Re: [PATCH -mm -v4 04/21] mm, THP, swap: Support PMD swap mapping in swapcache_free_cluster()

[Dave Hansen <dave.hansen@linux.intel.com>]

> +#ifdef CONFIG_THP_SWAP
> +static inline int cluster_swapcount(struct swap_cluster_info *ci)
> +{
> +	if (!ci || !cluster_is_huge(ci))
> +		return 0;
> +
> +	return cluster_count(ci) - SWAPFILE_CLUSTER;
> +}
> +#else
> +#define cluster_swapcount(ci)			0
> +#endif

Dumb questions, round 2:  On a CONFIG_THP_SWAP=n build, presumably,
cluster_is_huge()=0 always, so cluster_swapout() always returns 0.  Right?

So, why the #ifdef?

>  /*
>   * It's possible scan_swap_map() uses a free cluster in the middle of free
>   * cluster list. Avoiding such abuse to avoid list corruption.
> @@ -905,6 +917,7 @@ static void swap_free_cluster(struct swap_info_struct *si, unsigned long idx)
>  	struct swap_cluster_info *ci;
>  
>  	ci = lock_cluster(si, offset);
> +	memset(si->swap_map + offset, 0, SWAPFILE_CLUSTER);
>  	cluster_set_count_flag(ci, 0, 0);
>  	free_cluster(si, idx);
>  	unlock_cluster(ci);

This is another case of gloriously comment-free code, but stuff that
_was_ covered in the changelog.  I'd much rather have code comments than
changelog comments.  Could we fix that?

I'm generally finding it quite hard to review this because I keep having
to refer back to the changelog to see if what you are doing matches what
you said you were doing.

> @@ -1288,24 +1301,30 @@ static void swapcache_free_cluster(swp_entry_t entry)
>  
>  	ci = lock_cluster(si, offset);
>  	VM_BUG_ON(!cluster_is_huge(ci));
> +	VM_BUG_ON(!is_cluster_offset(offset));
> +	VM_BUG_ON(cluster_count(ci) < SWAPFILE_CLUSTER);
>  	map = si->swap_map + offset;
> -	for (i = 0; i < SWAPFILE_CLUSTER; i++) {
> -		val = map[i];
> -		VM_BUG_ON(!(val & SWAP_HAS_CACHE));
> -		if (val == SWAP_HAS_CACHE)
> -			free_entries++;
> +	if (!cluster_swapcount(ci)) {
> +		for (i = 0; i < SWAPFILE_CLUSTER; i++) {
> +			val = map[i];
> +			VM_BUG_ON(!(val & SWAP_HAS_CACHE));
> +			if (val == SWAP_HAS_CACHE)
> +				free_entries++;
> +		}
> +		if (free_entries != SWAPFILE_CLUSTER)
> +			cluster_clear_huge(ci);
>  	}

Also, I'll point out that cluster_swapcount() continues the horrific
naming of cluster_couunt(), not saying what the count is *of*.  The
return value doesn't help much:

	return cluster_count(ci) - SWAPFILE_CLUSTER;