Transport layer protocol ----- UDP protocol

Transport layer

Talk about port number again

1. effect ： It's actually an identifier , Used to distinguish different processes in the application layer .
But in communication , The port number is very important . A five tuple information is a necessary information in communication , The port number accounts for two , The success of a communication , Five tuples are indispensable , Therefore, you must know both the source port number and the opposite port number .

2. Representation of port number ： The port number is represented by 16 Bit by bit , That is to say 2 Bytes , The scope of identification is 0~65535.
among ： There are different divisions for different intervals .

• Be familiar with port number ：0~1023. stay IANA（ Internet address code distribution organization ） These port numbers are assigned to TCP\IP Some important application protocols in the architecture .（ for example ：FTP The port number of the protocol is 20/21,HTTP The port number of the protocol is 80,HTTPS The port number of the protocol is 443,DNS The port number of the protocol is 53 And so on. ）
• Registration port number ：1024~49151. Use for applications that are not familiar with port numbers . And if you want to use these port numbers , Must be in IANA Go up and register according to the prescribed procedures , Prevent mixing .
• Short port number ：49152~65535. Left to the customer process for a short time . When a customer communicates , The system will automatically assign an unused port number to the customer in this data （ This is what we did before tcp Socket programming and udp Socket programming , Reason for not binding port number to client ）. When the server process receives the message sent by the client process , We know the dynamic port number used by the client process . And after the communication , This port number can be used by other customers in the future .

Be careful ： The port number only has local meaning , Just to identify the processes of the application layer of this computer , And on the Internet , It doesn't matter if you have the same port number in different computers .

3. Except for the port number , In fact, there is an agreement number

Pass the above figure , You can see it ,ip In addition to the source ip Address and destination ip Out of address , One more TCP6, And this TCP6 Medium 6 Namely TCP Protocol fields for .（ namely ,6 Namely tcp The agreement number of ）
And for UDP agreement , Its agreement number is actually 17.
You can see from the above figure that , Port number in ip Duan disappeared , And in the tcp In the datagram , Therefore, the port number exists between the application layer and the transport layer , For the agreement number , It exists between the transport layer and the network layer .

4. Reuse of sender and reuse of receiver
Here's the picture ：

among ： For the picture , On the left is the sender , On the right is the receiver , The sender multiplexes the data according to the transmitted protocol , The receiving party divides the data according to the received data , Finally get the result you want .

Know the well-known port number

above , Some well-known port numbers have been introduced , In fact, there are many , Many of the port numbers we often use are as follows ：

• ssh The server ：22 port
• ftp The server ：21 port
• telnet The server ：23 port
• http The server ：80 port
• https The server ：443 port

There are still a lot of it , We can query the well-known port number through this command ：cat/etc/services
The query results are as follows ：

therefore , If we want to write a program to use the port number , We should try our best to avoid using the fixed port number of others , Prevent conflicts .

Two questions

1. A process can bind How many port numbers ？

answer ： A process can bind multiple port numbers .

reason ： Because multiple file descriptors can be opened in one process , And each file descriptor corresponds to a port number , So a process can bind multiple port numbers .

2. Whether a port number can be used by multiple processes bind Well ？

answer ： A port number cannot be bound by multiple processes . namely ： A port number can only be bound by one process .（ But the above also has some special points , It is a parent process, which is bound with a port number , But he also fork A sub process comes out , This subprocess also has this port number . So in this way, multiple processes are bound to the same port number . however , Different processes that are not parent-child processes , It is absolutely impossible to bind the same port number ）

reason ： Because the binding of port number only occurs on the server , In general , The server has multiple ports , One port only corresponds to one service . And most of the time （ With the exception of the parent-child process mentioned above ）, A service corresponds to a process , There are no two processes serving a port number at the same time .

netstat and pidof

1.netstat command

①： grammar ：netstat [ Options ]
②： function ： Check the network status .
Common options ：

• -n： Reject alias , Can display all of the numbers into numbers .
• -l ： Only those with LISTEN（ Monitoring ） Service status .（ be based on TCP signal communication ）
• -p： Displays the name of the program that established the relevant connection .
• -t ： Show only tcp Related options .
• -u： Show only udp Related options .
• -a： Show all options , Default not to show LISTEN dependent .

