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.
21 #include <sys/types.h>
25 #include <uuid/uuid.h>
26 #include "kerncompat.h"
27 #include "radix-tree.h"
31 #include "transaction.h"
34 #include "print-tree.h"
35 #include "rbtree-utils.h"
37 /* specified errno for check_tree_block */
38 #define BTRFS_BAD_BYTENR (-1)
39 #define BTRFS_BAD_FSID (-2)
41 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
44 struct btrfs_fs_devices *fs_devices;
45 int ret = BTRFS_BAD_FSID;
47 if (buf->start != btrfs_header_bytenr(buf))
48 return BTRFS_BAD_BYTENR;
50 fs_devices = root->fs_info->fs_devices;
52 if (!memcmp_extent_buffer(buf, fs_devices->fsid,
58 fs_devices = fs_devices->seed;
63 static void print_tree_block_err(struct btrfs_root *root,
64 struct extent_buffer *eb,
67 char fs_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
68 char found_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
69 u8 buf[BTRFS_UUID_SIZE];
73 read_extent_buffer(eb, buf, btrfs_header_fsid(),
75 uuid_unparse(buf, found_uuid);
76 uuid_unparse(root->fs_info->fsid, fs_uuid);
77 fprintf(stderr, "fsid mismatch, want=%s, have=%s\n",
80 case BTRFS_BAD_BYTENR:
81 fprintf(stderr, "bytenr mismatch, want=%llu, have=%llu\n",
82 eb->start, btrfs_header_bytenr(eb));
87 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
89 return crc32c(seed, data, len);
92 void btrfs_csum_final(u32 crc, char *result)
94 *(__le32 *)result = ~cpu_to_le32(crc);
97 static int __csum_tree_block_size(struct extent_buffer *buf, u16 csum_size,
98 int verify, int silent)
104 result = malloc(csum_size * sizeof(char));
108 len = buf->len - BTRFS_CSUM_SIZE;
109 crc = crc32c(crc, buf->data + BTRFS_CSUM_SIZE, len);
110 btrfs_csum_final(crc, result);
113 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
115 printk("checksum verify failed on %llu found %08X wanted %08X\n",
116 (unsigned long long)buf->start,
118 *((u32*)(char *)buf->data));
123 write_extent_buffer(buf, result, 0, csum_size);
129 int csum_tree_block_size(struct extent_buffer *buf, u16 csum_size, int verify)
131 return __csum_tree_block_size(buf, csum_size, verify, 0);
134 int verify_tree_block_csum_silent(struct extent_buffer *buf, u16 csum_size)
136 return __csum_tree_block_size(buf, csum_size, 1, 1);
139 int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
143 btrfs_super_csum_size(root->fs_info->super_copy);
144 return csum_tree_block_size(buf, csum_size, verify);
147 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
148 u64 bytenr, u32 blocksize)
150 return find_extent_buffer(&root->fs_info->extent_cache,
154 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
155 u64 bytenr, u32 blocksize)
157 return alloc_extent_buffer(&root->fs_info->extent_cache, bytenr,
161 void readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
164 struct extent_buffer *eb;
166 struct btrfs_multi_bio *multi = NULL;
167 struct btrfs_device *device;
169 eb = btrfs_find_tree_block(root, bytenr, blocksize);
170 if (!(eb && btrfs_buffer_uptodate(eb, parent_transid)) &&
171 !btrfs_map_block(&root->fs_info->mapping_tree, READ,
172 bytenr, &length, &multi, 0, NULL)) {
173 device = multi->stripes[0].dev;
175 blocksize = min(blocksize, (u32)(64 * 1024));
176 readahead(device->fd, multi->stripes[0].physical, blocksize);
179 free_extent_buffer(eb);
183 static int verify_parent_transid(struct extent_io_tree *io_tree,
184 struct extent_buffer *eb, u64 parent_transid,
189 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
192 if (extent_buffer_uptodate(eb) &&
193 btrfs_header_generation(eb) == parent_transid) {
197 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
198 (unsigned long long)eb->start,
199 (unsigned long long)parent_transid,
200 (unsigned long long)btrfs_header_generation(eb));
202 eb->flags |= EXTENT_BAD_TRANSID;
203 printk("Ignoring transid failure\n");
209 clear_extent_buffer_uptodate(io_tree, eb);
215 int read_whole_eb(struct btrfs_fs_info *info, struct extent_buffer *eb, int mirror)
217 unsigned long offset = 0;
218 struct btrfs_multi_bio *multi = NULL;
