Re: [PATCH v3 20/34] riscv: Implement the new page table range API

[Palmer Dabbelt <palmer@dabbelt.com>]

On Tue, 28 Feb 2023 13:37:23 PST (-0800), willy@infradead.org wrote:
> Add set_ptes(), update_mmu_cache_range() and flush_dcache_folio().
> Change the PG_dcache_clean flag from being per-page to per-folio.
>
> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
> Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
> Cc: Paul Walmsley <paul.walmsley@sifive.com>
> Cc: Palmer Dabbelt <palmer@dabbelt.com>
> Cc: Albert Ou <aou@eecs.berkeley.edu>
> Cc: linux-riscv@lists.infradead.org
> ---
>  arch/riscv/include/asm/cacheflush.h | 19 +++++++++----------
>  arch/riscv/include/asm/pgtable.h    | 26 +++++++++++++++++++-------
>  arch/riscv/mm/cacheflush.c          | 11 ++---------
>  3 files changed, 30 insertions(+), 26 deletions(-)
>
> diff --git a/arch/riscv/include/asm/cacheflush.h b/arch/riscv/include/asm/cacheflush.h
> index 03e3b95ae6da..10e5e96f09b5 100644
> --- a/arch/riscv/include/asm/cacheflush.h
> +++ b/arch/riscv/include/asm/cacheflush.h
> @@ -15,20 +15,19 @@ static inline void local_flush_icache_all(void)
>
>  #define PG_dcache_clean PG_arch_1
>
> -static inline void flush_dcache_page(struct page *page)
> +static inline void flush_dcache_folio(struct folio *folio)
>  {
> -	/*
> -	 * HugeTLB pages are always fully mapped and only head page will be
> -	 * set PG_dcache_clean (see comments in flush_icache_pte()).
> -	 */
> -	if (PageHuge(page))
> -		page = compound_head(page);
> -
> -	if (test_bit(PG_dcache_clean, &page->flags))
> -		clear_bit(PG_dcache_clean, &page->flags);
> +	if (test_bit(PG_dcache_clean, &folio->flags))
> +		clear_bit(PG_dcache_clean, &folio->flags);
>  }
> +#define flush_dcache_folio flush_dcache_folio
>  #define ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE 1
>
> +static inline void flush_dcache_page(struct page *page)
> +{
> +	flush_dcache_folio(page_folio(page));
> +}
> +
>  /*
>   * RISC-V doesn't have an instruction to flush parts of the instruction cache,
>   * so instead we just flush the whole thing.
> diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
> index b516f3b59616..3a3a776fc047 100644
> --- a/arch/riscv/include/asm/pgtable.h
> +++ b/arch/riscv/include/asm/pgtable.h
> @@ -405,8 +405,8 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
>
>
>  /* Commit new configuration to MMU hardware */
> -static inline void update_mmu_cache(struct vm_area_struct *vma,
> -	unsigned long address, pte_t *ptep)
> +static inline void update_mmu_cache_range(struct vm_area_struct *vma,
> +		unsigned long address, pte_t *ptep, unsigned int nr)
>  {
>  	/*
>  	 * The kernel assumes that TLBs don't cache invalid entries, but
> @@ -415,8 +415,11 @@ static inline void update_mmu_cache(struct vm_area_struct *vma,
>  	 * Relying on flush_tlb_fix_spurious_fault would suffice, but
>  	 * the extra traps reduce performance.  So, eagerly SFENCE.VMA.
>  	 */
> -	local_flush_tlb_page(address);
> +	while (nr--)
> +		local_flush_tlb_page(address + nr * PAGE_SIZE);
>  }
> +#define update_mmu_cache(vma, addr, ptep) \
> +	update_mmu_cache_range(vma, addr, ptep, 1)
>
>  #define __HAVE_ARCH_UPDATE_MMU_TLB
>  #define update_mmu_tlb update_mmu_cache
> @@ -456,12 +459,21 @@ static inline void __set_pte_at(struct mm_struct *mm,
>  	set_pte(ptep, pteval);
>  }
>
> -static inline void set_pte_at(struct mm_struct *mm,
> -	unsigned long addr, pte_t *ptep, pte_t pteval)
> +static inline void set_ptes(struct mm_struct *mm, unsigned long addr,
> +		pte_t *ptep, pte_t pteval, unsigned int nr)
>  {
> -	page_table_check_ptes_set(mm, addr, ptep, pteval, 1);
> -	__set_pte_at(mm, addr, ptep, pteval);
> +	page_table_check_ptes_set(mm, addr, ptep, pteval, nr);
> +
> +	for (;;) {
> +		__set_pte_at(mm, addr, ptep, pteval);
> +		if (--nr == 0)
> +			break;
> +		ptep++;
> +		addr += PAGE_SIZE;
> +		pte_val(pteval) += 1 << _PAGE_PFN_SHIFT;
> +	}
>  }
> +#define set_pte_at(mm, addr, ptep, pte) set_ptes(mm, addr, ptep, pte, 1)
>
>  static inline void pte_clear(struct mm_struct *mm,
>  	unsigned long addr, pte_t *ptep)
> diff --git a/arch/riscv/mm/cacheflush.c b/arch/riscv/mm/cacheflush.c
> index fcd6145fbead..e36a851e5788 100644
> --- a/arch/riscv/mm/cacheflush.c
> +++ b/arch/riscv/mm/cacheflush.c
> @@ -81,16 +81,9 @@ void flush_icache_mm(struct mm_struct *mm, bool local)
>  #ifdef CONFIG_MMU
>  void flush_icache_pte(pte_t pte)
>  {
> -	struct page *page = pte_page(pte);
> +	struct folio *folio = page_folio(pte_page(pte));
>
> -	/*
> -	 * HugeTLB pages are always fully mapped, so only setting head page's
> -	 * PG_dcache_clean flag is enough.
> -	 */
> -	if (PageHuge(page))
> -		page = compound_head(page);
> -
> -	if (!test_bit(PG_dcache_clean, &page->flags)) {
> +	if (!test_bit(PG_dcache_clean, &folio->flags)) {
>  		flush_icache_all();
>  		set_bit(PG_dcache_clean, &page->flags);
>  	}

Acked-by: Palmer Dabbelt <palmer@rivosinc.com>