基于变量访问序模式的中断数据竞争检测方法

发布时间：2018-07-29 20:09

【摘要】：在航天嵌入式软件等中断驱动型软件中,中断数据竞争问题十分突出.然而,中断在并发语义、同步机制、调度机制等方面与线程(任务)有诸多不同,具有Ad-hoc特征,难以统一刻画,因此,主流的数据竞争检测方法并不适用.以航天嵌入式软件数据竞争案例库为基础进行了系统分析,提出刻画有害中断数据竞争的7种缺陷模式.针对其中最常见且最难解决的单变量访问序模式,基于抽象解释,提出一种支持过程间分析、中断并发分析的高效检测方法.设计并实现了相应的检测工具Space DRC.实验结果表明,Space DRC能够在145ms内检测出约21 400行程序中的真实数据竞争.Space DRC已经在多个航天重点型号中进行了应用,使得中断数据竞争专项分析的效率提高了至少5倍,并且降低了问题遗漏率.
[Abstract]:In space embedded software and other interrupt driven software, interrupt data competition is very prominent. However, interrupt is different from thread (task) in many aspects such as concurrent semantics, synchronization mechanism, scheduling mechanism and so on. It has the characteristics of Ad-hoc and is difficult to describe uniformly. Therefore, the mainstream data competition detection method is not suitable. Based on the case base of data competition in aerospace embedded software, seven defect models are proposed to describe the harmful interrupt data competition. Aiming at the most common and most difficult to solve the single variable access order pattern, an efficient detection method based on abstract interpretation is proposed to support inter-process analysis and interrupt concurrent analysis. The corresponding detection tool Space DRC is designed and implemented. The experimental results show that Space DRC can detect real data competition in about 21,400 lines of programs in 145ms. Space DRC has been applied in many key spaceflight models, and the efficiency of special analysis for interrupting data competition has been improved by at least five times. It also reduces the problem omission rate.
【作者单位】：北京控制工程研究所;北京轩宇信息技术有限公司;中国空间技术研究院;
【基金】：国家自然科学基金(91118007)~~
【分类号】：TP311.11

【相似文献】