2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
60 /* Mask out flags that are inappropriate for the given type of inode. */
61 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
65 else if (S_ISREG(mode
))
66 return flags
& ~FS_DIRSYNC_FL
;
68 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
72 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
74 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
76 unsigned int iflags
= 0;
78 if (flags
& BTRFS_INODE_SYNC
)
80 if (flags
& BTRFS_INODE_IMMUTABLE
)
81 iflags
|= FS_IMMUTABLE_FL
;
82 if (flags
& BTRFS_INODE_APPEND
)
83 iflags
|= FS_APPEND_FL
;
84 if (flags
& BTRFS_INODE_NODUMP
)
85 iflags
|= FS_NODUMP_FL
;
86 if (flags
& BTRFS_INODE_NOATIME
)
87 iflags
|= FS_NOATIME_FL
;
88 if (flags
& BTRFS_INODE_DIRSYNC
)
89 iflags
|= FS_DIRSYNC_FL
;
90 if (flags
& BTRFS_INODE_NODATACOW
)
91 iflags
|= FS_NOCOW_FL
;
93 if ((flags
& BTRFS_INODE_COMPRESS
) && !(flags
& BTRFS_INODE_NOCOMPRESS
))
94 iflags
|= FS_COMPR_FL
;
95 else if (flags
& BTRFS_INODE_NOCOMPRESS
)
96 iflags
|= FS_NOCOMP_FL
;
102 * Update inode->i_flags based on the btrfs internal flags.
104 void btrfs_update_iflags(struct inode
*inode
)
106 struct btrfs_inode
*ip
= BTRFS_I(inode
);
108 inode
->i_flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
110 if (ip
->flags
& BTRFS_INODE_SYNC
)
111 inode
->i_flags
|= S_SYNC
;
112 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
113 inode
->i_flags
|= S_IMMUTABLE
;
114 if (ip
->flags
& BTRFS_INODE_APPEND
)
115 inode
->i_flags
|= S_APPEND
;
116 if (ip
->flags
& BTRFS_INODE_NOATIME
)
117 inode
->i_flags
|= S_NOATIME
;
118 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
119 inode
->i_flags
|= S_DIRSYNC
;
123 * Inherit flags from the parent inode.
125 * Currently only the compression flags and the cow flags are inherited.
127 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
134 flags
= BTRFS_I(dir
)->flags
;
136 if (flags
& BTRFS_INODE_NOCOMPRESS
) {
137 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_COMPRESS
;
138 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NOCOMPRESS
;
139 } else if (flags
& BTRFS_INODE_COMPRESS
) {
140 BTRFS_I(inode
)->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
141 BTRFS_I(inode
)->flags
|= BTRFS_INODE_COMPRESS
;
144 if (flags
& BTRFS_INODE_NODATACOW
) {
145 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATACOW
;
146 if (S_ISREG(inode
->i_mode
))
147 BTRFS_I(inode
)->flags
|= BTRFS_INODE_NODATASUM
;
150 btrfs_update_iflags(inode
);
153 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
155 struct btrfs_inode
*ip
= BTRFS_I(file_inode(file
));
156 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
158 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
163 static int check_flags(unsigned int flags
)
165 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
166 FS_NOATIME_FL
| FS_NODUMP_FL
| \
167 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
168 FS_NOCOMP_FL
| FS_COMPR_FL
|
172 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
178 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
180 struct inode
*inode
= file_inode(file
);
181 struct btrfs_inode
*ip
= BTRFS_I(inode
);
182 struct btrfs_root
*root
= ip
->root
;
183 struct btrfs_trans_handle
*trans
;
184 unsigned int flags
, oldflags
;
187 unsigned int i_oldflags
;
190 if (btrfs_root_readonly(root
))
193 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
196 ret
= check_flags(flags
);
200 if (!inode_owner_or_capable(inode
))
203 ret
= mnt_want_write_file(file
);
207 mutex_lock(&inode
->i_mutex
);
209 ip_oldflags
= ip
->flags
;
210 i_oldflags
= inode
->i_flags
;
211 mode
= inode
->i_mode
;
213 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
214 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
215 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
216 if (!capable(CAP_LINUX_IMMUTABLE
)) {
222 if (flags
& FS_SYNC_FL
)
223 ip
->flags
|= BTRFS_INODE_SYNC
;
225 ip
->flags
&= ~BTRFS_INODE_SYNC
;
226 if (flags
& FS_IMMUTABLE_FL
)
227 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
229 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
230 if (flags
& FS_APPEND_FL
)
231 ip
->flags
|= BTRFS_INODE_APPEND
;
233 ip
->flags
&= ~BTRFS_INODE_APPEND
;
234 if (flags
& FS_NODUMP_FL
)
235 ip
->flags
|= BTRFS_INODE_NODUMP
;
237 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
238 if (flags
& FS_NOATIME_FL
)
239 ip
->flags
|= BTRFS_INODE_NOATIME
;
241 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
242 if (flags
& FS_DIRSYNC_FL
)
243 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
245 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
246 if (flags
& FS_NOCOW_FL
) {
249 * It's safe to turn csums off here, no extents exist.
250 * Otherwise we want the flag to reflect the real COW
251 * status of the file and will not set it.
253 if (inode
->i_size
== 0)
254 ip
->flags
|= BTRFS_INODE_NODATACOW
255 | BTRFS_INODE_NODATASUM
;
257 ip
->flags
|= BTRFS_INODE_NODATACOW
;
261 * Revert back under same assuptions as above
264 if (inode
->i_size
== 0)
265 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
266 | BTRFS_INODE_NODATASUM
);
268 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
273 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
274 * flag may be changed automatically if compression code won't make
277 if (flags
& FS_NOCOMP_FL
) {
278 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
279 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
280 } else if (flags
& FS_COMPR_FL
) {
281 ip
->flags
|= BTRFS_INODE_COMPRESS
;
282 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
284 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
287 trans
= btrfs_start_transaction(root
, 1);
289 ret
= PTR_ERR(trans
);
293 btrfs_update_iflags(inode
);
294 inode_inc_iversion(inode
);
295 inode
->i_ctime
= CURRENT_TIME
;
296 ret
= btrfs_update_inode(trans
, root
, inode
);
298 btrfs_end_transaction(trans
, root
);
301 ip
->flags
= ip_oldflags
;
302 inode
->i_flags
= i_oldflags
;
306 mutex_unlock(&inode
->i_mutex
);
307 mnt_drop_write_file(file
);
311 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
313 struct inode
*inode
= file_inode(file
);
315 return put_user(inode
->i_generation
, arg
);
318 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
320 struct btrfs_fs_info
*fs_info
= btrfs_sb(fdentry(file
)->d_sb
);
321 struct btrfs_device
*device
;
322 struct request_queue
*q
;
323 struct fstrim_range range
;
324 u64 minlen
= ULLONG_MAX
;
326 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
329 if (!capable(CAP_SYS_ADMIN
))
333 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
337 q
= bdev_get_queue(device
->bdev
);
338 if (blk_queue_discard(q
)) {
340 minlen
= min((u64
)q
->limits
.discard_granularity
,
348 if (copy_from_user(&range
, arg
, sizeof(range
)))
350 if (range
.start
> total_bytes
||
351 range
.len
< fs_info
->sb
->s_blocksize
)
354 range
.len
= min(range
.len
, total_bytes
- range
.start
);
355 range
.minlen
= max(range
.minlen
, minlen
);
356 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
360 if (copy_to_user(arg
, &range
, sizeof(range
)))
366 static noinline
int create_subvol(struct inode
*dir
,
367 struct dentry
*dentry
,
368 char *name
, int namelen
,
370 struct btrfs_qgroup_inherit
*inherit
)
372 struct btrfs_trans_handle
*trans
;
373 struct btrfs_key key
;
374 struct btrfs_root_item root_item
;
375 struct btrfs_inode_item
*inode_item
;
376 struct extent_buffer
*leaf
;
377 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
378 struct btrfs_root
*new_root
;
379 struct btrfs_block_rsv block_rsv
;
380 struct timespec cur_time
= CURRENT_TIME
;
384 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
389 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
393 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
395 * The same as the snapshot creation, please see the comment
396 * of create_snapshot().
