Re: [PATCH mm-unstable 6/8] mm/hugetlb: convert alloc_migrate_huge_page to folios

[Sidhartha Kumar]

On 1/6/23 6:54 PM, Mike Kravetz wrote:
> On 01/03/23 13:13, Sidhartha Kumar wrote:
> > Change alloc_huge_page_nodemask() to alloc_hugetlb_folio_nodemask() and
> > alloc_migrate_huge_page() to alloc_migrate_hugetlb_folio(). Both functions
> > now return a folio rather than a page.
>
> >   /* mempolicy aware migration callback */
> > @@ -2357,16 +2357,16 @@ struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
> >   {
> >   	struct mempolicy *mpol;
> >   	nodemask_t *nodemask;
> > -	struct page *page;
> > +	struct folio *folio;
> >   	gfp_t gfp_mask;
> >   	int node;
> >   
> >   	gfp_mask = htlb_alloc_mask(h);
> >   	node = huge_node(vma, address, gfp_mask, &mpol, &nodemask);
> > -	page = alloc_huge_page_nodemask(h, node, nodemask, gfp_mask);
> > +	folio = alloc_hugetlb_folio_nodemask(h, node, nodemask, gfp_mask);
> >   	mpol_cond_put(mpol);
> >   
> > -	return page;
> > +	return &folio->page;
>
> Is it possible that folio could be NULL here and cause addressing exception?
>
> > diff --git a/mm/migrate.c b/mm/migrate.c
> > index 6932b3d5a9dd..fab706b78be1 100644
> > --- a/mm/migrate.c
> > +++ b/mm/migrate.c
> > @@ -1622,6 +1622,7 @@ struct page *alloc_migration_target(struct page *page, unsigned long private)
> >   	struct migration_target_control *mtc;
> >   	gfp_t gfp_mask;
> >   	unsigned int order = 0;
> > +	struct folio *hugetlb_folio = NULL;
> >   	struct folio *new_folio = NULL;
> >   	int nid;
> >   	int zidx;
> > @@ -1636,7 +1637,9 @@ struct page *alloc_migration_target(struct page *page, unsigned long private)
> >   		struct hstate *h = folio_hstate(folio);
> >   
> >   		gfp_mask = htlb_modify_alloc_mask(h, gfp_mask);
> > -		return alloc_huge_page_nodemask(h, nid, mtc->nmask, gfp_mask);
> > +		hugetlb_folio = alloc_hugetlb_folio_nodemask(h, nid,
> > +						mtc->nmask, gfp_mask);
> > +		return &hugetlb_folio->page;
>
> and, here as well?

Hi Mike,

It is possible that the folio could be null but I believe these 
instances would not cause an addressing exception because as described 
in [1], &folio->page is safe even if the folio is NULL as the page 
offset is at 0.


[1] https://lore.kernel.org/lkml/Y7h4jsv6jl0XSIsk@xxxxxxxxxxxxxxxxxxxx/T/