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 "kerncompat.h"
26 #include "radix-tree.h"
30 #include "transaction.h"
33 #include "print-tree.h"
34 #include "rbtree-utils.h"
36 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
39 struct btrfs_fs_devices *fs_devices;
42 if (buf->start != btrfs_header_bytenr(buf)) {
43 printk("Check tree block failed, want=%Lu, have=%Lu\n",
44 buf->start, btrfs_header_bytenr(buf));
48 fs_devices = root->fs_info->fs_devices;
50 if (!memcmp_extent_buffer(buf, fs_devices->fsid,
56 fs_devices = fs_devices->seed;
61 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
63 return crc32c(seed, data, len);
66 void btrfs_csum_final(u32 crc, char *result)
68 *(__le32 *)result = ~cpu_to_le32(crc);
71 static int __csum_tree_block_size(struct extent_buffer *buf, u16 csum_size,
72 int verify, int silent)
78 result = malloc(csum_size * sizeof(char));
82 len = buf->len - BTRFS_CSUM_SIZE;
83 crc = crc32c(crc, buf->data + BTRFS_CSUM_SIZE, len);
84 btrfs_csum_final(crc, result);
87 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
89 printk("checksum verify failed on %llu found %08X wanted %08X\n",
90 (unsigned long long)buf->start,
92 *((u32*)(char *)buf->data));
97 write_extent_buffer(buf, result, 0, csum_size);
103 int csum_tree_block_size(struct extent_buffer *buf, u16 csum_size, int verify)
105 return __csum_tree_block_size(buf, csum_size, verify, 0);
108 int verify_tree_block_csum_silent(struct extent_buffer *buf, u16 csum_size)
110 return __csum_tree_block_size(buf, csum_size, 1, 1);
113 int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
117 btrfs_super_csum_size(root->fs_info->super_copy);
118 return csum_tree_block_size(buf, csum_size, verify);
121 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
122 u64 bytenr, u32 blocksize)
124 return find_extent_buffer(&root->fs_info->extent_cache,
128 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
129 u64 bytenr, u32 blocksize)
131 return alloc_extent_buffer(&root->fs_info->extent_cache, bytenr,
135 void readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
138 struct extent_buffer *eb;
140 struct btrfs_multi_bio *multi = NULL;
141 struct btrfs_device *device;
143 eb = btrfs_find_tree_block(root, bytenr, blocksize);
144 if (!(eb && btrfs_buffer_uptodate(eb, parent_transid)) &&
145 !btrfs_map_block(&root->fs_info->mapping_tree, READ,
146 bytenr, &length, &multi, 0, NULL)) {
147 device = multi->stripes[0].dev;
149 blocksize = min(blocksize, (u32)(64 * 1024));
150 readahead(device->fd, multi->stripes[0].physical, blocksize);
153 free_extent_buffer(eb);
157 static int verify_parent_transid(struct extent_io_tree *io_tree,
158 struct extent_buffer *eb, u64 parent_transid,
163 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
166 if (extent_buffer_uptodate(eb) &&
167 btrfs_header_generation(eb) == parent_transid) {
171 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
172 (unsigned long long)eb->start,
173 (unsigned long long)parent_transid,
174 (unsigned long long)btrfs_header_generation(eb));
176 eb->flags |= EXTENT_BAD_TRANSID;
177 printk("Ignoring transid failure\n");
183 clear_extent_buffer_uptodate(io_tree, eb);
189 int read_whole_eb(struct btrfs_fs_info *info, struct extent_buffer *eb, int mirror)
191 unsigned long offset = 0;
192 struct btrfs_multi_bio *multi = NULL;
193 struct btrfs_device *device;
196 unsigned long bytes_left = eb->len;
199 read_len = bytes_left;
202 if (!info->on_restoring &&
203 eb->start != BTRFS_SUPER_INFO_OFFSET) {
204 ret = btrfs_map_block(&info->mapping_tree, READ,
205 eb->start + offset, &read_len, &multi,
208 printk("Couldn't map the block %Lu\n", eb->start + offset);
212 device = multi->stripes[0].dev;
214 if (device->fd == 0) {
221 eb->dev_bytenr = multi->stripes[0].physical;
225 /* special case for restore metadump */
226 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
227 if (device->devid == 1)
232 eb->dev_bytenr = eb->start;
236 if (read_len > bytes_left)
237 read_len = bytes_left;
239 ret = read_extent_from_disk(eb, offset, read_len);
243 bytes_left -= read_len;
248 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
249 u32 blocksize, u64 parent_transid)
252 struct extent_buffer *eb;
253 u64 best_transid = 0;
259 eb = btrfs_find_create_tree_block(root, bytenr, blocksize);
263 if (btrfs_buffer_uptodate(eb, parent_transid))
267 ret = read_whole_eb(root->fs_info, eb, mirror_num);
268 if (ret == 0 && check_tree_block(root, eb) == 0 &&
269 csum_tree_block(root, eb, 1) == 0 &&
270 verify_parent_transid(eb->tree, eb, parent_transid, ignore)
272 if (eb->flags & EXTENT_BAD_TRANSID &&
273 list_empty(&eb->recow)) {
274 list_add_tail(&eb->recow,
275 &root->fs_info->recow_ebs);
278 btrfs_set_buffer_uptodate(eb);
282 if (check_tree_block(root, eb))
283 printk("read block failed check_tree_block\n");
285 printk("Csum didn't match\n");
288 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
290 if (num_copies == 1) {
294 if (btrfs_header_generation(eb) > best_transid && mirror_num) {
295 best_transid = btrfs_header_generation(eb);
296 good_mirror = mirror_num;
299 if (mirror_num > num_copies) {
300 mirror_num = good_mirror;
305 free_extent_buffer(eb);
309 int write_and_map_eb(struct btrfs_trans_handle *trans,
310 struct btrfs_root *root,
311 struct extent_buffer *eb)
316 u64 *raid_map = NULL;
317 struct btrfs_multi_bio *multi = NULL;
321 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
322 eb->start, &length, &multi, 0, &raid_map);
325 ret = write_raid56_with_parity(root->fs_info, eb, multi,
328 } else while (dev_nr < multi->num_stripes) {
330 eb->fd = multi->stripes[dev_nr].dev->fd;
331 eb->dev_bytenr = multi->stripes[dev_nr].physical;
332 multi->stripes[dev_nr].dev->total_ios++;
334 ret = write_extent_to_disk(eb);
341 static int write_tree_block(struct btrfs_trans_handle *trans,
342 struct btrfs_root *root,
343 struct extent_buffer *eb)
345 if (check_tree_block(root, eb))
348 if (!btrfs_buffer_uptodate(eb, trans->transid))
351 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
352 csum_tree_block(root, eb, 0);
354 return write_and_map_eb(trans, root, eb);
357 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
358 u32 stripesize, struct btrfs_root *root,
359 struct btrfs_fs_info *fs_info, u64 objectid)
362 root->commit_root = NULL;
363 root->sectorsize = sectorsize;
364 root->nodesize = nodesize;
365 root->leafsize = leafsize;
366 root->stripesize = stripesize;
368 root->track_dirty = 0;
370 root->fs_info = fs_info;
371 root->objectid = objectid;
372 root->last_trans = 0;
373 root->highest_inode = 0;
374 root->last_inode_alloc = 0;
376 INIT_LIST_HEAD(&root->dirty_list);
377 memset(&root->root_key, 0, sizeof(root->root_key));
378 memset(&root->root_item, 0, sizeof(root->root_item));
379 root->root_key.objectid = objectid;
383 static int update_cowonly_root(struct btrfs_trans_handle *trans,
384 struct btrfs_root *root)
388 struct btrfs_root *tree_root = root->fs_info->tree_root;
390 btrfs_write_dirty_block_groups(trans, root);
392 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
393 if (old_root_bytenr == root->node->start)
395 btrfs_set_root_bytenr(&root->root_item,
397 btrfs_set_root_generation(&root->root_item,
