flush()方法,刷新内存队列,将数据写入到对端。flush()方法和write()方法在正常情况下,流程差不多,例如在pipeline中对事件的传播,从tail节点传播到head节点,最后由Unsafe处理。然而两者Unsafe的处理方式不同。
- write方法将数据写到内存队列中。
- flush方法刷新内存队列,将其中数据写入对端。
(还有些差异后文提)
AbstractChannel 对 flush() 方法的实现:
@Override
public Channel flush() {
pipeline.flush();
return this;
}
上面调用对应的 ChannelPipeline的flush() 方法,将 flush 事件在 pipeline 上传播。
DefaultChannelPipeline的flush() 方法:
@Override
public final ChannelPipeline flush() {
tail.flush();
return this;
}
tail.flush();将flush事件在pipeline中,从尾结点向头结点传播:
TailContext 对 flush() 方法的实现,是从 AbstractChannelHandlerContext 抽象类继承
@Override
public ChannelHandlerContext flush() {
// 获得下一个 Outbound 节点
final AbstractChannelHandlerContext next = findContextOutbound();
EventExecutor executor = next.executor();
// 在 EventLoop 的线程中
if (executor.inEventLoop()) {
// 执行 flush 事件到下一个节点
next.invokeFlush();
// 不在 EventLoop 的线程中
} else {
// 创建 flush 任务
Runnable task = next.invokeFlushTask;
if (task == null) {
next.invokeFlushTask = task = new Runnable() {
@Override
//不在EventLoop线程中
public void run() {
next.invokeFlush();
}
};
}
// 提交到 EventLoop 的线程中,执行该任务
safeExecute(executor, task, channel().voidPromise(), null);
}
return this;
}
在 pipeline 中,flush 事件最终会到达 HeadContext 节点。而 HeadContext 的 flush() 方法会处理该事件:
@Override
public void flush(ChannelHandlerContext ctx) throws Exception {
unsafe.flush();
}
最终会传播 flush 事件到 head 节点,刷新内存队列,将其中的数据写入到对端。
在这之前和write操作非常相似。
在上面方法内部,会调用 AbstractUnsafe的flush() 方法,刷新内存队列,将其中的数据写入到对端:
@Override
public final void flush() {
assertEventLoop();
// 内存队列为 null ,一般是 Channel 已经关闭,所以直接返回。
ChannelOutboundBuffer outboundBuffer = this.outboundBuffer;
if (outboundBuffer == null) {
return;
}
// 标记内存队列开始 flush
outboundBuffer.addFlush();
// 执行 flush
flush0();
}
flush0():
/**
* 是否正在 flush 中,即正在调用 {@link #flush0()} 中
*/
private boolean inFlush0;
@SuppressWarnings("deprecation")
protected void flush0() {
// 正在 flush 中,所以直接返回。
if (inFlush0) {
// Avoid re-entrance
return;
}
//if里
// 内存队列为 null ,一般是 Channel 已经关闭,所以直接返回。
// 内存队列为空,无需 flush ,所以直接返回
final ChannelOutboundBuffer outboundBuffer = this.outboundBuffer;
if (outboundBuffer == null || outboundBuffer.isEmpty()) {
return;
}
// 标记正在 flush 中。
inFlush0 = true;
// 若未激活,通知 flush 失败
// Mark all pending write requests as failure if the channel is inactive.
if (!isActive()) {
try {
if (isOpen()) {
outboundBuffer.failFlushed(FLUSH0_NOT_YET_CONNECTED_EXCEPTION, true);
} else {
// Do not trigger channelWritabilityChanged because the channel is closed already.
outboundBuffer.failFlushed(FLUSH0_CLOSED_CHANNEL_EXCEPTION, false);
}
} finally {
// 标记不在 flush 中。
inFlush0 = false;
}
return;
}
// 执行真正的写入到对端
try {
doWrite(outboundBuffer);
} catch (Throwable t) {
if (t instanceof IOException && config().isAutoClose()) {
/**
* Just call {@link #close(ChannelPromise, Throwable, boolean)} here which will take care of
* failing all flushed messages and also ensure the actual close of the underlying transport
* will happen before the promises are notified.
*
* This is needed as otherwise {@link #isActive()} , {@link #isOpen()} and {@link #isWritable()}
* may still return {@code true} even if the channel should be closed as result of the exception.
*/
close(voidPromise(), t, FLUSH0_CLOSED_CHANNEL_EXCEPTION, false);
} else {
try {
shutdownOutput(voidPromise(), t);
} catch (Throwable t2) {
close(voidPromise(), t2, FLUSH0_CLOSED_CHANNEL_EXCEPTION, false);
}
}
} finally {
// 标记不在 flush 中。
inFlush0 = false;
}
}
实际上,AbstractNioUnsafe 重写了 flush0() 方法:
@Override
protected final void flush0() {
// Flush immediately only when there's no pending flush.
// If there's a pending flush operation, event loop will call forceFlush() later,
// and thus there's no need to call it now.
if (!isFlushPending()) {
super.flush0();
}
}
在执行父类 AbstractUnsafe 的 flush0() 方法时,先调用 AbstractNioUnsafe的isFlushPending() 判断,是否已经处于 flush 准备中:
private boolean isFlushPending() {
SelectionKey selectionKey = selectionKey();
return selectionKey.isValid() // 合法
&& (selectionKey.interestOps() & SelectionKey.OP_WRITE) != 0; // 对 SelectionKey.OP_WRITE 事件不感兴趣。
}
若在异常情况下,(Channel在大多数情况下可写,所以不需要专门注册SelectionKey.OP_WRITE 事件,所以在Netty的实现中,默认Channel可写。当写入失败的时候,再去注册SelectionKey.OP_WRITE 事件)flush()方法若写入数据到Channel失败,就会通过注册SelectionKey.OP_WRITE 事件,然后再轮训到Channel可写时,再“回调”forceFlush()方法。(这里和write()有些差异)
AbstractChannel的doWrite(ChannelOutboundBuffer in) 抽象方法,执行真正的写入到对端。定义在 AbstractChannel 抽象类中:
/**
* Flush the content of the given buffer to the remote peer.
*/
protected abstract void doWrite(ChannelOutboundBuffer in) throws Exception;
NioSocketChannel 对该抽象方法,实现代码如下:
@Override
protected void doWrite(ChannelOutboundBuffer in) throws Exception {
SocketChannel ch = javaChannel();
// 获得自旋写入次数
int writeSpinCount = config().getWriteSpinCount();
do {
// 内存队列为空,结束循环,直接返回
if (in.isEmpty()) {
// 取消对 SelectionKey.OP_WRITE 的感兴趣
// All written so clear OP_WRITE
clearOpWrite();
// Directly return here so incompleteWrite(...) is not called.
return;
}
// 获得每次写入的最大字节数
// Ensure the pending writes are made of ByteBufs only.
int maxBytesPerGatheringWrite = ((NioSocketChannelConfig) config).getMaxBytesPerGatheringWrite();
// 从内存队列中,获得要写入的 ByteBuffer 数组
ByteBuffer[] nioBuffers = in.nioBuffers(1024, maxBytesPerGatheringWrite);
// 写入的 ByteBuffer 数组的个数
int nioBufferCnt = in.nioBufferCount();
// 写入 ByteBuffer 数组,到对端
// Always us nioBuffers() to workaround data-corruption.
// See https://github.com/netty/netty/issues/2761
switch (nioBufferCnt) {
case 0:
// TODO 1014 扣 doWrite0 的细节
// We have something else beside ByteBuffers to write so fallback to normal writes.
writeSpinCount -= doWrite0(in);
break;
case 1: {
// Only one ByteBuf so use non-gathering write
// Zero length buffers are not added to nioBuffers by ChannelOutboundBuffer, so there is no need
// to check if the total size of all the buffers is non-zero.
ByteBuffer buffer = nioBuffers[0];
int attemptedBytes = buffer.remaining();
// 执行 NIO write 调用,写入单个 ByteBuffer 对象到对端
final int localWrittenBytes = ch.write(buffer);
// 写入字节小于等于 0 ,说明 NIO Channel 不可写,所以注册 SelectionKey.OP_WRITE ,等待 NIO Channel 可写,并返回以结束循环
if (localWrittenBytes <= 0) {
incompleteWrite(true);
return;
}
// 调整每次写入的最大字节数
adjustMaxBytesPerGatheringWrite(attemptedBytes, localWrittenBytes, maxBytesPerGatheringWrite);
// 从内存队列中,移除已经写入的数据( 消息 )
in.removeBytes(localWrittenBytes);
// 写入次数减一
--writeSpinCount;
break;
}
default: {
// Zero length buffers are not added to nioBuffers by ChannelOutboundBuffer, so there is no need
// to check if the total size of all the buffers is non-zero.
// We limit the max amount to int above so cast is safe
long attemptedBytes = in.nioBufferSize();
// 执行 NIO write 调用,写入多个 ByteBuffer 到对端
final long localWrittenBytes = ch.write(nioBuffers, 0, nioBufferCnt);
// 写入字节小于等于 0 ,说明 NIO Channel 不可写,所以注册 SelectionKey.OP_WRITE ,等待 NIO Channel 可写,并返回以结束循环
if (localWrittenBytes <= 0) {
incompleteWrite(true);
return;
}
// 调整每次写入的最大字节数
// Casting to int is safe because we limit the total amount of data in the nioBuffers to int above.
adjustMaxBytesPerGatheringWrite((int) attemptedBytes, (int) localWrittenBytes, maxBytesPerGatheringWrite);
// 从内存队列中,移除已经写入的数据( 消息 )
in.removeBytes(localWrittenBytes);
// 写入次数减一
--writeSpinCount;
break;
}
}
} while (writeSpinCount > 0); // 循环自旋写入
// 内存队列中的数据未完全写入,说明 NIO Channel 不可写,所以注册 SelectionKey.OP_WRITE ,等待 NIO Channel 可写
incompleteWrite(writeSpinCount < 0);
}
上面在 Channel 不可写的时候,会注册 SelectionKey.OP_WRITE ,等待 NIO Channel 可写。而后会”回调” forceFlush() 方法,该方法内部也会调用 doWrite(ChannelOutboundBuffer in) 方法。所以在完成内部队列的数据向对端写入时候,需要调用 clearOpWrite() 方法:
protected final void clearOpWrite() {
final SelectionKey key = selectionKey();
// Check first if the key is still valid as it may be canceled as part of the deregistration
// from the EventLoop
// See https://github.com/netty/netty/issues/2104
if (!key.isValid()) { // 合法
return;
}
final int interestOps = key.interestOps();
// 若注册了 SelectionKey.OP_WRITE ,则进行取消
if ((interestOps & SelectionKey.OP_WRITE) != 0) {
key.interestOps(interestOps & ~SelectionKey.OP_WRITE);
}
}
调用 AbstractNioByteChannel中incompleteWrite(true) 方法:
protected final void incompleteWrite(boolean setOpWrite) {
// Did not write completely.
// true ,注册对 SelectionKey.OP_WRITE 事件感兴趣
if (setOpWrite) {
setOpWrite();
// false ,取消对 SelectionKey.OP_WRITE 事件感兴趣
} else {
// It is possible that we have set the write OP, woken up by NIO because the socket is writable, and then
// use our write quantum. In this case we no longer want to set the write OP because the socket is still
// writable (as far as we know). We will find out next time we attempt to write if the socket is writable
// and set the write OP if necessary.
clearOpWrite();
// Schedule flush again later so other tasks can be picked up in the meantime
// 立即发起下一次 flush 任务
eventLoop().execute(flushTask);
}
}
setOpWrite 为 true ,调用 setOpWrite() 方法,注册对 SelectionKey.OP_WRITE 事件感兴趣。(说明 Channel 此时是不可写的,那么调用父类 AbstractUnsafe 的 flush0() 方法,也没有意义,所以就不调用。)
protected final void setOpWrite() {
final SelectionKey key = selectionKey();
// Check first if the key is still valid as it may be canceled as part of the deregistration
// from the EventLoop
// See https://github.com/netty/netty/issues/2104
if (!key.isValid()) { // 合法
return;
}
final int interestOps = key.interestOps();
// 注册 SelectionKey.OP_WRITE 事件的感兴趣
if ((interestOps & SelectionKey.OP_WRITE) == 0) {
key.interestOps(interestOps | SelectionKey.OP_WRITE);
}
}
setOpWrite 为 false ,调用 clearOpWrite() 方法,取消对 SelectionKey.OP_WRITE 事件感兴趣。而后,立即发起下一次 flush 任务。
io.netty.channel.ChannelOutboundBuffer 内存队列
在 write 操作时,将数据写到 ChannelOutboundBuffer 中。
在 flush 操作时,将 ChannelOutboundBuffer 的数据写入到对端。
在 write 操作时,将数据写到 ChannelOutboundBuffer 中,都会产生一个 Entry 对象:
/**
* Recycler 对象,用于重用 Entry 对象
*/
private static final Recycler<Entry> RECYCLER = new Recycler<Entry>() {
@Override
protected Entry newObject(Handle<Entry> handle) {
return new Entry(handle);
}
};
/**
* Recycler 处理器
*/
private final Handle<Entry> handle;
/**
* 下一条 Entry,通过它,形成 ChannelOutboundBuffer 内部的链式存储每条写入数据的数据结构
*/
Entry next;
/**
* 写入的消息( 数据 )
*/
Object msg;
/**
* {@link #msg} 转化的 NIO
ByteBuffer[] bufs;
/**
* {@link #msg} 转化的 NIO ByteBuffer 对象
*/
ByteBuffer buf;
/**
* Promise 对象
*/
ChannelPromise promise;
/**
* 已写入的字节数
*/
long progress;
/**
* 长度,可读字节数数。
*/
long total;
/**
* 每个 Entry 预计占用的内存大小,计算方式为消息( {@link #msg} )的字节数 + Entry 对象自身占用内存的大小。
*/
int pendingSize;
/**
* {@link #msg} 转化的 NIO ByteBuffer 的数量。
*
* 当 = 1 时,使用 {@link #buf}
* 当 > 1 时,使用 {@link #bufs}
*/
int count = -1;
/**
* 是否取消写入对端
*/
boolean cancelled;
private Entry(Handle<Entry> handle) {
this.handle = handle;
}
newInstance(Object msg, int size, long total, ChannelPromise promise) 静态方法,创建 Entry 对象:
static Entry newInstance(Object msg, int size, long total, ChannelPromise promise) {
// 通过 Recycler 重用对象
Entry entry = RECYCLER.get();
// 初始化属性
entry.msg = msg;
entry.pendingSize = size + CHANNEL_OUTBOUND_BUFFER_ENTRY_OVERHEAD;
entry.total = total;
entry.promise = promise;
return entry;
}
recycle() 方法,回收 Entry 对象,为下次重用该对象。
void recycle() {
// 重置属性
next = null;
bufs = null;
buf = null;
msg = null;
promise = null;
progress = 0;
total = 0;
pendingSize = 0;
count = -1;
cancelled = false;
// 回收 Entry 对象
handle.recycle(this);
}
recycleAndGetNext() 方法,获得下一个 Entry 对象,并回收当前 Entry 对象
Entry recycleAndGetNext() {
// 获得下一个 Entry 对象
Entry next = this.next;
// 回收当前 Entry 对象
recycle();
// 返回下一个 Entry 对象
return next;
}
cancel() 方法,标记 Entry 对象,取消写入到对端。在 ChannelOutboundBuffer 里,Entry 数组是通过链式的方式进行组织,而当某个 Entry 对象( 节点 )如果需要取消写入到对端,是通过设置 canceled = true 来标记删除:
int cancel() {
if (!cancelled) {
// 标记取消
cancelled = true;
int pSize = pendingSize;
// 释放消息( 数据 )相关的资源
// release message and replace with an empty buffer
ReferenceCountUtil.safeRelease(msg);
// 设置为空 ByteBuf
msg = Unpooled.EMPTY_BUFFER;
// 置空属性
pendingSize = 0;
total = 0;
progress = 0;
bufs = null;
buf = null;
// 返回 pSize
return pSize;
}
return 0;
}
addMessage(Object msg, int size, ChannelPromise promise) 方法,写入消息( 数据 )到内存队列。promise 只有在真正完成写入到对端操作,才会进行通知:
/**
* Add given message to this {@link ChannelOutboundBuffer}. The given {@link ChannelPromise} will be notified once
* the message was written.
*/
public void addMessage(Object msg, int size, ChannelPromise promise) {
// 创建新 Entry 对象
Entry entry = Entry.newInstance(msg, size, total(msg), promise);
// 若 tailEntry 为空,将 flushedEntry 也设置为空。防御型编程,实际不会出现
if (tailEntry == null) {
flushedEntry = null;
// 若 tailEntry 非空,将原 tailEntry 指向新 Entry
} else {
Entry tail = tailEntry;
tail.next = entry;
}
// 更新 tailEntry 为新 Entry
tailEntry = entry;
// 若 unflushedEntry 为空,更新为新 Entry
if (unflushedEntry == null) {
unflushedEntry = entry;
}
// 增加 totalPendingSize 计数
// increment pending bytes after adding message to the unflushed arrays.
// See https://github.com/netty/netty/issues/1619
incrementPendingOutboundBytes(entry.pendingSize, false);
}
addFlush() 方法,标记内存队列每个 Entry 对象,开始 flush 。代码如下:
public void addFlush() {
// There is no need to process all entries if there was already a flush before and no new messages
// where added in the meantime.
//
// See https://github.com/netty/netty/issues/2577
Entry entry = unflushedEntry;
if (entry != null) {
// 若 flushedEntry 为空,赋值为 unflushedEntry ,用于记录第一个( 开始 ) flush 的 Entry 。
if (flushedEntry == null) {
// there is no flushedEntry yet, so start with the entry
flushedEntry = entry;
}
// 计算 flush 的数量,并设置每个 Entry 对应的 Promise 不可取消
do {
// 增加 flushed
flushed ++;
// 设置 Promise 不可取消
if (!entry.promise.setUncancellable()) { // 设置失败
// 减少 totalPending 计数
// Was cancelled so make sure we free up memory and notify about the freed bytes
int pending = entry.cancel();
decrementPendingOutboundBytes(pending, false, true);
}
// 获得下一个 Entry
entry = entry.next;
} while (entry != null);
// 设置 unflushedEntry 为空,表示所有都 flush
// All flushed so reset unflushedEntry
unflushedEntry = null;
}
}
close(…) 方法,关闭 ChannelOutboundBuffer ,进行后续的处理:
void close(ClosedChannelException cause) {
close(cause, false);
}
void close(final Throwable cause, final boolean allowChannelOpen) {
// 正在通知 flush 失败中
if (inFail) {
// 提交 EventLoop 的线程中,执行关闭
channel.eventLoop().execute(new Runnable() {
@Override
public void run() {
close(cause, allowChannelOpen);
}
});
// 返回
return;
}
// 标记正在通知 flush 失败中
inFail = true;
if (!allowChannelOpen && channel.isOpen()) {
throw new IllegalStateException("close() must be invoked after the channel is closed.");
}
if (!isEmpty()) {
throw new IllegalStateException("close() must be invoked after all flushed writes are handled.");
}
// Release all unflushed messages.
try {
// 从 unflushedEntry 节点,开始向下遍历
Entry e = unflushedEntry;
while (e != null) {
// 减少 totalPendingSize
// Just decrease; do not trigger any events via decrementPendingOutboundBytes()
int size = e.pendingSize;
TOTAL_PENDING_SIZE_UPDATER.addAndGet(this, -size);
if (!e.cancelled) {
// 释放消息( 数据 )相关的资源
ReferenceCountUtil.safeRelease(e.msg);
// 通知 Promise 执行失败
safeFail(e.promise, cause);
}
// 回收当前节点,并获得下一个 Entry 节点
e = e.recycleAndGetNext();
}
} finally {
// 标记在在通知 flush 失败中
inFail = false;
}
// 清除 NIO ByteBuff 数组的缓存。
clearNioBuffers();
}
ChannelOutboundBuffer 写入控制,当我们不断调用 addMessage(Object msg, int size, ChannelPromise promise) 方法,添加消息到 ChannelOutboundBuffer 内存队列中,如果不及时 flush 写到对端( 例如程序一直未调用 Channel中flush() 方法,或者对端接收数据比较慢导致 Channel 不可写 ),可能会导致 OOM 内存溢出。所以,在 ChannelOutboundBuffer 使用 totalPendingSize 属性,存储所有 Entry 预计占用的内存大小。
- 在 totalPendingSize 大于高水位阀值时( ChannelConfig.writeBufferHighWaterMark ,默认值为 64 KB ),关闭写开关( unwritable )
- 在 totalPendingSize 小于低水位阀值时( ChannelConfig.writeBufferLowWaterMark ,默认值为 32 KB ),打开写开关( unwritable )
思路在这,偷个懒,看了看就不写啦~