一个超经典 WinForm 卡死问题的再反思

一:背景

1.讲故事

这篇文章起源于昨天的一位朋友发给我的dump文件,说它的程序出现了卡死,看了下程序的主线程栈,居然又碰到了 OnUserPreferenceChanged 导致的挂死问题,真的是经典中的经典,线程栈如下:


0:000:x86> !clrstack
OS Thread Id: 0x4eb688 (0)
Child SP       IP Call Site
002fed38 0000002b [HelperMethodFrame_1OBJ: 002fed38] System.Threading.WaitHandle.WaitOneNative(System.Runtime.InteropServices.SafeHandle, UInt32, Boolean, Boolean)
002fee1c 5cddad21 System.Threading.WaitHandle.InternalWaitOne(System.Runtime.InteropServices.SafeHandle, Int64, Boolean, Boolean)
002fee34 5cddace8 System.Threading.WaitHandle.WaitOne(Int32, Boolean)
002fee48 538d876c System.Windows.Forms.Control.WaitForWaitHandle(System.Threading.WaitHandle)
002fee88 53c5214a System.Windows.Forms.Control.MarshaledInvoke(System.Windows.Forms.Control, System.Delegate, System.Object[], Boolean)
002fee8c 538dab4b [InlinedCallFrame: 002fee8c] 
002fef14 538dab4b System.Windows.Forms.Control.Invoke(System.Delegate, System.Object[])
002fef48 53b03bc6 System.Windows.Forms.WindowsFormsSynchronizationContext.Send(System.Threading.SendOrPostCallback, System.Object)
002fef60 5c774708 Microsoft.Win32.SystemEvents+SystemEventInvokeInfo.Invoke(Boolean, System.Object[])
002fef94 5c6616ec Microsoft.Win32.SystemEvents.RaiseEvent(Boolean, System.Object, System.Object[])
002fefe8 5c660cd4 Microsoft.Win32.SystemEvents.OnUserPreferenceChanged(Int32, IntPtr, IntPtr)
002ff008 5c882c98 Microsoft.Win32.SystemEvents.WindowProc(IntPtr, Int32, IntPtr, IntPtr)
...

说实话,这种dump从去年看到今年,应该不下五次了,都看烦了,其形成原因是:

  • 未在主线程中生成用户控件,导致用 WindowsFormsSynchronizationContext.Send 跨线程封送时,对方无法响应请求进而挂死

虽然知道原因,但有一个非常大的遗憾就是在 dump 中找不到到底是哪一个控件,只能笼统的告诉朋友,让其洞察下代码是哪里用了工作线程创建了 用户控件, 有些朋友根据这个信息成功的找到,也有朋友因为各种原因没有找到,比较遗憾。

为了不让这些朋友的遗憾延续下去,这一篇做一个系统归纳,希望能助这些朋友一臂之力。

二:解决方案

1. 背景

这个问题的形成详情,我在去年的一篇文章为:记一次 .NET 某新能源汽车锂电池检测程序 UI挂死分析 //www.cnblogs.com/huangxincheng/p/15245554.html 中已经做过分享,因为 dump 中找不到问题的 Control,所以也留下了一些遗憾,这一篇就做个补充。

2. 问题突破点分析

熟悉 WinForm 底层的朋友应该知道,一旦在 工作线程 上创建了 Control 控件,框架会自动给这个线程配备一个 WindowsFormsSynchronizationContext 和其底层的 MarshalingControl ,这个是有源码支撑的,大家可以找下 Control 的构造函数,简化后的源码如下:


public class Control : Component
{
    internal Control(bool autoInstallSyncContext)
    {
        //***

        if (autoInstallSyncContext)
        {
            WindowsFormsSynchronizationContext.InstallIfNeeded();
        }
    }
}

public sealed class WindowsFormsSynchronizationContext : SynchronizationContext, IDisposable
{
    private Control controlToSendTo;

    private WeakReference destinationThreadRef;

    public WindowsFormsSynchronizationContext()
    {
        DestinationThread = Thread.CurrentThread;
        Application.ThreadContext threadContext = Application.ThreadContext.FromCurrent();
        if (threadContext != null)
        {
            controlToSendTo = threadContext.MarshalingControl;
        }
    }

