERR(pages
[i
])) {
2834 int err
= PTR_ERR(pages
[i
]);
2845 static int lock_extent_range(struct inode
*inode
, u64 off
, u64 len
,
2846 bool retry_range_locking
)
2849 * Do any pending delalloc/csum calculations on inode, one way or
2850 * another, and lock file content.
2851 * The locking order is:
2854 * 2) range in the inode's io tree
2857 struct btrfs_ordered_extent
*ordered
;
2858 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2859 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2862 ordered
->file_offset
+ ordered
->len
<= off
||
2863 ordered
->file_offset
>= off
+ len
) &&
2864 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2865 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
)) {
2867 btrfs_put_ordered_extent(ordered
);
2870 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2872 btrfs_put_ordered_extent(ordered
);
2873 if (!retry_range_locking
)
2875 btrfs_wait_ordered_range(inode
, off
, len
);
2880 static void btrfs_double_inode_unlock(struct inode
*inode1
, struct inode
*inode2
)
2882 inode_unlock(inode1
);
2883 inode_unlock(inode2
);
2886 static void btrfs_double_inode_lock(struct inode
*inode1
, struct inode
*inode2
)
2888 if (inode1
< inode2
)
2889 swap(inode1
, inode2
);
2891 inode_lock_nested(inode1
, I_MUTEX_PARENT
);
2892 inode_lock_nested(inode2
, I_MUTEX_CHILD
);
2895 static void btrfs_double_extent_unlock(struct inode
*inode1
, u64 loff1
,
2896 struct inode
*inode2
, u64 loff2
, u64 len
)
2898 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2899 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2902 static int btrfs_double_extent_lock(struct inode
*inode1
, u64 loff1
,
2903 struct inode
*inode2
, u64 loff2
, u64 len
,
2904 bool retry_range_locking
)
2908 if (inode1
< inode2
) {
2909 swap(inode1
, inode2
);
2912 ret
= lock_extent_range(inode1
, loff1
, len
, retry_range_locking
);
2915 ret
= lock_extent_range(inode2
, loff2
, len
, retry_range_locking
);
2917 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
,
2924 struct page
**src_pages
;
2925 struct page
**dst_pages
;
2928 static void btrfs_cmp_data_free(struct cmp_pages
*cmp
)
2933 for (i
= 0; i
< cmp
->num_pages
; i
++) {
2934 pg
= cmp
->src_pages
[i
];
2937 page_cache_release(pg
);
2939 pg
= cmp
->dst_pages
[i
];
2942 page_cache_release(pg
);
2945 kfree(cmp
->src_pages
);
2946 kfree(cmp
->dst_pages
);
2949 static int btrfs_cmp_data_prepare(struct inode
*src
, u64 loff
,
2950 struct inode
*dst
, u64 dst_loff
,
2951 u64 len
, struct cmp_pages
*cmp
)
2954 int num_pages
= PAGE_CACHE_ALIGN(len
) >> PAGE_CACHE_SHIFT
;
2955 struct page
**src_pgarr
, **dst_pgarr
;
2958 * We must gather up all the pages before we initiate our
2959 * extent locking. We use an array for the page pointers. Size
2960 * of the array is bounded by len, which is in turn bounded by
2961 * BTRFS_MAX_DEDUPE_LEN.
2963 src_pgarr
= kzalloc(num_pages
* sizeof(struct page
*), GFP_NOFS
);
2964 dst_pgarr
= kzalloc(num_pages
* sizeof(struct page
*), GFP_NOFS
);
2965 if (!src_pgarr
|| !dst_pgarr
) {
2970 cmp
->num_pages
= num_pages
;
2971 cmp
->src_pages
= src_pgarr
;
2972 cmp
->dst_pages
= dst_pgarr
;
2974 ret
= gather_extent_pages(src
, cmp
->src_pages
, cmp
->num_pages
, loff
);
2978 ret
= gather_extent_pages(dst
, cmp
->dst_pages
, cmp
->num_pages
, dst_loff
);
2982 btrfs_cmp_data_free(cmp
);
2986 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
2987 u64 dst_loff
, u64 len
, struct cmp_pages
*cmp
)
2991 struct page
*src_page
, *dst_page
;
2992 unsigned int cmp_len
= PAGE_CACHE_SIZE
;
2993 void *addr
, *dst_addr
;
2997 if (len
< PAGE_CACHE_SIZE
)
3000 BUG_ON(i
>= cmp
->num_pages
);
3002 src_page
= cmp
->src_pages
[i
];
3003 dst_page
= cmp
->dst_pages
[i
];
3004 ASSERT(PageLocked(src_page
));
3005 ASSERT(PageLocked(dst_page
));
3007 addr
= kmap_atomic(src_page
);
3008 dst_addr
= kmap_atomic(dst_page
);
3010 flush_dcache_page(src_page
);
3011 flush_dcache_page(dst_page
);
3013 if (memcmp(addr
, dst_addr
, cmp_len
))
3016 kunmap_atomic(addr
);
3017 kunmap_atomic(dst_addr
);
3029 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64
*plen
,
3033 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
3035 if (off
+ olen
> inode
->i_size
|| off
+ olen
< off
)
3038 /* if we extend to eof, continue to block boundary */
3039 if (off
+ len
== inode
->i_size
)
3040 *plen
= len
= ALIGN(inode
->i_size
, bs
) - off
;
3042 /* Check that we are block aligned - btrfs_clone() requires this */
3043 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
3049 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 olen
,
3050 struct inode
*dst
, u64 dst_loff
)
3054 struct cmp_pages cmp
;
3056 u64 same_lock_start
= 0;
3057 u64 same_lock_len
= 0;
3068 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3073 * Single inode case wants the same checks, except we
3074 * don't want our length pushed out past i_size as
3075 * comparing that data range makes no sense.
3077 * extent_same_check_offsets() will do this for an
3078 * unaligned length at i_size, so catch it here and
3079 * reject the request.
3081 * This effectively means we require aligned extents
3082 * for the single-inode case, whereas the other cases
3083 * allow an unaligned length so long as it ends at
3091 /* Check for overlapping ranges */
3092 if (dst_loff
+ len
> loff
&& dst_loff
< loff
+ len
) {
3097 same_lock_start
= min_t(u64
, loff
, dst_loff
);
3098 same_lock_len
= max_t(u64
, loff
, dst_loff
) + len
- same_lock_start
;
3100 btrfs_double_inode_lock(src
, dst
);
3102 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3106 ret
= extent_same_check_offsets(dst
, dst_loff
, &len
, olen
);
3111 /* don't make the dst file partly checksummed */
3112 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3113 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
3119 ret
= btrfs_cmp_data_prepare(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3124 ret
= lock_extent_range(src
, same_lock_start
, same_lock_len
,
3127 ret
= btrfs_double_extent_lock(src
, loff
, dst
, dst_loff
, len
,
3130 * If one of the inodes has dirty pages in the respective range or
3131 * ordered extents, we need to flush dellaloc and wait for all ordered
3132 * extents in the range. We must unlock the pages and the ranges in the
3133 * io trees to avoid deadlocks when flushing delalloc (requires locking
3134 * pages) and when waiting for ordered extents to complete (they require
3137 if (ret
== -EAGAIN
) {
3139 * Ranges in the io trees already unlocked. Now unlock all
3140 * pages before waiting for all IO to complete.
