Re: Build failure in -next due to 'rtc: cmos: move mc146818rtc code out of asm-generic/rtc.h'

[Guenter Roeck <linux@roeck-us.net>]

On 06/03/2016 05:11 AM, Arnd Bergmann wrote:
> On Thursday, June 2, 2016 5:58:48 PM CEST Guenter Roeck wrote:
>> Hi Arnd,
>>
>> sparc32:allmodconfig fails to build in next-20160602 as follows.
>>
>> In file included from drivers/block/floppy.c:185:0:
>> include/linux/mc146818rtc.h: In function 'mc146818_is_updating':
>> include/linux/mc146818rtc.h:138:9: error: 'rtc_port' undeclared (first use in this function)
>> include/linux/mc146818rtc.h:138:9: note: each undeclared identifier is reported only once for each function it appears in
>> include/linux/mc146818rtc.h: In function 'mc146818_get_time':
>> include/linux/mc146818rtc.h:172:17: error: 'rtc_port' undeclared (first use in this function)
>> include/linux/mc146818rtc.h: In function 'mc146818_set_time':
>> include/linux/mc146818rtc.h:278:8: error: 'rtc_port' undeclared (first use in this function)
>> scripts/Makefile.build:295: recipe for target 'drivers/block/floppy.o' failed
>>
>> Bisect points to commit fd09cc80165c ("rtc: cmos: move mc146818rtc code out of
>> asm-generic/rtc.h"). Bisct log is attached.
>>
>> Note that removing the offending include file works for sparc32 but
>> not for x86; it results in a missing definition of CMOS_READ().
>
> I have trouble reproducing this, but I think the problem here is that there
> are two definitions of CMOS_READ() in sparc, and we pick up the wrong one
> here: there is no 'rtc_port' variable outside of drivers/char/rtc.c, and we
> don't build that file any more on sparc32, so we probably want this patch:
>
> diff --git a/arch/sparc/include/asm/io_32.h b/arch/sparc/include/asm/io_32.h
> index 57f26c398dc9..4dd268a3a8b0 100644
> --- a/arch/sparc/include/asm/io_32.h
> +++ b/arch/sparc/include/asm/io_32.h
> @@ -140,16 +140,6 @@ void ioport_unmap(void __iomem *);
>   struct pci_dev;
>   void pci_iounmap(struct pci_dev *dev, void __iomem *);
>
> -
> -
> -/*
> - * At the moment, we do not use CMOS_READ anywhere outside of rtc.c,
> - * so rtc_port is static in it. This should not change unless a new
> - * hardware pops up.
> - */
> -#define RTC_PORT(x)   (rtc_port + (x))
> -#define RTC_ALWAYS_BCD  0
> -
>   static inline int sbus_can_dma_64bit(void)
>   {
>   	return 0; /* actually, sparc_cpu_model==sun4d */
>
>
>
> Can you check if this fixes the problem?
>

Yes, it does. I did test builds with both sparc:allmodconfig and sparc64:allmodconfig
after applying above diff; both passed.

Guenter