标准STUN判断NAT类型的过程及改进

2024-04-20 21:08
文章标签 类型 判断 过程 nat 改进 标准 stun

本文主要是介绍标准STUN判断NAT类型的过程及改进，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

这里基于stund的实现，来研究标准STUN协议，判断NatType的过程。

stund用于判断NatType的接口的用法

首先来看stund中用于判断NatType的接口的用法。这里主要来看stund中的STUN客户端client.cxx的实现。client.cxx是一个常规的C/C++ app，这个app的主要code如下：

void usage() {cerr << "Usage:" << endl<< "    ./client stunServerHostname [testNumber] [-v] [-p srcPort] ""[-i nicAddr1] [-i nicAddr2] [-i nicAddr3] " << endl<< "For example, if the STUN server was larry.gloo.net, you could do:" << endl<< "    ./client larry.gloo.net" << endl<< "The testNumber is just used for special tests." << endl<< " test 1 runs test 1 from the RFC. For example:" << endl<< "    ./client larry.gloo.net 0" << endl << endl << endl;
}#define MAX_NIC 3
StunAddress4 stunServerAddr;int main(int argc, char* argv[]) {assert( sizeof(UInt8 ) == 1);assert( sizeof(UInt16) == 2);assert( sizeof(UInt32) == 4);initNetwork();cout << "STUN client version " << STUN_VERSION << endl;int testNum = 0;bool verbose = false;stunServerAddr.addr = 0;int srcPort = 0;StunAddress4 sAddr[MAX_NIC];int retval[MAX_NIC];int numNic = 0;for (int i = 0; i < MAX_NIC; i++) {sAddr[i].addr = 0;sAddr[i].port = 0;retval[i] = 0;}for (int arg = 1; arg < argc; arg++) {if (!strcmp(argv[arg], "-v")) {verbose = true;} else if (!strcmp(argv[arg], "-i")) {arg++;if (argc <= arg) {usage();exit(-1);}if (numNic >= MAX_NIC) {cerr << "Can not have more than " << MAX_NIC <<" -i options" << endl;usage();exit(-1);}stunParseServerName(argv[arg], sAddr[numNic++]);} else if (!strcmp(argv[arg], "-p")) {arg++;if (argc <= arg) {usage();exit(-1);}srcPort = strtol(argv[arg], NULL, 10);} else {char* ptr;int t = strtol(argv[arg], &ptr, 10);if (*ptr == 0) {// conversion workedtestNum = t;cout << "running test number " << testNum << endl;} else {bool ret = stunParseServerName(argv[arg], stunServerAddr);if (ret != true) {cerr << argv[arg] << " is not a valid host name " << endl;usage();exit(-1);}}}}if (srcPort == 0) {srcPort = stunRandomPort();}if (numNic == 0) {// use defaultnumNic = 1;}for (int nic = 0; nic < numNic; nic++) {sAddr[nic].port = srcPort;if (stunServerAddr.addr == 0) {usage();exit(-1);}if (testNum == 0) {bool presPort = false;bool hairpin = false;NatType stype = stunNatType(stunServerAddr, verbose, &presPort, &hairpin, srcPort, &sAddr[nic]);if (nic == 0) {cout << "Primary: ";} else {cout << "Secondary: ";}switch (stype) {case StunTypeFailure:cout << "Some stun error detetecting NAT type";retval[nic] = -1;exit(-1);break;case StunTypeUnknown:cout << "Some unknown type error detetecting NAT type";retval[nic] = 0xEE;break;case StunTypeOpen:cout << "Open";retval[nic] = 0x00;break;case StunTypeIndependentFilter:cout << "Independent Mapping, Independent Filter";if (presPort)cout << ", preserves ports";elsecout << ", random port";if (hairpin)cout << ", will hairpin";elsecout << ", no hairpin";retval[nic] = 0x02;break;case StunTypeDependentFilter:cout << "Independent Mapping, Address Dependent Filter";if (presPort)cout << ", preserves ports";elsecout << ", random port";if (hairpin)cout << ", will hairpin";elsecout << ", no hairpin";retval[nic] = 0x04;break;case StunTypePortDependedFilter:cout << "Independent Mapping, Port Dependent Filter";if (presPort)cout << ", preserves ports";elsecout << ", random port";if (hairpin)cout << ", will hairpin";elsecout << ", no hairpin";retval[nic] = 0x06;break;case StunTypeDependentMapping:cout << "Dependent Mapping";if (presPort)cout << ", preserves ports";elsecout << ", random port";if (hairpin)cout << ", will hairpin";elsecout << ", no hairpin";retval[nic] = 0x08;break;case StunTypeFirewall:cout << "Firewall";retval[nic] = 0x0A;break;case StunTypeBlocked:cout << "Blocked or could not reach STUN server";retval[nic] = 0x0C;break;default:cout << stype;cout << "Unkown NAT type";retval[nic] = 0x0E;  // Unknown NAT typebreak;}cout << "\t";cout.flush();if (!hairpin) {retval[nic] |= 0x10;}if (presPort) {retval[nic] |= 0x01;}} else if (testNum == 100) {