399 root->root_item.level = btrfs_header_level(root->node);
400 ret = btrfs_update_root(trans, tree_root,
404 btrfs_write_dirty_block_groups(trans, root);
409 static int commit_tree_roots(struct btrfs_trans_handle *trans,
410 struct btrfs_fs_info *fs_info)
412 struct btrfs_root *root;
413 struct list_head *next;
414 struct extent_buffer *eb;
417 if (fs_info->readonly)
420 eb = fs_info->tree_root->node;
421 extent_buffer_get(eb);
422 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
423 free_extent_buffer(eb);
427 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
428 next = fs_info->dirty_cowonly_roots.next;
430 root = list_entry(next, struct btrfs_root, dirty_list);
431 update_cowonly_root(trans, root);
432 free_extent_buffer(root->commit_root);
433 root->commit_root = NULL;
439 static int __commit_transaction(struct btrfs_trans_handle *trans,
440 struct btrfs_root *root)
444 struct extent_buffer *eb;
445 struct extent_io_tree *tree = &root->fs_info->extent_cache;
449 ret = find_first_extent_bit(tree, 0, &start, &end,
453 while(start <= end) {
454 eb = find_first_extent_buffer(tree, start);
455 BUG_ON(!eb || eb->start != start);
456 ret = write_tree_block(trans, root, eb);
459 clear_extent_buffer_dirty(eb);
460 free_extent_buffer(eb);
466 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
467 struct btrfs_root *root)
469 u64 transid = trans->transid;
471 struct btrfs_fs_info *fs_info = root->fs_info;
473 if (root->commit_root == root->node)
475 if (root == root->fs_info->tree_root)
478 free_extent_buffer(root->commit_root);
479 root->commit_root = NULL;
481 btrfs_set_root_bytenr(&root->root_item, root->node->start);
482 btrfs_set_root_generation(&root->root_item, trans->transid);
483 root->root_item.level = btrfs_header_level(root->node);
484 ret = btrfs_update_root(trans, root->fs_info->tree_root,
485 &root->root_key, &root->root_item);
488 ret = commit_tree_roots(trans, fs_info);
490 ret = __commit_transaction(trans, root);
492 write_ctree_super(trans, root);
493 btrfs_finish_extent_commit(trans, fs_info->extent_root,
494 &fs_info->pinned_extents);
495 btrfs_free_transaction(root, trans);
496 free_extent_buffer(root->commit_root);
497 root->commit_root = NULL;
498 fs_info->running_transaction = NULL;
499 fs_info->last_trans_committed = transid;
503 static int find_and_setup_root(struct btrfs_root *tree_root,
504 struct btrfs_fs_info *fs_info,
505 u64 objectid, struct btrfs_root *root)
511 __setup_root(tree_root->nodesize, tree_root->leafsize,
512 tree_root->sectorsize, tree_root->stripesize,
513 root, fs_info, objectid);
514 ret = btrfs_find_last_root(tree_root, objectid,
515 &root->root_item, &root->root_key);
519 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
520 generation = btrfs_root_generation(&root->root_item);
521 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
522 blocksize, generation);
523 if (!extent_buffer_uptodate(root->node))
529 static int find_and_setup_log_root(struct btrfs_root *tree_root,
530 struct btrfs_fs_info *fs_info,
531 struct btrfs_super_block *disk_super)
534 u64 blocknr = btrfs_super_log_root(disk_super);
535 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
545 blocksize = btrfs_level_size(tree_root,
546 btrfs_super_log_root_level(disk_super));
548 __setup_root(tree_root->nodesize, tree_root->leafsize,
549 tree_root->sectorsize, tree_root->stripesize,
550 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
552 log_root->node = read_tree_block(tree_root, blocknr,
554 btrfs_super_generation(disk_super) + 1);
556 fs_info->log_root_tree = log_root;
558 if (!extent_buffer_uptodate(log_root->node)) {
