其他分享
首页 > 其他分享> > [原]调试实战——使用windbg调试TerminateThread导致的死锁

[原]调试实战——使用windbg调试TerminateThread导致的死锁

作者:互联网

原调试debugwindbg死锁deadlock

前言

项目里的一个升级程序偶尔会死锁,查看dump后发现是死在了ShellExecuteExW里。经验少,不知道为什么,于是在高端调试论坛里发帖求助,链接如下http://advdbg.org/forums/6520/ShowPost.aspx

根据张银奎老师的描述可知,应该是拥有关键段的线程意外结束了。仔细检查项目中的代码,发现程序中有使用TerminateThread()来强制杀线程的代码。很可疑,于是写了一个测试程序,还原了这个问题。

{% note info %}

这也是几年前在项目中遇到的一个问题,我对之前的笔记进行了整理重新发布于此。

{% endnote %}

问题重现

重现方法

主程序会加载一个DLL,并调用该DLL的导出函数创建一个线程,然后调用TerminateThread()强制杀死这个线程,然后调用RunProcess()(内部封装了对ShellExecuteEx()的调用)执行一个新进程,会卡死在ShellExecuteEx()。为了让问题更容易重现,特地在DllMain()的参数ul_reason_for_callDLL_THREAD_DETACH时,强制睡眠了5秒。

代码摘录

主工程 testTerminateThread

//testTerminateThread.cpp
#include "stdafx.h"
#include "windows.h"
#include "process.h"

typedef HANDLE (*pfnGenerateThread)();

HANDLE RunProcess(const TCHAR* app_name, const TCHAR* cmd)
{
    SHELLEXECUTEINFO shex = {sizeof(SHELLEXECUTEINFO)};
    shex.fMask = SEE_MASK_NOCLOSEPROCESS;
    shex.lpVerb = _T("open");
    shex.lpFile = app_name; 
    shex.lpParameters = cmd; 
    shex.lpDirectory = NULL; 
    shex.nShow = SW_NORMAL;

    if (!::ShellExecuteEx(&shex))
    {
        return INVALID_HANDLE_VALUE;
    }

    return shex.hProcess;
}

int _tmain(int argc, _TCHAR* argv[])
{
    while ( 1 )
    {
        HMODULE hModule = LoadLibrary(_T("testDll.dll"));
        if ( NULL == hModule )
            return 0;

        pfnGenerateThread pfn = (pfnGenerateThread)GetProcAddress(hModule, "GenerateThread");
        if ( NULL == pfn )
            return 0;

        HANDLE hThread = pfn();

        // give thread time to start up
        Sleep(1000);
        
        // terminate thread.
        BOOL bOk = TerminateThread(hThread, 0);

        // dead lock in this function...
        RunProcess(argv[0], NULL);

        FreeLibrary(hModule);
    }

    return 0;
}

DLL工程 testDll

// DllMain.cpp
#include "stdafx.h"
#include "windows.h"

BOOL APIENTRY DllMain( HMODULE hModule,
                       DWORD  ul_reason_for_call,
                       LPVOID lpReserved
                     )
{

    switch (ul_reason_for_call)
    {
    case DLL_PROCESS_ATTACH:
        OutputDebugString(L"====> DLL_PROCESS_ATTACH called.\n");
        break;
    case DLL_THREAD_ATTACH:
        OutputDebugString(L"----> DLL_THREAD_ATTACH called.\n");
        break;
    case DLL_THREAD_DETACH:
        OutputDebugString(L"<---- DLL_THREAD_DETACH called.\n");
        // with LdrpLoaderLock held! sleep 5 seconds. 
        Sleep(5000);
        break;
    case DLL_PROCESS_DETACH:
        OutputDebugString(L"<==== DLL_PROCESS_DETACH called.\n");
        break;
    }
    return TRUE;
}
// testDll.cpp
#include "stdafx.h"
#include "stdio.h"
#include "process.h"
#include "windows.h"

void OutputCurrentThreadId()
{
    TCHAR szBuffer[1024];
    swprintf_s(szBuffer, L"thread [0x%x], running & exiting...\n", GetCurrentThreadId());
    OutputDebugString(szBuffer);
    return;
}

unsigned __stdcall testProc(void *)
{
    OutputCurrentThreadId();
    return 0;
}

HANDLE GenerateThread()
{
    HANDLE hThread = (HANDLE)_beginthreadex(NULL, 0, &testProc, NULL, 0, NULL);
    return hThread;
}

debug-deadlock-caused-by-TerminateThread-Demo.zip 10.63 KB

问题分析

运行测试程序前先打开DbgView监视调试信息,然后运行测试程序。

DebugView
DebugView

从日志可知,我们启动的测试线程的线程id0x1400

当程序hang住后,使用windbg附加。附加成功后,先运行~*kvn查看线程及每个线程的的调用栈信息。发现只有一个0号线程(1号线程是windbg附加到进程时产生的)。