可以看到这个app主要做了3件事情：

  1. 解析参数。主要从参数中获得STUN server的地址，及本地用于发送数据包所用的UDP端口号。

  2. 调用stunNatType()函数判断NatType。判断NatType的全部逻辑都在这个函数里。

  3. 将stunNatType()函数返回的NatType进行格式化并打印输出，以便于人的阅读。

接着来看stunNatType()函数的实现

stunNatType()函数的实现

stunNatType()函数的实现如下：

NatType stunNatType(StunAddress4& dest, bool verbose, bool* preservePort,  // if set, is return for if NAT preservers ports or notbool* hairpin,  // if set, is the return for if NAT will hairpin packetsint port,  // port to use for the test, 0 to choose random portStunAddress4* sAddr  // NIC to use) {assert( dest.addr != 0);assert( dest.port != 0);if (hairpin) {*hairpin = false;}if (port == 0) {port = stunRandomPort();}UInt32 interfaceIp = 0;if (sAddr) {interfaceIp = sAddr->addr;}Socket myFd1 = openPort(port, interfaceIp, verbose);Socket myFd2 = openPort(port + 1, interfaceIp, verbose);if ((myFd1 == INVALID_SOCKET) || (myFd2 == INVALID_SOCKET)) {cerr << "Some problem opening port/interface to send on" << endl;return StunTypeFailure;}assert( myFd1 != INVALID_SOCKET);assert( myFd2 != INVALID_SOCKET);bool respTestI = false;bool isNat = true;StunAddress4 testImappedAddr;bool respTestI2 = false;bool mappedIpSame = true;StunAddress4 testI2mappedAddr;StunAddress4 testI2dest = dest;bool respTestII = false;bool respTestIII = false;bool respTestHairpin = false;bool respTestPreservePort = false;memset(&testImappedAddr, 0, sizeof(testImappedAddr));StunAtrString username;StunAtrString password;username.sizeValue = 0;password.sizeValue = 0;#ifdef USE_TLS stunGetUserNameAndPassword( dest, username, password );
#endifint count = 0;while (count < 7) {struct timeval tv;fd_set fdSet;
#ifdef WIN32unsigned int fdSetSize;
#elseint fdSetSize;
#endifFD_ZERO(&fdSet);fdSetSize = 0;FD_SET(myFd1, &fdSet);fdSetSize = (myFd1 + 1 > fdSetSize) ? myFd1 + 1 : fdSetSize;FD_SET(myFd2, &fdSet);fdSetSize = (myFd2 + 1 > fdSetSize) ? myFd2 + 1 : fdSetSize;tv.tv_sec = 0;tv.tv_usec = 150 * 1000;  // 150 msif (count == 0)tv.tv_usec = 0;int err = select(fdSetSize, &fdSet, NULL, NULL, &tv);int e = getErrno();if (err == SOCKET_ERROR) {// error occuredcerr << "Error " << e << " " << strerror(e) << " in select" << endl;return StunTypeFailure;} else if (err == 0) {// timeout occuredcount++;if (!respTestI) {stunSendTest(myFd1, dest, username, password, 1, verbose);}if ((!respTestI2) && respTestI) {// check the address to send to if validif ((testI2dest.addr != 0) && (testI2dest.port != 0)) {stunSendTest(myFd1, testI2dest, username, password, 10, verbose);}}if (!respTestII) {stunSendTest(myFd2, dest, username, password, 2, verbose);}if (!respTestIII) {stunSendTest(myFd2, dest, username, password, 3, verbose);}if (respTestI && (!respTestHairpin)) {if ((testImappedAddr.addr != 0) && (testImappedAddr.port != 0)) {stunSendTest(myFd1, testImappedAddr, username, password, 11, verbose);}}} else {//if (verbose) clog << "-----------------------------------------" << endl;assert( err>0);// data is avialbe on some fdfor (int i = 0; i < 2; i++) {Socket myFd;if (i == 0) {myFd = myFd1;} else {myFd = myFd2;}if (myFd != INVALID_SOCKET) {if (FD_ISSET(myFd,&fdSet)) {char msg[STUN_MAX_MESSAGE_SIZE];int msgLen = sizeof(msg);StunAddress4 from;getMessage(myFd, msg, &msgLen, &from.addr, &from.port, verbose);StunMessage resp;memset(&resp, 0, sizeof(StunMessage));stunParseMessage(msg, msgLen, resp, verbose);if (verbose) {clog << "Received message of type " << resp.msgHdr.msgType << "  id="<< (int) (resp.msgHdr.id.octet[0]) << endl;}switch (resp.msgHdr.id.octet[0]) {case 1: {if (!respTestI) {testImappedAddr.addr = resp.mappedAddress.ipv4.addr;testImappedAddr.port = resp.mappedAddress.ipv4.port;respTestPreservePort = (testImappedAddr.port == port);if (preservePort) {*preservePort = respTestPreservePort;}testI2dest.addr = resp.changedAddress.ipv4.addr;if (sAddr) {sAddr->port = testImappedAddr.port;sAddr->addr = testImappedAddr.addr;}count = 0;}respTestI = true;}break;case 2: {respTestII = true;}break;case 3: {respTestIII = true;}break;case 10: {if (!respTestI2) {testI2mappedAddr.addr = resp.mappedAddress.ipv4.addr;testI2mappedAddr.port = resp.mappedAddress.ipv4.port;mappedIpSame = false;if ((testI2mappedAddr.addr == testImappedAddr.addr)&& (testI2mappedAddr.port == testImappedAddr.port)) {mappedIpSame = true;}}respTestI2 = true;}break;case 11: {if (hairpin) {*hairpin = true;}respTestHairpin = true;}break;}}}}}}// see if we can bind to this address//cerr << "try binding to " << testImappedAddr << endl;Socket s = openPort(0/*use ephemeral*/, testImappedAddr.addr, false);if (s != INVALID_SOCKET) {closesocket(s);isNat = false;//cerr << "binding worked" << endl;} else {isNat = true;//cerr << "binding failed" << endl;}if (verbose) {clog << "test I = " << respTestI << endl;clog << "test II = " << respTestII << endl;clog << "test III = " << respTestIII << endl;clog << "test I(2) = " << respTestI2 << endl;clog << "is nat  = " << isNat << endl;clog << "mapped IP same = " << mappedIpSame << endl;clog << "hairpin = " << respTestHairpin << endl;clog << "preserver port = " << respTestPreservePort << endl;}#if 0// implement logic flow chart from draft RFCif (respTestI) {if (isNat) {if (respTestII) {return StunTypeConeNat;} else {if (mappedIpSame) {if (respTestIII) {return StunTypeRestrictedNat;} else {return StunTypePortRestrictedNat;}} else {return StunTypeSymNat;}}} else {if (respTestII) {return StunTypeOpen;} else {return StunTypeSymFirewall;}}} else {return StunTypeBlocked;}
#elseif (respTestI) {  // not blockedif (isNat) {if (mappedIpSame) {if (respTestII) {return StunTypeIndependentFilter;} else {if (respTestIII) {return StunTypeDependentFilter;} else {return StunTypePortDependedFilter;}}} else {  // mappedIp is not samereturn StunTypeDependentMapping;}} else {  // isNat is falseif (respTestII) {return StunTypeOpen;} else {return StunTypeFirewall;}}} else {return StunTypeBlocked;}
#endifreturn StunTypeUnknown;
}

