最近几周，Android的NSD实现让我感到非常恼火：
从用户角度出发，会出现以下问题：
设备之间的发现方式完全是不确定的。如果我启动基于
NsdManager的应用程序，仅涉及两个设备时它可以更或多少地工作。但是，如果第三个设备加入，它很少会发现前两个设备，而前两个设备则看不到第三个设备。如果我退出应用程序（它们将优雅地注销NSD监听器）并以不同的顺序重新启动它们，则发现模式并不完全相同，但是类似。在我的家庭网络中，所发现的设备的IP解析基本上按预期工作。但是在工作场所，即使只使用两个设备（A和B），设备A有时也会使用A的IP地址和B的端口解析出B的服务，并反之亦然。因此，在较低级别（可能是NsdManager）上似乎将IP地址和服务名称混淆了。
我已经在Google Code上为此提交了一个错误报告(https://code.google.com/p/android/issues/detail?id=201314&thanks=201314&ts=1455814995)。我也在这里发布，希望能够获得更多反馈；也许我在Nsd助手类中有些错误。
首先，在无尽的调试后，我现在在logcat中找到了Android底层的NsdService本身可能存在故障的迹象，而MDnsDS似乎正常工作。但我还不确定...
下面是一些说明问题的日志输出（为了易读性，某些消息已被过滤）：
02-18 16:57:02.327: D/NsdService(628): startMDnsDaemon
02-18 16:57:02.327: D/MDnsDS(187): Starting MDNSD
02-18 16:57:02.529: D/NsdService(628): New client listening to asynchronous messages
02-18 16:57:02.529: D/NsdService(628): New client, channel: com.android.internal.util.AsyncChannel@1fa188ce messenger: android.os.Messenger@cca33ef
02-18 16:57:02.532: D/NsdService(628): Register service
02-18 16:57:02.532: D/NsdService(628): registerService: 106 name: TuSync-0.57392, type: _tusync._tcp., host: /::, port: 57392
02-18 16:57:02.533: D/MDnsDS(187): serviceRegister(106, (null), TuSync-0.57392, _tusync._tcp., (null), (null), 57392, 0, <binary>)
02-18 16:57:02.533: D/MDnsDS(187): serviceRegister successful
02-18 16:57:02.534: D/NsdService(628): Register 1 106
02-18 16:57:04.083: D/MDnsDS(187): register succeeded for 106 as TuSync-0.57392
02-18 16:57:04.087: D/NsdService(628): SERVICE_REGISTERED Raw: 606 106 "TuSync-0.57392"
02-18 16:57:04.109: D/NsdService(628): Discover services
02-18 16:57:04.109: D/NsdService(628): discoverServices: 107 _tusync._tcp.
02-18 16:57:04.110: D/MDnsDS(187): discover((null), _tusync._tcp., (null), 107, 0)
02-18 16:57:04.110: D/MDnsDS(187): discover successful
02-18 16:57:04.110: D/NsdService(628): Discover 2 107_tusync._tcp.
02-18 16:57:04.333: D/MDnsDS(187): Discover found new serviceName TuSync-0.57392, regType _tusync._tcp. and domain local. for 107
02-18 16:57:04.334: D/NsdService(628): SERVICE_FOUND Raw: 603 107 "TuSync-0.57392" _tusync._tcp. local.
02-18 16:57:04.338: D/NsdService(628): Resolve service
02-18 16:57:04.338: D/NsdService(628): resolveService: 108 name: TuSync-0.57392, type: _tusync._tcp., host: null, port: 0
02-18 16:57:04.339: D/MDnsDS(187): resolveService(108, (null), TuSync-0.57392, _tusync._tcp., local.)
02-18 16:57:04.345: D/MDnsDS(187): startMonitoring 108
02-18 16:57:04.345: D/MDnsDS(187): resolveService successful
02-18 16:57:04.346: D/MDnsDS(187): resolve succeeded for 108 finding TuSync-0\.57392._tusync._tcp.local. at Android-3.local.:57392 with txtLen 1
02-18 16:57:04.347: D/NsdService(628): SERVICE_RESOLVED Raw: 608 108 "TuSync-0\\.57392._tusync._tcp.local." "Android-3.local." 57392 1
02-18 16:57:04.347: D/NsdService(628): stopResolveService: 108
02-18 16:57:04.347: D/MDnsDS(187): Stopping resolve with ref 0xb5c4734c
02-18 16:57:04.349: D/NsdService(628): getAdddrInfo: 109
02-18 16:57:04.349: D/MDnsDS(187): getAddrInfo(109, (null) 0, Android-3.local.)
