Re: [PATCH v3] mm/page_owner: fix for deferred struct page init

[Qian Cai]

On Thu, 2018-12-20 at 14:00 -0700, William Kucharski wrote:
> > On Dec 20, 2018, at 1:31 PM, Qian Cai <cai@xxxxxx> wrote:
> > 
> > diff --git a/mm/page_ext.c b/mm/page_ext.c
> > index ae44f7adbe07..d76fd51e312a 100644
> > --- a/mm/page_ext.c
> > +++ b/mm/page_ext.c
> > @@ -399,9 +399,8 @@ void __init page_ext_init(void)
> > 			Â* -------------pfn-------------->
> > 			Â* N0 | N1 | N2 | N0 | N1 | N2|....
> > 			Â*
> > -			Â* Take into account DEFERRED_STRUCT_PAGE_INIT.
> > 			Â*/
> > -			if (early_pfn_to_nid(pfn) != nid)
> > +			if (pfn_to_nid(pfn) != nid)
> > 				continue;
> > 			if (init_section_page_ext(pfn, nid))
> > 				goto oom;
> > --Â
> > 2.17.2 (Apple Git-113)
> > 
> 
> Is there any danger in the fact that in the CONFIG_NUMA case in mmzone.h
> (around line 1261), pfn_to_nid() calls page_to_nid(), possibly causing the
> same issue seen in v2?
> 

No. If CONFIG_DEFERRED_STRUCT_PAGE_INIT=y, page_ext_init() is called after
page_alloc_init_late() where all the memory has already been initialized,
soÂpage_to_nid() will work then.