NIO框架之MINA源码解析(一):背景
NIO框架之MINA源码解析(二):mina核心引擎
NIO框架之MINA源码解析(三):底层通信与责任链模式应用
1、粘包与段包
粘包:指TCP协议中,发送方发送的若干包数据到接收方接收时粘成一包,从接收缓冲区看,后一包数据的头紧接着前一包数据的尾。
造成的可能原因:
发送端需要等缓冲区满才发送出去,造成粘包
接收方不及时接收缓冲区的包,造成多个包接收
断包:也就是数据不全,比如包太大,就把包分解成多个小包,多次发送,导致每次接收数据都不全。
2、消息传输的格式
消息长度+消息头+消息体 即前N个字节用于存储消息的长度,用于判断当前消息什么时候结束。
消息头+消息体 即固定长度的消息,前几个字节为消息头,后面的是消息头。
在MINA中用的是
消息长度+消息体 即前4个字节用于存储消息的长度,用于判断当前消息什么时候结束。
3、编码与解码
在网络中,信息的传输都是通过字节的形式传输的,而我们在编写自己的代码时,则都是具体的对象,那么要想我们的对象能够在网络中传输,就需要编码与解码。
编码:即把我们的消息编码成二进制形式,能以字节的形式在网络中传输。
解码:即把我们收到的字节解码成我们代码中的对象。
在MINA中对象的编码与解码用的都是JDK提供的ObjectOutputStream来实现的。
4、MINA中消息的处理实现
消息的接受处理,我们常用的是TCP协议,而TCP协议会分片的,在下面的代码中,具体功能就是循环从通道里面读取数据,直到没有数据可读,或者buffer满了,然后就把接受到的数据发给解码工厂进行处理。
4.1、消息的接收
//class AbstractPollingIoProcessor
private void read(S session) {
IoSessionConfig config = session.getConfig();
int bufferSize = config.getReadBufferSize();
IoBuffer buf = IoBuffer.allocate(bufferSize);
final boolean hasFragmentation = session.getTransportMetadata().hasFragmentation();
try {
int readBytes = 0;
int ret;
try {
//是否有分片 tcp传输会有分片,即把大消息分片成多个小消息再传输
if (hasFragmentation) {
//read方法非阻塞,没有读到数据的时候返回0
while ((ret = read(session, buf)) > 0) {
readBytes += ret;
//buffer 满了
if (!buf.hasRemaining()) {
break;
}
}
} else {
ret = read(session, buf);
if (ret > 0) {
readBytes = ret;
}
}
} finally {
buf.flip();
}
if (readBytes > 0) {
IoFilterChain filterChain = session.getFilterChain();
//处理消息
filterChain.fireMessageReceived(buf);
buf = null;
if (hasFragmentation) {
if (readBytes << 1 < config.getReadBufferSize()) {
session.decreaseReadBufferSize();
} else if (readBytes == config.getReadBufferSize()) {
session.increaseReadBufferSize();
}
}
}
if (ret < 0) {
scheduleRemove(session);
}
} catch (Throwable e) {
if (e instanceof IOException) {
if (!(e instanceof PortUnreachableException)
|| !AbstractDatagramSessionConfig.class.isAssignableFrom(config.getClass())
|| ((AbstractDatagramSessionConfig) config).isCloseOnPortUnreachable()) {
scheduleRemove(session);
}
}
IoFilterChain filterChain = session.getFilterChain();
filterChain.fireExceptionCaught(e);
}
}
4.2、解码与编码
//class AbstractIoBuffer
public Object getObject(final ClassLoader classLoader) throws ClassNotFoundException {
//首先判断当前buffer中消息长度是否完整,不完整的话直接返回
if (!prefixedDataAvailable(4)) {
throw new BufferUnderflowException();
}
//消息长度
int length = getInt();
if (length <= 4) {
throw new BufferDataException("Object length should be greater than 4: " + length);
}
int oldLimit = limit();
//limit到消息结尾处
limit(position() + length);
try {
ObjectInputStream in = new ObjectInputStream(asInputStream()) {
@Override
protected ObjectStreamClass readClassDescriptor() throws IOException, ClassNotFoundException {
int type = read();
if (type < 0) {
throw new EOFException();
}
switch (type) {
case 0: // NON-Serializable class or Primitive types
return super.readClassDescriptor();
case 1: // Serializable class
String className = readUTF();
Class<?> clazz = Class.forName(className, true, classLoader);
return ObjectStreamClass.lookup(clazz);
default:
throw new StreamCorruptedException("Unexpected class descriptor type: " + type);
}
}
@Override
protected Class<?> resolveClass(ObjectStreamClass desc) throws IOException, ClassNotFoundException {
String name = desc.getName();
try {
return Class.forName(name, false, classLoader);
} catch (ClassNotFoundException ex) {
return super.resolveClass(desc);
}
}
};
return in.readObject();
} catch (IOException e) {
throw new BufferDataException(e);
} finally {
limit(oldLimit);
}
}
//判断当前消息是否完整
public boolean prefixedDataAvailable(int prefixLength, int maxDataLength) {
if (remaining() < prefixLength) {
return false;
}
int dataLength;
switch (prefixLength) {
case 1:
dataLength = getUnsigned(position());
break;
case 2:
dataLength = getUnsignedShort(position());
