2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
69 struct btrfs_ioctl_timespec_32
{
72 } __attribute__ ((__packed__
));
74 struct btrfs_ioctl_received_subvol_args_32
{
75 char uuid
[BTRFS_UUID_SIZE
]; /* in */
76 __u64 stransid
; /* in */
77 __u64 rtransid
; /* out */
78 struct btrfs_ioctl_timespec_32 stime
; /* in */
79 struct btrfs_ioctl_timespec_32 rtime
; /* out */
81 __u64 reserved
[16]; /* in */
82 } __attribute__ ((__packed__
));
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
89 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
90 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
,
93 /* Mask out flags that are inappropriate for the given type of inode. */
94 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
98 else if (S_ISREG(mode
))
99 return flags
& ~FS_DIRSYNC_FL
;
101 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
105 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
107 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
109 unsigned int iflags
= 0;
111 if (flags
& BTRFS_INODE_SYNC
)
112 iflags
|= FS_SYNC_FL
;
113 if (flags
& BTRFS_INODE_IMMUTABLE
)
114 iflags
|= FS_IMMUTABLE_FL
;
115 if (flags
& BTRFS_INODE_APPEND
)
116 iflags
|= FS_APPEND_FL
;
117 if (flags
& BTRFS_INODE_NODUMP
)
118 iflags
|= FS_NODUMP_FL
;
119 if (flags
& BTRFS_INODE_NOATIME
)
120 iflags
|= FS_NOATIME_FL
;
121 if (flags
& BTRFS_INODE_DIRSYNC
)
122 iflags
|= FS_DIRSYNC_FL
;
123 if (flags
& BTRFS_INODE_NODATACOW
)
124 iflags
|= FS_NOCOW_FL
;
126 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
127 iflags
|= FS_COMPR_FL
;
128 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
129 iflags
|= FS_NOCOMP_FL
;
135 * Update inode->i_flags based on the btrfs internal flags.
137 void btrfs_update_iflags(struct inode
*inode
)
139 struct btrfs_inode
*ip
= BTRFS_I(inode
);
140 unsigned int new_fl
= 0;
142 if (ip
->flags
& BTRFS_INODE_SYNC
)
144 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
145 new_fl
|= S_IMMUTABLE
;
146 if (ip
->flags
& BTRFS_INODE_APPEND
)
148 if (ip
->flags
& BTRFS_INODE_NOATIME
)
150 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
153 set_mask_bits(&inode
->i_flags
,
154 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
,
159 * Inherit flags from the parent inode.
161 * Currently only the compression flags and the cow flags are inherited.
163 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
170 flags
= BTRFS_I(dir
)->flags
;
172 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
173 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
174 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
175 } else if (flags
& BTRFS_INODE_COMPRESS
) {
176 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
177 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
180 if (flags
& BTRFS_INODE_NODATACOW
) {
181 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
182 if (S_ISREG(inode
->i_mode
))
183 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
186 btrfs_update_iflags(inode
);
189 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
191 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
192 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
194 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
199 static int check_flags(unsigned int flags
)
201 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
202 FS_NOATIME_FL
| FS_NODUMP_FL
| \
203 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
204 FS_NOCOMP_FL
| FS_COMPR_FL
|
208 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
214 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
216 struct inode
*inode
= file_inode(file
);
217 struct btrfs_inode
*ip
= BTRFS_I(inode
);
218 struct btrfs_root
*root
= ip
->root
;
219 struct btrfs_trans_handle
*trans
;
220 unsigned int flags
, oldflags
;
223 unsigned int i_oldflags
;
226 if (!inode_owner_or_capable(inode
))
229 if (btrfs_root_readonly(root
))
232 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
235 ret
= check_flags(flags
);
239 ret
= mnt_want_write_file(file
);
243 mutex_lock(&inode
->i_mutex
);
245 ip_oldflags
= ip
->flags
;
246 i_oldflags
= inode
->i_flags
;
247 mode
= inode
->i_mode
;
249 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
250 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
251 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
252 if (!capable(CAP_LINUX_IMMUTABLE
)) {
258 if (flags
& FS_SYNC_FL
)
259 ip
->flags
|= BTRFS_INODE_SYNC
;
261 ip
->flags
&= ~BTRFS_INODE_SYNC
;
262 if (flags
& FS_IMMUTABLE_FL
)
263 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
265 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
266 if (flags
& FS_APPEND_FL
)
267 ip
->flags
|= BTRFS_INODE_APPEND
;
269 ip
->flags
&= ~BTRFS_INODE_APPEND
;
270 if (flags
& FS_NODUMP_FL
)
271 ip
->flags
|= BTRFS_INODE_NODUMP
;
273 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
274 if (flags
& FS_NOATIME_FL
)
275 ip
->flags
|= BTRFS_INODE_NOATIME
;
277 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
278 if (flags
& FS_DIRSYNC_FL
)
279 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
281 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
282 if (flags
& FS_NOCOW_FL
) {
285 * It's safe to turn csums off here, no extents exist.
286 * Otherwise we want the flag to reflect the real COW
287 * status of the file and will not set it.
289 if (inode
->i_size
== 0)
290 ip
->flags
|= BTRFS_INODE_NODATACOW
291 | BTRFS_INODE_NODATASUM
;
293 ip
->flags
|= BTRFS_INODE_NODATACOW
;
297 * Revert back under same assuptions as above
300 if (inode
->i_size
== 0)
301 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
302 | BTRFS_INODE_NODATASUM
);
304 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
309 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
310 * flag may be changed automatically if compression code won't make
313 if (flags
& FS_NOCOMP_FL
) {
314 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
315 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
317 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
318 if (ret
&& ret
!= -ENODATA
)
320 } else if (flags
& FS_COMPR_FL
) {
323 ip
->flags
|= BTRFS_INODE_COMPRESS
;
324 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
326 if (root
->fs_info
->compress_type
== BTRFS_COMPRESS_LZO
)
330 ret
= btrfs_set_prop(inode
, "btrfs.compression",
331 comp
, strlen(comp
), 0);
336 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
337 if (ret
&& ret
!= -ENODATA
)
339 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
342 trans
= btrfs_start_transaction(root
, 1);
344 ret
= PTR_ERR(trans
);
348 btrfs_update_iflags(inode
);
349 inode_inc_iversion(inode
);
350 inode
->i_ctime
= CURRENT_TIME
;
351 ret
= btrfs_update_inode(trans
, root
, inode
);
353 btrfs_end_transaction(trans
, root
);
356 ip
->flags
= ip_oldflags
;
357 inode
->i_flags
= i_oldflags
;
361 mutex_unlock(&inode
->i_mutex
);
362 mnt_drop_write_file(file
);
366 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
368 struct inode
*inode
= file_inode(file
);
370 return put_user(inode
->i_generation
, arg
);
373 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
375 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
376 struct btrfs_device
*device
;
377 struct request_queue
*q
;
378 struct fstrim_range range
;
379 u64 minlen
= ULLONG_MAX
;
381 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
384 if (!capable(CAP_SYS_ADMIN
))
388 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
392 q
= bdev_get_queue(device
->bdev
);
393 if (blk_queue_discard(q
)) {
395 minlen
= min((u64
)q
->limits
.discard_granularity
,
403 if (copy_from_user(&range
, arg
, sizeof(range
)))
405 if (range
.start
> total_bytes
||
406 range
.len
< fs_info
->sb
->s_blocksize
)
409 range
.len
= min(range
.len
, total_bytes
- range
.start
);
410 range
.minlen
= max(range
.minlen
, minlen
);
411 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
415 if (copy_to_user(arg
, &range
, sizeof(range
)))
421 int btrfs_is_empty_uuid(u8
*uuid
)
425 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
432 static noinline
int create_subvol(struct inode
*dir
,
433 struct dentry
*dentry
,
434 char *name
, int namelen
,
436 struct btrfs_qgroup_inherit
*inherit
)
438 struct btrfs_trans_handle
*trans
;
439 struct btrfs_key key
;
440 struct btrfs_root_item root_item
;
441 struct btrfs_inode_item
*inode_item
;
442 struct extent_buffer
*leaf
;
443 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
444 struct btrfs_root
*new_root
;
445 struct btrfs_block_rsv block_rsv
;
446 struct timespec cur_time
= CURRENT_TIME
;
451 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
456 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
461 * Don't create subvolume whose level is not zero. Or qgroup will be
462 * screwed up since it assume subvolme qgroup's level to be 0.
464 if (btrfs_qgroup_level(objectid
))
467 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
469 * The same as the snapshot creation, please see the comment
470 * of create_snapshot().
472 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
473 8, &qgroup_reserved
, false);
477 trans
= btrfs_start_transaction(root
, 0);
479 ret
= PTR_ERR(trans
);
480 btrfs_subvolume_release_metadata(root
, &block_rsv
,
484 trans
->block_rsv
= &block_rsv
;
485 trans
->bytes_reserved
= block_rsv
.size
;
487 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
491 leaf
= btrfs_alloc_tree_block(trans
, root
, 0, objectid
, NULL
, 0, 0, 0);
497 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
498 btrfs_set_header_bytenr(leaf
, leaf
->start
);
499 btrfs_set_header_generation(leaf
, trans
->transid
);
500 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
501 btrfs_set_header_owner(leaf
, objectid
);
503 write_extent_buffer(leaf
, root
->fs_info
->fsid
, btrfs_header_fsid(),
505 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
506 btrfs_header_chunk_tree_uuid(leaf
),
508 btrfs_mark_buffer_dirty(leaf
);
510 memset(&root_item
, 0, sizeof(root_item
));
512 inode_item
= &root_item
.inode
;
513 btrfs_set_stack_inode_generation(inode_item
, 1);
514 btrfs_set_stack_inode_size(inode_item
, 3);
515 btrfs_set_stack_inode_nlink(inode_item
, 1);
516 btrfs_set_stack_inode_nbytes(inode_item
, root
->nodesize
);
517 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
519 btrfs_set_root_flags(&root_item
, 0);
520 btrfs_set_root_limit(&root_item
, 0);
521 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
523 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
524 btrfs_set_root_generation(&root_item
, trans
->transid
);
525 btrfs_set_root_level(&root_item
, 0);
526 btrfs_set_root_refs(&root_item
, 1);
527 btrfs_set_root_used(&root_item
, leaf
->len
);
528 btrfs_set_root_last_snapshot(&root_item
, 0);
530 btrfs_set_root_generation_v2(&root_item
,
531 btrfs_root_generation(&root_item
));
532 uuid_le_gen(&new_uuid
);
533 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
534 btrfs_set_stack_timespec_sec(&root_item
.otime
, cur_time
.tv_sec
);
535 btrfs_set_stack_timespec_nsec(&root_item
.otime
, cur_time
.tv_nsec
);
536 root_item
.ctime
= root_item
.otime
;
537 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
538 btrfs_set_root_otransid(&root_item
, trans
->transid
);
540 btrfs_tree_unlock(leaf
);
541 free_extent_buffer(leaf
);
544 btrfs_set_root_dirid(&root_item
, new_dirid
);
546 key
.objectid
= objectid
;
548 key
.type
= BTRFS_ROOT_ITEM_KEY
;
549 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
554 key
.offset
= (u64
)-1;
555 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
556 if (IS_ERR(new_root
)) {
557 ret
= PTR_ERR(new_root
);
558 btrfs_abort_transaction(trans
, root
, ret
);
562 btrfs_record_root_in_trans(trans
, new_root
);
564 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
566 /* We potentially lose an unused inode item here */
567 btrfs_abort_transaction(trans
, root
, ret
);
572 * insert the directory item
574 ret
= btrfs_set_inode_index(dir
, &index
);
576 btrfs_abort_transaction(trans
, root
, ret
);
580 ret
= btrfs_insert_dir_item(trans
, root
,
581 name
, namelen
, dir
, &key
,
582 BTRFS_FT_DIR
, index
);
584 btrfs_abort_transaction(trans
, root
, ret
);
588 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
589 ret
= btrfs_update_inode(trans
, root
, dir
);
592 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
593 objectid
, root
->root_key
.objectid
,
594 btrfs_ino(dir
), index
, name
, namelen
);
597 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
598 root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
601 btrfs_abort_transaction(trans
, root
, ret
);
604 trans
->block_rsv
= NULL
;
605 trans
->bytes_reserved
= 0;
606 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
609 *async_transid
= trans
->transid
;
610 err
= btrfs_commit_transaction_async(trans
, root
, 1);
612 err
= btrfs_commit_transaction(trans
, root
);
614 err
= btrfs_commit_transaction(trans
, root
);
620 inode
= btrfs_lookup_dentry(dir
, dentry
);
622 return PTR_ERR(inode
);
623 d_instantiate(dentry
, inode
);
628 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root
*root
)
634 prepare_to_wait(&root
->subv_writers
->wait
, &wait
,
635 TASK_UNINTERRUPTIBLE
);
637 writers
= percpu_counter_sum(&root
->subv_writers
->counter
);
641 finish_wait(&root
->subv_writers
->wait
, &wait
);
645 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
646 struct dentry
*dentry
, char *name
, int namelen
,
647 u64
*async_transid
, bool readonly
,
648 struct btrfs_qgroup_inherit
*inherit
)
651 struct btrfs_pending_snapshot
*pending_snapshot
;
652 struct btrfs_trans_handle
*trans
;
655 if (!test_bit(BTRFS_ROOT_REF_COWS
, &root
->state
))
658 atomic_inc(&root
->will_be_snapshoted
);
659 smp_mb__after_atomic();
660 btrfs_wait_for_no_snapshoting_writes(root
);
662 ret
= btrfs_start_delalloc_inodes(root
, 0);
666 btrfs_wait_ordered_extents(root
, -1);
668 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
669 if (!pending_snapshot
) {
674 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
675 BTRFS_BLOCK_RSV_TEMP
);
677 * 1 - parent dir inode
680 * 2 - root ref/backref
681 * 1 - root of snapshot
684 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
685 &pending_snapshot
->block_rsv
, 8,
686 &pending_snapshot
->qgroup_reserved
,
691 pending_snapshot
->dentry
= dentry
;
692 pending_snapshot
->root
= root
;
693 pending_snapshot
->readonly
= readonly
;
694 pending_snapshot
->dir
= dir
;
695 pending_snapshot
->inherit
= inherit
;
697 trans
= btrfs_start_transaction(root
, 0);
699 ret
= PTR_ERR(trans
);
703 spin_lock(&root
->fs_info
->trans_lock
);
704 list_add(&pending_snapshot
->list
,
705 &trans
->transaction
->pending_snapshots
);
706 spin_unlock(&root
->fs_info
->trans_lock
);
708 *async_transid
= trans
->transid
;
709 ret
= btrfs_commit_transaction_async(trans
,
710 root
->fs_info
->extent_root
, 1);
712 ret
= btrfs_commit_transaction(trans
, root
);
714 ret
= btrfs_commit_transaction(trans
,
715 root
->fs_info
->extent_root
);
720 ret
= pending_snapshot
->error
;
724 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
728 inode
= btrfs_lookup_dentry(d_inode(dentry
->d_parent
), dentry
);
730 ret
= PTR_ERR(inode
);
734 d_instantiate(dentry
, inode
);
737 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
738 &pending_snapshot
->block_rsv
,
739 pending_snapshot
->qgroup_reserved
);
741 kfree(pending_snapshot
);
743 if (atomic_dec_and_test(&root
->will_be_snapshoted
))
744 wake_up_atomic_t(&root
->will_be_snapshoted
);
748 /* copy of may_delete in fs/namei.c()
749 * Check whether we can remove a link victim from directory dir, check
750 * whether the type of victim is right.
