Re: [PATCH v7 1/3] iommufd: Add data structure for Intel VT-d stage-1 cache invalidation

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Jan 03, 2024 at 09:06:23AM -0800, Nicolin Chen wrote:
> On Wed, Jan 03, 2024 at 12:58:48PM -0400, Jason Gunthorpe wrote:
> > On Wed, Jan 03, 2024 at 08:48:46AM -0800, Nicolin Chen wrote:
> > > > You can pass the ctx to the invalidate op, it is already implied
> > > > because the passed iommu_domain is linked to a single iommufd ctx.
> > > 
> > > The device virtual id lookup API needs something similar, yet it
> > > likely needs a viommu pointer (or its id) instead? As the table
> > > is attached to a viommu while an ictx can have multiple viommus,
> > > right?
> > 
> > Yes, when we get to an API for that it will have to be some op
> > 'invalidate_viommu(..)' and it can get the necessary pointers.
> 
> OK! I will try that first.
> 
> > The viommu object will have to be some driver object like the
> > iommu_domain.
> 
> I drafted something like this, linking it to struct iommu_device:
> 
> +struct iommufd_viommu {
> +       struct iommufd_object obj;
> +       struct iommufd_ctx *ictx;
> +       struct iommu_device *iommu_dev;
> +       struct iommufd_hwpt_paging *hwpt;
> +       /* array of struct iommufd_device, indexed by device virtual id */
> +       struct xarray device_ids;
> +};

The driver would have to create it and there would be some driver
specific enclosing struct to go with it

Perhaps device_ids goes in the driver specific struct, I don't know.

Not sure it should have hwpt at all, probably vmid should come from
the driver specific struct in some driver specific way

Jason