ISSN 1000-1239 CN 11-1777/TP

Journal of Computer Research and Development ›› 2021, Vol. 58 ›› Issue (3): 651-667.doi: 10.7544/issn1000-1239.2021.20200186

Previous Articles     Next Articles

Kernel Configuration Infographic Based on Multi-Label and Its Application

Hou Pengpeng1,2, Zhang Heng1, Wu Yanjun1, Yu Jiageng1, Tai Yang1, Miao Yuxia1   

  1. 1(Intelligent Software Research Center (Institute of Software, Chinese Academy of Sciences), Beijing 100190);2(University of Chinese Academy of Sciences, Beijing 100049)
  • Online:2021-03-01
  • Supported by: 
    This work was supported by the National Natural Science Foundation of China (62002350), the Strategic Priority Research Program of Chinese Academy of Sciences (XDC05040200), and the Key-Area Research and Development Program of Guangdong Province (2019B010154004).

Abstract: The Linux kernel provides flexible configuration items to customize the kernel for various application scenarios. However, the number of kernel configuration items is huge and growing rapidly, and the values of configuration items often change in different kernel versions, even professional kernel teams face many challenges when setting the values of configuration items. This paper presents an infographic containing a variety of information for kernel configuration items. The infographic contains the dependencies among configuration items, function labels, performance labels, security labels, and configuration item enable rates. In addition, the infographic provides a visualization interface, which is more intuitive, efficient and user-friendly. The infographic can be widely used in scenarios such as kernel startup optimization, kernel size tailoring, kernel security enhancement, kernel performance optimization, kernel configuration item abnormality detection, kernel configuration item intelliqient question and answering, and kernel configuration item recommendation. To verify the validity of the infographic, we have designed a configuration item-oriented retrieval framework KCIR based on the infographic, which implements query expansion based on multi-label information and text extension based on the dependencies between kernel configuration items, and our experiments demonstrate that KCIR is more effective than the traditional retrieval frameworks. The use of the infographic in the information retrieval field illustrates its effectiveness and practicality.

Key words: kernel config, multi-label, Linux kernel, information retrieval, visualization

CLC Number: