Boosting Software Fault Injection for Dependability Analysis of Real-Time Embedded Applications

The design of complex embedded systems deployed in safety-critical or mission-critical applications mandates the availability of methods to validate the system dependability across the whole design flow. In this article we introduce a fault injection approach, based on loadable kernel modules and running under the Linux operating system, which can be adopted as soon as a running prototype of the systems is available. Moreover, for the purpose of decoupling dependability analysis from hardware availability, we also propose the adoption of hardware virtualization. Extensive experimental results show that statistical analysis made on top of virtual prototypes are in good agreement with the information disclosed by fault detection trends of real platforms, even under real-time constraints.