Verification in the Light and Large: Large-Scale Verification for Fast-Moving Open Source C Projects

Abstract

A lightweight approach to the verification of software at large scales, called "symbolic approximation", has been developed in order to achieve post-hoc verification of C code in fast-moving open source projects such as the Linux kernel. While other approaches such as model-checking were known to be technically sound beforehand and were subsequently scaled up to treat larger problems, symbolic approximation was created to first of all work expediently at very large scales and absolute formal correctness is being adduced only afterwards. In this paper we report just a few new details and advances in the technology but supply new global reasoning supporting its correctness.