2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
63 #include "compression.h"
66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
71 struct btrfs_ioctl_timespec_32
{
74 } __attribute__ ((__packed__
));
76 struct btrfs_ioctl_received_subvol_args_32
{
77 char uuid
[BTRFS_UUID_SIZE
]; /* in */
78 __u64 stransid
; /* in */
79 __u64 rtransid
; /* out */
80 struct btrfs_ioctl_timespec_32 stime
; /* in */
81 struct btrfs_ioctl_timespec_32 rtime
; /* out */
83 __u64 reserved
[16]; /* in */
84 } __attribute__ ((__packed__
));
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
91 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
92 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
,
95 /* Mask out flags that are inappropriate for the given type of inode. */
96 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
100 else if (S_ISREG(mode
))
101 return flags
& ~FS_DIRSYNC_FL
;
103 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
107 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
109 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
111 unsigned int iflags
= 0;
113 if (flags
& BTRFS_INODE_SYNC
)
114 iflags
|= FS_SYNC_FL
;
115 if (flags
& BTRFS_INODE_IMMUTABLE
)
116 iflags
|= FS_IMMUTABLE_FL
;
117 if (flags
& BTRFS_INODE_APPEND
)
118 iflags
|= FS_APPEND_FL
;
119 if (flags
& BTRFS_INODE_NODUMP
)
120 iflags
|= FS_NODUMP_FL
;
121 if (flags
& BTRFS_INODE_NOATIME
)
122 iflags
|= FS_NOATIME_FL
;
123 if (flags
& BTRFS_INODE_DIRSYNC
)
124 iflags
|= FS_DIRSYNC_FL
;
125 if (flags
& BTRFS_INODE_NODATACOW
)
126 iflags
|= FS_NOCOW_FL
;
128 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
129 iflags
|= FS_COMPR_FL
;
130 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
131 iflags
|= FS_NOCOMP_FL
;
137 * Update inode->i_flags based on the btrfs internal flags.
139 void btrfs_update_iflags(struct inode
*inode
)
141 struct btrfs_inode
*ip
= BTRFS_I(inode
);
142 unsigned int new_fl
= 0;
144 if (ip
->flags
& BTRFS_INODE_SYNC
)
146 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
147 new_fl
|= S_IMMUTABLE
;
148 if (ip
->flags
& BTRFS_INODE_APPEND
)
150 if (ip
->flags
& BTRFS_INODE_NOATIME
)
152 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
155 set_mask_bits(&inode
->i_flags
,
156 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
,
161 * Inherit flags from the parent inode.
163 * Currently only the compression flags and the cow flags are inherited.
165 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
172 flags
= BTRFS_I(dir
)->flags
;
174 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
175 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
176 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
177 } else if (flags
& BTRFS_INODE_COMPRESS
) {
178 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
179 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
182 if (flags
& BTRFS_INODE_NODATACOW
) {
183 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
184 if (S_ISREG(inode
->i_mode
))
185 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
188 btrfs_update_iflags(inode
);
191 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
193 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
194 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
196 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
201 static int check_flags(unsigned int flags
)
203 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
204 FS_NOATIME_FL
| FS_NODUMP_FL
| \
205 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
206 FS_NOCOMP_FL
| FS_COMPR_FL
|
210 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
216 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
218 struct inode
*inode
= file_inode(file
);
219 struct btrfs_inode
*ip
= BTRFS_I(inode
);
220 struct btrfs_root
*root
= ip
->root
;
221 struct btrfs_trans_handle
*trans
;
222 unsigned int flags
, oldflags
;
225 unsigned int i_oldflags
;
228 if (!inode_owner_or_capable(inode
))
231 if (btrfs_root_readonly(root
))
234 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
237 ret
= check_flags(flags
);
241 ret
= mnt_want_write_file(file
);
247 ip_oldflags
= ip
->flags
;
248 i_oldflags
= inode
->i_flags
;
249 mode
= inode
->i_mode
;
251 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
252 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
253 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
254 if (!capable(CAP_LINUX_IMMUTABLE
)) {
260 if (flags
& FS_SYNC_FL
)
261 ip
->flags
|= BTRFS_INODE_SYNC
;
263 ip
->flags
&= ~BTRFS_INODE_SYNC
;
264 if (flags
& FS_IMMUTABLE_FL
)
265 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
267 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
268 if (flags
& FS_APPEND_FL
)
269 ip
->flags
|= BTRFS_INODE_APPEND
;
271 ip
->flags
&= ~BTRFS_INODE_APPEND
;
272 if (flags
& FS_NODUMP_FL
)
273 ip
->flags
|= BTRFS_INODE_NODUMP
;
275 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
276 if (flags
& FS_NOATIME_FL
)
277 ip
->flags
|= BTRFS_INODE_NOATIME
;
279 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
280 if (flags
& FS_DIRSYNC_FL
)
281 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
283 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
284 if (flags
& FS_NOCOW_FL
) {
287 * It's safe to turn csums off here, no extents exist.
288 * Otherwise we want the flag to reflect the real COW
289 * status of the file and will not set it.
291 if (inode
->i_size
== 0)
292 ip
->flags
|= BTRFS_INODE_NODATACOW
293 | BTRFS_INODE_NODATASUM
;
295 ip
->flags
|= BTRFS_INODE_NODATACOW
;
299 * Revert back under same assuptions as above
302 if (inode
->i_size
== 0)
303 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
304 | BTRFS_INODE_NODATASUM
);
306 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
311 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
312 * flag may be changed automatically if compression code won't make
315 if (flags
& FS_NOCOMP_FL
) {
316 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
317 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
319 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
320 if (ret
&& ret
!= -ENODATA
)
322 } else if (flags
& FS_COMPR_FL
) {
325 ip
->flags
|= BTRFS_INODE_COMPRESS
;
326 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
328 if (root
->fs_info
->compress_type
== BTRFS_COMPRESS_LZO
)
332 ret
= btrfs_set_prop(inode
, "btrfs.compression",
333 comp
, strlen(comp
), 0);
338 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
339 if (ret
&& ret
!= -ENODATA
)
341 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
344 trans
= btrfs_start_transaction(root
, 1);
346 ret
= PTR_ERR(trans
);
350 btrfs_update_iflags(inode
);
351 inode_inc_iversion(inode
);
352 inode
->i_ctime
= current_fs_time(inode
->i_sb
);
353 ret
= btrfs_update_inode(trans
, root
, inode
);
355 btrfs_end_transaction(trans
, root
);
358 ip
->flags
= ip_oldflags
;
359 inode
->i_flags
= i_oldflags
;
364 mnt_drop_write_file(file
);
368 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
370 struct inode
*inode
= file_inode(file
);
372 return put_user(inode
->i_generation
, arg
);
375 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
377 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
378 struct btrfs_device
*device
;
379 struct request_queue
*q
;
380 struct fstrim_range range
;
381 u64 minlen
= ULLONG_MAX
;
383 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
386 if (!capable(CAP_SYS_ADMIN
))
390 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
394 q
= bdev_get_queue(device
->bdev
);
395 if (blk_queue_discard(q
)) {
397 minlen
= min((u64
)q
->limits
.discard_granularity
,
405 if (copy_from_user(&range
, arg
, sizeof(range
)))
407 if (range
.start
> total_bytes
||
408 range
.len
< fs_info
->sb
->s_blocksize
)
411 range
.len
= min(range
.len
, total_bytes
- range
.start
);
412 range
.minlen
= max(range
.minlen
, minlen
);
413 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
417 if (copy_to_user(arg
, &range
, sizeof(range
)))
423 int btrfs_is_empty_uuid(u8
*uuid
)
427 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
434 static noinline
int create_subvol(struct inode
*dir
,
435 struct dentry
*dentry
,
436 char *name
, int namelen
,
438 struct btrfs_qgroup_inherit
*inherit
)
440 struct btrfs_trans_handle
*trans
;
441 struct btrfs_key key
;
442 struct btrfs_root_item root_item
;
443 struct btrfs_inode_item
*inode_item
;
444 struct extent_buffer
*leaf
;
445 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
446 struct btrfs_root
*new_root
;
447 struct btrfs_block_rsv block_rsv
;
448 struct timespec cur_time
= current_fs_time(dir
->i_sb
);
453 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
458 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
463 * Don't create subvolume whose level is not zero. Or qgroup will be
464 * screwed up since it assume subvolme qgroup's level to be 0.
466 if (btrfs_qgroup_level(objectid
))
469 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
471 * The same as the snapshot creation, please see the comment
472 * of create_snapshot().
474 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
475 8, &qgroup_reserved
, false);
479 trans
= btrfs_start_transaction(root
, 0);
481 ret
= PTR_ERR(trans
);
482 btrfs_subvolume_release_metadata(root
, &block_rsv
,
486 trans
->block_rsv
= &block_rsv
;
487 trans
->bytes_reserved
= block_rsv
.size
;
489 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
493 leaf
= btrfs_alloc_tree_block(trans
, root
, 0, objectid
, NULL
, 0, 0, 0);
499 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
500 btrfs_set_header_bytenr(leaf
, leaf
->start
);
501 btrfs_set_header_generation(leaf
, trans
->transid
);
502 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
503 btrfs_set_header_owner(leaf
, objectid
);
505 write_extent_buffer(leaf
, root
->fs_info
->fsid
, btrfs_header_fsid(),
507 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
508 btrfs_header_chunk_tree_uuid(leaf
),
510 btrfs_mark_buffer_dirty(leaf
);
512 memset(&root_item
, 0, sizeof(root_item
));
514 inode_item
= &root_item
.inode
;
515 btrfs_set_stack_inode_generation(inode_item
, 1);
516 btrfs_set_stack_inode_size(inode_item
, 3);
517 btrfs_set_stack_inode_nlink(inode_item
, 1);
518 btrfs_set_stack_inode_nbytes(inode_item
, root
->nodesize
);
519 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
521 btrfs_set_root_flags(&root_item
, 0);
522 btrfs_set_root_limit(&root_item
, 0);
523 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
525 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
526 btrfs_set_root_generation(&root_item
, trans
->transid
);
527 btrfs_set_root_level(&root_item
, 0);
528 btrfs_set_root_refs(&root_item
, 1);
529 btrfs_set_root_used(&root_item
, leaf
->len
);
530 btrfs_set_root_last_snapshot(&root_item
, 0);
532 btrfs_set_root_generation_v2(&root_item
,
533 btrfs_root_generation(&root_item
));
534 uuid_le_gen(&new_uuid
);
535 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
536 btrfs_set_stack_timespec_sec(&root_item
.otime
, cur_time
.tv_sec
);
537 btrfs_set_stack_timespec_nsec(&root_item
.otime
, cur_time
.tv_nsec
);
538 root_item
.ctime
= root_item
.otime
;
539 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
540 btrfs_set_root_otransid(&root_item
, trans
->transid
);
542 btrfs_tree_unlock(leaf
);
543 free_extent_buffer(leaf
);
546 btrfs_set_root_dirid(&root_item
, new_dirid
);
548 key
.objectid
= objectid
;
550 key
.type
= BTRFS_ROOT_ITEM_KEY
;
551 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
556 key
.offset
= (u64
)-1;
557 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
558 if (IS_ERR(new_root
)) {
559 ret
= PTR_ERR(new_root
);
560 btrfs_abort_transaction(trans
, root
, ret
);
564 btrfs_record_root_in_trans(trans
, new_root
);
566 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
568 /* We potentially lose an unused inode item here */
569 btrfs_abort_transaction(trans
, root
, ret
);
573 mutex_lock(&new_root
->objectid_mutex
);
574 new_root
->highest_objectid
= new_dirid
;
575 mutex_unlock(&new_root
->objectid_mutex
);
578 * insert the directory item
580 ret
= btrfs_set_inode_index(dir
, &index
);
582 btrfs_abort_transaction(trans
, root
, ret
);
586 ret
= btrfs_insert_dir_item(trans
, root
,
587 name
, namelen
, dir
, &key
,
588 BTRFS_FT_DIR
, index
);
590 btrfs_abort_transaction(trans
, root
, ret
);
594 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
595 ret
= btrfs_update_inode(trans
, root
, dir
);
598 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
599 objectid
, root
->root_key
.objectid
,
600 btrfs_ino(dir
), index
, name
, namelen
);
603 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
604 root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
607 btrfs_abort_transaction(trans
, root
, ret
);
610 trans
->block_rsv
= NULL
;
611 trans
->bytes_reserved
= 0;
612 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
615 *async_transid
= trans
->transid
;
616 err
= btrfs_commit_transaction_async(trans
, root
, 1);
618 err
= btrfs_commit_transaction(trans
, root
);
620 err
= btrfs_commit_transaction(trans
, root
);
626 inode
= btrfs_lookup_dentry(dir
, dentry
);
628 return PTR_ERR(inode
);
629 d_instantiate(dentry
, inode
);
634 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root
*root
)
640 prepare_to_wait(&root
->subv_writers
->wait
, &wait
,
641 TASK_UNINTERRUPTIBLE
);
643 writers
= percpu_counter_sum(&root
->subv_writers
->counter
);
647 finish_wait(&root
->subv_writers
->wait
, &wait
);
651 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
652 struct dentry
*dentry
, char *name
, int namelen
,
653 u64
*async_transid
, bool readonly
,
654 struct btrfs_qgroup_inherit
*inherit
)
657 struct btrfs_pending_snapshot
*pending_snapshot
;
658 struct btrfs_trans_handle
*trans
;
661 if (!test_bit(BTRFS_ROOT_REF_COWS
, &root
->state
))
664 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
665 if (!pending_snapshot
)
668 pending_snapshot
->root_item
= kzalloc(sizeof(struct btrfs_root_item
),
670 pending_snapshot
->path
= btrfs_alloc_path();
671 if (!pending_snapshot
->root_item
|| !pending_snapshot
->path
) {
676 atomic_inc(&root
->will_be_snapshoted
);
677 smp_mb__after_atomic();
678 btrfs_wait_for_no_snapshoting_writes(root
);
680 ret
= btrfs_start_delalloc_inodes(root
, 0);
684 btrfs_wait_ordered_extents(root
, -1);
686 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
687 BTRFS_BLOCK_RSV_TEMP
);
689 * 1 - parent dir inode
692 * 2 - root ref/backref
693 * 1 - root of snapshot
696 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
697 &pending_snapshot
->block_rsv
, 8,
698 &pending_snapshot
->qgroup_reserved
,
703 pending_snapshot
->dentry
= dentry
;
704 pending_snapshot
->root
= root
;
705 pending_snapshot
->readonly
= readonly
;
706 pending_snapshot
->dir
= dir
;
707 pending_snapshot
->inherit
= inherit
;
709 trans
= btrfs_start_transaction(root
, 0);
711 ret
= PTR_ERR(trans
);
715 spin_lock(&root
->fs_info
->trans_lock
);
716 list_add(&pending_snapshot
->list
,
717 &trans
->transaction
->pending_snapshots
);
718 spin_unlock(&root
->fs_info
->trans_lock
);
720 *async_transid
= trans
->transid
;
721 ret
= btrfs_commit_transaction_async(trans
,
722 root
->fs_info
->extent_root
, 1);
724 ret
= btrfs_commit_transaction(trans
, root
);
726 ret
= btrfs_commit_transaction(trans
,
727 root
->fs_info
->extent_root
);
732 ret
= pending_snapshot
->error
;
736 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
740 inode
= btrfs_lookup_dentry(d_inode(dentry
->d_parent
), dentry
);
742 ret
= PTR_ERR(inode
);
746 d_instantiate(dentry
, inode
);
749 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
750 &pending_snapshot
->block_rsv
,
751 pending_snapshot
->qgroup_reserved
);
753 if (atomic_dec_and_test(&root
->will_be_snapshoted
))
754 wake_up_atomic_t(&root
->will_be_snapshoted
);
756 kfree(pending_snapshot
->root_item
);
757 btrfs_free_path(pending_snapshot
->path
);
758 kfree(pending_snapshot
);
763 /* copy of may_delete in fs/namei.c()
764 * Check whether we can remove a link victim from directory dir, check
765 * whether the type of victim is right.
