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)
40 #define BTRFS_BAD_LEVEL (-3)
41 #define BTRFS_BAD_NRITEMS (-4)
43 /* Calculate max possible nritems for a leaf/node */
44 static u32 max_nritems(u8 level, u32 nodesize)
48 return ((nodesize - sizeof(struct btrfs_header)) /
49 sizeof(struct btrfs_item));
50 return ((nodesize - sizeof(struct btrfs_header)) /
51 sizeof(struct btrfs_key_ptr));
54 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
57 struct btrfs_fs_devices *fs_devices;
58 int ret = BTRFS_BAD_FSID;
60 if (buf->start != btrfs_header_bytenr(buf))
61 return BTRFS_BAD_BYTENR;
62 if (btrfs_header_level(buf) >= BTRFS_MAX_LEVEL)
63 return BTRFS_BAD_LEVEL;
64 if (btrfs_header_nritems(buf) > max_nritems(btrfs_header_level(buf),
66 return BTRFS_BAD_NRITEMS;
68 fs_devices = root->fs_info->fs_devices;
70 if (root->fs_info->ignore_fsid_mismatch ||
71 !memcmp_extent_buffer(buf, fs_devices->fsid,
77 fs_devices = fs_devices->seed;
82 static void print_tree_block_error(struct btrfs_root *root,
83 struct extent_buffer *eb,
86 char fs_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
87 char found_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
88 u8 buf[BTRFS_UUID_SIZE];
92 read_extent_buffer(eb, buf, btrfs_header_fsid(),
94 uuid_unparse(buf, found_uuid);
95 uuid_unparse(root->fs_info->fsid, fs_uuid);
96 fprintf(stderr, "fsid mismatch, want=%s, have=%s\n",
99 case BTRFS_BAD_BYTENR:
100 fprintf(stderr, "bytenr mismatch, want=%llu, have=%llu\n",
101 eb->start, btrfs_header_bytenr(eb));
103 case BTRFS_BAD_LEVEL:
104 fprintf(stderr, "bad level, %u > %u\n",
105 btrfs_header_level(eb), BTRFS_MAX_LEVEL);
107 case BTRFS_BAD_NRITEMS:
108 fprintf(stderr, "invalid nr_items: %u\n",
109 btrfs_header_nritems(eb));
114 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
116 return crc32c(seed, data, len);
119 void btrfs_csum_final(u32 crc, char *result)
121 *(__le32 *)result = ~cpu_to_le32(crc);
124 static int __csum_tree_block_size(struct extent_buffer *buf, u16 csum_size,
125 int verify, int silent)
131 result = malloc(csum_size * sizeof(char));
135 len = buf->len - BTRFS_CSUM_SIZE;
136 crc = crc32c(crc, buf->data + BTRFS_CSUM_SIZE, len);
137 btrfs_csum_final(crc, result);
140 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
142 printk("checksum verify failed on %llu found %08X wanted %08X\n",
143 (unsigned long long)buf->start,
145 *((u32*)(char *)buf->data));
150 write_extent_buffer(buf, result, 0, csum_size);
156 int csum_tree_block_size(struct extent_buffer *buf, u16 csum_size, int verify)
158 return __csum_tree_block_size(buf, csum_size, verify, 0);
161 int verify_tree_block_csum_silent(struct extent_buffer *buf, u16 csum_size)
163 return __csum_tree_block_size(buf, csum_size, 1, 1);
166 int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
170 btrfs_super_csum_size(root->fs_info->super_copy);
171 if (verify && root->fs_info->suppress_check_block_errors)
172 return verify_tree_block_csum_silent(buf, csum_size);
173 return csum_tree_block_size(buf, csum_size, verify);
176 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
177 u64 bytenr, u32 blocksize)
179 return find_extent_buffer(&root->fs_info->extent_cache,
183 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
184 u64 bytenr, u32 blocksize)
186 return alloc_extent_buffer(&root->fs_info->extent_cache, bytenr,
190 void readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
193 struct extent_buffer *eb;
195 struct btrfs_multi_bio *multi = NULL;
196 struct btrfs_device *device;
198 eb = btrfs_find_tree_block(root, bytenr, blocksize);
199 if (!(eb && btrfs_buffer_uptodate(eb, parent_transid)) &&
200 !btrfs_map_block(&root->fs_info->mapping_tree, READ,
201 bytenr, &length, &multi, 0, NULL)) {
202 device = multi->stripes[0].dev;
204 blocksize = min(blocksize, (u32)(64 * 1024));
205 readahead(device->fd, multi->stripes[0].physical, blocksize);
208 free_extent_buffer(eb);
212 static int verify_parent_transid(struct extent_io_tree *io_tree,
213 struct extent_buffer *eb, u64 parent_transid,
218 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
221 if (extent_buffer_uptodate(eb) &&
222 btrfs_header_generation(eb) == parent_transid) {
226 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
