发生ANR之后的事情

触发ANR条件

• Service Timeout:前台服务20s内未完成执行，后台服务超过200s
• BroadcastQueue Timeout：前台广播在10s内未完成,后台广播超过60s未完成
• ContentProvider Timeout：内容提供者,在publish过超时10s
• InputDispatching Timeout: 输入事件分发超时5s，包括按键和触摸事件

对应最后执行ANR记录的方法是：

• ActiveServices.serviceTimeout()
• BroadcastQueue.AppNotResponding.run().appNotResponding()
• AMS.appNotRespondingViaProvider()
• AMS.inputDispatchingTimedOut()

理解为什么会出发ANR，必须明白四大组件的启动流程。这个我们以后分析。

分析ANR触发之后记录问题

可以将anr目录下的文件放到电脑上进行查看
adb pull data/anr .

属性系统可以通过adb shell getprop dalvik.vm.stack-trace-file这种方式查找对应的属性值

当触发ANR之后会调用AppErrors.appNotResponding()方法

final void appNotResponding(ProcessRecord app, ActivityRecord activity,
        ActivityRecord parent, boolean aboveSystem, final String annotation) {
        ArrayList<Integer> firstPids = new ArrayList<Integer>(5);
        SparseArray<Boolean> lastPids = new SparseArray<Boolean>(20);
        ...
        //记录ANR时间
        long anrTime = SystemClock.uptimeMillis();
        //更新CPU状态
        if (ActivityManagerService.MONITOR_CPU_USAGE) {
            mService.updateCpuStatsNow();
        }
        //特定场景下忽略ANR
        synchronized (mService) {
            if (mService.mShuttingDown) {
                Slog.i(TAG, "During shutdown skipping ANR: " + app + " " + annotation);
                return;
            } else if (app.notResponding) {
                Slog.i(TAG, "Skipping duplicate ANR: " + app + " " + annotation);
                return;
            } else if (app.crashing) {
                Slog.i(TAG, "Crashing app skipping ANR: " + app + " " + annotation);
                return;
            }        
        }
        //为了防止多次对相同app的anr执行重复代码，在此处标注记录，属于上面的特定情况种的一种
        app.notResponding = true;
        //记录ANR信息到Event Log中
        EventLog.writeEvent(EventLogTags.AM_ANR, app.userId, app.pid,
                app.processName, app.info.flags, annotation);        
        //添加当前app到firstpids列表中
        firstPids.add(app.pid);
        //如果可能添加父进程到firstpids列表种
        int parentPid = app.pid;
        ...
        // 将ANR信息存在info变量中，后续打印到LOGCAT，这部分的信息会以ActivityManager为Tag打印出来，包含了ANR的进程，出现原因以及当时的CPU状态，这些对分析ANR是非常重要的信息
        StringBuilder info = new StringBuilder();
        info.setLength(0);
        info.append("ANR in ").append(app.processName);
        if (activity != null && activity.shortComponentName != null) {
            info.append(" (").append(activity.shortComponentName).append(")");
        }
        info.append("\n");
        info.append("PID: ").append(app.pid).append("\n");
        if (annotation != null) {
            info.append("Reason: ").append(annotation).append("\n");
        }
        if (parent != null && parent != activity) {
            info.append("Parent: ").append(parent.shortComponentName).append("\n");
        }
        //将ANR信息输出到traces文件，分为两种，一种带native层信息，一种不带
        ProcessCpuTracker processCpuTracker = new ProcessCpuTracker(true;
        String[] nativeProcs = NATIVE_STACKS_OF_INTEREST;
        // don't dump native PIDs for background ANRs
        File tracesFile = null;
        if (isSilentANR) {
            //这里返回了一个文件，这里的文件路径是：`/data/anr/traces.txt`
            //查找方法：adb shell getprop dalvik.vm.stack-trace-file
            tracesFile = mService.dumpStackTraces(true, firstPids, null, lastPids,
                null);
        } else {
            tracesFile = mService.dumpStackTraces(true, firstPids, processCpuTracker, lastPids,
                nativeProcs);
        }

        //再次更新CPU信息，并且输出到SystemLog中
        String cpuInfo = null;
        if (ActivityManagerService.MONITOR_CPU_USAGE) {
            mService.updateCpuStatsNow();
            synchronized (mService.mProcessCpuTracker) {
                cpuInfo = mService.mProcessCpuTracker.printCurrentState(anrTime);
            }
            info.append(processCpuTracker.printCurrentLoad());
            info.append(cpuInfo);
        }

        info.append(processCpuTracker.printCurrentState(anrTime));
        Slog.e(TAG, info.toString());
        //上面的信息已经对应的ANR信息写入/data/anr/traces.txt中