可以看到这个函数主要做了几件事：

  1. 打开了两个UDP socket。后续会通过这两个socket来进行数据包的发送，并最终根据这些数据包的响应数据包的情况来判断NatType。

  2. 向STUN server发送请求。调用stunSendTest()函数发送了5种不同类型的消息，各个消息之间的差异也仅仅在与stunSendTest()函数的testNum参数不同。这里我们也用testNum来区分不同的消息，我们称它们分别为类型1，类型2，类型3，类型10及类型11的消息。
    其中类型10和类型11的消息依赖于类型1的消息的响应，但类型2和类型3的消息的发送则与类型1的消息的发送及响应相互独立，因而它们可以与类型1的消息并行的发送。

  3. 接收发送的消息的响应。
    从类型1的消息的响应中获得的东西比较多。类型10和类型11的消息要发送的目标地址，都来源于类型1的消息的响应。
    类型10的消息发向类型1的消息的响应的changedAddress地址。这个地址是STUN server的副IP地址及端口号。
    类型11的消息则发向类型1的消息的响应的testImappedAddr地址，这个地址是发送消息的地址的出口公网地址，向这个消息发送消息实际是向本节点在发送消息，这么做的实际目的是为了测试节点所连接的NAT是否支持消息的回传，或者说测试NAT是否是hairpin的。即如果这个类型11的消息通过NAT并最终被发送给本节点且本节点接收到了这个消息，则说明本节点所连接的NAT是hairpin的。
    STUN终端会从类型10的消息的响应中获得相同的本地网络地址到另外的网络地址（IP地址与类型1的目标IP地址不同）的出口公网地址，并用这个地址与类型1的响应中携带的那个出口公网地址进行比较，以此来判断当前节点所连接的NAT是否是对称型的。
    除了类型1和类型10之外，发送其它的消息主要就是看看是否能获得对应的响应。

  4. 根据发送的这5种不同类型的消息的响应来判断当前节点所连接的NAT的类型并返回给调用者。

下面我们再用几张图来详细地说明，这些消息都发到了哪里，而响应又是从哪里返回回来的。

先说明一下，stund的STUN Server需要部署在一台具有双网卡且每个网卡都有一个自己公网IP地址的主机上。STUN Server的两个IP可以称为IPAddr1（primary IP）和IPAddr2（alt IP），两个端口可以称为Port1（primary port）和Port2（alt port），这两个端口默认分别为3478和3479。STUN Server会打开4个sockets，每个IP两个分别对应两个不同的端口。

首先是消息1：


160644_Zta3_919237.png

消息1从客户端的第一个端口Port1发向STUN Server的IPAddr1:Port1，响应中则会携带客户端发送消息的端口的出口网络地址，及IPAddr2:Port2，以为后续发送消息10及消息11做准备。

消息2：


161232_AfFx_919237.png

消息2从客户端的第二个端口，发向STUN Server的IPAddr1:Port1，这个消息请求STUN Server将响应从它的IPAddr2:Port1发送回来，也就是相对于接收数据包的网络地址而言切换一下IP地址的网络地址。

发送这个消息的目的是什么呢？这个消息的响应如果能接收到的话，说明当前节点连接的NAT的类型为全锥型的，说明NAT对于发向其内部的主机的数据包几乎没有限制。

这里为什么要从第二个端口发送消息呢？这主要是因为，类型10的消息会发向IPAddr2:Port1，这实际上会对消息2的响应的接收产生干扰。如果一个地址向IPAddr2:Port1发送了消息，即使当前节点连接的NAT的类型不是全锥型的，从IPAddr2:Port1发回来的消息也可能被接收到。

消息3：


161329_M7lo_919237.png

消息3同样从客户端的第二个端口发出，且同样发向STUN Server的IPAddr1:Port1，但这个消息请求STUN Server将响应从它的IPAddr1:Port2发送回来，也就是相对于接收数据包的网络地址而言切换一下端口的网络地址。

在消息2的响应接收不到的情况下，如果消息3的响应可以接收到，说明NAT对传入给内部主机的包是限制IP而不限制端口的，也就是说当前节点连接的NAT的类型是IP限制型的。

消息4：


161413_TSdS_919237.png

针对多主机部署的STUN Server优化

由上面的过程，不难看到，STUN Server的部署有一个比较大的限制，即要求部署的主机具有双网卡，这对于我们当前遍地云主机的环境而言，部署起来是不那么方便的。主要是对于类型2的消息，客户端请求STUN Server切换一下IP地址将消息发回来。

