Re: [PATCH V2 1/3] arm64/mm: Drop __HAVE_ARCH_HUGE_PTEP_GET

[Will Deacon]

On Wed, May 06, 2020 at 11:42:12AM +0530, Anshuman Khandual wrote:
> Platform specific huge_ptep_get() is required only when fetching the huge
> PTE involves more than just dereferencing the page table pointer. This is
> not the case on arm64 platform. Hence huge_ptep_pte() can be dropped along
> with it's __HAVE_ARCH_HUGE_PTEP_GET subscription.
> 
> Cc: Catalin Marinas <catalin.marinas@xxxxxxx>
> Cc: Will Deacon <will@xxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> Cc: linux-mm@xxxxxxxxx
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> Signed-off-by: Anshuman Khandual <anshuman.khandual@xxxxxxx>
> ---
>  arch/arm64/include/asm/hugetlb.h | 6 ------
>  1 file changed, 6 deletions(-)
> 
> diff --git a/arch/arm64/include/asm/hugetlb.h b/arch/arm64/include/asm/hugetlb.h
> index 2eb6c234d594..b88878ddc88b 100644
> --- a/arch/arm64/include/asm/hugetlb.h
> +++ b/arch/arm64/include/asm/hugetlb.h
> @@ -17,12 +17,6 @@
>  extern bool arch_hugetlb_migration_supported(struct hstate *h);
>  #endif
>  
> -#define __HAVE_ARCH_HUGE_PTEP_GET
> -static inline pte_t huge_ptep_get(pte_t *ptep)
> -{
> -	return READ_ONCE(*ptep);
> -}

Hmm, I'm nervous about dropping the READ_ONCE() here. We added them after
running into page-table issues with THP [1] and it's really important to
use them consistently to avoid hitting that again.

If the generic code used READ_ONCE(), I'd be happy to switch to it.

Will

[1] https://lore.kernel.org/lkml/1506527369-19535-1-git-send-email-will.deacon@xxxxxxx/