Re: [PATCH] mm: delete duplicate order checking, when stealing whole pageblock

Re: [PATCH] mm: delete duplicate order checking, when stealing whole pageblock

[Andrew Morton]

On Fri, 11 Jun 2021 14:38:34 +0800 chengkaitao <pilgrimtao@gmail.com> wrote:

> From: chengkaitao <pilgrimtao@gmail.com>
> 
> 1. Already has (order >= pageblock_order / 2) here, we don't neet
> (order >= pageblock_order)
> 2. set function can_steal_fallback to inline
> 
> ...
>
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2619,18 +2619,8 @@ static void change_pageblock_range(struct page *pageblock_page,
>   * is worse than movable allocations stealing from unmovable and reclaimable
>   * pageblocks.
>   */
> -static bool can_steal_fallback(unsigned int order, int start_mt)
> +static inline bool can_steal_fallback(unsigned int order, int start_mt)
>  {
> -	/*
> -	 * Leaving this order check is intended, although there is
> -	 * relaxed order check in next check. The reason is that
> -	 * we can actually steal whole pageblock if this condition met,
> -	 * but, below check doesn't guarantee it and that is just heuristic
> -	 * so could be changed anytime.
> -	 */
> -	if (order >= pageblock_order)
> -		return true;
> -
>  	if (order >= pageblock_order / 2 ||
>  		start_mt == MIGRATE_RECLAIMABLE ||
>  		start_mt == MIGRATE_UNMOVABLE ||

Well, that redundant check was put there deliberately, as the comment
explains.

The reasoning is perhaps a little dubious, but it seems that the
compiler has optimized away the redundant check anyway (your patch
doesn't alter code size).