记一次 .NET 某医疗住院系统 崩溃分析

一：背景

1. 讲故事

最近收到了两起程序崩溃的dump，查了下都是经典的 double free 造成的，蛮有意思，这里就抽一篇出来分享一下经验供后面的学习者避坑吧。

二：WinDbg 分析

1. 崩溃点在哪里

windbg 带了一个自动化分析命令 !analyze -v 可以帮助我们找到崩溃时的程序指令地址以及崩溃的代码，这对我们分析问题非常有帮助。

 0:090> !analyze -v ******************************************************************************* *                                                                             * *                        Exception Analysis                                   * *                                                                             * ******************************************************************************* CONTEXT:  (.ecxr) rax=00007ffec265d6e0 rbx=00000000c0000374 rcx=00000053653fe4f0 rdx=00007ffec1d3e9a0 rsi=0000000000000001 rdi=00007ffed7b827f0 rip=00007ffed7b1b349 rsp=00000053653fed10 rbp=000001c14fd20000  r8=000001c11957d9a0  r9=0000000000000033 r10=000001c453dbc7f0 r11=00007ffeb94db004 r12=0000000000000001 r13=000001c12e8526d0 r14=0000000000000000 r15=000001ce25531c60 iopl=0         nv up ei pl nz na po nc cs=0033  ss=002b  ds=002b  es=002b  fs=0053  gs=002b             efl=00000206 ntdll!RtlReportFatalFailure+0x9: 00007ffe`d7b1b349 eb00            jmp     ntdll!RtlReportFatalFailure+0xb (00007ffe`d7b1b34b) Resetting default scope  EXCEPTION_RECORD:  (.exr -1) ExceptionAddress: 00007ffed7b1b349 (ntdll!RtlReportFatalFailure+0x0000000000000009)    ExceptionCode: c0000374   ExceptionFlags: 00000001 NumberParameters: 1    Parameter[0]: 00007ffed7b827f0  PROCESS_NAME:  w3wp.exe ...  

熟悉 windows ntheap 的朋友应该知道，ExceptionCode: c0000374 是经典的 堆破坏 状态码，那到底是谁破坏了呢？

2. 到底是谁破坏了NT堆

windows 给了 ntheap 强大的调试支持，默认开启了 Termination on corruption 破坏检测，也就是说当你使用 !heap -s 的时候会显示具体破坏的详情记录，输出如下：

 0:090> !heap -s   ************************************************************************************************************************                                               NT HEAP STATS BELOW ************************************************************************************************************************ ************************************************************** *                                                            * *                  HEAP ERROR DETECTED                       * *                                                            * **************************************************************  Details:  Heap address:  000001c14fd20000 Error address: 000001ce25531c50 Error type: HEAP_FAILURE_BLOCK_NOT_BUSY Details:    The caller performed an operation (such as a free             or a size check) that is illegal on a free block. Follow-up:  Check the error's stack trace to find the culprit.   Stack trace: Stack trace at 0x00007ffed7b82848     00007ffed7abe109: ntdll!RtlpLogHeapFailure+0x45     00007ffed7acbb0e: ntdll!RtlFreeHeap+0x9d3ce     00007ffeb093276f: OraOps12!ssmem_free+0xf     00007ffeb0943077: OraOps12!OpsMetFreeValCtx+0xd7     00007ffeb093cdd8: OraOps12!OpsDacDispose+0x2b8     00007ffe655e4374: +0x655e4374  LFH Key                   : 0x5baf44f8068da60f Termination on corruption : ENABLED           Heap     Flags   Reserv  Commit  Virt   Free  List   UCR  Virt  Lock  Fast                              (k)     (k)    (k)     (k) length      blocks cont. heap  ------------------------------------------------------------------------------------- 000001c14fd20000 00000002 1021576 964388 1020020  19222  6063   166    2    82f   LFH 000001c14fc70000 00008000      64      4     64      2     1     1    0      0       ...  

上面的 Error type: HEAP_FAILURE_BLOCK_NOT_BUSY 表示是一个 double free，从 Stack trace 看是 OpsDacDispose 方法造成的，应该和 Oracle 相关，这就比较迷了。。。

3. 是托管层触发的吗

是不是托管层触发的呢？这就需要理解 Windows 独有的 SEH 异常处理机制，也就是说 Windows 的异常都会在 内核态 走一圈，画个图如下：

只要找到 t1 时刻的崩溃点，然后观察线程栈即可，代码如下：

 0:090> .ecxr rax=00007ffec265d6e0 rbx=00000000c0000374 rcx=00000053653fe4f0 rdx=00007ffec1d3e9a0 rsi=0000000000000001 rdi=00007ffed7b827f0 rip=00007ffed7b1b349 rsp=00000053653fed10 rbp=000001c14fd20000  r8=000001c11957d9a0  r9=0000000000000033 r10=000001c453dbc7f0 r11=00007ffeb94db004 r12=0000000000000001 r13=000001c12e8526d0 r14=0000000000000000 r15=000001ce25531c60 iopl=0         nv up ei pl nz na po nc cs=0033  ss=002b  ds=002b  es=002b  fs=0053  gs=002b             efl=00000206 ntdll!RtlReportFatalFailure+0x9: 00007ffe`d7b1b349 eb00            jmp     ntdll!RtlReportFatalFailure+0xb (00007ffe`d7b1b34b) 0:090> k   *** Stack trace for last set context - .thread/.cxr resets it  # Child-SP          RetAddr               Call Site 00 00000053`653fed10 00007ffe`d7b1b313     ntdll!RtlReportFatalFailure+0x9 01 00000053`653fed60 00007ffe`d7b23b9e     ntdll!RtlReportCriticalFailure+0x97 02 00000053`653fee50 00007ffe`d7b23eaa     ntdll!RtlpHeapHandleError+0x12 03 00000053`653fee80 00007ffe`d7abe109     ntdll!RtlpHpHeapHandleError+0x7a 04 00000053`653feeb0 00007ffe`d7acbb0e     ntdll!RtlpLogHeapFailure+0x45 05 00000053`653feee0 00007ffe`b093276f     ntdll!RtlFreeHeap+0x9d3ce 06 00000053`653fef80 00007ffe`b0943077     OraOps12!ssmem_free+0xf 07 00000053`653fefb0 00007ffe`b093cdd8     OraOps12!OpsMetFreeValCtx+0xd7 08 00000053`653fefe0 00007ffe`655e4374     OraOps12!OpsDacDispose+0x2b8 09 00000053`653ff060 00007ffe`655e31cf     0x00007ffe`655e4374 0a 00000053`653ff150 00007ffe`6a80969d     0x00007ffe`655e31cf 0b 00000053`653ff1f0 00007ffe`c4b96d06     0x00007ffe`6a80969d 0c 00000053`653ff220 00007ffe`c4c30e81     clr!FastCallFinalizeWorker+0x6 0d 00000053`653ff250 00007ffe`c4c30e09     clr!FastCallFinalize+0x55 0e 00000053`653ff2a0 00007ffe`c4c30d3a     clr!MethodTable::CallFinalizer+0xb5 0f 00000053`653ff2f0 00007ffe`c4c30bf5     clr!CallFinalizer+0x5e 10 00000053`653ff330 00007ffe`c4c304dc     clr!FinalizerThread::DoOneFinalization+0x95 11 00000053`653ff410 00007ffe`c4c31777     clr!FinalizerThread::FinalizeAllObjects+0xbf 12 00000053`653ff450 00007ffe`c4b97d01     clr!FinalizerThread::FinalizeAllObjects_Wrapper+0x18 13 00000053`653ff480 00007ffe`c4b97c70     clr!ManagedThreadBase_DispatchInner+0x39 14 00000053`653ff4c0 00007ffe`c4b97bad     clr!ManagedThreadBase_DispatchMiddle+0x6c 15 00000053`653ff5c0 00007ffe`c4b9ac34     clr!ManagedThreadBase_DispatchOuter+0x75 16 00000053`653ff650 00007ffe`c4bf5271     clr!ManagedThreadBase_DispatchInCorrectAD+0x15 17 00000053`653ff680 00007ffe`c4b9ac72     clr!Thread::DoADCallBack+0x109 18 00000053`653ff830 00007ffe`c4c3172a     clr!ManagedThreadBase_DispatchInner+0x82 19 00000053`653ff870 00007ffe`c4c304dc     clr!FinalizerThread::DoOneFinalization+0x1f1 1a 00000053`653ff950 00007ffe`c4c3062b     clr!FinalizerThread::FinalizeAllObjects+0xbf 1b 00000053`653ff990 00007ffe`c4b97d01     clr!FinalizerThread::FinalizerThreadWorker+0xbb 1c 00000053`653ff9d0 00007ffe`c4b97c70     clr!ManagedThreadBase_DispatchInner+0x39 1d 00000053`653ffa10 00007ffe`c4b97bad     clr!ManagedThreadBase_DispatchMiddle+0x6c 1e 00000053`653ffb10 00007ffe`c4cf4d4a     clr!ManagedThreadBase_DispatchOuter+0x75 1f 00000053`653ffba0 00007ffe`c4d5044f     clr!FinalizerThread::FinalizerThreadStart+0x126 20 00000053`653ffc40 00007ffe`d6157e94     clr!Thread::intermediateThreadProc+0x86 21 00000053`653ffd00 00007ffe`d7a87ad1     kernel32!BaseThreadInitThunk+0x14 22 00000053`653ffd30 00000000`00000000     ntdll!RtlUserThreadStart+0x21  0:090> !clrstack  OS Thread Id: 0x5634 (90)         Child SP               IP Call Site 00000053653ff0b8 00007ffed7abf0e4 [InlinedCallFrame: 00000053653ff0b8] Oracle.DataAccess.Client.OpsDac.Dispose(IntPtr, IntPtr, IntPtr, IntPtr ByRef, Oracle.DataAccess.Client.OpoMetValCtx*, Oracle.DataAccess.Client.OpoDacValCtx* ByRef, Oracle.DataAccess.Client.OpoSqlValCtx*, Int32, Int32) 00000053653ff0b8 00007ffe655e4374 [InlinedCallFrame: 00000053653ff0b8] Oracle.DataAccess.Client.OpsDac.Dispose(IntPtr, IntPtr, IntPtr, IntPtr ByRef, Oracle.DataAccess.Client.OpoMetValCtx*, Oracle.DataAccess.Client.OpoDacValCtx* ByRef, Oracle.DataAccess.Client.OpoSqlValCtx*, Int32, Int32) 00000053653ff060 00007ffe655e4374 DomainNeutralILStubClass.IL_STUB_PInvoke(IntPtr, IntPtr, IntPtr, IntPtr ByRef, Oracle.DataAccess.Client.OpoMetValCtx*, Oracle.DataAccess.Client.OpoDacValCtx* ByRef, Oracle.DataAccess.Client.OpoSqlValCtx*, Int32, Int32) 00000053653ff150 00007ffe655e31cf Oracle.DataAccess.Client.OracleDataReader.Dispose(Boolean) 00000053653ff1f0 00007ffe6a80969d Oracle.DataAccess.Client.OracleDataReader.Finalize() 00000053653ff608 00007ffec4b96d06 [DebuggerU2MCatchHandlerFrame: 00000053653ff608]  00000053653ff788 00007ffec4b96d06 [ContextTransitionFrame: 00000053653ff788]  00000053653ff8d0 00007ffec4b96d06 [GCFrame: 00000053653ff8d0]  00000053653ffb58 00007ffec4b96d06 [DebuggerU2MCatchHandlerFrame: 00000053653ffb58]   

