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>
46 #include "transaction.h"
47 #include "btrfs_inode.h"
49 #include "print-tree.h"
53 /* Mask out flags that are inappropriate for the given type of inode. */
54 static inline __u32
btrfs_mask_flags(umode_t mode
, __u32 flags
)
58 else if (S_ISREG(mode
))
59 return flags
& ~FS_DIRSYNC_FL
;
61 return flags
& (FS_NODUMP_FL
| FS_NOATIME_FL
);
65 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
67 static unsigned int btrfs_flags_to_ioctl(unsigned int flags
)
69 unsigned int iflags
= 0;
71 if (flags
& BTRFS_INODE_SYNC
)
73 if (flags
& BTRFS_INODE_IMMUTABLE
)
74 iflags
|= FS_IMMUTABLE_FL
;
75 if (flags
& BTRFS_INODE_APPEND
)
76 iflags
|= FS_APPEND_FL
;
77 if (flags
& BTRFS_INODE_NODUMP
)
78 iflags
|= FS_NODUMP_FL
;
79 if (flags
& BTRFS_INODE_NOATIME
)
80 iflags
|= FS_NOATIME_FL
;
81 if (flags
& BTRFS_INODE_DIRSYNC
)
82 iflags
|= FS_DIRSYNC_FL
;
88 * Update inode->i_flags based on the btrfs internal flags.
90 void btrfs_update_iflags(struct inode
*inode
)
92 struct btrfs_inode
*ip
= BTRFS_I(inode
);
94 inode
->i_flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
96 if (ip
->flags
& BTRFS_INODE_SYNC
)
97 inode
->i_flags
|= S_SYNC
;
98 if (ip
->flags
& BTRFS_INODE_IMMUTABLE
)
99 inode
->i_flags
|= S_IMMUTABLE
;
100 if (ip
->flags
& BTRFS_INODE_APPEND
)
101 inode
->i_flags
|= S_APPEND
;
102 if (ip
->flags
& BTRFS_INODE_NOATIME
)
103 inode
->i_flags
|= S_NOATIME
;
104 if (ip
->flags
& BTRFS_INODE_DIRSYNC
)
105 inode
->i_flags
|= S_DIRSYNC
;
109 * Inherit flags from the parent inode.
111 * Unlike extN we don't have any flags we don't want to inherit currently.
113 void btrfs_inherit_iflags(struct inode
*inode
, struct inode
*dir
)
120 flags
= BTRFS_I(dir
)->flags
;
122 if (S_ISREG(inode
->i_mode
))
123 flags
&= ~BTRFS_INODE_DIRSYNC
;
124 else if (!S_ISDIR(inode
->i_mode
))
125 flags
&= (BTRFS_INODE_NODUMP
| BTRFS_INODE_NOATIME
);
127 BTRFS_I(inode
)->flags
= flags
;
128 btrfs_update_iflags(inode
);
131 static int btrfs_ioctl_getflags(struct file
*file
, void __user
*arg
)
133 struct btrfs_inode
*ip
= BTRFS_I(file
->f_path
.dentry
->d_inode
);
134 unsigned int flags
= btrfs_flags_to_ioctl(ip
->flags
);
136 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
141 static int btrfs_ioctl_setflags(struct file
*file
, void __user
*arg
)
143 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
144 struct btrfs_inode
*ip
= BTRFS_I(inode
);
145 struct btrfs_root
*root
= ip
->root
;
146 struct btrfs_trans_handle
*trans
;
147 unsigned int flags
, oldflags
;
150 if (btrfs_root_readonly(root
))
153 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
156 if (flags
& ~(FS_IMMUTABLE_FL
| FS_APPEND_FL
| \
157 FS_NOATIME_FL
| FS_NODUMP_FL
| \
158 FS_SYNC_FL
| FS_DIRSYNC_FL
))
161 if (!is_owner_or_cap(inode
))
164 mutex_lock(&inode
->i_mutex
);
166 flags
= btrfs_mask_flags(inode
->i_mode
, flags
);
167 oldflags
= btrfs_flags_to_ioctl(ip
->flags
);
168 if ((flags
^ oldflags
) & (FS_APPEND_FL
| FS_IMMUTABLE_FL
)) {
169 if (!capable(CAP_LINUX_IMMUTABLE
)) {
175 ret
= mnt_want_write(file
->f_path
.mnt
);
179 if (flags
& FS_SYNC_FL
)
180 ip
->flags
|= BTRFS_INODE_SYNC
;
182 ip
->flags
&= ~BTRFS_INODE_SYNC
;
183 if (flags
& FS_IMMUTABLE_FL
)
184 ip
->flags
|= BTRFS_INODE_IMMUTABLE
;
186 ip
->flags
&= ~BTRFS_INODE_IMMUTABLE
;
187 if (flags
& FS_APPEND_FL
)
188 ip
->flags
|= BTRFS_INODE_APPEND
;
190 ip
->flags
&= ~BTRFS_INODE_APPEND
;
191 if (flags
& FS_NODUMP_FL
)
192 ip
->flags
|= BTRFS_INODE_NODUMP
;
194 ip
->flags
&= ~BTRFS_INODE_NODUMP
;
195 if (flags
& FS_NOATIME_FL
)
196 ip
->flags
|= BTRFS_INODE_NOATIME
;
198 ip
->flags
&= ~BTRFS_INODE_NOATIME
;
199 if (flags
& FS_DIRSYNC_FL
)
200 ip
->flags
|= BTRFS_INODE_DIRSYNC
;
202 ip
->flags
&= ~BTRFS_INODE_DIRSYNC
;
205 trans
= btrfs_join_transaction(root
, 1);
206 BUG_ON(IS_ERR(trans
));
208 ret
= btrfs_update_inode(trans
, root
, inode
);
211 btrfs_update_iflags(inode
);
212 inode
->i_ctime
= CURRENT_TIME
;
213 btrfs_end_transaction(trans
, root
);
215 mnt_drop_write(file
->f_path
.mnt
);
217 mutex_unlock(&inode
->i_mutex
);
221 static int btrfs_ioctl_getversion(struct file
*file
, int __user
*arg
)
223 struct inode
*inode
= file
->f_path
.dentry
->d_inode
;
225 return put_user(inode
->i_generation
, arg
);
228 static noinline
int create_subvol(struct btrfs_root
*root
,
229 struct dentry
*dentry
,
230 char *name
, int namelen
,
233 struct btrfs_trans_handle
*trans
;
234 struct btrfs_key key
;
235 struct btrfs_root_item root_item
;
236 struct btrfs_inode_item
*inode_item
;
237 struct extent_buffer
*leaf
;
238 struct btrfs_root
*new_root
;
239 struct dentry
*parent
= dget_parent(dentry
);
244 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
247 ret
= btrfs_find_free_objectid(NULL
, root
->fs_info
->tree_root
,
254 dir
= parent
->d_inode
;
262 trans
= btrfs_start_transaction(root
, 6);
265 return PTR_ERR(trans
);
268 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
269 0, objectid
, NULL
, 0, 0, 0);
275 memset_extent_buffer(leaf
, 0, 0, sizeof(struct btrfs_header
));
276 btrfs_set_header_bytenr(leaf
, leaf
->start
);
277 btrfs_set_header_generation(leaf
, trans
->transid
);
278 btrfs_set_header_backref_rev(leaf
, BTRFS_MIXED_BACKREF_REV
);
279 btrfs_set_header_owner(leaf
, objectid
);
281 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
282 (unsigned long)btrfs_header_fsid(leaf
),
284 write_extent_buffer(leaf
, root
->fs_info
->chunk_tree_uuid
,
285 (unsigned long)btrfs_header_chunk_tree_uuid(leaf
),
287 btrfs_mark_buffer_dirty(leaf
);
289 inode_item
= &root_item
.inode
;
290 memset(inode_item
, 0, sizeof(*inode_item
));
291 inode_item
->generation
= cpu_to_le64(1);
292 inode_item
->size
= cpu_to_le64(3);
293 inode_item
->nlink
= cpu_to_le32(1);
294 inode_item
->nbytes
= cpu_to_le64(root
->leafsize
);
295 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
297 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
298 btrfs_set_root_generation(&root_item
, trans
->transid
);
299 btrfs_set_root_level(&root_item
, 0);
300 btrfs_set_root_refs(&root_item
, 1);
301 btrfs_set_root_used(&root_item
, leaf
->len
);
302 btrfs_set_root_last_snapshot(&root_item
, 0);
304 memset(&root_item
.drop_progress
, 0, sizeof(root_item
.drop_progress
));
305 root_item
.drop_level
= 0;
307 btrfs_tree_unlock(leaf
);
308 free_extent_buffer(leaf
);
311 btrfs_set_root_dirid(&root_item
, new_dirid
);
313 key
.objectid
= objectid
;
315 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
316 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
321 key
.offset
= (u64
)-1;
322 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &key
);
323 BUG_ON(IS_ERR(new_root
));
325 btrfs_record_root_in_trans(trans
, new_root
);
