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
->f_path
.dentry
->d_inode
);
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
->f_path
.dentry
->d_inode
;
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
->f_path
.dentry
->d_inode
;
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 btrfs_root
*root
,
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
*new_root
;
378 struct dentry
*parent
= dentry
->d_parent
;
380 struct timespec cur_time
= CURRENT_TIME
;
384 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
388 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
392 dir
= parent
->d_inode
;
400 trans
= btrfs_start_transaction(root
, 6);
402 return PTR_ERR(trans
);
404 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
, inherit
);
408 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
409 0, objectid
, NULL
, 0, 0, 0);
415 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
416 btrfs_set_header_bytenr(leaf
, leaf
->start
);
417 btrfs_set_header_generation(leaf
, trans
->transid
);
418 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
419 btrfs_set_header_owner(leaf
, objectid
);
421 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
422 (unsigned long)btrfs_header_fsid(leaf
),
424 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
425 (unsigned long)btrfs_header_chunk_tree_uuid(leaf
),
427 btrfs_mark_buffer_dirty(leaf
);
429 memset(&root_item
, 0, sizeof(root_item
));
431 inode_item
= &root_item
.inode
;
432 inode_item
->generation
= cpu_to_le64(1);
433 inode_item
->size
= cpu_to_le64(3);
434 inode_item
->nlink
= cpu_to_le32(1);
435 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
436 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
439 root_item
.byte_limit
= 0;
440 inode_item
->flags
= cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT
);
442 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
443 btrfs_set_root_generation(&root_item
, trans
->transid
);
444 btrfs_set_root_level(&root_item
, 0);
445 btrfs_set_root_refs(&root_item
, 1);
446 btrfs_set_root_used(&root_item
, leaf
->len
);
447 btrfs_set_root_last_snapshot(&root_item
, 0);
449 btrfs_set_root_generation_v2(&root_item
,
450 btrfs_root_generation(&root_item
));
451 uuid_le_gen(&new_uuid
);
452 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
453 root_item
.otime
.sec
= cpu_to_le64(cur_time
.tv_sec
);
454 root_item
.otime
.nsec
= cpu_to_le32(cur_time
.tv_nsec
);
455 root_item
.ctime
= root_item
.otime
;
456 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
457 btrfs_set_root_otransid(&root_item
, trans
->transid
);
459 btrfs_tree_unlock(leaf
);
460 free_extent_buffer(leaf
);
463 btrfs_set_root_dirid(&root_item
, new_dirid
);
465 key
.objectid
= objectid
;
467 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
468 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
473 key
.offset
= (u64
)-1;
474 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
475 if (IS_ERR(new_root
)) {
476 btrfs_abort_transaction(trans
, root
, PTR_ERR(new_root
));
477 ret
= PTR_ERR(new_root
);
481 btrfs_record_root_in_trans(trans
, new_root
);
483 ret
= btrfs_create_subvol_root(trans
, new_root
, new_dirid
);
485 /* We potentially lose an unused inode item here */
486 btrfs_abort_transaction(trans
, root
, ret
);
491 * insert the directory item
493 ret
= btrfs_set_inode_index(dir
, &index
);
495 btrfs_abort_transaction(trans
, root
, ret
);
499 ret
= btrfs_insert_dir_item(trans
, root
,
500 name
, namelen
, dir
, &key
,
501 BTRFS_FT_DIR
, index
);
503 btrfs_abort_transaction(trans
, root
, ret
);
507 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
508 ret
= btrfs_update_inode(trans
, root
, dir
);
511 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
512 objectid
, root
->root_key
.objectid
,
513 btrfs_ino(dir
), index
, name
, namelen
);
519 *async_transid
= trans
->transid
;
520 err
= btrfs_commit_transaction_async(trans
, root
, 1);
522 err
= btrfs_commit_transaction(trans
, root
);
528 d_instantiate(dentry
, btrfs_lookup_dentry(dir
, dentry
));
533 static int create_snapshot(struct btrfs_root
*root
, struct dentry
*dentry
,
534 char *name
, int namelen
, u64
*async_transid
,
535 bool readonly
, struct btrfs_qgroup_inherit
*inherit
)
538 struct btrfs_pending_snapshot
*pending_snapshot
;
539 struct btrfs_trans_handle
*trans
;
545 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
546 if (!pending_snapshot
)
549 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
550 BTRFS_BLOCK_RSV_TEMP
);
551 pending_snapshot
->dentry
= dentry
;
552 pending_snapshot
->root
= root
;
553 pending_snapshot
->readonly
= readonly
;
554 pending_snapshot
->inherit
= inherit
;
556 trans
= btrfs_start_transaction(root
->fs_info
->extent_root
, 6);
558 ret
= PTR_ERR(trans
);
562 ret
= btrfs_snap_reserve_metadata(trans
, pending_snapshot
);
565 spin_lock(&root
->fs_info
->trans_lock
);
566 list_add(&pending_snapshot
->list
,
567 &trans
->transaction
->pending_snapshots
);
568 spin_unlock(&root
->fs_info
->trans_lock
);
570 *async_transid
= trans
->transid
;
571 ret
= btrfs_commit_transaction_async(trans
,
572 root
->fs_info
->extent_root
, 1);
574 ret
= btrfs_commit_transaction(trans
,
575 root
->fs_info
->extent_root
);
578 /* cleanup_transaction has freed this for us */
580 pending_snapshot
= NULL
;
584 ret
= pending_snapshot
->error
;
588 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
592 inode
= btrfs_lookup_dentry(dentry
->d_parent
->d_inode
, dentry
);
594 ret
= PTR_ERR(inode
);
598 d_instantiate(dentry
, inode
);
601 kfree(pending_snapshot
);
605 /* copy of check_sticky in fs/namei.c()
606 * It's inline, so penalty for filesystems that don't use sticky bit is
609 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
611 kuid_t fsuid
= current_fsuid();
613 if (!(dir
->i_mode
& S_ISVTX
))
615 if (uid_eq(inode
->i_uid
, fsuid
))
617 if (uid_eq(dir
->i_uid
, fsuid
))
619 return !capable(CAP_FOWNER
);
622 /* copy of may_delete in fs/namei.c()
623 * Check whether we can remove a link victim from directory dir, check
624 * whether the type of victim is right.
625 * 1. We can't do it if dir is read-only (done in permission())
626 * 2. We should have write and exec permissions on dir
627 * 3. We can't remove anything from append-only dir
628 * 4. We can't do anything with immutable dir (done in permission())
629 * 5. If the sticky bit on dir is set we should either
630 * a. be owner of dir, or
631 * b. be owner of victim, or
632 * c. have CAP_FOWNER capability
633 * 6. If the victim is append-only or immutable we can't do antyhing with
634 * links pointing to it.
635 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
636 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
637 * 9. We can't remove a root or mountpoint.
638 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
639 * nfs_async_unlink().
642 static int btrfs_may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
646 if (!victim
->d_inode
)
649 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
650 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
652 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
657 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
658 IS_APPEND(victim
->d_inode
)||
659 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
662 if (!S_ISDIR(victim
->d_inode
->i_mode
))
666 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
670 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
675 /* copy of may_create in fs/namei.c() */
676 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
682 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
686 * Create a new subvolume below @parent. This is largely modeled after
687 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
688 * inside this filesystem so it's quite a bit simpler.
690 static noinline
int btrfs_mksubvol(struct path
*parent
,
691 char *name
, int namelen
,
692 struct btrfs_root
*snap_src
,
693 u64
*async_transid
, bool readonly
,
694 struct btrfs_qgroup_inherit
*inherit
)
696 struct inode
*dir
= parent
->dentry
->d_inode
;
697 struct dentry
*dentry
;
700 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
702 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
703 error
= PTR_ERR(dentry
);
711 error
= btrfs_may_create(dir
, dentry
);
716 * even if this name doesn't exist, we may get hash collisions.
717 * check for them now when we can safely fail
719 error
= btrfs_check_dir_item_collision(BTRFS_I(dir
)->root
,
725 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
727 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
731 error
= create_snapshot(snap_src
, dentry
, name
, namelen
,
732 async_transid
, readonly
, inherit
);
734 error
= create_subvol(BTRFS_I(dir
)->root
, dentry
,
735 name
, namelen
, async_transid
, inherit
);
738 fsnotify_mkdir(dir
, dentry
);
740 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
744 mutex_unlock(&dir
->i_mutex
);
749 * When we're defragging a range, we don't want to kick it off again
750 * if it is really just waiting for delalloc to send it down.
751 * If we find a nice big extent or delalloc range for the bytes in the
752 * file you want to defrag, we return 0 to let you know to skip this
755 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, int thresh
)
757 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
758 struct extent_map
*em
= NULL
;
759 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
762 read_lock(&em_tree
->lock
);
763 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
764 read_unlock(&em_tree
->lock
);
767 end
= extent_map_end(em
);
769 if (end
- offset
> thresh
)
772 /* if we already have a nice delalloc here, just stop */
774 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
775 thresh
, EXTENT_DELALLOC
, 1);
782 * helper function to walk through a file and find extents
783 * newer than a specific transid, and smaller than thresh.
785 * This is used by the defragging code to find new and small
788 static int find_new_extents(struct btrfs_root
*root
,
789 struct inode
*inode
, u64 newer_than
,
790 u64
*off
, int thresh
)
792 struct btrfs_path
*path
;
793 struct btrfs_key min_key
;
794 struct btrfs_key max_key
;
795 struct extent_buffer
*leaf
;
796 struct btrfs_file_extent_item
*extent
;
799 u64 ino
= btrfs_ino(inode
);
801 path
= btrfs_alloc_path();
805 min_key
.objectid
= ino
;
806 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
807 min_key
.offset
= *off
;
809 max_key
.objectid
= ino
;
810 max_key
.type
= (u8
)-1;
811 max_key
.offset
= (u64
)-1;
813 path
->keep_locks
= 1;
816 ret
= btrfs_search_forward(root
, &min_key
, &max_key
,
820 if (min_key
.objectid
!= ino
)
822 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
825 leaf
= path
->nodes
[0];
826 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
827 struct btrfs_file_extent_item
);
829 type
= btrfs_file_extent_type(leaf
, extent
);
830 if (type
== BTRFS_FILE_EXTENT_REG
&&
831 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
832 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
833 *off
= min_key
.offset
;
834 btrfs_free_path(path
);
838 if (min_key
.offset
== (u64
)-1)
842 btrfs_release_path(path
);
845 btrfs_free_path(path
);
849 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
851 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
852 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
853 struct extent_map
*em
;
854 u64 len
= PAGE_CACHE_SIZE
;
857 * hopefully we have this extent in the tree already, try without
858 * the full extent lock
860 read_lock(&em_tree
->lock
);
861 em
= lookup_extent_mapping(em_tree
, start
, len
);
862 read_unlock(&em_tree
->lock
);
865 /* get the big lock and read metadata off disk */
866 lock_extent(io_tree
, start
, start
+ len
- 1);
867 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
868 unlock_extent(io_tree
, start
, start
+ len
- 1);
877 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
879 struct extent_map
*next
;
882 /* this is the last extent */
883 if (em
->start
+ em
->len
>= i_size_read(inode
))
886 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
887 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
890 free_extent_map(next
);
894 static int should_defrag_range(struct inode
*inode
, u64 start
, int thresh
,
895 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
898 struct extent_map
*em
;
900 bool next_mergeable
= true;
903 * make sure that once we start defragging an extent, we keep on
906 if (start
< *defrag_end
)
911 em
= defrag_lookup_extent(inode
, start
);
915 /* this will cover holes, and inline extents */
916 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
921 next_mergeable
= defrag_check_next_extent(inode
, em
);
924 * we hit a real extent, if it is big or the next extent is not a
925 * real extent, don't bother defragging it
927 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
928 (em
->len
>= thresh
|| !next_mergeable
))
932 * last_len ends up being a counter of how many bytes we've defragged.
933 * every time we choose not to defrag an extent, we reset *last_len
934 * so that the next tiny extent will force a defrag.
936 * The end result of this is that tiny extents before a single big
937 * extent will force at least part of that big extent to be defragged.
940 *defrag_end
= extent_map_end(em
);
943 *skip
= extent_map_end(em
);
952 * it doesn't do much good to defrag one or two pages
953 * at a time. This pulls in a nice chunk of pages
956 * It also makes sure the delalloc code has enough
957 * dirty data to avoid making new small extents as part
960 * It's a good idea to start RA on this range
961 * before calling this.
963 static int cluster_pages_for_defrag(struct inode
*inode
,
965 unsigned long start_index
,
968 unsigned long file_end
;
969 u64 isize
= i_size_read(inode
);
976 struct btrfs_ordered_extent
*ordered
;
977 struct extent_state
*cached_state
= NULL
;
978 struct extent_io_tree
*tree
;
979 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
981 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
982 if (!isize
|| start_index
> file_end
)
985 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
987 ret
= btrfs_delalloc_reserve_space(inode
,
988 page_cnt
<< PAGE_CACHE_SHIFT
);
992 tree
= &BTRFS_I(inode
)->io_tree
;
994 /* step one, lock all the pages */
995 for (i
= 0; i
< page_cnt
; i
++) {
998 page
= find_or_create_page(inode
->i_mapping
,
999 start_index
+ i
, mask
);
1003 page_start
= page_offset(page
);
1004 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
1006 lock_extent(tree
, page_start
, page_end
);
1007 ordered
= btrfs_lookup_ordered_extent(inode
,
1009 unlock_extent(tree
, page_start
, page_end
);
1014 btrfs_start_ordered_extent(inode
, ordered
, 1);
1015 btrfs_put_ordered_extent(ordered
);
1018 * we unlocked the page above, so we need check if
1019 * it was released or not.
1021 if (page
->mapping
!= inode
->i_mapping
) {
1023 page_cache_release(page
);
1028 if (!PageUptodate(page
)) {
1029 btrfs_readpage(NULL
, page
);
1031 if (!PageUptodate(page
)) {
1033 page_cache_release(page
);
1039 if (page
->mapping
!= inode
->i_mapping
) {
1041 page_cache_release(page
);
1051 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1055 * so now we have a nice long stream of locked
1056 * and up to date pages, lets wait on them
1058 for (i
= 0; i
< i_done
; i
++)
1059 wait_on_page_writeback(pages
[i
]);
1061 page_start
= page_offset(pages
[0]);
1062 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1064 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1065 page_start
, page_end
- 1, 0, &cached_state
);
1066 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1067 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1068 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1069 &cached_state
, GFP_NOFS
);
1071 if (i_done
!= page_cnt
) {
1072 spin_lock(&BTRFS_I(inode
)->lock
);
1073 BTRFS_I(inode
)->outstanding_extents
++;
1074 spin_unlock(&BTRFS_I(inode
)->lock
);
1075 btrfs_delalloc_release_space(inode
,
1076 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1080 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1081 &cached_state
, GFP_NOFS
);
1083 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1084 page_start
, page_end
- 1, &cached_state
,
1087 for (i
= 0; i
< i_done
; i
++) {
1088 clear_page_dirty_for_io(pages
[i
]);
1089 ClearPageChecked(pages
[i
]);
1090 set_page_extent_mapped(pages
[i
]);
1091 set_page_dirty(pages
[i
]);
1092 unlock_page(pages
[i
]);
1093 page_cache_release(pages
[i
]);
1097 for (i
= 0; i
< i_done
; i
++) {
1098 unlock_page(pages
[i
]);
1099 page_cache_release(pages
[i
]);
1101 btrfs_delalloc_release_space(inode
, page_cnt
<< PAGE_CACHE_SHIFT
);
1106 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1107 struct btrfs_ioctl_defrag_range_args
*range
,
1108 u64 newer_than
, unsigned long max_to_defrag
)
1110 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1111 struct file_ra_state
*ra
= NULL
;
1112 unsigned long last_index
;
1113 u64 isize
= i_size_read(inode
);
1117 u64 newer_off
= range
->start
;
1119 unsigned long ra_index
= 0;
1121 int defrag_count
= 0;
1122 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1123 int extent_thresh
= range
->extent_thresh
;
1124 int max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
1125 int cluster
= max_cluster
;
1126 u64 new_align
= ~((u64
)128 * 1024 - 1);
1127 struct page
**pages
= NULL
;
1129 if (extent_thresh
== 0)
1130 extent_thresh
= 256 * 1024;
1132 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1133 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1135 if (range
->compress_type
)
1136 compress_type
= range
->compress_type
;
1143 * if we were not given a file, allocate a readahead
1147 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1150 file_ra_state_init(ra
, inode
->i_mapping
);
1155 pages
= kmalloc(sizeof(struct page
*) * max_cluster
,
1162 /* find the last page to defrag */
1163 if (range
->start
+ range
->len
> range
->start
) {
1164 last_index
= min_t(u64
, isize
- 1,
1165 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1167 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1171 ret
= find_new_extents(root
, inode
, newer_than
,
1172 &newer_off
, 64 * 1024);
1174 range
->start
= newer_off
;
1176 * we always align our defrag to help keep
1177 * the extents in the file evenly spaced
1179 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1183 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1186 max_to_defrag
= last_index
+ 1;
1189 * make writeback starts from i, so the defrag range can be
1190 * written sequentially.
