使用Windbg在内存转储中查找调用的目标线程

本文关键字:调用 目标 线程 查找 Windbg 内存 转储 使用 | 更新日期: 2023-09-27 17:53:39

背景

一位客户报告C#应用程序挂起。在应用程序挂起时,我有一个内存转储。内存转储显示显示进度表单的主UI线程和正在运行的多个后台线程。其中一个后台线程正在尝试控制。调用回主线程以更新表单。CLR堆栈如下所示:

System.Threading.WaitHandle.WaitOneNative(Microsoft.Win32.SafeHandles.SafeWaitHandle, UInt32, Boolean, Boolean) 
System.Threading.WaitHandle.WaitOne(Int64, Boolean) 
System.Threading.WaitHandle.WaitOne(Int32, Boolean) 
System.Windows.Forms.Control.WaitForWaitHandle(System.Threading.WaitHandle) 
System.Windows.Forms.Control.MarshaledInvoke(System.Windows.Forms.Control, System.Delegate, System.Object[], Boolean) 
System.Windows.Forms.Control.Invoke(System.Delegate, System.Object[]) 
profdata.com.Library.frmAsyncExec.SetMessageText(System.String) 
profdata.com.Library.frmAsyncExec.SetMessage(System.String, System.String) 
profdata.com.Library.frmAsyncExec.DoAsyncProcess(System.Object) 
System.Threading.ThreadHelper.ThreadStart_Context(System.Object) 
System.Threading.ExecutionContext.runTryCode(System.Object) 
System.Runtime.CompilerServices.RuntimeHelpers.ExecuteCodeWithGuaranteedCleanup(TryCode, CleanupCode, System.Object) 
System.Threading.ExecutionContext.RunInternal(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object) 
System.Threading.ExecutionContext.Run(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object) 
System.Threading.ThreadHelper.ThreadStart(System.Object) 

UI线程处于模式循环中,正在等待一条消息:

System.Windows.Forms.UnsafeNativeMethods.WaitMessage() 
System.Windows.Forms.Application+ComponentManager.System.Windows.Forms.UnsafeNativeMethods.IMsoComponentManager.FPushMessageLoop(Int32, Int32, Int32) 
System.Windows.Forms.Application+ThreadContext.RunMessageLoopInner(Int32, System.Windows.Forms.ApplicationContext) 
System.Windows.Forms.Application+ThreadContext.RunMessageLoop(Int32, System.Windows.Forms.ApplicationContext) 
System.Windows.Forms.Application.RunDialog(System.Windows.Forms.Form) 
System.Windows.Forms.Form.ShowDialog(System.Windows.Forms.IWin32Window) 
profdata.com.Library.frmAppletBaseForm.ShowDialog(System.Windows.Forms.IWin32Window) 
profdata.com.Library.frmAsyncExec.ExecuteProcess(System.Windows.Forms.IWin32Window) 

我验证了消息队列中唯一的消息(我可以看到(是一条绘制消息。消息15对应于WM_PAINT:

0:000> !dso
OS Thread Id: 0x126c (0)
ESP/REG  Object   Name
ebx      01a688d0 System.Windows.Forms.Application+ThreadContext
esi      01a6eb9c System.Windows.Forms.Application+ComponentManager+ComponentHashtableEntry
edi      1b0c05e0 System.Collections.Hashtable+HashtableEnumerator
0036e244 1b08430c System.Windows.Forms.NativeMethods+MSG[]
0:000> !da 1b08430c 
Name: System.Windows.Forms.NativeMethods+MSG[]
MethodTable: 67e0592c
EEClass: 67be89b8
Size: 40(0x28) bytes
Array: Rank 1, Number of elements 1, Type VALUETYPE
Element Methodtable: 67e059dc
0:000> !dumpvc 67e059dc 1b084314
Name: System.Windows.Forms.NativeMethods+MSG
MethodTable 67e059dc
EEClass: 67bbd880
Size: 36(0x24) bytes
 (C:'Windows'assembly'GAC_MSIL'System.Windows.Forms'2.0.0.0__b77a5c561934e089'System.Windows.Forms.dll)
Fields:
      MT    Field   Offset                 Type VT     Attr    Value Name
691b35f0  4002ba0        0        System.IntPtr  1 instance    b02ee hwnd
691b2f74  4002ba1        4         System.Int32  1 instance       15 message
691b35f0  4002ba2        8        System.IntPtr  1 instance        0 wParam
691b35f0  4002ba3        c        System.IntPtr  1 instance        0 lParam
691b2f74  4002ba4       10         System.Int32  1 instance 264314085 time
691b2f74  4002ba5       14         System.Int32  1 instance      188 pt_x
691b2f74  4002ba6       18         System.Int32  1 instance      386 pt_y

我知道控制。MarshaledInvoke被实现为已知消息的PostMessage:

private object MarshaledInvoke(Control caller, Delegate method, object[] args, bool synchronous)
{
    ...
    System.Windows.Forms.UnsafeNativeMethods.PostMessage(new HandleRef(this, this.Handle), threadCallbackMessage, IntPtr.Zero, IntPtr.Zero);
    ...
}

MSDN这样评价PostMessage:

在与创建指定窗口并在不等待的情况下返回的线程以便线程处理消息。

我怀疑这是一个典型的后台线程访问UI对象的情况,已知该对象会导致应用程序挂起。我很清楚这是一件坏事。

由于有多个后台线程,我想确定哪个后台线程搞砸了。

问题

是否可以确定对Control调用的目标线程。使用内存转储中的信息调用?

