Re: [PATCH -next 2/8] mm: page_idle: Convert page idle to use folios

[SeongJae Park]

On Tue, 13 Dec 2022 17:27:29 +0800 Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> wrote:

> This change and rename all page related function to use folio,
> which save many compound_head() calls and kernel text.
> 
> Signed-off-by: Kefeng Wang <wangkefeng.wang@xxxxxxxxxx>
> ---
>  mm/page_idle.c | 91 ++++++++++++++++++++++++--------------------------
>  1 file changed, 44 insertions(+), 47 deletions(-)
> 
> diff --git a/mm/page_idle.c b/mm/page_idle.c
> index bc08332a609c..38641f94cf49 100644
> --- a/mm/page_idle.c
> +++ b/mm/page_idle.c
> @@ -19,34 +19,33 @@
>  #define BITMAP_CHUNK_BITS	(BITMAP_CHUNK_SIZE * BITS_PER_BYTE)
>  
>  /*
> - * Idle page tracking only considers user memory pages, for other types of
> - * pages the idle flag is always unset and an attempt to set it is silently
> + * Idle folio tracking only considers user memory folios, for other types of
> + * folios the idle flag is always unset and an attempt to set it is silently
>   * ignored.

I think this might make future readers confused, as the code and comments will
call this subsystem as Idle _folio_ tracking while the user interface and the
document calls it as Idle _page_ tracking.

>   *
> - * We treat a page as a user memory page if it is on an LRU list, because it is
> - * always safe to pass such a page to rmap_walk(), which is essential for idle
> - * page tracking. With such an indicator of user pages we can skip isolated
> - * pages, but since there are not usually many of them, it will hardly affect
> + * We treat a folio as a user memory folio if it is on an LRU list, because it is
> + * always safe to pass such a folio to rmap_walk(), which is essential for idle
> + * folio tracking. With such an indicator of user folios we can skip isolated
> + * folios, but since there are not usually many of them, it will hardly affect
>   * the overall result.
>   *
> - * This function tries to get a user memory page by pfn as described above.
> + * This function tries to get a user memory folio by pfn as described above.
>   */
> -static struct page *page_idle_get_page(unsigned long pfn)
> +static struct folio *folio_idle_get_folio(unsigned long pfn)
>  {
> -	struct page *page = pfn_to_online_page(pfn);
> +	struct folio *folio = pfn_to_online_folio(pfn);
>  
> -	if (!page || !PageLRU(page) ||
> -	    !get_page_unless_zero(page))
> +	if (!folio || !folio_test_lru(folio) || !folio_try_get(folio))
>  		return NULL;

As Matthew also mentioned, you changed the behavior.  It looks like a right
behavior if we call this function folio_idle_get_folio(), but it shouldn't.
page_idle_get_folio() would be a better name in my humble opinion.

So, unless we're gonna change the behavior of this subsystem (which wouldn't be
the case as it would break users), I think it would be better to keep the name
of this subsystem as Idle page tracking and avoid unnecessary renaming changes.


Thanks,
SJ