Pipeline details of IPC (interprocess communication)

linux and windows There are

Pipes and shared memory （ One of the fastest IPC） difference ：

1. Pipeline one communication and four data copies ： By user space buf Copy data to the kernel -> The kernel copies the data into memory -> Memory to kernel -> Kernel to user space buf. Shared memory copies data only twice ： User space to memory -> Memory to user space .

2. Pipelines are implemented with circular queues , Continuous data transmission can be unlimited . The data size of shared memory is fixed each time ;

3. Shared memory can be accessed randomly , Pipes can only be read and written sequentially ;

 

Anonymous pipeline

Communicating between processes with a common ancestor . signal communication Follow the process and die , So it is Only in memory Temporary files for . Because of this file No file name , Cannot be opened by a non parent process , It can only be used for communication between kinship processes , Therefore, the communication pipeline formed by files without names is called “ Anonymous pipeline ”.

 

principle ： Creating a child process copies some resources of the parent process, including files . If the parent process creates a file before creating the child process , Then the descriptor of this file will be shared by the child processes created later , Father 、 Child processes can communicate through this file . If both sides of the communication can only read , The other side can only write , Then this file is a pipeline that can only send messages in one direction .

 

Realization ： A process calls a function pipe() Create a pipe .pipe() The function is to create a memory file , But unlike the functions that create ordinary files , function pipe() Two file descriptors of this file will be returned for the process . among , A read-only file descriptor , A file descriptor that writes only attributes , That is, the process passes fildes[0] You can only read files , And by fildes[1] You can only write files . After determining the transmission direction , Closed by one party fildes[0], Closed by one party （close()） File descriptor fildes[1], Thus, the connection of the pipeline becomes a one-way transmission pipeline in the following cases .

 

The parent process creates a pipeline

After the parent process creates the child process , Pipes are also shared

The communicating party turns off reading , One side closes the write . Method ：close（） File descriptor

limitations ： Built in memory , So its capacity can't be very large ; A stream of unformatted bytes is transmitted , It is required that both parties using the pipeline must agree on the data format to be transmitted .

 

 

name pipes

On the actual file system Real documents , So it can communicate between any process .

Because the named pipeline strictly follows the principle of first in first out to transmit data , So this kind of pipe is also called FIFO file .

The synchronization between communication processes needs to be ensured by the pipeline itself , Named pipes are also a can only One way access The file of .

principle ： in other words , Named pipes provide a pathname to associate with , With FIFO The file form of exists in the file system , Generate a physical file in the file system , Other processes only need to access the file path , Can communicate with each other through pipelines . Open the pipeline file in read-only mode on the read data side , Open the pipeline file in write only mode on the write data side .

Difference from ordinary files ：

Ordinary files cannot realize byte stream management , Multi process access resource conflict ;
FIFO Files are managed by byte stream , Follow the first in, first out principle , Does not involve shared resource access .
The operation flow is ：mkfifo -> open -> read（write） -> close ->unlink.