2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #define _XOPEN_SOURCE 600
24 #include <sys/types.h>
28 #include "kerncompat.h"
29 #include "radix-tree.h"
33 #include "transaction.h"
36 #include "print-tree.h"
38 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
41 struct btrfs_fs_devices *fs_devices;
44 if (buf->start != btrfs_header_bytenr(buf)) {
45 printk("Check tree block failed, want=%Lu, have=%Lu\n",
46 buf->start, btrfs_header_bytenr(buf));
50 fs_devices = root->fs_info->fs_devices;
52 if (!memcmp_extent_buffer(buf, fs_devices->fsid,
58 fs_devices = fs_devices->seed;
63 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
65 return crc32c(seed, data, len);
68 void btrfs_csum_final(u32 crc, char *result)
70 *(__le32 *)result = ~cpu_to_le32(crc);
73 static int __csum_tree_block_size(struct extent_buffer *buf, u16 csum_size,
74 int verify, int silent)
80 result = malloc(csum_size * sizeof(char));
84 len = buf->len - BTRFS_CSUM_SIZE;
85 crc = crc32c(crc, buf->data + BTRFS_CSUM_SIZE, len);
86 btrfs_csum_final(crc, result);
89 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
91 printk("checksum verify failed on %llu found %08X wanted %08X\n",
92 (unsigned long long)buf->start,
94 *((u32*)(char *)buf->data));
99 write_extent_buffer(buf, result, 0, csum_size);
105 int csum_tree_block_size(struct extent_buffer *buf, u16 csum_size, int verify)
107 return __csum_tree_block_size(buf, csum_size, verify, 0);
110 int verify_tree_block_csum_silent(struct extent_buffer *buf, u16 csum_size)
112 return __csum_tree_block_size(buf, csum_size, 1, 1);
115 int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
119 btrfs_super_csum_size(root->fs_info->super_copy);
120 return csum_tree_block_size(buf, csum_size, verify);
123 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
124 u64 bytenr, u32 blocksize)
126 return find_extent_buffer(&root->fs_info->extent_cache,
130 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
131 u64 bytenr, u32 blocksize)
133 return alloc_extent_buffer(&root->fs_info->extent_cache, bytenr,
137 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
141 struct extent_buffer *eb;
143 struct btrfs_multi_bio *multi = NULL;
144 struct btrfs_device *device;
146 eb = btrfs_find_tree_block(root, bytenr, blocksize);
147 if (eb && btrfs_buffer_uptodate(eb, parent_transid)) {
148 free_extent_buffer(eb);
153 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
154 bytenr, &length, &multi, 0, NULL);
156 device = multi->stripes[0].dev;
158 blocksize = min(blocksize, (u32)(64 * 1024));
159 readahead(device->fd, multi->stripes[0].physical, blocksize);
164 static int verify_parent_transid(struct extent_io_tree *io_tree,
165 struct extent_buffer *eb, u64 parent_transid,
170 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
173 if (extent_buffer_uptodate(eb) &&
174 btrfs_header_generation(eb) == parent_transid) {
178 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
179 (unsigned long long)eb->start,
180 (unsigned long long)parent_transid,
181 (unsigned long long)btrfs_header_generation(eb));
183 eb->flags |= EXTENT_BAD_TRANSID;
184 printk("Ignoring transid failure\n");
190 clear_extent_buffer_uptodate(io_tree, eb);
196 int read_whole_eb(struct btrfs_fs_info *info, struct extent_buffer *eb, int mirror)
198 unsigned long offset = 0;
199 struct btrfs_multi_bio *multi = NULL;
200 struct btrfs_device *device;
203 unsigned long bytes_left = eb->len;
206 read_len = bytes_left;
209 if (!info->on_restoring) {
210 ret = btrfs_map_block(&info->mapping_tree, READ,
211 eb->start + offset, &read_len, &multi,
214 printk("Couldn't map the block %Lu\n", eb->start + offset);