766 * 1. We can't do it if dir is read-only (done in permission())
767 * 2. We should have write and exec permissions on dir
768 * 3. We can't remove anything from append-only dir
769 * 4. We can't do anything with immutable dir (done in permission())
770 * 5. If the sticky bit on dir is set we should either
771 * a. be owner of dir, or
772 * b. be owner of victim, or
773 * c. have CAP_FOWNER capability
774 * 6. If the victim is append-only or immutable we can't do antyhing with
775 * links pointing to it.
776 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
777 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
778 * 9. We can't remove a root or mountpoint.
779 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
780 * nfs_async_unlink().
783 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
787 if (d_really_is_negative(victim
))
790 BUG_ON(d_inode(victim
->d_parent
) != dir
);
791 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
793 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
798 if (check_sticky(dir
, d_inode(victim
)) || IS_APPEND(d_inode(victim
)) ||
799 IS_IMMUTABLE(d_inode(victim
)) || IS_SWAPFILE(d_inode(victim
)))
802 if (!d_is_dir(victim
))
806 } else if (d_is_dir(victim
))
810 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
815 /* copy of may_create in fs/namei.c() */
816 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
818 if (d_really_is_positive(child
))
822 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
826 * Create a new subvolume below @parent. This is largely modeled after
827 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
828 * inside this filesystem so it's quite a bit simpler.
830 static noinline
int btrfs_mksubvol(struct path
*parent
,
831 char *name
, int namelen
,
832 struct btrfs_root
*snap_src
,
833 u64
*async_transid
, bool readonly
,
834 struct btrfs_qgroup_inherit
*inherit
)
836 struct inode
*dir
= d_inode(parent
->dentry
);
837 struct dentry
*dentry
;
840 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
844 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
845 error
= PTR_ERR(dentry
);
849 error
= btrfs_may_create(dir
, dentry
);
854 * even if this name doesn't exist, we may get hash collisions.
855 * check for them now when we can safely fail
857 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
863 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
865 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
869 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
870 async_transid
, readonly
, inherit
);
872 error
= create_subvol(dir
, dentry
, name
, namelen
,
873 async_transid
, inherit
);
876 fsnotify_mkdir(dir
, dentry
);
878 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
887 * When we're defragging a range, we don't want to kick it off again
888 * if it is really just waiting for delalloc to send it down.
889 * If we find a nice big extent or delalloc range for the bytes in the
890 * file you want to defrag, we return 0 to let you know to skip this
893 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, u32 thresh
)
895 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
896 struct extent_map
*em
= NULL
;
897 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
900 read_lock(&em_tree
->lock
);
901 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_SIZE
);
902 read_unlock(&em_tree
->lock
);
905 end
= extent_map_end(em
);
907 if (end
- offset
> thresh
)
910 /* if we already have a nice delalloc here, just stop */
912 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
913 thresh
, EXTENT_DELALLOC
, 1);
920 * helper function to walk through a file and find extents
921 * newer than a specific transid, and smaller than thresh.
923 * This is used by the defragging code to find new and small
926 static int find_new_extents(struct btrfs_root
*root
,
927 struct inode
*inode
, u64 newer_than
,
928 u64
*off
, u32 thresh
)
930 struct btrfs_path
*path
;
931 struct btrfs_key min_key
;
932 struct extent_buffer
*leaf
;
933 struct btrfs_file_extent_item
*extent
;
936 u64 ino
= btrfs_ino(inode
);
938 path
= btrfs_alloc_path();
942 min_key
.objectid
= ino
;
943 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
944 min_key
.offset
= *off
;
947 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
951 if (min_key
.objectid
!= ino
)
953 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
956 leaf
= path
->nodes
[0];
957 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
958 struct btrfs_file_extent_item
);
960 type
= btrfs_file_extent_type(leaf
, extent
);
961 if (type
== BTRFS_FILE_EXTENT_REG
&&
962 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
963 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
964 *off
= min_key
.offset
;
965 btrfs_free_path(path
);
970 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
971 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
975 if (min_key
.offset
== (u64
)-1)
979 btrfs_release_path(path
);
982 btrfs_free_path(path
);
986 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
988 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
989 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
990 struct extent_map
*em
;
994 * hopefully we have this extent in the tree already, try without
995 * the full extent lock
997 read_lock(&em_tree
->lock
);
998 em
= lookup_extent_mapping(em_tree
, start
, len
);
999 read_unlock(&em_tree
->lock
);
1002 struct extent_state
*cached
= NULL
;
1003 u64 end
= start
+ len
- 1;
1005 /* get the big lock and read metadata off disk */
1006 lock_extent_bits(io_tree
, start
, end
, &cached
);
1007 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
1008 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1017 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1019 struct extent_map
*next
;
1022 /* this is the last extent */
1023 if (em
->start
+ em
->len
>= i_size_read(inode
))
1026 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1027 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
1029 else if ((em
->block_start
+ em
->block_len
== next
->block_start
) &&
1030 (em
->block_len
> SZ_128K
&& next
->block_len
> SZ_128K
))
1033 free_extent_map(next
);
1037 static int should_defrag_range(struct inode
*inode
, u64 start
, u32 thresh
,
1038 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1041 struct extent_map
*em
;
1043 bool next_mergeable
= true;
1044 bool prev_mergeable
= true;
1047 * make sure that once we start defragging an extent, we keep on
1050 if (start
< *defrag_end
)
1055 em
= defrag_lookup_extent(inode
, start
);
1059 /* this will cover holes, and inline extents */
1060 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1066 prev_mergeable
= false;
1068 next_mergeable
= defrag_check_next_extent(inode
, em
);
1070 * we hit a real extent, if it is big or the next extent is not a
1071 * real extent, don't bother defragging it
1073 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1074 (em
->len
>= thresh
|| (!next_mergeable
&& !prev_mergeable
)))
1078 * last_len ends up being a counter of how many bytes we've defragged.
1079 * every time we choose not to defrag an extent, we reset *last_len
1080 * so that the next tiny extent will force a defrag.
1082 * The end result of this is that tiny extents before a single big
1083 * extent will force at least part of that big extent to be defragged.
1086 *defrag_end
= extent_map_end(em
);
1089 *skip
= extent_map_end(em
);
1093 free_extent_map(em
);
1098 * it doesn't do much good to defrag one or two pages
1099 * at a time. This pulls in a nice chunk of pages
1100 * to COW and defrag.
1102 * It also makes sure the delalloc code has enough
1103 * dirty data to avoid making new small extents as part
1106 * It's a good idea to start RA on this range
1107 * before calling this.
1109 static int cluster_pages_for_defrag(struct inode
*inode
,
1110 struct page
**pages
,
1111 unsigned long start_index
,
1112 unsigned long num_pages
)
1114 unsigned long file_end
;
1115 u64 isize
= i_size_read(inode
);
1122 struct btrfs_ordered_extent
*ordered
;
1123 struct extent_state
*cached_state
= NULL
;
1124 struct extent_io_tree
*tree
;
1125 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1127 file_end
= (isize
- 1) >> PAGE_SHIFT
;
1128 if (!isize
|| start_index
> file_end
)
1131 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1133 ret
= btrfs_delalloc_reserve_space(inode
,
1134 start_index
<< PAGE_SHIFT
,
1135 page_cnt
<< PAGE_SHIFT
);
1139 tree
= &BTRFS_I(inode
)->io_tree
;
1141 /* step one, lock all the pages */
1142 for (i
= 0; i
< page_cnt
; i
++) {
1145 page
= find_or_create_page(inode
->i_mapping
,
1146 start_index
+ i
, mask
);
1150 page_start
= page_offset(page
);
1151 page_end
= page_start
+ PAGE_SIZE
- 1;
1153 lock_extent_bits(tree
, page_start
, page_end
,
1155 ordered
= btrfs_lookup_ordered_extent(inode
,
1157 unlock_extent_cached(tree
, page_start
, page_end
,
1158 &cached_state
, GFP_NOFS
);
1163 btrfs_start_ordered_extent(inode
, ordered
, 1);
1164 btrfs_put_ordered_extent(ordered
);
1167 * we unlocked the page above, so we need check if
1168 * it was released or not.
1170 if (page
->mapping
!= inode
->i_mapping
) {
1177 if (!PageUptodate(page
)) {
1178 btrfs_readpage(NULL
, page
);
1180 if (!PageUptodate(page
)) {
1188 if (page
->mapping
!= inode
->i_mapping
) {
1200 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1204 * so now we have a nice long stream of locked
1205 * and up to date pages, lets wait on them
1207 for (i
= 0; i
< i_done
; i
++)
1208 wait_on_page_writeback(pages
[i
]);
1210 page_start
= page_offset(pages
[0]);
1211 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_SIZE
;
1213 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1214 page_start
, page_end
- 1, &cached_state
);
1215 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1216 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1217 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1218 &cached_state
, GFP_NOFS
);
1220 if (i_done
!= page_cnt
) {
1221 spin_lock(&BTRFS_I(inode
)->lock
);
1222 BTRFS_I(inode
)->outstanding_extents
++;
1223 spin_unlock(&BTRFS_I(inode
)->lock
);
1224 btrfs_delalloc_release_space(inode
,
1225 start_index
<< PAGE_SHIFT
,
1226 (page_cnt
- i_done
) << PAGE_SHIFT
);
1230 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1231 &cached_state
, GFP_NOFS
);
1233 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1234 page_start
, page_end
- 1, &cached_state
,
1237 for (i
= 0; i
< i_done
; i
++) {
1238 clear_page_dirty_for_io(pages
[i
]);
1239 ClearPageChecked(pages
[i
]);
1240 set_page_extent_mapped(pages
[i
]);
1241 set_page_dirty(pages
[i
]);
1242 unlock_page(pages
[i
]);
1247 for (i
= 0; i
< i_done
; i
++) {
1248 unlock_page(pages
[i
]);
1251 btrfs_delalloc_release_space(inode
,
1252 start_index
<< PAGE_SHIFT
,
1253 page_cnt
<< PAGE_SHIFT
);
1258 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1259 struct btrfs_ioctl_defrag_range_args
*range
,
1260 u64 newer_than
, unsigned long max_to_defrag
)
1262 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1263 struct file_ra_state
*ra
= NULL
;
1264 unsigned long last_index
;
1265 u64 isize
= i_size_read(inode
);
1269 u64 newer_off
= range
->start
;
1271 unsigned long ra_index
= 0;
1273 int defrag_count
= 0;
1274 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1275 u32 extent_thresh
= range
->extent_thresh
;
1276 unsigned long max_cluster
= SZ_256K
>> PAGE_SHIFT
;
1277 unsigned long cluster
= max_cluster
;
1278 u64 new_align
= ~((u64
)SZ_128K
- 1);
1279 struct page
**pages
= NULL
;
1284 if (range
->start
>= isize
)
1287 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1288 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1290 if (range
->compress_type
)
1291 compress_type
= range
->compress_type
;
1294 if (extent_thresh
== 0)
1295 extent_thresh
= SZ_256K
;
1298 * if we were not given a file, allocate a readahead
1302 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1305 file_ra_state_init(ra
, inode
->i_mapping
);
1310 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1317 /* find the last page to defrag */
1318 if (range
->start
+ range
->len
> range
->start
) {
1319 last_index
= min_t(u64
, isize
- 1,
1320 range
->start
+ range
->len
- 1) >> PAGE_SHIFT
;
1322 last_index
= (isize
- 1) >> PAGE_SHIFT
;
1326 ret
= find_new_extents(root
, inode
, newer_than
,
1327 &newer_off
, SZ_64K
);
1329 range
->start
= newer_off
;
1331 * we always align our defrag to help keep
1332 * the extents in the file evenly spaced
1334 i
= (newer_off
& new_align
) >> PAGE_SHIFT
;
1338 i
= range
->start
>> PAGE_SHIFT
;
1341 max_to_defrag
= last_index
- i
+ 1;
1344 * make writeback starts from i, so the defrag range can be
1345 * written sequentially.
1347 if (i
< inode
->i_mapping
->writeback_index
)
1348 inode
->i_mapping
->writeback_index
= i
;
1350 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1351 (i
< DIV_ROUND_UP(i_size_read(inode
), PAGE_SIZE
))) {
1353 * make sure we stop running if someone unmounts
1356 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1359 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1360 btrfs_debug(root
->fs_info
, "defrag_file cancelled");
1365 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_SHIFT
,
1366 extent_thresh
, &last_len
, &skip
,
1367 &defrag_end
, range
->flags
&
1368 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1371 * the should_defrag function tells us how much to skip
1372 * bump our counter by the suggested amount
1374 next
= DIV_ROUND_UP(skip
, PAGE_SIZE
);
1375 i
= max(i
+ 1, next
);
1380 cluster
= (PAGE_ALIGN(defrag_end
) >>
1382 cluster
= min(cluster
, max_cluster
);
1384 cluster
= max_cluster
;
1387 if (i
+ cluster
> ra_index
) {
1388 ra_index
= max(i
, ra_index
);
1389 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1391 ra_index
+= cluster
;
1395 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1396 BTRFS_I(inode
)->force_compress
= compress_type
;
1397 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1399 inode_unlock(inode
);
1403 defrag_count
+= ret
;
1404 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1405 inode_unlock(inode
);
1408 if (newer_off
== (u64
)-1)
1414 newer_off
= max(newer_off
+ 1,
1415 (u64
)i
<< PAGE_SHIFT
);
1417 ret
= find_new_extents(root
, inode
, newer_than
,
1418 &newer_off
, SZ_64K
);
1420 range
->start
= newer_off
;
1421 i
= (newer_off
& new_align
) >> PAGE_SHIFT
;
1428 last_len
+= ret
<< PAGE_SHIFT
;
1436 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1437 filemap_flush(inode
->i_mapping
);
1438 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1439 &BTRFS_I(inode
)->runtime_flags
))
1440 filemap_flush(inode
->i_mapping
);
1443 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1444 /* the filemap_flush will queue IO into the worker threads, but
1445 * we have to make sure the IO is actually started and that
1446 * ordered extents get created before we return
1448 atomic_inc(&root
->fs_info
->async_submit_draining
);
1449 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1450 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1451 wait_event(root
->fs_info
->async_submit_wait
,
1452 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1453 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1455 atomic_dec(&root
->fs_info
->async_submit_draining
);
1458 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1459 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1465 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1467 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1468 inode_unlock(inode
);
1476 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1482 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1483 struct btrfs_ioctl_vol_args
*vol_args
;
1484 struct btrfs_trans_handle
*trans
;
1485 struct btrfs_device
*device
= NULL
;
1488 char *devstr
= NULL
;
1492 if (!capable(CAP_SYS_ADMIN
))
1495 ret
= mnt_want_write_file(file
);
1499 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1501 mnt_drop_write_file(file
);
1502 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1505 mutex_lock(&root
->fs_info
->volume_mutex
);
1506 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1507 if (IS_ERR(vol_args
)) {
1508 ret
= PTR_ERR(vol_args
);
1512 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1514 sizestr
= vol_args
->name
;
1515 devstr
= strchr(sizestr
, ':');
1517 sizestr
= devstr
+ 1;
1519 devstr
= vol_args
->name
;
1520 ret
= kstrtoull(devstr
, 10, &devid
);
1527 btrfs_info(root
->fs_info
, "resizing devid %llu", devid
);
1530 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1532 btrfs_info(root
->fs_info
, "resizer unable to find device %llu",
1538 if (!device
->writeable
) {
1539 btrfs_info(root
->fs_info
,
1540 "resizer unable to apply on readonly device %llu",
1546 if (!strcmp(sizestr
, "max"))
1547 new_size
= device
->bdev
->bd_inode
->i_size
;
1549 if (sizestr
[0] == '-') {
1552 } else if (sizestr
[0] == '+') {
1556 new_size
= memparse(sizestr
, &retptr
);
1557 if (*retptr
!= '\0' || new_size
== 0) {
1563 if (device
->is_tgtdev_for_dev_replace
) {
1568 old_size
= btrfs_device_get_total_bytes(device
);
1571 if (new_size
> old_size
) {
1575 new_size
= old_size
- new_size
;
1576 } else if (mod
> 0) {
1577 if (new_size
> ULLONG_MAX
- old_size
) {
1581 new_size
= old_size
+ new_size
;
1584 if (new_size
< SZ_256M
) {
1588 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1593 new_size
= div_u64(new_size
, root
->sectorsize
);
1594 new_size
*= root
->sectorsize
;
1596 btrfs_info_in_rcu(root
->fs_info
, "new size for %s is %llu",
1597 rcu_str_deref(device
->name
), new_size
);
1599 if (new_size
> old_size
) {
1600 trans
= btrfs_start_transaction(root
, 0);
1601 if (IS_ERR(trans
)) {
1602 ret
= PTR_ERR(trans
);
1605 ret
= btrfs_grow_device(trans
, device
, new_size
);
1606 btrfs_commit_transaction(trans
, root
);
1607 } else if (new_size
< old_size
) {
1608 ret
= btrfs_shrink_device(device
, new_size
);
1609 } /* equal, nothing need to do */
1614 mutex_unlock(&root
->fs_info
->volume_mutex
);
1615 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1616 mnt_drop_write_file(file
);
1620 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1621 char *name
, unsigned long fd
, int subvol
,
1622 u64
*transid
, bool readonly
,
1623 struct btrfs_qgroup_inherit
*inherit
)
1628 ret
= mnt_want_write_file(file
);
1632 namelen
= strlen(name
);
1633 if (strchr(name
, '/')) {
1635 goto out_drop_write
;
1638 if (name
[0] == '.' &&
1639 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1641 goto out_drop_write
;
1645 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1646 NULL
, transid
, readonly
, inherit
);
1648 struct fd src
= fdget(fd
);
1649 struct inode
*src_inode
;
1652 goto out_drop_write
;
1655 src_inode
= file_inode(src
.file
);
1656 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1657 btrfs_info(BTRFS_I(file_inode(file
))->root
->fs_info
,
1658 "Snapshot src from another FS");
1660 } else if (!inode_owner_or_capable(src_inode
)) {
1662 * Subvolume creation is not restricted, but snapshots
1663 * are limited to own subvolumes only
1667 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1668 BTRFS_I(src_inode
)->root
,
1669 transid
, readonly
, inherit
);
1674 mnt_drop_write_file(file
);
1679 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1680 void __user
*arg
, int subvol
)
1682 struct btrfs_ioctl_vol_args
*vol_args
;
1685 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1686 if (IS_ERR(vol_args
))
1687 return PTR_ERR(vol_args
);
1688 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1690 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1691 vol_args
->fd
, subvol
,
1698 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1699 void __user
*arg
, int subvol
)
1701 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1705 bool readonly
= false;
1706 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1708 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1709 if (IS_ERR(vol_args
))
1710 return PTR_ERR(vol_args
);
1711 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1713 if (vol_args
->flags
&
1714 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1715 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1720 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1722 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1724 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1725 if (vol_args
->size
> PAGE_SIZE
) {
1729 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1730 if (IS_ERR(inherit
)) {
1731 ret
= PTR_ERR(inherit
);
1736 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1737 vol_args
->fd
, subvol
, ptr
,
1742 if (ptr
&& copy_to_user(arg
+
1743 offsetof(struct btrfs_ioctl_vol_args_v2
,
1755 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1758 struct inode
*inode
= file_inode(file
);
1759 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1763 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1766 down_read(&root
->fs_info
->subvol_sem
);
1767 if (btrfs_root_readonly(root
))
1768 flags
|= BTRFS_SUBVOL_RDONLY
;
1769 up_read(&root
->fs_info
->subvol_sem
);
1771 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1777 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1780 struct inode
*inode
= file_inode(file
);
1781 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1782 struct btrfs_trans_handle
*trans
;
1787 if (!inode_owner_or_capable(inode
))
1790 ret
= mnt_want_write_file(file
);
1794 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1796 goto out_drop_write
;
1799 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1801 goto out_drop_write
;
1804 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1806 goto out_drop_write
;
1809 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1811 goto out_drop_write
;
1814 down_write(&root
->fs_info
->subvol_sem
);
1817 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1820 root_flags
= btrfs_root_flags(&root
->root_item
);
1821 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1822 btrfs_set_root_flags(&root
->root_item
,
1823 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1826 * Block RO -> RW transition if this subvolume is involved in
1829 spin_lock(&root
->root_item_lock
);
1830 if (root
->send_in_progress
== 0) {
1831 btrfs_set_root_flags(&root
->root_item
,
1832 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1833 spin_unlock(&root
->root_item_lock
);
1835 spin_unlock(&root
->root_item_lock
);
1836 btrfs_warn(root
->fs_info
,
1837 "Attempt to set subvolume %llu read-write during send",
1838 root
->root_key
.objectid
);
1844 trans
= btrfs_start_transaction(root
, 1);
1845 if (IS_ERR(trans
)) {
1846 ret
= PTR_ERR(trans
);
1850 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1851 &root
->root_key
, &root
->root_item
);
1853 btrfs_commit_transaction(trans
, root
);
1856 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1858 up_write(&root
->fs_info
->subvol_sem
);
1860 mnt_drop_write_file(file
);
1866 * helper to check if the subvolume references other subvolumes
1868 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1870 struct btrfs_path
*path
;
1871 struct btrfs_dir_item
*di
;
1872 struct btrfs_key key
;
1876 path
= btrfs_alloc_path();
1880 /* Make sure this root isn't set as the default subvol */
1881 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
1882 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
, path
,
1883 dir_id
, "default", 7, 0);
1884 if (di
&& !IS_ERR(di
)) {
1885 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1886 if (key
.objectid
== root
->root_key
.objectid
) {
1888 btrfs_err(root
->fs_info
, "deleting default subvolume "
1889 "%llu is not allowed", key
.objectid
);
1892 btrfs_release_path(path
);
1895 key
.objectid
= root
->root_key
.objectid
;
1896 key
.type
= BTRFS_ROOT_REF_KEY
;
1897 key
.offset
= (u64
)-1;
1899 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1906 if (path
->slots
[0] > 0) {
1908 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1909 if (key
.objectid
== root
->root_key
.objectid
&&
1910 key
.type
== BTRFS_ROOT_REF_KEY
)
1914 btrfs_free_path(path
);
1918 static noinline
int key_in_sk(struct btrfs_key
*key
,
1919 struct btrfs_ioctl_search_key
*sk
)
1921 struct btrfs_key test
;
1924 test
.objectid
= sk
->min_objectid
;
1925 test
.type
= sk
->min_type
;
1926 test
.offset
= sk
->min_offset
;
1928 ret
= btrfs_comp_cpu_keys(key
, &test
);
1932 test
.objectid
= sk
->max_objectid
;
1933 test
.type
= sk
->max_type
;
1934 test
.offset
= sk
->max_offset
;
1936 ret
= btrfs_comp_cpu_keys(key
, &test
);
1942 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1943 struct btrfs_path
*path
,
1944 struct btrfs_key
*key
,
1945 struct btrfs_ioctl_search_key
*sk
,
1948 unsigned long *sk_offset
,
1952 struct extent_buffer
*leaf
;
1953 struct btrfs_ioctl_search_header sh
;
1954 struct btrfs_key test
;
1955 unsigned long item_off
;
1956 unsigned long item_len
;
1962 leaf
= path
->nodes
[0];
1963 slot
= path
->slots
[0];
1964 nritems
= btrfs_header_nritems(leaf
);
1966 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1970 found_transid
= btrfs_header_generation(leaf
);
1972 for (i
= slot
; i
< nritems
; i
++) {
1973 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1974 item_len
= btrfs_item_size_nr(leaf
, i
);
1976 btrfs_item_key_to_cpu(leaf
, key
, i
);
1977 if (!key_in_sk(key
, sk
))
1980 if (sizeof(sh
) + item_len
> *buf_size
) {
1987 * return one empty item back for v1, which does not
1991 *buf_size
= sizeof(sh
) + item_len
;
1996 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
2001 sh
.objectid
= key
->objectid
;
2002 sh
.offset
= key
->offset
;
2003 sh
.type
= key
->type
;
2005 sh
.transid
= found_transid
;
2007 /* copy search result header */
2008 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2013 *sk_offset
+= sizeof(sh
);
2016 char __user
*up
= ubuf
+ *sk_offset
;
2018 if (read_extent_buffer_to_user(leaf
, up
,
2019 item_off
, item_len
)) {
2024 *sk_offset
+= item_len
;
2028 if (ret
) /* -EOVERFLOW from above */
2031 if (*num_found
>= sk
->nr_items
) {
2038 test
.objectid
= sk
->max_objectid
;
2039 test
.type
= sk
->max_type
;
2040 test
.offset
= sk
->max_offset
;
2041 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
2043 else if (key
->offset
< (u64
)-1)
2045 else if (key
->type
< (u8
)-1) {
2048 } else if (key
->objectid
< (u64
)-1) {
2056 * 0: all items from this leaf copied, continue with next
2057 * 1: * more items can be copied, but unused buffer is too small
2058 * * all items were found
2059 * Either way, it will stops the loop which iterates to the next
2061 * -EOVERFLOW: item was to large for buffer
2062 * -EFAULT: could not copy extent buffer back to userspace
2067 static noinline
int search_ioctl(struct inode
*inode
,
2068 struct btrfs_ioctl_search_key
*sk
,
2072 struct btrfs_root
*root
;
2073 struct btrfs_key key
;
2074 struct btrfs_path
*path
;
2075 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
2078 unsigned long sk_offset
= 0;
2080 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2081 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2085 path
= btrfs_alloc_path();
2089 if (sk
->tree_id
== 0) {
2090 /* search the root of the inode that was passed */
2091 root
= BTRFS_I(inode
)->root
;
2093 key
.objectid
= sk
->tree_id
;
2094 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2095 key
.offset
= (u64
)-1;
2096 root
= btrfs_read_fs_root_no_name(info
, &key
);
2098 btrfs_free_path(path
);
2103 key
.objectid
= sk
->min_objectid
;
2104 key
.type
= sk
->min_type
;
2105 key
.offset
= sk
->min_offset
;
2108 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2114 ret
= copy_to_sk(root
, path
, &key
, sk
, buf_size
, ubuf
,
2115 &sk_offset
, &num_found
);
2116 btrfs_release_path(path
);
2124 sk
->nr_items
= num_found
;
2125 btrfs_free_path(path
);
2129 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2132 struct btrfs_ioctl_search_args __user
*uargs
;
2133 struct btrfs_ioctl_search_key sk
;
2134 struct inode
*inode
;
2138 if (!capable(CAP_SYS_ADMIN
))
2141 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2143 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2146 buf_size
= sizeof(uargs
->buf
);
2148 inode
= file_inode(file
);
2149 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2152 * In the origin implementation an overflow is handled by returning a
2153 * search header with a len of zero, so reset ret.
2155 if (ret
== -EOVERFLOW
)
2158 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2163 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2166 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2167 struct btrfs_ioctl_search_args_v2 args
;
2168 struct inode
*inode
;
2171 const size_t buf_limit
= SZ_16M
;
2173 if (!capable(CAP_SYS_ADMIN
))
2176 /* copy search header and buffer size */
2177 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2178 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2181 buf_size
= args
.buf_size
;
2183 if (buf_size
< sizeof(struct btrfs_ioctl_search_header
))
2186 /* limit result size to 16MB */
2187 if (buf_size
> buf_limit
)
2188 buf_size
= buf_limit
;
2190 inode
= file_inode(file
);
2191 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2192 (char *)(&uarg
->buf
[0]));
2193 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2195 else if (ret
== -EOVERFLOW
&&
2196 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2203 * Search INODE_REFs to identify path name of 'dirid' directory
2204 * in a 'tree_id' tree. and sets path name to 'name'.