398 ret
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
399 7, &qgroup_reserved
);
403 trans
= btrfs_start_transaction(root
, 0);
405 ret
= PTR_ERR(trans
);
408 trans
->block_rsv
= &block_rsv
;
409 trans
->bytes_reserved
= block_rsv
.size
;
411 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
415 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
416 0, objectid
, NULL
, 0, 0, 0);
422 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
423 btrfs_set_header_bytenr(leaf
, leaf
->start
);
424 btrfs_set_header_generation(leaf
, trans
->transid
);
425 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
426 btrfs_set_header_owner(leaf
, objectid
);
428 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
429 (unsigned long)btrfs_header_fsid(leaf
),
431 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
432 (unsigned long)btrfs_header_chunk_tree_uuid(leaf
),
434 btrfs_mark_buffer_dirty(leaf
);
436 memset(&root_item
, 0, sizeof(root_item
));
438 inode_item
= &root_item
.inode
;
439 inode_item
->generation
= cpu_to_le64(1);
440 inode_item
->size
= cpu_to_le64(3);
441 inode_item
->nlink
= cpu_to_le32(1);
442 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
443 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
446 root_item
.byte_limit
= 0;
447 inode_item
->flags
= cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT
);
449 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
450 btrfs_set_root_generation(&root_item
, trans
->transid
);
451 btrfs_set_root_level(&root_item
, 0);
452 btrfs_set_root_refs(&root_item
, 1);
453 btrfs_set_root_used(&root_item
, leaf
->len
);
454 btrfs_set_root_last_snapshot(&root_item
, 0);
456 btrfs_set_root_generation_v2(&root_item
,
457 btrfs_root_generation(&root_item
));
458 uuid_le_gen(&new_uuid
);
459 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
460 root_item
.otime
.sec
= cpu_to_le64(cur_time
.tv_sec
);
461 root_item
.otime
.nsec
= cpu_to_le32(cur_time
.tv_nsec
);
462 root_item
.ctime
= root_item
.otime
;
463 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
464 btrfs_set_root_otransid(&root_item
, trans
->transid
);
466 btrfs_tree_unlock(leaf
);
467 free_extent_buffer(leaf
);
470 btrfs_set_root_dirid(&root_item
, new_dirid
);
472 key
.objectid
= objectid
;
474 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
475 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
480 key
.offset
= (u64
)-1;
481 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
482 if (IS_ERR(new_root
)) {
483 btrfs_abort_transaction(trans
, root
, PTR_ERR(new_root
));
484 ret
= PTR_ERR(new_root
);
488 btrfs_record_root_in_trans(trans
, new_root
);
490 ret
= btrfs_create_subvol_root(trans
, new_root
, new_dirid
);
492 /* We potentially lose an unused inode item here */
493 btrfs_abort_transaction(trans
, root
, ret
);
498 * insert the directory item
500 ret
= btrfs_set_inode_index(dir
, &index
);
502 btrfs_abort_transaction(trans
, root
, ret
);
506 ret
= btrfs_insert_dir_item(trans
, root
,
507 name
, namelen
, dir
, &key
,
508 BTRFS_FT_DIR
, index
);
510 btrfs_abort_transaction(trans
, root
, ret
);
514 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
515 ret
= btrfs_update_inode(trans
, root
, dir
);
518 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
519 objectid
, root
->root_key
.objectid
,
520 btrfs_ino(dir
), index
, name
, namelen
);
525 trans
->block_rsv
= NULL
;
526 trans
->bytes_reserved
= 0;
528 *async_transid
= trans
->transid
;
529 err
= btrfs_commit_transaction_async(trans
, root
, 1);
531 err
= btrfs_commit_transaction(trans
, root
);
533 err
= btrfs_commit_transaction(trans
, root
);
539 d_instantiate(dentry
, btrfs_lookup_dentry(dir
, dentry
));
541 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
545 static int create_snapshot(struct btrfs_root
*root
, struct inode
*dir
,
546 struct dentry
*dentry
, char *name
, int namelen
,
547 u64
*async_transid
, bool readonly
,
548 struct btrfs_qgroup_inherit
*inherit
)
551 struct btrfs_pending_snapshot
*pending_snapshot
;
552 struct btrfs_trans_handle
*trans
;
558 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
559 if (!pending_snapshot
)
562 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
563 BTRFS_BLOCK_RSV_TEMP
);
565 * 1 - parent dir inode
568 * 2 - root ref/backref
569 * 1 - root of snapshot
571 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
572 &pending_snapshot
->block_rsv
, 7,
573 &pending_snapshot
->qgroup_reserved
);
577 pending_snapshot
->dentry
= dentry
;
578 pending_snapshot
->root
= root
;
579 pending_snapshot
->readonly
= readonly
;
580 pending_snapshot
->dir
= dir
;
581 pending_snapshot
->inherit
= inherit
;
583 trans
= btrfs_start_transaction(root
, 0);
585 ret
= PTR_ERR(trans
);
589 spin_lock(&root
->fs_info
->trans_lock
);
590 list_add(&pending_snapshot
->list
,
591 &trans
->transaction
->pending_snapshots
);
592 spin_unlock(&root
->fs_info
->trans_lock
);
594 *async_transid
= trans
->transid
;
595 ret
= btrfs_commit_transaction_async(trans
,
596 root
->fs_info
->extent_root
, 1);
598 ret
= btrfs_commit_transaction(trans
, root
);
600 ret
= btrfs_commit_transaction(trans
,
601 root
->fs_info
->extent_root
);
606 ret
= pending_snapshot
->error
;
610 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
614 inode
= btrfs_lookup_dentry(dentry
->d_parent
->d_inode
, dentry
);
616 ret
= PTR_ERR(inode
);
620 d_instantiate(dentry
, inode
);
623 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
624 &pending_snapshot
->block_rsv
,
625 pending_snapshot
->qgroup_reserved
);
627 kfree(pending_snapshot
);
631 /* copy of check_sticky in fs/namei.c()
632 * It's inline, so penalty for filesystems that don't use sticky bit is
635 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
637 kuid_t fsuid
= current_fsuid();
639 if (!(dir
->i_mode
& S_ISVTX
))
641 if (uid_eq(inode
->i_uid
, fsuid
))
643 if (uid_eq(dir
->i_uid
, fsuid
))
645 return !capable(CAP_FOWNER
);
648 /* copy of may_delete in fs/namei.c()
649 * Check whether we can remove a link victim from directory dir, check
650 * whether the type of victim is right.
651 * 1. We can't do it if dir is read-only (done in permission())
652 * 2. We should have write and exec permissions on dir
653 * 3. We can't remove anything from append-only dir
654 * 4. We can't do anything with immutable dir (done in permission())
655 * 5. If the sticky bit on dir is set we should either
656 * a. be owner of dir, or
657 * b. be owner of victim, or
658 * c. have CAP_FOWNER capability
659 * 6. If the victim is append-only or immutable we can't do antyhing with
660 * links pointing to it.
661 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
662 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
663 * 9. We can't remove a root or mountpoint.
664 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
665 * nfs_async_unlink().
668 static int btrfs_may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
672 if (!victim
->d_inode
)
675 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
676 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
678 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
683 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
684 IS_APPEND(victim
->d_inode
)||
685 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
688 if (!S_ISDIR(victim
->d_inode
->i_mode
))
692 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
696 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
701 /* copy of may_create in fs/namei.c() */
702 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
708 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
712 * Create a new subvolume below @parent. This is largely modeled after
713 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
714 * inside this filesystem so it's quite a bit simpler.
716 static noinline
int btrfs_mksubvol(struct path
*parent
,
717 char *name
, int namelen
,
718 struct btrfs_root
*snap_src
,
719 u64
*async_transid
, bool readonly
,
720 struct btrfs_qgroup_inherit
*inherit
)
722 struct inode
*dir
= parent
->dentry
->d_inode
;
723 struct dentry
*dentry
;
726 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
730 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
731 error
= PTR_ERR(dentry
);
739 error
= btrfs_may_create(dir
, dentry
);
744 * even if this name doesn't exist, we may get hash collisions.
745 * check for them now when we can safely fail
747 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
753 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
755 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
759 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
760 async_transid
, readonly
, inherit
);
762 error
= create_subvol(dir
, dentry
, name
, namelen
,
763 async_transid
, inherit
);
766 fsnotify_mkdir(dir
, dentry
);
768 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
772 mutex_unlock(&dir
->i_mutex
);
777 * When we're defragging a range, we don't want to kick it off again
778 * if it is really just waiting for delalloc to send it down.
779 * If we find a nice big extent or delalloc range for the bytes in the
780 * file you want to defrag, we return 0 to let you know to skip this
783 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, int thresh
)
785 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
786 struct extent_map
*em
= NULL
;
787 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
790 read_lock(&em_tree
->lock
);
791 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
792 read_unlock(&em_tree
->lock
);
795 end
= extent_map_end(em
);
797 if (end
- offset
> thresh
)
800 /* if we already have a nice delalloc here, just stop */
802 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
803 thresh
, EXTENT_DELALLOC
, 1);
810 * helper function to walk through a file and find extents
811 * newer than a specific transid, and smaller than thresh.
813 * This is used by the defragging code to find new and small
816 static int find_new_extents(struct btrfs_root
*root
,
817 struct inode
*inode
, u64 newer_than
,
818 u64
*off
, int thresh
)
820 struct btrfs_path
*path
;
821 struct btrfs_key min_key
;
822 struct btrfs_key max_key
;
823 struct extent_buffer
*leaf
;
824 struct btrfs_file_extent_item
*extent
;
827 u64 ino
= btrfs_ino(inode
);
829 path
= btrfs_alloc_path();
833 min_key
.objectid
= ino
;
834 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
835 min_key
.offset
= *off
;
837 max_key
.objectid
= ino
;
838 max_key
.type
= (u8
)-1;
839 max_key
.offset
= (u64
)-1;
841 path
->keep_locks
= 1;
844 ret
= btrfs_search_forward(root
, &min_key
, &max_key
,
848 if (min_key
.objectid
!= ino
)
850 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
853 leaf
= path
->nodes
[0];
854 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
855 struct btrfs_file_extent_item
);
857 type
= btrfs_file_extent_type(leaf
, extent
);
858 if (type
== BTRFS_FILE_EXTENT_REG
&&
859 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
860 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
861 *off
= min_key
.offset
;
862 btrfs_free_path(path
);
866 if (min_key
.offset
== (u64
)-1)
870 btrfs_release_path(path
);
873 btrfs_free_path(path
);
877 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
879 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
880 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
881 struct extent_map
*em
;
882 u64 len
= PAGE_CACHE_SIZE
;
885 * hopefully we have this extent in the tree already, try without
886 * the full extent lock
888 read_lock(&em_tree
->lock
);
889 em
= lookup_extent_mapping(em_tree
, start
, len
);
890 read_unlock(&em_tree
->lock
);
893 /* get the big lock and read metadata off disk */
894 lock_extent(io_tree
, start
, start
+ len
- 1);
895 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
896 unlock_extent(io_tree
, start
, start
+ len
- 1);
905 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
907 struct extent_map
*next
;
910 /* this is the last extent */
911 if (em
->start
+ em
->len
>= i_size_read(inode
))
914 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
915 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
918 free_extent_map(next
);
922 static int should_defrag_range(struct inode
*inode
, u64 start
, int thresh
,
923 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
926 struct extent_map
*em
;
928 bool next_mergeable
= true;
931 * make sure that once we start defragging an extent, we keep on
934 if (start
< *defrag_end
)
939 em
= defrag_lookup_extent(inode
, start
);
943 /* this will cover holes, and inline extents */
944 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
949 next_mergeable
= defrag_check_next_extent(inode
, em
);
952 * we hit a real extent, if it is big or the next extent is not a
953 * real extent, don't bother defragging it
955 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
956 (em
->len
>= thresh
|| !next_mergeable
))
960 * last_len ends up being a counter of how many bytes we've defragged.
961 * every time we choose not to defrag an extent, we reset *last_len
962 * so that the next tiny extent will force a defrag.
964 * The end result of this is that tiny extents before a single big
965 * extent will force at least part of that big extent to be defragged.
968 *defrag_end
= extent_map_end(em
);
971 *skip
= extent_map_end(em
);
980 * it doesn't do much good to defrag one or two pages
981 * at a time. This pulls in a nice chunk of pages
984 * It also makes sure the delalloc code has enough
985 * dirty data to avoid making new small extents as part
988 * It's a good idea to start RA on this range
989 * before calling this.
991 static int cluster_pages_for_defrag(struct inode
*inode
,
993 unsigned long start_index
,
996 unsigned long file_end
;
997 u64 isize
= i_size_read(inode
);
1004 struct btrfs_ordered_extent
*ordered
;
1005 struct extent_state
*cached_state
= NULL
;
1006 struct extent_io_tree
*tree
;
1007 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1009 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1010 if (!isize
|| start_index
> file_end
)
1013 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1015 ret
= btrfs_delalloc_reserve_space(inode
,
1016 page_cnt
<< PAGE_CACHE_SHIFT
);
1020 tree
= &BTRFS_I(inode
)->io_tree
;
1022 /* step one, lock all the pages */
1023 for (i
= 0; i
< page_cnt
; i
++) {
1026 page
= find_or_create_page(inode
->i_mapping
,
1027 start_index
+ i
, mask
);
1031 page_start
= page_offset(page
);
1032 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1034 lock_extent(tree
, page_start
, page_end
);
1035 ordered
= btrfs_lookup_ordered_extent(inode
,
1037 unlock_extent(tree
, page_start
, page_end
);
1042 btrfs_start_ordered_extent(inode
, ordered
, 1);
1043 btrfs_put_ordered_extent(ordered
);
1046 * we unlocked the page above, so we need check if
1047 * it was released or not.
