Re: [PATCH -next] riscv: mm: fix build errors caused by mk_pmd()

["Sunnanyong (Nanyong Sun, Intelligent Computing Solution Development Dep)" <sunnanyong@huawei.com>]

On 2021/5/30 9:21, Palmer Dabbelt wrote:

> On Sat, 29 May 2021 15:45:04 PDT (-0700), Palmer Dabbelt wrote:
>> On Mon, 24 May 2021 00:02:20 PDT (-0700), sunnanyong@huawei.com wrote:
>>> With "riscv: mm: add THP support on 64-bit", mk_pmd() function
>>> introduce build errors,
>>> 1.build with CONFIG_ARCH_RV32I=y:
>>> arch/riscv/include/asm/pgtable.h: In function 'mk_pmd':
>>> arch/riscv/include/asm/pgtable.h:513:9: error: implicit declaration 
>>> of function 'pfn_pmd';
>>>  did you mean 'pfn_pgd'? [-Werror=implicit-function-declaration]
>>>
>>> 2.build with CONFIG_SPARSEMEM=y && CONFIG_SPARSEMEM_VMEMMAP=n
>>> arch/riscv/include/asm/pgtable.h: In function 'mk_pmd':
>>> include/asm-generic/memory_model.h:64:14: error: implicit 
>>> declaration of function 'page_to_section';
>>>  did you mean 'present_section'? 
>>> [-Werror=implicit-function-declaration]
>>>
>>> Use macro definition instead of inline function for mk_pmd
>>> to fix the above problems.It is similar to the mk_pte macro.
>>>
>>> Reported-by: kernel test robot <lkp@intel.com>
>>> Signed-off-by: Nanyong Sun <sunnanyong@huawei.com>
>>> ---
>>>  arch/riscv/include/asm/pgtable.h | 5 +----
>>>  1 file changed, 1 insertion(+), 4 deletions(-)
>>>
>>> diff --git a/arch/riscv/include/asm/pgtable.h 
>>> b/arch/riscv/include/asm/pgtable.h
>>> index 4b708ae08910..f35d9c90d4cb 100644
>>> --- a/arch/riscv/include/asm/pgtable.h
>>> +++ b/arch/riscv/include/asm/pgtable.h
>>> @@ -508,10 +508,7 @@ static inline unsigned long pmd_pfn(pmd_t pmd)
>>>      return ((__pmd_to_phys(pmd) & PMD_MASK) >> PAGE_SHIFT);
>>>  }
>>>
>>> -static inline pmd_t mk_pmd(struct page *page, pgprot_t prot)
>>> -{
>>> -    return pfn_pmd(page_to_pfn(page), prot);
>>> -}
>>> +#define mk_pmd(page, prot)    pfn_pmd(page_to_pfn(page), prot)
>>>
>>>  static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
>>>  {
>>
>> Thanks, this is on for-next.
>
> Sorry, I wasn't looking closely enough.  I think the actual problem 
> here is that mk_pmd() should be in pgtable-64.h.
> .

Moving mk_pmd() to pgtable-64.h can only solve the first problem, but 
not the second.

The second one need mk_pmd() defined as a macro, like this previous 
solution:

https://patchwork.kernel.org/project/linux-riscv/patch/20191023032302.160388-1-wangkefeng.wang@huawei.com/

Or do i need to move the '#define mk_pmd(page, prot) ...' to pgtable-64.h ?