CODEWITHSUNDEEP
clinked-listlinux-kernel
Adam
Adam

Reputation: 10036

Counting the number of siblings a process has, Linux kernel

I'm using the list_for_each macro and the sibling field of the current task_struct to print out the number of siblings current has.

Here's the code (inf.num_siblings tracks the number of siblings):

    struct list_head *curr_node;
    ....
    list_for_each(curr_node, &(current->sibling)) {
                inf.num_siblings++;
            }
    ...
    //print inf.num_siblings

current->sibling points to a node in a circularly linked list, so I think the code above should count the number of siblings/node in said list correctly.

This code seems to mostly work. If I run a program in the background that just loops forever before running the "count siblings" code, then the number of siblings for the current process increases by one compared with the count that existed before I ran the loop code. However, if I run a process that calls fork() once before starting to loop, rather than seeing the sibling count increase by two as I would expect, it only increases by one. Why is that?

Here is the code that forks then loops:

#include <sys/types.h>
#include <unistd.h> 

int main(int argc, char *argv[])
{   
    fork();

    while(1){
        sleep(3);
    }

    return 0;
}

Upvotes: 1

Views: 819

Answers (2)

Harry
Harry

Reputation: 11648

When you call fork you create a child process not a sibling. When you fork you'll increase the current->children count not the current->sibling count :)

Upvotes: 2

WreckeR
WreckeR

Reputation: 419

From: http://man7.org/linux/man-pages/man2/fork.2.html

When you called fork() the new process you created is a child and not a sibling of the process in which you called fork(). The sibling is correctly not getting incremented in this case.

Now, if you call fork() a bunch of times, and track the sibling count for the first child process, you will see the sibling count go up.

Upvotes: 0

Related Questions