1049 if (page
->mapping
!= inode
->i_mapping
) {
1051 page_cache_release(page
);
1056 if (!PageUptodate(page
)) {
1057 btrfs_readpage(NULL
, page
);
1059 if (!PageUptodate(page
)) {
1061 page_cache_release(page
);
1067 if (page
->mapping
!= inode
->i_mapping
) {
1069 page_cache_release(page
);
1079 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1083 * so now we have a nice long stream of locked
1084 * and up to date pages, lets wait on them
1086 for (i
= 0; i
< i_done
; i
++)
1087 wait_on_page_writeback(pages
[i
]);
1089 page_start
= page_offset(pages
[0]);
1090 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1092 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1093 page_start
, page_end
- 1, 0, &cached_state
);
1094 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1095 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1096 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1097 &cached_state
, GFP_NOFS
);
1099 if (i_done
!= page_cnt
) {
1100 spin_lock(&BTRFS_I(inode
)->lock
);
1101 BTRFS_I(inode
)->outstanding_extents
++;
1102 spin_unlock(&BTRFS_I(inode
)->lock
);
1103 btrfs_delalloc_release_space(inode
,
1104 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1108 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1109 &cached_state
, GFP_NOFS
);
1111 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1112 page_start
, page_end
- 1, &cached_state
,
1115 for (i
= 0; i
< i_done
; i
++) {
1116 clear_page_dirty_for_io(pages
[i
]);
1117 ClearPageChecked(pages
[i
]);
1118 set_page_extent_mapped(pages
[i
]);
1119 set_page_dirty(pages
[i
]);
1120 unlock_page(pages
[i
]);
1121 page_cache_release(pages
[i
]);
1125 for (i
= 0; i
< i_done
; i
++) {
1126 unlock_page(pages
[i
]);
1127 page_cache_release(pages
[i
]);
1129 btrfs_delalloc_release_space(inode
, page_cnt
<< PAGE_CACHE_SHIFT
);
1134 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1135 struct btrfs_ioctl_defrag_range_args
*range
,
1136 u64 newer_than
, unsigned long max_to_defrag
)
1138 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1139 struct file_ra_state
*ra
= NULL
;
1140 unsigned long last_index
;
1141 u64 isize
= i_size_read(inode
);
1145 u64 newer_off
= range
->start
;
1147 unsigned long ra_index
= 0;
1149 int defrag_count
= 0;
1150 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1151 int extent_thresh
= range
->extent_thresh
;
1152 int max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
1153 int cluster
= max_cluster
;
1154 u64 new_align
= ~((u64
)128 * 1024 - 1);
1155 struct page
**pages
= NULL
;
1160 if (range
->start
>= isize
)
1163 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1164 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1166 if (range
->compress_type
)
1167 compress_type
= range
->compress_type
;
1170 if (extent_thresh
== 0)
1171 extent_thresh
= 256 * 1024;
1174 * if we were not given a file, allocate a readahead
1178 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1181 file_ra_state_init(ra
, inode
->i_mapping
);
1186 pages
= kmalloc(sizeof(struct page
*) * max_cluster
,
1193 /* find the last page to defrag */
1194 if (range
->start
+ range
->len
> range
->start
) {
1195 last_index
= min_t(u64
, isize
- 1,
1196 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1198 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1202 ret
= find_new_extents(root
, inode
, newer_than
,
1203 &newer_off
, 64 * 1024);
1205 range
->start
= newer_off
;
1207 * we always align our defrag to help keep
1208 * the extents in the file evenly spaced
1210 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1214 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1217 max_to_defrag
= last_index
+ 1;
1220 * make writeback starts from i, so the defrag range can be
1221 * written sequentially.
1223 if (i
< inode
->i_mapping
->writeback_index
)
1224 inode
->i_mapping
->writeback_index
= i
;
1226 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1227 (i
< (i_size_read(inode
) + PAGE_CACHE_SIZE
- 1) >>
1228 PAGE_CACHE_SHIFT
)) {
1230 * make sure we stop running if someone unmounts
1233 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1236 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1237 printk(KERN_DEBUG
"btrfs: defrag_file cancelled\n");
1242 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1243 extent_thresh
, &last_len
, &skip
,
1244 &defrag_end
, range
->flags
&
1245 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1248 * the should_defrag function tells us how much to skip
1249 * bump our counter by the suggested amount
1251 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
1252 i
= max(i
+ 1, next
);
1257 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1258 PAGE_CACHE_SHIFT
) - i
;
1259 cluster
= min(cluster
, max_cluster
);
1261 cluster
= max_cluster
;
1264 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1265 BTRFS_I(inode
)->force_compress
= compress_type
;
1267 if (i
+ cluster
> ra_index
) {
1268 ra_index
= max(i
, ra_index
);
1269 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1271 ra_index
+= max_cluster
;
1274 mutex_lock(&inode
->i_mutex
);
1275 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1277 mutex_unlock(&inode
->i_mutex
);
1281 defrag_count
+= ret
;
1282 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1283 mutex_unlock(&inode
->i_mutex
);
1286 if (newer_off
== (u64
)-1)
1292 newer_off
= max(newer_off
+ 1,
1293 (u64
)i
<< PAGE_CACHE_SHIFT
);
1295 ret
= find_new_extents(root
, inode
,
1296 newer_than
, &newer_off
,
1299 range
->start
= newer_off
;
1300 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1307 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1315 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
))
1316 filemap_flush(inode
->i_mapping
);
1318 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1319 /* the filemap_flush will queue IO into the worker threads, but
1320 * we have to make sure the IO is actually started and that
1321 * ordered extents get created before we return
1323 atomic_inc(&root
->fs_info
->async_submit_draining
);
1324 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1325 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1326 wait_event(root
->fs_info
->async_submit_wait
,
1327 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1328 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1330 atomic_dec(&root
->fs_info
->async_submit_draining
);
1332 mutex_lock(&inode
->i_mutex
);
1333 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1334 mutex_unlock(&inode
->i_mutex
);
1337 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1338 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1350 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1356 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1357 struct btrfs_ioctl_vol_args
*vol_args
;
1358 struct btrfs_trans_handle
*trans
;
1359 struct btrfs_device
*device
= NULL
;
1361 char *devstr
= NULL
;
1365 if (!capable(CAP_SYS_ADMIN
))
1368 ret
= mnt_want_write_file(file
);
1372 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1374 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1375 mnt_drop_write_file(file
);
1379 mutex_lock(&root
->fs_info
->volume_mutex
);
1380 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1381 if (IS_ERR(vol_args
)) {
1382 ret
= PTR_ERR(vol_args
);
1386 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1388 sizestr
= vol_args
->name
;
1389 devstr
= strchr(sizestr
, ':');
1392 sizestr
= devstr
+ 1;
1394 devstr
= vol_args
->name
;
1395 devid
= simple_strtoull(devstr
, &end
, 10);
1400 printk(KERN_INFO
"btrfs: resizing devid %llu\n",
1401 (unsigned long long)devid
);
1404 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1406 printk(KERN_INFO
"btrfs: resizer unable to find device %llu\n",
1407 (unsigned long long)devid
);
1412 if (!device
->writeable
) {
1413 printk(KERN_INFO
"btrfs: resizer unable to apply on "
1414 "readonly device %llu\n",
1415 (unsigned long long)devid
);
1420 if (!strcmp(sizestr
, "max"))
1421 new_size
= device
->bdev
->bd_inode
->i_size
;
1423 if (sizestr
[0] == '-') {
1426 } else if (sizestr
[0] == '+') {
1430 new_size
= memparse(sizestr
, NULL
);
1431 if (new_size
== 0) {
1437 if (device
->is_tgtdev_for_dev_replace
) {
1442 old_size
= device
->total_bytes
;
1445 if (new_size
> old_size
) {
1449 new_size
= old_size
- new_size
;
1450 } else if (mod
> 0) {
1451 new_size
= old_size
+ new_size
;
1454 if (new_size
< 256 * 1024 * 1024) {
1458 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1463 do_div(new_size
, root
->sectorsize
);
1464 new_size
*= root
->sectorsize
;
1466 printk_in_rcu(KERN_INFO
"btrfs: new size for %s is %llu\n",
1467 rcu_str_deref(device
->name
),
1468 (unsigned long long)new_size
);
1470 if (new_size
> old_size
) {
1471 trans
= btrfs_start_transaction(root
, 0);
1472 if (IS_ERR(trans
)) {
1473 ret
= PTR_ERR(trans
);
1476 ret
= btrfs_grow_device(trans
, device
, new_size
);
1477 btrfs_commit_transaction(trans
, root
);
1478 } else if (new_size
< old_size
) {
1479 ret
= btrfs_shrink_device(device
, new_size
);
1480 } /* equal, nothing need to do */
1485 mutex_unlock(&root
->fs_info
->volume_mutex
);
1486 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1487 mnt_drop_write_file(file
);
1491 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1492 char *name
, unsigned long fd
, int subvol
,
1493 u64
*transid
, bool readonly
,
1494 struct btrfs_qgroup_inherit
*inherit
)
1499 if (!S_ISDIR(file_inode(file
)->i_mode
))
1502 ret
= mnt_want_write_file(file
);
1506 namelen
= strlen(name
);
1507 if (strchr(name
, '/')) {
1509 goto out_drop_write
;
1512 if (name
[0] == '.' &&
1513 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1515 goto out_drop_write
;
1519 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1520 NULL
, transid
, readonly
, inherit
);
1522 struct fd src
= fdget(fd
);
1523 struct inode
*src_inode
;
1526 goto out_drop_write
;
1529 src_inode
= file_inode(src
.file
);
1530 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1531 printk(KERN_INFO
"btrfs: Snapshot src from "
1534 } else if (!inode_owner_or_capable(src_inode
)) {
1536 * Subvolume creation is not restricted, but snapshots
1537 * are limited to own subvolumes only
1541 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1542 BTRFS_I(src_inode
)->root
,
1543 transid
, readonly
, inherit
);
1548 mnt_drop_write_file(file
);
1553 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1554 void __user
*arg
, int subvol
)
1556 struct btrfs_ioctl_vol_args
*vol_args
;
1559 if (!S_ISDIR(file_inode(file
)->i_mode
))
1562 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1563 if (IS_ERR(vol_args
))
1564 return PTR_ERR(vol_args
);
1565 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1567 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1568 vol_args
->fd
, subvol
,
1575 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1576 void __user
*arg
, int subvol
)
1578 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1582 bool readonly
= false;
1583 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1585 if (!S_ISDIR(file_inode(file
)->i_mode
))
1588 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1589 if (IS_ERR(vol_args
))
1590 return PTR_ERR(vol_args
);
1591 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1593 if (vol_args
->flags
&
1594 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1595 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1600 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1602 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1604 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1605 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1609 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1610 if (IS_ERR(inherit
)) {
1611 ret
= PTR_ERR(inherit
);
1616 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1617 vol_args
->fd
, subvol
, ptr
,
1620 if (ret
== 0 && ptr
&&
1622 offsetof(struct btrfs_ioctl_vol_args_v2
,
1623 transid
), ptr
, sizeof(*ptr
)))
1631 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1634 struct inode
*inode
= file_inode(file
);
1635 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1639 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1642 down_read(&root
->fs_info
->subvol_sem
);
1643 if (btrfs_root_readonly(root
))
1644 flags
|= BTRFS_SUBVOL_RDONLY
;
1645 up_read(&root
->fs_info
->subvol_sem
);
1647 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1653 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1656 struct inode
*inode
= file_inode(file
);
1657 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1658 struct btrfs_trans_handle
*trans
;
1663 ret
= mnt_want_write_file(file
);
1667 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1669 goto out_drop_write
;
1672 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1674 goto out_drop_write
;
1677 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1679 goto out_drop_write
;
1682 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1684 goto out_drop_write
;
1687 if (!inode_owner_or_capable(inode
)) {
1689 goto out_drop_write
;
1692 down_write(&root
->fs_info
->subvol_sem
);
1695 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1698 root_flags
= btrfs_root_flags(&root
->root_item
);
1699 if (flags
& BTRFS_SUBVOL_RDONLY
)
1700 btrfs_set_root_flags(&root
->root_item
,
1701 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1703 btrfs_set_root_flags(&root
->root_item
,
1704 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1706 trans
= btrfs_start_transaction(root
, 1);
1707 if (IS_ERR(trans
)) {
1708 ret
= PTR_ERR(trans
);
1712 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1713 &root
->root_key
, &root
->root_item
);
1715 btrfs_commit_transaction(trans
, root
);
1718 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1720 up_write(&root
->fs_info
->subvol_sem
);
1722 mnt_drop_write_file(file
);
1728 * helper to check if the subvolume references other subvolumes
1730 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1732 struct btrfs_path
*path
;
1733 struct btrfs_key key
;
1736 path
= btrfs_alloc_path();
1740 key
.objectid
= root
->root_key
.objectid
;
1741 key
.type
= BTRFS_ROOT_REF_KEY
;
1742 key
.offset
= (u64
)-1;
1744 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1751 if (path
->slots
[0] > 0) {
1753 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1754 if (key
.objectid
== root
->root_key
.objectid
&&
1755 key
.type
== BTRFS_ROOT_REF_KEY
)
1759 btrfs_free_path(path
);
1763 static noinline
int key_in_sk(struct btrfs_key
*key
,
1764 struct btrfs_ioctl_search_key
*sk
)
1766 struct btrfs_key test
;
1769 test
.objectid
= sk
->min_objectid
;
1770 test
.type
= sk
->min_type
;
1771 test
.offset
= sk
->min_offset
;
1773 ret
= btrfs_comp_cpu_keys(key
, &test
);
1777 test
.objectid
= sk
->max_objectid
;
1778 test
.type
= sk
->max_type
;
1779 test
.offset
= sk
->max_offset
;
1781 ret
= btrfs_comp_cpu_keys(key
, &test
);
1787 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1788 struct btrfs_path
*path
,
1789 struct btrfs_key
*key
,
1790 struct btrfs_ioctl_search_key
*sk
,
1792 unsigned long *sk_offset
,
1796 struct extent_buffer
*leaf
;
1797 struct btrfs_ioctl_search_header sh
;
1798 unsigned long item_off
;
1799 unsigned long item_len
;
1805 leaf
= path
->nodes
[0];
1806 slot
= path
->slots
[0];
1807 nritems
= btrfs_header_nritems(leaf
);
1809 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1813 found_transid
= btrfs_header_generation(leaf
);
1815 for (i
= slot
; i
< nritems
; i
++) {
1816 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1817 item_len
= btrfs_item_size_nr(leaf
, i
);
1819 btrfs_item_key_to_cpu(leaf
, key
, i
);
1820 if (!key_in_sk(key
, sk
))
1823 if (sizeof(sh
) + item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1826 if (sizeof(sh
) + item_len
+ *sk_offset
>
1827 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1832 sh
.objectid
= key
->objectid
;
1833 sh
.offset
= key
->offset
;
1834 sh
.type
= key
->type
;
1836 sh
.transid
= found_transid
;
1838 /* copy search result header */
1839 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1840 *sk_offset
+= sizeof(sh
);
1843 char *p
= buf
+ *sk_offset
;
1845 read_extent_buffer(leaf
, p
,
1846 item_off
, item_len
);
1847 *sk_offset
+= item_len
;
1851 if (*num_found
>= sk
->nr_items
)
1856 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1858 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1861 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
1871 static noinline
int search_ioctl(struct inode
*inode
,
1872 struct btrfs_ioctl_search_args
*args
)
1874 struct btrfs_root
*root
;
1875 struct btrfs_key key
;
1876 struct btrfs_key max_key
;
1877 struct btrfs_path
*path
;
1878 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
1879 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
1882 unsigned long sk_offset
= 0;
1884 path
= btrfs_alloc_path();
1888 if (sk
->tree_id
== 0) {
1889 /* search the root of the inode that was passed */
1890 root
= BTRFS_I(inode
)->root
;
1892 key
.objectid
= sk
->tree_id
;
1893 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1894 key
.offset
= (u64
)-1;
1895 root
= btrfs_read_fs_root_no_name(info
, &key
);
1897 printk(KERN_ERR
"could not find root %llu\n",
1899 btrfs_free_path(path
);
1904 key
.objectid
= sk
->min_objectid
;
1905 key
.type
= sk
->min_type
;
1906 key
.offset
= sk
->min_offset
;
1908 max_key
.objectid
= sk
->max_objectid
;
1909 max_key
.type
= sk
->max_type
;
1910 max_key
.offset
= sk
->max_offset
;
1912 path
->keep_locks
= 1;
1915 ret
= btrfs_search_forward(root
, &key
, &max_key
, path
,
1922 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
1923 &sk_offset
, &num_found
);
1924 btrfs_release_path(path
);
1925 if (ret
|| num_found
>= sk
->nr_items
)
1931 sk
->nr_items
= num_found
;
1932 btrfs_free_path(path
);
1936 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
1939 struct btrfs_ioctl_search_args
*args
;
1940 struct inode
*inode
;
1943 if (!capable(CAP_SYS_ADMIN
))
1946 args
= memdup_user(argp
, sizeof(*args
));
1948 return PTR_ERR(args
);
1950 inode
= file_inode(file
);
1951 ret
= search_ioctl(inode
, args
);
1952 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1959 * Search INODE_REFs to identify path name of 'dirid' directory
1960 * in a 'tree_id' tree. and sets path name to 'name'.
1962 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1963 u64 tree_id
, u64 dirid
, char *name
)
1965 struct btrfs_root
*root
;
1966 struct btrfs_key key
;
1972 struct btrfs_inode_ref
*iref
;
1973 struct extent_buffer
*l
;
1974 struct btrfs_path
*path
;
1976 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1981 path
= btrfs_alloc_path();
1985 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
1987 key
.objectid
= tree_id
;
1988 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1989 key
.offset
= (u64
)-1;
1990 root
= btrfs_read_fs_root_no_name(info
, &key
);
1992 printk(KERN_ERR
"could not find root %llu\n", tree_id
);
1997 key
.objectid
= dirid
;
1998 key
.type
= BTRFS_INODE_REF_KEY
;
1999 key
.offset
= (u64
)-1;
2002 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
2007 slot
= path
->slots
[0];
2008 if (ret
> 0 && slot
> 0)
2010 btrfs_item_key_to_cpu(l
, &key
, slot
);
2012 if (ret
> 0 && (key
.objectid
!= dirid
||
2013 key
.type
!= BTRFS_INODE_REF_KEY
)) {
2018 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2019 len
= btrfs_inode_ref_name_len(l
, iref
);
2021 total_len
+= len
+ 1;
2026 read_extent_buffer(l
, ptr
,(unsigned long)(iref
+ 1), len
);
2028 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2031 btrfs_release_path(path
);
2032 key
.objectid
= key
.offset
;
2033 key
.offset
= (u64
)-1;
2034 dirid
= key
.objectid
;
2038 memmove(name
, ptr
, total_len
);
2039 name
[total_len
]='\0';
2042 btrfs_free_path(path
);
2046 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2049 struct btrfs_ioctl_ino_lookup_args
*args
;
2050 struct inode
*inode
;
2053 if (!capable(CAP_SYS_ADMIN
))
2056 args
= memdup_user(argp
, sizeof(*args
));
2058 return PTR_ERR(args
);
2060 inode
= file_inode(file
);
2062 if (args
->treeid
== 0)
2063 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2065 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2066 args
->treeid
, args
->objectid
,
2069 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2076 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2079 struct dentry
*parent
= fdentry(file
);
2080 struct dentry
*dentry
;
2081 struct inode
*dir
= parent
->d_inode
;
2082 struct inode
*inode
;
2083 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2084 struct btrfs_root
*dest
= NULL
;
2085 struct btrfs_ioctl_vol_args
*vol_args
;
2086 struct btrfs_trans_handle
*trans
;
2087 struct btrfs_block_rsv block_rsv
;
2088 u64 qgroup_reserved
;
2093 if (!S_ISDIR(dir
->i_mode
))
2096 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2097 if (IS_ERR(vol_args
))
2098 return PTR_ERR(vol_args
);
2100 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2101 namelen
= strlen(vol_args
->name
);
2102 if (strchr(vol_args
->name
, '/') ||
2103 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2108 err
= mnt_want_write_file(file
);
2112 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2114 goto out_drop_write
;
2115 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2116 if (IS_ERR(dentry
)) {
2117 err
= PTR_ERR(dentry
);
2118 goto out_unlock_dir
;
2121 if (!dentry
->d_inode
) {
2126 inode
= dentry
->d_inode
;
2127 dest
= BTRFS_I(inode
)->root
;
2128 if (!capable(CAP_SYS_ADMIN
)){
2130 * Regular user. Only allow this with a special mount
2131 * option, when the user has write+exec access to the
2132 * subvol root, and when rmdir(2) would have been
2135 * Note that this is _not_ check that the subvol is
2136 * empty or doesn't contain data that we wouldn't
2137 * otherwise be able to delete.
