Re: [PATCH 3/6] swap: Unify normal/huge code path in swap_page_trans_huge_swapped()

[Daniel Jordan <daniel.m.jordan@oracle.com>]

On Fri, Jul 13, 2018 at 07:36:33AM +0800, Huang, Ying wrote:
> diff --git a/mm/swapfile.c b/mm/swapfile.c
> index 75c84aa763a3..160f78072667 100644
> --- a/mm/swapfile.c
> +++ b/mm/swapfile.c
> @@ -270,7 +270,10 @@ static inline void cluster_set_null(struct swap_cluster_info *info)
>  
>  static inline bool cluster_is_huge(struct swap_cluster_info *info)
>  {
> -	return info->flags & CLUSTER_FLAG_HUGE;
> +	if (IS_ENABLED(CONFIG_THP_SWAP))
> +		return info->flags & CLUSTER_FLAG_HUGE;
> +	else
> +		return false;
>  }
>  
>  static inline void cluster_clear_huge(struct swap_cluster_info *info)
> @@ -1489,9 +1492,6 @@ static bool swap_page_trans_huge_swapped(struct swap_info_struct *si,
>  	int i;
>  	bool ret = false;
>  
> -	if (!IS_ENABLED(CONFIG_THP_SWAP))
> -		return swap_swapcount(si, entry) != 0;

This tests the value returned from swap_count,

> -
>  	ci = lock_cluster_or_swap_info(si, offset);
>  	if (!ci || !cluster_is_huge(ci)) {
>  		if (map[roffset] != SWAP_HAS_CACHE)

and now we're testing

                    map[roffset] != SWAP_HAS_CACHE

instead.  The two seem to mean the same thing here, since the swap slot hasn't
been freed to the global pool and so can't be 0, but it might be better for
consistency and clarity to use swap_count here, and a few lines down too

        for (i = 0; i < SWAPFILE_CLUSTER; i++) {                                     
                if (map[offset + i] != SWAP_HAS_CACHE) {                             

since swap_count seems to be used everywhere else for this.