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 ret
= mnt_want_write_file(file
);
1503 namelen
= strlen(name
);
1504 if (strchr(name
, '/')) {
1506 goto out_drop_write
;
1509 if (name
[0] == '.' &&
1510 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1512 goto out_drop_write
;
1516 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1517 NULL
, transid
, readonly
, inherit
);
1519 struct fd src
= fdget(fd
);
1520 struct inode
*src_inode
;
1523 goto out_drop_write
;
1526 src_inode
= file_inode(src
.file
);
1527 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1528 printk(KERN_INFO
"btrfs: Snapshot src from "
1532 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1533 BTRFS_I(src_inode
)->root
,
1534 transid
, readonly
, inherit
);
1539 mnt_drop_write_file(file
);
1544 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1545 void __user
*arg
, int subvol
)
1547 struct btrfs_ioctl_vol_args
*vol_args
;
1550 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1551 if (IS_ERR(vol_args
))
1552 return PTR_ERR(vol_args
);
1553 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1555 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1556 vol_args
->fd
, subvol
,
1563 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1564 void __user
*arg
, int subvol
)
1566 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1570 bool readonly
= false;
1571 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1573 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1574 if (IS_ERR(vol_args
))
1575 return PTR_ERR(vol_args
);
1576 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1578 if (vol_args
->flags
&
1579 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1580 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1585 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1587 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1589 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1590 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1594 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1595 if (IS_ERR(inherit
)) {
1596 ret
= PTR_ERR(inherit
);
1601 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1602 vol_args
->fd
, subvol
, ptr
,
1605 if (ret
== 0 && ptr
&&
1607 offsetof(struct btrfs_ioctl_vol_args_v2
,
1608 transid
), ptr
, sizeof(*ptr
)))
1616 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1619 struct inode
*inode
= file_inode(file
);
1620 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1624 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1627 down_read(&root
->fs_info
->subvol_sem
);
1628 if (btrfs_root_readonly(root
))
1629 flags
|= BTRFS_SUBVOL_RDONLY
;
1630 up_read(&root
->fs_info
->subvol_sem
);
1632 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1638 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1641 struct inode
*inode
= file_inode(file
);
1642 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1643 struct btrfs_trans_handle
*trans
;
1648 ret
= mnt_want_write_file(file
);
1652 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1654 goto out_drop_write
;
1657 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1659 goto out_drop_write
;
1662 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1664 goto out_drop_write
;
1667 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1669 goto out_drop_write
;
1672 if (!inode_owner_or_capable(inode
)) {
1674 goto out_drop_write
;
1677 down_write(&root
->fs_info
->subvol_sem
);
1680 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1683 root_flags
= btrfs_root_flags(&root
->root_item
);
1684 if (flags
& BTRFS_SUBVOL_RDONLY
)
1685 btrfs_set_root_flags(&root
->root_item
,
1686 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1688 btrfs_set_root_flags(&root
->root_item
,
1689 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1691 trans
= btrfs_start_transaction(root
, 1);
1692 if (IS_ERR(trans
)) {
1693 ret
= PTR_ERR(trans
);
1697 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1698 &root
->root_key
, &root
->root_item
);
1700 btrfs_commit_transaction(trans
, root
);
1703 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1705 up_write(&root
->fs_info
->subvol_sem
);
1707 mnt_drop_write_file(file
);
1713 * helper to check if the subvolume references other subvolumes
1715 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1717 struct btrfs_path
*path
;
1718 struct btrfs_key key
;
1721 path
= btrfs_alloc_path();
1725 key
.objectid
= root
->root_key
.objectid
;
1726 key
.type
= BTRFS_ROOT_REF_KEY
;
1727 key
.offset
= (u64
)-1;
1729 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1736 if (path
->slots
[0] > 0) {
1738 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1739 if (key
.objectid
== root
->root_key
.objectid
&&
1740 key
.type
== BTRFS_ROOT_REF_KEY
)
1744 btrfs_free_path(path
);
1748 static noinline
int key_in_sk(struct btrfs_key
*key
,
1749 struct btrfs_ioctl_search_key
*sk
)
1751 struct btrfs_key test
;
1754 test
.objectid
= sk
->min_objectid
;
1755 test
.type
= sk
->min_type
;
1756 test
.offset
= sk
->min_offset
;
1758 ret
= btrfs_comp_cpu_keys(key
, &test
);
1762 test
.objectid
= sk
->max_objectid
;
1763 test
.type
= sk
->max_type
;
1764 test
.offset
= sk
->max_offset
;
1766 ret
= btrfs_comp_cpu_keys(key
, &test
);
1772 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1773 struct btrfs_path
*path
,
1774 struct btrfs_key
*key
,
1775 struct btrfs_ioctl_search_key
*sk
,
1777 unsigned long *sk_offset
,
1781 struct extent_buffer
*leaf
;
1782 struct btrfs_ioctl_search_header sh
;
1783 unsigned long item_off
;
1784 unsigned long item_len
;
1790 leaf
= path
->nodes
[0];
1791 slot
= path
->slots
[0];
1792 nritems
= btrfs_header_nritems(leaf
);
1794 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1798 found_transid
= btrfs_header_generation(leaf
);
1800 for (i
= slot
; i
< nritems
; i
++) {
1801 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1802 item_len
= btrfs_item_size_nr(leaf
, i
);
1804 if (item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1807 if (sizeof(sh
) + item_len
+ *sk_offset
>
1808 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1813 btrfs_item_key_to_cpu(leaf
, key
, i
);
1814 if (!key_in_sk(key
, sk
))
1817 sh
.objectid
= key
->objectid
;
1818 sh
.offset
= key
->offset
;
1819 sh
.type
= key
->type
;
1821 sh
.transid
= found_transid
;
1823 /* copy search result header */
1824 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1825 *sk_offset
+= sizeof(sh
);
1828 char *p
= buf
+ *sk_offset
;
1830 read_extent_buffer(leaf
, p
,
1831 item_off
, item_len
);
1832 *sk_offset
+= item_len
;
1836 if (*num_found
>= sk
->nr_items
)
1841 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1843 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1846 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
1856 static noinline
int search_ioctl(struct inode
*inode
,
1857 struct btrfs_ioctl_search_args
*args
)
1859 struct btrfs_root
*root
;
1860 struct btrfs_key key
;
1861 struct btrfs_key max_key
;
1862 struct btrfs_path
*path
;
1863 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
1864 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
1867 unsigned long sk_offset
= 0;
1869 path
= btrfs_alloc_path();
1873 if (sk
->tree_id
== 0) {
1874 /* search the root of the inode that was passed */
1875 root
= BTRFS_I(inode
)->root
;
1877 key
.objectid
= sk
->tree_id
;
1878 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1879 key
.offset
= (u64
)-1;
1880 root
= btrfs_read_fs_root_no_name(info
, &key
);
1882 printk(KERN_ERR
"could not find root %llu\n",
1884 btrfs_free_path(path
);
1889 key
.objectid
= sk
->min_objectid
;
1890 key
.type
= sk
->min_type
;
1891 key
.offset
= sk
->min_offset
;
1893 max_key
.objectid
= sk
->max_objectid
;
1894 max_key
.type
= sk
->max_type
;
1895 max_key
.offset
= sk
->max_offset
;
1897 path
->keep_locks
= 1;
1900 ret
= btrfs_search_forward(root
, &key
, &max_key
, path
,
1907 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
1908 &sk_offset
, &num_found
);
1909 btrfs_release_path(path
);
1910 if (ret
|| num_found
>= sk
->nr_items
)
1916 sk
->nr_items
= num_found
;
1917 btrfs_free_path(path
);
1921 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
1924 struct btrfs_ioctl_search_args
*args
;
1925 struct inode
*inode
;
1928 if (!capable(CAP_SYS_ADMIN
))
1931 args
= memdup_user(argp
, sizeof(*args
));
1933 return PTR_ERR(args
);
1935 inode
= file_inode(file
);
1936 ret
= search_ioctl(inode
, args
);
1937 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1944 * Search INODE_REFs to identify path name of 'dirid' directory
1945 * in a 'tree_id' tree. and sets path name to 'name'.
