Memory leak detection runtime-service for embedded Linux devices

Abstract

In practice, software is often deployed with several hidden bugs despite various test processes and static analyses. Such bugs often cause a phenomenon called software aging that refers to the accumulation of errors occurring in long running software systems that results in a decrease of performance and an increases of the probability to crash the entire system. This paper presents a mechanism to detect memory leaks in embedded systems software at runtime that can be used to counter software aging. In particular, it presents and compares two measurement based algorithms to identify memory leaks on Embedded Linux devices at runtime. Both algorithms have been implemented and evaluated using an industrial room controller targeting building automation. Based on this approach we were able to identify an existing memory leak we were unaware of beforehand.