Core Scheduling - Taming Hyper-Threads to Be Secure

Explore a comprehensive conference talk from the Linux Plumbers Conference that delves into the development, testing, and profiling efforts of core scheduling in the Linux community. Learn about the concept of core scheduling as a solution to security vulnerabilities in CPU architecture, particularly those targeting hyperthreading (SMT). Discover how this approach aims to keep SMT enabled while ensuring that only mutually trusted applications run concurrently on a core. Examine various proof-of-concept designs, their implementation details, and performance implications. Gain insights into the profiling techniques used to assess the correctness and performance of core scheduling patches, including the use of kernel features for obtaining time-sensitive data. Understand the challenges in defining security boundaries, policy implementation, and performance optimization in core scheduling. Analyze early performance results and explore future directions for this critical security feature in Linux systems.

Intro
A brief history of side-channel attacks
Core Scheduling: Concepts
Core Scheduling : task match
Core Scheduling: History
Core Scheduling: KVM based approach
Core Scheduling Generic Approach
Core Scheduling Implementation details
Core Scheduling: Iterations
Core Scheduling: Implementation Issues
Core Scheduling: vruntime comparison
Vruntime comparison corner cases after normalization
Forced idle corner case example
Proposed Solutions
Testing methodology
Co-scheduling stats example
Performance validation
Early performance results: CPU
Early performance results: mixed resources
Core Scheduling : Post v3 and beyond