Re: [PATCH RFC 3/4] mm: guest_memfd: Add option to remove guest private memory from direct map

[Manwaring, Derek]

On 2024-08-07 17:16-0700 Derek Manwaring wrote:
> All that said, we're also dependent on hardware not being subject to
> L1TF-style issues for the currently proposed non-CoCo method to be
> effective. We're simply clearing the Present bit while the physmap PTE
> still points to the guest physical page.

I was wrong here. The set_direct_map_invalid_noflush implementation
moves through __change_page_attr and pfn_pte, eventually arriving at
flip_protnone_guard where the PFN is inverted & thus no longer valid for
pages marked not present. So we do benefit from that prior work's extra
protection against L1TF.

Thank you for finding this, Patrick.

Derek