2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
69 struct btrfs_ioctl_timespec_32
{
72 } __attribute__ ((__packed__
));
74 struct btrfs_ioctl_received_subvol_args_32
{
75 char uuid
[BTRFS_UUID_SIZE
]; /* in */
76 __u64 stransid
; /* in */
77 __u64 rtransid
; /* out */
78 struct btrfs_ioctl_timespec_32 stime
; /* in */
79 struct btrfs_ioctl_timespec_32 rtime
; /* out */
81 __u64 reserved
[16]; /* in */
82 } __attribute__ ((__packed__
));
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
89 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
90 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
);
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
97 else if (S_ISREG(mode
))
98 return flags
& ~FS_DIRSYNC_FL
;
100 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
108 unsigned int iflags
= 0;
110 if (flags
& BTRFS_INODE_SYNC
)
111 iflags
|= FS_SYNC_FL
;
112 if (flags
& BTRFS_INODE_IMMUTABLE
)
113 iflags
|= FS_IMMUTABLE_FL
;
114 if (flags
& BTRFS_INODE_APPEND
)
115 iflags
|= FS_APPEND_FL
;
116 if (flags
& BTRFS_INODE_NODUMP
)
117 iflags
|= FS_NODUMP_FL
;
118 if (flags
& BTRFS_INODE_NOATIME
)
119 iflags
|= FS_NOATIME_FL
;
120 if (flags
& BTRFS_INODE_DIRSYNC
)
121 iflags
|= FS_DIRSYNC_FL
;
122 if (flags
& BTRFS_INODE_NODATACOW
)
123 iflags
|= FS_NOCOW_FL
;
125 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
126 iflags
|= FS_COMPR_FL
;
127 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
128 iflags
|= FS_NOCOMP_FL
;
134 * Update inode->i_flags based on the btrfs internal flags.
136 void btrfs_update_iflags(struct inode
*inode
)
138 struct btrfs_inode
*ip
= BTRFS_I(inode
);
139 unsigned int new_fl
= 0;
141 if (ip
->flags
& BTRFS_INODE_SYNC
)
143 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
144 new_fl
|= S_IMMUTABLE
;
145 if (ip
->flags
& BTRFS_INODE_APPEND
)
147 if (ip
->flags
& BTRFS_INODE_NOATIME
)
149 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
152 set_mask_bits(&inode
->i_flags
,
153 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
,
158 * Inherit flags from the parent inode.
160 * Currently only the compression flags and the cow flags are inherited.
162 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
169 flags
= BTRFS_I(dir
)->flags
;
171 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
172 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
173 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
174 } else if (flags
& BTRFS_INODE_COMPRESS
) {
175 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
176 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
179 if (flags
& BTRFS_INODE_NODATACOW
) {
180 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
181 if (S_ISREG(inode
->i_mode
))
182 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
185 btrfs_update_iflags(inode
);
188 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
190 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
191 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
193 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
198 static int check_flags(unsigned int flags
)
200 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
201 FS_NOATIME_FL
| FS_NODUMP_FL
| \
202 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
203 FS_NOCOMP_FL
| FS_COMPR_FL
|
207 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
213 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
215 struct inode
*inode
= file_inode(file
);
216 struct btrfs_inode
*ip
= BTRFS_I(inode
);
217 struct btrfs_root
*root
= ip
->root
;
218 struct btrfs_trans_handle
*trans
;
219 unsigned int flags
, oldflags
;
222 unsigned int i_oldflags
;
225 if (!inode_owner_or_capable(inode
))
228 if (btrfs_root_readonly(root
))
231 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
234 ret
= check_flags(flags
);
238 ret
= mnt_want_write_file(file
);
242 mutex_lock(&inode
->i_mutex
);
244 ip_oldflags
= ip
->flags
;
245 i_oldflags
= inode
->i_flags
;
246 mode
= inode
->i_mode
;
248 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
249 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
250 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
251 if (!capable(CAP_LINUX_IMMUTABLE
)) {
257 if (flags
& FS_SYNC_FL
)
258 ip
->flags
|= BTRFS_INODE_SYNC
;
260 ip
->flags
&= ~BTRFS_INODE_SYNC
;
261 if (flags
& FS_IMMUTABLE_FL
)
262 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
264 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
265 if (flags
& FS_APPEND_FL
)
266 ip
->flags
|= BTRFS_INODE_APPEND
;
268 ip
->flags
&= ~BTRFS_INODE_APPEND
;
269 if (flags
& FS_NODUMP_FL
)
270 ip
->flags
|= BTRFS_INODE_NODUMP
;
272 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
273 if (flags
& FS_NOATIME_FL
)
274 ip
->flags
|= BTRFS_INODE_NOATIME
;
276 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
277 if (flags
& FS_DIRSYNC_FL
)
278 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
280 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
281 if (flags
& FS_NOCOW_FL
) {
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
288 if (inode
->i_size
== 0)
289 ip
->flags
|= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM
;
292 ip
->flags
|= BTRFS_INODE_NODATACOW
;
296 * Revert back under same assuptions as above
299 if (inode
->i_size
== 0)
300 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM
);
303 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
312 if (flags
& FS_NOCOMP_FL
) {
313 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
314 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
316 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
317 if (ret
&& ret
!= -ENODATA
)
319 } else if (flags
& FS_COMPR_FL
) {
322 ip
->flags
|= BTRFS_INODE_COMPRESS
;
323 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
325 if (root
->fs_info
->compress_type
== BTRFS_COMPRESS_LZO
)
329 ret
= btrfs_set_prop(inode
, "btrfs.compression",
330 comp
, strlen(comp
), 0);
335 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
338 trans
= btrfs_start_transaction(root
, 1);
340 ret
= PTR_ERR(trans
);
344 btrfs_update_iflags(inode
);
345 inode_inc_iversion(inode
);
346 inode
->i_ctime
= CURRENT_TIME
;
347 ret
= btrfs_update_inode(trans
, root
, inode
);
349 btrfs_end_transaction(trans
, root
);
352 ip
->flags
= ip_oldflags
;
353 inode
->i_flags
= i_oldflags
;
357 mutex_unlock(&inode
->i_mutex
);
358 mnt_drop_write_file(file
);
362 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
364 struct inode
*inode
= file_inode(file
);
366 return put_user(inode
->i_generation
, arg
);
369 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
371 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
372 struct btrfs_device
*device
;
373 struct request_queue
*q
;
374 struct fstrim_range range
;
375 u64 minlen
= ULLONG_MAX
;
377 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
380 if (!capable(CAP_SYS_ADMIN
))
384 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
388 q
= bdev_get_queue(device
->bdev
);
389 if (blk_queue_discard(q
)) {
391 minlen
= min((u64
)q
->limits
.discard_granularity
,
399 if (copy_from_user(&range
, arg
, sizeof(range
)))
401 if (range
.start
> total_bytes
||
402 range
.len
< fs_info
->sb
->s_blocksize
)
405 range
.len
= min(range
.len
, total_bytes
- range
.start
);
406 range
.minlen
= max(range
.minlen
, minlen
);
407 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
411 if (copy_to_user(arg
, &range
, sizeof(range
)))
417 int btrfs_is_empty_uuid(u8
*uuid
)
421 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
428 static noinline
int create_subvol(struct inode
*dir
,
429 struct dentry
*dentry
,
430 char *name
, int namelen
,
432 struct btrfs_qgroup_inherit
*inherit
)
434 struct btrfs_trans_handle
*trans
;
435 struct btrfs_key key
;
436 struct btrfs_root_item root_item
;
437 struct btrfs_inode_item
*inode_item
;
438 struct extent_buffer
*leaf
;
439 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
440 struct btrfs_root
*new_root
;
441 struct btrfs_block_rsv block_rsv
;
442 struct timespec cur_time
= CURRENT_TIME
;
447 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
452 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
456 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
458 * The same as the snapshot creation, please see the comment
459 * of create_snapshot().
461 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
462 8, &qgroup_reserved
, false);
466 trans
= btrfs_start_transaction(root
, 0);
468 ret
= PTR_ERR(trans
);
469 btrfs_subvolume_release_metadata(root
, &block_rsv
,
473 trans
->block_rsv
= &block_rsv
;
474 trans
->bytes_reserved
= block_rsv
.size
;
476 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
480 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
481 0, objectid
, NULL
, 0, 0, 0);
487 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
488 btrfs_set_header_bytenr(leaf
, leaf
->start
);
489 btrfs_set_header_generation(leaf
, trans
->transid
);
490 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
491 btrfs_set_header_owner(leaf
, objectid
);
493 write_extent_buffer(leaf
, root
->fs_info
->fsid
, btrfs_header_fsid(),
495 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
496 btrfs_header_chunk_tree_uuid(leaf
),
498 btrfs_mark_buffer_dirty(leaf
);
500 memset(&root_item
, 0, sizeof(root_item
));
502 inode_item
= &root_item
.inode
;
503 btrfs_set_stack_inode_generation(inode_item
, 1);
504 btrfs_set_stack_inode_size(inode_item
, 3);
505 btrfs_set_stack_inode_nlink(inode_item
, 1);
506 btrfs_set_stack_inode_nbytes(inode_item
, root
->leafsize
);
507 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
509 btrfs_set_root_flags(&root_item
, 0);
510 btrfs_set_root_limit(&root_item
, 0);
511 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
513 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
514 btrfs_set_root_generation(&root_item
, trans
->transid
);
515 btrfs_set_root_level(&root_item
, 0);
516 btrfs_set_root_refs(&root_item
, 1);
517 btrfs_set_root_used(&root_item
, leaf
->len
);
518 btrfs_set_root_last_snapshot(&root_item
, 0);
520 btrfs_set_root_generation_v2(&root_item
,
521 btrfs_root_generation(&root_item
));
522 uuid_le_gen(&new_uuid
);
523 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
524 btrfs_set_stack_timespec_sec(&root_item
.otime
, cur_time
.tv_sec
);
525 btrfs_set_stack_timespec_nsec(&root_item
.otime
, cur_time
.tv_nsec
);
526 root_item
.ctime
= root_item
.otime
;
527 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
528 btrfs_set_root_otransid(&root_item
, trans
->transid
);
530 btrfs_tree_unlock(leaf
);
531 free_extent_buffer(leaf
);
534 btrfs_set_root_dirid(&root_item
, new_dirid
);
536 key
.objectid
= objectid
;
538 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
539 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
544 key
.offset
= (u64
)-1;
545 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
546 if (IS_ERR(new_root
)) {
547 btrfs_abort_transaction(trans
, root
, PTR_ERR(new_root
));
548 ret
= PTR_ERR(new_root
);
552 btrfs_record_root_in_trans(trans
, new_root
);
554 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
556 /* We potentially lose an unused inode item here */
557 btrfs_abort_transaction(trans
, root
, ret
);
562 * insert the directory item
564 ret
= btrfs_set_inode_index(dir
, &index
);
566 btrfs_abort_transaction(trans
, root
, ret
);
570 ret
= btrfs_insert_dir_item(trans
, root
,
571 name
, namelen
, dir
, &key
,
572 BTRFS_FT_DIR
, index
);
574 btrfs_abort_transaction(trans
, root
, ret
);
578 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
579 ret
= btrfs_update_inode(trans
, root
, dir
);
582 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
583 objectid
, root
->root_key
.objectid
,
584 btrfs_ino(dir
), index
, name
, namelen
);
587 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
588 root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
591 btrfs_abort_transaction(trans
, root
, ret
);
594 trans
->block_rsv
= NULL
;
595 trans
->bytes_reserved
= 0;
596 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
599 *async_transid
= trans
->transid
;
600 err
= btrfs_commit_transaction_async(trans
, root
, 1);
602 err
= btrfs_commit_transaction(trans
, root
);
604 err
= btrfs_commit_transaction(trans
, root
);
610 inode
= btrfs_lookup_dentry(dir
, dentry
);
612 return PTR_ERR(inode
);
613 d_instantiate(dentry
, inode
);
618 static void btrfs_wait_nocow_write(struct btrfs_root
*root
)
624 prepare_to_wait(&root
->subv_writers
->wait
, &wait
,
625 TASK_UNINTERRUPTIBLE
);
627 writers
= percpu_counter_sum(&root
->subv_writers
->counter
);
631 finish_wait(&root
->subv_writers
->wait
, &wait
);
635 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
636 struct dentry
*dentry
, char *name
, int namelen
,
637 u64
*async_transid
, bool readonly
,
638 struct btrfs_qgroup_inherit
*inherit
)
641 struct btrfs_pending_snapshot
*pending_snapshot
;
642 struct btrfs_trans_handle
*trans
;
645 if (!test_bit(BTRFS_ROOT_REF_COWS
, &root
->state
))
648 atomic_inc(&root
->will_be_snapshoted
);
649 smp_mb__after_atomic();
650 btrfs_wait_nocow_write(root
);
652 ret
= btrfs_start_delalloc_inodes(root
, 0);
656 btrfs_wait_ordered_extents(root
, -1);
658 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
659 if (!pending_snapshot
) {
664 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
665 BTRFS_BLOCK_RSV_TEMP
);
667 * 1 - parent dir inode
670 * 2 - root ref/backref
671 * 1 - root of snapshot
674 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
675 &pending_snapshot
->block_rsv
, 8,
676 &pending_snapshot
->qgroup_reserved
,
681 pending_snapshot
->dentry
= dentry
;
682 pending_snapshot
->root
= root
;
683 pending_snapshot
->readonly
= readonly
;
684 pending_snapshot
->dir
= dir
;
685 pending_snapshot
->inherit
= inherit
;
687 trans
= btrfs_start_transaction(root
, 0);
689 ret
= PTR_ERR(trans
);
693 spin_lock(&root
->fs_info
->trans_lock
);
694 list_add(&pending_snapshot
->list
,
695 &trans
->transaction
->pending_snapshots
);
696 spin_unlock(&root
->fs_info
->trans_lock
);
698 *async_transid
= trans
->transid
;
699 ret
= btrfs_commit_transaction_async(trans
,
700 root
->fs_info
->extent_root
, 1);
702 ret
= btrfs_commit_transaction(trans
, root
);
704 ret
= btrfs_commit_transaction(trans
,
705 root
->fs_info
->extent_root
);
710 ret
= pending_snapshot
->error
;
714 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
719 * If orphan cleanup did remove any orphans, it means the tree was
720 * modified and therefore the commit root is not the same as the
721 * current root anymore. This is a problem, because send uses the
722 * commit root and therefore can see inode items that don't exist
723 * in the current root anymore, and for example make calls to
724 * btrfs_iget, which will do tree lookups based on the current root
725 * and not on the commit root. Those lookups will fail, returning a
726 * -ESTALE error, and making send fail with that error. So make sure
727 * a send does not see any orphans we have just removed, and that it
728 * will see the same inodes regardless of whether a transaction
729 * commit happened before it started (meaning that the commit root
730 * will be the same as the current root) or not.
