move some things for udp. added a class for kademlia.

1 files changed, 530 insertions, 37 deletions

diff --git a/net.cpp b/net.cpp
index 6e9efff1..0a22a7e6 100644
--- a/net.cpp
+++ b/net.cpp
@@ -20,8 +20,8 @@
  */
 
 #include <boost/test/unit_test.hpp>
-#include <libdevcrypto/Common.h>
 #include <libdevcore/Worker.h>
+#include <libdevcrypto/Common.h>
 #include <libp2p/UDP.h>
 using namespace std;
 using namespace dev;
@@ -29,58 +29,551 @@ using namespace dev::p2p;
 namespace ba = boost::asio;
 namespace bi = ba::ip;
 
-class TestA: UDPSocketEvents, public Worker
+/**
+ * Ping packet: Check if node is alive.
+ * PingNode is cached and regenerated after expiration - t, where t is timeout.
+ *
+ * signature: Signature of message.
+ * ipAddress: Our IP address.
+ * port: Our port.
+ * expiration: Triggers regeneration of packet. May also provide control over synchronization.
+ *
+ * Ping is used to implement evict. When a new node is seen for
+ * a given bucket which is full, the least-responsive node is pinged.
+ * If the pinged node doesn't respond then it is removed and the new
+ * node is inserted.
+ */
+struct PingNode: RLPDatagram
+{
+	bytes ipAddress;
+	uint16_t port;
+	uint64_t expiration;
+
+	Signature signature;
+	
+//	void streamRLP(RLPStream& _s) const { _s.appendList(3); _s << ipAddress << port << expiration; }
+};
+
+struct Pong: RLPDatagram
 {
+	// todo: weak-signed pong
+	Address from;
+	uint64_t replyTo;	/// expiration from PingNode
+	
+	void streamRLP(RLPStream& _s) const { _s.appendList(2); _s << from << replyTo; }
+};
+
+/**
+ * FindNeighbors Packet: Request k-nodes, closest to the target.
+ * FindNeighbors is cached and regenerated after expiration - t, where t is timeout.
+ *
+ * signature: Signature of message.
+ * target: Address of NodeId. The responding node will send back nodes closest to the target.
+ * expiration: Triggers regeneration of packet. May also provide control over synchronization.
+ *
+ */
+struct FindNeighbors: RLPDatagram
+{
+	h160 target;
+	uint64_t expiration;
+	
+	Signature signature;
+	
+	void streamRLP(RLPStream& _s) const { _s.appendList(2); _s << target << expiration; }
+};
+
+/**
+ * Node Packet: Multiple node packets are sent in response to FindNeighbors.
+ */
+struct Neighbors: RLPDatagram
+{
+	struct Node
+	{
+		bytes ipAddress;
+		uint16_t port;
+		NodeId node;
+//		void streamRLP(RLPStream& _s) const { _s.appendList(3); _s << ipAddress << port << node; }
+	};
+	
+	std::set<Node> nodes;
+	h256 nonce;
+	
+	Signature signature;
+	
+//	void streamRLP(RLPStream& _s) const { _s.appendList(2); _s.appendList(nodes.size()); for (auto& n: nodes) n.streamRLP(_s); _s << nonce; }
+};
+
+/**
+ * NodeTable using S/Kademlia system for node discovery and preference.
+ * untouched buckets are refreshed if they have not been touched within an hour
+ *
+ * Thread-safety is ensured by modifying NodeEntry details via 
+ * shared_ptr replacement instead of mutating values.
+ *
+ * @todo don't try to evict node if node isRequired. (support for makeRequired)
+ * @todo optimize (use tree for state (or set w/custom compare for cache))
+ * @todo constructor support for m_node, m_secret
+ * @todo use s_bitsPerStep for find and refresh/ping
+ * @todo exclude bucket from refresh if we have node as peer
+ * @todo restore nodes
+ */
+class NodeTable: UDPSocketEvents, public std::enable_shared_from_this<NodeTable>
+{
+	using nodeSocket = UDPSocket<NodeTable, 1024>;
+	using timePoint = std::chrono::steady_clock::time_point;
+	
+	static unsigned const s_bucketSize = 16;	// Denoted by k in [Kademlia]. Number of nodes stored in each bucket.
