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 read_extent_data(struct btrfs_root *root, char *data,
370 u64 logical, u64 *len, int mirror)
373 struct btrfs_multi_bio *multi = NULL;
374 struct btrfs_fs_info *info = root->fs_info;
375 struct btrfs_device *device;
379 ret = btrfs_map_block(&info->mapping_tree, READ, logical, len,
380 &multi, mirror, NULL);
382 fprintf(stderr, "Couldn't map the block %llu\n",
386 device = multi->stripes[0].dev;
393 ret = pread64(device->fd, data, *len, multi->stripes[0].physical);
403 int write_and_map_eb(struct btrfs_trans_handle *trans,
404 struct btrfs_root *root,
405 struct extent_buffer *eb)
410 u64 *raid_map = NULL;
411 struct btrfs_multi_bio *multi = NULL;
415 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
416 eb->start, &length, &multi, 0, &raid_map);
419 ret = write_raid56_with_parity(root->fs_info, eb, multi,
422 } else while (dev_nr < multi->num_stripes) {
424 eb->fd = multi->stripes[dev_nr].dev->fd;
425 eb->dev_bytenr = multi->stripes[dev_nr].physical;
426 multi->stripes[dev_nr].dev->total_ios++;
428 ret = write_extent_to_disk(eb);
435 int write_tree_block(struct btrfs_trans_handle *trans,
436 struct btrfs_root *root,
437 struct extent_buffer *eb)
439 if (check_tree_block(root, eb)) {
440 print_tree_block_error(root, eb, check_tree_block(root, eb));
444 if (trans && !btrfs_buffer_uptodate(eb, trans->transid))
447 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
448 csum_tree_block(root, eb, 0);
450 return write_and_map_eb(trans, root, eb);
453 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
454 u32 stripesize, struct btrfs_root *root,
455 struct btrfs_fs_info *fs_info, u64 objectid)
458 root->commit_root = NULL;
459 root->sectorsize = sectorsize;
460 root->nodesize = nodesize;
461 root->leafsize = leafsize;
462 root->stripesize = stripesize;
464 root->track_dirty = 0;
466 root->fs_info = fs_info;
467 root->objectid = objectid;
468 root->last_trans = 0;
469 root->highest_inode = 0;
470 root->last_inode_alloc = 0;
472 INIT_LIST_HEAD(&root->dirty_list);
473 INIT_LIST_HEAD(&root->orphan_data_extents);
474 memset(&root->root_key, 0, sizeof(root->root_key));
475 memset(&root->root_item, 0, sizeof(root->root_item));
476 root->root_key.objectid = objectid;
480 static int update_cowonly_root(struct btrfs_trans_handle *trans,
481 struct btrfs_root *root)
485 struct btrfs_root *tree_root = root->fs_info->tree_root;
487 btrfs_write_dirty_block_groups(trans, root);
489 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
490 if (old_root_bytenr == root->node->start)
492 btrfs_set_root_bytenr(&root->root_item,
494 btrfs_set_root_generation(&root->root_item,
496 root->root_item.level = btrfs_header_level(root->node);
497 ret = btrfs_update_root(trans, tree_root,
501 btrfs_write_dirty_block_groups(trans, root);
506 static int commit_tree_roots(struct btrfs_trans_handle *trans,
507 struct btrfs_fs_info *fs_info)
509 struct btrfs_root *root;
510 struct list_head *next;
511 struct extent_buffer *eb;
514 if (fs_info->readonly)
517 eb = fs_info->tree_root->node;
518 extent_buffer_get(eb);
519 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
520 free_extent_buffer(eb);
524 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
525 next = fs_info->dirty_cowonly_roots.next;
527 root = list_entry(next, struct btrfs_root, dirty_list);
528 update_cowonly_root(trans, root);
529 free_extent_buffer(root->commit_root);
530 root->commit_root = NULL;
536 static int __commit_transaction(struct btrfs_trans_handle *trans,
537 struct btrfs_root *root)
541 struct extent_buffer *eb;
542 struct extent_io_tree *tree = &root->fs_info->extent_cache;
546 ret = find_first_extent_bit(tree, 0, &start, &end,
550 while(start <= end) {
551 eb = find_first_extent_buffer(tree, start);
552 BUG_ON(!eb || eb->start != start);
553 ret = write_tree_block(trans, root, eb);
556 clear_extent_buffer_dirty(eb);