1947 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1948 u64 tree_id
, u64 dirid
, char *name
)
1950 struct btrfs_root
*root
;
1951 struct btrfs_key key
;
1957 struct btrfs_inode_ref
*iref
;
1958 struct extent_buffer
*l
;
1959 struct btrfs_path
*path
;
1961 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1966 path
= btrfs_alloc_path();
1970 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
1972 key
.objectid
= tree_id
;
1973 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1974 key
.offset
= (u64
)-1;
1975 root
= btrfs_read_fs_root_no_name(info
, &key
);
1977 printk(KERN_ERR
"could not find root %llu\n", tree_id
);
1982 key
.objectid
= dirid
;
1983 key
.type
= BTRFS_INODE_REF_KEY
;
1984 key
.offset
= (u64
)-1;
1987 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1992 slot
= path
->slots
[0];
1993 if (ret
> 0 && slot
> 0)
1995 btrfs_item_key_to_cpu(l
, &key
, slot
);
1997 if (ret
> 0 && (key
.objectid
!= dirid
||
1998 key
.type
!= BTRFS_INODE_REF_KEY
)) {
2003 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2004 len
= btrfs_inode_ref_name_len(l
, iref
);
2006 total_len
+= len
+ 1;
2011 read_extent_buffer(l
, ptr
,(unsigned long)(iref
+ 1), len
);
2013 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2016 btrfs_release_path(path
);
2017 key
.objectid
= key
.offset
;
2018 key
.offset
= (u64
)-1;
2019 dirid
= key
.objectid
;
2023 memmove(name
, ptr
, total_len
);
2024 name
[total_len
]='\0';
2027 btrfs_free_path(path
);
2031 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2034 struct btrfs_ioctl_ino_lookup_args
*args
;
2035 struct inode
*inode
;
2038 if (!capable(CAP_SYS_ADMIN
))
2041 args
= memdup_user(argp
, sizeof(*args
));
2043 return PTR_ERR(args
);
2045 inode
= file_inode(file
);
2047 if (args
->treeid
== 0)
2048 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2050 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2051 args
->treeid
, args
->objectid
,
2054 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2061 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2064 struct dentry
*parent
= fdentry(file
);
2065 struct dentry
*dentry
;
2066 struct inode
*dir
= parent
->d_inode
;
2067 struct inode
*inode
;
2068 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2069 struct btrfs_root
*dest
= NULL
;
2070 struct btrfs_ioctl_vol_args
*vol_args
;
2071 struct btrfs_trans_handle
*trans
;
2072 struct btrfs_block_rsv block_rsv
;
2073 u64 qgroup_reserved
;
2078 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2079 if (IS_ERR(vol_args
))
2080 return PTR_ERR(vol_args
);
2082 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2083 namelen
= strlen(vol_args
->name
);
2084 if (strchr(vol_args
->name
, '/') ||
2085 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2090 err
= mnt_want_write_file(file
);
2094 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2097 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2098 if (IS_ERR(dentry
)) {
2099 err
= PTR_ERR(dentry
);
2100 goto out_unlock_dir
;
2103 if (!dentry
->d_inode
) {
2108 inode
= dentry
->d_inode
;
2109 dest
= BTRFS_I(inode
)->root
;
2110 if (!capable(CAP_SYS_ADMIN
)){
2112 * Regular user. Only allow this with a special mount
2113 * option, when the user has write+exec access to the
2114 * subvol root, and when rmdir(2) would have been
2117 * Note that this is _not_ check that the subvol is
2118 * empty or doesn't contain data that we wouldn't
2119 * otherwise be able to delete.
