Re: [PATCH 01/11] xen: reserve flags for internal usage in xen_domctl_createdomain

[Jan Beulich <jbeulich@xxxxxxxx>]

On 11.10.2021 12:45, Julien Grall wrote:
> Hi Jan,
> 
> On 28/09/2021 13:05, Jan Beulich wrote:
>> On 23.09.2021 11:54, Julien Grall wrote:
>>> On 23/09/2021 08:11, Penny Zheng wrote:
>>>> From: Stefano Stabellini <stefano.stabellini@xxxxxxxxxx>
>>>>
>>>> We are passing an extra special boolean flag at domain creation to
>>>> specify whether we want to the domain to be privileged (i.e. dom0) or
>>>> not. Another flag will be introduced later in this series.
>>>>
>>>> Reserve bits 16-31 from the existing flags bitfield in struct
>>>> xen_domctl_createdomain for internal Xen usage.
>>>
>>> I am a bit split with this approach. This feels a bit of a hack to
>>> reserve bits for internal purpose in external headers. But at the same
>>> time I can see how this is easier to deal with it over repurposing the
>>> last argument of domain_create().
>>
>> I actually have trouble seeing why that's easier. It is a common thing
>> to widen a bool to "unsigned int flags" when more than one control is
>> needed.
> 
> I was suggesting this is easier for the following two reasons:
>    1) All the option flags (internal and external) are in a single place.
>    2) Reduce the risk to make a mistake when widening the field. In 
> particular in the context of backporting. Although, this looks unlikely 
> here.
> 
>> Plus this makes things needlessly harder once (in the future)
>> the low 16 bits are exhausted in the public interface.
> 
> That's why I suggested this sounds like a hack. At the same time the 
> split between external vs internal option is a bit more a pain for the 
> developper. So I didn't feel pushing for one vs the other. That said, I 
> will not argue against if you want to push for repurposing the last 
> argument.

In which case - Penny, would you please change the patch accordingly?

Jan