因而一种用于stund的STUN Server的优化设计应运而生，结构如下图：


162407_B2RO_919237.png

这种设计主要是让STUN Server只绑定一个IP上的两个端口，同时在STUN之间建立一个通信信道，以便于类型2的消息能得到合适的处理。

针对多主机部署的STUN Server的优化当前实现的状况：
Github主页：https://github.com/hanpfei/stund

STUN消息的格式

具体可多主机部署的STUN Server要如何设计？这还要从STUN消息的具体格式说起。接着来看下STUN消息的具体格式。

首先是客户端发送的请求的格式。我们可以通过stunSendTest()函数的实现来对这个问题做一番了解：

static void stunSendTest(Socket myFd, StunAddress4& dest, const StunAtrString& username, const StunAtrString& password,int testNum, bool verbose) {assert( dest.addr != 0);assert( dest.port != 0);bool changePort = false;bool changeIP = false;bool discard = false;switch (testNum) {case 1:case 10:case 11:break;case 2://changePort=true;changeIP = true;break;case 3:changePort = true;break;case 4:changeIP = true;break;case 5:discard = true;break;default:cerr << "Test " << testNum << " is unkown\n";assert(0);}StunMessage req;memset(&req, 0, sizeof(StunMessage));stunBuildReqSimple(&req, username, changePort, changeIP, testNum);char buf[STUN_MAX_MESSAGE_SIZE];int len = STUN_MAX_MESSAGE_SIZE;len = stunEncodeMessage(req, buf, len, password, verbose);if (verbose) {clog << "About to send msg of len " << len << " to " << dest << endl;}sendMessage(myFd, buf, len, dest.addr, dest.port, verbose);// add some delay so the packets don't get sent too quickly
#ifdef WIN32 // !cj! TODO - should fix this up in windowsclock_t now = clock();assert( CLOCKS_PER_SEC == 1000 );while ( clock() <= now+10 ) {};
#elseusleep(10 * 1000);
#endif}

从这里似乎也得不到太多STUN消息格式的具体信息，细节都被放在stunBuildReqSimple()和stunEncodeMessage()两个函数中了，接着来看这两个函数的实现：

static char*
encodeAtrChangeRequest(char* ptr, const StunAtrChangeRequest& atr) {ptr = encode16(ptr, ChangeRequest);ptr = encode16(ptr, 4);ptr = encode32(ptr, atr.value);return ptr;
}. . . . . .unsigned int stunEncodeMessage(const StunMessage& msg, char* buf, unsigned int bufLen, const StunAtrString& password,bool verbose) {assert(bufLen >= sizeof(StunMsgHdr));char* ptr = buf;ptr = encode16(ptr, msg.msgHdr.msgType);char* lengthp = ptr;ptr = encode16(ptr, 0);ptr = encode(ptr, reinterpret_cast<const char*>(msg.msgHdr.id.octet), sizeof(msg.msgHdr.id));if (verbose)clog << "Encoding stun message: " << endl;if (msg.hasMappedAddress) {if (verbose)clog << "Encoding MappedAddress: " << msg.mappedAddress.ipv4 << endl;ptr = encodeAtrAddress4(ptr, MappedAddress, msg.mappedAddress);}if (msg.hasResponseAddress) {if (verbose)clog << "Encoding ResponseAddress: " << msg.responseAddress.ipv4 << endl;ptr = encodeAtrAddress4(ptr, ResponseAddress, msg.responseAddress);}if (msg.hasChangeRequest) {if (verbose)clog << "Encoding ChangeRequest: " << msg.changeRequest.value << endl;ptr = encodeAtrChangeRequest(ptr, msg.changeRequest);}if (msg.hasSourceAddress) {if (verbose)clog << "Encoding SourceAddress: " << msg.sourceAddress.ipv4 << endl;ptr = encodeAtrAddress4(ptr, SourceAddress, msg.sourceAddress);}if (msg.hasChangedAddress) {if (verbose)clog << "Encoding ChangedAddress: " << msg.changedAddress.ipv4 << endl;ptr = encodeAtrAddress4(ptr, ChangedAddress, msg.changedAddress);}if (msg.hasUsername) {if (verbose)clog << "Encoding Username: " << msg.username.value << endl;ptr = encodeAtrString(ptr, Username, msg.username);}if (msg.hasPassword) {if (verbose)clog << "Encoding Password: " << msg.password.value << endl;ptr = encodeAtrString(ptr, Password, msg.password);}if (msg.hasErrorCode) {if (verbose)clog << "Encoding ErrorCode: class=" << int(msg.errorCode.errorClass) << " number="<< int(msg.errorCode.number) << " reason=" << msg.errorCode.reason << endl;ptr = encodeAtrError(ptr, msg.errorCode);}if (msg.hasUnknownAttributes) {if (verbose)clog << "Encoding UnknownAttribute: ???" << endl;ptr = encodeAtrUnknown(ptr, msg.unknownAttributes);}if (msg.hasReflectedFrom) {if (verbose)clog << "Encoding ReflectedFrom: " << msg.reflectedFrom.ipv4 << endl;ptr = encodeAtrAddress4(ptr, ReflectedFrom, msg.reflectedFrom);}if (msg.hasXorMappedAddress) {if (verbose)clog << "Encoding XorMappedAddress: " << msg.xorMappedAddress.ipv4 << endl;ptr = encodeAtrAddress4(ptr, XorMappedAddress, msg.xorMappedAddress);}if (msg.xorOnly) {if (verbose)clog << "Encoding xorOnly: " << endl;ptr = encodeXorOnly(ptr);}if (msg.hasServerName) {if (verbose)clog << "Encoding ServerName: " << msg.serverName.value << endl;ptr = encodeAtrString(ptr, ServerName, msg.serverName);}if (msg.hasSecondaryAddress) {if (verbose)clog << "Encoding SecondaryAddress: " << msg.secondaryAddress.ipv4 << endl;ptr = encodeAtrAddress4(ptr, SecondaryAddress, msg.secondaryAddress);}if (password.sizeValue > 0) {if (verbose)clog << "HMAC with password: " << password.value << endl;StunAtrIntegrity integrity;computeHmac(integrity.hash, buf, int(ptr - buf), password.value, password.sizeValue);ptr = encodeAtrIntegrity(ptr, integrity);}if (verbose)clog << endl;encode16(lengthp, UInt16(ptr - buf - sizeof(StunMsgHdr)));return int(ptr - buf);
}void stunBuildReqSimple(StunMessage* msg, const StunAtrString& username, bool changePort, bool changeIp,unsigned int id) {assert( msg);memset(msg, 0, sizeof(*msg));msg->msgHdr.msgType = BindRequestMsg;for (int i = 0; i < 16; i = i + 4) {assert(i+3<16);int r = stunRand();msg->msgHdr.id.octet[i + 0] = r >> 0;msg->msgHdr.id.octet[i + 1] = r >> 8;msg->msgHdr.id.octet[i + 2] = r >> 16;msg->msgHdr.id.octet[i + 3] = r >> 24;}if (id != 0) {msg->msgHdr.id.octet[0] = id;}msg->hasChangeRequest = true;msg->changeRequest.value = (changeIp ? ChangeIpFlag : 0) | (changePort ? ChangePortFlag : 0);if (username.sizeValue > 0) {msg->hasUsername = true;msg->username = username;}
}

