Re: [PATCH v2 1/6] rtc: m68k: provide rtc_class_ops directly

[Arnd Bergmann <arnd@arndb.de>]

On Wednesday 27 April 2016 09:47:56 Geert Uytterhoeven wrote:
> > --- a/arch/m68k/kernel/time.c
> > +++ b/arch/m68k/kernel/time.c
> > @@ -86,7 +86,24 @@ void read_persistent_clock(struct timespec *ts)
> >         }
> >  }
> >
> > -#ifdef CONFIG_ARCH_USES_GETTIMEOFFSET
> > +#if defined(CONFIG_ARCH_USES_GETTIMEOFFSET) && defined(CONFIG_RTC_DRV_GENERIC)
> 
> s/defined/IS_ENABLED/ for the modular case.

Thanks, fixed in all three architectures/

> > @@ -95,7 +112,10 @@ static int __init rtc_init(void)
> >         if (!mach_hwclk)
> >                 return -ENODEV;
> >
> > -       pdev = platform_device_register_simple("rtc-generic", -1, NULL, 0);
> > +       /* or just call devm_rtc_device_register instead? */
> 
> I guess this comment is a bogus leftover? There's no "dev" parameter to
> pass to devm_rtc_device_register() here.

Sort of. When I wrote it, I thought that a NULL argument would work,
and I later found out myself that it doesn't.

I'll drop the comment there, but there are still a few ways we (probably
not me, but whoever is interested) could take this further:

- register both the device and the driver here, and call
  devm_rtc_device_register from the probe function so we can move away
  from drivers/rtc/rtc-generic.c

- do this separately for mac, mvme147, mvme16x, sun3, q40 and sun3x

- move the six implementations into drivers/rtc as standalone drivers.

One (AFAIK) unsolved problem here is the question of how to handle
read_boot_clock/read_persistent_clock/update_persistent_clock in this
case. This is a really odd API that is implemented in various ways
on a few major architectures, and not at all on others, so it's all
highly inconsistent.

	Arnd