559 free_extent_buffer(log_root->node);
561 fs_info->log_root_tree = NULL;
568 int btrfs_free_fs_root(struct btrfs_root *root)
571 free_extent_buffer(root->node);
572 if (root->commit_root)
573 free_extent_buffer(root->commit_root);
578 static void __free_fs_root(struct rb_node *node)
580 struct btrfs_root *root;
582 root = container_of(node, struct btrfs_root, rb_node);
583 btrfs_free_fs_root(root);
586 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
588 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
589 struct btrfs_key *location)
591 struct btrfs_root *root;
592 struct btrfs_root *tree_root = fs_info->tree_root;
593 struct btrfs_path *path;
594 struct extent_buffer *l;
599 root = malloc(sizeof(*root));
601 return ERR_PTR(-ENOMEM);
602 memset(root, 0, sizeof(*root));
603 if (location->offset == (u64)-1) {
604 ret = find_and_setup_root(tree_root, fs_info,
605 location->objectid, root);
613 __setup_root(tree_root->nodesize, tree_root->leafsize,
614 tree_root->sectorsize, tree_root->stripesize,
615 root, fs_info, location->objectid);
617 path = btrfs_alloc_path();
619 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
626 read_extent_buffer(l, &root->root_item,
627 btrfs_item_ptr_offset(l, path->slots[0]),
628 sizeof(root->root_item));
629 memcpy(&root->root_key, location, sizeof(*location));
632 btrfs_free_path(path);
637 generation = btrfs_root_generation(&root->root_item);
638 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
639 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
640 blocksize, generation);
643 return ERR_PTR(-EIO);
650 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
653 u64 objectid = *((u64 *)data);
654 struct btrfs_root *root;
656 root = rb_entry(node, struct btrfs_root, rb_node);
657 if (objectid > root->objectid)
659 else if (objectid < root->objectid)
665 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
666 struct rb_node *node2)
668 struct btrfs_root *root;
670 root = rb_entry(node2, struct btrfs_root, rb_node);
671 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
674 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
675 struct btrfs_key *location)
677 struct btrfs_root *root;
678 struct rb_node *node;
680 u64 objectid = location->objectid;
682 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
683 return fs_info->tree_root;
684 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
685 return fs_info->extent_root;
686 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
687 return fs_info->chunk_root;
688 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
689 return fs_info->dev_root;
690 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
691 return fs_info->csum_root;
692 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
693 return fs_info->quota_root;
695 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
696 location->offset != (u64)-1);
698 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
699 btrfs_fs_roots_compare_objectids, NULL);
701 return container_of(node, struct btrfs_root, rb_node);
703 root = btrfs_read_fs_root_no_cache(fs_info, location);
707 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
708 btrfs_fs_roots_compare_roots);
713 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
715 free(fs_info->tree_root);
716 free(fs_info->extent_root);
717 free(fs_info->chunk_root);
718 free(fs_info->dev_root);
719 free(fs_info->csum_root);
720 free(fs_info->quota_root);
721 free(fs_info->super_copy);
722 free(fs_info->log_root_tree);
726 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
728 struct btrfs_fs_info *fs_info;