327 ret
= btrfs_create_subvol_root(trans
, new_root
, new_dirid
,
328 BTRFS_I(dir
)->block_group
);
330 * insert the directory item
332 ret
= btrfs_set_inode_index(dir
, &index
);
335 ret
= btrfs_insert_dir_item(trans
, root
,
336 name
, namelen
, dir
->i_ino
, &key
,
337 BTRFS_FT_DIR
, index
);
341 btrfs_i_size_write(dir
, dir
->i_size
+ namelen
* 2);
342 ret
= btrfs_update_inode(trans
, root
, dir
);
345 ret
= btrfs_add_root_ref(trans
, root
->fs_info
->tree_root
,
346 objectid
, root
->root_key
.objectid
,
347 dir
->i_ino
, index
, name
, namelen
);
351 d_instantiate(dentry
, btrfs_lookup_dentry(dir
, dentry
));
355 *async_transid
= trans
->transid
;
356 err
= btrfs_commit_transaction_async(trans
, root
, 1);
358 err
= btrfs_commit_transaction(trans
, root
);
365 static int create_snapshot(struct btrfs_root
*root
, struct dentry
*dentry
,
366 char *name
, int namelen
, u64
*async_transid
,
370 struct dentry
*parent
;
371 struct btrfs_pending_snapshot
*pending_snapshot
;
372 struct btrfs_trans_handle
*trans
;
378 pending_snapshot
= kzalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
379 if (!pending_snapshot
)
382 btrfs_init_block_rsv(&pending_snapshot
->block_rsv
);
383 pending_snapshot
->dentry
= dentry
;
384 pending_snapshot
->root
= root
;
385 pending_snapshot
->readonly
= readonly
;
387 trans
= btrfs_start_transaction(root
->fs_info
->extent_root
, 5);
389 ret
= PTR_ERR(trans
);
393 ret
= btrfs_snap_reserve_metadata(trans
, pending_snapshot
);
396 list_add(&pending_snapshot
->list
,
397 &trans
->transaction
->pending_snapshots
);
399 *async_transid
= trans
->transid
;
400 ret
= btrfs_commit_transaction_async(trans
,
401 root
->fs_info
->extent_root
, 1);
403 ret
= btrfs_commit_transaction(trans
,
404 root
->fs_info
->extent_root
);
408 ret
= pending_snapshot
->error
;
412 btrfs_orphan_cleanup(pending_snapshot
->snap
);
414 parent
= dget_parent(dentry
);
415 inode
= btrfs_lookup_dentry(parent
->d_inode
, dentry
);
418 ret
= PTR_ERR(inode
);
422 d_instantiate(dentry
, inode
);
425 kfree(pending_snapshot
);
429 /* copy of check_sticky in fs/namei.c()
430 * It's inline, so penalty for filesystems that don't use sticky bit is
433 static inline int btrfs_check_sticky(struct inode
*dir
, struct inode
*inode
)
435 uid_t fsuid
= current_fsuid();
437 if (!(dir
->i_mode
& S_ISVTX
))
439 if (inode
->i_uid
== fsuid
)
441 if (dir
->i_uid
== fsuid
)
443 return !capable(CAP_FOWNER
);
446 /* copy of may_delete in fs/namei.c()
447 * Check whether we can remove a link victim from directory dir, check
448 * whether the type of victim is right.
449 * 1. We can't do it if dir is read-only (done in permission())
450 * 2. We should have write and exec permissions on dir
451 * 3. We can't remove anything from append-only dir
452 * 4. We can't do anything with immutable dir (done in permission())
453 * 5. If the sticky bit on dir is set we should either
454 * a. be owner of dir, or
455 * b. be owner of victim, or
456 * c. have CAP_FOWNER capability
457 * 6. If the victim is append-only or immutable we can't do antyhing with
458 * links pointing to it.
459 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
460 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
461 * 9. We can't remove a root or mountpoint.
462 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
463 * nfs_async_unlink().
466 static int btrfs_may_delete(struct inode
*dir
,struct dentry
*victim
,int isdir
)
470 if (!victim
->d_inode
)
473 BUG_ON(victim
->d_parent
->d_inode
!= dir
);
474 audit_inode_child(victim
, dir
);
476 error
= inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
481 if (btrfs_check_sticky(dir
, victim
->d_inode
)||
482 IS_APPEND(victim
->d_inode
)||
483 IS_IMMUTABLE(victim
->d_inode
) || IS_SWAPFILE(victim
->d_inode
))
486 if (!S_ISDIR(victim
->d_inode
->i_mode
))
490 } else if (S_ISDIR(victim
->d_inode
->i_mode
))
494 if (victim
->d_flags
& DCACHE_NFSFS_RENAMED
)
499 /* copy of may_create in fs/namei.c() */
500 static inline int btrfs_may_create(struct inode
*dir
, struct dentry
*child
)
506 return inode_permission(dir
, MAY_WRITE
| MAY_EXEC
);
510 * Create a new subvolume below @parent. This is largely modeled after
511 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
512 * inside this filesystem so it's quite a bit simpler.
514 static noinline
int btrfs_mksubvol(struct path
*parent
,
515 char *name
, int namelen
,
516 struct btrfs_root
*snap_src
,
517 u64
*async_transid
, bool readonly
)
519 struct inode
*dir
= parent
->dentry
->d_inode
;
520 struct dentry
*dentry
;
523 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
525 dentry
= lookup_one_len(name
, parent
->dentry
, namelen
);
526 error
= PTR_ERR(dentry
);
534 error
= mnt_want_write(parent
->mnt
);
538 error
= btrfs_may_create(dir
, dentry
);
542 down_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
544 if (btrfs_root_refs(&BTRFS_I(dir
)->root
->root_item
) == 0)
548 error
= create_snapshot(snap_src
, dentry
,
549 name
, namelen
, async_transid
, readonly
);
551 error
= create_subvol(BTRFS_I(dir
)->root
, dentry
,
552 name
, namelen
, async_transid
);
555 fsnotify_mkdir(dir
, dentry
);
557 up_read(&BTRFS_I(dir
)->root
->fs_info
->subvol_sem
);
559 mnt_drop_write(parent
->mnt
);
563 mutex_unlock(&dir
->i_mutex
);
567 static int should_defrag_range(struct inode
*inode
, u64 start
, u64 len
,
568 int thresh
, u64
*last_len
, u64
*skip
,
571 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
572 struct extent_map
*em
= NULL
;
573 struct extent_map_tree
*em_tree
= &BTRFS_I(inode
)->extent_tree
;
581 * make sure that once we start defragging and extent, we keep on
584 if (start
< *defrag_end
)
590 * hopefully we have this extent in the tree already, try without
591 * the full extent lock
593 read_lock(&em_tree
->lock
);
594 em
= lookup_extent_mapping(em_tree
, start
, len
);
595 read_unlock(&em_tree
->lock
);
598 /* get the big lock and read metadata off disk */
599 lock_extent(io_tree
, start
, start
+ len
- 1, GFP_NOFS
);
600 em
= btrfs_get_extent(inode
, NULL
, 0, start
, len
, 0);
601 unlock_extent(io_tree
, start
, start
+ len
- 1, GFP_NOFS
);
607 /* this will cover holes, and inline extents */
608 if (em
->block_start
>= EXTENT_MAP_LAST_BYTE
)
612 * we hit a real extent, if it is big don't bother defragging it again
614 if ((*last_len
== 0 || *last_len
>= thresh
) && em
->len
>= thresh
)
618 * last_len ends up being a counter of how many bytes we've defragged.
619 * every time we choose not to defrag an extent, we reset *last_len
620 * so that the next tiny extent will force a defrag.
622 * The end result of this is that tiny extents before a single big
623 * extent will force at least part of that big extent to be defragged.
