Transient Fault Detection and Recovery Mechanisms in μC/OS-II

Abstract

In avionics, satellites are widely used in meteorology, navigation and investigation. Satellites in space, however, are subject to radiation that causes transient fault. This often leads to single event upset on the logic state of device, undermining the stability and the correctness of the system. For example, transient fault can cause errors in the program execution flow, changing the state of or even crashing the system. In order to solve these problems, this paper puts forward a coarse-grained error detection scheme based on function-call relationships. We instrument signature codes at function entry and exit points at compile time to perform dynamic detection at runtime. We apply this method in the μC/OS-II kernel on a DSP platform. The coarse-grained error detection technology can reduce storage overhead effectively compared with basic block-based detection technology. For the moment, this method could be used in imbedded operating systems μC/OS-II, and it can simulate a program flow error caused by transient fault with the method of fault injection. With this method, it can help to detect the occurrence of an error and guarantee the normal running of the system using recovery mechanism. Finally, the result shows that technology of transient fault detection which is based on function call relationship could detect errors effectively, which guarantees the reliability and security of the running system.