Fix pidfile cleanup when receiving signal

When not using --pidfile option, pidfile_fd is initialized to -1.
However if we are interrupted by signal, when we check the value during
cleanup, the value is 0(!) and we try to delete a null pidfile and log
this error:

    ^CINFO | Received signal: Interrupt: 2
    ERROR| Failed to remove pidfile: "(null)": Bad address

This does not make sense, but we use sigsetjmp(3) for handling signals
and its manual mentions:

    The sigsetjmp()/siglongjmp() function pairs save and restore the
    signal mask if the argument savemask is non-zero; otherwise, only
    the register set and the stack are saved.

It seems that pidfile_fd is restored to the value at the time
sigsetjmp() was called, which may be 0 (or any other value on the
stack). I think this is an old bug, exposed by logging an error when
unlink() fails.

Fix by creating the pidfile and binding listen_fd before calling
sigsetjmp().  When sigsetjmp() restores the stack the value of
pidfile_fd and listen_fd are not affected.

Signed-off-by: Nir Soffer <nirsof@gmail.com>

Author: nirs

main.c

@@ -406,16 +406,6 @@ int main(int argc, char *argv[]) {
     WARN("Seems running with SETUID. This is insecure and highly discouraged: See README.md");
   }
 
-  if (sigsetjmp(jmpbuf, 1) != 0) {
-    goto done;
-  }
-  signal(SIGHUP, signalhandler);
-  signal(SIGINT, signalhandler);
-  signal(SIGTERM, signalhandler);
-
-  // We will receive EPIPE on the socket.
-  signal(SIGPIPE, SIG_IGN);
-
   int pidfile_fd = -1;
   if (cliopt->pidfile != NULL) {
     pidfile_fd = create_pidfile(cliopt->pidfile);
@@ -432,6 +422,16 @@ int main(int argc, char *argv[]) {
     goto done;
   }
 
+  if (sigsetjmp(jmpbuf, 1) != 0) {
+    goto done;
+  }
+  signal(SIGHUP, signalhandler);
+  signal(SIGINT, signalhandler);
+  signal(SIGTERM, signalhandler);
+
+  // We will receive EPIPE on the socket.
+  signal(SIGPIPE, SIG_IGN);
+
   state.sem = dispatch_semaphore_create(1);
 
   // Queue for vm connections, allowing processing vms requests in parallel.