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 #define _XOPEN_SOURCE 600
24 #include <sys/types.h>
28 #include "kerncompat.h"
29 #include "radix-tree.h"
33 #include "transaction.h"
36 #include "print-tree.h"
38 static int check_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
)
41 struct btrfs_fs_devices
*fs_devices
;
44 if (buf
->start
!= btrfs_header_bytenr(buf
)) {
45 printk("Check tree block failed, want=%Lu, have=%Lu\n",
46 buf
->start
, btrfs_header_bytenr(buf
));
50 fs_devices
= root
->fs_info
->fs_devices
;
52 if (!memcmp_extent_buffer(buf
, fs_devices
->fsid
,
58 fs_devices
= fs_devices
->seed
;
63 u32
btrfs_csum_data(struct btrfs_root
*root
, char *data
, u32 seed
, size_t len
)
65 return crc32c(seed
, data
, len
);
68 void btrfs_csum_final(u32 crc
, char *result
)
70 *(__le32
*)result
= ~cpu_to_le32(crc
);
73 static int __csum_tree_block_size(struct extent_buffer
*buf
, u16 csum_size
,
74 int verify
, int silent
)
80 result
= malloc(csum_size
* sizeof(char));
84 len
= buf
->len
- BTRFS_CSUM_SIZE
;
85 crc
= crc32c(crc
, buf
->data
+ BTRFS_CSUM_SIZE
, len
);
86 btrfs_csum_final(crc
, result
);
89 if (memcmp_extent_buffer(buf
, result
, 0, csum_size
)) {
91 printk("checksum verify failed on %llu found %08X wanted %08X\n",
92 (unsigned long long)buf
->start
,
94 *((u32
*)(char *)buf
->data
));
99 write_extent_buffer(buf
, result
, 0, csum_size
);
105 int csum_tree_block_size(struct extent_buffer
*buf
, u16 csum_size
, int verify
)
107 return __csum_tree_block_size(buf
, csum_size
, verify
, 0);
110 int verify_tree_block_csum_silent(struct extent_buffer
*buf
, u16 csum_size
)
112 return __csum_tree_block_size(buf
, csum_size
, 1, 1);
115 int csum_tree_block(struct btrfs_root
*root
, struct extent_buffer
*buf
,
119 btrfs_super_csum_size(root
->fs_info
->super_copy
);
120 return csum_tree_block_size(buf
, csum_size
, verify
);
123 struct extent_buffer
*btrfs_find_tree_block(struct btrfs_root
*root
,
124 u64 bytenr
, u32 blocksize
)
126 return find_extent_buffer(&root
->fs_info
->extent_cache
,
130 struct extent_buffer
*btrfs_find_create_tree_block(struct btrfs_root
*root
,
131 u64 bytenr
, u32 blocksize
)
133 return alloc_extent_buffer(&root
->fs_info
->extent_cache
, bytenr
,
137 int readahead_tree_block(struct btrfs_root
*root
, u64 bytenr
, u32 blocksize
,
141 struct extent_buffer
*eb
;
143 struct btrfs_multi_bio
*multi
= NULL
;
144 struct btrfs_device
*device
;
146 eb
= btrfs_find_tree_block(root
, bytenr
, blocksize
);
147 if (eb
&& btrfs_buffer_uptodate(eb
, parent_transid
)) {
148 free_extent_buffer(eb
);
153 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, READ
,
154 bytenr
, &length
, &multi
, 0, NULL
);
156 device
= multi
->stripes
[0].dev
;
158 blocksize
= min(blocksize
, (u32
)(64 * 1024));
159 readahead(device
->fd
, multi
->stripes
[0].physical
, blocksize
);
164 static int verify_parent_transid(struct extent_io_tree
*io_tree
,
165 struct extent_buffer
*eb
, u64 parent_transid
,
170 if (!parent_transid
|| btrfs_header_generation(eb
) == parent_transid
)
173 if (extent_buffer_uptodate(eb
) &&
174 btrfs_header_generation(eb
) == parent_transid
) {
178 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
179 (unsigned long long)eb
->start
,
180 (unsigned long long)parent_transid
,
181 (unsigned long long)btrfs_header_generation(eb
));
183 eb
->flags
|= EXTENT_BAD_TRANSID
;
184 printk("Ignoring transid failure\n");
190 clear_extent_buffer_uptodate(io_tree
, eb
);
196 int read_whole_eb(struct btrfs_fs_info
*info
, struct extent_buffer
*eb
, int mirror
)
198 unsigned long offset
= 0;
199 struct btrfs_multi_bio
*multi
= NULL
;
200 struct btrfs_device
*device
;
203 unsigned long bytes_left
= eb
->len
;
206 read_len
= bytes_left
;
209 if (!info
->on_restoring
) {
210 ret
= btrfs_map_block(&info
->mapping_tree
, READ
,
211 eb
->start
+ offset
, &read_len
, &multi
,
214 printk("Couldn't map the block %Lu\n", eb
->start
+ offset
);
218 device
= multi
->stripes
[0].dev
;
220 if (device
->fd
== 0) {
227 eb
->dev_bytenr
= multi
->stripes
[0].physical
;
231 /* special case for restore metadump */
232 list_for_each_entry(device
, &info
->fs_devices
->devices
, dev_list
) {
233 if (device
->devid
== 1)