218 device = multi->stripes[0].dev;
220 if (device->fd == 0) {
227 eb->dev_bytenr = multi->stripes[0].physical;
231 /* special case for restore metadump */
232 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
233 if (device->devid == 1)
238 eb->dev_bytenr = eb->start;
242 if (read_len > bytes_left)
243 read_len = bytes_left;
245 ret = read_extent_from_disk(eb, offset, read_len);
249 bytes_left -= read_len;
254 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
255 u32 blocksize, u64 parent_transid)
258 struct extent_buffer *eb;
259 u64 best_transid = 0;
265 eb = btrfs_find_create_tree_block(root, bytenr, blocksize);
269 if (btrfs_buffer_uptodate(eb, parent_transid))
273 ret = read_whole_eb(root->fs_info, eb, mirror_num);
274 if (ret == 0 && check_tree_block(root, eb) == 0 &&
275 csum_tree_block(root, eb, 1) == 0 &&
276 verify_parent_transid(eb->tree, eb, parent_transid, ignore)
278 if (eb->flags & EXTENT_BAD_TRANSID &&
279 list_empty(&eb->recow)) {
280 list_add_tail(&eb->recow,
281 &root->fs_info->recow_ebs);
284 btrfs_set_buffer_uptodate(eb);
288 if (check_tree_block(root, eb))
289 printk("read block failed check_tree_block\n");
291 printk("Csum didn't match\n");
294 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
296 if (num_copies == 1) {
300 if (btrfs_header_generation(eb) > best_transid && mirror_num) {
301 best_transid = btrfs_header_generation(eb);
302 good_mirror = mirror_num;
305 if (mirror_num > num_copies) {
306 mirror_num = good_mirror;
311 free_extent_buffer(eb);
315 int write_and_map_eb(struct btrfs_trans_handle *trans,
316 struct btrfs_root *root,
317 struct extent_buffer *eb)
322 u64 *raid_map = NULL;
323 struct btrfs_multi_bio *multi = NULL;
327 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
328 eb->start, &length, &multi, 0, &raid_map);
331 ret = write_raid56_with_parity(root->fs_info, eb, multi,
334 } else while (dev_nr < multi->num_stripes) {
336 eb->fd = multi->stripes[dev_nr].dev->fd;
337 eb->dev_bytenr = multi->stripes[dev_nr].physical;
338 multi->stripes[dev_nr].dev->total_ios++;
340 ret = write_extent_to_disk(eb);
347 static int write_tree_block(struct btrfs_trans_handle *trans,
348 struct btrfs_root *root,
349 struct extent_buffer *eb)
351 if (check_tree_block(root, eb))
354 if (!btrfs_buffer_uptodate(eb, trans->transid))
357 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
358 csum_tree_block(root, eb, 0);
360 return write_and_map_eb(trans, root, eb);
363 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
364 u32 stripesize, struct btrfs_root *root,
365 struct btrfs_fs_info *fs_info, u64 objectid)
368 root->commit_root = NULL;
369 root->sectorsize = sectorsize;
370 root->nodesize = nodesize;
371 root->leafsize = leafsize;
372 root->stripesize = stripesize;
374 root->track_dirty = 0;
376 root->fs_info = fs_info;
377 root->objectid = objectid;
378 root->last_trans = 0;
379 root->highest_inode = 0;
380 root->last_inode_alloc = 0;
382 INIT_LIST_HEAD(&root->dirty_list);
383 memset(&root->root_key, 0, sizeof(root->root_key));
384 memset(&root->root_item, 0, sizeof(root->root_item));
385 root->root_key.objectid = objectid;
389 static int update_cowonly_root(struct btrfs_trans_handle *trans,
390 struct btrfs_root *root)
394 struct btrfs_root *tree_root = root->fs_info->tree_root;
396 btrfs_write_dirty_block_groups(trans, root);
398 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
399 if (old_root_bytenr == root->node->start)
401 btrfs_set_root_bytenr(&root->root_item,
403 btrfs_set_root_generation(&root->root_item,
405 root->root_item.level = btrfs_header_level(root->node);
406 ret = btrfs_update_root(trans, tree_root,
410 btrfs_write_dirty_block_groups(trans, root);
415 static int commit_tree_roots(struct btrfs_trans_handle *trans,
416 struct btrfs_fs_info *fs_info)
418 struct btrfs_root *root;
419 struct list_head *next;
420 struct extent_buffer *eb;
423 if (fs_info->readonly)
426 eb = fs_info->tree_root->node;
427 extent_buffer_get(eb);
428 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
429 free_extent_buffer(eb);
433 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
434 next = fs_info->dirty_cowonly_roots.next;
436 root = list_entry(next, struct btrfs_root, dirty_list);
437 update_cowonly_root(trans, root);
438 free_extent_buffer(root->commit_root);
439 root->commit_root = NULL;
445 static int __commit_transaction(struct btrfs_trans_handle *trans,
446 struct btrfs_root *root)
450 struct extent_buffer *eb;
451 struct extent_io_tree *tree = &root->fs_info->extent_cache;
455 ret = find_first_extent_bit(tree, 0, &start, &end,
459 while(start <= end) {
460 eb = find_first_extent_buffer(tree, start);
461 BUG_ON(!eb || eb->start != start);
462 ret = write_tree_block(trans, root, eb);
465 clear_extent_buffer_dirty(eb);
466 free_extent_buffer(eb);
472 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
473 struct btrfs_root *root)
475 u64 transid = trans->transid;
477 struct btrfs_fs_info *fs_info = root->fs_info;
479 if (root->commit_root == root->node)
482 free_extent_buffer(root->commit_root);
483 root->commit_root = NULL;
485 btrfs_set_root_bytenr(&root->root_item, root->node->start);
486 btrfs_set_root_generation(&root->root_item, trans->transid);
487 root->root_item.level = btrfs_header_level(root->node);
488 ret = btrfs_update_root(trans, root->fs_info->tree_root,
489 &root->root_key, &root->root_item);
492 ret = commit_tree_roots(trans, fs_info);
494 ret = __commit_transaction(trans, root);
496 write_ctree_super(trans, root);
497 btrfs_finish_extent_commit(trans, fs_info->extent_root,
498 &fs_info->pinned_extents);
499 btrfs_free_transaction(root, trans);
500 free_extent_buffer(root->commit_root);
501 root->commit_root = NULL;
502 fs_info->running_transaction = NULL;
503 fs_info->last_trans_committed = transid;
507 static int find_and_setup_root(struct btrfs_root *tree_root,
508 struct btrfs_fs_info *fs_info,
509 u64 objectid, struct btrfs_root *root)
515 __setup_root(tree_root->nodesize, tree_root->leafsize,
516 tree_root->sectorsize, tree_root->stripesize,
517 root, fs_info, objectid);
518 ret = btrfs_find_last_root(tree_root, objectid,
519 &root->root_item, &root->root_key);
523 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
524 generation = btrfs_root_generation(&root->root_item);
525 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
526 blocksize, generation);
527 if (!extent_buffer_uptodate(root->node))
533 static int find_and_setup_log_root(struct btrfs_root *tree_root,
534 struct btrfs_fs_info *fs_info,
535 struct btrfs_super_block *disk_super)
538 u64 blocknr = btrfs_super_log_root(disk_super);
539 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
549 blocksize = btrfs_level_size(tree_root,
550 btrfs_super_log_root_level(disk_super));
552 __setup_root(tree_root->nodesize, tree_root->leafsize,
553 tree_root->sectorsize, tree_root->stripesize,
554 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
556 log_root->node = read_tree_block(tree_root, blocknr,
558 btrfs_super_generation(disk_super) + 1);
560 fs_info->log_root_tree = log_root;
562 if (!extent_buffer_uptodate(log_root->node)) {
563 free_extent_buffer(log_root->node);
565 fs_info->log_root_tree = NULL;
572 int btrfs_free_fs_root(struct btrfs_root *root)
575 free_extent_buffer(root->node);