1192 if (i
< inode
->i_mapping
->writeback_index
)
1193 inode
->i_mapping
->writeback_index
= i
;
1195 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1196 (i
< (i_size_read(inode
) + PAGE_CACHE_SIZE
- 1) >>
1197 PAGE_CACHE_SHIFT
)) {
1199 * make sure we stop running if someone unmounts
1202 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1205 if (btrfs_defrag_cancelled(root
->fs_info
)) {
1206 printk(KERN_DEBUG
"btrfs: defrag_file cancelled\n");
1211 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1212 extent_thresh
, &last_len
, &skip
,
1213 &defrag_end
, range
->flags
&
1214 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1217 * the should_defrag function tells us how much to skip
1218 * bump our counter by the suggested amount
1220 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
1221 i
= max(i
+ 1, next
);
1226 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1227 PAGE_CACHE_SHIFT
) - i
;
1228 cluster
= min(cluster
, max_cluster
);
1230 cluster
= max_cluster
;
1233 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1234 BTRFS_I(inode
)->force_compress
= compress_type
;
1236 if (i
+ cluster
> ra_index
) {
1237 ra_index
= max(i
, ra_index
);
1238 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1240 ra_index
+= max_cluster
;
1243 mutex_lock(&inode
->i_mutex
);
1244 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1246 mutex_unlock(&inode
->i_mutex
);
1250 defrag_count
+= ret
;
1251 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, ret
);
1252 mutex_unlock(&inode
->i_mutex
);
1255 if (newer_off
== (u64
)-1)
1261 newer_off
= max(newer_off
+ 1,
1262 (u64
)i
<< PAGE_CACHE_SHIFT
);
1264 ret
= find_new_extents(root
, inode
,
1265 newer_than
, &newer_off
,
1268 range
->start
= newer_off
;
1269 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1276 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1284 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
))
1285 filemap_flush(inode
->i_mapping
);
1287 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1288 /* the filemap_flush will queue IO into the worker threads, but
1289 * we have to make sure the IO is actually started and that
1290 * ordered extents get created before we return
1292 atomic_inc(&root
->fs_info
->async_submit_draining
);
1293 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1294 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1295 wait_event(root
->fs_info
->async_submit_wait
,
1296 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1297 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1299 atomic_dec(&root
->fs_info
->async_submit_draining
);
1301 mutex_lock(&inode
->i_mutex
);
1302 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1303 mutex_unlock(&inode
->i_mutex
);
1306 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1307 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1319 static noinline
int btrfs_ioctl_resize(struct file
*file
,
1325 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
1326 struct btrfs_ioctl_vol_args
*vol_args
;
1327 struct btrfs_trans_handle
*trans
;
1328 struct btrfs_device
*device
= NULL
;
1330 char *devstr
= NULL
;
1334 if (!capable(CAP_SYS_ADMIN
))
1337 ret
= mnt_want_write_file(file
);
1341 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1343 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1344 mnt_drop_write_file(file
);
1348 mutex_lock(&root
->fs_info
->volume_mutex
);
1349 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1350 if (IS_ERR(vol_args
)) {
1351 ret
= PTR_ERR(vol_args
);
1355 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1357 sizestr
= vol_args
->name
;
1358 devstr
= strchr(sizestr
, ':');
1361 sizestr
= devstr
+ 1;
1363 devstr
= vol_args
->name
;
1364 devid
= simple_strtoull(devstr
, &end
, 10);
1369 printk(KERN_INFO
"btrfs: resizing devid %llu\n",
1370 (unsigned long long)devid
);
1373 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1375 printk(KERN_INFO
"btrfs: resizer unable to find device %llu\n",
1376 (unsigned long long)devid
);
1381 if (!device
->writeable
) {
1382 printk(KERN_INFO
"btrfs: resizer unable to apply on "
1383 "readonly device %llu\n",
1384 (unsigned long long)devid
);
1389 if (!strcmp(sizestr
, "max"))
1390 new_size
= device
->bdev
->bd_inode
->i_size
;
1392 if (sizestr
[0] == '-') {
1395 } else if (sizestr
[0] == '+') {
1399 new_size
= memparse(sizestr
, NULL
);
1400 if (new_size
== 0) {
1406 if (device
->is_tgtdev_for_dev_replace
) {
1411 old_size
= device
->total_bytes
;
1414 if (new_size
> old_size
) {
1418 new_size
= old_size
- new_size
;
1419 } else if (mod
> 0) {
1420 new_size
= old_size
+ new_size
;
1423 if (new_size
< 256 * 1024 * 1024) {
1427 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1432 do_div(new_size
, root
->sectorsize
);
1433 new_size
*= root
->sectorsize
;
1435 printk_in_rcu(KERN_INFO
"btrfs: new size for %s is %llu\n",
1436 rcu_str_deref(device
->name
),
1437 (unsigned long long)new_size
);
1439 if (new_size
> old_size
) {
1440 trans
= btrfs_start_transaction(root
, 0);
1441 if (IS_ERR(trans
)) {
1442 ret
= PTR_ERR(trans
);
1445 ret
= btrfs_grow_device(trans
, device
, new_size
);
1446 btrfs_commit_transaction(trans
, root
);
1447 } else if (new_size
< old_size
) {
1448 ret
= btrfs_shrink_device(device
, new_size
);
1449 } /* equal, nothing need to do */
1454 mutex_unlock(&root
->fs_info
->volume_mutex
);
1455 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1456 mnt_drop_write_file(file
);
1460 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1461 char *name
, unsigned long fd
, int subvol
,
1462 u64
*transid
, bool readonly
,
1463 struct btrfs_qgroup_inherit
*inherit
)
1468 ret
= mnt_want_write_file(file
);
1472 namelen
= strlen(name
);
1473 if (strchr(name
, '/')) {
1475 goto out_drop_write
;
1478 if (name
[0] == '.' &&
1479 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1481 goto out_drop_write
;
1485 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1486 NULL
, transid
, readonly
, inherit
);
1488 struct fd src
= fdget(fd
);
1489 struct inode
*src_inode
;
1492 goto out_drop_write
;
1495 src_inode
= src
.file
->f_path
.dentry
->d_inode
;
1496 if (src_inode
->i_sb
!= file
->f_path
.dentry
->d_inode
->i_sb
) {
1497 printk(KERN_INFO
"btrfs: Snapshot src from "
1501 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1502 BTRFS_I(src_inode
)->root
,
1503 transid
, readonly
, inherit
);
1508 mnt_drop_write_file(file
);
1513 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1514 void __user
*arg
, int subvol
)
1516 struct btrfs_ioctl_vol_args
*vol_args
;
1519 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1520 if (IS_ERR(vol_args
))
1521 return PTR_ERR(vol_args
);
1522 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1524 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1525 vol_args
->fd
, subvol
,
1532 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1533 void __user
*arg
, int subvol
)
1535 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1539 bool readonly
= false;
1540 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1542 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1543 if (IS_ERR(vol_args
))
1544 return PTR_ERR(vol_args
);
1545 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1547 if (vol_args
->flags
&
1548 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1549 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1554 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1556 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1558 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1559 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1563 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1564 if (IS_ERR(inherit
)) {
1565 ret
= PTR_ERR(inherit
);
1570 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1571 vol_args
->fd
, subvol
, ptr
,
1574 if (ret
== 0 && ptr
&&
1576 offsetof(struct btrfs_ioctl_vol_args_v2
,
1577 transid
), ptr
, sizeof(*ptr
)))
1585 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1588 struct inode
*inode
= fdentry(file
)->d_inode
;
1589 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1593 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1596 down_read(&root
->fs_info
->subvol_sem
);
1597 if (btrfs_root_readonly(root
))
1598 flags
|= BTRFS_SUBVOL_RDONLY
;
1599 up_read(&root
->fs_info
->subvol_sem
);
1601 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1607 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1610 struct inode
*inode
= fdentry(file
)->d_inode
;
1611 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1612 struct btrfs_trans_handle
*trans
;
1617 ret
= mnt_want_write_file(file
);
1621 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1623 goto out_drop_write
;
1626 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1628 goto out_drop_write
;
1631 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1633 goto out_drop_write
;
1636 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1638 goto out_drop_write
;
1641 if (!inode_owner_or_capable(inode
)) {
1643 goto out_drop_write
;
1646 down_write(&root
->fs_info
->subvol_sem
);
1649 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1652 root_flags
= btrfs_root_flags(&root
->root_item
);
1653 if (flags
& BTRFS_SUBVOL_RDONLY
)
1654 btrfs_set_root_flags(&root
->root_item
,
1655 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1657 btrfs_set_root_flags(&root
->root_item
,
1658 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1660 trans
= btrfs_start_transaction(root
, 1);
1661 if (IS_ERR(trans
)) {
1662 ret
= PTR_ERR(trans
);
1666 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1667 &root
->root_key
, &root
->root_item
);
1669 btrfs_commit_transaction(trans
, root
);
1672 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1674 up_write(&root
->fs_info
->subvol_sem
);
1676 mnt_drop_write_file(file
);
1682 * helper to check if the subvolume references other subvolumes
1684 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1686 struct btrfs_path
*path
;
1687 struct btrfs_key key
;
1690 path
= btrfs_alloc_path();
1694 key
.objectid
= root
->root_key
.objectid
;
1695 key
.type
= BTRFS_ROOT_REF_KEY
;
1696 key
.offset
= (u64
)-1;
1698 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1705 if (path
->slots
[0] > 0) {
1707 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1708 if (key
.objectid
== root
->root_key
.objectid
&&
1709 key
.type
== BTRFS_ROOT_REF_KEY
)
1713 btrfs_free_path(path
);
1717 static noinline
int key_in_sk(struct btrfs_key
*key
,
1718 struct btrfs_ioctl_search_key
*sk
)
1720 struct btrfs_key test
;
1723 test
.objectid
= sk
->min_objectid
;
1724 test
.type
= sk
->min_type
;
1725 test
.offset
= sk
->min_offset
;
1727 ret
= btrfs_comp_cpu_keys(key
, &test
);
1731 test
.objectid
= sk
->max_objectid
;
1732 test
.type
= sk
->max_type
;
1733 test
.offset
= sk
->max_offset
;
1735 ret
= btrfs_comp_cpu_keys(key
, &test
);
1741 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1742 struct btrfs_path
*path
,
1743 struct btrfs_key
*key
,
1744 struct btrfs_ioctl_search_key
*sk
,
1746 unsigned long *sk_offset
,
1750 struct extent_buffer
*leaf
;
1751 struct btrfs_ioctl_search_header sh
;
1752 unsigned long item_off
;
1753 unsigned long item_len
;
1759 leaf
= path
->nodes
[0];
1760 slot
= path
->slots
[0];
1761 nritems
= btrfs_header_nritems(leaf
);
1763 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1767 found_transid
= btrfs_header_generation(leaf
);
1769 for (i
= slot
; i
< nritems
; i
++) {
1770 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1771 item_len
= btrfs_item_size_nr(leaf
, i
);
1773 if (item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1776 if (sizeof(sh
) + item_len
+ *sk_offset
>
1777 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1782 btrfs_item_key_to_cpu(leaf
, key
, i
);
1783 if (!key_in_sk(key
, sk
))
1786 sh
.objectid
= key
->objectid
;
1787 sh
.offset
= key
->offset
;
1788 sh
.type
= key
->type
;
1790 sh
.transid
= found_transid
;
1792 /* copy search result header */
1793 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1794 *sk_offset
+= sizeof(sh
);
1797 char *p
= buf
+ *sk_offset
;
1799 read_extent_buffer(leaf
, p
,
1800 item_off
, item_len
);
1801 *sk_offset
+= item_len
;
1805 if (*num_found
>= sk
->nr_items
)
1810 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1812 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1815 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
1825 static noinline
int search_ioctl(struct inode
*inode
,
1826 struct btrfs_ioctl_search_args
*args
)
1828 struct btrfs_root
*root
;
1829 struct btrfs_key key
;
1830 struct btrfs_key max_key
;
1831 struct btrfs_path
*path
;
1832 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
1833 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
1836 unsigned long sk_offset
= 0;
1838 path
= btrfs_alloc_path();
1842 if (sk
->tree_id
== 0) {
1843 /* search the root of the inode that was passed */
1844 root
= BTRFS_I(inode
)->root
;
1846 key
.objectid
= sk
->tree_id
;
1847 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1848 key
.offset
= (u64
)-1;
1849 root
= btrfs_read_fs_root_no_name(info
, &key
);
1851 printk(KERN_ERR
"could not find root %llu\n",
1853 btrfs_free_path(path
);
1858 key
.objectid
= sk
->min_objectid
;
1859 key
.type
= sk
->min_type
;
1860 key
.offset
= sk
->min_offset
;
1862 max_key
.objectid
= sk
->max_objectid
;
1863 max_key
.type
= sk
->max_type
;
1864 max_key
.offset
= sk
->max_offset
;
1866 path
->keep_locks
= 1;
1869 ret
= btrfs_search_forward(root
, &key
, &max_key
, path
,
1876 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
1877 &sk_offset
, &num_found
);
1878 btrfs_release_path(path
);
1879 if (ret
|| num_found
>= sk
->nr_items
)
1885 sk
->nr_items
= num_found
;
1886 btrfs_free_path(path
);
1890 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
1893 struct btrfs_ioctl_search_args
*args
;
1894 struct inode
*inode
;
1897 if (!capable(CAP_SYS_ADMIN
))
1900 args
= memdup_user(argp
, sizeof(*args
));
1902 return PTR_ERR(args
);
1904 inode
= fdentry(file
)->d_inode
;
1905 ret
= search_ioctl(inode
, args
);
1906 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1913 * Search INODE_REFs to identify path name of 'dirid' directory
1914 * in a 'tree_id' tree. and sets path name to 'name'.
