Re: [PATCH] mm/debug_vm_pgtable: Replace pte_mkhuge() with arch_make_huge_pte()

[Anshuman Khandual]

On 3/2/23 12:31, Christophe Leroy wrote:
> 
> 
> Le 02/03/2023 à 04:18, Anshuman Khandual a écrit :
>> Since the following commit arch_make_huge_pte() should be used directly in
>> generic memory subsystem as a platform provided page table helper, instead
>> of pte_mkhuge(). Change hugetlb_basic_tests() to call arch_make_huge_pte()
>> directly, and update its relevant documentation entry as required.
>>
>> 'commit 16785bd77431 ("mm: merge pte_mkhuge() call into arch_make_huge_pte()")'
>>
>> Cc: Jonathan Corbet <corbet@xxxxxxx>
>> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
>> Cc: "Mike Rapoport (IBM)" <rppt@xxxxxxxxxx>
>> Cc: linux-doc@xxxxxxxxxxxxxxx
>> Cc: linux-kernel@xxxxxxxxxxxxxxx
>> Cc: linux-mm@xxxxxxxxx
>> Reported-by: Christophe Leroy <christophe.leroy@xxxxxxxxxx>
>> Link: https://lore.kernel.org/all/1ea45095-0926-a56a-a273-816709e9075e@xxxxxxxxxx/
>> Signed-off-by: Anshuman Khandual <anshuman.khandual@xxxxxxx>
>> ---
>> This applies on latest mainline.
>>
>>   Documentation/mm/arch_pgtable_helpers.rst | 2 +-
>>   mm/debug_vm_pgtable.c                     | 2 +-
>>   2 files changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/Documentation/mm/arch_pgtable_helpers.rst b/Documentation/mm/arch_pgtable_helpers.rst
>> index 30d9a09f01f4..af3891f895b0 100644
>> --- a/Documentation/mm/arch_pgtable_helpers.rst
>> +++ b/Documentation/mm/arch_pgtable_helpers.rst
>> @@ -214,7 +214,7 @@ HugeTLB Page Table Helpers
>>   +---------------------------+--------------------------------------------------+
>>   | pte_huge                  | Tests a HugeTLB                                  |
>>   +---------------------------+--------------------------------------------------+
>> -| pte_mkhuge                | Creates a HugeTLB                                |
>> +| arch_make_huge_pte        | Creates a HugeTLB                                |
>>   +---------------------------+--------------------------------------------------+
>>   | huge_pte_dirty            | Tests a dirty HugeTLB                            |
>>   +---------------------------+--------------------------------------------------+
>> diff --git a/mm/debug_vm_pgtable.c b/mm/debug_vm_pgtable.c
>> index af59cc7bd307..92bed5bd5879 100644
>> --- a/mm/debug_vm_pgtable.c
>> +++ b/mm/debug_vm_pgtable.c
>> @@ -934,7 +934,7 @@ static void __init hugetlb_basic_tests(struct pgtable_debug_args *args)
>>   #ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
>>   	pte = pfn_pte(args->fixed_pmd_pfn, args->page_prot);
>>   
>> -	WARN_ON(!pte_huge(pte_mkhuge(pte)));
>> +	WARN_ON(!pte_huge(arch_make_huge_pte(pte)));
> 
> Did you build it ?

Ahh bad ! Built on arm64 which does not subscribe ARCH_WANT_GENERAL_HUGETLB.

> 
> arch_make_huge_pte() need 3 arguments.

Could call it with constants such as PMD_SHIFT and standard vm_flags without
involving real arg based values (as they are not initialized for HugeTLB).
But wondering if platform arch_make_huge_pte() implementation might complain.

arch_make_huge_pte(pte, PMD_SHIFT, VM_READ | VM_WRITE)