Explore nested virtualization on the IBM POWER9 processor in this 46-minute conference talk from linux.conf.au. Dive into the rationale, implementation details, and challenges of developing software support for running virtual machines within virtual machines. Learn about the changes made to KVM and QEMU, the current state of the project, and future work. Discover the increased hardware support for nested virtualization in POWER9, understand the entry path and memory management processes, and compare KVM-PR and KVM-HV approaches. Gain insights into nested guest address translation, partition scoped PTE generation, and migration of nested guests. Explore performance considerations and find out just how deep nesting can go. Get practical information on how to use this feature and learn about ongoing development efforts.
Overview
Syllabus
Intro
Nested KVM on the POWER9 Processor
What is POWERS
KVM Relevant Changes in POWER9
What is KVM?
Some Terminology
Process of Running a Guest
Entry Path
Memory Management
What is Nested KVM?
KVM-PR VS KVM-HV
Why would you want to do this?
Nested KVM-HV - The Puzzle Pieces
Nested Entry/Exit Path
Nested Guest Address Translation
Partition Scoped PTE Generation
Process Scoped Invalidations
Partition Scoped Invalidations
Nested Nested
Migration of Nested Guests
Migration Between Levels
Passthrough of Emulated MMIO
Performance
But How Deep can I Nest?
State of the Code
How to Use It
Future Work
Taught by
linux.conf.au
