Data Flow Analysis for Sequential Storage Structures

Sequential storage structures such as array and continuous memory block allocated dynamically by malloc are widely used in C programs. But traditional data flow analysis fails to adequately describe their structures and operations. When pointers are used to access the sequential storage structures in C programs, existing data flow analysis methods basically pay attention to only points-to information and do not consider the numerical properties offset. In addition, it does not consider the unsafe problem caused by out of bounds when offset occurs, which leads to inaccurate analysis for sequential storage structure. To improve the precision for analyzing sequential storage structures, an abstract memory model SeqMM is proposed to describe sequential storage structures, which can effectively describe points-to relationships and offset. Then there are three operations are summarized, such as the pointer-related transfer operation, predicate operation, and loop operation traversing sequential storage structures, and it is also considered that whether the index is out of bounds to ensure the security of operation execution when analyzing these operations. After that, mapping rules are introduced for parameters referencing sequential storage structure to corresponding arguments. Finally, a memory leak detection algorithm is proposed to detect memory leak in 5 open-source projects. The experimental results indicate that SeqMM can effectively describe sequential storage structure and various operations in C programs, and the results of data flow analysis can be used to detect memory leaks when a reasonable balance between accuracy and efficiency occurs. © Copyright 2020, Institute of Software, the Chinese Academy of Sciences. All rights reserved.