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/pagemap.h>
25 #include <linux/highmem.h>
26 #include <linux/time.h>
27 #include <linux/init.h>
28 #include <linux/string.h>
29 #include <linux/smp_lock.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mpage.h>
32 #include <linux/swap.h>
33 #include <linux/writeback.h>
34 #include <linux/statfs.h>
35 #include <linux/compat.h>
36 #include <linux/bit_spinlock.h>
37 #include <linux/version.h>
38 #include <linux/xattr.h>
41 #include "transaction.h"
42 #include "btrfs_inode.h"
44 #include "print-tree.h"
50 static noinline
int create_subvol(struct btrfs_root
*root
, char *name
,
53 struct btrfs_trans_handle
*trans
;
55 struct btrfs_root_item root_item
;
56 struct btrfs_inode_item
*inode_item
;
57 struct extent_buffer
*leaf
;
58 struct btrfs_root
*new_root
= root
;
63 u64 new_dirid
= BTRFS_FIRST_FREE_OBJECTID
;
66 mutex_lock(&root
->fs_info
->fs_mutex
);
67 ret
= btrfs_check_free_space(root
, 1, 0);
71 trans
= btrfs_start_transaction(root
, 1);
74 ret
= btrfs_find_free_objectid(trans
, root
->fs_info
->tree_root
,
79 leaf
= btrfs_alloc_free_block(trans
, root
, root
->leafsize
,
80 objectid
, trans
->transid
, 0, 0,
85 btrfs_set_header_nritems(leaf
, 0);
86 btrfs_set_header_level(leaf
, 0);
87 btrfs_set_header_bytenr(leaf
, leaf
->start
);
88 btrfs_set_header_generation(leaf
, trans
->transid
);
89 btrfs_set_header_owner(leaf
, objectid
);
91 write_extent_buffer(leaf
, root
->fs_info
->fsid
,
92 (unsigned long)btrfs_header_fsid(leaf
),
94 btrfs_mark_buffer_dirty(leaf
);
96 inode_item
= &root_item
.inode
;
97 memset(inode_item
, 0, sizeof(*inode_item
));
98 inode_item
->generation
= cpu_to_le64(1);
99 inode_item
->size
= cpu_to_le64(3);
100 inode_item
->nlink
= cpu_to_le32(1);
101 inode_item
->nblocks
= cpu_to_le64(1);
102 inode_item
->mode
= cpu_to_le32(S_IFDIR
| 0755);
104 btrfs_set_root_bytenr(&root_item
, leaf
->start
);
105 btrfs_set_root_level(&root_item
, 0);
106 btrfs_set_root_refs(&root_item
, 1);
107 btrfs_set_root_used(&root_item
, 0);
109 memset(&root_item
.drop_progress
, 0, sizeof(root_item
.drop_progress
));
110 root_item
.drop_level
= 0;
112 btrfs_tree_unlock(leaf
);
113 free_extent_buffer(leaf
);
116 btrfs_set_root_dirid(&root_item
, new_dirid
);
118 key
.objectid
= objectid
;
120 btrfs_set_key_type(&key
, BTRFS_ROOT_ITEM_KEY
);
121 ret
= btrfs_insert_root(trans
, root
->fs_info
->tree_root
, &key
,
127 * insert the directory item
129 key
.offset
= (u64
)-1;
130 dir
= root
->fs_info
->sb
->s_root
->d_inode
;
131 ret
= btrfs_insert_dir_item(trans
, root
->fs_info
->tree_root
,
132 name
, namelen
, dir
->i_ino
, &key
,
137 ret
= btrfs_insert_inode_ref(trans
, root
->fs_info
->tree_root
,
138 name
, namelen
, objectid
,
139 root
->fs_info
->sb
->s_root
->d_inode
->i_ino
);
143 ret
= btrfs_commit_transaction(trans
, root
);
147 new_root
= btrfs_read_fs_root(root
->fs_info
, &key
, name
, namelen
);
150 trans
= btrfs_start_transaction(new_root
, 1);
153 ret
= btrfs_create_subvol_root(new_root
, trans
, new_dirid
,
154 BTRFS_I(dir
)->block_group
);
158 /* Invalidate existing dcache entry for new subvolume. */
159 btrfs_invalidate_dcache_root(root
, name
, namelen
);
162 nr
= trans
->blocks_used
;
163 err
= btrfs_commit_transaction(trans
, new_root
);
167 mutex_unlock(&root
->fs_info
->fs_mutex
);
168 btrfs_btree_balance_dirty(root
, nr
);
169 btrfs_throttle(root
);
173 static int create_snapshot(struct btrfs_root
*root
, char *name
, int namelen
)
175 struct btrfs_pending_snapshot
*pending_snapshot
;
176 struct btrfs_trans_handle
*trans
;
179 unsigned long nr
= 0;
184 mutex_lock(&root
->fs_info
->fs_mutex
);
185 ret
= btrfs_check_free_space(root
, 1, 0);
189 pending_snapshot
= kmalloc(sizeof(*pending_snapshot
), GFP_NOFS
);
190 if (!pending_snapshot
) {
194 pending_snapshot
->name
= kmalloc(namelen
+ 1, GFP_NOFS
);
195 if (!pending_snapshot
->name
) {
197 kfree(pending_snapshot
);
200 memcpy(pending_snapshot
->name
, name
, namelen
);
201 pending_snapshot
->name
[namelen
] = '\0';
202 trans
= btrfs_start_transaction(root
, 1);
204 pending_snapshot
->root
= root
;
205 list_add(&pending_snapshot
->list
,
206 &trans
->transaction
->pending_snapshots
);
207 ret
= btrfs_update_inode(trans
, root
, root
->inode
);
208 err
= btrfs_commit_transaction(trans
, root
);
211 mutex_unlock(&root
->fs_info
->fs_mutex
);
212 btrfs_btree_balance_dirty(root
, nr
);
213 btrfs_throttle(root
);
217 int btrfs_defrag_file(struct file
*file
)
219 struct inode
*inode
= fdentry(file
)->d_inode
;
220 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
221 struct extent_io_tree
*io_tree
= &BTRFS_I(inode
)->io_tree
;
223 unsigned long last_index
;
224 unsigned long ra_pages
= root
->fs_info
->bdi
.ra_pages
;
225 unsigned long total_read
= 0;
231 mutex_lock(&root
->fs_info
->fs_mutex
);
232 ret
= btrfs_check_free_space(root
, inode
->i_size
, 0);
233 mutex_unlock(&root
->fs_info
->fs_mutex
);
237 mutex_lock(&inode
->i_mutex
);
238 last_index
= inode
->i_size
>> PAGE_CACHE_SHIFT
;
239 for (i
= 0; i
<= last_index
; i
++) {
240 if (total_read
% ra_pages
== 0) {
241 btrfs_force_ra(inode
->i_mapping
, &file
->f_ra
, file
, i
,
242 min(last_index
, i
+ ra_pages
- 1));
245 page
= grab_cache_page(inode
->i_mapping
, i
);
248 if (!PageUptodate(page
)) {
249 btrfs_readpage(NULL
, page
);
251 if (!PageUptodate(page
)) {
253 page_cache_release(page
);
258 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
259 ClearPageDirty(page
);
261 cancel_dirty_page(page
, PAGE_CACHE_SIZE
);
263 wait_on_page_writeback(page
);
264 set_page_extent_mapped(page
);
266 page_start
= (u64
)page
->index
<< PAGE_CACHE_SHIFT
;
267 page_end
= page_start
+ PAGE_CACHE_SIZE
- 1;
269 lock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
270 set_extent_delalloc(io_tree
, page_start
,
273 unlock_extent(io_tree
, page_start
, page_end
, GFP_NOFS
);
274 set_page_dirty(page
);
276 page_cache_release(page
);
277 balance_dirty_pages_ratelimited_nr(inode
->i_mapping
, 1);
281 mutex_unlock(&inode
->i_mutex
);
286 * Called inside transaction, so use GFP_NOFS
289 static int btrfs_ioctl_resize(struct btrfs_root
*root
, void __user
*arg
)
294 struct btrfs_ioctl_vol_args
*vol_args
;
295 struct btrfs_trans_handle
*trans