557 free_extent_buffer(eb);
563 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
564 struct btrfs_root *root)
566 u64 transid = trans->transid;
568 struct btrfs_fs_info *fs_info = root->fs_info;
570 if (root->commit_root == root->node)
572 if (root == root->fs_info->tree_root)
575 free_extent_buffer(root->commit_root);
576 root->commit_root = NULL;
578 btrfs_set_root_bytenr(&root->root_item, root->node->start);
579 btrfs_set_root_generation(&root->root_item, trans->transid);
580 root->root_item.level = btrfs_header_level(root->node);
581 ret = btrfs_update_root(trans, root->fs_info->tree_root,
582 &root->root_key, &root->root_item);
585 ret = commit_tree_roots(trans, fs_info);
587 ret = __commit_transaction(trans, root);
589 write_ctree_super(trans, root);
590 btrfs_finish_extent_commit(trans, fs_info->extent_root,
591 &fs_info->pinned_extents);
592 btrfs_free_transaction(root, trans);
593 free_extent_buffer(root->commit_root);
594 root->commit_root = NULL;
595 fs_info->running_transaction = NULL;
596 fs_info->last_trans_committed = transid;
600 static int find_and_setup_root(struct btrfs_root *tree_root,
601 struct btrfs_fs_info *fs_info,
602 u64 objectid, struct btrfs_root *root)
608 __setup_root(tree_root->nodesize, tree_root->leafsize,
609 tree_root->sectorsize, tree_root->stripesize,
610 root, fs_info, objectid);
611 ret = btrfs_find_last_root(tree_root, objectid,
612 &root->root_item, &root->root_key);
616 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
617 generation = btrfs_root_generation(&root->root_item);
618 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
619 blocksize, generation);
620 if (!extent_buffer_uptodate(root->node))
626 static int find_and_setup_log_root(struct btrfs_root *tree_root,
627 struct btrfs_fs_info *fs_info,
628 struct btrfs_super_block *disk_super)
631 u64 blocknr = btrfs_super_log_root(disk_super);
632 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
642 blocksize = btrfs_level_size(tree_root,
643 btrfs_super_log_root_level(disk_super));
645 __setup_root(tree_root->nodesize, tree_root->leafsize,
646 tree_root->sectorsize, tree_root->stripesize,
647 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
649 log_root->node = read_tree_block(tree_root, blocknr,
651 btrfs_super_generation(disk_super) + 1);
653 fs_info->log_root_tree = log_root;
655 if (!extent_buffer_uptodate(log_root->node)) {
656 free_extent_buffer(log_root->node);
658 fs_info->log_root_tree = NULL;
665 int btrfs_free_fs_root(struct btrfs_root *root)
668 free_extent_buffer(root->node);
669 if (root->commit_root)
670 free_extent_buffer(root->commit_root);
675 static void __free_fs_root(struct rb_node *node)
677 struct btrfs_root *root;
679 root = container_of(node, struct btrfs_root, rb_node);
680 btrfs_free_fs_root(root);
683 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
685 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
686 struct btrfs_key *location)
688 struct btrfs_root *root;
689 struct btrfs_root *tree_root = fs_info->tree_root;
690 struct btrfs_path *path;
691 struct extent_buffer *l;
696 root = malloc(sizeof(*root));
698 return ERR_PTR(-ENOMEM);
699 memset(root, 0, sizeof(*root));
700 if (location->offset == (u64)-1) {
701 ret = find_and_setup_root(tree_root, fs_info,
702 location->objectid, root);
710 __setup_root(tree_root->nodesize, tree_root->leafsize,
711 tree_root->sectorsize, tree_root->stripesize,
712 root, fs_info, location->objectid);
714 path = btrfs_alloc_path();
716 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
723 read_extent_buffer(l, &root->root_item,
724 btrfs_item_ptr_offset(l, path->slots[0]),
725 sizeof(root->root_item));
726 memcpy(&root->root_key, location, sizeof(*location));
729 btrfs_free_path(path);
734 generation = btrfs_root_generation(&root->root_item);
735 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
736 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
