LinKRID: Vetting Imbalance Reference Counting in Linux kernel with Symbolic Execution

Authors: 

Jian Liu, {CAS-KLONAT, BKLONSPT}, Institute of Information Engineering, Chinese Academy of Sciences and School of Cyber Security, University of Chinese Academy of Sciences; Lin Yi, {CAS-KLONAT, BKLONSPT}, Institute of Information Engineering, Chinese Academy of Sciences; Weiteng Chen, Chengyu Song, and Zhiyun Qian, UC Riverside; Qiuping Yi, Beijing University of Posts and Telecommunications and Beijing Key Lab of Intelligent Telecommunication Software and Multimedia

Abstract: 

Linux kernel employs reference counters, which record the number of references to a shared kernel object, to track its lifecycle and prevent memory errors like use-after-free. However, the usage of reference counters can be tricky and often error-prone, especially considering unique kernel conventions of managing reference counters (e.g., external vs. internal reference counters). In this paper, we aim to automatically discover incorrect usage of reference counters, overcoming two key challenges: (1) scalability and (2) the aforementioned unique kernel conventions. Specifically, we develop a tiered program analysis based solution to efficiently and precisely check the imbalances between the change in the actual number of references and the corresponding reference counter. We apply our tool to the 4.14.0 kernel (with allyesconfig) and find 118 bugs, out of which 87 are new. The result shows our tool is scalable and effective.

Open Access Media

USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.

BibTeX
@inproceedings {277162,
author = {Jian Liu and Lin Yi and Weiteng Chen and Chengyu Song and Zhiyun Qian and Qiuping Yi},
title = {{LinKRID}: Vetting Imbalance Reference Counting in Linux kernel with Symbolic Execution},
booktitle = {31st USENIX Security Symposium (USENIX Security 22)},
year = {2022},
isbn = {978-1-939133-31-1},
address = {Boston, MA},
pages = {125--142},
url = {https://www.usenix.org/conference/usenixsecurity22/presentation/liu-jian},
publisher = {USENIX Association},
month = aug
}

Presentation Video