Re: [PATCH RFC 3/6] mm/zsmalloc: use a proper page type

[David Hildenbrand]

On 22.05.24 23:03, David Hildenbrand wrote:
> Let's clean it up: use a proper page type and store our data (offset
> into a page) in the lower 16 bit as documented.
>
> We'll have to restrict ourselves to <= 64KB base page size (so the offset
> fits into 16 bit), which sounds reasonable. Unfortunately, we don't have
> any space to store it elsewhere for now.
>
> Based on this, we should do a proper "struct zsdesc" conversion, as
> proposed in [1].
>
> This removes the last _mapcount/page_type offender.
>
> [1] https://lore.kernel.org/all/20231130101242.2590384-1-42.hyeyoo@xxxxxxxxx/
>
> Cc: Hyeonggon Yoo <42.hyeyoo@xxxxxxxxx>
> Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
> ---
>   include/linux/page-flags.h |  3 +++
>   mm/Kconfig                 |  1 +
>   mm/zsmalloc.c              | 23 +++++++++++++++++++----
>   3 files changed, 23 insertions(+), 4 deletions(-)
>
> diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
> index ed9ac4b5233d..ccaf16656de9 100644
> --- a/include/linux/page-flags.h
> +++ b/include/linux/page-flags.h
> @@ -959,6 +959,7 @@ PAGEFLAG_FALSE(HasHWPoisoned, has_hwpoisoned)
>   #define PG_guard	0x00080000
>   #define PG_hugetlb	0x00100800
>   #define PG_slab		0x00200000
> +#define PG_zsmalloc	0x00400000
>   
>   #define PageType(page, flag)						\
>   	((page->page_type & (PAGE_TYPE_BASE | flag)) == PAGE_TYPE_BASE)
> @@ -1073,6 +1074,8 @@ FOLIO_TYPE_OPS(hugetlb, hugetlb)
>   FOLIO_TEST_FLAG_FALSE(hugetlb)
>   #endif
>   
> +PAGE_TYPE_OPS(Zsmalloc, zsmalloc, zsmalloc)
> +
>   /**
>    * PageHuge - Determine if the page belongs to hugetlbfs
>    * @page: The page to test.
> diff --git a/mm/Kconfig b/mm/Kconfig
> index b4cb45255a54..0371d79b1b75 100644
> --- a/mm/Kconfig
> +++ b/mm/Kconfig
> @@ -190,6 +190,7 @@ config ZSMALLOC
>   	tristate
>   	prompt "N:1 compression allocator (zsmalloc)" if ZSWAP
>   	depends on MMU
> +	depends on PAGE_SIZE_LESS_THAN_256KB # we want <= 64KB
>   	help
>   	  zsmalloc is a slab-based memory allocator designed to store
>   	  pages of various compression levels efficiently. It achieves
> diff --git a/mm/zsmalloc.c b/mm/zsmalloc.c
> index b42d3545ca85..6f0032e06242 100644
> --- a/mm/zsmalloc.c
> +++ b/mm/zsmalloc.c
> @@ -20,7 +20,8 @@
>    *	page->index: links together all component pages of a zspage
>    *		For the huge page, this is always 0, so we use this field
>    *		to store handle.
> - *	page->page_type: first object offset in a subpage of zspage
> + *	page->page_type: PG_zsmalloc, lower 16 bit locate the first object
> + *		offset in a subpage of a zspage
>    *
>    * Usage of struct page flags:
>    *	PG_private: identifies the first component page
> @@ -450,14 +451,22 @@ static inline struct page *get_first_page(struct zspage *zspage)
>   	return first_page;
>   }
>   
> +static inline void reset_first_obj_offset(struct page *page)
> +{
> +	page->page_type |= 0xffff;
> +}
> +
>   static inline unsigned int get_first_obj_offset(struct page *page)
>   {
> -	return page->page_type;
> +	return page->page_type & 0xffff;
>   }
>   
>   static inline void set_first_obj_offset(struct page *page, unsigned int offset)
>   {
> -	page->page_type = offset;
> +	BUILD_BUG_ON(PAGE_SIZE & ~0xffff);

Buildbot is correctly complaining with PAGE_SIZE=64KiB.

We must check BUILD_BUG_ON((PAGE_SIZE -1) & ~0xffff);

--
Cheers,

David / dhildenb