730 fs_info = malloc(sizeof(struct btrfs_fs_info));
734 memset(fs_info, 0, sizeof(struct btrfs_fs_info));
736 fs_info->tree_root = malloc(sizeof(struct btrfs_root));
737 fs_info->extent_root = malloc(sizeof(struct btrfs_root));
738 fs_info->chunk_root = malloc(sizeof(struct btrfs_root));
739 fs_info->dev_root = malloc(sizeof(struct btrfs_root));
740 fs_info->csum_root = malloc(sizeof(struct btrfs_root));
741 fs_info->quota_root = malloc(sizeof(struct btrfs_root));
742 fs_info->super_copy = malloc(BTRFS_SUPER_INFO_SIZE);
744 if (!fs_info->tree_root || !fs_info->extent_root ||
745 !fs_info->chunk_root || !fs_info->dev_root ||
746 !fs_info->csum_root || !fs_info->quota_root ||
747 !fs_info->super_copy)
750 memset(fs_info->super_copy, 0, BTRFS_SUPER_INFO_SIZE);
751 memset(fs_info->tree_root, 0, sizeof(struct btrfs_root));
752 memset(fs_info->extent_root, 0, sizeof(struct btrfs_root));
753 memset(fs_info->chunk_root, 0, sizeof(struct btrfs_root));
754 memset(fs_info->dev_root, 0, sizeof(struct btrfs_root));
755 memset(fs_info->csum_root, 0, sizeof(struct btrfs_root));
756 memset(fs_info->quota_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 static int setup_root_or_create_block(struct btrfs_fs_info *fs_info,
835 enum btrfs_open_ctree_flags flags,
836 struct btrfs_root *info_root,
837 u64 objectid, char *str)
839 struct btrfs_super_block *sb = fs_info->super_copy;
840 struct btrfs_root *root = fs_info->tree_root;
841 u32 leafsize = btrfs_super_leafsize(sb);
844 ret = find_and_setup_root(root, fs_info, objectid, info_root);
846 printk("Couldn't setup %s tree\n", str);
847 if (!(flags & OPEN_CTREE_PARTIAL))
850 * Need a blank node here just so we don't screw up in the
851 * million of places that assume a root has a valid ->node
854 btrfs_find_create_tree_block(info_root, 0, leafsize);
855 if (!info_root->node)
857 clear_extent_buffer_uptodate(NULL, info_root->node);
863 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
864 enum btrfs_open_ctree_flags flags)
866 struct btrfs_super_block *sb = fs_info->super_copy;
867 struct btrfs_root *root;
868 struct btrfs_key key;
877 nodesize = btrfs_super_nodesize(sb);
878 leafsize = btrfs_super_leafsize(sb);
879 sectorsize = btrfs_super_sectorsize(sb);
880 stripesize = btrfs_super_stripesize(sb);
882 root = fs_info->tree_root;
883 __setup_root(nodesize, leafsize, sectorsize, stripesize,
884 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
885 blocksize = btrfs_level_size(root, btrfs_super_root_level(sb));
886 generation = btrfs_super_generation(sb);
888 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
889 root_tree_bytenr = btrfs_super_root(sb);
890 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
891 struct btrfs_root_backup *backup;
892 int index = find_best_backup_root(sb);
893 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
894 fprintf(stderr, "Invalid backup root number\n");
897 backup = fs_info->super_copy->super_roots + index;
898 root_tree_bytenr = btrfs_backup_tree_root(backup);
899 generation = btrfs_backup_tree_root_gen(backup);
902 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
904 if (!extent_buffer_uptodate(root->node)) {
905 fprintf(stderr, "Couldn't read tree root\n");
909 ret = setup_root_or_create_block(fs_info, flags, fs_info->extent_root,
910 BTRFS_EXTENT_TREE_OBJECTID, "extent");
913 fs_info->extent_root->track_dirty = 1;
915 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
918 printk("Couldn't setup device tree\n");
921 fs_info->dev_root->track_dirty = 1;
923 ret = setup_root_or_create_block(fs_info, flags, fs_info->csum_root,
