Re: [PATCH] net: cpts: Condition WARN_ON on PTP_1588_CLOCK

[Arnd Bergmann]

On Mon, Apr 20, 2020 at 7:00 PM Richard Cochran
<richardcochran@xxxxxxxxx> wrote:
>
> On Mon, Apr 20, 2020 at 04:38:32PM +0200, Arnd Bergmann wrote:
> >
> > I suspect we should move all of them back. This was an early user
> > of 'imply', but the meaning of that keyword has now changed
> > in the latest Kconfig.
>
> Can you please explain the justification for changing the meaning?
>
> It was a big PITA for me to support this in the first place, and now
> we are back to square one?

I don't understand it either. Apparently it didn't always do what users
expected, though the new definition seems less useful to me, as it
only changes the default.

> > Something else is wrong if you need IS_ERR_OR_NULL(). Any
> > kernel interface should either return an negative error code when
> > something goes wrong, or should return NULL for all errors, but
> > not mix the two.
>
> On the contrary, this is exactly what the whole "imply" thing
> demanded.
>
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 172) #if IS_REACHABLE(CONFIG_PTP_1588_CLOCK)
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 173)
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 174) /**
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 175)  * ptp_clock_register() - register a PTP hardware clock driver
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 176)  *
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 177)  * @info:   Structure describing the new clock.
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 178)  * @parent: Pointer to the parent device of the new clock.
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 179)  *
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 180)  * Returns a valid pointer on success or PTR_ERR on failure.  If PHC
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 181)  * support is missing at the configuration level, this function
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 182)  * returns NULL, and drivers are expected to gracefully handle that
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 183)  * case separately.
> d1cbfd771ce82 (Nicolas Pitre       2016-11-11 184)  */

The key here is "gracefully". The second patch from Clay just turns NULL into
 -EOPNOTSUPP and treats the compile-time condition into a runtime error.
I don't see a point in allowing the driver to compile if it just always returns
an error. The two callers then go on to print a message for any error
and just keep going.

      Arnd