227 (unsigned long long)eb->start,
228 (unsigned long long)parent_transid,
229 (unsigned long long)btrfs_header_generation(eb));
231 eb->flags |= EXTENT_BAD_TRANSID;
232 printk("Ignoring transid failure\n");
238 clear_extent_buffer_uptodate(io_tree, eb);
244 int read_whole_eb(struct btrfs_fs_info *info, struct extent_buffer *eb, int mirror)
246 unsigned long offset = 0;
247 struct btrfs_multi_bio *multi = NULL;
248 struct btrfs_device *device;
251 unsigned long bytes_left = eb->len;
254 read_len = bytes_left;
257 if (!info->on_restoring &&
258 eb->start != BTRFS_SUPER_INFO_OFFSET) {
259 ret = btrfs_map_block(&info->mapping_tree, READ,
260 eb->start + offset, &read_len, &multi,
263 printk("Couldn't map the block %Lu\n", eb->start + offset);
267 device = multi->stripes[0].dev;
269 if (device->fd == 0) {
276 eb->dev_bytenr = multi->stripes[0].physical;
280 /* special case for restore metadump */
281 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
282 if (device->devid == 1)
287 eb->dev_bytenr = eb->start;
291 if (read_len > bytes_left)
292 read_len = bytes_left;
294 ret = read_extent_from_disk(eb, offset, read_len);
298 bytes_left -= read_len;
303 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
304 u32 blocksize, u64 parent_transid)
307 struct extent_buffer *eb;
308 u64 best_transid = 0;
314 eb = btrfs_find_create_tree_block(root, bytenr, blocksize);
316 return ERR_PTR(-ENOMEM);
318 if (btrfs_buffer_uptodate(eb, parent_transid))
322 ret = read_whole_eb(root->fs_info, eb, mirror_num);
323 if (ret == 0 && csum_tree_block(root, eb, 1) == 0 &&
324 check_tree_block(root, eb) == 0 &&
325 verify_parent_transid(eb->tree, eb, parent_transid, ignore)
327 if (eb->flags & EXTENT_BAD_TRANSID &&
328 list_empty(&eb->recow)) {
329 list_add_tail(&eb->recow,
330 &root->fs_info->recow_ebs);
333 btrfs_set_buffer_uptodate(eb);
337 if (check_tree_block(root, eb)) {
338 if (!root->fs_info->suppress_check_block_errors)
339 print_tree_block_error(root, eb,
340 check_tree_block(root, eb));
342 if (!root->fs_info->suppress_check_block_errors)
343 fprintf(stderr, "Csum didn't match\n");
348 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
350 if (num_copies == 1) {
354 if (btrfs_header_generation(eb) > best_transid && mirror_num) {
355 best_transid = btrfs_header_generation(eb);
356 good_mirror = mirror_num;
359 if (mirror_num > num_copies) {
360 mirror_num = good_mirror;
365 free_extent_buffer(eb);
369 int write_and_map_eb(struct btrfs_trans_handle *trans,
370 struct btrfs_root *root,
371 struct extent_buffer *eb)
376 u64 *raid_map = NULL;
377 struct btrfs_multi_bio *multi = NULL;
381 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
382 eb->start, &length, &multi, 0, &raid_map);
385 ret = write_raid56_with_parity(root->fs_info, eb, multi,
388 } else while (dev_nr < multi->num_stripes) {
390 eb->fd = multi->stripes[dev_nr].dev->fd;
391 eb->dev_bytenr = multi->stripes[dev_nr].physical;
392 multi->stripes[dev_nr].dev->total_ios++;
394 ret = write_extent_to_disk(eb);
401 int write_tree_block(struct btrfs_trans_handle *trans,
402 struct btrfs_root *root,
403 struct extent_buffer *eb)
405 if (check_tree_block(root, eb)) {
406 print_tree_block_error(root, eb, check_tree_block(root, eb));
410 if (trans && !btrfs_buffer_uptodate(eb, trans->transid))
413 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
414 csum_tree_block(root, eb, 0);
416 return write_and_map_eb(trans, root, eb);
419 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
420 u32 stripesize, struct btrfs_root *root,
421 struct btrfs_fs_info *fs_info, u64 objectid)
424 root->commit_root = NULL;
425 root->sectorsize = sectorsize;
426 root->nodesize = nodesize;
427 root->leafsize = leafsize;
428 root->stripesize = stripesize;
430 root->track_dirty = 0;
432 root->fs_info = fs_info;
433 root->objectid = objectid;
434 root->last_trans = 0;
435 root->highest_inode = 0;
436 root->last_inode_alloc = 0;
438 INIT_LIST_HEAD(&root->dirty_list);
439 INIT_LIST_HEAD(&root->orphan_data_extents);
440 memset(&root->root_key, 0, sizeof(root->root_key));
441 memset(&root->root_item, 0, sizeof(root->root_item));
442 root->root_key.objectid = objectid;
446 static int update_cowonly_root(struct btrfs_trans_handle *trans,