219 struct btrfs_device *device;
222 unsigned long bytes_left = eb->len;
225 read_len = bytes_left;
228 if (!info->on_restoring &&
229 eb->start != BTRFS_SUPER_INFO_OFFSET) {
230 ret = btrfs_map_block(&info->mapping_tree, READ,
231 eb->start + offset, &read_len, &multi,
234 printk("Couldn't map the block %Lu\n", eb->start + offset);
238 device = multi->stripes[0].dev;
240 if (device->fd == 0) {
247 eb->dev_bytenr = multi->stripes[0].physical;
251 /* special case for restore metadump */
252 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
253 if (device->devid == 1)
258 eb->dev_bytenr = eb->start;
262 if (read_len > bytes_left)
263 read_len = bytes_left;
265 ret = read_extent_from_disk(eb, offset, read_len);
269 bytes_left -= read_len;
274 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
275 u32 blocksize, u64 parent_transid)
278 struct extent_buffer *eb;
279 u64 best_transid = 0;
285 eb = btrfs_find_create_tree_block(root, bytenr, blocksize);
287 return ERR_PTR(-ENOMEM);
289 if (btrfs_buffer_uptodate(eb, parent_transid))
293 ret = read_whole_eb(root->fs_info, eb, mirror_num);
294 if (ret == 0 && check_tree_block(root, eb) == 0 &&
295 csum_tree_block(root, eb, 1) == 0 &&
296 verify_parent_transid(eb->tree, eb, parent_transid, ignore)
298 if (eb->flags & EXTENT_BAD_TRANSID &&
299 list_empty(&eb->recow)) {
300 list_add_tail(&eb->recow,
301 &root->fs_info->recow_ebs);
304 btrfs_set_buffer_uptodate(eb);
308 if (check_tree_block(root, eb))
309 print_tree_block_err(root, eb,
310 check_tree_block(root, eb));
312 printk("Csum didn't match\n");
316 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
318 if (num_copies == 1) {
322 if (btrfs_header_generation(eb) > best_transid && mirror_num) {
323 best_transid = btrfs_header_generation(eb);
324 good_mirror = mirror_num;
327 if (mirror_num > num_copies) {
328 mirror_num = good_mirror;
333 free_extent_buffer(eb);
337 int write_and_map_eb(struct btrfs_trans_handle *trans,
338 struct btrfs_root *root,
339 struct extent_buffer *eb)
344 u64 *raid_map = NULL;
345 struct btrfs_multi_bio *multi = NULL;
349 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
350 eb->start, &length, &multi, 0, &raid_map);
353 ret = write_raid56_with_parity(root->fs_info, eb, multi,
356 } else while (dev_nr < multi->num_stripes) {
358 eb->fd = multi->stripes[dev_nr].dev->fd;
359 eb->dev_bytenr = multi->stripes[dev_nr].physical;
360 multi->stripes[dev_nr].dev->total_ios++;
362 ret = write_extent_to_disk(eb);
369 static int write_tree_block(struct btrfs_trans_handle *trans,
370 struct btrfs_root *root,
371 struct extent_buffer *eb)
373 if (check_tree_block(root, eb)) {
374 print_tree_block_err(root, eb, check_tree_block(root, eb));
378 if (!btrfs_buffer_uptodate(eb, trans->transid))
381 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
382 csum_tree_block(root, eb, 0);
384 return write_and_map_eb(trans, root, eb);
387 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
388 u32 stripesize, struct btrfs_root *root,
389 struct btrfs_fs_info *fs_info, u64 objectid)
392 root->commit_root = NULL;
393 root->sectorsize = sectorsize;
394 root->nodesize = nodesize;
395 root->leafsize = leafsize;
396 root->stripesize = stripesize;
398 root->track_dirty = 0;
400 root->fs_info = fs_info;
401 root->objectid = objectid;
402 root->last_trans = 0;
403 root->highest_inode = 0;
404 root->last_inode_alloc = 0;
406 INIT_LIST_HEAD(&root->dirty_list);
407 INIT_LIST_HEAD(&root->orphan_data_extents);
408 memset(&root->root_key, 0, sizeof(root->root_key));
409 memset(&root->root_item, 0, sizeof(root->root_item));
410 root->root_key.objectid = objectid;
414 static int update_cowonly_root(struct btrfs_trans_handle *trans,
415 struct btrfs_root *root)
419 struct btrfs_root *tree_root = root->fs_info->tree_root;
421 btrfs_write_dirty_block_groups(trans, root);
423 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
424 if (old_root_bytenr == root->node->start)
426 btrfs_set_root_bytenr(&root->root_item,
428 btrfs_set_root_generation(&root->root_item,
