android wifiservice enable流程

转载链接：http://www.xuebuyuan.com/2139877.html

这一章来分析一下从Settings打开wifi后，framework所有的流程以及状态转换，调用的主要函数就是WiFiMananger的setWifiEnabled(boolean enabled)，这个函数在前面介绍的WifiService启动流程中也有提到，当SystemServer

创建了个WifiService后，就会调用它的checkAndStartWifi，在这个函数里面，也同样会调用到setWifiEnabled(boolean enabled)，这是为了恢复用户在开机之前的wifi状态。下面来详细分析WifiMananger的setWifiEnabled函数。先来看看总体的流程图



对照上面的流程图，我们从WifiMananger的setWifiEnabled来分析一下代码：

public boolean setWifiEnabled(boolean enabled) {
try {
return mService.setWifiEnabled(enabled);
} catch (RemoteException e) {
return false;
}
}

WifiService.java
public synchronized boolean setWifiEnabled(boolean enable) {
enforceChangePermission();
Slog.d(TAG, "setWifiEnabled: " + enable + " pid=" + Binder.getCallingPid()
+ ", uid=" + Binder.getCallingUid());
if (DBG) {
Slog.e(TAG, "Invoking mWifiStateMachine.setWifiEnabled\n");
}

/*
* Caller might not have WRITE_SECURE_SETTINGS,
* only CHANGE_WIFI_STATE is enforced
*/

long ident = Binder.clearCallingIdentity();
try {
if (! mSettingsStore.handleWifiToggled(enable)) {
// Nothing to do if wifi cannot be toggled
return true;
}
} finally {
Binder.restoreCallingIdentity(ident);
}

mWifiController.sendMessage(CMD_WIFI_TOGGLED);
return true;
}

在WifiService里面，先把传进来的参数写到WifiSettingsStore里面，WifiSettingsStore调用Settings提供的ContentProviders写到Sqlite的DB里面，所以我们看到mWifiController.sendMessage(CMD_WIFI_TOGGLED)是没有把这个参数传递过去的。接着去看WifiController里面如何处理CMD_WIFI_TOGGLED，由前面的WifiService启动流程分析，我们知道CMD_WIFI_TOGGLED将由WifiController的ApStaDisabledState来处理：

class ApStaDisabledState extends State {
private int mDeferredEnableSerialNumber = 0;
private boolean mHaveDeferredEnable = false;
private long mDisabledTimestamp;

@Override
public void enter() {
mWifiStateMachine.setSupplicantRunning(false);
// Supplicant can't restart right away, so not the time we switched off
mDisabledTimestamp = SystemClock.elapsedRealtime();
mDeferredEnableSerialNumber++;
mHaveDeferredEnable = false;
}
@Override
public boolean processMessage(Message msg) {
switch (msg.what) {
case CMD_WIFI_TOGGLED:
case CMD_AIRPLANE_TOGGLED:
if (mSettingsStore.isWifiToggleEnabled()) {
if (doDeferEnable(msg)) {
if (mHaveDeferredEnable) {
//  have 2 toggles now, inc serial number an ignore both
mDeferredEnableSerialNumber++;
}
mHaveDeferredEnable = !mHaveDeferredEnable;
break;
}
if (mDeviceIdle == false) {
transitionTo(mDeviceActiveState);
} else {
checkLocksAndTransitionWhenDeviceIdle();
}
}
break;

WifiController的ApStaDisabledState的处理很简单，只是简单的transition到DeviceActiveState，因为DeviceActiveState的父State是StaEnabledState，由StateMachine的知识，我们先到StaEnabledState和DeviceActiveState的enter()函数来看看：

class StaEnabledState extends State {
@Override
public void enter() {
mWifiStateMachine.setSupplicantRunning(true);
}

class DeviceActiveState extends State {
@Override
public void enter() {
mWifiStateMachine.setOperationalMode(WifiStateMachine.CONNECT_MODE);
mWifiStateMachine.setDriverStart(true);
mWifiStateMachine.setHighPerfModeEnabled(false);
}

