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"
65 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
66 * structures are incorrect, as the timespec structure from userspace
67 * is 4 bytes too small. We define these alternatives here to teach
68 * the kernel about the 32-bit struct packing.
70 struct btrfs_ioctl_timespec_32
{
73 } __attribute__ ((__packed__
));
75 struct btrfs_ioctl_received_subvol_args_32
{
76 char uuid
[BTRFS_UUID_SIZE
]; /* in */
77 __u64 stransid
; /* in */
78 __u64 rtransid
; /* out */
79 struct btrfs_ioctl_timespec_32 stime
; /* in */
80 struct btrfs_ioctl_timespec_32 rtime
; /* out */
82 __u64 reserved
[16]; /* in */
83 } __attribute__ ((__packed__
));
85 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
86 struct btrfs_ioctl_received_subvol_args_32)
90 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
91 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
,
94 /* Mask out flags that are inappropriate for the given type of inode. */
95 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
99 else if (S_ISREG(mode
))
100 return flags
& ~FS_DIRSYNC_FL
;
102 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
106 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
110 unsigned int iflags
= 0;
112 if (flags
& BTRFS_INODE_SYNC
)
113 iflags
|= FS_SYNC_FL
;
114 if (flags
& BTRFS_INODE_IMMUTABLE
)
115 iflags
|= FS_IMMUTABLE_FL
;
116 if (flags
& BTRFS_INODE_APPEND
)
117 iflags
|= FS_APPEND_FL
;
118 if (flags
& BTRFS_INODE_NODUMP
)
119 iflags
|= FS_NODUMP_FL
;
120 if (flags
& BTRFS_INODE_NOATIME
)
121 iflags
|= FS_NOATIME_FL
;
122 if (flags
& BTRFS_INODE_DIRSYNC
)
123 iflags
|= FS_DIRSYNC_FL
;
124 if (flags
& BTRFS_INODE_NODATACOW
)
125 iflags
|= FS_NOCOW_FL
;
127 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
128 iflags
|= FS_COMPR_FL
;
129 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
130 iflags
|= FS_NOCOMP_FL
;
136 * Update inode->i_flags based on the btrfs internal flags.
138 void btrfs_update_iflags(struct inode
*inode
)
140 struct btrfs_inode
*ip
= BTRFS_I(inode
);
141 unsigned int new_fl
= 0;
143 if (ip
->flags
& BTRFS_INODE_SYNC
)
145 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
146 new_fl
|= S_IMMUTABLE
;
147 if (ip
->flags
& BTRFS_INODE_APPEND
)
149 if (ip
->flags
& BTRFS_INODE_NOATIME
)
151 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
154 set_mask_bits(&inode
->i_flags
,
155 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
,
160 * Inherit flags from the parent inode.
162 * Currently only the compression flags and the cow flags are inherited.
164 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
171 flags
= BTRFS_I(dir
)->flags
;
173 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
174 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
175 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
176 } else if (flags
& BTRFS_INODE_COMPRESS
) {
177 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
178 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
181 if (flags
& BTRFS_INODE_NODATACOW
) {
182 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
183 if (S_ISREG(inode
->i_mode
))
184 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
187 btrfs_update_iflags(inode
);
190 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
192 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
193 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
195 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
200 static int check_flags(unsigned int flags
)
202 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
203 FS_NOATIME_FL
| FS_NODUMP_FL
| \
204 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
205 FS_NOCOMP_FL
| FS_COMPR_FL
|
209 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
215 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
217 struct inode
*inode
= file_inode(file
);
218 struct btrfs_inode
*ip
= BTRFS_I(inode
);
219 struct btrfs_root
*root
= ip
->root
;
220 struct btrfs_trans_handle
*trans
;
221 unsigned int flags
, oldflags
;
224 unsigned int i_oldflags
;
227 if (!inode_owner_or_capable(inode
))
230 if (btrfs_root_readonly(root
))
233 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
236 ret
= check_flags(flags
);
240 ret
= mnt_want_write_file(file
);
246 ip_oldflags
= ip
->flags
;
247 i_oldflags
= inode
->i_flags
;
248 mode
= inode
->i_mode
;
250 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
251 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
252 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
253 if (!capable(CAP_LINUX_IMMUTABLE
)) {
259 if (flags
& FS_SYNC_FL
)
260 ip
->flags
|= BTRFS_INODE_SYNC
;
262 ip
->flags
&= ~BTRFS_INODE_SYNC
;
263 if (flags
& FS_IMMUTABLE_FL
)
264 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
266 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
267 if (flags
& FS_APPEND_FL
)
268 ip
->flags
|= BTRFS_INODE_APPEND
;
270 ip
->flags
&= ~BTRFS_INODE_APPEND
;
271 if (flags
& FS_NODUMP_FL
)
272 ip
->flags
|= BTRFS_INODE_NODUMP
;
274 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
275 if (flags
& FS_NOATIME_FL
)
276 ip
->flags
|= BTRFS_INODE_NOATIME
;
278 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
279 if (flags
& FS_DIRSYNC_FL
)
280 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
282 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
283 if (flags
& FS_NOCOW_FL
) {
286 * It's safe to turn csums off here, no extents exist.
287 * Otherwise we want the flag to reflect the real COW
288 * status of the file and will not set it.
290 if (inode
->i_size
== 0)
291 ip
->flags
|= BTRFS_INODE_NODATACOW
292 | BTRFS_INODE_NODATASUM
;
294 ip
->flags
|= BTRFS_INODE_NODATACOW
;
298 * Revert back under same assuptions as above
301 if (inode
->i_size
== 0)
302 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
303 | BTRFS_INODE_NODATASUM
);
305 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
310 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
311 * flag may be changed automatically if compression code won't make
314 if (flags
& FS_NOCOMP_FL
) {
315 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
316 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
318 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
319 if (ret
&& ret
!= -ENODATA
)
321 } else if (flags
& FS_COMPR_FL
) {
324 ip
->flags
|= BTRFS_INODE_COMPRESS
;
325 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
327 if (root
->fs_info
->compress_type
== BTRFS_COMPRESS_LZO
)
331 ret
= btrfs_set_prop(inode
, "btrfs.compression",
332 comp
, strlen(comp
), 0);
337 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
338 if (ret
&& ret
!= -ENODATA
)
340 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
343 trans
= btrfs_start_transaction(root
, 1);
345 ret
= PTR_ERR(trans
);
349 btrfs_update_iflags(inode
);
350 inode_inc_iversion(inode
);
351 inode
->i_ctime
= current_fs_time(inode
->i_sb
);
352 ret
= btrfs_update_inode(trans
, root
, inode
);
354 btrfs_end_transaction(trans
, root
);
357 ip
->flags
= ip_oldflags
;
358 inode
->i_flags
= i_oldflags
;
363 mnt_drop_write_file(file
);
367 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
369 struct inode
*inode
= file_inode(file
);
371 return put_user(inode
->i_generation
, arg
);
374 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
376 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
377 struct btrfs_device
*device
;
378 struct request_queue
*q
;
379 struct fstrim_range range
;
380 u64 minlen
= ULLONG_MAX
;
382 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
385 if (!capable(CAP_SYS_ADMIN
))
389 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
393 q
= bdev_get_queue(device
->bdev
);
394 if (blk_queue_discard(q
)) {
396 minlen
= min((u64
)q
->limits
.discard_granularity
,
404 if (copy_from_user(&range
, arg
, sizeof(range
)))
406 if (range
.start
> total_bytes
||
407 range
.len
< fs_info
->sb
->s_blocksize
)
410 range
.len
= min(range
.len
, total_bytes
- range
.start
);
411 range
.minlen
= max(range
.minlen
, minlen
);
412 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
416 if (copy_to_user(arg
, &range
, sizeof(range
)))
422 int btrfs_is_empty_uuid(u8
*uuid
)
426 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
433 static noinline
int create_subvol(struct inode
*dir
,
434 struct dentry
*dentry
,
435 char *name
, int namelen
,
437 struct btrfs_qgroup_inherit
*inherit
)
439 struct btrfs_trans_handle
*trans
;
440 struct btrfs_key key
;
441 struct btrfs_root_item root_item
;
442 struct btrfs_inode_item
*inode_item
;
443 struct extent_buffer
*leaf
;
444 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
445 struct btrfs_root
*new_root
;
446 struct btrfs_block_rsv block_rsv
;
447 struct timespec cur_time
= current_fs_time(dir
->i_sb
);
452 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
457 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
462 * Don't create subvolume whose level is not zero. Or qgroup will be
463 * screwed up since it assume subvolme qgroup's level to be 0.
465 if (btrfs_qgroup_level(objectid
))
468 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
470 * The same as the snapshot creation, please see the comment
471 * of create_snapshot().
473 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
474 8, &qgroup_reserved
, false);
478 trans
= btrfs_start_transaction(root
, 0);
480 ret
= PTR_ERR(trans
);
481 btrfs_subvolume_release_metadata(root
, &block_rsv
,
485 trans
->block_rsv
= &block_rsv
;
486 trans
->bytes_reserved
= block_rsv
.size
;
488 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
492 leaf
= btrfs_alloc_tree_block(trans
, root
, 0, objectid
, NULL
, 0, 0, 0);
498 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
499 btrfs_set_header_bytenr(leaf
, leaf
->start
);
500 btrfs_set_header_generation(leaf
, trans
->transid
);
501 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
502 btrfs_set_header_owner(leaf
, objectid
);
504 write_extent_buffer(leaf
, root
->fs_info
->fsid
, btrfs_header_fsid(),
506 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
507 btrfs_header_chunk_tree_uuid(leaf
),
509 btrfs_mark_buffer_dirty(leaf
);
511 memset(&root_item
, 0, sizeof(root_item
));
513 inode_item
= &root_item
.inode
;
514 btrfs_set_stack_inode_generation(inode_item
, 1);
515 btrfs_set_stack_inode_size(inode_item
, 3);
516 btrfs_set_stack_inode_nlink(inode_item
, 1);
517 btrfs_set_stack_inode_nbytes(inode_item
, root
->nodesize
);
518 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
520 btrfs_set_root_flags(&root_item
, 0);
521 btrfs_set_root_limit(&root_item
, 0);
522 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
524 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
525 btrfs_set_root_generation(&root_item
, trans
->transid
);
526 btrfs_set_root_level(&root_item
, 0);
527 btrfs_set_root_refs(&root_item
, 1);
528 btrfs_set_root_used(&root_item
, leaf
->len
);
529 btrfs_set_root_last_snapshot(&root_item
, 0);
531 btrfs_set_root_generation_v2(&root_item
,
532 btrfs_root_generation(&root_item
));
533 uuid_le_gen(&new_uuid
);
534 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
535 btrfs_set_stack_timespec_sec(&root_item
.otime
, cur_time
.tv_sec
);
536 btrfs_set_stack_timespec_nsec(&root_item
.otime
, cur_time
.tv_nsec
);
537 root_item
.ctime
= root_item
.otime
;
538 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
539 btrfs_set_root_otransid(&root_item
, trans
->transid
);
541 btrfs_tree_unlock(leaf
);
542 free_extent_buffer(leaf
);
545 btrfs_set_root_dirid(&root_item
, new_dirid
);
547 key
.objectid
= objectid
;
549 key
.type
= BTRFS_ROOT_ITEM_KEY
;
550 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
555 key
.offset
= (u64
)-1;
556 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
557 if (IS_ERR(new_root
)) {
558 ret
= PTR_ERR(new_root
);
559 btrfs_abort_transaction(trans
, root
, ret
);
563 btrfs_record_root_in_trans(trans
, new_root
);
565 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
567 /* We potentially lose an unused inode item here */
568 btrfs_abort_transaction(trans
, root
, ret
);
572 mutex_lock(&new_root
->objectid_mutex
);
573 new_root
->highest_objectid
= new_dirid
;
574 mutex_unlock(&new_root
->objectid_mutex
);
577 * insert the directory item
579 ret
= btrfs_set_inode_index(dir
, &index
);
581 btrfs_abort_transaction(trans
, root
, ret
);
585 ret
= btrfs_insert_dir_item(trans
, root
,
586 name
, namelen
, dir
, &key
,
587 BTRFS_FT_DIR
, index
);
589 btrfs_abort_transaction(trans
, root
, ret
);
593 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
594 ret
= btrfs_update_inode(trans
, root
, dir
);
597 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
598 objectid
, root
->root_key
.objectid
,
599 btrfs_ino(dir
), index
, name
, namelen
);
602 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
603 root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
606 btrfs_abort_transaction(trans
, root
, ret
);
609 trans
->block_rsv
= NULL
;
610 trans
->bytes_reserved
= 0;
611 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
614 *async_transid
= trans
->transid
;
615 err
= btrfs_commit_transaction_async(trans
, root
, 1);
617 err
= btrfs_commit_transaction(trans
, root
);
619 err
= btrfs_commit_transaction(trans
, root
);
625 inode
= btrfs_lookup_dentry(dir
, dentry
);
627 return PTR_ERR(inode
);
628 d_instantiate(dentry
, inode
);
633 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root
*root
)
639 prepare_to_wait(&root
->subv_writers
->wait
, &wait
,
640 TASK_UNINTERRUPTIBLE
);
642 writers
= percpu_counter_sum(&root
->subv_writers
->counter
);
646 finish_wait(&root
->subv_writers
->wait
, &wait
);
650 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
651 struct dentry
*dentry
, char *name
, int namelen
,
652 u64
*async_transid
, bool readonly
,
653 struct btrfs_qgroup_inherit
*inherit
)
656 struct btrfs_pending_snapshot
*pending_snapshot
;
657 struct btrfs_trans_handle
*trans
;
660 if (!test_bit(BTRFS_ROOT_REF_COWS
, &root
->state
))
663 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
664 if (!pending_snapshot
)
667 pending_snapshot
->root_item
= kzalloc(sizeof(struct btrfs_root_item
),
669 pending_snapshot
->path
= btrfs_alloc_path();
670 if (!pending_snapshot
->root_item
|| !pending_snapshot
->path
) {
675 atomic_inc(&root
->will_be_snapshoted
);
676 smp_mb__after_atomic();
677 btrfs_wait_for_no_snapshoting_writes(root
);
679 ret
= btrfs_start_delalloc_inodes(root
, 0);
683 btrfs_wait_ordered_extents(root
, -1);
685 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
686 BTRFS_BLOCK_RSV_TEMP
);
688 * 1 - parent dir inode
691 * 2 - root ref/backref
692 * 1 - root of snapshot
695 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
696 &pending_snapshot
->block_rsv
, 8,
697 &pending_snapshot
->qgroup_reserved
,
702 pending_snapshot
->dentry
= dentry
;
703 pending_snapshot
->root
= root
;
704 pending_snapshot
->readonly
= readonly
;
705 pending_snapshot
->dir
= dir
;
706 pending_snapshot
->inherit
= inherit
;
708 trans
= btrfs_start_transaction(root
, 0);
710 ret
= PTR_ERR(trans
);
714 spin_lock(&root
->fs_info
->trans_lock
);
715 list_add(&pending_snapshot
->list
,
716 &trans
->transaction
->pending_snapshots
);
717 spin_unlock(&root
->fs_info
->trans_lock
);
719 *async_transid
= trans
->transid
;
720 ret
= btrfs_commit_transaction_async(trans
,
721 root
->fs_info
->extent_root
, 1);
723 ret
= btrfs_commit_transaction(trans
, root
);
725 ret
= btrfs_commit_transaction(trans
,
726 root
->fs_info
->extent_root
);
731 ret
= pending_snapshot
->error
;
735 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
739 inode
= btrfs_lookup_dentry(d_inode(dentry
->d_parent
), dentry
);
741 ret
= PTR_ERR(inode
);
745 d_instantiate(dentry
, inode
);
748 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
749 &pending_snapshot
->block_rsv
,
750 pending_snapshot
->qgroup_reserved
);
752 if (atomic_dec_and_test(&root
->will_be_snapshoted
))
753 wake_up_atomic_t(&root
->will_be_snapshoted
);
755 kfree(pending_snapshot
->root_item
);
756 btrfs_free_path(pending_snapshot
->path
);
757 kfree(pending_snapshot
);
762 /* copy of may_delete in fs/namei.c()
763 * Check whether we can remove a link victim from directory dir, check
764 * whether the type of victim is right.
765 * 1. We can't do it if dir is read-only (done in permission())
766 * 2. We should have write and exec permissions on dir
767 * 3. We can't remove anything from append-only dir
768 * 4. We can't do anything with immutable dir (done in permission())
769 * 5. If the sticky bit on dir is set we should either
770 * a. be owner of dir, or
771 * b. be owner of victim, or
772 * c. have CAP_FOWNER capability
773 * 6. If the victim is append-only or immutable we can't do antyhing with
774 * links pointing to it.
775 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
776 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
777 * 9. We can't remove a root or mountpoint.
778 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
779 * nfs_async_unlink().
782 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
786 if (d_really_is_negative(victim
))
789 BUG_ON(d_inode(victim
->d_parent
) != dir
);
790 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
792 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
797 if (check_sticky(dir
, d_inode(victim
)) || IS_APPEND(d_inode(victim
)) ||
798 IS_IMMUTABLE(d_inode(victim
)) || IS_SWAPFILE(d_inode(victim
)))
801 if (!d_is_dir(victim
))
805 } else if (d_is_dir(victim
))
809 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
814 /* copy of may_create in fs/namei.c() */
815 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
817 if (d_really_is_positive(child
))
821 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
825 * Create a new subvolume below @parent. This is largely modeled after
826 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
827 * inside this filesystem so it's quite a bit simpler.
829 static noinline
int btrfs_mksubvol(struct path
*parent
,
830 char *name
, int namelen
,
831 struct btrfs_root
*snap_src
,
832 u64
*async_transid
, bool readonly
,
833 struct btrfs_qgroup_inherit
*inherit
)
835 struct inode
*dir
= d_inode(parent
->dentry
);
836 struct dentry
*dentry
;
839 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
843 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
844 error
= PTR_ERR(dentry
);
848 error
= btrfs_may_create(dir
, dentry
);
853 * even if this name doesn't exist, we may get hash collisions.
854 * check for them now when we can safely fail
856 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
862 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
864 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
868 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
869 async_transid
, readonly
, inherit
);
871 error
= create_subvol(dir
, dentry
, name
, namelen
,
872 async_transid
, inherit
);
875 fsnotify_mkdir(dir
, dentry
);
877 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
886 * When we're defragging a range, we don't want to kick it off again
887 * if it is really just waiting for delalloc to send it down.
888 * If we find a nice big extent or delalloc range for the bytes in the
889 * file you want to defrag, we return 0 to let you know to skip this
892 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, u32 thresh
)
894 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
895 struct extent_map
*em
= NULL
;
896 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
899 read_lock(&em_tree
->lock
);
900 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
901 read_unlock(&em_tree
->lock
);
904 end
= extent_map_end(em
);
906 if (end
- offset
> thresh
)
909 /* if we already have a nice delalloc here, just stop */
911 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
912 thresh
, EXTENT_DELALLOC
, 1);
919 * helper function to walk through a file and find extents
920 * newer than a specific transid, and smaller than thresh.
922 * This is used by the defragging code to find new and small
925 static int find_new_extents(struct btrfs_root
*root
,
926 struct inode
*inode
, u64 newer_than
,
927 u64
*off
, u32 thresh
)
929 struct btrfs_path
*path
;
930 struct btrfs_key min_key
;
931 struct extent_buffer
*leaf
;
932 struct btrfs_file_extent_item
*extent
;
935 u64 ino
= btrfs_ino(inode
);
937 path
= btrfs_alloc_path();
941 min_key
.objectid
= ino
;
942 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
943 min_key
.offset
= *off
;
946 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
950 if (min_key
.objectid
!= ino
)
952 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
955 leaf
= path
->nodes
[0];
956 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
957 struct btrfs_file_extent_item
);
959 type
= btrfs_file_extent_type(leaf
, extent
);
960 if (type
== BTRFS_FILE_EXTENT_REG
&&
961 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
962 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
963 *off
= min_key
.offset
;
964 btrfs_free_path(path
);
969 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
970 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
974 if (min_key
.offset
== (u64
)-1)
978 btrfs_release_path(path
);
981 btrfs_free_path(path
);
985 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
987 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
988 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
989 struct extent_map
*em
;
990 u64 len
= PAGE_CACHE_SIZE
;
993 * hopefully we have this extent in the tree already, try without
994 * the full extent lock
996 read_lock(&em_tree
->lock
);
997 em
= lookup_extent_mapping(em_tree
, start
, len
);
998 read_unlock(&em_tree
->lock
);
1001 struct extent_state
*cached
= NULL
;
1002 u64 end
= start
+ len
- 1;
1004 /* get the big lock and read metadata off disk */
1005 lock_extent_bits(io_tree
, start
, end
, &cached
);
1006 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
1007 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1016 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1018 struct extent_map
*next
;
1021 /* this is the last extent */
1022 if (em
->start
+ em
->len
>= i_size_read(inode
))
1025 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1026 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
1028 else if ((em
->block_start
+ em
->block_len
== next
->block_start
) &&
1029 (em
->block_len
> SZ_128K
&& next
->block_len
> SZ_128K
))
1032 free_extent_map(next
);
1036 static int should_defrag_range(struct inode
*inode
, u64 start
, u32 thresh
,
1037 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1040 struct extent_map
*em
;
1042 bool next_mergeable
= true;
1043 bool prev_mergeable
= true;
1046 * make sure that once we start defragging an extent, we keep on
1049 if (start
< *defrag_end
)
1054 em
= defrag_lookup_extent(inode
, start
);
1058 /* this will cover holes, and inline extents */
1059 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1065 prev_mergeable
= false;
1067 next_mergeable
= defrag_check_next_extent(inode
, em
);
1069 * we hit a real extent, if it is big or the next extent is not a
1070 * real extent, don't bother defragging it
1072 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1073 (em
->len
>= thresh
|| (!next_mergeable
&& !prev_mergeable
)))
1077 * last_len ends up being a counter of how many bytes we've defragged.
1078 * every time we choose not to defrag an extent, we reset *last_len
1079 * so that the next tiny extent will force a defrag.
1081 * The end result of this is that tiny extents before a single big
1082 * extent will force at least part of that big extent to be defragged.
1085 *defrag_end
= extent_map_end(em
);
1088 *skip
= extent_map_end(em
);
1092 free_extent_map(em
);
1097 * it doesn't do much good to defrag one or two pages
1098 * at a time. This pulls in a nice chunk of pages
1099 * to COW and defrag.
1101 * It also makes sure the delalloc code has enough
1102 * dirty data to avoid making new small extents as part
1105 * It's a good idea to start RA on this range
1106 * before calling this.
1108 static int cluster_pages_for_defrag(struct inode
*inode
,
1109 struct page
**pages
,
1110 unsigned long start_index
,
1111 unsigned long num_pages
)
1113 unsigned long file_end
;
1114 u64 isize
= i_size_read(inode
);
1121 struct btrfs_ordered_extent
*ordered
;
1122 struct extent_state
*cached_state
= NULL
;
1123 struct extent_io_tree
*tree
;
1124 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1126 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1127 if (!isize
|| start_index
> file_end
)
1130 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1132 ret
= btrfs_delalloc_reserve_space(inode
,
1133 start_index
<< PAGE_CACHE_SHIFT
,
1134 page_cnt
<< PAGE_CACHE_SHIFT
);
1138 tree
= &BTRFS_I(inode
)->io_tree
;
1140 /* step one, lock all the pages */
1141 for (i
= 0; i
< page_cnt
; i
++) {
1144 page
= find_or_create_page(inode
->i_mapping
,
1145 start_index
+ i
, mask
);
1149 page_start
= page_offset(page
);
1150 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1152 lock_extent_bits(tree
, page_start
, page_end
,
1154 ordered
= btrfs_lookup_ordered_extent(inode
,
1156 unlock_extent_cached(tree
, page_start
, page_end
,
1157 &cached_state
, GFP_NOFS
);
1162 btrfs_start_ordered_extent(inode
, ordered
, 1);
1163 btrfs_put_ordered_extent(ordered
);
1166 * we unlocked the page above, so we need check if
1167 * it was released or not.
1169 if (page
->mapping
!= inode
->i_mapping
) {
1171 page_cache_release(page
);
1176 if (!PageUptodate(page
)) {
1177 btrfs_readpage(NULL
, page
);
1179 if (!PageUptodate(page
)) {
1181 page_cache_release(page
);
1187 if (page
->mapping
!= inode
->i_mapping
) {
1189 page_cache_release(page
);
1199 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1203 * so now we have a nice long stream of locked
1204 * and up to date pages, lets wait on them
1206 for (i
= 0; i
< i_done
; i
++)
1207 wait_on_page_writeback(pages
[i
]);
1209 page_start
= page_offset(pages
[0]);
1210 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1212 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1213 page_start
, page_end
- 1, &cached_state
);
1214 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1215 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1216 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1217 &cached_state
, GFP_NOFS
);
1219 if (i_done
!= page_cnt
) {
1220 spin_lock(&BTRFS_I(inode
)->lock
);
1221 BTRFS_I(inode
)->outstanding_extents
++;
1222 spin_unlock(&BTRFS_I(inode
)->lock
);
1223 btrfs_delalloc_release_space(inode
,
1224 start_index
<< PAGE_CACHE_SHIFT
,
1225 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1229 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1230 &cached_state
, GFP_NOFS
);
1232 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1233 page_start
, page_end
- 1, &cached_state
,
1236 for (i
= 0; i
< i_done
; i
++) {
1237 clear_page_dirty_for_io(pages
[i
]);
1238 ClearPageChecked(pages
[i
]);
1239 set_page_extent_mapped(pages
[i
]);
1240 set_page_dirty(pages
[i
]);
1241 unlock_page(pages
[i
]);
1242 page_cache_release(pages
[i
]);
1246 for (i
= 0; i
< i_done
; i
++) {
1247 unlock_page(pages
[i
]);
1248 page_cache_release(pages
[i
]);
1250 btrfs_delalloc_release_space(inode
,
1251 start_index
<< PAGE_CACHE_SHIFT
,
1252 page_cnt
<< PAGE_CACHE_SHIFT
);
1257 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1258 struct btrfs_ioctl_defrag_range_args
*range
,
1259 u64 newer_than
, unsigned long max_to_defrag
)
1261 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1262 struct file_ra_state
*ra
= NULL
;
1263 unsigned long last_index
;
1264 u64 isize
= i_size_read(inode
);
1268 u64 newer_off
= range
->start
;
1270 unsigned long ra_index
= 0;
1272 int defrag_count
= 0;
1273 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1274 u32 extent_thresh
= range
->extent_thresh
;
1275 unsigned long max_cluster
= SZ_256K
>> PAGE_CACHE_SHIFT
;
1276 unsigned long cluster
= max_cluster
;
1277 u64 new_align
= ~((u64
)SZ_128K
- 1);
1278 struct page
**pages
= NULL
;
1283 if (range
->start
>= isize
)
1286 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1287 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1289 if (range
->compress_type
)
1290 compress_type
= range
->compress_type
;
1293 if (extent_thresh
== 0)
1294 extent_thresh
= SZ_256K
;
1297 * if we were not given a file, allocate a readahead
1301 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1304 file_ra_state_init(ra
, inode
->i_mapping
);
1309 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1316 /* find the last page to defrag */
1317 if (range
->start
+ range
->len
> range
->start
) {
1318 last_index
= min_t(u64
, isize
- 1,
1319 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1321 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1325 ret
= find_new_extents(root
, inode
, newer_than
,
1326 &newer_off
, SZ_64K
);
1328 range
->start
= newer_off
;
1330 * we always align our defrag to help keep
1331 * the extents in the file evenly spaced
1333 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1337 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1340 max_to_defrag
= last_index
- i
+ 1;
1343 * make writeback starts from i, so the defrag range can be
1344 * written sequentially.
