Uploaded November 2026 | Updated November 2026, 1 day ago
In this episode, Bruno Verachten talks with Baptiste Le Duc and Oleksii Kurochko from Vates about porting Xen to RISC-V.
RISC-V's H-extension makes the ISA a good fit for virtualization, but what does it take to run a real type-1 hypervisor on it? This talk follows the work of porting Xen to RISC-V, from the ISA features it depends on to guests booting on real hardware.
We start with Xen's architecture and its two deployment models, dom0less (static partitioning, guests started by the hypervisor at boot) and the classic dom0 setup. Then we look at how RISC-V supports virtualization: VS-mode and two-stage address translation (and how Xen builds its p2m on top of them), how guest interrupts are delivered through AIA and the older PLIC, and how Xen provides the SBI interface guests expect from firmware. We also cover which devices Xen emulates and which it passes through directly, using the UART as an example, plus the hypercall interface.
Next, we list the extensions a server-class platform needs to host a hypervisor well (H, AIA, Zbb, Svpbmt, Sstc, IOMMU) and explain what changes when moving from QEMU to real silicon. A live demo shows Xen running guests on RISC-V. We finish with the upstreaming roadmap and how to get involved.
SHORT OUTLINE OF THE CONVERSATION:
1. Intros: Vates, XCP-ng, and why RISC-V matters for an open, sovereign stack from silicon to cloud
2. Target platform: which extensions a hypervisor needs (H, AIA/PLIC, Zbb, Svpbmt, Sstc, IOMMU)
3. What supporting Xen on RISC-V means:
- virtualization vs passthrough (e.g., the UART) p2m
- Interrupt virtualization: support of AIA and PLIC
- Hypercalls
- SBI virtualization
- From QEMU to real hardware
4. Roadmap and how to contribute
5. Q&A
In this episode, Bruno Verachten talks with Baptiste Le Duc and Oleksii Kurochko from Vates about porting Xen to RISC-V.
RISC-V's H-extension makes the ISA a good fit for virtualization, but what does it take to run a real type-1 hypervisor on it? This talk follows the work of porting Xen to RISC-V, from the ISA features it depends on to guests booting on real hardware.
We start with Xen's architecture and its two deployment models, dom0less (static partitioning, guests started by the hypervisor at boot) and the classic dom0 setup. Then we look at how RISC-V supports virtualization: VS-mode and two-stage address translation (and how Xen builds its p2m on top of them), how guest interrupts are delivered through AIA and the older PLIC, and how Xen provides the SBI interface guests expect from firmware. We also cover which devices Xen emulates and which it passes through directly, using the UART as an example, plus the hypercall interface.
Next, we list the extensions a server-class platform needs to host a hypervisor well (H, AIA, Zbb, Svpbmt, Sstc, IOMMU) and explain what changes when moving from QEMU to real silicon. A live demo shows Xen running guests on RISC-V. We finish with the upstreaming roadmap and how to get involved.
SHORT OUTLINE OF THE CONVERSATION:
1. Intros: Vates, XCP-ng, and why RISC-V matters for an open, sovereign stack from silicon to cloud
2. Target platform: which extensions a hypervisor needs (H, AIA/PLIC, Zbb, Svpbmt, Sstc, IOMMU)
3. What supporting Xen on RISC-V means:
- virtualization vs passthrough (e.g., the UART) p2m
- Interrupt virtualization: support of AIA and PLIC
- Hypercalls
- SBI virtualization
- From QEMU to real hardware
4. Roadmap and how to contribute
5. Q&A










