Re: [RFC PATCH v1 13/13] mm: splice local lists onto the front of the LRU

[Daniel Jordan]

On 02/02/2018 10:22 AM, Laurent Dufour wrote:
> On 01/02/2018 00:04, daniel.m.jordan@xxxxxxxxxx wrote:
...snip...
> > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > index 99a54df760e3..6911626f29b2 100644
> > --- a/mm/memcontrol.c
> > +++ b/mm/memcontrol.c
> > @@ -2077,6 +2077,7 @@ static void lock_page_lru(struct page *page, int *isolated)
> >
> >   		lruvec = mem_cgroup_page_lruvec(page, zone->zone_pgdat);
> >   		ClearPageLRU(page);
> > +		smp_rmb(); /* Pairs with smp_wmb in __pagevec_lru_add */
>
> Why not include the call to smp_rmb() in del_page_from_lru_list() instead
> of spreading smp_rmb() before calls to del_page_from_lru_list() ?

Yes, this is what I should have done.  The memory barriers came from 
another patch I squashed in and I didn't look back to see how obvious 
the encapsulation was.