1916 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1917 u64 tree_id
, u64 dirid
, char *name
)
1919 struct btrfs_root
*root
;
1920 struct btrfs_key key
;
1926 struct btrfs_inode_ref
*iref
;
1927 struct extent_buffer
*l
;
1928 struct btrfs_path
*path
;
1930 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1935 path
= btrfs_alloc_path();
1939 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
1941 key
.objectid
= tree_id
;
1942 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1943 key
.offset
= (u64
)-1;
1944 root
= btrfs_read_fs_root_no_name(info
, &key
);
1946 printk(KERN_ERR
"could not find root %llu\n", tree_id
);
1951 key
.objectid
= dirid
;
1952 key
.type
= BTRFS_INODE_REF_KEY
;
1953 key
.offset
= (u64
)-1;
1956 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1961 slot
= path
->slots
[0];
1962 if (ret
> 0 && slot
> 0)
1964 btrfs_item_key_to_cpu(l
, &key
, slot
);
1966 if (ret
> 0 && (key
.objectid
!= dirid
||
1967 key
.type
!= BTRFS_INODE_REF_KEY
)) {
1972 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
1973 len
= btrfs_inode_ref_name_len(l
, iref
);
1975 total_len
+= len
+ 1;
1980 read_extent_buffer(l
, ptr
,(unsigned long)(iref
+ 1), len
);
1982 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
1985 btrfs_release_path(path
);
1986 key
.objectid
= key
.offset
;
1987 key
.offset
= (u64
)-1;
1988 dirid
= key
.objectid
;
1992 memmove(name
, ptr
, total_len
);
1993 name
[total_len
]='\0';
1996 btrfs_free_path(path
);
2000 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
2003 struct btrfs_ioctl_ino_lookup_args
*args
;
2004 struct inode
*inode
;
2007 if (!capable(CAP_SYS_ADMIN
))
2010 args
= memdup_user(argp
, sizeof(*args
));
2012 return PTR_ERR(args
);
2014 inode
= fdentry(file
)->d_inode
;
2016 if (args
->treeid
== 0)
2017 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
2019 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
2020 args
->treeid
, args
->objectid
,
2023 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2030 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2033 struct dentry
*parent
= fdentry(file
);
2034 struct dentry
*dentry
;
2035 struct inode
*dir
= parent
->d_inode
;
2036 struct inode
*inode
;
2037 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2038 struct btrfs_root
*dest
= NULL
;
2039 struct btrfs_ioctl_vol_args
*vol_args
;
2040 struct btrfs_trans_handle
*trans
;
2045 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2046 if (IS_ERR(vol_args
))
2047 return PTR_ERR(vol_args
);
2049 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2050 namelen
= strlen(vol_args
->name
);
2051 if (strchr(vol_args
->name
, '/') ||
2052 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2057 err
= mnt_want_write_file(file
);
2061 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2062 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2063 if (IS_ERR(dentry
)) {
2064 err
= PTR_ERR(dentry
);
2065 goto out_unlock_dir
;
2068 if (!dentry
->d_inode
) {
2073 inode
= dentry
->d_inode
;
2074 dest
= BTRFS_I(inode
)->root
;
2075 if (!capable(CAP_SYS_ADMIN
)){
2077 * Regular user. Only allow this with a special mount
2078 * option, when the user has write+exec access to the
2079 * subvol root, and when rmdir(2) would have been
2082 * Note that this is _not_ check that the subvol is
2083 * empty or doesn't contain data that we wouldn't
2084 * otherwise be able to delete.
2086 * Users who want to delete empty subvols should try
2090 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2094 * Do not allow deletion if the parent dir is the same
2095 * as the dir to be deleted. That means the ioctl
2096 * must be called on the dentry referencing the root
2097 * of the subvol, not a random directory contained
2104 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2109 /* check if subvolume may be deleted by a user */
2110 err
= btrfs_may_delete(dir
, dentry
, 1);
2114 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2119 mutex_lock(&inode
->i_mutex
);
2120 err
= d_invalidate(dentry
);
2124 down_write(&root
->fs_info
->subvol_sem
);
2126 err
= may_destroy_subvol(dest
);
2130 trans
= btrfs_start_transaction(root
, 0);
2131 if (IS_ERR(trans
)) {
2132 err
= PTR_ERR(trans
);
2135 trans
->block_rsv
= &root
->fs_info
->global_block_rsv
;
2137 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2138 dest
->root_key
.objectid
,
2139 dentry
->d_name
.name
,
2140 dentry
->d_name
.len
);
2143 btrfs_abort_transaction(trans
, root
, ret
);
2147 btrfs_record_root_in_trans(trans
, dest
);
2149 memset(&dest
->root_item
.drop_progress
, 0,
2150 sizeof(dest
->root_item
.drop_progress
));
2151 dest
->root_item
.drop_level
= 0;
2152 btrfs_set_root_refs(&dest
->root_item
, 0);
2154 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
2155 ret
= btrfs_insert_orphan_item(trans
,
2156 root
->fs_info
->tree_root
,
2157 dest
->root_key
.objectid
);
2159 btrfs_abort_transaction(trans
, root
, ret
);
2165 ret
= btrfs_end_transaction(trans
, root
);
2168 inode
->i_flags
|= S_DEAD
;
2170 up_write(&root
->fs_info
->subvol_sem
);
2172 mutex_unlock(&inode
->i_mutex
);
2174 shrink_dcache_sb(root
->fs_info
->sb
);
2175 btrfs_invalidate_inodes(dest
);
2179 if (dest
->cache_inode
) {
2180 iput(dest
->cache_inode
);
2181 dest
->cache_inode
= NULL
;
2187 mutex_unlock(&dir
->i_mutex
);
2188 mnt_drop_write_file(file
);
2194 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2196 struct inode
*inode
= fdentry(file
)->d_inode
;
2197 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2198 struct btrfs_ioctl_defrag_range_args
*range
;
2201 ret
= mnt_want_write_file(file
);
2205 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2207 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2208 mnt_drop_write_file(file
);
2212 if (btrfs_root_readonly(root
)) {
2217 switch (inode
->i_mode
& S_IFMT
) {
2219 if (!capable(CAP_SYS_ADMIN
)) {
2223 ret
= btrfs_defrag_root(root
);
2226 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
);
2229 if (!(file
->f_mode
& FMODE_WRITE
)) {
2234 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2241 if (copy_from_user(range
, argp
,
2247 /* compression requires us to start the IO */
2248 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2249 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2250 range
->extent_thresh
= (u32
)-1;
2253 /* the rest are all set to zero by kzalloc */
2254 range
->len
= (u64
)-1;
2256 ret
= btrfs_defrag_file(fdentry(file
)->d_inode
, file
,
2266 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2267 mnt_drop_write_file(file
);
2271 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2273 struct btrfs_ioctl_vol_args
*vol_args
;
2276 if (!capable(CAP_SYS_ADMIN
))
2279 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2281 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2285 mutex_lock(&root
->fs_info
->volume_mutex
);
2286 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2287 if (IS_ERR(vol_args
)) {
2288 ret
= PTR_ERR(vol_args
);
2292 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2293 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2297 mutex_unlock(&root
->fs_info
->volume_mutex
);
2298 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2302 static long btrfs_ioctl_rm_dev(struct file
*file
, void __user
*arg
)
2304 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
2305 struct btrfs_ioctl_vol_args
*vol_args
;
2308 if (!capable(CAP_SYS_ADMIN
))
2311 ret
= mnt_want_write_file(file
);
2315 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2317 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2318 mnt_drop_write_file(file
);
2322 mutex_lock(&root
->fs_info
->volume_mutex
);
2323 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2324 if (IS_ERR(vol_args
)) {
2325 ret
= PTR_ERR(vol_args
);
2329 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2330 ret
= btrfs_rm_device(root
, vol_args
->name
);
2334 mutex_unlock(&root
->fs_info
->volume_mutex
);
2335 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2336 mnt_drop_write_file(file
);
2340 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2342 struct btrfs_ioctl_fs_info_args
*fi_args
;
2343 struct btrfs_device
*device
;
2344 struct btrfs_device
*next
;
2345 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2348 if (!capable(CAP_SYS_ADMIN
))
2351 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2355 fi_args
->num_devices
= fs_devices
->num_devices
;
2356 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2358 mutex_lock(&fs_devices
->device_list_mutex
);
2359 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2360 if (device
->devid
> fi_args
->max_id
)
2361 fi_args
->max_id
= device
->devid
;
2363 mutex_unlock(&fs_devices
->device_list_mutex
);
2365 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2372 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2374 struct btrfs_ioctl_dev_info_args
*di_args
;
2375 struct btrfs_device
*dev
;
2376 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2378 char *s_uuid
= NULL
;
2379 char empty_uuid
[BTRFS_UUID_SIZE
] = {0};
2381 if (!capable(CAP_SYS_ADMIN
))
2384 di_args
= memdup_user(arg
, sizeof(*di_args
));
2385 if (IS_ERR(di_args
))
2386 return PTR_ERR(di_args
);
2388 if (memcmp(empty_uuid
, di_args
->uuid
, BTRFS_UUID_SIZE
) != 0)
2389 s_uuid
= di_args
->uuid
;
2391 mutex_lock(&fs_devices
->device_list_mutex
);
2392 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2393 mutex_unlock(&fs_devices
->device_list_mutex
);
2400 di_args
->devid
= dev
->devid
;
2401 di_args
->bytes_used
= dev
->bytes_used
;
2402 di_args
->total_bytes
= dev
->total_bytes
;
2403 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2405 struct rcu_string
*name
;
2408 name
= rcu_dereference(dev
->name
);
2409 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2411 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2413 di_args
->path
[0] = '\0';
2417 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2424 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
2425 u64 off
, u64 olen
, u64 destoff
)
2427 struct inode
*inode
= fdentry(file
)->d_inode
;
2428 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2431 struct btrfs_trans_handle
*trans
;
2432 struct btrfs_path
*path
;
2433 struct extent_buffer
*leaf
;
2435 struct btrfs_key key
;
2440 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
2444 * - split compressed inline extents. annoying: we need to
2445 * decompress into destination's address_space (the file offset
2446 * may change, so source mapping won't do), then recompress (or
2447 * otherwise reinsert) a subrange.