由这些函数的实现，当不难理出来STUN请求消息的格式大体为：


170658_ZxCq_919237.png

整体来看，STUN请求消息分为两个部分，一部分是Header，另一部分是Attr的List。

而Header又包含消息的类型，消息不包含Header的长度，及一个128位16字节的id。在stund中，id的首个字节保存了消息的类型。STUN Server会原封不动的将客户端发过去的消息的id包含在响应中发回给客户端，在stund中，使用了id的首个字节用以区分发出去的不同类型的消息的响应。

Attr的List则是一系列的Attr。Attr的结构大体为，先是一个16位的AttrType，然后是16位的Attr值长度，接着便是Attr的值，而Attr的值所占字节数因Attr的不同而不同。对于判断NatType这个case而言，AttrList中只有一个Attr，及类型为ChangeRequest的Attr，它有一个32位4字节的值。这个Attr用于告诉STUN Server，响应应该从哪个网络地址发回来。

看完了STUN请求消息的格式之后，接着再来看STUN响应消息的格式。这个我们可以从stunServerProcessMsg()函数的实现来了解：

bool stunServerProcessMsg(char* buf, unsigned int bufLen, StunAddress4& from, StunAddress4& secondary,StunAddress4& myAddr, StunAddress4& altAddr, StunMessage* resp, StunAddress4* destination,StunAtrString* hmacPassword, bool* changePort, bool* changeIp, bool verbose) {// set up information for default responsememset(resp, 0, sizeof(*resp));*changeIp = false;*changePort = false;StunMessage req;bool ok = stunParseMessage(buf, bufLen, req, verbose);if (!ok) {      // Complete garbage, drop it on the floorif (verbose)clog << "Request did not parse" << endl;return false;}if (verbose)clog << "Request parsed ok" << endl;StunAddress4 mapped = req.mappedAddress.ipv4;StunAddress4 respondTo = req.responseAddress.ipv4;UInt32 flags = req.changeRequest.value;switch (req.msgHdr.msgType) {case SharedSecretRequestMsg:if (verbose)clog << "Received SharedSecretRequestMsg on udp. send error 433." << endl;// !cj! - should fix so you know if this came over TLS or UDPstunCreateSharedSecretResponse(req, from, *resp);//stunCreateSharedSecretErrorResponse(*resp, 4, 33, "this request must be over TLS");return true;case BindRequestMsg:if (!req.hasMessageIntegrity) {if (verbose)clog << "BindRequest does not contain MessageIntegrity" << endl;if (0) {  // !jf! mustAuthenticateif (verbose)clog << "Received BindRequest with no MessageIntegrity. Sending 401." << endl;stunCreateErrorResponse(*resp, 4, 1, "Missing MessageIntegrity");return true;}} else {if (!req.hasUsername) {if (verbose)clog << "No UserName. Send 432." << endl;stunCreateErrorResponse(*resp, 4, 32, "No UserName and contains MessageIntegrity");return true;} else {if (verbose)clog << "Validating username: " << req.username.value << endl;// !jf! could retrieve associated password from provisioning hereif (strcmp(req.username.value, "test") == 0) {if (0) {// !jf! if the credentials are stalestunCreateErrorResponse(*resp, 4, 30, "Stale credentials on BindRequest");return true;} else {if (verbose)clog << "Validating MessageIntegrity" << endl;// need access to shared secretunsigned char hmac[20];
#ifndef NOSSLunsigned int hmacSize=20;HMAC(EVP_sha1(),"1234", 4,reinterpret_cast<const unsigned char*>(buf), bufLen-20-4,hmac, &hmacSize);assert(hmacSize == 20);
#endifif (memcmp(buf, hmac, 20) != 0) {if (verbose)clog << "MessageIntegrity is bad. Sending " << endl;stunCreateErrorResponse(*resp, 4, 3, "Unknown username. Try test with password 1234");return true;}// need to compute this later after message is filled inresp->hasMessageIntegrity = true;assert(req.hasUsername);resp->hasUsername = true;resp->username = req.username;  // copy username in}} else {if (verbose)clog << "Invalid username: " << req.username.value << "Send 430." << endl;}}}// TODO !jf! should check for unknown attributes here and send 420 listing the// unknown attributes.if (respondTo.port == 0)respondTo = from;if (mapped.port == 0)mapped = from;*changeIp = (flags & ChangeIpFlag) ? true : false;*changePort = (flags & ChangePortFlag) ? true : false;if (verbose) {clog << "Request is valid:" << endl;clog << "\t flags=" << flags << endl;clog << "\t changeIp=" << *changeIp << endl;clog << "\t changePort=" << *changePort << endl;clog << "\t from = " << from << endl;clog << "\t respond to = " << respondTo << endl;clog << "\t mapped = " << mapped << endl;}// form the outgoing messageresp->msgHdr.msgType = BindResponseMsg;for (int i = 0; i < 16; i++) {resp->msgHdr.id.octet[i] = req.msgHdr.id.octet[i];}if (req.xorOnly == false) {resp->hasMappedAddress = true;resp->mappedAddress.ipv4.port = mapped.port;resp->mappedAddress.ipv4.addr = mapped.addr;}if (1) {  // do xorMapped address or notresp->hasXorMappedAddress = true;UInt16 id16 = req.msgHdr.id.octet[0] << 8 | req.msgHdr.id.octet[1];UInt32 id32 = req.msgHdr.id.octet[0] << 24 | req.msgHdr.id.octet[1] << 16 | req.msgHdr.id.octet[2] << 8| req.msgHdr.id.octet[3];resp->xorMappedAddress.ipv4.port = mapped.port ^ id16;resp->xorMappedAddress.ipv4.addr = mapped.addr ^ id32;}resp->hasSourceAddress = true;resp->sourceAddress.ipv4.port = (*changePort) ? altAddr.port : myAddr.port;resp->sourceAddress.ipv4.addr = (*changeIp) ? altAddr.addr : myAddr.addr;resp->hasChangedAddress = true;resp->changedAddress.ipv4.port = altAddr.port;resp->changedAddress.ipv4.addr = altAddr.addr;if (secondary.port != 0) {resp->hasSecondaryAddress = true;resp->secondaryAddress.ipv4.port = secondary.port;resp->secondaryAddress.ipv4.addr = secondary.addr;}if (req.hasUsername && req.username.sizeValue > 0) {// copy username inresp->hasUsername = true;assert( req.username.sizeValue % 4 == 0);assert( req.username.sizeValue < STUN_MAX_STRING);memcpy(resp->username.value, req.username.value, req.username.sizeValue);resp->username.sizeValue = req.username.sizeValue;}if (1) {  // add ServerNameresp->hasServerName = true;const char serverName[] = "Vovida.org " STUN_VERSION;  // must pad to mult of 4assert( sizeof(serverName) < STUN_MAX_STRING);//cerr << "sizeof serverName is "  << sizeof(serverName) << endl;assert( sizeof(serverName)%4 == 0);memcpy(resp->serverName.value, serverName, sizeof(serverName));resp->serverName.sizeValue = sizeof(serverName);}if (req.hasMessageIntegrity & req.hasUsername) {// this creates the password that will be used in the HMAC when then// messages is sentstunCreatePassword(req.username, hmacPassword);}if (req.hasUsername && (req.username.sizeValue > 64)) {UInt32 source;assert( sizeof(int) == sizeof(UInt32));sscanf(req.username.value, "%x", &source);resp->hasReflectedFrom = true;resp->reflectedFrom.ipv4.port = 0;resp->reflectedFrom.ipv4.addr = source;}destination->port = respondTo.port;destination->addr = respondTo.addr;return true;default:if (verbose)clog << "Unknown or unsupported request " << endl;return false;}assert(0);return false;
}