上面分别是调用WifiStateMachine的四个函数，这四个函数都是给WifiStateMachine发送四个消息，分别是CMD_START_SUPPLICANT、SET_OPERATIONAL_MODE、CMD_START_DRIVER和SET_HTGH_PERF_MODE，如上图中的Figure 1所示。接着我们到WifiStateMachine的InitialState中去看看如何处理CMD_START_SUPPLICANT：

public boolean processMessage(Message message) {
switch (message.what) {
case CMD_START_SUPPLICANT:
if (mWifiNative.loadDriver()) {
try {
mNwService.wifiFirmwareReload(mInterfaceName, "STA");
} catch (Exception e) {
loge("Failed to reload STA firmware " + e);
// continue
}

try {
mNwService.disableIpv6(mInterfaceName);
} catch (RemoteException re) {
loge("Unable to change interface settings: " + re);
} catch (IllegalStateException ie) {
loge("Unable to change interface settings: " + ie);
}

mWifiMonitor.killSupplicant(mP2pSupported);
if(mWifiNative.startSupplicant(mP2pSupported)) {
setWifiState(WIFI_STATE_ENABLING);
if (DBG) log("Supplicant start successful");
mWifiMonitor.startMonitoring();
transitionTo(mSupplicantStartingState);
} else {
loge("Failed to start supplicant!");
}
} else {
loge("Failed to load driver");
}
break;

这里主要调用WifiNative的loadDriver和startSupplicant两个函数去加载wifi driver和启动wpa_supplicant，当启动成功wpa_supplicant后，就会调用WifiMonitor的startMonitoring去和wpa_supplicant建立socket连接，并不断的从wpa_supplicant收event。wpa_supplicant是一个独立的运行程序，它和应用程序之间通过socket来通信，主要存在两个socket连接，一个用来向wpa_supplicant发送命令，另一个是wpa_supplicant用来向应用程序通知event，应用程序在收到event后可以知道当前的连接状态来进行下一步动作。我们进入到WifiMonitor.startMonitoring这个函数看看：

public synchronized void startMonitoring(String iface) {
WifiMonitor m = mIfaceMap.get(iface);
if (m == null) {
Log.e(TAG, "startMonitor called with unknown iface=" + iface);
return;
}

Log.d(TAG, "startMonitoring(" + iface + ") with mConnected = " + mConnected);

if (mConnected) {
m.mMonitoring = true;
m.mWifiStateMachine.sendMessage(SUP_CONNECTION_EVENT);
} else {
if (DBG) Log.d(TAG, "connecting to supplicant");
int connectTries = 0;
while (true) {
if (mWifiNative.connectToSupplicant()) {
m.mMonitoring = true;
m.mWifiStateMachine.sendMessage(SUP_CONNECTION_EVENT);
new MonitorThread(mWifiNative, this).start();
mConnected = true;
break;
}
if (connectTries++ < 5) {
try {
Thread.sleep(1000);
} catch (InterruptedException ignore) {
}
} else {
mIfaceMap.remove(iface);
m.mWifiStateMachine.sendMessage(SUP_DISCONNECTION_EVENT);
Log.e(TAG, "startMonitoring(" + iface + ") failed!");
break;
}
}
}
}

这个方法里面主要调用WifiNative的connenctToSupplicant去和wpa_supplicant建立socket连接，然后给WifiStateMachine发送一个SUP_CONNECTION_EVENT消息，最后新建一个MonitorThread运行，MonitorThread就是一个循环，不断的从wpa_supplicant收event，然后进行解析，并dispatch到不同的函数去处理，后面我们再来分析MonitorThread的流程。回到WifiStateMachine的InitialState中去看看如何处理CMD_START_SUPPLICANT的流程中来，当startMonitoring结束后，WifiStateMachine就跳转到SupplicantStartingState，这时的WifiStateMachine和MessageQueue里面的消息队列上图中的Figure

2。

接着来看当WifiStateMachine处理完SUP_CONNECTION_EVENT消息后，马上会收到SET_OPERATIONAL_MODE和CMD_START_DRIVER消息，这两个消息都会被SupplicantStartingState延迟处理，SET_HTGH_PERF_MODE会被DefaultState处理。接着SupplicantStartingState会收到SUP_CONNECTION_EVENT，处理代码如下：

public boolean processMessage(Message message) {
switch(message.what) {
case WifiMonitor.SUP_CONNECTION_EVENT:
if (DBG) log("Supplicant connection established");
setWifiState(WIFI_STATE_ENABLED);
mSupplicantRestartCount = 0;
/* Reset the supplicant state to indicate the supplicant
* state is not known at this time */
mSupplicantStateTracker.sendMessage(CMD_RESET_SUPPLICANT_STATE);
/* Initialize data structures */
mLastBssid = null;
mLastNetworkId = WifiConfiguration.INVALID_NETWORK_ID;
mLastSignalLevel = -1;

mWifiInfo.setMacAddress(mWifiNative.getMacAddress());
mWifiConfigStore.loadAndEnableAllNetworks();
initializeWpsDetails();

sendSupplicantConnectionChangedBroadcast(true);
transitionTo(mDriverStartedState);
break;

