前面的博客中,我们已经分析过,当Android中的进程要使用电量时,需要向PMS申请WakeLock;当进程完成工作后,需要释放对应的WakeLock。
PMS收到申请和释放WakeLock的请求后,均需要调用updatePowerStateLocked来更新电源的状态,该函数是PMS的核心方法。
接下来,我们就结合代码,看一下updatePowerStateLocked的工作流程。
/**
* Updates the global power state based on dirty bits recorded in mDirty.
*
* This is the main function that performs power state transitions.
* We centralize them here so that we can recompute the power state completely
* each time something important changes, and ensure that we do it the same
* way each time. The point is to gather all of the transition logic here.
*/
private void updatePowerStateLocked() {
//未启动完毕或mDirty没有记录变化
if (!mSystemReady || mDirty == 0) {
return;
}
..........
try {
// Basic state updates.
// 1、更新基本状态
updateIsPoweredLocked(mDirty);
updateStayOnLocked(mDirty);
updateScreenBrightnessBoostLocked(mDirty);
// Update wakefulness.
// Loop because the wake lock and user activity computations are influenced
// by changes in wakefulness.
// 2、更新wakelock和用户活动
final long now = SystemClock.uptimeMillis();
int dirtyPhase2 = 0;
for (;;) {
int dirtyPhase1 = mDirty;
dirtyPhase2 |= dirtyPhase1;
mDirty = 0;
updateWakeLockSummaryLocked(dirtyPhase1);
updateUserActivitySummaryLocked(now, dirtyPhase1);
if (!updateWakefulnessLocked(dirtyPhase1)) {
break;
}
}
// Update display power state.
// 3、更新display power state
boolean displayBecameReady = updateDisplayPowerStateLocked(dirtyPhase2);
// Update dream state (depends on display ready signal).
// 4、更新dream state
updateDreamLocked(dirtyPhase2, displayBecameReady);
// Send notifications, if needed.
finishWakefulnessChangeIfNeededLocked();
// Update suspend blocker.
// Because we might release the last suspend blocker here, we need to make sure
// we finished everything else first!
// 5、更新suspend blocker
updateSuspendBlockerLocked();
} finally {
..........
}
}
在PMS中有个很重要的变量mDirty,该变量按位存储PMS中的各种变化状态。
例如,之前介绍PMS的acquire WakeLock流程时,就进行了以下操作:
.........
mDirty |= DIRTY_WAKE_LOCKS;
........
每当PMS检测到一些重要事件发生时,就会更新mDirty的相应的位。
从updatePowerStateLocked的代码可以看出,它将根据mDirty中的信息,来更新手机中的电源状态。
根据Android源码中的注释,可以看出updatePowerStateLocked的工作主要分为几个步骤,接下来我们一个一个步骤的来进行分析。
一、更新基本状态信息
1、updateIsPoweredLocked
我们先来看看updateIsPoweredLocked函数:
private void updateIsPoweredLocked(int dirty) {
//DIRTY_BATTERY_STATE位置1时,表示终端的电源状态发生了改变
if ((dirty & DIRTY_BATTERY_STATE) != 0) {
//记录过去的状态
final boolean wasPowered = mIsPowered;
final int oldPlugType = mPlugType;
final boolean oldLevelLow = mBatteryLevelLow;
//得到终端现在是否在充电
mIsPowered = mBatteryManagerInternal.isPowered(BatteryManager.BATTERY_PLUGGED_ANY);
//得到充电的类型
mPlugType = mBatteryManagerInternal.getPlugType();
//得到当前的电量
mBatteryLevel = mBatteryManagerInternal.getBatteryLevel();
//判断是否为低电量
mBatteryLevelLow = mBatteryManagerInternal.getBatteryLevelLow();
//是否充电的状态发生改变,或者充电类型发生改变
if (wasPowered != mIsPowered || oldPlugType != mPlugType) {
mDirty |= DIRTY_IS_POWERED;
// Update wireless dock detection state.
//无线充电相关,暂时不用管
final boolean dockedOnWirelessCharger = mWirelessChargerDetector.update(
mIsPowered, mPlugType, mBatteryLevel);
final long now = SystemClock.uptimeMillis();
//判断插拔充电器或者USB是否需要唤醒屏幕
if (shouldWakeUpWhenPluggedOrUnpluggedLocked(wasPowered, oldPlugType,
dockedOnWirelessCharger)) {
//之前的博客中已经分析过这个函数,主要是做好唤醒终端屏幕前的准备工作
wakeUpNoUpdateLocked(now, "android.server.power:POWER", Process.SYSTEM_UID,
mContext.getOpPackageName(), Process.SYSTEM_UID);
}
//触发一次用户活动,修改PMS中记录用户活动事件的时间,同时通知BatteryStatsService等
userActivityNoUpdateLocked(
now, PowerManager.USER_ACTIVITY_EVENT_OTHER, 0, Process.SYSTEM_UID);
// Tell the notifier whether wireless charging has started so that
// it can provide feedback to the user.
//无线充电相关的通知,暂时可以不管
if (dockedOnWirelessCharger) {
mNotifier.onWirelessChargingStarted();
}
}
if (wasPowered != mIsPowered || oldLevelLow != mBatteryLevelLow) {
//结束低电的状态
if (oldLevelLow != mBatteryLevelLow && !mBatteryLevelLow) {
........
//从命名来看,该标志用于决定终端在低电模式下是否“打盹”(接近休眠)
mAutoLowPowerModeSnoozing = false;
}
//更新低电模式相关的操作
updateLowPowerModeLocked();
}
}
}
从以上代码可以看出updateIsPoweredLocked主要用于:
更新PMS中的一些变量,包括记录终端是否在充电、充电的类型、电池的电量及电池电量是否处于低电状态;
当电源的充电状态,或者充电类型发生变化,判断出现插拔充电器等操作时,是否需要点亮或熄灭屏幕;
当电源充电状态发生变化,或者终端是否处于低电量的标志发生变化的时候,终端调用updateLowPowerModeLocked()更新低电模式相关的操作。
我们跟进一下updateLowPowerModeLocked函数:
private void updateLowPowerModeLocked() {
//处于充电状态,并且设置过低电模式的标志位
if (mIsPowered && mLowPowerModeSetting) {
........
// Turn setting off if powered
//更新数据库,关闭低电模式
Settings.Global.putInt(mContext.getContentResolver(),
Settings.Global.LOW_POWER_MODE, 0);
mLowPowerModeSetting = false;
}
//判断是否可以进入自动省电模式
//要求是:未充电 && 进行了自动省电的配置 && 没有设置低电“打盹” && 电池电量低
final boolean autoLowPowerModeEnabled = !mIsPowered && mAutoLowPowerModeConfigured
&& !mAutoLowPowerModeSnoozing && mBatteryLevelLow;
//当前是否为低电模式
final boolean lowPowerModeEnabled = mLowPowerModeSetting || autoLowPowerModeEnabled;
if (mLowPowerModeEnabled != lowPowerModeEnabled) {
mLowPowerModeEnabled = lowPowerModeEnabled;
//调用底层动态库的powerHint函数
powerHintInternal(POWER_HINT_LOW_POWER, lowPowerModeEnabled ? 1 : 0);
//开机完成后才能执行的Runnable对象
postAfterBootCompleted(new Runnable() {
//发送低电模式CHANGING的广播
Intent intent = new Intent(PowerManager.ACTION_POWER_SAVE_MODE_CHANGING)
.putExtra(PowerManager.EXTRA_POWER_SAVE_MODE, mLowPowerModeEnabled)
.addFlags(Intent.FLAG_RECEIVER_REGISTERED_ONLY);
mContext.sendBroadcast(intent);
//PMS提供了registerLowPowerModeObserver的接口
//其它进程可以调用该接口,注册观察者
synchronized (mLock) {
listeners = new ArrayList(
mLowPowerModeListeners);
}
for (int i=0; i//调用回调接口的onLowPowerModeChanged函数,通知其它进程低电模式发生改变
listeners.get(i).onLowPowerModeChanged(lowPowerModeEnabled);
}
//再次发送CHANGED广播
intent = new Intent(PowerManager.ACTION_POWER_SAVE_MODE_CHANGED);
intent.addFlags(Intent.FLAG_RECEIVER_REGISTERED_ONLY);
mContext.sendBroadcast(intent);
// Send internal version that requires signature permission.
mContext.sendBroadcastAsUser(new Intent(
PowerManager.ACTION_POWER_SAVE_MODE_CHANGED_INTERNAL), UserHandle.ALL,
Manifest.permission.DEVICE_POWER);
});
}
}
从上面的代码可以看出updateLowPowerModeLocked函数,
首先判断手机是否在充电,如果手机在充电,退出LowPowerMode模式,同时更新数据库;
当手机的低电量模式发生了变化,就发送广播进行通知,并回调关于监听该模式变化的观察者的接口。
例如:UI对应的APK收到低电量省电模式的广播,就会弹出低电量省电模式的提醒界面。
可以看出这一部分除了更新PMS中的一些变量外,关注的重点还是集中在:
充电状态是否改变;
充电状态的改变,将引出对充电器插拔是否需要亮屏的考虑;
同样,充电状态的改变,将引出对终端的低电模式是否发生改变的考虑。
从这个角度来看,updateIsPoweredLocked函数的命名是实至名归的。
2、updateStayOnLocked
现在我们看看基本状态更新第二部分的updateStayOnLocked函数:
/**
* Updates the value of mStayOn.
* Sets DIRTY_STAY_ON if a change occurred.
*/
private void updateStayOnLocked(int dirty) {
//电源状态或电源设置发生了改变
if ((dirty & (DIRTY_BATTERY_STATE | DIRTY_SETTINGS)) != 0) {
final boolean wasStayOn = mStayOn;
//设置了充电器插入时亮屏(分为AC充电亮屏、USB充电亮屏或无线充电亮屏)
if (mStayOnWhilePluggedInSetting != 0
//判断mMaximumScreenOffTimeoutFromDeviceAdmin的是否处于0与Integer.MAX_VALUE之间
//Android给出的注释是:
//The maximum allowable screen off timeout according to the device
// administration policy
//初始为Integer.MAX_VALUE,因此这里是要求其它进程没有设置这个值
//应该对应于强制息屏时间
&& !isMaximumScreenOffTimeoutFromDeviceAdminEnforcedLocked()) {
//判断是否充电亮屏,定义于BatteryService.java中
//从代码来看,只要mStayOnWhilePluggedInSetting设置了,就会亮屏
mStayOn = mBatteryManagerInternal.isPowered(mStayOnWhilePluggedInSetting);
} else {
mStayOn = false;
}
if (mStayOn != wasStayOn) {
mDirty |= DIRTY_STAY_ON;
}
}
}
这一部分的代码功能比较单一,主要用于更新变量mStayOn的值。
如果mStayOn如果为true,则屏幕保持长亮的状态。
3、updateScreenBrightnessBoostLocked
Android手机定义了一个最大屏幕亮度,用户可以手动或者让终端自动确定最大的屏幕亮度。
updateScreenBrightnessBoostLocked函数主要用于:更新终端可处于最大屏幕亮度的时间。
为了比较好的理解updateScreenBrightnessBoostLocked函数,
我们可以先分析一下与之相关的,PMS提供的对外的接口boostScreenBrightness。
该方法的作用是让屏幕在一段时间内保持最大的亮度,使屏幕在强光下有更好的可读性。
public void boostScreenBrightness(long eventTime) {
..........
mContext.enforceCallingOrSelfPermission(
android.Manifest.permission.DEVICE_POWER, null);
final int uid = Binder.getCallingUid();
final long ident = Binder.clearCallingIdentity();
try {
boostScreenBrightnessInternal(eventTime, uid);
} finally {
Binder.restoreCallingIdentity(ident);
}
}
private void boostScreenBrightnessInternal(long eventTime, int uid) {
synchronized (mLock) {
//系统没有准备好或者当前为Asleep状态, 不处理新到的事件
if (!mSystemReady || mWakefulness == WAKEFULNESS_ASLEEP
//过时的事件不处理
|| eventTime < mLastScreenBrightnessBoostTime) {
return;
}
..............
//记录事件到来的事件,也可以认为是终端处于最亮屏幕状态的起始时间
mLastScreenBrightnessBoostTime = eventTime;
//设置最亮屏幕的标志位true
if (!mScreenBrightnessBoostInProgress) {
mScreenBrightnessBoostInProgress = true;
//发送广播
mNotifier.onScreenBrightnessBoostChanged();
}
//修改mDirty的值,表示最大屏幕亮度的状态发生了变化
mDirty |= DIRTY_SCREEN_BRIGHTNESS_BOOST;
//记录
userActivityNoUpdateLocked(eventTime,
PowerManager.USER_ACTIVITY_EVENT_OTHER, 0, uid);
//更新电源的状态信息
updatePowerStateLocked();
}
}
从上面的代码可以看出,该函数:
首先,使用mLastScreenBrightnessBoostTime变量记录了终端处于最大屏幕亮度的起始时间;
然后,将最大屏幕亮度的标志位置为true,并修改mDirty标志位,以表示最大屏幕亮度的状态发生了变化;
最后,调用updatePowerStateLocked方法更新电源状态信息。
我们已经知道,updatePowerStateLocked将会调用到updateScreenBrightnessBoostLocked。
接下来,我们看看updateScreenBrightnessBoostLocked对应的代码:
private void updateScreenBrightnessBoostLocked(int dirty) {
//根据mDirty的标志位来判断终端屏幕最大可用亮度的状态是否发生了变化
if ((dirty & DIRTY_SCREEN_BRIGHTNESS_BOOST) != 0) {
//上面的代码已经提到过,当boostScreenBrightness接口被调用时,mScreenBrightnessBoostInProgress置为true
if (mScreenBrightnessBoostInProgress) {
//移除旧的超时事件
final long now = SystemClock.uptimeMillis();
mHandler.removeMessages(MSG_SCREEN_BRIGHTNESS_BOOST_TIMEOUT);
//终端处于最大屏幕亮度的时间,在sleep的时间之后,说明终端还未息屏之类的
if (mLastScreenBrightnessBoostTime > mLastSleepTime) {
//此时,重新计算终端可处于最大屏幕亮度的时间
final long boostTimeout = mLastScreenBrightnessBoostTime +
SCREEN_BRIGHTNESS_BOOST_TIMEOUT;
if (boostTimeout > now) {
Message msg = mHandler.obtainMessage(MSG_SCREEN_BRIGHTNESS_BOOST_TIMEOUT);
msg.setAsynchronous(true);
//发送延迟的超时事件
//当屏幕离开最大亮度状态时,该事件将被发送
//当该事件被处理时,会再次进入到updateScreenBrightnessBoostLocked函数
mHandler.sendMessageAtTime(msg, boostTimeout);
return;
}
}
//进入到这个分支时,说明屏幕处于最大亮度状态的时间已经超时了
//将该标志置为false
mScreenBrightnessBoostInProgress = false;
//发送广播
mNotifier.onScreenBrightnessBoostChanged();
//触发一次用户活动,写入mDirty标志位,同时做一些其它记录
userActivityNoUpdateLocked(now,
PowerManager.USER_ACTIVITY_EVENT_OTHER, 0, Process.SYSTEM_UID);
}
}
}
至此,PMS第一阶段更新基本状态信息的流程结束。
二、更新wakelock和用户活动
for (;;) {
int dirtyPhase1 = mDirty;
dirtyPhase2 |= dirtyPhase1;
mDirty = 0;
updateWakeLockSummaryLocked(dirtyPhase1);
updateUserActivitySummaryLocked(now, dirtyPhase1);
//updateWakefulnessLocked将决定系统是否进入休眠或dreaming状态
//主要是更新DIRTY_WAKEFULNESS位,如果不需要更新,则返回false
if (!updateWakefulnessLocked(dirtyPhase1)) {
break;
}
}
1、updateWakeLockSummaryLocked
updateWakeLockSummaryLocked函数根据PMS当前持有的所有WakeLock,得到当前终端整体的信息,保存到mWakeLockSummary变量中。
/**
* Updates the value of mWakeLockSummary to summarize the state of all active wake locks.
* Note that most wake-locks are ignored when the system is asleep.
*/
private void updateWakeLockSummaryLocked(int dirty) {
//PMS持有的WakeLock发生变化,或者唤醒状态发生变化时,才重新进行更新mWakeLockSummary
//例如:调用PMS的acquireWakeLock时,就会将dirty的DIRTY_WAKE_LOCKS位置1
if ((dirty & (DIRTY_WAKE_LOCKS | DIRTY_WAKEFULNESS)) != 0) {
mWakeLockSummary = 0;
final int numWakeLocks = mWakeLocks.size();
for (int i = 0; i < numWakeLocks; i++) {
final WakeLock wakeLock = mWakeLocks.get(i);
//这里只关注WakeLock的level
//下面的代码其实就是实现每个level WakeLock对应的注释信息
switch (wakeLock.mFlags & PowerManager.WAKE_LOCK_LEVEL_MASK) {
case PowerManager.PARTIAL_WAKE_LOCK:
//在分析PMS acquireWakeLock的流程时,已经提到过
//在doze模式下,不在白名单内的非系统应用申请PARTIAL_WAKE_LOCK时,将被disabled
if (!wakeLock.mDisabled) {
// We only respect this if the wake lock is not disabled.
mWakeLockSummary |= WAKE_LOCK_CPU;
}
break;
case PowerManager.FULL_WAKE_LOCK:
mWakeLockSummary |= WAKE_LOCK_SCREEN_BRIGHT | WAKE_LOCK_BUTTON_BRIGHT;
break;
case PowerManager.SCREEN_BRIGHT_WAKE_LOCK:
mWakeLockSummary |= WAKE_LOCK_SCREEN_BRIGHT;
break;
case PowerManager.SCREEN_DIM_WAKE_LOCK:
mWakeLockSummary |= WAKE_LOCK_SCREEN_DIM;
break;
case PowerManager.PROXIMITY_SCREEN_OFF_WAKE_LOCK:
mWakeLockSummary |= WAKE_LOCK_PROXIMITY_SCREEN_OFF;
break;
case PowerManager.DOZE_WAKE_LOCK:
mWakeLockSummary |= WAKE_LOCK_DOZE;
break;
case PowerManager.DRAW_WAKE_LOCK:
mWakeLockSummary |= WAKE_LOCK_DRAW;
break;
}
}
// Cancel wake locks that make no sense based on the current state.
//从下面的代码可以看出,PMS中的mWakefulness变量记录了终端当前的状态
//下面就是移除在特定状态下,没有意义的WakeLock
if (mWakefulness != WAKEFULNESS_DOZING) {
//如果不是Dozing状态,移除相应的wakeLock标志位
mWakeLockSummary &= ~(WAKE_LOCK_DOZE | WAKE_LOCK_DRAW);
}
if (mWakefulness == WAKEFULNESS_ASLEEP
|| (mWakeLockSummary & WAKE_LOCK_DOZE) != 0) {
//如果当前为Asleep或者有Doze的wakeLock锁的时候,应该移除掉屏幕亮度相关的wakeLock锁
mWakeLockSummary &= ~(WAKE_LOCK_SCREEN_BRIGHT | WAKE_LOCK_SCREEN_DIM
| WAKE_LOCK_BUTTON_BRIGHT);
if (mWakefulness == WAKEFULNESS_ASLEEP) {
//休眠时,sensor不再需要监听终端是否靠近物体,以触发亮灭屏
mWakeLockSummary &= ~WAKE_LOCK_PROXIMITY_SCREEN_OFF;
}
}
// Infer implied wake locks where necessary based on the current state.