751 * 1. We can't do it if dir is read-only (done in permission())
752 * 2. We should have write and exec permissions on dir
753 * 3. We can't remove anything from append-only dir
754 * 4. We can't do anything with immutable dir (done in permission())
755 * 5. If the sticky bit on dir is set we should either
756 * a. be owner of dir, or
757 * b. be owner of victim, or
758 * c. have CAP_FOWNER capability
759 * 6. If the victim is append-only or immutable we can't do antyhing with
760 * links pointing to it.
761 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
762 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
763 * 9. We can't remove a root or mountpoint.
764 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
765 * nfs_async_unlink().
768 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
772 if (d_really_is_negative(victim
))
775 BUG_ON(d_inode(victim
->d_parent
) != dir
);
776 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
778 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
783 if (check_sticky(dir
, d_inode(victim
)) || IS_APPEND(d_inode(victim
)) ||
784 IS_IMMUTABLE(d_inode(victim
)) || IS_SWAPFILE(d_inode(victim
)))
787 if (!d_is_dir(victim
))
791 } else if (d_is_dir(victim
))
795 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
800 /* copy of may_create in fs/namei.c() */
801 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
803 if (d_really_is_positive(child
))
807 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
811 * Create a new subvolume below @parent. This is largely modeled after
812 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
813 * inside this filesystem so it's quite a bit simpler.
815 static noinline
int btrfs_mksubvol(struct path
*parent
,
816 char *name
, int namelen
,
817 struct btrfs_root
*snap_src
,
818 u64
*async_transid
, bool readonly
,
819 struct btrfs_qgroup_inherit
*inherit
)
821 struct inode
*dir
= d_inode(parent
->dentry
);
822 struct dentry
*dentry
;
825 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
829 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
830 error
= PTR_ERR(dentry
);
835 if (d_really_is_positive(dentry
))
838 error
= btrfs_may_create(dir
, dentry
);
843 * even if this name doesn't exist, we may get hash collisions.
844 * check for them now when we can safely fail
846 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
852 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
854 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
858 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
859 async_transid
, readonly
, inherit
);
861 error
= create_subvol(dir
, dentry
, name
, namelen
,
862 async_transid
, inherit
);
865 fsnotify_mkdir(dir
, dentry
);
867 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
871 mutex_unlock(&dir
->i_mutex
);
876 * When we're defragging a range, we don't want to kick it off again
877 * if it is really just waiting for delalloc to send it down.
878 * If we find a nice big extent or delalloc range for the bytes in the
879 * file you want to defrag, we return 0 to let you know to skip this
882 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, u32 thresh
)
884 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
885 struct extent_map
*em
= NULL
;
886 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
889 read_lock(&em_tree
->lock
);
890 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
891 read_unlock(&em_tree
->lock
);
894 end
= extent_map_end(em
);
896 if (end
- offset
> thresh
)
899 /* if we already have a nice delalloc here, just stop */
901 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
902 thresh
, EXTENT_DELALLOC
, 1);
909 * helper function to walk through a file and find extents
910 * newer than a specific transid, and smaller than thresh.
912 * This is used by the defragging code to find new and small
915 static int find_new_extents(struct btrfs_root
*root
,
916 struct inode
*inode
, u64 newer_than
,
917 u64
*off
, u32 thresh
)
919 struct btrfs_path
*path
;
920 struct btrfs_key min_key
;
921 struct extent_buffer
*leaf
;
922 struct btrfs_file_extent_item
*extent
;
925 u64 ino
= btrfs_ino(inode
);
927 path
= btrfs_alloc_path();
931 min_key
.objectid
= ino
;
932 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
933 min_key
.offset
= *off
;
936 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
940 if (min_key
.objectid
!= ino
)
942 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
945 leaf
= path
->nodes
[0];
946 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
947 struct btrfs_file_extent_item
);
949 type
= btrfs_file_extent_type(leaf
, extent
);
950 if (type
== BTRFS_FILE_EXTENT_REG
&&
951 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
952 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
953 *off
= min_key
.offset
;
954 btrfs_free_path(path
);
959 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
960 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
964 if (min_key
.offset
== (u64
)-1)
968 btrfs_release_path(path
);
971 btrfs_free_path(path
);
975 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
977 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
978 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
979 struct extent_map
*em
;
980 u64 len
= PAGE_CACHE_SIZE
;
983 * hopefully we have this extent in the tree already, try without
984 * the full extent lock
986 read_lock(&em_tree
->lock
);
987 em
= lookup_extent_mapping(em_tree
, start
, len
);
988 read_unlock(&em_tree
->lock
);
991 struct extent_state
*cached
= NULL
;
992 u64 end
= start
+ len
- 1;
994 /* get the big lock and read metadata off disk */
995 lock_extent_bits(io_tree
, start
, end
, 0, &cached
);
996 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
997 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1006 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1008 struct extent_map
*next
;
1011 /* this is the last extent */
1012 if (em
->start
+ em
->len
>= i_size_read(inode
))
1015 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1016 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
1018 else if ((em
->block_start
+ em
->block_len
== next
->block_start
) &&
1019 (em
->block_len
> 128 * 1024 && next
->block_len
> 128 * 1024))
1022 free_extent_map(next
);
1026 static int should_defrag_range(struct inode
*inode
, u64 start
, u32 thresh
,
1027 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1030 struct extent_map
*em
;
1032 bool next_mergeable
= true;
1033 bool prev_mergeable
= true;
1036 * make sure that once we start defragging an extent, we keep on
1039 if (start
< *defrag_end
)
1044 em
= defrag_lookup_extent(inode
, start
);
1048 /* this will cover holes, and inline extents */
1049 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1055 prev_mergeable
= false;
1057 next_mergeable
= defrag_check_next_extent(inode
, em
);
1059 * we hit a real extent, if it is big or the next extent is not a
1060 * real extent, don't bother defragging it
1062 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1063 (em
->len
>= thresh
|| (!next_mergeable
&& !prev_mergeable
)))
1067 * last_len ends up being a counter of how many bytes we've defragged.
1068 * every time we choose not to defrag an extent, we reset *last_len
1069 * so that the next tiny extent will force a defrag.
1071 * The end result of this is that tiny extents before a single big
1072 * extent will force at least part of that big extent to be defragged.
1075 *defrag_end
= extent_map_end(em
);
1078 *skip
= extent_map_end(em
);
1082 free_extent_map(em
);
1087 * it doesn't do much good to defrag one or two pages
1088 * at a time. This pulls in a nice chunk of pages
1089 * to COW and defrag.
1091 * It also makes sure the delalloc code has enough
1092 * dirty data to avoid making new small extents as part
1095 * It's a good idea to start RA on this range
1096 * before calling this.
1098 static int cluster_pages_for_defrag(struct inode
*inode
,
1099 struct page
**pages
,
1100 unsigned long start_index
,
1101 unsigned long num_pages
)
1103 unsigned long file_end
;
1104 u64 isize
= i_size_read(inode
);
1111 struct btrfs_ordered_extent
*ordered
;
1112 struct extent_state
*cached_state
= NULL
;
1113 struct extent_io_tree
*tree
;
1114 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1116 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1117 if (!isize
|| start_index
> file_end
)
1120 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1122 ret
= btrfs_delalloc_reserve_space(inode
,
1123 start_index
<< PAGE_CACHE_SHIFT
,
1124 page_cnt
<< PAGE_CACHE_SHIFT
);
1128 tree
= &BTRFS_I(inode
)->io_tree
;
1130 /* step one, lock all the pages */
1131 for (i
= 0; i
< page_cnt
; i
++) {
1134 page
= find_or_create_page(inode
->i_mapping
,
1135 start_index
+ i
, mask
);
1139 page_start
= page_offset(page
);
1140 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1142 lock_extent_bits(tree
, page_start
, page_end
,
1144 ordered
= btrfs_lookup_ordered_extent(inode
,
1146 unlock_extent_cached(tree
, page_start
, page_end
,
1147 &cached_state
, GFP_NOFS
);
1152 btrfs_start_ordered_extent(inode
, ordered
, 1);
1153 btrfs_put_ordered_extent(ordered
);
1156 * we unlocked the page above, so we need check if
1157 * it was released or not.
1159 if (page
->mapping
!= inode
->i_mapping
) {
1161 page_cache_release(page
);
1166 if (!PageUptodate(page
)) {
1167 btrfs_readpage(NULL
, page
);
1169 if (!PageUptodate(page
)) {
1171 page_cache_release(page
);
1177 if (page
->mapping
!= inode
->i_mapping
) {
1179 page_cache_release(page
);
1189 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1193 * so now we have a nice long stream of locked
1194 * and up to date pages, lets wait on them
1196 for (i
= 0; i
< i_done
; i
++)
1197 wait_on_page_writeback(pages
[i
]);
1199 page_start
= page_offset(pages
[0]);
1200 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1202 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1203 page_start
, page_end
- 1, 0, &cached_state
);
1204 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1205 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1206 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1207 &cached_state
, GFP_NOFS
);
1209 if (i_done
!= page_cnt
) {
1210 spin_lock(&BTRFS_I(inode
)->lock
);
1211 BTRFS_I(inode
)->outstanding_extents
++;
1212 spin_unlock(&BTRFS_I(inode
)->lock
);
1213 btrfs_delalloc_release_space(inode
,
1214 start_index
<< PAGE_CACHE_SHIFT
,
1215 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1219 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1220 &cached_state
, GFP_NOFS
);
1222 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1223 page_start
, page_end
- 1, &cached_state
,
1226 for (i
= 0; i
< i_done
; i
++) {
1227 clear_page_dirty_for_io(pages
[i
]);
1228 ClearPageChecked(pages
[i
]);
1229 set_page_extent_mapped(pages
[i
]);
1230 set_page_dirty(pages
[i
]);
1231 unlock_page(pages
[i
]);
1232 page_cache_release(pages
[i
]);
1236 for (i
= 0; i
< i_done
; i
++) {
1237 unlock_page(pages
[i
]);
1238 page_cache_release(pages
[i
]);
1240 btrfs_delalloc_release_space(inode
,
1241 start_index
<< PAGE_CACHE_SHIFT
,
1242 page_cnt
<< PAGE_CACHE_SHIFT
);
1247 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1248 struct btrfs_ioctl_defrag_range_args
*range
,
1249 u64 newer_than
, unsigned long max_to_defrag
)
1251 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1252 struct file_ra_state
*ra
= NULL
;
1253 unsigned long last_index
;
1254 u64 isize
= i_size_read(inode
);
1258 u64 newer_off
= range
->start
;
1260 unsigned long ra_index
= 0;
1262 int defrag_count
= 0;
1263 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1264 u32 extent_thresh
= range
->extent_thresh
;
1265 unsigned long max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
1266 unsigned long cluster
= max_cluster
;
1267 u64 new_align
= ~((u64
)128 * 1024 - 1);
1268 struct page
**pages
= NULL
;
1273 if (range
->start
>= isize
)
1276 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1277 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1279 if (range
->compress_type
)
1280 compress_type
= range
->compress_type
;
1283 if (extent_thresh
== 0)
1284 extent_thresh
= 256 * 1024;
1287 * if we were not given a file, allocate a readahead
1291 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1294 file_ra_state_init(ra
, inode
->i_mapping
);
1299 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1306 /* find the last page to defrag */
1307 if (range
->start
+ range
->len
> range
->start
) {
1308 last_index
= min_t(u64
, isize
- 1,
1309 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1311 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1315 ret
= find_new_extents(root
, inode
, newer_than
,
1316 &newer_off
, 64 * 1024);
1318 range
->start
= newer_off
;
1320 * we always align our defrag to help keep
1321 * the extents in the file evenly spaced
1323 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1327 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1330 max_to_defrag
= last_index
- i
+ 1;
1333 * make writeback starts from i, so the defrag range can be
1334 * written sequentially.