732 if (readonly
&& pending_snapshot
->snap
->node
!=
733 pending_snapshot
->snap
->commit_root
) {
734 trans
= btrfs_join_transaction(pending_snapshot
->snap
);
735 if (IS_ERR(trans
) && PTR_ERR(trans
) != -ENOENT
) {
736 ret
= PTR_ERR(trans
);
739 if (!IS_ERR(trans
)) {
740 ret
= btrfs_commit_transaction(trans
,
741 pending_snapshot
->snap
);
747 inode
= btrfs_lookup_dentry(dentry
->d_parent
->d_inode
, dentry
);
749 ret
= PTR_ERR(inode
);
753 d_instantiate(dentry
, inode
);
756 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
757 &pending_snapshot
->block_rsv
,
758 pending_snapshot
->qgroup_reserved
);
760 kfree(pending_snapshot
);
762 atomic_dec(&root
->will_be_snapshoted
);
766 /* copy of check_sticky in fs/namei.c()
767 * It's inline, so penalty for filesystems that don't use sticky bit is
770 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
772 kuid_t fsuid
= current_fsuid();
774 if (!(dir
->i_mode
& S_ISVTX
))
776 if (uid_eq(inode
->i_uid
, fsuid
))
778 if (uid_eq(dir
->i_uid
, fsuid
))
780 return !capable(CAP_FOWNER
);
783 /* copy of may_delete in fs/namei.c()
784 * Check whether we can remove a link victim from directory dir, check
785 * whether the type of victim is right.
786 * 1. We can't do it if dir is read-only (done in permission())
787 * 2. We should have write and exec permissions on dir
788 * 3. We can't remove anything from append-only dir
789 * 4. We can't do anything with immutable dir (done in permission())
790 * 5. If the sticky bit on dir is set we should either
791 * a. be owner of dir, or
792 * b. be owner of victim, or
793 * c. have CAP_FOWNER capability
794 * 6. If the victim is append-only or immutable we can't do antyhing with
795 * links pointing to it.
796 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
797 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
798 * 9. We can't remove a root or mountpoint.
799 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
800 * nfs_async_unlink().
803 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
807 if (!victim
->d_inode
)
810 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
811 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
813 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
818 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
819 IS_APPEND(victim
->d_inode
)||
820 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
823 if (!S_ISDIR(victim
->d_inode
->i_mode
))
827 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
831 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
836 /* copy of may_create in fs/namei.c() */
837 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
843 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
847 * Create a new subvolume below @parent. This is largely modeled after
848 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
849 * inside this filesystem so it's quite a bit simpler.
851 static noinline
int btrfs_mksubvol(struct path
*parent
,
852 char *name
, int namelen
,
853 struct btrfs_root
*snap_src
,
854 u64
*async_transid
, bool readonly
,
855 struct btrfs_qgroup_inherit
*inherit
)
857 struct inode
*dir
= parent
->dentry
->d_inode
;
858 struct dentry
*dentry
;
861 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
865 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
866 error
= PTR_ERR(dentry
);
874 error
= btrfs_may_create(dir
, dentry
);
879 * even if this name doesn't exist, we may get hash collisions.
880 * check for them now when we can safely fail
882 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
888 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
890 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
894 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
895 async_transid
, readonly
, inherit
);
897 error
= create_subvol(dir
, dentry
, name
, namelen
,
898 async_transid
, inherit
);
901 fsnotify_mkdir(dir
, dentry
);
903 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
907 mutex_unlock(&dir
->i_mutex
);
912 * When we're defragging a range, we don't want to kick it off again
913 * if it is really just waiting for delalloc to send it down.
914 * If we find a nice big extent or delalloc range for the bytes in the
915 * file you want to defrag, we return 0 to let you know to skip this
918 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, int thresh
)
920 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
921 struct extent_map
*em
= NULL
;
922 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
925 read_lock(&em_tree
->lock
);
926 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
927 read_unlock(&em_tree
->lock
);
930 end
= extent_map_end(em
);
932 if (end
- offset
> thresh
)
935 /* if we already have a nice delalloc here, just stop */
937 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
938 thresh
, EXTENT_DELALLOC
, 1);
945 * helper function to walk through a file and find extents
946 * newer than a specific transid, and smaller than thresh.
948 * This is used by the defragging code to find new and small
951 static int find_new_extents(struct btrfs_root
*root
,
952 struct inode
*inode
, u64 newer_than
,
953 u64
*off
, int thresh
)
955 struct btrfs_path
*path
;
956 struct btrfs_key min_key
;
957 struct extent_buffer
*leaf
;
958 struct btrfs_file_extent_item
*extent
;
961 u64 ino
= btrfs_ino(inode
);
963 path
= btrfs_alloc_path();
967 min_key
.objectid
= ino
;
968 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
969 min_key
.offset
= *off
;
972 path
->keep_locks
= 1;
973 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
976 path
->keep_locks
= 0;
977 btrfs_unlock_up_safe(path
, 1);
979 if (min_key
.objectid
!= ino
)
981 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
984 leaf
= path
->nodes
[0];
985 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
986 struct btrfs_file_extent_item
);
988 type
= btrfs_file_extent_type(leaf
, extent
);
989 if (type
== BTRFS_FILE_EXTENT_REG
&&
990 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
991 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
992 *off
= min_key
.offset
;
993 btrfs_free_path(path
);
998 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
999 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
1003 if (min_key
.offset
== (u64
)-1)
1007 btrfs_release_path(path
);
1010 btrfs_free_path(path
);
1014 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
1016 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
1017 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1018 struct extent_map
*em
;
1019 u64 len
= PAGE_CACHE_SIZE
;
1022 * hopefully we have this extent in the tree already, try without
1023 * the full extent lock
1025 read_lock(&em_tree
->lock
);
1026 em
= lookup_extent_mapping(em_tree
, start
, len
);
1027 read_unlock(&em_tree
->lock
);
1030 struct extent_state
*cached
= NULL
;
1031 u64 end
= start
+ len
- 1;
1033 /* get the big lock and read metadata off disk */
1034 lock_extent_bits(io_tree
, start
, end
, 0, &cached
);
1035 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
1036 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1045 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1047 struct extent_map
*next
;
1050 /* this is the last extent */
1051 if (em
->start
+ em
->len
>= i_size_read(inode
))
1054 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1055 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
||
1056 (em
->block_start
+ em
->block_len
== next
->block_start
))
1059 free_extent_map(next
);
1063 static int should_defrag_range(struct inode
*inode
, u64 start
, int thresh
,
1064 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1067 struct extent_map
*em
;
1069 bool next_mergeable
= true;
1072 * make sure that once we start defragging an extent, we keep on
1075 if (start
< *defrag_end
)
1080 em
= defrag_lookup_extent(inode
, start
);
1084 /* this will cover holes, and inline extents */
1085 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1090 next_mergeable
= defrag_check_next_extent(inode
, em
);
1093 * we hit a real extent, if it is big or the next extent is not a
1094 * real extent, don't bother defragging it
1096 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1097 (em
->len
>= thresh
|| !next_mergeable
))
1101 * last_len ends up being a counter of how many bytes we've defragged.
1102 * every time we choose not to defrag an extent, we reset *last_len
1103 * so that the next tiny extent will force a defrag.
1105 * The end result of this is that tiny extents before a single big
1106 * extent will force at least part of that big extent to be defragged.
1109 *defrag_end
= extent_map_end(em
);
1112 *skip
= extent_map_end(em
);
1116 free_extent_map(em
);
1121 * it doesn't do much good to defrag one or two pages
1122 * at a time. This pulls in a nice chunk of pages
1123 * to COW and defrag.
1125 * It also makes sure the delalloc code has enough
1126 * dirty data to avoid making new small extents as part
1129 * It's a good idea to start RA on this range
1130 * before calling this.
1132 static int cluster_pages_for_defrag(struct inode
*inode
,
1133 struct page
**pages
,
1134 unsigned long start_index
,
1135 unsigned long num_pages
)
1137 unsigned long file_end
;
1138 u64 isize
= i_size_read(inode
);
1145 struct btrfs_ordered_extent
*ordered
;
1146 struct extent_state
*cached_state
= NULL
;
1147 struct extent_io_tree
*tree
;
1148 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1150 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1151 if (!isize
|| start_index
> file_end
)
1154 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1156 ret
= btrfs_delalloc_reserve_space(inode
,
1157 page_cnt
<< PAGE_CACHE_SHIFT
);
1161 tree
= &BTRFS_I(inode
)->io_tree
;
1163 /* step one, lock all the pages */
1164 for (i
= 0; i
< page_cnt
; i
++) {
1167 page
= find_or_create_page(inode
->i_mapping
,
1168 start_index
+ i
, mask
);
1172 page_start
= page_offset(page
);
1173 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1175 lock_extent_bits(tree
, page_start
, page_end
,
1177 ordered
= btrfs_lookup_ordered_extent(inode
,
1179 unlock_extent_cached(tree
, page_start
, page_end
,
1180 &cached_state
, GFP_NOFS
);
1185 btrfs_start_ordered_extent(inode
, ordered
, 1);
1186 btrfs_put_ordered_extent(ordered
);
1189 * we unlocked the page above, so we need check if
1190 * it was released or not.
1192 if (page
->mapping
!= inode
->i_mapping
) {
1194 page_cache_release(page
);
1199 if (!PageUptodate(page
)) {
1200 btrfs_readpage(NULL
, page
);
1202 if (!PageUptodate(page
)) {
1204 page_cache_release(page
);
1210 if (page
->mapping
!= inode
->i_mapping
) {
1212 page_cache_release(page
);
1222 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1226 * so now we have a nice long stream of locked
1227 * and up to date pages, lets wait on them
1229 for (i
= 0; i
< i_done
; i
++)
1230 wait_on_page_writeback(pages
[i
]);
1232 page_start
= page_offset(pages
[0]);
1233 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1235 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1236 page_start
, page_end
- 1, 0, &cached_state
);
1237 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1238 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1239 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1240 &cached_state
, GFP_NOFS
);
1242 if (i_done
!= page_cnt
) {
1243 spin_lock(&BTRFS_I(inode
)->lock
);
1244 BTRFS_I(inode
)->outstanding_extents
++;
1245 spin_unlock(&BTRFS_I(inode
)->lock
);
1246 btrfs_delalloc_release_space(inode
,
1247 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1251 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1252 &cached_state
, GFP_NOFS
);
1254 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1255 page_start
, page_end
- 1, &cached_state
,
1258 for (i
= 0; i
< i_done
; i
++) {
1259 clear_page_dirty_for_io(pages
[i
]);
1260 ClearPageChecked(pages
[i
]);
1261 set_page_extent_mapped(pages
[i
]);
1262 set_page_dirty(pages
[i
]);
1263 unlock_page(pages
[i
]);
1264 page_cache_release(pages
[i
]);
1268 for (i
= 0; i
< i_done
; i
++) {
1269 unlock_page(pages
[i
]);
1270 page_cache_release(pages
[i
]);
1272 btrfs_delalloc_release_space(inode
, page_cnt
<< PAGE_CACHE_SHIFT
);
1277 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1278 struct btrfs_ioctl_defrag_range_args
*range
,
1279 u64 newer_than
, unsigned long max_to_defrag
)
1281 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1282 struct file_ra_state
*ra
= NULL
;
1283 unsigned long last_index
;
1284 u64 isize
= i_size_read(inode
);
1288 u64 newer_off
= range
->start
;
1290 unsigned long ra_index
= 0;
1292 int defrag_count
= 0;
1293 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1294 int extent_thresh
= range
->extent_thresh
;
1295 unsigned long max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
1296 unsigned long cluster
= max_cluster
;
1297 u64 new_align
= ~((u64
)128 * 1024 - 1);
1298 struct page
**pages
= NULL
;
1303 if (range
->start
>= isize
)
1306 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1307 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1309 if (range
->compress_type
)
1310 compress_type
= range
->compress_type
;
1313 if (extent_thresh
== 0)
1314 extent_thresh
= 256 * 1024;
1317 * if we were not given a file, allocate a readahead
1321 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1324 file_ra_state_init(ra
, inode
->i_mapping
);
1329 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1336 /* find the last page to defrag */
1337 if (range
->start
+ range
->len
> range
->start
) {
1338 last_index
= min_t(u64
, isize
- 1,
1339 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1341 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1345 ret
= find_new_extents(root
, inode
, newer_than
,
1346 &newer_off
, 64 * 1024);
1348 range
->start
= newer_off
;
1350 * we always align our defrag to help keep
1351 * the extents in the file evenly spaced
1353 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1357 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1360 max_to_defrag
= last_index
+ 1;
1363 * make writeback starts from i, so the defrag range can be
1364 * written sequentially.