//根据当前的状态,及PMS持有的WakeLock,推断出隐含的持锁需求
//例如:当PMS持有亮屏锁WAKE_LOCK_SCREEN_BRIGHT时,若当前终端为唤醒态
//那么CPU显然也需要处于唤醒态
if ((mWakeLockSummary & (WAKE_LOCK_SCREEN_BRIGHT | WAKE_LOCK_SCREEN_DIM)) != 0) {
if (mWakefulness == WAKEFULNESS_AWAKE) {
mWakeLockSummary |= WAKE_LOCK_CPU | WAKE_LOCK_STAY_AWAKE;
} else if (mWakefulness == WAKEFULNESS_DREAMING) {
mWakeLockSummary |= WAKE_LOCK_CPU;
}
}
if ((mWakeLockSummary & WAKE_LOCK_DRAW) != 0) {
mWakeLockSummary |= WAKE_LOCK_CPU;
}
...................
}
}
结合每个WakeLock level的注释信息,以上代码还是比较好理解的。
这里唯一需要说明的是,Android定义一个mWakeLockSummary变量的原因是:
PMS将WakeLock定义为不同进程的请求信息,这些请求信息对CPU、屏幕和键盘有不同的需求。
对于每一种资源而言,只要有一个申请满足获取条件,PMS就需要为终端分配该申请对应的资源。
例如:假设PMS有20个WakeLock,只有1个申请亮屏,另外19个只申请CPU唤醒,PMS仍然需要保持终端亮屏。
因此,mWakeLockSummary就提供了一种整合多个WakeLock请求的功能,方便PMS进行集中的控制。
2、updateUserActivitySummaryLocked
updateUserActivitySummaryLocked主要根据用户最后的活动来决定当前屏幕的状态。
/**
* Updates the value of mUserActivitySummary to summarize the user requested
* state of the system such as whether the screen should be bright or dim.
* Note that user activity is ignored when the system is asleep.
*/
private void updateUserActivitySummaryLocked(long now, int dirty) {
if ((dirty & (DIRTY_WAKE_LOCKS | DIRTY_USER_ACTIVITY
| DIRTY_WAKEFULNESS | DIRTY_SETTINGS)) != 0) {
mHandler.removeMessages(MSG_USER_ACTIVITY_TIMEOUT);
long nextTimeout = 0;
if (mWakefulness == WAKEFULNESS_AWAKE
|| mWakefulness == WAKEFULNESS_DREAMING
|| mWakefulness == WAKEFULNESS_DOZING) {
//获取进入休眠状态的时间sleepTimeout
//getSleepTimeoutLocked中会判断休眠时间和屏幕熄灭时间的关系
//如果休眠时间sleepTimeout小于屏幕熄灭时间screenOfftime,
//则休眠时间被调整为屏幕熄灭时间,因为屏幕亮屏状态下,终端不能进入休眠
final int sleepTimeout = getSleepTimeoutLocked();
//获取屏幕熄灭的时间
final int screenOffTimeout = getScreenOffTimeoutLocked(sleepTimeout);
//获取屏幕变暗的时间
final int screenDimDuration = getScreenDimDurationLocked(screenOffTimeout);
//当Window Manager判定用户inactive时,将此标志置为true
final boolean userInactiveOverride = mUserInactiveOverrideFromWindowManager;
//类似于之前的mWakeLockSummary,将当前的用户事件,转化为PMS可以处理的屏幕状态
mUserActivitySummary = 0;
//在唤醒的状态下,发生过用户事件
if (mLastUserActivityTime >= mLastWakeTime) {
//重新计算出屏幕需要变暗的时间
nextTimeout = mLastUserActivityTime
+ screenOffTimeout - screenDimDuration;
if (now < nextTimeout) {
//如果没有到达需要变暗的时间,那么当前屏幕的状态为USER_ACTIVITY_SCREEN_BRIGHT(亮屏)
mUserActivitySummary = USER_ACTIVITY_SCREEN_BRIGHT;
} else {
//到达变暗的时间,则计算出屏幕熄灭的时间
nextTimeout = mLastUserActivityTime + screenOffTimeout;
if (now < nextTimeout) {
//还没到熄灭的时间,则当前屏幕的状态为USER_ACTIVITY_SCREEN_DIM(暗屏)
mUserActivitySummary = USER_ACTIVITY_SCREEN_DIM;
}
}
}
//注意mUserActivitySummary为0才会进入下面的分支
//即上面改变mUserActivitySummary的条件不满足时,才会进入这个分支(例如:唤醒状态下,没发生过改变屏幕状态的UserActivity)
if (mUserActivitySummary == 0
//mLastUserActivityTimeNoChangeLights表示用户最后的活动不会改变屏幕当前的状态
&& mLastUserActivityTimeNoChangeLights >= mLastWakeTime) {
//计算下次屏幕熄灭的时间
nextTimeout = mLastUserActivityTimeNoChangeLights + screenOffTimeout;
//还未到达熄屏时间
if (now < nextTimeout) {
if (mDisplayPowerRequest.policy == DisplayPowerRequest.POLICY_BRIGHT) {
//当前屏幕是亮屏,仍然设置为亮屏
mUserActivitySummary = USER_ACTIVITY_SCREEN_BRIGHT;
} else if (mDisplayPowerRequest.policy == DisplayPowerRequest.POLICY_DIM) {
//当前屏幕是变暗,仍然设置为变暗
mUserActivitySummary = USER_ACTIVITY_SCREEN_DIM;
}
}
}
if (mUserActivitySummary == 0) {
//若定义了有效的休眠时间
if (sleepTimeout >= 0) {
//计算用户最后的活动时间
final long anyUserActivity = Math.max(mLastUserActivityTime,
mLastUserActivityTimeNoChangeLights);
//只有在唤醒状态下,进行了用户活动,才会重新更新休眠时间 (此时,应该是有过用户活动,但过了息屏时间了)
if (anyUserActivity >= mLastWakeTime) {
nextTimeout = anyUserActivity + sleepTimeout;
if (now < nextTimeout) {
//走到这个分支,应该是屏幕已经熄灭,但还未到达休眠状态,先进入dream态
mUserActivitySummary = USER_ACTIVITY_SCREEN_DREAM;
}
}
} else {
//直接进入dream态,后文的updateWakefulnessLocked将判断是否休眠
mUserActivitySummary = USER_ACTIVITY_SCREEN_DREAM;
nextTimeout = -1;
}
}
//如果屏幕未进入dream态,但Window Manager判定用户inactive,则进入下面分支
if (mUserActivitySummary != USER_ACTIVITY_SCREEN_DREAM && userInactiveOverride) {
//如果屏幕未熄灭
if ((mUserActivitySummary &
(USER_ACTIVITY_SCREEN_BRIGHT | USER_ACTIVITY_SCREEN_DIM)) != 0) {
// Device is being kept awake by recent user activity
if (nextTimeout >= now && mOverriddenTimeout == -1) {
// Save when the next timeout would have occurred
mOverriddenTimeout = nextTimeout;
}
}
//Window Manager的权限很大,如果它判断用户inactive,直接进入dream态
mUserActivitySummary = USER_ACTIVITY_SCREEN_DREAM;
nextTimeout = -1;
}
//根据nextTimeOut延迟发送信息,信息被处理后,将重新调用updatePowerStateLocked,于是再次进入到该方法
//通过不断进入该方法,不断评估是否根据用户动作亮、熄屏等
if (mUserActivitySummary != 0 && nextTimeout >= 0) {
Message msg = mHandler.obtainMessage(MSG_USER_ACTIVITY_TIMEOUT);
msg.setAsynchronous(true);
mHandler.sendMessageAtTime(msg, nextTimeout);
}
} else {
mUserActivitySummary = 0;
}
..........