576 if (root->commit_root)
577 free_extent_buffer(root->commit_root);
582 static void __free_fs_root(struct rb_node *node)
584 struct btrfs_root *root;
586 root = container_of(node, struct btrfs_root, rb_node);
587 btrfs_free_fs_root(root);
590 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
592 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
593 struct btrfs_key *location)
595 struct btrfs_root *root;
596 struct btrfs_root *tree_root = fs_info->tree_root;
597 struct btrfs_path *path;
598 struct extent_buffer *l;
603 root = malloc(sizeof(*root));
605 return ERR_PTR(-ENOMEM);
606 memset(root, 0, sizeof(*root));
607 if (location->offset == (u64)-1) {
608 ret = find_and_setup_root(tree_root, fs_info,
609 location->objectid, root);
617 __setup_root(tree_root->nodesize, tree_root->leafsize,
618 tree_root->sectorsize, tree_root->stripesize,
619 root, fs_info, location->objectid);
621 path = btrfs_alloc_path();
623 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
630 read_extent_buffer(l, &root->root_item,
631 btrfs_item_ptr_offset(l, path->slots[0]),
632 sizeof(root->root_item));
633 memcpy(&root->root_key, location, sizeof(*location));
636 btrfs_release_path(path);
637 btrfs_free_path(path);
642 generation = btrfs_root_generation(&root->root_item);
643 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
644 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
645 blocksize, generation);
648 return ERR_PTR(-EIO);
655 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
658 u64 objectid = *((u64 *)data);
659 struct btrfs_root *root;
661 root = rb_entry(node, struct btrfs_root, rb_node);
662 if (objectid > root->objectid)
664 else if (objectid < root->objectid)
670 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
671 struct rb_node *node2)
673 struct btrfs_root *root;
675 root = rb_entry(node2, struct btrfs_root, rb_node);
676 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
679 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
680 struct btrfs_key *location)
682 struct btrfs_root *root;
683 struct rb_node *node;
685 u64 objectid = location->objectid;
687 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
688 return fs_info->tree_root;
689 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
690 return fs_info->extent_root;
691 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
692 return fs_info->chunk_root;
693 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
694 return fs_info->dev_root;
695 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
696 return fs_info->csum_root;
697 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
698 return fs_info->csum_root;
700 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
701 location->offset != (u64)-1);
703 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
704 btrfs_fs_roots_compare_objectids, NULL);
706 return container_of(node, struct btrfs_root, rb_node);
708 root = btrfs_read_fs_root_no_cache(fs_info, location);
712 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
713 btrfs_fs_roots_compare_roots);
718 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
720 free(fs_info->tree_root);
721 free(fs_info->extent_root);
722 free(fs_info->chunk_root);
723 free(fs_info->dev_root);
724 free(fs_info->csum_root);
725 free(fs_info->quota_root);
726 free(fs_info->super_copy);
727 free(fs_info->log_root_tree);
731 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
733 struct btrfs_fs_info *fs_info;
735 fs_info = malloc(sizeof(struct btrfs_fs_info));
739 memset(fs_info, 0, sizeof(struct btrfs_fs_info));
741 fs_info->tree_root = malloc(sizeof(struct btrfs_root));