由这个函数的实现，我们不难看出STUN Server发回给客户端的响应的消息格式与请求的格式大体一样，但消息的具体内容有一些区别。消息的格式大体为：


174120_pPyU_919237.png

这个消息里的内容要多一点。

了解了STUN客户端和STUN Server间交互的这些UDP数据包的格式之后，我们就可以确定可双主机部署的STUN Server间通信的消息的格式了。

仔细来看stunServerProcessMsg()，我们注意到，STUN server响应发送的目标地址，以及返回给客户端的它的出口公网地址也就是mappedAddress也没有限定只能是from地址，这些值也可以来源于请求消息。

借助于stund的这些良好设计，可以大大简化我们的可双主机部署的STUN server的设计与实现。STUN server间的消息格式可以为：


182922_6Bzw_919237.png

也就是说，当STUN Server收到类型2的消息时，构造一个格式如上图的消息，并将该消息转发给另为一个STUN Server。其中MappedAddress和ResponseAddress Attr的值都是消息的from地址，即客户端发送消息的端口的出口公网地址。

经过对stunServerProcessMsg()的一番改造，终于可以实现STUN Server的多主机部署，其改造后的实现为：

bool stunServerProcessMsg(StunServerInfo& info, char* buf, unsigned int bufLen, StunAddress4& from,StunAddress4& secondary, StunAddress4& myAddr, StunAddress4& altAddr, StunMessage* resp,StunAddress4* destination, StunAtrString* hmacPassword, bool* changePort, bool* changeIp,bool verbose) {// set up information for default responsememset(resp, 0, sizeof(*resp));*changeIp = false;*changePort = false;StunMessage req;bool ok = stunParseMessage(buf, bufLen, req, verbose);if (!ok) {      // Complete garbage, drop it on the floorif (verbose)clog << "Request did not parse" << endl;return false;}if (verbose)clog << "Request parsed ok" << endl;StunAddress4 mapped = req.mappedAddress.ipv4;StunAddress4 respondTo = req.responseAddress.ipv4;UInt32 flags = req.changeRequest.value;switch (req.msgHdr.msgType) {case SharedSecretRequestMsg:if (verbose)clog << "Received SharedSecretRequestMsg on udp. send error 433." << endl;// !cj! - should fix so you know if this came over TLS or UDPstunCreateSharedSecretResponse(req, from, *resp);//stunCreateSharedSecretErrorResponse(*resp, 4, 33, "this request must be over TLS");return true;case BindRequestMsg:if (!req.hasMessageIntegrity) {if (verbose)clog << "BindRequest does not contain MessageIntegrity" << endl;if (0) {  // !jf! mustAuthenticateif (verbose)clog << "Received BindRequest with no MessageIntegrity. Sending 401." << endl;stunCreateErrorResponse(*resp, 4, 1, "Missing MessageIntegrity");return true;}} else {if (!req.hasUsername) {if (verbose)clog << "No UserName. Send 432." << endl;stunCreateErrorResponse(*resp, 4, 32, "No UserName and contains MessageIntegrity");return true;} else {if (verbose)clog << "Validating username: " << req.username.value << endl;// !jf! could retrieve associated password from provisioning hereif (strcmp(req.username.value, "test") == 0) {if (0) {// !jf! if the credentials are stalestunCreateErrorResponse(*resp, 4, 30, "Stale credentials on BindRequest");return true;} else {if (verbose)clog << "Validating MessageIntegrity" << endl;// need access to shared secretunsigned char hmac[20];
#ifndef NOSSLunsigned int hmacSize=20;HMAC(EVP_sha1(),"1234", 4,reinterpret_cast<const unsigned char*>(buf), bufLen-20-4,hmac, &hmacSize);assert(hmacSize == 20);
#endifif (memcmp(buf, hmac, 20) != 0) {if (verbose)clog << "MessageIntegrity is bad. Sending " << endl;stunCreateErrorResponse(*resp, 4, 3, "Unknown username. Try test with password 1234");return true;}// need to compute this later after message is filled inresp->hasMessageIntegrity = true;assert(req.hasUsername);resp->hasUsername = true;resp->username = req.username;  // copy username in}} else {if (verbose)clog << "Invalid username: " << req.username.value << "Send 430." << endl;}}}// TODO !jf! should check for unknown attributes here and send 420 listing the// unknown attributes.if (respondTo.port == 0)respondTo = from;if (mapped.port == 0)mapped = from;*changeIp = (flags & ChangeIpFlag) ? true : false;*changePort = (flags & ChangePortFlag) ? true : false;if (verbose) {clog << "Request is valid:" << endl;clog << "\t flags=" << flags << endl;clog << "\t changeIp=" << *changeIp << endl;clog << "\t changePort=" << *changePort << endl;clog << "\t from = " << from << endl;clog << "\t respond to = " << respondTo << endl;clog << "\t mapped = " << mapped << endl;}// form the outgoing messagefor (int i = 0; i < 16; i++) {resp->msgHdr.id.octet[i] = req.msgHdr.id.octet[i];}if (*changeIp && info.altIpFd == INVALID_SOCKET) {resp->msgHdr.msgType = req.msgHdr.msgType;*changeIp = false;*changePort = false;resp->hasChangeRequest = true;resp->changeRequest.value = changePort ? ChangePortFlag : 0;resp->hasMappedAddress = true;resp->mappedAddress.ipv4.port = mapped.port;resp->mappedAddress.ipv4.addr = mapped.addr;resp->hasResponseAddress = true;resp->responseAddress.ipv4.port = from.port;resp->responseAddress.ipv4.addr = from.addr;respondTo.port = info.myAddr.port;respondTo.addr = info.altAddr.addr;if (verbose) {clog << "\t respondTo change = " << respondTo << endl;}} else {resp->msgHdr.msgType = BindResponseMsg;if (req.xorOnly == false) {resp->hasMappedAddress = true;resp->mappedAddress.ipv4.port = mapped.port;resp->mappedAddress.ipv4.addr = mapped.addr;}if (1) {  // do xorMapped address or notresp->hasXorMappedAddress = true;UInt16 id16 = req.msgHdr.id.octet[0] << 8 | req.msgHdr.id.octet[1];UInt32 id32 = req.msgHdr.id.octet[0] << 24 | req.msgHdr.id.octet[1] << 16| req.msgHdr.id.octet[2] << 8 | req.msgHdr.id.octet[3];resp->xorMappedAddress.ipv4.port = mapped.port ^ id16;resp->xorMappedAddress.ipv4.addr = mapped.addr ^ id32;}resp->hasSourceAddress = true;resp->sourceAddress.ipv4.port = (*changePort) ? altAddr.port : myAddr.port;resp->sourceAddress.ipv4.addr = (*changeIp) ? altAddr.addr : myAddr.addr;resp->hasChangedAddress = true;resp->changedAddress.ipv4.port = altAddr.port;resp->changedAddress.ipv4.addr = altAddr.addr;if (secondary.port != 0) {resp->hasSecondaryAddress = true;resp->secondaryAddress.ipv4.port = secondary.port;resp->secondaryAddress.ipv4.addr = secondary.addr;}if (req.hasUsername && req.username.sizeValue > 0) {// copy username inresp->hasUsername = true;assert( req.username.sizeValue % 4 == 0);assert( req.username.sizeValue < STUN_MAX_STRING);memcpy(resp->username.value, req.username.value, req.username.sizeValue);resp->username.sizeValue = req.username.sizeValue;}if (1) {  // add ServerNameresp->hasServerName = true;const char serverName[] = "Vovida.org " STUN_VERSION;  // must pad to mult of 4assert( sizeof(serverName) < STUN_MAX_STRING);//cerr << "sizeof serverName is "  << sizeof(serverName) << endl;assert( sizeof(serverName)%4 == 0);memcpy(resp->serverName.value, serverName, sizeof(serverName));resp->serverName.sizeValue = sizeof(serverName);}if (req.hasMessageIntegrity & req.hasUsername) {// this creates the password that will be used in the HMAC when then// messages is sentstunCreatePassword(req.username, hmacPassword);}if (req.hasUsername && (req.username.sizeValue > 64)) {UInt32 source;assert( sizeof(int) == sizeof(UInt32));sscanf(req.username.value, "%x", &source);resp->hasReflectedFrom = true;resp->reflectedFrom.ipv4.port = 0;resp->reflectedFrom.ipv4.addr = source;}}destination->port = respondTo.port;destination->addr = respondTo.addr;return true;default:if (verbose)clog << "Unknown or unsupported request " << endl;return false;}assert(0);return false;
}

