Detection of ESD-Induced Soft Failures by Analyzing Linux Kernel Function Calls

Electrostatic discharge (ESD) into a functioning system can cause temporary upsets - soft failures. Subtle soft failures can reduce the reliability of system and cannot be detected by conventional methods based on external equipment or operation system (OS) log. This paper presents a proof of concept for a novel methodology for detection of ESD-induced soft failures through analyzing kernel function trace recordings of the operation system. The method is based on recording Linux kernel function calls during normal operation and after ESD stress injection. The recorded information is visualized in forms of graphical maps of function execution and system call distribution for each process to highlight ESD induced changes. The experimental data shows that soft failures manifest themselves as changes in the function maps and the call distribution within the observed processes. This novel method is capable of detecting subtle system upsets which are not observable for the user through standard I/O or attached equipment.