记一次 Visual Studio 2022 卡死分析
一：背景
1. 讲故事
最近不知道咋了，各种程序有问题都寻上我了，你说 .NET 程序有问题找我能理解，Windows 崩溃找我，我也可以试试看，毕竟对 Windows 内核也知道一丢丢，那 Visual Studio 有问题找我就说不过去了，但又不好拒绝，就让朋友发下卡死的 dump 我看一看。
二：WinDbg 分析
1. 到底是哪里的卡死
因为 VS 是窗体程序，所以在卡死的时候看下主线程便知，使用 ~0s;!clrstack 即可。
0:000> k
# Child-SP RetAddr Call Site
00 0000004b`acaf9b90 000001ed`309f0f28 0x00007ffb`1b77bfe8
01 0000004b`acaf9b98 00007ffb`4a03e397 0x000001ed`309f0f28
02 0000004b`acaf9ba0 00007ffb`4a04c08e PresentationFramework_ni!System.Windows.Controls.ItemContainerGenerator.DoLinearSearch+0x1d7
03 0000004b`acaf9c70 00007ffb`4ab3bd36 PresentationFramework_ni!System.Windows.Controls.ItemContainerGenerator.ContainerFromItem+0x8e
04 0000004b`acaf9ce0 00007ffb`4ab3bd6e PresentationFramework_ni!System.Windows.Automation.Peers.ItemAutomationPeer.GetWrapper+0xc6
05 0000004b`acaf9d20 00007ffb`4ab3c94f PresentationFramework_ni!System.Windows.Automation.Peers.ItemAutomationPeer.GetWrapperPeer+0xe
06 0000004b`acaf9d60 00007ffb`4ba3f72c PresentationFramework_ni!System.Windows.Automation.Peers.ItemAutomationPeer.IsControlElementCore+0xf
07 0000004b`acaf9d90 00007ffb`4ba42026 PresentationCore_ni!System.Windows.Automation.Peers.AutomationPeer.IsControlElement+0x3c
08 0000004b`acaf9de0 00007ffb`4ba41e8b PresentationCore_ni!System.Windows.Automation.Peers.AutomationPeer.IsControlElement+0x36
09 0000004b`acaf9e20 00007ffb`4bcc5632 PresentationCore_ni!System.Windows.Automation.Peers.AutomationPeer.GetPropertyValue+0x7b
0a 0000004b`acaf9e70 00007ffb`4c182cf8 PresentationCore_ni!MS.Internal.Automation.ElementUtil.<>c__DisplayClass11_0.<Invoke>b__0+0x32
0b 0000004b`acaf9eb0 00007ffb`4c182bf6 WindowsBase_ni!System.Windows.Threading.ExceptionWrapper.InternalRealCall+0x68
0c 0000004b`acaf9f20 00007ffb`4c180202 WindowsBase_ni!System.Windows.Threading.ExceptionWrapper.TryCatchWhen+0x36
0d 0000004b`acaf9f70 00007ffb`4bca4423 WindowsBase_ni!System.Windows.Threading.Dispatcher.LegacyInvokeImpl+0x172
0e 0000004b`acafa010 00007ffb`480629e1 PresentationCore_ni!MS.Internal.Automation.ElementUtil.Invoke+0xb3
0f 0000004b`acafa070 00007ffb`7af71059 UIAutomationTypes_ni+0x729e1
10 0000004b`acafa0e0 00007ffb`7ae13eba clr!COMToCLRDispatchHelper+0x39
11 0000004b`acafa110 00007ffb`7af70fb7 clr!COMToCLRWorker+0x1ea
12 0000004b`acafa1b0 00007ffb`625b4cc9 clr!GenericComCallStub+0x57
...