        //给底层发送信号Process.SIGNAL_QUIT=3
        if (tracesFile == null) {
            Process.sendSignal(app.pid, Process.SIGNAL_QUIT);
        }
        //将traces文件 和 CPU使用率信息保存到dropbox，即data/system/dropbox目录
        //命名：system_server/system_app/data_app + type+...比如下面
        //[email protected] 
        //[email protected]
        mService.addErrorToDropBox("anr", app, app.processName, activity, parent, annotation,
                cpuInfo, tracesFile, null);        

        synchronized (mService) {
            mService.mBatteryStatsService.noteProcessAnr(app.processName, app.uid);
            //如果是后台ANR则直接杀掉结束
            if (isSilentANR) {
                app.kill("bg anr", true);
                return;
            }
            //设置app的not响应状态，并查找errorReportReceiver
            makeAppNotRespondingLocked(app,
                    activity != null ? activity.shortComponentName : null,
                    annotation != null ? "ANR " + annotation : "ANR",
                    info.toString());

            //弹出ANR对话框
            Message msg = Message.obtain();
            HashMap<String, Object> map = new HashMap<String, Object>();
            msg.what = ActivityManagerService.SHOW_NOT_RESPONDING_UI_MSG;
            msg.obj = map;
            msg.arg1 = aboveSystem ? 1 : 0;
            map.put("app", app);
            if (activity != null) {
                map.put("activity", activity);
            }
            //向ui线程发送，内容为SHOW_NOT_RESPONDING_MSG的消息
            mService.mUiHandler.sendMessage(msg);
        }
}

我们来小节一下上面发生了什么：

• 立刻更新了CPU的信息
  java
/** 2721 cpu (total|1|6),(user|1|6),(system|1|6),(iowait|1|6),(irq|1|6),(softirq|1|6) */
public static final int CPU = 2721;
给event_log中写入值

• 忽略一些anr
• 在event_log中打印am_anr的信息，这个是anr立刻发生的记录
• 将ANR信息存在info变量中，后续打印到LOGCAT，这部分的信息会以ActivityManager为Tag打印出来，包含了ANR的进程，出现原因以及当时的CPU状态，这些对分析ANR是非常重要的信息
• 将ANR信息输出到data/anr/traces文件
• 没有输出到traces文件的时候，给底层发送一个rocess.SIGNAL_QUIT=3信号
• 将traces文件 和 CPU使用率信息保存到dropbox，即data/system/dropbox目录
• 如果是后台ANR则直接杀掉结束
• 弹出ANR对话框

细节

怎么样就将信息保存到了/data/anr/traces.txt了

1.AMS.dumpStackTraces

public static File dumpStackTraces(boolean clearTraces, ArrayList<Integer> firstPids,
        ProcessCpuTracker processCpuTracker, SparseArray<Boolean> lastPids, String[] nativeProcs) {
    //tracesPath = "data/anr/traces.txt"
    String tracesPath = SystemProperties.get("dalvik.vm.stack-trace-file", null);
    if (tracesPath == null || tracesPath.length() == 0) {
        return null;
    }

    File tracesFile = new File(tracesPath);
    try {
        if (clearTraces && tracesFile.exists()) tracesFile.delete();
        tracesFile.createNewFile();
        FileUtils.setPermissions(tracesFile.getPath(), 0666, -1, -1); // -rw-rw-rw-
    } catch (IOException e) {
        Slog.w(TAG, "Unable to prepare ANR traces file: " + tracesPath, e);
        return null;
    }
    //[2]
    dumpStackTraces(tracesPath, firstPids, processCpuTracker, lastPids, nativeProcs);
    return tracesFile;
}

2.

private static void dumpStackTraces(String tracesPath, ArrayList<Integer> firstPids,
        ProcessCpuTracker processCpuTracker, SparseArray<Boolean> lastPids, String[] nativeProcs){
    FileObserver observer = new FileObserver(tracesPath, FileObserver.CLOSE_WRITE) {
        @Override
        public synchronized void onEvent(int event, String path) { notify(); }
    };

    try {
        observer.startWatching();

        // 获取发生ANR进程的pid,然后遍历这些进程给进程发送Process.SIGNAL_QUIT=3的信号
        if (firstPids != null) {
            try {
                int num = firstPids.size();
                for (int i = 0; i < num; i++) {
                    synchronized (observer) {
                        final long sime = SystemClock.elapsedRealtime();
                        Process.sendSignal(firstPids.get(i), Process.SIGNAL_QUIT);
                        observer.wait(1000);  // Wait for write-close, give up after 1 sec
                    }
                }
            } catch (InterruptedException e) {
                Slog.wtf(TAG, e);
            }
        }

