Develop Reference

Hight CPU usage perf .net application

I'm trying to profile my .net application written in C# which uses 100% of cpu. Application is very big, contains tons of code, so it is impossible to provide whole project code. I tried to get threads stack for application threads that uses 25% CPU (1 core), and often i got this:
ntoskrnl.exe!KeSynchronizeExecution+0x2246
ntoskrnl.exe!KeWaitForMultipleObjects+0x135e
ntoskrnl.exe!KeWaitForMultipleObjects+0xdd9
ntoskrnl.exe!KeWaitForMutexObject+0x373
ntoskrnl.exe!KeStallWhileFrozen+0x1977
ntoskrnl.exe!_misaligned_access+0x13f9
ntoskrnl.exe!KeWaitForMultipleObjects+0x152f
ntoskrnl.exe!KeWaitForMultipleObjects+0xdd9
ntoskrnl.exe!KeWaitForMutexObject+0x373
ntoskrnl.exe!NtWaitForSingleObject+0xb2
ntoskrnl.exe!setjmpex+0x34a3
ntdll.dll!ZwWaitForSingleObject+0xa
KERNELBASE.dll!WaitForSingleObjectEx+0x98
clr.dll!GetMetaDataInternalInterface+0x25b1f
clr.dll!GetMetaDataInternalInterface+0x25ad3
clr.dll!GetMetaDataInternalInterface+0x25a92
clr.dll!GetMetaDataInternalInterface+0x39106
clr.dll!GetMetaDataInternalInterface+0x39a81
clr.dll!GetMetaDataInternalInterface+0x394ad
clr.dll!GetMetaDataInternalInterface+0x39979
clr.dll!GetMetaDataInternalInterface+0x398c1
clr.dll!GetMetaDataInternalInterface+0x3539a
clr.dll!ClrCreateManagedInstance+0x2747
KERNEL32.dll!BaseThreadInitThunk+0x22
ntdll.dll!RtlUserThreadStart+0x34
Can anyone explain to me why thread with this call stack consumes 1 core
of my CPU?
What does this 'KeSynchronizeExecution'?
How to avoid hight CPU usage in such situations?

Just trying to help you here, I am not an expert.
The ntoskrnl.exe!KeSynchronizeExecution routine synchronizes the execution of the specified routine with the interrupt service routine (ISR) that is assigned to a set of one or more interrupt objects.
The ntoskrnl.exe!KeWaitForMultipleObjects routine puts the current thread into an alertable or nonalertable wait state until any or all of a number of dispatcher objects are set to a signaled state or (optionally) until the wait times out.
The ntoskrnl.exe!KeWaitForMutexObject routine puts the current thread into an alertable or nonalertable wait state until the given mutex object is set to a signaled state or (optionally) until the wait times out.
I think ntoskrnl.exe!KEStallWhileFrozen routine is called when waits for multiple objects routines are not resolved.
ntoskrnl.exe!_misaligned_access routine is an alert when cpu cannot read misaligned data. Seems Misaligned memory accesses can incur enormous performance losses on targets that do not support them in hardware. Ref: https://msdn.microsoft.com/en-us/library/ms253949(v=vs.80).aspx. Also check the Avoiding Alignment Errors section.
ntoskrnl.exe!NtWaitForSingleObject waits until the specified object attains a state of signaled.
A call to the setjmp function saves the current instruction address as well as other CPU registers. A subsequent call to the longjmp function restores the instruction pointer and registers, and execution resumes at the point just after the setjmp call.
ntdll.dll!ZwWaitForSingleObject routine waits until the specified object attains a state of Signaled. An optional time-out can also be specified.
KERNELBASE.dll!WaitForSingleObjectEx waits until the specified object is in the signaled state, an I/O completion routine or asynchronous procedure call (APC) is queued to the thread, or the time-out interval elapses.
clr.dll!GetMetaDataInternalInterface gets a pointer to an internal interface instance that is used to read and write metadata in memory.

I used Jetbrains solution before for this. And it could really find optimization points easy. My advice use : https://www.jetbrains.com/profiler/ and find which process and methods high usage cpu. also you can find memory etc.
I know you can install it trial. Install it trial and solve your problem.

I have just saw a similar problem in my application.
I was able to profile it using PerfView application.
In my case it was lock inside Dictionary.Insert method. Trying to access dictionary from multiple threads at the same time causes infinit operations in these threads.
Looks like one CPU Core goes to 25% usage and no chance to unlock it.