1366 if (i
< inode
->i_mapping
->writeback_index
)
1367 inode
->i_mapping
->writeback_index
= i
;
1369 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1370 (i
< (i_size_read(inode
) + PAGE_CACHE_SIZE
- 1) >>
1371 PAGE_CACHE_SHIFT
)) {
1373 * make sure we stop running if someone unmounts
1376 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1379 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1380 printk(KERN_DEBUG
"BTRFS: defrag_file cancelled\n");
1385 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1386 extent_thresh
, &last_len
, &skip
,
1387 &defrag_end
, range
->flags
&
1388 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1391 * the should_defrag function tells us how much to skip
1392 * bump our counter by the suggested amount
1394 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
1395 i
= max(i
+ 1, next
);
1400 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1401 PAGE_CACHE_SHIFT
) - i
;
1402 cluster
= min(cluster
, max_cluster
);
1404 cluster
= max_cluster
;
1407 if (i
+ cluster
> ra_index
) {
1408 ra_index
= max(i
, ra_index
);
1409 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1411 ra_index
+= max_cluster
;
1414 mutex_lock(&inode
->i_mutex
);
1415 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1416 BTRFS_I(inode
)->force_compress
= compress_type
;
1417 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1419 mutex_unlock(&inode
->i_mutex
);
1423 defrag_count
+= ret
;
1424 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1425 mutex_unlock(&inode
->i_mutex
);
1428 if (newer_off
== (u64
)-1)
1434 newer_off
= max(newer_off
+ 1,
1435 (u64
)i
<< PAGE_CACHE_SHIFT
);
1437 ret
= find_new_extents(root
, inode
,
1438 newer_than
, &newer_off
,
1441 range
->start
= newer_off
;
1442 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1449 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1457 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1458 filemap_flush(inode
->i_mapping
);
1459 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1460 &BTRFS_I(inode
)->runtime_flags
))
1461 filemap_flush(inode
->i_mapping
);
1464 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1465 /* the filemap_flush will queue IO into the worker threads, but
1466 * we have to make sure the IO is actually started and that
1467 * ordered extents get created before we return
1469 atomic_inc(&root
->fs_info
->async_submit_draining
);
1470 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1471 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1472 wait_event(root
->fs_info
->async_submit_wait
,
1473 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1474 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1476 atomic_dec(&root
->fs_info
->async_submit_draining
);
1479 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1480 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1486 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1487 mutex_lock(&inode
->i_mutex
);
1488 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1489 mutex_unlock(&inode
->i_mutex
);
1497 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1503 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1504 struct btrfs_ioctl_vol_args
*vol_args
;
1505 struct btrfs_trans_handle
*trans
;
1506 struct btrfs_device
*device
= NULL
;
1509 char *devstr
= NULL
;
1513 if (!capable(CAP_SYS_ADMIN
))
1516 ret
= mnt_want_write_file(file
);
1520 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1522 mnt_drop_write_file(file
);
1523 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1526 mutex_lock(&root
->fs_info
->volume_mutex
);
1527 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1528 if (IS_ERR(vol_args
)) {
1529 ret
= PTR_ERR(vol_args
);
1533 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1535 sizestr
= vol_args
->name
;
1536 devstr
= strchr(sizestr
, ':');
1538 sizestr
= devstr
+ 1;
1540 devstr
= vol_args
->name
;
1541 ret
= kstrtoull(devstr
, 10, &devid
);
1548 btrfs_info(root
->fs_info
, "resizing devid %llu", devid
);
1551 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1553 btrfs_info(root
->fs_info
, "resizer unable to find device %llu",
1559 if (!device
->writeable
) {
1560 btrfs_info(root
->fs_info
,
1561 "resizer unable to apply on readonly device %llu",
1567 if (!strcmp(sizestr
, "max"))
1568 new_size
= device
->bdev
->bd_inode
->i_size
;
1570 if (sizestr
[0] == '-') {
1573 } else if (sizestr
[0] == '+') {
1577 new_size
= memparse(sizestr
, &retptr
);
1578 if (*retptr
!= '\0' || new_size
== 0) {
1584 if (device
->is_tgtdev_for_dev_replace
) {
1589 old_size
= device
->total_bytes
;
1592 if (new_size
> old_size
) {
1596 new_size
= old_size
- new_size
;
1597 } else if (mod
> 0) {
1598 if (new_size
> ULLONG_MAX
- old_size
) {
1602 new_size
= old_size
+ new_size
;
1605 if (new_size
< 256 * 1024 * 1024) {
1609 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1614 do_div(new_size
, root
->sectorsize
);
1615 new_size
*= root
->sectorsize
;
1617 printk_in_rcu(KERN_INFO
"BTRFS: new size for %s is %llu\n",
1618 rcu_str_deref(device
->name
), new_size
);
1620 if (new_size
> old_size
) {
1621 trans
= btrfs_start_transaction(root
, 0);
1622 if (IS_ERR(trans
)) {
1623 ret
= PTR_ERR(trans
);
1626 ret
= btrfs_grow_device(trans
, device
, new_size
);
1627 btrfs_commit_transaction(trans
, root
);
1628 } else if (new_size
< old_size
) {
1629 ret
= btrfs_shrink_device(device
, new_size
);
1630 } /* equal, nothing need to do */
1635 mutex_unlock(&root
->fs_info
->volume_mutex
);
1636 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1637 mnt_drop_write_file(file
);
1641 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1642 char *name
, unsigned long fd
, int subvol
,
1643 u64
*transid
, bool readonly
,
1644 struct btrfs_qgroup_inherit
*inherit
)
1649 ret
= mnt_want_write_file(file
);
1653 namelen
= strlen(name
);
1654 if (strchr(name
, '/')) {
1656 goto out_drop_write
;
1659 if (name
[0] == '.' &&
1660 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1662 goto out_drop_write
;
1666 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1667 NULL
, transid
, readonly
, inherit
);
1669 struct fd src
= fdget(fd
);
1670 struct inode
*src_inode
;
1673 goto out_drop_write
;
1676 src_inode
= file_inode(src
.file
);
1677 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1678 btrfs_info(BTRFS_I(src_inode
)->root
->fs_info
,
1679 "Snapshot src from another FS");
1681 } else if (!inode_owner_or_capable(src_inode
)) {
1683 * Subvolume creation is not restricted, but snapshots
1684 * are limited to own subvolumes only
1688 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1689 BTRFS_I(src_inode
)->root
,
1690 transid
, readonly
, inherit
);
1695 mnt_drop_write_file(file
);
1700 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1701 void __user
*arg
, int subvol
)
1703 struct btrfs_ioctl_vol_args
*vol_args
;
1706 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1707 if (IS_ERR(vol_args
))
1708 return PTR_ERR(vol_args
);
1709 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1711 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1712 vol_args
->fd
, subvol
,
1719 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1720 void __user
*arg
, int subvol
)
1722 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1726 bool readonly
= false;
1727 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1729 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1730 if (IS_ERR(vol_args
))
1731 return PTR_ERR(vol_args
);
1732 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1734 if (vol_args
->flags
&
1735 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1736 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1741 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1743 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1745 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1746 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1750 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1751 if (IS_ERR(inherit
)) {
1752 ret
= PTR_ERR(inherit
);
1757 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1758 vol_args
->fd
, subvol
, ptr
,
1761 if (ret
== 0 && ptr
&&
1763 offsetof(struct btrfs_ioctl_vol_args_v2
,
1764 transid
), ptr
, sizeof(*ptr
)))
1772 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1775 struct inode
*inode
= file_inode(file
);
1776 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1780 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1783 down_read(&root
->fs_info
->subvol_sem
);
1784 if (btrfs_root_readonly(root
))
1785 flags
|= BTRFS_SUBVOL_RDONLY
;
1786 up_read(&root
->fs_info
->subvol_sem
);
1788 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1794 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1797 struct inode
*inode
= file_inode(file
);
1798 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1799 struct btrfs_trans_handle
*trans
;
1804 if (!inode_owner_or_capable(inode
))
1807 ret
= mnt_want_write_file(file
);
1811 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1813 goto out_drop_write
;
1816 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1818 goto out_drop_write
;
1821 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1823 goto out_drop_write
;
1826 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1828 goto out_drop_write
;
1831 down_write(&root
->fs_info
->subvol_sem
);
1834 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1837 root_flags
= btrfs_root_flags(&root
->root_item
);
1838 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1839 btrfs_set_root_flags(&root
->root_item
,
1840 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1843 * Block RO -> RW transition if this subvolume is involved in
1846 spin_lock(&root
->root_item_lock
);
1847 if (root
->send_in_progress
== 0) {
1848 btrfs_set_root_flags(&root
->root_item
,
1849 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1850 spin_unlock(&root
->root_item_lock
);
1852 spin_unlock(&root
->root_item_lock
);
1853 btrfs_warn(root
->fs_info
,
1854 "Attempt to set subvolume %llu read-write during send",
1855 root
->root_key
.objectid
);
1861 trans
= btrfs_start_transaction(root
, 1);
1862 if (IS_ERR(trans
)) {
1863 ret
= PTR_ERR(trans
);
1867 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1868 &root
->root_key
, &root
->root_item
);
1870 btrfs_commit_transaction(trans
, root
);
1873 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1875 up_write(&root
->fs_info
->subvol_sem
);
1877 mnt_drop_write_file(file
);
1883 * helper to check if the subvolume references other subvolumes
1885 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1887 struct btrfs_path
*path
;
1888 struct btrfs_dir_item
*di
;
1889 struct btrfs_key key
;
1893 path
= btrfs_alloc_path();
1897 /* Make sure this root isn't set as the default subvol */
1898 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
1899 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
, path
,
1900 dir_id
, "default", 7, 0);
1901 if (di
&& !IS_ERR(di
)) {
1902 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1903 if (key
.objectid
== root
->root_key
.objectid
) {
1905 btrfs_err(root
->fs_info
, "deleting default subvolume "
1906 "%llu is not allowed", key
.objectid
);
1909 btrfs_release_path(path
);
1912 key
.objectid
= root
->root_key
.objectid
;
1913 key
.type
= BTRFS_ROOT_REF_KEY
;
1914 key
.offset
= (u64
)-1;
1916 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1923 if (path
->slots
[0] > 0) {
1925 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1926 if (key
.objectid
== root
->root_key
.objectid
&&
1927 key
.type
== BTRFS_ROOT_REF_KEY
)
1931 btrfs_free_path(path
);
1935 static noinline
int key_in_sk(struct btrfs_key
*key
,
1936 struct btrfs_ioctl_search_key
*sk
)
1938 struct btrfs_key test
;
1941 test
.objectid
= sk
->min_objectid
;
1942 test
.type
= sk
->min_type
;
1943 test
.offset
= sk
->min_offset
;
1945 ret
= btrfs_comp_cpu_keys(key
, &test
);
1949 test
.objectid
= sk
->max_objectid
;
1950 test
.type
= sk
->max_type
;
1951 test
.offset
= sk
->max_offset
;
1953 ret
= btrfs_comp_cpu_keys(key
, &test
);
1959 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1960 struct btrfs_path
*path
,
1961 struct btrfs_key
*key
,
1962 struct btrfs_ioctl_search_key
*sk
,
1965 unsigned long *sk_offset
,
1969 struct extent_buffer
*leaf
;
1970 struct btrfs_ioctl_search_header sh
;
1971 unsigned long item_off
;
1972 unsigned long item_len
;
1978 leaf
= path
->nodes
[0];
1979 slot
= path
->slots
[0];
1980 nritems
= btrfs_header_nritems(leaf
);
1982 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1986 found_transid
= btrfs_header_generation(leaf
);
1988 for (i
= slot
; i
< nritems
; i
++) {
1989 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1990 item_len
= btrfs_item_size_nr(leaf
, i
);
1992 btrfs_item_key_to_cpu(leaf
, key
, i
);
1993 if (!key_in_sk(key
, sk
))
1996 if (sizeof(sh
) + item_len
> *buf_size
) {
2003 * return one empty item back for v1, which does not
2007 *buf_size
= sizeof(sh
) + item_len
;
2012 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
2017 sh
.objectid
= key
->objectid
;
2018 sh
.offset
= key
->offset
;
2019 sh
.type
= key
->type
;
2021 sh
.transid
= found_transid
;
2023 /* copy search result header */
2024 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2029 *sk_offset
+= sizeof(sh
);
2032 char __user
*up
= ubuf
+ *sk_offset
;
2034 if (read_extent_buffer_to_user(leaf
, up
,
2035 item_off
, item_len
)) {
2040 *sk_offset
+= item_len
;
2044 if (ret
) /* -EOVERFLOW from above */
2047 if (*num_found
>= sk
->nr_items
) {
2054 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
2056 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
2059 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
2067 * 0: all items from this leaf copied, continue with next
2068 * 1: * more items can be copied, but unused buffer is too small
2069 * * all items were found
2070 * Either way, it will stops the loop which iterates to the next
2072 * -EOVERFLOW: item was to large for buffer
2073 * -EFAULT: could not copy extent buffer back to userspace
2078 static noinline
int search_ioctl(struct inode
*inode
,
2079 struct btrfs_ioctl_search_key
*sk
,
2083 struct btrfs_root
*root
;
2084 struct btrfs_key key
;
2085 struct btrfs_path
*path
;
2086 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
2089 unsigned long sk_offset
= 0;
2091 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2092 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2096 path
= btrfs_alloc_path();
2100 if (sk
->tree_id
== 0) {
2101 /* search the root of the inode that was passed */
2102 root
= BTRFS_I(inode
)->root
;
2104 key
.objectid
= sk
->tree_id
;
2105 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2106 key
.offset
= (u64
)-1;
2107 root
= btrfs_read_fs_root_no_name(info
, &key
);
2109 printk(KERN_ERR
"BTRFS: could not find root %llu\n",
2111 btrfs_free_path(path
);
2116 key
.objectid
= sk
->min_objectid
;
2117 key
.type
= sk
->min_type
;
2118 key
.offset
= sk
->min_offset
;
2120 path
->keep_locks
= 1;
2123 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2129 ret
= copy_to_sk(root
, path
, &key
, sk
, buf_size
, ubuf
,
2130 &sk_offset
, &num_found
);
2131 btrfs_release_path(path
);
2139 sk
->nr_items
= num_found
;
2140 btrfs_free_path(path
);
2144 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2147 struct btrfs_ioctl_search_args __user
*uargs
;
2148 struct btrfs_ioctl_search_key sk
;
2149 struct inode
*inode
;
2153 if (!capable(CAP_SYS_ADMIN
))
2156 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2158 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2161 buf_size
= sizeof(uargs
->buf
);
2163 inode
= file_inode(file
);
2164 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2167 * In the origin implementation an overflow is handled by returning a
2168 * search header with a len of zero, so reset ret.