}
}
从上面的代码可以看出,在该函数中用mUserActivitySummary变量存储当前屏幕的状态。
一共有3中基本状态:
* USER_ACTIVITY_SCREEN_BRIGHT 点亮屏幕
* USER_ACTIVITY_SCREEN_DIM 屏幕变暗
* USER_ACTIVITY_SCREEN_DREAM 屏保状态
从代码可以看出,屏幕变化和userActivity活动有关,它根据最后的userActivity活动的时间决定点亮屏幕、调暗屏幕或熄灭屏幕。
之前的很多方法中都会调用userActivityNoUpdateLocked方法。该方法将触发一次用户活动,以更新用户活动的时间,这样屏幕变暗和熄灭时间就会重新进行计算。
这也就是为什么用户一直操作手机,屏幕不会熄灭或者变暗的原因。
大图地址
整体来讲,个人感觉这个函数的代码写的还是挺绕的,因此还是作一个图记录一下。
大家有兴趣可以看一下。
3、updateWakefulnessLocked
从之前的代码可以看出,updateWakefulnessLocked将决定第二阶段的电源状态更新是否结束。
我们看一下updateWakefulnessLocked函数:
/**
* Updates the wakefulness of the device.
*
* This is the function that decides whether the device should start dreaming
* based on the current wake locks and user activity state. It may modify mDirty
* if the wakefulness changes.
*
* Returns true if the wakefulness changed and we need to restart power state calculation.
*/
private boolean updateWakefulnessLocked(int dirty) {
boolean changed = false;
//下面的条件还是比较容易满足的,基本上只要之前的流程更改过mDirty就会进入分支
if ((dirty & (DIRTY_WAKE_LOCKS | DIRTY_USER_ACTIVITY | DIRTY_BOOT_COMPLETED
| DIRTY_WAKEFULNESS | DIRTY_STAY_ON | DIRTY_PROXIMITY_POSITIVE
| DIRTY_DOCK_STATE)) != 0) {
//如果当前的状态是唤醒的,isItBedTimeYetLocked判定不能再保持唤醒态
if (mWakefulness == WAKEFULNESS_AWAKE && isItBedTimeYetLocked()) {
................
final long time = SystemClock.uptimeMillis();
//主要根据设置信息,判断是否满足进入Dream状态的条件
if (shouldNapAtBedTimeLocked()) {
//将mWakefullness的值置为WAKEFULNESS_DREAMING,修改mDirty变量,并进行通知等
changed = napNoUpdateLocked(time, Process.SYSTEM_UID);
} else {
//将mWakefullness的值置为WAKEFULNESS_DOZING
//如果系统设置了跳过Dozing态,则将mWakefullness置为WAKEFULNESS_ASLEEP
//同时修改mDirty变量,并进行通知等
changed = goToSleepNoUpdateLocked(time,
PowerManager.GO_TO_SLEEP_REASON_TIMEOUT, 0, Process.SYSTEM_UID);
}
//注意:napNoUpdateLocked和goToSleepNoUpdateLocked函数正常执行后,
//都会将mSandmanSummoned(被"睡魔"眷顾了)置为true
}
}
return changed;
}
从上面的代码可以看出,如果终端可以一直保持唤醒状态,或一开始就是非唤醒态,
那么mWakefulness不会发生改变,第二阶段的for循环将会break;
如果终端要从唤醒态变为非唤醒态,那么for循环将再运行一次,即重新计算一次mWakeLockSummary和mUserActivitySummary。
这么做的原因是:updateWakeLockSummaryLocked和updateUserActivitySummaryLocked函数的一些计算,与终端是否处于唤醒状态,即mWakefulness的值有关。
由于这两个函数并不会修改mWakefulness,因此在这一次运行时,updateWakefulnessLocked将返回false,即第二阶段结束。
因此,我们可以得出结论:更新电源状态的第二阶段,正常情况下最多运行两次。
在第二阶段的最后,我们看一下isItBedTimeYetLocked函数:
/**
* Returns true if the device should go to sleep now.
* Also used when exiting a dream to determine whether we should go back
* to being fully awake or else go to sleep for good.
*/
private boolean isItBedTimeYetLocked() {
//主要由isBeingKeptAwakeLocked决定
return mBootCompleted && !isBeingKeptAwakeLocked();
}
/**
* Returns true if the device is being kept awake by a wake lock, user activity
* or the stay on while powered setting. We also keep the phone awake when
* the proximity sensor returns a positive result so that the device does not
* lock while in a phone call. This function only controls whether the device
* will go to sleep or dream which is independent of whether it will be allowed
* to suspend.
*/
//根据状态,判断终端是否应该处于唤醒状态
private boolean isBeingKeptAwakeLocked() {
return mStayOn
|| mProximityPositive
|| (mWakeLockSummary & WAKE_LOCK_STAY_AWAKE) != 0
|| (mUserActivitySummary & (USER_ACTIVITY_SCREEN_BRIGHT
| USER_ACTIVITY_SCREEN_DIM)) != 0
|| mScreenBrightnessBoostInProgress;
}
参考原生代码的注释,这一部分代码还是比较好理解的。
三、更新display power state
第三阶段将负责更新屏幕的显示状态。
/**
* Updates the display power state asynchronously.
* When the update is finished, mDisplayReady will be set to true. The display
* controller posts a message to tell us when the actual display power state
* has been updated so we come back here to double-check and finish up.
*
* This function recalculates the display power state each time.