2121 * Users who want to delete empty subvols should try
2125 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2129 * Do not allow deletion if the parent dir is the same
2130 * as the dir to be deleted. That means the ioctl
2131 * must be called on the dentry referencing the root
2132 * of the subvol, not a random directory contained
2139 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2144 /* check if subvolume may be deleted by a user */
2145 err
= btrfs_may_delete(dir
, dentry
, 1);
2149 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2154 mutex_lock(&inode
->i_mutex
);
2155 err
= d_invalidate(dentry
);
2159 down_write(&root
->fs_info
->subvol_sem
);
2161 err
= may_destroy_subvol(dest
);
2165 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2167 * One for dir inode, two for dir entries, two for root
2170 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2171 5, &qgroup_reserved
);
2175 trans
= btrfs_start_transaction(root
, 0);
2176 if (IS_ERR(trans
)) {
2177 err
= PTR_ERR(trans
);
2180 trans
->block_rsv
= &block_rsv
;
2181 trans
->bytes_reserved
= block_rsv
.size
;
2183 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2184 dest
->root_key
.objectid
,
2185 dentry
->d_name
.name
,
2186 dentry
->d_name
.len
);
2189 btrfs_abort_transaction(trans
, root
, ret
);
2193 btrfs_record_root_in_trans(trans
, dest
);
2195 memset(&dest
->root_item
.drop_progress
, 0,
2196 sizeof(dest
->root_item
.drop_progress
));
2197 dest
->root_item
.drop_level
= 0;
2198 btrfs_set_root_refs(&dest
->root_item
, 0);
2200 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
2201 ret
= btrfs_insert_orphan_item(trans
,
2202 root
->fs_info
->tree_root
,
2203 dest
->root_key
.objectid
);
2205 btrfs_abort_transaction(trans
, root
, ret
);
2211 trans
->block_rsv
= NULL
;
2212 trans
->bytes_reserved
= 0;
2213 ret
= btrfs_end_transaction(trans
, root
);
2216 inode
->i_flags
|= S_DEAD
;
2218 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2220 up_write(&root
->fs_info
->subvol_sem
);
2222 mutex_unlock(&inode
->i_mutex
);
2224 shrink_dcache_sb(root
->fs_info
->sb
);
2225 btrfs_invalidate_inodes(dest
);
2229 if (dest
->cache_inode
) {
2230 iput(dest
->cache_inode
);
2231 dest
->cache_inode
= NULL
;
2237 mutex_unlock(&dir
->i_mutex
);
2238 mnt_drop_write_file(file
);
2244 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2246 struct inode
*inode
= file_inode(file
);
2247 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2248 struct btrfs_ioctl_defrag_range_args
*range
;
2251 ret
= mnt_want_write_file(file
);
2255 if (btrfs_root_readonly(root
)) {
2260 switch (inode
->i_mode
& S_IFMT
) {
2262 if (!capable(CAP_SYS_ADMIN
)) {
2266 ret
= btrfs_defrag_root(root
);
2269 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2272 if (!(file
->f_mode
& FMODE_WRITE
)) {
2277 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2284 if (copy_from_user(range
, argp
,
2290 /* compression requires us to start the IO */
2291 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2292 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2293 range
->extent_thresh
= (u32
)-1;
2296 /* the rest are all set to zero by kzalloc */
2297 range
->len
= (u64
)-1;
2299 ret
= btrfs_defrag_file(file_inode(file
), file
,
2309 mnt_drop_write_file(file
);
2313 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2315 struct btrfs_ioctl_vol_args
*vol_args
;
2318 if (!capable(CAP_SYS_ADMIN
))
2321 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2323 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2327 mutex_lock(&root
->fs_info
->volume_mutex
);
2328 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2329 if (IS_ERR(vol_args
)) {
2330 ret
= PTR_ERR(vol_args
);
2334 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2335 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2339 mutex_unlock(&root
->fs_info
->volume_mutex
);
2340 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2344 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2346 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2347 struct btrfs_ioctl_vol_args
*vol_args
;
2350 if (!capable(CAP_SYS_ADMIN
))
2353 ret
= mnt_want_write_file(file
);
2357 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2359 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2360 mnt_drop_write_file(file
);
2364 mutex_lock(&root
->fs_info
->volume_mutex
);
2365 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2366 if (IS_ERR(vol_args
)) {
2367 ret
= PTR_ERR(vol_args
);
2371 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2372 ret
= btrfs_rm_device(root
, vol_args
->name
);
2376 mutex_unlock(&root
->fs_info
->volume_mutex
);
2377 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2378 mnt_drop_write_file(file
);
2382 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2384 struct btrfs_ioctl_fs_info_args
*fi_args
;
2385 struct btrfs_device
*device
;
2386 struct btrfs_device
*next
;
2387 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2390 if (!capable(CAP_SYS_ADMIN
))
2393 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2397 fi_args
->num_devices
= fs_devices
->num_devices
;
2398 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2400 mutex_lock(&fs_devices
->device_list_mutex
);
2401 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2402 if (device
->devid
> fi_args
->max_id
)
2403 fi_args
->max_id
= device
->devid
;
2405 mutex_unlock(&fs_devices
->device_list_mutex
);
2407 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2414 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2416 struct btrfs_ioctl_dev_info_args
*di_args
;
2417 struct btrfs_device
*dev
;
2418 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2420 char *s_uuid
= NULL
;
2421 char empty_uuid
[BTRFS_UUID_SIZE
] = {0};
2423 if (!capable(CAP_SYS_ADMIN
))
2426 di_args
= memdup_user(arg
, sizeof(*di_args
));
2427 if (IS_ERR(di_args
))
2428 return PTR_ERR(di_args
);
2430 if (memcmp(empty_uuid
, di_args
->uuid
, BTRFS_UUID_SIZE
) != 0)
2431 s_uuid
= di_args
->uuid
;
2433 mutex_lock(&fs_devices
->device_list_mutex
);
2434 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2435 mutex_unlock(&fs_devices
->device_list_mutex
);
2442 di_args
->devid
= dev
->devid
;
2443 di_args
->bytes_used
= dev
->bytes_used
;
2444 di_args
->total_bytes
= dev
->total_bytes
;
2445 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2447 struct rcu_string
*name
;
2450 name
= rcu_dereference(dev
->name
);
2451 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2453 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2455 di_args
->path
[0] = '\0';
2459 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2466 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
2467 u64 off
, u64 olen
, u64 destoff
)
2469 struct inode
*inode
= file_inode(file
);
2470 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2473 struct btrfs_trans_handle
*trans
;
2474 struct btrfs_path
*path
;
2475 struct extent_buffer
*leaf
;
2477 struct btrfs_key key
;
2482 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
2486 * - split compressed inline extents. annoying: we need to
2487 * decompress into destination's address_space (the file offset
2488 * may change, so source mapping won't do), then recompress (or
2489 * otherwise reinsert) a subrange.
2490 * - allow ranges within the same file to be cloned (provided
2491 * they don't overlap)?
