On Mon, 7 Dec 2020, Roman Gushchin wrote:
> On Mon, Dec 07, 2020 at 04:02:54PM +0100, Michal Hocko wrote:
> >
> > As I've said the THP accounting change makes more sense to me because it
> > allows future changes which are already undergoing so there is more
> > merit in those.
>
> +1
> And this part is absolutely trivial.
It does need to be recognized that, with these changes, every THP stats
update overflows the per-cpu counter, resorting to atomic global updates.
And I'd like to see that mentioned in the commit message.
But this change is consistent with 4.7's 8f182270dfec ("mm/swap.c: flush
lru pvecs on compound page arrival"): we accepted greater overhead for
greater accuracy back then, so I think it's okay to do so for THP stats.
Hugh