2170 if (ret
== -EOVERFLOW
)
2173 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2178 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2181 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2182 struct btrfs_ioctl_search_args_v2 args
;
2183 struct inode
*inode
;
2186 const size_t buf_limit
= 16 * 1024 * 1024;
2188 if (!capable(CAP_SYS_ADMIN
))
2191 /* copy search header and buffer size */
2192 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2193 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2196 buf_size
= args
.buf_size
;
2198 if (buf_size
< sizeof(struct btrfs_ioctl_search_header
))
2201 /* limit result size to 16MB */
2202 if (buf_size
> buf_limit
)
2203 buf_size
= buf_limit
;
2205 inode
= file_inode(file
);
2206 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2207 (char *)(&uarg
->buf
[0]));
2208 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2210 else if (ret
== -EOVERFLOW
&&
2211 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2218 * Search INODE_REFs to identify path name of 'dirid' directory
2219 * in a 'tree_id' tree. and sets path name to 'name'.
2221 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2222 u64 tree_id
, u64 dirid
, char *name
)
2224 struct btrfs_root
*root
;
2225 struct btrfs_key key
;
2231 struct btrfs_inode_ref
*iref
;
2232 struct extent_buffer
*l
;
2233 struct btrfs_path
*path
;
2235 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2240 path
= btrfs_alloc_path();
2244 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2246 key
.objectid
= tree_id
;
2247 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2248 key
.offset
= (u64
)-1;
2249 root
= btrfs_read_fs_root_no_name(info
, &key
);
2251 printk(KERN_ERR
"BTRFS: could not find root %llu\n", tree_id
);
2256 key
.objectid
= dirid
;
2257 key
.type
= BTRFS_INODE_REF_KEY
;
2258 key
.offset
= (u64
)-1;
2261 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2265 ret
= btrfs_previous_item(root
, path
, dirid
,
2266 BTRFS_INODE_REF_KEY
);
2276 slot
= path
->slots
[0];
2277 btrfs_item_key_to_cpu(l
, &key
, slot
);
2279 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2280 len
= btrfs_inode_ref_name_len(l
, iref
);
2282 total_len
+= len
+ 1;
2284 ret
= -ENAMETOOLONG
;
2289 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2291 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2294 btrfs_release_path(path
);
2295 key
.objectid
= key
.offset
;
2296 key
.offset
= (u64
)-1;
2297 dirid
= key
.objectid
;
2299 memmove(name
, ptr
, total_len
);
2300 name
[total_len
] = '\0';
2303 btrfs_free_path(path
);
2307 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2310 struct btrfs_ioctl_ino_lookup_args
*args
;
2311 struct inode
*inode
;
2314 if (!capable(CAP_SYS_ADMIN
))
2317 args
= memdup_user(argp
, sizeof(*args
));
2319 return PTR_ERR(args
);
2321 inode
= file_inode(file
);
2323 if (args
->treeid
== 0)
2324 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2326 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2327 args
->treeid
, args
->objectid
,
2330 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2337 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2340 struct dentry
*parent
= file
->f_path
.dentry
;
2341 struct dentry
*dentry
;
2342 struct inode
*dir
= parent
->d_inode
;
2343 struct inode
*inode
;
2344 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2345 struct btrfs_root
*dest
= NULL
;
2346 struct btrfs_ioctl_vol_args
*vol_args
;
2347 struct btrfs_trans_handle
*trans
;
2348 struct btrfs_block_rsv block_rsv
;
2350 u64 qgroup_reserved
;
2355 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2356 if (IS_ERR(vol_args
))
2357 return PTR_ERR(vol_args
);
2359 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2360 namelen
= strlen(vol_args
->name
);
2361 if (strchr(vol_args
->name
, '/') ||
2362 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2367 err
= mnt_want_write_file(file
);
2372 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2374 goto out_drop_write
;
2375 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2376 if (IS_ERR(dentry
)) {
2377 err
= PTR_ERR(dentry
);
2378 goto out_unlock_dir
;
2381 if (!dentry
->d_inode
) {
2386 inode
= dentry
->d_inode
;
2387 dest
= BTRFS_I(inode
)->root
;
2388 if (!capable(CAP_SYS_ADMIN
)) {
2390 * Regular user. Only allow this with a special mount
2391 * option, when the user has write+exec access to the
2392 * subvol root, and when rmdir(2) would have been
2395 * Note that this is _not_ check that the subvol is
2396 * empty or doesn't contain data that we wouldn't
2397 * otherwise be able to delete.
2399 * Users who want to delete empty subvols should try
2403 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2407 * Do not allow deletion if the parent dir is the same
2408 * as the dir to be deleted. That means the ioctl
2409 * must be called on the dentry referencing the root
2410 * of the subvol, not a random directory contained
2417 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2422 /* check if subvolume may be deleted by a user */
2423 err
= btrfs_may_delete(dir
, dentry
, 1);
2427 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2432 mutex_lock(&inode
->i_mutex
);
2435 * Don't allow to delete a subvolume with send in progress. This is
2436 * inside the i_mutex so the error handling that has to drop the bit
2437 * again is not run concurrently.
2439 spin_lock(&dest
->root_item_lock
);
2440 root_flags
= btrfs_root_flags(&dest
->root_item
);
2441 if (dest
->send_in_progress
== 0) {
2442 btrfs_set_root_flags(&dest
->root_item
,
2443 root_flags
| BTRFS_ROOT_SUBVOL_DEAD
);
2444 spin_unlock(&dest
->root_item_lock
);
2446 spin_unlock(&dest
->root_item_lock
);
2447 btrfs_warn(root
->fs_info
,
2448 "Attempt to delete subvolume %llu during send",
2449 dest
->root_key
.objectid
);
2454 err
= d_invalidate(dentry
);
2458 down_write(&root
->fs_info
->subvol_sem
);
2460 err
= may_destroy_subvol(dest
);
2464 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2466 * One for dir inode, two for dir entries, two for root
2469 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2470 5, &qgroup_reserved
, true);
2474 trans
= btrfs_start_transaction(root
, 0);
2475 if (IS_ERR(trans
)) {
2476 err
= PTR_ERR(trans
);
2479 trans
->block_rsv
= &block_rsv
;
2480 trans
->bytes_reserved
= block_rsv
.size
;
2482 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2483 dest
->root_key
.objectid
,
2484 dentry
->d_name
.name
,
2485 dentry
->d_name
.len
);
2488 btrfs_abort_transaction(trans
, root
, ret
);
2492 btrfs_record_root_in_trans(trans
, dest
);
2494 memset(&dest
->root_item
.drop_progress
, 0,
2495 sizeof(dest
->root_item
.drop_progress
));
2496 dest
->root_item
.drop_level
= 0;
2497 btrfs_set_root_refs(&dest
->root_item
, 0);
2499 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED
, &dest
->state
)) {
2500 ret
= btrfs_insert_orphan_item(trans
,
2501 root
->fs_info
->tree_root
,
2502 dest
->root_key
.objectid
);
2504 btrfs_abort_transaction(trans
, root
, ret
);
2510 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2511 dest
->root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
2512 dest
->root_key
.objectid
);
2513 if (ret
&& ret
!= -ENOENT
) {
2514 btrfs_abort_transaction(trans
, root
, ret
);
2518 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2519 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2520 dest
->root_item
.received_uuid
,
2521 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2522 dest
->root_key
.objectid
);
2523 if (ret
&& ret
!= -ENOENT
) {
2524 btrfs_abort_transaction(trans
, root
, ret
);
2531 trans
->block_rsv
= NULL
;
2532 trans
->bytes_reserved
= 0;
2533 ret
= btrfs_end_transaction(trans
, root
);
2536 inode
->i_flags
|= S_DEAD
;
2538 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2540 up_write(&root
->fs_info
->subvol_sem
);
2543 spin_lock(&dest
->root_item_lock
);
2544 root_flags
= btrfs_root_flags(&dest
->root_item
);
2545 btrfs_set_root_flags(&dest
->root_item
,
2546 root_flags
& ~BTRFS_ROOT_SUBVOL_DEAD
);
2547 spin_unlock(&dest
->root_item_lock
);
2549 mutex_unlock(&inode
->i_mutex
);
2551 shrink_dcache_sb(root
->fs_info
->sb
);
2552 btrfs_invalidate_inodes(dest
);
2554 ASSERT(dest
->send_in_progress
== 0);
2557 if (dest
->cache_inode
) {
2558 iput(dest
->cache_inode
);
2559 dest
->cache_inode
= NULL
;
2565 mutex_unlock(&dir
->i_mutex
);
2567 mnt_drop_write_file(file
);
2573 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2575 struct inode
*inode
= file_inode(file
);
2576 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2577 struct btrfs_ioctl_defrag_range_args
*range
;
2580 ret
= mnt_want_write_file(file
);
2584 if (btrfs_root_readonly(root
)) {
2589 switch (inode
->i_mode
& S_IFMT
) {
2591 if (!capable(CAP_SYS_ADMIN
)) {
2595 ret
= btrfs_defrag_root(root
);
2598 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2601 if (!(file
->f_mode
& FMODE_WRITE
)) {
2606 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2613 if (copy_from_user(range
, argp
,
2619 /* compression requires us to start the IO */
2620 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2621 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2622 range
->extent_thresh
= (u32
)-1;
2625 /* the rest are all set to zero by kzalloc */
2626 range
->len
= (u64
)-1;
2628 ret
= btrfs_defrag_file(file_inode(file
), file
,
2638 mnt_drop_write_file(file
);
2642 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2644 struct btrfs_ioctl_vol_args
*vol_args
;
2647 if (!capable(CAP_SYS_ADMIN
))
2650 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2652 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2655 mutex_lock(&root
->fs_info
->volume_mutex
);
2656 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2657 if (IS_ERR(vol_args
)) {
2658 ret
= PTR_ERR(vol_args
);
2662 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2663 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2667 mutex_unlock(&root
->fs_info
->volume_mutex
);
2668 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2672 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2674 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2675 struct btrfs_ioctl_vol_args
*vol_args
;
2678 if (!capable(CAP_SYS_ADMIN
))
2681 ret
= mnt_want_write_file(file
);
2685 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2686 if (IS_ERR(vol_args
)) {
2687 ret
= PTR_ERR(vol_args
);
2691 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2693 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2695 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2699 mutex_lock(&root
->fs_info
->volume_mutex
);
2700 ret
= btrfs_rm_device(root
, vol_args
->name
);
2701 mutex_unlock(&root
->fs_info
->volume_mutex
);
2702 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2706 mnt_drop_write_file(file
);
2710 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2712 struct btrfs_ioctl_fs_info_args
*fi_args
;
2713 struct btrfs_device
*device
;
2714 struct btrfs_device
*next
;
2715 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2718 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2722 mutex_lock(&fs_devices
->device_list_mutex
);
2723 fi_args
->num_devices
= fs_devices
->num_devices
;
2724 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2726 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2727 if (device
->devid
> fi_args
->max_id
)
2728 fi_args
->max_id
= device
->devid
;
2730 mutex_unlock(&fs_devices
->device_list_mutex
);
2732 fi_args
->nodesize
= root
->fs_info
->super_copy
->nodesize
;
2733 fi_args
->sectorsize
= root
->fs_info
->super_copy
->sectorsize
;
2734 fi_args
->clone_alignment
= root
->fs_info
->super_copy
->sectorsize
;
2736 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2743 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2745 struct btrfs_ioctl_dev_info_args
*di_args
;
2746 struct btrfs_device
*dev
;
2747 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2749 char *s_uuid
= NULL
;
2751 di_args
= memdup_user(arg
, sizeof(*di_args
));
2752 if (IS_ERR(di_args
))
2753 return PTR_ERR(di_args
);
2755 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2756 s_uuid
= di_args
->uuid
;
2758 mutex_lock(&fs_devices
->device_list_mutex
);
2759 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2766 di_args
->devid
= dev
->devid
;
2767 di_args
->bytes_used
= dev
->bytes_used
;
2768 di_args
->total_bytes
= dev
->total_bytes
;
2769 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2771 struct rcu_string
*name
;
2774 name
= rcu_dereference(dev
->name
);
2775 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2777 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2779 di_args
->path
[0] = '\0';
2783 mutex_unlock(&fs_devices
->device_list_mutex
);
2784 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2791 static struct page
*extent_same_get_page(struct inode
*inode
, u64 off
)
2795 struct extent_io_tree
*tree
= &BTRFS_I(inode
)->io_tree
;
2797 index
= off
>> PAGE_CACHE_SHIFT
;
2799 page
= grab_cache_page(inode
->i_mapping
, index
);
2803 if (!PageUptodate(page
)) {
2804 if (extent_read_full_page_nolock(tree
, page
, btrfs_get_extent
,
2808 if (!PageUptodate(page
)) {
2810 page_cache_release(page
);
2819 static inline void lock_extent_range(struct inode
*inode
, u64 off
, u64 len
)
2821 /* do any pending delalloc/csum calc on src, one way or
2822 another, and lock file content */
2824 struct btrfs_ordered_extent
*ordered
;
2825 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2826 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2829 ordered
->file_offset
+ ordered
->len
<= off
||
2830 ordered
->file_offset
>= off
+ len
) &&
2831 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2832 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
)) {
2834 btrfs_put_ordered_extent(ordered
);
2837 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2839 btrfs_put_ordered_extent(ordered
);
2840 btrfs_wait_ordered_range(inode
, off
, len
);
2844 static void btrfs_double_unlock(struct inode
*inode1
, u64 loff1
,
2845 struct inode
*inode2
, u64 loff2
, u64 len
)
2847 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2848 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2850 mutex_unlock(&inode1
->i_mutex
);
2851 mutex_unlock(&inode2
->i_mutex
);
2854 static void btrfs_double_lock(struct inode
*inode1
, u64 loff1
,
2855 struct inode
*inode2
, u64 loff2
, u64 len
)
2857 if (inode1
< inode2
) {
2858 swap(inode1
, inode2
);
2862 mutex_lock_nested(&inode1
->i_mutex
, I_MUTEX_PARENT
);
2863 lock_extent_range(inode1
, loff1
, len
);
2864 if (inode1
!= inode2
) {
2865 mutex_lock_nested(&inode2
->i_mutex
, I_MUTEX_CHILD
);
2866 lock_extent_range(inode2
, loff2
, len
);
2870 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
2871 u64 dst_loff
, u64 len
)
2874 struct page
*src_page
, *dst_page
;
2875 unsigned int cmp_len
= PAGE_CACHE_SIZE
;
2876 void *addr
, *dst_addr
;
2879 if (len
< PAGE_CACHE_SIZE
)
2882 src_page
= extent_same_get_page(src
, loff
);
2885 dst_page
= extent_same_get_page(dst
, dst_loff
);
2887 page_cache_release(src_page
);
2890 addr
= kmap_atomic(src_page
);
2891 dst_addr
= kmap_atomic(dst_page
);
2893 flush_dcache_page(src_page
);
2894 flush_dcache_page(dst_page
);
2896 if (memcmp(addr
, dst_addr
, cmp_len
))
2897 ret
= BTRFS_SAME_DATA_DIFFERS
;
2899 kunmap_atomic(addr
);
2900 kunmap_atomic(dst_addr
);
2901 page_cache_release(src_page
);
2902 page_cache_release(dst_page
);
2908 dst_loff
+= cmp_len
;
2915 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64 len
)
2917 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
2919 if (off
+ len
> inode
->i_size
|| off
+ len
< off
)
2921 /* Check that we are block aligned - btrfs_clone() requires this */
2922 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
2928 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 len
,
2929 struct inode
*dst
, u64 dst_loff
)
2934 * btrfs_clone() can't handle extents in the same file
2935 * yet. Once that works, we can drop this check and replace it
2936 * with a check for the same inode, but overlapping extents.
