Re: [PATCH 8/8] Per-container pages reclamation

[Balbir Singh]

Pavel Emelianov wrote:
> Implement try_to_free_pages_in_container() to free the
> pages in container that has run out of memory.
>
> The scan_control->isolate_pages() function isolates the
> container pages only.
>
> +#ifdef CONFIG_RSS_CONTAINER
> +unsigned long try_to_free_pages_in_container(struct rss_container *cnt)
> +{
> +	struct scan_control sc = {
> +		.gfp_mask = GFP_KERNEL,
> +		.may_writepage = 1,
> +		.swap_cluster_max = 1,
> +		.may_swap = 1,
> +		.swappiness = vm_swappiness,
> +		.cnt = cnt,
> +		.isolate_pages = isolate_pages_in_container,
> +	};
> +	int node;
> +	struct zone **zones;
> +
> +	for_each_node(node) {
> +#ifdef CONFIG_HIGHMEM
> +		zones = NODE_DATA(node)->node_zonelists[ZONE_HIGHMEM].zones;
> +		if (do_try_to_free_pages(zones, sc.gfp_mask, &sc))
> +			return 1;
> +#endif
> +		zones = NODE_DATA(node)->node_zonelists[ZONE_NORMAL].zones;
> +		if (do_try_to_free_pages(zones, sc.gfp_mask, &sc))
> +			return 1;
> +	}
> +	return 0;
> +}
> +#endif
> +

Hi, Pavel,

This patch fixes a bug in the RSS controller, where we walk through all
nodes during reclaim in try_to_free_pages_in_container(). Instead of
for_each_node(), we now use for_each_online_node() so that we do not
end up with invalid zones from nodes that are not online.

Signed-off-by: Balbir Singh <balbir@xxxxxxxxxxxxxxxxxx>
Signed-off-by: Vaidyanathan Srinivasan <svaidy@xxxxxxxxxxxxxxxxxx>
---

 mm/vmscan.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff -puN mm/vmscan.c~rss-fix-nodescan mm/vmscan.c
--- linux-2.6.20/mm/vmscan.c~rss-fix-nodescan	2007-05-02 14:47:09.000000000 
+0530
+++ linux-2.6.20-balbir/mm/vmscan.c	2007-05-02 14:47:25.000000000 +0530
@@ -1127,7 +1127,7 @@ unsigned long try_to_free_pages_in_conta
 	int node;
 	struct zone **zones;

-	for_each_node(node) {
+	for_each_online_node(node) {
 #ifdef CONFIG_HIGHMEM
 		zones = NODE_DATA(node)->node_zonelists[ZONE_HIGHMEM].zones;
 		if (do_try_to_free_pages(zones, sc.gfp_mask, &sc))
_

--
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL
-
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/