Re: [PATCH 02/10] mm: provide free_reserved_pages(), removing x86 variant

["David Hildenbrand (Arm)" <david@kernel.org>]

On 7/8/26 16:10, David Hildenbrand (Arm) wrote:
> Let's extend free_reserved_page() in page_alloc.c to free_reserved_pages(),
> dropping the custom x86 variant. The common-code variant will consume an
> order, so adjust the x86 callers accordingly.
> 
> Make free_reserved_pages() assume that we are freeing ordinary
> high-order pages, just with the special "reserved" flavor. The target
> use case for now is freeing vmemmap PMD pages.
> 
> Set the refcount directly to 0 (instead of 1) and call
> __free_frozen_pages(). Add some kerneldoc. Use a single
> adjust_managed_page_count() call.
> 
> Signed-off-by: David Hildenbrand (Arm) <david@kernel.org>
> ---
>  arch/x86/mm/init_64.c |  8 +-------
>  include/linux/mm.h    |  8 ++++++--
>  mm/page_alloc.c       | 33 ++++++++++++++++++++++++++-------
>  3 files changed, 33 insertions(+), 16 deletions(-)
> 
> diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
> index 69e36f02a663a..320e742b0bb3f 100644
> --- a/arch/x86/mm/init_64.c
> +++ b/arch/x86/mm/init_64.c
> @@ -1000,12 +1000,6 @@ int arch_add_memory(int nid, u64 start, u64 size,
>  	return add_pages(nid, start_pfn, nr_pages, params);
>  }
>  
> -static void free_reserved_pages(struct page *page, unsigned long nr_pages)
> -{
> -	while (nr_pages--)
> -		free_reserved_page(page++);
> -}
> -
>  static void __meminit free_pagetable(struct page *page)
>  {
>  	/* bootmem page has reserved flag */
> @@ -1038,7 +1032,7 @@ static void __meminit free_vmemmap_pages(struct page *page, unsigned int order,
>  			while (nr_pages--)
>  				put_page_bootmem(page++);
>  		} else {
> -			free_reserved_pages(page, nr_pages);
> +			free_reserved_pages(page, order);
>  		}
>  	} else {
>  		__free_pages(page, order);
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 113e1752548c2..62b261cccf16f 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -3993,8 +3993,12 @@ extern unsigned long free_reserved_area(void *start, void *end,
>  
>  extern void adjust_managed_page_count(struct page *page, long count);
>  
> -/* Free the reserved page into the buddy system, so it gets managed. */
> -void free_reserved_page(struct page *page);
> +void free_reserved_pages(struct page *page, unsigned int order);
> +
> +static inline void free_reserved_page(struct page *page)
> +{
> +	free_reserved_pages(page, 0);
> +}
>  
>  static inline void mark_page_reserved(struct page *page)
>  {
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index a63733dac659e..9311d1f3793ad 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -6381,15 +6381,34 @@ void adjust_managed_page_count(struct page *page, long count)
>  }
>  EXPORT_SYMBOL(adjust_managed_page_count);
>  
> -void free_reserved_page(struct page *page)
> +/**
> + * free_reserved_pages - free reserved pages
> + * @page: First page to free.
> + * @order: The page order to free.
> + *
> + * Free pages allocated through memblock during boot, letting the buddy
> + * manage them from now on.
> + *
> + * @page must be naturally aligned to the order and the order must not
> + * exceed MAX_PAGE_ORDER. All pages must be reserved.
> + */
> +void free_reserved_pages(struct page *page, unsigned int order)
>  {
> -	clear_page_tag_ref(page);
> -	ClearPageReserved(page);
> -	init_page_count(page);
> -	__free_page(page);
> -	adjust_managed_page_count(page, 1);
> +	const unsigned long nr_pages = 1UL << order;
> +	int i;
> +
> +	VM_WARN_ON_ONCE(!IS_ALIGNED(page_to_pfn(page), nr_pages));
> +	VM_WARN_ON_ONCE(order > MAX_PAGE_ORDER);
> +
> +	for (i = 0; i < nr_pages; i++) {
> +		clear_page_tag_ref(page + i);
> +		ClearPageReserved(page + i);
> +		set_page_count(page + i, 0);

Sashiko raises in reply to #4 that PFN walkers trying to grab pages could see
their reference getting overridden.

While that is not something we should be worrying about (or fixing in old code),
I think it's best to ClearPageReserved() after set_page_count(), so someone
actually trying to grab a reference would have the chance to rely on
PageReserved from being a reliable indication.

-- 
Cheers,

David