1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/kernel.h>
8 #include <linux/file.h>
10 #include <linux/fsnotify.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/time.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/namei.h>
18 #include <linux/writeback.h>
19 #include <linux/compat.h>
20 #include <linux/security.h>
21 #include <linux/xattr.h>
23 #include <linux/slab.h>
24 #include <linux/blkdev.h>
25 #include <linux/uuid.h>
26 #include <linux/btrfs.h>
27 #include <linux/uaccess.h>
28 #include <linux/iversion.h>
31 #include "transaction.h"
32 #include "btrfs_inode.h"
33 #include "print-tree.h"
36 #include "inode-map.h"
38 #include "rcu-string.h"
40 #include "dev-replace.h"
45 #include "compression.h"
48 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
49 * structures are incorrect, as the timespec structure from userspace
50 * is 4 bytes too small. We define these alternatives here to teach
51 * the kernel about the 32-bit struct packing.
53 struct btrfs_ioctl_timespec_32
{
56 } __attribute__ ((__packed__
));
58 struct btrfs_ioctl_received_subvol_args_32
{
59 char uuid
[BTRFS_UUID_SIZE
]; /* in */
60 __u64 stransid
; /* in */
61 __u64 rtransid
; /* out */
62 struct btrfs_ioctl_timespec_32 stime
; /* in */
63 struct btrfs_ioctl_timespec_32 rtime
; /* out */
65 __u64 reserved
[16]; /* in */
66 } __attribute__ ((__packed__
));
68 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
69 struct btrfs_ioctl_received_subvol_args_32)
72 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
73 struct btrfs_ioctl_send_args_32
{
74 __s64 send_fd
; /* in */
75 __u64 clone_sources_count
; /* in */
76 compat_uptr_t clone_sources
; /* in */
77 __u64 parent_root
; /* in */
79 __u64 reserved
[4]; /* in */
80 } __attribute__ ((__packed__
));
82 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
83 struct btrfs_ioctl_send_args_32)
86 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
87 u64 off
, u64 olen
, u64 olen_aligned
, u64 destoff
,
90 /* Mask out flags that are inappropriate for the given type of inode. */
91 static unsigned int btrfs_mask_fsflags_for_type(struct inode
*inode
,
94 if (S_ISDIR(inode
->i_mode
))
96 else if (S_ISREG(inode
->i_mode
))
97 return flags
& ~FS_DIRSYNC_FL
;
99 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
103 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
106 static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags
)
108 unsigned int iflags
= 0;
110 if (flags
& BTRFS_INODE_SYNC
)
111 iflags
|= FS_SYNC_FL
;
112 if (flags
& BTRFS_INODE_IMMUTABLE
)
113 iflags
|= FS_IMMUTABLE_FL
;
114 if (flags
& BTRFS_INODE_APPEND
)
115 iflags
|= FS_APPEND_FL
;
116 if (flags
& BTRFS_INODE_NODUMP
)
117 iflags
|= FS_NODUMP_FL
;
118 if (flags
& BTRFS_INODE_NOATIME
)
119 iflags
|= FS_NOATIME_FL
;
120 if (flags
& BTRFS_INODE_DIRSYNC
)
121 iflags
|= FS_DIRSYNC_FL
;
122 if (flags
& BTRFS_INODE_NODATACOW
)
123 iflags
|= FS_NOCOW_FL
;
125 if (flags
& BTRFS_INODE_NOCOMPRESS
)
126 iflags
|= FS_NOCOMP_FL
;
127 else if (flags
& BTRFS_INODE_COMPRESS
)
128 iflags
|= FS_COMPR_FL
;
134 * Update inode->i_flags based on the btrfs internal flags.
136 void btrfs_sync_inode_flags_to_i_flags(struct inode
*inode
)
138 struct btrfs_inode
*binode
= BTRFS_I(inode
);
139 unsigned int new_fl
= 0;
141 if (binode
->flags
& BTRFS_INODE_SYNC
)
143 if (binode
->flags
& BTRFS_INODE_IMMUTABLE
)
144 new_fl
|= S_IMMUTABLE
;
145 if (binode
->flags
& BTRFS_INODE_APPEND
)
147 if (binode
->flags
& BTRFS_INODE_NOATIME
)
149 if (binode
->flags
& BTRFS_INODE_DIRSYNC
)
152 set_mask_bits(&inode
->i_flags
,
153 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
,
157 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
159 struct btrfs_inode
*binode
= BTRFS_I(file_inode(file
));
160 unsigned int flags
= btrfs_inode_flags_to_fsflags(binode
->flags
);
162 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
167 /* Check if @flags are a supported and valid set of FS_*_FL flags */
168 static int check_fsflags(unsigned int flags
)
170 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
171 FS_NOATIME_FL
| FS_NODUMP_FL
| \
172 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
173 FS_NOCOMP_FL
| FS_COMPR_FL
|
177 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
183 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
185 struct inode
*inode
= file_inode(file
);
186 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
187 struct btrfs_inode
*binode
= BTRFS_I(inode
);
188 struct btrfs_root
*root
= binode
->root
;
189 struct btrfs_trans_handle
*trans
;
190 unsigned int fsflags
, old_fsflags
;
193 unsigned int old_i_flags
;
196 if (!inode_owner_or_capable(inode
))
199 if (btrfs_root_readonly(root
))
202 if (copy_from_user(&fsflags
, arg
, sizeof(fsflags
)))
205 ret
= check_fsflags(fsflags
);
209 ret
= mnt_want_write_file(file
);
215 old_flags
= binode
->flags
;
216 old_i_flags
= inode
->i_flags
;
217 mode
= inode
->i_mode
;
219 fsflags
= btrfs_mask_fsflags_for_type(inode
, fsflags
);
220 old_fsflags
= btrfs_inode_flags_to_fsflags(binode
->flags
);
221 if ((fsflags
^ old_fsflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
222 if (!capable(CAP_LINUX_IMMUTABLE
)) {
228 if (fsflags
& FS_SYNC_FL
)
229 binode
->flags
|= BTRFS_INODE_SYNC
;
231 binode
->flags
&= ~BTRFS_INODE_SYNC
;
232 if (fsflags
& FS_IMMUTABLE_FL
)
233 binode
->flags
|= BTRFS_INODE_IMMUTABLE
;
235 binode
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
236 if (fsflags
& FS_APPEND_FL
)
237 binode
->flags
|= BTRFS_INODE_APPEND
;
239 binode
->flags
&= ~BTRFS_INODE_APPEND
;
240 if (fsflags
& FS_NODUMP_FL
)
241 binode
->flags
|= BTRFS_INODE_NODUMP
;
243 binode
->flags
&= ~BTRFS_INODE_NODUMP
;
244 if (fsflags
& FS_NOATIME_FL
)
245 binode
->flags
|= BTRFS_INODE_NOATIME
;
247 binode
->flags
&= ~BTRFS_INODE_NOATIME
;
248 if (fsflags
& FS_DIRSYNC_FL
)
249 binode
->flags
|= BTRFS_INODE_DIRSYNC
;
251 binode
->flags
&= ~BTRFS_INODE_DIRSYNC
;
252 if (fsflags
& FS_NOCOW_FL
) {
255 * It's safe to turn csums off here, no extents exist.
256 * Otherwise we want the flag to reflect the real COW
257 * status of the file and will not set it.
259 if (inode
->i_size
== 0)
260 binode
->flags
|= BTRFS_INODE_NODATACOW
261 | BTRFS_INODE_NODATASUM
;
263 binode
->flags
|= BTRFS_INODE_NODATACOW
;
267 * Revert back under same assumptions as above
270 if (inode
->i_size
== 0)
271 binode
->flags
&= ~(BTRFS_INODE_NODATACOW
272 | BTRFS_INODE_NODATASUM
);
274 binode
->flags
&= ~BTRFS_INODE_NODATACOW
;
279 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
280 * flag may be changed automatically if compression code won't make
283 if (fsflags
& FS_NOCOMP_FL
) {
284 binode
->flags
&= ~BTRFS_INODE_COMPRESS
;
285 binode
->flags
|= BTRFS_INODE_NOCOMPRESS
;
287 /* set no-compression no need to validate prop here */
288 ret
= btrfs_set_prop_trans(inode
, "btrfs.compression", NULL
,
290 if (ret
&& ret
!= -ENODATA
)
292 } else if (fsflags
& FS_COMPR_FL
) {
295 if (IS_SWAPFILE(inode
)) {
300 binode
->flags
|= BTRFS_INODE_COMPRESS
;
301 binode
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
303 /* compress_type is already validated during mount options */
304 comp
= btrfs_compress_type2str(fs_info
->compress_type
);
305 if (!comp
|| comp
[0] == 0)
306 comp
= btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB
);
308 ret
= btrfs_set_prop_trans(inode
, "btrfs.compression", comp
,
314 /* reset prop, no need of validate prop here */
315 ret
= btrfs_set_prop_trans(inode
, "btrfs.compression", NULL
,
317 if (ret
&& ret
!= -ENODATA
)
319 binode
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
322 trans
= btrfs_start_transaction(root
, 1);
324 ret
= PTR_ERR(trans
);
328 btrfs_sync_inode_flags_to_i_flags(inode
);
329 inode_inc_iversion(inode
);
330 inode
->i_ctime
= current_time(inode
);
331 ret
= btrfs_update_inode(trans
, root
, inode
);
333 btrfs_end_transaction(trans
);
336 binode
->flags
= old_flags
;
337 inode
->i_flags
= old_i_flags
;
342 mnt_drop_write_file(file
);
347 * Translate btrfs internal inode flags to xflags as expected by the
348 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
351 static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags
)
353 unsigned int xflags
= 0;
355 if (flags
& BTRFS_INODE_APPEND
)
356 xflags
|= FS_XFLAG_APPEND
;
357 if (flags
& BTRFS_INODE_IMMUTABLE
)
358 xflags
|= FS_XFLAG_IMMUTABLE
;
359 if (flags
& BTRFS_INODE_NOATIME
)
360 xflags
|= FS_XFLAG_NOATIME
;
361 if (flags
& BTRFS_INODE_NODUMP
)
362 xflags
|= FS_XFLAG_NODUMP
;
363 if (flags
& BTRFS_INODE_SYNC
)
364 xflags
|= FS_XFLAG_SYNC
;
369 /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
370 static int check_xflags(unsigned int flags
)
372 if (flags
& ~(FS_XFLAG_APPEND
| FS_XFLAG_IMMUTABLE
| FS_XFLAG_NOATIME
|
373 FS_XFLAG_NODUMP
| FS_XFLAG_SYNC
))
379 * Set the xflags from the internal inode flags. The remaining items of fsxattr
382 static int btrfs_ioctl_fsgetxattr(struct file
*file
, void __user
*arg
)
384 struct btrfs_inode
*binode
= BTRFS_I(file_inode(file
));
387 memset(&fa
, 0, sizeof(fa
));
388 fa
.fsx_xflags
= btrfs_inode_flags_to_xflags(binode
->flags
);
390 if (copy_to_user(arg
, &fa
, sizeof(fa
)))
396 static int btrfs_ioctl_fssetxattr(struct file
*file
, void __user
*arg
)
398 struct inode
*inode
= file_inode(file
);
399 struct btrfs_inode
*binode
= BTRFS_I(inode
);
400 struct btrfs_root
*root
= binode
->root
;
401 struct btrfs_trans_handle
*trans
;
404 unsigned old_i_flags
;
407 if (!inode_owner_or_capable(inode
))
410 if (btrfs_root_readonly(root
))
413 memset(&fa
, 0, sizeof(fa
));
414 if (copy_from_user(&fa
, arg
, sizeof(fa
)))
417 ret
= check_xflags(fa
.fsx_xflags
);
421 if (fa
.fsx_extsize
!= 0 || fa
.fsx_projid
!= 0 || fa
.fsx_cowextsize
!= 0)
424 ret
= mnt_want_write_file(file
);
430 old_flags
= binode
->flags
;
431 old_i_flags
= inode
->i_flags
;
433 /* We need the capabilities to change append-only or immutable inode */
434 if (((old_flags
& (BTRFS_INODE_APPEND
| BTRFS_INODE_IMMUTABLE
)) ||
435 (fa
.fsx_xflags
& (FS_XFLAG_APPEND
| FS_XFLAG_IMMUTABLE
))) &&
436 !capable(CAP_LINUX_IMMUTABLE
)) {
441 if (fa
.fsx_xflags
& FS_XFLAG_SYNC
)
442 binode
->flags
|= BTRFS_INODE_SYNC
;
444 binode
->flags
&= ~BTRFS_INODE_SYNC
;
445 if (fa
.fsx_xflags
& FS_XFLAG_IMMUTABLE
)
446 binode
->flags
|= BTRFS_INODE_IMMUTABLE
;
448 binode
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
449 if (fa
.fsx_xflags
& FS_XFLAG_APPEND
)
450 binode
->flags
|= BTRFS_INODE_APPEND
;
452 binode
->flags
&= ~BTRFS_INODE_APPEND
;
453 if (fa
.fsx_xflags
& FS_XFLAG_NODUMP
)
454 binode
->flags
|= BTRFS_INODE_NODUMP
;
456 binode
->flags
&= ~BTRFS_INODE_NODUMP
;
457 if (fa
.fsx_xflags
& FS_XFLAG_NOATIME
)
458 binode
->flags
|= BTRFS_INODE_NOATIME
;
460 binode
->flags
&= ~BTRFS_INODE_NOATIME
;
462 /* 1 item for the inode */
463 trans
= btrfs_start_transaction(root
, 1);
465 ret
= PTR_ERR(trans
);
469 btrfs_sync_inode_flags_to_i_flags(inode
);
470 inode_inc_iversion(inode
);
471 inode
->i_ctime
= current_time(inode
);
472 ret
= btrfs_update_inode(trans
, root
, inode
);
474 btrfs_end_transaction(trans
);
478 binode
->flags
= old_flags
;
479 inode
->i_flags
= old_i_flags
;
483 mnt_drop_write_file(file
);
488 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
490 struct inode
*inode
= file_inode(file
);
492 return put_user(inode
->i_generation
, arg
);
495 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
497 struct inode
*inode
= file_inode(file
);
498 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
499 struct btrfs_device
*device
;
500 struct request_queue
*q
;
501 struct fstrim_range range
;
502 u64 minlen
= ULLONG_MAX
;
506 if (!capable(CAP_SYS_ADMIN
))
510 * If the fs is mounted with nologreplay, which requires it to be
511 * mounted in RO mode as well, we can not allow discard on free space
512 * inside block groups, because log trees refer to extents that are not
513 * pinned in a block group's free space cache (pinning the extents is
514 * precisely the first phase of replaying a log tree).
516 if (btrfs_test_opt(fs_info
, NOLOGREPLAY
))
520 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
524 q
= bdev_get_queue(device
->bdev
);
525 if (blk_queue_discard(q
)) {
527 minlen
= min_t(u64
, q
->limits
.discard_granularity
,
535 if (copy_from_user(&range
, arg
, sizeof(range
)))
539 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
540 * block group is in the logical address space, which can be any
541 * sectorsize aligned bytenr in the range [0, U64_MAX].
543 if (range
.len
< fs_info
->sb
->s_blocksize
)
546 range
.minlen
= max(range
.minlen
, minlen
);
547 ret
= btrfs_trim_fs(fs_info
, &range
);
551 if (copy_to_user(arg
, &range
, sizeof(range
)))
557 int btrfs_is_empty_uuid(u8
*uuid
)
561 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
568 static noinline
int create_subvol(struct inode
*dir
,
569 struct dentry
*dentry
,
570 const char *name
, int namelen
,
572 struct btrfs_qgroup_inherit
*inherit
)
574 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
575 struct btrfs_trans_handle
*trans
;
576 struct btrfs_key key
;
577 struct btrfs_root_item
*root_item
;
578 struct btrfs_inode_item
*inode_item
;
579 struct extent_buffer
*leaf
;
580 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
581 struct btrfs_root
*new_root
;
582 struct btrfs_block_rsv block_rsv
;
583 struct timespec64 cur_time
= current_time(dir
);
588 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
592 root_item
= kzalloc(sizeof(*root_item
), GFP_KERNEL
);
596 ret
= btrfs_find_free_objectid(fs_info
->tree_root
, &objectid
);
601 * Don't create subvolume whose level is not zero. Or qgroup will be
602 * screwed up since it assumes subvolume qgroup's level to be 0.
604 if (btrfs_qgroup_level(objectid
)) {
609 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
611 * The same as the snapshot creation, please see the comment
612 * of create_snapshot().
