Re: [PATCH v3 03/13] VMX: implement suppress #VE.

["Jan Beulich" <JBeulich@suse.com>]

>>> On 01.07.15 at 20:09, <edmund.h.white@intel.com> wrote:
> @@ -232,6 +235,15 @@ static int ept_set_middle_entry(struct p2m_domain *p2m, ept_entry_t *ept_entry)
>      /* Manually set A bit to avoid overhead of MMU having to write it later. */
>      ept_entry->a = 1;
>  
> +    ept_entry->suppress_ve = 1;
> +
> +    table = __map_domain_page(pg);
> +
> +    for ( i = 0; i < EPT_PAGETABLE_ENTRIES; i++ )
> +        table[i].suppress_ve = 1;
> +
> +    unmap_domain_page(table);

For the moment I can certainly agree to it being done this way, but
it's inefficient and should be cleaned up: There shouldn't be two
mappings of the page being allocated (one in hap_alloc() and the
other being added here). I suppose the easiest would be to pass
an optional callback pointer to p2m_alloc_ptp(). Or, to also cover
the case below in ept_p2m_init() (i.e. p2m_alloc_table()) a new
optional hook in struct p2m_domain could be added for that purpose.
Albeit ...

> @@ -1134,6 +1151,13 @@ int ept_p2m_init(struct p2m_domain *p2m)
>          p2m->flush_hardware_cached_dirty = ept_flush_pml_buffers;
>      }
>  
> +    table = map_domain_page(pagetable_get_pfn(p2m_get_pagetable(p2m)));
> +
> +    for ( i = 0; i < EPT_PAGETABLE_ENTRIES; i++ )
> +        table[i].suppress_ve = 1;
> +
> +    unmap_domain_page(table);

... why is this needed? Bit 63 is documented to be ignored in PML4Es
(just like in all other intermediate page tables).

Jan