2139 * Users who want to delete empty subvols should try
2143 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2147 * Do not allow deletion if the parent dir is the same
2148 * as the dir to be deleted. That means the ioctl
2149 * must be called on the dentry referencing the root
2150 * of the subvol, not a random directory contained
2157 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2162 /* check if subvolume may be deleted by a user */
2163 err
= btrfs_may_delete(dir
, dentry
, 1);
2167 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2172 mutex_lock(&inode
->i_mutex
);
2173 err
= d_invalidate(dentry
);
2177 down_write(&root
->fs_info
->subvol_sem
);
2179 err
= may_destroy_subvol(dest
);
2183 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2185 * One for dir inode, two for dir entries, two for root
2188 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2189 5, &qgroup_reserved
);
2193 trans
= btrfs_start_transaction(root
, 0);
2194 if (IS_ERR(trans
)) {
2195 err
= PTR_ERR(trans
);
2198 trans
->block_rsv
= &block_rsv
;
2199 trans
->bytes_reserved
= block_rsv
.size
;
2201 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2202 dest
->root_key
.objectid
,
2203 dentry
->d_name
.name
,
2204 dentry
->d_name
.len
);
2207 btrfs_abort_transaction(trans
, root
, ret
);
2211 btrfs_record_root_in_trans(trans
, dest
);
2213 memset(&dest
->root_item
.drop_progress
, 0,
2214 sizeof(dest
->root_item
.drop_progress
));
2215 dest
->root_item
.drop_level
= 0;
2216 btrfs_set_root_refs(&dest
->root_item
, 0);
2218 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
2219 ret
= btrfs_insert_orphan_item(trans
,
2220 root
->fs_info
->tree_root
,
2221 dest
->root_key
.objectid
);
2223 btrfs_abort_transaction(trans
, root
, ret
);
2229 trans
->block_rsv
= NULL
;
2230 trans
->bytes_reserved
= 0;
2231 ret
= btrfs_end_transaction(trans
, root
);
2234 inode
->i_flags
|= S_DEAD
;
2236 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2238 up_write(&root
->fs_info
->subvol_sem
);
2240 mutex_unlock(&inode
->i_mutex
);
2242 shrink_dcache_sb(root
->fs_info
->sb
);
2243 btrfs_invalidate_inodes(dest
);
2247 if (dest
->cache_inode
) {
2248 iput(dest
->cache_inode
);
2249 dest
->cache_inode
= NULL
;
2255 mutex_unlock(&dir
->i_mutex
);
2257 mnt_drop_write_file(file
);
2263 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2265 struct inode
*inode
= file_inode(file
);
2266 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2267 struct btrfs_ioctl_defrag_range_args
*range
;
2270 ret
= mnt_want_write_file(file
);
2274 if (btrfs_root_readonly(root
)) {
2279 switch (inode
->i_mode
& S_IFMT
) {
2281 if (!capable(CAP_SYS_ADMIN
)) {
2285 ret
= btrfs_defrag_root(root
);
2288 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2291 if (!(file
->f_mode
& FMODE_WRITE
)) {
2296 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2303 if (copy_from_user(range
, argp
,
2309 /* compression requires us to start the IO */
2310 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2311 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2312 range
->extent_thresh
= (u32
)-1;
2315 /* the rest are all set to zero by kzalloc */
2316 range
->len
= (u64
)-1;
2318 ret
= btrfs_defrag_file(file_inode(file
), file
,
2328 mnt_drop_write_file(file
);
2332 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2334 struct btrfs_ioctl_vol_args
*vol_args
;
2337 if (!capable(CAP_SYS_ADMIN
))
2340 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2342 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2346 mutex_lock(&root
->fs_info
->volume_mutex
);
2347 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2348 if (IS_ERR(vol_args
)) {
2349 ret
= PTR_ERR(vol_args
);
2353 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2354 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2358 mutex_unlock(&root
->fs_info
->volume_mutex
);
2359 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2363 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2365 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2366 struct btrfs_ioctl_vol_args
*vol_args
;
2369 if (!capable(CAP_SYS_ADMIN
))
2372 ret
= mnt_want_write_file(file
);
2376 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2378 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2379 mnt_drop_write_file(file
);
2383 mutex_lock(&root
->fs_info
->volume_mutex
);
2384 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2385 if (IS_ERR(vol_args
)) {
2386 ret
= PTR_ERR(vol_args
);
2390 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2391 ret
= btrfs_rm_device(root
, vol_args
->name
);
2395 mutex_unlock(&root
->fs_info
->volume_mutex
);
2396 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2397 mnt_drop_write_file(file
);
2401 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2403 struct btrfs_ioctl_fs_info_args
*fi_args
;
2404 struct btrfs_device
*device
;
2405 struct btrfs_device
*next
;
2406 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2409 if (!capable(CAP_SYS_ADMIN
))
2412 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2416 fi_args
->num_devices
= fs_devices
->num_devices
;
2417 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2419 mutex_lock(&fs_devices
->device_list_mutex
);
2420 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2421 if (device
->devid
> fi_args
->max_id
)
2422 fi_args
->max_id
= device
->devid
;
2424 mutex_unlock(&fs_devices
->device_list_mutex
);
2426 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2433 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2435 struct btrfs_ioctl_dev_info_args
*di_args
;
2436 struct btrfs_device
*dev
;
2437 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2439 char *s_uuid
= NULL
;
2440 char empty_uuid
[BTRFS_UUID_SIZE
] = {0};
2442 if (!capable(CAP_SYS_ADMIN
))
2445 di_args
= memdup_user(arg
, sizeof(*di_args
));
2446 if (IS_ERR(di_args
))
2447 return PTR_ERR(di_args
);
2449 if (memcmp(empty_uuid
, di_args
->uuid
, BTRFS_UUID_SIZE
) != 0)
2450 s_uuid
= di_args
->uuid
;
2452 mutex_lock(&fs_devices
->device_list_mutex
);
2453 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2460 di_args
->devid
= dev
->devid
;
2461 di_args
->bytes_used
= dev
->bytes_used
;
2462 di_args
->total_bytes
= dev
->total_bytes
;
2463 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2465 struct rcu_string
*name
;
2468 name
= rcu_dereference(dev
->name
);
2469 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2471 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2473 di_args
->path
[0] = '\0';
2477 mutex_unlock(&fs_devices
->device_list_mutex
);
2478 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2485 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
2486 u64 off
, u64 olen
, u64 destoff
)
2488 struct inode
*inode
= file_inode(file
);
2489 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2492 struct btrfs_trans_handle
*trans
;
2493 struct btrfs_path
*path
;
2494 struct extent_buffer
*leaf
;
2496 struct btrfs_key key
;
2501 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
2505 * - split compressed inline extents. annoying: we need to
2506 * decompress into destination's address_space (the file offset
2507 * may change, so source mapping won't do), then recompress (or
2508 * otherwise reinsert) a subrange.
2509 * - allow ranges within the same file to be cloned (provided
2510 * they don't overlap)?
2513 /* the destination must be opened for writing */
2514 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
2517 if (btrfs_root_readonly(root
))
2520 ret
= mnt_want_write_file(file
);
2524 src_file
= fdget(srcfd
);
2525 if (!src_file
.file
) {
2527 goto out_drop_write
;
2531 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
2534 src
= file_inode(src_file
.file
);
2540 /* the src must be open for reading */
2541 if (!(src_file
.file
->f_mode
& FMODE_READ
))
2544 /* don't make the dst file partly checksummed */
2545 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2546 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
2550 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
2554 if (src
->i_sb
!= inode
->i_sb
)
2558 buf
= vmalloc(btrfs_level_size(root
, 0));
2562 path
= btrfs_alloc_path();
2570 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
2571 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
2573 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
2574 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
2577 /* determine range to clone */
2579 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
2582 olen
= len
= src
->i_size
- off
;
2583 /* if we extend to eof, continue to block boundary */
2584 if (off
+ len
== src
->i_size
)
2585 len
= ALIGN(src
->i_size
, bs
) - off
;
2592 /* verify the end result is block aligned */
2593 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
2594 !IS_ALIGNED(destoff
, bs
))
2597 if (destoff
> inode
->i_size
) {
2598 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
2603 /* truncate page cache pages from target inode range */
2604 truncate_inode_pages_range(&inode
->i_data
, destoff
,
2605 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
2607 /* do any pending delalloc/csum calc on src, one way or
2608 another, and lock file content */
2610 struct btrfs_ordered_extent
*ordered
;
2611 lock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2612 ordered
= btrfs_lookup_first_ordered_extent(src
, off
+ len
- 1);
2614 !test_range_bit(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1,
2615 EXTENT_DELALLOC
, 0, NULL
))
2617 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2619 btrfs_put_ordered_extent(ordered
);
2620 btrfs_wait_ordered_range(src
, off
, len
);
2624 key
.objectid
= btrfs_ino(src
);
2625 key
.type
= BTRFS_EXTENT_DATA_KEY
;
2630 * note the key will change type as we walk through the
2633 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
2638 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2639 if (path
->slots
[0] >= nritems
) {
2640 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
2645 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2647 leaf
= path
->nodes
[0];
2648 slot
= path
->slots
[0];
2650 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2651 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
2652 key
.objectid
!= btrfs_ino(src
))
2655 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
2656 struct btrfs_file_extent_item
*extent
;
2659 struct btrfs_key new_key
;
2660 u64 disko
= 0, diskl
= 0;
2661 u64 datao
