Why does sigaction say that the signals sent by the shortcut keys come from PID 0?

Question

I do dome research regarding signal processing. In this particular case, I'm interested in SIGTSTP on Linux (3.0.101 on SLES 11). I wrote a small program that catches SIGTSTP and prints the parent PID and PID of the process that sent the signal.

Here is what I see:

If I use kill -TSTP, then the PID of the submit process is the PID of the shell in which I ran the kill command, as expected.

If I type ctrl + z in the shell, the PID of the send process is 0, but I expected the PID of the shell in which I pressed ctrl + z (and I ran the trap program)

Does anyone know why this is happening? Shouldn't there be 0 PIDs of some special kernel-only processes? The Sigaction documentation reports that signals sent by kill will fill in the si_pid field, but command line shortcuts are not mentioned. Perhaps si_pid = 0 means "unspecified sender".

This is my capture program.

#include<stdio.h>
#include<stdlib.h>
#include<signal.h>
#include<unistd.h>
#include<string.h>

void tstp_handler(int num, siginfo_t* info, void* context)
{
    pid_t ppid = getppid();
    printf("\nReceived TSTP. pid %d ppid %d\n", info->si_pid, ppid);
}

int main(void)
{
    struct sigaction action;
    memset(&action, 0, sizeof(struct sigaction));
    action.sa_sigaction = tstp_handler;
    action.sa_flags = SA_SIGINFO;
    sigaction(SIGTSTP, &action, NULL);

    printf("Registered\n");
    printf("My PID is %d\n", getpid());
    while(1) 
        sleep(1);
    return 0;
}

Thanks and best regards

Answer 1

Please refer to the manual regarding struct siginfo:

si_signo, si_errno si_code .        (si_errno Linux.)         , ,         [...]

, info->si_pid . .

, kill si_pid, PID kill.