一次 .NET Core 中玩锁的经历：ManualResetEventSlim, Semaphore 与 SemaphoreSlim

最近同事对  .net core memcached 缓存客户端 EnyimMemcachedCore 进行了高并发下的压力测试，发现在 linux 上高并发下使用 async 异步方法读取缓存数据会出现大量失败的情况，比如在一次测试中，100万次读取缓存，只有12次成功，999988次失败，好恐怖。如果改为同步方法，没有一次失败，100%成功。奇怪的是，同样的压力测试程序在 Windows 上异步读取却没问题，100%成功。

排查后发现是2个地方使用的锁引起的，一个是 ManualResetEventSlim ，一个是 Semaphore ，这2个锁是在同步方法中使用的，但 aync 异步方法中调用了这2个同步方法，我们来分别看一下。

使用 ManualResetEventSlim 是在创建 Socket 连接时用于控制连接超时

var args = new SocketAsyncEventArgs();
using (var mres = new ManualResetEventSlim())
{
    args.Completed += (s, e) => mres.Set();
    if (socket.ConnectAsync(args))
    {
        if (!mres.Wait(timeout))
        {
            throw new TimeoutException("Could not connect to " + endpoint);
        }
    }
}

使用 Semaphore 是在从 EnyimMemcachedCore 自己实现的 Socket 连接池获取 Socket 连接时

if (!this.semaphore.WaitOne(this.queueTimeout))
{
    message = "Pool is full, timeouting. " + _endPoint;
    if (_isDebugEnabled) _logger.LogDebug(message);
    result.Fail(message, new TimeoutException());

    // everyone is so busy
    return result;
}

为了弃用这个2个锁造成的异步并发问题，采取了下面2个改进措施：

1）对于 ManualResetEventSlim ，参考 corefx 中 SqlClient 的 SNITcpHandle 的实现，改用 CancellationTokenSource 控制连接超时

var cts = new CancellationTokenSource();
cts.CancelAfter(timeout);
void Cancel()
{
    if (!socket.Connected)
    {
        socket.Dispose();
    }
}
cts.Token.Register(Cancel);

socket.Connect(endpoint);
if (socket.Connected)
{
    connected = true;
}
else
{
    socket.Dispose();
}

2）对于 Semaphore ，根据同事提交的 PR ，将 Semaphore 换成 SemaphoreSlim ，用 SemaphoreSlim.WaitAsync 方法等待信号量锁

if (!await this.semaphore.WaitAsync(this.queueTimeout))
{
    message = "Pool is full, timeouting. " + _endPoint;
    if (_isDebugEnabled) _logger.LogDebug(message);
    result.Fail(message, new TimeoutException());

    // everyone is so busy
    return result;
}

改进后，压力测试结果立马与同步方法一样，100% 成功！

为什么会这样？

我们到 github 的 coreclr 仓库（针对 .net core 2.2）中看看 ManualResetEventSlim 与 Semaphore 的实现源码，看能否找到一些线索。

(一）

先看看 ManualResetEventSlim.Wait 方法的实现代码（523开始）：

1）先 SpinWait 等待

var spinner = new SpinWait();
while (spinner.Count < spinCount)
{
    spinner.SpinOnce(sleep1Threshold: -);

    if (IsSet)
    {
        return true;
    }
}

SpinWait 等待时间比较短，不会造成长时间阻塞线程。

在高并发下大量线程在争抢锁，所以大量线程在这个阶段等不到锁。

2）然后 Monitor.Wait 等待

try
{
    // ** the actual wait **
    if (!Monitor.Wait(m_lock, realMillisecondsTimeout))
        return false; //return immediately if the timeout has expired.
}
finally
{
    // Clean up: we're done waiting.
    Waiters = Waiters - ;
}

Monitor.Wait 对应的实现代码

[MethodImplAttribute(MethodImplOptions.InternalCall)]
private static extern bool ObjWait(bool exitContext, int millisecondsTimeout, object obj);

public static bool Wait(object obj, int millisecondsTimeout, bool exitContext)
{
    if (obj == null)
        throw (new ArgumentNullException(nameof(obj)));
    return ObjWait(exitContext, millisecondsTimeout, obj);
}

最终调用的是一个本地库的 ObjWait 方法。

查阅一下 Monitor.Wait 方法的帮助文档：

Releases the lock on an object and blocks the current thread until it reacquires the lock. If the specified time-out interval elapses, the thread enters the ready queue.

Monitor.Wait 的确会阻塞当前线程，这在异步高并发下会带来问题，详见一码阻塞，万码等待：ASP.NET Core 同步方法调用异步方法“死锁”的真相。

（二）

再看看 Semaphore 的实现代码，它继承自 WaitHandle ， Semaphore.Wait 实际调用的是 WaitHandle.Wait ，后者调用的是 WaitOneNative ，这是一个本地库的方法

[MethodImplAttribute(MethodImplOptions.InternalCall)]
private static extern int WaitOneNative(SafeHandle waitableSafeHandle, uint millisecondsTimeout, bool hasThreadAffinity, bool exitContext);

.net core 3.0 中有些变化，这里调用的是 WaitOneCore 方法

[MethodImpl(MethodImplOptions.InternalCall)]
private static extern int WaitOneCore(IntPtr waitHandle, int millisecondsTimeout);

查阅一下 WaitHandle.Wait 方法的帮助文档：

Blocks the current thread until the current WaitHandle receives a signal, using a 32-bit signed integer to specify the time interval in milliseconds.

WaitHandle.Wait 也会阻塞当前线程。

2个地方在等待锁时都会阻塞线程，难怪高并发下会出问题。

(三）

接着阅读 SemaphoreSlim 的源码学习它是如何在 WaitAsync 中实现异步等待锁的？

public Task<bool> WaitAsync(int millisecondsTimeout, CancellationToken cancellationToken)
{
    //...

    lock (m_lockObj!)
    {
        // If there are counts available, allow this waiter to succeed.
        if (m_currentCount > )
        {
            --m_currentCount;
            if (m_waitHandle != null && m_currentCount == ) m_waitHandle.Reset();
            return s_trueTask;
        }
        else if (millisecondsTimeout == )
        {
            // No counts, if timeout is zero fail fast
            return s_falseTask;
        }
        // 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
        {
            Debug.Assert(m_currentCount == , "m_currentCount should never be negative");
            var asyncWaiter = CreateAndAddAsyncWaiter();
            return (millisecondsTimeout == Timeout.Infinite && !cancellationToken.CanBeCanceled) ?
                asyncWaiter :
                WaitUntilCountOrTimeoutAsync(asyncWaiter, millisecondsTimeout, cancellationToken);
        }
    }
}

重点看 else 部分的代码，SemaphoreSlim.WaitAsync 造了一个专门用于等待锁的 Task —— TaskNode ，CreateAndAddAsyncWaiter 就用于创建 TaskNode 的实例

private TaskNode CreateAndAddAsyncWaiter()
{
    // Create the task
    var task = new TaskNode();

    // Add it to the linked list
    if (m_asyncHead == null)
    {
        m_asyncHead = task;
        m_asyncTail = task;
    }
    else
    {
        m_asyncTail.Next = task;
        task.Prev = m_asyncTail;
        m_asyncTail = task;
    }

    // Hand it back
    return task;
}

从上面的代码看到 TaskNode 用到了链表，神奇的等锁专用 Task —— TaskNode 是如何实现的呢？

private sealed class TaskNode : Task<bool>
{
    internal TaskNode? Prev, Next;
    internal TaskNode() : base((object?)null, TaskCreationOptions.RunContinuationsAsynchronously) { }
}

好简单！

那 SemaphoreSlim.WaitAsync 如何用 TaskNode 实现指定了超时时间的锁等待？

看 WaitUntilCountOrTimeoutAsync 方法的实现源码：

private async Task<bool> WaitUntilCountOrTimeoutAsync(TaskNode asyncWaiter, int millisecondsTimeout, CancellationToken cancellationToken)
{
    // Wait until either the task is completed, timeout occurs, or cancellation is requested.
    // We need to ensure that the Task.Delay task is appropriately cleaned up if the await
    // completes due to the asyncWaiter completing, so we use our own token that we can explicitly
    // cancel, and we chain the caller's supplied token into it.
    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);
}

用 Task.WhenAny 等待 TaskNode 与 Task.Delay ，等其中任一者先完成，简单到可怕。

又一次通过 .net core 源码欣赏了高手是怎么玩转 Task 的。

【2019-5-6更新】

今天将 Task.WhenAny + Task.Delay 的招式用到了异步连接 Socket 的超时控制中

var connTask = _socket.ConnectAsync(_endpoint);
if (await Task.WhenAny(connTask, Task.Delay(_connectionTimeout)) == connTask)
{
    await connTask;
}