*/
private boolean updateDisplayPowerStateLocked(int dirty) {
final boolean oldDisplayReady = mDisplayReady;
//mDirty满足条件时,进入以下分支
if ((dirty & (DIRTY_WAKE_LOCKS | DIRTY_USER_ACTIVITY | DIRTY_WAKEFULNESS
| DIRTY_ACTUAL_DISPLAY_POWER_STATE_UPDATED | DIRTY_BOOT_COMPLETED
| DIRTY_SETTINGS | DIRTY_SCREEN_BRIGHTNESS_BOOST)) != 0) {
//根据mWakefullness、mWakeLockSummary、mUserActivitySummary等,决定屏幕的policy
//policy定义为DisplayPowerRequest.(POLICY_OFF、POLICY_DOZE、POLICY_BRIGHT和POLICY_DIM)
mDisplayPowerRequest.policy = getDesiredScreenPolicyLocked();
// Determine appropriate screen brightness and auto-brightness adjustments.
//决定屏幕的亮度
.................
// Update display power request.
// 更新mDisplayPowerRequest的参数
...................
//实际上调用DisplayPowerController的requestPowerState函数
//在初始时,PMS注册了mDisplayPowerCallbacks到DisplayPowerController中,
//当更新完成后,会回调定义的接口,重新updatePowerStateLocked
mDisplayReady = mDisplayManagerInternal.requestPowerState(mDisplayPowerRequest,
mRequestWaitForNegativeProximity);
...................
}
return mDisplayReady && !oldDisplayReady;
}
我们跟进一下DisplayPowerController的requestPowerState函数:
/**
* Requests a new power state.
* The controller makes a copy of the provided object and then
* begins adjusting the power state to match what was requested.
*/
public boolean requestPowerState(DisplayPowerRequest request,
boolean waitForNegativeProximity) {
.........
synchronized (mLock) {
boolean changed = false;
//新需求增加:proximity sensor需要检测距离
if (waitForNegativeProximity
//原来没有这个需求
&& !mPendingWaitForNegativeProximityLocked) {
mPendingWaitForNegativeProximityLocked = true;
changed = true;
}
//以下表示,参数中的Request对于DisplayPowerController而言,是一个新的需求
if (mPendingRequestLocked == null) {
mPendingRequestLocked = new DisplayPowerRequest(request);
changed = true;
} else if (!mPendingRequestLocked.equals(request)) {
mPendingRequestLocked.copyFrom(request);
changed = true;
}
if (changed) {
//一但有新的需求,mDisplayReadyLocked就是false,表示屏幕有待调整
mDisplayReadyLocked = false;
}
//有新需求,同时有对应的request
if (changed && !mPendingRequestChangedLocked) {
mPendingRequestChangedLocked = true;
//发送消息,更新屏幕状态
//最终通过DisplayPowerController的updatePowerState函数,进行屏幕状态更新
//这部分代码也极其复杂,暂时不在这里展开分析
//更新屏幕状态后,将回调PMS的接口
sendUpdatePowerStateLocked();
}
return mDisplayReadyLocked;
}
}
根据requestPowerState的代码,我们知道:
当PMS传入一个新的mDisplayPowerRequest时,requestPowerState应该返回为false;当DisplayPowerController按照mDisplayPowerRequest修改完屏幕状态,再次进入回到updateDisplayPowerStateLocked函数,调用requestPowerState时才会返回true。
这一阶段的代码,我们只是分析了整个过程的冰山一角,并没有分析更新屏幕状态的实际操作。
但从现有的代码可以看出,PMS的作用仅仅是维护终端电源相关状态,实际的工作还是通过类似发送Request的方式,让其它的服务协助完成。
例如:在整个阶段,PMS根据之前得到信息,构造出DisplayPowerRequest,然后发送给DisplayPowerController进行实际的处理。
当DisplayPowerController完成实际的工作(部分工作还依赖于PhoneWindowManager)后,再通知PMS进行复查。
因此PMS的定位,确实可以用一个”Manager”来形容;
负责整个终端信息的搜集和维护,然后将相应的工作指派给具体的“员工”执行;
“员工”执行完毕后,向”Manager”汇报;
“Manager”检查工作的完成情况后,然后做出下一步的指示。
四、更新dream state
updateDreamLocked函数主要用于更新屏保状态,当设备进入或者退出屏保的时候都会触发这个方法:
private void updateDreamLocked(int dirty, boolean displayBecameReady) {
if ((dirty & (DIRTY_WAKEFULNESS
| DIRTY_USER_ACTIVITY
| DIRTY_WAKE_LOCKS
| DIRTY_BOOT_COMPLETED
| DIRTY_SETTINGS
| DIRTY_IS_POWERED
| DIRTY_STAY_ON
| DIRTY_PROXIMITY_POSITIVE
| DIRTY_BATTERY_STATE)) != 0 || displayBecameReady) {
if (mDisplayReady) {
//mDirty满足条件,同时屏幕状态调整完毕,才进入下一步
scheduleSandmanLocked();
}
}
private void scheduleSandmanLocked() {
if (!mSandmanScheduled) {
//mSandmanScheduled的作用就是让MessageQueue中仅保留一个MSG_SANDMAN
mSandmanScheduled = true;
//由handleSandman处理
Message msg = mHandler.obtainMessage(MSG_SANDMAN);
msg.setAsynchronous(true);
mHandler.sendMessage(msg);
}
}
handleSandman函数比较复杂,主要用于决定设备是否应该停留在dreaming或dozing状态。
我们分段介绍该函数的功能。
1、决定是否可以进入屏保状态
/**
* Called when the device enters or exits a dreaming or dozing state.
*/
private void handleSandman() {
final boolean startDreaming;
final int wakefulness;
synchronized (mLock) {
mSandmanScheduled = false;
wakefulness = mWakefulness;
//前面提到过,当updateWakefulnessLocked判断进入dozing或sleep状态时,
//会将mSandmanSummoned置为true
//mDisplayReady主要确保前面屏幕状态更新完毕
if (mSandmanSummoned && mDisplayReady) {
//判断device是否可以dream或dozing
startDreaming = canDreamLocked() || canDozeLocked();
mSandmanSummoned = false;
} else {
startDreaming = false;
}
}
..........
这段代码主要用于确定,设备是否可以看是dreaming。
除去前置条件的限制外,此处的结果主要由canDreamLocked和canDozeLocked决定。
我们分别看看这两个函数:
/**
* Returns true if the device is allowed to dream in its current state.
