记一次 .NET 某医联体管理系统崩溃分析

news/2025/11/12 17:57:28/文章来源:https://www.cnblogs.com/huangxincheng/p/19214907

一：背景

1. 讲故事

这段时间都在跑外卖，感觉好久都没写文章了，今天继续给大家带来一篇崩溃类的生产事故，这是微信上有位老朋友找到我的，让我帮忙看下为啥崩溃了，dump也在手，接下来就可以一顿分析。

二：崩溃分析

1. 为什么会崩溃

双击打开dump文件，会看到崩溃信息通览，参考如下：


Executable search path is: 
Windows 10 Version 17763 MP (48 procs) Free x64
Product: Server, suite: TerminalServer DataCenter SingleUserTS
Edition build lab: 17763.1.amd64fre.rs5_release.180914-1434
Debug session time: Fri Oct 31 17:38:42.000 2025 (UTC + 8:00)
System Uptime: 14 days 2:42:29.643
Process Uptime: 0 days 0:00:58.000
................................................................
.......................................
Loading unloaded module list
.
This dump file has an exception of interest stored in it.
The stored exception information can be accessed via .ecxr.
(5a74.6250): Unknown exception - code c0000374 (first/second chance not available)
For analysis of this file, run !analyze -v
ntdll!NtWaitForMultipleObjects+0x14:
00007ffe`57baf0e4 c3              ret

从卦中看崩溃码是 c0000374，即 ntheap 损坏，哈哈，到这里一下子就把范围给缩小了。

2. 为什么ntheap 损坏

那为什么ntheap会损坏呢？可以使用 .ecxr 切到崩溃时的调用栈，观察崩溃行为。


0:032> .ecxr0:032> k*** Stack trace for last set context - .thread/.cxr resets it# Child-SP          RetAddr               Call Site
00 000000b4`8503ede0 00007ffe`57c0b313     ntdll!RtlReportFatalFailure+0x9
01 000000b4`8503ee30 00007ffe`57c13b9e     ntdll!RtlReportCriticalFailure+0x97
02 000000b4`8503ef20 00007ffe`57c13eaa     ntdll!RtlpHeapHandleError+0x12
03 000000b4`8503ef50 00007ffe`57bae109     ntdll!RtlpHpHeapHandleError+0x7a
04 000000b4`8503ef80 00007ffe`57bbbb0e     ntdll!RtlpLogHeapFailure+0x45
05 000000b4`8503efb0 00007ffe`17d17b3f     ntdll!RtlFreeHeap+0x9d3ce
06 000000b4`8503f050 00007ffe`541392af     AcLayers!NS_FaultTolerantHeap::APIHook_RtlFreeHeap+0x41f
07 000000b4`8503f0b0 00007ffe`3773b17e     KERNELBASE!LocalFree+0x2f
08 000000b4`8503f0f0 00007ffe`37661d12     mscorlib_ni+0x58b17e
09 000000b4`8503f1a0 00007ffd`e49fe127     mscorlib_ni!System.Runtime.InteropServices.Marshal.FreeHGlobal+0x22 [f:\dd\ndp\clr\src\BCL\system\runtime\interopservices\marshal.cs @ 1212] 
...0:032> !clrstack
OS Thread Id: 0x6250 (32)Child SP               IP Call Site
000000b48503f118 00007ffe57baf0e4 [InlinedCallFrame: 000000b48503f118] Microsoft.Win32.Win32Native.LocalFree(IntPtr)
000000b48503f118 00007ffe3773b17e [InlinedCallFrame: 000000b48503f118] Microsoft.Win32.Win32Native.LocalFree(IntPtr)
000000b48503f0f0 00007ffe3773b17e DomainNeutralILStubClass.IL_STUB_PInvoke(IntPtr)
000000b48503f1a0 00007ffe37661d12 System.Runtime.InteropServices.Marshal.FreeHGlobal(IntPtr) [f:\dd\ndp\clr\src\BCL\system\runtime\interopservices\marshal.cs @ 1212]
000000b48503f1e0 00007ffde49fe127 b.B+A.MoveNext()
000000b48503f240 00007ffe376b3423 System.Threading.ExecutionContext.RunInternal(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object, Boolean) [f:\dd\ndp\clr\src\BCL\system\threading\executioncontext.cs @ 954]
000000b48503f310 00007ffe376b32b4 System.Threading.ExecutionContext.Run(System.Threading.ExecutionContext, System.Threading.ContextCallback, System.Object, Boolean) [f:\dd\ndp\clr\src\BCL\system\threading\executioncontext.cs @ 902]
...
000000b48503f5c0 00007ffde49fb04e DomainBoundILStubClass.IL_STUB_ReversePInvoke(Int32, Int32, Int64)

