Re: [PATCH v1] ptp: add ToD device driver for Intel FPGA cards

[Andy Shevchenko <andriy.shevchenko@linux.intel.com>]

On Tue, Mar 21, 2023 at 02:28:15PM +0000, Zhang, Tianfei wrote:
> > From: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> > Sent: Tuesday, March 21, 2023 9:03 PM
> > To: Nicolas Pitre <nico@fluxnic.net>
> > On Mon, Mar 20, 2023 at 04:53:07PM -0400, Nicolas Pitre wrote:
> > > On Mon, 20 Mar 2023, Richard Cochran wrote:
> > > > On Mon, Mar 20, 2023 at 09:43:30AM -0400, Nicolas Pitre wrote:
> > > >
> > > > > Alternatively the above commit can be reverted if no one else
> > > > > cares. I personally gave up on the idea of a slimmed down Linux
> > > > > kernel a while ago.
> > > >
> > > > Does this mean I can restore the posix clocks back into the core
> > > > unconditionally?
> > >
> > > This only means _I_ no longer care. I'm not speaking for others (e.g.
> > > OpenWRT or the like) who might still rely on splitting it out.
> > > Maybe Andy wants to "fix" it?
> > 
> > I would be a good choice, if I have an access to the hardware at hand to test.
> > That said, I think Richard himself can try to finish that feature (optional PTP) and on
> > my side I can help with reviewing the code (just Cc me when needed).
> 
> Handle NULL as a valid parameter (object) to their respective APIs looks a
> good idea, but this will be a big change and need fully test for all devices,

Since it's core change, so a few devices that cover 100% APIs that should
handle optional PTP. I don't think it requires enormous work.

> we can add it on the TODO list.

It would be a good idea.

> So for this patch, are you agree we continue use the existing
> ptp_clock_register() API, for example, change my driver like below:

The problem is that it will increase the technical debt.
What about sending with NULL handled variant together with an RFC/RFT
that finishes the optional PTP support?

>       dt->ptp_clock = ptp_clock_register(&dt->ptp_clock_ops, dev);

	ret = PTR_ERR_OR_ZERO(...);
	if (ret)
		return ...

?

>       if (IS_ERR_OR_NULL(dt->ptp_clock))
>               return dev_err_probe(dt->dev, IS_ERR_OR_NULL(dt->ptp_clock),
>                                    "Unable to register PTP clock\n");


-- 
With Best Regards,
Andy Shevchenko