1346 if (i
< inode
->i_mapping
->writeback_index
)
1347 inode
->i_mapping
->writeback_index
= i
;
1349 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1350 (i
< DIV_ROUND_UP(i_size_read(inode
), PAGE_CACHE_SIZE
))) {
1352 * make sure we stop running if someone unmounts
1355 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1358 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1359 btrfs_debug(root
->fs_info
, "defrag_file cancelled");
1364 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1365 extent_thresh
, &last_len
, &skip
,
1366 &defrag_end
, range
->flags
&
1367 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1370 * the should_defrag function tells us how much to skip
1371 * bump our counter by the suggested amount
1373 next
= DIV_ROUND_UP(skip
, PAGE_CACHE_SIZE
);
1374 i
= max(i
+ 1, next
);
1379 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1380 PAGE_CACHE_SHIFT
) - i
;
1381 cluster
= min(cluster
, max_cluster
);
1383 cluster
= max_cluster
;
1386 if (i
+ cluster
> ra_index
) {
1387 ra_index
= max(i
, ra_index
);
1388 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1390 ra_index
+= cluster
;
1394 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1395 BTRFS_I(inode
)->force_compress
= compress_type
;
1396 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1398 inode_unlock(inode
);
1402 defrag_count
+= ret
;
1403 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1404 inode_unlock(inode
);
1407 if (newer_off
== (u64
)-1)
1413 newer_off
= max(newer_off
+ 1,
1414 (u64
)i
<< PAGE_CACHE_SHIFT
);
1416 ret
= find_new_extents(root
, inode
, newer_than
,
1417 &newer_off
, SZ_64K
);
1419 range
->start
= newer_off
;
1420 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1427 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1435 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1436 filemap_flush(inode
->i_mapping
);
1437 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1438 &BTRFS_I(inode
)->runtime_flags
))
1439 filemap_flush(inode
->i_mapping
);
1442 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1443 /* the filemap_flush will queue IO into the worker threads, but
1444 * we have to make sure the IO is actually started and that
1445 * ordered extents get created before we return
1447 atomic_inc(&root
->fs_info
->async_submit_draining
);
1448 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1449 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1450 wait_event(root
->fs_info
->async_submit_wait
,
1451 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1452 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1454 atomic_dec(&root
->fs_info
->async_submit_draining
);
1457 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1458 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1464 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1466 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1467 inode_unlock(inode
);
1475 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1481 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1482 struct btrfs_ioctl_vol_args
*vol_args
;
1483 struct btrfs_trans_handle
*trans
;
1484 struct btrfs_device
*device
= NULL
;
1487 char *devstr
= NULL
;
1491 if (!capable(CAP_SYS_ADMIN
))
1494 ret
= mnt_want_write_file(file
);
1498 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1500 mnt_drop_write_file(file
);
1501 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1504 mutex_lock(&root
->fs_info
->volume_mutex
);
1505 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1506 if (IS_ERR(vol_args
)) {
1507 ret
= PTR_ERR(vol_args
);
1511 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1513 sizestr
= vol_args
->name
;
1514 devstr
= strchr(sizestr
, ':');
1516 sizestr
= devstr
+ 1;
1518 devstr
= vol_args
->name
;
1519 ret
= kstrtoull(devstr
, 10, &devid
);
1526 btrfs_info(root
->fs_info
, "resizing devid %llu", devid
);
1529 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1531 btrfs_info(root
->fs_info
, "resizer unable to find device %llu",
1537 if (!device
->writeable
) {
1538 btrfs_info(root
->fs_info
,
1539 "resizer unable to apply on readonly device %llu",
1545 if (!strcmp(sizestr
, "max"))
1546 new_size
= device
->bdev
->bd_inode
->i_size
;
1548 if (sizestr
[0] == '-') {
1551 } else if (sizestr
[0] == '+') {
1555 new_size
= memparse(sizestr
, &retptr
);
1556 if (*retptr
!= '\0' || new_size
== 0) {
1562 if (device
->is_tgtdev_for_dev_replace
) {
1567 old_size
= btrfs_device_get_total_bytes(device
);
1570 if (new_size
> old_size
) {
1574 new_size
= old_size
- new_size
;
1575 } else if (mod
> 0) {
1576 if (new_size
> ULLONG_MAX
- old_size
) {
1580 new_size
= old_size
+ new_size
;
1583 if (new_size
< SZ_256M
) {
1587 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1592 new_size
= div_u64(new_size
, root
->sectorsize
);
1593 new_size
*= root
->sectorsize
;
1595 btrfs_info_in_rcu(root
->fs_info
, "new size for %s is %llu",
1596 rcu_str_deref(device
->name
), new_size
);
1598 if (new_size
> old_size
) {
1599 trans
= btrfs_start_transaction(root
, 0);
1600 if (IS_ERR(trans
)) {
1601 ret
= PTR_ERR(trans
);
1604 ret
= btrfs_grow_device(trans
, device
, new_size
);
1605 btrfs_commit_transaction(trans
, root
);
1606 } else if (new_size
< old_size
) {
1607 ret
= btrfs_shrink_device(device
, new_size
);
1608 } /* equal, nothing need to do */
1613 mutex_unlock(&root
->fs_info
->volume_mutex
);
1614 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1615 mnt_drop_write_file(file
);
1619 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1620 char *name
, unsigned long fd
, int subvol
,
1621 u64
*transid
, bool readonly
,
1622 struct btrfs_qgroup_inherit
*inherit
)
1627 ret
= mnt_want_write_file(file
);
1631 namelen
= strlen(name
);
1632 if (strchr(name
, '/')) {
1634 goto out_drop_write
;
1637 if (name
[0] == '.' &&
1638 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1640 goto out_drop_write
;
1644 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1645 NULL
, transid
, readonly
, inherit
);
1647 struct fd src
= fdget(fd
);
1648 struct inode
*src_inode
;
1651 goto out_drop_write
;
1654 src_inode
= file_inode(src
.file
);
1655 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1656 btrfs_info(BTRFS_I(src_inode
)->root
->fs_info
,
1657 "Snapshot src from another FS");
1659 } else if (!inode_owner_or_capable(src_inode
)) {
1661 * Subvolume creation is not restricted, but snapshots
1662 * are limited to own subvolumes only
1666 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1667 BTRFS_I(src_inode
)->root
,
1668 transid
, readonly
, inherit
);
1673 mnt_drop_write_file(file
);
1678 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1679 void __user
*arg
, int subvol
)
1681 struct btrfs_ioctl_vol_args
*vol_args
;
1684 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1685 if (IS_ERR(vol_args
))
1686 return PTR_ERR(vol_args
);
1687 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1689 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1690 vol_args
->fd
, subvol
,
1697 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1698 void __user
*arg
, int subvol
)
1700 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1704 bool readonly
= false;
1705 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1707 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1708 if (IS_ERR(vol_args
))
1709 return PTR_ERR(vol_args
);
1710 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1712 if (vol_args
->flags
&
1713 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1714 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1719 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1721 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1723 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1724 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1728 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1729 if (IS_ERR(inherit
)) {
1730 ret
= PTR_ERR(inherit
);
1735 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1736 vol_args
->fd
, subvol
, ptr
,
1741 if (ptr
&& copy_to_user(arg
+
1742 offsetof(struct btrfs_ioctl_vol_args_v2
,
1754 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1757 struct inode
*inode
= file_inode(file
);
1758 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1762 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1765 down_read(&root
->fs_info
->subvol_sem
);
1766 if (btrfs_root_readonly(root
))
1767 flags
|= BTRFS_SUBVOL_RDONLY
;
1768 up_read(&root
->fs_info
->subvol_sem
);
1770 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1776 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1779 struct inode
*inode
= file_inode(file
);
1780 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1781 struct btrfs_trans_handle
*trans
;
1786 if (!inode_owner_or_capable(inode
))
1789 ret
= mnt_want_write_file(file
);
1793 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1795 goto out_drop_write
;
1798 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1800 goto out_drop_write
;
1803 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1805 goto out_drop_write
;
1808 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1810 goto out_drop_write
;
1813 down_write(&root
->fs_info
->subvol_sem
);
1816 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1819 root_flags
= btrfs_root_flags(&root
->root_item
);
1820 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1821 btrfs_set_root_flags(&root
->root_item
,
1822 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1825 * Block RO -> RW transition if this subvolume is involved in
1828 spin_lock(&root
->root_item_lock
);
1829 if (root
->send_in_progress
== 0) {
1830 btrfs_set_root_flags(&root
->root_item
,
1831 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1832 spin_unlock(&root
->root_item_lock
);
1834 spin_unlock(&root
->root_item_lock
);
1835 btrfs_warn(root
->fs_info
,
1836 "Attempt to set subvolume %llu read-write during send",
1837 root
->root_key
.objectid
);
1843 trans
= btrfs_start_transaction(root
, 1);
1844 if (IS_ERR(trans
)) {
1845 ret
= PTR_ERR(trans
);
1849 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1850 &root
->root_key
, &root
->root_item
);
1852 btrfs_commit_transaction(trans
, root
);
1855 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1857 up_write(&root
->fs_info
->subvol_sem
);
1859 mnt_drop_write_file(file
);
1865 * helper to check if the subvolume references other subvolumes
1867 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1869 struct btrfs_path
*path
;
1870 struct btrfs_dir_item
*di
;
1871 struct btrfs_key key
;
1875 path
= btrfs_alloc_path();
1879 /* Make sure this root isn't set as the default subvol */
1880 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
1881 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
, path
,
1882 dir_id
, "default", 7, 0);
1883 if (di
&& !IS_ERR(di
)) {
1884 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1885 if (key
.objectid
== root
->root_key
.objectid
) {
1887 btrfs_err(root
->fs_info
, "deleting default subvolume "
1888 "%llu is not allowed", key
.objectid
);
1891 btrfs_release_path(path
);
1894 key
.objectid
= root
->root_key
.objectid
;
1895 key
.type
= BTRFS_ROOT_REF_KEY
;
1896 key
.offset
= (u64
)-1;
1898 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1905 if (path
->slots
[0] > 0) {
1907 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1908 if (key
.objectid
== root
->root_key
.objectid
&&
1909 key
.type
== BTRFS_ROOT_REF_KEY
)
1913 btrfs_free_path(path
);
1917 static noinline
int key_in_sk(struct btrfs_key
*key
,
1918 struct btrfs_ioctl_search_key
*sk
)
1920 struct btrfs_key test
;
1923 test
.objectid
= sk
->min_objectid
;
1924 test
.type
= sk
->min_type
;
1925 test
.offset
= sk
->min_offset
;
1927 ret
= btrfs_comp_cpu_keys(key
, &test
);
1931 test
.objectid
= sk
->max_objectid
;
1932 test
.type
= sk
->max_type
;
1933 test
.offset
= sk
->max_offset
;
1935 ret
= btrfs_comp_cpu_keys(key
, &test
);
1941 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1942 struct btrfs_path
*path
,
1943 struct btrfs_key
*key
,
1944 struct btrfs_ioctl_search_key
*sk
,
1947 unsigned long *sk_offset
,
1951 struct extent_buffer
*leaf
;
1952 struct btrfs_ioctl_search_header sh
;
1953 struct btrfs_key test
;
1954 unsigned long item_off
;
1955 unsigned long item_len
;
1961 leaf
= path
->nodes
[0];
1962 slot
= path
->slots
[0];
1963 nritems
= btrfs_header_nritems(leaf
);
1965 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1969 found_transid
= btrfs_header_generation(leaf
);
1971 for (i
= slot
; i
< nritems
; i
++) {
1972 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1973 item_len
= btrfs_item_size_nr(leaf
, i
);
1975 btrfs_item_key_to_cpu(leaf
, key
, i
);
1976 if (!key_in_sk(key
, sk
))
1979 if (sizeof(sh
) + item_len
> *buf_size
) {
1986 * return one empty item back for v1, which does not
1990 *buf_size
= sizeof(sh
) + item_len
;
1995 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
2000 sh
.objectid
= key
->objectid
;
2001 sh
.offset
= key
->offset
;
2002 sh
.type
= key
->type
;
2004 sh
.transid
= found_transid
;
2006 /* copy search result header */
2007 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2012 *sk_offset
+= sizeof(sh
);
2015 char __user
*up
= ubuf
+ *sk_offset
;
2017 if (read_extent_buffer_to_user(leaf
, up
,
2018 item_off
, item_len
)) {
2023 *sk_offset
+= item_len
;
2027 if (ret
) /* -EOVERFLOW from above */
2030 if (*num_found
>= sk
->nr_items
) {
2037 test
.objectid
= sk
->max_objectid
;
2038 test
.type
= sk
->max_type
;
2039 test
.offset
= sk
->max_offset
;
2040 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
2042 else if (key
->offset
< (u64
)-1)
2044 else if (key
->type
< (u8
)-1) {
2047 } else if (key
->objectid
< (u64
)-1) {
2055 * 0: all items from this leaf copied, continue with next
2056 * 1: * more items can be copied, but unused buffer is too small
2057 * * all items were found
2058 * Either way, it will stops the loop which iterates to the next
2060 * -EOVERFLOW: item was to large for buffer
2061 * -EFAULT: could not copy extent buffer back to userspace
2066 static noinline
int search_ioctl(struct inode
*inode
,
2067 struct btrfs_ioctl_search_key
*sk
,
2071 struct btrfs_root
*root
;
2072 struct btrfs_key key
;
2073 struct btrfs_path
*path
;
2074 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
2077 unsigned long sk_offset
= 0;
2079 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2080 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2084 path
= btrfs_alloc_path();
2088 if (sk
->tree_id
== 0) {
2089 /* search the root of the inode that was passed */
2090 root
= BTRFS_I(inode
)->root
;
2092 key
.objectid
= sk
->tree_id
;
2093 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2094 key
.offset
= (u64
)-1;
2095 root
= btrfs_read_fs_root_no_name(info
, &key
);
2097 btrfs_free_path(path
);
2102 key
.objectid
= sk
->min_objectid
;
2103 key
.type
= sk
->min_type
;
2104 key
.offset
= sk
->min_offset
;
2107 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2113 ret
= copy_to_sk(root
, path
, &key
, sk
, buf_size
, ubuf
,
2114 &sk_offset
, &num_found
);
2115 btrfs_release_path(path
);
2123 sk
->nr_items
= num_found
;
2124 btrfs_free_path(path
);
2128 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2131 struct btrfs_ioctl_search_args __user
*uargs
;
2132 struct btrfs_ioctl_search_key sk
;
2133 struct inode
*inode
;
2137 if (!capable(CAP_SYS_ADMIN
))
2140 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2142 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2145 buf_size
= sizeof(uargs
->buf
);
2147 inode
= file_inode(file
);
2148 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2151 * In the origin implementation an overflow is handled by returning a
2152 * search header with a len of zero, so reset ret.
2154 if (ret
== -EOVERFLOW
)
2157 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2162 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2165 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2166 struct btrfs_ioctl_search_args_v2 args
;
2167 struct inode
*inode
;
2170 const size_t buf_limit
= SZ_16M
;
2172 if (!capable(CAP_SYS_ADMIN
))
2175 /* copy search header and buffer size */
2176 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2177 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2180 buf_size
= args
.buf_size
;
2182 if (buf_size
< sizeof(struct btrfs_ioctl_search_header
))
2185 /* limit result size to 16MB */
2186 if (buf_size
> buf_limit
)
2187 buf_size
= buf_limit
;
2189 inode
= file_inode(file
);
2190 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2191 (char *)(&uarg
->buf
[0]));
2192 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2194 else if (ret
== -EOVERFLOW
&&
2195 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2202 * Search INODE_REFs to identify path name of 'dirid' directory
2203 * in a 'tree_id' tree. and sets path name to 'name'.
