Software Based Error Recovery Techniques For Aerospace Applications At Source Code Level

Transient faults in the circuit caused by the radiation of high energetic particles, known as soft errors, has severely affected the reliability and lifespan of the device of aircraft. To solve the problem and improve the reliability of aerospace application, hardware-enhanced devices with radiation resistance are widely used on the aerospace computing platform. However, there are also constraints of these devices in many aspects, such as design, material and consumption, which leads to the high cost.Compared with the hardware technology, software technology to reinforce the program does not need to modify hardware architecture and reduce the consumption and cost. Meanwhile performance issues can be ensured by using commercial product. Therefore software fault-tolerance technique gets widespread attention among the aerospace sectors and becomes an important issue in research. It is of great significance for the development of our aerospace industry to study the software fault-tolerance technique against the hardware transient faults.Based on the software fault-tolerance method, the paper concentrates on the program recovery of source-code level. The main work is as follows:1. We proposed a control flow recovery method, SICER, on source-code level. SICER can effectively detect the incorrect jump of inter-functions, inter-blocks and intra-block, and recover from the error of inter-blocks and intra-block. The method is based on the label detection and redundancy recovery with different levels inside and out of functions. In this way, configurability of the method can be improved by different division of basic blocks.2. We propose a data flow recovery method, SIDER, based on the data coding and checking. SIDER is a forward recovery method with less time overhead to redundant data for recovery, due to the high-efficiency of proper encoding algorithm compared to normal backup. Moreover, SIDER uses a new variable criticality analysis method to guarantee the configurability and set protection priorities for the variables on condition of limited performance overhead.3. We reinforce the source code of a certain aerospace application with the recovery method in the paper. To test the effectiveness of our method, we conduct the overhead analysis and the fault injection experiment based on the CCStudio, for which we design and implement the RFISD(Random Fault Injection with Script Debug). The result shows that the reliability of the target application is greatly improved with an error reduction of 41% and the overhead in time and space is controlled with 140.2% and 230.9%.