A Hybrid Testing System on P4 Cloud Gateway

Xin Yang, Enge Song, Fuliang Li, Wanli Cao, Tian Pan, Yang Song, Ye Yang, Yongwang Wu, Jianyuan Lu, Bowen Yang, Xing Li, Biao Lyu, Rong Wen, Xingwei Wang, Shunmin Zhu

Abstract: P4 gateway is the hub of cloud network, and with the growth of cloud traffic scale, the reliability of a cloud gateway must be guaranteed by testing. The existing cloud gateway testing solutions, such as symbolic execution and fuzzing, perform either insufficient scalability or poor coverage, leading to difficulties when deploying in production. To balance scalability and coverage, we present a hybrid testing system for P4 gateway that uses path probability as a tradeoff metric to combine fuzzing and symbolic execution. In addition, we applied a graph centrality strategy and intermediate result reuse operations to improve efficiency, and implemented a lightweight instrumentation tool for measuring coverage. We conducted evaluations on both open-source and production-scale P4 programs. The evaluations indicate that the prototype system achieved 100% coverage on most programs with high efficiency. At present, the system is still undergoing continuous testing and optimization in production.