Tencent Interview: Do you understand process communication?

[[432787]]

This article is reprinted from the WeChat public account "CS Guide", the author is Dabai. Please contact the CS Guide public account for reprinting this article.

noon

On a sunny afternoon, Dabai came to the leader's office.

Dabai: Boss, my tooth has been hurting since last night. I think I need to go to the hospital...

Leader: Okay, go ahead, take care of yourself...

That afternoon

Tencent Building, in a conference room~

Dabai: Hello, interviewer. I am Dabai! Although I have just worked for less than a year, I have strong programming basics and rich project experience...

Interviewer: OK, I have a general idea. Now let me ask you some general questions... No, let’s discuss some underlying principles.

Dabai: (Inner thoughts: No problem, I’m good at this. Haven’t you read my series on clichés?) Okay, interviewer!

Interviewer: Let me ask you a simple question first. Can you explain the difference between a process and a thread?

Dabai: OK, a process can contain many threads!

Interviewer: No more?

Dabai: It’s gone.

Interviewer: I am almost done with your interview, please go back and wait for the conclusion...

The interview officially begins

Dabai: No, no, no, I'm just kidding. Actually, there is a big difference between the two. Let's start with the process. I have written an article about Linux process creation before...

Interviewer: Excuse me, I have read your article and it is well written. I want to ask you something that you did not mention in the article. Do you understand process communication?

Dabai: I understand. There are mainly the following process communication methods: pipes, message queues, shared memory, semaphores, and sockets.

Interviewer: Do you usually use pipes for process communication?

Dabai: Yes, I have used it, often. There are two types of pipes: anonymous pipes and named pipes. I will introduce them to you in detail.

Let's take an example. The following Linux command uses pipes.

 echo "I'm dabai" | tee a. out  

The specific function of the above command is to output "I'm dabai" in the Linux console and pass the output information as input to the a.out file. It can be seen that the left side of I is input and the right side is output. The pipe represented by "|" is created with the command and will be automatically destroyed after the command is executed.

Another way is to create a named pipe explicitly through mkfifo.

 mkfifo dabaipipe

The above command creates a pipe named "dabaipipe". We can write information to the pipe, for example:

 echo "hello! I'm dabai" > dabaipipie

If you want to read the information in the pipeline, just use the following command!

 cat < dabaipipe 
 
 # Output hello! I'm dabai

Interviewer: Wow! Not bad. Let me ask you when to use anonymous pipes and when to use named pipes.

Dabai: Actually, this should start with the shortcomings of anonymous pipes. Anonymous pipes have two shortcomings, which are both its shortcomings and its characteristics.

(1) Anonymous pipes only support communication between parent processes and child processes.

(2) Anonymous pipes do not support communication between two processes across a network

It is precisely because of these two characteristics that the system overhead required by anonymous pipes is much smaller than that of named pipes. (Files in anonymous pipes are not written to disk, while files in named pipes are written to disk)

So it is obvious when to use anonymous pipes and when to use named pipes. (Praise me now)

Interviewer: Wait, what did you say? You said anonymous pipes only support communication between parent and child processes? Then in the command "echo "I'm dabai" | tee a.out" you just ran, who is the child process of whom?

Dabai: (Inner thoughts: This interviewer is really unethical! Fortunately, I was prepared and was waiting for you to make the mistake. I can only talk to you about this until the end of the interview)

Well, the process of creating an anonymous pipe is very interesting. Let me tell you about it, so that I can answer your question.

To create an anonymous pipe, we need to follow a few steps.

1. We need to create the "pipe" of the anonymous pipe, which is actually a cache in the kernel.

2. In order to communicate, we need to create a process, and then create two file descriptors in the process to point to the two ends of the pipe (one pointing to the input end and one pointing to the output end).

At this point, our anonymous pipe looks like this: If a process wants to input data into the pipe, it writes it through the input port descriptor, and if it wants to get data from the pipe, it reads it through the output port descriptor.

From the above figure, we can also see that there is no inter-process communication. So how to achieve communication between the parent process and the child process? In my article about processes, I talked about how the parent process creates the child process. Do you remember? It is to use fork. The parent process creates the child process through fork, and the child process will copy the parent process code, so that the input terminal descriptor of the parent process is also copied to the child process. At this time, both the parent process and the child process will point to the pipe at the same time.

We then close the input of the child process and the output of the parent process, and that's it!

As for the command "echo "I'm dabai" | tee a.out ", | the two sides are not the parent and child processes, so how do they complete the process communication? In fact, both processes are forked through the shell process. Then the two processes are connected to the input and output ends of the pipe respectively. Then just close the redundant input and output.

Finally, I don’t need to draw the structure diagram of the anonymous pipeline, right? (Inner thought: This interview has lasted for more than an hour, and I still have to go back to the company for free dinner and take a taxi after get off work at 9 o’clock)

Interviewer: Why don’t you draw it?

Dabai: Well then... it will be like this! (Inner thought: Aren’t you in a hurry to eat?)

Interviewer: You are good, young man. I am quite satisfied with you. I have passed you. I wanted to ask you about other ways of process communication, but I have to get off work. Please wait for the next interviewer to arrange an interview time for you!

After adding WeChat with the interviewer, I looked at the time. It was OK. I could go back to the company to collect the meal allowance and go home. It was another fulfilling day. Recently, I learned several other ways of process communication and shared them with the next interviewer.

References:

Geek Time "Interesting Talk about Linux Operating System" Link: http://gk.link/a/10zn1

"Linux System Programming, Network Programming" Chapter 7 Interprocess Communication (Local IPC) Link: https://www.bilibili.com/video/BV1fE411v7Bb

JavaGuide Operating System Link: https://snailclimb.gitee.io/javaguide/#/docs/cs-basics/operating-system/basis