A Hyper-V Based pvIOMMU Built on Generic I/O Page Table
summary
A Hyper-V pvIOMMU prototype for Linux guests uses the community Generic I/O Page Table framework and drives DMA remapping from the guest’s stage-1 nested translation rather than trap-and-emulate map/unmap calls, removing per-map hypervisor overhead. Proofs-of-concept show both virtio-iommu and Hyper-V pvIOMMU running atop this design, patches are posted, and expected throughput/latency wins replace emulation costs; open work includes modeling IOTLB invalidation and adding a domain-selective IOTLB flush hypercall and command-queue integration.
AI-written summary · about