Deterministic Replay for Multi-Core VxWorks Applications
Junjie Liu, Xiaopeng Gao, Bo Jiang, Shunkun Yang, Zhenyu Zhang · 2017
Debugging is an important and yet expensive activity in software development. With the adoption of multi-core computing platform, the debugging of multi-core bugs on real-time platform becomes challenging. Those multi-core concurrency bugs are often called Heisenbugs, because they manifest randomly during replay. As a result, it would be difficult for a programmer to use cyclic debugging to locate the fault. In this work, we present a deterministic replay tool for multi-core real-time VxWorks applications. We combine static analysis and source code instrumentation to help record interleaved access to shared memory for multi-tasking VxWorks applications. Our experimental results show that our tool can effectively reply multi-core real-time VxWorks applications with reasonable overhead.