;
296 struct btrfs_device
*device
= NULL
;
303 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
308 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
312 namelen
= strlen(vol_args
->name
);
313 if (namelen
> BTRFS_VOL_NAME_MAX
) {
318 mutex_lock(&root
->fs_info
->fs_mutex
);
319 sizestr
= vol_args
->name
;
320 devstr
= strchr(sizestr
, ':');
323 sizestr
= devstr
+ 1;
325 devstr
= vol_args
->name
;
326 devid
= simple_strtoull(devstr
, &end
, 10);
327 printk(KERN_INFO
"resizing devid %llu\n", devid
);
329 device
= btrfs_find_device(root
, devid
, NULL
);
331 printk(KERN_INFO
"resizer unable to find device %llu\n", devid
);
335 if (!strcmp(sizestr
, "max"))
336 new_size
= device
->bdev
->bd_inode
->i_size
;
338 if (sizestr
[0] == '-') {
341 } else if (sizestr
[0] == '+') {
345 new_size
= btrfs_parse_size(sizestr
);
352 old_size
= device
->total_bytes
;
355 if (new_size
> old_size
) {
359 new_size
= old_size
- new_size
;
360 } else if (mod
> 0) {
361 new_size
= old_size
+ new_size
;
364 if (new_size
< 256 * 1024 * 1024) {
368 if (new_size
> device
->bdev
->bd_inode
->i_size
) {
373 do_div(new_size
, root
->sectorsize
);
374 new_size
*= root
->sectorsize
;
376 printk(KERN_INFO
"new size for %s is %llu\n",
377 device
->name
, (unsigned long long)new_size
);
379 if (new_size
> old_size
) {
380 trans
= btrfs_start_transaction(root
, 1);
381 ret
= btrfs_grow_device(trans
, device
, new_size
);
382 btrfs_commit_transaction(trans
, root
);
384 ret
= btrfs_shrink_device(device
, new_size
);
388 mutex_unlock(&root
->fs_info
->fs_mutex
);
394 static noinline
int btrfs_ioctl_snap_create(struct btrfs_root
*root
,
397 struct btrfs_ioctl_vol_args
*vol_args
;
398 struct btrfs_dir_item
*di
;
399 struct btrfs_path
*path
;
404 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
409 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
414 namelen
= strlen(vol_args
->name
);
415 if (namelen
> BTRFS_VOL_NAME_MAX
) {
419 if (strchr(vol_args
->name
, '/')) {
424 path
= btrfs_alloc_path();
430 root_dirid
= root
->fs_info
->sb
->s_root
->d_inode
->i_ino
,
431 mutex_lock(&root
->fs_info
->fs_mutex
);
432 di
= btrfs_lookup_dir_item(NULL
, root
->fs_info
->tree_root
,
434 vol_args
->name
, namelen
, 0);
435 mutex_unlock(&root
->fs_info
->fs_mutex
);
436 btrfs_free_path(path
);
438 if (di
&& !IS_ERR(di
)) {
448 if (root
== root
->fs_info
->tree_root
)
449 ret
= create_subvol(root
, vol_args
->name
, namelen
);
451 ret
= create_snapshot(root
, vol_args
->name
, namelen
);
457 static int btrfs_ioctl_defrag(struct file
*file
)
459 struct inode
*inode
= fdentry(file
)->d_inode
;
460 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
462 switch (inode
->i_mode
& S_IFMT
) {
464 mutex_lock(&root
->fs_info
->fs_mutex
);
465 btrfs_defrag_root(root
, 0);
466 btrfs_defrag_root(root
->fs_info
->extent_root
, 0);
467 mutex_unlock(&root
->fs_info
->fs_mutex
);
470 btrfs_defrag_file(file
);
477 long btrfs_ioctl_add_dev(struct btrfs_root
*root
, void __user
*arg
)
479 struct btrfs_ioctl_vol_args
*vol_args
;
482 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
