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>
29 #include <linux/fileattr.h>
30 #include <linux/fsverity.h>
31 #include <linux/sched/xacct.h>
35 #include "transaction.h"
36 #include "btrfs_inode.h"
37 #include "print-tree.h"
41 #include "rcu-string.h"
43 #include "dev-replace.h"
48 #include "compression.h"
49 #include "space-info.h"
50 #include "delalloc-space.h"
51 #include "block-group.h"
54 #include "accessors.h"
55 #include "extent-tree.h"
56 #include "root-tree.h"
59 #include "uuid-tree.h"
66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
71 struct btrfs_ioctl_timespec_32
{
74 } __attribute__ ((__packed__
));
76 struct btrfs_ioctl_received_subvol_args_32
{
77 char uuid
[BTRFS_UUID_SIZE
]; /* in */
78 __u64 stransid
; /* in */
79 __u64 rtransid
; /* out */
80 struct btrfs_ioctl_timespec_32 stime
; /* in */
81 struct btrfs_ioctl_timespec_32 rtime
; /* out */
83 __u64 reserved
[16]; /* in */
84 } __attribute__ ((__packed__
));
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
90 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
91 struct btrfs_ioctl_send_args_32
{
92 __s64 send_fd
; /* in */
93 __u64 clone_sources_count
; /* in */
94 compat_uptr_t clone_sources
; /* in */
95 __u64 parent_root
; /* in */
97 __u32 version
; /* in */
98 __u8 reserved
[28]; /* in */
99 } __attribute__ ((__packed__
));
101 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
102 struct btrfs_ioctl_send_args_32)
104 struct btrfs_ioctl_encoded_io_args_32
{
106 compat_ulong_t iovcnt
;
111 __u64 unencoded_offset
;
117 #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
118 struct btrfs_ioctl_encoded_io_args_32)
119 #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
120 struct btrfs_ioctl_encoded_io_args_32)
123 /* Mask out flags that are inappropriate for the given type of inode. */
124 static unsigned int btrfs_mask_fsflags_for_type(struct inode
*inode
,
127 if (S_ISDIR(inode
->i_mode
))
129 else if (S_ISREG(inode
->i_mode
))
130 return flags
& ~FS_DIRSYNC_FL
;
132 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
136 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
139 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode
*binode
)
141 unsigned int iflags
= 0;
142 u32 flags
= binode
->flags
;
143 u32 ro_flags
= binode
->ro_flags
;
145 if (flags
& BTRFS_INODE_SYNC
)
146 iflags
|= FS_SYNC_FL
;
147 if (flags
& BTRFS_INODE_IMMUTABLE
)
148 iflags
|= FS_IMMUTABLE_FL
;
149 if (flags
& BTRFS_INODE_APPEND
)
150 iflags
|= FS_APPEND_FL
;
151 if (flags
& BTRFS_INODE_NODUMP
)
152 iflags
|= FS_NODUMP_FL
;
153 if (flags
& BTRFS_INODE_NOATIME
)
154 iflags
|= FS_NOATIME_FL
;
155 if (flags
& BTRFS_INODE_DIRSYNC
)
156 iflags
|= FS_DIRSYNC_FL
;
157 if (flags
& BTRFS_INODE_NODATACOW
)
158 iflags
|= FS_NOCOW_FL
;
159 if (ro_flags
& BTRFS_INODE_RO_VERITY
)
160 iflags
|= FS_VERITY_FL
;
162 if (flags
& BTRFS_INODE_NOCOMPRESS
)
163 iflags
|= FS_NOCOMP_FL
;
164 else if (flags
& BTRFS_INODE_COMPRESS
)
165 iflags
|= FS_COMPR_FL
;
171 * Update inode->i_flags based on the btrfs internal flags.
173 void btrfs_sync_inode_flags_to_i_flags(struct inode
*inode
)
175 struct btrfs_inode
*binode
= BTRFS_I(inode
);
176 unsigned int new_fl
= 0;
178 if (binode
->flags
& BTRFS_INODE_SYNC
)
180 if (binode
->flags
& BTRFS_INODE_IMMUTABLE
)
181 new_fl
|= S_IMMUTABLE
;
182 if (binode
->flags
& BTRFS_INODE_APPEND
)
184 if (binode
->flags
& BTRFS_INODE_NOATIME
)
186 if (binode
->flags
& BTRFS_INODE_DIRSYNC
)
188 if (binode
->ro_flags
& BTRFS_INODE_RO_VERITY
)
191 set_mask_bits(&inode
->i_flags
,
192 S_SYNC
| S_APPEND
| S_IMMUTABLE
| S_NOATIME
| S_DIRSYNC
|
197 * Check if @flags are a supported and valid set of FS_*_FL flags and that
198 * the old and new flags are not conflicting
200 static int check_fsflags(unsigned int old_flags
, unsigned int flags
)
202 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
203 FS_NOATIME_FL
| FS_NODUMP_FL
| \
204 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
205 FS_NOCOMP_FL
| FS_COMPR_FL
|
209 /* COMPR and NOCOMP on new/old are valid */
210 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
213 if ((flags
& FS_COMPR_FL
) && (flags
& FS_NOCOW_FL
))
216 /* NOCOW and compression options are mutually exclusive */
217 if ((old_flags
& FS_NOCOW_FL
) && (flags
& (FS_COMPR_FL
| FS_NOCOMP_FL
)))
219 if ((flags
& FS_NOCOW_FL
) && (old_flags
& (FS_COMPR_FL
| FS_NOCOMP_FL
)))
225 static int check_fsflags_compatible(struct btrfs_fs_info
*fs_info
,
228 if (btrfs_is_zoned(fs_info
) && (flags
& FS_NOCOW_FL
))
235 * Set flags/xflags from the internal inode flags. The remaining items of
236 * fsxattr are zeroed.
238 int btrfs_fileattr_get(struct dentry
*dentry
, struct fileattr
*fa
)
240 struct btrfs_inode
*binode
= BTRFS_I(d_inode(dentry
));
242 fileattr_fill_flags(fa
, btrfs_inode_flags_to_fsflags(binode
));
246 int btrfs_fileattr_set(struct mnt_idmap
*idmap
,
247 struct dentry
*dentry
, struct fileattr
*fa
)
249 struct inode
*inode
= d_inode(dentry
);
250 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
251 struct btrfs_inode
*binode
= BTRFS_I(inode
);
252 struct btrfs_root
*root
= binode
->root
;
253 struct btrfs_trans_handle
*trans
;
254 unsigned int fsflags
, old_fsflags
;
256 const char *comp
= NULL
;
259 if (btrfs_root_readonly(root
))
262 if (fileattr_has_fsx(fa
))
265 fsflags
= btrfs_mask_fsflags_for_type(inode
, fa
->flags
);
266 old_fsflags
= btrfs_inode_flags_to_fsflags(binode
);
267 ret
= check_fsflags(old_fsflags
, fsflags
);
271 ret
= check_fsflags_compatible(fs_info
, fsflags
);
275 binode_flags
= binode
->flags
;
276 if (fsflags
& FS_SYNC_FL
)
277 binode_flags
|= BTRFS_INODE_SYNC
;
279 binode_flags
&= ~BTRFS_INODE_SYNC
;
280 if (fsflags
& FS_IMMUTABLE_FL
)
281 binode_flags
|= BTRFS_INODE_IMMUTABLE
;
283 binode_flags
&= ~BTRFS_INODE_IMMUTABLE
;
284 if (fsflags
& FS_APPEND_FL
)
285 binode_flags
|= BTRFS_INODE_APPEND
;
287 binode_flags
&= ~BTRFS_INODE_APPEND
;
288 if (fsflags
& FS_NODUMP_FL
)
289 binode_flags
|= BTRFS_INODE_NODUMP
;
291 binode_flags
&= ~BTRFS_INODE_NODUMP
;
292 if (fsflags
& FS_NOATIME_FL
)
293 binode_flags
|= BTRFS_INODE_NOATIME
;
295 binode_flags
&= ~BTRFS_INODE_NOATIME
;
297 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
298 if (!fa
->flags_valid
) {
299 /* 1 item for the inode */
300 trans
= btrfs_start_transaction(root
, 1);
302 return PTR_ERR(trans
);
306 if (fsflags
& FS_DIRSYNC_FL
)
307 binode_flags
|= BTRFS_INODE_DIRSYNC
;
309 binode_flags
&= ~BTRFS_INODE_DIRSYNC
;
310 if (fsflags
& FS_NOCOW_FL
) {
311 if (S_ISREG(inode
->i_mode
)) {
313 * It's safe to turn csums off here, no extents exist.
314 * Otherwise we want the flag to reflect the real COW
315 * status of the file and will not set it.
317 if (inode
->i_size
== 0)
318 binode_flags
|= BTRFS_INODE_NODATACOW
|
319 BTRFS_INODE_NODATASUM
;
321 binode_flags
|= BTRFS_INODE_NODATACOW
;
325 * Revert back under same assumptions as above
327 if (S_ISREG(inode
->i_mode
)) {
328 if (inode
->i_size
== 0)
329 binode_flags
&= ~(BTRFS_INODE_NODATACOW
|
330 BTRFS_INODE_NODATASUM
);
332 binode_flags
&= ~BTRFS_INODE_NODATACOW
;
337 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
338 * flag may be changed automatically if compression code won't make
341 if (fsflags
& FS_NOCOMP_FL
) {
342 binode_flags
&= ~BTRFS_INODE_COMPRESS
;
343 binode_flags
|= BTRFS_INODE_NOCOMPRESS
;
344 } else if (fsflags
& FS_COMPR_FL
) {
346 if (IS_SWAPFILE(inode
))
349 binode_flags
|= BTRFS_INODE_COMPRESS
;
350 binode_flags
&= ~BTRFS_INODE_NOCOMPRESS
;
352 comp
= btrfs_compress_type2str(fs_info
->compress_type
);
353 if (!comp
|| comp
[0] == 0)
354 comp
= btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB
);
356 binode_flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
363 trans
= btrfs_start_transaction(root
, 3);
365 return PTR_ERR(trans
);
368 ret
= btrfs_set_prop(trans
, inode
, "btrfs.compression", comp
,
371 btrfs_abort_transaction(trans
, ret
);
375 ret
= btrfs_set_prop(trans
, inode
, "btrfs.compression", NULL
,
377 if (ret
&& ret
!= -ENODATA
) {
378 btrfs_abort_transaction(trans
, ret
);
384 binode
->flags
= binode_flags
;
385 btrfs_sync_inode_flags_to_i_flags(inode
);
386 inode_inc_iversion(inode
);
387 inode_set_ctime_current(inode
);
388 ret
= btrfs_update_inode(trans
, BTRFS_I(inode
));
391 btrfs_end_transaction(trans
);
396 * Start exclusive operation @type, return true on success
398 bool btrfs_exclop_start(struct btrfs_fs_info
*fs_info
,
399 enum btrfs_exclusive_operation type
)
403 spin_lock(&fs_info
->super_lock
);
404 if (fs_info
->exclusive_operation
== BTRFS_EXCLOP_NONE
) {
405 fs_info
->exclusive_operation
= type
;
408 spin_unlock(&fs_info
->super_lock
);
414 * Conditionally allow to enter the exclusive operation in case it's compatible
415 * with the running one. This must be paired with btrfs_exclop_start_unlock and
416 * btrfs_exclop_finish.
419 * - the same type is already running
420 * - when trying to add a device and balance has been paused
421 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
422 * must check the condition first that would allow none -> @type
424 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info
*fs_info
,
425 enum btrfs_exclusive_operation type
)
427 spin_lock(&fs_info
->super_lock
);
428 if (fs_info
->exclusive_operation
== type
||
429 (fs_info
->exclusive_operation
== BTRFS_EXCLOP_BALANCE_PAUSED
&&
430 type
== BTRFS_EXCLOP_DEV_ADD
))
433 spin_unlock(&fs_info
->super_lock
);
437 void btrfs_exclop_start_unlock(struct btrfs_fs_info
*fs_info
)
439 spin_unlock(&fs_info
->super_lock
);
442 void btrfs_exclop_finish(struct btrfs_fs_info
*fs_info
)
444 spin_lock(&fs_info
->super_lock
);
445 WRITE_ONCE(fs_info
->exclusive_operation
, BTRFS_EXCLOP_NONE
);
446 spin_unlock(&fs_info
->super_lock
);
447 sysfs_notify(&fs_info
->fs_devices
->fsid_kobj
, NULL
, "exclusive_operation");
450 void btrfs_exclop_balance(struct btrfs_fs_info
*fs_info
,
451 enum btrfs_exclusive_operation op
)
454 case BTRFS_EXCLOP_BALANCE_PAUSED
:
455 spin_lock(&fs_info
->super_lock
);
456 ASSERT(fs_info
->exclusive_operation
== BTRFS_EXCLOP_BALANCE
||
457 fs_info
->exclusive_operation
== BTRFS_EXCLOP_DEV_ADD
||
458 fs_info
->exclusive_operation
== BTRFS_EXCLOP_NONE
||
459 fs_info
->exclusive_operation
== BTRFS_EXCLOP_BALANCE_PAUSED
);
460 fs_info
->exclusive_operation
= BTRFS_EXCLOP_BALANCE_PAUSED
;
461 spin_unlock(&fs_info
->super_lock
);
463 case BTRFS_EXCLOP_BALANCE
:
464 spin_lock(&fs_info
->super_lock
);
465 ASSERT(fs_info
->exclusive_operation
== BTRFS_EXCLOP_BALANCE_PAUSED
);
466 fs_info
->exclusive_operation
= BTRFS_EXCLOP_BALANCE
;
467 spin_unlock(&fs_info
->super_lock
);
471 "invalid exclop balance operation %d requested", op
);
475 static int btrfs_ioctl_getversion(struct inode
*inode
, int __user
*arg
)
477 return put_user(inode
->i_generation
, arg
);
480 static noinline
int btrfs_ioctl_fitrim(struct btrfs_fs_info
*fs_info
,
483 struct btrfs_device
*device
;
484 struct fstrim_range range
;
485 u64 minlen
= ULLONG_MAX
;
489 if (!capable(CAP_SYS_ADMIN
))
493 * btrfs_trim_block_group() depends on space cache, which is not
494 * available in zoned filesystem. So, disallow fitrim on a zoned
495 * filesystem for now.
497 if (btrfs_is_zoned(fs_info
))
501 * If the fs is mounted with nologreplay, which requires it to be
502 * mounted in RO mode as well, we can not allow discard on free space
503 * inside block groups, because log trees refer to extents that are not
504 * pinned in a block group's free space cache (pinning the extents is
505 * precisely the first phase of replaying a log tree).
507 if (btrfs_test_opt(fs_info
, NOLOGREPLAY
))
511 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
513 if (!device
->bdev
|| !bdev_max_discard_sectors(device
->bdev
))
516 minlen
= min_t(u64
, bdev_discard_granularity(device
->bdev
),
523 if (copy_from_user(&range
, arg
, sizeof(range
)))
527 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
528 * block group is in the logical address space, which can be any
529 * sectorsize aligned bytenr in the range [0, U64_MAX].
531 if (range
.len
< fs_info
->sb
->s_blocksize
)
534 range
.minlen
= max(range
.minlen
, minlen
);
535 ret
= btrfs_trim_fs(fs_info
, &range
);
539 if (copy_to_user(arg
, &range
, sizeof(range
)))
545 int __pure
btrfs_is_empty_uuid(u8
*uuid
)
549 for (i
= 0; i
< BTRFS_UUID_SIZE
; i
++) {
557 * Calculate the number of transaction items to reserve for creating a subvolume
558 * or snapshot, not including the inode, directory entries, or parent directory.