从调用栈来看，原来是 终结器线程 在调用 OracleDataReader.Dispose() 方法的时候抛的异常，这个结果还是挺意外的，也就是说这个问题不是用户代码造成的，真的是 Oracle 这个 OraOps12.dll 造成的。。。

接下来用 lm 观察下这个 dll 的详情信息，输出如下：

 0:090> lmDvmOraOps12 Browse full module list start             end                 module name 00007ffe`b0920000 00007ffe`b098c000   OraOps12 C (export symbols)       OraOps12.dll     Loaded symbol image file: OraOps12.dll     Image path: C:ODACxxxxOraOps12.dll     Image name: OraOps12.dll     Browse all global symbols  functions  data     Timestamp:        Sat Sep 26 23:16:56 2015 (5606B6E8)     CheckSum:         00000000     ImageSize:        0006C000     File version:     2.121.2.0     Product version:  2.121.2.0     File flags:       0 (Mask 3F)     File OS:          4 Unknown Win32     File type:        2.0 Dll     File date:        00000000.00000000     Translations:     0409.04b0     Information from resource tables:         CompanyName:      Oracle Corporation         ProductName:      Oracle Data Provider for .NET         InternalName:     OraOps         OriginalFilename: OraOps12.dll         ProductVersion:   2.121.2.0 ODAC RELEASE 4         FileVersion:      2.121.2.0         FileDescription:  Oracle Provider Services         LegalCopyright:   Copyright © 2014  

虽然对 Oracle 不熟，但从 Timestamp: Sat Sep 26 23:16:56 2015 来看应该是一个比较老的 DLL 了，所以给到朋友的建议就是升级 OraOps12.dll 。

4. 是否有同行者

有时候直接让朋友升级dll有点缺少底气，最好就是找到一些同行者，经过一顿搜索，还真有同行者，又多了一份说服力，网址： https://techcommunity.microsoft.com/t5/iis-support-blog/w3wp-exe-crash-exception-code-0xc0000005/ba-p/334351 。

三：总结

在百加dump的分析旅程中，碰到和 Oracle SDK 相关的也有 3+ 起了，可能也许这些 SDK 在对接 .NET 上还不是特别稳健，大家在使用上尽量更新到最新版本吧，且用且珍惜！