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"
37 #include "rbtree-utils.h"
39 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
42 struct btrfs_fs_devices *fs_devices;
45 if (buf->start != btrfs_header_bytenr(buf)) {
46 printk("Check tree block failed, want=%Lu, have=%Lu\n",
47 buf->start, btrfs_header_bytenr(buf));
51 fs_devices = root->fs_info->fs_devices;
53 if (!memcmp_extent_buffer(buf, fs_devices->fsid,
59 fs_devices = fs_devices->seed;
64 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
66 return crc32c(seed, data, len);
69 void btrfs_csum_final(u32 crc, char *result)
71 *(__le32 *)result = ~cpu_to_le32(crc);
74 static int __csum_tree_block_size(struct extent_buffer *buf, u16 csum_size,
75 int verify, int silent)
81 result = malloc(csum_size * sizeof(char));
85 len = buf->len - BTRFS_CSUM_SIZE;
86 crc = crc32c(crc, buf->data + BTRFS_CSUM_SIZE, len);
87 btrfs_csum_final(crc, result);
90 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
92 printk("checksum verify failed on %llu found %08X wanted %08X\n",
93 (unsigned long long)buf->start,
95 *((u32*)(char *)buf->data));
100 write_extent_buffer(buf, result, 0, csum_size);
106 int csum_tree_block_size(struct extent_buffer *buf, u16 csum_size, int verify)
108 return __csum_tree_block_size(buf, csum_size, verify, 0);
111 int verify_tree_block_csum_silent(struct extent_buffer *buf, u16 csum_size)
113 return __csum_tree_block_size(buf, csum_size, 1, 1);
116 int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
120 btrfs_super_csum_size(root->fs_info->super_copy);
121 return csum_tree_block_size(buf, csum_size, verify);
124 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
125 u64 bytenr, u32 blocksize)
127 return find_extent_buffer(&root->fs_info->extent_cache,
131 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
132 u64 bytenr, u32 blocksize)
134 return alloc_extent_buffer(&root->fs_info->extent_cache, bytenr,
138 void 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 !btrfs_map_block(&root->fs_info->mapping_tree, READ,
149 bytenr, &length, &multi, 0, NULL)) {
150 device = multi->stripes[0].dev;
152 blocksize = min(blocksize, (u32)(64 * 1024));
153 readahead(device->fd, multi->stripes[0].physical, blocksize);
156 free_extent_buffer(eb);
160 static int verify_parent_transid(struct extent_io_tree *io_tree,
161 struct extent_buffer *eb, u64 parent_transid,
166 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
169 if (extent_buffer_uptodate(eb) &&
170 btrfs_header_generation(eb) == parent_transid) {
174 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
175 (unsigned long long)eb->start,
176 (unsigned long long)parent_transid,
177 (unsigned long long)btrfs_header_generation(eb));
179 eb->flags |= EXTENT_BAD_TRANSID;
180 printk("Ignoring transid failure\n");
186 clear_extent_buffer_uptodate(io_tree, eb);
192 int read_whole_eb(struct btrfs_fs_info *info, struct extent_buffer *eb, int mirror)
194 unsigned long offset = 0;
195 struct btrfs_multi_bio *multi = NULL;
196 struct btrfs_device *device;
199 unsigned long bytes_left = eb->len;
202 read_len = bytes_left;
205 if (!info->on_restoring &&
206 eb->start != BTRFS_SUPER_INFO_OFFSET) {
207 ret = btrfs_map_block(&info->mapping_tree, READ,
208 eb->start + offset, &read_len, &multi,
211 printk("Couldn't map the block %Lu\n", eb->start + offset);
215 device = multi->stripes[0].dev;
217 if (device->fd == 0) {
224 eb->dev_bytenr = multi->stripes[0].physical;
228 /* special case for restore metadump */
229 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
230 if (device->devid == 1)
235 eb->dev_bytenr = eb->start;
239 if (read_len > bytes_left)
240 read_len = bytes_left;
242 ret = read_extent_from_disk(eb, offset, read_len);
246 bytes_left -= read_len;
