Re: [PATCH net-next v19 06/14] mm: page_frag: reuse existing space for 'size' and 'pfmemalloc'

[Alexander Duyck <alexander.duyck@gmail.com>]

On Tue, Oct 1, 2024 at 12:59 AM Yunsheng Lin <yunshenglin0825@gmail.com> wrote:
>
> Currently there is one 'struct page_frag' for every 'struct
> sock' and 'struct task_struct', we are about to replace the
> 'struct page_frag' with 'struct page_frag_cache' for them.
> Before begin the replacing, we need to ensure the size of
> 'struct page_frag_cache' is not bigger than the size of
> 'struct page_frag', as there may be tens of thousands of
> 'struct sock' and 'struct task_struct' instances in the
> system.
>
> By or'ing the page order & pfmemalloc with lower bits of
> 'va' instead of using 'u16' or 'u32' for page size and 'u8'
> for pfmemalloc, we are able to avoid 3 or 5 bytes space waste.
> And page address & pfmemalloc & order is unchanged for the
> same page in the same 'page_frag_cache' instance, it makes
> sense to fit them together.
>
> After this patch, the size of 'struct page_frag_cache' should be
> the same as the size of 'struct page_frag'.
>
> CC: Alexander Duyck <alexander.duyck@gmail.com>
> Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
> ---
>  include/linux/mm_types_task.h   | 19 +++++----
>  include/linux/page_frag_cache.h | 26 +++++++++++-
>  mm/page_frag_cache.c            | 75 +++++++++++++++++++++++----------
>  3 files changed, 88 insertions(+), 32 deletions(-)
>
> diff --git a/include/linux/mm_types_task.h b/include/linux/mm_types_task.h
> index 0ac6daebdd5c..a82aa80c0ba4 100644
> --- a/include/linux/mm_types_task.h
> +++ b/include/linux/mm_types_task.h
> @@ -47,18 +47,21 @@ struct page_frag {
>  #define PAGE_FRAG_CACHE_MAX_SIZE       __ALIGN_MASK(32768, ~PAGE_MASK)
>  #define PAGE_FRAG_CACHE_MAX_ORDER      get_order(PAGE_FRAG_CACHE_MAX_SIZE)
>  struct page_frag_cache {
> -       void *va;
> -#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
> +       /* encoded_page consists of the virtual address, pfmemalloc bit and
> +        * order of a page.
> +        */
> +       unsigned long encoded_page;
> +
> +       /* we maintain a pagecount bias, so that we dont dirty cache line
> +        * containing page->_refcount every time we allocate a fragment.
> +        */
> +#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) && (BITS_PER_LONG <= 32)
>         __u16 offset;
> -       __u16 size;
> +       __u16 pagecnt_bias;
>  #else
>         __u32 offset;
> +       __u32 pagecnt_bias;
>  #endif
> -       /* we maintain a pagecount bias, so that we dont dirty cache line
> -        * containing page->_refcount every time we allocate a fragment.
> -        */
> -       unsigned int            pagecnt_bias;
> -       bool pfmemalloc;
>  };
>
>  /* Track pages that require TLB flushes */
> diff --git a/include/linux/page_frag_cache.h b/include/linux/page_frag_cache.h
> index 0a52f7a179c8..75aaad6eaea2 100644
> --- a/include/linux/page_frag_cache.h
> +++ b/include/linux/page_frag_cache.h
> @@ -3,18 +3,40 @@
>  #ifndef _LINUX_PAGE_FRAG_CACHE_H
>  #define _LINUX_PAGE_FRAG_CACHE_H
>
> +#include <linux/bits.h>
>  #include <linux/log2.h>
>  #include <linux/mm_types_task.h>
>  #include <linux/types.h>
>
> +#if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE)
> +/* Use a full byte here to enable assembler optimization as the shift
> + * operation is usually expecting a byte.
> + */
> +#define PAGE_FRAG_CACHE_ORDER_MASK             GENMASK(7, 0)
> +#define PAGE_FRAG_CACHE_PFMEMALLOC_SHIFT       8
> +#define PAGE_FRAG_CACHE_PFMEMALLOC_BIT         BIT(PAGE_FRAG_CACHE_PFMEMALLOC_SHIFT)
> +#else
> +/* Compiler should be able to figure out we don't read things as any value
> + * ANDed with 0 is 0.
> + */
> +#define PAGE_FRAG_CACHE_ORDER_MASK             0
> +#define PAGE_FRAG_CACHE_PFMEMALLOC_SHIFT       0
> +#define PAGE_FRAG_CACHE_PFMEMALLOC_BIT         BIT(PAGE_FRAG_CACHE_PFMEMALLOC_SHIFT)
> +#endif
> +

Minor nit on this. You probably only need to have
PAGE_FRAG_CACHE_ORDER_SHIFT defined in the ifdef. The PFMEMALLOC bit
code is the same in both so you could pull it out.

Also depending on how you defined it you could just define the
PFMEMALLOC_BIT as the ORDER_MASK + 1.