[英]Ensuring no packet loss between TCP client and server
I am writing a Java TCP client which sends chunks of data to a C server. 我正在编写一个Java TCP客户端,它将数据块发送到C服务器。 The client-server worked very well on my development PC.
客户端 - 服务器在我的开发PC上运行良好。 This code upon deployment on a hardware board showed packet loss.
在硬件板上部署时,此代码显示数据包丢失。 I only have the logs with me and I know that the server did not receive all packets.
我只有我的日志,我知道服务器没有收到所有数据包。 I do not have the hardware to test.
我没有要测试的硬件。 Therefore, at the first level, I want to be very sure client code does send all the required data.
因此,在第一级,我想非常确定客户端代码确实发送了所有必需的数据。
Here is my code(the client part in Java). 这是我的代码(Java中的客户端部分)。 How do I make sure this is done?
我如何确保这样做? Is there some resend commands with timings etc?
是否有一些重定时命令等命令?
Socket mySocket = new Socket("10.0.0.2",2800);
OutputStream os = mySocket.getOutputStream();
System.out.println(" Sending 8 byte Header Msg with length of following data to Server");
os.write(hdr, 0, 8);
os.flush();
System.out.println(" Sending Data ");
start = 0;
for(int index=0; index < ((rbuffer.length/chucksize)+1); index++){
if(start + chucksize > rbuffer.length) {
System.arraycopy(rbuffer, start, val, 0, rbuffer.length - start);
} else {
System.arraycopy(rbuffer, start, val, 0, chucksize);
}
start += chucksize ;
os.write(val,0,chucksize);
os.flush();
}
Here is the C snippet which receives this data: 以下是接收此数据的C片段:
while ((bytes_received = recv(connected, rMsg, sizeof(rMsg),0)) > 0){
if (bytes_received > 0) // zero indicates end of transmission */
{
/* get length of message (2 bytes) */
tmpVal = 0;
tmpVal |= rMsg[idx++];
tmpVal = tmpVal << 8;
tmpVal |= rMsg[idx++];
msg_len = tmpVal;
len = msg_len;
//printf("msg_len = %d\n", len);
printf("length of following message from header message : %d\n", len);
char echoBuffer[RCVBUFSIZE] ;
memset(echoBuffer, 0, RCVBUFSIZE);
int recvMsgsize = 0;
plain=(char *)malloc(len+1);
if (!plain)
{
fprintf(stderr, "Memory error!");
}
for( i = RCVBUFSIZE; i < (len+RCVBUFSIZE); i=i+RCVBUFSIZE){
if(i>=len){
recvMsgSize = recv(connected, echoBuffer, (len - (i-RCVBUFSIZE)), 0);
memcpy(&plain[k], echoBuffer, recvMsgSize);
k = k+recvMsgSize;
}
else{
recvMsgSize = recv(connected, echoBuffer, RCVBUFSIZE, 0);
memcpy(&plain[k], echoBuffer, recvMsgSize);
k = k+recvMsgSize;
}
}
}//closing if
}//closing while
First of all there is no such thing as packet loss in TCP/IP. 首先,没有TCP / IP中的丢包。 This protocol was designed to reliably send a stream of bytes in correct order.
该协议旨在以正确的顺序可靠地发送字节流。 So the problem must be with your application or the other side.
所以问题必须在你的应用程序或另一方。
I am not really in a mood to analyze this whole arraycopy()
madness (C anyone?), but why aren't you just sending the whole rbuffer
in one go through BufferedOutputStream
? 我真的不想分析整个
arraycopy()
疯狂(C任何人?),但为什么你不是只通过BufferedOutputStream
发送整个rbuffer
?
OutputStream os = new BufferedOutputStream(mySocket.getOutputStream());
and then: 然后:
os.write(rbuffer);
Believe me, BufferedOutputStream
is doing the exact same thing (collecting bytes into chunks and sending them in one go). 相信我,
BufferedOutputStream
正在做同样的事情(将字节收集到块中并一次发送)。 Or maybe I am missing something? 或许我错过了什么?
I changed the C side program in the following way and it now works: 我通过以下方式更改了C端程序,现在可以正常工作:
printf("length of following message from header message : %d\n", len);
plain=(char *)malloc(len+1);
if (!plain)
{
fprintf(stderr, "Memory error!");
}
memset(plain, 0, len+1);
int remain = len;
k= 0;
while (remain){
int toGet = remain > RCVBUFSIZE ? RCVBUFSIZE : remain;
remain -= toGet;
int recvd = 0;
while(recvd < toGet) {
if((recvMsgSize = recv(connected, echoBuffer, toGet-recvd, 0)) < 0){
printf("error receiving data\n");
}
memcpy(&plain[k], echoBuffer, recvMsgSize);
k += recvMsgSize;
printf("Total data accumulated after recv input %d\n", k);
recvd += recvMsgSize;
}
}
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.