2205 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2206 u64 tree_id
, u64 dirid
, char *name
)
2208 struct btrfs_root
*root
;
2209 struct btrfs_key key
;
2215 struct btrfs_inode_ref
*iref
;
2216 struct extent_buffer
*l
;
2217 struct btrfs_path
*path
;
2219 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2224 path
= btrfs_alloc_path();
2228 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2230 key
.objectid
= tree_id
;
2231 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2232 key
.offset
= (u64
)-1;
2233 root
= btrfs_read_fs_root_no_name(info
, &key
);
2235 btrfs_err(info
, "could not find root %llu", tree_id
);
2240 key
.objectid
= dirid
;
2241 key
.type
= BTRFS_INODE_REF_KEY
;
2242 key
.offset
= (u64
)-1;
2245 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2249 ret
= btrfs_previous_item(root
, path
, dirid
,
2250 BTRFS_INODE_REF_KEY
);
2260 slot
= path
->slots
[0];
2261 btrfs_item_key_to_cpu(l
, &key
, slot
);
2263 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2264 len
= btrfs_inode_ref_name_len(l
, iref
);
2266 total_len
+= len
+ 1;
2268 ret
= -ENAMETOOLONG
;
2273 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2275 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2278 btrfs_release_path(path
);
2279 key
.objectid
= key
.offset
;
2280 key
.offset
= (u64
)-1;
2281 dirid
= key
.objectid
;
2283 memmove(name
, ptr
, total_len
);
2284 name
[total_len
] = '\0';
2287 btrfs_free_path(path
);
2291 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2294 struct btrfs_ioctl_ino_lookup_args
*args
;
2295 struct inode
*inode
;
2298 args
= memdup_user(argp
, sizeof(*args
));
2300 return PTR_ERR(args
);
2302 inode
= file_inode(file
);
2305 * Unprivileged query to obtain the containing subvolume root id. The
2306 * path is reset so it's consistent with btrfs_search_path_in_tree.
2308 if (args
->treeid
== 0)
2309 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2311 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2316 if (!capable(CAP_SYS_ADMIN
)) {
2321 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2322 args
->treeid
, args
->objectid
,
2326 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2333 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2336 struct dentry
*parent
= file
->f_path
.dentry
;
2337 struct dentry
*dentry
;
2338 struct inode
*dir
= d_inode(parent
);
2339 struct inode
*inode
;
2340 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2341 struct btrfs_root
*dest
= NULL
;
2342 struct btrfs_ioctl_vol_args
*vol_args
;
2343 struct btrfs_trans_handle
*trans
;
2344 struct btrfs_block_rsv block_rsv
;
2346 u64 qgroup_reserved
;
2351 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2352 if (IS_ERR(vol_args
))
2353 return PTR_ERR(vol_args
);
2355 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2356 namelen
= strlen(vol_args
->name
);
2357 if (strchr(vol_args
->name
, '/') ||
2358 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2363 err
= mnt_want_write_file(file
);
2368 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2370 goto out_drop_write
;
2371 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2372 if (IS_ERR(dentry
)) {
2373 err
= PTR_ERR(dentry
);
2374 goto out_unlock_dir
;
2377 if (d_really_is_negative(dentry
)) {
2382 inode
= d_inode(dentry
);
2383 dest
= BTRFS_I(inode
)->root
;
2384 if (!capable(CAP_SYS_ADMIN
)) {
2386 * Regular user. Only allow this with a special mount
2387 * option, when the user has write+exec access to the
2388 * subvol root, and when rmdir(2) would have been
2391 * Note that this is _not_ check that the subvol is
2392 * empty or doesn't contain data that we wouldn't
2393 * otherwise be able to delete.
2395 * Users who want to delete empty subvols should try
2399 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2403 * Do not allow deletion if the parent dir is the same
2404 * as the dir to be deleted. That means the ioctl
2405 * must be called on the dentry referencing the root
2406 * of the subvol, not a random directory contained
2413 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2418 /* check if subvolume may be deleted by a user */
2419 err
= btrfs_may_delete(dir
, dentry
, 1);
2423 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2431 * Don't allow to delete a subvolume with send in progress. This is
2432 * inside the i_mutex so the error handling that has to drop the bit
2433 * again is not run concurrently.
2435 spin_lock(&dest
->root_item_lock
);
2436 root_flags
= btrfs_root_flags(&dest
->root_item
);
2437 if (dest
->send_in_progress
== 0) {
2438 btrfs_set_root_flags(&dest
->root_item
,
2439 root_flags
| BTRFS_ROOT_SUBVOL_DEAD
);
2440 spin_unlock(&dest
->root_item_lock
);
2442 spin_unlock(&dest
->root_item_lock
);
2443 btrfs_warn(root
->fs_info
,
2444 "Attempt to delete subvolume %llu during send",
2445 dest
->root_key
.objectid
);
2447 goto out_unlock_inode
;
2450 down_write(&root
->fs_info
->subvol_sem
);
2452 err
= may_destroy_subvol(dest
);
2456 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2458 * One for dir inode, two for dir entries, two for root
2461 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2462 5, &qgroup_reserved
, true);
2466 trans
= btrfs_start_transaction(root
, 0);
2467 if (IS_ERR(trans
)) {
2468 err
= PTR_ERR(trans
);
2471 trans
->block_rsv
= &block_rsv
;
2472 trans
->bytes_reserved
= block_rsv
.size
;
2474 btrfs_record_snapshot_destroy(trans
, dir
);
2476 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2477 dest
->root_key
.objectid
,
2478 dentry
->d_name
.name
,
2479 dentry
->d_name
.len
);
2482 btrfs_abort_transaction(trans
, root
, ret
);
2486 btrfs_record_root_in_trans(trans
, dest
);
2488 memset(&dest
->root_item
.drop_progress
, 0,
2489 sizeof(dest
->root_item
.drop_progress
));
2490 dest
->root_item
.drop_level
= 0;
2491 btrfs_set_root_refs(&dest
->root_item
, 0);
2493 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED
, &dest
->state
)) {
2494 ret
= btrfs_insert_orphan_item(trans
,
2495 root
->fs_info
->tree_root
,
2496 dest
->root_key
.objectid
);
2498 btrfs_abort_transaction(trans
, root
, ret
);
2504 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2505 dest
->root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
2506 dest
->root_key
.objectid
);
2507 if (ret
&& ret
!= -ENOENT
) {
2508 btrfs_abort_transaction(trans
, root
, ret
);
2512 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2513 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2514 dest
->root_item
.received_uuid
,
2515 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2516 dest
->root_key
.objectid
);
2517 if (ret
&& ret
!= -ENOENT
) {
2518 btrfs_abort_transaction(trans
, root
, ret
);
2525 trans
->block_rsv
= NULL
;
2526 trans
->bytes_reserved
= 0;
2527 ret
= btrfs_end_transaction(trans
, root
);
2530 inode
->i_flags
|= S_DEAD
;
2532 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2534 up_write(&root
->fs_info
->subvol_sem
);
2536 spin_lock(&dest
->root_item_lock
);
2537 root_flags
= btrfs_root_flags(&dest
->root_item
);
2538 btrfs_set_root_flags(&dest
->root_item
,
2539 root_flags
& ~BTRFS_ROOT_SUBVOL_DEAD
);
2540 spin_unlock(&dest
->root_item_lock
);
2543 inode_unlock(inode
);
2545 d_invalidate(dentry
);
2546 btrfs_invalidate_inodes(dest
);
2548 ASSERT(dest
->send_in_progress
== 0);
2551 if (dest
->ino_cache_inode
) {
2552 iput(dest
->ino_cache_inode
);
2553 dest
->ino_cache_inode
= NULL
;
2561 mnt_drop_write_file(file
);
2567 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2569 struct inode
*inode
= file_inode(file
);
2570 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2571 struct btrfs_ioctl_defrag_range_args
*range
;
2574 ret
= mnt_want_write_file(file
);
2578 if (btrfs_root_readonly(root
)) {
2583 switch (inode
->i_mode
& S_IFMT
) {
2585 if (!capable(CAP_SYS_ADMIN
)) {
2589 ret
= btrfs_defrag_root(root
);
2592 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2595 if (!(file
->f_mode
& FMODE_WRITE
)) {
2600 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2607 if (copy_from_user(range
, argp
,
2613 /* compression requires us to start the IO */
2614 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2615 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2616 range
->extent_thresh
= (u32
)-1;
2619 /* the rest are all set to zero by kzalloc */
2620 range
->len
= (u64
)-1;
2622 ret
= btrfs_defrag_file(file_inode(file
), file
,
2632 mnt_drop_write_file(file
);
2636 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2638 struct btrfs_ioctl_vol_args
*vol_args
;
2641 if (!capable(CAP_SYS_ADMIN
))
2644 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2646 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2649 mutex_lock(&root
->fs_info
->volume_mutex
);
2650 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2651 if (IS_ERR(vol_args
)) {
2652 ret
= PTR_ERR(vol_args
);
2656 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2657 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2660 btrfs_info(root
->fs_info
, "disk added %s",vol_args
->name
);
2664 mutex_unlock(&root
->fs_info
->volume_mutex
);
2665 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2669 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2671 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2672 struct btrfs_ioctl_vol_args
*vol_args
;
2675 if (!capable(CAP_SYS_ADMIN
))
2678 ret
= mnt_want_write_file(file
);
2682 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2683 if (IS_ERR(vol_args
)) {
2684 ret
= PTR_ERR(vol_args
);
2688 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2690 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2692 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2696 mutex_lock(&root
->fs_info
->volume_mutex
);
2697 ret
= btrfs_rm_device(root
, vol_args
->name
);
2698 mutex_unlock(&root
->fs_info
->volume_mutex
);
2699 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2702 btrfs_info(root
->fs_info
, "disk deleted %s",vol_args
->name
);
2707 mnt_drop_write_file(file
);
2711 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2713 struct btrfs_ioctl_fs_info_args
*fi_args
;
2714 struct btrfs_device
*device
;
2715 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2718 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2722 mutex_lock(&fs_devices
->device_list_mutex
);
2723 fi_args
->num_devices
= fs_devices
->num_devices
;
2724 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2726 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
2727 if (device
->devid
> fi_args
->max_id
)
2728 fi_args
->max_id
= device
->devid
;
2730 mutex_unlock(&fs_devices
->device_list_mutex
);
2732 fi_args
->nodesize
= root
->fs_info
->super_copy
->nodesize
;
2733 fi_args
->sectorsize
= root
->fs_info
->super_copy
->sectorsize
;
2734 fi_args
->clone_alignment
= root
->fs_info
->super_copy
->sectorsize
;
2736 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2743 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2745 struct btrfs_ioctl_dev_info_args
*di_args
;
2746 struct btrfs_device
*dev
;
2747 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2749 char *s_uuid
= NULL
;
2751 di_args
= memdup_user(arg
, sizeof(*di_args
));
2752 if (IS_ERR(di_args
))
2753 return PTR_ERR(di_args
);
2755 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2756 s_uuid
= di_args
->uuid
;
2758 mutex_lock(&fs_devices
->device_list_mutex
);
2759 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2766 di_args
->devid
= dev
->devid
;
2767 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
2768 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
2769 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2771 struct rcu_string
*name
;
2774 name
= rcu_dereference(dev
->name
);
2775 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2777 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2779 di_args
->path
[0] = '\0';
2783 mutex_unlock(&fs_devices
->device_list_mutex
);
2784 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2791 static struct page
*extent_same_get_page(struct inode
*inode
, pgoff_t index
)
2795 page
= grab_cache_page(inode
->i_mapping
, index
);
2797 return ERR_PTR(-ENOMEM
);
2799 if (!PageUptodate(page
)) {
2802 ret
= btrfs_readpage(NULL
, page
);
2804 return ERR_PTR(ret
);
2806 if (!PageUptodate(page
)) {
2808 page_cache_release(page
);
2809 return ERR_PTR(-EIO
);
2811 if (page
->mapping
!= inode
->i_mapping
) {
2813 page_cache_release(page
);
2814 return ERR_PTR(-EAGAIN
);
2821 static int gather_extent_pages(struct inode
*inode
, struct page
**pages
,
2822 int num_pages
, u64 off
)
2825 pgoff_t index
= off
>> PAGE_CACHE_SHIFT
;
2827 for (i
= 0; i
< num_pages
; i
++) {
2829 pages
[i
] = extent_same_get_page(inode
, index
+ i
);
2830 if (IS_ERR(pages
[i
])) {
2831 int err
= PTR_ERR(pages
[i
]);
2842 static int lock_extent_range(struct inode
*inode
, u64 off
, u64 len
,
2843 bool retry_range_locking
)
2846 * Do any pending delalloc/csum calculations on inode, one way or
2847 * another, and lock file content.
2848 * The locking order is:
2851 * 2) range in the inode's io tree
2854 struct btrfs_ordered_extent
*ordered
;
2855 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2856 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2859 ordered
->file_offset
+ ordered
->len
<= off
||
2860 ordered
->file_offset
>= off
+ len
) &&
2861 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2862 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
)) {
2864 btrfs_put_ordered_extent(ordered
);
2867 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2869 btrfs_put_ordered_extent(ordered
);
2870 if (!retry_range_locking
)
2872 btrfs_wait_ordered_range(inode
, off
, len
);
2877 static void btrfs_double_inode_unlock(struct inode
*inode1
, struct inode
*inode2
)
2879 inode_unlock(inode1
);
2880 inode_unlock(inode2
);
2883 static void btrfs_double_inode_lock(struct inode
*inode1
, struct inode
*inode2
)
2885 if (inode1
< inode2
)
2886 swap(inode1
, inode2
);
2888 inode_lock_nested(inode1
, I_MUTEX_PARENT
);
2889 inode_lock_nested(inode2
, I_MUTEX_CHILD
);
2892 static void btrfs_double_extent_unlock(struct inode
*inode1
, u64 loff1
,
2893 struct inode
*inode2
, u64 loff2
, u64 len
)
2895 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2896 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2899 static int btrfs_double_extent_lock(struct inode
*inode1
, u64 loff1
,
2900 struct inode
*inode2
, u64 loff2
, u64 len
,
2901 bool retry_range_locking
)
2905 if (inode1
< inode2
) {
2906 swap(inode1
, inode2
);
2909 ret
= lock_extent_range(inode1
, loff1
, len
, retry_range_locking
);
2912 ret
= lock_extent_range(inode2
, loff2
, len
, retry_range_locking
);
2914 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
,
2921 struct page
**src_pages
;
2922 struct page
**dst_pages
;
2925 static void btrfs_cmp_data_free(struct cmp_pages
*cmp
)
2930 for (i
= 0; i
< cmp
->num_pages
; i
++) {
2931 pg
= cmp
->src_pages
[i
];
2934 page_cache_release(pg
);
2936 pg
= cmp
->dst_pages
[i
];
2939 page_cache_release(pg
);
2942 kfree(cmp
->src_pages
);
2943 kfree(cmp
->dst_pages
);
2946 static int btrfs_cmp_data_prepare(struct inode
*src
, u64 loff
,
2947 struct inode
*dst
, u64 dst_loff
,
2948 u64 len
, struct cmp_pages
*cmp
)
2951 int num_pages
= PAGE_CACHE_ALIGN(len
) >> PAGE_CACHE_SHIFT
;
2952 struct page
**src_pgarr
, **dst_pgarr
;
2955 * We must gather up all the pages before we initiate our
2956 * extent locking. We use an array for the page pointers. Size
2957 * of the array is bounded by len, which is in turn bounded by
2958 * BTRFS_MAX_DEDUPE_LEN.
