Re: Major mke2fs slowdown (reproducable, bisected)

[Andi Kleen <andi@firstfloor.org>]

Alexey Dobriyan <adobriyan@gmail.com> writes:
>  
> +/* Return the page with the lowest PFN in the list */
> +static struct page *min_page(struct list_head *list)
> +{
> +	unsigned long min_pfn = -1UL;
> +	struct page *min_page = NULL, *page;;
> +
> +	list_for_each_entry(page, list, lru) {
> +		unsigned long pfn = page_to_pfn(page);
> +		if (pfn < min_pfn) {
> +			min_pfn = pfn;
> +			min_page = page;
> +		}
> +	}
> +
> +	return min_page;
> +}
> +
>  /* Remove an element from the buddy allocator from the fallback list */
>  static struct page *__rmqueue_fallback(struct zone *zone, int order,
>  						int start_migratetype)
> @@ -795,8 +812,11 @@ retry:
>  			if (list_empty(&area->free_list[migratetype]))
>  				continue;
>  
> +			/* Bias kernel allocations towards low pfns */
>  			page = list_entry(area->free_list[migratetype].next,
>  					struct page, lru);
> +			if (unlikely(start_migratetype != MIGRATE_MOVABLE))
> +				page = min_page(&area->free_list[migratetype]);

Do I misread this, or does it really turn the O(1) buddy allocation into
a "search whole free list" algorithm?  Even as fallback that looks like
a quite extreme thing to do.

-Andi