742 fs_info->extent_root = malloc(sizeof(struct btrfs_root));
743 fs_info->chunk_root = malloc(sizeof(struct btrfs_root));
744 fs_info->dev_root = malloc(sizeof(struct btrfs_root));
745 fs_info->csum_root = malloc(sizeof(struct btrfs_root));
746 fs_info->quota_root = malloc(sizeof(struct btrfs_root));
747 fs_info->super_copy = malloc(BTRFS_SUPER_INFO_SIZE);
749 if (!fs_info->tree_root || !fs_info->extent_root ||
750 !fs_info->chunk_root || !fs_info->dev_root ||
751 !fs_info->csum_root || !fs_info->quota_root ||
752 !fs_info->super_copy)
755 memset(fs_info->super_copy, 0, BTRFS_SUPER_INFO_SIZE);
756 memset(fs_info->tree_root, 0, sizeof(struct btrfs_root));
757 memset(fs_info->extent_root, 0, sizeof(struct btrfs_root));
758 memset(fs_info->chunk_root, 0, sizeof(struct btrfs_root));
759 memset(fs_info->dev_root, 0, sizeof(struct btrfs_root));
760 memset(fs_info->csum_root, 0, sizeof(struct btrfs_root));
761 memset(fs_info->quota_root, 0, sizeof(struct btrfs_root));
763 extent_io_tree_init(&fs_info->extent_cache);
764 extent_io_tree_init(&fs_info->free_space_cache);
765 extent_io_tree_init(&fs_info->block_group_cache);
766 extent_io_tree_init(&fs_info->pinned_extents);
767 extent_io_tree_init(&fs_info->pending_del);
768 extent_io_tree_init(&fs_info->extent_ins);
769 fs_info->fs_root_tree = RB_ROOT;
770 cache_tree_init(&fs_info->mapping_tree.cache_tree);
772 mutex_init(&fs_info->fs_mutex);
773 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
774 INIT_LIST_HEAD(&fs_info->space_info);
775 INIT_LIST_HEAD(&fs_info->recow_ebs);
778 fs_info->readonly = 1;
780 fs_info->super_bytenr = sb_bytenr;
781 fs_info->data_alloc_profile = (u64)-1;
782 fs_info->metadata_alloc_profile = (u64)-1;
783 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
786 btrfs_free_fs_info(fs_info);
790 int btrfs_check_fs_compatibility(struct btrfs_super_block *sb, int writable)
794 features = btrfs_super_incompat_flags(sb) &
795 ~BTRFS_FEATURE_INCOMPAT_SUPP;
797 printk("couldn't open because of unsupported "
798 "option features (%Lx).\n",
799 (unsigned long long)features);
803 features = btrfs_super_incompat_flags(sb);
804 if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
805 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
806 btrfs_set_super_incompat_flags(sb, features);
809 features = btrfs_super_compat_ro_flags(sb) &
810 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
811 if (writable && features) {
812 printk("couldn't open RDWR because of unsupported "
813 "option features (%Lx).\n",
814 (unsigned long long)features);
820 static int find_best_backup_root(struct btrfs_super_block *super)
822 struct btrfs_root_backup *backup;
823 u64 orig_gen = btrfs_super_generation(super);
828 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
829 backup = super->super_roots + i;
830 if (btrfs_backup_tree_root_gen(backup) != orig_gen &&
831 btrfs_backup_tree_root_gen(backup) > gen) {
833 gen = btrfs_backup_tree_root_gen(backup);
839 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
840 enum btrfs_open_ctree_flags flags)
842 struct btrfs_super_block *sb = fs_info->super_copy;
843 struct btrfs_root *root;
844 struct btrfs_key key;
853 nodesize = btrfs_super_nodesize(sb);
854 leafsize = btrfs_super_leafsize(sb);
855 sectorsize = btrfs_super_sectorsize(sb);
856 stripesize = btrfs_super_stripesize(sb);
858 root = fs_info->tree_root;
859 __setup_root(nodesize, leafsize, sectorsize, stripesize,
860 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
861 blocksize = btrfs_level_size(root, btrfs_super_root_level(sb));
862 generation = btrfs_super_generation(sb);
864 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