3142 btrfs_cmp_data_free(&cmp
);
3144 btrfs_wait_ordered_range(src
, same_lock_start
,
3147 btrfs_wait_ordered_range(src
, loff
, len
);
3148 btrfs_wait_ordered_range(dst
, dst_loff
, len
);
3154 /* ranges in the io trees already unlocked */
3155 btrfs_cmp_data_free(&cmp
);
3159 /* pass original length for comparison so we stay within i_size */
3160 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3162 ret
= btrfs_clone(src
, dst
, loff
, olen
, len
, dst_loff
, 1);
3165 unlock_extent(&BTRFS_I(src
)->io_tree
, same_lock_start
,
3166 same_lock_start
+ same_lock_len
- 1);
3168 btrfs_double_extent_unlock(src
, loff
, dst
, dst_loff
, len
);
3170 btrfs_cmp_data_free(&cmp
);
3175 btrfs_double_inode_unlock(src
, dst
);
3180 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3182 ssize_t
btrfs_dedupe_file_range(struct file
*src_file
, u64 loff
, u64 olen
,
3183 struct file
*dst_file
, u64 dst_loff
)
3185 struct inode
*src
= file_inode(src_file
);
3186 struct inode
*dst
= file_inode(dst_file
);
3187 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
3190 if (olen
> BTRFS_MAX_DEDUPE_LEN
)
3191 olen
= BTRFS_MAX_DEDUPE_LEN
;
3193 if (WARN_ON_ONCE(bs
< PAGE_CACHE_SIZE
)) {
3195 * Btrfs does not support blocksize < page_size. As a
3196 * result, btrfs_cmp_data() won't correctly handle
3197 * this situation without an update.
3202 res
= btrfs_extent_same(src
, loff
, olen
, dst
, dst_loff
);
3208 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3209 struct inode
*inode
,
3215 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3218 inode_inc_iversion(inode
);
3219 if (!no_time_update
)
3220 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3222 * We round up to the block size at eof when determining which
3223 * extents to clone above, but shouldn't round up the file size.
3225 if (endoff
> destoff
+ olen
)
3226 endoff
= destoff
+ olen
;
3227 if (endoff
> inode
->i_size
)
3228 btrfs_i_size_write(inode
, endoff
);
3230 ret
= btrfs_update_inode(trans
, root
, inode
);
3232 btrfs_abort_transaction(trans
, root
, ret
);
3233 btrfs_end_transaction(trans
, root
);
3236 ret
= btrfs_end_transaction(trans
, root
);
3241 static void clone_update_extent_map(struct inode
*inode
,
3242 const struct btrfs_trans_handle
*trans
,
3243 const struct btrfs_path
*path
,
3244 const u64 hole_offset
,
3247 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
3248 struct extent_map
*em
;
3251 em
= alloc_extent_map();
3253 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3254 &BTRFS_I(inode
)->runtime_flags
);
3259 struct btrfs_file_extent_item
*fi
;
3261 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3262 struct btrfs_file_extent_item
);
3263 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3264 em
->generation
= -1;
3265 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3266 BTRFS_FILE_EXTENT_INLINE
)
3267 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3268 &BTRFS_I(inode
)->runtime_flags
);
3270 em
->start
= hole_offset
;
3272 em
->ram_bytes
= em
->len
;
3273 em
->orig_start
= hole_offset
;
3274 em
->block_start
= EXTENT_MAP_HOLE
;
3276 em
->orig_block_len
= 0;
3277 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3278 em
->generation
= trans
->transid
;
3282 write_lock(&em_tree
->lock
);
3283 ret
= add_extent_mapping(em_tree
, em
, 1);
3284 write_unlock(&em_tree
->lock
);
3285 if (ret
!= -EEXIST
) {
3286 free_extent_map(em
);
3289 btrfs_drop_extent_cache(inode
, em
->start
,
3290 em
->start
+ em
->len
- 1, 0);
3294 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3295 &BTRFS_I(inode
)->runtime_flags
);
3299 * Make sure we do not end up inserting an inline extent into a file that has
3300 * already other (non-inline) extents. If a file has an inline extent it can
3301 * not have any other extents and the (single) inline extent must start at the
3302 * file offset 0. Failing to respect these rules will lead to file corruption,
3303 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3305 * We can have extents that have been already written to disk or we can have
3306 * dirty ranges still in delalloc, in which case the extent maps and items are
3307 * created only when we run delalloc, and the delalloc ranges might fall outside
3308 * the range we are currently locking in the inode's io tree. So we check the
3309 * inode's i_size because of that (i_size updates are done while holding the
3310 * i_mutex, which we are holding here).
3311 * We also check to see if the inode has a size not greater than "datal" but has
3312 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3313 * protected against such concurrent fallocate calls by the i_mutex).
3315 * If the file has no extents but a size greater than datal, do not allow the
3316 * copy because we would need turn the inline extent into a non-inline one (even
3317 * with NO_HOLES enabled). If we find our destination inode only has one inline
3318 * extent, just overwrite it with the source inline extent if its size is less
3319 * than the source extent's size, or we could copy the source inline extent's
3320 * data into the destination inode's inline extent if the later is greater then
3323 static int clone_copy_inline_extent(struct inode
*src
,
3325 struct btrfs_trans_handle
*trans
,
3326 struct btrfs_path
*path
,
3327 struct btrfs_key
*new_key
,
3328 const u64 drop_start
,
3334 struct btrfs_root
*root
= BTRFS_I(dst
)->root
;
3335 const u64 aligned_end
= ALIGN(new_key
->offset
+ datal
,
3338 struct btrfs_key key
;
3340 if (new_key
->offset
> 0)
3343 key
.objectid
= btrfs_ino(dst
);
3344 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3346 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
3349 } else if (ret
> 0) {
3350 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3351 ret
= btrfs_next_leaf(root
, path
);
3355 goto copy_inline_extent
;
3357 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
3358 if (key
.objectid
== btrfs_ino(dst
) &&
3359 key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3360 ASSERT(key
.offset
> 0);
3363 } else if (i_size_read(dst
) <= datal
) {
3364 struct btrfs_file_extent_item
*ei
;
3368 * If the file size is <= datal, make sure there are no other
3369 * extents following (can happen do to an fallocate call with
3370 * the flag FALLOC_FL_KEEP_SIZE).
3372 ei
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3373 struct btrfs_file_extent_item
);
3375 * If it's an inline extent, it can not have other extents
3378 if (btrfs_file_extent_type(path
->nodes
[0], ei
) ==
3379 BTRFS_FILE_EXTENT_INLINE
)
3380 goto copy_inline_extent
;
3382 ext_len
= btrfs_file_extent_num_bytes(path
->nodes
[0], ei
);
3383 if (ext_len
> aligned_end
)
3386 ret
= btrfs_next_item(root
, path
);
3389 } else if (ret
== 0) {
3390 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3392 if (key
.objectid
== btrfs_ino(dst
) &&
3393 key
.type
== BTRFS_EXTENT_DATA_KEY
)
3400 * We have no extent items, or we have an extent at offset 0 which may
3401 * or may not be inlined. All these cases are dealt the same way.
3403 if (i_size_read(dst
) > datal
) {
3405 * If the destination inode has an inline extent...
3406 * This would require copying the data from the source inline
3407 * extent into the beginning of the destination's inline extent.
3408 * But this is really complex, both extents can be compressed
3409 * or just one of them, which would require decompressing and
3410 * re-compressing data (which could increase the new compressed
3411 * size, not allowing the compressed data to fit anymore in an
3413 * So just don't support this case for now (it should be rare,
3414 * we are not really saving space when cloning inline extents).