447 struct btrfs_root *root)
451 struct btrfs_root *tree_root = root->fs_info->tree_root;
453 btrfs_write_dirty_block_groups(trans, root);
455 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
456 if (old_root_bytenr == root->node->start)
458 btrfs_set_root_bytenr(&root->root_item,
460 btrfs_set_root_generation(&root->root_item,
462 root->root_item.level = btrfs_header_level(root->node);
463 ret = btrfs_update_root(trans, tree_root,
467 btrfs_write_dirty_block_groups(trans, root);
472 static int commit_tree_roots(struct btrfs_trans_handle *trans,
473 struct btrfs_fs_info *fs_info)
475 struct btrfs_root *root;
476 struct list_head *next;
477 struct extent_buffer *eb;
480 if (fs_info->readonly)
483 eb = fs_info->tree_root->node;
484 extent_buffer_get(eb);
485 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
486 free_extent_buffer(eb);
490 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
491 next = fs_info->dirty_cowonly_roots.next;
493 root = list_entry(next, struct btrfs_root, dirty_list);
494 update_cowonly_root(trans, root);
495 free_extent_buffer(root->commit_root);
496 root->commit_root = NULL;
502 static int __commit_transaction(struct btrfs_trans_handle *trans,
503 struct btrfs_root *root)
507 struct extent_buffer *eb;
508 struct extent_io_tree *tree = &root->fs_info->extent_cache;
512 ret = find_first_extent_bit(tree, 0, &start, &end,
516 while(start <= end) {
517 eb = find_first_extent_buffer(tree, start);
518 BUG_ON(!eb || eb->start != start);
519 ret = write_tree_block(trans, root, eb);
522 clear_extent_buffer_dirty(eb);
523 free_extent_buffer(eb);
529 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
530 struct btrfs_root *root)
532 u64 transid = trans->transid;
534 struct btrfs_fs_info *fs_info = root->fs_info;
536 if (root->commit_root == root->node)
538 if (root == root->fs_info->tree_root)
541 free_extent_buffer(root->commit_root);
542 root->commit_root = NULL;
544 btrfs_set_root_bytenr(&root->root_item, root->node->start);
545 btrfs_set_root_generation(&root->root_item, trans->transid);
546 root->root_item.level = btrfs_header_level(root->node);
547 ret = btrfs_update_root(trans, root->fs_info->tree_root,
548 &root->root_key, &root->root_item);
551 ret = commit_tree_roots(trans, fs_info);
553 ret = __commit_transaction(trans, root);
555 write_ctree_super(trans, root);
556 btrfs_finish_extent_commit(trans, fs_info->extent_root,
557 &fs_info->pinned_extents);
558 btrfs_free_transaction(root, trans);
559 free_extent_buffer(root->commit_root);
560 root->commit_root = NULL;
561 fs_info->running_transaction = NULL;
562 fs_info->last_trans_committed = transid;
566 static int find_and_setup_root(struct btrfs_root *tree_root,
567 struct btrfs_fs_info *fs_info,
568 u64 objectid, struct btrfs_root *root)
574 __setup_root(tree_root->nodesize, tree_root->leafsize,
575 tree_root->sectorsize, tree_root->stripesize,
576 root, fs_info, objectid);
577 ret = btrfs_find_last_root(tree_root, objectid,
578 &root->root_item, &root->root_key);
582 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
583 generation = btrfs_root_generation(&root->root_item);
584 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
585 blocksize, generation);
586 if (!extent_buffer_uptodate(root->node))
592 static int find_and_setup_log_root(struct btrfs_root *tree_root,
593 struct btrfs_fs_info *fs_info,
594 struct btrfs_super_block *disk_super)
597 u64 blocknr = btrfs_super_log_root(disk_super);
598 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
608 blocksize = btrfs_level_size(tree_root,
609 btrfs_super_log_root_level(disk_super));
611 __setup_root(tree_root->nodesize, tree_root->leafsize,
612 tree_root->sectorsize, tree_root->stripesize,
613 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
615 log_root->node = read_tree_block(tree_root, blocknr,
617 btrfs_super_generation(disk_super) + 1);
619 fs_info->log_root_tree = log_root;
621 if (!extent_buffer_uptodate(log_root->node)) {
622 free_extent_buffer(log_root->node);
624 fs_info->log_root_tree = NULL;
631 int btrfs_free_fs_root(struct btrfs_root *root)
634 free_extent_buffer(root->node);
635 if (root->commit_root)
636 free_extent_buffer(root->commit_root);
