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
, false);
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 ret
= btrfs_start_delalloc_inodes(root
, 0);
562 btrfs_wait_ordered_extents(root
, 0);
564 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
565 if (!pending_snapshot
)
568 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
569 BTRFS_BLOCK_RSV_TEMP
);
571 * 1 - parent dir inode
574 * 2 - root ref/backref
575 * 1 - root of snapshot
577 ret
= btrfs_subvolume_reserve_metadata(BTRFS_I(dir
)->root
,
578 &pending_snapshot
->block_rsv
, 7,
579 &pending_snapshot
->qgroup_reserved
,
584 pending_snapshot
->dentry
= dentry
;
585 pending_snapshot
->root
= root
;
586 pending_snapshot
->readonly
= readonly
;
587 pending_snapshot
->dir
= dir
;
588 pending_snapshot
->inherit
= inherit
;
590 trans
= btrfs_start_transaction(root
, 0);
592 ret
= PTR_ERR(trans
);
596 spin_lock(&root
->fs_info
->trans_lock
);
597 list_add(&pending_snapshot
->list
,
598 &trans
->transaction
->pending_snapshots
);
599 spin_unlock(&root
->fs_info
->trans_lock
);
601 *async_transid
= trans
->transid
;
602 ret
= btrfs_commit_transaction_async(trans
,
603 root
->fs_info
->extent_root
, 1);
605 ret
= btrfs_commit_transaction(trans
, root
);
607 ret
= btrfs_commit_transaction(trans
,
608 root
->fs_info
->extent_root
);
613 ret
= pending_snapshot
->error
;
617 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
621 inode
= btrfs_lookup_dentry(dentry
->d_parent
->d_inode
, dentry
);
623 ret
= PTR_ERR(inode
);
627 d_instantiate(dentry
, inode
);
630 btrfs_subvolume_release_metadata(BTRFS_I(dir
)->root
,
631 &pending_snapshot
->block_rsv
,
632 pending_snapshot
->qgroup_reserved
);
634 kfree(pending_snapshot
);
638 /* copy of check_sticky in fs/namei.c()
639 * It's inline, so penalty for filesystems that don't use sticky bit is
642 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
644 kuid_t fsuid
= current_fsuid();
646 if (!(dir
->i_mode
& S_ISVTX
))
648 if (uid_eq(inode
->i_uid
, fsuid
))
650 if (uid_eq(dir
->i_uid
, fsuid
))
652 return !capable(CAP_FOWNER
);
655 /* copy of may_delete in fs/namei.c()
656 * Check whether we can remove a link victim from directory dir, check
657 * whether the type of victim is right.
658 * 1. We can't do it if dir is read-only (done in permission())
659 * 2. We should have write and exec permissions on dir
660 * 3. We can't remove anything from append-only dir
661 * 4. We can't do anything with immutable dir (done in permission())
662 * 5. If the sticky bit on dir is set we should either
663 * a. be owner of dir, or
664 * b. be owner of victim, or
665 * c. have CAP_FOWNER capability
666 * 6. If the victim is append-only or immutable we can't do antyhing with
667 * links pointing to it.
668 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
669 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
670 * 9. We can't remove a root or mountpoint.
671 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
672 * nfs_async_unlink().
675 static int btrfs_may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
679 if (!victim
->d_inode
)
682 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
683 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
685 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
690 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
691 IS_APPEND(victim
->d_inode
)||
692 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
695 if (!S_ISDIR(victim
->d_inode
->i_mode
))
699 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
703 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
708 /* copy of may_create in fs/namei.c() */
709 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
715 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
719 * Create a new subvolume below @parent. This is largely modeled after
720 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
721 * inside this filesystem so it's quite a bit simpler.
723 static noinline
int btrfs_mksubvol(struct path
*parent
,
724 char *name
, int namelen
,
725 struct btrfs_root
*snap_src
,
726 u64
*async_transid
, bool readonly
,
727 struct btrfs_qgroup_inherit
*inherit
)
729 struct inode
*dir
= parent
->dentry
->d_inode
;
730 struct dentry
*dentry
;
733 error
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
737 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
738 error
= PTR_ERR(dentry
);
746 error
= btrfs_may_create(dir
, dentry
);
751 * even if this name doesn't exist, we may get hash collisions.
752 * check for them now when we can safely fail
754 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
760 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
762 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
766 error
= create_snapshot(snap_src
, dir
, dentry
, name
, namelen
,
767 async_transid
, readonly
, inherit
);
769 error
= create_subvol(dir
, dentry
, name
, namelen
,
770 async_transid
, inherit
);
773 fsnotify_mkdir(dir
, dentry
);
775 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
779 mutex_unlock(&dir
->i_mutex
);
784 * When we're defragging a range, we don't want to kick it off again
785 * if it is really just waiting for delalloc to send it down.
786 * If we find a nice big extent or delalloc range for the bytes in the
787 * file you want to defrag, we return 0 to let you know to skip this
790 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, int thresh
)
792 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
793 struct extent_map
*em
= NULL
;
794 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
797 read_lock(&em_tree
->lock
);
798 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
799 read_unlock(&em_tree
->lock
);
802 end
= extent_map_end(em
);
804 if (end
- offset
> thresh
)
807 /* if we already have a nice delalloc here, just stop */
809 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
810 thresh
, EXTENT_DELALLOC
, 1);
817 * helper function to walk through a file and find extents
818 * newer than a specific transid, and smaller than thresh.
820 * This is used by the defragging code to find new and small
823 static int find_new_extents(struct btrfs_root
*root
,
824 struct inode
*inode
, u64 newer_than
,
825 u64
*off
, int thresh
)
827 struct btrfs_path
*path
;
828 struct btrfs_key min_key
;
829 struct btrfs_key max_key
;
830 struct extent_buffer
*leaf
;
831 struct btrfs_file_extent_item
*extent
;
834 u64 ino
= btrfs_ino(inode
);
836 path
= btrfs_alloc_path();
840 min_key
.objectid
= ino
;
841 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
842 min_key
.offset
= *off
;
844 max_key
.objectid
= ino
;
845 max_key
.type
= (u8
)-1;
846 max_key
.offset
= (u64
)-1;
848 path
->keep_locks
= 1;
851 ret
= btrfs_search_forward(root
, &min_key
, &max_key
,
855 if (min_key
.objectid
!= ino
)
857 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
860 leaf
= path
->nodes
[0];
861 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
862 struct btrfs_file_extent_item
);
864 type
= btrfs_file_extent_type(leaf
, extent
);
865 if (type
== BTRFS_FILE_EXTENT_REG
&&
866 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
867 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
868 *off
= min_key
.offset
;
869 btrfs_free_path(path
);
873 if (min_key
.offset
== (u64
)-1)
877 btrfs_release_path(path
);
880 btrfs_free_path(path
);
884 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
886 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
887 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
888 struct extent_map
*em
;
889 u64 len
= PAGE_CACHE_SIZE
;
892 * hopefully we have this extent in the tree already, try without
893 * the full extent lock
895 read_lock(&em_tree
->lock
);
896 em
= lookup_extent_mapping(em_tree
, start
, len
);
897 read_unlock(&em_tree
->lock
);
900 /* get the big lock and read metadata off disk */
901 lock_extent(io_tree
, start
, start
+ len
- 1);
902 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
903 unlock_extent(io_tree
, start
, start
+ len
- 1);
912 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
914 struct extent_map
*next
;
917 /* this is the last extent */
918 if (em
->start
+ em
->len
>= i_size_read(inode
))
921 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
922 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
925 free_extent_map(next
);
929 static int should_defrag_range(struct inode
*inode
, u64 start
, int thresh
,
930 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
933 struct extent_map
*em
;
935 bool next_mergeable
= true;
938 * make sure that once we start defragging an extent, we keep on
941 if (start
< *defrag_end
)
946 em
= defrag_lookup_extent(inode
, start
);
950 /* this will cover holes, and inline extents */
951 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
956 next_mergeable
= defrag_check_next_extent(inode
, em
);
959 * we hit a real extent, if it is big or the next extent is not a
960 * real extent, don't bother defragging it
962 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
963 (em
->len
>= thresh
|| !next_mergeable
))
967 * last_len ends up being a counter of how many bytes we've defragged.
968 * every time we choose not to defrag an extent, we reset *last_len
969 * so that the next tiny extent will force a defrag.
971 * The end result of this is that tiny extents before a single big
972 * extent will force at least part of that big extent to be defragged.
975 *defrag_end
= extent_map_end(em
);
978 *skip
= extent_map_end(em
);
987 * it doesn't do much good to defrag one or two pages
988 * at a time. This pulls in a nice chunk of pages
991 * It also makes sure the delalloc code has enough
992 * dirty data to avoid making new small extents as part
995 * It's a good idea to start RA on this range
996 * before calling this.
998 static int cluster_pages_for_defrag(struct inode
*inode
,
1000 unsigned long start_index
,
1003 unsigned long file_end
;
1004 u64 isize
= i_size_read(inode
);
1011 struct btrfs_ordered_extent
*ordered
;
1012 struct extent_state
*cached_state
= NULL
;
1013 struct extent_io_tree
*tree
;
1014 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
1016 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1017 if (!isize
|| start_index
> file_end
)
1020 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
1022 ret
= btrfs_delalloc_reserve_space(inode
,
1023 page_cnt
<< PAGE_CACHE_SHIFT
);
1027 tree
= &BTRFS_I(inode
)->io_tree
;
1029 /* step one, lock all the pages */
1030 for (i
= 0; i
< page_cnt
; i
++) {
1033 page
= find_or_create_page(inode
->i_mapping
,
1034 start_index
+ i
, mask
);
1038 page_start
= page_offset(page
);
1039 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1041 lock_extent(tree
, page_start
, page_end
);
1042 ordered
= btrfs_lookup_ordered_extent(inode
,
1044 unlock_extent(tree
, page_start
, page_end
);
1049 btrfs_start_ordered_extent(inode
, ordered
, 1);
1050 btrfs_put_ordered_extent(ordered
);
1053 * we unlocked the page above, so we need check if
1054 * it was released or not.