2960 src_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
2961 dst_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
2962 if (!src_pgarr
|| !dst_pgarr
) {
2967 cmp
->num_pages
= num_pages
;
2968 cmp
->src_pages
= src_pgarr
;
2969 cmp
->dst_pages
= dst_pgarr
;
2971 ret
= gather_extent_pages(src
, cmp
->src_pages
, cmp
->num_pages
, loff
);
2975 ret
= gather_extent_pages(dst
, cmp
->dst_pages
, cmp
->num_pages
, dst_loff
);
2979 btrfs_cmp_data_free(cmp
);
2983 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
2984 u64 dst_loff
, u64 len
, struct cmp_pages
*cmp
)
2988 struct page
*src_page
, *dst_page
;
2989 unsigned int cmp_len
= PAGE_CACHE_SIZE
;
2990 void *addr
, *dst_addr
;
2994 if (len
< PAGE_CACHE_SIZE
)
2997 BUG_ON(i
>= cmp
->num_pages
);
2999 src_page
= cmp
->src_pages
[i
];
3000 dst_page
= cmp
->dst_pages
[i
];
3001 ASSERT(PageLocked(src_page
));
3002 ASSERT(PageLocked(dst_page
));
3004 addr
= kmap_atomic(src_page
);
3005 dst_addr
= kmap_atomic(dst_page
);
3007 flush_dcache_page(src_page
);
3008 flush_dcache_page(dst_page
);
3010 if (memcmp(addr
, dst_addr
, cmp_len
))
3013 kunmap_atomic(addr
);
3014 kunmap_atomic(dst_addr
);
3026 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64
*plen
,
3030 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
3032 if (off
+ olen
> inode
->i_size
|| off
+ olen
< off
)
3035 /* if we extend to eof, continue to block boundary */
3036 if (off
+ len
== inode
->i_size
)
3037 *plen
= len
= ALIGN(inode
->i_size
, bs
) - off
;
3039 /* Check that we are block aligned - btrfs_clone() requires this */
3040 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
3046 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 olen
,
3047 struct inode
*dst
, u64 dst_loff
)
3051 struct cmp_pages cmp
;
3053 u64 same_lock_start
= 0;
3054 u64 same_lock_len
= 0;
3065 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3070 * Single inode case wants the same checks, except we
3071 * don't want our length pushed out past i_size as
3072 * comparing that data range makes no sense.
3074 * extent_same_check_offsets() will do this for an
3075 * unaligned length at i_size, so catch it here and
3076 * reject the request.
3078 * This effectively means we require aligned extents
3079 * for the single-inode case, whereas the other cases
3080 * allow an unaligned length so long as it ends at
3088 /* Check for overlapping ranges */
3089 if (dst_loff
+ len
> loff
&& dst_loff
< loff
+ len
) {
3094 same_lock_start
= min_t(u64
, loff
, dst_loff
);
3095 same_lock_len
= max_t(u64
, loff
, dst_loff
) + len
- same_lock_start
;
3097 btrfs_double_inode_lock(src
, dst
);
3099 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3103 ret
= extent_same_check_offsets(dst
, dst_loff
, &len
, olen
);
3108 /* don't make the dst file partly checksummed */
3109 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3110 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
3116 ret
= btrfs_cmp_data_prepare(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3121 ret
= lock_extent_range(src
, same_lock_start
, same_lock_len
,
3124 ret
= btrfs_double_extent_lock(src
, loff
, dst
, dst_loff
, len
,
3127 * If one of the inodes has dirty pages in the respective range or
3128 * ordered extents, we need to flush dellaloc and wait for all ordered
3129 * extents in the range. We must unlock the pages and the ranges in the
3130 * io trees to avoid deadlocks when flushing delalloc (requires locking
3131 * pages) and when waiting for ordered extents to complete (they require
3134 if (ret
== -EAGAIN
) {
3136 * Ranges in the io trees already unlocked. Now unlock all
3137 * pages before waiting for all IO to complete.
3139 btrfs_cmp_data_free(&cmp
);
3141 btrfs_wait_ordered_range(src
, same_lock_start
,
3144 btrfs_wait_ordered_range(src
, loff
, len
);
3145 btrfs_wait_ordered_range(dst
, dst_loff
, len
);
3151 /* ranges in the io trees already unlocked */
3152 btrfs_cmp_data_free(&cmp
);
3156 /* pass original length for comparison so we stay within i_size */
3157 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3159 ret
= btrfs_clone(src
, dst
, loff
, olen
, len
, dst_loff
, 1);
3162 unlock_extent(&BTRFS_I(src
)->io_tree
, same_lock_start
,
3163 same_lock_start
+ same_lock_len
- 1);
3165 btrfs_double_extent_unlock(src
, loff
, dst
, dst_loff
, len
);
3167 btrfs_cmp_data_free(&cmp
);
3172 btrfs_double_inode_unlock(src
, dst
);
3177 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3179 ssize_t
btrfs_dedupe_file_range(struct file
*src_file
, u64 loff
, u64 olen
,
3180 struct file
*dst_file
, u64 dst_loff
)
3182 struct inode
*src
= file_inode(src_file
);
3183 struct inode
*dst
= file_inode(dst_file
);
3184 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
3187 if (olen
> BTRFS_MAX_DEDUPE_LEN
)
3188 olen
= BTRFS_MAX_DEDUPE_LEN
;
3190 if (WARN_ON_ONCE(bs
< PAGE_CACHE_SIZE
)) {
3192 * Btrfs does not support blocksize < page_size. As a
3193 * result, btrfs_cmp_data() won't correctly handle
3194 * this situation without an update.
3199 res
= btrfs_extent_same(src
, loff
, olen
, dst
, dst_loff
);
3205 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3206 struct inode
*inode
,
3212 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3215 inode_inc_iversion(inode
);
3216 if (!no_time_update
)
3217 inode
->i_mtime
= inode
->i_ctime
= current_fs_time(inode
->i_sb
);
3219 * We round up to the block size at eof when determining which
3220 * extents to clone above, but shouldn't round up the file size.
3222 if (endoff
> destoff
+ olen
)
3223 endoff
= destoff
+ olen
;
3224 if (endoff
> inode
->i_size
)
3225 btrfs_i_size_write(inode
, endoff
);
3227 ret
= btrfs_update_inode(trans
, root
, inode
);
3229 btrfs_abort_transaction(trans
, root
, ret
);
3230 btrfs_end_transaction(trans
, root
);
3233 ret
= btrfs_end_transaction(trans
, root
);
3238 static void clone_update_extent_map(struct inode
*inode
,
3239 const struct btrfs_trans_handle
*trans
,
3240 const struct btrfs_path
*path
,
3241 const u64 hole_offset
,
3244 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
3245 struct extent_map
*em
;
3248 em
= alloc_extent_map();
3250 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3251 &BTRFS_I(inode
)->runtime_flags
);
3256 struct btrfs_file_extent_item
*fi
;
3258 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3259 struct btrfs_file_extent_item
);
3260 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3261 em
->generation
= -1;
3262 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3263 BTRFS_FILE_EXTENT_INLINE
)
3264 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3265 &BTRFS_I(inode
)->runtime_flags
);
3267 em
->start
= hole_offset
;
3269 em
->ram_bytes
= em
->len
;
3270 em
->orig_start
= hole_offset
;
3271 em
->block_start
= EXTENT_MAP_HOLE
;
3273 em
->orig_block_len
= 0;
3274 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3275 em
->generation
= trans
->transid
;
3279 write_lock(&em_tree
->lock
);
3280 ret
= add_extent_mapping(em_tree
, em
, 1);
3281 write_unlock(&em_tree
->lock
);
3282 if (ret
!= -EEXIST
) {
3283 free_extent_map(em
);
3286 btrfs_drop_extent_cache(inode
, em
->start
,
3287 em
->start
+ em
->len
- 1, 0);
3291 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3292 &BTRFS_I(inode
)->runtime_flags
);
3296 * Make sure we do not end up inserting an inline extent into a file that has
3297 * already other (non-inline) extents. If a file has an inline extent it can
3298 * not have any other extents and the (single) inline extent must start at the
3299 * file offset 0. Failing to respect these rules will lead to file corruption,
3300 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3302 * We can have extents that have been already written to disk or we can have
3303 * dirty ranges still in delalloc, in which case the extent maps and items are
3304 * created only when we run delalloc, and the delalloc ranges might fall outside
3305 * the range we are currently locking in the inode's io tree. So we check the
3306 * inode's i_size because of that (i_size updates are done while holding the
3307 * i_mutex, which we are holding here).
3308 * We also check to see if the inode has a size not greater than "datal" but has
3309 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3310 * protected against such concurrent fallocate calls by the i_mutex).
3312 * If the file has no extents but a size greater than datal, do not allow the
3313 * copy because we would need turn the inline extent into a non-inline one (even
3314 * with NO_HOLES enabled). If we find our destination inode only has one inline
3315 * extent, just overwrite it with the source inline extent if its size is less
3316 * than the source extent's size, or we could copy the source inline extent's
3317 * data into the destination inode's inline extent if the later is greater then
3320 static int clone_copy_inline_extent(struct inode
*src
,
3322 struct btrfs_trans_handle
*trans
,
3323 struct btrfs_path
*path
,
3324 struct btrfs_key
*new_key
,
3325 const u64 drop_start
,
3331 struct btrfs_root
*root
= BTRFS_I(dst
)->root
;
3332 const u64 aligned_end
= ALIGN(new_key
->offset
+ datal
,
3335 struct btrfs_key key
;
3337 if (new_key
->offset
> 0)
3340 key
.objectid
= btrfs_ino(dst
);
3341 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3343 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
3346 } else if (ret
> 0) {
3347 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3348 ret
= btrfs_next_leaf(root
, path
);
3352 goto copy_inline_extent
;
3354 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
3355 if (key
.objectid
== btrfs_ino(dst
) &&
3356 key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3357 ASSERT(key
.offset
> 0);
3360 } else if (i_size_read(dst
) <= datal
) {
3361 struct btrfs_file_extent_item
*ei
;
3365 * If the file size is <= datal, make sure there are no other
3366 * extents following (can happen do to an fallocate call with
3367 * the flag FALLOC_FL_KEEP_SIZE).
3369 ei
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3370 struct btrfs_file_extent_item
);
3372 * If it's an inline extent, it can not have other extents
3375 if (btrfs_file_extent_type(path
->nodes
[0], ei
) ==
3376 BTRFS_FILE_EXTENT_INLINE
)
3377 goto copy_inline_extent
;
3379 ext_len
= btrfs_file_extent_num_bytes(path
->nodes
[0], ei
);
3380 if (ext_len
> aligned_end
)
3383 ret
= btrfs_next_item(root
, path
);
3386 } else if (ret
== 0) {
3387 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3389 if (key
.objectid
== btrfs_ino(dst
) &&
3390 key
.type
== BTRFS_EXTENT_DATA_KEY
)
3397 * We have no extent items, or we have an extent at offset 0 which may
3398 * or may not be inlined. All these cases are dealt the same way.
3400 if (i_size_read(dst
) > datal
) {
3402 * If the destination inode has an inline extent...
3403 * This would require copying the data from the source inline
3404 * extent into the beginning of the destination's inline extent.
3405 * But this is really complex, both extents can be compressed
3406 * or just one of them, which would require decompressing and
3407 * re-compressing data (which could increase the new compressed
3408 * size, not allowing the compressed data to fit anymore in an
3410 * So just don't support this case for now (it should be rare,
3411 * we are not really saving space when cloning inline extents).
3416 btrfs_release_path(path
);
3417 ret
= btrfs_drop_extents(trans
, root
, dst
, drop_start
, aligned_end
, 1);
3420 ret
= btrfs_insert_empty_item(trans
, root
, path
, new_key
, size
);
3425 const u32 start
= btrfs_file_extent_calc_inline_size(0);
3427 memmove(inline_data
+ start
, inline_data
+ start
+ skip
, datal
);
3430 write_extent_buffer(path
->nodes
[0], inline_data
,
3431 btrfs_item_ptr_offset(path
->nodes
[0],
3434 inode_add_bytes(dst
, datal
);
3440 * btrfs_clone() - clone a range from inode file to another
3442 * @src: Inode to clone from
3443 * @inode: Inode to clone to
3444 * @off: Offset within source to start clone from
3445 * @olen: Original length, passed by user, of range to clone
3446 * @olen_aligned: Block-aligned value of olen
3447 * @destoff: Offset within @inode to start clone
3448 * @no_time_update: Whether to update mtime/ctime on the target inode
3450 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3451 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3452 const u64 destoff
, int no_time_update
)
3454 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3455 struct btrfs_path
*path
= NULL
;
3456 struct extent_buffer
*leaf
;
3457 struct btrfs_trans_handle
*trans
;
3459 struct btrfs_key key
;
3463 const u64 len
= olen_aligned
;
3464 u64 last_dest_end
= destoff
;
3467 buf
= vmalloc(root
->nodesize
);
3471 path
= btrfs_alloc_path();
3477 path
->reada
= READA_FORWARD
;
3479 key
.objectid
= btrfs_ino(src
);
3480 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3484 u64 next_key_min_offset
= key
.offset
+ 1;
3487 * note the key will change type as we walk through the
3490 path
->leave_spinning
= 1;
3491 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3496 * First search, if no extent item that starts at offset off was
3497 * found but the previous item is an extent item, it's possible
3498 * it might overlap our target range, therefore process it.
3500 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3501 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3502 path
->slots
[0] - 1);
3503 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3507 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3509 if (path
->slots
[0] >= nritems
) {
3510 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3515 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3517 leaf
= path
->nodes
[0];
3518 slot
= path
->slots
[0];
3520 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3521 if (key
.type
> BTRFS_EXTENT_DATA_KEY
||
3522 key
.objectid
!= btrfs_ino(src
))
3525 if (key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3526 struct btrfs_file_extent_item
*extent
;
3529 struct btrfs_key new_key
;
3530 u64 disko
= 0, diskl
= 0;
3531 u64 datao
= 0, datal
= 0;
3535 extent
= btrfs_item_ptr(leaf
, slot
,
3536 struct btrfs_file_extent_item
);
3537 comp
= btrfs_file_extent_compression(leaf
, extent
);
3538 type
= btrfs_file_extent_type(leaf
, extent
);
3539 if (type
== BTRFS_FILE_EXTENT_REG
||
3540 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3541 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3543 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3545 datao
= btrfs_file_extent_offset(leaf
, extent
);
3546 datal
= btrfs_file_extent_num_bytes(leaf
,
3548 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3549 /* take upper bound, may be compressed */
3550 datal
= btrfs_file_extent_ram_bytes(leaf
,
3555 * The first search might have left us at an extent
3556 * item that ends before our target range's start, can
3557 * happen if we have holes and NO_HOLES feature enabled.
3559 if (key
.offset
+ datal
<= off
) {
3562 } else if (key
.offset
>= off
+ len
) {
3565 next_key_min_offset
= key
.offset
+ datal
;
3566 size
= btrfs_item_size_nr(leaf
, slot
);
3567 read_extent_buffer(leaf
, buf
,
3568 btrfs_item_ptr_offset(leaf
, slot
),
3571 btrfs_release_path(path
);
3572 path
->leave_spinning
= 0;
3574 memcpy(&new_key
, &key
, sizeof(new_key
));
3575 new_key
.objectid
= btrfs_ino(inode
);
3576 if (off
<= key
.offset
)
3577 new_key
.offset
= key
.offset
+ destoff
- off
;
3579 new_key
.offset
= destoff
;
3582 * Deal with a hole that doesn't have an extent item
3583 * that represents it (NO_HOLES feature enabled).
3584 * This hole is either in the middle of the cloning
3585 * range or at the beginning (fully overlaps it or
3586 * partially overlaps it).
3588 if (new_key
.offset
!= last_dest_end
)
3589 drop_start
= last_dest_end
;
3591 drop_start
= new_key
.offset
;
3594 * 1 - adjusting old extent (we may have to split it)
3595 * 1 - add new extent
3598 trans
= btrfs_start_transaction(root
, 3);
3599 if (IS_ERR(trans
)) {
3600 ret
= PTR_ERR(trans
);
3604 if (type
== BTRFS_FILE_EXTENT_REG
||
3605 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3607 * a | --- range to clone ---| b
3608 * | ------------- extent ------------- |
3611 /* subtract range b */
3612 if (key
.offset
+ datal
> off
+ len
)
3613 datal
= off
+ len
- key
.offset
;
3615 /* subtract range a */
3616 if (off
> key
.offset
) {
3617 datao
+= off
- key
.offset
;
3618 datal
-= off
- key
.offset
;
3621 ret
= btrfs_drop_extents(trans
, root
, inode
,
3623 new_key
.offset
+ datal
,
3626 if (ret
!= -EOPNOTSUPP
)
3627 btrfs_abort_transaction(trans
,
3629 btrfs_end_transaction(trans
, root
);
3633 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3636 btrfs_abort_transaction(trans
, root
,
3638 btrfs_end_transaction(trans
, root
);
3642 leaf
= path
->nodes
[0];
3643 slot
= path
->slots
[0];
3644 write_extent_buffer(leaf
, buf
,
3645 btrfs_item_ptr_offset(leaf
, slot
),
3648 extent
= btrfs_item_ptr(leaf
, slot
,
3649 struct btrfs_file_extent_item
);
3651 /* disko == 0 means it's a hole */
3655 btrfs_set_file_extent_offset(leaf
, extent
,
3657 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3661 inode_add_bytes(inode
, datal
);
3662 ret
= btrfs_inc_extent_ref(trans
, root
,
3664 root
->root_key
.objectid
,
3666 new_key
.offset
- datao
);
3668 btrfs_abort_transaction(trans
,
3671 btrfs_end_transaction(trans
,
3677 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3681 if (off
> key
.offset
) {
3682 skip
= off
- key
.offset
;
3683 new_key
.offset
+= skip
;
3686 if (key
.offset
+ datal
> off
+ len
)
3687 trim
= key
.offset
+ datal
- (off
+ len
);
3689 if (comp
&& (skip
|| trim
)) {
3691 btrfs_end_transaction(trans
, root
);
3694 size
-= skip
+ trim
;
3695 datal
-= skip
+ trim
;
3697 ret
= clone_copy_inline_extent(src
, inode
,
3704 if (ret
!= -EOPNOTSUPP
)
3705 btrfs_abort_transaction(trans
,
3708 btrfs_end_transaction(trans
, root
);
3711 leaf
= path
->nodes
[0];
3712 slot
= path
->slots
[0];
3715 /* If we have an implicit hole (NO_HOLES feature). */
3716 if (drop_start
< new_key
.offset
)
3717 clone_update_extent_map(inode
, trans
,
3719 new_key
.offset
- drop_start
);
3721 clone_update_extent_map(inode
, trans
, path
, 0, 0);
3723 btrfs_mark_buffer_dirty(leaf
);
3724 btrfs_release_path(path
);
3726 last_dest_end
= ALIGN(new_key
.offset
+ datal
,
3728 ret
= clone_finish_inode_update(trans
, inode
,
3734 if (new_key
.offset
+ datal
>= destoff
+ len
)
3737 btrfs_release_path(path
);
3738 key
.offset
= next_key_min_offset
;
3742 if (last_dest_end
< destoff
+ len
) {
3744 * We have an implicit hole (NO_HOLES feature is enabled) that
3745 * fully or partially overlaps our cloning range at its end.
3747 btrfs_release_path(path
);
3750 * 1 - remove extent(s)
3753 trans
= btrfs_start_transaction(root
, 2);
3754 if (IS_ERR(trans
)) {
3755 ret
= PTR_ERR(trans
);
3758 ret
= btrfs_drop_extents(trans
, root
, inode
,
3759 last_dest_end
, destoff
+ len
, 1);
3761 if (ret
!= -EOPNOTSUPP
)
3762 btrfs_abort_transaction(trans
, root
, ret
);
3763 btrfs_end_transaction(trans
, root
);
3766 clone_update_extent_map(inode
, trans
, NULL
, last_dest_end
,
3767 destoff
+ len
- last_dest_end
);
3768 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3769 destoff
, olen
, no_time_update
);
3773 btrfs_free_path(path
);
3778 static noinline
int btrfs_clone_files(struct file
*file
, struct file
*file_src
,
3779 u64 off
, u64 olen
, u64 destoff
)
3781 struct inode
*inode
= file_inode(file
);
3782 struct inode
*src
= file_inode(file_src
);
3783 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3786 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
3787 int same_inode
= src
== inode
;
3791 * - split compressed inline extents. annoying: we need to
3792 * decompress into destination's address_space (the file offset
3793 * may change, so source mapping won't do), then recompress (or
3794 * otherwise reinsert) a subrange.
3796 * - split destination inode's inline extents. The inline extents can
3797 * be either compressed or non-compressed.
3800 if (btrfs_root_readonly(root
))
3803 if (file_src
->f_path
.mnt
!= file
->f_path
.mnt
||
3804 src
->i_sb
!= inode
->i_sb
)
3807 /* don't make the dst file partly checksummed */
3808 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3809 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3812 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3816 btrfs_double_inode_lock(src
, inode
);
3821 /* determine range to clone */
3823 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3826 olen
= len
= src
->i_size
- off
;
3827 /* if we extend to eof, continue to block boundary */
3828 if (off
+ len
== src
->i_size
)
3829 len
= ALIGN(src
->i_size
, bs
) - off
;
3836 /* verify the end result is block aligned */
3837 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3838 !IS_ALIGNED(destoff
, bs
))
3841 /* verify if ranges are overlapped within the same file */
3843 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3847 if (destoff
> inode
->i_size
) {
3848 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3854 * Lock the target range too. Right after we replace the file extent
3855 * items in the fs tree (which now point to the cloned data), we might
3856 * have a worker replace them with extent items relative to a write
3857 * operation that was issued before this clone operation (i.e. confront
3858 * with inode.c:btrfs_finish_ordered_io).
3861 u64 lock_start
= min_t(u64
, off
, destoff
);
3862 u64 lock_len
= max_t(u64
, off
, destoff
) + len
- lock_start
;
3864 ret
= lock_extent_range(src
, lock_start
, lock_len
, true);
3866 ret
= btrfs_double_extent_lock(src
, off
, inode
, destoff
, len
,
3871 /* ranges in the io trees already unlocked */
3875 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
, 0);
3878 u64 lock_start
= min_t(u64
, off
, destoff
);
3879 u64 lock_end
= max_t(u64
, off
, destoff
) + len
- 1;
3881 unlock_extent(&BTRFS_I(src
)->io_tree
, lock_start
, lock_end
);
3883 btrfs_double_extent_unlock(src
, off
, inode
, destoff
, len
);
3886 * Truncate page cache pages so that future reads will see the cloned
3887 * data immediately and not the previous data.
3889 truncate_inode_pages_range(&inode
->i_data
,
3890 round_down(destoff
, PAGE_CACHE_SIZE
),
3891 round_up(destoff
+ len
, PAGE_CACHE_SIZE
) - 1);
3894 btrfs_double_inode_unlock(src
, inode
);
3900 ssize_t
btrfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
3901 struct file
*file_out
, loff_t pos_out
,
3902 size_t len
, unsigned int flags
)
3906 ret
= btrfs_clone_files(file_out
, file_in
, pos_in
, len
, pos_out
);
3912 int btrfs_clone_file_range(struct file
*src_file
, loff_t off
,
3913 struct file
*dst_file
, loff_t destoff
, u64 len
)
3915 return btrfs_clone_files(dst_file
, src_file
, off
, len
, destoff
);
3919 * there are many ways the trans_start and trans_end ioctls can lead
3920 * to deadlocks. They should only be used by applications that
3921 * basically own the machine, and have a very in depth understanding
3922 * of all the possible deadlocks and enospc problems.
3924 static long btrfs_ioctl_trans_start(struct file
*file
)
3926 struct inode
*inode
= file_inode(file
);
3927 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3928 struct btrfs_trans_handle
*trans
;
3932 if (!capable(CAP_SYS_ADMIN
))
3936 if (file
->private_data
)
3940 if (btrfs_root_readonly(root
))
3943 ret
= mnt_want_write_file(file
);
3947 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
3950 trans
= btrfs_start_ioctl_transaction(root
);
3954 file
->private_data
= trans
;
3958 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3959 mnt_drop_write_file(file
);
3964 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
3966 struct inode
*inode
= file_inode(file
);
3967 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3968 struct btrfs_root
*new_root
;
3969 struct btrfs_dir_item
*di
;
3970 struct btrfs_trans_handle
*trans
;
3971 struct btrfs_path
*path
;
3972 struct btrfs_key location
;
3973 struct btrfs_disk_key disk_key
;
3978 if (!capable(CAP_SYS_ADMIN
))
3981 ret
= mnt_want_write_file(file
);
3985 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
3991 objectid
= BTRFS_FS_TREE_OBJECTID
;
3993 location
.objectid
= objectid
;
3994 location
.type
= BTRFS_ROOT_ITEM_KEY
;
3995 location
.offset
= (u64
)-1;
3997 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
3998 if (IS_ERR(new_root
)) {
3999 ret
= PTR_ERR(new_root
);
4003 path
= btrfs_alloc_path();
4008 path
->leave_spinning
= 1;
4010 trans
= btrfs_start_transaction(root
, 1);
4011 if (IS_ERR(trans
)) {
4012 btrfs_free_path(path
);
4013 ret
= PTR_ERR(trans
);
4017 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
4018 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
4019 dir_id
, "default", 7, 1);
4020 if (IS_ERR_OR_NULL(di
)) {
4021 btrfs_free_path(path
);
4022 btrfs_end_transaction(trans
, root
);
4023 btrfs_err(new_root
->fs_info
, "Umm, you don't have the default dir"
4024 "item, this isn't going to work");
4029 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
4030 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
4031 btrfs_mark_buffer_dirty(path
->nodes
[0]);
4032 btrfs_free_path(path
);
4034 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
4035 btrfs_end_transaction(trans
, root
);
4037 mnt_drop_write_file(file
);
4041 void btrfs_get_block_group_info(struct list_head
*groups_list
,
4042 struct btrfs_ioctl_space_info
*space
)
4044 struct btrfs_block_group_cache
*block_group
;
4046 space
->total_bytes
= 0;
4047 space
->used_bytes
= 0;
4049 list_for_each_entry(block_group
, groups_list
, list
) {
4050 space
->flags
= block_group
->flags
;
4051 space
->total_bytes
+= block_group
->key
.offset
;
4052 space
->used_bytes
+=
4053 btrfs_block_group_used(&block_group
->item
);
4057 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
4059 struct btrfs_ioctl_space_args space_args
;
4060 struct btrfs_ioctl_space_info space
;
4061 struct btrfs_ioctl_space_info
*dest
;
4062 struct btrfs_ioctl_space_info
*dest_orig
;
4063 struct btrfs_ioctl_space_info __user
*user_dest
;
4064 struct btrfs_space_info
*info
;
4065 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
4066 BTRFS_BLOCK_GROUP_SYSTEM
,
4067 BTRFS_BLOCK_GROUP_METADATA
,
4068 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
4075 if (copy_from_user(&space_args
,
4076 (struct btrfs_ioctl_space_args __user
*)arg
,
4077 sizeof(space_args
)))
4080 for (i
= 0; i
< num_types
; i
++) {
4081 struct btrfs_space_info
*tmp
;
4085 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4087 if (tmp
->flags
== types
[i
]) {
4097 down_read(&info
->groups_sem
);
4098 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4099 if (!list_empty(&info
->block_groups
[c
]))
4102 up_read(&info
->groups_sem
);
4106 * Global block reserve, exported as a space_info
4110 /* space_slots == 0 means they are asking for a count */
4111 if (space_args
.space_slots
== 0) {
4112 space_args
.total_spaces
= slot_count
;
4116 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
4118 alloc_size
= sizeof(*dest
) * slot_count
;
4120 /* we generally have at most 6 or so space infos, one for each raid
4121 * level. So, a whole page should be more than enough for everyone
4123 if (alloc_size
> PAGE_CACHE_SIZE
)
4126 space_args
.total_spaces
= 0;
4127 dest
= kmalloc(alloc_size
, GFP_KERNEL
);
4132 /* now we have a buffer to copy into */
4133 for (i
= 0; i
< num_types
; i
++) {
4134 struct btrfs_space_info
*tmp
;
4141 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4143 if (tmp
->flags
== types
[i
]) {
4152 down_read(&info
->groups_sem
);
4153 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4154 if (!list_empty(&info
->block_groups
[c
])) {
4155 btrfs_get_block_group_info(
4156 &info
->block_groups
[c
], &space
);
4157 memcpy(dest
, &space
, sizeof(space
));
4159 space_args
.total_spaces
++;
4165 up_read(&info
->groups_sem
);
4169 * Add global block reserve
4172 struct btrfs_block_rsv
*block_rsv
= &root
->fs_info
->global_block_rsv
;
4174 spin_lock(&block_rsv
->lock
);
4175 space
.total_bytes
= block_rsv
->size
;
4176 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4177 spin_unlock(&block_rsv
->lock
);
4178 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4179 memcpy(dest
, &space
, sizeof(space
));
4180 space_args
.total_spaces
++;
4183 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4184 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4186 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4191 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4198 * there are many ways the trans_start and trans_end ioctls can lead
4199 * to deadlocks. They should only be used by applications that
4200 * basically own the machine, and have a very in depth understanding
4201 * of all the possible deadlocks and enospc problems.
4203 long btrfs_ioctl_trans_end(struct file
*file
)
4205 struct inode
*inode
= file_inode(file
);
4206 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4207 struct btrfs_trans_handle
*trans
;
4209 trans
= file
->private_data
;
4212 file
->private_data
= NULL
;
4214 btrfs_end_transaction(trans
, root
);
4216 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4218 mnt_drop_write_file(file
);
4222 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4225 struct btrfs_trans_handle
*trans
;
4229 trans
= btrfs_attach_transaction_barrier(root
);
4230 if (IS_ERR(trans
)) {
4231 if (PTR_ERR(trans
) != -ENOENT
)
4232 return PTR_ERR(trans
);
4234 /* No running transaction, don't bother */
4235 transid
= root
->fs_info
->last_trans_committed
;
4238 transid
= trans
->transid
;
4239 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
4241 btrfs_end_transaction(trans
, root
);
4246 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4251 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
4257 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4260 transid
= 0; /* current trans */
4262 return btrfs_wait_for_commit(root
, transid
);
4265 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4267 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4268 struct btrfs_ioctl_scrub_args
*sa
;
4271 if (!capable(CAP_SYS_ADMIN
))
4274 sa
= memdup_user(arg
, sizeof(*sa
));
4278 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4279 ret
= mnt_want_write_file(file
);
4284 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4285 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4288 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4291 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4292 mnt_drop_write_file(file
);
4298 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
4300 if (!capable(CAP_SYS_ADMIN
))
4303 return btrfs_scrub_cancel(root
->fs_info
);
4306 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
4309 struct btrfs_ioctl_scrub_args
*sa
;
4312 if (!capable(CAP_SYS_ADMIN
))
4315 sa
= memdup_user(arg
, sizeof(*sa
));
4319 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
4321 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4328 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
4331 struct btrfs_ioctl_get_dev_stats
*sa
;
4334 sa
= memdup_user(arg
, sizeof(*sa
));
4338 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4343 ret
= btrfs_get_dev_stats(root
, sa
);
4345 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4352 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
4354 struct btrfs_ioctl_dev_replace_args
*p
;
4357 if (!capable(CAP_SYS_ADMIN
))
4360 p
= memdup_user(arg
, sizeof(*p
));
4365 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4366 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
) {
4371 &root
->fs_info
->mutually_exclusive_operation_running
,
4373 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4375 ret
= btrfs_dev_replace_start(root
, p
);
4377 &root
->fs_info
->mutually_exclusive_operation_running
,
4381 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4382 btrfs_dev_replace_status(root
->fs_info
, p
);
4385 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4386 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
4393 if (copy_to_user(arg
, p
, sizeof(*p
)))
4400 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4406 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4407 struct inode_fs_paths
*ipath
= NULL
;
4408 struct btrfs_path
*path
;
4410 if (!capable(CAP_DAC_READ_SEARCH
))
4413 path
= btrfs_alloc_path();
4419 ipa
= memdup_user(arg
, sizeof(*ipa
));
4426 size
= min_t(u32
, ipa
->size
, 4096);
4427 ipath
= init_ipath(size
, root
, path
);
4428 if (IS_ERR(ipath
)) {
4429 ret
= PTR_ERR(ipath
);
4434 ret
= paths_from_inode(ipa
->inum
, ipath
);
4438 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4439 rel_ptr
= ipath
->fspath
->val
[i
] -
4440 (u64
)(unsigned long)ipath
->fspath
->val
;
4441 ipath
->fspath
->val
[i
] = rel_ptr
;
4444 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
4445 (void *)(unsigned long)ipath
->fspath
, size
);
4452 btrfs_free_path(path
);
4459 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4461 struct btrfs_data_container
*inodes
= ctx
;
4462 const size_t c
= 3 * sizeof(u64
);
4464 if (inodes
->bytes_left
>= c
) {
4465 inodes
->bytes_left
-= c
;
4466 inodes
->val
[inodes
->elem_cnt
] = inum
;
4467 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4468 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4469 inodes
->elem_cnt
+= 3;
4471 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4472 inodes
->bytes_left
= 0;
4473 inodes
->elem_missed
+= 3;
4479 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
4484 struct btrfs_ioctl_logical_ino_args
*loi
;
4485 struct btrfs_data_container
*inodes
= NULL
;
4486 struct btrfs_path
*path
= NULL
;
4488 if (!capable(CAP_SYS_ADMIN
))
4491 loi
= memdup_user(arg
, sizeof(*loi
));
4498 path
= btrfs_alloc_path();
4504 size
= min_t(u32
, loi
->size
, SZ_64K
);
4505 inodes
= init_data_container(size
);
4506 if (IS_ERR(inodes
)) {
4507 ret
= PTR_ERR(inodes
);
4512 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
4513 build_ino_list
, inodes
);
4519 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
4520 (void *)(unsigned long)inodes
, size
);
4525 btrfs_free_path(path
);
4532 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
4533 struct btrfs_ioctl_balance_args
*bargs
)
4535 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4537 bargs
->flags
= bctl
->flags
;
4539 if (atomic_read(&fs_info
->balance_running
))
4540 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4541 if (atomic_read(&fs_info
->balance_pause_req
))
4542 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4543 if (atomic_read(&fs_info
->balance_cancel_req
))
4544 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4546 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4547 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4548 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4551 spin_lock(&fs_info
->balance_lock
);
4552 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4553 spin_unlock(&fs_info
->balance_lock
);
4555 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4559 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4561 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4562 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4563 struct btrfs_ioctl_balance_args
*bargs
;
4564 struct btrfs_balance_control
*bctl
;
4565 bool need_unlock
; /* for mut. excl. ops lock */
4568 if (!capable(CAP_SYS_ADMIN
))
4571 ret
= mnt_want_write_file(file
);
4576 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4577 mutex_lock(&fs_info
->volume_mutex
);
4578 mutex_lock(&fs_info
->balance_mutex
);
4584 * mut. excl. ops lock is locked. Three possibilites:
4585 * (1) some other op is running
4586 * (2) balance is running
4587 * (3) balance is paused -- special case (think resume)
4589 mutex_lock(&fs_info
->balance_mutex
);
4590 if (fs_info
->balance_ctl
) {
4591 /* this is either (2) or (3) */
4592 if (!atomic_read(&fs_info
->balance_running
)) {
4593 mutex_unlock(&fs_info
->balance_mutex
);
4594 if (!mutex_trylock(&fs_info
->volume_mutex
))
4596 mutex_lock(&fs_info
->balance_mutex
);
4598 if (fs_info
->balance_ctl
&&
4599 !atomic_read(&fs_info
->balance_running
)) {
4601 need_unlock
= false;
4605 mutex_unlock(&fs_info
->balance_mutex
);
4606 mutex_unlock(&fs_info
->volume_mutex
);
4610 mutex_unlock(&fs_info
->balance_mutex
);
4616 mutex_unlock(&fs_info
->balance_mutex
);
4617 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4622 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4625 bargs
= memdup_user(arg
, sizeof(*bargs
));
4626 if (IS_ERR(bargs
)) {
4627 ret
= PTR_ERR(bargs
);
4631 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4632 if (!fs_info
->balance_ctl
) {
4637 bctl
= fs_info
->balance_ctl
;
4638 spin_lock(&fs_info
->balance_lock
);
4639 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4640 spin_unlock(&fs_info
->balance_lock
);
4648 if (fs_info
->balance_ctl
) {
4653 bctl
= kzalloc(sizeof(*bctl
), GFP_KERNEL
);
4659 bctl
->fs_info
= fs_info
;
4661 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4662 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4663 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4665 bctl
->flags
= bargs
->flags
;
4667 /* balance everything - no filters */
4668 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4671 if (bctl
->flags
& ~(BTRFS_BALANCE_ARGS_MASK
| BTRFS_BALANCE_TYPE_MASK
)) {
4678 * Ownership of bctl and mutually_exclusive_operation_running
4679 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4680 * or, if restriper was paused all the way until unmount, in
4681 * free_fs_info. mutually_exclusive_operation_running is
4682 * cleared in __cancel_balance.
4684 need_unlock
= false;
4686 ret
= btrfs_balance(bctl
, bargs
);
4690 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4699 mutex_unlock(&fs_info
->balance_mutex
);
4700 mutex_unlock(&fs_info
->volume_mutex
);
4702 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4704 mnt_drop_write_file(file
);
4708 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4710 if (!capable(CAP_SYS_ADMIN
))
4714 case BTRFS_BALANCE_CTL_PAUSE
:
4715 return btrfs_pause_balance(root
->fs_info
);
4716 case BTRFS_BALANCE_CTL_CANCEL
:
4717 return btrfs_cancel_balance(root
->fs_info
);
4723 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4726 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4727 struct btrfs_ioctl_balance_args
*bargs
;
4730 if (!capable(CAP_SYS_ADMIN
))
4733 mutex_lock(&fs_info
->balance_mutex
);
4734 if (!fs_info
->balance_ctl
) {
4739 bargs
= kzalloc(sizeof(*bargs
), GFP_KERNEL
);
4745 update_ioctl_balance_args(fs_info
, 1, bargs
);
4747 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4752 mutex_unlock(&fs_info
->balance_mutex
);
4756 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4758 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4759 struct btrfs_ioctl_quota_ctl_args
*sa
;
4760 struct btrfs_trans_handle
*trans
= NULL
;
4764 if (!capable(CAP_SYS_ADMIN
))
4767 ret
= mnt_want_write_file(file
);
4771 sa
= memdup_user(arg
, sizeof(*sa
));
4777 down_write(&root
->fs_info
->subvol_sem
);
4778 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
4779 if (IS_ERR(trans
)) {
4780 ret
= PTR_ERR(trans
);
4785 case BTRFS_QUOTA_CTL_ENABLE
:
4786 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
4788 case BTRFS_QUOTA_CTL_DISABLE
:
4789 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
4796 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
4801 up_write(&root
->fs_info
->subvol_sem
);
4803 mnt_drop_write_file(file
);
4807 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4809 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4810 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4811 struct btrfs_trans_handle
*trans
;
4815 if (!capable(CAP_SYS_ADMIN
))
4818 ret
= mnt_want_write_file(file
);
4822 sa
= memdup_user(arg
, sizeof(*sa
));
4828 trans
= btrfs_join_transaction(root
);
4829 if (IS_ERR(trans
)) {
4830 ret
= PTR_ERR(trans
);
4834 /* FIXME: check if the IDs really exist */
4836 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
4839 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
4843 /* update qgroup status and info */
4844 err
= btrfs_run_qgroups(trans
, root
->fs_info
);
4846 btrfs_std_error(root
->fs_info
, ret
,
4847 "failed to update qgroup status and info\n");
4848 err
= btrfs_end_transaction(trans
, root
);
4855 mnt_drop_write_file(file
);
4859 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4861 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4862 struct btrfs_ioctl_qgroup_create_args
*sa
;
4863 struct btrfs_trans_handle
*trans
;
4867 if (!capable(CAP_SYS_ADMIN
))
4870 ret
= mnt_want_write_file(file
);
4874 sa
= memdup_user(arg
, sizeof(*sa
));
4880 if (!sa
->qgroupid
) {
4885 trans
= btrfs_join_transaction(root
);
4886 if (IS_ERR(trans
)) {
4887 ret
= PTR_ERR(trans
);
4891 /* FIXME: check if the IDs really exist */
4893 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4895 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4898 err
= btrfs_end_transaction(trans
, root
);
4905 mnt_drop_write_file(file
);
4909 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4911 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4912 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4913 struct btrfs_trans_handle
*trans
;
4918 if (!capable(CAP_SYS_ADMIN
))
4921 ret
= mnt_want_write_file(file
);
4925 sa
= memdup_user(arg
, sizeof(*sa
));
4931 trans
= btrfs_join_transaction(root
);
4932 if (IS_ERR(trans
)) {
4933 ret
= PTR_ERR(trans
);
4937 qgroupid
= sa
->qgroupid
;
4939 /* take the current subvol as qgroup */
4940 qgroupid
= root
->root_key
.objectid
;
4943 /* FIXME: check if the IDs really exist */
4944 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
4946 err
= btrfs_end_transaction(trans
, root
);
4953 mnt_drop_write_file(file
);
4957 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
4959 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4960 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4963 if (!capable(CAP_SYS_ADMIN
))
4966 ret
= mnt_want_write_file(file
);
4970 qsa
= memdup_user(arg
, sizeof(*qsa
));
4981 ret
= btrfs_qgroup_rescan(root
->fs_info
);
4986 mnt_drop_write_file(file
);
4990 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
4992 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4993 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4996 if (!capable(CAP_SYS_ADMIN
))
4999 qsa
= kzalloc(sizeof(*qsa
), GFP_KERNEL
);
5003 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
5005 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
5008 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
5015 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
5017 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5019 if (!capable(CAP_SYS_ADMIN
))
5022 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
5025 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
5026 struct btrfs_ioctl_received_subvol_args
*sa
)
5028 struct inode
*inode
= file_inode(file
);
5029 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5030 struct btrfs_root_item
*root_item
= &root
->root_item
;
5031 struct btrfs_trans_handle
*trans
;
5032 struct timespec ct
