Re: [PATCH] x86/ept: fix shattering of special pages

[Jan Beulich <jbeulich@xxxxxxxx>]

On 27.06.2022 12:01, Roger Pau Monne wrote:
> The current logic in epte_get_entry_emt() will split any page marked
> as special with order greater than zero, without checking whether the
> super page is all special.
> 
> Fix this by only splitting the page only if it's not all marked as
> special, in order to prevent unneeded super page shuttering.
> 
> Fixes: ca24b2ffdb ('x86/hvm: set 'ipat' in EPT for special pages')
> Signed-off-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>
> ---
> Cc: Paul Durrant <paul@xxxxxxx>
> ---
> It would seem weird to me to have a super page entry in EPT with
> ranges marked as special and not the full page.  I guess it's better
> to be safe, but I don't see an scenario where we could end up in that
> situation.
> 
> I've been trying to find a clarification in the original patch
> submission about how it's possible to have such super page EPT entry,
> but haven't been able to find any justification.

I think the loop there was added "just in case", whereas in reality
it was only single special pages that would ever be mapped. So ...

> Might be nice to expand the commit message as to why it's possible to
> have such mixed super page entries that would need splitting.

... there may not be anything to add. I don't mind this being re-done,
hence
Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>
Yet I'm not sure I understand what case this actually fixes (and
hence why you did add a Fixes: tag) - are there cases where non-
order-0 special pages are mapped in reality?

As to the Fixes: tag - I think we mean to follow Linux there and
hence want 12-digit hashes to be specified. See also
docs/process/sending-patches.pandoc. (But no need to resend just
for this.)

Jan