251 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
252 u32 blocksize, u64 parent_transid)
255 struct extent_buffer *eb;
256 u64 best_transid = 0;
262 eb = btrfs_find_create_tree_block(root, bytenr, blocksize);
266 if (btrfs_buffer_uptodate(eb, parent_transid))
270 ret = read_whole_eb(root->fs_info, eb, mirror_num);
271 if (ret == 0 && check_tree_block(root, eb) == 0 &&
272 csum_tree_block(root, eb, 1) == 0 &&
273 verify_parent_transid(eb->tree, eb, parent_transid, ignore)
275 if (eb->flags & EXTENT_BAD_TRANSID &&
276 list_empty(&eb->recow)) {
277 list_add_tail(&eb->recow,
278 &root->fs_info->recow_ebs);
281 btrfs_set_buffer_uptodate(eb);
285 if (check_tree_block(root, eb))
286 printk("read block failed check_tree_block\n");
288 printk("Csum didn't match\n");
291 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
293 if (num_copies == 1) {
297 if (btrfs_header_generation(eb) > best_transid && mirror_num) {
298 best_transid = btrfs_header_generation(eb);
299 good_mirror = mirror_num;
302 if (mirror_num > num_copies) {
303 mirror_num = good_mirror;
308 free_extent_buffer(eb);
312 int write_and_map_eb(struct btrfs_trans_handle *trans,
313 struct btrfs_root *root,
314 struct extent_buffer *eb)
319 u64 *raid_map = NULL;
320 struct btrfs_multi_bio *multi = NULL;
324 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
325 eb->start, &length, &multi, 0, &raid_map);
328 ret = write_raid56_with_parity(root->fs_info, eb, multi,
331 } else while (dev_nr < multi->num_stripes) {
333 eb->fd = multi->stripes[dev_nr].dev->fd;
334 eb->dev_bytenr = multi->stripes[dev_nr].physical;
335 multi->stripes[dev_nr].dev->total_ios++;
337 ret = write_extent_to_disk(eb);
344 static int write_tree_block(struct btrfs_trans_handle *trans,
345 struct btrfs_root *root,
346 struct extent_buffer *eb)
348 if (check_tree_block(root, eb))
351 if (!btrfs_buffer_uptodate(eb, trans->transid))
354 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
355 csum_tree_block(root, eb, 0);
357 return write_and_map_eb(trans, root, eb);
360 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
361 u32 stripesize, struct btrfs_root *root,
362 struct btrfs_fs_info *fs_info, u64 objectid)
365 root->commit_root = NULL;
366 root->sectorsize = sectorsize;
367 root->nodesize = nodesize;
368 root->leafsize = leafsize;
369 root->stripesize = stripesize;
371 root->track_dirty = 0;
373 root->fs_info = fs_info;
374 root->objectid = objectid;
375 root->last_trans = 0;
376 root->highest_inode = 0;
377 root->last_inode_alloc = 0;
379 INIT_LIST_HEAD(&root->dirty_list);
380 memset(&root->root_key, 0, sizeof(root->root_key));
381 memset(&root->root_item, 0, sizeof(root->root_item));
382 root->root_key.objectid = objectid;
386 static int update_cowonly_root(struct btrfs_trans_handle *trans,
387 struct btrfs_root *root)
391 struct btrfs_root *tree_root = root->fs_info->tree_root;
393 btrfs_write_dirty_block_groups(trans, root);
395 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
396 if (old_root_bytenr == root->node->start)
398 btrfs_set_root_bytenr(&root->root_item,
400 btrfs_set_root_generation(&root->root_item,
402 root->root_item.level = btrfs_header_level(root->node);
403 ret = btrfs_update_root(trans, tree_root,
407 btrfs_write_dirty_block_groups(trans, root);
412 static int commit_tree_roots(struct btrfs_trans_handle *trans,
413 struct btrfs_fs_info *fs_info)
415 struct btrfs_root *root;
416 struct list_head *next;
417 struct extent_buffer *eb;
420 if (fs_info->readonly)
423 eb = fs_info->tree_root->node;
424 extent_buffer_get(eb);
425 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
426 free_extent_buffer(eb);
430 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
431 next = fs_info->dirty_cowonly_roots.next;
