Netty源码分析之ByteBuf引用计数
引用计数是一种常用的内存管理机制,是指将资源的被引用次数保存起来,当被引用次数变为零时就将其释放的过程。Netty在4.x版本开始使用引用计数机制进行部分对象的管理,其实现思路并不是特别复杂,它主要涉及跟踪某个对象被引用的次数。在Netty具体代码中需要通过引用计数进行内存管理的对象,会基于ReferenceCounted接口实现,其中引用计数大于0时则代表该对象被引用不会释放,当引用计数减少到0时,该对象就会被释放。通过引用计数机制,Netty可以很好的实现内存管理,引用计数减少到0时要么直接释放内存,要么放回内存池中重复利用。
1、基本示例
下面先通过一个简单示例看下Netty中引用计数机制的使用
@Override public void channelRead(ChannelHandlerContext ctx, Object msg) { ByteBuf recvBuffer = (ByteBuf) msg;// 申请ByteBuf 需要主动释放 if(recvBuffer.isDirect()){ System.err.println(true); } PooledByteBufAllocator allocator = new PooledByteBufAllocator(true); ByteBuf sendBuffer = allocator.buffer();//申请池化直接内存 System.err.println("sendBuffer的引用计数:"+sendBuffer.refCnt()); sendBuffer.retain(); System.err.println("sendBuffer的引用计数:"+sendBuffer.refCnt()); sendBuffer.release(); System.err.println("sendBuffer的引用计数:"+sendBuffer.refCnt());
try { byte[] bytesReady = new byte[recvBuffer.readableBytes()]; recvBuffer.readBytes(bytesReady); System.out.println("channelRead收到数据:"+ BytesUtils.toHexString(bytesReady)); byte[] sendBytes = new byte[] {0x7E,0x01,0x02,0x7e}; sendBuffer.writeBytes(sendBytes); ctx.writeAndFlush(sendBuffer); System.err.println("sendBuffer的引用计数:"+sendBuffer.refCnt()); }catch (Exception e) { // TODO: handle exception System.err.println(e.getMessage()); }finally { System.err.println("recvBuffer的引用计数:"+recvBuffer.refCnt()); recvBuffer.release(); //此处需要释放 System.err.println("recvBuffer的引用计数:"+recvBuffer.refCnt()); } }
输出结果如下,通过示例可以看出retain方法会增加计数引用,release方法会减少计数引用
true sendBuffer的引用计数:1 sendBuffer的引用计数:2 sendBuffer的引用计数:1 sendBuffer的引用计数:0 recvBuffer的引用计数:1 recvBuffer的引用计数:0
AbstractReferenceCountedByteBuf实现了对ByteBuf的内存管理,以实现内存的回收、释放或者重复利用 ,AbstractReferenceCountedByteBuf的继承实现关系如下图所示
2、ReferenceCounted接口定义
首先是ReferenceCounted接口的定义
public interface ReferenceCounted { /** * Returns the reference count of this object. If {@code 0}, it means this object has been deallocated. * 返回对象的引用计数 */ int refCnt(); /** * Increases the reference count by {@code 1}. * 增加引用计数 */ ReferenceCounted retain(); /** * Increases the reference count by the specified {@code increment}. * 引用计数增加指定值 */ ReferenceCounted retain(int increment); /** * Records the current access location of this object for debugging purposes. * If this object is determined to be leaked, the information recorded by this operation will be provided to you * via {@link ResourceLeakDetector}. This method is a shortcut to {@link #touch(Object) touch(null)}. * 记录该对象的当前访问位置,用于调试。 * 如果确定该对象被泄露,将提供此操作记录的信息给您 */ ReferenceCounted touch(); /** * Records the current access location of this object with an additional arbitrary information for debugging * purposes. If this object is determined to be leaked, the information recorded by this operation will be * provided to you via {@link ResourceLeakDetector}. * 记录该对象的当前访问位置,附加信息用于调试。 * 如果确定该对象被泄露,将提供此操作记录的信息给您 */ ReferenceCounted touch(Object hint); /** * Decreases the reference count by {@code 1} and deallocates this object if the reference count reaches at * {@code 0}. * * @return {@code true} if and only if the reference count became {@code 0} and this object has been deallocated * 引用计数减少,如果计数变为了0,则释放对象资源 * 如果对象资源被释放,则返回true,否则返回false */ boolean release(); /** * Decreases the reference count by the specified {@code decrement} and deallocates this object if the reference * count reaches at {@code 0}. * * @return {@code true} if and only if the reference count became {@code 0} and this object has been deallocated * 引用计数-指定值,如果计数变为了0,则释放对象资源或交回到对象池 * 如果对象资源被释放,则返回true,否则返回false */ boolean release(int decrement); }
3、AbstractReferenceCountedByteBuf源码分析
AbstractReferenceCountedByteBuf对ReferenceCounted进行了具体实现,retain与release两个方法通过CAS方式对引用计数refcnt进行操作,下面对其源码进行简单分析
初始化
引用计数初始值refCnt 使用关键字volatile修饰,保证线程的可见性,同时使用偶数,引用增加通过位移操作实现,提高运算效率。
采用 AtomicIntegerFieldUpdater 对象,通过CAS方式更新refCnt,以实现线程安全,避免加锁,提高效率。
private static final long REFCNT_FIELD_OFFSET; //采用 AtomicIntegerFieldUpdater 对象,CAS方式更新refCnt private static final AtomicIntegerFieldUpdater<AbstractReferenceCountedByteBuf> refCntUpdater = AtomicIntegerFieldUpdater.newUpdater(AbstractReferenceCountedByteBuf.class, "refCnt"); //refCnt 实际值为偶数,采用位移操作提高效率 // even => "real" refcount is (refCnt >>> 1); odd => "real" refcount is 0 @SuppressWarnings("unused") private volatile int refCnt = 2;
retain操作
上面示例中每调用一次retain方法,引用计数就会累加一次,我们看下源码中retain的具体实现
public ByteBuf retain() { return retain0(1); } @Override public ByteBuf retain(int increment) { return retain0(checkPositive(increment, "increment")); } //计数器增值操作 private ByteBuf retain0(final int increment) { // all changes to the raw count are 2x the "real" change int adjustedIncrement = increment << 1; // overflow OK here 真正的计数都是2倍递增 int oldRef = refCntUpdater.getAndAdd(this, adjustedIncrement); //通过CAS方式递增并获取原值 if ((oldRef & 1) != 0) {//判断奇偶,正常情况这里应该都是偶数 throw new IllegalReferenceCountException(0, increment); } // don't pass 0! 如果计数小于等于0,以及整型范围越界(0x7fffffff+1)抛出异常 if ((oldRef <= 0 && oldRef + adjustedIncrement >= 0) || (oldRef >= 0 && oldRef + adjustedIncrement < oldRef)) { // overflow case refCntUpdater.getAndAdd(this, -adjustedIncrement); throw new IllegalReferenceCountException(realRefCnt(oldRef), increment); } return this; }
release操作
通过调用release方法,对引用计数做减值操作,源码中release的具体实现要注意的是由于引用计数以2倍递增,所以引用次数= 引用计数/2,当decrement=refcnt/2 也就是引用次数=释放次数时,代表ByteBuf不再被引用,执行内存释放或放回内存池的操作。
//计数器减值操作 private boolean release0(int decrement) { int rawCnt = nonVolatileRawCnt(), realCnt = toLiveRealCnt(rawCnt, decrement); //对计数器进行除以2操作,也就是引用次数 /** * /这里如注意 你传入的减值参数decrement = realCnt 时 等同于 引用次数=释放次数,直接进行释放操作 */ if (decrement == realCnt) { if (refCntUpdater.compareAndSet(this, rawCnt, 1)) { //CAS方式置为1 deallocate();//内存释放或放回内存池 return true; } return retryRelease0(decrement);//进入具体操作 } return releaseNonFinal0(decrement, rawCnt, realCnt); } private boolean releaseNonFinal0(int decrement, int rawCnt, int realCnt) { //如果decrement 小于 realCnt,通过CAS方式减去decrement*2 if (decrement < realCnt // all changes to the raw count are 2x the "real" change && refCntUpdater.compareAndSet(this, rawCnt, rawCnt - (decrement << 1))) { return false; } return retryRelease0(decrement); } private boolean retryRelease0(int decrement) { for (;;) { int rawCnt = refCntUpdater.get(this), realCnt = toLiveRealCnt(rawCnt, decrement); if (decrement == realCnt) { if (refCntUpdater.compareAndSet(this, rawCnt, 1)) { deallocate(); return true; } } else if (decrement < realCnt) {//如果decrement 小于 realCnt,通过CAS方式减去decrement*2 // all changes to the raw count are 2x the "real" change if (refCntUpdater.compareAndSet(this, rawCnt, rawCnt - (decrement << 1))) { return false; } } else { throw new IllegalReferenceCountException(realCnt, -decrement); } Thread.yield(); // this benefits throughput under high contention } } /** * Like {@link #realRefCnt(int)} but throws if refCnt == 0 */ private static int toLiveRealCnt(int rawCnt, int decrement) { if ((rawCnt & 1) == 0) { return rawCnt >>> 1; } // odd rawCnt => already deallocated throw new IllegalReferenceCountException(0, -decrement); }
4、总结
以上我们围绕AbstractReferenceCountedByteBuf对Netty引用计数的具体实现进行了分析,可以看到Netty在实现引用计数的同时,结合CAS、位移计算等方式,保证了运算效率和线程安全,在实际项目中我们遇到类似应用场景也都可以借鉴参考,如数据发送次数,商品剩余数量等计数场景的实现。希望本文对大家能有所帮助,其中如有不足与不正确的地方还望指正与海涵,十分感谢。
关注微信公众号,查看更多技术文章。