Examples are as follows ：
see tcp process

see udp The process is as follows ：

among ： Above picture
①Proto： It represents the protocol type .
②Local Address： It means the source side ip Address and source port number .
③Foreign Address： It means the destination ip Address and destination port number .
④State： Indicates the current established state .

3.pidof command

① grammar ：pidof [ Process name ].
② function ： View by process name id.
for example ：

Upper figure , adopt bash, Check their pid Number .

UDP agreement

UDP Protocol segment format

Please look at the chart below. ：

For one udp Simulation diagram of user datagram , among udp The bytes occupied by the header are 8 byte , And it contains four components ： Source port , Destination port , length , Inspection and . Every possession 2 Bytes .
among ：

• Source port ： Where does the data come from .
• Destination port ： Where does the data go .
• length ： Length of data .（ The whole datagram contains the first ）
• Inspection and ： If udp Message check and error , Will discard the message .

The internal structure is as follows ：

It means these four parts .

UDP Characteristics

UDP The process of transmission , Similar to sending a letter , So it has the following characteristics ：

1. There is no connection ： When communicating with each other , No connection needed （ Just know each other's addresses ）, Just send data directly .（ Just like sending a letter , I only know your address , I can send you a letter ）

2. Support one-to-one , One to many , Many to one and many to many interactive communication ： Use udp In the LAN of the Protocol , Any host can send broadcast information to other main sentences （ That is, all hosts in the LAN can receive ）, You can also send multicast to one of the multicast groups （ Multicast group is a group composed of several hosts under the same LAN ）, Unicast can also be sent to one of the hosts .

3. The application of direct packet delivery layer ： The datagram sent by the sender to the transmission layer , Use UDP The agreement will directly add a UDP The first one is sent .

Be careful ：

• Application layer to UDP How long data ,udp Will not split , Will not merge , Add directly udp The first part is then sent directly .
• If the sender calls once sendto, And the one sent is 100 Bytes . Then the corresponding end should only call once recvfrom, Once received 100 Bytes , Instead of cycling ten times , Every time I receive 10 Bytes .
• We noticed that , We talked about , Two bytes are used to save one udp Length of datagram , So one udp The maximum byte length of a datagram is 64k, And it also contains the first . therefore , For the data part , When we transmit , Try to be as small as possible 64k, If the data we want to transmit is greater than 64k, Then you need to subcontract manually at the application layer , Send... Multiple times , And manually assemble at the receiving end .（ Based on strict whole transmission , That is, if the buffer of the receiver is too small , There is no room for the data sent by the sender , Then this data will be discarded ）

4. Do your best to deliver , It's unreliable ： If in the process of transmission , The user datagram receives interference and generates error code , The receiving party will act in accordance with udp Header field calibration information , Check the error code , If there is a bit error , The side discards only this datagram , And nothing else （ Because there is no retransmission mechanism , There is no confirmation mechanism ）.

And so udp About the buffer ：

• udp There is no real buffer , call sendto Will be handed over directly to the kernel , The kernel transmits the data to the network layer protocol for subsequent transmission .
• udp With buffer , The function of this buffer cannot guarantee the received udp Message sequence and sending udp Report in the same order , And if the buffer is full , Then the newspaper coming later will be discarded .
• udp Of socket, Not only can you read , You can also write , It's a full duplex communication socket.

among ：udp The test and error detection method of is binary summation algorithm .
Binary summation algorithm ：

①： The sender organizes the data , Set the check and position of the header field to 0, And then from the 0 Bytes （ Including the first , Is the first byte of the first part ） Start to reverse sum each byte , beyond 16 Bit , Then cut off the high order , Let it be low 16 Bits continue to sum .

②： After completion, fill the checksum into the inspection sum field .

③： The receiver receives the data , Sum the data in reverse code from beginning to end , Now , If the data is correct , So the end result should be 0.

UDP The protocol corresponding to the application layer

• NFS： Network file system .
• TFTP： Simple file transfer protocol .
• DHCP： Dynamic Host Configuration Protocol .
• BOOTP： Boot protocol .（ Used for setting startup without disk ）
• DNS： Domain name resolution .

Of course , It also includes writing by yourself udp The protocol customized at the application layer when the program .