+//	const unsigned s_bitsPerStep = 5;			// @todo Denoted by b in [Kademlia]. Bits by which address space will be divided for find responses.
+	static unsigned const s_alpha = 3;		// Denoted by \alpha in [Kademlia]. Number of concurrent FindNeighbors requests.
+	const unsigned s_findTimout = 300;		// How long to wait between find queries.
+//	const unsigned s_siblings = 5;			// @todo Denoted by s in [S/Kademlia]. User-defined by sub-protocols.
+	const unsigned s_bucketRefresh = 3600;		// Refresh interval prevents bucket from becoming stale. [Kademlia]
+	const unsigned s_bits = sizeof(Address);	// Denoted by n.
+	const unsigned s_buckets = 8 * s_bits - 1;
+	const unsigned s_evictionCheckInterval = 75;	// Interval by which eviction timeouts are checked.
+	const unsigned s_pingTimeout = 500;
+	static size_t const s_tableSize = Address::size * 8 - 1; // Address::size
+	
+public:
+	static unsigned dist(Address const& _a, Address const& _b) { u160 d = _a ^ _b; unsigned ret; for (ret = 0; d >>= 1; ++ret) {}; return ret; }
+	
+protected:
+	struct NodeDefaultEndpoint
+	{
+		NodeDefaultEndpoint(bi::udp::endpoint _udp): udp(_udp) {}
+		bi::udp::endpoint udp;
+	};
+	
+	struct NodeEntry
+	{
+		NodeEntry(Address _id, Public _pubk, bi::udp::endpoint _udp): id(_id), pubk(_pubk), endpoint(NodeDefaultEndpoint(_udp)), distance(0) {}
+		NodeEntry(NodeEntry _src, Address _id, Public _pubk, bi::udp::endpoint _udp): id(_id), pubk(_pubk), endpoint(NodeDefaultEndpoint(_udp)), distance(dist(_src.id,_id)) {}
+		NodeEntry(NodeEntry _src, Address _id, Public _pubk, NodeDefaultEndpoint _gw): id(_id), pubk(_pubk), endpoint(_gw), distance(dist(_src.id,_id)) {}
+		Address id;
+		Public pubk;
+		NodeDefaultEndpoint endpoint;		///< How we've previously connected to this node. (must match node's reported endpoint)
+		const unsigned distance;
+		timePoint activePing;
+	};
+	
+	struct NodeBucket
+	{
+		unsigned distance;
+		timePoint modified;
+		std::list<std::weak_ptr<NodeEntry>> nodes;
+	};
+	
+	using EvictionTimeout = std::pair<std::pair<Address,timePoint>,Address>;
+
 public:
-	TestA(): Worker("test",0), m_io(), m_socket(new UDPSocket<TestA, 1024>(m_io, *this, 30300)) {}
-	~TestA() { m_io.stop(); stopWorking(); }
+	NodeTable(ba::io_service& _io):
+		m_node(NodeEntry(Address(), Public(), bi::udp::endpoint())),
+		m_socket(new nodeSocket(_io, *this, 30300)),
+		m_socketPtr(m_socket.get()),
+		m_io(_io),
+		m_bucketRefreshTimer(m_io),
+		m_evictionCheckTimer(m_io)
+	{
+		for (unsigned i = 0; i < s_buckets; i++)
+			m_state[i].distance = i, m_state[i].modified = chrono::steady_clock::now() - chrono::seconds(1);
+		doRefreshBuckets(boost::system::error_code());
+	}
+	
+	~NodeTable() {
+		m_evictionCheckTimer.cancel();
+		m_bucketRefreshTimer.cancel();
+		m_socketPtr->disconnect();
+	}
+	
+	void join() { doFindNode(m_node.id); }
+	
+	std::list<Address> nodes() const
+	{
+		std::list<Address> nodes;
+		Guard l(x_nodes);
+		for (auto& i: m_nodes)
+			nodes.push_back(i.second->id);
+		return std::move(nodes);
+	}
+	
+	NodeEntry operator[](Address _id)
+	{
+		Guard l(x_nodes);
+		return *m_nodes[_id];
+	}
+	
+protected:
+	void requestNeighbors(NodeEntry const& _node, Address _target) const
+	{
+		FindNeighbors p;
+		p.target = _target;
+		
+		p.to = _node.endpoint.udp;
+		p.seal(m_secret);
+		m_socketPtr->send(p);
+	}
+	
+	/// Dispatches udp requests in order to populate node table to be as close as possible to _node.
+	void doFindNode(Address _node, unsigned _round = 0, std::shared_ptr<std::set<std::shared_ptr<NodeEntry>>> _tried = std::shared_ptr<std::set<std::shared_ptr<NodeEntry>>>())
+	{
