Re: [Xen-devel] [PATCH v1 for-4.6 2/2] xen: prevent access to HPET from Dom0

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 19.12.14 at 12:32, <andrew.cooper3@xxxxxxxxxx> wrote:
> On 19/12/14 09:11, Jan Beulich wrote:
>>>>> On 18.12.14 at 19:51, <andrew.cooper3@xxxxxxxxxx> wrote:
>>> On 18/12/14 18:27, Roger Pau Monne wrote:
>>>> Prevent Dom0 from accessing HPET MMIO region by adding it to the list of
>>>> denied memory regions.
>>>>
>>>> Signed-off-by: Roger Pau Monnà <roger.pau@xxxxxxxxxx>
>>>> Cc: Jan Beulich <jbeulich@xxxxxxxx>
>>>> Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
>>> Apologies that this reply is split between patch 0 and 2 - I replied to
>>> your cover letter before reading this patch.
>>>
>>> Denying access is only valid if acpi_table_hpet.flags & 
>>> ACPI_HPET_PAGE_PROTECT4 is true.
>> Having just checked (as an example) the most modern Intel box I
>> have direct access to, I wonder how many systems actually supply
>> other than 0 here. Perhaps we ought to at once add a command
>> line option to trigger the denial?
> 
> I also can't find a server which sets this flag.  I wonder how many
> systems actually have other things sitting in the remainder of the page.

One would think (or should I say hope) that there's at least nothing
with read side effects anywhere, or else Linux'es exposing of the
page to user mode would be a security problem. Perhaps we should
also limit Dom0 mappings to r/o when we can't hide the page
altogether.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel