[btree_node_checker] first draft

Spin-off from btree_damage_visitor

Makefile.in

@@ -67,6 +67,7 @@ SOURCE=\
 	persistent-data/data-structures/bitset.cc \
 	persistent-data/data-structures/bloom_filter.cc \
 	persistent-data/data-structures/btree.cc \
+	persistent-data/data-structures/btree_node_checker.cc \
 	persistent-data/error_set.cc \
 	persistent-data/file_utils.cc \
 	persistent-data/hex_dump.cc \

persistent-data/data-structures/btree_node_checker.cc

@@ -0,0 +1,127 @@
+#include "btree_node_checker.h"
+
+#include <sstream>
+
+using persistent_data::btree_detail::btree_node_checker;
+
+//----------------------------------------------------------------
+
+btree_node_checker::error_type btree_node_checker::get_last_error() {
+	return last_error_;
+}
+
+std::string btree_node_checker::get_last_error_string() {
+	switch (last_error_) {
+	case BLOCK_NR_MISMATCH:
+		return block_nr_mismatch_string();
+	case VALUE_SIZES_MISMATCH:
+		return value_sizes_mismatch_string();
+	case MAX_ENTRIES_TOO_LARGE:
+		return max_entries_too_large_string();
+	case MAX_ENTRIES_NOT_DIVISIBLE:
+		return max_entries_not_divisible_string();
+	case NR_ENTRIES_TOO_LARGE:
+		return nr_entries_too_large_string();
+	case NR_ENTRIES_TOO_SMALL:
+		return nr_entries_too_small_string();
+	case KEYS_OUT_OF_ORDER:
+		return keys_out_of_order_string();
+	case PARENT_KEY_MISMATCH:
+		return parent_key_mismatch_string();
+	case LEAF_KEY_OVERLAPPED:
+		return leaf_key_overlapped_string();
+	default:
+		return std::string();
+	}
+}
+
+void btree_node_checker::reset() {
+	last_error_ = NO_ERROR;
+}
+
+std::string btree_node_checker::block_nr_mismatch_string() {
+	std::ostringstream out;
+	out << "block number mismatch: actually "
+	    << error_location_
+	    << ", claims " << error_block_nr_;
+
+	return out.str();
+}
+
+std::string btree_node_checker::value_sizes_mismatch_string() {
+	std::ostringstream out;
+	out << "value size mismatch: expected " << error_value_sizes_[1]
+	    << ", but got " << error_value_sizes_[0]
+	    << ". This is not the btree you are looking for."
+	    << " (block " << error_location_ << ")";
+
+	return out.str();
+}
+
+std::string btree_node_checker::max_entries_too_large_string() {
+	std::ostringstream out;
+	out << "max entries too large: " << error_max_entries_
+	    << " (block " << error_location_ << ")";
+
+	return out.str();
+}
+
+std::string btree_node_checker::max_entries_not_divisible_string() {
+	std::ostringstream out;
+	out << "max entries is not divisible by 3: " << error_max_entries_
+	    << " (block " << error_location_ << ")";
+
+	return out.str();
+}
+
+std::string btree_node_checker::nr_entries_too_large_string() {
+	std::ostringstream out;
+	out << "bad nr_entries: "
+	    << error_nr_entries_ << " < "
+	    << error_max_entries_
+	    << " (block " << error_location_ << ")";
+
+	return out.str();
+}
+
+std::string btree_node_checker::nr_entries_too_small_string() {
+	std::ostringstream out;
+	out << "too few entries in btree_node: "
+	    << error_nr_entries_
+	    << ", expected at least "
+	    << (error_max_entries_ / 3)
+	    << " (block " << error_location_
+	    << ", max_entries = " << error_max_entries_ << ")";
+
+	return out.str();
+}
+
+std::string btree_node_checker::keys_out_of_order_string() {
+	std::ostringstream out;
+	out << "keys are out of order, "
+	    << error_keys_[0] << " <= " << error_keys_[1]
+	    << " (block " << error_location_ << ")";
+
+	return out.str();
+}
+
+std::string btree_node_checker::parent_key_mismatch_string() {
+	std::ostringstream out;
+	out << "parent key mismatch: parent was " << error_keys_[1]
+	    << ", but lowest in node was " << error_keys_[0]
+	    << " (block " << error_location_ << ")";
+
+	return out.str();
+}
+
+std::string btree_node_checker::leaf_key_overlapped_string() {
+	std::ostringstream out;
+	out << "the last key of the previous leaf was " << error_keys_[1]
+	    << " and the first key of this leaf is " << error_keys_[0]
+	    << " (block " << error_location_ << ")";
+
+	return out.str();
+}
+
+//----------------------------------------------------------------
+