614 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
, 8, false);
618 trans
= btrfs_start_transaction(root
, 0);
620 ret
= PTR_ERR(trans
);
621 btrfs_subvolume_release_metadata(fs_info
, &block_rsv
);
624 trans
->block_rsv
= &block_rsv
;
625 trans
->bytes_reserved
= block_rsv
.size
;
627 ret
= btrfs_qgroup_inherit(trans
, 0, objectid
, inherit
);
631 leaf
= btrfs_alloc_tree_block(trans
, root
, 0, objectid
, NULL
, 0, 0, 0);
637 btrfs_mark_buffer_dirty(leaf
);
639 inode_item
= &root_item
->inode
;
640 btrfs_set_stack_inode_generation(inode_item
, 1);
641 btrfs_set_stack_inode_size(inode_item
, 3);
642 btrfs_set_stack_inode_nlink(inode_item
, 1);
643 btrfs_set_stack_inode_nbytes(inode_item
,
645 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
647 btrfs_set_root_flags(root_item
, 0);
648 btrfs_set_root_limit(root_item
, 0);
649 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
651 btrfs_set_root_bytenr(root_item
, leaf
->start
);
652 btrfs_set_root_generation(root_item
, trans
->transid
);
653 btrfs_set_root_level(root_item
, 0);
654 btrfs_set_root_refs(root_item
, 1);
655 btrfs_set_root_used(root_item
, leaf
->len
);
656 btrfs_set_root_last_snapshot(root_item
, 0);
658 btrfs_set_root_generation_v2(root_item
,
659 btrfs_root_generation(root_item
));
660 uuid_le_gen(&new_uuid
);
661 memcpy(root_item
->uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
662 btrfs_set_stack_timespec_sec(&root_item
->otime
, cur_time
.tv_sec
);
663 btrfs_set_stack_timespec_nsec(&root_item
->otime
, cur_time
.tv_nsec
);
664 root_item
->ctime
= root_item
->otime
;
665 btrfs_set_root_ctransid(root_item
, trans
->transid
);
666 btrfs_set_root_otransid(root_item
, trans
->transid
);
668 btrfs_tree_unlock(leaf
);
669 free_extent_buffer(leaf
);
672 btrfs_set_root_dirid(root_item
, new_dirid
);
674 key
.objectid
= objectid
;
676 key
.type
= BTRFS_ROOT_ITEM_KEY
;
677 ret
= btrfs_insert_root(trans
, fs_info
->tree_root
, &key
,
682 key
.offset
= (u64
)-1;
683 new_root
= btrfs_read_fs_root_no_name(fs_info
, &key
);
684 if (IS_ERR(new_root
)) {
685 ret
= PTR_ERR(new_root
);
686 btrfs_abort_transaction(trans
, ret
);
690 btrfs_record_root_in_trans(trans
, new_root
);
692 ret
= btrfs_create_subvol_root(trans
, new_root
, root
, new_dirid
);
694 /* We potentially lose an unused inode item here */
695 btrfs_abort_transaction(trans
, ret
);
699 mutex_lock(&new_root
->objectid_mutex
);
700 new_root
->highest_objectid
= new_dirid
;
701 mutex_unlock(&new_root
->objectid_mutex
);
704 * insert the directory item
706 ret
= btrfs_set_inode_index(BTRFS_I(dir
), &index
);
708 btrfs_abort_transaction(trans
, ret
);
712 ret
= btrfs_insert_dir_item(trans
, name
, namelen
, BTRFS_I(dir
), &key
,
713 BTRFS_FT_DIR
, index
);
715 btrfs_abort_transaction(trans
, ret
);
719 btrfs_i_size_write(BTRFS_I(dir
), dir
->i_size
+ namelen
* 2);
720 ret
= btrfs_update_inode(trans
, root
, dir
);
723 ret
= btrfs_add_root_ref(trans
, objectid
, root
->root_key
.objectid
,
724 btrfs_ino(BTRFS_I(dir
)), index
, name
, namelen
);
727 ret
= btrfs_uuid_tree_add(trans
, root_item
->uuid
,
728 BTRFS_UUID_KEY_SUBVOL
, objectid
);
730 btrfs_abort_transaction(trans
, ret
);
734 trans
->block_rsv
= NULL
;
735 trans
->bytes_reserved
= 0;
736 btrfs_subvolume_release_metadata(fs_info
, &block_rsv
);
739 *async_transid
= trans
->transid
;
740 err
= btrfs_commit_transaction_async(trans
, 1);
742 err
= btrfs_commit_transaction(trans
);
744 err
= btrfs_commit_transaction(trans
);
750 inode
= btrfs_lookup_dentry(dir
, dentry
);
752 return PTR_ERR(inode
);
753 d_instantiate(dentry
, inode
);
762 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
763 struct dentry
*dentry
,
764 u64
*async_transid
, bool readonly
,
765 struct btrfs_qgroup_inherit
*inherit
)
767 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
769 struct btrfs_pending_snapshot
*pending_snapshot
;
770 struct btrfs_trans_handle
*trans
;
772 bool snapshot_force_cow
= false;
774 if (!test_bit(BTRFS_ROOT_REF_COWS
, &root
->state
))
777 if (atomic_read(&root
->nr_swapfiles
)) {
779 "cannot snapshot subvolume with active swapfile");
783 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_KERNEL
);
784 if (!pending_snapshot
)
787 pending_snapshot
->root_item
= kzalloc(sizeof(struct btrfs_root_item
),
789 pending_snapshot
->path
= btrfs_alloc_path();
790 if (!pending_snapshot
->root_item
|| !pending_snapshot
->path
) {
796 * Force new buffered writes to reserve space even when NOCOW is
797 * possible. This is to avoid later writeback (running dealloc) to
798 * fallback to COW mode and unexpectedly fail with ENOSPC.
800 atomic_inc(&root
->will_be_snapshotted
);
801 smp_mb__after_atomic();
802 /* wait for no snapshot writes */
803 wait_event(root
->subv_writers
->wait
,
804 percpu_counter_sum(&root
->subv_writers
->counter
) == 0);
806 ret
= btrfs_start_delalloc_snapshot(root
);
811 * All previous writes have started writeback in NOCOW mode, so now
812 * we force future writes to fallback to COW mode during snapshot
815 atomic_inc(&root
->snapshot_force_cow
);
816 snapshot_force_cow
= true;
818 btrfs_wait_ordered_extents(root
, U64_MAX
, 0, (u64
)-1);
820 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
821 BTRFS_BLOCK_RSV_TEMP
);
823 * 1 - parent dir inode
826 * 2 - root ref/backref
827 * 1 - root of snapshot
830 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
831 &pending_snapshot
->block_rsv
, 8,
836 pending_snapshot
->dentry
= dentry
;
837 pending_snapshot
->root
= root
;
838 pending_snapshot
->readonly
= readonly
;
839 pending_snapshot
->dir
= dir
;
840 pending_snapshot
->inherit
= inherit
;
842 trans
= btrfs_start_transaction(root
, 0);
844 ret
= PTR_ERR(trans
);
848 spin_lock(&fs_info
->trans_lock
);
849 list_add(&pending_snapshot
->list
,
850 &trans
->transaction
->pending_snapshots
);
851 spin_unlock(&fs_info
->trans_lock
);
853 *async_transid
= trans
->transid
;
854 ret
= btrfs_commit_transaction_async(trans
, 1);
856 ret
= btrfs_commit_transaction(trans
);
858 ret
= btrfs_commit_transaction(trans
);
863 ret
= pending_snapshot
->error
;
867 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
871 inode
= btrfs_lookup_dentry(d_inode(dentry
->d_parent
), dentry
);
873 ret
= PTR_ERR(inode
);
877 d_instantiate(dentry
, inode
);
880 btrfs_subvolume_release_metadata(fs_info
, &pending_snapshot
->block_rsv
);
882 if (snapshot_force_cow
)
883 atomic_dec(&root
->snapshot_force_cow
);
884 if (atomic_dec_and_test(&root
->will_be_snapshotted
))
885 wake_up_var(&root
->will_be_snapshotted
);
887 kfree(pending_snapshot
->root_item
);
888 btrfs_free_path(pending_snapshot
->path
);
889 kfree(pending_snapshot
);
894 /* copy of may_delete in fs/namei.c()
895 * Check whether we can remove a link victim from directory dir, check
896 * whether the type of victim is right.
897 * 1. We can't do it if dir is read-only (done in permission())
898 * 2. We should have write and exec permissions on dir
899 * 3. We can't remove anything from append-only dir
900 * 4. We can't do anything with immutable dir (done in permission())
901 * 5. If the sticky bit on dir is set we should either
902 * a. be owner of dir, or
903 * b. be owner of victim, or
904 * c. have CAP_FOWNER capability
905 * 6. If the victim is append-only or immutable we can't do anything with
906 * links pointing to it.
907 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
908 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
909 * 9. We can't remove a root or mountpoint.
910 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
911 * nfs_async_unlink().
914 static int btrfs_may_delete(struct inode
*dir
, struct dentry
*victim
, int isdir
)
918 if (d_really_is_negative(victim
))
921 BUG_ON(d_inode(victim
->d_parent
) != dir
);
922 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
924 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
929 if (check_sticky(dir
, d_inode(victim
)) || IS_APPEND(d_inode(victim
)) ||
930 IS_IMMUTABLE(d_inode(victim
)) || IS_SWAPFILE(d_inode(victim
)))
933 if (!d_is_dir(victim
))
937 } else if (d_is_dir(victim
))
941 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
946 /* copy of may_create in fs/namei.c() */
947 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
949 if (d_really_is_positive(child
))
953 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
957 * Create a new subvolume below @parent. This is largely modeled after
958 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
959 * inside this filesystem so it's quite a bit simpler.
961 static noinline
int btrfs_mksubvol(const struct path
*parent
,
962 const char *name
, int namelen
,
963 struct btrfs_root
*snap_src
,
964 u64
*async_transid
, bool readonly
,
965 struct btrfs_qgroup_inherit
*inherit
)
967 struct inode
*dir
= d_inode(parent
->dentry
);
968 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
969 struct dentry
*dentry
;
972 error
= down_write_killable_nested(&dir
->i_rwsem
, I_MUTEX_PARENT
);
976 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
977 error
= PTR_ERR(dentry
);
981 error
= btrfs_may_create(dir
, dentry
);
986 * even if this name doesn't exist, we may get hash collisions.
987 * check for them now when we can safely fail
989 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
995 down_read(&fs_info
->subvol_sem
);
997 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
1001 error
= create_snapshot(snap_src
, dir
, dentry
,
1002 async_transid
, readonly
, inherit
);
1004 error
= create_subvol(dir
, dentry
, name
, namelen
,
1005 async_transid
, inherit
);
1008 fsnotify_mkdir(dir
, dentry
);
1010 up_read(&fs_info
->subvol_sem
);
1019 * When we're defragging a range, we don't want to kick it off again
1020 * if it is really just waiting for delalloc to send it down.
1021 * If we find a nice big extent or delalloc range for the bytes in the
1022 * file you want to defrag, we return 0 to let you know to skip this
1025 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, u32 thresh
)
1027 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1028 struct extent_map
*em
= NULL
;
1029 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
1032 read_lock(&em_tree
->lock
);
1033 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_SIZE
);
1034 read_unlock(&em_tree
->lock
);
1037 end
= extent_map_end(em
);
1038 free_extent_map(em
);
1039 if (end
- offset
> thresh
)
1042 /* if we already have a nice delalloc here, just stop */
1044 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
1045 thresh
, EXTENT_DELALLOC
, 1);
1052 * helper function to walk through a file and find extents
1053 * newer than a specific transid, and smaller than thresh.
1055 * This is used by the defragging code to find new and small
1058 static int find_new_extents(struct btrfs_root
*root
,
1059 struct inode
*inode
, u64 newer_than
,
1060 u64
*off
, u32 thresh
)
1062 struct btrfs_path
*path
;
1063 struct btrfs_key min_key
;
1064 struct extent_buffer
*leaf
;
1065 struct btrfs_file_extent_item
*extent
;
1068 u64 ino
= btrfs_ino(BTRFS_I(inode
));
1070 path
= btrfs_alloc_path();
1074 min_key
.objectid
= ino
;
1075 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
1076 min_key
.offset
= *off
;
1079 ret
= btrfs_search_forward(root
, &min_key
, path
, newer_than
);
1083 if (min_key
.objectid
!= ino
)
1085 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
1088 leaf
= path
->nodes
[0];
1089 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
1090 struct btrfs_file_extent_item
);
1092 type
= btrfs_file_extent_type(leaf
, extent
);
1093 if (type
== BTRFS_FILE_EXTENT_REG
&&
1094 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
1095 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
1096 *off
= min_key
.offset
;
1097 btrfs_free_path(path
);
1102 if (path
->slots
[0] < btrfs_header_nritems(leaf
)) {
1103 btrfs_item_key_to_cpu(leaf
, &min_key
, path
->slots
[0]);
1107 if (min_key
.offset
== (u64
)-1)
1111 btrfs_release_path(path
);
1114 btrfs_free_path(path
);
1118 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
1120 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
1121 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
1122 struct extent_map
*em
;
1123 u64 len
= PAGE_SIZE
;
1126 * hopefully we have this extent in the tree already, try without
1127 * the full extent lock
1129 read_lock(&em_tree
->lock
);
1130 em
= lookup_extent_mapping(em_tree
, start
, len
);
1131 read_unlock(&em_tree
->lock
);
1134 struct extent_state
*cached
= NULL
;
1135 u64 end
= start
+ len
- 1;
1137 /* get the big lock and read metadata off disk */
1138 lock_extent_bits(io_tree
, start
, end
, &cached
);
1139 em
= btrfs_get_extent(BTRFS_I(inode
), NULL
, 0, start
, len
, 0);
1140 unlock_extent_cached(io_tree
, start
, end
, &cached
);
1149 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
1151 struct extent_map
*next
;
1154 /* this is the last extent */
1155 if (em
->start
+ em
->len
>= i_size_read(inode
))
1158 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
1159 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
1161 else if ((em
->block_start
+ em
->block_len
== next
->block_start
) &&
1162 (em
->block_len
> SZ_128K
&& next
->block_len
> SZ_128K
))
1165 free_extent_map(next
);
1169 static int should_defrag_range(struct inode
*inode
, u64 start
, u32 thresh
,
1170 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
1173 struct extent_map
*em
;
1175 bool next_mergeable
= true;
1176 bool prev_mergeable
= true;
1179 * make sure that once we start defragging an extent, we keep on
1182 if (start
< *defrag_end
)
1187 em
= defrag_lookup_extent(inode
, start
);
1191 /* this will cover holes, and inline extents */
1192 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
1198 prev_mergeable
= false;
1200 next_mergeable
= defrag_check_next_extent(inode
, em
);
1202 * we hit a real extent, if it is big or the next extent is not a
1203 * real extent, don't bother defragging it
1205 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
1206 (em
->len
>= thresh
|| (!next_mergeable
&& !prev_mergeable
)))
1210 * last_len ends up being a counter of how many bytes we've defragged.
1211 * every time we choose not to defrag an extent, we reset *last_len
1212 * so that the next tiny extent will force a defrag.
1214 * The end result of this is that tiny extents before a single big
1215 * extent will force at least part of that big extent to be defragged.
1218 *defrag_end
= extent_map_end(em
);
1221 *skip
= extent_map_end(em
);
1225 free_extent_map(em
);
1230 * it doesn't do much good to defrag one or two pages
1231 * at a time. This pulls in a nice chunk of pages
1232 * to COW and defrag.
1234 * It also makes sure the delalloc code has enough
1235 * dirty data to avoid making new small extents as part
1238 * It's a good idea to start RA on this range
1239 * before calling this.
1241 static int cluster_pages_for_defrag(struct inode
*inode
,
1242 struct page
**pages
,
1243 unsigned long start_index
,
1244 unsigned long num_pages
)
1246 unsigned long file_end
;
1247 u64 isize
= i_size_read(inode
);
1254 struct btrfs_ordered_extent
*ordered
;
1255 struct extent_state
*cached_state
= NULL
;
1256 struct extent_io_tree
*tree
;
1257 struct extent_changeset
*data_reserved
= NULL
;
1258 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1260 file_end
= (isize
- 1) >> PAGE_SHIFT
;
1261 if (!isize
|| start_index
> file_end
)
1264 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1266 ret
= btrfs_delalloc_reserve_space(inode
, &data_reserved
,
1267 start_index
<< PAGE_SHIFT
,
1268 page_cnt
<< PAGE_SHIFT
);
1272 tree
= &BTRFS_I(inode
)->io_tree
;
1274 /* step one, lock all the pages */
1275 for (i
= 0; i
< page_cnt
; i
++) {
1278 page
= find_or_create_page(inode
->i_mapping
,
1279 start_index
+ i
, mask
);
1283 page_start
= page_offset(page
);
1284 page_end
= page_start
+ PAGE_SIZE
- 1;
1286 lock_extent_bits(tree
, page_start
, page_end
,
1288 ordered
= btrfs_lookup_ordered_extent(inode
,
1290 unlock_extent_cached(tree
, page_start
, page_end
,
1296 btrfs_start_ordered_extent(inode
, ordered
, 1);
1297 btrfs_put_ordered_extent(ordered
);
1300 * we unlocked the page above, so we need check if
1301 * it was released or not.