= current_fs_time(inode
->i_sb
);
5034 int received_uuid_changed
;
5036 if (!inode_owner_or_capable(inode
))
5039 ret
= mnt_want_write_file(file
);
5043 down_write(&root
->fs_info
->subvol_sem
);
5045 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
5050 if (btrfs_root_readonly(root
)) {
5057 * 2 - uuid items (received uuid + subvol uuid)
5059 trans
= btrfs_start_transaction(root
, 3);
5060 if (IS_ERR(trans
)) {
5061 ret
= PTR_ERR(trans
);
5066 sa
->rtransid
= trans
->transid
;
5067 sa
->rtime
.sec
= ct
.tv_sec
;
5068 sa
->rtime
.nsec
= ct
.tv_nsec
;
5070 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
5072 if (received_uuid_changed
&&
5073 !btrfs_is_empty_uuid(root_item
->received_uuid
))
5074 btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
5075 root_item
->received_uuid
,
5076 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5077 root
->root_key
.objectid
);
5078 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
5079 btrfs_set_root_stransid(root_item
, sa
->stransid
);
5080 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
5081 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
5082 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
5083 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
5084 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
5086 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
5087 &root
->root_key
, &root
->root_item
);
5089 btrfs_end_transaction(trans
, root
);
5092 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
5093 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
5095 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5096 root
->root_key
.objectid
);
5097 if (ret
< 0 && ret
!= -EEXIST
) {
5098 btrfs_abort_transaction(trans
, root
, ret
);
5102 ret
= btrfs_commit_transaction(trans
, root
);
5104 btrfs_abort_transaction(trans
, root
, ret
);
5109 up_write(&root
->fs_info
->subvol_sem
);
5110 mnt_drop_write_file(file
);
5115 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
5118 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
5119 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
5122 args32
= memdup_user(arg
, sizeof(*args32
));
5123 if (IS_ERR(args32
)) {
5124 ret
= PTR_ERR(args32
);
5129 args64
= kmalloc(sizeof(*args64
), GFP_KERNEL
);
5135 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
5136 args64
->stransid
= args32
->stransid
;
5137 args64
->rtransid
= args32
->rtransid
;
5138 args64
->stime
.sec
= args32
->stime
.sec
;
5139 args64
->stime
.nsec
= args32
->stime
.nsec
;
5140 args64
->rtime
.sec
= args32
->rtime
.sec
;
5141 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
5142 args64
->flags
= args32
->flags
;
5144 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
5148 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
5149 args32
->stransid
= args64
->stransid
;
5150 args32
->rtransid
= args64
->rtransid
;
5151 args32
->stime
.sec
= args64
->stime
.sec
;
5152 args32
->stime
.nsec
= args64
->stime
.nsec
;
5153 args32
->rtime
.sec
= args64
->rtime
.sec
;
5154 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
5155 args32
->flags
= args64
->flags
;
5157 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
5168 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
5171 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
5174 sa
= memdup_user(arg
, sizeof(*sa
));
5181 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5186 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5195 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5197 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5200 char label
[BTRFS_LABEL_SIZE
];
5202 spin_lock(&root
->fs_info
->super_lock
);
5203 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5204 spin_unlock(&root
->fs_info
->super_lock
);
5206 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5208 if (len
== BTRFS_LABEL_SIZE
) {
5209 btrfs_warn(root
->fs_info
,
5210 "label is too long, return the first %zu bytes", --len
);
5213 ret
= copy_to_user(arg
, label
, len
);
5215 return ret
? -EFAULT
: 0;
5218 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5220 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5221 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5222 struct btrfs_trans_handle
*trans
;
5223 char label
[BTRFS_LABEL_SIZE
];
5226 if (!capable(CAP_SYS_ADMIN
))
5229 if (copy_from_user(label
, arg
, sizeof(label
)))
5232 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5233 btrfs_err(root
->fs_info
, "unable to set label with more than %d bytes",
5234 BTRFS_LABEL_SIZE
- 1);
5238 ret
= mnt_want_write_file(file
);
5242 trans
= btrfs_start_transaction(root
, 0);
5243 if (IS_ERR(trans
)) {
5244 ret
= PTR_ERR(trans
);
5248 spin_lock(&root
->fs_info
->super_lock
);
5249 strcpy(super_block
->label
, label
);
5250 spin_unlock(&root
->fs_info
->super_lock
);
5251 ret
= btrfs_commit_transaction(trans
, root
);
5254 mnt_drop_write_file(file
);
5258 #define INIT_FEATURE_FLAGS(suffix) \
5259 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5260 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5261 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5263 int btrfs_ioctl_get_supported_features(void __user
*arg
)
5265 static const struct btrfs_ioctl_feature_flags features
[3] = {
5266 INIT_FEATURE_FLAGS(SUPP
),
5267 INIT_FEATURE_FLAGS(SAFE_SET
),
5268 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5271 if (copy_to_user(arg
, &features
, sizeof(features
)))
5277 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5279 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5280 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5281 struct btrfs_ioctl_feature_flags features
;
5283 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5284 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5285 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5287 if (copy_to_user(arg
, &features
, sizeof(features
)))
5293 static int check_feature_bits(struct btrfs_root
*root
,
5294 enum btrfs_feature_set set
,
5295 u64 change_mask
, u64 flags
, u64 supported_flags
,
5296 u64 safe_set
, u64 safe_clear
)
5298 const char *type
= btrfs_feature_set_names
[set
];
5300 u64 disallowed
, unsupported
;
5301 u64 set_mask
= flags
& change_mask
;
5302 u64 clear_mask
= ~flags
& change_mask
;
5304 unsupported
= set_mask
& ~supported_flags
;
5306 names
= btrfs_printable_features(set
, unsupported
);
5308 btrfs_warn(root
->fs_info
,
5309 "this kernel does not support the %s feature bit%s",
5310 names
, strchr(names
, ',') ? "s" : "");
5313 btrfs_warn(root
->fs_info
,
5314 "this kernel does not support %s bits 0x%llx",
5319 disallowed
= set_mask
& ~safe_set
;
5321 names
= btrfs_printable_features(set
, disallowed
);
5323 btrfs_warn(root
->fs_info
,
5324 "can't set the %s feature bit%s while mounted",
5325 names
, strchr(names
, ',') ? "s" : "");
5328 btrfs_warn(root
->fs_info
,
5329 "can't set %s bits 0x%llx while mounted",
5334 disallowed
= clear_mask
& ~safe_clear
;
5336 names
= btrfs_printable_features(set
, disallowed
);
5338 btrfs_warn(root
->fs_info
,
5339 "can't clear the %s feature bit%s while mounted",
5340 names
, strchr(names
, ',') ? "s" : "");
5343 btrfs_warn(root
->fs_info
,
5344 "can't clear %s bits 0x%llx while mounted",
5352 #define check_feature(root, change_mask, flags, mask_base) \
5353 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5354 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5355 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5356 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5358 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5360 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5361 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5362 struct btrfs_ioctl_feature_flags flags
[2];
5363 struct btrfs_trans_handle
*trans
;
5367 if (!capable(CAP_SYS_ADMIN
))
5370 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5374 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5375 !flags
[0].incompat_flags
)
5378 ret
= check_feature(root
, flags
[0].compat_flags
,
5379 flags
[1].compat_flags
, COMPAT
);
5383 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
5384 flags
[1].compat_ro_flags
, COMPAT_RO
);
5388 ret
= check_feature(root
, flags
[0].incompat_flags
,
5389 flags
[1].incompat_flags
, INCOMPAT
);
5393 trans
= btrfs_start_transaction(root
, 0);
5395 return PTR_ERR(trans
);
5397 spin_lock(&root
->fs_info
->super_lock
);
5398 newflags
= btrfs_super_compat_flags(super_block
);
5399 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5400 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5401 btrfs_set_super_compat_flags(super_block
, newflags
);
5403 newflags
= btrfs_super_compat_ro_flags(super_block
);
5404 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5405 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5406 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5408 newflags
= btrfs_super_incompat_flags(super_block
);
5409 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5410 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5411 btrfs_set_super_incompat_flags(super_block
, newflags
);
5412 spin_unlock(&root
->fs_info
->super_lock
);
5414 return btrfs_commit_transaction(trans
, root
);
5417 long btrfs_ioctl(struct file
*file
, unsigned int
5418 cmd
, unsigned long arg
)
5420 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5421 void __user
*argp
= (void __user
*)arg
;
5424 case FS_IOC_GETFLAGS
:
5425 return btrfs_ioctl_getflags(file
, argp
);
5426 case FS_IOC_SETFLAGS
:
5427 return btrfs_ioctl_setflags(file
, argp
);
5428 case FS_IOC_GETVERSION
:
5429 return btrfs_ioctl_getversion(file
, argp
);
5431 return btrfs_ioctl_fitrim(file
, argp
);
5432 case BTRFS_IOC_SNAP_CREATE
:
5433 return btrfs_ioctl_snap_create(file
, argp
, 0);
5434 case BTRFS_IOC_SNAP_CREATE_V2
:
5435 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5436 case BTRFS_IOC_SUBVOL_CREATE
:
5437 return btrfs_ioctl_snap_create(file
, argp
, 1);
5438 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5439 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5440 case BTRFS_IOC_SNAP_DESTROY
:
5441 return btrfs_ioctl_snap_destroy(file
, argp
);
5442 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5443 return btrfs_ioctl_subvol_getflags(file
, argp
);
5444 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5445 return btrfs_ioctl_subvol_setflags(file
, argp
);
5446 case BTRFS_IOC_DEFAULT_SUBVOL
:
5447 return btrfs_ioctl_default_subvol(file
, argp
);
5448 case BTRFS_IOC_DEFRAG
:
5449 return btrfs_ioctl_defrag(file
, NULL
);
5450 case BTRFS_IOC_DEFRAG_RANGE
:
5451 return btrfs_ioctl_defrag(file
, argp
);
5452 case BTRFS_IOC_RESIZE
:
5453 return btrfs_ioctl_resize(file
, argp
);
5454 case BTRFS_IOC_ADD_DEV
:
5455 return btrfs_ioctl_add_dev(root
, argp
);
5456 case BTRFS_IOC_RM_DEV
:
5457 return btrfs_ioctl_rm_dev(file
, argp
);
5458 case BTRFS_IOC_FS_INFO
:
5459 return btrfs_ioctl_fs_info(root
, argp
);
5460 case BTRFS_IOC_DEV_INFO
:
5461 return btrfs_ioctl_dev_info(root
, argp
);
5462 case BTRFS_IOC_BALANCE
:
5463 return btrfs_ioctl_balance(file
, NULL
);
5464 case BTRFS_IOC_TRANS_START
:
5465 return btrfs_ioctl_trans_start(file
);
5466 case BTRFS_IOC_TRANS_END
:
5467 return btrfs_ioctl_trans_end(file
);
5468 case BTRFS_IOC_TREE_SEARCH
:
5469 return btrfs_ioctl_tree_search(file
, argp
);
5470 case BTRFS_IOC_TREE_SEARCH_V2
:
5471 return btrfs_ioctl_tree_search_v2(file
, argp
);
5472 case BTRFS_IOC_INO_LOOKUP
:
5473 return btrfs_ioctl_ino_lookup(file
, argp
);
5474 case BTRFS_IOC_INO_PATHS
:
5475 return btrfs_ioctl_ino_to_path(root
, argp
);
5476 case BTRFS_IOC_LOGICAL_INO
:
5477 return btrfs_ioctl_logical_to_ino(root
, argp
);
5478 case BTRFS_IOC_SPACE_INFO
:
5479 return btrfs_ioctl_space_info(root
, argp
);
5480 case BTRFS_IOC_SYNC
: {
5483 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
5486 ret
= btrfs_sync_fs(file_inode(file
)->i_sb
, 1);
5488 * The transaction thread may want to do more work,
5489 * namely it pokes the cleaner ktread that will start
5490 * processing uncleaned subvols.
5492 wake_up_process(root
->fs_info
->transaction_kthread
);
5495 case BTRFS_IOC_START_SYNC
:
5496 return btrfs_ioctl_start_sync(root
, argp
);
5497 case BTRFS_IOC_WAIT_SYNC
:
5498 return btrfs_ioctl_wait_sync(root
, argp
);
5499 case BTRFS_IOC_SCRUB
:
5500 return btrfs_ioctl_scrub(file
, argp
);
5501 case BTRFS_IOC_SCRUB_CANCEL
:
5502 return btrfs_ioctl_scrub_cancel(root
, argp
);
5503 case BTRFS_IOC_SCRUB_PROGRESS
:
5504 return btrfs_ioctl_scrub_progress(root
, argp
);
5505 case BTRFS_IOC_BALANCE_V2
:
5506 return btrfs_ioctl_balance(file
, argp
);
5507 case BTRFS_IOC_BALANCE_CTL
:
5508 return btrfs_ioctl_balance_ctl(root
, arg
);
5509 case BTRFS_IOC_BALANCE_PROGRESS
:
5510 return btrfs_ioctl_balance_progress(root
, argp
);
5511 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5512 return btrfs_ioctl_set_received_subvol(file
, argp
);
5514 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5515 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5517 case BTRFS_IOC_SEND
:
5518 return btrfs_ioctl_send(file
, argp
);
5519 case BTRFS_IOC_GET_DEV_STATS
:
5520 return btrfs_ioctl_get_dev_stats(root
, argp
);
5521 case BTRFS_IOC_QUOTA_CTL
:
5522 return btrfs_ioctl_quota_ctl(file
, argp
);
5523 case BTRFS_IOC_QGROUP_ASSIGN
:
5524 return btrfs_ioctl_qgroup_assign(file
, argp
);
5525 case BTRFS_IOC_QGROUP_CREATE
:
5526 return btrfs_ioctl_qgroup_create(file
, argp
);
5527 case BTRFS_IOC_QGROUP_LIMIT
:
5528 return btrfs_ioctl_qgroup_limit(file
, argp
);
5529 case BTRFS_IOC_QUOTA_RESCAN
:
5530 return btrfs_ioctl_quota_rescan(file
, argp
);
5531 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5532 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5533 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5534 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5535 case BTRFS_IOC_DEV_REPLACE
:
5536 return btrfs_ioctl_dev_replace(root
, argp
);
5537 case BTRFS_IOC_GET_FSLABEL
:
5538 return btrfs_ioctl_get_fslabel(file
, argp
);
5539 case BTRFS_IOC_SET_FSLABEL
:
5540 return btrfs_ioctl_set_fslabel(file
, argp
);
5541 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5542 return btrfs_ioctl_get_supported_features(argp
);
5543 case BTRFS_IOC_GET_FEATURES
:
5544 return btrfs_ioctl_get_features(file
, argp
);
5545 case BTRFS_IOC_SET_FEATURES
:
5546 return btrfs_ioctl_set_features(file
, argp
);