1056 if (page
->mapping
!= inode
->i_mapping
) {
1058 page_cache_release(page
);
1063 if (!PageUptodate(page
)) {
1064 btrfs_readpage(NULL
, page
);
1066 if (!PageUptodate(page
)) {
1068 page_cache_release(page
);
1074 if (page
->mapping
!= inode
->i_mapping
) {
1076 page_cache_release(page
);
1086 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1090 * so now we have a nice long stream of locked
1091 * and up to date pages, lets wait on them
1093 for (i
= 0; i
< i_done
; i
++)
1094 wait_on_page_writeback(pages
[i
]);
1096 page_start
= page_offset(pages
[0]);
1097 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1099 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1100 page_start
, page_end
- 1, 0, &cached_state
);
1101 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1102 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1103 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1104 &cached_state
, GFP_NOFS
);
1106 if (i_done
!= page_cnt
) {
1107 spin_lock(&BTRFS_I(inode
)->lock
);
1108 BTRFS_I(inode
)->outstanding_extents
++;
1109 spin_unlock(&BTRFS_I(inode
)->lock
);
1110 btrfs_delalloc_release_space(inode
,
1111 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1115 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1116 &cached_state
, GFP_NOFS
);
1118 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1119 page_start
, page_end
- 1, &cached_state
,
1122 for (i
= 0; i
< i_done
; i
++) {
1123 clear_page_dirty_for_io(pages
[i
]);
1124 ClearPageChecked(pages
[i
]);
1125 set_page_extent_mapped(pages
[i
]);
1126 set_page_dirty(pages
[i
]);
1127 unlock_page(pages
[i
]);
1128 page_cache_release(pages
[i
]);
1132 for (i
= 0; i
< i_done
; i
++) {
1133 unlock_page(pages
[i
]);
1134 page_cache_release(pages
[i
]);
1136 btrfs_delalloc_release_space(inode
, page_cnt
<< PAGE_CACHE_SHIFT
);
1141 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1142 struct btrfs_ioctl_defrag_range_args
*range
,
1143 u64 newer_than
, unsigned long max_to_defrag
)
1145 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1146 struct file_ra_state
*ra
= NULL
;
1147 unsigned long last_index
;
1148 u64 isize
= i_size_read(inode
);
1152 u64 newer_off
= range
->start
;
1154 unsigned long ra_index
= 0;
1156 int defrag_count
= 0;
1157 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1158 int extent_thresh
= range
->extent_thresh
;
1159 int max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
1160 int cluster
= max_cluster
;
1161 u64 new_align
= ~((u64
)128 * 1024 - 1);
1162 struct page
**pages
= NULL
;
1167 if (range
->start
>= isize
)
1170 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1171 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1173 if (range
->compress_type
)
1174 compress_type
= range
->compress_type
;
1177 if (extent_thresh
== 0)
1178 extent_thresh
= 256 * 1024;
1181 * if we were not given a file, allocate a readahead
1185 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1188 file_ra_state_init(ra
, inode
->i_mapping
);
1193 pages
= kmalloc(sizeof(struct page
*) * max_cluster
,
1200 /* find the last page to defrag */
1201 if (range
->start
+ range
->len
> range
->start
) {
1202 last_index
= min_t(u64
, isize
- 1,
1203 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1205 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1209 ret
= find_new_extents(root
, inode
, newer_than
,
1210 &newer_off
, 64 * 1024);
1212 range
->start
= newer_off
;
1214 * we always align our defrag to help keep
1215 * the extents in the file evenly spaced
1217 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1221 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1224 max_to_defrag
= last_index
+ 1;
1227 * make writeback starts from i, so the defrag range can be
1228 * written sequentially.
1230 if (i
< inode
->i_mapping
->writeback_index
)
1231 inode
->i_mapping
->writeback_index
= i
;
1233 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1234 (i
< (i_size_read(inode
) + PAGE_CACHE_SIZE
- 1) >>
1235 PAGE_CACHE_SHIFT
)) {
1237 * make sure we stop running if someone unmounts
1240 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1243 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1244 printk(KERN_DEBUG
"btrfs: defrag_file cancelled\n");
1249 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1250 extent_thresh
, &last_len
, &skip
,
1251 &defrag_end
, range
->flags
&
1252 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1255 * the should_defrag function tells us how much to skip
1256 * bump our counter by the suggested amount
1258 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
1259 i
= max(i
+ 1, next
);
1264 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1265 PAGE_CACHE_SHIFT
) - i
;
1266 cluster
= min(cluster
, max_cluster
);
1268 cluster
= max_cluster
;
1271 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1272 BTRFS_I(inode
)->force_compress
= compress_type
;
1274 if (i
+ cluster
> ra_index
) {
1275 ra_index
= max(i
, ra_index
);
1276 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1278 ra_index
+= max_cluster
;
1281 mutex_lock(&inode
->i_mutex
);
1282 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1284 mutex_unlock(&inode
->i_mutex
);
1288 defrag_count
+= ret
;
1289 balance_dirty_pages_ratelimited(inode
->i_mapping
);
1290 mutex_unlock(&inode
->i_mutex
);
1293 if (newer_off
== (u64
)-1)
1299 newer_off
= max(newer_off
+ 1,
1300 (u64
)i
<< PAGE_CACHE_SHIFT
);
1302 ret
= find_new_extents(root
, inode
,
1303 newer_than
, &newer_off
,
1306 range
->start
= newer_off
;
1307 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1314 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1322 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
))
1323 filemap_flush(inode
->i_mapping
);
1325 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1326 /* the filemap_flush will queue IO into the worker threads, but
1327 * we have to make sure the IO is actually started and that
1328 * ordered extents get created before we return
1330 atomic_inc(&root
->fs_info
->async_submit_draining
);
1331 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1332 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1333 wait_event(root
->fs_info
->async_submit_wait
,
1334 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1335 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1337 atomic_dec(&root
->fs_info
->async_submit_draining
);
1339 mutex_lock(&inode
->i_mutex
);
1340 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1341 mutex_unlock(&inode
->i_mutex
);
1344 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1345 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1357 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1363 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
1364 struct btrfs_ioctl_vol_args
*vol_args
;
1365 struct btrfs_trans_handle
*trans
;
1366 struct btrfs_device
*device
= NULL
;
1368 char *devstr
= NULL
;
1372 if (!capable(CAP_SYS_ADMIN
))
1375 ret
= mnt_want_write_file(file
);
1379 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1381 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1382 mnt_drop_write_file(file
);
1386 mutex_lock(&root
->fs_info
->volume_mutex
);
1387 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1388 if (IS_ERR(vol_args
)) {
1389 ret
= PTR_ERR(vol_args
);
1393 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1395 sizestr
= vol_args
->name
;
1396 devstr
= strchr(sizestr
, ':');
1399 sizestr
= devstr
+ 1;
1401 devstr
= vol_args
->name
;
1402 devid
= simple_strtoull(devstr
, &end
, 10);
1407 printk(KERN_INFO
"btrfs: resizing devid %llu\n",
1408 (unsigned long long)devid
);
1411 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1413 printk(KERN_INFO
"btrfs: resizer unable to find device %llu\n",
1414 (unsigned long long)devid
);
1419 if (!device
->writeable
) {
1420 printk(KERN_INFO
"btrfs: resizer unable to apply on "
1421 "readonly device %llu\n",
1422 (unsigned long long)devid
);
1427 if (!strcmp(sizestr
, "max"))
1428 new_size
= device
->bdev
->bd_inode
->i_size
;
1430 if (sizestr
[0] == '-') {
1433 } else if (sizestr
[0] == '+') {
1437 new_size
= memparse(sizestr
, NULL
);
1438 if (new_size
== 0) {
1444 if (device
->is_tgtdev_for_dev_replace
) {
1449 old_size
= device
->total_bytes
;
1452 if (new_size
> old_size
) {
1456 new_size
= old_size
- new_size
;
1457 } else if (mod
> 0) {
1458 new_size
= old_size
+ new_size
;
1461 if (new_size
< 256 * 1024 * 1024) {
1465 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1470 do_div(new_size
, root
->sectorsize
);
1471 new_size
*= root
->sectorsize
;
1473 printk_in_rcu(KERN_INFO
"btrfs: new size for %s is %llu\n",
1474 rcu_str_deref(device
->name
),
1475 (unsigned long long)new_size
);
1477 if (new_size
> old_size
) {
1478 trans
= btrfs_start_transaction(root
, 0);
1479 if (IS_ERR(trans
)) {
1480 ret
= PTR_ERR(trans
);
1483 ret
= btrfs_grow_device(trans
, device
, new_size
);
1484 btrfs_commit_transaction(trans
, root
);
1485 } else if (new_size
< old_size
) {
1486 ret
= btrfs_shrink_device(device
, new_size
);
1487 } /* equal, nothing need to do */
1492 mutex_unlock(&root
->fs_info
->volume_mutex
);
1493 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1494 mnt_drop_write_file(file
);
1498 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1499 char *name
, unsigned long fd
, int subvol
,
1500 u64
*transid
, bool readonly
,
1501 struct btrfs_qgroup_inherit
*inherit
)
1506 ret
= mnt_want_write_file(file
);
1510 namelen
= strlen(name
);
1511 if (strchr(name
, '/')) {
1513 goto out_drop_write
;
1516 if (name
[0] == '.' &&
1517 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1519 goto out_drop_write
;
1523 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1524 NULL
, transid
, readonly
, inherit
);
1526 struct fd src
= fdget(fd
);
1527 struct inode
*src_inode
;
1530 goto out_drop_write
;
1533 src_inode
= file_inode(src
.file
);
1534 if (src_inode
->i_sb
!= file_inode(file
)->i_sb
) {
1535 printk(KERN_INFO
"btrfs: Snapshot src from "
1539 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1540 BTRFS_I(src_inode
)->root
,
1541 transid
, readonly
, inherit
);
1546 mnt_drop_write_file(file
);
1551 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1552 void __user
*arg
, int subvol
)
1554 struct btrfs_ioctl_vol_args
*vol_args
;
1557 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1558 if (IS_ERR(vol_args
))
1559 return PTR_ERR(vol_args
);
1560 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1562 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1563 vol_args
->fd
, subvol
,
1570 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1571 void __user
*arg
, int subvol
)
1573 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1577 bool readonly
= false;
1578 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1580 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1581 if (IS_ERR(vol_args
))
1582 return PTR_ERR(vol_args
);
1583 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1585 if (vol_args
->flags
&
1586 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1587 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1592 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1594 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1596 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1597 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1601 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1602 if (IS_ERR(inherit
)) {
1603 ret
= PTR_ERR(inherit
);
1608 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1609 vol_args
->fd
, subvol
, ptr
,
1612 if (ret
== 0 && ptr
&&
1614 offsetof(struct btrfs_ioctl_vol_args_v2
,
1615 transid
), ptr
, sizeof(*ptr
)))
1623 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1626 struct inode
*inode
= file_inode(file
);
1627 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1631 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1634 down_read(&root
->fs_info
->subvol_sem
);
1635 if (btrfs_root_readonly(root
))
1636 flags
|= BTRFS_SUBVOL_RDONLY
;
1637 up_read(&root
->fs_info
->subvol_sem
);
1639 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1645 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1648 struct inode
*inode
= file_inode(file
);
1649 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1650 struct btrfs_trans_handle
*trans
;
1655 ret
= mnt_want_write_file(file
);
1659 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1661 goto out_drop_write
;
1664 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1666 goto out_drop_write
;
1669 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1671 goto out_drop_write
;
1674 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1676 goto out_drop_write
;
1679 if (!inode_owner_or_capable(inode
)) {
1681 goto out_drop_write
;
1684 down_write(&root
->fs_info
->subvol_sem
);
1687 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1690 root_flags
= btrfs_root_flags(&root
->root_item
);
1691 if (flags
& BTRFS_SUBVOL_RDONLY
)
1692 btrfs_set_root_flags(&root
->root_item
,
1693 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1695 btrfs_set_root_flags(&root
->root_item
,
1696 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1698 trans
= btrfs_start_transaction(root
, 1);
1699 if (IS_ERR(trans
)) {
1700 ret
= PTR_ERR(trans
);
1704 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1705 &root
->root_key
, &root
->root_item
);
1707 btrfs_commit_transaction(trans
, root
);
1710 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1712 up_write(&root
->fs_info
->subvol_sem
);
1714 mnt_drop_write_file(file
);
1720 * helper to check if the subvolume references other subvolumes
1722 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1724 struct btrfs_path
*path
;
1725 struct btrfs_key key
;
1728 path
= btrfs_alloc_path();
1732 key
.objectid
= root
->root_key
.objectid
;
1733 key
.type
= BTRFS_ROOT_REF_KEY
;
1734 key
.offset
= (u64
)-1;
1736 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1743 if (path
->slots
[0] > 0) {
1745 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1746 if (key
.objectid
== root
->root_key
.objectid
&&
1747 key
.type
== BTRFS_ROOT_REF_KEY
)
1751 btrfs_free_path(path
);
1755 static noinline
int key_in_sk(struct btrfs_key
*key
,
1756 struct btrfs_ioctl_search_key
*sk
)
1758 struct btrfs_key test
;
1761 test
.objectid
= sk
->min_objectid
;
1762 test
.type
= sk
->min_type
;
1763 test
.offset
= sk
->min_offset
;
1765 ret
= btrfs_comp_cpu_keys(key
, &test
);
1769 test
.objectid
= sk
->max_objectid
;
1770 test
.type
= sk
->max_type
;
1771 test
.offset
= sk
->max_offset
;
1773 ret
= btrfs_comp_cpu_keys(key
, &test
);
1779 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1780 struct btrfs_path
*path
,
1781 struct btrfs_key
*key
,
1782 struct btrfs_ioctl_search_key
*sk
,
1784 unsigned long *sk_offset
,
1788 struct extent_buffer
*leaf
;
1789 struct btrfs_ioctl_search_header sh
;
1790 unsigned long item_off
;
1791 unsigned long item_len
;
1797 leaf
= path
->nodes
[0];
1798 slot
= path
->slots
[0];
1799 nritems
= btrfs_header_nritems(leaf
);
1801 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1805 found_transid
= btrfs_header_generation(leaf
);
1807 for (i
= slot
; i
< nritems
; i
++) {
1808 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1809 item_len
= btrfs_item_size_nr(leaf
, i
);
1811 btrfs_item_key_to_cpu(leaf
, key
, i
);
1812 if (!key_in_sk(key
, sk
))
1815 if (sizeof(sh
) + item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1818 if (sizeof(sh
) + item_len
+ *sk_offset
>
1819 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1824 sh
.objectid
= key
->objectid
;
1825 sh
.offset
= key
->offset
;
1826 sh
.type
= key
->type
;
1828 sh
.transid
= found_transid
;
1830 /* copy search result header */
1831 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1832 *sk_offset
+= sizeof(sh
);
1835 char *p
= buf
+ *sk_offset
;
1837 read_extent_buffer(leaf
, p
,
1838 item_off
, item_len
);
1839 *sk_offset
+= item_len
;
1843 if (*num_found
>= sk
->nr_items
)
1848 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1850 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1853 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
1863 static noinline
int search_ioctl(struct inode
*inode
,
1864 struct btrfs_ioctl_search_args
*args
)
1866 struct btrfs_root
*root
;
1867 struct btrfs_key key
;
1868 struct btrfs_key max_key
;
1869 struct btrfs_path
*path
;
1870 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
1871 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
1874 unsigned long sk_offset
= 0;
1876 path
= btrfs_alloc_path();
1880 if (sk
->tree_id
== 0) {
1881 /* search the root of the inode that was passed */
1882 root
= BTRFS_I(inode
)->root
;
1884 key
.objectid
= sk
->tree_id
;
1885 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1886 key
.offset
= (u64
)-1;
1887 root
= btrfs_read_fs_root_no_name(info
, &key
);
1889 printk(KERN_ERR
"could not find root %llu\n",
1891 btrfs_free_path(path
);
1896 key
.objectid
= sk
->min_objectid
;
1897 key
.type
= sk
->min_type
;
1898 key
.offset
= sk
->min_offset
;
1900 max_key
.objectid
= sk
->max_objectid
;
1901 max_key
.type
= sk
->max_type
;
1902 max_key
.offset
= sk
->max_offset
;
1904 path
->keep_locks
= 1;
1907 ret
= btrfs_search_forward(root
, &key
, &max_key
, path
,
1914 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
1915 &sk_offset
, &num_found
);
1916 btrfs_release_path(path
);
1917 if (ret
|| num_found
>= sk
->nr_items
)
1923 sk
->nr_items
= num_found
;
1924 btrfs_free_path(path
);
1928 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
1931 struct btrfs_ioctl_search_args
*args
;
1932 struct inode
*inode
;
1935 if (!capable(CAP_SYS_ADMIN
))
1938 args
= memdup_user(argp
, sizeof(*args
));
1940 return PTR_ERR(args
);
1942 inode
= file_inode(file
);
1943 ret
= search_ioctl(inode
, args
);
1944 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1951 * Search INODE_REFs to identify path name of 'dirid' directory
1952 * in a 'tree_id' tree. and sets path name to 'name'.
