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>
48 #include "transaction.h"
49 #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
;
147 btrfs_update_iflags(inode
);
150 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
152 struct btrfs_inode
*ip
= BTRFS_I(file
->f_path
.dentry
->d_inode
);
153 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
155 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
160 static int check_flags(unsigned int flags
)
162 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
163 FS_NOATIME_FL
| FS_NODUMP_FL
| \
164 FS_SYNC_FL
| FS_DIRSYNC_FL
| \
165 FS_NOCOMP_FL
| FS_COMPR_FL
|
169 if ((flags
& FS_NOCOMP_FL
) && (flags
& FS_COMPR_FL
))
175 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
177 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
178 struct btrfs_inode
*ip
= BTRFS_I(inode
);
179 struct btrfs_root
*root
= ip
->root
;
180 struct btrfs_trans_handle
*trans
;
181 unsigned int flags
, oldflags
;
184 unsigned int i_oldflags
;
187 if (btrfs_root_readonly(root
))
190 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
193 ret
= check_flags(flags
);
197 if (!inode_owner_or_capable(inode
))
200 ret
= mnt_want_write_file(file
);
204 mutex_lock(&inode
->i_mutex
);
206 ip_oldflags
= ip
->flags
;
207 i_oldflags
= inode
->i_flags
;
208 mode
= inode
->i_mode
;
210 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
211 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
212 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
213 if (!capable(CAP_LINUX_IMMUTABLE
)) {
219 if (flags
& FS_SYNC_FL
)
220 ip
->flags
|= BTRFS_INODE_SYNC
;
222 ip
->flags
&= ~BTRFS_INODE_SYNC
;
223 if (flags
& FS_IMMUTABLE_FL
)
224 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
226 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
227 if (flags
& FS_APPEND_FL
)
228 ip
->flags
|= BTRFS_INODE_APPEND
;
230 ip
->flags
&= ~BTRFS_INODE_APPEND
;
231 if (flags
& FS_NODUMP_FL
)
232 ip
->flags
|= BTRFS_INODE_NODUMP
;
234 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
235 if (flags
& FS_NOATIME_FL
)
236 ip
->flags
|= BTRFS_INODE_NOATIME
;
238 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
239 if (flags
& FS_DIRSYNC_FL
)
240 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
242 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
243 if (flags
& FS_NOCOW_FL
) {
246 * It's safe to turn csums off here, no extents exist.
247 * Otherwise we want the flag to reflect the real COW
248 * status of the file and will not set it.
250 if (inode
->i_size
== 0)
251 ip
->flags
|= BTRFS_INODE_NODATACOW
252 | BTRFS_INODE_NODATASUM
;
254 ip
->flags
|= BTRFS_INODE_NODATACOW
;
258 * Revert back under same assuptions as above
261 if (inode
->i_size
== 0)
262 ip
->flags
&= ~(BTRFS_INODE_NODATACOW
263 | BTRFS_INODE_NODATASUM
);
265 ip
->flags
&= ~BTRFS_INODE_NODATACOW
;
270 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
271 * flag may be changed automatically if compression code won't make
274 if (flags
& FS_NOCOMP_FL
) {
275 ip
->flags
&= ~BTRFS_INODE_COMPRESS
;
276 ip
->flags
|= BTRFS_INODE_NOCOMPRESS
;
277 } else if (flags
& FS_COMPR_FL
) {
278 ip
->flags
|= BTRFS_INODE_COMPRESS
;
279 ip
->flags
&= ~BTRFS_INODE_NOCOMPRESS
;
281 ip
->flags
&= ~(BTRFS_INODE_COMPRESS
| BTRFS_INODE_NOCOMPRESS
);
284 trans
= btrfs_start_transaction(root
, 1);
286 ret
= PTR_ERR(trans
);
290 btrfs_update_iflags(inode
);
291 inode_inc_iversion(inode
);
292 inode
->i_ctime
= CURRENT_TIME
;
293 ret
= btrfs_update_inode(trans
, root
, inode
);
295 btrfs_end_transaction(trans
, root
);
298 ip
->flags
= ip_oldflags
;
299 inode
->i_flags
= i_oldflags
;
303 mutex_unlock(&inode
->i_mutex
);
304 mnt_drop_write_file(file
);
308 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
310 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
312 return put_user(inode
->i_generation
, arg
);
315 static noinline
int btrfs_ioctl_fitrim(struct file
*file
, void __user
*arg
)
317 struct btrfs_fs_info
*fs_info
= btrfs_sb(fdentry(file
)->d_sb
);
318 struct btrfs_device
*device
;
319 struct request_queue
*q
;
320 struct fstrim_range range
;
321 u64 minlen
= ULLONG_MAX
;
323 u64 total_bytes
= btrfs_super_total_bytes(fs_info
->super_copy
);
326 if (!capable(CAP_SYS_ADMIN
))
330 list_for_each_entry_rcu(device
, &fs_info
->fs_devices
->devices
,
334 q
= bdev_get_queue(device
->bdev
);
335 if (blk_queue_discard(q
)) {
337 minlen
= min((u64
)q
->limits
.discard_granularity
,
345 if (copy_from_user(&range
, arg
, sizeof(range
)))
347 if (range
.start
> total_bytes
||
348 range
.len
< fs_info
->sb
->s_blocksize
)
351 range
.len
= min(range
.len
, total_bytes
- range
.start
);
352 range
.minlen
= max(range
.minlen
, minlen
);
353 ret
= btrfs_trim_fs(fs_info
->tree_root
, &range
);
357 if (copy_to_user(arg
, &range
, sizeof(range
)))
363 static noinline
int create_subvol(struct btrfs_root
*root
,
364 struct dentry
*dentry
,
365 char *name
, int namelen
,
367 struct btrfs_qgroup_inherit
**inherit
)
369 struct btrfs_trans_handle
*trans
;
370 struct btrfs_key key
;
371 struct btrfs_root_item root_item
;
372 struct btrfs_inode_item
*inode_item
;
373 struct extent_buffer
*leaf
;
374 struct btrfs_root
*new_root
;
375 struct dentry
*parent
= dentry
->d_parent
;
377 struct timespec cur_time
= CURRENT_TIME
;
381 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
385 ret
= btrfs_find_free_objectid(root
->fs_info
->tree_root
, &objectid
);
389 dir
= parent
->d_inode
;
397 trans
= btrfs_start_transaction(root
, 6);
399 return PTR_ERR(trans
);
401 ret
= btrfs_qgroup_inherit(trans
, root
->fs_info
, 0, objectid
,
402 inherit
? *inherit
: NULL
);
406 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
407 0, objectid
, NULL
, 0, 0, 0);
413 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
414 btrfs_set_header_bytenr(leaf
, leaf
->start
);
415 btrfs_set_header_generation(leaf
, trans
->transid
);
416 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
417 btrfs_set_header_owner(leaf
, objectid
);
419 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
420 (unsigned long)btrfs_header_fsid(leaf
),
422 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
423 (unsigned long)btrfs_header_chunk_tree_uuid(leaf
),
425 btrfs_mark_buffer_dirty(leaf
);
427 memset(&root_item
, 0, sizeof(root_item
));
429 inode_item
= &root_item
.inode
;
430 inode_item
->generation
= cpu_to_le64(1);
431 inode_item
->size
= cpu_to_le64(3);
432 inode_item
->nlink
= cpu_to_le32(1);
433 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
434 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
437 root_item
.byte_limit
= 0;
438 inode_item
->flags
= cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT
);
440 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
441 btrfs_set_root_generation(&root_item
, trans
->transid
);
442 btrfs_set_root_level(&root_item
, 0);
443 btrfs_set_root_refs(&root_item
, 1);
444 btrfs_set_root_used(&root_item
, leaf
->len
);
445 btrfs_set_root_last_snapshot(&root_item
, 0);
447 btrfs_set_root_generation_v2(&root_item
,
448 btrfs_root_generation(&root_item
));
449 uuid_le_gen(&new_uuid
);
450 memcpy(root_item
.uuid
, new_uuid
.b
, BTRFS_UUID_SIZE
);
451 root_item
.otime
.sec
= cpu_to_le64(cur_time
.tv_sec
);
452 root_item
.otime
.nsec
= cpu_to_le32(cur_time
.tv_nsec
);
453 root_item
.ctime
= root_item
.otime
;
454 btrfs_set_root_ctransid(&root_item
, trans
->transid
);
455 btrfs_set_root_otransid(&root_item
, trans
->transid
);
457 btrfs_tree_unlock(leaf
);
458 free_extent_buffer(leaf
);
461 btrfs_set_root_dirid(&root_item
, new_dirid
);
463 key
.objectid
= objectid
;
465 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
466 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
471 key
.offset
= (u64
)-1;
472 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
473 if (IS_ERR(new_root
)) {
474 btrfs_abort_transaction(trans
, root
, PTR_ERR(new_root
));
475 ret
= PTR_ERR(new_root
);
479 btrfs_record_root_in_trans(trans
, new_root
);
481 ret
= btrfs_create_subvol_root(trans
, new_root
, new_dirid
);
483 /* We potentially lose an unused inode item here */
484 btrfs_abort_transaction(trans
, root
, ret
);
489 * insert the directory item
491 ret
= btrfs_set_inode_index(dir
, &index
);
493 btrfs_abort_transaction(trans
, root
, ret
);
497 ret
= btrfs_insert_dir_item(trans
, root
,
498 name
, namelen
, dir
, &key
,
499 BTRFS_FT_DIR
, index
);
501 btrfs_abort_transaction(trans
, root
, ret
);
505 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
506 ret
= btrfs_update_inode(trans
, root
, dir
);
509 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
510 objectid
, root
->root_key
.objectid
,
511 btrfs_ino(dir
), index
, name
, namelen
);
515 d_instantiate(dentry
, btrfs_lookup_dentry(dir
, dentry
));
518 *async_transid
= trans
->transid
;
519 err
= btrfs_commit_transaction_async(trans
, root
, 1);
521 err
= btrfs_commit_transaction(trans
, root
);
528 static int create_snapshot(struct btrfs_root
*root
, struct dentry
*dentry
,
529 char *name
, int namelen
, u64
*async_transid
,
530 bool readonly
, struct btrfs_qgroup_inherit
**inherit
)
533 struct btrfs_pending_snapshot
*pending_snapshot
;
534 struct btrfs_trans_handle
*trans
;
540 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
541 if (!pending_snapshot
)
544 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
,
545 BTRFS_BLOCK_RSV_TEMP
);
546 pending_snapshot
->dentry
= dentry
;
547 pending_snapshot
->root
= root
;
548 pending_snapshot
->readonly
= readonly
;
550 pending_snapshot
->inherit
= *inherit
;
551 *inherit
= NULL
; /* take responsibility to free it */
554 trans
= btrfs_start_transaction(root
->fs_info
->extent_root
, 6);
556 ret
= PTR_ERR(trans
);
560 ret
= btrfs_snap_reserve_metadata(trans
, pending_snapshot
);
563 spin_lock(&root
->fs_info
->trans_lock
);
564 list_add(&pending_snapshot
->list
,
565 &trans
->transaction
->pending_snapshots
);
566 spin_unlock(&root
->fs_info
->trans_lock
);
568 *async_transid
= trans
->transid
;
569 ret
= btrfs_commit_transaction_async(trans
,
570 root
->fs_info
->extent_root
, 1);
572 ret
= btrfs_commit_transaction(trans
,
573 root
->fs_info
->extent_root
);
576 /* cleanup_transaction has freed this for us */
578 pending_snapshot
= NULL
;
582 ret
= pending_snapshot
->error
;
586 ret
= btrfs_orphan_cleanup(pending_snapshot
->snap
);
590 inode
= btrfs_lookup_dentry(dentry
->d_parent
->d_inode
, dentry
);
592 ret
= PTR_ERR(inode
);
596 d_instantiate(dentry
, inode
);
599 kfree(pending_snapshot
);
603 /* copy of check_sticky in fs/namei.c()
604 * It's inline, so penalty for filesystems that don't use sticky bit is
607 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
609 kuid_t fsuid
= current_fsuid();
611 if (!(dir
->i_mode
& S_ISVTX
))
613 if (uid_eq(inode
->i_uid
, fsuid
))
615 if (uid_eq(dir
->i_uid
, fsuid
))
617 return !capable(CAP_FOWNER
);
620 /* copy of may_delete in fs/namei.c()
621 * Check whether we can remove a link victim from directory dir, check
622 * whether the type of victim is right.
