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/compat.h>
37 #include <linux/bit_spinlock.h>
38 #include <linux/security.h>
39 #include <linux/xattr.h>
40 #include <linux/vmalloc.h>
41 #include <linux/slab.h>
42 #include <linux/blkdev.h>
43 #include <linux/uuid.h>
44 #include <linux/btrfs.h>
45 #include <linux/uaccess.h>
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
53 #include "inode-map.h"
55 #include "rcu-string.h"
57 #include "dev-replace.h"
62 #include "compression.h"
65 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
66 * structures are incorrect, as the timespec structure from userspace
67 * is 4 bytes too small. We define these alternatives here to teach
68 * the kernel about the 32-bit struct packing.
70 struct btrfs_ioctl_timespec_32
{
73 } __attribute__ ((__packed__
));
75 struct btrfs_ioctl_received_subvol_args_32
{
76 char uuid
[BTRFS_UUID_SIZE
]; /* in */
77 __u64 stransid
; /* in */
78 __u64 rtransid
; /* out */
79 struct btrfs_ioctl_timespec_32 stime
; /* in */
80 struct btrfs_ioctl_timespec_32 rtime
; /* out */
82 __u64 reserved
[16]; /* in */
83 } __attribute__ ((__packed__
));
85 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
86 struct btrfs_ioctl_received_subvol_args_32)
90 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
91 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
,
94 /* Mask out flags that are inappropriate for the given type of inode. */
95 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
99 else if (S_ISREG(mode
))
100 return flags
& ~FS_DIRSYNC_FL
;
102 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
106 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
110 unsigned int iflags
= 0;
112 if (flags
& BTRFS_INODE_SYNC
)
113 iflags
|= FS_SYNC_FL
;
114 if (flags
& BTRFS_INODE_IMMUTABLE
)
115 iflags
|= FS_IMMUTABLE_FL
;
116 if (flags
& BTRFS_INODE_APPEND
)
117 iflags
|= FS_APPEND_FL
;
118 if (flags
& BTRFS_INODE_NODUMP
)
119 iflags
|= FS_NODUMP_FL
;
120 if (flags
& BTRFS_INODE_NOATIME
)
121 iflags
|= FS_NOATIME_FL
;
122 if (flags
& BTRFS_INODE_DIRSYNC
)
123 iflags
|= FS_DIRSYNC_FL
;
124 if (flags
& BTRFS_INODE_NODATACOW
)
125 iflags
|= FS_NOCOW_FL
;
127 if (flags
& BTRFS_INODE_NOCOMPRESS
)
128 iflags
|= FS_NOCOMP_FL
;
129 else if (flags
& BTRFS_INODE_COMPRESS
)
130 iflags
|= FS_COMPR_FL
;
136 * Update inode->i_flags based on the btrfs internal flags.
138 void btrfs_update_iflags(struct inode
*inode
)
140 struct btrfs_inode
*ip
= BTRFS_I(inode
);
141 unsigned int new_fl
= 0;
143 if (ip
->flags
& BTRFS_INODE_SYNC
)
145 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
146 new_fl
|= S_IMMUTABLE
;
147 if (ip
->flags
& BTRFS_INODE_APPEND
)
149 if (ip
->flags
& BTRFS_INODE_NOATIME
)
151 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
154 set_mask_bits(&inode
->i_flags
,
155 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
,
160 * Inherit flags from the parent inode.
162 * Currently only the compression flags and the cow flags are inherited.
164 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
171 flags
= BTRFS_I(dir
)->flags
;
173 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
174 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
175 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
176 } else if (flags
& BTRFS_INODE_COMPRESS
) {
177 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
178 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
181 if (flags
& BTRFS_INODE_NODATACOW
) {
182 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
183 if (S_ISREG(inode
->i_mode
))
184 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
187 btrfs_update_iflags(inode
);
190 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
192 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
193 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
195 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
200 static int check_flags(unsigned int flags
)
202 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
203 FS_NOATIME_FL
| FS_NODUMP_FL
| \
204 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
205 FS_NOCOMP_FL
| FS_COMPR_FL
|
209 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
215 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
217 struct inode
*inode
= file_inode(file
);
218 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
219 struct btrfs_inode
*ip
= BTRFS_I(inode
);
220 struct btrfs_root
*root
= ip
->root
;
221 struct btrfs_trans_handle
*trans
;
222 unsigned int flags
, oldflags
;
225 unsigned int i_oldflags
;
228 if (!inode_owner_or_capable(inode
))
231 if (btrfs_root_readonly(root
))
234 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
237 ret
= check_flags(flags
);
241 ret
= mnt_want_write_file(file
);
247 ip_oldflags
= ip
->flags
;
248 i_oldflags
= inode
->i_flags
;
249 mode
= inode
->i_mode
;
251 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
252 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
253 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
254 if (!capable(CAP_LINUX_IMMUTABLE
)) {
260 if (flags
& FS_SYNC_FL
)
261 ip
->flags
|= BTRFS_INODE_SYNC
;
263 ip
->flags
&= ~BTRFS_INODE_SYNC
;
264 if (flags
& FS_IMMUTABLE_FL
)
265 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
267 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
268 if (flags
& FS_APPEND_FL
)
269 ip
->flags
|= BTRFS_INODE_APPEND
;
271 ip
->flags
&= ~BTRFS_INODE_APPEND
;
272 if (flags
& FS_NODUMP_FL
)
273 ip
->flags
|= BTRFS_INODE_NODUMP
;
275 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
276 if (flags
& FS_NOATIME_FL
)
277 ip
->flags
|= BTRFS_INODE_NOATIME
;
279 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
280 if (flags
& FS_DIRSYNC_FL
)
281 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
283 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
284 if (flags
& FS_NOCOW_FL
) {
287 * It's safe to turn csums off here, no extents exist.
288 * Otherwise we want the flag to reflect the real COW
289 * status of the file and will not set it.
291 if (inode
->i_size
== 0)
292 ip
->flags
|= BTRFS_INODE_NODATACOW
293 | BTRFS_INODE_NODATASUM
;
295 ip
->flags
|= BTRFS_INODE_NODATACOW
;
299 * Revert back under same assumptions as above
302 if (inode
->i_size
== 0)
303 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
304 | BTRFS_INODE_NODATASUM
);
306 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
311 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
312 * flag may be changed automatically if compression code won't make
315 if (flags
& FS_NOCOMP_FL
) {
316 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
317 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
319 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
320 if (ret
&& ret
!= -ENODATA
)
322 } else if (flags
& FS_COMPR_FL
) {
325 ip
->flags
|= BTRFS_INODE_COMPRESS
;
326 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
328 if (fs_info
->compress_type
== BTRFS_COMPRESS_LZO
)
332 ret
= btrfs_set_prop(inode
, "btrfs.compression",
333 comp
, strlen(comp
), 0);
338 ret
= btrfs_set_prop(inode
, "btrfs.compression", NULL
, 0, 0);
339 if (ret
&& ret
!= -ENODATA
)
341 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
344 trans
= btrfs_start_transaction(root
, 1);
346 ret
= PTR_ERR(trans
);
350 btrfs_update_iflags(inode
);
351 inode_inc_iversion(inode
);
352 inode
->i_ctime
= current_time(inode
);
353 ret
= btrfs_update_inode(trans
, root
, inode
);
355 btrfs_end_transaction(trans
, root
);
358 ip
->flags
= ip_oldflags
;
359 inode
->i_flags
= i_oldflags
;
364 mnt_drop_write_file(file
);
368 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
370 struct inode
*inode
= file_inode(file
);
372 return put_user(inode
->i_generation
, arg
);
375 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
377 struct inode
*inode
= file_inode(file
);
378 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
379 struct btrfs_device
*device
;
380 struct request_queue
*q
;
381 struct fstrim_range range
;
382 u64 minlen
= ULLONG_MAX
;
384 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
387 if (!capable(CAP_SYS_ADMIN
))
391 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
395 q
= bdev_get_queue(device
->bdev
);
396 if (blk_queue_discard(q
)) {
398 minlen
= min((u64
)q
->limits
.discard_granularity
,
406 if (copy_from_user(&range
, arg
, sizeof(range
)))
408 if (range
.start
> total_bytes
||
409 range
.len
< fs_info
->sb
->s_blocksize
)
412 range
.len
= min(range
.len
, total_bytes
- range
.start
);
413 range
.minlen
= max(range
.minlen
, minlen
);
414 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
418 if (copy_to_user(arg
, &range
, sizeof(range
)))
424 int btrfs_is_empty_uuid(u8
*uuid
)
428 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
435 static noinline
int create_subvol(struct inode
*dir
,
436 struct dentry
*dentry
,
437 char *name
, int namelen
,
439 struct btrfs_qgroup_inherit
*inherit
)
441 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
442 struct btrfs_trans_handle
*trans
;
443 struct btrfs_key key
;
444 struct btrfs_root_item
*root_item
;
445 struct btrfs_inode_item
*inode_item
;
446 struct extent_buffer
*leaf
;
447 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
448 struct btrfs_root
*new_root
;
449 struct btrfs_block_rsv block_rsv
;
450 struct timespec cur_time
= current_time(dir
);
455 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
460 root_item
= kzalloc(sizeof(*root_item
), GFP_KERNEL
);
464 ret
= btrfs_find_free_objectid(fs_info
->tree_root
, &objectid
);
469 * Don't create subvolume whose level is not zero. Or qgroup will be
470 * screwed up since it assumes subvolume qgroup's level to be 0.
472 if (btrfs_qgroup_level(objectid
)) {
477 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
479 * The same as the snapshot creation, please see the comment
480 * of create_snapshot().
482 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
483 8, &qgroup_reserved
, false);
487 trans
= btrfs_start_transaction(root
, 0);
489 ret
= PTR_ERR(trans
);
490 btrfs_subvolume_release_metadata(root
, &block_rsv
,
494 trans
->block_rsv
= &block_rsv
;
495 trans
->bytes_reserved
= block_rsv
.size
;
497 ret
= btrfs_qgroup_inherit(trans
, fs_info
, 0, objectid
, inherit
);
501 leaf
= btrfs_alloc_tree_block(trans
, root
, 0, objectid
, NULL
, 0, 0, 0);
507 memzero_extent_buffer(leaf
, 0, sizeof(struct btrfs_header
));
508 btrfs_set_header_bytenr(leaf
, leaf
->start
);
509 btrfs_set_header_generation(leaf
, trans
->transid
);
510 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
511 btrfs_set_header_owner(leaf
, objectid
);
513 write_extent_buffer_fsid(leaf
, fs_info
->fsid
);
514 write_extent_buffer_chunk_tree_uuid(leaf
, fs_info
->chunk_tree_uuid
);
515 btrfs_mark_buffer_dirty(leaf
);
517 inode_item
= &root_item
->inode
;
518 btrfs_set_stack_inode_generation(inode_item
, 1);
519 btrfs_set_stack_inode_size(inode_item
, 3);
520 btrfs_set_stack_inode_nlink(inode_item
, 1);
521 btrfs_set_stack_inode_nbytes(inode_item
,
523 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
525 btrfs_set_root_flags(root_item
, 0);
526 btrfs_set_root_limit(root_item
, 0);
527 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
529 btrfs_set_root_bytenr(root_item
, leaf
->start
);
530 btrfs_set_root_generation(root_item
, trans
->transid
);
531 btrfs_set_root_level(root_item
, 0);
532 btrfs_set_root_refs(root_item
, 1);
533 btrfs_set_root_used(root_item
, leaf
->len
);
534 btrfs_set_root_last_snapshot(root_item
, 0);
536 btrfs_set_root_generation_v2(root_item
,
537 btrfs_root_generation(root_item
));
538 uuid_le_gen(&new_uuid
);
539 memcpy(root_item
->uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
540 btrfs_set_stack_timespec_sec(&root_item
->otime
, cur_time
.tv_sec
);
541 btrfs_set_stack_timespec_nsec(&root_item
->otime
, cur_time
.tv_nsec
);
542 root_item
->ctime
= root_item
->otime
;
543 btrfs_set_root_ctransid(root_item
, trans
->transid
);
544 btrfs_set_root_otransid(root_item
, trans
->transid
);
546 btrfs_tree_unlock(leaf
);
547 free_extent_buffer(leaf
);
550 btrfs_set_root_dirid(root_item
, new_dirid
);
552 key
.objectid
= objectid
;
554 key
.type
= BTRFS_ROOT_ITEM_KEY
;
555 ret
= btrfs_insert_root(trans
, fs_info
->tree_root
, &key
,
560 key
.offset
= (u64
)-1;
561 new_root
= btrfs_read_fs_root_no_name(fs_info
, &key
);
562 if (IS_ERR(new_root
)) {
563 ret
= PTR_ERR(new_root
);
564 btrfs_abort_transaction(trans
, ret
);
568 btrfs_record_root_in_trans(trans
, new_root
);
570 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
572 /* We potentially lose an unused inode item here */
573 btrfs_abort_transaction(trans
, ret
);
577 mutex_lock(&new_root
->objectid_mutex
);
578 new_root
->highest_objectid
= new_dirid
;
579 mutex_unlock(&new_root
->objectid_mutex
);
582 * insert the directory item
584 ret
= btrfs_set_inode_index(dir
, &index
);
586 btrfs_abort_transaction(trans
, ret
);
590 ret
= btrfs_insert_dir_item(trans
, root
,
591 name
, namelen
, dir
, &key
,
592 BTRFS_FT_DIR
, index
);
594 btrfs_abort_transaction(trans
, ret
);
598 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
599 ret
= btrfs_update_inode(trans
, root
, dir
);
602 ret
= btrfs_add_root_ref(trans
, fs_info
,
603 objectid
, root
->root_key
.objectid
,
604 btrfs_ino(dir
), index
, name
, namelen
);
607 ret
= btrfs_uuid_tree_add(trans
, fs_info
, root_item
->uuid
,
608 BTRFS_UUID_KEY_SUBVOL
, objectid
);
610 btrfs_abort_transaction(trans
, ret
);
614 trans
->block_rsv
= NULL
;
615 trans
->bytes_reserved
= 0;
616 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
619 *async_transid
= trans
->transid
;
620 err
= btrfs_commit_transaction_async(trans
, root
, 1);
622 err
= btrfs_commit_transaction(trans
, root
);
624 err
= btrfs_commit_transaction(trans
, root
);
630 inode
= btrfs_lookup_dentry(dir
, dentry
);
632 return PTR_ERR(inode
);
633 d_instantiate(dentry
, inode
);
642 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root
*root
)
648 prepare_to_wait(&root
->subv_writers
->wait
, &wait
,
649 TASK_UNINTERRUPTIBLE
);
651 writers
= percpu_counter_sum(&root
->subv_writers
->counter
);
655 finish_wait(&root
->subv_writers
->wait
, &wait
);
659 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
660 struct dentry
*dentry
, char *name
, int namelen
,
661 u64
*async_transid
, bool readonly
,
662 struct btrfs_qgroup_inherit
*inherit
)
664 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
666 struct btrfs_pending_snapshot
*pending_snapshot
;
667 struct btrfs_trans_handle
*trans
;
670 if (!test_bit(BTRFS_ROOT_REF_COWS
, &root
->state
))
673 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
674 if (!pending_snapshot
)
677 pending_snapshot
->root_item
= kzalloc(sizeof(struct btrfs_root_item
),
679 pending_snapshot
->path
= btrfs_alloc_path();
680 if (!pending_snapshot
->root_item
|| !pending_snapshot
->path
) {
685 atomic_inc(&root
->will_be_snapshoted
);
686 smp_mb__after_atomic();
687 btrfs_wait_for_no_snapshoting_writes(root
);
689 ret
= btrfs_start_delalloc_inodes(root
, 0);
693 btrfs_wait_ordered_extents(root
, -1, 0, (u64
)-1);
695 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
696 BTRFS_BLOCK_RSV_TEMP
);
698 * 1 - parent dir inode
701 * 2 - root ref/backref
702 * 1 - root of snapshot
705 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
706 &pending_snapshot
->block_rsv
, 8,
707 &pending_snapshot
->qgroup_reserved
,
712 pending_snapshot
->dentry
= dentry
;
713 pending_snapshot
->root
= root
;
714 pending_snapshot
->readonly
= readonly
;
715 pending_snapshot
->dir
= dir
;
716 pending_snapshot
->inherit
= inherit
;
718 trans
= btrfs_start_transaction(root
, 0);
720 ret
= PTR_ERR(trans
);
724 spin_lock(&fs_info
->trans_lock
);
725 list_add(&pending_snapshot
->list
,
726 &trans
->transaction
->pending_snapshots
);
727 spin_unlock(&fs_info
->trans_lock
);
729 *async_transid
= trans
->transid
;
730 ret
= btrfs_commit_transaction_async(trans
,
731 fs_info
->extent_root
, 1);
733 ret
= btrfs_commit_transaction(trans
, root
);
735 ret
= btrfs_commit_transaction(trans
,
736 fs_info
->extent_root
);
741 ret
= pending_snapshot
->error
;
745 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
749 inode
= btrfs_lookup_dentry(d_inode(dentry
->d_parent
), dentry
);
751 ret
= PTR_ERR(inode
);
755 d_instantiate(dentry
, inode
);
758 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
759 &pending_snapshot
->block_rsv
,
760 pending_snapshot
->qgroup_reserved
);
762 if (atomic_dec_and_test(&root
->will_be_snapshoted
))
763 wake_up_atomic_t(&root
->will_be_snapshoted
);
765 kfree(pending_snapshot
->root_item
);
766 btrfs_free_path(pending_snapshot
->path
);
767 kfree(pending_snapshot
);
772 /* copy of may_delete in fs/namei.c()
773 * Check whether we can remove a link victim from directory dir, check
774 * whether the type of victim is right.