2448 * - allow ranges within the same file to be cloned (provided
2449 * they don't overlap)?
2452 /* the destination must be opened for writing */
2453 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
2456 if (btrfs_root_readonly(root
))
2459 ret
= mnt_want_write_file(file
);
2463 src_file
= fdget(srcfd
);
2464 if (!src_file
.file
) {
2466 goto out_drop_write
;
2470 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
2473 src
= src_file
.file
->f_dentry
->d_inode
;
2479 /* the src must be open for reading */
2480 if (!(src_file
.file
->f_mode
& FMODE_READ
))
2483 /* don't make the dst file partly checksummed */
2484 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2485 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
2489 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
2493 if (src
->i_sb
!= inode
->i_sb
)
2497 buf
= vmalloc(btrfs_level_size(root
, 0));
2501 path
= btrfs_alloc_path();
2509 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
2510 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
2512 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
2513 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
2516 /* determine range to clone */
2518 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
2521 olen
= len
= src
->i_size
- off
;
2522 /* if we extend to eof, continue to block boundary */
2523 if (off
+ len
== src
->i_size
)
2524 len
= ALIGN(src
->i_size
, bs
) - off
;
2526 /* verify the end result is block aligned */
2527 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
2528 !IS_ALIGNED(destoff
, bs
))
2531 if (destoff
> inode
->i_size
) {
2532 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
2537 /* truncate page cache pages from target inode range */
2538 truncate_inode_pages_range(&inode
->i_data
, destoff
,
2539 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
2541 /* do any pending delalloc/csum calc on src, one way or
2542 another, and lock file content */
2544 struct btrfs_ordered_extent
*ordered
;
2545 lock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2546 ordered
= btrfs_lookup_first_ordered_extent(src
, off
+ len
- 1);
2548 !test_range_bit(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1,
2549 EXTENT_DELALLOC
, 0, NULL
))
2551 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2553 btrfs_put_ordered_extent(ordered
);
2554 btrfs_wait_ordered_range(src
, off
, len
);
2558 key
.objectid
= btrfs_ino(src
);
2559 key
.type
= BTRFS_EXTENT_DATA_KEY
;
2564 * note the key will change type as we walk through the
2567 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
2572 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2573 if (path
->slots
[0] >= nritems
) {
2574 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
2579 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2581 leaf
= path
->nodes
[0];
2582 slot
= path
->slots
[0];
2584 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2585 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
2586 key
.objectid
!= btrfs_ino(src
))
2589 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
2590 struct btrfs_file_extent_item
*extent
;
2593 struct btrfs_key new_key
;
2594 u64 disko
= 0, diskl
= 0;
2595 u64 datao
= 0, datal
= 0;
2599 size
= btrfs_item_size_nr(leaf
, slot
);
2600 read_extent_buffer(leaf
, buf
,
2601 btrfs_item_ptr_offset(leaf
, slot
),
2604 extent
= btrfs_item_ptr(leaf
, slot
,
2605 struct btrfs_file_extent_item
);
2606 comp
= btrfs_file_extent_compression(leaf
, extent
);
2607 type
= btrfs_file_extent_type(leaf
, extent
);
2608 if (type
== BTRFS_FILE_EXTENT_REG
||
2609 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2610 disko
= btrfs_file_extent_disk_bytenr(leaf
,
2612 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
2614 datao
= btrfs_file_extent_offset(leaf
, extent
);
2615 datal
= btrfs_file_extent_num_bytes(leaf
,
2617 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2618 /* take upper bound, may be compressed */
2619 datal
= btrfs_file_extent_ram_bytes(leaf
,
2622 btrfs_release_path(path
);
2624 if (key
.offset
+ datal
<= off
||
2625 key
.offset
>= off
+ len
- 1)
2628 memcpy(&new_key
, &key
, sizeof(new_key
));
2629 new_key
.objectid
= btrfs_ino(inode
);
2630 if (off
<= key
.offset
)
2631 new_key
.offset
= key
.offset
+ destoff
- off
;
2633 new_key
.offset
= destoff
;
2636 * 1 - adjusting old extent (we may have to split it)
2637 * 1 - add new extent
2640 trans
= btrfs_start_transaction(root
, 3);
2641 if (IS_ERR(trans
)) {
2642 ret
= PTR_ERR(trans
);
2646 if (type
== BTRFS_FILE_EXTENT_REG
||
2647 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2649 * a | --- range to clone ---| b
2650 * | ------------- extent ------------- |
2653 /* substract range b */
2654 if (key
.offset
+ datal
> off
+ len
)
2655 datal
= off
+ len
- key
.offset
;
2657 /* substract range a */
2658 if (off
> key
.offset
) {
2659 datao
+= off
- key
.offset
;
2660 datal
-= off
- key
.offset
;
2663 ret
= btrfs_drop_extents(trans
, root
, inode
,
2665 new_key
.offset
+ datal
,
2668 btrfs_abort_transaction(trans
, root
,
2670 btrfs_end_transaction(trans
, root
);
2674 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2677 btrfs_abort_transaction(trans
, root
,
2679 btrfs_end_transaction(trans
, root
);
2683 leaf
= path
->nodes
[0];
2684 slot
= path
->slots
[0];
2685 write_extent_buffer(leaf
, buf
,
2686 btrfs_item_ptr_offset(leaf
, slot
),
2689 extent
= btrfs_item_ptr(leaf
, slot
,
2690 struct btrfs_file_extent_item
);
2692 /* disko == 0 means it's a hole */
2696 btrfs_set_file_extent_offset(leaf
, extent
,
2698 btrfs_set_file_extent_num_bytes(leaf
, extent
,
2701 inode_add_bytes(inode
, datal
);
2702 ret
= btrfs_inc_extent_ref(trans
, root
,
2704 root
->root_key
.objectid
,
2706 new_key
.offset
- datao
,
2709 btrfs_abort_transaction(trans
,
2712 btrfs_end_transaction(trans
,
2718 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2721 if (off
> key
.offset
) {
2722 skip
= off
- key
.offset
;
2723 new_key
.offset
+= skip
;
2726 if (key
.offset
+ datal
> off
+ len
)
2727 trim
= key
.offset
+ datal
- (off
+ len
);
2729 if (comp
&& (skip
|| trim
)) {
2731 btrfs_end_transaction(trans
, root
);
2734 size
-= skip
+ trim
;
2735 datal
-= skip
+ trim
;
2737 ret
= btrfs_drop_extents(trans
, root
, inode
,
2739 new_key
.offset
+ datal
,
2742 btrfs_abort_transaction(trans
, root
,
2744 btrfs_end_transaction(trans
, root
);
2748 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2751 btrfs_abort_transaction(trans
, root
,
2753 btrfs_end_transaction(trans
, root
);
2759 btrfs_file_extent_calc_inline_size(0);
2760 memmove(buf
+start
, buf
+start
+skip
,
2764 leaf
= path
->nodes
[0];
2765 slot
= path
->slots
[0];
2766 write_extent_buffer(leaf
, buf
,
2767 btrfs_item_ptr_offset(leaf
, slot
),
2769 inode_add_bytes(inode
, datal
);
2772 btrfs_mark_buffer_dirty(leaf
);
2773 btrfs_release_path(path
);
2775 inode_inc_iversion(inode
);
2776 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2779 * we round up to the block size at eof when
2780 * determining which extents to clone above,
2781 * but shouldn't round up the file size
2783 endoff
= new_key
.offset
+ datal
;
2784 if (endoff
> destoff
+olen
)
2785 endoff
= destoff
+olen
;
2786 if (endoff
> inode
->i_size
)
2787 btrfs_i_size_write(inode
, endoff
);
2789 ret
= btrfs_update_inode(trans
, root
, inode
);
2791 btrfs_abort_transaction(trans
, root
, ret
);
2792 btrfs_end_transaction(trans
, root
);
2795 ret
= btrfs_end_transaction(trans
, root
);
2798 btrfs_release_path(path
);
2803 btrfs_release_path(path
);
2804 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2806 mutex_unlock(&src
->i_mutex
);
2807 mutex_unlock(&inode
->i_mutex
);
2809 btrfs_free_path(path
);
2813 mnt_drop_write_file(file
);
2817 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
2819 struct btrfs_ioctl_clone_range_args args
;
2821 if (copy_from_user(&args
, argp
, sizeof(args
)))
2823 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
2824 args
.src_length
, args
.dest_offset
);
2828 * there are many ways the trans_start and trans_end ioctls can lead
2829 * to deadlocks. They should only be used by applications that
2830 * basically own the machine, and have a very in depth understanding
2831 * of all the possible deadlocks and enospc problems.
2833 static long btrfs_ioctl_trans_start(struct file
*file
)
2835 struct inode
*inode
= fdentry(file
)->d_inode
;
2836 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2837 struct btrfs_trans_handle
*trans
;
2841 if (!capable(CAP_SYS_ADMIN
))
2845 if (file
->private_data
)
2849 if (btrfs_root_readonly(root
))
2852 ret
= mnt_want_write_file(file
);
2856 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
2859 trans
= btrfs_start_ioctl_transaction(root
);
2863 file
->private_data
= trans
;
2867 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
2868 mnt_drop_write_file(file
);
2873 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2875 struct inode
*inode
= fdentry(file
)->d_inode
;
2876 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2877 struct btrfs_root
*new_root
;
2878 struct btrfs_dir_item
*di
;
2879 struct btrfs_trans_handle
*trans
;
2880 struct btrfs_path
*path
;
2881 struct btrfs_key location
;
2882 struct btrfs_disk_key disk_key
;
2887 if (!capable(CAP_SYS_ADMIN
))
2890 ret
= mnt_want_write_file(file
);
2894 if (copy_from_user(&objectid
, argp
, sizeof(objectid
))) {
2900 objectid
= root
->root_key
.objectid
;
2902 location
.objectid
= objectid
;
2903 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2904 location
.offset
= (u64
)-1;
2906 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
2907 if (IS_ERR(new_root
)) {
2908 ret
= PTR_ERR(new_root
);
2912 if (btrfs_root_refs(&new_root
->root_item
) == 0) {
2917 path
= btrfs_alloc_path();
2922 path
->leave_spinning
= 1;
2924 trans
= btrfs_start_transaction(root
, 1);
2925 if (IS_ERR(trans
)) {
2926 btrfs_free_path(path
);
2927 ret
= PTR_ERR(trans
);
2931 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
2932 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
2933 dir_id
, "default", 7, 1);
2934 if (IS_ERR_OR_NULL(di
)) {
2935 btrfs_free_path(path
);
2936 btrfs_end_transaction(trans
, root
);
2937 printk(KERN_ERR
"Umm, you don't have the default dir item, "
2938 "this isn't going to work\n");
2943 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
2944 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
2945 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2946 btrfs_free_path(path
);
2948 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
2949 btrfs_end_transaction(trans
, root
);
2951 mnt_drop_write_file(file
);
2955 void btrfs_get_block_group_info(struct list_head
*groups_list
,
2956 struct btrfs_ioctl_space_info
*space
)
2958 struct btrfs_block_group_cache
*block_group
;
2960 space
->total_bytes
= 0;
2961 space
->used_bytes
= 0;
2963 list_for_each_entry(block_group
, groups_list
, list
) {
2964 space
->flags
= block_group
->flags
;
2965 space
->total_bytes
+= block_group
->key
.offset
;
2966 space
->used_bytes
+=
2967 btrfs_block_group_used(&block_group
->item
);
2971 long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
2973 struct btrfs_ioctl_space_args space_args
;
2974 struct btrfs_ioctl_space_info space
;
2975 struct btrfs_ioctl_space_info
*dest
;
2976 struct btrfs_ioctl_space_info
*dest_orig
;
2977 struct btrfs_ioctl_space_info __user
*user_dest
;
2978 struct btrfs_space_info
*info
;
2979 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
2980 BTRFS_BLOCK_GROUP_SYSTEM
,
2981 BTRFS_BLOCK_GROUP_METADATA
,
2982 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
2989 if (copy_from_user(&space_args
,
2990 (struct btrfs_ioctl_space_args __user
*)arg
,
2991 sizeof(space_args
)))
2994 for (i
= 0; i
< num_types
; i
++) {
2995 struct btrfs_space_info
*tmp
;
2999 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3001 if (tmp
->flags
== types
[i
]) {
3011 down_read(&info
->groups_sem
);
3012 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3013 if (!list_empty(&info
->block_groups
[c
]))
3016 up_read(&info
->groups_sem
);
3019 /* space_slots == 0 means they are asking for a count */
3020 if (space_args
.space_slots
== 0) {
3021 space_args
.total_spaces
= slot_count
;
3025 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
3027 alloc_size
= sizeof(*dest
) * slot_count
;
3029 /* we generally have at most 6 or so space infos, one for each raid
3030 * level. So, a whole page should be more than enough for everyone
3032 if (alloc_size
> PAGE_CACHE_SIZE
)
3035 space_args
.total_spaces
= 0;
3036 dest
= kmalloc(alloc_size
, GFP_NOFS
);
3041 /* now we have a buffer to copy into */
3042 for (i
= 0; i
< num_types
; i
++) {
3043 struct btrfs_space_info
*tmp
;
3050 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
3052 if (tmp
->flags
== types
[i
]) {
3061 down_read(&info
->groups_sem
);
3062 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3063 if (!list_empty(&info
->block_groups
[c
])) {
3064 btrfs_get_block_group_info(
3065 &info
->block_groups
[c
], &space
);
3066 memcpy(dest
, &space
, sizeof(space
));
3068 space_args
.total_spaces
++;
3074 up_read(&info
->groups_sem
);
3077 user_dest
= (struct btrfs_ioctl_space_info __user
*)
3078 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
3080 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
3085 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
3092 * there are many ways the trans_start and trans_end ioctls can lead
3093 * to deadlocks. They should only be used by applications that
3094 * basically own the machine, and have a very in depth understanding
3095 * of all the possible deadlocks and enospc problems.
3097 long btrfs_ioctl_trans_end(struct file
*file
)
3099 struct inode
*inode
= fdentry(file
)->d_inode
;
3100 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3101 struct btrfs_trans_handle
*trans
;
3103 trans
= file
->private_data
;
3106 file
->private_data
= NULL
;
3108 btrfs_end_transaction(trans
, root
);
3110 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3112 mnt_drop_write_file(file
);
3116 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
3119 struct btrfs_trans_handle
*trans
;
3123 trans
= btrfs_attach_transaction_barrier(root
);
3124 if (IS_ERR(trans
)) {
3125 if (PTR_ERR(trans
) != -ENOENT
)
3126 return PTR_ERR(trans
);
3128 /* No running transaction, don't bother */
3129 transid
= root
->fs_info
->last_trans_committed
;
3132 transid
= trans
->transid
;
3133 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
3135 btrfs_end_transaction(trans
, root
);
3140 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
3145 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
3151 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
3154 transid
= 0; /* current trans */
3156 return btrfs_wait_for_commit(root
, transid
);
3159 static long btrfs_ioctl_scrub(struct file
*file
, void __user
*arg
)
3161 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3162 struct btrfs_ioctl_scrub_args
*sa
;
3165 if (!capable(CAP_SYS_ADMIN
))
3168 sa
= memdup_user(arg
, sizeof(*sa
));
3172 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
)) {
3173 ret
= mnt_want_write_file(file
);
3178 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
3179 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
3182 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3185 if (!(sa
->flags
& BTRFS_SCRUB_READONLY
))
3186 mnt_drop_write_file(file
);
3192 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
3194 if (!capable(CAP_SYS_ADMIN
))
3197 return btrfs_scrub_cancel(root
->fs_info
);
3200 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
3203 struct btrfs_ioctl_scrub_args
*sa
;
3206 if (!capable(CAP_SYS_ADMIN
))
3209 sa
= memdup_user(arg
, sizeof(*sa
));
3213 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
3215 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3222 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
3225 struct btrfs_ioctl_get_dev_stats
*sa
;
3228 sa
= memdup_user(arg
, sizeof(*sa
));
3232 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
3237 ret
= btrfs_get_dev_stats(root
, sa
);
3239 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3246 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
3248 struct btrfs_ioctl_dev_replace_args
*p
;
3251 if (!capable(CAP_SYS_ADMIN
))
3254 p
= memdup_user(arg
, sizeof(*p
));
3259 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
3261 &root
->fs_info
->mutually_exclusive_operation_running
,
3263 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3266 ret
= btrfs_dev_replace_start(root
, p
);
3268 &root
->fs_info
->mutually_exclusive_operation_running
,
3272 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
3273 btrfs_dev_replace_status(root
->fs_info
, p
);
3276 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