1336 if (i
< inode
->i_mapping
->writeback_index
)
1337 inode
->i_mapping
->writeback_index
= i
;
1339 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1340 (i
< DIV_ROUND_UP(i_size_read(inode
), PAGE_CACHE_SIZE
))) {
1342 * make sure we stop running if someone unmounts
1345 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1348 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1349 btrfs_debug(root
->fs_info
, "defrag_file cancelled");
1354 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1355 extent_thresh
, &last_len
, &skip
,
1356 &defrag_end
, range
->flags
&
1357 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1360 * the should_defrag function tells us how much to skip
1361 * bump our counter by the suggested amount
1363 next
= DIV_ROUND_UP(skip
, PAGE_CACHE_SIZE
);
1364 i
= max(i
+ 1, next
);
1369 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1370 PAGE_CACHE_SHIFT
) - i
;
1371 cluster
= min(cluster
, max_cluster
);
1373 cluster
= max_cluster
;
1376 if (i
+ cluster
> ra_index
) {
1377 ra_index
= max(i
, ra_index
);
1378 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1380 ra_index
+= cluster
;
1383 mutex_lock(&inode
->i_mutex
);
1384 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1385 BTRFS_I(inode
)->force_compress
= compress_type
;
1386 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1388 mutex_unlock(&inode
->i_mutex
);
1392 defrag_count
+= ret
;
1393 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1394 mutex_unlock(&inode
->i_mutex
);
1397 if (newer_off
== (u64
)-1)
1403 newer_off
= max(newer_off
+ 1,
1404 (u64
)i
<< PAGE_CACHE_SHIFT
);
1406 ret
= find_new_extents(root
, inode
,
1407 newer_than
, &newer_off
,
1410 range
->start
= newer_off
;
1411 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1418 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1426 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1427 filemap_flush(inode
->i_mapping
);
1428 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1429 &BTRFS_I(inode
)->runtime_flags
))
1430 filemap_flush(inode
->i_mapping
);
1433 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1434 /* the filemap_flush will queue IO into the worker threads, but
1435 * we have to make sure the IO is actually started and that
1436 * ordered extents get created before we return
1438 atomic_inc(&root
->fs_info
->async_submit_draining
);
1439 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1440 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1441 wait_event(root
->fs_info
->async_submit_wait
,
1442 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1443 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1445 atomic_dec(&root
->fs_info
->async_submit_draining
);
1448 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1449 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1455 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1456 mutex_lock(&inode
->i_mutex
);
1457 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1458 mutex_unlock(&inode
->i_mutex
);
1466 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1472 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1473 struct btrfs_ioctl_vol_args
*vol_args
;
1474 struct btrfs_trans_handle
*trans
;
1475 struct btrfs_device
*device
= NULL
;
1478 char *devstr
= NULL
;
1482 if (!capable(CAP_SYS_ADMIN
))
1485 ret
= mnt_want_write_file(file
);
1489 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1491 mnt_drop_write_file(file
);
1492 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1495 mutex_lock(&root
->fs_info
->volume_mutex
);
1496 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1497 if (IS_ERR(vol_args
)) {
1498 ret
= PTR_ERR(vol_args
);
1502 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1504 sizestr
= vol_args
->name
;
1505 devstr
= strchr(sizestr
, ':');
1507 sizestr
= devstr
+ 1;
1509 devstr
= vol_args
->name
;
1510 ret
= kstrtoull(devstr
, 10, &devid
);
1517 btrfs_info(root
->fs_info
, "resizing devid %llu", devid
);
1520 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1522 btrfs_info(root
->fs_info
, "resizer unable to find device %llu",
1528 if (!device
->writeable
) {
1529 btrfs_info(root
->fs_info
,
1530 "resizer unable to apply on readonly device %llu",
1536 if (!strcmp(sizestr
, "max"))
1537 new_size
= device
->bdev
->bd_inode
->i_size
;
1539 if (sizestr
[0] == '-') {
1542 } else if (sizestr
[0] == '+') {
1546 new_size
= memparse(sizestr
, &retptr
);
1547 if (*retptr
!= '\0' || new_size
== 0) {
1553 if (device
->is_tgtdev_for_dev_replace
) {
1558 old_size
= btrfs_device_get_total_bytes(device
);
1561 if (new_size
> old_size
) {
1565 new_size
= old_size
- new_size
;
1566 } else if (mod
> 0) {
1567 if (new_size
> ULLONG_MAX
- old_size
) {
1571 new_size
= old_size
+ new_size
;
1574 if (new_size
< 256 * 1024 * 1024) {
1578 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1583 new_size
= div_u64(new_size
, root
->sectorsize
);
1584 new_size
*= root
->sectorsize
;
1586 btrfs_info_in_rcu(root
->fs_info
, "new size for %s is %llu",
1587 rcu_str_deref(device
->name
), new_size
);
1589 if (new_size
> old_size
) {
1590 trans
= btrfs_start_transaction(root
, 0);
1591 if (IS_ERR(trans
)) {
1592 ret
= PTR_ERR(trans
);
1595 ret
= btrfs_grow_device(trans
, device
, new_size
);
1596 btrfs_commit_transaction(trans
, root
);
1597 } else if (new_size
< old_size
) {
1598 ret
= btrfs_shrink_device(device
, new_size
);
1599 } /* equal, nothing need to do */
1604 mutex_unlock(&root
->fs_info
->volume_mutex
);
1605 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1606 mnt_drop_write_file(file
);
1610 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1611 char *name
, unsigned long fd
, int subvol
,
1612 u64
*transid
, bool readonly
,
1613 struct btrfs_qgroup_inherit
*inherit
)
1618 ret
= mnt_want_write_file(file
);
1622 namelen
= strlen(name
);
1623 if (strchr(name
, '/')) {
1625 goto out_drop_write
;
1628 if (name
[0] == '.' &&
1629 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1631 goto out_drop_write
;
1635 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1636 NULL
, transid
, readonly
, inherit
);
1638 struct fd src
= fdget(fd
);
1639 struct inode
*src_inode
;
1642 goto out_drop_write
;
1645 src_inode
= file_inode(src
.file
);
1646 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1647 btrfs_info(BTRFS_I(src_inode
)->root
->fs_info
,
1648 "Snapshot src from another FS");
1650 } else if (!inode_owner_or_capable(src_inode
)) {
1652 * Subvolume creation is not restricted, but snapshots
1653 * are limited to own subvolumes only
1657 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1658 BTRFS_I(src_inode
)->root
,
1659 transid
, readonly
, inherit
);
1664 mnt_drop_write_file(file
);
1669 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1670 void __user
*arg
, int subvol
)
1672 struct btrfs_ioctl_vol_args
*vol_args
;
1675 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1676 if (IS_ERR(vol_args
))
1677 return PTR_ERR(vol_args
);
1678 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1680 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1681 vol_args
->fd
, subvol
,
1688 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1689 void __user
*arg
, int subvol
)
1691 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1695 bool readonly
= false;
1696 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1698 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1699 if (IS_ERR(vol_args
))
1700 return PTR_ERR(vol_args
);
1701 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1703 if (vol_args
->flags
&
1704 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1705 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1710 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1712 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1714 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1715 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1719 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1720 if (IS_ERR(inherit
)) {
1721 ret
= PTR_ERR(inherit
);
1726 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1727 vol_args
->fd
, subvol
, ptr
,
1732 if (ptr
&& copy_to_user(arg
+
1733 offsetof(struct btrfs_ioctl_vol_args_v2
,
1745 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1748 struct inode
*inode
= file_inode(file
);
1749 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1753 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1756 down_read(&root
->fs_info
->subvol_sem
);
1757 if (btrfs_root_readonly(root
))
1758 flags
|= BTRFS_SUBVOL_RDONLY
;
1759 up_read(&root
->fs_info
->subvol_sem
);
1761 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1767 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1770 struct inode
*inode
= file_inode(file
);
1771 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1772 struct btrfs_trans_handle
*trans
;
1777 if (!inode_owner_or_capable(inode
))
1780 ret
= mnt_want_write_file(file
);
1784 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1786 goto out_drop_write
;
1789 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1791 goto out_drop_write
;
1794 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1796 goto out_drop_write
;
1799 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1801 goto out_drop_write
;
1804 down_write(&root
->fs_info
->subvol_sem
);
1807 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1810 root_flags
= btrfs_root_flags(&root
->root_item
);
1811 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1812 btrfs_set_root_flags(&root
->root_item
,
1813 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1816 * Block RO -> RW transition if this subvolume is involved in
1819 spin_lock(&root
->root_item_lock
);
1820 if (root
->send_in_progress
== 0) {
1821 btrfs_set_root_flags(&root
->root_item
,
1822 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1823 spin_unlock(&root
->root_item_lock
);
1825 spin_unlock(&root
->root_item_lock
);
1826 btrfs_warn(root
->fs_info
,
1827 "Attempt to set subvolume %llu read-write during send",
1828 root
->root_key
.objectid
);
1834 trans
= btrfs_start_transaction(root
, 1);
1835 if (IS_ERR(trans
)) {
1836 ret
= PTR_ERR(trans
);
1840 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1841 &root
->root_key
, &root
->root_item
);
1843 btrfs_commit_transaction(trans
, root
);
1846 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1848 up_write(&root
->fs_info
->subvol_sem
);
1850 mnt_drop_write_file(file
);
1856 * helper to check if the subvolume references other subvolumes
1858 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1860 struct btrfs_path
*path
;
1861 struct btrfs_dir_item
*di
;
1862 struct btrfs_key key
;
1866 path
= btrfs_alloc_path();
1870 /* Make sure this root isn't set as the default subvol */
1871 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
1872 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
, path
,
1873 dir_id
, "default", 7, 0);
1874 if (di
&& !IS_ERR(di
)) {
1875 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1876 if (key
.objectid
== root
->root_key
.objectid
) {
1878 btrfs_err(root
->fs_info
, "deleting default subvolume "
1879 "%llu is not allowed", key
.objectid
);
1882 btrfs_release_path(path
);
1885 key
.objectid
= root
->root_key
.objectid
;
1886 key
.type
= BTRFS_ROOT_REF_KEY
;
1887 key
.offset
= (u64
)-1;
1889 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1896 if (path
->slots
[0] > 0) {
1898 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1899 if (key
.objectid
== root
->root_key
.objectid
&&
1900 key
.type
== BTRFS_ROOT_REF_KEY
)
1904 btrfs_free_path(path
);
1908 static noinline
int key_in_sk(struct btrfs_key
*key
,
1909 struct btrfs_ioctl_search_key
*sk
)
1911 struct btrfs_key test
;
1914 test
.objectid
= sk
->min_objectid
;
1915 test
.type
= sk
->min_type
;
1916 test
.offset
= sk
->min_offset
;
1918 ret
= btrfs_comp_cpu_keys(key
, &test
);
1922 test
.objectid
= sk
->max_objectid
;
1923 test
.type
= sk
->max_type
;
1924 test
.offset
= sk
->max_offset
;
1926 ret
= btrfs_comp_cpu_keys(key
, &test
);
1932 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1933 struct btrfs_path
*path
,
1934 struct btrfs_key
*key
,
1935 struct btrfs_ioctl_search_key
*sk
,
1938 unsigned long *sk_offset
,
1942 struct extent_buffer
*leaf
;
1943 struct btrfs_ioctl_search_header sh
;
1944 struct btrfs_key test
;
1945 unsigned long item_off
;
1946 unsigned long item_len
;
1952 leaf
= path
->nodes
[0];
1953 slot
= path
->slots
[0];
1954 nritems
= btrfs_header_nritems(leaf
);
1956 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1960 found_transid
= btrfs_header_generation(leaf
);
1962 for (i
= slot
; i
< nritems
; i
++) {
1963 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1964 item_len
= btrfs_item_size_nr(leaf
, i
);
1966 btrfs_item_key_to_cpu(leaf
, key
, i
);
1967 if (!key_in_sk(key
, sk
))
1970 if (sizeof(sh
) + item_len
> *buf_size
) {
1977 * return one empty item back for v1, which does not
1981 *buf_size
= sizeof(sh
) + item_len
;
1986 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
1991 sh
.objectid
= key
->objectid
;
1992 sh
.offset
= key
->offset
;
1993 sh
.type
= key
->type
;
1995 sh
.transid
= found_transid
;
1997 /* copy search result header */
1998 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2003 *sk_offset
+= sizeof(sh
);
2006 char __user
*up
= ubuf
+ *sk_offset
;
2008 if (read_extent_buffer_to_user(leaf
, up
,
2009 item_off
, item_len
)) {
2014 *sk_offset
+= item_len
;
2018 if (ret
) /* -EOVERFLOW from above */
2021 if (*num_found
>= sk
->nr_items
) {
2028 test
.objectid
= sk
->max_objectid
;
2029 test
.type
= sk
->max_type
;
2030 test
.offset
= sk
->max_offset
;
2031 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
2033 else if (key
->offset
< (u64
)-1)
2035 else if (key
->type
< (u8
)-1) {
2038 } else if (key
->objectid
< (u64
)-1) {
2046 * 0: all items from this leaf copied, continue with next
2047 * 1: * more items can be copied, but unused buffer is too small
2048 * * all items were found
2049 * Either way, it will stops the loop which iterates to the next
2051 * -EOVERFLOW: item was to large for buffer
2052 * -EFAULT: could not copy extent buffer back to userspace
2057 static noinline
int search_ioctl(struct inode
*inode
,
2058 struct btrfs_ioctl_search_key
*sk
,
2062 struct btrfs_root
*root
;
2063 struct btrfs_key key
;
2064 struct btrfs_path
*path
;
2065 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
2068 unsigned long sk_offset
= 0;
2070 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2071 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2075 path
= btrfs_alloc_path();
2079 if (sk
->tree_id
== 0) {
2080 /* search the root of the inode that was passed */
2081 root
= BTRFS_I(inode
)->root
;
2083 key
.objectid
= sk
->tree_id
;
2084 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2085 key
.offset
= (u64
)-1;
2086 root
= btrfs_read_fs_root_no_name(info
, &key
);
2088 btrfs_err(info
, "could not find root %llu",
2090 btrfs_free_path(path
);
2095 key
.objectid
= sk
->min_objectid
;
2096 key
.type
= sk
->min_type
;
2097 key
.offset
= sk
->min_offset
;
2100 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2106 ret
= copy_to_sk(root
, path
, &key
, sk
, buf_size
, ubuf
,
2107 &sk_offset
, &num_found
);
2108 btrfs_release_path(path
);
2116 sk
->nr_items
= num_found
;
2117 btrfs_free_path(path
);
2121 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2124 struct btrfs_ioctl_search_args __user
*uargs
;
2125 struct btrfs_ioctl_search_key sk
;
2126 struct inode
*inode
;
2130 if (!capable(CAP_SYS_ADMIN
))
2133 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2135 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2138 buf_size
= sizeof(uargs
->buf
);
2140 inode
= file_inode(file
);
2141 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2144 * In the origin implementation an overflow is handled by returning a
2145 * search header with a len of zero, so reset ret.
2147 if (ret
== -EOVERFLOW
)
2150 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2155 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2158 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2159 struct btrfs_ioctl_search_args_v2 args
;
2160 struct inode
*inode
;
2163 const size_t buf_limit
= 16 * 1024 * 1024;
2165 if (!capable(CAP_SYS_ADMIN
))
2168 /* copy search header and buffer size */
2169 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2170 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2173 buf_size
= args
.buf_size
;
2175 if (buf_size
< sizeof(struct btrfs_ioctl_search_header
))
2178 /* limit result size to 16MB */
2179 if (buf_size
> buf_limit
)
2180 buf_size
= buf_limit
;
2182 inode
= file_inode(file
);
2183 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2184 (char *)(&uarg
->buf
[0]));
2185 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2187 else if (ret
== -EOVERFLOW
&&
2188 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2195 * Search INODE_REFs to identify path name of 'dirid' directory
2196 * in a 'tree_id' tree. and sets path name to 'name'.
2198 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2199 u64 tree_id
, u64 dirid
, char *name
)
2201 struct btrfs_root
*root
;
2202 struct btrfs_key key
;
2208 struct btrfs_inode_ref
*iref
;
2209 struct extent_buffer
*l
;
2210 struct btrfs_path
*path
;
2212 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2217 path
= btrfs_alloc_path();
2221 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2223 key
.objectid
= tree_id
;
2224 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2225 key
.offset
= (u64
)-1;
2226 root
= btrfs_read_fs_root_no_name(info
, &key
);
2228 btrfs_err(info
, "could not find root %llu", tree_id
);
2233 key
.objectid
= dirid
;
2234 key
.type
= BTRFS_INODE_REF_KEY
;
2235 key
.offset
= (u64
)-1;
2238 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2242 ret
= btrfs_previous_item(root
, path
, dirid
,
2243 BTRFS_INODE_REF_KEY
);
2253 slot
= path
->slots
[0];
2254 btrfs_item_key_to_cpu(l
, &key
, slot
);
2256 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2257 len
= btrfs_inode_ref_name_len(l
, iref
);
2259 total_len
+= len
+ 1;
2261 ret
= -ENAMETOOLONG
;
2266 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2268 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2271 btrfs_release_path(path
);
2272 key
.objectid
= key
.offset
;
2273 key
.offset
= (u64
)-1;
2274 dirid
= key
.objectid
;
2276 memmove(name
, ptr
, total_len
);
2277 name
[total_len
] = '\0';
2280 btrfs_free_path(path
);
2284 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2287 struct btrfs_ioctl_ino_lookup_args
*args
;
2288 struct inode
*inode
;
2291 args
= memdup_user(argp
, sizeof(*args
));
2293 return PTR_ERR(args
);
2295 inode
= file_inode(file
);
2298 * Unprivileged query to obtain the containing subvolume root id. The
2299 * path is reset so it's consistent with btrfs_search_path_in_tree.
2301 if (args
->treeid
== 0)
2302 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2304 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2309 if (!capable(CAP_SYS_ADMIN
)) {
2314 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2315 args
->treeid
, args
->objectid
,
2319 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2326 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2329 struct dentry
*parent
= file
->f_path
.dentry
;
2330 struct dentry
*dentry
;
2331 struct inode
*dir
= d_inode(parent
);
2332 struct inode
*inode
;
2333 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2334 struct btrfs_root
*dest
= NULL
;
2335 struct btrfs_ioctl_vol_args
*vol_args
;
2336 struct btrfs_trans_handle
*trans
;
2337 struct btrfs_block_rsv block_rsv
;
2339 u64 qgroup_reserved
;
2344 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2345 if (IS_ERR(vol_args
))
2346 return PTR_ERR(vol_args
);
2348 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2349 namelen
= strlen(vol_args
->name
);
2350 if (strchr(vol_args
->name
, '/') ||
2351 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2356 err
= mnt_want_write_file(file
);
2361 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2363 goto out_drop_write
;
2364 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2365 if (IS_ERR(dentry
)) {
2366 err
= PTR_ERR(dentry
);
2367 goto out_unlock_dir
;
2370 if (d_really_is_negative(dentry
)) {
2375 inode
= d_inode(dentry
);
2376 dest
= BTRFS_I(inode
)->root
;
2377 if (!capable(CAP_SYS_ADMIN
)) {
2379 * Regular user. Only allow this with a special mount
2380 * option, when the user has write+exec access to the
2381 * subvol root, and when rmdir(2) would have been
2384 * Note that this is _not_ check that the subvol is
2385 * empty or doesn't contain data that we wouldn't
2386 * otherwise be able to delete.
2388 * Users who want to delete empty subvols should try
2392 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2396 * Do not allow deletion if the parent dir is the same
2397 * as the dir to be deleted. That means the ioctl
2398 * must be called on the dentry referencing the root
2399 * of the subvol, not a random directory contained
2406 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2411 /* check if subvolume may be deleted by a user */
2412 err
= btrfs_may_delete(dir
, dentry
, 1);
2416 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2421 mutex_lock(&inode
->i_mutex
);
2424 * Don't allow to delete a subvolume with send in progress. This is
2425 * inside the i_mutex so the error handling that has to drop the bit
2426 * again is not run concurrently.
2428 spin_lock(&dest
->root_item_lock
);
2429 root_flags
= btrfs_root_flags(&dest
->root_item
);
2430 if (dest
->send_in_progress
== 0) {
2431 btrfs_set_root_flags(&dest
->root_item
,
2432 root_flags
| BTRFS_ROOT_SUBVOL_DEAD
);
2433 spin_unlock(&dest
->root_item_lock
);
2435 spin_unlock(&dest
->root_item_lock
);
2436 btrfs_warn(root
->fs_info
,
2437 "Attempt to delete subvolume %llu during send",
2438 dest
->root_key
.objectid
);
2440 goto out_unlock_inode
;
2443 down_write(&root
->fs_info
->subvol_sem
);
2445 err
= may_destroy_subvol(dest
);
2449 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2451 * One for dir inode, two for dir entries, two for root
2454 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2455 5, &qgroup_reserved
, true);
2459 trans
= btrfs_start_transaction(root
, 0);
2460 if (IS_ERR(trans
)) {
2461 err
= PTR_ERR(trans
);
2464 trans
->block_rsv
= &block_rsv
;
2465 trans
->bytes_reserved
= block_rsv
.size
;
2467 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2468 dest
->root_key
.objectid
,
2469 dentry
->d_name
.name
,
2470 dentry
->d_name
.len
);
2473 btrfs_abort_transaction(trans
, root
, ret
);
2477 btrfs_record_root_in_trans(trans
, dest
);
2479 memset(&dest
->root_item
.drop_progress
, 0,
2480 sizeof(dest
->root_item
.drop_progress
));
2481 dest
->root_item
.drop_level
= 0;
2482 btrfs_set_root_refs(&dest
->root_item
, 0);
2484 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED
, &dest
->state
)) {
2485 ret
= btrfs_insert_orphan_item(trans
,
2486 root
->fs_info
->tree_root
,
2487 dest
->root_key
.objectid
);
2489 btrfs_abort_transaction(trans
, root
, ret
);
2495 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2496 dest
->root_item
.uuid
, BTRFS_UUID_KEY_SUBVOL
,
2497 dest
->root_key
.objectid
);
2498 if (ret
&& ret
!= -ENOENT
) {
2499 btrfs_abort_transaction(trans
, root
, ret
);
2503 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2504 ret
= btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
2505 dest
->root_item
.received_uuid
,
2506 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2507 dest
->root_key
.objectid
);
2508 if (ret
&& ret
!= -ENOENT
) {
2509 btrfs_abort_transaction(trans
, root
, ret
);
2516 trans
->block_rsv
= NULL
;
2517 trans
->bytes_reserved
= 0;
2518 ret
= btrfs_end_transaction(trans
, root
);
2521 inode
->i_flags
|= S_DEAD
;
2523 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2525 up_write(&root
->fs_info
->subvol_sem
);
2527 spin_lock(&dest
->root_item_lock
);
2528 root_flags
= btrfs_root_flags(&dest
->root_item
);
2529 btrfs_set_root_flags(&dest
->root_item
,
2530 root_flags
& ~BTRFS_ROOT_SUBVOL_DEAD
);
2531 spin_unlock(&dest
->root_item_lock
);
2534 mutex_unlock(&inode
->i_mutex
);
2536 d_invalidate(dentry
);
2537 btrfs_invalidate_inodes(dest
);
2539 ASSERT(dest
->send_in_progress
== 0);
2542 if (dest
->ino_cache_inode
) {
2543 iput(dest
->ino_cache_inode
);
2544 dest
->ino_cache_inode
= NULL
;
2550 mutex_unlock(&dir
->i_mutex
);
2552 mnt_drop_write_file(file
);
2558 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2560 struct inode
*inode
= file_inode(file
);
2561 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2562 struct btrfs_ioctl_defrag_range_args
*range
;
2565 ret
= mnt_want_write_file(file
);
2569 if (btrfs_root_readonly(root
)) {
2574 switch (inode
->i_mode
& S_IFMT
) {
2576 if (!capable(CAP_SYS_ADMIN
)) {
2580 ret
= btrfs_defrag_root(root
);
2583 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2586 if (!(file
->f_mode
& FMODE_WRITE
)) {
2591 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2598 if (copy_from_user(range
, argp
,
2604 /* compression requires us to start the IO */
2605 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2606 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2607 range
->extent_thresh
= (u32
)-1;
2610 /* the rest are all set to zero by kzalloc */
2611 range
->len
= (u64
)-1;
2613 ret
= btrfs_defrag_file(file_inode(file
), file
,
2623 mnt_drop_write_file(file
);
2627 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2629 struct btrfs_ioctl_vol_args
*vol_args
;
2632 if (!capable(CAP_SYS_ADMIN
))
2635 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2637 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2640 mutex_lock(&root
->fs_info
->volume_mutex
);
2641 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2642 if (IS_ERR(vol_args
)) {
2643 ret
= PTR_ERR(vol_args
);
2647 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2648 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2651 btrfs_info(root
->fs_info
, "disk added %s",vol_args
->name
);
2655 mutex_unlock(&root
->fs_info
->volume_mutex
);
2656 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2660 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2662 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2663 struct btrfs_ioctl_vol_args
*vol_args
;
2666 if (!capable(CAP_SYS_ADMIN
))
2669 ret
= mnt_want_write_file(file
);
2673 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2674 if (IS_ERR(vol_args
)) {
2675 ret
= PTR_ERR(vol_args
);
2679 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2681 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2683 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2687 mutex_lock(&root
->fs_info
->volume_mutex
);
2688 ret
= btrfs_rm_device(root
, vol_args
->name
);
2689 mutex_unlock(&root
->fs_info
->volume_mutex
);
2690 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2693 btrfs_info(root
->fs_info
, "disk deleted %s",vol_args
->name
);
2698 mnt_drop_write_file(file
);
2702 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2704 struct btrfs_ioctl_fs_info_args
*fi_args
;
2705 struct btrfs_device
*device
;
2706 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2709 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2713 mutex_lock(&fs_devices
->device_list_mutex
);
2714 fi_args
->num_devices
= fs_devices
->num_devices
;
2715 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2717 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
2718 if (device
->devid
> fi_args
->max_id
)
2719 fi_args
->max_id
= device
->devid
;
2721 mutex_unlock(&fs_devices
->device_list_mutex
);
2723 fi_args
->nodesize
= root
->fs_info
->super_copy
->nodesize
;
2724 fi_args
->sectorsize
= root
->fs_info
->super_copy
->sectorsize
;
2725 fi_args
->clone_alignment
= root
->fs_info
->super_copy
->sectorsize
;
2727 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2734 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2736 struct btrfs_ioctl_dev_info_args
*di_args
;
2737 struct btrfs_device
*dev
;
2738 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2740 char *s_uuid
= NULL
;
2742 di_args
= memdup_user(arg
, sizeof(*di_args
));
2743 if (IS_ERR(di_args
))
2744 return PTR_ERR(di_args
);
2746 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2747 s_uuid
= di_args
->uuid
;
2749 mutex_lock(&fs_devices
->device_list_mutex
);
2750 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2757 di_args
->devid
= dev
->devid
;
2758 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
2759 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
2760 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2762 struct rcu_string
*name
;
2765 name
= rcu_dereference(dev
->name
);
2766 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2768 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2770 di_args
->path
[0] = '\0';
2774 mutex_unlock(&fs_devices
->device_list_mutex
);
2775 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2782 static struct page
*extent_same_get_page(struct inode
*inode
, pgoff_t index
)
2785 struct extent_io_tree
*tree
= &BTRFS_I(inode
)->io_tree
;
2787 page
= grab_cache_page(inode
->i_mapping
, index
);
2791 if (!PageUptodate(page
)) {
2792 if (extent_read_full_page_nolock(tree
, page
, btrfs_get_extent
,
2796 if (!PageUptodate(page
)) {
2798 page_cache_release(page
);
2807 static int gather_extent_pages(struct inode
*inode
, struct page
**pages
,
2808 int num_pages
, u64 off
)
2811 pgoff_t index
= off
>> PAGE_CACHE_SHIFT
;
2813 for (i
= 0; i
< num_pages
; i
++) {
2814 pages
[i
] = extent_same_get_page(inode
, index
+ i
);
2821 static inline void lock_extent_range(struct inode
*inode
, u64 off
, u64 len
)
2823 /* do any pending delalloc/csum calc on src, one way or
2824 another, and lock file content */
2826 struct btrfs_ordered_extent
*ordered
;
2827 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2828 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2831 ordered
->file_offset
+ ordered
->len
<= off
||
2832 ordered
->file_offset
>= off
+ len
) &&
2833 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2834 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
)) {
2836 btrfs_put_ordered_extent(ordered
);
2839 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2841 btrfs_put_ordered_extent(ordered
);
2842 btrfs_wait_ordered_range(inode
, off
, len
);
2846 static void btrfs_double_inode_unlock(struct inode
*inode1
, struct inode
*inode2
)
2848 mutex_unlock(&inode1
->i_mutex
);
2849 mutex_unlock(&inode2
->i_mutex
);
2852 static void btrfs_double_inode_lock(struct inode
*inode1
, struct inode
*inode2
)
2854 if (inode1
< inode2
)
2855 swap(inode1
, inode2
);
2857 mutex_lock_nested(&inode1
->i_mutex
, I_MUTEX_PARENT
);
2858 mutex_lock_nested(&inode2
->i_mutex
, I_MUTEX_CHILD
);
2861 static void btrfs_double_extent_unlock(struct inode
*inode1
, u64 loff1
,
2862 struct inode
*inode2
, u64 loff2
, u64 len
)
2864 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2865 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2868 static void btrfs_double_extent_lock(struct inode
*inode1
, u64 loff1
,
2869 struct inode
*inode2
, u64 loff2
, u64 len
)
2871 if (inode1
< inode2
) {
2872 swap(inode1
, inode2
);
2875 lock_extent_range(inode1
, loff1
, len
);
2876 lock_extent_range(inode2
, loff2
, len
);
2881 struct page
**src_pages
;
2882 struct page
**dst_pages
;
2885 static void btrfs_cmp_data_free(struct cmp_pages
*cmp
)
2890 for (i
= 0; i
< cmp
->num_pages
; i
++) {
2891 pg
= cmp
->src_pages
[i
];
2893 page_cache_release(pg
);
2894 pg
= cmp
->dst_pages
[i
];
2896 page_cache_release(pg
);
2898 kfree(cmp
->src_pages
);
2899 kfree(cmp
->dst_pages
);
2902 static int btrfs_cmp_data_prepare(struct inode
*src
, u64 loff
,
2903 struct inode
*dst
, u64 dst_loff
,
2904 u64 len
, struct cmp_pages
*cmp
)
2907 int num_pages
= PAGE_CACHE_ALIGN(len
) >> PAGE_CACHE_SHIFT
;
2908 struct page
**src_pgarr
, **dst_pgarr
;
2911 * We must gather up all the pages before we initiate our
2912 * extent locking. We use an array for the page pointers. Size
2913 * of the array is bounded by len, which is in turn bounded by
2914 * BTRFS_MAX_DEDUPE_LEN.
2916 src_pgarr
= kzalloc(num_pages
* sizeof(struct page
*), GFP_NOFS
);
2917 dst_pgarr
= kzalloc(num_pages
* sizeof(struct page
*), GFP_NOFS
);
2918 if (!src_pgarr
|| !dst_pgarr
) {
2923 cmp
->num_pages
= num_pages
;
2924 cmp
->src_pages
= src_pgarr
;
2925 cmp
->dst_pages
= dst_pgarr
;
2927 ret
= gather_extent_pages(src
, cmp
->src_pages
, cmp
->num_pages
, loff
);
2931 ret
= gather_extent_pages(dst
, cmp
->dst_pages
, cmp
->num_pages
, dst_loff
);
2935 btrfs_cmp_data_free(cmp
);
2939 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
2940 u64 dst_loff
, u64 len
, struct cmp_pages
*cmp
)
2944 struct page
*src_page
, *dst_page
;
2945 unsigned int cmp_len
= PAGE_CACHE_SIZE
;
2946 void *addr
, *dst_addr
;
2950 if (len
< PAGE_CACHE_SIZE
)
2953 BUG_ON(i
>= cmp
->num_pages
);
2955 src_page
= cmp
->src_pages
[i
];
2956 dst_page
= cmp
->dst_pages
[i
];
2958 addr
= kmap_atomic(src_page
);
2959 dst_addr
= kmap_atomic(dst_page
);
2961 flush_dcache_page(src_page
);
2962 flush_dcache_page(dst_page
);
2964 if (memcmp(addr
, dst_addr
, cmp_len
))
2965 ret
= BTRFS_SAME_DATA_DIFFERS
;
2967 kunmap_atomic(addr
);
2968 kunmap_atomic(dst_addr
);
2980 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64
*plen
,
2984 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
2986 if (off
+ olen
> inode
->i_size
|| off
+ olen
< off
)
2989 /* if we extend to eof, continue to block boundary */
2990 if (off
+ len
== inode
->i_size
)
2991 *plen
= len
= ALIGN(inode
->i_size
, bs
) - off
;
2993 /* Check that we are block aligned - btrfs_clone() requires this */
2994 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
3000 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 olen
,
3001 struct inode
*dst
, u64 dst_loff
)
3005 struct cmp_pages cmp
;
3007 u64 same_lock_start
= 0;
3008 u64 same_lock_len
= 0;
3017 mutex_lock(&src
->i_mutex
);
3019 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3024 * Single inode case wants the same checks, except we
3025 * don't want our length pushed out past i_size as
3026 * comparing that data range makes no sense.
3028 * extent_same_check_offsets() will do this for an
3029 * unaligned length at i_size, so catch it here and
3030 * reject the request.
3032 * This effectively means we require aligned extents
3033 * for the single-inode case, whereas the other cases
3034 * allow an unaligned length so long as it ends at
3042 /* Check for overlapping ranges */
3043 if (dst_loff
+ len
> loff
&& dst_loff
< loff
+ len
) {
3048 same_lock_start
= min_t(u64
, loff
, dst_loff
);
3049 same_lock_len
= max_t(u64
, loff
, dst_loff
) + len
- same_lock_start
;
3051 btrfs_double_inode_lock(src
, dst
);
3053 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3057 ret
= extent_same_check_offsets(dst
, dst_loff
, &len
, olen
);
3062 /* don't make the dst file partly checksummed */
3063 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3064 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
3069 ret
= btrfs_cmp_data_prepare(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3074 lock_extent_range(src
, same_lock_start
, same_lock_len
);
3076 btrfs_double_extent_lock(src
, loff
, dst
, dst_loff
, len
);
3078 /* pass original length for comparison so we stay within i_size */
3079 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3081 ret
= btrfs_clone(src
, dst
, loff
, olen
, len
, dst_loff
, 1);
3084 unlock_extent(&BTRFS_I(src
)->io_tree
, same_lock_start
,
3085 same_lock_start
+ same_lock_len
- 1);
3087 btrfs_double_extent_unlock(src
, loff
, dst
, dst_loff
, len
);
3089 btrfs_cmp_data_free(&cmp
);
3092 mutex_unlock(&src
->i_mutex
);
3094 btrfs_double_inode_unlock(src
, dst
);
3099 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
3101 static long btrfs_ioctl_file_extent_same(struct file
*file
,
3102 struct btrfs_ioctl_same_args __user
*argp
)
3104 struct btrfs_ioctl_same_args
*same
= NULL
;
3105 struct btrfs_ioctl_same_extent_info
*info
;
3106 struct inode
*src
= file_inode(file
);
3112 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
3113 bool is_admin
= capable(CAP_SYS_ADMIN
);
3116 if (!(file
->f_mode
& FMODE_READ
))
3119 ret
= mnt_want_write_file(file
);
3123 if (get_user(count
, &argp
->dest_count
)) {
3128 size
= offsetof(struct btrfs_ioctl_same_args __user
, info
[count
]);
3130 same
= memdup_user(argp
, size
);
3133 ret
= PTR_ERR(same
);
3138 off
= same
->logical_offset
;
3142 * Limit the total length we will dedupe for each operation.
3143 * This is intended to bound the total time spent in this
3144 * ioctl to something sane.
3146 if (len
> BTRFS_MAX_DEDUPE_LEN
)
3147 len
= BTRFS_MAX_DEDUPE_LEN
;
3149 if (WARN_ON_ONCE(bs
< PAGE_CACHE_SIZE
)) {
3151 * Btrfs does not support blocksize < page_size. As a
3152 * result, btrfs_cmp_data() won't correctly handle
3153 * this situation without an update.
3160 if (S_ISDIR(src
->i_mode
))
3164 if (!S_ISREG(src
->i_mode
))
3167 /* pre-format output fields to sane values */
3168 for (i
= 0; i
< count
; i
++) {
3169 same
->info
[i
].bytes_deduped
= 0ULL;
3170 same
->info
[i
].status
= 0;
3173 for (i
= 0, info
= same
->info
; i
< count
; i
++, info
++) {
3175 struct fd dst_file
= fdget(info
->fd
);
3176 if (!dst_file
.file
) {
3177 info
->status
= -EBADF
;
3180 dst
= file_inode(dst_file
.file
);
3182 if (!(is_admin
|| (dst_file
.file
->f_mode
& FMODE_WRITE
))) {
3183 info
->status
= -EINVAL
;
3184 } else if (file
->f_path
.mnt
!= dst_file
.file
->f_path
.mnt
) {
3185 info
->status
= -EXDEV
;
3186 } else if (S_ISDIR(dst
->i_mode
)) {
3187 info
->status
= -EISDIR
;
3188 } else if (!S_ISREG(dst
->i_mode
)) {
3189 info
->status
= -EACCES
;
3191 info
->status
= btrfs_extent_same(src
, off
, len
, dst
,
3192 info
->logical_offset
);
3193 if (info
->status
== 0)
3194 info
->bytes_deduped
+= len
;
3199 ret
= copy_to_user(argp
, same
, size
);
3204 mnt_drop_write_file(file
);
3209 /* Helper to check and see if this root currently has a ref on the given disk
3210 * bytenr. If it does then we need to update the quota for this root. This
3211 * doesn't do anything if quotas aren't enabled.
3213 static int check_ref(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
3216 struct seq_list tree_mod_seq_elem
= SEQ_LIST_INIT(tree_mod_seq_elem
);
3217 struct ulist
*roots
;
3218 struct ulist_iterator uiter
;
3219 struct ulist_node
*root_node
= NULL
;
3222 if (!root
->fs_info
->quota_enabled
)
3225 btrfs_get_tree_mod_seq(root
->fs_info
, &tree_mod_seq_elem
);
3226 ret
= btrfs_find_all_roots(trans
, root
->fs_info
, disko
,
3227 tree_mod_seq_elem
.seq
, &roots
);
3231 ULIST_ITER_INIT(&uiter
);
3232 while ((root_node
= ulist_next(roots
, &uiter
))) {
3233 if (root_node
->val
== root
->objectid
) {
3240 btrfs_put_tree_mod_seq(root
->fs_info
, &tree_mod_seq_elem
);
3244 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3245 struct inode
*inode
,
3251 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3254 inode_inc_iversion(inode
);
3255 if (!no_time_update
)
3256 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
3258 * We round up to the block size at eof when determining which
3259 * extents to clone above, but shouldn't round up the file size.
3261 if (endoff
> destoff
+ olen
)
3262 endoff
= destoff
+ olen
;
3263 if (endoff
> inode
->i_size
)
3264 btrfs_i_size_write(inode
, endoff
);
3266 ret
= btrfs_update_inode(trans
, root
, inode
);
3268 btrfs_abort_transaction(trans
, root
, ret
);
3269 btrfs_end_transaction(trans
, root
);
3272 ret
= btrfs_end_transaction(trans
, root
);
3277 static void clone_update_extent_map(struct inode
*inode
,
3278 const struct btrfs_trans_handle
*trans
,
3279 const struct btrfs_path
*path
,
3280 const u64 hole_offset
,
3283 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
3284 struct extent_map
*em
;
3287 em
= alloc_extent_map();
3289 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3290 &BTRFS_I(inode
)->runtime_flags
);
3295 struct btrfs_file_extent_item
*fi
;
3297 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3298 struct btrfs_file_extent_item
);
3299 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3300 em
->generation
= -1;
3301 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3302 BTRFS_FILE_EXTENT_INLINE
)
3303 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3304 &BTRFS_I(inode
)->runtime_flags
);
3306 em
->start
= hole_offset
;
3308 em
->ram_bytes
= em
->len
;
3309 em
->orig_start
= hole_offset
;
3310 em
->block_start
= EXTENT_MAP_HOLE
;
3312 em
->orig_block_len
= 0;
3313 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3314 em
->generation
= trans
->transid
;
3318 write_lock(&em_tree
->lock
);
3319 ret
= add_extent_mapping(em_tree
, em
, 1);
3320 write_unlock(&em_tree
->lock
);
3321 if (ret
!= -EEXIST
) {
3322 free_extent_map(em
);
3325 btrfs_drop_extent_cache(inode
, em
->start
,
3326 em
->start
+ em
->len
- 1, 0);
3330 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3331 &BTRFS_I(inode
)->runtime_flags
);
3335 * Make sure we do not end up inserting an inline extent into a file that has
3336 * already other (non-inline) extents. If a file has an inline extent it can
3337 * not have any other extents and the (single) inline extent must start at the
3338 * file offset 0. Failing to respect these rules will lead to file corruption,
3339 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3341 * We can have extents that have been already written to disk or we can have
3342 * dirty ranges still in delalloc, in which case the extent maps and items are
3343 * created only when we run delalloc, and the delalloc ranges might fall outside
3344 * the range we are currently locking in the inode's io tree. So we check the
3345 * inode's i_size because of that (i_size updates are done while holding the
3346 * i_mutex, which we are holding here).
3347 * We also check to see if the inode has a size not greater than "datal" but has
3348 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3349 * protected against such concurrent fallocate calls by the i_mutex).
3351 * If the file has no extents but a size greater than datal, do not allow the
3352 * copy because we would need turn the inline extent into a non-inline one (even
3353 * with NO_HOLES enabled). If we find our destination inode only has one inline
3354 * extent, just overwrite it with the source inline extent if its size is less
3355 * than the source extent's size, or we could copy the source inline extent's
3356 * data into the destination inode's inline extent if the later is greater then
3359 static int clone_copy_inline_extent(struct inode
*src
,
3361 struct btrfs_trans_handle
*trans
,
3362 struct btrfs_path
*path
,
3363 struct btrfs_key
*new_key
,
3364 const u64 drop_start
,
3370 struct btrfs_root
*root
= BTRFS_I(dst
)->root
;
3371 const u64 aligned_end
= ALIGN(new_key
->offset
+ datal
,
3374 struct btrfs_key key
;
3376 if (new_key
->offset
> 0)
3379 key
.objectid
= btrfs_ino(dst
);
3380 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3382 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
3385 } else if (ret
> 0) {
3386 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3387 ret
= btrfs_next_leaf(root
, path
);
3391 goto copy_inline_extent
;
3393 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
3394 if (key
.objectid
== btrfs_ino(dst
) &&
3395 key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3396 ASSERT(key
.offset
> 0);
3399 } else if (i_size_read(dst
) <= datal
) {
3400 struct btrfs_file_extent_item
*ei
;
3404 * If the file size is <= datal, make sure there are no other
3405 * extents following (can happen do to an fallocate call with
3406 * the flag FALLOC_FL_KEEP_SIZE).
3408 ei
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3409 struct btrfs_file_extent_item
);
3411 * If it's an inline extent, it can not have other extents
3414 if (btrfs_file_extent_type(path
->nodes
[0], ei
) ==
3415 BTRFS_FILE_EXTENT_INLINE
)
3416 goto copy_inline_extent
;
3418 ext_len
= btrfs_file_extent_num_bytes(path
->nodes
[0], ei
);
3419 if (ext_len
> aligned_end
)
3422 ret
= btrfs_next_item(root
, path
);
3425 } else if (ret
== 0) {
3426 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3428 if (key
.objectid
== btrfs_ino(dst
) &&
3429 key
.type
== BTRFS_EXTENT_DATA_KEY
)
3436 * We have no extent items, or we have an extent at offset 0 which may
3437 * or may not be inlined. All these cases are dealt the same way.
3439 if (i_size_read(dst
) > datal
) {
3441 * If the destination inode has an inline extent...
3442 * This would require copying the data from the source inline
3443 * extent into the beginning of the destination's inline extent.
3444 * But this is really complex, both extents can be compressed
3445 * or just one of them, which would require decompressing and
3446 * re-compressing data (which could increase the new compressed
3447 * size, not allowing the compressed data to fit anymore in an
3449 * So just don't support this case for now (it should be rare,
3450 * we are not really saving space when cloning inline extents).
3455 btrfs_release_path(path
);
3456 ret
= btrfs_drop_extents(trans
, root
, dst
, drop_start
, aligned_end
, 1);
3459 ret
= btrfs_insert_empty_item(trans
, root
, path
, new_key
, size
);
3464 const u32 start
= btrfs_file_extent_calc_inline_size(0);
3466 memmove(inline_data
+ start
, inline_data
+ start
+ skip
, datal
);
3469 write_extent_buffer(path
->nodes
[0], inline_data
,
3470 btrfs_item_ptr_offset(path
->nodes
[0],
3473 inode_add_bytes(dst
, datal
);
3479 * btrfs_clone() - clone a range from inode file to another
3481 * @src: Inode to clone from
3482 * @inode: Inode to clone to
3483 * @off: Offset within source to start clone from
3484 * @olen: Original length, passed by user, of range to clone
3485 * @olen_aligned: Block-aligned value of olen
3486 * @destoff: Offset within @inode to start clone
3487 * @no_time_update: Whether to update mtime/ctime on the target inode
3489 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3490 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3491 const u64 destoff
, int no_time_update
)
3493 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3494 struct btrfs_path
*path
= NULL
;
3495 struct extent_buffer
*leaf
;
3496 struct btrfs_trans_handle
*trans
;
3498 struct btrfs_key key
;
3503 const u64 len
= olen_aligned
;
3505 u64 last_dest_end
= destoff
;
3508 buf
= vmalloc(root
->nodesize
);
3512 path
= btrfs_alloc_path();
3520 key
.objectid
= btrfs_ino(src
);
3521 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3525 u64 next_key_min_offset
= key
.offset
+ 1;
3528 * note the key will change type as we walk through the
3531 path
->leave_spinning
= 1;
3532 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3537 * First search, if no extent item that starts at offset off was
3538 * found but the previous item is an extent item, it's possible
3539 * it might overlap our target range, therefore process it.
