React Native仅Android7.x触发ModuleHolder.getModule类ANR问题

问题现象 #

近期仅在Android 7/7.1/7.1.1系统上偶现ANR问题，具体堆栈信息如下：

Input dispatching timed out (Waiting because no window has focus but there is a focused application that may eventually add a window when it finishes starting up.)

"main" prio=5 tid=1 Waiting

at java.lang.Object.wait! (Native method)
  at com.facebook.react.bridge.ModuleHolder.getModule (ModuleHolder.java:14)
  at com.facebook.react.bridge.NativeModuleRegistry.getModule (NativeModuleRegistry.java:11)
  at com.facebook.react.bridge.CatalystInstanceImpl.getNativeModule (CatalystInstanceImpl.java:5)
  at com.facebook.react.bridge.CatalystInstanceImpl.getNativeModule (CatalystInstanceImpl.java:1)
  at com.facebook.internal.instrument.InstrumentUtility.getUIManager (InstrumentUtility.java:4)
  at com.facebook.react.ReactInstanceManager.attachRootViewToInstance (ReactInstanceManager.java:4)
  at com.facebook.react.ReactRootView.attachToReactInstanceManager (ReactRootView.java:12)
  at com.facebook.react.ReactRootView.startReactApplication (ReactRootView.java:9)
  at com.myapp.ReactActivity.onCreate (ReactActivity.java:191)
  at android.app.Activity.performCreate (Activity.java:6845)
  at android.app.Instrumentation.callActivityOnCreate (Instrumentation.java:1119)
  at android.app.ActivityThread.performLaunchActivity (ActivityThread.java:2700)
  at android.app.ActivityThread.handleLaunchActivity (ActivityThread.java:2808)
  at android.app.ActivityThread.-wrap12 (ActivityThread.java)
  at android.app.ActivityThread$H.handleMessage (ActivityThread.java:1541)
  at android.os.Handler.dispatchMessage (Handler.java:102)
  at android.os.Looper.loop (Looper.java:165)
  at android.app.ActivityThread.main (ActivityThread.java:6375)
  at java.lang.reflect.Method.invoke! (Native method)
  at com.android.internal.os.ZygoteInit$MethodAndArgsCaller.run (ZygoteInit.java:912)
  at com.android.internal.os.ZygoteInit.main (ZygoteInit.java:802)

"Jit thread pool worker thread 0" prio=5 tid=2 Native (still starting up)
  #00  pc 000000000001c0ac  /system/lib64/libc.so (syscall+28)
  #00  pc 00000000000e6dd4  /system/lib64/libart.so (art::ConditionVariable::WaitHoldingLocks(art::Thread*)+160)
  #00  pc 0000000000463cc4  /system/lib64/libart.so (art::ThreadPool::GetTask(art::Thread*)+252)
  #00  pc 0000000000463180  /system/lib64/libart.so (art::ThreadPoolWorker::Run()+124)
  #00  pc 0000000000462ab0  /system/lib64/libart.so (art::ThreadPoolWorker::Callback(void*)+116)
  #00  pc 0000000000068440  /system/lib64/libc.so (__pthread_start(void*)+196)
  #00  pc 000000000001dd70  /system/lib64/libc.so (__start_thread+16)

"HeapTaskDaemon" daemon prio=5 tid=7 Blocked

  #00  pc 000000000001beec  /system/lib64/libc.so (syscall+28)
  #00  pc 00000000000e8148  /system/lib64/libart.so (art::ConditionVariable::WaitHoldingLocks(art::Thread*)+160)
  #00  pc 0000000000240ee8  /system/lib64/libart.so (art::gc::TaskProcessor::GetTask(art::Thread*)+360)
  #00  pc 00000000002417c4  /system/lib64/libart.so (art::gc::TaskProcessor::RunAllTasks(art::Thread*)+92)
  #00  pc 00000000001e5530  /system/framework/arm64/boot-core-libart.oat (Java_dalvik_system_VMRuntime_runHeapTasks__+124)
  at dalvik.system.VMRuntime.runHeapTasks (Native method)
  at java.lang.Daemons$HeapTaskDaemon.run (Daemons.java:435)
  at java.lang.Thread.run (Thread.java:760)

....Many more thread in WAITING state

环境信息 #

系统信息：
OS: macOS 11.4
CPU: (12) x64 Intel(R) Core(TM) i7-9750H CPU @2.60GHz
内存: 115.29 MB / 16.00 GB
Shell: 5.8 - /bin/zsh
二进制工具：
Node: 14.5.0 - ~/.nvm/versions/node/v14.5.0/bin/node
Yarn: 1.22.10 - /usr/local/bin/yarn
npm: 6.14.5 - ~/.nvm/versions/node/v14.5.0/bin/npm
Watchman: 4.9.0 - /usr/local/bin/watchman
SDK信息：
iOS SDK: 支持平台：DriverKit 21.0.1, iOS 15.0, macOS 12.0, tvOS 15.0, watchO 8.0
IDE信息：
Xcode: 13.1/13A1030d - /usr/bin/xcodebuild
npm依赖包：
react: 16.9.0 => 16.9.0
react-native: 0.61.4 => 0.61.4
npm全局包：
react-native-cli: 2.0.1

复现条件 #

目前无法稳定复现，已知问题仅在Android7.x系统上偶发，触发时机为调用rootView.startReactApplication(...)时。在Application的onCreate生命周期中，通过createReactContextInBackground()初始化了两个不同的React Instance Manager分别承载两个不同业务功能，第二个实例是在本次新版本中新增引入的。

根因分析 #

1. 主线程卡在ModuleHolder.getModule的native wait调用上，是React Native 0.61.x版本多ReactInstanceManager场景下的已知锁竞争问题。两个实例并行初始化时，NativeModule初始化流程会争抢全局锁，当主线程调用startReactApplication需要获取UIManager模块时，锁被另一个初始化线程持有，就会进入无限等待。
2. Android 7.x的ART虚拟机锁调度、堆任务优先级策略和高版本系统有差异，HeapTaskDaemon等系统线程更容易抢占锁资源，放大了锁等待的概率，最终触发ANR超时。

修复方案 #

方案1：调整实例初始化时机 #

不要在Application onCreate中同时初始化两个ReactInstanceManager，采用懒加载策略：

• 保留核心业务对应的实例在Application层预初始化
• 第二个新增实例延迟到对应业务场景启动前再初始化，或者通过IdleHandler在主线程空闲时触发初始化，避免和主流程争抢资源。

方案2：修复React Native版本缺陷 #

你当前使用的0.61.4版本存在多实例锁死的已知缺陷，可二选一处理：

• 升级React Native到0.63.0及以上版本，官方已在该版本修复了ModuleHolder.getModule的死锁问题
• 若暂时无法升级版本，可自行给ModuleHolder类打补丁，移除初始化过程中不必要的同步锁，或者给wait调用添加超时逻辑，避免主线程无限等待。

方案3：Android 7.x专属优化 #

针对Android 7.x系统做适配，禁用堆任务守护线程抢占锁：

if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.N && Build.VERSION.SDK_INT <= Build.VERSION_CODES.N_MR1) {
    try {
        Class<?> vmRuntimeClazz = Class.forName("dalvik.system.VMRuntime");
        Method getRuntimeMethod = vmRuntimeClazz.getMethod("getRuntime");
        Object runtimeInstance = getRuntimeMethod.invoke(null);
        Method setHeapTaskDaemonEnabledMethod = vmRuntimeClazz.getMethod("setHeapTaskDaemonEnabled", boolean.class);
        setHeapTaskDaemonEnabledMethod.invoke(runtimeInstance, false);
    } catch (Exception e) {
        // 异常不影响主流程，可直接忽略
    }
}

将上述代码添加到Application的onCreate方法中即可。

方案4：调用逻辑优化 #

调用startReactApplication前先判断对应ReactInstanceManager的上下文是否初始化完成，若还在初始化流程中，先展示占位加载UI，等初始化成功回调触发后再绑定RootView，避免主线程阻塞。

内容的提问来源于stack exchange，提问作者Rishabh876