3277 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
3284 if (copy_to_user(arg
, p
, sizeof(*p
)))
3291 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
3297 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
3298 struct inode_fs_paths
*ipath
= NULL
;
3299 struct btrfs_path
*path
;
3301 if (!capable(CAP_DAC_READ_SEARCH
))
3304 path
= btrfs_alloc_path();
3310 ipa
= memdup_user(arg
, sizeof(*ipa
));
3317 size
= min_t(u32
, ipa
->size
, 4096);
3318 ipath
= init_ipath(size
, root
, path
);
3319 if (IS_ERR(ipath
)) {
3320 ret
= PTR_ERR(ipath
);
3325 ret
= paths_from_inode(ipa
->inum
, ipath
);
3329 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
3330 rel_ptr
= ipath
->fspath
->val
[i
] -
3331 (u64
)(unsigned long)ipath
->fspath
->val
;
3332 ipath
->fspath
->val
[i
] = rel_ptr
;
3335 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
3336 (void *)(unsigned long)ipath
->fspath
, size
);
3343 btrfs_free_path(path
);
3350 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
3352 struct btrfs_data_container
*inodes
= ctx
;
3353 const size_t c
= 3 * sizeof(u64
);
3355 if (inodes
->bytes_left
>= c
) {
3356 inodes
->bytes_left
-= c
;
3357 inodes
->val
[inodes
->elem_cnt
] = inum
;
3358 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
3359 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
3360 inodes
->elem_cnt
+= 3;
3362 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
3363 inodes
->bytes_left
= 0;
3364 inodes
->elem_missed
+= 3;
3370 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
3375 struct btrfs_ioctl_logical_ino_args
*loi
;
3376 struct btrfs_data_container
*inodes
= NULL
;
3377 struct btrfs_path
*path
= NULL
;
3379 if (!capable(CAP_SYS_ADMIN
))
3382 loi
= memdup_user(arg
, sizeof(*loi
));
3389 path
= btrfs_alloc_path();
3395 size
= min_t(u32
, loi
->size
, 64 * 1024);
3396 inodes
= init_data_container(size
);
3397 if (IS_ERR(inodes
)) {
3398 ret
= PTR_ERR(inodes
);
3403 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
3404 build_ino_list
, inodes
);
3410 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
3411 (void *)(unsigned long)inodes
, size
);
3416 btrfs_free_path(path
);
3423 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
3424 struct btrfs_ioctl_balance_args
*bargs
)
3426 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
3428 bargs
->flags
= bctl
->flags
;
3430 if (atomic_read(&fs_info
->balance_running
))
3431 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
3432 if (atomic_read(&fs_info
->balance_pause_req
))
3433 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
3434 if (atomic_read(&fs_info
->balance_cancel_req
))
3435 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
3437 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
3438 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
3439 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
3442 spin_lock(&fs_info
->balance_lock
);
3443 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3444 spin_unlock(&fs_info
->balance_lock
);
3446 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3450 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
3452 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3453 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3454 struct btrfs_ioctl_balance_args
*bargs
;
3455 struct btrfs_balance_control
*bctl
;
3456 bool need_unlock
; /* for mut. excl. ops lock */
3459 if (!capable(CAP_SYS_ADMIN
))
3462 ret
= mnt_want_write_file(file
);
3467 if (!atomic_xchg(&fs_info
->mutually_exclusive_operation_running
, 1)) {
3468 mutex_lock(&fs_info
->volume_mutex
);
3469 mutex_lock(&fs_info
->balance_mutex
);
3475 * mut. excl. ops lock is locked. Three possibilites:
3476 * (1) some other op is running
3477 * (2) balance is running
3478 * (3) balance is paused -- special case (think resume)
3480 mutex_lock(&fs_info
->balance_mutex
);
3481 if (fs_info
->balance_ctl
) {
3482 /* this is either (2) or (3) */
3483 if (!atomic_read(&fs_info
->balance_running
)) {
3484 mutex_unlock(&fs_info
->balance_mutex
);
3485 if (!mutex_trylock(&fs_info
->volume_mutex
))
3487 mutex_lock(&fs_info
->balance_mutex
);
3489 if (fs_info
->balance_ctl
&&
3490 !atomic_read(&fs_info
->balance_running
)) {
3492 need_unlock
= false;
3496 mutex_unlock(&fs_info
->balance_mutex
);
3497 mutex_unlock(&fs_info
->volume_mutex
);
3501 mutex_unlock(&fs_info
->balance_mutex
);
3507 mutex_unlock(&fs_info
->balance_mutex
);
3508 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3514 BUG_ON(!atomic_read(&fs_info
->mutually_exclusive_operation_running
));
3517 bargs
= memdup_user(arg
, sizeof(*bargs
));
3518 if (IS_ERR(bargs
)) {
3519 ret
= PTR_ERR(bargs
);
3523 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
3524 if (!fs_info
->balance_ctl
) {
3529 bctl
= fs_info
->balance_ctl
;
3530 spin_lock(&fs_info
->balance_lock
);
3531 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
3532 spin_unlock(&fs_info
->balance_lock
);
3540 if (fs_info
->balance_ctl
) {
3545 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
3551 bctl
->fs_info
= fs_info
;
3553 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
3554 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
3555 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
3557 bctl
->flags
= bargs
->flags
;
3559 /* balance everything - no filters */
3560 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
3565 * Ownership of bctl and mutually_exclusive_operation_running
3566 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3567 * or, if restriper was paused all the way until unmount, in
3568 * free_fs_info. mutually_exclusive_operation_running is
3569 * cleared in __cancel_balance.
3571 need_unlock
= false;
3573 ret
= btrfs_balance(bctl
, bargs
);
3576 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3583 mutex_unlock(&fs_info
->balance_mutex
);
3584 mutex_unlock(&fs_info
->volume_mutex
);
3586 atomic_set(&fs_info
->mutually_exclusive_operation_running
, 0);
3588 mnt_drop_write_file(file
);
3592 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
3594 if (!capable(CAP_SYS_ADMIN
))
3598 case BTRFS_BALANCE_CTL_PAUSE
:
3599 return btrfs_pause_balance(root
->fs_info
);
3600 case BTRFS_BALANCE_CTL_CANCEL
:
3601 return btrfs_cancel_balance(root
->fs_info
);
3607 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
3610 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3611 struct btrfs_ioctl_balance_args
*bargs
;
3614 if (!capable(CAP_SYS_ADMIN
))
3617 mutex_lock(&fs_info
->balance_mutex
);
3618 if (!fs_info
->balance_ctl
) {
3623 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
3629 update_ioctl_balance_args(fs_info
, 1, bargs
);
3631 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3636 mutex_unlock(&fs_info
->balance_mutex
);
3640 static long btrfs_ioctl_quota_ctl(struct file
*file
, void __user
*arg
)
3642 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3643 struct btrfs_ioctl_quota_ctl_args
*sa
;
3644 struct btrfs_trans_handle
*trans
= NULL
;
3648 if (!capable(CAP_SYS_ADMIN
))
3651 ret
= mnt_want_write_file(file
);
3655 sa
= memdup_user(arg
, sizeof(*sa
));
3661 if (sa
->cmd
!= BTRFS_QUOTA_CTL_RESCAN
) {
3662 trans
= btrfs_start_transaction(root
, 2);
3663 if (IS_ERR(trans
)) {
3664 ret
= PTR_ERR(trans
);
3670 case BTRFS_QUOTA_CTL_ENABLE
:
3671 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
3673 case BTRFS_QUOTA_CTL_DISABLE
:
3674 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
3676 case BTRFS_QUOTA_CTL_RESCAN
:
3677 ret
= btrfs_quota_rescan(root
->fs_info
);
3684 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3688 err
= btrfs_commit_transaction(trans
, root
);
3695 mnt_drop_write_file(file
);
3699 static long btrfs_ioctl_qgroup_assign(struct file
*file
, void __user
*arg
)
3701 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3702 struct btrfs_ioctl_qgroup_assign_args
*sa
;
3703 struct btrfs_trans_handle
*trans
;
3707 if (!capable(CAP_SYS_ADMIN
))
3710 ret
= mnt_want_write_file(file
);
3714 sa
= memdup_user(arg
, sizeof(*sa
));
3720 trans
= btrfs_join_transaction(root
);
3721 if (IS_ERR(trans
)) {
3722 ret
= PTR_ERR(trans
);
3726 /* FIXME: check if the IDs really exist */
3728 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
3731 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
3735 err
= btrfs_end_transaction(trans
, root
);
3742 mnt_drop_write_file(file
);
3746 static long btrfs_ioctl_qgroup_create(struct file
*file
, void __user
*arg
)
3748 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3749 struct btrfs_ioctl_qgroup_create_args
*sa
;