2941 btrfs_double_lock(src
, loff
, dst
, dst_loff
, len
);
2943 ret
= extent_same_check_offsets(src
, loff
, len
);
2947 ret
= extent_same_check_offsets(dst
, dst_loff
, len
);
2951 /* don't make the dst file partly checksummed */
2952 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2953 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
2958 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, len
);
2960 ret
= btrfs_clone(src
, dst
, loff
, len
, len
, dst_loff
);
2963 btrfs_double_unlock(src
, loff
, dst
, dst_loff
, len
);
2968 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2970 static long btrfs_ioctl_file_extent_same(struct file
*file
,
2971 struct btrfs_ioctl_same_args __user
*argp
)
2973 struct btrfs_ioctl_same_args
*same
;
2974 struct btrfs_ioctl_same_extent_info
*info
;
2975 struct inode
*src
= file_inode(file
);
2981 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
2982 bool is_admin
= capable(CAP_SYS_ADMIN
);
2985 if (!(file
->f_mode
& FMODE_READ
))
2988 ret
= mnt_want_write_file(file
);
2992 if (get_user(count
, &argp
->dest_count
)) {
2997 size
= offsetof(struct btrfs_ioctl_same_args __user
, info
[count
]);
2999 same
= memdup_user(argp
, size
);
3002 ret
= PTR_ERR(same
);
3006 off
= same
->logical_offset
;
3010 * Limit the total length we will dedupe for each operation.
3011 * This is intended to bound the total time spent in this
3012 * ioctl to something sane.
3014 if (len
> BTRFS_MAX_DEDUPE_LEN
)
3015 len
= BTRFS_MAX_DEDUPE_LEN
;
3017 if (WARN_ON_ONCE(bs
< PAGE_CACHE_SIZE
)) {
3019 * Btrfs does not support blocksize < page_size. As a
3020 * result, btrfs_cmp_data() won't correctly handle
3021 * this situation without an update.
3028 if (S_ISDIR(src
->i_mode
))
3032 if (!S_ISREG(src
->i_mode
))
3035 /* pre-format output fields to sane values */
3036 for (i
= 0; i
< count
; i
++) {
3037 same
->info
[i
].bytes_deduped
= 0ULL;
3038 same
->info
[i
].status
= 0;
3041 for (i
= 0, info
= same
->info
; i
< count
; i
++, info
++) {
3043 struct fd dst_file
= fdget(info
->fd
);
3044 if (!dst_file
.file
) {
3045 info
->status
= -EBADF
;
3048 dst
= file_inode(dst_file
.file
);
3050 if (!(is_admin
|| (dst_file
.file
->f_mode
& FMODE_WRITE
))) {
3051 info
->status
= -EINVAL
;
3052 } else if (file
->f_path
.mnt
!= dst_file
.file
->f_path
.mnt
) {
3053 info
->status
= -EXDEV
;
3054 } else if (S_ISDIR(dst
->i_mode
)) {
3055 info
->status
= -EISDIR
;
3056 } else if (!S_ISREG(dst
->i_mode
)) {
3057 info
->status
= -EACCES
;
3059 info
->status
= btrfs_extent_same(src
, off
, len
, dst
,
3060 info
->logical_offset
);
3061 if (info
->status
== 0)
3062 info
->bytes_deduped
+= len
;
3067 ret
= copy_to_user(argp
, same
, size
);
3072 mnt_drop_write_file(file
);
3076 /* Helper to check and see if this root currently has a ref on the given disk
3077 * bytenr. If it does then we need to update the quota for this root. This
3078 * doesn't do anything if quotas aren't enabled.
3080 static int check_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
3083 struct seq_list tree_mod_seq_elem
= {};
3084 struct ulist
*roots
;
3085 struct ulist_iterator uiter
;
3086 struct ulist_node
*root_node
= NULL
;
3089 if (!root
->fs_info
->quota_enabled
)
3092 btrfs_get_tree_mod_seq(root
->fs_info
, &tree_mod_seq_elem
);
3093 ret
= btrfs_find_all_roots(trans
, root
->fs_info
, disko
,
3094 tree_mod_seq_elem
.seq
, &roots
);
3098 ULIST_ITER_INIT(&uiter
);
3099 while ((root_node
= ulist_next(roots
, &uiter
))) {
3100 if (root_node
->val
== root
->objectid
) {
3107 btrfs_put_tree_mod_seq(root
->fs_info
, &tree_mod_seq_elem
);
3111 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3112 struct inode
*inode
,
3117 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3120 inode_inc_iversion(inode
);
3121 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3123 * We round up to the block size at eof when determining which
3124 * extents to clone above, but shouldn't round up the file size.
3126 if (endoff
> destoff
+ olen
)
3127 endoff
= destoff
+ olen
;
3128 if (endoff
> inode
->i_size
)
3129 btrfs_i_size_write(inode
, endoff
);
3131 ret
= btrfs_update_inode(trans
, root
, inode
);
3133 btrfs_abort_transaction(trans
, root
, ret
);
3134 btrfs_end_transaction(trans
, root
);
3137 ret
= btrfs_end_transaction(trans
, root
);
3142 static void clone_update_extent_map(struct inode
*inode
,
3143 const struct btrfs_trans_handle
*trans
,
3144 const struct btrfs_path
*path
,
3145 const u64 hole_offset
,
3148 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
3149 struct extent_map
*em
;
3152 em
= alloc_extent_map();
3154 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3155 &BTRFS_I(inode
)->runtime_flags
);
3160 struct btrfs_file_extent_item
*fi
;
3162 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3163 struct btrfs_file_extent_item
);
3164 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3165 em
->generation
= -1;
3166 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3167 BTRFS_FILE_EXTENT_INLINE
)
3168 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3169 &BTRFS_I(inode
)->runtime_flags
);
3171 em
->start
= hole_offset
;
3173 em
->ram_bytes
= em
->len
;
3174 em
->orig_start
= hole_offset
;
3175 em
->block_start
= EXTENT_MAP_HOLE
;
3177 em
->orig_block_len
= 0;
3178 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3179 em
->generation
= trans
->transid
;
3183 write_lock(&em_tree
->lock
);
3184 ret
= add_extent_mapping(em_tree
, em
, 1);
3185 write_unlock(&em_tree
->lock
);
3186 if (ret
!= -EEXIST
) {
3187 free_extent_map(em
);
3190 btrfs_drop_extent_cache(inode
, em
->start
,
3191 em
->start
+ em
->len
- 1, 0);
3195 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3196 &BTRFS_I(inode
)->runtime_flags
);
3200 * btrfs_clone() - clone a range from inode file to another
3202 * @src: Inode to clone from
3203 * @inode: Inode to clone to
3204 * @off: Offset within source to start clone from
3205 * @olen: Original length, passed by user, of range to clone
3206 * @olen_aligned: Block-aligned value of olen, extent_same uses
3207 * identical values here
3208 * @destoff: Offset within @inode to start clone
3210 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3211 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3214 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3215 struct btrfs_path
*path
= NULL
;
3216 struct extent_buffer
*leaf
;
3217 struct btrfs_trans_handle
*trans
;
3219 struct btrfs_key key
;
3224 const u64 len
= olen_aligned
;
3226 u64 last_dest_end
= destoff
;
3229 buf
= vmalloc(btrfs_level_size(root
, 0));
3233 path
= btrfs_alloc_path();
3241 key
.objectid
= btrfs_ino(src
);
3242 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3247 * note the key will change type as we walk through the
3250 path
->leave_spinning
= 1;
3251 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3256 * First search, if no extent item that starts at offset off was
3257 * found but the previous item is an extent item, it's possible
3258 * it might overlap our target range, therefore process it.
3260 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3261 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3262 path
->slots
[0] - 1);
3263 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3267 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3270 if (path
->slots
[0] >= nritems
) {
3271 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3276 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3278 leaf
= path
->nodes
[0];
3279 slot
= path
->slots
[0];
3281 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3282 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
3283 key
.objectid
!= btrfs_ino(src
))
3286 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
3287 struct btrfs_file_extent_item
*extent
;
3290 struct btrfs_key new_key
;
3291 u64 disko
= 0, diskl
= 0;
3292 u64 datao
= 0, datal
= 0;
3296 extent
= btrfs_item_ptr(leaf
, slot
,
3297 struct btrfs_file_extent_item
);
3298 comp
= btrfs_file_extent_compression(leaf
, extent
);
3299 type
= btrfs_file_extent_type(leaf
, extent
);
3300 if (type
== BTRFS_FILE_EXTENT_REG
||
3301 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3302 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3304 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3306 datao
= btrfs_file_extent_offset(leaf
, extent
);
3307 datal
= btrfs_file_extent_num_bytes(leaf
,
3309 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3310 /* take upper bound, may be compressed */
3311 datal
= btrfs_file_extent_ram_bytes(leaf
,
3316 * The first search might have left us at an extent
3317 * item that ends before our target range's start, can
3318 * happen if we have holes and NO_HOLES feature enabled.
3320 if (key
.offset
+ datal
<= off
) {
3323 } else if (key
.offset
>= off
+ len
) {
3327 size
= btrfs_item_size_nr(leaf
, slot
);
3328 read_extent_buffer(leaf
, buf
,
3329 btrfs_item_ptr_offset(leaf
, slot
),
3332 btrfs_release_path(path
);
3333 path
->leave_spinning
= 0;
3335 memcpy(&new_key
, &key
, sizeof(new_key
));
3336 new_key
.objectid
= btrfs_ino(inode
);
3337 if (off
<= key
.offset
)
3338 new_key
.offset
= key
.offset
+ destoff
- off
;
3340 new_key
.offset
= destoff
;
3343 * Deal with a hole that doesn't have an extent item
3344 * that represents it (NO_HOLES feature enabled).
3345 * This hole is either in the middle of the cloning
3346 * range or at the beginning (fully overlaps it or
3347 * partially overlaps it).