+		if (!m_socketPtr->isOpen() || _round == 7)
+			return;
+
+		auto nearest = findNearest(_node);
+		std::list<std::shared_ptr<NodeEntry>> tried;
+		for (unsigned i = 0; i < nearest.size() && tried.size() < s_alpha; i++)
+			if (!_tried->count(nearest[i]))
+			{
+				tried.push_back(nearest[i]);
+				requestNeighbors(*nearest[i], _node);
+			}
+			else
+				continue;
+		
+		while (auto n = tried.front())
+		{
+			_tried->insert(n);
+			tried.pop_front();
+		}
+		
+		auto self(shared_from_this());
+		m_evictionCheckTimer.expires_from_now(boost::posix_time::milliseconds(s_findTimout));
+		m_evictionCheckTimer.async_wait([this, self, _node, _round, _tried](boost::system::error_code const& _ec)
+		{
+			if (_ec)
+				return;
+			doFindNode(_node, _round + 1, _tried);
+		});
+	}
+	
+	std::vector<std::shared_ptr<NodeEntry>> findNearest(Address _target)
+	{
+		// send s_alpha FindNeighbors packets to nodes we know, closest to target
+		unsigned head = dist(m_node.id, _target);
+		unsigned tail = (head - 1) % (s_tableSize - 1);
+		
+		// todo: optimize with tree
+		std::map<unsigned, std::list<std::shared_ptr<NodeEntry>>> found;
+		unsigned count = 0;
+		
+		// if d is 0, then we roll look forward, if last, we reverse, else, spread from d
+		if (head != 0 && tail != s_tableSize)
+			while (head != tail && count < s_bucketSize)
+			{
+				Guard l(x_state);
+				for (auto& n: m_state[head].nodes)
+					if (auto p = n.lock())
+					{
+						if (count < s_bucketSize)
+							found[dist(_target, p->id)].push_back(p);
+						else
+							break;
+					}
+				
+				if (count < s_bucketSize && head)
+					for (auto& n: m_state[tail].nodes)
+						if (auto p = n.lock())
+						{
+							if (count < s_bucketSize)
+								found[dist(_target, p->id)].push_back(p);
+							else
+								break;
+						}
+				head++;
+				tail = (tail - 1) % (s_tableSize - 1);
+			}
+		else if (head == 0)
+			while (head < s_bucketSize && count < s_bucketSize)
+			{
+				Guard l(x_state);
+				for (auto& n: m_state[head].nodes)
+					if (auto p = n.lock())
+					{
+						if (count < s_bucketSize)
+							found[dist(_target, p->id)].push_back(p);
+						else
+							break;
+					}
+				head--;
+			}
+		else if (tail == s_tableSize - 1)
+			while (tail > 0 && count < s_bucketSize)
+			{
+				Guard l(x_state);
+				for (auto& n: m_state[tail].nodes)
+					if (auto p = n.lock())
+					{
+						if (count < s_bucketSize)
+							found[dist(_target, p->id)].push_back(p);
+						else
+							break;
+					}
+				tail--;
+			}
+		
+		std::vector<std::shared_ptr<NodeEntry>> ret;
+		for (auto& nodes: found)
+			for (auto& n: nodes.second)
+				ret.push_back(n);
+		return std::move(ret);
+	}
+	
+	void ping(bi::address _address, unsigned _port) const
+	{
+		PingNode p;
+		string ip = m_node.endpoint.udp.address().to_string();
+		p.ipAddress = asBytes(ip);
+		p.port = m_node.endpoint.udp.port();
+//		p.expiration;
+		p.seal(m_secret);
+		m_socketPtr->send(p);
+	}
+	
+	void ping(NodeEntry* _n) const
+	{
+		if (_n && _n->endpoint.udp.address().is_v4())
+			ping(_n->endpoint.udp.address(), _n->endpoint.udp.port());
+	}
+	
+	void evict(std::shared_ptr<NodeEntry> _leastSeen, std::shared_ptr<NodeEntry> _new)
+	{
+		if (!m_socketPtr->isOpen())
+			return;
+		
+		Guard l(x_evictions);
+		m_evictions.push_back(EvictionTimeout(make_pair(_leastSeen->id,chrono::steady_clock::now()), _new->id));
+		if (m_evictions.size() == 1)
+			doCheckEvictions(boost::system::error_code());
+		
+		m_evictions.push_back(EvictionTimeout(make_pair(_leastSeen->id,chrono::steady_clock::now()), _new->id));
+		ping(_leastSeen.get());
+	}
 	