641 static void __free_fs_root(struct rb_node *node)
643 struct btrfs_root *root;
645 root = container_of(node, struct btrfs_root, rb_node);
646 btrfs_free_fs_root(root);
649 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
651 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
652 struct btrfs_key *location)
654 struct btrfs_root *root;
655 struct btrfs_root *tree_root = fs_info->tree_root;
656 struct btrfs_path *path;
657 struct extent_buffer *l;
662 root = malloc(sizeof(*root));
664 return ERR_PTR(-ENOMEM);
665 memset(root, 0, sizeof(*root));
666 if (location->offset == (u64)-1) {
667 ret = find_and_setup_root(tree_root, fs_info,
668 location->objectid, root);
676 __setup_root(tree_root->nodesize, tree_root->leafsize,
677 tree_root->sectorsize, tree_root->stripesize,
678 root, fs_info, location->objectid);
680 path = btrfs_alloc_path();
682 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
689 read_extent_buffer(l, &root->root_item,
690 btrfs_item_ptr_offset(l, path->slots[0]),
691 sizeof(root->root_item));
692 memcpy(&root->root_key, location, sizeof(*location));
695 btrfs_free_path(path);
700 generation = btrfs_root_generation(&root->root_item);
701 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
702 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
703 blocksize, generation);
704 if (!extent_buffer_uptodate(root->node)) {
706 return ERR_PTR(-EIO);
713 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
716 u64 objectid = *((u64 *)data);
717 struct btrfs_root *root;
719 root = rb_entry(node, struct btrfs_root, rb_node);
720 if (objectid > root->objectid)
722 else if (objectid < root->objectid)
728 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
729 struct rb_node *node2)
731 struct btrfs_root *root;
733 root = rb_entry(node2, struct btrfs_root, rb_node);
734 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
737 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
738 struct btrfs_key *location)
740 struct btrfs_root *root;
741 struct rb_node *node;
743 u64 objectid = location->objectid;
745 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
746 return fs_info->tree_root;
747 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
748 return fs_info->extent_root;
749 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
750 return fs_info->chunk_root;
751 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
752 return fs_info->dev_root;
753 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
754 return fs_info->csum_root;
755 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
756 return fs_info->quota_root;
758 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
759 location->offset != (u64)-1);
761 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
762 btrfs_fs_roots_compare_objectids, NULL);
764 return container_of(node, struct btrfs_root, rb_node);
766 root = btrfs_read_fs_root_no_cache(fs_info, location);
770 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
771 btrfs_fs_roots_compare_roots);
776 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
778 free(fs_info->tree_root);
779 free(fs_info->extent_root);
780 free(fs_info->chunk_root);
781 free(fs_info->dev_root);
782 free(fs_info->csum_root);
783 free(fs_info->quota_root);
784 free(fs_info->super_copy);
785 free(fs_info->log_root_tree);
789 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
791 struct btrfs_fs_info *fs_info;
793 fs_info = malloc(sizeof(struct btrfs_fs_info));
797 memset(fs_info, 0, sizeof(struct btrfs_fs_info));
799 fs_info->tree_root = malloc(sizeof(struct btrfs_root));
800 fs_info->extent_root = malloc(sizeof(struct btrfs_root));
801 fs_info->chunk_root = malloc(sizeof(struct btrfs_root));
802 fs_info->dev_root = malloc(sizeof(struct btrfs_root));
803 fs_info->csum_root = malloc(sizeof(struct btrfs_root));
804 fs_info->quota_root = malloc(sizeof(struct btrfs_root));
805 fs_info->super_copy = malloc(BTRFS_SUPER_INFO_SIZE);
807 if (!fs_info->tree_root || !fs_info->extent_root ||
808 !fs_info->chunk_root || !fs_info->dev_root ||
809 !fs_info->csum_root || !fs_info->quota_root ||
810 !fs_info->super_copy)
813 memset(fs_info->super_copy, 0, BTRFS_SUPER_INFO_SIZE);
814 memset(fs_info->tree_root, 0, sizeof(struct btrfs_root));