02-18 16:57:04.350: D/MDnsDS(187): getAddrInfo successful
02-18 16:57:04.352: D/MDnsDS(187): getAddrInfo succeeded for 109: 109 "Android-3.local." 120 10.0.0.4
02-18 16:57:04.352: D/MDnsDS(187): getAddrInfo succeeded for 109: 109 "Android-3.local." 120 fe80::204:4bff:fe2c:6c87
02-18 16:57:04.354: D/NsdService(628): SERVICE_GET_ADDR_SUCCESS Raw: 612 109 "Android-3.local." 120 10.0.0.4
02-18 16:57:04.354: D/NsdService(628): stopGetAdddrInfo: 109
02-18 16:57:04.355: D/MDnsDS(187): Stopping getaddrinfo with ref 0xb5c472d4
02-18 16:57:04.364: E/NsdService(628): Unique id with no client mapping: 109
02-18 16:57:04.364: E/NsdService(628): Unhandled { when=-10ms what=393242 obj=com.android.server.NsdService$NativeEvent@86af300 target=com.android.internal.util.StateMachine$SmHandler }
02-18 16:57:04.627: D/MDnsDS(187): Discover found new serviceName TuSync-0.36230, regType _tusync._tcp. and domain local. for 107
02-18 16:57:04.632: D/MDnsDS(187): Discover found new serviceName TuSync-0.60493, regType _tusync._tcp. and domain local. for 107
02-18 16:57:04.633: D/NsdService(628): SERVICE_FOUND Raw: 603 107 "TuSync-0.36230" _tusync._tcp. local.
02-18 16:57:04.634: D/NsdService(628): SERVICE_FOUND Raw: 603 107 "TuSync-0.60493" _tusync._tcp. local.
02-18 16:57:04.635: D/NsdService(628): Resolve service
02-18 16:57:04.635: D/NsdService(628): resolveService: 110 name: TuSync-0.36230, type: _tusync._tcp., host: null, port: 0
02-18 16:57:04.636: D/MDnsDS(187): resolveService(110, (null), TuSync-0.36230, _tusync._tcp., local.)
02-18 16:57:04.637: D/MDnsDS(187): resolve succeeded for 110 finding TuSync-0\.36230._tusync._tcp.local. at Android.local.:36230 with txtLen 1
02-18 16:57:04.638: D/NsdService(628): Resolve service
02-18 16:57:04.638: D/NsdService(628): SERVICE_RESOLVED Raw: 608 110 "TuSync-0\\.36230._tusync._tcp.local." "Android.local." 36230 1
02-18 16:57:04.639: D/NsdService(628): stopResolveService: 110
02-18 16:57:04.639: D/MDnsDS(187): Stopping resolve with ref 0xb5c473c4
02-18 16:57:04.643: D/MDnsDS(187): getAddrInfo succeeded for 111: 111 "Android.local." 120 10.0.0.5
02-18 16:57:04.643: D/MDnsDS(187): getAddrInfo succeeded for 111: 111 "Android.local." 120 fe80::204:4bff:fe26:8483
02-18 16:57:04.644: D/NsdService(628): SERVICE_GET_ADDR_SUCCESS Raw: 612 111 "Android.local." 120 10.0.0.5
02-18 16:57:04.644: D/NsdService(628): stopGetAdddrInfo: 111
02-18 16:57:04.645: D/MDnsDS(187): Stopping getaddrinfo with ref 0xb5c47364
02-18 16:57:04.645: D/MDnsDS(187): Going to poll with pollCount 3
02-18 16:57:04.658: E/NsdService(628): Unique id with no client mapping: 111
02-18 16:57:04.658: E/NsdService(628): Unhandled { when=-14ms what=393242 obj=com.android.server.NsdService$NativeEvent@1d93a739 target=com.android.internal.util.StateMachine$SmHandler }
关于上下文的一些说明：
- 我的NSD服务类型是_tusync._tcp.
- 我为所有节点创建格式为TuSync-0.[本地端口号]的唯一服务名称，以防止命名冲突和方便调试。
- 在这个测试场景中，有三个设备。日志记录设备的IP是10.0.0.4，端口是57392。
日志显示，底层的MDnsDS守护程序正确发现并解析了所有节点。然而，上面的NsdService没有传播所有节点的解析。似乎在16:57:04.627发生了一个ID冲突，其中两个设备的对等体(TuSync-0.36230和TuSync-0.60493)都被分配了107个内部ID (如果我仅仅通过查看日志来正确解释机制的话)。我向NsdManager注册的discoveryListener在发现这两个节点时会得到通知，但只有一个能够解析成功，另一个会触发错误:
02-18 16:57:04.638: E/NsdHelper(6370): Resolve failed with error code:
3. Service: name: TuSync-0.60493, type: _tusync._tcp., host: null, port: 0
我还遇到了其他情况，在 NsdService 在日志中发出“SERVICE_FOUND Raw”消息后，我的发现监听器没有收到通知。下面是一个示例日志（严重过滤;与上面相同的测试设置）:
02-18 17:54:06.692: D/MDnsDS(187): Starting MDNSD
02-18 17:54:06.896: D/NsdService(628): registerService: 112 name: TuSync-0.57392, type: _tusync._tcp., host: /::, port: 57392
02-18 17:54:06.896: D/MDnsDS(187): serviceRegister(112, (null), TuSync-0.57392, _tusync._tcp., (null), (null), 57392, 0, <binary>)
02-18 17:54:06.896: D/MDnsDS(187): serviceRegister successful
02-18 17:54:08.802: D/NsdService(628): SERVICE_REGISTERED Raw: 606 112 "TuSync-0.57392"
02-18 17:54:08.820: D/NsdService(628): Discover services
02-18 17:54:09.050: D/MDnsDS(187): Discover found new serviceName TuSync-0.57392, regType _tusync._tcp. and domain local. for 113
02-18 17:54:09.050: D/NsdService(628): SERVICE_FOUND Raw: 603 113 "TuSync-0.57392" _tusync._tcp. local.
02-18 17:54:09.211: D/MDnsDS(187): Discover found new serviceName TuSync-0.60493, regType _tusync._tcp. and domain local. for 113
02-18 17:54:09.212: D/NsdService(628): SERVICE_FOUND Raw: 603 113 "TuSync-0.60493" _tusync._tcp. local.
02-18 17:54:09.215: D/NsdService(628): resolveService: 116 name: TuSync-0.60493, type: _tusync._tcp., host: null, port: 0
02-18 17:54:09.216: D/MDnsDS(187): resolveService(116, (null), TuSync-0.60493, _tusync._tcp., local.)
02-18 17:54:09.217: D/MDnsDS(187): resolve succeeded for 116 finding TuSync-0\.60493._tusync._tcp.local. at Android-2.local.:60493 with txtLen 1
02-18 17:54:09.219: D/NsdService(628): SERVICE_RESOLVED Raw: 608 116 "TuSync-0\\.60493._tusync._tcp.local." "Android-2.local." 60493 1
02-18 17:54:09.228: D/MDnsDS(187): getAddrInfo succeeded for 117: 117 "Android-2.local." 120 10.0.0.6
02-18 17:54:09.228: D/MDnsDS(187): getAddrInfo succeeded for 117: 117 "Android-2.local." 120 fe80::c643:8fff:fec5:5648
02-18 17:54:09.229: D/NsdService(628): SERVICE_GET_ADDR_SUCCESS Raw: 612 117 "Android-2.local." 120 10.0.0.6
02-18 17:54:09.244: D/MDnsDS(187): Discover found new serviceName TuSync-0.36230, regType _tusync._tcp. and domain local. for 113
02-18 17:54:09.251: E/NsdService(628): Unique id with no client mapping: 117
02-18 17:54:09.251: E/NsdService(628): Unhandled { when=-22ms what=393242 obj=com.android.server.NsdService$NativeEvent@1e992653 target=com.android.internal.util.StateMachine$SmHandler }
02-18 17:54:09.255: D/NsdService(628): SERVICE_FOUND Raw: 603 113 "TuSync-0.36230" _tusync._tcp. local.