2206 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2207 u64 tree_id
, u64 dirid
, char *name
)
2209 struct btrfs_root
*root
;
2210 struct btrfs_key key
;
2216 struct btrfs_inode_ref
*iref
;
2217 struct extent_buffer
*l
;
2218 struct btrfs_path
*path
;
2220 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2225 path
= btrfs_alloc_path();
2229 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2231 key
.objectid
= tree_id
;
2232 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2233 key
.offset
= (u64
)-1;
2234 root
= btrfs_read_fs_root_no_name(info
, &key
);
2236 btrfs_err(info
, "could not find root %llu", tree_id
);
2241 key
.objectid
= dirid
;
2242 key
.type
= BTRFS_INODE_REF_KEY
;
2243 key
.offset
= (u64
)-1;
2246 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2250 ret
= btrfs_previous_item(root
, path
, dirid
,
2251 BTRFS_INODE_REF_KEY
);
2261 slot
= path
->slots
[0];
2262 btrfs_item_key_to_cpu(l
, &key
, slot
);
2264 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2265 len
= btrfs_inode_ref_name_len(l
, iref
);
2267 total_len
+= len
+ 1;
2269 ret
= -ENAMETOOLONG
;
2274 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2276 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2279 btrfs_release_path(path
);
2280 key
.objectid
= key
.offset
;
2281 key
.offset
= (u64
)-1;
2282 dirid
= key
.objectid
;
2284 memmove(name
, ptr
, total_len
);
2285 name
[total_len
] = '\0';
2288 btrfs_free_path(path
);
2292 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2295 struct btrfs_ioctl_ino_lookup_args
*args
;
2296 struct inode
*inode
;
2299 args
= memdup_user(argp
, sizeof(*args
));
2301 return PTR_ERR(args
);
2303 inode
= file_inode(file
);
2306 * Unprivileged query to obtain the containing subvolume root id. The
2307 * path is reset so it's consistent with btrfs_search_path_in_tree.
2309 if (args
->treeid
== 0)
2310 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2312 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2317 if (!capable(CAP_SYS_ADMIN
)) {
2322 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2323 args
->treeid
, args
->objectid
,
2327 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2334 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2337 struct dentry
*parent
= file
->f_path
.dentry
;
2338 struct dentry
*dentry
;
2339 struct inode
*dir
= d_inode(parent
);
2340 struct inode
*inode
;
2341 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2342 struct btrfs_root
*dest
= NULL
;
2343 struct btrfs_ioctl_vol_args
*vol_args
;
2344 struct btrfs_trans_handle
*trans
;
2345 struct btrfs_block_rsv block_rsv
;
2347 u64 qgroup_reserved
;
2352 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2353 if (IS_ERR(vol_args
))
2354 return PTR_ERR(vol_args
);
2356 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2357 namelen
= strlen(vol_args
->name
);
2358 if (strchr(vol_args
->name
, '/') ||
2359 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2364 err
= mnt_want_write_file(file
);
2369 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2371 goto out_drop_write
;
2372 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2373 if (IS_ERR(dentry
)) {
2374 err
= PTR_ERR(dentry
);
2375 goto out_unlock_dir
;
2378 if (d_really_is_negative(dentry
)) {
2383 inode
= d_inode(dentry
);
2384 dest
= BTRFS_I(inode
)->root
;
2385 if (!capable(CAP_SYS_ADMIN
)) {
2387 * Regular user. Only allow this with a special mount
2388 * option, when the user has write+exec access to the
2389 * subvol root, and when rmdir(2) would have been
2392 * Note that this is _not_ check that the subvol is
2393 * empty or doesn't contain data that we wouldn't
2394 * otherwise be able to delete.
2396 * Users who want to delete empty subvols should try
2400 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2404 * Do not allow deletion if the parent dir is the same
2405 * as the dir to be deleted. That means the ioctl
2406 * must be called on the dentry referencing the root
2407 * of the subvol, not a random directory contained
2414 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2419 /* check if subvolume may be deleted by a user */
2420 err
= btrfs_may_delete(dir
, dentry
, 1);
2424 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2432 * Don't allow to delete a subvolume with send in progress. This is
2433 * inside the i_mutex so the error handling that has to drop the bit
2434 * again is not run concurrently.
2436 spin_lock(&dest
->root_item_lock
);
2437 root_flags
= btrfs_root_flags(&dest
->root_item
);
2438 if (dest
->send_in_progress
== 0) {
2439 btrfs_set_root_flags(&dest
->root_item
,
2440 root_flags
| BTRFS_ROOT_SUBVOL_DEAD
);
2441 spin_unlock(&dest
->root_item_lock
);
2443 spin_unlock(&dest
->root_item_lock
);
2444 btrfs_warn(root
->fs_info
,
2445 "Attempt to delete subvolume %llu during send",
2446 dest
->root_key
.objectid
);
2448 goto out_unlock_inode
;
2451 down_write(&root
->fs_info
->subvol_sem
);
2453 err
= may_destroy_subvol(dest
);
2457 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2459 * One for dir inode, two for dir entries, two for root
2462 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2463 5, &qgroup_reserved
, true);
2467 trans
= btrfs_start_transaction(root
, 0);
2468 if (IS_ERR(trans
)) {
2469 err
= PTR_ERR(trans
);
2472 trans
->block_rsv
= &block_rsv
;
2473 trans
->bytes_reserved
= block_rsv
.size
;
2475 btrfs_record_snapshot_destroy(trans
, dir
);
2477 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2478 dest
->root_key
.objectid
,
2479 dentry
->d_name
.name
,
2480 dentry
->d_name
.len
);
2483 btrfs_abort_transaction(trans
, root
, ret
);
2487 btrfs_record_root_in_trans(trans
, dest
);
2489 memset(&dest
->root_item
.drop_progress
, 0,
2490 sizeof(dest
->root_item
.drop_progress
));
2491 dest
->root_item
.drop_level
= 0;
2492 btrfs_set_root_refs(&dest
->root_item
, 0);
2494 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED
, &dest
->state
)) {
2495 ret
= btrfs_insert_orphan_item(trans
,
2496 root
->fs_info
->tree_root
,
2497 dest
->root_key
.objectid
);
2499 btrfs_abort_transaction(trans
, root
, ret
);
2505 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2506 dest
->root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
2507 dest
->root_key
.objectid
);
2508 if (ret
&& ret
!= -ENOENT
) {
2509 btrfs_abort_transaction(trans
, root
, ret
);
2513 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2514 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2515 dest
->root_item
.received_uuid
,
2516 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2517 dest
->root_key
.objectid
);
2518 if (ret
&& ret
!= -ENOENT
) {
2519 btrfs_abort_transaction(trans
, root
, ret
);
2526 trans
->block_rsv
= NULL
;
2527 trans
->bytes_reserved
= 0;
2528 ret
= btrfs_end_transaction(trans
, root
);
2531 inode
->i_flags
|= S_DEAD
;
2533 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2535 up_write(&root
->fs_info
->subvol_sem
);
2537 spin_lock(&dest
->root_item_lock
);
2538 root_flags
= btrfs_root_flags(&dest
->root_item
);
2539 btrfs_set_root_flags(&dest
->root_item
,
2540 root_flags
& ~BTRFS_ROOT_SUBVOL_DEAD
);
2541 spin_unlock(&dest
->root_item_lock
);
2544 inode_unlock(inode
);
2546 d_invalidate(dentry
);
2547 btrfs_invalidate_inodes(dest
);
2549 ASSERT(dest
->send_in_progress
== 0);
2552 if (dest
->ino_cache_inode
) {
2553 iput(dest
->ino_cache_inode
);
2554 dest
->ino_cache_inode
= NULL
;
2562 mnt_drop_write_file(file
);
2568 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2570 struct inode
*inode
= file_inode(file
);
2571 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2572 struct btrfs_ioctl_defrag_range_args
*range
;
2575 ret
= mnt_want_write_file(file
);
2579 if (btrfs_root_readonly(root
)) {
2584 switch (inode
->i_mode
& S_IFMT
) {
2586 if (!capable(CAP_SYS_ADMIN
)) {
2590 ret
= btrfs_defrag_root(root
);
2593 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2596 if (!(file
->f_mode
& FMODE_WRITE
)) {
2601 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2608 if (copy_from_user(range
, argp
,
2614 /* compression requires us to start the IO */
2615 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2616 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2617 range
->extent_thresh
= (u32
)-1;
2620 /* the rest are all set to zero by kzalloc */
2621 range
->len
= (u64
)-1;
2623 ret
= btrfs_defrag_file(file_inode(file
), file
,
2633 mnt_drop_write_file(file
);
2637 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2639 struct btrfs_ioctl_vol_args
*vol_args
;
2642 if (!capable(CAP_SYS_ADMIN
))
2645 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2647 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2650 mutex_lock(&root
->fs_info
->volume_mutex
);
2651 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2652 if (IS_ERR(vol_args
)) {
2653 ret
= PTR_ERR(vol_args
);
2657 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2658 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2661 btrfs_info(root
->fs_info
, "disk added %s",vol_args
->name
);
2665 mutex_unlock(&root
->fs_info
->volume_mutex
);
2666 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2670 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2672 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2673 struct btrfs_ioctl_vol_args
*vol_args
;
2676 if (!capable(CAP_SYS_ADMIN
))
2679 ret
= mnt_want_write_file(file
);
2683 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2684 if (IS_ERR(vol_args
)) {
2685 ret
= PTR_ERR(vol_args
);
2689 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2691 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2693 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2697 mutex_lock(&root
->fs_info
->volume_mutex
);
2698 ret
= btrfs_rm_device(root
, vol_args
->name
);
2699 mutex_unlock(&root
->fs_info
->volume_mutex
);
2700 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2703 btrfs_info(root
->fs_info
, "disk deleted %s",vol_args
->name
);
2708 mnt_drop_write_file(file
);
2712 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2714 struct btrfs_ioctl_fs_info_args
*fi_args
;
2715 struct btrfs_device
*device
;
2716 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2719 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2723 mutex_lock(&fs_devices
->device_list_mutex
);
2724 fi_args
->num_devices
= fs_devices
->num_devices
;
2725 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2727 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
2728 if (device
->devid
> fi_args
->max_id
)
2729 fi_args
->max_id
= device
->devid
;
2731 mutex_unlock(&fs_devices
->device_list_mutex
);
2733 fi_args
->nodesize
= root
->fs_info
->super_copy
->nodesize
;
2734 fi_args
->sectorsize
= root
->fs_info
->super_copy
->sectorsize
;
2735 fi_args
->clone_alignment
= root
->fs_info
->super_copy
->sectorsize
;
2737 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2744 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2746 struct btrfs_ioctl_dev_info_args
*di_args
;
2747 struct btrfs_device
*dev
;
2748 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2750 char *s_uuid
= NULL
;
2752 di_args
= memdup_user(arg
, sizeof(*di_args
));
2753 if (IS_ERR(di_args
))
2754 return PTR_ERR(di_args
);
2756 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2757 s_uuid
= di_args
->uuid
;
2759 mutex_lock(&fs_devices
->device_list_mutex
);
2760 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2767 di_args
->devid
= dev
->devid
;
2768 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
2769 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
2770 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2772 struct rcu_string
*name
;
2775 name
= rcu_dereference(dev
->name
);
2776 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2778 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2780 di_args
->path
[0] = '\0';
2784 mutex_unlock(&fs_devices
->device_list_mutex
);
2785 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2792 static struct page
*extent_same_get_page(struct inode
*inode
, pgoff_t index
)
2796 page
= grab_cache_page(inode
->i_mapping
, index
);
2798 return ERR_PTR(-ENOMEM
);
2800 if (!PageUptodate(page
)) {
2803 ret
= btrfs_readpage(NULL
, page
);
2805 return ERR_PTR(ret
);
2807 if (!PageUptodate(page
)) {
2810 return ERR_PTR(-EIO
);
2812 if (page
->mapping
!= inode
->i_mapping
) {
2815 return ERR_PTR(-EAGAIN
);
2822 static int gather_extent_pages(struct inode
*inode
, struct page
**pages
,
2823 int num_pages
, u64 off
)
2826 pgoff_t index
= off
>> PAGE_SHIFT
;
2828 for (i
= 0; i
< num_pages
; i
++) {
2830 pages
[i
] = extent_same_get_page(inode
, index
+ i
);
2831 if (IS_ERR(pages
[i
])) {
2832 int err
= PTR_ERR(pages
[i
]);
2843 static int lock_extent_range(struct inode
*inode
, u64 off
, u64 len
,
2844 bool retry_range_locking
)
2847 * Do any pending delalloc/csum calculations on inode, one way or
2848 * another, and lock file content.
2849 * The locking order is:
2852 * 2) range in the inode's io tree
2855 struct btrfs_ordered_extent
*ordered
;
2856 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2857 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2860 ordered
->file_offset
+ ordered
->len
<= off
||
2861 ordered
->file_offset
>= off
+ len
) &&
2862 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2863 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
)) {
2865 btrfs_put_ordered_extent(ordered
);
2868 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2870 btrfs_put_ordered_extent(ordered
);
2871 if (!retry_range_locking
)
2873 btrfs_wait_ordered_range(inode
, off
, len
);
2878 static void btrfs_double_inode_unlock(struct inode
*inode1
, struct inode
*inode2
)
2880 inode_unlock(inode1
);
2881 inode_unlock(inode2
);
2884 static void btrfs_double_inode_lock(struct inode
*inode1
, struct inode
*inode2
)
2886 if (inode1
< inode2
)
2887 swap(inode1
, inode2
);
2889 inode_lock_nested(inode1
, I_MUTEX_PARENT
);
2890 inode_lock_nested(inode2
, I_MUTEX_CHILD
);
2893 static void btrfs_double_extent_unlock(struct inode
*inode1
, u64 loff1
,
2894 struct inode
*inode2
, u64 loff2
, u64 len
)
2896 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2897 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2900 static int btrfs_double_extent_lock(struct inode
*inode1
, u64 loff1
,
2901 struct inode
*inode2
, u64 loff2
, u64 len
,
2902 bool retry_range_locking
)
2906 if (inode1
< inode2
) {
2907 swap(inode1
, inode2
);
2910 ret
= lock_extent_range(inode1
, loff1
, len
, retry_range_locking
);
2913 ret
= lock_extent_range(inode2
, loff2
, len
, retry_range_locking
);
2915 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
,
2922 struct page
**src_pages
;
2923 struct page
**dst_pages
;
2926 static void btrfs_cmp_data_free(struct cmp_pages
*cmp
)
2931 for (i
= 0; i
< cmp
->num_pages
; i
++) {
2932 pg
= cmp
->src_pages
[i
];
2937 pg
= cmp
->dst_pages
[i
];
2943 kfree(cmp
->src_pages
);
2944 kfree(cmp
->dst_pages
);
2947 static int btrfs_cmp_data_prepare(struct inode
*src
, u64 loff
,
2948 struct inode
*dst
, u64 dst_loff
,
2949 u64 len
, struct cmp_pages
*cmp
)
2952 int num_pages
= PAGE_ALIGN(len
) >> PAGE_SHIFT
;
2953 struct page
**src_pgarr
, **dst_pgarr
;
2956 * We must gather up all the pages before we initiate our
2957 * extent locking. We use an array for the page pointers. Size
2958 * of the array is bounded by len, which is in turn bounded by
2959 * BTRFS_MAX_DEDUPE_LEN.
