First off: this is not a Unix/Linux system. I am working on an IBM AS/400 V7R1 (C++ 98) and do not have access to fork(). Nevertheless, I do have spawnp() to start new child processes and the AS/400 supports the notion of process groups.
In my system, I have a "head" program that starts X number of children. This head calls accept() on incoming connections and immediately gives the socket away to one of the child process via sendmsg(). The children are all sitting on recvmsg(). For the head program, it goes something like this:
rc = socketpair(AF_UNIX, SOCK_DGRAM, 0, pair_sd);
if (rc != 0) {
perror("socketpair() failed");
close(listen_sd);
exit(-1);
}
server_sd = pair_sd[0];
worker_sd = pair_sd[1];
// do some other stuff, set up arguments for spawnp()...
// ...
spawn_fdmap[0] = worker_sd;
for (int i = 0; i < numOfChildren; i++) {
pid = spawnp(spawn_argv[0], 1, spawn_fdmap, &inherit, spawn_argv, spawn_envp);
if (pid < 0) {
CERR << "errno=" << errno << ", " << strerror(errno) << endl;
CERR << "command line [";
for (int x = 0; spawn_argv[x] != 0; ++x) {
cerr << spawn_argv[x] << " ";
}
cerr << ']' << endl;
close(listen_sd);
exit(-1);
}
else {
CERR << "Child worker PID = " << pid << endl;
child_pids.push_back(pid);
}
}
// Close down the worker side of the socketpair.
close(worker_sd);
I've got a reason/scheme to start additional child processes after initial program start. I plan to send the head program some signal which would cause the spawnp() call to execute again. The "close(worker_sd)" has me concerned though. Can I call spawnp() again after I've closed the worker socket? It's just a number, after all. Is it OK to keep the worker_sd open?