Re: [PATCH v2 12/12] iommu/amd: Introduce IOMMUFD vIOMMU support for AMD

[Jason Gunthorpe <jgg@nvidia.com>]

On Thu, Oct 02, 2025 at 01:05:32PM -0700, Nicolin Chen wrote:
> > +/*
> > + * See include/linux/iommufd.h
> > + * struct iommufd_viommu_ops - vIOMMU specific operations
> > + */
> > +const struct iommufd_viommu_ops amd_viommu_ops = {
> > +	.alloc_domain_nested = amd_iommu_alloc_domain_nested,
> > +};
> 
> Unfortunately, a viommu_ops with alloc_domain_nested is incomplete,
> IMHO. If this series gets merged alone, it declares that the kernel
> now supports AMD IOMMU's virtualization, which actually won't work
> without a cache invalidation op (SW) or hw_queue (HW-acceleration).

Yeah, I'd move this patch to a series implementing viommu fully.

I'm pretty sure this series is OK as is without it, but it would be
good to at least share a sketch of the viommu support to be sure
before merging it.

Jason