815 memset(fs_info->extent_root, 0, sizeof(struct btrfs_root));
816 memset(fs_info->chunk_root, 0, sizeof(struct btrfs_root));
817 memset(fs_info->dev_root, 0, sizeof(struct btrfs_root));
818 memset(fs_info->csum_root, 0, sizeof(struct btrfs_root));
819 memset(fs_info->quota_root, 0, sizeof(struct btrfs_root));
821 extent_io_tree_init(&fs_info->extent_cache);
822 extent_io_tree_init(&fs_info->free_space_cache);
823 extent_io_tree_init(&fs_info->block_group_cache);
824 extent_io_tree_init(&fs_info->pinned_extents);
825 extent_io_tree_init(&fs_info->pending_del);
826 extent_io_tree_init(&fs_info->extent_ins);
827 fs_info->excluded_extents = NULL;
829 fs_info->fs_root_tree = RB_ROOT;
830 cache_tree_init(&fs_info->mapping_tree.cache_tree);
832 mutex_init(&fs_info->fs_mutex);
833 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
834 INIT_LIST_HEAD(&fs_info->space_info);
835 INIT_LIST_HEAD(&fs_info->recow_ebs);
838 fs_info->readonly = 1;
840 fs_info->super_bytenr = sb_bytenr;
841 fs_info->data_alloc_profile = (u64)-1;
842 fs_info->metadata_alloc_profile = (u64)-1;
843 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
846 btrfs_free_fs_info(fs_info);
850 int btrfs_check_fs_compatibility(struct btrfs_super_block *sb, int writable)
854 features = btrfs_super_incompat_flags(sb) &
855 ~BTRFS_FEATURE_INCOMPAT_SUPP;
857 printk("couldn't open because of unsupported "
858 "option features (%Lx).\n",
859 (unsigned long long)features);
863 features = btrfs_super_incompat_flags(sb);
864 if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
865 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
866 btrfs_set_super_incompat_flags(sb, features);
869 features = btrfs_super_compat_ro_flags(sb) &
870 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
871 if (writable && features) {
872 printk("couldn't open RDWR because of unsupported "
873 "option features (%Lx).\n",
874 (unsigned long long)features);
880 static int find_best_backup_root(struct btrfs_super_block *super)
882 struct btrfs_root_backup *backup;
883 u64 orig_gen = btrfs_super_generation(super);
888 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
889 backup = super->super_roots + i;
890 if (btrfs_backup_tree_root_gen(backup) != orig_gen &&
891 btrfs_backup_tree_root_gen(backup) > gen) {
893 gen = btrfs_backup_tree_root_gen(backup);
899 static int setup_root_or_create_block(struct btrfs_fs_info *fs_info,
900 enum btrfs_open_ctree_flags flags,
901 struct btrfs_root *info_root,
902 u64 objectid, char *str)
904 struct btrfs_super_block *sb = fs_info->super_copy;
905 struct btrfs_root *root = fs_info->tree_root;
906 u32 leafsize = btrfs_super_leafsize(sb);
909 ret = find_and_setup_root(root, fs_info, objectid, info_root);
911 printk("Couldn't setup %s tree\n", str);
912 if (!(flags & OPEN_CTREE_PARTIAL))
915 * Need a blank node here just so we don't screw up in the
916 * million of places that assume a root has a valid ->node
919 btrfs_find_create_tree_block(info_root, 0, leafsize);
920 if (!info_root->node)
922 clear_extent_buffer_uptodate(NULL, info_root->node);
928 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
929 enum btrfs_open_ctree_flags flags)
931 struct btrfs_super_block *sb = fs_info->super_copy;
932 struct btrfs_root *root;
933 struct btrfs_key key;
942 nodesize = btrfs_super_nodesize(sb);
943 leafsize = btrfs_super_leafsize(sb);
944 sectorsize = btrfs_super_sectorsize(sb);
945 stripesize = btrfs_super_stripesize(sb);
947 root = fs_info->tree_root;
948 __setup_root(nodesize, leafsize, sectorsize, stripesize,
949 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
950 blocksize = btrfs_level_size(root, btrfs_super_root_level(sb));
951 generation = btrfs_super_generation(sb);
953 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
954 root_tree_bytenr = btrfs_super_root(sb);
955 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
956 struct btrfs_root_backup *backup;
957 int index = find_best_backup_root(sb);
958 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
959 fprintf(stderr, "Invalid backup root number\n");
962 backup = fs_info->super_copy->super_roots + index;
963 root_tree_bytenr = btrfs_backup_tree_root(backup);
964 generation = btrfs_backup_tree_root_gen(backup);
