idle(): recheck runq just before cpu status check

There is a problem which causes wait4(2) hang when wait4(2) called by a
process races with its child process termination.
This is a quick fix for this problem. See idle().

refs #271

kernel/process.c

@@ -1789,6 +1789,38 @@ static void idle(void)
 	while (1) {
 		schedule();
 		cpu_disable_interrupt();
+		/*
+		 * XXX: KLUDGE: It is desirable to be resolved in schedule().
+		 *
+		 * There is a problem which causes wait4(2) hang when
+		 * wait4(2) called by a process races with its child process
+		 * termination. This is a quick fix for this problem.
+		 *
+		 * The problem occurrd in the following sequence.
+		 * 1) The parent process called schedule() from sys_wait4() to
+		 *    wait for an event generated by the child process.
+		 * 2) schedule() resumed the idle process because there was no
+		 *    runnable process in run queue.
+		 * 3) At the moment, the child process began to end. It set
+		 *    the parent process runnable, and sent an interrupt to
+		 *    the parent process's cpu. But this interrupt had no
+		 *    effect because the parent process's cpu had not halted.
+		 * 4) The idle process was resumed, and halted for waiting for
+		 *    the interrupt that had already been handled.
+		 */
+		if (v->status == CPU_STATUS_IDLE) {
+			long s;
+			struct process *p;
+
+			s = ihk_mc_spinlock_lock(&v->runq_lock);
+			list_for_each_entry(p, &v->runq, sched_list) {
+				if (p->status == PS_RUNNING) {
+					v->status = CPU_STATUS_RUNNING;
+					break;
+				}
+			}
+			ihk_mc_spinlock_unlock(&v->runq_lock, s);
+		}
 		if (v->status == CPU_STATUS_IDLE) {
 			cpu_safe_halt();
 		}