627 *defrag_end
= extent_map_end(em
);
630 *skip
= extent_map_end(em
);
638 static int btrfs_defrag_file(struct file
*file
,
639 struct btrfs_ioctl_defrag_range_args
*range
)
641 struct inode
*inode
= fdentry(file
)->d_inode
;
642 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
643 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
644 struct btrfs_ordered_extent
*ordered
;
646 struct btrfs_super_block
*disk_super
;
647 unsigned long last_index
;
648 unsigned long ra_pages
= root
->fs_info
->bdi
.ra_pages
;
649 unsigned long total_read
= 0;
658 int compress_type
= BTRFS_COMPRESS_ZLIB
;
660 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
) {
661 if (range
->compress_type
> BTRFS_COMPRESS_TYPES
)
663 if (range
->compress_type
)
664 compress_type
= range
->compress_type
;
667 if (inode
->i_size
== 0)
670 if (range
->start
+ range
->len
> range
->start
) {
671 last_index
= min_t(u64
, inode
->i_size
- 1,
672 range
->start
+ range
->len
- 1) >> PAGE_CACHE_SHIFT
;
674 last_index
= (inode
->i_size
- 1) >> PAGE_CACHE_SHIFT
;
677 i
= range
->start
>> PAGE_CACHE_SHIFT
;
678 while (i
<= last_index
) {
679 if (!should_defrag_range(inode
, (u64
)i
<< PAGE_CACHE_SHIFT
,
681 range
->extent_thresh
,
686 * the should_defrag function tells us how much to skip
687 * bump our counter by the suggested amount
689 next
= (skip
+ PAGE_CACHE_SIZE
- 1) >> PAGE_CACHE_SHIFT
;
690 i
= max(i
+ 1, next
);
694 if (total_read
% ra_pages
== 0) {
695 btrfs_force_ra(inode
->i_mapping
, &file
->f_ra
, file
, i
,
696 min(last_index
, i
+ ra_pages
- 1));
699 mutex_lock(&inode
->i_mutex
);
700 if (range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)
701 BTRFS_I(inode
)->force_compress
= compress_type
;
703 ret
= btrfs_delalloc_reserve_space(inode
, PAGE_CACHE_SIZE
);
707 if (inode
->i_size
== 0 ||
708 i
> ((inode
->i_size
- 1) >> PAGE_CACHE_SHIFT
)) {
710 goto err_reservations
;
713 page
= grab_cache_page(inode
->i_mapping
, i
);
716 goto err_reservations
;
719 if (!PageUptodate(page
)) {
720 btrfs_readpage(NULL
, page
);
722 if (!PageUptodate(page
)) {
724 page_cache_release(page
);
726 goto err_reservations
;
730 if (page
->mapping
!= inode
->i_mapping
) {
732 page_cache_release(page
);
736 wait_on_page_writeback(page
);
738 if (PageDirty(page
)) {
739 btrfs_delalloc_release_space(inode
, PAGE_CACHE_SIZE
);
743 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
744 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
745 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
747 ordered
= btrfs_lookup_ordered_extent(inode
, page_start
);
749 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
751 page_cache_release(page
);
752 btrfs_start_ordered_extent(inode
, ordered
, 1);
753 btrfs_put_ordered_extent(ordered
);
756 set_page_extent_mapped(page
);
759 * this makes sure page_mkwrite is called on the
760 * page if it is dirtied again later
762 clear_page_dirty_for_io(page
);
763 clear_extent_bits(&BTRFS_I(inode
)->io_tree
, page_start
,
764 page_end
, EXTENT_DIRTY
| EXTENT_DELALLOC
|
765 EXTENT_DO_ACCOUNTING
, GFP_NOFS
);
767 btrfs_set_extent_delalloc(inode
, page_start
, page_end
, NULL
);
768 ClearPageChecked(page
);
769 set_page_dirty(page
);
770 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
774 page_cache_release(page
);
775 mutex_unlock(&inode
->i_mutex
);
777 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, 1);
781 if ((range
->flags
& BTRFS_DEFRAG_RANGE_START_IO
))
782 filemap_flush(inode
->i_mapping
);
784 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
785 /* the filemap_flush will queue IO into the worker threads, but
786 * we have to make sure the IO is actually started and that
787 * ordered extents get created before we return
789 atomic_inc(&root
->fs_info
->async_submit_draining
);
790 while (atomic_read(&root
->fs_info
->nr_async_submits
) ||
791 atomic_read(&root
->fs_info
->async_delalloc_pages
)) {
792 wait_event(root
->fs_info
->async_submit_wait
,
793 (atomic_read(&root
->fs_info
->nr_async_submits
) == 0 &&
794 atomic_read(&root
->fs_info
->async_delalloc_pages
) == 0));
796 atomic_dec(&root
->fs_info
->async_submit_draining
);
798 mutex_lock(&inode
->i_mutex
);
799 BTRFS_I(inode
)->force_compress
= BTRFS_COMPRESS_NONE
;
800 mutex_unlock(&inode
->i_mutex
);
803 disk_super
= &root
->fs_info
->super_copy
;
804 features
= btrfs_super_incompat_flags(disk_super
);
805 if (range
->compress_type
== BTRFS_COMPRESS_LZO
) {
806 features
|= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO
;
807 btrfs_set_super_incompat_flags(disk_super
, features
);
813 btrfs_delalloc_release_space(inode
, PAGE_CACHE_SIZE
);
815 mutex_unlock(&inode
->i_mutex
);
819 static noinline
int btrfs_ioctl_resize(struct btrfs_root
*root
,
825 struct btrfs_ioctl_vol_args
*vol_args
;
826 struct btrfs_trans_handle
*trans
;
827 struct btrfs_device
*device
= NULL
;
833 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
836 if (!capable(CAP_SYS_ADMIN
))
839 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
840 if (IS_ERR(vol_args
))
841 return PTR_ERR(vol_args
);
843 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
845 mutex_lock(&root
->fs_info
->volume_mutex
);
846 sizestr
= vol_args
->name
;
847 devstr
= strchr(sizestr
, ':');
850 sizestr
= devstr
+ 1;
852 devstr
= vol_args
->name
;
853 devid
= simple_strtoull(devstr
, &end
, 10);
854 printk(KERN_INFO
"resizing devid %llu\n",
855 (unsigned long long)devid
);
857 device
= btrfs_find_device(root
, devid
, NULL
, NULL
);
859 printk(KERN_INFO
"resizer unable to find device %llu\n",
860 (unsigned long long)devid
);
864 if (!strcmp(sizestr
, "max"))
865 new_size
= device
->bdev
->bd_inode
->i_size
;
867 if (sizestr
[0] == '-') {
870 } else if (sizestr
[0] == '+') {
874 new_size
= memparse(sizestr
, NULL
);
881 old_size
= device
->total_bytes
;
884 if (new_size
> old_size
) {
888 new_size
= old_size
- new_size
;
889 } else if (mod
> 0) {
890 new_size
= old_size
+ new_size
;
893 if (new_size
< 256 * 1024 * 1024) {
897 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
902 do_div(new_size
, root
->sectorsize
);
903 new_size
*= root
->sectorsize
;
905 printk(KERN_INFO
"new size for %s is %llu\n",
906 device
->name
, (unsigned long long)new_size
);
908 if (new_size
> old_size
) {
909 trans
= btrfs_start_transaction(root
, 0);
911 ret
= PTR_ERR(trans
);
914 ret
= btrfs_grow_device(trans
, device
, new_size
);
915 btrfs_commit_transaction(trans
, root
);
917 ret
= btrfs_shrink_device(device
, new_size
);
921 mutex_unlock(&root
->fs_info
->volume_mutex
);
926 static noinline
int btrfs_ioctl_snap_create_transid(struct file
*file
,
933 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
934 struct file
*src_file
;
938 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
941 namelen
= strlen(name
);
942 if (strchr(name
, '/')) {
948 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
949 NULL
, transid
, readonly
);
951 struct inode
*src_inode
;
958 src_inode
= src_file
->f_path
.dentry
->d_inode
;
959 if (src_inode
->i_sb
!= file
->f_path
.dentry
->d_inode
->i_sb
) {
960 printk(KERN_INFO
"btrfs: Snapshot src from "
966 ret
= btrfs_mksubvol(&file
->f_path
, name
, namelen
,
967 BTRFS_I(src_inode
)->root
,
975 static noinline
int btrfs_ioctl_snap_create(struct file
*file
,
976 void __user
*arg