到目前为止的工作

我得到了后台线程的堆栈对象的转储:

0:000> ~20s
0:020> !dso
OS Thread Id: 0x5d8 (20)
ESP/REG  Object   Name
ecx      01a2b2d0 System.Runtime.Remoting.Contexts.Context
0c42e49c 1b083e20 System.Threading.ManualResetEvent
0c42e524 01a2006c System.Collections.Hashtable
0c42e560 1b083e38 Microsoft.Win32.SafeHandles.SafeWaitHandle
0c42e5fc 1b044128 System.Windows.Forms.PropertyStore
0c42e608 1b060b68 System.Collections.Queue
0c42e610 1b083de0 System.Windows.Forms.Control+ThreadMethodEntry
0c42e65c 1b043b30 profdata.com.DailyReporting.frmDRAsyncExec
0c42e66c 1b083cb0 System.Object[]    (System.Object[])
0c42e670 1b083c90 profdata.com.Library.frmAsyncExec+SetMessageDelegate
0c42e674 1b083cc4 System.Windows.Forms.Control+MultithreadSafeCallScope
0c42e694 1b043b30 profdata.com.DailyReporting.frmDRAsyncExec

我试过浏览ThreadMethodEntry和Context对象,但我不确定我在寻找什么:

0:020> !do 1b083de0 
Name: System.Windows.Forms.Control+ThreadMethodEntry
MethodTable: 67e05380
EEClass: 67be8454
Size: 52(0x34) bytes
 (C:'Windows'assembly'GAC_MSIL'System.Windows.Forms'2.0.0.0__b77a5c561934e089'System.Windows.Forms.dll)
Fields:
      MT    Field   Offset                 Type VT     Attr    Value Name
67dff750  40011b9        4 ...ows.Forms.Control  0 instance 1b043b30 caller
67dff750  40011ba        8 ...ows.Forms.Control  0 instance 1b043b30 marshaler
691b118c  40011bb        c      System.Delegate  0 instance 1b083c90 method
691844f8  40011bc       10      System.Object[]  0 instance 1b083cb0 args
691b0944  40011bd       14        System.Object  0 instance 00000000 retVal
691b0ebc  40011be       18     System.Exception  0 instance 00000000 exception
691847f4  40011bf       2c       System.Boolean  1 instance        1 synchronous
691847f4  40011c0       2d       System.Boolean  1 instance        0 isCompleted
69197b54  40011c1       1c ....ManualResetEvent  0 instance 1b083e20 resetEvent
691b0944  40011c2       20        System.Object  0 instance 1b083e14 invokeSyncObject
691ada1c  40011c3       24 ....ExecutionContext  0 instance 1b083cd0 executionContext
691aa00c  40011c4       28 ...ronizationContext  0 instance 00000000 syncContext
0:020> !do 01a2b2d0 
Name: System.Runtime.Remoting.Contexts.Context
MethodTable: 691a210c
EEClass: 68fcf620
Size: 60(0x3c) bytes
 (C:'Windows'assembly'GAC_32'mscorlib'2.0.0.0__b77a5c561934e089'mscorlib.dll)
Fields:
      MT    Field   Offset                 Type VT     Attr    Value Name
691844f8  4001f60        4      System.Object[]  0 instance 01a2b3a0 _ctxProps
691a219c  4001f61        8 ...micPropertyHolder  0 instance 00000000 _dphCtx
691a75b4  4001f62        c ...em.LocalDataStore  0 instance 00000000 _localDataStore
69196f18  4001f63       10 ...ging.IMessageSink  0 instance 00000000 _serverContextChain
69196f18  4001f64       14 ...ging.IMessageSink  0 instance 00000000 _clientContextChain
691b1508  4001f65       18     System.AppDomain  0 instance 01a01298 _appDomain
691844f8  4001f66       1c      System.Object[]  0 instance 00000000 _ctxStatics
691b35f0  4001f67       20        System.IntPtr  1 instance   551520 _internalContext
691b2f74  4001f68       24         System.Int32  1 instance        0 _ctxID
691b2f74  4001f69       28         System.Int32  1 instance        3 _ctxFlags
691b2f74  4001f6a       2c         System.Int32  1 instance        1 _numCtxProps
691b2f74  4001f6b       30         System.Int32  1 instance        0 _ctxStaticsCurrentBucket
691b2f74  4001f6c       34         System.Int32  1 instance        0 _ctxStaticsFreeIndex
691a219c  4001f6d      654 ...micPropertyHolder  0   shared   static _dphGlobal
    >> Domain:Value  00557ff8:01a2b068 <<
69189944  4001f6e      658 ...LocalDataStoreMgr  0   shared   static _localDataStoreMgr
    >> Domain:Value  00557ff8:01a2b07c <<
691b2f74  4001f6f      b40         System.Int32  1   shared   static _ctxIDCounter
    >> Domain:Value  00557ff8:0 <<

或者可能有一种更原生的方法,包括找到窗口句柄与特定线程的关联,但如果是这样,我不知道如何做到这一点。

使用Windbg在内存转储中查找调用的目标线程

您需要在ThreadMethodEntry中转储在UI线程上执行的委托(方法字段(。其他成员只是为了让UI线程知道将回调发送到哪个控件,以及在哪个同步上下文中执行回调。

还有一些标志可以检查是否需要调用,如果调用失败,则有一个异常字段,如果是通过invoke调用完成的,则可以将其整理回调用方。你可以放心地忽略那里的其他东西。

有关如何在手头只有委托实例的情况下找到调用了哪个方法的更多信息,请访问此处:http://geekswithblogs.net/akraus1/archive/2012/05/20/149699.aspx

由于您的回调甚至没有执行,因此您也应该查看其他线程。这种神秘的挂起通常与WM_SETTING_CHANGE窗口消息一起发生,其中非UI线程试图调用您的控件(请参阅http://ikriv.com/dev/dotnet/MysteriousHang.html#BeginInvokeDance)。