775 * 1. We can't do it if dir is read-only (done in permission())
776 * 2. We should have write and exec permissions on dir
777 * 3. We can't remove anything from append-only dir
778 * 4. We can't do anything with immutable dir (done in permission())
779 * 5. If the sticky bit on dir is set we should either
780 * a. be owner of dir, or
781 * b. be owner of victim, or
782 * c. have CAP_FOWNER capability
783 * 6. If the victim is append-only or immutable we can't do anything with
784 * links pointing to it.
785 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
786 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
787 * 9. We can't remove a root or mountpoint.
788 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
789 * nfs_async_unlink().
792 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
796 if (d_really_is_negative(victim
))
799 BUG_ON(d_inode(victim
->d_parent
) != dir
);
800 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
802 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
807 if (check_sticky(dir
, d_inode(victim
)) || IS_APPEND(d_inode(victim
)) ||
808 IS_IMMUTABLE(d_inode(victim
)) || IS_SWAPFILE(d_inode(victim
)))
811 if (!d_is_dir(victim
))
815 } else if (d_is_dir(victim
))
819 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
824 /* copy of may_create in fs/namei.c() */
825 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
827 if (d_really_is_positive(child
))
831 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
835 * Create a new subvolume below @parent. This is largely modeled after
836 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
837 * inside this filesystem so it's quite a bit simpler.
839 static noinline
int btrfs_mksubvol(struct path
*parent
,
840 char *name
, int namelen
,
841 struct btrfs_root
*snap_src
,
842 u64
*async_transid
, bool readonly
,
843 struct btrfs_qgroup_inherit
*inherit
)
845 struct inode
*dir
= d_inode(parent
->dentry
);
846 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
847 struct dentry
*dentry
;
850 error
= down_write_killable_nested(&dir
->i_rwsem
, I_MUTEX_PARENT
);
854 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
855 error
= PTR_ERR(dentry
);
859 error
= btrfs_may_create(dir
, dentry
);
864 * even if this name doesn't exist, we may get hash collisions.
865 * check for them now when we can safely fail
867 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
873 down_read(&fs_info
->subvol_sem
);
875 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
879 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
880 async_transid
, readonly
, inherit
);
882 error
= create_subvol(dir
, dentry
, name
, namelen
,
883 async_transid
, inherit
);
886 fsnotify_mkdir(dir
, dentry
);
888 up_read(&fs_info
->subvol_sem
);
897 * When we're defragging a range, we don't want to kick it off again
898 * if it is really just waiting for delalloc to send it down.
899 * If we find a nice big extent or delalloc range for the bytes in the
900 * file you want to defrag, we return 0 to let you know to skip this
903 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, u32 thresh
)
905 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
906 struct extent_map
*em
= NULL
;
907 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
910 read_lock(&em_tree
->lock
);
911 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_SIZE
);
912 read_unlock(&em_tree
->lock
);
915 end
= extent_map_end(em
);
917 if (end
- offset
> thresh
)
920 /* if we already have a nice delalloc here, just stop */
922 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
923 thresh
, EXTENT_DELALLOC
, 1);
930 * helper function to walk through a file and find extents
931 * newer than a specific transid, and smaller than thresh.
933 * This is used by the defragging code to find new and small
936 static int find_new_extents(struct btrfs_root
*root
,
937 struct inode
*inode
, u64 newer_than
,
938 u64
*off
, u32 thresh
)
940 struct btrfs_path
*path
;
941 struct btrfs_key min_key
;
942 struct extent_buffer
*leaf
;
943 struct btrfs_file_extent_item
*extent
;
946 u64 ino
= btrfs_ino(inode
);
948 path
= btrfs_alloc_path();
952 min_key
.objectid
= ino
;
953 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
954 min_key
.offset
= *off
;
957 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
961 if (min_key
.objectid
!= ino
)
963 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
966 leaf
= path
->nodes
[0];
967 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
968 struct btrfs_file_extent_item
);
970 type
= btrfs_file_extent_type(leaf
, extent
);
971 if (type
== BTRFS_FILE_EXTENT_REG
&&
972 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
973 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
974 *off
= min_key
.offset
;
975 btrfs_free_path(path
);
980 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
981 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
985 if (min_key
.offset
== (u64
)-1)
989 btrfs_release_path(path
);
992 btrfs_free_path(path
);
996 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
998 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
999 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1000 struct extent_map
*em
;
1001 u64 len
= PAGE_SIZE
;
1004 * hopefully we have this extent in the tree already, try without
1005 * the full extent lock
1007 read_lock(&em_tree
->lock
);
1008 em
= lookup_extent_mapping(em_tree
, start
, len
);
1009 read_unlock(&em_tree
->lock
);
1012 struct extent_state
*cached
= NULL
;
1013 u64 end
= start
+ len
- 1;
1015 /* get the big lock and read metadata off disk */
1016 lock_extent_bits(io_tree
, start
, end
, &cached
);
1017 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
1018 unlock_extent_cached(io_tree
, start
, end
, &cached
, GFP_NOFS
);
1027 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1029 struct extent_map
*next
;
1032 /* this is the last extent */
1033 if (em
->start
+ em
->len
>= i_size_read(inode
))
1036 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1037 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
1039 else if ((em
->block_start
+ em
->block_len
== next
->block_start
) &&
1040 (em
->block_len
> SZ_128K
&& next
->block_len
> SZ_128K
))
1043 free_extent_map(next
);
1047 static int should_defrag_range(struct inode
*inode
, u64 start
, u32 thresh
,
1048 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1051 struct extent_map
*em
;
1053 bool next_mergeable
= true;
1054 bool prev_mergeable
= true;
1057 * make sure that once we start defragging an extent, we keep on
1060 if (start
< *defrag_end
)
1065 em
= defrag_lookup_extent(inode
, start
);
1069 /* this will cover holes, and inline extents */
1070 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1076 prev_mergeable
= false;
1078 next_mergeable
= defrag_check_next_extent(inode
, em
);
1080 * we hit a real extent, if it is big or the next extent is not a
1081 * real extent, don't bother defragging it
1083 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1084 (em
->len
>= thresh
|| (!next_mergeable
&& !prev_mergeable
)))
1088 * last_len ends up being a counter of how many bytes we've defragged.
1089 * every time we choose not to defrag an extent, we reset *last_len
1090 * so that the next tiny extent will force a defrag.
1092 * The end result of this is that tiny extents before a single big
1093 * extent will force at least part of that big extent to be defragged.
1096 *defrag_end
= extent_map_end(em
);
1099 *skip
= extent_map_end(em
);
1103 free_extent_map(em
);
1108 * it doesn't do much good to defrag one or two pages
1109 * at a time. This pulls in a nice chunk of pages
1110 * to COW and defrag.
1112 * It also makes sure the delalloc code has enough
1113 * dirty data to avoid making new small extents as part
1116 * It's a good idea to start RA on this range
1117 * before calling this.
1119 static int cluster_pages_for_defrag(struct inode
*inode
,
1120 struct page
**pages
,
1121 unsigned long start_index
,
1122 unsigned long num_pages
)
1124 unsigned long file_end
;
1125 u64 isize
= i_size_read(inode
);
1132 struct btrfs_ordered_extent
*ordered
;
1133 struct extent_state
*cached_state
= NULL
;
1134 struct extent_io_tree
*tree
;
1135 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1137 file_end
= (isize
- 1) >> PAGE_SHIFT
;
1138 if (!isize
|| start_index
> file_end
)
1141 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1143 ret
= btrfs_delalloc_reserve_space(inode
,
1144 start_index
<< PAGE_SHIFT
,
1145 page_cnt
<< PAGE_SHIFT
);
1149 tree
= &BTRFS_I(inode
)->io_tree
;
1151 /* step one, lock all the pages */
1152 for (i
= 0; i
< page_cnt
; i
++) {
1155 page
= find_or_create_page(inode
->i_mapping
,
1156 start_index
+ i
, mask
);
1160 page_start
= page_offset(page
);
1161 page_end
= page_start
+ PAGE_SIZE
- 1;
1163 lock_extent_bits(tree
, page_start
, page_end
,
1165 ordered
= btrfs_lookup_ordered_extent(inode
,
1167 unlock_extent_cached(tree
, page_start
, page_end
,
1168 &cached_state
, GFP_NOFS
);
1173 btrfs_start_ordered_extent(inode
, ordered
, 1);
1174 btrfs_put_ordered_extent(ordered
);
1177 * we unlocked the page above, so we need check if
1178 * it was released or not.
1180 if (page
->mapping
!= inode
->i_mapping
) {
1187 if (!PageUptodate(page
)) {
1188 btrfs_readpage(NULL
, page
);
1190 if (!PageUptodate(page
)) {
1198 if (page
->mapping
!= inode
->i_mapping
) {
1210 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1214 * so now we have a nice long stream of locked
1215 * and up to date pages, lets wait on them
1217 for (i
= 0; i
< i_done
; i
++)
1218 wait_on_page_writeback(pages
[i
]);
1220 page_start
= page_offset(pages
[0]);
1221 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_SIZE
;
1223 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1224 page_start
, page_end
- 1, &cached_state
);
1225 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1226 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1227 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1228 &cached_state
, GFP_NOFS
);
1230 if (i_done
!= page_cnt
) {
1231 spin_lock(&BTRFS_I(inode
)->lock
);
1232 BTRFS_I(inode
)->outstanding_extents
++;
1233 spin_unlock(&BTRFS_I(inode
)->lock
);
1234 btrfs_delalloc_release_space(inode
,
1235 start_index
<< PAGE_SHIFT
,
1236 (page_cnt
- i_done
) << PAGE_SHIFT
);
1240 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1243 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1244 page_start
, page_end
- 1, &cached_state
,
1247 for (i
= 0; i
< i_done
; i
++) {
1248 clear_page_dirty_for_io(pages
[i
]);
1249 ClearPageChecked(pages
[i
]);
1250 set_page_extent_mapped(pages
[i
]);
1251 set_page_dirty(pages
[i
]);
1252 unlock_page(pages
[i
]);
1257 for (i
= 0; i
< i_done
; i
++) {
1258 unlock_page(pages
[i
]);
1261 btrfs_delalloc_release_space(inode
,
1262 start_index
<< PAGE_SHIFT
,
1263 page_cnt
<< PAGE_SHIFT
);
1268 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1269 struct btrfs_ioctl_defrag_range_args
*range
,
1270 u64 newer_than
, unsigned long max_to_defrag
)
1272 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1273 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1274 struct file_ra_state
*ra
= NULL
;
1275 unsigned long last_index
;
1276 u64 isize
= i_size_read(inode
);
1280 u64 newer_off
= range
->start
;
1282 unsigned long ra_index
= 0;
1284 int defrag_count
= 0;
1285 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1286 u32 extent_thresh
= range
->extent_thresh
;
1287 unsigned long max_cluster
= SZ_256K
>> PAGE_SHIFT
;
1288 unsigned long cluster
= max_cluster
;
1289 u64 new_align
= ~((u64
)SZ_128K
- 1);
1290 struct page
**pages
= NULL
;
1295 if (range
->start
>= isize
)
1298 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1299 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1301 if (range
->compress_type
)
1302 compress_type
= range
->compress_type
;
1305 if (extent_thresh
== 0)
1306 extent_thresh
= SZ_256K
;
1309 * if we were not given a file, allocate a readahead
1313 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1316 file_ra_state_init(ra
, inode
->i_mapping
);
1321 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*),
1328 /* find the last page to defrag */
1329 if (range
->start
+ range
->len
> range
->start
) {
1330 last_index
= min_t(u64
, isize
- 1,
1331 range
->start
+ range
->len
- 1) >> PAGE_SHIFT
;
1333 last_index
= (isize
- 1) >> PAGE_SHIFT
;
1337 ret
= find_new_extents(root
, inode
, newer_than
,
1338 &newer_off
, SZ_64K
);
1340 range
->start
= newer_off
;
1342 * we always align our defrag to help keep
1343 * the extents in the file evenly spaced
1345 i
= (newer_off
& new_align
) >> PAGE_SHIFT
;
1349 i
= range
->start
>> PAGE_SHIFT
;
1352 max_to_defrag
= last_index
- i
+ 1;
1355 * make writeback starts from i, so the defrag range can be
1356 * written sequentially.