238 eb
->dev_bytenr
= eb
->start
;
242 if (read_len
> bytes_left
)
243 read_len
= bytes_left
;
245 ret
= read_extent_from_disk(eb
, offset
, read_len
);
249 bytes_left
-= read_len
;
254 struct extent_buffer
*read_tree_block(struct btrfs_root
*root
, u64 bytenr
,
255 u32 blocksize
, u64 parent_transid
)
258 struct extent_buffer
*eb
;
259 u64 best_transid
= 0;
265 eb
= btrfs_find_create_tree_block(root
, bytenr
, blocksize
);
269 if (btrfs_buffer_uptodate(eb
, parent_transid
))
273 ret
= read_whole_eb(root
->fs_info
, eb
, mirror_num
);
274 if (ret
== 0 && check_tree_block(root
, eb
) == 0 &&
275 csum_tree_block(root
, eb
, 1) == 0 &&
276 verify_parent_transid(eb
->tree
, eb
, parent_transid
, ignore
)
278 if (eb
->flags
& EXTENT_BAD_TRANSID
&&
279 list_empty(&eb
->recow
)) {
280 list_add_tail(&eb
->recow
,
281 &root
->fs_info
->recow_ebs
);
284 btrfs_set_buffer_uptodate(eb
);
288 if (check_tree_block(root
, eb
))
289 printk("read block failed check_tree_block\n");
291 printk("Csum didn't match\n");
294 num_copies
= btrfs_num_copies(&root
->fs_info
->mapping_tree
,
296 if (num_copies
== 1) {
300 if (btrfs_header_generation(eb
) > best_transid
&& mirror_num
) {
301 best_transid
= btrfs_header_generation(eb
);
302 good_mirror
= mirror_num
;
305 if (mirror_num
> num_copies
) {
306 mirror_num
= good_mirror
;
311 free_extent_buffer(eb
);
315 int write_and_map_eb(struct btrfs_trans_handle
*trans
,
316 struct btrfs_root
*root
,
317 struct extent_buffer
*eb
)
322 u64
*raid_map
= NULL
;
323 struct btrfs_multi_bio
*multi
= NULL
;
327 ret
= btrfs_map_block(&root
->fs_info
->mapping_tree
, WRITE
,
328 eb
->start
, &length
, &multi
, 0, &raid_map
);
331 ret
= write_raid56_with_parity(root
->fs_info
, eb
, multi
,
334 } else while (dev_nr
< multi
->num_stripes
) {
336 eb
->fd
= multi
->stripes
[dev_nr
].dev
->fd
;
337 eb
->dev_bytenr
= multi
->stripes
[dev_nr
].physical
;
338 multi
->stripes
[dev_nr
].dev
->total_ios
++;
340 ret
= write_extent_to_disk(eb
);
347 static int write_tree_block(struct btrfs_trans_handle
*trans
,
348 struct btrfs_root
*root
,
349 struct extent_buffer
*eb
)
351 if (check_tree_block(root
, eb
))
354 if (!btrfs_buffer_uptodate(eb
, trans
->transid
))
357 btrfs_set_header_flag(eb
, BTRFS_HEADER_FLAG_WRITTEN
);
358 csum_tree_block(root
, eb
, 0);
360 return write_and_map_eb(trans
, root
, eb
);
363 int __setup_root(u32 nodesize
, u32 leafsize
, u32 sectorsize
,
364 u32 stripesize
, struct btrfs_root
*root
,
365 struct btrfs_fs_info
*fs_info
, u64 objectid
)
368 root
->commit_root
= NULL
;
369 root
->sectorsize
= sectorsize
;
370 root
->nodesize
= nodesize
;
371 root
->leafsize
= leafsize
;
372 root
->stripesize
= stripesize
;
374 root
->track_dirty
= 0;
376 root
->fs_info
= fs_info
;
377 root
->objectid
= objectid
;
378 root
->last_trans
= 0;
379 root
->highest_inode
= 0;
380 root
->last_inode_alloc
= 0;
382 INIT_LIST_HEAD(&root
->dirty_list
);
383 memset(&root
->root_key
, 0, sizeof(root
->root_key
));
384 memset(&root
->root_item
, 0, sizeof(root
->root_item
));
385 root
->root_key
.objectid
= objectid
;
389 static int update_cowonly_root(struct btrfs_trans_handle
*trans
,
390 struct btrfs_root
*root
)
394 struct btrfs_root
*tree_root
= root
->fs_info
->tree_root
;
396 btrfs_write_dirty_block_groups(trans
, root
);
398 old_root_bytenr
= btrfs_root_bytenr(&root
->root_item
);
399 if (old_root_bytenr
== root
->node
->start
)
401 btrfs_set_root_bytenr(&root
->root_item
,
403 btrfs_set_root_generation(&root
->root_item
,
405 root
->root_item
.level
= btrfs_header_level(root
->node
);
406 ret
= btrfs_update_root(trans
, tree_root
,
410 btrfs_write_dirty_block_groups(trans
, root
);
415 static int commit_tree_roots(struct btrfs_trans_handle
*trans
,
416 struct btrfs_fs_info
*fs_info
)
418 struct btrfs_root
*root
;
419 struct list_head
*next
;
420 struct extent_buffer
*eb
;
423 if (fs_info
->readonly
)
426 eb
= fs_info
->tree_root
->node
;
427 extent_buffer_get(eb
);
428 ret
= btrfs_cow_block(trans
, fs_info
->tree_root
, eb
, NULL
, 0, &eb
);
429 free_extent_buffer(eb
);
433 while(!list_empty(&fs_info
->dirty_cowonly_roots
)) {
434 next
= fs_info
->dirty_cowonly_roots
.next
;
436 root
= list_entry(next
, struct btrfs_root
