Re: [PATCH] vmscan: cleanup the scan batching code

[Peter Zijlstra]

On Sat, 2009-05-02 at 10:31 +0800, Wu Fengguang wrote:
> The vmscan batching logic is twisting. Move it into a standalone
> function nr_scan_try_batch() and document it.  No behavior change.
> 
> CC: Nick Piggin <npiggin@xxxxxxx>
> CC: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
> CC: Christoph Lameter <cl@xxxxxxxxxxxxxxxxxxxx>
> Signed-off-by: Wu Fengguang <fengguang.wu@xxxxxxxxx>

Acked-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>

> ---
>  include/linux/mmzone.h |    4 ++--
>  mm/page_alloc.c        |    2 +-
>  mm/vmscan.c            |   39 ++++++++++++++++++++++++++++-----------
>  mm/vmstat.c            |    8 ++++----
>  4 files changed, 35 insertions(+), 18 deletions(-)
> 
> --- mm.orig/include/linux/mmzone.h
> +++ mm/include/linux/mmzone.h
> @@ -323,9 +323,9 @@ struct zone {
>  
>  	/* Fields commonly accessed by the page reclaim scanner */
>  	spinlock_t		lru_lock;	
> -	struct {
> +	struct zone_lru {
>  		struct list_head list;
> -		unsigned long nr_scan;
> +		unsigned long nr_saved_scan;	/* accumulated for batching */
>  	} lru[NR_LRU_LISTS];
>  
>  	struct zone_reclaim_stat reclaim_stat;
> --- mm.orig/mm/vmscan.c
> +++ mm/mm/vmscan.c
> @@ -1450,6 +1450,26 @@ static void get_scan_ratio(struct zone *
>  	percent[1] = 100 - percent[0];
>  }
>  
> +/*
> + * Smallish @nr_to_scan's are deposited in @nr_saved_scan,
> + * until we collected @swap_cluster_max pages to scan.
> + */
> +static unsigned long nr_scan_try_batch(unsigned long nr_to_scan,
> +				       unsigned long *nr_saved_scan,
> +				       unsigned long swap_cluster_max)
> +{
> +	unsigned long nr;
> +
> +	*nr_saved_scan += nr_to_scan;
> +	nr = *nr_saved_scan;
> +
> +	if (nr >= swap_cluster_max)
> +		*nr_saved_scan = 0;
> +	else
> +		nr = 0;
> +
> +	return nr;
> +}
>  
>  /*
>   * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
> @@ -1475,14 +1495,11 @@ static void shrink_zone(int priority, st
>  			scan >>= priority;
>  			scan = (scan * percent[file]) / 100;
>  		}
> -		if (scanning_global_lru(sc)) {
> -			zone->lru[l].nr_scan += scan;
> -			nr[l] = zone->lru[l].nr_scan;
> -			if (nr[l] >= swap_cluster_max)
> -				zone->lru[l].nr_scan = 0;
> -			else
> -				nr[l] = 0;
> -		} else
> +		if (scanning_global_lru(sc))
> +			nr[l] = nr_scan_try_batch(scan,
> +						  &zone->lru[l].nr_saved_scan,
> +						  swap_cluster_max);
> +		else
>  			nr[l] = scan;
>  	}
>  
> @@ -2079,11 +2096,11 @@ static void shrink_all_zones(unsigned lo
>  						l == LRU_ACTIVE_FILE))
>  				continue;
>  
> -			zone->lru[l].nr_scan += (lru_pages >> prio) + 1;
> -			if (zone->lru[l].nr_scan >= nr_pages || pass > 3) {
> +			zone->lru[l].nr_saved_scan += (lru_pages >> prio) + 1;
> +			if (zone->lru[l].nr_saved_scan >= nr_pages || pass > 3) {
>  				unsigned long nr_to_scan;
>  
> -				zone->lru[l].nr_scan = 0;
> +				zone->lru[l].nr_saved_scan = 0;
>  				nr_to_scan = min(nr_pages, lru_pages);
>  				nr_reclaimed += shrink_list(l, nr_to_scan, zone,
>  								sc, prio);
> --- mm.orig/mm/vmstat.c
> +++ mm/mm/vmstat.c
> @@ -729,10 +729,10 @@ static void zoneinfo_show_print(struct s
>  		   zone->pages_low,
>  		   zone->pages_high,
>  		   zone->pages_scanned,
> -		   zone->lru[LRU_ACTIVE_ANON].nr_scan,
> -		   zone->lru[LRU_INACTIVE_ANON].nr_scan,
> -		   zone->lru[LRU_ACTIVE_FILE].nr_scan,
> -		   zone->lru[LRU_INACTIVE_FILE].nr_scan,
> +		   zone->lru[LRU_ACTIVE_ANON].nr_saved_scan,
> +		   zone->lru[LRU_INACTIVE_ANON].nr_saved_scan,
> +		   zone->lru[LRU_ACTIVE_FILE].nr_saved_scan,
> +		   zone->lru[LRU_INACTIVE_FILE].nr_saved_scan,
>  		   zone->spanned_pages,
>  		   zone->present_pages);
>  
> --- mm.orig/mm/page_alloc.c
> +++ mm/mm/page_alloc.c
> @@ -3544,7 +3544,7 @@ static void __paginginit free_area_init_
>  		zone_pcp_init(zone);
>  		for_each_lru(l) {
>  			INIT_LIST_HEAD(&zone->lru[l].list);
> -			zone->lru[l].nr_scan = 0;
> +			zone->lru[l].nr_saved_scan = 0;
>  		}
>  		zone->reclaim_stat.recent_rotated[0] = 0;
>  		zone->reclaim_stat.recent_rotated[1] = 0;
--
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/