737 blocksize, generation);
738 if (!extent_buffer_uptodate(root->node)) {
740 return ERR_PTR(-EIO);
747 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
750 u64 objectid = *((u64 *)data);
751 struct btrfs_root *root;
753 root = rb_entry(node, struct btrfs_root, rb_node);
754 if (objectid > root->objectid)
756 else if (objectid < root->objectid)
762 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
763 struct rb_node *node2)
765 struct btrfs_root *root;
767 root = rb_entry(node2, struct btrfs_root, rb_node);
768 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
771 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
772 struct btrfs_key *location)
774 struct btrfs_root *root;
775 struct rb_node *node;
777 u64 objectid = location->objectid;
779 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
780 return fs_info->tree_root;
781 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
782 return fs_info->extent_root;
783 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
784 return fs_info->chunk_root;
785 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
786 return fs_info->dev_root;
787 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
788 return fs_info->csum_root;
789 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
790 return fs_info->quota_root;
792 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
793 location->offset != (u64)-1);
795 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
796 btrfs_fs_roots_compare_objectids, NULL);
798 return container_of(node, struct btrfs_root, rb_node);
800 root = btrfs_read_fs_root_no_cache(fs_info, location);
804 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
805 btrfs_fs_roots_compare_roots);
810 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
812 free(fs_info->tree_root);
813 free(fs_info->extent_root);
814 free(fs_info->chunk_root);
815 free(fs_info->dev_root);
816 free(fs_info->csum_root);
817 free(fs_info->quota_root);
818 free(fs_info->super_copy);
819 free(fs_info->log_root_tree);
823 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
825 struct btrfs_fs_info *fs_info;
827 fs_info = malloc(sizeof(struct btrfs_fs_info));
831 memset(fs_info, 0, sizeof(struct btrfs_fs_info));
833 fs_info->tree_root = malloc(sizeof(struct btrfs_root));
834 fs_info->extent_root = malloc(sizeof(struct btrfs_root));
835 fs_info->chunk_root = malloc(sizeof(struct btrfs_root));
836 fs_info->dev_root = malloc(sizeof(struct btrfs_root));
837 fs_info->csum_root = malloc(sizeof(struct btrfs_root));
838 fs_info->quota_root = malloc(sizeof(struct btrfs_root));
839 fs_info->super_copy = malloc(BTRFS_SUPER_INFO_SIZE);
841 if (!fs_info->tree_root || !fs_info->extent_root ||
842 !fs_info->chunk_root || !fs_info->dev_root ||
843 !fs_info->csum_root || !fs_info->quota_root ||
844 !fs_info->super_copy)
847 memset(fs_info->super_copy, 0, BTRFS_SUPER_INFO_SIZE);
848 memset(fs_info->tree_root, 0, sizeof(struct btrfs_root));
849 memset(fs_info->extent_root, 0, sizeof(struct btrfs_root));
850 memset(fs_info->chunk_root, 0, sizeof(struct btrfs_root));
851 memset(fs_info->dev_root, 0, sizeof(struct btrfs_root));
852 memset(fs_info->csum_root, 0, sizeof(struct btrfs_root));
853 memset(fs_info->quota_root, 0, sizeof(struct btrfs_root));
855 extent_io_tree_init(&fs_info->extent_cache);
856 extent_io_tree_init(&fs_info->free_space_cache);
857 extent_io_tree_init(&fs_info->block_group_cache);
858 extent_io_tree_init(&fs_info->pinned_extents);
859 extent_io_tree_init(&fs_info->pending_del);
860 extent_io_tree_init(&fs_info->extent_ins);
861 fs_info->excluded_extents = NULL;
863 fs_info->fs_root_tree = RB_ROOT;
864 cache_tree_init(&fs_info->mapping_tree.cache_tree);
866 mutex_init(&fs_info->fs_mutex);
867 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
868 INIT_LIST_HEAD(&fs_info->space_info);
869 INIT_LIST_HEAD(&fs_info->recow_ebs);
872 fs_info->readonly = 1;