, dirty_list
);
437 update_cowonly_root(trans
, root
);
438 free_extent_buffer(root
->commit_root
);
439 root
->commit_root
= NULL
;
445 static int __commit_transaction(struct btrfs_trans_handle
*trans
,
446 struct btrfs_root
*root
)
450 struct extent_buffer
*eb
;
451 struct extent_io_tree
*tree
= &root
->fs_info
->extent_cache
;
455 ret
= find_first_extent_bit(tree
, 0, &start
, &end
,
459 while(start
<= end
) {
460 eb
= find_first_extent_buffer(tree
, start
);
461 BUG_ON(!eb
|| eb
->start
!= start
);
462 ret
= write_tree_block(trans
, root
, eb
);
465 clear_extent_buffer_dirty(eb
);
466 free_extent_buffer(eb
);
472 int btrfs_commit_transaction(struct btrfs_trans_handle
*trans
,
473 struct btrfs_root
*root
)
475 u64 transid
= trans
->transid
;
477 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
479 if (root
->commit_root
== root
->node
)
482 free_extent_buffer(root
->commit_root
);
483 root
->commit_root
= NULL
;
485 btrfs_set_root_bytenr(&root
->root_item
, root
->node
->start
);
486 btrfs_set_root_generation(&root
->root_item
, trans
->transid
);
487 root
->root_item
.level
= btrfs_header_level(root
->node
);
488 ret
= btrfs_update_root(trans
, root
->fs_info
->tree_root
,
489 &root
->root_key
, &root
->root_item
);
492 ret
= commit_tree_roots(trans
, fs_info
);
494 ret
= __commit_transaction(trans
, root
);
496 write_ctree_super(trans
, root
);
497 btrfs_finish_extent_commit(trans
, fs_info
->extent_root
,
498 &fs_info
->pinned_extents
);
499 btrfs_free_transaction(root
, trans
);
500 free_extent_buffer(root
->commit_root
);
501 root
->commit_root
= NULL
;
502 fs_info
->running_transaction
= NULL
;
503 fs_info
->last_trans_committed
= transid
;
507 static int find_and_setup_root(struct btrfs_root
*tree_root
,
508 struct btrfs_fs_info
*fs_info
,
509 u64 objectid
, struct btrfs_root
*root
)
515 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
516 tree_root
->sectorsize
, tree_root
->stripesize
,
517 root
, fs_info
, objectid
);
518 ret
= btrfs_find_last_root(tree_root
, objectid
,
519 &root
->root_item
, &root
->root_key
);
523 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
524 generation
= btrfs_root_generation(&root
->root_item
);
525 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
526 blocksize
, generation
);
527 if (!extent_buffer_uptodate(root
->node
))
533 static int find_and_setup_log_root(struct btrfs_root
*tree_root
,
534 struct btrfs_fs_info
*fs_info
,
535 struct btrfs_super_block
*disk_super
)
538 u64 blocknr
= btrfs_super_log_root(disk_super
);
539 struct btrfs_root
*log_root
= malloc(sizeof(struct btrfs_root
));
549 blocksize
= btrfs_level_size(tree_root
,
550 btrfs_super_log_root_level(disk_super
));
552 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
553 tree_root
->sectorsize
, tree_root
->stripesize
,
554 log_root
, fs_info
, BTRFS_TREE_LOG_OBJECTID
);
556 log_root
->node
= read_tree_block(tree_root
, blocknr
,
558 btrfs_super_generation(disk_super
) + 1);
560 fs_info
->log_root_tree
= log_root
;
562 if (!extent_buffer_uptodate(log_root
->node
)) {
563 free_extent_buffer(log_root
->node
);
565 fs_info
->log_root_tree
= NULL
;
573 int btrfs_free_fs_root(struct btrfs_root
*root
)
576 free_extent_buffer(root
->node
);
577 if (root
->commit_root
)
578 free_extent_buffer(root
->commit_root
);
583 static void __free_fs_root(struct rb_node
*node
)
585 struct btrfs_root
*root
;
587 root
= container_of(node
, struct btrfs_root
, rb_node
);
588 btrfs_free_fs_root(root
);
591 FREE_RB_BASED_TREE(fs_roots
, __free_fs_root
);
593 struct btrfs_root
*btrfs_read_fs_root_no_cache(struct btrfs_fs_info
*fs_info
,
594 struct btrfs_key
*location
)
596 struct btrfs_root
*root
;
597 struct btrfs_root
*tree_root
= fs_info
->tree_root
;
598 struct btrfs_path
*path
;
599 struct extent_buffer
*l
;
604 root
= malloc(sizeof(*root
));
606 return ERR_PTR(-ENOMEM
);
607 memset(root
, 0, sizeof(*root
));
608 if (location
->offset
== (u64
)-1) {
609 ret
= find_and_setup_root(tree_root
, fs_info
,
610 location
->objectid
, root
);
618 __setup_root(tree_root
->nodesize
, tree_root
->leafsize
,
619 tree_root
->sectorsize
, tree_root
->stripesize
,
620 root
, fs_info
, location
->objectid
);
622 path