924 BTRFS_CSUM_TREE_OBJECTID, "csum");
927 fs_info->csum_root->track_dirty = 1;
929 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
930 fs_info->quota_root);
932 fs_info->quota_enabled = 1;
934 ret = find_and_setup_log_root(root, fs_info, sb);
936 printk("Couldn't setup log root tree\n");
937 if (!(flags & OPEN_CTREE_PARTIAL))
941 fs_info->generation = generation;
942 fs_info->last_trans_committed = generation;
943 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
944 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
945 btrfs_read_block_groups(fs_info->tree_root);
947 key.objectid = BTRFS_FS_TREE_OBJECTID;
948 key.type = BTRFS_ROOT_ITEM_KEY;
949 key.offset = (u64)-1;
950 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
952 if (IS_ERR(fs_info->fs_root))
957 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
959 if (fs_info->quota_root)
960 free_extent_buffer(fs_info->quota_root->node);
961 if (fs_info->csum_root)
962 free_extent_buffer(fs_info->csum_root->node);
963 if (fs_info->dev_root)
964 free_extent_buffer(fs_info->dev_root->node);
965 if (fs_info->extent_root)
966 free_extent_buffer(fs_info->extent_root->node);
967 if (fs_info->tree_root)
968 free_extent_buffer(fs_info->tree_root->node);
969 if (fs_info->log_root_tree)
970 free_extent_buffer(fs_info->log_root_tree->node);
971 if (fs_info->chunk_root)
972 free_extent_buffer(fs_info->chunk_root->node);
975 static void free_map_lookup(struct cache_extent *ce)
977 struct map_lookup *map;
979 map = container_of(ce, struct map_lookup, ce);
983 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
985 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
987 while (!list_empty(&fs_info->recow_ebs)) {
988 struct extent_buffer *eb;
989 eb = list_first_entry(&fs_info->recow_ebs,
990 struct extent_buffer, recow);
991 list_del_init(&eb->recow);
992 free_extent_buffer(eb);
994 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
995 extent_io_tree_cleanup(&fs_info->extent_cache);
996 extent_io_tree_cleanup(&fs_info->free_space_cache);
997 extent_io_tree_cleanup(&fs_info->block_group_cache);
998 extent_io_tree_cleanup(&fs_info->pinned_extents);
999 extent_io_tree_cleanup(&fs_info->pending_del);
1000 extent_io_tree_cleanup(&fs_info->extent_ins);
1003 int btrfs_scan_fs_devices(int fd, const char *path,
1004 struct btrfs_fs_devices **fs_devices,
1005 u64 sb_bytenr, int super_recover)
1012 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1014 seek_ret = lseek(fd, 0, SEEK_END);
1018 dev_size = seek_ret;
1019 lseek(fd, 0, SEEK_SET);
1020 if (sb_bytenr > dev_size) {
1021 fprintf(stderr, "Superblock bytenr is larger than device size\n");
1025 ret = btrfs_scan_one_device(fd, path, fs_devices,
1026 &total_devs, sb_bytenr, super_recover);
1028 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1032 if (total_devs != 1) {
1033 ret = btrfs_scan_lblkid();
1040 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info)
1042 struct btrfs_super_block *sb = fs_info->super_copy;
1051 nodesize = btrfs_super_nodesize(sb);
1052 leafsize = btrfs_super_leafsize(sb);
1053 sectorsize = btrfs_super_sectorsize(sb);
1054 stripesize = btrfs_super_stripesize(sb);
1056 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1057 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1059 ret = btrfs_read_sys_array(fs_info->chunk_root);
1063 blocksize = btrfs_level_size(fs_info->chunk_root,
1064 btrfs_super_chunk_root_level(sb));
1065 generation = btrfs_super_chunk_root_generation(sb);
1067 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1068 btrfs_super_chunk_root(sb),
1069 blocksize, generation);