623 * 1. We can't do it if dir is read-only (done in permission())
624 * 2. We should have write and exec permissions on dir
625 * 3. We can't remove anything from append-only dir
626 * 4. We can't do anything with immutable dir (done in permission())
627 * 5. If the sticky bit on dir is set we should either
628 * a. be owner of dir, or
629 * b. be owner of victim, or
630 * c. have CAP_FOWNER capability
631 * 6. If the victim is append-only or immutable we can't do antyhing with
632 * links pointing to it.
633 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
634 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
635 * 9. We can't remove a root or mountpoint.
636 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
637 * nfs_async_unlink().
640 static int btrfs_may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
644 if (!victim
->d_inode
)
647 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
648 audit_inode_child(dir
, victim
, AUDIT_TYPE_CHILD_DELETE
);
650 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
655 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
656 IS_APPEND(victim
->d_inode
)||
657 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
660 if (!S_ISDIR(victim
->d_inode
->i_mode
))
664 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
668 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
673 /* copy of may_create in fs/namei.c() */
674 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
680 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
684 * Create a new subvolume below @parent. This is largely modeled after
685 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
686 * inside this filesystem so it's quite a bit simpler.
688 static noinline
int btrfs_mksubvol(struct path
*parent
,
689 char *name
, int namelen
,
690 struct btrfs_root
*snap_src
,
691 u64
*async_transid
, bool readonly
,
692 struct btrfs_qgroup_inherit
**inherit
)
694 struct inode
*dir
= parent
->dentry
->d_inode
;
695 struct dentry
*dentry
;
698 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
700 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
701 error
= PTR_ERR(dentry
);
709 error
= btrfs_may_create(dir
, dentry
);
713 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
715 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
719 error
= create_snapshot(snap_src
, dentry
, name
, namelen
,
720 async_transid
, readonly
, inherit
);
722 error
= create_subvol(BTRFS_I(dir
)->root
, dentry
,
723 name
, namelen
, async_transid
, inherit
);
726 fsnotify_mkdir(dir
, dentry
);
728 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
732 mutex_unlock(&dir
->i_mutex
);
737 * When we're defragging a range, we don't want to kick it off again
738 * if it is really just waiting for delalloc to send it down.
739 * If we find a nice big extent or delalloc range for the bytes in the
740 * file you want to defrag, we return 0 to let you know to skip this
743 static int check_defrag_in_cache(struct inode
*inode
, u64 offset
, int thresh
)
745 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
746 struct extent_map
*em
= NULL
;
747 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
750 read_lock(&em_tree
->lock
);
751 em
= lookup_extent_mapping(em_tree
, offset
, PAGE_CACHE_SIZE
);
752 read_unlock(&em_tree
->lock
);
755 end
= extent_map_end(em
);
757 if (end
- offset
> thresh
)
760 /* if we already have a nice delalloc here, just stop */
762 end
= count_range_bits(io_tree
, &offset
, offset
+ thresh
,
763 thresh
, EXTENT_DELALLOC
, 1);
770 * helper function to walk through a file and find extents
771 * newer than a specific transid, and smaller than thresh.
773 * This is used by the defragging code to find new and small
776 static int find_new_extents(struct btrfs_root
*root
,
777 struct inode
*inode
, u64 newer_than
,
778 u64
*off
, int thresh
)
780 struct btrfs_path
*path
;
781 struct btrfs_key min_key
;
782 struct btrfs_key max_key
;
783 struct extent_buffer
*leaf
;
784 struct btrfs_file_extent_item
*extent
;
787 u64 ino
= btrfs_ino(inode
);
789 path
= btrfs_alloc_path();
793 min_key
.objectid
= ino
;
794 min_key
.type
= BTRFS_EXTENT_DATA_KEY
;
795 min_key
.offset
= *off
;
797 max_key
.objectid
= ino
;
798 max_key
.type
= (u8
)-1;
799 max_key
.offset
= (u64
)-1;
801 path
->keep_locks
= 1;
804 ret
= btrfs_search_forward(root
, &min_key
, &max_key
,
805 path
, 0, newer_than
);
808 if (min_key
.objectid
!= ino
)
810 if (min_key
.type
!= BTRFS_EXTENT_DATA_KEY
)
813 leaf
= path
->nodes
[0];
814 extent
= btrfs_item_ptr(leaf
, path
->slots
[0],
815 struct btrfs_file_extent_item
);
817 type
= btrfs_file_extent_type(leaf
, extent
);
818 if (type
== BTRFS_FILE_EXTENT_REG
&&
819 btrfs_file_extent_num_bytes(leaf
, extent
) < thresh
&&
820 check_defrag_in_cache(inode
, min_key
.offset
, thresh
)) {
821 *off
= min_key
.offset
;
822 btrfs_free_path(path
);
826 if (min_key
.offset
== (u64
)-1)
830 btrfs_release_path(path
);
833 btrfs_free_path(path
);
837 static struct extent_map
*defrag_lookup_extent(struct inode
*inode
, u64 start
)
839 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
840 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
841 struct extent_map
*em
;
842 u64 len
= PAGE_CACHE_SIZE
;
845 * hopefully we have this extent in the tree already, try without
846 * the full extent lock
848 read_lock(&em_tree
->lock
);
849 em
= lookup_extent_mapping(em_tree
, start
, len
);
850 read_unlock(&em_tree
->lock
);
853 /* get the big lock and read metadata off disk */
854 lock_extent(io_tree
, start
, start
+ len
- 1);
855 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
856 unlock_extent(io_tree
, start
, start
+ len
- 1);
865 static bool defrag_check_next_extent(struct inode
*inode
, struct extent_map
*em
)
867 struct extent_map
*next
;
870 /* this is the last extent */
871 if (em
->start
+ em
->len
>= i_size_read(inode
))
874 next
= defrag_lookup_extent(inode
, em
->start
+ em
->len
);
875 if (!next
|| next
->block_start
>= EXTENT_MAP_LAST_BYTE
)
878 free_extent_map(next
);
882 static int should_defrag_range(struct inode
*inode
, u64 start
, int thresh
,
883 u64
*last_len
, u64
*skip
, u64
*defrag_end
,
886 struct extent_map
*em
;
888 bool next_mergeable
= true;
891 * make sure that once we start defragging an extent, we keep on
894 if (start
< *defrag_end
)
899 em
= defrag_lookup_extent(inode
, start
);
903 /* this will cover holes, and inline extents */
904 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
) {
909 next_mergeable
= defrag_check_next_extent(inode
, em
);
912 * we hit a real extent, if it is big or the next extent is not a
913 * real extent, don't bother defragging it
915 if (!compress
&& (*last_len
== 0 || *last_len
>= thresh
) &&
916 (em
->len
>= thresh
|| !next_mergeable
))
920 * last_len ends up being a counter of how many bytes we've defragged.
921 * every time we choose not to defrag an extent, we reset *last_len
922 * so that the next tiny extent will force a defrag.
924 * The end result of this is that tiny extents before a single big
925 * extent will force at least part of that big extent to be defragged.
928 *defrag_end
= extent_map_end(em
);
931 *skip
= extent_map_end(em
);
940 * it doesn't do much good to defrag one or two pages
941 * at a time. This pulls in a nice chunk of pages
944 * It also makes sure the delalloc code has enough
945 * dirty data to avoid making new small extents as part
948 * It's a good idea to start RA on this range
949 * before calling this.
951 static int cluster_pages_for_defrag(struct inode
*inode
,
953 unsigned long start_index
,
956 unsigned long file_end
;
957 u64 isize
= i_size_read(inode
);
964 struct btrfs_ordered_extent
*ordered
;
965 struct extent_state
*cached_state
= NULL
;
966 struct extent_io_tree
*tree
;
967 gfp_t mask
= btrfs_alloc_write_mask(inode
->i_mapping
);
969 file_end
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
970 if (!isize
|| start_index
> file_end
)
973 page_cnt
= min_t(u64
, (u64
)num_pages
, (u64
)file_end
- start_index
+ 1);
975 ret
= btrfs_delalloc_reserve_space(inode
,
976 page_cnt
<< PAGE_CACHE_SHIFT
);
980 tree
= &BTRFS_I(inode
)->io_tree
;
982 /* step one, lock all the pages */
983 for (i
= 0; i
< page_cnt
; i
++) {
986 page
= find_or_create_page(inode
->i_mapping
,
987 start_index
+ i
, mask
);
991 page_start
= page_offset(page
);
992 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
994 lock_extent(tree
, page_start
, page_end
);
995 ordered
= btrfs_lookup_ordered_extent(inode
,
997 unlock_extent(tree
, page_start
, page_end
);
1002 btrfs_start_ordered_extent(inode
, ordered
, 1);
1003 btrfs_put_ordered_extent(ordered
);
1006 * we unlocked the page above, so we need check if
1007 * it was released or not.
