CAREER: Taming the size, complexity and longevity of OS kernels via enhanced OS kernel extensions

点击翻译按钮获取中文摘要

Operating system (OS) kernels are the most fundamental softwarecomponent society depends on, underpinning the computing industry,from the clouds and datacenters that manage our data to the smartphones and other devices we all use daily. Unfortunately, they havegrown to be complex and have not kept pace with advances inprogramming languages and modular software design, instead strugglingto meet modern security and reliability threats to our computinginfrastructure. Moreover, there appears to be no clear path toincrementally evolve them and their unchecked complexity has become asignificant barrier to entry for students and practitioners to learnor innovate at the kernel level. This project seeks to remedy thisissue by create a safe kernel extension framework within the legacy OSkernel that enables it to be replaced in part or in its entirety withsafe components. Through this extension framework, the project willenable a practical, incrementally-adoptable mechanism forundergraduate, graduate students and practitioners to evolve kernelsto be safer and more reliable, improving national security and leadingto higher productivity for society and the national economy.This project aims to achieve three goals for the OS kernel: securityand robustness through language safety, robustness and incrementalupdate through componentization, and practical relevance throughbuilding upon existing kernel extension frameworks. The key insight inour approach is that a safe kernel extension framework can be createdwithin the legacy OS kernel that enables any part of it to be replacedvia bypass extensions, evolving the kernel towards a safe,componentized architecture, or in its entirety with a safe andspecialized in-situ kernel. Our integrated research and educationplan involves three broad research thrusts. First, we will exploreexpressiveness/safety tradeoffs for kernel extensions, and will designa new Rust-based kernel extension environment that provides similarsafety guarantees with a more expressive programming environment thanthe popular eBPF kernel extension framework in Linux. Second, we willexplore to what extent our enhanced safe kernel extensions can replaceindividual kernel components with bypass extensions and explore newcomponent boundaries in Linux. Finally, we will explore to what extentan entirely new in-situ kernel can be created from extensions,ultimately in the context of Linux. Our educational plan integratesthe research in courses and other initiatives involving open sourceand industry.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.