3419 btrfs_release_path(path
);
3420 ret
= btrfs_drop_extents(trans
, root
, dst
, drop_start
, aligned_end
, 1);
3423 ret
= btrfs_insert_empty_item(trans
, root
, path
, new_key
, size
);
3428 const u32 start
= btrfs_file_extent_calc_inline_size(0);
3430 memmove(inline_data
+ start
, inline_data
+ start
+ skip
, datal
);
3433 write_extent_buffer(path
->nodes
[0], inline_data
,
3434 btrfs_item_ptr_offset(path
->nodes
[0],
3437 inode_add_bytes(dst
, datal
);
3443 * btrfs_clone() - clone a range from inode file to another
3445 * @src: Inode to clone from
3446 * @inode: Inode to clone to
3447 * @off: Offset within source to start clone from
3448 * @olen: Original length, passed by user, of range to clone
3449 * @olen_aligned: Block-aligned value of olen
3450 * @destoff: Offset within @inode to start clone
3451 * @no_time_update: Whether to update mtime/ctime on the target inode
3453 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3454 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3455 const u64 destoff
, int no_time_update
)
3457 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3458 struct btrfs_path
*path
= NULL
;
3459 struct extent_buffer
*leaf
;
3460 struct btrfs_trans_handle
*trans
;
3462 struct btrfs_key key
;
3466 const u64 len
= olen_aligned
;
3467 u64 last_dest_end
= destoff
;
3470 buf
= vmalloc(root
->nodesize
);
3474 path
= btrfs_alloc_path();
3480 path
->reada
= READA_FORWARD
;
3482 key
.objectid
= btrfs_ino(src
);
3483 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3487 u64 next_key_min_offset
= key
.offset
+ 1;
3490 * note the key will change type as we walk through the
3493 path
->leave_spinning
= 1;
3494 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3499 * First search, if no extent item that starts at offset off was
3500 * found but the previous item is an extent item, it's possible
3501 * it might overlap our target range, therefore process it.
3503 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3504 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3505 path
->slots
[0] - 1);
3506 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3510 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3512 if (path
->slots
[0] >= nritems
) {
3513 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3518 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3520 leaf
= path
->nodes
[0];
3521 slot
= path
->slots
[0];
3523 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3524 if (key
.type
> BTRFS_EXTENT_DATA_KEY
||
3525 key
.objectid
!= btrfs_ino(src
))
3528 if (key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3529 struct btrfs_file_extent_item
*extent
;
3532 struct btrfs_key new_key
;
3533 u64 disko
= 0, diskl
= 0;
3534 u64 datao
= 0, datal
= 0;
3538 extent
= btrfs_item_ptr(leaf
, slot
,
3539 struct btrfs_file_extent_item
);
3540 comp
= btrfs_file_extent_compression(leaf
, extent
);
3541 type
= btrfs_file_extent_type(leaf
, extent
);
3542 if (type
== BTRFS_FILE_EXTENT_REG
||
3543 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3544 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3546 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3548 datao
= btrfs_file_extent_offset(leaf
, extent
);
3549 datal
= btrfs_file_extent_num_bytes(leaf
,
3551 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3552 /* take upper bound, may be compressed */
3553 datal
= btrfs_file_extent_ram_bytes(leaf
,
3558 * The first search might have left us at an extent
3559 * item that ends before our target range's start, can
3560 * happen if we have holes and NO_HOLES feature enabled.
3562 if (key
.offset
+ datal
<= off
) {
3565 } else if (key
.offset
>= off
+ len
) {
3568 next_key_min_offset
= key
.offset
+ datal
;
3569 size
= btrfs_item_size_nr(leaf
, slot
);
3570 read_extent_buffer(leaf
, buf
,
3571 btrfs_item_ptr_offset(leaf
, slot
),
3574 btrfs_release_path(path
);
3575 path
->leave_spinning
= 0;
3577 memcpy(&new_key
, &key
, sizeof(new_key
));
3578 new_key
.objectid
= btrfs_ino(inode
);
3579 if (off
<= key
.offset
)
3580 new_key
.offset
= key
.offset
+ destoff
- off
;
3582 new_key
.offset
= destoff
;
3585 * Deal with a hole that doesn't have an extent item
3586 * that represents it (NO_HOLES feature enabled).
3587 * This hole is either in the middle of the cloning
3588 * range or at the beginning (fully overlaps it or
3589 * partially overlaps it).
3591 if (new_key
.offset
!= last_dest_end
)
3592 drop_start
= last_dest_end
;
3594 drop_start
= new_key
.offset
;
3597 * 1 - adjusting old extent (we may have to split it)
3598 * 1 - add new extent
3601 trans
= btrfs_start_transaction(root
, 3);
3602 if (IS_ERR(trans
)) {
3603 ret
= PTR_ERR(trans
);
3607 if (type
== BTRFS_FILE_EXTENT_REG
||
3608 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3610 * a | --- range to clone ---| b
3611 * | ------------- extent ------------- |
3614 /* subtract range b */
3615 if (key
.offset
+ datal
> off
+ len
)
3616 datal
= off
+ len
- key
.offset
;
3618 /* subtract range a */
3619 if (off
> key
.offset
) {
3620 datao
+= off
- key
.offset
;
3621 datal
-= off
- key
.offset
;
3624 ret
= btrfs_drop_extents(trans
, root
, inode
,
3626 new_key
.offset
+ datal
,
3629 if (ret
!= -EOPNOTSUPP
)
3630 btrfs_abort_transaction(trans
,
3632 btrfs_end_transaction(trans
, root
);
3636 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3639 btrfs_abort_transaction(trans
, root
,
3641 btrfs_end_transaction(trans
, root
);
3645 leaf
= path
->nodes
[0];
3646 slot
= path
->slots
[0];
3647 write_extent_buffer(leaf
, buf
,
3648 btrfs_item_ptr_offset(leaf
, slot
),
3651 extent
= btrfs_item_ptr(leaf
, slot
,
3652 struct btrfs_file_extent_item
);
3654 /* disko == 0 means it's a hole */
3658 btrfs_set_file_extent_offset(leaf
, extent
,
3660 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3664 inode_add_bytes(inode
, datal
);
3665 ret
= btrfs_inc_extent_ref(trans
, root
,
3667 root
->root_key
.objectid
,
3669 new_key
.offset
- datao
);
3671 btrfs_abort_transaction(trans
,
3674 btrfs_end_transaction(trans
,
3680 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3684 if (off
> key
.offset
) {
3685 skip
= off
- key
.offset
;
3686 new_key
.offset
+= skip
;
3689 if (key
.offset
+ datal
> off
+ len
)
3690 trim
= key
.offset
+ datal
- (off
+ len
);
3692 if (comp
&& (skip
|| trim
)) {
3694 btrfs_end_transaction(trans
, root
);
3697 size
-= skip
+ trim
;
3698 datal
-= skip
+ trim
;
3700 ret
= clone_copy_inline_extent(src
, inode
,
3707 if (ret
!= -EOPNOTSUPP
)
3708 btrfs_abort_transaction(trans
,
3711 btrfs_end_transaction(trans
, root
);
3714 leaf
= path
->nodes
[0];
3715 slot
= path
->slots
[0];
3718 /* If we have an implicit hole (NO_HOLES feature). */
3719 if (drop_start
< new_key
.offset
)
3720 clone_update_extent_map(inode
, trans
,
3722 new_key
.offset
- drop_start
);
3724 clone_update_extent_map(inode
, trans
, path
, 0, 0);
3726 btrfs_mark_buffer_dirty(leaf
);
3727 btrfs_release_path(path
);
3729 last_dest_end
= ALIGN(new_key
.offset
+ datal
,
3731 ret
= clone_finish_inode_update(trans
, inode
,
3737 if (new_key
.offset
+ datal
>= destoff
+ len
)
3740 btrfs_release_path(path
);
3741 key
.offset
= next_key_min_offset
;
3745 if (last_dest_end
< destoff
+ len
) {
3747 * We have an implicit hole (NO_HOLES feature is enabled) that
3748 * fully or partially overlaps our cloning range at its end.