430 root->root_item.level = btrfs_header_level(root->node);
431 ret = btrfs_update_root(trans, tree_root,
435 btrfs_write_dirty_block_groups(trans, root);
440 static int commit_tree_roots(struct btrfs_trans_handle *trans,
441 struct btrfs_fs_info *fs_info)
443 struct btrfs_root *root;
444 struct list_head *next;
445 struct extent_buffer *eb;
448 if (fs_info->readonly)
451 eb = fs_info->tree_root->node;
452 extent_buffer_get(eb);
453 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
454 free_extent_buffer(eb);
458 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
459 next = fs_info->dirty_cowonly_roots.next;
461 root = list_entry(next, struct btrfs_root, dirty_list);
462 update_cowonly_root(trans, root);
463 free_extent_buffer(root->commit_root);
464 root->commit_root = NULL;
470 static int __commit_transaction(struct btrfs_trans_handle *trans,
471 struct btrfs_root *root)
475 struct extent_buffer *eb;
476 struct extent_io_tree *tree = &root->fs_info->extent_cache;
480 ret = find_first_extent_bit(tree, 0, &start, &end,
484 while(start <= end) {
485 eb = find_first_extent_buffer(tree, start);
486 BUG_ON(!eb || eb->start != start);
487 ret = write_tree_block(trans, root, eb);
490 clear_extent_buffer_dirty(eb);
491 free_extent_buffer(eb);
497 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
498 struct btrfs_root *root)
500 u64 transid = trans->transid;
502 struct btrfs_fs_info *fs_info = root->fs_info;
504 if (root->commit_root == root->node)
506 if (root == root->fs_info->tree_root)
509 free_extent_buffer(root->commit_root);
510 root->commit_root = NULL;
512 btrfs_set_root_bytenr(&root->root_item, root->node->start);
513 btrfs_set_root_generation(&root->root_item, trans->transid);
514 root->root_item.level = btrfs_header_level(root->node);
515 ret = btrfs_update_root(trans, root->fs_info->tree_root,
516 &root->root_key, &root->root_item);
519 ret = commit_tree_roots(trans, fs_info);
521 ret = __commit_transaction(trans, root);
523 write_ctree_super(trans, root);
524 btrfs_finish_extent_commit(trans, fs_info->extent_root,
525 &fs_info->pinned_extents);
526 btrfs_free_transaction(root, trans);
527 free_extent_buffer(root->commit_root);
528 root->commit_root = NULL;
529 fs_info->running_transaction = NULL;
530 fs_info->last_trans_committed = transid;
534 static int find_and_setup_root(struct btrfs_root *tree_root,
535 struct btrfs_fs_info *fs_info,
536 u64 objectid, struct btrfs_root *root)
542 __setup_root(tree_root->nodesize, tree_root->leafsize,
543 tree_root->sectorsize, tree_root->stripesize,
544 root, fs_info, objectid);
545 ret = btrfs_find_last_root(tree_root, objectid,
546 &root->root_item, &root->root_key);
550 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
551 generation = btrfs_root_generation(&root->root_item);
552 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
553 blocksize, generation);
554 if (!extent_buffer_uptodate(root->node))
560 static int find_and_setup_log_root(struct btrfs_root *tree_root,
561 struct btrfs_fs_info *fs_info,
562 struct btrfs_super_block *disk_super)
565 u64 blocknr = btrfs_super_log_root(disk_super);
566 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
576 blocksize = btrfs_level_size(tree_root,
577 btrfs_super_log_root_level(disk_super));
579 __setup_root(tree_root->nodesize, tree_root->leafsize,
580 tree_root->sectorsize, tree_root->stripesize,
581 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
583 log_root->node = read_tree_block(tree_root, blocknr,
585 btrfs_super_generation(disk_super) + 1);
587 fs_info->log_root_tree = log_root;
589 if (!extent_buffer_uptodate(log_root->node)) {
590 free_extent_buffer(log_root->node);
592 fs_info->log_root_tree = NULL;
599 int btrfs_free_fs_root(struct btrfs_root *root)
602 free_extent_buffer(root->node);
603 if (root->commit_root)
604 free_extent_buffer(root->commit_root);
609 static void __free_fs_root(struct rb_node *node)
611 struct btrfs_root *root;
613 root = container_of(node, struct btrfs_root, rb_node);
614 btrfs_free_fs_root(root);
