C# SemaphoreSlim 实现

当多个任务或线程并行运行时，难以避免的对某些有限的资源进行并发的访问。可以考虑使用信号量来进行这方面的控制（System.Threading.Semaphore）是表示一个Windows内核的信号量对象。如果预计等待的时间较短，可以考虑使用SemaphoreSlim，它则带来的开销更小。.NetFrameWork中的信号量通过跟踪进入和离开的任务或线程来协调对资源的访问。信号量需要知道资源的最大数量，当一个任务进入时，资源计数器会被减1，当计数器为0时，如果有任务访问资源，它会被阻塞，直到有任务离开为止。
如果需要有跨进程或AppDomain的同步时，可以考虑使用Semaphore。Semaphore是取得的Windows 内核的信号量，所以在整个系统中是有效的。它主要的接口是Release和WaitOne，使用的方式和SemaphoreSlim是一致的。
信号量Semaphore是另外一个CLR中的内核同步对象。在.net中，类Semaphore封装了这个对象。与标准的排他锁对象（Monitor，Mutex，SpinLock）不同的是，它不是一个排他的锁对象，它与SemaphoreSlim，ReaderWriteLock等一样允许多个有限的线程同时访问共享内存资源。

Semaphore就好像一个栅栏，有一定的容量，当里面的线程数量到达设置的最大值时候，就没有线程可以进去。然后，如果一个线程工作完成以后出来了，那下一个线程就可以进去了。Semaphore的WaitOne或Release等操作分别将自动地递减或者递增信号量的当前计数值。当线程试图对计数值已经为0的信号量执行WaitOne操作时，线程将阻塞直到计数值大于0。在构造Semaphore时，最少需要2个参数。信号量的初始容量和最大的容量。

Semaphore的WaitOne或者Release方法的调用大约会耗费1微秒的系统时间，而优化后的SemaphoreSlim则需要大致四分之一微秒。在计算中大量频繁使用它的时候SemaphoreSlim还是优势明显，加上SemaphoreSlim还丰富了不少接口，更加方便我们进行控制，所以在4.0以后的多线程开发中，推荐使用SemaphoreSlim。SemaphoreSlim的实现如下：

public class SemaphoreSlim : IDisposable
{
private volatile int m_currentCount; //可用数的资源数，<=0开始阻塞
private readonly int m_maxCount;
private volatile int m_waitCount; //阻塞的线程数
private object m_lockObj;
private volatile ManualResetEvent m_waitHandle;
private const int NO_MAXIMUM = Int32.MaxValue;
//Head of list representing asynchronous waits on the semaphore.
private TaskNode m_asyncHead;
// Tail of list representing asynchronous waits on the semaphore.
private TaskNode m_asyncTail;
// A pre-completed task with Result==true
private readonly static Task<bool> s_trueTask =
new Task<bool>(false, true, (TaskCreationOptions)InternalTaskOptions.DoNotDispose, default(CancellationToken));

public SemaphoreSlim(int initialCount) : this(initialCount, NO_MAXIMUM){ }
public SemaphoreSlim(int initialCount, int maxCount)
{
if (initialCount < 0 || initialCount > maxCount)
{
throw new ArgumentOutOfRangeException("initialCount", initialCount, GetResourceString("SemaphoreSlim_ctor_InitialCountWrong"));
}
if (maxCount <= 0)
{
throw new ArgumentOutOfRangeException("maxCount", maxCount, GetResourceString("SemaphoreSlim_ctor_MaxCountWrong"));
}
m_maxCount = maxCount;
m_lockObj = new object();
m_currentCount = initialCount;
}
public void Wait(){Wait(Timeout.Infinite, new CancellationToken());}
public bool Wait(int millisecondsTimeout, CancellationToken cancellationToken)
{
CheckDispose();
if (millisecondsTimeout < -1)
{
throw new ArgumentOutOfRangeException("totalMilliSeconds", millisecondsTimeout, GetResourceString("SemaphoreSlim_Wait_TimeoutWrong"));
}
cancellationToken.ThrowIfCancellationRequested();
uint startTime = 0;
if (millisecondsTimeout != Timeout.Infinite && millisecondsTimeout > 0)
{
startTime = TimeoutHelper.GetTime();
}

bool waitSuccessful = false;
Task<bool> asyncWaitTask = null;
bool lockTaken = false;

CancellationTokenRegistration cancellationTokenRegistration = cancellationToken.InternalRegisterWithoutEC(s_cancellationTokenCanceledEventHandler, this);
try
{
SpinWait spin = new SpinWait();
while (m_currentCount == 0 && !spin.NextSpinWillYield)
{
spin.SpinOnce();
}
try { }
finally
{
Monitor.Enter(m_lockObj, ref lockTaken);
if (lockTaken)
{
m_waitCount++;
}
}