1009 if (page
->mapping
!= inode
->i_mapping
) {
1011 page_cache_release(page
);
1016 if (!PageUptodate(page
)) {
1017 btrfs_readpage(NULL
, page
);
1019 if (!PageUptodate(page
)) {
1021 page_cache_release(page
);
1027 if (page
->mapping
!= inode
->i_mapping
) {
1029 page_cache_release(page
);
1039 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1043 * so now we have a nice long stream of locked
1044 * and up to date pages, lets wait on them
1046 for (i
= 0; i
< i_done
; i
++)
1047 wait_on_page_writeback(pages
[i
]);
1049 page_start
= page_offset(pages
[0]);
1050 page_end
= page_offset(pages
[i_done
- 1]) + PAGE_CACHE_SIZE
;
1052 lock_extent_bits(&BTRFS_I(inode
)->io_tree
,
1053 page_start
, page_end
- 1, 0, &cached_state
);
1054 clear_extent_bit(&BTRFS_I(inode
)->io_tree
, page_start
,
1055 page_end
- 1, EXTENT_DIRTY
| EXTENT_DELALLOC
|
1056 EXTENT_DO_ACCOUNTING
| EXTENT_DEFRAG
, 0, 0,
1057 &cached_state
, GFP_NOFS
);
1059 if (i_done
!= page_cnt
) {
1060 spin_lock(&BTRFS_I(inode
)->lock
);
1061 BTRFS_I(inode
)->outstanding_extents
++;
1062 spin_unlock(&BTRFS_I(inode
)->lock
);
1063 btrfs_delalloc_release_space(inode
,
1064 (page_cnt
- i_done
) << PAGE_CACHE_SHIFT
);
1068 set_extent_defrag(&BTRFS_I(inode
)->io_tree
, page_start
, page_end
- 1,
1069 &cached_state
, GFP_NOFS
);
1071 unlock_extent_cached(&BTRFS_I(inode
)->io_tree
,
1072 page_start
, page_end
- 1, &cached_state
,
1075 for (i
= 0; i
< i_done
; i
++) {
1076 clear_page_dirty_for_io(pages
[i
]);
1077 ClearPageChecked(pages
[i
]);
1078 set_page_extent_mapped(pages
[i
]);
1079 set_page_dirty(pages
[i
]);
1080 unlock_page(pages
[i
]);
1081 page_cache_release(pages
[i
]);
1085 for (i
= 0; i
< i_done
; i
++) {
1086 unlock_page(pages
[i
]);
1087 page_cache_release(pages
[i
]);
1089 btrfs_delalloc_release_space(inode
, page_cnt
<< PAGE_CACHE_SHIFT
);
1094 int btrfs_defrag_file(struct inode
*inode
, struct file
*file
,
1095 struct btrfs_ioctl_defrag_range_args
*range
,
1096 u64 newer_than
, unsigned long max_to_defrag
)
1098 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1099 struct file_ra_state
*ra
= NULL
;
1100 unsigned long last_index
;
1101 u64 isize
= i_size_read(inode
);
1105 u64 newer_off
= range
->start
;
1107 unsigned long ra_index
= 0;
1109 int defrag_count
= 0;
1110 int compress_type
= BTRFS_COMPRESS_ZLIB
;
1111 int extent_thresh
= range
->extent_thresh
;
1112 int max_cluster
= (256 * 1024) >> PAGE_CACHE_SHIFT
;
1113 int cluster
= max_cluster
;
1114 u64 new_align
= ~((u64
)128 * 1024 - 1);
1115 struct page
**pages
= NULL
;
1117 if (extent_thresh
== 0)
1118 extent_thresh
= 256 * 1024;
1120 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
1121 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
1123 if (range
->compress_type
)
1124 compress_type
= range
->compress_type
;
1131 * if we were not given a file, allocate a readahead
1135 ra
= kzalloc(sizeof(*ra
), GFP_NOFS
);
1138 file_ra_state_init(ra
, inode
->i_mapping
);
1143 pages
= kmalloc(sizeof(struct page
*) * max_cluster
,
1150 /* find the last page to defrag */
1151 if (range
->start
+ range
->len
> range
->start
) {
1152 last_index
= min_t(u64
, isize
- 1,
1153 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
1155 last_index
= (isize
- 1) >> PAGE_CACHE_SHIFT
;
1159 ret
= find_new_extents(root
, inode
, newer_than
,
1160 &newer_off
, 64 * 1024);
1162 range
->start
= newer_off
;
1164 * we always align our defrag to help keep
1165 * the extents in the file evenly spaced
1167 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1171 i
= range
->start
>> PAGE_CACHE_SHIFT
;
1174 max_to_defrag
= last_index
+ 1;
1177 * make writeback starts from i, so the defrag range can be
1178 * written sequentially.
1180 if (i
< inode
->i_mapping
->writeback_index
)
1181 inode
->i_mapping
->writeback_index
= i
;
1183 while (i
<= last_index
&& defrag_count
< max_to_defrag
&&
1184 (i
< (i_size_read(inode
) + PAGE_CACHE_SIZE
- 1) >>
1185 PAGE_CACHE_SHIFT
)) {
1187 * make sure we stop running if someone unmounts
1190 if (!(inode
->i_sb
->s_flags
& MS_ACTIVE
))
1193 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
1194 extent_thresh
, &last_len
, &skip
,
1195 &defrag_end
, range
->flags
&
1196 BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1199 * the should_defrag function tells us how much to skip
1200 * bump our counter by the suggested amount
1202 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
1203 i
= max(i
+ 1, next
);
1208 cluster
= (PAGE_CACHE_ALIGN(defrag_end
) >>
1209 PAGE_CACHE_SHIFT
) - i
;
1210 cluster
= min(cluster
, max_cluster
);
1212 cluster
= max_cluster
;
1215 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
1216 BTRFS_I(inode
)->force_compress
= compress_type
;
1218 if (i
+ cluster
> ra_index
) {
1219 ra_index
= max(i
, ra_index
);
1220 btrfs_force_ra(inode
->i_mapping
, ra
, file
, ra_index
,
1222 ra_index
+= max_cluster
;
1225 mutex_lock(&inode
->i_mutex
);
1226 ret
= cluster_pages_for_defrag(inode
, pages
, i
, cluster
);
1228 mutex_unlock(&inode
->i_mutex
);
1232 defrag_count
+= ret
;
1233 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, ret
);
1234 mutex_unlock(&inode
->i_mutex
);
1237 if (newer_off
== (u64
)-1)
1243 newer_off
= max(newer_off
+ 1,
1244 (u64
)i
<< PAGE_CACHE_SHIFT
);
1246 ret
= find_new_extents(root
, inode
,
1247 newer_than
, &newer_off
,
1250 range
->start
= newer_off
;
1251 i
= (newer_off
& new_align
) >> PAGE_CACHE_SHIFT
;
1258 last_len
+= ret
<< PAGE_CACHE_SHIFT
;
1266 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
))
1267 filemap_flush(inode
->i_mapping
);
1269 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1270 /* the filemap_flush will queue IO into the worker threads, but
1271 * we have to make sure the IO is actually started and that
1272 * ordered extents get created before we return
1274 atomic_inc(&root
->fs_info
->async_submit_draining
);
1275 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
1276 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
1277 wait_event(root
->fs_info
->async_submit_wait
,
1278 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
1279 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
1281 atomic_dec(&root
->fs_info
->async_submit_draining
);
1283 mutex_lock(&inode
->i_mutex
);
1284 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
1285 mutex_unlock(&inode
->i_mutex
);
1288 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
1289 btrfs_set_fs_incompat(root
->fs_info
, COMPRESS_LZO
);
1301 static noinline
int btrfs_ioctl_resize(struct btrfs_root
*root
,
1307 struct btrfs_ioctl_vol_args
*vol_args
;
1308 struct btrfs_trans_handle
*trans
;
1309 struct btrfs_device
*device
= NULL
;
1311 char *devstr
= NULL
;
1315 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
1318 if (!capable(CAP_SYS_ADMIN
))
1321 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
1323 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1324 return -EINPROGRESS
;
1327 mutex_lock(&root
->fs_info
->volume_mutex
);
1328 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1329 if (IS_ERR(vol_args
)) {
1330 ret
= PTR_ERR(vol_args
);
1334 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1336 sizestr
= vol_args
->name
;
1337 devstr
= strchr(sizestr
, ':');
1340 sizestr
= devstr
+ 1;
1342 devstr
= vol_args
->name
;
1343 devid
= simple_strtoull(devstr
, &end
, 10);
1344 printk(KERN_INFO
"btrfs: resizing devid %llu\n",
1345 (unsigned long long)devid
);
1347 device
= btrfs_find_device(root
->fs_info
, devid
, NULL
, NULL
);
1349 printk(KERN_INFO
"btrfs: resizer unable to find device %llu\n",
1350 (unsigned long long)devid
);
1354 if (device
->fs_devices
&& device
->fs_devices
->seeding
) {
1355 printk(KERN_INFO
"btrfs: resizer unable to apply on "
1356 "seeding device %llu\n",
1357 (unsigned long long)devid
);
1362 if (!strcmp(sizestr
, "max"))
1363 new_size
= device
->bdev
->bd_inode
->i_size
;
1365 if (sizestr
[0] == '-') {
1368 } else if (sizestr
[0] == '+') {
1372 new_size
= memparse(sizestr
, NULL
);
1373 if (new_size
== 0) {
1379 if (device
->is_tgtdev_for_dev_replace
) {
1384 old_size
= device
->total_bytes
;
1387 if (new_size
> old_size
) {
1391 new_size
= old_size
- new_size
;
1392 } else if (mod
> 0) {
1393 new_size
= old_size
+ new_size
;
1396 if (new_size
< 256 * 1024 * 1024) {
1400 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
1405 do_div(new_size
, root
->sectorsize
);
1406 new_size
*= root
->sectorsize
;
1408 printk_in_rcu(KERN_INFO
"btrfs: new size for %s is %llu\n",
1409 rcu_str_deref(device
->name
),
1410 (unsigned long long)new_size
);
1412 if (new_size
> old_size
) {
1413 trans
= btrfs_start_transaction(root
, 0);
1414 if (IS_ERR(trans
)) {
1415 ret
= PTR_ERR(trans
);
1418 ret
= btrfs_grow_device(trans
, device
, new_size
);
1419 btrfs_commit_transaction(trans
, root
);
1420 } else if (new_size
< old_size
) {
1421 ret
= btrfs_shrink_device(device
, new_size
);
1422 } /* equal, nothing need to do */
1427 mutex_unlock(&root
->fs_info
->volume_mutex
);
1428 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
1432 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
1433 char *name
, unsigned long fd
, int subvol
,
1434 u64
*transid
, bool readonly
,
1435 struct btrfs_qgroup_inherit
**inherit
)
1440 ret
= mnt_want_write_file(file
);
1444 namelen
= strlen(name
);
1445 if (strchr(name
, '/')) {
1447 goto out_drop_write
;
1450 if (name
[0] == '.' &&
1451 (namelen
== 1 || (name
[1] == '.' && namelen
== 2))) {
1453 goto out_drop_write
;
1457 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1458 NULL
, transid
, readonly
, inherit
);
1460 struct fd src
= fdget(fd
);
1461 struct inode
*src_inode
;
1464 goto out_drop_write
;
1467 src_inode
= src
.file
->f_path
.dentry
->d_inode
;
1468 if (src_inode
->i_sb
!= file
->f_path
.dentry
->d_inode
->i_sb
) {
1469 printk(KERN_INFO
"btrfs: Snapshot src from "
1473 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
1474 BTRFS_I(src_inode
)->root
,
1475 transid
, readonly
, inherit
);
1480 mnt_drop_write_file(file
);
1485 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
1486 void __user
*arg
, int subvol
)
1488 struct btrfs_ioctl_vol_args
*vol_args
;
1491 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1492 if (IS_ERR(vol_args
))
1493 return PTR_ERR(vol_args
);
1494 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1496 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1497 vol_args
->fd
, subvol
,
1504 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
1505 void __user
*arg
, int subvol
)
1507 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1511 bool readonly
= false;
1512 struct btrfs_qgroup_inherit
*inherit
= NULL
;
1514 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1515 if (IS_ERR(vol_args
))
1516 return PTR_ERR(vol_args
);
1517 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1519 if (vol_args
->flags
&
1520 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
|
1521 BTRFS_SUBVOL_QGROUP_INHERIT
)) {
1526 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1528 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1530 if (vol_args
->flags
& BTRFS_SUBVOL_QGROUP_INHERIT
) {
1531 if (vol_args
->size
> PAGE_CACHE_SIZE
) {
1535 inherit
= memdup_user(vol_args
->qgroup_inherit
, vol_args
->size
);
1536 if (IS_ERR(inherit
)) {
1537 ret
= PTR_ERR(inherit
);
1542 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1543 vol_args
->fd
, subvol
, ptr
,
1544 readonly
, &inherit
);
1546 if (ret
== 0 && ptr
&&
1548 offsetof(struct btrfs_ioctl_vol_args_v2
,
1549 transid
), ptr
, sizeof(*ptr
)))
1557 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1560 struct inode
*inode
= fdentry(file
)->d_inode
;
1561 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1565 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
)
1568 down_read(&root
->fs_info
->subvol_sem
);
1569 if (btrfs_root_readonly(root
))
1570 flags
|= BTRFS_SUBVOL_RDONLY
;
1571 up_read(&root
->fs_info
->subvol_sem
);
1573 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1579 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1582 struct inode
*inode
= fdentry(file
)->d_inode
;
1583 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1584 struct btrfs_trans_handle
*trans
;
1589 ret
= mnt_want_write_file(file
);
1593 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
1595 goto out_drop_write
;
1598 if (copy_from_user(&flags
, arg
, sizeof(flags
))) {
1600 goto out_drop_write
;
1603 if (flags
& BTRFS_SUBVOL_CREATE_ASYNC
) {
1605 goto out_drop_write
;
1608 if (flags
& ~BTRFS_SUBVOL_RDONLY
) {
1610 goto out_drop_write
;
1613 if (!inode_owner_or_capable(inode
)) {
1615 goto out_drop_write
;
1618 down_write(&root
->fs_info
->subvol_sem
);
1621 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1624 root_flags
= btrfs_root_flags(&root
->root_item
);
1625 if (flags
& BTRFS_SUBVOL_RDONLY
)
1626 btrfs_set_root_flags(&root
->root_item
,
1627 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1629 btrfs_set_root_flags(&root
->root_item
,
1630 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1632 trans
= btrfs_start_transaction(root
, 1);
1633 if (IS_ERR(trans
)) {
1634 ret
= PTR_ERR(trans
);
1638 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
1639 &root
->root_key
, &root
->root_item
);
1641 btrfs_commit_transaction(trans
, root
);
1644 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1646 up_write(&root
->fs_info
->subvol_sem
);
1648 mnt_drop_write_file(file
);
1654 * helper to check if the subvolume references other subvolumes
1656 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1658 struct btrfs_path
*path
;
1659 struct btrfs_key key
;
1662 path
= btrfs_alloc_path();
1666 key
.objectid
= root
->root_key
.objectid
;
1667 key
.type
= BTRFS_ROOT_REF_KEY
;
1668 key
.offset
= (u64
)-1;
1670 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1677 if (path
->slots
[0] > 0) {
1679 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1680 if (key
.objectid
== root
->root_key
.objectid
&&
1681 key
.type
== BTRFS_ROOT_REF_KEY
)
1685 btrfs_free_path(path
);
1689 static noinline
int key_in_sk(struct btrfs_key
*key
,
1690 struct btrfs_ioctl_search_key
*sk
)
1692 struct btrfs_key test
;
1695 test
.objectid
= sk
->min_objectid
;
1696 test
.type
= sk
->min_type
;
1697 test
.offset
= sk
->min_offset
;
1699 ret
= btrfs_comp_cpu_keys(key
, &test
);
1703 test
.objectid
= sk
->max_objectid
;
1704 test
.type
= sk
->max_type
;
1705 test
.offset
= sk
->max_offset
;
1707 ret
= btrfs_comp_cpu_keys(key
, &test
);
1713 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1714 struct btrfs_path
*path
,
1715 struct btrfs_key
*key
,
1716 struct btrfs_ioctl_search_key
*sk
,
1718 unsigned long *sk_offset
,
1722 struct extent_buffer
*leaf
;
1723 struct btrfs_ioctl_search_header sh
;
1724 unsigned long item_off
;
1725 unsigned long item_len
;
1731 leaf
= path
->nodes
[0];
1732 slot
= path
->slots
[0];
1733 nritems
= btrfs_header_nritems(leaf
);
1735 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1739 found_transid
= btrfs_header_generation(leaf
);
1741 for (i
= slot
; i
< nritems
; i
++) {
1742 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1743 item_len
= btrfs_item_size_nr(leaf
, i
);
1745 if (item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1748 if (sizeof(sh
) + item_len
+ *sk_offset
>
1749 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1754 btrfs_item_key_to_cpu(leaf
, key
, i
);
1755 if (!key_in_sk(key
, sk
))
1758 sh
.objectid
= key
->objectid
;
1759 sh
.offset
= key
->offset
;
1760 sh
.type
= key
->type
;
1762 sh
.transid
= found_transid
;
1764 /* copy search result header */
1765 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1766 *sk_offset
+= sizeof(sh
);
1769 char *p
= buf
+ *sk_offset
;
1771 read_extent_buffer(leaf
, p
,
1772 item_off
, item_len
);
1773 *sk_offset
+= item_len
;
1777 if (*num_found
>= sk
->nr_items
)
1782 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1784 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1787 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
1797 static noinline
int search_ioctl(struct inode
*inode
,
1798 struct btrfs_ioctl_search_args
*args
)
1800 struct btrfs_root
*root
;
1801 struct btrfs_key key
;
1802 struct btrfs_key max_key
;
1803 struct btrfs_path
*path
;
1804 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
1805 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
1808 unsigned long sk_offset
= 0;
1810 path
= btrfs_alloc_path();
1814 if (sk
->tree_id
== 0) {
1815 /* search the root of the inode that was passed */
1816 root
= BTRFS_I(inode
)->root
;
1818 key
.objectid
= sk
->tree_id
;
1819 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1820 key
.offset
= (u64
)-1;
1821 root
= btrfs_read_fs_root_no_name(info
, &key
);
1823 printk(KERN_ERR
"could not find root %llu\n",
1825 btrfs_free_path(path
);
1830 key
.objectid
= sk
->min_objectid
;
1831 key
.type
= sk
->min_type
;
1832 key
.offset
= sk
->min_offset
;
1834 max_key
.objectid
= sk
->max_objectid
;
1835 max_key
.type
= sk
->max_type
;
1836 max_key
.offset
= sk
->max_offset
;
1838 path
->keep_locks
= 1;
1841 ret
= btrfs_search_forward(root
, &key
, &max_key
, path
, 0,
1848 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
1849 &sk_offset
, &num_found
);
1850 btrfs_release_path(path
);
1851 if (ret
|| num_found
>= sk
->nr_items
)
1857 sk
->nr_items
= num_found
;
1858 btrfs_free_path(path
);
1862 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
1865 struct btrfs_ioctl_search_args
*args
;
1866 struct inode
*inode
;
1869 if (!capable(CAP_SYS_ADMIN
))
1872 args
= memdup_user(argp
, sizeof(*args
));
1874 return PTR_ERR(args
);
1876 inode
= fdentry(file
)->d_inode
;
1877 ret
= search_ioctl(inode
, args
);
1878 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1885 * Search INODE_REFs to identify path name of 'dirid' directory
1886 * in a 'tree_id' tree. and sets path name to 'name'.
1888 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1889 u64 tree_id
, u64 dirid
, char *name
)
1891 struct btrfs_root
*root
;
1892 struct btrfs_key key
;
1898 struct btrfs_inode_ref
*iref
;
1899 struct extent_buffer
*l
;
1900 struct btrfs_path
*path
;
1902 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1907 path
= btrfs_alloc_path();
1911 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
1913 key
.objectid
= tree_id
;
1914 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1915 key
.offset
= (u64
)-1;
1916 root
= btrfs_read_fs_root_no_name(info
, &key
);
1918 printk(KERN_ERR
"could not find root %llu\n", tree_id
);
1923 key
.objectid
= dirid
;
1924 key
.type
= BTRFS_INODE_REF_KEY
;
1925 key
.offset
= (u64
)-1;
1928 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1933 slot
= path
->slots
[0];
1934 if (ret
> 0 && slot
> 0)
1936 btrfs_item_key_to_cpu(l
, &key
, slot
);
1938 if (ret
> 0 && (key
.objectid
!= dirid
||
1939 key
.type
!= BTRFS_INODE_REF_KEY
)) {
1944 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
1945 len
= btrfs_inode_ref_name_len(l
, iref
);
1947 total_len
+= len
+ 1;
1952 read_extent_buffer(l
, ptr
,(unsigned long)(iref
+ 1), len
);
1954 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
1957 btrfs_release_path(path
);
1958 key
.objectid
= key
.offset
;
1959 key
.offset
= (u64
)-1;
1960 dirid
= key
.objectid
;
1964 memmove(name
, ptr
, total_len
);
1965 name
[total_len
]='\0';
1968 btrfs_free_path(path
);
1972 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
1975 struct btrfs_ioctl_ino_lookup_args
*args
;
1976 struct inode
*inode
;
1979 if (!capable(CAP_SYS_ADMIN
))
1982 args
= memdup_user(argp
, sizeof(*args
));
1984 return PTR_ERR(args
);
1986 inode
= fdentry(file
)->d_inode
;
1988 if (args
->treeid
== 0)
1989 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
1991 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
1992 args
->treeid
, args
->objectid
,
1995 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
2002 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
2005 struct dentry
*parent
= fdentry(file
);
2006 struct dentry
*dentry
;
2007 struct inode
*dir
= parent
->d_inode
;
2008 struct inode
*inode
;
2009 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
2010 struct btrfs_root
*dest
= NULL
;
2011 struct btrfs_ioctl_vol_args
*vol_args
;
2012 struct btrfs_trans_handle
*trans
;
2017 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2018 if (IS_ERR(vol_args
))
2019 return PTR_ERR(vol_args
);
2021 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2022 namelen
= strlen(vol_args
->name
);
2023 if (strchr(vol_args
->name
, '/') ||
2024 strncmp(vol_args
->name
, "..", namelen
) == 0) {
2029 err
= mnt_want_write_file(file
);
2033 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
2034 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
2035 if (IS_ERR(dentry
)) {
2036 err
= PTR_ERR(dentry
);
2037 goto out_unlock_dir
;
2040 if (!dentry
->d_inode
) {
2045 inode
= dentry
->d_inode
;
2046 dest
= BTRFS_I(inode
)->root
;
2047 if (!capable(CAP_SYS_ADMIN
)){
2049 * Regular user. Only allow this with a special mount
2050 * option, when the user has write+exec access to the
2051 * subvol root, and when rmdir(2) would have been
2054 * Note that this is _not_ check that the subvol is
2055 * empty or doesn't contain data that we wouldn't
2056 * otherwise be able to delete.
2058 * Users who want to delete empty subvols should try
2062 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
2066 * Do not allow deletion if the parent dir is the same
2067 * as the dir to be deleted. That means the ioctl
2068 * must be called on the dentry referencing the root
2069 * of the subvol, not a random directory contained
2076 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
2080 /* check if subvolume may be deleted by a non-root user */
2081 err
= btrfs_may_delete(dir
, dentry
, 1);
2086 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
2091 mutex_lock(&inode
->i_mutex
);
2092 err
= d_invalidate(dentry
);
2096 down_write(&root
->fs_info
->subvol_sem
);
2098 err
= may_destroy_subvol(dest
);
2102 trans
= btrfs_start_transaction(root
, 0);
2103 if (IS_ERR(trans
)) {
2104 err
= PTR_ERR(trans
);
2107 trans
->block_rsv
= &root
->fs_info
->global_block_rsv
;
2109 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
2110 dest
->root_key
.objectid
,
2111 dentry
->d_name
.name
,
2112 dentry
->d_name
.len
);
2115 btrfs_abort_transaction(trans
, root
, ret
);
2119 btrfs_record_root_in_trans(trans
, dest
);
2121 memset(&dest
->root_item
.drop_progress
, 0,
2122 sizeof(dest
->root_item
.drop_progress
));
2123 dest
->root_item
.drop_level
= 0;
2124 btrfs_set_root_refs(&dest
->root_item
, 0);
2126 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
2127 ret
= btrfs_insert_orphan_item(trans
,
2128 root
->fs_info
->tree_root
,
2129 dest
->root_key
.objectid
);
2131 btrfs_abort_transaction(trans
, root
, ret
);
2137 ret
= btrfs_end_transaction(trans
, root
);
2140 inode
->i_flags
|= S_DEAD
;
2142 up_write(&root
->fs_info
->subvol_sem
);
2144 mutex_unlock(&inode
->i_mutex
);
2146 shrink_dcache_sb(root
->fs_info
->sb
);
2147 btrfs_invalidate_inodes(dest
);
2153 mutex_unlock(&dir
->i_mutex
);
2154 mnt_drop_write_file(file
);
2160 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
2162 struct inode
*inode
= fdentry(file
)->d_inode
;
2163 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2164 struct btrfs_ioctl_defrag_range_args
*range
;
2167 if (btrfs_root_readonly(root
))
2170 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2172 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2173 return -EINPROGRESS
;
2175 ret
= mnt_want_write_file(file
);
2177 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
,
2182 switch (inode
->i_mode
& S_IFMT
) {
2184 if (!capable(CAP_SYS_ADMIN
)) {
2188 ret
= btrfs_defrag_root(root
, 0);
2191 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
2194 if (!(file
->f_mode
& FMODE_WRITE
)) {
2199 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
2206 if (copy_from_user(range
, argp
,
2212 /* compression requires us to start the IO */
2213 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
2214 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
2215 range
->extent_thresh
= (u32
)-1;
2218 /* the rest are all set to zero by kzalloc */
2219 range
->len
= (u64
)-1;
2221 ret
= btrfs_defrag_file(fdentry(file
)->d_inode
, file
,
2231 mnt_drop_write_file(file
);
2232 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2236 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
2238 struct btrfs_ioctl_vol_args
*vol_args
;
2241 if (!capable(CAP_SYS_ADMIN
))
2244 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2246 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2247 return -EINPROGRESS
;
2250 mutex_lock(&root
->fs_info
->volume_mutex
);
2251 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2252 if (IS_ERR(vol_args
)) {
2253 ret
= PTR_ERR(vol_args
);
2257 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2258 ret
= btrfs_init_new_device(root
, vol_args
->name
);
2262 mutex_unlock(&root
->fs_info
->volume_mutex
);
2263 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2267 static long btrfs_ioctl_rm_dev(struct btrfs_root
*root
, void __user
*arg
)
2269 struct btrfs_ioctl_vol_args
*vol_args
;
2272 if (!capable(CAP_SYS_ADMIN
))
2275 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
2278 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
2280 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2281 return -EINPROGRESS
;
2284 mutex_lock(&root
->fs_info
->volume_mutex
);
2285 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
2286 if (IS_ERR(vol_args
)) {
2287 ret
= PTR_ERR(vol_args
);
2291 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
2292 ret
= btrfs_rm_device(root
, vol_args
->name
);
2296 mutex_unlock(&root
->fs_info
->volume_mutex
);
2297 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
, 0);
2301 static long btrfs_ioctl_fs_info(struct btrfs_root
*root
, void __user
*arg
)
2303 struct btrfs_ioctl_fs_info_args
*fi_args
;
2304 struct btrfs_device
*device
;
2305 struct btrfs_device
*next
;
2306 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2309 if (!capable(CAP_SYS_ADMIN
))
2312 fi_args
= kzalloc(sizeof(*fi_args
), GFP_KERNEL
);
2316 fi_args
->num_devices
= fs_devices
->num_devices
;
2317 memcpy(&fi_args
->fsid
, root
->fs_info
->fsid
, sizeof(fi_args
->fsid
));
2319 mutex_lock(&fs_devices
->device_list_mutex
);
2320 list_for_each_entry_safe(device
, next
, &fs_devices
->devices
, dev_list
) {
2321 if (device
->devid
> fi_args
->max_id
)
2322 fi_args
->max_id
= device
->devid
;
2324 mutex_unlock(&fs_devices
->device_list_mutex
);
2326 if (copy_to_user(arg
, fi_args
, sizeof(*fi_args
)))
2333 static long btrfs_ioctl_dev_info(struct btrfs_root
*root
, void __user
*arg
)
2335 struct btrfs_ioctl_dev_info_args
*di_args
;
2336 struct btrfs_device
*dev
;
2337 struct btrfs_fs_devices
*fs_devices
= root
->fs_info
->fs_devices
;
2339 char *s_uuid
= NULL
;
2340 char empty_uuid
[BTRFS_UUID_SIZE
] = {0};
2342 if (!capable(CAP_SYS_ADMIN
))
2345 di_args
= memdup_user(arg
, sizeof(*di_args
));
2346 if (IS_ERR(di_args
))
2347 return PTR_ERR(di_args
);
2349 if (memcmp(empty_uuid
, di_args
->uuid
, BTRFS_UUID_SIZE
) != 0)
2350 s_uuid
= di_args
->uuid
;
2352 mutex_lock(&fs_devices
->device_list_mutex
);
2353 dev
= btrfs_find_device(root
->fs_info
, di_args
->devid
, s_uuid
, NULL
);
2354 mutex_unlock(&fs_devices
->device_list_mutex
);
2361 di_args
->devid
= dev
->devid
;
2362 di_args
->bytes_used
= dev
->bytes_used
;
2363 di_args
->total_bytes
= dev
->total_bytes
;
2364 memcpy(di_args
->uuid
, dev
->uuid
, sizeof(di_args
->uuid
));
2366 struct rcu_string
*name
;
2369 name
= rcu_dereference(dev
->name
);
2370 strncpy(di_args
->path
, name
->str
, sizeof(di_args
->path
));
2372 di_args
->path
[sizeof(di_args
->path
) - 1] = 0;
2374 di_args
->path
[0] = '\0';
2378 if (ret
== 0 && copy_to_user(arg
, di_args
, sizeof(*di_args
)))
2385 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
2386 u64 off
, u64 olen
, u64 destoff
)
2388 struct inode
*inode
= fdentry(file
)->d_inode
;
2389 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2392 struct btrfs_trans_handle
*trans
;
2393 struct btrfs_path
*path
;
2394 struct extent_buffer
*leaf
;
2396 struct btrfs_key key
;
2401 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
2405 * - split compressed inline extents. annoying: we need to
2406 * decompress into destination's address_space (the file offset
2407 * may change, so source mapping won't do), then recompress (or
2408 * otherwise reinsert) a subrange.
2409 * - allow ranges within the same file to be cloned (provided
2410 * they don't overlap)?
2413 /* the destination must be opened for writing */
2414 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
2417 if (btrfs_root_readonly(root
))
2420 ret
= mnt_want_write_file(file
);
2424 src_file
= fdget(srcfd
);
2425 if (!src_file
.file
) {
2427 goto out_drop_write
;
2431 if (src_file
.file
->f_path
.mnt
!= file
->f_path
.mnt
)
2434 src
= src_file
.file
->f_dentry
->d_inode
;
2440 /* the src must be open for reading */
2441 if (!(src_file
.file
->f_mode
& FMODE_READ
))
2444 /* don't make the dst file partly checksummed */
2445 if ((BTRFS_I(src
)->flags
& BTRFS_INODE_NODATASUM
) !=
2446 (BTRFS_I(inode
)->flags
& BTRFS_INODE_NODATASUM
))
2450 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
2454 if (src
->i_sb
!= inode
->i_sb
)
2458 buf
= vmalloc(btrfs_level_size(root
, 0));
2462 path
= btrfs_alloc_path();
2470 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
2471 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
2473 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
2474 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
2477 /* determine range to clone */
2479 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
2482 olen
= len
= src
->i_size
- off
;
2483 /* if we extend to eof, continue to block boundary */
2484 if (off
+ len
== src
->i_size
)
2485 len
= ALIGN(src
->i_size
, bs
) - off
;
2487 /* verify the end result is block aligned */
2488 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
2489 !IS_ALIGNED(destoff
, bs
))
2492 if (destoff
> inode
->i_size
) {
2493 ret
= btrfs_cont_expand(inode
, inode
->i_size
, destoff
);
2498 /* truncate page cache pages from target inode range */
2499 truncate_inode_pages_range(&inode
->i_data
, destoff
,
2500 PAGE_CACHE_ALIGN(destoff
+ len
) - 1);
2502 /* do any pending delalloc/csum calc on src, one way or
2503 another, and lock file content */
2505 struct btrfs_ordered_extent
*ordered
;
2506 lock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2507 ordered
= btrfs_lookup_first_ordered_extent(src
, off
+ len
- 1);
2509 !test_range_bit(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1,
2510 EXTENT_DELALLOC
, 0, NULL
))
2512 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2514 btrfs_put_ordered_extent(ordered
);
2515 btrfs_wait_ordered_range(src
, off
, len
);
2519 key
.objectid
= btrfs_ino(src
);
2520 key
.type
= BTRFS_EXTENT_DATA_KEY
;
2525 * note the key will change type as we walk through the
2528 ret
= btrfs_search_slot(NULL
, BTRFS_I(src
)->root
, &key
, path
,
2533 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2534 if (path
->slots
[0] >= nritems
) {
2535 ret
= btrfs_next_leaf(BTRFS_I(src
)->root
, path
);
2540 nritems
= btrfs_header_nritems(path
->nodes
[0]);
2542 leaf
= path
->nodes
[0];
2543 slot
= path
->slots
[0];
2545 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
2546 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
2547 key
.objectid
!= btrfs_ino(src
))
2550 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
2551 struct btrfs_file_extent_item
*extent
;
2554 struct btrfs_key new_key
;