617 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
619 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
620 struct btrfs_key *location)
622 struct btrfs_root *root;
623 struct btrfs_root *tree_root = fs_info->tree_root;
624 struct btrfs_path *path;
625 struct extent_buffer *l;
630 root = malloc(sizeof(*root));
632 return ERR_PTR(-ENOMEM);
633 memset(root, 0, sizeof(*root));
634 if (location->offset == (u64)-1) {
635 ret = find_and_setup_root(tree_root, fs_info,
636 location->objectid, root);
644 __setup_root(tree_root->nodesize, tree_root->leafsize,
645 tree_root->sectorsize, tree_root->stripesize,
646 root, fs_info, location->objectid);
648 path = btrfs_alloc_path();
650 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
657 read_extent_buffer(l, &root->root_item,
658 btrfs_item_ptr_offset(l, path->slots[0]),
659 sizeof(root->root_item));
660 memcpy(&root->root_key, location, sizeof(*location));
663 btrfs_free_path(path);
668 generation = btrfs_root_generation(&root->root_item);
669 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
670 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
671 blocksize, generation);
672 if (!extent_buffer_uptodate(root->node)) {
674 return ERR_PTR(-EIO);
681 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
684 u64 objectid = *((u64 *)data);
685 struct btrfs_root *root;
687 root = rb_entry(node, struct btrfs_root, rb_node);
688 if (objectid > root->objectid)
690 else if (objectid < root->objectid)
696 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
697 struct rb_node *node2)
699 struct btrfs_root *root;
701 root = rb_entry(node2, struct btrfs_root, rb_node);
702 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
705 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
706 struct btrfs_key *location)
708 struct btrfs_root *root;
709 struct rb_node *node;
711 u64 objectid = location->objectid;
713 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
714 return fs_info->tree_root;
715 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
716 return fs_info->extent_root;
717 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
718 return fs_info->chunk_root;
719 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
720 return fs_info->dev_root;
721 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
722 return fs_info->csum_root;
723 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
724 return fs_info->quota_root;
726 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
727 location->offset != (u64)-1);
729 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
730 btrfs_fs_roots_compare_objectids, NULL);
732 return container_of(node, struct btrfs_root, rb_node);
734 root = btrfs_read_fs_root_no_cache(fs_info, location);
738 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
739 btrfs_fs_roots_compare_roots);
744 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
746 free(fs_info->tree_root);
747 free(fs_info->extent_root);
748 free(fs_info->chunk_root);
749 free(fs_info->dev_root);
750 free(fs_info->csum_root);
751 free(fs_info->quota_root);
752 free(fs_info->super_copy);
753 free(fs_info->log_root_tree);
757 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
759 struct btrfs_fs_info *fs_info;
761 fs_info = malloc(sizeof(struct btrfs_fs_info));
765 memset(fs_info, 0, sizeof(struct btrfs_fs_info));
767 fs_info->tree_root = malloc(sizeof(struct btrfs_root));
768 fs_info->extent_root = malloc(sizeof(struct btrfs_root));
769 fs_info->chunk_root = malloc(sizeof(struct btrfs_root));
770 fs_info->dev_root = malloc(sizeof(struct btrfs_root));
771 fs_info->csum_root = malloc(sizeof(struct btrfs_root));
772 fs_info->quota_root = malloc(sizeof(struct btrfs_root));
773 fs_info->super_copy = malloc(BTRFS_SUPER_INFO_SIZE);
775 if (!fs_info->tree_root || !fs_info->extent_root ||
776 !fs_info->chunk_root || !fs_info->dev_root ||
777 !fs_info->csum_root || !fs_info->quota_root ||
778 !fs_info->super_copy)
781 memset(fs_info->super_copy, 0, BTRFS_SUPER_INFO_SIZE);
782 memset(fs_info->tree_root, 0, sizeof(struct btrfs_root));
783 memset(fs_info->extent_root, 0, sizeof(struct btrfs_root));
784 memset(fs_info->chunk_root, 0, sizeof(struct btrfs_root));
785 memset(fs_info->dev_root, 0, sizeof(struct btrfs_root));
786 memset(fs_info->csum_root, 0, sizeof(struct btrfs_root));
787 memset(fs_info->quota_root, 0, sizeof(struct btrfs_root));
789 extent_io_tree_init(&fs_info->extent_cache);
790 extent_io_tree_init(&fs_info->free_space_cache);
791 extent_io_tree_init(&fs_info->block_group_cache);
792 extent_io_tree_init(&fs_info->pinned_extents);
793 extent_io_tree_init(&fs_info->pending_del);
794 extent_io_tree_init(&fs_info->extent_ins);
795 fs_info->excluded_extents = NULL;
797 fs_info->fs_root_tree = RB_ROOT;
798 cache_tree_init(&fs_info->mapping_tree.cache_tree);
800 mutex_init(&fs_info->fs_mutex);
801 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
802 INIT_LIST_HEAD(&fs_info->space_info);
803 INIT_LIST_HEAD(&fs_info->recow_ebs);
806 fs_info->readonly = 1;
808 fs_info->super_bytenr = sb_bytenr;
809 fs_info->data_alloc_profile = (u64)-1;
810 fs_info->metadata_alloc_profile = (u64)-1;
811 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
814 btrfs_free_fs_info(fs_info);
818 int btrfs_check_fs_compatibility(struct btrfs_super_block *sb, int writable)
822 features = btrfs_super_incompat_flags(sb) &
823 ~BTRFS_FEATURE_INCOMPAT_SUPP;
825 printk("couldn't open because of unsupported "
826 "option features (%Lx).\n",
827 (unsigned long long)features);
831 features = btrfs_super_incompat_flags(sb);
832 if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
833 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
834 btrfs_set_super_incompat_flags(sb, features);
837 features = btrfs_super_compat_ro_flags(sb) &
838 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
839 if (writable && features) {
840 printk("couldn't open RDWR because of unsupported "
841 "option features (%Lx).\n",
842 (unsigned long long)features);
848 static int find_best_backup_root(struct btrfs_super_block *super)
850 struct btrfs_root_backup *backup;
851 u64 orig_gen = btrfs_super_generation(super);
856 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
857 backup = super->super_roots + i;
858 if (btrfs_backup_tree_root_gen(backup) != orig_gen &&
859 btrfs_backup_tree_root_gen(backup) > gen) {
861 gen = btrfs_backup_tree_root_gen(backup);
867 static int setup_root_or_create_block(struct btrfs_fs_info *fs_info,
868 enum btrfs_open_ctree_flags flags,
869 struct btrfs_root *info_root,
870 u64 objectid, char *str)
872 struct btrfs_super_block *sb = fs_info->super_copy;
873 struct btrfs_root *root = fs_info->tree_root;
874 u32 leafsize = btrfs_super_leafsize(sb);
877 ret = find_and_setup_root(root, fs_info, objectid, info_root);
879 printk("Couldn't setup %s tree\n", str);
880 if (!(flags & OPEN_CTREE_PARTIAL))
883 * Need a blank node here just so we don't screw up in the
884 * million of places that assume a root has a valid ->node
887 btrfs_find_create_tree_block(info_root, 0, leafsize);
888 if (!info_root->node)
890 clear_extent_buffer_uptodate(NULL, info_root->node);
896 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
897 enum btrfs_open_ctree_flags flags)
899 struct btrfs_super_block *sb = fs_info->super_copy;
900 struct btrfs_root *root;
901 struct btrfs_key key;
910 nodesize = btrfs_super_nodesize(sb);
911 leafsize = btrfs_super_leafsize(sb);
912 sectorsize = btrfs_super_sectorsize(sb);
913 stripesize = btrfs_super_stripesize(sb);
915 root = fs_info->tree_root;
916 __setup_root(nodesize, leafsize, sectorsize, stripesize,
917 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
918 blocksize = btrfs_level_size(root, btrfs_super_root_level(sb));
919 generation = btrfs_super_generation(sb);
921 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
922 root_tree_bytenr = btrfs_super_root(sb);
923 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
924 struct btrfs_root_backup *backup;
925 int index = find_best_backup_root(sb);
926 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
927 fprintf(stderr, "Invalid backup root number\n");
930 backup = fs_info->super_copy->super_roots + index;
931 root_tree_bytenr = btrfs_backup_tree_root(backup);
932 generation = btrfs_backup_tree_root_gen(backup);
935 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
937 if (!extent_buffer_uptodate(root->node)) {
938 fprintf(stderr, "Couldn't read tree root\n");
942 ret = setup_root_or_create_block(fs_info, flags, fs_info->extent_root,
943 BTRFS_EXTENT_TREE_OBJECTID, "extent");
946 fs_info->extent_root->track_dirty = 1;
948 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
951 printk("Couldn't setup device tree\n");
954 fs_info->dev_root->track_dirty = 1;
956 ret = setup_root_or_create_block(fs_info, flags, fs_info->csum_root,
957 BTRFS_CSUM_TREE_OBJECTID, "csum");
960 fs_info->csum_root->track_dirty = 1;
962 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
963 fs_info->quota_root);
965 fs_info->quota_enabled = 1;
967 ret = find_and_setup_log_root(root, fs_info, sb);
969 printk("Couldn't setup log root tree\n");
970 if (!(flags & OPEN_CTREE_PARTIAL))
974 fs_info->generation = generation;
975 fs_info->last_trans_committed = generation;
976 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
977 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
978 btrfs_read_block_groups(fs_info->tree_root);
980 key.objectid = BTRFS_FS_TREE_OBJECTID;
981 key.type = BTRFS_ROOT_ITEM_KEY;
982 key.offset = (u64)-1;
983 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
985 if (IS_ERR(fs_info->fs_root))
990 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
992 if (fs_info->quota_root)
993 free_extent_buffer(fs_info->quota_root->node);
994 if (fs_info->csum_root)
995 free_extent_buffer(fs_info->csum_root->node);
996 if (fs_info->dev_root)
997 free_extent_buffer(fs_info->dev_root->node);
998 if (fs_info->extent_root)
999 free_extent_buffer(fs_info->extent_root->node);
1000 if (fs_info->tree_root)
1001 free_extent_buffer(fs_info->tree_root->node);
1002 if (fs_info->log_root_tree)
1003 free_extent_buffer(fs_info->log_root_tree->node);
1004 if (fs_info->chunk_root)
1005 free_extent_buffer(fs_info->chunk_root->node);
1008 static void free_map_lookup(struct cache_extent *ce)
1010 struct map_lookup *map;
1012 map = container_of(ce, struct map_lookup, ce);
1016 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
1018 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
1020 while (!list_empty(&fs_info->recow_ebs)) {
1021 struct extent_buffer *eb;
1022 eb = list_first_entry(&fs_info->recow_ebs,
1023 struct extent_buffer, recow);
1024 list_del_init(&eb->recow);
1025 free_extent_buffer(eb);
1027 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
1028 extent_io_tree_cleanup(&fs_info->extent_cache);
1029 extent_io_tree_cleanup(&fs_info->free_space_cache);
1030 extent_io_tree_cleanup(&fs_info->block_group_cache);
1031 extent_io_tree_cleanup(&fs_info->pinned_extents);
1032 extent_io_tree_cleanup(&fs_info->pending_del);
1033 extent_io_tree_cleanup(&fs_info->extent_ins);
1036 int btrfs_scan_fs_devices(int fd, const char *path,
1037 struct btrfs_fs_devices **fs_devices,
1038 u64 sb_bytenr, int super_recover,
1046 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1048 seek_ret = lseek(fd, 0, SEEK_END);
1052 dev_size = seek_ret;
1053 lseek(fd, 0, SEEK_SET);
1054 if (sb_bytenr > dev_size) {
1055 fprintf(stderr, "Superblock bytenr is larger than device size\n");
1059 ret = btrfs_scan_one_device(fd, path, fs_devices,
1060 &total_devs, sb_bytenr, super_recover);
1062 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1066 if (!skip_devices && total_devs != 1) {
1067 ret = btrfs_scan_lblkid();
1074 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info)
1076 struct btrfs_super_block *sb = fs_info->super_copy;
1085 nodesize = btrfs_super_nodesize(sb);
1086 leafsize = btrfs_super_leafsize(sb);
1087 sectorsize = btrfs_super_sectorsize(sb);
1088 stripesize = btrfs_super_stripesize(sb);
