Chaînage automatisé de programmes BPF par application de correctifs

Abstract

BPF is an important technology at the core of the Linux kernel that has received an increase in attention and contribution from multiple fields of computer science like networking and security. BPF allows execution of precompiled programs directly inside the Linux kernel for significant performance improvements over other approaches. These programs can monitor a running system, filter network packets, route network packets, etc. In many use cases, there is a need to run multiple orthogonal programs sequentially, like packet filtering or packet routing. Allowing the automation of program chaining will allow even more notable improvements to the ecosystem surrounding this technology by improving how programs can be coupled together to acheive more complex use cases. In this work, we propose a novel approach to chain the execution of BPF programs without harming performance and offering better compatibility compared to similar approaches. We compare our novel approach with XDP Dispatcher and manual program chaining. For all approaches, we compare performances by measuring the transfer speed, run time and CPU cycles used for multiple LTS versions of the Linux kernel. We show that our approach is a viable alternative that does not suffer from performance costs and argue its boons on more subjective matters like compatibility, program orthogonality, etc. We hope this work will aid in the development of new tools in the BPF ecosystem.

Date	26 Sept 2023
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Abdelouahed Gherbi (Supervisor)