[RFC] Proposal: RISC-V Architecture Support in KubeVirt (Initial Contribution)

[JinCheng NI]

Hi everyone,

Our team has initiated work on adding RISC-V architecture support to KubeVirt as an independent effort. We're still early in the process, but we'd like to share what we've done so far and get feedback from the community on next steps.

🔗 Source Repository:
https://code.risc-verse.isrc.ac.cn/risc-verse/wg-cloudcomputing/kubevirt-riscv

Motivation

RISC-V is gaining attention in research and emerging cloud platforms, but KubeVirt does not yet support it.
We've been exploring basic RISC-V integration on real hardware and emulators, and hope to contribute a minimal, working foundation.

While we understand RISC-V is still early in cloud adoption, we believe starting this support now may benefit future experimental and academic use cases. We're happy to align with community priorities and keep the scope minimal and maintainable.

Current Progress

• Successfully adapted the Bazel build system for RISC-V architecture
• Completed cross-compilation of core components for riscv64 using the new toolchain
• Replaced and adapted container image build scripts and base distributions to support RISC-V
• Updated relevant KubeVirt core source code to handle RISC-V-specific characteristics

Questions for the Community

1. Would the community be open to incorporating RISC-V as a supported architecture in KubeVirt?
2. KubeVirt currently uses container images based on distributions (e.g., CentOS Stream) that lack official RISC-V support. Would it be acceptable to use alternative distributions for RISC-V builds (e.g., Fedora or openEuler)?
3. Are there technical or roadmap concerns we should consider before submitting formal PRs?

If the community expresses interest, we would be happy to contribute this effort incrementally and maintain RISC-V support long-term. We're also willing to align with KubeVirt's architectural principles and contribution guidelines.

Looking forward to your feedback and guidance.

Best regards,
Jincheng Ni
Contact: nijin...@iscas.ac.cn

[Edmund Jeghan]

Hi Jin, 

I am new to this I went to the Web page it looks pretty awesome very easy to understand. I also saw that you were asking for people to use the apps or create agents. You can explain to me what to do on the apps. 

Regards, 

Edmundzen 

--
You received this message because you are subscribed to the Google Groups "kubevirt-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubevirt-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/kubevirt-dev/86cf1ac4-fdee-401c-ba49-b1d8c9cc6f8an%40googlegroups.com.

[Daniel Hiller]

Hey,

speaking from a CI perspective, we definitely lack testing infrastructure to support this - thus we can't guarantee over the course of RISC-V integration that other PRs might break the RISC-V support again.

Is there some way to receive bare metal hardware to be able to run a subset of the test cases so we make sure everything works as expected?

Best,

Daniel

--

You received this message because you are subscribed to the Google Groups "kubevirt-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubevirt-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/kubevirt-dev/86cf1ac4-fdee-401c-ba49-b1d8c9cc6f8an%40googlegroups.com.

--

-- 

Best,

Daniel

[倪金成]

Hi Daniel,


Thanks a lot for your feedback — that's completely understandable.

We recognize that without CI integration, it's hard to ensure long-term reliability for a new architecture like RISC-V. We’ll check internally to see if there's a possibility to provide access to RISC-V hardware (bare metal or remote).

We’ll follow up once we have something concrete to offer or propose.

Appreciate your time and support!


Best regards,  
Jincheng Ni

-----Original Messages-----
From:"'Daniel Hiller' via kubevirt-dev" <kubevi...@googlegroups.com>
Sent Time:2025-05-30 15:50:24 (Friday)
To: "JinCheng NI" <nijin...@iscas.ac.cn>
Cc: kubevirt-dev <kubevi...@googlegroups.com>
Subject: Re: [kubevirt-dev] [RFC] Proposal: RISC-V Architecture Support in KubeVirt (Initial Contribution)

You received this message because you are subscribed to a topic in the Google Groups "kubevirt-dev" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/kubevirt-dev/dObRKg8Stwo/unsubscribe.
To unsubscribe from this group and all its topics, send an email to kubevirt-dev...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/kubevirt-dev/CAK%2BeyL43wyDKYY6oDNt3aUawwbyB1BGMSSrR3piR%3D8ceG5hc%3DQ%40mail.gmail.com.

[倪金成]

Hi Daniel,

Thanks again for your response.

At the moment, we do not have access to RISC-V servers that support the H extension required for KVM-based virtualization, as such hardware is not yet commercially available in a form suitable for CI deployment.

As a temporary approach, we are exploring the use of QEMU-based emulation to run a limited set of tests. While this is not a complete substitute for hardware-backed testing, we believe it can still help catch regressions and provide a basic level of integration assurance during the early development phase.

In parallel, we’re closely tracking hardware developments in the RISC-V space. If things go as expected, we anticipate having access to real RISC-V hardware with KVM support sometime next year. At that point, we’d be glad to revisit this and explore options to provide the community with access to physical hardware for CI integration.

Best regards,

Jincheng Ni

-----Original Messages-----
From:倪金成 <nijin...@iscas.ac.cn>
Sent Time:2025-05-31 23:38:28 (Saturday)
To: "Daniel Hiller" <dhi...@redhat.com>
Cc: kubevirt-dev <kubevi...@googlegroups.com>
Subject: Re: Re: [kubevirt-dev] [RFC] Proposal: RISC-V Architecture Support in KubeVirt (Initial Contribution)

To view this discussion visit https://groups.google.com/d/msgid/kubevirt-dev/29c687b9.2298a.19726fe9692.Coremail.nijincheng%40iscas.ac.cn.