3750 btrfs_release_path(path
);
3753 * 1 - remove extent(s)
3756 trans
= btrfs_start_transaction(root
, 2);
3757 if (IS_ERR(trans
)) {
3758 ret
= PTR_ERR(trans
);
3761 ret
= btrfs_drop_extents(trans
, root
, inode
,
3762 last_dest_end
, destoff
+ len
, 1);
3764 if (ret
!= -EOPNOTSUPP
)
3765 btrfs_abort_transaction(trans
, root
, ret
);
3766 btrfs_end_transaction(trans
, root
);
3769 clone_update_extent_map(inode
, trans
, NULL
, last_dest_end
,
3770 destoff
+ len
- last_dest_end
);
3771 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3772 destoff
, olen
, no_time_update
);
3776 btrfs_free_path(path
);
3781 static noinline
int btrfs_clone_files(struct file
*file
, struct file
*file_src
,
3782 u64 off
, u64 olen
, u64 destoff
)
3784 struct inode
*inode
= file_inode(file
);
3785 struct inode
*src
= file_inode(file_src
);
3786 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3789 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
3790 int same_inode
= src
== inode
;
3794 * - split compressed inline extents. annoying: we need to
3795 * decompress into destination's address_space (the file offset
3796 * may change, so source mapping won't do), then recompress (or
3797 * otherwise reinsert) a subrange.
3799 * - split destination inode's inline extents. The inline extents can
3800 * be either compressed or non-compressed.
3803 if (btrfs_root_readonly(root
))
3806 if (file_src
->f_path
.mnt
!= file
->f_path
.mnt
||
3807 src
->i_sb
!= inode
->i_sb
)
3810 /* don't make the dst file partly checksummed */
3811 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3812 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3815 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3819 btrfs_double_inode_lock(src
, inode
);
3824 /* determine range to clone */
3826 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3829 olen
= len
= src
->i_size
- off
;
3830 /* if we extend to eof, continue to block boundary */
3831 if (off
+ len
== src
->i_size
)
3832 len
= ALIGN(src
->i_size
, bs
) - off
;
3839 /* verify the end result is block aligned */
3840 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3841 !IS_ALIGNED(destoff
, bs
))
3844 /* verify if ranges are overlapped within the same file */
3846 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3850 if (destoff
> inode
->i_size
) {
3851 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3857 * Lock the target range too. Right after we replace the file extent
3858 * items in the fs tree (which now point to the cloned data), we might
3859 * have a worker replace them with extent items relative to a write
3860 * operation that was issued before this clone operation (i.e. confront
3861 * with inode.c:btrfs_finish_ordered_io).
3864 u64 lock_start
= min_t(u64
, off
, destoff
);
3865 u64 lock_len
= max_t(u64
, off
, destoff
) + len
- lock_start
;
3867 ret
= lock_extent_range(src
, lock_start
, lock_len
, true);
3869 ret
= btrfs_double_extent_lock(src
, off
, inode
, destoff
, len
,
3874 /* ranges in the io trees already unlocked */
3878 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
, 0);
3881 u64 lock_start
= min_t(u64
, off
, destoff
);
3882 u64 lock_end
= max_t(u64
, off
, destoff
) + len
- 1;
3884 unlock_extent(&BTRFS_I(src
)->io_tree
, lock_start
, lock_end
);
3886 btrfs_double_extent_unlock(src
, off
, inode
, destoff
, len
);
3889 * Truncate page cache pages so that future reads will see the cloned
3890 * data immediately and not the previous data.
3892 truncate_inode_pages_range(&inode
->i_data
, destoff
,
3893 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
3896 btrfs_double_inode_unlock(src
, inode
);
3902 ssize_t
btrfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
3903 struct file
*file_out
, loff_t pos_out
,
3904 size_t len
, unsigned int flags
)
3908 ret
= btrfs_clone_files(file_out
, file_in
, pos_in
, len
, pos_out
);
3914 int btrfs_clone_file_range(struct file
*src_file
, loff_t off
,
3915 struct file
*dst_file
, loff_t destoff
, u64 len
)
3917 return btrfs_clone_files(dst_file
, src_file
, off
, len
, destoff
);
3921 * there are many ways the trans_start and trans_end ioctls can lead
3922 * to deadlocks. They should only be used by applications that
3923 * basically own the machine, and have a very in depth understanding
3924 * of all the possible deadlocks and enospc problems.
3926 static long btrfs_ioctl_trans_start(struct file
*file
)
3928 struct inode
*inode
= file_inode(file
);
3929 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3930 struct btrfs_trans_handle
*trans
;
3934 if (!capable(CAP_SYS_ADMIN
))
3938 if (file
->private_data
)
3942 if (btrfs_root_readonly(root
))
3945 ret
= mnt_want_write_file(file
);
3949 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
3952 trans
= btrfs_start_ioctl_transaction(root
);
3956 file
->private_data
= trans
;
3960 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3961 mnt_drop_write_file(file
);
3966 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
3968 struct inode
*inode
= file_inode(file
);
3969 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3970 struct btrfs_root
*new_root
;
3971 struct btrfs_dir_item
*di
;
3972 struct btrfs_trans_handle
*trans
;
3973 struct btrfs_path
*path
;
3974 struct btrfs_key location
;
3975 struct btrfs_disk_key disk_key
;
3980 if (!capable(CAP_SYS_ADMIN
))
3983 ret
= mnt_want_write_file(file
);
3987 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
3993 objectid
= BTRFS_FS_TREE_OBJECTID
;
3995 location
.objectid
= objectid
;
3996 location
.type
= BTRFS_ROOT_ITEM_KEY
;
3997 location
.offset
= (u64
)-1;
3999 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
4000 if (IS_ERR(new_root
)) {
4001 ret
= PTR_ERR(new_root
);
4005 path
= btrfs_alloc_path();
4010 path
->leave_spinning
= 1;
4012 trans
= btrfs_start_transaction(root
, 1);
4013 if (IS_ERR(trans
)) {
4014 btrfs_free_path(path
);
4015 ret
= PTR_ERR(trans
);
4019 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
4020 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
4021 dir_id
, "default", 7, 1);
4022 if (IS_ERR_OR_NULL(di
)) {
4023 btrfs_free_path(path
);
4024 btrfs_end_transaction(trans
, root
);
4025 btrfs_err(new_root
->fs_info
, "Umm, you don't have the default dir"
4026 "item, this isn't going to work");
4031 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
4032 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
4033 btrfs_mark_buffer_dirty(path
->nodes
[0]);
4034 btrfs_free_path(path
);
4036 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
4037 btrfs_end_transaction(trans
, root
);
4039 mnt_drop_write_file(file
);
4043 void btrfs_get_block_group_info(struct list_head
*groups_list
,
4044 struct btrfs_ioctl_space_info
*space
)
4046 struct btrfs_block_group_cache
*block_group
;
4048 space
->total_bytes
= 0;
4049 space
->used_bytes
= 0;
4051 list_for_each_entry(block_group
, groups_list
, list
) {
4052 space
->flags
= block_group
->flags
;
4053 space
->total_bytes
+= block_group
->key
.offset
;
4054 space
->used_bytes
+=
4055 btrfs_block_group_used(&block_group
->item
);
4059 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
4061 struct btrfs_ioctl_space_args space_args
;
4062 struct btrfs_ioctl_space_info space
;
4063 struct btrfs_ioctl_space_info
*dest
;
4064 struct btrfs_ioctl_space_info
*dest_orig
;
4065 struct btrfs_ioctl_space_info __user
*user_dest
;
4066 struct btrfs_space_info
*info
;
4067 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
4068 BTRFS_BLOCK_GROUP_SYSTEM
,
4069 BTRFS_BLOCK_GROUP_METADATA
,
4070 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
4077 if (copy_from_user(&space_args
,
4078 (struct btrfs_ioctl_space_args __user
*)arg
,
4079 sizeof(space_args
)))
4082 for (i
= 0; i
< num_types
; i
++) {
4083 struct btrfs_space_info
*tmp
;