2494 /* the destination must be opened for writing */
2495 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
2498 if (btrfs_root_readonly(root
))
2501 ret
= mnt_want_write_file(file
);
2505 src_file
= fdget(srcfd
);
2506 if (!src_file
.file
) {
2508 goto out_drop_write
;
2512 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
2515 src
= file_inode(src_file
.file
);
2521 /* the src must be open for reading */
2522 if (!(src_file
.file
->f_mode
& FMODE_READ
))
2525 /* don't make the dst file partly checksummed */
2526 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2527 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
2531 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
2535 if (src
->i_sb
!= inode
->i_sb
)
2539 buf
= vmalloc(btrfs_level_size(root
, 0));
2543 path
= btrfs_alloc_path();
2551 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
2552 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
2554 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
2555 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
2558 /* determine range to clone */
2560 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
2563 olen
= len
= src
->i_size
- off
;
2564 /* if we extend to eof, continue to block boundary */
2565 if (off
+ len
== src
->i_size
)
2566 len
= ALIGN(src
->i_size
, bs
) - off
;
2568 /* verify the end result is block aligned */
2569 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
2570 !IS_ALIGNED(destoff
, bs
))
2573 if (destoff
> inode
->i_size
) {
2574 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
2579 /* truncate page cache pages from target inode range */
2580 truncate_inode_pages_range(&inode
->i_data
, destoff
,
2581 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
2583 /* do any pending delalloc/csum calc on src, one way or
2584 another, and lock file content */
2586 struct btrfs_ordered_extent
*ordered
;
2587 lock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2588 ordered
= btrfs_lookup_first_ordered_extent(src
, off
+ len
- 1);
2590 !test_range_bit(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1,
2591 EXTENT_DELALLOC
, 0, NULL
))
2593 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2595 btrfs_put_ordered_extent(ordered
);
2596 btrfs_wait_ordered_range(src
, off
, len
);
2600 key
.objectid
= btrfs_ino(src
);
2601 key
.type
= BTRFS_EXTENT_DATA_KEY
;
2606 * note the key will change type as we walk through the
2609 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
2614 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2615 if (path
->slots
[0] >= nritems
) {
2616 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
2621 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2623 leaf
= path
->nodes
[0];
2624 slot
= path
->slots
[0];
2626 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2627 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
2628 key
.objectid
!= btrfs_ino(src
))
2631 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
2632 struct btrfs_file_extent_item
*extent
;
2635 struct btrfs_key new_key
;
2636 u64 disko
= 0, diskl
= 0;
2637 u64 datao
= 0, datal
= 0;
2641 size
= btrfs_item_size_nr(leaf
, slot
);
2642 read_extent_buffer(leaf
, buf
,
2643 btrfs_item_ptr_offset(leaf
, slot
),
2646 extent
= btrfs_item_ptr(leaf
, slot
,
2647 struct btrfs_file_extent_item
);
2648 comp
= btrfs_file_extent_compression(leaf
, extent
);
2649 type
= btrfs_file_extent_type(leaf
, extent
);
2650 if (type
== BTRFS_FILE_EXTENT_REG
||
2651 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2652 disko
= btrfs_file_extent_disk_bytenr(leaf
,
2654 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
2656 datao
= btrfs_file_extent_offset(leaf
, extent
);
2657 datal
= btrfs_file_extent_num_bytes(leaf
,
2659 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2660 /* take upper bound, may be compressed */
2661 datal
= btrfs_file_extent_ram_bytes(leaf
,
2664 btrfs_release_path(path
);
2666 if (key
.offset
+ datal
<= off
||
2667 key
.offset
>= off
+ len
- 1)
2670 memcpy(&new_key
, &key
, sizeof(new_key
));
2671 new_key
.objectid
= btrfs_ino(inode
);
2672 if (off
<= key
.offset
)
2673 new_key
.offset
= key
.offset
+ destoff
- off
;
2675 new_key
.offset
= destoff
;
2678 * 1 - adjusting old extent (we may have to split it)
2679 * 1 - add new extent
2682 trans
= btrfs_start_transaction(root
, 3);
2683 if (IS_ERR(trans
)) {
2684 ret
= PTR_ERR(trans
);
2688 if (type
== BTRFS_FILE_EXTENT_REG
||
2689 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2691 * a | --- range to clone ---| b
2692 * | ------------- extent ------------- |
2695 /* substract range b */
2696 if (key
.offset
+ datal
> off
+ len
)
2697 datal
= off
+ len
- key
.offset
;
2699 /* substract range a */
2700 if (off
> key
.offset
) {
2701 datao
+= off
- key
.offset
;
2702 datal
-= off
- key
.offset
;
2705 ret
= btrfs_drop_extents(trans
, root
, inode
,
2707 new_key
.offset
+ datal
,
2710 btrfs_abort_transaction(trans
, root
,
2712 btrfs_end_transaction(trans
, root
);
2716 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2719 btrfs_abort_transaction(trans
, root
,
2721 btrfs_end_transaction(trans
, root
);
2725 leaf
= path
->nodes
[0];
2726 slot
= path
->slots
[0];
2727 write_extent_buffer(leaf
, buf
,
2728 btrfs_item_ptr_offset(leaf
, slot
),
2731 extent
= btrfs_item_ptr(leaf
, slot
,
2732 struct btrfs_file_extent_item
);
2734 /* disko == 0 means it's a hole */
2738 btrfs_set_file_extent_offset(leaf
, extent
,
2740 btrfs_set_file_extent_num_bytes(leaf
, extent
,
2743 inode_add_bytes(inode
, datal
);
2744 ret
= btrfs_inc_extent_ref(trans
, root
,
2746 root
->root_key
.objectid
,
2748 new_key
.offset
- datao
,
2751 btrfs_abort_transaction(trans
,
2754 btrfs_end_transaction(trans
,
2760 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2763 if (off
> key
.offset
) {
2764 skip
= off
- key
.offset
;
2765 new_key
.offset
+= skip
;
2768 if (key
.offset
+ datal
> off
+ len
)
2769 trim
= key
.offset
+ datal
- (off
+ len
);
2771 if (comp
&& (skip
|| trim
)) {
2773 btrfs_end_transaction(trans
, root
);
2776 size
-= skip
+ trim
;
2777 datal
-= skip
+ trim
;
2779 ret
= btrfs_drop_extents(trans
, root
, inode
,
2781 new_key
.offset
+ datal
,
2784 btrfs_abort_transaction(trans
, root
,
2786 btrfs_end_transaction(trans
, root
);
2790 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2793 btrfs_abort_transaction(trans
, root
,
2795 btrfs_end_transaction(trans
, root
);
2801 btrfs_file_extent_calc_inline_size(0);
2802 memmove(buf
+start
, buf
+start
+skip
,
2806 leaf
= path
->nodes
[0];
2807 slot
= path
->slots
[0];
2808 write_extent_buffer(leaf
, buf
,
2809 btrfs_item_ptr_offset(leaf
, slot
),
2811 inode_add_bytes(inode
, datal
);
2814 btrfs_mark_buffer_dirty(leaf
);
2815 btrfs_release_path(path
);
2817 inode_inc_iversion(inode
);
2818 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2821 * we round up to the block size at eof when
2822 * determining which extents to clone above,
2823 * but shouldn't round up the file size
2825 endoff
= new_key
.offset
+ datal
;
2826 if (endoff
> destoff
+olen
)
2827 endoff
= destoff
+olen
;
2828 if (endoff
> inode
->i_size
)
2829 btrfs_i_size_write(inode
, endoff
);
2831 ret
= btrfs_update_inode(trans
, root
, inode
);
2833 btrfs_abort_transaction(trans
, root
, ret
);
2834 btrfs_end_transaction(trans
, root
);
2837 ret
= btrfs_end_transaction(trans
, root
);
2840 btrfs_release_path(path
);
2845 btrfs_release_path(path
);
2846 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2848 mutex_unlock(&src
->i_mutex
);
2849 mutex_unlock(&inode
->i_mutex
);
2851 btrfs_free_path(path
);
2855 mnt_drop_write_file(file
);
2859 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
2861 struct btrfs_ioctl_clone_range_args args
;
2863 if (copy_from_user(&args
, argp
, sizeof(args
)))
2865 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
2866 args
.src_length
, args
.dest_offset
);
2870 * there are many ways the trans_start and trans_end ioctls can lead
2871 * to deadlocks. They should only be used by applications that
2872 * basically own the machine, and have a very in depth understanding
2873 * of all the possible deadlocks and enospc problems.