1303 if (page
->mapping
!= inode
->i_mapping
) {
1310 if (!PageUptodate(page
)) {
1311 btrfs_readpage(NULL
, page
);
1313 if (!PageUptodate(page
)) {
1321 if (page
->mapping
!= inode
->i_mapping
) {
1333 if (!(inode
->i_sb
->s_flags
& SB_ACTIVE
))
1337 * so now we have a nice long stream of locked
1338 * and up to date pages, lets wait on them
1340 for (i
= 0; i
< i_done
; i
++)
1341 wait_on_page_writeback(pages
[i
]);
1343 page_start
= page_offset(pages
[0]);
1344 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_SIZE
;
1346 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1347 page_start
, page_end
- 1, &cached_state
);
1348 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1349 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1350 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1353 if (i_done
!= page_cnt
) {
1354 spin_lock(&BTRFS_I(inode
)->lock
);
1355 btrfs_mod_outstanding_extents(BTRFS_I(inode
), 1);
1356 spin_unlock(&BTRFS_I(inode
)->lock
);
1357 btrfs_delalloc_release_space(inode
, data_reserved
,
1358 start_index
<< PAGE_SHIFT
,
1359 (page_cnt
- i_done
) << PAGE_SHIFT
, true);
1363 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1366 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1367 page_start
, page_end
- 1, &cached_state
);
1369 for (i
= 0; i
< i_done
; i
++) {
1370 clear_page_dirty_for_io(pages
[i
]);
1371 ClearPageChecked(pages
[i
]);
1372 set_page_extent_mapped(pages
[i
]);
1373 set_page_dirty(pages
[i
]);
1374 unlock_page(pages
[i
]);
1377 btrfs_delalloc_release_extents(BTRFS_I(inode
), page_cnt
<< PAGE_SHIFT
,
1379 extent_changeset_free(data_reserved
);
1382 for (i
= 0; i
< i_done
; i
++) {
1383 unlock_page(pages
[i
]);
1386 btrfs_delalloc_release_space(inode
, data_reserved
,
1387 start_index
<< PAGE_SHIFT
,
1388 page_cnt
<< PAGE_SHIFT
, true);
1389 btrfs_delalloc_release_extents(BTRFS_I(inode
), page_cnt
<< PAGE_SHIFT
,
1391 extent_changeset_free(data_reserved
);
1396 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1397 struct btrfs_ioctl_defrag_range_args
*range
,
1398 u64 newer_than
, unsigned long max_to_defrag
)
1400 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1401 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1402 struct file_ra_state
*ra
= NULL
;
1403 unsigned long last_index
;
1404 u64 isize
= i_size_read(inode
);
1408 u64 newer_off
= range
->start
;
1410 unsigned long ra_index
= 0;
1412 int defrag_count
= 0;
1413 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1414 u32 extent_thresh
= range
->extent_thresh
;
1415 unsigned long max_cluster
= SZ_256K
>> PAGE_SHIFT
;
1416 unsigned long cluster
= max_cluster
;
1417 u64 new_align
= ~((u64
)SZ_128K
- 1);
1418 struct page
**pages
= NULL
;
1419 bool do_compress
= range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
;
1424 if (range
->start
>= isize
)
1428 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1430 if (range
->compress_type
)
1431 compress_type
= range
->compress_type
;
1434 if (extent_thresh
== 0)
1435 extent_thresh
= SZ_256K
;
1438 * If we were not given a file, allocate a readahead context. As
1439 * readahead is just an optimization, defrag will work without it so
1440 * we don't error out.
1443 ra
= kzalloc(sizeof(*ra
), GFP_KERNEL
);
1445 file_ra_state_init(ra
, inode
->i_mapping
);
1450 pages
= kmalloc_array(max_cluster
, sizeof(struct page
*), GFP_KERNEL
);
1456 /* find the last page to defrag */
1457 if (range
->start
+ range
->len
> range
->start
) {
1458 last_index
= min_t(u64
, isize
- 1,
1459 range
->start
+ range
->len
- 1) >> PAGE_SHIFT
;
1461 last_index
= (isize
- 1) >> PAGE_SHIFT
;
1465 ret
= find_new_extents(root
, inode
, newer_than
,
1466 &newer_off
, SZ_64K
);
1468 range
->start
= newer_off
;
1470 * we always align our defrag to help keep
1471 * the extents in the file evenly spaced
1473 i
= (newer_off
& new_align
) >> PAGE_SHIFT
;
1477 i
= range
->start
>> PAGE_SHIFT
;
1480 max_to_defrag
= last_index
- i
+ 1;
1483 * make writeback starts from i, so the defrag range can be
1484 * written sequentially.
1486 if (i
< inode
->i_mapping
->writeback_index
)
1487 inode
->i_mapping
->writeback_index
= i
;
1489 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1490 (i
< DIV_ROUND_UP(i_size_read(inode
), PAGE_SIZE
))) {
1492 * make sure we stop running if someone unmounts
1495 if (!(inode
->i_sb
->s_flags
& SB_ACTIVE
))
1498 if (btrfs_defrag_cancelled(fs_info
)) {
1499 btrfs_debug(fs_info
, "defrag_file cancelled");
1504 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_SHIFT
,
1505 extent_thresh
, &last_len
, &skip
,
1506 &defrag_end
, do_compress
)){
1509 * the should_defrag function tells us how much to skip
1510 * bump our counter by the suggested amount
1512 next
= DIV_ROUND_UP(skip
, PAGE_SIZE
);
1513 i
= max(i
+ 1, next
);
1518 cluster
= (PAGE_ALIGN(defrag_end
) >>
1520 cluster
= min(cluster
, max_cluster
);
1522 cluster
= max_cluster
;
1525 if (i
+ cluster
> ra_index
) {
1526 ra_index
= max(i
, ra_index
);
1528 page_cache_sync_readahead(inode
->i_mapping
, ra
,
1529 file
, ra_index
, cluster
);
1530 ra_index
+= cluster
;
1534 if (IS_SWAPFILE(inode
)) {
1538 BTRFS_I(inode
)->defrag_compress
= compress_type
;
1539 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1542 inode_unlock(inode
);
1546 defrag_count
+= ret
;
1547 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1548 inode_unlock(inode
);
1551 if (newer_off
== (u64
)-1)
1557 newer_off
= max(newer_off
+ 1,
1558 (u64
)i
<< PAGE_SHIFT
);
1560 ret
= find_new_extents(root
, inode
, newer_than
,
1561 &newer_off
, SZ_64K
);
1563 range
->start
= newer_off
;
1564 i
= (newer_off
& new_align
) >> PAGE_SHIFT
;
1571 last_len
+= ret
<< PAGE_SHIFT
;
1579 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
)) {
1580 filemap_flush(inode
->i_mapping
);
1581 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT
,
1582 &BTRFS_I(inode
)->runtime_flags
))
1583 filemap_flush(inode
->i_mapping
);
1586 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1587 btrfs_set_fs_incompat(fs_info
, COMPRESS_LZO
);
1588 } else if (range
->compress_type
== BTRFS_COMPRESS_ZSTD
) {
1589 btrfs_set_fs_incompat(fs_info
, COMPRESS_ZSTD
);
1597 BTRFS_I(inode
)->defrag_compress
= BTRFS_COMPRESS_NONE
;
1598 inode_unlock(inode
);
1606 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1609 struct inode
*inode
= file_inode(file
);
1610 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1614 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1615 struct btrfs_ioctl_vol_args
*vol_args
;
1616 struct btrfs_trans_handle
*trans
;
1617 struct btrfs_device
*device
= NULL
;
1620 char *devstr
= NULL
;
1624 if (!capable(CAP_SYS_ADMIN
))
1627 ret
= mnt_want_write_file(file
);
1631 if (test_and_set_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
)) {
1632 mnt_drop_write_file(file
);
1633 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1636 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1637 if (IS_ERR(vol_args
)) {
1638 ret
= PTR_ERR(vol_args
);
1642 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1644 sizestr
= vol_args
->name
;
1645 devstr
= strchr(sizestr
, ':');
1647 sizestr
= devstr
+ 1;
1649 devstr
= vol_args
->name
;
1650 ret
= kstrtoull(devstr
, 10, &devid
);
1657 btrfs_info(fs_info
, "resizing devid %llu", devid
);
1660 device
= btrfs_find_device(fs_info
->fs_devices
, devid
, NULL
, NULL
, true);
1662 btrfs_info(fs_info
, "resizer unable to find device %llu",
1668 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE
, &device
->dev_state
)) {
1670 "resizer unable to apply on readonly device %llu",
1676 if (!strcmp(sizestr
, "max"))
1677 new_size
= device
->bdev
->bd_inode
->i_size
;
1679 if (sizestr
[0] == '-') {
1682 } else if (sizestr
[0] == '+') {
1686 new_size
= memparse(sizestr
, &retptr
);
1687 if (*retptr
!= '\0' || new_size
== 0) {
1693 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT
, &device
->dev_state
)) {
1698 old_size
= btrfs_device_get_total_bytes(device
);
1701 if (new_size
> old_size
) {
1705 new_size
= old_size
- new_size
;
1706 } else if (mod
> 0) {
1707 if (new_size
> ULLONG_MAX
- old_size
) {
1711 new_size
= old_size
+ new_size
;
1714 if (new_size
< SZ_256M
) {
1718 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1723 new_size
= round_down(new_size
, fs_info
->sectorsize
);
1725 btrfs_info_in_rcu(fs_info
, "new size for %s is %llu",
1726 rcu_str_deref(device
->name
), new_size
);
1728 if (new_size
> old_size
) {
1729 trans
= btrfs_start_transaction(root
, 0);
1730 if (IS_ERR(trans
)) {
1731 ret
= PTR_ERR(trans
);
1734 ret
= btrfs_grow_device(trans
, device
, new_size
);
1735 btrfs_commit_transaction(trans
);
1736 } else if (new_size
< old_size
) {
1737 ret
= btrfs_shrink_device(device
, new_size
);
1738 } /* equal, nothing need to do */
1743 clear_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
);
1744 mnt_drop_write_file(file
);
1748 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1749 const char *name
, unsigned long fd
, int subvol
,
1750 u64
*transid
, bool readonly
,
1751 struct btrfs_qgroup_inherit
*inherit
)
1756 if (!S_ISDIR(file_inode(file
)->i_mode
))
1759 ret
= mnt_want_write_file(file
);
1763 namelen
= strlen(name
);
1764 if (strchr(name
, '/')) {
1766 goto out_drop_write
;
1769 if (name
[0] == '.' &&
1770 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1772 goto out_drop_write
;
1776 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1777 NULL
, transid
, readonly
, inherit
);
1779 struct fd src
= fdget(fd
);
1780 struct inode
*src_inode
;
1783 goto out_drop_write
;
1786 src_inode
= file_inode(src
.file
);
1787 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1788 btrfs_info(BTRFS_I(file_inode(file
))->root
->fs_info
,
1789 "Snapshot src from another FS");
1791 } else if (!inode_owner_or_capable(src_inode
)) {
1793 * Subvolume creation is not restricted, but snapshots
1794 * are limited to own subvolumes only
1798 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1799 BTRFS_I(src_inode
)->root
,
1800 transid
, readonly
, inherit
);
1805 mnt_drop_write_file(file
);
1810 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1811 void __user
*arg
, int subvol
)
1813 struct btrfs_ioctl_vol_args
*vol_args
;
1816 if (!S_ISDIR(file_inode(file
)->i_mode
))
1819 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1820 if (IS_ERR(vol_args
))
1821 return PTR_ERR(vol_args
);
1822 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1824 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1825 vol_args
->fd
, subvol
,
1832 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1833 void __user
*arg
, int subvol
)
1835 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1839 bool readonly
= false;
1840 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1842 if (!S_ISDIR(file_inode(file
)->i_mode
))
1845 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1846 if (IS_ERR(vol_args
))
1847 return PTR_ERR(vol_args
);
1848 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1850 if (vol_args
->flags
&
1851 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1852 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1857 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1859 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1861 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1862 if (vol_args
->size
> PAGE_SIZE
) {
1866 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1867 if (IS_ERR(inherit
)) {
1868 ret
= PTR_ERR(inherit
);
1873 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1874 vol_args
->fd
, subvol
, ptr
,
1879 if (ptr
&& copy_to_user(arg
+
1880 offsetof(struct btrfs_ioctl_vol_args_v2
,
1892 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1895 struct inode
*inode
= file_inode(file
);
1896 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1897 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1901 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
)
1904 down_read(&fs_info
->subvol_sem
);
1905 if (btrfs_root_readonly(root
))
1906 flags
|= BTRFS_SUBVOL_RDONLY
;
1907 up_read(&fs_info
->subvol_sem
);
1909 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1915 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1918 struct inode
*inode
= file_inode(file
);
1919 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1920 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1921 struct btrfs_trans_handle
*trans
;
1926 if (!inode_owner_or_capable(inode
))
1929 ret
= mnt_want_write_file(file
);
1933 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
) {
1935 goto out_drop_write
;
1938 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1940 goto out_drop_write
;
1943 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1945 goto out_drop_write
;
1948 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1950 goto out_drop_write
;
1953 down_write(&fs_info
->subvol_sem
);
1956 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1959 root_flags
= btrfs_root_flags(&root
->root_item
);
1960 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1961 btrfs_set_root_flags(&root
->root_item
,
1962 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1965 * Block RO -> RW transition if this subvolume is involved in
1968 spin_lock(&root
->root_item_lock
);
1969 if (root
->send_in_progress
== 0) {
1970 btrfs_set_root_flags(&root
->root_item
,
1971 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1972 spin_unlock(&root
->root_item_lock
);
1974 spin_unlock(&root
->root_item_lock
);
1976 "Attempt to set subvolume %llu read-write during send",
1977 root
->root_key
.objectid
);
1983 trans
= btrfs_start_transaction(root
, 1);
1984 if (IS_ERR(trans
)) {
1985 ret
= PTR_ERR(trans
);
1989 ret
= btrfs_update_root(trans
, fs_info
->tree_root
,
1990 &root
->root_key
, &root
->root_item
);
1992 btrfs_end_transaction(trans
);
1996 ret
= btrfs_commit_transaction(trans
);
2000 btrfs_set_root_flags(&root
->root_item
, root_flags
);
2002 up_write(&fs_info
->subvol_sem
);
2004 mnt_drop_write_file(file
);
2009 static noinline
int key_in_sk(struct btrfs_key
*key
,
2010 struct btrfs_ioctl_search_key
*sk
)
2012 struct btrfs_key test
;
2015 test
.objectid
= sk
->min_objectid
;
2016 test
.type
= sk
->min_type
;
2017 test
.offset
= sk
->min_offset
;
2019 ret
= btrfs_comp_cpu_keys(key
, &test
);
2023 test
.objectid
= sk
->max_objectid
;
2024 test
.type
= sk
->max_type
;
2025 test
.offset
= sk
->max_offset
;
2027 ret
= btrfs_comp_cpu_keys(key
, &test
);
2033 static noinline
int copy_to_sk(struct btrfs_path
*path
,
2034 struct btrfs_key
*key
,
2035 struct btrfs_ioctl_search_key
*sk
,
2038 unsigned long *sk_offset
,
2042 struct extent_buffer
*leaf
;
2043 struct btrfs_ioctl_search_header sh
;
2044 struct btrfs_key test
;
2045 unsigned long item_off
;
2046 unsigned long item_len
;
2052 leaf
= path
->nodes
[0];
2053 slot
= path
->slots
[0];
2054 nritems
= btrfs_header_nritems(leaf
);
2056 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
2060 found_transid
= btrfs_header_generation(leaf
);
2062 for (i
= slot
; i
< nritems
; i
++) {
2063 item_off
= btrfs_item_ptr_offset(leaf
, i
);
2064 item_len
= btrfs_item_size_nr(leaf
, i
);
2066 btrfs_item_key_to_cpu(leaf
, key
, i
);
2067 if (!key_in_sk(key
, sk
))
2070 if (sizeof(sh
) + item_len
> *buf_size
) {
2077 * return one empty item back for v1, which does not
2081 *buf_size
= sizeof(sh
) + item_len
;
2086 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
2091 sh
.objectid
= key
->objectid
;
2092 sh
.offset
= key
->offset
;
2093 sh
.type
= key
->type
;
2095 sh
.transid
= found_transid
;
2097 /* copy search result header */
2098 if (copy_to_user(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
2103 *sk_offset
+= sizeof(sh
);
2106 char __user
*up
= ubuf
+ *sk_offset
;
2108 if (read_extent_buffer_to_user(leaf
, up
,
2109 item_off
, item_len
)) {
2114 *sk_offset
+= item_len
;
2118 if (ret
) /* -EOVERFLOW from above */
2121 if (*num_found
>= sk
->nr_items
) {
2128 test
.objectid
= sk
->max_objectid
;
2129 test
.type
= sk
->max_type
;
2130 test
.offset
= sk
->max_offset
;
2131 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
2133 else if (key
->offset
< (u64
)-1)
2135 else if (key
->type
< (u8
)-1) {
2138 } else if (key
->objectid
< (u64
)-1) {
2146 * 0: all items from this leaf copied, continue with next
2147 * 1: * more items can be copied, but unused buffer is too small
2148 * * all items were found
2149 * Either way, it will stops the loop which iterates to the next
2151 * -EOVERFLOW: item was to large for buffer
2152 * -EFAULT: could not copy extent buffer back to userspace
2157 static noinline
int search_ioctl(struct inode
*inode
,
2158 struct btrfs_ioctl_search_key
*sk
,
2162 struct btrfs_fs_info
*info
= btrfs_sb(inode
->i_sb
);
2163 struct btrfs_root
*root
;
2164 struct btrfs_key key
;
2165 struct btrfs_path
*path
;
2168 unsigned long sk_offset
= 0;
2170 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
2171 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
2175 path
= btrfs_alloc_path();
2179 if (sk
->tree_id
== 0) {
2180 /* search the root of the inode that was passed */
2181 root
= BTRFS_I(inode
)->root
;
2183 key
.objectid
= sk
->tree_id
;
2184 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2185 key
.offset
= (u64
)-1;
2186 root
= btrfs_read_fs_root_no_name(info
, &key
);
2188 btrfs_free_path(path
);
2189 return PTR_ERR(root
);
2193 key
.objectid
= sk
->min_objectid
;
2194 key
.type
= sk
->min_type
;
2195 key
.offset
= sk
->min_offset
;
2198 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
2204 ret
= copy_to_sk(path
, &key
, sk
, buf_size
, ubuf
,
2205 &sk_offset
, &num_found
);
2206 btrfs_release_path(path
);
2214 sk
->nr_items
= num_found
;
2215 btrfs_free_path(path
);
2219 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
2222 struct btrfs_ioctl_search_args __user
*uargs
;
2223 struct btrfs_ioctl_search_key sk
;
2224 struct inode
*inode
;
2228 if (!capable(CAP_SYS_ADMIN
))
2231 uargs
= (struct btrfs_ioctl_search_args __user
*)argp
;
2233 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
2236 buf_size
= sizeof(uargs
->buf
);
2238 inode
= file_inode(file
);
2239 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
2242 * In the origin implementation an overflow is handled by returning a
2243 * search header with a len of zero, so reset ret.
2245 if (ret
== -EOVERFLOW
)
2248 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
2253 static noinline
int btrfs_ioctl_tree_search_v2(struct file
*file
,
2256 struct btrfs_ioctl_search_args_v2 __user
*uarg
;
2257 struct btrfs_ioctl_search_args_v2 args
;
2258 struct inode
*inode
;
2261 const size_t buf_limit
= SZ_16M
;
2263 if (!capable(CAP_SYS_ADMIN
))
2266 /* copy search header and buffer size */
2267 uarg
= (struct btrfs_ioctl_search_args_v2 __user
*)argp
;
2268 if (copy_from_user(&args
, uarg
, sizeof(args
)))
2271 buf_size
= args
.buf_size
;
2273 /* limit result size to 16MB */
2274 if (buf_size
> buf_limit
)
2275 buf_size
= buf_limit
;
2277 inode
= file_inode(file
);
2278 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
2279 (char __user
*)(&uarg
->buf
[0]));
2280 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
2282 else if (ret
== -EOVERFLOW
&&
2283 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
2290 * Search INODE_REFs to identify path name of 'dirid' directory
2291 * in a 'tree_id' tree. and sets path name to 'name'.
2293 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
2294 u64 tree_id
, u64 dirid
, char *name
)
2296 struct btrfs_root
*root
;
2297 struct btrfs_key key
;
2303 struct btrfs_inode_ref
*iref
;
2304 struct extent_buffer
*l
;
2305 struct btrfs_path
*path
;
2307 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
2312 path
= btrfs_alloc_path();
2316 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
- 1];
2318 key
.objectid
= tree_id
;
2319 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2320 key
.offset
= (u64
)-1;
2321 root
= btrfs_read_fs_root_no_name(info
, &key
);
2323 ret
= PTR_ERR(root
);
2327 key
.objectid
= dirid
;
2328 key
.type
= BTRFS_INODE_REF_KEY
;
2329 key
.offset
= (u64
)-1;
2332 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2336 ret
= btrfs_previous_item(root
, path
, dirid
,
2337 BTRFS_INODE_REF_KEY
);
2347 slot
= path
->slots
[0];
2348 btrfs_item_key_to_cpu(l
, &key
, slot
);
2350 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2351 len
= btrfs_inode_ref_name_len(l
, iref
);
2353 total_len
+= len
+ 1;
2355 ret
= -ENAMETOOLONG
;
2360 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
2362 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2365 btrfs_release_path(path
);
2366 key
.objectid
= key
.offset
;
2367 key
.offset
= (u64
)-1;
2368 dirid
= key
.objectid
;
2370 memmove(name
, ptr
, total_len
);
2371 name
[total_len
] = '\0';
2374 btrfs_free_path(path
);
2378 static int btrfs_search_path_in_tree_user(struct inode
*inode
,
2379 struct btrfs_ioctl_ino_lookup_user_args
*args
)
2381 struct btrfs_fs_info
*fs_info
= BTRFS_I(inode
)->root
->fs_info
;
2382 struct super_block
*sb
= inode
->i_sb
;
2383 struct btrfs_key upper_limit
= BTRFS_I(inode
)->location
;
2384 u64 treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2385 u64 dirid
= args
->dirid
;
2386 unsigned long item_off
;
2387 unsigned long item_len
;
2388 struct btrfs_inode_ref
*iref
;
2389 struct btrfs_root_ref
*rref
;
2390 struct btrfs_root
*root
;
2391 struct btrfs_path
*path
;
2392 struct btrfs_key key
, key2
;
2393 struct extent_buffer
*leaf
;
2394 struct inode
*temp_inode
;
2401 path
= btrfs_alloc_path();
2406 * If the bottom subvolume does not exist directly under upper_limit,
2407 * construct the path in from the bottom up.
2409 if (dirid
!= upper_limit
.objectid
) {
2410 ptr
= &args
->path
[BTRFS_INO_LOOKUP_USER_PATH_MAX
- 1];
2412 key
.objectid
= treeid
;
2413 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2414 key
.offset
= (u64
)-1;
2415 root
= btrfs_read_fs_root_no_name(fs_info
, &key
);
2417 ret
= PTR_ERR(root
);
2421 key
.objectid
= dirid
;
2422 key
.type
= BTRFS_INODE_REF_KEY
;
2423 key
.offset
= (u64
)-1;
2425 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2428 } else if (ret
> 0) {
2429 ret
= btrfs_previous_item(root
, path
, dirid
,
2430 BTRFS_INODE_REF_KEY
);
2433 } else if (ret
> 0) {
2439 leaf
= path
->nodes
[0];
2440 slot
= path
->slots
[0];
2441 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2443 iref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_inode_ref
);
2444 len
= btrfs_inode_ref_name_len(leaf
, iref
);
2446 total_len
+= len
+ 1;
2447 if (ptr
< args
->path
) {
2448 ret
= -ENAMETOOLONG
;
2453 read_extent_buffer(leaf
, ptr
,
2454 (unsigned long)(iref
+ 1), len
);
2456 /* Check the read+exec permission of this directory */
2457 ret
= btrfs_previous_item(root
, path
, dirid
,
2458 BTRFS_INODE_ITEM_KEY
);
2461 } else if (ret
> 0) {
2466 leaf
= path
->nodes
[0];
2467 slot
= path
->slots
[0];
2468 btrfs_item_key_to_cpu(leaf
, &key2
, slot
);
2469 if (key2
.objectid
!= dirid
) {
2474 temp_inode
= btrfs_iget(sb
, &key2
, root
, NULL
);
2475 if (IS_ERR(temp_inode
)) {
2476 ret
= PTR_ERR(temp_inode
);
2479 ret
= inode_permission(temp_inode
, MAY_READ
| MAY_EXEC
);
2486 if (key
.offset
== upper_limit
.objectid
)
2488 if (key
.objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2493 btrfs_release_path(path
);
2494 key
.objectid
= key
.offset
;
2495 key
.offset
= (u64
)-1;
2496 dirid
= key
.objectid
;
2499 memmove(args
->path
, ptr
, total_len
);
2500 args
->path
[total_len
] = '\0';
2501 btrfs_release_path(path
);
2504 /* Get the bottom subvolume's name from ROOT_REF */
2505 root
= fs_info
->tree_root
;
2506 key
.objectid
= treeid
;
2507 key
.type
= BTRFS_ROOT_REF_KEY
;
2508 key
.offset
= args
->treeid
;
2509 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2512 } else if (ret
> 0) {
2517 leaf
= path
->nodes
[0];
2518 slot
= path
->slots
[0];
2519 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2521 item_off
= btrfs_item_ptr_offset(leaf
, slot
);
2522 item_len
= btrfs_item_size_nr(leaf
, slot
);
2523 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2524 rref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_root_ref
);
2525 if (args
->dirid
!= btrfs_root_ref_dirid(leaf
, rref
)) {
2530 /* Copy subvolume's name */
2531 item_off
+= sizeof(struct btrfs_root_ref
);
2532 item_len
-= sizeof(struct btrfs_root_ref
);
2533 read_extent_buffer(leaf
, args
->name
, item_off
, item_len
);
2534 args
->name
[item_len
] = 0;
2537 btrfs_free_path(path
);
2541 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2544 struct btrfs_ioctl_ino_lookup_args
*args
;
2545 struct inode
*inode
;
2548 args
= memdup_user(argp
, sizeof(*args
));
2550 return PTR_ERR(args
);
2552 inode
= file_inode(file
);
2555 * Unprivileged query to obtain the containing subvolume root id. The
2556 * path is reset so it's consistent with btrfs_search_path_in_tree.
2558 if (args
->treeid
== 0)
2559 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2561 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2566 if (!capable(CAP_SYS_ADMIN
)) {
2571 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2572 args
->treeid
, args
->objectid
,
2576 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2584 * Version of ino_lookup ioctl (unprivileged)
2586 * The main differences from ino_lookup ioctl are:
2588 * 1. Read + Exec permission will be checked using inode_permission() during
2589 * path construction. -EACCES will be returned in case of failure.
2590 * 2. Path construction will be stopped at the inode number which corresponds
2591 * to the fd with which this ioctl is called. If constructed path does not
2592 * exist under fd's inode, -EACCES will be returned.
2593 * 3. The name of bottom subvolume is also searched and filled.
2595 static int btrfs_ioctl_ino_lookup_user(struct file
*file
, void __user
*argp
)
2597 struct btrfs_ioctl_ino_lookup_user_args
*args
;
2598 struct inode
*inode
;
2601 args
= memdup_user(argp
, sizeof(*args
));
2603 return PTR_ERR(args
);
2605 inode
= file_inode(file
);
2607 if (args
->dirid
== BTRFS_FIRST_FREE_OBJECTID
&&
2608 BTRFS_I(inode
)->location
.objectid
!= BTRFS_FIRST_FREE_OBJECTID
) {
2610 * The subvolume does not exist under fd with which this is
2617 ret
= btrfs_search_path_in_tree_user(inode
, args
);
2619 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2626 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2627 static int btrfs_ioctl_get_subvol_info(struct file
*file
, void __user
*argp
)
2629 struct btrfs_ioctl_get_subvol_info_args
*subvol_info
;
2630 struct btrfs_fs_info
*fs_info
;
2631 struct btrfs_root
*root
;
2632 struct btrfs_path
*path
;
2633 struct btrfs_key key
;
2634 struct btrfs_root_item
*root_item
;
2635 struct btrfs_root_ref
*rref
;
2636 struct extent_buffer
*leaf
;
2637 unsigned long item_off
;
2638 unsigned long item_len
;
2639 struct inode
*inode
;
2643 path
= btrfs_alloc_path();
2647 subvol_info
= kzalloc(sizeof(*subvol_info
), GFP_KERNEL
);
2649 btrfs_free_path(path
);
2653 inode
= file_inode(file
);
2654 fs_info
= BTRFS_I(inode
)->root
->fs_info
;
2656 /* Get root_item of inode's subvolume */
2657 key
.objectid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2658 key
.type
= BTRFS_ROOT_ITEM_KEY
;
2659 key
.offset
= (u64
)-1;
2660 root
= btrfs_read_fs_root_no_name(fs_info
, &key
);
2662 ret
= PTR_ERR(root
);
2665 root_item
= &root
->root_item
;
2667 subvol_info
->treeid
= key
.objectid
;
2669 subvol_info
->generation
= btrfs_root_generation(root_item
);
2670 subvol_info
->flags
= btrfs_root_flags(root_item
);
2672 memcpy(subvol_info
->uuid
, root_item
->uuid
, BTRFS_UUID_SIZE
);
2673 memcpy(subvol_info
->parent_uuid
, root_item
->parent_uuid
,
2675 memcpy(subvol_info
->received_uuid
, root_item
->received_uuid
,
2678 subvol_info
->ctransid
= btrfs_root_ctransid(root_item
);
2679 subvol_info
->ctime
.sec
= btrfs_stack_timespec_sec(&root_item
->ctime
);
2680 subvol_info
->ctime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->ctime
);
2682 subvol_info
->otransid
= btrfs_root_otransid(root_item
);
2683 subvol_info
->otime
.sec
= btrfs_stack_timespec_sec(&root_item
->otime
);
2684 subvol_info
->otime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->otime
);
2686 subvol_info
->stransid
= btrfs_root_stransid(root_item
);
2687 subvol_info
->stime
.sec
= btrfs_stack_timespec_sec(&root_item
->stime
);
2688 subvol_info
->stime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->stime
);
2690 subvol_info
->rtransid
= btrfs_root_rtransid(root_item
);
2691 subvol_info
->rtime
.sec
= btrfs_stack_timespec_sec(&root_item
->rtime
);
2692 subvol_info
->rtime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->rtime
);
2694 if (key
.objectid
!= BTRFS_FS_TREE_OBJECTID
) {
2695 /* Search root tree for ROOT_BACKREF of this subvolume */
2696 root
= fs_info
->tree_root
;
2698 key
.type
= BTRFS_ROOT_BACKREF_KEY
;
2700 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2703 } else if (path
->slots
[0] >=
2704 btrfs_header_nritems(path
->nodes
[0])) {
2705 ret
= btrfs_next_leaf(root
, path
);
2708 } else if (ret
> 0) {
2714 leaf
= path
->nodes
[0];
2715 slot
= path
->slots
[0];
2716 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2717 if (key
.objectid
== subvol_info
->treeid
&&
2718 key
.type
== BTRFS_ROOT_BACKREF_KEY
) {
2719 subvol_info
->parent_id
= key
.offset
;
2721 rref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_root_ref
);
2722 subvol_info
->dirid
= btrfs_root_ref_dirid(leaf
, rref
);
2724 item_off
= btrfs_item_ptr_offset(leaf
, slot
)
2725 + sizeof(struct btrfs_root_ref
);
2726 item_len
= btrfs_item_size_nr(leaf
, slot
)
2727 - sizeof(struct btrfs_root_ref
);
2728 read_extent_buffer(leaf
, subvol_info
->name
,
2729 item_off
, item_len
);
2736 if (copy_to_user(argp
, subvol_info
, sizeof(*subvol_info
)))
2740 btrfs_free_path(path
);
2741 kzfree(subvol_info
);
2746 * Return ROOT_REF information of the subvolume containing this inode
2747 * except the subvolume name.
2749 static int btrfs_ioctl_get_subvol_rootref(struct file
*file
, void __user
*argp
)
2751 struct btrfs_ioctl_get_subvol_rootref_args
*rootrefs
;
2752 struct btrfs_root_ref
*rref
;
2753 struct btrfs_root
*root
;
2754 struct btrfs_path
*path
;
2755 struct btrfs_key key
;
2756 struct extent_buffer
*leaf
;
2757 struct inode
*inode
;
2763 path
= btrfs_alloc_path();
2767 rootrefs
= memdup_user(argp
, sizeof(*rootrefs
));
2768 if (IS_ERR(rootrefs
)) {
2769 btrfs_free_path(path
);
2770 return PTR_ERR(rootrefs
);
2773 inode
= file_inode(file
);
2774 root
= BTRFS_I(inode
)->root
->fs_info
->tree_root
;
2775 objectid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2777 key
.objectid
= objectid
;
2778 key
.type
= BTRFS_ROOT_REF_KEY
;
2779 key
.offset
= rootrefs
->min_treeid
;
2782 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2785 } else if (path
->slots
[0] >=
2786 btrfs_header_nritems(path
->nodes
[0])) {
2787 ret
= btrfs_next_leaf(root
, path
);
2790 } else if (ret
> 0) {
2796 leaf
= path
->nodes
[0];
2797 slot
= path
->slots
[0];
2799 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2800 if (key
.objectid
!= objectid
|| key
.type
!= BTRFS_ROOT_REF_KEY
) {
2805 if (found
== BTRFS_MAX_ROOTREF_BUFFER_NUM
) {
2810 rref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_root_ref
);
2811 rootrefs
->rootref
[found
].treeid
= key
.offset
;
2812 rootrefs
->rootref
[found
].dirid
=
2813 btrfs_root_ref_dirid(leaf
, rref
);
2816 ret
= btrfs_next_item(root
, path
);
2819 } else if (ret
> 0) {
2826 if (!ret
|| ret
== -EOVERFLOW
) {
2827 rootrefs
->num_items
= found
;
2828 /* update min_treeid for next search */
2830 rootrefs
->min_treeid
=
2831 rootrefs
->rootref
[found
- 1].treeid
+ 1;
2832 if (copy_to_user(argp
, rootrefs
, sizeof(*rootrefs
)))
2837 btrfs_free_path(path
);
2842 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2845 struct dentry
*parent
= file
->f_path
.dentry
;
2846 struct btrfs_fs_info
*fs_info
= btrfs_sb(parent
->d_sb
);
2847 struct dentry
*dentry
;
2848 struct inode
*dir
= d_inode(parent
);
2849 struct inode
*inode
;
2850 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2851 struct btrfs_root
*dest
= NULL
;
2852 struct btrfs_ioctl_vol_args
*vol_args
;
2856 if (!S_ISDIR(dir
->i_mode
))
2859 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2860 if (IS_ERR(vol_args
))
2861 return PTR_ERR(vol_args
);
2863 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2864 namelen
= strlen(vol_args
->name
);
2865 if (strchr(vol_args
->name
, '/') ||
2866 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2871 err
= mnt_want_write_file(file
);
2876 err
= down_write_killable_nested(&dir
->i_rwsem
, I_MUTEX_PARENT
);
2878 goto out_drop_write
;
2879 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2880 if (IS_ERR(dentry
)) {
2881 err
= PTR_ERR(dentry
);
2882 goto out_unlock_dir
;
2885 if (d_really_is_negative(dentry
)) {
2890 inode
= d_inode(dentry
);
2891 dest
= BTRFS_I(inode
)->root
;
2892 if (!capable(CAP_SYS_ADMIN
)) {
2894 * Regular user. Only allow this with a special mount
2895 * option, when the user has write+exec access to the
2896 * subvol root, and when rmdir(2) would have been
2899 * Note that this is _not_ check that the subvol is
2900 * empty or doesn't contain data that we wouldn't
2901 * otherwise be able to delete.