433 root = list_entry(next, struct btrfs_root, dirty_list);
434 update_cowonly_root(trans, root);
435 free_extent_buffer(root->commit_root);
436 root->commit_root = NULL;
442 static int __commit_transaction(struct btrfs_trans_handle *trans,
443 struct btrfs_root *root)
447 struct extent_buffer *eb;
448 struct extent_io_tree *tree = &root->fs_info->extent_cache;
452 ret = find_first_extent_bit(tree, 0, &start, &end,
456 while(start <= end) {
457 eb = find_first_extent_buffer(tree, start);
458 BUG_ON(!eb || eb->start != start);
459 ret = write_tree_block(trans, root, eb);
462 clear_extent_buffer_dirty(eb);
463 free_extent_buffer(eb);
469 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
470 struct btrfs_root *root)
472 u64 transid = trans->transid;
474 struct btrfs_fs_info *fs_info = root->fs_info;
476 if (root->commit_root == root->node)
479 free_extent_buffer(root->commit_root);
480 root->commit_root = NULL;
482 btrfs_set_root_bytenr(&root->root_item, root->node->start);
483 btrfs_set_root_generation(&root->root_item, trans->transid);
484 root->root_item.level = btrfs_header_level(root->node);
485 ret = btrfs_update_root(trans, root->fs_info->tree_root,
486 &root->root_key, &root->root_item);
489 ret = commit_tree_roots(trans, fs_info);
491 ret = __commit_transaction(trans, root);
493 write_ctree_super(trans, root);
494 btrfs_finish_extent_commit(trans, fs_info->extent_root,
495 &fs_info->pinned_extents);
496 btrfs_free_transaction(root, trans);
497 free_extent_buffer(root->commit_root);
498 root->commit_root = NULL;
499 fs_info->running_transaction = NULL;
500 fs_info->last_trans_committed = transid;
504 static int find_and_setup_root(struct btrfs_root *tree_root,
505 struct btrfs_fs_info *fs_info,
506 u64 objectid, struct btrfs_root *root)
512 __setup_root(tree_root->nodesize, tree_root->leafsize,
513 tree_root->sectorsize, tree_root->stripesize,
514 root, fs_info, objectid);
515 ret = btrfs_find_last_root(tree_root, objectid,
516 &root->root_item, &root->root_key);
520 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
521 generation = btrfs_root_generation(&root->root_item);
522 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
523 blocksize, generation);
524 if (!extent_buffer_uptodate(root->node))
530 static int find_and_setup_log_root(struct btrfs_root *tree_root,
531 struct btrfs_fs_info *fs_info,
532 struct btrfs_super_block *disk_super)
535 u64 blocknr = btrfs_super_log_root(disk_super);
536 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
546 blocksize = btrfs_level_size(tree_root,
547 btrfs_super_log_root_level(disk_super));
549 __setup_root(tree_root->nodesize, tree_root->leafsize,
550 tree_root->sectorsize, tree_root->stripesize,
551 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
553 log_root->node = read_tree_block(tree_root, blocknr,
555 btrfs_super_generation(disk_super) + 1);
557 fs_info->log_root_tree = log_root;
559 if (!extent_buffer_uptodate(log_root->node)) {
560 free_extent_buffer(log_root->node);
562 fs_info->log_root_tree = NULL;
569 int btrfs_free_fs_root(struct btrfs_root *root)
572 free_extent_buffer(root->node);
573 if (root->commit_root)
574 free_extent_buffer(root->commit_root);
579 static void __free_fs_root(struct rb_node *node)
581 struct btrfs_root *root;
583 root = container_of(node, struct btrfs_root, rb_node);
584 btrfs_free_fs_root(root);
587 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
589 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
590 struct btrfs_key *location)
592 struct btrfs_root *root;
593 struct btrfs_root *tree_root = fs_info->tree_root;
594 struct btrfs_path *path;
595 struct extent_buffer *l;
600 root = malloc(sizeof(*root));
602 return ERR_PTR(-ENOMEM);
