Re: [PATCH v10 1/3] x86/tlb: introduce a flush HVM ASIDs flag

[Jan Beulich <jbeulich@suse.com>]

On 21.04.2020 15:51, Roger Pau Monné wrote:
> On Tue, Apr 21, 2020 at 02:59:10PM +0200, Jan Beulich wrote:
>> On 21.04.2020 12:43, Roger Pau Monné wrote:
>>> On Tue, Apr 21, 2020 at 12:21:12PM +0200, Jan Beulich wrote:
>>>> On 16.04.2020 15:59, Roger Pau Monne wrote:
>>>>> Introduce a specific flag to request a HVM guest linear TLB flush,
>>>>> which is an ASID/VPID tickle that forces a guest linear to guest
>>>>> physical TLB flush for all HVM guests.
>>>>>
>>>>> This was previously unconditionally done in each pre_flush call, but
>>>>> that's not required: HVM guests not using shadow don't require linear
>>>>> TLB flushes as Xen doesn't modify the guest page tables in that case
>>>>> (ie: when using HAP).
>>>>
>>>> I'm afraid I'm being confused by this: Even in shadow mode Xen
>>>> doesn't modify guest page tables, does it?
>>>
>>> I'm also confused now. It's my understand that when running in shadow
>>> mode guest page tables are not actually used, and the guest uses Xen's
>>> crafted shadow tables instead, which are based on the original guest
>>> page tables suitably adjusted by Xen in order to do the p2m
>>> translation in the HVM case, or the needed PTE adjustments in the PV
>>> case.
>>>
>>> So guest page tables are not modified, but are also not used as they
>>> are never loaded into cr3.
>>
>> This matches my understanding.
> 
> Please bear with me, as I'm not sure if your question was because you
> think the paragraph is not clear and/or should be expanded.
> 
> The point of the paragraph you mention was to have a difference
> between guests running in shadow mode vs guests running in HAP mode.
> Maybe I should use guest loaded page pages, to differentiate between
> guest created page tables and the page tables actually loaded in cr3
> in guest mode?

How about using "the pages tables the guest runs on"?

Jan