487 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
491 ret
= btrfs_init_new_device(root
, vol_args
->name
);
498 long btrfs_ioctl_rm_dev(struct btrfs_root
*root
, void __user
*arg
)
500 struct btrfs_ioctl_vol_args
*vol_args
;
503 vol_args
= kmalloc(sizeof(*vol_args
), GFP_NOFS
);
508 if (copy_from_user(vol_args
, arg
, sizeof(*vol_args
))) {
512 ret
= btrfs_rm_device(root
, vol_args
->name
);
519 int dup_item_to_inode(struct btrfs_trans_handle
*trans
,
520 struct btrfs_root
*root
,
521 struct btrfs_path
*path
,
522 struct extent_buffer
*leaf
,
524 struct btrfs_key
*key
,
528 int len
= btrfs_item_size_nr(leaf
, slot
);
529 struct btrfs_key ckey
= *key
;
532 dup
= kmalloc(len
, GFP_NOFS
);
536 read_extent_buffer(leaf
, dup
, btrfs_item_ptr_offset(leaf
, slot
), len
);
537 btrfs_release_path(root
, path
);
539 ckey
.objectid
= destino
;
540 ret
= btrfs_insert_item(trans
, root
, &ckey
, dup
, len
);
545 long btrfs_ioctl_clone(struct file
*file
, unsigned long src_fd
)
547 struct inode
*inode
= fdentry(file
)->d_inode
;
548 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
549 struct file
*src_file
;
551 struct btrfs_trans_handle
*trans
;
554 struct btrfs_path
*path
;
555 struct btrfs_key key
;
556 struct extent_buffer
*leaf
;
560 src_file
= fget(src_fd
);
563 src
= src_file
->f_dentry
->d_inode
;
566 if (src
->i_sb
!= inode
->i_sb
)
570 mutex_lock(&inode
->i_mutex
);
571 mutex_lock(&src
->i_mutex
);
573 mutex_lock(&src
->i_mutex
);
574 mutex_lock(&inode
->i_mutex
);
581 /* do any pending delalloc/csum calc on src, one way or
582 another, and lock file content */
584 filemap_write_and_wait(src
->i_mapping
);
585 lock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
586 if (BTRFS_I(src
)->delalloc_bytes
== 0)
588 unlock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
591 mutex_lock(&root
->fs_info
->fs_mutex
);
592 trans
= btrfs_start_transaction(root
, 0);
593 path
= btrfs_alloc_path();
599 key
.type
= BTRFS_EXTENT_DATA_KEY
;
600 key
.objectid
= src
->i_ino
;
606 * note the key will change type as we walk through the
609 ret
= btrfs_search_slot(trans
, root
, &key
, path
, 0, 0);
613 if (path
->slots
[0] >= btrfs_header_nritems(path
->nodes
[0])) {
614 ret
= btrfs_next_leaf(root
, path
);
620 leaf
= path
->nodes
[0];
621 slot
= path
->slots
[0];
622 btrfs_item_key_to_cpu(leaf
, &key
, slot
);
623 nritems
= btrfs_header_nritems(leaf
);
625 if (btrfs_key_type(&key
) > BTRFS_CSUM_ITEM_KEY
||
626 key
.objectid
!= src
->i_ino
)
629 if (btrfs_key_type(&key
) == BTRFS_EXTENT_DATA_KEY
) {
630 struct btrfs_file_extent_item
*extent
;
633 extent
= btrfs_item_ptr(leaf
, slot
,
634 struct btrfs_file_extent_item
);
635 found_type
= btrfs_file_extent_type(leaf
, extent
);
636 if (found_type
== BTRFS_FILE_EXTENT_REG
) {
637 u64 len
= btrfs_file_extent_num_bytes(leaf
,
639 u64 ds
= btrfs_file_extent_disk_bytenr(leaf
,
641 u64 dl
= btrfs_file_extent_disk_num_bytes(leaf
,
643 u64 off
= btrfs_file_extent_offset(leaf
,
645 btrfs_insert_file_extent(trans
, root
,
648 /* ds == 0 means there's a hole */
650 btrfs_inc_extent_ref(trans
, root
,
652 root
->root_key
.objectid
,
656 pos
= key
