Android进程启动

Android进程启动可以参考gityuan的相关介绍，本文记录下遇到的问题，修改方法，以及引申出的其他问题

一、遇到的问题

Android O上，用户正常操作过程中，很低概率会遇到system_server卡死，看traces.txt中system_server主线程卡死状态

"main" prio=5 tid=1 Blocked  | group="main" sCount=1 dsCount=0 flags=1 obj=0x72b927d0 self=0x7eed0bea00  | sysTid=2196 nice=-2 cgrp=default sched=0/0 handle=0x7ef17a99b0  | state=S schedstat=( 0 0 0 ) utm=292 stm=117 core=1 HZ=100  | stack=0x7ff187a000-0x7ff187c000 stackSize=8MB  | held mutexes=  at com.android.server.am.ActivityManagerService.checkContentProviderAccess(ActivityManagerService.java:11204)  - waiting to lock <0x07934e4e> (a com.android.server.am.ActivityManagerService) held by thread 103  at com.android.server.am.ActivityManagerService$LocalService.checkContentProviderAccess(ActivityManagerService.java:23909)  at com.android.server.content.ContentService.notifyChange(ContentService.java:368)  at android.content.ContentResolver.notifyChange(ContentResolver.java:2046)  at com.android.providers.settings.SettingsProvider$SettingsRegistry$MyHandler.handleMessage(SettingsProvider.java:2877)  at android.os.Handler.dispatchMessage(Handler.java:105)  at android.os.Looper.loop(Looper.java:164)  at com.android.server.SystemServer.run(SystemServer.java:425)  at com.android.server.SystemServer.main(SystemServer.java:264)  at java.lang.reflect.Method.invoke(Native method)  at com.android.internal.os.Zygote$MethodAndArgsCaller.run(Zygote.java:240)  at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:820)……"Binder:2196_1A" prio=5 tid=103 Native  | group="main" sCount=1 dsCount=0 flags=1 obj=0x13f0b4c8 self=0x7ed3a92c00  | sysTid=11637 nice=-2 cgrp=default sched=1073741824/0 handle=0x7eca5ff4f0  | state=S schedstat=( 0 0 0 ) utm=622 stm=333 core=1 HZ=100  | stack=0x7eca505000-0x7eca507000 stackSize=1005KB  | held mutexes=  kernel: (couldn't read /proc/self/task/11637/stack)  native: #00 pc 0000000000069184  /system/lib64/libc.so (recvmsg+4)  native: #01 pc 000000000011844c  /system/lib64/libandroid_runtime.so (???)  native: #02 pc 00000000001180b4  /system/lib64/libandroid_runtime.so (???)  native: #03 pc 00000000005c74a4  /system/framework/arm64/boot-framework.oat (Java_android_net_LocalSocketImpl_readba_1native___3BIILjava_io_FileDescriptor_2+196)  at android.net.LocalSocketImpl.readba_native(Native method)  at android.net.LocalSocketImpl.-wrap1(LocalSocketImpl.java:-1)  at android.net.LocalSocketImpl$SocketInputStream.read(LocalSocketImpl.java:110)  - locked <0x04da015a> (a java.lang.Object)  at java.io.DataInputStream.readFully(DataInputStream.java:198)  at java.io.DataInputStream.readInt(DataInputStream.java:389)  at android.os.ZygoteProcess.zygoteSendArgsAndGetResult(ZygoteProcess.java:297)  at android.os.ZygoteProcess.startViaZygote(ZygoteProcess.java:431)  - locked <0x02d82f8b> (a java.lang.Object)  at android.os.ZygoteProcess.start(ZygoteProcess.java:207)  at android.os.Process.startWebView(Process.java:470)  at com.android.server.am.ActivityManagerService.startProcessLocked(ActivityManagerService.java:3929)  at com.android.server.am.ActivityManagerService.startProcessLocked(ActivityManagerService.java:3760)  at com.android.server.am.ActivityManagerService.startProcessLocked(ActivityManagerService.java:3649)  at com.android.server.am.ActiveServices.bringUpServiceLocked(ActiveServices.java:2161)  at com.android.server.am.ActiveServices.bindServiceLocked(ActiveServices.java:1439)  at com.android.server.am.ActivityManagerService.bindService(ActivityManagerService.java:18465)  - locked <0x07934e4e> (a com.android.server.am.ActivityManagerService)  at android.app.IActivityManager$Stub.onTransact(IActivityManager.java:596)  at com.android.server.am.ActivityManagerService.onTransact(ActivityManagerService.java:2944)