1358 if (i
< inode
->i_mapping
->writeback_index
)
1359 inode
->i_mapping
->writeback_index
= i
;
1361 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1362 (i
< DIV_ROUND_UP(i_size_read(inode
), PAGE_SIZE
))) {
1364 * make sure we stop running if someone unmounts
1367 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1370 if (btrfs_defrag_cancelled(fs_info
)) {
1371 btrfs_debug(fs_info
, "defrag_file cancelled");
1376 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_SHIFT
,
1377 extent_thresh
, &last_len
, &skip
,
1378 &defrag_end
, range
->flags
&
1379 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1382 * the should_defrag function tells us how much to skip
1383 * bump our counter by the suggested amount
1385 next
= DIV_ROUND_UP(skip
, PAGE_SIZE
);
1386 i
= max(i
+ 1, next
);
1391 cluster
= (PAGE_ALIGN(defrag_end
) >>
1393 cluster
= min(cluster
, max_cluster
);
1395 cluster
= max_cluster
;
1398 if (i
+ cluster
> ra_index
) {
1399 ra_index
= max(i
, ra_index
);
1400 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1402 ra_index
+= cluster
;
1406 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1407 BTRFS_I(inode
)->force_compress
= compress_type
;
1408 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1410 inode_unlock(inode
);
1414 defrag_count
+= ret
;
1415 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1416 inode_unlock(inode
);
1419 if (newer_off
== (u64
)-1)
1425 newer_off
= max(newer_off
+ 1,
1426 (u64
)i
<< PAGE_SHIFT
);
1428 ret
= find_new_extents(root
, inode
, newer_than
,
1429 &newer_off
, SZ_64K
);
1431 range
->start
= newer_off
;
1432 i
= (newer_off
& new_align
) >> PAGE_SHIFT
;
1439 last_len
+= ret
<< PAGE_SHIFT
;
1447 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1448 filemap_flush(inode
->i_mapping
);
1449 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1450 &BTRFS_I(inode
)->runtime_flags
))
1451 filemap_flush(inode
->i_mapping
);
1454 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1455 /* the filemap_flush will queue IO into the worker threads, but
1456 * we have to make sure the IO is actually started and that
1457 * ordered extents get created before we return
1459 atomic_inc(&fs_info
->async_submit_draining
);
1460 while (atomic_read(&fs_info
->nr_async_submits
) ||
1461 atomic_read(&fs_info
->async_delalloc_pages
)) {
1462 wait_event(fs_info
->async_submit_wait
,
1463 (atomic_read(&fs_info
->nr_async_submits
) == 0 &&
1464 atomic_read(&fs_info
->async_delalloc_pages
) == 0));
1466 atomic_dec(&fs_info
->async_submit_draining
);
1469 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1470 btrfs_set_fs_incompat(fs_info
, COMPRESS_LZO
);
1476 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1478 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1479 inode_unlock(inode
);
1487 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1490 struct inode
*inode
= file_inode(file
);
1491 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1495 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1496 struct btrfs_ioctl_vol_args
*vol_args
;
1497 struct btrfs_trans_handle
*trans
;
1498 struct btrfs_device
*device
= NULL
;
1501 char *devstr
= NULL
;
1505 if (!capable(CAP_SYS_ADMIN
))
1508 ret
= mnt_want_write_file(file
);
1512 if (atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
1513 mnt_drop_write_file(file
);
1514 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1517 mutex_lock(&fs_info
->volume_mutex
);
1518 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1519 if (IS_ERR(vol_args
)) {
1520 ret
= PTR_ERR(vol_args
);
1524 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1526 sizestr
= vol_args
->name
;
1527 devstr
= strchr(sizestr
, ':');
1529 sizestr
= devstr
+ 1;
1531 devstr
= vol_args
->name
;
1532 ret
= kstrtoull(devstr
, 10, &devid
);
1539 btrfs_info(fs_info
, "resizing devid %llu", devid
);
1542 device
= btrfs_find_device(fs_info
, devid
, NULL
, NULL
);
1544 btrfs_info(fs_info
, "resizer unable to find device %llu",
1550 if (!device
->writeable
) {
1552 "resizer unable to apply on readonly device %llu",
1558 if (!strcmp(sizestr
, "max"))
1559 new_size
= device
->bdev
->bd_inode
->i_size
;
1561 if (sizestr
[0] == '-') {
1564 } else if (sizestr
[0] == '+') {
1568 new_size
= memparse(sizestr
, &retptr
);
1569 if (*retptr
!= '\0' || new_size
== 0) {
1575 if (device
->is_tgtdev_for_dev_replace
) {
1580 old_size
= btrfs_device_get_total_bytes(device
);
1583 if (new_size
> old_size
) {
1587 new_size
= old_size
- new_size
;
1588 } else if (mod
> 0) {
1589 if (new_size
> ULLONG_MAX
- old_size
) {
1593 new_size
= old_size
+ new_size
;
1596 if (new_size
< SZ_256M
) {
1600 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1605 new_size
= div_u64(new_size
, fs_info
->sectorsize
);
1606 new_size
*= fs_info
->sectorsize
;
1608 btrfs_info_in_rcu(fs_info
, "new size for %s is %llu",
1609 rcu_str_deref(device
->name
), new_size
);
1611 if (new_size
> old_size
) {
1612 trans
= btrfs_start_transaction(root
, 0);
1613 if (IS_ERR(trans
)) {
1614 ret
= PTR_ERR(trans
);
1617 ret
= btrfs_grow_device(trans
, device
, new_size
);
1618 btrfs_commit_transaction(trans
, root
);
1619 } else if (new_size
< old_size
) {
1620 ret
= btrfs_shrink_device(device
, new_size
);
1621 } /* equal, nothing need to do */
1626 mutex_unlock(&fs_info
->volume_mutex
);
1627 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
1628 mnt_drop_write_file(file
);
1632 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1633 char *name
, unsigned long fd
, int subvol
,
1634 u64
*transid
, bool readonly
,
1635 struct btrfs_qgroup_inherit
*inherit
)
1640 if (!S_ISDIR(file_inode(file
)->i_mode
))
1643 ret
= mnt_want_write_file(file
);
1647 namelen
= strlen(name
);
1648 if (strchr(name
, '/')) {
1650 goto out_drop_write
;
1653 if (name
[0] == '.' &&
1654 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1656 goto out_drop_write
;
1660 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1661 NULL
, transid
, readonly
, inherit
);
1663 struct fd src
= fdget(fd
);
1664 struct inode
*src_inode
;
1667 goto out_drop_write
;
1670 src_inode
= file_inode(src
.file
);
1671 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1672 btrfs_info(BTRFS_I(file_inode(file
))->root
->fs_info
,
1673 "Snapshot src from another FS");
1675 } else if (!inode_owner_or_capable(src_inode
)) {
1677 * Subvolume creation is not restricted, but snapshots
1678 * are limited to own subvolumes only
1682 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1683 BTRFS_I(src_inode
)->root
,
1684 transid
, readonly
, inherit
);
1689 mnt_drop_write_file(file
);
1694 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1695 void __user
*arg
, int subvol
)
1697 struct btrfs_ioctl_vol_args
*vol_args
;
1700 if (!S_ISDIR(file_inode(file
)->i_mode
))
1703 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1704 if (IS_ERR(vol_args
))
1705 return PTR_ERR(vol_args
);
1706 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1708 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1709 vol_args
->fd
, subvol
,
1716 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1717 void __user
*arg
, int subvol
)
1719 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1723 bool readonly
= false;
1724 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1726 if (!S_ISDIR(file_inode(file
)->i_mode
))
1729 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1730 if (IS_ERR(vol_args
))
1731 return PTR_ERR(vol_args
);
1732 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1734 if (vol_args
->flags
&
1735 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1736 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1741 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1743 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1745 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1746 if (vol_args
->size
> PAGE_SIZE
) {
1750 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1751 if (IS_ERR(inherit
)) {
1752 ret
= PTR_ERR(inherit
);
1757 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1758 vol_args
->fd
, subvol
, ptr
,
1763 if (ptr
&& copy_to_user(arg
+
1764 offsetof(struct btrfs_ioctl_vol_args_v2
,
1776 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1779 struct inode
*inode
= file_inode(file
);
1780 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1781 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1785 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1788 down_read(&fs_info
->subvol_sem
);
1789 if (btrfs_root_readonly(root
))
1790 flags
|= BTRFS_SUBVOL_RDONLY
;
1791 up_read(&fs_info
->subvol_sem
);
1793 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1799 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1802 struct inode
*inode
= file_inode(file
);
1803 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1804 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1805 struct btrfs_trans_handle
*trans
;
1810 if (!inode_owner_or_capable(inode
))
1813 ret
= mnt_want_write_file(file
);
1817 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1819 goto out_drop_write
;
1822 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1824 goto out_drop_write
;
1827 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1829 goto out_drop_write
;
1832 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1834 goto out_drop_write
;
1837 down_write(&fs_info
->subvol_sem
);
1840 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1843 root_flags
= btrfs_root_flags(&root
->root_item
);
1844 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1845 btrfs_set_root_flags(&root
->root_item
,
1846 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1849 * Block RO -> RW transition if this subvolume is involved in
1852 spin_lock(&root
->root_item_lock
);
1853 if (root
->send_in_progress
== 0) {
1854 btrfs_set_root_flags(&root
->root_item
,
1855 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1856 spin_unlock(&root
->root_item_lock
);
1858 spin_unlock(&root
->root_item_lock
);
1860 "Attempt to set subvolume %llu read-write during send",
1861 root
->root_key
.objectid
);
1867 trans
= btrfs_start_transaction(root
, 1);
1868 if (IS_ERR(trans
)) {
1869 ret
= PTR_ERR(trans
);
1873 ret
= btrfs_update_root(trans
, fs_info
->tree_root
,
1874 &root
->root_key
, &root
->root_item
);
1876 btrfs_commit_transaction(trans
, root
);
1879 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1881 up_write(&fs_info
->subvol_sem
);
1883 mnt_drop_write_file(file
);
1889 * helper to check if the subvolume references other subvolumes
1891 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1893 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1894 struct btrfs_path
*path
;
1895 struct btrfs_dir_item
*di
;
1896 struct btrfs_key key
;
1900 path
= btrfs_alloc_path();
1904 /* Make sure this root isn't set as the default subvol */
1905 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
1906 di
= btrfs_lookup_dir_item(NULL
, fs_info
->tree_root
, path
,
1907 dir_id
, "default", 7, 0);
1908 if (di
&& !IS_ERR(di
)) {
1909 btrfs_dir_item_key_to_cpu(path
->nodes
[0], di
, &key
);
1910 if (key
.objectid
== root
->root_key
.objectid
) {
1913 "deleting default subvolume %llu is not allowed",
1917 btrfs_release_path(path
);
1920 key
.objectid
= root
->root_key
.objectid
;
1921 key
.type
= BTRFS_ROOT_REF_KEY
;
1922 key
.offset
= (u64
)-1;
1924 ret
= btrfs_search_slot(NULL
, fs_info
->tree_root
, &key
, path
, 0, 0);
1930 if (path
->slots
[0] > 0) {
1932 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1933 if (key
.objectid
== root
->root_key
.objectid
&&
1934 key
.type
== BTRFS_ROOT_REF_KEY
)
1938 btrfs_free_path(path
);
1942 static noinline
int key_in_sk(struct btrfs_key
*key
,
1943 struct btrfs_ioctl_search_key
*sk
)
1945 struct btrfs_key test
;
1948 test
.objectid
= sk
->min_objectid
;
1949 test
.type
= sk
->min_type
;
1950 test
.offset
= sk
->min_offset
;
1952 ret
= btrfs_comp_cpu_keys(key
, &test
);
1956 test
.objectid
= sk
->max_objectid
;
1957 test
.type
= sk
->max_type
;
1958 test
.offset
= sk
->max_offset
;
1960 ret
= btrfs_comp_cpu_keys(key
, &test
);
1966 static noinline
int copy_to_sk(struct btrfs_path
*path
,
1967 struct btrfs_key
*key
,
1968 struct btrfs_ioctl_search_key
*sk
,
1971 unsigned long *sk_offset
,
1975 struct extent_buffer
*leaf
;
1976 struct btrfs_ioctl_search_header sh
;
1977 struct btrfs_key test
;
1978 unsigned long item_off
;
1979 unsigned long item_len
;
1985 leaf
= path
->nodes
[0];
1986 slot
= path
->slots
[0];
1987 nritems
= btrfs_header_nritems(leaf
);
1989 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1993 found_transid
= btrfs_header_generation(leaf
);
1995 for (i
= slot
; i
< nritems
; i
++) {
1996 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1997 item_len
= btrfs_item_size_nr(leaf
, i
);
1999 btrfs_item_key_to_cpu(leaf
, key
, i
);
2000 if (!key_in_sk(key
, sk
))
2003 if (sizeof(sh
) + item_len
> *buf_size
) {
2010 * return one empty item back for v1, which does not
2014 *buf_size
= sizeof(sh
) + item_len
;
2019 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
2024 sh
.objectid
= key
->objectid
;
2025 sh
.offset
= key
->offset
;
2026 sh
.type
= key
->type
;
2028 sh
.transid
= found_transid
;
2030 /* copy search result header */
2031 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2036 *sk_offset
+= sizeof(sh
);
2039 char __user
*up
= ubuf
+ *sk_offset
;
2041 if (read_extent_buffer_to_user(leaf
, up
,
2042 item_off
, item_len
)) {
2047 *sk_offset
+= item_len
;
2051 if (ret
) /* -EOVERFLOW from above */
2054 if (*num_found
>= sk
->nr_items
) {
2061 test
.objectid
= sk
->max_objectid
;
2062 test
.type
= sk
->max_type
;
2063 test
.offset
= sk
->max_offset
;
2064 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
2066 else if (key
->offset
< (u64
)-1)
2068 else if (key
->type
< (u8
)-1) {
2071 } else if (key
->objectid
< (u64
)-1) {
2079 * 0: all items from this leaf copied, continue with next
2080 * 1: * more items can be copied, but unused buffer is too small
2081 * * all items were found
2082 * Either way, it will stops the loop which iterates to the next
2084 * -EOVERFLOW: item was to large for buffer
2085 * -EFAULT: could not copy extent buffer back to userspace
2090 static noinline
int search_ioctl(struct inode
*inode
,
2091 struct btrfs_ioctl_search_key
*sk
,
2095 struct btrfs_fs_info
*info
= btrfs_sb(inode
->i_sb
);
2096 struct btrfs_root
*root
;
2097 struct btrfs_key key
;
2098 struct btrfs_path
*path
;
2101 unsigned long sk_offset
= 0;
2103 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2104 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2108 path
= btrfs_alloc_path();
2112 if (sk
->tree_id
== 0) {
2113 /* search the root of the inode that was passed */
2114 root
= BTRFS_I(inode
)->root
;
2116 key
.objectid
= sk
->tree_id
;
2117 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2118 key
.offset
= (u64
)-1;
2119 root
= btrfs_read_fs_root_no_name(info
, &key
);
2121 btrfs_free_path(path
);
2126 key
.objectid
= sk
->min_objectid
;
2127 key
.type
= sk
->min_type
;
2128 key
.offset
= sk
->min_offset
;
2131 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2137 ret
= copy_to_sk(path
, &key
, sk
, buf_size
, ubuf
,
2138 &sk_offset
, &num_found
);
2139 btrfs_release_path(path
);
2147 sk
->nr_items
= num_found
;
2148 btrfs_free_path(path
);
2152 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2155 struct btrfs_ioctl_search_args __user
*uargs
;
2156 struct btrfs_ioctl_search_key sk
;
2157 struct inode
*inode
;
2161 if (!capable(CAP_SYS_ADMIN
))
2164 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2166 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2169 buf_size
= sizeof(uargs
->buf
);
2171 inode
= file_inode(file
);
2172 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2175 * In the origin implementation an overflow is handled by returning a
2176 * search header with a len of zero, so reset ret.
2178 if (ret
== -EOVERFLOW
)
2181 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2186 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2189 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2190 struct btrfs_ioctl_search_args_v2 args
;
2191 struct inode
*inode
;
2194 const size_t buf_limit
= SZ_16M
;
2196 if (!capable(CAP_SYS_ADMIN
))
2199 /* copy search header and buffer size */
2200 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2201 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2204 buf_size
= args
.buf_size
;
2206 if (buf_size
< sizeof(struct btrfs_ioctl_search_header
))
2209 /* limit result size to 16MB */
2210 if (buf_size
> buf_limit
)
2211 buf_size
= buf_limit
;
2213 inode
= file_inode(file
);
2214 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2215 (char *)(&uarg
->buf
[0]));
2216 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2218 else if (ret
== -EOVERFLOW
&&
2219 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2226 * Search INODE_REFs to identify path name of 'dirid' directory
2227 * in a 'tree_id' tree. and sets path name to 'name'.