967 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
969 if (!extent_buffer_uptodate(root->node)) {
970 fprintf(stderr, "Couldn't read tree root\n");
974 ret = setup_root_or_create_block(fs_info, flags, fs_info->extent_root,
975 BTRFS_EXTENT_TREE_OBJECTID, "extent");
978 fs_info->extent_root->track_dirty = 1;
980 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
983 printk("Couldn't setup device tree\n");
986 fs_info->dev_root->track_dirty = 1;
988 ret = setup_root_or_create_block(fs_info, flags, fs_info->csum_root,
989 BTRFS_CSUM_TREE_OBJECTID, "csum");
992 fs_info->csum_root->track_dirty = 1;
994 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
995 fs_info->quota_root);
997 fs_info->quota_enabled = 1;
999 ret = find_and_setup_log_root(root, fs_info, sb);
1001 printk("Couldn't setup log root tree\n");
1002 if (!(flags & OPEN_CTREE_PARTIAL))
1006 fs_info->generation = generation;
1007 fs_info->last_trans_committed = generation;
1008 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
1009 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
1010 btrfs_read_block_groups(fs_info->tree_root);
1012 key.objectid = BTRFS_FS_TREE_OBJECTID;
1013 key.type = BTRFS_ROOT_ITEM_KEY;
1014 key.offset = (u64)-1;
1015 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
1017 if (IS_ERR(fs_info->fs_root))
1022 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
1024 if (fs_info->quota_root)
1025 free_extent_buffer(fs_info->quota_root->node);
1026 if (fs_info->csum_root)
1027 free_extent_buffer(fs_info->csum_root->node);
1028 if (fs_info->dev_root)
1029 free_extent_buffer(fs_info->dev_root->node);
1030 if (fs_info->extent_root)
1031 free_extent_buffer(fs_info->extent_root->node);
1032 if (fs_info->tree_root)
1033 free_extent_buffer(fs_info->tree_root->node);
1034 if (fs_info->log_root_tree)
1035 free_extent_buffer(fs_info->log_root_tree->node);
1036 if (fs_info->chunk_root)
1037 free_extent_buffer(fs_info->chunk_root->node);
1040 static void free_map_lookup(struct cache_extent *ce)
1042 struct map_lookup *map;
1044 map = container_of(ce, struct map_lookup, ce);
1048 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
1050 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
1052 while (!list_empty(&fs_info->recow_ebs)) {
1053 struct extent_buffer *eb;
1054 eb = list_first_entry(&fs_info->recow_ebs,
1055 struct extent_buffer, recow);
1056 list_del_init(&eb->recow);
1057 free_extent_buffer(eb);
1059 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
1060 extent_io_tree_cleanup(&fs_info->extent_cache);
1061 extent_io_tree_cleanup(&fs_info->free_space_cache);
1062 extent_io_tree_cleanup(&fs_info->block_group_cache);
1063 extent_io_tree_cleanup(&fs_info->pinned_extents);
1064 extent_io_tree_cleanup(&fs_info->pending_del);
1065 extent_io_tree_cleanup(&fs_info->extent_ins);
1068 int btrfs_scan_fs_devices(int fd, const char *path,
1069 struct btrfs_fs_devices **fs_devices,
1070 u64 sb_bytenr, int super_recover,
1078 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1080 seek_ret = lseek(fd, 0, SEEK_END);
1084 dev_size = seek_ret;
1085 lseek(fd, 0, SEEK_SET);
1086 if (sb_bytenr > dev_size) {
1087 fprintf(stderr, "Superblock bytenr is larger than device size\n");
1091 ret = btrfs_scan_one_device(fd, path, fs_devices,
1092 &total_devs, sb_bytenr, super_recover);
1094 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1098 if (!skip_devices && total_devs != 1) {
1099 ret = btrfs_scan_lblkid();
1106 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info)
1108 struct btrfs_super_block *sb = fs_info->super_copy;
1117 nodesize = btrfs_super_nodesize(sb);
1118 leafsize = btrfs_super_leafsize(sb);
1119 sectorsize = btrfs_super_sectorsize(sb);
1120 stripesize = btrfs_super_stripesize(sb);
1122 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1123 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1125 ret = btrfs_read_sys_array(fs_info->chunk_root);
1129 blocksize = btrfs_level_size(fs_info->chunk_root,
1130 btrfs_super_chunk_root_level(sb));
1131 generation = btrfs_super_chunk_root_generation(sb);
1133 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1134 btrfs_super_chunk_root(sb),
1135 blocksize, generation);
