Re: [PATCH v2 2/6] ARM: config: multi v7: Clean up enabled by default options

On Fri, Jun 11, 2021 at 5:29 AM Joel Stanley [off-list ref] wrote:

On Thu, 10 Jun 2021 at 07:58, Arnd Bergmann [off-list ref] wrote:

quoted

quoted

PTP_1588_CLOCK is implied by enabled drivers STMMAC_ETH, IGB, E1000E and FEC.

I still feel a little uneasy about not enabling PTP_1588_CLOCK in case
we get rid of that
'imply', the rest looks good.

If we got rid of the imply, would it become a select?

I realise that's bad form, but if these drivers need that functionality...

That would need some further discussion. I think we have two distinct
cases:

- some drivers can use the PTP_1588_CLOCK feature if that symbol
  is built-in, or both the driver and the ptp support are loadable modules.
  If the drivers are built-in, but PTP_1588_CLOCK=m, they currently
  silently break, which is unfortunate and often unexpected.
  The 'imply' keyword was originally intended to work around this
  problem but doesn't do that any more.

- Some drivers have a separate Kconfig symbol for enabling PTP
  support within a driver that is controlled separately. These tend to
  use 'depends on PTP_1588_CLOCK'.

I would hope that we can at least be consistent with 'select'
vs 'depends on' here, but either solution is possible.  One option
that would work is to have a global setting that decides whether
PTP support is available to drivers at all, and then another
symbol that gets selected by drivers to decide whether it ends
up built-in or in a loadable module.

Happy to re-roll with that change dropped, just let me know.

That would be my preference, yes.

        Arnd

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel