Using fault injection for testing Linux kernel components

Abstract

The paper presents methods aimed to extend coverage of existing tests by systematic and targeted fault injection in Linux kernel. The main goal is to test if kernel components correctly handle abnormal situations. As long as such situations happen seldom and unpredictably, it is difficult to debug bugs in the code handling them. Currently the only approach used in practice is random fault injection. The paper presents new methods that inject faults systematically. It allows to achieve better coverage and to make test results more predictable. The paper describes results of evaluation of the methods proposed and random fault injection.