在SUP_CONNECTION_EVENT的处理流程中，主要是调用WifiConfigStore的loadAndEnableAllNetworks函数来加载并enable用户之前连接过并保存的AP，然后会初始化一些Wps相关的信息，最后transition到DriverStartedState上，如上图的Figure 3。再来看DriverStartedState的enter函数，这里面有一些重要的流程：

class DriverStartedState extends State {
@Override
public void enter() {

/* set country code */
setCountryCode();
/* set frequency band of operation */
setFrequencyBand();
/* initialize network state */
setNetworkDetailedState(DetailedState.DISCONNECTED);

mDhcpActive = false;

startBatchedScan();

if (mOperationalMode != CONNECT_MODE) {
mWifiNative.disconnect();
mWifiConfigStore.disableAllNetworks();
if (mOperationalMode == SCAN_ONLY_WITH_WIFI_OFF_MODE) {
setWifiState(WIFI_STATE_DISABLED);
}
transitionTo(mScanModeState);
} else {
/* Driver stop may have disabled networks, enable right after start */
mWifiConfigStore.enableAllNetworks();

if (DBG) log("Attempting to reconnect to wifi network ..");
mWifiNative.reconnect();

// Status pulls in the current supplicant state and network connection state
// events over the monitor connection. This helps framework sync up with
// current supplicant state
mWifiNative.status();
transitionTo(mDisconnectedState);
}
if (mP2pSupported) {
if (mOperationalMode == CONNECT_MODE) {
mWifiP2pChannel.sendMessage(WifiStateMachine.CMD_ENABLE_P2P);
} else {
// P2P statemachine starts in disabled state, and is not enabled until
// CMD_ENABLE_P2P is sent from here; so, nothing needs to be done to
// keep it disabled.
}
}

}

enter函数的代码比较多，上面是精简后的代码，上面主要分为两条分支，一是OperationalMode != CONNECT_MODE，一种是OperationalMode = CONNECT_MODE，根据官方的解释，OperationalMode一共有三种，分别如下：

1.CONNECT_MODE，这种模式下，STA可以scan并连接热点

2.SCAN_ONLY_MODE，这种模式下，STA只能扫描热点

3.SCAN_ONLY_WIFI_OFF_MODE，在这种模式下，当wifi是toggle off的情况下，也可以进行scan

这三种模式默认的是CONNECT_MODE，后面两种模式现在用到的不多，但按照Google的设计，后面可能会有很多的app会用到，比如利用热点来进行点位，这个应用其实这需要能够scan，并不需要链接热点。那我们接看这OperationalMode = CONNECT_MODE的流程，它直接transition 到DisconnectedState，如上图中的Figure 4，在transition到DisconnectedState之前，还会向WifiNative下reconnect的命令，用于重新连接上次连接但没有forget的AP，即开机后自动连上AP。如平台支持P2P，这里还将会给WifiP2pService发送CMD_ENABLE_P2P的消息，以后在学习P2P模块的时候再来分析。

到这里，Wifi toggle on的流程就分析完了。

WiFi toggled off流程分析

和Wifi toggle on的流程一样，当用户从Settings关闭wifi时，会调用WiFiMananger的setWifiEnabled(false)，接着WifiService会给WifiController发送一个CMD_WIFI_TOGGLED消息，由前面启动Wifi的流程来看，StaEabledState会处理这个消息，先看看整体的流程图：



进入到StaEabledState的processMessage来看看代码流程：

public boolean processMessage(Message msg) {
switch (msg.what) {
case CMD_WIFI_TOGGLED:
if (! mSettingsStore.isWifiToggleEnabled()) {
if (mSettingsStore.isScanAlwaysAvailable()) {
transitionTo(mStaDisabledWithScanState);
} else {
transitionTo(mApStaDisabledState);
}
}
break;

WifiController会transition 到ApStaDisabledState中，这样WifiController这个状态机就回到的最开始的初始状态了，进到ApStaDisabledState的enter函数分析如何对WifiStateMachine这个状态机做处理：

class ApStaDisabledState extends State {
private int mDeferredEnableSerialNumber = 0;
private boolean mHaveDeferredEnable = false;
private long mDisabledTimestamp;

@Override
public void enter() {
mWifiStateMachine.setSupplicantRunning(false);
// Supplicant can't restart right away, so not the time we switched off
mDisabledTimestamp = SystemClock.elapsedRealtime();
mDeferredEnableSerialNumber++;
mHaveDeferredEnable = false;
}

