Re: [PATCH mm-unstable 3/8] mm/hugetlb: convert dequeue_hugetlb_page_node functions to folios

[Mike Kravetz]

On 01/03/23 21:00, Matthew Wilcox wrote:
> On Tue, Jan 03, 2023 at 01:13:35PM -0600, Sidhartha Kumar wrote:
> > +static struct folio *dequeue_hugetlb_folio_node_exact(struct hstate *h,
> > +								int nid)
> >  {
> >  	struct page *page;
> > +	struct folio *folio;
> >  	bool pin = !!(current->flags & PF_MEMALLOC_PIN);
> >  
> >  	lockdep_assert_held(&hugetlb_lock);
> >  	list_for_each_entry(page, &h->hugepage_freelists[nid], lru) {
> > -		if (pin && !is_longterm_pinnable_page(page))
> > +		folio = page_folio(page);
> 
> I'd argue that you can pull folios directly off the hugepage_freelists.
> Since they're attached through the 'lru', you know they're not tail
> pages, because lru.prev aliases with compound_head.

Yes, then we can get rid of the local variable *page.

A quick grep shows only the routine __mem_cgroup_uncharge_list() does
this today.
-- 
Mike Kravetz