Re: [PATCH v4 02/12] iommu: Introduce iommu_paging_domain_alloc_flags()

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Oct 23, 2024 at 03:26:48PM +0100, Robin Murphy wrote:
> On 2024-10-23 11:41 am, Vasant Hegde wrote:
> > From: Jason Gunthorpe <jgg@ziepe.ca>
> > 
> > Currently drivers calls iommu_paging_domain_alloc(dev) to get an
> > UNMANAGED domain. This is not sufficient to support PASID with
> > UNMANAGED domain as some HW like AMD requires certain page table type
> > to support PASIDs.
> > 
> > Also the domain_alloc_paging op only passes device as param for domain
> > allocation. This is not sufficient for AMD driver to decide the right
> > page table.
> > 
> > Instead of extending ops->domain_alloc_paging() it was decided to
> > enhance ops->domain_alloc_user() so that caller can pass various
> > additional flags.
> 
> Wait, what? So iommu_paging_domain_alloc_flags() opaquely may or may not
> actually allocate a regular paging domain or a user domain 

There is no such thing as a "user domain".

> depending on the flags? But then per patch #3 we allocate default
> domains - the most kernel-owned of all - as *user* domains,

We are going to rename domain_alloc_user() once this lands and the
dust settled, it's purpose has now expanded. I'm going to try to do it
for rc7 when all the branches are together.

> return an unfinished non-user domain, in spite of the explicitly documented
> behaviour:

> | Unlike @domain_alloc, it is called only by IOMMUFD and
> | must fully initialize the new domain before return.
> 
> ???

That remark about "called only by iommufd" should be removed in this
series.

> This is insane. Please implement it in manner which is not insane.

There are more steps, calm down.

Jason