Deadline Analysis of Interrupt-driven Software

Real-time, reactive, and embedded systems are increasingly used throughout society (e.g., flight control, railway signaling, vehicle management, medical devices, and many others). For real-time, interrupt-driven software, timely interrupt handling is part of correctness. It is vital for software verification in such systems to check that all specified deadlines for interrupt handling are met. Such verification is a daunting task because of the large number of different possible interrupt arrival scenarios. For example, for a Z86-based microcontroller, there can be up to six interrupt sources and each interrupt can arrive during any clock cycle. Verification of such systems has traditionally relied upon lengthy and tedious testing; even under the best of circumstances, testing is likely to cover only a fraction of the state space in interrupt-driven systems. This paper presents the Zilog architecture resource bounding infrastructure (ZARBI), a tool for deadline analysis of interrupt-driven Z86-based software. The main idea is to use static analysis to significantly decrease the required testing effort by automatically identifying and isolating the segments of code that need the most testing. Our tool combines multiresolution static analysis and testing oracles in such a way that only the oracles need to be verified by testing. Each oracle specifies the worst-case execution time from one program point to another, which is then used by the static analysis to improve precision. For six commercial microcontroller systems, our experiments show that a moderate number of testing oracles are sufficient to do precise deadline analysis