1136 if (!extent_buffer_uptodate(fs_info->chunk_root->node)) {
1137 fprintf(stderr, "Couldn't read chunk root\n");
1141 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1142 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1144 fprintf(stderr, "Couldn't read chunk tree\n");
1151 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1153 u64 root_tree_bytenr,
1154 enum btrfs_open_ctree_flags flags)
1156 struct btrfs_fs_info *fs_info;
1157 struct btrfs_super_block *disk_super;
1158 struct btrfs_fs_devices *fs_devices = NULL;
1159 struct extent_buffer *eb;
1164 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1166 /* try to drop all the caches */
1167 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1168 fprintf(stderr, "Warning, could not drop caches\n");
1170 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1172 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1175 if (flags & OPEN_CTREE_RESTORE)
1176 fs_info->on_restoring = 1;
1177 if (flags & OPEN_CTREE_SUPPRESS_CHECK_BLOCK_ERRORS)
1178 fs_info->suppress_check_block_errors = 1;
1179 if (flags & OPEN_CTREE_IGNORE_FSID_MISMATCH)
1180 fs_info->ignore_fsid_mismatch = 1;
1182 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr,
1183 (flags & OPEN_CTREE_RECOVER_SUPER),
1184 (flags & OPEN_CTREE_NO_DEVICES));
1188 fs_info->fs_devices = fs_devices;
1189 if (flags & OPEN_CTREE_WRITES)
1194 if (flags & OPEN_CTREE_EXCLUSIVE)
1197 ret = btrfs_open_devices(fs_devices, oflags);
1201 disk_super = fs_info->super_copy;
1202 if (!(flags & OPEN_CTREE_RECOVER_SUPER))
1203 ret = btrfs_read_dev_super(fs_devices->latest_bdev,
1204 disk_super, sb_bytenr, 1);
1206 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr, 0);
1208 printk("No valid btrfs found\n");
1212 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_CHANGING_FSID &&
1213 !fs_info->ignore_fsid_mismatch) {
1214 fprintf(stderr, "ERROR: Filesystem UUID change in progress\n");
1218 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1220 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1221 flags & OPEN_CTREE_WRITES);
1225 ret = btrfs_setup_chunk_tree_and_device_map(fs_info);
1229 eb = fs_info->chunk_root->node;
1230 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1231 btrfs_header_chunk_tree_uuid(eb),
1234 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1235 if (ret && !(flags & __OPEN_CTREE_RETURN_CHUNK_ROOT))
1241 btrfs_release_all_roots(fs_info);
1242 btrfs_cleanup_all_caches(fs_info);
1244 btrfs_close_devices(fs_devices);
1246 btrfs_free_fs_info(fs_info);
1250 struct btrfs_fs_info *open_ctree_fs_info(const char *filename,
1251 u64 sb_bytenr, u64 root_tree_bytenr,
1252 enum btrfs_open_ctree_flags flags)
1255 struct btrfs_fs_info *info;
1256 int oflags = O_CREAT | O_RDWR;
1258 if (!(flags & OPEN_CTREE_WRITES))
1261 fp = open(filename, oflags, 0600);
1263 fprintf (stderr, "Could not open %s\n", filename);
1266 info = __open_ctree_fd(fp, filename, sb_bytenr, root_tree_bytenr,
1272 struct btrfs_root *open_ctree(const char *filename, u64 sb_bytenr,
1273 enum btrfs_open_ctree_flags flags)
1275 struct btrfs_fs_info *info;
1277 info = open_ctree_fs_info(filename, sb_bytenr, 0, flags);
1280 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1281 return info->chunk_root;
1282 return info->fs_root;
1285 struct btrfs_root *open_ctree_fd(int fp, const char *path, u64 sb_bytenr,
1286 enum btrfs_open_ctree_flags flags)
1288 struct btrfs_fs_info *info;
1289 info = __open_ctree_fd(fp, path, sb_bytenr, 0, flags);
1292 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1293 return info->chunk_root;
1294 return info->fs_root;
1297 int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
1300 u8 fsid[BTRFS_FSID_SIZE];
1301 int fsid_is_initialized = 0;
1302 struct btrfs_super_block buf;
1305 int max_super = super_recover ? BTRFS_SUPER_MIRROR_MAX : 1;
1309 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1310 ret = pread64(fd, &buf, sizeof(buf), sb_bytenr);
1311 if (ret < sizeof(buf))
1314 if (btrfs_super_bytenr(&buf) != sb_bytenr ||
1315 btrfs_super_magic(&buf) != BTRFS_MAGIC)
1318 memcpy(sb, &buf, sizeof(*sb));
1323 * we would like to check all the supers, but that would make
1324 * a btrfs mount succeed after a mkfs from a different FS.