2229 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2230 u64 tree_id
, u64 dirid
, char *name
)
2232 struct btrfs_root
*root
;
2233 struct btrfs_key key
;
2239 struct btrfs_inode_ref
*iref
;
2240 struct extent_buffer
*l
;
2241 struct btrfs_path
*path
;
2243 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2248 path
= btrfs_alloc_path();
2252 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
2254 key
.objectid
= tree_id
;
2255 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2256 key
.offset
= (u64
)-1;
2257 root
= btrfs_read_fs_root_no_name(info
, &key
);
2259 btrfs_err(info
, "could not find root %llu", tree_id
);
2264 key
.objectid
= dirid
;
2265 key
.type
= BTRFS_INODE_REF_KEY
;
2266 key
.offset
= (u64
)-1;
2269 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2273 ret
= btrfs_previous_item(root
, path
, dirid
,
2274 BTRFS_INODE_REF_KEY
);
2284 slot
= path
->slots
[0];
2285 btrfs_item_key_to_cpu(l
, &key
, slot
);
2287 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2288 len
= btrfs_inode_ref_name_len(l
, iref
);
2290 total_len
+= len
+ 1;
2292 ret
= -ENAMETOOLONG
;
2297 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2299 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2302 btrfs_release_path(path
);
2303 key
.objectid
= key
.offset
;
2304 key
.offset
= (u64
)-1;
2305 dirid
= key
.objectid
;
2307 memmove(name
, ptr
, total_len
);
2308 name
[total_len
] = '\0';
2311 btrfs_free_path(path
);
2315 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2318 struct btrfs_ioctl_ino_lookup_args
*args
;
2319 struct inode
*inode
;
2322 args
= memdup_user(argp
, sizeof(*args
));
2324 return PTR_ERR(args
);
2326 inode
= file_inode(file
);
2329 * Unprivileged query to obtain the containing subvolume root id. The
2330 * path is reset so it's consistent with btrfs_search_path_in_tree.
2332 if (args
->treeid
== 0)
2333 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2335 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2340 if (!capable(CAP_SYS_ADMIN
)) {
2345 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2346 args
->treeid
, args
->objectid
,
2350 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2357 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2360 struct dentry
*parent
= file
->f_path
.dentry
;
2361 struct btrfs_fs_info
*fs_info
= btrfs_sb(parent
->d_sb
);
2362 struct dentry
*dentry
;
2363 struct inode
*dir
= d_inode(parent
);
2364 struct inode
*inode
;
2365 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2366 struct btrfs_root
*dest
= NULL
;
2367 struct btrfs_ioctl_vol_args
*vol_args
;
2368 struct btrfs_trans_handle
*trans
;
2369 struct btrfs_block_rsv block_rsv
;
2371 u64 qgroup_reserved
;
2376 if (!S_ISDIR(dir
->i_mode
))
2379 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2380 if (IS_ERR(vol_args
))
2381 return PTR_ERR(vol_args
);
2383 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2384 namelen
= strlen(vol_args
->name
);
2385 if (strchr(vol_args
->name
, '/') ||
2386 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2391 err
= mnt_want_write_file(file
);
2396 err
= down_write_killable_nested(&dir
->i_rwsem
, I_MUTEX_PARENT
);
2398 goto out_drop_write
;
2399 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2400 if (IS_ERR(dentry
)) {
2401 err
= PTR_ERR(dentry
);
2402 goto out_unlock_dir
;
2405 if (d_really_is_negative(dentry
)) {
2410 inode
= d_inode(dentry
);
2411 dest
= BTRFS_I(inode
)->root
;
2412 if (!capable(CAP_SYS_ADMIN
)) {
2414 * Regular user. Only allow this with a special mount
2415 * option, when the user has write+exec access to the
2416 * subvol root, and when rmdir(2) would have been
2419 * Note that this is _not_ check that the subvol is
2420 * empty or doesn't contain data that we wouldn't
2421 * otherwise be able to delete.
2423 * Users who want to delete empty subvols should try
2427 if (!btrfs_test_opt(fs_info
, USER_SUBVOL_RM_ALLOWED
))
2431 * Do not allow deletion if the parent dir is the same
2432 * as the dir to be deleted. That means the ioctl
2433 * must be called on the dentry referencing the root
2434 * of the subvol, not a random directory contained
2441 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2446 /* check if subvolume may be deleted by a user */
2447 err
= btrfs_may_delete(dir
, dentry
, 1);
2451 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2459 * Don't allow to delete a subvolume with send in progress. This is
2460 * inside the i_mutex so the error handling that has to drop the bit
2461 * again is not run concurrently.
2463 spin_lock(&dest
->root_item_lock
);
2464 root_flags
= btrfs_root_flags(&dest
->root_item
);
2465 if (dest
->send_in_progress
== 0) {
2466 btrfs_set_root_flags(&dest
->root_item
,
2467 root_flags
| BTRFS_ROOT_SUBVOL_DEAD
);
2468 spin_unlock(&dest
->root_item_lock
);
2470 spin_unlock(&dest
->root_item_lock
);
2472 "Attempt to delete subvolume %llu during send",
2473 dest
->root_key
.objectid
);
2475 goto out_unlock_inode
;
2478 down_write(&fs_info
->subvol_sem
);
2480 err
= may_destroy_subvol(dest
);
2484 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2486 * One for dir inode, two for dir entries, two for root
2489 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2490 5, &qgroup_reserved
, true);
2494 trans
= btrfs_start_transaction(root
, 0);
2495 if (IS_ERR(trans
)) {
2496 err
= PTR_ERR(trans
);
2499 trans
->block_rsv
= &block_rsv
;
2500 trans
->bytes_reserved
= block_rsv
.size
;
2502 btrfs_record_snapshot_destroy(trans
, dir
);
2504 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2505 dest
->root_key
.objectid
,
2506 dentry
->d_name
.name
,
2507 dentry
->d_name
.len
);
2510 btrfs_abort_transaction(trans
, ret
);
2514 btrfs_record_root_in_trans(trans
, dest
);
2516 memset(&dest
->root_item
.drop_progress
, 0,
2517 sizeof(dest
->root_item
.drop_progress
));
2518 dest
->root_item
.drop_level
= 0;
2519 btrfs_set_root_refs(&dest
->root_item
, 0);
2521 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED
, &dest
->state
)) {
2522 ret
= btrfs_insert_orphan_item(trans
,
2524 dest
->root_key
.objectid
);
2526 btrfs_abort_transaction(trans
, ret
);
2532 ret
= btrfs_uuid_tree_rem(trans
, fs_info
, dest
->root_item
.uuid
,
2533 BTRFS_UUID_KEY_SUBVOL
,
2534 dest
->root_key
.objectid
);
2535 if (ret
&& ret
!= -ENOENT
) {
2536 btrfs_abort_transaction(trans
, ret
);
2540 if (!btrfs_is_empty_uuid(dest
->root_item
.received_uuid
)) {
2541 ret
= btrfs_uuid_tree_rem(trans
, fs_info
,
2542 dest
->root_item
.received_uuid
,
2543 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
2544 dest
->root_key
.objectid
);
2545 if (ret
&& ret
!= -ENOENT
) {
2546 btrfs_abort_transaction(trans
, ret
);
2553 trans
->block_rsv
= NULL
;
2554 trans
->bytes_reserved
= 0;
2555 ret
= btrfs_end_transaction(trans
, root
);
2558 inode
->i_flags
|= S_DEAD
;
2560 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2562 up_write(&fs_info
->subvol_sem
);
2564 spin_lock(&dest
->root_item_lock
);
2565 root_flags
= btrfs_root_flags(&dest
->root_item
);
2566 btrfs_set_root_flags(&dest
->root_item
,
2567 root_flags
& ~BTRFS_ROOT_SUBVOL_DEAD
);
2568 spin_unlock(&dest
->root_item_lock
);
2571 inode_unlock(inode
);
2573 d_invalidate(dentry
);
2574 btrfs_invalidate_inodes(dest
);
2576 ASSERT(dest
->send_in_progress
== 0);
2579 if (dest
->ino_cache_inode
) {
2580 iput(dest
->ino_cache_inode
);
2581 dest
->ino_cache_inode
= NULL
;
2589 mnt_drop_write_file(file
);
2595 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2597 struct inode
*inode
= file_inode(file
);
2598 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2599 struct btrfs_ioctl_defrag_range_args
*range
;
2602 ret
= mnt_want_write_file(file
);
2606 if (btrfs_root_readonly(root
)) {
2611 switch (inode
->i_mode
& S_IFMT
) {
2613 if (!capable(CAP_SYS_ADMIN
)) {
2617 ret
= btrfs_defrag_root(root
);
2620 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2623 if (!(file
->f_mode
& FMODE_WRITE
)) {
2628 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2635 if (copy_from_user(range
, argp
,
2641 /* compression requires us to start the IO */
2642 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2643 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2644 range
->extent_thresh
= (u32
)-1;
2647 /* the rest are all set to zero by kzalloc */
2648 range
->len
= (u64
)-1;
2650 ret
= btrfs_defrag_file(file_inode(file
), file
,
2660 mnt_drop_write_file(file
);
2664 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2666 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
2667 struct btrfs_ioctl_vol_args
*vol_args
;
2670 if (!capable(CAP_SYS_ADMIN
))
2673 if (atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1))
2674 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2676 mutex_lock(&fs_info
->volume_mutex
);
2677 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2678 if (IS_ERR(vol_args
)) {
2679 ret
= PTR_ERR(vol_args
);
2683 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2684 ret
= btrfs_init_new_device(root
->fs_info
, vol_args
->name
);
2687 btrfs_info(fs_info
, "disk added %s", vol_args
->name
);
2691 mutex_unlock(&fs_info
->volume_mutex
);
2692 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
2696 static long btrfs_ioctl_rm_dev_v2(struct file
*file
, void __user
*arg
)
2698 struct inode
*inode
= file_inode(file
);
2699 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
2700 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2701 struct btrfs_ioctl_vol_args_v2
*vol_args
;
2704 if (!capable(CAP_SYS_ADMIN
))
2707 ret
= mnt_want_write_file(file
);
2711 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2712 if (IS_ERR(vol_args
)) {
2713 ret
= PTR_ERR(vol_args
);
2717 /* Check for compatibility reject unknown flags */
2718 if (vol_args
->flags
& ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED
)
2721 if (atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
2722 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2726 mutex_lock(&fs_info
->volume_mutex
);
2727 if (vol_args
->flags
& BTRFS_DEVICE_SPEC_BY_ID
) {
2728 ret
= btrfs_rm_device(root
, NULL
, vol_args
->devid
);
2730 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
2731 ret
= btrfs_rm_device(root
, vol_args
->name
, 0);
2733 mutex_unlock(&fs_info
->volume_mutex
);
2734 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
2737 if (vol_args
->flags
& BTRFS_DEVICE_SPEC_BY_ID
)
2738 btrfs_info(fs_info
, "device deleted: id %llu",
2741 btrfs_info(fs_info
, "device deleted: %s",
2747 mnt_drop_write_file(file
);
2751 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2753 struct inode
*inode
= file_inode(file
);
2754 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
2755 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2756 struct btrfs_ioctl_vol_args
*vol_args
;
2759 if (!capable(CAP_SYS_ADMIN
))
2762 ret
= mnt_want_write_file(file
);
2766 if (atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
2767 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2768 goto out_drop_write
;
2771 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2772 if (IS_ERR(vol_args
)) {
2773 ret
= PTR_ERR(vol_args
);
2777 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2778 mutex_lock(&fs_info
->volume_mutex
);
2779 ret
= btrfs_rm_device(root
, vol_args
->name
, 0);
2780 mutex_unlock(&fs_info
->volume_mutex
);
2783 btrfs_info(fs_info
, "disk deleted %s", vol_args
->name
);
2786 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
2788 mnt_drop_write_file(file
);
2793 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2795 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
2796 struct btrfs_ioctl_fs_info_args
*fi_args
;
2797 struct btrfs_device
*device
;
2798 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
2801 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2805 mutex_lock(&fs_devices
->device_list_mutex
);
2806 fi_args
->num_devices
= fs_devices
->num_devices
;
2807 memcpy(&fi_args
->fsid
, fs_info
->fsid
, sizeof(fi_args
->fsid
));
2809 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
2810 if (device
->devid
> fi_args
->max_id
)
2811 fi_args
->max_id
= device
->devid
;
2813 mutex_unlock(&fs_devices
->device_list_mutex
);
2815 fi_args
->nodesize
= fs_info
->super_copy
->nodesize
;
2816 fi_args
->sectorsize
= fs_info
->super_copy
->sectorsize
;
2817 fi_args
->clone_alignment
= fs_info
->super_copy
->sectorsize
;
2819 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2826 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2828 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
2829 struct btrfs_ioctl_dev_info_args
*di_args
;
2830 struct btrfs_device
*dev
;
2831 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
2833 char *s_uuid
= NULL
;
2835 di_args
= memdup_user(arg
, sizeof(*di_args
));
2836 if (IS_ERR(di_args
))
2837 return PTR_ERR(di_args
);
2839 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2840 s_uuid
= di_args
->uuid
;
2842 mutex_lock(&fs_devices
->device_list_mutex
);
2843 dev
= btrfs_find_device(fs_info
, di_args
->devid
, s_uuid
, NULL
);
2850 di_args
->devid
= dev
->devid
;
2851 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
2852 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
2853 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2855 struct rcu_string
*name
;
2858 name
= rcu_dereference(dev
->name
);
2859 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2861 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2863 di_args
->path
[0] = '\0';
2867 mutex_unlock(&fs_devices
->device_list_mutex
);
2868 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2875 static struct page
*extent_same_get_page(struct inode
*inode
, pgoff_t index
)
2879 page
= grab_cache_page(inode
->i_mapping
, index
);
2881 return ERR_PTR(-ENOMEM
);
2883 if (!PageUptodate(page
)) {
2886 ret
= btrfs_readpage(NULL
, page
);
2888 return ERR_PTR(ret
);
2890 if (!PageUptodate(page
)) {
2893 return ERR_PTR(-EIO
);
2895 if (page
->mapping
!= inode
->i_mapping
) {
2898 return ERR_PTR(-EAGAIN
);
2905 static int gather_extent_pages(struct inode
*inode
, struct page
**pages
,
2906 int num_pages
, u64 off
)
2909 pgoff_t index
= off
>> PAGE_SHIFT
;
2911 for (i
= 0; i
< num_pages
; i
++) {
2913 pages
[i
] = extent_same_get_page(inode
, index
+ i
);
2914 if (IS_ERR(pages
[i
])) {
2915 int err
= PTR_ERR(pages
[i
]);
2926 static int lock_extent_range(struct inode
*inode
, u64 off
, u64 len
,
2927 bool retry_range_locking
)
2930 * Do any pending delalloc/csum calculations on inode, one way or
2931 * another, and lock file content.
2932 * The locking order is:
2935 * 2) range in the inode's io tree
2938 struct btrfs_ordered_extent
*ordered
;
2939 lock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2940 ordered
= btrfs_lookup_first_ordered_extent(inode
,
2943 ordered
->file_offset
+ ordered
->len
<= off
||
2944 ordered
->file_offset
>= off
+ len
) &&
2945 !test_range_bit(&BTRFS_I(inode
)->io_tree
, off
,
2946 off
+ len
- 1, EXTENT_DELALLOC
, 0, NULL
)) {
2948 btrfs_put_ordered_extent(ordered
);
2951 unlock_extent(&BTRFS_I(inode
)->io_tree
, off
, off
+ len
- 1);
2953 btrfs_put_ordered_extent(ordered
);
2954 if (!retry_range_locking
)
2956 btrfs_wait_ordered_range(inode
, off
, len
);
2961 static void btrfs_double_inode_unlock(struct inode
*inode1
, struct inode
*inode2
)
2963 inode_unlock(inode1
);
2964 inode_unlock(inode2
);
2967 static void btrfs_double_inode_lock(struct inode
*inode1
, struct inode
*inode2
)
2969 if (inode1
< inode2
)
2970 swap(inode1
, inode2
);
2972 inode_lock_nested(inode1
, I_MUTEX_PARENT
);
2973 inode_lock_nested(inode2
, I_MUTEX_CHILD
);
2976 static void btrfs_double_extent_unlock(struct inode
*inode1
, u64 loff1
,
2977 struct inode
*inode2
, u64 loff2
, u64 len
)
2979 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
2980 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
2983 static int btrfs_double_extent_lock(struct inode
*inode1
, u64 loff1
,
2984 struct inode
*inode2
, u64 loff2
, u64 len
,
2985 bool retry_range_locking
)
2989 if (inode1
< inode2
) {
2990 swap(inode1
, inode2
);
2993 ret
= lock_extent_range(inode1
, loff1
, len
, retry_range_locking
);
2996 ret
= lock_extent_range(inode2
, loff2
, len
, retry_range_locking
);
2998 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
,
3005 struct page
**src_pages
;
3006 struct page
**dst_pages
;
3009 static void btrfs_cmp_data_free(struct cmp_pages
*cmp
)
3014 for (i
= 0; i
< cmp
->num_pages
; i
++) {
3015 pg
= cmp
->src_pages
[i
];
3020 pg
= cmp
->dst_pages
[i
];
3026 kfree(cmp
->src_pages
);
3027 kfree(cmp
->dst_pages
);
3030 static int btrfs_cmp_data_prepare(struct inode
*src
, u64 loff
,
3031 struct inode
*dst
, u64 dst_loff
,
3032 u64 len
, struct cmp_pages
*cmp
)
3035 int num_pages
= PAGE_ALIGN(len
) >> PAGE_SHIFT
;
3036 struct page
**src_pgarr
, **dst_pgarr
;
3039 * We must gather up all the pages before we initiate our
3040 * extent locking. We use an array for the page pointers. Size
3041 * of the array is bounded by len, which is in turn bounded by
3042 * BTRFS_MAX_DEDUPE_LEN.
