A runtime fault analysis method in embedded software using signal hooking and information tagging

Reproducing or debugging a fault in embedded software is a difficult activity, due to the complicated operating environment and various input data that have an effect on the software execution path. We propose a testing method to identify fault type and locate debugging points by detecting faults while the system is running, rather than analyzing faults through reproducing them after program fail. Our method is hooking the execution of modules related to runtime errors, selectively collecting execution information that is necessary to detect a fault. We have developed an automated testing tool called RopheTR, and performed experiments to compare with representative runtime error detection tools and RopheTR. Experimental results show that RopheTR has the highest fault detection rate of 91.8%, and has excellent accuracy in providing code location for debugging and fault type, compared with other tools. Moreover, our method exhibited the least slow down rate of 0.03x on average, and the slightest memory usage increase of 0.13x. Our method is a very light-weight test method suitable for embedded software that has severe resource constraints of real-time operating system.