Fix container exit detection in systemd scope environments

When container processes are not direct children of conmon conmon fails to
detect container exits because it never receives SIGCHLD signals.

This fixes issue where conmon processes remain running after container exit
in certain systemd cgroup manager configurations.

Fixes: #545

Signed-off-by: Jindrich Novy <jnovy@redhat.com>

Author: jnovy

src/ctr_exit.c

@@ -46,6 +46,25 @@ static void check_child_processes(GHashTable *pid_to_handler, GHashTable *cache)
 			continue;
 
 		if (pid < 0 && errno == ECHILD) {
+			/* Before quitting, check if container_pid is still alive.
+			 * In some systemd configurations, the container process may not be
+			 * a direct child, so we won't receive SIGCHLD when it exits.
+			 * Use kill(pid, 0) to check if the process still exists. */
+			if (container_pid > 0) {
+				if (kill(container_pid, 0) == 0) {
+					/* Container process is still alive but not our child.
+					 * Don't quit the main loop yet. */
+					ninfof("Container process %d is still alive but not a direct child", container_pid);
+					return;
+				} else if (errno == ESRCH) {
+					/* Container process has exited */
+					ninfof("Container process %d has exited (detected via kill probe)", container_pid);
+					/* Simulate container exit callback */
+					container_status = 0; /* We can't get the real exit status */
+					container_pid = -1;
+					/* Fall through to quit the main loop */
+				}
+			}
 			g_main_loop_quit(main_loop);
 			return;
 		}