3044 src_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
3045 dst_pgarr
= kcalloc(num_pages
, sizeof(struct page
*), GFP_KERNEL
);
3046 if (!src_pgarr
|| !dst_pgarr
) {
3051 cmp
->num_pages
= num_pages
;
3052 cmp
->src_pages
= src_pgarr
;
3053 cmp
->dst_pages
= dst_pgarr
;
3055 ret
= gather_extent_pages(src
, cmp
->src_pages
, cmp
->num_pages
, loff
);
3059 ret
= gather_extent_pages(dst
, cmp
->dst_pages
, cmp
->num_pages
, dst_loff
);
3063 btrfs_cmp_data_free(cmp
);
3067 static int btrfs_cmp_data(struct inode
*src
, u64 loff
, struct inode
*dst
,
3068 u64 dst_loff
, u64 len
, struct cmp_pages
*cmp
)
3072 struct page
*src_page
, *dst_page
;
3073 unsigned int cmp_len
= PAGE_SIZE
;
3074 void *addr
, *dst_addr
;
3078 if (len
< PAGE_SIZE
)
3081 BUG_ON(i
>= cmp
->num_pages
);
3083 src_page
= cmp
->src_pages
[i
];
3084 dst_page
= cmp
->dst_pages
[i
];
3085 ASSERT(PageLocked(src_page
));
3086 ASSERT(PageLocked(dst_page
));
3088 addr
= kmap_atomic(src_page
);
3089 dst_addr
= kmap_atomic(dst_page
);
3091 flush_dcache_page(src_page
);
3092 flush_dcache_page(dst_page
);
3094 if (memcmp(addr
, dst_addr
, cmp_len
))
3097 kunmap_atomic(addr
);
3098 kunmap_atomic(dst_addr
);
3110 static int extent_same_check_offsets(struct inode
*inode
, u64 off
, u64
*plen
,
3114 u64 bs
= BTRFS_I(inode
)->root
->fs_info
->sb
->s_blocksize
;
3116 if (off
+ olen
> inode
->i_size
|| off
+ olen
< off
)
3119 /* if we extend to eof, continue to block boundary */
3120 if (off
+ len
== inode
->i_size
)
3121 *plen
= len
= ALIGN(inode
->i_size
, bs
) - off
;
3123 /* Check that we are block aligned - btrfs_clone() requires this */
3124 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
))
3130 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 olen
,
3131 struct inode
*dst
, u64 dst_loff
)
3135 struct cmp_pages cmp
;
3137 u64 same_lock_start
= 0;
3138 u64 same_lock_len
= 0;
3149 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3152 ret
= extent_same_check_offsets(src
, dst_loff
, &len
, olen
);
3157 * Single inode case wants the same checks, except we
3158 * don't want our length pushed out past i_size as
3159 * comparing that data range makes no sense.
3161 * extent_same_check_offsets() will do this for an
3162 * unaligned length at i_size, so catch it here and
3163 * reject the request.
3165 * This effectively means we require aligned extents
3166 * for the single-inode case, whereas the other cases
3167 * allow an unaligned length so long as it ends at
3175 /* Check for overlapping ranges */
3176 if (dst_loff
+ len
> loff
&& dst_loff
< loff
+ len
) {
3181 same_lock_start
= min_t(u64
, loff
, dst_loff
);
3182 same_lock_len
= max_t(u64
, loff
, dst_loff
) + len
- same_lock_start
;
3184 btrfs_double_inode_lock(src
, dst
);
3186 ret
= extent_same_check_offsets(src
, loff
, &len
, olen
);
3190 ret
= extent_same_check_offsets(dst
, dst_loff
, &len
, olen
);
3195 /* don't make the dst file partly checksummed */
3196 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3197 (BTRFS_I(dst
)->flags
& BTRFS_INODE_NODATASUM
)) {
3203 ret
= btrfs_cmp_data_prepare(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3208 ret
= lock_extent_range(src
, same_lock_start
, same_lock_len
,
3211 ret
= btrfs_double_extent_lock(src
, loff
, dst
, dst_loff
, len
,
3214 * If one of the inodes has dirty pages in the respective range or
3215 * ordered extents, we need to flush dellaloc and wait for all ordered
3216 * extents in the range. We must unlock the pages and the ranges in the
3217 * io trees to avoid deadlocks when flushing delalloc (requires locking
3218 * pages) and when waiting for ordered extents to complete (they require
3221 if (ret
== -EAGAIN
) {
3223 * Ranges in the io trees already unlocked. Now unlock all
3224 * pages before waiting for all IO to complete.
3226 btrfs_cmp_data_free(&cmp
);
3228 btrfs_wait_ordered_range(src
, same_lock_start
,
3231 btrfs_wait_ordered_range(src
, loff
, len
);
3232 btrfs_wait_ordered_range(dst
, dst_loff
, len
);
3238 /* ranges in the io trees already unlocked */
3239 btrfs_cmp_data_free(&cmp
);
3243 /* pass original length for comparison so we stay within i_size */
3244 ret
= btrfs_cmp_data(src
, loff
, dst
, dst_loff
, olen
, &cmp
);
3246 ret
= btrfs_clone(src
, dst
, loff
, olen
, len
, dst_loff
, 1);
3249 unlock_extent(&BTRFS_I(src
)->io_tree
, same_lock_start
,
3250 same_lock_start
+ same_lock_len
- 1);
3252 btrfs_double_extent_unlock(src
, loff
, dst
, dst_loff
, len
);
3254 btrfs_cmp_data_free(&cmp
);
3259 btrfs_double_inode_unlock(src
, dst
);
3264 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3266 ssize_t
btrfs_dedupe_file_range(struct file
*src_file
, u64 loff
, u64 olen
,
3267 struct file
*dst_file
, u64 dst_loff
)
3269 struct inode
*src
= file_inode(src_file
);
3270 struct inode
*dst
= file_inode(dst_file
);
3271 u64 bs
= BTRFS_I(src
)->root
->fs_info
->sb
->s_blocksize
;
3274 if (olen
> BTRFS_MAX_DEDUPE_LEN
)
3275 olen
= BTRFS_MAX_DEDUPE_LEN
;
3277 if (WARN_ON_ONCE(bs
< PAGE_SIZE
)) {
3279 * Btrfs does not support blocksize < page_size. As a
3280 * result, btrfs_cmp_data() won't correctly handle
3281 * this situation without an update.
3286 res
= btrfs_extent_same(src
, loff
, olen
, dst
, dst_loff
);
3292 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3293 struct inode
*inode
,
3299 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3302 inode_inc_iversion(inode
);
3303 if (!no_time_update
)
3304 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
3306 * We round up to the block size at eof when determining which
3307 * extents to clone above, but shouldn't round up the file size.
3309 if (endoff
> destoff
+ olen
)
3310 endoff
= destoff
+ olen
;
3311 if (endoff
> inode
->i_size
)
3312 btrfs_i_size_write(inode
, endoff
);
3314 ret
= btrfs_update_inode(trans
, root
, inode
);
3316 btrfs_abort_transaction(trans
, ret
);
3317 btrfs_end_transaction(trans
, root
);
3320 ret
= btrfs_end_transaction(trans
, root
);
3325 static void clone_update_extent_map(struct inode
*inode
,
3326 const struct btrfs_trans_handle
*trans
,
3327 const struct btrfs_path
*path
,
3328 const u64 hole_offset
,
3331 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
3332 struct extent_map
*em
;
3335 em
= alloc_extent_map();
3337 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3338 &BTRFS_I(inode
)->runtime_flags
);
3343 struct btrfs_file_extent_item
*fi
;
3345 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3346 struct btrfs_file_extent_item
);
3347 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3348 em
->generation
= -1;
3349 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3350 BTRFS_FILE_EXTENT_INLINE
)
3351 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3352 &BTRFS_I(inode
)->runtime_flags
);
3354 em
->start
= hole_offset
;
3356 em
->ram_bytes
= em
->len
;
3357 em
->orig_start
= hole_offset
;
3358 em
->block_start
= EXTENT_MAP_HOLE
;
3360 em
->orig_block_len
= 0;
3361 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3362 em
->generation
= trans
->transid
;
3366 write_lock(&em_tree
->lock
);
3367 ret
= add_extent_mapping(em_tree
, em
, 1);
3368 write_unlock(&em_tree
->lock
);
3369 if (ret
!= -EEXIST
) {
3370 free_extent_map(em
);
3373 btrfs_drop_extent_cache(inode
, em
->start
,
3374 em
->start
+ em
->len
- 1, 0);
3378 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3379 &BTRFS_I(inode
)->runtime_flags
);
3383 * Make sure we do not end up inserting an inline extent into a file that has
3384 * already other (non-inline) extents. If a file has an inline extent it can
3385 * not have any other extents and the (single) inline extent must start at the
3386 * file offset 0. Failing to respect these rules will lead to file corruption,
3387 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3389 * We can have extents that have been already written to disk or we can have
3390 * dirty ranges still in delalloc, in which case the extent maps and items are
3391 * created only when we run delalloc, and the delalloc ranges might fall outside
3392 * the range we are currently locking in the inode's io tree. So we check the
3393 * inode's i_size because of that (i_size updates are done while holding the
3394 * i_mutex, which we are holding here).
3395 * We also check to see if the inode has a size not greater than "datal" but has
3396 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3397 * protected against such concurrent fallocate calls by the i_mutex).
3399 * If the file has no extents but a size greater than datal, do not allow the
3400 * copy because we would need turn the inline extent into a non-inline one (even
3401 * with NO_HOLES enabled). If we find our destination inode only has one inline
3402 * extent, just overwrite it with the source inline extent if its size is less
3403 * than the source extent's size, or we could copy the source inline extent's
3404 * data into the destination inode's inline extent if the later is greater then
3407 static int clone_copy_inline_extent(struct inode
*src
,
3409 struct btrfs_trans_handle
*trans
,
3410 struct btrfs_path
*path
,
3411 struct btrfs_key
*new_key
,
3412 const u64 drop_start
,
3418 struct btrfs_fs_info
*fs_info
= btrfs_sb(dst
->i_sb
);
3419 struct btrfs_root
*root
= BTRFS_I(dst
)->root
;
3420 const u64 aligned_end
= ALIGN(new_key
->offset
+ datal
,
3421 fs_info
->sectorsize
);
3423 struct btrfs_key key
;
3425 if (new_key
->offset
> 0)
3428 key
.objectid
= btrfs_ino(dst
);
3429 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3431 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
3434 } else if (ret
> 0) {
3435 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3436 ret
= btrfs_next_leaf(root
, path
);
3440 goto copy_inline_extent
;
3442 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
3443 if (key
.objectid
== btrfs_ino(dst
) &&
3444 key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3445 ASSERT(key
.offset
> 0);
3448 } else if (i_size_read(dst
) <= datal
) {
3449 struct btrfs_file_extent_item
*ei
;
3453 * If the file size is <= datal, make sure there are no other
3454 * extents following (can happen do to an fallocate call with
3455 * the flag FALLOC_FL_KEEP_SIZE).
3457 ei
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3458 struct btrfs_file_extent_item
);
3460 * If it's an inline extent, it can not have other extents
3463 if (btrfs_file_extent_type(path
->nodes
[0], ei
) ==
3464 BTRFS_FILE_EXTENT_INLINE
)
3465 goto copy_inline_extent
;
3467 ext_len
= btrfs_file_extent_num_bytes(path
->nodes
[0], ei
);
3468 if (ext_len
> aligned_end
)
3471 ret
= btrfs_next_item(root
, path
);
3474 } else if (ret
== 0) {
3475 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3477 if (key
.objectid
== btrfs_ino(dst
) &&
3478 key
.type
== BTRFS_EXTENT_DATA_KEY
)
3485 * We have no extent items, or we have an extent at offset 0 which may
3486 * or may not be inlined. All these cases are dealt the same way.
3488 if (i_size_read(dst
) > datal
) {
3490 * If the destination inode has an inline extent...
3491 * This would require copying the data from the source inline
3492 * extent into the beginning of the destination's inline extent.
3493 * But this is really complex, both extents can be compressed
3494 * or just one of them, which would require decompressing and
3495 * re-compressing data (which could increase the new compressed
3496 * size, not allowing the compressed data to fit anymore in an
3498 * So just don't support this case for now (it should be rare,
3499 * we are not really saving space when cloning inline extents).
3504 btrfs_release_path(path
);
3505 ret
= btrfs_drop_extents(trans
, root
, dst
, drop_start
, aligned_end
, 1);
3508 ret
= btrfs_insert_empty_item(trans
, root
, path
, new_key
, size
);
3513 const u32 start
= btrfs_file_extent_calc_inline_size(0);
3515 memmove(inline_data
+ start
, inline_data
+ start
+ skip
, datal
);
3518 write_extent_buffer(path
->nodes
[0], inline_data
,
3519 btrfs_item_ptr_offset(path
->nodes
[0],
3522 inode_add_bytes(dst
, datal
);
3528 * btrfs_clone() - clone a range from inode file to another
3530 * @src: Inode to clone from
3531 * @inode: Inode to clone to
3532 * @off: Offset within source to start clone from
3533 * @olen: Original length, passed by user, of range to clone
3534 * @olen_aligned: Block-aligned value of olen
3535 * @destoff: Offset within @inode to start clone
3536 * @no_time_update: Whether to update mtime/ctime on the target inode
3538 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3539 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3540 const u64 destoff
, int no_time_update
)
3542 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3543 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3544 struct btrfs_path
*path
= NULL
;
3545 struct extent_buffer
*leaf
;
3546 struct btrfs_trans_handle
*trans
;
3548 struct btrfs_key key
;
3552 const u64 len
= olen_aligned
;
3553 u64 last_dest_end
= destoff
;
3556 buf
= kmalloc(fs_info
->nodesize
, GFP_KERNEL
| __GFP_NOWARN
);
3558 buf
= vmalloc(fs_info
->nodesize
);
3563 path
= btrfs_alloc_path();
3569 path
->reada
= READA_FORWARD
;
3571 key
.objectid
= btrfs_ino(src
);
3572 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3576 u64 next_key_min_offset
= key
.offset
+ 1;
3579 * note the key will change type as we walk through the
3582 path
->leave_spinning
= 1;
3583 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3588 * First search, if no extent item that starts at offset off was
3589 * found but the previous item is an extent item, it's possible
3590 * it might overlap our target range, therefore process it.