874 fs_info->super_bytenr = sb_bytenr;
875 fs_info->data_alloc_profile = (u64)-1;
876 fs_info->metadata_alloc_profile = (u64)-1;
877 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
880 btrfs_free_fs_info(fs_info);
884 int btrfs_check_fs_compatibility(struct btrfs_super_block *sb, int writable)
888 features = btrfs_super_incompat_flags(sb) &
889 ~BTRFS_FEATURE_INCOMPAT_SUPP;
891 printk("couldn't open because of unsupported "
892 "option features (%Lx).\n",
893 (unsigned long long)features);
897 features = btrfs_super_incompat_flags(sb);
898 if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
899 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
900 btrfs_set_super_incompat_flags(sb, features);
903 features = btrfs_super_compat_ro_flags(sb) &
904 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
905 if (writable && features) {
906 printk("couldn't open RDWR because of unsupported "
907 "option features (%Lx).\n",
908 (unsigned long long)features);
914 static int find_best_backup_root(struct btrfs_super_block *super)
916 struct btrfs_root_backup *backup;
917 u64 orig_gen = btrfs_super_generation(super);
922 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
923 backup = super->super_roots + i;
924 if (btrfs_backup_tree_root_gen(backup) != orig_gen &&
925 btrfs_backup_tree_root_gen(backup) > gen) {
927 gen = btrfs_backup_tree_root_gen(backup);
933 static int setup_root_or_create_block(struct btrfs_fs_info *fs_info,
934 enum btrfs_open_ctree_flags flags,
935 struct btrfs_root *info_root,
936 u64 objectid, char *str)
938 struct btrfs_super_block *sb = fs_info->super_copy;
939 struct btrfs_root *root = fs_info->tree_root;
940 u32 leafsize = btrfs_super_leafsize(sb);
943 ret = find_and_setup_root(root, fs_info, objectid, info_root);
945 printk("Couldn't setup %s tree\n", str);
946 if (!(flags & OPEN_CTREE_PARTIAL))
949 * Need a blank node here just so we don't screw up in the
950 * million of places that assume a root has a valid ->node
953 btrfs_find_create_tree_block(info_root, 0, leafsize);
954 if (!info_root->node)
956 clear_extent_buffer_uptodate(NULL, info_root->node);
962 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
963 enum btrfs_open_ctree_flags flags)
965 struct btrfs_super_block *sb = fs_info->super_copy;
966 struct btrfs_root *root;
967 struct btrfs_key key;
976 nodesize = btrfs_super_nodesize(sb);
977 leafsize = btrfs_super_leafsize(sb);
978 sectorsize = btrfs_super_sectorsize(sb);
979 stripesize = btrfs_super_stripesize(sb);
981 root = fs_info->tree_root;
982 __setup_root(nodesize, leafsize, sectorsize, stripesize,
983 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
984 blocksize = btrfs_level_size(root, btrfs_super_root_level(sb));
985 generation = btrfs_super_generation(sb);
987 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
988 root_tree_bytenr = btrfs_super_root(sb);
989 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
990 struct btrfs_root_backup *backup;
991 int index = find_best_backup_root(sb);
992 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
993 fprintf(stderr, "Invalid backup root number\n");
996 backup = fs_info->super_copy->super_roots + index;
997 root_tree_bytenr = btrfs_backup_tree_root(backup);
998 generation = btrfs_backup_tree_root_gen(backup);
1001 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
1003 if (!extent_buffer_uptodate(root->node)) {
1004 fprintf(stderr, "Couldn't read tree root\n");
1008 ret = setup_root_or_create_block(fs_info, flags, fs_info->extent_root,
1009 BTRFS_EXTENT_TREE_OBJECTID, "extent");
1012 fs_info->extent_root->track_dirty = 1;
1014 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
1017 printk("Couldn't setup device tree\n");
1020 fs_info->dev_root->track_dirty = 1;
1022 ret = setup_root_or_create_block(fs_info, flags, fs_info->csum_root,
1023 BTRFS_CSUM_TREE_OBJECTID, "csum");
1026 fs_info->csum_root->track_dirty = 1;