从卦中可以清晰的看到是 b.B+A.MoveNext 方法中调用了 FreeHGlobal 导致的NTHeap崩溃，如果你经验比较足的话，看到这个 FreeHGlobal 就应该想到 double free 问题，这是一个经典的问题。

3. 何为 double free

双释放即对一个 block 块进行二次释放，windows 的 RtlFreeHeap 方法会在业务逻辑中对这种情况直接判为异常，接下来你或许想知道这个 block 的地址是什么？这个可以用 !heap -s 观察，参考代码如下：


0:032> !heap -s************************************************************************************************************************NT HEAP STATS BELOW
************************************************************************************************************************Details:Heap address:  0000028c75bb0000
Error address: 0000028c786018a0
Error type: HEAP_FAILURE_BLOCK_NOT_BUSY
Details:    The caller performed an operation (such as a freeor 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 0x00007ffe57c7284800007ffe57bae109: ntdll!RtlpLogHeapFailure+0x4500007ffe57bbbb0e: ntdll!RtlFreeHeap+0x9d3ce00007ffe17d17b3f: AcLayers!NS_FaultTolerantHeap::APIHook_RtlFreeHeap+0x41f00007ffe541392af: KERNELBASE!LocalFree+0x2f00007ffe3773b17e: mscorlib_ni+0x58b17e00007ffe37661d12: mscorlib_ni!System.Runtime.InteropServices.Marshal.FreeHGlobal+0x2200007ffde49fe127: +0xe49fe127LFH Key                   : 0x765363a7204cf973
Termination on corruption : ENABLEDHeap     Flags   Reserv  Commit  Virt   Free  List   UCR  Virt  Lock  Fast (k)     (k)    (k)     (k) length      blocks cont. heap 
-------------------------------------------------------------------------------------
0000028c75bb0000 00000002   17920   9256  16364   2120   214     5    1      a   LFHExternal fragmentation  23 % (214 free blocks)
0000028c75b40000 00008000      64      4     64      2     1     1    0      0      
0000028c75de0000 00001002    2636    132   1080     20     5     2    0      0   LFH
0000028c76190000 00001002    4680   2268   3124   1420    40     3    0      0   LFHExternal fragmentation  62 % (40 free blocks)
0000028c76130000 00001002    2636    472   1080      5    27     2    0      0   LFH
0000028c767f0000 00041002      60      8     60      5     1     1    0      0      
0000028c77020000 00041002      60     16     60      2     2     1    0      0      
-------------------------------------------------------------------------------------

从卦中可以看到 Heap address: 0000028c75bb0000 即为 block 地址，接下来使用 !heap -x 0000028c786018a0 观察这个 block 块的状态，可以看到此时确实是 free 的。


0:032> !heap -x 0000028c786018a0
Entry             User              Heap              Segment               Size  PrevSize  Unused    Flags
-------------------------------------------------------------------------------------------------------------
0000028c786018a0  0000028c786018b0  0000028c75bb0000  0000028c785c80d0        e0      -            0  LFH;free

到这里问题的成因我们是完全搞清楚了，接下来就是反推问题代码的时候了。