3592 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3593 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3594 path
->slots
[0] - 1);
3595 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3599 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3601 if (path
->slots
[0] >= nritems
) {
3602 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3607 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3609 leaf
= path
->nodes
[0];
3610 slot
= path
->slots
[0];
3612 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3613 if (key
.type
> BTRFS_EXTENT_DATA_KEY
||
3614 key
.objectid
!= btrfs_ino(src
))
3617 if (key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3618 struct btrfs_file_extent_item
*extent
;
3621 struct btrfs_key new_key
;
3622 u64 disko
= 0, diskl
= 0;
3623 u64 datao
= 0, datal
= 0;
3627 extent
= btrfs_item_ptr(leaf
, slot
,
3628 struct btrfs_file_extent_item
);
3629 comp
= btrfs_file_extent_compression(leaf
, extent
);
3630 type
= btrfs_file_extent_type(leaf
, extent
);
3631 if (type
== BTRFS_FILE_EXTENT_REG
||
3632 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3633 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3635 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3637 datao
= btrfs_file_extent_offset(leaf
, extent
);
3638 datal
= btrfs_file_extent_num_bytes(leaf
,
3640 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3641 /* take upper bound, may be compressed */
3642 datal
= btrfs_file_extent_ram_bytes(leaf
,
3647 * The first search might have left us at an extent
3648 * item that ends before our target range's start, can
3649 * happen if we have holes and NO_HOLES feature enabled.
3651 if (key
.offset
+ datal
<= off
) {
3654 } else if (key
.offset
>= off
+ len
) {
3657 next_key_min_offset
= key
.offset
+ datal
;
3658 size
= btrfs_item_size_nr(leaf
, slot
);
3659 read_extent_buffer(leaf
, buf
,
3660 btrfs_item_ptr_offset(leaf
, slot
),
3663 btrfs_release_path(path
);
3664 path
->leave_spinning
= 0;
3666 memcpy(&new_key
, &key
, sizeof(new_key
));
3667 new_key
.objectid
= btrfs_ino(inode
);
3668 if (off
<= key
.offset
)
3669 new_key
.offset
= key
.offset
+ destoff
- off
;
3671 new_key
.offset
= destoff
;
3674 * Deal with a hole that doesn't have an extent item
3675 * that represents it (NO_HOLES feature enabled).
3676 * This hole is either in the middle of the cloning
3677 * range or at the beginning (fully overlaps it or
3678 * partially overlaps it).
3680 if (new_key
.offset
!= last_dest_end
)
3681 drop_start
= last_dest_end
;
3683 drop_start
= new_key
.offset
;
3686 * 1 - adjusting old extent (we may have to split it)
3687 * 1 - add new extent
3690 trans
= btrfs_start_transaction(root
, 3);
3691 if (IS_ERR(trans
)) {
3692 ret
= PTR_ERR(trans
);
3696 if (type
== BTRFS_FILE_EXTENT_REG
||
3697 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3699 * a | --- range to clone ---| b
3700 * | ------------- extent ------------- |
3703 /* subtract range b */
3704 if (key
.offset
+ datal
> off
+ len
)
3705 datal
= off
+ len
- key
.offset
;
3707 /* subtract range a */
3708 if (off
> key
.offset
) {
3709 datao
+= off
- key
.offset
;
3710 datal
-= off
- key
.offset
;
3713 ret
= btrfs_drop_extents(trans
, root
, inode
,
3715 new_key
.offset
+ datal
,
3718 if (ret
!= -EOPNOTSUPP
)
3719 btrfs_abort_transaction(trans
,
3721 btrfs_end_transaction(trans
, root
);
3725 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3728 btrfs_abort_transaction(trans
, ret
);
3729 btrfs_end_transaction(trans
, root
);
3733 leaf
= path
->nodes
[0];
3734 slot
= path
->slots
[0];
3735 write_extent_buffer(leaf
, buf
,
3736 btrfs_item_ptr_offset(leaf
, slot
),
3739 extent
= btrfs_item_ptr(leaf
, slot
,
3740 struct btrfs_file_extent_item
);
3742 /* disko == 0 means it's a hole */
3746 btrfs_set_file_extent_offset(leaf
, extent
,
3748 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3752 inode_add_bytes(inode
, datal
);
3753 ret
= btrfs_inc_extent_ref(trans
, root
,
3755 root
->root_key
.objectid
,
3757 new_key
.offset
- datao
);
3759 btrfs_abort_transaction(trans
,
3761 btrfs_end_transaction(trans
,
3767 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3771 if (off
> key
.offset
) {
3772 skip
= off
- key
.offset
;
3773 new_key
.offset
+= skip
;
3776 if (key
.offset
+ datal
> off
+ len
)
3777 trim
= key
.offset
+ datal
- (off
+ len
);
3779 if (comp
&& (skip
|| trim
)) {
3781 btrfs_end_transaction(trans
, root
);
3784 size
-= skip
+ trim
;
3785 datal
-= skip
+ trim
;
3787 ret
= clone_copy_inline_extent(src
, inode
,
3794 if (ret
!= -EOPNOTSUPP
)
3795 btrfs_abort_transaction(trans
,
3797 btrfs_end_transaction(trans
, root
);
3800 leaf
= path
->nodes
[0];
3801 slot
= path
->slots
[0];
3804 /* If we have an implicit hole (NO_HOLES feature). */
3805 if (drop_start
< new_key
.offset
)
3806 clone_update_extent_map(inode
, trans
,
3808 new_key
.offset
- drop_start
);
3810 clone_update_extent_map(inode
, trans
, path
, 0, 0);
3812 btrfs_mark_buffer_dirty(leaf
);
3813 btrfs_release_path(path
);
3815 last_dest_end
= ALIGN(new_key
.offset
+ datal
,
3816 fs_info
->sectorsize
);
3817 ret
= clone_finish_inode_update(trans
, inode
,
3823 if (new_key
.offset
+ datal
>= destoff
+ len
)
3826 btrfs_release_path(path
);
3827 key
.offset
= next_key_min_offset
;
3829 if (fatal_signal_pending(current
)) {
3836 if (last_dest_end
< destoff
+ len
) {
3838 * We have an implicit hole (NO_HOLES feature is enabled) that
3839 * fully or partially overlaps our cloning range at its end.
3841 btrfs_release_path(path
);
3844 * 1 - remove extent(s)
3847 trans
= btrfs_start_transaction(root
, 2);
3848 if (IS_ERR(trans
)) {
3849 ret
= PTR_ERR(trans
);
3852 ret
= btrfs_drop_extents(trans
, root
, inode
,
3853 last_dest_end
, destoff
+ len
, 1);
3855 if (ret
!= -EOPNOTSUPP
)
3856 btrfs_abort_transaction(trans
, ret
);
3857 btrfs_end_transaction(trans
, root
);
3860 clone_update_extent_map(inode
, trans
, NULL
, last_dest_end
,
3861 destoff
+ len
- last_dest_end
);
3862 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3863 destoff
, olen
, no_time_update
);
3867 btrfs_free_path(path
);
3872 static noinline
int btrfs_clone_files(struct file
*file
, struct file
*file_src
,
3873 u64 off
, u64 olen
, u64 destoff
)
3875 struct inode
*inode
= file_inode(file
);
3876 struct inode
*src
= file_inode(file_src
);
3877 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3878 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3881 u64 bs
= fs_info
->sb
->s_blocksize
;
3882 int same_inode
= src
== inode
;
3886 * - split compressed inline extents. annoying: we need to
3887 * decompress into destination's address_space (the file offset
3888 * may change, so source mapping won't do), then recompress (or
3889 * otherwise reinsert) a subrange.
3891 * - split destination inode's inline extents. The inline extents can
3892 * be either compressed or non-compressed.
3895 if (btrfs_root_readonly(root
))
3898 if (file_src
->f_path
.mnt
!= file
->f_path
.mnt
||
3899 src
->i_sb
!= inode
->i_sb
)
3902 /* don't make the dst file partly checksummed */
3903 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
3904 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
3907 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
3911 btrfs_double_inode_lock(src
, inode
);
3916 /* determine range to clone */
3918 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
3921 olen
= len
= src
->i_size
- off
;
3922 /* if we extend to eof, continue to block boundary */
3923 if (off
+ len
== src
->i_size
)
3924 len
= ALIGN(src
->i_size
, bs
) - off
;
3931 /* verify the end result is block aligned */
3932 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
3933 !IS_ALIGNED(destoff
, bs
))
3936 /* verify if ranges are overlapped within the same file */
3938 if (destoff
+ len
> off
&& destoff
< off
+ len
)
3942 if (destoff
> inode
->i_size
) {
3943 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3949 * Lock the target range too. Right after we replace the file extent
3950 * items in the fs tree (which now point to the cloned data), we might
3951 * have a worker replace them with extent items relative to a write
3952 * operation that was issued before this clone operation (i.e. confront
3953 * with inode.c:btrfs_finish_ordered_io).
3956 u64 lock_start
= min_t(u64
, off
, destoff
);
3957 u64 lock_len
= max_t(u64
, off
, destoff
) + len
- lock_start
;
3959 ret
= lock_extent_range(src
, lock_start
, lock_len
, true);
3961 ret
= btrfs_double_extent_lock(src
, off
, inode
, destoff
, len
,
3966 /* ranges in the io trees already unlocked */
3970 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
, 0);
3973 u64 lock_start
= min_t(u64
, off
, destoff
);
3974 u64 lock_end
= max_t(u64
, off
, destoff
) + len
- 1;
3976 unlock_extent(&BTRFS_I(src
)->io_tree
, lock_start
, lock_end
);
3978 btrfs_double_extent_unlock(src
, off
, inode
, destoff
, len
);
3981 * Truncate page cache pages so that future reads will see the cloned
3982 * data immediately and not the previous data.
3984 truncate_inode_pages_range(&inode
->i_data
,
3985 round_down(destoff
, PAGE_SIZE
),
3986 round_up(destoff
+ len
, PAGE_SIZE
) - 1);
3989 btrfs_double_inode_unlock(src
, inode
);
3995 ssize_t
btrfs_copy_file_range(struct file
*file_in
, loff_t pos_in
,
3996 struct file
*file_out
, loff_t pos_out
,
3997 size_t len
, unsigned int flags
)
4001 ret
= btrfs_clone_files(file_out
, file_in
, pos_in
, len
, pos_out
);
4007 int btrfs_clone_file_range(struct file
*src_file
, loff_t off
,
4008 struct file
*dst_file
, loff_t destoff
, u64 len
)
4010 return btrfs_clone_files(dst_file
, src_file
, off
, len
, destoff
);
4014 * there are many ways the trans_start and trans_end ioctls can lead
4015 * to deadlocks. They should only be used by applications that
4016 * basically own the machine, and have a very in depth understanding
4017 * of all the possible deadlocks and enospc problems.
4019 static long btrfs_ioctl_trans_start(struct file
*file
)
4021 struct inode
*inode
= file_inode(file
);
4022 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4023 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4024 struct btrfs_trans_handle
*trans
;
4028 if (!capable(CAP_SYS_ADMIN
))
4032 if (file
->private_data
)
4036 if (btrfs_root_readonly(root
))
4039 ret
= mnt_want_write_file(file
);
4043 atomic_inc(&fs_info
->open_ioctl_trans
);
4046 trans
= btrfs_start_ioctl_transaction(root
);
4050 file
->private_data
= trans
;
4054 atomic_dec(&fs_info
->open_ioctl_trans
);
4055 mnt_drop_write_file(file
);
4060 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
4062 struct inode
*inode
= file_inode(file
);
4063 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4064 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4065 struct btrfs_root
*new_root
;
4066 struct btrfs_dir_item
*di
;
4067 struct btrfs_trans_handle
*trans
;
4068 struct btrfs_path
*path
;
4069 struct btrfs_key location
;
4070 struct btrfs_disk_key disk_key
;
4075 if (!capable(CAP_SYS_ADMIN
))
4078 ret
= mnt_want_write_file(file
);
4082 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
4088 objectid
= BTRFS_FS_TREE_OBJECTID
;
4090 location
.objectid
= objectid
;
4091 location
.type
= BTRFS_ROOT_ITEM_KEY
;
4092 location
.offset
= (u64
)-1;
4094 new_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
4095 if (IS_ERR(new_root
)) {
4096 ret
= PTR_ERR(new_root
);
4100 path
= btrfs_alloc_path();
4105 path
->leave_spinning
= 1;
4107 trans
= btrfs_start_transaction(root
, 1);
4108 if (IS_ERR(trans
)) {
4109 btrfs_free_path(path
);
4110 ret
= PTR_ERR(trans
);
4114 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
4115 di
= btrfs_lookup_dir_item(trans
, fs_info
->tree_root
, path
,
4116 dir_id
, "default", 7, 1);
4117 if (IS_ERR_OR_NULL(di
)) {
4118 btrfs_free_path(path
);
4119 btrfs_end_transaction(trans
, root
);
4121 "Umm, you don't have the default diritem, this isn't going to work");
4126 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
4127 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
4128 btrfs_mark_buffer_dirty(path
->nodes
[0]);
4129 btrfs_free_path(path
);
4131 btrfs_set_fs_incompat(fs_info
, DEFAULT_SUBVOL
);
4132 btrfs_end_transaction(trans
, root
);
4134 mnt_drop_write_file(file
);
4138 void btrfs_get_block_group_info(struct list_head
*groups_list
,
4139 struct btrfs_ioctl_space_info
*space
)
4141 struct btrfs_block_group_cache
*block_group
;
4143 space
->total_bytes
= 0;
4144 space
->used_bytes
= 0;
4146 list_for_each_entry(block_group
, groups_list
, list
) {
4147 space
->flags
= block_group
->flags
;
4148 space
->total_bytes
+= block_group
->key
.offset
;
4149 space
->used_bytes
+=
4150 btrfs_block_group_used(&block_group
->item
);
4154 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
4156 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4157 struct btrfs_ioctl_space_args space_args
;
4158 struct btrfs_ioctl_space_info space
;
4159 struct btrfs_ioctl_space_info
*dest
;
4160 struct btrfs_ioctl_space_info
*dest_orig
;
4161 struct btrfs_ioctl_space_info __user
*user_dest
;
4162 struct btrfs_space_info
*info
;
4163 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
4164 BTRFS_BLOCK_GROUP_SYSTEM
,
4165 BTRFS_BLOCK_GROUP_METADATA
,
4166 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
4173 if (copy_from_user(&space_args
,
4174 (struct btrfs_ioctl_space_args __user
*)arg
,
4175 sizeof(space_args
)))
4178 for (i
= 0; i
< num_types
; i
++) {
4179 struct btrfs_space_info
*tmp
;
4183 list_for_each_entry_rcu(tmp
, &fs_info
->space_info
,
4185 if (tmp
->flags
== types
[i
]) {
4195 down_read(&info
->groups_sem
);
4196 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4197 if (!list_empty(&info
->block_groups
[c
]))
4200 up_read(&info
->groups_sem
);
4204 * Global block reserve, exported as a space_info
4208 /* space_slots == 0 means they are asking for a count */
4209 if (space_args
.space_slots
== 0) {
4210 space_args
.total_spaces
= slot_count
;
4214 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
4216 alloc_size
= sizeof(*dest
) * slot_count
;
4218 /* we generally have at most 6 or so space infos, one for each raid
4219 * level. So, a whole page should be more than enough for everyone
4221 if (alloc_size
> PAGE_SIZE
)
4224 space_args
.total_spaces
= 0;
4225 dest
= kmalloc(alloc_size
, GFP_KERNEL
);
4230 /* now we have a buffer to copy into */
4231 for (i
= 0; i
< num_types
; i
++) {
4232 struct btrfs_space_info
*tmp
;
4239 list_for_each_entry_rcu(tmp
, &fs_info
->space_info
,
4241 if (tmp
->flags
== types
[i
]) {
4250 down_read(&info
->groups_sem
);
4251 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4252 if (!list_empty(&info
->block_groups
[c
])) {
4253 btrfs_get_block_group_info(
4254 &info
->block_groups
[c
], &space
);
4255 memcpy(dest
, &space
, sizeof(space
));
4257 space_args
.total_spaces
++;
4263 up_read(&info
->groups_sem
);
4267 * Add global block reserve
4270 struct btrfs_block_rsv
*block_rsv
= &fs_info
->global_block_rsv
;
4272 spin_lock(&block_rsv
->lock
);
4273 space
.total_bytes
= block_rsv
->size
;
4274 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4275 spin_unlock(&block_rsv
->lock
);
4276 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4277 memcpy(dest
, &space
, sizeof(space
));
4278 space_args
.total_spaces
++;
4281 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4282 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4284 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4289 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4296 * there are many ways the trans_start and trans_end ioctls can lead
4297 * to deadlocks. They should only be used by applications that
4298 * basically own the machine, and have a very in depth understanding
4299 * of all the possible deadlocks and enospc problems.