1070 if (!fs_info->chunk_root->node ||
1071 !extent_buffer_uptodate(fs_info->chunk_root->node)) {
1072 fprintf(stderr, "Couldn't read chunk root\n");
1076 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1077 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1079 fprintf(stderr, "Couldn't read chunk tree\n");
1086 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1088 u64 root_tree_bytenr,
1089 enum btrfs_open_ctree_flags flags)
1091 struct btrfs_fs_info *fs_info;
1092 struct btrfs_super_block *disk_super;
1093 struct btrfs_fs_devices *fs_devices = NULL;
1094 struct extent_buffer *eb;
1099 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1101 /* try to drop all the caches */
1102 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1103 fprintf(stderr, "Warning, could not drop caches\n");
1105 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1107 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1110 if (flags & OPEN_CTREE_RESTORE)
1111 fs_info->on_restoring = 1;
1113 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr,
1114 (flags & OPEN_CTREE_RECOVER_SUPER));
1118 fs_info->fs_devices = fs_devices;
1119 if (flags & OPEN_CTREE_WRITES)
1124 if (flags & OPEN_CTREE_EXCLUSIVE)
1127 ret = btrfs_open_devices(fs_devices, oflags);
1131 disk_super = fs_info->super_copy;
1132 if (!(flags & OPEN_CTREE_RECOVER_SUPER))
1133 ret = btrfs_read_dev_super(fs_devices->latest_bdev,
1134 disk_super, sb_bytenr, 1);
1136 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr, 0);
1138 printk("No valid btrfs found\n");
1142 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1144 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1145 flags & OPEN_CTREE_WRITES);
1149 ret = btrfs_setup_chunk_tree_and_device_map(fs_info);
1153 eb = fs_info->chunk_root->node;
1154 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1155 btrfs_header_chunk_tree_uuid(eb),
1158 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1165 btrfs_release_all_roots(fs_info);
1166 btrfs_cleanup_all_caches(fs_info);
1168 btrfs_close_devices(fs_devices);
1170 btrfs_free_fs_info(fs_info);
1174 struct btrfs_fs_info *open_ctree_fs_info(const char *filename,
1175 u64 sb_bytenr, u64 root_tree_bytenr,
1176 enum btrfs_open_ctree_flags flags)
1179 struct btrfs_fs_info *info;
1180 int oflags = O_CREAT | O_RDWR;
1182 if (!(flags & OPEN_CTREE_WRITES))
1185 fp = open(filename, oflags, 0600);
1187 fprintf (stderr, "Could not open %s\n", filename);
1190 info = __open_ctree_fd(fp, filename, sb_bytenr, root_tree_bytenr,
1196 struct btrfs_root *open_ctree(const char *filename, u64 sb_bytenr,
1197 enum btrfs_open_ctree_flags flags)
1199 struct btrfs_fs_info *info;
1201 info = open_ctree_fs_info(filename, sb_bytenr, 0, flags);
1204 return info->fs_root;
1207 struct btrfs_root *open_ctree_fd(int fp, const char *path, u64 sb_bytenr,
1208 enum btrfs_open_ctree_flags flags)
1210 struct btrfs_fs_info *info;
1211 info = __open_ctree_fd(fp, path, sb_bytenr, 0, flags);
1214 return info->fs_root;
1217 int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
1220 u8 fsid[BTRFS_FSID_SIZE];
1221 int fsid_is_initialized = 0;
1222 struct btrfs_super_block buf;
1225 int max_super = super_recover ? BTRFS_SUPER_MIRROR_MAX : 1;
1229 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1230 ret = pread64(fd, &buf, sizeof(buf), sb_bytenr);
1231 if (ret < sizeof(buf))
1234 if (btrfs_super_bytenr(&buf) != sb_bytenr ||
1235 btrfs_super_magic(&buf) != BTRFS_MAGIC)
1238 memcpy(sb, &buf, sizeof(*sb));
1243 * we would like to check all the supers, but that would make
1244 * a btrfs mount succeed after a mkfs from a different FS.