560 static unsigned int create_subvol_num_items(struct btrfs_qgroup_inherit
*inherit
)
563 * 1 to add root block
566 * 1 to add root backref
568 * 1 to add qgroup info
569 * 1 to add qgroup limit
571 * Ideally the last two would only be accounted if qgroups are enabled,
572 * but that can change between now and the time we would insert them.
574 unsigned int num_items
= 7;
577 /* 2 to add qgroup relations for each inherited qgroup */
578 num_items
+= 2 * inherit
->num_qgroups
;
583 static noinline
int create_subvol(struct mnt_idmap
*idmap
,
584 struct inode
*dir
, struct dentry
*dentry
,
585 struct btrfs_qgroup_inherit
*inherit
)
587 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
588 struct btrfs_trans_handle
*trans
;
589 struct btrfs_key key
;
590 struct btrfs_root_item
*root_item
;
591 struct btrfs_inode_item
*inode_item
;
592 struct extent_buffer
*leaf
;
593 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
594 struct btrfs_root
*new_root
;
595 struct btrfs_block_rsv block_rsv
;
596 struct timespec64 cur_time
= current_time(dir
);
597 struct btrfs_new_inode_args new_inode_args
= {
602 unsigned int trans_num_items
;
607 root_item
= kzalloc(sizeof(*root_item
), GFP_KERNEL
);
611 ret
= btrfs_get_free_objectid(fs_info
->tree_root
, &objectid
);
616 * Don't create subvolume whose level is not zero. Or qgroup will be
617 * screwed up since it assumes subvolume qgroup's level to be 0.
619 if (btrfs_qgroup_level(objectid
)) {
624 ret
= get_anon_bdev(&anon_dev
);
628 new_inode_args
.inode
= btrfs_new_subvol_inode(idmap
, dir
);
629 if (!new_inode_args
.inode
) {
633 ret
= btrfs_new_inode_prepare(&new_inode_args
, &trans_num_items
);
636 trans_num_items
+= create_subvol_num_items(inherit
);
638 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
639 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
640 trans_num_items
, false);
642 goto out_new_inode_args
;
644 trans
= btrfs_start_transaction(root
, 0);
646 ret
= PTR_ERR(trans
);
647 btrfs_subvolume_release_metadata(root
, &block_rsv
);
648 goto out_new_inode_args
;
650 trans
->block_rsv
= &block_rsv
;
651 trans
->bytes_reserved
= block_rsv
.size
;
652 /* Tree log can't currently deal with an inode which is a new root. */
653 btrfs_set_log_full_commit(trans
);
655 ret
= btrfs_qgroup_inherit(trans
, 0, objectid
, root
->root_key
.objectid
, inherit
);
659 leaf
= btrfs_alloc_tree_block(trans
, root
, 0, objectid
, NULL
, 0, 0, 0,
660 0, BTRFS_NESTING_NORMAL
);
666 btrfs_mark_buffer_dirty(trans
, leaf
);
668 inode_item
= &root_item
->inode
;
669 btrfs_set_stack_inode_generation(inode_item
, 1);
670 btrfs_set_stack_inode_size(inode_item
, 3);
671 btrfs_set_stack_inode_nlink(inode_item
, 1);
672 btrfs_set_stack_inode_nbytes(inode_item
,
674 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
676 btrfs_set_root_flags(root_item
, 0);
677 btrfs_set_root_limit(root_item
, 0);
678 btrfs_set_stack_inode_flags(inode_item
, BTRFS_INODE_ROOT_ITEM_INIT
);
680 btrfs_set_root_bytenr(root_item
, leaf
->start
);
681 btrfs_set_root_generation(root_item
, trans
->transid
);
682 btrfs_set_root_level(root_item
, 0);
683 btrfs_set_root_refs(root_item
, 1);
684 btrfs_set_root_used(root_item
, leaf
->len
);
685 btrfs_set_root_last_snapshot(root_item
, 0);
687 btrfs_set_root_generation_v2(root_item
,
688 btrfs_root_generation(root_item
));
689 generate_random_guid(root_item
->uuid
);
690 btrfs_set_stack_timespec_sec(&root_item
->otime
, cur_time
.tv_sec
);
691 btrfs_set_stack_timespec_nsec(&root_item
->otime
, cur_time
.tv_nsec
);
692 root_item
->ctime
= root_item
->otime
;
693 btrfs_set_root_ctransid(root_item
, trans
->transid
);
694 btrfs_set_root_otransid(root_item
, trans
->transid
);
696 btrfs_tree_unlock(leaf
);
698 btrfs_set_root_dirid(root_item
, BTRFS_FIRST_FREE_OBJECTID
);
700 key
.objectid
= objectid
;
702 key
.type
= BTRFS_ROOT_ITEM_KEY
;
703 ret
= btrfs_insert_root(trans
, fs_info
->tree_root
, &key
,
707 * Since we don't abort the transaction in this case, free the
708 * tree block so that we don't leak space and leave the
709 * filesystem in an inconsistent state (an extent item in the
710 * extent tree with a backreference for a root that does not
713 btrfs_tree_lock(leaf
);
714 btrfs_clear_buffer_dirty(trans
, leaf
);
715 btrfs_tree_unlock(leaf
);
716 btrfs_free_tree_block(trans
, objectid
, leaf
, 0, 1);
717 free_extent_buffer(leaf
);
721 free_extent_buffer(leaf
);
724 new_root
= btrfs_get_new_fs_root(fs_info
, objectid
, anon_dev
);
725 if (IS_ERR(new_root
)) {
726 ret
= PTR_ERR(new_root
);
727 btrfs_abort_transaction(trans
, ret
);
730 /* anon_dev is owned by new_root now. */
732 BTRFS_I(new_inode_args
.inode
)->root
= new_root
;
733 /* ... and new_root is owned by new_inode_args.inode now. */
735 ret
= btrfs_record_root_in_trans(trans
, new_root
);
737 btrfs_abort_transaction(trans
, ret
);
741 ret
= btrfs_uuid_tree_add(trans
, root_item
->uuid
,
742 BTRFS_UUID_KEY_SUBVOL
, objectid
);
744 btrfs_abort_transaction(trans
, ret
);
748 ret
= btrfs_create_new_inode(trans
, &new_inode_args
);
750 btrfs_abort_transaction(trans
, ret
);
754 d_instantiate_new(dentry
, new_inode_args
.inode
);
755 new_inode_args
.inode
= NULL
;
758 trans
->block_rsv
= NULL
;
759 trans
->bytes_reserved
= 0;
760 btrfs_subvolume_release_metadata(root
, &block_rsv
);
762 btrfs_end_transaction(trans
);
764 btrfs_new_inode_args_destroy(&new_inode_args
);
766 iput(new_inode_args
.inode
);
769 free_anon_bdev(anon_dev
);
775 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
776 struct dentry
*dentry
, bool readonly
,
777 struct btrfs_qgroup_inherit
*inherit
)
779 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
781 struct btrfs_pending_snapshot
*pending_snapshot
;
782 unsigned int trans_num_items
;
783 struct btrfs_trans_handle
*trans
;
786 /* We do not support snapshotting right now. */
787 if (btrfs_fs_incompat(fs_info
, EXTENT_TREE_V2
)) {
789 "extent tree v2 doesn't support snapshotting yet");
793 if (btrfs_root_refs(&root
->root_item
) == 0)
796 if (!test_bit(BTRFS_ROOT_SHAREABLE
, &root
->state
))
799 if (atomic_read(&root
->nr_swapfiles
)) {
801 "cannot snapshot subvolume with active swapfile");
805 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_KERNEL
);
806 if (!pending_snapshot
)
809 ret
= get_anon_bdev(&pending_snapshot
->anon_dev
);
812 pending_snapshot
->root_item
= kzalloc(sizeof(struct btrfs_root_item
),
814 pending_snapshot
->path
= btrfs_alloc_path();
815 if (!pending_snapshot
->root_item
|| !pending_snapshot
->path
) {
820 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
821 BTRFS_BLOCK_RSV_TEMP
);
825 * 1 to update parent inode item
827 trans_num_items
= create_subvol_num_items(inherit
) + 3;
828 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
829 &pending_snapshot
->block_rsv
,
830 trans_num_items
, false);
834 pending_snapshot
->dentry
= dentry
;
835 pending_snapshot
->root
= root
;
836 pending_snapshot
->readonly
= readonly
;
837 pending_snapshot
->dir
= dir
;
838 pending_snapshot
->inherit
= inherit
;
840 trans
= btrfs_start_transaction(root
, 0);
842 ret
= PTR_ERR(trans
);
846 trans
->pending_snapshot
= pending_snapshot
;
848 ret
= btrfs_commit_transaction(trans
);
852 ret
= pending_snapshot
->error
;
856 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
860 inode
= btrfs_lookup_dentry(d_inode(dentry
->d_parent
), dentry
);
862 ret
= PTR_ERR(inode
);
866 d_instantiate(dentry
, inode
);
868 pending_snapshot
->anon_dev
= 0;
870 /* Prevent double freeing of anon_dev */
871 if (ret
&& pending_snapshot
->snap
)
872 pending_snapshot
->snap
->anon_dev
= 0;
873 btrfs_put_root(pending_snapshot
->snap
);
874 btrfs_subvolume_release_metadata(root
, &pending_snapshot
->block_rsv
);
876 if (pending_snapshot
->anon_dev
)
877 free_anon_bdev(pending_snapshot
->anon_dev
);
878 kfree(pending_snapshot
->root_item
);
879 btrfs_free_path(pending_snapshot
->path
);
880 kfree(pending_snapshot
);
885 /* copy of may_delete in fs/namei.c()
886 * Check whether we can remove a link victim from directory dir, check
887 * whether the type of victim is right.
888 * 1. We can't do it if dir is read-only (done in permission())
889 * 2. We should have write and exec permissions on dir
890 * 3. We can't remove anything from append-only dir
891 * 4. We can't do anything with immutable dir (done in permission())
892 * 5. If the sticky bit on dir is set we should either
893 * a. be owner of dir, or
894 * b. be owner of victim, or
895 * c. have CAP_FOWNER capability
896 * 6. If the victim is append-only or immutable we can't do anything with
897 * links pointing to it.
898 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
899 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
900 * 9. We can't remove a root or mountpoint.
901 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
902 * nfs_async_unlink().
905 static int btrfs_may_delete(struct mnt_idmap
*idmap
,
906 struct inode
*dir
, struct dentry
*victim
, int isdir
)
910 if (d_really_is_negative(victim
))
913 BUG_ON(d_inode(victim
->d_parent
) != dir
);
914 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
916 error
= inode_permission(idmap
, dir
, MAY_WRITE
| MAY_EXEC
);
921 if (check_sticky(idmap
, dir
, d_inode(victim
)) ||
922 IS_APPEND(d_inode(victim
)) || IS_IMMUTABLE(d_inode(victim
)) ||
923 IS_SWAPFILE(d_inode(victim
)))
926 if (!d_is_dir(victim
))
930 } else if (d_is_dir(victim
))
934 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
939 /* copy of may_create in fs/namei.c() */
940 static inline int btrfs_may_create(struct mnt_idmap
*idmap
,
941 struct inode
*dir
, struct dentry
*child
)
943 if (d_really_is_positive(child
))
947 if (!fsuidgid_has_mapping(dir
->i_sb
, idmap
))
949 return inode_permission(idmap
, dir
, MAY_WRITE
| MAY_EXEC
);
953 * Create a new subvolume below @parent. This is largely modeled after
954 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
955 * inside this filesystem so it's quite a bit simpler.
957 static noinline
int btrfs_mksubvol(const struct path
*parent
,
958 struct mnt_idmap
*idmap
,
959 const char *name
, int namelen
,
960 struct btrfs_root
*snap_src
,
962 struct btrfs_qgroup_inherit
*inherit
)
964 struct inode
*dir
= d_inode(parent
->dentry
);
965 struct btrfs_fs_info
*fs_info
= btrfs_sb(dir
->i_sb
);
966 struct dentry
*dentry
;
967 struct fscrypt_str name_str
= FSTR_INIT((char *)name
, namelen
);
970 error
= down_write_killable_nested(&dir
->i_rwsem
, I_MUTEX_PARENT
);
974 dentry
= lookup_one(idmap
, name
, parent
->dentry
, namelen
);
975 error
= PTR_ERR(dentry
);
979 error
= btrfs_may_create(idmap
, dir
, dentry
);
984 * even if this name doesn't exist, we may get hash collisions.
985 * check for them now when we can safely fail
987 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
988 dir
->i_ino
, &name_str
);
992 down_read(&fs_info
->subvol_sem
);
994 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
998 error
= create_snapshot(snap_src
, dir
, dentry
, readonly
, inherit
);
1000 error
= create_subvol(idmap
, dir
, dentry
, inherit
);
1003 fsnotify_mkdir(dir
, dentry
);
1005 up_read(&fs_info
->subvol_sem
);
1009 btrfs_inode_unlock(BTRFS_I(dir
), 0);
1013 static noinline
int btrfs_mksnapshot(const struct path
*parent
,
1014 struct mnt_idmap
*idmap
,
1015 const char *name
, int namelen
,
1016 struct btrfs_root
*root
,
1018 struct btrfs_qgroup_inherit
*inherit
)
1021 bool snapshot_force_cow
= false;
1024 * Force new buffered writes to reserve space even when NOCOW is
1025 * possible. This is to avoid later writeback (running dealloc) to
1026 * fallback to COW mode and unexpectedly fail with ENOSPC.
1028 btrfs_drew_read_lock(&root
->snapshot_lock
);
1030 ret
= btrfs_start_delalloc_snapshot(root
, false);
1035 * All previous writes have started writeback in NOCOW mode, so now
1036 * we force future writes to fallback to COW mode during snapshot
1039 atomic_inc(&root
->snapshot_force_cow
);
1040 snapshot_force_cow
= true;
1042 btrfs_wait_ordered_extents(root
, U64_MAX
, 0, (u64
)-1);
1044 ret
= btrfs_mksubvol(parent
, idmap
, name
, namelen
,
1045 root
, readonly
, inherit
);
1047 if (snapshot_force_cow
)
1048 atomic_dec(&root
->snapshot_force_cow
);
1049 btrfs_drew_read_unlock(&root
->snapshot_lock
);
1054 * Try to start exclusive operation @type or cancel it if it's running.
1057 * 0 - normal mode, newly claimed op started
1058 * >0 - normal mode, something else is running,
1059 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1060 * ECANCELED - cancel mode, successful cancel
1061 * ENOTCONN - cancel mode, operation not running anymore
1063 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info
*fs_info
,
1064 enum btrfs_exclusive_operation type
, bool cancel
)
1067 /* Start normal op */
1068 if (!btrfs_exclop_start(fs_info
, type
))
1069 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
1070 /* Exclusive operation is now claimed */
1074 /* Cancel running op */
1075 if (btrfs_exclop_start_try_lock(fs_info
, type
)) {
1077 * This blocks any exclop finish from setting it to NONE, so we
1078 * request cancellation. Either it runs and we will wait for it,
1079 * or it has finished and no waiting will happen.