通过线程状态看，system_server卡在起动进程的过程中，在等待zygote通过socket写进程pid，告诉system_server启动应用的进程号

二、问题分析

一图胜千言万语，图片来自gityuan

Android进程启动_第1张图片

socket有问题，或者zygote有问题？

三、措施

看源码中也有添加超时的建议，但是目前没添加，而且只有Android O存在问题，Android N或者Android P都没有遇到

那就自己尝试添加，思路是在从zygote读pid的时候，添加循环读，一次进程启动的需要读取数据长度为5
因而判断条件为：当前系统uptime < 最大超时时间 && 读取长度 < 5

最大超时时间 = 第一次进来的系统uptime + 超时

注意：

必须要用系统uptime，不能简单使用currentTimeMillis，否则时间出现跳变会导致报下面异常

09-06 10:17:09.476  1000  2255  2509 E ZygoteProcess: Starting VM process through Zygote failed09-06 10:17:09.476  1000  2255  2509 E ActivityManager: Failure starting process com.****.******.****09-06 10:17:09.476  1000  2255  2509 E ActivityManager: java.lang.RuntimeException: Starting VM process through Zygote failed09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at android.os.ZygoteProcess.start(ZygoteProcess.java:214)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at android.os.Process.start(Process.java:453)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at com.android.server.am.ActivityManagerService.startProcessLocked(ActivityManagerService.java:3951)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at com.android.server.am.ActivityManagerService.startProcessLocked(ActivityManagerService.java:3777)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at com.android.server.am.ActivityManagerService.startProcessLocked(ActivityManagerService.java:3666)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at com.android.server.am.ActiveServices.bringUpServiceLocked(ActiveServices.java:2161)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at com.android.server.am.ActiveServices.bindServiceLocked(ActiveServices.java:1439)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at com.android.server.am.ActivityManagerService.bindService(ActivityManagerService.java:18520)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at android.app.IActivityManager$Stub.onTransact(IActivityManager.java:596)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at com.android.server.am.ActivityManagerService.onTransact(ActivityManagerService.java:2961)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at android.os.Binder.execTransact(Binder.java:674)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: Caused by: android.os.ZygoteStartFailedEx: fork() failed09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at android.os.ZygoteProcess.zygoteSendArgsAndGetResult(ZygoteProcess.java:319)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at android.os.ZygoteProcess.startViaZygote(ZygoteProcess.java:469)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: at android.os.ZygoteProcess.start(ZygoteProcess.java:208)09-06 10:17:09.476  1000  2255  2509 E ActivityManager: ... 10 more09-06 10:17:09.649  1000  2255  2846 W ActivityManager: No pending application record for pid 2927 (IApplicationThread [email protected]); dropping process

此后一直无法启动成功，重复出现No pending application record for pid错误，因为由于时间跳变导致认为超时抛出上面异常了，但是其实应用还在正常启动中，会回调attachApplication方法，但此时拿到的pid是下一个正在启动进程pid，导致在mPidSelfLocked中找不到，返回null

四、引申问题

上面问题是时间跳变模拟到的，并不是真正出问题的时候，真正出问题的情况是如何的呢？

或者上面理解有什么不对的么？

参考：

gityuan

fork机制

更多相关文章

随机推荐