Re: [PATCH v4 28/41] vfio/iommufd: Implement the iommufd backend

[Jason Gunthorpe]

On Wed, Nov 08, 2023 at 01:25:34PM +0000, Duan, Zhenzhong wrote:

> >I was expecting that hwpt manipulation would be done exclusively
> >inside the device-specific vIOMMU userspace driver. Generic code paths
> >that don't have that knowledge should use the IOAS for everything
> 
> Yes, this way we don't need to distinguish between mdev and real device,
> just attach to IOAS. But lose the benefit that same hwpt could be passed
> into vIOMMU to be used as S2 hwpt in nesting.

If you have a nesting capable vIOMMU driver then it should be
creating the HWPTs and managing them in its layer. Maybe the core code
provides some helpers.

Obviously you can't link a mdev to a nesting vIOMMU driver in the
first place. Mdev should be connected to a different IOMMU driver that
doesn't use HWPT at all.

I think it will make alot of trouble to put the hwpt in the wrong
layer as there shouldn't really be much generic code touching it.

Jason