3541 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3542 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3543 path
->slots
[0] - 1);
3544 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3548 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3551 if (path
->slots
[0] >= nritems
) {
3552 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3557 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3559 leaf
= path
->nodes
[0];
3560 slot
= path
->slots
[0];
3562 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3563 if (key
.type
> BTRFS_EXTENT_DATA_KEY
||
3564 key
.objectid
!= btrfs_ino(src
))
3567 if (key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3568 struct btrfs_file_extent_item
*extent
;
3571 struct btrfs_key new_key
;
3572 u64 disko
= 0, diskl
= 0;
3573 u64 datao
= 0, datal
= 0;
3577 extent
= btrfs_item_ptr(leaf
, slot
,
3578 struct btrfs_file_extent_item
);
3579 comp
= btrfs_file_extent_compression(leaf
, extent
);
3580 type
= btrfs_file_extent_type(leaf
, extent
);
3581 if (type
== BTRFS_FILE_EXTENT_REG
||
3582 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3583 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3585 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3587 datao
= btrfs_file_extent_offset(leaf
, extent
);
3588 datal
= btrfs_file_extent_num_bytes(leaf
,
3590 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3591 /* take upper bound, may be compressed */
3592 datal
= btrfs_file_extent_ram_bytes(leaf
,
3597 * The first search might have left us at an extent
3598 * item that ends before our target range's start, can
3599 * happen if we have holes and NO_HOLES feature enabled.
3601 if (key
.offset
+ datal
<= off
) {
3604 } else if (key
.offset
>= off
+ len
) {
3607 next_key_min_offset
= key
.offset
+ datal
;
3608 size
= btrfs_item_size_nr(leaf
, slot
);
3609 read_extent_buffer(leaf
, buf
,
3610 btrfs_item_ptr_offset(leaf
, slot
),
3613 btrfs_release_path(path
);
3614 path
->leave_spinning
= 0;
3616 memcpy(&new_key
, &key
, sizeof(new_key
));
3617 new_key
.objectid
= btrfs_ino(inode
);
3618 if (off
<= key
.offset
)
3619 new_key
.offset
= key
.offset
+ destoff
- off
;
3621 new_key
.offset
= destoff
;
3624 * Deal with a hole that doesn't have an extent item
3625 * that represents it (NO_HOLES feature enabled).
3626 * This hole is either in the middle of the cloning
3627 * range or at the beginning (fully overlaps it or
3628 * partially overlaps it).
3630 if (new_key
.offset
!= last_dest_end
)
3631 drop_start
= last_dest_end
;
3633 drop_start
= new_key
.offset
;
3636 * 1 - adjusting old extent (we may have to split it)
3637 * 1 - add new extent
3640 trans
= btrfs_start_transaction(root
, 3);
3641 if (IS_ERR(trans
)) {
3642 ret
= PTR_ERR(trans
);
3646 if (type
== BTRFS_FILE_EXTENT_REG
||
3647 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3649 * a | --- range to clone ---| b
3650 * | ------------- extent ------------- |
3653 /* subtract range b */
3654 if (key
.offset
+ datal
> off
+ len
)
3655 datal
= off
+ len
- key
.offset
;
3657 /* subtract range a */
3658 if (off
> key
.offset
) {
3659 datao
+= off
- key
.offset
;
3660 datal
-= off
- key
.offset
;
3663 ret
= btrfs_drop_extents(trans
, root
, inode
,
3665 new_key
.offset
+ datal
,
3668 if (ret
!= -EOPNOTSUPP
)
3669 btrfs_abort_transaction(trans
,
3671 btrfs_end_transaction(trans
, root
);
3675 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3678 btrfs_abort_transaction(trans
, root
,
3680 btrfs_end_transaction(trans
, root
);
3684 leaf
= path
->nodes
[0];
3685 slot
= path
->slots
[0];
3686 write_extent_buffer(leaf
, buf
,
3687 btrfs_item_ptr_offset(leaf
, slot
),
3690 extent
= btrfs_item_ptr(leaf
, slot
,
3691 struct btrfs_file_extent_item
);
3693 /* disko == 0 means it's a hole */
3697 btrfs_set_file_extent_offset(leaf
, extent
,
3699 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3703 * We need to look up the roots that point at
3704 * this bytenr and see if the new root does. If
3705 * it does not we need to make sure we update
3706 * quotas appropriately.
3708 if (disko
&& root
!= BTRFS_I(src
)->root
&&
3709 disko
!= last_disko
) {
3710 no_quota
= check_ref(trans
, root
,
3713 btrfs_abort_transaction(trans
,
3716 btrfs_end_transaction(trans
,
3724 inode_add_bytes(inode
, datal
);
3725 ret
= btrfs_inc_extent_ref(trans
, root
,
3727 root
->root_key
.objectid
,
3729 new_key
.offset
- datao
,
3732 btrfs_abort_transaction(trans
,
3735 btrfs_end_transaction(trans
,
3741 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3745 if (off
> key
.offset
) {
3746 skip
= off
- key
.offset
;
3747 new_key
.offset
+= skip
;
3750 if (key
.offset
+ datal
> off
+ len
)
3751 trim
= key
.offset
+ datal
- (off
+ len
);
3753 if (comp
&& (skip
|| trim
)) {
3755 btrfs_end_transaction(trans
, root
);
3758 size
-= skip
+ trim
;
3759 datal
-= skip
+ trim
;
3761 ret
= clone_copy_inline_extent(src
, inode
,
3768 if (ret
!= -EOPNOTSUPP
)
3769 btrfs_abort_transaction(trans
,
3772 btrfs_end_transaction(trans
, root
);
3775 leaf
= path
->nodes
[0];
3776 slot
= path
->slots
[0];
3779 /* If we have an implicit hole (NO_HOLES feature). */
3780 if (drop_start
< new_key
.offset
)
3781 clone_update_extent_map(inode
, trans
,
3783 new_key
.offset
- drop_start
);
3785 clone_update_extent_map(inode
, trans
, path
, 0, 0);
3787 btrfs_mark_buffer_dirty(leaf
);
3788 btrfs_release_path(path
);
3790 last_dest_end
= ALIGN(new_key
.offset
+ datal
,
3792 ret
= clone_finish_inode_update(trans
, inode
,
3798 if (new_key
.offset
+ datal
>= destoff
+ len
)
3801 btrfs_release_path(path
);
3802 key
.offset
= next_key_min_offset
;
3806 if (last_dest_end
< destoff
+ len
) {
3808 * We have an implicit hole (NO_HOLES feature is enabled) that
3809 * fully or partially overlaps our cloning range at its end.
3811 btrfs_release_path(path
);
3814 * 1 - remove extent(s)
3817 trans
= btrfs_start_transaction(root
, 2);
3818 if (IS_ERR(trans
)) {
3819 ret
= PTR_ERR(trans
);
3822 ret
= btrfs_drop_extents(trans
, root
, inode
,
3823 last_dest_end
, destoff
+ len
, 1);
3825 if (ret
!= -EOPNOTSUPP
)
3826 btrfs_abort_transaction(trans
, root
, ret
);
3827 btrfs_end_transaction(trans
, root
);
3830 clone_update_extent_map(inode
, trans
, NULL
, last_dest_end
,
3831 destoff
+ len
- last_dest_end
);
3832 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3833 destoff
, olen
, no_time_update
);
3837 btrfs_free_path(path
);
3842 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
3843 u64 off
, u64 olen
, u64 destoff
)
3845 struct inode
*inode
= file_inode(file
);
3846 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3851 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
3856 * - split compressed inline extents. annoying: we need to
3857 * decompress into destination's address_space (the file offset
3858 * may change, so source mapping won't do), then recompress (or
3859 * otherwise reinsert) a subrange.
3861 * - split destination inode's inline extents. The inline extents can
3862 * be either compressed or non-compressed.
3865 /* the destination must be opened for writing */
3866 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
3869 if (btrfs_root_readonly(root
))
3872 ret
= mnt_want_write_file(file
);
3876 src_file
= fdget(srcfd
);
3877 if (!src_file
.file
) {
3879 goto out_drop_write
;
3883 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
3886 src
= file_inode(src_file
.file
);
3892 /* the src must be open for reading */
3893 if (!(src_file
.file
->f_mode
& FMODE_READ
))
3896 /* don't make the dst file partly checksummed */
3897 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3898 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3902 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3906 if (src
->i_sb
!= inode
->i_sb
)
3910 btrfs_double_inode_lock(src
, inode
);
3912 mutex_lock(&src
->i_mutex
);
3915 /* determine range to clone */
3917 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3920 olen
= len
= src
->i_size
- off
;
3921 /* if we extend to eof, continue to block boundary */
3922 if (off
+ len
== src
->i_size
)
3923 len
= ALIGN(src
->i_size
, bs
) - off
;
3930 /* verify the end result is block aligned */
3931 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3932 !IS_ALIGNED(destoff
, bs
))
3935 /* verify if ranges are overlapped within the same file */
3937 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3941 if (destoff
> inode
->i_size
) {
3942 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3948 * Lock the target range too. Right after we replace the file extent
3949 * items in the fs tree (which now point to the cloned data), we might
3950 * have a worker replace them with extent items relative to a write
3951 * operation that was issued before this clone operation (i.e. confront
3952 * with inode.c:btrfs_finish_ordered_io).
3955 u64 lock_start
= min_t(u64
, off
, destoff
);
3956 u64 lock_len
= max_t(u64
, off
, destoff
) + len
- lock_start
;
3958 lock_extent_range(src
, lock_start
, lock_len
);
3960 btrfs_double_extent_lock(src
, off
, inode
, destoff
, len
);
3963 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
, 0);
3966 u64 lock_start
= min_t(u64
, off
, destoff
);
3967 u64 lock_end
= max_t(u64
, off
, destoff
) + len
- 1;
3969 unlock_extent(&BTRFS_I(src
)->io_tree
, lock_start
, lock_end
);
3971 btrfs_double_extent_unlock(src
, off
, inode
, destoff
, len
);
3974 * Truncate page cache pages so that future reads will see the cloned
3975 * data immediately and not the previous data.
3977 truncate_inode_pages_range(&inode
->i_data
, destoff
,
3978 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
3981 btrfs_double_inode_unlock(src
, inode
);
3983 mutex_unlock(&src
->i_mutex
);
3987 mnt_drop_write_file(file
);
3991 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
3993 struct btrfs_ioctl_clone_range_args args
;
3995 if (copy_from_user(&args
, argp
, sizeof(args
)))
3997 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
3998 args
.src_length
, args
.dest_offset
);
4002 * there are many ways the trans_start and trans_end ioctls can lead
4003 * to deadlocks. They should only be used by applications that
4004 * basically own the machine, and have a very in depth understanding
4005 * of all the possible deadlocks and enospc problems.
4007 static long btrfs_ioctl_trans_start(struct file
*file
)
4009 struct inode
*inode
= file_inode(file
);
4010 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4011 struct btrfs_trans_handle
*trans
;
4015 if (!capable(CAP_SYS_ADMIN
))
4019 if (file
->private_data
)
4023 if (btrfs_root_readonly(root
))
4026 ret
= mnt_want_write_file(file
);
4030 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
4033 trans
= btrfs_start_ioctl_transaction(root
);
4037 file
->private_data
= trans
;
4041 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4042 mnt_drop_write_file(file
);
4047 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
4049 struct inode
*inode
= file_inode(file
);
4050 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4051 struct btrfs_root
*new_root
;
4052 struct btrfs_dir_item
*di
;
4053 struct btrfs_trans_handle
*trans
;
4054 struct btrfs_path
*path
;
4055 struct btrfs_key location
;
4056 struct btrfs_disk_key disk_key
;
4061 if (!capable(CAP_SYS_ADMIN
))
4064 ret
= mnt_want_write_file(file
);
4068 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
4074 objectid
= BTRFS_FS_TREE_OBJECTID
;
4076 location
.objectid
= objectid
;
4077 location
.type
= BTRFS_ROOT_ITEM_KEY
;
4078 location
.offset
= (u64
)-1;
4080 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
4081 if (IS_ERR(new_root
)) {
4082 ret
= PTR_ERR(new_root
);
4086 path
= btrfs_alloc_path();
4091 path
->leave_spinning
= 1;
4093 trans
= btrfs_start_transaction(root
, 1);
4094 if (IS_ERR(trans
)) {
4095 btrfs_free_path(path
);
4096 ret
= PTR_ERR(trans
);
4100 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
4101 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
4102 dir_id
, "default", 7, 1);
4103 if (IS_ERR_OR_NULL(di
)) {
4104 btrfs_free_path(path
);
4105 btrfs_end_transaction(trans
, root
);
4106 btrfs_err(new_root
->fs_info
, "Umm, you don't have the default dir"
4107 "item, this isn't going to work");
4112 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
4113 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
4114 btrfs_mark_buffer_dirty(path
->nodes
[0]);
4115 btrfs_free_path(path
);
4117 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
4118 btrfs_end_transaction(trans
, root
);
4120 mnt_drop_write_file(file
);
4124 void btrfs_get_block_group_info(struct list_head
*groups_list
,
4125 struct btrfs_ioctl_space_info
*space
)
4127 struct btrfs_block_group_cache
*block_group
;
4129 space
->total_bytes
= 0;
4130 space
->used_bytes
= 0;
4132 list_for_each_entry(block_group
, groups_list
, list
) {
4133 space
->flags
= block_group
->flags
;
4134 space
->total_bytes
+= block_group
->key
.offset
;
4135 space
->used_bytes
+=
4136 btrfs_block_group_used(&block_group
->item
);
4140 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
4142 struct btrfs_ioctl_space_args space_args
;
4143 struct btrfs_ioctl_space_info space
;
4144 struct btrfs_ioctl_space_info
*dest
;
4145 struct btrfs_ioctl_space_info
*dest_orig
;
4146 struct btrfs_ioctl_space_info __user
*user_dest
;
4147 struct btrfs_space_info
*info
;
4148 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
4149 BTRFS_BLOCK_GROUP_SYSTEM
,
4150 BTRFS_BLOCK_GROUP_METADATA
,
4151 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
4158 if (copy_from_user(&space_args
,
4159 (struct btrfs_ioctl_space_args __user
*)arg
,
4160 sizeof(space_args
)))
4163 for (i
= 0; i
< num_types
; i
++) {
4164 struct btrfs_space_info
*tmp
;
4168 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4170 if (tmp
->flags
== types
[i
]) {
4180 down_read(&info
->groups_sem
);
4181 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4182 if (!list_empty(&info
->block_groups
[c
]))
4185 up_read(&info
->groups_sem
);
4189 * Global block reserve, exported as a space_info
4193 /* space_slots == 0 means they are asking for a count */
4194 if (space_args
.space_slots
== 0) {
4195 space_args
.total_spaces
= slot_count
;
4199 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
4201 alloc_size
= sizeof(*dest
) * slot_count
;
4203 /* we generally have at most 6 or so space infos, one for each raid
4204 * level. So, a whole page should be more than enough for everyone
4206 if (alloc_size
> PAGE_CACHE_SIZE
)
4209 space_args
.total_spaces
= 0;
4210 dest
= kmalloc(alloc_size
, GFP_NOFS
);
4215 /* now we have a buffer to copy into */
4216 for (i
= 0; i
< num_types
; i
++) {
4217 struct btrfs_space_info
*tmp
;
4224 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
4226 if (tmp
->flags
== types
[i
]) {
4235 down_read(&info
->groups_sem
);
4236 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4237 if (!list_empty(&info
->block_groups
[c
])) {
4238 btrfs_get_block_group_info(
4239 &info
->block_groups
[c
], &space
);
4240 memcpy(dest
, &space
, sizeof(space
));
4242 space_args
.total_spaces
++;
4248 up_read(&info
->groups_sem
);
4252 * Add global block reserve
4255 struct btrfs_block_rsv
*block_rsv
= &root
->fs_info
->global_block_rsv
;
4257 spin_lock(&block_rsv
->lock
);
4258 space
.total_bytes
= block_rsv
->size
;
4259 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4260 spin_unlock(&block_rsv
->lock
);
4261 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4262 memcpy(dest
, &space
, sizeof(space
));
4263 space_args
.total_spaces
++;
4266 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4267 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4269 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4274 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4281 * there are many ways the trans_start and trans_end ioctls can lead
4282 * to deadlocks. They should only be used by applications that
4283 * basically own the machine, and have a very in depth understanding
4284 * of all the possible deadlocks and enospc problems.