2875 static long btrfs_ioctl_trans_start(struct file
*file
)
2877 struct inode
*inode
= file_inode(file
);
2878 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2879 struct btrfs_trans_handle
*trans
;
2883 if (!capable(CAP_SYS_ADMIN
))
2887 if (file
->private_data
)
2891 if (btrfs_root_readonly(root
))
2894 ret
= mnt_want_write_file(file
);
2898 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
2901 trans
= btrfs_start_ioctl_transaction(root
);
2905 file
->private_data
= trans
;
2909 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
2910 mnt_drop_write_file(file
);
2915 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2917 struct inode
*inode
= file_inode(file
);
2918 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2919 struct btrfs_root
*new_root
;
2920 struct btrfs_dir_item
*di
;
2921 struct btrfs_trans_handle
*trans
;
2922 struct btrfs_path
*path
;
2923 struct btrfs_key location
;
2924 struct btrfs_disk_key disk_key
;
2929 if (!capable(CAP_SYS_ADMIN
))
2932 ret
= mnt_want_write_file(file
);
2936 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
2942 objectid
= root
->root_key
.objectid
;
2944 location
.objectid
= objectid
;
2945 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2946 location
.offset
= (u64
)-1;
2948 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
2949 if (IS_ERR(new_root
)) {
2950 ret
= PTR_ERR(new_root
);
2954 if (btrfs_root_refs(&new_root
->root_item
) == 0) {
2959 path
= btrfs_alloc_path();
2964 path
->leave_spinning
= 1;
2966 trans
= btrfs_start_transaction(root
, 1);
2967 if (IS_ERR(trans
)) {
2968 btrfs_free_path(path
);
2969 ret
= PTR_ERR(trans
);
2973 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
2974 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
2975 dir_id
, "default", 7, 1);
2976 if (IS_ERR_OR_NULL(di
)) {
2977 btrfs_free_path(path
);
2978 btrfs_end_transaction(trans
, root
);
2979 printk(KERN_ERR
"Umm, you don't have the default dir item, "
2980 "this isn't going to work\n");
2985 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
2986 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
2987 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2988 btrfs_free_path(path
);
2990 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
2991 btrfs_end_transaction(trans
, root
);
2993 mnt_drop_write_file(file
);
2997 void btrfs_get_block_group_info(struct list_head
*groups_list
,
2998 struct btrfs_ioctl_space_info
*space
)
3000 struct btrfs_block_group_cache
*block_group
;
3002 space
->total_bytes
= 0;
3003 space
->used_bytes
= 0;
3005 list_for_each_entry(block_group
, groups_list
, list
) {
3006 space
->flags
= block_group
->flags
;
3007 space
->total_bytes
+= block_group
->key
.offset
;
3008 space
->used_bytes
+=
3009 btrfs_block_group_used(&block_group
->item
);
3013 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
3015 struct btrfs_ioctl_space_args space_args
;
3016 struct btrfs_ioctl_space_info space
;
3017 struct btrfs_ioctl_space_info
*dest
;
3018 struct btrfs_ioctl_space_info
*dest_orig
;
3019 struct btrfs_ioctl_space_info __user
*user_dest
;
3020 struct btrfs_space_info
*info
;
3021 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
3022 BTRFS_BLOCK_GROUP_SYSTEM
,
3023 BTRFS_BLOCK_GROUP_METADATA
,
3024 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
3031 if (copy_from_user(&space_args
,
3032 (struct btrfs_ioctl_space_args __user
*)arg
,
3033 sizeof(space_args
)))
3036 for (i
= 0; i
< num_types
; i
++) {
3037 struct btrfs_space_info
*tmp
;
3041 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3043 if (tmp
->flags
== types
[i
]) {
3053 down_read(&info
->groups_sem
);
3054 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3055 if (!list_empty(&info
->block_groups
[c
]))
3058 up_read(&info
->groups_sem
);
3061 /* space_slots == 0 means they are asking for a count */
3062 if (space_args
.space_slots
== 0) {
3063 space_args
.total_spaces
= slot_count
;
3067 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
3069 alloc_size
= sizeof(*dest
) * slot_count
;
3071 /* we generally have at most 6 or so space infos, one for each raid
3072 * level. So, a whole page should be more than enough for everyone
3074 if (alloc_size
> PAGE_CACHE_SIZE
)
3077 space_args
.total_spaces
= 0;
3078 dest
= kmalloc(alloc_size
, GFP_NOFS
);
3083 /* now we have a buffer to copy into */
3084 for (i
= 0; i
< num_types
; i
++) {
3085 struct btrfs_space_info
*tmp
;
3092 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3094 if (tmp
->flags
== types
[i
]) {
3103 down_read(&info
->groups_sem
);
3104 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3105 if (!list_empty(&info
->block_groups
[c
])) {
3106 btrfs_get_block_group_info(
3107 &info
->block_groups
[c
], &space
);
3108 memcpy(dest
, &space
, sizeof(space
));
3110 space_args
.total_spaces
++;
3116 up_read(&info
->groups_sem
);
3119 user_dest
= (struct btrfs_ioctl_space_info __user
*)
3120 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
3122 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
3127 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
3134 * there are many ways the trans_start and trans_end ioctls can lead
3135 * to deadlocks. They should only be used by applications that
3136 * basically own the machine, and have a very in depth understanding
3137 * of all the possible deadlocks and enospc problems.
