RE: [PATCH v6 2/6] iommufd: Add IOMMU_HWPT_INVALIDATE

[Liu, Yi L]

> From: Jason Gunthorpe <jgg@xxxxxxxxxx>
> Sent: Tuesday, December 12, 2023 10:40 PM
> 
> On Tue, Dec 12, 2023 at 01:45:16PM +0000, Liu, Yi L wrote:
> > > From: Jason Gunthorpe <jgg@xxxxxxxxxx>
> > > Sent: Monday, December 11, 2023 9:22 PM
> > >
> > > On Mon, Dec 11, 2023 at 03:53:40PM +0800, Yi Liu wrote:
> > >
> > > > > >  From that thread Jason mentioned to make the invalidation format
> > > > > > part of domain allocation. If that is the direction to go then there
> > > > > > won't be multiple invalidation formats per hwpt. The user should
> > > > > > create multiple hwpt's per invalidation format (though mixing
> > > > > > formats in one virtual platform is very unlikely)?
> > > > >
> > > > > I think we could do either, but I have a vauge cleanness preference
> > > > > that the enums are just different? That would follow a pretty typical
> > > > > pattern for a structure tag to reflect the content of the structure.
> > > >
> > > > Is this still following the direction to make the invalidation format
> > > > part of domain allocation?
> > >
> > > I think you should make it seperate
> >
> > Sure. I'll add a separate enum for cache invalidation format. Just to
> > see if it is good to pass down the invalidation format in the hwpt
> > alloc path? So iommu driver can check if the invalidation format
> > can be used for the hwpt to be allocated.
> 
> I would skip it in the invalidation. If necessary drivers can push a 0
> length invalidation to do 'try and fail' discovery of supported types.

I think you mean keep passing the req_type in cache invalidation path
instead of the way I proposed. For the 'try and fail' discovery, we should
allow a zero-length array, is it? If the req_type is supported by the iommu
driver, then return successful, otherwise fail the ioctl. Is it?

Regards,
Yi Liu