4286 long btrfs_ioctl_trans_end(struct file
*file
)
4288 struct inode
*inode
= file_inode(file
);
4289 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4290 struct btrfs_trans_handle
*trans
;
4292 trans
= file
->private_data
;
4295 file
->private_data
= NULL
;
4297 btrfs_end_transaction(trans
, root
);
4299 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4301 mnt_drop_write_file(file
);
4305 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4308 struct btrfs_trans_handle
*trans
;
4312 trans
= btrfs_attach_transaction_barrier(root
);
4313 if (IS_ERR(trans
)) {
4314 if (PTR_ERR(trans
) != -ENOENT
)
4315 return PTR_ERR(trans
);
4317 /* No running transaction, don't bother */
4318 transid
= root
->fs_info
->last_trans_committed
;
4321 transid
= trans
->transid
;
4322 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
4324 btrfs_end_transaction(trans
, root
);
4329 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4334 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
4340 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4343 transid
= 0; /* current trans */
4345 return btrfs_wait_for_commit(root
, transid
);
4348 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4350 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4351 struct btrfs_ioctl_scrub_args
*sa
;
4354 if (!capable(CAP_SYS_ADMIN
))
4357 sa
= memdup_user(arg
, sizeof(*sa
));
4361 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4362 ret
= mnt_want_write_file(file
);
4367 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4368 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4371 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4374 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4375 mnt_drop_write_file(file
);
4381 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
4383 if (!capable(CAP_SYS_ADMIN
))
4386 return btrfs_scrub_cancel(root
->fs_info
);
4389 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
4392 struct btrfs_ioctl_scrub_args
*sa
;
4395 if (!capable(CAP_SYS_ADMIN
))
4398 sa
= memdup_user(arg
, sizeof(*sa
));
4402 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
4404 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4411 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
4414 struct btrfs_ioctl_get_dev_stats
*sa
;
4417 sa
= memdup_user(arg
, sizeof(*sa
));
4421 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4426 ret
= btrfs_get_dev_stats(root
, sa
);
4428 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4435 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
4437 struct btrfs_ioctl_dev_replace_args
*p
;
4440 if (!capable(CAP_SYS_ADMIN
))
4443 p
= memdup_user(arg
, sizeof(*p
));
4448 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4449 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
) {
4454 &root
->fs_info
->mutually_exclusive_operation_running
,
4456 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4458 ret
= btrfs_dev_replace_start(root
, p
);
4460 &root
->fs_info
->mutually_exclusive_operation_running
,
4464 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4465 btrfs_dev_replace_status(root
->fs_info
, p
);
4468 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4469 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
4476 if (copy_to_user(arg
, p
, sizeof(*p
)))
4483 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4489 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4490 struct inode_fs_paths
*ipath
= NULL
;
4491 struct btrfs_path
*path
;
4493 if (!capable(CAP_DAC_READ_SEARCH
))
4496 path
= btrfs_alloc_path();
4502 ipa
= memdup_user(arg
, sizeof(*ipa
));
4509 size
= min_t(u32
, ipa
->size
, 4096);
4510 ipath
= init_ipath(size
, root
, path
);
4511 if (IS_ERR(ipath
)) {
4512 ret
= PTR_ERR(ipath
);
4517 ret
= paths_from_inode(ipa
->inum
, ipath
);
4521 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4522 rel_ptr
= ipath
->fspath
->val
[i
] -
4523 (u64
)(unsigned long)ipath
->fspath
->val
;
4524 ipath
->fspath
->val
[i
] = rel_ptr
;
4527 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
4528 (void *)(unsigned long)ipath
->fspath
, size
);
4535 btrfs_free_path(path
);
4542 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4544 struct btrfs_data_container
*inodes
= ctx
;
4545 const size_t c
= 3 * sizeof(u64
);
4547 if (inodes
->bytes_left
>= c
) {
4548 inodes
->bytes_left
-= c
;
4549 inodes
->val
[inodes
->elem_cnt
] = inum
;
4550 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4551 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4552 inodes
->elem_cnt
+= 3;
4554 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4555 inodes
->bytes_left
= 0;
4556 inodes
->elem_missed
+= 3;
4562 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
4567 struct btrfs_ioctl_logical_ino_args
*loi
;
4568 struct btrfs_data_container
*inodes
= NULL
;
4569 struct btrfs_path
*path
= NULL
;
4571 if (!capable(CAP_SYS_ADMIN
))
4574 loi
= memdup_user(arg
, sizeof(*loi
));
4581 path
= btrfs_alloc_path();
4587 size
= min_t(u32
, loi
->size
, 64 * 1024);
4588 inodes
= init_data_container(size
);
4589 if (IS_ERR(inodes
)) {
4590 ret
= PTR_ERR(inodes
);
4595 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
4596 build_ino_list
, inodes
);
4602 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
4603 (void *)(unsigned long)inodes
, size
);
4608 btrfs_free_path(path
);
4615 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
4616 struct btrfs_ioctl_balance_args
*bargs
)
4618 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4620 bargs
->flags
= bctl
->flags
;
4622 if (atomic_read(&fs_info
->balance_running
))
4623 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4624 if (atomic_read(&fs_info
->balance_pause_req
))
4625 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4626 if (atomic_read(&fs_info
->balance_cancel_req
))
4627 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4629 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4630 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4631 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4634 spin_lock(&fs_info
->balance_lock
);
4635 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4636 spin_unlock(&fs_info
->balance_lock
);
4638 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4642 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4644 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4645 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4646 struct btrfs_ioctl_balance_args
*bargs
;
4647 struct btrfs_balance_control
*bctl
;
4648 bool need_unlock
; /* for mut. excl. ops lock */
4651 if (!capable(CAP_SYS_ADMIN
))
4654 ret
= mnt_want_write_file(file
);
4659 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4660 mutex_lock(&fs_info
->volume_mutex
);
4661 mutex_lock(&fs_info
->balance_mutex
);
4667 * mut. excl. ops lock is locked. Three possibilites:
4668 * (1) some other op is running
4669 * (2) balance is running
4670 * (3) balance is paused -- special case (think resume)
4672 mutex_lock(&fs_info
->balance_mutex
);
4673 if (fs_info
->balance_ctl
) {
4674 /* this is either (2) or (3) */
4675 if (!atomic_read(&fs_info
->balance_running
)) {
4676 mutex_unlock(&fs_info
->balance_mutex
);
4677 if (!mutex_trylock(&fs_info
->volume_mutex
))
4679 mutex_lock(&fs_info
->balance_mutex
);
4681 if (fs_info
->balance_ctl
&&
4682 !atomic_read(&fs_info
->balance_running
)) {
4684 need_unlock
= false;
4688 mutex_unlock(&fs_info
->balance_mutex
);
4689 mutex_unlock(&fs_info
->volume_mutex
);
4693 mutex_unlock(&fs_info
->balance_mutex
);
4699 mutex_unlock(&fs_info
->balance_mutex
);
4700 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4705 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4708 bargs
= memdup_user(arg
, sizeof(*bargs
));
4709 if (IS_ERR(bargs
)) {
4710 ret
= PTR_ERR(bargs
);
4714 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4715 if (!fs_info
->balance_ctl
) {
4720 bctl
= fs_info
->balance_ctl
;
4721 spin_lock(&fs_info
->balance_lock
);
4722 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4723 spin_unlock(&fs_info
->balance_lock
);
4731 if (fs_info
->balance_ctl
) {
4736 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
4742 bctl
->fs_info
= fs_info
;
4744 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4745 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4746 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4748 bctl
->flags
= bargs
->flags
;
4750 /* balance everything - no filters */
4751 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4756 * Ownership of bctl and mutually_exclusive_operation_running
4757 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4758 * or, if restriper was paused all the way until unmount, in
4759 * free_fs_info. mutually_exclusive_operation_running is
4760 * cleared in __cancel_balance.
4762 need_unlock
= false;
4764 ret
= btrfs_balance(bctl
, bargs
);
4767 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4774 mutex_unlock(&fs_info
->balance_mutex
);
4775 mutex_unlock(&fs_info
->volume_mutex
);
4777 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4779 mnt_drop_write_file(file
);
4783 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4785 if (!capable(CAP_SYS_ADMIN
))
4789 case BTRFS_BALANCE_CTL_PAUSE
:
4790 return btrfs_pause_balance(root
->fs_info
);
4791 case BTRFS_BALANCE_CTL_CANCEL
:
4792 return btrfs_cancel_balance(root
->fs_info
);
4798 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4801 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4802 struct btrfs_ioctl_balance_args
*bargs
;
4805 if (!capable(CAP_SYS_ADMIN
))
4808 mutex_lock(&fs_info
->balance_mutex
);
4809 if (!fs_info
->balance_ctl
) {
4814 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
4820 update_ioctl_balance_args(fs_info
, 1, bargs
);
4822 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4827 mutex_unlock(&fs_info
->balance_mutex
);
4831 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4833 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4834 struct btrfs_ioctl_quota_ctl_args
*sa
;
4835 struct btrfs_trans_handle
*trans
= NULL
;
4839 if (!capable(CAP_SYS_ADMIN
))
4842 ret
= mnt_want_write_file(file
);
4846 sa
= memdup_user(arg
, sizeof(*sa
));
4852 down_write(&root
->fs_info
->subvol_sem
);
4853 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
4854 if (IS_ERR(trans
)) {
4855 ret
= PTR_ERR(trans
);
4860 case BTRFS_QUOTA_CTL_ENABLE
:
4861 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
4863 case BTRFS_QUOTA_CTL_DISABLE
:
4864 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
4871 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
4876 up_write(&root
->fs_info
->subvol_sem
);
4878 mnt_drop_write_file(file
);
4882 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4884 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4885 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4886 struct btrfs_trans_handle
*trans
;
4890 if (!capable(CAP_SYS_ADMIN
))
4893 ret
= mnt_want_write_file(file
);
4897 sa
= memdup_user(arg
, sizeof(*sa
));
4903 trans
= btrfs_join_transaction(root
);
4904 if (IS_ERR(trans
)) {
4905 ret
= PTR_ERR(trans
);
4909 /* FIXME: check if the IDs really exist */
4911 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
4914 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
4918 /* update qgroup status and info */
4919 err
= btrfs_run_qgroups(trans
, root
->fs_info
);
4921 btrfs_std_error(root
->fs_info
, ret
,
4922 "failed to update qgroup status and info\n");
4923 err
= btrfs_end_transaction(trans
, root
);
4930 mnt_drop_write_file(file
);
4934 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4936 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4937 struct btrfs_ioctl_qgroup_create_args
*sa
;
4938 struct btrfs_trans_handle
*trans
;
4942 if (!capable(CAP_SYS_ADMIN
))
4945 ret
= mnt_want_write_file(file
);
4949 sa
= memdup_user(arg
, sizeof(*sa
));
4955 if (!sa
->qgroupid
) {
4960 trans
= btrfs_join_transaction(root
);
4961 if (IS_ERR(trans
)) {
4962 ret
= PTR_ERR(trans
);
4966 /* FIXME: check if the IDs really exist */
4968 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4970 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
4973 err
= btrfs_end_transaction(trans
, root
);
4980 mnt_drop_write_file(file
);
4984 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4986 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4987 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4988 struct btrfs_trans_handle
*trans
;
4993 if (!capable(CAP_SYS_ADMIN
))
4996 ret
= mnt_want_write_file(file
);
5000 sa
= memdup_user(arg
, sizeof(*sa
));
5006 trans
= btrfs_join_transaction(root
);
5007 if (IS_ERR(trans
)) {
5008 ret
= PTR_ERR(trans
);
5012 qgroupid
= sa
->qgroupid
;
5014 /* take the current subvol as qgroup */
5015 qgroupid
= root
->root_key
.objectid
;
5018 /* FIXME: check if the IDs really exist */
5019 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
5021 err
= btrfs_end_transaction(trans
, root
);
5028 mnt_drop_write_file(file
);
5032 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
5034 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5035 struct btrfs_ioctl_quota_rescan_args
*qsa
;
5038 if (!capable(CAP_SYS_ADMIN
))
5041 ret
= mnt_want_write_file(file
);
5045 qsa
= memdup_user(arg
, sizeof(*qsa
));
5056 ret
= btrfs_qgroup_rescan(root
->fs_info
);
5061 mnt_drop_write_file(file
);
5065 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
5067 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5068 struct btrfs_ioctl_quota_rescan_args
*qsa
;
5071 if (!capable(CAP_SYS_ADMIN
))
5074 qsa
= kzalloc(sizeof(*qsa
), GFP_NOFS
);
5078 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
5080 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
5083 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
5090 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
5092 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5094 if (!capable(CAP_SYS_ADMIN
))
5097 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
5100 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
5101 struct btrfs_ioctl_received_subvol_args
*sa
)
5103 struct inode
*inode
= file_inode(file
);
5104 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5105 struct btrfs_root_item
*root_item
= &root
->root_item
;
5106 struct btrfs_trans_handle
*trans
;
5107 struct timespec ct
= CURRENT_TIME
;
5109 int received_uuid_changed
;
5111 if (!inode_owner_or_capable(inode
))
5114 ret
= mnt_want_write_file(file
);
5118 down_write(&root
->fs_info