    internal static void InstallIfNeeded()
    {
        try
        {
            SynchronizationContext synchronizationContext = AsyncOperationManager.SynchronizationContext;
            if (synchronizationContext == null || synchronizationContext.GetType() == typeof(SynchronizationContext))
            {
                AsyncOperationManager.SynchronizationContext = new WindowsFormsSynchronizationContext();
            }
        }
        finally
        {
            inSyncContextInstallation = false;
        }
    }
}

public sealed class WindowsFormsSynchronizationContext : SynchronizationContext, IDisposable
{
    public WindowsFormsSynchronizationContext()
    {
        DestinationThread = Thread.CurrentThread;
        Application.ThreadContext threadContext = Application.ThreadContext.FromCurrent();
        if (threadContext != null)
        {
            controlToSendTo = threadContext.MarshalingControl;
        }
    }
}

internal sealed class ThreadContext
{
    internal Control MarshalingControl
    {
        get
        {
            lock (this)
            {
                if (marshalingControl == null)
                {
                    marshalingControl = new MarshalingControl();
                }
                return marshalingControl;
            }
        }
    }
}

这段代码可以挖到下面两点信息。

  1. 一旦 Control 创建在工作线程上,那这个线程就会安装一个 WindowsFormsSynchronizationContext 变量,比如此时就存在两个对象了。

0:000:x86> !dso
OS Thread Id: 0x4eb688 (0)
ESP/REG  Object   Name
002FEC40 025a0fb0 System.Windows.Forms.WindowsFormsSynchronizationContext
...
002FEF44 0260992c System.Object[]    (System.Object[])
002FEF48 02d69164 System.Windows.Forms.WindowsFormsSynchronizationContext
...

  1. 工作线程ID 会记录在内部的 destinationThreadRef 字段中,我们试探下 02d69164

0:000:x86> !do 02d69164
Name:        System.Windows.Forms.WindowsFormsSynchronizationContext
Fields:
      MT    Field   Offset                 Type VT     Attr    Value Name
...
533c2204  4002522        8 ...ows.Forms.Control  0 instance 02d69218 controlToSendTo
5cef92d0  4002523        c System.WeakReference  0 instance 02d69178 destinationThreadRef

0:000:x86> !DumpObj /d 02d69178
Name:        System.WeakReference
MethodTable: 5cef92d0
EEClass:     5cabf0cc
Size:        12(0xc) bytes
File:        C:\Windows\Microsoft.Net\assembly\GAC_32\mscorlib\v4.0_4.0.0.0__b77a5c561934e089\mscorlib.dll
Fields:
      MT    Field   Offset                 Type VT     Attr    Value Name
5cee2bdc  400070a        4        System.IntPtr  1 instance   111828 m_handle

0:000:x86> !do poi(111828)
Name:        System.Threading.Thread
Fields:
      MT    Field   Offset                 Type VT     Attr    Value Name
5cee1638  40018ca       28         System.Int32  1 instance        9 m_ManagedThreadId 
...

从上面的输出中可以看到,9号线程 曾经创建了不该创建的 Control,所以找出这个 Control 就是解决问题的关键,这也是最难的。

3. 如何找到问题 Control

以我目前的技术实力,从 dump 中确实找不到,但我可以运行时监测,突破点就是一旦这个 Control 在工作线程中创建,底层会安排一个 WindowsFormsSynchronizationContext 以及 MarshalingControl 对象,我们拦截他们的生成构造就好了。

为了方便讲述,先上一段测试代码,在 backgroundWorker1_DoWork 方法中创建一个 Button 控件。


namespace WindowsFormsApp1
{
    public partial class Form1 : Form
    {
        public Form1()
        {
            InitializeComponent();
        }

        private void backgroundWorker1_DoWork(object sender, DoWorkEventArgs e)
        {
            Button btn = new Button();
        }

        private void Form1_Load(object sender, EventArgs e)
        {

        }

        private void button1_Click(object sender, EventArgs e)
        {
            backgroundWorker1.RunWorkerAsync();
        }
    }
}