4301 long btrfs_ioctl_trans_end(struct file
*file
)
4303 struct inode
*inode
= file_inode(file
);
4304 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4305 struct btrfs_trans_handle
*trans
;
4307 trans
= file
->private_data
;
4310 file
->private_data
= NULL
;
4312 btrfs_end_transaction(trans
, root
);
4314 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
4316 mnt_drop_write_file(file
);
4320 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4323 struct btrfs_trans_handle
*trans
;
4327 trans
= btrfs_attach_transaction_barrier(root
);
4328 if (IS_ERR(trans
)) {
4329 if (PTR_ERR(trans
) != -ENOENT
)
4330 return PTR_ERR(trans
);
4332 /* No running transaction, don't bother */
4333 transid
= root
->fs_info
->last_trans_committed
;
4336 transid
= trans
->transid
;
4337 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
4339 btrfs_end_transaction(trans
, root
);
4344 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4349 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
4355 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4358 transid
= 0; /* current trans */
4360 return btrfs_wait_for_commit(root
, transid
);
4363 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4365 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
4366 struct btrfs_ioctl_scrub_args
*sa
;
4369 if (!capable(CAP_SYS_ADMIN
))
4372 sa
= memdup_user(arg
, sizeof(*sa
));
4376 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4377 ret
= mnt_want_write_file(file
);
4382 ret
= btrfs_scrub_dev(fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4383 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4386 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4389 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4390 mnt_drop_write_file(file
);
4396 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
4398 if (!capable(CAP_SYS_ADMIN
))
4401 return btrfs_scrub_cancel(root
->fs_info
);
4404 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
4407 struct btrfs_ioctl_scrub_args
*sa
;
4410 if (!capable(CAP_SYS_ADMIN
))
4413 sa
= memdup_user(arg
, sizeof(*sa
));
4417 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
4419 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4426 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
4429 struct btrfs_ioctl_get_dev_stats
*sa
;
4432 sa
= memdup_user(arg
, sizeof(*sa
));
4436 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4441 ret
= btrfs_get_dev_stats(root
, sa
);
4443 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
4450 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
4452 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4453 struct btrfs_ioctl_dev_replace_args
*p
;
4456 if (!capable(CAP_SYS_ADMIN
))
4459 p
= memdup_user(arg
, sizeof(*p
));
4464 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4465 if (fs_info
->sb
->s_flags
& MS_RDONLY
) {
4470 &fs_info
->mutually_exclusive_operation_running
, 1)) {
4471 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4473 ret
= btrfs_dev_replace_by_ioctl(root
, p
);
4475 &fs_info
->mutually_exclusive_operation_running
, 0);
4478 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4479 btrfs_dev_replace_status(fs_info
, p
);
4482 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4483 ret
= btrfs_dev_replace_cancel(fs_info
, p
);
4490 if (copy_to_user(arg
, p
, sizeof(*p
)))
4497 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4503 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4504 struct inode_fs_paths
*ipath
= NULL
;
4505 struct btrfs_path
*path
;
4507 if (!capable(CAP_DAC_READ_SEARCH
))
4510 path
= btrfs_alloc_path();
4516 ipa
= memdup_user(arg
, sizeof(*ipa
));
4523 size
= min_t(u32
, ipa
->size
, 4096);
4524 ipath
= init_ipath(size
, root
, path
);
4525 if (IS_ERR(ipath
)) {
4526 ret
= PTR_ERR(ipath
);
4531 ret
= paths_from_inode(ipa
->inum
, ipath
);
4535 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4536 rel_ptr
= ipath
->fspath
->val
[i
] -
4537 (u64
)(unsigned long)ipath
->fspath
->val
;
4538 ipath
->fspath
->val
[i
] = rel_ptr
;
4541 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
4542 (void *)(unsigned long)ipath
->fspath
, size
);
4549 btrfs_free_path(path
);
4556 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4558 struct btrfs_data_container
*inodes
= ctx
;
4559 const size_t c
= 3 * sizeof(u64
);
4561 if (inodes
->bytes_left
>= c
) {
4562 inodes
->bytes_left
-= c
;
4563 inodes
->val
[inodes
->elem_cnt
] = inum
;
4564 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4565 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4566 inodes
->elem_cnt
+= 3;
4568 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4569 inodes
->bytes_left
= 0;
4570 inodes
->elem_missed
+= 3;
4576 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
4581 struct btrfs_ioctl_logical_ino_args
*loi
;
4582 struct btrfs_data_container
*inodes
= NULL
;
4583 struct btrfs_path
*path
= NULL
;
4585 if (!capable(CAP_SYS_ADMIN
))
4588 loi
= memdup_user(arg
, sizeof(*loi
));
4590 return PTR_ERR(loi
);
4592 path
= btrfs_alloc_path();
4598 size
= min_t(u32
, loi
->size
, SZ_64K
);
4599 inodes
= init_data_container(size
);
4600 if (IS_ERR(inodes
)) {
4601 ret
= PTR_ERR(inodes
);
4606 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
4607 build_ino_list
, inodes
);
4613 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
4614 (void *)(unsigned long)inodes
, size
);
4619 btrfs_free_path(path
);
4626 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
4627 struct btrfs_ioctl_balance_args
*bargs
)
4629 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4631 bargs
->flags
= bctl
->flags
;
4633 if (atomic_read(&fs_info
->balance_running
))
4634 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4635 if (atomic_read(&fs_info
->balance_pause_req
))
4636 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4637 if (atomic_read(&fs_info
->balance_cancel_req
))
4638 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4640 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4641 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4642 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4645 spin_lock(&fs_info
->balance_lock
);
4646 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4647 spin_unlock(&fs_info
->balance_lock
);
4649 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4653 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4655 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4656 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4657 struct btrfs_ioctl_balance_args
*bargs
;
4658 struct btrfs_balance_control
*bctl
;
4659 bool need_unlock
; /* for mut. excl. ops lock */
4662 if (!capable(CAP_SYS_ADMIN
))
4665 ret
= mnt_want_write_file(file
);
4670 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
4671 mutex_lock(&fs_info
->volume_mutex
);
4672 mutex_lock(&fs_info
->balance_mutex
);
4678 * mut. excl. ops lock is locked. Three possibilities:
4679 * (1) some other op is running
4680 * (2) balance is running
4681 * (3) balance is paused -- special case (think resume)
4683 mutex_lock(&fs_info
->balance_mutex
);
4684 if (fs_info
->balance_ctl
) {
4685 /* this is either (2) or (3) */
4686 if (!atomic_read(&fs_info
->balance_running
)) {
4687 mutex_unlock(&fs_info
->balance_mutex
);
4688 if (!mutex_trylock(&fs_info
->volume_mutex
))
4690 mutex_lock(&fs_info
->balance_mutex
);
4692 if (fs_info
->balance_ctl
&&
4693 !atomic_read(&fs_info
->balance_running
)) {
4695 need_unlock
= false;
4699 mutex_unlock(&fs_info
->balance_mutex
);
4700 mutex_unlock(&fs_info
->volume_mutex
);
4704 mutex_unlock(&fs_info
->balance_mutex
);
4710 mutex_unlock(&fs_info
->balance_mutex
);
4711 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4716 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
4719 bargs
= memdup_user(arg
, sizeof(*bargs
));
4720 if (IS_ERR(bargs
)) {
4721 ret
= PTR_ERR(bargs
);
4725 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4726 if (!fs_info
->balance_ctl
) {
4731 bctl
= fs_info
->balance_ctl
;
4732 spin_lock(&fs_info
->balance_lock
);
4733 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4734 spin_unlock(&fs_info
->balance_lock
);
4742 if (fs_info
->balance_ctl
) {
4747 bctl
= kzalloc(sizeof(*bctl
), GFP_KERNEL
);
4753 bctl
->fs_info
= fs_info
;
4755 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4756 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4757 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4759 bctl
->flags
= bargs
->flags
;
4761 /* balance everything - no filters */
4762 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4765 if (bctl
->flags
& ~(BTRFS_BALANCE_ARGS_MASK
| BTRFS_BALANCE_TYPE_MASK
)) {
4772 * Ownership of bctl and mutually_exclusive_operation_running
4773 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4774 * or, if restriper was paused all the way until unmount, in
4775 * free_fs_info. mutually_exclusive_operation_running is
4776 * cleared in __cancel_balance.
4778 need_unlock
= false;
4780 ret
= btrfs_balance(bctl
, bargs
);
4784 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4793 mutex_unlock(&fs_info
->balance_mutex
);
4794 mutex_unlock(&fs_info
->volume_mutex
);
4796 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
4798 mnt_drop_write_file(file
);
4802 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
4804 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4806 if (!capable(CAP_SYS_ADMIN
))
4810 case BTRFS_BALANCE_CTL_PAUSE
:
4811 return btrfs_pause_balance(fs_info
);
4812 case BTRFS_BALANCE_CTL_CANCEL
:
4813 return btrfs_cancel_balance(fs_info
);
4819 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
4822 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4823 struct btrfs_ioctl_balance_args
*bargs
;
4826 if (!capable(CAP_SYS_ADMIN
))
4829 mutex_lock(&fs_info
->balance_mutex
);
4830 if (!fs_info
->balance_ctl
) {
4835 bargs
= kzalloc(sizeof(*bargs
), GFP_KERNEL
);
4841 update_ioctl_balance_args(fs_info
, 1, bargs
);
4843 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4848 mutex_unlock(&fs_info
->balance_mutex
);
4852 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4854 struct inode
*inode
= file_inode(file
);
4855 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4856 struct btrfs_ioctl_quota_ctl_args
*sa
;
4857 struct btrfs_trans_handle
*trans
= NULL
;
4861 if (!capable(CAP_SYS_ADMIN
))
4864 ret
= mnt_want_write_file(file
);
4868 sa
= memdup_user(arg
, sizeof(*sa
));
4874 down_write(&fs_info
->subvol_sem
);
4875 trans
= btrfs_start_transaction(fs_info
->tree_root
, 2);
4876 if (IS_ERR(trans
)) {
4877 ret
= PTR_ERR(trans
);
4882 case BTRFS_QUOTA_CTL_ENABLE
:
4883 ret
= btrfs_quota_enable(trans
, fs_info
);
4885 case BTRFS_QUOTA_CTL_DISABLE
:
4886 ret
= btrfs_quota_disable(trans
, fs_info
);
4893 err
= btrfs_commit_transaction(trans
, fs_info
->tree_root
);
4898 up_write(&fs_info
->subvol_sem
);
4900 mnt_drop_write_file(file
);
4904 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4906 struct inode
*inode
= file_inode(file
);
4907 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4908 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4909 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4910 struct btrfs_trans_handle
*trans
;
4914 if (!capable(CAP_SYS_ADMIN
))
4917 ret
= mnt_want_write_file(file
);
4921 sa
= memdup_user(arg
, sizeof(*sa
));
4927 trans
= btrfs_join_transaction(root
);
4928 if (IS_ERR(trans
)) {
4929 ret
= PTR_ERR(trans
);
4933 /* FIXME: check if the IDs really exist */
4935 ret
= btrfs_add_qgroup_relation(trans
, fs_info
,
4938 ret
= btrfs_del_qgroup_relation(trans
, fs_info
,
4942 /* update qgroup status and info */
4943 err
= btrfs_run_qgroups(trans
, fs_info
);
4945 btrfs_handle_fs_error(fs_info
, err
,
4946 "failed to update qgroup status and info");
4947 err
= btrfs_end_transaction(trans
, root
);
4954 mnt_drop_write_file(file
);
4958 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4960 struct inode
*inode
= file_inode(file
);
4961 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4962 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4963 struct btrfs_ioctl_qgroup_create_args
*sa
;
4964 struct btrfs_trans_handle
*trans
;
4968 if (!capable(CAP_SYS_ADMIN
))
4971 ret
= mnt_want_write_file(file
);
4975 sa
= memdup_user(arg
, sizeof(*sa
));
4981 if (!sa
->qgroupid
) {
4986 trans
= btrfs_join_transaction(root
);
4987 if (IS_ERR(trans
)) {
4988 ret
= PTR_ERR(trans
);
4992 /* FIXME: check if the IDs really exist */
4994 ret
= btrfs_create_qgroup(trans
, fs_info
, sa
->qgroupid
);
4996 ret
= btrfs_remove_qgroup(trans
, fs_info
, sa
->qgroupid
);
4999 err
= btrfs_end_transaction(trans
, root
);
5006 mnt_drop_write_file(file
);
5010 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
5012 struct inode
*inode
= file_inode(file
);
5013 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5014 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5015 struct btrfs_ioctl_qgroup_limit_args
*sa
;
5016 struct btrfs_trans_handle
*trans
;
5021 if (!capable(CAP_SYS_ADMIN
))
5024 ret
= mnt_want_write_file(file
);
5028 sa
= memdup_user(arg
, sizeof(*sa
));
5034 trans
= btrfs_join_transaction(root
);
5035 if (IS_ERR(trans
)) {
5036 ret
= PTR_ERR(trans
);
5040 qgroupid
= sa
->qgroupid
;
5042 /* take the current subvol as qgroup */
5043 qgroupid
= root
->root_key
.objectid
;
5046 /* FIXME: check if the IDs really exist */
5047 ret
= btrfs_limit_qgroup(trans
, fs_info
, qgroupid
, &sa
->lim
);
5049 err
= btrfs_end_transaction(trans
, root
);
5056 mnt_drop_write_file(file
);
5060 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
5062 struct inode
*inode
= file_inode(file
);
5063 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5064 struct btrfs_ioctl_quota_rescan_args
*qsa
;
5067 if (!capable(CAP_SYS_ADMIN
))
5070 ret
= mnt_want_write_file(file
);
5074 qsa
= memdup_user(arg
, sizeof(*qsa
));
5085 ret
= btrfs_qgroup_rescan(fs_info
);
5090 mnt_drop_write_file(file
);
5094 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
5096 struct inode
*inode
= file_inode(file
);
5097 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5098 struct btrfs_ioctl_quota_rescan_args
*qsa
;
5101 if (!capable(CAP_SYS_ADMIN
))
5104 qsa
= kzalloc(sizeof(*qsa
), GFP_KERNEL
);
5108 if (fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
5110 qsa
->progress
= fs_info
->qgroup_rescan_progress
.objectid
;
5113 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
5120 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
5122 struct inode
*inode
= file_inode(file
);
5123 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5125 if (!capable(CAP_SYS_ADMIN
))
5128 return btrfs_qgroup_wait_for_completion(fs_info
, true);
5131 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
5132 struct btrfs_ioctl_received_subvol_args
*sa
)
5134 struct inode
*inode
= file_inode(file
);
5135 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5136 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5137 struct btrfs_root_item
*root_item
= &root
->root_item
;
5138 struct btrfs_trans_handle
*trans
;