, int subvol
)
978 struct btrfs_ioctl_vol_args
*vol_args
;
981 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
982 if (IS_ERR(vol_args
))
983 return PTR_ERR(vol_args
);
984 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
986 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
987 vol_args
->fd
, subvol
,
994 static noinline
int btrfs_ioctl_snap_create_v2(struct file
*file
,
995 void __user
*arg
, int subvol
)
997 struct btrfs_ioctl_vol_args_v2
*vol_args
;
1001 bool readonly
= false;
1003 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1004 if (IS_ERR(vol_args
))
1005 return PTR_ERR(vol_args
);
1006 vol_args
->name
[BTRFS_SUBVOL_NAME_MAX
] = '\0';
1008 if (vol_args
->flags
&
1009 ~(BTRFS_SUBVOL_CREATE_ASYNC
| BTRFS_SUBVOL_RDONLY
)) {
1014 if (vol_args
->flags
& BTRFS_SUBVOL_CREATE_ASYNC
)
1016 if (vol_args
->flags
& BTRFS_SUBVOL_RDONLY
)
1019 ret
= btrfs_ioctl_snap_create_transid(file
, vol_args
->name
,
1020 vol_args
->fd
, subvol
,
1023 if (ret
== 0 && ptr
&&
1025 offsetof(struct btrfs_ioctl_vol_args_v2
,
1026 transid
), ptr
, sizeof(*ptr
)))
1033 static noinline
int btrfs_ioctl_subvol_getflags(struct file
*file
,
1036 struct inode
*inode
= fdentry(file
)->d_inode
;
1037 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1041 if (inode
->i_ino
!= BTRFS_FIRST_FREE_OBJECTID
)
1044 down_read(&root
->fs_info
->subvol_sem
);
1045 if (btrfs_root_readonly(root
))
1046 flags
|= BTRFS_SUBVOL_RDONLY
;
1047 up_read(&root
->fs_info
->subvol_sem
);
1049 if (copy_to_user(arg
, &flags
, sizeof(flags
)))
1055 static noinline
int btrfs_ioctl_subvol_setflags(struct file
*file
,
1058 struct inode
*inode
= fdentry(file
)->d_inode
;
1059 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1060 struct btrfs_trans_handle
*trans
;
1065 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
1068 if (inode
->i_ino
!= BTRFS_FIRST_FREE_OBJECTID
)
1071 if (copy_from_user(&flags
, arg
, sizeof(flags
)))
1074 if (flags
& ~BTRFS_SUBVOL_CREATE_ASYNC
)
1077 if (flags
& ~BTRFS_SUBVOL_RDONLY
)
1080 down_write(&root
->fs_info
->subvol_sem
);
1083 if (!!(flags
& BTRFS_SUBVOL_RDONLY
) == btrfs_root_readonly(root
))
1086 root_flags
= btrfs_root_flags(&root
->root_item
);
1087 if (flags
& BTRFS_SUBVOL_RDONLY
)
1088 btrfs_set_root_flags(&root
->root_item
,
1089 root_flags
| BTRFS_ROOT_SUBVOL_RDONLY
);
1091 btrfs_set_root_flags(&root
->root_item
,
1092 root_flags
& ~BTRFS_ROOT_SUBVOL_RDONLY
);
1094 trans
= btrfs_start_transaction(root
, 1);
1095 if (IS_ERR(trans
)) {
1096 ret
= PTR_ERR(trans
);
1100 ret
= btrfs_update_root(trans
, root
,
1101 &root
->root_key
, &root
->root_item
);
1103 btrfs_commit_transaction(trans
, root
);
1106 btrfs_set_root_flags(&root
->root_item
, root_flags
);
1108 up_write(&root
->fs_info
->subvol_sem
);
1113 * helper to check if the subvolume references other subvolumes
1115 static noinline
int may_destroy_subvol(struct btrfs_root
*root
)
1117 struct btrfs_path
*path
;
1118 struct btrfs_key key
;
1121 path
= btrfs_alloc_path();
1125 key
.objectid
= root
->root_key
.objectid
;
1126 key
.type
= BTRFS_ROOT_REF_KEY
;
1127 key
.offset
= (u64
)-1;
1129 ret
= btrfs_search_slot(NULL
, root
->fs_info
->tree_root
,
1136 if (path
->slots
[0] > 0) {
1138 btrfs_item_key_to_cpu(path
->nodes
[0], &key
, path
->slots
[0]);
1139 if (key
.objectid
== root
->root_key
.objectid
&&
1140 key
.type
== BTRFS_ROOT_REF_KEY
)
1144 btrfs_free_path(path
);
1148 static noinline
int key_in_sk(struct btrfs_key
*key
,
1149 struct btrfs_ioctl_search_key
*sk
)
1151 struct btrfs_key test
;
1154 test
.objectid
= sk
->min_objectid
;
1155 test
.type
= sk
->min_type
;
1156 test
.offset
= sk
->min_offset
;
1158 ret
= btrfs_comp_cpu_keys(key
, &test
);
1162 test
.objectid
= sk
->max_objectid
;
1163 test
.type
= sk
->max_type
;
1164 test
.offset
= sk
->max_offset
;
1166 ret
= btrfs_comp_cpu_keys(key
, &test
);
1172 static noinline
int copy_to_sk(struct btrfs_root
*root
,
1173 struct btrfs_path
*path
,
1174 struct btrfs_key
*key
,
1175 struct btrfs_ioctl_search_key
*sk
,
1177 unsigned long *sk_offset
,
1181 struct extent_buffer
*leaf
;
1182 struct btrfs_ioctl_search_header sh
;
1183 unsigned long item_off
;
1184 unsigned long item_len
;
1191 leaf
= path
->nodes
[0];
1192 slot
= path
->slots
[0];
1193 nritems
= btrfs_header_nritems(leaf
);
1195 if (btrfs_header_generation(leaf
) > sk
->max_transid
) {
1199 found_transid
= btrfs_header_generation(leaf
);
1201 for (i
= slot
; i
< nritems
; i
++) {
1202 item_off
= btrfs_item_ptr_offset(leaf
, i
);
1203 item_len
= btrfs_item_size_nr(leaf
, i
);
1205 if (item_len
> BTRFS_SEARCH_ARGS_BUFSIZE
)
1208 if (sizeof(sh
) + item_len
+ *sk_offset
>
1209 BTRFS_SEARCH_ARGS_BUFSIZE
) {
1214 btrfs_item_key_to_cpu(leaf
, key
, i
);
1215 if (!key_in_sk(key
, sk
))
1218 sh
.objectid
= key
->objectid
;
1219 sh
.offset
= key
->offset
;
1220 sh
.type
= key
->type
;
1222 sh
.transid
= found_transid
;
1224 /* copy search result header */
1225 memcpy(buf
+ *sk_offset
, &sh
, sizeof(sh
));
1226 *sk_offset
+= sizeof(sh
);
1229 char *p
= buf
+ *sk_offset
;
1231 read_extent_buffer(leaf
, p
,
1232 item_off
, item_len
);
1233 *sk_offset
+= item_len
;
1237 if (*num_found
>= sk
->nr_items
)
1242 if (key
->offset
< (u64
)-1 && key
->offset
< sk
->max_offset
)
1244 else if (key
->type
< (u8
)-1 && key
->type
< sk
->max_type
) {
1247 } else if (key
->objectid
< (u64
)-1 && key
->objectid
< sk
->max_objectid
) {
1254 *num_found
+= found
;
1258 static noinline
int search_ioctl(struct inode
*inode
,
1259 struct btrfs_ioctl_search_args
*args
)
1261 struct btrfs_root
*root
;
1262 struct btrfs_key key
;
1263 struct btrfs_key max_key
;
1264 struct btrfs_path
*path
;
1265 struct btrfs_ioctl_search_key
*sk
= &args
->key
;
1266 struct btrfs_fs_info
*info
= BTRFS_I(inode
)->root
->fs_info
;
1269 unsigned long sk_offset
= 0;
1271 path
= btrfs_alloc_path();
1275 if (sk
->tree_id
== 0) {
1276 /* search the root of the inode that was passed */
1277 root
= BTRFS_I(inode
)->root
;
1279 key
.objectid
= sk
->tree_id
;
1280 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1281 key
.offset
= (u64
)-1;
1282 root
= btrfs_read_fs_root_no_name(info
, &key
);
1284 printk(KERN_ERR
"could not find root %llu\n",
1286 btrfs_free_path(path
);
1291 key
.objectid
= sk
->min_objectid
;
1292 key
.type
= sk
->min_type
;
1293 key
.offset
= sk
->min_offset
;
1295 max_key
.objectid
= sk
->max_objectid
;
1296 max_key
.type
= sk
->max_type
;
1297 max_key
.offset
= sk
->max_offset
;
1299 path
->keep_locks
= 1;
1302 ret
= btrfs_search_forward(root
, &key
, &max_key
, path
, 0,
1309 ret
= copy_to_sk(root
, path
, &key
, sk
, args
->buf
,
1310 &sk_offset
, &num_found
);
1311 btrfs_release_path(root
, path
);
1312 if (ret
|| num_found
>= sk
->nr_items
)
1318 sk
->nr_items
= num_found
;
1319 btrfs_free_path(path
);
1323 static noinline
int btrfs_ioctl_tree_search(struct file
*file
,
1326 struct btrfs_ioctl_search_args
*args
;
1327 struct inode
*inode
;
1330 if (!capable(CAP_SYS_ADMIN
))
1333 args
= memdup_user(argp
, sizeof(*args
));
1335 return PTR_ERR(args
);
1337 inode
= fdentry(file
)->d_inode
;
1338 ret
= search_ioctl(inode
, args
);
1339 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1346 * Search INODE_REFs to identify path name of 'dirid' directory
1347 * in a 'tree_id' tree. and sets path name to 'name'.
1349 static noinline
int btrfs_search_path_in_tree(struct btrfs_fs_info
*info
,
1350 u64 tree_id
, u64 dirid
, char *name
)
1352 struct btrfs_root
*root
;
1353 struct btrfs_key key
;
1359 struct btrfs_inode_ref
*iref
;
1360 struct extent_buffer
*l
;
1361 struct btrfs_path
*path
;
1363 if (dirid
== BTRFS_FIRST_FREE_OBJECTID
) {
1368 path
= btrfs_alloc_path();
1372 ptr
= &name
[BTRFS_INO_LOOKUP_PATH_MAX
];
1374 key
.objectid
= tree_id
;
1375 key
.type
= BTRFS_ROOT_ITEM_KEY
;
1376 key
.offset
= (u64
)-1;
1377 root
= btrfs_read_fs_root_no_name(info
, &key
);
1379 printk(KERN_ERR
"could not find root %llu\n", tree_id
);
1384 key
.objectid
= dirid
;
1385 key
.type
= BTRFS_INODE_REF_KEY
;
1386 key
.offset
= (u64
)-1;
1389 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1394 slot
= path
->slots
[0];
1395 if (ret
> 0 && slot
> 0)
1397 btrfs_item_key_to_cpu(l
, &key
, slot
);
1399 if (ret
> 0 && (key
.objectid
!= dirid
||
1400 key
.type
!= BTRFS_INODE_REF_KEY
)) {
1405 iref
= btrfs_item_ptr(l
, slot
, struct btrfs_inode_ref
);
1406 len
= btrfs_inode_ref_name_len(l
, iref
);
1408 total_len
+= len
+ 1;
1413 read_extent_buffer(l
, ptr
,(unsigned long)(iref
+ 1), len
);
1415 if (key
.offset
== BTRFS_FIRST_FREE_OBJECTID
)
1418 btrfs_release_path(root
, path
);
1419 key
.objectid
= key
.offset
;
1420 key
.offset
= (u64
)-1;
1421 dirid
= key
.objectid
;
1426 memcpy(name
, ptr
, total_len
);
1427 name
[total_len
]='\0';
1430 btrfs_free_path(path
);
1434 static noinline
int btrfs_ioctl_ino_lookup(struct file
*file
,
1437 struct btrfs_ioctl_ino_lookup_args
*args
;
1438 struct inode
*inode
;
1441 if (!capable(CAP_SYS_ADMIN
))
1444 args
= memdup_user(argp
, sizeof(*args
));
1446 return PTR_ERR(args
);
1448 inode
= fdentry(file
)->d_inode
;
1450 if (args
->treeid
== 0)
1451 args
->treeid
= BTRFS_I(inode
)->root
->root_key
.objectid
;
1453 ret
= btrfs_search_path_in_tree(BTRFS_I(inode
)->root
->fs_info
,
1454 args
->treeid
, args
->objectid
,
1457 if (ret
== 0 && copy_to_user(argp
, args
, sizeof(*args
)))
1464 static noinline
int btrfs_ioctl_snap_destroy(struct file
*file
,
1467 struct dentry
*parent
= fdentry(file
);
1468 struct dentry
*dentry
;
1469 struct inode
*dir
= parent
->d_inode
;
1470 struct inode
*inode
;
1471 struct btrfs_root
*root
= BTRFS_I(dir
)->root
;
1472 struct btrfs_root
*dest
= NULL
;
1473 struct btrfs_ioctl_vol_args
*vol_args
;
1474 struct btrfs_trans_handle
*trans
;
1479 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1480 if (IS_ERR(vol_args
))
1481 return PTR_ERR(vol_args
);
1483 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1484 namelen
= strlen(vol_args
->name
);
1485 if (strchr(vol_args
->name
, '/') ||
1486 strncmp(vol_args
->name
, "..", namelen
) == 0) {
1491 err
= mnt_want_write(file
->f_path
.mnt
);
1495 mutex_lock_nested(&dir
->i_mutex
, I_MUTEX_PARENT
);
1496 dentry
= lookup_one_len(vol_args
->name
, parent
, namelen
);
1497 if (IS_ERR(dentry
)) {
1498 err
= PTR_ERR(dentry
);
1499 goto out_unlock_dir
;
1502 if (!dentry
->d_inode
) {
1507 inode
= dentry
->d_inode
;
1508 dest
= BTRFS_I(inode
)->root
;
1509 if (!capable(CAP_SYS_ADMIN
)){
1511 * Regular user. Only allow this with a special mount
1512 * option, when the user has write+exec access to the
1513 * subvol root, and when rmdir(2) would have been
1516 * Note that this is _not_ check that the subvol is
1517 * empty or doesn't contain data that we wouldn't
1518 * otherwise be able to delete.