在这种情况下，发现的对等方10.0.0.5（端口36230）不触发discoveryListener通知。在最后一条日志消息之后，什么也没有发生。因此，我的记录节点10.0.0.4只发现另一台对等方10.0.0.6：60493。
较少的类似错误报告使我想知道我是否是唯一遇到这些问题的人，还是NsdManager完全不稳定并且根本没有人使用它？
供参考，这是我的辅助类的代码--它类似于Android NSD聊天教程，但由于该教程似乎会引起其他一些错误，所以我试图进行改进。
public final class NsdHelper {
public static final String TAG = "NsdHelper";
private final Context mContext;
private final NsdManager mNsdManager;
private final String mBaseServiceName; // Base component of the service name, e.g. "service_xy"
private String mServiceName; // Service name of the local node, may be updated upon peer detection with service name conflicts, e.g. to "service_xy (2)"
private final String mServiceType;
private final NsdHandler mNsdHandler;
private MyRegistrationListener mRegistrationListener;
private final Object mRegistrationLock = new Object();
private MyDiscoveryListener mDiscoveryListener;
private final Object mDiscoveryLock = new Object();
private final Object mResolveLock = new Object();
private final Semaphore mResolveSemaphore;
public NsdHelper(Context context, String baseServiceName, String serviceName, String serviceType, NsdHandler nsdHandler) {
mContext = context;
mNsdManager = (NsdManager) context.getSystemService(Context.NSD_SERVICE);
mNsdHandler = nsdHandler;
mBaseServiceName = baseServiceName;
mServiceName = serviceName;
mServiceType = serviceType;
mResolveSemaphore = new Semaphore(10, true);
}
/*********************
* Lifecycle methods *
*********************/
public void registerLocalService(final int port) {
NsdServiceInfo localServiceInfo = new NsdServiceInfo();
localServiceInfo.setServiceName(mServiceName);
localServiceInfo.setServiceType(mServiceType);
localServiceInfo.setPort(port);
synchronized (mRegistrationLock) {
if (mRegistrationListener == null) {
mRegistrationListener = new MyRegistrationListener();
// try {
mNsdManager.registerService(
localServiceInfo, NsdManager.PROTOCOL_DNS_SD, mRegistrationListener);
/*} catch (Exception e) {
MLog.e(TAG, "Exception registering service; trying to unregister.", e);
unregisterLocalService();
mNsdHandler.onRegistrationFailed(localServiceInfo, 0);
}*/
} else {
MLog.w(TAG, "registerLocalService called while service registration already in progress or service already registered.");
}
}
}
public void unregisterLocalService() {
synchronized (mRegistrationLock) {
if (mRegistrationListener != null) {
// try {
mNsdManager.unregisterService(mRegistrationListener);
/*} catch (IllegalArgumentException e) {
MLog.w(TAG, "Exception trying to unregister registrationListener.");
}*/
mRegistrationListener = null;
} else {
MLog.w(TAG, "unregisterLocalService called while service not yet registered or already unregistered.");
}
}
}
public void startDiscovery() {
synchronized(mDiscoveryLock) {
if(mDiscoveryListener == null) {
mDiscoveryListener = new MyDiscoveryListener();
mNsdManager.discoverServices(
mServiceType, NsdManager.PROTOCOL_DNS_SD, mDiscoveryListener);
} else {
MLog.w(TAG, "StartDiscovery called while discovery is already in progress.");
}
}
}
public void stopDiscovery() {
synchronized (mDiscoveryLock) {
if (mDiscoveryListener != null) {
mNsdManager.stopServiceDiscovery(mDiscoveryListener);
mDiscoveryListener = null;
} else {
MLog.w(TAG, "StopDiscovery called while no discovery is in progress.");
}
}
}
public void tearDown() {
MLog.v(TAG, "NsdHelper: tearDown()");
stopDiscovery();
unregisterLocalService(); // TODO this causes an exception, when the listener is already unregistered
}
/**
* Returns the current service name of the service.
