Btrfs: check inconsistence between chunk and block group

With btrfs-corrupt-block, one can drop one chunk item and mounting
will end up with a panic in btrfs_full_stripe_len().

This doesn't not remove the BUG_ON, but instead checks it a bit
earlier when we find the block group item.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

fs/btrfs/extent-tree.c

@@ -9887,7 +9887,22 @@ static int find_first_block_group(struct btrfs_root *root,
 
 		if (found_key.objectid >= key->objectid &&
 		    found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
-			ret = 0;
+			struct extent_map_tree *em_tree;
+			struct extent_map *em;
+
+			em_tree = &root->fs_info->mapping_tree.map_tree;
+			read_lock(&em_tree->lock);
+			em = lookup_extent_mapping(em_tree, found_key.objectid,
+						   found_key.offset);
+			read_unlock(&em_tree->lock);
+			if (!em) {
+				btrfs_err(root->fs_info,
+			"logical %llu len %llu found bg but no related chunk",
+					  found_key.objectid, found_key.offset);
+				ret = -ENOENT;
+			} else {
+				ret = 0;
+			}
 			goto out;
 		}
 		path->slots[0]++;