0:000:x86> !clrstackOS Thread Id: 0x4a08 (0)Child SP IP Call Site012fe784 0000002b [HelperMethodFrame_1OBJ: 012fe784] System.Threading.WaitHandle.WaitOneNative(System.Runtime.InteropServices.SafeHandle, UInt32, Boolean, Boolean)012fe868 7115d952 System.Threading.WaitHandle.InternalWaitOne(System.Runtime.InteropServices.SafeHandle, Int64, Boolean, Boolean) [f:\dd\ndp\clr\src\BCL\system\threading\waithandle.cs @ 243]012fe880 7115d919 System.Threading.WaitHandle.WaitOne(Int32, Boolean) [f:\dd\ndp\clr\src\BCL\system\threading\waithandle.cs @ 194]012fe894 711e89bf System.Threading.WaitHandle.WaitOne(Int32) [f:\dd\ndp\clr\src\BCL\system\threading\waithandle.cs @ 220]012fe89c 6fb186b8 System.Threading.ReaderWriterLockSlim.WaitOnEvent(System.Threading.EventWaitHandle, UInt32 ByRef, TimeoutTracker, EnterLockType)012fe8e0 6fb17892 System.Threading.ReaderWriterLockSlim.TryEnterReadLockCore(TimeoutTracker)012fe920 6fb17562 System.Threading.ReaderWriterLockSlim.TryEnterReadLock(TimeoutTracker)012fe94c 0325f49f xxx.QuyitpjK0dXKR6IyqH(System.Object)012fe964 0325ee8a xxx.RWAutoLock..ctor(System.Threading.ReaderWriterLockSlim, Boolean)...
从卦中的线程栈数据来看,貌似是卡在一个读写锁TryEnterReadLock 上,根据读写锁的规则,一定有人实行了一个 WriteLock 并且出不来,接下来便是探求持有这个 lock 的线程。
2. 到底谁在持有如果是 lock ,相信很多朋友都知道用 !syncblk 命令,那读写锁用什么命令呢?说实话我也搞不清楚,只能先挖挖 ReaderWriterLockSlim 类本身,看看有没有什么新创造。

0:000:x86> !DumpObj /d 03526f38Name: System.Threading.ReaderWriterLockSlimMethodTable: 6f947428EEClass: 6f9a92dcSize: 72(0x48) bytesFile: C:\Windows\Microsoft.Net\assembly\GAC_MSIL\System.Core\v4.0_4.0.0.0__b77a5c561934e089\System.Core.dllFields: MT Field Offset Type VT Attr Value Name70da878c 40004aa 38 System.Boolean 1 instance 0 _fIsReentrant6f92fa28 40004ab 3c ...LockSlim+SpinLock 1 instance 03526f74 _spinLock70dfba4c 40004ac 1c System.UInt32 1 instance 20 _numWriteWaiters70dfba4c 40004ad 20 System.UInt32 1 instance 1 _numReadWaiters70dfba4c 40004ae 24 System.UInt32 1 instance 0 _numWriteUpgradeWaiters70dfba4c 40004af 28 System.UInt32 1 instance 0 _numUpgradeWaiters6f93d764 40004b0 39 System.Byte 1 instance 0 _waiterStates70da42a8 40004b1 2c System.Int32 1 instance -1 _upgradeLockOwnerId70da42a8 40004b2 30 System.Int32 1 instance 11 _writeLockOwnerId70da6924 40004b3 c ...g.EventWaitHandle 0 instance 034844d0 _writeEvent70da6924 40004b4 10 ...g.EventWaitHandle 0 instance 042a69c8 _readEvent70da6924 40004b5 14 ...g.EventWaitHandle 0 instance 00000000 _upgradeEvent70da6924 40004b6 18 ...g.EventWaitHandle 0 instance 00000000 _waitUpgradeEvent70da150c 40004b8 4 System.Int64 1 instance 367 _lockID70da878c 40004ba 3a System.Boolean 1 instance 0 _fUpgradeThreadHoldingRead70dfba4c 40004bc 34 System.UInt32 1 instance 3221225472 _owners70da878c 40004c2 3b System.Boolean 1 instance 0 _fDisposed70da42a8 40004a9 4dc System.Int32 1 static 4 ProcessorCount70da150c 40004b7 4d4 System.Int64 1 static 1882 s_nextLockID6f942b7c 40004b9 0 ...ReaderWriterCount 0 TLstatic t_rwc
结合源码剖析,创造上面的 _writeLockOwnerId=11 便是持有锁的线程ID,找到持有线程就好办了,把这个 managedid=11 转成 dbgid 再不雅观察。

0:000:x86> !t 13 11 47bc 0a0702c0 1029220 Preemptive 00000000:00000000 01425ed0 0 MTA (Threadpool Worker) 0:013:x86> !clrstackOS Thread Id: 0x47bc (13)Child SP IP Call Site07e4f1ac 0000002b [InlinedCallFrame: 07e4f1ac] 07e4f1a4 09e38597 DomainBoundILStubClass.IL_STUB_PInvoke(IntPtr)07e4f1ac 09e38334 [InlinedCallFrame: 07e4f1ac] System.Data.SQLite.UnsafeNativeMethods.sqlite3_step(IntPtr)07e4f1dc 09e38334 System.Data.SQLite.SQLite3.Step(System.Data.SQLite.SQLiteStatement)07e4f228 09e36fe8 System.Data.SQLite.SQLiteDataReader.NextResult()07e4f250 09e36ceb System.Data.SQLite.SQLiteDataReader..ctor(System.Data.SQLite.SQLiteCommand, System.Data.CommandBehavior)07e4f270 09e367ce System.Data.SQLite.SQLiteCommand.ExecuteReader(System.Data.CommandBehavior)07e4f284 09e36732 System.Data.SQLite.SQLiteCommand.ExecuteNonQuery(System.Data.CommandBehavior)07e4f2b0 09e366e6 System.Data.SQLite.SQLiteCommand.ExecuteNonQuery()07e4f2bc 09e350dc SqlSugar.AdoProvider.ExecuteCommand(System.String, SqlSugar.SugarParameter[])07e4f388 13189518 SqlSugar.InsertableProvider`1[[System.__Canon, mscorlib]].ExecuteCommand()07e4f420 0181ac4a xxx.OperateLog+d__8.MoveNext()...0:013:x86> kCvRegToMachine(x86) conversion failure for 0x14fX86MachineInfo::SetVal: unknown register 0 requested # ChildEBP RetAddr 00 07e4ede0 76c9ad10 ntdll_76ed0000!NtFlushBuffersFile+0xc01 07e4ede0 6b27af8c KERNELBASE!FlushFileBuffers+0x30WARNING: Stack unwind information not available. Following frames may be wrong.02 07e4edf0 6b270256 SQLite_Interop!SI768767362ea03a94+0xf73c03 07e4ee1c 6b267938 SQLite_Interop!SI768767362ea03a94+0x4a0604 07e4ee38 6b2599e1 SQLite_Interop!SI83d1cf4976f57337+0x84c805 07e4ee80 6b25902b SQLite_Interop!SIa3401e98cbad673e+0x320106 07e4ee98 6b25258c SQLite_Interop!SIa3401e98cbad673e+0x284b07 07e4f168 6b255a05 SQLite_Interop!SI327cfc7a6b1fd1fb+0x633c08 07e4f19c 09e38597 SQLite_Interop!SI9c6d7cd7b7d38055+0x255
结合卦中的读写信息,大概知道了原来是用写锁来写sqlite,后者卡在缓冲区刷新函数 NtFlushBuffersFile 上,方法署名如下:
NTSTATUS NtFlushBuffersFile( HANDLE FileHandle, IO_STATUS_BLOCK IoStatusBlock);
有些朋友可能想看一下到底怎么写的,那就大略的反编译一下代码:
到这里基本就搞清楚了,由于 13号 线程持有了 写锁,导致主线程要用读锁操作 sqlite 时进行了永劫光等待。
办理办法就比较大略了,主线程尽可能的只做UI更新的操作,不要让他触发各种锁,否则就有等锁的概率发生。
3. NtFlushBuffersFile 怎么了有些朋友可能要问为什么 NtFlushBuffersFile 函数会卡去世不返回,要想找到这个答案,须要看下反汇编。
0:013:x86> uf ntdll_76ed0000!NtFlushBuffersFilentdll_76ed0000!NtFlushBuffersFile:76f41ad0 b84b000000 mov eax,4Bh76f41ad5 ba7071f576 mov edx,offset ntdll_76ed0000!Wow64SystemServiceCall (76f57170)76f41ada ffd2 call edx76f41adc c20800 ret 80:013:x86> u 76F57170hntdll_76ed0000!Wow64SystemServiceCall:76f57170 ff252892ff76 jmp dword ptr [ntdll_76ed0000!Wow64Transition (76ff9228)]0:013:x86> u 76ec7000wow64cpu!KiFastSystemCall:76ec7000 ea0970ec763300 jmp 0033:76EC700976ec7007 0000 add byte ptr [eax],al76ec7009 41 inc ecx76ec700a ffa7f8000000 jmp dword ptr [edi+0F8h]
从汇编代码看,NtFlushBuffersFile 通过 KiFastSystemCall 进入内核态了,用户态dump是没法看内核态的,以是也无法连续深究下去。
不过也可以看下这个线程过往的 GetLastError() 值,可能有些收成,利用 !gle 命令。
0:013:x86> !gleLastErrorValue: (Win32) 0x26 (38) - <Unable to get error code text>LastStatusValue: (NTSTATUS) 0xc0000008 - <Unable to get error code text>
根据上面的状态码,去msdn上搜一下详细信息。
从缺点解释看,可能是这个sqlite文件有什么问题,又是句柄无效,又是读到头了,疑惑是操作sqlite 的时候涌现了文件破坏。
现在转头看看,如果想对 Sqlite 进行并发读写,开启下 Write-Ahead Logging 模式该当就可以了,不须要在程序里面进行读写掌握。
以是终极的建议便是:
开启WAL模式删掉读写掌握三:总结这次卡去世事件还是挺故意思的,熟习了下 ReaderWriterLockSlim 又对 sqlite 有了一个新的认识。