3349 if (new_key
.offset
!= last_dest_end
)
3350 drop_start
= last_dest_end
;
3352 drop_start
= new_key
.offset
;
3355 * 1 - adjusting old extent (we may have to split it)
3356 * 1 - add new extent
3359 trans
= btrfs_start_transaction(root
, 3);
3360 if (IS_ERR(trans
)) {
3361 ret
= PTR_ERR(trans
);
3365 if (type
== BTRFS_FILE_EXTENT_REG
||
3366 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3368 * a | --- range to clone ---| b
3369 * | ------------- extent ------------- |
3372 /* subtract range b */
3373 if (key
.offset
+ datal
> off
+ len
)
3374 datal
= off
+ len
- key
.offset
;
3376 /* subtract range a */
3377 if (off
> key
.offset
) {
3378 datao
+= off
- key
.offset
;
3379 datal
-= off
- key
.offset
;
3382 ret
= btrfs_drop_extents(trans
, root
, inode
,
3384 new_key
.offset
+ datal
,
3387 if (ret
!= -EOPNOTSUPP
)
3388 btrfs_abort_transaction(trans
,
3390 btrfs_end_transaction(trans
, root
);
3394 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3397 btrfs_abort_transaction(trans
, root
,
3399 btrfs_end_transaction(trans
, root
);
3403 leaf
= path
->nodes
[0];
3404 slot
= path
->slots
[0];
3405 write_extent_buffer(leaf
, buf
,
3406 btrfs_item_ptr_offset(leaf
, slot
),
3409 extent
= btrfs_item_ptr(leaf
, slot
,
3410 struct btrfs_file_extent_item
);
3412 /* disko == 0 means it's a hole */
3416 btrfs_set_file_extent_offset(leaf
, extent
,
3418 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3422 * We need to look up the roots that point at
3423 * this bytenr and see if the new root does. If
3424 * it does not we need to make sure we update
3425 * quotas appropriately.
3427 if (disko
&& root
!= BTRFS_I(src
)->root
&&
3428 disko
!= last_disko
) {
3429 no_quota
= check_ref(trans
, root
,
3432 btrfs_abort_transaction(trans
,
3435 btrfs_end_transaction(trans
,
3443 inode_add_bytes(inode
, datal
);
3444 ret
= btrfs_inc_extent_ref(trans
, root
,
3446 root
->root_key
.objectid
,
3448 new_key
.offset
- datao
,
3451 btrfs_abort_transaction(trans
,
3454 btrfs_end_transaction(trans
,
3460 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3463 u64 aligned_end
= 0;
3465 if (off
> key
.offset
) {
3466 skip
= off
- key
.offset
;
3467 new_key
.offset
+= skip
;
3470 if (key
.offset
+ datal
> off
+ len
)
3471 trim
= key
.offset
+ datal
- (off
+ len
);
3473 if (comp
&& (skip
|| trim
)) {
3475 btrfs_end_transaction(trans
, root
);
3478 size
-= skip
+ trim
;
3479 datal
-= skip
+ trim
;
3481 aligned_end
= ALIGN(new_key
.offset
+ datal
,
3483 ret
= btrfs_drop_extents(trans
, root
, inode
,
3488 if (ret
!= -EOPNOTSUPP
)
3489 btrfs_abort_transaction(trans
,
3491 btrfs_end_transaction(trans
, root
);
3495 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3498 btrfs_abort_transaction(trans
, root
,
3500 btrfs_end_transaction(trans
, root
);
3506 btrfs_file_extent_calc_inline_size(0);
3507 memmove(buf
+start
, buf
+start
+skip
,
3511 leaf
= path
->nodes
[0];
3512 slot
= path
->slots
[0];
3513 write_extent_buffer(leaf
, buf
,
3514 btrfs_item_ptr_offset(leaf
, slot
),
3516 inode_add_bytes(inode
, datal
);
3519 /* If we have an implicit hole (NO_HOLES feature). */
3520 if (drop_start
< new_key
.offset
)
3521 clone_update_extent_map(inode
, trans
,
3523 new_key
.offset
- drop_start
);
3525 clone_update_extent_map(inode
, trans
, path
, 0, 0);
3527 btrfs_mark_buffer_dirty(leaf
);
3528 btrfs_release_path(path
);
3530 last_dest_end
= new_key
.offset
+ datal
;
3531 ret
= clone_finish_inode_update(trans
, inode
,
3536 if (new_key
.offset
+ datal
>= destoff
+ len
)
3539 btrfs_release_path(path
);
3544 if (last_dest_end
< destoff
+ len
) {
3546 * We have an implicit hole (NO_HOLES feature is enabled) that
3547 * fully or partially overlaps our cloning range at its end.
3549 btrfs_release_path(path
);
3552 * 1 - remove extent(s)
3555 trans
= btrfs_start_transaction(root
, 2);
3556 if (IS_ERR(trans
)) {
3557 ret
= PTR_ERR(trans
);
3560 ret
= btrfs_drop_extents(trans
, root
, inode
,
3561 last_dest_end
, destoff
+ len
, 1);
3563 if (ret
!= -EOPNOTSUPP
)
3564 btrfs_abort_transaction(trans
, root
, ret
);
3565 btrfs_end_transaction(trans
, root
);
3568 clone_update_extent_map(inode
, trans
, NULL
, last_dest_end
,
3569 destoff
+ len
- last_dest_end
);
3570 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3575 btrfs_free_path(path
);
3580 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
3581 u64 off
, u64 olen
, u64 destoff
)
3583 struct inode
*inode
= file_inode(file
);
3584 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3589 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
3594 * - split compressed inline extents. annoying: we need to
3595 * decompress into destination's address_space (the file offset
3596 * may change, so source mapping won't do), then recompress (or
3597 * otherwise reinsert) a subrange.
3599 * - split destination inode's inline extents. The inline extents can
3600 * be either compressed or non-compressed.
3603 /* the destination must be opened for writing */
3604 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
3607 if (btrfs_root_readonly(root
))
3610 ret
= mnt_want_write_file(file
);
3614 src_file
= fdget(srcfd
);
3615 if (!src_file
.file
) {
3617 goto out_drop_write
;
3621 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
3624 src
= file_inode(src_file
.file
);
3630 /* the src must be open for reading */
3631 if (!(src_file
.file
->f_mode
& FMODE_READ
))
3634 /* don't make the dst file partly checksummed */
3635 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3636 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3640 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3644 if (src
->i_sb
!= inode
->i_sb
)
3649 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
3650 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
3652 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
3653 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
3656 mutex_lock(&src
->i_mutex
);
3659 /* determine range to clone */
3661 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3664 olen
= len
= src
->i_size
- off
;
3665 /* if we extend to eof, continue to block boundary */
3666 if (off
+ len
== src
->i_size
)
3667 len
= ALIGN(src
->i_size
, bs
) - off
;
3669 /* verify the end result is block aligned */
3670 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3671 !IS_ALIGNED(destoff
, bs
))
3674 /* verify if ranges are overlapped within the same file */
3676 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3680 if (destoff
> inode
->i_size
) {
3681 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3687 * Lock the target range too. Right after we replace the file extent
3688 * items in the fs tree (which now point to the cloned data), we might
3689 * have a worker replace them with extent items relative to a write
3690 * operation that was issued before this clone operation (i.e. confront
3691 * with inode.c:btrfs_finish_ordered_io).
3694 u64 lock_start
= min_t(u64
, off
, destoff
);
3695 u64 lock_len
= max_t(u64
, off
, destoff
) + len
- lock_start
;
3697 lock_extent_range(src
, lock_start
, lock_len
);
3699 lock_extent_range(src
, off
, len
);
3700 lock_extent_range(inode
, destoff
, len
);
3703 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
);
3706 u64 lock_start
= min_t(u64
, off
, destoff
);
3707 u64 lock_end
= max_t(u64
, off
, destoff
) + len
- 1;
3709 unlock_extent(&BTRFS_I(src
)->io_tree
, lock_start
, lock_end
);
3711 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
3712 unlock_extent(&BTRFS_I(inode
)->io_tree
, destoff
,
3716 * Truncate page cache pages so that future reads will see the cloned
3717 * data immediately and not the previous data.
3719 truncate_inode_pages_range(&inode
->i_data
, destoff
,
3720 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
3724 mutex_unlock(&src
->i_mutex
);
3725 mutex_unlock(&inode
->i_mutex
);
3727 mutex_unlock(&inode
->i_mutex
);
3728 mutex_unlock(&src
->i_mutex
);
3731 mutex_unlock(&src
->i_mutex
);
3736 mnt_drop_write_file(file
);
3740 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
3742 struct btrfs_ioctl_clone_range_args args
;
3744 if (copy_from_user(&args
, argp
, sizeof(args
)))
3746 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
3747 args
.src_length
, args
.dest_offset
);
3751 * there are many ways the trans_start and trans_end ioctls can lead
3752 * to deadlocks. They should only be used by applications that
3753 * basically own the machine, and have a very in depth understanding
3754 * of all the possible deadlocks and enospc problems.
3756 static long btrfs_ioctl_trans_start(struct file
*file
)
3758 struct inode
*inode
= file_inode(file
);
3759 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3760 struct btrfs_trans_handle
*trans
;
3764 if (!capable(CAP_SYS_ADMIN
))
3768 if (file
->private_data
)
3772 if (btrfs_root_readonly(root
))
3775 ret
= mnt_want_write_file(file
);
3779 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
3782 trans
= btrfs_start_ioctl_transaction(root
);
3786 file
->private_data
= trans
;
3790 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3791 mnt_drop_write_file(file
);
3796 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
3798 struct inode
*inode
= file_inode(file
);
3799 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3800 struct btrfs_root
*new_root
;
3801 struct btrfs_dir_item
*di
;
3802 struct btrfs_trans_handle
*trans
;
3803 struct btrfs_path
*path
;
3804 struct btrfs_key location
;
3805 struct btrfs_disk_key disk_key
;
3810 if (!capable(CAP_SYS_ADMIN
))
3813 ret
= mnt_want_write_file(file
);
3817 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
3823 objectid
= BTRFS_FS_TREE_OBJECTID
;
3825 location
.objectid
= objectid
;
3826 location
.type
= BTRFS_ROOT_ITEM_KEY
;
3827 location
.offset
= (u64
)-1;
3829 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
3830 if (IS_ERR(new_root
)) {
3831 ret
= PTR_ERR(new_root
);
3835 path
= btrfs_alloc_path();
3840 path
->leave_spinning
= 1;
3842 trans
= btrfs_start_transaction(root
, 1);
3843 if (IS_ERR(trans
)) {
3844 btrfs_free_path(path
);
3845 ret
= PTR_ERR(trans
);
3849 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
3850 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
3851 dir_id
, "default", 7, 1);
3852 if (IS_ERR_OR_NULL(di
)) {
3853 btrfs_free_path(path
);
3854 btrfs_end_transaction(trans
, root
);
3855 btrfs_err(new_root
->fs_info
, "Umm, you don't have the default dir"
3856 "item, this isn't going to work");
3861 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
3862 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
3863 btrfs_mark_buffer_dirty(path
->nodes
[0]);
3864 btrfs_free_path(path
);
3866 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
3867 btrfs_end_transaction(trans
, root
);
3869 mnt_drop_write_file(file
);
3873 void btrfs_get_block_group_info(struct list_head
*groups_list
,
3874 struct btrfs_ioctl_space_info
*space
)
3876 struct btrfs_block_group_cache
*block_group
;
3878 space
->total_bytes
= 0;
3879 space
->used_bytes
= 0;
3881 list_for_each_entry(block_group
, groups_list
, list
) {
3882 space
->flags
= block_group
->flags
;
3883 space
->total_bytes
+= block_group
->key
.offset
;
3884 space
->used_bytes
+=
3885 btrfs_block_group_used(&block_group
->item
);
3889 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
3891 struct btrfs_ioctl_space_args space_args
;
3892 struct btrfs_ioctl_space_info space
;
3893 struct btrfs_ioctl_space_info
*dest
;
3894 struct btrfs_ioctl_space_info
*dest_orig
;
3895 struct btrfs_ioctl_space_info __user
*user_dest
;
3896 struct btrfs_space_info
*info
;
3897 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
3898 BTRFS_BLOCK_GROUP_SYSTEM
,
3899 BTRFS_BLOCK_GROUP_METADATA
,
3900 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
3907 if (copy_from_user(&space_args
,
3908 (struct btrfs_ioctl_space_args __user
*)arg
,
3909 sizeof(space_args
)))
3912 for (i
= 0; i
< num_types
; i
++) {
3913 struct btrfs_space_info
*tmp
;
3917 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3919 if (tmp
->flags
== types
[i
]) {
3929 down_read(&info
->groups_sem
);
3930 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3931 if (!list_empty(&info
->block_groups
[c
]))
3934 up_read(&info
->groups_sem
);
3938 * Global block reserve, exported as a space_info
3942 /* space_slots == 0 means they are asking for a count */
3943 if (space_args
.space_slots
== 0) {
3944 space_args
.total_spaces
= slot_count
;
3948 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
3950 alloc_size
= sizeof(*dest
) * slot_count
;
3952 /* we generally have at most 6 or so space infos, one for each raid
3953 * level. So, a whole page should be more than enough for everyone
3955 if (alloc_size
> PAGE_CACHE_SIZE
)
3958 space_args
.total_spaces
= 0;
3959 dest
= kmalloc(alloc_size
, GFP_NOFS
);
3964 /* now we have a buffer to copy into */
3965 for (i
= 0; i
< num_types
; i
++) {
3966 struct btrfs_space_info
*tmp
;
3973 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3975 if (tmp
->flags
== types
[i
]) {
3984 down_read(&info
->groups_sem
);
3985 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3986 if (!list_empty(&info
->block_groups
[c
])) {
3987 btrfs_get_block_group_info(
3988 &info
->block_groups
[c
], &space
);
3989 memcpy(dest
, &space
, sizeof(space
));
3991 space_args
.total_spaces
++;
3997 up_read(&info
->groups_sem
);
4001 * Add global block reserve
4004 struct btrfs_block_rsv
*block_rsv
= &root
->fs_info
->global_block_rsv
;
4006 spin_lock(&block_rsv
->lock
);
4007 space
.total_bytes
= block_rsv
->size
;
4008 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4009 spin_unlock(&block_rsv
->lock
);
4010 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4011 memcpy(dest
, &space
, sizeof(space
));
4012 space_args
.total_spaces
++;
4015 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4016 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4018 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4023 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4030 * there are many ways the trans_start and trans_end ioctls can lead
4031 * to deadlocks. They should only be used by applications that
4032 * basically own the machine, and have a very in depth understanding
4033 * of all the possible deadlocks and enospc problems.
