论文部分内容阅读
硬件瞬态故障是运行于高辐照宇宙环境下的航天计算机面临的最主要挑战之一.其中,控制流错误是这种故障影响系统可靠性的重要体现.首先在程序控制流图的基础上,利用图着色算法对基本块进行分类,然后基于基本块的格式化标签提出一种有效的控制流检测方法ECCFS,并针对基本块内部和过程间两个控制流检测问题分别给出其扩展解决方法.检测效能分析和故障注入实验的结果都表明,除了伪分支和几个边界检测盲点之外,ECCFS能够检测出绝大部分的控制流错误.与两个有代表性的控制流检测方法相比较,ECCFS在错误检测率和性能开销等方面都具有一定的优势.
The hardware transient fault is one of the most important challenges that a spacecraft operating in a high-radiation cosmic environment is faced with. Control flow error is an important manifestation of the reliability of this fault affecting the system.Firstly, based on the flow chart of program control, Graphical coloring algorithm is used to classify the basic blocks. Then, an effective control flow detection method ECCFS is proposed based on the basic block formatted tags. An extended solution is given respectively to two control flow detection problems in the basic block and between processes. The results of both detection performance analysis and fault injection experiments show that ECCFS can detect most control flow errors except for the fake branch and several border detection blind spots.Compared with the two representative control flow detection methods, ECCFS has some advantages in terms of error detection rate and performance overhead.