1

我正在编写在我们的多用户多线程应用程序服务器中实现某种“进程内分析器”组件的代码。它基本上工作完美 - 显示每个线程的堆栈跟踪以及一些附加信息,例如该线程的 CPU 利用率、SQL 等待时间等,每 X 秒刷新一次。我差点签到,但突然在调试会话中我注意到我的应用程序挂起!它并非一直发生,而是随机发生,暗示了一些并发问题。经过仔细检查,我发现该应用程序在调用“new StackTrace(one_of_my_threads, false)”期间挂起。这是调试器的堆栈:

[Managed to Native Transition]  
mscorlib.dll!System.RuntimeType.RuntimeTypeCache.MemberInfoCache<System.Reflection.RuntimeMethodInfo>.AddMethod(System.RuntimeTypeHandle declaringType = {System.RuntimeTypeHandle}, System.RuntimeMethodHandle method, System.RuntimeType.RuntimeTypeCache.CacheType cacheType) + 0x88 bytes   
mscorlib.dll!System.RuntimeType.RuntimeTypeCache.GetMethod(System.RuntimeTypeHandle declaringType, System.RuntimeMethodHandle method) + 0x2d bytes  
mscorlib.dll!System.RuntimeType.GetMethodBase(System.RuntimeTypeHandle reflectedTypeHandle, System.RuntimeMethodHandle methodHandle) + 0xf5 bytes   
mscorlib.dll!System.Diagnostics.StackFrameHelper.GetMethodBase(int i) + 0x4e bytes  
mscorlib.dll!System.Diagnostics.StackTrace.CaptureStackTrace(int iSkip, bool fNeedFileInfo = false, System.Threading.Thread targetThread, System.Exception e = null) + 0xb8 bytes   

mscorlib.dll!System.Diagnostics.StackTrace.StackTrace(System.Threading.Thread targetThread, bool needFileInfo) + 0x18 bytes 

>Almog.Next.Tools.dll!Almog.Next.Services.ThreadWorkUnit.GetStackTrace(System.Threading.Thread th = {System.Threading.Thread}, Almog.Next.Services.StackDetalization details = Simplified) Line 175 + 0x2c bytes    

Almog.Next.Tools.dll!Almog.Next.Services.ThreadWorkUnit.InternalCalculate(Almog.Next.Services.StackDetalization sdetails = Simplified) Line 131 + 0x1b bytes

Almog.Next.Tools.dll!Almog.Next.Services.ThreadWorkUnit.Calculate(Almog.Next.Services.StackDetalization sdetails = Simplified) Line 90 + 0xc bytes

Almog.Next.Tools.dll!Almog.Next.Services.InProcessProfiler.GetThreadWorks(Almog.Next.Services.StackDetalization stackLevel = Simplified) Line 41 + 0xe bytes

Almog.Next.Tools.dll!Almog.Next.CommonControls.ShowWorkloadForm.timer_Tick(object sender = {Interval = 250}, System.EventArgs e = {System.EventArgs}) Line 40 + 0x9 bytes

System.Windows.Forms.dll!System.Windows.Forms.Timer.OnTick(System.EventArgs e) + 0x17 bytes 

System.Windows.Forms.dll!System.Windows.Forms.Timer.TimerNativeWindow.WndProc(ref System.Windows.Forms.Message m) + 0x34 bytes  

System.Windows.Forms.dll!System.Windows.Forms.NativeWindow.Callback(System.IntPtr hWnd, int msg = 275, System.IntPtr wparam, System.IntPtr lparam) + 0x5a bytes 

[Native to Managed Transition]  

[Managed to Native Transition]  

System.Windows.Forms.dll!System.Windows.Forms.Application.ComponentManager.System.Windows.Forms.UnsafeNativeMethods.IMsoComponentManager.FPushMessageLoop(int dwComponentID, int reason = -1, int pvLoopData = 0) + 0x24e bytes 

System.Windows.Forms.dll!System.Windows.Forms.Application.ThreadContext.RunMessageLoopInner(int reason = -1, System.Windows.Forms.ApplicationContext context = {System.Windows.Forms.ApplicationContext}) + 0x177 bytes 

System.Windows.Forms.dll!System.Windows.Forms.Application.ThreadContext.RunMessageLoop(int reason, System.Windows.Forms.ApplicationContext context) + 0x61 bytes    

System.Windows.Forms.dll!System.Windows.Forms.Application.Run(System.Windows.Forms.Form mainForm) + 0x31 bytes  

NextServer.exe!Almog.Next.Server.Program.Main(string[] ParamStr = {string[0]}) Line 24 + 0x1d bytes

我还注意到其他线程看起来也被阻塞了,主要是在分配新对象时(调用“new”)。线程示例之一,挂在这行代码上:

return new TransactionController(this, transactionMode, isolationLevel);

我相信每个“新”都与 MemberInfoCache 有关,但即使在使用 Reflector 查看 AddMethod() 的代码后,我也无法弄清楚我能做什么。请帮忙!我完全被困住了。

这是读取堆栈跟踪的代码。没有什么不寻常的,恕我直言:...

if (th == Thread.CurrentThread)
                throw new NextException("Internal error: it seems the thread {0} once called BeginWork() is the same thread which calls GetThreadWorks() now...", th.ManagedThreadId);
pragma warning disable 612, 618
            th.Suspend();
            var trace = new StackTrace(th, details == StackDetalization.Detailed);
            th.Resume();
pragma warning restore 612, 618
            var sb = new StringBuilder();
            int frameCount = trace.FrameCount;
            for (int i = 0; i < frameCount; i++)
            {                

...

提前致谢...

4

1 回答 1

0

如果您挂起的线程持有一个锁并且 StackTrace 正在使用该锁,您将死锁。

不推荐使用 Thread.Suspend的原因是:您基本上不能安全地使用它。如果您确切知道挂起的线程正在运行什么代码,包括任何 BCL 内容,您才能使用它。你不能知道。

我不知道这个问题的解决方案。我猜线程需要暂停才能捕获堆栈跟踪?

于 2012-04-26T10:41:54.987 回答