2903 * Users who want to delete empty subvols should try
2907 if (!btrfs_test_opt(fs_info
, USER_SUBVOL_RM_ALLOWED
))
2911 * Do not allow deletion if the parent dir is the same
2912 * as the dir to be deleted. That means the ioctl
2913 * must be called on the dentry referencing the root
2914 * of the subvol, not a random directory contained
2921 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2926 /* check if subvolume may be deleted by a user */
2927 err
= btrfs_may_delete(dir
, dentry
, 1);
2931 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
) {
2937 err
= btrfs_delete_subvolume(dir
, dentry
);
2938 inode_unlock(inode
);
2947 mnt_drop_write_file(file
);
2953 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2955 struct inode
*inode
= file_inode(file
);
2956 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2957 struct btrfs_ioctl_defrag_range_args
*range
;
2960 ret
= mnt_want_write_file(file
);
2964 if (btrfs_root_readonly(root
)) {
2969 switch (inode
->i_mode
& S_IFMT
) {
2971 if (!capable(CAP_SYS_ADMIN
)) {
2975 ret
= btrfs_defrag_root(root
);
2979 * Note that this does not check the file descriptor for write
2980 * access. This prevents defragmenting executables that are
2981 * running and allows defrag on files open in read-only mode.
2983 if (!capable(CAP_SYS_ADMIN
) &&
2984 inode_permission(inode
, MAY_WRITE
)) {
2989 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2996 if (copy_from_user(range
, argp
,
3002 /* compression requires us to start the IO */
3003 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
3004 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
3005 range
->extent_thresh
= (u32
)-1;
3008 /* the rest are all set to zero by kzalloc */
3009 range
->len
= (u64
)-1;
3011 ret
= btrfs_defrag_file(file_inode(file
), file
,
3012 range
, BTRFS_OLDEST_GENERATION
, 0);
3021 mnt_drop_write_file(file
);
3025 static long btrfs_ioctl_add_dev(struct btrfs_fs_info
*fs_info
, void __user
*arg
)
3027 struct btrfs_ioctl_vol_args
*vol_args
;
3030 if (!capable(CAP_SYS_ADMIN
))
3033 if (test_and_set_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
))
3034 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
3036 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
3037 if (IS_ERR(vol_args
)) {
3038 ret
= PTR_ERR(vol_args
);
3042 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
3043 ret
= btrfs_init_new_device(fs_info
, vol_args
->name
);
3046 btrfs_info(fs_info
, "disk added %s", vol_args
->name
);
3050 clear_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
);
3054 static long btrfs_ioctl_rm_dev_v2(struct file
*file
, void __user
*arg
)
3056 struct inode
*inode
= file_inode(file
);
3057 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3058 struct btrfs_ioctl_vol_args_v2
*vol_args
;
3061 if (!capable(CAP_SYS_ADMIN
))
3064 ret
= mnt_want_write_file(file
);
3068 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
3069 if (IS_ERR(vol_args
)) {
3070 ret
= PTR_ERR(vol_args
);
3074 /* Check for compatibility reject unknown flags */
3075 if (vol_args
->flags
& ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED
) {
3080 if (test_and_set_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
)) {
3081 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
3085 if (vol_args
->flags
& BTRFS_DEVICE_SPEC_BY_ID
) {
3086 ret
= btrfs_rm_device(fs_info
, NULL
, vol_args
->devid
);
3088 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
3089 ret
= btrfs_rm_device(fs_info
, vol_args
->name
, 0);
3091 clear_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
);
3094 if (vol_args
->flags
& BTRFS_DEVICE_SPEC_BY_ID
)
3095 btrfs_info(fs_info
, "device deleted: id %llu",
3098 btrfs_info(fs_info
, "device deleted: %s",
3104 mnt_drop_write_file(file
);
3108 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
3110 struct inode
*inode
= file_inode(file
);
3111 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3112 struct btrfs_ioctl_vol_args
*vol_args
;
3115 if (!capable(CAP_SYS_ADMIN
))
3118 ret
= mnt_want_write_file(file
);
3122 if (test_and_set_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
)) {
3123 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
3124 goto out_drop_write
;
3127 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
3128 if (IS_ERR(vol_args
)) {
3129 ret
= PTR_ERR(vol_args
);
3133 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
3134 ret
= btrfs_rm_device(fs_info
, vol_args
->name
, 0);
3137 btrfs_info(fs_info
, "disk deleted %s", vol_args
->name
);
3140 clear_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
);
3142 mnt_drop_write_file(file
);
3147 static long btrfs_ioctl_fs_info(struct btrfs_fs_info
*fs_info
,
3150 struct btrfs_ioctl_fs_info_args
*fi_args
;
3151 struct btrfs_device
*device
;
3152 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
3155 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
3160 fi_args
->num_devices
= fs_devices
->num_devices
;
3162 list_for_each_entry_rcu(device
, &fs_devices
->devices
, dev_list
) {
3163 if (device
->devid
> fi_args
->max_id
)
3164 fi_args
->max_id
= device
->devid
;
3168 memcpy(&fi_args
->fsid
, fs_devices
->fsid
, sizeof(fi_args
->fsid
));
3169 fi_args
->nodesize
= fs_info
->nodesize
;
3170 fi_args
->sectorsize
= fs_info
->sectorsize
;
3171 fi_args
->clone_alignment
= fs_info
->sectorsize
;
3173 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
3180 static long btrfs_ioctl_dev_info(struct btrfs_fs_info
*fs_info
,
3183 struct btrfs_ioctl_dev_info_args
*di_args
;
3184 struct btrfs_device
*dev
;
3186 char *s_uuid
= NULL
;
3188 di_args
= memdup_user(arg
, sizeof(*di_args
));
3189 if (IS_ERR(di_args
))
3190 return PTR_ERR(di_args
);
3192 if (!btrfs_is_empty_uuid(di_args
->uuid
))
3193 s_uuid
= di_args
->uuid
;
3196 dev
= btrfs_find_device(fs_info
->fs_devices
, di_args
->devid
, s_uuid
,
3204 di_args
->devid
= dev
->devid
;
3205 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
3206 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
3207 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
3209 strncpy(di_args
->path
, rcu_str_deref(dev
->name
),
3210 sizeof(di_args
->path
) - 1);
3211 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
3213 di_args
->path
[0] = '\0';
3218 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
3225 static void btrfs_double_extent_unlock(struct inode
*inode1
, u64 loff1
,
3226 struct inode
*inode2
, u64 loff2
, u64 len
)
3228 unlock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
3229 unlock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
3232 static void btrfs_double_extent_lock(struct inode
*inode1
, u64 loff1
,
3233 struct inode
*inode2
, u64 loff2
, u64 len
)
3235 if (inode1
< inode2
) {
3236 swap(inode1
, inode2
);
3238 } else if (inode1
== inode2
&& loff2
< loff1
) {
3241 lock_extent(&BTRFS_I(inode1
)->io_tree
, loff1
, loff1
+ len
- 1);
3242 lock_extent(&BTRFS_I(inode2
)->io_tree
, loff2
, loff2
+ len
- 1);
3245 static int btrfs_extent_same_range(struct inode
*src
, u64 loff
, u64 len
,
3246 struct inode
*dst
, u64 dst_loff
)
3251 * Lock destination range to serialize with concurrent readpages() and
3252 * source range to serialize with relocation.
3254 btrfs_double_extent_lock(src
, loff
, dst
, dst_loff
, len
);
3255 ret
= btrfs_clone(src
, dst
, loff
, len
, len
, dst_loff
, 1);
3256 btrfs_double_extent_unlock(src
, loff
, dst
, dst_loff
, len
);
3261 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3263 static int btrfs_extent_same(struct inode
*src
, u64 loff
, u64 olen
,
3264 struct inode
*dst
, u64 dst_loff
)
3267 u64 i
, tail_len
, chunk_count
;
3268 struct btrfs_root
*root_dst
= BTRFS_I(dst
)->root
;
3270 spin_lock(&root_dst
->root_item_lock
);
3271 if (root_dst
->send_in_progress
) {
3272 btrfs_warn_rl(root_dst
->fs_info
,
3273 "cannot deduplicate to root %llu while send operations are using it (%d in progress)",
3274 root_dst
->root_key
.objectid
,
3275 root_dst
->send_in_progress
);
3276 spin_unlock(&root_dst
->root_item_lock
);
3279 root_dst
->dedupe_in_progress
++;
3280 spin_unlock(&root_dst
->root_item_lock
);
3282 tail_len
= olen
% BTRFS_MAX_DEDUPE_LEN
;
3283 chunk_count
= div_u64(olen
, BTRFS_MAX_DEDUPE_LEN
);
3285 for (i
= 0; i
< chunk_count
; i
++) {
3286 ret
= btrfs_extent_same_range(src
, loff
, BTRFS_MAX_DEDUPE_LEN
,
3291 loff
+= BTRFS_MAX_DEDUPE_LEN
;
3292 dst_loff
+= BTRFS_MAX_DEDUPE_LEN
;
3296 ret
= btrfs_extent_same_range(src
, loff
, tail_len
, dst
,
3299 spin_lock(&root_dst
->root_item_lock
);
3300 root_dst
->dedupe_in_progress
--;
3301 spin_unlock(&root_dst
->root_item_lock
);
3306 static int clone_finish_inode_update(struct btrfs_trans_handle
*trans
,
3307 struct inode
*inode
,
3313 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3316 inode_inc_iversion(inode
);
3317 if (!no_time_update
)
3318 inode
->i_mtime
= inode
->i_ctime
= current_time(inode
);
3320 * We round up to the block size at eof when determining which
3321 * extents to clone above, but shouldn't round up the file size.
3323 if (endoff
> destoff
+ olen
)
3324 endoff
= destoff
+ olen
;
3325 if (endoff
> inode
->i_size
)
3326 btrfs_i_size_write(BTRFS_I(inode
), endoff
);
3328 ret
= btrfs_update_inode(trans
, root
, inode
);
3330 btrfs_abort_transaction(trans
, ret
);
3331 btrfs_end_transaction(trans
);
3334 ret
= btrfs_end_transaction(trans
);
3339 static void clone_update_extent_map(struct btrfs_inode
*inode
,
3340 const struct btrfs_trans_handle
*trans
,
3341 const struct btrfs_path
*path
,
3342 const u64 hole_offset
,
3345 struct extent_map_tree
*em_tree
= &inode
->extent_tree
;
3346 struct extent_map
*em
;
3349 em
= alloc_extent_map();
3351 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
, &inode
->runtime_flags
);
3356 struct btrfs_file_extent_item
*fi
;
3358 fi
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3359 struct btrfs_file_extent_item
);
3360 btrfs_extent_item_to_extent_map(inode
, path
, fi
, false, em
);
3361 em
->generation
= -1;
3362 if (btrfs_file_extent_type(path
->nodes
[0], fi
) ==
3363 BTRFS_FILE_EXTENT_INLINE
)
3364 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
,
3365 &inode
->runtime_flags
);
3367 em
->start
= hole_offset
;
3369 em
->ram_bytes
= em
->len
;
3370 em
->orig_start
= hole_offset
;
3371 em
->block_start
= EXTENT_MAP_HOLE
;
3373 em
->orig_block_len
= 0;
3374 em
->compress_type
= BTRFS_COMPRESS_NONE
;
3375 em
->generation
= trans
->transid
;
3379 write_lock(&em_tree
->lock
);
3380 ret
= add_extent_mapping(em_tree
, em
, 1);
3381 write_unlock(&em_tree
->lock
);
3382 if (ret
!= -EEXIST
) {
3383 free_extent_map(em
);
3386 btrfs_drop_extent_cache(inode
, em
->start
,
3387 em
->start
+ em
->len
- 1, 0);
3391 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC
, &inode
->runtime_flags
);
3395 * Make sure we do not end up inserting an inline extent into a file that has
3396 * already other (non-inline) extents. If a file has an inline extent it can
3397 * not have any other extents and the (single) inline extent must start at the
3398 * file offset 0. Failing to respect these rules will lead to file corruption,
3399 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3401 * We can have extents that have been already written to disk or we can have
3402 * dirty ranges still in delalloc, in which case the extent maps and items are
3403 * created only when we run delalloc, and the delalloc ranges might fall outside
3404 * the range we are currently locking in the inode's io tree. So we check the
3405 * inode's i_size because of that (i_size updates are done while holding the
3406 * i_mutex, which we are holding here).
3407 * We also check to see if the inode has a size not greater than "datal" but has
3408 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3409 * protected against such concurrent fallocate calls by the i_mutex).
3411 * If the file has no extents but a size greater than datal, do not allow the
3412 * copy because we would need turn the inline extent into a non-inline one (even
3413 * with NO_HOLES enabled). If we find our destination inode only has one inline
3414 * extent, just overwrite it with the source inline extent if its size is less
3415 * than the source extent's size, or we could copy the source inline extent's
3416 * data into the destination inode's inline extent if the later is greater then
3419 static int clone_copy_inline_extent(struct inode
*dst
,
3420 struct btrfs_trans_handle
*trans
,
3421 struct btrfs_path
*path
,
3422 struct btrfs_key
*new_key
,
3423 const u64 drop_start
,
3429 struct btrfs_fs_info
*fs_info
= btrfs_sb(dst
->i_sb
);
3430 struct btrfs_root
*root
= BTRFS_I(dst
)->root
;
3431 const u64 aligned_end
= ALIGN(new_key
->offset
+ datal
,
3432 fs_info
->sectorsize
);
3434 struct btrfs_key key
;
3436 if (new_key
->offset
> 0)
3439 key
.objectid
= btrfs_ino(BTRFS_I(dst
));
3440 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3442 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
3445 } else if (ret
> 0) {
3446 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
3447 ret
= btrfs_next_leaf(root
, path
);
3451 goto copy_inline_extent
;
3453 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
3454 if (key
.objectid
== btrfs_ino(BTRFS_I(dst
)) &&
3455 key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3456 ASSERT(key
.offset
> 0);
3459 } else if (i_size_read(dst
) <= datal
) {
3460 struct btrfs_file_extent_item
*ei
;
3464 * If the file size is <= datal, make sure there are no other
3465 * extents following (can happen do to an fallocate call with
3466 * the flag FALLOC_FL_KEEP_SIZE).
3468 ei
= btrfs_item_ptr(path
->nodes
[0], path
->slots
[0],
3469 struct btrfs_file_extent_item
);
3471 * If it's an inline extent, it can not have other extents
3474 if (btrfs_file_extent_type(path
->nodes
[0], ei
) ==
3475 BTRFS_FILE_EXTENT_INLINE
)
3476 goto copy_inline_extent
;
3478 ext_len
= btrfs_file_extent_num_bytes(path
->nodes
[0], ei
);
3479 if (ext_len
> aligned_end
)
3482 ret
= btrfs_next_item(root
, path
);
3485 } else if (ret
== 0) {
3486 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3488 if (key
.objectid
== btrfs_ino(BTRFS_I(dst
)) &&
3489 key
.type
== BTRFS_EXTENT_DATA_KEY
)
3496 * We have no extent items, or we have an extent at offset 0 which may
3497 * or may not be inlined. All these cases are dealt the same way.
3499 if (i_size_read(dst
) > datal
) {
3501 * If the destination inode has an inline extent...
3502 * This would require copying the data from the source inline
3503 * extent into the beginning of the destination's inline extent.
3504 * But this is really complex, both extents can be compressed
3505 * or just one of them, which would require decompressing and
3506 * re-compressing data (which could increase the new compressed
3507 * size, not allowing the compressed data to fit anymore in an
3509 * So just don't support this case for now (it should be rare,
3510 * we are not really saving space when cloning inline extents).
3515 btrfs_release_path(path
);
3516 ret
= btrfs_drop_extents(trans
, root
, dst
, drop_start
, aligned_end
, 1);
3519 ret
= btrfs_insert_empty_item(trans
, root
, path
, new_key
, size
);
3524 const u32 start
= btrfs_file_extent_calc_inline_size(0);
3526 memmove(inline_data
+ start
, inline_data
+ start
+ skip
, datal
);
3529 write_extent_buffer(path
->nodes
[0], inline_data
,
3530 btrfs_item_ptr_offset(path
->nodes
[0],
3533 inode_add_bytes(dst
, datal
);
3539 * btrfs_clone() - clone a range from inode file to another
3541 * @src: Inode to clone from
3542 * @inode: Inode to clone to
3543 * @off: Offset within source to start clone from
3544 * @olen: Original length, passed by user, of range to clone
3545 * @olen_aligned: Block-aligned value of olen
3546 * @destoff: Offset within @inode to start clone
3547 * @no_time_update: Whether to update mtime/ctime on the target inode
3549 static int btrfs_clone(struct inode
*src
, struct inode
*inode
,
3550 const u64 off
, const u64 olen
, const u64 olen_aligned
,
3551 const u64 destoff
, int no_time_update
)
3553 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3554 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3555 struct btrfs_path
*path
= NULL
;
3556 struct extent_buffer
*leaf
;
3557 struct btrfs_trans_handle
*trans
;
3559 struct btrfs_key key
;
3563 const u64 len
= olen_aligned
;
3564 u64 last_dest_end
= destoff
;
3567 buf
= kvmalloc(fs_info
->nodesize
, GFP_KERNEL
);
3571 path
= btrfs_alloc_path();
3577 path
->reada
= READA_FORWARD
;
3579 key
.objectid
= btrfs_ino(BTRFS_I(src
));
3580 key
.type
= BTRFS_EXTENT_DATA_KEY
;
3584 u64 next_key_min_offset
= key
.offset
+ 1;
3587 * note the key will change type as we walk through the
3590 path
->leave_spinning
= 1;
3591 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
3596 * First search, if no extent item that starts at offset off was
3597 * found but the previous item is an extent item, it's possible
3598 * it might overlap our target range, therefore process it.
3600 if (key
.offset
== off
&& ret
> 0 && path
->slots
[0] > 0) {
3601 btrfs_item_key_to_cpu(path
->nodes
[0], &key
,
3602 path
->slots
[0] - 1);
3603 if (key
.type
== BTRFS_EXTENT_DATA_KEY
)
3607 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3609 if (path
->slots
[0] >= nritems
) {
3610 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
3615 nritems
= btrfs_header_nritems(path
->nodes
[0]);
3617 leaf
= path
->nodes
[0];
3618 slot
= path
->slots
[0];
3620 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
3621 if (key
.type
> BTRFS_EXTENT_DATA_KEY
||
3622 key
.objectid
!= btrfs_ino(BTRFS_I(src
)))
3625 if (key
.type
== BTRFS_EXTENT_DATA_KEY
) {
3626 struct btrfs_file_extent_item
*extent
;
3629 struct btrfs_key new_key
;
3630 u64 disko
= 0, diskl
= 0;
3631 u64 datao
= 0, datal
= 0;
3635 extent
= btrfs_item_ptr(leaf
, slot
,
3636 struct btrfs_file_extent_item
);
3637 comp
= btrfs_file_extent_compression(leaf
, extent
);
3638 type
= btrfs_file_extent_type(leaf
, extent
);
3639 if (type
== BTRFS_FILE_EXTENT_REG
||
3640 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3641 disko
= btrfs_file_extent_disk_bytenr(leaf
,
3643 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
3645 datao
= btrfs_file_extent_offset(leaf
, extent
);
3646 datal
= btrfs_file_extent_num_bytes(leaf
,
3648 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3649 /* take upper bound, may be compressed */
3650 datal
= btrfs_file_extent_ram_bytes(leaf
,
3655 * The first search might have left us at an extent
3656 * item that ends before our target range's start, can
3657 * happen if we have holes and NO_HOLES feature enabled.
3659 if (key
.offset
+ datal
<= off
) {
3662 } else if (key
.offset
>= off
+ len
) {
3665 next_key_min_offset
= key
.offset
+ datal
;
3666 size
= btrfs_item_size_nr(leaf
, slot
);
3667 read_extent_buffer(leaf
, buf
,
3668 btrfs_item_ptr_offset(leaf
, slot
),
3671 btrfs_release_path(path
);
3672 path
->leave_spinning
= 0;
3674 memcpy(&new_key
, &key
, sizeof(new_key
));
3675 new_key
.objectid
= btrfs_ino(BTRFS_I(inode
));
3676 if (off
<= key
.offset
)
3677 new_key
.offset
= key
.offset
+ destoff
- off
;
3679 new_key
.offset
= destoff
;
3682 * Deal with a hole that doesn't have an extent item
3683 * that represents it (NO_HOLES feature enabled).
3684 * This hole is either in the middle of the cloning
3685 * range or at the beginning (fully overlaps it or
3686 * partially overlaps it).
3688 if (new_key
.offset
!= last_dest_end
)
3689 drop_start
= last_dest_end
;
3691 drop_start
= new_key
.offset
;
3694 * 1 - adjusting old extent (we may have to split it)
3695 * 1 - add new extent
3698 trans
= btrfs_start_transaction(root
, 3);
3699 if (IS_ERR(trans
)) {
3700 ret
= PTR_ERR(trans
);
3704 if (type
== BTRFS_FILE_EXTENT_REG
||
3705 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
3707 * a | --- range to clone ---| b
3708 * | ------------- extent ------------- |
3711 /* subtract range b */
3712 if (key
.offset
+ datal
> off
+ len
)
3713 datal
= off
+ len
- key
.offset
;
3715 /* subtract range a */
3716 if (off
> key
.offset
) {
3717 datao
+= off
- key
.offset
;
3718 datal
-= off
- key
.offset
;
3721 ret
= btrfs_drop_extents(trans
, root
, inode
,
3723 new_key
.offset
+ datal
,
3726 if (ret
!= -EOPNOTSUPP
)
3727 btrfs_abort_transaction(trans
,
3729 btrfs_end_transaction(trans
);
3733 ret
= btrfs_insert_empty_item(trans
, root
, path
,
3736 btrfs_abort_transaction(trans
, ret
);
3737 btrfs_end_transaction(trans
);
3741 leaf
= path
->nodes
[0];
3742 slot
= path
->slots
[0];
3743 write_extent_buffer(leaf
, buf
,
3744 btrfs_item_ptr_offset(leaf
, slot
),
3747 extent
= btrfs_item_ptr(leaf
, slot
,
3748 struct btrfs_file_extent_item
);
3750 /* disko == 0 means it's a hole */
3754 btrfs_set_file_extent_offset(leaf
, extent
,
3756 btrfs_set_file_extent_num_bytes(leaf
, extent
,
3760 struct btrfs_ref ref
= { 0 };
3761 inode_add_bytes(inode
, datal
);
3762 btrfs_init_generic_ref(&ref
,
3763 BTRFS_ADD_DELAYED_REF
, disko
,
3765 btrfs_init_data_ref(&ref
,
3766 root
->root_key
.objectid
,
3767 btrfs_ino(BTRFS_I(inode
)),
3768 new_key
.offset
- datao
);
3769 ret
= btrfs_inc_extent_ref(trans
, &ref
);
3771 btrfs_abort_transaction(trans
,
3773 btrfs_end_transaction(trans
);
3778 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
3782 if (off
> key
.offset
) {
3783 skip
= off
- key
.offset
;
3784 new_key
.offset
+= skip
;
3787 if (key
.offset
+ datal
> off
+ len
)
3788 trim
= key
.offset
+ datal
- (off
+ len
);
3790 if (comp
&& (skip
|| trim
)) {
3792 btrfs_end_transaction(trans
);
3795 size
-= skip
+ trim
;
3796 datal
-= skip
+ trim
;
3798 ret
= clone_copy_inline_extent(inode
,
3805 if (ret
!= -EOPNOTSUPP
)
3806 btrfs_abort_transaction(trans
,
3808 btrfs_end_transaction(trans
);
3811 leaf
= path
->nodes
[0];
3812 slot
= path
->slots
[0];
3815 /* If we have an implicit hole (NO_HOLES feature). */
3816 if (drop_start
< new_key
.offset
)
3817 clone_update_extent_map(BTRFS_I(inode
), trans
,
3819 new_key
.offset
- drop_start
);
3821 clone_update_extent_map(BTRFS_I(inode
), trans
,
3824 btrfs_mark_buffer_dirty(leaf
);
3825 btrfs_release_path(path
);
3827 last_dest_end
= ALIGN(new_key
.offset
+ datal
,
3828 fs_info
->sectorsize
);
3829 ret
= clone_finish_inode_update(trans
, inode
,
3835 if (new_key
.offset
+ datal
>= destoff
+ len
)
3838 btrfs_release_path(path
);
3839 key
.offset
= next_key_min_offset
;
3841 if (fatal_signal_pending(current
)) {
3848 if (last_dest_end
< destoff
+ len
) {
3850 * We have an implicit hole (NO_HOLES feature is enabled) that
3851 * fully or partially overlaps our cloning range at its end.
3853 btrfs_release_path(path
);
3856 * 1 - remove extent(s)
3859 trans
= btrfs_start_transaction(root
, 2);
3860 if (IS_ERR(trans
)) {
3861 ret
= PTR_ERR(trans
);
3864 ret
= btrfs_drop_extents(trans
, root
, inode
,
3865 last_dest_end
, destoff
+ len
, 1);
3867 if (ret
!= -EOPNOTSUPP
)
3868 btrfs_abort_transaction(trans
, ret
);
3869 btrfs_end_transaction(trans
);
3872 clone_update_extent_map(BTRFS_I(inode
), trans
, NULL
,
3874 destoff
+ len
- last_dest_end
);
3875 ret
= clone_finish_inode_update(trans
, inode
, destoff
+ len
,
3876 destoff
, olen
, no_time_update
);
3880 btrfs_free_path(path
);
3885 static noinline
int btrfs_clone_files(struct file
*file
, struct file
*file_src
,
3886 u64 off
, u64 olen
, u64 destoff
)
3888 struct inode
*inode
= file_inode(file
);
3889 struct inode
*src
= file_inode(file_src
);
3890 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3893 u64 bs
= fs_info
->sb
->s_blocksize
;
3897 * - split compressed inline extents. annoying: we need to
3898 * decompress into destination's address_space (the file offset
3899 * may change, so source mapping won't do), then recompress (or
3900 * otherwise reinsert) a subrange.
3902 * - split destination inode's inline extents. The inline extents can
3903 * be either compressed or non-compressed.
3907 * VFS's generic_remap_file_range_prep() protects us from cloning the
3908 * eof block into the middle of a file, which would result in corruption
3909 * if the file size is not blocksize aligned. So we don't need to check
3910 * for that case here.
3912 if (off
+ len
== src
->i_size
)
3913 len
= ALIGN(src
->i_size
, bs
) - off
;
3915 if (destoff
> inode
->i_size
) {
3916 const u64 wb_start
= ALIGN_DOWN(inode
->i_size
, bs
);
3918 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
3922 * We may have truncated the last block if the inode's size is
3923 * not sector size aligned, so we need to wait for writeback to
3924 * complete before proceeding further, otherwise we can race
3925 * with cloning and attempt to increment a reference to an
3926 * extent that no longer exists (writeback completed right after
3927 * we found the previous extent covering eof and before we
3928 * attempted to increment its reference count).
3930 ret
= btrfs_wait_ordered_range(inode
, wb_start
,
3931 destoff
- wb_start
);
3937 * Lock destination range to serialize with concurrent readpages() and
3938 * source range to serialize with relocation.
3940 btrfs_double_extent_lock(src
, off
, inode
, destoff
, len
);
3941 ret
= btrfs_clone(src
, inode
, off
, olen
, len
, destoff
, 0);
3942 btrfs_double_extent_unlock(src
, off
, inode
, destoff
, len
);
3944 * Truncate page cache pages so that future reads will see the cloned
3945 * data immediately and not the previous data.
3947 truncate_inode_pages_range(&inode
->i_data
,
3948 round_down(destoff
, PAGE_SIZE
),
3949 round_up(destoff
+ len
, PAGE_SIZE
) - 1);
3954 static int btrfs_remap_file_range_prep(struct file
*file_in
, loff_t pos_in
,
3955 struct file
*file_out
, loff_t pos_out
,
3956 loff_t
*len
, unsigned int remap_flags
)
3958 struct inode
*inode_in
= file_inode(file_in
);
3959 struct inode
*inode_out
= file_inode(file_out
);
3960 u64 bs
= BTRFS_I(inode_out
)->root
->fs_info
->sb
->s_blocksize
;
3961 bool same_inode
= inode_out
== inode_in
;
3965 if (!(remap_flags
& REMAP_FILE_DEDUP
)) {
3966 struct btrfs_root
*root_out
= BTRFS_I(inode_out
)->root
;
3968 if (btrfs_root_readonly(root_out
))
3971 if (file_in
->f_path
.mnt
!= file_out
->f_path
.mnt
||
3972 inode_in
->i_sb
!= inode_out
->i_sb
)
3976 /* don't make the dst file partly checksummed */
3977 if ((BTRFS_I(inode_in
)->flags
& BTRFS_INODE_NODATASUM
) !=
3978 (BTRFS_I(inode_out
)->flags
& BTRFS_INODE_NODATASUM
)) {
3983 * Now that the inodes are locked, we need to start writeback ourselves
3984 * and can not rely on the writeback from the VFS's generic helper
3985 * generic_remap_file_range_prep() because:
3987 * 1) For compression we must call filemap_fdatawrite_range() range
3988 * twice (btrfs_fdatawrite_range() does it for us), and the generic
3989 * helper only calls it once;
3991 * 2) filemap_fdatawrite_range(), called by the generic helper only
3992 * waits for the writeback to complete, i.e. for IO to be done, and
3993 * not for the ordered extents to complete. We need to wait for them
3994 * to complete so that new file extent items are in the fs tree.
3996 if (*len
== 0 && !(remap_flags
& REMAP_FILE_DEDUP
))
3997 wb_len
= ALIGN(inode_in
->i_size
, bs
) - ALIGN_DOWN(pos_in
, bs
);
3999 wb_len
= ALIGN(*len
, bs
);
4002 * Since we don't lock ranges, wait for ongoing lockless dio writes (as
4003 * any in progress could create its ordered extents after we wait for
4004 * existing ordered extents below).
4006 inode_dio_wait(inode_in
);
4008 inode_dio_wait(inode_out
);
4010 ret
= btrfs_wait_ordered_range(inode_in
, ALIGN_DOWN(pos_in
, bs
),
4014 ret
= btrfs_wait_ordered_range(inode_out
, ALIGN_DOWN(pos_out
, bs
),
4019 return generic_remap_file_range_prep(file_in
, pos_in
, file_out
, pos_out
,
4023 loff_t
btrfs_remap_file_range(struct file
*src_file
, loff_t off
,
4024 struct file
*dst_file
, loff_t destoff
, loff_t len
,
4025 unsigned int remap_flags
)
4027 struct inode
*src_inode
= file_inode(src_file
);
4028 struct inode
*dst_inode
= file_inode(dst_file
);
4029 bool same_inode
= dst_inode
== src_inode
;
4032 if (remap_flags
& ~(REMAP_FILE_DEDUP
| REMAP_FILE_ADVISORY
))
4036 inode_lock(src_inode
);
4038 lock_two_nondirectories(src_inode
, dst_inode
);
4040 ret
= btrfs_remap_file_range_prep(src_file
, off
, dst_file
, destoff
,
4042 if (ret
< 0 || len
== 0)
4045 if (remap_flags
& REMAP_FILE_DEDUP
)
4046 ret
= btrfs_extent_same(src_inode
, off
, len
, dst_inode
, destoff
);
4048 ret
= btrfs_clone_files(dst_file
, src_file
, off
, len
, destoff
);
4052 inode_unlock(src_inode
);
4054 unlock_two_nondirectories(src_inode
, dst_inode
);
4056 return ret
< 0 ? ret
: len
;
4059 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
4061 struct inode
*inode
= file_inode(file
);
4062 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4063 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4064 struct btrfs_root
*new_root
;
4065 struct btrfs_dir_item
*di
;
4066 struct btrfs_trans_handle
*trans
;
4067 struct btrfs_path
*path
;
4068 struct btrfs_key location
;
4069 struct btrfs_disk_key disk_key
;
4074 if (!capable(CAP_SYS_ADMIN
))
4077 ret
= mnt_want_write_file(file
);
4081 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
4087 objectid
= BTRFS_FS_TREE_OBJECTID
;
4089 location
.objectid
= objectid
;
4090 location
.type
= BTRFS_ROOT_ITEM_KEY
;
4091 location
.offset
= (u64
)-1;
4093 new_root
= btrfs_read_fs_root_no_name(fs_info
, &location
);
4094 if (IS_ERR(new_root
)) {
4095 ret
= PTR_ERR(new_root
);
4098 if (!is_fstree(new_root
->root_key
.objectid
)) {
4103 path
= btrfs_alloc_path();
4108 path
->leave_spinning
= 1;
4110 trans
= btrfs_start_transaction(root
, 1);
4111 if (IS_ERR(trans
)) {
4112 btrfs_free_path(path
);
4113 ret
= PTR_ERR(trans
);
4117 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
4118 di
= btrfs_lookup_dir_item(trans
, fs_info
->tree_root
, path
,
4119 dir_id
, "default", 7, 1);
4120 if (IS_ERR_OR_NULL(di
)) {
4121 btrfs_free_path(path
);
4122 btrfs_end_transaction(trans
);
4124 "Umm, you don't have the default diritem, this isn't going to work");
4129 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
4130 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
4131 btrfs_mark_buffer_dirty(path
->nodes
[0]);
4132 btrfs_free_path(path
);
4134 btrfs_set_fs_incompat(fs_info
, DEFAULT_SUBVOL
);
4135 btrfs_end_transaction(trans
);
4137 mnt_drop_write_file(file
);
4141 static void get_block_group_info(struct list_head
*groups_list
,
4142 struct btrfs_ioctl_space_info
*space
)
4144 struct btrfs_block_group_cache
*block_group
;
4146 space
->total_bytes
= 0;
4147 space
->used_bytes
= 0;
4149 list_for_each_entry(block_group
, groups_list
, list
) {
4150 space
->flags
= block_group
->flags
;
4151 space
->total_bytes
+= block_group
->key
.offset
;
4152 space
->used_bytes
+=
4153 btrfs_block_group_used(&block_group
->item
);
4157 static long btrfs_ioctl_space_info(struct btrfs_fs_info
*fs_info
,
4160 struct btrfs_ioctl_space_args space_args
;
4161 struct btrfs_ioctl_space_info space
;
4162 struct btrfs_ioctl_space_info
*dest
;
4163 struct btrfs_ioctl_space_info
*dest_orig
;
4164 struct btrfs_ioctl_space_info __user
*user_dest
;
4165 struct btrfs_space_info
*info
;
4166 static const u64 types
[] = {
4167 BTRFS_BLOCK_GROUP_DATA
,
4168 BTRFS_BLOCK_GROUP_SYSTEM
,
4169 BTRFS_BLOCK_GROUP_METADATA
,
4170 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
4178 if (copy_from_user(&space_args
,
4179 (struct btrfs_ioctl_space_args __user
*)arg
,
4180 sizeof(space_args
)))
4183 for (i
= 0; i
< num_types
; i
++) {
4184 struct btrfs_space_info
*tmp
;
4188 list_for_each_entry_rcu(tmp
, &fs_info
->space_info
,
4190 if (tmp
->flags
== types
[i
]) {
4200 down_read(&info
->groups_sem
);
4201 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4202 if (!list_empty(&info
->block_groups
[c
]))
4205 up_read(&info
->groups_sem
);
4209 * Global block reserve, exported as a space_info
4213 /* space_slots == 0 means they are asking for a count */
4214 if (space_args
.space_slots
== 0) {
4215 space_args
.total_spaces
= slot_count
;
4219 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
4221 alloc_size
= sizeof(*dest
) * slot_count
;
4223 /* we generally have at most 6 or so space infos, one for each raid
4224 * level. So, a whole page should be more than enough for everyone
4226 if (alloc_size
> PAGE_SIZE
)
4229 space_args
.total_spaces
= 0;
4230 dest
= kmalloc(alloc_size
, GFP_KERNEL
);
4235 /* now we have a buffer to copy into */
4236 for (i
= 0; i
< num_types
; i
++) {
4237 struct btrfs_space_info
*tmp
;
4244 list_for_each_entry_rcu(tmp
, &fs_info
->space_info
,
4246 if (tmp
->flags
== types
[i
]) {
4255 down_read(&info
->groups_sem
);
4256 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
4257 if (!list_empty(&info
->block_groups
[c
])) {
4258 get_block_group_info(&info
->block_groups
[c
],
4260 memcpy(dest
, &space
, sizeof(space
));
4262 space_args
.total_spaces
++;
4268 up_read(&info
->groups_sem
);
4272 * Add global block reserve
4275 struct btrfs_block_rsv
*block_rsv
= &fs_info
->global_block_rsv
;
4277 spin_lock(&block_rsv
->lock
);
4278 space
.total_bytes
= block_rsv
->size
;
4279 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
4280 spin_unlock(&block_rsv
->lock
);
4281 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
4282 memcpy(dest
, &space
, sizeof(space
));
4283 space_args
.total_spaces
++;
4286 user_dest
= (struct btrfs_ioctl_space_info __user
*)
4287 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
4289 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
4294 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
4300 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
4303 struct btrfs_trans_handle
*trans
;
4307 trans
= btrfs_attach_transaction_barrier(root
);
4308 if (IS_ERR(trans
)) {
4309 if (PTR_ERR(trans
) != -ENOENT
)
4310 return PTR_ERR(trans
);
4312 /* No running transaction, don't bother */
4313 transid
= root
->fs_info
->last_trans_committed
;
4316 transid
= trans
->transid
;
4317 ret
= btrfs_commit_transaction_async(trans
, 0);
4319 btrfs_end_transaction(trans
);
4324 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
4329 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_fs_info
*fs_info
,
4335 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
4338 transid
= 0; /* current trans */
4340 return btrfs_wait_for_commit(fs_info
, transid
);
4343 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
4345 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
4346 struct btrfs_ioctl_scrub_args
*sa
;
4349 if (!capable(CAP_SYS_ADMIN
))
4352 sa
= memdup_user(arg
, sizeof(*sa
));
4356 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
4357 ret
= mnt_want_write_file(file
);
4362 ret
= btrfs_scrub_dev(fs_info
, sa
->devid
, sa
->start
, sa
->end
,
4363 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
4366 if (ret
== 0 && copy_to_user(arg
, sa
, sizeof(*sa
)))
4369 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
4370 mnt_drop_write_file(file
);
4376 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info
*fs_info
)
4378 if (!capable(CAP_SYS_ADMIN
))
4381 return btrfs_scrub_cancel(fs_info
);
4384 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info
*fs_info
,
4387 struct btrfs_ioctl_scrub_args
*sa
;
4390 if (!capable(CAP_SYS_ADMIN
))
4393 sa
= memdup_user(arg
, sizeof(*sa
));
4397 ret
= btrfs_scrub_progress(fs_info
, sa
->devid
, &sa
->progress
);
4399 if (ret
== 0 && copy_to_user(arg
, sa
, sizeof(*sa
)))
4406 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info
*fs_info
,
4409 struct btrfs_ioctl_get_dev_stats
*sa
;
4412 sa
= memdup_user(arg
, sizeof(*sa
));
4416 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
4421 ret
= btrfs_get_dev_stats(fs_info
, sa
);
4423 if (ret
== 0 && copy_to_user(arg
, sa
, sizeof(*sa
)))
4430 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info
*fs_info
,
4433 struct btrfs_ioctl_dev_replace_args
*p
;
4436 if (!capable(CAP_SYS_ADMIN
))
4439 p
= memdup_user(arg
, sizeof(*p
));
4444 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
4445 if (sb_rdonly(fs_info
->sb
)) {
4449 if (test_and_set_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
)) {
4450 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4452 ret
= btrfs_dev_replace_by_ioctl(fs_info
, p
);
4453 clear_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
);
4456 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
4457 btrfs_dev_replace_status(fs_info
, p
);
4460 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
4461 p
->result
= btrfs_dev_replace_cancel(fs_info
);
4469 if ((ret
== 0 || ret
== -ECANCELED
) && copy_to_user(arg
, p
, sizeof(*p
)))
4476 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
4482 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
4483 struct inode_fs_paths
*ipath
= NULL
;
4484 struct btrfs_path
*path
;
4486 if (!capable(CAP_DAC_READ_SEARCH
))
4489 path
= btrfs_alloc_path();
4495 ipa
= memdup_user(arg
, sizeof(*ipa
));
4502 size
= min_t(u32
, ipa
->size
, 4096);
4503 ipath
= init_ipath(size
, root
, path
);
4504 if (IS_ERR(ipath
)) {
4505 ret
= PTR_ERR(ipath
);
4510 ret
= paths_from_inode(ipa
->inum
, ipath
);
4514 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
4515 rel_ptr
= ipath
->fspath
->val
[i
] -
4516 (u64
)(unsigned long)ipath
->fspath
->val
;
4517 ipath
->fspath
->val
[i
] = rel_ptr
;
4520 ret
= copy_to_user((void __user
*)(unsigned long)ipa
->fspath
,
4521 ipath
->fspath
, size
);
4528 btrfs_free_path(path
);
4535 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
4537 struct btrfs_data_container
*inodes
= ctx
;
4538 const size_t c
= 3 * sizeof(u64
);
4540 if (inodes
->bytes_left
>= c
) {
4541 inodes
->bytes_left
-= c
;
4542 inodes
->val
[inodes
->elem_cnt
] = inum
;
4543 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
4544 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
4545 inodes
->elem_cnt
+= 3;
4547 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
4548 inodes
->bytes_left
= 0;
4549 inodes
->elem_missed
+= 3;
4555 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info
*fs_info
,
4556 void __user
*arg
, int version
)
4560 struct btrfs_ioctl_logical_ino_args
*loi
;
4561 struct btrfs_data_container
*inodes
= NULL
;
4562 struct btrfs_path
*path
= NULL
;
4565 if (!capable(CAP_SYS_ADMIN
))
4568 loi
= memdup_user(arg
, sizeof(*loi
));
4570 return PTR_ERR(loi
);
4573 ignore_offset
= false;
4574 size
= min_t(u32
, loi
->size
, SZ_64K
);
4576 /* All reserved bits must be 0 for now */
4577 if (memchr_inv(loi
->reserved
, 0, sizeof(loi
->reserved
))) {
4581 /* Only accept flags we have defined so far */
4582 if (loi
->flags
& ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET
)) {
4586 ignore_offset
= loi
->flags
& BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET
;
4587 size
= min_t(u32
, loi
->size
, SZ_16M
);
4590 path
= btrfs_alloc_path();
4596 inodes
= init_data_container(size
);
4597 if (IS_ERR(inodes
)) {
4598 ret
= PTR_ERR(inodes
);
4603 ret
= iterate_inodes_from_logical(loi
->logical
, fs_info
, path
,
4604 build_ino_list
, inodes
, ignore_offset
);
4610 ret
= copy_to_user((void __user
*)(unsigned long)loi
->inodes
, inodes
,
4616 btrfs_free_path(path
);
4624 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
,
4625 struct btrfs_ioctl_balance_args
*bargs
)
4627 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
4629 bargs
->flags
= bctl
->flags
;
4631 if (test_bit(BTRFS_FS_BALANCE_RUNNING
, &fs_info
->flags
))
4632 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
4633 if (atomic_read(&fs_info
->balance_pause_req
))
4634 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
4635 if (atomic_read(&fs_info
->balance_cancel_req
))
4636 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
4638 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
4639 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
4640 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
4642 spin_lock(&fs_info
->balance_lock
);
4643 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
4644 spin_unlock(&fs_info
->balance_lock
);
4647 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
4649 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4650 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
4651 struct btrfs_ioctl_balance_args
*bargs
;
4652 struct btrfs_balance_control
*bctl
;
4653 bool need_unlock
; /* for mut. excl. ops lock */
4656 if (!capable(CAP_SYS_ADMIN
))
4659 ret
= mnt_want_write_file(file
);
4664 if (!test_and_set_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
)) {
4665 mutex_lock(&fs_info
->balance_mutex
);
4671 * mut. excl. ops lock is locked. Three possibilities:
4672 * (1) some other op is running
4673 * (2) balance is running
4674 * (3) balance is paused -- special case (think resume)
4676 mutex_lock(&fs_info
->balance_mutex
);
4677 if (fs_info
->balance_ctl
) {
4678 /* this is either (2) or (3) */
4679 if (!test_bit(BTRFS_FS_BALANCE_RUNNING
, &fs_info
->flags
)) {
4680 mutex_unlock(&fs_info
->balance_mutex
);
4682 * Lock released to allow other waiters to continue,
4683 * we'll reexamine the status again.
4685 mutex_lock(&fs_info
->balance_mutex
);
4687 if (fs_info
->balance_ctl
&&
4688 !test_bit(BTRFS_FS_BALANCE_RUNNING
, &fs_info
->flags
)) {
4690 need_unlock
= false;
4694 mutex_unlock(&fs_info
->balance_mutex
);
4698 mutex_unlock(&fs_info
->balance_mutex
);
4704 mutex_unlock(&fs_info
->balance_mutex
);
4705 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
4710 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
));
4713 bargs
= memdup_user(arg
, sizeof(*bargs
));
4714 if (IS_ERR(bargs
)) {
4715 ret
= PTR_ERR(bargs
);
4719 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
4720 if (!fs_info
->balance_ctl
) {
4725 bctl
= fs_info
->balance_ctl
;
4726 spin_lock(&fs_info
->balance_lock
);
4727 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
4728 spin_unlock(&fs_info
->balance_lock
);
4736 if (fs_info
->balance_ctl
) {
4741 bctl
= kzalloc(sizeof(*bctl
), GFP_KERNEL
);
4748 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
4749 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
4750 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
4752 bctl
->flags
= bargs
->flags
;
4754 /* balance everything - no filters */
4755 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
4758 if (bctl
->flags
& ~(BTRFS_BALANCE_ARGS_MASK
| BTRFS_BALANCE_TYPE_MASK
)) {
4765 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP goes to
4766 * btrfs_balance. bctl is freed in reset_balance_state, or, if
4767 * restriper was paused all the way until unmount, in free_fs_info.
4768 * The flag should be cleared after reset_balance_state.
4770 need_unlock
= false;
4772 ret
= btrfs_balance(fs_info
, bctl
, bargs
);
4775 if ((ret
== 0 || ret
== -ECANCELED
) && arg
) {
4776 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4785 mutex_unlock(&fs_info
->balance_mutex
);
4787 clear_bit(BTRFS_FS_EXCL_OP
, &fs_info
->flags
);
4789 mnt_drop_write_file(file
);
4793 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info
*fs_info
, int cmd
)
4795 if (!capable(CAP_SYS_ADMIN
))
4799 case BTRFS_BALANCE_CTL_PAUSE
:
4800 return btrfs_pause_balance(fs_info
);
4801 case BTRFS_BALANCE_CTL_CANCEL
:
4802 return btrfs_cancel_balance(fs_info
);
4808 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info
*fs_info
,
4811 struct btrfs_ioctl_balance_args
*bargs
;
4814 if (!capable(CAP_SYS_ADMIN
))
4817 mutex_lock(&fs_info
->balance_mutex
);
4818 if (!fs_info
->balance_ctl
) {
4823 bargs
= kzalloc(sizeof(*bargs
), GFP_KERNEL
);
4829 btrfs_update_ioctl_balance_args(fs_info
, bargs
);
4831 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
4836 mutex_unlock(&fs_info
->balance_mutex
);
4840 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
4842 struct inode
*inode
= file_inode(file
);
4843 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4844 struct btrfs_ioctl_quota_ctl_args
*sa
;
4847 if (!capable(CAP_SYS_ADMIN
))
4850 ret
= mnt_want_write_file(file
);
4854 sa
= memdup_user(arg
, sizeof(*sa
));
4860 down_write(&fs_info
->subvol_sem
);
4863 case BTRFS_QUOTA_CTL_ENABLE
:
4864 ret
= btrfs_quota_enable(fs_info
);
4866 case BTRFS_QUOTA_CTL_DISABLE
:
4867 ret
= btrfs_quota_disable(fs_info
);
4875 up_write(&fs_info
->subvol_sem
);
4877 mnt_drop_write_file(file
);
4881 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
4883 struct inode
*inode
= file_inode(file
);
4884 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4885 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4886 struct btrfs_ioctl_qgroup_assign_args
*sa
;
4887 struct btrfs_trans_handle
*trans
;
4891 if (!capable(CAP_SYS_ADMIN
))
4894 ret
= mnt_want_write_file(file
);
4898 sa
= memdup_user(arg
, sizeof(*sa
));
4904 trans
= btrfs_join_transaction(root
);
4905 if (IS_ERR(trans
)) {
4906 ret
= PTR_ERR(trans
);
4911 ret
= btrfs_add_qgroup_relation(trans
, sa
->src
, sa
->dst
);
4913 ret
= btrfs_del_qgroup_relation(trans
, sa
->src
, sa
->dst
);
4916 /* update qgroup status and info */
4917 err
= btrfs_run_qgroups(trans
);
4919 btrfs_handle_fs_error(fs_info
, err
,
4920 "failed to update qgroup status and info");
4921 err
= btrfs_end_transaction(trans
);
4928 mnt_drop_write_file(file
);
4932 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
4934 struct inode
*inode
= file_inode(file
);
4935 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4936 struct btrfs_ioctl_qgroup_create_args
*sa
;
4937 struct btrfs_trans_handle
*trans
;
4941 if (!capable(CAP_SYS_ADMIN
))
4944 ret
= mnt_want_write_file(file
);
4948 sa
= memdup_user(arg
, sizeof(*sa
));
4954 if (!sa
->qgroupid
) {
4959 trans
= btrfs_join_transaction(root
);
4960 if (IS_ERR(trans
)) {
4961 ret
= PTR_ERR(trans
);
4966 ret
= btrfs_create_qgroup(trans
, sa
->qgroupid
);
4968 ret
= btrfs_remove_qgroup(trans
, sa
->qgroupid
);
4971 err
= btrfs_end_transaction(trans
);
4978 mnt_drop_write_file(file
);
4982 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
4984 struct inode
*inode
= file_inode(file
);
4985 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4986 struct btrfs_ioctl_qgroup_limit_args
*sa
;
4987 struct btrfs_trans_handle
*trans
;
4992 if (!capable(CAP_SYS_ADMIN
))
4995 ret
= mnt_want_write_file(file
);
4999 sa
= memdup_user(arg
, sizeof(*sa
));
5005 trans
= btrfs_join_transaction(root
);
5006 if (IS_ERR(trans
)) {
5007 ret
= PTR_ERR(trans
);
5011 qgroupid
= sa
->qgroupid
;
5013 /* take the current subvol as qgroup */
5014 qgroupid
= root
->root_key
.objectid
;
5017 ret
= btrfs_limit_qgroup(trans
, qgroupid
, &sa
->lim
);
5019 err
= btrfs_end_transaction(trans
);
5026 mnt_drop_write_file(file
);
5030 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
5032 struct inode
*inode
= file_inode(file
);
5033 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5034 struct btrfs_ioctl_quota_rescan_args
*qsa
;
5037 if (!capable(CAP_SYS_ADMIN
))
5040 ret
= mnt_want_write_file(file
);
5044 qsa
= memdup_user(arg
, sizeof(*qsa
));
5055 ret
= btrfs_qgroup_rescan(fs_info
);
5060 mnt_drop_write_file(file
);
5064 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
5066 struct inode
*inode
= file_inode(file
);
5067 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5068 struct btrfs_ioctl_quota_rescan_args
*qsa
;
5071 if (!capable(CAP_SYS_ADMIN
))
5074 qsa
= kzalloc(sizeof(*qsa
), GFP_KERNEL
);
5078 if (fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
5080 qsa
->progress
= fs_info
->qgroup_rescan_progress
.objectid
;
5083 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
5090 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
5092 struct inode
*inode
= file_inode(file
);
5093 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5095 if (!capable(CAP_SYS_ADMIN
))
5098 return btrfs_qgroup_wait_for_completion(fs_info
, true);
5101 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
5102 struct btrfs_ioctl_received_subvol_args
*sa
)
5104 struct inode
*inode
= file_inode(file
);
5105 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5106 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5107 struct btrfs_root_item
*root_item
= &root
->root_item
;
5108 struct btrfs_trans_handle
*trans
;
5109 struct timespec64 ct
= current_time(inode
);
5111 int received_uuid_changed
;
5113 if (!inode_owner_or_capable(inode
))
5116 ret
= mnt_want_write_file(file
);
5120 down_write(&fs_info
->subvol_sem
);
5122 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
) {
5127 if (btrfs_root_readonly(root
)) {
5134 * 2 - uuid items (received uuid + subvol uuid)
5136 trans
= btrfs_start_transaction(root
, 3);
5137 if (IS_ERR(trans
)) {
5138 ret
= PTR_ERR(trans
);
5143 sa
->rtransid
= trans
->transid
;
5144 sa
->rtime
.sec
= ct
.tv_sec
;
5145 sa
->rtime
.nsec
= ct
.tv_nsec
;
5147 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
5149 if (received_uuid_changed
&&
5150 !btrfs_is_empty_uuid(root_item
->received_uuid
)) {
5151 ret
= btrfs_uuid_tree_remove(trans
, root_item
->received_uuid
,
5152 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5153 root
->root_key
.objectid
);
5154 if (ret
&& ret
!= -ENOENT
) {
5155 btrfs_abort_transaction(trans
, ret
);
5156 btrfs_end_transaction(trans
);
5160 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
5161 btrfs_set_root_stransid(root_item
, sa
->stransid
);
5162 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
5163 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
5164 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
5165 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
5166 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
5168 ret
= btrfs_update_root(trans
, fs_info
->tree_root
,
5169 &root
->root_key
, &root
->root_item
);
5171 btrfs_end_transaction(trans
);
5174 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
5175 ret
= btrfs_uuid_tree_add(trans
, sa
->uuid
,
5176 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
5177 root
->root_key
.objectid
);
5178 if (ret
< 0 && ret
!= -EEXIST
) {
5179 btrfs_abort_transaction(trans
, ret
);
5180 btrfs_end_transaction(trans
);
5184 ret
= btrfs_commit_transaction(trans
);
5186 up_write(&fs_info
->subvol_sem
);
5187 mnt_drop_write_file(file
);
5192 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
5195 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
5196 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
5199 args32
= memdup_user(arg
, sizeof(*args32
));
5201 return PTR_ERR(args32
);
5203 args64
= kmalloc(sizeof(*args64
), GFP_KERNEL
);
5209 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
5210 args64
->stransid
= args32
->stransid
;
5211 args64
->rtransid
= args32
->rtransid
;
5212 args64
->stime
.sec
= args32
->stime
.sec
;
5213 args64
->stime
.nsec
= args32
->stime
.nsec
;
5214 args64
->rtime
.sec
= args32
->rtime
.sec
;
5215 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
5216 args64
->flags
= args32
->flags
;
5218 ret
= _btrfs_ioctl_set_received_subvol(file
, args64
);
5222 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
5223 args32
->stransid
= args64
->stransid
;
5224 args32
->rtransid
= args64
->rtransid
;
5225 args32
->stime
.sec
= args64
->stime
.sec
;
5226 args32
->stime
.nsec
= args64
->stime
.nsec
;
5227 args32
->rtime
.sec
= args64
->rtime
.sec
;
5228 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
5229 args32
->flags
= args64
->flags
;
5231 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
5242 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
5245 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
5248 sa
= memdup_user(arg
, sizeof(*sa
));
5252 ret
= _btrfs_ioctl_set_received_subvol(file
, sa
);
5257 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
5266 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
5268 struct inode
*inode
= file_inode(file
);
5269 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5272 char label
[BTRFS_LABEL_SIZE
];
5274 spin_lock(&fs_info
->super_lock
);
5275 memcpy(label
, fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
5276 spin_unlock(&fs_info
->super_lock
);
5278 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
5280 if (len
== BTRFS_LABEL_SIZE
) {
5282 "label is too long, return the first %zu bytes",
5286 ret
= copy_to_user(arg
, label
, len
);
5288 return ret
? -EFAULT
: 0;
5291 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
5293 struct inode
*inode
= file_inode(file
);
5294 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5295 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5296 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
5297 struct btrfs_trans_handle
*trans
;
5298 char label
[BTRFS_LABEL_SIZE
];
5301 if (!capable(CAP_SYS_ADMIN
))
5304 if (copy_from_user(label
, arg
, sizeof(label
)))
5307 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
5309 "unable to set label with more than %d bytes",
5310 BTRFS_LABEL_SIZE
- 1);
5314 ret
= mnt_want_write_file(file
);
5318 trans
= btrfs_start_transaction(root
, 0);
5319 if (IS_ERR(trans
)) {
5320 ret
= PTR_ERR(trans
);
5324 spin_lock(&fs_info
->super_lock
);
5325 strcpy(super_block
->label
, label
);
5326 spin_unlock(&fs_info
->super_lock
);
5327 ret
= btrfs_commit_transaction(trans
);
5330 mnt_drop_write_file(file
);
5334 #define INIT_FEATURE_FLAGS(suffix) \
5335 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5336 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5337 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5339 int btrfs_ioctl_get_supported_features(void __user
*arg
)
5341 static const struct btrfs_ioctl_feature_flags features
[3] = {
5342 INIT_FEATURE_FLAGS(SUPP
),
5343 INIT_FEATURE_FLAGS(SAFE_SET
),
5344 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
5347 if (copy_to_user(arg
, &features
, sizeof(features
)))
5353 static int btrfs_ioctl_get_features(struct file
*file
, void __user
*arg
)
5355 struct inode
*inode
= file_inode(file
);
5356 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5357 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
5358 struct btrfs_ioctl_feature_flags features
;
5360 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
5361 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
5362 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
5364 if (copy_to_user(arg
, &features
, sizeof(features
)))
5370 static int check_feature_bits(struct btrfs_fs_info
*fs_info
,
5371 enum btrfs_feature_set set
,
5372 u64 change_mask
, u64 flags
, u64 supported_flags
,
5373 u64 safe_set
, u64 safe_clear
)
5375 const char *type
= btrfs_feature_set_names
[set
];
5377 u64 disallowed
, unsupported
;
5378 u64 set_mask
= flags
& change_mask
;
5379 u64 clear_mask
= ~flags
& change_mask
;
5381 unsupported
= set_mask
& ~supported_flags
;
5383 names
= btrfs_printable_features(set
, unsupported
);
5386 "this kernel does not support the %s feature bit%s",
5387 names
, strchr(names
, ',') ? "s" : "");
5391 "this kernel does not support %s bits 0x%llx",
5396 disallowed
= set_mask
& ~safe_set
;
5398 names
= btrfs_printable_features(set
, disallowed
);
5401 "can't set the %s feature bit%s while mounted",
5402 names
, strchr(names
, ',') ? "s" : "");
5406 "can't set %s bits 0x%llx while mounted",
5411 disallowed
= clear_mask
& ~safe_clear
;
5413 names
= btrfs_printable_features(set
, disallowed
);
5416 "can't clear the %s feature bit%s while mounted",
5417 names
, strchr(names
, ',') ? "s" : "");
5421 "can't clear %s bits 0x%llx while mounted",
5429 #define check_feature(fs_info, change_mask, flags, mask_base) \
5430 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5431 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5432 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5433 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5435 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
5437 struct inode
*inode
= file_inode(file
);
5438 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5439 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5440 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
5441 struct btrfs_ioctl_feature_flags flags
[2];
5442 struct btrfs_trans_handle
*trans
;
5446 if (!capable(CAP_SYS_ADMIN
))
5449 if (copy_from_user(flags
, arg
, sizeof(flags
)))
5453 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
5454 !flags
[0].incompat_flags
)
5457 ret
= check_feature(fs_info
, flags
[0].compat_flags
,
5458 flags
[1].compat_flags
, COMPAT
);
5462 ret
= check_feature(fs_info
, flags
[0].compat_ro_flags
,
5463 flags
[1].compat_ro_flags
, COMPAT_RO
);
5467 ret
= check_feature(fs_info
, flags
[0].incompat_flags
,
5468 flags
[1].incompat_flags
, INCOMPAT
);
5472 ret
= mnt_want_write_file(file
);
5476 trans
= btrfs_start_transaction(root
, 0);
5477 if (IS_ERR(trans
)) {
5478 ret
= PTR_ERR(trans
);
5479 goto out_drop_write
;
5482 spin_lock(&fs_info
->super_lock
);
5483 newflags
= btrfs_super_compat_flags(super_block
);
5484 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
5485 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
5486 btrfs_set_super_compat_flags(super_block
, newflags
);
5488 newflags
= btrfs_super_compat_ro_flags(super_block
);
5489 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
5490 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
5491 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
5493 newflags
= btrfs_super_incompat_flags(super_block
);
5494 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
5495 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
5496 btrfs_set_super_incompat_flags(super_block
, newflags
);
5497 spin_unlock(&fs_info
->super_lock
);
5499 ret
= btrfs_commit_transaction(trans
);
5501 mnt_drop_write_file(file
);
5506 static int _btrfs_ioctl_send(struct file
*file
, void __user
*argp
, bool compat
)
5508 struct btrfs_ioctl_send_args
*arg
;
5512 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5513 struct btrfs_ioctl_send_args_32 args32
;
5515 ret
= copy_from_user(&args32
, argp
, sizeof(args32
));
5518 arg
= kzalloc(sizeof(*arg
), GFP_KERNEL
);
5521 arg
->send_fd
= args32
.send_fd
;
5522 arg
->clone_sources_count
= args32
.clone_sources_count
;
5523 arg
->clone_sources
= compat_ptr(args32
.clone_sources
);
5524 arg
->parent_root
= args32
.parent_root
;
5525 arg
->flags
= args32
.flags
;
5526 memcpy(arg
->reserved
, args32
.reserved
,
5527 sizeof(args32
.reserved
));
5532 arg
= memdup_user(argp
, sizeof(*arg
));
5534 return PTR_ERR(arg
);
5536 ret
= btrfs_ioctl_send(file
, arg
);
5541 long btrfs_ioctl(struct file
*file
, unsigned int
5542 cmd
, unsigned long arg
)
5544 struct inode
*inode
= file_inode(file
);
5545 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
5546 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
5547 void __user
*argp
= (void __user
*)arg
;
5550 case FS_IOC_GETFLAGS
:
5551 return btrfs_ioctl_getflags(file
, argp
);
5552 case FS_IOC_SETFLAGS
:
5553 return btrfs_ioctl_setflags(file
, argp
);
5554 case FS_IOC_GETVERSION
:
5555 return btrfs_ioctl_getversion(file
, argp
);
5557 return btrfs_ioctl_fitrim(file
, argp
);
5558 case BTRFS_IOC_SNAP_CREATE
:
5559 return btrfs_ioctl_snap_create(file
, argp
, 0);
5560 case BTRFS_IOC_SNAP_CREATE_V2
:
5561 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
5562 case BTRFS_IOC_SUBVOL_CREATE
:
5563 return btrfs_ioctl_snap_create(file
, argp
, 1);
5564 case BTRFS_IOC_SUBVOL_CREATE_V2
:
5565 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
5566 case BTRFS_IOC_SNAP_DESTROY
:
5567 return btrfs_ioctl_snap_destroy(file
, argp
);
5568 case BTRFS_IOC_SUBVOL_GETFLAGS
:
5569 return btrfs_ioctl_subvol_getflags(file
, argp
);
5570 case BTRFS_IOC_SUBVOL_SETFLAGS
:
5571 return btrfs_ioctl_subvol_setflags(file
, argp
);
5572 case BTRFS_IOC_DEFAULT_SUBVOL
:
5573 return btrfs_ioctl_default_subvol(file
, argp
);
5574 case BTRFS_IOC_DEFRAG
:
5575 return btrfs_ioctl_defrag(file
, NULL
);
5576 case BTRFS_IOC_DEFRAG_RANGE
:
5577 return btrfs_ioctl_defrag(file
, argp
);
5578 case BTRFS_IOC_RESIZE
:
5579 return btrfs_ioctl_resize(file
, argp
);
5580 case BTRFS_IOC_ADD_DEV
:
5581 return btrfs_ioctl_add_dev(fs_info
, argp
);
5582 case BTRFS_IOC_RM_DEV
:
5583 return btrfs_ioctl_rm_dev(file
, argp
);
5584 case BTRFS_IOC_RM_DEV_V2
:
5585 return btrfs_ioctl_rm_dev_v2(file
, argp
);
5586 case BTRFS_IOC_FS_INFO
:
5587 return btrfs_ioctl_fs_info(fs_info
, argp
);
5588 case BTRFS_IOC_DEV_INFO
:
5589 return btrfs_ioctl_dev_info(fs_info
, argp
);
5590 case BTRFS_IOC_BALANCE
:
5591 return btrfs_ioctl_balance(file
, NULL
);
5592 case BTRFS_IOC_TREE_SEARCH
:
5593 return btrfs_ioctl_tree_search(file
, argp
);
5594 case BTRFS_IOC_TREE_SEARCH_V2
:
5595 return btrfs_ioctl_tree_search_v2(file
, argp
);
5596 case BTRFS_IOC_INO_LOOKUP
:
5597 return btrfs_ioctl_ino_lookup(file
, argp
);
5598 case BTRFS_IOC_INO_PATHS
:
5599 return btrfs_ioctl_ino_to_path(root
, argp
);
5600 case BTRFS_IOC_LOGICAL_INO
:
5601 return btrfs_ioctl_logical_to_ino(fs_info
, argp
, 1);
5602 case BTRFS_IOC_LOGICAL_INO_V2
:
5603 return btrfs_ioctl_logical_to_ino(fs_info
, argp
, 2);
5604 case BTRFS_IOC_SPACE_INFO
:
5605 return btrfs_ioctl_space_info(fs_info
, argp
);
5606 case BTRFS_IOC_SYNC
: {
5609 ret
= btrfs_start_delalloc_roots(fs_info
, -1);
5612 ret
= btrfs_sync_fs(inode
->i_sb
, 1);
5614 * The transaction thread may want to do more work,
5615 * namely it pokes the cleaner kthread that will start
5616 * processing uncleaned subvols.
5618 wake_up_process(fs_info
->transaction_kthread
);
5621 case BTRFS_IOC_START_SYNC
:
5622 return btrfs_ioctl_start_sync(root
, argp
);
5623 case BTRFS_IOC_WAIT_SYNC
:
5624 return btrfs_ioctl_wait_sync(fs_info
, argp
);
5625 case BTRFS_IOC_SCRUB
:
5626 return btrfs_ioctl_scrub(file
, argp
);
5627 case BTRFS_IOC_SCRUB_CANCEL
:
5628 return btrfs_ioctl_scrub_cancel(fs_info
);
5629 case BTRFS_IOC_SCRUB_PROGRESS
:
5630 return btrfs_ioctl_scrub_progress(fs_info
, argp
);
5631 case BTRFS_IOC_BALANCE_V2
:
5632 return btrfs_ioctl_balance(file
, argp
);
5633 case BTRFS_IOC_BALANCE_CTL
:
5634 return btrfs_ioctl_balance_ctl(fs_info
, arg
);
5635 case BTRFS_IOC_BALANCE_PROGRESS
:
5636 return btrfs_ioctl_balance_progress(fs_info
, argp
);
5637 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
5638 return btrfs_ioctl_set_received_subvol(file
, argp
);
5640 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
5641 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
5643 case BTRFS_IOC_SEND
:
5644 return _btrfs_ioctl_send(file
, argp
, false);
5645 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5646 case BTRFS_IOC_SEND_32
:
5647 return _btrfs_ioctl_send(file
, argp
, true);
5649 case BTRFS_IOC_GET_DEV_STATS
:
5650 return btrfs_ioctl_get_dev_stats(fs_info
, argp
);
5651 case BTRFS_IOC_QUOTA_CTL
:
5652 return btrfs_ioctl_quota_ctl(file
, argp
);
5653 case BTRFS_IOC_QGROUP_ASSIGN
:
5654 return btrfs_ioctl_qgroup_assign(file
, argp
);
5655 case BTRFS_IOC_QGROUP_CREATE
:
5656 return btrfs_ioctl_qgroup_create(file
, argp
);
5657 case BTRFS_IOC_QGROUP_LIMIT
:
5658 return btrfs_ioctl_qgroup_limit(file
, argp
);
5659 case BTRFS_IOC_QUOTA_RESCAN
:
5660 return btrfs_ioctl_quota_rescan(file
, argp
);
5661 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
5662 return btrfs_ioctl_quota_rescan_status(file
, argp
);
5663 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
5664 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
5665 case BTRFS_IOC_DEV_REPLACE
:
5666 return btrfs_ioctl_dev_replace(fs_info
, argp
);
5667 case BTRFS_IOC_GET_FSLABEL
:
5668 return btrfs_ioctl_get_fslabel(file
, argp
);
5669 case BTRFS_IOC_SET_FSLABEL
:
5670 return btrfs_ioctl_set_fslabel(file
, argp
);
5671 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
5672 return btrfs_ioctl_get_supported_features(argp
);
5673 case BTRFS_IOC_GET_FEATURES
:
5674 return btrfs_ioctl_get_features(file
, argp
);
5675 case BTRFS_IOC_SET_FEATURES
:
5676 return btrfs_ioctl_set_features(file
, argp
);
5677 case FS_IOC_FSGETXATTR
:
5678 return btrfs_ioctl_fsgetxattr(file
, argp
);
5679 case FS_IOC_FSSETXATTR
:
5680 return btrfs_ioctl_fssetxattr(file
, argp
);
5681 case BTRFS_IOC_GET_SUBVOL_INFO
:
5682 return btrfs_ioctl_get_subvol_info(file
, argp
);
5683 case BTRFS_IOC_GET_SUBVOL_ROOTREF
:
5684 return btrfs_ioctl_get_subvol_rootref(file
, argp
);
5685 case BTRFS_IOC_INO_LOOKUP_USER
:
5686 return btrfs_ioctl_ino_lookup_user(file
, argp
);
5692 #ifdef CONFIG_COMPAT
5693 long btrfs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
5696 * These all access 32-bit values anyway so no further
5697 * handling is necessary.
5700 case FS_IOC32_GETFLAGS
:
5701 cmd
= FS_IOC_GETFLAGS
;
5703 case FS_IOC32_SETFLAGS
:
5704 cmd
= FS_IOC_SETFLAGS
;
5706 case FS_IOC32_GETVERSION
:
5707 cmd
= FS_IOC_GETVERSION
;
5711 return btrfs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));