主要的改动即是在发现客户端请求改变IP地址发回响应时，构造如上图中的消息，并发给另一个STUN Server。从而，对于消息2，数据包的流转过程大体如下：


140802_Enu2_919237.png

Done。

这篇关于标准STUN判断NAT类型的过程及改进的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！


http://www.chinasem.cn/article/921323。

相关文章

浅析Spring Security认证过程

浅析Spring Security认证过程

类图 为了方便理解Spring Security认证流程，特意画了如下的类图，包含相关的核心认证类 概述 核心验证器 AuthenticationManager 该对象提供了认证方法的入口，接收一个Authentiaton对象作为参数; public interface AuthenticationManager {Authentication authenticate(Authenti
阅读更多...
作业提交过程之HDFSMapReduce

作业提交过程之HDFSMapReduce

作业提交全过程详解 （1）作业提交 第1步：Client调用job.waitForCompletion方法，向整个集群提交MapReduce作业。 第2步：Client向RM申请一个作业id。 第3步：RM给Client返回该job资源的提交路径和作业id。 第4步：Client提交jar包、切片信息和配置文件到指定的资源提交路径。 第5步：Client提交完资源后，向RM申请运行MrAp
阅读更多...
零基础学习Redis(10) -- zset类型命令使用

零基础学习Redis(10) -- zset类型命令使用

zset是有序集合，内部除了存储元素外，还会存储一个score，存储在zset中的元素会按照score的大小升序排列，不同元素的score可以重复，score相同的元素会按照元素的字典序排列。 1. zset常用命令 1.1 zadd  zadd key [NX | XX] [GT | LT]   [CH] [INCR] score member [score member ...]
阅读更多...
poj 3259 uva 558 Wormholes(bellman最短路负权回路判断)

poj 3259 uva 558 Wormholes(bellman最短路负权回路判断)

poj 3259： 题意：John的农场里n块地，m条路连接两块地，w个虫洞，虫洞是一条单向路，不但会把你传送到目的地，而且时间会倒退Ts。 任务是求你会不会在从某块地出发后又回来，看到了离开之前的自己。 判断树中是否存在负权回路就ok了。 bellman代码： #include<stdio.h>const int MaxN = 501;//农场数const int
阅读更多...
【机器学习】高斯过程的基本概念和应用领域以及在python中的实例

【机器学习】高斯过程的基本概念和应用领域以及在python中的实例

引言 高斯过程（Gaussian Process，简称GP）是一种概率模型，用于描述一组随机变量的联合概率分布，其中任何一个有限维度的子集都具有高斯分布 文章目录 引言一、高斯过程1.1 基本定义1.1.1 随机过程1.1.2 高斯分布 1.2 高斯过程的特性1.2.1 联合高斯性1.2.2 均值函数1.2.3 协方差函数（或核函数） 1.3 核函数1.4 高斯过程回归（Gauss
阅读更多...
zoj 1721 判断2条线段（完全）相交

zoj 1721 判断2条线段（完全）相交

给出起点，终点，与一些障碍线段。 求起点到终点的最短路。 枚举2点的距离，然后最短路。 2点可达条件：没有线段与这2点所构成的线段（完全）相交。 const double eps = 1e-8 ;double add(double x , double y){if(fabs(x+y) < eps*(fabs(x) + fabs(y))) return 0 ;return x + y ;
阅读更多...
POJ1269 判断2条直线的位置关系

POJ1269 判断2条直线的位置关系

题目大意：给两个点能够确定一条直线，题目给出两条直线（由4个点确定），要求判断出这两条直线的关系：平行，同线，相交。如果相交还要求出交点坐标。 解题思路： 先判断两条直线p1p2, q1q2是否共线， 如果不是，再判断 直线 是否平行， 如果还不是， 则两直线相交。  判断共线：  p1p2q1 共线 且 p1p2q2 共线 ，共线用叉乘为 0  来判断，  判断 平行：  p1p
阅读更多...
Codeforces Round #113 (Div. 2) B 判断多边形是否在凸包内

Codeforces Round #113 (Div. 2) B 判断多边形是否在凸包内

题目点击打开链接 凸多边形A， 多边形B， 判断B是否严格在A内。  注意AB有重点 。  将A,B上的点合在一起求凸包，如果凸包上的点是B的某个点，则B肯定不在A内。 或者说B上的某点在凸包的边上则也说明B不严格在A里面。 这个处理有个巧妙的方法，只需在求凸包的时候， <=  改成< 也就是说凸包一条边上的所有点都重复点都记录在凸包里面了。 另外不能去重点。 int
阅读更多...
自定义类型：结构体(续)

自定义类型：结构体(续)

目录 一. 结构体的内存对齐 1.1 为什么存在内存对齐？ 1.2 修改默认对齐数 二. 结构体传参 三. 结构体实现位段 一. 结构体的内存对齐 在前面的文章里我们已经讲过一部分的内存对齐的知识，并举出了两个例子，我们再举出两个例子继续说明： struct S3{double a;int b;char c;};int mian(){printf("%zd\n",s
阅读更多...
【编程底层思考】垃圾收集机制，GC算法，垃圾收集器类型概述

【编程底层思考】垃圾收集机制，GC算法，垃圾收集器类型概述

Java的垃圾收集（Garbage Collection，GC）机制是Java语言的一大特色，它负责自动管理内存的回收，释放不再使用的对象所占用的内存。以下是对Java垃圾收集机制的详细介绍： 一、垃圾收集机制概述： 对象存活判断：垃圾收集器定期检查堆内存中的对象，判断哪些对象是“垃圾”，即不再被任何引用链直接或间接引用的对象。内存回收：将判断为垃圾的对象占用的内存进行回收，以便重新使用。
阅读更多...

Copyright China编程 23002807@qq.com

沪ICP备2023033072号-2