        // 接下来收集本地pids的堆栈
        if (nativeProcs != null) {
            int[] pids = Process.getPidsForCommands(nativeProcs);
            if (pids != null) {
                for (int pid : pids) {
                    final long sime = SystemClock.elapsedRealtime();
                    Debug.dumpNativeBacktraceToFileTimeout(pid, tracesPath, 10);//[3]输出native进程的trace并且限制超时时间
                }
            }
        }
        if (processCpuTracker != null) {
            processCpuTracker.init();
            System.gc();
            processCpuTracker.update();
            try {
                synchronized (processCpuTracker) {
                    processCpuTracker.wait(500); // measure over 1/2 second.
                }
            } catch (InterruptedException e) {
            }
            processCpuTracker.update();

            //从lastPids中选取CPU使用率 top 5的进程，输出这些进程的stacks
            final int N = processCpuTracker.countWorkingStats();
            int numProcs = 0;
            for (int i=0; i<N && numProcs<5; i++) {
                ProcessCpuTracker.Stats stats = processCpuTracker.getWorkingStats(i);
                if (lastPids.indexOfKey(stats.pid) >= 0) {
                    numProcs++;
                    try {
                        synchronized (observer) {
                            final long stime = SystemClock.elapsedRealtime();
                            Process.sendSignal(stats.pid, Process.SIGNAL_QUIT);
                            observer.wait(1000);  // Wait for write-close, give up after 1 sec

                        }
                    } catch (InterruptedException e) {
                        Slog.wtf(TAG, e);
                    }
                } else if (DEBUG_ANR) {
                    Slog.d(TAG, "Skipping next CPU consuming process, not a java proc: "
                            + stats.pid);
                }
            }
        }
    } finally {
        observer.stopWatching();
    }
}

小结：

• 收集发生anr进程的调用栈

  • 发生anr的进程
  • anr进程的父进程（anr进程是由于AMS生成，AMS在system_server进程中，system_server进程是anr的父进程）
  • mLruProcesses中所有的persistent进程

• 收集Native进程的调用栈

  • "/system/bin/audioserver"
  • "/system/bin/cameraserver"
  • "/system/bin/drmserver"
  • "/system/bin/mediadrmserver"
  • "/system/bin/mediaserver"
  • "/system/bin/sdcard"
  • "/system/bin/surfaceflinger"
  • "media.codec" // system/bin/mediacodec
  • "media.extractor" // system/bin/mediaextractor
  • "com.android.bluetooth" // Bluetooth service
• 收集lastPids进程的stacks
  
  • 收集前五名

注意收集信息等待的时间

3.Debug.dumpNativeBacktraceToFileTimeout()

static void android_os_Debug_dumpNativeBacktraceToFileTimeout(JNIEnv* env, jobject clazz,
    jint pid, jstring fileName, jint timeoutSecs)
{
    if (fileName == NULL) {
        jniThrowNullPointerException(env, "file == null");
        return;
    }
    const jchar* str = env->GetStringCritical(fileName, 0);
    String8 fileName8;
    if (str) {
        fileName8 = String8(reinterpret_cast<const char16_t*>(str),
                            env->GetStringLength(fileName));
        env->ReleaseStringCritical(fileName, str);
    }
    //打开文件(data/anr/traces.txt)
    int fd = open(fileName8.string(), O_CREAT | O_WRONLY | O_NOFOLLOW | O_CLOEXEC | O_APPEND, 0666);
    if (fd < 0) {
        fprintf(stderr, "Can't open %s: %s\n", fileName8.string(), strerror(errno));
        return;
    }

    dump_backtrace_to_file_timeout(pid, fd, timeoutSecs);//[4]

    close(fd);
}

4.dump_backtrace_to_file_timeout()

int dump_backtrace_to_file_timeout(pid_t tid, int fd, int timeout_secs) {
  //发送dump请求得到sock_fd
  int sock_fd = make_dump_request(DEBUGGER_ACTION_DUMP_BACKTRACE, tid, timeout_secs);
  if (sock_fd < 0) {
    return -1;
  }
  int result = 0;
  char buffer[1024];
  ssize_t n;
  int flag = 0;
  //从sock_fd中读取信息写入data/anr/traces.txt中
  while ((n = TEMP_FAILURE_RETRY(read(sock_fd, buffer, sizeof(buffer)))) > 0) {
    flag = 1;
    if (TEMP_FAILURE_RETRY(write(fd, buffer, n)) != n) {
      result = -1;
      break;
    }
  }
  close(sock_fd);
  ...
  return result;
}

主要是通过给底层发送DEBUGGER_ACTION_DUMP_BACKTRACE来请求dump的sock_fd句柄，底层调用dump_backtraces()来获取信息，从而写入data/anr/traces.txt文件中

总结

当发生anr的时候，距离ANR最近的时间是am_anr这个日志的时间，然后会打印各种信息有底层dump的，有进程的调用栈信息等等。最后将trances.txt写入data/system/dropbox目录下，并且重命名，规则见上文。

补充

其中Process.sendSignal(stats.pid, Process.SIGNAL_QUIT);发出退出进程信号