Taming Deadlocks in Multithreaded Programs

Abstract

Many real-world multithreaded programs contain deadlock bugs. These bugs should be detected and corrected. Many existing detection strategies are not consistently scalable to handle large-scale applications. Many existing dynamic confirmation strategies may not reveal detectable deadlocks with high probability. And many existing runtime deadlock-tolerant strategies may incur high runtime overhead and may not prevent the same deadlock from re-occurring. This paper presents the current progress of our project on dynamic deadlock detection, confirmation, and resolution. It also describes a test harness framework developed to support our proposed approach.