5139 struct timespec ct
= current_time(inode
);
5141 int received_uuid_changed
;
5143 if (!inode_owner_or_capable(inode
))
5146 ret
= mnt_want_write_file(file
);
5150 down_write(&fs_info
->subvol_sem
);
5152 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
5157 if (btrfs_root_readonly(root
)) {
5164 * 2 - uuid items (received uuid + subvol uuid)
5166 trans
= btrfs_start_transaction(root
, 3);
5167 if (IS_ERR(trans
)) {
5168 ret
= PTR_ERR(trans
);
5173 sa
->rtransid
= trans
->transid
;
5174 sa
->rtime
.sec
= ct
.tv_sec
;
5175 sa
->rtime
.nsec
= ct
.tv_nsec
;
5177 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
5179 if (received_uuid_changed
&&
5180 !btrfs_is_empty_uuid(root_item
->received_uuid
))
5181 btrfs_uuid_tree_rem(trans
, fs_info
, root_item
->received_uuid
,
5182 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5183 root
->root_key
.objectid
);
5184 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
5185 btrfs_set_root_stransid(root_item
, sa
->stransid
);
5186 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
5187 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
5188 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
5189 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
5190 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
5192 ret
= btrfs_update_root(trans
, fs_info
->tree_root
,
5193 &root
->root_key
, &root
->root_item
);
5195 btrfs_end_transaction(trans
, root
);
5198 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
5199 ret
= btrfs_uuid_tree_add(trans
, fs_info
, sa
->uuid
,
5200 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5201 root
->root_key
.objectid
);
5202 if (ret
< 0 && ret
!= -EEXIST
) {
5203 btrfs_abort_transaction(trans
, ret
);
5207 ret
= btrfs_commit_transaction(trans
, root
);
5209 btrfs_abort_transaction(trans
, ret
);
5214 up_write(&fs_info
->subvol_sem
);
5215 mnt_drop_write_file(file
);
5220 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
5223 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
5224 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
5227 args32
= memdup_user(arg
, sizeof(*args32
));
5229 return PTR_ERR(args32
);
5231 args64
= kmalloc(sizeof(*args64
), GFP_KERNEL
);
5237 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
5238 args64
->stransid
= args32
->stransid
;
5239 args64
->rtransid
= args32
->rtransid
;
5240 args64
->stime
.sec
= args32
->stime
.sec
;
5241 args64
->stime
.nsec
= args32
->stime
.nsec
;
5242 args64
->rtime
.sec
= args32
->rtime
.sec
;
5243 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
5244 args64
->flags
= args32
->flags
;
5246 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
5250 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
5251 args32
->stransid
= args64
->stransid
;
5252 args32
->rtransid
= args64
->rtransid
;
5253 args32
->stime
.sec
= args64
->stime
.sec
;
5254 args32
->stime
.nsec
= args64
->stime
.nsec
;
5255 args32
->rtime
.sec
= args64
->rtime
.sec
;
5256 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
5257 args32
->flags
= args64
->flags
;
5259 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
5270 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
5273 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
5276 sa
= memdup_user(arg
, sizeof(*sa
));
5280 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5285 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5294 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5296 struct inode
*inode
= file_inode(file
);
5297 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5300 char label
[BTRFS_LABEL_SIZE
];
5302 spin_lock(&fs_info
->super_lock
);
5303 memcpy(label
, fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5304 spin_unlock(&fs_info
->super_lock
);
5306 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5308 if (len
== BTRFS_LABEL_SIZE
) {
5310 "label is too long, return the first %zu bytes",
5314 ret
= copy_to_user(arg
, label
, len
);
5316 return ret
? -EFAULT
: 0;
5319 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5321 struct inode
*inode
= file_inode(file
);
5322 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5323 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5324 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
5325 struct btrfs_trans_handle
*trans
;
5326 char label
[BTRFS_LABEL_SIZE
];
5329 if (!capable(CAP_SYS_ADMIN
))
5332 if (copy_from_user(label
, arg
, sizeof(label
)))
5335 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5337 "unable to set label with more than %d bytes",
5338 BTRFS_LABEL_SIZE
- 1);
5342 ret
= mnt_want_write_file(file
);
5346 trans
= btrfs_start_transaction(root
, 0);
5347 if (IS_ERR(trans
)) {
5348 ret
= PTR_ERR(trans
);
5352 spin_lock(&fs_info
->super_lock
);
5353 strcpy(super_block
->label
, label
);
5354 spin_unlock(&fs_info
->super_lock
);
5355 ret
= btrfs_commit_transaction(trans
, root
);
5358 mnt_drop_write_file(file
);
5362 #define INIT_FEATURE_FLAGS(suffix) \
5363 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5364 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5365 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5367 int btrfs_ioctl_get_supported_features(void __user
*arg
)
5369 static const struct btrfs_ioctl_feature_flags features
[3] = {
5370 INIT_FEATURE_FLAGS(SUPP
),
5371 INIT_FEATURE_FLAGS(SAFE_SET
),
5372 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5375 if (copy_to_user(arg
, &features
, sizeof(features
)))
5381 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5383 struct inode
*inode
= file_inode(file
);
5384 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5385 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
5386 struct btrfs_ioctl_feature_flags features
;
5388 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5389 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5390 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5392 if (copy_to_user(arg
, &features
, sizeof(features
)))
5398 static int check_feature_bits(struct btrfs_root
*root
,
5399 enum btrfs_feature_set set
,
5400 u64 change_mask
, u64 flags
, u64 supported_flags
,
5401 u64 safe_set
, u64 safe_clear
)
5403 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
5404 const char *type
= btrfs_feature_set_names
[set
];
5406 u64 disallowed
, unsupported
;
5407 u64 set_mask
= flags
& change_mask
;
5408 u64 clear_mask
= ~flags
& change_mask
;
5410 unsupported
= set_mask
& ~supported_flags
;
5412 names
= btrfs_printable_features(set
, unsupported
);
5415 "this kernel does not support the %s feature bit%s",
5416 names
, strchr(names
, ',') ? "s" : "");
5420 "this kernel does not support %s bits 0x%llx",
5425 disallowed
= set_mask
& ~safe_set
;
5427 names
= btrfs_printable_features(set
, disallowed
);
5430 "can't set the %s feature bit%s while mounted",
5431 names
, strchr(names
, ',') ? "s" : "");
5435 "can't set %s bits 0x%llx while mounted",
5440 disallowed
= clear_mask
& ~safe_clear
;
5442 names
= btrfs_printable_features(set
, disallowed
);
5445 "can't clear the %s feature bit%s while mounted",
5446 names
, strchr(names
, ',') ? "s" : "");
5450 "can't clear %s bits 0x%llx while mounted",
5458 #define check_feature(root, change_mask, flags, mask_base) \
5459 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5460 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5461 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5462 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5464 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5466 struct inode
*inode
= file_inode(file
);
5467 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5468 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5469 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
5470 struct btrfs_ioctl_feature_flags flags
[2];
5471 struct btrfs_trans_handle
*trans
;
5475 if (!capable(CAP_SYS_ADMIN
))
5478 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5482 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5483 !flags
[0].incompat_flags
)
5486 ret
= check_feature(root
, flags
[0].compat_flags
,
5487 flags
[1].compat_flags
, COMPAT
);
5491 ret
= check_feature(root
, flags
[0].compat_ro_flags
,
5492 flags
[1].compat_ro_flags
, COMPAT_RO
);
5496 ret
= check_feature(root
, flags
[0].incompat_flags
,
5497 flags
[1].incompat_flags
, INCOMPAT
);
5501 ret
= mnt_want_write_file(file
);
5505 trans
= btrfs_start_transaction(root
, 0);
5506 if (IS_ERR(trans
)) {
5507 ret
= PTR_ERR(trans
);
5508 goto out_drop_write
;
5511 spin_lock(&fs_info
->super_lock
);
5512 newflags
= btrfs_super_compat_flags(super_block
);
5513 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5514 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5515 btrfs_set_super_compat_flags(super_block
, newflags
);
5517 newflags
= btrfs_super_compat_ro_flags(super_block
);
5518 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5519 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5520 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5522 newflags
= btrfs_super_incompat_flags(super_block
);
5523 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5524 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5525 btrfs_set_super_incompat_flags(super_block
, newflags
);
5526 spin_unlock(&fs_info
->super_lock
);
5528 ret
= btrfs_commit_transaction(trans
, root
);
5530 mnt_drop_write_file(file
);
5535 long btrfs_ioctl(struct file
*file
, unsigned int
5536 cmd
, unsigned long arg
)
5538 struct inode
*inode
= file_inode(file
);
5539 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5540 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5541 void __user
*argp
= (void __user
*)arg
;
5544 case FS_IOC_GETFLAGS
:
5545 return btrfs_ioctl_getflags(file
, argp
);
5546 case FS_IOC_SETFLAGS
:
5547 return btrfs_ioctl_setflags(file
, argp
);
5548 case FS_IOC_GETVERSION
:
5549 return btrfs_ioctl_getversion(file
, argp
);
5551 return btrfs_ioctl_fitrim(file
, argp
);
5552 case BTRFS_IOC_SNAP_CREATE
:
5553 return btrfs_ioctl_snap_create(file
, argp
, 0);
5554 case BTRFS_IOC_SNAP_CREATE_V2
:
5555 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5556 case BTRFS_IOC_SUBVOL_CREATE
:
5557 return btrfs_ioctl_snap_create(file
, argp
, 1);
5558 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5559 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5560 case BTRFS_IOC_SNAP_DESTROY
:
5561 return btrfs_ioctl_snap_destroy(file
, argp
);
5562 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5563 return btrfs_ioctl_subvol_getflags(file
, argp
);
5564 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5565 return btrfs_ioctl_subvol_setflags(file
, argp
);
5566 case BTRFS_IOC_DEFAULT_SUBVOL
:
5567 return btrfs_ioctl_default_subvol(file
, argp
);
5568 case BTRFS_IOC_DEFRAG
:
5569 return btrfs_ioctl_defrag(file
, NULL
);
5570 case BTRFS_IOC_DEFRAG_RANGE
:
5571 return btrfs_ioctl_defrag(file
, argp
);
5572 case BTRFS_IOC_RESIZE
:
5573 return btrfs_ioctl_resize(file
, argp
);
5574 case BTRFS_IOC_ADD_DEV
:
5575 return btrfs_ioctl_add_dev(root
, argp
);
5576 case BTRFS_IOC_RM_DEV
:
5577 return btrfs_ioctl_rm_dev(file
, argp
);
5578 case BTRFS_IOC_RM_DEV_V2
:
5579 return btrfs_ioctl_rm_dev_v2(file
, argp
);
5580 case BTRFS_IOC_FS_INFO
:
5581 return btrfs_ioctl_fs_info(root
, argp
);
5582 case BTRFS_IOC_DEV_INFO
:
5583 return btrfs_ioctl_dev_info(root
, argp
);
5584 case BTRFS_IOC_BALANCE
:
5585 return btrfs_ioctl_balance(file
, NULL
);
5586 case BTRFS_IOC_TRANS_START
:
5587 return btrfs_ioctl_trans_start(file
);
5588 case BTRFS_IOC_TRANS_END
:
5589 return btrfs_ioctl_trans_end(file
);
5590 case BTRFS_IOC_TREE_SEARCH
:
5591 return btrfs_ioctl_tree_search(file
, argp
);
5592 case BTRFS_IOC_TREE_SEARCH_V2
:
5593 return btrfs_ioctl_tree_search_v2(file
, argp
);
5594 case BTRFS_IOC_INO_LOOKUP
:
5595 return btrfs_ioctl_ino_lookup(file
, argp
);
5596 case BTRFS_IOC_INO_PATHS
:
5597 return btrfs_ioctl_ino_to_path(root
, argp
);
5598 case BTRFS_IOC_LOGICAL_INO
:
5599 return btrfs_ioctl_logical_to_ino(root
, argp
);
5600 case BTRFS_IOC_SPACE_INFO
:
5601 return btrfs_ioctl_space_info(root
, argp
);
5602 case BTRFS_IOC_SYNC
: {
5605 ret
= btrfs_start_delalloc_roots(fs_info
, 0, -1);
5608 ret
= btrfs_sync_fs(inode
->i_sb
, 1);
5610 * The transaction thread may want to do more work,
5611 * namely it pokes the cleaner kthread that will start
5612 * processing uncleaned subvols.
5614 wake_up_process(fs_info
->transaction_kthread
);
5617 case BTRFS_IOC_START_SYNC
:
5618 return btrfs_ioctl_start_sync(root
, argp
);
5619 case BTRFS_IOC_WAIT_SYNC
:
5620 return btrfs_ioctl_wait_sync(root
, argp
);
5621 case BTRFS_IOC_SCRUB
:
5622 return btrfs_ioctl_scrub(file
, argp
);
5623 case BTRFS_IOC_SCRUB_CANCEL
:
5624 return btrfs_ioctl_scrub_cancel(root
, argp
);
5625 case BTRFS_IOC_SCRUB_PROGRESS
:
5626 return btrfs_ioctl_scrub_progress(root
, argp
);
5627 case BTRFS_IOC_BALANCE_V2
:
5628 return btrfs_ioctl_balance(file
, argp
);
5629 case BTRFS_IOC_BALANCE_CTL
:
5630 return btrfs_ioctl_balance_ctl(root
, arg
);
5631 case BTRFS_IOC_BALANCE_PROGRESS
:
5632 return btrfs_ioctl_balance_progress(root
, argp
);
5633 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5634 return btrfs_ioctl_set_received_subvol(file
, argp
);
5636 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5637 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5639 case BTRFS_IOC_SEND
:
5640 return btrfs_ioctl_send(file
, argp
);
5641 case BTRFS_IOC_GET_DEV_STATS
:
5642 return btrfs_ioctl_get_dev_stats(root
, argp
);
5643 case BTRFS_IOC_QUOTA_CTL
:
5644 return btrfs_ioctl_quota_ctl(file
, argp
);
5645 case BTRFS_IOC_QGROUP_ASSIGN
:
5646 return btrfs_ioctl_qgroup_assign(file
, argp
);
5647 case BTRFS_IOC_QGROUP_CREATE
:
5648 return btrfs_ioctl_qgroup_create(file
, argp
);
5649 case BTRFS_IOC_QGROUP_LIMIT
:
5650 return btrfs_ioctl_qgroup_limit(file
, argp
);
5651 case BTRFS_IOC_QUOTA_RESCAN
:
5652 return btrfs_ioctl_quota_rescan(file
, argp
);
5653 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5654 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5655 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5656 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5657 case BTRFS_IOC_DEV_REPLACE
:
5658 return btrfs_ioctl_dev_replace(root
, argp
);
5659 case BTRFS_IOC_GET_FSLABEL
:
5660 return btrfs_ioctl_get_fslabel(file
, argp
);
5661 case BTRFS_IOC_SET_FSLABEL
:
5662 return btrfs_ioctl_set_fslabel(file
, argp
);
5663 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5664 return btrfs_ioctl_get_supported_features(argp
);
5665 case BTRFS_IOC_GET_FEATURES
:
5666 return btrfs_ioctl_get_features(file
, argp
);
5667 case BTRFS_IOC_SET_FEATURES
:
5668 return btrfs_ioctl_set_features(file
, argp
);
5674 #ifdef CONFIG_COMPAT
5675 long btrfs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
5678 case FS_IOC32_GETFLAGS
:
5679 cmd
= FS_IOC_GETFLAGS
;
5681 case FS_IOC32_SETFLAGS
:
5682 cmd
= FS_IOC_SETFLAGS
;
5684 case FS_IOC32_GETVERSION
:
5685 cmd
= FS_IOC_GETVERSION
;
5688 return -ENOIOCTLCMD
;
5691 return btrfs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));