这里主要调用WifiStateMachine的setSupplicantRunning(false)，这个函数直接给WifiStateMachine发送一个CMD_STOP_SUPPLICAN消息，由前面的知识，我们知道SupplicantStartedState会处理这个消息，进到具体代码中分析：

public boolean processMessage(Message message) {
switch(message.what) {
case CMD_STOP_SUPPLICANT:   /* Supplicant stopped by user */
if (mP2pSupported) {
transitionTo(mWaitForP2pDisableState);
} else {
transitionTo(mSupplicantStoppingState);
}
break;

如果平台支持P2P，这里会跳转到WaitForP2pDisableState中，如果不支持则会跳转到SupplicantStoppingState中。这里我们看mP2pSupported为true的情况，因为现在大多数平台都应该支持P2P了，另外，在WaitForP2pDisableState中处理完P2P相关的内容后，也会跳转到SupplicantStoppingState中来。我们来分析WaitForP2pDisableState的enter函数：

class WaitForP2pDisableState extends State {
private State mTransitionToState;
@Override
public void enter() {
switch (getCurrentMessage().what) {
case WifiMonitor.SUP_DISCONNECTION_EVENT:
mTransitionToState = mInitialState;
break;
case CMD_DELAYED_STOP_DRIVER:
mTransitionToState = mDriverStoppingState;
break;
case CMD_STOP_SUPPLICANT:
mTransitionToState = mSupplicantStoppingState;
break;
default:
mTransitionToState = mDriverStoppingState;
break;
}
mWifiP2pChannel.sendMessage(WifiStateMachine.CMD_DISABLE_P2P_REQ);
}
@Override
public boolean processMessage(Message message) {
switch(message.what) {
case WifiStateMachine.CMD_DISABLE_P2P_RSP:
transitionTo(mTransitionToState);
break;

mTransitionToState根据当前处理消息的不同种类，记录下次将要跳转到的State信息，然后给WifiP2pService发送一个CMD_DISABLE_P2P_REQ消息，P2P的内容后面再来分析，当P2P处理完CMD_DISABLE_P2P_REQ后，就会给WifiStateMachine发送一个CMD_DISABLE_P2P_RSP消息，在WaitForP2pDisableState收到这个消息后，就会跳转到SupplicantStoppingState中，再到这个函数的enter去看看：

class SupplicantStoppingState extends State {
@Override
public void enter() {
/* Send any reset commands to supplicant before shutting it down */
handleNetworkDisconnect();
if (mDhcpStateMachine != null) {
mDhcpStateMachine.doQuit();
}

if (DBG) log("stopping supplicant");
mWifiMonitor.stopSupplicant();

/* Send ourselves a delayed message to indicate failure after a wait time */
sendMessageDelayed(obtainMessage(CMD_STOP_SUPPLICANT_FAILED,
++mSupplicantStopFailureToken, 0), SUPPLICANT_RESTART_INTERVAL_MSECS);
setWifiState(WIFI_STATE_DISABLING);
mSupplicantStateTracker.sendMessage(CMD_RESET_SUPPLICANT_STATE);
}
@Override
public boolean processMessage(Message message) {
switch(message.what) {
case WifiMonitor.SUP_CONNECTION_EVENT:
loge("Supplicant connection received while stopping");
break;
case WifiMonitor.SUP_DISCONNECTION_EVENT:
if (DBG) log("Supplicant connection lost");
handleSupplicantConnectionLoss();
transitionTo(mInitialState);
break;
case CMD_STOP_SUPPLICANT_FAILED:
if (message.arg1 == mSupplicantStopFailureToken) {
loge("Timed out on a supplicant stop, kill and proceed");
handleSupplicantConnectionLoss();
transitionTo(mInitialState);
}
break;

在SupplicantStoppingState主要调用handleNetworkDisconnect和stopSupplicant函数，handleNetworkDisconnect主要工作是停掉stopDhcp和clear一些状态信息；WifiMonitor的stopSupplicant用于停掉wpa_supplicant，就是向wpa_supplicant发送一个TERMINATE命令，当wpa_supplicant收到TERMINATE命令会，就会给调用者发送CTRL-EVENT-TERMINATING这个event，当WifiMonitor收到这个event后，又会给WifiStateMachine发送SUP_DISCONNECTION_EVENT消息。回到SupplicantStoppingState收到这个消息后，就可以去结束掉wpa_supplicant进程并断开与它的socket连接，并且transition到InitialState，这样WifiStateMachine也恢复到最初的状态了。另外，从上面enter函数可以看到，还会发送一个CMD_STOP_SUPPLICANT_FAILED消息给自己，如果调用stopSupplicant不成功，间隔SUPPLICANT_RESTART_INTERVAL_MSECS毫秒后，也会走到和SUP_DISCONNECTION_EVENT消息处理一样的流程中来。