1520 * Users who want to delete empty subvols should try
1524 if (!btrfs_test_opt(root
, USER_SUBVOL_RM_ALLOWED
))
1528 * Do not allow deletion if the parent dir is the same
1529 * as the dir to be deleted. That means the ioctl
1530 * must be called on the dentry referencing the root
1531 * of the subvol, not a random directory contained
1538 err
= inode_permission(inode
, MAY_WRITE
| MAY_EXEC
);
1542 /* check if subvolume may be deleted by a non-root user */
1543 err
= btrfs_may_delete(dir
, dentry
, 1);
1548 if (inode
->i_ino
!= BTRFS_FIRST_FREE_OBJECTID
) {
1553 mutex_lock(&inode
->i_mutex
);
1554 err
= d_invalidate(dentry
);
1558 down_write(&root
->fs_info
->subvol_sem
);
1560 err
= may_destroy_subvol(dest
);
1564 trans
= btrfs_start_transaction(root
, 0);
1565 if (IS_ERR(trans
)) {
1566 err
= PTR_ERR(trans
);
1569 trans
->block_rsv
= &root
->fs_info
->global_block_rsv
;
1571 ret
= btrfs_unlink_subvol(trans
, root
, dir
,
1572 dest
->root_key
.objectid
,
1573 dentry
->d_name
.name
,
1574 dentry
->d_name
.len
);
1577 btrfs_record_root_in_trans(trans
, dest
);
1579 memset(&dest
->root_item
.drop_progress
, 0,
1580 sizeof(dest
->root_item
.drop_progress
));
1581 dest
->root_item
.drop_level
= 0;
1582 btrfs_set_root_refs(&dest
->root_item
, 0);
1584 if (!xchg(&dest
->orphan_item_inserted
, 1)) {
1585 ret
= btrfs_insert_orphan_item(trans
,
1586 root
->fs_info
->tree_root
,
1587 dest
->root_key
.objectid
);
1591 ret
= btrfs_end_transaction(trans
, root
);
1593 inode
->i_flags
|= S_DEAD
;
1595 up_write(&root
->fs_info
->subvol_sem
);
1597 mutex_unlock(&inode
->i_mutex
);
1599 shrink_dcache_sb(root
->fs_info
->sb
);
1600 btrfs_invalidate_inodes(dest
);
1606 mutex_unlock(&dir
->i_mutex
);
1607 mnt_drop_write(file
->f_path
.mnt
);
1613 static int btrfs_ioctl_defrag(struct file
*file
, void __user
*argp
)
1615 struct inode
*inode
= fdentry(file
)->d_inode
;
1616 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1617 struct btrfs_ioctl_defrag_range_args
*range
;
1620 if (btrfs_root_readonly(root
))
1623 ret
= mnt_want_write(file
->f_path
.mnt
);
1627 switch (inode
->i_mode
& S_IFMT
) {
1629 if (!capable(CAP_SYS_ADMIN
)) {
1633 ret
= btrfs_defrag_root(root
, 0);
1636 ret
= btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
1639 if (!(file
->f_mode
& FMODE_WRITE
)) {
1644 range
= kzalloc(sizeof(*range
), GFP_KERNEL
);
1651 if (copy_from_user(range
, argp
,
1657 /* compression requires us to start the IO */
1658 if ((range
->flags
& BTRFS_DEFRAG_RANGE_COMPRESS
)) {
1659 range
->flags
|= BTRFS_DEFRAG_RANGE_START_IO
;
1660 range
->extent_thresh
= (u32
)-1;
1663 /* the rest are all set to zero by kzalloc */
1664 range
->len
= (u64
)-1;
1666 ret
= btrfs_defrag_file(file
, range
);
1673 mnt_drop_write(file
->f_path
.mnt
);
1677 static long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
1679 struct btrfs_ioctl_vol_args
*vol_args
;
1682 if (!capable(CAP_SYS_ADMIN
))
1685 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1686 if (IS_ERR(vol_args
))
1687 return PTR_ERR(vol_args
);
1689 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1690 ret
= btrfs_init_new_device(root
, vol_args
->name
);
1696 static long btrfs_ioctl_rm_dev(struct btrfs_root
*root
, void __user
*arg
)
1698 struct btrfs_ioctl_vol_args
*vol_args
;
1701 if (!capable(CAP_SYS_ADMIN
))
1704 if (root
->fs_info
->sb
->s_flags
& MS_RDONLY
)
1707 vol_args
= memdup_user(arg
, sizeof(*vol_args
));
1708 if (IS_ERR(vol_args
))
1709 return PTR_ERR(vol_args
);
1711 vol_args
->name
[BTRFS_PATH_NAME_MAX
] = '\0';
1712 ret
= btrfs_rm_device(root
, vol_args
->name
);
1718 static noinline
long btrfs_ioctl_clone(struct file
*file
, unsigned long srcfd
,
1719 u64 off
, u64 olen
, u64 destoff
)
1721 struct inode
*inode
= fdentry(file
)->d_inode
;
1722 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
1723 struct file
*src_file
;
1725 struct btrfs_trans_handle
*trans
;
1726 struct btrfs_path
*path
;
1727 struct extent_buffer
*leaf
;
1729 struct btrfs_key key
;
1734 u64 bs
= root
->fs_info
->sb
->s_blocksize
;
1739 * - split compressed inline extents. annoying: we need to
1740 * decompress into destination's address_space (the file offset
1741 * may change, so source mapping won't do), then recompress (or
1742 * otherwise reinsert) a subrange.
1743 * - allow ranges within the same file to be cloned (provided
1744 * they don't overlap)?
1747 /* the destination must be opened for writing */
1748 if (!(file
->f_mode
& FMODE_WRITE
) || (file
->f_flags
& O_APPEND
))
1751 if (btrfs_root_readonly(root
))
1754 ret
= mnt_want_write(file
->f_path
.mnt
);
1758 src_file
= fget(srcfd
);
1761 goto out_drop_write
;
1764 src
= src_file
->f_dentry
->d_inode
;
1770 /* the src must be open for reading */
1771 if (!(src_file
->f_mode
& FMODE_READ
))
1775 if (S_ISDIR(src
->i_mode
) || S_ISDIR(inode
->i_mode
))
1779 if (src
->i_sb
!= inode
->i_sb
|| BTRFS_I(src
)->root
!= root
)
1783 buf
= vmalloc(btrfs_level_size(root
, 0));
1787 path
= btrfs_alloc_path();
1795 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_PARENT
);
1796 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_CHILD
);
1798 mutex_lock_nested(&src
->i_mutex
, I_MUTEX_PARENT
);
1799 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_CHILD
);
1802 /* determine range to clone */
1804 if (off
+ len
> src
->i_size
|| off
+ len
< off
)
1807 olen
= len
= src
->i_size
- off
;
1808 /* if we extend to eof, continue to block boundary */
1809 if (off
+ len
== src
->i_size
)
1810 len
= ALIGN(src
->i_size
, bs
) - off
;
1812 /* verify the end result is block aligned */
1813 if (!IS_ALIGNED(off
, bs
) || !IS_ALIGNED(off
+ len
, bs
) ||
1814 !IS_ALIGNED(destoff
, bs
))
1817 /* do any pending delalloc/csum calc on src, one way or
1818 another, and lock file content */
1820 struct btrfs_ordered_extent
*ordered
;
1821 lock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+len
, GFP_NOFS
);
1822 ordered
= btrfs_lookup_first_ordered_extent(src
, off
+len
);
1824 !test_range_bit(&BTRFS_I(src
)->io_tree
, off
, off
+len
,
1825 EXTENT_DELALLOC
, 0, NULL
))
1827 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+len
, GFP_NOFS
);
1829 btrfs_put_ordered_extent(ordered
);
1830 btrfs_wait_ordered_range(src
, off
, len
);
1834 key
.objectid
= src
->i_ino
;
1835 key
.type
= BTRFS_EXTENT_DATA_KEY
;
1840 * note the key will change type as we walk through the
1843 ret
= btrfs_search_slot(NULL
, root
, &key
, path
, 0, 0);
1847 nritems
= btrfs_header_nritems(path
->nodes
[0]);
1848 if (path
->slots
[0] >= nritems
) {
1849 ret
= btrfs_next_leaf(root
, path
);
1854 nritems
= btrfs_header_nritems(path
->nodes
[0]);
1856 leaf
= path
->nodes
[0];
1857 slot
= path
->slots
[0];
1859 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
1860 if (btrfs_key_type(&key
) > BTRFS_EXTENT_DATA_KEY
||
1861 key
.objectid
!= src
->i_ino
)
1864 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
1865 struct btrfs_file_extent_item
*extent
;
1868 struct btrfs_key new_key
;
1869 u64 disko
= 0, diskl
= 0;
1870 u64 datao
= 0, datal
= 0;
1874 size
= btrfs_item_size_nr(leaf
, slot
);
1875 read_extent_buffer(leaf
, buf
,
1876 btrfs_item_ptr_offset(leaf
, slot
),
1879 extent
= btrfs_item_ptr(leaf
, slot
,
1880 struct btrfs_file_extent_item
);
1881 comp
= btrfs_file_extent_compression(leaf
, extent
);
1882 type
= btrfs_file_extent_type(leaf
, extent
);
1883 if (type
== BTRFS_FILE_EXTENT_REG
||
1884 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
1885 disko
= btrfs_file_extent_disk_bytenr(leaf
,
1887 diskl
= btrfs_file_extent_disk_num_bytes(leaf
,
1889 datao
= btrfs_file_extent_offset(leaf
, extent
);
1890 datal
= btrfs_file_extent_num_bytes(leaf
,
1892 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
1893 /* take upper bound, may be compressed */
1894 datal
= btrfs_file_extent_ram_bytes(leaf
,
1897 btrfs_release_path(root
, path
);
1899 if (key
.offset
+ datal
<= off
||
1900 key
.offset
>= off
+len
)
1903 memcpy(&new_key
, &key
, sizeof(new_key
));
1904 new_key
.objectid
= inode
->i_ino
;
1905 if (off
<= key
.offset
)
1906 new_key
.offset
= key
.offset
+ destoff
- off
;
1908 new_key
.offset
= destoff
;
1910 trans
= btrfs_start_transaction(root
, 1);
1911 if (IS_ERR(trans
)) {
1912 ret
= PTR_ERR(trans
);
1916 if (type
== BTRFS_FILE_EXTENT_REG
||
1917 type
== BTRFS_FILE_EXTENT_PREALLOC
) {
1918 if (off
> key
.offset
) {
1919 datao
+= off
- key
.offset
;
1920 datal
-= off
- key
.offset
;
1923 if (key
.offset
+ datal
> off
+ len
)
1924 datal
= off
+ len
- key
.offset
;
1926 ret
= btrfs_drop_extents(trans
, inode
,
1928 new_key
.offset
+ datal
,
1932 ret
= btrfs_insert_empty_item(trans
, root
, path
,
1936 leaf
= path
->nodes
[0];
1937 slot
= path
->slots
[0];
1938 write_extent_buffer(leaf
, buf
,
1939 btrfs_item_ptr_offset(leaf
, slot
),
1942 extent
= btrfs_item_ptr(leaf
, slot
,
1943 struct btrfs_file_extent_item
);
1945 /* disko == 0 means it's a hole */
1949 btrfs_set_file_extent_offset(leaf
, extent
,
1951 btrfs_set_file_extent_num_bytes(leaf
, extent
,
1954 inode_add_bytes(inode
, datal
);
1955 ret
= btrfs_inc_extent_ref(trans
, root
,
1957 root
->root_key
.objectid
,
1959 new_key
.offset
- datao
);
1962 } else if (type
== BTRFS_FILE_EXTENT_INLINE
) {
1965 if (off
> key
.offset
) {
1966 skip
= off
- key
.offset
;
1967 new_key
.offset
+= skip
;
1970 if (key
.offset
+ datal
> off
+len
)
1971 trim
= key
.offset
+ datal
- (off
+len
);
1973 if (comp
&& (skip
|| trim
)) {
1975 btrfs_end_transaction(trans
, root
);
1978 size
-= skip
+ trim
;
1979 datal
-= skip
+ trim
;
1981 ret
= btrfs_drop_extents(trans
, inode
,
1983 new_key
.offset
+ datal
,
1987 ret
= btrfs_insert_empty_item(trans
, root
, path
,
1993 btrfs_file_extent_calc_inline_size(0);
1994 memmove(buf
+start
, buf
+start
+skip
,
1998 leaf
= path
->nodes
[0];
1999 slot
= path
->slots
[0];
2000 write_extent_buffer(leaf
, buf
,
2001 btrfs_item_ptr_offset(leaf
, slot
),
2003 inode_add_bytes(inode
, datal
);
2006 btrfs_mark_buffer_dirty(leaf
);
2007 btrfs_release_path(root
, path
);
2009 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
2012 * we round up to the block size at eof when
2013 * determining which extents to clone above,
2014 * but shouldn't round up the file size
2016 endoff
= new_key
.offset
+ datal
;
2017 if (endoff
> destoff
+olen
)
2018 endoff
= destoff
+olen
;
2019 if (endoff
> inode
->i_size
)
2020 btrfs_i_size_write(inode
, endoff
);
2022 BTRFS_I(inode
)->flags
= BTRFS_I(src
)->flags
;
2023 ret
= btrfs_update_inode(trans
, root
, inode
);
2025 btrfs_end_transaction(trans
, root
);
2028 btrfs_release_path(root
, path
);
2033 btrfs_release_path(root
, path
);
2034 unlock_extent(&BTRFS_I(src
)->io_tree
, off
, off
+len
, GFP_NOFS
);
2036 mutex_unlock(&src
->i_mutex
);
2037 mutex_unlock(&inode
->i_mutex
);
2039 btrfs_free_path(path
);
2043 mnt_drop_write(file
->f_path
.mnt
);
2047 static long btrfs_ioctl_clone_range(struct file
*file
, void __user
*argp
)
2049 struct btrfs_ioctl_clone_range_args args
;
2051 if (copy_from_user(&args
, argp
, sizeof(args
)))
2053 return btrfs_ioctl_clone(file
, args
.src_fd
, args
.src_offset
,
2054 args
.src_length
, args
.dest_offset
);
2058 * there are many ways the trans_start and trans_end ioctls can lead
2059 * to deadlocks. They should only be used by applications that
2060 * basically own the machine, and have a very in depth understanding
2061 * of all the possible deadlocks and enospc problems.
2063 static long btrfs_ioctl_trans_start(struct file
*file
)
2065 struct inode
*inode
= fdentry(file
)->d_inode
;
2066 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2067 struct btrfs_trans_handle
*trans
;
2071 if (!capable(CAP_SYS_ADMIN
))
2075 if (file
->private_data
)
2079 if (btrfs_root_readonly(root
))
2082 ret
= mnt_want_write(file
->f_path
.mnt
);
2086 mutex_lock(&root
->fs_info
->trans_mutex
);
2087 root
->fs_info
->open_ioctl_trans
++;
2088 mutex_unlock(&root
->fs_info
->trans_mutex
);
2091 trans
= btrfs_start_ioctl_transaction(root
, 0);
2095 file
->private_data
= trans
;
2099 mutex_lock(&root
->fs_info
->trans_mutex
);
2100 root
->fs_info
->open_ioctl_trans
--;
2101 mutex_unlock(&root
->fs_info
->trans_mutex
);
2102 mnt_drop_write(file
->f_path
.mnt
);
2107 static long btrfs_ioctl_default_subvol(struct file
*file
, void __user
*argp
)
2109 struct inode
*inode
= fdentry(file
)->d_inode
;
2110 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2111 struct btrfs_root
*new_root
;
2112 struct btrfs_dir_item
*di
;
2113 struct btrfs_trans_handle
*trans
;
2114 struct btrfs_path
*path
;
2115 struct btrfs_key location
;
2116 struct btrfs_disk_key disk_key
;
2117 struct btrfs_super_block
*disk_super
;
2122 if (!capable(CAP_SYS_ADMIN
))
2125 if (copy_from_user(&objectid
, argp
, sizeof(objectid
)))
2129 objectid
= root
->root_key
.objectid
;
2131 location
.objectid
= objectid
;
2132 location
.type
= BTRFS_ROOT_ITEM_KEY
;
2133 location
.offset
= (u64
)-1;
2135 new_root
= btrfs_read_fs_root_no_name(root
->fs_info
, &location
);
2136 if (IS_ERR(new_root
))
2137 return PTR_ERR(new_root
);
2139 if (btrfs_root_refs(&new_root
->root_item
) == 0)
2142 path
= btrfs_alloc_path();
2145 path
->leave_spinning
= 1;
2147 trans
= btrfs_start_transaction(root
, 1);
2148 if (IS_ERR(trans
)) {
2149 btrfs_free_path(path
);
2150 return PTR_ERR(trans
);
2153 dir_id
= btrfs_super_root_dir(&root
->fs_info
->super_copy
);
2154 di
= btrfs_lookup_dir_item(trans
, root
->fs_info
->tree_root
, path
,
2155 dir_id
, "default", 7, 1);
2156 if (IS_ERR_OR_NULL(di
)) {
2157 btrfs_free_path(path
);
2158 btrfs_end_transaction(trans
, root
);
2159 printk(KERN_ERR
"Umm, you don't have the default dir item, "
2160 "this isn't going to work\n");
2164 btrfs_cpu_key_to_disk(&disk_key
, &new_root
->root_key
);
2165 btrfs_set_dir_item_key(path
->nodes
[0], di
, &disk_key
);
2166 btrfs_mark_buffer_dirty(path
->nodes
[0]);
2167 btrfs_free_path(path
);
2169 disk_super
= &root
->fs_info
