Re: [PATCH 9/9] mm: swap: replace cluster_swap_free_nr() with swap_entries_put_[map/cache]()

[Tim Chen <tim.c.chen@linux.intel.com>]

On Fri, 2025-03-14 at 05:05 +0800, Kemeng Shi wrote:
> Replace cluster_swap_free_nr() with swap_entries_put_[map/cache]() to
> remove repeat code and leverage batch-remove for entries with last flag.
> 
> Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
> ---
>  mm/swapfile.c | 21 ++-------------------
>  1 file changed, 2 insertions(+), 19 deletions(-)
> 
> diff --git a/mm/swapfile.c b/mm/swapfile.c
> index 343b34eb2a81..c27cf09d84a6 100644
> --- a/mm/swapfile.c
> +++ b/mm/swapfile.c
> @@ -1570,21 +1570,6 @@ static void swap_entries_free(struct swap_info_struct *si,
>  	__swap_entries_free(si, ci, entry, nr_pages);
>  }
>  
> -static void cluster_swap_free_nr(struct swap_info_struct *si,
> -		unsigned long offset, int nr_pages,
> -		unsigned char usage)
> -{
> -	struct swap_cluster_info *ci;
> -	unsigned long end = offset + nr_pages;
> -
> -	ci = lock_cluster(si, offset);
> -	do {
> -		swap_entry_put_locked(si, ci, swp_entry(si->type, offset),
> -				      usage);
> -	} while (++offset < end);
> -	unlock_cluster(ci);
> -}
> -
>  /*
>   * Caller has made sure that the swap device corresponding to entry
>   * is still around or has not been recycled.
> @@ -1601,7 +1586,7 @@ void swap_free_nr(swp_entry_t entry, int nr_pages)
>  
>  	while (nr_pages) {
>  		nr = min_t(int, nr_pages, SWAPFILE_CLUSTER - offset % SWAPFILE_CLUSTER);
> -		cluster_swap_free_nr(sis, offset, nr, 1);
> +		swap_entries_put_map(sis, swp_entry(sis->type, offset), nr);
>  		offset += nr;
>  		nr_pages -= nr;
>  	}
> @@ -3632,9 +3617,7 @@ int swapcache_prepare(swp_entry_t entry, int nr)
>  
>  void swapcache_clear(struct swap_info_struct *si, swp_entry_t entry, int nr)
>  {
> -	unsigned long offset = swp_offset(entry);
> -
> -	cluster_swap_free_nr(si, offset, nr, SWAP_HAS_CACHE);
> +	swap_entries_put_cache(si, entry, nr);


swap_entries_put_cache() assumes nr does not cross cluster boundary
as we only lock the cluster associated with the beginning entry.

Current callers to swapcache_clear() like
do_swap_page() and shmem_swapin_folio, call swapcache_clear()
only for pages within a folio so we do fall in a cluster so
we are okay.

Perhaps we should document with a comment that
caller to swapcache_clear() should use the function only for pages in
a folio so the pages don't cross clusters for future users
of swapcache_clear().

Otherwise the patch looks good.

Tim

>  }
>  
>  struct swap_info_struct *swp_swap_info(swp_entry_t entry)