1954 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1955 u64 tree_id
, u64 dirid
, char *name
)
1957 struct btrfs_root
*root
;
1958 struct btrfs_key key
;
1964 struct btrfs_inode_ref
*iref
;
1965 struct extent_buffer
*l
;
1966 struct btrfs_path
*path
;
1968 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1973 path
= btrfs_alloc_path();
1977 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
1979 key
.objectid
= tree_id
;
1980 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1981 key
.offset
= (u64
)-1;
1982 root
= btrfs_read_fs_root_no_name(info
, &key
);
1984 printk(KERN_ERR
"could not find root %llu\n", tree_id
);
1989 key
.objectid
= dirid
;
1990 key
.type
= BTRFS_INODE_REF_KEY
;
1991 key
.offset
= (u64
)-1;
1994 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1999 slot
= path
->slots
[0];
2000 if (ret
> 0 && slot
> 0)
2002 btrfs_item_key_to_cpu(l
, &key
, slot
);
2004 if (ret
> 0 && (key
.objectid
!= dirid
||
2005 key
.type
!= BTRFS_INODE_REF_KEY
)) {
2010 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
2011 len
= btrfs_inode_ref_name_len(l
, iref
);
2013 total_len
+= len
+ 1;
2018 read_extent_buffer(l
, ptr
,(unsigned long)(iref
+ 1), len
);
2020 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
2023 btrfs_release_path(path
);
2024 key
.objectid
= key
.offset
;
2025 key
.offset
= (u64
)-1;
2026 dirid
= key
.objectid
;
2030 memmove(name
, ptr
, total_len
);
2031 name
[total_len
]='\0';
2034 btrfs_free_path(path
);
2038 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2041 struct btrfs_ioctl_ino_lookup_args
*args
;
2042 struct inode
*inode
;
2045 if (!capable(CAP_SYS_ADMIN
))
2048 args
= memdup_user(argp
, sizeof(*args
));
2050 return PTR_ERR(args
);
2052 inode
= file_inode(file
);
2054 if (args
->treeid
== 0)
2055 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2057 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2058 args
->treeid
, args
->objectid
,
2061 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2068 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2071 struct dentry
*parent
= fdentry(file
);
2072 struct dentry
*dentry
;
2073 struct inode
*dir
= parent
->d_inode
;
2074 struct inode
*inode
;
2075 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2076 struct btrfs_root
*dest
= NULL
;
2077 struct btrfs_ioctl_vol_args
*vol_args
;
2078 struct btrfs_trans_handle
*trans
;
2079 struct btrfs_block_rsv block_rsv
;
2080 u64 qgroup_reserved
;
2085 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2086 if (IS_ERR(vol_args
))
2087 return PTR_ERR(vol_args
);
2089 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2090 namelen
= strlen(vol_args
->name
);
2091 if (strchr(vol_args
->name
, '/') ||
2092 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2097 err
= mnt_want_write_file(file
);
2101 err
= mutex_lock_killable_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2104 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2105 if (IS_ERR(dentry
)) {
2106 err
= PTR_ERR(dentry
);
2107 goto out_unlock_dir
;
2110 if (!dentry
->d_inode
) {
2115 inode
= dentry
->d_inode
;
2116 dest
= BTRFS_I(inode
)->root
;
2117 if (!capable(CAP_SYS_ADMIN
)){
2119 * Regular user. Only allow this with a special mount
2120 * option, when the user has write+exec access to the
2121 * subvol root, and when rmdir(2) would have been
2124 * Note that this is _not_ check that the subvol is
2125 * empty or doesn't contain data that we wouldn't
2126 * otherwise be able to delete.
2128 * Users who want to delete empty subvols should try
2132 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2136 * Do not allow deletion if the parent dir is the same
2137 * as the dir to be deleted. That means the ioctl
2138 * must be called on the dentry referencing the root
2139 * of the subvol, not a random directory contained
2146 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2151 /* check if subvolume may be deleted by a user */
2152 err
= btrfs_may_delete(dir
, dentry
, 1);
2156 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2161 mutex_lock(&inode
->i_mutex
);
2162 err
= d_invalidate(dentry
);
2166 down_write(&root
->fs_info
->subvol_sem
);
2168 err
= may_destroy_subvol(dest
);
2172 btrfs_init_block_rsv(&block_rsv
, BTRFS_BLOCK_RSV_TEMP
);
2174 * One for dir inode, two for dir entries, two for root
2177 err
= btrfs_subvolume_reserve_metadata(root
, &block_rsv
,
2178 5, &qgroup_reserved
, true);
2182 trans
= btrfs_start_transaction(root
, 0);
2183 if (IS_ERR(trans
)) {
2184 err
= PTR_ERR(trans
);
2187 trans
->block_rsv
= &block_rsv
;
2188 trans
->bytes_reserved
= block_rsv
.size
;
2190 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2191 dest
->root_key
.objectid
,
2192 dentry
->d_name
.name
,
2193 dentry
->d_name
.len
);
2196 btrfs_abort_transaction(trans
, root
, ret
);
2200 btrfs_record_root_in_trans(trans
, dest
);
2202 memset(&dest
->root_item
.drop_progress
, 0,
2203 sizeof(dest
->root_item
.drop_progress
));
2204 dest
->root_item
.drop_level
= 0;
2205 btrfs_set_root_refs(&dest
->root_item
, 0);
2207 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
2208 ret
= btrfs_insert_orphan_item(trans
,
2209 root
->fs_info
->tree_root
,
2210 dest
->root_key
.objectid
);
2212 btrfs_abort_transaction(trans
, root
, ret
);
2218 trans
->block_rsv
= NULL
;
2219 trans
->bytes_reserved
= 0;
2220 ret
= btrfs_end_transaction(trans
, root
);
2223 inode
->i_flags
|= S_DEAD
;
2225 btrfs_subvolume_release_metadata(root
, &block_rsv
, qgroup_reserved
);
2227 up_write(&root
->fs_info
->subvol_sem
);
2229 mutex_unlock(&inode
->i_mutex
);
2231 shrink_dcache_sb(root
->fs_info
->sb
);
2232 btrfs_invalidate_inodes(dest
);
2236 if (dest
->cache_inode
) {
2237 iput(dest
->cache_inode
);
2238 dest
->cache_inode
= NULL
;
2244 mutex_unlock(&dir
->i_mutex
);
2245 mnt_drop_write_file(file
);
2251 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2253 struct inode
*inode
= file_inode(file
);
2254 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2255 struct btrfs_ioctl_defrag_range_args
*range
;
2258 ret
= mnt_want_write_file(file
);
2262 if (btrfs_root_readonly(root
)) {
2267 switch (inode
->i_mode
& S_IFMT
) {
2269 if (!capable(CAP_SYS_ADMIN
)) {
2273 ret
= btrfs_defrag_root(root
);
2276 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2279 if (!(file
->f_mode
& FMODE_WRITE
)) {
2284 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2291 if (copy_from_user(range
, argp
,
2297 /* compression requires us to start the IO */
2298 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2299 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2300 range
->extent_thresh
= (u32
)-1;
2303 /* the rest are all set to zero by kzalloc */
2304 range
->len
= (u64
)-1;
2306 ret
= btrfs_defrag_file(file_inode(file
), file
,
2316 mnt_drop_write_file(file
);
2320 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2322 struct btrfs_ioctl_vol_args
*vol_args
;
2325 if (!capable(CAP_SYS_ADMIN
))
2328 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2330 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2334 mutex_lock(&root
->fs_info
->volume_mutex
);
2335 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2336 if (IS_ERR(vol_args
)) {
2337 ret
= PTR_ERR(vol_args
);
2341 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2342 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2346 mutex_unlock(&root
->fs_info
->volume_mutex
);
2347 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2351 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2353 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
2354 struct btrfs_ioctl_vol_args
*vol_args
;
2357 if (!capable(CAP_SYS_ADMIN
))
2360 ret
= mnt_want_write_file(file
);
2364 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2365 if (IS_ERR(vol_args
)) {
2366 ret
= PTR_ERR(vol_args
);
2370 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2372 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2374 ret
= BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS
;
2378 mutex_lock(&root
->fs_info
->volume_mutex
);
2379 ret
= btrfs_rm_device(root
, vol_args
->name
);
2380 mutex_unlock(&root
->fs_info
->volume_mutex
);
2381 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2385 mnt_drop_write_file(file
);
2389 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2391 struct btrfs_ioctl_fs_info_args
*fi_args
;
2392 struct btrfs_device
*device
;
2393 struct btrfs_device
*next
;
2394 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2397 if (!capable(CAP_SYS_ADMIN
))
2400 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2404 fi_args
->num_devices
= fs_devices
->num_devices
;
2405 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2407 mutex_lock(&fs_devices
->device_list_mutex
);
2408 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2409 if (device
->devid
> fi_args
->max_id
)
2410 fi_args
->max_id
= device
->devid
;
2412 mutex_unlock(&fs_devices
->device_list_mutex
);
2414 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2421 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2423 struct btrfs_ioctl_dev_info_args
*di_args
;
2424 struct btrfs_device
*dev
;
2425 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2427 char *s_uuid
= NULL
;
2428 char empty_uuid
[BTRFS_UUID_SIZE
] = {0};
2430 if (!capable(CAP_SYS_ADMIN
))
2433 di_args
= memdup_user(arg
, sizeof(*di_args
));
2434 if (IS_ERR(di_args
))
2435 return PTR_ERR(di_args
);
2437 if (memcmp(empty_uuid
, di_args
->uuid
, BTRFS_UUID_SIZE
) != 0)
2438 s_uuid
= di_args
->uuid
;
2440 mutex_lock(&fs_devices
->device_list_mutex
);
2441 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2448 di_args
->devid
= dev
->devid
;
2449 di_args
->bytes_used
= dev
->bytes_used
;
2450 di_args
->total_bytes
= dev
->total_bytes
;
2451 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2453 struct rcu_string
*name
;
2456 name
= rcu_dereference(dev
->name
);
2457 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2459 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2461 di_args
->path
[0] = '\0';
2465 mutex_unlock(&fs_devices
->device_list_mutex
);
2466 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2473 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
2474 u64 off
, u64 olen
, u64 destoff
)
2476 struct inode
*inode
= file_inode(file
);
2477 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2480 struct btrfs_trans_handle
*trans
;
2481 struct btrfs_path
*path
;
2482 struct extent_buffer
*leaf
;
2484 struct btrfs_key key
;
2489 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
2494 * - split compressed inline extents. annoying: we need to
2495 * decompress into destination's address_space (the file offset
2496 * may change, so source mapping won't do), then recompress (or
2497 * otherwise reinsert) a subrange.
2498 * - allow ranges within the same file to be cloned (provided
2499 * they don't overlap)?
2502 /* the destination must be opened for writing */
2503 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
2506 if (btrfs_root_readonly(root
))
2509 ret
= mnt_want_write_file(file
);
2513 src_file
= fdget(srcfd
);
2514 if (!src_file
.file
) {
2516 goto out_drop_write
;
2520 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
2523 src
= file_inode(src_file
.file
);
2529 /* the src must be open for reading */
2530 if (!(src_file
.file
->f_mode
& FMODE_READ
))
2533 /* don't make the dst file partly checksummed */
2534 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2535 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
2539 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
2543 if (src
->i_sb
!= inode
->i_sb
)
2547 buf
= vmalloc(btrfs_level_size(root
, 0));
2551 path
= btrfs_alloc_path();
2560 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
2561 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
2563 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
2564 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
2567 mutex_lock(&src
->i_mutex
);
2570 /* determine range to clone */
2572 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
2575 olen
= len
= src
->i_size
- off
;
2576 /* if we extend to eof, continue to block boundary */
2577 if (off
+ len
== src
->i_size
)
2578 len
= ALIGN(src
->i_size
, bs
) - off
;
2580 /* verify the end result is block aligned */
2581 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
2582 !IS_ALIGNED(destoff
, bs
))
2585 /* verify if ranges are overlapped within the same file */
2587 if (destoff
+ len
> off
&& destoff
< off
+ len
)
2591 if (destoff
> inode
->i_size
) {
2592 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
2597 /* truncate page cache pages from target inode range */
2598 truncate_inode_pages_range(&inode
->i_data
, destoff
,
2599 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
2601 /* do any pending delalloc/csum calc on src, one way or
2602 another, and lock file content */
2604 struct btrfs_ordered_extent
*ordered
;
2605 lock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2606 ordered
= btrfs_lookup_first_ordered_extent(src
, off
+ len
- 1);
2608 !test_range_bit(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1,
2609 EXTENT_DELALLOC
, 0, NULL
))
2611 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2613 btrfs_put_ordered_extent(ordered
);
2614 btrfs_wait_ordered_range(src
, off
, len
);
2618 key
.objectid
= btrfs_ino(src
);
2619 key
.type
= BTRFS_EXTENT_DATA_KEY
;
2624 * note the key will change type as we walk through the
2627 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
2632 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2633 if (path
->slots
[0] >= nritems
) {
2634 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
2639 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2641 leaf
= path
->nodes
[0];
2642 slot
= path
->slots
[0];
2644 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2645 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
2646 key
.objectid
!= btrfs_ino(src
))
2649 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
2650 struct btrfs_file_extent_item
*extent
;
2653 struct btrfs_key new_key
;
2654 u64 disko
= 0, diskl
= 0;
2655 u64 datao
= 0, datal
= 0;
2659 size
= btrfs_item_size_nr(leaf
, slot
);
2660 read_extent_buffer(leaf
, buf
,
2661 btrfs_item_ptr_offset(leaf
, slot
),
2664 extent
= btrfs_item_ptr(leaf
, slot
,
2665 struct btrfs_file_extent_item
);
2666 comp
= btrfs_file_extent_compression(leaf
, extent
);
2667 type
= btrfs_file_extent_type(leaf
, extent
);
2668 if (type
== BTRFS_FILE_EXTENT_REG
||
2669 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2670 disko
= btrfs_file_extent_disk_bytenr(leaf
,
2672 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
2674 datao
= btrfs_file_extent_offset(leaf
, extent
);
2675 datal
= btrfs_file_extent_num_bytes(leaf
,
2677 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2678 /* take upper bound, may be compressed */
2679 datal
= btrfs_file_extent_ram_bytes(leaf
,
2682 btrfs_release_path(path
);
2684 if (key
.offset
+ datal
<= off
||
2685 key
.offset
>= off
+ len
- 1)
2688 memcpy(&new_key
, &key
, sizeof(new_key
));
2689 new_key
.objectid
= btrfs_ino(inode
);
2690 if (off
<= key
.offset
)
2691 new_key
.offset
= key
.offset
+ destoff
- off
;
2693 new_key
.offset
= destoff
;
2696 * 1 - adjusting old extent (we may have to split it)
2697 * 1 - add new extent
2700 trans
= btrfs_start_transaction(root
, 3);
2701 if (IS_ERR(trans
)) {
2702 ret
= PTR_ERR(trans
);
2706 if (type
== BTRFS_FILE_EXTENT_REG
||
2707 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2709 * a | --- range to clone ---| b
2710 * | ------------- extent ------------- |
2713 /* substract range b */
2714 if (key
.offset
+ datal
> off
+ len
)
2715 datal
= off
+ len
- key
.offset
;
2717 /* substract range a */
2718 if (off
> key
.offset
) {
2719 datao
+= off
- key
.offset
;
2720 datal
-= off
- key
.offset
;
2723 ret
= btrfs_drop_extents(trans
, root
, inode
,
2725 new_key
.offset
+ datal
,
2728 btrfs_abort_transaction(trans
, root
,
2730 btrfs_end_transaction(trans
, root
);
2734 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2737 btrfs_abort_transaction(trans
, root
,
2739 btrfs_end_transaction(trans
, root
);
2743 leaf
= path
->nodes
[0];
2744 slot
= path
->slots
[0];
2745 write_extent_buffer(leaf
, buf
,
2746 btrfs_item_ptr_offset(leaf
, slot
),
2749 extent
= btrfs_item_ptr(leaf
, slot
,
2750 struct btrfs_file_extent_item
);
2752 /* disko == 0 means it's a hole */
2756 btrfs_set_file_extent_offset(leaf
, extent
,
2758 btrfs_set_file_extent_num_bytes(leaf
, extent
,
2761 inode_add_bytes(inode
, datal
);
2762 ret
= btrfs_inc_extent_ref(trans
, root
,
2764 root
->root_key
.objectid
,
2766 new_key
.offset
- datao
,
2769 btrfs_abort_transaction(trans
,
2772 btrfs_end_transaction(trans
,
2778 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2781 if (off
> key
.offset
) {
2782 skip
= off
- key
.offset
;
2783 new_key
.offset
+= skip
;
2786 if (key
.offset
+ datal
> off
+ len
)
2787 trim
= key
.offset
+ datal
- (off
+ len
);
2789 if (comp
&& (skip
|| trim
)) {
2791 btrfs_end_transaction(trans
, root
);
2794 size
-= skip
+ trim
;
2795 datal
-= skip
+ trim
;
2797 ret
= btrfs_drop_extents(trans
, root
, inode
,
2799 new_key
.offset
+ datal
,
2802 btrfs_abort_transaction(trans
, root
,
2804 btrfs_end_transaction(trans
, root
);
2808 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2811 btrfs_abort_transaction(trans
, root
,
2813 btrfs_end_transaction(trans
, root
);
2819 btrfs_file_extent_calc_inline_size(0);
2820 memmove(buf
+start
, buf
+start
+skip
,
2824 leaf
= path
->nodes
[0];
2825 slot
= path
->slots
[0];
2826 write_extent_buffer(leaf
, buf
,
2827 btrfs_item_ptr_offset(leaf
, slot
),
2829 inode_add_bytes(inode
, datal
);
2832 btrfs_mark_buffer_dirty(leaf
);
2833 btrfs_release_path(path
);
2835 inode_inc_iversion(inode
);
2836 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2839 * we round up to the block size at eof when
2840 * determining which extents to clone above,
2841 * but shouldn't round up the file size
2843 endoff
= new_key
.offset
+ datal
;
2844 if (endoff
> destoff
+olen
)
2845 endoff
= destoff
+olen
;
2846 if (endoff
> inode
->i_size
)
2847 btrfs_i_size_write(inode
, endoff
);
2849 ret
= btrfs_update_inode(trans
, root
, inode
);
2851 btrfs_abort_transaction(trans
, root
, ret
);
2852 btrfs_end_transaction(trans
, root
);
2855 ret
= btrfs_end_transaction(trans
, root
);
2858 btrfs_release_path(path
);
2863 btrfs_release_path(path
);
2864 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2866 mutex_unlock(&src
->i_mutex
);
2868 mutex_unlock(&inode
->i_mutex
);
2870 btrfs_free_path(path
);
2874 mnt_drop_write_file(file
);
2878 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
2880 struct btrfs_ioctl_clone_range_args args
;
2882 if (copy_from_user(&args
, argp
, sizeof(args
)))
2884 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
2885 args
.src_length
, args
.dest_offset
);
2889 * there are many ways the trans_start and trans_end ioctls can lead
2890 * to deadlocks. They should only be used by applications that
2891 * basically own the machine, and have a very in depth understanding
2892 * of all the possible deadlocks and enospc problems.
2894 static long btrfs_ioctl_trans_start(struct file
*file
)
2896 struct inode
*inode
= file_inode(file
);
2897 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2898 struct btrfs_trans_handle
*trans
;
2902 if (!capable(CAP_SYS_ADMIN
))
2906 if (file
->private_data
)
2910 if (btrfs_root_readonly(root
))
2913 ret
= mnt_want_write_file(file
);
2917 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
2920 trans
= btrfs_start_ioctl_transaction(root
);
2924 file
->private_data
= trans
;
2928 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
2929 mnt_drop_write_file(file
);
2934 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2936 struct inode
*inode
= file_inode(file
);
2937 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2938 struct btrfs_root
*new_root
;
2939 struct btrfs_dir_item
*di
;
2940 struct btrfs_trans_handle
*trans
;
2941 struct btrfs_path
*path
;
2942 struct btrfs_key location
;
2943 struct btrfs_disk_key disk_key
;
2948 if (!capable(CAP_SYS_ADMIN
))
2951 ret
= mnt_want_write_file(file
);
2955 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
2961 objectid
= root
->root_key
.objectid
;
2963 location
.objectid
= objectid
;
2964 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2965 location
.offset
= (u64
)-1;
2967 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
2968 if (IS_ERR(new_root
)) {
2969 ret
= PTR_ERR(new_root
);
2973 path
= btrfs_alloc_path();
2978 path
->leave_spinning
= 1;
2980 trans
= btrfs_start_transaction(root
, 1);
2981 if (IS_ERR(trans
)) {
2982 btrfs_free_path(path
);
2983 ret
= PTR_ERR(trans
);
2987 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
2988 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
2989 dir_id
, "default", 7, 1);
2990 if (IS_ERR_OR_NULL(di
)) {
2991 btrfs_free_path(path
);
2992 btrfs_end_transaction(trans
, root
);
2993 printk(KERN_ERR
"Umm, you don't have the default dir item, "
2994 "this isn't going to work\n");
2999 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
3000 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
3001 btrfs_mark_buffer_dirty(path
->nodes
[0]);
3002 btrfs_free_path(path
);
3004 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
3005 btrfs_end_transaction(trans
, root
);
3007 mnt_drop_write_file(file
);
3011 void btrfs_get_block_group_info(struct list_head
*groups_list
,
3012 struct btrfs_ioctl_space_info
*space
)
3014 struct btrfs_block_group_cache
*block_group
;
3016 space
->total_bytes
= 0;
3017 space
->used_bytes
= 0;
3019 list_for_each_entry(block_group
, groups_list
, list
) {
3020 space
->flags
= block_group
->flags
;
3021 space
->total_bytes
+= block_group
->key
.offset
;
3022 space
->used_bytes
+=
3023 btrfs_block_group_used(&block_group
->item
);
3027 static long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
3029 struct btrfs_ioctl_space_args space_args
;
3030 struct btrfs_ioctl_space_info space
;
3031 struct btrfs_ioctl_space_info
*dest
;
3032 struct btrfs_ioctl_space_info
*dest_orig
;
3033 struct btrfs_ioctl_space_info __user
*user_dest
;
3034 struct btrfs_space_info
*info
;
3035 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
3036 BTRFS_BLOCK_GROUP_SYSTEM
,
3037 BTRFS_BLOCK_GROUP_METADATA
,
3038 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
3045 if (copy_from_user(&space_args
,
3046 (struct btrfs_ioctl_space_args __user
*)arg
,
3047 sizeof(space_args
)))
3050 for (i
= 0; i
< num_types
; i
++) {
3051 struct btrfs_space_info
*tmp
;
3055 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3057 if (tmp
->flags
== types
[i
]) {
3067 down_read(&info
->groups_sem
);
3068 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3069 if (!list_empty(&info
->block_groups
[c
]))
3072 up_read(&info
->groups_sem
);
3075 /* space_slots == 0 means they are asking for a count */
3076 if (space_args
.space_slots
== 0) {
3077 space_args
.total_spaces
= slot_count
;
3081 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
3083 alloc_size
= sizeof(*dest
) * slot_count
;
3085 /* we generally have at most 6 or so space infos, one for each raid
3086 * level. So, a whole page should be more than enough for everyone
3088 if (alloc_size
> PAGE_CACHE_SIZE
)
3091 space_args
.total_spaces
= 0;
3092 dest
= kmalloc(alloc_size
, GFP_NOFS
);
3097 /* now we have a buffer to copy into */
3098 for (i
= 0; i
< num_types
; i
++) {
3099 struct btrfs_space_info
*tmp
;
3106 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3108 if (tmp
->flags
== types
[i
]) {
3117 down_read(&info
->groups_sem
);
3118 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3119 if (!list_empty(&info
->block_groups
[c
])) {
3120 btrfs_get_block_group_info(
3121 &info
->block_groups
[c
], &space
);
3122 memcpy(dest
, &space
, sizeof(space
));
3124 space_args
.total_spaces
++;
3130 up_read(&info
->groups_sem
);
3133 user_dest
= (struct btrfs_ioctl_space_info __user
*)
3134 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
3136 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
3141 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
3148 * there are many ways the trans_start and trans_end ioctls can lead
3149 * to deadlocks. They should only be used by applications that
3150 * basically own the machine, and have a very in depth understanding
3151 * of all the possible deadlocks and enospc problems.
3153 long btrfs_ioctl_trans_end(struct file
*file
)
3155 struct inode
*inode
= file_inode(file
);
3156 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3157 struct btrfs_trans_handle
*trans
;
3159 trans
= file
->private_data
;
3162 file
->private_data
= NULL
;
3164 btrfs_end_transaction(trans
, root
);
3166 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3168 mnt_drop_write_file(file
);
3172 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
3175 struct btrfs_trans_handle
*trans
;
3179 trans
= btrfs_attach_transaction_barrier(root
);
3180 if (IS_ERR(trans
)) {
3181 if (PTR_ERR(trans
) != -ENOENT
)
3182 return PTR_ERR(trans
);
3184 /* No running transaction, don't bother */
3185 transid
= root
->fs_info
->last_trans_committed
;
3188 transid
= trans
->transid
;
3189 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
3191 btrfs_end_transaction(trans
, root
);
3196 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
3201 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
3207 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
3210 transid
= 0; /* current trans */
3212 return btrfs_wait_for_commit(root
, transid
);
3215 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
3217 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3218 struct btrfs_ioctl_scrub_args
*sa
;
3221 if (!capable(CAP_SYS_ADMIN
))
3224 sa
= memdup_user(arg
, sizeof(*sa
));
3228 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
3229 ret
= mnt_want_write_file(file
);
3234 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
3235 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
3238 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3241 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
3242 mnt_drop_write_file(file
);
3248 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
3250 if (!capable(CAP_SYS_ADMIN
))
3253 return btrfs_scrub_cancel(root
->fs_info
);
3256 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
3259 struct btrfs_ioctl_scrub_args
*sa
;
3262 if (!capable(CAP_SYS_ADMIN
))
3265 sa
= memdup_user(arg
, sizeof(*sa
));
3269 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
3271 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3278 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
3281 struct btrfs_ioctl_get_dev_stats
*sa
;
3284 sa
= memdup_user(arg
, sizeof(*sa
));
3288 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
3293 ret
= btrfs_get_dev_stats(root
, sa
);
3295 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3302 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
3304 struct btrfs_ioctl_dev_replace_args
*p
;
3307 if (!capable(CAP_SYS_ADMIN
))
3310 p
= memdup_user(arg
, sizeof(*p
));
3315 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
3317 &root
->fs_info
->mutually_exclusive_operation_running
,
3319 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3322 ret
= btrfs_dev_replace_start(root
, p
);
3324 &root
->fs_info
->mutually_exclusive_operation_running
,
3328 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
3329 btrfs_dev_replace_status(root
->fs_info
, p
);
3332 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
3333 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
3340 if (copy_to_user(arg
, p
, sizeof(*p
)))
3347 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
3353 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
3354 struct inode_fs_paths
*ipath
= NULL
;
3355 struct btrfs_path
*path
;
3357 if (!capable(CAP_DAC_READ_SEARCH
))
3360 path
= btrfs_alloc_path();
3366 ipa
= memdup_user(arg
, sizeof(*ipa
));
3373 size
= min_t(u32
, ipa
->size
, 4096);
3374 ipath
= init_ipath(size
, root
, path
);
3375 if (IS_ERR(ipath
)) {
3376 ret
= PTR_ERR(ipath
);
3381 ret
= paths_from_inode(ipa
->inum
, ipath
);
3385 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
3386 rel_ptr
= ipath
->fspath
->val
[i
] -
3387 (u64
)(unsigned long)ipath
->fspath
->val
;
3388 ipath
->fspath
->val
[i
] = rel_ptr
;
3391 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
3392 (void *)(unsigned long)ipath
->fspath
, size
);
3399 btrfs_free_path(path
);
3406 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
3408 struct btrfs_data_container
*inodes
= ctx
;
3409 const size_t c
= 3 * sizeof(u64
);
3411 if (inodes
->bytes_left
>= c
) {
3412 inodes
->bytes_left
-= c
;
3413 inodes
->val
[inodes
->elem_cnt
] = inum
;
3414 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
3415 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
3416 inodes
->elem_cnt
+= 3;
3418 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
3419 inodes
->bytes_left
= 0;
3420 inodes
->elem_missed
+= 3;
3426 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
3431 struct btrfs_ioctl_logical_ino_args
*loi
;
3432 struct btrfs_data_container
*inodes
= NULL
;
3433 struct btrfs_path
*path
= NULL
;
3435 if (!capable(CAP_SYS_ADMIN
))
3438 loi
= memdup_user(arg
, sizeof(*loi
));
3445 path
= btrfs_alloc_path();
3451 size
= min_t(u32
, loi
->size
, 64 * 1024);
3452 inodes
= init_data_container(size
);
3453 if (IS_ERR(inodes
)) {
3454 ret
= PTR_ERR(inodes
);
3459 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
3460 build_ino_list
, inodes
);
3466 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
3467 (void *)(unsigned long)inodes
, size
);
3472 btrfs_free_path(path
);
3479 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
3480 struct btrfs_ioctl_balance_args
*bargs
)
3482 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
3484 bargs
->flags
= bctl
->flags
;
3486 if (atomic_read(&fs_info
->balance_running
))
3487 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
3488 if (atomic_read(&fs_info
->balance_pause_req
))
3489 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
3490 if (atomic_read(&fs_info
->balance_cancel_req
))
3491 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
3493 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
3494 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
3495 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
3498 spin_lock(&fs_info
->balance_lock
);
3499 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3500 spin_unlock(&fs_info
->balance_lock
);
3502 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3506 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
3508 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3509 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3510 struct btrfs_ioctl_balance_args
*bargs
;
3511 struct btrfs_balance_control
*bctl
;
3512 bool need_unlock
; /* for mut. excl. ops lock */
3515 if (!capable(CAP_SYS_ADMIN
))
3518 ret
= mnt_want_write_file(file
);
3523 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
3524 mutex_lock(&fs_info
->volume_mutex
);
3525 mutex_lock(&fs_info
->balance_mutex
);
3531 * mut. excl. ops lock is locked. Three possibilites:
3532 * (1) some other op is running
3533 * (2) balance is running
3534 * (3) balance is paused -- special case (think resume)
3536 mutex_lock(&fs_info
->balance_mutex
);
3537 if (fs_info
->balance_ctl
) {
3538 /* this is either (2) or (3) */
3539 if (!atomic_read(&fs_info
->balance_running
)) {
3540 mutex_unlock(&fs_info
->balance_mutex
);
3541 if (!mutex_trylock(&fs_info
->volume_mutex
))
3543 mutex_lock(&fs_info
->balance_mutex
);
3545 if (fs_info
->balance_ctl
&&
3546 !atomic_read(&fs_info
->balance_running
)) {
3548 need_unlock
= false;
3552 mutex_unlock(&fs_info
->balance_mutex
);
3553 mutex_unlock(&fs_info
->volume_mutex
);
3557 mutex_unlock(&fs_info
->balance_mutex
);
3563 mutex_unlock(&fs_info
->balance_mutex
);
3564 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3570 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
3573 bargs
= memdup_user(arg
, sizeof(*bargs
));
3574 if (IS_ERR(bargs
)) {
3575 ret
= PTR_ERR(bargs
);
3579 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
3580 if (!fs_info
->balance_ctl
) {
3585 bctl
= fs_info
->balance_ctl
;
3586 spin_lock(&fs_info
->balance_lock
);
3587 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
3588 spin_unlock(&fs_info
->balance_lock
);
3596 if (fs_info
->balance_ctl
) {
3601 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
3607 bctl
->fs_info
= fs_info
;
3609 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
3610 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
3611 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
3613 bctl
->flags
= bargs
->flags
;
3615 /* balance everything - no filters */
3616 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
3621 * Ownership of bctl and mutually_exclusive_operation_running
3622 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3623 * or, if restriper was paused all the way until unmount, in
3624 * free_fs_info. mutually_exclusive_operation_running is
3625 * cleared in __cancel_balance.
3627 need_unlock
= false;
3629 ret
= btrfs_balance(bctl
, bargs
);
3632 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3639 mutex_unlock(&fs_info
->balance_mutex
);
3640 mutex_unlock(&fs_info
->volume_mutex
);
3642 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
3644 mnt_drop_write_file(file
);
3648 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
3650 if (!capable(CAP_SYS_ADMIN
))
3654 case BTRFS_BALANCE_CTL_PAUSE
:
3655 return btrfs_pause_balance(root
->fs_info
);
3656 case BTRFS_BALANCE_CTL_CANCEL
:
3657 return btrfs_cancel_balance(root
->fs_info
);
3663 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
3666 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3667 struct btrfs_ioctl_balance_args
*bargs
;
3670 if (!capable(CAP_SYS_ADMIN
))
3673 mutex_lock(&fs_info
->balance_mutex
);
3674 if (!fs_info
->balance_ctl
) {
3679 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
3685 update_ioctl_balance_args(fs_info
, 1, bargs
);
3687 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3692 mutex_unlock(&fs_info
->balance_mutex
);
3696 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
3698 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3699 struct btrfs_ioctl_quota_ctl_args
*sa
;
3700 struct btrfs_trans_handle
*trans
= NULL
;
3704 if (!capable(CAP_SYS_ADMIN
))
3707 ret
= mnt_want_write_file(file
);
3711 sa
= memdup_user(arg
, sizeof(*sa
));
3717 down_write(&root
->fs_info
->subvol_sem
);
3718 trans
= btrfs_start_transaction(root
->fs_info
->tree_root
, 2);
3719 if (IS_ERR(trans
)) {
3720 ret
= PTR_ERR(trans
);
3725 case BTRFS_QUOTA_CTL_ENABLE
:
3726 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
3728 case BTRFS_QUOTA_CTL_DISABLE
:
3729 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
3736 err
= btrfs_commit_transaction(trans
, root
->fs_info
->tree_root
);
3741 up_write(&root
->fs_info
->subvol_sem
);
3743 mnt_drop_write_file(file
);
3747 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
3749 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3750 struct btrfs_ioctl_qgroup_assign_args
*sa
;
3751 struct btrfs_trans_handle
*trans
;
3755 if (!capable(CAP_SYS_ADMIN
))
3758 ret
= mnt_want_write_file(file
);
3762 sa
= memdup_user(arg
, sizeof(*sa
));
3768 trans
= btrfs_join_transaction(root
);
3769 if (IS_ERR(trans
)) {
3770 ret
= PTR_ERR(trans
);
3774 /* FIXME: check if the IDs really exist */
3776 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
3779 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
3783 err
= btrfs_end_transaction(trans
, root
);
3790 mnt_drop_write_file(file
);
3794 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
3796 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3797 struct btrfs_ioctl_qgroup_create_args
*sa
;
3798 struct btrfs_trans_handle
*trans
;
3802 if (!capable(CAP_SYS_ADMIN
))
3805 ret
= mnt_want_write_file(file
);
3809 sa
= memdup_user(arg
, sizeof(*sa
));
3815 if (!sa
->qgroupid
) {
3820 trans
= btrfs_join_transaction(root
);
3821 if (IS_ERR(trans
)) {
3822 ret
= PTR_ERR(trans
);
3826 /* FIXME: check if the IDs really exist */
3828 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
,
3831 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
3834 err
= btrfs_end_transaction(trans
, root
);
3841 mnt_drop_write_file(file
);
3845 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
3847 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3848 struct btrfs_ioctl_qgroup_limit_args
*sa
;
3849 struct btrfs_trans_handle
*trans
;
3854 if (!capable(CAP_SYS_ADMIN
))
3857 ret
= mnt_want_write_file(file
);
3861 sa
= memdup_user(arg
, sizeof(*sa
));
3867 trans
= btrfs_join_transaction(root
);
3868 if (IS_ERR(trans
)) {
3869 ret
= PTR_ERR(trans
);
3873 qgroupid
= sa
->qgroupid
;
3875 /* take the current subvol as qgroup */
3876 qgroupid
= root
->root_key
.objectid
;
3879 /* FIXME: check if the IDs really exist */
3880 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
3882 err
= btrfs_end_transaction(trans
, root
);
3889 mnt_drop_write_file(file
);
3893 static long btrfs_ioctl_quota_rescan(struct file
*file
, void __user
*arg
)
3895 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3896 struct btrfs_ioctl_quota_rescan_args
*qsa
;
3899 if (!capable(CAP_SYS_ADMIN
))
3902 ret
= mnt_want_write_file(file
);
3906 qsa
= memdup_user(arg
, sizeof(*qsa
));
3917 ret
= btrfs_qgroup_rescan(root
->fs_info
);
3922 mnt_drop_write_file(file
);
3926 static long btrfs_ioctl_quota_rescan_status(struct file
*file
, void __user
*arg
)
3928 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
3929 struct btrfs_ioctl_quota_rescan_args
*qsa
;
3932 if (!capable(CAP_SYS_ADMIN
))
3935 qsa
= kzalloc(sizeof(*qsa
), GFP_NOFS
);
3939 if (root
->fs_info
->qgroup_flags
& BTRFS_QGROUP_STATUS_FLAG_RESCAN
) {
3941 qsa
->progress
= root
->fs_info
->qgroup_rescan_progress
.objectid
;
3944 if (copy_to_user(arg
, qsa
, sizeof(*qsa
)))
3951 static long btrfs_ioctl_quota_rescan_wait(struct file
*file
, void __user
*arg
)
3953 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3955 if (!capable(CAP_SYS_ADMIN
))
3958 return btrfs_qgroup_wait_for_completion(root
->fs_info
);
3961 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
3964 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
3965 struct inode
*inode
= file_inode(file
);
3966 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3967 struct btrfs_root_item
*root_item
= &root
->root_item
;
3968 struct btrfs_trans_handle
*trans
;
3969 struct timespec ct
= CURRENT_TIME
;
3972 ret
= mnt_want_write_file(file
);
3976 down_write(&root
->fs_info
->subvol_sem
);
3978 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
3983 if (btrfs_root_readonly(root
)) {
3988 if (!inode_owner_or_capable(inode
)) {
3993 sa
= memdup_user(arg
, sizeof(*sa
));
4000 trans
= btrfs_start_transaction(root
, 1);
4001 if (IS_ERR(trans
)) {
4002 ret
= PTR_ERR(trans
);
4007 sa
->rtransid
= trans
->transid
;
4008 sa
->rtime
.sec
= ct
.tv_sec
;
4009 sa
->rtime
.nsec
= ct
.tv_nsec
;
4011 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
4012 btrfs_set_root_stransid(root_item
, sa
->stransid
);
4013 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
4014 root_item
->stime
.sec
= cpu_to_le64(sa
->stime
.sec
);
4015 root_item
->stime
.nsec
= cpu_to_le32(sa
->stime
.nsec
);
4016 root_item
->rtime
.sec
= cpu_to_le64(sa
->rtime
.sec
);
4017 root_item
->rtime
.nsec
= cpu_to_le32(sa
->rtime
.nsec
);
4019 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
4020 &root
->root_key
, &root
->root_item
);
4022 btrfs_end_transaction(trans
, root
);
4026 ret
= btrfs_commit_transaction(trans
, root
);
4031 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
4037 up_write(&root
->fs_info
->subvol_sem
);
4038 mnt_drop_write_file(file
);
4042 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
4044 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4045 const char *label
= root
->fs_info
->super_copy
->label
;
4046 size_t len
= strnlen(label
, BTRFS_LABEL_SIZE
);
4049 if (len
== BTRFS_LABEL_SIZE
) {
4050 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4054 mutex_lock(&root
->fs_info
->volume_mutex
);
4055 ret
= copy_to_user(arg
, label
, len
);
4056 mutex_unlock(&root
->fs_info
->volume_mutex
);
4058 return ret
? -EFAULT
: 0;
4061 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
4063 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4064 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
4065 struct btrfs_trans_handle
*trans
;
4066 char label
[BTRFS_LABEL_SIZE
];
4069 if (!capable(CAP_SYS_ADMIN
))
4072 if (copy_from_user(label
, arg
, sizeof(label
)))
4075 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
4076 pr_err("btrfs: unable to set label with more than %d bytes\n",
4077 BTRFS_LABEL_SIZE
- 1);
4081 ret
= mnt_want_write_file(file
);
4085 mutex_lock(&root
->fs_info
->volume_mutex
);
4086 trans
= btrfs_start_transaction(root
, 0);
4087 if (IS_ERR(trans
)) {
4088 ret
= PTR_ERR(trans
);
4092 strcpy(super_block
->label
, label
);
4093 ret
= btrfs_end_transaction(trans
, root
);
4096 mutex_unlock(&root
->fs_info
->volume_mutex
);
4097 mnt_drop_write_file(file
);
4101 long btrfs_ioctl(struct file
*file
, unsigned int
4102 cmd
, unsigned long arg
)
4104 struct btrfs_root
*root
= BTRFS_I(file_inode(file
))->root
;
4105 void __user
*argp
= (void __user
*)arg
;
4108 case FS_IOC_GETFLAGS
:
4109 return btrfs_ioctl_getflags(file
, argp
);
4110 case FS_IOC_SETFLAGS
:
4111 return btrfs_ioctl_setflags(file
, argp
);
4112 case FS_IOC_GETVERSION
:
4113 return btrfs_ioctl_getversion(file
, argp
);
4115 return btrfs_ioctl_fitrim(file
, argp
);
4116 case BTRFS_IOC_SNAP_CREATE
:
4117 return btrfs_ioctl_snap_create(file
, argp
, 0);
4118 case BTRFS_IOC_SNAP_CREATE_V2
:
4119 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
4120 case BTRFS_IOC_SUBVOL_CREATE
:
4121 return btrfs_ioctl_snap_create(file
, argp
, 1);
4122 case BTRFS_IOC_SUBVOL_CREATE_V2
:
4123 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
4124 case BTRFS_IOC_SNAP_DESTROY
:
4125 return btrfs_ioctl_snap_destroy(file
, argp
);
4126 case BTRFS_IOC_SUBVOL_GETFLAGS
:
4127 return btrfs_ioctl_subvol_getflags(file
, argp
);
4128 case BTRFS_IOC_SUBVOL_SETFLAGS
:
4129 return btrfs_ioctl_subvol_setflags(file
, argp
);
4130 case BTRFS_IOC_DEFAULT_SUBVOL
:
4131 return btrfs_ioctl_default_subvol(file
, argp
);
4132 case BTRFS_IOC_DEFRAG
:
4133 return btrfs_ioctl_defrag(file
, NULL
);
4134 case BTRFS_IOC_DEFRAG_RANGE
:
4135 return btrfs_ioctl_defrag(file
, argp
);
4136 case BTRFS_IOC_RESIZE
:
4137 return btrfs_ioctl_resize(file
, argp
);
4138 case BTRFS_IOC_ADD_DEV
:
4139 return btrfs_ioctl_add_dev(root
, argp
);
4140 case BTRFS_IOC_RM_DEV
:
4141 return btrfs_ioctl_rm_dev(file
, argp
);
4142 case BTRFS_IOC_FS_INFO
:
4143 return btrfs_ioctl_fs_info(root
, argp
);
4144 case BTRFS_IOC_DEV_INFO
:
4145 return btrfs_ioctl_dev_info(root
, argp
);
4146 case BTRFS_IOC_BALANCE
:
4147 return btrfs_ioctl_balance(file
, NULL
);
4148 case BTRFS_IOC_CLONE
:
4149 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
4150 case BTRFS_IOC_CLONE_RANGE
:
4151 return btrfs_ioctl_clone_range(file
, argp
);
4152 case BTRFS_IOC_TRANS_START
:
4153 return btrfs_ioctl_trans_start(file
);
4154 case BTRFS_IOC_TRANS_END
:
4155 return btrfs_ioctl_trans_end(file
);
4156 case BTRFS_IOC_TREE_SEARCH
:
4157 return btrfs_ioctl_tree_search(file
, argp
);
4158 case BTRFS_IOC_INO_LOOKUP
:
4159 return btrfs_ioctl_ino_lookup(file
, argp
);
4160 case BTRFS_IOC_INO_PATHS
:
4161 return btrfs_ioctl_ino_to_path(root
, argp
);
4162 case BTRFS_IOC_LOGICAL_INO
:
4163 return btrfs_ioctl_logical_to_ino(root
, argp
);
4164 case BTRFS_IOC_SPACE_INFO
:
4165 return btrfs_ioctl_space_info(root
, argp
);
4166 case BTRFS_IOC_SYNC
:
4167 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
4169 case BTRFS_IOC_START_SYNC
:
4170 return btrfs_ioctl_start_sync(root
, argp
);
4171 case BTRFS_IOC_WAIT_SYNC
:
4172 return btrfs_ioctl_wait_sync(root
, argp
);
4173 case BTRFS_IOC_SCRUB
:
4174 return btrfs_ioctl_scrub(file
, argp
);
4175 case BTRFS_IOC_SCRUB_CANCEL
:
4176 return btrfs_ioctl_scrub_cancel(root
, argp
);
4177 case BTRFS_IOC_SCRUB_PROGRESS
:
4178 return btrfs_ioctl_scrub_progress(root
, argp
);
4179 case BTRFS_IOC_BALANCE_V2
:
4180 return btrfs_ioctl_balance(file
, argp
);
4181 case BTRFS_IOC_BALANCE_CTL
:
4182 return btrfs_ioctl_balance_ctl(root
, arg
);
4183 case BTRFS_IOC_BALANCE_PROGRESS
:
4184 return btrfs_ioctl_balance_progress(root
, argp
);
4185 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
4186 return btrfs_ioctl_set_received_subvol(file
, argp
);
4187 case BTRFS_IOC_SEND
:
4188 return btrfs_ioctl_send(file
, argp
);
4189 case BTRFS_IOC_GET_DEV_STATS
:
4190 return btrfs_ioctl_get_dev_stats(root
, argp
);
4191 case BTRFS_IOC_QUOTA_CTL
:
4192 return btrfs_ioctl_quota_ctl(file
, argp
);
4193 case BTRFS_IOC_QGROUP_ASSIGN
:
4194 return btrfs_ioctl_qgroup_assign(file
, argp
);
4195 case BTRFS_IOC_QGROUP_CREATE
:
4196 return btrfs_ioctl_qgroup_create(file
, argp
);
4197 case BTRFS_IOC_QGROUP_LIMIT
:
4198 return btrfs_ioctl_qgroup_limit(file
, argp
);
4199 case BTRFS_IOC_QUOTA_RESCAN
:
4200 return btrfs_ioctl_quota_rescan(file
, argp
);
4201 case BTRFS_IOC_QUOTA_RESCAN_STATUS
:
4202 return btrfs_ioctl_quota_rescan_status(file
, argp
);
4203 case BTRFS_IOC_QUOTA_RESCAN_WAIT
:
4204 return btrfs_ioctl_quota_rescan_wait(file
, argp
);
4205 case BTRFS_IOC_DEV_REPLACE
:
4206 return btrfs_ioctl_dev_replace(root
, argp
);
4207 case BTRFS_IOC_GET_FSLABEL
:
4208 return btrfs_ioctl_get_fslabel(file
, argp
);
4209 case BTRFS_IOC_SET_FSLABEL
:
4210 return btrfs_ioctl_set_fslabel(file
, argp
);