.offset
+ len
;
657 } else if (found_type
== BTRFS_FILE_EXTENT_INLINE
) {
658 ret
= dup_item_to_inode(trans
, root
, path
,
663 pos
= key
.offset
+ btrfs_item_size_nr(leaf
,
666 } else if (btrfs_key_type(&key
) == BTRFS_CSUM_ITEM_KEY
) {
667 ret
= dup_item_to_inode(trans
, root
, path
, leaf
,
668 slot
, &key
, inode
->i_ino
);
674 btrfs_release_path(root
, path
);
679 btrfs_free_path(path
);
681 inode
->i_blocks
= src
->i_blocks
;
682 i_size_write(inode
, src
->i_size
);
683 btrfs_update_inode(trans
, root
, inode
);
685 unlock_extent(&BTRFS_I(src
)->io_tree
, 0, (u64
)-1, GFP_NOFS
);
687 btrfs_end_transaction(trans
, root
);
688 mutex_unlock(&root
->fs_info
->fs_mutex
);
691 mutex_unlock(&src
->i_mutex
);
692 mutex_unlock(&inode
->i_mutex
);
699 * there are many ways the trans_start and trans_end ioctls can lead
700 * to deadlocks. They should only be used by applications that
701 * basically own the machine, and have a very in depth understanding
702 * of all the possible deadlocks and enospc problems.
704 long btrfs_ioctl_trans_start(struct file
*file
)
706 struct inode
*inode
= fdentry(file
)->d_inode
;
707 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
708 struct btrfs_trans_handle
*trans
;
711 if (!capable(CAP_SYS_ADMIN
))
714 mutex_lock(&root
->fs_info
->fs_mutex
);
715 if (file
->private_data
) {
719 trans
= btrfs_start_transaction(root
, 0);
721 file
->private_data
= trans
;
724 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
726 mutex_unlock(&root
->fs_info
->fs_mutex
);
731 * there are many ways the trans_start and trans_end ioctls can lead
732 * to deadlocks. They should only be used by applications that
733 * basically own the machine, and have a very in depth understanding
734 * of all the possible deadlocks and enospc problems.
736 long btrfs_ioctl_trans_end(struct file
*file
)
738 struct inode
*inode
= fdentry(file
)->d_inode
;
739 struct btrfs_root
*root
= BTRFS_I(inode
)->root
;
740 struct btrfs_trans_handle
*trans
;
743 mutex_lock(&root
->fs_info
->fs_mutex
);
744 trans
= file
->private_data
;
749 btrfs_end_transaction(trans
, root
);
750 file
->private_data
= 0;
752 mutex_unlock(&root
->fs_info
->fs_mutex
);
756 long btrfs_ioctl(struct file
*file
, unsigned int
757 cmd
, unsigned long arg
)
759 struct btrfs_root
*root
= BTRFS_I(fdentry(file
)->d_inode
)->root
;
762 case BTRFS_IOC_SNAP_CREATE
:
763 return btrfs_ioctl_snap_create(root
, (void __user
*)arg
);
764 case BTRFS_IOC_DEFRAG
:
765 return btrfs_ioctl_defrag(file
);
766 case BTRFS_IOC_RESIZE
:
767 return btrfs_ioctl_resize(root
, (void __user
*)arg
);
768 case BTRFS_IOC_ADD_DEV
:
769 return btrfs_ioctl_add_dev(root
, (void __user
*)arg
);
770 case BTRFS_IOC_RM_DEV
:
771 return btrfs_ioctl_rm_dev(root
, (void __user
*)arg
);
772 case BTRFS_IOC_BALANCE
:
773 return btrfs_balance(root
->fs_info
->dev_root
);
774 case BTRFS_IOC_CLONE
:
775 return btrfs_ioctl_clone(file
, arg
);
776 case BTRFS_IOC_TRANS_START
:
777 return btrfs_ioctl_trans_start(file
);
778 case BTRFS_IOC_TRANS_END
:
779 return btrfs_ioctl_trans_end(file
);
781 btrfs_sync_fs(file
->f_dentry
->d_sb
, 1);