= btrfs_alloc_path();
624 ret
= btrfs_search_slot(NULL
, tree_root
, location
, path
, 0, 0);
631 read_extent_buffer(l
, &root
->root_item
,
632 btrfs_item_ptr_offset(l
, path
->slots
[0]),
633 sizeof(root
->root_item
));
634 memcpy(&root
->root_key
, location
, sizeof(*location
));
637 btrfs_release_path(path
);
638 btrfs_free_path(path
);
643 generation
= btrfs_root_generation(&root
->root_item
);
644 blocksize
= btrfs_level_size(root
, btrfs_root_level(&root
->root_item
));
645 root
->node
= read_tree_block(root
, btrfs_root_bytenr(&root
->root_item
),
646 blocksize
, generation
);
649 return ERR_PTR(-EIO
);
656 static int btrfs_fs_roots_compare_objectids(struct rb_node
*node
,
659 u64 objectid
= *((u64
*)data
);
660 struct btrfs_root
*root
;
662 root
= rb_entry(node
, struct btrfs_root
, rb_node
);
663 if (objectid
> root
->objectid
)
665 else if (objectid
< root
->objectid
)
671 static int btrfs_fs_roots_compare_roots(struct rb_node
*node1
,
672 struct rb_node
*node2
)
674 struct btrfs_root
*root
;
676 root
= rb_entry(node2
, struct btrfs_root
, rb_node
);
677 return btrfs_fs_roots_compare_objectids(node1
, (void *)&root
->objectid
);
680 struct btrfs_root
*btrfs_read_fs_root(struct btrfs_fs_info
*fs_info
,
681 struct btrfs_key
*location
)
683 struct btrfs_root
*root
;
684 struct rb_node
*node
;
686 u64 objectid
= location
->objectid
;
688 if (location
->objectid
== BTRFS_ROOT_TREE_OBJECTID
)
689 return fs_info
->tree_root
;
690 if (location
->objectid
== BTRFS_EXTENT_TREE_OBJECTID
)
691 return fs_info
->extent_root
;
692 if (location
->objectid
== BTRFS_CHUNK_TREE_OBJECTID
)
693 return fs_info
->chunk_root
;
694 if (location
->objectid
== BTRFS_DEV_TREE_OBJECTID
)
695 return fs_info
->dev_root
;
696 if (location
->objectid
== BTRFS_CSUM_TREE_OBJECTID
)
697 return fs_info
->csum_root
;
699 BUG_ON(location
->objectid
== BTRFS_TREE_RELOC_OBJECTID
||
700 location
->offset
!= (u64
)-1);
702 node
= rb_search(&fs_info
->fs_root_tree
, (void *)&objectid
,
703 btrfs_fs_roots_compare_objectids
, NULL
);
705 return container_of(node
, struct btrfs_root
, rb_node
);
707 root
= btrfs_read_fs_root_no_cache(fs_info
, location
);
711 ret
= rb_insert(&fs_info
->fs_root_tree
, &root
->rb_node
,
712 btrfs_fs_roots_compare_roots
);
717 void btrfs_free_fs_info(struct btrfs_fs_info
*fs_info
)
719 free(fs_info
->tree_root
);
720 free(fs_info
->extent_root
);
721 free(fs_info
->chunk_root
);
722 free(fs_info
->dev_root
);
723 free(fs_info
->csum_root
);
724 free(fs_info
->super_copy
);
725 free(fs_info
->log_root_tree
);
729 struct btrfs_fs_info
*btrfs_new_fs_info(int writable
, u64 sb_bytenr
)
731 struct btrfs_fs_info
*fs_info
;
733 fs_info
= malloc(sizeof(struct btrfs_fs_info
));
737 memset(fs_info
, 0, sizeof(struct btrfs_fs_info
));
739 fs_info
->tree_root
= malloc(sizeof(struct btrfs_root
));
740 fs_info
->extent_root
= malloc(sizeof(struct btrfs_root
));
741 fs_info
->chunk_root
= malloc(sizeof(struct btrfs_root
));
742 fs_info
->dev_root
= malloc(sizeof(struct btrfs_root
));
743 fs_info
->csum_root
= malloc(sizeof(struct btrfs_root
));
744 fs_info
->super_copy
= malloc(BTRFS_SUPER_INFO_SIZE
);
746 if (!fs_info
->tree_root
|| !fs_info
->extent_root
||
747 !fs_info
->chunk_root
|| !fs_info
->dev_root
||
748 !fs_info
->csum_root
|| !fs_info
->super_copy
)
751 memset(fs_info
->super_copy
, 0, BTRFS_SUPER_INFO_SIZE
);
752 memset(fs_info
->tree_root
, 0, sizeof(struct btrfs_root
));
753 memset(fs_info
->extent_root
, 0, sizeof(struct btrfs_root
));
754 memset(fs_info
->chunk_root
, 0, sizeof(struct btrfs_root
));
755 memset(fs_info
->dev_root
, 0, sizeof(struct btrfs_root
));
756 memset(fs_info
->csum_root
, 0, sizeof(struct btrfs_root
));
758 extent_io_tree_init(&fs_info
->extent_cache
);
759 extent_io_tree_init(&fs_info
->free_space_cache
);
760 extent_io_tree_init(&fs_info
->block_group_cache
);
761 extent_io_tree_init(&fs_info
->pinned_extents
);
762 extent_io_tree_init(&fs_info
->pending_del
);
763 extent_io_tree_init(&fs_info
->extent_ins
);
764 fs_info
->fs_root_tree
= RB_ROOT
;
765 cache_tree_init(&fs_info
->mapping_tree
.cache_tree
);
767 mutex_init(&fs_info
->fs_mutex
);
768 INIT_LIST_HEAD(&fs_info
->dirty_cowonly_roots
);