603 memset(root, 0, sizeof(*root));
604 if (location->offset == (u64)-1) {
605 ret = find_and_setup_root(tree_root, fs_info,
606 location->objectid, root);
614 __setup_root(tree_root->nodesize, tree_root->leafsize,
615 tree_root->sectorsize, tree_root->stripesize,
616 root, fs_info, location->objectid);
618 path = btrfs_alloc_path();
620 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
627 read_extent_buffer(l, &root->root_item,
628 btrfs_item_ptr_offset(l, path->slots[0]),
629 sizeof(root->root_item));
630 memcpy(&root->root_key, location, sizeof(*location));
633 btrfs_free_path(path);
638 generation = btrfs_root_generation(&root->root_item);
639 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
640 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
641 blocksize, generation);
644 return ERR_PTR(-EIO);
651 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
654 u64 objectid = *((u64 *)data);
655 struct btrfs_root *root;
657 root = rb_entry(node, struct btrfs_root, rb_node);
658 if (objectid > root->objectid)
660 else if (objectid < root->objectid)
666 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
667 struct rb_node *node2)
669 struct btrfs_root *root;
671 root = rb_entry(node2, struct btrfs_root, rb_node);
672 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
675 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
676 struct btrfs_key *location)
678 struct btrfs_root *root;
679 struct rb_node *node;
681 u64 objectid = location->objectid;
683 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
684 return fs_info->tree_root;
685 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
686 return fs_info->extent_root;
687 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
688 return fs_info->chunk_root;
689 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
690 return fs_info->dev_root;
691 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
692 return fs_info->csum_root;
693 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
694 return fs_info->csum_root;
696 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
697 location->offset != (u64)-1);
699 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
700 btrfs_fs_roots_compare_objectids, NULL);
702 return container_of(node, struct btrfs_root, rb_node);
704 root = btrfs_read_fs_root_no_cache(fs_info, location);
708 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
709 btrfs_fs_roots_compare_roots);
714 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
716 free(fs_info->tree_root);
717 free(fs_info->extent_root);
718 free(fs_info->chunk_root);
719 free(fs_info->dev_root);
720 free(fs_info->csum_root);
721 free(fs_info->quota_root);
722 free(fs_info->super_copy);
723 free(fs_info->log_root_tree);
727 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
729 struct btrfs_fs_info *fs_info;
731 fs_info = malloc(sizeof(struct btrfs_fs_info));
735 memset(fs_info, 0, sizeof(struct btrfs_fs_info));
737 fs_info->tree_root = malloc(sizeof(struct btrfs_root));
738 fs_info->extent_root = malloc(sizeof(struct btrfs_root));
739 fs_info->chunk_root = malloc(sizeof(struct btrfs_root));
740 fs_info->dev_root = malloc(sizeof(struct btrfs_root));
741 fs_info->csum_root = malloc(sizeof(struct btrfs_root));
742 fs_info->quota_root = malloc(sizeof(struct btrfs_root));
743 fs_info->super_copy = malloc(BTRFS_SUPER_INFO_SIZE);
745 if (!fs_info->tree_root || !fs_info->extent_root ||
746 !fs_info->chunk_root || !fs_info->dev_root ||
747 !fs_info->csum_root || !fs_info->quota_root ||
748 !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));
757 memset(fs_info->quota_root, 0, sizeof(struct btrfs_root));
759 extent_io_tree_init(&fs_info->extent_cache);
760 extent_io_tree_init(&fs_info->free_space_cache);
