[PATCH] mtd: aspeed: fix compile warning in aspeed_smc_read_from_ahb()

[Cédric Le Goater]

On 02/09/2017 04:04 PM, Marek Vasut wrote:
> On 02/09/2017 10:59 AM, Cédric Le Goater wrote:
>> On 02/09/2017 10:05 AM, Marek Vasut wrote:
>>> On 02/09/2017 12:37 AM, Brian Norris wrote:
>>>> On Thu, Jan 19, 2017 at 10:13:18AM +0100, Cédric Le Goater wrote:
>>>>> The first argument of ioread32_rep() and ioread8_rep is not
>>>>> const. Change aspeed_smc_read_from_ahb() prototype to fix compile
>>>>> warning :
>>>>>
>>>>>    drivers/mtd/spi-nor/aspeed-smc.c: In function 'aspeed_smc_read_from_ahb':
>>>>>    drivers/mtd/spi-nor/aspeed-smc.c:212:16: warning: passing argument 1 of 'ioread32_rep' discards 'const' qualifier from pointer target type [-Wdiscarded-qualifiers]
>>>>>       ioread32_rep(src, buf, len >> 2);
>>>>
>>>> But why isn't that param constant? Seems like that's a bug in the
>>>> ioread* APIs. Some of their implementations are const, but some are not.
>>>>
>>>> I see for instance that the implementation in lib/iomap.c is not const,
>>>> where there's zero reason it shouldn't be (if we also fix, e.g., its
>>>> mmio_insl()).
>>>
>>> Maybe it's worth fixing the ioread*_rep then ? Then again, that seems to
>>> be way far off the scope of this patch.
>>
>> Looking closer at where this comes from, the compile warning was 
>> reported on x86 which pulls a different file for the definitions 
>> of the ioread routines.
> 
> Well yes, intel is a special snowflake in the io accessor area.
> 
>> On x86, arch/x86/include/asm/io.h includes <asm-generic/iomap.h>
>> which contains prototypes of the ioread routines without 'const'.
>>
>> On ARM, arch/arm/include/asm/io.h includes <asm-generic/io.h>
>> which contains the 'const' definitions.
> 
> Can you check if any of those arches needs it without const ?

ok. I have checked over ppc, all defconfigs compile fine. 
I will send a patch to linuxppc to see how we can merged it 
in powerpc-next. The other arches would then be checked 
by the Intel 0-day bot.

C. 


> 
>> So may be, we should just drop that patch and fix iomap.h instead,
>> but quite a few arches use it :
>>
>> arch/sh/include/asm/io.h:#include <asm-generic/iomap.h>
>> arch/mips/include/asm/io.h:#include <asm-generic/iomap.h>
>> arch/ia64/include/asm/io.h:#include <asm-generic/iomap.h>
>> arch/cris/include/asm/io.h:#include <asm-generic/iomap.h>
>> arch/x86/include/asm/io.h:#include <asm-generic/iomap.h>
>> arch/alpha/include/asm/io.h:#include <asm-generic/iomap.h>
>> arch/powerpc/include/asm/io.h:#include <asm-generic/iomap.h>
>> arch/m68k/include/asm/io_mm.h:#include <asm-generic/iomap.h>
>> arch/m68k/include/asm/io_no.h:#include <asm-generic/iomap.h>
>> arch/parisc/include/asm/io.h:#include <asm-generic/iomap.h>
>> include/asm-generic/io.h:#include <asm-generic/iomap.h>
>>
>> Thanks,
>>
>> C.
>>
> 
>