3139 long btrfs_ioctl_trans_end(struct file
*file
)
3141 struct inode
*inode
= file_inode(file
);
3142 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3143 struct btrfs_trans_handle
*trans
;
3145 trans
= file
->private_data
;
3148 file
->private_data
= NULL
;
3150 btrfs_end_transaction(trans
, root
);
3152 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3154 mnt_drop_write_file(file
);
3158 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
3161 struct btrfs_trans_handle
*trans
;
3165 trans
= btrfs_attach_transaction_barrier(root
);
3166 if (IS_ERR(trans
)) {
3167 if (PTR_ERR(trans
) != -ENOENT
)
3168 return PTR_ERR(trans
);
3170 /* No running transaction, don't bother */
3171 transid
= root
->fs_info
->last_trans_committed
;
3174 transid
= trans
->transid
;
3175 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
3177 btrfs_end_transaction(trans
, root
);
3182 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
3187 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
3193 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
3196 transid
= 0; /* current trans */
3198 return btrfs_wait_for_commit(root
, transid
);
3201 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
3203 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3204 struct btrfs_ioctl_scrub_args
*sa
;
3207 if (!capable(CAP_SYS_ADMIN
))
3210 sa
= memdup_user(arg
, sizeof(*sa
));
3214 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
3215 ret
= mnt_want_write_file(file
);
3220 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
3221 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
3224 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3227 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
3228 mnt_drop_write_file(file
);
3234 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
3236 if (!capable(CAP_SYS_ADMIN
))
3239 return btrfs_scrub_cancel(root
->fs_info
);
3242 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
3245 struct btrfs_ioctl_scrub_args
*sa
;
3248 if (!capable(CAP_SYS_ADMIN
))
3251 sa
= memdup_user(arg
, sizeof(*sa
));
3255 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
3257 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3264 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
3267 struct btrfs_ioctl_get_dev_stats
*sa
;
3270 sa
= memdup_user(arg
, sizeof(*sa
));
3274 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
3279 ret
= btrfs_get_dev_stats(root
, sa
);
3281 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3288 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
3290 struct btrfs_ioctl_dev_replace_args
*p
;
3293 if (!capable(CAP_SYS_ADMIN
))
3296 p
= memdup_user(arg
, sizeof(*p
));
3301 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
3303 &root
->fs_info
->mutually_exclusive_operation_running
,
3305 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3308 ret
= btrfs_dev_replace_start(root
, p
);
3310 &root
->fs_info
->mutually_exclusive_operation_running
,
3314 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
3315 btrfs_dev_replace_status(root
->fs_info
, p
);
3318 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
3319 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
3326 if (copy_to_user(arg
, p
, sizeof(*p
)))
3333 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
3339 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
3340 struct inode_fs_paths
*ipath
= NULL
;
3341 struct btrfs_path
*path
;
3343 if (!capable(CAP_DAC_READ_SEARCH
))
3346 path
= btrfs_alloc_path();
3352 ipa
= memdup_user(arg
, sizeof(*ipa
));
3359 size
= min_t(u32
, ipa
->size
, 4096);
3360 ipath
= init_ipath(size
, root
, path
);
3361 if (IS_ERR(ipath
)) {
3362 ret
= PTR_ERR(ipath
);
3367 ret
= paths_from_inode(ipa
->inum
, ipath
);
3371 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
3372 rel_ptr
= ipath
->fspath
->val
[i
] -
3373 (u64
)(unsigned long)ipath
->fspath
->val
;
3374 ipath
->fspath
->val
[i
] = rel_ptr
;
3377 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
3378 (void *)(unsigned long)ipath
->fspath
, size
);
3385 btrfs_free_path(path
);
3392 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
3394 struct btrfs_data_container
*inodes
= ctx
;
3395 const size_t c
= 3 * sizeof(u64
);
3397 if (inodes
->bytes_left
>= c
) {
3398 inodes
->bytes_left
-= c
;
3399 inodes
->val
[inodes
->elem_cnt
] = inum
;
3400 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
3401 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
3402 inodes
->elem_cnt
+= 3;
3404 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
3405 inodes
->bytes_left
= 0;
3406 inodes
->elem_missed
+= 3;
3412 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
3417 struct btrfs_ioctl_logical_ino_args
*loi
;
3418 struct btrfs_data_container
*inodes
= NULL
;
3419 struct btrfs_path
*path
= NULL
;
3421 if (!capable(CAP_SYS_ADMIN
))
3424 loi
= memdup_user(arg
, sizeof(*loi
));
3431 path
= btrfs_alloc_path();
3437 size
= min_t(u32
, loi
->size
, 64 * 1024);
3438 inodes
= init_data_container(size
);
3439 if (IS_ERR(inodes
)) {
3440 ret
= PTR_ERR(inodes
);
3445 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
3446 build_ino_list
, inodes
);
3452 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
3453 (void *)(unsigned long)inodes
, size
);
3458 btrfs_free_path(path
);
3465 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
3466 struct btrfs_ioctl_balance_args
*bargs
)
3468 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
3470 bargs
->flags
= bctl
->flags
;
3472 if (atomic_read(&fs_info
->balance_running
))
3473 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
3474 if (atomic_read(&fs_info
->balance_pause_req
))
3475 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
3476 if (atomic_read(&fs_info
->balance_cancel_req
))
3477 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
3479 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
3480 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
3481 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
3484 spin_lock(&fs_info
->balance_lock
);
3485 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3486 spin_unlock(&fs_info
->balance_lock
);
3488 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3492 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
3494 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3495 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3496 struct btrfs_ioctl_balance_args
*bargs
;
3497 struct btrfs_balance_control
*bctl
;
3498 bool need_unlock
; /* for mut. excl. ops lock */
3501 if (!capable(CAP_SYS_ADMIN
))
3504 ret
= mnt_want_write_file(file
);
3509 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
3510 mutex_lock(&fs_info
->volume_mutex
);
3511 mutex_lock(&fs_info
->balance_mutex
);
3517 * mut. excl. ops lock is locked. Three possibilites:
3518 * (1) some other op is running
3519 * (2) balance is running
3520 * (3) balance is paused -- special case (think resume)
3522 mutex_lock(&fs_info
->balance_mutex
);
3523 if (fs_info
->balance_ctl
) {
3524 /* this is either (2) or (3) */
3525 if (!atomic_read(&fs_info
->balance_running
)) {
3526 mutex_unlock(&fs_info
->balance_mutex
);
3527 if (!mutex_trylock(&fs_info
->volume_mutex
))
3529 mutex_lock(&fs_info
->balance_mutex
);
3531 if (fs_info
->balance_ctl
&&
3532 !atomic_read(&fs_info
->balance_running
)) {
3534 need_unlock
= false;
3538 mutex_unlock(&fs_info
->balance_mutex
);
3539 mutex_unlock(&fs_info
->volume_mutex
);
3543 mutex_unlock(&fs_info
->balance_mutex
);
3549 mutex_unlock(&fs_info
->balance_mutex
);
3550 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3556 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
3559 bargs
= memdup_user(arg
, sizeof(*bargs
));
3560 if (IS_ERR(bargs
)) {
3561 ret
= PTR_ERR(bargs
);
3565 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
3566 if (!fs_info
->balance_ctl
) {
3571 bctl
= fs_info
->balance_ctl
;
3572 spin_lock(&fs_info
->balance_lock
);
3573 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
3574 spin_unlock(&fs_info
->balance_lock
);
3582 if (fs_info
->balance_ctl
) {
3587 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
3593 bctl
->fs_info
= fs_info
;
3595 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
3596 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
3597 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
3599 bctl
->flags
= bargs
->flags
;
3601 /* balance everything - no filters */
3602 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
3607 * Ownership of bctl and mutually_exclusive_operation_running
3608 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3609 * or, if restriper was paused all the way until unmount, in
3610 * free_fs_info. mutually_exclusive_operation_running is
3611 * cleared in __cancel_balance.
