Using model checking to debug device firmware

Author:

Kumar Sanjeev1,Li Kai1

Affiliation:

1. Princeton University

Abstract

Device firmware is a piece of concurrent software that achieves high performance at the cost of software complexity. They contain subtle race conditions that make them difficult to debug using traditional debugging techniques. The problem is further compounded by the lack of debugging support on the devices. This is a serious problem because the device firmware is trusted by the operating system.Model checkers are designed to systematically verify properties of concurrent systems. Therefore, model checking is a promising approach to debugging device firmware. However, model checking involves an exponential search. Consequently, the models have to be small to allow effective model checking.This paper describes the abstraction techniques used by the ESP compiler to extract abstract models from device firmware written in ESP. The abstract models are small because they discard some of the details in the firmware that is irrelevant to the particular property being verified. The programmer is required to specify the abstractions to be performed. The ESP compiler uses the abstraction specification to extract models conservatively. Therefore, every bug in the original program will be present in the extracted model.This paper also presents our experience with using Spin model checker to develop and debug VMMC firmware for the Myrinet network interfaces. An earlier version of the ESP compiler yielded models that were too large to check for system-wide properties like absence of deadlocks. The new version of the compiler generated abstract models that were used to identify several subtle bugs in the firmware. So far, we have not encountered any bugs that were not caught by Spin.

Publisher

Association for Computing Machinery (ACM)

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3