->super_copy
;
2170 features
= btrfs_super_incompat_flags(disk_super
);
2171 if (!(features
& BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL
)) {
2172 features
|= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL
;
2173 btrfs_set_super_incompat_flags(disk_super
, features
);
2175 btrfs_end_transaction(trans
, root
);
2180 static void get_block_group_info(struct list_head
*groups_list
,
2181 struct btrfs_ioctl_space_info
*space
)
2183 struct btrfs_block_group_cache
*block_group
;
2185 space
->total_bytes
= 0;
2186 space
->used_bytes
= 0;
2188 list_for_each_entry(block_group
, groups_list
, list
) {
2189 space
->flags
= block_group
->flags
;
2190 space
->total_bytes
+= block_group
->key
.offset
;
2191 space
->used_bytes
+=
2192 btrfs_block_group_used(&block_group
->item
);
2196 long btrfs_ioctl_space_info(struct btrfs_root
*root
, void __user
*arg
)
2198 struct btrfs_ioctl_space_args space_args
;
2199 struct btrfs_ioctl_space_info space
;
2200 struct btrfs_ioctl_space_info
*dest
;
2201 struct btrfs_ioctl_space_info
*dest_orig
;
2202 struct btrfs_ioctl_space_info
*user_dest
;
2203 struct btrfs_space_info
*info
;
2204 u64 types
[] = {BTRFS_BLOCK_GROUP_DATA
,
2205 BTRFS_BLOCK_GROUP_SYSTEM
,
2206 BTRFS_BLOCK_GROUP_METADATA
,
2207 BTRFS_BLOCK_GROUP_DATA
| BTRFS_BLOCK_GROUP_METADATA
};
2214 if (copy_from_user(&space_args
,
2215 (struct btrfs_ioctl_space_args __user
*)arg
,
2216 sizeof(space_args
)))
2219 for (i
= 0; i
< num_types
; i
++) {
2220 struct btrfs_space_info
*tmp
;
2224 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
2226 if (tmp
->flags
== types
[i
]) {
2236 down_read(&info
->groups_sem
);
2237 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
2238 if (!list_empty(&info
->block_groups
[c
]))
2241 up_read(&info
->groups_sem
);
2244 /* space_slots == 0 means they are asking for a count */
2245 if (space_args
.space_slots
== 0) {
2246 space_args
.total_spaces
= slot_count
;
2250 slot_count
= min_t(u64
, space_args
.space_slots
, slot_count
);
2252 alloc_size
= sizeof(*dest
) * slot_count
;
2254 /* we generally have at most 6 or so space infos, one for each raid
2255 * level. So, a whole page should be more than enough for everyone
2257 if (alloc_size
> PAGE_CACHE_SIZE
)
2260 space_args
.total_spaces
= 0;
2261 dest
= kmalloc(alloc_size
, GFP_NOFS
);
2266 /* now we have a buffer to copy into */
2267 for (i
= 0; i
< num_types
; i
++) {
2268 struct btrfs_space_info
*tmp
;
2275 list_for_each_entry_rcu(tmp
, &root
->fs_info
->space_info
,
2277 if (tmp
->flags
== types
[i
]) {
2286 down_read(&info
->groups_sem
);
2287 for (c
= 0; c
< BTRFS_NR_RAID_TYPES
; c
++) {
2288 if (!list_empty(&info
->block_groups
[c
])) {
2289 get_block_group_info(&info
->block_groups
[c
],
2291 memcpy(dest
, &space
, sizeof(space
));
2293 space_args
.total_spaces
++;
2299 up_read(&info
->groups_sem
);
2302 user_dest
= (struct btrfs_ioctl_space_info
*)
2303 (arg
+ sizeof(struct btrfs_ioctl_space_args
));
2305 if (copy_to_user(user_dest
, dest_orig
, alloc_size
))
2310 if (ret
== 0 && copy_to_user(arg
, &space_args
, sizeof(space_args
)))
2317 * there are many ways the trans_start and trans_end ioctls can lead
2318 * to deadlocks. They should only be used by applications that
2319 * basically own the machine, and have a very in depth understanding
2320 * of all the possible deadlocks and enospc problems.
2322 long btrfs_ioctl_trans_end(struct file
*file
)
2324 struct inode
*inode
= fdentry(file
)->d_inode
;
2325 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
2326 struct btrfs_trans_handle
*trans
;
2328 trans
= file
->private_data
;
2331 file
->private_data
= NULL
;
2333 btrfs_end_transaction(trans
, root
);
2335 mutex_lock(&root
->fs_info
->trans_mutex
);
2336 root
->fs_info
->open_ioctl_trans
--;
2337 mutex_unlock(&root
->fs_info
->trans_mutex
);
2339 mnt_drop_write(file
->f_path
.mnt
);
2343 static noinline
long btrfs_ioctl_start_sync(struct file
*file
, void __user
*argp
)
2345 struct btrfs_root
*root
= BTRFS_I(file
->f_dentry
->d_inode
)->root
;
2346 struct btrfs_trans_handle
*trans
;
2349 trans
= btrfs_start_transaction(root
, 0);
2351 return PTR_ERR(trans
);
2352 transid
= trans
->transid
;
2353 btrfs_commit_transaction_async(trans
, root
, 0);
2356 if (copy_to_user(argp
, &transid
, sizeof(transid
)))
2361 static noinline
long btrfs_ioctl_wait_sync(struct file
*file
, void __user
*argp
)
2363 struct btrfs_root
*root
= BTRFS_I(file
->f_dentry
->d_inode
)->root
;
2367 if (copy_from_user(&transid
, argp
, sizeof(transid
)))
2370 transid
= 0; /* current trans */
2372 return btrfs_wait_for_commit(root
, transid
);
2375 long btrfs_ioctl(struct file
*file
, unsigned int
2376 cmd
, unsigned long arg
)
2378 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
2379 void __user
*argp
= (void __user
*)arg
;
2382 case FS_IOC_GETFLAGS
:
2383 return btrfs_ioctl_getflags(file
, argp
);
2384 case FS_IOC_SETFLAGS
:
2385 return btrfs_ioctl_setflags(file
, argp
);
2386 case FS_IOC_GETVERSION
:
2387 return btrfs_ioctl_getversion(file
, argp
);
2388 case BTRFS_IOC_SNAP_CREATE
:
2389 return btrfs_ioctl_snap_create(file
, argp
, 0);
2390 case BTRFS_IOC_SNAP_CREATE_V2
:
2391 return btrfs_ioctl_snap_create_v2(file
, argp
, 0);
2392 case BTRFS_IOC_SUBVOL_CREATE
:
2393 return btrfs_ioctl_snap_create(file
, argp
, 1);
2394 case BTRFS_IOC_SNAP_DESTROY
:
2395 return btrfs_ioctl_snap_destroy(file
, argp
);
2396 case BTRFS_IOC_SUBVOL_GETFLAGS
:
2397 return btrfs_ioctl_subvol_getflags(file
, argp
);
2398 case BTRFS_IOC_SUBVOL_SETFLAGS
:
2399 return btrfs_ioctl_subvol_setflags(file
, argp
);
2400 case BTRFS_IOC_DEFAULT_SUBVOL
:
2401 return btrfs_ioctl_default_subvol(file
, argp
);
2402 case BTRFS_IOC_DEFRAG
:
2403 return btrfs_ioctl_defrag(file
, NULL
);
2404 case BTRFS_IOC_DEFRAG_RANGE
:
2405 return btrfs_ioctl_defrag(file
, argp
);
2406 case BTRFS_IOC_RESIZE
:
2407 return btrfs_ioctl_resize(root
, argp
);
2408 case BTRFS_IOC_ADD_DEV
:
2409 return btrfs_ioctl_add_dev(root
, argp
);
2410 case BTRFS_IOC_RM_DEV
:
2411 return btrfs_ioctl_rm_dev(root
, argp
);
2412 case BTRFS_IOC_BALANCE
:
2413 return btrfs_balance(root
->fs_info
->dev_root
);
2414 case BTRFS_IOC_CLONE
:
2415 return btrfs_ioctl_clone(file
, arg
, 0, 0, 0);
2416 case BTRFS_IOC_CLONE_RANGE
:
2417 return btrfs_ioctl_clone_range(file
, argp
);
2418 case BTRFS_IOC_TRANS_START
:
2419 return btrfs_ioctl_trans_start(file
);
2420 case BTRFS_IOC_TRANS_END
:
2421 return btrfs_ioctl_trans_end(file
);
2422 case BTRFS_IOC_TREE_SEARCH
:
2423 return btrfs_ioctl_tree_search(file
, argp
);
2424 case BTRFS_IOC_INO_LOOKUP
:
2425 return btrfs_ioctl_ino_lookup(file
, argp
);
2426 case BTRFS_IOC_SPACE_INFO
:
2427 return btrfs_ioctl_space_info(root
, argp
);
2428 case BTRFS_IOC_SYNC
:
2429 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);
2431 case BTRFS_IOC_START_SYNC
:
2432 return btrfs_ioctl_start_sync(file
, argp
);
2433 case BTRFS_IOC_WAIT_SYNC
:
2434 return btrfs_ioctl_wait_sync(file
, argp
);