2961 src_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
2962 dst_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
2963 if (!src_pgarr
|| !dst_pgarr
) {
2968 cmp
->num_pages
= num_pages
;
2969 cmp
->src_pages
= src_pgarr
;
2970 cmp
->dst_pages
= dst_pgarr
;
2972 ret
= gather_extent_pages(src
, cmp
->src_pages
, cmp
->num_pages
, loff
);
2976 ret
= gather_extent_pages(dst
, cmp
->dst_pages
, cmp
->num_pages
, dst_loff
);
2980 btrfs_cmp_data_free(cmp
);
2984 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
2985 u64 dst_loff
, u64 len
, struct cmp_pages
*cmp
)
2989 struct page
*src_page
, *dst_page
;
2990 unsigned int cmp_len
= PAGE_SIZE
;
2991 void *addr
, *dst_addr
;
2995 if (len
< PAGE_SIZE
)
2998 BUG_ON(i
>= cmp
->num_pages
);
3000 src_page
= cmp
->src_pages
[i
];
3001 dst_page
= cmp
->dst_pages
[i
];
3002 ASSERT(PageLocked(src_page
));
3003 ASSERT(PageLocked(dst_page
));
3005 addr
= kmap_atomic(src_page
);
3006 dst_addr
= kmap_atomic(dst_page
);
3008 flush_dcache_page(src_page
);
3009 flush_dcache_page(dst_page
);
3011 if (memcmp(addr
, dst_addr
, cmp_len
))
3014 kunmap_atomic(addr
);
3015 kunmap_atomic(dst_addr
);
3027 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64
*plen
,
3031 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
3033 if (off
+ olen
> inode
->i_size
|| off
+ olen
< off
)
3036 /* if we extend to eof, continue to block boundary */
3037 if (off
+ len
== inode
->i_size
)
3038 *plen
= len
= ALIGN(inode
->i_size
, bs
) - off
;
3040 /* Check that we are block aligned - btrfs_clone() requires this */
3041 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
3047 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 olen
,
3048 struct inode
*dst
, u64 dst_loff
)
3052 struct cmp_pages cmp
;
3054 u64 same_lock_start
= 0;
3055 u64 same_lock_len
= 0;
3066 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3069 ret
= extent_same_check_offsets(src
, dst_loff
, &len
, olen
);
3074 * Single inode case wants the same checks, except we
3075 * don't want our length pushed out past i_size as
3076 * comparing that data range makes no sense.
3078 * extent_same_check_offsets() will do this for an
3079 * unaligned length at i_size, so catch it here and
3080 * reject the request.
3082 * This effectively means we require aligned extents
3083 * for the single-inode case, whereas the other cases
3084 * allow an unaligned length so long as it ends at
3092 /* Check for overlapping ranges */
3093 if (dst_loff
+ len
> loff
&& dst_loff
< loff
+ len
) {
3098 same_lock_start
= min_t(u64
, loff
, dst_loff
);
3099 same_lock_len
= max_t(u64
, loff
, dst_loff
) + len
- same_lock_start
;
3101 btrfs_double_inode_lock(src
, dst
);
3103 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3107 ret
= extent_same_check_offsets(dst
, dst_loff
, &len
, olen
);
3112 /* don't make the dst file partly checksummed */
3113 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3114 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
3120 ret
= btrfs_cmp_data_prepare(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3125 ret
= lock_extent_range(src
, same_lock_start
, same_lock_len
,
3128 ret
= btrfs_double_extent_lock(src
, loff
, dst
, dst_loff
, len
,
3131 * If one of the inodes has dirty pages in the respective range or
3132 * ordered extents, we need to flush dellaloc and wait for all ordered
3133 * extents in the range. We must unlock the pages and the ranges in the
3134 * io trees to avoid deadlocks when flushing delalloc (requires locking
3135 * pages) and when waiting for ordered extents to complete (they require
3138 if (ret
== -EAGAIN
) {
3140 * Ranges in the io trees already unlocked. Now unlock all
3141 * pages before waiting for all IO to complete.
3143 btrfs_cmp_data_free(&cmp
);
3145 btrfs_wait_ordered_range(src
, same_lock_start
,
3148 btrfs_wait_ordered_range(src
, loff
, len
);
3149 btrfs_wait_ordered_range(dst
, dst_loff
, len
);
3155 /* ranges in the io trees already unlocked */
3156 btrfs_cmp_data_free(&cmp
);
3160 /* pass original length for comparison so we stay within i_size */
3161 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3163 ret
= btrfs_clone(src
, dst
, loff
, olen
, len
, dst_loff
, 1);
3166 unlock_extent(&BTRFS_I(src
)->io_tree
, same_lock_start
,
3167 same_lock_start
+ same_lock_len
- 1);
3169 btrfs_double_extent_unlock(src
, loff
, dst
, dst_loff
, len
);
3171 btrfs_cmp_data_free(&cmp
);
3176 btrfs_double_inode_unlock(src
, dst
);
3181 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3183 ssize_t
btrfs_dedupe_file_range(struct file
*src_file
, u64 loff
, u64 olen
,
3184 struct file
*dst_file
, u64 dst_loff
)
3186 struct inode
*src
= file_inode(src_file
);
3187 struct inode
*dst
= file_inode(dst_file
);
3188 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
3191 if (olen
> BTRFS_MAX_DEDUPE_LEN
)
3192 olen
= BTRFS_MAX_DEDUPE_LEN
;
3194 if (WARN_ON_ONCE(bs
< PAGE_SIZE
)) {
3196 * Btrfs does not support blocksize < page_size. As a
3197 * result, btrfs_cmp_data() won't correctly handle
3198 * this situation without an update.
3203 res
= btrfs_extent_same(src
, loff
, olen
, dst
, dst_loff
);
3209 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3210 struct inode
*inode
,
3216 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3219 inode_inc_iversion(inode
);
3220 if (!no_time_update
)
3221 inode
->i_mtime
= inode
->i_ctime
= current_fs_time(inode
->i_sb
);
3223 * We round up to the block size at eof when determining which
3224 * extents to clone above, but shouldn't round up the file size.
3226 if (endoff
> destoff
+ olen
)
3227 endoff
= destoff
+ olen
;
3228 if (endoff
> inode
->i_size
)
3229 btrfs_i_size_write(inode
, endoff
);
3231 ret
= btrfs_update_inode(trans
, root
, inode
);
3233 btrfs_abort_transaction(trans
, root
, ret
);
3234 btrfs_end_transaction(trans
, root
);
3237 ret
= btrfs_end_transaction(trans
, root
);
3242 static void clone_update_extent_map(struct inode
*inode
,
3243 const struct btrfs_trans_handle
*trans
,
3244 const struct btrfs_path
*path
,
3245 const u64 hole_offset
,
3248 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
3249 struct extent_map
*em
;
3252 em
= alloc_extent_map();
3254 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3255 &BTRFS_I(inode
)->runtime_flags
);
3260 struct btrfs_file_extent_item
*fi
;
3262 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3263 struct btrfs_file_extent_item
);
3264 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3265 em
->generation
= -1;
3266 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3267 BTRFS_FILE_EXTENT_INLINE
)
3268 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3269 &BTRFS_I(inode
)->runtime_flags
);
3271 em
->start
= hole_offset
;
3273 em
->ram_bytes
= em
->len
;
3274 em
->orig_start
= hole_offset
;
3275 em
->block_start
= EXTENT_MAP_HOLE
;
3277 em
->orig_block_len
= 0;
3278 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3279 em
->generation
= trans
->transid
;
3283 write_lock(&em_tree
->lock
);
3284 ret
= add_extent_mapping(em_tree
, em
, 1);
3285 write_unlock(&em_tree
->lock
);
3286 if (ret
!= -EEXIST
) {
3287 free_extent_map(em
);
3290 btrfs_drop_extent_cache(inode
, em
->start
,
3291 em
->start
+ em
->len
- 1, 0);
3295 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3296 &BTRFS_I(inode
)->runtime_flags
);
3300 * Make sure we do not end up inserting an inline extent into a file that has
3301 * already other (non-inline) extents. If a file has an inline extent it can
3302 * not have any other extents and the (single) inline extent must start at the
3303 * file offset 0. Failing to respect these rules will lead to file corruption,
3304 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3306 * We can have extents that have been already written to disk or we can have
3307 * dirty ranges still in delalloc, in which case the extent maps and items are
3308 * created only when we run delalloc, and the delalloc ranges might fall outside
3309 * the range we are currently locking in the inode's io tree. So we check the
3310 * inode's i_size because of that (i_size updates are done while holding the
3311 * i_mutex, which we are holding here).
3312 * We also check to see if the inode has a size not greater than "datal" but has
3313 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3314 * protected against such concurrent fallocate calls by the i_mutex).
3316 * If the file has no extents but a size greater than datal, do not allow the
3317 * copy because we would need turn the inline extent into a non-inline one (even
3318 * with NO_HOLES enabled). If we find our destination inode only has one inline
3319 * extent, just overwrite it with the source inline extent if its size is less
3320 * than the source extent's size, or we could copy the source inline extent's
3321 * data into the destination inode's inline extent if the later is greater then
3324 static int clone_copy_inline_extent(struct inode
*src
,
3326 struct btrfs_trans_handle
*trans
,
3327 struct btrfs_path
*path
,
3328 struct btrfs_key
*new_key
,
3329 const u64 drop_start
,
3335 struct btrfs_root
*root
= BTRFS_I(dst
)->root
;
3336 const u64 aligned_end
= ALIGN(new_key
->offset
+ datal
,
3339 struct btrfs_key key
;
3341 if (new_key
->offset
> 0)
3344 key
.objectid
= btrfs_ino(dst
);
3345 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3347 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
3350 } else if (ret
> 0) {
3351 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3352 ret
= btrfs_next_leaf(root
, path
);
3356 goto copy_inline_extent
;
3358 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
3359 if (key
.objectid
== btrfs_ino(dst
) &&
3360 key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3361 ASSERT(key
.offset
> 0);
3364 } else if (i_size_read(dst
) <= datal
) {
3365 struct btrfs_file_extent_item
*ei
;
3369 * If the file size is <= datal, make sure there are no other
3370 * extents following (can happen do to an fallocate call with
3371 * the flag FALLOC_FL_KEEP_SIZE).
3373 ei
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3374 struct btrfs_file_extent_item
);
3376 * If it's an inline extent, it can not have other extents
3379 if (btrfs_file_extent_type(path
->nodes
[0], ei
) ==
3380 BTRFS_FILE_EXTENT_INLINE
)
3381 goto copy_inline_extent
;
3383 ext_len
= btrfs_file_extent_num_bytes(path
->nodes
[0], ei
);
3384 if (ext_len
> aligned_end
)
3387 ret
= btrfs_next_item(root
, path
);
3390 } else if (ret
== 0) {
3391 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3393 if (key
.objectid
== btrfs_ino(dst
) &&
3394 key
.type
== BTRFS_EXTENT_DATA_KEY
)
3401 * We have no extent items, or we have an extent at offset 0 which may
3402 * or may not be inlined. All these cases are dealt the same way.
3404 if (i_size_read(dst
) > datal
) {
3406 * If the destination inode has an inline extent...
3407 * This would require copying the data from the source inline
3408 * extent into the beginning of the destination's inline extent.
3409 * But this is really complex, both extents can be compressed
3410 * or just one of them, which would require decompressing and
3411 * re-compressing data (which could increase the new compressed
3412 * size, not allowing the compressed data to fit anymore in an
3414 * So just don't support this case for now (it should be rare,
3415 * we are not really saving space when cloning inline extents).
3420 btrfs_release_path(path
);
3421 ret
= btrfs_drop_extents(trans
, root
, dst
, drop_start
, aligned_end
, 1);
3424 ret
= btrfs_insert_empty_item(trans
, root
, path
, new_key
, size
);
3429 const u32 start
= btrfs_file_extent_calc_inline_size(0);
3431 memmove(inline_data
+ start
, inline_data
+ start
+ skip
, datal
);
3434 write_extent_buffer(path
->nodes
[0], inline_data
,
3435 btrfs_item_ptr_offset(path
->nodes
[0],
3438 inode_add_bytes(dst
, datal
);
3444 * btrfs_clone() - clone a range from inode file to another
3446 * @src: Inode to clone from
3447 * @inode: Inode to clone to
3448 * @off: Offset within source to start clone from
3449 * @olen: Original length, passed by user, of range to clone
3450 * @olen_aligned: Block-aligned value of olen
3451 * @destoff: Offset within @inode to start clone
3452 * @no_time_update: Whether to update mtime/ctime on the target inode
3454 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3455 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3456 const u64 destoff
, int no_time_update
)
3458 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3459 struct btrfs_path
*path
= NULL
;
3460 struct extent_buffer
*leaf
;
3461 struct btrfs_trans_handle
*trans
;
3463 struct btrfs_key key
;
3467 const u64 len
= olen_aligned
;
3468 u64 last_dest_end
= destoff
;
3471 buf
= kmalloc(root
->nodesize
, GFP_KERNEL
| __GFP_NOWARN
);
3473 buf
= vmalloc(root
->nodesize
);
3478 path
= btrfs_alloc_path();
3484 path
->reada
= READA_FORWARD
;
3486 key
.objectid
= btrfs_ino(src
);
3487 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3491 u64 next_key_min_offset
= key
.offset
+ 1;
3494 * note the key will change type as we walk through the
3497 path
->leave_spinning
= 1;
3498 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3503 * First search, if no extent item that starts at offset off was
3504 * found but the previous item is an extent item, it's possible
3505 * it might overlap our target range, therefore process it.
