Re: [PATCH v2 1/1] eventfd: implementation of EFD_MASK flag

[Martin Sustrik <sustrik@250bpm.com>]

On 15/02/13 18:32, Andy Lutomirski wrote:
> On Thu, Feb 14, 2013 at 9:24 PM, Andrew Morton
> <akpm@linux-foundation.org>  wrote:
>> On Fri, 15 Feb 2013 04:42:27 +0100 Martin Sustrik<sustrik@250bpm.com>  wrote:
>>
>>>> This is a non-back-compatible userspace interface change.  A procfs
>>>> file which previously displayed
>>>>
>>>>      eventfd-count: nnnn
>>>>
>>>> can now also display
>>>>
>>>>      eventfd-mask: nnnn
>>>>
>>>> So existing userspace could misbehave.
>>>>
>>>> Please fully describe the proposed interface change in the changelog.
>>>> That description should include the full pathname of the procfs file
>>>> and example before-and-after output and a discussion of whether and why
>>>> the risk to existing userspace is acceptable.
>>>
>>> I am not sure what the policy is here. Is not printing out the state of
>>> the object acceptable way to maintain backward compatibility? If not so,
>>> does new type of object require new procfs file, which, AFAIU, is the
>>> only way to retain full backward compatibility?
>>
>> Adding a new file is the only way I can think of to preserve the API.
>> But from Andy's comment is sounds like we don't have to worry a lot
>> about back-compatibility.
>>
>
> I'm not even convinced there's an issue in the first place (other than
> the fact that use of this feature will break old criu, regardless of
> /proc changes).  The fdinfo files already vary by descriptor type.
> Anything that screws up if unexpected fields are present is already
> screwed.

Ok then. I'll leave the relevant code as is.

Martin