Re: [PATCHv4 net-next] ptp/ioctl: support MONOTONIC_RAW timestamps for PTP_SYS_OFFSET_EXTENDED
From: Thomas Gleixner
Date: Sat May 11 2024 - 06:00:15 EST
On Fri, May 10 2024 at 09:45, Mahesh Bandewar (महेश बंडेवार) wrote:
> On Fri, May 10, 2024 at 12:50 AM Thomas Gleixner <tglx@linutronixde> wrote:
>>
>> On Thu, May 09 2024 at 21:07, Richard Cochran wrote:
>>
>> > Thomas,
>> >
>> > On Wed, May 08, 2024 at 09:38:58AM +0200, Thomas Gleixner wrote:
>> >> On Tue, May 07 2024 at 21:44, Richard Cochran wrote:
>> >> > On Thu, May 02, 2024 at 02:10:47PM -0700, Mahesh Bandewar wrote:
>> >> >> + * History:
>> >> >> + * v1: Initial implementation.
>> >> >> + *
>> >> >> + * v2: Use the first word of the reserved-field for @clockid. That's
>> >> >> + * backward compatible since v1 expects all three reserved words
>> >> >> + * (@rsv[3]) to be 0 while the clockid (first word in v2) for
>> >> >> + * CLOCK_REALTIME is '0'.
>> >
>> > ..
>> >
>> >> I agree that it wants to be in the commit message, but having the
>> >> version information in the kernel-doc which describes the UAPI is
>> >> sensible and useful. That's where I'd look first and asking a user to
>> >> dig up this information on lore is not really helpful.
>> >
>> > But writing "v1, v2" doesn't make sense for this code. There never
>> > was a "v1" for this ioctl. At the very least, the change should be
>> > identified by kernel version (or git SHA).
>>
>> Adding the git SHA before committing the change is going to be
>> challenging :)
>
> Instead of v1/v2, how about I can make it 'prior to kernel 6.10' and
> 'from 6.10 onwards' etc. (as Richard proposed)?
Sure