= 0, datal
= 0;
2665 size
= btrfs_item_size_nr(leaf
, slot
);
2666 read_extent_buffer(leaf
, buf
,
2667 btrfs_item_ptr_offset(leaf
, slot
),
2670 extent
= btrfs_item_ptr(leaf
, slot
,
2671 struct btrfs_file_extent_item
);
2672 comp
= btrfs_file_extent_compression(leaf
, extent
);
2673 type
= btrfs_file_extent_type(leaf
, extent
);
2674 if (type
== BTRFS_FILE_EXTENT_REG
||
2675 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2676 disko
= btrfs_file_extent_disk_bytenr(leaf
,
2678 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
2680 datao
= btrfs_file_extent_offset(leaf
, extent
);
2681 datal
= btrfs_file_extent_num_bytes(leaf
,
2683 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2684 /* take upper bound, may be compressed */
2685 datal
= btrfs_file_extent_ram_bytes(leaf
,
2688 btrfs_release_path(path
);
2690 if (key
.offset
+ datal
<= off
||
2691 key
.offset
>= off
+ len
- 1)
2694 memcpy(&new_key
, &key
, sizeof(new_key
));
2695 new_key
.objectid
= btrfs_ino(inode
);
2696 if (off
<= key
.offset
)
2697 new_key
.offset
= key
.offset
+ destoff
- off
;
2699 new_key
.offset
= destoff
;
2702 * 1 - adjusting old extent (we may have to split it)
2703 * 1 - add new extent
2706 trans
= btrfs_start_transaction(root
, 3);
2707 if (IS_ERR(trans
)) {
2708 ret
= PTR_ERR(trans
);
2712 if (type
== BTRFS_FILE_EXTENT_REG
||
2713 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2715 * a | --- range to clone ---| b
2716 * | ------------- extent ------------- |
2719 /* substract range b */
2720 if (key
.offset
+ datal
> off
+ len
)
2721 datal
= off
+ len
- key
.offset
;
2723 /* substract range a */
2724 if (off
> key
.offset
) {
2725 datao
+= off
- key
.offset
;
2726 datal
-= off
- key
.offset
;
2729 ret
= btrfs_drop_extents(trans
, root
, inode
,
2731 new_key
.offset
+ datal
,
2734 btrfs_abort_transaction(trans
, root
,
2736 btrfs_end_transaction(trans
, root
);
2740 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2743 btrfs_abort_transaction(trans
, root
,
2745 btrfs_end_transaction(trans
, root
);
2749 leaf
= path
->nodes
[0];
2750 slot
= path
->slots
[0];
2751 write_extent_buffer(leaf
, buf
,
2752 btrfs_item_ptr_offset(leaf
, slot
),
2755 extent
= btrfs_item_ptr(leaf
, slot
,
2756 struct btrfs_file_extent_item
);
2758 /* disko == 0 means it's a hole */
2762 btrfs_set_file_extent_offset(leaf
, extent
,
2764 btrfs_set_file_extent_num_bytes(leaf
, extent
,
2767 inode_add_bytes(inode
, datal
);
2768 ret
= btrfs_inc_extent_ref(trans
, root
,
2770 root
->root_key
.objectid
,
2772 new_key
.offset
- datao
,
2775 btrfs_abort_transaction(trans
,
2778 btrfs_end_transaction(trans
,
2784 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2787 if (off
> key
.offset
) {
2788 skip
= off
- key
.offset
;
2789 new_key
.offset
+= skip
;
2792 if (key
.offset
+ datal
> off
+ len
)
2793 trim
= key
.offset
+ datal
- (off
+ len
);
2795 if (comp
&& (skip
|| trim
)) {
2797 btrfs_end_transaction(trans
, root
);
2800 size
-= skip
+ trim
;
2801 datal
-= skip
+ trim
;
2803 ret
= btrfs_drop_extents(trans
, root
, inode
,
2805 new_key
.offset
+ datal
,
2808 btrfs_abort_transaction(trans
, root
,
2810 btrfs_end_transaction(trans
, root
);
2814 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2817 btrfs_abort_transaction(trans
, root
,
2819 btrfs_end_transaction(trans
, root
);
2825 btrfs_file_extent_calc_inline_size(0);
2826 memmove(buf
+start
, buf
+start
+skip
,
2830 leaf
= path
->nodes
[0];
2831 slot
= path
->slots
[0];
2832 write_extent_buffer(leaf
, buf
,
2833 btrfs_item_ptr_offset(leaf
, slot
),
2835 inode_add_bytes(inode
, datal
);
2838 btrfs_mark_buffer_dirty(leaf
);
2839 btrfs_release_path(path
);
2841 inode_inc_iversion(inode
);
2842 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2845 * we round up to the block size at eof when
2846 * determining which extents to clone above,
2847 * but shouldn't round up the file size
2849 endoff
= new_key
.offset
+ datal
;
2850 if (endoff
> destoff
+olen
)
2851 endoff
= destoff
+olen
;
2852 if (endoff
> inode
->i_size
)
2853 btrfs_i_size_write(inode
, endoff
);
2855 ret
= btrfs_update_inode(trans
, root
, inode
);
2857 btrfs_abort_transaction(trans
, root
, ret
);
2858 btrfs_end_transaction(trans
, root
);
2861 ret
= btrfs_end_transaction(trans
, root
);
2864 btrfs_release_path(path
);
2869 btrfs_release_path(path
);
2870 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2872 mutex_unlock(&src
->i_mutex
);
2873 mutex_unlock(&inode
->i_mutex
);
2875 btrfs_free_path(path
);
2879 mnt_drop_write_file(file
);
2883 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
2885 struct btrfs_ioctl_clone_range_args args
;
2887 if (copy_from_user(&args
, argp
, sizeof(args
)))
2889 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
2890 args
.src_length
, args
.dest_offset
);
2894 * there are many ways the trans_start and trans_end ioctls can lead
2895 * to deadlocks. They should only be used by applications that
2896 * basically own the machine, and have a very in depth understanding
2897 * of all the possible deadlocks and enospc problems.
2899 static long btrfs_ioctl_trans_start(struct file
*file
)
2901 struct inode
*inode
= file_inode(file
);
2902 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2903 struct btrfs_trans_handle
*trans
;
2907 if (!capable(CAP_SYS_ADMIN
))
2911 if (file
->private_data
)
2915 if (btrfs_root_readonly(root
))
2918 ret
= mnt_want_write_file(file
);
2922 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
2925 trans
= btrfs_start_ioctl_transaction(root
);
2929 file
->private_data
= trans
;
2933 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
2934 mnt_drop_write_file(file
);
2939 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2941 struct inode
*inode
= file_inode(file
);
2942 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2943 struct btrfs_root
*new_root
;
2944 struct btrfs_dir_item
*di
;
2945 struct btrfs_trans_handle
*trans
;
2946 struct btrfs_path
*path
;
2947 struct btrfs_key location
;
2948 struct btrfs_disk_key disk_key
;
2953 if (!capable(CAP_SYS_ADMIN
))
2956 ret
= mnt_want_write_file(file
);
2960 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
2966 objectid
= root
->root_key
.objectid
;
2968 location
.objectid
= objectid
;
2969 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2970 location
.offset
= (u64
)-1;
2972 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
2973 if (IS_ERR(new_root
)) {
2974 ret
= PTR_ERR(new_root
);
2977 if (!is_fstree(new_root
->objectid
)) {
2982 if (btrfs_root_refs(&new_root
->root_item
) == 0) {
2987 path
= btrfs_alloc_path();
2992 path
->leave_spinning
= 1;
2994 trans
= btrfs_start_transaction(root
, 1);
2995 if (IS_ERR(trans
)) {
2996 btrfs_free_path(path
);
2997 ret
= PTR_ERR(trans
);
3001 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
3002 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
3003 dir_id
, "default", 7, 1);
3004 if (IS_ERR_OR_NULL(di
)) {
3005 btrfs_free_path(path
);
3006 btrfs_end_transaction(trans
, root
);
3007 printk(KERN_ERR
"Umm, you don't have the default dir item, "
3008 "this isn't going to work\n");
3013 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
3014 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
3015 btrfs_mark_buffer_dirty(path
->nodes
[0]);
3016 btrfs_free_path(path
);
3018 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
3019 btrfs_end_transaction(trans
, root
);
3021 mnt_drop_write_file(file
);
3025 void btrfs_get_block_group_info(struct list_head
*groups_list
,
3026 struct btrfs_ioctl_space_info
*space
)
3028 struct btrfs_block_group_cache
*block_group
;
3030 space
->total_bytes
= 0;
3031 space
->used_bytes
= 0;
3033 list_for_each_entry(block_group
, groups_list
, list
) {
3034 space
->flags
= block_group
->flags
;
3035 space
->total_bytes
+= block_group
->key
.offset
;
3036 space
->used_bytes
+=
3037 btrfs_block_group_used(&block_group
->item
);
3041 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
3043 struct btrfs_ioctl_space_args space_args
;
3044 struct btrfs_ioctl_space_info space
;
3045 struct btrfs_ioctl_space_info
*dest
;
3046 struct btrfs_ioctl_space_info
*dest_orig
;
3047 struct btrfs_ioctl_space_info __user
*user_dest
;
3048 struct btrfs_space_info
*info
;
3049 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
3050 BTRFS_BLOCK_GROUP_SYSTEM
,
3051 BTRFS_BLOCK_GROUP_METADATA
,
3052 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
3059 if (copy_from_user(&space_args
,
3060 (struct btrfs_ioctl_space_args __user
*)arg
,
3061 sizeof(space_args
)))
3064 for (i
= 0; i
< num_types
; i
++) {
3065 struct btrfs_space_info
*tmp
;
3069 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3071 if (tmp
->flags
== types
[i
]) {
3081 down_read(&info
->groups_sem
);
3082 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3083 if (!list_empty(&info
->block_groups
[c
]))
3086 up_read(&info
->groups_sem
);
3089 /* space_slots == 0 means they are asking for a count */
3090 if (space_args
.space_slots
== 0) {
3091 space_args
.total_spaces
= slot_count
;
3095 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
3097 alloc_size
= sizeof(*dest
) * slot_count
;
3099 /* we generally have at most 6 or so space infos, one for each raid
3100 * level. So, a whole page should be more than enough for everyone
3102 if (alloc_size
> PAGE_CACHE_SIZE
)
3105 space_args
.total_spaces
= 0;
3106 dest
= kmalloc(alloc_size
, GFP_NOFS
);
3111 /* now we have a buffer to copy into */
3112 for (i
= 0; i
< num_types
; i
++) {
3113 struct btrfs_space_info
*tmp
;
3120 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3122 if (tmp
->flags
== types
[i
]) {
3131 down_read(&info
->groups_sem
);
3132 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3133 if (!list_empty(&info
->block_groups
[c
])) {
3134 btrfs_get_block_group_info(
3135 &info
->block_groups
[c
], &space
);
3136 memcpy(dest
, &space
, sizeof(space
));
3138 space_args
.total_spaces
++;
3144 up_read(&info
->groups_sem
);
3147 user_dest
= (struct btrfs_ioctl_space_info __user
*)
3148 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
3150 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
3155 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
3162 * there are many ways the trans_start and trans_end ioctls can lead
3163 * to deadlocks. They should only be used by applications that
3164 * basically own the machine, and have a very in depth understanding
3165 * of all the possible deadlocks and enospc problems.
3167 long btrfs_ioctl_trans_end(struct file
*file
)
3169 struct inode
*inode
= file_inode(file
);
3170 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3171 struct btrfs_trans_handle
*trans
;
3173 trans
= file
->private_data
;
3176 file
->private_data
= NULL
;
3178 btrfs_end_transaction(trans
, root
);
3180 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3182 mnt_drop_write_file(file
);
3186 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
3189 struct btrfs_trans_handle
*trans
;
3193 trans
= btrfs_attach_transaction_barrier(root
);
3194 if (IS_ERR(trans
)) {
3195 if (PTR_ERR(trans
) != -ENOENT
)
3196 return PTR_ERR(trans
);
3198 /* No running transaction, don't bother */
3199 transid
= root
->fs_info
->last_trans_committed
;
3202 transid
= trans
->transid
;
3203 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
3205 btrfs_end_transaction(trans
, root
);
3210 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
3215 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
3221 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
3224 transid
= 0; /* current trans */
3226 return btrfs_wait_for_commit(root
, transid
);
3229 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
3231 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3232 struct btrfs_ioctl_scrub_args
*sa
;
3235 if (!capable(CAP_SYS_ADMIN
))
3238 sa
= memdup_user(arg
, sizeof(*sa
));
3242 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
3243 ret
= mnt_want_write_file(file
);
3248 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
3249 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
3252 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3255 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
3256 mnt_drop_write_file(file
);
3262 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
3264 if (!capable(CAP_SYS_ADMIN
))
3267 return btrfs_scrub_cancel(root
->fs_info
);
3270 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
3273 struct btrfs_ioctl_scrub_args
*sa
;
3276 if (!capable(CAP_SYS_ADMIN
))
3279 sa
= memdup_user(arg
, sizeof(*sa
));
3283 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
3285 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3292 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
3295 struct btrfs_ioctl_get_dev_stats
*sa
;
3298 sa
= memdup_user(arg
, sizeof(*sa
));
3302 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
3307 ret
= btrfs_get_dev_stats(root
, sa
);
3309 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3316 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
3318 struct btrfs_ioctl_dev_replace_args
*p
;
3321 if (!capable(CAP_SYS_ADMIN
))
3324 p
= memdup_user(arg
, sizeof(*p
));
3329 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
3330 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
3334 &root
->fs_info
->mutually_exclusive_operation_running
,
3336 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3339 ret
= btrfs_dev_replace_start(root
, p
);
3341 &root
->fs_info
->mutually_exclusive_operation_running
,
3345 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
3346 btrfs_dev_replace_status(root
->fs_info
, p
);
3349 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
3350 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
3357 if (copy_to_user(arg
, p
, sizeof(*p
)))
3364 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
3370 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
3371 struct inode_fs_paths
*ipath
= NULL
;
3372 struct btrfs_path
*path
;
3374 if (!capable(CAP_DAC_READ_SEARCH
))
3377 path
= btrfs_alloc_path();
3383 ipa
= memdup_user(arg
, sizeof(*ipa
));
3390 size
= min_t(u32
, ipa
->size
, 4096);
3391 ipath
= init_ipath(size
, root
, path
);
3392 if (IS_ERR(ipath
)) {
3393 ret
= PTR_ERR(ipath
);
3398 ret
= paths_from_inode(ipa
->inum
, ipath
);
3402 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
3403 rel_ptr
= ipath
->fspath
->val
[i
] -
3404 (u64
)(unsigned long)ipath
->fspath
->val
;
3405 ipath
->fspath
->val
[i
] = rel_ptr
;
3408 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
3409 (void *)(unsigned long)ipath
->fspath
, size
);
3416 btrfs_free_path(path
);
3423 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
3425 struct btrfs_data_container
*inodes
= ctx
;
3426 const size_t c
= 3 * sizeof(u64
);
3428 if (inodes
->bytes_left
>= c
) {
3429 inodes
->bytes_left
-= c
;
3430 inodes
->val
[inodes
->elem_cnt
] = inum
;
3431 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
3432 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
3433 inodes
->elem_cnt
+= 3;
3435 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
3436 inodes
->bytes_left
= 0;
3437 inodes
->elem_missed
+= 3;
3443 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
3448 struct btrfs_ioctl_logical_ino_args
*loi
;
3449 struct btrfs_data_container
*inodes
= NULL
;
3450 struct btrfs_path
*path
= NULL
;
3452 if (!capable(CAP_SYS_ADMIN
))
3455 loi
= memdup_user(arg
, sizeof(*loi
));
3462 path
= btrfs_alloc_path();
3468 size
= min_t(u32
, loi
->size
, 64 * 1024);
3469 inodes
= init_data_container(size
);
3470 if (IS_ERR(inodes
)) {
3471 ret
= PTR_ERR(inodes
);
3476 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
3477 build_ino_list
, inodes
);
3483 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
3484 (void *)(unsigned long)inodes
, size
);
3489 btrfs_free_path(path
);
3496 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
3497 struct btrfs_ioctl_balance_args
*bargs
)
3499 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
3501 bargs
->flags
= bctl
->flags
;
3503 if (atomic_read(&fs_info
->balance_running
))
3504 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
3505 if (atomic_read(&fs_info
->balance_pause_req
))
3506 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
3507 if (atomic_read(&fs_info
->balance_cancel_req
))
3508 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
3510 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
3511 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
3512 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
3515 spin_lock(&fs_info
->balance_lock
);
3516 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3517 spin_unlock(&fs_info
->balance_lock
);
3519 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3523 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
3525 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3526 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3527 struct btrfs_ioctl_balance_args
*bargs
;
3528 struct btrfs_balance_control
*bctl
;
3529 bool need_unlock
; /* for mut. excl. ops lock */
3532 if (!capable(CAP_SYS_ADMIN
))
3535 ret
= mnt_want_write_file(file
);
3540 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
3541 mutex_lock(&fs_info
->volume_mutex
);
3542 mutex_lock(&fs_info
->balance_mutex
);
3548 * mut. excl. ops lock is locked. Three possibilites:
3549 * (1) some other op is running
3550 * (2) balance is running
3551 * (3) balance is paused -- special case (think resume)
3553 mutex_lock(&fs_info
->balance_mutex
);
3554 if (fs_info
->balance_ctl
) {
3555 /* this is either (2) or (3) */
3556 if (!atomic_read(&fs_info
->balance_running
)) {
3557 mutex_unlock(&fs_info
->balance_mutex
);
3558 if (!mutex_trylock(&fs_info
->volume_mutex
))
3560 mutex_lock(&fs_info
->balance_mutex
);
3562 if (fs_info
->balance_ctl
&&
3563 !atomic_read(&fs_info
->balance_running
)) {
3565 need_unlock
= false;
3569 mutex_unlock(&fs_info
->balance_mutex
);
3570 mutex_unlock(&fs_info
->volume_mutex
);
3574 mutex_unlock(&fs_info
->balance_mutex
);
3580 mutex_unlock(&fs_info
->balance_mutex
);
3581 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3587 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
3590 bargs
= memdup_user(arg
, sizeof(*bargs
));
3591 if (IS_ERR(bargs
)) {
3592 ret
= PTR_ERR(bargs
);
3596 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
3597 if (!fs_info
->balance_ctl
) {
3602 bctl
= fs_info
->balance_ctl
;
3603 spin_lock(&fs_info
->balance_lock
);
3604 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
3605 spin_unlock(&fs_info
->balance_lock
);
3613 if (fs_info
->balance_ctl
) {
3618 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
3624 bctl
->fs_info
= fs_info
;
3626 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
3627 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
3628 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
3630 bctl
->flags
= bargs
->flags
;
3632 /* balance everything - no filters */
3633 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
3638 * Ownership of bctl and mutually_exclusive_operation_running
3639 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3640 * or, if restriper was paused all the way until unmount, in
3641 * free_fs_info. mutually_exclusive_operation_running is
3642 * cleared in __cancel_balance.
3644 need_unlock
= false;
3646 ret
= btrfs_balance(bctl
, bargs
);
3649 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3656 mutex_unlock(&fs_info
->balance_mutex
);
3657 mutex_unlock(&fs_info
->volume_mutex
);
3659 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
3661 mnt_drop_write_file(file
);
3665 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
3667 if (!capable(CAP_SYS_ADMIN
))
3671 case BTRFS_BALANCE_CTL_PAUSE
:
3672 return btrfs_pause_balance(root
->fs_info
);
3673 case BTRFS_BALANCE_CTL_CANCEL
:
3674 return btrfs_cancel_balance(root
->fs_info
);
3680 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
3683 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3684 struct btrfs_ioctl_balance_args
*bargs
;
3687 if (!capable(CAP_SYS_ADMIN
))
3690 mutex_lock(&fs_info
->balance_mutex
);
3691 if (!fs_info
->balance_ctl
) {
3696 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
3702 update_ioctl_balance_args(fs_info
, 1, bargs
);
3704 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3709 mutex_unlock(&fs_info
->balance_mutex
);
3713 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
3715 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3716 struct btrfs_ioctl_quota_ctl_args
*sa
;
3717 struct btrfs_trans_handle
*trans
= NULL
;
3721 if (!capable(CAP_SYS_ADMIN
))
3724 ret
= mnt_want_write_file(file
);
3728 sa
= memdup_user(arg
, sizeof(*sa
));
3734 down_write(&root
->fs_info
->subvol_sem
);
3735 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
3736 if (IS_ERR(trans
)) {
3737 ret
= PTR_ERR(trans
);
3742 case BTRFS_QUOTA_CTL_ENABLE
:
3743 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
3745 case BTRFS_QUOTA_CTL_DISABLE
:
