[PATCH] mtd: nand: xway: fix build undefined MODULE_DEVICE_TABLE()

[John Crispin]

On 01/12/2016 14:47, Boris Brezillon wrote:
> On Thu, 1 Dec 2016 14:02:55 +0100
> John Crispin <john at phrozen.org> wrote:
> 
>> On 01/12/2016 13:36, Boris Brezillon wrote:
>>> Hi Hauke,
>>>
>>> On Wed, 30 Nov 2016 23:51:10 +0100
>>> Hauke Mehrtens <hauke at hauke-m.de> wrote:
>>>   
>>>> The header file with the definition of MODULE_DEVICE_TABLE() was
>>>> missing, add include for linux/module.h to fix the problem in 4.9.  
>>>
>>> I tried to enable this driver as a module, and the build failed because
>>> of a missing symbol (see the following patch).
>>> Now, if it's not supposed to be compiled as a module, then you should
>>> modify the Kconfig accordingly.
>>>
>>> Regards,
>>>
>>> Boris
>>>   
>>> --->8---  
>>> diff --git a/arch/mips/lantiq/xway/sysctrl.c
>>> b/arch/mips/lantiq/xway/sysctrl.c index 236193b5210b..29e753556597
>>> 100644 --- a/arch/mips/lantiq/xway/sysctrl.c
>>> +++ b/arch/mips/lantiq/xway/sysctrl.c
>>> @@ -156,6 +156,7 @@ static void __iomem *pmu_membase;
>>>  static void __iomem *ltq_xbar_membase;
>>>  void __iomem *ltq_cgu_membase;
>>>  void __iomem *ltq_ebu_membase;
>>> +EXPORT_SYMBOL(ltq_ebu_membase);  
>>
>> Hi,
>>
>> i would be against exporting the membase pointers as global symbols. i
>> already find it annoying that cgu and ebu are not static
> 
> Yep, that's also my opinion.
> 
> Let's patch the Kconfig entry instead. BTW, if you don't compile the
> driver as a module, then you don't need the MODULE_XX() calls, and you
> can get rid of the module.h header.
>

agreed. not sure how hauke tested his patch in the first place.

	John