padata: use __this_cpu_read per-cpu helper

For bottom halves off, __this_cpu_read is better.
Signed-off-by: NShan Wei <davidshan@tencent.com>
Reviewed-by: NChristoph Lameter <cl@linux.com>
Acked-by: NSteffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

kernel/padata.c

@@ -171,7 +171,7 @@ static struct padata_priv *padata_get_next(struct parallel_data *pd)
 {
 	int cpu, num_cpus;
 	unsigned int next_nr, next_index;
-	struct padata_parallel_queue *queue, *next_queue;
+	struct padata_parallel_queue *next_queue;
 	struct padata_priv *padata;
 	struct padata_list *reorder;
 
@@ -204,8 +204,7 @@ static struct padata_priv *padata_get_next(struct parallel_data *pd)
 		goto out;
 	}
 
-	queue = per_cpu_ptr(pd->pqueue, smp_processor_id());
-	if (queue->cpu_index == next_queue->cpu_index) {
+	if (__this_cpu_read(pd->pqueue->cpu_index) == next_queue->cpu_index) {
 		padata = ERR_PTR(-ENODATA);
 		goto out;
 	}