// If there are any async waiters, for fairness we'll get in line behind
if (m_asyncHead != null)
{
Contract.Assert(m_asyncTail != null, "tail should not be null if head isn't");
asyncWaitTask = WaitAsync(millisecondsTimeout, cancellationToken);
}
// There are no async waiters, so we can proceed with normal synchronous waiting.
else
{
// If the count > 0 we are good to move on.
// If not, then wait if we were given allowed some wait duration
OperationCanceledException oce = null;
if (m_currentCount == 0)
{
if (millisecondsTimeout == 0)
{
return false;
}
// Prepare for the main wait...
// wait until the count become greater than zero or the timeout is expired
try
{
waitSuccessful = WaitUntilCountOrTimeout(millisecondsTimeout, startTime, cancellationToken);
}
catch (OperationCanceledException e) { oce = e; }
}

Contract.Assert(!waitSuccessful || m_currentCount > 0, "If the wait was successful, there should be count available.");
if (m_currentCount > 0)
{
waitSuccessful = true;
m_currentCount--;
}
else if (oce != null)
{
throw oce;
}
if (m_waitHandle != null && m_currentCount == 0)
{
m_waitHandle.Reset();
}
}
}
finally
{
// Release the lock
if (lockTaken)
{
m_waitCount--;
Monitor.Exit(m_lockObj);
}

// Unregister the cancellation callback.
cancellationTokenRegistration.Dispose();
}
return (asyncWaitTask != null) ? asyncWaitTask.GetAwaiter().GetResult() : waitSuccessful;
}

private bool WaitUntilCountOrTimeout(int millisecondsTimeout, uint startTime, CancellationToken cancellationToken)
{
int remainingWaitMilliseconds = Timeout.Infinite;
//Wait on the monitor as long as the count is zero
while (m_currentCount == 0)
{
// If cancelled, we throw. Trying to wait could lead to deadlock.
cancellationToken.ThrowIfCancellationRequested();
if (millisecondsTimeout != Timeout.Infinite)
{
remainingWaitMilliseconds = TimeoutHelper.UpdateTimeOut(startTime, millisecondsTimeout);
if (remainingWaitMilliseconds <= 0)
{
// The thread has expires its timeout
return false;
}
}
// ** the actual wait **
if (!Monitor.Wait(m_lockObj, remainingWaitMilliseconds))
{
return false;
}
}
return true;
}
public Task<bool> WaitAsync(int millisecondsTimeout, CancellationToken cancellationToken)
{
CheckDispose();
// Validate input
if (millisecondsTimeout < -1)
{
throw new ArgumentOutOfRangeException("totalMilliSeconds", millisecondsTimeout, GetResourceString("SemaphoreSlim_Wait_TimeoutWrong"));
}
// Bail early for cancellation
if (cancellationToken.IsCancellationRequested)
return Task.FromCancellation<bool>(cancellationToken);

lock (m_lockObj)
{
// If there are counts available, allow this waiter to succeed.
if (m_currentCount > 0)
{
--m_currentCount;
if (m_waitHandle != null && m_currentCount == 0) m_waitHandle.Reset();
return s_trueTask;
}
// If there aren't, create and return a task to the caller.
// The task will be completed either when they've successfully acquired
// the semaphore or when the timeout expired or cancellation was requested.
else
{
Contract.Assert(m_currentCount == 0, "m_currentCount should never be negative");
var asyncWaiter = CreateAndAddAsyncWaiter();
return (millisecondsTimeout == Timeout.Infinite && !cancellationToken.CanBeCanceled) ?
asyncWaiter :
WaitUntilCountOrTimeoutAsync(asyncWaiter, millisecondsTimeout, cancellationToken);
}
}
}

/// <summary>Creates a new task and stores it into the async waiters list.</summary>
/// <returns>The created task.</returns>
private TaskNode CreateAndAddAsyncWaiter()
{
Contract.Assert(Monitor.IsEntered(m_lockObj), "Requires the lock be held");
// Create the task
var task = new TaskNode();
// Add it to the linked list
if (m_asyncHead == null)
{
Contract.Assert(m_asyncTail == null, "If head is null, so too should be tail");
m_asyncHead = task;
m_asyncTail = task;
}
else
{
Contract.Assert(m_asyncTail != null, "If head is not null, neither should be tail");
m_asyncTail.Next = task;
task.Prev = m_asyncTail;
m_asyncTail = task;
}
// Hand it back
return task;
}

private async Task<bool> WaitUntilCountOrTimeoutAsync(TaskNode asyncWaiter, int millisecondsTimeout, CancellationToken cancellationToken)
{
Contract.Assert(asyncWaiter != null, "Waiter should have been constructed");
Contract.Assert(Monitor.IsEntered(m_lockObj), "Requires the lock be held");
using (var cts = cancellationToken.CanBeCanceled ?
CancellationTokenSource.CreateLinkedTokenSource(cancellationToken, default(CancellationToken)) :
new CancellationTokenSource())
{
var waitCompleted = Task.WhenAny(asyncWaiter, Task.Delay(millisecondsTimeout, cts.Token));
if (asyncWaiter == await waitCompleted.ConfigureAwait(false))
{
cts.Cancel(); // ensure that the Task.Delay task is cleaned up
return true; // successfully acquired
}
}