*/
private boolean canDreamLocked() {
//mWakefulness等于WAKEFULNESS_DREAMING
if (mWakefulness != WAKEFULNESS_DREAMING
//设备支持dreaming
|| !mDreamsSupportedConfig
//设置开关开启
|| !mDreamsEnabledSetting
//屏幕熄灭
|| !mDisplayPowerRequest.isBrightOrDim()
|| (mUserActivitySummary & (USER_ACTIVITY_SCREEN_BRIGHT
| USER_ACTIVITY_SCREEN_DIM | USER_ACTIVITY_SCREEN_DREAM)) == 0
//初始化完成
|| !mBootCompleted) {
return false;
}
//以上条件均满足,才能进入后面的判断
//不处于唤醒态
if (!isBeingKeptAwakeLocked()) {
//没充电,电源选项也未配置,不可dreaming
if (!mIsPowered && !mDreamsEnabledOnBatteryConfig) {
return false;
}
//没充电,且电池电量过低,不可dreaming
if (!mIsPowered
&& mDreamsBatteryLevelMinimumWhenNotPoweredConfig >= 0
&& mBatteryLevel < mDreamsBatteryLevelMinimumWhenNotPoweredConfig) {
return false;
}
//充电,但电池电量过低,不可dreaming
if (mIsPowered
&& mDreamsBatteryLevelMinimumWhenPoweredConfig >= 0
&& mBatteryLevel < mDreamsBatteryLevelMinimumWhenPoweredConfig) {
return false;
}
//充电和未充电分别有一个最低的dreaming电量门限
}
return true;
}
从上面的代码可以看出,dreaming除了对终端当前的状态、配置项有关外,在非唤醒状态下还与当前的电池电量有关系。
canDozeLocked函数相对简单:
private boolean canDozeLocked() {
return mWakefulness == WAKEFULNESS_DOZING;
}
2、在必要时,进入屏保状态
// Start dreaming if needed.
final boolean isDreaming;
if (mDreamManager != null) {
if (startDreaming) {
//结束旧梦
mDreamManager.stopDream(false /*immediate*/);
//开启新梦
mDreamManager.startDream(wakefulness == WAKEFULNESS_DOZING);
}
//startDream成功后,一般isDreaming就会返回true
isDreaming = mDreamManager.isDreaming();
} else {
isDreaming = false;
}
决定了是否可以进入屏保状态后,这一部分就开始进行实际的工作。
mDreamManager为DreamManagerService的Binder代理。
我们重点看看DreamManagerService的startDream函数,stopDream的工作内容与startDream相反,不做细致分析:
//定义于DreamManagerService的内部类中
public void startDream(boolean doze) {
startDreamInternal(doze);
}
//定义于DreamManagerService
private void startDreamInternal(boolean doze) {
final int userId = ActivityManager.getCurrentUser();
//个人觉得这里应该是获取屏保对象
final ComponentName dream = chooseDreamForUser(doze, userId);
if (dream != null) {
synchronized (mLock) {
startDreamLocked(dream, false /*isTest*/, doze, userId);
}
}
}
private void startDreamLocked(final ComponentName name,
final boolean isTest, final boolean canDoze, final int userId) {
//申请的屏保与当前的一致,不用进行修改
if (Objects.equal(mCurrentDreamName, name)
&& mCurrentDreamIsTest == isTest
&& mCurrentDreamCanDoze == canDoze
&& mCurrentDreamUserId == userId) {
return;
}
//立即停止当前的屏保
stopDreamLocked(true /*immediate*/);
final Binder newToken = new Binder();
mCurrentDreamToken = newToken;
mCurrentDreamName = name;
mCurrentDreamIsTest = isTest;
mCurrentDreamCanDoze = canDoze;
mCurrentDreamUserId = userId;
mHandler.post(new Runnable() {
@Override
public void run() {
//调用DreamController的startDream函数
mController.startDream(newToken, name, isTest, canDoze, userId);
}
});
}
//定义于DreamController中
public void startDream(Binder token, ComponentName name,
boolean isTest, boolean canDoze, int userId) {
//移除当前屏保并回调通知
stopDream(true /*immediate*/);
.........
try {
..............
//记录dream
mCurrentDream = new DreamRecord(token, name, isTest, canDoze, userId);
mDreamStartTime = SystemClock.elapsedRealtime();
..............
//做好屏幕相关的准备工作
try {
mIWindowManager.addWindowToken(token, WindowManager.LayoutParams.TYPE_DREAM);
} catch (RemoteException ex) {
Slog.e(TAG, "Unable to add window token for dream.", ex);
stopDream(true /*immediate*/);
return;
}
Intent intent = new Intent(DreamService.SERVICE_INTERFACE);
intent.setComponent(name);
intent.addFlags(Intent.FLAG_ACTIVITY_EXCLUDE_FROM_RECENTS);
try {
//拉起屏保服务
if (!mContext.bindServiceAsUser(intent, mCurrentDream,
Context.BIND_AUTO_CREATE | Context.BIND_FOREGROUND_SERVICE,
new UserHandle(userId))) {
Slog.e(TAG, "Unable to bind dream service: " + intent);
stopDream(true /*immediate*/);
return;
} catch (SecurityException ex) {
............
stopDream(true /*immediate*/);
return;
}
mCurrentDream.mBound = true;
//在DREAM_CONNECTION_TIMEOUT到期时,bind服务还未成功,runnable就负责结束dream
mHandler.postDelayed(mStopUnconnectedDreamRunnable, DREAM_CONNECTION_TIMEOUT);
} finally {
........
}
}
从这部分代码我们知道了,所谓的屏保其实就是拉起一个特殊的服务。
3、更新屏保状态
// Update dream state.
synchronized (mLock) {
// Remember the initial battery level when the dream started.
if (startDreaming && isDreaming) {
mBatteryLevelWhenDreamStarted = mBatteryLevel;
................
}
// If preconditions changed, wait for the next iteration to determine
// whether the dream should continue (or be restarted).
//例如:mDisplayReady为false时, mSandmanSummoned保持为false
if (mSandmanSummoned || mWakefulness != wakefulness) {
return; // wait for next cycle
}
.............
// Determine whether the dream should continue.
if (wakefulness == WAKEFULNESS_DREAMING) {
if (isDreaming && canDreamLocked()) {
if (mDreamsBatteryLevelDrainCutoffConfig >= 0
//下面这句我是懵逼的,这不是必然成立的么?
//也就是只要配置了mDreamsBatteryLevelDrainCutoffConfig就会成立
//按注释来讲,这里好歹重新取一次mBatteryLevel啊!!!怀疑是bug点
&& mBatteryLevel < mBatteryLevelWhenDreamStarted
- mDreamsBatteryLevelDrainCutoffConfig
&& !isBeingKeptAwakeLocked()) {
// If the user activity timeout expired and the battery appears
// to be draining faster than it is charging then stop dreaming
// and go to sleep.
