Netty源码分析第七章: 编码器和写数据
第四节: 刷新buffer队列
上一小节学习了writeAndFlush的write方法, 这一小节我们剖析flush方法
通过前面的学习我们知道, flush方法通过事件传递, 最终会传递到HeadContext的flush方法:
public void flush(ChannelHandlerContext ctx) throws Exception { ???unsafe.flush();}这里最终会调用AbstractUnsafe的flush方法:
public final void flush() { ???assertEventLoop(); ???ChannelOutboundBuffer outboundBuffer = this.outboundBuffer; ???if (outboundBuffer == null) { ???????return; ???} ???outboundBuffer.addFlush(); ???flush0();}这里首先也是拿到ChannelOutboundBuffer对象
然后我们看这一步:
outboundBuffer.addFlush();
这一步同样也是调整ChannelOutboundBuffer的指针
跟进addFlush方法:
public void addFlush() { ???Entry entry = unflushedEntry; ???if (entry != null) { ????????if (flushedEntry == null) { ???????????flushedEntry = entry; ???????} ???????do { ???????????flushed ++; ???????????if (!entry.promise.setUncancellable()) { ???????????????int pending = entry.cancel(); ???????????????decrementPendingOutboundBytes(pending, false, true); ???????????} ???????????entry = entry.next; ???????} while (entry != null); ???????unflushedEntry = null; ???}}首先声明一个entry指向unflushedEntry, 也就是第一个未flush的entry
通常情况下unflushedEntry是不为空的, 所以进入if
再未刷新前flushedEntry通常为空, 所以会执行到flushedEntry = entry
也就是flushedEntry指向entry
经过上述操作, 缓冲区的指针情况如图所示:
7-4-1
然后通过do-while将, 不断寻找unflushedEntry后面的节点, 直到没有节点为止
flushed自增代表需要刷新多少个节点
循环中我们关注这一步
decrementPendingOutboundBytes(pending, false, true);
这一步也是统计缓冲区中的字节数, 但是是和上一小节的incrementPendingOutboundBytes正好是相反, 因为这里是刷新, 所以这里要减掉刷新后的字节数,
我们跟到方法中:
private void decrementPendingOutboundBytes(long size, boolean invokeLater, boolean notifyWritability) { ???if (size == 0) { ???????return; ???} ???//从总的大小减去 ???long newWriteBufferSize = TOTAL_PENDING_SIZE_UPDATER.addAndGet(this, -size); ???//直到减到小于某一个阈值32个字节 ???if (notifyWritability && newWriteBufferSize < channel.config().getWriteBufferLowWaterMark()) { ???????//设置写状态 ???????setWritable(invokeLater); ???}}同样TOTAL_PENDING_SIZE_UPDATER代表缓冲区的字节数, 这里的addAndGet中参数是-size, 也就是减掉size的长度
再看 if (notifyWritability && newWriteBufferSize < channel.config().getWriteBufferLowWaterMark())
getWriteBufferLowWaterMark()代表写buffer的第水位值, 也就是32k, 如果写buffer的长度小于这个数, 就通过setWritable方法设置写状态
也就是通道由原来的不可写改成可写
回到addFlush方法:
遍历do-while循环结束之后, 将unflushedEntry指为空, 代表所有的entry都是可写的
经过上述操作, 缓冲区的指针情况如下图所示:
7-4-2
回到AbstractUnsafe的flush方法:
指针调整完之后, 我们跟到flush0()方法中:
protected void flush0() { ???if (inFlush0) { ???????return; ???} ???final ChannelOutboundBuffer outboundBuffer = this.outboundBuffer; ???if (outboundBuffer == null || outboundBuffer.isEmpty()) { ???????return; ???} ???inFlush0 = true; ???if (!isActive()) { ???????try { ???????????if (isOpen()) { ???????????????outboundBuffer.failFlushed(FLUSH0_NOT_YET_CONNECTED_EXCEPTION, true); ???????????} else { ???????????????outboundBuffer.failFlushed(FLUSH0_CLOSED_CHANNEL_EXCEPTION, false); ???????????} ???????} finally { ???????????inFlush0 = false; ???????} ???????return; ???} ???try { ???????doWrite(outboundBuffer); ???} catch (Throwable t) { ???????if (t instanceof IOException && config().isAutoClose()) { ???????????close(voidPromise(), t, FLUSH0_CLOSED_CHANNEL_EXCEPTION, false); ???????} else { ???????????outboundBuffer.failFlushed(t, true); ???????} ???} finally { ???????inFlush0 = false; ???}}if (inFlush0) 表示判断当前flush是否在进行中, 如果在进行中, 则返回, 避免重复进入
我们重点关注doWrite方法
跟到AbstractNioByteChannel的doWrite方法中去:
protected void doWrite(ChannelOutboundBuffer in) throws Exception { ???int writeSpinCount = -1; ???boolean setOpWrite = false; ???for (;;) { ???????//每次拿到当前节点 ???????Object msg = in.current(); ???????if (msg == null) { ???????????clearOpWrite(); ???????????return; ???????} ???????if (msg instanceof ByteBuf) { ???????????//转化成ByteBuf ???????????ByteBuf buf = (ByteBuf) msg; ???????????//如果没有可写的值 ???????????int readableBytes = buf.readableBytes(); ???????????if (readableBytes == 0) { ???????????????//移除 ???????????????in.remove(); ???????????????continue; ???????????} ????????????boolean done = false; ???????????long flushedAmount = 0; ???????????if (writeSpinCount == -1) { ???????????????writeSpinCount = config().getWriteSpinCount(); ???????????} ???????????for (int i = writeSpinCount - 1; i >= 0; i --) { ???????????????//将buf写入到socket里面 ???????????????//localFlushedAmount代表向jdk底层写了多少字节 ???????????????int localFlushedAmount = doWriteBytes(buf); ???????????????//如果一个字节没写, 直接break ???????????????if (localFlushedAmount == 0) { ???????????????????setOpWrite = true; ???????????????????break; ???????????????} ???????????????//统计总共写了多少字节 ???????????????flushedAmount += localFlushedAmount; ???????????????//如果buffer全部写到jdk底层 ???????????????if (!buf.isReadable()) { ???????????????????//标记全写道 ???????????????????done = true; ???????????????????break; ???????????????} ???????????} ???????????in.progress(flushedAmount); ???????????if (done) { ???????????????//移除当前对象 ???????????????in.remove(); ???????????} else { ???????????????break; ???????????} ???????} else if (msg instanceof FileRegion) { ???????????//代码省略 ???????} else { ???????????throw new Error(); ???????} ???} ???incompleteWrite(setOpWrite);}首先是一个无限for循环
Object msg = in.current() 这一步是拿到flushedEntry指向的entry中的msg
跟到current()方法中:
public Object current() { ????Entry entry = flushedEntry; ???if (entry == null) { ???????return null; ???} ???return entry.msg;}这里直接拿到flushedEntry指向的entry中关联的msg, 也就是一个ByteBuf
回到doWrite方法:
如果msg为null, 说明没有可以刷新的entry, 则调用clearOpWrite()方法清除写标识
如果msg不为null, 则会判断是否是ByteBuf类型, 如果是ByteBuf, 就进入if块中的逻辑
if块中首先将msg转化为ByteBuf, 然后判断ByteBuf是否可读, 如果不可读, 则通过in.remove()将当前的byteBuf所关联的entry移除, 然后跳过这次循环进入下次循环
remove方法稍后分析, 这里我们先继续往下看
boolean done = false 这里设置一个标识, 标识刷新操作是否执行完成, 这里默认值为false代表走到这里没有执行完成
writeSpinCount = config().getWriteSpinCount() 这里是获得一个写操作的循环次数, 默认是16
然后根据这个循环次数, 进行循环的写操作
在循环中, 关注这一步:
int localFlushedAmount = doWriteBytes(buf);
这一步就是将buf的内容写到channel中, 并返回写的字节数, 这里会调用NioSocketChannel的doWriteBytes
我们跟到doWriteBytes方法中:
protected int doWriteBytes(ByteBuf buf) throws Exception { ????final int expectedWrittenBytes = buf.readableBytes(); ???return buf.readBytes(javaChannel(), expectedWrittenBytes);}这里首先拿到buf的可读字节数, 然后通过readBytes将可读字节写入到jdk底层的channel中
回到doWrite方法:
将内容写的jdk底层的channel之后, 如果一个字节都没写, 说明现在channel可能不可写, 将setOpWrite设置为true, 用于标识写操作位, 并退出循环
如果已经写出字节, 则通过 flushedAmount += localFlushedAmount 累加写出的字节数
然后根据是buf是否没有可读字节数判断是否buf的数据已经写完, 如果写完, 将done设置为true, 说明写操作完成, 并退出循环
因为有时候不一定一次就能将byteBuf所有的字节写完, 所以这里会继续通过循环进行写出, 直到循环到16次
如果ByteBuf内容完全写完, 会通过in.remove()将当前entry移除掉
我们跟到remove方法中:
public boolean remove() { ???//拿到当前第一个flush的entry ???Entry e = flushedEntry; ???if (e == null) { ???????clearNioBuffers(); ???????return false; ???} ???Object msg = e.msg; ???ChannelPromise promise = e.promise; ???int size = e.pendingSize; ???removeEntry(e); ???if (!e.cancelled) { ???????ReferenceCountUtil.safeRelease(msg); ???????safeSuccess(promise); ???????decrementPendingOutboundBytes(size, false, true); ???} ???e.recycle(); ???return true;}首先拿到当前的flushedEntry
我们重点关注removeEntry这步, 跟进去:
private void removeEntry(Entry e) { ????if (-- flushed == 0) { ???????//位置为空 ???????flushedEntry = null; ???????//如果是最后一个节点 ???????if (e == tailEntry) { ???????????//全部设置为空 ???????????tailEntry = null; ???????????unflushedEntry = null; ???????} ???} else { ???????//移动到下一个节点 ???????flushedEntry = e.next; ???}}if (-- flushed == 0) 表示当前节点是否为需要刷新的最后一个节点, 如果是, 则flushedEntry指针设置为空
如果当前节点是tailEntry节点, 说明当前节点是最后一个节点, 将tailEntry和unflushedEntry两个指针全部设置为空
如果当前节点不是需要刷新的最后的一个节点, 则通过 flushedEntry = e.nex t这步将flushedEntry指针移动到下一个节点
以上就是flush操作的相关逻辑
Netty源码分析第7章(编码器和写数据)---->第4节: 刷新buffer队列
原文地址：https://www.cnblogs.com/xiangnan6122/p/10208247.html