761 extent_io_tree_init(&fs_info->block_group_cache);
762 extent_io_tree_init(&fs_info->pinned_extents);
763 extent_io_tree_init(&fs_info->pending_del);
764 extent_io_tree_init(&fs_info->extent_ins);
765 fs_info->fs_root_tree = RB_ROOT;
766 cache_tree_init(&fs_info->mapping_tree.cache_tree);
768 mutex_init(&fs_info->fs_mutex);
769 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
770 INIT_LIST_HEAD(&fs_info->space_info);
771 INIT_LIST_HEAD(&fs_info->recow_ebs);
774 fs_info->readonly = 1;
776 fs_info->super_bytenr = sb_bytenr;
777 fs_info->data_alloc_profile = (u64)-1;
778 fs_info->metadata_alloc_profile = (u64)-1;
779 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
782 btrfs_free_fs_info(fs_info);
786 int btrfs_check_fs_compatibility(struct btrfs_super_block *sb, int writable)
790 features = btrfs_super_incompat_flags(sb) &
791 ~BTRFS_FEATURE_INCOMPAT_SUPP;
793 printk("couldn't open because of unsupported "
794 "option features (%Lx).\n",
795 (unsigned long long)features);
799 features = btrfs_super_incompat_flags(sb);
800 if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
801 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
802 btrfs_set_super_incompat_flags(sb, features);
805 features = btrfs_super_compat_ro_flags(sb) &
806 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
807 if (writable && features) {
808 printk("couldn't open RDWR because of unsupported "
809 "option features (%Lx).\n",
810 (unsigned long long)features);
816 static int find_best_backup_root(struct btrfs_super_block *super)
818 struct btrfs_root_backup *backup;
819 u64 orig_gen = btrfs_super_generation(super);
824 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
825 backup = super->super_roots + i;
826 if (btrfs_backup_tree_root_gen(backup) != orig_gen &&
827 btrfs_backup_tree_root_gen(backup) > gen) {
829 gen = btrfs_backup_tree_root_gen(backup);
835 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
836 enum btrfs_open_ctree_flags flags)
838 struct btrfs_super_block *sb = fs_info->super_copy;
839 struct btrfs_root *root;
840 struct btrfs_key key;
849 nodesize = btrfs_super_nodesize(sb);
850 leafsize = btrfs_super_leafsize(sb);
851 sectorsize = btrfs_super_sectorsize(sb);
852 stripesize = btrfs_super_stripesize(sb);
854 root = fs_info->tree_root;
855 __setup_root(nodesize, leafsize, sectorsize, stripesize,
856 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
857 blocksize = btrfs_level_size(root, btrfs_super_root_level(sb));
858 generation = btrfs_super_generation(sb);
860 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
861 root_tree_bytenr = btrfs_super_root(sb);
862 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
863 struct btrfs_root_backup *backup;
864 int index = find_best_backup_root(sb);
865 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
866 fprintf(stderr, "Invalid backup root number\n");
869 backup = fs_info->super_copy->super_roots + index;
870 root_tree_bytenr = btrfs_backup_tree_root(backup);
871 generation = btrfs_backup_tree_root_gen(backup);
874 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
876 if (!extent_buffer_uptodate(root->node)) {
877 fprintf(stderr, "Couldn't read tree root\n");
881 ret = find_and_setup_root(root, fs_info, BTRFS_EXTENT_TREE_OBJECTID,
882 fs_info->extent_root);
884 printk("Couldn't setup extent tree\n");
885 if (!(flags & OPEN_CTREE_PARTIAL))
887 /* Need a blank node here just so we don't screw up in the
888 * million of places that assume a root has a valid ->node
890 fs_info->extent_root->node =
891 btrfs_find_create_tree_block(fs_info->extent_root, 0,
893 if (!fs_info->extent_root->node)
895 clear_extent_buffer_uptodate(NULL, fs_info->extent_root->node);
897 fs_info->extent_root->track_dirty = 1;