* @return
*/
public String getServiceName() {
return mServiceName;
}
/**
* Convenience method to initiate service resolution
* @param serviceInfo NsdServiceInfo object for the service to be resolved
*/
private void resolveService(NsdServiceInfo serviceInfo) {
try {
MLog.vv(TAG, "Resolving service: acquiring semaphore.");
mResolveSemaphore.acquire();
MLog.vv(TAG, "Resolving service: semaphore acquired.");
} catch (InterruptedException e) {
MLog.w(TAG, "resolveService: Waiting for acquisition of semaphore interrupted.");
}
mNsdManager.resolveService(serviceInfo, new MyResolveListener(serviceInfo.getServiceName()));
}
/*************
* Listeners *
*************/
private class MyDiscoveryListener implements NsdManager.DiscoveryListener {
@Override
public void onDiscoveryStarted(String regType) {
MLog.d(TAG, "Service discovery started");
mNsdHandler.onDiscoveryStarted();
}
@Override
public void onServiceFound(NsdServiceInfo serviceInfo) {
MLog.d(TAG, "Discovered service: " + serviceInfo);
// Protocol matches?
if (!serviceInfo.getServiceType().equals(mServiceType)) {
MLog.v(TAG, "Discovered: other serviceType: " + serviceInfo.getServiceType());
}
// Make sure, that service name matches, and just resolve remote host
else if (serviceInfo.getServiceName().contains(mBaseServiceName)){
MLog.d(TAG, "Discovered: correct serviceType: " + mBaseServiceName);
resolveService(serviceInfo);
}
else {
// Other service name, log anyway
MLog.d(TAG, "Discovered: service with different serviceName: " + serviceInfo.getServiceName() + ". Ignoring.");
}
}
@Override
public void onServiceLost(NsdServiceInfo service) {
MLog.e(TAG, "Service lost: " + service);
mNsdHandler.onRemotePeerLost(service);
}
@Override
public void onDiscoveryStopped(String serviceType) {
MLog.v(TAG, "Discovery stopped: " + serviceType);
mNsdHandler.onDiscoveryStopped();
}
@Override
public void onStartDiscoveryFailed(String serviceType, int errorCode) {
MLog.e(TAG, "Discovery starting failed. Error code: " + errorCode);
synchronized (mDiscoveryLock) {
mDiscoveryListener = null; // just throw away the discovery listener, explicit stopping of the discovery should not be needed according to
// https://code.google.com/p/android/issues/detail?id=99510&q=nsd&colspec=ID%20Type%20Status%20Owner%20Summary%20Stars
}
}
@Override
public void onStopDiscoveryFailed(String serviceType, int errorCode) {
MLog.e(TAG, "Discovery stopping failed. Error code: " + errorCode);
// try again
// mNsdManager.stopServiceDiscovery(this); // This should not be needed according to https://code.google.com/p/android/issues/detail?id=99510&q=nsd&colspec=ID%20Type%20Status%20Owner%20Summary%20Stars
}
};
private class MyRegistrationListener implements NsdManager.RegistrationListener {
@Override
public void onServiceRegistered(NsdServiceInfo nsdServiceInfo) {
MLog.d(TAG, "Service registered. NsdServiceInfo: " + nsdServiceInfo);
boolean nameChanged = false;
// Update service name of this node (might change due to automatic conflict resolution!)