// If we get here, the wait has timed out or been canceled.

// If the await completed synchronously, we still hold the lock.  If it didn't,
// we no longer hold the lock.  As such, acquire it.
lock (m_lockObj)
{
// Remove the task from the list.  If we're successful in doing so,
// we know that no one else has tried to complete this waiter yet,
// so we can safely cancel or timeout.
if (RemoveAsyncWaiter(asyncWaiter))
{
cancellationToken.ThrowIfCancellationRequested(); // cancellation occurred
return false; // timeout occurred
}
}

// The waiter had already been removed, which means it's already completed or is about to
// complete, so let it, and don't return until it does.
return await asyncWaiter.ConfigureAwait(false) await asyncWaiter.ConfigureAwait(false);
}
public int Release(){ return Release(1);}

public int Release(int releaseCount)
{
CheckDispose();

// Validate input
if (releaseCount < 1)
{
throw new ArgumentOutOfRangeException( "releaseCount", releaseCount, GetResourceString("SemaphoreSlim_Release_CountWrong"));
}
int returnCount;

lock (m_lockObj)
{
// Read the m_currentCount into a local variable to avoid unnecessary volatile accesses inside the lock.
int currentCount = m_currentCount;
returnCount = currentCount;

// If the release count would result exceeding the maximum count, throw SemaphoreFullException.
if (m_maxCount - currentCount < releaseCount)
{
throw new SemaphoreFullException();
}

// Increment the count by the actual release count
currentCount += releaseCount;

// Signal to any synchronous waiters
int waitCount = m_waitCount;
if (currentCount == 1 || waitCount == 1)
{
Monitor.Pulse(m_lockObj);
}
else if (waitCount > 1)
{
Monitor.PulseAll(m_lockObj);
}

// Now signal to any asynchronous waiters, if there are any.  While we've already
// signaled the synchronous waiters, we still hold the lock, and thus
// they won't have had an opportunity to acquire this yet.  So, when releasing
// asynchronous waiters, we assume that all synchronous waiters will eventually
// acquire the semaphore.  That could be a faulty assumption if those synchronous
// waits are canceled, but the wait code path will handle that.
if (m_asyncHead != null)
{
Contract.Assert(m_asyncTail != null, "tail should not be null if head isn't null");
int maxAsyncToRelease = currentCount - waitCount;
while (maxAsyncToRelease > 0 && m_asyncHead != null)
{
--currentCount;
--maxAsyncToRelease;

// Get the next async waiter to release and queue it to be completed
var waiterTask = m_asyncHead;
RemoveAsyncWaiter(waiterTask); // ensures waiterTask.Next/Prev are null
QueueWaiterTask(waiterTask);
}
}
m_currentCount = currentCount;

// Exposing wait handle if it is not null
if (m_waitHandle != null && returnCount == 0 && currentCount > 0)
{
m_waitHandle.Set();
}
}

// And return the count
return returnCount;
}

///Removes the waiter task from the linked list.</summary>
private bool RemoveAsyncWaiter(TaskNode task)
{
Contract.Requires(task != null, "Expected non-null task");
Contract.Assert(Monitor.IsEntered(m_lockObj), "Requires the lock be held");

// Is the task in the list?  To be in the list, either it's the head or it has a predecessor that's in the list.
bool wasInList = m_asyncHead == task || task.Prev != null;

// Remove it from the linked list
if (task.Next != null) task.Next.Prev = task.Prev;
if (task.Prev != null) task.Prev.Next = task.Next;
if (m_asyncHead == task) m_asyncHead = task.Next;
if (m_asyncTail == task) m_asyncTail = task.Prev;
Contract.Assert((m_asyncHead == null) == (m_asyncTail == null), "Head is null iff tail is null");

// Make sure not to leak
task.Next = task.Prev = null;