1028 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
1029 fs_info->quota_root);
1031 fs_info->quota_enabled = 1;
1033 ret = find_and_setup_log_root(root, fs_info, sb);
1035 printk("Couldn't setup log root tree\n");
1036 if (!(flags & OPEN_CTREE_PARTIAL))
1040 fs_info->generation = generation;
1041 fs_info->last_trans_committed = generation;
1042 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
1043 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
1044 btrfs_read_block_groups(fs_info->tree_root);
1046 key.objectid = BTRFS_FS_TREE_OBJECTID;
1047 key.type = BTRFS_ROOT_ITEM_KEY;
1048 key.offset = (u64)-1;
1049 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
1051 if (IS_ERR(fs_info->fs_root))
1056 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
1058 if (fs_info->quota_root)
1059 free_extent_buffer(fs_info->quota_root->node);
1060 if (fs_info->csum_root)
1061 free_extent_buffer(fs_info->csum_root->node);
1062 if (fs_info->dev_root)
1063 free_extent_buffer(fs_info->dev_root->node);
1064 if (fs_info->extent_root)
1065 free_extent_buffer(fs_info->extent_root->node);
1066 if (fs_info->tree_root)
1067 free_extent_buffer(fs_info->tree_root->node);
1068 if (fs_info->log_root_tree)
1069 free_extent_buffer(fs_info->log_root_tree->node);
1070 if (fs_info->chunk_root)
1071 free_extent_buffer(fs_info->chunk_root->node);
1074 static void free_map_lookup(struct cache_extent *ce)
1076 struct map_lookup *map;
1078 map = container_of(ce, struct map_lookup, ce);
1082 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
1084 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
1086 while (!list_empty(&fs_info->recow_ebs)) {
1087 struct extent_buffer *eb;
1088 eb = list_first_entry(&fs_info->recow_ebs,
1089 struct extent_buffer, recow);
1090 list_del_init(&eb->recow);
1091 free_extent_buffer(eb);
1093 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
1094 extent_io_tree_cleanup(&fs_info->extent_cache);
1095 extent_io_tree_cleanup(&fs_info->free_space_cache);
1096 extent_io_tree_cleanup(&fs_info->block_group_cache);
1097 extent_io_tree_cleanup(&fs_info->pinned_extents);
1098 extent_io_tree_cleanup(&fs_info->pending_del);
1099 extent_io_tree_cleanup(&fs_info->extent_ins);
1102 int btrfs_scan_fs_devices(int fd, const char *path,
1103 struct btrfs_fs_devices **fs_devices,
1104 u64 sb_bytenr, int super_recover,
1112 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1114 seek_ret = lseek(fd, 0, SEEK_END);
1118 dev_size = seek_ret;
1119 lseek(fd, 0, SEEK_SET);
1120 if (sb_bytenr > dev_size) {
1121 fprintf(stderr, "Superblock bytenr is larger than device size\n");
1125 ret = btrfs_scan_one_device(fd, path, fs_devices,
1126 &total_devs, sb_bytenr, super_recover);
1128 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1132 if (!skip_devices && total_devs != 1) {
1133 ret = btrfs_scan_lblkid();
1140 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info)
1142 struct btrfs_super_block *sb = fs_info->super_copy;
1151 nodesize = btrfs_super_nodesize(sb);
1152 leafsize = btrfs_super_leafsize(sb);
1153 sectorsize = btrfs_super_sectorsize(sb);
1154 stripesize = btrfs_super_stripesize(sb);
1156 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1157 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1159 ret = btrfs_read_sys_array(fs_info->chunk_root);
1163 blocksize = btrfs_level_size(fs_info->chunk_root,
1164 btrfs_super_chunk_root_level(sb));
1165 generation = btrfs_super_chunk_root_generation(sb);
1167 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1168 btrfs_super_chunk_root(sb),
1169 blocksize, generation);
1170 if (!extent_buffer_uptodate(fs_info->chunk_root->node)) {
1171 fprintf(stderr, "Couldn't read chunk root\n");
1175 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1176 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1178 fprintf(stderr, "Couldn't read chunk tree\n");
1185 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1187 u64 root_tree_bytenr,