if(!mServiceName.equals(nsdServiceInfo.getServiceName())){
mServiceName = nsdServiceInfo.getServiceName();
nameChanged = true;
MLog.d(TAG, "Local service name updated to: " + mServiceName);
}
// Notify
if (mNsdHandler != null) {
mNsdHandler.onRegistrationSuccess(nsdServiceInfo);
if (nameChanged) {
mNsdHandler.onLocalServiceNameChanged(mServiceName);
}
} else {
MLog.w(TAG, "onServiceRegistered: NsdHandler is null.");
}
}
@Override
public void onRegistrationFailed(NsdServiceInfo arg0, int arg1) {
MLog.w(TAG, "Service registration failed with error code " + arg1 + ".");
if (mNsdHandler == null) {
MLog.w(TAG, "onRegistrationFailed: NsdHandler is null.");
return;
}
mNsdHandler.onRegistrationFailed(arg0, arg1);
}
@Override
public void onServiceUnregistered(NsdServiceInfo arg0) {
MLog.d(TAG, "Service unregistered.");
if (mNsdHandler == null) {
MLog.w(TAG, "onServiceUnRegistered: NsdHandler is null.");
return;
}
}
@Override
public void onUnregistrationFailed(NsdServiceInfo serviceInfo, int errorCode) {
MLog.w(TAG, "Service unregistering failed.");
if (mNsdHandler == null) {
MLog.w(TAG, "onUnRegistrationFailed: NsdHandler is null.");
return;
}
}
};
private class MyResolveListener implements NsdManager.ResolveListener {
private final String mServiceName;
public MyResolveListener(String serviceName) {
mServiceName = serviceName;
}
@Override
public void onResolveFailed(final NsdServiceInfo serviceInfo, int errorCode) {
// Release resource
mResolveSemaphore.release();
MLog.e(TAG, "Resolve failed with error code: " + errorCode + ". Service: " + serviceInfo);
if((serviceInfo.getServiceName() != null) && (!serviceInfo.getServiceName().equals(mServiceName))) {
MLog.e(TAG, "Service name changed: " + mServiceName + " => " + serviceInfo.getServiceName());
}
}
@Override
public void onServiceResolved(final NsdServiceInfo serviceInfo) {
// Release resource
mResolveSemaphore.release();
MLog.v(TAG, "Resolve succeeded. Service: " + serviceInfo + ", Address: " + serviceInfo.getHost().getHostAddress() + ":" + serviceInfo.getPort());
if((serviceInfo.getServiceName() != null) && (!serviceInfo.getServiceName().equals(mServiceName))) {
MLog.w(TAG, "Service name changed: " + mServiceName + " => " + serviceInfo.getServiceName());
}
mNsdHandler.onNewRemotePeerResolved(serviceInfo);
}
};
/**
* Interface for handlers that deal just with essential NSD events.
* @author Alexander Fischl (alexander.fischl@semeion.net)
*/
public interface NsdHandler {
/**
* Called, when the NSD manager registered the service successfully.
* @param nsdServiceInfo
*/
public void onRegistrationSuccess(final NsdServiceInfo nsdServiceInfo);
/**
* Called, when the NSD registration was unsuccessful.
*/
public void onRegistrationFailed(final NsdServiceInfo nsdServiceInfo, final int errorCode);
/**
* Called, when the NSD manager discovers a new peer. Services registered on the
* local machine DO NOT trigger this call!
* @param nsdServiceInfo
*/
public void onNewRemotePeerDiscovered(final NsdServiceInfo nsdServiceInfo);
/**
* Called, when the NSD manager resolves a new peer, yielding the connection data.
* Services registered on the local machine DO NOT trigger this call!
* @param nsdServiceInfo
*/
public void onNewRemotePeerResolved(final NsdServiceInfo nsdServiceInfo);
/**
* Called, when the NSD manager loses an already discovered peer.
* @param nsdServiceInfo
*/
public void onRemotePeerLost(final NsdServiceInfo nsdServiceInfo);
/**
* Called, when the local service name needs to be updated (e.g. due to
* conflict resolution when the local service is registered, and the chosen service
* name is already taken by another node in the network.)
* @param newLocalServiceName
*/
public void onLocalServiceNameChanged(String newLocalServiceName);
/**
* Called, when the service discovery has successfully started.
*/
public void onDiscoveryStarted();
/**
* Called, when the service discovery was halted.
*/
public void onDiscoveryStopped();
}
}
注意，我甚至实现了一个信号量，可以将其设置为1，以防止同时解析多个服务，因为其他人报告说有问题的并行解析。但是，将它设置为1不起作用，因为有时正在进行的解析既不成功也不失败；这会导致信号量没有被释放，而NsdManager线程在下一个解析请求处永久卡住。
还有其他人遇到这样的问题吗？如果那些成功使用NsdManager的人也能发表评论，我会很高兴-至少这意味着我面临的是我可以解决的问题 :)
我已经考虑放弃NSD并实现自己的广播/组播发现机制。理论上这可能很容易，但我读过Android上的组播也很麻烦，因为一些设备会阻止它...