769 INIT_LIST_HEAD(&fs_info
->space_info
);
770 INIT_LIST_HEAD(&fs_info
->recow_ebs
);
773 fs_info
->readonly
= 1;
775 fs_info
->super_bytenr
= sb_bytenr
;
776 fs_info
->data_alloc_profile
= (u64
)-1;
777 fs_info
->metadata_alloc_profile
= (u64
)-1;
778 fs_info
->system_alloc_profile
= fs_info
->metadata_alloc_profile
;
781 btrfs_free_fs_info(fs_info
);
785 int btrfs_check_fs_compatibility(struct btrfs_super_block
*sb
, int writable
)
789 features
= btrfs_super_incompat_flags(sb
) &
790 ~BTRFS_FEATURE_INCOMPAT_SUPP
;
792 printk("couldn't open because of unsupported "
793 "option features (%Lx).\n",
794 (unsigned long long)features
);
798 features
= btrfs_super_incompat_flags(sb
);
799 if (!(features
& BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF
)) {
800 features
|= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF
;
801 btrfs_set_super_incompat_flags(sb
, features
);
804 features
= btrfs_super_compat_ro_flags(sb
) &
805 ~BTRFS_FEATURE_COMPAT_RO_SUPP
;
806 if (writable
&& features
) {
807 printk("couldn't open RDWR because of unsupported "
808 "option features (%Lx).\n",
809 (unsigned long long)features
);
815 static int find_best_backup_root(struct btrfs_super_block
*super
)
817 struct btrfs_root_backup
*backup
;
818 u64 orig_gen
= btrfs_super_generation(super
);
823 for (i
= 0; i
< BTRFS_NUM_BACKUP_ROOTS
; i
++) {
824 backup
= super
->super_roots
+ i
;
825 if (btrfs_backup_tree_root_gen(backup
) != orig_gen
&&
826 btrfs_backup_tree_root_gen(backup
) > gen
) {
828 gen
= btrfs_backup_tree_root_gen(backup
);
834 int btrfs_setup_all_roots(struct btrfs_fs_info
*fs_info
, u64 root_tree_bytenr
,
835 enum btrfs_open_ctree_flags flags
)
837 struct btrfs_super_block
*sb
= fs_info
->super_copy
;
838 struct btrfs_root
*root
;
839 struct btrfs_key key
;
848 nodesize
= btrfs_super_nodesize(sb
);
849 leafsize
= btrfs_super_leafsize(sb
);
850 sectorsize
= btrfs_super_sectorsize(sb
);
851 stripesize
= btrfs_super_stripesize(sb
);
853 root
= fs_info
->tree_root
;
854 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
855 root
, fs_info
, BTRFS_ROOT_TREE_OBJECTID
);
856 blocksize
= btrfs_level_size(root
, btrfs_super_root_level(sb
));
857 generation
= btrfs_super_generation(sb
);
859 if (!root_tree_bytenr
&& !(flags
& OPEN_CTREE_BACKUP_ROOT
)) {
860 root_tree_bytenr
= btrfs_super_root(sb
);
861 } else if (flags
& OPEN_CTREE_BACKUP_ROOT
) {
862 struct btrfs_root_backup
*backup
;
863 int index
= find_best_backup_root(sb
);
864 if (index
>= BTRFS_NUM_BACKUP_ROOTS
) {
865 fprintf(stderr
, "Invalid backup root number\n");
868 backup
= fs_info
->super_copy
->super_roots
+ index
;
869 root_tree_bytenr
= btrfs_backup_tree_root(backup
);
870 generation
= btrfs_backup_tree_root_gen(backup
);
873 root
->node
= read_tree_block(root
, root_tree_bytenr
, blocksize
,
875 if (!extent_buffer_uptodate(root
->node
)) {
876 fprintf(stderr
, "Couldn't read tree root\n");
880 ret
= find_and_setup_root(root
, fs_info
, BTRFS_EXTENT_TREE_OBJECTID
,
881 fs_info
->extent_root
);
883 printk("Couldn't setup extent tree\n");
884 if (!(flags
& OPEN_CTREE_PARTIAL
))
886 /* Need a blank node here just so we don't screw up in the
887 * million of places that assume a root has a valid ->node
889 fs_info
->extent_root
->node
=
890 btrfs_find_create_tree_block(fs_info
->extent_root
, 0,
892 if (!fs_info
->extent_root
->node
)
894 clear_extent_buffer_uptodate(NULL
, fs_info
->extent_root
->node
);
896 fs_info
->extent_root
->track_dirty
= 1;
898 ret
= find_and_setup_root(root
, fs_info
, BTRFS_DEV_TREE_OBJECTID
,
901 printk("Couldn't setup device tree\n");
904 fs_info
->dev_root
->track_dirty
= 1;
906 ret
= find_and_setup_root(root
, fs_info
, BTRFS_CSUM_TREE_OBJECTID
,
909 printk("Couldn't setup csum tree\n");
910 if (!(flags
& OPEN_CTREE_PARTIAL
))
913 fs_info
->csum_root
->track_dirty
= 1;
915 ret
= find_and_setup_log_root(root
, fs_info
, sb
);
917 printk("Couldn't setup log root tree\n");
921 fs_info
->generation
= generation
;
922 fs_info
->last_trans_committed
= generation
;
923 if (extent_buffer_uptodate(fs_info
->extent_root
->node
) &&
924 !(flags
& OPEN_CTREE_NO_BLOCK_GROUPS
))
925 btrfs_read_block_groups(fs_info
->tree_root
);
927 key
.objectid
= BTRFS_FS_TREE_OBJECTID
;
928 key
.type
= BTRFS_ROOT_ITEM_KEY
;
929 key
.offset
= (u64
)-1;
930 fs_info
->fs_root
= btrfs_read_fs_root(fs_info
, &key
);
932 if (!fs_info
->fs_root
)
937 void btrfs_release_all_roots(struct btrfs_fs_info
*fs_info
)
939 if (fs_info
->csum_root
)
940 free_extent_buffer(fs_info
->csum_root
->node
);
941 if (fs_info
->dev_root
)
942 free_extent_buffer(fs_info
->dev_root
->node
);
943 if (fs_info
->extent_root
)
944 free_extent_buffer(fs_info
->extent_root
->node
);
945 if (fs_info
->tree_root
)
946 free_extent_buffer(fs_info
->tree_root
->node
);
947 if (fs_info
->log_root_tree
)
948 free_extent_buffer(fs_info
->log_root_tree
->node
);
949 if (fs_info
->chunk_root
)
950 free_extent_buffer(fs_info
->chunk_root
->node
);
953 static void free_map_lookup(struct cache_extent
*ce
)
955 struct map_lookup
*map
;
957 map
= container_of(ce
, struct map_lookup
, ce
);
961 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache
, free_map_lookup
);
963 void btrfs_cleanup_all_caches(struct btrfs_fs_info
*fs_info
)
965 while (!list_empty(&fs_info
->recow_ebs
)) {
966 struct extent_buffer
*eb
;
967 eb
= list_first_entry(&fs_info
->recow_ebs
,
968 struct extent_buffer
, recow
);
969 list_del_init(&eb
->recow
);
970 free_extent_buffer(eb
);
972 free_mapping_cache_tree(&fs_info
->mapping_tree
.cache_tree
);
973 extent_io_tree_cleanup(&fs_info
->extent_cache
);
974 extent_io_tree_cleanup(&fs_info
->free_space_cache
);
975 extent_io_tree_cleanup(&fs_info
->block_group_cache
);
976 extent_io_tree_cleanup(&fs_info
->pinned_extents
);
977 extent_io_tree_cleanup(&fs_info
->pending_del
);
978 extent_io_tree_cleanup(&fs_info
->extent_ins
);
981 int btrfs_scan_fs_devices(int fd
, const char *path
,
982 struct btrfs_fs_devices
**fs_devices
,
983 u64 sb_bytenr
, int run_ioctl
)
988 sb_bytenr
= BTRFS_SUPER_INFO_OFFSET
;
990 ret
= btrfs_scan_one_device(fd
, path
, fs_devices
,
991 &total_devs
, sb_bytenr
);
993 fprintf(stderr
, "No valid Btrfs found on %s\n", path
);
997 if (total_devs
!= 1) {
998 ret
= btrfs_scan_for_fsid(run_ioctl
);
1005 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info
*fs_info
)
1007 struct btrfs_super_block
*sb
= fs_info
->super_copy
;
1016 nodesize
= btrfs_super_nodesize(sb
);
1017 leafsize
= btrfs_super_leafsize(sb
);
1018 sectorsize
= btrfs_super_sectorsize(sb
);
1019 stripesize
= btrfs_super_stripesize(sb
);
1021 __setup_root(nodesize
, leafsize
, sectorsize
, stripesize
,
1022 fs_info
->chunk_root
, fs_info
, BTRFS_CHUNK_TREE_OBJECTID
);
1024 ret
= btrfs_read_sys_array(fs_info
->chunk_root
);
1028 blocksize
= btrfs_level_size(fs_info
->chunk_root
,
1029 btrfs_super_chunk_root_level(sb
));
1030 generation
= btrfs_super_chunk_root_generation(sb
);
1032 fs_info
->chunk_root
->node
= read_tree_block(fs_info
->chunk_root
,
1033 btrfs_super_chunk_root(sb
),
1034 blocksize
, generation
);
1035 if (!fs_info
->chunk_root
->node
||
1036 !extent_buffer_uptodate(fs_info
->chunk_root
->node
)) {
1037 fprintf(stderr
, "Couldn't read chunk root\n");
1041 if (!(btrfs_super_flags(sb
) & BTRFS_SUPER_FLAG_METADUMP
)) {
1042 ret
= btrfs_read_chunk_tree(fs_info
->chunk_root
);
1044 fprintf(stderr
, "Couldn't read chunk tree\n");
1051 static struct btrfs_fs_info
*__open_ctree_fd(int fp
, const char *path
,
1053 u64 root_tree_bytenr
,
1054 enum btrfs_open_ctree_flags flags
)
1056 struct btrfs_fs_info
*fs_info
;
1057 struct btrfs_super_block
*disk_super
;
1058 struct btrfs_fs_devices
*fs_devices
= NULL
;
1059 struct extent_buffer
*eb
;
1064 sb_bytenr
= BTRFS_SUPER_INFO_OFFSET
;
1066 /* try to drop all the caches */
1067 if (posix_fadvise(fp
, 0, 0, POSIX_FADV_DONTNEED
))
1068 fprintf(stderr
, "Warning, could not drop caches\n");
1070 fs_info
= btrfs_new_fs_info(flags
& OPEN_CTREE_WRITES
, sb_bytenr
);
1072 fprintf(stderr
, "Failed to allocate memory for fs_info\n");
1075 if (flags
& OPEN_CTREE_RESTORE
)
1076 fs_info
->on_restoring
= 1;
1078 ret
= btrfs_scan_fs_devices(fp
, path
, &fs_devices
, sb_bytenr
,
1079 !(flags
& OPEN_CTREE_RECOVER_SUPER
));
1083 fs_info
->fs_devices
= fs_devices
;
1084 if (flags
& OPEN_CTREE_WRITES
)
1089 if (flags
& OPEN_CTREE_EXCLUSIVE
)
1092 ret
= btrfs_open_devices(fs_devices
, oflags
