Unifying KVM API for protected VM and utilities
summary
Proposes adding VFIO-side region fragments so a portion of a device BAR gets a separate mapping with different attributes, driven by customer demand for low‑latency DPDK-style message buffers; plans RFC patches to expose this and validation hooks, with an option to unmap the original BAR fragment or map whole BARs instead. Notes architecture differences (x86 MTRRs/PAT, ARM behavior), risks around speculative reads and cacheability, and asks Intel/AMD/ARM sign‑off before relaxing guest restrictions.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 1:05Proposed solution for VFIO regions
- 3:35Use case and adapter specifics
- 5:23Challenges and open questions
- 8:12Guest and host mapping complexities
- 10:20Historical KVM restrictions
- 12:05Next steps and RFC patches
related talks
50mPanel Discussion: KVM-based Virtualization Contributor Q&A
36mShared Virtual Addressing for High Performance Arm Infrastructure Platforms
38mPanel Discussion: KVM-based Virtualization Contributor Q&A
26mCooperative DMA in a memory-oversubscribed environment
31mHardware Friendly Vhost vDPA: Towards an Efficient and Migratable Device Model
22mAMD-vIOMMU: A Hardware-assisted Virtual IOMMU Technology
37mHardware Error Handling Improvement for Reliable KVM Hypervisor
44mKernel TEE subsystem BoF