1081 atomic_inc(&fs_info
->reloc_cancel_req
);
1082 btrfs_exclop_start_unlock(fs_info
);
1084 if (test_bit(BTRFS_FS_RELOC_RUNNING
, &fs_info
->flags
))
1085 wait_on_bit(&fs_info
->flags
, BTRFS_FS_RELOC_RUNNING
,
1086 TASK_INTERRUPTIBLE
);
1091 /* Something else is running or none */
1095 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1098 BTRFS_DEV_LOOKUP_ARGS(args
);
1099 struct inode
*inode
= file_inode(file
);
1100 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1104 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1105 struct btrfs_ioctl_vol_args
*vol_args
;
1106 struct btrfs_trans_handle
*trans
;
1107 struct btrfs_device
*device
= NULL
;
1110 char *devstr
= NULL
;
1115 if (!capable(CAP_SYS_ADMIN
))
1118 ret
= mnt_want_write_file(file
);
1123 * Read the arguments before checking exclusivity to be able to
1124 * distinguish regular resize and cancel
1126 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1127 if (IS_ERR(vol_args
)) {
1128 ret
= PTR_ERR(vol_args
);
1131 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1132 sizestr
= vol_args
->name
;
1133 cancel
= (strcmp("cancel", sizestr
) == 0);
1134 ret
= exclop_start_or_cancel_reloc(fs_info
, BTRFS_EXCLOP_RESIZE
, cancel
);
1137 /* Exclusive operation is now claimed */
1139 devstr
= strchr(sizestr
, ':');
1141 sizestr
= devstr
+ 1;
1143 devstr
= vol_args
->name
;
1144 ret
= kstrtoull(devstr
, 10, &devid
);
1151 btrfs_info(fs_info
, "resizing devid %llu", devid
);
1155 device
= btrfs_find_device(fs_info
->fs_devices
, &args
);
1157 btrfs_info(fs_info
, "resizer unable to find device %llu",
1163 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE
, &device
->dev_state
)) {
1165 "resizer unable to apply on readonly device %llu",
1171 if (!strcmp(sizestr
, "max"))
1172 new_size
= bdev_nr_bytes(device
->bdev
);
1174 if (sizestr
[0] == '-') {
1177 } else if (sizestr
[0] == '+') {
1181 new_size
= memparse(sizestr
, &retptr
);
1182 if (*retptr
!= '\0' || new_size
== 0) {
1188 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT
, &device
->dev_state
)) {
1193 old_size
= btrfs_device_get_total_bytes(device
);
1196 if (new_size
> old_size
) {
1200 new_size
= old_size
- new_size
;
1201 } else if (mod
> 0) {
1202 if (new_size
> ULLONG_MAX
- old_size
) {
1206 new_size
= old_size
+ new_size
;
1209 if (new_size
< SZ_256M
) {
1213 if (new_size
> bdev_nr_bytes(device
->bdev
)) {
1218 new_size
= round_down(new_size
, fs_info
->sectorsize
);
1220 if (new_size
> old_size
) {
1221 trans
= btrfs_start_transaction(root
, 0);
1222 if (IS_ERR(trans
)) {
1223 ret
= PTR_ERR(trans
);
1226 ret
= btrfs_grow_device(trans
, device
, new_size
);
1227 btrfs_commit_transaction(trans
);
1228 } else if (new_size
< old_size
) {
1229 ret
= btrfs_shrink_device(device
, new_size
);
1230 } /* equal, nothing need to do */
1232 if (ret
== 0 && new_size
!= old_size
)
1233 btrfs_info_in_rcu(fs_info
,
1234 "resize device %s (devid %llu) from %llu to %llu",
1235 btrfs_dev_name(device
), device
->devid
,
1236 old_size
, new_size
);
1238 btrfs_exclop_finish(fs_info
);
1242 mnt_drop_write_file(file
);
1246 static noinline
int __btrfs_ioctl_snap_create(struct file
*file
,
1247 struct mnt_idmap
*idmap
,
1248 const char *name
, unsigned long fd
, int subvol
,
1250 struct btrfs_qgroup_inherit
*inherit
)
1255 if (!S_ISDIR(file_inode(file
)->i_mode
))
1258 ret
= mnt_want_write_file(file
);
1262 namelen
= strlen(name
);
1263 if (strchr(name
, '/')) {
1265 goto out_drop_write
;
1268 if (name
[0] == '.' &&
1269 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1271 goto out_drop_write
;
1275 ret
= btrfs_mksubvol(&file
->f_path
, idmap
, name
,
1276 namelen
, NULL
, readonly
, inherit
);
1278 struct fd src
= fdget(fd
);
1279 struct inode
*src_inode
;
1282 goto out_drop_write
;
1285 src_inode
= file_inode(src
.file
);
1286 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1287 btrfs_info(BTRFS_I(file_inode(file
))->root
->fs_info
,
1288 "Snapshot src from another FS");
1290 } else if (!inode_owner_or_capable(idmap
, src_inode
)) {
1292 * Subvolume creation is not restricted, but snapshots
1293 * are limited to own subvolumes only
1297 ret
= btrfs_mksnapshot(&file
->f_path
, idmap
,
1299 BTRFS_I(src_inode
)->root
,
1305 mnt_drop_write_file(file
);
1310 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1311 void __user
*arg
, int subvol
)
1313 struct btrfs_ioctl_vol_args
*vol_args
;
1316 if (!S_ISDIR(file_inode(file
)->i_mode
))
1319 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1320 if (IS_ERR(vol_args
))
1321 return PTR_ERR(vol_args
);
1322 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1324 ret
= __btrfs_ioctl_snap_create(file
, file_mnt_idmap(file
),
1325 vol_args
->name
, vol_args
->fd
, subvol
,
1332 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1333 void __user
*arg
, int subvol
)
1335 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1337 bool readonly
= false;
1338 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1340 if (!S_ISDIR(file_inode(file
)->i_mode
))
1343 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1344 if (IS_ERR(vol_args
))
1345 return PTR_ERR(vol_args
);
1346 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1348 if (vol_args
->flags
& ~BTRFS_SUBVOL_CREATE_ARGS_MASK
) {
1353 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1355 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1358 if (vol_args
->size
< sizeof(*inherit
) ||
1359 vol_args
->size
> PAGE_SIZE
) {
1363 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1364 if (IS_ERR(inherit
)) {
1365 ret
= PTR_ERR(inherit
);
1369 if (inherit
->num_qgroups
> PAGE_SIZE
||
1370 inherit
->num_ref_copies
> PAGE_SIZE
||
1371 inherit
->num_excl_copies
> PAGE_SIZE
) {
1376 nums
= inherit
->num_qgroups
+ 2 * inherit
->num_ref_copies
+
1377 2 * inherit
->num_excl_copies
;
1378 if (vol_args
->size
!= struct_size(inherit
, qgroups
, nums
)) {
1384 ret
= __btrfs_ioctl_snap_create(file
, file_mnt_idmap(file
),
1385 vol_args
->name
, vol_args
->fd
, subvol
,
1396 static noinline
int btrfs_ioctl_subvol_getflags(struct inode
*inode
,
1399 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1400 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1404 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
)
1407 down_read(&fs_info
->subvol_sem
);
1408 if (btrfs_root_readonly(root
))
1409 flags
|= BTRFS_SUBVOL_RDONLY
;
1410 up_read(&fs_info
->subvol_sem
);
1412 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1418 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1421 struct inode
*inode
= file_inode(file
);
1422 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
1423 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1424 struct btrfs_trans_handle
*trans
;
1429 if (!inode_owner_or_capable(file_mnt_idmap(file
), inode
))
1432 ret
= mnt_want_write_file(file
);
1436 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
) {
1438 goto out_drop_write
;
1441 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1443 goto out_drop_write
;
1446 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1448 goto out_drop_write
;
1451 down_write(&fs_info
->subvol_sem
);
1454 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1457 root_flags
= btrfs_root_flags(&root
->root_item
);
1458 if (flags
& BTRFS_SUBVOL_RDONLY
) {
1459 btrfs_set_root_flags(&root
->root_item
,
1460 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1463 * Block RO -> RW transition if this subvolume is involved in
1466 spin_lock(&root
->root_item_lock
);
1467 if (root
->send_in_progress
== 0) {
1468 btrfs_set_root_flags(&root
->root_item
,
1469 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1470 spin_unlock(&root
->root_item_lock
);
1472 spin_unlock(&root
->root_item_lock
);
1474 "Attempt to set subvolume %llu read-write during send",
1475 root
->root_key
.objectid
);
1481 trans
= btrfs_start_transaction(root
, 1);
1482 if (IS_ERR(trans
)) {
1483 ret
= PTR_ERR(trans
);
1487 ret
= btrfs_update_root(trans
, fs_info
->tree_root
,
1488 &root
->root_key
, &root
->root_item
);
1490 btrfs_end_transaction(trans
);
1494 ret
= btrfs_commit_transaction(trans
);
1498 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1500 up_write(&fs_info
->subvol_sem
);
1502 mnt_drop_write_file(file
);
1507 static noinline
int key_in_sk(struct btrfs_key
*key
,
1508 struct btrfs_ioctl_search_key
*sk
)
1510 struct btrfs_key test
;
1513 test
.objectid
= sk
->min_objectid
;
1514 test
.type
= sk
->min_type
;
1515 test
.offset
= sk
->min_offset
;
1517 ret
= btrfs_comp_cpu_keys(key
, &test
);
1521 test
.objectid
= sk
->max_objectid
;
1522 test
.type
= sk
->max_type
;
1523 test
.offset
= sk
->max_offset
;
1525 ret
= btrfs_comp_cpu_keys(key
, &test
);
1531 static noinline
int copy_to_sk(struct btrfs_path
*path
,
1532 struct btrfs_key
*key
,
1533 struct btrfs_ioctl_search_key
*sk
,
1536 unsigned long *sk_offset
,
1540 struct extent_buffer
*leaf
;
1541 struct btrfs_ioctl_search_header sh
;
1542 struct btrfs_key test
;
1543 unsigned long item_off
;
1544 unsigned long item_len
;
1550 leaf
= path
->nodes
[0];
1551 slot
= path
->slots
[0];
1552 nritems
= btrfs_header_nritems(leaf
);
1554 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1558 found_transid
= btrfs_header_generation(leaf
);
1560 for (i
= slot
; i
< nritems
; i
++) {
1561 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1562 item_len
= btrfs_item_size(leaf
, i
);
1564 btrfs_item_key_to_cpu(leaf
, key
, i
);
1565 if (!key_in_sk(key
, sk
))
1568 if (sizeof(sh
) + item_len
> *buf_size
) {
1575 * return one empty item back for v1, which does not
1579 *buf_size
= sizeof(sh
) + item_len
;
1584 if (sizeof(sh
) + item_len
+ *sk_offset
> *buf_size
) {
1589 sh
.objectid
= key
->objectid
;
1590 sh
.offset
= key
->offset
;
1591 sh
.type
= key
->type
;
1593 sh
.transid
= found_transid
;
1596 * Copy search result header. If we fault then loop again so we
1597 * can fault in the pages and -EFAULT there if there's a
1598 * problem. Otherwise we'll fault and then copy the buffer in
1599 * properly this next time through
1601 if (copy_to_user_nofault(ubuf
+ *sk_offset
, &sh
, sizeof(sh
))) {
1606 *sk_offset
+= sizeof(sh
);
1609 char __user
*up
= ubuf
+ *sk_offset
;
1611 * Copy the item, same behavior as above, but reset the
1612 * * sk_offset so we copy the full thing again.
1614 if (read_extent_buffer_to_user_nofault(leaf
, up
,
1615 item_off
, item_len
)) {
1617 *sk_offset
-= sizeof(sh
);
1621 *sk_offset
+= item_len
;
1625 if (ret
) /* -EOVERFLOW from above */
1628 if (*num_found
>= sk
->nr_items
) {
1635 test
.objectid
= sk
->max_objectid
;
1636 test
.type
= sk
->max_type
;
1637 test
.offset
= sk
->max_offset
;
1638 if (btrfs_comp_cpu_keys(key
, &test
) >= 0)
1640 else if (key
->offset
< (u64
)-1)
1642 else if (key
->type
< (u8
)-1) {
1645 } else if (key
->objectid
< (u64
)-1) {
1653 * 0: all items from this leaf copied, continue with next
1654 * 1: * more items can be copied, but unused buffer is too small
1655 * * all items were found
1656 * Either way, it will stops the loop which iterates to the next
1658 * -EOVERFLOW: item was to large for buffer
1659 * -EFAULT: could not copy extent buffer back to userspace
1664 static noinline
int search_ioctl(struct inode
*inode
,
1665 struct btrfs_ioctl_search_key
*sk
,
1669 struct btrfs_fs_info
*info
= btrfs_sb(inode
->i_sb
);
1670 struct btrfs_root
*root
;
1671 struct btrfs_key key
;
1672 struct btrfs_path
*path
;
1675 unsigned long sk_offset
= 0;
1677 if (*buf_size
< sizeof(struct btrfs_ioctl_search_header
)) {
1678 *buf_size
= sizeof(struct btrfs_ioctl_search_header
);
1682 path
= btrfs_alloc_path();
1686 if (sk
->tree_id
== 0) {
1687 /* search the root of the inode that was passed */
1688 root
= btrfs_grab_root(BTRFS_I(inode
)->root
);
1690 root
= btrfs_get_fs_root(info
, sk
->tree_id
, true);
1692 btrfs_free_path(path
);
1693 return PTR_ERR(root
);
1697 key
.objectid
= sk
->min_objectid
;
1698 key
.type
= sk
->min_type
;
1699 key
.offset
= sk
->min_offset
;
1704 * Ensure that the whole user buffer is faulted in at sub-page
1705 * granularity, otherwise the loop may live-lock.
1707 if (fault_in_subpage_writeable(ubuf
+ sk_offset
,
1708 *buf_size
- sk_offset
))
1711 ret
= btrfs_search_forward(root
, &key
, path
, sk
->min_transid
);
1717 ret
= copy_to_sk(path
, &key
, sk
, buf_size
, ubuf
,
1718 &sk_offset
, &num_found
);
1719 btrfs_release_path(path
);
1727 sk
->nr_items
= num_found
;
1728 btrfs_put_root(root
);
1729 btrfs_free_path(path
);
1733 static noinline
int btrfs_ioctl_tree_search(struct inode
*inode
,
1736 struct btrfs_ioctl_search_args __user
*uargs
= argp
;
1737 struct btrfs_ioctl_search_key sk
;
1741 if (!capable(CAP_SYS_ADMIN
))
1744 if (copy_from_user(&sk
, &uargs
->key
, sizeof(sk
)))
1747 buf_size
= sizeof(uargs
->buf
);
1749 ret
= search_ioctl(inode
, &sk
, &buf_size
, uargs
->buf
);
1752 * In the origin implementation an overflow is handled by returning a
1753 * search header with a len of zero, so reset ret.
1755 if (ret
== -EOVERFLOW
)
1758 if (ret
== 0 && copy_to_user(&uargs
->key
, &sk
, sizeof(sk
)))
1763 static noinline
int btrfs_ioctl_tree_search_v2(struct inode
*inode
,
1766 struct btrfs_ioctl_search_args_v2 __user
*uarg
= argp
;
1767 struct btrfs_ioctl_search_args_v2 args
;
1770 const u64 buf_limit
= SZ_16M
;
1772 if (!capable(CAP_SYS_ADMIN
))
1775 /* copy search header and buffer size */
1776 if (copy_from_user(&args
, uarg
, sizeof(args
)))
1779 buf_size
= args
.buf_size
;
1781 /* limit result size to 16MB */
1782 if (buf_size
> buf_limit
)
1783 buf_size
= buf_limit
;
1785 ret
= search_ioctl(inode
, &args
.key
, &buf_size
,
1786 (char __user
*)(&uarg
->buf
[0]));
1787 if (ret
== 0 && copy_to_user(&uarg
->key
, &args
.key
, sizeof(args
.key
)))
1789 else if (ret
== -EOVERFLOW
&&
1790 copy_to_user(&uarg
->buf_size
, &buf_size
, sizeof(buf_size
)))
1797 * Search INODE_REFs to identify path name of 'dirid' directory
1798 * in a 'tree_id' tree. and sets path name to 'name'.
1800 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1801 u64 tree_id
, u64 dirid
, char *name
)
1803 struct btrfs_root
*root
;
1804 struct btrfs_key key
;
1810 struct btrfs_inode_ref
*iref
;
1811 struct extent_buffer
*l
;
1812 struct btrfs_path
*path
;
1814 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1819 path
= btrfs_alloc_path();
1823 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
- 1];
1825 root
= btrfs_get_fs_root(info
, tree_id
, true);
1827 ret
= PTR_ERR(root
);
1832 key
.objectid
= dirid
;
1833 key
.type
= BTRFS_INODE_REF_KEY
;
1834 key
.offset
= (u64
)-1;
1837 ret
= btrfs_search_backwards(root
, &key
, path
);
1846 slot
= path
->slots
[0];
1848 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
1849 len
= btrfs_inode_ref_name_len(l
, iref
);
1851 total_len
+= len
+ 1;
1853 ret
= -ENAMETOOLONG
;
1858 read_extent_buffer(l
, ptr
, (unsigned long)(iref
+ 1), len
);
1860 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
1863 btrfs_release_path(path
);
1864 key
.objectid
= key
.offset
;
1865 key
.offset
= (u64
)-1;
1866 dirid
= key
.objectid
;
1868 memmove(name
, ptr
, total_len
);
1869 name
[total_len
] = '\0';
1872 btrfs_put_root(root
);
1873 btrfs_free_path(path
);
1877 static int btrfs_search_path_in_tree_user(struct mnt_idmap
*idmap
,
1878 struct inode
*inode
,
1879 struct btrfs_ioctl_ino_lookup_user_args
*args
)
1881 struct btrfs_fs_info
*fs_info
= BTRFS_I(inode
)->root
->fs_info
;
1882 struct super_block
*sb
= inode
->i_sb
;
1883 struct btrfs_key upper_limit
= BTRFS_I(inode
)->location
;
1884 u64 treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
1885 u64 dirid
= args
->dirid
;
1886 unsigned long item_off
;
1887 unsigned long item_len
;
1888 struct btrfs_inode_ref
*iref
;
1889 struct btrfs_root_ref
*rref
;
1890 struct btrfs_root
*root
= NULL
;
1891 struct btrfs_path
*path
;
1892 struct btrfs_key key
, key2
;
1893 struct extent_buffer
*leaf
;
1894 struct inode
*temp_inode
;
1901 path
= btrfs_alloc_path();
1906 * If the bottom subvolume does not exist directly under upper_limit,
1907 * construct the path in from the bottom up.
1909 if (dirid
!= upper_limit
.objectid
) {
1910 ptr
= &args
->path
[BTRFS_INO_LOOKUP_USER_PATH_MAX
- 1];
1912 root
= btrfs_get_fs_root(fs_info
, treeid
, true);
1914 ret
= PTR_ERR(root
);
1918 key
.objectid
= dirid
;
1919 key
.type
= BTRFS_INODE_REF_KEY
;
1920 key
.offset
= (u64
)-1;
1922 ret
= btrfs_search_backwards(root
, &key
, path
);
1930 leaf
= path
->nodes
[0];
1931 slot
= path
->slots
[0];
1933 iref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_inode_ref
);
1934 len
= btrfs_inode_ref_name_len(leaf
, iref
);
1936 total_len
+= len
+ 1;
1937 if (ptr
< args
->path
) {
1938 ret
= -ENAMETOOLONG
;
1943 read_extent_buffer(leaf
, ptr
,
1944 (unsigned long)(iref
+ 1), len
);
1946 /* Check the read+exec permission of this directory */
1947 ret
= btrfs_previous_item(root
, path
, dirid
,
1948 BTRFS_INODE_ITEM_KEY
);
1951 } else if (ret
> 0) {
1956 leaf
= path
->nodes
[0];
1957 slot
= path
->slots
[0];
1958 btrfs_item_key_to_cpu(leaf
, &key2
, slot
);
1959 if (key2
.objectid
!= dirid
) {
1965 * We don't need the path anymore, so release it and
1966 * avoid deadlocks and lockdep warnings in case
1967 * btrfs_iget() needs to lookup the inode from its root
1968 * btree and lock the same leaf.
1970 btrfs_release_path(path
);
1971 temp_inode
= btrfs_iget(sb
, key2
.objectid
, root
);
1972 if (IS_ERR(temp_inode
)) {
1973 ret
= PTR_ERR(temp_inode
);
1976 ret
= inode_permission(idmap
, temp_inode
,
1977 MAY_READ
| MAY_EXEC
);
1984 if (key
.offset
== upper_limit
.objectid
)
1986 if (key
.objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
1991 key
.objectid
= key
.offset
;
1992 key
.offset
= (u64
)-1;
1993 dirid
= key
.objectid
;
1996 memmove(args
->path
, ptr
, total_len
);
1997 args
->path
[total_len
] = '\0';
1998 btrfs_put_root(root
);
2000 btrfs_release_path(path
);
2003 /* Get the bottom subvolume's name from ROOT_REF */
2004 key
.objectid
= treeid
;
2005 key
.type
= BTRFS_ROOT_REF_KEY
;
2006 key
.offset
= args
->treeid
;
2007 ret
= btrfs_search_slot(NULL
, fs_info
->tree_root
, &key
, path
, 0, 0);
2010 } else if (ret
> 0) {
2015 leaf
= path
->nodes
[0];
2016 slot
= path
->slots
[0];
2017 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2019 item_off
= btrfs_item_ptr_offset(leaf
, slot
);
2020 item_len
= btrfs_item_size(leaf
, slot
);
2021 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2022 rref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_root_ref
);
2023 if (args
->dirid
!= btrfs_root_ref_dirid(leaf
, rref
)) {
2028 /* Copy subvolume's name */
2029 item_off
+= sizeof(struct btrfs_root_ref
);
2030 item_len
-= sizeof(struct btrfs_root_ref
);
2031 read_extent_buffer(leaf
, args
->name
, item_off
, item_len
);
2032 args
->name
[item_len
] = 0;
2035 btrfs_put_root(root
);
2037 btrfs_free_path(path
);
2041 static noinline
int btrfs_ioctl_ino_lookup(struct btrfs_root
*root
,
2044 struct btrfs_ioctl_ino_lookup_args
*args
;
2047 args
= memdup_user(argp
, sizeof(*args
));
2049 return PTR_ERR(args
);
2052 * Unprivileged query to obtain the containing subvolume root id. The
2053 * path is reset so it's consistent with btrfs_search_path_in_tree.
2055 if (args
->treeid
== 0)
2056 args
->treeid
= root
->root_key
.objectid
;
2058 if (args
->objectid
== BTRFS_FIRST_FREE_OBJECTID
) {
2063 if (!capable(CAP_SYS_ADMIN
)) {
2068 ret
= btrfs_search_path_in_tree(root
->fs_info
,
2069 args
->treeid
, args
->objectid
,
2073 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2081 * Version of ino_lookup ioctl (unprivileged)
2083 * The main differences from ino_lookup ioctl are:
2085 * 1. Read + Exec permission will be checked using inode_permission() during
2086 * path construction. -EACCES will be returned in case of failure.
2087 * 2. Path construction will be stopped at the inode number which corresponds
2088 * to the fd with which this ioctl is called. If constructed path does not
2089 * exist under fd's inode, -EACCES will be returned.
2090 * 3. The name of bottom subvolume is also searched and filled.
2092 static int btrfs_ioctl_ino_lookup_user(struct file
*file
, void __user
*argp
)
2094 struct btrfs_ioctl_ino_lookup_user_args
*args
;
2095 struct inode
*inode
;
2098 args
= memdup_user(argp
, sizeof(*args
));
2100 return PTR_ERR(args
);
2102 inode
= file_inode(file
);
2104 if (args
->dirid
== BTRFS_FIRST_FREE_OBJECTID
&&
2105 BTRFS_I(inode
)->location
.objectid
!= BTRFS_FIRST_FREE_OBJECTID
) {
2107 * The subvolume does not exist under fd with which this is
2114 ret
= btrfs_search_path_in_tree_user(file_mnt_idmap(file
), inode
, args
);
2116 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2123 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2124 static int btrfs_ioctl_get_subvol_info(struct inode
*inode
, void __user
*argp
)
2126 struct btrfs_ioctl_get_subvol_info_args
*subvol_info
;
2127 struct btrfs_fs_info
*fs_info
;
2128 struct btrfs_root
*root
;
2129 struct btrfs_path
*path
;
2130 struct btrfs_key key
;
2131 struct btrfs_root_item
*root_item
;
2132 struct btrfs_root_ref
*rref
;
2133 struct extent_buffer
*leaf
;
2134 unsigned long item_off
;
2135 unsigned long item_len
;
2139 path
= btrfs_alloc_path();
2143 subvol_info
= kzalloc(sizeof(*subvol_info
), GFP_KERNEL
);
2145 btrfs_free_path(path
);
2149 fs_info
= BTRFS_I(inode
)->root
->fs_info
;
2151 /* Get root_item of inode's subvolume */
2152 key
.objectid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2153 root
= btrfs_get_fs_root(fs_info
, key
.objectid
, true);
2155 ret
= PTR_ERR(root
);
2158 root_item
= &root
->root_item
;
2160 subvol_info
->treeid
= key
.objectid
;
2162 subvol_info
->generation
= btrfs_root_generation(root_item
);
2163 subvol_info
->flags
= btrfs_root_flags(root_item
);
2165 memcpy(subvol_info
->uuid
, root_item
->uuid
, BTRFS_UUID_SIZE
);
2166 memcpy(subvol_info
->parent_uuid
, root_item
->parent_uuid
,
2168 memcpy(subvol_info
->received_uuid
, root_item
->received_uuid
,
2171 subvol_info
->ctransid
= btrfs_root_ctransid(root_item
);
2172 subvol_info
->ctime
.sec
= btrfs_stack_timespec_sec(&root_item
->ctime
);
2173 subvol_info
->ctime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->ctime
);
2175 subvol_info
->otransid
= btrfs_root_otransid(root_item
);
2176 subvol_info
->otime
.sec
= btrfs_stack_timespec_sec(&root_item
->otime
);
2177 subvol_info
->otime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->otime
);
2179 subvol_info
->stransid
= btrfs_root_stransid(root_item
);
2180 subvol_info
->stime
.sec
= btrfs_stack_timespec_sec(&root_item
->stime
);
2181 subvol_info
->stime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->stime
);
2183 subvol_info
->rtransid
= btrfs_root_rtransid(root_item
);
2184 subvol_info
->rtime
.sec
= btrfs_stack_timespec_sec(&root_item
->rtime
);
2185 subvol_info
->rtime
.nsec
= btrfs_stack_timespec_nsec(&root_item
->rtime
);
2187 if (key
.objectid
!= BTRFS_FS_TREE_OBJECTID
) {
2188 /* Search root tree for ROOT_BACKREF of this subvolume */
2189 key
.type
= BTRFS_ROOT_BACKREF_KEY
;
2191 ret
= btrfs_search_slot(NULL
, fs_info
->tree_root
, &key
, path
, 0, 0);
2194 } else if (path
->slots
[0] >=
2195 btrfs_header_nritems(path
->nodes
[0])) {
2196 ret
= btrfs_next_leaf(fs_info
->tree_root
, path
);
2199 } else if (ret
> 0) {
2205 leaf
= path
->nodes
[0];
2206 slot
= path
->slots
[0];
2207 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2208 if (key
.objectid
== subvol_info
->treeid
&&
2209 key
.type
== BTRFS_ROOT_BACKREF_KEY
) {
2210 subvol_info
->parent_id
= key
.offset
;
2212 rref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_root_ref
);
2213 subvol_info
->dirid
= btrfs_root_ref_dirid(leaf
, rref
);
2215 item_off
= btrfs_item_ptr_offset(leaf
, slot
)
2216 + sizeof(struct btrfs_root_ref
);
2217 item_len
= btrfs_item_size(leaf
, slot
)
2218 - sizeof(struct btrfs_root_ref
);
2219 read_extent_buffer(leaf
, subvol_info
->name
,
2220 item_off
, item_len
);
2227 btrfs_free_path(path
);
2229 if (copy_to_user(argp
, subvol_info
, sizeof(*subvol_info
)))
2233 btrfs_put_root(root
);
2235 btrfs_free_path(path
);
2241 * Return ROOT_REF information of the subvolume containing this inode
2242 * except the subvolume name.
2244 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root
*root
,
2247 struct btrfs_ioctl_get_subvol_rootref_args
*rootrefs
;
2248 struct btrfs_root_ref
*rref
;
2249 struct btrfs_path
*path
;
2250 struct btrfs_key key
;
2251 struct extent_buffer
*leaf
;
2257 path
= btrfs_alloc_path();
2261 rootrefs
= memdup_user(argp
, sizeof(*rootrefs
));
2262 if (IS_ERR(rootrefs
)) {
2263 btrfs_free_path(path
);
2264 return PTR_ERR(rootrefs
);
2267 objectid
= root
->root_key
.objectid
;
2268 key
.objectid
= objectid
;
2269 key
.type
= BTRFS_ROOT_REF_KEY
;
2270 key
.offset
= rootrefs
->min_treeid
;
2273 root
= root
->fs_info
->tree_root
;
2274 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2277 } else if (path
->slots
[0] >=
2278 btrfs_header_nritems(path
->nodes
[0])) {
2279 ret
= btrfs_next_leaf(root
, path
);
2282 } else if (ret
> 0) {
2288 leaf
= path
->nodes
[0];
2289 slot
= path
->slots
[0];
2291 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2292 if (key
.objectid
!= objectid
|| key
.type
!= BTRFS_ROOT_REF_KEY
) {
2297 if (found
== BTRFS_MAX_ROOTREF_BUFFER_NUM
) {
2302 rref
= btrfs_item_ptr(leaf
, slot
, struct btrfs_root_ref
);
2303 rootrefs
->rootref
[found
].treeid
= key
.offset
;
2304 rootrefs
->rootref
[found
].dirid
=
2305 btrfs_root_ref_dirid(leaf
, rref
);
2308 ret
= btrfs_next_item(root
, path
);
2311 } else if (ret
> 0) {
2318 btrfs_free_path(path
);
2320 if (!ret
|| ret
== -EOVERFLOW
) {
2321 rootrefs
->num_items
= found
;
2322 /* update min_treeid for next search */
2324 rootrefs
->min_treeid
=
2325 rootrefs
->rootref
[found
- 1].treeid
+ 1;
2326 if (copy_to_user(argp
, rootrefs
, sizeof(*rootrefs
)))
2335 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2339 struct dentry
*parent
= file
->f_path
.dentry
;
2340 struct btrfs_fs_info
*fs_info
= btrfs_sb(parent
->d_sb
);
2341 struct dentry
*dentry
;
2342 struct inode
*dir
= d_inode(parent
);
2343 struct inode
*inode
;
2344 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2345 struct btrfs_root
*dest
= NULL
;
2346 struct btrfs_ioctl_vol_args
*vol_args
= NULL
;
2347 struct btrfs_ioctl_vol_args_v2
*vol_args2
= NULL
;
2348 struct mnt_idmap
*idmap
= file_mnt_idmap(file
);
2349 char *subvol_name
, *subvol_name_ptr
= NULL
;
2352 bool destroy_parent
= false;
2354 /* We don't support snapshots with extent tree v2 yet. */
2355 if (btrfs_fs_incompat(fs_info
, EXTENT_TREE_V2
)) {
2357 "extent tree v2 doesn't support snapshot deletion yet");
2362 vol_args2
= memdup_user(arg
, sizeof(*vol_args2
));
2363 if (IS_ERR(vol_args2
))
2364 return PTR_ERR(vol_args2
);
2366 if (vol_args2
->flags
& ~BTRFS_SUBVOL_DELETE_ARGS_MASK
) {
2372 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2373 * name, same as v1 currently does.
2375 if (!(vol_args2
->flags
& BTRFS_SUBVOL_SPEC_BY_ID
)) {
2376 vol_args2
->name
[BTRFS_SUBVOL_NAME_MAX
] = 0;
2377 subvol_name
= vol_args2
->name
;
2379 err
= mnt_want_write_file(file
);
2383 struct inode
*old_dir
;
2385 if (vol_args2
->subvolid
< BTRFS_FIRST_FREE_OBJECTID
) {
2390 err
= mnt_want_write_file(file
);
2394 dentry
= btrfs_get_dentry(fs_info
->sb
,
2395 BTRFS_FIRST_FREE_OBJECTID
,
2396 vol_args2
->subvolid
, 0);
2397 if (IS_ERR(dentry
)) {
2398 err
= PTR_ERR(dentry
);
2399 goto out_drop_write
;
2403 * Change the default parent since the subvolume being
2404 * deleted can be outside of the current mount point.
2406 parent
= btrfs_get_parent(dentry
);
2409 * At this point dentry->d_name can point to '/' if the
2410 * subvolume we want to destroy is outsite of the
2411 * current mount point, so we need to release the
2412 * current dentry and execute the lookup to return a new
2413 * one with ->d_name pointing to the
2414 * <mount point>/subvol_name.
2417 if (IS_ERR(parent
)) {
2418 err
= PTR_ERR(parent
);
2419 goto out_drop_write
;
2422 dir
= d_inode(parent
);
2425 * If v2 was used with SPEC_BY_ID, a new parent was
2426 * allocated since the subvolume can be outside of the
2427 * current mount point. Later on we need to release this
2428 * new parent dentry.
2430 destroy_parent
= true;
2433 * On idmapped mounts, deletion via subvolid is
2434 * restricted to subvolumes that are immediate
2435 * ancestors of the inode referenced by the file
2436 * descriptor in the ioctl. Otherwise the idmapping
2437 * could potentially be abused to delete subvolumes
2438 * anywhere in the filesystem the user wouldn't be able
2439 * to delete without an idmapped mount.
2441 if (old_dir
!= dir
&& idmap
!= &nop_mnt_idmap
) {
2446 subvol_name_ptr
= btrfs_get_subvol_name_from_objectid(
2447 fs_info
, vol_args2
->subvolid
);
2448 if (IS_ERR(subvol_name_ptr
)) {
2449 err
= PTR_ERR(subvol_name_ptr
);
2452 /* subvol_name_ptr is already nul terminated */
2453 subvol_name
= (char *)kbasename(subvol_name_ptr
);
2456 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2457 if (IS_ERR(vol_args
))
2458 return PTR_ERR(vol_args
);
2460 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = 0;
2461 subvol_name
= vol_args
->name
;
2463 err
= mnt_want_write_file(file
);
2468 subvol_namelen
= strlen(subvol_name
);
2470 if (strchr(subvol_name
, '/') ||
2471 strncmp(subvol_name
, "..", subvol_namelen
) == 0) {
2473 goto free_subvol_name
;
2476 if (!S_ISDIR(dir
->i_mode
)) {
2478 goto free_subvol_name
;
2481 err
= down_write_killable_nested(&dir
->i_rwsem
, I_MUTEX_PARENT
);
2483 goto free_subvol_name
;
2484 dentry
= lookup_one(idmap
, subvol_name
, parent
, subvol_namelen
);
2485 if (IS_ERR(dentry
)) {
2486 err
= PTR_ERR(dentry
);
2487 goto out_unlock_dir
;
2490 if (d_really_is_negative(dentry
)) {
2495 inode
= d_inode(dentry
);
2496 dest
= BTRFS_I(inode
)->root
;
2497 if (!capable(CAP_SYS_ADMIN
)) {
2499 * Regular user. Only allow this with a special mount
2500 * option, when the user has write+exec access to the
2501 * subvol root, and when rmdir(2) would have been
2504 * Note that this is _not_ check that the subvol is
2505 * empty or doesn't contain data that we wouldn't
2506 * otherwise be able to delete.
2508 * Users who want to delete empty subvols should try
2512 if (!btrfs_test_opt(fs_info
, USER_SUBVOL_RM_ALLOWED
))
2516 * Do not allow deletion if the parent dir is the same
2517 * as the dir to be deleted. That means the ioctl
2518 * must be called on the dentry referencing the root
2519 * of the subvol, not a random directory contained
2526 err
= inode_permission(idmap
, inode
, MAY_WRITE
| MAY_EXEC
);
2531 /* check if subvolume may be deleted by a user */
2532 err
= btrfs_may_delete(idmap
, dir
, dentry
, 1);
2536 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
) {
2541 btrfs_inode_lock(BTRFS_I(inode
), 0);
2542 err
= btrfs_delete_subvolume(BTRFS_I(dir
), dentry
);
2543 btrfs_inode_unlock(BTRFS_I(inode
), 0);
2545 d_delete_notify(dir
, dentry
);
2550 btrfs_inode_unlock(BTRFS_I(dir
), 0);
2552 kfree(subvol_name_ptr
);
2557 mnt_drop_write_file(file
);
2564 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2566 struct inode
*inode
= file_inode(file
);
2567 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2568 struct btrfs_ioctl_defrag_range_args range
= {0};
2571 ret
= mnt_want_write_file(file
);
2575 if (btrfs_root_readonly(root
)) {
2580 switch (inode
->i_mode
& S_IFMT
) {
2582 if (!capable(CAP_SYS_ADMIN
)) {
2586 ret
= btrfs_defrag_root(root
);
2590 * Note that this does not check the file descriptor for write
2591 * access. This prevents defragmenting executables that are
2592 * running and allows defrag on files open in read-only mode.
2594 if (!capable(CAP_SYS_ADMIN
) &&
2595 inode_permission(&nop_mnt_idmap
, inode
, MAY_WRITE
)) {
2601 if (copy_from_user(&range
, argp
, sizeof(range
))) {
2605 /* compression requires us to start the IO */
2606 if ((range
.flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2607 range
.flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2608 range
.extent_thresh
= (u32
)-1;
2611 /* the rest are all set to zero by kzalloc */
2612 range
.len
= (u64
)-1;
2614 ret
= btrfs_defrag_file(file_inode(file
), &file
->f_ra
,
2615 &range
, BTRFS_OLDEST_GENERATION
, 0);
2623 mnt_drop_write_file(file
);
2627 static long btrfs_ioctl_add_dev(struct btrfs_fs_info
*fs_info
, void __user
*arg
)
2629 struct btrfs_ioctl_vol_args
*vol_args
;
2630 bool restore_op
= false;
2633 if (!capable(CAP_SYS_ADMIN
))
2636 if (btrfs_fs_incompat(fs_info
, EXTENT_TREE_V2
)) {
2637 btrfs_err(fs_info
, "device add not supported on extent tree v2 yet");
2641 if (fs_info
->fs_devices
->temp_fsid
) {
2643 "device add not supported on cloned temp-fsid mount");
2647 if (!btrfs_exclop_start(fs_info
, BTRFS_EXCLOP_DEV_ADD
)) {
2648 if (!btrfs_exclop_start_try_lock(fs_info
, BTRFS_EXCLOP_DEV_ADD
))
2649 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2652 * We can do the device add because we have a paused balanced,
2653 * change the exclusive op type and remember we should bring
2654 * back the paused balance
2656 fs_info
->exclusive_operation
= BTRFS_EXCLOP_DEV_ADD
;
2657 btrfs_exclop_start_unlock(fs_info
);
2661 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2662 if (IS_ERR(vol_args
)) {
2663 ret
= PTR_ERR(vol_args
);
2667 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2668 ret
= btrfs_init_new_device(fs_info
, vol_args
->name
);
2671 btrfs_info(fs_info
, "disk added %s", vol_args
->name
);
2676 btrfs_exclop_balance(fs_info
, BTRFS_EXCLOP_BALANCE_PAUSED
);
2678 btrfs_exclop_finish(fs_info
);
2682 static long btrfs_ioctl_rm_dev_v2(struct file
*file
, void __user
*arg
)
2684 BTRFS_DEV_LOOKUP_ARGS(args
);
2685 struct inode
*inode
= file_inode(file
);
2686 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
2687 struct btrfs_ioctl_vol_args_v2
*vol_args
;
2688 struct bdev_handle
*bdev_handle
= NULL
;
2690 bool cancel
= false;
2692 if (!capable(CAP_SYS_ADMIN
))
2695 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2696 if (IS_ERR(vol_args
))
2697 return PTR_ERR(vol_args
);
2699 if (vol_args
->flags
& ~BTRFS_DEVICE_REMOVE_ARGS_MASK
) {
2704 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
2705 if (vol_args
->flags
& BTRFS_DEVICE_SPEC_BY_ID
) {
2706 args
.devid
= vol_args
->devid
;
2707 } else if (!strcmp("cancel", vol_args
->name
)) {
2710 ret
= btrfs_get_dev_args_from_path(fs_info
, &args
, vol_args
->name
);
2715 ret
= mnt_want_write_file(file
);
2719 ret
= exclop_start_or_cancel_reloc(fs_info
, BTRFS_EXCLOP_DEV_REMOVE
,
2724 /* Exclusive operation is now claimed */
2725 ret
= btrfs_rm_device(fs_info
, &args
, &bdev_handle
);
2727 btrfs_exclop_finish(fs_info
);
2730 if (vol_args
->flags
& BTRFS_DEVICE_SPEC_BY_ID
)
2731 btrfs_info(fs_info
, "device deleted: id %llu",
2734 btrfs_info(fs_info
, "device deleted: %s",
2738 mnt_drop_write_file(file
);
2740 bdev_release(bdev_handle
);
2742 btrfs_put_dev_args_from_path(&args
);
2747 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2749 BTRFS_DEV_LOOKUP_ARGS(args
);
2750 struct inode
*inode
= file_inode(file
);
2751 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
2752 struct btrfs_ioctl_vol_args
*vol_args
;
2753 struct bdev_handle
*bdev_handle
= NULL
;
2755 bool cancel
= false;
2757 if (!capable(CAP_SYS_ADMIN
))
2760 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2761 if (IS_ERR(vol_args
))
2762 return PTR_ERR(vol_args
);
2764 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2765 if (!strcmp("cancel", vol_args
->name
)) {
2768 ret
= btrfs_get_dev_args_from_path(fs_info
, &args
, vol_args
->name
);
2773 ret
= mnt_want_write_file(file
);
2777 ret
= exclop_start_or_cancel_reloc(fs_info
, BTRFS_EXCLOP_DEV_REMOVE
,
2780 ret
= btrfs_rm_device(fs_info
, &args
, &bdev_handle
);
2782 btrfs_info(fs_info
, "disk deleted %s", vol_args
->name
);
2783 btrfs_exclop_finish(fs_info
);
2786 mnt_drop_write_file(file
);
2788 bdev_release(bdev_handle
);
2790 btrfs_put_dev_args_from_path(&args
);
2795 static long btrfs_ioctl_fs_info(struct btrfs_fs_info
*fs_info
,
2798 struct btrfs_ioctl_fs_info_args
*fi_args
;
2799 struct btrfs_device
*device
;
2800 struct btrfs_fs_devices
*fs_devices
= fs_info
->fs_devices
;
2804 fi_args
= memdup_user(arg
, sizeof(*fi_args
));
2805 if (IS_ERR(fi_args
))
2806 return PTR_ERR(fi_args
);
2808 flags_in
= fi_args
->flags
;
2809 memset(fi_args
, 0, sizeof(*fi_args
));
2812 fi_args
->num_devices
= fs_devices
->num_devices
;
2814 list_for_each_entry_rcu(device
, &fs_devices
->devices
, dev_list
) {
2815 if (device
->devid
> fi_args
->max_id
)
2816 fi_args
->max_id
= device
->devid
;
2820 memcpy(&fi_args
->fsid
, fs_devices
->fsid
, sizeof(fi_args
->fsid
));
2821 fi_args
->nodesize
= fs_info
->nodesize
;
2822 fi_args
->sectorsize
= fs_info
->sectorsize
;
2823 fi_args
->clone_alignment
= fs_info
->sectorsize
;
2825 if (flags_in
& BTRFS_FS_INFO_FLAG_CSUM_INFO
) {
2826 fi_args
->csum_type
= btrfs_super_csum_type(fs_info
->super_copy
);
2827 fi_args
->csum_size
= btrfs_super_csum_size(fs_info
->super_copy
);
2828 fi_args
->flags
|= BTRFS_FS_INFO_FLAG_CSUM_INFO
;
2831 if (flags_in
& BTRFS_FS_INFO_FLAG_GENERATION
) {
2832 fi_args
->generation
= btrfs_get_fs_generation(fs_info
);
2833 fi_args
->flags
|= BTRFS_FS_INFO_FLAG_GENERATION
;
2836 if (flags_in
& BTRFS_FS_INFO_FLAG_METADATA_UUID
) {
2837 memcpy(&fi_args
->metadata_uuid
, fs_devices
->metadata_uuid
,
2838 sizeof(fi_args
->metadata_uuid
));
2839 fi_args
->flags
|= BTRFS_FS_INFO_FLAG_METADATA_UUID
;
2842 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2849 static long btrfs_ioctl_dev_info(struct btrfs_fs_info
*fs_info
,
2852 BTRFS_DEV_LOOKUP_ARGS(args
);
2853 struct btrfs_ioctl_dev_info_args
*di_args
;
2854 struct btrfs_device
*dev
;
2857 di_args
= memdup_user(arg
, sizeof(*di_args
));
2858 if (IS_ERR(di_args
))
2859 return PTR_ERR(di_args
);
2861 args
.devid
= di_args
->devid
;
2862 if (!btrfs_is_empty_uuid(di_args
->uuid
))
2863 args
.uuid
= di_args
->uuid
;
2866 dev
= btrfs_find_device(fs_info
->fs_devices
, &args
);
2872 di_args
->devid
= dev
->devid
;
2873 di_args
->bytes_used
= btrfs_device_get_bytes_used(dev
);
2874 di_args
->total_bytes
= btrfs_device_get_total_bytes(dev
);
2875 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2876 memcpy(di_args
->fsid
, dev
->fs_devices
->fsid
, BTRFS_UUID_SIZE
);
2878 strscpy(di_args
->path
, btrfs_dev_name(dev
), sizeof(di_args
->path
));
2880 di_args
->path
[0] = '\0';
2884 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2891 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2893 struct inode
*inode
= file_inode(file
);
2894 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
2895 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2896 struct btrfs_root
*new_root
;
2897 struct btrfs_dir_item
*di
;
2898 struct btrfs_trans_handle
*trans
;
2899 struct btrfs_path
*path
= NULL
;
2900 struct btrfs_disk_key disk_key
;
2901 struct fscrypt_str name
= FSTR_INIT("default", 7);
2906 if (!capable(CAP_SYS_ADMIN
))
2909 ret
= mnt_want_write_file(file
);
2913 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
2919 objectid
= BTRFS_FS_TREE_OBJECTID
;
2921 new_root
= btrfs_get_fs_root(fs_info
, objectid
, true);
2922 if (IS_ERR(new_root
)) {
2923 ret
= PTR_ERR(new_root
);
2926 if (!is_fstree(new_root
->root_key
.objectid
)) {
2931 path
= btrfs_alloc_path();
2937 trans
= btrfs_start_transaction(root
, 1);
2938 if (IS_ERR(trans
)) {
2939 ret
= PTR_ERR(trans
);
2943 dir_id
= btrfs_super_root_dir(fs_info
->super_copy
);
2944 di
= btrfs_lookup_dir_item(trans
, fs_info
->tree_root
, path
,
2946 if (IS_ERR_OR_NULL(di
)) {
2947 btrfs_release_path(path
);
2948 btrfs_end_transaction(trans
);
2950 "Umm, you don't have the default diritem, this isn't going to work");
2955 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
2956 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
2957 btrfs_mark_buffer_dirty(trans
, path
->nodes
[0]);
2958 btrfs_release_path(path
);
2960 btrfs_set_fs_incompat(fs_info
, DEFAULT_SUBVOL
);
2961 btrfs_end_transaction(trans
);
2963 btrfs_put_root(new_root
);
2964 btrfs_free_path(path
);
2966 mnt_drop_write_file(file
);
2970 static void get_block_group_info(struct list_head
*groups_list
,
2971 struct btrfs_ioctl_space_info
*space
)
2973 struct btrfs_block_group
*block_group
;
2975 space
->total_bytes
= 0;
2976 space
->used_bytes
= 0;
2978 list_for_each_entry(block_group
, groups_list
, list
) {
2979 space
->flags
= block_group
->flags
;
2980 space
->total_bytes
+= block_group
->length
;
2981 space
->used_bytes
+= block_group
->used
;
2985 static long btrfs_ioctl_space_info(struct btrfs_fs_info
*fs_info
,
2988 struct btrfs_ioctl_space_args space_args
= { 0 };
2989 struct btrfs_ioctl_space_info space
;
2990 struct btrfs_ioctl_space_info
*dest
;
2991 struct btrfs_ioctl_space_info
*dest_orig
;
2992 struct btrfs_ioctl_space_info __user
*user_dest
;
2993 struct btrfs_space_info
*info
;
2994 static const u64 types
[] = {
2995 BTRFS_BLOCK_GROUP_DATA
,
2996 BTRFS_BLOCK_GROUP_SYSTEM
,
2997 BTRFS_BLOCK_GROUP_METADATA
,
2998 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
3006 if (copy_from_user(&space_args
,
3007 (struct btrfs_ioctl_space_args __user
*)arg
,
3008 sizeof(space_args
)))
3011 for (i
= 0; i
< num_types
; i
++) {
3012 struct btrfs_space_info
*tmp
;
3015 list_for_each_entry(tmp
, &fs_info
->space_info
, list
) {
3016 if (tmp
->flags
== types
[i
]) {
3025 down_read(&info
->groups_sem
);
3026 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3027 if (!list_empty(&info
->block_groups
[c
]))
3030 up_read(&info
->groups_sem
);
3034 * Global block reserve, exported as a space_info
3038 /* space_slots == 0 means they are asking for a count */
3039 if (space_args
.space_slots
== 0) {
3040 space_args
.total_spaces
= slot_count
;
3044 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
3046 alloc_size
= sizeof(*dest
) * slot_count
;
3048 /* we generally have at most 6 or so space infos, one for each raid
3049 * level. So, a whole page should be more than enough for everyone
3051 if (alloc_size
> PAGE_SIZE
)
3054 space_args
.total_spaces
= 0;
3055 dest
= kmalloc(alloc_size
, GFP_KERNEL
);
3060 /* now we have a buffer to copy into */
3061 for (i
= 0; i
< num_types
; i
++) {
3062 struct btrfs_space_info
*tmp
;
3068 list_for_each_entry(tmp
, &fs_info
->space_info
, list
) {
3069 if (tmp
->flags
== types
[i
]) {
3077 down_read(&info
->groups_sem
);
3078 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3079 if (!list_empty(&info
->block_groups
[c
])) {
3080 get_block_group_info(&info
->block_groups
[c
],
3082 memcpy(dest
, &space
, sizeof(space
));
3084 space_args
.total_spaces
++;
3090 up_read(&info
->groups_sem
);
3094 * Add global block reserve
3097 struct btrfs_block_rsv
*block_rsv
= &fs_info
->global_block_rsv
;
3099 spin_lock(&block_rsv
->lock
);
3100 space
.total_bytes
= block_rsv
->size
;
3101 space
.used_bytes
= block_rsv
->size
- block_rsv
->reserved
;
3102 spin_unlock(&block_rsv
->lock
);
3103 space
.flags
= BTRFS_SPACE_INFO_GLOBAL_RSV
;
3104 memcpy(dest
, &space
, sizeof(space
));
3105 space_args
.total_spaces
++;
3108 user_dest
= (struct btrfs_ioctl_space_info __user
*)
3109 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
3111 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
3116 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
3122 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
3125 struct btrfs_trans_handle
*trans
;
3129 * Start orphan cleanup here for the given root in case it hasn't been
3130 * started already by other means. Errors are handled in the other
3131 * functions during transaction commit.
3133 btrfs_orphan_cleanup(root
);
3135 trans
= btrfs_attach_transaction_barrier(root
);
3136 if (IS_ERR(trans
)) {
3137 if (PTR_ERR(trans
) != -ENOENT
)
3138 return PTR_ERR(trans
);
3140 /* No running transaction, don't bother */
3141 transid
= btrfs_get_last_trans_committed(root
->fs_info
);
3144 transid
= trans
->transid
;
3145 btrfs_commit_transaction_async(trans
);
3148 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
3153 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_fs_info
*fs_info
,
3156 /* By default wait for the current transaction. */
3160 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
3163 return btrfs_wait_for_commit(fs_info
, transid
);
3166 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
3168 struct btrfs_fs_info
*fs_info
= btrfs_sb(file_inode(file
)->i_sb
);
3169 struct btrfs_ioctl_scrub_args
*sa
;
3172 if (!capable(CAP_SYS_ADMIN
))
3175 if (btrfs_fs_incompat(fs_info
, EXTENT_TREE_V2
)) {
3176 btrfs_err(fs_info
, "scrub is not supported on extent tree v2 yet");
3180 sa
= memdup_user(arg
, sizeof(*sa
));
3184 if (sa
->flags
& ~BTRFS_SCRUB_SUPPORTED_FLAGS
) {
3189 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
3190 ret
= mnt_want_write_file(file
);
3195 ret
= btrfs_scrub_dev(fs_info
, sa
->devid
, sa
->start
, sa
->end
,
3196 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
3200 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3201 * error. This is important as it allows user space to know how much
3202 * progress scrub has done. For example, if scrub is canceled we get
3203 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3204 * space. Later user space can inspect the progress from the structure
3205 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3206 * previously (btrfs-progs does this).
3207 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3208 * then return -EFAULT to signal the structure was not copied or it may
3209 * be corrupt and unreliable due to a partial copy.
3211 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3214 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
3215 mnt_drop_write_file(file
);
3221 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info
*fs_info
)
3223 if (!capable(CAP_SYS_ADMIN
))
3226 return btrfs_scrub_cancel(fs_info
);
3229 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info
*fs_info
,
3232 struct btrfs_ioctl_scrub_args
*sa
;
3235 if (!capable(CAP_SYS_ADMIN
))
3238 sa
= memdup_user(arg
, sizeof(*sa
));
3242 ret
= btrfs_scrub_progress(fs_info
, sa
->devid
, &sa
->progress
);
3244 if (ret
== 0 && copy_to_user(arg
, sa
, sizeof(*sa
)))
3251 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info
*fs_info
,
3254 struct btrfs_ioctl_get_dev_stats
*sa
;
3257 sa
= memdup_user(arg
, sizeof(*sa
));
3261 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
3266 ret
= btrfs_get_dev_stats(fs_info
, sa
);
3268 if (ret
== 0 && copy_to_user(arg
, sa
, sizeof(*sa
)))
3275 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info
*fs_info
,
3278 struct btrfs_ioctl_dev_replace_args
*p
;
3281 if (!capable(CAP_SYS_ADMIN
))
3284 if (btrfs_fs_incompat(fs_info
, EXTENT_TREE_V2
)) {
3285 btrfs_err(fs_info
, "device replace not supported on extent tree v2 yet");
3289 p
= memdup_user(arg
, sizeof(*p
));
3294 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
3295 if (sb_rdonly(fs_info
->sb
)) {
3299 if (!btrfs_exclop_start(fs_info
, BTRFS_EXCLOP_DEV_REPLACE
)) {
3300 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
3302 ret
= btrfs_dev_replace_by_ioctl(fs_info
, p
);
3303 btrfs_exclop_finish(fs_info
);
3306 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
3307 btrfs_dev_replace_status(fs_info
, p
);
3310 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
3311 p
->result
= btrfs_dev_replace_cancel(fs_info
);
3319 if ((ret
== 0 || ret
== -ECANCELED
) && copy_to_user(arg
, p
, sizeof(*p
)))
3326 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
3332 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
3333 struct inode_fs_paths
*ipath
= NULL
;
3334 struct btrfs_path
*path
;
3336 if (!capable(CAP_DAC_READ_SEARCH
))
3339 path
= btrfs_alloc_path();
3345 ipa
= memdup_user(arg
, sizeof(*ipa
));
3352 size
= min_t(u32
, ipa
->size
, 4096);
3353 ipath
= init_ipath(size
, root
, path
);
3354 if (IS_ERR(ipath
)) {
3355 ret
= PTR_ERR(ipath
);
3360 ret
= paths_from_inode(ipa
->inum
, ipath
);
3364 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
3365 rel_ptr
= ipath
->fspath
->val
[i
] -
3366 (u64
)(unsigned long)ipath
->fspath
->val
;
3367 ipath
->fspath
->val
[i
] = rel_ptr
;
3370 btrfs_free_path(path
);
3372 ret
= copy_to_user((void __user
*)(unsigned long)ipa
->fspath
,
3373 ipath
->fspath
, size
);
3380 btrfs_free_path(path
);
3387 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info
*fs_info
,
3388 void __user
*arg
, int version
)
3392 struct btrfs_ioctl_logical_ino_args
*loi
;
3393 struct btrfs_data_container
*inodes
= NULL
;
3394 struct btrfs_path
*path
= NULL
;
3397 if (!capable(CAP_SYS_ADMIN
))
3400 loi
= memdup_user(arg
, sizeof(*loi
));
3402 return PTR_ERR(loi
);
3405 ignore_offset
= false;
3406 size
= min_t(u32
, loi
->size
, SZ_64K
);
3408 /* All reserved bits must be 0 for now */
3409 if (memchr_inv(loi
->reserved
, 0, sizeof(loi
->reserved
))) {
3413 /* Only accept flags we have defined so far */
3414 if (loi
->flags
& ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET
)) {
3418 ignore_offset
= loi
->flags
& BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET
;
3419 size
= min_t(u32
, loi
->size
, SZ_16M
);
3422 inodes
= init_data_container(size
);
3423 if (IS_ERR(inodes
)) {
3424 ret
= PTR_ERR(inodes
);
3428 path
= btrfs_alloc_path();
3433 ret
= iterate_inodes_from_logical(loi
->logical
, fs_info
, path
,
3434 inodes
, ignore_offset
);
3435 btrfs_free_path(path
);
3441 ret
= copy_to_user((void __user
*)(unsigned long)loi
->inodes
, inodes
,
3454 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
,
3455 struct btrfs_ioctl_balance_args
*bargs
)
3457 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
3459 bargs
->flags
= bctl
->flags
;
3461 if (test_bit(BTRFS_FS_BALANCE_RUNNING
, &fs_info
->flags
))
3462 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
3463 if (atomic_read(&fs_info
->balance_pause_req
))
3464 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
3465 if (atomic_read(&fs_info
->balance_cancel_req
))
3466 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
3468 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
3469 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
3470 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
3472 spin_lock(&fs_info
->balance_lock
);
3473 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3474 spin_unlock(&fs_info
->balance_lock
);
3478 * Try to acquire fs_info::balance_mutex as well as set BTRFS_EXLCOP_BALANCE as
3481 * @fs_info: the filesystem
3482 * @excl_acquired: ptr to boolean value which is set to false in case balance
3485 * Return 0 on success in which case both fs_info::balance is acquired as well
3486 * as exclusive ops are blocked. In case of failure return an error code.
3488 static int btrfs_try_lock_balance(struct btrfs_fs_info
*fs_info
, bool *excl_acquired
)
3493 * Exclusive operation is locked. Three possibilities:
3494 * (1) some other op is running
3495 * (2) balance is running
3496 * (3) balance is paused -- special case (think resume)
3499 if (btrfs_exclop_start(fs_info
, BTRFS_EXCLOP_BALANCE
)) {
3500 *excl_acquired
= true;
3501 mutex_lock(&fs_info
->balance_mutex
);
3505 mutex_lock(&fs_info
->balance_mutex
);
3506 if (fs_info
->balance_ctl
) {
3507 /* This is either (2) or (3) */
3508 if (test_bit(BTRFS_FS_BALANCE_RUNNING
, &fs_info
->flags
)) {
3514 mutex_unlock(&fs_info
->balance_mutex
);
3516 * Lock released to allow other waiters to
3517 * continue, we'll reexamine the status again.
3519 mutex_lock(&fs_info
->balance_mutex
);
3521 if (fs_info
->balance_ctl
&&
3522 !test_bit(BTRFS_FS_BALANCE_RUNNING
, &fs_info
->flags
)) {
3524 *excl_acquired
= false;
3530 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
3534 mutex_unlock(&fs_info
->balance_mutex
);
3538 mutex_unlock(&fs_info
->balance_mutex
);
3539 *excl_acquired
= false;
3543 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
3545 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3546 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3547 struct btrfs_ioctl_balance_args
*bargs
;
3548 struct btrfs_balance_control
*bctl
;
3549 bool need_unlock
= true;
3552 if (!capable(CAP_SYS_ADMIN
))
3555 ret
= mnt_want_write_file(file
);
3559 bargs
= memdup_user(arg
, sizeof(*bargs
));
3560 if (IS_ERR(bargs
)) {
3561 ret
= PTR_ERR(bargs
);
3566 ret
= btrfs_try_lock_balance(fs_info
, &need_unlock
);
3570 lockdep_assert_held(&fs_info
->balance_mutex
);
3572 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
3573 if (!fs_info
->balance_ctl
) {
3578 bctl
= fs_info
->balance_ctl
;
3579 spin_lock(&fs_info
->balance_lock
);
3580 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
3581 spin_unlock(&fs_info
->balance_lock
);
3582 btrfs_exclop_balance(fs_info
, BTRFS_EXCLOP_BALANCE
);
3587 if (bargs
->flags
& ~(BTRFS_BALANCE_ARGS_MASK
| BTRFS_BALANCE_TYPE_MASK
)) {
3592 if (fs_info
->balance_ctl
) {
3597 bctl
= kzalloc(sizeof(*bctl
), GFP_KERNEL
);
3603 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
3604 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
3605 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
3607 bctl
->flags
= bargs
->flags
;
3610 * Ownership of bctl and exclusive operation goes to btrfs_balance.
3611 * bctl is freed in reset_balance_state, or, if restriper was paused
3612 * all the way until unmount, in free_fs_info. The flag should be
3613 * cleared after reset_balance_state.
3615 need_unlock
= false;
3617 ret
= btrfs_balance(fs_info
, bctl
, bargs
);
3620 if (ret
== 0 || ret
== -ECANCELED
) {
3621 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3627 mutex_unlock(&fs_info
->balance_mutex
);
3629 btrfs_exclop_finish(fs_info
);
3631 mnt_drop_write_file(file
);
3636 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info
*fs_info
, int cmd
)
3638 if (!capable(CAP_SYS_ADMIN
))
3642 case BTRFS_BALANCE_CTL_PAUSE
:
3643 return btrfs_pause_balance(fs_info
);
3644 case BTRFS_BALANCE_CTL_CANCEL
:
3645 return btrfs_cancel_balance(fs_info
);
3651 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info
*fs_info
,
3654 struct btrfs_ioctl_balance_args
*bargs
;
3657 if (!capable(CAP_SYS_ADMIN
))
3660 mutex_lock(&fs_info
->balance_mutex
);
3661 if (!fs_info
->balance_ctl
) {
3666 bargs
= kzalloc(sizeof(*bargs
), GFP_KERNEL
);
3672 btrfs_update_ioctl_balance_args(fs_info
, bargs
);
3674 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3679 mutex_unlock(&fs_info
->balance_mutex
);
3683 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
3685 struct inode
*inode
= file_inode(file
);
3686 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3687 struct btrfs_ioctl_quota_ctl_args
*sa
;
3690 if (!capable(CAP_SYS_ADMIN
))
3693 ret
= mnt_want_write_file(file
);
3697 sa
= memdup_user(arg
, sizeof(*sa
));
3703 down_write(&fs_info
->subvol_sem
);
3706 case BTRFS_QUOTA_CTL_ENABLE
:
3707 case BTRFS_QUOTA_CTL_ENABLE_SIMPLE_QUOTA
:
3708 ret
= btrfs_quota_enable(fs_info
, sa
);
3710 case BTRFS_QUOTA_CTL_DISABLE
:
3711 ret
= btrfs_quota_disable(fs_info
);
3719 up_write(&fs_info
->subvol_sem
);
3721 mnt_drop_write_file(file
);
3725 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
3727 struct inode
*inode
= file_inode(file
);
3728 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3729 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3730 struct btrfs_ioctl_qgroup_assign_args
*sa
;
3731 struct btrfs_trans_handle
*trans
;
3735 if (!capable(CAP_SYS_ADMIN
))
3738 ret
= mnt_want_write_file(file
);
3742 sa
= memdup_user(arg
, sizeof(*sa
));
3748 trans
= btrfs_join_transaction(root
);
3749 if (IS_ERR(trans
)) {
3750 ret
= PTR_ERR(trans
);
3755 ret
= btrfs_add_qgroup_relation(trans
, sa
->src
, sa
->dst
);
3757 ret
= btrfs_del_qgroup_relation(trans
, sa
->src
, sa
->dst
);
3760 /* update qgroup status and info */
3761 mutex_lock(&fs_info
->qgroup_ioctl_lock
);
3762 err
= btrfs_run_qgroups(trans
);
3763 mutex_unlock(&fs_info
->qgroup_ioctl_lock
);
3765 btrfs_handle_fs_error(fs_info
, err
,
3766 "failed to update qgroup status and info");
3767 err
= btrfs_end_transaction(trans
);
3774 mnt_drop_write_file(file
);
3778 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
3780 struct inode
*inode
= file_inode(file
);
3781 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3782 struct btrfs_ioctl_qgroup_create_args
*sa
;
3783 struct btrfs_trans_handle
*trans
;
3787 if (!capable(CAP_SYS_ADMIN
))
3790 ret
= mnt_want_write_file(file
);
3794 sa
= memdup_user(arg
, sizeof(*sa
));
3800 if (!sa
->qgroupid
) {
3805 trans
= btrfs_join_transaction(root
);
3806 if (IS_ERR(trans
)) {
3807 ret
= PTR_ERR(trans
);
3812 ret
= btrfs_create_qgroup(trans
, sa
->qgroupid
);
3814 ret
= btrfs_remove_qgroup(trans
, sa
->qgroupid
);
3817 err
= btrfs_end_transaction(trans
);
3824 mnt_drop_write_file(file
);
3828 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
3830 struct inode
*inode
= file_inode(file
);
3831 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3832 struct btrfs_ioctl_qgroup_limit_args
*sa
;
3833 struct btrfs_trans_handle
*trans
;
3838 if (!capable(CAP_SYS_ADMIN
))
3841 ret
= mnt_want_write_file(file
);
3845 sa
= memdup_user(arg
, sizeof(*sa
));
3851 trans
= btrfs_join_transaction(root
);
3852 if (IS_ERR(trans
)) {
3853 ret
= PTR_ERR(trans
);
3857 qgroupid
= sa
->qgroupid
;
3859 /* take the current subvol as qgroup */
3860 qgroupid
= root
->root_key
.objectid
;
3863 ret
= btrfs_limit_qgroup(trans
, qgroupid
, &sa
->lim
);
3865 err
= btrfs_end_transaction(trans
);
3872 mnt_drop_write_file(file
);
3876 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
3878 struct inode
*inode
= file_inode(file
);
3879 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3880 struct btrfs_ioctl_quota_rescan_args
*qsa
;
3883 if (!capable(CAP_SYS_ADMIN
))
3886 ret
= mnt_want_write_file(file
);
3890 qsa
= memdup_user(arg
, sizeof(*qsa
));
3901 ret
= btrfs_qgroup_rescan(fs_info
);
3906 mnt_drop_write_file(file
);
3910 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info
*fs_info
,
3913 struct btrfs_ioctl_quota_rescan_args qsa
= {0};
3915 if (!capable(CAP_SYS_ADMIN
))
3918 if (fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
3920 qsa
.progress
= fs_info
->qgroup_rescan_progress
.objectid
;
3923 if (copy_to_user(arg
, &qsa
, sizeof(qsa
)))
3929 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info
*fs_info
,
3932 if (!capable(CAP_SYS_ADMIN
))
3935 return btrfs_qgroup_wait_for_completion(fs_info
, true);
3938 static long _btrfs_ioctl_set_received_subvol(struct file
*file
,
3939 struct mnt_idmap
*idmap
,
3940 struct btrfs_ioctl_received_subvol_args
*sa
)
3942 struct inode
*inode
= file_inode(file
);
3943 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
3944 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3945 struct btrfs_root_item
*root_item
= &root
->root_item
;
3946 struct btrfs_trans_handle
*trans
;
3947 struct timespec64 ct
= current_time(inode
);
3949 int received_uuid_changed
;
3951 if (!inode_owner_or_capable(idmap
, inode
))
3954 ret
= mnt_want_write_file(file
);
3958 down_write(&fs_info
->subvol_sem
);
3960 if (btrfs_ino(BTRFS_I(inode
)) != BTRFS_FIRST_FREE_OBJECTID
) {
3965 if (btrfs_root_readonly(root
)) {
3972 * 2 - uuid items (received uuid + subvol uuid)
3974 trans
= btrfs_start_transaction(root
, 3);
3975 if (IS_ERR(trans
)) {
3976 ret
= PTR_ERR(trans
);
3981 sa
->rtransid
= trans
->transid
;
3982 sa
->rtime
.sec
= ct
.tv_sec
;
3983 sa
->rtime
.nsec
= ct
.tv_nsec
;
3985 received_uuid_changed
= memcmp(root_item
->received_uuid
, sa
->uuid
,
3987 if (received_uuid_changed
&&
3988 !btrfs_is_empty_uuid(root_item
->received_uuid
)) {
3989 ret
= btrfs_uuid_tree_remove(trans
, root_item
->received_uuid
,
3990 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
3991 root
->root_key
.objectid
);
3992 if (ret
&& ret
!= -ENOENT
) {
3993 btrfs_abort_transaction(trans
, ret
);
3994 btrfs_end_transaction(trans
);
3998 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
3999 btrfs_set_root_stransid(root_item
, sa
->stransid
);
4000 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
4001 btrfs_set_stack_timespec_sec(&root_item
->stime
, sa
->stime
.sec
);
4002 btrfs_set_stack_timespec_nsec(&root_item
->stime
, sa
->stime
.nsec
);
4003 btrfs_set_stack_timespec_sec(&root_item
->rtime
, sa
->rtime
.sec
);
4004 btrfs_set_stack_timespec_nsec(&root_item
->rtime
, sa
->rtime
.nsec
);
4006 ret
= btrfs_update_root(trans
, fs_info
->tree_root
,
4007 &root
->root_key
, &root
->root_item
);
4009 btrfs_end_transaction(trans
);
4012 if (received_uuid_changed
&& !btrfs_is_empty_uuid(sa
->uuid
)) {
4013 ret
= btrfs_uuid_tree_add(trans
, sa
->uuid
,
4014 BTRFS_UUID_KEY_RECEIVED_SUBVOL
,
4015 root
->root_key
.objectid
);
4016 if (ret
< 0 && ret
!= -EEXIST
) {
4017 btrfs_abort_transaction(trans
, ret
);
4018 btrfs_end_transaction(trans
);
4022 ret
= btrfs_commit_transaction(trans
);
4024 up_write(&fs_info
->subvol_sem
);
4025 mnt_drop_write_file(file
);
4030 static long btrfs_ioctl_set_received_subvol_32(struct file
*file
,
4033 struct btrfs_ioctl_received_subvol_args_32
*args32
= NULL
;
4034 struct btrfs_ioctl_received_subvol_args
*args64
= NULL
;
4037 args32
= memdup_user(arg
, sizeof(*args32
));
4039 return PTR_ERR(args32
);
4041 args64
= kmalloc(sizeof(*args64
), GFP_KERNEL
);
4047 memcpy(args64
->uuid
, args32
->uuid
, BTRFS_UUID_SIZE
);
4048 args64
->stransid
= args32
->stransid
;
4049 args64
->rtransid
= args32
->rtransid
;
4050 args64
->stime
.sec
= args32
->stime
.sec
;
4051 args64
->stime
.nsec
= args32
->stime
.nsec
;
4052 args64
->rtime
.sec
= args32
->rtime
.sec
;
4053 args64
->rtime
.nsec
= args32
->rtime
.nsec
;
4054 args64
->flags
= args32
->flags
;
4056 ret
= _btrfs_ioctl_set_received_subvol(file
, file_mnt_idmap(file
), args64
);
4060 memcpy(args32
->uuid
, args64
->uuid
, BTRFS_UUID_SIZE
);
4061 args32
->stransid
= args64
->stransid
;
4062 args32
->rtransid
= args64
->rtransid
;
4063 args32
->stime
.sec
= args64
->stime
.sec
;
4064 args32
->stime
.nsec
= args64
->stime
.nsec
;
4065 args32
->rtime
.sec
= args64
->rtime
.sec
;
4066 args32
->rtime
.nsec
= args64
->rtime
.nsec
;
4067 args32
->flags
= args64
->flags
;
4069 ret
= copy_to_user(arg
, args32
, sizeof(*args32
));
4080 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
4083 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
4086 sa
= memdup_user(arg
, sizeof(*sa
));
4090 ret
= _btrfs_ioctl_set_received_subvol(file
, file_mnt_idmap(file
), sa
);
4095 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
4104 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info
*fs_info
,
4109 char label
[BTRFS_LABEL_SIZE
];
4111 spin_lock(&fs_info
->super_lock
);
4112 memcpy(label
, fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
4113 spin_unlock(&fs_info
->super_lock
);
4115 len
= strnlen(label
, BTRFS_LABEL_SIZE
);
4117 if (len
== BTRFS_LABEL_SIZE
) {
4119 "label is too long, return the first %zu bytes",
4123 ret
= copy_to_user(arg
, label
, len
);
4125 return ret
? -EFAULT
: 0;
4128 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
4130 struct inode
*inode
= file_inode(file
);
4131 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4132 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4133 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
4134 struct btrfs_trans_handle
*trans
;
4135 char label
[BTRFS_LABEL_SIZE
];
4138 if (!capable(CAP_SYS_ADMIN
))
4141 if (copy_from_user(label
, arg
, sizeof(label
)))
4144 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
4146 "unable to set label with more than %d bytes",
4147 BTRFS_LABEL_SIZE
- 1);
4151 ret
= mnt_want_write_file(file
);
4155 trans
= btrfs_start_transaction(root
, 0);
4156 if (IS_ERR(trans
)) {
4157 ret
= PTR_ERR(trans
);
4161 spin_lock(&fs_info
->super_lock
);
4162 strcpy(super_block
->label
, label
);
4163 spin_unlock(&fs_info
->super_lock
);
4164 ret
= btrfs_commit_transaction(trans
);
4167 mnt_drop_write_file(file
);
4171 #define INIT_FEATURE_FLAGS(suffix) \
4172 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4173 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4174 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4176 int btrfs_ioctl_get_supported_features(void __user
*arg
)
4178 static const struct btrfs_ioctl_feature_flags features
[3] = {
4179 INIT_FEATURE_FLAGS(SUPP
),
4180 INIT_FEATURE_FLAGS(SAFE_SET
),
4181 INIT_FEATURE_FLAGS(SAFE_CLEAR
)
4184 if (copy_to_user(arg
, &features
, sizeof(features
)))
4190 static int btrfs_ioctl_get_features(struct btrfs_fs_info
*fs_info
,
4193 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
4194 struct btrfs_ioctl_feature_flags features
;
4196 features
.compat_flags
= btrfs_super_compat_flags(super_block
);
4197 features
.compat_ro_flags
= btrfs_super_compat_ro_flags(super_block
);
4198 features
.incompat_flags
= btrfs_super_incompat_flags(super_block
);
4200 if (copy_to_user(arg
, &features
, sizeof(features
)))
4206 static int check_feature_bits(struct btrfs_fs_info
*fs_info
,
4207 enum btrfs_feature_set set
,
4208 u64 change_mask
, u64 flags
, u64 supported_flags
,
4209 u64 safe_set
, u64 safe_clear
)
4211 const char *type
= btrfs_feature_set_name(set
);
4213 u64 disallowed
, unsupported
;
4214 u64 set_mask
= flags
& change_mask
;
4215 u64 clear_mask
= ~flags
& change_mask
;
4217 unsupported
= set_mask
& ~supported_flags
;
4219 names
= btrfs_printable_features(set
, unsupported
);
4222 "this kernel does not support the %s feature bit%s",
4223 names
, strchr(names
, ',') ? "s" : "");
4227 "this kernel does not support %s bits 0x%llx",
4232 disallowed
= set_mask
& ~safe_set
;
4234 names
= btrfs_printable_features(set
, disallowed
);
4237 "can't set the %s feature bit%s while mounted",
4238 names
, strchr(names
, ',') ? "s" : "");
4242 "can't set %s bits 0x%llx while mounted",
4247 disallowed
= clear_mask
& ~safe_clear
;
4249 names
= btrfs_printable_features(set
, disallowed
);
4252 "can't clear the %s feature bit%s while mounted",
4253 names
, strchr(names
, ',') ? "s" : "");
4257 "can't clear %s bits 0x%llx while mounted",
4265 #define check_feature(fs_info, change_mask, flags, mask_base) \
4266 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4267 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4268 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4269 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4271 static int btrfs_ioctl_set_features(struct file
*file
, void __user
*arg
)
4273 struct inode
*inode
= file_inode(file
);
4274 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4275 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4276 struct btrfs_super_block
*super_block
= fs_info
->super_copy
;
4277 struct btrfs_ioctl_feature_flags flags
[2];
4278 struct btrfs_trans_handle
*trans
;
4282 if (!capable(CAP_SYS_ADMIN
))
4285 if (copy_from_user(flags
, arg
, sizeof(flags
)))
4289 if (!flags
[0].compat_flags
&& !flags
[0].compat_ro_flags
&&
4290 !flags
[0].incompat_flags
)
4293 ret
= check_feature(fs_info
, flags
[0].compat_flags
,
4294 flags
[1].compat_flags
, COMPAT
);
4298 ret
= check_feature(fs_info
, flags
[0].compat_ro_flags
,
4299 flags
[1].compat_ro_flags
, COMPAT_RO
);
4303 ret
= check_feature(fs_info
, flags
[0].incompat_flags
,
4304 flags
[1].incompat_flags
, INCOMPAT
);
4308 ret
= mnt_want_write_file(file
);
4312 trans
= btrfs_start_transaction(root
, 0);
4313 if (IS_ERR(trans
)) {
4314 ret
= PTR_ERR(trans
);
4315 goto out_drop_write
;
4318 spin_lock(&fs_info
->super_lock
);
4319 newflags
= btrfs_super_compat_flags(super_block
);
4320 newflags
|= flags
[0].compat_flags
& flags
[1].compat_flags
;
4321 newflags
&= ~(flags
[0].compat_flags
& ~flags
[1].compat_flags
);
4322 btrfs_set_super_compat_flags(super_block
, newflags
);
4324 newflags
= btrfs_super_compat_ro_flags(super_block
);
4325 newflags
|= flags
[0].compat_ro_flags
& flags
[1].compat_ro_flags
;
4326 newflags
&= ~(flags
[0].compat_ro_flags
& ~flags
[1].compat_ro_flags
);
4327 btrfs_set_super_compat_ro_flags(super_block
, newflags
);
4329 newflags
= btrfs_super_incompat_flags(super_block
);
4330 newflags
|= flags
[0].incompat_flags
& flags
[1].incompat_flags
;
4331 newflags
&= ~(flags
[0].incompat_flags
& ~flags
[1].incompat_flags
);
4332 btrfs_set_super_incompat_flags(super_block
, newflags
);
4333 spin_unlock(&fs_info
->super_lock
);
4335 ret
= btrfs_commit_transaction(trans
);
4337 mnt_drop_write_file(file
);
4342 static int _btrfs_ioctl_send(struct inode
*inode
, void __user
*argp
, bool compat
)
4344 struct btrfs_ioctl_send_args
*arg
;
4348 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4349 struct btrfs_ioctl_send_args_32 args32
= { 0 };
4351 ret
= copy_from_user(&args32
, argp
, sizeof(args32
));
4354 arg
= kzalloc(sizeof(*arg
), GFP_KERNEL
);
4357 arg
->send_fd
= args32
.send_fd
;
4358 arg
->clone_sources_count
= args32
.clone_sources_count
;
4359 arg
->clone_sources
= compat_ptr(args32
.clone_sources
);
4360 arg
->parent_root
= args32
.parent_root
;
4361 arg
->flags
= args32
.flags
;
4362 arg
->version
= args32
.version
;
4363 memcpy(arg
->reserved
, args32
.reserved
,
4364 sizeof(args32
.reserved
));
4369 arg
= memdup_user(argp
, sizeof(*arg
));
4371 return PTR_ERR(arg
);
4373 ret
= btrfs_ioctl_send(inode
, arg
);
4378 static int btrfs_ioctl_encoded_read(struct file
*file
, void __user
*argp
,
4381 struct btrfs_ioctl_encoded_io_args args
= { 0 };
4382 size_t copy_end_kernel
= offsetofend(struct btrfs_ioctl_encoded_io_args
,
4385 struct iovec iovstack
[UIO_FASTIOV
];
4386 struct iovec
*iov
= iovstack
;
4387 struct iov_iter iter
;
4392 if (!capable(CAP_SYS_ADMIN
)) {
4398 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4399 struct btrfs_ioctl_encoded_io_args_32 args32
;
4401 copy_end
= offsetofend(struct btrfs_ioctl_encoded_io_args_32
,
4403 if (copy_from_user(&args32
, argp
, copy_end
)) {
4407 args
.iov
= compat_ptr(args32
.iov
);
4408 args
.iovcnt
= args32
.iovcnt
;
4409 args
.offset
= args32
.offset
;
4410 args
.flags
= args32
.flags
;
4415 copy_end
= copy_end_kernel
;
4416 if (copy_from_user(&args
, argp
, copy_end
)) {
4421 if (args
.flags
!= 0) {
4426 ret
= import_iovec(ITER_DEST
, args
.iov
, args
.iovcnt
, ARRAY_SIZE(iovstack
),
4431 if (iov_iter_count(&iter
) == 0) {
4436 ret
= rw_verify_area(READ
, file
, &pos
, args
.len
);
4440 init_sync_kiocb(&kiocb
, file
);
4443 ret
= btrfs_encoded_read(&kiocb
, &iter
, &args
);
4445 fsnotify_access(file
);
4446 if (copy_to_user(argp
+ copy_end
,
4447 (char *)&args
+ copy_end_kernel
,
4448 sizeof(args
) - copy_end_kernel
))
4456 add_rchar(current
, ret
);
4461 static int btrfs_ioctl_encoded_write(struct file
*file
, void __user
*argp
, bool compat
)
4463 struct btrfs_ioctl_encoded_io_args args
;
4464 struct iovec iovstack
[UIO_FASTIOV
];
4465 struct iovec
*iov
= iovstack
;
4466 struct iov_iter iter
;
4471 if (!capable(CAP_SYS_ADMIN
)) {
4476 if (!(file
->f_mode
& FMODE_WRITE
)) {
4482 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4483 struct btrfs_ioctl_encoded_io_args_32 args32
;
4485 if (copy_from_user(&args32
, argp
, sizeof(args32
))) {
4489 args
.iov
= compat_ptr(args32
.iov
);
4490 args
.iovcnt
= args32
.iovcnt
;
4491 args
.offset
= args32
.offset
;
4492 args
.flags
= args32
.flags
;
4493 args
.len
= args32
.len
;
4494 args
.unencoded_len
= args32
.unencoded_len
;
4495 args
.unencoded_offset
= args32
.unencoded_offset
;
4496 args
.compression
= args32
.compression
;
4497 args
.encryption
= args32
.encryption
;
4498 memcpy(args
.reserved
, args32
.reserved
, sizeof(args
.reserved
));
4503 if (copy_from_user(&args
, argp
, sizeof(args
))) {
4510 if (args
.flags
!= 0)
4512 if (memchr_inv(args
.reserved
, 0, sizeof(args
.reserved
)))
4514 if (args
.compression
== BTRFS_ENCODED_IO_COMPRESSION_NONE
&&
4515 args
.encryption
== BTRFS_ENCODED_IO_ENCRYPTION_NONE
)
4517 if (args
.compression
>= BTRFS_ENCODED_IO_COMPRESSION_TYPES
||
4518 args
.encryption
>= BTRFS_ENCODED_IO_ENCRYPTION_TYPES
)
4520 if (args
.unencoded_offset
> args
.unencoded_len
)
4522 if (args
.len
> args
.unencoded_len
- args
.unencoded_offset
)
4525 ret
= import_iovec(ITER_SOURCE
, args
.iov
, args
.iovcnt
, ARRAY_SIZE(iovstack
),
4530 file_start_write(file
);
4532 if (iov_iter_count(&iter
) == 0) {
4537 ret
= rw_verify_area(WRITE
, file
, &pos
, args
.len
);
4541 init_sync_kiocb(&kiocb
, file
);
4542 ret
= kiocb_set_rw_flags(&kiocb
, 0);
4547 ret
= btrfs_do_write_iter(&kiocb
, &iter
, &args
);
4549 fsnotify_modify(file
);
4552 file_end_write(file
);
4556 add_wchar(current
, ret
);
4561 long btrfs_ioctl(struct file
*file
, unsigned int
4562 cmd
, unsigned long arg
)
4564 struct inode
*inode
= file_inode(file
);
4565 struct btrfs_fs_info
*fs_info
= btrfs_sb(inode
->i_sb
);
4566 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
4567 void __user
*argp
= (void __user
*)arg
;
4570 case FS_IOC_GETVERSION
:
4571 return btrfs_ioctl_getversion(inode
, argp
);
4572 case FS_IOC_GETFSLABEL
:
4573 return btrfs_ioctl_get_fslabel(fs_info
, argp
);
4574 case FS_IOC_SETFSLABEL
:
4575 return btrfs_ioctl_set_fslabel(file
, argp
);
4577 return btrfs_ioctl_fitrim(fs_info
, argp
);
4578 case BTRFS_IOC_SNAP_CREATE
:
4579 return btrfs_ioctl_snap_create(file
, argp
, 0);
4580 case BTRFS_IOC_SNAP_CREATE_V2
:
4581 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
4582 case BTRFS_IOC_SUBVOL_CREATE
:
4583 return btrfs_ioctl_snap_create(file
, argp
, 1);
4584 case BTRFS_IOC_SUBVOL_CREATE_V2
:
4585 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
4586 case BTRFS_IOC_SNAP_DESTROY
:
4587 return btrfs_ioctl_snap_destroy(file
, argp
, false);
4588 case BTRFS_IOC_SNAP_DESTROY_V2
:
4589 return btrfs_ioctl_snap_destroy(file
, argp
, true);
4590 case BTRFS_IOC_SUBVOL_GETFLAGS
:
4591 return btrfs_ioctl_subvol_getflags(inode
, argp
);
4592 case BTRFS_IOC_SUBVOL_SETFLAGS
:
4593 return btrfs_ioctl_subvol_setflags(file
, argp
);
4594 case BTRFS_IOC_DEFAULT_SUBVOL
:
4595 return btrfs_ioctl_default_subvol(file
, argp
);
4596 case BTRFS_IOC_DEFRAG
:
4597 return btrfs_ioctl_defrag(file
, NULL
);
4598 case BTRFS_IOC_DEFRAG_RANGE
:
4599 return btrfs_ioctl_defrag(file
, argp
);
4600 case BTRFS_IOC_RESIZE
:
4601 return btrfs_ioctl_resize(file
, argp
);
4602 case BTRFS_IOC_ADD_DEV
:
4603 return btrfs_ioctl_add_dev(fs_info
, argp
);
4604 case BTRFS_IOC_RM_DEV
:
4605 return btrfs_ioctl_rm_dev(file
, argp
);
4606 case BTRFS_IOC_RM_DEV_V2
:
4607 return btrfs_ioctl_rm_dev_v2(file
, argp
);
4608 case BTRFS_IOC_FS_INFO
:
4609 return btrfs_ioctl_fs_info(fs_info
, argp
);
4610 case BTRFS_IOC_DEV_INFO
:
4611 return btrfs_ioctl_dev_info(fs_info
, argp
);
4612 case BTRFS_IOC_TREE_SEARCH
:
4613 return btrfs_ioctl_tree_search(inode
, argp
);
4614 case BTRFS_IOC_TREE_SEARCH_V2
:
4615 return btrfs_ioctl_tree_search_v2(inode
, argp
);
4616 case BTRFS_IOC_INO_LOOKUP
:
4617 return btrfs_ioctl_ino_lookup(root
, argp
);
4618 case BTRFS_IOC_INO_PATHS
:
4619 return btrfs_ioctl_ino_to_path(root
, argp
);
4620 case BTRFS_IOC_LOGICAL_INO
:
4621 return btrfs_ioctl_logical_to_ino(fs_info
, argp
, 1);
4622 case BTRFS_IOC_LOGICAL_INO_V2
:
4623 return btrfs_ioctl_logical_to_ino(fs_info
, argp
, 2);
4624 case BTRFS_IOC_SPACE_INFO
:
4625 return btrfs_ioctl_space_info(fs_info
, argp
);
4626 case BTRFS_IOC_SYNC
: {
4629 ret
= btrfs_start_delalloc_roots(fs_info
, LONG_MAX
, false);
4632 ret
= btrfs_sync_fs(inode
->i_sb
, 1);
4634 * The transaction thread may want to do more work,
4635 * namely it pokes the cleaner kthread that will start
4636 * processing uncleaned subvols.
4638 wake_up_process(fs_info
->transaction_kthread
);
4641 case BTRFS_IOC_START_SYNC
:
4642 return btrfs_ioctl_start_sync(root
, argp
);
4643 case BTRFS_IOC_WAIT_SYNC
:
4644 return btrfs_ioctl_wait_sync(fs_info
, argp
);
4645 case BTRFS_IOC_SCRUB
:
4646 return btrfs_ioctl_scrub(file
, argp
);
4647 case BTRFS_IOC_SCRUB_CANCEL
:
4648 return btrfs_ioctl_scrub_cancel(fs_info
);
4649 case BTRFS_IOC_SCRUB_PROGRESS
:
4650 return btrfs_ioctl_scrub_progress(fs_info
, argp
);
4651 case BTRFS_IOC_BALANCE_V2
:
4652 return btrfs_ioctl_balance(file
, argp
);
4653 case BTRFS_IOC_BALANCE_CTL
:
4654 return btrfs_ioctl_balance_ctl(fs_info
, arg
);
4655 case BTRFS_IOC_BALANCE_PROGRESS
:
4656 return btrfs_ioctl_balance_progress(fs_info
, argp
);
4657 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
4658 return btrfs_ioctl_set_received_subvol(file
, argp
);
4660 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32
:
4661 return btrfs_ioctl_set_received_subvol_32(file
, argp
);
4663 case BTRFS_IOC_SEND
:
4664 return _btrfs_ioctl_send(inode
, argp
, false);
4665 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4666 case BTRFS_IOC_SEND_32
:
4667 return _btrfs_ioctl_send(inode
, argp
, true);
4669 case BTRFS_IOC_GET_DEV_STATS
:
4670 return btrfs_ioctl_get_dev_stats(fs_info
, argp
);
4671 case BTRFS_IOC_QUOTA_CTL
:
4672 return btrfs_ioctl_quota_ctl(file
, argp
);
4673 case BTRFS_IOC_QGROUP_ASSIGN
:
4674 return btrfs_ioctl_qgroup_assign(file
, argp
);
4675 case BTRFS_IOC_QGROUP_CREATE
:
4676 return btrfs_ioctl_qgroup_create(file
, argp
);
4677 case BTRFS_IOC_QGROUP_LIMIT
:
4678 return btrfs_ioctl_qgroup_limit(file
, argp
);
4679 case BTRFS_IOC_QUOTA_RESCAN
:
4680 return btrfs_ioctl_quota_rescan(file
, argp
);
4681 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
4682 return btrfs_ioctl_quota_rescan_status(fs_info
, argp
);
4683 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
4684 return btrfs_ioctl_quota_rescan_wait(fs_info
, argp
);
4685 case BTRFS_IOC_DEV_REPLACE
:
4686 return btrfs_ioctl_dev_replace(fs_info
, argp
);
4687 case BTRFS_IOC_GET_SUPPORTED_FEATURES
:
4688 return btrfs_ioctl_get_supported_features(argp
);
4689 case BTRFS_IOC_GET_FEATURES
:
4690 return btrfs_ioctl_get_features(fs_info
, argp
);
4691 case BTRFS_IOC_SET_FEATURES
:
4692 return btrfs_ioctl_set_features(file
, argp
);
4693 case BTRFS_IOC_GET_SUBVOL_INFO
:
4694 return btrfs_ioctl_get_subvol_info(inode
, argp
);
4695 case BTRFS_IOC_GET_SUBVOL_ROOTREF
:
4696 return btrfs_ioctl_get_subvol_rootref(root
, argp
);
4697 case BTRFS_IOC_INO_LOOKUP_USER
:
4698 return btrfs_ioctl_ino_lookup_user(file
, argp
);
4699 case FS_IOC_ENABLE_VERITY
:
4700 return fsverity_ioctl_enable(file
, (const void __user
*)argp
);
4701 case FS_IOC_MEASURE_VERITY
:
4702 return fsverity_ioctl_measure(file
, argp
);
4703 case BTRFS_IOC_ENCODED_READ
:
4704 return btrfs_ioctl_encoded_read(file
, argp
, false);
4705 case BTRFS_IOC_ENCODED_WRITE
:
4706 return btrfs_ioctl_encoded_write(file
, argp
, false);
4707 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4708 case BTRFS_IOC_ENCODED_READ_32
:
4709 return btrfs_ioctl_encoded_read(file
, argp
, true);
4710 case BTRFS_IOC_ENCODED_WRITE_32
:
4711 return btrfs_ioctl_encoded_write(file
, argp
, true);
4718 #ifdef CONFIG_COMPAT
4719 long btrfs_compat_ioctl(struct file
*file
, unsigned int cmd
, unsigned long arg
)
4722 * These all access 32-bit values anyway so no further
4723 * handling is necessary.
4726 case FS_IOC32_GETVERSION
:
4727 cmd
= FS_IOC_GETVERSION
;
4731 return btrfs_ioctl(file
, cmd
, (unsigned long) compat_ptr(arg
));