Re: [PATCH V7 2/2] xen/gnttab: Store frame GFN in struct page_info on Arm

[Jan Beulich <jbeulich@xxxxxxxx>]

On 11.10.2022 15:33, Julien Grall wrote:
> Hi Jan,
> 
> On 11/10/2022 14:28, Jan Beulich wrote:
>> On 11.10.2022 15:01, Julien Grall wrote:
>>> Hi Jan,
>>>
>>> On 11/10/2022 12:59, Jan Beulich wrote:
>>>> On 16.07.2022 16:56, Oleksandr Tyshchenko wrote:
>>>>> From: Oleksandr Tyshchenko <oleksandr_tyshchenko@xxxxxxxx>
>>>>>
>>>>> Rework Arm implementation to store grant table frame GFN
>>>>> in struct page_info directly instead of keeping it in
>>>>> standalone status/shared arrays. This patch is based on
>>>>> the assumption that a grant table page is a xenheap page.
>>>>>
>>>>> To cover 64-bit/40-bit IPA on Arm64/Arm32 we need the space
>>>>> to hold 52-bit/28-bit + extra bit value respectively. In order
>>>>> to not grow the size of struct page_info borrow the required
>>>>> amount of bits from type_info's count portion which current
>>>>> context won't suffer (currently only 1 bit is used on Arm).
>>>>
>>>> I'm afraid this isn't true: There's no requirement for a guest to pass
>>>> all different GFNs to VCPUOP_register_vcpu_info, yet map_vcpu_info()
>>>> tries to obtain a reference for every vCPU.
>>>
>>> AFAIU, this would be a reference of the **count_info** not **type_info**
>>> (which BTW will never be incremented on Arm because we have no type
>>> support).
>>
>> I should have said "obtain a writable type reference".
> 
> Thanks for the clarification.
> 
>>
>>> The commit message is only referring to the 'type_info's count'. So...
>>>
>>>> With my adding of GFN
>>>> (really gaddr) based registration of the runstate area (already
>>>> looking towards 4.18) the maximum possible count is to further grow.
>>>
>>> ... I am not sure which problem you are referring too.
>>
>> Wow - a mere stub (but not inline) function to make the build happy.
>> Then why is the description talking about one bit that's needed on
>> Arm?
> 
> Because share_xen_page_with_guest() will always set the type info's 
> count to 1.
> 
> TBH I don't exactly know why we set it. I always assumed this was a 
> requirement for the common code but never checked.

I don't think there is any such requirement. In fact there are only
very few uses of type_info in common code. By also setting
PGT_count_mask to zero you could actually have the compiler eliminate
some dead code ...

Jan