4035 long btrfs_ioctl_trans_end(struct file
*file
)
4037 struct inode
*inode
= file_inode(file
);
4038 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4039 struct btrfs_trans_handle
*trans
;
4041 trans
= file
->private_data
;
4044 file
->private_data
= NULL
;
4046 btrfs_end_transaction(trans
, root
);
4048 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4050 mnt_drop_write_file(file
);
4054 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4057 struct btrfs_trans_handle
*trans
;
4061 trans
= btrfs_attach_transaction_barrier(root
);
4062 if (IS_ERR(trans
)) {
4063 if (PTR_ERR(trans
) != -ENOENT
)
4064 return PTR_ERR(trans
);
4066 /* No running transaction, don't bother */
4067 transid
= root
->fs_info
->last_trans_committed
;
4070 transid
= trans
->transid
;
4071 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
4073 btrfs_end_transaction(trans
, root
);
4078 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4083 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
4089 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4092 transid
= 0; /* current trans */
4094 return btrfs_wait_for_commit(root
, transid
);
4097 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4099 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4100 struct btrfs_ioctl_scrub_args
*sa
;
4103 if (!capable(CAP_SYS_ADMIN
))
4106 sa
= memdup_user(arg
, sizeof(*sa
));
4110 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4111 ret
= mnt_want_write_file(file
);
4116 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4117 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4120 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4123 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4124 mnt_drop_write_file(file
);
4130 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
4132 if (!capable(CAP_SYS_ADMIN
))
4135 return btrfs_scrub_cancel(root
->fs_info
);
4138 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
4141 struct btrfs_ioctl_scrub_args
*sa
;
4144 if (!capable(CAP_SYS_ADMIN
))
4147 sa
= memdup_user(arg
, sizeof(*sa
));
4151 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
4153 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4160 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
4163 struct btrfs_ioctl_get_dev_stats
*sa
;
4166 sa
= memdup_user(arg
, sizeof(*sa
));
4170 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4175 ret
= btrfs_get_dev_stats(root
, sa
);
4177 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4184 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
4186 struct btrfs_ioctl_dev_replace_args
*p
;
4189 if (!capable(CAP_SYS_ADMIN
))
4192 p
= memdup_user(arg
, sizeof(*p
));
4197 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4198 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
) {
4203 &root
->fs_info
->mutually_exclusive_operation_running
,
4205 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4207 ret
= btrfs_dev_replace_start(root
, p
);
4209 &root
->fs_info
->mutually_exclusive_operation_running
,
4213 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4214 btrfs_dev_replace_status(root
->fs_info
, p
);
4217 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4218 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
4225 if (copy_to_user(arg
, p
, sizeof(*p
)))
4232 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4238 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4239 struct inode_fs_paths
*ipath
= NULL
;
4240 struct btrfs_path
*path
;
4242 if (!capable(CAP_DAC_READ_SEARCH
))
4245 path
= btrfs_alloc_path();
4251 ipa
= memdup_user(arg
, sizeof(*ipa
));
4258 size
= min_t(u32
, ipa
->size
, 4096);
4259 ipath
= init_ipath(size
, root
, path
);
4260 if (IS_ERR(ipath
)) {
4261 ret
= PTR_ERR(ipath
);
4266 ret
= paths_from_inode(ipa
->inum
, ipath
);
4270 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4271 rel_ptr
= ipath
->fspath
->val
[i
] -
4272 (u64
)(unsigned long)ipath
->fspath
->val
;
4273 ipath
->fspath
->val
[i
] = rel_ptr
;
4276 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
4277 (void *)(unsigned long)ipath
->fspath
, size
);
4284 btrfs_free_path(path
);
4291 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4293 struct btrfs_data_container
*inodes
= ctx
;
4294 const size_t c
= 3 * sizeof(u64
);
4296 if (inodes
->bytes_left
>= c
) {
4297 inodes
->bytes_left
-= c
;
4298 inodes
->val
[inodes
->elem_cnt
] = inum
;
4299 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4300 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4301 inodes
->elem_cnt
+= 3;
4303 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4304 inodes
->bytes_left
= 0;
4305 inodes
->elem_missed
+= 3;
4311 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
4316 struct btrfs_ioctl_logical_ino_args
*loi
;
4317 struct btrfs_data_container
*inodes
= NULL
;
4318 struct btrfs_path
*path
= NULL
;
4320 if (!capable(CAP_SYS_ADMIN
))
4323 loi
= memdup_user(arg
, sizeof(*loi
));
4330 path
= btrfs_alloc_path();
4336 size
= min_t(u32
, loi
->size
, 64 * 1024);
4337 inodes
= init_data_container(size
);
4338 if (IS_ERR(inodes
)) {
4339 ret
= PTR_ERR(inodes
);
4344 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
4345 build_ino_list
, inodes
);
4351 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
4352 (void *)(unsigned long)inodes
, size
);
4357 btrfs_free_path(path
);
4364 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
4365 struct btrfs_ioctl_balance_args
*bargs
)
4367 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4369 bargs
->flags
= bctl
->flags
;
4371 if (atomic_read(&fs_info
->balance_running
))
4372 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4373 if (atomic_read(&fs_info
->balance_pause_req
))
4374 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4375 if (atomic_read(&fs_info
->balance_cancel_req
))
4376 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4378 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4379 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4380 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4383 spin_lock(&fs_info
->balance_lock
);
4384 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4385 spin_unlock(&fs_info
->balance_lock
);
4387 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4391 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4393 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4394 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4395 struct btrfs_ioctl_balance_args
*bargs
;
4396 struct btrfs_balance_control
*bctl
;
4397 bool need_unlock
; /* for mut. excl. ops lock */
4400 if (!capable(CAP_SYS_ADMIN
))
4403 ret
= mnt_want_write_file(file
);
4408 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4409 mutex_lock(&fs_info
->volume_mutex
);
4410 mutex_lock(&fs_info
->balance_mutex
);
4416 * mut. excl. ops lock is locked. Three possibilites:
4417 * (1) some other op is running
4418 * (2) balance is running
4419 * (3) balance is paused -- special case (think resume)
4421 mutex_lock(&fs_info
->balance_mutex
);
4422 if (fs_info
->balance_ctl
) {
4423 /* this is either (2) or (3) */
4424 if (!atomic_read(&fs_info
->balance_running
)) {
4425 mutex_unlock(&fs_info
->balance_mutex
);
4426 if (!mutex_trylock(&fs_info
->volume_mutex
))
4428 mutex_lock(&fs_info
->balance_mutex
);
4430 if (fs_info
->balance_ctl
&&
4431 !atomic_read(&fs_info
->balance_running
)) {
4433 need_unlock
= false;
4437 mutex_unlock(&fs_info
->balance_mutex
);
4438 mutex_unlock(&fs_info
->volume_mutex
);
4442 mutex_unlock(&fs_info
->balance_mutex
);
4448 mutex_unlock(&fs_info
->balance_mutex
);
4449 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4454 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4457 bargs
= memdup_user(arg
, sizeof(*bargs
));
4458 if (IS_ERR(bargs
)) {
4459 ret
= PTR_ERR(bargs
);
4463 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4464 if (!fs_info
->balance_ctl
) {
4469 bctl
= fs_info
->balance_ctl
;
4470 spin_lock(&fs_info
->balance_lock
);
4471 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4472 spin_unlock(&fs_info
->balance_lock
);
4480 if (fs_info
->balance_ctl
) {
4485 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
4491 bctl
->fs_info
= fs_info
;
4493 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4494 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4495 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4497 bctl
->flags
= bargs
->flags
;
4499 /* balance everything - no filters */
4500 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4505 * Ownership of bctl and mutually_exclusive_operation_running
4506 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4507 * or, if restriper was paused all the way until unmount, in
4508 * free_fs_info. mutually_exclusive_operation_running is
4509 * cleared in __cancel_balance.
4511 need_unlock
= false;
4513 ret
= btrfs_balance(bctl
, bargs
);
4516 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4523 mutex_unlock(&fs_info
->balance_mutex
);
4524 mutex_unlock(&fs_info
->volume_mutex
);
4526 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4528 mnt_drop_write_file(file
);
4532 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4534 if (!capable(CAP_SYS_ADMIN
))
4538 case BTRFS_BALANCE_CTL_PAUSE
:
4539 return btrfs_pause_balance(root
->fs_info
);
4540 case BTRFS_BALANCE_CTL_CANCEL
:
4541 return btrfs_cancel_balance(root
->fs_info
);
4547 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4550 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4551 struct btrfs_ioctl_balance_args
*bargs
;
4554 if (!capable(CAP_SYS_ADMIN
))
4557 mutex_lock(&fs_info
->balance_mutex
);
4558 if (!fs_info
->balance_ctl
) {
4563 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
4569 update_ioctl_balance_args(fs_info
, 1, bargs
);
4571 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4576 mutex_unlock(&fs_info
->balance_mutex
);
4580 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4582 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4583 struct btrfs_ioctl_quota_ctl_args
*sa
;
4584 struct btrfs_trans_handle
*trans
= NULL
;
4588 if (!capable(CAP_SYS_ADMIN
))
4591 ret
= mnt_want_write_file(file
);
4595 sa
= memdup_user(arg
, sizeof(*sa
));
4601 down_write(&root
->fs_info
->subvol_sem
);
4602 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
4603 if (IS_ERR(trans
)) {
4604 ret
= PTR_ERR(trans
);
4609 case BTRFS_QUOTA_CTL_ENABLE
:
4610 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
4612 case BTRFS_QUOTA_CTL_DISABLE
:
4613 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
4620 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
4625 up_write(&root
->fs_info
->subvol_sem
);
4627 mnt_drop_write_file(file
);
4631 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4633 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4634 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4635 struct btrfs_trans_handle
*trans
;
4639 if (!capable(CAP_SYS_ADMIN
))
4642 ret
= mnt_want_write_file(file
);
4646 sa
= memdup_user(arg
, sizeof(*sa
));
4652 trans
= btrfs_join_transaction(root
);
4653 if (IS_ERR(trans
)) {
4654 ret
= PTR_ERR(trans
);
4658 /* FIXME: check if the IDs really exist */
4660 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
4663 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
4667 err
= btrfs_end_transaction(trans
, root
);
4674 mnt_drop_write_file(file
);
4678 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4680 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4681 struct btrfs_ioctl_qgroup_create_args
*sa
;
4682 struct btrfs_trans_handle
*trans
;
4686 if (!capable(CAP_SYS_ADMIN
))
4689 ret
= mnt_want_write_file(file
);
4693 sa
= memdup_user(arg
, sizeof(*sa
));
4699 if (!sa
->qgroupid
) {
4704 trans
= btrfs_join_transaction(root
);
4705 if (IS_ERR(trans
)) {
4706 ret
= PTR_ERR(trans
);
4710 /* FIXME: check if the IDs really exist */
4712 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
,
4715 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4718 err
= btrfs_end_transaction(trans
, root
);
4725 mnt_drop_write_file(file
);
4729 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4731 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4732 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4733 struct btrfs_trans_handle
*trans
;
4738 if (!capable(CAP_SYS_ADMIN
))
4741 ret
= mnt_want_write_file(file
);
4745 sa
= memdup_user(arg
, sizeof(*sa
));
4751 trans
= btrfs_join_transaction(root
);
4752 if (IS_ERR(trans
)) {
4753 ret
= PTR_ERR(trans
);
4757 qgroupid
= sa
->qgroupid
;
4759 /* take the current subvol as qgroup */
4760 qgroupid
= root
->root_key
.objectid
;
4763 /* FIXME: check if the IDs really exist */
4764 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
4766 err
= btrfs_end_transaction(trans
, root
);
4773 mnt_drop_write_file(file
);
4777 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
4779 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4780 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4783 if (!capable(CAP_SYS_ADMIN
))
4786 ret
= mnt_want_write_file(file
);
4790 qsa
= memdup_user(arg
, sizeof(*qsa
));
4801 ret
= btrfs_qgroup_rescan(root
->fs_info
);
4806 mnt_drop_write_file(file
);
4810 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
4812 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4813 struct btrfs_ioctl_quota_rescan_args
*qsa
;
4816 if (!capable(CAP_SYS_ADMIN
))
4819 qsa
= kzalloc(sizeof(*qsa
), GFP_NOFS
);
4823 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
4825 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
4828 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
4835 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
4837 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4839 if (!capable(CAP_SYS_ADMIN
))
4842 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
4845 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
4846 struct btrfs_ioctl_received_subvol_args
*sa
)
4848 struct inode
*inode
= file_inode(file
);
4849 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4850 struct btrfs_root_item
*root_item
= &root
->root_item
;
4851 struct btrfs_trans_handle
*trans
;
4852 struct timespec ct
= CURRENT_TIME
;
4854 int received_uuid_changed
;
4856 if (!inode_owner_or_capable(inode
))
4859 ret
= mnt_want_write_file(file
);
4863 down_write(&root