Wait for ANY thread to finish, not ALL

I'm starting multiple threads and would like to know when any of then finishes. I know the following code:
foreach (Thread t in threads)
t.Join();
But it will only wait for all threads together. That's much too late. I need to know when one thread finishes, even when other threads are still running. I'm looking for something equivalent to WaitAny only for threads. But I can't add code to all threads I'm monitoring, so using signals or other synchronisation objects is not an option.
Some clarification: I'm working on a logging/tracing tool that should log the application's activity. I can insert log statements when a thread starts, but I can't insert a log statement on every possible way out of the thread (multiple exit points, exceptions etc.). So I'd like to register the new thread and then be notified when it finishes to write a log entry. I could asynchronously Join on every thread, but that means a second thread for every monitored thread which may seem a bit much overhead. Threads are used by various means, be it a BackgroundWorker, Task or pool thread. In its essence, it's a thread and I'd like to know when it's done. The exact thread mechanism is defined by the application, not the logging solution.

Instead of Threads use Tasks. It has the method WaitAny.
Task.WaitAny
As you can read here,
More efficient and more scalable use of system resources.
More programmatic control than is possible with a thread or work item.

In my opinion WaitHandle.WaitAny is the best solution, since you don't like to use it for some xyz reason you can try something like this.
Take the advantage of Thread.Join(int) method which takes millisecond timeout and returns true when thread is terminated or false when timed out.
List<Thread> threads = new List<Thread>();
while (!threads.Any(x=> x.Join(100)))
{
}
You can alter the timeout of Join If you know how long it will take.

My answer is based on your clarification that all you have is Thread.Current. Disclaimer: IMO, what you're trying to do is a hack, thus my idea by all means is a hack too.
So, use reflection to obtain the set of native Win32 handles for your desired threads. Your are looking for Thread.GetNativeHandle method which is internal, so you call it like thread.GetType().InvokeMember("GetNativeHandle", BindingFlags.InvokeMethod | BindingFlags.Instance | BindingFlags.NonPublic, ...). Use a reflection tool of your choice or Framework sources to learn more about it. Once you've got the handles, go on with one of the following options:
Set up your own implementation of SynchronizationContext (derive from it) and use SynchronizationContext.WaitHelper(waitAll: false) to wait for your unmanaged handles.
Use the raw Win32 API like WaitForMultipleObjects or CoWaitForMultipleObjects (depending on whether you need to pump messages).
Perform the wait on a separate child or pool thread.
[EDITED] Depending on the execution environment of your target threads, this hack may not work, because one-to-one mapping between managed and unmanaged threads is not guaranteed:
It is possible to determine the Windows thread that is executing the code for a managed thread and to retrieve its handle. However, it still doesn't make sense to call the SetThreadAffinityMask function for this Windows thread, because the managed scheduler can continue the execution of a managed thread in another Windows thread.
It appears however, this may be an implication only for custom CLR hosts. Also, it appears to be possible to control managed thread affinity with Thread.BeginThreadAffinity and Thread.EndThreadAffinity.

You could use a background worker for your working threads.
Then hook all the RunWorkerCompleted events to a method that will wait for them.
If you want that to be synched to the code where you're currently waiting for the join, then the problem is reduced to just synchronizing that single event method to that place in code.
Better yet, I'd suggest to do what you're doing asynchronously without blocking, and just do what you want in the event.

Would you consider wrapping your thread invocations with another 'logging' thread? That way you could log synchronously before & after the thread run.
Something like this pseudo-code:
int threadLogger(<parms>) {
log("starting thread");
retcode = ActualThreadBody(<parms>);
log("exiting thread");
return retcode;
}
If you have more information on the thread started, you could log that as well.
You could also take the thread function as a parameter in the case where you have multiple types of threads to start, which it sounds like you do.

Is there any way to work around OS loader lock deadlocks caused by third-party libraries?

I have an interesting problem that I haven't seen documented anywhere else (at least not this specific issue).
This issue is a combination of COM, VB6, and .NET and making them play nice.
Here's what I have:
A legacy VB6 ActiveX DLL (written by us)
A multi-threaded Windows service written in C# that processes requests from clients over the network and sends back results. It does this by creating a new STA thread to handle each request. Each request-handler thread instantiates a COM object (defined in the ActiveX DLL) to process the request and get the result (a string of XML is passed in, and it returns a string of XML back), explicitly releases the COM object, and exits. The service then sends the result back to the client.
All of the network code is handled using asynchronous networking (i.e. thread pool threads).
And yes, I know this is already a risky thing to be doing in the first place, since VB6 isn't very friendly with multi-threaded applications to begin with, but unfortunately it's what I am stuck with for the moment.
I've already fixed a number of things that were causing deadlocks in the code (for example, making sure the COM objects are actually created and called from a separate STA thread, making sure to explicitly release the COM objects before the thread exits to prevent deadlocks that were occurring between the garbage collector and the COM Interop code, etc.), but there is one deadlock scenario that I just can't seem to solve.
With some help from WinDbg, I was able to figure out what is happening, but I'm not sure how (or if) there is a way around this particular deadlock.
What's happening
If one request-handler thread is exiting, and another request-handler thread is starting at the same time, a deadlock can occur because of the way the VB6 runtime initialization and termination routines seem to work.
The deadlock occurs in the following scenario:
The new thread that is starting up is in the middle of creating a new instance of the (VB6) COM object to process an incoming request. At this point, the COM runtime is in the middle of a call to retrieve the object's class factory. The class factory implementation is in the VB6 runtime itself (MSVBVM60.dll). That is, its calling the VB6 runtime's DllGetClassObject function. This, in turn, calls an internal runtime function (MSVBVM60!CThreadPool::InitRuntime), which acquires a mutex and enters a critical section to do part of its work. At this point, it's about to call LoadLibrary to load oleaut32.dll into the process, while holding this mutex. So, now it's holding this internal VB6 runtime mutex and waiting for the OS loader lock.
The thread that is exiting is already running inside the loader lock, because it's done executing managed code and is executing inside the KERNEL32!ExitThread function. Specifically, it's in the middle of handling the DLL_THREAD_DETECH message for MSVBVM60.dll on that thread, which in turn calls a method to terminate the VB6 runtime on the thread (MSVBVM60!CThreadPool::TerminateRuntime). Now, this thread tries to acquire the same mutex that the other thread being initialized already has.
A classic deadlock. Thread A has L1 and wants L2, but Thread B has L2 and needs L1.
The problem (if you've followed me this far) is I don't have any control over what the VB6 runtime is doing in its internal thread initialization and teardown routines.
In theory, if I could force the VB6 runtime initialization code to run inside the OS loader lock, I would prevent the deadlock, because I am fairly certain the mutex the VB6 runtime is holding is specifically only used inside the initialization and termination routines.
Requirements
I can't make the COM calls from a single STA thread, because then the service won't be able to handle concurrent requests. I can't have a long-running request block other client requests either. This is why I create one STA thread per-request.
I need to create a new instance of the COM object on each thread, because I need to make sure each instance has its own copy of global variables in the VB6 code (VB6 gives each thread its own copy of all global variables).
Solutions I've tried that didn't work
Converted ActiveX DLL to ActiveX EXE
First, I tried the obvious solution and created an ActiveX EXE (out-of-process server) to handle the COM calls. Initially, I compiled it so that a new ActiveX EXE (process) was created for each incoming request, and I also tried it with the Thread Per Object compile option (one process instance is created, and it creates each object on a new thread within the ActiveX EXE).
This fixes the deadlock issue with respect to the VB6 runtime, because the VB6 runtime never gets loaded into the .NET code proper. However, this led to a different problem: if concurrent requests come into the service, the ActiveX EXE tends to fail randomly with RPC_E_SERVERFAULT errors. I assume this is because the COM marshalling and/or the VB6 runtime can't deal with concurrent object creation/destruction, or concurrent method calls, inside the ActiveX EXE.
Force the VB6 code to run inside the OS loader lock
Next, I switched back to using an ActiveX DLL for the COM class. To force the VB6 runtime to run its thread initialization code inside the OS loader lock, I created a native (Win32) C++ DLL, with code to handle DLL_THREAD_ATTACH in DllMain. The DLL_THREAD_ATTACH code calls CoInitialize and then instantiates a dummy VB6 class to force the VB6 runtime to be loaded and force the runtime initialization routine to run on the thread.
When the Windows service starts, I use LoadLibrary to load this C++ DLL into memory, so that any threads created by the service will execute that DLL's DLL_THREAD_ATTACH code.
The problem is that this code runs for every thread the service creates, including the .NET garbage collector thread and the thread-pool threads used by the async networking code, which doesn't end well (this just seems to cause the threads to never start properly, and I imagine initializing COM on the GC and thread-pool threads is in general just a very bad idea).
Addendum
I just realized why this is a bad idea (and probably part of the reason it didn't work): it isn't safe to call LoadLibrary when you are holding the loader lock. See Remarks section in this MSDN article: http://msdn.microsoft.com/en-us/library/ms682583%28VS.85%29.aspx, specifically:
Threads in DllMain hold the loader lock so no additional DLLs can be dynamically loaded or initialized.
Is there any way to workaround these issues?
So, my question is, is there any way to work around the original deadlock issue?
The only other thing I can think of is to create my own lock object and surround the code that instantiates the COM object in a .NET lock block, but then I have no way (that I know of) to put the same lock around the (operating system's) thread exit code.
Is there a more obvious solution to this issue, or am I plain out of luck here?

As long as all of your modules work in one process, you can hook Windows API by replacing some system calls with your wrappers. Then, you can wrap the calls in a single critical section to avoid deadlock.
There are several libraries and samples to achieve that, the technique is commonly known as detouring:
http://www.codeproject.com/Articles/30140/API-Hooking-with-MS-Detours
http://research.microsoft.com/en-us/projects/detours/
And of course the implementation of wrappers should be done in native code, preferably C++. .NET detours work too for high-level API functions such as MessageBox, but if you try to reimplement LoadLibrary API call in .NET then you may get a cyclic dependency issue because .NET runtime internally uses LoadLibrary function during execution and does this often.
So the solution looks like this to me: a separate .DLL module which is loaded at the very start of your application. The module fixes the deadlock problem by patching several VB and Windows API calls with your own wrappers. All wrappers do one thing: wrap the call in critical section and invoke the original API function to do the real job.

EDIT: in retrospect, I don't think this will work. The problem is that the deadlock can occur at any time that a Win32 thread exits, and since Win32 threads don't map 1:1 to .NET threads, we can't (within .NET) force Win32 threads to acquire the lock before exiting. In addition to the possibility of the .NET thread that is exiting being switched to a different OS thread, there are presumably OS threads not associated with any .NET thread (garbage collection and the like) which may start and exit at random.
The only other thing I can think of is to create my own lock object
and surround the code that instantiates the COM object in a .NET lock
block, but then I have no way (that I know of) to put the same lock
around the (operating system's) thread exit code.
That sounds like a promising approach. I gather from this that you
are able to modify the service's code, and you say each thread
explicitly releases the COM object before exiting, so presumably you
could claim a lock at this point, either just before explicitly
releasing the COM object or just after. The secret is to choose a
type of lock that is implicitly released once the thread holding it
has exited, such as a Win32 mutex.
It is likely that a Win32 mutex object does not become abandoned until
the thread has completed all DLL_THREAD_DETACH calls, although I don't
know whether this behaviour is documented. I'm not familiar with
locking in .NET but my guess is that they are unlikely to be suitable,
because even if the right kind of lock exists, it would be likely to
be considered abandoned as soon as the thread reaches the end of the
managed code section, i.e., before the calls to DLL_THREAD_DETACH.
If Win32 mutex objects don't do the trick (or if you very reasonably
prefer not to rely on undocumented behaviour) you might need to
implement the lock yourself. One way to do this would be to use
OpenThread to get a handle to the current thread and save this in your
lock object, along with an event or similar object. If the lock has
been claimed and you want to wait for it to be available, use
WaitForMultipleObjects to wait until either the thread handle or the
event is signaled. If the event is signaled this means the lock has
been explicitly released, if the thread handle is signaled it was
implicitly released by the thread exiting. Obviously implementing
this involves a lot of tricky details (for example: when a thread
explicitly releases the lock, you can't close the thread handle
because another thread might be waiting on it, so you'll have to close
it when the lock is next claimed instead) but it shouldn't be too
difficult to sort these out.

I don't see any reason why you couldn't load an extra instance of the ActiveX control in your startup code and just hang onto the reference. Presto, no more loader lock issues since the VB6 runtime never shuts down.

Since I'm still exploring my options, I wanted to still see if I could implement a solution in pure .NET code without using any native code, for the sake of simplicity. I'm not sure if this is a fool-proof solution yet, because I'm still trying to figure out whether it actually gives me the mutual exclusion I need, or if it just looks like it does.
Any thoughts or comments are welcome.
The relevant part of the code is below. Some notes:
The HandleRpcRequest method is called from a thread-pool thread when a new message is received from a remote client
This fires off a separate STA thread so that it can make the COM call safely
DbRequestProxy is a thin wrapper class around the real COM class I'm using
I used a ManualResetEvent (_safeForNewThread) to provide the mutual exclusion. The basic idea is that this event stays unsignaled (blocking other threads) if any one particular thread is about to exit (and hence potentially about to terminate the VB6 runtime). The event is only signaled again after the current thread completely terminates (after the Join call finishes). This way multiple request-handler threads can still execute concurrently unless an existing thread is exiting.
So far, I think this code is correct and guarantees that two threads can't deadlock in the VB6 runtime initialization/termination code anymore, while still allowing them to execute concurrently for most of their execution time, but I could be missing something here.
public class ClientHandler {
private static ManualResetEvent _safeForNewThread = new ManualResetEvent(true);
private void HandleRpcRequest(string request)
{
Thread rpcThread = new Thread(delegate()
{
DbRequestProxy dbRequest = null;
try
{
Thread.BeginThreadAffinity();
string response = null;
// Creates a COM object. The VB6 runtime initializes itself here.
// Other threads can be executing here at the same time without fear
// of a deadlock, because the VB6 runtime lock is re-entrant.
dbRequest = new DbRequestProxy();
// Call the COM object
response = dbRequest.ProcessDBRequest(request);
// Send response back to client
_messenger.Send(Messages.RpcResponse(response), true);
}
catch (Exception ex)
{
_messenger.Send(Messages.Error(ex.ToString()));
}
finally
{
if (dbRequest != null)
{
// Force release of COM objects and VB6 globals
// to prevent a different deadlock scenario with VB6
// and the .NET garbage collector/finalizer threads
dbRequest.Dispose();
}
// Other request threads cannot start right now, because
// we're exiting this thread, which will detach the VB6 runtime
// when the underlying native thread exits
_safeForNewThread.Reset();
Thread.EndThreadAffinity();
}
});
// Make sure we can start a new thread (i.e. another thread
// isn't in the middle of exiting...)
_safeForNewThread.WaitOne();
// Put the thread into an STA, start it up, and wait for
// it to end. If other requests come in, they'll get picked
// up by other thread-pool threads, so we won't usually be blocking anyone
// by doing this (although we are blocking a thread-pool thread, so
// hopefully we don't block for *too* long).
rpcThread.SetApartmentState(ApartmentState.STA);
rpcThread.Start();
rpcThread.Join();
// Since we've joined the thread, we know at this point
// that any DLL_THREAD_DETACH notifications have been handled
// and that the underlying native thread has completely terminated.
// Hence, other threads can safely be started.
_safeForNewThread.Set();
}
}

I had written a rather complex code using VB6,VC6 about 20 years ago and I need to port it to visual studio.net.
I simply took the functions as I had written them along with the header files corrected all the compile errors (which were MANY) and then tried to load it. got "loaderlock closed"
I then decided to redo all the files starting from those that few other files depended upon and then worked my way up and as I went I included only the header files that that particular file required. The result it loads now just fine. no more loaderlock closed.
the lesson for me is don't include any more header files in a particular cpp file than is absolutely necessary.
hope this helps
from a very happy camper!!
david

Threading issues in C# from external process

I have this simple code:
public void Run()
{
var invokerThread = new Thread(new ThreadStart(RunOnBackground));
invokerThread.Start();
}
private void RunOnBackground()
{
Trace.WriteLine("hi");
...
}
Unfortunately when running this code (from third party process) the thread doesn't really run.
Either in process explorer and in VS debugger I see that the thread is created and its state is "Running".
The main thread's apartment is STA and I've tried both STA and MTA on internal thread.
When I add to the Run() method at the end invokerThread.Join(); then the thread does run. But then again it doesn't really help.
What am I missing?
Edit: Here is some more information regarding the code hosting -
Run() method is called via COM interop from a process which is also managed executable assembly (the reason COM interop is used is because all other components in the system are native).
The method RunOnBackground() includes some more code after the tracing and generally its execution lasts between 10 - 20 seconds, including starting another process and waiting for its termination. Also I have some other areas in the code where I write some debug information to the Trace. While debugging the code, Run() runs as usual and after invokerThread.Start(); invokerThread's state is "Running" (though breakpoints inside the RunOnBackground() method don't stop).
When I add invokerThread.Join() at the end of the Run() method the debugger goes to RunOnBackground() after the Join().

There's some crucial information missing about what RunOnBackground() really does. This is otherwise a good match for what happens when you use apartment threaded COM objects on a worker thread. COM automatically marshals any method call on such an object from the worker thread to the STA thread on which it was created.
That can only work well when the STA thread observes STA threading requirements. It must pump a message loop and cannot block. Breaking those rules makes deadlock very likely, the worker thread call cannot complete until the STA thread dispatches the marshaled call. A sure sign that this is what is going on is seeing Thread.Join() solve the problem. It pumps a message loop internally in the CLR when it is called on an STA thread.
To diagnose this, you'll need Debug + Windows + Threads to see what that worker thread is blocking on. If my guess is right, it will be buried deep inside of the COM plumbing code, waiting for the marshaled call to complete. You can only see this by enabling unmanaged code debugging and setting up the Microsoft Symbol Server so you get debugging symbols for the plumbing code and get a reliable stack trace.
Fixing this is going to be difficult. You cannot magically flip a switch and make code run on a thread when it has explicitly stated that it doesn't support multi-threading. It is imperative that you create the instance of the COM object on the same thread that calls its methods. And that thread has to be an STA thread. Check this sample code for the approach. If you don't control the creation of the COM object then you're stuck.

I may say something stupid, but here is what I saw in MSDN Threads.
Look at the examples section at the end.
The output of the example is interesting, you can see there that the Thread created and started only starts executing when the main thread does a Sleep(0) or a Thread.Join().
It seems to be what exactly happens to you, doesn't it ?
Maybe try with a Sleep(0) on your main thread to really launch your working Thread.
Another workaround would be to use the BackGroundWorker.
As its name says it, it works on the Background and is really easy to use. It may be of much use to you.

What happens if an asynchronous delegate call never returns?

I found a decent looking example of how to call a delegate asynchronously with a timeout... http://www.eggheadcafe.com/tutorials/aspnet/847c94bf-4b8d-4a66-9ae5-5b61f049019f/basics-make-any-method-c.aspx. In summary it uses WaitOne with a timeout to determine if the call does not return before the timeout expires.
I also know that you should have an EndInvoke to match each BeginInvoke.
So what happens if the wait timeout expires? We (presumably) DON'T want to call EndInvoke as that will block. The code can go on to do 'other things', but have we leaked anything? Is there some poor thread someplace blocked waiting for a return that's never going to happen? Have we leaked some memory where the result-that-will-never-return was going to be placed?

I think this post talks about it very well:
From the post:
You can't terminate an executing async delegate if it's not your thread, but you can if it is. If you use the common BeginInvoke type methods, you get a thread pool thread managed by the framework. If you use the Thread() class you get you own thread to manage, start, suspend, etc. as you like.
Developing asychronously requires that one decide who will manage the threads. The many different methods that execute asynchronously are using the ThreadPool threads behind the scenes.
Since you can't/shouldn't terminate a thread pool thread then you must design you code to communicate with the thread so that it can exit. The MSDN examples for the BackgroundWorker component demonstrates this kind of communication.
Sometimes your code may have the thread blocking waiting for IO. Here you would normally use a multiple object wait instead, waiting for IO or for a ManualResetEvent.
So in short, youll need to find a way to manage the threads yourself if there is a possibility of timing out and you want the thread to end.

You need to call EndInvoke().
Here is a link talking about what happens with EndInvoke():
Is EndInvoke() optional, sort-of optional, or definitely not optional?
Here is a link to the article in the accepted answer.
We had all been talking about the 'fire and forget' technique with asynchronous delegate invocation in various public forums. Many DevelopMentor instructors had written articles and example code showing the technique, and we had all described it in class. And of course it was in Don's book by then too. So when Microsoft eventually remembered to let the outside world know that this technique is not in fact legal, it was rather astonishing.
An MSDN link on the asynchronous pattern.

You will leak the resources held by the thread. There will be various bits of .NET remoting plumbing objects like the AsyncResult. Several unmanaged handles associated with the thread. All peanuts compared to the one megabyte of virtual memory address space you'll leak, held by the thread stack.
You cannot abort the thread in any way, the leak is permanent. When you have to deal with badly behaving code like this, your only good resource is to run it in a separate process so you can get Windows to clean up the shrapnel when you shoot the process in the head with Process.Kill(). Even that is not guaranteed, these kind of freezes tend to be associated with misbehaving device drivers. Process.Kill won't terminate a device driver thread. Easy to see: trying to abort the process with Taskmgr.exe will leave it running with one Handle. You have some hope if that doesn't happen.