3507 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3508 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3509 path
->slots
[0] - 1);
3510 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3514 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3516 if (path
->slots
[0] >= nritems
) {
3517 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3522 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3524 leaf
= path
->nodes
[0];
3525 slot
= path
->slots
[0];
3527 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3528 if (key
.type
> BTRFS_EXTENT_DATA_KEY
||
3529 key
.objectid
!= btrfs_ino(src
))
3532 if (key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3533 struct btrfs_file_extent_item
*extent
;
3536 struct btrfs_key new_key
;
3537 u64 disko
= 0, diskl
= 0;
3538 u64 datao
= 0, datal
= 0;
3542 extent
= btrfs_item_ptr(leaf
, slot
,
3543 struct btrfs_file_extent_item
);
3544 comp
= btrfs_file_extent_compression(leaf
, extent
);
3545 type
= btrfs_file_extent_type(leaf
, extent
);
3546 if (type
== BTRFS_FILE_EXTENT_REG
||
3547 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3548 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3550 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3552 datao
= btrfs_file_extent_offset(leaf
, extent
);
3553 datal
= btrfs_file_extent_num_bytes(leaf
,
3555 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3556 /* take upper bound, may be compressed */
3557 datal
= btrfs_file_extent_ram_bytes(leaf
,
3562 * The first search might have left us at an extent
3563 * item that ends before our target range's start, can
3564 * happen if we have holes and NO_HOLES feature enabled.
3566 if (key
.offset
+ datal
<= off
) {
3569 } else if (key
.offset
>= off
+ len
) {
3572 next_key_min_offset
= key
.offset
+ datal
;
3573 size
= btrfs_item_size_nr(leaf
, slot
);
3574 read_extent_buffer(leaf
, buf
,
3575 btrfs_item_ptr_offset(leaf
, slot
),
3578 btrfs_release_path(path
);
3579 path
->leave_spinning
= 0;
3581 memcpy(&new_key
, &key
, sizeof(new_key
));
3582 new_key
.objectid
= btrfs_ino(inode
);
3583 if (off
<= key
.offset
)
3584 new_key
.offset
= key
.offset
+ destoff
- off
;
3586 new_key
.offset
= destoff
;
3589 * Deal with a hole that doesn't have an extent item
3590 * that represents it (NO_HOLES feature enabled).
3591 * This hole is either in the middle of the cloning
3592 * range or at the beginning (fully overlaps it or
3593 * partially overlaps it).
3595 if (new_key
.offset
!= last_dest_end
)
3596 drop_start
= last_dest_end
;
3598 drop_start
= new_key
.offset
;
3601 * 1 - adjusting old extent (we may have to split it)
3602 * 1 - add new extent
3605 trans
= btrfs_start_transaction(root
, 3);
3606 if (IS_ERR(trans
)) {
3607 ret
= PTR_ERR(trans
);
3611 if (type
== BTRFS_FILE_EXTENT_REG
||
3612 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3614 * a | --- range to clone ---| b
3615 * | ------------- extent ------------- |
3618 /* subtract range b */
3619 if (key
.offset
+ datal
> off
+ len
)
3620 datal
= off
+ len
- key
.offset
;
3622 /* subtract range a */
3623 if (off
> key
.offset
) {
3624 datao
+= off
- key
.offset
;
3625 datal
-= off
- key
.offset
;
3628 ret
= btrfs_drop_extents(trans
, root
, inode
,
3630 new_key
.offset
+ datal
,
3633 if (ret
!= -EOPNOTSUPP
)
3634 btrfs_abort_transaction(trans
,
3636 btrfs_end_transaction(trans
, root
);
3640 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3643 btrfs_abort_transaction(trans
, root
,
3645 btrfs_end_transaction(trans
, root
);
3649 leaf
= path
->nodes
[0];
3650 slot
= path
->slots
[0];
3651 write_extent_buffer(leaf
, buf
,
3652 btrfs_item_ptr_offset(leaf
, slot
),
3655 extent
= btrfs_item_ptr(leaf
, slot
,
3656 struct btrfs_file_extent_item
);
3658 /* disko == 0 means it's a hole */
3662 btrfs_set_file_extent_offset(leaf
, extent
,
3664 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3668 inode_add_bytes(inode
, datal
);
3669 ret
= btrfs_inc_extent_ref(trans
, root
,
3671 root
->root_key
.objectid
,
3673 new_key
.offset
- datao
);
3675 btrfs_abort_transaction(trans
,
3678 btrfs_end_transaction(trans
,
3684 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3688 if (off
> key
.offset
) {
3689 skip
= off
- key
.offset
;
3690 new_key
.offset
+= skip
;
3693 if (key
.offset
+ datal
> off
+ len
)
3694 trim
= key
.offset
+ datal
- (off
+ len
);
3696 if (comp
&& (skip
|| trim
)) {
3698 btrfs_end_transaction(trans
, root
);
3701 size
-= skip
+ trim
;
3702 datal
-= skip
+ trim
;
3704 ret
= clone_copy_inline_extent(src
, inode
,
3711 if (ret
!= -EOPNOTSUPP
)
3712 btrfs_abort_transaction(trans
,
3715 btrfs_end_transaction(trans
, root
);
3718 leaf
= path
->nodes
[0];
3719 slot
= path
->slots
[0];
3722 /* If we have an implicit hole (NO_HOLES feature). */
3723 if (drop_start
< new_key
.offset
)
3724 clone_update_extent_map(inode
, trans
,
3726 new_key
.offset
- drop_start
);
3728 clone_update_extent_map(inode
, trans
, path
, 0, 0);
3730 btrfs_mark_buffer_dirty(leaf
);
3731 btrfs_release_path(path
);
3733 last_dest_end
= ALIGN(new_key
.offset
+ datal
,
3735 ret
= clone_finish_inode_update(trans
, inode
,
3741 if (new_key
.offset
+ datal
>= destoff
+ len
)
3744 btrfs_release_path(path
);
3745 key
.offset
= next_key_min_offset
;
3749 if (last_dest_end
< destoff
+ len
) {
3751 * We have an implicit hole (NO_HOLES feature is enabled) that
3752 * fully or partially overlaps our cloning range at its end.
3754 btrfs_release_path(path
);
3757 * 1 - remove extent(s)
3760 trans
= btrfs_start_transaction(root
, 2);
3761 if (IS_ERR(trans
)) {
3762 ret
= PTR_ERR(trans
);
3765 ret
= btrfs_drop_extents(trans
, root
, inode
,
3766 last_dest_end
, destoff
+ len
, 1);
3768 if (ret
!= -EOPNOTSUPP
)
3769 btrfs_abort_transaction(trans
, root
, ret
);
3770 btrfs_end_transaction(trans
, root
);
3773 clone_update_extent_map(inode
, trans
, NULL
, last_dest_end
,
3774 destoff
+ len
- last_dest_end
);
3775 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3776 destoff
, olen
, no_time_update
);
3780 btrfs_free_path(path
);
3785 static noinline
int btrfs_clone_files(struct file
*file
, struct file
*file_src
,
3786 u64 off
, u64 olen
, u64 destoff
)
3788 struct inode
*inode
= file_inode(file
);
3789 struct inode
*src
= file_inode(file_src
);
3790 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3793 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
3794 int same_inode
= src
== inode
;
3798 * - split compressed inline extents. annoying: we need to
3799 * decompress into destination's address_space (the file offset
3800 * may change, so source mapping won't do), then recompress (or
3801 * otherwise reinsert) a subrange.
3803 * - split destination inode's inline extents. The inline extents can
3804 * be either compressed or non-compressed.
3807 if (btrfs_root_readonly(root
))
3810 if (file_src
->f_path
.mnt
!= file
->f_path
.mnt
||
3811 src
->i_sb
!= inode
->i_sb
)
3814 /* don't make the dst file partly checksummed */
3815 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3816 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3819 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3823 btrfs_double_inode_lock(src
, inode
);
3828 /* determine range to clone */
3830 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3833 olen
= len
= src
->i_size
- off
;
3834 /* if we extend to eof, continue to block boundary */
3835 if (off
+ len
== src
->i_size
)
3836 len
= ALIGN(src
->i_size
, bs
) - off
;
3843 /* verify the end result is block aligned */
3844 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3845 !IS_ALIGNED(destoff
, bs
))
3848 /* verify if ranges are overlapped within the same file */
3850 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3854 if (destoff
> inode
->i_size
) {
3855 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3861 * Lock the target range too. Right after we replace the file extent
3862 * items in the fs tree (which now point to the cloned data), we might
3863 * have a worker replace them with extent items relative to a write
3864 * operation that was issued before this clone operation (i.e. confront
3865 * with inode.c:btrfs_finish_ordered_io).
3868 u64 lock_start
= min_t(u64
, off
, destoff
);
3869 u64 lock_len
= max_t(u64
, off
, destoff
) + len
- lock_start
;
3871 ret
= lock_extent_range(src
, lock_start
, lock_len
, true);
3873 ret
= btrfs_double_extent_lock(src
, off
, inode
, destoff
, len
,
3878 /* ranges in the io trees already unlocked */
3882 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
, 0);
3885 u64 lock_start
= min_t(u64
, off
, destoff
);
3886 u64 lock_end
= max_t(u64
, off
, destoff
) + len
- 1;
3888 unlock_extent(&BTRFS_I(src
)->io_tree
, lock_start
, lock_end
);
3890 btrfs_double_extent_unlock(src
, off
, inode
, destoff
, len
);
3893 * Truncate page cache pages so that future reads will see the cloned
3894 * data immediately and not the previous data.
3896 truncate_inode_pages_range(&inode
->i_data
,
3897 round_down(destoff
, PAGE_SIZE
),
3898 round_up(destoff
+ len
, PAGE_SIZE
) - 1);
3901 btrfs_double_inode_unlock(src
, inode
);
3907 ssize_t
btrfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
3908 struct file
*file_out
, loff_t pos_out
,
3909 size_t len
, unsigned int flags
)
3913 ret
= btrfs_clone_files(file_out
, file_in
, pos_in
, len
, pos_out
);
3919 int btrfs_clone_file_range(struct file
*src_file
, loff_t off
,
3920 struct file
*dst_file
, loff_t destoff
, u64 len
)
3922 return btrfs_clone_files(dst_file
, src_file
, off
, len
, destoff
);
3926 * there are many ways the trans_start and trans_end ioctls can lead
3927 * to deadlocks. They should only be used by applications that
3928 * basically own the machine, and have a very in depth understanding
3929 * of all the possible deadlocks and enospc problems.
3931 static long btrfs_ioctl_trans_start(struct file
*file
)
3933 struct inode
*inode
= file_inode(file
);
3934 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3935 struct btrfs_trans_handle
*trans
;
3939 if (!capable(CAP_SYS_ADMIN
))
3943 if (file
->private_data
)
3947 if (btrfs_root_readonly(root
))
3950 ret
= mnt_want_write_file(file
);
3954 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
3957 trans
= btrfs_start_ioctl_transaction(root
);
3961 file
->private_data
= trans
;
3965 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3966 mnt_drop_write_file(file
);
3971 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
3973 struct inode
*inode
= file_inode(file
);
3974 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3975 struct btrfs_root
*new_root
;
3976 struct btrfs_dir_item
*di
;
3977 struct btrfs_trans_handle
*trans
;
3978 struct btrfs_path
*path
;
3979 struct btrfs_key location
;
3980 struct btrfs_disk_key disk_key
;
3985 if (!capable(CAP_SYS_ADMIN
))
3988 ret
= mnt_want_write_file(file
);
3992 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
3998 objectid
= BTRFS_FS_TREE_OBJECTID
;
4000 location
.objectid
= objectid
;
4001 location
.type
= BTRFS_ROOT_ITEM_KEY
;
4002 location
.offset
= (u64
)-1;
4004 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
4005 if (IS_ERR(new_root
)) {
4006 ret
= PTR_ERR(new_root
);
4010 path
= btrfs_alloc_path();
4015 path
->leave_spinning
= 1;
4017 trans
= btrfs_start_transaction(root
, 1);
4018 if (IS_ERR(trans
)) {
4019 btrfs_free_path(path
);
4020 ret
= PTR_ERR(trans
);
4024 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
4025 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
4026 dir_id
, "default", 7, 1);
4027 if (IS_ERR_OR_NULL(di
)) {
4028 btrfs_free_path(path
);
4029 btrfs_end_transaction(trans
, root
);
4030 btrfs_err(new_root
->fs_info
, "Umm, you don't have the default dir"
4031 "item, this isn't going to work");
4036 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
4037 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
4038 btrfs_mark_buffer_dirty(path
->nodes
[0]);
4039 btrfs_free_path(path
);
4041 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
4042 btrfs_end_transaction(trans
, root
);
4044 mnt_drop_write_file(file
);
4048 void btrfs_get_block_group_info(struct list_head
*groups_list
,
4049 struct btrfs_ioctl_space_info
*space
)
4051 struct btrfs_block_group_cache
*block_group
;
4053 space
->total_bytes
= 0;
4054 space
->used_bytes
= 0;
4056 list_for_each_entry(block_group
, groups_list
, list
) {
4057 space
->flags
= block_group
->flags
;
4058 space
->total_bytes
+= block_group
->key
.offset
;
4059 space
->used_bytes
+=
4060 btrfs_block_group_used(&block_group
->item
);
4064 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
4066 struct btrfs_ioctl_space_args space_args
;
4067 struct btrfs_ioctl_space_info space
;
4068 struct btrfs_ioctl_space_info
*dest
;
4069 struct btrfs_ioctl_space_info
*dest_orig
;
4070 struct btrfs_ioctl_space_info __user
*user_dest
;
4071 struct btrfs_space_info
*info
;
4072 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
4073 BTRFS_BLOCK_GROUP_SYSTEM
,
4074 BTRFS_BLOCK_GROUP_METADATA
,
4075 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
4082 if (copy_from_user(&space_args
,
4083 (struct btrfs_ioctl_space_args __user
*)arg
,
4084 sizeof(space_args
)))
4087 for (i
= 0; i
< num_types
; i
++) {
4088 struct btrfs_space_info
*tmp
;
4092 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4094 if (tmp
->flags
== types
[i
]) {
4104 down_read(&info
->groups_sem
);
4105 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4106 if (!list_empty(&info
->block_groups
[c
]))
4109 up_read(&info
->groups_sem
);
4113 * Global block reserve, exported as a space_info
4117 /* space_slots == 0 means they are asking for a count */
4118 if (space_args
.space_slots
== 0) {
4119 space_args
.total_spaces
= slot_count
;
4123 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
4125 alloc_size
= sizeof(*dest
) * slot_count
;
4127 /* we generally have at most 6 or so space infos, one for each raid
4128 * level. So, a whole page should be more than enough for everyone
4130 if (alloc_size
> PAGE_SIZE
)
4133 space_args
.total_spaces
= 0;
4134 dest
= kmalloc(alloc_size
, GFP_KERNEL
);
4139 /* now we have a buffer to copy into */
4140 for (i
= 0; i
< num_types
; i
++) {
4141 struct btrfs_space_info
*tmp
;
4148 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4150 if (tmp
->flags
== types
[i
]) {
4159 down_read(&info
->groups_sem
);
4160 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4161 if (!list_empty(&info
->block_groups
[c
])) {
4162 btrfs_get_block_group_info(
4163 &info
->block_groups
[c
], &space
);
4164 memcpy(dest
, &space
, sizeof(space
));
4166 space_args
.total_spaces
++;
4172 up_read(&info
->groups_sem
);
4176 * Add global block reserve
4179 struct btrfs_block_rsv
*block_rsv
= &root
->fs_info
->global_block_rsv
;
4181 spin_lock(&block_rsv
->lock
);
4182 space
.total_bytes
= block_rsv
->size
;
4183 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4184 spin_unlock(&block_rsv
->lock
);
4185 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4186 memcpy(dest
, &space
, sizeof(space
));
4187 space_args
.total_spaces
++;
4190 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4191 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4193 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4198 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4205 * there are many ways the trans_start and trans_end ioctls can lead
4206 * to deadlocks. They should only be used by applications that
4207 * basically own the machine, and have a very in depth understanding
4208 * of all the possible deadlocks and enospc problems.