865 root_tree_bytenr = btrfs_super_root(sb);
866 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
867 struct btrfs_root_backup *backup;
868 int index = find_best_backup_root(sb);
869 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
870 fprintf(stderr, "Invalid backup root number\n");
873 backup = fs_info->super_copy->super_roots + index;
874 root_tree_bytenr = btrfs_backup_tree_root(backup);
875 generation = btrfs_backup_tree_root_gen(backup);
878 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
880 if (!extent_buffer_uptodate(root->node)) {
881 fprintf(stderr, "Couldn't read tree root\n");
885 ret = find_and_setup_root(root, fs_info, BTRFS_EXTENT_TREE_OBJECTID,
886 fs_info->extent_root);
888 printk("Couldn't setup extent tree\n");
889 if (!(flags & OPEN_CTREE_PARTIAL))
891 /* Need a blank node here just so we don't screw up in the
892 * million of places that assume a root has a valid ->node
894 fs_info->extent_root->node =
895 btrfs_find_create_tree_block(fs_info->extent_root, 0,
897 if (!fs_info->extent_root->node)
899 clear_extent_buffer_uptodate(NULL, fs_info->extent_root->node);
901 fs_info->extent_root->track_dirty = 1;
903 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
906 printk("Couldn't setup device tree\n");
909 fs_info->dev_root->track_dirty = 1;
911 ret = find_and_setup_root(root, fs_info, BTRFS_CSUM_TREE_OBJECTID,
914 printk("Couldn't setup csum tree\n");
915 if (!(flags & OPEN_CTREE_PARTIAL))
918 fs_info->csum_root->track_dirty = 1;
920 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
921 fs_info->quota_root);
923 fs_info->quota_enabled = 1;
925 ret = find_and_setup_log_root(root, fs_info, sb);
927 printk("Couldn't setup log root tree\n");
931 fs_info->generation = generation;
932 fs_info->last_trans_committed = generation;
933 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
934 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
935 btrfs_read_block_groups(fs_info->tree_root);
937 key.objectid = BTRFS_FS_TREE_OBJECTID;
938 key.type = BTRFS_ROOT_ITEM_KEY;
939 key.offset = (u64)-1;
940 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
942 if (IS_ERR(fs_info->fs_root))
947 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
949 if (fs_info->quota_root)
950 free_extent_buffer(fs_info->quota_root->node);
951 if (fs_info->csum_root)
952 free_extent_buffer(fs_info->csum_root->node);
953 if (fs_info->dev_root)
954 free_extent_buffer(fs_info->dev_root->node);
955 if (fs_info->extent_root)
956 free_extent_buffer(fs_info->extent_root->node);
957 if (fs_info->tree_root)
958 free_extent_buffer(fs_info->tree_root->node);
959 if (fs_info->log_root_tree)
960 free_extent_buffer(fs_info->log_root_tree->node);
961 if (fs_info->chunk_root)
962 free_extent_buffer(fs_info->chunk_root->node);
965 static void free_map_lookup(struct cache_extent *ce)
967 struct map_lookup *map;
969 map = container_of(ce, struct map_lookup, ce);
973 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
975 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
977 while (!list_empty(&fs_info->recow_ebs)) {
978 struct extent_buffer *eb;
979 eb = list_first_entry(&fs_info->recow_ebs,
980 struct extent_buffer, recow);
981 list_del_init(&eb->recow);
982 free_extent_buffer(eb);
984 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
985 extent_io_tree_cleanup(&fs_info->extent_cache);
986 extent_io_tree_cleanup(&fs_info->free_space_cache);
987 extent_io_tree_cleanup(&fs_info->block_group_cache);
988 extent_io_tree_cleanup(&fs_info->pinned_extents);
989 extent_io_tree_cleanup(&fs_info->pending_del);
990 extent_io_tree_cleanup(&fs_info->extent_ins);
993 int btrfs_scan_fs_devices(int fd, const char *path,
