Detect Peripheral Hardware Faults Using I / O - state - based Dynamic Value Invariants ∗

Hardware system has been expected to become increasingly vulnerable to faults due to continuously increasing function complexity and decreasing feature size. Using I/O-state-based dynamic value invariants, one of software visible symptoms, can probabilistically detect peripheral hardware faults. This paper explores a software solution that watches for anomalous dynamic value invariant behaviors to indicate the presence of peripheral hardware faults with low cost. The approach extracts I/Ostate-based dynamic value invariants of real commodity software, and detects faults by checking any data inconsistencies arising in an application’s behavior. We implemented the proof of concept in a full system simulator Bochs-P86. The experimentation with Windows XP shows that the approach is effective in detecting peripheral hardware faults. Four forms of dynamic value invariants all have over 46% coverage rate, detect more than 60% faults within 1000 instructions latency, and achieve less than 1.2% false positive rate.