mm, vmscan: do not count freed pages as PGDEACTIVATE

PGDEACTIVATE represents the number of pages moved from the active list
to the inactive list.  At least this sounds like the original motivation
of the counter.  move_active_pages_to_lru, however, counts pages which
got freed in the mean time as deactivated as well.  This is a very rare
event and counting them as deactivation in itself is not harmful but it
makes the code more convoluted than necessary - we have to count both
all pages and those which are freed which is a bit confusing.

After this patch the PGDEACTIVATE should have a slightly more clear
semantic and only count those pages which are moved from the active to
the inactive list which is a plus.

Link: http://lkml.kernel.org/r/20170112211221.17636-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Suggested-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This commit is contained in:
Michal Hocko 2017-02-22 15:45:55 -08:00 committed by Linus Torvalds
parent bc71226b06
commit f0958906cd

View File

@ -1878,7 +1878,6 @@ static unsigned move_active_pages_to_lru(struct lruvec *lruvec,
enum lru_list lru)
{
struct pglist_data *pgdat = lruvec_pgdat(lruvec);
unsigned long pgmoved = 0;
struct page *page;
int nr_pages;
int nr_moved = 0;
@ -1893,7 +1892,6 @@ static unsigned move_active_pages_to_lru(struct lruvec *lruvec,
nr_pages = hpage_nr_pages(page);
update_lru_size(lruvec, lru, page_zonenum(page), nr_pages);
list_move(&page->lru, &lruvec->lists[lru]);
pgmoved += nr_pages;
if (put_page_testzero(page)) {
__ClearPageLRU(page);
@ -1913,7 +1911,7 @@ static unsigned move_active_pages_to_lru(struct lruvec *lruvec,
}
if (!is_active_lru(lru))
__count_vm_events(PGDEACTIVATE, pgmoved);
__count_vm_events(PGDEACTIVATE, nr_moved);
return nr_moved;
}