1325 * So, we need to add a special mount option to scan for
1326 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1329 for (i = 0; i < max_super; i++) {
1330 bytenr = btrfs_sb_offset(i);
1331 ret = pread64(fd, &buf, sizeof(buf), bytenr);
1332 if (ret < sizeof(buf))
1335 if (btrfs_super_bytenr(&buf) != bytenr )
1337 /* if magic is NULL, the device was removed */
1338 if (btrfs_super_magic(&buf) == 0 && i == 0)
1340 if (btrfs_super_magic(&buf) != BTRFS_MAGIC)
1343 if (!fsid_is_initialized) {
1344 memcpy(fsid, buf.fsid, sizeof(fsid));
1345 fsid_is_initialized = 1;
1346 } else if (memcmp(fsid, buf.fsid, sizeof(fsid))) {
1348 * the superblocks (the original one and
1349 * its backups) contain data of different
1350 * filesystems -> the super cannot be trusted
1355 if (btrfs_super_generation(&buf) > transid) {
1356 memcpy(sb, &buf, sizeof(*sb));
1357 transid = btrfs_super_generation(&buf);
1361 return transid > 0 ? 0 : -1;
1364 static int write_dev_supers(struct btrfs_root *root,
1365 struct btrfs_super_block *sb,
1366 struct btrfs_device *device)
1372 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1373 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1375 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1376 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1377 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1380 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1381 * zero filled, we can use it directly
1383 ret = pwrite64(device->fd, root->fs_info->super_copy,
1384 BTRFS_SUPER_INFO_SIZE,
1385 root->fs_info->super_bytenr);
1386 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1390 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1391 bytenr = btrfs_sb_offset(i);
1392 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1395 btrfs_set_super_bytenr(sb, bytenr);
1398 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1399 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1400 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1403 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1404 * zero filled, we can use it directly
1406 ret = pwrite64(device->fd, root->fs_info->super_copy,
1407 BTRFS_SUPER_INFO_SIZE, bytenr);
1408 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1414 int write_all_supers(struct btrfs_root *root)
1416 struct list_head *cur;
1417 struct list_head *head = &root->fs_info->fs_devices->devices;
1418 struct btrfs_device *dev;
1419 struct btrfs_super_block *sb;
1420 struct btrfs_dev_item *dev_item;
1424 sb = root->fs_info->super_copy;
1425 dev_item = &sb->dev_item;
1426 list_for_each(cur, head) {
1427 dev = list_entry(cur, struct btrfs_device, dev_list);
1428 if (!dev->writeable)
1431 btrfs_set_stack_device_generation(dev_item, 0);
1432 btrfs_set_stack_device_type(dev_item, dev->type);
1433 btrfs_set_stack_device_id(dev_item, dev->devid);
1434 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1435 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1436 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1437 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1438 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1439 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1440 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1442 flags = btrfs_super_flags(sb);
1443 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1445 ret = write_dev_supers(root, sb, dev);
1451 int write_ctree_super(struct btrfs_trans_handle *trans,
1452 struct btrfs_root *root)
1455 struct btrfs_root *tree_root = root->fs_info->tree_root;
1456 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1458 if (root->fs_info->readonly)
1461 btrfs_set_super_generation(root->fs_info->super_copy,
1463 btrfs_set_super_root(root->fs_info->super_copy,
1464 tree_root->node->start);
1465 btrfs_set_super_root_level(root->fs_info->super_copy,
1466 btrfs_header_level(tree_root->node));
1467 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1468 chunk_root->node->start);
1469 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1470 btrfs_header_level(chunk_root->node));
1471 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1472 btrfs_header_generation(chunk_root->node));
1474 ret = write_all_supers(root);
1476 fprintf(stderr, "failed to write new super block err %d\n", ret);
1480 int close_ctree(struct btrfs_root *root)
1483 struct btrfs_trans_handle *trans;
1484 struct btrfs_fs_info *fs_info = root->fs_info;
1486 if (fs_info->last_trans_committed !=
1487 fs_info->generation) {
1488 trans = btrfs_start_transaction(root, 1);
1489 btrfs_commit_transaction(trans, root);
1490 trans = btrfs_start_transaction(root, 1);
1491 ret = commit_tree_roots(trans, fs_info);
1493 ret = __commit_transaction(trans, root);
1495 write_ctree_super(trans, root);
1496 btrfs_free_transaction(root, trans);
1498 btrfs_free_block_groups(fs_info);
1500 free_fs_roots_tree(&fs_info->fs_root_tree);
1502 btrfs_release_all_roots(fs_info);
1503 btrfs_close_devices(fs_info->fs_devices);
1504 btrfs_cleanup_all_caches(fs_info);
1505 btrfs_free_fs_info(fs_info);
1509 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1510 struct extent_buffer *eb)
1512 return clear_extent_buffer_dirty(eb);
1515 int wait_on_tree_block_writeback(struct btrfs_root *root,
1516 struct extent_buffer *eb)
1521 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1523 set_extent_buffer_dirty(eb);
1526 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1530 ret = extent_buffer_uptodate(buf);
1534 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1538 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1540 return set_extent_buffer_uptodate(eb);