接下来在 MarshalingControl 的构造函数上下一个bp断点来自动化记录,观察 new Button 的时候是否命中。


0:007> !name2ee System_Windows_Forms_ni System.Windows.Forms.Application+MarshalingControl..ctor
Module:      5b9b1000
Assembly:    System.Windows.Forms.dll
Token:       0600554a
MethodDesc:  5b9fe594
Name:        System.Windows.Forms.Application+MarshalingControl..ctor()
JITTED Code Address: 5bb5d1a4
0:007> bp 5bb5d1a4 "!clrstack; gc"
0:007> g
OS Thread Id: 0x249c (9)
Child SP       IP Call Site
067ff2f0 5bb5d1a4 System.Windows.Forms.Application+MarshalingControl..ctor()
067ff2f4 5bb70224 System.Windows.Forms.Application+ThreadContext.get_MarshalingControl()
067ff324 5bb6fe5d System.Windows.Forms.WindowsFormsSynchronizationContext..ctor()
067ff338 5bb6fd4d System.Windows.Forms.WindowsFormsSynchronizationContext.InstallIfNeeded()
067ff364 5bb6e9a0 System.Windows.Forms.Control..ctor(Boolean)
067ff41c 5bbcd5cc System.Windows.Forms.ButtonBase..ctor()
067ff428 5bbcd531 System.Windows.Forms.Button..ctor()
067ff434 02342500 WindowsFormsApp1.Form1.backgroundWorker1_DoWork(System.Object, System.ComponentModel.DoWorkEventArgs)
067ff488 630ee649 System.ComponentModel.BackgroundWorker.OnDoWork(System.ComponentModel.DoWorkEventArgs) [f:\dd\NDP\fx\src\compmod\system\componentmodel\BackgroundWorker.cs @ 107]
067ff49c 630ee55d System.ComponentModel.BackgroundWorker.WorkerThreadStart(System.Object) [f:\dd\NDP\fx\src\compmod\system\componentmodel\BackgroundWorker.cs @ 245]
067ff6a0 7c69f036 [HelperMethodFrame_PROTECTOBJ: 067ff6a0] System.Runtime.Remoting.Messaging.StackBuilderSink._PrivateProcessMessage(IntPtr, System.Object[], System.Object, System.Object[] ByRef)
067ff95c 6197c82c System.Runtime.Remoting.Messaging.StackBuilderSink.AsyncProcessMessage(System.Runtime.Remoting.Messaging.IMessage, System.Runtime.Remoting.Messaging.IMessageSink)
067ff9b0 61978274 System.Runtime.Remoting.Proxies.AgileAsyncWorkerItem.DoAsyncCall() [f:\dd\ndp\clr\src\BCL\system\runtime\remoting\remotingproxy.cs @ 760]
067ff9bc 61978238 System.Runtime.Remoting.Proxies.AgileAsyncWorkerItem.ThreadPoolCallBack(System.Object) [f:\dd\ndp\clr\src\BCL\system\runtime\remoting\remotingproxy.cs @ 753]
067ff9c0 6104e7b4 System.Threading.QueueUserWorkItemCallback.WaitCallback_Context(System.Object) [f:\dd\ndp\clr\src\BCL\system\threading\threadpool.cs @ 1274]
067ff9c8 61078604 System.Threading.ExecutionContext.RunInternal(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object, Boolean) [f:\dd\ndp\clr\src\BCL\system\threading\executioncontext.cs @ 980]
067ffa34 61078537 System.Threading.ExecutionContext.Run(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object, Boolean) [f:\dd\ndp\clr\src\BCL\system\threading\executioncontext.cs @ 928]
067ffa48 6104f445 System.Threading.QueueUserWorkItemCallback.System.Threading.IThreadPoolWorkItem.ExecuteWorkItem() [f:\dd\ndp\clr\src\BCL\system\threading\threadpool.cs @ 1252]
067ffa5c 6104eb7d System.Threading.ThreadPoolWorkQueue.Dispatch() [f:\dd\ndp\clr\src\BCL\system\threading\threadpool.cs @ 820]
067ffaac 6104e9db System.Threading._ThreadPoolWaitCallback.PerformWaitCallback() [f:\dd\ndp\clr\src\BCL\system\threading\threadpool.cs @ 1161]
067ffccc 7c69f036 [DebuggerU2MCatchHandlerFrame: 067ffccc] 

从线程栈可以清晰的追踪到原来是 backgroundWorker1_DoWork 下的 Button 创建的,这就是问题的根源。。。

三:总结

在我一百多dump的分析旅程中,这个问题真的太高频了,补充此篇真心希望能帮助这些朋友在焦虑中找到问题Control, 一毫之善,与人方便。

Tags: