我发现现有的解释太官样文章了。下面是我的解释:

STA: If a thread creates a COM object that's set to STA (when calling CoCreateXXX you can pass a flag that sets the COM object to STA mode), then only this thread can access this COM object (that's what STA means - Single Threaded Apartment), other thread trying to call methods on this COM object is under the hood silently turned into delivering messages to the thread that creates(owns) the COM object. This is very much like the fact that only the thread that created a UI control can access it directly. And this mechanism is meant to prevent complicated lock/unlock operations.

MTA: 如果一个线程创建了一个设置为MTA的COM对象，那么几乎每个线程都可以直接调用该对象上的方法。

这差不多就是要点了。虽然从技术上讲，还有一些细节我没有提到，比如在“STA”段落中，创造者线程本身必须是STA。但这几乎是你了解STA/MTA/NA所需要知道的全部内容。

每个承载COM或OLE控件的EXE都定义了它的单元状态。单元状态默认为STA(对于大多数程序应该是STA)。

STA -所有OLE控件必须位于STA中。STA意味着你的com对象必须总是在UI线程上操作，不能传递给其他线程(就像MFC中的任何UI元素一样)。但是，您的程序仍然可以有许多线程。

MTA -你可以在程序中的任何线程上操作COM对象。

调用COM对象dll的代码(例如，读取私有数据文件)可能在用户界面中正常工作，但却神秘地挂在服务上。原因是。net 2.0的用户界面假设是STA(线程安全的)，而服务假设是MTA((在此之前，服务假设是STA)。必须为服务中的每个COM调用创建一个STA线程会增加大量开销。

Side note: If you are using some PowerShell 2.0 snap-ins, you need to launch PowerShell version 3 or greater with -MTA option to use them. PowerShell 2 apartment model is MTA versus later versions use STA as default. Other point is bitness. Normal calls in apartment are not marshalled (direct calls), so if your caller is x64 then callee must be also x64. Only way around this is to use remote procedure call (RPC), which add huge amount of overhead (spawn a new 32-bit process to load snap-in DLL and query result by some means).For developer: always publish type library - it make your COM object discovery and usage much easier! Every interface should be public and unique - implementation can be proprietary or open source.

另一种情况

例子:

 IStorage_vtbl** reference; // you got it by some means of factory
 
 
 public unsafe int OpenStorage(char* pwcsName, IStorage pstgPriority, uint grfMode, char** snbExclude, uint reserved, IStorage* ppstg)
 {
     IStorage_vtbl** @this = (IStorage_vtbl**)reference;
     IStorage_vtbl* vtbl = *@this;
     if (vtbl == null)
         throw new InvalidComObjectException();
     Delegate genericDelegate = Marshal.GetDelegateForFunctionPointer(vtbl->method_6, typeof(delegate_6));
     delegate_6 method = (delegate_6)genericDelegate;
     return method(@this, pwcsName, pstgPriority, grfMode, snbExclude, reserved, ppstg);
 }
 

这段代码只是为COM子系统的实际调用添加了“This”指针 那么，这个调用是打开IStorage STA或MTA的实例吗?

我发现现有的解释太官样文章了。下面是我的解释:

STA: If a thread creates a COM object that's set to STA (when calling CoCreateXXX you can pass a flag that sets the COM object to STA mode), then only this thread can access this COM object (that's what STA means - Single Threaded Apartment), other thread trying to call methods on this COM object is under the hood silently turned into delivering messages to the thread that creates(owns) the COM object. This is very much like the fact that only the thread that created a UI control can access it directly. And this mechanism is meant to prevent complicated lock/unlock operations.

MTA: 如果一个线程创建了一个设置为MTA的COM对象，那么几乎每个线程都可以直接调用该对象上的方法。

这差不多就是要点了。虽然从技术上讲，还有一些细节我没有提到，比如在“STA”段落中，创造者线程本身必须是STA。但这几乎是你了解STA/MTA/NA所需要知道的全部内容。

根据我的理解，“Apartment”是用来保护COM对象不受多线程问题的影响。

If a COM object is not thread-safe, it should declare it as a STA object. Then only the thread who creates it can access it. The creation thread should declare itself as a STA thread. Under the hood, the thread stores the STA information in its TLS(Thread Local Storage). We call this behavior as that the thread enters a STA apartment. When other threads want to access this COM object, it should marshal the access to the creation thread. Basically, the creation thread uses messages mechanism to process the in-bound calls.

如果一个COM对象是线程安全的，它应该声明为一个MTA对象。MTA对象可以被多线程访问。