->subvol_sem
);
5120 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
5125 if (btrfs_root_readonly(root
)) {
5132 * 2 - uuid items (received uuid + subvol uuid)
5134 trans
= btrfs_start_transaction(root
, 3);
5135 if (IS_ERR(trans
)) {
5136 ret
= PTR_ERR(trans
);
5141 sa
->rtransid
= trans
->transid
;
5142 sa
->rtime
.sec
= ct
.tv_sec
;
5143 sa
->rtime
.nsec
= ct
.tv_nsec
;
5145 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
5147 if (received_uuid_changed
&&
5148 !btrfs_is_empty_uuid(root_item
->received_uuid
))
5149 btrfs_uuid_tree_rem(trans
, root
->fs_info
->uuid_root
,
5150 root_item
->received_uuid
,
5151 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5152 root
->root_key
.objectid
);
5153 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
5154 btrfs_set_root_stransid(root_item
, sa
->stransid
);
5155 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
5156 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
5157 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
5158 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
5159 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
5161 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
5162 &root
->root_key
, &root
->root_item
);
5164 btrfs_end_transaction(trans
, root
);
5167 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
5168 ret
= btrfs_uuid_tree_add(trans
, root
->fs_info
->uuid_root
,
5170 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5171 root
->root_key
.objectid
);
5172 if (ret
< 0 && ret
!= -EEXIST
) {
5173 btrfs_abort_transaction(trans
, root
, ret
);
5177 ret
= btrfs_commit_transaction(trans
, root
);
5179 btrfs_abort_transaction(trans
, root
, ret
);
5184 up_write(&root
->fs_info
->subvol_sem
);
5185 mnt_drop_write_file(file
);
5190 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
5193 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
5194 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
5197 args32
= memdup_user(arg
, sizeof(*args32
));
5198 if (IS_ERR(args32
)) {
5199 ret
= PTR_ERR(args32
);
5204 args64
= kmalloc(sizeof(*args64
), GFP_NOFS
);
5210 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
5211 args64
->stransid
= args32
->stransid
;
5212 args64
->rtransid
= args32
->rtransid
;
5213 args64
->stime
.sec
= args32
->stime
.sec
;
5214 args64
->stime
.nsec
= args32
->stime
.nsec
;
5215 args64
->rtime
.sec
= args32
->rtime
.sec
;
5216 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
5217 args64
->flags
= args32
->flags
;
5219 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
5223 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
5224 args32
->stransid
= args64
->stransid
;
5225 args32
->rtransid
= args64
->rtransid
;
5226 args32
->stime
.sec
= args64
->stime
.sec
;
5227 args32
->stime
.nsec
= args64
->stime
.nsec
;
5228 args32
->rtime
.sec
= args64
->rtime
.sec
;
5229 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
5230 args32
->flags
= args64
->flags
;
5232 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
5243 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
5246 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
5249 sa
= memdup_user(arg
, sizeof(*sa
));
5256 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5261 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5270 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5272 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5275 char label
[BTRFS_LABEL_SIZE
];
5277 spin_lock(&root
->fs_info
->super_lock
);
5278 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5279 spin_unlock(&root
->fs_info
->super_lock
);
5281 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5283 if (len
== BTRFS_LABEL_SIZE
) {
5284 btrfs_warn(root
->fs_info
,
5285 "label is too long, return the first %zu bytes", --len
);
5288 ret
= copy_to_user(arg
, label
, len
);
5290 return ret
? -EFAULT
: 0;
5293 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5295 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5296 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5297 struct btrfs_trans_handle
*trans
;
5298 char label
[BTRFS_LABEL_SIZE
];
5301 if (!capable(CAP_SYS_ADMIN
))
5304 if (copy_from_user(label
, arg
, sizeof(label
)))
5307 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5308 btrfs_err(root
->fs_info
, "unable to set label with more than %d bytes",
5309 BTRFS_LABEL_SIZE
- 1);
5313 ret
= mnt_want_write_file(file
);
5317 trans
= btrfs_start_transaction(root
, 0);
5318 if (IS_ERR(trans
)) {
5319 ret
= PTR_ERR(trans
);
5323 spin_lock(&root
->fs_info
->super_lock
);
5324 strcpy(super_block
->label
, label
);
5325 spin_unlock(&root
->fs_info
->super_lock
);
5326 ret
= btrfs_commit_transaction(trans
, root
);
5329 mnt_drop_write_file(file
);
5333 #define INIT_FEATURE_FLAGS(suffix) \
5334 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5335 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5336 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5338 static int btrfs_ioctl_get_supported_features(struct file
*file
,
5341 static struct btrfs_ioctl_feature_flags features
[3] = {
5342 INIT_FEATURE_FLAGS(SUPP
),
5343 INIT_FEATURE_FLAGS(SAFE_SET
),
5344 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5347 if (copy_to_user(arg
, &features
, sizeof(features
)))
5353 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5355 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5356 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5357 struct btrfs_ioctl_feature_flags features
;
5359 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5360 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5361 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5363 if (copy_to_user(arg
, &features
, sizeof(features
)))
5369 static int check_feature_bits(struct btrfs_root
*root
,
5370 enum btrfs_feature_set set
,
5371 u64 change_mask
, u64 flags
, u64 supported_flags
,
5372 u64 safe_set
, u64 safe_clear
)
5374 const char *type
= btrfs_feature_set_names
[set
];
5376 u64 disallowed
, unsupported
;
5377 u64 set_mask
= flags
& change_mask
;
5378 u64 clear_mask
= ~flags
& change_mask
;
5380 unsupported
= set_mask
& ~supported_flags
;
5382 names
= btrfs_printable_features(set
, unsupported
);
5384 btrfs_warn(root
->fs_info
,
5385 "this kernel does not support the %s feature bit%s",
5386 names
, strchr(names
, ',') ? "s" : "");
5389 btrfs_warn(root
->fs_info
,
5390 "this kernel does not support %s bits 0x%llx",
5395 disallowed
= set_mask
& ~safe_set
;
5397 names
= btrfs_printable_features(set
, disallowed
);
5399 btrfs_warn(root
->fs_info
,
5400 "can't set the %s feature bit%s while mounted",
5401 names
, strchr(names
, ',') ? "s" : "");
5404 btrfs_warn(root
->fs_info
,
5405 "can't set %s bits 0x%llx while mounted",
5410 disallowed
= clear_mask
& ~safe_clear
;
5412 names
= btrfs_printable_features(set
, disallowed
);
5414 btrfs_warn(root
->fs_info
,
5415 "can't clear the %s feature bit%s while mounted",
5416 names
, strchr(names
, ',') ? "s" : "");
5419 btrfs_warn(root
->fs_info
,
5420 "can't clear %s bits 0x%llx while mounted",
5428 #define check_feature(root, change_mask, flags, mask_base) \
5429 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5430 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5431 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5432 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5434 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5436 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5437 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
5438 struct btrfs_ioctl_feature_flags flags
[2];
5439 struct btrfs_trans_handle
*trans
;
5443 if (!capable(CAP_SYS_ADMIN
))
5446 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5450 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5451 !flags
[0].incompat_flags
)
5454 ret
= check_feature(root
, flags
[0].compat_flags
,
5455 flags
[1].compat_flags
, COMPAT
);
5459 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
5460 flags
[1].compat_ro_flags
, COMPAT_RO
);
5464 ret
= check_feature(root
, flags
[0].incompat_flags
,
5465 flags
[1].incompat_flags
, INCOMPAT
);
5469 trans
= btrfs_start_transaction(root
, 0);
5471 return PTR_ERR(trans
);
5473 spin_lock(&root
->fs_info
->super_lock
);
5474 newflags
= btrfs_super_compat_flags(super_block
);
5475 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5476 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5477 btrfs_set_super_compat_flags(super_block
, newflags
);
5479 newflags
= btrfs_super_compat_ro_flags(super_block
);
5480 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5481 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5482 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5484 newflags
= btrfs_super_incompat_flags(super_block
);
5485 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5486 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5487 btrfs_set_super_incompat_flags(super_block
, newflags
);
5488 spin_unlock(&root
->fs_info
->super_lock
);
5490 return btrfs_commit_transaction(trans
, root
);
5493 long btrfs_ioctl(struct file
*file
, unsigned int
5494 cmd
, unsigned long arg
)
5496 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
5497 void __user
*argp
= (void __user
*)arg
;
5500 case FS_IOC_GETFLAGS
:
5501 return btrfs_ioctl_getflags(file
, argp
);
5502 case FS_IOC_SETFLAGS
:
5503 return btrfs_ioctl_setflags(file
, argp
);
5504 case FS_IOC_GETVERSION
:
5505 return btrfs_ioctl_getversion(file
, argp
);
5507 return btrfs_ioctl_fitrim(file
, argp
);
5508 case BTRFS_IOC_SNAP_CREATE
:
5509 return btrfs_ioctl_snap_create(file
, argp
, 0);
5510 case BTRFS_IOC_SNAP_CREATE_V2
:
5511 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5512 case BTRFS_IOC_SUBVOL_CREATE
:
5513 return btrfs_ioctl_snap_create(file
, argp
, 1);
5514 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5515 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5516 case BTRFS_IOC_SNAP_DESTROY
:
5517 return btrfs_ioctl_snap_destroy(file
, argp
);
5518 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5519 return btrfs_ioctl_subvol_getflags(file
, argp
);
5520 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5521 return btrfs_ioctl_subvol_setflags(file
, argp
);
5522 case BTRFS_IOC_DEFAULT_SUBVOL
:
5523 return btrfs_ioctl_default_subvol(file
, argp
);
5524 case BTRFS_IOC_DEFRAG
:
5525 return btrfs_ioctl_defrag(file
, NULL
);
5526 case BTRFS_IOC_DEFRAG_RANGE
:
5527 return btrfs_ioctl_defrag(file
, argp
);
5528 case BTRFS_IOC_RESIZE
:
5529 return btrfs_ioctl_resize(file
, argp
);
5530 case BTRFS_IOC_ADD_DEV
:
5531 return btrfs_ioctl_add_dev(root
, argp
);
5532 case BTRFS_IOC_RM_DEV
:
5533 return btrfs_ioctl_rm_dev(file
, argp
);
5534 case BTRFS_IOC_FS_INFO
:
5535 return btrfs_ioctl_fs_info(root
, argp
);
5536 case BTRFS_IOC_DEV_INFO
:
5537 return btrfs_ioctl_dev_info(root
, argp
);
5538 case BTRFS_IOC_BALANCE
:
5539 return btrfs_ioctl_balance(file
, NULL
);
5540 case BTRFS_IOC_CLONE
:
5541 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
5542 case BTRFS_IOC_CLONE_RANGE
:
5543 return btrfs_ioctl_clone_range(file
, argp
);
5544 case BTRFS_IOC_TRANS_START
:
5545 return btrfs_ioctl_trans_start(file
);
5546 case BTRFS_IOC_TRANS_END
:
5547 return btrfs_ioctl_trans_end(file
);
5548 case BTRFS_IOC_TREE_SEARCH
:
5549 return btrfs_ioctl_tree_search(file
, argp
);
5550 case BTRFS_IOC_TREE_SEARCH_V2
:
5551 return btrfs_ioctl_tree_search_v2(file
, argp
);
5552 case BTRFS_IOC_INO_LOOKUP
:
5553 return btrfs_ioctl_ino_lookup(file
, argp
);
5554 case BTRFS_IOC_INO_PATHS
:
5555 return btrfs_ioctl_ino_to_path(root
, argp
);
5556 case BTRFS_IOC_LOGICAL_INO
:
5557 return btrfs_ioctl_logical_to_ino(root
, argp
);
5558 case BTRFS_IOC_SPACE_INFO
:
5559 return btrfs_ioctl_space_info(root
, argp
);
5560 case BTRFS_IOC_SYNC
: {
5563 ret
= btrfs_start_delalloc_roots(root
->fs_info
, 0, -1);
5566 ret
= btrfs_sync_fs(file_inode(file
)->i_sb
, 1);
5568 * The transaction thread may want to do more work,
5569 * namely it pokes the cleaner ktread that will start
5570 * processing uncleaned subvols.
5572 wake_up_process(root
->fs_info
->transaction_kthread
);
5575 case BTRFS_IOC_START_SYNC
:
5576 return btrfs_ioctl_start_sync(root
, argp
);
5577 case BTRFS_IOC_WAIT_SYNC
:
5578 return btrfs_ioctl_wait_sync(root
, argp
);
5579 case BTRFS_IOC_SCRUB
:
5580 return btrfs_ioctl_scrub(file
, argp
);
5581 case BTRFS_IOC_SCRUB_CANCEL
:
5582 return btrfs_ioctl_scrub_cancel(root
, argp
);
5583 case BTRFS_IOC_SCRUB_PROGRESS
:
5584 return btrfs_ioctl_scrub_progress(root
, argp
);
5585 case BTRFS_IOC_BALANCE_V2
:
5586 return btrfs_ioctl_balance(file
, argp
);
5587 case BTRFS_IOC_BALANCE_CTL
:
5588 return btrfs_ioctl_balance_ctl(root
, arg
);
5589 case BTRFS_IOC_BALANCE_PROGRESS
:
5590 return btrfs_ioctl_balance_progress(root
, argp
);
5591 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5592 return btrfs_ioctl_set_received_subvol(file
, argp
);
5594 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5595 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5597 case BTRFS_IOC_SEND
:
5598 return btrfs_ioctl_send(file
, argp
);
5599 case BTRFS_IOC_GET_DEV_STATS
:
5600 return btrfs_ioctl_get_dev_stats(root
, argp
);
5601 case BTRFS_IOC_QUOTA_CTL
:
5602 return btrfs_ioctl_quota_ctl(file
, argp
);
5603 case BTRFS_IOC_QGROUP_ASSIGN
:
5604 return btrfs_ioctl_qgroup_assign(file
, argp
);
5605 case BTRFS_IOC_QGROUP_CREATE
:
5606 return btrfs_ioctl_qgroup_create(file
, argp
);
5607 case BTRFS_IOC_QGROUP_LIMIT
:
5608 return btrfs_ioctl_qgroup_limit(file
, argp
);
5609 case BTRFS_IOC_QUOTA_RESCAN
:
5610 return btrfs_ioctl_quota_rescan(file
, argp
);
5611 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5612 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5613 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5614 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5615 case BTRFS_IOC_DEV_REPLACE
:
5616 return btrfs_ioctl_dev_replace(root
, argp
);
5617 case BTRFS_IOC_GET_FSLABEL
:
5618 return btrfs_ioctl_get_fslabel(file
, argp
);
5619 case BTRFS_IOC_SET_FSLABEL
:
5620 return btrfs_ioctl_set_fslabel(file
, argp
);
5621 case BTRFS_IOC_FILE_EXTENT_SAME
:
5622 return btrfs_ioctl_file_extent_same(file
, argp
);
5623 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5624 return btrfs_ioctl_get_supported_features(file
, argp
);
5625 case BTRFS_IOC_GET_FEATURES
:
5626 return btrfs_ioctl_get_features(file
, argp
);
5627 case BTRFS_IOC_SET_FEATURES
:
5628 return btrfs_ioctl_set_features(file
, argp
);