Re: Re: [PATCH V4 net-bugfixs] net/ethernet: Update ret when ptp_clock is ERROR

On Sat, Nov 14, 2020 at 4:14 PM Richard Cochran
[off-list ref] wrote:

On Fri, Nov 13, 2020 at 05:21:43PM +0100, Arnd Bergmann wrote:

quoted

I've prototyped a patch that I think makes this more sensible
again: https://pastebin.com/AQ5nWS9e

I like the behavior described in the text.

Instead of this ...

     - if a built-in driver calls PTP interface functions but fails
       to select HAVE_PTP_1588_CLOCK or depend on PTP_1588_CLOCK,
       and PTP support is a loadable module, we get a link error
       instead of having an unusable clock.

how about simply deleting the #else clause of

    --- a/include/linux/ptp_clock_kernel.h
    +++ b/include/linux/ptp_clock_kernel.h
    @@ -173,7 +173,7 @@ struct ptp_clock_event {
      };
     };

    -#if IS_REACHABLE(CONFIG_PTP_1588_CLOCK)
    +#if IS_ENABLED(CONFIG_PTP_1588_CLOCK)

so that invalid configurations throw a compile time error instead?

I was trying to still allow PTP clocks to be disabled, either when
building a kernel that doesn't need it, or when posix timers are
disabled. Leaving out the #else path would break all drivers that
have PTP support in the main ethernet driver file rather than
conditionally compiling it based on a Kconfig symbol that depends
on CONFIG_PTP_1588_CLOCK.

       Arnd