1188 enum btrfs_open_ctree_flags flags)
1190 struct btrfs_fs_info *fs_info;
1191 struct btrfs_super_block *disk_super;
1192 struct btrfs_fs_devices *fs_devices = NULL;
1193 struct extent_buffer *eb;
1198 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1200 /* try to drop all the caches */
1201 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1202 fprintf(stderr, "Warning, could not drop caches\n");
1204 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1206 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1209 if (flags & OPEN_CTREE_RESTORE)
1210 fs_info->on_restoring = 1;
1211 if (flags & OPEN_CTREE_SUPPRESS_CHECK_BLOCK_ERRORS)
1212 fs_info->suppress_check_block_errors = 1;
1213 if (flags & OPEN_CTREE_IGNORE_FSID_MISMATCH)
1214 fs_info->ignore_fsid_mismatch = 1;
1216 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr,
1217 (flags & OPEN_CTREE_RECOVER_SUPER),
1218 (flags & OPEN_CTREE_NO_DEVICES));
1222 fs_info->fs_devices = fs_devices;
1223 if (flags & OPEN_CTREE_WRITES)
1228 if (flags & OPEN_CTREE_EXCLUSIVE)
1231 ret = btrfs_open_devices(fs_devices, oflags);
1235 disk_super = fs_info->super_copy;
1236 if (!(flags & OPEN_CTREE_RECOVER_SUPER))
1237 ret = btrfs_read_dev_super(fs_devices->latest_bdev,
1238 disk_super, sb_bytenr, 1);
1240 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr, 0);
1242 printk("No valid btrfs found\n");
1246 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_CHANGING_FSID &&
1247 !fs_info->ignore_fsid_mismatch) {
1248 fprintf(stderr, "ERROR: Filesystem UUID change in progress\n");
1252 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1254 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1255 flags & OPEN_CTREE_WRITES);
1259 ret = btrfs_setup_chunk_tree_and_device_map(fs_info);
1263 eb = fs_info->chunk_root->node;
1264 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1265 btrfs_header_chunk_tree_uuid(eb),
1268 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1269 if (ret && !(flags & __OPEN_CTREE_RETURN_CHUNK_ROOT))
1275 btrfs_release_all_roots(fs_info);
1276 btrfs_cleanup_all_caches(fs_info);
1278 btrfs_close_devices(fs_devices);
1280 btrfs_free_fs_info(fs_info);
1284 struct btrfs_fs_info *open_ctree_fs_info(const char *filename,
1285 u64 sb_bytenr, u64 root_tree_bytenr,
1286 enum btrfs_open_ctree_flags flags)
1289 struct btrfs_fs_info *info;
1290 int oflags = O_CREAT | O_RDWR;
1292 if (!(flags & OPEN_CTREE_WRITES))
1295 fp = open(filename, oflags, 0600);
1297 fprintf (stderr, "Could not open %s\n", filename);
1300 info = __open_ctree_fd(fp, filename, sb_bytenr, root_tree_bytenr,
1306 struct btrfs_root *open_ctree(const char *filename, u64 sb_bytenr,
1307 enum btrfs_open_ctree_flags flags)
1309 struct btrfs_fs_info *info;
1311 info = open_ctree_fs_info(filename, sb_bytenr, 0, flags);
1314 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1315 return info->chunk_root;
1316 return info->fs_root;
1319 struct btrfs_root *open_ctree_fd(int fp, const char *path, u64 sb_bytenr,
1320 enum btrfs_open_ctree_flags flags)
1322 struct btrfs_fs_info *info;
1323 info = __open_ctree_fd(fp, path, sb_bytenr, 0, flags);
1326 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1327 return info->chunk_root;
1328 return info->fs_root;
1331 int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
1334 u8 fsid[BTRFS_FSID_SIZE];
1335 int fsid_is_initialized = 0;
1336 struct btrfs_super_block buf;
1339 int max_super = super_recover ? BTRFS_SUPER_MIRROR_MAX : 1;
1343 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1344 ret = pread64(fd, &buf, sizeof(buf), sb_bytenr);
1345 if (ret < sizeof(buf))
1348 if (btrfs_super_bytenr(&buf) != sb_bytenr ||
1349 btrfs_super_magic(&buf) != BTRFS_MAGIC)
1352 memcpy(sb, &buf, sizeof(*sb));
1357 * we would like to check all the supers, but that would make
1358 * a btrfs mount succeed after a mkfs from a different FS.