2555 u64 disko
= 0, diskl
= 0;
2556 u64 datao
= 0, datal
= 0;
2560 size
= btrfs_item_size_nr(leaf
, slot
);
2561 read_extent_buffer(leaf
, buf
,
2562 btrfs_item_ptr_offset(leaf
, slot
),
2565 extent
= btrfs_item_ptr(leaf
, slot
,
2566 struct btrfs_file_extent_item
);
2567 comp
= btrfs_file_extent_compression(leaf
, extent
);
2568 type
= btrfs_file_extent_type(leaf
, extent
);
2569 if (type
== BTRFS_FILE_EXTENT_REG
||
2570 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2571 disko
= btrfs_file_extent_disk_bytenr(leaf
,
2573 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
2575 datao
= btrfs_file_extent_offset(leaf
, extent
);
2576 datal
= btrfs_file_extent_num_bytes(leaf
,
2578 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2579 /* take upper bound, may be compressed */
2580 datal
= btrfs_file_extent_ram_bytes(leaf
,
2583 btrfs_release_path(path
);
2585 if (key
.offset
+ datal
<= off
||
2586 key
.offset
>= off
+ len
- 1)
2589 memcpy(&new_key
, &key
, sizeof(new_key
));
2590 new_key
.objectid
= btrfs_ino(inode
);
2591 if (off
<= key
.offset
)
2592 new_key
.offset
= key
.offset
+ destoff
- off
;
2594 new_key
.offset
= destoff
;
2597 * 1 - adjusting old extent (we may have to split it)
2598 * 1 - add new extent
2601 trans
= btrfs_start_transaction(root
, 3);
2602 if (IS_ERR(trans
)) {
2603 ret
= PTR_ERR(trans
);
2607 if (type
== BTRFS_FILE_EXTENT_REG
||
2608 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
2610 * a | --- range to clone ---| b
2611 * | ------------- extent ------------- |
2614 /* substract range b */
2615 if (key
.offset
+ datal
> off
+ len
)
2616 datal
= off
+ len
- key
.offset
;
2618 /* substract range a */
2619 if (off
> key
.offset
) {
2620 datao
+= off
- key
.offset
;
2621 datal
-= off
- key
.offset
;
2624 ret
= btrfs_drop_extents(trans
, root
, inode
,
2626 new_key
.offset
+ datal
,
2629 btrfs_abort_transaction(trans
, root
,
2631 btrfs_end_transaction(trans
, root
);
2635 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2638 btrfs_abort_transaction(trans
, root
,
2640 btrfs_end_transaction(trans
, root
);
2644 leaf
= path
->nodes
[0];
2645 slot
= path
->slots
[0];
2646 write_extent_buffer(leaf
, buf
,
2647 btrfs_item_ptr_offset(leaf
, slot
),
2650 extent
= btrfs_item_ptr(leaf
, slot
,
2651 struct btrfs_file_extent_item
);
2653 /* disko == 0 means it's a hole */
2657 btrfs_set_file_extent_offset(leaf
, extent
,
2659 btrfs_set_file_extent_num_bytes(leaf
, extent
,
2662 inode_add_bytes(inode
, datal
);
2663 ret
= btrfs_inc_extent_ref(trans
, root
,
2665 root
->root_key
.objectid
,
2667 new_key
.offset
- datao
,
2670 btrfs_abort_transaction(trans
,
2673 btrfs_end_transaction(trans
,
2679 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
2682 if (off
> key
.offset
) {
2683 skip
= off
- key
.offset
;
2684 new_key
.offset
+= skip
;
2687 if (key
.offset
+ datal
> off
+ len
)
2688 trim
= key
.offset
+ datal
- (off
+ len
);
2690 if (comp
&& (skip
|| trim
)) {
2692 btrfs_end_transaction(trans
, root
);
2695 size
-= skip
+ trim
;
2696 datal
-= skip
+ trim
;
2698 ret
= btrfs_drop_extents(trans
, root
, inode
,
2700 new_key
.offset
+ datal
,
2703 btrfs_abort_transaction(trans
, root
,
2705 btrfs_end_transaction(trans
, root
);
2709 ret
= btrfs_insert_empty_item(trans
, root
, path
,
2712 btrfs_abort_transaction(trans
, root
,
2714 btrfs_end_transaction(trans
, root
);
2720 btrfs_file_extent_calc_inline_size(0);
2721 memmove(buf
+start
, buf
+start
+skip
,
2725 leaf
= path
->nodes
[0];
2726 slot
= path
->slots
[0];
2727 write_extent_buffer(leaf
, buf
,
2728 btrfs_item_ptr_offset(leaf
, slot
),
2730 inode_add_bytes(inode
, datal
);
2733 btrfs_mark_buffer_dirty(leaf
);
2734 btrfs_release_path(path
);
2736 inode_inc_iversion(inode
);
2737 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2740 * we round up to the block size at eof when
2741 * determining which extents to clone above,
2742 * but shouldn't round up the file size
2744 endoff
= new_key
.offset
+ datal
;
2745 if (endoff
> destoff
+olen
)
2746 endoff
= destoff
+olen
;
2747 if (endoff
> inode
->i_size
)
2748 btrfs_i_size_write(inode
, endoff
);
2750 ret
= btrfs_update_inode(trans
, root
, inode
);
2752 btrfs_abort_transaction(trans
, root
, ret
);
2753 btrfs_end_transaction(trans
, root
);
2756 ret
= btrfs_end_transaction(trans
, root
);
2759 btrfs_release_path(path
);
2764 btrfs_release_path(path
);
2765 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+ len
- 1);
2767 mutex_unlock(&src
->i_mutex
);
2768 mutex_unlock(&inode
->i_mutex
);
2770 btrfs_free_path(path
);
2774 mnt_drop_write_file(file
);
2778 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
2780 struct btrfs_ioctl_clone_range_args args
;
2782 if (copy_from_user(&args
, argp
, sizeof(args
)))
2784 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
2785 args
.src_length
, args
.dest_offset
);
2789 * there are many ways the trans_start and trans_end ioctls can lead
2790 * to deadlocks. They should only be used by applications that
2791 * basically own the machine, and have a very in depth understanding
2792 * of all the possible deadlocks and enospc problems.
2794 static long btrfs_ioctl_trans_start(struct file
*file
)
2796 struct inode
*inode
= fdentry(file
)->d_inode
;
2797 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2798 struct btrfs_trans_handle
*trans
;
2802 if (!capable(CAP_SYS_ADMIN
))
2806 if (file
->private_data
)
2810 if (btrfs_root_readonly(root
))
2813 ret
= mnt_want_write_file(file
);
2817 atomic_inc(&root
->fs_info
->open_ioctl_trans
);
2820 trans
= btrfs_start_ioctl_transaction(root
);
2824 file
->private_data
= trans
;
2828 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
2829 mnt_drop_write_file(file
);
2834 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2836 struct inode
*inode
= fdentry(file
)->d_inode
;
2837 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2838 struct btrfs_root
*new_root
;
2839 struct btrfs_dir_item
*di
;
2840 struct btrfs_trans_handle
*trans
;
2841 struct btrfs_path
*path
;
2842 struct btrfs_key location
;
2843 struct btrfs_disk_key disk_key
;
2847 if (!capable(CAP_SYS_ADMIN
))
2850 if (copy_from_user(&objectid
, argp
, sizeof(objectid
)))
2854 objectid
= root
->root_key
.objectid
;
2856 location
.objectid
= objectid
;
2857 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2858 location
.offset
= (u64
)-1;
2860 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
2861 if (IS_ERR(new_root
))
2862 return PTR_ERR(new_root
);
2864 if (btrfs_root_refs(&new_root
->root_item
) == 0)
2867 path
= btrfs_alloc_path();
2870 path
->leave_spinning
= 1;
2872 trans
= btrfs_start_transaction(root
, 1);
2873 if (IS_ERR(trans
)) {
2874 btrfs_free_path(path
);
2875 return PTR_ERR(trans
);
2878 dir_id
= btrfs_super_root_dir(root
->fs_info
->super_copy
);
2879 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
2880 dir_id
, "default", 7, 1);
2881 if (IS_ERR_OR_NULL(di
)) {
2882 btrfs_free_path(path
);
2883 btrfs_end_transaction(trans
, root
);
2884 printk(KERN_ERR
"Umm, you don't have the default dir item, "
2885 "this isn't going to work\n");
2889 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
2890 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
2891 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2892 btrfs_free_path(path
);
2894 btrfs_set_fs_incompat(root
->fs_info
, DEFAULT_SUBVOL
);
2895 btrfs_end_transaction(trans
, root
);
2900 void btrfs_get_block_group_info(struct list_head
*groups_list
,
2901 struct btrfs_ioctl_space_info
*space
)
2903 struct btrfs_block_group_cache
*block_group
;
2905 space
->total_bytes
= 0;
2906 space
->used_bytes
= 0;
2908 list_for_each_entry(block_group
, groups_list
, list
) {
2909 space
->flags
= block_group
->flags
;
2910 space
->total_bytes
+= block_group
->key
.offset
;
2911 space
->used_bytes
+=
2912 btrfs_block_group_used(&block_group
->item
);
2916 long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
2918 struct btrfs_ioctl_space_args space_args
;
2919 struct btrfs_ioctl_space_info space
;
2920 struct btrfs_ioctl_space_info
*dest
;
2921 struct btrfs_ioctl_space_info
*dest_orig
;
2922 struct btrfs_ioctl_space_info __user
*user_dest
;
2923 struct btrfs_space_info
*info
;
2924 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
2925 BTRFS_BLOCK_GROUP_SYSTEM
,
2926 BTRFS_BLOCK_GROUP_METADATA
,
2927 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
2934 if (copy_from_user(&space_args
,
2935 (struct btrfs_ioctl_space_args __user
*)arg
,
2936 sizeof(space_args
)))
2939 for (i
= 0; i
< num_types
; i
++) {
2940 struct btrfs_space_info
*tmp
;
2944 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
2946 if (tmp
->flags
== types
[i
]) {
2956 down_read(&info
->groups_sem
);
2957 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
2958 if (!list_empty(&info
->block_groups
[c
]))
2961 up_read(&info
->groups_sem
);
2964 /* space_slots == 0 means they are asking for a count */
2965 if (space_args
.space_slots
== 0) {
2966 space_args
.total_spaces
= slot_count
;
2970 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
2972 alloc_size
= sizeof(*dest
) * slot_count
;
2974 /* we generally have at most 6 or so space infos, one for each raid
2975 * level. So, a whole page should be more than enough for everyone
2977 if (alloc_size
> PAGE_CACHE_SIZE
)
2980 space_args
.total_spaces
= 0;
2981 dest
= kmalloc(alloc_size
, GFP_NOFS
);
2986 /* now we have a buffer to copy into */
2987 for (i
= 0; i
< num_types
; i
++) {
2988 struct btrfs_space_info
*tmp
;
2995 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
2997 if (tmp
->flags
== types
[i
]) {
3006 down_read(&info
->groups_sem
);
3007 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
3008 if (!list_empty(&info
->block_groups
[c
])) {
3009 btrfs_get_block_group_info(
3010 &info
->block_groups
[c
], &space
);
3011 memcpy(dest
, &space
, sizeof(space
));
3013 space_args
.total_spaces
++;
3019 up_read(&info
->groups_sem
);
3022 user_dest
= (struct btrfs_ioctl_space_info __user
*)
3023 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
3025 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
3030 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
3037 * there are many ways the trans_start and trans_end ioctls can lead
3038 * to deadlocks. They should only be used by applications that
3039 * basically own the machine, and have a very in depth understanding
3040 * of all the possible deadlocks and enospc problems.
3042 long btrfs_ioctl_trans_end(struct file
*file
)
3044 struct inode
*inode
= fdentry(file
)->d_inode
;
3045 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3046 struct btrfs_trans_handle
*trans
;
3048 trans
= file
->private_data
;
3051 file
->private_data
= NULL
;
3053 btrfs_end_transaction(trans
, root
);
3055 atomic_dec(&root
->fs_info
->open_ioctl_trans
);
3057 mnt_drop_write_file(file
);
3061 static noinline
long btrfs_ioctl_start_sync(struct btrfs_root
*root
,
3064 struct btrfs_trans_handle
*trans
;
3068 trans
= btrfs_attach_transaction(root
);
3069 if (IS_ERR(trans
)) {
3070 if (PTR_ERR(trans
) != -ENOENT
)
3071 return PTR_ERR(trans
);
3073 /* No running transaction, don't bother */
3074 transid
= root
->fs_info
->last_trans_committed
;
3077 transid
= trans
->transid
;
3078 ret
= btrfs_commit_transaction_async(trans
, root
, 0);
3080 btrfs_end_transaction(trans
, root
);
3085 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
3090 static noinline
long btrfs_ioctl_wait_sync(struct btrfs_root
*root
,
3096 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
3099 transid
= 0; /* current trans */
3101 return btrfs_wait_for_commit(root
, transid
);
3104 static long btrfs_ioctl_scrub(struct btrfs_root
*root
, void __user
*arg
)
3107 struct btrfs_ioctl_scrub_args
*sa
;
3109 if (!capable(CAP_SYS_ADMIN
))
3112 sa
= memdup_user(arg
, sizeof(*sa
));
3116 ret
= btrfs_scrub_dev(root
->fs_info
, sa
->devid
, sa
->start
, sa
->end
,
3117 &sa
->progress
, sa
->flags
& BTRFS_SCRUB_READONLY
,
3120 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3127 static long btrfs_ioctl_scrub_cancel(struct btrfs_root
*root
, void __user
*arg
)
3129 if (!capable(CAP_SYS_ADMIN
))
3132 return btrfs_scrub_cancel(root
->fs_info
);
3135 static long btrfs_ioctl_scrub_progress(struct btrfs_root
*root
,
3138 struct btrfs_ioctl_scrub_args
*sa
;
3141 if (!capable(CAP_SYS_ADMIN
))
3144 sa
= memdup_user(arg
, sizeof(*sa
));
3148 ret
= btrfs_scrub_progress(root
, sa
->devid
, &sa
->progress
);
3150 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3157 static long btrfs_ioctl_get_dev_stats(struct btrfs_root
*root
,
3160 struct btrfs_ioctl_get_dev_stats
*sa
;
3163 sa
= memdup_user(arg
, sizeof(*sa
));
3167 if ((sa
->flags
& BTRFS_DEV_STATS_RESET
) && !capable(CAP_SYS_ADMIN
)) {
3172 ret
= btrfs_get_dev_stats(root
, sa
);
3174 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3181 static long btrfs_ioctl_dev_replace(struct btrfs_root
*root
, void __user
*arg
)
3183 struct btrfs_ioctl_dev_replace_args
*p
;
3186 if (!capable(CAP_SYS_ADMIN
))
3189 p
= memdup_user(arg
, sizeof(*p
));
3194 case BTRFS_IOCTL_DEV_REPLACE_CMD_START
:
3196 &root
->fs_info
->mutually_exclusive_operation_running
,
3198 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3201 ret
= btrfs_dev_replace_start(root
, p
);
3203 &root
->fs_info
->mutually_exclusive_operation_running
,
3207 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS
:
3208 btrfs_dev_replace_status(root
->fs_info
, p
);
3211 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL
:
3212 ret
= btrfs_dev_replace_cancel(root
->fs_info
, p
);
3219 if (copy_to_user(arg
, p
, sizeof(*p
)))
3226 static long btrfs_ioctl_ino_to_path(struct btrfs_root
*root
, void __user
*arg
)
3232 struct btrfs_ioctl_ino_path_args
*ipa
= NULL
;
3233 struct inode_fs_paths
*ipath
= NULL
;
3234 struct btrfs_path
*path
;
3236 if (!capable(CAP_SYS_ADMIN
))
3239 path
= btrfs_alloc_path();
3245 ipa
= memdup_user(arg
, sizeof(*ipa
));
3252 size
= min_t(u32
, ipa
->size
, 4096);
3253 ipath
= init_ipath(size
, root
, path
);
3254 if (IS_ERR(ipath
)) {
3255 ret
= PTR_ERR(ipath
);
3260 ret
= paths_from_inode(ipa
->inum
, ipath
);
3264 for (i
= 0; i
< ipath
->fspath
->elem_cnt
; ++i
) {
3265 rel_ptr
= ipath
->fspath
->val
[i
] -
3266 (u64
)(unsigned long)ipath
->fspath
->val
;
3267 ipath
->fspath
->val
[i
] = rel_ptr
;
3270 ret
= copy_to_user((void *)(unsigned long)ipa
->fspath
,
3271 (void *)(unsigned long)ipath
->fspath
, size
);
3278 btrfs_free_path(path
);
3285 static int build_ino_list(u64 inum
, u64 offset
, u64 root
, void *ctx
)
3287 struct btrfs_data_container
*inodes
= ctx
;
3288 const size_t c
= 3 * sizeof(u64
);
3290 if (inodes
->bytes_left
>= c
) {
3291 inodes
->bytes_left
-= c
;
3292 inodes
->val
[inodes
->elem_cnt
] = inum
;
3293 inodes
->val
[inodes
->elem_cnt
+ 1] = offset
;
3294 inodes
->val
[inodes
->elem_cnt
+ 2] = root
;
3295 inodes
->elem_cnt
+= 3;
3297 inodes
->bytes_missing
+= c
- inodes
->bytes_left
;
3298 inodes
->bytes_left
= 0;
3299 inodes
->elem_missed
+= 3;
3305 static long btrfs_ioctl_logical_to_ino(struct btrfs_root
*root
,
3310 struct btrfs_ioctl_logical_ino_args
*loi
;
3311 struct btrfs_data_container
*inodes
= NULL
;
3312 struct btrfs_path
*path
= NULL
;
3314 if (!capable(CAP_SYS_ADMIN
))
3317 loi
= memdup_user(arg
, sizeof(*loi
));
3324 path
= btrfs_alloc_path();
3330 size
= min_t(u32
, loi
->size
, 64 * 1024);
3331 inodes
= init_data_container(size
);
3332 if (IS_ERR(inodes
)) {
3333 ret
= PTR_ERR(inodes
);
3338 ret
= iterate_inodes_from_logical(loi
->logical
, root
->fs_info
, path
,
3339 build_ino_list
, inodes
);
3345 ret
= copy_to_user((void *)(unsigned long)loi
->inodes
,
3346 (void *)(unsigned long)inodes
, size
);
3351 btrfs_free_path(path
);
3358 void update_ioctl_balance_args(struct btrfs_fs_info
*fs_info
, int lock
,
3359 struct btrfs_ioctl_balance_args
*bargs
)
3361 struct btrfs_balance_control
*bctl
= fs_info
->balance_ctl
;
3363 bargs
->flags
= bctl
->flags
;
3365 if (atomic_read(&fs_info
->balance_running
))
3366 bargs
->state
|= BTRFS_BALANCE_STATE_RUNNING
;
3367 if (atomic_read(&fs_info
->balance_pause_req
))
3368 bargs
->state
|= BTRFS_BALANCE_STATE_PAUSE_REQ
;
3369 if (atomic_read(&fs_info
->balance_cancel_req
))
3370 bargs
->state
|= BTRFS_BALANCE_STATE_CANCEL_REQ
;
3372 memcpy(&bargs
->data
, &bctl
->data
, sizeof(bargs
->data
));
3373 memcpy(&bargs
->meta
, &bctl
->meta
, sizeof(bargs
->meta
));
3374 memcpy(&bargs
->sys
, &bctl
->sys
, sizeof(bargs
->sys
));
3377 spin_lock(&fs_info
->balance_lock
);
3378 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3379 spin_unlock(&fs_info
->balance_lock
);
3381 memcpy(&bargs
->stat
, &bctl
->stat
, sizeof(bargs
->stat
));
3385 static long btrfs_ioctl_balance(struct file
*file
, void __user
*arg
)
3387 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3388 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3389 struct btrfs_ioctl_balance_args
*bargs
;
3390 struct btrfs_balance_control
*bctl
;
3392 int need_to_clear_lock
= 0;
3394 if (!capable(CAP_SYS_ADMIN
))
3397 ret
= mnt_want_write_file(file
);
3401 mutex_lock(&fs_info
->volume_mutex
);
3402 mutex_lock(&fs_info
->balance_mutex
);
3405 bargs
= memdup_user(arg
, sizeof(*bargs
));
3406 if (IS_ERR(bargs
)) {
3407 ret
= PTR_ERR(bargs
);
3411 if (bargs
->flags
& BTRFS_BALANCE_RESUME
) {
3412 if (!fs_info
->balance_ctl
) {
3417 bctl
= fs_info
->balance_ctl
;
3418 spin_lock(&fs_info
->balance_lock
);
3419 bctl
->flags
|= BTRFS_BALANCE_RESUME
;
3420 spin_unlock(&fs_info
->balance_lock
);
3428 if (atomic_xchg(&root
->fs_info
->mutually_exclusive_operation_running
,
3430 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3434 need_to_clear_lock
= 1;
3436 bctl
= kzalloc(sizeof(*bctl
), GFP_NOFS
);
3442 bctl
->fs_info
= fs_info
;
3444 memcpy(&bctl
->data
, &bargs
->data
, sizeof(bctl
->data
));
3445 memcpy(&bctl
->meta
, &bargs
->meta
, sizeof(bctl
->meta
));
3446 memcpy(&bctl
->sys
, &bargs
->sys
, sizeof(bctl
->sys
));
3448 bctl
->flags
= bargs
->flags
;
3450 /* balance everything - no filters */
3451 bctl
->flags
|= BTRFS_BALANCE_TYPE_MASK
;
3455 ret
= btrfs_balance(bctl
, bargs
);
3457 * bctl is freed in __cancel_balance or in free_fs_info if
3458 * restriper was paused all the way until unmount
3461 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3468 if (need_to_clear_lock
)
3469 atomic_set(&root
->fs_info
->mutually_exclusive_operation_running
,
3471 mutex_unlock(&fs_info
->balance_mutex
);
3472 mutex_unlock(&fs_info
->volume_mutex
);
3473 mnt_drop_write_file(file
);
3477 static long btrfs_ioctl_balance_ctl(struct btrfs_root
*root
, int cmd
)
3479 if (!capable(CAP_SYS_ADMIN
))
3483 case BTRFS_BALANCE_CTL_PAUSE
:
3484 return btrfs_pause_balance(root
->fs_info
);
3485 case BTRFS_BALANCE_CTL_CANCEL
:
3486 return btrfs_cancel_balance(root
->fs_info
);
3492 static long btrfs_ioctl_balance_progress(struct btrfs_root
*root
,
3495 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
3496 struct btrfs_ioctl_balance_args
*bargs
;
3499 if (!capable(CAP_SYS_ADMIN
))
3502 mutex_lock(&fs_info
->balance_mutex
);
3503 if (!fs_info
->balance_ctl
) {
3508 bargs