1090 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1091 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1093 ret = btrfs_read_sys_array(fs_info->chunk_root);
1097 blocksize = btrfs_level_size(fs_info->chunk_root,
1098 btrfs_super_chunk_root_level(sb));
1099 generation = btrfs_super_chunk_root_generation(sb);
1101 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1102 btrfs_super_chunk_root(sb),
1103 blocksize, generation);
1104 if (!extent_buffer_uptodate(fs_info->chunk_root->node)) {
1105 fprintf(stderr, "Couldn't read chunk root\n");
1109 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1110 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1112 fprintf(stderr, "Couldn't read chunk tree\n");
1119 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1121 u64 root_tree_bytenr,
1122 enum btrfs_open_ctree_flags flags)
1124 struct btrfs_fs_info *fs_info;
1125 struct btrfs_super_block *disk_super;
1126 struct btrfs_fs_devices *fs_devices = NULL;
1127 struct extent_buffer *eb;
1132 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1134 /* try to drop all the caches */
1135 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1136 fprintf(stderr, "Warning, could not drop caches\n");
1138 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1140 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1143 if (flags & OPEN_CTREE_RESTORE)
1144 fs_info->on_restoring = 1;
1146 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr,
1147 (flags & OPEN_CTREE_RECOVER_SUPER),
1148 (flags & OPEN_CTREE_NO_DEVICES));
1152 fs_info->fs_devices = fs_devices;
1153 if (flags & OPEN_CTREE_WRITES)
1158 if (flags & OPEN_CTREE_EXCLUSIVE)
1161 ret = btrfs_open_devices(fs_devices, oflags);
1165 disk_super = fs_info->super_copy;
1166 if (!(flags & OPEN_CTREE_RECOVER_SUPER))
1167 ret = btrfs_read_dev_super(fs_devices->latest_bdev,
1168 disk_super, sb_bytenr, 1);
1170 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr, 0);
1172 printk("No valid btrfs found\n");
1176 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1178 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1179 flags & OPEN_CTREE_WRITES);
1183 ret = btrfs_setup_chunk_tree_and_device_map(fs_info);
1187 eb = fs_info->chunk_root->node;
1188 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1189 btrfs_header_chunk_tree_uuid(eb),
1192 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1199 btrfs_release_all_roots(fs_info);
1200 btrfs_cleanup_all_caches(fs_info);
1202 btrfs_close_devices(fs_devices);
1204 btrfs_free_fs_info(fs_info);
1208 struct btrfs_fs_info *open_ctree_fs_info(const char *filename,
1209 u64 sb_bytenr, u64 root_tree_bytenr,
1210 enum btrfs_open_ctree_flags flags)
1213 struct btrfs_fs_info *info;
1214 int oflags = O_CREAT | O_RDWR;
1216 if (!(flags & OPEN_CTREE_WRITES))
1219 fp = open(filename, oflags, 0600);
1221 fprintf (stderr, "Could not open %s\n", filename);
1224 info = __open_ctree_fd(fp, filename, sb_bytenr, root_tree_bytenr,
1230 struct btrfs_root *open_ctree(const char *filename, u64 sb_bytenr,
1231 enum btrfs_open_ctree_flags flags)
1233 struct btrfs_fs_info *info;
1235 info = open_ctree_fs_info(filename, sb_bytenr, 0, flags);
1238 return info->fs_root;
1241 struct btrfs_root *open_ctree_fd(int fp, const char *path, u64 sb_bytenr,
1242 enum btrfs_open_ctree_flags flags)
1244 struct btrfs_fs_info *info;
1245 info = __open_ctree_fd(fp, path, sb_bytenr, 0, flags);
1248 return info->fs_root;
1251 int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
1254 u8 fsid[BTRFS_FSID_SIZE];
1255 int fsid_is_initialized = 0;
1256 struct btrfs_super_block buf;
1259 int max_super = super_recover ? BTRFS_SUPER_MIRROR_MAX : 1;
1263 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1264 ret = pread64(fd, &buf, sizeof(buf), sb_bytenr);
1265 if (ret < sizeof(buf))
1268 if (btrfs_super_bytenr(&buf) != sb_bytenr ||
1269 btrfs_super_magic(&buf) != BTRFS_MAGIC)
1272 memcpy(sb, &buf, sizeof(*sb));
1277 * we would like to check all the supers, but that would make
1278 * a btrfs mount succeed after a mkfs from a different FS.
