mm: do not iterate over NR_CPUS in __zone_pcp_update()

__zone_pcp_update() iterates over NR_CPUS instead of limiting the access
to the possible cpus.  This might result in access to uninitialized areas
as the per cpu allocator only populates the per cpu memory for possible
cpus.

This problem was created as a result of the dynamic allocation of pagesets
from percpu memory that went in during the merge window - commit
99dcc3e ("this_cpu: Page allocator
conversion").

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Christoph Lameter <cl@linux-foundation.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page_alloc.c

@@ -3224,7 +3224,7 @@ static int __zone_pcp_update(void *data)
 	int cpu;
 	unsigned long batch = zone_batchsize(zone), flags;
 
-	for (cpu = 0; cpu < NR_CPUS; cpu++) {
+	for_each_possible_cpu(cpu) {
 		struct per_cpu_pageset *pset;
 		struct per_cpu_pages *pcp;