Re: [patch 2/2] mm: vmscan: drop nr_force_scan[] from get_scan_count

[Minchan Kim]

On Fri, Aug 12, 2011 at 5:31 AM, Johannes Weiner <jweiner@xxxxxxxxxx> wrote:
> The nr_force_scan[] tuple holds the effective scan numbers for anon
> and file pages in case the situation called for a forced scan and the
> regularly calculated scan numbers turned out zero.
>
> However, the effective scan number can always be assumed to be
> SWAP_CLUSTER_MAX right before the division into anon and file. ÂThe
> numerators and denominator are properly set up for all cases, be it
> force scan for just file, just anon, or both, to do the right thing.
>
> Signed-off-by: Johannes Weiner <jweiner@xxxxxxxxxx>

Reviewed-by: Minchan Kim <minchan.kim@xxxxxxxxx>

There is a nitpick at below.

> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> Cc: Michal Hocko <mhocko@xxxxxxx>
> Cc: Ying Han <yinghan@xxxxxxxxxx>
> Cc: Balbir Singh <bsingharora@xxxxxxxxx>
> Cc: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
> Cc: Daisuke Nishimura <nishimura@xxxxxxxxxxxxxxxxx>
> Cc: Mel Gorman <mel@xxxxxxxxx>
> ---
> Âmm/vmscan.c | Â 24 ++----------------------
> Â1 files changed, 2 insertions(+), 22 deletions(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 96061d7..45f0986 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1831,7 +1831,6 @@ static void get_scan_count(struct zone *zone, struct scan_control *sc,
> Â Â Â Âenum lru_list l;
> Â Â Â Âint noswap = 0;
> Â Â Â Âbool force_scan = false;
> - Â Â Â unsigned long nr_force_scan[2];
>
> Â Â Â Â/* kswapd does zone balancing and need to scan this zone */
> Â Â Â Âif (scanning_global_lru(sc) && current_is_kswapd())
> @@ -1846,8 +1845,6 @@ static void get_scan_count(struct zone *zone, struct scan_control *sc,
> Â Â Â Â Â Â Â Âfraction[0] = 0;
> Â Â Â Â Â Â Â Âfraction[1] = 1;
> Â Â Â Â Â Â Â Âdenominator = 1;
> - Â Â Â Â Â Â Â nr_force_scan[0] = 0;
> - Â Â Â Â Â Â Â nr_force_scan[1] = SWAP_CLUSTER_MAX;
> Â Â Â Â Â Â Â Âgoto out;
> Â Â Â Â}
>
> @@ -1864,8 +1861,6 @@ static void get_scan_count(struct zone *zone, struct scan_control *sc,
> Â Â Â Â Â Â Â Â Â Â Â Âfraction[0] = 1;
> Â Â Â Â Â Â Â Â Â Â Â Âfraction[1] = 0;
> Â Â Â Â Â Â Â Â Â Â Â Âdenominator = 1;
> - Â Â Â Â Â Â Â Â Â Â Â nr_force_scan[0] = SWAP_CLUSTER_MAX;
> - Â Â Â Â Â Â Â Â Â Â Â nr_force_scan[1] = 0;
> Â Â Â Â Â Â Â Â Â Â Â Âgoto out;
> Â Â Â Â Â Â Â Â}
> Â Â Â Â}
> @@ -1914,11 +1909,6 @@ static void get_scan_count(struct zone *zone, struct scan_control *sc,
> Â Â Â Âfraction[0] = ap;
> Â Â Â Âfraction[1] = fp;
> Â Â Â Âdenominator = ap + fp + 1;
> - Â Â Â if (force_scan) {
> - Â Â Â Â Â Â Â unsigned long scan = SWAP_CLUSTER_MAX;
> - Â Â Â Â Â Â Â nr_force_scan[0] = div64_u64(scan * ap, denominator);
> - Â Â Â Â Â Â Â nr_force_scan[1] = div64_u64(scan * fp, denominator);
> - Â Â Â }
> Âout:
> Â Â Â Âfor_each_evictable_lru(l) {
> Â Â Â Â Â Â Â Âint file = is_file_lru(l);
> @@ -1927,20 +1917,10 @@ out:
> Â Â Â Â Â Â Â Âscan = zone_nr_lru_pages(zone, sc, l);
> Â Â Â Â Â Â Â Âif (priority || noswap) {
> Â Â Â Â Â Â Â Â Â Â Â Âscan >>= priority;
> + Â Â Â Â Â Â Â Â Â Â Â if (!scan && force_scan)
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â scan = SWAP_CLUSTER_MAX;
> Â Â Â Â Â Â Â Â Â Â Â Âscan = div64_u64(scan * fraction[file], denominator);
> Â Â Â Â Â Â Â Â}
> -
> - Â Â Â Â Â Â Â /*
> - Â Â Â Â Â Â Â Â* If zone is small or memcg is small, nr[l] can be 0.
> - Â Â Â Â Â Â Â Â* This results no-scan on this priority and priority drop down.
> - Â Â Â Â Â Â Â Â* For global direct reclaim, it can visit next zone and tend
> - Â Â Â Â Â Â Â Â* not to have problems. For global kswapd, it's for zone
> - Â Â Â Â Â Â Â Â* balancing and it need to scan a small amounts. When using
> - Â Â Â Â Â Â Â Â* memcg, priority drop can cause big latency. So, it's better
> - Â Â Â Â Â Â Â Â* to scan small amount. See may_noscan above.
> - Â Â Â Â Â Â Â Â*/

Please move this comment with tidy-up at where making force_scan true.
Of course, we can find it by git log[246e87a9393] but as I looked the
git log, it explain this comment indirectly and it's not
understandable to newbies. I think this comment is more understandable
than changelog in git.



-- 
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/