Re: [PATCH v3 0/6] hw: Add Cnuas virtual devices

["Michael S. Tsirkin" <mst@redhat.com>]

On Sat, Oct 31, 2026 at 12:05:14PM +0000, Weqaar Janjua wrote:
> On Fri, Oct 30, 2026 at 07:19:05PM -0400, Michael S. Tsirkin wrote:
> > I see. Then a simple way for you to get some eyes on this, and a path
> > towards upstreaming, could be to try and add this rdma functionality
> > in the virtio spec.
> > Interface changes might be required during the standardization
> > process.
> 
> Thanks Michael, that is a helpful pointer.
> 
> The intended use of the Cnuas emulated PCIe devices is the reason they
> were not designed as virtio devices.
> 
> Cnuas is intended for teaching and research of Open Compute ORV3
> (and future versions as it evolves) rack-scale data center systems, where the
> PCIe devices and their drivers are what is being studied. CnuasNIC and
> CnuasGPU are small, fully documented PCIe devices in the spirit of
> edu, but with production-like features such as configurable link speed and
> width, a 64-bit prefetchable BAR (CnuasGPU), and a switch transport
> that joins several guests into one fabric. Their guest drivers are
> full netdev and RDMA core drivers that run rping, perftest and QEMU
> RDMA migration unmodified. Just to clarify these aren't edu extensions 
> because they are different PCI classes (network and co-processor), and 
> edu's fixed register map is used by existing course material. My concern with
> virtio device is that it would hide the layer we want people to work with.


People being the students?

> Also, that in the ORV3 rack model, each sled carries its own NIC and accelerator
> as separate PCIe functions with their own driver, next to the virtio
> devices the guest uses for its disk and management network.
> 
> I understand that a virtio-rdma device could definitely be useful in its own right, 
> I am happy to look at it separately since the CnuasSwitch backend could serve 
> both. If you are ok with it, for this series I would like to keep the devices as 
> they are, in the same spirit as the other QEMU-only devices in the 1b36 range, and 
> keep looking for a reviewer for patches 2 and 3 (Jason, Peter, Zhijian, Akihiko, 
> any of you?).
> 
> Thanks,
> 
> Weqaar


-- 
MST