Understanding and Detecting Disordered Error Handling with Precise Function Pairing


Qiushi Wu, Aditya Pakki, Navid Emamdoost, Stephen McCamant, and Kangjie Lu, University of Minnesota


Software programs may frequently encounter various errors such as allocation failures. Error handling aims to gracefully deal with the errors to avoid security and reliability issues, thus it is prevalent and vital. However, because of its complexity and corner cases, error handling itself is often erroneous, and prior research has primarily focused on finding bugs in the handling part, such as incorrect error-code returning or missing error propagation.

In this paper, we propose and investigate a class of bugs in error-handling code from a different perspective. In particular, we find that programs often perform "cleanup" operations before the actual error handling, such as freeing memory or decreasing refcount. Critical bugs occur when these operations are performed (1) in an incorrect order, (2) redundantly, or (3) inadequately. We refer to such bugs as Disordered Error Handling (DiEH). Our investigation reveals that DiEH bugs are not only common but can also cause security problems such as privilege escalation, memory corruption, and denial-of-service. Based on the findings from the investigation, we then develop a system, HERO (Handling ERrors Orderly), to automatically detect DiEH. The core of HERO is a novel technique that precisely pairs both common and custom functions based on the unique error-handling structures, which allows us to infer expected cleanup functions. With HERO, we found 239 DiEH bugs in the Linux kernel, the FreeBSD kernel, and OpenSSL, which can cause security and reliability issues. The evaluation results show that DiEH is critical and widely exists in system software, and HERO is effective in detecting DiEH. We also believe that the precise function pairing is of independent interest in other research areas such as temporal-rule inference and race detection.

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.

@inproceedings {263886,
title = {Understanding and Detecting Disordered Error Handling with Precise Function Pairing},
booktitle = {30th {USENIX} Security Symposium ({USENIX} Security 21)},
year = {2021},
address = {Vancouver, B.C.},
url = {https://www.usenix.org/conference/usenixsecurity21/presentation/wu-qiushi},
publisher = {{USENIX} Association},
month = aug,