On 14.10.2024 10:00, Jürgen Groß wrote:
> On 14.10.24 09:46, Jan Beulich wrote:
>> On 14.10.2024 09:36, Jürgen Groß wrote:
>>> On 14.10.24 09:14, Jan Beulich wrote:
>>>> On 14.10.2024 09:06, Juergen Gross wrote:
>>>>> The recent addition of the XEN_DOMCTL_dt_overlay function was missing
>>>>> the related update of XEN_DOMCTL_INTERFACE_VERSION, as it has been the
>>>>> first interface change of the 4.20 release cycle.
>>>>>
>>>>> Fixes: 4c733873b5c2 ("xen/arm: Add XEN_DOMCTL_dt_overlay and device
>>>>> attachment to domains")
>>>>
>>>> I'm confused: That change (a) pre-dates the branching of 4.20 and (b)
>>>> bumped the version ...
>>>>
>>>>> --- a/xen/include/public/domctl.h
>>>>> +++ b/xen/include/public/domctl.h
>>>>> @@ -21,7 +21,7 @@
>>>>> #include "hvm/save.h"
>>>>> #include "memory.h"
>>>>>
>>>>> -#define XEN_DOMCTL_INTERFACE_VERSION 0x00000017
>>>>> +#define XEN_DOMCTL_INTERFACE_VERSION 0x00000018
>>>>
>>>> ... from 0x16 to 0x17. And for no apparent reason, as plain additions don't
>>>> require a bump. Did you maybe mean to reference a different commit?
>>>
>>> Oh, indeed. I wanted to reference d6e9a2aab39e.
>>>
>>> And regarding to "plain additions don't require a bump": 4c733873b5c2 did
>>> a plain addition and bumped the version.
>>
>> Right, hence why I said "for no apparent reason".
>
> There seems to be a lack of documentation in this regard.
>
> Julien explicitly asked for the bump for that addition.
Julien - why was that? Bumps are needed only for backwards incompatible
changes. Plain additions therefore never require a bump. As as we get
better with properly checking e.g. padding fields, the frequency of
required bumps should also further reduce.
> I'm fine with dropping my patch if others agree that the bump isn't needed.
> In that case I'll send another one adding a comment for the mechanics of
> interface version bump in domctl.h and sysctl.h.
Oh, yes, please feel free to do so.
Jan
