daemon: restore SIGTERM default handler in child process

process_io() loop doesn't check flag set by SIGTERM handler in the
parent qrexec-daemon process. This results in child processes (for
example handling socket service calls to dom0) effectively ignore
SIGTERM. This subsequently cause system shutdown to wait (until finally
sending SIGKILL).

Fix this by restoring default SIGTERM handler in the child process.

Fixes: f3a5784 "Avoid qrexec-client for VM -> dom0 calls"

Author: marmarek

daemon/qrexec-daemon.c

@@ -1202,6 +1202,7 @@ static void handle_execute_service(
 {
     int policy_pending_slot;
     pid_t pid;
+    struct sigaction sa = { .sa_handler = SIG_DFL };
 
     policy_pending_slot = find_policy_pending_slot();
     if (policy_pending_slot < 0) {
@@ -1217,6 +1218,8 @@ static void handle_execute_service(
         case 0:
             if (atexit(null_exit))
                 _exit(QREXEC_EXIT_PROBLEM);
+            if (sigaction(SIGTERM, &sa, NULL))
+                LOG(WARNING, "Failed to restore SIGTERM handler: %d", errno);
             handle_execute_service_child(remote_domain_id, remote_domain_name,
                                          target_domain, service_name, request_id);
             abort();