->fs_info
->subvol_sem
);
4865 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
4870 if (btrfs_root_readonly(root
)) {
4877 * 2 - uuid items (received uuid + subvol uuid)
4879 trans
= btrfs_start_transaction(root
, 3);
4880 if (IS_ERR(trans
)) {
4881 ret
= PTR_ERR(trans
);
4886 sa
->rtransid
= trans
->transid
;
4887 sa
->rtime
.sec
= ct
.tv_sec
;
4888 sa
->rtime
.nsec
= ct
.tv_nsec
;
4890 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
4892 if (received_uuid_changed
&&
4893 !btrfs_is_empty_uuid(root_item
->received_uuid
))
4894 btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
4895 root_item
->received_uuid
,
4896 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
4897 root
->root_key
.objectid
);
4898 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
4899 btrfs_set_root_stransid(root_item
, sa
->stransid
);
4900 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
4901 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
4902 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
4903 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
4904 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
4906 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
4907 &root
->root_key
, &root
->root_item
);
4909 btrfs_end_transaction(trans
, root
);
4912 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
4913 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
4915 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
4916 root
->root_key
.objectid
);
4917 if (ret
< 0 && ret
!= -EEXIST
) {
4918 btrfs_abort_transaction(trans
, root
, ret
);
4922 ret
= btrfs_commit_transaction(trans
, root
);
4924 btrfs_abort_transaction(trans
, root
, ret
);
4929 up_write(&root
->fs_info
->subvol_sem
);
4930 mnt_drop_write_file(file
);
4935 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
4938 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
4939 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
4942 args32
= memdup_user(arg
, sizeof(*args32
));
4943 if (IS_ERR(args32
)) {
4944 ret
= PTR_ERR(args32
);
4949 args64
= kmalloc(sizeof(*args64
), GFP_NOFS
);
4955 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
4956 args64
->stransid
= args32
->stransid
;
4957 args64
->rtransid
= args32
->rtransid
;
4958 args64
->stime
.sec
= args32
->stime
.sec
;
4959 args64
->stime
.nsec
= args32
->stime
.nsec
;
4960 args64
->rtime
.sec
= args32
->rtime
.sec
;
4961 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
4962 args64
->flags
= args32
->flags
;
4964 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
4968 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
4969 args32
->stransid
= args64
->stransid
;
4970 args32
->rtransid
= args64
->rtransid
;
4971 args32
->stime
.sec
= args64
->stime
.sec
;
4972 args32
->stime
.nsec
= args64
->stime
.nsec
;
4973 args32
->rtime
.sec
= args64
->rtime
.sec
;
4974 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
4975 args32
->flags
= args64
->flags
;
4977 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
4988 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
4991 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
4994 sa
= memdup_user(arg
, sizeof(*sa
));
5001 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5006 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5015 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5017 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5020 char label
[BTRFS_LABEL_SIZE
];
5022 spin_lock(&root
->fs_info
->super_lock
);
5023 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5024 spin_unlock(&root
->fs_info
->super_lock
);
5026 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5028 if (len
== BTRFS_LABEL_SIZE
) {
5029 btrfs_warn(root
->fs_info
,
5030 "label is too long, return the first %zu bytes", --len
);
5033 ret
= copy_to_user(arg
, label
, len
);
5035 return ret
? -EFAULT
: 0;
5038 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5040 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5041 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5042 struct btrfs_trans_handle
*trans
;
5043 char label
[BTRFS_LABEL_SIZE
];
5046 if (!capable(CAP_SYS_ADMIN
))
5049 if (copy_from_user(label
, arg
, sizeof(label
)))
5052 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5053 btrfs_err(root
->fs_info
, "unable to set label with more than %d bytes",
5054 BTRFS_LABEL_SIZE
- 1);
5058 ret
= mnt_want_write_file(file
);
5062 trans
= btrfs_start_transaction(root
, 0);
5063 if (IS_ERR(trans
)) {
5064 ret
= PTR_ERR(trans
);
5068 spin_lock(&root
->fs_info
->super_lock
);
5069 strcpy(super_block
->label
, label
);
5070 spin_unlock(&root
->fs_info
->super_lock
);
5071 ret
= btrfs_commit_transaction(trans
, root
);
5074 mnt_drop_write_file(file
);
5078 #define INIT_FEATURE_FLAGS(suffix) \
5079 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5080 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5081 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5083 static int btrfs_ioctl_get_supported_features(struct file
*file
,
5086 static struct btrfs_ioctl_feature_flags features
[3] = {
5087 INIT_FEATURE_FLAGS(SUPP
),
5088 INIT_FEATURE_FLAGS(SAFE_SET
),
5089 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5092 if (copy_to_user(arg
, &features
, sizeof(features
)))
5098 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5100 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5101 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5102 struct btrfs_ioctl_feature_flags features
;
5104 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5105 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5106 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5108 if (copy_to_user(arg
, &features
, sizeof(features
)))
5114 static int check_feature_bits(struct btrfs_root
*root
,
5115 enum btrfs_feature_set set
,
5116 u64 change_mask
, u64 flags
, u64 supported_flags
,
5117 u64 safe_set
, u64 safe_clear
)
5119 const char *type
= btrfs_feature_set_names
[set
];
5121 u64 disallowed
, unsupported
;
5122 u64 set_mask
= flags
& change_mask
;
5123 u64 clear_mask
= ~flags
& change_mask
;
5125 unsupported
= set_mask
& ~supported_flags
;
5127 names
= btrfs_printable_features(set
, unsupported
);
5129 btrfs_warn(root
->fs_info
,
5130 "this kernel does not support the %s feature bit%s",
5131 names
, strchr(names
, ',') ? "s" : "");
5134 btrfs_warn(root
->fs_info
,
5135 "this kernel does not support %s bits 0x%llx",
5140 disallowed
= set_mask
& ~safe_set
;
5142 names
= btrfs_printable_features(set
, disallowed
);
5144 btrfs_warn(root
->fs_info
,
5145 "can't set the %s feature bit%s while mounted",
5146 names
, strchr(names
, ',') ? "s" : "");
5149 btrfs_warn(root
->fs_info
,
5150 "can't set %s bits 0x%llx while mounted",
5155 disallowed
= clear_mask
& ~safe_clear
;
5157 names
= btrfs_printable_features(set
, disallowed
);
5159 btrfs_warn(root
->fs_info
,
5160 "can't clear the %s feature bit%s while mounted",
5161 names
, strchr(names
, ',') ? "s" : "");
5164 btrfs_warn(root
->fs_info
,
5165 "can't clear %s bits 0x%llx while mounted",
5173 #define check_feature(root, change_mask, flags, mask_base) \
5174 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5175 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5176 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5177 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5179 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5181 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5182 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5183 struct btrfs_ioctl_feature_flags flags
[2];
5184 struct btrfs_trans_handle
*trans
;
5188 if (!capable(CAP_SYS_ADMIN
))
5191 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5195 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5196 !flags
[0].incompat_flags
)
5199 ret
= check_feature(root
, flags
[0].compat_flags
,
5200 flags
[1].compat_flags
, COMPAT
);
5204 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
5205 flags
[1].compat_ro_flags
, COMPAT_RO
);
5209 ret
= check_feature(root
, flags
[0].incompat_flags
,
5210 flags
[1].incompat_flags
, INCOMPAT
);
5214 trans
= btrfs_start_transaction(root
, 0);
5216 return PTR_ERR(trans
);
5218 spin_lock(&root
->fs_info
->super_lock
);
5219 newflags
= btrfs_super_compat_flags(super_block
);
5220 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5221 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5222 btrfs_set_super_compat_flags(super_block
, newflags
);
5224 newflags
= btrfs_super_compat_ro_flags(super_block
);
5225 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5226 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5227 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5229 newflags
= btrfs_super_incompat_flags(super_block
);
5230 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5231 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5232 btrfs_set_super_incompat_flags(super_block
, newflags
);
5233 spin_unlock(&root
->fs_info
->super_lock
);
5235 return btrfs_commit_transaction(trans
, root
);
5238 long btrfs_ioctl(struct file
*file
, unsigned int
5239 cmd
, unsigned long arg
)
5241 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5242 void __user
*argp
= (void __user
*)arg
;
5245 case FS_IOC_GETFLAGS
:
5246 return btrfs_ioctl_getflags(file
, argp
);
5247 case FS_IOC_SETFLAGS
:
5248 return btrfs_ioctl_setflags(file
, argp
);
5249 case FS_IOC_GETVERSION
:
5250 return btrfs_ioctl_getversion(file
, argp
);
5252 return btrfs_ioctl_fitrim(file
, argp
);
5253 case BTRFS_IOC_SNAP_CREATE
:
5254 return btrfs_ioctl_snap_create(file
, argp
, 0);
5255 case BTRFS_IOC_SNAP_CREATE_V2
:
5256 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5257 case BTRFS_IOC_SUBVOL_CREATE
:
5258 return btrfs_ioctl_snap_create(file
, argp
, 1);
5259 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5260 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5261 case BTRFS_IOC_SNAP_DESTROY
:
5262 return btrfs_ioctl_snap_destroy(file
, argp
);
5263 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5264 return btrfs_ioctl_subvol_getflags(file
, argp
);
5265 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5266 return btrfs_ioctl_subvol_setflags(file
, argp
);
5267 case BTRFS_IOC_DEFAULT_SUBVOL
:
5268 return btrfs_ioctl_default_subvol(file
, argp
);
5269 case BTRFS_IOC_DEFRAG
:
5270 return btrfs_ioctl_defrag(file
, NULL
);
5271 case BTRFS_IOC_DEFRAG_RANGE
:
5272 return btrfs_ioctl_defrag(file
, argp
);
5273 case BTRFS_IOC_RESIZE
:
5274 return btrfs_ioctl_resize(file
, argp
);
5275 case BTRFS_IOC_ADD_DEV
:
5276 return btrfs_ioctl_add_dev(root
, argp
);
5277 case BTRFS_IOC_RM_DEV
:
5278 return btrfs_ioctl_rm_dev(file
, argp
);
5279 case BTRFS_IOC_FS_INFO
:
5280 return btrfs_ioctl_fs_info(root
, argp
);
5281 case BTRFS_IOC_DEV_INFO
:
5282 return btrfs_ioctl_dev_info(root
, argp
);
5283 case BTRFS_IOC_BALANCE
:
5284 return btrfs_ioctl_balance(file
, NULL
);
5285 case BTRFS_IOC_CLONE
:
5286 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
5287 case BTRFS_IOC_CLONE_RANGE
:
5288 return btrfs_ioctl_clone_range(file
, argp
);
5289 case BTRFS_IOC_TRANS_START
:
5290 return btrfs_ioctl_trans_start(file
);
5291 case BTRFS_IOC_TRANS_END
:
5292 return btrfs_ioctl_trans_end(file
);
5293 case BTRFS_IOC_TREE_SEARCH
:
5294 return btrfs_ioctl_tree_search(file
, argp
);
5295 case BTRFS_IOC_TREE_SEARCH_V2
:
5296 return btrfs_ioctl_tree_search_v2(file
, argp
);
5297 case BTRFS_IOC_INO_LOOKUP
:
5298 return btrfs_ioctl_ino_lookup(file
, argp
);
5299 case BTRFS_IOC_INO_PATHS
:
5300 return btrfs_ioctl_ino_to_path(root
, argp
);
5301 case BTRFS_IOC_LOGICAL_INO
:
5302 return btrfs_ioctl_logical_to_ino(root
, argp
);
5303 case BTRFS_IOC_SPACE_INFO
:
5304 return btrfs_ioctl_space_info(root
, argp
);
5305 case BTRFS_IOC_SYNC
: {
5308 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
5311 ret
= btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
5314 case BTRFS_IOC_START_SYNC
:
5315 return btrfs_ioctl_start_sync(root
, argp
);
5316 case BTRFS_IOC_WAIT_SYNC
:
5317 return btrfs_ioctl_wait_sync(root
, argp
);
5318 case BTRFS_IOC_SCRUB
:
5319 return btrfs_ioctl_scrub(file
, argp
);
5320 case BTRFS_IOC_SCRUB_CANCEL
:
5321 return btrfs_ioctl_scrub_cancel(root
, argp
);
5322 case BTRFS_IOC_SCRUB_PROGRESS
:
5323 return btrfs_ioctl_scrub_progress(root
, argp
);
5324 case BTRFS_IOC_BALANCE_V2
:
5325 return btrfs_ioctl_balance(file
, argp
);
5326 case BTRFS_IOC_BALANCE_CTL
:
5327 return btrfs_ioctl_balance_ctl(root
, arg
);
5328 case BTRFS_IOC_BALANCE_PROGRESS
:
5329 return btrfs_ioctl_balance_progress(root
, argp
);
5330 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5331 return btrfs_ioctl_set_received_subvol(file
, argp
);
5333 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5334 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5336 case BTRFS_IOC_SEND
:
5337 return btrfs_ioctl_send(file
, argp
);
5338 case BTRFS_IOC_GET_DEV_STATS
:
5339 return btrfs_ioctl_get_dev_stats(root
, argp
);
5340 case BTRFS_IOC_QUOTA_CTL
:
5341 return btrfs_ioctl_quota_ctl(file
, argp
);
5342 case BTRFS_IOC_QGROUP_ASSIGN
:
5343 return btrfs_ioctl_qgroup_assign(file
, argp
);
5344 case BTRFS_IOC_QGROUP_CREATE
:
5345 return btrfs_ioctl_qgroup_create(file
, argp
);
5346 case BTRFS_IOC_QGROUP_LIMIT
:
5347 return btrfs_ioctl_qgroup_limit(file
, argp
);
5348 case BTRFS_IOC_QUOTA_RESCAN
:
5349 return btrfs_ioctl_quota_rescan(file
, argp
);
5350 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5351 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5352 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5353 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5354 case BTRFS_IOC_DEV_REPLACE
:
5355 return btrfs_ioctl_dev_replace(root
, argp
);
5356 case BTRFS_IOC_GET_FSLABEL
:
5357 return btrfs_ioctl_get_fslabel(file
, argp
);
5358 case BTRFS_IOC_SET_FSLABEL
:
5359 return btrfs_ioctl_set_fslabel(file
, argp
);
5360 case BTRFS_IOC_FILE_EXTENT_SAME
:
5361 return btrfs_ioctl_file_extent_same(file
, argp
);
5362 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5363 return btrfs_ioctl_get_supported_features(file
, argp
);
5364 case BTRFS_IOC_GET_FEATURES
:
5365 return btrfs_ioctl_get_features(file
, argp
);
5366 case BTRFS_IOC_SET_FEATURES
:
5367 return btrfs_ioctl_set_features(file
, argp
);