Re: [RFC PATCH 12/21] KVM: IOMMUFD: MEMFD: Map private pages

[Jason Gunthorpe <jgg@nvidia.com>]

On Thu, Sep 05, 2024 at 01:53:27PM -0700, Dan Williams wrote:

> As mentioned in another thread this entry into the LOCKED state is
> likely nearly as violent as hotplug event since the DMA layer currently
> has no concept of a device having a foot in the secure world and the
> shared world at the same time.

There is also something of a complicated situation where the VM also
must validate that it has received the complete and correct device
before it can lock it. Ie that the MMIO ranges belong to the device,
the DMA goes to the right place (the vRID:pRID is trusted), and so on.

Further, the vIOMMU, and it's parameters, in the VM must also be
validated and trusted before the VM can lock the device. The VM and
the trusted world must verify they have the exclusive control over the
translation.

This is where AMDs model of having the hypervisor control things get a
little bit confusing for me. I suppose there will be some way that the
confidential VM asks the trusted world to control the secure DTE such
that it can select between GCR3, BLOCKED and IDENTITY.

Regardless, I think everyone will need some metadata from the vIOMMU
world into the trusted world to do all of this.

Jason