Re: [PATCH RFC 2/2] mm: add PMD-level huge page support for remap_pfn_range()

[David Hildenbrand <david@redhat.com>]

On 23.09.25 15:31, Yin Tirui wrote:
> Add PMD-level huge page support to remap_pfn_range(), automatically
> creating huge mappings when prerequisites are satisfied (size, alignment,
> architecture support, etc.) and falling back to normal page mappings
> otherwise.
> 
> Implement special huge PMD splitting by utilizing the pgtable deposit/
> withdraw mechanism. When splitting is needed, the deposited pgtable is
> withdrawn and populated with individual PTEs created from the original
> huge mapping, using pte_clrhuge() to clear huge page attributes.
> 
> Update arch_needs_pgtable_deposit() to return true when PMD pfnmap
> support is enabled, ensuring proper pgtable management for huge
> pfnmap operations.
> 
> Introduce pfnmap_max_page_shift parameter to control maximum page
> size and "nohugepfnmap" boot option to disable huge pfnmap entirely.

Why? If an arch supports it we should just do it. Or what's the reason 
behind that?

> 
> Signed-off-by: Yin Tirui <yintirui@huawei.com>
> ---
>   include/linux/pgtable.h |  6 +++-
>   mm/huge_memory.c        | 22 ++++++++----
>   mm/memory.c             | 74 ++++++++++++++++++++++++++++++++++++-----
>   3 files changed, 85 insertions(+), 17 deletions(-)
> 
> diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
> index 4c035637eeb7..4028318552ca 100644
> --- a/include/linux/pgtable.h
> +++ b/include/linux/pgtable.h
> @@ -1025,7 +1025,11 @@ extern pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm, pmd_t *pmdp);
>   #endif
>   
>   #ifndef arch_needs_pgtable_deposit
> -#define arch_needs_pgtable_deposit() (false)
> +#define arch_needs_pgtable_deposit arch_needs_pgtable_deposit
> +static inline bool arch_needs_pgtable_deposit(void)
> +{
> +	return IS_ENABLED(CONFIG_ARCH_SUPPORTS_PMD_PFNMAP);
> +}
>   #endif
>   
>   #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 9c38a95e9f09..9f20adcbbb55 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -2857,14 +2857,22 @@ static void __split_huge_pmd_locked(struct vm_area_struct *vma, pmd_t *pmd,
>   
>   	if (!vma_is_anonymous(vma)) {
>   		old_pmd = pmdp_huge_clear_flush(vma, haddr, pmd);
> -		/*
> -		 * We are going to unmap this huge page. So
> -		 * just go ahead and zap it
> -		 */
> -		if (arch_needs_pgtable_deposit())
> -			zap_deposited_table(mm, pmd);

Are you sure we can just entirely remove this block for 
!vma_is_anonymous(vma)?

-- 
Cheers

David / dhildenb