3613 need_unlock
= false;
3615 ret
= btrfs_balance(bctl
, bargs
);
3618 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3625 mutex_unlock(&fs_info
->balance_mutex
);
3626 mutex_unlock(&fs_info
->volume_mutex
);
3628 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
3630 mnt_drop_write_file(file
);
3634 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
3636 if (!capable(CAP_SYS_ADMIN
))
3640 case BTRFS_BALANCE_CTL_PAUSE
:
3641 return btrfs_pause_balance(root
->fs_info
);
3642 case BTRFS_BALANCE_CTL_CANCEL
:
3643 return btrfs_cancel_balance(root
->fs_info
);
3649 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
3652 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3653 struct btrfs_ioctl_balance_args
*bargs
;
3656 if (!capable(CAP_SYS_ADMIN
))
3659 mutex_lock(&fs_info
->balance_mutex
);
3660 if (!fs_info
->balance_ctl
) {
3665 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
3671 update_ioctl_balance_args(fs_info
, 1, bargs
);
3673 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3678 mutex_unlock(&fs_info
->balance_mutex
);
3682 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
3684 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3685 struct btrfs_ioctl_quota_ctl_args
*sa
;
3686 struct btrfs_trans_handle
*trans
= NULL
;
3690 if (!capable(CAP_SYS_ADMIN
))
3693 ret
= mnt_want_write_file(file
);
3697 sa
= memdup_user(arg
, sizeof(*sa
));
3703 down_write(&root
->fs_info
->subvol_sem
);
3704 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
3705 if (IS_ERR(trans
)) {
3706 ret
= PTR_ERR(trans
);
3711 case BTRFS_QUOTA_CTL_ENABLE
:
3712 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
3714 case BTRFS_QUOTA_CTL_DISABLE
:
3715 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
3722 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3725 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
3730 up_write(&root
->fs_info
->subvol_sem
);
3732 mnt_drop_write_file(file
);
3736 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
3738 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3739 struct btrfs_ioctl_qgroup_assign_args
*sa
;
3740 struct btrfs_trans_handle
*trans
;
3744 if (!capable(CAP_SYS_ADMIN
))
3747 ret
= mnt_want_write_file(file
);
3751 sa
= memdup_user(arg
, sizeof(*sa
));
3757 trans
= btrfs_join_transaction(root
);
3758 if (IS_ERR(trans
)) {
3759 ret
= PTR_ERR(trans
);
3763 /* FIXME: check if the IDs really exist */
3765 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
3768 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
3772 err
= btrfs_end_transaction(trans
, root
);
3779 mnt_drop_write_file(file
);
3783 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
3785 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3786 struct btrfs_ioctl_qgroup_create_args
*sa
;
3787 struct btrfs_trans_handle
*trans
;
3791 if (!capable(CAP_SYS_ADMIN
))
3794 ret
= mnt_want_write_file(file
);
3798 sa
= memdup_user(arg
, sizeof(*sa
));
3804 if (!sa
->qgroupid
) {
3809 trans
= btrfs_join_transaction(root
);
3810 if (IS_ERR(trans
)) {
3811 ret
= PTR_ERR(trans
);
3815 /* FIXME: check if the IDs really exist */
3817 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
,
3820 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
3823 err
= btrfs_end_transaction(trans
, root
);
3830 mnt_drop_write_file(file
);
3834 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
3836 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3837 struct btrfs_ioctl_qgroup_limit_args
*sa
;
3838 struct btrfs_trans_handle
*trans
;
3843 if (!capable(CAP_SYS_ADMIN
))
3846 ret
= mnt_want_write_file(file
);
3850 sa
= memdup_user(arg
, sizeof(*sa
));
3856 trans
= btrfs_join_transaction(root
);
3857 if (IS_ERR(trans
)) {
3858 ret
= PTR_ERR(trans
);
3862 qgroupid
= sa
->qgroupid
;
3864 /* take the current subvol as qgroup */
3865 qgroupid
= root
->root_key
.objectid
;
3868 /* FIXME: check if the IDs really exist */
3869 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
3871 err
= btrfs_end_transaction(trans
, root
);
3878 mnt_drop_write_file(file
);
3882 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
3884 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3885 struct btrfs_ioctl_quota_rescan_args
*qsa
;
3888 if (!capable(CAP_SYS_ADMIN
))
3891 ret
= mnt_want_write_file(file
);
3895 qsa
= memdup_user(arg
, sizeof(*qsa
));
3906 ret
= btrfs_qgroup_rescan(root
->fs_info
);
3911 mnt_drop_write_file(file
);
3915 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
3917 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3918 struct btrfs_ioctl_quota_rescan_args
*qsa
;
3921 if (!capable(CAP_SYS_ADMIN
))
3924 qsa
= kzalloc(sizeof(*qsa
), GFP_NOFS
);
3928 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
3930 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
3933 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
3940 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
3943 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
3944 struct inode
*inode
= file_inode(file
);
3945 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3946 struct btrfs_root_item
*root_item
= &root
->root_item
;
3947 struct btrfs_trans_handle
*trans
;
3948 struct timespec ct
= CURRENT_TIME
;
3951 ret
= mnt_want_write_file(file
);
3955 down_write(&root
->fs_info
->subvol_sem
);
3957 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
3962 if (btrfs_root_readonly(root
)) {
3967 if (!inode_owner_or_capable(inode
)) {
3972 sa
= memdup_user(arg
, sizeof(*sa
));
3979 trans
= btrfs_start_transaction(root
, 1);
3980 if (IS_ERR(trans
)) {
3981 ret
= PTR_ERR(trans
);
3986 sa
->rtransid
= trans
->transid
;
3987 sa
->rtime
.sec
= ct
.tv_sec
;
3988 sa
->rtime
.nsec
= ct
.tv_nsec
;
3990 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
3991 btrfs_set_root_stransid(root_item
, sa
->stransid
);
3992 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
3993 root_item
->stime
.sec
= cpu_to_le64(sa
->stime
.sec
);
3994 root_item
->stime
.nsec
= cpu_to_le32(sa
->stime
.nsec
);
3995 root_item
->rtime
.sec
= cpu_to_le64(sa
->rtime
.sec
);
3996 root_item
->rtime
.nsec
= cpu_to_le32(sa
->rtime
.nsec
);