// Return whether the task was in the list
return wasInList;
}
private static void QueueWaiterTask(TaskNode waiterTask)
{
ThreadPool.UnsafeQueueCustomWorkItem(waiterTask, forceGlobal: false);
}
public int CurrentCount
{
get { return m_currentCount; }
}
public WaitHandle AvailableWaitHandle
{
get
{
CheckDispose();
if (m_waitHandle != null)
return m_waitHandle;
lock (m_lockObj)
{
if (m_waitHandle == null)
{
m_waitHandle = new ManualResetEvent(m_currentCount != 0);
}
}
return m_waitHandle;
}
}
private sealed class TaskNode : Task<bool>, IThreadPoolWorkItem
{
internal TaskNode Prev, Next;
internal TaskNode() : base() {}

[SecurityCritical]
void IThreadPoolWorkItem.ExecuteWorkItem()
{
bool setSuccessfully = TrySetResult(true);
Contract.Assert(setSuccessfully, "Should have been able to complete task");
}

[SecurityCritical]
void IThreadPoolWorkItem.MarkAborted(ThreadAbortException tae) { /* nop */ }
}
}

SemaphoreSlim类有几个私有字段很重要，m_currentCount表示可用资源，如果m_currentCount>0每次调用Wait都会减1，当m_currentCount<=0时再次调用Wait方法就会阻塞。每次调用Release方法m_currentCount都会加1.m_maxCount表示最大可用资源数，是在构造函数中指定的。m_waitCount表示当前阻塞的线程数。TaskNode m_asyncHead，m_asyncTail这2个变量主要用于异步方法。

我们首先来看看Wait方法，这里还有它的异步版本WaitAsync。在Wait方法中首先检查m_currentCount是否为0，如果是我们用SpinWait自旋10次；任意一次Wait都需要锁住m_lockObj对象，m_asyncHead != null表示当前已经存在异步的对象，所以我们调用WaitAsync方法，如果没有那么我们调用WaitUntilCountOrTimeout方法，该方法在m_currentCount==0会阻塞到到m_currentCount不为0或者超时；看到WaitUntilCountOrTimeout方法中【if (!Monitor.Wait(m_lockObj, remainingWaitMilliseconds))】，就很明了Wait方法中【CancellationTokenRegistration cancellationTokenRegistration = cancellationToken.InternalRegisterWithoutEC(s_cancellationTokenCanceledEventHandler, this)】存在的原因了，确实很巧妙【这里和ManualResetEventSlim相似】。现在我们回到WaitAsync方法，该方法也是首先检查m_currentCount是否大于0，大于直接返回。否者调用CreateAndAddAsyncWaiter创建一个Task<bool>【Task<bool>是一个链表结构】，如果没有取消且超时大于-1，那么就调用WaitUntilCountOrTimeoutAsync方法，该方法首先包装一个Task【var waitCompleted = Task.WhenAny(asyncWaiter, Task.Delay(millisecondsTimeout, cts.Token))】然后等待线程【await waitCompleted.ConfigureAwait(false)】返回的是asyncWaiter或者另一个Delay的Task。如果返回的不是asyncWaiter说明已经超时需要调用RemoveAsyncWaiter，然后返回 await asyncWaiter.ConfigureAwait(false)，如果返回的是asyncWaiter，那么就调用Cancel方法。那么这里的asyncWaiter.ConfigureAwait(false)什么时候退出了【或者说不阻塞】，这就要看Release中的QueueWaiterTask方法了。

QueueWaiterTask方法或调用TaskNode的ExecuteWorkItem方法。
那现在我们来看看Release方法，该方法会把currentCount加1，然后把等待线程转为就绪线程【Monitor.Pulse(m_lockObj)或 Monitor.PulseAll(m_lockObj)】，如果存在异步的话，看看还可以释放几个异步task【 int maxAsyncToRelease = currentCount - waitCount】，这里Release的注释很重要，只是没怎么明白，现释同步的waiters，然后在释放异步的waiters，但是释放同步后锁的资源没有释放，在释放异步的waiters时候是把currentCount减1，这样感觉异步waiters优先获取资源。也不知道我的理解是否正确？
1）当ConfigureAwait(true)，代码由同步执行进入异步执行时，当前同步执行的线程上下文信息（比如HttpConext.Current，Thread.CurrentThread.CurrentCulture）就会被捕获并保存至SynchronizationContext中，供异步执行中使用，并且供异步执行完成之后（await之后的代码）的同步执行中使用（虽然await之后是同步执行的，但是发生了线程切换，会在另外一个线程中执行「ASP.NET场景」）。这个捕获当然是有代价的，当时我们误以为性能问题是这个地方的开销引起，但实际上这个开销很小，在我们的应用场景不至于会带来性能问题。

2）当Configurewait(flase)，则不进行线程上下文信息的捕获，async方法中与await之后的代码执行时就无法获取await之前的线程的上下文信息，在ASP.NET中最直接的影响就是HttpConext.Current的值为null。