[Andrea Bolognani]

On Wed, Jun 11, 2025 at 03:20:38PM +0800, 倪金成 wrote:
> At the moment, we do not have access to RISC-V servers that support
> the H extension required for KVM-based virtualization, as such
> hardware is not yet commercially available in a form suitable for
> CI deployment.
>
> As a temporary approach, we are exploring the use of QEMU-based
> emulation to run a limited set of tests. While this is not a
> complete substitute for hardware-backed testing, we believe it can
> still help catch regressions and provide a basic level of
> integration assurance during the early development phase.
>
> In parallel, we’re closely tracking hardware developments in the
> RISC-V space. If things go as expected, we anticipate having access
> to real RISC-V hardware with KVM support sometime next year. At
> that point, we’d be glad to revisit this and explore options to
> provide the community with access to physical hardware for CI
> integration.

As an additional note, on the software side KubeVirt uses CentOS
Stream 9 as the base OS for its container images, and that OS doesn't
have RISC-V support.

However, that is changing for CentOS Stream 10: see the recent
announcements[1][2].

Hopefully at some point next year we'll see all factors converging:
hardware with H extension becoming commonplace, CentOS Stream 10
being built on RISC-V, and KubeVirt switching to CentOS Stream 10.

When that happens, adding RISC-V support to KubeVirt will become a
much more interesting proposition.


[1] https://blog.centos.org/2025/05/initial-centos-support-for-risc-v/
[2] https://blog.centos.org/2025/05/centos-for-risc-v-now-available/
--
Andrea Bolognani / Red Hat / Virtualization

[倪金成]

Hi Andrea,

Thanks a lot for sharing the latest information on CentOS.
We're also following the evolution of RISC-V support in CentOS Stream 10 — the current progress is encouraging.
We understand that upstream integration might make more sense once the base image story (e.g. CentOS Stream 10) becomes more stable. In the meantime, if the community considers it helpful to explore RISC-V support in a separate context or staging branch, we’d be happy to assist or contribute in whatever way is most appropriate.
Appreciate your insight — and we’ll continue to follow the community’s direction as things evolve.

Best regards,
Jincheng Ni

&gt; -----Original Messages-----
&gt; From: "'Andrea Bolognani' via kubevirt-dev" <kubevi...@googlegroups.com>
&gt; Sent Time: 2025-06-11 17:51:42 (Wednesday)
&gt; To: 倪金成 <nijin...@iscas.ac.cn>
&gt; Cc: "Daniel Hiller" <dhi...@redhat.com>, kubevirt-dev <kubevi...@googlegroups.com>
&gt; Subject: Re: [kubevirt-dev] [RFC] Proposal: RISC-V Architecture Support in KubeVirt (Initial Contribution)
&gt;
&gt; On Wed, Jun 11, 2025 at 03:20:38PM +0800, 倪金成 wrote:
&gt; &gt; At the moment, we do not have access to RISC-V servers that support
&gt; &gt; the H extension required for KVM-based virtualization, as such
&gt; &gt; hardware is not yet commercially available in a form suitable for
&gt; &gt; CI deployment.
&gt; &gt;
&gt; &gt; As a temporary approach, we are exploring the use of QEMU-based
&gt; &gt; emulation to run a limited set of tests. While this is not a
&gt; &gt; complete substitute for hardware-backed testing, we believe it can
&gt; &gt; still help catch regressions and provide a basic level of
&gt; &gt; integration assurance during the early development phase.
&gt; &gt;
&gt; &gt; In parallel, we’re closely tracking hardware developments in the
&gt; &gt; RISC-V space. If things go as expected, we anticipate having access
&gt; &gt; to real RISC-V hardware with KVM support sometime next year. At
&gt; &gt; that point, we’d be glad to revisit this and explore options to
&gt; &gt; provide the community with access to physical hardware for CI
&gt; &gt; integration.
&gt;
&gt; As an additional note, on the software side KubeVirt uses CentOS
&gt; Stream 9 as the base OS for its container images, and that OS doesn't
&gt; have RISC-V support.
&gt;
&gt; However, that is changing for CentOS Stream 10: see the recent
&gt; announcements[1][2].
&gt;
&gt; Hopefully at some point next year we'll see all factors converging:
&gt; hardware with H extension becoming commonplace, CentOS Stream 10
&gt; being built on RISC-V, and KubeVirt switching to CentOS Stream 10.
&gt;
&gt; When that happens, adding RISC-V support to KubeVirt will become a
&gt; much more interesting proposition.
&gt;
&gt;
&gt; [1] https://blog.centos.org/2025/05/initial-centos-support-for-risc-v/
&gt; [2] https://blog.centos.org/2025/05/centos-for-risc-v-now-available/
&gt; --
&gt; Andrea Bolognani / Red Hat / Virtualization
&gt;
&gt; --
&gt; You received this message because you are subscribed to the Google Groups "kubevirt-dev" group.
&gt; To unsubscribe from this group and stop receiving emails from it, send an email to kubevirt-dev...@googlegroups.com.
&gt; To view this discussion visit https://groups.google.com/d/msgid/kubevirt-dev/CABJz62N7TB8sEVx_t%3DWH-AU5g5Pxxm5K0FwrFT%2BBG5E7cGCx%2Bg%40mail.gmail.com.
</kubevi...@googlegroups.com></dhi...@redhat.com></nijin...@iscas.ac.cn></kubevi...@googlegroups.com>