A Linux kernel boot test and configuration consistency intelligent repair method oriented towards configuration item repair
By generating candidate configuration files, parsing structured logs, modeling Kconfig semantics, and making automated repair decisions, the complexity and stability issues in the Linux kernel configuration item repair process are resolved, achieving efficient and reliable configuration consistency repair and stability verification.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING UNIV OF INFORMATION SCI & TECH
- Filing Date
- 2026-03-31
- Publication Date
- 2026-06-30
AI Technical Summary
In the process of fixing Linux kernel configuration items, existing technologies have problems such as complex configuration constraints, difficulty in error location, reliance on manual experience for repair, high iteration costs, insufficient stability assurance, and lack of engineering-based methods for locating cross-version regressions.
Candidate configuration files are generated using randomization or batching strategies. A closed-loop approach is adopted, combining reproducible startup testing, Kconfig semantic constraint modeling, automated repair decision-making, and stability verification. Automated repair is achieved through structured log parsing, unified dependency graph construction, and configuration item classification.
It significantly improves the efficiency and stability of configuration item repair, reduces reliance on manual experience, reduces redundant attempts, increases repair convergence speed, and reduces runtime instability risks through coverage-guided kernel fuzzing, while providing an engineering-based path for cross-version regression localization.
Smart Images

Figure CN121935167B_ABST