On Fri, Oct 14, 2016 at 05:03:55PM +0200, Michal Hocko wrote:
> On Fri 14-10-16 23:44:48, Minchan Kim wrote:
> > On Fri, Oct 14, 2016 at 03:53:34PM +0200, Michal Hocko wrote:
> > > On Fri 14-10-16 22:46:04, Minchan Kim wrote:
> > > [...]
> > > > > > > Why don't you simply mimic what shrink_inactive_list does? Aka
> > > > > > > count the
> > > > > > > number of isolated pages and then account them when appropriate?
> > > > > > >
> > > > > > I think i am correcting clearly wrong part. So, there is no need to
> > > > > > describe it too detailed. It's a misunderstanding, and i will add
> > > > > > more comments as you suggest.
> > > > >
> > > > > OK, so could you explain why you prefer to relyon __PageMovable rather
> > > > > than do a trivial counting during the isolation?
> > > >
> > > > I don't get it. Could you elaborate it a bit more?
> > >
> > > It is really simple. You can count the number of file and anonymous
> > > pages while they are isolated and then account them to NR_ISOLATED_*
> > > later. Basically the same thing we do during the reclaim. We absolutely
> > > do not have to rely on __PageMovable and make this code more complex
> > > than necessary.
> > I don't understand your point.
> > isolate_migratepages_block can isolate any movable pages, for instance,
> > anon, file and non-lru and they are isolated into cc->migratepges.
> > Then, acct_isolated accounts them to NR_ISOLATED_*.
> > Isn't it same with the one you suggested?
> > The problem is we should identify which pages is non-lru movable first.
> > If it's not non-lru, it means the page is either anon or file so we
> > can account them.
> > That's exactly waht Ming Ling did.
> > Sorry if I didn't get your point. Maybe, it would be better to give
> > pseudo code out of your mind for better understanding rather than
> > several ping-ping with vague words.
> diff --git a/mm/compaction.c b/mm/compaction.c
> index 0409a4ad6ea1..6584705a46f6 100644
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -685,7 +685,8 @@ static bool too_many_isolated(struct zone *zone)
> static unsigned long
> isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
> - unsigned long end_pfn, isolate_mode_t isolate_mode)
> + unsigned long end_pfn, isolate_mode_t isolate_mode,
> + unsigned long *isolated_file, unsigned long
> struct zone *zone = cc->zone;
> unsigned long nr_scanned = 0, nr_isolated = 0;
> @@ -866,6 +867,10 @@ isolate_migratepages_block(struct compact_control *cc,
> unsigned long low_pfn,
> /* Successfully isolated */
> del_page_from_lru_list(page, lruvec, page_lru(page));
> + if (page_is_file_cache(page))
> + (*isolated_file)++;
> + else
> + (*isolated_anon)++;
> list_add(&page->lru, &cc->migratepages);
> Makes more sense?
It is doable for isolation part. IOW, maybe we can make acct_isolated
simple with those counters but we need to handle migrate, putback part.
If you want to remove the check of __PageMoable with those counter, it
means we should pass the counter on every functions related migration
where isolate, migrate, putback parts.