} else {
return; // continue dreaming
}
}
// Dream has ended or will be stopped. Update the power state.
if (isItBedTimeYetLocked()) {
//休眠
goToSleepNoUpdateLocked(SystemClock.uptimeMillis(),
PowerManager.GO_TO_SLEEP_REASON_TIMEOUT, 0, Process.SYSTEM_UID);
updatePowerStateLocked();
} else {
//唤醒
wakeUpNoUpdateLocked(SystemClock.uptimeMillis(), "android.server.power:DREAM",
Process.SYSTEM_UID, mContext.getOpPackageName(), Process.SYSTEM_UID);
updatePowerStateLocked();
}
} else if (wakefulness == WAKEFULNESS_DOZING) {
if (isDreaming) {
return; // continue dozing
}
// Doze has ended or will be stopped. Update the power state.
reallyGoToSleepNoUpdateLocked(SystemClock.uptimeMillis(), Process.SYSTEM_UID);
updatePowerStateLocked();
}
}
// Stop dream.
//执行到这里说明退出了dreaming状态,如果之前拉起过屏保服务,此时应该停止它
if (isDreaming) {
mDreamManager.stopDream(false /*immediate*/);
}
...........
以上是PMS更新屏保状态的基本流程,整体来看相当的繁琐。
我们还是用一个图来整体整理一下:
大图链接
这部分代码最后太乱,每次更新状态后都会重新调用updatePowerStateLocked,然后再次进入到handleSandman函数中。
这种反复地递归调用,比较难以把控。
五、更新suspend blocker
updateSuspendBlockerLocked函数主要根据之前流程的执行结果,持有或者释放CPU和屏幕的锁。
我们一起来看看对应的函数:
private void updateSuspendBlockerLocked() {
//根据是否有CPU的wakelock,来决定cpu是保持否唤醒
final boolean needWakeLockSuspendBlocker = ((mWakeLockSummary & WAKE_LOCK_CPU) != 0);
//根据前面屏幕相关的状态,来决定是否需要持有屏幕的锁
final boolean needDisplaySuspendBlocker = needDisplaySuspendBlockerLocked();
//屏幕如果不需要保持开启状态,那么可自动熄灭
final boolean autoSuspend = !needDisplaySuspendBlocker;
//应该是表示屏幕是否是可交互的
final boolean interactive = mDisplayPowerRequest.isBrightOrDim();
// Disable auto-suspend if needed.
//autoSuspend为false,说明屏幕还需要点亮
if (!autoSuspend && mDecoupleHalAutoSuspendModeFromDisplayConfig) {
//通过native函数,调用底层的autosuspend_disable
setHalAutoSuspendModeLocked(false);
}
// First acquire suspend blockers if needed.
//在需要的情况下,获取CPU和屏幕的锁
if (needWakeLockSuspendBlocker && !mHoldingWakeLockSuspendBlocker) {
mWakeLockSuspendBlocker.acquire();
mHoldingWakeLockSuspendBlocker = true;
}
if (needDisplaySuspendBlocker && !mHoldingDisplaySuspendBlocker) {
mDisplaySuspendBlocker.acquire();
mHoldingDisplaySuspendBlocker = true;
}
// Inform the power HAL about interactive mode.
if (mDecoupleHalInteractiveModeFromDisplayConfig) {
if (interactive || mDisplayReady) {
//调用底层动态库的setInteractive函数,决定终端是否可以进行交互
setHalInteractiveModeLocked(interactive);
}
}
// Then release suspend blockers if needed.
//如果不需要,则释放CPU和屏幕的锁
if (!needWakeLockSuspendBlocker && mHoldingWakeLockSuspendBlocker) {
mWakeLockSuspendBlocker.release();
mHoldingWakeLockSuspendBlocker = false;
}
if (!needDisplaySuspendBlocker && mHoldingDisplaySuspendBlocker) {
mDisplaySuspendBlocker.release();
mHoldingDisplaySuspendBlocker = false;
}
// Enable auto-suspend if needed.
//如果需要设置自动休眠模式
if (autoSuspend && mDecoupleHalAutoSuspendModeFromDisplayConfig) {
setHalAutoSuspendModeLocked(true);
}
}
从上面的代码可以看出PMS是非常依赖于native层的,真实的持锁、释放锁、设置交互状态等工作,均是移交到native层进行操作。
我们以mWakeLockSuspendBlocker的处理流程为例,看看native的调用过程:
..........
mWakeLockSuspendBlocker = createSuspendBlockerLocked("PowerManagerService.WakeLocks");
.........
之前的博客也提到过,PMS在其构造函数中调用createSuspendBlockerLocked函数,创建出了mWakeLockSuspendBlocker:
private SuspendBlocker createSuspendBlockerLocked(String name) {
//实际对象为PMS内部类SuspendBlockerImpl
SuspendBlocker suspendBlocker = new SuspendBlockerImpl(name);
mSuspendBlockers.add(suspendBlocker);
return suspendBlocker;
}
从上面的代码,我们知道当PMS需要获取底层锁时,调用的是SuspendBlockerImpl的acquire函数:
public void acquire() {
synchronized (this) {
mReferenceCount += 1;
if (mReferenceCount == 1) {
.......
//调用到了native层
nativeAcquireSuspendBlocker(mName);
}
}
}
在native层的com_android_server_power_PowerManagerService.cpp中,对应的native函数为:
static void nativeAcquireSuspendBlocker(JNIEnv *env, jclass /* clazz */, jstring nameStr) {
ScopedUtfChars name(env, nameStr);
//获取的是PARTIAL_WAKE_LOCK的类型,即保持CPU唤醒的
acquire_wake_lock(PARTIAL_WAKE_LOCK, name.c_str());
}
从这里的代码我们不难发现,尽管PMS定义了不同的WakeLock等级,但当通过PMS的native函数调用HAL层函数acquire_wake_lock时,使用的都是PARTIAL_WAKE_LOCK。
个人觉得这是可以理解的,当其它进程向PMS申请保持屏幕唤醒的Framework层WakeLock后,PMS在Framework层就进行了对应的处理,例如将请求信息等地交给DisplayPowerController等处理。因此,对于底层的HAL层而言,只需要关注CPU是否需要保持唤醒即可。
HAL层函数acquire_wake_lock,最后会向/sys/power/wake_lock节点进行write操作。
总结
至此,updatePowerStateLocked的基本流程介绍完毕,大体上如下图所示:
通过其中的源码,我们也能看出仅管理当前的状态,涉及的细节就非常的琐碎。
而屏幕和CPU的实际控制,还牵扯到大量其它对象和HAL层代码。
Android电源的管理实际上是基于Linux电源管理策略的,因此若要真正掌握,还需要对Linux的电源管理策略作进一步的了解。
由于个人水平有限,目前还无法高屋建瓴地整体分析宏观的电源管理架构,细节也有一些遗漏。
后续争取以此博客为基础,不断迭代,以求更进一步地了解PMS的知识。