我丢，这线程栈一看有意外发现哈，这 PresentationCore_ni 不是 WPF 的专用库嘛，下面还有 clr ，看样子 VS 的UI是 WPF 写的，顿时有一种亲切感，那既然是 .NET 程序我还是可以分析的。
进一步观察线程栈，可以看到它没有非托管的部分，诸如：user32.dll，ntdll.dll，也就说明此时的卡死只是托管层面，接下来使用 sos 专有的 !clrstack 观察，删减后如下：
0:000> !clrstack
OS Thread Id: 0x8144 (0)
Child SP IP Call Site
0000004bacafa220 00007ffb1b77bfe8 [ComMethodFrame: 0000004bacafa220]
0000004bacafb0d8 00007ffb1b77bfe8 [InlinedCallFrame: 0000004bacafb0d8] MS.Win32.UnsafeNativeMethods.OleSetClipboard(System.Runtime.InteropServices.ComTypes.IDataObject)
0000004bacafb0d8 00007ffb4c357e22 [InlinedCallFrame: 0000004bacafb0d8] MS.Win32.UnsafeNativeMethods.OleSetClipboard(System.Runtime.InteropServices.ComTypes.IDataObject)
0000004bacafb0a0 00007ffb4c357e22 DomainNeutralILStubClass.IL_STUB_PInvoke(System.Runtime.InteropServices.ComTypes.IDataObject)
0000004bacafb180 00007ffb4ba0f5ed System.Windows.Clipboard.CriticalSetDataObject(System.Object, Boolean)
0000004bacafb1c0 00007ffae3b1bd6a Microsoft.VisualStudio.Text.Utilities.WpfClipboardService.SetData(System.String, System.String, System.String, Boolean, System.String, Boolean, Boolean)
0000004bacafb210 00007ffae3b1bc1c Microsoft.VisualStudio.Text.Operations.Implementation.EditorOperations.CopyToClipboard(System.String, System.String, Boolean, Boolean)
0000004bacafb270 00007ffae3b199a2 Microsoft.VisualStudio.Text.Operations.Implementation.EditorOperations+c__DisplayClass176_0.b__0()
0000004bacafb2a0 00007ffae3b196ca Microsoft.VisualStudio.Text.Operations.Implementation.EditorOperations.CopySelection()
...
0000004bacafcc20 00007ffb2edd765b JetBrains.ReSharper.Feature.Services.Clipboard.CopyPasteAssistManager.DoCopyOrCut(JetBrains.Application.DataContext.IDataContext, JetBrains.Application.UI.Actions.DelegateExecute)
0000004bacafcc70 00007ffb2edd7628 JetBrains.ReSharper.Feature.Services.Clipboard.CopyPasteAssistManager.DoCopyOrCut(JetBrains.Application.DataContext.IDataContext, JetBrains.Application.UI.Actions.DelegateExecute)
0000004bacafcd00 00007ffb2edd5407 JetBrains.ReSharper.InplaceRefactorings.CutCopyPaste.CopyPasteManager.DoCopyOrCut(JetBrains.Application.DataContext.IDataContext, Boolean, JetBrains.Application.UI.Actions.DelegateExecute)
0000004bacafcd70 00007ffb2edd53c4 JetBrains.ReSharper.InplaceRefactorings.CutCopyPaste.CopyPasteManager.DoCopyOrCut(JetBrains.Application.DataContext.IDataContext, Boolean, JetBrains.Application.UI.Actions.DelegateExecute)
0000004bacafce50 00007ffb2edd4d88 JetBrains.ReSharper.Feature.Services.Clipboard.ClipboardActionHandler.Execute(JetBrains.Application.DataContext.IDataContext, JetBrains.Application.UI.Actions.DelegateExecute)
...
0000004bacafdc40 00007ffb2268a2da DomainNeutralILStubClass.IL_STUB_COMtoCLR(IntPtr, Int32, Int32, Int64, Int64)
0000004bacafde00 00007ffb7af71011 [ComMethodFrame: 0000004bacafde00]
从卦中看，是在处理剪贴板OleSetClipboard的逻辑中一直出不来，而且还有一个外来的 JetBrains.ReSharper 插件，看样子朋友的某些操作让 Resharper 介入了。
为了进一步验证是不是 Resharper 导致的，可以根据 ip 找到所属的模块。
0:000> !ip2md 00007ffb2edd5407
MethodDesc: 00007ffb28e984b0
Method Name: JetBrains.ReSharper.InplaceRefactorings.CutCopyPaste.CopyPasteManager.DoCopyOrCut(JetBrains.Application.DataContext.IDataContext, Boolean, JetBrains.Application.UI.Actions.DelegateExecute)
Class: 00007ffb28ea1b40
MethodTable: 00007ffb28e98578
mdToken: 00000000060000aa
Module: 00007ffb201c48b8
IsJitted: yes
CodeAddr: 00007ffb2edd4f60
Transparency: Critical
0:000> !DumpModule /d 00007ffb201c48b8
Name: C:\Users\Administrator\AppData\Local\JetBrains\Installations\ReSharperPlatformVs17_265273ed_001\JetBrains.ReSharper.InplaceRefactorings.dll
Attributes: PEFile SupportsUpdateableMethods
Assembly: 000001edb0cce780
LoaderHeap: 0000000000000000
TypeDefToMethodTableMap: 00007ffb201d0020
TypeRefToMethodTableMap: 00007ffb201d03a8
MethodDefToDescMap: 00007ffb201d10a0
FieldDefToDescMap: 00007ffb201d21c0
MemberRefToDescMap: 0000000000000000
FileReferencesMap: 00007ffb201d2ae0
AssemblyReferencesMap: 00007ffb201d2ae8
MetaData start address: 000001edb6b8a960 (103320 bytes)
从卦中的 Name 来看，再一次确认了 ReSharper 的问题。
2. ReSharper 是阻塞还是死锁
本着 4S 店只换不修的思路，让朋友直接卸载掉VS中的 ReSharper 肯定是没问题的，但为了兴趣继续探究下 Resharper 正在做什么？
从汇编代码看，当前正准备做两个 if 判断，而且都是 true，最后跳转到 IGeneratorHost.View 属性中，不管怎么说，这里还是不断的处理，所以我觉得这里的阻塞要么是 死循环 出不来，要么还需要再等等。
可能有些朋友好奇，Resharper 塞入到剪贴板中到底是什么数据，要想挖这个信息，可以看汇编从线程栈提取，这次就不搞这么复杂了，换个思路吧，先看 CriticalSetDataObject 方法源码，输出如下：
internal static void CriticalSetDataObject(object data, bool copy)
{
IComDataObject dataObject;
if (data is DataObject)
{
dataObject = (DataObject)data;
}
else if (data is IComDataObject)
{
SecurityHelper.DemandUnmanagedCode();
dataObject = (IComDataObject)data;
}
else
{
dataObject = new DataObject(data);
}
...
}
接下来用 !dso 看下有没有类似的 DataObject 和 IComDataObject 对象，输出如下：
0:000> !dso
OS Thread Id: 0x8144 (0)
RSP/REG Object Name
...
0000004BACAFB138 000001ed401e9ec8 System.Windows.DataObject
...
0:000> !mdt 000001ed401e9ec8
000001ed401e9ec8 (System.Windows.DataObject)
_innerData:000001ed401e9ee0 (System.Windows.DataObject+DataStore)
0:000> !mdt 000001ed401e9ee0
000001ed401e9ee0 (System.Windows.DataObject+DataStore)
_data:000001ed401e9ef8 (System.Collections.Hashtable)
0:000> !mdt 000001ed401e9ef8
000001ed401e9ef8 (System.Collections.Hashtable)
buckets:000001ed401e9f48 (System.Collections.Hashtable+bucket[], Elements: 3, ElementMT=00007ffb79123af8)
count:0x2 (System.Int32)
occupancy:0x1 (System.Int32)
loadsize:0x2 (System.Int32)
loadFactor:0.720000 (System.Single)
version:0x2 (System.Int32)
isWriterInProgress:false (System.Boolean)
keys:NULL (System.Collections.ICollection)
values:NULL (System.Collections.ICollection)
_keycomparer:NULL (System.Collections.IEqualityComparer)
_syncRoot:NULL (System.Object)
expand all 2 items
0:000> !mdt 000001ed401e9f48
000001ed401e9f48 (System.Collections.Hashtable+bucket[], Elements: 3, ElementMT=00007ffb79123af8)
expand all 3 items
0:000> !mdt -e:2 000001ed401e9f48
000001ed401e9f48 (System.Collections.Hashtable+bucket[], Elements: 3, ElementMT=00007ffb79123af8)
[0] (System.Collections.Hashtable+bucket) VALTYPE (MT=00007ffb79123af8, ADDR=000001ed401e9f58)
key:000001ed310ea6f0 (System.String) Length=11, String="UnicodeText"
val:000001ed401e9fd0 (System.Windows.DataObject+DataStore+DataStoreEntry[], Elements: 1)
hash_coll:0xf337c502 (System.Int32)
[1] (System.Collections.Hashtable+bucket) VALTYPE (MT=00007ffb79123af8, ADDR=000001ed401e9f70)
key:000001ed310eaa08 (System.String) Length=16, String="Rich Text Format"
val:000001ed401ea018 (System.Windows.DataObject+DataStore+DataStoreEntry[], Elements: 1)
hash_coll:0x30818946 (System.Int32)
[2] (System.Collections.Hashtable+bucket) VALTYPE (MT=00007ffb79123af8, ADDR=000001ed401e9f88)
key:NULL (System.Object)
val:NULL (System.Object)
hash_coll:0x0 (System.Int32)
increase depth
0:000> !mdt 000001ed401ea018
000001ed401ea018 (System.Windows.DataObject+DataStore+DataStoreEntry[], Elements: 1)
expand all 1 items
0:000> !mdt -e:2 000001ed401ea018
000001ed401ea018 (System.Windows.DataObject+DataStore+DataStoreEntry[], Elements: 1)
[0] 000001ed401e9ff0 (System.Windows.DataObject+DataStore+DataStoreEntry)
_data:000001ed401635e8 (System.String) Length=1165, String="{\rtf\ansi{\fonttbl{\f0 NSimSun;}}{\colortbl;\red0\green0\blue255;\red0\green0\blue0;\red0\green128\blue0;}\f0 \fs19 \cf1 \cb0 \highlight0 var\cf2 hostname = System.Net.Dns.GetHostName();\par System.Net.IPAddress[] hostaddrs = System.Net.Dns.GetHostAddresses(hostname);\par \par \cf1 var\cf2 localIPList = \cf1 new\cf2 List<IPAddress>();\par \cf1 for\cf2 (\cf1 int\cf2 i = 0; i < hostaddrs.Length; i++)\par \{\par \cf1 if\cf2 (System.Net.Sockets.AddressFamily.InterNetwork == hostaddrs[i].AddressFamily)\par \{\par \cf1 if\cf2 (hostaddrs[i].ToString().Equals(_localIP))\par \{\par localIPList.Insert(0, hostaddrs[i]);\cf3 //\uc1\u20248?\uc1\u20808?\uc1\u20351?\uc1\u29992?\uc1\u19978?\uc1\u27425?\uc1\u-28706?\uc1\u25509?IP\cf2 \par \}\par \cf1 else\cf2 \par \{\par localIPList.Add(hostaddrs[i]);\par \}\par \}\par \}}"
_autoConvert:true (System.Boolean)
_aspect:0x1 (System.Runtime.InteropServices.ComTypes.DVASPECT)
_index:0x0 (System.Int32)
increase depth
简单整理了下大概是这样的代码。
var hostname = System.Net.Dns.GetHostName();
System.Net.IPAddress[] hostaddrs = System.Net.Dns.GetHostAddresses(hostname);
var localIPList = new List<IPAddress>();
for (int i = 0; i < hostaddrs.Length; i++)
{
if (System.Net.Sockets.AddressFamily.InterNetwork == hostaddrs[i].AddressFamily)
{
if (hostaddrs[i].ToString().Equals(_localIP))
{
localIPList.Insert(0, hostaddrs[i]);//优先使用上次连接IP
}
else
{
localIPList.Add(hostaddrs[i]);
}
}
}
有了这些信息，还是先让朋友把 Reshaper 卸载掉看看，据朋友反馈在 Resharper 官方的 issue 里找到了解决方案，禁用了如下选项，暂时没有出现任何问题，截图如下：
三：总结
综合朋友的反馈，这次VS的卡死就是他按下了 Ctrl+C 复制这段代码的时候，Resharper 插件介入，然后在处理富文本时出问题了，不知道大家可踩过类似的坑，算是给后来人一点定位经验吧。
