In this paper, we propose a new technique, called lazy locking technique, for improving real-time performance of embedded Linux. It employs the policy that entering a critical section is allowed only when the operation does not disturb future execution of the real-time application. To execute this policy, we introduce timer interrupt prediction and lock hold time acquisition. In addition, we newly designed and implemented a simply, yet efficient high resolution timer. We implemented the prototype on Linux 2.4.18. Experimental results show that worst-case OS latency of real-time process is reduced to 23% of the original one, at the expense of slowdown of non-real-time process by 20%. Though we focus only on embedded Linux, our technique is useful to all kinds of real-time operating system in which critical section is significantly long.