);
1097 disk_super
= fs_info
->super_copy
;
1098 if (!(flags
& OPEN_CTREE_RECOVER_SUPER
))
1099 ret
= btrfs_read_dev_super(fs_devices
->latest_bdev
,
1100 disk_super
, sb_bytenr
);
1102 ret
= btrfs_read_dev_super(fp
, disk_super
, sb_bytenr
);
1104 printk("No valid btrfs found\n");
1108 memcpy(fs_info
->fsid
, &disk_super
->fsid
, BTRFS_FSID_SIZE
);
1110 ret
= btrfs_check_fs_compatibility(fs_info
->super_copy
,
1111 flags
& OPEN_CTREE_WRITES
);
1115 ret
= btrfs_setup_chunk_tree_and_device_map(fs_info
);
1119 eb
= fs_info
->chunk_root
->node
;
1120 read_extent_buffer(eb
, fs_info
->chunk_tree_uuid
,
1121 btrfs_header_chunk_tree_uuid(eb
),
1124 ret
= btrfs_setup_all_roots(fs_info
, root_tree_bytenr
, flags
);
1131 if (flags
& OPEN_CTREE_PARTIAL
)
1134 btrfs_release_all_roots(fs_info
);
1135 btrfs_cleanup_all_caches(fs_info
);
1137 btrfs_close_devices(fs_devices
);
1139 btrfs_free_fs_info(fs_info
);
1143 struct btrfs_fs_info
*open_ctree_fs_info(const char *filename
,
1144 u64 sb_bytenr
, u64 root_tree_bytenr
,
1145 enum btrfs_open_ctree_flags flags
)
1148 struct btrfs_fs_info
*info
;
1149 int oflags
= O_CREAT
| O_RDWR
;
1151 if (!(flags
& OPEN_CTREE_WRITES
))
1154 fp
= open(filename
, oflags
, 0600);
1156 fprintf (stderr
, "Could not open %s\n", filename
);
1159 info
= __open_ctree_fd(fp
, filename
, sb_bytenr
, root_tree_bytenr
,
1165 struct btrfs_root
*open_ctree(const char *filename
, u64 sb_bytenr
,
1166 enum btrfs_open_ctree_flags flags
)
1168 struct btrfs_fs_info
*info
;
1170 info
= open_ctree_fs_info(filename
, sb_bytenr
, 0, flags
);
1173 return info
->fs_root
;
1176 struct btrfs_root
*open_ctree_fd(int fp
, const char *path
, u64 sb_bytenr
,
1177 enum btrfs_open_ctree_flags flags
)
1179 struct btrfs_fs_info
*info
;
1180 info
= __open_ctree_fd(fp
, path
, sb_bytenr
, 0, flags
);
1183 return info
->fs_root
;
1186 int btrfs_read_dev_super(int fd
, struct btrfs_super_block
*sb
, u64 sb_bytenr
)
1188 u8 fsid
[BTRFS_FSID_SIZE
];
1189 int fsid_is_initialized
= 0;
1190 struct btrfs_super_block buf
;
1196 if (sb_bytenr
!= BTRFS_SUPER_INFO_OFFSET
) {
1197 ret
= pread64(fd
, &buf
, sizeof(buf
), sb_bytenr
);
1198 if (ret
< sizeof(buf
))
1201 if (btrfs_super_bytenr(&buf
) != sb_bytenr
||
1202 btrfs_super_magic(&buf
) != BTRFS_MAGIC
)
1205 memcpy(sb
, &buf
, sizeof(*sb
));
1210 * we would like to check all the supers, but that would make
1211 * a btrfs mount succeed after a mkfs from a different FS.
1212 * So, we need to add a special mount option to scan for
1213 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1216 for (i
= 0; i
< 1; i
++) {
1217 bytenr
= btrfs_sb_offset(i
);
1218 ret
= pread64(fd
, &buf
, sizeof(buf
), bytenr
);
1219 if (ret
< sizeof(buf
))
1222 if (btrfs_super_bytenr(&buf
) != bytenr
)
1224 /* if magic is NULL, the device was removed */
1225 if (btrfs_super_magic(&buf
) == 0 && i
== 0)
1227 if (btrfs_super_magic(&buf
) != BTRFS_MAGIC
)
1230 if (!fsid_is_initialized
) {
1231 memcpy(fsid
, buf
.fsid
, sizeof(fsid
));
1232 fsid_is_initialized
= 1;
1233 } else if (memcmp(fsid
, buf
.fsid
, sizeof(fsid
))) {
1235 * the superblocks (the original one and
1236 * its backups) contain data of different
1237 * filesystems -> the super cannot be trusted
1242 if (btrfs_super_generation(&buf
) > transid
) {
1243 memcpy(sb
, &buf
, sizeof(*sb
));
1244 transid
= btrfs_super_generation(&buf
);
1248 return transid
> 0 ? 0 : -1;
1251 static int write_dev_supers(struct btrfs_root
*root
,
1252 struct btrfs_super_block
*sb
,
1253 struct btrfs_device
*device
)
1259 if (root
->fs_info
->super_bytenr
!= BTRFS_SUPER_INFO_OFFSET
) {
1260 btrfs_set_super_bytenr(sb
, root
->fs_info
->super_bytenr
);
1262 crc
= btrfs_csum_data(NULL
, (char *)sb
+ BTRFS_CSUM_SIZE
, crc
,
1263 BTRFS_SUPER_INFO_SIZE
- BTRFS_CSUM_SIZE
);
1264 btrfs_csum_final(crc
, (char *)&sb
->csum
[0]);
1267 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1268 * zero filled, we can use it directly
1270 ret
= pwrite64(device
->fd
, root
->fs_info
->super_copy
,
1271 BTRFS_SUPER_INFO_SIZE
,
1272 root
->fs_info
->super_bytenr
);
1273 BUG_ON(ret
!= BTRFS_SUPER_INFO_SIZE
);
1277 for (i
= 0; i
< BTRFS_SUPER_MIRROR_MAX
; i
++) {
1278 bytenr
= btrfs_sb_offset(i
);