0:001> ~*kvn

   0  Id: 18c0.1008 Suspend: 1 Teb: 7ffdf000 Unfrozen
 # ChildEBP RetAddr  Args to Child              
00 002bf614 775a6a64 77592278 00000064 00000000 ntdll!KiFastSystemCallRet (FPO: [0,0,0])
01 002bf618 77592278 00000064 00000000 00000000 ntdll!NtWaitForSingleObject+0xc (FPO: [3,0,0])
02 002bf67c 7759215c 00000000 00000000 00000001 ntdll!RtlpWaitOnCriticalSection+0x13e (FPO: [Non-Fpo])
03 002bf6a4 775c00e1 77637340 77bf1b77 00000000 ntdll!RtlEnterCriticalSection+0x150 (FPO: [Non-Fpo])
04 002bf6dc 75587bc3 00000001 00000000 002bf704 ntdll!LdrLockLoaderLock+0xe4 (FPO: [Non-Fpo])
05 002bf728 7679215d 00000000 002bf73c 00000104 KERNELBASE!GetModuleFileNameW+0x75 (FPO: [Non-Fpo])
06 002bf948 76792112 002bfbb0 002bf968 7ffdb000 SHELL32!InRunDllProcess+0x39 (FPO: [Non-Fpo])
*** WARNING: Unable to verify checksum for C:\Users\BianChengNan\Documents\Visual Studio 2012\Projects\testTerminateThread\Debug\testTerminateThread.exe
07 002bf95c 013714db 002bfa44 002bfcbc 002bfbc0 SHELL32!ShellExecuteExW+0x51 (FPO: [Non-Fpo])
08 002bfbb0 01371685 000ac518 00000000 00000000 testTerminateThread!RunProcess+0xdb (FPO: [Non-Fpo]) (CONV: cdecl) [c:\users\bianchengnan\documents\visual studio 2012\projects\testterminatethread\testterminatethread\testterminatethread.cpp @ 28]
09 002bfcbc 01371c69 00000001 000ac510 000ae660 testTerminateThread!wmain+0xc5 (FPO: [Non-Fpo]) (CONV: cdecl) [c:\users\bianchengnan\documents\visual studio 2012\projects\testterminatethread\testterminatethread\testterminatethread.cpp @ 59]
0a 002bfd0c 01371e5d 002bfd20 758ced6c 7ffdb000 testTerminateThread!__tmainCRTStartup+0x199 (FPO: [Non-Fpo]) (CONV: cdecl) [f:\dd\vctools\crt_bld\self_x86\crt\src\crtexe.c @ 533]
0b 002bfd14 758ced6c 7ffdb000 002bfd60 775c37eb testTerminateThread!wmainCRTStartup+0xd (FPO: [Non-Fpo]) (CONV: cdecl) [f:\dd\vctools\crt_bld\self_x86\crt\src\crtexe.c @ 377]
0c 002bfd20 775c37eb 7ffdb000 77bf10cb 00000000 kernel32!BaseThreadInitThunk+0xe (FPO: [Non-Fpo])
0d 002bfd60 775c37be 01371082 7ffdb000 00000000 ntdll!__RtlUserThreadStart+0x70 (FPO: [Non-Fpo])
0e 002bfd78 00000000 01371082 7ffdb000 00000000 ntdll!_RtlUserThreadStart+0x1b (FPO: [Non-Fpo])

#  1  Id: 18c0.193c Suspend: 1 Teb: 7ffde000 Unfrozen
 # ChildEBP RetAddr  Args to Child              
00 0133fbac 775ff20f 76a71677 00000000 00000000 ntdll!DbgBreakPoint (FPO: [0,0,0])
01 0133fbdc 758ced6c 00000000 0133fc28 775c37eb ntdll!DbgUiRemoteBreakin+0x3c (FPO: [Non-Fpo])
02 0133fbe8 775c37eb 00000000 76a71183 00000000 kernel32!BaseThreadInitThunk+0xe (FPO: [Non-Fpo])
03 0133fc28 775c37be 775ff1d3 00000000 00000000 ntdll!__RtlUserThreadStart+0x70 (FPO: [Non-Fpo])
04 0133fc40 00000000 775ff1d3 00000000 00000000 ntdll!_RtlUserThreadStart+0x1b (FPO: [Non-Fpo])

通过调用栈,我们发现程序卡在了ShellExecuteExW里。

运行!cs -l看下输出结果:

0:001> !cs -l
-----------------------------------------
DebugInfo          = 0x77637540
Critical section   = 0x77637340 (ntdll!LdrpLoaderLock+0x0)
LOCKED
LockCount          = 0x1
WaiterWoken        = No
OwningThread       = 0x00001400
RecursionCount     = 0x1
LockSemaphore      = 0x64
SpinCount          = 0x00000000

注意OwningThread的值0x00001400 正是我们生成的测试线程,与我们在DbgView里看到的线程id一致。但是该线程已经被我们杀死了,它在被杀死前获得了进程加载锁0x77637340 (ntdll!LdrpLoaderLock+0x0)

至此,真相大白。

总结

参考资料

标签:Non,windbg,00000000,FPO,ntdll,死锁,线程,Fpo,调试
来源: https://www.cnblogs.com/bianchengnan/p/12158723.html