4087 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4089 if (tmp
->flags
== types
[i
]) {
4099 down_read(&info
->groups_sem
);
4100 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4101 if (!list_empty(&info
->block_groups
[c
]))
4104 up_read(&info
->groups_sem
);
4108 * Global block reserve, exported as a space_info
4112 /* space_slots == 0 means they are asking for a count */
4113 if (space_args
.space_slots
== 0) {
4114 space_args
.total_spaces
= slot_count
;
4118 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
4120 alloc_size
= sizeof(*dest
) * slot_count
;
4122 /* we generally have at most 6 or so space infos, one for each raid
4123 * level. So, a whole page should be more than enough for everyone
4125 if (alloc_size
> PAGE_CACHE_SIZE
)
4128 space_args
.total_spaces
= 0;
4129 dest
= kmalloc(alloc_size
, GFP_KERNEL
);
4134 /* now we have a buffer to copy into */
4135 for (i
= 0; i
< num_types
; i
++) {
4136 struct btrfs_space_info
*tmp
;
4143 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4145 if (tmp
->flags
== types
[i
]) {
4154 down_read(&info
->groups_sem
);
4155 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4156 if (!list_empty(&info
->block_groups
[c
])) {
4157 btrfs_get_block_group_info(
4158 &info
->block_groups
[c
], &space
);
4159 memcpy(dest
, &space
, sizeof(space
));
4161 space_args
.total_spaces
++;
4167 up_read(&info
->groups_sem
);
4171 * Add global block reserve
4174 struct btrfs_block_rsv
*block_rsv
= &root
->fs_info
->global_block_rsv
;
4176 spin_lock(&block_rsv
->lock
);
4177 space
.total_bytes
= block_rsv
->size
;
4178 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4179 spin_unlock(&block_rsv
->lock
);
4180 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4181 memcpy(dest
, &space
, sizeof(space
));
4182 space_args
.total_spaces
++;
4185 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4186 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4188 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4193 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4200 * there are many ways the trans_start and trans_end ioctls can lead
4201 * to deadlocks. They should only be used by applications that
4202 * basically own the machine, and have a very in depth understanding
4203 * of all the possible deadlocks and enospc problems.
4205 long btrfs_ioctl_trans_end(struct file
*file
)
4207 struct inode
*inode
= file_inode(file
);
4208 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4209 struct btrfs_trans_handle
*trans
;
4211 trans
= file
->private_data
;
4214 file
->private_data
= NULL
;
4216 btrfs_end_transaction(trans
, root
);
4218 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4220 mnt_drop_write_file(file
);
4224 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4227 struct btrfs_trans_handle
*trans
;
4231 trans
= btrfs_attach_transaction_barrier(root
);
4232 if (IS_ERR(trans
)) {
4233 if (PTR_ERR(trans
) != -ENOENT
)
4234 return PTR_ERR(trans
);
4236 /* No running transaction, don't bother */
4237 transid
= root
->fs_info
->last_trans_committed
;
4240 transid
= trans
->transid
;
4241 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
4243 btrfs_end_transaction(trans
, root
);
4248 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4253 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
4259 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4262 transid
= 0; /* current trans */
4264 return btrfs_wait_for_commit(root
, transid
);
4267 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4269 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4270 struct btrfs_ioctl_scrub_args
*sa
;
4273 if (!capable(CAP_SYS_ADMIN
))
4276 sa
= memdup_user(arg
, sizeof(*sa
));
4280 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4281 ret
= mnt_want_write_file(file
);
4286 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4287 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4290 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4293 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4294 mnt_drop_write_file(file
);
4300 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
4302 if (!capable(CAP_SYS_ADMIN
))
4305 return btrfs_scrub_cancel(root
->fs_info
);
4308 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
4311 struct btrfs_ioctl_scrub_args
*sa
;
4314 if (!capable(CAP_SYS_ADMIN
))
4317 sa
= memdup_user(arg
, sizeof(*sa
));
4321 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
4323 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4330 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
4333 struct btrfs_ioctl_get_dev_stats
*sa
;
4336 sa
= memdup_user(arg
, sizeof(*sa
));
4340 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4345 ret
= btrfs_get_dev_stats(root
, sa
);
4347 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4354 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
4356 struct btrfs_ioctl_dev_replace_args
*p
;
4359 if (!capable(CAP_SYS_ADMIN
))
4362 p
= memdup_user(arg
, sizeof(*p
));
4367 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4368 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
) {
4373 &root
->fs_info
->mutually_exclusive_operation_running
,
4375 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4377 ret
= btrfs_dev_replace_start(root
, p
);
4379 &root
->fs_info
->mutually_exclusive_operation_running
,
4383 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4384 btrfs_dev_replace_status(root
->fs_info
, p
);
4387 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4388 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
4395 if (copy_to_user(arg
, p
, sizeof(*p
)))
4402 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4408 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4409 struct inode_fs_paths
*ipath
= NULL
;
4410 struct btrfs_path
*path
;
4412 if (!capable(CAP_DAC_READ_SEARCH
))
4415 path
= btrfs_alloc_path();
4421 ipa
= memdup_user(arg
, sizeof(*ipa
));
4428 size
= min_t(u32
, ipa
->size
, 4096);
4429 ipath
= init_ipath(size
, root
, path
);
4430 if (IS_ERR(ipath
)) {
4431 ret
= PTR_ERR(ipath
);
4436 ret
= paths_from_inode(ipa
->inum
, ipath
);
4440 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4441 rel_ptr
= ipath
->fspath
->val
[i
] -
4442 (u64
)(unsigned long)ipath
->fspath
->val
;
4443 ipath
->fspath
->val
[i
] = rel_ptr
;
4446 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
4447 (void *)(unsigned long)ipath
->fspath
, size
);
4454 btrfs_free_path(path
);
4461 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4463 struct btrfs_data_container
*inodes
= ctx
;
4464 const size_t c
= 3 * sizeof(u64
);
4466 if (inodes
->bytes_left
>= c
) {
4467 inodes
->bytes_left
-= c
;
4468 inodes
->val
[inodes
->elem_cnt
] = inum
;
4469 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4470 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4471 inodes
->elem_cnt
+= 3;
4473 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4474 inodes
->bytes_left
= 0;
4475 inodes
->elem_missed
+= 3;
4481 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
4486 struct btrfs_ioctl_logical_ino_args
*loi
;
4487 struct btrfs_data_container
*inodes
= NULL
;
4488 struct btrfs_path
*path
= NULL
;
4490 if (!capable(CAP_SYS_ADMIN
))
4493 loi
= memdup_user(arg
, sizeof(*loi
));
4500 path
= btrfs_alloc_path();
4506 size
= min_t(u32
, loi
->size
, SZ_64K
);
4507 inodes
= init_data_container(size
);
4508 if (IS_ERR(inodes
)) {
4509 ret
= PTR_ERR(inodes
);
4514 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
4515 build_ino_list
, inodes
);
4521 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
4522 (void *)(unsigned long)inodes
, size
);
4527 btrfs_free_path(path
);
4534 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
4535 struct btrfs_ioctl_balance_args
*bargs
)
4537 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4539 bargs
->flags
= bctl
->flags
;
4541 if (atomic_read(&fs_info
->balance_running
))
4542 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4543 if (atomic_read(&fs_info
->balance_pause_req
))
4544 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4545 if (atomic_read(&fs_info
->balance_cancel_req
))
4546 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4548 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4549 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4550 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4553 spin_lock(&fs_info
->balance_lock
);
4554 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4555 spin_unlock(&fs_info
->balance_lock
);
4557 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4561 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4563 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4564 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4565 struct btrfs_ioctl_balance_args
*bargs
;
4566 struct btrfs_balance_control
*bctl
;
4567 bool need_unlock
; /* for mut. excl. ops lock */
4570 if (!capable(CAP_SYS_ADMIN
))
4573 ret
= mnt_want_write_file(file
);
4578 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4579 mutex_lock(&fs_info
->volume_mutex
);
4580 mutex_lock(&fs_info
->balance_mutex
);
4586 * mut. excl. ops lock is locked. Three possibilites:
4587 * (1) some other op is running
4588 * (2) balance is running
4589 * (3) balance is paused -- special case (think resume)
4591 mutex_lock(&fs_info
->balance_mutex
);
4592 if (fs_info
->balance_ctl
) {
4593 /* this is either (2) or (3) */
4594 if (!atomic_read(&fs_info
->balance_running
)) {
4595 mutex_unlock(&fs_info
->balance_mutex
);
4596 if (!mutex_trylock(&fs_info
->volume_mutex
))
4598 mutex_lock(&fs_info
->balance_mutex
);
4600 if (fs_info
->balance_ctl
&&
4601 !atomic_read(&fs_info
->balance_running
)) {
4603 need_unlock
= false;
4607 mutex_unlock(&fs_info
->balance_mutex
);
4608 mutex_unlock(&fs_info
->volume_mutex
);
4612 mutex_unlock(&fs_info
->balance_mutex
);
4618 mutex_unlock(&fs_info
->balance_mutex
);
4619 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4624 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4627 bargs
= memdup_user(arg
, sizeof(*bargs
));
4628 if (IS_ERR(bargs
)) {
4629 ret
= PTR_ERR(bargs
);
4633 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4634 if (!fs_info
->balance_ctl
) {
4639 bctl
= fs_info
->balance_ctl
;
4640 spin_lock(&fs_info
->balance_lock
);
4641 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4642 spin_unlock(&fs_info
->balance_lock
);
4650 if (fs_info
->balance_ctl
) {
4655 bctl
= kzalloc(sizeof(*bctl
), GFP_KERNEL
);
4661 bctl
->fs_info
= fs_info
;
4663 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4664 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4665 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4667 bctl
->flags
= bargs
->flags
;
4669 /* balance everything - no filters */
4670 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4673 if (bctl
->flags
& ~(BTRFS_BALANCE_ARGS_MASK
| BTRFS_BALANCE_TYPE_MASK
)) {
4680 * Ownership of bctl and mutually_exclusive_operation_running
4681 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4682 * or, if restriper was paused all the way until unmount, in
4683 * free_fs_info. mutually_exclusive_operation_running is
4684 * cleared in __cancel_balance.
4686 need_unlock
= false;
4688 ret
= btrfs_balance(bctl
, bargs
);
4692 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4701 mutex_unlock(&fs_info
->balance_mutex
);
4702 mutex_unlock(&fs_info
->volume_mutex
);
4704 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4706 mnt_drop_write_file(file
);
4710 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4712 if (!capable(CAP_SYS_ADMIN
))
4716 case BTRFS_BALANCE_CTL_PAUSE
:
4717 return btrfs_pause_balance(root
->fs_info
);
4718 case BTRFS_BALANCE_CTL_CANCEL
:
4719 return btrfs_cancel_balance(root
->fs_info
);
4725 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4728 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4729 struct btrfs_ioctl_balance_args
*bargs
;
4732 if (!capable(CAP_SYS_ADMIN
))
4735 mutex_lock(&fs_info
->balance_mutex
);
4736 if (!fs_info
->balance_ctl
) {
4741 bargs
= kzalloc(sizeof(*bargs
), GFP_KERNEL
);
4747 update_ioctl_balance_args(fs_info
, 1, bargs
);
4749 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4754 mutex_unlock(&fs_info
->balance_mutex
);
4758 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4760 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4761 struct btrfs_ioctl_quota_ctl_args
*sa
;
4762 struct btrfs_trans_handle
*trans
= NULL
;
4766 if (!capable(CAP_SYS_ADMIN
))
4769 ret
= mnt_want_write_file(file
);
4773 sa
= memdup_user(arg
, sizeof(*sa
));
4779 down_write(&root
->fs_info
->subvol_sem
);
4780 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
4781 if (IS_ERR(trans
)) {
4782 ret
= PTR_ERR(trans
);
4787 case BTRFS_QUOTA_CTL_ENABLE
:
4788 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
4790 case BTRFS_QUOTA_CTL_DISABLE
:
4791 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
4798 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
4803 up_write(&root
->fs_info
->subvol_sem
);
4805 mnt_drop_write_file(file
);
4809 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4811 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4812 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4813 struct btrfs_trans_handle
*trans
;
4817 if (!capable(CAP_SYS_ADMIN
))
4820 ret
= mnt_want_write_file(file
);
4824 sa
= memdup_user(arg
, sizeof(*sa
));
4830 trans
= btrfs_join_transaction(root
);
4831 if (IS_ERR(trans
)) {
4832 ret
= PTR_ERR(trans
);
4836 /* FIXME: check if the IDs really exist */
4838 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
4841 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
4845 /* update qgroup status and info */
4846 err
= btrfs_run_qgroups(trans
, root
->fs_info
);
4848 btrfs_std_error(root
->fs_info
, ret
,
4849 "failed to update qgroup status and info\n");
4850 err
= btrfs_end_transaction(trans
, root
);
4857 mnt_drop_write_file(file
);
4861 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4863 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4864 struct btrfs_ioctl_qgroup_create_args
*sa
;
4865 struct btrfs_trans_handle
*trans
;
4869 if (!capable(CAP_SYS_ADMIN
))
4872 ret
= mnt_want_write_file(file
);
4876 sa
= memdup_user(arg
, sizeof(*sa
));
4882 if (!sa
->qgroupid
) {
4887 trans
= btrfs_join_transaction(root
);
4888 if (IS_ERR(trans
)) {
4889 ret
= PTR_ERR(trans
);
4893 /* FIXME: check if the IDs really exist */
4895 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4897 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4900 err
= btrfs_end_transaction(trans
, root
);
4907 mnt_drop_write_file(file
);
4911 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4913 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4914 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4915 struct btrfs_trans_handle
*trans
;
4920 if (!capable(CAP_SYS_ADMIN
))
4923 ret
= mnt_want_write_file(file
);
4927 sa
= memdup_user(arg
, sizeof(*sa
));
4933 trans
= btrfs_join_transaction(root
);
4934 if (IS_ERR(trans
)) {
4935 ret
= PTR_ERR(trans
);
4939 qgroupid
= sa
->qgroupid
;
4941 /* take the current subvol as qgroup */
4942 qgroupid
= root
->root_key
.objectid
;
4945 /* FIXME: check if the IDs really exist */
4946 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
4948 err
= btrfs_end_transaction(trans
, root
);
4955 mnt_drop_write_file(file
);
4959 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
4961 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4962 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4965 if (!capable(CAP_SYS_ADMIN
))
4968 ret
= mnt_want_write_file(file
);
4972 qsa
= memdup_user(arg
, sizeof(*qsa
));
4983 ret
= btrfs_qgroup_rescan(root
->fs_info
);
4988 mnt_drop_write_file(file
);
4992 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
4994 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4995 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4998 if (!capable(CAP_SYS_ADMIN
))
5001 qsa
= kzalloc(sizeof(*qsa
), GFP_KERNEL
);
5005 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
5007 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
5010 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
5017 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
5019 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5021 if (!capable(CAP_SYS_ADMIN
))
5024 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
5027 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
5028 struct btrfs_ioctl_received_subvol_args
*sa
)
5030 struct inode
*inode
= file_inode(file
);
5031 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5032 struct btrfs_root_item
*root_item
= &root
->root_item
;
5033 struct btrfs_trans_handle
*trans
;
5034 struct timespec ct
= CURRENT_TIME
;
5036 int received_uuid_changed
;
5038 if (!inode_owner_or_capable(inode
))
5041 ret
= mnt_want_write_file(file
);
5045 down_write(&root
->fs_info
->subvol_sem
);
5047 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
5052 if (btrfs_root_readonly(root
)) {
5059 * 2 - uuid items (received uuid + subvol uuid)
5061 trans
= btrfs_start_transaction(root
, 3);
5062 if (IS_ERR(trans
)) {
5063 ret
= PTR_ERR(trans
);
5068 sa
->rtransid
= trans
->transid
;
5069 sa
->rtime
.sec
= ct
.tv_sec
;
5070 sa
->rtime
.nsec
= ct
.tv_nsec
;
5072 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
5074 if (received_uuid_changed
&&
5075 !btrfs_is_empty_uuid(root_item
->received_uuid
))
5076 btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
5077 root_item
->received_uuid
,
5078 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5079 root
->root_key
.objectid
);
5080 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
5081 btrfs_set_root_stransid(root_item
, sa
->stransid
);
5082 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
5083 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
5084 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
5085 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
5086 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
5088 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
5089 &root
->root_key
, &root
->root_item
);
5091 btrfs_end_transaction(trans
, root
);
5094 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
5095 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
5097 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5098 root
->root_key
.objectid
);
5099 if (ret
< 0 && ret
!= -EEXIST
) {
5100 btrfs_abort_transaction(trans
, root
, ret
);
5104 ret
= btrfs_commit_transaction(trans
, root
);
5106 btrfs_abort_transaction(trans
, root
, ret
);
5111 up_write(&root
->fs_info
->subvol_sem
);
5112 mnt_drop_write_file(file
);
5117 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
5120 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
5121 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
5124 args32
= memdup_user(arg
, sizeof(*args32
));
5125 if (IS_ERR(args32
)) {
5126 ret
= PTR_ERR(args32
);
5131 args64
= kmalloc(sizeof(*args64
), GFP_KERNEL
);
5137 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
5138 args64
->stransid
= args32
->stransid
;
5139 args64
->rtransid
= args32
->rtransid
;
5140 args64
->stime
.sec
= args32
->stime
.sec
;
5141 args64
->stime
.nsec
= args32
->stime
.nsec
;
5142 args64
->rtime
.sec
= args32
->rtime
.sec
;
5143 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
5144 args64
->flags
= args32
->flags
;
5146 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
5150 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
5151 args32
->stransid
= args64
->stransid
;
5152 args32
->rtransid
= args64
->rtransid
;
5153 args32
->stime
.sec
= args64
->stime
.sec
;
5154 args32
->stime
.nsec
= args64
->stime
.nsec
;
5155 args32
->rtime
.sec
= args64
->rtime
.sec
;
5156 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
5157 args32
->flags
= args64
->flags
;
5159 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
5170 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
5173 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
5176 sa
= memdup_user(arg
, sizeof(*sa
));
5183 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5188 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5197 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5199 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5202 char label
[BTRFS_LABEL_SIZE
];
5204 spin_lock(&root
->fs_info
->super_lock
);
5205 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5206 spin_unlock(&root
->fs_info
->super_lock
);
5208 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5210 if (len
== BTRFS_LABEL_SIZE
) {
5211 btrfs_warn(root
->fs_info
,
5212 "label is too long, return the first %zu bytes", --len
);
5215 ret
= copy_to_user(arg
, label
, len
);
5217 return ret
? -EFAULT
: 0;
5220 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5222 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5223 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5224 struct btrfs_trans_handle
*trans
;
5225 char label
[BTRFS_LABEL_SIZE
];
5228 if (!capable(CAP_SYS_ADMIN
))
5231 if (copy_from_user(label
, arg
, sizeof(label
)))
5234 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5235 btrfs_err(root
->fs_info
, "unable to set label with more than %d bytes",
5236 BTRFS_LABEL_SIZE
- 1);
5240 ret
= mnt_want_write_file(file
);
5244 trans
= btrfs_start_transaction(root
, 0);
5245 if (IS_ERR(trans
)) {
5246 ret
= PTR_ERR(trans
);
5250 spin_lock(&root
->fs_info
->super_lock
);
5251 strcpy(super_block
->label
, label
);
5252 spin_unlock(&root
->fs_info
->super_lock
);
5253 ret
= btrfs_commit_transaction(trans
, root
);
5256 mnt_drop_write_file(file
);
5260 #define INIT_FEATURE_FLAGS(suffix) \
5261 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5262 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5263 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5265 static int btrfs_ioctl_get_supported_features(struct file
*file
,
5268 static const struct btrfs_ioctl_feature_flags features
[3] = {
5269 INIT_FEATURE_FLAGS(SUPP
),
5270 INIT_FEATURE_FLAGS(SAFE_SET
),
5271 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5274 if (copy_to_user(arg
, &features
, sizeof(features
)))
5280 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5282 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5283 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5284 struct btrfs_ioctl_feature_flags features
;
5286 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5287 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5288 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5290 if (copy_to_user(arg
, &features
, sizeof(features
)))
5296 static int check_feature_bits(struct btrfs_root
*root
,
5297 enum btrfs_feature_set set
,
5298 u64 change_mask
, u64 flags
, u64 supported_flags
,
5299 u64 safe_set
, u64 safe_clear
)
5301 const char *type
= btrfs_feature_set_names
[set
];
5303 u64 disallowed
, unsupported
;
5304 u64 set_mask
= flags
& change_mask
;
5305 u64 clear_mask
= ~flags
& change_mask
;
5307 unsupported
= set_mask
& ~supported_flags
;
5309 names
= btrfs_printable_features(set
, unsupported
);
5311 btrfs_warn(root
->fs_info
,
5312 "this kernel does not support the %s feature bit%s",
5313 names
, strchr(names
, ',') ? "s" : "");
5316 btrfs_warn(root
->fs_info
,
5317 "this kernel does not support %s bits 0x%llx",
5322 disallowed
= set_mask
& ~safe_set
;
5324 names
= btrfs_printable_features(set
, disallowed
);
5326 btrfs_warn(root
->fs_info
,
5327 "can't set the %s feature bit%s while mounted",
5328 names
, strchr(names
, ',') ? "s" : "");
5331 btrfs_warn(root
->fs_info
,
5332 "can't set %s bits 0x%llx while mounted",
5337 disallowed
= clear_mask
& ~safe_clear
;
5339 names
= btrfs_printable_features(set
, disallowed
);
5341 btrfs_warn(root
->fs_info
,
5342 "can't clear the %s feature bit%s while mounted",
5343 names
, strchr(names
, ',') ? "s" : "");
5346 btrfs_warn(root
->fs_info
,
5347 "can't clear %s bits 0x%llx while mounted",
5355 #define check_feature(root, change_mask, flags, mask_base) \
5356 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5357 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5358 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5359 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5361 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5363 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5364 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5365 struct btrfs_ioctl_feature_flags flags
[2];
5366 struct btrfs_trans_handle
*trans
;
5370 if (!capable(CAP_SYS_ADMIN
))
5373 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5377 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5378 !flags
[0].incompat_flags
)
5381 ret
= check_feature(root
, flags
[0].compat_flags
,
5382 flags
[1].compat_flags
, COMPAT
);
5386 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
5387 flags
[1].compat_ro_flags
, COMPAT_RO
);
5391 ret
= check_feature(root
, flags
[0].incompat_flags
,
5392 flags
[1].incompat_flags
, INCOMPAT
);
5396 trans
= btrfs_start_transaction(root
, 0);
5398 return PTR_ERR(trans
);
5400 spin_lock(&root
->fs_info
->super_lock
);
5401 newflags
= btrfs_super_compat_flags(super_block
);
5402 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5403 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5404 btrfs_set_super_compat_flags(super_block
, newflags
);
5406 newflags
= btrfs_super_compat_ro_flags(super_block
);
5407 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5408 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5409 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5411 newflags
= btrfs_super_incompat_flags(super_block
);
5412 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5413 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5414 btrfs_set_super_incompat_flags(super_block
, newflags
);
5415 spin_unlock(&root
->fs_info
->super_lock
);
5417 return btrfs_commit_transaction(trans
, root
);
5420 long btrfs_ioctl(struct file
*file
, unsigned int
5421 cmd
, unsigned long arg
)
5423 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5424 void __user
*argp
= (void __user
*)arg
;
5427 case FS_IOC_GETFLAGS
:
5428 return btrfs_ioctl_getflags(file
, argp
);
5429 case FS_IOC_SETFLAGS
:
5430 return btrfs_ioctl_setflags(file
, argp
);
5431 case FS_IOC_GETVERSION
:
5432 return btrfs_ioctl_getversion(file
, argp
);
5434 return btrfs_ioctl_fitrim(file
, argp
);
5435 case BTRFS_IOC_SNAP_CREATE
:
5436 return btrfs_ioctl_snap_create(file
, argp
, 0);
5437 case BTRFS_IOC_SNAP_CREATE_V2
:
5438 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5439 case BTRFS_IOC_SUBVOL_CREATE
:
5440 return btrfs_ioctl_snap_create(file
, argp
, 1);
5441 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5442 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5443 case BTRFS_IOC_SNAP_DESTROY
:
5444 return btrfs_ioctl_snap_destroy(file
, argp
);
5445 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5446 return btrfs_ioctl_subvol_getflags(file
, argp
);
5447 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5448 return btrfs_ioctl_subvol_setflags(file
, argp
);
5449 case BTRFS_IOC_DEFAULT_SUBVOL
:
5450 return btrfs_ioctl_default_subvol(file
, argp
);
5451 case BTRFS_IOC_DEFRAG
:
5452 return btrfs_ioctl_defrag(file
, NULL
);
5453 case BTRFS_IOC_DEFRAG_RANGE
:
5454 return btrfs_ioctl_defrag(file
, argp
);
5455 case BTRFS_IOC_RESIZE
:
5456 return btrfs_ioctl_resize(file
, argp
);
5457 case BTRFS_IOC_ADD_DEV
:
5458 return btrfs_ioctl_add_dev(root
, argp
);
5459 case BTRFS_IOC_RM_DEV
:
5460 return btrfs_ioctl_rm_dev(file
, argp
);
5461 case BTRFS_IOC_FS_INFO
:
5462 return btrfs_ioctl_fs_info(root
, argp
);
5463 case BTRFS_IOC_DEV_INFO
:
5464 return btrfs_ioctl_dev_info(root
, argp
);
5465 case BTRFS_IOC_BALANCE
:
5466 return btrfs_ioctl_balance(file
, NULL
);
5467 case BTRFS_IOC_TRANS_START
:
5468 return btrfs_ioctl_trans_start(file
);
5469 case BTRFS_IOC_TRANS_END
:
5470 return btrfs_ioctl_trans_end(file
);
5471 case BTRFS_IOC_TREE_SEARCH
:
5472 return btrfs_ioctl_tree_search(file
, argp
);
5473 case BTRFS_IOC_TREE_SEARCH_V2
:
5474 return btrfs_ioctl_tree_search_v2(file
, argp
);
5475 case BTRFS_IOC_INO_LOOKUP
:
5476 return btrfs_ioctl_ino_lookup(file
, argp
);
5477 case BTRFS_IOC_INO_PATHS
:
5478 return btrfs_ioctl_ino_to_path(root
, argp
);
5479 case BTRFS_IOC_LOGICAL_INO
:
5480 return btrfs_ioctl_logical_to_ino(root
, argp
);
5481 case BTRFS_IOC_SPACE_INFO
:
5482 return btrfs_ioctl_space_info(root
, argp
);
5483 case BTRFS_IOC_SYNC
: {
5486 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
5489 ret
= btrfs_sync_fs(file_inode(file
)->i_sb
, 1);
5491 * The transaction thread may want to do more work,
5492 * namely it pokes the cleaner ktread that will start
5493 * processing uncleaned subvols.
5495 wake_up_process(root
->fs_info
->transaction_kthread
);
5498 case BTRFS_IOC_START_SYNC
:
5499 return btrfs_ioctl_start_sync(root
, argp
);
5500 case BTRFS_IOC_WAIT_SYNC
:
5501 return btrfs_ioctl_wait_sync(root
, argp
);
5502 case BTRFS_IOC_SCRUB
:
5503 return btrfs_ioctl_scrub(file
, argp
);
5504 case BTRFS_IOC_SCRUB_CANCEL
:
5505 return btrfs_ioctl_scrub_cancel(root
, argp
);
5506 case BTRFS_IOC_SCRUB_PROGRESS
:
5507 return btrfs_ioctl_scrub_progress(root
, argp
);
5508 case BTRFS_IOC_BALANCE_V2
:
5509 return btrfs_ioctl_balance(file
, argp
);
5510 case BTRFS_IOC_BALANCE_CTL
:
5511 return btrfs_ioctl_balance_ctl(root
, arg
);
5512 case BTRFS_IOC_BALANCE_PROGRESS
:
5513 return btrfs_ioctl_balance_progress(root
, argp
);
5514 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5515 return btrfs_ioctl_set_received_subvol(file
, argp
);
5517 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5518 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5520 case BTRFS_IOC_SEND
:
5521 return btrfs_ioctl_send(file
, argp
);
5522 case BTRFS_IOC_GET_DEV_STATS
:
5523 return btrfs_ioctl_get_dev_stats(root
, argp
);
5524 case BTRFS_IOC_QUOTA_CTL
:
5525 return btrfs_ioctl_quota_ctl(file
, argp
);
5526 case BTRFS_IOC_QGROUP_ASSIGN
:
5527 return btrfs_ioctl_qgroup_assign(file
, argp
);
5528 case BTRFS_IOC_QGROUP_CREATE
:
5529 return btrfs_ioctl_qgroup_create(file
, argp
);
5530 case BTRFS_IOC_QGROUP_LIMIT
:
5531 return btrfs_ioctl_qgroup_limit(file
, argp
);
5532 case BTRFS_IOC_QUOTA_RESCAN
:
5533 return btrfs_ioctl_quota_rescan(file
, argp
);
5534 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5535 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5536 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5537 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5538 case BTRFS_IOC_DEV_REPLACE
:
5539 return btrfs_ioctl_dev_replace(root
, argp
);
5540 case BTRFS_IOC_GET_FSLABEL
:
5541 return btrfs_ioctl_get_fslabel(file
, argp
);
5542 case BTRFS_IOC_SET_FSLABEL
:
5543 return btrfs_ioctl_set_fslabel(file
, argp
);
5544 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5545 return btrfs_ioctl_get_supported_features(file
, argp
);
5546 case BTRFS_IOC_GET_FEATURES
:
5547 return btrfs_ioctl_get_features(file
, argp
);
5548 case BTRFS_IOC_SET_FEATURES
:
5549 return btrfs_ioctl_set_features(file
, argp
);