1245 * So, we need to add a special mount option to scan for
1246 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1249 for (i = 0; i < max_super; i++) {
1250 bytenr = btrfs_sb_offset(i);
1251 ret = pread64(fd, &buf, sizeof(buf), bytenr);
1252 if (ret < sizeof(buf))
1255 if (btrfs_super_bytenr(&buf) != bytenr )
1257 /* if magic is NULL, the device was removed */
1258 if (btrfs_super_magic(&buf) == 0 && i == 0)
1260 if (btrfs_super_magic(&buf) != BTRFS_MAGIC)
1263 if (!fsid_is_initialized) {
1264 memcpy(fsid, buf.fsid, sizeof(fsid));
1265 fsid_is_initialized = 1;
1266 } else if (memcmp(fsid, buf.fsid, sizeof(fsid))) {
1268 * the superblocks (the original one and
1269 * its backups) contain data of different
1270 * filesystems -> the super cannot be trusted
1275 if (btrfs_super_generation(&buf) > transid) {
1276 memcpy(sb, &buf, sizeof(*sb));
1277 transid = btrfs_super_generation(&buf);
1281 return transid > 0 ? 0 : -1;
1284 static int write_dev_supers(struct btrfs_root *root,
1285 struct btrfs_super_block *sb,
1286 struct btrfs_device *device)
1292 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1293 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1295 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1296 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1297 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1300 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1301 * zero filled, we can use it directly
1303 ret = pwrite64(device->fd, root->fs_info->super_copy,
1304 BTRFS_SUPER_INFO_SIZE,
1305 root->fs_info->super_bytenr);
1306 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1310 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1311 bytenr = btrfs_sb_offset(i);
1312 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1315 btrfs_set_super_bytenr(sb, bytenr);
1318 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1319 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1320 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1323 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1324 * zero filled, we can use it directly
1326 ret = pwrite64(device->fd, root->fs_info->super_copy,
1327 BTRFS_SUPER_INFO_SIZE, bytenr);
1328 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1334 int write_all_supers(struct btrfs_root *root)
1336 struct list_head *cur;
1337 struct list_head *head = &root->fs_info->fs_devices->devices;
1338 struct btrfs_device *dev;
1339 struct btrfs_super_block *sb;
1340 struct btrfs_dev_item *dev_item;
1344 sb = root->fs_info->super_copy;
1345 dev_item = &sb->dev_item;
1346 list_for_each(cur, head) {
1347 dev = list_entry(cur, struct btrfs_device, dev_list);
1348 if (!dev->writeable)
1351 btrfs_set_stack_device_generation(dev_item, 0);
1352 btrfs_set_stack_device_type(dev_item, dev->type);
1353 btrfs_set_stack_device_id(dev_item, dev->devid);
1354 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1355 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1356 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1357 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1358 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1359 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1360 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1362 flags = btrfs_super_flags(sb);
1363 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1365 ret = write_dev_supers(root, sb, dev);
1371 int write_ctree_super(struct btrfs_trans_handle *trans,
1372 struct btrfs_root *root)
1375 struct btrfs_root *tree_root = root->fs_info->tree_root;
1376 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1378 if (root->fs_info->readonly)
1381 btrfs_set_super_generation(root->fs_info->super_copy,
1383 btrfs_set_super_root(root->fs_info->super_copy,
1384 tree_root->node->start);
1385 btrfs_set_super_root_level(root->fs_info->super_copy,
1386 btrfs_header_level(tree_root->node));
1387 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1388 chunk_root->node->start);
1389 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1390 btrfs_header_level(chunk_root->node));
1391 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1392 btrfs_header_generation(chunk_root->node));
1394 ret = write_all_supers(root);
1396 fprintf(stderr, "failed to write new super block err %d\n", ret);
1400 int close_ctree(struct btrfs_root *root)
1403 struct btrfs_trans_handle *trans;
1404 struct btrfs_fs_info *fs_info = root->fs_info;
1406 if (fs_info->last_trans_committed !=
1407 fs_info->generation) {
1408 trans = btrfs_start_transaction(root, 1);
1409 btrfs_commit_transaction(trans, root);
1410 trans = btrfs_start_transaction(root, 1);
1411 ret = commit_tree_roots(trans, fs_info);
1413 ret = __commit_transaction(trans, root);
1415 write_ctree_super(trans, root);
1416 btrfs_free_transaction(root, trans);
1418 btrfs_free_block_groups(fs_info);
1420 free_fs_roots_tree(&fs_info->fs_root_tree);
1422 btrfs_release_all_roots(fs_info);
1423 btrfs_close_devices(fs_info->fs_devices);
1424 btrfs_cleanup_all_caches(fs_info);
1425 btrfs_free_fs_info(fs_info);
1429 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1430 struct extent_buffer *eb)
1432 return clear_extent_buffer_dirty(eb);
1435 int wait_on_tree_block_writeback(struct btrfs_root *root,
1436 struct extent_buffer *eb)
1441 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1443 set_extent_buffer_dirty(eb);
1446 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1450 ret = extent_buffer_uptodate(buf);
1454 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1458 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1460 return set_extent_buffer_uptodate(eb);