无阻塞io是使用单线程或者只使用少量的多线程,每个连接共用一个线程,当处于等待(没有事件)的时候线程资源可以释放出来处理别的请求,通过事件驱动模型当有accept/read/write等事件发生后通知(唤醒)主线程分配资源来处理相关事件。java.nio.channels.Selector就是在该模型中事件的观察者,可以将多个SocketChannel的事件注册到一个Selector上,当没有事件发生时Selector处于阻塞状态,当SocketChannel有accept/read/write等事件发生时唤醒Selector。
这个Selector是使用了单线程模型,主要用来描述事件驱动模型,要优化性能需要一个好的线程模型来使用,目前比较好的nio框架有Netty,apache的mina等。线程模型这块后面再分享,这里重点研究Selector的阻塞和唤醒原理。
先看一段简单的Selector使用的代码
selector = Selector.open();
ServerSocketChannel ssc = ServerSocketChannel.open();
ssc.configureBlocking(false);
ssc.socket().bind(new InetSocketAddress(port));
ssc.register(selector, SelectionKey.OP_ACCEPT);
while (true) {
// select()阻塞,等待有事件发生唤醒
int selected = selector.select();
if (selected > 0) {
Iterator selectedKeys = selector.selectedKeys().iterator();
while (selectedKeys.hasNext()) {
SelectionKey key = selectedKeys.next();
if ((key.readyOps() & SelectionKey.OP_ACCEPT) == SelectionKey.OP_ACCEPT) {
// 处理 accept 事件
} else if ((key.readyOps() & SelectionKey.OP_READ) == SelectionKey.OP_READ) {
// 处理 read 事件
} else if ((key.readyOps() & SelectionKey.OP_WRITE) == SelectionKey.OP_WRITE) {
// 处理 write 事件
}
selectedKeys.remove();
}
}
}
代码中关键的几个点在:
Selector.open();
selector.select();
阻塞后唤醒可以通过注册在selector上的socket有事件发生 或者 selector.select(timeOut)超时 或者 selector.wakeup()主动唤醒;
整个阻塞和唤醒的过程涉及到的点非常多,先上一张梳理出的整体图,再进入源码会比较容易理解
现在通过openjdk中的源码来解析上图中的每一个环节:
1. Selector.open()
Selector.java
-----
public static Selector open() throws IOException {
return SelectorProvider.provider().openSelector();
}
先看看SelectorProvider.provider()做了什么:
SelectorProvider.java
-----
public static SelectorProvider provider() {
synchronized (lock) {
if (provider != null)
return provider;
return (SelectorProvider)AccessController
.doPrivileged(new PrivilegedAction() {
public Object run() {
if (loadProviderFromProperty())
return provider;
if (loadProviderAsService())
return provider;
provider = sun.nio.ch.DefaultSelectorProvider.create();
return provider;
}
});
}
}
其中provider = sun.nio.ch.DefaultSelectorProvider.create();会根据操作系统来返回不同的实现类,windows平台就返回WindowsSelectorProvider;
这里主要以windows的实现来梳理整个流程,拿到provider后来看openSelector()中的实现
WindowsSelectorProvider.java
----
public AbstractSelector openSelector() throws IOException {
return new WindowsSelectorImpl(this);
}
WindowsSelectorImpl.java
----
WindowsSelectorImpl(SelectorProvider sp) throws IOException {
super(sp);
pollWrapper = new PollArrayWrapper(INIT_CAP);
wakeupPipe = Pipe.open();
wakeupSourceFd = ((SelChImpl)wakeupPipe.source()).getFDVal();
// Disable the Nagle algorithm so that the wakeup is more immediate
SinkChannelImpl sink = (SinkChannelImpl)wakeupPipe.sink();
(sink.sc).socket().setTcpNoDelay(true);
wakeupSinkFd = ((SelChImpl)sink).getFDVal();
pollWrapper.addWakeupSocket(wakeupSourceFd, 0);
}
这段代码中做了如下几个事情
Pipe.open()打开一个管道(打开管道的实现后面再看);拿到wakeupSourceFd和wakeupSinkFd两个文件描述符;把唤醒端的文件描述符(wakeupSourceFd)放到pollWrapper里;
那么为什么需要一个管道,这个管道是怎么实现的?接下来看Pipe.open()做了什么
Pipe.java
----
public static Pipe open() throws IOException {
return SelectorProvider.provider().openPipe();
}
同样,SelectorProvider.provider()也是获取操作系统相关的实现
SelectorProvider.java
----
public Pipe openPipe() throws IOException {
return new PipeImpl(this);
}
这里还是看windows下的实现
PipeImpl.java
----
PipeImpl(final SelectorProvider sp) throws IOException {
try {
AccessController.doPrivileged(new Initializer(sp));
} catch (PrivilegedActionException x) {
throw (IOException)x.getCause();
}
}
创建了一个PipeImpl对象, AccessController.doPrivileged调用后紧接着会执行initializer的run方法
PipeImpl.Initializer
-----
public Object run() throws IOException {
ServerSocketChannel ssc = null;
SocketChannel sc1 = null;
SocketChannel sc2 = null;
try {
// loopback address
InetAddress lb = InetAddress.getByName("127.0.0.1");
assert(lb.isLoopbackAddress());
// bind ServerSocketChannel to a port on the loopback address
ssc = ServerSocketChannel.open();
ssc.socket().bind(new InetSocketAddress(lb, 0));
// Establish connection (assumes connections are eagerly
// accepted)
InetSocketAddress sa
= new InetSocketAddress(lb, ssc.socket().getLocalPort());
sc1 = SocketChannel.open(sa);
ByteBuffer bb = ByteBuffer.allocate(8);
long secret = rnd.nextLong();
bb.putLong(secret).flip();
sc1.write(bb);
// Get a connection and verify it is legitimate
for (;;) {
sc2 = ssc.accept();
bb.clear();
sc2.read(bb);
bb.rewind();
if (bb.getLong() == secret)
break;
sc2.close();
}
// Create source and sink channels
source = new SourceChannelImpl(sp, sc1);
sink = new SinkChannelImpl(sp, sc2);
} catch (IOException e) {
try {
if (sc1 != null)
sc1.close();
if (sc2 != null)
sc2.close();
} catch (IOException e2) { }
IOException x = new IOException("Unable to establish"
+ " loopback connection");
x.initCause(e);
throw x;
} finally {
try {
if (ssc != null)
ssc.close();
} catch (IOException e2) { }
}
return null;
}
这里即为上图中最下面那部分创建pipe的过程,windows下的实现是创建两个本地的socketChannel,然后连接(链接的过程通过写一个随机long做两个socket的链接校验),两个socketChannel分别实现了管道的source与sink端。
source端由前面提到的WindowsSelectorImpl放到了pollWrapper中(pollWrapper.addWakeupSocket(wakeupSourceFd, 0))
PollArrayWrapper.java
----
private AllocatedNativeObject pollArray; // The fd array
// Adds Windows wakeup socket at a given index.
void addWakeupSocket(int fdVal, int index) {
putDescriptor(index, fdVal);
putEventOps(index, POLLIN);
}
// Access methods for fd structures
void putDescriptor(int i, int fd) {
pollArray.putInt(SIZE_POLLFD * i + FD_OFFSET, fd);
}
void putEventOps(int i, int event) {
pollArray.putShort(SIZE_POLLFD * i + EVENT_OFFSET, (short)event);
}
这里将source的POLLIN事件标识为感兴趣的,当sink端有数据写入时,source对应的文件描述符wakeupSourceFd就会处于就绪状态
Java代码 收藏代码
AllocatedNativeObject.java
----
class AllocatedNativeObject extends NativeObject
AllocatedNativeObject(int size, boolean pageAligned) {
super(size, pageAligned);
}
NativeObject.java
----
protected NativeObject(int size, boolean pageAligned) {
if (!pageAligned) {
this.allocationAddress = unsafe.allocateMemory(size);
this.address = this.allocationAddress;
} else {
int ps = pageSize();
long a = unsafe.allocateMemory(size + ps);
this.allocationAddress = a;
this.address = a + ps - (a & (ps - 1));
}
}
从以上可以看到pollArray是通过unsafe.allocateMemory(size + ps)分配的一块系统内存
到这里完成了Selector.open(),主要完成建立Pipe,并把pipe的wakeupSourceFd放入pollArray中,这个pollArray是Selector的枢纽。这里是以Windows的实现来看,在windows下通过两个链接的socketChannel实现了Pipe,linux下则是直接使用系统的pipe。
2. serverSocketChannel.register(selector, SelectionKey.OP_ACCEPT);
AbstractSelectableChannel.java --> register() --> SelectorImpl.java
----
protected final SelectionKey register(AbstractSelectableChannel ch,int ops,Object attachment)
{
if (!(ch instanceof SelChImpl))
throw new IllegalSelectorException();
SelectionKeyImpl k = new SelectionKeyImpl((SelChImpl)ch, this);
k.attach(attachment);
synchronized (publicKeys) {
implRegister(k);
}
k.interestOps(ops);
return k;
}
关键是implRegister(k);
WindowsSelectorImpl.java
----
protected void implRegister(SelectionKeyImpl ski) {
growIfNeeded();
channelArray[totalChannels] = ski;
ski.setIndex(totalChannels);
fdMap.put(ski);
keys.add(ski);
pollWrapper.addEntry(totalChannels, ski);
totalChannels++;
}
PollArrayWrapper.java
----
void addEntry(int index, SelectionKeyImpl ski) {
putDescriptor(index, ski.channel.getFDVal());
}
这里把socketChannel的文件描述符放到pollArray中。
3. selector.select();
SelectorImpl.java
----
public int select(long timeout) throws IOException
{
if (timeout < 0)
throw new IllegalArgumentException("Negative timeout");
return lockAndDoSelect((timeout == 0) ? -1 : timeout);
}
private int lockAndDoSelect(long timeout) throws IOException {
synchronized (this) {if (!isOpen()) throw new ClosedSelectorException(); synchronized (publicKeys) { synchronized (publicSelectedKeys) { return doSelect(timeout); } } } }
其中的doSelector又回到我们的Windows实现:
WindowsSelectorImpl.java
----
protected int doSelect(long timeout) throws IOException {
if (channelArray == null)
throw new ClosedSelectorException();
this.timeout = timeout; // set selector timeout
processDeregisterQueue();
if (interruptTriggered) {
resetWakeupSocket();
return 0;
}
// Calculate number of helper threads needed for poll. If necessary
// threads are created here and start waiting on startLock
adjustThreadsCount();
finishLock.reset(); // reset finishLock
// Wakeup helper threads, waiting on startLock, so they start polling.
// Redundant threads will exit here after wakeup.
startLock.startThreads();
// do polling in the main thread. Main thread is responsible for
// first MAX_SELECTABLE_FDS entries in pollArray.
try {
begin();
try {
subSelector.poll();
} catch (IOException e) {
finishLock.setException(e); // Save this exception
}
// Main thread is out of poll(). Wakeup others and wait for them
if (threads.size() > 0)
finishLock.waitForHelperThreads();
} finally {
end();
}
// Done with poll(). Set wakeupSocket to nonsignaled for the next run.
finishLock.checkForException();
processDeregisterQueue();
int updated = updateSelectedKeys();
// Done with poll(). Set wakeupSocket to nonsignaled for the next run.
resetWakeupSocket();
return updated;
}
private int poll() throws IOException{ // poll for the main thread
return poll0(pollWrapper.pollArrayAddress,
Math.min(totalChannels, MAX_SELECTABLE_FDS),
readFds, writeFds, exceptFds, timeout);
}
private native int poll0(long pollAddress, int numfds, int[] readFds, int[] writeFds, int[] exceptFds, long timeout);
其他的都是一些准备工作,关键是subSelector.poll(),最后调用了native的poll0,并把pollWrapper.pollArrayAddress作为参数传给poll0,那么poll0对pollArray做了什么:
WindowsSelectorImpl.c
----
Java_sun_nio_ch_WindowsSelectorImpl_00024SubSelector_poll0(JNIEnv *env, jobject this,
jlong pollAddress, jint numfds,
jintArray returnReadFds, jintArray returnWriteFds,
jintArray returnExceptFds, jlong timeout)
{
// 代码.... 此处省略一万字
/* Call select */
if ((result = select(0 , &readfds, &writefds, &exceptfds, tv)) == SOCKET_ERROR) {
// 代码.... 此处省略一万字
for (i = 0; i < numfds; i++) {
// 代码.... 此处省略一万字
}
}
}
代码已经忘得差不多了,但这里可以看到实现思路是调用c的select方法,这里的select对应于内核中的sys_select调用,sys_select首先将第二三四个参数指向的fd_set拷贝到内核,然后对每个被SET的描述符调用进行poll,并记录在临时结果中(fdset),如果有事件发生,select会将临时结果写到用户空间并返回;当轮询一遍后没有任何事件发生时,如果指定了超时时间,则select会睡眠到超时,睡眠结束后再进行一次轮询,并将临时结果写到用户空间,然后返回。
这里的select就是轮询pollArray中的FD,看有没有事件发生,如果有事件发生收集所有发生事件的FD,退出阻塞。
关于select系统调用参考了《select、poll、epoll的比较》这篇文章,同时看到nio的select在不同平台上的实现不同,在linux上通过epoll可以不用轮询,在第一次调用后,事件信息就会与对应的epoll描述符关联起来,待的描述符上注册回调函数,当事件发生时,回调函数负责把发生的事件存储在就绪事件链表中,最后写到用户空间。
到这里已经比较清楚了,退出阻塞的方式有:regist在selector上的socketChannel处于就绪状态(放在pollArray中的socketChannel的FD就绪) 或者 第1节中放在pollArray中的wakeupSourceFd就绪。前者(socketChannel)就绪唤醒应证了文章开始的阻塞->事件驱动->唤醒的过程,后者(wakeupSourceFd)就是下面要看的主动wakeup。
4. selector.wakeup()
WindowsSelectorImpl.java
----
public Selector wakeup() {
synchronized (interruptLock) {
if (!interruptTriggered) {
setWakeupSocket();
interruptTriggered = true;
}
}
return this;
}
// Sets Windows wakeup socket to a signaled state.
private void setWakeupSocket() {
setWakeupSocket0(wakeupSinkFd);
}
private native void setWakeupSocket0(int wakeupSinkFd);
native实现摘要:
WindowsSelectorImpl.c
----
Java_sun_nio_ch_WindowsSelectorImpl_setWakeupSocket0(JNIEnv *env, jclass this,
jint scoutFd)
{
/* Write one byte into the pipe */
send(scoutFd, (char*)&POLLIN, 1, 0);
}
这里完成了向最开始建立的pipe的sink端写入了一个字节,source文件描述符就会处于就绪状态,poll方法会返回,从而导致select方法返回。(原来自己建立一个socket链着自己另外一个socket就是为了干这事)