Blasting Linux Code
Computer programs can only run reliably if the underlying operating system is free of errors. In this paper we evaluate, from a practitioner’s point of view, the utility of the popular software model checker Blast for revealing errors in Linux kernel code. The emphasis is on important errors related to memory safety in and locking behaviour of device drivers. Our conducted case studies show that, while Blast’s abstraction and refinement techniques are efficient and powerful, the tool has deficiencies regarding usability and support for analysing pointers, which are likely to prevent kernel developers from using it.
Unable to display preview. Download preview PDF.