1359 * So, we need to add a special mount option to scan for
1360 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1363 for (i = 0; i < max_super; i++) {
1364 bytenr = btrfs_sb_offset(i);
1365 ret = pread64(fd, &buf, sizeof(buf), bytenr);
1366 if (ret < sizeof(buf))
1369 if (btrfs_super_bytenr(&buf) != bytenr )
1371 /* if magic is NULL, the device was removed */
1372 if (btrfs_super_magic(&buf) == 0 && i == 0)
1374 if (btrfs_super_magic(&buf) != BTRFS_MAGIC)
1377 if (!fsid_is_initialized) {
1378 memcpy(fsid, buf.fsid, sizeof(fsid));
1379 fsid_is_initialized = 1;
1380 } else if (memcmp(fsid, buf.fsid, sizeof(fsid))) {
1382 * the superblocks (the original one and
1383 * its backups) contain data of different
1384 * filesystems -> the super cannot be trusted
1389 if (btrfs_super_generation(&buf) > transid) {
1390 memcpy(sb, &buf, sizeof(*sb));
1391 transid = btrfs_super_generation(&buf);
1395 return transid > 0 ? 0 : -1;
1398 static int write_dev_supers(struct btrfs_root *root,
1399 struct btrfs_super_block *sb,
1400 struct btrfs_device *device)
1406 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1407 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1409 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1410 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1411 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1414 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1415 * zero filled, we can use it directly
1417 ret = pwrite64(device->fd, root->fs_info->super_copy,
1418 BTRFS_SUPER_INFO_SIZE,
1419 root->fs_info->super_bytenr);
1420 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1424 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1425 bytenr = btrfs_sb_offset(i);
1426 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1429 btrfs_set_super_bytenr(sb, bytenr);
1432 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1433 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1434 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1437 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1438 * zero filled, we can use it directly
1440 ret = pwrite64(device->fd, root->fs_info->super_copy,
1441 BTRFS_SUPER_INFO_SIZE, bytenr);
1442 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1448 int write_all_supers(struct btrfs_root *root)
1450 struct list_head *cur;
1451 struct list_head *head = &root->fs_info->fs_devices->devices;
1452 struct btrfs_device *dev;
1453 struct btrfs_super_block *sb;
1454 struct btrfs_dev_item *dev_item;
1458 sb = root->fs_info->super_copy;
1459 dev_item = &sb->dev_item;
1460 list_for_each(cur, head) {
1461 dev = list_entry(cur, struct btrfs_device, dev_list);
1462 if (!dev->writeable)
1465 btrfs_set_stack_device_generation(dev_item, 0);
1466 btrfs_set_stack_device_type(dev_item, dev->type);
1467 btrfs_set_stack_device_id(dev_item, dev->devid);
1468 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1469 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1470 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1471 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1472 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1473 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1474 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1476 flags = btrfs_super_flags(sb);
1477 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1479 ret = write_dev_supers(root, sb, dev);
1485 int write_ctree_super(struct btrfs_trans_handle *trans,
1486 struct btrfs_root *root)
1489 struct btrfs_root *tree_root = root->fs_info->tree_root;
1490 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1492 if (root->fs_info->readonly)
1495 btrfs_set_super_generation(root->fs_info->super_copy,
1497 btrfs_set_super_root(root->fs_info->super_copy,
1498 tree_root->node->start);
1499 btrfs_set_super_root_level(root->fs_info->super_copy,
1500 btrfs_header_level(tree_root->node));
1501 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1502 chunk_root->node->start);
1503 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1504 btrfs_header_level(chunk_root->node));
1505 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1506 btrfs_header_generation(chunk_root->node));
1508 ret = write_all_supers(root);
1510 fprintf(stderr, "failed to write new super block err %d\n", ret);
1514 int close_ctree(struct btrfs_root *root)
1517 struct btrfs_trans_handle *trans;
1518 struct btrfs_fs_info *fs_info = root->fs_info;
1520 if (fs_info->last_trans_committed !=
1521 fs_info->generation) {
1522 trans = btrfs_start_transaction(root, 1);
1523 btrfs_commit_transaction(trans, root);
1524 trans = btrfs_start_transaction(root, 1);
1525 ret = commit_tree_roots(trans, fs_info);
1527 ret = __commit_transaction(trans, root);
1529 write_ctree_super(trans, root);
1530 btrfs_free_transaction(root, trans);
1532 btrfs_free_block_groups(fs_info);
1534 free_fs_roots_tree(&fs_info->fs_root_tree);
1536 btrfs_release_all_roots(fs_info);
1537 btrfs_close_devices(fs_info->fs_devices);
1538 btrfs_cleanup_all_caches(fs_info);
1539 btrfs_free_fs_info(fs_info);
1543 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1544 struct extent_buffer *eb)
1546 return clear_extent_buffer_dirty(eb);
1549 int wait_on_tree_block_writeback(struct btrfs_root *root,
1550 struct extent_buffer *eb)
1555 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1557 set_extent_buffer_dirty(eb);
1560 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1564 ret = extent_buffer_uptodate(buf);
1568 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1572 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1574 return set_extent_buffer_uptodate(eb);