Re: [RFC PATCH 0/9] iommu/iommufd: Add hypervisor external attach

[Jason Gunthorpe]

On Wed, Nov 04, 2026 at 10:23:22AM +0000, Mostafa Saleh wrote:

> > IMHO, I don't see any change needed for the host-side attach between hw
> > nested and collapsed single stage, if guest and host attach are
> > split. I think it is up to pKVM or MSHV to handle guest virtual iommu
> > mapping for both cases.
> 
> Although the hypervisor (pKVM, MSHV) is managing the guest IOMMU in that
> case, I believe the host should be able to decide the guest virtual
> toplology, as that will impact the firmware tables populated from the
> host for example, as in the nested case the guest will see an
> architectural SMMUv3 while in the collapsed case it will see a simple
> HVC based IOMMU.

Yes, this is how RMM/etc are working in CC land.

Untrusted hypervisor defines the vPCI topology, trusted hypervisor
implements it as instructed.

Jason