= kzalloc(sizeof(*bargs
), GFP_NOFS
);
3514 update_ioctl_balance_args(fs_info
, 1, bargs
);
3516 if (copy_to_user(arg
, bargs
, sizeof(*bargs
)))
3521 mutex_unlock(&fs_info
->balance_mutex
);
3525 static long btrfs_ioctl_quota_ctl(struct btrfs_root
*root
, void __user
*arg
)
3527 struct btrfs_ioctl_quota_ctl_args
*sa
;
3528 struct btrfs_trans_handle
*trans
= NULL
;
3532 if (!capable(CAP_SYS_ADMIN
))
3535 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
3538 sa
= memdup_user(arg
, sizeof(*sa
));
3542 if (sa
->cmd
!= BTRFS_QUOTA_CTL_RESCAN
) {
3543 trans
= btrfs_start_transaction(root
, 2);
3544 if (IS_ERR(trans
)) {
3545 ret
= PTR_ERR(trans
);
3551 case BTRFS_QUOTA_CTL_ENABLE
:
3552 ret
= btrfs_quota_enable(trans
, root
->fs_info
);
3554 case BTRFS_QUOTA_CTL_DISABLE
:
3555 ret
= btrfs_quota_disable(trans
, root
->fs_info
);
3557 case BTRFS_QUOTA_CTL_RESCAN
:
3558 ret
= btrfs_quota_rescan(root
->fs_info
);
3565 if (copy_to_user(arg
, sa
, sizeof(*sa
)))
3569 err
= btrfs_commit_transaction(trans
, root
);
3579 static long btrfs_ioctl_qgroup_assign(struct btrfs_root
*root
, void __user
*arg
)
3581 struct btrfs_ioctl_qgroup_assign_args
*sa
;
3582 struct btrfs_trans_handle
*trans
;
3586 if (!capable(CAP_SYS_ADMIN
))
3589 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
3592 sa
= memdup_user(arg
, sizeof(*sa
));
3596 trans
= btrfs_join_transaction(root
);
3597 if (IS_ERR(trans
)) {
3598 ret
= PTR_ERR(trans
);
3602 /* FIXME: check if the IDs really exist */
3604 ret
= btrfs_add_qgroup_relation(trans
, root
->fs_info
,
3607 ret
= btrfs_del_qgroup_relation(trans
, root
->fs_info
,
3611 err
= btrfs_end_transaction(trans
, root
);
3620 static long btrfs_ioctl_qgroup_create(struct btrfs_root
*root
, void __user
*arg
)
3622 struct btrfs_ioctl_qgroup_create_args
*sa
;
3623 struct btrfs_trans_handle
*trans
;
3627 if (!capable(CAP_SYS_ADMIN
))
3630 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
3633 sa
= memdup_user(arg
, sizeof(*sa
));
3637 trans
= btrfs_join_transaction(root
);
3638 if (IS_ERR(trans
)) {
3639 ret
= PTR_ERR(trans
);
3643 /* FIXME: check if the IDs really exist */
3645 ret
= btrfs_create_qgroup(trans
, root
->fs_info
, sa
->qgroupid
,
3648 ret
= btrfs_remove_qgroup(trans
, root
->fs_info
, sa
->qgroupid
);
3651 err
= btrfs_end_transaction(trans
, root
);
3660 static long btrfs_ioctl_qgroup_limit(struct btrfs_root
*root
, void __user
*arg
)
3662 struct btrfs_ioctl_qgroup_limit_args
*sa
;
3663 struct btrfs_trans_handle
*trans
;
3668 if (!capable(CAP_SYS_ADMIN
))
3671 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
3674 sa
= memdup_user(arg
, sizeof(*sa
));
3678 trans
= btrfs_join_transaction(root
);
3679 if (IS_ERR(trans
)) {
3680 ret
= PTR_ERR(trans
);
3684 qgroupid
= sa
->qgroupid
;
3686 /* take the current subvol as qgroup */
3687 qgroupid
= root
->root_key
.objectid
;
3690 /* FIXME: check if the IDs really exist */
3691 ret
= btrfs_limit_qgroup(trans
, root
->fs_info
, qgroupid
, &sa
->lim
);
3693 err
= btrfs_end_transaction(trans
, root
);
3702 static long btrfs_ioctl_set_received_subvol(struct file
*file
,
3705 struct btrfs_ioctl_received_subvol_args
*sa
= NULL
;
3706 struct inode
*inode
= fdentry(file
)->d_inode
;
3707 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
3708 struct btrfs_root_item
*root_item
= &root
->root_item
;
3709 struct btrfs_trans_handle
*trans
;
3710 struct timespec ct
= CURRENT_TIME
;
3713 ret
= mnt_want_write_file(file
);
3717 down_write(&root
->fs_info
->subvol_sem
);
3719 if (btrfs_ino(inode
) != BTRFS_FIRST_FREE_OBJECTID
) {
3724 if (btrfs_root_readonly(root
)) {
3729 if (!inode_owner_or_capable(inode
)) {
3734 sa
= memdup_user(arg
, sizeof(*sa
));
3741 trans
= btrfs_start_transaction(root
, 1);
3742 if (IS_ERR(trans
)) {
3743 ret
= PTR_ERR(trans
);
3748 sa
->rtransid
= trans
->transid
;
3749 sa
->rtime
.sec
= ct
.tv_sec
;
3750 sa
->rtime
.nsec
= ct
.tv_nsec
;
3752 memcpy(root_item
->received_uuid
, sa
->uuid
, BTRFS_UUID_SIZE
);
3753 btrfs_set_root_stransid(root_item
, sa
->stransid
);
3754 btrfs_set_root_rtransid(root_item
, sa
->rtransid
);
3755 root_item
->stime
.sec
= cpu_to_le64(sa
->stime
.sec
);
3756 root_item
->stime
.nsec
= cpu_to_le32(sa
->stime
.nsec
);
3757 root_item
->rtime
.sec
= cpu_to_le64(sa
->rtime
.sec
);
3758 root_item
->rtime
.nsec
= cpu_to_le32(sa
->rtime
.nsec
);
3760 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
3761 &root
->root_key
, &root
->root_item
);
3763 btrfs_end_transaction(trans
, root
);
3767 ret
= btrfs_commit_transaction(trans
, root
);
3772 ret
= copy_to_user(arg
, sa
, sizeof(*sa
));
3778 up_write(&root
->fs_info
->subvol_sem
);
3779 mnt_drop_write_file(file
);
3783 long btrfs_ioctl(struct file
*file
, unsigned int
3784 cmd
, unsigned long arg
)
3786 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
3787 void __user
*argp
= (void __user
*)arg
;
3790 case FS_IOC_GETFLAGS
:
3791 return btrfs_ioctl_getflags(file
, argp
);
3792 case FS_IOC_SETFLAGS
:
3793 return btrfs_ioctl_setflags(file
, argp
);
3794 case FS_IOC_GETVERSION
:
3795 return btrfs_ioctl_getversion(file
, argp
);
3797 return btrfs_ioctl_fitrim(file
, argp
);
3798 case BTRFS_IOC_SNAP_CREATE
:
3799 return btrfs_ioctl_snap_create(file
, argp
, 0);
3800 case BTRFS_IOC_SNAP_CREATE_V2
:
3801 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
3802 case BTRFS_IOC_SUBVOL_CREATE
:
3803 return btrfs_ioctl_snap_create(file
, argp
, 1);
3804 case BTRFS_IOC_SUBVOL_CREATE_V2
:
3805 return btrfs_ioctl_snap_create_v2(file
, argp
, 1);
3806 case BTRFS_IOC_SNAP_DESTROY
:
3807 return btrfs_ioctl_snap_destroy(file
, argp
);
3808 case BTRFS_IOC_SUBVOL_GETFLAGS
:
3809 return btrfs_ioctl_subvol_getflags(file
, argp
);
3810 case BTRFS_IOC_SUBVOL_SETFLAGS
:
3811 return btrfs_ioctl_subvol_setflags(file
, argp
);
3812 case BTRFS_IOC_DEFAULT_SUBVOL
:
3813 return btrfs_ioctl_default_subvol(file
, argp
);
3814 case BTRFS_IOC_DEFRAG
:
3815 return btrfs_ioctl_defrag(file
, NULL
);
3816 case BTRFS_IOC_DEFRAG_RANGE
:
3817 return btrfs_ioctl_defrag(file
, argp
);
3818 case BTRFS_IOC_RESIZE
:
3819 return btrfs_ioctl_resize(root
, argp
);
3820 case BTRFS_IOC_ADD_DEV
:
3821 return btrfs_ioctl_add_dev(root
, argp
);
3822 case BTRFS_IOC_RM_DEV
:
3823 return btrfs_ioctl_rm_dev(root
, argp
);
3824 case BTRFS_IOC_FS_INFO
:
3825 return btrfs_ioctl_fs_info(root
, argp
);
3826 case BTRFS_IOC_DEV_INFO
:
3827 return btrfs_ioctl_dev_info(root
, argp
);
3828 case BTRFS_IOC_BALANCE
:
3829 return btrfs_ioctl_balance(file
, NULL
);
3830 case BTRFS_IOC_CLONE
:
3831 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
3832 case BTRFS_IOC_CLONE_RANGE
:
3833 return btrfs_ioctl_clone_range(file
, argp
);
3834 case BTRFS_IOC_TRANS_START
:
3835 return btrfs_ioctl_trans_start(file
);
3836 case BTRFS_IOC_TRANS_END
:
3837 return btrfs_ioctl_trans_end(file
);
3838 case BTRFS_IOC_TREE_SEARCH
:
3839 return btrfs_ioctl_tree_search(file
, argp
);
3840 case BTRFS_IOC_INO_LOOKUP
:
3841 return btrfs_ioctl_ino_lookup(file
, argp
);
3842 case BTRFS_IOC_INO_PATHS
:
3843 return btrfs_ioctl_ino_to_path(root
, argp
);
3844 case BTRFS_IOC_LOGICAL_INO
:
3845 return btrfs_ioctl_logical_to_ino(root
, argp
);
3846 case BTRFS_IOC_SPACE_INFO
:
3847 return btrfs_ioctl_space_info(root
, argp
);
3848 case BTRFS_IOC_SYNC
:
3849 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
3851 case BTRFS_IOC_START_SYNC
:
3852 return btrfs_ioctl_start_sync(root
, argp
);
3853 case BTRFS_IOC_WAIT_SYNC
:
3854 return btrfs_ioctl_wait_sync(root
, argp
);
3855 case BTRFS_IOC_SCRUB
:
3856 return btrfs_ioctl_scrub(root
, argp
);
3857 case BTRFS_IOC_SCRUB_CANCEL
:
3858 return btrfs_ioctl_scrub_cancel(root
, argp
);
3859 case BTRFS_IOC_SCRUB_PROGRESS
:
3860 return btrfs_ioctl_scrub_progress(root
, argp
);
3861 case BTRFS_IOC_BALANCE_V2
:
3862 return btrfs_ioctl_balance(file
, argp
);
3863 case BTRFS_IOC_BALANCE_CTL
:
3864 return btrfs_ioctl_balance_ctl(root
, arg
);
3865 case BTRFS_IOC_BALANCE_PROGRESS
:
3866 return btrfs_ioctl_balance_progress(root
, argp
);
3867 case BTRFS_IOC_SET_RECEIVED_SUBVOL
:
3868 return btrfs_ioctl_set_received_subvol(file
, argp
);
3869 case BTRFS_IOC_SEND
:
3870 return btrfs_ioctl_send(file
, argp
);
3871 case BTRFS_IOC_GET_DEV_STATS
:
3872 return btrfs_ioctl_get_dev_stats(root
, argp
);
3873 case BTRFS_IOC_QUOTA_CTL
:
3874 return btrfs_ioctl_quota_ctl(root
, argp
);
3875 case BTRFS_IOC_QGROUP_ASSIGN
:
3876 return btrfs_ioctl_qgroup_assign(root
, argp
);
3877 case BTRFS_IOC_QGROUP_CREATE
:
3878 return btrfs_ioctl_qgroup_create(root
, argp
);
3879 case BTRFS_IOC_QGROUP_LIMIT
:
3880 return btrfs_ioctl_qgroup_limit(root
, argp
);
3881 case BTRFS_IOC_DEV_REPLACE
:
3882 return btrfs_ioctl_dev_replace(root
, argp
);