Re: [PATCH] arm64/migration: Define arm64_hugetlb_valid_size()

[Anshuman Khandual]

On 2/10/22 12:04 AM, Catalin Marinas wrote:
> On Mon, Feb 07, 2022 at 07:01:08AM +0530, Anshuman Khandual wrote:
>> diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c
>> index ffb9c229610a..dcdc4c0c3bd8 100644
>> --- a/arch/arm64/mm/hugetlbpage.c
>> +++ b/arch/arm64/mm/hugetlbpage.c
>> @@ -56,24 +56,12 @@ void __init arm64_hugetlb_cma_reserve(void)
>>  }
>>  #endif /* CONFIG_CMA */
>>  
>> +static bool arm64_hugetlb_valid_size(unsigned long size);
>> +
>>  #ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
>>  bool arch_hugetlb_migration_supported(struct hstate *h)
>>  {
>> -	size_t pagesize = huge_page_size(h);
>> -
>> -	switch (pagesize) {
>> -#ifndef __PAGETABLE_PMD_FOLDED
>> -	case PUD_SIZE:
>> -		return pud_sect_supported();
>> -#endif
>> -	case PMD_SIZE:
>> -	case CONT_PMD_SIZE:
>> -	case CONT_PTE_SIZE:
>> -		return true;
>> -	}
>> -	pr_warn("%s: unrecognized huge page size 0x%lx\n",
>> -			__func__, pagesize);
>> -	return false;
>> +	return arm64_hugetlb_valid_size(huge_page_size(h));
>>  }
>>  #endif
>>  
>> @@ -504,7 +492,7 @@ static int __init hugetlbpage_init(void)
>>  }
>>  arch_initcall(hugetlbpage_init);
>>  
>> -bool __init arch_hugetlb_valid_size(unsigned long size)
>> +static bool arm64_hugetlb_valid_size(unsigned long size)
>>  {
>>  	switch (size) {
>>  #ifndef __PAGETABLE_PMD_FOLDED
>> @@ -517,5 +505,12 @@ bool __init arch_hugetlb_valid_size(unsigned long size)
>>  		return true;
>>  	}
>>  
>> +	pr_warn("%s: unrecognized huge page size 0x%lx\n",
>> +		__func__, size);
>>  	return false;
>>  }
> 
> We already have the warnings in the caller of arch_hugetlb_valid_size(),
> I wouldn't add another here. You can keep it in
> arch_hugetlb_migration_supported() though.

Sure, make sense. Will change it to something like this.

bool arch_hugetlb_migration_supported(struct hstate *h)
{
        size_t pagesize = huge_page_size(h);

        if (!arm64_hugetlb_valid_size(pagesize)) {
                pr_warn("%s: unrecognized huge page size 0x%lx\n",
                        __func__, pagesize);
                return false;
        }
        return true;
}

> 
>> +
>> +bool __init arch_hugetlb_valid_size(unsigned long size)
>> +{
>> +	return arm64_hugetlb_valid_size(size);
>> +}
> 
> What's wrong with keeping the arch_hugetlb_valid_size() name and just
> removing __init?
> 

When arch_hugetlb_migration_supported() calls arch_hugetlb_valid_size(),

With __init for arch_hugetlb_valid_size(), there is a build problem.

WARNING: modpost: vmlinux.o(.text.unlikely+0xf18): Section mismatch in
reference from the function arch_hugetlb_migration_supported() to the
function .init.text:arch_hugetlb_valid_size()
The function arch_hugetlb_migration_supported() references
the function __init arch_hugetlb_valid_size().
This is often because arch_hugetlb_migration_supported lacks a __init 
annotation or the annotation of arch_hugetlb_valid_size is wrong.

Without __init for arch_hugetlb_valid_size(), there is the same problem.

WARNING: modpost: vmlinux.o(.text.unlikely+0xf18): Section mismatch in
reference from the function arch_hugetlb_migration_supported() to the
function .init.text:arch_hugetlb_valid_size()
The function arch_hugetlb_migration_supported() references
the function __init arch_hugetlb_valid_size().
This is often because arch_hugetlb_migration_supported lacks a __init 
annotation or the annotation of arch_hugetlb_valid_size is wrong.

This might be because generic prototype for arch_hugetlb_valid_size() has
__init attribute, which arch_hugetlb_migration_supported() does not have.

Hence a local static helper arm64_hugetlb_valid_size() which both these
generic functions can call into, looked appropriate.