persistent-data/data-structures/btree_node_checker.h

@@ -0,0 +1,206 @@
+#ifndef BTREE_NODE_CHECKER_H
+#define BTREE_NODE_CHECKER_H
+
+#include "block-cache/block_cache.h"
+#include "persistent-data/block.h"
+#include "persistent-data/data-structures/btree.h"
+#include "persistent-data/data-structures/btree_disk_structures.h"
+
+#include <boost/optional.hpp>
+#include <string>
+
+using bcache::block_address;
+
+//----------------------------------------------------------------
+
+namespace persistent_data {
+	namespace btree_detail {
+		class btree_node_checker {
+		public:
+			enum error_type {
+				NO_ERROR,
+				BLOCK_NR_MISMATCH,
+				VALUE_SIZES_MISMATCH,
+				MAX_ENTRIES_TOO_LARGE,
+				MAX_ENTRIES_NOT_DIVISIBLE,
+				NR_ENTRIES_TOO_LARGE,
+				NR_ENTRIES_TOO_SMALL,
+				KEYS_OUT_OF_ORDER,
+				VALUE_SIZE_MISMATCH,
+				PARENT_KEY_MISMATCH,
+				LEAF_KEY_OVERLAPPED,
+			};
+
+			btree_node_checker():
+				last_error_(NO_ERROR),
+				error_location_(0),
+				error_block_nr_(0),
+				error_nr_entries_(0),
+				error_max_entries_(0),
+				error_value_sizes_{0, 0},
+				error_keys_{0, 0} {
+			}
+
+			template <typename ValueTraits>
+			bool check_block_nr(btree_detail::node_ref<ValueTraits> const &n) {
+				if (n.get_location() != n.get_block_nr()) {
+					last_error_ = BLOCK_NR_MISMATCH;
+					error_block_nr_ = n.get_block_nr();
+					error_location_ = n.get_location();
+
+					return false;
+				}
+
+				return true;
+			}
+
+			template <typename ValueTraits>
+			bool check_value_size(btree_detail::node_ref<ValueTraits> const &n) {
+				if (!n.value_sizes_match()) {
+					last_error_ = VALUE_SIZES_MISMATCH;
+					error_location_ = n.get_location();
+					error_value_sizes_[0] = n.get_value_size();
+					error_value_sizes_[1] = sizeof(typename ValueTraits::disk_type);
+					return false;
+				}
+
+				return true;
+			}
+
+			template <typename ValueTraits>
+			bool check_max_entries(btree_detail::node_ref<ValueTraits> const &n) {
+				size_t elt_size = sizeof(uint64_t) + n.get_value_size();
+				if (elt_size * n.get_max_entries() + sizeof(node_header) > MD_BLOCK_SIZE) {
+					last_error_ = MAX_ENTRIES_TOO_LARGE;
+					error_location_ = n.get_location();
+					error_max_entries_ = n.get_max_entries();
+
+					return false;
+				}
+
+				if (n.get_max_entries() % 3) {
+					last_error_ = MAX_ENTRIES_NOT_DIVISIBLE;
+					error_location_ = n.get_location();
+					error_max_entries_ = n.get_max_entries();
+
+					return false;
+				}
+
+				return true;
+			}
+
+			template <typename ValueTraits>
+			bool check_nr_entries(btree_detail::node_ref<ValueTraits> const &n,
+					      bool is_root) {
+				if (n.get_nr_entries() > n.get_max_entries()) {
+					last_error_ = NR_ENTRIES_TOO_LARGE;
+					error_location_ = n.get_location();
+					error_nr_entries_ = n.get_nr_entries();
+					error_max_entries_ = n.get_max_entries();
+
+					return false;
+				}
+
+				block_address min = n.get_max_entries() / 3;
+				if (!is_root && (n.get_nr_entries() < min)) {
+					last_error_ = NR_ENTRIES_TOO_SMALL;
+					error_location_ = n.get_location();
+					error_nr_entries_ = n.get_nr_entries();
+					error_max_entries_ = n.get_max_entries();
+
+					return false;
+				}
+
+				return true;
+			}
+
+			template <typename ValueTraits>
+			bool check_ordered_keys(btree_detail::node_ref<ValueTraits> const &n) {
+				unsigned nr_entries = n.get_nr_entries();
+
+				if (nr_entries == 0)
+					return true; // can only happen if a root node
+
+				uint64_t last_key = n.key_at(0);
+
+				for (unsigned i = 1; i < nr_entries; i++) {
+					uint64_t k = n.key_at(i);
+					if (k <= last_key) {
+						last_error_ = KEYS_OUT_OF_ORDER;
+						error_location_ = n.get_location();
+						error_keys_[0] = k;
+						error_keys_[1] = last_key;
+
+						return false;
+					}
+					last_key = k;
+				}
+
+				return true;
+			}
+
+			template <typename ValueTraits>
+			bool check_parent_key(btree_detail::node_ref<ValueTraits> const &n,
+					      boost::optional<uint64_t> key) {
+				if (!key)
+					return true;
+
+				if (*key > n.key_at(0)) {
+					last_error_ = PARENT_KEY_MISMATCH;
+					error_location_ = n.get_location();
+					error_keys_[0] = n.key_at(0);
+					error_keys_[1] = *key;
+
+					return false;
+				}
+
+				return true;
+			}
+
+			template <typename ValueTraits>
+			bool check_leaf_key(btree_detail::node_ref<ValueTraits> const &n,
+					    boost::optional<uint64_t> key) {
+				if (n.get_nr_entries() == 0)
+					return true; // can only happen if a root node
+
+				if (key && *key >= n.key_at(0)) {
+					last_error_ = LEAF_KEY_OVERLAPPED;
+					error_location_ = n.get_location();
+					error_keys_[0] = n.key_at(0);
+					error_keys_[1] = *key;
+
+					return false;
+				}
+
+				return true;
+			}
+
+			error_type get_last_error();
+			std::string get_last_error_string();
+			void reset();
+
+		private:
+			std::string block_nr_mismatch_string();
+			std::string value_sizes_mismatch_string();
+			std::string max_entries_too_large_string();
+			std::string max_entries_not_divisible_string();
+			std::string nr_entries_too_large_string();
+			std::string nr_entries_too_small_string();
+			std::string keys_out_of_order_string();
+			std::string parent_key_mismatch_string();
+			std::string leaf_key_overlapped_string();
+
+			error_type last_error_;
+			block_address error_location_;
+			block_address error_block_nr_;
+			uint32_t error_nr_entries_;
+			uint32_t error_max_entries_;
+			uint32_t error_value_sizes_[2];
+			uint64_t error_keys_[2];
+		};
+	}
+}
+
+//----------------------------------------------------------------
+
+#endif