一:背景
1. 讲故事
很多.NET开发者在学习高级调试的时候,使用sos的命令输出会发现这里也看不懂那里也看不懂,比如截图中的这位朋友。
.NET高级调试属于一个偏冷门的领域,国内可观测的资料比较少,所以很多东西需要你自己去探究源代码,然后用各种调试工具去验证,相关源代码如下:
- coreclr: https://github.com/dotnet/runtime
- windows: https://github.com/reactos/reactos
- linux: https://www.kernel.org/
这一篇权当抛砖引玉,来引导一下如何去探索。
二:WinDbg 分析
1. 复原现象
为了方便讲述先上一段简单的测试代码,来观察一下线程列表中所谓的 lockcount 列。
internal class Program { static void Main(string[] args) { Console.WriteLine("hello world..."); Debugger.Break(); } }
接下来用 windbg 附加一下,使用 !threads
命令观察输出。
0:000> !t ThreadCount: 3 UnstartedThread: 0 BackgroundThread: 2 PendingThread: 0 DeadThread: 0 Hosted Runtime: no Lock DBG ID OSID ThreadOBJ State GC Mode GC Alloc Context Domain Count Apt Exception 0 1 14c8 000001A8B862DBB0 2a020 Preemptive 000001A8BA1DDEE0:000001A8BA1DF478 000001a8b8623c20 0 MTA 5 2 71e0 000001A8D25740B0 2b220 Preemptive 0000000000000000:0000000000000000 000001a8b8623c20 0 MTA (Finalizer) 7 3 31b0 000001A8B86E7730 102a220 Preemptive 0000000000000000:0000000000000000 000001a8b8623c20 0 MTA (Threadpool Worker)
从上面的卦中可以看到 Lock Count=0
,那这里的值取自源码的哪里呢?
2. 取自哪里
大家要知道每一行记录都是在coreclr Thread
中摘取出来的字段,言外之意就是可以把 Thread 所有信息给展示出来,可以用 dt 命令,简化后如下:
0:000> dt coreclr!Thread 000001A8B862DBB0 +0x000 __VFN_table : 0x00007ffd`d2f93c70 =00007ffd`d30ce8d8 m_DetachCount : 0n0 =00007ffd`d30ce8d4 m_ActiveDetachCount : 0n0 =00007ffd`d30cf4dc m_threadsAtUnsafePlaces : Volatile<long> +0x008 m_State : Volatile<enum Thread::ThreadState> +0x00c m_fPreemptiveGCDisabled : Volatile<unsigned long> +0x010 m_pFrame : 0x0000006f`df57e328 Frame +0x018 m_pDomain : 0x000001a8`b8623c20 AppDomain +0x020 m_dwLockCount : 0 +0x024 m_ThreadId : 1 ...
通过仔细分析卦中信息,可以发现这个 lockcount 其实就是 m_dwLockCount
字段,知道是这个字段之后接下来就是寻找coreclr源码啦,截图如下:
从源码注释中看的非常清楚,这个字段可用来跟踪5中锁。
- critical section
- spin lock
- syncblock lock
- EE Crst
- GC lock
接下来搜索下源码看看 m_dwLockCount 是怎么更新的,可以发现如下两处,并且都是和同步块索引相关,截图如下:
inline void Thread::IncLockCount() { LIMITED_METHOD_CONTRACT; _ASSERTE(GetThread() == this); m_dwLockCount++; _ASSERTE(m_dwLockCount != 0 || HasThreadStateNC(TSNC_UnbalancedLocks)); } inline void Thread::DecLockCount() { LIMITED_METHOD_CONTRACT; _ASSERTE(GetThread() == this); _ASSERTE(m_dwLockCount > 0 || HasThreadStateNC(TSNC_UnbalancedLocks)); m_dwLockCount--; }
接下来要做的一件事就是如何自增 m_dwLockCount 值 来模拟达到这位朋友的 1024。
3. 如何模拟复现
其实当你知道是跟踪这五种锁,要模拟就非常简单了,为了方便讲述上一段测试代码,利用 lock 让 m_dwLockCount 字段不断自增,参考如下:
internal class Program { public static object[] locks; static void Main(string[] args) { locks = Enumerable.Range(0, 100).Select(i => new object()).ToArray(); foreach (var item in locks) { Monitor.Enter(item); } Debugger.Break(); Console.ReadLine(); } }
接下来用 windbg 附加观察一下。
0:000> !t ThreadCount: 3 UnstartedThread: 0 BackgroundThread: 2 PendingThread: 0 DeadThread: 0 Hosted Runtime: no Lock DBG ID OSID ThreadOBJ State GC Mode GC Alloc Context Domain Count Apt Exception 0 1 6c90 00000239840DDBB0 2a020 Preemptive 0000023985B2BB38:0000023985B2D478 00000239840d3c20 100 MTA 5 2 2178 00000239859640B0 2b220 Preemptive 0000000000000000:0000000000000000 00000239840d3c20 0 MTA (Finalizer) 7 3 6d00 0000023984197700 102a220 Preemptive 0000000000000000:0000000000000000 00000239840d3c20 0 MTA (Threadpool Worker) 0:000> dt coreclr!Thread 00000239840DDBB0 +0x000 __VFN_table : 0x00007ffd`d2f93c70 =00007ffd`d30ce8d8 m_DetachCount : 0n0 =00007ffd`d30ce8d4 m_ActiveDetachCount : 0n0 =00007ffd`d30cf4dc m_threadsAtUnsafePlaces : Volatile<long> +0x008 m_State : Volatile<enum Thread::ThreadState> +0x00c m_fPreemptiveGCDisabled : Volatile<unsigned long> +0x010 m_pFrame : 0x00000074`a937e888 Frame +0x018 m_pDomain : 0x00000239`840d3c20 AppDomain +0x020 m_dwLockCount : 0x64
这里顺便提一下,如果你想实时观察 m_dwLockCount 更改的线程栈信息,可以在 Thread+0x20
处下一个ba硬件断点即可。
0:000> !t ThreadCount: 3 UnstartedThread: 0 BackgroundThread: 2 PendingThread: 0 DeadThread: 0 Hosted Runtime: no Lock DBG ID OSID ThreadOBJ State GC Mode GC Alloc Context Domain Count Apt Exception 0 1 6188 00000290D550DBB0 2a020 Preemptive 00000290D6F4AD20:00000290D6F4B478 00000290d5503c20 0 MTA 5 2 686c 00000290EF3F40B0 2b220 Preemptive 0000000000000000:0000000000000000 00000290d5503c20 0 MTA (Finalizer) 7 3 312c 00000290D55C7740 102a220 Preemptive 0000000000000000:0000000000000000 00000290d5503c20 0 MTA (Threadpool Worker) 0:000> ba w4 00000290D550DBB0+0x20 0:000> g ModLoad: 00007ffd`cff80000 00007ffd`cffd1000 C:\Program Files\dotnet\shared\Microsoft.NETCore.App\3.1.32\System.Collections.dll ModLoad: 00007ffd`d00d0000 00007ffd`d0103000 C:\Program Files\dotnet\shared\Microsoft.NETCore.App\3.1.32\System.Runtime.Extensions.dll Breakpoint 1 hit coreclr!JIT_MonEnter_Portable+0x65: 00007ffd`d298c8a5 4883c430 add rsp,30h 0:000> kL 5 # Child-SP RetAddr Call Site 00 000000b5`f917e780 00007ffd`72e917f4 coreclr!JIT_MonEnter_Portable+0x65 01 000000b5`f917e7c0 00007ffd`d29b6ca3 0x00007ffd`72e917f4 02 000000b5`f917e860 00007ffd`d294cc62 coreclr!CallDescrWorkerInternal+0x83 03 (Inline Function) --------`-------- coreclr!CallDescrWorkerWithHandler+0x57 04 000000b5`f917e8a0 00007ffd`d2953c29 coreclr!MethodDescCallSite::CallTargetWorker+0x196 ...
三:总结
还是那句话,国内.NET高级调试
方面的资料甚少,要想理解sos命令输出是什么意思,切记多看源码,其实本篇重要的不是知识,而是告诉你如何去探究新知的方法。