break;
case 4:
dataLength = getInt(position());
break;
default:
throw new IllegalArgumentException("prefixLength: " + prefixLength);
}
if (dataLength < 0 || dataLength > maxDataLength) {
throw new BufferDataException("dataLength: " + dataLength);
}
//判断当前消息是否完整
return remaining() - prefixLength >= dataLength;
}
//编码
public IoBuffer putObject(Object o) {
int oldPos = position();
skip(4); // Make a room for the length field.预留4个字节用于存储消息长度
try {
ObjectOutputStream out = new ObjectOutputStream(asOutputStream()) {
@Override
protected void writeClassDescriptor(ObjectStreamClass desc) throws IOException {
try {
Class<?> clz = Class.forName(desc.getName());
if (!Serializable.class.isAssignableFrom(clz)) { // NON-Serializable class
write(0);
super.writeClassDescriptor(desc);
} else { // Serializable class
write(1);
writeUTF(desc.getName());
}
} catch (ClassNotFoundException ex) { // Primitive types
write(0);
super.writeClassDescriptor(desc);
}
}
};
out.writeObject(o);
out.flush();
} catch (IOException e) {
throw new BufferDataException(e);
}
// Fill the length field
int newPos = position();
position(oldPos);
//存储消息长度
putInt(newPos - oldPos - 4);
position(newPos);
return this;
}
4.3、断包与粘包处理
// class CumulativeProtocolDecoder
public void decode(IoSession session, IoBuffer in, ProtocolDecoderOutput out) throws Exception {
//是否有分片,tcp 有分片
if (!session.getTransportMetadata().hasFragmentation()) {
while (in.hasRemaining()) {
if (!doDecode(session, in, out)) {
break;
}
}
return;
}
// 1、断包处理
// 2、处理粘包
boolean usingSessionBuffer = true;
//session中是否有断包情况(上次处理后),断包保存在session中
IoBuffer buf = (IoBuffer) session.getAttribute(BUFFER);
// If we have a session buffer, append data to that; otherwise
// use the buffer read from the network directly.
if (buf != null) {//有断包,则把当前包拼接到断包里面
boolean appended = false;
// Make sure that the buffer is auto-expanded.
if (buf.isAutoExpand()) {
try {
buf.put(in);
appended = true;
} catch (IllegalStateException e) {
// A user called derivation method (e.g. slice()),
// which disables auto-expansion of the parent buffer.
} catch (IndexOutOfBoundsException e) {
// A user disabled auto-expansion.
}
}
if (appended) {
buf.flip();
} else {
// Reallocate the buffer if append operation failed due to
// derivation or disabled auto-expansion.
buf.flip();
IoBuffer newBuf = IoBuffer.allocate(buf.remaining() + in.remaining()).setAutoExpand(true);
newBuf.order(buf.order());
newBuf.put(buf);
newBuf.put(in);
newBuf.flip();
buf = newBuf;
// Update the session attribute.
session.setAttribute(BUFFER, buf);
}
} else {
buf = in;
usingSessionBuffer = false;
}
//2 粘包处理,可能buffer中有多个消息,需要多次处理(解码)每个消息,直到消息处理完,或者剩下的消息不是一个完整的消息或者buffer没有数据了
for (;;) {
int oldPos = buf.position();
boolean decoded = doDecode(session, buf, out);
if (decoded) {//解码 成功
if (buf.position() == oldPos) {
throw new IllegalStateException("doDecode() can't return true when buffer is not consumed.");
}
//buffer空了
if (!buf.hasRemaining()) {//buffer没有数据了
break;
}
} else {//剩下的消息不是一个完整的消息,断包出现了
break;
}
}
// if there is any data left that cannot be decoded, we store
// it in a buffer in the session and next time this decoder is
// invoked the session buffer gets appended to
if (buf.hasRemaining()) {//剩下的消息不是一个完整的消息,断包出现了
//如果断包已经保存在session中,则更新buffer,没有的话,就把剩下的断包保存在session中
if (usingSessionBuffer && buf.isAutoExpand()) {
buf.compact();
} else {
storeRemainingInSession(buf, session);
}
} else {
if (usingSessionBuffer) {
removeSessionBuffer(session);
}
}
}
//class ObjectSerializationDecoder
protected boolean doDecode(IoSession session, IoBuffer in, ProtocolDecoderOutput out) throws Exception {
//首先判断当前buffer中消息长度是否完整,不完整的话直接返回
if (!in.prefixedDataAvailable(4, maxObjectSize)) {
return false;
}
out.write(in.getObject(classLoader));
return true;
}