Method and system for detecting parallel program shared data class bug

Abstract

The invention provides a method for detecting a parallel program shared data class bug. The method includes the steps of defining instant Fetter sets for dynamic examples of all instructions in software invariants, obtaining containing sets and obligation sets of the instant Fetter sets, conducting pruning on the containing sets and the obligation sets to obtain pruned results, storing the pruned results in an invariant database, detecting each instant Fetter set through a detection assembly to find out whether the first instant Fetter set which does not satisfy the corresponding containing set or the corresponding obligation set occurs or not, if yes, judging that the first instant Fetter set contains the bug and generating a violation report through the detection assembly, and suspending a thread which runs at present so that the detected bug can be prevented from running, wherein the software invariants refer to instruction sets with the pre-fixed execution sequence in a program. According to the method, the potential shared data class bug in the program can be automatically detected and prevented. The invention further provides a system for detecting the parallel program shared data class bug.

Description

technical field [0001] The invention relates to the technical field of software engineering, in particular to a method and system for detecting loopholes in parallel program sharing data. Background technique [0002] Generally speaking, the research on data race type vulnerabilities can be divided into two parts, lock-set approach (such as Eraser) and time series-based method (happened-before) (such as SigRace). Although data race vulnerabilities have been well researched, it is not enough to solve all data race vulnerabilities perfectly. The reasons are: 1) A large part of data race vulnerabilities do not cause any substantial consequences. That is, they are benign; 2) Atomic violations without data races are not uncommon; 3) The vast majority of sequential violations cannot be resolved by simply wrapping them in a critical section. [0003] Due to the limitation of tools for detecting data race vulnerabilities, the research on how to solve atomicity violation vulnerabil...

AI Technical Summary

Problems solved by technology

There are still the following problems with this approach: 1) Even after fairly adequate testing, there is still no guarantee that all vulnerabilities have been found; 2) Misinterpretation of a reported vulnerability may lead to the introduction of new vulnerabilities by fixes ;3) The time interval between the discovery of the vulnerability and its fix is ​​too long

However, despite the use of a new programming model, the vulnerability has not been eliminated

Images