Messages in this thread | | | Date | Sat, 15 Feb 2014 21:25:39 -0800 (PST) | From | Hugh Dickins <> | Subject | Re: [RFC PATCH] mm/vmscan: remove two un-needed mem_cgroup_page_lruvec() call |
| |
On Sun, 16 Feb 2014, Weijie Yang wrote: > On Sun, Feb 16, 2014 at 12:00 PM, Hugh Dickins <hughd@google.com> wrote: > > On Sun, 16 Feb 2014, Weijie Yang wrote: > > > >> In putback_inactive_pages() and move_active_pages_to_lru(), > >> lruvec is already an input parameter and pages are all from this lruvec, > >> therefore there is no need to call mem_cgroup_page_lruvec() in loop. > >> > >> Signed-off-by: Weijie Yang <weijie.yang@samsung.com> > > > > Looks plausible but I believe it's incorrect. The lruvec passed in > > is the one we took the pages from, but there's a small but real chance > > that the page has become uncharged meanwhile, and should now be put back > > on the root_mem_cgroup's lruvec instead of the original memcg's lruvec. > > Hi Hugh, > > Thanks for your review. > Frankly speaking, I am not very sure about it, that is why I add a RFC tag here. > So, do we need update the reclaim_stat meanwhile as we change the lruvec?
No, it's not worth bothering about, it's only for stats and this is an unlikely case; whereas wrong memcg can be a significant correctness issue.
Hugh
| |