On 12.04.2021 16:03, Jan Beulich wrote:
> Since it was brought up in the discussion of v1: I think the end
> goal is to be for mm/p2m.c to become a HVM-only file. Any "wrappers"
> also trying to take care of !paging_mode_translate() guests ought to
> be moved elsewhere. To see what actually still needs taking care of,
> incrementally extending the #ifdef CONFIG_HVM regions there is the
> way to go imo.
>
> Compared to v1 there are many new patches here plus build fixes to
> two of the three remaining ones from v1.
>
> 01: p2m: set_{foreign,mmio}_p2m_entry() are HVM-only
> 02: p2m: {,un}map_mmio_regions() are HVM-only
> 03: mm: the gva_to_gfn() hook is HVM-only
> 04: AMD/IOMMU: guest IOMMU support is for HVM only
> 05: p2m: change_entry_type_* hooks are HVM-only
> 06: p2m: hardware-log-dirty related hooks are HVM-only
> 07: p2m: the recalc hook is HVM-only
> 08: mem-access is HVM-only
> 09: make mem-paging configuarable and default it to off for being unsupported
> 10: p2m: {get,set}_entry hooks and p2m-pt.c are HVM-only
> 11: p2m: write_p2m_entry_{pre,post} hooks are HVM-only
> 12: p2m: re-arrange struct p2m_domain
Besides patch 8, which has gone in, I've got two acks from Tim for
shadow code changes and contradicting feedback on patch 9, but
nothing else.
Thanks for acks or otherwise,
Jan