1279 if (bytenr
+ BTRFS_SUPER_INFO_SIZE
> device
->total_bytes
)
1282 btrfs_set_super_bytenr(sb
, bytenr
);
1285 crc
= btrfs_csum_data(NULL
, (char *)sb
+ BTRFS_CSUM_SIZE
, crc
,
1286 BTRFS_SUPER_INFO_SIZE
- BTRFS_CSUM_SIZE
);
1287 btrfs_csum_final(crc
, (char *)&sb
->csum
[0]);
1290 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1291 * zero filled, we can use it directly
1293 ret
= pwrite64(device
->fd
, root
->fs_info
->super_copy
,
1294 BTRFS_SUPER_INFO_SIZE
, bytenr
);
1295 BUG_ON(ret
!= BTRFS_SUPER_INFO_SIZE
);
1301 int write_all_supers(struct btrfs_root
*root
)
1303 struct list_head
*cur
;
1304 struct list_head
*head
= &root
->fs_info
->fs_devices
->devices
;
1305 struct btrfs_device
*dev
;
1306 struct btrfs_super_block
*sb
;
1307 struct btrfs_dev_item
*dev_item
;
1311 sb
= root
->fs_info
->super_copy
;
1312 dev_item
= &sb
->dev_item
;
1313 list_for_each(cur
, head
) {
1314 dev
= list_entry(cur
, struct btrfs_device
, dev_list
);
1315 if (!dev
->writeable
)
1318 btrfs_set_stack_device_generation(dev_item
, 0);
1319 btrfs_set_stack_device_type(dev_item
, dev
->type
);
1320 btrfs_set_stack_device_id(dev_item
, dev
->devid
);
1321 btrfs_set_stack_device_total_bytes(dev_item
, dev
->total_bytes
);
1322 btrfs_set_stack_device_bytes_used(dev_item
, dev
->bytes_used
);
1323 btrfs_set_stack_device_io_align(dev_item
, dev
->io_align
);
1324 btrfs_set_stack_device_io_width(dev_item
, dev
->io_width
);
1325 btrfs_set_stack_device_sector_size(dev_item
, dev
->sector_size
);
1326 memcpy(dev_item
->uuid
, dev
->uuid
, BTRFS_UUID_SIZE
);
1327 memcpy(dev_item
->fsid
, dev
->fs_devices
->fsid
, BTRFS_UUID_SIZE
);
1329 flags
= btrfs_super_flags(sb
);
1330 btrfs_set_super_flags(sb
, flags
| BTRFS_HEADER_FLAG_WRITTEN
);
1332 ret
= write_dev_supers(root
, sb
, dev
);
1338 int write_ctree_super(struct btrfs_trans_handle
*trans
,
1339 struct btrfs_root
*root
)
1342 struct btrfs_root
*tree_root
= root
->fs_info
->tree_root
;
1343 struct btrfs_root
*chunk_root
= root
->fs_info
->chunk_root
;
1345 if (root
->fs_info
->readonly
)
1348 btrfs_set_super_generation(root
->fs_info
->super_copy
,
1350 btrfs_set_super_root(root
->fs_info
->super_copy
,
1351 tree_root
->node
->start
);
1352 btrfs_set_super_root_level(root
->fs_info
->super_copy
,
1353 btrfs_header_level(tree_root
->node
));
1354 btrfs_set_super_chunk_root(root
->fs_info
->super_copy
,
1355 chunk_root
->node
->start
);
1356 btrfs_set_super_chunk_root_level(root
->fs_info
->super_copy
,
1357 btrfs_header_level(chunk_root
->node
));
1358 btrfs_set_super_chunk_root_generation(root
->fs_info
->super_copy
,
1359 btrfs_header_generation(chunk_root
->node
));
1361 ret
= write_all_supers(root
);
1363 fprintf(stderr
, "failed to write new super block err %d\n", ret
);
1367 int close_ctree(struct btrfs_root
*root
)
1370 struct btrfs_trans_handle
*trans
;
1371 struct btrfs_fs_info
*fs_info
= root
->fs_info
;
1373 if (fs_info
->last_trans_committed
!=
1374 fs_info
->generation
) {
1375 trans
= btrfs_start_transaction(root
, 1);
1376 btrfs_commit_transaction(trans
, root
);
1377 trans
= btrfs_start_transaction(root
, 1);
1378 ret
= commit_tree_roots(trans
, fs_info
);
1380 ret
= __commit_transaction(trans
, root
);
1382 write_ctree_super(trans
, root
);
1383 btrfs_free_transaction(root
, trans
);
1385 btrfs_free_block_groups(fs_info
);
1387 free_fs_roots_tree(&fs_info
->fs_root_tree
);
1389 btrfs_release_all_roots(fs_info
);
1390 btrfs_close_devices(fs_info
->fs_devices
);
1391 btrfs_cleanup_all_caches(fs_info
);
1392 btrfs_free_fs_info(fs_info
);
1396 int clean_tree_block(struct btrfs_trans_handle
*trans
, struct btrfs_root
*root
,
1397 struct extent_buffer
*eb
)
1399 return clear_extent_buffer_dirty(eb
);
1402 int wait_on_tree_block_writeback(struct btrfs_root
*root
,
1403 struct extent_buffer
*eb
)
1408 void btrfs_mark_buffer_dirty(struct extent_buffer
*eb
)
1410 set_extent_buffer_dirty(eb
);
1413 int btrfs_buffer_uptodate(struct extent_buffer
*buf
, u64 parent_transid
)
1417 ret
= extent_buffer_uptodate(buf
);
1421 ret
= verify_parent_transid(buf
->tree
, buf
, parent_transid
, 1);
1425 int btrfs_set_buffer_uptodate(struct extent_buffer
*eb
)
1427 return set_extent_buffer_uptodate(eb
);