[RFC 0/6] iommu: generic api migration and grouping

[Ohad Ben-Cohen]

On Mon, Jun 6, 2011 at 10:20 PM, Roedel, Joerg <Joerg.Roedel at amd.com> wrote:
> Well, it certainly makes sense to have a single implementation for this.
> But I want to hide this complexity to the user of the IOMMU-API. The
> best choice is to put this into the layer between the IOMMU-API and the
> backend implementation.

I agree.

The IOMMU API should take physically contiguous regions from the user,
split them up according to page-sizes (/alignment requirements)
supported by the hardware, and then tell the underlying implementation
what to map where.

> That interface is not put into stone. There were other complains about
> the ->unmap part recently, so there is certainly room for improvement
> there.

Once the supported page sizes are exposed to the framework, the
current ->unmap API should probably be enough. 'va' + 'order' sounds
like all the information an implementation needs to unmap a page.