4210 long btrfs_ioctl_trans_end(struct file
*file
)
4212 struct inode
*inode
= file_inode(file
);
4213 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4214 struct btrfs_trans_handle
*trans
;
4216 trans
= file
->private_data
;
4219 file
->private_data
= NULL
;
4221 btrfs_end_transaction(trans
, root
);
4223 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4225 mnt_drop_write_file(file
);
4229 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4232 struct btrfs_trans_handle
*trans
;
4236 trans
= btrfs_attach_transaction_barrier(root
);
4237 if (IS_ERR(trans
)) {
4238 if (PTR_ERR(trans
) != -ENOENT
)
4239 return PTR_ERR(trans
);
4241 /* No running transaction, don't bother */
4242 transid
= root
->fs_info
->last_trans_committed
;
4245 transid
= trans
->transid
;
4246 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
4248 btrfs_end_transaction(trans
, root
);
4253 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4258 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
4264 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4267 transid
= 0; /* current trans */
4269 return btrfs_wait_for_commit(root
, transid
);
4272 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4274 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4275 struct btrfs_ioctl_scrub_args
*sa
;
4278 if (!capable(CAP_SYS_ADMIN
))
4281 sa
= memdup_user(arg
, sizeof(*sa
));
4285 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4286 ret
= mnt_want_write_file(file
);
4291 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4292 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4295 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4298 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4299 mnt_drop_write_file(file
);
4305 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
4307 if (!capable(CAP_SYS_ADMIN
))
4310 return btrfs_scrub_cancel(root
->fs_info
);
4313 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
4316 struct btrfs_ioctl_scrub_args
*sa
;
4319 if (!capable(CAP_SYS_ADMIN
))
4322 sa
= memdup_user(arg
, sizeof(*sa
));
4326 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
4328 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4335 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
4338 struct btrfs_ioctl_get_dev_stats
*sa
;
4341 sa
= memdup_user(arg
, sizeof(*sa
));
4345 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4350 ret
= btrfs_get_dev_stats(root
, sa
);
4352 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4359 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
4361 struct btrfs_ioctl_dev_replace_args
*p
;
4364 if (!capable(CAP_SYS_ADMIN
))
4367 p
= memdup_user(arg
, sizeof(*p
));
4372 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4373 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
) {
4378 &root
->fs_info
->mutually_exclusive_operation_running
,
4380 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4382 ret
= btrfs_dev_replace_start(root
, p
);
4384 &root
->fs_info
->mutually_exclusive_operation_running
,
4388 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4389 btrfs_dev_replace_status(root
->fs_info
, p
);
4392 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4393 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
4400 if (copy_to_user(arg
, p
, sizeof(*p
)))
4407 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4413 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4414 struct inode_fs_paths
*ipath
= NULL
;
4415 struct btrfs_path
*path
;
4417 if (!capable(CAP_DAC_READ_SEARCH
))
4420 path
= btrfs_alloc_path();
4426 ipa
= memdup_user(arg
, sizeof(*ipa
));
4433 size
= min_t(u32
, ipa
->size
, 4096);
4434 ipath
= init_ipath(size
, root
, path
);
4435 if (IS_ERR(ipath
)) {
4436 ret
= PTR_ERR(ipath
);
4441 ret
= paths_from_inode(ipa
->inum
, ipath
);
4445 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4446 rel_ptr
= ipath
->fspath
->val
[i
] -
4447 (u64
)(unsigned long)ipath
->fspath
->val
;
4448 ipath
->fspath
->val
[i
] = rel_ptr
;
4451 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
4452 (void *)(unsigned long)ipath
->fspath
, size
);
4459 btrfs_free_path(path
);
4466 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4468 struct btrfs_data_container
*inodes
= ctx
;
4469 const size_t c
= 3 * sizeof(u64
);
4471 if (inodes
->bytes_left
>= c
) {
4472 inodes
->bytes_left
-= c
;
4473 inodes
->val
[inodes
->elem_cnt
] = inum
;
4474 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4475 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4476 inodes
->elem_cnt
+= 3;
4478 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4479 inodes
->bytes_left
= 0;
4480 inodes
->elem_missed
+= 3;
4486 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
4491 struct btrfs_ioctl_logical_ino_args
*loi
;
4492 struct btrfs_data_container
*inodes
= NULL
;
4493 struct btrfs_path
*path
= NULL
;
4495 if (!capable(CAP_SYS_ADMIN
))
4498 loi
= memdup_user(arg
, sizeof(*loi
));
4505 path
= btrfs_alloc_path();
4511 size
= min_t(u32
, loi
->size
, SZ_64K
);
4512 inodes
= init_data_container(size
);
4513 if (IS_ERR(inodes
)) {
4514 ret
= PTR_ERR(inodes
);
4519 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
4520 build_ino_list
, inodes
);
4526 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
4527 (void *)(unsigned long)inodes
, size
);
4532 btrfs_free_path(path
);
4539 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
4540 struct btrfs_ioctl_balance_args
*bargs
)
4542 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4544 bargs
->flags
= bctl
->flags
;
4546 if (atomic_read(&fs_info
->balance_running
))
4547 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4548 if (atomic_read(&fs_info
->balance_pause_req
))
4549 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4550 if (atomic_read(&fs_info
->balance_cancel_req
))
4551 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4553 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4554 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4555 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4558 spin_lock(&fs_info
->balance_lock
);
4559 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4560 spin_unlock(&fs_info
->balance_lock
);
4562 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4566 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4568 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4569 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4570 struct btrfs_ioctl_balance_args
*bargs
;
4571 struct btrfs_balance_control
*bctl
;
4572 bool need_unlock
; /* for mut. excl. ops lock */
4575 if (!capable(CAP_SYS_ADMIN
))
4578 ret
= mnt_want_write_file(file
);
4583 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4584 mutex_lock(&fs_info
->volume_mutex
);
4585 mutex_lock(&fs_info
->balance_mutex
);
4591 * mut. excl. ops lock is locked. Three possibilites:
4592 * (1) some other op is running
4593 * (2) balance is running
4594 * (3) balance is paused -- special case (think resume)
4596 mutex_lock(&fs_info
->balance_mutex
);
4597 if (fs_info
->balance_ctl
) {
4598 /* this is either (2) or (3) */
4599 if (!atomic_read(&fs_info
->balance_running
)) {
4600 mutex_unlock(&fs_info
->balance_mutex
);
4601 if (!mutex_trylock(&fs_info
->volume_mutex
))
4603 mutex_lock(&fs_info
->balance_mutex
);
4605 if (fs_info
->balance_ctl
&&
4606 !atomic_read(&fs_info
->balance_running
)) {
4608 need_unlock
= false;
4612 mutex_unlock(&fs_info
->balance_mutex
);
4613 mutex_unlock(&fs_info
->volume_mutex
);
4617 mutex_unlock(&fs_info
->balance_mutex
);
4623 mutex_unlock(&fs_info
->balance_mutex
);
4624 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4629 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4632 bargs
= memdup_user(arg
, sizeof(*bargs
));
4633 if (IS_ERR(bargs
)) {
4634 ret
= PTR_ERR(bargs
);
4638 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4639 if (!fs_info
->balance_ctl
) {
4644 bctl
= fs_info
->balance_ctl
;
4645 spin_lock(&fs_info
->balance_lock
);
4646 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4647 spin_unlock(&fs_info
->balance_lock
);
4655 if (fs_info
->balance_ctl
) {
4660 bctl
= kzalloc(sizeof(*bctl
), GFP_KERNEL
);
4666 bctl
->fs_info
= fs_info
;
4668 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4669 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4670 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4672 bctl
->flags
= bargs
->flags
;
4674 /* balance everything - no filters */
4675 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4678 if (bctl
->flags
& ~(BTRFS_BALANCE_ARGS_MASK
| BTRFS_BALANCE_TYPE_MASK
)) {
4685 * Ownership of bctl and mutually_exclusive_operation_running
4686 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4687 * or, if restriper was paused all the way until unmount, in
4688 * free_fs_info. mutually_exclusive_operation_running is
4689 * cleared in __cancel_balance.
4691 need_unlock
= false;
4693 ret
= btrfs_balance(bctl
, bargs
);
4697 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4706 mutex_unlock(&fs_info
->balance_mutex
);
4707 mutex_unlock(&fs_info
->volume_mutex
);
4709 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4711 mnt_drop_write_file(file
);
4715 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4717 if (!capable(CAP_SYS_ADMIN
))
4721 case BTRFS_BALANCE_CTL_PAUSE
:
4722 return btrfs_pause_balance(root
->fs_info
);
4723 case BTRFS_BALANCE_CTL_CANCEL
:
4724 return btrfs_cancel_balance(root
->fs_info
);
4730 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4733 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4734 struct btrfs_ioctl_balance_args
*bargs
;
4737 if (!capable(CAP_SYS_ADMIN
))
4740 mutex_lock(&fs_info
->balance_mutex
);
4741 if (!fs_info
->balance_ctl
) {
4746 bargs
= kzalloc(sizeof(*bargs
), GFP_KERNEL
);
4752 update_ioctl_balance_args(fs_info
, 1, bargs
);
4754 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4759 mutex_unlock(&fs_info
->balance_mutex
);
4763 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4765 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4766 struct btrfs_ioctl_quota_ctl_args
*sa
;
4767 struct btrfs_trans_handle
*trans
= NULL
;
4771 if (!capable(CAP_SYS_ADMIN
))
4774 ret
= mnt_want_write_file(file
);
4778 sa
= memdup_user(arg
, sizeof(*sa
));
4784 down_write(&root
->fs_info
->subvol_sem
);
4785 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
4786 if (IS_ERR(trans
)) {
4787 ret
= PTR_ERR(trans
);
4792 case BTRFS_QUOTA_CTL_ENABLE
:
4793 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
4795 case BTRFS_QUOTA_CTL_DISABLE
:
4796 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
4803 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
4808 up_write(&root
->fs_info
->subvol_sem
);
4810 mnt_drop_write_file(file
);
4814 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4816 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4817 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4818 struct btrfs_trans_handle
*trans
;
4822 if (!capable(CAP_SYS_ADMIN
))
4825 ret
= mnt_want_write_file(file
);
4829 sa
= memdup_user(arg
, sizeof(*sa
));
4835 trans
= btrfs_join_transaction(root
);
4836 if (IS_ERR(trans
)) {
4837 ret
= PTR_ERR(trans
);
4841 /* FIXME: check if the IDs really exist */
4843 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
4846 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
4850 /* update qgroup status and info */
4851 err
= btrfs_run_qgroups(trans
, root
->fs_info
);
4853 btrfs_std_error(root
->fs_info
, ret
,
4854 "failed to update qgroup status and info\n");
4855 err
= btrfs_end_transaction(trans
, root
);
4862 mnt_drop_write_file(file
);
4866 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4868 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4869 struct btrfs_ioctl_qgroup_create_args
*sa
;
4870 struct btrfs_trans_handle
*trans
;
4874 if (!capable(CAP_SYS_ADMIN
))
4877 ret
= mnt_want_write_file(file
);
4881 sa
= memdup_user(arg
, sizeof(*sa
));
4887 if (!sa
->qgroupid
) {
4892 trans
= btrfs_join_transaction(root
);
4893 if (IS_ERR(trans
)) {
4894 ret
= PTR_ERR(trans
);
4898 /* FIXME: check if the IDs really exist */
4900 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4902 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4905 err
= btrfs_end_transaction(trans
, root
);
4912 mnt_drop_write_file(file
);
4916 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4918 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4919 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4920 struct btrfs_trans_handle
*trans
;
4925 if (!capable(CAP_SYS_ADMIN
))
4928 ret
= mnt_want_write_file(file
);
4932 sa
= memdup_user(arg
, sizeof(*sa
));
4938 trans
= btrfs_join_transaction(root
);
4939 if (IS_ERR(trans
)) {
4940 ret
= PTR_ERR(trans
);
4944 qgroupid
= sa
->qgroupid
;
4946 /* take the current subvol as qgroup */
4947 qgroupid
= root
->root_key
.objectid
;
4950 /* FIXME: check if the IDs really exist */
4951 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
4953 err
= btrfs_end_transaction(trans
, root
);
4960 mnt_drop_write_file(file
);
4964 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
4966 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4967 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4970 if (!capable(CAP_SYS_ADMIN
))
4973 ret
= mnt_want_write_file(file
);
4977 qsa
= memdup_user(arg
, sizeof(*qsa
));
4988 ret
= btrfs_qgroup_rescan(root
->fs_info
);
4993 mnt_drop_write_file(file
);
4997 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
4999 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5000 struct btrfs_ioctl_quota_rescan_args
*qsa
;
5003 if (!capable(CAP_SYS_ADMIN
))
5006 qsa
= kzalloc(sizeof(*qsa
), GFP_KERNEL
);
5010 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
5012 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
5015 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
5022 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
5024 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5026 if (!capable(CAP_SYS_ADMIN
))
5029 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
5032 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
5033 struct btrfs_ioctl_received_subvol_args
*sa
)
5035 struct inode
*inode
= file_inode(file
);
5036 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5037 struct btrfs_root_item
*root_item
= &root
->root_item
;
5038 struct btrfs_trans_handle
*trans
;
5039 struct timespec ct
= current_fs_time(inode
->i_sb
);
5041 int received_uuid_changed
;
5043 if (!inode_owner_or_capable(inode
))
5046 ret
= mnt_want_write_file(file
);
5050 down_write(&root
->fs_info
->subvol_sem
);
5052 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
5057 if (btrfs_root_readonly(root
)) {
5064 * 2 - uuid items (received uuid + subvol uuid)
5066 trans
= btrfs_start_transaction(root
, 3);
5067 if (IS_ERR(trans
)) {
5068 ret
= PTR_ERR(trans
);
5073 sa
->rtransid
= trans
->transid
;
5074 sa
->rtime
.sec
= ct
.tv_sec
;
5075 sa
->rtime
.nsec
= ct
.tv_nsec
;
5077 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
5079 if (received_uuid_changed
&&
5080 !btrfs_is_empty_uuid(root_item
->received_uuid
))
5081 btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
5082 root_item
->received_uuid
,
5083 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5084 root
->root_key
.objectid
);
5085 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
5086 btrfs_set_root_stransid(root_item
, sa
->stransid
);
5087 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
5088 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
5089 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
5090 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
5091 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
5093 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
5094 &root
->root_key
, &root
->root_item
);
5096 btrfs_end_transaction(trans
, root
);
5099 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
5100 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
5102 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5103 root
->root_key
.objectid
);
5104 if (ret
< 0 && ret
!= -EEXIST
) {
5105 btrfs_abort_transaction(trans
, root
, ret
);
5109 ret
= btrfs_commit_transaction(trans
, root
);
5111 btrfs_abort_transaction(trans
, root
, ret
);
5116 up_write(&root
->fs_info
->subvol_sem
);
5117 mnt_drop_write_file(file
);
5122 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
5125 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
5126 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
5129 args32
= memdup_user(arg
, sizeof(*args32
));
5130 if (IS_ERR(args32
)) {
5131 ret
= PTR_ERR(args32
);
5136 args64
= kmalloc(sizeof(*args64
), GFP_KERNEL
);
5142 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
5143 args64
->stransid
= args32
->stransid
;
5144 args64
->rtransid
= args32
->rtransid
;
5145 args64
->stime
.sec
= args32
->stime
.sec
;
5146 args64
->stime
.nsec
= args32
->stime
.nsec
;
5147 args64
->rtime
.sec
= args32
->rtime
.sec
;
5148 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
5149 args64
->flags
= args32
->flags
;
5151 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
5155 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
5156 args32
->stransid
= args64
->stransid
;
5157 args32
->rtransid
= args64
->rtransid
;
5158 args32
->stime
.sec
= args64
->stime
.sec
;
5159 args32
->stime
.nsec
= args64
->stime
.nsec
;
5160 args32
->rtime
.sec
= args64
->rtime
.sec
;
5161 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
5162 args32
->flags
= args64
->flags
;
5164 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
5175 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
5178 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
5181 sa
= memdup_user(arg
, sizeof(*sa
));
5188 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5193 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5202 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5204 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5207 char label
[BTRFS_LABEL_SIZE
];
5209 spin_lock(&root
->fs_info
->super_lock
);
5210 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5211 spin_unlock(&root
->fs_info
->super_lock
);
5213 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5215 if (len
== BTRFS_LABEL_SIZE
) {
5216 btrfs_warn(root
->fs_info
,
5217 "label is too long, return the first %zu bytes", --len
);
5220 ret
= copy_to_user(arg
, label
, len
);
5222 return ret
? -EFAULT
: 0;
5225 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5227 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5228 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5229 struct btrfs_trans_handle
*trans
;
5230 char label
[BTRFS_LABEL_SIZE
];
5233 if (!capable(CAP_SYS_ADMIN
))
5236 if (copy_from_user(label
, arg
, sizeof(label
)))
5239 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5240 btrfs_err(root
->fs_info
, "unable to set label with more than %d bytes",
5241 BTRFS_LABEL_SIZE
- 1);
5245 ret
= mnt_want_write_file(file
);
5249 trans
= btrfs_start_transaction(root
, 0);
5250 if (IS_ERR(trans
)) {
5251 ret
= PTR_ERR(trans
);
5255 spin_lock(&root
->fs_info
->super_lock
);
5256 strcpy(super_block
->label
, label
);
5257 spin_unlock(&root
->fs_info
->super_lock
);
5258 ret
= btrfs_commit_transaction(trans
, root
);
5261 mnt_drop_write_file(file
);
5265 #define INIT_FEATURE_FLAGS(suffix) \
5266 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5267 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5268 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5270 int btrfs_ioctl_get_supported_features(void __user
*arg
)
5272 static const struct btrfs_ioctl_feature_flags features
[3] = {
5273 INIT_FEATURE_FLAGS(SUPP
),
5274 INIT_FEATURE_FLAGS(SAFE_SET
),
5275 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5278 if (copy_to_user(arg
, &features
, sizeof(features
)))
5284 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5286 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5287 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5288 struct btrfs_ioctl_feature_flags features
;
5290 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5291 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5292 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5294 if (copy_to_user(arg
, &features
, sizeof(features
)))
5300 static int check_feature_bits(struct btrfs_root
*root
,
5301 enum btrfs_feature_set set
,
5302 u64 change_mask
, u64 flags
, u64 supported_flags
,
5303 u64 safe_set
, u64 safe_clear
)
5305 const char *type
= btrfs_feature_set_names
[set
];
5307 u64 disallowed
, unsupported
;
5308 u64 set_mask
= flags
& change_mask
;
5309 u64 clear_mask
= ~flags
& change_mask
;
5311 unsupported
= set_mask
& ~supported_flags
;
5313 names
= btrfs_printable_features(set
, unsupported
);
5315 btrfs_warn(root
->fs_info
,
5316 "this kernel does not support the %s feature bit%s",
5317 names
, strchr(names
, ',') ? "s" : "");
5320 btrfs_warn(root
->fs_info
,
5321 "this kernel does not support %s bits 0x%llx",
5326 disallowed
= set_mask
& ~safe_set
;
5328 names
= btrfs_printable_features(set
, disallowed
);
5330 btrfs_warn(root
->fs_info
,
5331 "can't set the %s feature bit%s while mounted",
5332 names
, strchr(names
, ',') ? "s" : "");
5335 btrfs_warn(root
->fs_info
,
5336 "can't set %s bits 0x%llx while mounted",
5341 disallowed
= clear_mask
& ~safe_clear
;
5343 names
= btrfs_printable_features(set
, disallowed
);
5345 btrfs_warn(root
->fs_info
,
5346 "can't clear the %s feature bit%s while mounted",
5347 names
, strchr(names
, ',') ? "s" : "");
5350 btrfs_warn(root
->fs_info
,
5351 "can't clear %s bits 0x%llx while mounted",
5359 #define check_feature(root, change_mask, flags, mask_base) \
5360 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5361 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5362 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5363 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5365 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5367 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5368 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5369 struct btrfs_ioctl_feature_flags flags
[2];
5370 struct btrfs_trans_handle
*trans
;
5374 if (!capable(CAP_SYS_ADMIN
))
5377 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5381 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5382 !flags
[0].incompat_flags
)
5385 ret
= check_feature(root
, flags
[0].compat_flags
,
5386 flags
[1].compat_flags
, COMPAT
);
5390 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
5391 flags
[1].compat_ro_flags
, COMPAT_RO
);
5395 ret
= check_feature(root
, flags
[0].incompat_flags
,
5396 flags
[1].incompat_flags
, INCOMPAT
);
5400 trans
= btrfs_start_transaction(root
, 0);
5402 return PTR_ERR(trans
);
5404 spin_lock(&root
->fs_info
->super_lock
);
5405 newflags
= btrfs_super_compat_flags(super_block
);
5406 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5407 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5408 btrfs_set_super_compat_flags(super_block
, newflags
);
5410 newflags
= btrfs_super_compat_ro_flags(super_block
);
5411 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5412 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5413 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5415 newflags
= btrfs_super_incompat_flags(super_block
);
5416 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5417 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5418 btrfs_set_super_incompat_flags(super_block
, newflags
);
5419 spin_unlock(&root
->fs_info
->super_lock
);
5421 return btrfs_commit_transaction(trans
, root
);
5424 long btrfs_ioctl(struct file
*file
, unsigned int
5425 cmd
, unsigned long arg
)
5427 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5428 void __user
*argp
= (void __user
*)arg
;
5431 case FS_IOC_GETFLAGS
:
5432 return btrfs_ioctl_getflags(file
, argp
);
5433 case FS_IOC_SETFLAGS
:
5434 return btrfs_ioctl_setflags(file
, argp
);
5435 case FS_IOC_GETVERSION
:
5436 return btrfs_ioctl_getversion(file
, argp
);
5438 return btrfs_ioctl_fitrim(file
, argp
);
5439 case BTRFS_IOC_SNAP_CREATE
:
5440 return btrfs_ioctl_snap_create(file
, argp
, 0);
5441 case BTRFS_IOC_SNAP_CREATE_V2
:
5442 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5443 case BTRFS_IOC_SUBVOL_CREATE
:
5444 return btrfs_ioctl_snap_create(file
, argp
, 1);
5445 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5446 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5447 case BTRFS_IOC_SNAP_DESTROY
:
5448 return btrfs_ioctl_snap_destroy(file
, argp
);
5449 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5450 return btrfs_ioctl_subvol_getflags(file
, argp
);
5451 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5452 return btrfs_ioctl_subvol_setflags(file
, argp
);
5453 case BTRFS_IOC_DEFAULT_SUBVOL
:
5454 return btrfs_ioctl_default_subvol(file
, argp
);
5455 case BTRFS_IOC_DEFRAG
:
5456 return btrfs_ioctl_defrag(file
, NULL
);
5457 case BTRFS_IOC_DEFRAG_RANGE
:
5458 return btrfs_ioctl_defrag(file
, argp
);
5459 case BTRFS_IOC_RESIZE
:
5460 return btrfs_ioctl_resize(file
, argp
);
5461 case BTRFS_IOC_ADD_DEV
:
5462 return btrfs_ioctl_add_dev(root
, argp
);
5463 case BTRFS_IOC_RM_DEV
:
5464 return btrfs_ioctl_rm_dev(file
, argp
);
5465 case BTRFS_IOC_FS_INFO
:
5466 return btrfs_ioctl_fs_info(root
, argp
);
5467 case BTRFS_IOC_DEV_INFO
:
5468 return btrfs_ioctl_dev_info(root
, argp
);
5469 case BTRFS_IOC_BALANCE
:
5470 return btrfs_ioctl_balance(file
, NULL
);
5471 case BTRFS_IOC_TRANS_START
:
5472 return btrfs_ioctl_trans_start(file
);
5473 case BTRFS_IOC_TRANS_END
:
5474 return btrfs_ioctl_trans_end(file
);
5475 case BTRFS_IOC_TREE_SEARCH
:
5476 return btrfs_ioctl_tree_search(file
, argp
);
5477 case BTRFS_IOC_TREE_SEARCH_V2
:
5478 return btrfs_ioctl_tree_search_v2(file
, argp
);
5479 case BTRFS_IOC_INO_LOOKUP
:
5480 return btrfs_ioctl_ino_lookup(file
, argp
);
5481 case BTRFS_IOC_INO_PATHS
:
5482 return btrfs_ioctl_ino_to_path(root
, argp
);
5483 case BTRFS_IOC_LOGICAL_INO
:
5484 return btrfs_ioctl_logical_to_ino(root
, argp
);
5485 case BTRFS_IOC_SPACE_INFO
:
5486 return btrfs_ioctl_space_info(root
, argp
);
5487 case BTRFS_IOC_SYNC
: {
5490 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
5493 ret
= btrfs_sync_fs(file_inode(file
)->i_sb
, 1);
5495 * The transaction thread may want to do more work,
5496 * namely it pokes the cleaner ktread that will start
5497 * processing uncleaned subvols.
5499 wake_up_process(root
->fs_info
->transaction_kthread
);
5502 case BTRFS_IOC_START_SYNC
:
5503 return btrfs_ioctl_start_sync(root
, argp
);
5504 case BTRFS_IOC_WAIT_SYNC
:
5505 return btrfs_ioctl_wait_sync(root
, argp
);
5506 case BTRFS_IOC_SCRUB
:
5507 return btrfs_ioctl_scrub(file
, argp
);
5508 case BTRFS_IOC_SCRUB_CANCEL
:
5509 return btrfs_ioctl_scrub_cancel(root
, argp
);
5510 case BTRFS_IOC_SCRUB_PROGRESS
:
5511 return btrfs_ioctl_scrub_progress(root
, argp
);
5512 case BTRFS_IOC_BALANCE_V2
:
5513 return btrfs_ioctl_balance(file
, argp
);
5514 case BTRFS_IOC_BALANCE_CTL
:
5515 return btrfs_ioctl_balance_ctl(root
, arg
);
5516 case BTRFS_IOC_BALANCE_PROGRESS
:
5517 return btrfs_ioctl_balance_progress(root
, argp
);
5518 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5519 return btrfs_ioctl_set_received_subvol(file
, argp
);
5521 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5522 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5524 case BTRFS_IOC_SEND
:
5525 return btrfs_ioctl_send(file
, argp
);
5526 case BTRFS_IOC_GET_DEV_STATS
:
5527 return btrfs_ioctl_get_dev_stats(root
, argp
);
5528 case BTRFS_IOC_QUOTA_CTL
:
5529 return btrfs_ioctl_quota_ctl(file
, argp
);
5530 case BTRFS_IOC_QGROUP_ASSIGN
:
5531 return btrfs_ioctl_qgroup_assign(file
, argp
);
5532 case BTRFS_IOC_QGROUP_CREATE
:
5533 return btrfs_ioctl_qgroup_create(file
, argp
);
5534 case BTRFS_IOC_QGROUP_LIMIT
:
5535 return btrfs_ioctl_qgroup_limit(file
, argp
);
5536 case BTRFS_IOC_QUOTA_RESCAN
:
5537 return btrfs_ioctl_quota_rescan(file
, argp
);
5538 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5539 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5540 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5541 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5542 case BTRFS_IOC_DEV_REPLACE
:
5543 return btrfs_ioctl_dev_replace(root
, argp
);
5544 case BTRFS_IOC_GET_FSLABEL
:
5545 return btrfs_ioctl_get_fslabel(file
, argp
);
5546 case BTRFS_IOC_SET_FSLABEL
:
5547 return btrfs_ioctl_set_fslabel(file
, argp
);
5548 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5549 return btrfs_ioctl_get_supported_features(argp
);
5550 case BTRFS_IOC_GET_FEATURES
:
5551 return btrfs_ioctl_get_features(file
, argp
);
5552 case BTRFS_IOC_SET_FEATURES
:
5553 return btrfs_ioctl_set_features(file
, argp
);
5559 #ifdef CONFIG_COMPAT
5560 long btrfs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
5563 case FS_IOC32_GETFLAGS
:
5564 cmd
= FS_IOC_GETFLAGS
;
5566 case FS_IOC32_SETFLAGS
:
5567 cmd
= FS_IOC_SETFLAGS
;
5569 case FS_IOC32_GETVERSION
:
5570 cmd
= FS_IOC_GETVERSION
;
5573 return -ENOIOCTLCMD
;
5576 return btrfs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));