3746 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
3753 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3756 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
3761 up_write(&root
->fs_info
->subvol_sem
);
3763 mnt_drop_write_file(file
);
3767 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
3769 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3770 struct btrfs_ioctl_qgroup_assign_args
*sa
;
3771 struct btrfs_trans_handle
*trans
;
3775 if (!capable(CAP_SYS_ADMIN
))
3778 ret
= mnt_want_write_file(file
);
3782 sa
= memdup_user(arg
, sizeof(*sa
));
3788 trans
= btrfs_join_transaction(root
);
3789 if (IS_ERR(trans
)) {
3790 ret
= PTR_ERR(trans
);
3794 /* FIXME: check if the IDs really exist */
3796 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
3799 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
3803 err
= btrfs_end_transaction(trans
, root
);
3810 mnt_drop_write_file(file
);
3814 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
3816 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3817 struct btrfs_ioctl_qgroup_create_args
*sa
;
3818 struct btrfs_trans_handle
*trans
;
3822 if (!capable(CAP_SYS_ADMIN
))
3825 ret
= mnt_want_write_file(file
);
3829 sa
= memdup_user(arg
, sizeof(*sa
));
3835 if (!sa
->qgroupid
) {
3840 trans
= btrfs_join_transaction(root
);
3841 if (IS_ERR(trans
)) {
3842 ret
= PTR_ERR(trans
);
3846 /* FIXME: check if the IDs really exist */
3848 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
,
3851 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
3854 err
= btrfs_end_transaction(trans
, root
);
3861 mnt_drop_write_file(file
);
3865 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
3867 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3868 struct btrfs_ioctl_qgroup_limit_args
*sa
;
3869 struct btrfs_trans_handle
*trans
;
3874 if (!capable(CAP_SYS_ADMIN
))
3877 ret
= mnt_want_write_file(file
);
3881 sa
= memdup_user(arg
, sizeof(*sa
));
3887 trans
= btrfs_join_transaction(root
);
3888 if (IS_ERR(trans
)) {
3889 ret
= PTR_ERR(trans
);
3893 qgroupid
= sa
->qgroupid
;
3895 /* take the current subvol as qgroup */
3896 qgroupid
= root
->root_key
.objectid
;
3899 /* FIXME: check if the IDs really exist */
3900 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
3902 err
= btrfs_end_transaction(trans
, root
);
3909 mnt_drop_write_file(file
);
3913 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
3915 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3916 struct btrfs_ioctl_quota_rescan_args
*qsa
;
3919 if (!capable(CAP_SYS_ADMIN
))
3922 ret
= mnt_want_write_file(file
);
3926 qsa
= memdup_user(arg
, sizeof(*qsa
));
3937 ret
= btrfs_qgroup_rescan(root
->fs_info
);
3942 mnt_drop_write_file(file
);
3946 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
3948 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3949 struct btrfs_ioctl_quota_rescan_args
*qsa
;
3952 if (!capable(CAP_SYS_ADMIN
))
3955 qsa
= kzalloc(sizeof(*qsa
), GFP_NOFS
);
3959 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
3961 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
3964 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
3971 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
3974 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
3975 struct inode
*inode
= file_inode(file
);
3976 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3977 struct btrfs_root_item
*root_item
= &root
->root_item
;
3978 struct btrfs_trans_handle
*trans
;
3979 struct timespec ct
= CURRENT_TIME
;
3982 ret
= mnt_want_write_file(file
);
3986 down_write(&root
->fs_info
->subvol_sem
);
3988 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
3993 if (btrfs_root_readonly(root
)) {
3998 if (!inode_owner_or_capable(inode
)) {
4003 sa
= memdup_user(arg
, sizeof(*sa
));
4010 trans
= btrfs_start_transaction(root
, 1);
4011 if (IS_ERR(trans
)) {
4012 ret
= PTR_ERR(trans
);
4017 sa
->rtransid
= trans
->transid
;
4018 sa
->rtime
.sec
= ct
.tv_sec
;
4019 sa
->rtime
.nsec
= ct
.tv_nsec
;
4021 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
4022 btrfs_set_root_stransid(root_item
, sa
->stransid
);
4023 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
4024 root_item
->stime
.sec
= cpu_to_le64(sa
->stime
.sec
);
4025 root_item
->stime
.nsec
= cpu_to_le32(sa
->stime
.nsec
);
4026 root_item
->rtime
.sec
= cpu_to_le64(sa
->rtime
.sec
);
4027 root_item
->rtime
.nsec
= cpu_to_le32(sa
->rtime
.nsec
);
4029 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
4030 &root
->root_key
, &root
->root_item
);
4032 btrfs_end_transaction(trans
, root
);
4036 ret
= btrfs_commit_transaction(trans
, root
);
4041 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
4047 up_write(&root
->fs_info
->subvol_sem
);
4048 mnt_drop_write_file(file
);
4052 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
4054 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
4055 const char *label
= root
->fs_info
->super_copy
->label
;
4056 size_t len
= strnlen(label
, BTRFS_LABEL_SIZE
);
4059 if (len
== BTRFS_LABEL_SIZE
) {
4060 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4064 mutex_lock(&root
->fs_info
->volume_mutex
);
4065 ret
= copy_to_user(arg
, label
, len
);
4066 mutex_unlock(&root
->fs_info
->volume_mutex
);
4068 return ret
? -EFAULT
: 0;
4071 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
4073 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
4074 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
4075 struct btrfs_trans_handle
*trans
;
4076 char label
[BTRFS_LABEL_SIZE
];
4079 if (!capable(CAP_SYS_ADMIN
))
4082 if (copy_from_user(label
, arg
, sizeof(label
)))
4085 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
4086 pr_err("btrfs: unable to set label with more than %d bytes\n",
4087 BTRFS_LABEL_SIZE
- 1);
4091 ret
= mnt_want_write_file(file
);
4095 mutex_lock(&root
->fs_info
->volume_mutex
);
4096 trans
= btrfs_start_transaction(root
, 0);
4097 if (IS_ERR(trans
)) {
4098 ret
= PTR_ERR(trans
);
4102 strcpy(super_block
->label
, label
);
4103 ret
= btrfs_end_transaction(trans
, root
);
4106 mutex_unlock(&root
->fs_info
->volume_mutex
);
4107 mnt_drop_write_file(file
);
4111 long btrfs_ioctl(struct file
*file
, unsigned int
4112 cmd
, unsigned long arg
)
4114 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4115 void __user
*argp
= (void __user
*)arg
;
4118 case FS_IOC_GETFLAGS
:
4119 return btrfs_ioctl_getflags(file
, argp
);
4120 case FS_IOC_SETFLAGS
:
4121 return btrfs_ioctl_setflags(file
, argp
);
4122 case FS_IOC_GETVERSION
:
4123 return btrfs_ioctl_getversion(file
, argp
);
4125 return btrfs_ioctl_fitrim(file
, argp
);
4126 case BTRFS_IOC_SNAP_CREATE
:
4127 return btrfs_ioctl_snap_create(file
, argp
, 0);
4128 case BTRFS_IOC_SNAP_CREATE_V2
:
4129 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
4130 case BTRFS_IOC_SUBVOL_CREATE
:
4131 return btrfs_ioctl_snap_create(file
, argp
, 1);
4132 case BTRFS_IOC_SUBVOL_CREATE_V2
:
4133 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
4134 case BTRFS_IOC_SNAP_DESTROY
:
4135 return btrfs_ioctl_snap_destroy(file
, argp
);
4136 case BTRFS_IOC_SUBVOL_GETFLAGS
:
4137 return btrfs_ioctl_subvol_getflags(file
, argp
);
4138 case BTRFS_IOC_SUBVOL_SETFLAGS
:
4139 return btrfs_ioctl_subvol_setflags(file
, argp
);
4140 case BTRFS_IOC_DEFAULT_SUBVOL
:
4141 return btrfs_ioctl_default_subvol(file
, argp
);
4142 case BTRFS_IOC_DEFRAG
:
4143 return btrfs_ioctl_defrag(file
, NULL
);
4144 case BTRFS_IOC_DEFRAG_RANGE
:
4145 return btrfs_ioctl_defrag(file
, argp
);
4146 case BTRFS_IOC_RESIZE
:
4147 return btrfs_ioctl_resize(file
, argp
);
4148 case BTRFS_IOC_ADD_DEV
:
4149 return btrfs_ioctl_add_dev(root
, argp
);
4150 case BTRFS_IOC_RM_DEV
:
4151 return btrfs_ioctl_rm_dev(file
, argp
);
4152 case BTRFS_IOC_FS_INFO
:
4153 return btrfs_ioctl_fs_info(root
, argp
);
4154 case BTRFS_IOC_DEV_INFO
:
4155 return btrfs_ioctl_dev_info(root
, argp
);
4156 case BTRFS_IOC_BALANCE
:
4157 return btrfs_ioctl_balance(file
, NULL
);
4158 case BTRFS_IOC_CLONE
:
4159 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
4160 case BTRFS_IOC_CLONE_RANGE
:
4161 return btrfs_ioctl_clone_range(file
, argp
);
4162 case BTRFS_IOC_TRANS_START
:
4163 return btrfs_ioctl_trans_start(file
);
4164 case BTRFS_IOC_TRANS_END
:
4165 return btrfs_ioctl_trans_end(file
);
4166 case BTRFS_IOC_TREE_SEARCH
:
4167 return btrfs_ioctl_tree_search(file
, argp
);
4168 case BTRFS_IOC_INO_LOOKUP
:
4169 return btrfs_ioctl_ino_lookup(file
, argp
);
4170 case BTRFS_IOC_INO_PATHS
:
4171 return btrfs_ioctl_ino_to_path(root
, argp
);
4172 case BTRFS_IOC_LOGICAL_INO
:
4173 return btrfs_ioctl_logical_to_ino(root
, argp
);
4174 case BTRFS_IOC_SPACE_INFO
:
4175 return btrfs_ioctl_space_info(root
, argp
);
4176 case BTRFS_IOC_SYNC
:
4177 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
4179 case BTRFS_IOC_START_SYNC
:
4180 return btrfs_ioctl_start_sync(root
, argp
);
4181 case BTRFS_IOC_WAIT_SYNC
:
4182 return btrfs_ioctl_wait_sync(root
, argp
);
4183 case BTRFS_IOC_SCRUB
:
4184 return btrfs_ioctl_scrub(file
, argp
);
4185 case BTRFS_IOC_SCRUB_CANCEL
:
4186 return btrfs_ioctl_scrub_cancel(root
, argp
);
4187 case BTRFS_IOC_SCRUB_PROGRESS
:
4188 return btrfs_ioctl_scrub_progress(root
, argp
);
4189 case BTRFS_IOC_BALANCE_V2
:
4190 return btrfs_ioctl_balance(file
, argp
);
4191 case BTRFS_IOC_BALANCE_CTL
:
4192 return btrfs_ioctl_balance_ctl(root
, arg
);
4193 case BTRFS_IOC_BALANCE_PROGRESS
:
4194 return btrfs_ioctl_balance_progress(root
, argp
);
4195 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
4196 return btrfs_ioctl_set_received_subvol(file
, argp
);
4197 case BTRFS_IOC_SEND
:
4198 return btrfs_ioctl_send(file
, argp
);
4199 case BTRFS_IOC_GET_DEV_STATS
:
4200 return btrfs_ioctl_get_dev_stats(root
, argp
);
4201 case BTRFS_IOC_QUOTA_CTL
:
4202 return btrfs_ioctl_quota_ctl(file
, argp
);
4203 case BTRFS_IOC_QGROUP_ASSIGN
:
4204 return btrfs_ioctl_qgroup_assign(file
, argp
);
4205 case BTRFS_IOC_QGROUP_CREATE
:
4206 return btrfs_ioctl_qgroup_create(file
, argp
);
4207 case BTRFS_IOC_QGROUP_LIMIT
:
4208 return btrfs_ioctl_qgroup_limit(file
, argp
);
4209 case BTRFS_IOC_QUOTA_RESCAN
:
4210 return btrfs_ioctl_quota_rescan(file
, argp
);
4211 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
4212 return btrfs_ioctl_quota_rescan_status(file
, argp
);
4213 case BTRFS_IOC_DEV_REPLACE
:
4214 return btrfs_ioctl_dev_replace(root
, argp
);
4215 case BTRFS_IOC_GET_FSLABEL
:
4216 return btrfs_ioctl_get_fslabel(file
, argp
);
4217 case BTRFS_IOC_SET_FSLABEL
:
4218 return btrfs_ioctl_set_fslabel(file
, argp
);