Re: [PATCH 02/11] riscv: mm: Increment PFN in place when splitting mappings

[Alexandre Ghiti <alex@ghiti.fr>]

Hi Samuel,

On 02/11/2024 01:07, Samuel Holland wrote:
> The current code separates page table entry values into a PFN and a
> pgprot_t before incrementing the PFN and combining the two parts using
> pfn_pXX(). On some hardware with custom page table formats or memory
> aliases, the pfn_pXX() functions need to transform the PTE value, so
> these functions would need to apply the opposite transformation when
> breaking apart the PTE value.
>
> However, both transformations can be avoided by incrementing the PFN in
> place, as done by pte_advance_pfn() and set_ptes().
>
> Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
> ---
>
>   arch/riscv/mm/pageattr.c | 17 ++++++++---------
>   1 file changed, 8 insertions(+), 9 deletions(-)
>
> diff --git a/arch/riscv/mm/pageattr.c b/arch/riscv/mm/pageattr.c
> index 271d01a5ba4d..335060adc1a6 100644
> --- a/arch/riscv/mm/pageattr.c
> +++ b/arch/riscv/mm/pageattr.c
> @@ -109,9 +109,8 @@ static int __split_linear_mapping_pmd(pud_t *pudp,
>   			continue;
>   
>   		if (pmd_leaf(pmdp_get(pmdp))) {
> +			pte_t pte = pmd_pte(pmdp_get(pmdp));
>   			struct page *pte_page;
> -			unsigned long pfn = _pmd_pfn(pmdp_get(pmdp));
> -			pgprot_t prot = __pgprot(pmd_val(pmdp_get(pmdp)) & ~_PAGE_PFN_MASK);
>   			pte_t *ptep_new;
>   			int i;
>   
> @@ -121,7 +120,7 @@ static int __split_linear_mapping_pmd(pud_t *pudp,
>   
>   			ptep_new = (pte_t *)page_address(pte_page);
>   			for (i = 0; i < PTRS_PER_PTE; ++i, ++ptep_new)
> -				set_pte(ptep_new, pfn_pte(pfn + i, prot));
> +				set_pte(ptep_new, pte_advance_pfn(pte, i));
>   
>   			smp_wmb();
>   
> @@ -149,9 +148,8 @@ static int __split_linear_mapping_pud(p4d_t *p4dp,
>   			continue;
>   
>   		if (pud_leaf(pudp_get(pudp))) {
> +			pmd_t pmd = __pmd(pud_val(pudp_get(pudp)));


Nit: You could use pud_pte() here.


>   			struct page *pmd_page;
> -			unsigned long pfn = _pud_pfn(pudp_get(pudp));
> -			pgprot_t prot = __pgprot(pud_val(pudp_get(pudp)) & ~_PAGE_PFN_MASK);
>   			pmd_t *pmdp_new;
>   			int i;
>   
> @@ -162,7 +160,8 @@ static int __split_linear_mapping_pud(p4d_t *p4dp,
>   			pmdp_new = (pmd_t *)page_address(pmd_page);
>   			for (i = 0; i < PTRS_PER_PMD; ++i, ++pmdp_new)
>   				set_pmd(pmdp_new,
> -					pfn_pmd(pfn + ((i * PMD_SIZE) >> PAGE_SHIFT), prot));
> +					__pmd(pmd_val(pmd) +
> +					      (i << (PMD_SHIFT - PAGE_SHIFT + PFN_PTE_SHIFT))));


Nit: Here you could use pte_advance_pfn(pmd, i << (PMD_SHIFT - PAGE_SHIFT))


>   
>   			smp_wmb();
>   
> @@ -198,9 +197,8 @@ static int __split_linear_mapping_p4d(pgd_t *pgdp,
>   			continue;
>   
>   		if (p4d_leaf(p4dp_get(p4dp))) {
> +			pud_t pud = __pud(p4d_val(p4dp_get(p4dp)));
>   			struct page *pud_page;
> -			unsigned long pfn = _p4d_pfn(p4dp_get(p4dp));
> -			pgprot_t prot = __pgprot(p4d_val(p4dp_get(p4dp)) & ~_PAGE_PFN_MASK);
>   			pud_t *pudp_new;
>   			int i;
>   
> @@ -215,7 +213,8 @@ static int __split_linear_mapping_p4d(pgd_t *pgdp,
>   			pudp_new = (pud_t *)page_address(pud_page);
>   			for (i = 0; i < PTRS_PER_PUD; ++i, ++pudp_new)
>   				set_pud(pudp_new,
> -					pfn_pud(pfn + ((i * PUD_SIZE) >> PAGE_SHIFT), prot));
> +					__pud(pud_val(pud) +
> +					      (i << (PUD_SHIFT - PAGE_SHIFT + PFN_PTE_SHIFT))));


Nit: Ditto


>   
>   			/*
>   			 * Make sure the pud filling is not reordered with the


Other than the nits (which are up to you), you can add:

Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>

Thanks,

Alex