3750 struct btrfs_trans_handle
*trans
;
3754 if (!capable(CAP_SYS_ADMIN
))
3757 ret
= mnt_want_write_file(file
);
3761 sa
= memdup_user(arg
, sizeof(*sa
));
3767 if (!sa
->qgroupid
) {
3772 trans
= btrfs_join_transaction(root
);
3773 if (IS_ERR(trans
)) {
3774 ret
= PTR_ERR(trans
);
3778 /* FIXME: check if the IDs really exist */
3780 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
,
3783 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
3786 err
= btrfs_end_transaction(trans
, root
);
3793 mnt_drop_write_file(file
);
3797 static long btrfs_ioctl_qgroup_limit(struct file
*file
, void __user
*arg
)
3799 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3800 struct btrfs_ioctl_qgroup_limit_args
*sa
;
3801 struct btrfs_trans_handle
*trans
;
3806 if (!capable(CAP_SYS_ADMIN
))
3809 ret
= mnt_want_write_file(file
);
3813 sa
= memdup_user(arg
, sizeof(*sa
));
3819 trans
= btrfs_join_transaction(root
);
3820 if (IS_ERR(trans
)) {
3821 ret
= PTR_ERR(trans
);
3825 qgroupid
= sa
->qgroupid
;
3827 /* take the current subvol as qgroup */
3828 qgroupid
= root
->root_key
.objectid
;
3831 /* FIXME: check if the IDs really exist */
3832 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
3834 err
= btrfs_end_transaction(trans
, root
);
3841 mnt_drop_write_file(file
);
3845 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
3848 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
3849 struct inode
*inode
= fdentry(file
)->d_inode
;
3850 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3851 struct btrfs_root_item
*root_item
= &root
->root_item
;
3852 struct btrfs_trans_handle
*trans
;
3853 struct timespec ct
= CURRENT_TIME
;
3856 ret
= mnt_want_write_file(file
);
3860 down_write(&root
->fs_info
->subvol_sem
);
3862 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
3867 if (btrfs_root_readonly(root
)) {
3872 if (!inode_owner_or_capable(inode
)) {
3877 sa
= memdup_user(arg
, sizeof(*sa
));
3884 trans
= btrfs_start_transaction(root
, 1);
3885 if (IS_ERR(trans
)) {
3886 ret
= PTR_ERR(trans
);
3891 sa
->rtransid
= trans
->transid
;
3892 sa
->rtime
.sec
= ct
.tv_sec
;
3893 sa
->rtime
.nsec
= ct
.tv_nsec
;
3895 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
3896 btrfs_set_root_stransid(root_item
, sa
->stransid
);
3897 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
3898 root_item
->stime
.sec
= cpu_to_le64(sa
->stime
.sec
);
3899 root_item
->stime
.nsec
= cpu_to_le32(sa
->stime
.nsec
);
3900 root_item
->rtime
.sec
= cpu_to_le64(sa
->rtime
.sec
);
3901 root_item
->rtime
.nsec
= cpu_to_le32(sa
->rtime
.nsec
);
3903 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
3904 &root
->root_key
, &root
->root_item
);
3906 btrfs_end_transaction(trans
, root
);
3910 ret
= btrfs_commit_transaction(trans
, root
);
3915 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
3921 up_write(&root
->fs_info
->subvol_sem
);
3922 mnt_drop_write_file(file
);
3926 static int btrfs_ioctl_get_fslabel(struct file
*file
, void __user
*arg
)
3928 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3929 const char *label
= root
->fs_info
->super_copy
->label
;
3930 size_t len
= strnlen(label
, BTRFS_LABEL_SIZE
);
3933 if (len
== BTRFS_LABEL_SIZE
) {
3934 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
3938 mutex_lock(&root
->fs_info
->volume_mutex
);
3939 ret
= copy_to_user(arg
, label
, len
);
3940 mutex_unlock(&root
->fs_info
->volume_mutex
);
3942 return ret
? -EFAULT
: 0;
3945 static int btrfs_ioctl_set_fslabel(struct file
*file
, void __user
*arg
)
3947 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3948 struct btrfs_super_block
*super_block
= root
->fs_info
->super_copy
;
3949 struct btrfs_trans_handle
*trans
;
3950 char label
[BTRFS_LABEL_SIZE
];
3953 if (!capable(CAP_SYS_ADMIN
))
3956 if (copy_from_user(label
, arg
, sizeof(label
)))
3959 if (strnlen(label
, BTRFS_LABEL_SIZE
) == BTRFS_LABEL_SIZE
) {
3960 pr_err("btrfs: unable to set label with more than %d bytes\n",
3961 BTRFS_LABEL_SIZE
- 1);
3965 ret
= mnt_want_write_file(file
);
3969 mutex_lock(&root
->fs_info
->volume_mutex
);
3970 trans
= btrfs_start_transaction(root
, 0);
3971 if (IS_ERR(trans
)) {
3972 ret
= PTR_ERR(trans
);
3976 strcpy(super_block
->label
, label
);
3977 ret
= btrfs_end_transaction(trans
, root
);
3980 mutex_unlock(&root
->fs_info
->volume_mutex
);
3981 mnt_drop_write_file(file
);
3985 long btrfs_ioctl(struct file
*file
, unsigned int
3986 cmd
, unsigned long arg
)
3988 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3989 void __user
*argp
= (void __user
*)arg
;
3992 case FS_IOC_GETFLAGS
:
3993 return btrfs_ioctl_getflags(file
, argp
);
3994 case FS_IOC_SETFLAGS
:
3995 return btrfs_ioctl_setflags(file
, argp
);
3996 case FS_IOC_GETVERSION
:
3997 return btrfs_ioctl_getversion(file
, argp
);
3999 return btrfs_ioctl_fitrim(file
, argp
);
4000 case BTRFS_IOC_SNAP_CREATE
:
4001 return btrfs_ioctl_snap_create(file
, argp
, 0);
4002 case BTRFS_IOC_SNAP_CREATE_V2
:
4003 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
4004 case BTRFS_IOC_SUBVOL_CREATE
:
4005 return btrfs_ioctl_snap_create(file
, argp
, 1);
4006 case BTRFS_IOC_SUBVOL_CREATE_V2
:
4007 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
4008 case BTRFS_IOC_SNAP_DESTROY
:
4009 return btrfs_ioctl_snap_destroy(file
, argp
);
4010 case BTRFS_IOC_SUBVOL_GETFLAGS
:
4011 return btrfs_ioctl_subvol_getflags(file
, argp
);
4012 case BTRFS_IOC_SUBVOL_SETFLAGS
:
4013 return btrfs_ioctl_subvol_setflags(file
, argp
);
4014 case BTRFS_IOC_DEFAULT_SUBVOL
:
4015 return btrfs_ioctl_default_subvol(file
, argp
);
4016 case BTRFS_IOC_DEFRAG
:
4017 return btrfs_ioctl_defrag(file
, NULL
);
4018 case BTRFS_IOC_DEFRAG_RANGE
:
4019 return btrfs_ioctl_defrag(file
, argp
);
4020 case BTRFS_IOC_RESIZE
:
4021 return btrfs_ioctl_resize(file
, argp
);
4022 case BTRFS_IOC_ADD_DEV
:
4023 return btrfs_ioctl_add_dev(root
, argp
);
4024 case BTRFS_IOC_RM_DEV
:
4025 return btrfs_ioctl_rm_dev(file
, argp
);
4026 case BTRFS_IOC_FS_INFO
:
4027 return btrfs_ioctl_fs_info(root
, argp
);
4028 case BTRFS_IOC_DEV_INFO
:
4029 return btrfs_ioctl_dev_info(root
, argp
);
4030 case BTRFS_IOC_BALANCE
:
4031 return btrfs_ioctl_balance(file
, NULL
);
4032 case BTRFS_IOC_CLONE
:
4033 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
4034 case BTRFS_IOC_CLONE_RANGE
:
4035 return btrfs_ioctl_clone_range(file
, argp
);
4036 case BTRFS_IOC_TRANS_START
:
4037 return btrfs_ioctl_trans_start(file
);
4038 case BTRFS_IOC_TRANS_END
:
4039 return btrfs_ioctl_trans_end(file
);
4040 case BTRFS_IOC_TREE_SEARCH
:
4041 return btrfs_ioctl_tree_search(file
, argp
);
4042 case BTRFS_IOC_INO_LOOKUP
:
4043 return btrfs_ioctl_ino_lookup(file
, argp
);
4044 case BTRFS_IOC_INO_PATHS
:
4045 return btrfs_ioctl_ino_to_path(root
, argp
);
4046 case BTRFS_IOC_LOGICAL_INO
:
4047 return btrfs_ioctl_logical_to_ino(root
, argp
);
4048 case BTRFS_IOC_SPACE_INFO
:
4049 return btrfs_ioctl_space_info(root
, argp
);
4050 case BTRFS_IOC_SYNC
:
4051 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
4053 case BTRFS_IOC_START_SYNC
:
4054 return btrfs_ioctl_start_sync(root
, argp
);
4055 case BTRFS_IOC_WAIT_SYNC
:
4056 return btrfs_ioctl_wait_sync(root
, argp
);
4057 case BTRFS_IOC_SCRUB
:
4058 return btrfs_ioctl_scrub(file
, argp
);
4059 case BTRFS_IOC_SCRUB_CANCEL
:
4060 return btrfs_ioctl_scrub_cancel(root
, argp
);
4061 case BTRFS_IOC_SCRUB_PROGRESS
:
4062 return btrfs_ioctl_scrub_progress(root
, argp
);
4063 case BTRFS_IOC_BALANCE_V2
:
4064 return btrfs_ioctl_balance(file
, argp
);
4065 case BTRFS_IOC_BALANCE_CTL
:
4066 return btrfs_ioctl_balance_ctl(root
, arg
);
4067 case BTRFS_IOC_BALANCE_PROGRESS
:
4068 return btrfs_ioctl_balance_progress(root
, argp
);
4069 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
4070 return btrfs_ioctl_set_received_subvol(file
, argp
);
4071 case BTRFS_IOC_SEND
:
4072 return btrfs_ioctl_send(file
, argp
);
4073 case BTRFS_IOC_GET_DEV_STATS
:
4074 return btrfs_ioctl_get_dev_stats(root
, argp
);
4075 case BTRFS_IOC_QUOTA_CTL
:
4076 return btrfs_ioctl_quota_ctl(file
, argp
);
4077 case BTRFS_IOC_QGROUP_ASSIGN
:
4078 return btrfs_ioctl_qgroup_assign(file
, argp
);
4079 case BTRFS_IOC_QGROUP_CREATE
:
4080 return btrfs_ioctl_qgroup_create(file
, argp
);
4081 case BTRFS_IOC_QGROUP_LIMIT
:
4082 return btrfs_ioctl_qgroup_limit(file
, argp
);
4083 case BTRFS_IOC_DEV_REPLACE
:
4084 return btrfs_ioctl_dev_replace(root
, argp
);
4085 case BTRFS_IOC_GET_FSLABEL
:
4086 return btrfs_ioctl_get_fslabel(file
, argp
);
4087 case BTRFS_IOC_SET_FSLABEL
:
4088 return btrfs_ioctl_set_fslabel(file
, argp
);