Re: [PATCH v4 2/2] arm64: mm: hugetlb: Enable HUGETLB_PAGE_FREE_VMEMMAP for arm64

[Anshuman Khandual <anshuman.khandual@arm.com>]

On 3/31/22 12:26, Muchun Song wrote:
> --- a/arch/arm64/mm/flush.c
> +++ b/arch/arm64/mm/flush.c
> @@ -68,6 +68,19 @@ EXPORT_SYMBOL_GPL(__sync_icache_dcache);
>   */
>  void flush_dcache_page(struct page *page)
>  {
> +	/*
> +	 * Only the head page's flags of HugeTLB can be cleared since the tail
> +	 * vmemmap pages associated with each HugeTLB page are mapped with
> +	 * read-only when CONFIG_HUGETLB_PAGE_FREE_VMEMMAP is enabled (more
> +	 * details can refer to vmemmap_remap_pte()).  Although

Did you see real permission fault when flush_dcache_page() accessed remapped
tail pages, with readonly vmemmap ? OR this change is from code inspection ?

> +	 * __sync_icache_dcache() only set PG_dcache_clean flag on the head
> +	 * page struct, some tail page structs still can be seen the flag is

Sentence here needs restructuring ....               ^^^^^^^^^^^^^^^^^^
 
> +	 * set since the head vmemmap page frame is reused (more details can
> +	 * refer to the comments above page_fixed_fake_head()).
> +	 */
> +	if (hugetlb_free_vmemmap_enabled() && PageHuge(page))
> +		page = compound_head(page);

This should also be applicable to any other platform with both configs enabled
i.e ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE and ARCH_WANT_HUGETLB_PAGE_FREE_VMEMMAP ?
If yes, then how to ensure that the platforms change flush_dcache_page() before
subscribing into ARCH_WANT_HUGETLB_PAGE_FREE_VMEMMAP ?

> +
>  	if (test_bit(PG_dcache_clean, &page->flags))
>  		clear_bit(PG_dcache_clean, &page->flags);
>  }