Re: 2.6.16-rc1-mm2

["Serge E. Hallyn" <serue@us.ibm.com>]

Quoting Christoph Lameter (clameter@engr.sgi.com):
> On Mon, 23 Jan 2006, Serge E. Hallyn wrote:
> 
> > I don't understand why this wouldn't die on every architecture,
> > since node_to_cpumask is an inline function.
> 
> Its an array lookup on ia64.

Oh I see, sorry, I was looking at only partial lxr ouput.

Is the following patch an ok fix?

thanks
-serge

On alpha, powerpc, and i386, node_to_cpumask is an inline function
rather than a #define to an array lookup.

--
Signed-off-by: Serge Hallyn <serue@us.ibm.com>

Index: linux-2.6.15/mm/vmscan.c
===================================================================
--- linux-2.6.15.orig/mm/vmscan.c	2006-01-23 07:14:48.000000000 -0600
+++ linux-2.6.15/mm/vmscan.c	2006-01-23 07:26:51.000000000 -0600
@@ -1836,13 +1836,15 @@ int zone_reclaim(struct zone *zone, gfp_
 	struct task_struct *p = current;
 	struct reclaim_state reclaim_state;
 	struct scan_control sc;
+	cpumask_t mask;
 
 	if (time_before(jiffies,
 		zone->last_unsuccessful_zone_reclaim + ZONE_RECLAIM_INTERVAL))
 			return 0;
 
+	mask = node_to_cpumask(zone->zone_pgdat->node_id);
 	if (!(gfp_mask & __GFP_WAIT) ||
-		(!cpus_empty(node_to_cpumask(zone->zone_pgdat->node_id)) &&
+		(!cpus_empty(mask) &&
 			 zone->zone_pgdat->node_id != numa_node_id()) ||
 		zone->all_unreclaimable ||
 		atomic_read(&zone->reclaim_in_progress) > 0)