MLEE: Effective Detection of Memory Leaks on Early-Exit Paths in OS Kernels

Authors: 

Wenwen Wang, University of Georgia

Abstract: 

Memory leaks in operating system (OS) kernels can cause critical performance and security issues. However, it is quite challenging to detect memory leaks due to the inherent complexity and large-scale code base of real-world OS kernels. In this work, inspired by the observation that software bugs are often hidden in rarely-tested program paths, we focus on detecting memory leaks on early-exit (E-E) paths in OS kernels. To this end, we conduct a systematic study of memory management operations involved on E-E paths in OS kernels. Based on the findings, we design a novel leak detector for OS kernels: MLEE, which intelligently discovers memory leaks on E-E paths by cross-checking the presence of memory deallocations on different E-E paths and normal paths. MLEE successfully reports 120 new memory leak bugs in the Linux kernel. It is the first time these memory leaks are uncovered by a leak detector for OS kernels.

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 {273857,
author = {Wenwen Wang},
title = {{MLEE}: Effective Detection of Memory Leaks on {Early-Exit} Paths in {OS} Kernels},
booktitle = {2021 USENIX Annual Technical Conference (USENIX ATC 21)},
year = {2021},
isbn = {978-1-939133-23-6},
pages = {31--45},
url = {https://www.usenix.org/conference/atc21/presentation/wang-wenwen},
publisher = {USENIX Association},
month = jul
}

Presentation Video