1279 * So, we need to add a special mount option to scan for
1280 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1283 for (i = 0; i < max_super; i++) {
1284 bytenr = btrfs_sb_offset(i);
1285 ret = pread64(fd, &buf, sizeof(buf), bytenr);
1286 if (ret < sizeof(buf))
1289 if (btrfs_super_bytenr(&buf) != bytenr )
1291 /* if magic is NULL, the device was removed */
1292 if (btrfs_super_magic(&buf) == 0 && i == 0)
1294 if (btrfs_super_magic(&buf) != BTRFS_MAGIC)
1297 if (!fsid_is_initialized) {
1298 memcpy(fsid, buf.fsid, sizeof(fsid));
1299 fsid_is_initialized = 1;
1300 } else if (memcmp(fsid, buf.fsid, sizeof(fsid))) {
1302 * the superblocks (the original one and
1303 * its backups) contain data of different
1304 * filesystems -> the super cannot be trusted
1309 if (btrfs_super_generation(&buf) > transid) {
1310 memcpy(sb, &buf, sizeof(*sb));
1311 transid = btrfs_super_generation(&buf);
1315 return transid > 0 ? 0 : -1;
1318 static int write_dev_supers(struct btrfs_root *root,
1319 struct btrfs_super_block *sb,
1320 struct btrfs_device *device)
1326 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1327 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1329 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1330 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1331 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1334 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1335 * zero filled, we can use it directly
1337 ret = pwrite64(device->fd, root->fs_info->super_copy,
1338 BTRFS_SUPER_INFO_SIZE,
1339 root->fs_info->super_bytenr);
1340 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1344 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1345 bytenr = btrfs_sb_offset(i);
1346 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1349 btrfs_set_super_bytenr(sb, bytenr);
1352 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1353 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1354 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1357 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1358 * zero filled, we can use it directly
1360 ret = pwrite64(device->fd, root->fs_info->super_copy,
1361 BTRFS_SUPER_INFO_SIZE, bytenr);
1362 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1368 int write_all_supers(struct btrfs_root *root)
1370 struct list_head *cur;
1371 struct list_head *head = &root->fs_info->fs_devices->devices;
1372 struct btrfs_device *dev;
1373 struct btrfs_super_block *sb;
1374 struct btrfs_dev_item *dev_item;
1378 sb = root->fs_info->super_copy;
1379 dev_item = &sb->dev_item;
1380 list_for_each(cur, head) {
1381 dev = list_entry(cur, struct btrfs_device, dev_list);
1382 if (!dev->writeable)
1385 btrfs_set_stack_device_generation(dev_item, 0);
1386 btrfs_set_stack_device_type(dev_item, dev->type);
1387 btrfs_set_stack_device_id(dev_item, dev->devid);
1388 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1389 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1390 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1391 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1392 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1393 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1394 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1396 flags = btrfs_super_flags(sb);
1397 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1399 ret = write_dev_supers(root, sb, dev);
1405 int write_ctree_super(struct btrfs_trans_handle *trans,
1406 struct btrfs_root *root)
1409 struct btrfs_root *tree_root = root->fs_info->tree_root;
1410 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1412 if (root->fs_info->readonly)
1415 btrfs_set_super_generation(root->fs_info->super_copy,
1417 btrfs_set_super_root(root->fs_info->super_copy,
1418 tree_root->node->start);
1419 btrfs_set_super_root_level(root->fs_info->super_copy,
1420 btrfs_header_level(tree_root->node));
1421 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1422 chunk_root->node->start);
1423 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1424 btrfs_header_level(chunk_root->node));
1425 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1426 btrfs_header_generation(chunk_root->node));
1428 ret = write_all_supers(root);
1430 fprintf(stderr, "failed to write new super block err %d\n", ret);
1434 int close_ctree(struct btrfs_root *root)
1437 struct btrfs_trans_handle *trans;
1438 struct btrfs_fs_info *fs_info = root->fs_info;
1440 if (fs_info->last_trans_committed !=
1441 fs_info->generation) {
1442 trans = btrfs_start_transaction(root, 1);
1443 btrfs_commit_transaction(trans, root);
1444 trans = btrfs_start_transaction(root, 1);
1445 ret = commit_tree_roots(trans, fs_info);
1447 ret = __commit_transaction(trans, root);
1449 write_ctree_super(trans, root);
1450 btrfs_free_transaction(root, trans);
1452 btrfs_free_block_groups(fs_info);
1454 free_fs_roots_tree(&fs_info->fs_root_tree);
1456 btrfs_release_all_roots(fs_info);
1457 btrfs_close_devices(fs_info->fs_devices);
1458 btrfs_cleanup_all_caches(fs_info);
1459 btrfs_free_fs_info(fs_info);
1463 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1464 struct extent_buffer *eb)
1466 return clear_extent_buffer_dirty(eb);
1469 int wait_on_tree_block_writeback(struct btrfs_root *root,
1470 struct extent_buffer *eb)
1475 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1477 set_extent_buffer_dirty(eb);
1480 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1484 ret = extent_buffer_uptodate(buf);
1488 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1492 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1494 return set_extent_buffer_uptodate(eb);