Re: [PATCH] mm/hugetlb: fix surplus pages in dissolve_free_huge_page()

[David Hildenbrand <david@redhat.com>]

> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -2135,6 +2135,8 @@ int dissolve_free_hugetlb_folio(struct folio *folio)
>   
>   	if (!folio_ref_count(folio)) {
>   		struct hstate *h = folio_hstate(folio);
> +		bool adjust_surplus = false;
> +
>   		if (!available_huge_pages(h))
>   			goto out;
>   
> @@ -2157,7 +2159,9 @@ int dissolve_free_hugetlb_folio(struct folio *folio)
>   			goto retry;
>   		}
>   
> -		remove_hugetlb_folio(h, folio, false);
> +		if (h->surplus_huge_pages_node[folio_nid(folio)])
> +			adjust_surplus = true;

This change looks good to me

> +		remove_hugetlb_folio(h, folio, adjust_surplus);
>   		h->max_huge_pages--;
>   		spin_unlock_irq(&hugetlb_lock);
>   
> @@ -2177,7 +2181,7 @@ int dissolve_free_hugetlb_folio(struct folio *folio)
>   			rc = hugetlb_vmemmap_restore_folio(h, folio);
>   			if (rc) {
>   				spin_lock_irq(&hugetlb_lock);
> -				add_hugetlb_folio(h, folio, false);
> +				add_hugetlb_folio(h, folio, adjust_surplus);

I'm not quite sure here, though. We dropped the hugetlb_lock, can't some 
weird concurrent action result in us not having to adjust surplus page 
anymore?

-- 
Cheers,

David / dhildenb