+	void noteNode(Public _pubk, bi::udp::endpoint _endpoint)
+	{
+		Address id = right160(sha3(_pubk));
+		std::shared_ptr<NodeEntry> node;
+		{
+			Guard l(x_nodes);
+			auto n = m_nodes.find(id);
+			if (n == m_nodes.end())
+			{
+				m_nodes[id] = std::shared_ptr<NodeEntry>(new NodeEntry(m_node, id, _pubk, _endpoint));
+				node = m_nodes[id];
+			}
+			else
+				node = n->second;
+		}
+		
+		noteNode(node);
+	}
+	
+	void noteNode(std::shared_ptr<NodeEntry> _n)
+	{
+		std::shared_ptr<NodeEntry> contested;
+		{
+			NodeBucket s = bucket(_n.get());
+			Guard l(x_state);
+			s.nodes.remove_if([&_n](std::weak_ptr<NodeEntry> n)
+			{
+				auto p = n.lock();
+				if (!p || p == _n)
+					return true;
+				return false;
+			});
+
+			if (s.nodes.size() >= s_bucketSize)
+			{
+				contested = s.nodes.front().lock();
+				if (!contested)
+				{
+					s.nodes.pop_front();
+					s.nodes.push_back(_n);
+				}
+			}
+			else
+				s.nodes.push_back(_n);
+		}
+		
+		if (contested)
+			evict(contested, _n);
+	}
+	
+	void dropNode(std::shared_ptr<NodeEntry> _n)
+	{
+		NodeBucket s = bucket(_n.get());
+		{
+			Guard l(x_state);
+			s.nodes.remove_if([&_n](std::weak_ptr<NodeEntry> n) { return n.lock() == _n; });
+		}
+		Guard l(x_nodes);
+		m_nodes.erase(_n->id);
+	}
+	
+	NodeBucket const& bucket(NodeEntry* _n) const
+	{
+		return m_state[_n->distance];
+	}
+	
+	void onReceived(UDPSocketFace*, bi::udp::endpoint const& _from, bytesConstRef _packet)
+	{
+		RLP rlp(_packet);
+		
+		
+		// whenever a pong is received, first check if it's in m_evictions, if so, remove it
+		Guard l(x_evictions);
+	}
+	
+	void onDisconnected(UDPSocketFace*)
+	{
+		
+	}
+	
+	void doCheckEvictions(boost::system::error_code const& _ec)
+	{
+		if (_ec || !m_socketPtr->isOpen())
+			return;
+
+		m_evictionCheckTimer.expires_from_now(boost::posix_time::milliseconds(s_evictionCheckInterval));
+		auto self(shared_from_this());
+		m_evictionCheckTimer.async_wait([this, self](boost::system::error_code const& _ec)
+		{
+			if (_ec)
+				return;
+			
+			bool evictionsRemain = false;
+			std::list<shared_ptr<NodeEntry>> drop;
+			{
+				Guard l(x_evictions);
+				for (auto& e: m_evictions)
+					if (chrono::steady_clock::now() - e.first.second > chrono::milliseconds(s_pingTimeout))
+					{
+						Guard l(x_nodes);
+						drop.push_back(m_nodes[e.second]);
+					}
+				evictionsRemain = m_evictions.size() - drop.size() > 0;
+			}
+			
+			for (auto& n: drop)
+				dropNode(n);
+			
+			if (evictionsRemain)
+				doCheckEvictions(boost::system::error_code());
+		});
+	}
+	
+	void doRefreshBuckets(boost::system::error_code const& _ec)
+	{
+		cout << "refreshing buckets" << endl;
+		if (_ec)
+			return;
+		
+		// first check if there are any pending evictions
+
+		
+		bool connected = m_socketPtr->isOpen();
+		bool refreshed = false;
+		if (connected)
+		{
+			Guard l(x_state);
+			for (auto& d: m_state)
+				if (chrono::steady_clock::now() - d.modified > chrono::seconds(s_bucketRefresh))
+					while (!d.nodes.empty())
+					{
+						auto n = d.nodes.front();
+						if (auto p = n.lock())
+						{
+							refreshed = true;
+							ping(p.get());
+							break;
+						}
+						d.nodes.pop_front();
+					}
+		}
+
+		unsigned nextRefresh = connected ? (refreshed ? 200 : s_bucketRefresh*1000) : 10000;
+		auto runcb = [this](boost::system::error_code const& error) -> void { doRefreshBuckets(error); };
+		m_bucketRefreshTimer.expires_from_now(boost::posix_time::milliseconds(nextRefresh));
+		m_bucketRefreshTimer.async_wait(runcb);
+	}
+	
+private:
+	NodeEntry m_node;										///< This node.
+	Secret m_secret;											///< This nodes secret key.
+
+	mutable Mutex x_nodes;									///< Mutable for thread-safe copy in nodes() const.
+	std::map<Address, std::shared_ptr<NodeEntry>> m_nodes;		///< Address -> Node table (most common lookup path)
+
+	Mutex x_state;
+	std::array<NodeBucket, s_tableSize> m_state;				///< State table; logbinned nodes.
+
+	Mutex x_evictions;
+	std::deque<EvictionTimeout> m_evictions;					///< Eviction timeouts.
+	
+	shared_ptr<nodeSocket> m_socket;							///< Shared pointer for our UDPSocket; ASIO requires shared_ptr.
+	nodeSocket* m_socketPtr;									///< Set to m_socket.get().
+	ba::io_service& m_io;										///< Used by bucket refresh timer.
+	boost::asio::deadline_timer m_bucketRefreshTimer;			///< Timer which schedules and enacts bucket refresh.
+	boost::asio::deadline_timer m_evictionCheckTimer;			///< Timer for handling node evictions.
+};
+
+/**
+ * Only used for testing. Not useful beyond tests.
+ */
+class TestHost: public Worker
+{
+public:
+	TestHost(): Worker("test",0), m_io() {};
+	~TestHost() { m_io.stop(); stopWorking(); }
+	void start() { startWorking(); }
+	void doWork() { m_io.run(); }
+	
+protected:
+	ba::io_service m_io;
+};
+
+/**
+ * Only used for testing. Not useful beyond tests.
+ */
+class TestNodeHost: public TestHost
+{
+public:
+	TestNodeHost(): m_nodes(m_io) {};
+	~TestNodeHost() { m_io.stop(); stopWorking(); }
+	void start() { startWorking(); }
+	void doWork() { m_io.run(); }
+
+	NodeTable m_nodes;
+};
+
+class TestUDPSocket: UDPSocketEvents, public TestHost
+{
+public:
+	TestUDPSocket(): m_socket(new UDPSocket<TestUDPSocket, 1024>(m_io, *this, 30300)) {}
+	~TestUDPSocket() { m_io.stop(); stopWorking(); }
 	void start() { startWorking(); }
 	void doWork() { m_io.run(); }
 	
 	void onDisconnected(UDPSocketFace*) {};
-	void onReceived(UDPSocketFace*, bi::udp::endpoint const& _from, bytesConstRef _packet) { if (_packet.toString() == "AAAA") success = true; }
+	void onReceived(UDPSocketFace*, bi::udp::endpoint const&, bytesConstRef _packet) { if (_packet.toString() == "AAAA") success = true; }
 
-	ba::io_service m_io;
-	shared_ptr<UDPSocket<TestA, 1024>> m_socket;
+	shared_ptr<UDPSocket<TestUDPSocket, 1024>> m_socket;
 	
 	bool success = false;
 };
 
-//struct TestBProtocol: UDPSocketEvents
-//{
-//	void onDisconnected(UDPSocketFace*) {};
-//	void onReceived(UDPSocketFace*, bi::udp::endpoint const& _from, bytesConstRef _packet) { cout << "received TestBProtocol" << endl; };
-//};
-//
-//class TestB: TestBProtocol
-//{
-//public:
-//	TestB(): m_io(), m_socket(m_io, *this, 30300) {}
-////private:
-//	ba::io_service m_io;
-//	UDPSocket<TestBProtocol, 1024> m_socket;
-//};
-//
-//class TestC
-//{
-//public:
-//	TestC(): m_io(), m_socket(m_io, m_rpc, 30300) {}
-////private:
-//	ba::io_service m_io;
-//	TestBProtocol m_rpc;
-//	UDPSocket<TestBProtocol, 1024> m_socket;
-//};
-
 BOOST_AUTO_TEST_SUITE(p2p)
 
-BOOST_AUTO_TEST_CASE(test_txrx_one)
+BOOST_AUTO_TEST_CASE(kademlia)
 {
-	UDPDatagram d;
-	d.to = boost::asio::ip::udp::endpoint(boost::asio::ip::address::from_string("127.0.0.1"), 30300);
-	d.data = bytes({65,65,65,65});
+	TestNodeHost nodeHost;
 	
-	TestA a; a.m_socket->connect(); a.start();
+}
+
+BOOST_AUTO_TEST_CASE(test_txrx_one)
+{
+	UDPDatagram d(bi::udp::endpoint(boost::asio::ip::address::from_string("127.0.0.1"), 30300), bytes({65,65,65,65}));
+	TestUDPSocket a; a.m_socket->connect(); a.start();
 	a.m_socket->send(d);
 	sleep(1);
 	BOOST_REQUIRE_EQUAL(true, a.success);