Abstract

Most of the software model checker tools do not scale well on complicated software. Our goal was to develop a tool, which provides an adjustable balance between precise and slow software model checkers and fast and imprecise static analyzers. The key idea of the approach is an abstraction over the precise thread interaction and analysis for each thread in a separate way, but together with a specific environment, which models effects of other threads. The environment contains a description of potential actions over the shared data and synchronization primitives, and conditions for its application. Adjusting the precision of the environment, one can achieve a required balance between speed and precision of the complete analysis. A formal description of the suggested approach was performed within a Configurable Program Analysis theory. It allows formulating assumptions and proving the soundness of the approach under the assumptions. For efficient data race detection we use a specific memory model, which allows to distinguish memory domains into the disjoint set of regions, which correspond to a data types. An implementation of the suggested approach into the CPAchecker framework allows reusing an existed approaches with minimal changes. Implementation of additional techniques according to the extended theory allows to increase the precision of the analysis. Results of the evaluation allow confirming scalability and practical usability of the approach.

中文翻译：

---

操作系统组件正确同步的分析

摘要

大多数软件模型检查器工具在复杂软件上无法很好地扩展。我们的目标是开发一种工具，在精确和缓慢的软件模型检查器与快速和不精确的静态分析器之间提供可调整的平衡。该方法的关键思想是对每个线程的精确线程交互和分析的抽象，以一种单独的方式进行抽象，但要结合一个对其他线程的效果进行建模的特定环境。该环境包含对共享数据和同步原语可能采取的措施的描述，以及对其应用的条件。调整环境的精度，可以在完整分析的速度和精度之间达到所需的平衡。在可配置程序分析理论中对建议的方法进行了正式描述。它允许制定假设并在假设下证明方法的合理性。为了进行有效的数据竞争检测，我们使用特定的内存模型，该模型可以将内存域区分为不相交的区域集，这些区域对应于数据类型。将建议的方法实施到CPAchecker框架中后，可以以最小的更改重用现有方法。根据扩展理论实施其他技术可以提高分析的准确性。评估结果可以确认该方法的可扩展性和实际可用性。对应于一种数据类型。将建议的方法实施到CPAchecker框架中后，可以以最小的更改重用现有方法。根据扩展理论实施其他技术可以提高分析的准确性。评估结果可以确认该方法的可扩展性和实际可用性。对应于一种数据类型。将建议的方法实施到CPAchecker框架中后，可以以最小的更改重用现有方法。根据扩展理论实施其他技术可以提高分析的准确性。评估结果可以确认该方法的可扩展性和实际可用性。