994 struct btrfs_fs_devices **fs_devices,
995 u64 sb_bytenr, int run_ioctl)
1000 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1002 ret = btrfs_scan_one_device(fd, path, fs_devices,
1003 &total_devs, sb_bytenr);
1005 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1009 if (total_devs != 1) {
1010 ret = btrfs_scan_for_fsid(run_ioctl);
1017 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info)
1019 struct btrfs_super_block *sb = fs_info->super_copy;
1028 nodesize = btrfs_super_nodesize(sb);
1029 leafsize = btrfs_super_leafsize(sb);
1030 sectorsize = btrfs_super_sectorsize(sb);
1031 stripesize = btrfs_super_stripesize(sb);
1033 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1034 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1036 ret = btrfs_read_sys_array(fs_info->chunk_root);
1040 blocksize = btrfs_level_size(fs_info->chunk_root,
1041 btrfs_super_chunk_root_level(sb));
1042 generation = btrfs_super_chunk_root_generation(sb);
1044 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1045 btrfs_super_chunk_root(sb),
1046 blocksize, generation);
1047 if (!fs_info->chunk_root->node ||
1048 !extent_buffer_uptodate(fs_info->chunk_root->node)) {
1049 fprintf(stderr, "Couldn't read chunk root\n");
1053 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1054 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1056 fprintf(stderr, "Couldn't read chunk tree\n");
1063 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1065 u64 root_tree_bytenr,
1066 enum btrfs_open_ctree_flags flags)
1068 struct btrfs_fs_info *fs_info;
1069 struct btrfs_super_block *disk_super;
1070 struct btrfs_fs_devices *fs_devices = NULL;
1071 struct extent_buffer *eb;
1076 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1078 /* try to drop all the caches */
1079 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1080 fprintf(stderr, "Warning, could not drop caches\n");
1082 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1084 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1087 if (flags & OPEN_CTREE_RESTORE)
1088 fs_info->on_restoring = 1;
1090 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr,
1091 !(flags & OPEN_CTREE_RECOVER_SUPER));
1095 fs_info->fs_devices = fs_devices;
1096 if (flags & OPEN_CTREE_WRITES)
1101 if (flags & OPEN_CTREE_EXCLUSIVE)
1104 ret = btrfs_open_devices(fs_devices, oflags);
1108 disk_super = fs_info->super_copy;
1109 if (!(flags & OPEN_CTREE_RECOVER_SUPER))
1110 ret = btrfs_read_dev_super(fs_devices->latest_bdev,
1111 disk_super, sb_bytenr);
1113 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr);
1115 printk("No valid btrfs found\n");
1119 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1121 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1122 flags & OPEN_CTREE_WRITES);
1126 ret = btrfs_setup_chunk_tree_and_device_map(fs_info);
1130 eb = fs_info->chunk_root->node;
1131 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1132 btrfs_header_chunk_tree_uuid(eb),
1135 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1142 if (flags & OPEN_CTREE_PARTIAL)
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)
1199 u8 fsid[BTRFS_FSID_SIZE];
1200 int fsid_is_initialized = 0;
1201 struct btrfs_super_block buf;
1207 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1208 ret = pread64(fd, &buf, sizeof(buf), sb_bytenr);
1209 if (ret < sizeof(buf))
1212 if (btrfs_super_bytenr(&buf) != sb_bytenr ||
1213 btrfs_super_magic(&buf) != BTRFS_MAGIC)
1216 memcpy(sb, &buf, sizeof(*sb));
1221 * we would like to check all the supers, but that would make
1222 * a btrfs mount succeed after a mkfs from a different FS.
1223 * So, we need to add a special mount option to scan for
1224 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1227 for (i = 0; i < 1; i++) {
1228 bytenr = btrfs_sb_offset(i);
1229 ret = pread64(fd, &buf, sizeof(buf), bytenr);
1230 if (ret < sizeof(buf))
1233 if (btrfs_super_bytenr(&buf) != bytenr )
1235 /* if magic is NULL, the device was removed */
1236 if (btrfs_super_magic(&buf) == 0 && i == 0)
1238 if (btrfs_super_magic(&buf) != BTRFS_MAGIC)
1241 if (!fsid_is_initialized) {
1242 memcpy(fsid, buf.fsid, sizeof(fsid));
1243 fsid_is_initialized = 1;
1244 } else if (memcmp(fsid, buf.fsid, sizeof(fsid))) {
1246 * the superblocks (the original one and
1247 * its backups) contain data of different
1248 * filesystems -> the super cannot be trusted
1253 if (btrfs_super_generation(&buf) > transid) {
1254 memcpy(sb, &buf, sizeof(*sb));
1255 transid = btrfs_super_generation(&buf);
1259 return transid > 0 ? 0 : -1;
1262 static int write_dev_supers(struct btrfs_root *root,
1263 struct btrfs_super_block *sb,
1264 struct btrfs_device *device)
1270 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1271 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1273 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1274 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1275 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1278 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1279 * zero filled, we can use it directly
1281 ret = pwrite64(device->fd, root->fs_info->super_copy,
1282 BTRFS_SUPER_INFO_SIZE,
1283 root->fs_info->super_bytenr);
1284 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1288 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1289 bytenr = btrfs_sb_offset(i);
1290 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1293 btrfs_set_super_bytenr(sb, bytenr);
1296 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1297 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1298 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1301 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1302 * zero filled, we can use it directly
1304 ret = pwrite64(device->fd, root->fs_info->super_copy,
1305 BTRFS_SUPER_INFO_SIZE, bytenr);
1306 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1312 int write_all_supers(struct btrfs_root *root)
1314 struct list_head *cur;
1315 struct list_head *head = &root->fs_info->fs_devices->devices;
1316 struct btrfs_device *dev;
1317 struct btrfs_super_block *sb;
1318 struct btrfs_dev_item *dev_item;
1322 sb = root->fs_info->super_copy;
1323 dev_item = &sb->dev_item;
1324 list_for_each(cur, head) {
1325 dev = list_entry(cur, struct btrfs_device, dev_list);
1326 if (!dev->writeable)
1329 btrfs_set_stack_device_generation(dev_item, 0);
1330 btrfs_set_stack_device_type(dev_item, dev->type);
1331 btrfs_set_stack_device_id(dev_item, dev->devid);
1332 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1333 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1334 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1335 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1336 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1337 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1338 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1340 flags = btrfs_super_flags(sb);
1341 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1343 ret = write_dev_supers(root, sb, dev);
1349 int write_ctree_super(struct btrfs_trans_handle *trans,
1350 struct btrfs_root *root)
1353 struct btrfs_root *tree_root = root->fs_info->tree_root;
1354 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1356 if (root->fs_info->readonly)
1359 btrfs_set_super_generation(root->fs_info->super_copy,
1361 btrfs_set_super_root(root->fs_info->super_copy,
1362 tree_root->node->start);
1363 btrfs_set_super_root_level(root->fs_info->super_copy,
1364 btrfs_header_level(tree_root->node));
1365 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1366 chunk_root->node->start);
1367 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1368 btrfs_header_level(chunk_root->node));
1369 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1370 btrfs_header_generation(chunk_root->node));
1372 ret = write_all_supers(root);
1374 fprintf(stderr, "failed to write new super block err %d\n", ret);
1378 int close_ctree(struct btrfs_root *root)
1381 struct btrfs_trans_handle *trans;
1382 struct btrfs_fs_info *fs_info = root->fs_info;
1384 if (fs_info->last_trans_committed !=
1385 fs_info->generation) {
1386 trans = btrfs_start_transaction(root, 1);
1387 btrfs_commit_transaction(trans, root);
1388 trans = btrfs_start_transaction(root, 1);
1389 ret = commit_tree_roots(trans, fs_info);
1391 ret = __commit_transaction(trans, root);
1393 write_ctree_super(trans, root);
1394 btrfs_free_transaction(root, trans);
1396 btrfs_free_block_groups(fs_info);
1398 free_fs_roots_tree(&fs_info->fs_root_tree);
1400 btrfs_release_all_roots(fs_info);
1401 btrfs_close_devices(fs_info->fs_devices);
1402 btrfs_cleanup_all_caches(fs_info);
1403 btrfs_free_fs_info(fs_info);
1407 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1408 struct extent_buffer *eb)
1410 return clear_extent_buffer_dirty(eb);
1413 int wait_on_tree_block_writeback(struct btrfs_root *root,
1414 struct extent_buffer *eb)
1419 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1421 set_extent_buffer_dirty(eb);
1424 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1428 ret = extent_buffer_uptodate(buf);
1432 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1436 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1438 return set_extent_buffer_uptodate(eb);