899 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
902 printk("Couldn't setup device tree\n");
905 fs_info->dev_root->track_dirty = 1;
907 ret = find_and_setup_root(root, fs_info, BTRFS_CSUM_TREE_OBJECTID,
910 printk("Couldn't setup csum tree\n");
911 if (!(flags & OPEN_CTREE_PARTIAL))
913 /* do the same thing as extent tree rebuilding */
914 fs_info->csum_root->node =
915 btrfs_find_create_tree_block(fs_info->extent_root, 0,
917 if (!fs_info->csum_root->node)
919 clear_extent_buffer_uptodate(NULL, fs_info->csum_root->node);
921 fs_info->csum_root->track_dirty = 1;
923 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
924 fs_info->quota_root);
926 fs_info->quota_enabled = 1;
928 ret = find_and_setup_log_root(root, fs_info, sb);
930 printk("Couldn't setup log root tree\n");
934 fs_info->generation = generation;
935 fs_info->last_trans_committed = generation;
936 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
937 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
938 btrfs_read_block_groups(fs_info->tree_root);
940 key.objectid = BTRFS_FS_TREE_OBJECTID;
941 key.type = BTRFS_ROOT_ITEM_KEY;
942 key.offset = (u64)-1;
943 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
945 if (IS_ERR(fs_info->fs_root))
950 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
952 if (fs_info->quota_root)
953 free_extent_buffer(fs_info->quota_root->node);
954 if (fs_info->csum_root)
955 free_extent_buffer(fs_info->csum_root->node);
956 if (fs_info->dev_root)
957 free_extent_buffer(fs_info->dev_root->node);
958 if (fs_info->extent_root)
959 free_extent_buffer(fs_info->extent_root->node);
960 if (fs_info->tree_root)
961 free_extent_buffer(fs_info->tree_root->node);
962 if (fs_info->log_root_tree)
963 free_extent_buffer(fs_info->log_root_tree->node);
964 if (fs_info->chunk_root)
965 free_extent_buffer(fs_info->chunk_root->node);
968 static void free_map_lookup(struct cache_extent *ce)
970 struct map_lookup *map;
972 map = container_of(ce, struct map_lookup, ce);
976 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
978 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
980 while (!list_empty(&fs_info->recow_ebs)) {
981 struct extent_buffer *eb;
982 eb = list_first_entry(&fs_info->recow_ebs,
983 struct extent_buffer, recow);
984 list_del_init(&eb->recow);
985 free_extent_buffer(eb);
987 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
988 extent_io_tree_cleanup(&fs_info->extent_cache);
989 extent_io_tree_cleanup(&fs_info->free_space_cache);
990 extent_io_tree_cleanup(&fs_info->block_group_cache);
991 extent_io_tree_cleanup(&fs_info->pinned_extents);
992 extent_io_tree_cleanup(&fs_info->pending_del);
993 extent_io_tree_cleanup(&fs_info->extent_ins);
996 int btrfs_scan_fs_devices(int fd, const char *path,
997 struct btrfs_fs_devices **fs_devices,
998 u64 sb_bytenr, int super_recover)
1003 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1005 ret = btrfs_scan_one_device(fd, path, fs_devices,
1006 &total_devs, sb_bytenr, super_recover);
1008 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1012 if (total_devs != 1) {
1013 ret = btrfs_scan_lblkid(!BTRFS_UPDATE_KERNEL);
1020 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info)
1022 struct btrfs_super_block *sb = fs_info->super_copy;
1031 nodesize = btrfs_super_nodesize(sb);
1032 leafsize = btrfs_super_leafsize(sb);
1033 sectorsize = btrfs_super_sectorsize(sb);
1034 stripesize = btrfs_super_stripesize(sb);
1036 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1037 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1039 ret = btrfs_read_sys_array(fs_info->chunk_root);
1043 blocksize = btrfs_level_size(fs_info->chunk_root,
1044 btrfs_super_chunk_root_level(sb));
1045 generation = btrfs_super_chunk_root_generation(sb);
1047 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1048 btrfs_super_chunk_root(sb),
1049 blocksize, generation);
1050 if (!fs_info->chunk_root->node ||
1051 !extent_buffer_uptodate(fs_info->chunk_root->node)) {
1052 fprintf(stderr, "Couldn't read chunk root\n");
1056 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1057 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1059 fprintf(stderr, "Couldn't read chunk tree\n");
1066 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1068 u64 root_tree_bytenr,
1069 enum btrfs_open_ctree_flags flags)
1071 struct btrfs_fs_info *fs_info;
1072 struct btrfs_super_block *disk_super;
1073 struct btrfs_fs_devices *fs_devices = NULL;
1074 struct extent_buffer *eb;
1079 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1081 /* try to drop all the caches */
1082 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1083 fprintf(stderr, "Warning, could not drop caches\n");
1085 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1087 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1090 if (flags & OPEN_CTREE_RESTORE)
1091 fs_info->on_restoring = 1;
1093 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr,
1094 (flags & OPEN_CTREE_RECOVER_SUPER));
1098 fs_info->fs_devices = fs_devices;
1099 if (flags & OPEN_CTREE_WRITES)
1104 if (flags & OPEN_CTREE_EXCLUSIVE)
1107 ret = btrfs_open_devices(fs_devices, oflags);
1111 disk_super = fs_info->super_copy;
1112 if (!(flags & OPEN_CTREE_RECOVER_SUPER))
1113 ret = btrfs_read_dev_super(fs_devices->latest_bdev,
1114 disk_super, sb_bytenr, 1);
1116 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr, 0);
1118 printk("No valid btrfs found\n");
1122 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1124 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1125 flags & OPEN_CTREE_WRITES);
1129 ret = btrfs_setup_chunk_tree_and_device_map(fs_info);
1133 eb = fs_info->chunk_root->node;
1134 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1135 btrfs_header_chunk_tree_uuid(eb),
1138 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1145 btrfs_release_all_roots(fs_info);
1146 btrfs_cleanup_all_caches(fs_info);
1148 btrfs_close_devices(fs_devices);
1150 btrfs_free_fs_info(fs_info);
1154 struct btrfs_fs_info *open_ctree_fs_info(const char *filename,
1155 u64 sb_bytenr, u64 root_tree_bytenr,
1156 enum btrfs_open_ctree_flags flags)
1159 struct btrfs_fs_info *info;
1160 int oflags = O_CREAT | O_RDWR;
1162 if (!(flags & OPEN_CTREE_WRITES))
1165 fp = open(filename, oflags, 0600);
1167 fprintf (stderr, "Could not open %s\n", filename);
1170 info = __open_ctree_fd(fp, filename, sb_bytenr, root_tree_bytenr,
1176 struct btrfs_root *open_ctree(const char *filename, u64 sb_bytenr,
1177 enum btrfs_open_ctree_flags flags)
1179 struct btrfs_fs_info *info;
1181 info = open_ctree_fs_info(filename, sb_bytenr, 0, flags);
1184 return info->fs_root;
1187 struct btrfs_root *open_ctree_fd(int fp, const char *path, u64 sb_bytenr,
1188 enum btrfs_open_ctree_flags flags)
1190 struct btrfs_fs_info *info;
1191 info = __open_ctree_fd(fp, path, sb_bytenr, 0, flags);
1194 return info->fs_root;
1197 int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
1200 u8 fsid[BTRFS_FSID_SIZE];
1201 int fsid_is_initialized = 0;
1202 struct btrfs_super_block buf;
1205 int max_super = super_recover ? BTRFS_SUPER_MIRROR_MAX : 1;
1209 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1210 ret = pread64(fd, &buf, sizeof(buf), sb_bytenr);
1211 if (ret < sizeof(buf))
1214 if (btrfs_super_bytenr(&buf) != sb_bytenr ||
1215 btrfs_super_magic(&buf) != BTRFS_MAGIC)
1218 memcpy(sb, &buf, sizeof(*sb));
1223 * we would like to check all the supers, but that would make
1224 * a btrfs mount succeed after a mkfs from a different FS.
1225 * So, we need to add a special mount option to scan for
1226 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1229 for (i = 0; i < max_super; i++) {
1230 bytenr = btrfs_sb_offset(i);
1231 ret = pread64(fd, &buf, sizeof(buf), bytenr);
1232 if (ret < sizeof(buf))
1235 if (btrfs_super_bytenr(&buf) != bytenr )
1237 /* if magic is NULL, the device was removed */
1238 if (btrfs_super_magic(&buf) == 0 && i == 0)
1240 if (btrfs_super_magic(&buf) != BTRFS_MAGIC)
1243 if (!fsid_is_initialized) {
1244 memcpy(fsid, buf.fsid, sizeof(fsid));
1245 fsid_is_initialized = 1;
1246 } else if (memcmp(fsid, buf.fsid, sizeof(fsid))) {
1248 * the superblocks (the original one and
1249 * its backups) contain data of different
1250 * filesystems -> the super cannot be trusted
1255 if (btrfs_super_generation(&buf) > transid) {
1256 memcpy(sb, &buf, sizeof(*sb));
1257 transid = btrfs_super_generation(&buf);
1261 return transid > 0 ? 0 : -1;
1264 static int write_dev_supers(struct btrfs_root *root,
1265 struct btrfs_super_block *sb,
1266 struct btrfs_device *device)
1272 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1273 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1275 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1276 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1277 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1280 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1281 * zero filled, we can use it directly
1283 ret = pwrite64(device->fd, root->fs_info->super_copy,
1284 BTRFS_SUPER_INFO_SIZE,
1285 root->fs_info->super_bytenr);
1286 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1290 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1291 bytenr = btrfs_sb_offset(i);
1292 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1295 btrfs_set_super_bytenr(sb, bytenr);
1298 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1299 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1300 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1303 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1304 * zero filled, we can use it directly
1306 ret = pwrite64(device->fd, root->fs_info->super_copy,
1307 BTRFS_SUPER_INFO_SIZE, bytenr);
1308 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1314 int write_all_supers(struct btrfs_root *root)
1316 struct list_head *cur;
1317 struct list_head *head = &root->fs_info->fs_devices->devices;
1318 struct btrfs_device *dev;
1319 struct btrfs_super_block *sb;
1320 struct btrfs_dev_item *dev_item;
1324 sb = root->fs_info->super_copy;
1325 dev_item = &sb->dev_item;
1326 list_for_each(cur, head) {
1327 dev = list_entry(cur, struct btrfs_device, dev_list);
1328 if (!dev->writeable)
1331 btrfs_set_stack_device_generation(dev_item, 0);
1332 btrfs_set_stack_device_type(dev_item, dev->type);
1333 btrfs_set_stack_device_id(dev_item, dev->devid);
1334 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1335 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1336 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1337 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1338 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1339 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1340 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1342 flags = btrfs_super_flags(sb);
1343 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1345 ret = write_dev_supers(root, sb, dev);
1351 int write_ctree_super(struct btrfs_trans_handle *trans,
1352 struct btrfs_root *root)
1355 struct btrfs_root *tree_root = root->fs_info->tree_root;
1356 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1358 if (root->fs_info->readonly)
1361 btrfs_set_super_generation(root->fs_info->super_copy,
1363 btrfs_set_super_root(root->fs_info->super_copy,
1364 tree_root->node->start);
1365 btrfs_set_super_root_level(root->fs_info->super_copy,
1366 btrfs_header_level(tree_root->node));
1367 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1368 chunk_root->node->start);
1369 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1370 btrfs_header_level(chunk_root->node));
1371 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1372 btrfs_header_generation(chunk_root->node));
1374 ret = write_all_supers(root);
1376 fprintf(stderr, "failed to write new super block err %d\n", ret);
1380 int close_ctree(struct btrfs_root *root)
1383 struct btrfs_trans_handle *trans;
1384 struct btrfs_fs_info *fs_info = root->fs_info;
1386 if (fs_info->last_trans_committed !=
1387 fs_info->generation) {
1388 trans = btrfs_start_transaction(root, 1);
1389 btrfs_commit_transaction(trans, root);
1390 trans = btrfs_start_transaction(root, 1);
1391 ret = commit_tree_roots(trans, fs_info);
1393 ret = __commit_transaction(trans, root);
1395 write_ctree_super(trans, root);
1396 btrfs_free_transaction(root, trans);
1398 btrfs_free_block_groups(fs_info);
1400 free_fs_roots_tree(&fs_info->fs_root_tree);
1402 btrfs_release_all_roots(fs_info);
1403 btrfs_close_devices(fs_info->fs_devices);
1404 btrfs_cleanup_all_caches(fs_info);
1405 btrfs_free_fs_info(fs_info);
1409 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1410 struct extent_buffer *eb)
1412 return clear_extent_buffer_dirty(eb);
1415 int wait_on_tree_block_writeback(struct btrfs_root *root,
1416 struct extent_buffer *eb)
1421 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1423 set_extent_buffer_dirty(eb);
1426 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1430 ret = extent_buffer_uptodate(buf);
1434 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1438 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1440 return set_extent_buffer_uptodate(eb);