3998 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
3999 &root
->root_key
, &root
->root_item
);
4001 btrfs_end_transaction(trans
, root
);
4005 ret
= btrfs_commit_transaction(trans
, root
);
4010 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
4016 up_write(&root
->fs_info
->subvol_sem
);
4017 mnt_drop_write_file(file
);
4021 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
4023 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
4024 const char *label
= root
->fs_info
->super_copy
->label
;
4025 size_t len
= strnlen(label
, BTRFS_LABEL_SIZE
);
4028 if (len
== BTRFS_LABEL_SIZE
) {
4029 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4033 mutex_lock(&root
->fs_info
->volume_mutex
);
4034 ret
= copy_to_user(arg
, label
, len
);
4035 mutex_unlock(&root
->fs_info
->volume_mutex
);
4037 return ret
? -EFAULT
: 0;
4040 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
4042 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
4043 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
4044 struct btrfs_trans_handle
*trans
;
4045 char label
[BTRFS_LABEL_SIZE
];
4048 if (!capable(CAP_SYS_ADMIN
))
4051 if (copy_from_user(label
, arg
, sizeof(label
)))
4054 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
4055 pr_err("btrfs: unable to set label with more than %d bytes\n",
4056 BTRFS_LABEL_SIZE
- 1);
4060 ret
= mnt_want_write_file(file
);
4064 mutex_lock(&root
->fs_info
->volume_mutex
);
4065 trans
= btrfs_start_transaction(root
, 0);
4066 if (IS_ERR(trans
)) {
4067 ret
= PTR_ERR(trans
);
4071 strcpy(super_block
->label
, label
);
4072 ret
= btrfs_end_transaction(trans
, root
);
4075 mutex_unlock(&root
->fs_info
->volume_mutex
);
4076 mnt_drop_write_file(file
);
4080 long btrfs_ioctl(struct file
*file
, unsigned int
4081 cmd
, unsigned long arg
)
4083 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4084 void __user
*argp
= (void __user
*)arg
;
4087 case FS_IOC_GETFLAGS
:
4088 return btrfs_ioctl_getflags(file
, argp
);
4089 case FS_IOC_SETFLAGS
:
4090 return btrfs_ioctl_setflags(file
, argp
);
4091 case FS_IOC_GETVERSION
:
4092 return btrfs_ioctl_getversion(file
, argp
);
4094 return btrfs_ioctl_fitrim(file
, argp
);
4095 case BTRFS_IOC_SNAP_CREATE
:
4096 return btrfs_ioctl_snap_create(file
, argp
, 0);
4097 case BTRFS_IOC_SNAP_CREATE_V2
:
4098 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
4099 case BTRFS_IOC_SUBVOL_CREATE
:
4100 return btrfs_ioctl_snap_create(file
, argp
, 1);
4101 case BTRFS_IOC_SUBVOL_CREATE_V2
:
4102 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
4103 case BTRFS_IOC_SNAP_DESTROY
:
4104 return btrfs_ioctl_snap_destroy(file
, argp
);
4105 case BTRFS_IOC_SUBVOL_GETFLAGS
:
4106 return btrfs_ioctl_subvol_getflags(file
, argp
);
4107 case BTRFS_IOC_SUBVOL_SETFLAGS
:
4108 return btrfs_ioctl_subvol_setflags(file
, argp
);
4109 case BTRFS_IOC_DEFAULT_SUBVOL
:
4110 return btrfs_ioctl_default_subvol(file
, argp
);
4111 case BTRFS_IOC_DEFRAG
:
4112 return btrfs_ioctl_defrag(file
, NULL
);
4113 case BTRFS_IOC_DEFRAG_RANGE
:
4114 return btrfs_ioctl_defrag(file
, argp
);
4115 case BTRFS_IOC_RESIZE
:
4116 return btrfs_ioctl_resize(file
, argp
);
4117 case BTRFS_IOC_ADD_DEV
:
4118 return btrfs_ioctl_add_dev(root
, argp
);
4119 case BTRFS_IOC_RM_DEV
:
4120 return btrfs_ioctl_rm_dev(file
, argp
);
4121 case BTRFS_IOC_FS_INFO
:
4122 return btrfs_ioctl_fs_info(root
, argp
);
4123 case BTRFS_IOC_DEV_INFO
:
4124 return btrfs_ioctl_dev_info(root
, argp
);
4125 case BTRFS_IOC_BALANCE
:
4126 return btrfs_ioctl_balance(file
, NULL
);
4127 case BTRFS_IOC_CLONE
:
4128 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
4129 case BTRFS_IOC_CLONE_RANGE
:
4130 return btrfs_ioctl_clone_range(file
, argp
);
4131 case BTRFS_IOC_TRANS_START
:
4132 return btrfs_ioctl_trans_start(file
);
4133 case BTRFS_IOC_TRANS_END
:
4134 return btrfs_ioctl_trans_end(file
);
4135 case BTRFS_IOC_TREE_SEARCH
:
4136 return btrfs_ioctl_tree_search(file
, argp
);
4137 case BTRFS_IOC_INO_LOOKUP
:
4138 return btrfs_ioctl_ino_lookup(file
, argp
);
4139 case BTRFS_IOC_INO_PATHS
:
4140 return btrfs_ioctl_ino_to_path(root
, argp
);
4141 case BTRFS_IOC_LOGICAL_INO
:
4142 return btrfs_ioctl_logical_to_ino(root
, argp
);
4143 case BTRFS_IOC_SPACE_INFO
:
4144 return btrfs_ioctl_space_info(root
, argp
);
4145 case BTRFS_IOC_SYNC
:
4146 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
4148 case BTRFS_IOC_START_SYNC
:
4149 return btrfs_ioctl_start_sync(root
, argp
);
4150 case BTRFS_IOC_WAIT_SYNC
:
4151 return btrfs_ioctl_wait_sync(root
, argp
);
4152 case BTRFS_IOC_SCRUB
:
4153 return btrfs_ioctl_scrub(file
, argp
);
4154 case BTRFS_IOC_SCRUB_CANCEL
:
4155 return btrfs_ioctl_scrub_cancel(root
, argp
);
4156 case BTRFS_IOC_SCRUB_PROGRESS
:
4157 return btrfs_ioctl_scrub_progress(root
, argp
);
4158 case BTRFS_IOC_BALANCE_V2
:
4159 return btrfs_ioctl_balance(file
, argp
);
4160 case BTRFS_IOC_BALANCE_CTL
:
4161 return btrfs_ioctl_balance_ctl(root
, arg
);
4162 case BTRFS_IOC_BALANCE_PROGRESS
:
4163 return btrfs_ioctl_balance_progress(root
, argp
);
4164 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
4165 return btrfs_ioctl_set_received_subvol(file
, argp
);
4166 case BTRFS_IOC_SEND
:
4167 return btrfs_ioctl_send(file
, argp
);
4168 case BTRFS_IOC_GET_DEV_STATS
:
4169 return btrfs_ioctl_get_dev_stats(root
, argp
);
4170 case BTRFS_IOC_QUOTA_CTL
:
4171 return btrfs_ioctl_quota_ctl(file
, argp
);
4172 case BTRFS_IOC_QGROUP_ASSIGN
:
4173 return btrfs_ioctl_qgroup_assign(file
, argp
);
4174 case BTRFS_IOC_QGROUP_CREATE
:
4175 return btrfs_ioctl_qgroup_create(file
, argp
);
4176 case BTRFS_IOC_QGROUP_LIMIT
:
4177 return btrfs_ioctl_qgroup_limit(file
, argp
);
4178 case BTRFS_IOC_QUOTA_RESCAN
:
4179 return btrfs_ioctl_quota_rescan(file
, argp
);
4180 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
4181 return btrfs_ioctl_quota_rescan_status(file
, argp
);
4182 case BTRFS_IOC_DEV_REPLACE
:
4183 return btrfs_ioctl_dev_replace(root
, argp
);
4184 case BTRFS_IOC_GET_FSLABEL
:
4185 return btrfs_ioctl_get_fslabel(file
, argp
);
4186 case BTRFS_IOC_SET_FSLABEL
:
4187 return btrfs_ioctl_set_fslabel(file
, argp
);